From 6bb9e4bff5c6fd908d907222108ef5650d77972f Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 30 Jul 2008 12:29:37 -0700 Subject: SiS DRM: fix the memory allocator if the SIS FB is built as a module Fix the SIS DRM memory allocator if the SIS FB built as a module. The SIS DRM code initialises the mm allocation hooks, but _only_ if the SIS FB is not built as a module because it depends on CONFIG_FB_SIS, and that's unset if the SIS FB is not built in. It must check CONFIG_FB_SIS_MODULE as well. Signed-off-by: David Howells Cc: Dave Airlie Signed-off-by: Andrew Morton Signed-off-by: Dave Airlie --- drivers/gpu/drm/sis/sis_mm.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/sis/sis_mm.c b/drivers/gpu/drm/sis/sis_mm.c index b3878770fce..c4debfdc1aa 100644 --- a/drivers/gpu/drm/sis/sis_mm.c +++ b/drivers/gpu/drm/sis/sis_mm.c @@ -41,7 +41,7 @@ #define AGP_TYPE 1 -#if defined(CONFIG_FB_SIS) +#if defined(CONFIG_FB_SIS) || defined(CONFIG_FB_SIS_MODULE) /* fb management via fb device */ #define SIS_MM_ALIGN_SHIFT 0 @@ -75,12 +75,12 @@ static unsigned long sis_sman_mm_offset(void *private, void *ref) return ~((unsigned long)ref); } -#else /* CONFIG_FB_SIS */ +#else /* CONFIG_FB_SIS[_MODULE] */ #define SIS_MM_ALIGN_SHIFT 4 #define SIS_MM_ALIGN_MASK ( (1 << SIS_MM_ALIGN_SHIFT) - 1) -#endif /* CONFIG_FB_SIS */ +#endif /* CONFIG_FB_SIS[_MODULE] */ static int sis_fb_init(struct drm_device *dev, void *data, struct drm_file *file_priv) { @@ -89,7 +89,7 @@ static int sis_fb_init(struct drm_device *dev, void *data, struct drm_file *file int ret; mutex_lock(&dev->struct_mutex); -#if defined(CONFIG_FB_SIS) +#if defined(CONFIG_FB_SIS) || defined(CONFIG_FB_SIS_MODULE) { struct drm_sman_mm sman_mm; sman_mm.private = (void *)0xFFFFFFFF; -- cgit v1.2.3 From 48f185d0e0f3adde81117ead074e5e6ec5548449 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 30 Jul 2008 12:29:38 -0700 Subject: SiS DRM: fix a pointer cast warning Fix a pointer cast warning in the SIS DRM code. This was introduced in patch ce65a44de07f73ceda1749812b75086b7add408d. Signed-off-by: David Howells Cc: Dave Airlie Signed-off-by: Andrew Morton Signed-off-by: Dave Airlie --- drivers/gpu/drm/sis/sis_mm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/sis/sis_mm.c b/drivers/gpu/drm/sis/sis_mm.c index c4debfdc1aa..af22111397d 100644 --- a/drivers/gpu/drm/sis/sis_mm.c +++ b/drivers/gpu/drm/sis/sis_mm.c @@ -57,7 +57,7 @@ static void *sis_sman_mm_allocate(void *private, unsigned long size, if (req.size == 0) return NULL; else - return (void *)~req.offset; + return (void *)(unsigned long)~req.offset; } static void sis_sman_mm_free(void *private, void *ref) -- cgit v1.2.3 From 0790d5e148c0747499742a3c09ba5f1c07f9ed0d Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Wed, 30 Jul 2008 12:28:47 -0700 Subject: i915: remove settable use_mi_batchbuffer_start The driver can know what hardware requires MI_BATCH_BUFFER vs MI_BATCH_BUFFER_START; there's no reason to let user mode configure this. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 11 +---------- drivers/gpu/drm/i915/i915_drv.h | 1 - 2 files changed, 1 insertion(+), 11 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 9ac4720e647..ae0ba350188 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -159,13 +159,6 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) dev_priv->current_page = 0; dev_priv->sarea_priv->pf_current_page = dev_priv->current_page; - /* We are using separate values as placeholders for mechanisms for - * private backbuffer/depthbuffer usage. - */ - dev_priv->use_mi_batchbuffer_start = 0; - if (IS_I965G(dev)) /* 965 doesn't support older method */ - dev_priv->use_mi_batchbuffer_start = 1; - /* Allow hardware batchbuffers unless told otherwise. */ dev_priv->allow_batchbuffer = 1; @@ -486,7 +479,7 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev, return ret; } - if (dev_priv->use_mi_batchbuffer_start) { + if (!IS_I830(dev) && !IS_845G(dev)) { BEGIN_LP_RING(2); if (IS_I965G(dev)) { OUT_RING(MI_BATCH_BUFFER_START | (2 << 6) | MI_BATCH_NON_SECURE_I965); @@ -697,8 +690,6 @@ static int i915_setparam(struct drm_device *dev, void *data, switch (param->param) { case I915_SETPARAM_USE_MI_BATCHBUFFER_START: - if (!IS_I965G(dev)) - dev_priv->use_mi_batchbuffer_start = param->value; break; case I915_SETPARAM_TEX_LRU_LOG_GRANULARITY: dev_priv->tex_lru_log_granularity = param->value; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d7326d92a23..2d441d37da9 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -99,7 +99,6 @@ typedef struct drm_i915_private { int front_offset; int current_page; int page_flipping; - int use_mi_batchbuffer_start; wait_queue_head_t irq_queue; atomic_t irq_received; -- cgit v1.2.3 From 962d4fd7273e144ae003ddb90138ae4b80567c70 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Wed, 30 Jul 2008 12:36:08 -0700 Subject: i915: Ignore X server provided mmio address It is already correctly detected by the kernel for use in suspend/resume. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 12 ------------ 1 file changed, 12 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index ae0ba350188..28c812a0c63 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -121,13 +121,6 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) return -EINVAL; } - dev_priv->mmio_map = drm_core_findmap(dev, init->mmio_offset); - if (!dev_priv->mmio_map) { - i915_dma_cleanup(dev); - DRM_ERROR("can not find mmio map!\n"); - return -EINVAL; - } - dev_priv->sarea_priv = (drm_i915_sarea_t *) ((u8 *) dev_priv->sarea->handle + init->sarea_priv_offset); @@ -194,11 +187,6 @@ static int i915_dma_resume(struct drm_device * dev) return -EINVAL; } - if (!dev_priv->mmio_map) { - DRM_ERROR("can not find mmio map!\n"); - return -EINVAL; - } - if (dev_priv->ring.map.handle == NULL) { DRM_ERROR("can not ioremap virtual address for" " ring buffer\n"); -- cgit v1.2.3 From 585fb111348f7cdc30c6a1b903987612ddeafb23 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Tue, 29 Jul 2008 11:54:06 -0700 Subject: i915: Use more consistent names for regs, and store them in a separate file. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 32 +- drivers/gpu/drm/i915/i915_drv.c | 54 +- drivers/gpu/drm/i915/i915_drv.h | 836 +---------------------- drivers/gpu/drm/i915/i915_irq.c | 73 +- drivers/gpu/drm/i915/i915_reg.h | 1405 +++++++++++++++++++++++++++++++++++++++ 5 files changed, 1508 insertions(+), 892 deletions(-) create mode 100644 drivers/gpu/drm/i915/i915_reg.h (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 28c812a0c63..4d56dfd25e0 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -40,11 +40,11 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) { drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_ring_buffer_t *ring = &(dev_priv->ring); - u32 last_head = I915_READ(LP_RING + RING_HEAD) & HEAD_ADDR; + u32 last_head = I915_READ(PRB0_HEAD) & HEAD_ADDR; int i; for (i = 0; i < 10000; i++) { - ring->head = I915_READ(LP_RING + RING_HEAD) & HEAD_ADDR; + ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) ring->space += ring->Size; @@ -67,8 +67,8 @@ void i915_kernel_lost_context(struct drm_device * dev) drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_ring_buffer_t *ring = &(dev_priv->ring); - ring->head = I915_READ(LP_RING + RING_HEAD) & HEAD_ADDR; - ring->tail = I915_READ(LP_RING + RING_TAIL) & TAIL_ADDR; + ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; + ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR; ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) ring->space += ring->Size; @@ -98,13 +98,13 @@ static int i915_dma_cleanup(struct drm_device * dev) drm_pci_free(dev, dev_priv->status_page_dmah); dev_priv->status_page_dmah = NULL; /* Need to rewrite hardware status page */ - I915_WRITE(0x02080, 0x1ffff000); + I915_WRITE(HWS_PGA, 0x1ffff000); } if (dev_priv->status_gfx_addr) { dev_priv->status_gfx_addr = 0; drm_core_ioremapfree(&dev_priv->hws_map, dev); - I915_WRITE(0x2080, 0x1ffff000); + I915_WRITE(HWS_PGA, 0x1ffff000); } return 0; @@ -170,7 +170,7 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr; memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - I915_WRITE(0x02080, dev_priv->dma_status_page); + I915_WRITE(HWS_PGA, dev_priv->dma_status_page); } DRM_DEBUG("Enabled hardware status page\n"); return 0; @@ -201,9 +201,9 @@ static int i915_dma_resume(struct drm_device * dev) DRM_DEBUG("hw status page @ %p\n", dev_priv->hw_status_page); if (dev_priv->status_gfx_addr != 0) - I915_WRITE(0x02080, dev_priv->status_gfx_addr); + I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); else - I915_WRITE(0x02080, dev_priv->dma_status_page); + I915_WRITE(HWS_PGA, dev_priv->dma_status_page); DRM_DEBUG("Enabled hardware status page\n"); return 0; @@ -402,8 +402,8 @@ static void i915_emit_breadcrumb(struct drm_device *dev) dev_priv->sarea_priv->last_enqueue = dev_priv->counter = 1; BEGIN_LP_RING(4); - OUT_RING(CMD_STORE_DWORD_IDX); - OUT_RING(20); + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(5 << MI_STORE_DWORD_INDEX_SHIFT); OUT_RING(dev_priv->counter); OUT_RING(0); ADVANCE_LP_RING(); @@ -505,7 +505,7 @@ static int i915_dispatch_flip(struct drm_device * dev) i915_kernel_lost_context(dev); BEGIN_LP_RING(2); - OUT_RING(INST_PARSER_CLIENT | INST_OP_FLUSH | INST_FLUSH_MAP_CACHE); + OUT_RING(MI_FLUSH | MI_READ_FLUSH); OUT_RING(0); ADVANCE_LP_RING(); @@ -530,8 +530,8 @@ static int i915_dispatch_flip(struct drm_device * dev) dev_priv->sarea_priv->last_enqueue = dev_priv->counter++; BEGIN_LP_RING(4); - OUT_RING(CMD_STORE_DWORD_IDX); - OUT_RING(20); + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(5 << MI_STORE_DWORD_INDEX_SHIFT); OUT_RING(dev_priv->counter); OUT_RING(0); ADVANCE_LP_RING(); @@ -728,8 +728,8 @@ static int i915_set_status_page(struct drm_device *dev, void *data, dev_priv->hw_status_page = dev_priv->hws_map.handle; memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - I915_WRITE(0x02080, dev_priv->status_gfx_addr); - DRM_DEBUG("load hws 0x2080 with gfx mem 0x%x\n", + I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); + DRM_DEBUG("load hws HWS_PGA with gfx mem 0x%x\n", dev_priv->status_gfx_addr); DRM_DEBUG("load hws at %p\n", dev_priv->hw_status_page); return 0; diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 93aed1c38bd..6c99aab12da 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -279,13 +279,13 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) dev_priv->saveDSPASTRIDE = I915_READ(DSPASTRIDE); dev_priv->saveDSPASIZE = I915_READ(DSPASIZE); dev_priv->saveDSPAPOS = I915_READ(DSPAPOS); - dev_priv->saveDSPABASE = I915_READ(DSPABASE); + dev_priv->saveDSPAADDR = I915_READ(DSPAADDR); if (IS_I965G(dev)) { dev_priv->saveDSPASURF = I915_READ(DSPASURF); dev_priv->saveDSPATILEOFF = I915_READ(DSPATILEOFF); } i915_save_palette(dev, PIPE_A); - dev_priv->savePIPEASTAT = I915_READ(I915REG_PIPEASTAT); + dev_priv->savePIPEASTAT = I915_READ(PIPEASTAT); /* Pipe & plane B info */ dev_priv->savePIPEBCONF = I915_READ(PIPEBCONF); @@ -307,13 +307,13 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) dev_priv->saveDSPBSTRIDE = I915_READ(DSPBSTRIDE); dev_priv->saveDSPBSIZE = I915_READ(DSPBSIZE); dev_priv->saveDSPBPOS = I915_READ(DSPBPOS); - dev_priv->saveDSPBBASE = I915_READ(DSPBBASE); + dev_priv->saveDSPBADDR = I915_READ(DSPBADDR); if (IS_I965GM(dev) || IS_IGD_GM(dev)) { dev_priv->saveDSPBSURF = I915_READ(DSPBSURF); dev_priv->saveDSPBTILEOFF = I915_READ(DSPBTILEOFF); } i915_save_palette(dev, PIPE_B); - dev_priv->savePIPEBSTAT = I915_READ(I915REG_PIPEBSTAT); + dev_priv->savePIPEBSTAT = I915_READ(PIPEBSTAT); /* CRT state */ dev_priv->saveADPA = I915_READ(ADPA); @@ -328,9 +328,9 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) dev_priv->saveLVDS = I915_READ(LVDS); if (!IS_I830(dev) && !IS_845G(dev)) dev_priv->savePFIT_CONTROL = I915_READ(PFIT_CONTROL); - dev_priv->saveLVDSPP_ON = I915_READ(LVDSPP_ON); - dev_priv->saveLVDSPP_OFF = I915_READ(LVDSPP_OFF); - dev_priv->savePP_CYCLE = I915_READ(PP_CYCLE); + dev_priv->savePP_ON_DELAYS = I915_READ(PP_ON_DELAYS); + dev_priv->savePP_OFF_DELAYS = I915_READ(PP_OFF_DELAYS); + dev_priv->savePP_DIVISOR = I915_READ(PP_DIVISOR); /* FIXME: save TV & SDVO state */ @@ -341,19 +341,19 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) dev_priv->saveFBC_CONTROL = I915_READ(FBC_CONTROL); /* Interrupt state */ - dev_priv->saveIIR = I915_READ(I915REG_INT_IDENTITY_R); - dev_priv->saveIER = I915_READ(I915REG_INT_ENABLE_R); - dev_priv->saveIMR = I915_READ(I915REG_INT_MASK_R); + dev_priv->saveIIR = I915_READ(IIR); + dev_priv->saveIER = I915_READ(IER); + dev_priv->saveIMR = I915_READ(IMR); /* VGA state */ - dev_priv->saveVCLK_DIVISOR_VGA0 = I915_READ(VCLK_DIVISOR_VGA0); - dev_priv->saveVCLK_DIVISOR_VGA1 = I915_READ(VCLK_DIVISOR_VGA1); - dev_priv->saveVCLK_POST_DIV = I915_READ(VCLK_POST_DIV); + dev_priv->saveVGA0 = I915_READ(VGA0); + dev_priv->saveVGA1 = I915_READ(VGA1); + dev_priv->saveVGA_PD = I915_READ(VGA_PD); dev_priv->saveVGACNTRL = I915_READ(VGACNTRL); /* Clock gating state */ dev_priv->saveD_STATE = I915_READ(D_STATE); - dev_priv->saveDSPCLK_GATE_D = I915_READ(DSPCLK_GATE_D); + dev_priv->saveCG_2D_DIS = I915_READ(CG_2D_DIS); /* Cache mode state */ dev_priv->saveCACHE_MODE_0 = I915_READ(CACHE_MODE_0); @@ -363,7 +363,7 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) /* Scratch space */ for (i = 0; i < 16; i++) { - dev_priv->saveSWF0[i] = I915_READ(SWF0 + (i << 2)); + dev_priv->saveSWF0[i] = I915_READ(SWF00 + (i << 2)); dev_priv->saveSWF1[i] = I915_READ(SWF10 + (i << 2)); } for (i = 0; i < 3; i++) @@ -424,7 +424,7 @@ static int i915_resume(struct drm_device *dev) I915_WRITE(DSPASIZE, dev_priv->saveDSPASIZE); I915_WRITE(DSPAPOS, dev_priv->saveDSPAPOS); I915_WRITE(PIPEASRC, dev_priv->savePIPEASRC); - I915_WRITE(DSPABASE, dev_priv->saveDSPABASE); + I915_WRITE(DSPAADDR, dev_priv->saveDSPAADDR); I915_WRITE(DSPASTRIDE, dev_priv->saveDSPASTRIDE); if (IS_I965G(dev)) { I915_WRITE(DSPASURF, dev_priv->saveDSPASURF); @@ -436,7 +436,7 @@ static int i915_resume(struct drm_device *dev) i915_restore_palette(dev, PIPE_A); /* Enable the plane */ I915_WRITE(DSPACNTR, dev_priv->saveDSPACNTR); - I915_WRITE(DSPABASE, I915_READ(DSPABASE)); + I915_WRITE(DSPAADDR, I915_READ(DSPAADDR)); /* Pipe & plane B info */ if (dev_priv->saveDPLL_B & DPLL_VCO_ENABLE) { @@ -466,7 +466,7 @@ static int i915_resume(struct drm_device *dev) I915_WRITE(DSPBSIZE, dev_priv->saveDSPBSIZE); I915_WRITE(DSPBPOS, dev_priv->saveDSPBPOS); I915_WRITE(PIPEBSRC, dev_priv->savePIPEBSRC); - I915_WRITE(DSPBBASE, dev_priv->saveDSPBBASE); + I915_WRITE(DSPBADDR, dev_priv->saveDSPBADDR); I915_WRITE(DSPBSTRIDE, dev_priv->saveDSPBSTRIDE); if (IS_I965G(dev)) { I915_WRITE(DSPBSURF, dev_priv->saveDSPBSURF); @@ -478,7 +478,7 @@ static int i915_resume(struct drm_device *dev) i915_restore_palette(dev, PIPE_B); /* Enable the plane */ I915_WRITE(DSPBCNTR, dev_priv->saveDSPBCNTR); - I915_WRITE(DSPBBASE, I915_READ(DSPBBASE)); + I915_WRITE(DSPBADDR, I915_READ(DSPBADDR)); /* CRT state */ I915_WRITE(ADPA, dev_priv->saveADPA); @@ -493,9 +493,9 @@ static int i915_resume(struct drm_device *dev) I915_WRITE(PFIT_PGM_RATIOS, dev_priv->savePFIT_PGM_RATIOS); I915_WRITE(BLC_PWM_CTL, dev_priv->saveBLC_PWM_CTL); - I915_WRITE(LVDSPP_ON, dev_priv->saveLVDSPP_ON); - I915_WRITE(LVDSPP_OFF, dev_priv->saveLVDSPP_OFF); - I915_WRITE(PP_CYCLE, dev_priv->savePP_CYCLE); + I915_WRITE(PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS); + I915_WRITE(PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS); + I915_WRITE(PP_DIVISOR, dev_priv->savePP_DIVISOR); I915_WRITE(PP_CONTROL, dev_priv->savePP_CONTROL); /* FIXME: restore TV & SDVO state */ @@ -508,14 +508,14 @@ static int i915_resume(struct drm_device *dev) /* VGA state */ I915_WRITE(VGACNTRL, dev_priv->saveVGACNTRL); - I915_WRITE(VCLK_DIVISOR_VGA0, dev_priv->saveVCLK_DIVISOR_VGA0); - I915_WRITE(VCLK_DIVISOR_VGA1, dev_priv->saveVCLK_DIVISOR_VGA1); - I915_WRITE(VCLK_POST_DIV, dev_priv->saveVCLK_POST_DIV); + I915_WRITE(VGA0, dev_priv->saveVGA0); + I915_WRITE(VGA1, dev_priv->saveVGA1); + I915_WRITE(VGA_PD, dev_priv->saveVGA_PD); udelay(150); /* Clock gating state */ I915_WRITE (D_STATE, dev_priv->saveD_STATE); - I915_WRITE (DSPCLK_GATE_D, dev_priv->saveDSPCLK_GATE_D); + I915_WRITE(CG_2D_DIS, dev_priv->saveCG_2D_DIS); /* Cache mode state */ I915_WRITE (CACHE_MODE_0, dev_priv->saveCACHE_MODE_0 | 0xffff0000); @@ -524,7 +524,7 @@ static int i915_resume(struct drm_device *dev) I915_WRITE (MI_ARB_STATE, dev_priv->saveMI_ARB_STATE | 0xffff0000); for (i = 0; i < 16; i++) { - I915_WRITE(SWF0 + (i << 2), dev_priv->saveSWF0[i]); + I915_WRITE(SWF00 + (i << 2), dev_priv->saveSWF0[i]); I915_WRITE(SWF10 + (i << 2), dev_priv->saveSWF1[i+7]); } for (i = 0; i < 3; i++) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 2d441d37da9..afb51a390e1 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -30,6 +30,8 @@ #ifndef _I915_DRV_H_ #define _I915_DRV_H_ +#include "i915_reg.h" + /* General customization: */ @@ -138,7 +140,7 @@ typedef struct drm_i915_private { u32 saveDSPASTRIDE; u32 saveDSPASIZE; u32 saveDSPAPOS; - u32 saveDSPABASE; + u32 saveDSPAADDR; u32 saveDSPASURF; u32 saveDSPATILEOFF; u32 savePFIT_PGM_RATIOS; @@ -159,24 +161,24 @@ typedef struct drm_i915_private { u32 saveDSPBSTRIDE; u32 saveDSPBSIZE; u32 saveDSPBPOS; - u32 saveDSPBBASE; + u32 saveDSPBADDR; u32 saveDSPBSURF; u32 saveDSPBTILEOFF; - u32 saveVCLK_DIVISOR_VGA0; - u32 saveVCLK_DIVISOR_VGA1; - u32 saveVCLK_POST_DIV; + u32 saveVGA0; + u32 saveVGA1; + u32 saveVGA_PD; u32 saveVGACNTRL; u32 saveADPA; u32 saveLVDS; - u32 saveLVDSPP_ON; - u32 saveLVDSPP_OFF; + u32 savePP_ON_DELAYS; + u32 savePP_OFF_DELAYS; u32 saveDVOA; u32 saveDVOB; u32 saveDVOC; u32 savePP_ON; u32 savePP_OFF; u32 savePP_CONTROL; - u32 savePP_CYCLE; + u32 savePP_DIVISOR; u32 savePFIT_CONTROL; u32 save_palette_a[256]; u32 save_palette_b[256]; @@ -189,7 +191,7 @@ typedef struct drm_i915_private { u32 saveIMR; u32 saveCACHE_MODE_0; u32 saveD_STATE; - u32 saveDSPCLK_GATE_D; + u32 saveCG_2D_DIS; u32 saveMI_ARB_STATE; u32 saveSWF0[16]; u32 saveSWF1[16]; @@ -283,816 +285,26 @@ extern void i915_mem_release(struct drm_device * dev, if (I915_VERBOSE) DRM_DEBUG("ADVANCE_LP_RING %x\n", outring); \ dev_priv->ring.tail = outring; \ dev_priv->ring.space -= outcount * 4; \ - I915_WRITE(LP_RING + RING_TAIL, outring); \ + I915_WRITE(PRB0_TAIL, outring); \ } while(0) -extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); - -/* Extended config space */ -#define LBB 0xf4 - -/* VGA stuff */ - -#define VGA_ST01_MDA 0x3ba -#define VGA_ST01_CGA 0x3da - -#define VGA_MSR_WRITE 0x3c2 -#define VGA_MSR_READ 0x3cc -#define VGA_MSR_MEM_EN (1<<1) -#define VGA_MSR_CGA_MODE (1<<0) - -#define VGA_SR_INDEX 0x3c4 -#define VGA_SR_DATA 0x3c5 - -#define VGA_AR_INDEX 0x3c0 -#define VGA_AR_VID_EN (1<<5) -#define VGA_AR_DATA_WRITE 0x3c0 -#define VGA_AR_DATA_READ 0x3c1 - -#define VGA_GR_INDEX 0x3ce -#define VGA_GR_DATA 0x3cf -/* GR05 */ -#define VGA_GR_MEM_READ_MODE_SHIFT 3 -#define VGA_GR_MEM_READ_MODE_PLANE 1 -/* GR06 */ -#define VGA_GR_MEM_MODE_MASK 0xc -#define VGA_GR_MEM_MODE_SHIFT 2 -#define VGA_GR_MEM_A0000_AFFFF 0 -#define VGA_GR_MEM_A0000_BFFFF 1 -#define VGA_GR_MEM_B0000_B7FFF 2 -#define VGA_GR_MEM_B0000_BFFFF 3 - -#define VGA_DACMASK 0x3c6 -#define VGA_DACRX 0x3c7 -#define VGA_DACWX 0x3c8 -#define VGA_DACDATA 0x3c9 - -#define VGA_CR_INDEX_MDA 0x3b4 -#define VGA_CR_DATA_MDA 0x3b5 -#define VGA_CR_INDEX_CGA 0x3d4 -#define VGA_CR_DATA_CGA 0x3d5 - -#define GFX_OP_USER_INTERRUPT ((0<<29)|(2<<23)) -#define GFX_OP_BREAKPOINT_INTERRUPT ((0<<29)|(1<<23)) -#define CMD_REPORT_HEAD (7<<23) -#define CMD_STORE_DWORD_IDX ((0x21<<23) | 0x1) -#define CMD_OP_BATCH_BUFFER ((0x0<<29)|(0x30<<23)|0x1) - -#define INST_PARSER_CLIENT 0x00000000 -#define INST_OP_FLUSH 0x02000000 -#define INST_FLUSH_MAP_CACHE 0x00000001 - -#define BB1_START_ADDR_MASK (~0x7) -#define BB1_PROTECTED (1<<0) -#define BB1_UNPROTECTED (0<<0) -#define BB2_END_ADDR_MASK (~0x7) - -/* Framebuffer compression */ -#define FBC_CFB_BASE 0x03200 /* 4k page aligned */ -#define FBC_LL_BASE 0x03204 /* 4k page aligned */ -#define FBC_CONTROL 0x03208 -#define FBC_CTL_EN (1<<31) -#define FBC_CTL_PERIODIC (1<<30) -#define FBC_CTL_INTERVAL_SHIFT (16) -#define FBC_CTL_UNCOMPRESSIBLE (1<<14) -#define FBC_CTL_STRIDE_SHIFT (5) -#define FBC_CTL_FENCENO (1<<0) -#define FBC_COMMAND 0x0320c -#define FBC_CMD_COMPRESS (1<<0) -#define FBC_STATUS 0x03210 -#define FBC_STAT_COMPRESSING (1<<31) -#define FBC_STAT_COMPRESSED (1<<30) -#define FBC_STAT_MODIFIED (1<<29) -#define FBC_STAT_CURRENT_LINE (1<<0) -#define FBC_CONTROL2 0x03214 -#define FBC_CTL_FENCE_DBL (0<<4) -#define FBC_CTL_IDLE_IMM (0<<2) -#define FBC_CTL_IDLE_FULL (1<<2) -#define FBC_CTL_IDLE_LINE (2<<2) -#define FBC_CTL_IDLE_DEBUG (3<<2) -#define FBC_CTL_CPU_FENCE (1<<1) -#define FBC_CTL_PLANEA (0<<0) -#define FBC_CTL_PLANEB (1<<0) -#define FBC_FENCE_OFF 0x0321b - -#define FBC_LL_SIZE (1536) -#define FBC_LL_PAD (32) - -/* Interrupt bits: - */ -#define USER_INT_FLAG (1<<1) -#define VSYNC_PIPEB_FLAG (1<<5) -#define VSYNC_PIPEA_FLAG (1<<7) -#define HWB_OOM_FLAG (1<<13) /* binner out of memory */ - -#define I915REG_HWSTAM 0x02098 -#define I915REG_INT_IDENTITY_R 0x020a4 -#define I915REG_INT_MASK_R 0x020a8 -#define I915REG_INT_ENABLE_R 0x020a0 - -#define I915REG_PIPEASTAT 0x70024 -#define I915REG_PIPEBSTAT 0x71024 - -#define I915_VBLANK_INTERRUPT_ENABLE (1UL<<17) -#define I915_VBLANK_CLEAR (1UL<<1) - -#define SRX_INDEX 0x3c4 -#define SRX_DATA 0x3c5 -#define SR01 1 -#define SR01_SCREEN_OFF (1<<5) - -#define PPCR 0x61204 -#define PPCR_ON (1<<0) - -#define DVOB 0x61140 -#define DVOB_ON (1<<31) -#define DVOC 0x61160 -#define DVOC_ON (1<<31) -#define LVDS 0x61180 -#define LVDS_ON (1<<31) - -#define ADPA 0x61100 -#define ADPA_DPMS_MASK (~(3<<10)) -#define ADPA_DPMS_ON (0<<10) -#define ADPA_DPMS_SUSPEND (1<<10) -#define ADPA_DPMS_STANDBY (2<<10) -#define ADPA_DPMS_OFF (3<<10) - -#define NOPID 0x2094 -#define LP_RING 0x2030 -#define HP_RING 0x2040 -/* The binner has its own ring buffer: - */ -#define HWB_RING 0x2400 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x001FFFF8 -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define RING_START 0x08 -#define START_ADDR 0x0xFFFFF000 -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - -/* Instruction parser error reg: - */ -#define IPEIR 0x2088 - -/* Scratch pad debug 0 reg: - */ -#define SCPD0 0x209c - -/* Error status reg: - */ -#define ESR 0x20b8 - -/* Secondary DMA fetch address debug reg: - */ -#define DMA_FADD_S 0x20d4 - -/* Memory Interface Arbitration State - */ -#define MI_ARB_STATE 0x20e4 - -/* Cache mode 0 reg. - * - Manipulating render cache behaviour is central - * to the concept of zone rendering, tuning this reg can help avoid - * unnecessary render cache reads and even writes (for z/stencil) - * at beginning and end of scene. - * - * - To change a bit, write to this reg with a mask bit set and the - * bit of interest either set or cleared. EG: (BIT<<16) | BIT to set. - */ -#define Cache_Mode_0 0x2120 -#define CACHE_MODE_0 0x2120 -#define CM0_MASK_SHIFT 16 -#define CM0_IZ_OPT_DISABLE (1<<6) -#define CM0_ZR_OPT_DISABLE (1<<5) -#define CM0_DEPTH_EVICT_DISABLE (1<<4) -#define CM0_COLOR_EVICT_DISABLE (1<<3) -#define CM0_DEPTH_WRITE_DISABLE (1<<1) -#define CM0_RC_OP_FLUSH_DISABLE (1<<0) - - -/* Graphics flush control. A CPU write flushes the GWB of all writes. - * The data is discarded. - */ -#define GFX_FLSH_CNTL 0x2170 - -/* Binner control. Defines the location of the bin pointer list: - */ -#define BINCTL 0x2420 -#define BC_MASK (1 << 9) - -/* Binned scene info. - */ -#define BINSCENE 0x2428 -#define BS_OP_LOAD (1 << 8) -#define BS_MASK (1 << 22) - -/* Bin command parser debug reg: - */ -#define BCPD 0x2480 - -/* Bin memory control debug reg: - */ -#define BMCD 0x2484 - -/* Bin data cache debug reg: - */ -#define BDCD 0x2488 - -/* Binner pointer cache debug reg: - */ -#define BPCD 0x248c - -/* Binner scratch pad debug reg: - */ -#define BINSKPD 0x24f0 - -/* HWB scratch pad debug reg: - */ -#define HWBSKPD 0x24f4 - -/* Binner memory pool reg: - */ -#define BMP_BUFFER 0x2430 -#define BMP_PAGE_SIZE_4K (0 << 10) -#define BMP_BUFFER_SIZE_SHIFT 1 -#define BMP_ENABLE (1 << 0) - -/* Get/put memory from the binner memory pool: - */ -#define BMP_GET 0x2438 -#define BMP_PUT 0x2440 -#define BMP_OFFSET_SHIFT 5 - -/* 3D state packets: - */ -#define GFX_OP_RASTER_RULES ((0x3<<29)|(0x7<<24)) - -#define GFX_OP_SCISSOR ((0x3<<29)|(0x1c<<24)|(0x10<<19)) -#define SC_UPDATE_SCISSOR (0x1<<1) -#define SC_ENABLE_MASK (0x1<<0) -#define SC_ENABLE (0x1<<0) - -#define GFX_OP_LOAD_INDIRECT ((0x3<<29)|(0x1d<<24)|(0x7<<16)) - -#define GFX_OP_SCISSOR_INFO ((0x3<<29)|(0x1d<<24)|(0x81<<16)|(0x1)) -#define SCI_YMIN_MASK (0xffff<<16) -#define SCI_XMIN_MASK (0xffff<<0) -#define SCI_YMAX_MASK (0xffff<<16) -#define SCI_XMAX_MASK (0xffff<<0) - -#define GFX_OP_SCISSOR_ENABLE ((0x3<<29)|(0x1c<<24)|(0x10<<19)) -#define GFX_OP_SCISSOR_RECT ((0x3<<29)|(0x1d<<24)|(0x81<<16)|1) -#define GFX_OP_COLOR_FACTOR ((0x3<<29)|(0x1d<<24)|(0x1<<16)|0x0) -#define GFX_OP_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) -#define GFX_OP_MAP_INFO ((0x3<<29)|(0x1d<<24)|0x4) -#define GFX_OP_DESTBUFFER_VARS ((0x3<<29)|(0x1d<<24)|(0x85<<16)|0x0) -#define GFX_OP_DRAWRECT_INFO ((0x3<<29)|(0x1d<<24)|(0x80<<16)|(0x3)) - -#define GFX_OP_DRAWRECT_INFO_I965 ((0x7900<<16)|0x2) - -#define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|4) -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) -#define XY_SRC_COPY_BLT_SRC_TILED (1<<15) -#define XY_SRC_COPY_BLT_DST_TILED (1<<11) - -#define MI_BATCH_BUFFER ((0x30<<23)|1) -#define MI_BATCH_BUFFER_START (0x31<<23) -#define MI_BATCH_BUFFER_END (0xA<<23) -#define MI_BATCH_NON_SECURE (1) -#define MI_BATCH_NON_SECURE_I965 (1<<8) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) -#define MI_WAIT_FOR_PLANE_A_SCANLINES (1<<1) - -#define MI_LOAD_SCAN_LINES_INCL ((0x12<<23)) - -#define CMD_OP_DISPLAYBUFFER_INFO ((0x0<<29)|(0x14<<23)|2) -#define ASYNC_FLIP (1<<22) -#define DISPLAY_PLANE_A (0<<20) -#define DISPLAY_PLANE_B (1<<20) - -/* Display regs */ -#define DSPACNTR 0x70180 -#define DSPBCNTR 0x71180 -#define DISPPLANE_SEL_PIPE_MASK (1<<24) - -/* Define the region of interest for the binner: - */ -#define CMD_OP_BIN_CONTROL ((0x3<<29)|(0x1d<<24)|(0x84<<16)|4) - -#define CMD_OP_DESTBUFFER_INFO ((0x3<<29)|(0x1d<<24)|(0x8e<<16)|1) - -#define CMD_MI_FLUSH (0x04 << 23) -#define MI_NO_WRITE_FLUSH (1 << 2) -#define MI_READ_FLUSH (1 << 0) -#define MI_EXE_FLUSH (1 << 1) -#define MI_END_SCENE (1 << 4) /* flush binner and incr scene count */ -#define MI_SCENE_COUNT (1 << 3) /* just increment scene count */ - -#define BREADCRUMB_BITS 31 -#define BREADCRUMB_MASK ((1U << BREADCRUMB_BITS) - 1) - -#define READ_BREADCRUMB(dev_priv) (((volatile u32*)(dev_priv->hw_status_page))[5]) -#define READ_HWSP(dev_priv, reg) (((volatile u32*)(dev_priv->hw_status_page))[reg]) - -#define BLC_PWM_CTL 0x61254 -#define BACKLIGHT_MODULATION_FREQ_SHIFT (17) - -#define BLC_PWM_CTL2 0x61250 /** - * This is the most significant 15 bits of the number of backlight cycles in a - * complete cycle of the modulated backlight control. + * Reads a dword out of the status page, which is written to from the command + * queue by automatic updates, MI_REPORT_HEAD, MI_STORE_DATA_INDEX, or + * MI_STORE_DATA_IMM. * - * The actual value is this field multiplied by two. - */ -#define BACKLIGHT_MODULATION_FREQ_MASK (0x7fff << 17) -#define BLM_LEGACY_MODE (1 << 16) -/** - * This is the number of cycles out of the backlight modulation cycle for which - * the backlight is on. + * The following dwords have a reserved meaning: + * 0: ISR copy, updated when an ISR bit not set in the HWSTAM changes. + * 4: ring 0 head pointer + * 5: ring 1 head pointer (915-class) + * 6: ring 2 head pointer (915-class) * - * This field must be no greater than the number of cycles in the complete - * backlight modulation cycle. - */ -#define BACKLIGHT_DUTY_CYCLE_SHIFT (0) -#define BACKLIGHT_DUTY_CYCLE_MASK (0xffff) - -#define I915_GCFGC 0xf0 -#define I915_LOW_FREQUENCY_ENABLE (1 << 7) -#define I915_DISPLAY_CLOCK_190_200_MHZ (0 << 4) -#define I915_DISPLAY_CLOCK_333_MHZ (4 << 4) -#define I915_DISPLAY_CLOCK_MASK (7 << 4) - -#define I855_HPLLCC 0xc0 -#define I855_CLOCK_CONTROL_MASK (3 << 0) -#define I855_CLOCK_133_200 (0 << 0) -#define I855_CLOCK_100_200 (1 << 0) -#define I855_CLOCK_100_133 (2 << 0) -#define I855_CLOCK_166_250 (3 << 0) - -/* p317, 319 + * The area from dword 0x10 to 0x3ff is available for driver usage. */ -#define VCLK2_VCO_M 0x6008 /* treat as 16 bit? (includes msbs) */ -#define VCLK2_VCO_N 0x600a -#define VCLK2_VCO_DIV_SEL 0x6012 - -#define VCLK_DIVISOR_VGA0 0x6000 -#define VCLK_DIVISOR_VGA1 0x6004 -#define VCLK_POST_DIV 0x6010 -/** Selects a post divisor of 4 instead of 2. */ -# define VGA1_PD_P2_DIV_4 (1 << 15) -/** Overrides the p2 post divisor field */ -# define VGA1_PD_P1_DIV_2 (1 << 13) -# define VGA1_PD_P1_SHIFT 8 -/** P1 value is 2 greater than this field */ -# define VGA1_PD_P1_MASK (0x1f << 8) -/** Selects a post divisor of 4 instead of 2. */ -# define VGA0_PD_P2_DIV_4 (1 << 7) -/** Overrides the p2 post divisor field */ -# define VGA0_PD_P1_DIV_2 (1 << 5) -# define VGA0_PD_P1_SHIFT 0 -/** P1 value is 2 greater than this field */ -# define VGA0_PD_P1_MASK (0x1f << 0) - -/* PCI D state control register */ -#define D_STATE 0x6104 -#define DSPCLK_GATE_D 0x6200 - -/* I830 CRTC registers */ -#define HTOTAL_A 0x60000 -#define HBLANK_A 0x60004 -#define HSYNC_A 0x60008 -#define VTOTAL_A 0x6000c -#define VBLANK_A 0x60010 -#define VSYNC_A 0x60014 -#define PIPEASRC 0x6001c -#define BCLRPAT_A 0x60020 -#define VSYNCSHIFT_A 0x60028 - -#define HTOTAL_B 0x61000 -#define HBLANK_B 0x61004 -#define HSYNC_B 0x61008 -#define VTOTAL_B 0x6100c -#define VBLANK_B 0x61010 -#define VSYNC_B 0x61014 -#define PIPEBSRC 0x6101c -#define BCLRPAT_B 0x61020 -#define VSYNCSHIFT_B 0x61028 - -#define PP_STATUS 0x61200 -# define PP_ON (1 << 31) -/** - * Indicates that all dependencies of the panel are on: - * - * - PLL enabled - * - pipe enabled - * - LVDS/DVOB/DVOC on - */ -# define PP_READY (1 << 30) -# define PP_SEQUENCE_NONE (0 << 28) -# define PP_SEQUENCE_ON (1 << 28) -# define PP_SEQUENCE_OFF (2 << 28) -# define PP_SEQUENCE_MASK 0x30000000 -#define PP_CONTROL 0x61204 -# define POWER_TARGET_ON (1 << 0) - -#define LVDSPP_ON 0x61208 -#define LVDSPP_OFF 0x6120c -#define PP_CYCLE 0x61210 - -#define PFIT_CONTROL 0x61230 -# define PFIT_ENABLE (1 << 31) -# define PFIT_PIPE_MASK (3 << 29) -# define PFIT_PIPE_SHIFT 29 -# define VERT_INTERP_DISABLE (0 << 10) -# define VERT_INTERP_BILINEAR (1 << 10) -# define VERT_INTERP_MASK (3 << 10) -# define VERT_AUTO_SCALE (1 << 9) -# define HORIZ_INTERP_DISABLE (0 << 6) -# define HORIZ_INTERP_BILINEAR (1 << 6) -# define HORIZ_INTERP_MASK (3 << 6) -# define HORIZ_AUTO_SCALE (1 << 5) -# define PANEL_8TO6_DITHER_ENABLE (1 << 3) - -#define PFIT_PGM_RATIOS 0x61234 -# define PFIT_VERT_SCALE_MASK 0xfff00000 -# define PFIT_HORIZ_SCALE_MASK 0x0000fff0 - -#define PFIT_AUTO_RATIOS 0x61238 - - -#define DPLL_A 0x06014 -#define DPLL_B 0x06018 -# define DPLL_VCO_ENABLE (1 << 31) -# define DPLL_DVO_HIGH_SPEED (1 << 30) -# define DPLL_SYNCLOCK_ENABLE (1 << 29) -# define DPLL_VGA_MODE_DIS (1 << 28) -# define DPLLB_MODE_DAC_SERIAL (1 << 26) /* i915 */ -# define DPLLB_MODE_LVDS (2 << 26) /* i915 */ -# define DPLL_MODE_MASK (3 << 26) -# define DPLL_DAC_SERIAL_P2_CLOCK_DIV_10 (0 << 24) /* i915 */ -# define DPLL_DAC_SERIAL_P2_CLOCK_DIV_5 (1 << 24) /* i915 */ -# define DPLLB_LVDS_P2_CLOCK_DIV_14 (0 << 24) /* i915 */ -# define DPLLB_LVDS_P2_CLOCK_DIV_7 (1 << 24) /* i915 */ -# define DPLL_P2_CLOCK_DIV_MASK 0x03000000 /* i915 */ -# define DPLL_FPA01_P1_POST_DIV_MASK 0x00ff0000 /* i915 */ -/** - * The i830 generation, in DAC/serial mode, defines p1 as two plus this - * bitfield, or just 2 if PLL_P1_DIVIDE_BY_TWO is set. - */ -# define DPLL_FPA01_P1_POST_DIV_MASK_I830 0x001f0000 -/** - * The i830 generation, in LVDS mode, defines P1 as the bit number set within - * this field (only one bit may be set). - */ -# define DPLL_FPA01_P1_POST_DIV_MASK_I830_LVDS 0x003f0000 -# define DPLL_FPA01_P1_POST_DIV_SHIFT 16 -# define PLL_P2_DIVIDE_BY_4 (1 << 23) /* i830, required in DVO non-gang */ -# define PLL_P1_DIVIDE_BY_TWO (1 << 21) /* i830 */ -# define PLL_REF_INPUT_DREFCLK (0 << 13) -# define PLL_REF_INPUT_TVCLKINA (1 << 13) /* i830 */ -# define PLL_REF_INPUT_TVCLKINBC (2 << 13) /* SDVO TVCLKIN */ -# define PLLB_REF_INPUT_SPREADSPECTRUMIN (3 << 13) -# define PLL_REF_INPUT_MASK (3 << 13) -# define PLL_LOAD_PULSE_PHASE_SHIFT 9 -/* - * Parallel to Serial Load Pulse phase selection. - * Selects the phase for the 10X DPLL clock for the PCIe - * digital display port. The range is 4 to 13; 10 or more - * is just a flip delay. The default is 6 - */ -# define PLL_LOAD_PULSE_PHASE_MASK (0xf << PLL_LOAD_PULSE_PHASE_SHIFT) -# define DISPLAY_RATE_SELECT_FPA1 (1 << 8) - -/** - * SDVO multiplier for 945G/GM. Not used on 965. - * - * \sa DPLL_MD_UDI_MULTIPLIER_MASK - */ -# define SDVO_MULTIPLIER_MASK 0x000000ff -# define SDVO_MULTIPLIER_SHIFT_HIRES 4 -# define SDVO_MULTIPLIER_SHIFT_VGA 0 - -/** @defgroup DPLL_MD - * @{ - */ -/** Pipe A SDVO/UDI clock multiplier/divider register for G965. */ -#define DPLL_A_MD 0x0601c -/** Pipe B SDVO/UDI clock multiplier/divider register for G965. */ -#define DPLL_B_MD 0x06020 -/** - * UDI pixel divider, controlling how many pixels are stuffed into a packet. - * - * Value is pixels minus 1. Must be set to 1 pixel for SDVO. - */ -# define DPLL_MD_UDI_DIVIDER_MASK 0x3f000000 -# define DPLL_MD_UDI_DIVIDER_SHIFT 24 -/** UDI pixel divider for VGA, same as DPLL_MD_UDI_DIVIDER_MASK. */ -# define DPLL_MD_VGA_UDI_DIVIDER_MASK 0x003f0000 -# define DPLL_MD_VGA_UDI_DIVIDER_SHIFT 16 -/** - * SDVO/UDI pixel multiplier. - * - * SDVO requires that the bus clock rate be between 1 and 2 Ghz, and the bus - * clock rate is 10 times the DPLL clock. At low resolution/refresh rate - * modes, the bus rate would be below the limits, so SDVO allows for stuffing - * dummy bytes in the datastream at an increased clock rate, with both sides of - * the link knowing how many bytes are fill. - * - * So, for a mode with a dotclock of 65Mhz, we would want to double the clock - * rate to 130Mhz to get a bus rate of 1.30Ghz. The DPLL clock rate would be - * set to 130Mhz, and the SDVO multiplier set to 2x in this register and - * through an SDVO command. - * - * This register field has values of multiplication factor minus 1, with - * a maximum multiplier of 5 for SDVO. - */ -# define DPLL_MD_UDI_MULTIPLIER_MASK 0x00003f00 -# define DPLL_MD_UDI_MULTIPLIER_SHIFT 8 -/** SDVO/UDI pixel multiplier for VGA, same as DPLL_MD_UDI_MULTIPLIER_MASK. - * This best be set to the default value (3) or the CRT won't work. No, - * I don't entirely understand what this does... - */ -# define DPLL_MD_VGA_UDI_MULTIPLIER_MASK 0x0000003f -# define DPLL_MD_VGA_UDI_MULTIPLIER_SHIFT 0 -/** @} */ - -#define DPLL_TEST 0x606c -# define DPLLB_TEST_SDVO_DIV_1 (0 << 22) -# define DPLLB_TEST_SDVO_DIV_2 (1 << 22) -# define DPLLB_TEST_SDVO_DIV_4 (2 << 22) -# define DPLLB_TEST_SDVO_DIV_MASK (3 << 22) -# define DPLLB_TEST_N_BYPASS (1 << 19) -# define DPLLB_TEST_M_BYPASS (1 << 18) -# define DPLLB_INPUT_BUFFER_ENABLE (1 << 16) -# define DPLLA_TEST_N_BYPASS (1 << 3) -# define DPLLA_TEST_M_BYPASS (1 << 2) -# define DPLLA_INPUT_BUFFER_ENABLE (1 << 0) - -#define ADPA 0x61100 -#define ADPA_DAC_ENABLE (1<<31) -#define ADPA_DAC_DISABLE 0 -#define ADPA_PIPE_SELECT_MASK (1<<30) -#define ADPA_PIPE_A_SELECT 0 -#define ADPA_PIPE_B_SELECT (1<<30) -#define ADPA_USE_VGA_HVPOLARITY (1<<15) -#define ADPA_SETS_HVPOLARITY 0 -#define ADPA_VSYNC_CNTL_DISABLE (1<<11) -#define ADPA_VSYNC_CNTL_ENABLE 0 -#define ADPA_HSYNC_CNTL_DISABLE (1<<10) -#define ADPA_HSYNC_CNTL_ENABLE 0 -#define ADPA_VSYNC_ACTIVE_HIGH (1<<4) -#define ADPA_VSYNC_ACTIVE_LOW 0 -#define ADPA_HSYNC_ACTIVE_HIGH (1<<3) -#define ADPA_HSYNC_ACTIVE_LOW 0 - -#define FPA0 0x06040 -#define FPA1 0x06044 -#define FPB0 0x06048 -#define FPB1 0x0604c -# define FP_N_DIV_MASK 0x003f0000 -# define FP_N_DIV_SHIFT 16 -# define FP_M1_DIV_MASK 0x00003f00 -# define FP_M1_DIV_SHIFT 8 -# define FP_M2_DIV_MASK 0x0000003f -# define FP_M2_DIV_SHIFT 0 - - -#define PORT_HOTPLUG_EN 0x61110 -# define SDVOB_HOTPLUG_INT_EN (1 << 26) -# define SDVOC_HOTPLUG_INT_EN (1 << 25) -# define TV_HOTPLUG_INT_EN (1 << 18) -# define CRT_HOTPLUG_INT_EN (1 << 9) -# define CRT_HOTPLUG_FORCE_DETECT (1 << 3) - -#define PORT_HOTPLUG_STAT 0x61114 -# define CRT_HOTPLUG_INT_STATUS (1 << 11) -# define TV_HOTPLUG_INT_STATUS (1 << 10) -# define CRT_HOTPLUG_MONITOR_MASK (3 << 8) -# define CRT_HOTPLUG_MONITOR_COLOR (3 << 8) -# define CRT_HOTPLUG_MONITOR_MONO (2 << 8) -# define CRT_HOTPLUG_MONITOR_NONE (0 << 8) -# define SDVOC_HOTPLUG_INT_STATUS (1 << 7) -# define SDVOB_HOTPLUG_INT_STATUS (1 << 6) - -#define SDVOB 0x61140 -#define SDVOC 0x61160 -#define SDVO_ENABLE (1 << 31) -#define SDVO_PIPE_B_SELECT (1 << 30) -#define SDVO_STALL_SELECT (1 << 29) -#define SDVO_INTERRUPT_ENABLE (1 << 26) -/** - * 915G/GM SDVO pixel multiplier. - * - * Programmed value is multiplier - 1, up to 5x. - * - * \sa DPLL_MD_UDI_MULTIPLIER_MASK - */ -#define SDVO_PORT_MULTIPLY_MASK (7 << 23) -#define SDVO_PORT_MULTIPLY_SHIFT 23 -#define SDVO_PHASE_SELECT_MASK (15 << 19) -#define SDVO_PHASE_SELECT_DEFAULT (6 << 19) -#define SDVO_CLOCK_OUTPUT_INVERT (1 << 18) -#define SDVOC_GANG_MODE (1 << 16) -#define SDVO_BORDER_ENABLE (1 << 7) -#define SDVOB_PCIE_CONCURRENCY (1 << 3) -#define SDVO_DETECTED (1 << 2) -/* Bits to be preserved when writing */ -#define SDVOB_PRESERVE_MASK ((1 << 17) | (1 << 16) | (1 << 14)) -#define SDVOC_PRESERVE_MASK (1 << 17) - -/** @defgroup LVDS - * @{ - */ -/** - * This register controls the LVDS output enable, pipe selection, and data - * format selection. - * - * All of the clock/data pairs are force powered down by power sequencing. - */ -#define LVDS 0x61180 -/** - * Enables the LVDS port. This bit must be set before DPLLs are enabled, as - * the DPLL semantics change when the LVDS is assigned to that pipe. - */ -# define LVDS_PORT_EN (1 << 31) -/** Selects pipe B for LVDS data. Must be set on pre-965. */ -# define LVDS_PIPEB_SELECT (1 << 30) - -/** - * Enables the A0-A2 data pairs and CLKA, containing 18 bits of color data per - * pixel. - */ -# define LVDS_A0A2_CLKA_POWER_MASK (3 << 8) -# define LVDS_A0A2_CLKA_POWER_DOWN (0 << 8) -# define LVDS_A0A2_CLKA_POWER_UP (3 << 8) -/** - * Controls the A3 data pair, which contains the additional LSBs for 24 bit - * mode. Only enabled if LVDS_A0A2_CLKA_POWER_UP also indicates it should be - * on. - */ -# define LVDS_A3_POWER_MASK (3 << 6) -# define LVDS_A3_POWER_DOWN (0 << 6) -# define LVDS_A3_POWER_UP (3 << 6) -/** - * Controls the CLKB pair. This should only be set when LVDS_B0B3_POWER_UP - * is set. - */ -# define LVDS_CLKB_POWER_MASK (3 << 4) -# define LVDS_CLKB_POWER_DOWN (0 << 4) -# define LVDS_CLKB_POWER_UP (3 << 4) - -/** - * Controls the B0-B3 data pairs. This must be set to match the DPLL p2 - * setting for whether we are in dual-channel mode. The B3 pair will - * additionally only be powered up when LVDS_A3_POWER_UP is set. - */ -# define LVDS_B0B3_POWER_MASK (3 << 2) -# define LVDS_B0B3_POWER_DOWN (0 << 2) -# define LVDS_B0B3_POWER_UP (3 << 2) - -#define PIPEACONF 0x70008 -#define PIPEACONF_ENABLE (1<<31) -#define PIPEACONF_DISABLE 0 -#define PIPEACONF_DOUBLE_WIDE (1<<30) -#define I965_PIPECONF_ACTIVE (1<<30) -#define PIPEACONF_SINGLE_WIDE 0 -#define PIPEACONF_PIPE_UNLOCKED 0 -#define PIPEACONF_PIPE_LOCKED (1<<25) -#define PIPEACONF_PALETTE 0 -#define PIPEACONF_GAMMA (1<<24) -#define PIPECONF_FORCE_BORDER (1<<25) -#define PIPECONF_PROGRESSIVE (0 << 21) -#define PIPECONF_INTERLACE_W_FIELD_INDICATION (6 << 21) -#define PIPECONF_INTERLACE_FIELD_0_ONLY (7 << 21) - -#define DSPARB 0x70030 -#define DSPARB_CSTART_MASK (0x7f << 7) -#define DSPARB_CSTART_SHIFT 7 -#define DSPARB_BSTART_MASK (0x7f) -#define DSPARB_BSTART_SHIFT 0 - -#define PIPEBCONF 0x71008 -#define PIPEBCONF_ENABLE (1<<31) -#define PIPEBCONF_DISABLE 0 -#define PIPEBCONF_DOUBLE_WIDE (1<<30) -#define PIPEBCONF_DISABLE 0 -#define PIPEBCONF_GAMMA (1<<24) -#define PIPEBCONF_PALETTE 0 - -#define PIPEBGCMAXRED 0x71010 -#define PIPEBGCMAXGREEN 0x71014 -#define PIPEBGCMAXBLUE 0x71018 -#define PIPEBSTAT 0x71024 -#define PIPEBFRAMEHIGH 0x71040 -#define PIPEBFRAMEPIXEL 0x71044 - -#define DSPACNTR 0x70180 -#define DSPBCNTR 0x71180 -#define DISPLAY_PLANE_ENABLE (1<<31) -#define DISPLAY_PLANE_DISABLE 0 -#define DISPPLANE_GAMMA_ENABLE (1<<30) -#define DISPPLANE_GAMMA_DISABLE 0 -#define DISPPLANE_PIXFORMAT_MASK (0xf<<26) -#define DISPPLANE_8BPP (0x2<<26) -#define DISPPLANE_15_16BPP (0x4<<26) -#define DISPPLANE_16BPP (0x5<<26) -#define DISPPLANE_32BPP_NO_ALPHA (0x6<<26) -#define DISPPLANE_32BPP (0x7<<26) -#define DISPPLANE_STEREO_ENABLE (1<<25) -#define DISPPLANE_STEREO_DISABLE 0 -#define DISPPLANE_SEL_PIPE_MASK (1<<24) -#define DISPPLANE_SEL_PIPE_A 0 -#define DISPPLANE_SEL_PIPE_B (1<<24) -#define DISPPLANE_SRC_KEY_ENABLE (1<<22) -#define DISPPLANE_SRC_KEY_DISABLE 0 -#define DISPPLANE_LINE_DOUBLE (1<<20) -#define DISPPLANE_NO_LINE_DOUBLE 0 -#define DISPPLANE_STEREO_POLARITY_FIRST 0 -#define DISPPLANE_STEREO_POLARITY_SECOND (1<<18) -/* plane B only */ -#define DISPPLANE_ALPHA_TRANS_ENABLE (1<<15) -#define DISPPLANE_ALPHA_TRANS_DISABLE 0 -#define DISPPLANE_SPRITE_ABOVE_DISPLAYA 0 -#define DISPPLANE_SPRITE_ABOVE_OVERLAY (1) - -#define DSPABASE 0x70184 -#define DSPASTRIDE 0x70188 - -#define DSPBBASE 0x71184 -#define DSPBADDR DSPBBASE -#define DSPBSTRIDE 0x71188 - -#define DSPAKEYVAL 0x70194 -#define DSPAKEYMASK 0x70198 - -#define DSPAPOS 0x7018C /* reserved */ -#define DSPASIZE 0x70190 -#define DSPBPOS 0x7118C -#define DSPBSIZE 0x71190 - -#define DSPASURF 0x7019C -#define DSPATILEOFF 0x701A4 - -#define DSPBSURF 0x7119C -#define DSPBTILEOFF 0x711A4 - -#define VGACNTRL 0x71400 -# define VGA_DISP_DISABLE (1 << 31) -# define VGA_2X_MODE (1 << 30) -# define VGA_PIPE_B_SELECT (1 << 29) - -/* - * Some BIOS scratch area registers. The 845 (and 830?) store the amount - * of video memory available to the BIOS in SWF1. - */ - -#define SWF0 0x71410 - -/* - * 855 scratch registers. - */ -#define SWF10 0x70410 - -#define SWF30 0x72414 - -/* - * Overlay registers. These are overlay registers accessed via MMIO. - * Those loaded via the overlay register page are defined in i830_video.c. - */ -#define OVADD 0x30000 - -#define DOVSTA 0x30008 -#define OC_BUF (0x3<<20) +#define READ_HWSP(dev_priv, reg) (((volatile u32*)(dev_priv->hw_status_page))[reg]) +#define READ_BREADCRUMB(dev_priv) READ_HWSP(dev_priv, 5) -#define OGAMC5 0x30010 -#define OGAMC4 0x30014 -#define OGAMC3 0x30018 -#define OGAMC2 0x3001c -#define OGAMC1 0x30020 -#define OGAMC0 0x30024 -/* - * Palette registers - */ -#define PALETTE_A 0x0a000 -#define PALETTE_B 0x0a800 +extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define IS_I830(dev) ((dev)->pci_device == 0x3577) #define IS_845G(dev) ((dev)->pci_device == 0x2562) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index df036118b8b..4a2de789734 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -31,10 +31,6 @@ #include "i915_drm.h" #include "i915_drv.h" -#define USER_INT_FLAG (1<<1) -#define VSYNC_PIPEB_FLAG (1<<5) -#define VSYNC_PIPEA_FLAG (1<<7) - #define MAX_NOPID ((u32)~0) /** @@ -236,40 +232,43 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) u16 temp; u32 pipea_stats, pipeb_stats; - pipea_stats = I915_READ(I915REG_PIPEASTAT); - pipeb_stats = I915_READ(I915REG_PIPEBSTAT); + pipea_stats = I915_READ(PIPEASTAT); + pipeb_stats = I915_READ(PIPEBSTAT); - temp = I915_READ16(I915REG_INT_IDENTITY_R); + temp = I915_READ16(IIR); - temp &= (USER_INT_FLAG | VSYNC_PIPEA_FLAG | VSYNC_PIPEB_FLAG); + temp &= (I915_USER_INTERRUPT | + I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | + I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT); DRM_DEBUG("%s flag=%08x\n", __FUNCTION__, temp); if (temp == 0) return IRQ_NONE; - I915_WRITE16(I915REG_INT_IDENTITY_R, temp); - (void) I915_READ16(I915REG_INT_IDENTITY_R); + I915_WRITE16(IIR, temp); + (void) I915_READ16(IIR); DRM_READMEMORYBARRIER(); dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - if (temp & USER_INT_FLAG) + if (temp & I915_USER_INTERRUPT) DRM_WAKEUP(&dev_priv->irq_queue); - if (temp & (VSYNC_PIPEA_FLAG | VSYNC_PIPEB_FLAG)) { + if (temp & (I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | + I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT)) { int vblank_pipe = dev_priv->vblank_pipe; if ((vblank_pipe & (DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B)) == (DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B)) { - if (temp & VSYNC_PIPEA_FLAG) + if (temp & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) atomic_inc(&dev->vbl_received); - if (temp & VSYNC_PIPEB_FLAG) + if (temp & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) atomic_inc(&dev->vbl_received2); - } else if (((temp & VSYNC_PIPEA_FLAG) && + } else if (((temp & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) && (vblank_pipe & DRM_I915_VBLANK_PIPE_A)) || - ((temp & VSYNC_PIPEB_FLAG) && + ((temp & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) && (vblank_pipe & DRM_I915_VBLANK_PIPE_B))) atomic_inc(&dev->vbl_received); @@ -278,12 +277,12 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) if (dev_priv->swaps_pending > 0) drm_locked_tasklet(dev, i915_vblank_tasklet); - I915_WRITE(I915REG_PIPEASTAT, + I915_WRITE(PIPEASTAT, pipea_stats|I915_VBLANK_INTERRUPT_ENABLE| - I915_VBLANK_CLEAR); - I915_WRITE(I915REG_PIPEBSTAT, + PIPE_VBLANK_INTERRUPT_STATUS); + I915_WRITE(PIPEBSTAT, pipeb_stats|I915_VBLANK_INTERRUPT_ENABLE| - I915_VBLANK_CLEAR); + PIPE_VBLANK_INTERRUPT_STATUS); } return IRQ_HANDLED; @@ -304,12 +303,12 @@ static int i915_emit_irq(struct drm_device * dev) dev_priv->sarea_priv->last_enqueue = dev_priv->counter = 1; BEGIN_LP_RING(6); - OUT_RING(CMD_STORE_DWORD_IDX); - OUT_RING(20); + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(5 << MI_STORE_DWORD_INDEX_SHIFT); OUT_RING(dev_priv->counter); OUT_RING(0); OUT_RING(0); - OUT_RING(GFX_OP_USER_INTERRUPT); + OUT_RING(MI_USER_INTERRUPT); ADVANCE_LP_RING(); return dev_priv->counter; @@ -421,11 +420,11 @@ static void i915_enable_interrupt (struct drm_device *dev) flag = 0; if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_A) - flag |= VSYNC_PIPEA_FLAG; + flag |= I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B) - flag |= VSYNC_PIPEB_FLAG; + flag |= I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; - I915_WRITE16(I915REG_INT_ENABLE_R, USER_INT_FLAG | flag); + I915_WRITE16(IER, I915_USER_INTERRUPT | flag); } /* Set the vblank monitor pipe @@ -465,11 +464,11 @@ int i915_vblank_pipe_get(struct drm_device *dev, void *data, return -EINVAL; } - flag = I915_READ(I915REG_INT_ENABLE_R); + flag = I915_READ(IER); pipe->pipe = 0; - if (flag & VSYNC_PIPEA_FLAG) + if (flag & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) pipe->pipe |= DRM_I915_VBLANK_PIPE_A; - if (flag & VSYNC_PIPEB_FLAG) + if (flag & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) pipe->pipe |= DRM_I915_VBLANK_PIPE_B; return 0; @@ -587,9 +586,9 @@ void i915_driver_irq_preinstall(struct drm_device * dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - I915_WRITE16(I915REG_HWSTAM, 0xfffe); - I915_WRITE16(I915REG_INT_MASK_R, 0x0); - I915_WRITE16(I915REG_INT_ENABLE_R, 0x0); + I915_WRITE16(HWSTAM, 0xfffe); + I915_WRITE16(IMR, 0x0); + I915_WRITE16(IER, 0x0); } void i915_driver_irq_postinstall(struct drm_device * dev) @@ -614,10 +613,10 @@ void i915_driver_irq_uninstall(struct drm_device * dev) if (!dev_priv) return; - I915_WRITE16(I915REG_HWSTAM, 0xffff); - I915_WRITE16(I915REG_INT_MASK_R, 0xffff); - I915_WRITE16(I915REG_INT_ENABLE_R, 0x0); + I915_WRITE16(HWSTAM, 0xffff); + I915_WRITE16(IMR, 0xffff); + I915_WRITE16(IER, 0x0); - temp = I915_READ16(I915REG_INT_IDENTITY_R); - I915_WRITE16(I915REG_INT_IDENTITY_R, temp); + temp = I915_READ16(IIR); + I915_WRITE16(IIR, temp); } diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h new file mode 100644 index 00000000000..477c64e6ba2 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -0,0 +1,1405 @@ +/* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef _I915_REG_H_ +#define _I915_REG_H_ + +/* MCH MMIO space */ +/** 915-945 and GM965 MCH register controlling DRAM channel access */ +#define DCC 0x200 +#define DCC_ADDRESSING_MODE_SINGLE_CHANNEL (0 << 0) +#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_ASYMMETRIC (1 << 0) +#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_INTERLEAVED (2 << 0) +#define DCC_ADDRESSING_MODE_MASK (3 << 0) +#define DCC_CHANNEL_XOR_DISABLE (1 << 10) + +/** 965 MCH register controlling DRAM channel configuration */ +#define CHDECMISC 0x111 +#define CHDECMISC_FLEXMEMORY (1 << 1) + +/* + * The Bridge device's PCI config space has information about the + * fb aperture size and the amount of pre-reserved memory. + */ +#define INTEL_GMCH_CTRL 0x52 +#define INTEL_GMCH_ENABLED 0x4 +#define INTEL_GMCH_MEM_MASK 0x1 +#define INTEL_GMCH_MEM_64M 0x1 +#define INTEL_GMCH_MEM_128M 0 + +#define INTEL_855_GMCH_GMS_MASK (0x7 << 4) +#define INTEL_855_GMCH_GMS_DISABLED (0x0 << 4) +#define INTEL_855_GMCH_GMS_STOLEN_1M (0x1 << 4) +#define INTEL_855_GMCH_GMS_STOLEN_4M (0x2 << 4) +#define INTEL_855_GMCH_GMS_STOLEN_8M (0x3 << 4) +#define INTEL_855_GMCH_GMS_STOLEN_16M (0x4 << 4) +#define INTEL_855_GMCH_GMS_STOLEN_32M (0x5 << 4) + +#define INTEL_915G_GMCH_GMS_STOLEN_48M (0x6 << 4) +#define INTEL_915G_GMCH_GMS_STOLEN_64M (0x7 << 4) + +/* PCI config space */ + +#define HPLLCC 0xc0 /* 855 only */ +#define GC_CLOCK_CONTROL_MASK (3 << 0) +#define GC_CLOCK_133_200 (0 << 0) +#define GC_CLOCK_100_200 (1 << 0) +#define GC_CLOCK_100_133 (2 << 0) +#define GC_CLOCK_166_250 (3 << 0) +#define GCFGC 0xf0 /* 915+ only */ +#define GC_LOW_FREQUENCY_ENABLE (1 << 7) +#define GC_DISPLAY_CLOCK_190_200_MHZ (0 << 4) +#define GC_DISPLAY_CLOCK_333_MHZ (4 << 4) +#define GC_DISPLAY_CLOCK_MASK (7 << 4) +#define LBB 0xf4 + +/* VGA stuff */ + +#define VGA_ST01_MDA 0x3ba +#define VGA_ST01_CGA 0x3da + +#define VGA_MSR_WRITE 0x3c2 +#define VGA_MSR_READ 0x3cc +#define VGA_MSR_MEM_EN (1<<1) +#define VGA_MSR_CGA_MODE (1<<0) + +#define VGA_SR_INDEX 0x3c4 +#define VGA_SR_DATA 0x3c5 + +#define VGA_AR_INDEX 0x3c0 +#define VGA_AR_VID_EN (1<<5) +#define VGA_AR_DATA_WRITE 0x3c0 +#define VGA_AR_DATA_READ 0x3c1 + +#define VGA_GR_INDEX 0x3ce +#define VGA_GR_DATA 0x3cf +/* GR05 */ +#define VGA_GR_MEM_READ_MODE_SHIFT 3 +#define VGA_GR_MEM_READ_MODE_PLANE 1 +/* GR06 */ +#define VGA_GR_MEM_MODE_MASK 0xc +#define VGA_GR_MEM_MODE_SHIFT 2 +#define VGA_GR_MEM_A0000_AFFFF 0 +#define VGA_GR_MEM_A0000_BFFFF 1 +#define VGA_GR_MEM_B0000_B7FFF 2 +#define VGA_GR_MEM_B0000_BFFFF 3 + +#define VGA_DACMASK 0x3c6 +#define VGA_DACRX 0x3c7 +#define VGA_DACWX 0x3c8 +#define VGA_DACDATA 0x3c9 + +#define VGA_CR_INDEX_MDA 0x3b4 +#define VGA_CR_DATA_MDA 0x3b5 +#define VGA_CR_INDEX_CGA 0x3d4 +#define VGA_CR_DATA_CGA 0x3d5 + +/* + * Memory interface instructions used by the kernel + */ +#define MI_INSTR(opcode, flags) (((opcode) << 23) | (flags)) + +#define MI_NOOP MI_INSTR(0, 0) +#define MI_USER_INTERRUPT MI_INSTR(0x02, 0) +#define MI_WAIT_FOR_EVENT MI_INSTR(0x03, 0) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) +#define MI_WAIT_FOR_PLANE_A_SCANLINES (1<<1) +#define MI_FLUSH MI_INSTR(0x04, 0) +#define MI_READ_FLUSH (1 << 0) +#define MI_EXE_FLUSH (1 << 1) +#define MI_NO_WRITE_FLUSH (1 << 2) +#define MI_SCENE_COUNT (1 << 3) /* just increment scene count */ +#define MI_END_SCENE (1 << 4) /* flush binner and incr scene count */ +#define MI_BATCH_BUFFER_END MI_INSTR(0x0a, 0) +#define MI_REPORT_HEAD MI_INSTR(0x07, 0) +#define MI_LOAD_SCAN_LINES_INCL MI_INSTR(0x12, 0) +#define MI_STORE_DWORD_IMM MI_INSTR(0x20, 1) +#define MI_MEM_VIRTUAL (1 << 22) /* 965+ only */ +#define MI_STORE_DWORD_INDEX MI_INSTR(0x21, 1) +#define MI_STORE_DWORD_INDEX_SHIFT 2 +#define MI_LOAD_REGISTER_IMM MI_INSTR(0x22, 1) +#define MI_BATCH_BUFFER MI_INSTR(0x30, 1) +#define MI_BATCH_NON_SECURE (1) +#define MI_BATCH_NON_SECURE_I965 (1<<8) +#define MI_BATCH_BUFFER_START MI_INSTR(0x31, 0) + +/* + * 3D instructions used by the kernel + */ +#define GFX_INSTR(opcode, flags) ((0x3 << 29) | ((opcode) << 24) | (flags)) + +#define GFX_OP_RASTER_RULES ((0x3<<29)|(0x7<<24)) +#define GFX_OP_SCISSOR ((0x3<<29)|(0x1c<<24)|(0x10<<19)) +#define SC_UPDATE_SCISSOR (0x1<<1) +#define SC_ENABLE_MASK (0x1<<0) +#define SC_ENABLE (0x1<<0) +#define GFX_OP_LOAD_INDIRECT ((0x3<<29)|(0x1d<<24)|(0x7<<16)) +#define GFX_OP_SCISSOR_INFO ((0x3<<29)|(0x1d<<24)|(0x81<<16)|(0x1)) +#define SCI_YMIN_MASK (0xffff<<16) +#define SCI_XMIN_MASK (0xffff<<0) +#define SCI_YMAX_MASK (0xffff<<16) +#define SCI_XMAX_MASK (0xffff<<0) +#define GFX_OP_SCISSOR_ENABLE ((0x3<<29)|(0x1c<<24)|(0x10<<19)) +#define GFX_OP_SCISSOR_RECT ((0x3<<29)|(0x1d<<24)|(0x81<<16)|1) +#define GFX_OP_COLOR_FACTOR ((0x3<<29)|(0x1d<<24)|(0x1<<16)|0x0) +#define GFX_OP_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) +#define GFX_OP_MAP_INFO ((0x3<<29)|(0x1d<<24)|0x4) +#define GFX_OP_DESTBUFFER_VARS ((0x3<<29)|(0x1d<<24)|(0x85<<16)|0x0) +#define GFX_OP_DESTBUFFER_INFO ((0x3<<29)|(0x1d<<24)|(0x8e<<16)|1) +#define GFX_OP_DRAWRECT_INFO ((0x3<<29)|(0x1d<<24)|(0x80<<16)|(0x3)) +#define GFX_OP_DRAWRECT_INFO_I965 ((0x7900<<16)|0x2) +#define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|4) +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_MONO_SRC_COPY_IMM_BLT ((2<<29)|(0x71<<22)|5) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) +#define BLT_DEPTH_8 (0<<24) +#define BLT_DEPTH_16_565 (1<<24) +#define BLT_DEPTH_16_1555 (2<<24) +#define BLT_DEPTH_32 (3<<24) +#define BLT_ROP_GXCOPY (0xcc<<16) +#define XY_SRC_COPY_BLT_SRC_TILED (1<<15) /* 965+ only */ +#define XY_SRC_COPY_BLT_DST_TILED (1<<11) /* 965+ only */ +#define CMD_OP_DISPLAYBUFFER_INFO ((0x0<<29)|(0x14<<23)|2) +#define ASYNC_FLIP (1<<22) +#define DISPLAY_PLANE_A (0<<20) +#define DISPLAY_PLANE_B (1<<20) + +/* + * Instruction and interrupt control regs + */ + +#define PRB0_TAIL 0x02030 +#define PRB0_HEAD 0x02034 +#define PRB0_START 0x02038 +#define PRB0_CTL 0x0203c +#define TAIL_ADDR 0x001FFFF8 +#define HEAD_WRAP_COUNT 0xFFE00000 +#define HEAD_WRAP_ONE 0x00200000 +#define HEAD_ADDR 0x001FFFFC +#define RING_NR_PAGES 0x001FF000 +#define RING_REPORT_MASK 0x00000006 +#define RING_REPORT_64K 0x00000002 +#define RING_REPORT_128K 0x00000004 +#define RING_NO_REPORT 0x00000000 +#define RING_VALID_MASK 0x00000001 +#define RING_VALID 0x00000001 +#define RING_INVALID 0x00000000 +#define PRB1_TAIL 0x02040 /* 915+ only */ +#define PRB1_HEAD 0x02044 /* 915+ only */ +#define PRB1_START 0x02048 /* 915+ only */ +#define PRB1_CTL 0x0204c /* 915+ only */ +#define ACTHD_I965 0x02074 +#define HWS_PGA 0x02080 +#define HWS_ADDRESS_MASK 0xfffff000 +#define HWS_START_ADDRESS_SHIFT 4 +#define IPEIR 0x02088 +#define NOPID 0x02094 +#define HWSTAM 0x02098 +#define SCPD0 0x0209c /* 915+ only */ +#define IER 0x020a0 +#define IIR 0x020a4 +#define IMR 0x020a8 +#define ISR 0x020ac +#define I915_PIPE_CONTROL_NOTIFY_INTERRUPT (1<<18) +#define I915_DISPLAY_PORT_INTERRUPT (1<<17) +#define I915_RENDER_COMMAND_PARSER_ERROR_INTERRUPT (1<<15) +#define I915_GMCH_THERMAL_SENSOR_EVENT_INTERRUPT (1<<14) +#define I915_HWB_OOM_INTERRUPT (1<<13) +#define I915_SYNC_STATUS_INTERRUPT (1<<12) +#define I915_DISPLAY_PLANE_A_FLIP_PENDING_INTERRUPT (1<<11) +#define I915_DISPLAY_PLANE_B_FLIP_PENDING_INTERRUPT (1<<10) +#define I915_OVERLAY_PLANE_FLIP_PENDING_INTERRUPT (1<<9) +#define I915_DISPLAY_PLANE_C_FLIP_PENDING_INTERRUPT (1<<8) +#define I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT (1<<7) +#define I915_DISPLAY_PIPE_A_EVENT_INTERRUPT (1<<6) +#define I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT (1<<5) +#define I915_DISPLAY_PIPE_B_EVENT_INTERRUPT (1<<4) +#define I915_DEBUG_INTERRUPT (1<<2) +#define I915_USER_INTERRUPT (1<<1) +#define I915_ASLE_INTERRUPT (1<<0) +#define EIR 0x020b0 +#define EMR 0x020b4 +#define ESR 0x020b8 +#define INSTPM 0x020c0 +#define ACTHD 0x020c8 +#define FW_BLC 0x020d8 +#define FW_BLC_SELF 0x020e0 /* 915+ only */ +#define MI_ARB_STATE 0x020e4 /* 915+ only */ +#define CACHE_MODE_0 0x02120 /* 915+ only */ +#define CM0_MASK_SHIFT 16 +#define CM0_IZ_OPT_DISABLE (1<<6) +#define CM0_ZR_OPT_DISABLE (1<<5) +#define CM0_DEPTH_EVICT_DISABLE (1<<4) +#define CM0_COLOR_EVICT_DISABLE (1<<3) +#define CM0_DEPTH_WRITE_DISABLE (1<<1) +#define CM0_RC_OP_FLUSH_DISABLE (1<<0) +#define GFX_FLSH_CNTL 0x02170 /* 915+ only */ + +/* + * Framebuffer compression (915+ only) + */ + +#define FBC_CFB_BASE 0x03200 /* 4k page aligned */ +#define FBC_LL_BASE 0x03204 /* 4k page aligned */ +#define FBC_CONTROL 0x03208 +#define FBC_CTL_EN (1<<31) +#define FBC_CTL_PERIODIC (1<<30) +#define FBC_CTL_INTERVAL_SHIFT (16) +#define FBC_CTL_UNCOMPRESSIBLE (1<<14) +#define FBC_CTL_STRIDE_SHIFT (5) +#define FBC_CTL_FENCENO (1<<0) +#define FBC_COMMAND 0x0320c +#define FBC_CMD_COMPRESS (1<<0) +#define FBC_STATUS 0x03210 +#define FBC_STAT_COMPRESSING (1<<31) +#define FBC_STAT_COMPRESSED (1<<30) +#define FBC_STAT_MODIFIED (1<<29) +#define FBC_STAT_CURRENT_LINE (1<<0) +#define FBC_CONTROL2 0x03214 +#define FBC_CTL_FENCE_DBL (0<<4) +#define FBC_CTL_IDLE_IMM (0<<2) +#define FBC_CTL_IDLE_FULL (1<<2) +#define FBC_CTL_IDLE_LINE (2<<2) +#define FBC_CTL_IDLE_DEBUG (3<<2) +#define FBC_CTL_CPU_FENCE (1<<1) +#define FBC_CTL_PLANEA (0<<0) +#define FBC_CTL_PLANEB (1<<0) +#define FBC_FENCE_OFF 0x0321b + +#define FBC_LL_SIZE (1536) + +/* + * GPIO regs + */ +#define GPIOA 0x5010 +#define GPIOB 0x5014 +#define GPIOC 0x5018 +#define GPIOD 0x501c +#define GPIOE 0x5020 +#define GPIOF 0x5024 +#define GPIOG 0x5028 +#define GPIOH 0x502c +# define GPIO_CLOCK_DIR_MASK (1 << 0) +# define GPIO_CLOCK_DIR_IN (0 << 1) +# define GPIO_CLOCK_DIR_OUT (1 << 1) +# define GPIO_CLOCK_VAL_MASK (1 << 2) +# define GPIO_CLOCK_VAL_OUT (1 << 3) +# define GPIO_CLOCK_VAL_IN (1 << 4) +# define GPIO_CLOCK_PULLUP_DISABLE (1 << 5) +# define GPIO_DATA_DIR_MASK (1 << 8) +# define GPIO_DATA_DIR_IN (0 << 9) +# define GPIO_DATA_DIR_OUT (1 << 9) +# define GPIO_DATA_VAL_MASK (1 << 10) +# define GPIO_DATA_VAL_OUT (1 << 11) +# define GPIO_DATA_VAL_IN (1 << 12) +# define GPIO_DATA_PULLUP_DISABLE (1 << 13) + +/* + * Clock control & power management + */ + +#define VGA0 0x6000 +#define VGA1 0x6004 +#define VGA_PD 0x6010 +#define VGA0_PD_P2_DIV_4 (1 << 7) +#define VGA0_PD_P1_DIV_2 (1 << 5) +#define VGA0_PD_P1_SHIFT 0 +#define VGA0_PD_P1_MASK (0x1f << 0) +#define VGA1_PD_P2_DIV_4 (1 << 15) +#define VGA1_PD_P1_DIV_2 (1 << 13) +#define VGA1_PD_P1_SHIFT 8 +#define VGA1_PD_P1_MASK (0x1f << 8) +#define DPLL_A 0x06014 +#define DPLL_B 0x06018 +#define DPLL_VCO_ENABLE (1 << 31) +#define DPLL_DVO_HIGH_SPEED (1 << 30) +#define DPLL_SYNCLOCK_ENABLE (1 << 29) +#define DPLL_VGA_MODE_DIS (1 << 28) +#define DPLLB_MODE_DAC_SERIAL (1 << 26) /* i915 */ +#define DPLLB_MODE_LVDS (2 << 26) /* i915 */ +#define DPLL_MODE_MASK (3 << 26) +#define DPLL_DAC_SERIAL_P2_CLOCK_DIV_10 (0 << 24) /* i915 */ +#define DPLL_DAC_SERIAL_P2_CLOCK_DIV_5 (1 << 24) /* i915 */ +#define DPLLB_LVDS_P2_CLOCK_DIV_14 (0 << 24) /* i915 */ +#define DPLLB_LVDS_P2_CLOCK_DIV_7 (1 << 24) /* i915 */ +#define DPLL_P2_CLOCK_DIV_MASK 0x03000000 /* i915 */ +#define DPLL_FPA01_P1_POST_DIV_MASK 0x00ff0000 /* i915 */ + +#define I915_FIFO_UNDERRUN_STATUS (1UL<<31) +#define I915_CRC_ERROR_ENABLE (1UL<<29) +#define I915_CRC_DONE_ENABLE (1UL<<28) +#define I915_GMBUS_EVENT_ENABLE (1UL<<27) +#define I915_VSYNC_INTERRUPT_ENABLE (1UL<<25) +#define I915_DISPLAY_LINE_COMPARE_ENABLE (1UL<<24) +#define I915_DPST_EVENT_ENABLE (1UL<<23) +#define I915_LEGACY_BLC_EVENT_ENABLE (1UL<<22) +#define I915_ODD_FIELD_INTERRUPT_ENABLE (1UL<<21) +#define I915_EVEN_FIELD_INTERRUPT_ENABLE (1UL<<20) +#define I915_START_VBLANK_INTERRUPT_ENABLE (1UL<<18) /* 965 or later */ +#define I915_VBLANK_INTERRUPT_ENABLE (1UL<<17) +#define I915_OVERLAY_UPDATED_ENABLE (1UL<<16) +#define I915_CRC_ERROR_INTERRUPT_STATUS (1UL<<13) +#define I915_CRC_DONE_INTERRUPT_STATUS (1UL<<12) +#define I915_GMBUS_INTERRUPT_STATUS (1UL<<11) +#define I915_VSYNC_INTERRUPT_STATUS (1UL<<9) +#define I915_DISPLAY_LINE_COMPARE_STATUS (1UL<<8) +#define I915_DPST_EVENT_STATUS (1UL<<7) +#define I915_LEGACY_BLC_EVENT_STATUS (1UL<<6) +#define I915_ODD_FIELD_INTERRUPT_STATUS (1UL<<5) +#define I915_EVEN_FIELD_INTERRUPT_STATUS (1UL<<4) +#define I915_START_VBLANK_INTERRUPT_STATUS (1UL<<2) /* 965 or later */ +#define I915_VBLANK_INTERRUPT_STATUS (1UL<<1) +#define I915_OVERLAY_UPDATED_STATUS (1UL<<0) + +#define SRX_INDEX 0x3c4 +#define SRX_DATA 0x3c5 +#define SR01 1 +#define SR01_SCREEN_OFF (1<<5) + +#define PPCR 0x61204 +#define PPCR_ON (1<<0) + +#define DVOB 0x61140 +#define DVOB_ON (1<<31) +#define DVOC 0x61160 +#define DVOC_ON (1<<31) +#define LVDS 0x61180 +#define LVDS_ON (1<<31) + +#define ADPA 0x61100 +#define ADPA_DPMS_MASK (~(3<<10)) +#define ADPA_DPMS_ON (0<<10) +#define ADPA_DPMS_SUSPEND (1<<10) +#define ADPA_DPMS_STANDBY (2<<10) +#define ADPA_DPMS_OFF (3<<10) + +#define RING_TAIL 0x00 +#define TAIL_ADDR 0x001FFFF8 +#define RING_HEAD 0x04 +#define HEAD_WRAP_COUNT 0xFFE00000 +#define HEAD_WRAP_ONE 0x00200000 +#define HEAD_ADDR 0x001FFFFC +#define RING_START 0x08 +#define START_ADDR 0xFFFFF000 +#define RING_LEN 0x0C +#define RING_NR_PAGES 0x001FF000 +#define RING_REPORT_MASK 0x00000006 +#define RING_REPORT_64K 0x00000002 +#define RING_REPORT_128K 0x00000004 +#define RING_NO_REPORT 0x00000000 +#define RING_VALID_MASK 0x00000001 +#define RING_VALID 0x00000001 +#define RING_INVALID 0x00000000 + +/* Scratch pad debug 0 reg: + */ +#define DPLL_FPA01_P1_POST_DIV_MASK_I830 0x001f0000 +/* + * The i830 generation, in LVDS mode, defines P1 as the bit number set within + * this field (only one bit may be set). + */ +#define DPLL_FPA01_P1_POST_DIV_MASK_I830_LVDS 0x003f0000 +#define DPLL_FPA01_P1_POST_DIV_SHIFT 16 +/* i830, required in DVO non-gang */ +#define PLL_P2_DIVIDE_BY_4 (1 << 23) +#define PLL_P1_DIVIDE_BY_TWO (1 << 21) /* i830 */ +#define PLL_REF_INPUT_DREFCLK (0 << 13) +#define PLL_REF_INPUT_TVCLKINA (1 << 13) /* i830 */ +#define PLL_REF_INPUT_TVCLKINBC (2 << 13) /* SDVO TVCLKIN */ +#define PLLB_REF_INPUT_SPREADSPECTRUMIN (3 << 13) +#define PLL_REF_INPUT_MASK (3 << 13) +#define PLL_LOAD_PULSE_PHASE_SHIFT 9 +/* + * Parallel to Serial Load Pulse phase selection. + * Selects the phase for the 10X DPLL clock for the PCIe + * digital display port. The range is 4 to 13; 10 or more + * is just a flip delay. The default is 6 + */ +#define PLL_LOAD_PULSE_PHASE_MASK (0xf << PLL_LOAD_PULSE_PHASE_SHIFT) +#define DISPLAY_RATE_SELECT_FPA1 (1 << 8) +/* + * SDVO multiplier for 945G/GM. Not used on 965. + */ +#define SDVO_MULTIPLIER_MASK 0x000000ff +#define SDVO_MULTIPLIER_SHIFT_HIRES 4 +#define SDVO_MULTIPLIER_SHIFT_VGA 0 +#define DPLL_A_MD 0x0601c /* 965+ only */ +/* + * UDI pixel divider, controlling how many pixels are stuffed into a packet. + * + * Value is pixels minus 1. Must be set to 1 pixel for SDVO. + */ +#define DPLL_MD_UDI_DIVIDER_MASK 0x3f000000 +#define DPLL_MD_UDI_DIVIDER_SHIFT 24 +/* UDI pixel divider for VGA, same as DPLL_MD_UDI_DIVIDER_MASK. */ +#define DPLL_MD_VGA_UDI_DIVIDER_MASK 0x003f0000 +#define DPLL_MD_VGA_UDI_DIVIDER_SHIFT 16 +/* + * SDVO/UDI pixel multiplier. + * + * SDVO requires that the bus clock rate be between 1 and 2 Ghz, and the bus + * clock rate is 10 times the DPLL clock. At low resolution/refresh rate + * modes, the bus rate would be below the limits, so SDVO allows for stuffing + * dummy bytes in the datastream at an increased clock rate, with both sides of + * the link knowing how many bytes are fill. + * + * So, for a mode with a dotclock of 65Mhz, we would want to double the clock + * rate to 130Mhz to get a bus rate of 1.30Ghz. The DPLL clock rate would be + * set to 130Mhz, and the SDVO multiplier set to 2x in this register and + * through an SDVO command. + * + * This register field has values of multiplication factor minus 1, with + * a maximum multiplier of 5 for SDVO. + */ +#define DPLL_MD_UDI_MULTIPLIER_MASK 0x00003f00 +#define DPLL_MD_UDI_MULTIPLIER_SHIFT 8 +/* + * SDVO/UDI pixel multiplier for VGA, same as DPLL_MD_UDI_MULTIPLIER_MASK. + * This best be set to the default value (3) or the CRT won't work. No, + * I don't entirely understand what this does... + */ +#define DPLL_MD_VGA_UDI_MULTIPLIER_MASK 0x0000003f +#define DPLL_MD_VGA_UDI_MULTIPLIER_SHIFT 0 +#define DPLL_B_MD 0x06020 /* 965+ only */ +#define FPA0 0x06040 +#define FPA1 0x06044 +#define FPB0 0x06048 +#define FPB1 0x0604c +#define FP_N_DIV_MASK 0x003f0000 +#define FP_N_DIV_SHIFT 16 +#define FP_M1_DIV_MASK 0x00003f00 +#define FP_M1_DIV_SHIFT 8 +#define FP_M2_DIV_MASK 0x0000003f +#define FP_M2_DIV_SHIFT 0 +#define DPLL_TEST 0x606c +#define DPLLB_TEST_SDVO_DIV_1 (0 << 22) +#define DPLLB_TEST_SDVO_DIV_2 (1 << 22) +#define DPLLB_TEST_SDVO_DIV_4 (2 << 22) +#define DPLLB_TEST_SDVO_DIV_MASK (3 << 22) +#define DPLLB_TEST_N_BYPASS (1 << 19) +#define DPLLB_TEST_M_BYPASS (1 << 18) +#define DPLLB_INPUT_BUFFER_ENABLE (1 << 16) +#define DPLLA_TEST_N_BYPASS (1 << 3) +#define DPLLA_TEST_M_BYPASS (1 << 2) +#define DPLLA_INPUT_BUFFER_ENABLE (1 << 0) +#define D_STATE 0x6104 +#define CG_2D_DIS 0x6200 +#define CG_3D_DIS 0x6204 + +/* + * Palette regs + */ + +#define PALETTE_A 0x0a000 +#define PALETTE_B 0x0a800 + +/* + * Overlay regs + */ + +#define OVADD 0x30000 +#define DOVSTA 0x30008 +#define OC_BUF (0x3<<20) +#define OGAMC5 0x30010 +#define OGAMC4 0x30014 +#define OGAMC3 0x30018 +#define OGAMC2 0x3001c +#define OGAMC1 0x30020 +#define OGAMC0 0x30024 + +/* + * Display engine regs + */ + +/* Pipe A timing regs */ +#define HTOTAL_A 0x60000 +#define HBLANK_A 0x60004 +#define HSYNC_A 0x60008 +#define VTOTAL_A 0x6000c +#define VBLANK_A 0x60010 +#define VSYNC_A 0x60014 +#define PIPEASRC 0x6001c +#define BCLRPAT_A 0x60020 + +/* Pipe B timing regs */ +#define HTOTAL_B 0x61000 +#define HBLANK_B 0x61004 +#define HSYNC_B 0x61008 +#define VTOTAL_B 0x6100c +#define VBLANK_B 0x61010 +#define VSYNC_B 0x61014 +#define PIPEBSRC 0x6101c +#define BCLRPAT_B 0x61020 + +/* VGA port control */ +#define ADPA 0x61100 +#define ADPA_DAC_ENABLE (1<<31) +#define ADPA_DAC_DISABLE 0 +#define ADPA_PIPE_SELECT_MASK (1<<30) +#define ADPA_PIPE_A_SELECT 0 +#define ADPA_PIPE_B_SELECT (1<<30) +#define ADPA_USE_VGA_HVPOLARITY (1<<15) +#define ADPA_SETS_HVPOLARITY 0 +#define ADPA_VSYNC_CNTL_DISABLE (1<<11) +#define ADPA_VSYNC_CNTL_ENABLE 0 +#define ADPA_HSYNC_CNTL_DISABLE (1<<10) +#define ADPA_HSYNC_CNTL_ENABLE 0 +#define ADPA_VSYNC_ACTIVE_HIGH (1<<4) +#define ADPA_VSYNC_ACTIVE_LOW 0 +#define ADPA_HSYNC_ACTIVE_HIGH (1<<3) +#define ADPA_HSYNC_ACTIVE_LOW 0 +#define ADPA_DPMS_MASK (~(3<<10)) +#define ADPA_DPMS_ON (0<<10) +#define ADPA_DPMS_SUSPEND (1<<10) +#define ADPA_DPMS_STANDBY (2<<10) +#define ADPA_DPMS_OFF (3<<10) + +/* Hotplug control (945+ only) */ +#define PORT_HOTPLUG_EN 0x61110 +#define SDVOB_HOTPLUG_INT_EN (1 << 26) +#define SDVOC_HOTPLUG_INT_EN (1 << 25) +#define TV_HOTPLUG_INT_EN (1 << 18) +#define CRT_HOTPLUG_INT_EN (1 << 9) +#define CRT_HOTPLUG_FORCE_DETECT (1 << 3) + +#define PORT_HOTPLUG_STAT 0x61114 +#define CRT_HOTPLUG_INT_STATUS (1 << 11) +#define TV_HOTPLUG_INT_STATUS (1 << 10) +#define CRT_HOTPLUG_MONITOR_MASK (3 << 8) +#define CRT_HOTPLUG_MONITOR_COLOR (3 << 8) +#define CRT_HOTPLUG_MONITOR_MONO (2 << 8) +#define CRT_HOTPLUG_MONITOR_NONE (0 << 8) +#define SDVOC_HOTPLUG_INT_STATUS (1 << 7) +#define SDVOB_HOTPLUG_INT_STATUS (1 << 6) + +/* SDVO port control */ +#define SDVOB 0x61140 +#define SDVOC 0x61160 +#define SDVO_ENABLE (1 << 31) +#define SDVO_PIPE_B_SELECT (1 << 30) +#define SDVO_STALL_SELECT (1 << 29) +#define SDVO_INTERRUPT_ENABLE (1 << 26) +/** + * 915G/GM SDVO pixel multiplier. + * + * Programmed value is multiplier - 1, up to 5x. + * + * \sa DPLL_MD_UDI_MULTIPLIER_MASK + */ +#define SDVO_PORT_MULTIPLY_MASK (7 << 23) +#define SDVO_PORT_MULTIPLY_SHIFT 23 +#define SDVO_PHASE_SELECT_MASK (15 << 19) +#define SDVO_PHASE_SELECT_DEFAULT (6 << 19) +#define SDVO_CLOCK_OUTPUT_INVERT (1 << 18) +#define SDVOC_GANG_MODE (1 << 16) +#define SDVO_BORDER_ENABLE (1 << 7) +#define SDVOB_PCIE_CONCURRENCY (1 << 3) +#define SDVO_DETECTED (1 << 2) +/* Bits to be preserved when writing */ +#define SDVOB_PRESERVE_MASK ((1 << 17) | (1 << 16) | (1 << 14) | (1 << 26)) +#define SDVOC_PRESERVE_MASK ((1 << 17) | (1 << 26)) + +/* DVO port control */ +#define DVOA 0x61120 +#define DVOB 0x61140 +#define DVOC 0x61160 +#define DVO_ENABLE (1 << 31) +#define DVO_PIPE_B_SELECT (1 << 30) +#define DVO_PIPE_STALL_UNUSED (0 << 28) +#define DVO_PIPE_STALL (1 << 28) +#define DVO_PIPE_STALL_TV (2 << 28) +#define DVO_PIPE_STALL_MASK (3 << 28) +#define DVO_USE_VGA_SYNC (1 << 15) +#define DVO_DATA_ORDER_I740 (0 << 14) +#define DVO_DATA_ORDER_FP (1 << 14) +#define DVO_VSYNC_DISABLE (1 << 11) +#define DVO_HSYNC_DISABLE (1 << 10) +#define DVO_VSYNC_TRISTATE (1 << 9) +#define DVO_HSYNC_TRISTATE (1 << 8) +#define DVO_BORDER_ENABLE (1 << 7) +#define DVO_DATA_ORDER_GBRG (1 << 6) +#define DVO_DATA_ORDER_RGGB (0 << 6) +#define DVO_DATA_ORDER_GBRG_ERRATA (0 << 6) +#define DVO_DATA_ORDER_RGGB_ERRATA (1 << 6) +#define DVO_VSYNC_ACTIVE_HIGH (1 << 4) +#define DVO_HSYNC_ACTIVE_HIGH (1 << 3) +#define DVO_BLANK_ACTIVE_HIGH (1 << 2) +#define DVO_OUTPUT_CSTATE_PIXELS (1 << 1) /* SDG only */ +#define DVO_OUTPUT_SOURCE_SIZE_PIXELS (1 << 0) /* SDG only */ +#define DVO_PRESERVE_MASK (0x7<<24) +#define DVOA_SRCDIM 0x61124 +#define DVOB_SRCDIM 0x61144 +#define DVOC_SRCDIM 0x61164 +#define DVO_SRCDIM_HORIZONTAL_SHIFT 12 +#define DVO_SRCDIM_VERTICAL_SHIFT 0 + +/* LVDS port control */ +#define LVDS 0x61180 +/* + * Enables the LVDS port. This bit must be set before DPLLs are enabled, as + * the DPLL semantics change when the LVDS is assigned to that pipe. + */ +#define LVDS_PORT_EN (1 << 31) +/* Selects pipe B for LVDS data. Must be set on pre-965. */ +#define LVDS_PIPEB_SELECT (1 << 30) +/* + * Enables the A0-A2 data pairs and CLKA, containing 18 bits of color data per + * pixel. + */ +#define LVDS_A0A2_CLKA_POWER_MASK (3 << 8) +#define LVDS_A0A2_CLKA_POWER_DOWN (0 << 8) +#define LVDS_A0A2_CLKA_POWER_UP (3 << 8) +/* + * Controls the A3 data pair, which contains the additional LSBs for 24 bit + * mode. Only enabled if LVDS_A0A2_CLKA_POWER_UP also indicates it should be + * on. + */ +#define LVDS_A3_POWER_MASK (3 << 6) +#define LVDS_A3_POWER_DOWN (0 << 6) +#define LVDS_A3_POWER_UP (3 << 6) +/* + * Controls the CLKB pair. This should only be set when LVDS_B0B3_POWER_UP + * is set. + */ +#define LVDS_CLKB_POWER_MASK (3 << 4) +#define LVDS_CLKB_POWER_DOWN (0 << 4) +#define LVDS_CLKB_POWER_UP (3 << 4) +/* + * Controls the B0-B3 data pairs. This must be set to match the DPLL p2 + * setting for whether we are in dual-channel mode. The B3 pair will + * additionally only be powered up when LVDS_A3_POWER_UP is set. + */ +#define LVDS_B0B3_POWER_MASK (3 << 2) +#define LVDS_B0B3_POWER_DOWN (0 << 2) +#define LVDS_B0B3_POWER_UP (3 << 2) + +/* Panel power sequencing */ +#define PP_STATUS 0x61200 +#define PP_ON (1 << 31) +/* + * Indicates that all dependencies of the panel are on: + * + * - PLL enabled + * - pipe enabled + * - LVDS/DVOB/DVOC on + */ +#define PP_READY (1 << 30) +#define PP_SEQUENCE_NONE (0 << 28) +#define PP_SEQUENCE_ON (1 << 28) +#define PP_SEQUENCE_OFF (2 << 28) +#define PP_SEQUENCE_MASK 0x30000000 +#define PP_CONTROL 0x61204 +#define POWER_TARGET_ON (1 << 0) +#define PP_ON_DELAYS 0x61208 +#define PP_OFF_DELAYS 0x6120c +#define PP_DIVISOR 0x61210 + +/* Panel fitting */ +#define PFIT_CONTROL 0x61230 +#define PFIT_ENABLE (1 << 31) +#define PFIT_PIPE_MASK (3 << 29) +#define PFIT_PIPE_SHIFT 29 +#define VERT_INTERP_DISABLE (0 << 10) +#define VERT_INTERP_BILINEAR (1 << 10) +#define VERT_INTERP_MASK (3 << 10) +#define VERT_AUTO_SCALE (1 << 9) +#define HORIZ_INTERP_DISABLE (0 << 6) +#define HORIZ_INTERP_BILINEAR (1 << 6) +#define HORIZ_INTERP_MASK (3 << 6) +#define HORIZ_AUTO_SCALE (1 << 5) +#define PANEL_8TO6_DITHER_ENABLE (1 << 3) +#define PFIT_PGM_RATIOS 0x61234 +#define PFIT_VERT_SCALE_MASK 0xfff00000 +#define PFIT_HORIZ_SCALE_MASK 0x0000fff0 +#define PFIT_AUTO_RATIOS 0x61238 + +/* Backlight control */ +#define BLC_PWM_CTL 0x61254 +#define BACKLIGHT_MODULATION_FREQ_SHIFT (17) +#define BLC_PWM_CTL2 0x61250 /* 965+ only */ +/* + * This is the most significant 15 bits of the number of backlight cycles in a + * complete cycle of the modulated backlight control. + * + * The actual value is this field multiplied by two. + */ +#define BACKLIGHT_MODULATION_FREQ_MASK (0x7fff << 17) +#define BLM_LEGACY_MODE (1 << 16) +/* + * This is the number of cycles out of the backlight modulation cycle for which + * the backlight is on. + * + * This field must be no greater than the number of cycles in the complete + * backlight modulation cycle. + */ +#define BACKLIGHT_DUTY_CYCLE_SHIFT (0) +#define BACKLIGHT_DUTY_CYCLE_MASK (0xffff) + +/* TV port control */ +#define TV_CTL 0x68000 +/** Enables the TV encoder */ +# define TV_ENC_ENABLE (1 << 31) +/** Sources the TV encoder input from pipe B instead of A. */ +# define TV_ENC_PIPEB_SELECT (1 << 30) +/** Outputs composite video (DAC A only) */ +# define TV_ENC_OUTPUT_COMPOSITE (0 << 28) +/** Outputs SVideo video (DAC B/C) */ +# define TV_ENC_OUTPUT_SVIDEO (1 << 28) +/** Outputs Component video (DAC A/B/C) */ +# define TV_ENC_OUTPUT_COMPONENT (2 << 28) +/** Outputs Composite and SVideo (DAC A/B/C) */ +# define TV_ENC_OUTPUT_SVIDEO_COMPOSITE (3 << 28) +# define TV_TRILEVEL_SYNC (1 << 21) +/** Enables slow sync generation (945GM only) */ +# define TV_SLOW_SYNC (1 << 20) +/** Selects 4x oversampling for 480i and 576p */ +# define TV_OVERSAMPLE_4X (0 << 18) +/** Selects 2x oversampling for 720p and 1080i */ +# define TV_OVERSAMPLE_2X (1 << 18) +/** Selects no oversampling for 1080p */ +# define TV_OVERSAMPLE_NONE (2 << 18) +/** Selects 8x oversampling */ +# define TV_OVERSAMPLE_8X (3 << 18) +/** Selects progressive mode rather than interlaced */ +# define TV_PROGRESSIVE (1 << 17) +/** Sets the colorburst to PAL mode. Required for non-M PAL modes. */ +# define TV_PAL_BURST (1 << 16) +/** Field for setting delay of Y compared to C */ +# define TV_YC_SKEW_MASK (7 << 12) +/** Enables a fix for 480p/576p standard definition modes on the 915GM only */ +# define TV_ENC_SDP_FIX (1 << 11) +/** + * Enables a fix for the 915GM only. + * + * Not sure what it does. + */ +# define TV_ENC_C0_FIX (1 << 10) +/** Bits that must be preserved by software */ +# define TV_CTL_SAVE ((3 << 8) | (3 << 6)) +# define TV_FUSE_STATE_MASK (3 << 4) +/** Read-only state that reports all features enabled */ +# define TV_FUSE_STATE_ENABLED (0 << 4) +/** Read-only state that reports that Macrovision is disabled in hardware*/ +# define TV_FUSE_STATE_NO_MACROVISION (1 << 4) +/** Read-only state that reports that TV-out is disabled in hardware. */ +# define TV_FUSE_STATE_DISABLED (2 << 4) +/** Normal operation */ +# define TV_TEST_MODE_NORMAL (0 << 0) +/** Encoder test pattern 1 - combo pattern */ +# define TV_TEST_MODE_PATTERN_1 (1 << 0) +/** Encoder test pattern 2 - full screen vertical 75% color bars */ +# define TV_TEST_MODE_PATTERN_2 (2 << 0) +/** Encoder test pattern 3 - full screen horizontal 75% color bars */ +# define TV_TEST_MODE_PATTERN_3 (3 << 0) +/** Encoder test pattern 4 - random noise */ +# define TV_TEST_MODE_PATTERN_4 (4 << 0) +/** Encoder test pattern 5 - linear color ramps */ +# define TV_TEST_MODE_PATTERN_5 (5 << 0) +/** + * This test mode forces the DACs to 50% of full output. + * + * This is used for load detection in combination with TVDAC_SENSE_MASK + */ +# define TV_TEST_MODE_MONITOR_DETECT (7 << 0) +# define TV_TEST_MODE_MASK (7 << 0) + +#define TV_DAC 0x68004 +/** + * Reports that DAC state change logic has reported change (RO). + * + * This gets cleared when TV_DAC_STATE_EN is cleared +*/ +# define TVDAC_STATE_CHG (1 << 31) +# define TVDAC_SENSE_MASK (7 << 28) +/** Reports that DAC A voltage is above the detect threshold */ +# define TVDAC_A_SENSE (1 << 30) +/** Reports that DAC B voltage is above the detect threshold */ +# define TVDAC_B_SENSE (1 << 29) +/** Reports that DAC C voltage is above the detect threshold */ +# define TVDAC_C_SENSE (1 << 28) +/** + * Enables DAC state detection logic, for load-based TV detection. + * + * The PLL of the chosen pipe (in TV_CTL) must be running, and the encoder set + * to off, for load detection to work. + */ +# define TVDAC_STATE_CHG_EN (1 << 27) +/** Sets the DAC A sense value to high */ +# define TVDAC_A_SENSE_CTL (1 << 26) +/** Sets the DAC B sense value to high */ +# define TVDAC_B_SENSE_CTL (1 << 25) +/** Sets the DAC C sense value to high */ +# define TVDAC_C_SENSE_CTL (1 << 24) +/** Overrides the ENC_ENABLE and DAC voltage levels */ +# define DAC_CTL_OVERRIDE (1 << 7) +/** Sets the slew rate. Must be preserved in software */ +# define ENC_TVDAC_SLEW_FAST (1 << 6) +# define DAC_A_1_3_V (0 << 4) +# define DAC_A_1_1_V (1 << 4) +# define DAC_A_0_7_V (2 << 4) +# define DAC_A_OFF (3 << 4) +# define DAC_B_1_3_V (0 << 2) +# define DAC_B_1_1_V (1 << 2) +# define DAC_B_0_7_V (2 << 2) +# define DAC_B_OFF (3 << 2) +# define DAC_C_1_3_V (0 << 0) +# define DAC_C_1_1_V (1 << 0) +# define DAC_C_0_7_V (2 << 0) +# define DAC_C_OFF (3 << 0) + +/** + * CSC coefficients are stored in a floating point format with 9 bits of + * mantissa and 2 or 3 bits of exponent. The exponent is represented as 2**-n, + * where 2-bit exponents are unsigned n, and 3-bit exponents are signed n with + * -1 (0x3) being the only legal negative value. + */ +#define TV_CSC_Y 0x68010 +# define TV_RY_MASK 0x07ff0000 +# define TV_RY_SHIFT 16 +# define TV_GY_MASK 0x00000fff +# define TV_GY_SHIFT 0 + +#define TV_CSC_Y2 0x68014 +# define TV_BY_MASK 0x07ff0000 +# define TV_BY_SHIFT 16 +/** + * Y attenuation for component video. + * + * Stored in 1.9 fixed point. + */ +# define TV_AY_MASK 0x000003ff +# define TV_AY_SHIFT 0 + +#define TV_CSC_U 0x68018 +# define TV_RU_MASK 0x07ff0000 +# define TV_RU_SHIFT 16 +# define TV_GU_MASK 0x000007ff +# define TV_GU_SHIFT 0 + +#define TV_CSC_U2 0x6801c +# define TV_BU_MASK 0x07ff0000 +# define TV_BU_SHIFT 16 +/** + * U attenuation for component video. + * + * Stored in 1.9 fixed point. + */ +# define TV_AU_MASK 0x000003ff +# define TV_AU_SHIFT 0 + +#define TV_CSC_V 0x68020 +# define TV_RV_MASK 0x0fff0000 +# define TV_RV_SHIFT 16 +# define TV_GV_MASK 0x000007ff +# define TV_GV_SHIFT 0 + +#define TV_CSC_V2 0x68024 +# define TV_BV_MASK 0x07ff0000 +# define TV_BV_SHIFT 16 +/** + * V attenuation for component video. + * + * Stored in 1.9 fixed point. + */ +# define TV_AV_MASK 0x000007ff +# define TV_AV_SHIFT 0 + +#define TV_CLR_KNOBS 0x68028 +/** 2s-complement brightness adjustment */ +# define TV_BRIGHTNESS_MASK 0xff000000 +# define TV_BRIGHTNESS_SHIFT 24 +/** Contrast adjustment, as a 2.6 unsigned floating point number */ +# define TV_CONTRAST_MASK 0x00ff0000 +# define TV_CONTRAST_SHIFT 16 +/** Saturation adjustment, as a 2.6 unsigned floating point number */ +# define TV_SATURATION_MASK 0x0000ff00 +# define TV_SATURATION_SHIFT 8 +/** Hue adjustment, as an integer phase angle in degrees */ +# define TV_HUE_MASK 0x000000ff +# define TV_HUE_SHIFT 0 + +#define TV_CLR_LEVEL 0x6802c +/** Controls the DAC level for black */ +# define TV_BLACK_LEVEL_MASK 0x01ff0000 +# define TV_BLACK_LEVEL_SHIFT 16 +/** Controls the DAC level for blanking */ +# define TV_BLANK_LEVEL_MASK 0x000001ff +# define TV_BLANK_LEVEL_SHIFT 0 + +#define TV_H_CTL_1 0x68030 +/** Number of pixels in the hsync. */ +# define TV_HSYNC_END_MASK 0x1fff0000 +# define TV_HSYNC_END_SHIFT 16 +/** Total number of pixels minus one in the line (display and blanking). */ +# define TV_HTOTAL_MASK 0x00001fff +# define TV_HTOTAL_SHIFT 0 + +#define TV_H_CTL_2 0x68034 +/** Enables the colorburst (needed for non-component color) */ +# define TV_BURST_ENA (1 << 31) +/** Offset of the colorburst from the start of hsync, in pixels minus one. */ +# define TV_HBURST_START_SHIFT 16 +# define TV_HBURST_START_MASK 0x1fff0000 +/** Length of the colorburst */ +# define TV_HBURST_LEN_SHIFT 0 +# define TV_HBURST_LEN_MASK 0x0001fff + +#define TV_H_CTL_3 0x68038 +/** End of hblank, measured in pixels minus one from start of hsync */ +# define TV_HBLANK_END_SHIFT 16 +# define TV_HBLANK_END_MASK 0x1fff0000 +/** Start of hblank, measured in pixels minus one from start of hsync */ +# define TV_HBLANK_START_SHIFT 0 +# define TV_HBLANK_START_MASK 0x0001fff + +#define TV_V_CTL_1 0x6803c +/** XXX */ +# define TV_NBR_END_SHIFT 16 +# define TV_NBR_END_MASK 0x07ff0000 +/** XXX */ +# define TV_VI_END_F1_SHIFT 8 +# define TV_VI_END_F1_MASK 0x00003f00 +/** XXX */ +# define TV_VI_END_F2_SHIFT 0 +# define TV_VI_END_F2_MASK 0x0000003f + +#define TV_V_CTL_2 0x68040 +/** Length of vsync, in half lines */ +# define TV_VSYNC_LEN_MASK 0x07ff0000 +# define TV_VSYNC_LEN_SHIFT 16 +/** Offset of the start of vsync in field 1, measured in one less than the + * number of half lines. + */ +# define TV_VSYNC_START_F1_MASK 0x00007f00 +# define TV_VSYNC_START_F1_SHIFT 8 +/** + * Offset of the start of vsync in field 2, measured in one less than the + * number of half lines. + */ +# define TV_VSYNC_START_F2_MASK 0x0000007f +# define TV_VSYNC_START_F2_SHIFT 0 + +#define TV_V_CTL_3 0x68044 +/** Enables generation of the equalization signal */ +# define TV_EQUAL_ENA (1 << 31) +/** Length of vsync, in half lines */ +# define TV_VEQ_LEN_MASK 0x007f0000 +# define TV_VEQ_LEN_SHIFT 16 +/** Offset of the start of equalization in field 1, measured in one less than + * the number of half lines. + */ +# define TV_VEQ_START_F1_MASK 0x0007f00 +# define TV_VEQ_START_F1_SHIFT 8 +/** + * Offset of the start of equalization in field 2, measured in one less than + * the number of half lines. + */ +# define TV_VEQ_START_F2_MASK 0x000007f +# define TV_VEQ_START_F2_SHIFT 0 + +#define TV_V_CTL_4 0x68048 +/** + * Offset to start of vertical colorburst, measured in one less than the + * number of lines from vertical start. + */ +# define TV_VBURST_START_F1_MASK 0x003f0000 +# define TV_VBURST_START_F1_SHIFT 16 +/** + * Offset to the end of vertical colorburst, measured in one less than the + * number of lines from the start of NBR. + */ +# define TV_VBURST_END_F1_MASK 0x000000ff +# define TV_VBURST_END_F1_SHIFT 0 + +#define TV_V_CTL_5 0x6804c +/** + * Offset to start of vertical colorburst, measured in one less than the + * number of lines from vertical start. + */ +# define TV_VBURST_START_F2_MASK 0x003f0000 +# define TV_VBURST_START_F2_SHIFT 16 +/** + * Offset to the end of vertical colorburst, measured in one less than the + * number of lines from the start of NBR. + */ +# define TV_VBURST_END_F2_MASK 0x000000ff +# define TV_VBURST_END_F2_SHIFT 0 + +#define TV_V_CTL_6 0x68050 +/** + * Offset to start of vertical colorburst, measured in one less than the + * number of lines from vertical start. + */ +# define TV_VBURST_START_F3_MASK 0x003f0000 +# define TV_VBURST_START_F3_SHIFT 16 +/** + * Offset to the end of vertical colorburst, measured in one less than the + * number of lines from the start of NBR. + */ +# define TV_VBURST_END_F3_MASK 0x000000ff +# define TV_VBURST_END_F3_SHIFT 0 + +#define TV_V_CTL_7 0x68054 +/** + * Offset to start of vertical colorburst, measured in one less than the + * number of lines from vertical start. + */ +# define TV_VBURST_START_F4_MASK 0x003f0000 +# define TV_VBURST_START_F4_SHIFT 16 +/** + * Offset to the end of vertical colorburst, measured in one less than the + * number of lines from the start of NBR. + */ +# define TV_VBURST_END_F4_MASK 0x000000ff +# define TV_VBURST_END_F4_SHIFT 0 + +#define TV_SC_CTL_1 0x68060 +/** Turns on the first subcarrier phase generation DDA */ +# define TV_SC_DDA1_EN (1 << 31) +/** Turns on the first subcarrier phase generation DDA */ +# define TV_SC_DDA2_EN (1 << 30) +/** Turns on the first subcarrier phase generation DDA */ +# define TV_SC_DDA3_EN (1 << 29) +/** Sets the subcarrier DDA to reset frequency every other field */ +# define TV_SC_RESET_EVERY_2 (0 << 24) +/** Sets the subcarrier DDA to reset frequency every fourth field */ +# define TV_SC_RESET_EVERY_4 (1 << 24) +/** Sets the subcarrier DDA to reset frequency every eighth field */ +# define TV_SC_RESET_EVERY_8 (2 << 24) +/** Sets the subcarrier DDA to never reset the frequency */ +# define TV_SC_RESET_NEVER (3 << 24) +/** Sets the peak amplitude of the colorburst.*/ +# define TV_BURST_LEVEL_MASK 0x00ff0000 +# define TV_BURST_LEVEL_SHIFT 16 +/** Sets the increment of the first subcarrier phase generation DDA */ +# define TV_SCDDA1_INC_MASK 0x00000fff +# define TV_SCDDA1_INC_SHIFT 0 + +#define TV_SC_CTL_2 0x68064 +/** Sets the rollover for the second subcarrier phase generation DDA */ +# define TV_SCDDA2_SIZE_MASK 0x7fff0000 +# define TV_SCDDA2_SIZE_SHIFT 16 +/** Sets the increent of the second subcarrier phase generation DDA */ +# define TV_SCDDA2_INC_MASK 0x00007fff +# define TV_SCDDA2_INC_SHIFT 0 + +#define TV_SC_CTL_3 0x68068 +/** Sets the rollover for the third subcarrier phase generation DDA */ +# define TV_SCDDA3_SIZE_MASK 0x7fff0000 +# define TV_SCDDA3_SIZE_SHIFT 16 +/** Sets the increent of the third subcarrier phase generation DDA */ +# define TV_SCDDA3_INC_MASK 0x00007fff +# define TV_SCDDA3_INC_SHIFT 0 + +#define TV_WIN_POS 0x68070 +/** X coordinate of the display from the start of horizontal active */ +# define TV_XPOS_MASK 0x1fff0000 +# define TV_XPOS_SHIFT 16 +/** Y coordinate of the display from the start of vertical active (NBR) */ +# define TV_YPOS_MASK 0x00000fff +# define TV_YPOS_SHIFT 0 + +#define TV_WIN_SIZE 0x68074 +/** Horizontal size of the display window, measured in pixels*/ +# define TV_XSIZE_MASK 0x1fff0000 +# define TV_XSIZE_SHIFT 16 +/** + * Vertical size of the display window, measured in pixels. + * + * Must be even for interlaced modes. + */ +# define TV_YSIZE_MASK 0x00000fff +# define TV_YSIZE_SHIFT 0 + +#define TV_FILTER_CTL_1 0x68080 +/** + * Enables automatic scaling calculation. + * + * If set, the rest of the registers are ignored, and the calculated values can + * be read back from the register. + */ +# define TV_AUTO_SCALE (1 << 31) +/** + * Disables the vertical filter. + * + * This is required on modes more than 1024 pixels wide */ +# define TV_V_FILTER_BYPASS (1 << 29) +/** Enables adaptive vertical filtering */ +# define TV_VADAPT (1 << 28) +# define TV_VADAPT_MODE_MASK (3 << 26) +/** Selects the least adaptive vertical filtering mode */ +# define TV_VADAPT_MODE_LEAST (0 << 26) +/** Selects the moderately adaptive vertical filtering mode */ +# define TV_VADAPT_MODE_MODERATE (1 << 26) +/** Selects the most adaptive vertical filtering mode */ +# define TV_VADAPT_MODE_MOST (3 << 26) +/** + * Sets the horizontal scaling factor. + * + * This should be the fractional part of the horizontal scaling factor divided + * by the oversampling rate. TV_HSCALE should be less than 1, and set to: + * + * (src width - 1) / ((oversample * dest width) - 1) + */ +# define TV_HSCALE_FRAC_MASK 0x00003fff +# define TV_HSCALE_FRAC_SHIFT 0 + +#define TV_FILTER_CTL_2 0x68084 +/** + * Sets the integer part of the 3.15 fixed-point vertical scaling factor. + * + * TV_VSCALE should be (src height - 1) / ((interlace * dest height) - 1) + */ +# define TV_VSCALE_INT_MASK 0x00038000 +# define TV_VSCALE_INT_SHIFT 15 +/** + * Sets the fractional part of the 3.15 fixed-point vertical scaling factor. + * + * \sa TV_VSCALE_INT_MASK + */ +# define TV_VSCALE_FRAC_MASK 0x00007fff +# define TV_VSCALE_FRAC_SHIFT 0 + +#define TV_FILTER_CTL_3 0x68088 +/** + * Sets the integer part of the 3.15 fixed-point vertical scaling factor. + * + * TV_VSCALE should be (src height - 1) / (1/4 * (dest height - 1)) + * + * For progressive modes, TV_VSCALE_IP_INT should be set to zeroes. + */ +# define TV_VSCALE_IP_INT_MASK 0x00038000 +# define TV_VSCALE_IP_INT_SHIFT 15 +/** + * Sets the fractional part of the 3.15 fixed-point vertical scaling factor. + * + * For progressive modes, TV_VSCALE_IP_INT should be set to zeroes. + * + * \sa TV_VSCALE_IP_INT_MASK + */ +# define TV_VSCALE_IP_FRAC_MASK 0x00007fff +# define TV_VSCALE_IP_FRAC_SHIFT 0 + +#define TV_CC_CONTROL 0x68090 +# define TV_CC_ENABLE (1 << 31) +/** + * Specifies which field to send the CC data in. + * + * CC data is usually sent in field 0. + */ +# define TV_CC_FID_MASK (1 << 27) +# define TV_CC_FID_SHIFT 27 +/** Sets the horizontal position of the CC data. Usually 135. */ +# define TV_CC_HOFF_MASK 0x03ff0000 +# define TV_CC_HOFF_SHIFT 16 +/** Sets the vertical position of the CC data. Usually 21 */ +# define TV_CC_LINE_MASK 0x0000003f +# define TV_CC_LINE_SHIFT 0 + +#define TV_CC_DATA 0x68094 +# define TV_CC_RDY (1 << 31) +/** Second word of CC data to be transmitted. */ +# define TV_CC_DATA_2_MASK 0x007f0000 +# define TV_CC_DATA_2_SHIFT 16 +/** First word of CC data to be transmitted. */ +# define TV_CC_DATA_1_MASK 0x0000007f +# define TV_CC_DATA_1_SHIFT 0 + +#define TV_H_LUMA_0 0x68100 +#define TV_H_LUMA_59 0x681ec +#define TV_H_CHROMA_0 0x68200 +#define TV_H_CHROMA_59 0x682ec +#define TV_V_LUMA_0 0x68300 +#define TV_V_LUMA_42 0x683a8 +#define TV_V_CHROMA_0 0x68400 +#define TV_V_CHROMA_42 0x684a8 + +/* Display & cursor control */ + +/* Pipe A */ +#define PIPEADSL 0x70000 +#define PIPEACONF 0x70008 +#define PIPEACONF_ENABLE (1<<31) +#define PIPEACONF_DISABLE 0 +#define PIPEACONF_DOUBLE_WIDE (1<<30) +#define I965_PIPECONF_ACTIVE (1<<30) +#define PIPEACONF_SINGLE_WIDE 0 +#define PIPEACONF_PIPE_UNLOCKED 0 +#define PIPEACONF_PIPE_LOCKED (1<<25) +#define PIPEACONF_PALETTE 0 +#define PIPEACONF_GAMMA (1<<24) +#define PIPECONF_FORCE_BORDER (1<<25) +#define PIPECONF_PROGRESSIVE (0 << 21) +#define PIPECONF_INTERLACE_W_FIELD_INDICATION (6 << 21) +#define PIPECONF_INTERLACE_FIELD_0_ONLY (7 << 21) +#define PIPEASTAT 0x70024 +#define PIPE_FIFO_UNDERRUN_STATUS (1UL<<31) +#define PIPE_CRC_ERROR_ENABLE (1UL<<29) +#define PIPE_CRC_DONE_ENABLE (1UL<<28) +#define PIPE_GMBUS_EVENT_ENABLE (1UL<<27) +#define PIPE_HOTPLUG_INTERRUPT_ENABLE (1UL<<26) +#define PIPE_VSYNC_INTERRUPT_ENABLE (1UL<<25) +#define PIPE_DISPLAY_LINE_COMPARE_ENABLE (1UL<<24) +#define PIPE_DPST_EVENT_ENABLE (1UL<<23) +#define PIPE_LEGACY_BLC_EVENT_ENABLE (1UL<<22) +#define PIPE_ODD_FIELD_INTERRUPT_ENABLE (1UL<<21) +#define PIPE_EVEN_FIELD_INTERRUPT_ENABLE (1UL<<20) +#define PIPE_HOTPLUG_TV_INTERRUPT_ENABLE (1UL<<18) /* pre-965 */ +#define PIPE_START_VBLANK_INTERRUPT_ENABLE (1UL<<18) /* 965 or later */ +#define PIPE_VBLANK_INTERRUPT_ENABLE (1UL<<17) +#define PIPE_OVERLAY_UPDATED_ENABLE (1UL<<16) +#define PIPE_CRC_ERROR_INTERRUPT_STATUS (1UL<<13) +#define PIPE_CRC_DONE_INTERRUPT_STATUS (1UL<<12) +#define PIPE_GMBUS_INTERRUPT_STATUS (1UL<<11) +#define PIPE_HOTPLUG_INTERRUPT_STATUS (1UL<<10) +#define PIPE_VSYNC_INTERRUPT_STATUS (1UL<<9) +#define PIPE_DISPLAY_LINE_COMPARE_STATUS (1UL<<8) +#define PIPE_DPST_EVENT_STATUS (1UL<<7) +#define PIPE_LEGACY_BLC_EVENT_STATUS (1UL<<6) +#define PIPE_ODD_FIELD_INTERRUPT_STATUS (1UL<<5) +#define PIPE_EVEN_FIELD_INTERRUPT_STATUS (1UL<<4) +#define PIPE_HOTPLUG_TV_INTERRUPT_STATUS (1UL<<2) /* pre-965 */ +#define PIPE_START_VBLANK_INTERRUPT_STATUS (1UL<<2) /* 965 or later */ +#define PIPE_VBLANK_INTERRUPT_STATUS (1UL<<1) +#define PIPE_OVERLAY_UPDATED_STATUS (1UL<<0) + +#define DSPARB 0x70030 +#define DSPARB_CSTART_MASK (0x7f << 7) +#define DSPARB_CSTART_SHIFT 7 +#define DSPARB_BSTART_MASK (0x7f) +#define DSPARB_BSTART_SHIFT 0 +/* + * The two pipe frame counter registers are not synchronized, so + * reading a stable value is somewhat tricky. The following code + * should work: + * + * do { + * high1 = ((INREG(PIPEAFRAMEHIGH) & PIPE_FRAME_HIGH_MASK) >> + * PIPE_FRAME_HIGH_SHIFT; + * low1 = ((INREG(PIPEAFRAMEPIXEL) & PIPE_FRAME_LOW_MASK) >> + * PIPE_FRAME_LOW_SHIFT); + * high2 = ((INREG(PIPEAFRAMEHIGH) & PIPE_FRAME_HIGH_MASK) >> + * PIPE_FRAME_HIGH_SHIFT); + * } while (high1 != high2); + * frame = (high1 << 8) | low1; + */ +#define PIPEAFRAMEHIGH 0x70040 +#define PIPE_FRAME_HIGH_MASK 0x0000ffff +#define PIPE_FRAME_HIGH_SHIFT 0 +#define PIPEAFRAMEPIXEL 0x70044 +#define PIPE_FRAME_LOW_MASK 0xff000000 +#define PIPE_FRAME_LOW_SHIFT 24 +#define PIPE_PIXEL_MASK 0x00ffffff +#define PIPE_PIXEL_SHIFT 0 + +/* Cursor A & B regs */ +#define CURACNTR 0x70080 +#define CURSOR_MODE_DISABLE 0x00 +#define CURSOR_MODE_64_32B_AX 0x07 +#define CURSOR_MODE_64_ARGB_AX ((1 << 5) | CURSOR_MODE_64_32B_AX) +#define MCURSOR_GAMMA_ENABLE (1 << 26) +#define CURABASE 0x70084 +#define CURAPOS 0x70088 +#define CURSOR_POS_MASK 0x007FF +#define CURSOR_POS_SIGN 0x8000 +#define CURSOR_X_SHIFT 0 +#define CURSOR_Y_SHIFT 16 +#define CURBCNTR 0x700c0 +#define CURBBASE 0x700c4 +#define CURBPOS 0x700c8 + +/* Display A control */ +#define DSPACNTR 0x70180 +#define DISPLAY_PLANE_ENABLE (1<<31) +#define DISPLAY_PLANE_DISABLE 0 +#define DISPPLANE_GAMMA_ENABLE (1<<30) +#define DISPPLANE_GAMMA_DISABLE 0 +#define DISPPLANE_PIXFORMAT_MASK (0xf<<26) +#define DISPPLANE_8BPP (0x2<<26) +#define DISPPLANE_15_16BPP (0x4<<26) +#define DISPPLANE_16BPP (0x5<<26) +#define DISPPLANE_32BPP_NO_ALPHA (0x6<<26) +#define DISPPLANE_32BPP (0x7<<26) +#define DISPPLANE_STEREO_ENABLE (1<<25) +#define DISPPLANE_STEREO_DISABLE 0 +#define DISPPLANE_SEL_PIPE_MASK (1<<24) +#define DISPPLANE_SEL_PIPE_A 0 +#define DISPPLANE_SEL_PIPE_B (1<<24) +#define DISPPLANE_SRC_KEY_ENABLE (1<<22) +#define DISPPLANE_SRC_KEY_DISABLE 0 +#define DISPPLANE_LINE_DOUBLE (1<<20) +#define DISPPLANE_NO_LINE_DOUBLE 0 +#define DISPPLANE_STEREO_POLARITY_FIRST 0 +#define DISPPLANE_STEREO_POLARITY_SECOND (1<<18) +#define DSPAADDR 0x70184 +#define DSPASTRIDE 0x70188 +#define DSPAPOS 0x7018C /* reserved */ +#define DSPASIZE 0x70190 +#define DSPASURF 0x7019C /* 965+ only */ +#define DSPATILEOFF 0x701A4 /* 965+ only */ + +/* VBIOS flags */ +#define SWF00 0x71410 +#define SWF01 0x71414 +#define SWF02 0x71418 +#define SWF03 0x7141c +#define SWF04 0x71420 +#define SWF05 0x71424 +#define SWF06 0x71428 +#define SWF10 0x70410 +#define SWF11 0x70414 +#define SWF14 0x71420 +#define SWF30 0x72414 +#define SWF31 0x72418 +#define SWF32 0x7241c + +/* Pipe B */ +#define PIPEBDSL 0x71000 +#define PIPEBCONF 0x71008 +#define PIPEBSTAT 0x71024 +#define PIPEBFRAMEHIGH 0x71040 +#define PIPEBFRAMEPIXEL 0x71044 + +/* Display B control */ +#define DSPBCNTR 0x71180 +#define DISPPLANE_ALPHA_TRANS_ENABLE (1<<15) +#define DISPPLANE_ALPHA_TRANS_DISABLE 0 +#define DISPPLANE_SPRITE_ABOVE_DISPLAY 0 +#define DISPPLANE_SPRITE_ABOVE_OVERLAY (1) +#define DSPBADDR 0x71184 +#define DSPBSTRIDE 0x71188 +#define DSPBPOS 0x7118C +#define DSPBSIZE 0x71190 +#define DSPBSURF 0x7119C +#define DSPBTILEOFF 0x711A4 + +/* VBIOS regs */ +#define VGACNTRL 0x71400 +# define VGA_DISP_DISABLE (1 << 31) +# define VGA_2X_MODE (1 << 30) +# define VGA_PIPE_B_SELECT (1 << 29) + +#endif /* _I915_REG_H_ */ -- cgit v1.2.3 From ed4cb4142b242d8090d3811d5eb4abf6aa985bc8 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Tue, 29 Jul 2008 12:10:39 -0700 Subject: i915: Add support for MSI and interrupt mitigation. Previous attempts at interrupt mitigation had been foiled by i915_wait_irq's failure to update the sarea seqno value when the status page indicated that the seqno had already been passed. MSI support has been seen to cut CPU costs by up to 40% in some workloads by avoiding other expensive interrupt handlers for frequent graphics interrupts. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_irq.c | 20 +++--- drivers/gpu/drm/i915/i915_dma.c | 21 +++++- drivers/gpu/drm/i915/i915_drv.h | 6 ++ drivers/gpu/drm/i915/i915_irq.c | 153 ++++++++++++++++++++++++++++------------ 4 files changed, 145 insertions(+), 55 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index 53f0e5af1cc..61ed5158f78 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -63,7 +63,7 @@ int drm_irq_by_busid(struct drm_device *dev, void *data, p->devnum != PCI_SLOT(dev->pdev->devfn) || p->funcnum != PCI_FUNC(dev->pdev->devfn)) return -EINVAL; - p->irq = dev->irq; + p->irq = dev->pdev->irq; DRM_DEBUG("%d:%d:%d => IRQ %d\n", p->busnum, p->devnum, p->funcnum, p->irq); @@ -89,7 +89,7 @@ static int drm_irq_install(struct drm_device * dev) if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ)) return -EINVAL; - if (dev->irq == 0) + if (dev->pdev->irq == 0) return -EINVAL; mutex_lock(&dev->struct_mutex); @@ -107,7 +107,7 @@ static int drm_irq_install(struct drm_device * dev) dev->irq_enabled = 1; mutex_unlock(&dev->struct_mutex); - DRM_DEBUG("irq=%d\n", dev->irq); + DRM_DEBUG("irq=%d\n", dev->pdev->irq); if (drm_core_check_feature(dev, DRIVER_IRQ_VBL)) { init_waitqueue_head(&dev->vbl_queue); @@ -127,8 +127,12 @@ static int drm_irq_install(struct drm_device * dev) if (drm_core_check_feature(dev, DRIVER_IRQ_SHARED)) sh_flags = IRQF_SHARED; - ret = request_irq(dev->irq, dev->driver->irq_handler, + ret = request_irq(dev->pdev->irq, dev->driver->irq_handler, sh_flags, dev->devname, dev); + /* Expose the device irq number to drivers that want to export it for + * whatever reason. + */ + dev->irq = dev->pdev->irq; if (ret < 0) { mutex_lock(&dev->struct_mutex); dev->irq_enabled = 0; @@ -164,11 +168,11 @@ int drm_irq_uninstall(struct drm_device * dev) if (!irq_enabled) return -EINVAL; - DRM_DEBUG("irq=%d\n", dev->irq); + DRM_DEBUG("irq=%d\n", dev->pdev->irq); dev->driver->irq_uninstall(dev); - free_irq(dev->irq, dev); + free_irq(dev->pdev->irq, dev); dev->locked_tasklet_func = NULL; @@ -201,7 +205,7 @@ int drm_control(struct drm_device *dev, void *data, if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ)) return 0; if (dev->if_version < DRM_IF_VERSION(1, 2) && - ctl->irq != dev->irq) + ctl->irq != dev->pdev->irq) return -EINVAL; return drm_irq_install(dev); case DRM_UNINST_HANDLER: @@ -239,7 +243,7 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr int ret = 0; unsigned int flags, seq; - if ((!dev->irq) || (!dev->irq_enabled)) + if ((!dev->pdev->irq) || (!dev->irq_enabled)) return -EINVAL; if (vblwait->request.type & diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 4d56dfd25e0..27a1f78a7f1 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -84,7 +84,7 @@ static int i915_dma_cleanup(struct drm_device * dev) * may not have been called from userspace and after dev_private * is freed, it's too late. */ - if (dev->irq) + if (dev->irq_enabled) drm_irq_uninstall(dev); if (dev_priv->ring.virtual_start) { @@ -644,7 +644,7 @@ static int i915_getparam(struct drm_device *dev, void *data, switch (param->param) { case I915_PARAM_IRQ_ACTIVE: - value = dev->irq ? 1 : 0; + value = dev->irq_enabled; break; case I915_PARAM_ALLOW_BATCHBUFFER: value = dev_priv->allow_batchbuffer ? 1 : 0; @@ -763,6 +763,20 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) ret = drm_addmap(dev, base, size, _DRM_REGISTERS, _DRM_KERNEL | _DRM_DRIVER, &dev_priv->mmio_map); + + + /* On the 945G/GM, the chipset reports the MSI capability on the + * integrated graphics even though the support isn't actually there + * according to the published specs. It doesn't appear to function + * correctly in testing on 945G. + * This may be a side effect of MSI having been made available for PEG + * and the registers being closely associated. + */ + if (!IS_I945G(dev) && !IS_I945GM(dev)) + pci_enable_msi(dev->pdev); + + spin_lock_init(&dev_priv->user_irq_lock); + return ret; } @@ -770,6 +784,9 @@ int i915_driver_unload(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + if (dev->pdev->msi_enabled) + pci_disable_msi(dev->pdev); + if (dev_priv->mmio_map) drm_rmmap(dev, dev_priv->mmio_map); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index afb51a390e1..8daf0d84cd5 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -105,6 +105,12 @@ typedef struct drm_i915_private { wait_queue_head_t irq_queue; atomic_t irq_received; atomic_t irq_emitted; + /** Protects user_irq_refcount and irq_mask_reg */ + spinlock_t user_irq_lock; + /** Refcount for i915_user_irq_get() versus i915_user_irq_put(). */ + int user_irq_refcount; + /** Cached value of IMR to avoid reads in updating the bitfield */ + u32 irq_mask_reg; int tex_lru_log_granularity; int allow_batchbuffer; diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 4a2de789734..24d11ed5bbc 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -33,6 +33,31 @@ #define MAX_NOPID ((u32)~0) +/** These are the interrupts used by the driver */ +#define I915_INTERRUPT_ENABLE_MASK (I915_USER_INTERRUPT | \ + I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | \ + I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) + +static inline void +i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask) +{ + if ((dev_priv->irq_mask_reg & mask) != 0) { + dev_priv->irq_mask_reg &= ~mask; + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IMR); + } +} + +static inline void +i915_disable_irq(drm_i915_private_t *dev_priv, u32 mask) +{ + if ((dev_priv->irq_mask_reg & mask) != mask) { + dev_priv->irq_mask_reg |= mask; + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IMR); + } +} + /** * Emit blits for scheduled buffer swaps. * @@ -229,46 +254,50 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) { struct drm_device *dev = (struct drm_device *) arg; drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - u16 temp; u32 pipea_stats, pipeb_stats; + u32 iir; pipea_stats = I915_READ(PIPEASTAT); pipeb_stats = I915_READ(PIPEBSTAT); - temp = I915_READ16(IIR); - - temp &= (I915_USER_INTERRUPT | - I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | - I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT); + if (dev->pdev->msi_enabled) + I915_WRITE(IMR, ~0); + iir = I915_READ(IIR); - DRM_DEBUG("%s flag=%08x\n", __FUNCTION__, temp); + DRM_DEBUG("iir=%08x\n", iir); - if (temp == 0) + if (iir == 0) { + if (dev->pdev->msi_enabled) { + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IMR); + } return IRQ_NONE; + } - I915_WRITE16(IIR, temp); - (void) I915_READ16(IIR); - DRM_READMEMORYBARRIER(); + I915_WRITE(IIR, iir); + if (dev->pdev->msi_enabled) + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IIR); /* Flush posted writes */ dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - if (temp & I915_USER_INTERRUPT) + if (iir & I915_USER_INTERRUPT) DRM_WAKEUP(&dev_priv->irq_queue); - if (temp & (I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | - I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT)) { + if (iir & (I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | + I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT)) { int vblank_pipe = dev_priv->vblank_pipe; if ((vblank_pipe & (DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B)) == (DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B)) { - if (temp & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) + if (iir & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) atomic_inc(&dev->vbl_received); - if (temp & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) + if (iir & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) atomic_inc(&dev->vbl_received2); - } else if (((temp & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) && + } else if (((iir & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) && (vblank_pipe & DRM_I915_VBLANK_PIPE_A)) || - ((temp & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) && + ((iir & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) && (vblank_pipe & DRM_I915_VBLANK_PIPE_B))) atomic_inc(&dev->vbl_received); @@ -314,6 +343,27 @@ static int i915_emit_irq(struct drm_device * dev) return dev_priv->counter; } +static void i915_user_irq_get(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + + spin_lock(&dev_priv->user_irq_lock); + if (dev->irq_enabled && (++dev_priv->user_irq_refcount == 1)) + i915_enable_irq(dev_priv, I915_USER_INTERRUPT); + spin_unlock(&dev_priv->user_irq_lock); +} + +static void i915_user_irq_put(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + + spin_lock(&dev_priv->user_irq_lock); + BUG_ON(dev->irq_enabled && dev_priv->user_irq_refcount <= 0); + if (dev->irq_enabled && (--dev_priv->user_irq_refcount == 0)) + i915_disable_irq(dev_priv, I915_USER_INTERRUPT); + spin_unlock(&dev_priv->user_irq_lock); +} + static int i915_wait_irq(struct drm_device * dev, int irq_nr) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; @@ -322,13 +372,17 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) DRM_DEBUG("irq_nr=%d breadcrumb=%d\n", irq_nr, READ_BREADCRUMB(dev_priv)); - if (READ_BREADCRUMB(dev_priv) >= irq_nr) + if (READ_BREADCRUMB(dev_priv) >= irq_nr) { + dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); return 0; + } dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; + i915_user_irq_get(dev); DRM_WAIT_ON(ret, dev_priv->irq_queue, 3 * DRM_HZ, READ_BREADCRUMB(dev_priv) >= irq_nr); + i915_user_irq_put(dev); if (ret == -EBUSY) { DRM_ERROR("EBUSY -- rec: %d emitted: %d\n", @@ -413,20 +467,6 @@ int i915_irq_wait(struct drm_device *dev, void *data, return i915_wait_irq(dev, irqwait->irq_seq); } -static void i915_enable_interrupt (struct drm_device *dev) -{ - drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - u16 flag; - - flag = 0; - if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_A) - flag |= I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; - if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B) - flag |= I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; - - I915_WRITE16(IER, I915_USER_INTERRUPT | flag); -} - /* Set the vblank monitor pipe */ int i915_vblank_pipe_set(struct drm_device *dev, void *data, @@ -434,6 +474,7 @@ int i915_vblank_pipe_set(struct drm_device *dev, void *data, { drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_vblank_pipe_t *pipe = data; + u32 enable_mask = 0, disable_mask = 0; if (!dev_priv) { DRM_ERROR("called with no initialization\n"); @@ -445,9 +486,20 @@ int i915_vblank_pipe_set(struct drm_device *dev, void *data, return -EINVAL; } - dev_priv->vblank_pipe = pipe->pipe; + if (pipe->pipe & DRM_I915_VBLANK_PIPE_A) + enable_mask |= I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; + else + disable_mask |= I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; + + if (pipe->pipe & DRM_I915_VBLANK_PIPE_B) + enable_mask |= I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; + else + disable_mask |= I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; - i915_enable_interrupt (dev); + i915_enable_irq(dev_priv, enable_mask); + i915_disable_irq(dev_priv, disable_mask); + + dev_priv->vblank_pipe = pipe->pipe; return 0; } @@ -464,7 +516,7 @@ int i915_vblank_pipe_get(struct drm_device *dev, void *data, return -EINVAL; } - flag = I915_READ(IER); + flag = I915_READ(IMR); pipe->pipe = 0; if (flag & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) pipe->pipe |= DRM_I915_VBLANK_PIPE_A; @@ -586,9 +638,9 @@ void i915_driver_irq_preinstall(struct drm_device * dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - I915_WRITE16(HWSTAM, 0xfffe); - I915_WRITE16(IMR, 0x0); - I915_WRITE16(IER, 0x0); + I915_WRITE(HWSTAM, 0xfffe); + I915_WRITE(IMR, 0x0); + I915_WRITE(IER, 0x0); } void i915_driver_irq_postinstall(struct drm_device * dev) @@ -601,7 +653,18 @@ void i915_driver_irq_postinstall(struct drm_device * dev) if (!dev_priv->vblank_pipe) dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A; - i915_enable_interrupt(dev); + + /* Set initial unmasked IRQs to just the selected vblank pipes. */ + dev_priv->irq_mask_reg = ~0; + if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_A) + dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; + if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B) + dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; + + I915_WRITE(IMR, dev_priv->irq_mask_reg); + I915_WRITE(IER, I915_INTERRUPT_ENABLE_MASK); + (void) I915_READ(IER); + DRM_INIT_WAITQUEUE(&dev_priv->irq_queue); } @@ -613,10 +676,10 @@ void i915_driver_irq_uninstall(struct drm_device * dev) if (!dev_priv) return; - I915_WRITE16(HWSTAM, 0xffff); - I915_WRITE16(IMR, 0xffff); - I915_WRITE16(IER, 0x0); + I915_WRITE(HWSTAM, 0xffff); + I915_WRITE(IMR, 0xffff); + I915_WRITE(IER, 0x0); - temp = I915_READ16(IIR); - I915_WRITE16(IIR, temp); + temp = I915_READ(IIR); + I915_WRITE(IIR, temp); } -- cgit v1.2.3 From d3a6d4467ca44833bcb4ba1893a7aeaae939e4d5 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Wed, 30 Jul 2008 12:21:20 -0700 Subject: i915: Track progress inside of batchbuffers for determining wedgedness. This avoids early termination for long-running commands. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 27a1f78a7f1..aa9e8da9f71 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -40,11 +40,15 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) { drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_ring_buffer_t *ring = &(dev_priv->ring); + u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD; + u32 last_acthd = I915_READ(acthd_reg); + u32 acthd; u32 last_head = I915_READ(PRB0_HEAD) & HEAD_ADDR; int i; - for (i = 0; i < 10000; i++) { + for (i = 0; i < 100000; i++) { ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; + acthd = I915_READ(acthd_reg); ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) ring->space += ring->Size; @@ -55,8 +59,13 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) if (ring->head != last_head) i = 0; + if (acthd != last_acthd) + i = 0; last_head = ring->head; + last_acthd = acthd; + msleep_interruptible(10); + } return -EBUSY; -- cgit v1.2.3 From 398c9cb20b5c6c5d1313912b937d653a46fec578 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Wed, 30 Jul 2008 13:03:43 -0700 Subject: i915: Initialize hardware status page at device load when possible. Some chips were unstable with repeated setup/teardown of the hardware status page. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 86 +++++++++++++++++++++++++++-------------- 1 file changed, 57 insertions(+), 29 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index aa9e8da9f71..f1b5aa92fa8 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -71,6 +71,52 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) return -EBUSY; } +/** + * Sets up the hardware status page for devices that need a physical address + * in the register. + */ +int i915_init_phys_hws(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + /* Program Hardware Status Page */ + dev_priv->status_page_dmah = + drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE, 0xffffffff); + + if (!dev_priv->status_page_dmah) { + DRM_ERROR("Can not allocate hardware status page\n"); + return -ENOMEM; + } + dev_priv->hw_status_page = dev_priv->status_page_dmah->vaddr; + dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr; + + memset(dev_priv->hw_status_page, 0, PAGE_SIZE); + + I915_WRITE(HWS_PGA, dev_priv->dma_status_page); + DRM_DEBUG("Enabled hardware status page\n"); + return 0; +} + +/** + * Frees the hardware status page, whether it's a physical address or a virtual + * address set up by the X Server. + */ +void i915_free_hws(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + if (dev_priv->status_page_dmah) { + drm_pci_free(dev, dev_priv->status_page_dmah); + dev_priv->status_page_dmah = NULL; + } + + if (dev_priv->status_gfx_addr) { + dev_priv->status_gfx_addr = 0; + drm_core_ioremapfree(&dev_priv->hws_map, dev); + } + + /* Need to rewrite hardware status page */ + I915_WRITE(HWS_PGA, 0x1ffff000); +} + void i915_kernel_lost_context(struct drm_device * dev) { drm_i915_private_t *dev_priv = dev->dev_private; @@ -103,18 +149,9 @@ static int i915_dma_cleanup(struct drm_device * dev) dev_priv->ring.map.size = 0; } - if (dev_priv->status_page_dmah) { - drm_pci_free(dev, dev_priv->status_page_dmah); - dev_priv->status_page_dmah = NULL; - /* Need to rewrite hardware status page */ - I915_WRITE(HWS_PGA, 0x1ffff000); - } - - if (dev_priv->status_gfx_addr) { - dev_priv->status_gfx_addr = 0; - drm_core_ioremapfree(&dev_priv->hws_map, dev); - I915_WRITE(HWS_PGA, 0x1ffff000); - } + /* Clear the HWS virtual address at teardown */ + if (I915_NEED_GFX_HWS(dev)) + i915_free_hws(dev); return 0; } @@ -165,23 +202,6 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) */ dev_priv->allow_batchbuffer = 1; - /* Program Hardware Status Page */ - if (!I915_NEED_GFX_HWS(dev)) { - dev_priv->status_page_dmah = - drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE, 0xffffffff); - - if (!dev_priv->status_page_dmah) { - i915_dma_cleanup(dev); - DRM_ERROR("Can not allocate hardware status page\n"); - return -ENOMEM; - } - dev_priv->hw_status_page = dev_priv->status_page_dmah->vaddr; - dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr; - - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - I915_WRITE(HWS_PGA, dev_priv->dma_status_page); - } - DRM_DEBUG("Enabled hardware status page\n"); return 0; } @@ -773,6 +793,12 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) _DRM_KERNEL | _DRM_DRIVER, &dev_priv->mmio_map); + /* Init HWS */ + if (!I915_NEED_GFX_HWS(dev)) { + ret = i915_init_phys_hws(dev); + if (ret != 0) + return ret; + } /* On the 945G/GM, the chipset reports the MSI capability on the * integrated graphics even though the support isn't actually there @@ -796,6 +822,8 @@ int i915_driver_unload(struct drm_device *dev) if (dev->pdev->msi_enabled) pci_disable_msi(dev->pdev); + i915_free_hws(dev); + if (dev_priv->mmio_map) drm_rmmap(dev, dev_priv->mmio_map); -- cgit v1.2.3 From 8ee1c3db9075cb3211352e737e0feb98fd733b20 Mon Sep 17 00:00:00 2001 From: Matthew Garrett Date: Tue, 5 Aug 2008 19:37:25 +0100 Subject: Add Intel ACPI IGD OpRegion support This adds the support necessary for allowing ACPI backlight control to work on some newer Intel-based graphics systems. Tested on Thinkpad T61 and HP 2510p hardware. Signed-off-by: Matthew Garrett Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/Makefile | 2 +- drivers/gpu/drm/i915/i915_dma.c | 4 + drivers/gpu/drm/i915/i915_drv.c | 4 + drivers/gpu/drm/i915/i915_drv.h | 17 ++ drivers/gpu/drm/i915/i915_irq.c | 25 ++- drivers/gpu/drm/i915/i915_opregion.c | 371 +++++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_reg.h | 1 + 7 files changed, 415 insertions(+), 9 deletions(-) create mode 100644 drivers/gpu/drm/i915/i915_opregion.c (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index a9e60464df7..b032808d9c1 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -3,7 +3,7 @@ # Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher. ccflags-y := -Iinclude/drm -i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o +i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o i915_opregion.o i915-$(CONFIG_COMPAT) += i915_ioc32.o diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index f1b5aa92fa8..29115450ee4 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -810,6 +810,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) if (!IS_I945G(dev) && !IS_I945GM(dev)) pci_enable_msi(dev->pdev); + intel_opregion_init(dev); + spin_lock_init(&dev_priv->user_irq_lock); return ret; @@ -827,6 +829,8 @@ int i915_driver_unload(struct drm_device *dev) if (dev_priv->mmio_map) drm_rmmap(dev, dev_priv->mmio_map); + intel_opregion_free(dev); + drm_free(dev->dev_private, sizeof(drm_i915_private_t), DRM_MEM_DRIVER); diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 6c99aab12da..d95eca2bc45 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -371,6 +371,8 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) i915_save_vga(dev); + intel_opregion_free(dev); + if (state.event == PM_EVENT_SUSPEND) { /* Shut down the device */ pci_disable_device(dev->pdev); @@ -532,6 +534,8 @@ static int i915_resume(struct drm_device *dev) i915_restore_vga(dev); + intel_opregion_init(dev); + return 0; } diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 8daf0d84cd5..e4bd01c511a 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -82,6 +82,14 @@ typedef struct _drm_i915_vbl_swap { unsigned int sequence; } drm_i915_vbl_swap_t; +struct intel_opregion { + struct opregion_header *header; + struct opregion_acpi *acpi; + struct opregion_swsci *swsci; + struct opregion_asle *asle; + int enabled; +}; + typedef struct drm_i915_private { drm_local_map_t *sarea; drm_local_map_t *mmio_map; @@ -122,6 +130,8 @@ typedef struct drm_i915_private { drm_i915_vbl_swap_t vbl_swaps; unsigned int swaps_pending; + struct intel_opregion opregion; + /* Register state */ u8 saveLBB; u32 saveDSPACNTR; @@ -244,6 +254,7 @@ extern int i915_vblank_pipe_get(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int i915_vblank_swap(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern void i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask); /* i915_mem.c */ extern int i915_mem_alloc(struct drm_device *dev, void *data, @@ -258,6 +269,12 @@ extern void i915_mem_takedown(struct mem_block **heap); extern void i915_mem_release(struct drm_device * dev, struct drm_file *file_priv, struct mem_block *heap); +/* i915_opregion.c */ +extern int intel_opregion_init(struct drm_device *dev); +extern void intel_opregion_free(struct drm_device *dev); +extern void opregion_asle_intr(struct drm_device *dev); +extern void opregion_enable_asle(struct drm_device *dev); + #define I915_READ(reg) DRM_READ32(dev_priv->mmio_map, (reg)) #define I915_WRITE(reg,val) DRM_WRITE32(dev_priv->mmio_map, (reg), (val)) #define I915_READ16(reg) DRM_READ16(dev_priv->mmio_map, (reg)) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 24d11ed5bbc..ae7d3a82a6d 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -36,9 +36,11 @@ /** These are the interrupts used by the driver */ #define I915_INTERRUPT_ENABLE_MASK (I915_USER_INTERRUPT | \ I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | \ - I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) + I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT | \ + I915_ASLE_INTERRUPT | \ + I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) -static inline void +void i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask) { if ((dev_priv->irq_mask_reg & mask) != 0) { @@ -274,6 +276,9 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) return IRQ_NONE; } + I915_WRITE(PIPEASTAT, pipea_stats); + I915_WRITE(PIPEBSTAT, pipeb_stats); + I915_WRITE(IIR, iir); if (dev->pdev->msi_enabled) I915_WRITE(IMR, dev_priv->irq_mask_reg); @@ -306,14 +311,14 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) if (dev_priv->swaps_pending > 0) drm_locked_tasklet(dev, i915_vblank_tasklet); - I915_WRITE(PIPEASTAT, - pipea_stats|I915_VBLANK_INTERRUPT_ENABLE| - PIPE_VBLANK_INTERRUPT_STATUS); - I915_WRITE(PIPEBSTAT, - pipeb_stats|I915_VBLANK_INTERRUPT_ENABLE| - PIPE_VBLANK_INTERRUPT_STATUS); } + if (iir & I915_ASLE_INTERRUPT) + opregion_asle_intr(dev); + + if (iir & I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) + opregion_asle_intr(dev); + return IRQ_HANDLED; } @@ -661,10 +666,14 @@ void i915_driver_irq_postinstall(struct drm_device * dev) if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B) dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; + dev_priv->irq_mask_reg &= I915_INTERRUPT_ENABLE_MASK; + I915_WRITE(IMR, dev_priv->irq_mask_reg); I915_WRITE(IER, I915_INTERRUPT_ENABLE_MASK); (void) I915_READ(IER); + opregion_enable_asle(dev); + DRM_INIT_WAITQUEUE(&dev_priv->irq_queue); } diff --git a/drivers/gpu/drm/i915/i915_opregion.c b/drivers/gpu/drm/i915/i915_opregion.c new file mode 100644 index 00000000000..1787a0c7e3a --- /dev/null +++ b/drivers/gpu/drm/i915/i915_opregion.c @@ -0,0 +1,371 @@ +/* + * Copyright 2008 Intel Corporation + * Copyright 2008 Red Hat + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NON-INFRINGEMENT. IN NO EVENT SHALL INTEL AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#include + +#include "drmP.h" +#include "i915_drm.h" +#include "i915_drv.h" + +#define PCI_ASLE 0xe4 +#define PCI_LBPC 0xf4 +#define PCI_ASLS 0xfc + +#define OPREGION_SZ (8*1024) +#define OPREGION_HEADER_OFFSET 0 +#define OPREGION_ACPI_OFFSET 0x100 +#define OPREGION_SWSCI_OFFSET 0x200 +#define OPREGION_ASLE_OFFSET 0x300 +#define OPREGION_VBT_OFFSET 0x1000 + +#define OPREGION_SIGNATURE "IntelGraphicsMem" +#define MBOX_ACPI (1<<0) +#define MBOX_SWSCI (1<<1) +#define MBOX_ASLE (1<<2) + +struct opregion_header { + u8 signature[16]; + u32 size; + u32 opregion_ver; + u8 bios_ver[32]; + u8 vbios_ver[16]; + u8 driver_ver[16]; + u32 mboxes; + u8 reserved[164]; +} __attribute__((packed)); + +/* OpRegion mailbox #1: public ACPI methods */ +struct opregion_acpi { + u32 drdy; /* driver readiness */ + u32 csts; /* notification status */ + u32 cevt; /* current event */ + u8 rsvd1[20]; + u32 didl[8]; /* supported display devices ID list */ + u32 cpdl[8]; /* currently presented display list */ + u32 cadl[8]; /* currently active display list */ + u32 nadl[8]; /* next active devices list */ + u32 aslp; /* ASL sleep time-out */ + u32 tidx; /* toggle table index */ + u32 chpd; /* current hotplug enable indicator */ + u32 clid; /* current lid state*/ + u32 cdck; /* current docking state */ + u32 sxsw; /* Sx state resume */ + u32 evts; /* ASL supported events */ + u32 cnot; /* current OS notification */ + u32 nrdy; /* driver status */ + u8 rsvd2[60]; +} __attribute__((packed)); + +/* OpRegion mailbox #2: SWSCI */ +struct opregion_swsci { + u32 scic; /* SWSCI command|status|data */ + u32 parm; /* command parameters */ + u32 dslp; /* driver sleep time-out */ + u8 rsvd[244]; +} __attribute__((packed)); + +/* OpRegion mailbox #3: ASLE */ +struct opregion_asle { + u32 ardy; /* driver readiness */ + u32 aslc; /* ASLE interrupt command */ + u32 tche; /* technology enabled indicator */ + u32 alsi; /* current ALS illuminance reading */ + u32 bclp; /* backlight brightness to set */ + u32 pfit; /* panel fitting state */ + u32 cblv; /* current brightness level */ + u16 bclm[20]; /* backlight level duty cycle mapping table */ + u32 cpfm; /* current panel fitting mode */ + u32 epfm; /* enabled panel fitting modes */ + u8 plut[74]; /* panel LUT and identifier */ + u32 pfmb; /* PWM freq and min brightness */ + u8 rsvd[102]; +} __attribute__((packed)); + +/* ASLE irq request bits */ +#define ASLE_SET_ALS_ILLUM (1 << 0) +#define ASLE_SET_BACKLIGHT (1 << 1) +#define ASLE_SET_PFIT (1 << 2) +#define ASLE_SET_PWM_FREQ (1 << 3) +#define ASLE_REQ_MSK 0xf + +/* response bits of ASLE irq request */ +#define ASLE_ALS_ILLUM_FAIL (2<<10) +#define ASLE_BACKLIGHT_FAIL (2<<12) +#define ASLE_PFIT_FAIL (2<<14) +#define ASLE_PWM_FREQ_FAIL (2<<16) + +/* ASLE backlight brightness to set */ +#define ASLE_BCLP_VALID (1<<31) +#define ASLE_BCLP_MSK (~(1<<31)) + +/* ASLE panel fitting request */ +#define ASLE_PFIT_VALID (1<<31) +#define ASLE_PFIT_CENTER (1<<0) +#define ASLE_PFIT_STRETCH_TEXT (1<<1) +#define ASLE_PFIT_STRETCH_GFX (1<<2) + +/* PWM frequency and minimum brightness */ +#define ASLE_PFMB_BRIGHTNESS_MASK (0xff) +#define ASLE_PFMB_BRIGHTNESS_VALID (1<<8) +#define ASLE_PFMB_PWM_MASK (0x7ffffe00) +#define ASLE_PFMB_PWM_VALID (1<<31) + +#define ASLE_CBLV_VALID (1<<31) + +static u32 asle_set_backlight(struct drm_device *dev, u32 bclp) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct opregion_asle *asle = dev_priv->opregion.asle; + u32 blc_pwm_ctl, blc_pwm_ctl2; + + if (!(bclp & ASLE_BCLP_VALID)) + return ASLE_BACKLIGHT_FAIL; + + bclp &= ASLE_BCLP_MSK; + if (bclp < 0 || bclp > 255) + return ASLE_BACKLIGHT_FAIL; + + blc_pwm_ctl = I915_READ(BLC_PWM_CTL); + blc_pwm_ctl &= ~BACKLIGHT_DUTY_CYCLE_MASK; + blc_pwm_ctl2 = I915_READ(BLC_PWM_CTL2); + + if (blc_pwm_ctl2 & BLM_COMBINATION_MODE) + pci_write_config_dword(dev->pdev, PCI_LBPC, bclp); + else + I915_WRITE(BLC_PWM_CTL, blc_pwm_ctl | ((bclp * 0x101)-1)); + + asle->cblv = (bclp*0x64)/0xff | ASLE_CBLV_VALID; + + return 0; +} + +static u32 asle_set_als_illum(struct drm_device *dev, u32 alsi) +{ + /* alsi is the current ALS reading in lux. 0 indicates below sensor + range, 0xffff indicates above sensor range. 1-0xfffe are valid */ + return 0; +} + +static u32 asle_set_pwm_freq(struct drm_device *dev, u32 pfmb) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + if (pfmb & ASLE_PFMB_PWM_VALID) { + u32 blc_pwm_ctl = I915_READ(BLC_PWM_CTL); + u32 pwm = pfmb & ASLE_PFMB_PWM_MASK; + blc_pwm_ctl &= BACKLIGHT_DUTY_CYCLE_MASK; + pwm = pwm >> 9; + /* FIXME - what do we do with the PWM? */ + } + return 0; +} + +static u32 asle_set_pfit(struct drm_device *dev, u32 pfit) +{ + /* Panel fitting is currently controlled by the X code, so this is a + noop until modesetting support works fully */ + if (!(pfit & ASLE_PFIT_VALID)) + return ASLE_PFIT_FAIL; + return 0; +} + +void opregion_asle_intr(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct opregion_asle *asle = dev_priv->opregion.asle; + u32 asle_stat = 0; + u32 asle_req; + + if (!asle) + return; + + asle_req = asle->aslc & ASLE_REQ_MSK; + + if (!asle_req) { + DRM_DEBUG("non asle set request??\n"); + return; + } + + if (asle_req & ASLE_SET_ALS_ILLUM) + asle_stat |= asle_set_als_illum(dev, asle->alsi); + + if (asle_req & ASLE_SET_BACKLIGHT) + asle_stat |= asle_set_backlight(dev, asle->bclp); + + if (asle_req & ASLE_SET_PFIT) + asle_stat |= asle_set_pfit(dev, asle->pfit); + + if (asle_req & ASLE_SET_PWM_FREQ) + asle_stat |= asle_set_pwm_freq(dev, asle->pfmb); + + asle->aslc = asle_stat; +} + +#define ASLE_ALS_EN (1<<0) +#define ASLE_BLC_EN (1<<1) +#define ASLE_PFIT_EN (1<<2) +#define ASLE_PFMB_EN (1<<3) + +void opregion_enable_asle(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct opregion_asle *asle = dev_priv->opregion.asle; + + if (asle) { + u32 pipeb_stats = I915_READ(PIPEBSTAT); + if (IS_MOBILE(dev)) { + /* Many devices trigger events with a write to the + legacy backlight controller, so we need to ensure + that it's able to generate interrupts */ + I915_WRITE(PIPEBSTAT, pipeb_stats |= + I915_LEGACY_BLC_EVENT_ENABLE); + i915_enable_irq(dev_priv, I915_ASLE_INTERRUPT | + I915_DISPLAY_PIPE_B_EVENT_INTERRUPT); + } else + i915_enable_irq(dev_priv, I915_ASLE_INTERRUPT); + + asle->tche = ASLE_ALS_EN | ASLE_BLC_EN | ASLE_PFIT_EN | + ASLE_PFMB_EN; + asle->ardy = 1; + } +} + +#define ACPI_EV_DISPLAY_SWITCH (1<<0) +#define ACPI_EV_LID (1<<1) +#define ACPI_EV_DOCK (1<<2) + +static struct intel_opregion *system_opregion; + +int intel_opregion_video_event(struct notifier_block *nb, unsigned long val, + void *data) +{ + /* The only video events relevant to opregion are 0x80. These indicate + either a docking event, lid switch or display switch request. In + Linux, these are handled by the dock, button and video drivers. + We might want to fix the video driver to be opregion-aware in + future, but right now we just indicate to the firmware that the + request has been handled */ + + struct opregion_acpi *acpi; + + if (!system_opregion) + return NOTIFY_DONE; + + acpi = system_opregion->acpi; + acpi->csts = 0; + + return NOTIFY_OK; +} + +static struct notifier_block intel_opregion_notifier = { + .notifier_call = intel_opregion_video_event, +}; + +int intel_opregion_init(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_opregion *opregion = &dev_priv->opregion; + void *base; + u32 asls, mboxes; + int err = 0; + + pci_read_config_dword(dev->pdev, PCI_ASLS, &asls); + DRM_DEBUG("graphic opregion physical addr: 0x%x\n", asls); + if (asls == 0) { + DRM_DEBUG("ACPI OpRegion not supported!\n"); + return -ENOTSUPP; + } + + base = ioremap(asls, OPREGION_SZ); + if (!base) + return -ENOMEM; + + opregion->header = base; + if (memcmp(opregion->header->signature, OPREGION_SIGNATURE, 16)) { + DRM_DEBUG("opregion signature mismatch\n"); + err = -EINVAL; + goto err_out; + } + + mboxes = opregion->header->mboxes; + if (mboxes & MBOX_ACPI) { + DRM_DEBUG("Public ACPI methods supported\n"); + opregion->acpi = base + OPREGION_ACPI_OFFSET; + } else { + DRM_DEBUG("Public ACPI methods not supported\n"); + err = -ENOTSUPP; + goto err_out; + } + opregion->enabled = 1; + + if (mboxes & MBOX_SWSCI) { + DRM_DEBUG("SWSCI supported\n"); + opregion->swsci = base + OPREGION_SWSCI_OFFSET; + } + if (mboxes & MBOX_ASLE) { + DRM_DEBUG("ASLE supported\n"); + opregion->asle = base + OPREGION_ASLE_OFFSET; + } + + /* Notify BIOS we are ready to handle ACPI video ext notifs. + * Right now, all the events are handled by the ACPI video module. + * We don't actually need to do anything with them. */ + opregion->acpi->csts = 0; + opregion->acpi->drdy = 1; + + system_opregion = opregion; + register_acpi_notifier(&intel_opregion_notifier); + + return 0; + +err_out: + iounmap(opregion->header); + opregion->header = NULL; + return err; +} + +void intel_opregion_free(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_opregion *opregion = &dev_priv->opregion; + + if (!opregion->enabled) + return; + + opregion->acpi->drdy = 0; + + system_opregion = NULL; + unregister_acpi_notifier(&intel_opregion_notifier); + + /* just clear all opregion memory pointers now */ + iounmap(opregion->header); + opregion->header = NULL; + opregion->acpi = NULL; + opregion->swsci = NULL; + opregion->asle = NULL; + + opregion->enabled = 0; +} diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 477c64e6ba2..43ad2cb8b9f 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -740,6 +740,7 @@ #define BLC_PWM_CTL 0x61254 #define BACKLIGHT_MODULATION_FREQ_SHIFT (17) #define BLC_PWM_CTL2 0x61250 /* 965+ only */ +#define BLM_COMBINATION_MODE (1 << 30) /* * This is the most significant 15 bits of the number of backlight cycles in a * complete cycle of the modulated backlight control. -- cgit v1.2.3 From dfcf96d09cff63c9aaa8e7c98bbc71e5073b1377 Mon Sep 17 00:00:00 2001 From: Adrian Bunk Date: Sun, 24 Aug 2008 17:11:22 +1000 Subject: FB_SIS=m, DRM_SIS=y is not a legal configuration. Reported-by: Randy Dunlap Signed-off-by: Adrian Bunk Signed-off-by: Dave Airlie --- drivers/gpu/drm/Kconfig | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index 610d6fd5bb5..bf9003ee7ab 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -87,6 +87,7 @@ config DRM_MGA config DRM_SIS tristate "SiS video cards" depends on DRM && AGP + depends on FB_SIS || FB_SIS=n help Choose this option if you have a SiS 630 or compatible video chipset. If M is selected the module will be called sis. AGP -- cgit v1.2.3 From 38eda21189b414b1520ea7aa0e71220796f7008f Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 2 Sep 2008 10:06:06 +1000 Subject: drm: fix sysfs error path. Pointed out by Roel Kluin on dri-devel. Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_sysfs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_sysfs.c b/drivers/gpu/drm/drm_sysfs.c index af211a0ef17..1611b9bcbe7 100644 --- a/drivers/gpu/drm/drm_sysfs.c +++ b/drivers/gpu/drm/drm_sysfs.c @@ -184,7 +184,7 @@ int drm_sysfs_device_add(struct drm_minor *minor) err_out_files: if (i > 0) for (j = 0; j < i; j++) - device_remove_file(&minor->kdev, &device_attrs[i]); + device_remove_file(&minor->kdev, &device_attrs[j]); device_unregister(&minor->kdev); err_out: -- cgit v1.2.3 From 6b79d521e07aae155303a992245abb539974dbaa Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 2 Sep 2008 10:10:16 +1000 Subject: radeon: fix writeback across suspend/resume. Make writeback not get disabled on resume. Signed-off-by: Dave Airlie --- drivers/gpu/drm/radeon/radeon_cp.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c index 248ab4a7d39..e1678cae9cc 100644 --- a/drivers/gpu/drm/radeon/radeon_cp.c +++ b/drivers/gpu/drm/radeon/radeon_cp.c @@ -626,8 +626,6 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, dev_priv->ring.size_l2qw); #endif - /* Start with assuming that writeback doesn't work */ - dev_priv->writeback_works = 0; /* Initialize the scratch register pointer. This will cause * the scratch register values to be written out to memory @@ -674,6 +672,9 @@ static void radeon_test_writeback(drm_radeon_private_t * dev_priv) { u32 tmp; + /* Start with assuming that writeback doesn't work */ + dev_priv->writeback_works = 0; + /* Writeback doesn't seem to work everywhere, test it here and possibly * enable it if it appears to work */ -- cgit v1.2.3 From 317c35d1446f68b34d4de4e1100fc01680bd4877 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Mon, 25 Aug 2008 15:11:06 -0700 Subject: separate i915 suspend/resume functions into their own file [Patch against drm-next. Consider this a trial balloon for our new Linux development model.] This is a big chunk of code. Separating it out makes it easier to change without churn on the main i915_drv.c file (and there will be churn as we fix bugs and add things like kernel mode setting). Also makes it easier to share this file with BSD. Signed-off-by: Jesse Barnes Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/i915_drv.c | 462 +------------------------------- drivers/gpu/drm/i915/i915_drv.h | 11 + drivers/gpu/drm/i915/i915_suspend.c | 509 ++++++++++++++++++++++++++++++++++++ 4 files changed, 524 insertions(+), 461 deletions(-) create mode 100644 drivers/gpu/drm/i915/i915_suspend.c (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index b032808d9c1..c4bbda662e2 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -3,7 +3,8 @@ # Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher. ccflags-y := -Iinclude/drm -i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o i915_opregion.o +i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o i915_opregion.o \ + i915_suspend.o i915-$(CONFIG_COMPAT) += i915_ioc32.o diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index d95eca2bc45..eff66ed3e58 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -38,211 +38,9 @@ static struct pci_device_id pciidlist[] = { i915_PCI_IDS }; -enum pipe { - PIPE_A = 0, - PIPE_B, -}; - -static bool i915_pipe_enabled(struct drm_device *dev, enum pipe pipe) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - - if (pipe == PIPE_A) - return (I915_READ(DPLL_A) & DPLL_VCO_ENABLE); - else - return (I915_READ(DPLL_B) & DPLL_VCO_ENABLE); -} - -static void i915_save_palette(struct drm_device *dev, enum pipe pipe) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long reg = (pipe == PIPE_A ? PALETTE_A : PALETTE_B); - u32 *array; - int i; - - if (!i915_pipe_enabled(dev, pipe)) - return; - - if (pipe == PIPE_A) - array = dev_priv->save_palette_a; - else - array = dev_priv->save_palette_b; - - for(i = 0; i < 256; i++) - array[i] = I915_READ(reg + (i << 2)); -} - -static void i915_restore_palette(struct drm_device *dev, enum pipe pipe) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - unsigned long reg = (pipe == PIPE_A ? PALETTE_A : PALETTE_B); - u32 *array; - int i; - - if (!i915_pipe_enabled(dev, pipe)) - return; - - if (pipe == PIPE_A) - array = dev_priv->save_palette_a; - else - array = dev_priv->save_palette_b; - - for(i = 0; i < 256; i++) - I915_WRITE(reg + (i << 2), array[i]); -} - -static u8 i915_read_indexed(u16 index_port, u16 data_port, u8 reg) -{ - outb(reg, index_port); - return inb(data_port); -} - -static u8 i915_read_ar(u16 st01, u8 reg, u16 palette_enable) -{ - inb(st01); - outb(palette_enable | reg, VGA_AR_INDEX); - return inb(VGA_AR_DATA_READ); -} - -static void i915_write_ar(u8 st01, u8 reg, u8 val, u16 palette_enable) -{ - inb(st01); - outb(palette_enable | reg, VGA_AR_INDEX); - outb(val, VGA_AR_DATA_WRITE); -} - -static void i915_write_indexed(u16 index_port, u16 data_port, u8 reg, u8 val) -{ - outb(reg, index_port); - outb(val, data_port); -} - -static void i915_save_vga(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int i; - u16 cr_index, cr_data, st01; - - /* VGA color palette registers */ - dev_priv->saveDACMASK = inb(VGA_DACMASK); - /* DACCRX automatically increments during read */ - outb(0, VGA_DACRX); - /* Read 3 bytes of color data from each index */ - for (i = 0; i < 256 * 3; i++) - dev_priv->saveDACDATA[i] = inb(VGA_DACDATA); - - /* MSR bits */ - dev_priv->saveMSR = inb(VGA_MSR_READ); - if (dev_priv->saveMSR & VGA_MSR_CGA_MODE) { - cr_index = VGA_CR_INDEX_CGA; - cr_data = VGA_CR_DATA_CGA; - st01 = VGA_ST01_CGA; - } else { - cr_index = VGA_CR_INDEX_MDA; - cr_data = VGA_CR_DATA_MDA; - st01 = VGA_ST01_MDA; - } - - /* CRT controller regs */ - i915_write_indexed(cr_index, cr_data, 0x11, - i915_read_indexed(cr_index, cr_data, 0x11) & - (~0x80)); - for (i = 0; i <= 0x24; i++) - dev_priv->saveCR[i] = - i915_read_indexed(cr_index, cr_data, i); - /* Make sure we don't turn off CR group 0 writes */ - dev_priv->saveCR[0x11] &= ~0x80; - - /* Attribute controller registers */ - inb(st01); - dev_priv->saveAR_INDEX = inb(VGA_AR_INDEX); - for (i = 0; i <= 0x14; i++) - dev_priv->saveAR[i] = i915_read_ar(st01, i, 0); - inb(st01); - outb(dev_priv->saveAR_INDEX, VGA_AR_INDEX); - inb(st01); - - /* Graphics controller registers */ - for (i = 0; i < 9; i++) - dev_priv->saveGR[i] = - i915_read_indexed(VGA_GR_INDEX, VGA_GR_DATA, i); - - dev_priv->saveGR[0x10] = - i915_read_indexed(VGA_GR_INDEX, VGA_GR_DATA, 0x10); - dev_priv->saveGR[0x11] = - i915_read_indexed(VGA_GR_INDEX, VGA_GR_DATA, 0x11); - dev_priv->saveGR[0x18] = - i915_read_indexed(VGA_GR_INDEX, VGA_GR_DATA, 0x18); - - /* Sequencer registers */ - for (i = 0; i < 8; i++) - dev_priv->saveSR[i] = - i915_read_indexed(VGA_SR_INDEX, VGA_SR_DATA, i); -} - -static void i915_restore_vga(struct drm_device *dev) -{ - struct drm_i915_private *dev_priv = dev->dev_private; - int i; - u16 cr_index, cr_data, st01; - - /* MSR bits */ - outb(dev_priv->saveMSR, VGA_MSR_WRITE); - if (dev_priv->saveMSR & VGA_MSR_CGA_MODE) { - cr_index = VGA_CR_INDEX_CGA; - cr_data = VGA_CR_DATA_CGA; - st01 = VGA_ST01_CGA; - } else { - cr_index = VGA_CR_INDEX_MDA; - cr_data = VGA_CR_DATA_MDA; - st01 = VGA_ST01_MDA; - } - - /* Sequencer registers, don't write SR07 */ - for (i = 0; i < 7; i++) - i915_write_indexed(VGA_SR_INDEX, VGA_SR_DATA, i, - dev_priv->saveSR[i]); - - /* CRT controller regs */ - /* Enable CR group 0 writes */ - i915_write_indexed(cr_index, cr_data, 0x11, dev_priv->saveCR[0x11]); - for (i = 0; i <= 0x24; i++) - i915_write_indexed(cr_index, cr_data, i, dev_priv->saveCR[i]); - - /* Graphics controller regs */ - for (i = 0; i < 9; i++) - i915_write_indexed(VGA_GR_INDEX, VGA_GR_DATA, i, - dev_priv->saveGR[i]); - - i915_write_indexed(VGA_GR_INDEX, VGA_GR_DATA, 0x10, - dev_priv->saveGR[0x10]); - i915_write_indexed(VGA_GR_INDEX, VGA_GR_DATA, 0x11, - dev_priv->saveGR[0x11]); - i915_write_indexed(VGA_GR_INDEX, VGA_GR_DATA, 0x18, - dev_priv->saveGR[0x18]); - - /* Attribute controller registers */ - inb(st01); - for (i = 0; i <= 0x14; i++) - i915_write_ar(st01, i, dev_priv->saveAR[i], 0); - inb(st01); /* switch back to index mode */ - outb(dev_priv->saveAR_INDEX | 0x20, VGA_AR_INDEX); - inb(st01); - - /* VGA color palette registers */ - outb(dev_priv->saveDACMASK, VGA_DACMASK); - /* DACCRX automatically increments during read */ - outb(0, VGA_DACWX); - /* Read 3 bytes of color data from each index */ - for (i = 0; i < 256 * 3; i++) - outb(dev_priv->saveDACDATA[i], VGA_DACDATA); - -} - static int i915_suspend(struct drm_device *dev, pm_message_t state) { struct drm_i915_private *dev_priv = dev->dev_private; - int i; if (!dev || !dev_priv) { printk(KERN_ERR "dev: %p, dev_priv: %p\n", dev, dev_priv); @@ -254,122 +52,8 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) return 0; pci_save_state(dev->pdev); - pci_read_config_byte(dev->pdev, LBB, &dev_priv->saveLBB); - - /* Display arbitration control */ - dev_priv->saveDSPARB = I915_READ(DSPARB); - - /* Pipe & plane A info */ - dev_priv->savePIPEACONF = I915_READ(PIPEACONF); - dev_priv->savePIPEASRC = I915_READ(PIPEASRC); - dev_priv->saveFPA0 = I915_READ(FPA0); - dev_priv->saveFPA1 = I915_READ(FPA1); - dev_priv->saveDPLL_A = I915_READ(DPLL_A); - if (IS_I965G(dev)) - dev_priv->saveDPLL_A_MD = I915_READ(DPLL_A_MD); - dev_priv->saveHTOTAL_A = I915_READ(HTOTAL_A); - dev_priv->saveHBLANK_A = I915_READ(HBLANK_A); - dev_priv->saveHSYNC_A = I915_READ(HSYNC_A); - dev_priv->saveVTOTAL_A = I915_READ(VTOTAL_A); - dev_priv->saveVBLANK_A = I915_READ(VBLANK_A); - dev_priv->saveVSYNC_A = I915_READ(VSYNC_A); - dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A); - - dev_priv->saveDSPACNTR = I915_READ(DSPACNTR); - dev_priv->saveDSPASTRIDE = I915_READ(DSPASTRIDE); - dev_priv->saveDSPASIZE = I915_READ(DSPASIZE); - dev_priv->saveDSPAPOS = I915_READ(DSPAPOS); - dev_priv->saveDSPAADDR = I915_READ(DSPAADDR); - if (IS_I965G(dev)) { - dev_priv->saveDSPASURF = I915_READ(DSPASURF); - dev_priv->saveDSPATILEOFF = I915_READ(DSPATILEOFF); - } - i915_save_palette(dev, PIPE_A); - dev_priv->savePIPEASTAT = I915_READ(PIPEASTAT); - - /* Pipe & plane B info */ - dev_priv->savePIPEBCONF = I915_READ(PIPEBCONF); - dev_priv->savePIPEBSRC = I915_READ(PIPEBSRC); - dev_priv->saveFPB0 = I915_READ(FPB0); - dev_priv->saveFPB1 = I915_READ(FPB1); - dev_priv->saveDPLL_B = I915_READ(DPLL_B); - if (IS_I965G(dev)) - dev_priv->saveDPLL_B_MD = I915_READ(DPLL_B_MD); - dev_priv->saveHTOTAL_B = I915_READ(HTOTAL_B); - dev_priv->saveHBLANK_B = I915_READ(HBLANK_B); - dev_priv->saveHSYNC_B = I915_READ(HSYNC_B); - dev_priv->saveVTOTAL_B = I915_READ(VTOTAL_B); - dev_priv->saveVBLANK_B = I915_READ(VBLANK_B); - dev_priv->saveVSYNC_B = I915_READ(VSYNC_B); - dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A); - - dev_priv->saveDSPBCNTR = I915_READ(DSPBCNTR); - dev_priv->saveDSPBSTRIDE = I915_READ(DSPBSTRIDE); - dev_priv->saveDSPBSIZE = I915_READ(DSPBSIZE); - dev_priv->saveDSPBPOS = I915_READ(DSPBPOS); - dev_priv->saveDSPBADDR = I915_READ(DSPBADDR); - if (IS_I965GM(dev) || IS_IGD_GM(dev)) { - dev_priv->saveDSPBSURF = I915_READ(DSPBSURF); - dev_priv->saveDSPBTILEOFF = I915_READ(DSPBTILEOFF); - } - i915_save_palette(dev, PIPE_B); - dev_priv->savePIPEBSTAT = I915_READ(PIPEBSTAT); - - /* CRT state */ - dev_priv->saveADPA = I915_READ(ADPA); - /* LVDS state */ - dev_priv->savePP_CONTROL = I915_READ(PP_CONTROL); - dev_priv->savePFIT_PGM_RATIOS = I915_READ(PFIT_PGM_RATIOS); - dev_priv->saveBLC_PWM_CTL = I915_READ(BLC_PWM_CTL); - if (IS_I965G(dev)) - dev_priv->saveBLC_PWM_CTL2 = I915_READ(BLC_PWM_CTL2); - if (IS_MOBILE(dev) && !IS_I830(dev)) - dev_priv->saveLVDS = I915_READ(LVDS); - if (!IS_I830(dev) && !IS_845G(dev)) - dev_priv->savePFIT_CONTROL = I915_READ(PFIT_CONTROL); - dev_priv->savePP_ON_DELAYS = I915_READ(PP_ON_DELAYS); - dev_priv->savePP_OFF_DELAYS = I915_READ(PP_OFF_DELAYS); - dev_priv->savePP_DIVISOR = I915_READ(PP_DIVISOR); - - /* FIXME: save TV & SDVO state */ - - /* FBC state */ - dev_priv->saveFBC_CFB_BASE = I915_READ(FBC_CFB_BASE); - dev_priv->saveFBC_LL_BASE = I915_READ(FBC_LL_BASE); - dev_priv->saveFBC_CONTROL2 = I915_READ(FBC_CONTROL2); - dev_priv->saveFBC_CONTROL = I915_READ(FBC_CONTROL); - - /* Interrupt state */ - dev_priv->saveIIR = I915_READ(IIR); - dev_priv->saveIER = I915_READ(IER); - dev_priv->saveIMR = I915_READ(IMR); - - /* VGA state */ - dev_priv->saveVGA0 = I915_READ(VGA0); - dev_priv->saveVGA1 = I915_READ(VGA1); - dev_priv->saveVGA_PD = I915_READ(VGA_PD); - dev_priv->saveVGACNTRL = I915_READ(VGACNTRL); - - /* Clock gating state */ - dev_priv->saveD_STATE = I915_READ(D_STATE); - dev_priv->saveCG_2D_DIS = I915_READ(CG_2D_DIS); - - /* Cache mode state */ - dev_priv->saveCACHE_MODE_0 = I915_READ(CACHE_MODE_0); - - /* Memory Arbitration state */ - dev_priv->saveMI_ARB_STATE = I915_READ(MI_ARB_STATE); - - /* Scratch space */ - for (i = 0; i < 16; i++) { - dev_priv->saveSWF0[i] = I915_READ(SWF00 + (i << 2)); - dev_priv->saveSWF1[i] = I915_READ(SWF10 + (i << 2)); - } - for (i = 0; i < 3; i++) - dev_priv->saveSWF2[i] = I915_READ(SWF30 + (i << 2)); - - i915_save_vga(dev); + i915_save_state(dev); intel_opregion_free(dev); @@ -384,155 +68,13 @@ static int i915_suspend(struct drm_device *dev, pm_message_t state) static int i915_resume(struct drm_device *dev) { - struct drm_i915_private *dev_priv = dev->dev_private; - int i; - pci_set_power_state(dev->pdev, PCI_D0); pci_restore_state(dev->pdev); if (pci_enable_device(dev->pdev)) return -1; pci_set_master(dev->pdev); - pci_write_config_byte(dev->pdev, LBB, dev_priv->saveLBB); - - I915_WRITE(DSPARB, dev_priv->saveDSPARB); - - /* Pipe & plane A info */ - /* Prime the clock */ - if (dev_priv->saveDPLL_A & DPLL_VCO_ENABLE) { - I915_WRITE(DPLL_A, dev_priv->saveDPLL_A & - ~DPLL_VCO_ENABLE); - udelay(150); - } - I915_WRITE(FPA0, dev_priv->saveFPA0); - I915_WRITE(FPA1, dev_priv->saveFPA1); - /* Actually enable it */ - I915_WRITE(DPLL_A, dev_priv->saveDPLL_A); - udelay(150); - if (IS_I965G(dev)) - I915_WRITE(DPLL_A_MD, dev_priv->saveDPLL_A_MD); - udelay(150); - - /* Restore mode */ - I915_WRITE(HTOTAL_A, dev_priv->saveHTOTAL_A); - I915_WRITE(HBLANK_A, dev_priv->saveHBLANK_A); - I915_WRITE(HSYNC_A, dev_priv->saveHSYNC_A); - I915_WRITE(VTOTAL_A, dev_priv->saveVTOTAL_A); - I915_WRITE(VBLANK_A, dev_priv->saveVBLANK_A); - I915_WRITE(VSYNC_A, dev_priv->saveVSYNC_A); - I915_WRITE(BCLRPAT_A, dev_priv->saveBCLRPAT_A); - - /* Restore plane info */ - I915_WRITE(DSPASIZE, dev_priv->saveDSPASIZE); - I915_WRITE(DSPAPOS, dev_priv->saveDSPAPOS); - I915_WRITE(PIPEASRC, dev_priv->savePIPEASRC); - I915_WRITE(DSPAADDR, dev_priv->saveDSPAADDR); - I915_WRITE(DSPASTRIDE, dev_priv->saveDSPASTRIDE); - if (IS_I965G(dev)) { - I915_WRITE(DSPASURF, dev_priv->saveDSPASURF); - I915_WRITE(DSPATILEOFF, dev_priv->saveDSPATILEOFF); - } - - I915_WRITE(PIPEACONF, dev_priv->savePIPEACONF); - - i915_restore_palette(dev, PIPE_A); - /* Enable the plane */ - I915_WRITE(DSPACNTR, dev_priv->saveDSPACNTR); - I915_WRITE(DSPAADDR, I915_READ(DSPAADDR)); - - /* Pipe & plane B info */ - if (dev_priv->saveDPLL_B & DPLL_VCO_ENABLE) { - I915_WRITE(DPLL_B, dev_priv->saveDPLL_B & - ~DPLL_VCO_ENABLE); - udelay(150); - } - I915_WRITE(FPB0, dev_priv->saveFPB0); - I915_WRITE(FPB1, dev_priv->saveFPB1); - /* Actually enable it */ - I915_WRITE(DPLL_B, dev_priv->saveDPLL_B); - udelay(150); - if (IS_I965G(dev)) - I915_WRITE(DPLL_B_MD, dev_priv->saveDPLL_B_MD); - udelay(150); - - /* Restore mode */ - I915_WRITE(HTOTAL_B, dev_priv->saveHTOTAL_B); - I915_WRITE(HBLANK_B, dev_priv->saveHBLANK_B); - I915_WRITE(HSYNC_B, dev_priv->saveHSYNC_B); - I915_WRITE(VTOTAL_B, dev_priv->saveVTOTAL_B); - I915_WRITE(VBLANK_B, dev_priv->saveVBLANK_B); - I915_WRITE(VSYNC_B, dev_priv->saveVSYNC_B); - I915_WRITE(BCLRPAT_B, dev_priv->saveBCLRPAT_B); - - /* Restore plane info */ - I915_WRITE(DSPBSIZE, dev_priv->saveDSPBSIZE); - I915_WRITE(DSPBPOS, dev_priv->saveDSPBPOS); - I915_WRITE(PIPEBSRC, dev_priv->savePIPEBSRC); - I915_WRITE(DSPBADDR, dev_priv->saveDSPBADDR); - I915_WRITE(DSPBSTRIDE, dev_priv->saveDSPBSTRIDE); - if (IS_I965G(dev)) { - I915_WRITE(DSPBSURF, dev_priv->saveDSPBSURF); - I915_WRITE(DSPBTILEOFF, dev_priv->saveDSPBTILEOFF); - } - - I915_WRITE(PIPEBCONF, dev_priv->savePIPEBCONF); - - i915_restore_palette(dev, PIPE_B); - /* Enable the plane */ - I915_WRITE(DSPBCNTR, dev_priv->saveDSPBCNTR); - I915_WRITE(DSPBADDR, I915_READ(DSPBADDR)); - - /* CRT state */ - I915_WRITE(ADPA, dev_priv->saveADPA); - - /* LVDS state */ - if (IS_I965G(dev)) - I915_WRITE(BLC_PWM_CTL2, dev_priv->saveBLC_PWM_CTL2); - if (IS_MOBILE(dev) && !IS_I830(dev)) - I915_WRITE(LVDS, dev_priv->saveLVDS); - if (!IS_I830(dev) && !IS_845G(dev)) - I915_WRITE(PFIT_CONTROL, dev_priv->savePFIT_CONTROL); - - I915_WRITE(PFIT_PGM_RATIOS, dev_priv->savePFIT_PGM_RATIOS); - I915_WRITE(BLC_PWM_CTL, dev_priv->saveBLC_PWM_CTL); - I915_WRITE(PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS); - I915_WRITE(PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS); - I915_WRITE(PP_DIVISOR, dev_priv->savePP_DIVISOR); - I915_WRITE(PP_CONTROL, dev_priv->savePP_CONTROL); - - /* FIXME: restore TV & SDVO state */ - - /* FBC info */ - I915_WRITE(FBC_CFB_BASE, dev_priv->saveFBC_CFB_BASE); - I915_WRITE(FBC_LL_BASE, dev_priv->saveFBC_LL_BASE); - I915_WRITE(FBC_CONTROL2, dev_priv->saveFBC_CONTROL2); - I915_WRITE(FBC_CONTROL, dev_priv->saveFBC_CONTROL); - - /* VGA state */ - I915_WRITE(VGACNTRL, dev_priv->saveVGACNTRL); - I915_WRITE(VGA0, dev_priv->saveVGA0); - I915_WRITE(VGA1, dev_priv->saveVGA1); - I915_WRITE(VGA_PD, dev_priv->saveVGA_PD); - udelay(150); - - /* Clock gating state */ - I915_WRITE (D_STATE, dev_priv->saveD_STATE); - I915_WRITE(CG_2D_DIS, dev_priv->saveCG_2D_DIS); - - /* Cache mode state */ - I915_WRITE (CACHE_MODE_0, dev_priv->saveCACHE_MODE_0 | 0xffff0000); - - /* Memory arbitration state */ - I915_WRITE (MI_ARB_STATE, dev_priv->saveMI_ARB_STATE | 0xffff0000); - - for (i = 0; i < 16; i++) { - I915_WRITE(SWF00 + (i << 2), dev_priv->saveSWF0[i]); - I915_WRITE(SWF10 + (i << 2), dev_priv->saveSWF1[i+7]); - } - for (i = 0; i < 3; i++) - I915_WRITE(SWF30 + (i << 2), dev_priv->saveSWF2[i]); - - i915_restore_vga(dev); + i915_restore_state(dev); intel_opregion_init(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index e4bd01c511a..a82b487de7b 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -41,6 +41,11 @@ #define DRIVER_DESC "Intel Graphics" #define DRIVER_DATE "20060119" +enum pipe { + PIPE_A = 0, + PIPE_B, +}; + /* Interface history: * * 1.1: Original. @@ -269,6 +274,10 @@ extern void i915_mem_takedown(struct mem_block **heap); extern void i915_mem_release(struct drm_device * dev, struct drm_file *file_priv, struct mem_block *heap); +/* i915_suspend.c */ +extern int i915_save_state(struct drm_device *dev); +extern int i915_restore_state(struct drm_device *dev); + /* i915_opregion.c */ extern int intel_opregion_init(struct drm_device *dev); extern void intel_opregion_free(struct drm_device *dev); @@ -279,6 +288,8 @@ extern void opregion_enable_asle(struct drm_device *dev); #define I915_WRITE(reg,val) DRM_WRITE32(dev_priv->mmio_map, (reg), (val)) #define I915_READ16(reg) DRM_READ16(dev_priv->mmio_map, (reg)) #define I915_WRITE16(reg,val) DRM_WRITE16(dev_priv->mmio_map, (reg), (val)) +#define I915_READ8(reg) DRM_READ8(dev_priv->mmio_map, (reg)) +#define I915_WRITE8(reg,val) DRM_WRITE8(dev_priv->mmio_map, (reg), (val)) #define I915_VERBOSE 0 diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c new file mode 100644 index 00000000000..e0c1fe4d1b8 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_suspend.c @@ -0,0 +1,509 @@ +/* + * + * Copyright 2008 (c) Intel Corporation + * Jesse Barnes + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "drmP.h" +#include "drm.h" +#include "i915_drm.h" +#include "i915_drv.h" + +static bool i915_pipe_enabled(struct drm_device *dev, enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + if (pipe == PIPE_A) + return (I915_READ(DPLL_A) & DPLL_VCO_ENABLE); + else + return (I915_READ(DPLL_B) & DPLL_VCO_ENABLE); +} + +static void i915_save_palette(struct drm_device *dev, enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long reg = (pipe == PIPE_A ? PALETTE_A : PALETTE_B); + u32 *array; + int i; + + if (!i915_pipe_enabled(dev, pipe)) + return; + + if (pipe == PIPE_A) + array = dev_priv->save_palette_a; + else + array = dev_priv->save_palette_b; + + for(i = 0; i < 256; i++) + array[i] = I915_READ(reg + (i << 2)); +} + +static void i915_restore_palette(struct drm_device *dev, enum pipe pipe) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + unsigned long reg = (pipe == PIPE_A ? PALETTE_A : PALETTE_B); + u32 *array; + int i; + + if (!i915_pipe_enabled(dev, pipe)) + return; + + if (pipe == PIPE_A) + array = dev_priv->save_palette_a; + else + array = dev_priv->save_palette_b; + + for(i = 0; i < 256; i++) + I915_WRITE(reg + (i << 2), array[i]); +} + +static u8 i915_read_indexed(struct drm_device *dev, u16 index_port, u16 data_port, u8 reg) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_WRITE8(index_port, reg); + return I915_READ8(data_port); +} + +static u8 i915_read_ar(struct drm_device *dev, u16 st01, u8 reg, u16 palette_enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_READ8(st01); + I915_WRITE8(VGA_AR_INDEX, palette_enable | reg); + return I915_READ8(VGA_AR_DATA_READ); +} + +static void i915_write_ar(struct drm_device *dev, u16 st01, u8 reg, u8 val, u16 palette_enable) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_READ8(st01); + I915_WRITE8(VGA_AR_INDEX, palette_enable | reg); + I915_WRITE8(VGA_AR_DATA_WRITE, val); +} + +static void i915_write_indexed(struct drm_device *dev, u16 index_port, u16 data_port, u8 reg, u8 val) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + I915_WRITE8(index_port, reg); + I915_WRITE8(data_port, val); +} + +static void i915_save_vga(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + u16 cr_index, cr_data, st01; + + /* VGA color palette registers */ + dev_priv->saveDACMASK = I915_READ8(VGA_DACMASK); + /* DACCRX automatically increments during read */ + I915_WRITE8(VGA_DACRX, 0); + /* Read 3 bytes of color data from each index */ + for (i = 0; i < 256 * 3; i++) + dev_priv->saveDACDATA[i] = I915_READ8(VGA_DACDATA); + + /* MSR bits */ + dev_priv->saveMSR = I915_READ8(VGA_MSR_READ); + if (dev_priv->saveMSR & VGA_MSR_CGA_MODE) { + cr_index = VGA_CR_INDEX_CGA; + cr_data = VGA_CR_DATA_CGA; + st01 = VGA_ST01_CGA; + } else { + cr_index = VGA_CR_INDEX_MDA; + cr_data = VGA_CR_DATA_MDA; + st01 = VGA_ST01_MDA; + } + + /* CRT controller regs */ + i915_write_indexed(dev, cr_index, cr_data, 0x11, + i915_read_indexed(dev, cr_index, cr_data, 0x11) & + (~0x80)); + for (i = 0; i <= 0x24; i++) + dev_priv->saveCR[i] = + i915_read_indexed(dev, cr_index, cr_data, i); + /* Make sure we don't turn off CR group 0 writes */ + dev_priv->saveCR[0x11] &= ~0x80; + + /* Attribute controller registers */ + I915_READ8(st01); + dev_priv->saveAR_INDEX = I915_READ8(VGA_AR_INDEX); + for (i = 0; i <= 0x14; i++) + dev_priv->saveAR[i] = i915_read_ar(dev, st01, i, 0); + I915_READ8(st01); + I915_WRITE8(VGA_AR_INDEX, dev_priv->saveAR_INDEX); + I915_READ8(st01); + + /* Graphics controller registers */ + for (i = 0; i < 9; i++) + dev_priv->saveGR[i] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, i); + + dev_priv->saveGR[0x10] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x10); + dev_priv->saveGR[0x11] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x11); + dev_priv->saveGR[0x18] = + i915_read_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x18); + + /* Sequencer registers */ + for (i = 0; i < 8; i++) + dev_priv->saveSR[i] = + i915_read_indexed(dev, VGA_SR_INDEX, VGA_SR_DATA, i); +} + +static void i915_restore_vga(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + u16 cr_index, cr_data, st01; + + /* MSR bits */ + I915_WRITE8(VGA_MSR_WRITE, dev_priv->saveMSR); + if (dev_priv->saveMSR & VGA_MSR_CGA_MODE) { + cr_index = VGA_CR_INDEX_CGA; + cr_data = VGA_CR_DATA_CGA; + st01 = VGA_ST01_CGA; + } else { + cr_index = VGA_CR_INDEX_MDA; + cr_data = VGA_CR_DATA_MDA; + st01 = VGA_ST01_MDA; + } + + /* Sequencer registers, don't write SR07 */ + for (i = 0; i < 7; i++) + i915_write_indexed(dev, VGA_SR_INDEX, VGA_SR_DATA, i, + dev_priv->saveSR[i]); + + /* CRT controller regs */ + /* Enable CR group 0 writes */ + i915_write_indexed(dev, cr_index, cr_data, 0x11, dev_priv->saveCR[0x11]); + for (i = 0; i <= 0x24; i++) + i915_write_indexed(dev, cr_index, cr_data, i, dev_priv->saveCR[i]); + + /* Graphics controller regs */ + for (i = 0; i < 9; i++) + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, i, + dev_priv->saveGR[i]); + + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x10, + dev_priv->saveGR[0x10]); + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x11, + dev_priv->saveGR[0x11]); + i915_write_indexed(dev, VGA_GR_INDEX, VGA_GR_DATA, 0x18, + dev_priv->saveGR[0x18]); + + /* Attribute controller registers */ + I915_READ8(st01); /* switch back to index mode */ + for (i = 0; i <= 0x14; i++) + i915_write_ar(dev, st01, i, dev_priv->saveAR[i], 0); + I915_READ8(st01); /* switch back to index mode */ + I915_WRITE8(VGA_AR_INDEX, dev_priv->saveAR_INDEX | 0x20); + I915_READ8(st01); + + /* VGA color palette registers */ + I915_WRITE8(VGA_DACMASK, dev_priv->saveDACMASK); + /* DACCRX automatically increments during read */ + I915_WRITE8(VGA_DACWX, 0); + /* Read 3 bytes of color data from each index */ + for (i = 0; i < 256 * 3; i++) + I915_WRITE8(VGA_DACDATA, dev_priv->saveDACDATA[i]); + +} + +int i915_save_state(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + + pci_read_config_byte(dev->pdev, LBB, &dev_priv->saveLBB); + + /* Display arbitration control */ + dev_priv->saveDSPARB = I915_READ(DSPARB); + + /* Pipe & plane A info */ + dev_priv->savePIPEACONF = I915_READ(PIPEACONF); + dev_priv->savePIPEASRC = I915_READ(PIPEASRC); + dev_priv->saveFPA0 = I915_READ(FPA0); + dev_priv->saveFPA1 = I915_READ(FPA1); + dev_priv->saveDPLL_A = I915_READ(DPLL_A); + if (IS_I965G(dev)) + dev_priv->saveDPLL_A_MD = I915_READ(DPLL_A_MD); + dev_priv->saveHTOTAL_A = I915_READ(HTOTAL_A); + dev_priv->saveHBLANK_A = I915_READ(HBLANK_A); + dev_priv->saveHSYNC_A = I915_READ(HSYNC_A); + dev_priv->saveVTOTAL_A = I915_READ(VTOTAL_A); + dev_priv->saveVBLANK_A = I915_READ(VBLANK_A); + dev_priv->saveVSYNC_A = I915_READ(VSYNC_A); + dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A); + + dev_priv->saveDSPACNTR = I915_READ(DSPACNTR); + dev_priv->saveDSPASTRIDE = I915_READ(DSPASTRIDE); + dev_priv->saveDSPASIZE = I915_READ(DSPASIZE); + dev_priv->saveDSPAPOS = I915_READ(DSPAPOS); + dev_priv->saveDSPAADDR = I915_READ(DSPAADDR); + if (IS_I965G(dev)) { + dev_priv->saveDSPASURF = I915_READ(DSPASURF); + dev_priv->saveDSPATILEOFF = I915_READ(DSPATILEOFF); + } + i915_save_palette(dev, PIPE_A); + dev_priv->savePIPEASTAT = I915_READ(PIPEASTAT); + + /* Pipe & plane B info */ + dev_priv->savePIPEBCONF = I915_READ(PIPEBCONF); + dev_priv->savePIPEBSRC = I915_READ(PIPEBSRC); + dev_priv->saveFPB0 = I915_READ(FPB0); + dev_priv->saveFPB1 = I915_READ(FPB1); + dev_priv->saveDPLL_B = I915_READ(DPLL_B); + if (IS_I965G(dev)) + dev_priv->saveDPLL_B_MD = I915_READ(DPLL_B_MD); + dev_priv->saveHTOTAL_B = I915_READ(HTOTAL_B); + dev_priv->saveHBLANK_B = I915_READ(HBLANK_B); + dev_priv->saveHSYNC_B = I915_READ(HSYNC_B); + dev_priv->saveVTOTAL_B = I915_READ(VTOTAL_B); + dev_priv->saveVBLANK_B = I915_READ(VBLANK_B); + dev_priv->saveVSYNC_B = I915_READ(VSYNC_B); + dev_priv->saveBCLRPAT_A = I915_READ(BCLRPAT_A); + + dev_priv->saveDSPBCNTR = I915_READ(DSPBCNTR); + dev_priv->saveDSPBSTRIDE = I915_READ(DSPBSTRIDE); + dev_priv->saveDSPBSIZE = I915_READ(DSPBSIZE); + dev_priv->saveDSPBPOS = I915_READ(DSPBPOS); + dev_priv->saveDSPBADDR = I915_READ(DSPBADDR); + if (IS_I965GM(dev) || IS_IGD_GM(dev)) { + dev_priv->saveDSPBSURF = I915_READ(DSPBSURF); + dev_priv->saveDSPBTILEOFF = I915_READ(DSPBTILEOFF); + } + i915_save_palette(dev, PIPE_B); + dev_priv->savePIPEBSTAT = I915_READ(PIPEBSTAT); + + /* CRT state */ + dev_priv->saveADPA = I915_READ(ADPA); + + /* LVDS state */ + dev_priv->savePP_CONTROL = I915_READ(PP_CONTROL); + dev_priv->savePFIT_PGM_RATIOS = I915_READ(PFIT_PGM_RATIOS); + dev_priv->saveBLC_PWM_CTL = I915_READ(BLC_PWM_CTL); + if (IS_I965G(dev)) + dev_priv->saveBLC_PWM_CTL2 = I915_READ(BLC_PWM_CTL2); + if (IS_MOBILE(dev) && !IS_I830(dev)) + dev_priv->saveLVDS = I915_READ(LVDS); + if (!IS_I830(dev) && !IS_845G(dev)) + dev_priv->savePFIT_CONTROL = I915_READ(PFIT_CONTROL); + dev_priv->savePP_ON_DELAYS = I915_READ(PP_ON_DELAYS); + dev_priv->savePP_OFF_DELAYS = I915_READ(PP_OFF_DELAYS); + dev_priv->savePP_DIVISOR = I915_READ(PP_DIVISOR); + + /* FIXME: save TV & SDVO state */ + + /* FBC state */ + dev_priv->saveFBC_CFB_BASE = I915_READ(FBC_CFB_BASE); + dev_priv->saveFBC_LL_BASE = I915_READ(FBC_LL_BASE); + dev_priv->saveFBC_CONTROL2 = I915_READ(FBC_CONTROL2); + dev_priv->saveFBC_CONTROL = I915_READ(FBC_CONTROL); + + /* Interrupt state */ + dev_priv->saveIIR = I915_READ(IIR); + dev_priv->saveIER = I915_READ(IER); + dev_priv->saveIMR = I915_READ(IMR); + + /* VGA state */ + dev_priv->saveVGA0 = I915_READ(VGA0); + dev_priv->saveVGA1 = I915_READ(VGA1); + dev_priv->saveVGA_PD = I915_READ(VGA_PD); + dev_priv->saveVGACNTRL = I915_READ(VGACNTRL); + + /* Clock gating state */ + dev_priv->saveD_STATE = I915_READ(D_STATE); + dev_priv->saveCG_2D_DIS = I915_READ(CG_2D_DIS); + + /* Cache mode state */ + dev_priv->saveCACHE_MODE_0 = I915_READ(CACHE_MODE_0); + + /* Memory Arbitration state */ + dev_priv->saveMI_ARB_STATE = I915_READ(MI_ARB_STATE); + + /* Scratch space */ + for (i = 0; i < 16; i++) { + dev_priv->saveSWF0[i] = I915_READ(SWF00 + (i << 2)); + dev_priv->saveSWF1[i] = I915_READ(SWF10 + (i << 2)); + } + for (i = 0; i < 3; i++) + dev_priv->saveSWF2[i] = I915_READ(SWF30 + (i << 2)); + + i915_save_vga(dev); + + return 0; +} + +int i915_restore_state(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + int i; + + pci_write_config_byte(dev->pdev, LBB, dev_priv->saveLBB); + + I915_WRITE(DSPARB, dev_priv->saveDSPARB); + + /* Pipe & plane A info */ + /* Prime the clock */ + if (dev_priv->saveDPLL_A & DPLL_VCO_ENABLE) { + I915_WRITE(DPLL_A, dev_priv->saveDPLL_A & + ~DPLL_VCO_ENABLE); + DRM_UDELAY(150); + } + I915_WRITE(FPA0, dev_priv->saveFPA0); + I915_WRITE(FPA1, dev_priv->saveFPA1); + /* Actually enable it */ + I915_WRITE(DPLL_A, dev_priv->saveDPLL_A); + DRM_UDELAY(150); + if (IS_I965G(dev)) + I915_WRITE(DPLL_A_MD, dev_priv->saveDPLL_A_MD); + DRM_UDELAY(150); + + /* Restore mode */ + I915_WRITE(HTOTAL_A, dev_priv->saveHTOTAL_A); + I915_WRITE(HBLANK_A, dev_priv->saveHBLANK_A); + I915_WRITE(HSYNC_A, dev_priv->saveHSYNC_A); + I915_WRITE(VTOTAL_A, dev_priv->saveVTOTAL_A); + I915_WRITE(VBLANK_A, dev_priv->saveVBLANK_A); + I915_WRITE(VSYNC_A, dev_priv->saveVSYNC_A); + I915_WRITE(BCLRPAT_A, dev_priv->saveBCLRPAT_A); + + /* Restore plane info */ + I915_WRITE(DSPASIZE, dev_priv->saveDSPASIZE); + I915_WRITE(DSPAPOS, dev_priv->saveDSPAPOS); + I915_WRITE(PIPEASRC, dev_priv->savePIPEASRC); + I915_WRITE(DSPAADDR, dev_priv->saveDSPAADDR); + I915_WRITE(DSPASTRIDE, dev_priv->saveDSPASTRIDE); + if (IS_I965G(dev)) { + I915_WRITE(DSPASURF, dev_priv->saveDSPASURF); + I915_WRITE(DSPATILEOFF, dev_priv->saveDSPATILEOFF); + } + + I915_WRITE(PIPEACONF, dev_priv->savePIPEACONF); + + i915_restore_palette(dev, PIPE_A); + /* Enable the plane */ + I915_WRITE(DSPACNTR, dev_priv->saveDSPACNTR); + I915_WRITE(DSPAADDR, I915_READ(DSPAADDR)); + + /* Pipe & plane B info */ + if (dev_priv->saveDPLL_B & DPLL_VCO_ENABLE) { + I915_WRITE(DPLL_B, dev_priv->saveDPLL_B & + ~DPLL_VCO_ENABLE); + DRM_UDELAY(150); + } + I915_WRITE(FPB0, dev_priv->saveFPB0); + I915_WRITE(FPB1, dev_priv->saveFPB1); + /* Actually enable it */ + I915_WRITE(DPLL_B, dev_priv->saveDPLL_B); + DRM_UDELAY(150); + if (IS_I965G(dev)) + I915_WRITE(DPLL_B_MD, dev_priv->saveDPLL_B_MD); + DRM_UDELAY(150); + + /* Restore mode */ + I915_WRITE(HTOTAL_B, dev_priv->saveHTOTAL_B); + I915_WRITE(HBLANK_B, dev_priv->saveHBLANK_B); + I915_WRITE(HSYNC_B, dev_priv->saveHSYNC_B); + I915_WRITE(VTOTAL_B, dev_priv->saveVTOTAL_B); + I915_WRITE(VBLANK_B, dev_priv->saveVBLANK_B); + I915_WRITE(VSYNC_B, dev_priv->saveVSYNC_B); + I915_WRITE(BCLRPAT_B, dev_priv->saveBCLRPAT_B); + + /* Restore plane info */ + I915_WRITE(DSPBSIZE, dev_priv->saveDSPBSIZE); + I915_WRITE(DSPBPOS, dev_priv->saveDSPBPOS); + I915_WRITE(PIPEBSRC, dev_priv->savePIPEBSRC); + I915_WRITE(DSPBADDR, dev_priv->saveDSPBADDR); + I915_WRITE(DSPBSTRIDE, dev_priv->saveDSPBSTRIDE); + if (IS_I965G(dev)) { + I915_WRITE(DSPBSURF, dev_priv->saveDSPBSURF); + I915_WRITE(DSPBTILEOFF, dev_priv->saveDSPBTILEOFF); + } + + I915_WRITE(PIPEBCONF, dev_priv->savePIPEBCONF); + + i915_restore_palette(dev, PIPE_B); + /* Enable the plane */ + I915_WRITE(DSPBCNTR, dev_priv->saveDSPBCNTR); + I915_WRITE(DSPBADDR, I915_READ(DSPBADDR)); + + /* CRT state */ + I915_WRITE(ADPA, dev_priv->saveADPA); + + /* LVDS state */ + if (IS_I965G(dev)) + I915_WRITE(BLC_PWM_CTL2, dev_priv->saveBLC_PWM_CTL2); + if (IS_MOBILE(dev) && !IS_I830(dev)) + I915_WRITE(LVDS, dev_priv->saveLVDS); + if (!IS_I830(dev) && !IS_845G(dev)) + I915_WRITE(PFIT_CONTROL, dev_priv->savePFIT_CONTROL); + + I915_WRITE(PFIT_PGM_RATIOS, dev_priv->savePFIT_PGM_RATIOS); + I915_WRITE(BLC_PWM_CTL, dev_priv->saveBLC_PWM_CTL); + I915_WRITE(PP_ON_DELAYS, dev_priv->savePP_ON_DELAYS); + I915_WRITE(PP_OFF_DELAYS, dev_priv->savePP_OFF_DELAYS); + I915_WRITE(PP_DIVISOR, dev_priv->savePP_DIVISOR); + I915_WRITE(PP_CONTROL, dev_priv->savePP_CONTROL); + + /* FIXME: restore TV & SDVO state */ + + /* FBC info */ + I915_WRITE(FBC_CFB_BASE, dev_priv->saveFBC_CFB_BASE); + I915_WRITE(FBC_LL_BASE, dev_priv->saveFBC_LL_BASE); + I915_WRITE(FBC_CONTROL2, dev_priv->saveFBC_CONTROL2); + I915_WRITE(FBC_CONTROL, dev_priv->saveFBC_CONTROL); + + /* VGA state */ + I915_WRITE(VGACNTRL, dev_priv->saveVGACNTRL); + I915_WRITE(VGA0, dev_priv->saveVGA0); + I915_WRITE(VGA1, dev_priv->saveVGA1); + I915_WRITE(VGA_PD, dev_priv->saveVGA_PD); + DRM_UDELAY(150); + + /* Clock gating state */ + I915_WRITE (D_STATE, dev_priv->saveD_STATE); + I915_WRITE (CG_2D_DIS, dev_priv->saveCG_2D_DIS); + + /* Cache mode state */ + I915_WRITE (CACHE_MODE_0, dev_priv->saveCACHE_MODE_0 | 0xffff0000); + + /* Memory arbitration state */ + I915_WRITE (MI_ARB_STATE, dev_priv->saveMI_ARB_STATE | 0xffff0000); + + for (i = 0; i < 16; i++) { + I915_WRITE(SWF00 + (i << 2), dev_priv->saveSWF0[i]); + I915_WRITE(SWF10 + (i << 2), dev_priv->saveSWF1[i+7]); + } + for (i = 0; i < 3; i++) + I915_WRITE(SWF30 + (i << 2), dev_priv->saveSWF2[i]); + + i915_restore_vga(dev); + + return 0; +} + -- cgit v1.2.3 From b9bfdfe6703eb089839d48316a79c84924a3c335 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Mon, 25 Aug 2008 15:16:19 -0700 Subject: new chip name is GM45 Author: Zhenyu Wang i915: official name for GM45 chipset Signed-off-by: Zhenyu Wang Acked-by: Jesse Barnes Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_drv.h | 6 +++--- drivers/gpu/drm/i915/i915_suspend.c | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index a82b487de7b..71326ca9367 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -364,7 +364,7 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); #define IS_I965GM(dev) ((dev)->pci_device == 0x2A02) -#define IS_IGD_GM(dev) ((dev)->pci_device == 0x2A42) +#define IS_GM45(dev) ((dev)->pci_device == 0x2A42) #define IS_G4X(dev) ((dev)->pci_device == 0x2E02 || \ (dev)->pci_device == 0x2E12 || \ @@ -378,9 +378,9 @@ extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); IS_I945GM(dev) || IS_I965G(dev) || IS_G33(dev)) #define IS_MOBILE(dev) (IS_I830(dev) || IS_I85X(dev) || IS_I915GM(dev) || \ - IS_I945GM(dev) || IS_I965GM(dev) || IS_IGD_GM(dev)) + IS_I945GM(dev) || IS_I965GM(dev) || IS_GM45(dev)) -#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_IGD_GM(dev) || IS_G4X(dev)) +#define I915_NEED_GFX_HWS(dev) (IS_G33(dev) || IS_GM45(dev) || IS_G4X(dev)) #define PRIMARY_RINGBUFFER_SIZE (128*1024) diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c index e0c1fe4d1b8..603fe742ccd 100644 --- a/drivers/gpu/drm/i915/i915_suspend.c +++ b/drivers/gpu/drm/i915/i915_suspend.c @@ -292,7 +292,7 @@ int i915_save_state(struct drm_device *dev) dev_priv->saveDSPBSIZE = I915_READ(DSPBSIZE); dev_priv->saveDSPBPOS = I915_READ(DSPBPOS); dev_priv->saveDSPBADDR = I915_READ(DSPBADDR); - if (IS_I965GM(dev) || IS_IGD_GM(dev)) { + if (IS_I965GM(dev) || IS_GM45(dev)) { dev_priv->saveDSPBSURF = I915_READ(DSPBSURF); dev_priv->saveDSPBTILEOFF = I915_READ(DSPBTILEOFF); } -- cgit v1.2.3 From 2df68b439fcb97a4c55f81516206ef4ee325e28d Mon Sep 17 00:00:00 2001 From: David Howells Date: Tue, 2 Sep 2008 11:03:14 +1000 Subject: drm/cred: wrap task credential accesses in the drm driver. Wrap access to task credentials so that they can be separated more easily from the task_struct during the introduction of COW creds. Change most current->(|e|s|fs)[ug]id to current_(|e|s|fs)[ug]id(). Change some task->e?[ug]id to task_e?[ug]id(). In some places it makes more sense to use RCU directly rather than a convenient wrapper; these will be addressed by later patches. Signed-off-by: David Howells Reviewed-by: James Morris Acked-by: Serge Hallyn Signed-off-by: David Airlie --- drivers/gpu/drm/drm_fops.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c index 851a53f1acc..dcf8b4dc954 100644 --- a/drivers/gpu/drm/drm_fops.c +++ b/drivers/gpu/drm/drm_fops.c @@ -246,7 +246,7 @@ static int drm_open_helper(struct inode *inode, struct file *filp, memset(priv, 0, sizeof(*priv)); filp->private_data = priv; priv->filp = filp; - priv->uid = current->euid; + priv->uid = current_euid(); priv->pid = task_pid_nr(current); priv->minor = idr_find(&drm_minors_idr, minor_id); priv->ioctl_count = 0; -- cgit v1.2.3 From 0a3e67a4caac273a3bfc4ced3da364830b1ab241 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Tue, 30 Sep 2008 12:14:26 -0700 Subject: drm: Rework vblank-wait handling to allow interrupt reduction. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Previously, drivers supporting vblank interrupt waits would run the interrupt all the time, or all the time that any 3d client was running, preventing the CPU from sleeping for long when the system was otherwise idle. Now, interrupts are disabled any time that no client is waiting on a vblank event. The new method uses vblank counters on the chipsets when the interrupts are turned off, rather than counting interrupts, so that we can continue to present accurate vblank numbers. Co-author: Michel Dänzer Signed-off-by: Jesse Barnes Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_drv.c | 2 + drivers/gpu/drm/drm_irq.c | 447 ++++++++++++++++++++++++++++++------ drivers/gpu/drm/i915/i915_dma.c | 5 +- drivers/gpu/drm/i915/i915_drv.c | 11 +- drivers/gpu/drm/i915/i915_drv.h | 20 +- drivers/gpu/drm/i915/i915_irq.c | 395 ++++++++++++++++++++----------- drivers/gpu/drm/mga/mga_drv.c | 29 +-- drivers/gpu/drm/mga/mga_drv.h | 6 +- drivers/gpu/drm/mga/mga_irq.c | 74 ++++-- drivers/gpu/drm/r128/r128_drv.c | 29 +-- drivers/gpu/drm/r128/r128_drv.h | 11 +- drivers/gpu/drm/r128/r128_irq.c | 55 +++-- drivers/gpu/drm/radeon/radeon_cp.c | 2 +- drivers/gpu/drm/radeon/radeon_drv.c | 32 ++- drivers/gpu/drm/radeon/radeon_drv.h | 27 ++- drivers/gpu/drm/radeon/radeon_irq.c | 268 +++++++++++++-------- drivers/gpu/drm/via/via_drv.c | 26 ++- drivers/gpu/drm/via/via_drv.h | 16 +- drivers/gpu/drm/via/via_irq.c | 102 ++++---- 19 files changed, 1095 insertions(+), 462 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index 452c2d866ec..fb45fe7aeb8 100644 --- a/drivers/gpu/drm/drm_drv.c +++ b/drivers/gpu/drm/drm_drv.c @@ -116,6 +116,8 @@ static struct drm_ioctl_desc drm_ioctls[] = { DRM_IOCTL_DEF(DRM_IOCTL_WAIT_VBLANK, drm_wait_vblank, 0), + DRM_IOCTL_DEF(DRM_IOCTL_MODESET_CTL, drm_modeset_ctl, 0), + DRM_IOCTL_DEF(DRM_IOCTL_UPDATE_DRAW, drm_update_drawable_info, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), }; diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index 61ed5158f78..d0c13d954f5 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -71,19 +71,131 @@ int drm_irq_by_busid(struct drm_device *dev, void *data, return 0; } +static void vblank_disable_fn(unsigned long arg) +{ + struct drm_device *dev = (struct drm_device *)arg; + unsigned long irqflags; + int i; + + if (!dev->vblank_disable_allowed) + return; + + for (i = 0; i < dev->num_crtcs; i++) { + spin_lock_irqsave(&dev->vbl_lock, irqflags); + if (atomic_read(&dev->vblank_refcount[i]) == 0 && + dev->vblank_enabled[i]) { + DRM_DEBUG("disabling vblank on crtc %d\n", i); + dev->last_vblank[i] = + dev->driver->get_vblank_counter(dev, i); + dev->driver->disable_vblank(dev, i); + dev->vblank_enabled[i] = 0; + } + spin_unlock_irqrestore(&dev->vbl_lock, irqflags); + } +} + +static void drm_vblank_cleanup(struct drm_device *dev) +{ + /* Bail if the driver didn't call drm_vblank_init() */ + if (dev->num_crtcs == 0) + return; + + del_timer(&dev->vblank_disable_timer); + + vblank_disable_fn((unsigned long)dev); + + drm_free(dev->vbl_queue, sizeof(*dev->vbl_queue) * dev->num_crtcs, + DRM_MEM_DRIVER); + drm_free(dev->vbl_sigs, sizeof(*dev->vbl_sigs) * dev->num_crtcs, + DRM_MEM_DRIVER); + drm_free(dev->_vblank_count, sizeof(*dev->_vblank_count) * + dev->num_crtcs, DRM_MEM_DRIVER); + drm_free(dev->vblank_refcount, sizeof(*dev->vblank_refcount) * + dev->num_crtcs, DRM_MEM_DRIVER); + drm_free(dev->vblank_enabled, sizeof(*dev->vblank_enabled) * + dev->num_crtcs, DRM_MEM_DRIVER); + drm_free(dev->last_vblank, sizeof(*dev->last_vblank) * dev->num_crtcs, + DRM_MEM_DRIVER); + drm_free(dev->vblank_inmodeset, sizeof(*dev->vblank_inmodeset) * + dev->num_crtcs, DRM_MEM_DRIVER); + + dev->num_crtcs = 0; +} + +int drm_vblank_init(struct drm_device *dev, int num_crtcs) +{ + int i, ret = -ENOMEM; + + setup_timer(&dev->vblank_disable_timer, vblank_disable_fn, + (unsigned long)dev); + spin_lock_init(&dev->vbl_lock); + atomic_set(&dev->vbl_signal_pending, 0); + dev->num_crtcs = num_crtcs; + + dev->vbl_queue = drm_alloc(sizeof(wait_queue_head_t) * num_crtcs, + DRM_MEM_DRIVER); + if (!dev->vbl_queue) + goto err; + + dev->vbl_sigs = drm_alloc(sizeof(struct list_head) * num_crtcs, + DRM_MEM_DRIVER); + if (!dev->vbl_sigs) + goto err; + + dev->_vblank_count = drm_alloc(sizeof(atomic_t) * num_crtcs, + DRM_MEM_DRIVER); + if (!dev->_vblank_count) + goto err; + + dev->vblank_refcount = drm_alloc(sizeof(atomic_t) * num_crtcs, + DRM_MEM_DRIVER); + if (!dev->vblank_refcount) + goto err; + + dev->vblank_enabled = drm_calloc(num_crtcs, sizeof(int), + DRM_MEM_DRIVER); + if (!dev->vblank_enabled) + goto err; + + dev->last_vblank = drm_calloc(num_crtcs, sizeof(u32), DRM_MEM_DRIVER); + if (!dev->last_vblank) + goto err; + + dev->vblank_inmodeset = drm_calloc(num_crtcs, sizeof(int), + DRM_MEM_DRIVER); + if (!dev->vblank_inmodeset) + goto err; + + /* Zero per-crtc vblank stuff */ + for (i = 0; i < num_crtcs; i++) { + init_waitqueue_head(&dev->vbl_queue[i]); + INIT_LIST_HEAD(&dev->vbl_sigs[i]); + atomic_set(&dev->_vblank_count[i], 0); + atomic_set(&dev->vblank_refcount[i], 0); + } + + dev->vblank_disable_allowed = 0; + + return 0; + +err: + drm_vblank_cleanup(dev); + return ret; +} +EXPORT_SYMBOL(drm_vblank_init); + /** * Install IRQ handler. * * \param dev DRM device. - * \param irq IRQ number. * - * Initializes the IRQ related data, and setups drm_device::vbl_queue. Installs the handler, calling the driver + * Initializes the IRQ related data. Installs the handler, calling the driver * \c drm_driver_irq_preinstall() and \c drm_driver_irq_postinstall() functions * before and after the installation. */ -static int drm_irq_install(struct drm_device * dev) +int drm_irq_install(struct drm_device *dev) { - int ret; + int ret = 0; unsigned long sh_flags = 0; if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ)) @@ -109,17 +221,6 @@ static int drm_irq_install(struct drm_device * dev) DRM_DEBUG("irq=%d\n", dev->pdev->irq); - if (drm_core_check_feature(dev, DRIVER_IRQ_VBL)) { - init_waitqueue_head(&dev->vbl_queue); - - spin_lock_init(&dev->vbl_lock); - - INIT_LIST_HEAD(&dev->vbl_sigs); - INIT_LIST_HEAD(&dev->vbl_sigs2); - - dev->vbl_pending = 0; - } - /* Before installing handler */ dev->driver->irq_preinstall(dev); @@ -141,10 +242,16 @@ static int drm_irq_install(struct drm_device * dev) } /* After installing handler */ - dev->driver->irq_postinstall(dev); + ret = dev->driver->irq_postinstall(dev); + if (ret < 0) { + mutex_lock(&dev->struct_mutex); + dev->irq_enabled = 0; + mutex_unlock(&dev->struct_mutex); + } - return 0; + return ret; } +EXPORT_SYMBOL(drm_irq_install); /** * Uninstall the IRQ handler. @@ -174,11 +281,12 @@ int drm_irq_uninstall(struct drm_device * dev) free_irq(dev->pdev->irq, dev); + drm_vblank_cleanup(dev); + dev->locked_tasklet_func = NULL; return 0; } - EXPORT_SYMBOL(drm_irq_uninstall); /** @@ -217,6 +325,174 @@ int drm_control(struct drm_device *dev, void *data, } } +/** + * drm_vblank_count - retrieve "cooked" vblank counter value + * @dev: DRM device + * @crtc: which counter to retrieve + * + * Fetches the "cooked" vblank count value that represents the number of + * vblank events since the system was booted, including lost events due to + * modesetting activity. + */ +u32 drm_vblank_count(struct drm_device *dev, int crtc) +{ + return atomic_read(&dev->_vblank_count[crtc]); +} +EXPORT_SYMBOL(drm_vblank_count); + +/** + * drm_update_vblank_count - update the master vblank counter + * @dev: DRM device + * @crtc: counter to update + * + * Call back into the driver to update the appropriate vblank counter + * (specified by @crtc). Deal with wraparound, if it occurred, and + * update the last read value so we can deal with wraparound on the next + * call if necessary. + * + * Only necessary when going from off->on, to account for frames we + * didn't get an interrupt for. + * + * Note: caller must hold dev->vbl_lock since this reads & writes + * device vblank fields. + */ +static void drm_update_vblank_count(struct drm_device *dev, int crtc) +{ + u32 cur_vblank, diff; + + /* + * Interrupts were disabled prior to this call, so deal with counter + * wrap if needed. + * NOTE! It's possible we lost a full dev->max_vblank_count events + * here if the register is small or we had vblank interrupts off for + * a long time. + */ + cur_vblank = dev->driver->get_vblank_counter(dev, crtc); + diff = cur_vblank - dev->last_vblank[crtc]; + if (cur_vblank < dev->last_vblank[crtc]) { + diff += dev->max_vblank_count; + + DRM_DEBUG("last_vblank[%d]=0x%x, cur_vblank=0x%x => diff=0x%x\n", + crtc, dev->last_vblank[crtc], cur_vblank, diff); + } + + DRM_DEBUG("enabling vblank interrupts on crtc %d, missed %d\n", + crtc, diff); + + atomic_add(diff, &dev->_vblank_count[crtc]); +} + +/** + * drm_vblank_get - get a reference count on vblank events + * @dev: DRM device + * @crtc: which CRTC to own + * + * Acquire a reference count on vblank events to avoid having them disabled + * while in use. + * + * RETURNS + * Zero on success, nonzero on failure. + */ +int drm_vblank_get(struct drm_device *dev, int crtc) +{ + unsigned long irqflags; + int ret = 0; + + spin_lock_irqsave(&dev->vbl_lock, irqflags); + /* Going from 0->1 means we have to enable interrupts again */ + if (atomic_add_return(1, &dev->vblank_refcount[crtc]) == 1 && + !dev->vblank_enabled[crtc]) { + ret = dev->driver->enable_vblank(dev, crtc); + DRM_DEBUG("enabling vblank on crtc %d, ret: %d\n", crtc, ret); + if (ret) + atomic_dec(&dev->vblank_refcount[crtc]); + else { + dev->vblank_enabled[crtc] = 1; + drm_update_vblank_count(dev, crtc); + } + } + spin_unlock_irqrestore(&dev->vbl_lock, irqflags); + + return ret; +} +EXPORT_SYMBOL(drm_vblank_get); + +/** + * drm_vblank_put - give up ownership of vblank events + * @dev: DRM device + * @crtc: which counter to give up + * + * Release ownership of a given vblank counter, turning off interrupts + * if possible. + */ +void drm_vblank_put(struct drm_device *dev, int crtc) +{ + /* Last user schedules interrupt disable */ + if (atomic_dec_and_test(&dev->vblank_refcount[crtc])) + mod_timer(&dev->vblank_disable_timer, jiffies + 5*DRM_HZ); +} +EXPORT_SYMBOL(drm_vblank_put); + +/** + * drm_modeset_ctl - handle vblank event counter changes across mode switch + * @DRM_IOCTL_ARGS: standard ioctl arguments + * + * Applications should call the %_DRM_PRE_MODESET and %_DRM_POST_MODESET + * ioctls around modesetting so that any lost vblank events are accounted for. + * + * Generally the counter will reset across mode sets. If interrupts are + * enabled around this call, we don't have to do anything since the counter + * will have already been incremented. + */ +int drm_modeset_ctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_modeset_ctl *modeset = data; + unsigned long irqflags; + int crtc, ret = 0; + + /* If drm_vblank_init() hasn't been called yet, just no-op */ + if (!dev->num_crtcs) + goto out; + + crtc = modeset->crtc; + if (crtc >= dev->num_crtcs) { + ret = -EINVAL; + goto out; + } + + /* + * To avoid all the problems that might happen if interrupts + * were enabled/disabled around or between these calls, we just + * have the kernel take a reference on the CRTC (just once though + * to avoid corrupting the count if multiple, mismatch calls occur), + * so that interrupts remain enabled in the interim. + */ + switch (modeset->cmd) { + case _DRM_PRE_MODESET: + if (!dev->vblank_inmodeset[crtc]) { + dev->vblank_inmodeset[crtc] = 1; + drm_vblank_get(dev, crtc); + } + break; + case _DRM_POST_MODESET: + if (dev->vblank_inmodeset[crtc]) { + spin_lock_irqsave(&dev->vbl_lock, irqflags); + dev->vblank_disable_allowed = 1; + dev->vblank_inmodeset[crtc] = 0; + spin_unlock_irqrestore(&dev->vbl_lock, irqflags); + drm_vblank_put(dev, crtc); + } + break; + default: + ret = -EINVAL; + break; + } + +out: + return ret; +} + /** * Wait for VBLANK. * @@ -236,12 +512,12 @@ int drm_control(struct drm_device *dev, void *data, * * If a signal is not requested, then calls vblank_wait(). */ -int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_priv) +int drm_wait_vblank(struct drm_device *dev, void *data, + struct drm_file *file_priv) { union drm_wait_vblank *vblwait = data; - struct timeval now; int ret = 0; - unsigned int flags, seq; + unsigned int flags, seq, crtc; if ((!dev->pdev->irq) || (!dev->irq_enabled)) return -EINVAL; @@ -255,13 +531,17 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr } flags = vblwait->request.type & _DRM_VBLANK_FLAGS_MASK; + crtc = flags & _DRM_VBLANK_SECONDARY ? 1 : 0; - if (!drm_core_check_feature(dev, (flags & _DRM_VBLANK_SECONDARY) ? - DRIVER_IRQ_VBL2 : DRIVER_IRQ_VBL)) + if (crtc >= dev->num_crtcs) return -EINVAL; - seq = atomic_read((flags & _DRM_VBLANK_SECONDARY) ? &dev->vbl_received2 - : &dev->vbl_received); + ret = drm_vblank_get(dev, crtc); + if (ret) { + DRM_ERROR("failed to acquire vblank counter, %d\n", ret); + return ret; + } + seq = drm_vblank_count(dev, crtc); switch (vblwait->request.type & _DRM_VBLANK_TYPES_MASK) { case _DRM_VBLANK_RELATIVE: @@ -270,7 +550,8 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr case _DRM_VBLANK_ABSOLUTE: break; default: - return -EINVAL; + ret = -EINVAL; + goto done; } if ((flags & _DRM_VBLANK_NEXTONMISS) && @@ -280,8 +561,7 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr if (flags & _DRM_VBLANK_SIGNAL) { unsigned long irqflags; - struct list_head *vbl_sigs = (flags & _DRM_VBLANK_SECONDARY) - ? &dev->vbl_sigs2 : &dev->vbl_sigs; + struct list_head *vbl_sigs = &dev->vbl_sigs[crtc]; struct drm_vbl_sig *vbl_sig; spin_lock_irqsave(&dev->vbl_lock, irqflags); @@ -302,22 +582,29 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr } } - if (dev->vbl_pending >= 100) { + if (atomic_read(&dev->vbl_signal_pending) >= 100) { spin_unlock_irqrestore(&dev->vbl_lock, irqflags); - return -EBUSY; + ret = -EBUSY; + goto done; } - dev->vbl_pending++; - spin_unlock_irqrestore(&dev->vbl_lock, irqflags); - if (! - (vbl_sig = - drm_alloc(sizeof(struct drm_vbl_sig), DRM_MEM_DRIVER))) { - return -ENOMEM; + vbl_sig = drm_calloc(1, sizeof(struct drm_vbl_sig), + DRM_MEM_DRIVER); + if (!vbl_sig) { + ret = -ENOMEM; + goto done; + } + + ret = drm_vblank_get(dev, crtc); + if (ret) { + drm_free(vbl_sig, sizeof(struct drm_vbl_sig), + DRM_MEM_DRIVER); + return ret; } - memset((void *)vbl_sig, 0, sizeof(*vbl_sig)); + atomic_inc(&dev->vbl_signal_pending); vbl_sig->sequence = vblwait->request.sequence; vbl_sig->info.si_signo = vblwait->request.signal; @@ -331,20 +618,29 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr vblwait->reply.sequence = seq; } else { - if (flags & _DRM_VBLANK_SECONDARY) { - if (dev->driver->vblank_wait2) - ret = dev->driver->vblank_wait2(dev, &vblwait->request.sequence); - } else if (dev->driver->vblank_wait) - ret = - dev->driver->vblank_wait(dev, - &vblwait->request.sequence); - - do_gettimeofday(&now); - vblwait->reply.tval_sec = now.tv_sec; - vblwait->reply.tval_usec = now.tv_usec; + DRM_DEBUG("waiting on vblank count %d, crtc %d\n", + vblwait->request.sequence, crtc); + DRM_WAIT_ON(ret, dev->vbl_queue[crtc], 3 * DRM_HZ, + ((drm_vblank_count(dev, crtc) + - vblwait->request.sequence) <= (1 << 23))); + + if (ret != -EINTR) { + struct timeval now; + + do_gettimeofday(&now); + + vblwait->reply.tval_sec = now.tv_sec; + vblwait->reply.tval_usec = now.tv_usec; + vblwait->reply.sequence = drm_vblank_count(dev, crtc); + DRM_DEBUG("returning %d to client\n", + vblwait->reply.sequence); + } else { + DRM_DEBUG("vblank wait interrupted by signal\n"); + } } - done: +done: + drm_vblank_put(dev, crtc); return ret; } @@ -352,44 +648,57 @@ int drm_wait_vblank(struct drm_device *dev, void *data, struct drm_file *file_pr * Send the VBLANK signals. * * \param dev DRM device. + * \param crtc CRTC where the vblank event occurred * * Sends a signal for each task in drm_device::vbl_sigs and empties the list. * * If a signal is not requested, then calls vblank_wait(). */ -void drm_vbl_send_signals(struct drm_device * dev) +static void drm_vbl_send_signals(struct drm_device *dev, int crtc) { + struct drm_vbl_sig *vbl_sig, *tmp; + struct list_head *vbl_sigs; + unsigned int vbl_seq; unsigned long flags; - int i; spin_lock_irqsave(&dev->vbl_lock, flags); - for (i = 0; i < 2; i++) { - struct drm_vbl_sig *vbl_sig, *tmp; - struct list_head *vbl_sigs = i ? &dev->vbl_sigs2 : &dev->vbl_sigs; - unsigned int vbl_seq = atomic_read(i ? &dev->vbl_received2 : - &dev->vbl_received); + vbl_sigs = &dev->vbl_sigs[crtc]; + vbl_seq = drm_vblank_count(dev, crtc); - list_for_each_entry_safe(vbl_sig, tmp, vbl_sigs, head) { - if ((vbl_seq - vbl_sig->sequence) <= (1 << 23)) { - vbl_sig->info.si_code = vbl_seq; - send_sig_info(vbl_sig->info.si_signo, - &vbl_sig->info, vbl_sig->task); + list_for_each_entry_safe(vbl_sig, tmp, vbl_sigs, head) { + if ((vbl_seq - vbl_sig->sequence) <= (1 << 23)) { + vbl_sig->info.si_code = vbl_seq; + send_sig_info(vbl_sig->info.si_signo, + &vbl_sig->info, vbl_sig->task); - list_del(&vbl_sig->head); - - drm_free(vbl_sig, sizeof(*vbl_sig), - DRM_MEM_DRIVER); + list_del(&vbl_sig->head); - dev->vbl_pending--; - } - } + drm_free(vbl_sig, sizeof(*vbl_sig), + DRM_MEM_DRIVER); + atomic_dec(&dev->vbl_signal_pending); + drm_vblank_put(dev, crtc); + } } spin_unlock_irqrestore(&dev->vbl_lock, flags); } -EXPORT_SYMBOL(drm_vbl_send_signals); +/** + * drm_handle_vblank - handle a vblank event + * @dev: DRM device + * @crtc: where this event occurred + * + * Drivers should call this routine in their vblank interrupt handlers to + * update the vblank counter and send any signals that may be pending. + */ +void drm_handle_vblank(struct drm_device *dev, int crtc) +{ + atomic_inc(&dev->_vblank_count[crtc]); + DRM_WAKEUP(&dev->vbl_queue[crtc]); + drm_vbl_send_signals(dev, crtc); +} +EXPORT_SYMBOL(drm_handle_vblank); /** * Tasklet wrapper function. diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 29115450ee4..63c6803d471 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -673,7 +673,7 @@ static int i915_getparam(struct drm_device *dev, void *data, switch (param->param) { case I915_PARAM_IRQ_ACTIVE: - value = dev->irq_enabled; + value = dev->pdev->irq ? 1 : 0; break; case I915_PARAM_ALLOW_BATCHBUFFER: value = dev_priv->allow_batchbuffer ? 1 : 0; @@ -808,7 +808,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) * and the registers being closely associated. */ if (!IS_I945G(dev) && !IS_I945GM(dev)) - pci_enable_msi(dev->pdev); + if (pci_enable_msi(dev->pdev)) + DRM_ERROR("failed to enable MSI\n"); intel_opregion_init(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index eff66ed3e58..37af03f4db3 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -85,10 +85,8 @@ static struct drm_driver driver = { /* don't use mtrr's here, the Xserver or user space app should * deal with them for intel hardware. */ - .driver_features = - DRIVER_USE_AGP | DRIVER_REQUIRE_AGP | /* DRIVER_USE_MTRR |*/ - DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_IRQ_VBL | - DRIVER_IRQ_VBL2, + .driver_features = DRIVER_USE_AGP | DRIVER_REQUIRE_AGP | + DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED, .load = i915_driver_load, .unload = i915_driver_unload, .lastclose = i915_driver_lastclose, @@ -96,8 +94,9 @@ static struct drm_driver driver = { .suspend = i915_suspend, .resume = i915_resume, .device_is_agp = i915_driver_device_is_agp, - .vblank_wait = i915_driver_vblank_wait, - .vblank_wait2 = i915_driver_vblank_wait2, + .get_vblank_counter = i915_get_vblank_counter, + .enable_vblank = i915_enable_vblank, + .disable_vblank = i915_disable_vblank, .irq_preinstall = i915_driver_irq_preinstall, .irq_postinstall = i915_driver_irq_postinstall, .irq_uninstall = i915_driver_irq_uninstall, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 71326ca9367..d1a02bead45 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -83,10 +83,15 @@ struct mem_block { typedef struct _drm_i915_vbl_swap { struct list_head head; drm_drawable_t drw_id; - unsigned int pipe; + unsigned int plane; unsigned int sequence; } drm_i915_vbl_swap_t; +struct opregion_header; +struct opregion_acpi; +struct opregion_swsci; +struct opregion_asle; + struct intel_opregion { struct opregion_header *header; struct opregion_acpi *acpi; @@ -105,7 +110,7 @@ typedef struct drm_i915_private { drm_dma_handle_t *status_page_dmah; void *hw_status_page; dma_addr_t dma_status_page; - unsigned long counter; + uint32_t counter; unsigned int status_gfx_addr; drm_local_map_t hws_map; @@ -247,16 +252,17 @@ extern int i915_irq_emit(struct drm_device *dev, void *data, extern int i915_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_priv); -extern int i915_driver_vblank_wait(struct drm_device *dev, unsigned int *sequence); -extern int i915_driver_vblank_wait2(struct drm_device *dev, unsigned int *sequence); extern irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS); extern void i915_driver_irq_preinstall(struct drm_device * dev); -extern void i915_driver_irq_postinstall(struct drm_device * dev); +extern int i915_driver_irq_postinstall(struct drm_device *dev); extern void i915_driver_irq_uninstall(struct drm_device * dev); extern int i915_vblank_pipe_set(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int i915_vblank_pipe_get(struct drm_device *dev, void *data, struct drm_file *file_priv); +extern int i915_enable_vblank(struct drm_device *dev, int crtc); +extern void i915_disable_vblank(struct drm_device *dev, int crtc); +extern u32 i915_get_vblank_counter(struct drm_device *dev, int crtc); extern int i915_vblank_swap(struct drm_device *dev, void *data, struct drm_file *file_priv); extern void i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask); @@ -278,6 +284,10 @@ extern void i915_mem_release(struct drm_device * dev, extern int i915_save_state(struct drm_device *dev); extern int i915_restore_state(struct drm_device *dev); +/* i915_suspend.c */ +extern int i915_save_state(struct drm_device *dev); +extern int i915_restore_state(struct drm_device *dev); + /* i915_opregion.c */ extern int intel_opregion_init(struct drm_device *dev); extern void intel_opregion_free(struct drm_device *dev); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index ae7d3a82a6d..f8759597233 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -35,9 +35,8 @@ /** These are the interrupts used by the driver */ #define I915_INTERRUPT_ENABLE_MASK (I915_USER_INTERRUPT | \ - I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | \ - I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT | \ I915_ASLE_INTERRUPT | \ + I915_DISPLAY_PIPE_A_EVENT_INTERRUPT | \ I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) void @@ -60,6 +59,64 @@ i915_disable_irq(drm_i915_private_t *dev_priv, u32 mask) } } +/** + * i915_get_pipe - return the the pipe associated with a given plane + * @dev: DRM device + * @plane: plane to look for + * + * The Intel Mesa & 2D drivers call the vblank routines with a plane number + * rather than a pipe number, since they may not always be equal. This routine + * maps the given @plane back to a pipe number. + */ +static int +i915_get_pipe(struct drm_device *dev, int plane) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + u32 dspcntr; + + dspcntr = plane ? I915_READ(DSPBCNTR) : I915_READ(DSPACNTR); + + return dspcntr & DISPPLANE_SEL_PIPE_MASK ? 1 : 0; +} + +/** + * i915_get_plane - return the the plane associated with a given pipe + * @dev: DRM device + * @pipe: pipe to look for + * + * The Intel Mesa & 2D drivers call the vblank routines with a plane number + * rather than a plane number, since they may not always be equal. This routine + * maps the given @pipe back to a plane number. + */ +static int +i915_get_plane(struct drm_device *dev, int pipe) +{ + if (i915_get_pipe(dev, 0) == pipe) + return 0; + return 1; +} + +/** + * i915_pipe_enabled - check if a pipe is enabled + * @dev: DRM device + * @pipe: pipe to check + * + * Reading certain registers when the pipe is disabled can hang the chip. + * Use this routine to make sure the PLL is running and the pipe is active + * before reading such registers if unsure. + */ +static int +i915_pipe_enabled(struct drm_device *dev, int pipe) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + unsigned long pipeconf = pipe ? PIPEBCONF : PIPEACONF; + + if (I915_READ(pipeconf) & PIPEACONF_ENABLE) + return 1; + + return 0; +} + /** * Emit blits for scheduled buffer swaps. * @@ -71,8 +128,7 @@ static void i915_vblank_tasklet(struct drm_device *dev) unsigned long irqflags; struct list_head *list, *tmp, hits, *hit; int nhits, nrects, slice[2], upper[2], lower[2], i; - unsigned counter[2] = { atomic_read(&dev->vbl_received), - atomic_read(&dev->vbl_received2) }; + unsigned counter[2]; struct drm_drawable_info *drw; drm_i915_sarea_t *sarea_priv = dev_priv->sarea_priv; u32 cpp = dev_priv->cpp; @@ -94,6 +150,9 @@ static void i915_vblank_tasklet(struct drm_device *dev) src_pitch >>= 2; } + counter[0] = drm_vblank_count(dev, 0); + counter[1] = drm_vblank_count(dev, 1); + DRM_DEBUG("\n"); INIT_LIST_HEAD(&hits); @@ -106,12 +165,14 @@ static void i915_vblank_tasklet(struct drm_device *dev) list_for_each_safe(list, tmp, &dev_priv->vbl_swaps.head) { drm_i915_vbl_swap_t *vbl_swap = list_entry(list, drm_i915_vbl_swap_t, head); + int pipe = i915_get_pipe(dev, vbl_swap->plane); - if ((counter[vbl_swap->pipe] - vbl_swap->sequence) > (1<<23)) + if ((counter[pipe] - vbl_swap->sequence) > (1<<23)) continue; list_del(list); dev_priv->swaps_pending--; + drm_vblank_put(dev, pipe); spin_unlock(&dev_priv->swaps_lock); spin_lock(&dev->drw_lock); @@ -204,7 +265,7 @@ static void i915_vblank_tasklet(struct drm_device *dev) drm_i915_vbl_swap_t *swap_hit = list_entry(hit, drm_i915_vbl_swap_t, head); struct drm_clip_rect *rect; - int num_rects, pipe; + int num_rects, plane; unsigned short top, bottom; drw = drm_get_drawable_info(dev, swap_hit->drw_id); @@ -213,9 +274,9 @@ static void i915_vblank_tasklet(struct drm_device *dev) continue; rect = drw->rects; - pipe = swap_hit->pipe; - top = upper[pipe]; - bottom = lower[pipe]; + plane = swap_hit->plane; + top = upper[plane]; + bottom = lower[plane]; for (num_rects = drw->num_rects; num_rects--; rect++) { int y1 = max(rect->y1, top); @@ -252,22 +313,54 @@ static void i915_vblank_tasklet(struct drm_device *dev) } } +u32 i915_get_vblank_counter(struct drm_device *dev, int plane) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + unsigned long high_frame; + unsigned long low_frame; + u32 high1, high2, low, count; + int pipe; + + pipe = i915_get_pipe(dev, plane); + high_frame = pipe ? PIPEBFRAMEHIGH : PIPEAFRAMEHIGH; + low_frame = pipe ? PIPEBFRAMEPIXEL : PIPEAFRAMEPIXEL; + + if (!i915_pipe_enabled(dev, pipe)) { + DRM_ERROR("trying to get vblank count for disabled pipe %d\n", pipe); + return 0; + } + + /* + * High & low register fields aren't synchronized, so make sure + * we get a low value that's stable across two reads of the high + * register. + */ + do { + high1 = ((I915_READ(high_frame) & PIPE_FRAME_HIGH_MASK) >> + PIPE_FRAME_HIGH_SHIFT); + low = ((I915_READ(low_frame) & PIPE_FRAME_LOW_MASK) >> + PIPE_FRAME_LOW_SHIFT); + high2 = ((I915_READ(high_frame) & PIPE_FRAME_HIGH_MASK) >> + PIPE_FRAME_HIGH_SHIFT); + } while (high1 != high2); + + count = (high1 << 8) | low; + + return count; +} + irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) { struct drm_device *dev = (struct drm_device *) arg; drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - u32 pipea_stats, pipeb_stats; u32 iir; - - pipea_stats = I915_READ(PIPEASTAT); - pipeb_stats = I915_READ(PIPEBSTAT); + u32 pipea_stats, pipeb_stats; + int vblank = 0; if (dev->pdev->msi_enabled) I915_WRITE(IMR, ~0); iir = I915_READ(IIR); - DRM_DEBUG("iir=%08x\n", iir); - if (iir == 0) { if (dev->pdev->msi_enabled) { I915_WRITE(IMR, dev_priv->irq_mask_reg); @@ -276,48 +369,56 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) return IRQ_NONE; } - I915_WRITE(PIPEASTAT, pipea_stats); - I915_WRITE(PIPEBSTAT, pipeb_stats); - - I915_WRITE(IIR, iir); - if (dev->pdev->msi_enabled) - I915_WRITE(IMR, dev_priv->irq_mask_reg); - (void) I915_READ(IIR); /* Flush posted writes */ - - dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - - if (iir & I915_USER_INTERRUPT) - DRM_WAKEUP(&dev_priv->irq_queue); - - if (iir & (I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT | - I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT)) { - int vblank_pipe = dev_priv->vblank_pipe; - - if ((vblank_pipe & - (DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B)) - == (DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B)) { - if (iir & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) - atomic_inc(&dev->vbl_received); - if (iir & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) - atomic_inc(&dev->vbl_received2); - } else if (((iir & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) && - (vblank_pipe & DRM_I915_VBLANK_PIPE_A)) || - ((iir & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) && - (vblank_pipe & DRM_I915_VBLANK_PIPE_B))) - atomic_inc(&dev->vbl_received); + /* + * Clear the PIPE(A|B)STAT regs before the IIR otherwise + * we may get extra interrupts. + */ + if (iir & I915_DISPLAY_PIPE_A_EVENT_INTERRUPT) { + pipea_stats = I915_READ(PIPEASTAT); + if (!(dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_A)) + pipea_stats &= ~(PIPE_START_VBLANK_INTERRUPT_ENABLE | + PIPE_VBLANK_INTERRUPT_ENABLE); + else if (pipea_stats & (PIPE_START_VBLANK_INTERRUPT_STATUS| + PIPE_VBLANK_INTERRUPT_STATUS)) { + vblank++; + drm_handle_vblank(dev, i915_get_plane(dev, 0)); + } - DRM_WAKEUP(&dev->vbl_queue); - drm_vbl_send_signals(dev); + I915_WRITE(PIPEASTAT, pipea_stats); + } + if (iir & I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) { + pipeb_stats = I915_READ(PIPEBSTAT); + /* Ack the event */ + I915_WRITE(PIPEBSTAT, pipeb_stats); + + /* The vblank interrupt gets enabled even if we didn't ask for + it, so make sure it's shut down again */ + if (!(dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B)) + pipeb_stats &= ~(PIPE_START_VBLANK_INTERRUPT_ENABLE | + PIPE_VBLANK_INTERRUPT_ENABLE); + else if (pipeb_stats & (PIPE_START_VBLANK_INTERRUPT_STATUS| + PIPE_VBLANK_INTERRUPT_STATUS)) { + vblank++; + drm_handle_vblank(dev, i915_get_plane(dev, 1)); + } - if (dev_priv->swaps_pending > 0) - drm_locked_tasklet(dev, i915_vblank_tasklet); + if (pipeb_stats & I915_LEGACY_BLC_EVENT_STATUS) + opregion_asle_intr(dev); + I915_WRITE(PIPEBSTAT, pipeb_stats); } if (iir & I915_ASLE_INTERRUPT) opregion_asle_intr(dev); - if (iir & I915_DISPLAY_PIPE_B_EVENT_INTERRUPT) - opregion_asle_intr(dev); + dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); + + if (dev->pdev->msi_enabled) + I915_WRITE(IMR, dev_priv->irq_mask_reg); + I915_WRITE(IIR, iir); + (void) I915_READ(IIR); + + if (vblank && dev_priv->swaps_pending > 0) + drm_locked_tasklet(dev, i915_vblank_tasklet); return IRQ_HANDLED; } @@ -358,7 +459,7 @@ static void i915_user_irq_get(struct drm_device *dev) spin_unlock(&dev_priv->user_irq_lock); } -static void i915_user_irq_put(struct drm_device *dev) +void i915_user_irq_put(struct drm_device *dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; @@ -395,41 +496,10 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) } dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - return ret; -} - -static int i915_driver_vblank_do_wait(struct drm_device *dev, unsigned int *sequence, - atomic_t *counter) -{ - drm_i915_private_t *dev_priv = dev->dev_private; - unsigned int cur_vblank; - int ret = 0; - - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); - return -EINVAL; - } - - DRM_WAIT_ON(ret, dev->vbl_queue, 3 * DRM_HZ, - (((cur_vblank = atomic_read(counter)) - - *sequence) <= (1<<23))); - - *sequence = cur_vblank; return ret; } - -int i915_driver_vblank_wait(struct drm_device *dev, unsigned int *sequence) -{ - return i915_driver_vblank_do_wait(dev, sequence, &dev->vbl_received); -} - -int i915_driver_vblank_wait2(struct drm_device *dev, unsigned int *sequence) -{ - return i915_driver_vblank_do_wait(dev, sequence, &dev->vbl_received2); -} - /* Needs the lock as it touches the ring. */ int i915_irq_emit(struct drm_device *dev, void *data, @@ -472,40 +542,88 @@ int i915_irq_wait(struct drm_device *dev, void *data, return i915_wait_irq(dev, irqwait->irq_seq); } +int i915_enable_vblank(struct drm_device *dev, int plane) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + int pipe = i915_get_pipe(dev, plane); + u32 pipestat_reg = 0; + u32 pipestat; + + switch (pipe) { + case 0: + pipestat_reg = PIPEASTAT; + i915_enable_irq(dev_priv, I915_DISPLAY_PIPE_A_EVENT_INTERRUPT); + break; + case 1: + pipestat_reg = PIPEBSTAT; + i915_enable_irq(dev_priv, I915_DISPLAY_PIPE_B_EVENT_INTERRUPT); + break; + default: + DRM_ERROR("tried to enable vblank on non-existent pipe %d\n", + pipe); + break; + } + + if (pipestat_reg) { + pipestat = I915_READ(pipestat_reg); + if (IS_I965G(dev)) + pipestat |= PIPE_START_VBLANK_INTERRUPT_ENABLE; + else + pipestat |= PIPE_VBLANK_INTERRUPT_ENABLE; + /* Clear any stale interrupt status */ + pipestat |= (PIPE_START_VBLANK_INTERRUPT_STATUS | + PIPE_VBLANK_INTERRUPT_STATUS); + I915_WRITE(pipestat_reg, pipestat); + } + + return 0; +} + +void i915_disable_vblank(struct drm_device *dev, int plane) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + int pipe = i915_get_pipe(dev, plane); + u32 pipestat_reg = 0; + u32 pipestat; + + switch (pipe) { + case 0: + pipestat_reg = PIPEASTAT; + i915_disable_irq(dev_priv, I915_DISPLAY_PIPE_A_EVENT_INTERRUPT); + break; + case 1: + pipestat_reg = PIPEBSTAT; + i915_disable_irq(dev_priv, I915_DISPLAY_PIPE_B_EVENT_INTERRUPT); + break; + default: + DRM_ERROR("tried to disable vblank on non-existent pipe %d\n", + pipe); + break; + } + + if (pipestat_reg) { + pipestat = I915_READ(pipestat_reg); + pipestat &= ~(PIPE_START_VBLANK_INTERRUPT_ENABLE | + PIPE_VBLANK_INTERRUPT_ENABLE); + /* Clear any stale interrupt status */ + pipestat |= (PIPE_START_VBLANK_INTERRUPT_STATUS | + PIPE_VBLANK_INTERRUPT_STATUS); + I915_WRITE(pipestat_reg, pipestat); + } +} + /* Set the vblank monitor pipe */ int i915_vblank_pipe_set(struct drm_device *dev, void *data, struct drm_file *file_priv) { drm_i915_private_t *dev_priv = dev->dev_private; - drm_i915_vblank_pipe_t *pipe = data; - u32 enable_mask = 0, disable_mask = 0; if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } - if (pipe->pipe & ~(DRM_I915_VBLANK_PIPE_A|DRM_I915_VBLANK_PIPE_B)) { - DRM_ERROR("called with invalid pipe 0x%x\n", pipe->pipe); - return -EINVAL; - } - - if (pipe->pipe & DRM_I915_VBLANK_PIPE_A) - enable_mask |= I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; - else - disable_mask |= I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; - - if (pipe->pipe & DRM_I915_VBLANK_PIPE_B) - enable_mask |= I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; - else - disable_mask |= I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; - - i915_enable_irq(dev_priv, enable_mask); - i915_disable_irq(dev_priv, disable_mask); - - dev_priv->vblank_pipe = pipe->pipe; - return 0; } @@ -514,19 +632,13 @@ int i915_vblank_pipe_get(struct drm_device *dev, void *data, { drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_vblank_pipe_t *pipe = data; - u16 flag; if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } - flag = I915_READ(IMR); - pipe->pipe = 0; - if (flag & I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT) - pipe->pipe |= DRM_I915_VBLANK_PIPE_A; - if (flag & I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT) - pipe->pipe |= DRM_I915_VBLANK_PIPE_B; + pipe->pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B; return 0; } @@ -540,9 +652,10 @@ int i915_vblank_swap(struct drm_device *dev, void *data, drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_vblank_swap_t *swap = data; drm_i915_vbl_swap_t *vbl_swap; - unsigned int pipe, seqtype, curseq; + unsigned int pipe, seqtype, curseq, plane; unsigned long irqflags; struct list_head *list; + int ret; if (!dev_priv) { DRM_ERROR("%s called with no initialization\n", __func__); @@ -560,7 +673,8 @@ int i915_vblank_swap(struct drm_device *dev, void *data, return -EINVAL; } - pipe = (swap->seqtype & _DRM_VBLANK_SECONDARY) ? 1 : 0; + plane = (swap->seqtype & _DRM_VBLANK_SECONDARY) ? 1 : 0; + pipe = i915_get_pipe(dev, plane); seqtype = swap->seqtype & (_DRM_VBLANK_RELATIVE | _DRM_VBLANK_ABSOLUTE); @@ -579,7 +693,14 @@ int i915_vblank_swap(struct drm_device *dev, void *data, spin_unlock_irqrestore(&dev->drw_lock, irqflags); - curseq = atomic_read(pipe ? &dev->vbl_received2 : &dev->vbl_received); + /* + * We take the ref here and put it when the swap actually completes + * in the tasklet. + */ + ret = drm_vblank_get(dev, pipe); + if (ret) + return ret; + curseq = drm_vblank_count(dev, pipe); if (seqtype == _DRM_VBLANK_RELATIVE) swap->sequence += curseq; @@ -589,6 +710,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, swap->sequence = curseq + 1; } else { DRM_DEBUG("Missed target sequence\n"); + drm_vblank_put(dev, pipe); return -EINVAL; } } @@ -599,7 +721,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, vbl_swap = list_entry(list, drm_i915_vbl_swap_t, head); if (vbl_swap->drw_id == swap->drawable && - vbl_swap->pipe == pipe && + vbl_swap->plane == plane && vbl_swap->sequence == swap->sequence) { spin_unlock_irqrestore(&dev_priv->swaps_lock, irqflags); DRM_DEBUG("Already scheduled\n"); @@ -611,6 +733,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, if (dev_priv->swaps_pending >= 100) { DRM_DEBUG("Too many swaps queued\n"); + drm_vblank_put(dev, pipe); return -EBUSY; } @@ -618,13 +741,14 @@ int i915_vblank_swap(struct drm_device *dev, void *data, if (!vbl_swap) { DRM_ERROR("Failed to allocate memory to queue swap\n"); + drm_vblank_put(dev, pipe); return -ENOMEM; } DRM_DEBUG("\n"); vbl_swap->drw_id = swap->drawable; - vbl_swap->pipe = pipe; + vbl_swap->plane = plane; vbl_swap->sequence = swap->sequence; spin_lock_irqsave(&dev_priv->swaps_lock, irqflags); @@ -643,28 +767,32 @@ void i915_driver_irq_preinstall(struct drm_device * dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - I915_WRITE(HWSTAM, 0xfffe); - I915_WRITE(IMR, 0x0); + I915_WRITE(HWSTAM, 0xeffe); + I915_WRITE(IMR, 0xffffffff); I915_WRITE(IER, 0x0); } -void i915_driver_irq_postinstall(struct drm_device * dev) +int i915_driver_irq_postinstall(struct drm_device *dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + int ret, num_pipes = 2; spin_lock_init(&dev_priv->swaps_lock); INIT_LIST_HEAD(&dev_priv->vbl_swaps.head); dev_priv->swaps_pending = 0; - if (!dev_priv->vblank_pipe) - dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A; - /* Set initial unmasked IRQs to just the selected vblank pipes. */ dev_priv->irq_mask_reg = ~0; - if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_A) - dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; - if (dev_priv->vblank_pipe & DRM_I915_VBLANK_PIPE_B) - dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; + + ret = drm_vblank_init(dev, num_pipes); + if (ret) + return ret; + + dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A | DRM_I915_VBLANK_PIPE_B; + dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_A_VBLANK_INTERRUPT; + dev_priv->irq_mask_reg &= ~I915_DISPLAY_PIPE_B_VBLANK_INTERRUPT; + + dev->max_vblank_count = 0xffffff; /* only 24 bits of frame count */ dev_priv->irq_mask_reg &= I915_INTERRUPT_ENABLE_MASK; @@ -673,22 +801,29 @@ void i915_driver_irq_postinstall(struct drm_device * dev) (void) I915_READ(IER); opregion_enable_asle(dev); - DRM_INIT_WAITQUEUE(&dev_priv->irq_queue); + + return 0; } void i915_driver_irq_uninstall(struct drm_device * dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; - u16 temp; + u32 temp; if (!dev_priv) return; - I915_WRITE(HWSTAM, 0xffff); - I915_WRITE(IMR, 0xffff); + dev_priv->vblank_pipe = 0; + + I915_WRITE(HWSTAM, 0xffffffff); + I915_WRITE(IMR, 0xffffffff); I915_WRITE(IER, 0x0); + temp = I915_READ(PIPEASTAT); + I915_WRITE(PIPEASTAT, temp); + temp = I915_READ(PIPEBSTAT); + I915_WRITE(PIPEBSTAT, temp); temp = I915_READ(IIR); I915_WRITE(IIR, temp); } diff --git a/drivers/gpu/drm/mga/mga_drv.c b/drivers/gpu/drm/mga/mga_drv.c index 5572939fc7d..97ee566ef74 100644 --- a/drivers/gpu/drm/mga/mga_drv.c +++ b/drivers/gpu/drm/mga/mga_drv.c @@ -45,15 +45,16 @@ static struct pci_device_id pciidlist[] = { static struct drm_driver driver = { .driver_features = DRIVER_USE_AGP | DRIVER_USE_MTRR | DRIVER_PCI_DMA | - DRIVER_HAVE_DMA | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | - DRIVER_IRQ_VBL, + DRIVER_HAVE_DMA | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED, .dev_priv_size = sizeof(drm_mga_buf_priv_t), .load = mga_driver_load, .unload = mga_driver_unload, .lastclose = mga_driver_lastclose, .dma_quiescent = mga_driver_dma_quiescent, .device_is_agp = mga_driver_device_is_agp, - .vblank_wait = mga_driver_vblank_wait, + .get_vblank_counter = mga_get_vblank_counter, + .enable_vblank = mga_enable_vblank, + .disable_vblank = mga_disable_vblank, .irq_preinstall = mga_driver_irq_preinstall, .irq_postinstall = mga_driver_irq_postinstall, .irq_uninstall = mga_driver_irq_uninstall, @@ -64,20 +65,20 @@ static struct drm_driver driver = { .ioctls = mga_ioctls, .dma_ioctl = mga_dma_buffers, .fops = { - .owner = THIS_MODULE, - .open = drm_open, - .release = drm_release, - .ioctl = drm_ioctl, - .mmap = drm_mmap, - .poll = drm_poll, - .fasync = drm_fasync, + .owner = THIS_MODULE, + .open = drm_open, + .release = drm_release, + .ioctl = drm_ioctl, + .mmap = drm_mmap, + .poll = drm_poll, + .fasync = drm_fasync, #ifdef CONFIG_COMPAT - .compat_ioctl = mga_compat_ioctl, + .compat_ioctl = mga_compat_ioctl, #endif - }, + }, .pci_driver = { - .name = DRIVER_NAME, - .id_table = pciidlist, + .name = DRIVER_NAME, + .id_table = pciidlist, }, .name = DRIVER_NAME, diff --git a/drivers/gpu/drm/mga/mga_drv.h b/drivers/gpu/drm/mga/mga_drv.h index f6ebd24bd58..88257c276eb 100644 --- a/drivers/gpu/drm/mga/mga_drv.h +++ b/drivers/gpu/drm/mga/mga_drv.h @@ -120,6 +120,7 @@ typedef struct drm_mga_private { u32 clear_cmd; u32 maccess; + atomic_t vbl_received; /**< Number of vblanks received. */ wait_queue_head_t fence_queue; atomic_t last_fence_retired; u32 next_fence_to_post; @@ -181,11 +182,14 @@ extern int mga_warp_install_microcode(drm_mga_private_t * dev_priv); extern int mga_warp_init(drm_mga_private_t * dev_priv); /* mga_irq.c */ +extern int mga_enable_vblank(struct drm_device *dev, int crtc); +extern void mga_disable_vblank(struct drm_device *dev, int crtc); +extern u32 mga_get_vblank_counter(struct drm_device *dev, int crtc); extern int mga_driver_fence_wait(struct drm_device * dev, unsigned int *sequence); extern int mga_driver_vblank_wait(struct drm_device * dev, unsigned int *sequence); extern irqreturn_t mga_driver_irq_handler(DRM_IRQ_ARGS); extern void mga_driver_irq_preinstall(struct drm_device * dev); -extern void mga_driver_irq_postinstall(struct drm_device * dev); +extern int mga_driver_irq_postinstall(struct drm_device *dev); extern void mga_driver_irq_uninstall(struct drm_device * dev); extern long mga_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); diff --git a/drivers/gpu/drm/mga/mga_irq.c b/drivers/gpu/drm/mga/mga_irq.c index 9302cb8f0f8..bab42f41188 100644 --- a/drivers/gpu/drm/mga/mga_irq.c +++ b/drivers/gpu/drm/mga/mga_irq.c @@ -1,5 +1,6 @@ /* mga_irq.c -- IRQ handling for radeon -*- linux-c -*- - * + */ +/* * Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. * * The Weather Channel (TM) funded Tungsten Graphics to develop the @@ -35,6 +36,18 @@ #include "mga_drm.h" #include "mga_drv.h" +u32 mga_get_vblank_counter(struct drm_device *dev, int crtc) +{ + const drm_mga_private_t *const dev_priv = + (drm_mga_private_t *) dev->dev_private; + + if (crtc != 0) + return 0; + + return atomic_read(&dev_priv->vbl_received); +} + + irqreturn_t mga_driver_irq_handler(DRM_IRQ_ARGS) { struct drm_device *dev = (struct drm_device *) arg; @@ -47,9 +60,8 @@ irqreturn_t mga_driver_irq_handler(DRM_IRQ_ARGS) /* VBLANK interrupt */ if (status & MGA_VLINEPEN) { MGA_WRITE(MGA_ICLEAR, MGA_VLINEICLR); - atomic_inc(&dev->vbl_received); - DRM_WAKEUP(&dev->vbl_queue); - drm_vbl_send_signals(dev); + atomic_inc(&dev_priv->vbl_received); + drm_handle_vblank(dev, 0); handled = 1; } @@ -58,6 +70,7 @@ irqreturn_t mga_driver_irq_handler(DRM_IRQ_ARGS) const u32 prim_start = MGA_READ(MGA_PRIMADDRESS); const u32 prim_end = MGA_READ(MGA_PRIMEND); + MGA_WRITE(MGA_ICLEAR, MGA_SOFTRAPICLR); /* In addition to clearing the interrupt-pending bit, we @@ -72,28 +85,39 @@ irqreturn_t mga_driver_irq_handler(DRM_IRQ_ARGS) handled = 1; } - if (handled) { + if (handled) return IRQ_HANDLED; - } return IRQ_NONE; } -int mga_driver_vblank_wait(struct drm_device * dev, unsigned int *sequence) +int mga_enable_vblank(struct drm_device *dev, int crtc) { - unsigned int cur_vblank; - int ret = 0; + drm_mga_private_t *dev_priv = (drm_mga_private_t *) dev->dev_private; - /* Assume that the user has missed the current sequence number - * by about a day rather than she wants to wait for years - * using vertical blanks... - */ - DRM_WAIT_ON(ret, dev->vbl_queue, 3 * DRM_HZ, - (((cur_vblank = atomic_read(&dev->vbl_received)) - - *sequence) <= (1 << 23))); + if (crtc != 0) { + DRM_ERROR("tried to enable vblank on non-existent crtc %d\n", + crtc); + return 0; + } - *sequence = cur_vblank; + MGA_WRITE(MGA_IEN, MGA_VLINEIEN | MGA_SOFTRAPEN); + return 0; +} - return ret; + +void mga_disable_vblank(struct drm_device *dev, int crtc) +{ + if (crtc != 0) { + DRM_ERROR("tried to disable vblank on non-existent crtc %d\n", + crtc); + } + + /* Do *NOT* disable the vertical refresh interrupt. MGA doesn't have + * a nice hardware counter that tracks the number of refreshes when + * the interrupt is disabled, and the kernel doesn't know the refresh + * rate to calculate an estimate. + */ + /* MGA_WRITE(MGA_IEN, MGA_VLINEIEN | MGA_SOFTRAPEN); */ } int mga_driver_fence_wait(struct drm_device * dev, unsigned int *sequence) @@ -125,14 +149,22 @@ void mga_driver_irq_preinstall(struct drm_device * dev) MGA_WRITE(MGA_ICLEAR, ~0); } -void mga_driver_irq_postinstall(struct drm_device * dev) +int mga_driver_irq_postinstall(struct drm_device *dev) { drm_mga_private_t *dev_priv = (drm_mga_private_t *) dev->dev_private; + int ret; + + ret = drm_vblank_init(dev, 1); + if (ret) + return ret; DRM_INIT_WAITQUEUE(&dev_priv->fence_queue); - /* Turn on vertical blank interrupt and soft trap interrupt. */ - MGA_WRITE(MGA_IEN, MGA_VLINEIEN | MGA_SOFTRAPEN); + /* Turn on soft trap interrupt. Vertical blank interrupts are enabled + * in mga_enable_vblank. + */ + MGA_WRITE(MGA_IEN, MGA_SOFTRAPEN); + return 0; } void mga_driver_irq_uninstall(struct drm_device * dev) diff --git a/drivers/gpu/drm/r128/r128_drv.c b/drivers/gpu/drm/r128/r128_drv.c index 6108e7587e1..3265d53ba91 100644 --- a/drivers/gpu/drm/r128/r128_drv.c +++ b/drivers/gpu/drm/r128/r128_drv.c @@ -43,12 +43,13 @@ static struct pci_device_id pciidlist[] = { static struct drm_driver driver = { .driver_features = DRIVER_USE_AGP | DRIVER_USE_MTRR | DRIVER_PCI_DMA | DRIVER_SG | - DRIVER_HAVE_DMA | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | - DRIVER_IRQ_VBL, + DRIVER_HAVE_DMA | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED, .dev_priv_size = sizeof(drm_r128_buf_priv_t), .preclose = r128_driver_preclose, .lastclose = r128_driver_lastclose, - .vblank_wait = r128_driver_vblank_wait, + .get_vblank_counter = r128_get_vblank_counter, + .enable_vblank = r128_enable_vblank, + .disable_vblank = r128_disable_vblank, .irq_preinstall = r128_driver_irq_preinstall, .irq_postinstall = r128_driver_irq_postinstall, .irq_uninstall = r128_driver_irq_uninstall, @@ -59,21 +60,20 @@ static struct drm_driver driver = { .ioctls = r128_ioctls, .dma_ioctl = r128_cce_buffers, .fops = { - .owner = THIS_MODULE, - .open = drm_open, - .release = drm_release, - .ioctl = drm_ioctl, - .mmap = drm_mmap, - .poll = drm_poll, - .fasync = drm_fasync, + .owner = THIS_MODULE, + .open = drm_open, + .release = drm_release, + .ioctl = drm_ioctl, + .mmap = drm_mmap, + .poll = drm_poll, + .fasync = drm_fasync, #ifdef CONFIG_COMPAT - .compat_ioctl = r128_compat_ioctl, + .compat_ioctl = r128_compat_ioctl, #endif }, - .pci_driver = { - .name = DRIVER_NAME, - .id_table = pciidlist, + .name = DRIVER_NAME, + .id_table = pciidlist, }, .name = DRIVER_NAME, @@ -87,6 +87,7 @@ static struct drm_driver driver = { static int __init r128_init(void) { driver.num_ioctls = r128_max_ioctl; + return drm_init(&driver); } diff --git a/drivers/gpu/drm/r128/r128_drv.h b/drivers/gpu/drm/r128/r128_drv.h index 011105e51ac..5898b274279 100644 --- a/drivers/gpu/drm/r128/r128_drv.h +++ b/drivers/gpu/drm/r128/r128_drv.h @@ -29,7 +29,7 @@ * Rickard E. (Rik) Faith * Kevin E. Martin * Gareth Hughes - * Michel Dänzer + * Michel D�zer */ #ifndef __R128_DRV_H__ @@ -97,6 +97,8 @@ typedef struct drm_r128_private { u32 crtc_offset; u32 crtc_offset_cntl; + atomic_t vbl_received; + u32 color_fmt; unsigned int front_offset; unsigned int front_pitch; @@ -149,11 +151,12 @@ extern int r128_wait_ring(drm_r128_private_t * dev_priv, int n); extern int r128_do_cce_idle(drm_r128_private_t * dev_priv); extern int r128_do_cleanup_cce(struct drm_device * dev); -extern int r128_driver_vblank_wait(struct drm_device * dev, unsigned int *sequence); - +extern int r128_enable_vblank(struct drm_device *dev, int crtc); +extern void r128_disable_vblank(struct drm_device *dev, int crtc); +extern u32 r128_get_vblank_counter(struct drm_device *dev, int crtc); extern irqreturn_t r128_driver_irq_handler(DRM_IRQ_ARGS); extern void r128_driver_irq_preinstall(struct drm_device * dev); -extern void r128_driver_irq_postinstall(struct drm_device * dev); +extern int r128_driver_irq_postinstall(struct drm_device *dev); extern void r128_driver_irq_uninstall(struct drm_device * dev); extern void r128_driver_lastclose(struct drm_device * dev); extern void r128_driver_preclose(struct drm_device * dev, diff --git a/drivers/gpu/drm/r128/r128_irq.c b/drivers/gpu/drm/r128/r128_irq.c index c76fdca7662..d7349012a68 100644 --- a/drivers/gpu/drm/r128/r128_irq.c +++ b/drivers/gpu/drm/r128/r128_irq.c @@ -35,6 +35,16 @@ #include "r128_drm.h" #include "r128_drv.h" +u32 r128_get_vblank_counter(struct drm_device *dev, int crtc) +{ + const drm_r128_private_t *dev_priv = dev->dev_private; + + if (crtc != 0) + return 0; + + return atomic_read(&dev_priv->vbl_received); +} + irqreturn_t r128_driver_irq_handler(DRM_IRQ_ARGS) { struct drm_device *dev = (struct drm_device *) arg; @@ -46,30 +56,38 @@ irqreturn_t r128_driver_irq_handler(DRM_IRQ_ARGS) /* VBLANK interrupt */ if (status & R128_CRTC_VBLANK_INT) { R128_WRITE(R128_GEN_INT_STATUS, R128_CRTC_VBLANK_INT_AK); - atomic_inc(&dev->vbl_received); - DRM_WAKEUP(&dev->vbl_queue); - drm_vbl_send_signals(dev); + atomic_inc(&dev_priv->vbl_received); + drm_handle_vblank(dev, 0); return IRQ_HANDLED; } return IRQ_NONE; } -int r128_driver_vblank_wait(struct drm_device * dev, unsigned int *sequence) +int r128_enable_vblank(struct drm_device *dev, int crtc) { - unsigned int cur_vblank; - int ret = 0; + drm_r128_private_t *dev_priv = dev->dev_private; - /* Assume that the user has missed the current sequence number - * by about a day rather than she wants to wait for years - * using vertical blanks... - */ - DRM_WAIT_ON(ret, dev->vbl_queue, 3 * DRM_HZ, - (((cur_vblank = atomic_read(&dev->vbl_received)) - - *sequence) <= (1 << 23))); + if (crtc != 0) { + DRM_ERROR("%s: bad crtc %d\n", __func__, crtc); + return -EINVAL; + } - *sequence = cur_vblank; + R128_WRITE(R128_GEN_INT_CNTL, R128_CRTC_VBLANK_INT_EN); + return 0; +} + +void r128_disable_vblank(struct drm_device *dev, int crtc) +{ + if (crtc != 0) + DRM_ERROR("%s: bad crtc %d\n", __func__, crtc); - return ret; + /* + * FIXME: implement proper interrupt disable by using the vblank + * counter register (if available) + * + * R128_WRITE(R128_GEN_INT_CNTL, + * R128_READ(R128_GEN_INT_CNTL) & ~R128_CRTC_VBLANK_INT_EN); + */ } void r128_driver_irq_preinstall(struct drm_device * dev) @@ -82,12 +100,9 @@ void r128_driver_irq_preinstall(struct drm_device * dev) R128_WRITE(R128_GEN_INT_STATUS, R128_CRTC_VBLANK_INT_AK); } -void r128_driver_irq_postinstall(struct drm_device * dev) +int r128_driver_irq_postinstall(struct drm_device *dev) { - drm_r128_private_t *dev_priv = (drm_r128_private_t *) dev->dev_private; - - /* Turn on VBL interrupt */ - R128_WRITE(R128_GEN_INT_CNTL, R128_CRTC_VBLANK_INT_EN); + return drm_vblank_init(dev, 1); } void r128_driver_irq_uninstall(struct drm_device * dev) diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c index e1678cae9cc..6157cd4bb43 100644 --- a/drivers/gpu/drm/radeon/radeon_cp.c +++ b/drivers/gpu/drm/radeon/radeon_cp.c @@ -1287,7 +1287,7 @@ static int radeon_do_resume_cp(struct drm_device * dev) radeon_cp_init_ring_buffer(dev, dev_priv); radeon_do_engine_reset(dev); - radeon_enable_interrupt(dev); + radeon_irq_set_state(dev, RADEON_SW_INT_ENABLE, 1); DRM_DEBUG("radeon_do_resume_cp() complete\n"); diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index 637bd7faf13..71af746a4e4 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -52,6 +52,28 @@ static int dri_library_name(struct drm_device *dev, char *buf) "r300")); } +static int radeon_suspend(struct drm_device *dev, pm_message_t state) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + /* Disable *all* interrupts */ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) + RADEON_WRITE(R500_DxMODE_INT_MASK, 0); + RADEON_WRITE(RADEON_GEN_INT_CNTL, 0); + return 0; +} + +static int radeon_resume(struct drm_device *dev) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + /* Restore interrupt registers */ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) + RADEON_WRITE(R500_DxMODE_INT_MASK, dev_priv->r500_disp_irq_reg); + RADEON_WRITE(RADEON_GEN_INT_CNTL, dev_priv->irq_enable_reg); + return 0; +} + static struct pci_device_id pciidlist[] = { radeon_PCI_IDS }; @@ -59,8 +81,7 @@ static struct pci_device_id pciidlist[] = { static struct drm_driver driver = { .driver_features = DRIVER_USE_AGP | DRIVER_USE_MTRR | DRIVER_PCI_DMA | DRIVER_SG | - DRIVER_HAVE_IRQ | DRIVER_HAVE_DMA | DRIVER_IRQ_SHARED | - DRIVER_IRQ_VBL | DRIVER_IRQ_VBL2, + DRIVER_HAVE_IRQ | DRIVER_HAVE_DMA | DRIVER_IRQ_SHARED, .dev_priv_size = sizeof(drm_radeon_buf_priv_t), .load = radeon_driver_load, .firstopen = radeon_driver_firstopen, @@ -69,8 +90,11 @@ static struct drm_driver driver = { .postclose = radeon_driver_postclose, .lastclose = radeon_driver_lastclose, .unload = radeon_driver_unload, - .vblank_wait = radeon_driver_vblank_wait, - .vblank_wait2 = radeon_driver_vblank_wait2, + .suspend = radeon_suspend, + .resume = radeon_resume, + .get_vblank_counter = radeon_get_vblank_counter, + .enable_vblank = radeon_enable_vblank, + .disable_vblank = radeon_disable_vblank, .dri_library_name = dri_library_name, .irq_preinstall = radeon_driver_irq_preinstall, .irq_postinstall = radeon_driver_irq_postinstall, diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h index 09938169317..d7e9c6cc6a1 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.h +++ b/drivers/gpu/drm/radeon/radeon_drv.h @@ -378,17 +378,17 @@ extern void radeon_mem_release(struct drm_file *file_priv, struct mem_block *heap); /* radeon_irq.c */ +extern void radeon_irq_set_state(struct drm_device *dev, u32 mask, int state); extern int radeon_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int radeon_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_priv); extern void radeon_do_release(struct drm_device * dev); -extern int radeon_driver_vblank_wait(struct drm_device * dev, - unsigned int *sequence); -extern int radeon_driver_vblank_wait2(struct drm_device * dev, - unsigned int *sequence); +extern u32 radeon_get_vblank_counter(struct drm_device *dev, int crtc); +extern int radeon_enable_vblank(struct drm_device *dev, int crtc); +extern void radeon_disable_vblank(struct drm_device *dev, int crtc); extern irqreturn_t radeon_driver_irq_handler(DRM_IRQ_ARGS); extern void radeon_driver_irq_preinstall(struct drm_device * dev); -extern void radeon_driver_irq_postinstall(struct drm_device * dev); +extern int radeon_driver_irq_postinstall(struct drm_device *dev); extern void radeon_driver_irq_uninstall(struct drm_device * dev); extern void radeon_enable_interrupt(struct drm_device *dev); extern int radeon_vblank_crtc_get(struct drm_device *dev); @@ -397,19 +397,22 @@ extern int radeon_vblank_crtc_set(struct drm_device *dev, int64_t value); extern int radeon_driver_load(struct drm_device *dev, unsigned long flags); extern int radeon_driver_unload(struct drm_device *dev); extern int radeon_driver_firstopen(struct drm_device *dev); -extern void radeon_driver_preclose(struct drm_device * dev, struct drm_file *file_priv); -extern void radeon_driver_postclose(struct drm_device * dev, struct drm_file * filp); +extern void radeon_driver_preclose(struct drm_device *dev, + struct drm_file *file_priv); +extern void radeon_driver_postclose(struct drm_device *dev, + struct drm_file *file_priv); extern void radeon_driver_lastclose(struct drm_device * dev); -extern int radeon_driver_open(struct drm_device * dev, struct drm_file * filp_priv); +extern int radeon_driver_open(struct drm_device *dev, + struct drm_file *file_priv); extern long radeon_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); /* r300_cmdbuf.c */ extern void r300_init_reg_flags(struct drm_device *dev); -extern int r300_do_cp_cmdbuf(struct drm_device * dev, +extern int r300_do_cp_cmdbuf(struct drm_device *dev, struct drm_file *file_priv, - drm_radeon_kcmd_buffer_t * cmdbuf); + drm_radeon_kcmd_buffer_t *cmdbuf); /* Flags for stats.boxes */ @@ -623,6 +626,7 @@ extern int r300_do_cp_cmdbuf(struct drm_device * dev, # define RADEON_SW_INT_TEST (1 << 25) # define RADEON_SW_INT_TEST_ACK (1 << 25) # define RADEON_SW_INT_FIRE (1 << 26) +# define R500_DISPLAY_INT_STATUS (1 << 0) #define RADEON_HOST_PATH_CNTL 0x0130 # define RADEON_HDP_SOFT_RESET (1 << 26) @@ -1116,6 +1120,9 @@ extern int r300_do_cp_cmdbuf(struct drm_device * dev, #define R200_VAP_PVS_CNTL_1 0x22D0 +#define RADEON_CRTC_CRNT_FRAME 0x0214 +#define RADEON_CRTC2_CRNT_FRAME 0x0314 + #define R500_D1CRTC_STATUS 0x609c #define R500_D2CRTC_STATUS 0x689c #define R500_CRTC_V_BLANK (1<<0) diff --git a/drivers/gpu/drm/radeon/radeon_irq.c b/drivers/gpu/drm/radeon/radeon_irq.c index ee40d197deb..5079f7054a2 100644 --- a/drivers/gpu/drm/radeon/radeon_irq.c +++ b/drivers/gpu/drm/radeon/radeon_irq.c @@ -27,7 +27,7 @@ * * Authors: * Keith Whitwell - * Michel Dänzer + * Michel D�zer */ #include "drmP.h" @@ -35,12 +35,128 @@ #include "radeon_drm.h" #include "radeon_drv.h" -static __inline__ u32 radeon_acknowledge_irqs(drm_radeon_private_t * dev_priv, - u32 mask) +void radeon_irq_set_state(struct drm_device *dev, u32 mask, int state) { - u32 irqs = RADEON_READ(RADEON_GEN_INT_STATUS) & mask; + drm_radeon_private_t *dev_priv = dev->dev_private; + + if (state) + dev_priv->irq_enable_reg |= mask; + else + dev_priv->irq_enable_reg &= ~mask; + + RADEON_WRITE(RADEON_GEN_INT_CNTL, dev_priv->irq_enable_reg); +} + +static void r500_vbl_irq_set_state(struct drm_device *dev, u32 mask, int state) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + if (state) + dev_priv->r500_disp_irq_reg |= mask; + else + dev_priv->r500_disp_irq_reg &= ~mask; + + RADEON_WRITE(R500_DxMODE_INT_MASK, dev_priv->r500_disp_irq_reg); +} + +int radeon_enable_vblank(struct drm_device *dev, int crtc) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { + switch (crtc) { + case 0: + r500_vbl_irq_set_state(dev, R500_D1MODE_INT_MASK, 1); + break; + case 1: + r500_vbl_irq_set_state(dev, R500_D2MODE_INT_MASK, 1); + break; + default: + DRM_ERROR("tried to enable vblank on non-existent crtc %d\n", + crtc); + return EINVAL; + } + } else { + switch (crtc) { + case 0: + radeon_irq_set_state(dev, RADEON_CRTC_VBLANK_MASK, 1); + break; + case 1: + radeon_irq_set_state(dev, RADEON_CRTC2_VBLANK_MASK, 1); + break; + default: + DRM_ERROR("tried to enable vblank on non-existent crtc %d\n", + crtc); + return EINVAL; + } + } + + return 0; +} + +void radeon_disable_vblank(struct drm_device *dev, int crtc) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { + switch (crtc) { + case 0: + r500_vbl_irq_set_state(dev, R500_D1MODE_INT_MASK, 0); + break; + case 1: + r500_vbl_irq_set_state(dev, R500_D2MODE_INT_MASK, 0); + break; + default: + DRM_ERROR("tried to enable vblank on non-existent crtc %d\n", + crtc); + break; + } + } else { + switch (crtc) { + case 0: + radeon_irq_set_state(dev, RADEON_CRTC_VBLANK_MASK, 0); + break; + case 1: + radeon_irq_set_state(dev, RADEON_CRTC2_VBLANK_MASK, 0); + break; + default: + DRM_ERROR("tried to enable vblank on non-existent crtc %d\n", + crtc); + break; + } + } +} + +static inline u32 radeon_acknowledge_irqs(drm_radeon_private_t *dev_priv, u32 *r500_disp_int) +{ + u32 irqs = RADEON_READ(RADEON_GEN_INT_STATUS); + u32 irq_mask = RADEON_SW_INT_TEST; + + *r500_disp_int = 0; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { + /* vbl interrupts in a different place */ + + if (irqs & R500_DISPLAY_INT_STATUS) { + /* if a display interrupt */ + u32 disp_irq; + + disp_irq = RADEON_READ(R500_DISP_INTERRUPT_STATUS); + + *r500_disp_int = disp_irq; + if (disp_irq & R500_D1_VBLANK_INTERRUPT) + RADEON_WRITE(R500_D1MODE_VBLANK_STATUS, R500_VBLANK_ACK); + if (disp_irq & R500_D2_VBLANK_INTERRUPT) + RADEON_WRITE(R500_D2MODE_VBLANK_STATUS, R500_VBLANK_ACK); + } + irq_mask |= R500_DISPLAY_INT_STATUS; + } else + irq_mask |= RADEON_CRTC_VBLANK_STAT | RADEON_CRTC2_VBLANK_STAT; + + irqs &= irq_mask; + if (irqs) RADEON_WRITE(RADEON_GEN_INT_STATUS, irqs); + return irqs; } @@ -68,44 +184,33 @@ irqreturn_t radeon_driver_irq_handler(DRM_IRQ_ARGS) drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private; u32 stat; + u32 r500_disp_int; /* Only consider the bits we're interested in - others could be used * outside the DRM */ - stat = radeon_acknowledge_irqs(dev_priv, (RADEON_SW_INT_TEST_ACK | - RADEON_CRTC_VBLANK_STAT | - RADEON_CRTC2_VBLANK_STAT)); + stat = radeon_acknowledge_irqs(dev_priv, &r500_disp_int); if (!stat) return IRQ_NONE; stat &= dev_priv->irq_enable_reg; /* SW interrupt */ - if (stat & RADEON_SW_INT_TEST) { + if (stat & RADEON_SW_INT_TEST) DRM_WAKEUP(&dev_priv->swi_queue); - } /* VBLANK interrupt */ - if (stat & (RADEON_CRTC_VBLANK_STAT|RADEON_CRTC2_VBLANK_STAT)) { - int vblank_crtc = dev_priv->vblank_crtc; - - if ((vblank_crtc & - (DRM_RADEON_VBLANK_CRTC1 | DRM_RADEON_VBLANK_CRTC2)) == - (DRM_RADEON_VBLANK_CRTC1 | DRM_RADEON_VBLANK_CRTC2)) { - if (stat & RADEON_CRTC_VBLANK_STAT) - atomic_inc(&dev->vbl_received); - if (stat & RADEON_CRTC2_VBLANK_STAT) - atomic_inc(&dev->vbl_received2); - } else if (((stat & RADEON_CRTC_VBLANK_STAT) && - (vblank_crtc & DRM_RADEON_VBLANK_CRTC1)) || - ((stat & RADEON_CRTC2_VBLANK_STAT) && - (vblank_crtc & DRM_RADEON_VBLANK_CRTC2))) - atomic_inc(&dev->vbl_received); - - DRM_WAKEUP(&dev->vbl_queue); - drm_vbl_send_signals(dev); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { + if (r500_disp_int & R500_D1_VBLANK_INTERRUPT) + drm_handle_vblank(dev, 0); + if (r500_disp_int & R500_D2_VBLANK_INTERRUPT) + drm_handle_vblank(dev, 1); + } else { + if (stat & RADEON_CRTC_VBLANK_STAT) + drm_handle_vblank(dev, 0); + if (stat & RADEON_CRTC2_VBLANK_STAT) + drm_handle_vblank(dev, 1); } - return IRQ_HANDLED; } @@ -144,54 +249,31 @@ static int radeon_wait_irq(struct drm_device * dev, int swi_nr) return ret; } -static int radeon_driver_vblank_do_wait(struct drm_device * dev, - unsigned int *sequence, int crtc) +u32 radeon_get_vblank_counter(struct drm_device *dev, int crtc) { - drm_radeon_private_t *dev_priv = - (drm_radeon_private_t *) dev->dev_private; - unsigned int cur_vblank; - int ret = 0; - int ack = 0; - atomic_t *counter; + drm_radeon_private_t *dev_priv = dev->dev_private; + if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } - if (crtc == DRM_RADEON_VBLANK_CRTC1) { - counter = &dev->vbl_received; - ack |= RADEON_CRTC_VBLANK_STAT; - } else if (crtc == DRM_RADEON_VBLANK_CRTC2) { - counter = &dev->vbl_received2; - ack |= RADEON_CRTC2_VBLANK_STAT; - } else + if (crtc < 0 || crtc > 1) { + DRM_ERROR("Invalid crtc %d\n", crtc); return -EINVAL; + } - radeon_acknowledge_irqs(dev_priv, ack); - - dev_priv->stats.boxes |= RADEON_BOX_WAIT_IDLE; - - /* Assume that the user has missed the current sequence number - * by about a day rather than she wants to wait for years - * using vertical blanks... - */ - DRM_WAIT_ON(ret, dev->vbl_queue, 3 * DRM_HZ, - (((cur_vblank = atomic_read(counter)) - - *sequence) <= (1 << 23))); - - *sequence = cur_vblank; - - return ret; -} - -int radeon_driver_vblank_wait(struct drm_device *dev, unsigned int *sequence) -{ - return radeon_driver_vblank_do_wait(dev, sequence, DRM_RADEON_VBLANK_CRTC1); -} - -int radeon_driver_vblank_wait2(struct drm_device *dev, unsigned int *sequence) -{ - return radeon_driver_vblank_do_wait(dev, sequence, DRM_RADEON_VBLANK_CRTC2); + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) { + if (crtc == 0) + return RADEON_READ(R500_D1CRTC_FRAME_COUNT); + else + return RADEON_READ(R500_D2CRTC_FRAME_COUNT); + } else { + if (crtc == 0) + return RADEON_READ(RADEON_CRTC_CRNT_FRAME); + else + return RADEON_READ(RADEON_CRTC2_CRNT_FRAME); + } } /* Needs the lock as it touches the ring. @@ -234,46 +316,41 @@ int radeon_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_pr return radeon_wait_irq(dev, irqwait->irq_seq); } -void radeon_enable_interrupt(struct drm_device *dev) -{ - drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private; - - dev_priv->irq_enable_reg = RADEON_SW_INT_ENABLE; - if (dev_priv->vblank_crtc & DRM_RADEON_VBLANK_CRTC1) - dev_priv->irq_enable_reg |= RADEON_CRTC_VBLANK_MASK; - - if (dev_priv->vblank_crtc & DRM_RADEON_VBLANK_CRTC2) - dev_priv->irq_enable_reg |= RADEON_CRTC2_VBLANK_MASK; - - RADEON_WRITE(RADEON_GEN_INT_CNTL, dev_priv->irq_enable_reg); - dev_priv->irq_enabled = 1; -} - /* drm_dma.h hooks */ void radeon_driver_irq_preinstall(struct drm_device * dev) { drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private; + u32 dummy; /* Disable *all* interrupts */ + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) + RADEON_WRITE(R500_DxMODE_INT_MASK, 0); RADEON_WRITE(RADEON_GEN_INT_CNTL, 0); /* Clear bits if they're already high */ - radeon_acknowledge_irqs(dev_priv, (RADEON_SW_INT_TEST_ACK | - RADEON_CRTC_VBLANK_STAT | - RADEON_CRTC2_VBLANK_STAT)); + radeon_acknowledge_irqs(dev_priv, &dummy); } -void radeon_driver_irq_postinstall(struct drm_device * dev) +int radeon_driver_irq_postinstall(struct drm_device *dev) { drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private; + int ret; atomic_set(&dev_priv->swi_emitted, 0); DRM_INIT_WAITQUEUE(&dev_priv->swi_queue); - radeon_enable_interrupt(dev); + ret = drm_vblank_init(dev, 2); + if (ret) + return ret; + + dev->max_vblank_count = 0x001fffff; + + radeon_irq_set_state(dev, RADEON_SW_INT_ENABLE, 1); + + return 0; } void radeon_driver_irq_uninstall(struct drm_device * dev) @@ -285,6 +362,8 @@ void radeon_driver_irq_uninstall(struct drm_device * dev) dev_priv->irq_enabled = 0; + if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RS690) + RADEON_WRITE(R500_DxMODE_INT_MASK, 0); /* Disable *all* interrupts */ RADEON_WRITE(RADEON_GEN_INT_CNTL, 0); } @@ -293,18 +372,8 @@ void radeon_driver_irq_uninstall(struct drm_device * dev) int radeon_vblank_crtc_get(struct drm_device *dev) { drm_radeon_private_t *dev_priv = (drm_radeon_private_t *) dev->dev_private; - u32 flag; - u32 value; - - flag = RADEON_READ(RADEON_GEN_INT_CNTL); - value = 0; - - if (flag & RADEON_CRTC_VBLANK_MASK) - value |= DRM_RADEON_VBLANK_CRTC1; - if (flag & RADEON_CRTC2_VBLANK_MASK) - value |= DRM_RADEON_VBLANK_CRTC2; - return value; + return dev_priv->vblank_crtc; } int radeon_vblank_crtc_set(struct drm_device *dev, int64_t value) @@ -315,6 +384,5 @@ int radeon_vblank_crtc_set(struct drm_device *dev, int64_t value) return -EINVAL; } dev_priv->vblank_crtc = (unsigned int)value; - radeon_enable_interrupt(dev); return 0; } diff --git a/drivers/gpu/drm/via/via_drv.c b/drivers/gpu/drm/via/via_drv.c index 80c01cdfa37..0993b441fc4 100644 --- a/drivers/gpu/drm/via/via_drv.c +++ b/drivers/gpu/drm/via/via_drv.c @@ -40,11 +40,13 @@ static struct pci_device_id pciidlist[] = { static struct drm_driver driver = { .driver_features = DRIVER_USE_AGP | DRIVER_USE_MTRR | DRIVER_HAVE_IRQ | - DRIVER_IRQ_SHARED | DRIVER_IRQ_VBL, + DRIVER_IRQ_SHARED, .load = via_driver_load, .unload = via_driver_unload, .context_dtor = via_final_context, - .vblank_wait = via_driver_vblank_wait, + .get_vblank_counter = via_get_vblank_counter, + .enable_vblank = via_enable_vblank, + .disable_vblank = via_disable_vblank, .irq_preinstall = via_driver_irq_preinstall, .irq_postinstall = via_driver_irq_postinstall, .irq_uninstall = via_driver_irq_uninstall, @@ -59,17 +61,17 @@ static struct drm_driver driver = { .get_reg_ofs = drm_core_get_reg_ofs, .ioctls = via_ioctls, .fops = { - .owner = THIS_MODULE, - .open = drm_open, - .release = drm_release, - .ioctl = drm_ioctl, - .mmap = drm_mmap, - .poll = drm_poll, - .fasync = drm_fasync, - }, + .owner = THIS_MODULE, + .open = drm_open, + .release = drm_release, + .ioctl = drm_ioctl, + .mmap = drm_mmap, + .poll = drm_poll, + .fasync = drm_fasync, + }, .pci_driver = { - .name = DRIVER_NAME, - .id_table = pciidlist, + .name = DRIVER_NAME, + .id_table = pciidlist, }, .name = DRIVER_NAME, diff --git a/drivers/gpu/drm/via/via_drv.h b/drivers/gpu/drm/via/via_drv.h index 2daae81874c..cafcb844a22 100644 --- a/drivers/gpu/drm/via/via_drv.h +++ b/drivers/gpu/drm/via/via_drv.h @@ -75,6 +75,7 @@ typedef struct drm_via_private { struct timeval last_vblank; int last_vblank_valid; unsigned usec_per_vblank; + atomic_t vbl_received; drm_via_state_t hc_state; char pci_buf[VIA_PCI_BUF_SIZE]; const uint32_t *fire_offsets[VIA_FIRE_BUF_SIZE]; @@ -130,21 +131,24 @@ extern int via_init_context(struct drm_device * dev, int context); extern int via_final_context(struct drm_device * dev, int context); extern int via_do_cleanup_map(struct drm_device * dev); -extern int via_driver_vblank_wait(struct drm_device * dev, unsigned int *sequence); +extern u32 via_get_vblank_counter(struct drm_device *dev, int crtc); +extern int via_enable_vblank(struct drm_device *dev, int crtc); +extern void via_disable_vblank(struct drm_device *dev, int crtc); extern irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS); extern void via_driver_irq_preinstall(struct drm_device * dev); -extern void via_driver_irq_postinstall(struct drm_device * dev); +extern int via_driver_irq_postinstall(struct drm_device *dev); extern void via_driver_irq_uninstall(struct drm_device * dev); extern int via_dma_cleanup(struct drm_device * dev); extern void via_init_command_verifier(void); extern int via_driver_dma_quiescent(struct drm_device * dev); -extern void via_init_futex(drm_via_private_t * dev_priv); -extern void via_cleanup_futex(drm_via_private_t * dev_priv); -extern void via_release_futex(drm_via_private_t * dev_priv, int context); +extern void via_init_futex(drm_via_private_t *dev_priv); +extern void via_cleanup_futex(drm_via_private_t *dev_priv); +extern void via_release_futex(drm_via_private_t *dev_priv, int context); -extern void via_reclaim_buffers_locked(struct drm_device *dev, struct drm_file *file_priv); +extern void via_reclaim_buffers_locked(struct drm_device *dev, + struct drm_file *file_priv); extern void via_lastclose(struct drm_device *dev); extern void via_dmablit_handler(struct drm_device *dev, int engine, int from_irq); diff --git a/drivers/gpu/drm/via/via_irq.c b/drivers/gpu/drm/via/via_irq.c index c6bb978a110..1b966fbdb49 100644 --- a/drivers/gpu/drm/via/via_irq.c +++ b/drivers/gpu/drm/via/via_irq.c @@ -43,7 +43,7 @@ #define VIA_REG_INTERRUPT 0x200 /* VIA_REG_INTERRUPT */ -#define VIA_IRQ_GLOBAL (1 << 31) +#define VIA_IRQ_GLOBAL (1 << 31) #define VIA_IRQ_VBLANK_ENABLE (1 << 19) #define VIA_IRQ_VBLANK_PENDING (1 << 3) #define VIA_IRQ_HQV0_ENABLE (1 << 11) @@ -68,16 +68,15 @@ static maskarray_t via_pro_group_a_irqs[] = { {VIA_IRQ_HQV0_ENABLE, VIA_IRQ_HQV0_PENDING, 0x000003D0, 0x00008010, - 0x00000000}, + 0x00000000 }, {VIA_IRQ_HQV1_ENABLE, VIA_IRQ_HQV1_PENDING, 0x000013D0, 0x00008010, - 0x00000000}, + 0x00000000 }, {VIA_IRQ_DMA0_TD_ENABLE, VIA_IRQ_DMA0_TD_PENDING, VIA_PCI_DMA_CSR0, VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008}, {VIA_IRQ_DMA1_TD_ENABLE, VIA_IRQ_DMA1_TD_PENDING, VIA_PCI_DMA_CSR1, VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008}, }; -static int via_num_pro_group_a = - sizeof(via_pro_group_a_irqs) / sizeof(maskarray_t); +static int via_num_pro_group_a = ARRAY_SIZE(via_pro_group_a_irqs); static int via_irqmap_pro_group_a[] = {0, 1, -1, 2, -1, 3}; static maskarray_t via_unichrome_irqs[] = { @@ -86,14 +85,24 @@ static maskarray_t via_unichrome_irqs[] = { {VIA_IRQ_DMA1_TD_ENABLE, VIA_IRQ_DMA1_TD_PENDING, VIA_PCI_DMA_CSR1, VIA_DMA_CSR_TA | VIA_DMA_CSR_TD, 0x00000008} }; -static int via_num_unichrome = sizeof(via_unichrome_irqs) / sizeof(maskarray_t); +static int via_num_unichrome = ARRAY_SIZE(via_unichrome_irqs); static int via_irqmap_unichrome[] = {-1, -1, -1, 0, -1, 1}; + static unsigned time_diff(struct timeval *now, struct timeval *then) { return (now->tv_usec >= then->tv_usec) ? - now->tv_usec - then->tv_usec : - 1000000 - (then->tv_usec - now->tv_usec); + now->tv_usec - then->tv_usec : + 1000000 - (then->tv_usec - now->tv_usec); +} + +u32 via_get_vblank_counter(struct drm_device *dev, int crtc) +{ + drm_via_private_t *dev_priv = dev->dev_private; + if (crtc != 0) + return 0; + + return atomic_read(&dev_priv->vbl_received); } irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS) @@ -108,23 +117,22 @@ irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS) status = VIA_READ(VIA_REG_INTERRUPT); if (status & VIA_IRQ_VBLANK_PENDING) { - atomic_inc(&dev->vbl_received); - if (!(atomic_read(&dev->vbl_received) & 0x0F)) { + atomic_inc(&dev_priv->vbl_received); + if (!(atomic_read(&dev_priv->vbl_received) & 0x0F)) { do_gettimeofday(&cur_vblank); if (dev_priv->last_vblank_valid) { dev_priv->usec_per_vblank = - time_diff(&cur_vblank, - &dev_priv->last_vblank) >> 4; + time_diff(&cur_vblank, + &dev_priv->last_vblank) >> 4; } dev_priv->last_vblank = cur_vblank; dev_priv->last_vblank_valid = 1; } - if (!(atomic_read(&dev->vbl_received) & 0xFF)) { + if (!(atomic_read(&dev_priv->vbl_received) & 0xFF)) { DRM_DEBUG("US per vblank is: %u\n", dev_priv->usec_per_vblank); } - DRM_WAKEUP(&dev->vbl_queue); - drm_vbl_send_signals(dev); + drm_handle_vblank(dev, 0); handled = 1; } @@ -145,6 +153,7 @@ irqreturn_t via_driver_irq_handler(DRM_IRQ_ARGS) /* Acknowlege interrupts */ VIA_WRITE(VIA_REG_INTERRUPT, status); + if (handled) return IRQ_HANDLED; else @@ -163,31 +172,34 @@ static __inline__ void viadrv_acknowledge_irqs(drm_via_private_t * dev_priv) } } -int via_driver_vblank_wait(struct drm_device * dev, unsigned int *sequence) +int via_enable_vblank(struct drm_device *dev, int crtc) { - drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; - unsigned int cur_vblank; - int ret = 0; + drm_via_private_t *dev_priv = dev->dev_private; + u32 status; - DRM_DEBUG("\n"); - if (!dev_priv) { - DRM_ERROR("called with no initialization\n"); + if (crtc != 0) { + DRM_ERROR("%s: bad crtc %d\n", __func__, crtc); return -EINVAL; } - viadrv_acknowledge_irqs(dev_priv); + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status & VIA_IRQ_VBLANK_ENABLE); + + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) | 0x30); - /* Assume that the user has missed the current sequence number - * by about a day rather than she wants to wait for years - * using vertical blanks... - */ + return 0; +} - DRM_WAIT_ON(ret, dev->vbl_queue, 3 * DRM_HZ, - (((cur_vblank = atomic_read(&dev->vbl_received)) - - *sequence) <= (1 << 23))); +void via_disable_vblank(struct drm_device *dev, int crtc) +{ + drm_via_private_t *dev_priv = dev->dev_private; - *sequence = cur_vblank; - return ret; + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) & ~0x30); + + if (crtc != 0) + DRM_ERROR("%s: bad crtc %d\n", __func__, crtc); } static int @@ -239,6 +251,7 @@ via_driver_irq_wait(struct drm_device * dev, unsigned int irq, int force_sequenc return ret; } + /* * drm_dma.h hooks */ @@ -292,23 +305,25 @@ void via_driver_irq_preinstall(struct drm_device * dev) } } -void via_driver_irq_postinstall(struct drm_device * dev) +int via_driver_irq_postinstall(struct drm_device *dev) { drm_via_private_t *dev_priv = (drm_via_private_t *) dev->dev_private; u32 status; - DRM_DEBUG("\n"); - if (dev_priv) { - status = VIA_READ(VIA_REG_INTERRUPT); - VIA_WRITE(VIA_REG_INTERRUPT, status | VIA_IRQ_GLOBAL - | dev_priv->irq_enable_mask); + DRM_DEBUG("via_driver_irq_postinstall\n"); + if (!dev_priv) + return -EINVAL; - /* Some magic, oh for some data sheets ! */ + drm_vblank_init(dev, 1); + status = VIA_READ(VIA_REG_INTERRUPT); + VIA_WRITE(VIA_REG_INTERRUPT, status | VIA_IRQ_GLOBAL + | dev_priv->irq_enable_mask); - VIA_WRITE8(0x83d4, 0x11); - VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) | 0x30); + /* Some magic, oh for some data sheets ! */ + VIA_WRITE8(0x83d4, 0x11); + VIA_WRITE8(0x83d5, VIA_READ8(0x83d5) | 0x30); - } + return 0; } void via_driver_irq_uninstall(struct drm_device * dev) @@ -352,7 +367,8 @@ int via_wait_irq(struct drm_device *dev, void *data, struct drm_file *file_priv) switch (irqwait->request.type & ~VIA_IRQ_FLAGS_MASK) { case VIA_IRQ_RELATIVE: - irqwait->request.sequence += atomic_read(&cur_irq->irq_received); + irqwait->request.sequence += + atomic_read(&cur_irq->irq_received); irqwait->request.type &= ~_DRM_VBLANK_RELATIVE; case VIA_IRQ_ABSOLUTE: break; -- cgit v1.2.3 From 673a394b1e3b69be886ff24abfd6df97c52e8d08 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 30 Jul 2008 12:06:12 -0700 Subject: drm: Add GEM ("graphics execution manager") to i915 driver. GEM allows the creation of persistent buffer objects accessible by the graphics device through new ioctls for managing execution of commands on the device. The userland API is almost entirely driver-specific to ensure that any driver building on this model can easily map the interface to individual driver requirements. GEM is used by the 2d driver for managing its internal state allocations and will be used for pixmap storage to reduce memory consumption and enable zero-copy GLX_EXT_texture_from_pixmap, and in the 3d driver is used to enable GL_EXT_framebuffer_object and GL_ARB_pixel_buffer_object. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/Makefile | 5 +- drivers/gpu/drm/drm_agpsupport.c | 51 +- drivers/gpu/drm/drm_cache.c | 76 + drivers/gpu/drm/drm_drv.c | 4 + drivers/gpu/drm/drm_fops.c | 6 + drivers/gpu/drm/drm_gem.c | 420 ++++++ drivers/gpu/drm/drm_memory.c | 2 + drivers/gpu/drm/drm_mm.c | 5 +- drivers/gpu/drm/drm_proc.c | 135 +- drivers/gpu/drm/drm_stub.c | 10 + drivers/gpu/drm/i915/Makefile | 6 +- drivers/gpu/drm/i915/i915_dma.c | 94 +- drivers/gpu/drm/i915/i915_drv.c | 11 +- drivers/gpu/drm/i915/i915_drv.h | 253 +++- drivers/gpu/drm/i915/i915_gem.c | 2497 ++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_gem_debug.c | 201 +++ drivers/gpu/drm/i915/i915_gem_proc.c | 292 ++++ drivers/gpu/drm/i915/i915_gem_tiling.c | 256 ++++ drivers/gpu/drm/i915/i915_irq.c | 19 +- drivers/gpu/drm/i915/i915_reg.h | 37 +- 20 files changed, 4318 insertions(+), 62 deletions(-) create mode 100644 drivers/gpu/drm/drm_cache.c create mode 100644 drivers/gpu/drm/drm_gem.c create mode 100644 drivers/gpu/drm/i915/i915_gem.c create mode 100644 drivers/gpu/drm/i915/i915_gem_debug.c create mode 100644 drivers/gpu/drm/i915/i915_gem_proc.c create mode 100644 drivers/gpu/drm/i915/i915_gem_tiling.c (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/Makefile b/drivers/gpu/drm/Makefile index e9f9a97ae00..74da99495e2 100644 --- a/drivers/gpu/drm/Makefile +++ b/drivers/gpu/drm/Makefile @@ -4,8 +4,9 @@ ccflags-y := -Iinclude/drm -drm-y := drm_auth.o drm_bufs.o drm_context.o drm_dma.o drm_drawable.o \ - drm_drv.o drm_fops.o drm_ioctl.o drm_irq.o \ +drm-y := drm_auth.o drm_bufs.o drm_cache.o \ + drm_context.o drm_dma.o drm_drawable.o \ + drm_drv.o drm_fops.o drm_gem.o drm_ioctl.o drm_irq.o \ drm_lock.o drm_memory.o drm_proc.o drm_stub.o drm_vm.o \ drm_agpsupport.o drm_scatter.o ati_pcigart.o drm_pci.o \ drm_sysfs.o drm_hashtab.o drm_sman.o drm_mm.o diff --git a/drivers/gpu/drm/drm_agpsupport.c b/drivers/gpu/drm/drm_agpsupport.c index aefa5ac4c0b..2639be2db9e 100644 --- a/drivers/gpu/drm/drm_agpsupport.c +++ b/drivers/gpu/drm/drm_agpsupport.c @@ -33,6 +33,7 @@ #include "drmP.h" #include +#include #if __OS_HAS_AGP @@ -452,4 +453,52 @@ int drm_agp_unbind_memory(DRM_AGP_MEM * handle) return agp_unbind_memory(handle); } -#endif /* __OS_HAS_AGP */ +/** + * Binds a collection of pages into AGP memory at the given offset, returning + * the AGP memory structure containing them. + * + * No reference is held on the pages during this time -- it is up to the + * caller to handle that. + */ +DRM_AGP_MEM * +drm_agp_bind_pages(struct drm_device *dev, + struct page **pages, + unsigned long num_pages, + uint32_t gtt_offset) +{ + DRM_AGP_MEM *mem; + int ret, i; + + DRM_DEBUG("\n"); + + mem = drm_agp_allocate_memory(dev->agp->bridge, num_pages, + AGP_USER_MEMORY); + if (mem == NULL) { + DRM_ERROR("Failed to allocate memory for %ld pages\n", + num_pages); + return NULL; + } + + for (i = 0; i < num_pages; i++) + mem->memory[i] = phys_to_gart(page_to_phys(pages[i])); + mem->page_count = num_pages; + + mem->is_flushed = true; + ret = drm_agp_bind_memory(mem, gtt_offset / PAGE_SIZE); + if (ret != 0) { + DRM_ERROR("Failed to bind AGP memory: %d\n", ret); + agp_free_memory(mem); + return NULL; + } + + return mem; +} +EXPORT_SYMBOL(drm_agp_bind_pages); + +void drm_agp_chipset_flush(struct drm_device *dev) +{ + agp_flush_chipset(dev->agp->bridge); +} +EXPORT_SYMBOL(drm_agp_chipset_flush); + +#endif /* __OS_HAS_AGP */ diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c new file mode 100644 index 00000000000..9475f7d9901 --- /dev/null +++ b/drivers/gpu/drm/drm_cache.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright (c) 2006-2007 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include "drmP.h" + +#if defined(CONFIG_X86) +static void +drm_clflush_page(struct page *page) +{ + uint8_t *page_virtual; + unsigned int i; + + if (unlikely(page == NULL)) + return; + + page_virtual = kmap_atomic(page, KM_USER0); + for (i = 0; i < PAGE_SIZE; i += boot_cpu_data.x86_clflush_size) + clflush(page_virtual + i); + kunmap_atomic(page_virtual, KM_USER0); +} +#endif + +static void +drm_clflush_ipi_handler(void *null) +{ + wbinvd(); +} + +void +drm_clflush_pages(struct page *pages[], unsigned long num_pages) +{ + +#if defined(CONFIG_X86) + if (cpu_has_clflush) { + unsigned long i; + + mb(); + for (i = 0; i < num_pages; ++i) + drm_clflush_page(*pages++); + mb(); + + return; + } +#endif + + if (on_each_cpu(drm_clflush_ipi_handler, NULL, 1) != 0) + DRM_ERROR("Timed out waiting for cache flush.\n"); +} +EXPORT_SYMBOL(drm_clflush_pages); diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c index fb45fe7aeb8..96f416afc3f 100644 --- a/drivers/gpu/drm/drm_drv.c +++ b/drivers/gpu/drm/drm_drv.c @@ -119,6 +119,10 @@ static struct drm_ioctl_desc drm_ioctls[] = { DRM_IOCTL_DEF(DRM_IOCTL_MODESET_CTL, drm_modeset_ctl, 0), DRM_IOCTL_DEF(DRM_IOCTL_UPDATE_DRAW, drm_update_drawable_info, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), + + DRM_IOCTL_DEF(DRM_IOCTL_GEM_CLOSE, drm_gem_close_ioctl, 0), + DRM_IOCTL_DEF(DRM_IOCTL_GEM_FLINK, drm_gem_flink_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_IOCTL_GEM_OPEN, drm_gem_open_ioctl, DRM_AUTH), }; #define DRM_CORE_IOCTL_COUNT ARRAY_SIZE( drm_ioctls ) diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c index dcf8b4dc954..0d46627663b 100644 --- a/drivers/gpu/drm/drm_fops.c +++ b/drivers/gpu/drm/drm_fops.c @@ -256,6 +256,9 @@ static int drm_open_helper(struct inode *inode, struct file *filp, INIT_LIST_HEAD(&priv->lhead); + if (dev->driver->driver_features & DRIVER_GEM) + drm_gem_open(dev, priv); + if (dev->driver->open) { ret = dev->driver->open(dev, priv); if (ret < 0) @@ -400,6 +403,9 @@ int drm_release(struct inode *inode, struct file *filp) dev->driver->reclaim_buffers(dev, file_priv); } + if (dev->driver->driver_features & DRIVER_GEM) + drm_gem_release(dev, file_priv); + drm_fasync(-1, filp, 0); mutex_lock(&dev->ctxlist_mutex); diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c new file mode 100644 index 00000000000..434155b387e --- /dev/null +++ b/drivers/gpu/drm/drm_gem.c @@ -0,0 +1,420 @@ +/* + * Copyright © 2008 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include "drmP.h" + +/** @file drm_gem.c + * + * This file provides some of the base ioctls and library routines for + * the graphics memory manager implemented by each device driver. + * + * Because various devices have different requirements in terms of + * synchronization and migration strategies, implementing that is left up to + * the driver, and all that the general API provides should be generic -- + * allocating objects, reading/writing data with the cpu, freeing objects. + * Even there, platform-dependent optimizations for reading/writing data with + * the CPU mean we'll likely hook those out to driver-specific calls. However, + * the DRI2 implementation wants to have at least allocate/mmap be generic. + * + * The goal was to have swap-backed object allocation managed through + * struct file. However, file descriptors as handles to a struct file have + * two major failings: + * - Process limits prevent more than 1024 or so being used at a time by + * default. + * - Inability to allocate high fds will aggravate the X Server's select() + * handling, and likely that of many GL client applications as well. + * + * This led to a plan of using our own integer IDs (called handles, following + * DRM terminology) to mimic fds, and implement the fd syscalls we need as + * ioctls. The objects themselves will still include the struct file so + * that we can transition to fds if the required kernel infrastructure shows + * up at a later date, and as our interface with shmfs for memory allocation. + */ + +/** + * Initialize the GEM device fields + */ + +int +drm_gem_init(struct drm_device *dev) +{ + spin_lock_init(&dev->object_name_lock); + idr_init(&dev->object_name_idr); + atomic_set(&dev->object_count, 0); + atomic_set(&dev->object_memory, 0); + atomic_set(&dev->pin_count, 0); + atomic_set(&dev->pin_memory, 0); + atomic_set(&dev->gtt_count, 0); + atomic_set(&dev->gtt_memory, 0); + return 0; +} + +/** + * Allocate a GEM object of the specified size with shmfs backing store + */ +struct drm_gem_object * +drm_gem_object_alloc(struct drm_device *dev, size_t size) +{ + struct drm_gem_object *obj; + + BUG_ON((size & (PAGE_SIZE - 1)) != 0); + + obj = kcalloc(1, sizeof(*obj), GFP_KERNEL); + + obj->dev = dev; + obj->filp = shmem_file_setup("drm mm object", size, 0); + if (IS_ERR(obj->filp)) { + kfree(obj); + return NULL; + } + + kref_init(&obj->refcount); + kref_init(&obj->handlecount); + obj->size = size; + if (dev->driver->gem_init_object != NULL && + dev->driver->gem_init_object(obj) != 0) { + fput(obj->filp); + kfree(obj); + return NULL; + } + atomic_inc(&dev->object_count); + atomic_add(obj->size, &dev->object_memory); + return obj; +} +EXPORT_SYMBOL(drm_gem_object_alloc); + +/** + * Removes the mapping from handle to filp for this object. + */ +static int +drm_gem_handle_delete(struct drm_file *filp, int handle) +{ + struct drm_device *dev; + struct drm_gem_object *obj; + + /* This is gross. The idr system doesn't let us try a delete and + * return an error code. It just spews if you fail at deleting. + * So, we have to grab a lock around finding the object and then + * doing the delete on it and dropping the refcount, or the user + * could race us to double-decrement the refcount and cause a + * use-after-free later. Given the frequency of our handle lookups, + * we may want to use ida for number allocation and a hash table + * for the pointers, anyway. + */ + spin_lock(&filp->table_lock); + + /* Check if we currently have a reference on the object */ + obj = idr_find(&filp->object_idr, handle); + if (obj == NULL) { + spin_unlock(&filp->table_lock); + return -EINVAL; + } + dev = obj->dev; + + /* Release reference and decrement refcount. */ + idr_remove(&filp->object_idr, handle); + spin_unlock(&filp->table_lock); + + mutex_lock(&dev->struct_mutex); + drm_gem_object_handle_unreference(obj); + mutex_unlock(&dev->struct_mutex); + + return 0; +} + +/** + * Create a handle for this object. This adds a handle reference + * to the object, which includes a regular reference count. Callers + * will likely want to dereference the object afterwards. + */ +int +drm_gem_handle_create(struct drm_file *file_priv, + struct drm_gem_object *obj, + int *handlep) +{ + int ret; + + /* + * Get the user-visible handle using idr. + */ +again: + /* ensure there is space available to allocate a handle */ + if (idr_pre_get(&file_priv->object_idr, GFP_KERNEL) == 0) + return -ENOMEM; + + /* do the allocation under our spinlock */ + spin_lock(&file_priv->table_lock); + ret = idr_get_new_above(&file_priv->object_idr, obj, 1, handlep); + spin_unlock(&file_priv->table_lock); + if (ret == -EAGAIN) + goto again; + + if (ret != 0) + return ret; + + drm_gem_object_handle_reference(obj); + return 0; +} +EXPORT_SYMBOL(drm_gem_handle_create); + +/** Returns a reference to the object named by the handle. */ +struct drm_gem_object * +drm_gem_object_lookup(struct drm_device *dev, struct drm_file *filp, + int handle) +{ + struct drm_gem_object *obj; + + spin_lock(&filp->table_lock); + + /* Check if we currently have a reference on the object */ + obj = idr_find(&filp->object_idr, handle); + if (obj == NULL) { + spin_unlock(&filp->table_lock); + return NULL; + } + + drm_gem_object_reference(obj); + + spin_unlock(&filp->table_lock); + + return obj; +} +EXPORT_SYMBOL(drm_gem_object_lookup); + +/** + * Releases the handle to an mm object. + */ +int +drm_gem_close_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_gem_close *args = data; + int ret; + + if (!(dev->driver->driver_features & DRIVER_GEM)) + return -ENODEV; + + ret = drm_gem_handle_delete(file_priv, args->handle); + + return ret; +} + +/** + * Create a global name for an object, returning the name. + * + * Note that the name does not hold a reference; when the object + * is freed, the name goes away. + */ +int +drm_gem_flink_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_gem_flink *args = data; + struct drm_gem_object *obj; + int ret; + + if (!(dev->driver->driver_features & DRIVER_GEM)) + return -ENODEV; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EINVAL; + +again: + if (idr_pre_get(&dev->object_name_idr, GFP_KERNEL) == 0) + return -ENOMEM; + + spin_lock(&dev->object_name_lock); + if (obj->name) { + spin_unlock(&dev->object_name_lock); + return -EEXIST; + } + ret = idr_get_new_above(&dev->object_name_idr, obj, 1, + &obj->name); + spin_unlock(&dev->object_name_lock); + if (ret == -EAGAIN) + goto again; + + if (ret != 0) { + mutex_lock(&dev->struct_mutex); + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + return ret; + } + + /* + * Leave the reference from the lookup around as the + * name table now holds one + */ + args->name = (uint64_t) obj->name; + + return 0; +} + +/** + * Open an object using the global name, returning a handle and the size. + * + * This handle (of course) holds a reference to the object, so the object + * will not go away until the handle is deleted. + */ +int +drm_gem_open_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_gem_open *args = data; + struct drm_gem_object *obj; + int ret; + int handle; + + if (!(dev->driver->driver_features & DRIVER_GEM)) + return -ENODEV; + + spin_lock(&dev->object_name_lock); + obj = idr_find(&dev->object_name_idr, (int) args->name); + if (obj) + drm_gem_object_reference(obj); + spin_unlock(&dev->object_name_lock); + if (!obj) + return -ENOENT; + + ret = drm_gem_handle_create(file_priv, obj, &handle); + mutex_lock(&dev->struct_mutex); + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + if (ret) + return ret; + + args->handle = handle; + args->size = obj->size; + + return 0; +} + +/** + * Called at device open time, sets up the structure for handling refcounting + * of mm objects. + */ +void +drm_gem_open(struct drm_device *dev, struct drm_file *file_private) +{ + idr_init(&file_private->object_idr); + spin_lock_init(&file_private->table_lock); +} + +/** + * Called at device close to release the file's + * handle references on objects. + */ +static int +drm_gem_object_release_handle(int id, void *ptr, void *data) +{ + struct drm_gem_object *obj = ptr; + + drm_gem_object_handle_unreference(obj); + + return 0; +} + +/** + * Called at close time when the filp is going away. + * + * Releases any remaining references on objects by this filp. + */ +void +drm_gem_release(struct drm_device *dev, struct drm_file *file_private) +{ + mutex_lock(&dev->struct_mutex); + idr_for_each(&file_private->object_idr, + &drm_gem_object_release_handle, NULL); + + idr_destroy(&file_private->object_idr); + mutex_unlock(&dev->struct_mutex); +} + +/** + * Called after the last reference to the object has been lost. + * + * Frees the object + */ +void +drm_gem_object_free(struct kref *kref) +{ + struct drm_gem_object *obj = (struct drm_gem_object *) kref; + struct drm_device *dev = obj->dev; + + BUG_ON(!mutex_is_locked(&dev->struct_mutex)); + + if (dev->driver->gem_free_object != NULL) + dev->driver->gem_free_object(obj); + + fput(obj->filp); + atomic_dec(&dev->object_count); + atomic_sub(obj->size, &dev->object_memory); + kfree(obj); +} +EXPORT_SYMBOL(drm_gem_object_free); + +/** + * Called after the last handle to the object has been closed + * + * Removes any name for the object. Note that this must be + * called before drm_gem_object_free or we'll be touching + * freed memory + */ +void +drm_gem_object_handle_free(struct kref *kref) +{ + struct drm_gem_object *obj = container_of(kref, + struct drm_gem_object, + handlecount); + struct drm_device *dev = obj->dev; + + /* Remove any name for this object */ + spin_lock(&dev->object_name_lock); + if (obj->name) { + idr_remove(&dev->object_name_idr, obj->name); + spin_unlock(&dev->object_name_lock); + /* + * The object name held a reference to this object, drop + * that now. + */ + drm_gem_object_unreference(obj); + } else + spin_unlock(&dev->object_name_lock); + +} +EXPORT_SYMBOL(drm_gem_object_handle_free); + diff --git a/drivers/gpu/drm/drm_memory.c b/drivers/gpu/drm/drm_memory.c index 0177012845c..803bc9e7ce3 100644 --- a/drivers/gpu/drm/drm_memory.c +++ b/drivers/gpu/drm/drm_memory.c @@ -133,6 +133,7 @@ int drm_free_agp(DRM_AGP_MEM * handle, int pages) { return drm_agp_free_memory(handle) ? 0 : -EINVAL; } +EXPORT_SYMBOL(drm_free_agp); /** Wrapper around agp_bind_memory() */ int drm_bind_agp(DRM_AGP_MEM * handle, unsigned int start) @@ -145,6 +146,7 @@ int drm_unbind_agp(DRM_AGP_MEM * handle) { return drm_agp_unbind_memory(handle); } +EXPORT_SYMBOL(drm_unbind_agp); #else /* __OS_HAS_AGP */ static inline void *agp_remap(unsigned long offset, unsigned long size, diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c index dcff9e9b52e..217ad7dc707 100644 --- a/drivers/gpu/drm/drm_mm.c +++ b/drivers/gpu/drm/drm_mm.c @@ -169,6 +169,7 @@ struct drm_mm_node *drm_mm_get_block(struct drm_mm_node * parent, return child; } +EXPORT_SYMBOL(drm_mm_get_block); /* * Put a block. Merge with the previous and / or next block if they are free. @@ -217,6 +218,7 @@ void drm_mm_put_block(struct drm_mm_node * cur) drm_free(cur, sizeof(*cur), DRM_MEM_MM); } } +EXPORT_SYMBOL(drm_mm_put_block); struct drm_mm_node *drm_mm_search_free(const struct drm_mm * mm, unsigned long size, @@ -265,6 +267,7 @@ int drm_mm_clean(struct drm_mm * mm) return (head->next->next == head); } +EXPORT_SYMBOL(drm_mm_search_free); int drm_mm_init(struct drm_mm * mm, unsigned long start, unsigned long size) { @@ -273,7 +276,7 @@ int drm_mm_init(struct drm_mm * mm, unsigned long start, unsigned long size) return drm_mm_create_tail_node(mm, start, size); } - +EXPORT_SYMBOL(drm_mm_init); void drm_mm_takedown(struct drm_mm * mm) { diff --git a/drivers/gpu/drm/drm_proc.c b/drivers/gpu/drm/drm_proc.c index 93b1e0475c9..d490db4c0de 100644 --- a/drivers/gpu/drm/drm_proc.c +++ b/drivers/gpu/drm/drm_proc.c @@ -49,6 +49,10 @@ static int drm_queues_info(char *buf, char **start, off_t offset, int request, int *eof, void *data); static int drm_bufs_info(char *buf, char **start, off_t offset, int request, int *eof, void *data); +static int drm_gem_name_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data); +static int drm_gem_object_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data); #if DRM_DEBUG_CODE static int drm_vma_info(char *buf, char **start, off_t offset, int request, int *eof, void *data); @@ -60,13 +64,16 @@ static int drm_vma_info(char *buf, char **start, off_t offset, static struct drm_proc_list { const char *name; /**< file name */ int (*f) (char *, char **, off_t, int, int *, void *); /**< proc callback*/ + u32 driver_features; /**< Required driver features for this entry */ } drm_proc_list[] = { - {"name", drm_name_info}, - {"mem", drm_mem_info}, - {"vm", drm_vm_info}, - {"clients", drm_clients_info}, - {"queues", drm_queues_info}, - {"bufs", drm_bufs_info}, + {"name", drm_name_info, 0}, + {"mem", drm_mem_info, 0}, + {"vm", drm_vm_info, 0}, + {"clients", drm_clients_info, 0}, + {"queues", drm_queues_info, 0}, + {"bufs", drm_bufs_info, 0}, + {"gem_names", drm_gem_name_info, DRIVER_GEM}, + {"gem_objects", drm_gem_object_info, DRIVER_GEM}, #if DRM_DEBUG_CODE {"vma", drm_vma_info}, #endif @@ -90,8 +97,9 @@ static struct drm_proc_list { int drm_proc_init(struct drm_minor *minor, int minor_id, struct proc_dir_entry *root) { + struct drm_device *dev = minor->dev; struct proc_dir_entry *ent; - int i, j; + int i, j, ret; char name[64]; sprintf(name, "%d", minor_id); @@ -102,23 +110,42 @@ int drm_proc_init(struct drm_minor *minor, int minor_id, } for (i = 0; i < DRM_PROC_ENTRIES; i++) { + u32 features = drm_proc_list[i].driver_features; + + if (features != 0 && + (dev->driver->driver_features & features) != features) + continue; + ent = create_proc_entry(drm_proc_list[i].name, S_IFREG | S_IRUGO, minor->dev_root); if (!ent) { DRM_ERROR("Cannot create /proc/dri/%s/%s\n", name, drm_proc_list[i].name); - for (j = 0; j < i; j++) - remove_proc_entry(drm_proc_list[i].name, - minor->dev_root); - remove_proc_entry(name, root); - minor->dev_root = NULL; - return -1; + ret = -1; + goto fail; } ent->read_proc = drm_proc_list[i].f; ent->data = minor; } + if (dev->driver->proc_init) { + ret = dev->driver->proc_init(minor); + if (ret) { + DRM_ERROR("DRM: Driver failed to initialize " + "/proc/dri.\n"); + goto fail; + } + } + return 0; + fail: + + for (j = 0; j < i; j++) + remove_proc_entry(drm_proc_list[i].name, + minor->dev_root); + remove_proc_entry(name, root); + minor->dev_root = NULL; + return ret; } /** @@ -133,12 +160,16 @@ int drm_proc_init(struct drm_minor *minor, int minor_id, */ int drm_proc_cleanup(struct drm_minor *minor, struct proc_dir_entry *root) { + struct drm_device *dev = minor->dev; int i; char name[64]; if (!root || !minor->dev_root) return 0; + if (dev->driver->proc_cleanup) + dev->driver->proc_cleanup(minor); + for (i = 0; i < DRM_PROC_ENTRIES; i++) remove_proc_entry(drm_proc_list[i].name, minor->dev_root); sprintf(name, "%d", minor->index); @@ -480,6 +511,84 @@ static int drm_clients_info(char *buf, char **start, off_t offset, return ret; } +struct drm_gem_name_info_data { + int len; + char *buf; + int eof; +}; + +static int drm_gem_one_name_info(int id, void *ptr, void *data) +{ + struct drm_gem_object *obj = ptr; + struct drm_gem_name_info_data *nid = data; + + DRM_INFO("name %d size %d\n", obj->name, obj->size); + if (nid->eof) + return 0; + + nid->len += sprintf(&nid->buf[nid->len], + "%6d%9d%8d%9d\n", + obj->name, obj->size, + atomic_read(&obj->handlecount.refcount), + atomic_read(&obj->refcount.refcount)); + if (nid->len > DRM_PROC_LIMIT) { + nid->eof = 1; + return 0; + } + return 0; +} + +static int drm_gem_name_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + struct drm_gem_name_info_data nid; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + nid.len = sprintf(buf, " name size handles refcount\n"); + nid.buf = buf; + nid.eof = 0; + idr_for_each(&dev->object_name_idr, drm_gem_one_name_info, &nid); + + *start = &buf[offset]; + *eof = 0; + if (nid.len > request + offset) + return request; + *eof = 1; + return nid.len - offset; +} + +static int drm_gem_object_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("%d objects\n", atomic_read(&dev->object_count)); + DRM_PROC_PRINT("%d object bytes\n", atomic_read(&dev->object_memory)); + DRM_PROC_PRINT("%d pinned\n", atomic_read(&dev->pin_count)); + DRM_PROC_PRINT("%d pin bytes\n", atomic_read(&dev->pin_memory)); + DRM_PROC_PRINT("%d gtt bytes\n", atomic_read(&dev->gtt_memory)); + DRM_PROC_PRINT("%d gtt total\n", dev->gtt_total); + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + #if DRM_DEBUG_CODE static int drm__vma_info(char *buf, char **start, off_t offset, int request, diff --git a/drivers/gpu/drm/drm_stub.c b/drivers/gpu/drm/drm_stub.c index c2f584f3b46..82f4657b887 100644 --- a/drivers/gpu/drm/drm_stub.c +++ b/drivers/gpu/drm/drm_stub.c @@ -152,6 +152,15 @@ static int drm_fill_in_dev(struct drm_device * dev, struct pci_dev *pdev, goto error_out_unreg; } + if (driver->driver_features & DRIVER_GEM) { + retcode = drm_gem_init(dev); + if (retcode) { + DRM_ERROR("Cannot initialize graphics execution " + "manager (GEM)\n"); + goto error_out_unreg; + } + } + return 0; error_out_unreg: @@ -317,6 +326,7 @@ int drm_put_dev(struct drm_device * dev) int drm_put_minor(struct drm_minor **minor_p) { struct drm_minor *minor = *minor_p; + DRM_DEBUG("release secondary minor %d\n", minor->index); if (minor->type == DRM_MINOR_LEGACY) diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index c4bbda662e2..5ba78e4fd2b 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -4,7 +4,11 @@ ccflags-y := -Iinclude/drm i915-y := i915_drv.o i915_dma.o i915_irq.o i915_mem.o i915_opregion.o \ - i915_suspend.o + i915_suspend.o \ + i915_gem.o \ + i915_gem_debug.o \ + i915_gem_proc.o \ + i915_gem_tiling.o i915-$(CONFIG_COMPAT) += i915_ioc32.o diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 63c6803d471..f167ff68a4b 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -170,24 +170,31 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) dev_priv->sarea_priv = (drm_i915_sarea_t *) ((u8 *) dev_priv->sarea->handle + init->sarea_priv_offset); - dev_priv->ring.Start = init->ring_start; - dev_priv->ring.End = init->ring_end; - dev_priv->ring.Size = init->ring_size; - dev_priv->ring.tail_mask = dev_priv->ring.Size - 1; + if (init->ring_size != 0) { + if (dev_priv->ring.ring_obj != NULL) { + i915_dma_cleanup(dev); + DRM_ERROR("Client tried to initialize ringbuffer in " + "GEM mode\n"); + return -EINVAL; + } - dev_priv->ring.map.offset = init->ring_start; - dev_priv->ring.map.size = init->ring_size; - dev_priv->ring.map.type = 0; - dev_priv->ring.map.flags = 0; - dev_priv->ring.map.mtrr = 0; + dev_priv->ring.Size = init->ring_size; + dev_priv->ring.tail_mask = dev_priv->ring.Size - 1; - drm_core_ioremap(&dev_priv->ring.map, dev); + dev_priv->ring.map.offset = init->ring_start; + dev_priv->ring.map.size = init->ring_size; + dev_priv->ring.map.type = 0; + dev_priv->ring.map.flags = 0; + dev_priv->ring.map.mtrr = 0; - if (dev_priv->ring.map.handle == NULL) { - i915_dma_cleanup(dev); - DRM_ERROR("can not ioremap virtual address for" - " ring buffer\n"); - return -ENOMEM; + drm_core_ioremap(&dev_priv->ring.map, dev); + + if (dev_priv->ring.map.handle == NULL) { + i915_dma_cleanup(dev); + DRM_ERROR("can not ioremap virtual address for" + " ring buffer\n"); + return -ENOMEM; + } } dev_priv->ring.virtual_start = dev_priv->ring.map.handle; @@ -377,9 +384,10 @@ static int i915_emit_cmds(struct drm_device * dev, int __user * buffer, int dwor return 0; } -static int i915_emit_box(struct drm_device * dev, - struct drm_clip_rect __user * boxes, - int i, int DR1, int DR4) +int +i915_emit_box(struct drm_device *dev, + struct drm_clip_rect __user *boxes, + int i, int DR1, int DR4) { drm_i915_private_t *dev_priv = dev->dev_private; struct drm_clip_rect box; @@ -681,6 +689,9 @@ static int i915_getparam(struct drm_device *dev, void *data, case I915_PARAM_LAST_DISPATCH: value = READ_BREADCRUMB(dev_priv); break; + case I915_PARAM_HAS_GEM: + value = 1; + break; default: DRM_ERROR("Unknown parameter %d\n", param->param); return -EINVAL; @@ -784,6 +795,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) memset(dev_priv, 0, sizeof(drm_i915_private_t)); dev->dev_private = (void *)dev_priv; + dev_priv->dev = dev; /* Add register map (needed for suspend/resume) */ base = drm_get_resource_start(dev, mmio_bar); @@ -793,6 +805,8 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) _DRM_KERNEL | _DRM_DRIVER, &dev_priv->mmio_map); + i915_gem_load(dev); + /* Init HWS */ if (!I915_NEED_GFX_HWS(dev)) { ret = i915_init_phys_hws(dev); @@ -838,6 +852,25 @@ int i915_driver_unload(struct drm_device *dev) return 0; } +int i915_driver_open(struct drm_device *dev, struct drm_file *file_priv) +{ + struct drm_i915_file_private *i915_file_priv; + + DRM_DEBUG("\n"); + i915_file_priv = (struct drm_i915_file_private *) + drm_alloc(sizeof(*i915_file_priv), DRM_MEM_FILES); + + if (!i915_file_priv) + return -ENOMEM; + + file_priv->driver_priv = i915_file_priv; + + i915_file_priv->mm.last_gem_seqno = 0; + i915_file_priv->mm.last_gem_throttle_seqno = 0; + + return 0; +} + void i915_driver_lastclose(struct drm_device * dev) { drm_i915_private_t *dev_priv = dev->dev_private; @@ -845,6 +878,8 @@ void i915_driver_lastclose(struct drm_device * dev) if (!dev_priv) return; + i915_gem_lastclose(dev); + if (dev_priv->agp_heap) i915_mem_takedown(&(dev_priv->agp_heap)); @@ -857,6 +892,13 @@ void i915_driver_preclose(struct drm_device * dev, struct drm_file *file_priv) i915_mem_release(dev, file_priv, dev_priv->agp_heap); } +void i915_driver_postclose(struct drm_device *dev, struct drm_file *file_priv) +{ + struct drm_i915_file_private *i915_file_priv = file_priv->driver_priv; + + drm_free(i915_file_priv, sizeof(*i915_file_priv), DRM_MEM_FILES); +} + struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF(DRM_I915_INIT, i915_dma_init, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_FLUSH, i915_flush_ioctl, DRM_AUTH), @@ -875,6 +917,22 @@ struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF(DRM_I915_GET_VBLANK_PIPE, i915_vblank_pipe_get, DRM_AUTH ), DRM_IOCTL_DEF(DRM_I915_VBLANK_SWAP, i915_vblank_swap, DRM_AUTH), DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), + DRM_IOCTL_DEF(DRM_I915_GEM_UNPIN, i915_gem_unpin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), + DRM_IOCTL_DEF(DRM_I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_THROTTLE, i915_gem_throttle_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_ENTERVT, i915_gem_entervt_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_LEAVEVT, i915_gem_leavevt_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_CREATE, i915_gem_create_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_PREAD, i915_gem_pread_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_PWRITE, i915_gem_pwrite_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_MMAP, i915_gem_mmap_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_SET_DOMAIN, i915_gem_set_domain_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_SW_FINISH, i915_gem_sw_finish_ioctl, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_SET_TILING, i915_gem_set_tiling, 0), + DRM_IOCTL_DEF(DRM_I915_GEM_GET_TILING, i915_gem_get_tiling, 0), }; int i915_max_ioctl = DRM_ARRAY_SIZE(i915_ioctls); diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 37af03f4db3..a80ead21528 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -85,12 +85,15 @@ static struct drm_driver driver = { /* don't use mtrr's here, the Xserver or user space app should * deal with them for intel hardware. */ - .driver_features = DRIVER_USE_AGP | DRIVER_REQUIRE_AGP | - DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED, + .driver_features = + DRIVER_USE_AGP | DRIVER_REQUIRE_AGP | /* DRIVER_USE_MTRR |*/ + DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_GEM, .load = i915_driver_load, .unload = i915_driver_unload, + .open = i915_driver_open, .lastclose = i915_driver_lastclose, .preclose = i915_driver_preclose, + .postclose = i915_driver_postclose, .suspend = i915_suspend, .resume = i915_resume, .device_is_agp = i915_driver_device_is_agp, @@ -104,6 +107,10 @@ static struct drm_driver driver = { .reclaim_buffers = drm_core_reclaim_buffers, .get_map_ofs = drm_core_get_map_ofs, .get_reg_ofs = drm_core_get_reg_ofs, + .proc_init = i915_gem_proc_init, + .proc_cleanup = i915_gem_proc_cleanup, + .gem_init_object = i915_gem_init_object, + .gem_free_object = i915_gem_free_object, .ioctls = i915_ioctls, .fops = { .owner = THIS_MODULE, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d1a02bead45..87b071ab864 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -39,7 +39,7 @@ #define DRIVER_NAME "i915" #define DRIVER_DESC "Intel Graphics" -#define DRIVER_DATE "20060119" +#define DRIVER_DATE "20080730" enum pipe { PIPE_A = 0, @@ -60,16 +60,23 @@ enum pipe { #define DRIVER_MINOR 6 #define DRIVER_PATCHLEVEL 0 +#define WATCH_COHERENCY 0 +#define WATCH_BUF 0 +#define WATCH_EXEC 0 +#define WATCH_LRU 0 +#define WATCH_RELOC 0 +#define WATCH_INACTIVE 0 +#define WATCH_PWRITE 0 + typedef struct _drm_i915_ring_buffer { int tail_mask; - unsigned long Start; - unsigned long End; unsigned long Size; u8 *virtual_start; int head; int tail; int space; drm_local_map_t map; + struct drm_gem_object *ring_obj; } drm_i915_ring_buffer_t; struct mem_block { @@ -101,6 +108,8 @@ struct intel_opregion { }; typedef struct drm_i915_private { + struct drm_device *dev; + drm_local_map_t *sarea; drm_local_map_t *mmio_map; @@ -113,6 +122,7 @@ typedef struct drm_i915_private { uint32_t counter; unsigned int status_gfx_addr; drm_local_map_t hws_map; + struct drm_gem_object *hws_obj; unsigned int cpp; int back_offset; @@ -122,7 +132,6 @@ typedef struct drm_i915_private { wait_queue_head_t irq_queue; atomic_t irq_received; - atomic_t irq_emitted; /** Protects user_irq_refcount and irq_mask_reg */ spinlock_t user_irq_lock; /** Refcount for i915_user_irq_get() versus i915_user_irq_put(). */ @@ -230,8 +239,174 @@ typedef struct drm_i915_private { u8 saveDACMASK; u8 saveDACDATA[256*3]; /* 256 3-byte colors */ u8 saveCR[37]; + + struct { + struct drm_mm gtt_space; + + /** + * List of objects currently involved in rendering from the + * ringbuffer. + * + * A reference is held on the buffer while on this list. + */ + struct list_head active_list; + + /** + * List of objects which are not in the ringbuffer but which + * still have a write_domain which needs to be flushed before + * unbinding. + * + * A reference is held on the buffer while on this list. + */ + struct list_head flushing_list; + + /** + * LRU list of objects which are not in the ringbuffer and + * are ready to unbind, but are still in the GTT. + * + * A reference is not held on the buffer while on this list, + * as merely being GTT-bound shouldn't prevent its being + * freed, and we'll pull it off the list in the free path. + */ + struct list_head inactive_list; + + /** + * List of breadcrumbs associated with GPU requests currently + * outstanding. + */ + struct list_head request_list; + + /** + * We leave the user IRQ off as much as possible, + * but this means that requests will finish and never + * be retired once the system goes idle. Set a timer to + * fire periodically while the ring is running. When it + * fires, go retire requests. + */ + struct delayed_work retire_work; + + uint32_t next_gem_seqno; + + /** + * Waiting sequence number, if any + */ + uint32_t waiting_gem_seqno; + + /** + * Last seq seen at irq time + */ + uint32_t irq_gem_seqno; + + /** + * Flag if the X Server, and thus DRM, is not currently in + * control of the device. + * + * This is set between LeaveVT and EnterVT. It needs to be + * replaced with a semaphore. It also needs to be + * transitioned away from for kernel modesetting. + */ + int suspended; + + /** + * Flag if the hardware appears to be wedged. + * + * This is set when attempts to idle the device timeout. + * It prevents command submission from occuring and makes + * every pending request fail + */ + int wedged; + + /** Bit 6 swizzling required for X tiling */ + uint32_t bit_6_swizzle_x; + /** Bit 6 swizzling required for Y tiling */ + uint32_t bit_6_swizzle_y; + } mm; } drm_i915_private_t; +/** driver private structure attached to each drm_gem_object */ +struct drm_i915_gem_object { + struct drm_gem_object *obj; + + /** Current space allocated to this object in the GTT, if any. */ + struct drm_mm_node *gtt_space; + + /** This object's place on the active/flushing/inactive lists */ + struct list_head list; + + /** + * This is set if the object is on the active or flushing lists + * (has pending rendering), and is not set if it's on inactive (ready + * to be unbound). + */ + int active; + + /** + * This is set if the object has been written to since last bound + * to the GTT + */ + int dirty; + + /** AGP memory structure for our GTT binding. */ + DRM_AGP_MEM *agp_mem; + + struct page **page_list; + + /** + * Current offset of the object in GTT space. + * + * This is the same as gtt_space->start + */ + uint32_t gtt_offset; + + /** Boolean whether this object has a valid gtt offset. */ + int gtt_bound; + + /** How many users have pinned this object in GTT space */ + int pin_count; + + /** Breadcrumb of last rendering to the buffer. */ + uint32_t last_rendering_seqno; + + /** Current tiling mode for the object. */ + uint32_t tiling_mode; + + /** + * Flagging of which individual pages are valid in GEM_DOMAIN_CPU when + * GEM_DOMAIN_CPU is not in the object's read domain. + */ + uint8_t *page_cpu_valid; +}; + +/** + * Request queue structure. + * + * The request queue allows us to note sequence numbers that have been emitted + * and may be associated with active buffers to be retired. + * + * By keeping this list, we can avoid having to do questionable + * sequence-number comparisons on buffer last_rendering_seqnos, and associate + * an emission time with seqnos for tracking how far ahead of the GPU we are. + */ +struct drm_i915_gem_request { + /** GEM sequence number associated with this request. */ + uint32_t seqno; + + /** Time at which this request was emitted, in jiffies. */ + unsigned long emitted_jiffies; + + /** Cache domains that were flushed at the start of the request. */ + uint32_t flush_domains; + + struct list_head list; +}; + +struct drm_i915_file_private { + struct { + uint32_t last_gem_seqno; + uint32_t last_gem_throttle_seqno; + } mm; +}; + extern struct drm_ioctl_desc i915_ioctls[]; extern int i915_max_ioctl; @@ -239,18 +414,26 @@ extern int i915_max_ioctl; extern void i915_kernel_lost_context(struct drm_device * dev); extern int i915_driver_load(struct drm_device *, unsigned long flags); extern int i915_driver_unload(struct drm_device *); +extern int i915_driver_open(struct drm_device *dev, struct drm_file *file_priv); extern void i915_driver_lastclose(struct drm_device * dev); extern void i915_driver_preclose(struct drm_device *dev, struct drm_file *file_priv); +extern void i915_driver_postclose(struct drm_device *dev, + struct drm_file *file_priv); extern int i915_driver_device_is_agp(struct drm_device * dev); extern long i915_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); +extern int i915_emit_box(struct drm_device *dev, + struct drm_clip_rect __user *boxes, + int i, int DR1, int DR4); /* i915_irq.c */ extern int i915_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int i915_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_priv); +void i915_user_irq_get(struct drm_device *dev); +void i915_user_irq_put(struct drm_device *dev); extern irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS); extern void i915_driver_irq_preinstall(struct drm_device * dev); @@ -279,6 +462,67 @@ extern int i915_mem_destroy_heap(struct drm_device *dev, void *data, extern void i915_mem_takedown(struct mem_block **heap); extern void i915_mem_release(struct drm_device * dev, struct drm_file *file_priv, struct mem_block *heap); +/* i915_gem.c */ +int i915_gem_init_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_create_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_pread_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_pwrite_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_mmap_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_set_domain_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_execbuffer(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_pin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_unpin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_busy_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_throttle_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_entervt_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_set_tiling(struct drm_device *dev, void *data, + struct drm_file *file_priv); +int i915_gem_get_tiling(struct drm_device *dev, void *data, + struct drm_file *file_priv); +void i915_gem_load(struct drm_device *dev); +int i915_gem_proc_init(struct drm_minor *minor); +void i915_gem_proc_cleanup(struct drm_minor *minor); +int i915_gem_init_object(struct drm_gem_object *obj); +void i915_gem_free_object(struct drm_gem_object *obj); +int i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment); +void i915_gem_object_unpin(struct drm_gem_object *obj); +void i915_gem_lastclose(struct drm_device *dev); +uint32_t i915_get_gem_seqno(struct drm_device *dev); +void i915_gem_retire_requests(struct drm_device *dev); +void i915_gem_retire_work_handler(struct work_struct *work); +void i915_gem_clflush_object(struct drm_gem_object *obj); + +/* i915_gem_tiling.c */ +void i915_gem_detect_bit_6_swizzle(struct drm_device *dev); + +/* i915_gem_debug.c */ +void i915_gem_dump_object(struct drm_gem_object *obj, int len, + const char *where, uint32_t mark); +#if WATCH_INACTIVE +void i915_verify_inactive(struct drm_device *dev, char *file, int line); +#else +#define i915_verify_inactive(dev, file, line) +#endif +void i915_gem_object_check_coherency(struct drm_gem_object *obj, int handle); +void i915_gem_dump_object(struct drm_gem_object *obj, int len, + const char *where, uint32_t mark); +void i915_dump_lru(struct drm_device *dev, const char *where); /* i915_suspend.c */ extern int i915_save_state(struct drm_device *dev); @@ -347,6 +591,7 @@ extern void opregion_enable_asle(struct drm_device *dev); */ #define READ_HWSP(dev_priv, reg) (((volatile u32*)(dev_priv->hw_status_page))[reg]) #define READ_BREADCRUMB(dev_priv) READ_HWSP(dev_priv, 5) +#define I915_GEM_HWS_INDEX 0x10 extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c new file mode 100644 index 00000000000..90ae8a0369f --- /dev/null +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -0,0 +1,2497 @@ +/* + * Copyright © 2008 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt + * + */ + +#include "drmP.h" +#include "drm.h" +#include "i915_drm.h" +#include "i915_drv.h" +#include + +static int +i915_gem_object_set_domain(struct drm_gem_object *obj, + uint32_t read_domains, + uint32_t write_domain); +static int +i915_gem_object_set_domain_range(struct drm_gem_object *obj, + uint64_t offset, + uint64_t size, + uint32_t read_domains, + uint32_t write_domain); +static int +i915_gem_set_domain(struct drm_gem_object *obj, + struct drm_file *file_priv, + uint32_t read_domains, + uint32_t write_domain); +static int i915_gem_object_get_page_list(struct drm_gem_object *obj); +static void i915_gem_object_free_page_list(struct drm_gem_object *obj); +static int i915_gem_object_wait_rendering(struct drm_gem_object *obj); + +int +i915_gem_init_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_init *args = data; + + mutex_lock(&dev->struct_mutex); + + if (args->gtt_start >= args->gtt_end || + (args->gtt_start & (PAGE_SIZE - 1)) != 0 || + (args->gtt_end & (PAGE_SIZE - 1)) != 0) { + mutex_unlock(&dev->struct_mutex); + return -EINVAL; + } + + drm_mm_init(&dev_priv->mm.gtt_space, args->gtt_start, + args->gtt_end - args->gtt_start); + + dev->gtt_total = (uint32_t) (args->gtt_end - args->gtt_start); + + mutex_unlock(&dev->struct_mutex); + + return 0; +} + + +/** + * Creates a new mm object and returns a handle to it. + */ +int +i915_gem_create_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_create *args = data; + struct drm_gem_object *obj; + int handle, ret; + + args->size = roundup(args->size, PAGE_SIZE); + + /* Allocate the new object */ + obj = drm_gem_object_alloc(dev, args->size); + if (obj == NULL) + return -ENOMEM; + + ret = drm_gem_handle_create(file_priv, obj, &handle); + mutex_lock(&dev->struct_mutex); + drm_gem_object_handle_unreference(obj); + mutex_unlock(&dev->struct_mutex); + + if (ret) + return ret; + + args->handle = handle; + + return 0; +} + +/** + * Reads data from the object referenced by handle. + * + * On error, the contents of *data are undefined. + */ +int +i915_gem_pread_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_pread *args = data; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + ssize_t read; + loff_t offset; + int ret; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EBADF; + obj_priv = obj->driver_private; + + /* Bounds check source. + * + * XXX: This could use review for overflow issues... + */ + if (args->offset > obj->size || args->size > obj->size || + args->offset + args->size > obj->size) { + drm_gem_object_unreference(obj); + return -EINVAL; + } + + mutex_lock(&dev->struct_mutex); + + ret = i915_gem_object_set_domain_range(obj, args->offset, args->size, + I915_GEM_DOMAIN_CPU, 0); + if (ret != 0) { + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + } + + offset = args->offset; + + read = vfs_read(obj->filp, (char __user *)(uintptr_t)args->data_ptr, + args->size, &offset); + if (read != args->size) { + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + if (read < 0) + return read; + else + return -EINVAL; + } + + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + + return 0; +} + +static int +i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, + struct drm_i915_gem_pwrite *args, + struct drm_file *file_priv) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + ssize_t remain; + loff_t offset; + char __user *user_data; + char *vaddr; + int i, o, l; + int ret = 0; + unsigned long pfn; + unsigned long unwritten; + + user_data = (char __user *) (uintptr_t) args->data_ptr; + remain = args->size; + if (!access_ok(VERIFY_READ, user_data, remain)) + return -EFAULT; + + + mutex_lock(&dev->struct_mutex); + ret = i915_gem_object_pin(obj, 0); + if (ret) { + mutex_unlock(&dev->struct_mutex); + return ret; + } + ret = i915_gem_set_domain(obj, file_priv, + I915_GEM_DOMAIN_GTT, I915_GEM_DOMAIN_GTT); + if (ret) + goto fail; + + obj_priv = obj->driver_private; + offset = obj_priv->gtt_offset + args->offset; + obj_priv->dirty = 1; + + while (remain > 0) { + /* Operation in this page + * + * i = page number + * o = offset within page + * l = bytes to copy + */ + i = offset >> PAGE_SHIFT; + o = offset & (PAGE_SIZE-1); + l = remain; + if ((o + l) > PAGE_SIZE) + l = PAGE_SIZE - o; + + pfn = (dev->agp->base >> PAGE_SHIFT) + i; + +#ifdef CONFIG_HIGHMEM + /* kmap_atomic can't map IO pages on non-HIGHMEM kernels + */ + vaddr = kmap_atomic_pfn(pfn, KM_USER0); +#if WATCH_PWRITE + DRM_INFO("pwrite i %d o %d l %d pfn %ld vaddr %p\n", + i, o, l, pfn, vaddr); +#endif + unwritten = __copy_from_user_inatomic_nocache(vaddr + o, + user_data, l); + kunmap_atomic(vaddr, KM_USER0); + + if (unwritten) +#endif /* CONFIG_HIGHMEM */ + { + vaddr = ioremap(pfn << PAGE_SHIFT, PAGE_SIZE); +#if WATCH_PWRITE + DRM_INFO("pwrite slow i %d o %d l %d " + "pfn %ld vaddr %p\n", + i, o, l, pfn, vaddr); +#endif + if (vaddr == NULL) { + ret = -EFAULT; + goto fail; + } + unwritten = __copy_from_user(vaddr + o, user_data, l); +#if WATCH_PWRITE + DRM_INFO("unwritten %ld\n", unwritten); +#endif + iounmap(vaddr); + if (unwritten) { + ret = -EFAULT; + goto fail; + } + } + + remain -= l; + user_data += l; + offset += l; + } +#if WATCH_PWRITE && 1 + i915_gem_clflush_object(obj); + i915_gem_dump_object(obj, args->offset + args->size, __func__, ~0); + i915_gem_clflush_object(obj); +#endif + +fail: + i915_gem_object_unpin(obj); + mutex_unlock(&dev->struct_mutex); + + return ret; +} + +int +i915_gem_shmem_pwrite(struct drm_device *dev, struct drm_gem_object *obj, + struct drm_i915_gem_pwrite *args, + struct drm_file *file_priv) +{ + int ret; + loff_t offset; + ssize_t written; + + mutex_lock(&dev->struct_mutex); + + ret = i915_gem_set_domain(obj, file_priv, + I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU); + if (ret) { + mutex_unlock(&dev->struct_mutex); + return ret; + } + + offset = args->offset; + + written = vfs_write(obj->filp, + (char __user *)(uintptr_t) args->data_ptr, + args->size, &offset); + if (written != args->size) { + mutex_unlock(&dev->struct_mutex); + if (written < 0) + return written; + else + return -EINVAL; + } + + mutex_unlock(&dev->struct_mutex); + + return 0; +} + +/** + * Writes data to the object referenced by handle. + * + * On error, the contents of the buffer that were to be modified are undefined. + */ +int +i915_gem_pwrite_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_pwrite *args = data; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret = 0; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EBADF; + obj_priv = obj->driver_private; + + /* Bounds check destination. + * + * XXX: This could use review for overflow issues... + */ + if (args->offset > obj->size || args->size > obj->size || + args->offset + args->size > obj->size) { + drm_gem_object_unreference(obj); + return -EINVAL; + } + + /* We can only do the GTT pwrite on untiled buffers, as otherwise + * it would end up going through the fenced access, and we'll get + * different detiling behavior between reading and writing. + * pread/pwrite currently are reading and writing from the CPU + * perspective, requiring manual detiling by the client. + */ + if (obj_priv->tiling_mode == I915_TILING_NONE && + dev->gtt_total != 0) + ret = i915_gem_gtt_pwrite(dev, obj, args, file_priv); + else + ret = i915_gem_shmem_pwrite(dev, obj, args, file_priv); + +#if WATCH_PWRITE + if (ret) + DRM_INFO("pwrite failed %d\n", ret); +#endif + + drm_gem_object_unreference(obj); + + return ret; +} + +/** + * Called when user space prepares to use an object + */ +int +i915_gem_set_domain_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_set_domain *args = data; + struct drm_gem_object *obj; + int ret; + + if (!(dev->driver->driver_features & DRIVER_GEM)) + return -ENODEV; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EBADF; + + mutex_lock(&dev->struct_mutex); +#if WATCH_BUF + DRM_INFO("set_domain_ioctl %p(%d), %08x %08x\n", + obj, obj->size, args->read_domains, args->write_domain); +#endif + ret = i915_gem_set_domain(obj, file_priv, + args->read_domains, args->write_domain); + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + return ret; +} + +/** + * Called when user space has done writes to this buffer + */ +int +i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_sw_finish *args = data; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret = 0; + + if (!(dev->driver->driver_features & DRIVER_GEM)) + return -ENODEV; + + mutex_lock(&dev->struct_mutex); + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) { + mutex_unlock(&dev->struct_mutex); + return -EBADF; + } + +#if WATCH_BUF + DRM_INFO("%s: sw_finish %d (%p %d)\n", + __func__, args->handle, obj, obj->size); +#endif + obj_priv = obj->driver_private; + + /* Pinned buffers may be scanout, so flush the cache */ + if ((obj->write_domain & I915_GEM_DOMAIN_CPU) && obj_priv->pin_count) { + i915_gem_clflush_object(obj); + drm_agp_chipset_flush(dev); + } + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + return ret; +} + +/** + * Maps the contents of an object, returning the address it is mapped + * into. + * + * While the mapping holds a reference on the contents of the object, it doesn't + * imply a ref on the object itself. + */ +int +i915_gem_mmap_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_mmap *args = data; + struct drm_gem_object *obj; + loff_t offset; + unsigned long addr; + + if (!(dev->driver->driver_features & DRIVER_GEM)) + return -ENODEV; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EBADF; + + offset = args->offset; + + down_write(¤t->mm->mmap_sem); + addr = do_mmap(obj->filp, 0, args->size, + PROT_READ | PROT_WRITE, MAP_SHARED, + args->offset); + up_write(¤t->mm->mmap_sem); + mutex_lock(&dev->struct_mutex); + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + if (IS_ERR((void *)addr)) + return addr; + + args->addr_ptr = (uint64_t) addr; + + return 0; +} + +static void +i915_gem_object_free_page_list(struct drm_gem_object *obj) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int page_count = obj->size / PAGE_SIZE; + int i; + + if (obj_priv->page_list == NULL) + return; + + + for (i = 0; i < page_count; i++) + if (obj_priv->page_list[i] != NULL) { + if (obj_priv->dirty) + set_page_dirty(obj_priv->page_list[i]); + mark_page_accessed(obj_priv->page_list[i]); + page_cache_release(obj_priv->page_list[i]); + } + obj_priv->dirty = 0; + + drm_free(obj_priv->page_list, + page_count * sizeof(struct page *), + DRM_MEM_DRIVER); + obj_priv->page_list = NULL; +} + +static void +i915_gem_object_move_to_active(struct drm_gem_object *obj) +{ + struct drm_device *dev = obj->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + /* Add a reference if we're newly entering the active list. */ + if (!obj_priv->active) { + drm_gem_object_reference(obj); + obj_priv->active = 1; + } + /* Move from whatever list we were on to the tail of execution. */ + list_move_tail(&obj_priv->list, + &dev_priv->mm.active_list); +} + + +static void +i915_gem_object_move_to_inactive(struct drm_gem_object *obj) +{ + struct drm_device *dev = obj->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + i915_verify_inactive(dev, __FILE__, __LINE__); + if (obj_priv->pin_count != 0) + list_del_init(&obj_priv->list); + else + list_move_tail(&obj_priv->list, &dev_priv->mm.inactive_list); + + if (obj_priv->active) { + obj_priv->active = 0; + drm_gem_object_unreference(obj); + } + i915_verify_inactive(dev, __FILE__, __LINE__); +} + +/** + * Creates a new sequence number, emitting a write of it to the status page + * plus an interrupt, which will trigger i915_user_interrupt_handler. + * + * Must be called with struct_lock held. + * + * Returned sequence numbers are nonzero on success. + */ +static uint32_t +i915_add_request(struct drm_device *dev, uint32_t flush_domains) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_request *request; + uint32_t seqno; + int was_empty; + RING_LOCALS; + + request = drm_calloc(1, sizeof(*request), DRM_MEM_DRIVER); + if (request == NULL) + return 0; + + /* Grab the seqno we're going to make this request be, and bump the + * next (skipping 0 so it can be the reserved no-seqno value). + */ + seqno = dev_priv->mm.next_gem_seqno; + dev_priv->mm.next_gem_seqno++; + if (dev_priv->mm.next_gem_seqno == 0) + dev_priv->mm.next_gem_seqno++; + + BEGIN_LP_RING(4); + OUT_RING(MI_STORE_DWORD_INDEX); + OUT_RING(I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT); + OUT_RING(seqno); + + OUT_RING(MI_USER_INTERRUPT); + ADVANCE_LP_RING(); + + DRM_DEBUG("%d\n", seqno); + + request->seqno = seqno; + request->emitted_jiffies = jiffies; + request->flush_domains = flush_domains; + was_empty = list_empty(&dev_priv->mm.request_list); + list_add_tail(&request->list, &dev_priv->mm.request_list); + + if (was_empty) + schedule_delayed_work(&dev_priv->mm.retire_work, HZ); + return seqno; +} + +/** + * Command execution barrier + * + * Ensures that all commands in the ring are finished + * before signalling the CPU + */ +uint32_t +i915_retire_commands(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + uint32_t cmd = MI_FLUSH | MI_NO_WRITE_FLUSH; + uint32_t flush_domains = 0; + RING_LOCALS; + + /* The sampler always gets flushed on i965 (sigh) */ + if (IS_I965G(dev)) + flush_domains |= I915_GEM_DOMAIN_SAMPLER; + BEGIN_LP_RING(2); + OUT_RING(cmd); + OUT_RING(0); /* noop */ + ADVANCE_LP_RING(); + return flush_domains; +} + +/** + * Moves buffers associated only with the given active seqno from the active + * to inactive list, potentially freeing them. + */ +static void +i915_gem_retire_request(struct drm_device *dev, + struct drm_i915_gem_request *request) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + + /* Move any buffers on the active list that are no longer referenced + * by the ringbuffer to the flushing/inactive lists as appropriate. + */ + while (!list_empty(&dev_priv->mm.active_list)) { + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + + obj_priv = list_first_entry(&dev_priv->mm.active_list, + struct drm_i915_gem_object, + list); + obj = obj_priv->obj; + + /* If the seqno being retired doesn't match the oldest in the + * list, then the oldest in the list must still be newer than + * this seqno. + */ + if (obj_priv->last_rendering_seqno != request->seqno) + return; +#if WATCH_LRU + DRM_INFO("%s: retire %d moves to inactive list %p\n", + __func__, request->seqno, obj); +#endif + + if (obj->write_domain != 0) { + list_move_tail(&obj_priv->list, + &dev_priv->mm.flushing_list); + } else { + i915_gem_object_move_to_inactive(obj); + } + } + + if (request->flush_domains != 0) { + struct drm_i915_gem_object *obj_priv, *next; + + /* Clear the write domain and activity from any buffers + * that are just waiting for a flush matching the one retired. + */ + list_for_each_entry_safe(obj_priv, next, + &dev_priv->mm.flushing_list, list) { + struct drm_gem_object *obj = obj_priv->obj; + + if (obj->write_domain & request->flush_domains) { + obj->write_domain = 0; + i915_gem_object_move_to_inactive(obj); + } + } + + } +} + +/** + * Returns true if seq1 is later than seq2. + */ +static int +i915_seqno_passed(uint32_t seq1, uint32_t seq2) +{ + return (int32_t)(seq1 - seq2) >= 0; +} + +uint32_t +i915_get_gem_seqno(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + + return READ_HWSP(dev_priv, I915_GEM_HWS_INDEX); +} + +/** + * This function clears the request list as sequence numbers are passed. + */ +void +i915_gem_retire_requests(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + uint32_t seqno; + + seqno = i915_get_gem_seqno(dev); + + while (!list_empty(&dev_priv->mm.request_list)) { + struct drm_i915_gem_request *request; + uint32_t retiring_seqno; + + request = list_first_entry(&dev_priv->mm.request_list, + struct drm_i915_gem_request, + list); + retiring_seqno = request->seqno; + + if (i915_seqno_passed(seqno, retiring_seqno) || + dev_priv->mm.wedged) { + i915_gem_retire_request(dev, request); + + list_del(&request->list); + drm_free(request, sizeof(*request), DRM_MEM_DRIVER); + } else + break; + } +} + +void +i915_gem_retire_work_handler(struct work_struct *work) +{ + drm_i915_private_t *dev_priv; + struct drm_device *dev; + + dev_priv = container_of(work, drm_i915_private_t, + mm.retire_work.work); + dev = dev_priv->dev; + + mutex_lock(&dev->struct_mutex); + i915_gem_retire_requests(dev); + if (!list_empty(&dev_priv->mm.request_list)) + schedule_delayed_work(&dev_priv->mm.retire_work, HZ); + mutex_unlock(&dev->struct_mutex); +} + +/** + * Waits for a sequence number to be signaled, and cleans up the + * request and object lists appropriately for that event. + */ +int +i915_wait_request(struct drm_device *dev, uint32_t seqno) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + int ret = 0; + + BUG_ON(seqno == 0); + + if (!i915_seqno_passed(i915_get_gem_seqno(dev), seqno)) { + dev_priv->mm.waiting_gem_seqno = seqno; + i915_user_irq_get(dev); + ret = wait_event_interruptible(dev_priv->irq_queue, + i915_seqno_passed(i915_get_gem_seqno(dev), + seqno) || + dev_priv->mm.wedged); + i915_user_irq_put(dev); + dev_priv->mm.waiting_gem_seqno = 0; + } + if (dev_priv->mm.wedged) + ret = -EIO; + + if (ret && ret != -ERESTARTSYS) + DRM_ERROR("%s returns %d (awaiting %d at %d)\n", + __func__, ret, seqno, i915_get_gem_seqno(dev)); + + /* Directly dispatch request retiring. While we have the work queue + * to handle this, the waiter on a request often wants an associated + * buffer to have made it to the inactive list, and we would need + * a separate wait queue to handle that. + */ + if (ret == 0) + i915_gem_retire_requests(dev); + + return ret; +} + +static void +i915_gem_flush(struct drm_device *dev, + uint32_t invalidate_domains, + uint32_t flush_domains) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + uint32_t cmd; + RING_LOCALS; + +#if WATCH_EXEC + DRM_INFO("%s: invalidate %08x flush %08x\n", __func__, + invalidate_domains, flush_domains); +#endif + + if (flush_domains & I915_GEM_DOMAIN_CPU) + drm_agp_chipset_flush(dev); + + if ((invalidate_domains | flush_domains) & ~(I915_GEM_DOMAIN_CPU | + I915_GEM_DOMAIN_GTT)) { + /* + * read/write caches: + * + * I915_GEM_DOMAIN_RENDER is always invalidated, but is + * only flushed if MI_NO_WRITE_FLUSH is unset. On 965, it is + * also flushed at 2d versus 3d pipeline switches. + * + * read-only caches: + * + * I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if + * MI_READ_FLUSH is set, and is always flushed on 965. + * + * I915_GEM_DOMAIN_COMMAND may not exist? + * + * I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is + * invalidated when MI_EXE_FLUSH is set. + * + * I915_GEM_DOMAIN_VERTEX, which exists on 965, is + * invalidated with every MI_FLUSH. + * + * TLBs: + * + * On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND + * and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and + * I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER + * are flushed at any MI_FLUSH. + */ + + cmd = MI_FLUSH | MI_NO_WRITE_FLUSH; + if ((invalidate_domains|flush_domains) & + I915_GEM_DOMAIN_RENDER) + cmd &= ~MI_NO_WRITE_FLUSH; + if (!IS_I965G(dev)) { + /* + * On the 965, the sampler cache always gets flushed + * and this bit is reserved. + */ + if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER) + cmd |= MI_READ_FLUSH; + } + if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION) + cmd |= MI_EXE_FLUSH; + +#if WATCH_EXEC + DRM_INFO("%s: queue flush %08x to ring\n", __func__, cmd); +#endif + BEGIN_LP_RING(2); + OUT_RING(cmd); + OUT_RING(0); /* noop */ + ADVANCE_LP_RING(); + } +} + +/** + * Ensures that all rendering to the object has completed and the object is + * safe to unbind from the GTT or access from the CPU. + */ +static int +i915_gem_object_wait_rendering(struct drm_gem_object *obj) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int ret; + + /* If there are writes queued to the buffer, flush and + * create a new seqno to wait for. + */ + if (obj->write_domain & ~(I915_GEM_DOMAIN_CPU|I915_GEM_DOMAIN_GTT)) { + uint32_t write_domain = obj->write_domain; +#if WATCH_BUF + DRM_INFO("%s: flushing object %p from write domain %08x\n", + __func__, obj, write_domain); +#endif + i915_gem_flush(dev, 0, write_domain); + + i915_gem_object_move_to_active(obj); + obj_priv->last_rendering_seqno = i915_add_request(dev, + write_domain); + BUG_ON(obj_priv->last_rendering_seqno == 0); +#if WATCH_LRU + DRM_INFO("%s: flush moves to exec list %p\n", __func__, obj); +#endif + } + + /* If there is rendering queued on the buffer being evicted, wait for + * it. + */ + if (obj_priv->active) { +#if WATCH_BUF + DRM_INFO("%s: object %p wait for seqno %08x\n", + __func__, obj, obj_priv->last_rendering_seqno); +#endif + ret = i915_wait_request(dev, obj_priv->last_rendering_seqno); + if (ret != 0) + return ret; + } + + return 0; +} + +/** + * Unbinds an object from the GTT aperture. + */ +static int +i915_gem_object_unbind(struct drm_gem_object *obj) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int ret = 0; + +#if WATCH_BUF + DRM_INFO("%s:%d %p\n", __func__, __LINE__, obj); + DRM_INFO("gtt_space %p\n", obj_priv->gtt_space); +#endif + if (obj_priv->gtt_space == NULL) + return 0; + + if (obj_priv->pin_count != 0) { + DRM_ERROR("Attempting to unbind pinned buffer\n"); + return -EINVAL; + } + + /* Wait for any rendering to complete + */ + ret = i915_gem_object_wait_rendering(obj); + if (ret) { + DRM_ERROR("wait_rendering failed: %d\n", ret); + return ret; + } + + /* Move the object to the CPU domain to ensure that + * any possible CPU writes while it's not in the GTT + * are flushed when we go to remap it. This will + * also ensure that all pending GPU writes are finished + * before we unbind. + */ + ret = i915_gem_object_set_domain(obj, I915_GEM_DOMAIN_CPU, + I915_GEM_DOMAIN_CPU); + if (ret) { + DRM_ERROR("set_domain failed: %d\n", ret); + return ret; + } + + if (obj_priv->agp_mem != NULL) { + drm_unbind_agp(obj_priv->agp_mem); + drm_free_agp(obj_priv->agp_mem, obj->size / PAGE_SIZE); + obj_priv->agp_mem = NULL; + } + + BUG_ON(obj_priv->active); + + i915_gem_object_free_page_list(obj); + + if (obj_priv->gtt_space) { + atomic_dec(&dev->gtt_count); + atomic_sub(obj->size, &dev->gtt_memory); + + drm_mm_put_block(obj_priv->gtt_space); + obj_priv->gtt_space = NULL; + } + + /* Remove ourselves from the LRU list if present. */ + if (!list_empty(&obj_priv->list)) + list_del_init(&obj_priv->list); + + return 0; +} + +static int +i915_gem_evict_something(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret = 0; + + for (;;) { + /* If there's an inactive buffer available now, grab it + * and be done. + */ + if (!list_empty(&dev_priv->mm.inactive_list)) { + obj_priv = list_first_entry(&dev_priv->mm.inactive_list, + struct drm_i915_gem_object, + list); + obj = obj_priv->obj; + BUG_ON(obj_priv->pin_count != 0); +#if WATCH_LRU + DRM_INFO("%s: evicting %p\n", __func__, obj); +#endif + BUG_ON(obj_priv->active); + + /* Wait on the rendering and unbind the buffer. */ + ret = i915_gem_object_unbind(obj); + break; + } + + /* If we didn't get anything, but the ring is still processing + * things, wait for one of those things to finish and hopefully + * leave us a buffer to evict. + */ + if (!list_empty(&dev_priv->mm.request_list)) { + struct drm_i915_gem_request *request; + + request = list_first_entry(&dev_priv->mm.request_list, + struct drm_i915_gem_request, + list); + + ret = i915_wait_request(dev, request->seqno); + if (ret) + break; + + /* if waiting caused an object to become inactive, + * then loop around and wait for it. Otherwise, we + * assume that waiting freed and unbound something, + * so there should now be some space in the GTT + */ + if (!list_empty(&dev_priv->mm.inactive_list)) + continue; + break; + } + + /* If we didn't have anything on the request list but there + * are buffers awaiting a flush, emit one and try again. + * When we wait on it, those buffers waiting for that flush + * will get moved to inactive. + */ + if (!list_empty(&dev_priv->mm.flushing_list)) { + obj_priv = list_first_entry(&dev_priv->mm.flushing_list, + struct drm_i915_gem_object, + list); + obj = obj_priv->obj; + + i915_gem_flush(dev, + obj->write_domain, + obj->write_domain); + i915_add_request(dev, obj->write_domain); + + obj = NULL; + continue; + } + + DRM_ERROR("inactive empty %d request empty %d " + "flushing empty %d\n", + list_empty(&dev_priv->mm.inactive_list), + list_empty(&dev_priv->mm.request_list), + list_empty(&dev_priv->mm.flushing_list)); + /* If we didn't do any of the above, there's nothing to be done + * and we just can't fit it in. + */ + return -ENOMEM; + } + return ret; +} + +static int +i915_gem_object_get_page_list(struct drm_gem_object *obj) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int page_count, i; + struct address_space *mapping; + struct inode *inode; + struct page *page; + int ret; + + if (obj_priv->page_list) + return 0; + + /* Get the list of pages out of our struct file. They'll be pinned + * at this point until we release them. + */ + page_count = obj->size / PAGE_SIZE; + BUG_ON(obj_priv->page_list != NULL); + obj_priv->page_list = drm_calloc(page_count, sizeof(struct page *), + DRM_MEM_DRIVER); + if (obj_priv->page_list == NULL) { + DRM_ERROR("Faled to allocate page list\n"); + return -ENOMEM; + } + + inode = obj->filp->f_path.dentry->d_inode; + mapping = inode->i_mapping; + for (i = 0; i < page_count; i++) { + page = read_mapping_page(mapping, i, NULL); + if (IS_ERR(page)) { + ret = PTR_ERR(page); + DRM_ERROR("read_mapping_page failed: %d\n", ret); + i915_gem_object_free_page_list(obj); + return ret; + } + obj_priv->page_list[i] = page; + } + return 0; +} + +/** + * Finds free space in the GTT aperture and binds the object there. + */ +static int +i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment) +{ + struct drm_device *dev = obj->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + struct drm_mm_node *free_space; + int page_count, ret; + + if (alignment == 0) + alignment = PAGE_SIZE; + if (alignment & (PAGE_SIZE - 1)) { + DRM_ERROR("Invalid object alignment requested %u\n", alignment); + return -EINVAL; + } + + search_free: + free_space = drm_mm_search_free(&dev_priv->mm.gtt_space, + obj->size, alignment, 0); + if (free_space != NULL) { + obj_priv->gtt_space = drm_mm_get_block(free_space, obj->size, + alignment); + if (obj_priv->gtt_space != NULL) { + obj_priv->gtt_space->private = obj; + obj_priv->gtt_offset = obj_priv->gtt_space->start; + } + } + if (obj_priv->gtt_space == NULL) { + /* If the gtt is empty and we're still having trouble + * fitting our object in, we're out of memory. + */ +#if WATCH_LRU + DRM_INFO("%s: GTT full, evicting something\n", __func__); +#endif + if (list_empty(&dev_priv->mm.inactive_list) && + list_empty(&dev_priv->mm.flushing_list) && + list_empty(&dev_priv->mm.active_list)) { + DRM_ERROR("GTT full, but LRU list empty\n"); + return -ENOMEM; + } + + ret = i915_gem_evict_something(dev); + if (ret != 0) { + DRM_ERROR("Failed to evict a buffer %d\n", ret); + return ret; + } + goto search_free; + } + +#if WATCH_BUF + DRM_INFO("Binding object of size %d at 0x%08x\n", + obj->size, obj_priv->gtt_offset); +#endif + ret = i915_gem_object_get_page_list(obj); + if (ret) { + drm_mm_put_block(obj_priv->gtt_space); + obj_priv->gtt_space = NULL; + return ret; + } + + page_count = obj->size / PAGE_SIZE; + /* Create an AGP memory structure pointing at our pages, and bind it + * into the GTT. + */ + obj_priv->agp_mem = drm_agp_bind_pages(dev, + obj_priv->page_list, + page_count, + obj_priv->gtt_offset); + if (obj_priv->agp_mem == NULL) { + i915_gem_object_free_page_list(obj); + drm_mm_put_block(obj_priv->gtt_space); + obj_priv->gtt_space = NULL; + return -ENOMEM; + } + atomic_inc(&dev->gtt_count); + atomic_add(obj->size, &dev->gtt_memory); + + /* Assert that the object is not currently in any GPU domain. As it + * wasn't in the GTT, there shouldn't be any way it could have been in + * a GPU cache + */ + BUG_ON(obj->read_domains & ~(I915_GEM_DOMAIN_CPU|I915_GEM_DOMAIN_GTT)); + BUG_ON(obj->write_domain & ~(I915_GEM_DOMAIN_CPU|I915_GEM_DOMAIN_GTT)); + + return 0; +} + +void +i915_gem_clflush_object(struct drm_gem_object *obj) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + /* If we don't have a page list set up, then we're not pinned + * to GPU, and we can ignore the cache flush because it'll happen + * again at bind time. + */ + if (obj_priv->page_list == NULL) + return; + + drm_clflush_pages(obj_priv->page_list, obj->size / PAGE_SIZE); +} + +/* + * Set the next domain for the specified object. This + * may not actually perform the necessary flushing/invaliding though, + * as that may want to be batched with other set_domain operations + * + * This is (we hope) the only really tricky part of gem. The goal + * is fairly simple -- track which caches hold bits of the object + * and make sure they remain coherent. A few concrete examples may + * help to explain how it works. For shorthand, we use the notation + * (read_domains, write_domain), e.g. (CPU, CPU) to indicate the + * a pair of read and write domain masks. + * + * Case 1: the batch buffer + * + * 1. Allocated + * 2. Written by CPU + * 3. Mapped to GTT + * 4. Read by GPU + * 5. Unmapped from GTT + * 6. Freed + * + * Let's take these a step at a time + * + * 1. Allocated + * Pages allocated from the kernel may still have + * cache contents, so we set them to (CPU, CPU) always. + * 2. Written by CPU (using pwrite) + * The pwrite function calls set_domain (CPU, CPU) and + * this function does nothing (as nothing changes) + * 3. Mapped by GTT + * This function asserts that the object is not + * currently in any GPU-based read or write domains + * 4. Read by GPU + * i915_gem_execbuffer calls set_domain (COMMAND, 0). + * As write_domain is zero, this function adds in the + * current read domains (CPU+COMMAND, 0). + * flush_domains is set to CPU. + * invalidate_domains is set to COMMAND + * clflush is run to get data out of the CPU caches + * then i915_dev_set_domain calls i915_gem_flush to + * emit an MI_FLUSH and drm_agp_chipset_flush + * 5. Unmapped from GTT + * i915_gem_object_unbind calls set_domain (CPU, CPU) + * flush_domains and invalidate_domains end up both zero + * so no flushing/invalidating happens + * 6. Freed + * yay, done + * + * Case 2: The shared render buffer + * + * 1. Allocated + * 2. Mapped to GTT + * 3. Read/written by GPU + * 4. set_domain to (CPU,CPU) + * 5. Read/written by CPU + * 6. Read/written by GPU + * + * 1. Allocated + * Same as last example, (CPU, CPU) + * 2. Mapped to GTT + * Nothing changes (assertions find that it is not in the GPU) + * 3. Read/written by GPU + * execbuffer calls set_domain (RENDER, RENDER) + * flush_domains gets CPU + * invalidate_domains gets GPU + * clflush (obj) + * MI_FLUSH and drm_agp_chipset_flush + * 4. set_domain (CPU, CPU) + * flush_domains gets GPU + * invalidate_domains gets CPU + * wait_rendering (obj) to make sure all drawing is complete. + * This will include an MI_FLUSH to get the data from GPU + * to memory + * clflush (obj) to invalidate the CPU cache + * Another MI_FLUSH in i915_gem_flush (eliminate this somehow?) + * 5. Read/written by CPU + * cache lines are loaded and dirtied + * 6. Read written by GPU + * Same as last GPU access + * + * Case 3: The constant buffer + * + * 1. Allocated + * 2. Written by CPU + * 3. Read by GPU + * 4. Updated (written) by CPU again + * 5. Read by GPU + * + * 1. Allocated + * (CPU, CPU) + * 2. Written by CPU + * (CPU, CPU) + * 3. Read by GPU + * (CPU+RENDER, 0) + * flush_domains = CPU + * invalidate_domains = RENDER + * clflush (obj) + * MI_FLUSH + * drm_agp_chipset_flush + * 4. Updated (written) by CPU again + * (CPU, CPU) + * flush_domains = 0 (no previous write domain) + * invalidate_domains = 0 (no new read domains) + * 5. Read by GPU + * (CPU+RENDER, 0) + * flush_domains = CPU + * invalidate_domains = RENDER + * clflush (obj) + * MI_FLUSH + * drm_agp_chipset_flush + */ +static int +i915_gem_object_set_domain(struct drm_gem_object *obj, + uint32_t read_domains, + uint32_t write_domain) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + uint32_t invalidate_domains = 0; + uint32_t flush_domains = 0; + int ret; + +#if WATCH_BUF + DRM_INFO("%s: object %p read %08x -> %08x write %08x -> %08x\n", + __func__, obj, + obj->read_domains, read_domains, + obj->write_domain, write_domain); +#endif + /* + * If the object isn't moving to a new write domain, + * let the object stay in multiple read domains + */ + if (write_domain == 0) + read_domains |= obj->read_domains; + else + obj_priv->dirty = 1; + + /* + * Flush the current write domain if + * the new read domains don't match. Invalidate + * any read domains which differ from the old + * write domain + */ + if (obj->write_domain && obj->write_domain != read_domains) { + flush_domains |= obj->write_domain; + invalidate_domains |= read_domains & ~obj->write_domain; + } + /* + * Invalidate any read caches which may have + * stale data. That is, any new read domains. + */ + invalidate_domains |= read_domains & ~obj->read_domains; + if ((flush_domains | invalidate_domains) & I915_GEM_DOMAIN_CPU) { +#if WATCH_BUF + DRM_INFO("%s: CPU domain flush %08x invalidate %08x\n", + __func__, flush_domains, invalidate_domains); +#endif + /* + * If we're invaliding the CPU cache and flushing a GPU cache, + * then pause for rendering so that the GPU caches will be + * flushed before the cpu cache is invalidated + */ + if ((invalidate_domains & I915_GEM_DOMAIN_CPU) && + (flush_domains & ~(I915_GEM_DOMAIN_CPU | + I915_GEM_DOMAIN_GTT))) { + ret = i915_gem_object_wait_rendering(obj); + if (ret) + return ret; + } + i915_gem_clflush_object(obj); + } + + if ((write_domain | flush_domains) != 0) + obj->write_domain = write_domain; + + /* If we're invalidating the CPU domain, clear the per-page CPU + * domain list as well. + */ + if (obj_priv->page_cpu_valid != NULL && + (write_domain != 0 || + read_domains & I915_GEM_DOMAIN_CPU)) { + drm_free(obj_priv->page_cpu_valid, obj->size / PAGE_SIZE, + DRM_MEM_DRIVER); + obj_priv->page_cpu_valid = NULL; + } + obj->read_domains = read_domains; + + dev->invalidate_domains |= invalidate_domains; + dev->flush_domains |= flush_domains; +#if WATCH_BUF + DRM_INFO("%s: read %08x write %08x invalidate %08x flush %08x\n", + __func__, + obj->read_domains, obj->write_domain, + dev->invalidate_domains, dev->flush_domains); +#endif + return 0; +} + +/** + * Set the read/write domain on a range of the object. + * + * Currently only implemented for CPU reads, otherwise drops to normal + * i915_gem_object_set_domain(). + */ +static int +i915_gem_object_set_domain_range(struct drm_gem_object *obj, + uint64_t offset, + uint64_t size, + uint32_t read_domains, + uint32_t write_domain) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int ret, i; + + if (obj->read_domains & I915_GEM_DOMAIN_CPU) + return 0; + + if (read_domains != I915_GEM_DOMAIN_CPU || + write_domain != 0) + return i915_gem_object_set_domain(obj, + read_domains, write_domain); + + /* Wait on any GPU rendering to the object to be flushed. */ + if (obj->write_domain & ~(I915_GEM_DOMAIN_CPU | I915_GEM_DOMAIN_GTT)) { + ret = i915_gem_object_wait_rendering(obj); + if (ret) + return ret; + } + + if (obj_priv->page_cpu_valid == NULL) { + obj_priv->page_cpu_valid = drm_calloc(1, obj->size / PAGE_SIZE, + DRM_MEM_DRIVER); + } + + /* Flush the cache on any pages that are still invalid from the CPU's + * perspective. + */ + for (i = offset / PAGE_SIZE; i <= (offset + size - 1) / PAGE_SIZE; i++) { + if (obj_priv->page_cpu_valid[i]) + continue; + + drm_clflush_pages(obj_priv->page_list + i, 1); + + obj_priv->page_cpu_valid[i] = 1; + } + + return 0; +} + +/** + * Once all of the objects have been set in the proper domain, + * perform the necessary flush and invalidate operations. + * + * Returns the write domains flushed, for use in flush tracking. + */ +static uint32_t +i915_gem_dev_set_domain(struct drm_device *dev) +{ + uint32_t flush_domains = dev->flush_domains; + + /* + * Now that all the buffers are synced to the proper domains, + * flush and invalidate the collected domains + */ + if (dev->invalidate_domains | dev->flush_domains) { +#if WATCH_EXEC + DRM_INFO("%s: invalidate_domains %08x flush_domains %08x\n", + __func__, + dev->invalidate_domains, + dev->flush_domains); +#endif + i915_gem_flush(dev, + dev->invalidate_domains, + dev->flush_domains); + dev->invalidate_domains = 0; + dev->flush_domains = 0; + } + + return flush_domains; +} + +/** + * Pin an object to the GTT and evaluate the relocations landing in it. + */ +static int +i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, + struct drm_file *file_priv, + struct drm_i915_gem_exec_object *entry) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_relocation_entry reloc; + struct drm_i915_gem_relocation_entry __user *relocs; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int i, ret; + uint32_t last_reloc_offset = -1; + void *reloc_page = NULL; + + /* Choose the GTT offset for our buffer and put it there. */ + ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment); + if (ret) + return ret; + + entry->offset = obj_priv->gtt_offset; + + relocs = (struct drm_i915_gem_relocation_entry __user *) + (uintptr_t) entry->relocs_ptr; + /* Apply the relocations, using the GTT aperture to avoid cache + * flushing requirements. + */ + for (i = 0; i < entry->relocation_count; i++) { + struct drm_gem_object *target_obj; + struct drm_i915_gem_object *target_obj_priv; + uint32_t reloc_val, reloc_offset, *reloc_entry; + int ret; + + ret = copy_from_user(&reloc, relocs + i, sizeof(reloc)); + if (ret != 0) { + i915_gem_object_unpin(obj); + return ret; + } + + target_obj = drm_gem_object_lookup(obj->dev, file_priv, + reloc.target_handle); + if (target_obj == NULL) { + i915_gem_object_unpin(obj); + return -EBADF; + } + target_obj_priv = target_obj->driver_private; + + /* The target buffer should have appeared before us in the + * exec_object list, so it should have a GTT space bound by now. + */ + if (target_obj_priv->gtt_space == NULL) { + DRM_ERROR("No GTT space found for object %d\n", + reloc.target_handle); + drm_gem_object_unreference(target_obj); + i915_gem_object_unpin(obj); + return -EINVAL; + } + + if (reloc.offset > obj->size - 4) { + DRM_ERROR("Relocation beyond object bounds: " + "obj %p target %d offset %d size %d.\n", + obj, reloc.target_handle, + (int) reloc.offset, (int) obj->size); + drm_gem_object_unreference(target_obj); + i915_gem_object_unpin(obj); + return -EINVAL; + } + if (reloc.offset & 3) { + DRM_ERROR("Relocation not 4-byte aligned: " + "obj %p target %d offset %d.\n", + obj, reloc.target_handle, + (int) reloc.offset); + drm_gem_object_unreference(target_obj); + i915_gem_object_unpin(obj); + return -EINVAL; + } + + if (reloc.write_domain && target_obj->pending_write_domain && + reloc.write_domain != target_obj->pending_write_domain) { + DRM_ERROR("Write domain conflict: " + "obj %p target %d offset %d " + "new %08x old %08x\n", + obj, reloc.target_handle, + (int) reloc.offset, + reloc.write_domain, + target_obj->pending_write_domain); + drm_gem_object_unreference(target_obj); + i915_gem_object_unpin(obj); + return -EINVAL; + } + +#if WATCH_RELOC + DRM_INFO("%s: obj %p offset %08x target %d " + "read %08x write %08x gtt %08x " + "presumed %08x delta %08x\n", + __func__, + obj, + (int) reloc.offset, + (int) reloc.target_handle, + (int) reloc.read_domains, + (int) reloc.write_domain, + (int) target_obj_priv->gtt_offset, + (int) reloc.presumed_offset, + reloc.delta); +#endif + + target_obj->pending_read_domains |= reloc.read_domains; + target_obj->pending_write_domain |= reloc.write_domain; + + /* If the relocation already has the right value in it, no + * more work needs to be done. + */ + if (target_obj_priv->gtt_offset == reloc.presumed_offset) { + drm_gem_object_unreference(target_obj); + continue; + } + + /* Now that we're going to actually write some data in, + * make sure that any rendering using this buffer's contents + * is completed. + */ + i915_gem_object_wait_rendering(obj); + + /* As we're writing through the gtt, flush + * any CPU writes before we write the relocations + */ + if (obj->write_domain & I915_GEM_DOMAIN_CPU) { + i915_gem_clflush_object(obj); + drm_agp_chipset_flush(dev); + obj->write_domain = 0; + } + + /* Map the page containing the relocation we're going to + * perform. + */ + reloc_offset = obj_priv->gtt_offset + reloc.offset; + if (reloc_page == NULL || + (last_reloc_offset & ~(PAGE_SIZE - 1)) != + (reloc_offset & ~(PAGE_SIZE - 1))) { + if (reloc_page != NULL) + iounmap(reloc_page); + + reloc_page = ioremap(dev->agp->base + + (reloc_offset & ~(PAGE_SIZE - 1)), + PAGE_SIZE); + last_reloc_offset = reloc_offset; + if (reloc_page == NULL) { + drm_gem_object_unreference(target_obj); + i915_gem_object_unpin(obj); + return -ENOMEM; + } + } + + reloc_entry = (uint32_t *)((char *)reloc_page + + (reloc_offset & (PAGE_SIZE - 1))); + reloc_val = target_obj_priv->gtt_offset + reloc.delta; + +#if WATCH_BUF + DRM_INFO("Applied relocation: %p@0x%08x %08x -> %08x\n", + obj, (unsigned int) reloc.offset, + readl(reloc_entry), reloc_val); +#endif + writel(reloc_val, reloc_entry); + + /* Write the updated presumed offset for this entry back out + * to the user. + */ + reloc.presumed_offset = target_obj_priv->gtt_offset; + ret = copy_to_user(relocs + i, &reloc, sizeof(reloc)); + if (ret != 0) { + drm_gem_object_unreference(target_obj); + i915_gem_object_unpin(obj); + return ret; + } + + drm_gem_object_unreference(target_obj); + } + + if (reloc_page != NULL) + iounmap(reloc_page); + +#if WATCH_BUF + if (0) + i915_gem_dump_object(obj, 128, __func__, ~0); +#endif + return 0; +} + +/** Dispatch a batchbuffer to the ring + */ +static int +i915_dispatch_gem_execbuffer(struct drm_device *dev, + struct drm_i915_gem_execbuffer *exec, + uint64_t exec_offset) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_clip_rect __user *boxes = (struct drm_clip_rect __user *) + (uintptr_t) exec->cliprects_ptr; + int nbox = exec->num_cliprects; + int i = 0, count; + uint32_t exec_start, exec_len; + RING_LOCALS; + + exec_start = (uint32_t) exec_offset + exec->batch_start_offset; + exec_len = (uint32_t) exec->batch_len; + + if ((exec_start | exec_len) & 0x7) { + DRM_ERROR("alignment\n"); + return -EINVAL; + } + + if (!exec_start) + return -EINVAL; + + count = nbox ? nbox : 1; + + for (i = 0; i < count; i++) { + if (i < nbox) { + int ret = i915_emit_box(dev, boxes, i, + exec->DR1, exec->DR4); + if (ret) + return ret; + } + + if (IS_I830(dev) || IS_845G(dev)) { + BEGIN_LP_RING(4); + OUT_RING(MI_BATCH_BUFFER); + OUT_RING(exec_start | MI_BATCH_NON_SECURE); + OUT_RING(exec_start + exec_len - 4); + OUT_RING(0); + ADVANCE_LP_RING(); + } else { + BEGIN_LP_RING(2); + if (IS_I965G(dev)) { + OUT_RING(MI_BATCH_BUFFER_START | + (2 << 6) | + MI_BATCH_NON_SECURE_I965); + OUT_RING(exec_start); + } else { + OUT_RING(MI_BATCH_BUFFER_START | + (2 << 6)); + OUT_RING(exec_start | MI_BATCH_NON_SECURE); + } + ADVANCE_LP_RING(); + } + } + + /* XXX breadcrumb */ + return 0; +} + +/* Throttle our rendering by waiting until the ring has completed our requests + * emitted over 20 msec ago. + * + * This should get us reasonable parallelism between CPU and GPU but also + * relatively low latency when blocking on a particular request to finish. + */ +static int +i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file_priv) +{ + struct drm_i915_file_private *i915_file_priv = file_priv->driver_priv; + int ret = 0; + uint32_t seqno; + + mutex_lock(&dev->struct_mutex); + seqno = i915_file_priv->mm.last_gem_throttle_seqno; + i915_file_priv->mm.last_gem_throttle_seqno = + i915_file_priv->mm.last_gem_seqno; + if (seqno) + ret = i915_wait_request(dev, seqno); + mutex_unlock(&dev->struct_mutex); + return ret; +} + +int +i915_gem_execbuffer(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_file_private *i915_file_priv = file_priv->driver_priv; + struct drm_i915_gem_execbuffer *args = data; + struct drm_i915_gem_exec_object *exec_list = NULL; + struct drm_gem_object **object_list = NULL; + struct drm_gem_object *batch_obj; + int ret, i, pinned = 0; + uint64_t exec_offset; + uint32_t seqno, flush_domains; + +#if WATCH_EXEC + DRM_INFO("buffers_ptr %d buffer_count %d len %08x\n", + (int) args->buffers_ptr, args->buffer_count, args->batch_len); +#endif + + /* Copy in the exec list from userland */ + exec_list = drm_calloc(sizeof(*exec_list), args->buffer_count, + DRM_MEM_DRIVER); + object_list = drm_calloc(sizeof(*object_list), args->buffer_count, + DRM_MEM_DRIVER); + if (exec_list == NULL || object_list == NULL) { + DRM_ERROR("Failed to allocate exec or object list " + "for %d buffers\n", + args->buffer_count); + ret = -ENOMEM; + goto pre_mutex_err; + } + ret = copy_from_user(exec_list, + (struct drm_i915_relocation_entry __user *) + (uintptr_t) args->buffers_ptr, + sizeof(*exec_list) * args->buffer_count); + if (ret != 0) { + DRM_ERROR("copy %d exec entries failed %d\n", + args->buffer_count, ret); + goto pre_mutex_err; + } + + mutex_lock(&dev->struct_mutex); + + i915_verify_inactive(dev, __FILE__, __LINE__); + + if (dev_priv->mm.wedged) { + DRM_ERROR("Execbuf while wedged\n"); + mutex_unlock(&dev->struct_mutex); + return -EIO; + } + + if (dev_priv->mm.suspended) { + DRM_ERROR("Execbuf while VT-switched.\n"); + mutex_unlock(&dev->struct_mutex); + return -EBUSY; + } + + /* Zero the gloabl flush/invalidate flags. These + * will be modified as each object is bound to the + * gtt + */ + dev->invalidate_domains = 0; + dev->flush_domains = 0; + + /* Look up object handles and perform the relocations */ + for (i = 0; i < args->buffer_count; i++) { + object_list[i] = drm_gem_object_lookup(dev, file_priv, + exec_list[i].handle); + if (object_list[i] == NULL) { + DRM_ERROR("Invalid object handle %d at index %d\n", + exec_list[i].handle, i); + ret = -EBADF; + goto err; + } + + object_list[i]->pending_read_domains = 0; + object_list[i]->pending_write_domain = 0; + ret = i915_gem_object_pin_and_relocate(object_list[i], + file_priv, + &exec_list[i]); + if (ret) { + DRM_ERROR("object bind and relocate failed %d\n", ret); + goto err; + } + pinned = i + 1; + } + + /* Set the pending read domains for the batch buffer to COMMAND */ + batch_obj = object_list[args->buffer_count-1]; + batch_obj->pending_read_domains = I915_GEM_DOMAIN_COMMAND; + batch_obj->pending_write_domain = 0; + + i915_verify_inactive(dev, __FILE__, __LINE__); + + for (i = 0; i < args->buffer_count; i++) { + struct drm_gem_object *obj = object_list[i]; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + if (obj_priv->gtt_space == NULL) { + /* We evicted the buffer in the process of validating + * our set of buffers in. We could try to recover by + * kicking them everything out and trying again from + * the start. + */ + ret = -ENOMEM; + goto err; + } + + /* make sure all previous memory operations have passed */ + ret = i915_gem_object_set_domain(obj, + obj->pending_read_domains, + obj->pending_write_domain); + if (ret) + goto err; + } + + i915_verify_inactive(dev, __FILE__, __LINE__); + + /* Flush/invalidate caches and chipset buffer */ + flush_domains = i915_gem_dev_set_domain(dev); + + i915_verify_inactive(dev, __FILE__, __LINE__); + +#if WATCH_COHERENCY + for (i = 0; i < args->buffer_count; i++) { + i915_gem_object_check_coherency(object_list[i], + exec_list[i].handle); + } +#endif + + exec_offset = exec_list[args->buffer_count - 1].offset; + +#if WATCH_EXEC + i915_gem_dump_object(object_list[args->buffer_count - 1], + args->batch_len, + __func__, + ~0); +#endif + + (void)i915_add_request(dev, flush_domains); + + /* Exec the batchbuffer */ + ret = i915_dispatch_gem_execbuffer(dev, args, exec_offset); + if (ret) { + DRM_ERROR("dispatch failed %d\n", ret); + goto err; + } + + /* + * Ensure that the commands in the batch buffer are + * finished before the interrupt fires + */ + flush_domains = i915_retire_commands(dev); + + i915_verify_inactive(dev, __FILE__, __LINE__); + + /* + * Get a seqno representing the execution of the current buffer, + * which we can wait on. We would like to mitigate these interrupts, + * likely by only creating seqnos occasionally (so that we have + * *some* interrupts representing completion of buffers that we can + * wait on when trying to clear up gtt space). + */ + seqno = i915_add_request(dev, flush_domains); + BUG_ON(seqno == 0); + i915_file_priv->mm.last_gem_seqno = seqno; + for (i = 0; i < args->buffer_count; i++) { + struct drm_gem_object *obj = object_list[i]; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + i915_gem_object_move_to_active(obj); + obj_priv->last_rendering_seqno = seqno; +#if WATCH_LRU + DRM_INFO("%s: move to exec list %p\n", __func__, obj); +#endif + } +#if WATCH_LRU + i915_dump_lru(dev, __func__); +#endif + + i915_verify_inactive(dev, __FILE__, __LINE__); + + /* Copy the new buffer offsets back to the user's exec list. */ + ret = copy_to_user((struct drm_i915_relocation_entry __user *) + (uintptr_t) args->buffers_ptr, + exec_list, + sizeof(*exec_list) * args->buffer_count); + if (ret) + DRM_ERROR("failed to copy %d exec entries " + "back to user (%d)\n", + args->buffer_count, ret); +err: + if (object_list != NULL) { + for (i = 0; i < pinned; i++) + i915_gem_object_unpin(object_list[i]); + + for (i = 0; i < args->buffer_count; i++) + drm_gem_object_unreference(object_list[i]); + } + mutex_unlock(&dev->struct_mutex); + +pre_mutex_err: + drm_free(object_list, sizeof(*object_list) * args->buffer_count, + DRM_MEM_DRIVER); + drm_free(exec_list, sizeof(*exec_list) * args->buffer_count, + DRM_MEM_DRIVER); + + return ret; +} + +int +i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int ret; + + i915_verify_inactive(dev, __FILE__, __LINE__); + if (obj_priv->gtt_space == NULL) { + ret = i915_gem_object_bind_to_gtt(obj, alignment); + if (ret != 0) { + DRM_ERROR("Failure to bind: %d", ret); + return ret; + } + } + obj_priv->pin_count++; + + /* If the object is not active and not pending a flush, + * remove it from the inactive list + */ + if (obj_priv->pin_count == 1) { + atomic_inc(&dev->pin_count); + atomic_add(obj->size, &dev->pin_memory); + if (!obj_priv->active && + (obj->write_domain & ~(I915_GEM_DOMAIN_CPU | + I915_GEM_DOMAIN_GTT)) == 0 && + !list_empty(&obj_priv->list)) + list_del_init(&obj_priv->list); + } + i915_verify_inactive(dev, __FILE__, __LINE__); + + return 0; +} + +void +i915_gem_object_unpin(struct drm_gem_object *obj) +{ + struct drm_device *dev = obj->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + i915_verify_inactive(dev, __FILE__, __LINE__); + obj_priv->pin_count--; + BUG_ON(obj_priv->pin_count < 0); + BUG_ON(obj_priv->gtt_space == NULL); + + /* If the object is no longer pinned, and is + * neither active nor being flushed, then stick it on + * the inactive list + */ + if (obj_priv->pin_count == 0) { + if (!obj_priv->active && + (obj->write_domain & ~(I915_GEM_DOMAIN_CPU | + I915_GEM_DOMAIN_GTT)) == 0) + list_move_tail(&obj_priv->list, + &dev_priv->mm.inactive_list); + atomic_dec(&dev->pin_count); + atomic_sub(obj->size, &dev->pin_memory); + } + i915_verify_inactive(dev, __FILE__, __LINE__); +} + +int +i915_gem_pin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_pin *args = data; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret; + + mutex_lock(&dev->struct_mutex); + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) { + DRM_ERROR("Bad handle in i915_gem_pin_ioctl(): %d\n", + args->handle); + mutex_unlock(&dev->struct_mutex); + return -EBADF; + } + obj_priv = obj->driver_private; + + ret = i915_gem_object_pin(obj, args->alignment); + if (ret != 0) { + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + return ret; + } + + /* XXX - flush the CPU caches for pinned objects + * as the X server doesn't manage domains yet + */ + if (obj->write_domain & I915_GEM_DOMAIN_CPU) { + i915_gem_clflush_object(obj); + drm_agp_chipset_flush(dev); + obj->write_domain = 0; + } + args->offset = obj_priv->gtt_offset; + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + + return 0; +} + +int +i915_gem_unpin_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_pin *args = data; + struct drm_gem_object *obj; + + mutex_lock(&dev->struct_mutex); + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) { + DRM_ERROR("Bad handle in i915_gem_unpin_ioctl(): %d\n", + args->handle); + mutex_unlock(&dev->struct_mutex); + return -EBADF; + } + + i915_gem_object_unpin(obj); + + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + return 0; +} + +int +i915_gem_busy_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_busy *args = data; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + + mutex_lock(&dev->struct_mutex); + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) { + DRM_ERROR("Bad handle in i915_gem_busy_ioctl(): %d\n", + args->handle); + mutex_unlock(&dev->struct_mutex); + return -EBADF; + } + + obj_priv = obj->driver_private; + args->busy = obj_priv->active; + + drm_gem_object_unreference(obj); + mutex_unlock(&dev->struct_mutex); + return 0; +} + +int +i915_gem_throttle_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + return i915_gem_ring_throttle(dev, file_priv); +} + +int i915_gem_init_object(struct drm_gem_object *obj) +{ + struct drm_i915_gem_object *obj_priv; + + obj_priv = drm_calloc(1, sizeof(*obj_priv), DRM_MEM_DRIVER); + if (obj_priv == NULL) + return -ENOMEM; + + /* + * We've just allocated pages from the kernel, + * so they've just been written by the CPU with + * zeros. They'll need to be clflushed before we + * use them with the GPU. + */ + obj->write_domain = I915_GEM_DOMAIN_CPU; + obj->read_domains = I915_GEM_DOMAIN_CPU; + + obj->driver_private = obj_priv; + obj_priv->obj = obj; + INIT_LIST_HEAD(&obj_priv->list); + return 0; +} + +void i915_gem_free_object(struct drm_gem_object *obj) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + while (obj_priv->pin_count > 0) + i915_gem_object_unpin(obj); + + i915_gem_object_unbind(obj); + + drm_free(obj_priv->page_cpu_valid, 1, DRM_MEM_DRIVER); + drm_free(obj->driver_private, 1, DRM_MEM_DRIVER); +} + +static int +i915_gem_set_domain(struct drm_gem_object *obj, + struct drm_file *file_priv, + uint32_t read_domains, + uint32_t write_domain) +{ + struct drm_device *dev = obj->dev; + int ret; + uint32_t flush_domains; + + BUG_ON(!mutex_is_locked(&dev->struct_mutex)); + + ret = i915_gem_object_set_domain(obj, read_domains, write_domain); + if (ret) + return ret; + flush_domains = i915_gem_dev_set_domain(obj->dev); + + if (flush_domains & ~(I915_GEM_DOMAIN_CPU|I915_GEM_DOMAIN_GTT)) + (void) i915_add_request(dev, flush_domains); + + return 0; +} + +/** Unbinds all objects that are on the given buffer list. */ +static int +i915_gem_evict_from_list(struct drm_device *dev, struct list_head *head) +{ + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret; + + while (!list_empty(head)) { + obj_priv = list_first_entry(head, + struct drm_i915_gem_object, + list); + obj = obj_priv->obj; + + if (obj_priv->pin_count != 0) { + DRM_ERROR("Pinned object in unbind list\n"); + mutex_unlock(&dev->struct_mutex); + return -EINVAL; + } + + ret = i915_gem_object_unbind(obj); + if (ret != 0) { + DRM_ERROR("Error unbinding object in LeaveVT: %d\n", + ret); + mutex_unlock(&dev->struct_mutex); + return ret; + } + } + + + return 0; +} + +static int +i915_gem_idle(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + uint32_t seqno, cur_seqno, last_seqno; + int stuck, ret; + + if (dev_priv->mm.suspended) + return 0; + + /* Hack! Don't let anybody do execbuf while we don't control the chip. + * We need to replace this with a semaphore, or something. + */ + dev_priv->mm.suspended = 1; + + i915_kernel_lost_context(dev); + + /* Flush the GPU along with all non-CPU write domains + */ + i915_gem_flush(dev, ~(I915_GEM_DOMAIN_CPU|I915_GEM_DOMAIN_GTT), + ~(I915_GEM_DOMAIN_CPU|I915_GEM_DOMAIN_GTT)); + seqno = i915_add_request(dev, ~(I915_GEM_DOMAIN_CPU | + I915_GEM_DOMAIN_GTT)); + + if (seqno == 0) { + mutex_unlock(&dev->struct_mutex); + return -ENOMEM; + } + + dev_priv->mm.waiting_gem_seqno = seqno; + last_seqno = 0; + stuck = 0; + for (;;) { + cur_seqno = i915_get_gem_seqno(dev); + if (i915_seqno_passed(cur_seqno, seqno)) + break; + if (last_seqno == cur_seqno) { + if (stuck++ > 100) { + DRM_ERROR("hardware wedged\n"); + dev_priv->mm.wedged = 1; + DRM_WAKEUP(&dev_priv->irq_queue); + break; + } + } + msleep(10); + last_seqno = cur_seqno; + } + dev_priv->mm.waiting_gem_seqno = 0; + + i915_gem_retire_requests(dev); + + /* Active and flushing should now be empty as we've + * waited for a sequence higher than any pending execbuffer + */ + BUG_ON(!list_empty(&dev_priv->mm.active_list)); + BUG_ON(!list_empty(&dev_priv->mm.flushing_list)); + + /* Request should now be empty as we've also waited + * for the last request in the list + */ + BUG_ON(!list_empty(&dev_priv->mm.request_list)); + + /* Move all buffers out of the GTT. */ + ret = i915_gem_evict_from_list(dev, &dev_priv->mm.inactive_list); + if (ret) + return ret; + + BUG_ON(!list_empty(&dev_priv->mm.active_list)); + BUG_ON(!list_empty(&dev_priv->mm.flushing_list)); + BUG_ON(!list_empty(&dev_priv->mm.inactive_list)); + BUG_ON(!list_empty(&dev_priv->mm.request_list)); + return 0; +} + +static int +i915_gem_init_hws(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret; + + /* If we need a physical address for the status page, it's already + * initialized at driver load time. + */ + if (!I915_NEED_GFX_HWS(dev)) + return 0; + + obj = drm_gem_object_alloc(dev, 4096); + if (obj == NULL) { + DRM_ERROR("Failed to allocate status page\n"); + return -ENOMEM; + } + obj_priv = obj->driver_private; + + ret = i915_gem_object_pin(obj, 4096); + if (ret != 0) { + drm_gem_object_unreference(obj); + return ret; + } + + dev_priv->status_gfx_addr = obj_priv->gtt_offset; + dev_priv->hws_map.offset = dev->agp->base + obj_priv->gtt_offset; + dev_priv->hws_map.size = 4096; + dev_priv->hws_map.type = 0; + dev_priv->hws_map.flags = 0; + dev_priv->hws_map.mtrr = 0; + + drm_core_ioremap(&dev_priv->hws_map, dev); + if (dev_priv->hws_map.handle == NULL) { + DRM_ERROR("Failed to map status page.\n"); + memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); + drm_gem_object_unreference(obj); + return -EINVAL; + } + dev_priv->hws_obj = obj; + dev_priv->hw_status_page = dev_priv->hws_map.handle; + memset(dev_priv->hw_status_page, 0, PAGE_SIZE); + I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); + DRM_DEBUG("hws offset: 0x%08x\n", dev_priv->status_gfx_addr); + + return 0; +} + +static int +i915_gem_init_ringbuffer(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + int ret; + + ret = i915_gem_init_hws(dev); + if (ret != 0) + return ret; + + obj = drm_gem_object_alloc(dev, 128 * 1024); + if (obj == NULL) { + DRM_ERROR("Failed to allocate ringbuffer\n"); + return -ENOMEM; + } + obj_priv = obj->driver_private; + + ret = i915_gem_object_pin(obj, 4096); + if (ret != 0) { + drm_gem_object_unreference(obj); + return ret; + } + + /* Set up the kernel mapping for the ring. */ + dev_priv->ring.Size = obj->size; + dev_priv->ring.tail_mask = obj->size - 1; + + dev_priv->ring.map.offset = dev->agp->base + obj_priv->gtt_offset; + dev_priv->ring.map.size = obj->size; + dev_priv->ring.map.type = 0; + dev_priv->ring.map.flags = 0; + dev_priv->ring.map.mtrr = 0; + + drm_core_ioremap(&dev_priv->ring.map, dev); + if (dev_priv->ring.map.handle == NULL) { + DRM_ERROR("Failed to map ringbuffer.\n"); + memset(&dev_priv->ring, 0, sizeof(dev_priv->ring)); + drm_gem_object_unreference(obj); + return -EINVAL; + } + dev_priv->ring.ring_obj = obj; + dev_priv->ring.virtual_start = dev_priv->ring.map.handle; + + /* Stop the ring if it's running. */ + I915_WRITE(PRB0_CTL, 0); + I915_WRITE(PRB0_HEAD, 0); + I915_WRITE(PRB0_TAIL, 0); + I915_WRITE(PRB0_START, 0); + + /* Initialize the ring. */ + I915_WRITE(PRB0_START, obj_priv->gtt_offset); + I915_WRITE(PRB0_CTL, + ((obj->size - 4096) & RING_NR_PAGES) | + RING_NO_REPORT | + RING_VALID); + + /* Update our cache of the ring state */ + i915_kernel_lost_context(dev); + + return 0; +} + +static void +i915_gem_cleanup_ringbuffer(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + + if (dev_priv->ring.ring_obj == NULL) + return; + + drm_core_ioremapfree(&dev_priv->ring.map, dev); + + i915_gem_object_unpin(dev_priv->ring.ring_obj); + drm_gem_object_unreference(dev_priv->ring.ring_obj); + dev_priv->ring.ring_obj = NULL; + memset(&dev_priv->ring, 0, sizeof(dev_priv->ring)); + + if (dev_priv->hws_obj != NULL) { + i915_gem_object_unpin(dev_priv->hws_obj); + drm_gem_object_unreference(dev_priv->hws_obj); + dev_priv->hws_obj = NULL; + memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); + + /* Write high address into HWS_PGA when disabling. */ + I915_WRITE(HWS_PGA, 0x1ffff000); + } +} + +int +i915_gem_entervt_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + int ret; + + if (dev_priv->mm.wedged) { + DRM_ERROR("Reenabling wedged hardware, good luck\n"); + dev_priv->mm.wedged = 0; + } + + ret = i915_gem_init_ringbuffer(dev); + if (ret != 0) + return ret; + + mutex_lock(&dev->struct_mutex); + BUG_ON(!list_empty(&dev_priv->mm.active_list)); + BUG_ON(!list_empty(&dev_priv->mm.flushing_list)); + BUG_ON(!list_empty(&dev_priv->mm.inactive_list)); + BUG_ON(!list_empty(&dev_priv->mm.request_list)); + dev_priv->mm.suspended = 0; + mutex_unlock(&dev->struct_mutex); + return 0; +} + +int +i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + int ret; + + mutex_lock(&dev->struct_mutex); + ret = i915_gem_idle(dev); + if (ret == 0) + i915_gem_cleanup_ringbuffer(dev); + mutex_unlock(&dev->struct_mutex); + + return 0; +} + +void +i915_gem_lastclose(struct drm_device *dev) +{ + int ret; + drm_i915_private_t *dev_priv = dev->dev_private; + + mutex_lock(&dev->struct_mutex); + + if (dev_priv->ring.ring_obj != NULL) { + ret = i915_gem_idle(dev); + if (ret) + DRM_ERROR("failed to idle hardware: %d\n", ret); + + i915_gem_cleanup_ringbuffer(dev); + } + + mutex_unlock(&dev->struct_mutex); +} + +void +i915_gem_load(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + + INIT_LIST_HEAD(&dev_priv->mm.active_list); + INIT_LIST_HEAD(&dev_priv->mm.flushing_list); + INIT_LIST_HEAD(&dev_priv->mm.inactive_list); + INIT_LIST_HEAD(&dev_priv->mm.request_list); + INIT_DELAYED_WORK(&dev_priv->mm.retire_work, + i915_gem_retire_work_handler); + dev_priv->mm.next_gem_seqno = 1; + + i915_gem_detect_bit_6_swizzle(dev); +} diff --git a/drivers/gpu/drm/i915/i915_gem_debug.c b/drivers/gpu/drm/i915/i915_gem_debug.c new file mode 100644 index 00000000000..131c088f8c8 --- /dev/null +++ b/drivers/gpu/drm/i915/i915_gem_debug.c @@ -0,0 +1,201 @@ +/* + * Copyright © 2008 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Keith Packard + * + */ + +#include "drmP.h" +#include "drm.h" +#include "i915_drm.h" +#include "i915_drv.h" + +#if WATCH_INACTIVE +void +i915_verify_inactive(struct drm_device *dev, char *file, int line) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + + list_for_each_entry(obj_priv, &dev_priv->mm.inactive_list, list) { + obj = obj_priv->obj; + if (obj_priv->pin_count || obj_priv->active || + (obj->write_domain & ~(I915_GEM_DOMAIN_CPU | + I915_GEM_DOMAIN_GTT))) + DRM_ERROR("inactive %p (p %d a %d w %x) %s:%d\n", + obj, + obj_priv->pin_count, obj_priv->active, + obj->write_domain, file, line); + } +} +#endif /* WATCH_INACTIVE */ + + +#if WATCH_BUF | WATCH_EXEC | WATCH_PWRITE +static void +i915_gem_dump_page(struct page *page, uint32_t start, uint32_t end, + uint32_t bias, uint32_t mark) +{ + uint32_t *mem = kmap_atomic(page, KM_USER0); + int i; + for (i = start; i < end; i += 4) + DRM_INFO("%08x: %08x%s\n", + (int) (bias + i), mem[i / 4], + (bias + i == mark) ? " ********" : ""); + kunmap_atomic(mem, KM_USER0); + /* give syslog time to catch up */ + msleep(1); +} + +void +i915_gem_dump_object(struct drm_gem_object *obj, int len, + const char *where, uint32_t mark) +{ + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int page; + + DRM_INFO("%s: object at offset %08x\n", where, obj_priv->gtt_offset); + for (page = 0; page < (len + PAGE_SIZE-1) / PAGE_SIZE; page++) { + int page_len, chunk, chunk_len; + + page_len = len - page * PAGE_SIZE; + if (page_len > PAGE_SIZE) + page_len = PAGE_SIZE; + + for (chunk = 0; chunk < page_len; chunk += 128) { + chunk_len = page_len - chunk; + if (chunk_len > 128) + chunk_len = 128; + i915_gem_dump_page(obj_priv->page_list[page], + chunk, chunk + chunk_len, + obj_priv->gtt_offset + + page * PAGE_SIZE, + mark); + } + } +} +#endif + +#if WATCH_LRU +void +i915_dump_lru(struct drm_device *dev, const char *where) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv; + + DRM_INFO("active list %s {\n", where); + list_for_each_entry(obj_priv, &dev_priv->mm.active_list, + list) + { + DRM_INFO(" %p: %08x\n", obj_priv, + obj_priv->last_rendering_seqno); + } + DRM_INFO("}\n"); + DRM_INFO("flushing list %s {\n", where); + list_for_each_entry(obj_priv, &dev_priv->mm.flushing_list, + list) + { + DRM_INFO(" %p: %08x\n", obj_priv, + obj_priv->last_rendering_seqno); + } + DRM_INFO("}\n"); + DRM_INFO("inactive %s {\n", where); + list_for_each_entry(obj_priv, &dev_priv->mm.inactive_list, list) { + DRM_INFO(" %p: %08x\n", obj_priv, + obj_priv->last_rendering_seqno); + } + DRM_INFO("}\n"); +} +#endif + + +#if WATCH_COHERENCY +void +i915_gem_object_check_coherency(struct drm_gem_object *obj, int handle) +{ + struct drm_device *dev = obj->dev; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + int page; + uint32_t *gtt_mapping; + uint32_t *backing_map = NULL; + int bad_count = 0; + + DRM_INFO("%s: checking coherency of object %p@0x%08x (%d, %dkb):\n", + __func__, obj, obj_priv->gtt_offset, handle, + obj->size / 1024); + + gtt_mapping = ioremap(dev->agp->base + obj_priv->gtt_offset, + obj->size); + if (gtt_mapping == NULL) { + DRM_ERROR("failed to map GTT space\n"); + return; + } + + for (page = 0; page < obj->size / PAGE_SIZE; page++) { + int i; + + backing_map = kmap_atomic(obj_priv->page_list[page], KM_USER0); + + if (backing_map == NULL) { + DRM_ERROR("failed to map backing page\n"); + goto out; + } + + for (i = 0; i < PAGE_SIZE / 4; i++) { + uint32_t cpuval = backing_map[i]; + uint32_t gttval = readl(gtt_mapping + + page * 1024 + i); + + if (cpuval != gttval) { + DRM_INFO("incoherent CPU vs GPU at 0x%08x: " + "0x%08x vs 0x%08x\n", + (int)(obj_priv->gtt_offset + + page * PAGE_SIZE + i * 4), + cpuval, gttval); + if (bad_count++ >= 8) { + DRM_INFO("...\n"); + goto out; + } + } + } + kunmap_atomic(backing_map, KM_USER0); + backing_map = NULL; + } + + out: + if (backing_map != NULL) + kunmap_atomic(backing_map, KM_USER0); + iounmap(gtt_mapping); + + /* give syslog time to catch up */ + msleep(1); + + /* Directly flush the object, since we just loaded values with the CPU + * from the backing pages and we don't want to disturb the cache + * management that we're trying to observe. + */ + + i915_gem_clflush_object(obj); +} +#endif diff --git a/drivers/gpu/drm/i915/i915_gem_proc.c b/drivers/gpu/drm/i915/i915_gem_proc.c new file mode 100644 index 00000000000..15d4160415b --- /dev/null +++ b/drivers/gpu/drm/i915/i915_gem_proc.c @@ -0,0 +1,292 @@ +/* + * Copyright © 2008 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt + * Keith Packard + * + */ + +#include "drmP.h" +#include "drm.h" +#include "i915_drm.h" +#include "i915_drv.h" + +static int i915_gem_active_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("Active:\n"); + list_for_each_entry(obj_priv, &dev_priv->mm.active_list, + list) + { + struct drm_gem_object *obj = obj_priv->obj; + if (obj->name) { + DRM_PROC_PRINT(" %p(%d): %08x %08x %d\n", + obj, obj->name, + obj->read_domains, obj->write_domain, + obj_priv->last_rendering_seqno); + } else { + DRM_PROC_PRINT(" %p: %08x %08x %d\n", + obj, + obj->read_domains, obj->write_domain, + obj_priv->last_rendering_seqno); + } + } + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + +static int i915_gem_flushing_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("Flushing:\n"); + list_for_each_entry(obj_priv, &dev_priv->mm.flushing_list, + list) + { + struct drm_gem_object *obj = obj_priv->obj; + if (obj->name) { + DRM_PROC_PRINT(" %p(%d): %08x %08x %d\n", + obj, obj->name, + obj->read_domains, obj->write_domain, + obj_priv->last_rendering_seqno); + } else { + DRM_PROC_PRINT(" %p: %08x %08x %d\n", obj, + obj->read_domains, obj->write_domain, + obj_priv->last_rendering_seqno); + } + } + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + +static int i915_gem_inactive_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("Inactive:\n"); + list_for_each_entry(obj_priv, &dev_priv->mm.inactive_list, + list) + { + struct drm_gem_object *obj = obj_priv->obj; + if (obj->name) { + DRM_PROC_PRINT(" %p(%d): %08x %08x %d\n", + obj, obj->name, + obj->read_domains, obj->write_domain, + obj_priv->last_rendering_seqno); + } else { + DRM_PROC_PRINT(" %p: %08x %08x %d\n", obj, + obj->read_domains, obj->write_domain, + obj_priv->last_rendering_seqno); + } + } + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + +static int i915_gem_request_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_request *gem_request; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("Request:\n"); + list_for_each_entry(gem_request, &dev_priv->mm.request_list, + list) + { + DRM_PROC_PRINT(" %d @ %d %08x\n", + gem_request->seqno, + (int) (jiffies - gem_request->emitted_jiffies), + gem_request->flush_domains); + } + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + +static int i915_gem_seqno_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("Current sequence: %d\n", i915_get_gem_seqno(dev)); + DRM_PROC_PRINT("Waiter sequence: %d\n", + dev_priv->mm.waiting_gem_seqno); + DRM_PROC_PRINT("IRQ sequence: %d\n", dev_priv->mm.irq_gem_seqno); + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + + +static int i915_interrupt_info(char *buf, char **start, off_t offset, + int request, int *eof, void *data) +{ + struct drm_minor *minor = (struct drm_minor *) data; + struct drm_device *dev = minor->dev; + drm_i915_private_t *dev_priv = dev->dev_private; + int len = 0; + + if (offset > DRM_PROC_LIMIT) { + *eof = 1; + return 0; + } + + *start = &buf[offset]; + *eof = 0; + DRM_PROC_PRINT("Interrupt enable: %08x\n", + I915_READ(IER)); + DRM_PROC_PRINT("Interrupt identity: %08x\n", + I915_READ(IIR)); + DRM_PROC_PRINT("Interrupt mask: %08x\n", + I915_READ(IMR)); + DRM_PROC_PRINT("Pipe A stat: %08x\n", + I915_READ(PIPEASTAT)); + DRM_PROC_PRINT("Pipe B stat: %08x\n", + I915_READ(PIPEBSTAT)); + DRM_PROC_PRINT("Interrupts received: %d\n", + atomic_read(&dev_priv->irq_received)); + DRM_PROC_PRINT("Current sequence: %d\n", + i915_get_gem_seqno(dev)); + DRM_PROC_PRINT("Waiter sequence: %d\n", + dev_priv->mm.waiting_gem_seqno); + DRM_PROC_PRINT("IRQ sequence: %d\n", + dev_priv->mm.irq_gem_seqno); + if (len > request + offset) + return request; + *eof = 1; + return len - offset; +} + +static struct drm_proc_list { + /** file name */ + const char *name; + /** proc callback*/ + int (*f) (char *, char **, off_t, int, int *, void *); +} i915_gem_proc_list[] = { + {"i915_gem_active", i915_gem_active_info}, + {"i915_gem_flushing", i915_gem_flushing_info}, + {"i915_gem_inactive", i915_gem_inactive_info}, + {"i915_gem_request", i915_gem_request_info}, + {"i915_gem_seqno", i915_gem_seqno_info}, + {"i915_gem_interrupt", i915_interrupt_info}, +}; + +#define I915_GEM_PROC_ENTRIES ARRAY_SIZE(i915_gem_proc_list) + +int i915_gem_proc_init(struct drm_minor *minor) +{ + struct proc_dir_entry *ent; + int i, j; + + for (i = 0; i < I915_GEM_PROC_ENTRIES; i++) { + ent = create_proc_entry(i915_gem_proc_list[i].name, + S_IFREG | S_IRUGO, minor->dev_root); + if (!ent) { + DRM_ERROR("Cannot create /proc/dri/.../%s\n", + i915_gem_proc_list[i].name); + for (j = 0; j < i; j++) + remove_proc_entry(i915_gem_proc_list[i].name, + minor->dev_root); + return -1; + } + ent->read_proc = i915_gem_proc_list[i].f; + ent->data = minor; + } + return 0; +} + +void i915_gem_proc_cleanup(struct drm_minor *minor) +{ + int i; + + if (!minor->dev_root) + return; + + for (i = 0; i < I915_GEM_PROC_ENTRIES; i++) + remove_proc_entry(i915_gem_proc_list[i].name, minor->dev_root); +} diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c new file mode 100644 index 00000000000..0c1b3a0834e --- /dev/null +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -0,0 +1,256 @@ +/* + * Copyright © 2008 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + * Authors: + * Eric Anholt + * + */ + +#include "drmP.h" +#include "drm.h" +#include "i915_drm.h" +#include "i915_drv.h" + +/** @file i915_gem_tiling.c + * + * Support for managing tiling state of buffer objects. + * + * The idea behind tiling is to increase cache hit rates by rearranging + * pixel data so that a group of pixel accesses are in the same cacheline. + * Performance improvement from doing this on the back/depth buffer are on + * the order of 30%. + * + * Intel architectures make this somewhat more complicated, though, by + * adjustments made to addressing of data when the memory is in interleaved + * mode (matched pairs of DIMMS) to improve memory bandwidth. + * For interleaved memory, the CPU sends every sequential 64 bytes + * to an alternate memory channel so it can get the bandwidth from both. + * + * The GPU also rearranges its accesses for increased bandwidth to interleaved + * memory, and it matches what the CPU does for non-tiled. However, when tiled + * it does it a little differently, since one walks addresses not just in the + * X direction but also Y. So, along with alternating channels when bit + * 6 of the address flips, it also alternates when other bits flip -- Bits 9 + * (every 512 bytes, an X tile scanline) and 10 (every two X tile scanlines) + * are common to both the 915 and 965-class hardware. + * + * The CPU also sometimes XORs in higher bits as well, to improve + * bandwidth doing strided access like we do so frequently in graphics. This + * is called "Channel XOR Randomization" in the MCH documentation. The result + * is that the CPU is XORing in either bit 11 or bit 17 to bit 6 of its address + * decode. + * + * All of this bit 6 XORing has an effect on our memory management, + * as we need to make sure that the 3d driver can correctly address object + * contents. + * + * If we don't have interleaved memory, all tiling is safe and no swizzling is + * required. + * + * When bit 17 is XORed in, we simply refuse to tile at all. Bit + * 17 is not just a page offset, so as we page an objet out and back in, + * individual pages in it will have different bit 17 addresses, resulting in + * each 64 bytes being swapped with its neighbor! + * + * Otherwise, if interleaved, we have to tell the 3d driver what the address + * swizzling it needs to do is, since it's writing with the CPU to the pages + * (bit 6 and potentially bit 11 XORed in), and the GPU is reading from the + * pages (bit 6, 9, and 10 XORed in), resulting in a cumulative bit swizzling + * required by the CPU of XORing in bit 6, 9, 10, and potentially 11, in order + * to match what the GPU expects. + */ + +/** + * Detects bit 6 swizzling of address lookup between IGD access and CPU + * access through main memory. + */ +void +i915_gem_detect_bit_6_swizzle(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + uint32_t swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN; + uint32_t swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN; + + if (!IS_I9XX(dev)) { + /* As far as we know, the 865 doesn't have these bit 6 + * swizzling issues. + */ + swizzle_x = I915_BIT_6_SWIZZLE_NONE; + swizzle_y = I915_BIT_6_SWIZZLE_NONE; + } else if (!IS_I965G(dev) || IS_I965GM(dev)) { + uint32_t dcc; + + /* On 915-945 and GM965, channel interleave by the CPU is + * determined by DCC. The CPU will alternate based on bit 6 + * in interleaved mode, and the GPU will then also alternate + * on bit 6, 9, and 10 for X, but the CPU may also optionally + * alternate based on bit 17 (XOR not disabled and XOR + * bit == 17). + */ + dcc = I915_READ(DCC); + switch (dcc & DCC_ADDRESSING_MODE_MASK) { + case DCC_ADDRESSING_MODE_SINGLE_CHANNEL: + case DCC_ADDRESSING_MODE_DUAL_CHANNEL_ASYMMETRIC: + swizzle_x = I915_BIT_6_SWIZZLE_NONE; + swizzle_y = I915_BIT_6_SWIZZLE_NONE; + break; + case DCC_ADDRESSING_MODE_DUAL_CHANNEL_INTERLEAVED: + if (IS_I915G(dev) || IS_I915GM(dev) || + dcc & DCC_CHANNEL_XOR_DISABLE) { + swizzle_x = I915_BIT_6_SWIZZLE_9_10; + swizzle_y = I915_BIT_6_SWIZZLE_9; + } else if (IS_I965GM(dev)) { + /* GM965 only does bit 11-based channel + * randomization + */ + swizzle_x = I915_BIT_6_SWIZZLE_9_10_11; + swizzle_y = I915_BIT_6_SWIZZLE_9_11; + } else { + /* Bit 17 or perhaps other swizzling */ + swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN; + swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN; + } + break; + } + if (dcc == 0xffffffff) { + DRM_ERROR("Couldn't read from MCHBAR. " + "Disabling tiling.\n"); + swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN; + swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN; + } + } else { + /* The 965, G33, and newer, have a very flexible memory + * configuration. It will enable dual-channel mode + * (interleaving) on as much memory as it can, and the GPU + * will additionally sometimes enable different bit 6 + * swizzling for tiled objects from the CPU. + * + * Here's what I found on the G965: + * slot fill memory size swizzling + * 0A 0B 1A 1B 1-ch 2-ch + * 512 0 0 0 512 0 O + * 512 0 512 0 16 1008 X + * 512 0 0 512 16 1008 X + * 0 512 0 512 16 1008 X + * 1024 1024 1024 0 2048 1024 O + * + * We could probably detect this based on either the DRB + * matching, which was the case for the swizzling required in + * the table above, or from the 1-ch value being less than + * the minimum size of a rank. + */ + if (I915_READ16(C0DRB3) != I915_READ16(C1DRB3)) { + swizzle_x = I915_BIT_6_SWIZZLE_NONE; + swizzle_y = I915_BIT_6_SWIZZLE_NONE; + } else { + swizzle_x = I915_BIT_6_SWIZZLE_9_10; + swizzle_y = I915_BIT_6_SWIZZLE_9; + } + } + + dev_priv->mm.bit_6_swizzle_x = swizzle_x; + dev_priv->mm.bit_6_swizzle_y = swizzle_y; +} + +/** + * Sets the tiling mode of an object, returning the required swizzling of + * bit 6 of addresses in the object. + */ +int +i915_gem_set_tiling(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_set_tiling *args = data; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EINVAL; + obj_priv = obj->driver_private; + + mutex_lock(&dev->struct_mutex); + + if (args->tiling_mode == I915_TILING_NONE) { + obj_priv->tiling_mode = I915_TILING_NONE; + args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE; + } else { + if (args->tiling_mode == I915_TILING_X) + args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x; + else + args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y; + /* If we can't handle the swizzling, make it untiled. */ + if (args->swizzle_mode == I915_BIT_6_SWIZZLE_UNKNOWN) { + args->tiling_mode = I915_TILING_NONE; + args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE; + } + } + obj_priv->tiling_mode = args->tiling_mode; + + mutex_unlock(&dev->struct_mutex); + + drm_gem_object_unreference(obj); + + return 0; +} + +/** + * Returns the current tiling mode and required bit 6 swizzling for the object. + */ +int +i915_gem_get_tiling(struct drm_device *dev, void *data, + struct drm_file *file_priv) +{ + struct drm_i915_gem_get_tiling *args = data; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_gem_object *obj; + struct drm_i915_gem_object *obj_priv; + + obj = drm_gem_object_lookup(dev, file_priv, args->handle); + if (obj == NULL) + return -EINVAL; + obj_priv = obj->driver_private; + + mutex_lock(&dev->struct_mutex); + + args->tiling_mode = obj_priv->tiling_mode; + switch (obj_priv->tiling_mode) { + case I915_TILING_X: + args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x; + break; + case I915_TILING_Y: + args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y; + break; + case I915_TILING_NONE: + args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE; + break; + default: + DRM_ERROR("unknown tiling mode\n"); + } + + mutex_unlock(&dev->struct_mutex); + + drm_gem_object_unreference(obj); + + return 0; +} diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index f8759597233..f295bdf16e2 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -407,15 +407,20 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) I915_WRITE(PIPEBSTAT, pipeb_stats); } - if (iir & I915_ASLE_INTERRUPT) - opregion_asle_intr(dev); + I915_WRITE(IIR, iir); + if (dev->pdev->msi_enabled) + I915_WRITE(IMR, dev_priv->irq_mask_reg); + (void) I915_READ(IIR); /* Flush posted writes */ dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); - if (dev->pdev->msi_enabled) - I915_WRITE(IMR, dev_priv->irq_mask_reg); - I915_WRITE(IIR, iir); - (void) I915_READ(IIR); + if (iir & I915_USER_INTERRUPT) { + dev_priv->mm.irq_gem_seqno = i915_get_gem_seqno(dev); + DRM_WAKEUP(&dev_priv->irq_queue); + } + + if (iir & I915_ASLE_INTERRUPT) + opregion_asle_intr(dev); if (vblank && dev_priv->swaps_pending > 0) drm_locked_tasklet(dev, i915_vblank_tasklet); @@ -449,7 +454,7 @@ static int i915_emit_irq(struct drm_device * dev) return dev_priv->counter; } -static void i915_user_irq_get(struct drm_device *dev) +void i915_user_irq_get(struct drm_device *dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 43ad2cb8b9f..5c2d9f206d0 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -25,19 +25,6 @@ #ifndef _I915_REG_H_ #define _I915_REG_H_ -/* MCH MMIO space */ -/** 915-945 and GM965 MCH register controlling DRAM channel access */ -#define DCC 0x200 -#define DCC_ADDRESSING_MODE_SINGLE_CHANNEL (0 << 0) -#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_ASYMMETRIC (1 << 0) -#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_INTERLEAVED (2 << 0) -#define DCC_ADDRESSING_MODE_MASK (3 << 0) -#define DCC_CHANNEL_XOR_DISABLE (1 << 10) - -/** 965 MCH register controlling DRAM channel configuration */ -#define CHDECMISC 0x111 -#define CHDECMISC_FLEXMEMORY (1 << 1) - /* * The Bridge device's PCI config space has information about the * fb aperture size and the amount of pre-reserved memory. @@ -516,6 +503,30 @@ #define PALETTE_A 0x0a000 #define PALETTE_B 0x0a800 +/* MCH MMIO space */ + +/* + * MCHBAR mirror. + * + * This mirrors the MCHBAR MMIO space whose location is determined by + * device 0 function 0's pci config register 0x44 or 0x48 and matches it in + * every way. It is not accessible from the CP register read instructions. + * + */ +#define MCHBAR_MIRROR_BASE 0x10000 + +/** 915-945 and GM965 MCH register controlling DRAM channel access */ +#define DCC 0x10200 +#define DCC_ADDRESSING_MODE_SINGLE_CHANNEL (0 << 0) +#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_ASYMMETRIC (1 << 0) +#define DCC_ADDRESSING_MODE_DUAL_CHANNEL_INTERLEAVED (2 << 0) +#define DCC_ADDRESSING_MODE_MASK (3 << 0) +#define DCC_CHANNEL_XOR_DISABLE (1 << 10) + +/** 965 MCH register controlling DRAM channel configuration */ +#define C0DRB3 0x10206 +#define C1DRB3 0x10606 + /* * Overlay regs */ -- cgit v1.2.3 From ed4c9c4acf948b42b138747fcb8843ecb1a24ce4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kristian=20H=C3=B8gsberg?= Date: Wed, 20 Aug 2008 11:08:52 -0400 Subject: i915: Add chip set ID param. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Kristian Høgsberg Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index f167ff68a4b..ea85d71cab0 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -689,6 +689,9 @@ static int i915_getparam(struct drm_device *dev, void *data, case I915_PARAM_LAST_DISPATCH: value = READ_BREADCRUMB(dev_priv); break; + case I915_PARAM_CHIPSET_ID: + value = dev->pci_device; + break; case I915_PARAM_HAS_GEM: value = 1; break; -- cgit v1.2.3 From 546b0974c39657017407c86fe79811100b60700d Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Mon, 1 Sep 2008 16:45:29 -0700 Subject: i915: Use struct_mutex to protect ring in GEM mode. In the conversion for GEM, we had stopped using the hardware lock to protect ring usage, since it was all internal to the DRM now. However, some paths weren't converted to using struct_mutex to prevent multiple threads from concurrently working on the ring, in particular between the vblank swap handler and ioctls. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 28 ++++++++++++++++++++++------ drivers/gpu/drm/i915/i915_drv.h | 15 +++++++++++++++ drivers/gpu/drm/i915/i915_gem.c | 2 ++ drivers/gpu/drm/i915/i915_irq.c | 28 ++++++++++++++++++++++++---- 4 files changed, 63 insertions(+), 10 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index ea85d71cab0..d71c89f8802 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -588,9 +588,15 @@ static int i915_quiescent(struct drm_device * dev) static int i915_flush_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) { - LOCK_TEST_WITH_RETURN(dev, file_priv); + int ret; + + RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - return i915_quiescent(dev); + mutex_lock(&dev->struct_mutex); + ret = i915_quiescent(dev); + mutex_unlock(&dev->struct_mutex); + + return ret; } static int i915_batchbuffer(struct drm_device *dev, void *data, @@ -611,14 +617,16 @@ static int i915_batchbuffer(struct drm_device *dev, void *data, DRM_DEBUG("i915 batchbuffer, start %x used %d cliprects %d\n", batch->start, batch->used, batch->num_cliprects); - LOCK_TEST_WITH_RETURN(dev, file_priv); + RING_LOCK_TEST_WITH_RETURN(dev, file_priv); if (batch->num_cliprects && DRM_VERIFYAREA_READ(batch->cliprects, batch->num_cliprects * sizeof(struct drm_clip_rect))) return -EFAULT; + mutex_lock(&dev->struct_mutex); ret = i915_dispatch_batchbuffer(dev, batch); + mutex_unlock(&dev->struct_mutex); sarea_priv->last_dispatch = (int)hw_status[5]; return ret; @@ -637,7 +645,7 @@ static int i915_cmdbuffer(struct drm_device *dev, void *data, DRM_DEBUG("i915 cmdbuffer, buf %p sz %d cliprects %d\n", cmdbuf->buf, cmdbuf->sz, cmdbuf->num_cliprects); - LOCK_TEST_WITH_RETURN(dev, file_priv); + RING_LOCK_TEST_WITH_RETURN(dev, file_priv); if (cmdbuf->num_cliprects && DRM_VERIFYAREA_READ(cmdbuf->cliprects, @@ -647,7 +655,9 @@ static int i915_cmdbuffer(struct drm_device *dev, void *data, return -EFAULT; } + mutex_lock(&dev->struct_mutex); ret = i915_dispatch_cmdbuffer(dev, cmdbuf); + mutex_unlock(&dev->struct_mutex); if (ret) { DRM_ERROR("i915_dispatch_cmdbuffer failed\n"); return ret; @@ -660,11 +670,17 @@ static int i915_cmdbuffer(struct drm_device *dev, void *data, static int i915_flip_bufs(struct drm_device *dev, void *data, struct drm_file *file_priv) { + int ret; + DRM_DEBUG("%s\n", __func__); - LOCK_TEST_WITH_RETURN(dev, file_priv); + RING_LOCK_TEST_WITH_RETURN(dev, file_priv); - return i915_dispatch_flip(dev); + mutex_lock(&dev->struct_mutex); + ret = i915_dispatch_flip(dev); + mutex_unlock(&dev->struct_mutex); + + return ret; } static int i915_getparam(struct drm_device *dev, void *data, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 87b071ab864..8547f0aeafc 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -285,6 +285,9 @@ typedef struct drm_i915_private { */ struct delayed_work retire_work; + /** Work task for vblank-related ring access */ + struct work_struct vblank_work; + uint32_t next_gem_seqno; /** @@ -435,6 +438,7 @@ extern int i915_irq_wait(struct drm_device *dev, void *data, void i915_user_irq_get(struct drm_device *dev); void i915_user_irq_put(struct drm_device *dev); +extern void i915_gem_vblank_work_handler(struct work_struct *work); extern irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS); extern void i915_driver_irq_preinstall(struct drm_device * dev); extern int i915_driver_irq_postinstall(struct drm_device *dev); @@ -538,6 +542,17 @@ extern void intel_opregion_free(struct drm_device *dev); extern void opregion_asle_intr(struct drm_device *dev); extern void opregion_enable_asle(struct drm_device *dev); +/** + * Lock test for when it's just for synchronization of ring access. + * + * In that case, we don't need to do it when GEM is initialized as nobody else + * has access to the ring. + */ +#define RING_LOCK_TEST_WITH_RETURN(dev, file_priv) do { \ + if (((drm_i915_private_t *)dev->dev_private)->ring.ring_obj == NULL) \ + LOCK_TEST_WITH_RETURN(dev, file_priv); \ +} while (0) + #define I915_READ(reg) DRM_READ32(dev_priv->mmio_map, (reg)) #define I915_WRITE(reg,val) DRM_WRITE32(dev_priv->mmio_map, (reg), (val)) #define I915_READ16(reg) DRM_READ16(dev_priv->mmio_map, (reg)) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 90ae8a0369f..bb6e5a37efa 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2491,6 +2491,8 @@ i915_gem_load(struct drm_device *dev) INIT_LIST_HEAD(&dev_priv->mm.request_list); INIT_DELAYED_WORK(&dev_priv->mm.retire_work, i915_gem_retire_work_handler); + INIT_WORK(&dev_priv->mm.vblank_work, + i915_gem_vblank_work_handler); dev_priv->mm.next_gem_seqno = 1; i915_gem_detect_bit_6_swizzle(dev); diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index f295bdf16e2..d04c526410a 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -349,6 +349,21 @@ u32 i915_get_vblank_counter(struct drm_device *dev, int plane) return count; } +void +i915_gem_vblank_work_handler(struct work_struct *work) +{ + drm_i915_private_t *dev_priv; + struct drm_device *dev; + + dev_priv = container_of(work, drm_i915_private_t, + mm.vblank_work); + dev = dev_priv->dev; + + mutex_lock(&dev->struct_mutex); + i915_vblank_tasklet(dev); + mutex_unlock(&dev->struct_mutex); +} + irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) { struct drm_device *dev = (struct drm_device *) arg; @@ -422,8 +437,12 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) if (iir & I915_ASLE_INTERRUPT) opregion_asle_intr(dev); - if (vblank && dev_priv->swaps_pending > 0) - drm_locked_tasklet(dev, i915_vblank_tasklet); + if (vblank && dev_priv->swaps_pending > 0) { + if (dev_priv->ring.ring_obj == NULL) + drm_locked_tasklet(dev, i915_vblank_tasklet); + else + schedule_work(&dev_priv->mm.vblank_work); + } return IRQ_HANDLED; } @@ -514,14 +533,15 @@ int i915_irq_emit(struct drm_device *dev, void *data, drm_i915_irq_emit_t *emit = data; int result; - LOCK_TEST_WITH_RETURN(dev, file_priv); + RING_LOCK_TEST_WITH_RETURN(dev, file_priv); if (!dev_priv) { DRM_ERROR("called with no initialization\n"); return -EINVAL; } - + mutex_lock(&dev->struct_mutex); result = i915_emit_irq(dev); + mutex_unlock(&dev->struct_mutex); if (DRM_COPY_TO_USER(emit->irq_seq, &result, sizeof(int))) { DRM_ERROR("copy_to_user\n"); -- cgit v1.2.3 From c99b058f132388a666544d293392d52d1def6b12 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kristian=20H=C3=B8gsberg?= Date: Wed, 20 Aug 2008 11:20:13 -0400 Subject: i915: Make use of sarea_priv conditional. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We fail ioctls that depend on the sarea_priv with EINVAL. Signed-off-by: Kristian Høgsberg Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 21 ++++++++++++++------- drivers/gpu/drm/i915/i915_irq.c | 25 +++++++++++++++++-------- 2 files changed, 31 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index d71c89f8802..048da791ca6 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -55,7 +55,8 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) if (ring->space >= n) return 0; - dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; + if (dev_priv->sarea_priv) + dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; if (ring->head != last_head) i = 0; @@ -128,7 +129,7 @@ void i915_kernel_lost_context(struct drm_device * dev) if (ring->space < 0) ring->space += ring->Size; - if (ring->head == ring->tail) + if (ring->head == ring->tail && dev_priv->sarea_priv) dev_priv->sarea_priv->perf_boxes |= I915_BOX_RING_EMPTY; } @@ -433,10 +434,11 @@ static void i915_emit_breadcrumb(struct drm_device *dev) drm_i915_private_t *dev_priv = dev->dev_private; RING_LOCALS; - dev_priv->sarea_priv->last_enqueue = ++dev_priv->counter; - + dev_priv->counter++; if (dev_priv->counter > 0x7FFFFFFFUL) - dev_priv->sarea_priv->last_enqueue = dev_priv->counter = 1; + dev_priv->counter = 0; + if (dev_priv->sarea_priv) + dev_priv->sarea_priv->last_enqueue = dev_priv->counter; BEGIN_LP_RING(4); OUT_RING(MI_STORE_DWORD_INDEX); @@ -534,6 +536,9 @@ static int i915_dispatch_flip(struct drm_device * dev) drm_i915_private_t *dev_priv = dev->dev_private; RING_LOCALS; + if (!dev_priv->sarea_priv) + return -EINVAL; + DRM_DEBUG("%s: page=%d pfCurrentPage=%d\n", __func__, dev_priv->current_page, @@ -628,7 +633,8 @@ static int i915_batchbuffer(struct drm_device *dev, void *data, ret = i915_dispatch_batchbuffer(dev, batch); mutex_unlock(&dev->struct_mutex); - sarea_priv->last_dispatch = (int)hw_status[5]; + if (sarea_priv) + sarea_priv->last_dispatch = (int)hw_status[5]; return ret; } @@ -663,7 +669,8 @@ static int i915_cmdbuffer(struct drm_device *dev, void *data, return ret; } - sarea_priv->last_dispatch = (int)hw_status[5]; + if (sarea_priv) + sarea_priv->last_dispatch = (int)hw_status[5]; return 0; } diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index d04c526410a..ef03a59c1df 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -427,7 +427,9 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) I915_WRITE(IMR, dev_priv->irq_mask_reg); (void) I915_READ(IIR); /* Flush posted writes */ - dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); + if (dev_priv->sarea_priv) + dev_priv->sarea_priv->last_dispatch = + READ_BREADCRUMB(dev_priv); if (iir & I915_USER_INTERRUPT) { dev_priv->mm.irq_gem_seqno = i915_get_gem_seqno(dev); @@ -456,10 +458,11 @@ static int i915_emit_irq(struct drm_device * dev) DRM_DEBUG("\n"); - dev_priv->sarea_priv->last_enqueue = ++dev_priv->counter; - + dev_priv->counter++; if (dev_priv->counter > 0x7FFFFFFFUL) - dev_priv->sarea_priv->last_enqueue = dev_priv->counter = 1; + dev_priv->counter = 1; + if (dev_priv->sarea_priv) + dev_priv->sarea_priv->last_enqueue = dev_priv->counter; BEGIN_LP_RING(6); OUT_RING(MI_STORE_DWORD_INDEX); @@ -503,11 +506,15 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) READ_BREADCRUMB(dev_priv)); if (READ_BREADCRUMB(dev_priv) >= irq_nr) { - dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); + if (dev_priv->sarea_priv) { + dev_priv->sarea_priv->last_dispatch = + READ_BREADCRUMB(dev_priv); + } return 0; } - dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; + if (dev_priv->sarea_priv) + dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; i915_user_irq_get(dev); DRM_WAIT_ON(ret, dev_priv->irq_queue, 3 * DRM_HZ, @@ -519,7 +526,9 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) READ_BREADCRUMB(dev_priv), (int)dev_priv->counter); } - dev_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv); + if (dev_priv->sarea_priv) + dev_priv->sarea_priv->last_dispatch = + READ_BREADCRUMB(dev_priv); return ret; } @@ -682,7 +691,7 @@ int i915_vblank_swap(struct drm_device *dev, void *data, struct list_head *list; int ret; - if (!dev_priv) { + if (!dev_priv || !dev_priv->sarea_priv) { DRM_ERROR("%s called with no initialization\n", __func__); return -EINVAL; } -- cgit v1.2.3 From dbb19d302baa8ba518599701914f600935fc53fa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kristian=20H=C3=B8gsberg?= Date: Wed, 20 Aug 2008 11:04:27 -0400 Subject: i915 gem: install and uninstall irq handler in entervt and leavevt ioctls. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Kristian Høgsberg Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index bb6e5a37efa..5fe50343de1 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2443,6 +2443,9 @@ i915_gem_entervt_ioctl(struct drm_device *dev, void *data, BUG_ON(!list_empty(&dev_priv->mm.request_list)); dev_priv->mm.suspended = 0; mutex_unlock(&dev->struct_mutex); + + drm_irq_install(dev); + return 0; } @@ -2458,6 +2461,8 @@ i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, i915_gem_cleanup_ringbuffer(dev); mutex_unlock(&dev->struct_mutex); + drm_irq_uninstall(dev); + return 0; } -- cgit v1.2.3 From d4e7b898c12b2b458323243abdd4a215f8f8f090 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Tue, 9 Sep 2008 11:40:34 -0700 Subject: DRM: Return -EBADF on bad object in flink, and return curent name if it exists. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_gem.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index 434155b387e..ccd1afdede0 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -251,7 +251,7 @@ drm_gem_flink_ioctl(struct drm_device *dev, void *data, obj = drm_gem_object_lookup(dev, file_priv, args->handle); if (obj == NULL) - return -EINVAL; + return -EBADF; again: if (idr_pre_get(&dev->object_name_idr, GFP_KERNEL) == 0) @@ -259,8 +259,9 @@ again: spin_lock(&dev->object_name_lock); if (obj->name) { + args->name = obj->name; spin_unlock(&dev->object_name_lock); - return -EEXIST; + return 0; } ret = idr_get_new_above(&dev->object_name_idr, obj, 1, &obj->name); -- cgit v1.2.3 From 4f481ed22ec0d412336a13dc4477f6d0f3688882 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 10 Sep 2008 14:22:49 -0700 Subject: drm: Avoid oops in GEM execbuffers with bad arguments. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 5fe50343de1..29d9d21e941 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1763,6 +1763,10 @@ i915_gem_execbuffer(struct drm_device *dev, void *data, (int) args->buffers_ptr, args->buffer_count, args->batch_len); #endif + if (args->buffer_count < 1) { + DRM_ERROR("execbuf with %d buffers\n", args->buffer_count); + return -EINVAL; + } /* Copy in the exec list from userland */ exec_list = drm_calloc(sizeof(*exec_list), args->buffer_count, DRM_MEM_DRIVER); -- cgit v1.2.3 From 28af0a2767412937e8424364a8ece9b230bdbc83 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Mon, 15 Sep 2008 13:13:34 -0700 Subject: drm: G33-class hardware has a newer 965-style MCH (no DCC register). Fixes bad software fallback rendering in Mesa in dual-channel configurations. d9a2470012588dc5313a5ac8bb2f03575af00e99 Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 0c1b3a0834e..6b3f1e4a34a 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -96,7 +96,7 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) */ swizzle_x = I915_BIT_6_SWIZZLE_NONE; swizzle_y = I915_BIT_6_SWIZZLE_NONE; - } else if (!IS_I965G(dev) || IS_I965GM(dev)) { + } else if ((!IS_I965G(dev) && !IS_G33(dev)) || IS_I965GM(dev)) { uint32_t dcc; /* On 915-945 and GM965, channel interleave by the CPU is -- cgit v1.2.3 From bd88ee4c1b1c8fc8b78a0ba7b6235d230cea0d05 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Tue, 23 Sep 2008 14:50:57 -0700 Subject: drm: Use ioremap_wc in i915_driver instead of ioremap, since we always want WC. Fixes failure to map the ringbuffer when PAT tells us we don't get to do uncached on something that's already mapped WC, or something along those lines. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 29d9d21e941..6ecfd108eff 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -233,7 +233,7 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, if (unwritten) #endif /* CONFIG_HIGHMEM */ { - vaddr = ioremap(pfn << PAGE_SHIFT, PAGE_SIZE); + vaddr = ioremap_wc(pfn << PAGE_SHIFT, PAGE_SIZE); #if WATCH_PWRITE DRM_INFO("pwrite slow i %d o %d l %d " "pfn %ld vaddr %p\n", @@ -1612,9 +1612,10 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, if (reloc_page != NULL) iounmap(reloc_page); - reloc_page = ioremap(dev->agp->base + - (reloc_offset & ~(PAGE_SIZE - 1)), - PAGE_SIZE); + reloc_page = ioremap_wc(dev->agp->base + + (reloc_offset & + ~(PAGE_SIZE - 1)), + PAGE_SIZE); last_reloc_offset = reloc_offset; if (reloc_page == NULL) { drm_gem_object_unreference(target_obj); @@ -2318,7 +2319,9 @@ i915_gem_init_hws(struct drm_device *dev) dev_priv->hws_map.flags = 0; dev_priv->hws_map.mtrr = 0; - drm_core_ioremap(&dev_priv->hws_map, dev); + /* Ioremapping here is the wrong thing to do. We want cached access. + */ + drm_core_ioremap_wc(&dev_priv->hws_map, dev); if (dev_priv->hws_map.handle == NULL) { DRM_ERROR("Failed to map status page.\n"); memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); @@ -2369,7 +2372,7 @@ i915_gem_init_ringbuffer(struct drm_device *dev) dev_priv->ring.map.flags = 0; dev_priv->ring.map.mtrr = 0; - drm_core_ioremap(&dev_priv->ring.map, dev); + drm_core_ioremap_wc(&dev_priv->ring.map, dev); if (dev_priv->ring.map.handle == NULL) { DRM_ERROR("Failed to map ringbuffer.\n"); memset(&dev_priv->ring, 0, sizeof(dev_priv->ring)); -- cgit v1.2.3 From 3043c60c485ad694392d3f71bd7ef9f5c5f7cfdd Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Thu, 2 Oct 2008 12:24:47 -0700 Subject: drm: Clean up many sparse warnings in i915. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 16 +++++++--------- drivers/gpu/drm/i915/i915_drv.h | 14 +++++++------- drivers/gpu/drm/i915/i915_gem.c | 31 ++++++++++++++++++------------- 3 files changed, 32 insertions(+), 29 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 048da791ca6..13a6116b59f 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -76,7 +76,7 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) * Sets up the hardware status page for devices that need a physical address * in the register. */ -int i915_init_phys_hws(struct drm_device *dev) +static int i915_init_phys_hws(struct drm_device *dev) { drm_i915_private_t *dev_priv = dev->dev_private; /* Program Hardware Status Page */ @@ -101,7 +101,7 @@ int i915_init_phys_hws(struct drm_device *dev) * Frees the hardware status page, whether it's a physical address or a virtual * address set up by the X Server. */ -void i915_free_hws(struct drm_device *dev) +static void i915_free_hws(struct drm_device *dev) { drm_i915_private_t *dev_priv = dev->dev_private; if (dev_priv->status_page_dmah) { @@ -145,8 +145,8 @@ static int i915_dma_cleanup(struct drm_device * dev) if (dev_priv->ring.virtual_start) { drm_core_ioremapfree(&dev_priv->ring.map, dev); - dev_priv->ring.virtual_start = 0; - dev_priv->ring.map.handle = 0; + dev_priv->ring.virtual_start = NULL; + dev_priv->ring.map.handle = NULL; dev_priv->ring.map.size = 0; } @@ -827,9 +827,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) base = drm_get_resource_start(dev, mmio_bar); size = drm_get_resource_len(dev, mmio_bar); - ret = drm_addmap(dev, base, size, _DRM_REGISTERS, - _DRM_KERNEL | _DRM_DRIVER, - &dev_priv->mmio_map); + dev_priv->regs = ioremap(base, size); i915_gem_load(dev); @@ -867,8 +865,8 @@ int i915_driver_unload(struct drm_device *dev) i915_free_hws(dev); - if (dev_priv->mmio_map) - drm_rmmap(dev, dev_priv->mmio_map); + if (dev_priv->regs != NULL) + iounmap(dev_priv->regs); intel_opregion_free(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 8547f0aeafc..74011235f2c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -110,8 +110,8 @@ struct intel_opregion { typedef struct drm_i915_private { struct drm_device *dev; + void __iomem *regs; drm_local_map_t *sarea; - drm_local_map_t *mmio_map; drm_i915_sarea_t *sarea_priv; drm_i915_ring_buffer_t ring; @@ -553,12 +553,12 @@ extern void opregion_enable_asle(struct drm_device *dev); LOCK_TEST_WITH_RETURN(dev, file_priv); \ } while (0) -#define I915_READ(reg) DRM_READ32(dev_priv->mmio_map, (reg)) -#define I915_WRITE(reg,val) DRM_WRITE32(dev_priv->mmio_map, (reg), (val)) -#define I915_READ16(reg) DRM_READ16(dev_priv->mmio_map, (reg)) -#define I915_WRITE16(reg,val) DRM_WRITE16(dev_priv->mmio_map, (reg), (val)) -#define I915_READ8(reg) DRM_READ8(dev_priv->mmio_map, (reg)) -#define I915_WRITE8(reg,val) DRM_WRITE8(dev_priv->mmio_map, (reg), (val)) +#define I915_READ(reg) readl(dev_priv->regs + (reg)) +#define I915_WRITE(reg, val) writel(val, dev_priv->regs + (reg)) +#define I915_READ16(reg) readw(dev_priv->regs + (reg)) +#define I915_WRITE16(reg, val) writel(val, dev_priv->regs + (reg)) +#define I915_READ8(reg) readb(dev_priv->regs + (reg)) +#define I915_WRITE8(reg, val) writeb(val, dev_priv->regs + (reg)) #define I915_VERBOSE 0 diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 6ecfd108eff..6a89449f31d 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -176,7 +176,8 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, ssize_t remain; loff_t offset; char __user *user_data; - char *vaddr; + char __iomem *vaddr; + char *vaddr_atomic; int i, o, l; int ret = 0; unsigned long pfn; @@ -219,16 +220,20 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, pfn = (dev->agp->base >> PAGE_SHIFT) + i; #ifdef CONFIG_HIGHMEM - /* kmap_atomic can't map IO pages on non-HIGHMEM kernels + /* This is a workaround for the low performance of iounmap + * (approximate 10% cpu cost on normal 3D workloads). + * kmap_atomic on HIGHMEM kernels happens to let us map card + * memory without taking IPIs. When the vmap rework lands + * we should be able to dump this hack. */ - vaddr = kmap_atomic_pfn(pfn, KM_USER0); + vaddr_atomic = kmap_atomic_pfn(pfn, KM_USER0); #if WATCH_PWRITE DRM_INFO("pwrite i %d o %d l %d pfn %ld vaddr %p\n", - i, o, l, pfn, vaddr); + i, o, l, pfn, vaddr_atomic); #endif - unwritten = __copy_from_user_inatomic_nocache(vaddr + o, + unwritten = __copy_from_user_inatomic_nocache(vaddr_atomic + o, user_data, l); - kunmap_atomic(vaddr, KM_USER0); + kunmap_atomic(vaddr_atomic, KM_USER0); if (unwritten) #endif /* CONFIG_HIGHMEM */ @@ -271,7 +276,7 @@ fail: return ret; } -int +static int i915_gem_shmem_pwrite(struct drm_device *dev, struct drm_gem_object *obj, struct drm_i915_gem_pwrite *args, struct drm_file *file_priv) @@ -587,7 +592,7 @@ i915_add_request(struct drm_device *dev, uint32_t flush_domains) * Ensures that all commands in the ring are finished * before signalling the CPU */ -uint32_t +static uint32_t i915_retire_commands(struct drm_device *dev) { drm_i915_private_t *dev_priv = dev->dev_private; @@ -734,7 +739,7 @@ i915_gem_retire_work_handler(struct work_struct *work) * Waits for a sequence number to be signaled, and cleans up the * request and object lists appropriately for that event. */ -int +static int i915_wait_request(struct drm_device *dev, uint32_t seqno) { drm_i915_private_t *dev_priv = dev->dev_private; @@ -1483,7 +1488,7 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, struct drm_i915_gem_object *obj_priv = obj->driver_private; int i, ret; uint32_t last_reloc_offset = -1; - void *reloc_page = NULL; + void __iomem *reloc_page = NULL; /* Choose the GTT offset for our buffer and put it there. */ ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment); @@ -1500,8 +1505,8 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, for (i = 0; i < entry->relocation_count; i++) { struct drm_gem_object *target_obj; struct drm_i915_gem_object *target_obj_priv; - uint32_t reloc_val, reloc_offset, *reloc_entry; - int ret; + uint32_t reloc_val, reloc_offset; + uint32_t __iomem *reloc_entry; ret = copy_from_user(&reloc, relocs + i, sizeof(reloc)); if (ret != 0) { @@ -1624,7 +1629,7 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj, } } - reloc_entry = (uint32_t *)((char *)reloc_page + + reloc_entry = (uint32_t __iomem *)(reloc_page + (reloc_offset & (PAGE_SIZE - 1))); reloc_val = target_obj_priv->gtt_offset + reloc.delta; -- cgit v1.2.3 From 2bdf00b22154023ac312481583603f4724eb1401 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 7 Oct 2008 13:40:10 +1000 Subject: i915: fixup permissions on gem ioctls. init/entervt/leavevt should be root-only master ioctls. Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 13a6116b59f..9b967bf3aa6 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -941,14 +941,14 @@ struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF(DRM_I915_GET_VBLANK_PIPE, i915_vblank_pipe_get, DRM_AUTH ), DRM_IOCTL_DEF(DRM_I915_VBLANK_SWAP, i915_vblank_swap, DRM_AUTH), DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH), - DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH), DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_UNPIN, i915_gem_unpin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_AUTH), DRM_IOCTL_DEF(DRM_I915_GEM_THROTTLE, i915_gem_throttle_ioctl, DRM_AUTH), - DRM_IOCTL_DEF(DRM_I915_GEM_ENTERVT, i915_gem_entervt_ioctl, DRM_AUTH), - DRM_IOCTL_DEF(DRM_I915_GEM_LEAVEVT, i915_gem_leavevt_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_GEM_ENTERVT, i915_gem_entervt_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), + DRM_IOCTL_DEF(DRM_I915_GEM_LEAVEVT, i915_gem_leavevt_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_CREATE, i915_gem_create_ioctl, 0), DRM_IOCTL_DEF(DRM_I915_GEM_PREAD, i915_gem_pread_ioctl, 0), DRM_IOCTL_DEF(DRM_I915_GEM_PWRITE, i915_gem_pwrite_ioctl, 0), -- cgit v1.2.3 From e7d22bc3cb57126196c4f475d4e55aa44e151784 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 7 Oct 2008 13:40:36 +1000 Subject: i915: add missing return in error path. Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 6a89449f31d..a31ee3202f1 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -146,6 +146,7 @@ i915_gem_pread_ioctl(struct drm_device *dev, void *data, if (ret != 0) { drm_gem_object_unreference(obj); mutex_unlock(&dev->struct_mutex); + return ret; } offset = args->offset; -- cgit v1.2.3 From e0f0754ff6128570dcf38032f5bfb1f195e5bbee Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 7 Oct 2008 13:41:49 +1000 Subject: drm: wbinvd is cache coherent. doing an ipi for the wbinvd case isn't necessary. Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_cache.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c index 9475f7d9901..0e994a0e46d 100644 --- a/drivers/gpu/drm/drm_cache.c +++ b/drivers/gpu/drm/drm_cache.c @@ -47,12 +47,6 @@ drm_clflush_page(struct page *page) } #endif -static void -drm_clflush_ipi_handler(void *null) -{ - wbinvd(); -} - void drm_clflush_pages(struct page *pages[], unsigned long num_pages) { @@ -68,9 +62,8 @@ drm_clflush_pages(struct page *pages[], unsigned long num_pages) return; } -#endif - if (on_each_cpu(drm_clflush_ipi_handler, NULL, 1) != 0) - DRM_ERROR("Timed out waiting for cache flush.\n"); + wbinvd(); +#endif } EXPORT_SYMBOL(drm_clflush_pages); -- cgit v1.2.3 From 9bfbd5cb72c9edb8504a4a7a0aa89cdb2fcb4845 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Mon, 15 Sep 2008 15:00:33 -0700 Subject: drm: kill drm_device->irq Like the last patch but adds a macro to get at the irq value instead of dereferencing pdev directly. Should make things easier for the BSD guys and if we ever support non-PCI devices. Signed-off-by: Jesse Barnes Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_irq.c | 7 ++----- drivers/gpu/drm/drm_stub.c | 1 - drivers/gpu/drm/mga/mga_state.c | 2 +- drivers/gpu/drm/r128/r128_state.c | 2 +- drivers/gpu/drm/radeon/radeon_state.c | 2 +- drivers/gpu/drm/via/via_irq.c | 3 --- drivers/gpu/drm/via/via_mm.c | 3 +-- 7 files changed, 6 insertions(+), 14 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c index d0c13d954f5..4091b9e291f 100644 --- a/drivers/gpu/drm/drm_irq.c +++ b/drivers/gpu/drm/drm_irq.c @@ -228,12 +228,9 @@ int drm_irq_install(struct drm_device *dev) if (drm_core_check_feature(dev, DRIVER_IRQ_SHARED)) sh_flags = IRQF_SHARED; - ret = request_irq(dev->pdev->irq, dev->driver->irq_handler, + ret = request_irq(drm_dev_to_irq(dev), dev->driver->irq_handler, sh_flags, dev->devname, dev); - /* Expose the device irq number to drivers that want to export it for - * whatever reason. - */ - dev->irq = dev->pdev->irq; + if (ret < 0) { mutex_lock(&dev->struct_mutex); dev->irq_enabled = 0; diff --git a/drivers/gpu/drm/drm_stub.c b/drivers/gpu/drm/drm_stub.c index 82f4657b887..141e33004a7 100644 --- a/drivers/gpu/drm/drm_stub.c +++ b/drivers/gpu/drm/drm_stub.c @@ -107,7 +107,6 @@ static int drm_fill_in_dev(struct drm_device * dev, struct pci_dev *pdev, #ifdef __alpha__ dev->hose = pdev->sysdata; #endif - dev->irq = pdev->irq; if (drm_ht_create(&dev->map_hash, 12)) { return -ENOMEM; diff --git a/drivers/gpu/drm/mga/mga_state.c b/drivers/gpu/drm/mga/mga_state.c index d3f8aade07b..b710fab21cb 100644 --- a/drivers/gpu/drm/mga/mga_state.c +++ b/drivers/gpu/drm/mga/mga_state.c @@ -1022,7 +1022,7 @@ static int mga_getparam(struct drm_device *dev, void *data, struct drm_file *fil switch (param->param) { case MGA_PARAM_IRQ_NR: - value = dev->irq; + value = drm_dev_to_irq(dev); break; case MGA_PARAM_CARD_TYPE: value = dev_priv->chipset; diff --git a/drivers/gpu/drm/r128/r128_state.c b/drivers/gpu/drm/r128/r128_state.c index 51a9afce7b9..f7a5b574076 100644 --- a/drivers/gpu/drm/r128/r128_state.c +++ b/drivers/gpu/drm/r128/r128_state.c @@ -1629,7 +1629,7 @@ static int r128_getparam(struct drm_device *dev, void *data, struct drm_file *fi switch (param->param) { case R128_PARAM_IRQ_NR: - value = dev->irq; + value = drm_dev_to_irq(dev); break; default: return -EINVAL; diff --git a/drivers/gpu/drm/radeon/radeon_state.c b/drivers/gpu/drm/radeon/radeon_state.c index 11c146b4921..5d7153fcc7b 100644 --- a/drivers/gpu/drm/radeon/radeon_state.c +++ b/drivers/gpu/drm/radeon/radeon_state.c @@ -2997,7 +2997,7 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil value = GET_SCRATCH(2); break; case RADEON_PARAM_IRQ_NR: - value = dev->irq; + value = drm_dev_to_irq(dev); break; case RADEON_PARAM_GART_BASE: value = dev_priv->gart_vm_start; diff --git a/drivers/gpu/drm/via/via_irq.c b/drivers/gpu/drm/via/via_irq.c index 1b966fbdb49..665d319b927 100644 --- a/drivers/gpu/drm/via/via_irq.c +++ b/drivers/gpu/drm/via/via_irq.c @@ -354,9 +354,6 @@ int via_wait_irq(struct drm_device *dev, void *data, struct drm_file *file_priv) drm_via_irq_t *cur_irq = dev_priv->via_irqs; int force_sequence; - if (!dev->irq) - return -EINVAL; - if (irqwait->request.irq >= dev_priv->num_irqs) { DRM_ERROR("Trying to wait on unknown irq %d\n", irqwait->request.irq); diff --git a/drivers/gpu/drm/via/via_mm.c b/drivers/gpu/drm/via/via_mm.c index e64094916e4..f694cb5eded 100644 --- a/drivers/gpu/drm/via/via_mm.c +++ b/drivers/gpu/drm/via/via_mm.c @@ -93,8 +93,7 @@ int via_final_context(struct drm_device *dev, int context) /* Last context, perform cleanup */ if (dev->ctx_count == 1 && dev->dev_private) { DRM_DEBUG("Last Context\n"); - if (dev->irq) - drm_irq_uninstall(dev); + drm_irq_uninstall(dev); via_cleanup_futex(dev_priv); via_do_cleanup_map(dev); } -- cgit v1.2.3 From 630681d9a5314e6cf53d144f7f58b7c19862a7d3 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Mon, 6 Oct 2008 15:14:12 -0700 Subject: drm: Increment dev_priv->irq_received so i915_gem_interrupts count works. Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_irq.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index ef03a59c1df..baae511c785 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -372,6 +372,8 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS) u32 pipea_stats, pipeb_stats; int vblank = 0; + atomic_inc(&dev_priv->irq_received); + if (dev->pdev->msi_enabled) I915_WRITE(IMR, ~0); iir = I915_READ(IIR); -- cgit v1.2.3 From 0cdad7e88a23910a911a3339ff2d70f8f952d7b8 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Tue, 14 Oct 2008 17:19:38 -0700 Subject: i915: Use non-reserved status page index for breadcrumb Dwords 0 through 0x1f are reserved for use by the hardware. Move the GEM breadcrumb from 0x10 to 0x20 to keep out of this area. Signed-off-by: Keith Packard Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_drv.h | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 74011235f2c..da7c0072ad4 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -597,16 +597,18 @@ extern void opregion_enable_asle(struct drm_device *dev); * MI_STORE_DATA_IMM. * * The following dwords have a reserved meaning: - * 0: ISR copy, updated when an ISR bit not set in the HWSTAM changes. - * 4: ring 0 head pointer - * 5: ring 1 head pointer (915-class) - * 6: ring 2 head pointer (915-class) + * 0x00: ISR copy, updated when an ISR bit not set in the HWSTAM changes. + * 0x04: ring 0 head pointer + * 0x05: ring 1 head pointer (915-class) + * 0x06: ring 2 head pointer (915-class) + * 0x10-0x1b: Context status DWords (GM45) + * 0x1f: Last written status offset. (GM45) * - * The area from dword 0x10 to 0x3ff is available for driver usage. + * The area from dword 0x20 to 0x3ff is available for driver usage. */ #define READ_HWSP(dev_priv, reg) (((volatile u32*)(dev_priv->hw_status_page))[reg]) #define READ_BREADCRUMB(dev_priv) READ_HWSP(dev_priv, 5) -#define I915_GEM_HWS_INDEX 0x10 +#define I915_GEM_HWS_INDEX 0x20 extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); -- cgit v1.2.3 From 50aa253d820ad4577e2231202f2c8fd89f9dc4e6 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Tue, 14 Oct 2008 17:20:35 -0700 Subject: i915: Fix up ring initialization to cover G45 oddities G45 appears quite sensitive to ring initialization register writes, sometimes leaving the HEAD register with the START register contents. Check to make sure HEAD is reset correctly when START is written, and fix it up, screaming loudly. Signed-off-by: Keith Packard Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem.c | 37 +++++++++++++++++++++++++++++++++++-- 1 file changed, 35 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index a31ee3202f1..94b9bdce0c7 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -2350,6 +2350,7 @@ i915_gem_init_ringbuffer(struct drm_device *dev) struct drm_gem_object *obj; struct drm_i915_gem_object *obj_priv; int ret; + u32 head; ret = i915_gem_init_hws(dev); if (ret != 0) @@ -2390,17 +2391,49 @@ i915_gem_init_ringbuffer(struct drm_device *dev) /* Stop the ring if it's running. */ I915_WRITE(PRB0_CTL, 0); - I915_WRITE(PRB0_HEAD, 0); I915_WRITE(PRB0_TAIL, 0); - I915_WRITE(PRB0_START, 0); + I915_WRITE(PRB0_HEAD, 0); /* Initialize the ring. */ I915_WRITE(PRB0_START, obj_priv->gtt_offset); + head = I915_READ(PRB0_HEAD) & HEAD_ADDR; + + /* G45 ring initialization fails to reset head to zero */ + if (head != 0) { + DRM_ERROR("Ring head not reset to zero " + "ctl %08x head %08x tail %08x start %08x\n", + I915_READ(PRB0_CTL), + I915_READ(PRB0_HEAD), + I915_READ(PRB0_TAIL), + I915_READ(PRB0_START)); + I915_WRITE(PRB0_HEAD, 0); + + DRM_ERROR("Ring head forced to zero " + "ctl %08x head %08x tail %08x start %08x\n", + I915_READ(PRB0_CTL), + I915_READ(PRB0_HEAD), + I915_READ(PRB0_TAIL), + I915_READ(PRB0_START)); + } + I915_WRITE(PRB0_CTL, ((obj->size - 4096) & RING_NR_PAGES) | RING_NO_REPORT | RING_VALID); + head = I915_READ(PRB0_HEAD) & HEAD_ADDR; + + /* If the head is still not zero, the ring is dead */ + if (head != 0) { + DRM_ERROR("Ring initialization failed " + "ctl %08x head %08x tail %08x start %08x\n", + I915_READ(PRB0_CTL), + I915_READ(PRB0_HEAD), + I915_READ(PRB0_TAIL), + I915_READ(PRB0_START)); + return -EIO; + } + /* Update our cache of the ring state */ i915_kernel_lost_context(dev); -- cgit v1.2.3 From ba1eb1d825fdef40f69871caf8e5842d00efbbc5 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Tue, 14 Oct 2008 19:55:10 -0700 Subject: i915: Map status page cached for chips with GTT-based HWS location. This should improve performance by avoiding uncached reads by the CPU (the point of having a status page), and may improve stability. This patch only affects G33, GM45 and G45 chips as those are the only ones using GTT-based HWS mappings. Signed-off-by: Keith Packard Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/drm_agpsupport.c | 5 +++-- drivers/gpu/drm/i915/i915_drv.h | 3 +++ drivers/gpu/drm/i915/i915_gem.c | 28 +++++++++++++++------------- 3 files changed, 21 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_agpsupport.c b/drivers/gpu/drm/drm_agpsupport.c index 2639be2db9e..3d33b8252b5 100644 --- a/drivers/gpu/drm/drm_agpsupport.c +++ b/drivers/gpu/drm/drm_agpsupport.c @@ -464,7 +464,8 @@ DRM_AGP_MEM * drm_agp_bind_pages(struct drm_device *dev, struct page **pages, unsigned long num_pages, - uint32_t gtt_offset) + uint32_t gtt_offset, + u32 type) { DRM_AGP_MEM *mem; int ret, i; @@ -472,7 +473,7 @@ drm_agp_bind_pages(struct drm_device *dev, DRM_DEBUG("\n"); mem = drm_agp_allocate_memory(dev->agp->bridge, num_pages, - AGP_USER_MEMORY); + type); if (mem == NULL) { DRM_ERROR("Failed to allocate memory for %ld pages\n", num_pages); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index da7c0072ad4..eae4ed3956e 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -373,6 +373,9 @@ struct drm_i915_gem_object { /** Current tiling mode for the object. */ uint32_t tiling_mode; + /** AGP mapping type (AGP_USER_MEMORY or AGP_USER_CACHED_MEMORY */ + uint32_t agp_type; + /** * Flagging of which individual pages are valid in GEM_DOMAIN_CPU when * GEM_DOMAIN_CPU is not in the object's read domain. diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 94b9bdce0c7..99a86249bb1 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1159,7 +1159,8 @@ i915_gem_object_bind_to_gtt(struct drm_gem_object *obj, unsigned alignment) obj_priv->agp_mem = drm_agp_bind_pages(dev, obj_priv->page_list, page_count, - obj_priv->gtt_offset); + obj_priv->gtt_offset, + obj_priv->agp_type); if (obj_priv->agp_mem == NULL) { i915_gem_object_free_page_list(obj); drm_mm_put_block(obj_priv->gtt_space); @@ -2142,6 +2143,8 @@ int i915_gem_init_object(struct drm_gem_object *obj) obj->write_domain = I915_GEM_DOMAIN_CPU; obj->read_domains = I915_GEM_DOMAIN_CPU; + obj_priv->agp_type = AGP_USER_MEMORY; + obj->driver_private = obj_priv; obj_priv->obj = obj; INIT_LIST_HEAD(&obj_priv->list); @@ -2311,6 +2314,7 @@ i915_gem_init_hws(struct drm_device *dev) return -ENOMEM; } obj_priv = obj->driver_private; + obj_priv->agp_type = AGP_USER_CACHED_MEMORY; ret = i915_gem_object_pin(obj, 4096); if (ret != 0) { @@ -2319,25 +2323,18 @@ i915_gem_init_hws(struct drm_device *dev) } dev_priv->status_gfx_addr = obj_priv->gtt_offset; - dev_priv->hws_map.offset = dev->agp->base + obj_priv->gtt_offset; - dev_priv->hws_map.size = 4096; - dev_priv->hws_map.type = 0; - dev_priv->hws_map.flags = 0; - dev_priv->hws_map.mtrr = 0; - /* Ioremapping here is the wrong thing to do. We want cached access. - */ - drm_core_ioremap_wc(&dev_priv->hws_map, dev); - if (dev_priv->hws_map.handle == NULL) { + dev_priv->hw_status_page = kmap(obj_priv->page_list[0]); + if (dev_priv->hw_status_page == NULL) { DRM_ERROR("Failed to map status page.\n"); memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); drm_gem_object_unreference(obj); return -EINVAL; } dev_priv->hws_obj = obj; - dev_priv->hw_status_page = dev_priv->hws_map.handle; memset(dev_priv->hw_status_page, 0, PAGE_SIZE); I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); + I915_READ(HWS_PGA); /* posting read */ DRM_DEBUG("hws offset: 0x%08x\n", dev_priv->status_gfx_addr); return 0; @@ -2456,10 +2453,15 @@ i915_gem_cleanup_ringbuffer(struct drm_device *dev) memset(&dev_priv->ring, 0, sizeof(dev_priv->ring)); if (dev_priv->hws_obj != NULL) { - i915_gem_object_unpin(dev_priv->hws_obj); - drm_gem_object_unreference(dev_priv->hws_obj); + struct drm_gem_object *obj = dev_priv->hws_obj; + struct drm_i915_gem_object *obj_priv = obj->driver_private; + + kunmap(obj_priv->page_list[0]); + i915_gem_object_unpin(obj); + drm_gem_object_unreference(obj); dev_priv->hws_obj = NULL; memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); + dev_priv->hw_status_page = NULL; /* Write high address into HWS_PGA when disabling. */ I915_WRITE(HWS_PGA, 0x1ffff000); -- cgit v1.2.3 From 6dbe2772d6af067845bab57be490c302f4490ac7 Mon Sep 17 00:00:00 2001 From: Keith Packard Date: Tue, 14 Oct 2008 21:41:13 -0700 Subject: i915: Don't run retire work handler while suspended At leavevt and lastclose time, cancel any pending retire work handler invocation, and keep the retire work handler from requeuing itself if it is currently running. This patch restructures i915_gem_idle to perform all of these tasks instead of having both leavevt and lastclose call a sequence of functions. Signed-off-by: Keith Packard Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem.c | 50 +++++++++++++++++++++++------------------ 1 file changed, 28 insertions(+), 22 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 99a86249bb1..9ac73dd1b42 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -50,6 +50,9 @@ static int i915_gem_object_get_page_list(struct drm_gem_object *obj); static void i915_gem_object_free_page_list(struct drm_gem_object *obj); static int i915_gem_object_wait_rendering(struct drm_gem_object *obj); +static void +i915_gem_cleanup_ringbuffer(struct drm_device *dev); + int i915_gem_init_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) @@ -582,7 +585,7 @@ i915_add_request(struct drm_device *dev, uint32_t flush_domains) was_empty = list_empty(&dev_priv->mm.request_list); list_add_tail(&request->list, &dev_priv->mm.request_list); - if (was_empty) + if (was_empty && !dev_priv->mm.suspended) schedule_delayed_work(&dev_priv->mm.retire_work, HZ); return seqno; } @@ -731,7 +734,8 @@ i915_gem_retire_work_handler(struct work_struct *work) mutex_lock(&dev->struct_mutex); i915_gem_retire_requests(dev); - if (!list_empty(&dev_priv->mm.request_list)) + if (!dev_priv->mm.suspended && + !list_empty(&dev_priv->mm.request_list)) schedule_delayed_work(&dev_priv->mm.retire_work, HZ); mutex_unlock(&dev->struct_mutex); } @@ -2227,14 +2231,24 @@ i915_gem_idle(struct drm_device *dev) uint32_t seqno, cur_seqno, last_seqno; int stuck, ret; - if (dev_priv->mm.suspended) + mutex_lock(&dev->struct_mutex); + + if (dev_priv->mm.suspended || dev_priv->ring.ring_obj == NULL) { + mutex_unlock(&dev->struct_mutex); return 0; + } /* Hack! Don't let anybody do execbuf while we don't control the chip. * We need to replace this with a semaphore, or something. */ dev_priv->mm.suspended = 1; + /* Cancel the retire work handler, wait for it to finish if running + */ + mutex_unlock(&dev->struct_mutex); + cancel_delayed_work_sync(&dev_priv->mm.retire_work); + mutex_lock(&dev->struct_mutex); + i915_kernel_lost_context(dev); /* Flush the GPU along with all non-CPU write domains @@ -2284,13 +2298,19 @@ i915_gem_idle(struct drm_device *dev) /* Move all buffers out of the GTT. */ ret = i915_gem_evict_from_list(dev, &dev_priv->mm.inactive_list); - if (ret) + if (ret) { + mutex_unlock(&dev->struct_mutex); return ret; + } BUG_ON(!list_empty(&dev_priv->mm.active_list)); BUG_ON(!list_empty(&dev_priv->mm.flushing_list)); BUG_ON(!list_empty(&dev_priv->mm.inactive_list)); BUG_ON(!list_empty(&dev_priv->mm.request_list)); + + i915_gem_cleanup_ringbuffer(dev); + mutex_unlock(&dev->struct_mutex); + return 0; } @@ -2503,34 +2523,20 @@ i915_gem_leavevt_ioctl(struct drm_device *dev, void *data, { int ret; - mutex_lock(&dev->struct_mutex); ret = i915_gem_idle(dev); - if (ret == 0) - i915_gem_cleanup_ringbuffer(dev); - mutex_unlock(&dev->struct_mutex); - drm_irq_uninstall(dev); - return 0; + return ret; } void i915_gem_lastclose(struct drm_device *dev) { int ret; - drm_i915_private_t *dev_priv = dev->dev_private; - - mutex_lock(&dev->struct_mutex); - - if (dev_priv->ring.ring_obj != NULL) { - ret = i915_gem_idle(dev); - if (ret) - DRM_ERROR("failed to idle hardware: %d\n", ret); - - i915_gem_cleanup_ringbuffer(dev); - } - mutex_unlock(&dev->struct_mutex); + ret = i915_gem_idle(dev); + if (ret) + DRM_ERROR("failed to idle hardware: %d\n", ret); } void -- cgit v1.2.3 From b612eda98e4b4bae4c98a863f039bc89425f9039 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 15 Oct 2008 00:05:58 -0700 Subject: i915: GM45 has GM965-style MCH setup. Fixes tiling swizzling mode failures that manifest in glReadPixels(). Signed-off-by: Eric Anholt Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_gem_tiling.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 6b3f1e4a34a..e8b85ac4ca0 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -96,7 +96,8 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) */ swizzle_x = I915_BIT_6_SWIZZLE_NONE; swizzle_y = I915_BIT_6_SWIZZLE_NONE; - } else if ((!IS_I965G(dev) && !IS_G33(dev)) || IS_I965GM(dev)) { + } else if ((!IS_I965G(dev) && !IS_G33(dev)) || IS_I965GM(dev) || + IS_GM45(dev)) { uint32_t dcc; /* On 915-945 and GM965, channel interleave by the CPU is @@ -118,7 +119,7 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) dcc & DCC_CHANNEL_XOR_DISABLE) { swizzle_x = I915_BIT_6_SWIZZLE_9_10; swizzle_y = I915_BIT_6_SWIZZLE_9; - } else if (IS_I965GM(dev)) { + } else if (IS_I965GM(dev) || IS_GM45(dev)) { /* GM965 only does bit 11-based channel * randomization */ -- cgit v1.2.3 From f0738e92403466d45cfb5008da668260c77fff4b Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 16 Oct 2008 17:12:02 +1000 Subject: drm/radeon: add support for RS740 IGP chipsets. This adds support for the HS2100 IGP chipset. Signed-off-by: Dave Airlie --- drivers/gpu/drm/radeon/radeon_cp.c | 24 ++++++++++++++++-------- drivers/gpu/drm/radeon/radeon_drv.h | 4 +++- 2 files changed, 19 insertions(+), 9 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c index 6157cd4bb43..e6e0c2933ef 100644 --- a/drivers/gpu/drm/radeon/radeon_cp.c +++ b/drivers/gpu/drm/radeon/radeon_cp.c @@ -71,7 +71,8 @@ static u32 RS690_READ_MCIND(drm_radeon_private_t *dev_priv, int addr) static u32 IGP_READ_MCIND(drm_radeon_private_t *dev_priv, int addr) { - if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) return RS690_READ_MCIND(dev_priv, addr); else return RS480_READ_MCIND(dev_priv, addr); @@ -82,7 +83,8 @@ u32 radeon_read_fb_location(drm_radeon_private_t *dev_priv) if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV515) return R500_READ_MCIND(dev_priv, RV515_MC_FB_LOCATION); - else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) return RS690_READ_MCIND(dev_priv, RS690_MC_FB_LOCATION); else if ((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_RV515) return R500_READ_MCIND(dev_priv, R520_MC_FB_LOCATION); @@ -94,7 +96,8 @@ static void radeon_write_fb_location(drm_radeon_private_t *dev_priv, u32 fb_loc) { if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV515) R500_WRITE_MCIND(RV515_MC_FB_LOCATION, fb_loc); - else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) RS690_WRITE_MCIND(RS690_MC_FB_LOCATION, fb_loc); else if ((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_RV515) R500_WRITE_MCIND(R520_MC_FB_LOCATION, fb_loc); @@ -106,7 +109,8 @@ static void radeon_write_agp_location(drm_radeon_private_t *dev_priv, u32 agp_lo { if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV515) R500_WRITE_MCIND(RV515_MC_AGP_LOCATION, agp_loc); - else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) RS690_WRITE_MCIND(RS690_MC_AGP_LOCATION, agp_loc); else if ((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_RV515) R500_WRITE_MCIND(R520_MC_AGP_LOCATION, agp_loc); @@ -122,7 +126,8 @@ static void radeon_write_agp_base(drm_radeon_private_t *dev_priv, u64 agp_base) if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV515) { R500_WRITE_MCIND(RV515_MC_AGP_BASE, agp_base_lo); R500_WRITE_MCIND(RV515_MC_AGP_BASE_2, agp_base_hi); - } else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) { + } else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) { RS690_WRITE_MCIND(RS690_MC_AGP_BASE, agp_base_lo); RS690_WRITE_MCIND(RS690_MC_AGP_BASE_2, agp_base_hi); } else if ((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_RV515) { @@ -364,8 +369,9 @@ static void radeon_cp_load_microcode(drm_radeon_private_t * dev_priv) RADEON_WRITE(RADEON_CP_ME_RAM_DATAL, R420_cp_microcode[i][0]); } - } else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) { - DRM_INFO("Loading RS690 Microcode\n"); + } else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) { + DRM_INFO("Loading RS690/RS740 Microcode\n"); for (i = 0; i < 256; i++) { RADEON_WRITE(RADEON_CP_ME_RAM_DATAH, RS690_cp_microcode[i][1]); @@ -720,7 +726,8 @@ static void radeon_set_igpgart(drm_radeon_private_t * dev_priv, int on) dev_priv->gart_size); temp = IGP_READ_MCIND(dev_priv, RS480_MC_MISC_CNTL); - if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) + if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) IGP_WRITE_MCIND(RS480_MC_MISC_CNTL, (RS480_GART_INDEX_REG_EN | RS690_BLOCK_GFX_D3_EN)); else @@ -813,6 +820,7 @@ static void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on) u32 tmp; if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740) || (dev_priv->flags & RADEON_IS_IGPGART)) { radeon_set_igpgart(dev_priv, on); return; diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h index d7e9c6cc6a1..9d752bc0602 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.h +++ b/drivers/gpu/drm/radeon/radeon_drv.h @@ -125,6 +125,7 @@ enum radeon_family { CHIP_RV410, CHIP_RS480, CHIP_RS690, + CHIP_RS740, CHIP_RV515, CHIP_R520, CHIP_RV530, @@ -1207,7 +1208,8 @@ do { \ #define IGP_WRITE_MCIND(addr, val) \ do { \ - if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) \ + if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || \ + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) \ RS690_WRITE_MCIND(addr, val); \ else \ RS480_WRITE_MCIND(addr, val); \ -- cgit v1.2.3 From b2ceddfa52cbeb244b90096f1e8d3e9f7e0ce299 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 17 Oct 2008 09:19:33 +1000 Subject: radeon: add RS400 family support. This adds support for the RS400 family of IGPs for Intel CPUs. Signed-off-by: Dave Airlie --- drivers/gpu/drm/radeon/radeon_cp.c | 6 ++++-- drivers/gpu/drm/radeon/radeon_drv.h | 1 + 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c index e6e0c2933ef..a83d7615ba7 100644 --- a/drivers/gpu/drm/radeon/radeon_cp.c +++ b/drivers/gpu/drm/radeon/radeon_cp.c @@ -133,9 +133,10 @@ static void radeon_write_agp_base(drm_radeon_private_t *dev_priv, u64 agp_base) } else if ((dev_priv->flags & RADEON_FAMILY_MASK) > CHIP_RV515) { R500_WRITE_MCIND(R520_MC_AGP_BASE, agp_base_lo); R500_WRITE_MCIND(R520_MC_AGP_BASE_2, agp_base_hi); - } else if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS480) { + } else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS400) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS480)) { RADEON_WRITE(RADEON_AGP_BASE, agp_base_lo); - RADEON_WRITE(RS480_AGP_BASE_2, 0); + RADEON_WRITE(RS480_AGP_BASE_2, agp_base_hi); } else { RADEON_WRITE(RADEON_AGP_BASE, agp_base_lo); if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R200) @@ -352,6 +353,7 @@ static void radeon_cp_load_microcode(drm_radeon_private_t * dev_priv) ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_R350) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV350) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV380) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS400) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS480)) { DRM_INFO("Loading R300 Microcode\n"); for (i = 0; i < 256; i++) { diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h index 9d752bc0602..9278429af9e 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.h +++ b/drivers/gpu/drm/radeon/radeon_drv.h @@ -123,6 +123,7 @@ enum radeon_family { CHIP_RV380, CHIP_R420, CHIP_RV410, + CHIP_RS400, CHIP_RS480, CHIP_RS690, CHIP_RS740, -- cgit v1.2.3 From edc6f389f6ae9cb7621270a8ddbb1892bd8df125 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 17 Oct 2008 09:21:45 +1000 Subject: radeon: fix PCI bus mastering support enables. Someone noticed these registers moved around for later chips, so we redo the codepaths per-chip. PCIE chips don't appear to require explicit enables. Signed-off-by: Dave Airlie --- drivers/gpu/drm/radeon/radeon_cp.c | 16 ++++++++++++++-- drivers/gpu/drm/radeon/radeon_drv.h | 25 +++++++++++++++++++++++++ 2 files changed, 39 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c index a83d7615ba7..59a2132a8f5 100644 --- a/drivers/gpu/drm/radeon/radeon_cp.c +++ b/drivers/gpu/drm/radeon/radeon_cp.c @@ -363,6 +363,7 @@ static void radeon_cp_load_microcode(drm_radeon_private_t * dev_priv) R300_cp_microcode[i][0]); } } else if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_R420) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_R423) || ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV410)) { DRM_INFO("Loading R400 Microcode\n"); for (i = 0; i < 256; i++) { @@ -652,8 +653,18 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev, RADEON_WRITE(RADEON_SCRATCH_UMSK, 0x7); /* Turn on bus mastering */ - tmp = RADEON_READ(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS; - RADEON_WRITE(RADEON_BUS_CNTL, tmp); + if (((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS400) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690) || + ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS740)) { + /* rs400, rs690/rs740 */ + tmp = RADEON_READ(RADEON_BUS_CNTL) & ~RS400_BUS_MASTER_DIS; + RADEON_WRITE(RADEON_BUS_CNTL, tmp); + } else if (!(((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RV380) || + ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_R423))) { + /* r1xx, r2xx, r300, r(v)350, r420/r481, rs480 */ + tmp = RADEON_READ(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS; + RADEON_WRITE(RADEON_BUS_CNTL, tmp); + } /* PCIE cards appears to not need this */ dev_priv->sarea_priv->last_frame = dev_priv->scratch[0] = 0; RADEON_WRITE(RADEON_LAST_FRAME_REG, dev_priv->sarea_priv->last_frame); @@ -1719,6 +1730,7 @@ int radeon_driver_load(struct drm_device *dev, unsigned long flags) case CHIP_R300: case CHIP_R350: case CHIP_R420: + case CHIP_R423: case CHIP_RV410: case CHIP_RV515: case CHIP_R520: diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h index 9278429af9e..4dbb813910c 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.h +++ b/drivers/gpu/drm/radeon/radeon_drv.h @@ -122,6 +122,7 @@ enum radeon_family { CHIP_RV350, CHIP_RV380, CHIP_R420, + CHIP_R423, CHIP_RV410, CHIP_RS400, CHIP_RS480, @@ -439,8 +440,31 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, # define RADEON_SCISSOR_1_ENABLE (1 << 29) # define RADEON_SCISSOR_2_ENABLE (1 << 30) +/* + * PCIE radeons (rv370/rv380, rv410, r423/r430/r480, r5xx) + * don't have an explicit bus mastering disable bit. It's handled + * by the PCI D-states. PMI_BM_DIS disables D-state bus master + * handling, not bus mastering itself. + */ #define RADEON_BUS_CNTL 0x0030 +/* r1xx, r2xx, r300, r(v)350, r420/r481, rs480 */ # define RADEON_BUS_MASTER_DIS (1 << 6) +/* rs400, rs690/rs740 */ +# define RS400_BUS_MASTER_DIS (1 << 14) +# define RS400_MSI_REARM (1 << 20) +/* see RS480_MSI_REARM in AIC_CNTL for rs480 */ + +#define RADEON_BUS_CNTL1 0x0034 +# define RADEON_PMI_BM_DIS (1 << 2) +# define RADEON_PMI_INT_DIS (1 << 3) + +#define RV370_BUS_CNTL 0x004c +# define RV370_PMI_BM_DIS (1 << 5) +# define RV370_PMI_INT_DIS (1 << 6) + +#define RADEON_MSI_REARM_EN 0x0160 +/* rv370/rv380, rv410, r423/r430/r480, r5xx */ +# define RV370_MSI_REARM_EN (1 << 0) #define RADEON_CLOCK_CNTL_DATA 0x000c # define RADEON_PLL_WR_EN (1 << 7) @@ -913,6 +937,7 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev, #define RADEON_AIC_CNTL 0x01d0 # define RADEON_PCIGART_TRANSLATE_EN (1 << 0) +# define RS480_MSI_REARM (1 << 3) #define RADEON_AIC_STAT 0x01d4 #define RADEON_AIC_PT_BASE 0x01d8 #define RADEON_AIC_LO_ADDR 0x01dc -- cgit v1.2.3 From 9e0b97e37fddaf5419d8af24362015ab684eff7e Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Fri, 17 Oct 2008 09:29:14 +1000 Subject: drm: make CONFIG_DRM depend on CONFIG_SHMEM. This can be removed later when DRM doesn't depend on shmem. Signed-off-by: Dave Airlie --- drivers/gpu/drm/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index bf9003ee7ab..9097500de5f 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -6,7 +6,7 @@ # menuconfig DRM tristate "Direct Rendering Manager (XFree86 4.1.0 and higher DRI support)" - depends on (AGP || AGP=n) && PCI && !EMULATED_CMPXCHG + depends on (AGP || AGP=n) && PCI && !EMULATED_CMPXCHG && SHMEM help Kernel-level support for the Direct Rendering Infrastructure (DRI) introduced in XFree86 4.0. If you say Y here, you need to select -- cgit v1.2.3 From 4b40893918203ee1a1f6a114316c2a19c072e9bd Mon Sep 17 00:00:00 2001 From: Matthias Hopf Date: Sat, 18 Oct 2008 07:18:05 +1000 Subject: drm/i915: fix ioremap of a user address for non-root (CVE-2008-3831) Olaf Kirch noticed that the i915_set_status_page() function of the i915 kernel driver calls ioremap with an address offset that is supplied by userspace via ioctl. The function zeroes the mapped memory via memset and tells the hardware about the address. Turns out that access to that ioctl is not restricted to root so users could probably exploit that to do nasty things. We haven't tried to write actual exploit code though. It only affects the Intel G33 series and newer. Signed-off-by: Dave Airlie --- drivers/gpu/drm/i915/i915_dma.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 9b967bf3aa6..db34780edbb 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -940,7 +940,7 @@ struct drm_ioctl_desc i915_ioctls[] = { DRM_IOCTL_DEF(DRM_I915_SET_VBLANK_PIPE, i915_vblank_pipe_set, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY ), DRM_IOCTL_DEF(DRM_I915_GET_VBLANK_PIPE, i915_vblank_pipe_get, DRM_AUTH ), DRM_IOCTL_DEF(DRM_I915_VBLANK_SWAP, i915_vblank_swap, DRM_AUTH), - DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH), + DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY), DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH), DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY), -- cgit v1.2.3 From 1ae87786800b5e0411847974b211797b6ada63c4 Mon Sep 17 00:00:00 2001 From: Linus Torvalds Date: Mon, 20 Oct 2008 14:14:25 -0700 Subject: Fix sprintf format warnings in drm_proc.c Use "%zd" for size_t, and make sure to have a space between the numbers instead of depending on the field width. I don't like warnings in my default targeted build. Signed-off-by: Linus Torvalds --- drivers/gpu/drm/drm_proc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/drm_proc.c b/drivers/gpu/drm/drm_proc.c index d490db4c0de..ae73b7f7249 100644 --- a/drivers/gpu/drm/drm_proc.c +++ b/drivers/gpu/drm/drm_proc.c @@ -522,12 +522,12 @@ static int drm_gem_one_name_info(int id, void *ptr, void *data) struct drm_gem_object *obj = ptr; struct drm_gem_name_info_data *nid = data; - DRM_INFO("name %d size %d\n", obj->name, obj->size); + DRM_INFO("name %d size %zd\n", obj->name, obj->size); if (nid->eof) return 0; nid->len += sprintf(&nid->buf[nid->len], - "%6d%9d%8d%9d\n", + "%6d %8zd %7d %8d\n", obj->name, obj->size, atomic_read(&obj->handlecount.refcount), atomic_read(&obj->refcount.refcount)); -- cgit v1.2.3 From 9b7530cc329eb036cfa589930c270e85031f554c Mon Sep 17 00:00:00 2001 From: Linus Torvalds Date: Mon, 20 Oct 2008 14:16:43 -0700 Subject: i915: cleanup coding horrors in i915_gem_gtt_pwrite() Yes, this will probably be switched over to a cleaner model anyway, but in the meantime I don't want to see the 'unused variable' warnings that come from the disgusting #ifdef code. Make the special case be a nice inlien function of its own, clean up the code, and make the warning go away. I wish people didn't write code that gets (valid) warnings from the compiler, but I'll limit my fixes to code that I actually care about (in this case just because I see the warning and it annoys me). Signed-off-by: Linus Torvalds --- drivers/gpu/drm/i915/i915_gem.c | 59 +++++++++++++++++++++++++---------------- 1 file changed, 36 insertions(+), 23 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 9ac73dd1b42..49c5a1798ac 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -171,6 +171,36 @@ i915_gem_pread_ioctl(struct drm_device *dev, void *data, return 0; } +/* + * Try to write quickly with an atomic kmap. Return true on success. + * + * If this fails (which includes a partial write), we'll redo the whole + * thing with the slow version. + * + * This is a workaround for the low performance of iounmap (approximate + * 10% cpu cost on normal 3D workloads). kmap_atomic on HIGHMEM kernels + * happens to let us map card memory without taking IPIs. When the vmap + * rework lands we should be able to dump this hack. + */ +static inline int fast_user_write(unsigned long pfn, char __user *user_data, int l) +{ +#ifdef CONFIG_HIGHMEM + unsigned long unwritten; + char *vaddr_atomic; + + vaddr_atomic = kmap_atomic_pfn(pfn, KM_USER0); +#if WATCH_PWRITE + DRM_INFO("pwrite i %d o %d l %d pfn %ld vaddr %p\n", + i, o, l, pfn, vaddr_atomic); +#endif + unwritten = __copy_from_user_inatomic_nocache(vaddr_atomic + o, user_data, l); + kunmap_atomic(vaddr_atomic, KM_USER0); + return !unwritten; +#else + return 0; +#endif +} + static int i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, struct drm_i915_gem_pwrite *args, @@ -180,12 +210,7 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, ssize_t remain; loff_t offset; char __user *user_data; - char __iomem *vaddr; - char *vaddr_atomic; - int i, o, l; int ret = 0; - unsigned long pfn; - unsigned long unwritten; user_data = (char __user *) (uintptr_t) args->data_ptr; remain = args->size; @@ -209,6 +234,9 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, obj_priv->dirty = 1; while (remain > 0) { + unsigned long pfn; + int i, o, l; + /* Operation in this page * * i = page number @@ -223,25 +251,10 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, pfn = (dev->agp->base >> PAGE_SHIFT) + i; -#ifdef CONFIG_HIGHMEM - /* This is a workaround for the low performance of iounmap - * (approximate 10% cpu cost on normal 3D workloads). - * kmap_atomic on HIGHMEM kernels happens to let us map card - * memory without taking IPIs. When the vmap rework lands - * we should be able to dump this hack. - */ - vaddr_atomic = kmap_atomic_pfn(pfn, KM_USER0); -#if WATCH_PWRITE - DRM_INFO("pwrite i %d o %d l %d pfn %ld vaddr %p\n", - i, o, l, pfn, vaddr_atomic); -#endif - unwritten = __copy_from_user_inatomic_nocache(vaddr_atomic + o, - user_data, l); - kunmap_atomic(vaddr_atomic, KM_USER0); + if (!fast_user_write(pfn, user_data, l)) { + unsigned long unwritten; + char __iomem *vaddr; - if (unwritten) -#endif /* CONFIG_HIGHMEM */ - { vaddr = ioremap_wc(pfn << PAGE_SHIFT, PAGE_SIZE); #if WATCH_PWRITE DRM_INFO("pwrite slow i %d o %d l %d " -- cgit v1.2.3 From e8848a170fd432bdda176a2d568919d4bba90467 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Tue, 21 Oct 2008 00:47:45 +0200 Subject: fix CONFIG_HIGHMEM compile error in drivers/gpu/drm/i915/i915_gem.c commit 9b7530cc329eb036cfa589930c270e85031f554c ("i915: cleanup coding horrors in i915_gem_gtt_pwrite()") broke the i386 build for CONFIG_HIGHMEM=y. Caught by automatic testing http://www.tglx.de/autoqa-logs/000137-0006-0001.log Signed-off-by: Thomas Gleixner [ My bad. It's the same patch I sent out earlier, nobody noticed then either.. ] Signed-off-by: Linus Torvalds --- drivers/gpu/drm/i915/i915_gem.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 49c5a1798ac..dc2e6fdb6ca 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -182,7 +182,8 @@ i915_gem_pread_ioctl(struct drm_device *dev, void *data, * happens to let us map card memory without taking IPIs. When the vmap * rework lands we should be able to dump this hack. */ -static inline int fast_user_write(unsigned long pfn, char __user *user_data, int l) +static inline int fast_user_write(unsigned long pfn, char __user *user_data, + int l, int o) { #ifdef CONFIG_HIGHMEM unsigned long unwritten; @@ -251,7 +252,7 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj, pfn = (dev->agp->base >> PAGE_SHIFT) + i; - if (!fast_user_write(pfn, user_data, l)) { + if (!fast_user_write(pfn, user_data, l, o)) { unsigned long unwritten; char __iomem *vaddr; -- cgit v1.2.3 From e798ba57e9f423dddbf1bdeb20a62bdd0593890f Mon Sep 17 00:00:00 2001 From: Hugh Dickins Date: Tue, 21 Oct 2008 00:04:04 +0100 Subject: Export tiny shmem_file_setup for DRM-GEM We're trying to keep the !CONFIG_SHMEM tiny-shmem.c (using ramfs without swap) in synch with CONFIG_SHMEM shmem.c (and mpm is preparing patches to combine them). I was glad to see EXPORT_SYMBOL_GPL(shmem_file_setup) go into shmem.c, but why not support DRM-GEM when !CONFIG_SHMEM too? But caution says still depend on MMU, since !CONFIG_MMU is.. different. Signed-off-by: Hugh Dickins Acked-by: Matt Mackall Acked-by: Dave Airlie Signed-off-by: Linus Torvalds --- drivers/gpu/drm/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/Kconfig b/drivers/gpu/drm/Kconfig index 9097500de5f..a8b33c2ec8d 100644 --- a/drivers/gpu/drm/Kconfig +++ b/drivers/gpu/drm/Kconfig @@ -6,7 +6,7 @@ # menuconfig DRM tristate "Direct Rendering Manager (XFree86 4.1.0 and higher DRI support)" - depends on (AGP || AGP=n) && PCI && !EMULATED_CMPXCHG && SHMEM + depends on (AGP || AGP=n) && PCI && !EMULATED_CMPXCHG && MMU help Kernel-level support for the Direct Rendering Infrastructure (DRI) introduced in XFree86 4.0. If you say Y here, you need to select -- cgit v1.2.3