From 5e5a29bf2624a5984e1c36c3a2481ee91249ec9c Mon Sep 17 00:00:00 2001 From: Anders H Kaseorg Date: Sat, 28 Jun 2008 18:25:41 -0400 Subject: x86, 64-bit: patch paravirt inline replacements when loading modules small speedup. Paravirt replacements were added to the i386 module loader by commit 139ec7c416248b9ea227d21839235344edfee1e0. This adds the same code to the x86_64 module loader. Signed-off-by: Anders Kaseorg Acked-by: "H. Peter Anvin" Signed-off-by: Ingo Molnar --- arch/x86/kernel/module_64.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/module_64.c b/arch/x86/kernel/module_64.c index a888e67f587..0e867676b5a 100644 --- a/arch/x86/kernel/module_64.c +++ b/arch/x86/kernel/module_64.c @@ -150,7 +150,8 @@ int module_finalize(const Elf_Ehdr *hdr, const Elf_Shdr *sechdrs, struct module *me) { - const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL; + const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL, + *para = NULL; char *secstrings = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset; for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) { @@ -160,6 +161,8 @@ int module_finalize(const Elf_Ehdr *hdr, alt = s; if (!strcmp(".smp_locks", secstrings + s->sh_name)) locks= s; + if (!strcmp(".parainstructions", secstrings + s->sh_name)) + para = s; } if (alt) { @@ -175,6 +178,11 @@ int module_finalize(const Elf_Ehdr *hdr, tseg, tseg + text->sh_size); } + if (para) { + void *pseg = (void *)para->sh_addr; + apply_paravirt(pseg, pseg + para->sh_size); + } + return module_bug_finalize(hdr, sechdrs, me); } -- cgit v1.2.3 From 299a140dacaa514be5e567b5851c187c42ec38c4 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Tue, 8 Jul 2008 14:47:16 +0200 Subject: x86, AMD IOMMU: ignore detection of GART IOMMU One of the last IOMMU updates covered a bug in the AMD IOMMU code. The early detection code does not succeed if the GART is already detected. This patch fixes this. Cc: Robert Richter Cc: Bhavna Sarathy Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: Joerg Roedel Cc: Robert Richter Cc: Bhavna Sarathy Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 2a13e430437..bb0280077a3 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -828,7 +828,7 @@ static int __init early_amd_iommu_detect(struct acpi_table_header *table) void __init amd_iommu_detect(void) { - if (swiotlb || no_iommu || iommu_detected) + if (swiotlb || no_iommu || (iommu_detected && !gart_iommu_aperture)) return; if (acpi_table_parse("IVRS", early_amd_iommu_detect) == 0) { -- cgit v1.2.3 From ab6bc3e343fbe3be4a0f67225e849d0db6b4b7ac Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 5 Jul 2008 15:53:36 +0400 Subject: x86: idle process - add checking for NULL early param Signed-off-by: Cyrill Gorcunov Cc: akpm@linux-foundation.org Cc: andi@firstfloor.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/process.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index ba370dc8685..58325a6604a 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c @@ -164,6 +164,9 @@ void __cpuinit select_idle_routine(const struct cpuinfo_x86 *c) static int __init idle_setup(char *str) { + if (!str) + return -EINVAL; + if (!strcmp(str, "poll")) { printk("using polling idle threads.\n"); pm_idle = poll_idle; -- cgit v1.2.3 From d6cd7effcc5e0047faf15ab0a54c980f1a616a07 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 5 Jul 2008 15:53:37 +0400 Subject: x86: io delay - add checking for NULL early param Signed-off-by: Cyrill Gorcunov Cc: akpm@linux-foundation.org Cc: andi@firstfloor.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/io_delay.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/io_delay.c b/arch/x86/kernel/io_delay.c index 5921e5f0a64..1c3a66a67f8 100644 --- a/arch/x86/kernel/io_delay.c +++ b/arch/x86/kernel/io_delay.c @@ -103,6 +103,9 @@ void __init io_delay_init(void) static int __init io_delay_param(char *s) { + if (!s) + return -EINVAL; + if (!strcmp(s, "0x80")) io_delay_type = CONFIG_IO_DELAY_TYPE_0X80; else if (!strcmp(s, "0xed")) -- cgit v1.2.3 From 4d8cc874d7ed43eda72765e9c0e141e170fee4f3 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Sat, 5 Jul 2008 15:53:38 +0400 Subject: x86: smpboot maxcpus - add checking for NULL early param Signed-off-by: Cyrill Gorcunov Cc: akpm@linux-foundation.org Cc: andi@firstfloor.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/smpboot.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index fd933b5465b..e47bfac70c3 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -1452,7 +1452,8 @@ static int __init parse_maxcpus(char *arg) { extern unsigned int maxcpus; - maxcpus = simple_strtoul(arg, NULL, 0); + if (arg) + maxcpus = simple_strtoul(arg, NULL, 0); return 0; } early_param("maxcpus", parse_maxcpus); -- cgit v1.2.3 From 46a7fa270afbe5fddc6042a598cfe22977b0e989 Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Fri, 11 Jul 2008 10:23:42 +0900 Subject: x86: make only GART code include gart.h gart.h has only GART-specific stuff. Only GART code needs it. Other IOMMU stuff should include iommu.h instead of gart.h. Signed-off-by: FUJITA Tomonori Acked-by: Muli Ben-Yehuda Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu.c | 2 +- arch/x86/kernel/amd_iommu_init.c | 2 +- arch/x86/kernel/aperture_64.c | 1 + arch/x86/kernel/early-quirks.c | 5 +---- arch/x86/kernel/pci-calgary_64.c | 2 +- arch/x86/kernel/pci-dma.c | 2 +- arch/x86/kernel/pci-gart_64.c | 1 + arch/x86/kernel/pci-nommu.c | 2 +- arch/x86/kernel/pci-swiotlb_64.c | 2 +- arch/x86/kernel/setup.c | 2 +- 10 files changed, 10 insertions(+), 11 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index f2766d84c7a..cf2f74bcde5 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -23,7 +23,7 @@ #include #include #include -#include +#include #include #include diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 2a13e430437..66438284c69 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -25,7 +25,7 @@ #include #include #include -#include +#include /* * definitions for the ACPI scanning code diff --git a/arch/x86/kernel/aperture_64.c b/arch/x86/kernel/aperture_64.c index 9f907806c1a..44e21826db1 100644 --- a/arch/x86/kernel/aperture_64.c +++ b/arch/x86/kernel/aperture_64.c @@ -21,6 +21,7 @@ #include #include #include +#include #include #include #include diff --git a/arch/x86/kernel/early-quirks.c b/arch/x86/kernel/early-quirks.c index a4665f37cfc..510b8e36773 100644 --- a/arch/x86/kernel/early-quirks.c +++ b/arch/x86/kernel/early-quirks.c @@ -16,10 +16,7 @@ #include #include #include - -#ifdef CONFIG_GART_IOMMU -#include -#endif +#include static void __init fix_hypertransport_config(int num, int slot, int func) { diff --git a/arch/x86/kernel/pci-calgary_64.c b/arch/x86/kernel/pci-calgary_64.c index 6959b5c45df..151f2d171f7 100644 --- a/arch/x86/kernel/pci-calgary_64.c +++ b/arch/x86/kernel/pci-calgary_64.c @@ -36,7 +36,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index 8467ec2320f..f581a4b63b4 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -5,7 +5,7 @@ #include #include -#include +#include #include #include diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c index d0d18db5d2a..949ca985deb 100644 --- a/arch/x86/kernel/pci-gart_64.c +++ b/arch/x86/kernel/pci-gart_64.c @@ -32,6 +32,7 @@ #include #include #include +#include #include #include #include diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c index aec43d56f49..792b9179eff 100644 --- a/arch/x86/kernel/pci-nommu.c +++ b/arch/x86/kernel/pci-nommu.c @@ -7,7 +7,7 @@ #include #include -#include +#include #include #include diff --git a/arch/x86/kernel/pci-swiotlb_64.c b/arch/x86/kernel/pci-swiotlb_64.c index 82299cd1d04..20df839b9c2 100644 --- a/arch/x86/kernel/pci-swiotlb_64.c +++ b/arch/x86/kernel/pci-swiotlb_64.c @@ -5,7 +5,7 @@ #include #include -#include +#include #include #include diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 86fc2d62427..e5d208934bf 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -96,7 +96,7 @@ #include #include #include -#include +#include #include #include -- cgit v1.2.3 From ac7ded2adb2e43152fe7385ddd53bf45f5c92285 Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Fri, 11 Jul 2008 10:23:43 +0900 Subject: x86: remove ifdef CONFIG_GART_IOMMU in pci-dma.c Our way to handle gart_* functions for CONFIG_GART_IOMMU and !CONFIG_GART_IOMMU cases is inconsistent. We have some dummy gart_* functions in !CONFIG_GART_IOMMU case and also use ifdef CONFIG_GART_IOMMU tricks in pci-dma.c to call some gart_* functions in only CONFIG_GART_IOMMU case. This patch removes ifdef CONFIG_GART_IOMMU in pci-dma.c and always use dummy gart_* functions in iommu.h. Signed-off-by: FUJITA Tomonori Acked-by: Muli Ben-Yehuda Signed-off-by: Ingo Molnar --- arch/x86/kernel/pci-dma.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index f581a4b63b4..dd57c5bbe2d 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -114,9 +114,7 @@ void __init pci_iommu_alloc(void) * The order of these functions is important for * fall-back/fail-over reasons */ -#ifdef CONFIG_GART_IOMMU gart_iommu_hole_init(); -#endif #ifdef CONFIG_CALGARY_IOMMU detect_calgary(); @@ -184,9 +182,7 @@ static __init int iommu_setup(char *p) swiotlb = 1; #endif -#ifdef CONFIG_GART_IOMMU gart_parse_options(p); -#endif #ifdef CONFIG_CALGARY_IOMMU if (!strncmp(p, "calgary", 7)) @@ -508,9 +504,7 @@ static int __init pci_iommu_init(void) amd_iommu_init(); -#ifdef CONFIG_GART_IOMMU gart_iommu_init(); -#endif no_iommu_init(); return 0; -- cgit v1.2.3 From b8b48326f312026af12799917383c54c25d05482 Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Fri, 11 Jul 2008 10:23:44 +0900 Subject: x86: remove ifdef CONFIG_CALGARY_IOMMU in pci-dma.c asm-x86/calgary.h has dummy calgary_iommu_init() and detect_calgary() in !CONFIG_CALGARY_IOMMU case. So we don't need ifdef CONFIG_CALGARY_IOMMU in pci-dma.c. Signed-off-by: FUJITA Tomonori Acked-by: Muli Ben-Yehuda Cc: Alexis Bruemmer Signed-off-by: Ingo Molnar --- arch/x86/kernel/pci-dma.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index dd57c5bbe2d..f16cbbe424a 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -116,9 +116,7 @@ void __init pci_iommu_alloc(void) */ gart_iommu_hole_init(); -#ifdef CONFIG_CALGARY_IOMMU detect_calgary(); -#endif detect_intel_iommu(); @@ -496,9 +494,7 @@ EXPORT_SYMBOL(dma_free_coherent); static int __init pci_iommu_init(void) { -#ifdef CONFIG_CALGARY_IOMMU calgary_iommu_init(); -#endif intel_iommu_init(); -- cgit v1.2.3 From be54f9d1c8df93c4998e134a306652caaa58f67f Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Fri, 11 Jul 2008 10:23:45 +0900 Subject: x86: remove ifdef CONFIG_SWIOTLB in pci-dma.c As other IOMMUs do, this puts dummy pci_swiotlb_init() in swiotlb.h and remove ifdef CONFIG_SWIOTLB in pci-dma.c. Signed-off-by: FUJITA Tomonori Acked-by: Muli Ben-Yehuda Signed-off-by: Ingo Molnar --- arch/x86/kernel/pci-dma.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index f16cbbe424a..d12945de056 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -122,9 +122,7 @@ void __init pci_iommu_alloc(void) amd_iommu_detect(); -#ifdef CONFIG_SWIOTLB pci_swiotlb_init(); -#endif } #endif -- cgit v1.2.3 From b65233a9c1da587bf19ee161982f4f0ec59941c0 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:21 +0200 Subject: x86, AMD IOMMU: add comments to the initialization code This patch adds some comments to the AMD IOMMU initialization code to increase its readability. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 214 +++++++++++++++++++++++++++++++++++++-- 1 file changed, 206 insertions(+), 8 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index bb0280077a3..9ddb46d7c52 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -71,6 +71,17 @@ #define ACPI_DEVFLAG_LINT1 0x80 #define ACPI_DEVFLAG_ATSDIS 0x10000000 +/* + * ACPI table definitions + * + * These data structures are laid over the table to parse the important values + * out of it. + */ + +/* + * structure describing one IOMMU in the ACPI table. Typically followed by one + * or more ivhd_entrys. + */ struct ivhd_header { u8 type; u8 flags; @@ -83,6 +94,10 @@ struct ivhd_header { u32 reserved; } __attribute__((packed)); +/* + * A device entry describing which devices a specific IOMMU translates and + * which requestor ids they use. + */ struct ivhd_entry { u8 type; u16 devid; @@ -90,6 +105,10 @@ struct ivhd_entry { u32 ext; } __attribute__((packed)); +/* + * An AMD IOMMU memory definition structure. It defines things like exclusion + * ranges for devices and regions that should be unity mapped. + */ struct ivmd_header { u8 type; u8 flags; @@ -103,22 +122,66 @@ struct ivmd_header { static int __initdata amd_iommu_detected; -u16 amd_iommu_last_bdf; -struct list_head amd_iommu_unity_map; -unsigned amd_iommu_aperture_order = 26; -int amd_iommu_isolate; +u16 amd_iommu_last_bdf; /* largest PCI device id we have + to handle */ +struct list_head amd_iommu_unity_map; /* a list of required unity mappings + we find in ACPI */ +unsigned amd_iommu_aperture_order = 26; /* size of aperture in power of 2 */ +int amd_iommu_isolate; /* if 1, device isolation is enabled */ -struct list_head amd_iommu_list; +struct list_head amd_iommu_list; /* list of all AMD IOMMUs in the + system */ + +/* + * Pointer to the device table which is shared by all AMD IOMMUs + * it is indexed by the PCI device id or the HT unit id and contains + * information about the domain the device belongs to as well as the + * page table root pointer. + */ struct dev_table_entry *amd_iommu_dev_table; + +/* + * The alias table is a driver specific data structure which contains the + * mappings of the PCI device ids to the actual requestor ids on the IOMMU. + * More than one device can share the same requestor id. + */ u16 *amd_iommu_alias_table; + +/* + * The rlookup table is used to find the IOMMU which is responsible + * for a specific device. It is also indexed by the PCI device id. + */ struct amd_iommu **amd_iommu_rlookup_table; + +/* + * The pd table (protection domain table) is used to find the protection domain + * data structure a device belongs to. Indexed with the PCI device id too. + */ struct protection_domain **amd_iommu_pd_table; + +/* + * AMD IOMMU allows up to 2^16 differend protection domains. This is a bitmap + * to know which ones are already in use. + */ unsigned long *amd_iommu_pd_alloc_bitmap; -static u32 dev_table_size; -static u32 alias_table_size; -static u32 rlookup_table_size; +static u32 dev_table_size; /* size of the device table */ +static u32 alias_table_size; /* size of the alias table */ +static u32 rlookup_table_size; /* size if the rlookup table */ + +/**************************************************************************** + * + * AMD IOMMU MMIO register space handling functions + * + * These functions are used to program the IOMMU device registers in + * MMIO space required for that driver. + * + ****************************************************************************/ +/* + * This function set the exclusion range in the IOMMU. DMA accesses to the + * exclusion range are passed through untranslated + */ static void __init iommu_set_exclusion_range(struct amd_iommu *iommu) { u64 start = iommu->exclusion_start & PAGE_MASK; @@ -137,6 +200,7 @@ static void __init iommu_set_exclusion_range(struct amd_iommu *iommu) &entry, sizeof(entry)); } +/* Programs the physical address of the device table into the IOMMU hardware */ static void __init iommu_set_device_table(struct amd_iommu *iommu) { u32 entry; @@ -149,6 +213,7 @@ static void __init iommu_set_device_table(struct amd_iommu *iommu) &entry, sizeof(entry)); } +/* Generic functions to enable/disable certain features of the IOMMU. */ static void __init iommu_feature_enable(struct amd_iommu *iommu, u8 bit) { u32 ctrl; @@ -167,6 +232,7 @@ static void __init iommu_feature_disable(struct amd_iommu *iommu, u8 bit) writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET); } +/* Function to enable the hardware */ void __init iommu_enable(struct amd_iommu *iommu) { printk(KERN_INFO "AMD IOMMU: Enabling IOMMU at "); @@ -176,6 +242,10 @@ void __init iommu_enable(struct amd_iommu *iommu) iommu_feature_enable(iommu, CONTROL_IOMMU_EN); } +/* + * mapping and unmapping functions for the IOMMU MMIO space. Each AMD IOMMU in + * the system has one. + */ static u8 * __init iommu_map_mmio_space(u64 address) { u8 *ret; @@ -199,6 +269,19 @@ static void __init iommu_unmap_mmio_space(struct amd_iommu *iommu) release_mem_region(iommu->mmio_phys, MMIO_REGION_LENGTH); } +/**************************************************************************** + * + * The functions below belong to the first pass of AMD IOMMU ACPI table + * parsing. In this pass we try to find out the highest device id this + * code has to handle. Upon this information the size of the shared data + * structures is determined later. + * + ****************************************************************************/ + +/* + * This function reads the last device id the IOMMU has to handle from the PCI + * capability header for this IOMMU + */ static int __init find_last_devid_on_pci(int bus, int dev, int fn, int cap_ptr) { u32 cap; @@ -209,6 +292,10 @@ static int __init find_last_devid_on_pci(int bus, int dev, int fn, int cap_ptr) return 0; } +/* + * After reading the highest device id from the IOMMU PCI capability header + * this function looks if there is a higher device id defined in the ACPI table + */ static int __init find_last_devid_from_ivhd(struct ivhd_header *h) { u8 *p = (void *)h, *end = (void *)h; @@ -229,6 +316,7 @@ static int __init find_last_devid_from_ivhd(struct ivhd_header *h) case IVHD_DEV_RANGE_END: case IVHD_DEV_ALIAS: case IVHD_DEV_EXT_SELECT: + /* all the above subfield types refer to device ids */ UPDATE_LAST_BDF(dev->devid); break; default: @@ -242,6 +330,11 @@ static int __init find_last_devid_from_ivhd(struct ivhd_header *h) return 0; } +/* + * Iterate over all IVHD entries in the ACPI table and find the highest device + * id which we need to handle. This is the first of three functions which parse + * the ACPI table. So we check the checksum here. + */ static int __init find_last_devid_acpi(struct acpi_table_header *table) { int i; @@ -277,6 +370,20 @@ static int __init find_last_devid_acpi(struct acpi_table_header *table) return 0; } +/**************************************************************************** + * + * The following functions belong the the code path which parses the ACPI table + * the second time. In this ACPI parsing iteration we allocate IOMMU specific + * data structures, initialize the device/alias/rlookup table and also + * basically initialize the hardware. + * + ****************************************************************************/ + +/* + * Allocates the command buffer. This buffer is per AMD IOMMU. We can + * write commands to that buffer later and the IOMMU will execute them + * asynchronously + */ static u8 * __init alloc_command_buffer(struct amd_iommu *iommu) { u8 *cmd_buf = (u8 *)__get_free_pages(GFP_KERNEL, @@ -307,6 +414,7 @@ static void __init free_command_buffer(struct amd_iommu *iommu) get_order(CMD_BUFFER_SIZE)); } +/* sets a specific bit in the device table entry. */ static void set_dev_entry_bit(u16 devid, u8 bit) { int i = (bit >> 5) & 0x07; @@ -315,6 +423,10 @@ static void set_dev_entry_bit(u16 devid, u8 bit) amd_iommu_dev_table[devid].data[i] |= (1 << _bit); } +/* + * This function takes the device specific flags read from the ACPI + * table and sets up the device table entry with that information + */ static void __init set_dev_entry_from_acpi(u16 devid, u32 flags, u32 ext_flags) { if (flags & ACPI_DEVFLAG_INITPASS) @@ -333,11 +445,16 @@ static void __init set_dev_entry_from_acpi(u16 devid, u32 flags, u32 ext_flags) set_dev_entry_bit(devid, DEV_ENTRY_LINT1_PASS); } +/* Writes the specific IOMMU for a device into the rlookup table */ static void __init set_iommu_for_device(struct amd_iommu *iommu, u16 devid) { amd_iommu_rlookup_table[devid] = iommu; } +/* + * Reads the device exclusion range from ACPI and initialize IOMMU with + * it + */ static void __init set_device_exclusion_range(u16 devid, struct ivmd_header *m) { struct amd_iommu *iommu = amd_iommu_rlookup_table[devid]; @@ -346,12 +463,22 @@ static void __init set_device_exclusion_range(u16 devid, struct ivmd_header *m) return; if (iommu) { + /* + * We only can configure exclusion ranges per IOMMU, not + * per device. But we can enable the exclusion range per + * device. This is done here + */ set_dev_entry_bit(m->devid, DEV_ENTRY_EX); iommu->exclusion_start = m->range_start; iommu->exclusion_length = m->range_length; } } +/* + * This function reads some important data from the IOMMU PCI space and + * initializes the driver data structure with it. It reads the hardware + * capabilities and the first/last device entries + */ static void __init init_iommu_from_pci(struct amd_iommu *iommu) { int bus = PCI_BUS(iommu->devid); @@ -367,6 +494,10 @@ static void __init init_iommu_from_pci(struct amd_iommu *iommu) iommu->last_device = DEVID(MMIO_GET_BUS(range), MMIO_GET_LD(range)); } +/* + * Takes a pointer to an AMD IOMMU entry in the ACPI table and + * initializes the hardware and our data structures with it. + */ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, struct ivhd_header *h) { @@ -467,6 +598,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, } } +/* Initializes the device->iommu mapping for the driver */ static int __init init_iommu_devices(struct amd_iommu *iommu) { u16 i; @@ -494,6 +626,11 @@ static void __init free_iommu_all(void) } } +/* + * This function clues the initialization function for one IOMMU + * together and also allocates the command buffer and programs the + * hardware. It does NOT enable the IOMMU. This is done afterwards. + */ static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h) { spin_lock_init(&iommu->lock); @@ -521,6 +658,10 @@ static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h) return 0; } +/* + * Iterates over all IOMMU entries in the ACPI table, allocates the + * IOMMU structure and initializes it with init_iommu_one() + */ static int __init init_iommu_all(struct acpi_table_header *table) { u8 *p = (u8 *)table, *end = (u8 *)table; @@ -555,6 +696,14 @@ static int __init init_iommu_all(struct acpi_table_header *table) return 0; } +/**************************************************************************** + * + * The next functions belong to the third pass of parsing the ACPI + * table. In this last pass the memory mapping requirements are + * gathered (like exclusion and unity mapping reanges). + * + ****************************************************************************/ + static void __init free_unity_maps(void) { struct unity_map_entry *entry, *next; @@ -565,6 +714,7 @@ static void __init free_unity_maps(void) } } +/* called when we find an exclusion range definition in ACPI */ static int __init init_exclusion_range(struct ivmd_header *m) { int i; @@ -588,6 +738,7 @@ static int __init init_exclusion_range(struct ivmd_header *m) return 0; } +/* called for unity map ACPI definition */ static int __init init_unity_map_range(struct ivmd_header *m) { struct unity_map_entry *e = 0; @@ -619,6 +770,7 @@ static int __init init_unity_map_range(struct ivmd_header *m) return 0; } +/* iterates over all memory definitions we find in the ACPI table */ static int __init init_memory_definitions(struct acpi_table_header *table) { u8 *p = (u8 *)table, *end = (u8 *)table; @@ -642,6 +794,10 @@ static int __init init_memory_definitions(struct acpi_table_header *table) return 0; } +/* + * This function finally enables all IOMMUs found in the system after + * they have been initialized + */ static void __init enable_iommus(void) { struct amd_iommu *iommu; @@ -678,6 +834,34 @@ static struct sys_device device_amd_iommu = { .cls = &amd_iommu_sysdev_class, }; +/* + * This is the core init function for AMD IOMMU hardware in the system. + * This function is called from the generic x86 DMA layer initialization + * code. + * + * This function basically parses the ACPI table for AMD IOMMU (IVRS) + * three times: + * + * 1 pass) Find the highest PCI device id the driver has to handle. + * Upon this information the size of the data structures is + * determined that needs to be allocated. + * + * 2 pass) Initialize the data structures just allocated with the + * information in the ACPI table about available AMD IOMMUs + * in the system. It also maps the PCI devices in the + * system to specific IOMMUs + * + * 3 pass) After the basic data structures are allocated and + * initialized we update them with information about memory + * remapping requirements parsed out of the ACPI table in + * this last pass. + * + * After that the hardware is initialized and ready to go. In the last + * step we do some Linux specific things like registering the driver in + * the dma_ops interface and initializing the suspend/resume support + * functions. Finally it prints some information about AMD IOMMUs and + * the driver state and enables the hardware. + */ int __init amd_iommu_init(void) { int i, ret = 0; @@ -821,6 +1005,13 @@ free: goto out; } +/**************************************************************************** + * + * Early detect code. This code runs at IOMMU detection time in the DMA + * layer. It just looks if there is an IVRS ACPI table to detect AMD + * IOMMUs + * + ****************************************************************************/ static int __init early_amd_iommu_detect(struct acpi_table_header *table) { return 0; @@ -841,6 +1032,13 @@ void __init amd_iommu_detect(void) } } +/**************************************************************************** + * + * Parsing functions for the AMD IOMMU specific kernel command line + * options. + * + ****************************************************************************/ + static int __init parse_amd_iommu_options(char *str) { for (; *str; ++str) { -- cgit v1.2.3 From 431b2a2015337533f1a9e39a840266a8a2c93144 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:22 +0200 Subject: x86, AMD IOMMU: add comments to core code This patch adds comments about how the AMD IOMMU core code works for the DMA remapping functionality. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu.c | 201 +++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 199 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index f2766d84c7a..4bae96ca7c1 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -34,6 +34,9 @@ static DEFINE_RWLOCK(amd_iommu_devtable_lock); +/* + * general struct to manage commands send to an IOMMU + */ struct command { u32 data[4]; }; @@ -41,11 +44,22 @@ struct command { static int dma_ops_unity_map(struct dma_ops_domain *dma_dom, struct unity_map_entry *e); +/* returns !0 if the IOMMU is caching non-present entries in its TLB */ static int iommu_has_npcache(struct amd_iommu *iommu) { return iommu->cap & IOMMU_CAP_NPCACHE; } +/**************************************************************************** + * + * IOMMU command queuing functions + * + ****************************************************************************/ + +/* + * Writes the command to the IOMMUs command buffer and informs the + * hardware about the new command. Must be called with iommu->lock held. + */ static int __iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) { u32 tail, head; @@ -63,6 +77,10 @@ static int __iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) return 0; } +/* + * General queuing function for commands. Takes iommu->lock and calls + * __iommu_queue_command(). + */ static int iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) { unsigned long flags; @@ -75,6 +93,13 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) return ret; } +/* + * This function is called whenever we need to ensure that the IOMMU has + * completed execution of all commands we sent. It sends a + * COMPLETION_WAIT command and waits for it to finish. The IOMMU informs + * us about that by writing a value to a physical address we pass with + * the command. + */ static int iommu_completion_wait(struct amd_iommu *iommu) { int ret; @@ -101,6 +126,9 @@ static int iommu_completion_wait(struct amd_iommu *iommu) return 0; } +/* + * Command send function for invalidating a device table entry + */ static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid) { struct command cmd; @@ -116,6 +144,9 @@ static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid) return iommu_queue_command(iommu, &cmd); } +/* + * Generic command send function for invalidaing TLB entries + */ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu, u64 address, u16 domid, int pde, int s) { @@ -127,9 +158,9 @@ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu, cmd.data[1] |= domid; cmd.data[2] = LOW_U32(address); cmd.data[3] = HIGH_U32(address); - if (s) + if (s) /* size bit - we flush more than one 4kb page */ cmd.data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK; - if (pde) + if (pde) /* PDE bit - we wan't flush everything not only the PTEs */ cmd.data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK; iommu->need_sync = 1; @@ -137,6 +168,11 @@ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu, return iommu_queue_command(iommu, &cmd); } +/* + * TLB invalidation function which is called from the mapping functions. + * It invalidates a single PTE if the range to flush is within a single + * page. Otherwise it flushes the whole TLB of the IOMMU. + */ static int iommu_flush_pages(struct amd_iommu *iommu, u16 domid, u64 address, size_t size) { @@ -159,6 +195,20 @@ static int iommu_flush_pages(struct amd_iommu *iommu, u16 domid, return 0; } +/**************************************************************************** + * + * The functions below are used the create the page table mappings for + * unity mapped regions. + * + ****************************************************************************/ + +/* + * Generic mapping functions. It maps a physical address into a DMA + * address space. It allocates the page table pages if necessary. + * In the future it can be extended to a generic mapping function + * supporting all features of AMD IOMMU page tables like level skipping + * and full 64 bit address spaces. + */ static int iommu_map(struct protection_domain *dom, unsigned long bus_addr, unsigned long phys_addr, @@ -209,6 +259,10 @@ static int iommu_map(struct protection_domain *dom, return 0; } +/* + * This function checks if a specific unity mapping entry is needed for + * this specific IOMMU. + */ static int iommu_for_unity_map(struct amd_iommu *iommu, struct unity_map_entry *entry) { @@ -223,6 +277,12 @@ static int iommu_for_unity_map(struct amd_iommu *iommu, return 0; } +/* + * Init the unity mappings for a specific IOMMU in the system + * + * Basically iterates over all unity mapping entries and applies them to + * the default domain DMA of that IOMMU if necessary. + */ static int iommu_init_unity_mappings(struct amd_iommu *iommu) { struct unity_map_entry *entry; @@ -239,6 +299,10 @@ static int iommu_init_unity_mappings(struct amd_iommu *iommu) return 0; } +/* + * This function actually applies the mapping to the page table of the + * dma_ops domain. + */ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom, struct unity_map_entry *e) { @@ -261,6 +325,9 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom, return 0; } +/* + * Inits the unity mappings required for a specific device + */ static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom, u16 devid) { @@ -278,12 +345,26 @@ static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom, return 0; } +/**************************************************************************** + * + * The next functions belong to the address allocator for the dma_ops + * interface functions. They work like the allocators in the other IOMMU + * drivers. Its basically a bitmap which marks the allocated pages in + * the aperture. Maybe it could be enhanced in the future to a more + * efficient allocator. + * + ****************************************************************************/ static unsigned long dma_mask_to_pages(unsigned long mask) { return (mask >> PAGE_SHIFT) + (PAGE_ALIGN(mask & ~PAGE_MASK) >> PAGE_SHIFT); } +/* + * The address allocator core function. + * + * called with domain->lock held + */ static unsigned long dma_ops_alloc_addresses(struct device *dev, struct dma_ops_domain *dom, unsigned int pages) @@ -317,6 +398,11 @@ static unsigned long dma_ops_alloc_addresses(struct device *dev, return address; } +/* + * The address free function. + * + * called with domain->lock held + */ static void dma_ops_free_addresses(struct dma_ops_domain *dom, unsigned long address, unsigned int pages) @@ -325,6 +411,16 @@ static void dma_ops_free_addresses(struct dma_ops_domain *dom, iommu_area_free(dom->bitmap, address, pages); } +/**************************************************************************** + * + * The next functions belong to the domain allocation. A domain is + * allocated for every IOMMU as the default domain. If device isolation + * is enabled, every device get its own domain. The most important thing + * about domains is the page table mapping the DMA address space they + * contain. + * + ****************************************************************************/ + static u16 domain_id_alloc(void) { unsigned long flags; @@ -342,6 +438,10 @@ static u16 domain_id_alloc(void) return id; } +/* + * Used to reserve address ranges in the aperture (e.g. for exclusion + * ranges. + */ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom, unsigned long start_page, unsigned int pages) @@ -382,6 +482,10 @@ static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom) free_page((unsigned long)p1); } +/* + * Free a domain, only used if something went wrong in the + * allocation path and we need to free an already allocated page table + */ static void dma_ops_domain_free(struct dma_ops_domain *dom) { if (!dom) @@ -396,6 +500,11 @@ static void dma_ops_domain_free(struct dma_ops_domain *dom) kfree(dom); } +/* + * Allocates a new protection domain usable for the dma_ops functions. + * It also intializes the page table and the address allocator data + * structures required for the dma_ops interface + */ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu, unsigned order) { @@ -436,6 +545,7 @@ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu, dma_dom->bitmap[0] = 1; dma_dom->next_bit = 0; + /* Intialize the exclusion range if necessary */ if (iommu->exclusion_start && iommu->exclusion_start < dma_dom->aperture_size) { unsigned long startpage = iommu->exclusion_start >> PAGE_SHIFT; @@ -444,6 +554,11 @@ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu, dma_ops_reserve_addresses(dma_dom, startpage, pages); } + /* + * At the last step, build the page tables so we don't need to + * allocate page table pages in the dma_ops mapping/unmapping + * path. + */ num_pte_pages = dma_dom->aperture_size / (PAGE_SIZE * 512); dma_dom->pte_pages = kzalloc(num_pte_pages * sizeof(void *), GFP_KERNEL); @@ -472,6 +587,10 @@ free_dma_dom: return NULL; } +/* + * Find out the protection domain structure for a given PCI device. This + * will give us the pointer to the page table root for example. + */ static struct protection_domain *domain_for_device(u16 devid) { struct protection_domain *dom; @@ -484,6 +603,10 @@ static struct protection_domain *domain_for_device(u16 devid) return dom; } +/* + * If a device is not yet associated with a domain, this function does + * assigns it visible for the hardware + */ static void set_device_domain(struct amd_iommu *iommu, struct protection_domain *domain, u16 devid) @@ -508,6 +631,19 @@ static void set_device_domain(struct amd_iommu *iommu, iommu->need_sync = 1; } +/***************************************************************************** + * + * The next functions belong to the dma_ops mapping/unmapping code. + * + *****************************************************************************/ + +/* + * In the dma_ops path we only have the struct device. This function + * finds the corresponding IOMMU, the protection domain and the + * requestor id for a given device. + * If the device is not yet associated with a domain this is also done + * in this function. + */ static int get_device_resources(struct device *dev, struct amd_iommu **iommu, struct protection_domain **domain, @@ -522,6 +658,7 @@ static int get_device_resources(struct device *dev, pcidev = to_pci_dev(dev); _bdf = (pcidev->bus->number << 8) | pcidev->devfn; + /* device not translated by any IOMMU in the system? */ if (_bdf >= amd_iommu_last_bdf) { *iommu = NULL; *domain = NULL; @@ -547,6 +684,10 @@ static int get_device_resources(struct device *dev, return 1; } +/* + * This is the generic map function. It maps one 4kb page at paddr to + * the given address in the DMA address space for the domain. + */ static dma_addr_t dma_ops_domain_map(struct amd_iommu *iommu, struct dma_ops_domain *dom, unsigned long address, @@ -578,6 +719,9 @@ static dma_addr_t dma_ops_domain_map(struct amd_iommu *iommu, return (dma_addr_t)address; } +/* + * The generic unmapping function for on page in the DMA address space. + */ static void dma_ops_domain_unmap(struct amd_iommu *iommu, struct dma_ops_domain *dom, unsigned long address) @@ -597,6 +741,12 @@ static void dma_ops_domain_unmap(struct amd_iommu *iommu, *pte = 0ULL; } +/* + * This function contains common code for mapping of a physically + * contiguous memory region into DMA address space. It is uses by all + * mapping functions provided by this IOMMU driver. + * Must be called with the domain lock held. + */ static dma_addr_t __map_single(struct device *dev, struct amd_iommu *iommu, struct dma_ops_domain *dma_dom, @@ -628,6 +778,10 @@ out: return address; } +/* + * Does the reverse of the __map_single function. Must be called with + * the domain lock held too + */ static void __unmap_single(struct amd_iommu *iommu, struct dma_ops_domain *dma_dom, dma_addr_t dma_addr, @@ -652,6 +806,9 @@ static void __unmap_single(struct amd_iommu *iommu, dma_ops_free_addresses(dma_dom, dma_addr, pages); } +/* + * The exported map_single function for dma_ops. + */ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr, size_t size, int dir) { @@ -664,6 +821,7 @@ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr, get_device_resources(dev, &iommu, &domain, &devid); if (iommu == NULL || domain == NULL) + /* device not handled by any AMD IOMMU */ return (dma_addr_t)paddr; spin_lock_irqsave(&domain->lock, flags); @@ -683,6 +841,9 @@ out: return addr; } +/* + * The exported unmap_single function for dma_ops. + */ static void unmap_single(struct device *dev, dma_addr_t dma_addr, size_t size, int dir) { @@ -692,6 +853,7 @@ static void unmap_single(struct device *dev, dma_addr_t dma_addr, u16 devid; if (!get_device_resources(dev, &iommu, &domain, &devid)) + /* device not handled by any AMD IOMMU */ return; spin_lock_irqsave(&domain->lock, flags); @@ -706,6 +868,10 @@ static void unmap_single(struct device *dev, dma_addr_t dma_addr, spin_unlock_irqrestore(&domain->lock, flags); } +/* + * This is a special map_sg function which is used if we should map a + * device which is not handled by an AMD IOMMU in the system. + */ static int map_sg_no_iommu(struct device *dev, struct scatterlist *sglist, int nelems, int dir) { @@ -720,6 +886,10 @@ static int map_sg_no_iommu(struct device *dev, struct scatterlist *sglist, return nelems; } +/* + * The exported map_sg function for dma_ops (handles scatter-gather + * lists). + */ static int map_sg(struct device *dev, struct scatterlist *sglist, int nelems, int dir) { @@ -775,6 +945,10 @@ unmap: goto out; } +/* + * The exported map_sg function for dma_ops (handles scatter-gather + * lists). + */ static void unmap_sg(struct device *dev, struct scatterlist *sglist, int nelems, int dir) { @@ -804,6 +978,9 @@ static void unmap_sg(struct device *dev, struct scatterlist *sglist, spin_unlock_irqrestore(&domain->lock, flags); } +/* + * The exported alloc_coherent function for dma_ops. + */ static void *alloc_coherent(struct device *dev, size_t size, dma_addr_t *dma_addr, gfp_t flag) { @@ -851,6 +1028,11 @@ out: return virt_addr; } +/* + * The exported free_coherent function for dma_ops. + * FIXME: fix the generic x86 DMA layer so that it actually calls that + * function. + */ static void free_coherent(struct device *dev, size_t size, void *virt_addr, dma_addr_t dma_addr) { @@ -879,6 +1061,8 @@ free_mem: } /* + * The function for pre-allocating protection domains. + * * If the driver core informs the DMA layer if a driver grabs a device * we don't need to preallocate the protection domains anymore. * For now we have to. @@ -921,12 +1105,20 @@ static struct dma_mapping_ops amd_iommu_dma_ops = { .unmap_sg = unmap_sg, }; +/* + * The function which clues the AMD IOMMU driver into dma_ops. + */ int __init amd_iommu_init_dma_ops(void) { struct amd_iommu *iommu; int order = amd_iommu_aperture_order; int ret; + /* + * first allocate a default protection domain for every IOMMU we + * found in the system. Devices not assigned to any other + * protection domain will be assigned to the default one. + */ list_for_each_entry(iommu, &amd_iommu_list, list) { iommu->default_dom = dma_ops_domain_alloc(iommu, order); if (iommu->default_dom == NULL) @@ -936,6 +1128,10 @@ int __init amd_iommu_init_dma_ops(void) goto free_domains; } + /* + * If device isolation is enabled, pre-allocate the protection + * domains for each device. + */ if (amd_iommu_isolate) prealloc_protection_domains(); @@ -947,6 +1143,7 @@ int __init amd_iommu_init_dma_ops(void) gart_iommu_aperture = 0; #endif + /* Make the driver finally visible to the drivers */ dma_ops = &amd_iommu_dma_ops; return 0; -- cgit v1.2.3 From 8ea80d783efd0c50577ec8d69757ae54c408eacd Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:23 +0200 Subject: x86, AMD IOMMU: replace HIGH_U32 macro with upper_32_bits function Removes a driver specific macro and replaces it with a generic function already available in Linux. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index 4bae96ca7c1..9098f047c1a 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -109,7 +109,7 @@ static int iommu_completion_wait(struct amd_iommu *iommu) memset(&cmd, 0, sizeof(cmd)); cmd.data[0] = LOW_U32(ready_phys) | CMD_COMPL_WAIT_STORE_MASK; - cmd.data[1] = HIGH_U32(ready_phys); + cmd.data[1] = upper_32_bits(ready_phys); cmd.data[2] = 1; /* value written to 'ready' */ CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT); @@ -157,7 +157,7 @@ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu, CMD_SET_TYPE(&cmd, CMD_INV_IOMMU_PAGES); cmd.data[1] |= domid; cmd.data[2] = LOW_U32(address); - cmd.data[3] = HIGH_U32(address); + cmd.data[3] = upper_32_bits(address); if (s) /* size bit - we flush more than one 4kb page */ cmd.data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK; if (pde) /* PDE bit - we wan't flush everything not only the PTEs */ -- cgit v1.2.3 From 208ec8c94d818a3def0b424958493728871716d1 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:24 +0200 Subject: x86, AMD IOMMU: replace UPDATE_LAST_BDF macro with a function This patch replaces the UPDATE_LAST_BDF macro in the init code with the update_last_devid function. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 9ddb46d7c52..6e1c8ffc0c5 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -30,11 +30,6 @@ /* * definitions for the ACPI scanning code */ -#define UPDATE_LAST_BDF(x) do {\ - if ((x) > amd_iommu_last_bdf) \ - amd_iommu_last_bdf = (x); \ - } while (0); - #define DEVID(bus, devfn) (((bus) << 8) | (devfn)) #define PCI_BUS(x) (((x) >> 8) & 0xff) #define IVRS_HEADER_LENGTH 48 @@ -169,6 +164,12 @@ static u32 dev_table_size; /* size of the device table */ static u32 alias_table_size; /* size of the alias table */ static u32 rlookup_table_size; /* size if the rlookup table */ +static inline void update_last_devid(u16 devid) +{ + if (devid > amd_iommu_last_bdf) + amd_iommu_last_bdf = devid; +} + /**************************************************************************** * * AMD IOMMU MMIO register space handling functions @@ -287,7 +288,7 @@ static int __init find_last_devid_on_pci(int bus, int dev, int fn, int cap_ptr) u32 cap; cap = read_pci_config(bus, dev, fn, cap_ptr+MMIO_RANGE_OFFSET); - UPDATE_LAST_BDF(DEVID(MMIO_GET_BUS(cap), MMIO_GET_LD(cap))); + update_last_devid(DEVID(MMIO_GET_BUS(cap), MMIO_GET_LD(cap))); return 0; } @@ -317,7 +318,7 @@ static int __init find_last_devid_from_ivhd(struct ivhd_header *h) case IVHD_DEV_ALIAS: case IVHD_DEV_EXT_SELECT: /* all the above subfield types refer to device ids */ - UPDATE_LAST_BDF(dev->devid); + update_last_devid(dev->devid); break; default: break; -- cgit v1.2.3 From c571484e53f3e1d90bc5374528580c7419d28d4c Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:25 +0200 Subject: x86, AMD IOMMU: replace TBL_SIZE macro with a function This patch converts the TBL_SIZE macro in the init code to a function. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 6e1c8ffc0c5..1f148393cf7 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -33,7 +33,6 @@ #define DEVID(bus, devfn) (((bus) << 8) | (devfn)) #define PCI_BUS(x) (((x) >> 8) & 0xff) #define IVRS_HEADER_LENGTH 48 -#define TBL_SIZE(x) (1 << (PAGE_SHIFT + get_order(amd_iommu_last_bdf * (x)))) #define ACPI_IVHD_TYPE 0x10 #define ACPI_IVMD_TYPE_ALL 0x20 @@ -170,6 +169,14 @@ static inline void update_last_devid(u16 devid) amd_iommu_last_bdf = devid; } +static inline unsigned long tbl_size(int entry_size) +{ + unsigned shift = PAGE_SHIFT + + get_order(amd_iommu_last_bdf * entry_size); + + return 1UL << shift; +} + /**************************************************************************** * * AMD IOMMU MMIO register space handling functions @@ -884,9 +891,9 @@ int __init amd_iommu_init(void) if (acpi_table_parse("IVRS", find_last_devid_acpi) != 0) return -ENODEV; - dev_table_size = TBL_SIZE(DEV_TABLE_ENTRY_SIZE); - alias_table_size = TBL_SIZE(ALIAS_TABLE_ENTRY_SIZE); - rlookup_table_size = TBL_SIZE(RLOOKUP_TABLE_ENTRY_SIZE); + dev_table_size = tbl_size(DEV_TABLE_ENTRY_SIZE); + alias_table_size = tbl_size(ALIAS_TABLE_ENTRY_SIZE); + rlookup_table_size = tbl_size(RLOOKUP_TABLE_ENTRY_SIZE); ret = -ENOMEM; -- cgit v1.2.3 From 9a836de0c9944c42d006ec241712c72e74737c73 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:26 +0200 Subject: x86, AMD IOMMU: remove unnecessary free checks from init code This patch removes unnecessary checks before memory is released. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 27 ++++++++++----------------- 1 file changed, 10 insertions(+), 17 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 1f148393cf7..0f5a9115a69 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -417,9 +417,7 @@ static u8 * __init alloc_command_buffer(struct amd_iommu *iommu) static void __init free_command_buffer(struct amd_iommu *iommu) { - if (iommu->cmd_buf) - free_pages((unsigned long)iommu->cmd_buf, - get_order(CMD_BUFFER_SIZE)); + free_pages((unsigned long)iommu->cmd_buf, get_order(CMD_BUFFER_SIZE)); } /* sets a specific bit in the device table entry. */ @@ -987,24 +985,19 @@ out: return ret; free: - if (amd_iommu_pd_alloc_bitmap) - free_pages((unsigned long)amd_iommu_pd_alloc_bitmap, 1); + free_pages((unsigned long)amd_iommu_pd_alloc_bitmap, 1); - if (amd_iommu_pd_table) - free_pages((unsigned long)amd_iommu_pd_table, - get_order(rlookup_table_size)); + free_pages((unsigned long)amd_iommu_pd_table, + get_order(rlookup_table_size)); - if (amd_iommu_rlookup_table) - free_pages((unsigned long)amd_iommu_rlookup_table, - get_order(rlookup_table_size)); + free_pages((unsigned long)amd_iommu_rlookup_table, + get_order(rlookup_table_size)); - if (amd_iommu_alias_table) - free_pages((unsigned long)amd_iommu_alias_table, - get_order(alias_table_size)); + free_pages((unsigned long)amd_iommu_alias_table, + get_order(alias_table_size)); - if (amd_iommu_dev_table) - free_pages((unsigned long)amd_iommu_dev_table, - get_order(dev_table_size)); + free_pages((unsigned long)amd_iommu_dev_table, + get_order(dev_table_size)); free_iommu_all(); -- cgit v1.2.3 From 136f78a19cf94d469f31a4009c7c0ac2301fbbf0 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:27 +0200 Subject: x86, AMD IOMMU: add an emergency exit to the completion wait loop To make the loop waiting for the completion wait command not wait forever this patch adds a limit of cycles that loop. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index 9098f047c1a..7fa2d5d57dd 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -32,6 +32,8 @@ #define to_pages(addr, size) \ (round_up(((addr) & ~PAGE_MASK) + (size), PAGE_SIZE) >> PAGE_SHIFT) +#define EXIT_LOOP_COUNT 10000000 + static DEFINE_RWLOCK(amd_iommu_devtable_lock); /* @@ -106,6 +108,7 @@ static int iommu_completion_wait(struct amd_iommu *iommu) struct command cmd; volatile u64 ready = 0; unsigned long ready_phys = virt_to_phys(&ready); + unsigned long i = 0; memset(&cmd, 0, sizeof(cmd)); cmd.data[0] = LOW_U32(ready_phys) | CMD_COMPL_WAIT_STORE_MASK; @@ -120,8 +123,13 @@ static int iommu_completion_wait(struct amd_iommu *iommu) if (ret) return ret; - while (!ready) + while (!ready && (i < EXIT_LOOP_COUNT)) { + ++i; cpu_relax(); + } + + if (unlikely((i == EXIT_LOOP_COUNT) && printk_ratelimit())) + printk(KERN_WARNING "AMD IOMMU: Completion wait loop failed\n"); return 0; } -- cgit v1.2.3 From d64495366ff78fdbd5bd3176a7ada2f0c2cbfba6 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:28 +0200 Subject: x86, AMD IOMMU: rename struct command to iommu_cmd This patch gives the struct command a more descriptive and not so generic name. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index 7fa2d5d57dd..dec10e1a397 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -39,7 +39,7 @@ static DEFINE_RWLOCK(amd_iommu_devtable_lock); /* * general struct to manage commands send to an IOMMU */ -struct command { +struct iommu_cmd { u32 data[4]; }; @@ -62,7 +62,7 @@ static int iommu_has_npcache(struct amd_iommu *iommu) * Writes the command to the IOMMUs command buffer and informs the * hardware about the new command. Must be called with iommu->lock held. */ -static int __iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) +static int __iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd) { u32 tail, head; u8 *target; @@ -83,7 +83,7 @@ static int __iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) * General queuing function for commands. Takes iommu->lock and calls * __iommu_queue_command(). */ -static int iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) +static int iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd) { unsigned long flags; int ret; @@ -105,7 +105,7 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct command *cmd) static int iommu_completion_wait(struct amd_iommu *iommu) { int ret; - struct command cmd; + struct iommu_cmd cmd; volatile u64 ready = 0; unsigned long ready_phys = virt_to_phys(&ready); unsigned long i = 0; @@ -139,7 +139,7 @@ static int iommu_completion_wait(struct amd_iommu *iommu) */ static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid) { - struct command cmd; + struct iommu_cmd cmd; BUG_ON(iommu == NULL); @@ -158,7 +158,7 @@ static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid) static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu, u64 address, u16 domid, int pde, int s) { - struct command cmd; + struct iommu_cmd cmd; memset(&cmd, 0, sizeof(cmd)); address &= PAGE_MASK; -- cgit v1.2.3 From d0312b2142ac7665031755c1cc3dba827d4eb711 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:29 +0200 Subject: x86, AMD IOMMU: remove unneeded initializations from command buffer allocation This patch removes an unneeded initialization from the alloc_command_buffer function and replaces a memset with __GFP_ZERO. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 0f5a9115a69..0124995c7b5 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -394,17 +394,15 @@ static int __init find_last_devid_acpi(struct acpi_table_header *table) */ static u8 * __init alloc_command_buffer(struct amd_iommu *iommu) { - u8 *cmd_buf = (u8 *)__get_free_pages(GFP_KERNEL, + u8 *cmd_buf = (u8 *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(CMD_BUFFER_SIZE)); - u64 entry = 0; + u64 entry; if (cmd_buf == NULL) return NULL; iommu->cmd_buf_size = CMD_BUFFER_SIZE; - memset(cmd_buf, 0, CMD_BUFFER_SIZE); - entry = (u64)virt_to_phys(cmd_buf); entry |= MMIO_CMD_SIZE_512; memcpy_toio(iommu->mmio_base + MMIO_CMD_BUF_OFFSET, -- cgit v1.2.3 From 58a3bee567b588a84cdde05fecc45439b396362c Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:30 +0200 Subject: x86, AMD IOMMU: use true/false instead of 0/1 for bool value This patch replaces the integer values used for the bool variable in ACPI scanning code with true and false. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 0124995c7b5..316fe2eaeef 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -509,7 +509,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, u8 *end = p, flags = 0; u16 dev_i, devid = 0, devid_start = 0, devid_to = 0; u32 ext_flags = 0; - bool alias = 0; + bool alias = false; struct ivhd_entry *e; /* @@ -559,7 +559,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, devid_start = e->devid; flags = e->flags; ext_flags = 0; - alias = 0; + alias = false; break; case IVHD_DEV_ALIAS: devid = e->devid; @@ -572,7 +572,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, flags = e->flags; devid_to = e->ext >> 8; ext_flags = 0; - alias = 1; + alias = true; break; case IVHD_DEV_EXT_SELECT: devid = e->devid; @@ -582,7 +582,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, devid_start = e->devid; flags = e->flags; ext_flags = e->ext; - alias = 0; + alias = false; break; case IVHD_DEV_RANGE_END: devid = e->devid; -- cgit v1.2.3 From 2e22847fbe05f2543ccebd0c2df94d9cf3c52aa5 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:31 +0200 Subject: x86, AMD IOMMU: do runtime list initialization at compile time This patch changes the list initialization for the iommu list and the unity map list from runtime to compile time. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 316fe2eaeef..0c247032308 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -118,12 +118,12 @@ static int __initdata amd_iommu_detected; u16 amd_iommu_last_bdf; /* largest PCI device id we have to handle */ -struct list_head amd_iommu_unity_map; /* a list of required unity mappings +LIST_HEAD(amd_iommu_unity_map); /* a list of required unity mappings we find in ACPI */ unsigned amd_iommu_aperture_order = 26; /* size of aperture in power of 2 */ int amd_iommu_isolate; /* if 1, device isolation is enabled */ -struct list_head amd_iommu_list; /* list of all AMD IOMMUs in the +LIST_HEAD(amd_iommu_list); /* list of all AMD IOMMUs in the system */ /* @@ -673,8 +673,6 @@ static int __init init_iommu_all(struct acpi_table_header *table) struct amd_iommu *iommu; int ret; - INIT_LIST_HEAD(&amd_iommu_list); - end += table->length; p += IVRS_HEADER_LENGTH; @@ -780,8 +778,6 @@ static int __init init_memory_definitions(struct acpi_table_header *table) u8 *p = (u8 *)table, *end = (u8 *)table; struct ivmd_header *m; - INIT_LIST_HEAD(&amd_iommu_unity_map); - end += table->length; p += IVRS_HEADER_LENGTH; -- cgit v1.2.3 From 5dc8bff0f6d0dfeb1f1c6e694294ba7c33d099f1 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:32 +0200 Subject: x86, AMD IOMMU: replace memset with __GFP_ZERO for table allocation This patch removes the memset from the data structure initialization code and allocate the structures with the __GFP_ZERO flag. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 14 +++++--------- 1 file changed, 5 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 0c247032308..2efc3d59b7e 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -890,7 +890,7 @@ int __init amd_iommu_init(void) ret = -ENOMEM; /* Device table - directly used by all IOMMUs */ - amd_iommu_dev_table = (void *)__get_free_pages(GFP_KERNEL, + amd_iommu_dev_table = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(dev_table_size)); if (amd_iommu_dev_table == NULL) goto out; @@ -914,27 +914,23 @@ int __init amd_iommu_init(void) * Protection Domain table - maps devices to protection domains * This table has the same size as the rlookup_table */ - amd_iommu_pd_table = (void *)__get_free_pages(GFP_KERNEL, + amd_iommu_pd_table = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(rlookup_table_size)); if (amd_iommu_pd_table == NULL) goto free; - amd_iommu_pd_alloc_bitmap = (void *)__get_free_pages(GFP_KERNEL, + amd_iommu_pd_alloc_bitmap = (void *)__get_free_pages( + GFP_KERNEL | __GFP_ZERO, get_order(MAX_DOMAIN_ID/8)); if (amd_iommu_pd_alloc_bitmap == NULL) goto free; /* - * memory is allocated now; initialize the device table with all zeroes - * and let all alias entries point to itself + * let all alias entries point to itself */ - memset(amd_iommu_dev_table, 0, dev_table_size); for (i = 0; i < amd_iommu_last_bdf; ++i) amd_iommu_alias_table[i] = i; - memset(amd_iommu_pd_table, 0, rlookup_table_size); - memset(amd_iommu_pd_alloc_bitmap, 0, MAX_DOMAIN_ID / 8); - /* * never allocate domain 0 because its used as the non-allocated and * error value placeholder -- cgit v1.2.3 From 0906372e6cf372f3162481f24a0b8ccae0eff4d7 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:33 +0200 Subject: x86, AMD IOMMU: replace self made size parsing with memparse call This patch replaces the self-made parsing of the amd_iommu_size option with the generic memparse function call. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 18 ++++-------------- 1 file changed, 4 insertions(+), 14 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 2efc3d59b7e..e0ff9404e6c 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -1042,20 +1042,10 @@ static int __init parse_amd_iommu_options(char *str) static int __init parse_amd_iommu_size_options(char *str) { - for (; *str; ++str) { - if (strcmp(str, "32M") == 0) - amd_iommu_aperture_order = 25; - if (strcmp(str, "64M") == 0) - amd_iommu_aperture_order = 26; - if (strcmp(str, "128M") == 0) - amd_iommu_aperture_order = 27; - if (strcmp(str, "256M") == 0) - amd_iommu_aperture_order = 28; - if (strcmp(str, "512M") == 0) - amd_iommu_aperture_order = 29; - if (strcmp(str, "1G") == 0) - amd_iommu_aperture_order = 30; - } + unsigned order = PAGE_SHIFT + get_order(memparse(str, &str)); + + if ((order > 24) && (order < 31)) + amd_iommu_aperture_order = order; return 1; } -- cgit v1.2.3 From d591b0a3ae25f587d0c4da1e1d1a425143590790 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Fri, 11 Jul 2008 17:14:35 +0200 Subject: x86, AMD IOMMU: replace DEVID macro with a function This patch replaces the DEVID macro with a function and uses them where apropriate (also in the core code). Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu.c | 2 +- arch/x86/kernel/amd_iommu_init.c | 9 +++++---- 2 files changed, 6 insertions(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index dec10e1a397..8c3deb027d3 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -664,7 +664,7 @@ static int get_device_resources(struct device *dev, BUG_ON(!dev || dev->bus != &pci_bus_type || !dev->dma_mask); pcidev = to_pci_dev(dev); - _bdf = (pcidev->bus->number << 8) | pcidev->devfn; + _bdf = calc_devid(pcidev->bus->number, pcidev->devfn); /* device not translated by any IOMMU in the system? */ if (_bdf >= amd_iommu_last_bdf) { diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index e0ff9404e6c..9bf1b8111b0 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -30,7 +30,6 @@ /* * definitions for the ACPI scanning code */ -#define DEVID(bus, devfn) (((bus) << 8) | (devfn)) #define PCI_BUS(x) (((x) >> 8) & 0xff) #define IVRS_HEADER_LENGTH 48 @@ -295,7 +294,7 @@ static int __init find_last_devid_on_pci(int bus, int dev, int fn, int cap_ptr) u32 cap; cap = read_pci_config(bus, dev, fn, cap_ptr+MMIO_RANGE_OFFSET); - update_last_devid(DEVID(MMIO_GET_BUS(cap), MMIO_GET_LD(cap))); + update_last_devid(calc_devid(MMIO_GET_BUS(cap), MMIO_GET_LD(cap))); return 0; } @@ -494,8 +493,10 @@ static void __init init_iommu_from_pci(struct amd_iommu *iommu) iommu->cap = read_pci_config(bus, dev, fn, cap_ptr+MMIO_CAP_HDR_OFFSET); range = read_pci_config(bus, dev, fn, cap_ptr+MMIO_RANGE_OFFSET); - iommu->first_device = DEVID(MMIO_GET_BUS(range), MMIO_GET_FD(range)); - iommu->last_device = DEVID(MMIO_GET_BUS(range), MMIO_GET_LD(range)); + iommu->first_device = calc_devid(MMIO_GET_BUS(range), + MMIO_GET_FD(range)); + iommu->last_device = calc_devid(MMIO_GET_BUS(range), + MMIO_GET_LD(range)); } /* -- cgit v1.2.3 From 2510495e208e7a69b64fcf5cdf8966d873536d9e Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 11 Jul 2008 12:13:59 +0200 Subject: x86/pci: Removing pci-y in Makefile Cc: Sam Ravnborg Signed-off-by: Robert Richter Cc: Robert Richter Cc: Sam Ravnborg Signed-off-by: Ingo Molnar --- arch/x86/pci/Makefile | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/pci/Makefile b/arch/x86/pci/Makefile index e515e8db842..28451f41e0e 100644 --- a/arch/x86/pci/Makefile +++ b/arch/x86/pci/Makefile @@ -5,13 +5,13 @@ obj-$(CONFIG_PCI_MMCONFIG) += mmconfig_$(BITS).o direct.o mmconfig-shared.o obj-$(CONFIG_PCI_DIRECT) += direct.o obj-$(CONFIG_PCI_OLPC) += olpc.o -pci-y := fixup.o -pci-$(CONFIG_ACPI) += acpi.o -pci-y += legacy.o irq.o +obj-y += fixup.o +obj-$(CONFIG_ACPI) += acpi.o +obj-y += legacy.o irq.o -pci-$(CONFIG_X86_VISWS) += visws.o +obj-$(CONFIG_X86_VISWS) += visws.o -pci-$(CONFIG_X86_NUMAQ) += numa.o +obj-$(CONFIG_X86_NUMAQ) += numa.o -obj-y += $(pci-y) common.o early.o +obj-y += common.o early.o obj-y += amd_bus.o -- cgit v1.2.3 From 060b9708a0c04cf9af69c128ef7954b6f0a84180 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 11 Jul 2008 12:14:27 +0200 Subject: x86/pci: Changing subsystem initialization order for NUMA Cc: Yinghai Lu Signed-off-by: Robert Richter Cc: Robert Richter Cc: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/pci/legacy.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/pci/legacy.c b/arch/x86/pci/legacy.c index 132876cc6fc..60e8caa1356 100644 --- a/arch/x86/pci/legacy.c +++ b/arch/x86/pci/legacy.c @@ -57,14 +57,14 @@ static int __init pci_legacy_init(void) int __init pci_subsys_init(void) { +#ifdef CONFIG_X86_NUMAQ + pci_numa_init(); +#endif #ifdef CONFIG_ACPI pci_acpi_init(); #endif pci_legacy_init(); pcibios_irq_init(); -#ifdef CONFIG_X86_NUMAQ - pci_numa_init(); -#endif pcibios_init(); return 0; -- cgit v1.2.3 From 9314d301390ad0d96986da3d893a21e81a287982 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 11 Jul 2008 12:18:40 +0200 Subject: x86/pci: renamed: numa.c -> numaq_32.c Cc: Yinghai Lu Signed-off-by: Robert Richter Cc: Robert Richter Cc: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/pci/Makefile | 2 +- arch/x86/pci/numa.c | 178 ------------------------------------------------ arch/x86/pci/numaq_32.c | 178 ++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 179 insertions(+), 179 deletions(-) delete mode 100644 arch/x86/pci/numa.c create mode 100644 arch/x86/pci/numaq_32.c (limited to 'arch') diff --git a/arch/x86/pci/Makefile b/arch/x86/pci/Makefile index 28451f41e0e..d49202e740e 100644 --- a/arch/x86/pci/Makefile +++ b/arch/x86/pci/Makefile @@ -11,7 +11,7 @@ obj-y += legacy.o irq.o obj-$(CONFIG_X86_VISWS) += visws.o -obj-$(CONFIG_X86_NUMAQ) += numa.o +obj-$(CONFIG_X86_NUMAQ) += numaq_32.o obj-y += common.o early.o obj-y += amd_bus.o diff --git a/arch/x86/pci/numa.c b/arch/x86/pci/numa.c deleted file mode 100644 index 8b5ca196673..00000000000 --- a/arch/x86/pci/numa.c +++ /dev/null @@ -1,178 +0,0 @@ -/* - * numa.c - Low-level PCI access for NUMA-Q machines - */ - -#include -#include -#include -#include -#include -#include "pci.h" - -#define XQUAD_PORTIO_BASE 0xfe400000 -#define XQUAD_PORTIO_QUAD 0x40000 /* 256k per quad. */ - -#define BUS2QUAD(global) (mp_bus_id_to_node[global]) - -#define BUS2LOCAL(global) (mp_bus_id_to_local[global]) - -#define QUADLOCAL2BUS(quad,local) (quad_local_to_mp_bus_id[quad][local]) - -/* Where the IO area was mapped on multiquad, always 0 otherwise */ -void *xquad_portio; -EXPORT_SYMBOL(xquad_portio); - -#define XQUAD_PORT_ADDR(port, quad) (xquad_portio + (XQUAD_PORTIO_QUAD*quad) + port) - -#define PCI_CONF1_MQ_ADDRESS(bus, devfn, reg) \ - (0x80000000 | (BUS2LOCAL(bus) << 16) | (devfn << 8) | (reg & ~3)) - -static void write_cf8(unsigned bus, unsigned devfn, unsigned reg) -{ - unsigned val = PCI_CONF1_MQ_ADDRESS(bus, devfn, reg); - if (xquad_portio) - writel(val, XQUAD_PORT_ADDR(0xcf8, BUS2QUAD(bus))); - else - outl(val, 0xCF8); -} - -static int pci_conf1_mq_read(unsigned int seg, unsigned int bus, - unsigned int devfn, int reg, int len, u32 *value) -{ - unsigned long flags; - void *adr __iomem = XQUAD_PORT_ADDR(0xcfc, BUS2QUAD(bus)); - - if (!value || (bus >= MAX_MP_BUSSES) || (devfn > 255) || (reg > 255)) - return -EINVAL; - - spin_lock_irqsave(&pci_config_lock, flags); - - write_cf8(bus, devfn, reg); - - switch (len) { - case 1: - if (xquad_portio) - *value = readb(adr + (reg & 3)); - else - *value = inb(0xCFC + (reg & 3)); - break; - case 2: - if (xquad_portio) - *value = readw(adr + (reg & 2)); - else - *value = inw(0xCFC + (reg & 2)); - break; - case 4: - if (xquad_portio) - *value = readl(adr); - else - *value = inl(0xCFC); - break; - } - - spin_unlock_irqrestore(&pci_config_lock, flags); - - return 0; -} - -static int pci_conf1_mq_write(unsigned int seg, unsigned int bus, - unsigned int devfn, int reg, int len, u32 value) -{ - unsigned long flags; - void *adr __iomem = XQUAD_PORT_ADDR(0xcfc, BUS2QUAD(bus)); - - if ((bus >= MAX_MP_BUSSES) || (devfn > 255) || (reg > 255)) - return -EINVAL; - - spin_lock_irqsave(&pci_config_lock, flags); - - write_cf8(bus, devfn, reg); - - switch (len) { - case 1: - if (xquad_portio) - writeb(value, adr + (reg & 3)); - else - outb((u8)value, 0xCFC + (reg & 3)); - break; - case 2: - if (xquad_portio) - writew(value, adr + (reg & 2)); - else - outw((u16)value, 0xCFC + (reg & 2)); - break; - case 4: - if (xquad_portio) - writel(value, adr + reg); - else - outl((u32)value, 0xCFC); - break; - } - - spin_unlock_irqrestore(&pci_config_lock, flags); - - return 0; -} - -#undef PCI_CONF1_MQ_ADDRESS - -static struct pci_raw_ops pci_direct_conf1_mq = { - .read = pci_conf1_mq_read, - .write = pci_conf1_mq_write -}; - - -static void __devinit pci_fixup_i450nx(struct pci_dev *d) -{ - /* - * i450NX -- Find and scan all secondary buses on all PXB's. - */ - int pxb, reg; - u8 busno, suba, subb; - int quad = BUS2QUAD(d->bus->number); - - printk("PCI: Searching for i450NX host bridges on %s\n", pci_name(d)); - reg = 0xd0; - for(pxb=0; pxb<2; pxb++) { - pci_read_config_byte(d, reg++, &busno); - pci_read_config_byte(d, reg++, &suba); - pci_read_config_byte(d, reg++, &subb); - DBG("i450NX PXB %d: %02x/%02x/%02x\n", pxb, busno, suba, subb); - if (busno) { - /* Bus A */ - pci_scan_bus_with_sysdata(QUADLOCAL2BUS(quad, busno)); - } - if (suba < subb) { - /* Bus B */ - pci_scan_bus_with_sysdata(QUADLOCAL2BUS(quad, suba+1)); - } - } - pcibios_last_bus = -1; -} -DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82451NX, pci_fixup_i450nx); - -int __init pci_numa_init(void) -{ - int quad; - - if (!found_numaq) - return 0; - - raw_pci_ops = &pci_direct_conf1_mq; - - if (pcibios_scanned++) - return 0; - - pci_root_bus = pcibios_scan_root(0); - if (pci_root_bus) - pci_bus_add_devices(pci_root_bus); - if (num_online_nodes() > 1) - for_each_online_node(quad) { - if (quad == 0) - continue; - printk("Scanning PCI bus %d for quad %d\n", - QUADLOCAL2BUS(quad,0), quad); - pci_scan_bus_with_sysdata(QUADLOCAL2BUS(quad, 0)); - } - return 0; -} diff --git a/arch/x86/pci/numaq_32.c b/arch/x86/pci/numaq_32.c new file mode 100644 index 00000000000..8b5ca196673 --- /dev/null +++ b/arch/x86/pci/numaq_32.c @@ -0,0 +1,178 @@ +/* + * numa.c - Low-level PCI access for NUMA-Q machines + */ + +#include +#include +#include +#include +#include +#include "pci.h" + +#define XQUAD_PORTIO_BASE 0xfe400000 +#define XQUAD_PORTIO_QUAD 0x40000 /* 256k per quad. */ + +#define BUS2QUAD(global) (mp_bus_id_to_node[global]) + +#define BUS2LOCAL(global) (mp_bus_id_to_local[global]) + +#define QUADLOCAL2BUS(quad,local) (quad_local_to_mp_bus_id[quad][local]) + +/* Where the IO area was mapped on multiquad, always 0 otherwise */ +void *xquad_portio; +EXPORT_SYMBOL(xquad_portio); + +#define XQUAD_PORT_ADDR(port, quad) (xquad_portio + (XQUAD_PORTIO_QUAD*quad) + port) + +#define PCI_CONF1_MQ_ADDRESS(bus, devfn, reg) \ + (0x80000000 | (BUS2LOCAL(bus) << 16) | (devfn << 8) | (reg & ~3)) + +static void write_cf8(unsigned bus, unsigned devfn, unsigned reg) +{ + unsigned val = PCI_CONF1_MQ_ADDRESS(bus, devfn, reg); + if (xquad_portio) + writel(val, XQUAD_PORT_ADDR(0xcf8, BUS2QUAD(bus))); + else + outl(val, 0xCF8); +} + +static int pci_conf1_mq_read(unsigned int seg, unsigned int bus, + unsigned int devfn, int reg, int len, u32 *value) +{ + unsigned long flags; + void *adr __iomem = XQUAD_PORT_ADDR(0xcfc, BUS2QUAD(bus)); + + if (!value || (bus >= MAX_MP_BUSSES) || (devfn > 255) || (reg > 255)) + return -EINVAL; + + spin_lock_irqsave(&pci_config_lock, flags); + + write_cf8(bus, devfn, reg); + + switch (len) { + case 1: + if (xquad_portio) + *value = readb(adr + (reg & 3)); + else + *value = inb(0xCFC + (reg & 3)); + break; + case 2: + if (xquad_portio) + *value = readw(adr + (reg & 2)); + else + *value = inw(0xCFC + (reg & 2)); + break; + case 4: + if (xquad_portio) + *value = readl(adr); + else + *value = inl(0xCFC); + break; + } + + spin_unlock_irqrestore(&pci_config_lock, flags); + + return 0; +} + +static int pci_conf1_mq_write(unsigned int seg, unsigned int bus, + unsigned int devfn, int reg, int len, u32 value) +{ + unsigned long flags; + void *adr __iomem = XQUAD_PORT_ADDR(0xcfc, BUS2QUAD(bus)); + + if ((bus >= MAX_MP_BUSSES) || (devfn > 255) || (reg > 255)) + return -EINVAL; + + spin_lock_irqsave(&pci_config_lock, flags); + + write_cf8(bus, devfn, reg); + + switch (len) { + case 1: + if (xquad_portio) + writeb(value, adr + (reg & 3)); + else + outb((u8)value, 0xCFC + (reg & 3)); + break; + case 2: + if (xquad_portio) + writew(value, adr + (reg & 2)); + else + outw((u16)value, 0xCFC + (reg & 2)); + break; + case 4: + if (xquad_portio) + writel(value, adr + reg); + else + outl((u32)value, 0xCFC); + break; + } + + spin_unlock_irqrestore(&pci_config_lock, flags); + + return 0; +} + +#undef PCI_CONF1_MQ_ADDRESS + +static struct pci_raw_ops pci_direct_conf1_mq = { + .read = pci_conf1_mq_read, + .write = pci_conf1_mq_write +}; + + +static void __devinit pci_fixup_i450nx(struct pci_dev *d) +{ + /* + * i450NX -- Find and scan all secondary buses on all PXB's. + */ + int pxb, reg; + u8 busno, suba, subb; + int quad = BUS2QUAD(d->bus->number); + + printk("PCI: Searching for i450NX host bridges on %s\n", pci_name(d)); + reg = 0xd0; + for(pxb=0; pxb<2; pxb++) { + pci_read_config_byte(d, reg++, &busno); + pci_read_config_byte(d, reg++, &suba); + pci_read_config_byte(d, reg++, &subb); + DBG("i450NX PXB %d: %02x/%02x/%02x\n", pxb, busno, suba, subb); + if (busno) { + /* Bus A */ + pci_scan_bus_with_sysdata(QUADLOCAL2BUS(quad, busno)); + } + if (suba < subb) { + /* Bus B */ + pci_scan_bus_with_sysdata(QUADLOCAL2BUS(quad, suba+1)); + } + } + pcibios_last_bus = -1; +} +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82451NX, pci_fixup_i450nx); + +int __init pci_numa_init(void) +{ + int quad; + + if (!found_numaq) + return 0; + + raw_pci_ops = &pci_direct_conf1_mq; + + if (pcibios_scanned++) + return 0; + + pci_root_bus = pcibios_scan_root(0); + if (pci_root_bus) + pci_bus_add_devices(pci_root_bus); + if (num_online_nodes() > 1) + for_each_online_node(quad) { + if (quad == 0) + continue; + printk("Scanning PCI bus %d for quad %d\n", + QUADLOCAL2BUS(quad,0), quad); + pci_scan_bus_with_sysdata(QUADLOCAL2BUS(quad, 0)); + } + return 0; +} -- cgit v1.2.3 From e27cf3a2e151b79375efadf71a5d383ad416fb44 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 11 Jul 2008 12:18:41 +0200 Subject: x86/pci: renaming numa into numaq Cc: Yinghai Lu Signed-off-by: Robert Richter Cc: Robert Richter Cc: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/pci/legacy.c | 2 +- arch/x86/pci/numaq_32.c | 4 ++-- arch/x86/pci/pci.h | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/pci/legacy.c b/arch/x86/pci/legacy.c index 60e8caa1356..f405eb0b891 100644 --- a/arch/x86/pci/legacy.c +++ b/arch/x86/pci/legacy.c @@ -58,7 +58,7 @@ static int __init pci_legacy_init(void) int __init pci_subsys_init(void) { #ifdef CONFIG_X86_NUMAQ - pci_numa_init(); + pci_numaq_init(); #endif #ifdef CONFIG_ACPI pci_acpi_init(); diff --git a/arch/x86/pci/numaq_32.c b/arch/x86/pci/numaq_32.c index 8b5ca196673..f4b16dc11da 100644 --- a/arch/x86/pci/numaq_32.c +++ b/arch/x86/pci/numaq_32.c @@ -1,5 +1,5 @@ /* - * numa.c - Low-level PCI access for NUMA-Q machines + * numaq_32.c - Low-level PCI access for NUMA-Q machines */ #include @@ -151,7 +151,7 @@ static void __devinit pci_fixup_i450nx(struct pci_dev *d) } DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82451NX, pci_fixup_i450nx); -int __init pci_numa_init(void) +int __init pci_numaq_init(void) { int quad; diff --git a/arch/x86/pci/pci.h b/arch/x86/pci/pci.h index b2270a55b0c..36b8dd019fa 100644 --- a/arch/x86/pci/pci.h +++ b/arch/x86/pci/pci.h @@ -107,7 +107,7 @@ extern void __init dmi_check_skip_isa_align(void); /* some common used subsys_initcalls */ extern int __init pci_acpi_init(void); extern int __init pcibios_irq_init(void); -extern int __init pci_numa_init(void); +extern int __init pci_numaq_init(void); extern int __init pcibios_init(void); /* pci-mmconfig.c */ -- cgit v1.2.3 From 3cabf37f6167125cb5185db05f5061650f685ab7 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Fri, 11 Jul 2008 12:26:59 +0200 Subject: x86/pci: Changing subsystem init for visws I don't know, if this new code boots, but at least it compiles. Someone should really test it. Signed-off-by: Robert Richter Cc: Robert Richter Signed-off-by: Ingo Molnar --- arch/x86/pci/legacy.c | 3 +++ arch/x86/pci/pci.h | 1 + arch/x86/pci/visws.c | 23 +++++++---------------- 3 files changed, 11 insertions(+), 16 deletions(-) (limited to 'arch') diff --git a/arch/x86/pci/legacy.c b/arch/x86/pci/legacy.c index f405eb0b891..ec9ce35e44d 100644 --- a/arch/x86/pci/legacy.c +++ b/arch/x86/pci/legacy.c @@ -62,6 +62,9 @@ int __init pci_subsys_init(void) #endif #ifdef CONFIG_ACPI pci_acpi_init(); +#endif +#ifdef CONFIG_X86_VISWS + pci_visws_init(); #endif pci_legacy_init(); pcibios_irq_init(); diff --git a/arch/x86/pci/pci.h b/arch/x86/pci/pci.h index 36b8dd019fa..a2c55ee98af 100644 --- a/arch/x86/pci/pci.h +++ b/arch/x86/pci/pci.h @@ -107,6 +107,7 @@ extern void __init dmi_check_skip_isa_align(void); /* some common used subsys_initcalls */ extern int __init pci_acpi_init(void); extern int __init pcibios_irq_init(void); +extern int __init pci_visws_init(void); extern int __init pci_numaq_init(void); extern int __init pcibios_init(void); diff --git a/arch/x86/pci/visws.c b/arch/x86/pci/visws.c index 1a7bed492bb..42f4cb19fac 100644 --- a/arch/x86/pci/visws.c +++ b/arch/x86/pci/visws.c @@ -86,8 +86,14 @@ void __init pcibios_update_irq(struct pci_dev *dev, int irq) pci_write_config_byte(dev, PCI_INTERRUPT_LINE, irq); } -static int __init pci_visws_init(void) +int __init pci_visws_init(void) { + if (!is_visws_box()) + return -1; + + pcibios_enable_irq = &pci_visws_enable_irq; + pcibios_disable_irq = &pci_visws_disable_irq; + /* The VISWS supports configuration access type 1 only */ pci_probe = (pci_probe | PCI_PROBE_CONF1) & ~(PCI_PROBE_BIOS | PCI_PROBE_CONF2); @@ -105,18 +111,3 @@ static int __init pci_visws_init(void) pcibios_resource_survey(); return 0; } - -static __init int pci_subsys_init(void) -{ - if (!is_visws_box()) - return -1; - - pcibios_enable_irq = &pci_visws_enable_irq; - pcibios_disable_irq = &pci_visws_disable_irq; - - pci_visws_init(); - pcibios_init(); - - return 0; -} -subsys_initcall(pci_subsys_init); -- cgit v1.2.3 From d54191b85e294c46f05a2249b1f55ae54930bcc7 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 22 Apr 2008 15:09:30 +0200 Subject: Kprobe smoke test lockdep warning On Mon, 2008-04-21 at 18:54 -0400, Masami Hiramatsu wrote: > Thank you for reporting. > > Actually, kprobes tries to fixup thread's flags in post_kprobe_handler > (which is called from kprobe_exceptions_notify) by > trace_hardirqs_fixup_flags(pt_regs->flags). However, even the irq flag > is set in pt_regs->flags, true hardirq is still off until returning > from do_debug. Thus, lockdep assumes that hardirq is off without annotation. > > IMHO, one possible solution is that fixing hardirq flags right after > notify_die in do_debug instead of in post_kprobe_handler. My reply to BZ 10489: > [ 2.707509] Kprobe smoke test started > [ 2.709300] ------------[ cut here ]------------ > [ 2.709420] WARNING: at kernel/lockdep.c:2658 check_flags+0x4d/0x12c() > [ 2.709541] Modules linked in: > [ 2.709588] Pid: 1, comm: swapper Not tainted 2.6.25.jml.057 #1 > [ 2.709588] [] warn_on_slowpath+0x41/0x51 > [ 2.709588] [] ? save_stack_trace+0x1d/0x3b > [ 2.709588] [] ? save_trace+0x37/0x89 > [ 2.709588] [] ? kernel_map_pages+0x103/0x11c > [ 2.709588] [] ? native_sched_clock+0xca/0xea > [ 2.709588] [] ? mark_held_locks+0x41/0x5c > [ 2.709588] [] ? kprobe_exceptions_notify+0x322/0x3af > [ 2.709588] [] ? trace_hardirqs_on+0xf1/0x119 > [ 2.709588] [] ? kprobe_exceptions_notify+0x355/0x3af > [ 2.709588] [] check_flags+0x4d/0x12c > [ 2.709588] [] lock_release+0x58/0x195 > [ 2.709588] [] ? __atomic_notifier_call_chain+0x0/0x80 > [ 2.709588] [] __atomic_notifier_call_chain+0x5a/0x80 > [ 2.709588] [] atomic_notifier_call_chain+0xc/0xe > [ 2.709588] [] notify_die+0x2d/0x2f > [ 2.709588] [] do_debug+0x67/0xfe > [ 2.709588] [] debug_stack_correct+0x27/0x30 > [ 2.709588] [] ? kprobe_target+0x1/0x34 > [ 2.709588] [] ? init_test_probes+0x50/0x186 > [ 2.709588] [] init_kprobes+0x85/0x8c > [ 2.709588] [] kernel_init+0x13d/0x298 > [ 2.709588] [] ? kernel_init+0x0/0x298 > [ 2.709588] [] ? kernel_init+0x0/0x298 > [ 2.709588] [] kernel_thread_helper+0x7/0x10 > [ 2.709588] ======================= > [ 2.709588] ---[ end trace 778e504de7e3b1e3 ]--- > [ 2.709588] possible reason: unannotated irqs-off. > [ 2.709588] irq event stamp: 370065 > [ 2.709588] hardirqs last enabled at (370065): [] kprobe_exceptions_notify+0x322/0x3af > [ 2.709588] hardirqs last disabled at (370064): [] do_int3+0x1d/0x7d > [ 2.709588] softirqs last enabled at (370050): [] __do_softirq+0xfa/0x100 > [ 2.709588] softirqs last disabled at (370045): [] do_softirq+0x74/0xd9 > [ 2.714751] Kprobe smoke test passed successfully how I love this stuff... Ok, do_debug() is a trap, this can happen at any time regardless of the machine's IRQ state. So the first thing we do is fix up the IRQ state. Then we call this die notifier stuff; and return with messed up IRQ state... YAY. So, kprobes fudges it.. notify_die(DIE_DEBUG) kprobe_exceptions_notify() post_kprobe_handler() modify regs->flags trace_hardirqs_fixup_flags(regs->flags); <--- must be it So what's the use of modifying flags if they're not meant to take effect at some point. /me tries to reproduce issue; enable kprobes test thingy && boot OK, that reproduces.. So the below makes it work - but I'm not getting this code; at the time I wrote that stuff I CC'ed each and every kprobe maintainer listed in the usual places but got no reposonse - can some please explain this stuff to me? Are the saved flags only for the TF bit or are they made in full effect later (and if so, where) ? Signed-off-by: Peter Zijlstra Acked-by: Masami Hiramatsu Signed-off-by: Ingo Molnar --- arch/x86/kernel/kprobes.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c index b8c6743a13d..43c019f85f0 100644 --- a/arch/x86/kernel/kprobes.c +++ b/arch/x86/kernel/kprobes.c @@ -860,7 +860,6 @@ static int __kprobes post_kprobe_handler(struct pt_regs *regs) resume_execution(cur, regs, kcb); regs->flags |= kcb->kprobe_saved_flags; - trace_hardirqs_fixup_flags(regs->flags); if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) { kcb->kprobe_status = KPROBE_HIT_SSDONE; -- cgit v1.2.3 From a312b37b2a212fd2e227d1d6321f903b91b65ec7 Mon Sep 17 00:00:00 2001 From: Eduardo Habkost Date: Tue, 8 Jul 2008 15:06:23 -0700 Subject: x86/paravirt: call paravirt_pagetable_setup_{start, done} Call paravirt_pagetable_setup_{start,done} These paravirt_ops functions were not being called on x86_64. Signed-off-by: Eduardo Habkost Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/paravirt.c | 4 ++++ arch/x86/kernel/setup.c | 2 ++ arch/x86/xen/enlighten.c | 4 ++++ 3 files changed, 10 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index e0f571d58c1..2963ab5d91e 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -29,6 +29,7 @@ #include #include #include +#include #include #include #include @@ -373,6 +374,9 @@ struct pv_mmu_ops pv_mmu_ops = { #ifndef CONFIG_X86_64 .pagetable_setup_start = native_pagetable_setup_start, .pagetable_setup_done = native_pagetable_setup_done, +#else + .pagetable_setup_start = paravirt_nop, + .pagetable_setup_done = paravirt_nop, #endif .read_cr2 = native_read_cr2, diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 36c540d4ac4..8ce6a91ce10 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -819,7 +819,9 @@ void __init setup_arch(char **cmdline_p) vmi_init(); #endif + paravirt_pagetable_setup_start(swapper_pg_dir); paging_init(); + paravirt_pagetable_setup_done(swapper_pg_dir); #ifdef CONFIG_X86_64 map_vsyscall(); diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index bb508456ef5..eaab6c9b4a8 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -841,6 +841,7 @@ static __init void xen_set_pte_init(pte_t *ptep, pte_t pte) static __init void xen_pagetable_setup_start(pgd_t *base) { +#ifdef CONFIG_X86_32 pgd_t *xen_pgd = (pgd_t *)xen_start_info->pt_base; int i; @@ -886,6 +887,7 @@ static __init void xen_pagetable_setup_start(pgd_t *base) /* Unpin initial Xen pagetable */ pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(xen_start_info->pt_base))); +#endif /* CONFIG_X86_32 */ } void xen_setup_shared_info(void) @@ -927,9 +929,11 @@ static __init void xen_pagetable_setup_done(pgd_t *base) xen_setup_shared_info(); +#ifdef CONFIG_X86_32 /* Actually pin the pagetable down, but we can't set PG_pinned yet because the page structures don't exist yet. */ pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(base))); +#endif } static __init void xen_post_allocator_init(void) -- cgit v1.2.3 From c1f2f09ef66d5dadd5fe42ea909e708470c9636d Mon Sep 17 00:00:00 2001 From: Eduardo Habkost Date: Tue, 8 Jul 2008 15:06:24 -0700 Subject: pvops-64: call paravirt_post_allocator_init() on setup_arch() Signed-off-by: Eduardo Habkost Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/setup.c | 1 + arch/x86/mm/init_32.c | 2 -- arch/x86/xen/mmu.c | 8 +++++--- 3 files changed, 6 insertions(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 8ce6a91ce10..2ed504b97d4 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -822,6 +822,7 @@ void __init setup_arch(char **cmdline_p) paravirt_pagetable_setup_start(swapper_pg_dir); paging_init(); paravirt_pagetable_setup_done(swapper_pg_dir); + paravirt_post_allocator_init(); #ifdef CONFIG_X86_64 map_vsyscall(); diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c index 9689a5138e6..7113acd8ac4 100644 --- a/arch/x86/mm/init_32.c +++ b/arch/x86/mm/init_32.c @@ -868,8 +868,6 @@ void __init paging_init(void) */ sparse_init(); zone_sizes_init(); - - paravirt_post_allocator_init(); } /* diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index ff0aa74afaa..ebd6900e331 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -656,9 +656,11 @@ void xen_mm_pin_all(void) spin_unlock_irqrestore(&pgd_lock, flags); } -/* The init_mm pagetable is really pinned as soon as its created, but - that's before we have page structures to store the bits. So do all - the book-keeping now. */ +/* + * The init_mm pagetable is really pinned as soon as its created, but + * that's before we have page structures to store the bits. So do all + * the book-keeping now. + */ static __init int mark_pinned(struct page *page, enum pt_level level) { SetPagePinned(page); -- cgit v1.2.3 From 8840c0ccd763936a8e730ece118197a51be8dc8e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:25 -0700 Subject: x86_64: there's no need to preallocate level1_fixmap_pgt Early fixmap will allocate its own L1 pagetable page for fixmap mappings, so there's no need to preallocate one. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 6 ------ 1 file changed, 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index b07ac7b217c..4b6bda21837 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -362,12 +362,6 @@ NEXT_PAGE(level3_kernel_pgt) .quad level2_fixmap_pgt - __START_KERNEL_map + _PAGE_TABLE NEXT_PAGE(level2_fixmap_pgt) - .fill 506,8,0 - .quad level1_fixmap_pgt - __START_KERNEL_map + _PAGE_TABLE - /* 8MB reserved for vsyscalls + a 2MB hole = 4 + 1 entries */ - .fill 5,8,0 - -NEXT_PAGE(level1_fixmap_pgt) .fill 512,8,0 NEXT_PAGE(level2_ident_pgt) -- cgit v1.2.3 From 87b935a0ef9a1ddf62f2f0c0fc17b10654ff41cd Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:26 -0700 Subject: x86: clean up formatting of __switch_to process_64.c:__switch_to has some very old strange formatting, some of it dating back to pre-git. Fix it up. No functional changes. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/process_64.c | 56 ++++++++++++++++++++++---------------------- 1 file changed, 28 insertions(+), 28 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c index a8e53626ac9..e8a8e1b9981 100644 --- a/arch/x86/kernel/process_64.c +++ b/arch/x86/kernel/process_64.c @@ -537,8 +537,8 @@ static inline void __switch_to_xtra(struct task_struct *prev_p, struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct *next_p) { - struct thread_struct *prev = &prev_p->thread, - *next = &next_p->thread; + struct thread_struct *prev = &prev_p->thread; + struct thread_struct *next = &next_p->thread; int cpu = smp_processor_id(); struct tss_struct *tss = &per_cpu(init_tss, cpu); unsigned fsindex, gsindex; @@ -586,35 +586,34 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) /* * Switch FS and GS. + * + * Segment register != 0 always requires a reload. Also + * reload when it has changed. When prev process used 64bit + * base always reload to avoid an information leak. */ - { - /* segment register != 0 always requires a reload. - also reload when it has changed. - when prev process used 64bit base always reload - to avoid an information leak. */ - if (unlikely(fsindex | next->fsindex | prev->fs)) { - loadsegment(fs, next->fsindex); - /* check if the user used a selector != 0 - * if yes clear 64bit base, since overloaded base - * is always mapped to the Null selector - */ - if (fsindex) + if (unlikely(fsindex | next->fsindex | prev->fs)) { + loadsegment(fs, next->fsindex); + /* + * Check if the user used a selector != 0; if yes + * clear 64bit base, since overloaded base is always + * mapped to the Null selector + */ + if (fsindex) prev->fs = 0; - } - /* when next process has a 64bit base use it */ - if (next->fs) - wrmsrl(MSR_FS_BASE, next->fs); - prev->fsindex = fsindex; - - if (unlikely(gsindex | next->gsindex | prev->gs)) { - load_gs_index(next->gsindex); - if (gsindex) + } + /* when next process has a 64bit base use it */ + if (next->fs) + wrmsrl(MSR_FS_BASE, next->fs); + prev->fsindex = fsindex; + + if (unlikely(gsindex | next->gsindex | prev->gs)) { + load_gs_index(next->gsindex); + if (gsindex) prev->gs = 0; - } - if (next->gs) - wrmsrl(MSR_KERNEL_GS_BASE, next->gs); - prev->gsindex = gsindex; } + if (next->gs) + wrmsrl(MSR_KERNEL_GS_BASE, next->gs); + prev->gsindex = gsindex; /* Must be after DS reload */ unlazy_fpu(prev_p); @@ -627,7 +626,8 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) write_pda(pcurrent, next_p); write_pda(kernelstack, - (unsigned long)task_stack_page(next_p) + THREAD_SIZE - PDA_STACKOFFSET); + (unsigned long)task_stack_page(next_p) + + THREAD_SIZE - PDA_STACKOFFSET); #ifdef CONFIG_CC_STACKPROTECTOR write_pda(stack_canary, next_p->stack_canary); /* -- cgit v1.2.3 From cbcd79c2e5b496b84845618cef734b4c40736576 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:27 -0700 Subject: x86: use __page_aligned_data/bss Update arch/x86's use of page-aligned variables. The change to arch/x86/xen/mmu.c fixes an actual bug, but the rest are cleanups and to set a precedent. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common_64.c | 4 ++-- arch/x86/kernel/irq_32.c | 7 ++----- arch/x86/xen/mmu.c | 15 ++++++--------- 3 files changed, 10 insertions(+), 16 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index 7b8cc72feb4..15419cd3c5a 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -16,6 +16,7 @@ #include #include #include +#include #include #include #include @@ -517,8 +518,7 @@ void pda_init(int cpu) } char boot_exception_stacks[(N_EXCEPTION_STACKS - 1) * EXCEPTION_STKSZ + - DEBUG_STKSZ] -__attribute__((section(".bss.page_aligned"))); + DEBUG_STKSZ] __page_aligned_bss; extern asmlinkage void ignore_sysret(void); diff --git a/arch/x86/kernel/irq_32.c b/arch/x86/kernel/irq_32.c index 47a6f6f1247..1cf8c1fcc08 100644 --- a/arch/x86/kernel/irq_32.c +++ b/arch/x86/kernel/irq_32.c @@ -83,11 +83,8 @@ union irq_ctx { static union irq_ctx *hardirq_ctx[NR_CPUS] __read_mostly; static union irq_ctx *softirq_ctx[NR_CPUS] __read_mostly; -static char softirq_stack[NR_CPUS * THREAD_SIZE] - __attribute__((__section__(".bss.page_aligned"))); - -static char hardirq_stack[NR_CPUS * THREAD_SIZE] - __attribute__((__section__(".bss.page_aligned"))); +static char softirq_stack[NR_CPUS * THREAD_SIZE] __page_aligned_bss; +static char hardirq_stack[NR_CPUS * THREAD_SIZE] __page_aligned_bss; static void call_on_stack(void *func, void *stack) { diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index ebd6900e331..4fca9d88bef 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -46,6 +46,7 @@ #include #include #include +#include #include #include @@ -60,22 +61,18 @@ #define TOP_ENTRIES (MAX_DOMAIN_PAGES / P2M_ENTRIES_PER_PAGE) /* Placeholder for holes in the address space */ -static unsigned long p2m_missing[P2M_ENTRIES_PER_PAGE] - __attribute__((section(".data.page_aligned"))) = +static unsigned long p2m_missing[P2M_ENTRIES_PER_PAGE] __page_aligned_data = { [ 0 ... P2M_ENTRIES_PER_PAGE-1 ] = ~0UL }; /* Array of pointers to pages containing p2m entries */ -static unsigned long *p2m_top[TOP_ENTRIES] - __attribute__((section(".data.page_aligned"))) = +static unsigned long *p2m_top[TOP_ENTRIES] __page_aligned_data = { [ 0 ... TOP_ENTRIES - 1] = &p2m_missing[0] }; /* Arrays of p2m arrays expressed in mfns used for save/restore */ -static unsigned long p2m_top_mfn[TOP_ENTRIES] - __attribute__((section(".bss.page_aligned"))); +static unsigned long p2m_top_mfn[TOP_ENTRIES] __page_aligned_bss; -static unsigned long p2m_top_mfn_list[ - PAGE_ALIGN(TOP_ENTRIES / P2M_ENTRIES_PER_PAGE)] - __attribute__((section(".bss.page_aligned"))); +static unsigned long p2m_top_mfn_list[TOP_ENTRIES / P2M_ENTRIES_PER_PAGE] + __page_aligned_bss; static inline unsigned p2m_top_index(unsigned long pfn) { -- cgit v1.2.3 From 360c044eb1b985a9ef29d952276a3e14973bed93 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:28 -0700 Subject: x86_64: adjust exception frame in ia32entry The 32-bit compat int $0x80 entrypoint needs exception frame adjustment. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/ia32/ia32entry.S | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S index 20371d0635e..0ae1e77eae5 100644 --- a/arch/x86/ia32/ia32entry.S +++ b/arch/x86/ia32/ia32entry.S @@ -321,6 +321,7 @@ ENTRY(ia32_syscall) /*CFI_REL_OFFSET rflags,EFLAGS-RIP*/ /*CFI_REL_OFFSET cs,CS-RIP*/ CFI_REL_OFFSET rip,RIP-RIP + PARAVIRT_ADJUST_EXCEPTION_FRAME SWAPGS /* * No need to follow this irqs on/off section: the syscall -- cgit v1.2.3 From 7c33b1e6ee26d67551109aca04d46544d0ce55b1 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:29 -0700 Subject: x86_64: unstatic get_local_pda This allows Xen's xen_cpu_up() to allocate a pda for the new CPU. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/smpboot.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 687376ab07e..1deb3b624a7 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -768,7 +768,7 @@ static void __cpuinit do_fork_idle(struct work_struct *work) * * Must be called after the _cpu_pda pointer table is initialized. */ -static int __cpuinit get_local_pda(int cpu) +int __cpuinit get_local_pda(int cpu) { struct x8664_pda *oldpda, *newpda; unsigned long size = sizeof(struct x8664_pda); -- cgit v1.2.3 From 8ba6c2b0958c332d2f3336f4ca9c116ed81f38e9 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:30 -0700 Subject: xen: print backtrace on multicall failure Print a backtrace if a multicall fails, to help with debugging. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/multicalls.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/xen/multicalls.c b/arch/x86/xen/multicalls.c index 3c63c4da7ed..9efd1c6c977 100644 --- a/arch/x86/xen/multicalls.c +++ b/arch/x86/xen/multicalls.c @@ -76,6 +76,7 @@ void xen_mc_flush(void) if (ret) { printk(KERN_ERR "%d multicall(s) failed: cpu %d\n", ret, smp_processor_id()); + dump_stack(); for (i = 0; i < b->mcidx; i++) { printk(" call %2d/%d: op=%lu arg=[%lx] result=%ld\n", i+1, b->mcidx, -- cgit v1.2.3 From ad55db9fed6d6cd09333045945cb03ba2c070085 Mon Sep 17 00:00:00 2001 From: Isaku Yamahata Date: Tue, 8 Jul 2008 15:06:32 -0700 Subject: xen: add xen_arch_resume()/xen_timer_resume hook for ia64 support add xen_timer_resume() hook. Timer resume should be done after event channel is resumed. add xen_arch_resume() hook when ipi becomes usable after resume. After resume, some cpu specific resource must be reinitialized on ia64 that can't be set by another cpu. However available hooks is run once on only one cpu so that ipi has to be used. During stop_machine_run() ipi can't be used because interrupt is masked. So add another hook after stop_machine_run(). Another approach might be use resume hook which is run by device_resume(). However device_resume() may be executed on suspend error recovery path. So it is necessary to determine whether it is executed on real resume path or error recovery path. Signed-off-by: Isaku Yamahata Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/suspend.c | 5 ++++- arch/x86/xen/xen-ops.h | 1 - 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/suspend.c b/arch/x86/xen/suspend.c index 251669a932d..2a234db5949 100644 --- a/arch/x86/xen/suspend.c +++ b/arch/x86/xen/suspend.c @@ -38,8 +38,11 @@ void xen_post_suspend(int suspend_cancelled) xen_cpu_initialized_map = cpu_online_map; #endif xen_vcpu_restore(); - xen_timer_resume(); } } +void xen_arch_resume(void) +{ + /* nothing */ +} diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 6f4b1045c1c..77354d20425 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -37,7 +37,6 @@ void __init xen_time_init(void); unsigned long xen_get_wallclock(void); int xen_set_wallclock(unsigned long time); unsigned long long xen_sched_clock(void); -void xen_timer_resume(void); irqreturn_t xen_debug_interrupt(int irq, void *dev_id); -- cgit v1.2.3 From 851fa3c4e7b50d6a946d8b4c0a68683b5e56b2f1 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:33 -0700 Subject: xen: define set_pte from the outset We need set_pte to work from a relatively early point, so enable it from the start. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index eaab6c9b4a8..c5f0b40aa39 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -845,9 +845,6 @@ static __init void xen_pagetable_setup_start(pgd_t *base) pgd_t *xen_pgd = (pgd_t *)xen_start_info->pt_base; int i; - /* special set_pte for pagetable initialization */ - pv_mmu_ops.set_pte = xen_set_pte_init; - init_mm.pgd = base; /* * copy top-level of Xen-supplied pagetable into place. This @@ -1174,7 +1171,7 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .kmap_atomic_pte = xen_kmap_atomic_pte, #endif - .set_pte = NULL, /* see xen_pagetable_setup_* */ + .set_pte = xen_set_pte_init, .set_pte_at = xen_set_pte_at, .set_pmd = xen_set_pmd_hyper, -- cgit v1.2.3 From 48b5db20621388582ca11ac3c61d3403966dbe51 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:34 -0700 Subject: xen64: define asm/xen/interface for 64-bit Copy 64-bit definitions of various interface structures into place. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.h | 12 ------------ 1 file changed, 12 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index 297bf9f5b8b..7856e37f604 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -10,18 +10,6 @@ enum pt_level { PT_PTE }; -/* - * Page-directory addresses above 4GB do not fit into architectural %cr3. - * When accessing %cr3, or equivalent field in vcpu_guest_context, guests - * must use the following accessor macros to pack/unpack valid MFNs. - * - * Note that Xen is using the fact that the pagetable base is always - * page-aligned, and putting the 12 MSB of the address into the 12 LSB - * of cr3. - */ -#define xen_pfn_to_cr3(pfn) (((unsigned)(pfn) << 12) | ((unsigned)(pfn) >> 20)) -#define xen_cr3_to_pfn(cr3) (((unsigned)(cr3) >> 12) | ((unsigned)(cr3) << 20)) - void set_pte_mfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags); -- cgit v1.2.3 From 7077c33d81a8d790135ae87cd19e6efcb075c23a Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:35 -0700 Subject: xen: make ELF notes work for 32 and 64 bit Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/xen-head.S | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S index 7c0cf6320a0..a9cac9dc04b 100644 --- a/arch/x86/xen/xen-head.S +++ b/arch/x86/xen/xen-head.S @@ -5,7 +5,10 @@ #include #include + #include +#include + #include #include @@ -21,21 +24,21 @@ ENTRY(startup_xen) .pushsection .text .align PAGE_SIZE_asm ENTRY(hypercall_page) - .skip 0x1000 + .skip PAGE_SIZE_asm .popsection ELFNOTE(Xen, XEN_ELFNOTE_GUEST_OS, .asciz "linux") ELFNOTE(Xen, XEN_ELFNOTE_GUEST_VERSION, .asciz "2.6") ELFNOTE(Xen, XEN_ELFNOTE_XEN_VERSION, .asciz "xen-3.0") - ELFNOTE(Xen, XEN_ELFNOTE_VIRT_BASE, .long __PAGE_OFFSET) - ELFNOTE(Xen, XEN_ELFNOTE_ENTRY, .long startup_xen) - ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, .long hypercall_page) + ELFNOTE(Xen, XEN_ELFNOTE_VIRT_BASE, _ASM_PTR __PAGE_OFFSET) + ELFNOTE(Xen, XEN_ELFNOTE_ENTRY, _ASM_PTR startup_xen) + ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, _ASM_PTR hypercall_page) ELFNOTE(Xen, XEN_ELFNOTE_FEATURES, .asciz "!writable_page_tables|pae_pgdir_above_4gb") ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE, .asciz "yes") ELFNOTE(Xen, XEN_ELFNOTE_LOADER, .asciz "generic") ELFNOTE(Xen, XEN_ELFNOTE_L1_MFN_VALID, .quad _PAGE_PRESENT; .quad _PAGE_PRESENT) ELFNOTE(Xen, XEN_ELFNOTE_SUSPEND_CANCEL, .long 1) - ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW, .long __HYPERVISOR_VIRT_START) + ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW, _ASM_PTR __HYPERVISOR_VIRT_START) #endif /*CONFIG_XEN */ -- cgit v1.2.3 From f6e587325b3bc7e5c829a407ddc25b52c1e73851 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:38 -0700 Subject: xen64: add extra pv_mmu_ops We need extra pv_mmu_ops for 64-bit, to deal with the extra level of pagetable. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 33 ++++++++++++++++++++++++++++++- arch/x86/xen/mmu.c | 51 +++++++++++++++++++++++++++++++++++++++++++++++- arch/x86/xen/mmu.h | 15 ++++++++++++-- 3 files changed, 95 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index c5f0b40aa39..afb047e30bd 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -803,6 +803,18 @@ static void xen_release_pmd(u32 pfn) xen_release_ptpage(pfn, PT_PMD); } +#if PAGETABLE_LEVELS == 4 +static void xen_alloc_pud(struct mm_struct *mm, u32 pfn) +{ + xen_alloc_ptpage(mm, pfn, PT_PUD); +} + +static void xen_release_pud(u32 pfn) +{ + xen_release_ptpage(pfn, PT_PUD); +} +#endif + #ifdef CONFIG_HIGHPTE static void *xen_kmap_atomic_pte(struct page *page, enum km_type type) { @@ -922,6 +934,11 @@ static __init void xen_pagetable_setup_done(pgd_t *base) pv_mmu_ops.alloc_pmd = xen_alloc_pmd; pv_mmu_ops.release_pte = xen_release_pte; pv_mmu_ops.release_pmd = xen_release_pmd; +#if PAGETABLE_LEVELS == 4 + pv_mmu_ops.alloc_pud = xen_alloc_pud; + pv_mmu_ops.release_pud = xen_release_pud; +#endif + pv_mmu_ops.set_pte = xen_set_pte; xen_setup_shared_info(); @@ -937,6 +954,9 @@ static __init void xen_post_allocator_init(void) { pv_mmu_ops.set_pmd = xen_set_pmd; pv_mmu_ops.set_pud = xen_set_pud; +#if PAGETABLE_LEVELS == 4 + pv_mmu_ops.set_pgd = xen_set_pgd; +#endif xen_mark_init_mm_pinned(); } @@ -1185,15 +1205,26 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .make_pte = xen_make_pte, .make_pgd = xen_make_pgd, +#ifdef CONFIG_X86_PAE .set_pte_atomic = xen_set_pte_atomic, .set_pte_present = xen_set_pte_at, - .set_pud = xen_set_pud_hyper, .pte_clear = xen_pte_clear, .pmd_clear = xen_pmd_clear, +#endif /* CONFIG_X86_PAE */ + .set_pud = xen_set_pud_hyper, .make_pmd = xen_make_pmd, .pmd_val = xen_pmd_val, +#if PAGETABLE_LEVELS == 4 + .pud_val = xen_pud_val, + .make_pud = xen_make_pud, + .set_pgd = xen_set_pgd_hyper, + + .alloc_pud = xen_alloc_pte_init, + .release_pud = xen_release_pte_init, +#endif /* PAGETABLE_LEVELS == 4 */ + .activate_mm = xen_activate_mm, .dup_mmap = xen_dup_mmap, .exit_mmap = xen_exit_mmap, diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 4fca9d88bef..d0976b87cd2 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -438,14 +438,19 @@ void xen_set_pud(pud_t *ptr, pud_t val) void xen_set_pte(pte_t *ptep, pte_t pte) { +#ifdef CONFIG_X86_PAE ptep->pte_high = pte.pte_high; smp_wmb(); ptep->pte_low = pte.pte_low; +#else + *ptep = pte; +#endif } +#ifdef CONFIG_X86_PAE void xen_set_pte_atomic(pte_t *ptep, pte_t pte) { - set_64bit((u64 *)ptep, pte_val_ma(pte)); + set_64bit((u64 *)ptep, native_pte_val(pte)); } void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep) @@ -459,6 +464,7 @@ void xen_pmd_clear(pmd_t *pmdp) { set_pmd(pmdp, __pmd(0)); } +#endif /* CONFIG_X86_PAE */ pmd_t xen_make_pmd(pmdval_t pmd) { @@ -466,6 +472,49 @@ pmd_t xen_make_pmd(pmdval_t pmd) return native_make_pmd(pmd); } +#if PAGETABLE_LEVELS == 4 +pudval_t xen_pud_val(pud_t pud) +{ + return pte_mfn_to_pfn(pud.pud); +} + +pud_t xen_make_pud(pudval_t pud) +{ + pud = pte_pfn_to_mfn(pud); + + return native_make_pud(pud); +} + +void xen_set_pgd_hyper(pgd_t *ptr, pgd_t val) +{ + struct mmu_update u; + + preempt_disable(); + + xen_mc_batch(); + + u.ptr = virt_to_machine(ptr).maddr; + u.val = pgd_val_ma(val); + extend_mmu_update(&u); + + xen_mc_issue(PARAVIRT_LAZY_MMU); + + preempt_enable(); +} + +void xen_set_pgd(pgd_t *ptr, pgd_t val) +{ + /* If page is not pinned, we can just update the entry + directly */ + if (!page_pinned(ptr)) { + *ptr = val; + return; + } + + xen_set_pgd_hyper(ptr, val); +} +#endif /* PAGETABLE_LEVELS == 4 */ + /* (Yet another) pagetable walker. This one is intended for pinning a pagetable. This means that it walks a pagetable and calls the diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index 7856e37f604..19d544b0b6c 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -32,13 +32,24 @@ pgd_t xen_make_pgd(pgdval_t); void xen_set_pte(pte_t *ptep, pte_t pteval); void xen_set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval); + +#ifdef CONFIG_X86_PAE void xen_set_pte_atomic(pte_t *ptep, pte_t pte); +void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep); +void xen_pmd_clear(pmd_t *pmdp); +#endif /* CONFIG_X86_PAE */ + void xen_set_pmd(pmd_t *pmdp, pmd_t pmdval); void xen_set_pud(pud_t *ptr, pud_t val); void xen_set_pmd_hyper(pmd_t *pmdp, pmd_t pmdval); void xen_set_pud_hyper(pud_t *ptr, pud_t val); -void xen_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep); -void xen_pmd_clear(pmd_t *pmdp); + +#if PAGETABLE_LEVELS == 4 +pudval_t xen_pud_val(pud_t pud); +pud_t xen_make_pud(pudval_t pudval); +void xen_set_pgd(pgd_t *pgdp, pgd_t pgd); +void xen_set_pgd_hyper(pgd_t *pgdp, pgd_t pgd); +#endif pte_t xen_ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr, pte_t *ptep); void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr, -- cgit v1.2.3 From f5d36de069f4b343f64e858e7377cfc9c772c4fb Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:39 -0700 Subject: xen64: random ifdefs to mask out 32-bit only code Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index afb047e30bd..ada2e1a141d 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1296,6 +1296,7 @@ static const struct machine_ops __initdata xen_machine_ops = { static void __init xen_reserve_top(void) { +#ifdef CONFIG_X86_32 unsigned long top = HYPERVISOR_VIRT_START; struct xen_platform_parameters pp; @@ -1303,6 +1304,7 @@ static void __init xen_reserve_top(void) top = pp.virt_start; reserve_top_address(-top + 2 * PAGE_SIZE); +#endif /* CONFIG_X86_32 */ } /* First C function to be called on Xen boot */ @@ -1333,6 +1335,11 @@ asmlinkage void __init xen_start_kernel(void) machine_ops = xen_machine_ops; +#ifdef CONFIG_X86_64 + /* Disable until direct per-cpu data access. */ + have_vcpu_info_placement = 0; +#endif + #ifdef CONFIG_SMP smp_ops = xen_smp_ops; #endif @@ -1343,9 +1350,11 @@ asmlinkage void __init xen_start_kernel(void) pgd = (pgd_t *)xen_start_info->pt_base; +#ifdef CONFIG_X86_32 init_pg_tables_start = __pa(pgd); init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; max_pfn_mapped = (init_pg_tables_end + 512*1024) >> PAGE_SHIFT; +#endif init_mm.pgd = pgd; /* use the Xen pagetables to start */ @@ -1372,7 +1381,9 @@ asmlinkage void __init xen_start_kernel(void) /* set up basic CPUID stuff */ cpu_detect(&new_cpu_data); +#ifdef CONFIG_X86_32 new_cpu_data.hard_math = 1; +#endif new_cpu_data.x86_capability[0] = cpuid_edx(1); /* Poke various useful things into boot_params */ @@ -1388,5 +1399,9 @@ asmlinkage void __init xen_start_kernel(void) } /* Start the world */ +#ifdef CONFIG_X86_32 i386_start_kernel(); +#else + x86_64_start_kernel((char *)&boot_params); +#endif } -- cgit v1.2.3 From ce87b3d326de733c72b47662f106ee6cd699a20f Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:40 -0700 Subject: xen64: get active_mm from the pda x86_64 stores the active_mm in the pda, so fetch it from there. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index d0976b87cd2..2579e70cdd0 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -805,8 +805,15 @@ void xen_dup_mmap(struct mm_struct *oldmm, struct mm_struct *mm) static void drop_other_mm_ref(void *info) { struct mm_struct *mm = info; + struct mm_struct *active_mm; - if (__get_cpu_var(cpu_tlbstate).active_mm == mm) +#ifdef CONFIG_X86_64 + active_mm = read_pda(active_mm); +#else + active_mm = __get_cpu_var(cpu_tlbstate).active_mm; +#endif + + if (active_mm == mm) leave_mm(smp_processor_id()); /* If this cpu still has a stale cr3 reference, then make sure -- cgit v1.2.3 From a9e7062d7339f1a1df2b6d7e5d595c7d55b56bfb Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:41 -0700 Subject: xen: move smp setup into smp.c Move all the smp_ops setup into smp.c, allowing a lot of things to become static. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 19 +------------------ arch/x86/xen/smp.c | 34 ++++++++++++++++++++++++++-------- arch/x86/xen/xen-ops.h | 13 +++++-------- 3 files changed, 32 insertions(+), 34 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index ada2e1a141d..a85f447b8d0 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1237,21 +1237,6 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .set_fixmap = xen_set_fixmap, }; -#ifdef CONFIG_SMP -static const struct smp_ops xen_smp_ops __initdata = { - .smp_prepare_boot_cpu = xen_smp_prepare_boot_cpu, - .smp_prepare_cpus = xen_smp_prepare_cpus, - .cpu_up = xen_cpu_up, - .smp_cpus_done = xen_smp_cpus_done, - - .smp_send_stop = xen_smp_send_stop, - .smp_send_reschedule = xen_smp_send_reschedule, - - .send_call_func_ipi = xen_smp_send_call_function_ipi, - .send_call_func_single_ipi = xen_smp_send_call_function_single_ipi, -}; -#endif /* CONFIG_SMP */ - static void xen_reboot(int reason) { struct sched_shutdown r = { .reason = reason }; @@ -1340,9 +1325,7 @@ asmlinkage void __init xen_start_kernel(void) have_vcpu_info_placement = 0; #endif -#ifdef CONFIG_SMP - smp_ops = xen_smp_ops; -#endif + xen_smp_init(); /* Get mfn list */ if (!xen_feature(XENFEAT_auto_translated_physmap)) diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 233156f39b7..91fae8ff756 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -152,7 +152,7 @@ void __init xen_fill_possible_map(void) } } -void __init xen_smp_prepare_boot_cpu(void) +static void __init xen_smp_prepare_boot_cpu(void) { int cpu; @@ -176,7 +176,7 @@ void __init xen_smp_prepare_boot_cpu(void) xen_setup_vcpu_info_placement(); } -void __init xen_smp_prepare_cpus(unsigned int max_cpus) +static void __init xen_smp_prepare_cpus(unsigned int max_cpus) { unsigned cpu; @@ -276,7 +276,7 @@ cpu_initialize_context(unsigned int cpu, struct task_struct *idle) return 0; } -int __cpuinit xen_cpu_up(unsigned int cpu) +static int __cpuinit xen_cpu_up(unsigned int cpu) { struct task_struct *idle = idle_task(cpu); int rc; @@ -319,7 +319,7 @@ int __cpuinit xen_cpu_up(unsigned int cpu) return 0; } -void xen_smp_cpus_done(unsigned int max_cpus) +static void xen_smp_cpus_done(unsigned int max_cpus) { } @@ -335,12 +335,12 @@ static void stop_self(void *v) BUG(); } -void xen_smp_send_stop(void) +static void xen_smp_send_stop(void) { smp_call_function(stop_self, NULL, 0); } -void xen_smp_send_reschedule(int cpu) +static void xen_smp_send_reschedule(int cpu) { xen_send_IPI_one(cpu, XEN_RESCHEDULE_VECTOR); } @@ -355,7 +355,7 @@ static void xen_send_IPI_mask(cpumask_t mask, enum ipi_vector vector) xen_send_IPI_one(cpu, vector); } -void xen_smp_send_call_function_ipi(cpumask_t mask) +static void xen_smp_send_call_function_ipi(cpumask_t mask) { int cpu; @@ -370,7 +370,7 @@ void xen_smp_send_call_function_ipi(cpumask_t mask) } } -void xen_smp_send_call_function_single_ipi(int cpu) +static void xen_smp_send_call_function_single_ipi(int cpu) { xen_send_IPI_mask(cpumask_of_cpu(cpu), XEN_CALL_FUNCTION_SINGLE_VECTOR); } @@ -394,3 +394,21 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id) return IRQ_HANDLED; } + +static const struct smp_ops xen_smp_ops __initdata = { + .smp_prepare_boot_cpu = xen_smp_prepare_boot_cpu, + .smp_prepare_cpus = xen_smp_prepare_cpus, + .cpu_up = xen_cpu_up, + .smp_cpus_done = xen_smp_cpus_done, + + .smp_send_stop = xen_smp_send_stop, + .smp_send_reschedule = xen_smp_send_reschedule, + + .send_call_func_ipi = xen_smp_send_call_function_ipi, + .send_call_func_single_ipi = xen_smp_send_call_function_single_ipi, +}; + +void __init xen_smp_init(void) +{ + smp_ops = xen_smp_ops; +} diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 77354d20425..81a779fc9b2 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -47,17 +47,14 @@ void xen_mark_init_mm_pinned(void); void __init xen_fill_possible_map(void); void __init xen_setup_vcpu_info_placement(void); -void xen_smp_prepare_boot_cpu(void); -void xen_smp_prepare_cpus(unsigned int max_cpus); -int xen_cpu_up(unsigned int cpu); -void xen_smp_cpus_done(unsigned int max_cpus); -void xen_smp_send_stop(void); -void xen_smp_send_reschedule(int cpu); -void xen_smp_send_call_function_ipi(cpumask_t mask); -void xen_smp_send_call_function_single_ipi(int cpu); +#ifdef CONFIG_SMP +void xen_smp_init(void); extern cpumask_t xen_cpu_initialized_map; +#else +static inline void xen_smp_init(void) {} +#endif /* Declare an asm function, along with symbols needed to make it -- cgit v1.2.3 From 5b09b2876ed1a8e34a0da8f069575fc6174e2077 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:42 -0700 Subject: x86_64: add workaround for no %gs-based percpu As a stopgap until Mike Travis's x86-64 gs-based percpu patches are ready, provide workaround functions for x86_read/write_percpu for Xen's use. Specifically, this means that we can't really make use of vcpu placement, because we can't use a single gs-based memory access to get to vcpu fields. So disable all that for now. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/head64.c | 11 ++++++++--- arch/x86/xen/enlighten.c | 5 +++++ 2 files changed, 13 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c index c9781982914..1b318e903bf 100644 --- a/arch/x86/kernel/head64.c +++ b/arch/x86/kernel/head64.c @@ -39,6 +39,13 @@ static struct x8664_pda *__cpu_pda[NR_CPUS] __initdata; static struct x8664_pda *__cpu_pda[NR_CPUS] __read_mostly; #endif +void __init x86_64_init_pda(void) +{ + _cpu_pda = __cpu_pda; + cpu_pda(0) = &_boot_cpu_pda; + pda_init(0); +} + static void __init zap_identity_mappings(void) { pgd_t *pgd = pgd_offset_k(0UL); @@ -102,9 +109,7 @@ void __init x86_64_start_kernel(char * real_mode_data) early_printk("Kernel alive\n"); - _cpu_pda = __cpu_pda; - cpu_pda(0) = &_boot_cpu_pda; - pda_init(0); + x86_64_init_pda(); early_printk("Kernel really alive\n"); diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index a85f447b8d0..f3f11acf785 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -971,6 +971,7 @@ void xen_setup_vcpu_info_placement(void) /* xen_vcpu_setup managed to place the vcpu_info within the percpu area for all cpus, so make use of it */ +#ifdef CONFIG_X86_32 if (have_vcpu_info_placement) { printk(KERN_INFO "Xen: using vcpu_info placement\n"); @@ -980,6 +981,7 @@ void xen_setup_vcpu_info_placement(void) pv_irq_ops.irq_enable = xen_irq_enable_direct; pv_mmu_ops.read_cr2 = xen_read_cr2_direct; } +#endif } static unsigned xen_patch(u8 type, u16 clobbers, void *insnbuf, @@ -1000,10 +1002,12 @@ static unsigned xen_patch(u8 type, u16 clobbers, void *insnbuf, goto patch_site switch (type) { +#ifdef CONFIG_X86_32 SITE(pv_irq_ops, irq_enable); SITE(pv_irq_ops, irq_disable); SITE(pv_irq_ops, save_fl); SITE(pv_irq_ops, restore_fl); +#endif /* CONFIG_X86_32 */ #undef SITE patch_site: @@ -1323,6 +1327,7 @@ asmlinkage void __init xen_start_kernel(void) #ifdef CONFIG_X86_64 /* Disable until direct per-cpu data access. */ have_vcpu_info_placement = 0; + x86_64_init_pda(); #endif xen_smp_init(); -- cgit v1.2.3 From c7b75947f89d45493562ede6d9ee7311dfa5c4ce Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:43 -0700 Subject: xen64: smp.c compile hacking A number of random changes to make xen/smp.c compile in 64-bit mode. Signed-off-by: Jeremy Fitzhardinge a Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 7 +--- arch/x86/xen/smp.c | 98 +++++++++++++++++++++++++++++--------------------- arch/x86/xen/xen-ops.h | 2 -- 3 files changed, 58 insertions(+), 49 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index e0a39595bde..f52f3855fb6 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -98,7 +98,7 @@ void xen_enable_sysenter(void) /* Mask events on entry, even though they get enabled immediately */ static struct callback_register sysenter = { .type = CALLBACKTYPE_sysenter, - .address = { __KERNEL_CS, (unsigned long)xen_sysenter_target }, + .address = XEN_CALLBACK(__KERNEL_CS, xen_sysenter_target), .flags = CALLBACKF_mask_events, }; @@ -143,11 +143,6 @@ void __init xen_arch_setup(void) pm_idle = xen_idle; -#ifdef CONFIG_SMP - /* fill cpus_possible with all available cpus */ - xen_fill_possible_map(); -#endif - paravirt_disable_iospace(); fiddle_vdso(); diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 91fae8ff756..800bb2191e2 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -66,13 +66,21 @@ static __cpuinit void cpu_bringup_and_idle(void) int cpu = smp_processor_id(); cpu_init(); + preempt_disable(); + xen_enable_sysenter(); - preempt_disable(); - per_cpu(cpu_state, cpu) = CPU_ONLINE; + cpu = smp_processor_id(); + smp_store_cpu_info(cpu); + cpu_data(cpu).x86_max_cores = 1; + set_cpu_sibling_map(cpu); xen_setup_cpu_clockevents(); + cpu_set(cpu, cpu_online_map); + x86_write_percpu(cpu_state, CPU_ONLINE); + wmb(); + /* We can take interrupts now: we're officially "up". */ local_irq_enable(); @@ -141,7 +149,7 @@ static int xen_smp_intr_init(unsigned int cpu) return rc; } -void __init xen_fill_possible_map(void) +static void __init xen_fill_possible_map(void) { int i, rc; @@ -154,24 +162,12 @@ void __init xen_fill_possible_map(void) static void __init xen_smp_prepare_boot_cpu(void) { - int cpu; - BUG_ON(smp_processor_id() != 0); native_smp_prepare_boot_cpu(); /* We've switched to the "real" per-cpu gdt, so make sure the old memory can be recycled */ - make_lowmem_page_readwrite(&per_cpu__gdt_page); - - for_each_possible_cpu(cpu) { - cpus_clear(per_cpu(cpu_sibling_map, cpu)); - /* - * cpu_core_map lives in a per cpu area that is cleared - * when the per cpu array is allocated. - * - * cpus_clear(per_cpu(cpu_core_map, cpu)); - */ - } + make_lowmem_page_readwrite(&per_cpu_var(gdt_page)); xen_setup_vcpu_info_placement(); } @@ -180,17 +176,8 @@ static void __init xen_smp_prepare_cpus(unsigned int max_cpus) { unsigned cpu; - for_each_possible_cpu(cpu) { - cpus_clear(per_cpu(cpu_sibling_map, cpu)); - /* - * cpu_core_ map will be zeroed when the per - * cpu area is allocated. - * - * cpus_clear(per_cpu(cpu_core_map, cpu)); - */ - } - smp_store_cpu_info(0); + cpu_data(0).x86_max_cores = 1; set_cpu_sibling_map(0); if (xen_smp_intr_init(0)) @@ -225,7 +212,7 @@ static __cpuinit int cpu_initialize_context(unsigned int cpu, struct task_struct *idle) { struct vcpu_guest_context *ctxt; - struct gdt_page *gdt = &per_cpu(gdt_page, cpu); + struct desc_struct *gdt; if (cpu_test_and_set(cpu, xen_cpu_initialized_map)) return 0; @@ -234,12 +221,15 @@ cpu_initialize_context(unsigned int cpu, struct task_struct *idle) if (ctxt == NULL) return -ENOMEM; + gdt = get_cpu_gdt_table(cpu); + ctxt->flags = VGCF_IN_KERNEL; ctxt->user_regs.ds = __USER_DS; ctxt->user_regs.es = __USER_DS; - ctxt->user_regs.fs = __KERNEL_PERCPU; - ctxt->user_regs.gs = 0; ctxt->user_regs.ss = __KERNEL_DS; +#ifdef CONFIG_X86_32 + ctxt->user_regs.fs = __KERNEL_PERCPU; +#endif ctxt->user_regs.eip = (unsigned long)cpu_bringup_and_idle; ctxt->user_regs.eflags = 0x1000; /* IOPL_RING1 */ @@ -249,11 +239,11 @@ cpu_initialize_context(unsigned int cpu, struct task_struct *idle) ctxt->ldt_ents = 0; - BUG_ON((unsigned long)gdt->gdt & ~PAGE_MASK); - make_lowmem_page_readonly(gdt->gdt); + BUG_ON((unsigned long)gdt & ~PAGE_MASK); + make_lowmem_page_readonly(gdt); - ctxt->gdt_frames[0] = virt_to_mfn(gdt->gdt); - ctxt->gdt_ents = ARRAY_SIZE(gdt->gdt); + ctxt->gdt_frames[0] = virt_to_mfn(gdt); + ctxt->gdt_ents = GDT_ENTRIES; ctxt->user_regs.cs = __KERNEL_CS; ctxt->user_regs.esp = idle->thread.sp0 - sizeof(struct pt_regs); @@ -261,9 +251,11 @@ cpu_initialize_context(unsigned int cpu, struct task_struct *idle) ctxt->kernel_ss = __KERNEL_DS; ctxt->kernel_sp = idle->thread.sp0; +#ifdef CONFIG_X86_32 ctxt->event_callback_cs = __KERNEL_CS; - ctxt->event_callback_eip = (unsigned long)xen_hypervisor_callback; ctxt->failsafe_callback_cs = __KERNEL_CS; +#endif + ctxt->event_callback_eip = (unsigned long)xen_hypervisor_callback; ctxt->failsafe_callback_eip = (unsigned long)xen_failsafe_callback; per_cpu(xen_cr3, cpu) = __pa(swapper_pg_dir); @@ -287,11 +279,28 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) return rc; #endif +#ifdef CONFIG_X86_64 + /* Allocate node local memory for AP pdas */ + WARN_ON(cpu == 0); + if (cpu > 0) { + rc = get_local_pda(cpu); + if (rc) + return rc; + } +#endif + +#ifdef CONFIG_X86_32 init_gdt(cpu); per_cpu(current_task, cpu) = idle; irq_ctx_init(cpu); +#else + cpu_pda(cpu)->pcurrent = idle; + clear_tsk_thread_flag(idle, TIF_FORK); +#endif xen_setup_timer(cpu); + per_cpu(cpu_state, cpu) = CPU_UP_PREPARE; + /* make sure interrupts start blocked */ per_cpu(xen_vcpu, cpu)->evtchn_upcall_mask = 1; @@ -306,16 +315,14 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) if (rc) return rc; - smp_store_cpu_info(cpu); - set_cpu_sibling_map(cpu); - /* This must be done before setting cpu_online_map */ - wmb(); - - cpu_set(cpu, cpu_online_map); - rc = HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL); BUG_ON(rc); + while(per_cpu(cpu_state, cpu) != CPU_ONLINE) { + HYPERVISOR_sched_op(SCHEDOP_yield, 0); + barrier(); + } + return 0; } @@ -379,7 +386,11 @@ static irqreturn_t xen_call_function_interrupt(int irq, void *dev_id) { irq_enter(); generic_smp_call_function_interrupt(); +#ifdef CONFIG_X86_32 __get_cpu_var(irq_stat).irq_call_count++; +#else + add_pda(irq_call_count, 1); +#endif irq_exit(); return IRQ_HANDLED; @@ -389,7 +400,11 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id) { irq_enter(); generic_smp_call_function_single_interrupt(); +#ifdef CONFIG_X86_32 __get_cpu_var(irq_stat).irq_call_count++; +#else + add_pda(irq_call_count, 1); +#endif irq_exit(); return IRQ_HANDLED; @@ -411,4 +426,5 @@ static const struct smp_ops xen_smp_ops __initdata = { void __init xen_smp_init(void) { smp_ops = xen_smp_ops; + xen_fill_possible_map(); } diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index 81a779fc9b2..aca4a7803e2 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -44,8 +44,6 @@ bool xen_vcpu_stolen(int vcpu); void xen_mark_init_mm_pinned(void); -void __init xen_fill_possible_map(void); - void __init xen_setup_vcpu_info_placement(void); #ifdef CONFIG_SMP -- cgit v1.2.3 From 8c5e5ac32fe08793246709fbb94c055ec76a7c0e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:44 -0700 Subject: xen64: add xen-head code to head_64.S Add the Xen entrypoint and ELF notes to head_64.S. Adapts xen-head.S to compile either 32-bit or 64-bit. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/asm-offsets_64.c | 3 +++ arch/x86/kernel/head_64.S | 1 + arch/x86/xen/xen-head.S | 15 +++++++++++++-- 3 files changed, 17 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index bacf5deeec2..0f7e1f09aa0 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -131,5 +131,8 @@ int main(void) OFFSET(BP_loadflags, boot_params, hdr.loadflags); OFFSET(BP_hardware_subarch, boot_params, hdr.hardware_subarch); OFFSET(BP_version, boot_params, hdr.version); + + BLANK(); + DEFINE(PAGE_SIZE_asm, PAGE_SIZE); return 0; } diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 4b6bda21837..2240f823676 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -401,6 +401,7 @@ ENTRY(phys_base) /* This must match the first entry in level2_kernel_pgt */ .quad 0x0000000000000000 +#include "../../x86/xen/xen-head.S" .section .bss, "aw", @nobits .align L1_CACHE_BYTES diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S index a9cac9dc04b..63d49a523ed 100644 --- a/arch/x86/xen/xen-head.S +++ b/arch/x86/xen/xen-head.S @@ -8,15 +8,21 @@ #include #include +#include #include #include __INIT ENTRY(startup_xen) - movl %esi,xen_start_info cld - movl $(init_thread_union+THREAD_SIZE),%esp +#ifdef CONFIG_X86_32 + mov %esi,xen_start_info + mov $init_thread_union+THREAD_SIZE,%esp +#else + mov %rsi,xen_start_info + mov $init_thread_union+THREAD_SIZE,%rsp +#endif jmp xen_start_kernel __FINIT @@ -30,7 +36,11 @@ ENTRY(hypercall_page) ELFNOTE(Xen, XEN_ELFNOTE_GUEST_OS, .asciz "linux") ELFNOTE(Xen, XEN_ELFNOTE_GUEST_VERSION, .asciz "2.6") ELFNOTE(Xen, XEN_ELFNOTE_XEN_VERSION, .asciz "xen-3.0") +#ifdef CONFIG_X86_32 ELFNOTE(Xen, XEN_ELFNOTE_VIRT_BASE, _ASM_PTR __PAGE_OFFSET) +#else + ELFNOTE(Xen, XEN_ELFNOTE_VIRT_BASE, _ASM_PTR __START_KERNEL_map) +#endif ELFNOTE(Xen, XEN_ELFNOTE_ENTRY, _ASM_PTR startup_xen) ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, _ASM_PTR hypercall_page) ELFNOTE(Xen, XEN_ELFNOTE_FEATURES, .asciz "!writable_page_tables|pae_pgdir_above_4gb") @@ -40,5 +50,6 @@ ENTRY(hypercall_page) .quad _PAGE_PRESENT; .quad _PAGE_PRESENT) ELFNOTE(Xen, XEN_ELFNOTE_SUSPEND_CANCEL, .long 1) ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW, _ASM_PTR __HYPERVISOR_VIRT_START) + ELFNOTE(Xen, XEN_ELFNOTE_PADDR_OFFSET, _ASM_PTR 0) #endif /*CONFIG_XEN */ -- cgit v1.2.3 From 555cf2b5805a213ba262a2830c4d22ad635a249e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:45 -0700 Subject: xen64: add asm-offsets Add Xen vcpu_info offsets to asm-offsets_64. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/asm-offsets_64.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/asm-offsets_64.c b/arch/x86/kernel/asm-offsets_64.c index 0f7e1f09aa0..aa89387006f 100644 --- a/arch/x86/kernel/asm-offsets_64.c +++ b/arch/x86/kernel/asm-offsets_64.c @@ -18,6 +18,8 @@ #include #include +#include + #define __NO_STUBS 1 #undef __SYSCALL #undef _ASM_X86_64_UNISTD_H_ @@ -134,5 +136,11 @@ int main(void) BLANK(); DEFINE(PAGE_SIZE_asm, PAGE_SIZE); +#ifdef CONFIG_XEN + BLANK(); + OFFSET(XEN_vcpu_info_mask, vcpu_info, evtchn_upcall_mask); + OFFSET(XEN_vcpu_info_pending, vcpu_info, evtchn_upcall_pending); +#undef ENTRY +#endif return 0; } -- cgit v1.2.3 From cdacc1278b12d929f9a053c245ff3d16eb7af9f8 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:46 -0700 Subject: xen64: add 64-bit assembler Split xen-asm into 32- and 64-bit files, and implement the 64-bit variants. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/Makefile | 2 +- arch/x86/xen/xen-asm.S | 305 ---------------------------------------------- arch/x86/xen/xen-asm_32.S | 305 ++++++++++++++++++++++++++++++++++++++++++++++ arch/x86/xen/xen-asm_64.S | 141 +++++++++++++++++++++ 4 files changed, 447 insertions(+), 306 deletions(-) delete mode 100644 arch/x86/xen/xen-asm.S create mode 100644 arch/x86/xen/xen-asm_32.S create mode 100644 arch/x86/xen/xen-asm_64.S (limited to 'arch') diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile index 2ba2d164913..59c1e539aed 100644 --- a/arch/x86/xen/Makefile +++ b/arch/x86/xen/Makefile @@ -1,4 +1,4 @@ obj-y := enlighten.o setup.o multicalls.o mmu.o \ - time.o xen-asm.o grant-table.o suspend.o + time.o xen-asm_$(BITS).o grant-table.o suspend.o obj-$(CONFIG_SMP) += smp.o diff --git a/arch/x86/xen/xen-asm.S b/arch/x86/xen/xen-asm.S deleted file mode 100644 index 2497a30f41d..00000000000 --- a/arch/x86/xen/xen-asm.S +++ /dev/null @@ -1,305 +0,0 @@ -/* - Asm versions of Xen pv-ops, suitable for either direct use or inlining. - The inline versions are the same as the direct-use versions, with the - pre- and post-amble chopped off. - - This code is encoded for size rather than absolute efficiency, - with a view to being able to inline as much as possible. - - We only bother with direct forms (ie, vcpu in pda) of the operations - here; the indirect forms are better handled in C, since they're - generally too large to inline anyway. - */ - -#include - -#include -#include -#include -#include -#include - -#include - -#define RELOC(x, v) .globl x##_reloc; x##_reloc=v -#define ENDPATCH(x) .globl x##_end; x##_end=. - -/* Pseudo-flag used for virtual NMI, which we don't implement yet */ -#define XEN_EFLAGS_NMI 0x80000000 - -/* - Enable events. This clears the event mask and tests the pending - event status with one and operation. If there are pending - events, then enter the hypervisor to get them handled. - */ -ENTRY(xen_irq_enable_direct) - /* Unmask events */ - movb $0, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask - - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ - - /* Test for pending */ - testb $0xff, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_pending - jz 1f - -2: call check_events -1: -ENDPATCH(xen_irq_enable_direct) - ret - ENDPROC(xen_irq_enable_direct) - RELOC(xen_irq_enable_direct, 2b+1) - - -/* - Disabling events is simply a matter of making the event mask - non-zero. - */ -ENTRY(xen_irq_disable_direct) - movb $1, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask -ENDPATCH(xen_irq_disable_direct) - ret - ENDPROC(xen_irq_disable_direct) - RELOC(xen_irq_disable_direct, 0) - -/* - (xen_)save_fl is used to get the current interrupt enable status. - Callers expect the status to be in X86_EFLAGS_IF, and other bits - may be set in the return value. We take advantage of this by - making sure that X86_EFLAGS_IF has the right value (and other bits - in that byte are 0), but other bits in the return value are - undefined. We need to toggle the state of the bit, because - Xen and x86 use opposite senses (mask vs enable). - */ -ENTRY(xen_save_fl_direct) - testb $0xff, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask - setz %ah - addb %ah,%ah -ENDPATCH(xen_save_fl_direct) - ret - ENDPROC(xen_save_fl_direct) - RELOC(xen_save_fl_direct, 0) - - -/* - In principle the caller should be passing us a value return - from xen_save_fl_direct, but for robustness sake we test only - the X86_EFLAGS_IF flag rather than the whole byte. After - setting the interrupt mask state, it checks for unmasked - pending events and enters the hypervisor to get them delivered - if so. - */ -ENTRY(xen_restore_fl_direct) - testb $X86_EFLAGS_IF>>8, %ah - setz PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask - /* Preempt here doesn't matter because that will deal with - any pending interrupts. The pending check may end up being - run on the wrong CPU, but that doesn't hurt. */ - - /* check for unmasked and pending */ - cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_pending - jz 1f -2: call check_events -1: -ENDPATCH(xen_restore_fl_direct) - ret - ENDPROC(xen_restore_fl_direct) - RELOC(xen_restore_fl_direct, 2b+1) - -/* - We can't use sysexit directly, because we're not running in ring0. - But we can easily fake it up using iret. Assuming xen_sysexit - is jumped to with a standard stack frame, we can just strip it - back to a standard iret frame and use iret. - */ -ENTRY(xen_sysexit) - movl PT_EAX(%esp), %eax /* Shouldn't be necessary? */ - orl $X86_EFLAGS_IF, PT_EFLAGS(%esp) - lea PT_EIP(%esp), %esp - - jmp xen_iret -ENDPROC(xen_sysexit) - -/* - This is run where a normal iret would be run, with the same stack setup: - 8: eflags - 4: cs - esp-> 0: eip - - This attempts to make sure that any pending events are dealt - with on return to usermode, but there is a small window in - which an event can happen just before entering usermode. If - the nested interrupt ends up setting one of the TIF_WORK_MASK - pending work flags, they will not be tested again before - returning to usermode. This means that a process can end up - with pending work, which will be unprocessed until the process - enters and leaves the kernel again, which could be an - unbounded amount of time. This means that a pending signal or - reschedule event could be indefinitely delayed. - - The fix is to notice a nested interrupt in the critical - window, and if one occurs, then fold the nested interrupt into - the current interrupt stack frame, and re-process it - iteratively rather than recursively. This means that it will - exit via the normal path, and all pending work will be dealt - with appropriately. - - Because the nested interrupt handler needs to deal with the - current stack state in whatever form its in, we keep things - simple by only using a single register which is pushed/popped - on the stack. - */ -ENTRY(xen_iret) - /* test eflags for special cases */ - testl $(X86_EFLAGS_VM | XEN_EFLAGS_NMI), 8(%esp) - jnz hyper_iret - - push %eax - ESP_OFFSET=4 # bytes pushed onto stack - - /* Store vcpu_info pointer for easy access. Do it this - way to avoid having to reload %fs */ -#ifdef CONFIG_SMP - GET_THREAD_INFO(%eax) - movl TI_cpu(%eax),%eax - movl __per_cpu_offset(,%eax,4),%eax - mov per_cpu__xen_vcpu(%eax),%eax -#else - movl per_cpu__xen_vcpu, %eax -#endif - - /* check IF state we're restoring */ - testb $X86_EFLAGS_IF>>8, 8+1+ESP_OFFSET(%esp) - - /* Maybe enable events. Once this happens we could get a - recursive event, so the critical region starts immediately - afterwards. However, if that happens we don't end up - resuming the code, so we don't have to be worried about - being preempted to another CPU. */ - setz XEN_vcpu_info_mask(%eax) -xen_iret_start_crit: - - /* check for unmasked and pending */ - cmpw $0x0001, XEN_vcpu_info_pending(%eax) - - /* If there's something pending, mask events again so we - can jump back into xen_hypervisor_callback */ - sete XEN_vcpu_info_mask(%eax) - - popl %eax - - /* From this point on the registers are restored and the stack - updated, so we don't need to worry about it if we're preempted */ -iret_restore_end: - - /* Jump to hypervisor_callback after fixing up the stack. - Events are masked, so jumping out of the critical - region is OK. */ - je xen_hypervisor_callback - -1: iret -xen_iret_end_crit: -.section __ex_table,"a" - .align 4 - .long 1b,iret_exc -.previous - -hyper_iret: - /* put this out of line since its very rarely used */ - jmp hypercall_page + __HYPERVISOR_iret * 32 - - .globl xen_iret_start_crit, xen_iret_end_crit - -/* - This is called by xen_hypervisor_callback in entry.S when it sees - that the EIP at the time of interrupt was between xen_iret_start_crit - and xen_iret_end_crit. We're passed the EIP in %eax so we can do - a more refined determination of what to do. - - The stack format at this point is: - ---------------- - ss : (ss/esp may be present if we came from usermode) - esp : - eflags } outer exception info - cs } - eip } - ---------------- <- edi (copy dest) - eax : outer eax if it hasn't been restored - ---------------- - eflags } nested exception info - cs } (no ss/esp because we're nested - eip } from the same ring) - orig_eax }<- esi (copy src) - - - - - - - - - - fs } - es } - ds } SAVE_ALL state - eax } - : : - ebx }<- esp - ---------------- - - In order to deliver the nested exception properly, we need to shift - everything from the return addr up to the error code so it - sits just under the outer exception info. This means that when we - handle the exception, we do it in the context of the outer exception - rather than starting a new one. - - The only caveat is that if the outer eax hasn't been - restored yet (ie, it's still on stack), we need to insert - its value into the SAVE_ALL state before going on, since - it's usermode state which we eventually need to restore. - */ -ENTRY(xen_iret_crit_fixup) - /* - Paranoia: Make sure we're really coming from kernel space. - One could imagine a case where userspace jumps into the - critical range address, but just before the CPU delivers a GP, - it decides to deliver an interrupt instead. Unlikely? - Definitely. Easy to avoid? Yes. The Intel documents - explicitly say that the reported EIP for a bad jump is the - jump instruction itself, not the destination, but some virtual - environments get this wrong. - */ - movl PT_CS(%esp), %ecx - andl $SEGMENT_RPL_MASK, %ecx - cmpl $USER_RPL, %ecx - je 2f - - lea PT_ORIG_EAX(%esp), %esi - lea PT_EFLAGS(%esp), %edi - - /* If eip is before iret_restore_end then stack - hasn't been restored yet. */ - cmp $iret_restore_end, %eax - jae 1f - - movl 0+4(%edi),%eax /* copy EAX (just above top of frame) */ - movl %eax, PT_EAX(%esp) - - lea ESP_OFFSET(%edi),%edi /* move dest up over saved regs */ - - /* set up the copy */ -1: std - mov $PT_EIP / 4, %ecx /* saved regs up to orig_eax */ - rep movsl - cld - - lea 4(%edi),%esp /* point esp to new frame */ -2: jmp xen_do_upcall - - -/* - Force an event check by making a hypercall, - but preserve regs before making the call. - */ -check_events: - push %eax - push %ecx - push %edx - call force_evtchn_callback - pop %edx - pop %ecx - pop %eax - ret diff --git a/arch/x86/xen/xen-asm_32.S b/arch/x86/xen/xen-asm_32.S new file mode 100644 index 00000000000..2497a30f41d --- /dev/null +++ b/arch/x86/xen/xen-asm_32.S @@ -0,0 +1,305 @@ +/* + Asm versions of Xen pv-ops, suitable for either direct use or inlining. + The inline versions are the same as the direct-use versions, with the + pre- and post-amble chopped off. + + This code is encoded for size rather than absolute efficiency, + with a view to being able to inline as much as possible. + + We only bother with direct forms (ie, vcpu in pda) of the operations + here; the indirect forms are better handled in C, since they're + generally too large to inline anyway. + */ + +#include + +#include +#include +#include +#include +#include + +#include + +#define RELOC(x, v) .globl x##_reloc; x##_reloc=v +#define ENDPATCH(x) .globl x##_end; x##_end=. + +/* Pseudo-flag used for virtual NMI, which we don't implement yet */ +#define XEN_EFLAGS_NMI 0x80000000 + +/* + Enable events. This clears the event mask and tests the pending + event status with one and operation. If there are pending + events, then enter the hypervisor to get them handled. + */ +ENTRY(xen_irq_enable_direct) + /* Unmask events */ + movb $0, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask + + /* Preempt here doesn't matter because that will deal with + any pending interrupts. The pending check may end up being + run on the wrong CPU, but that doesn't hurt. */ + + /* Test for pending */ + testb $0xff, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_pending + jz 1f + +2: call check_events +1: +ENDPATCH(xen_irq_enable_direct) + ret + ENDPROC(xen_irq_enable_direct) + RELOC(xen_irq_enable_direct, 2b+1) + + +/* + Disabling events is simply a matter of making the event mask + non-zero. + */ +ENTRY(xen_irq_disable_direct) + movb $1, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask +ENDPATCH(xen_irq_disable_direct) + ret + ENDPROC(xen_irq_disable_direct) + RELOC(xen_irq_disable_direct, 0) + +/* + (xen_)save_fl is used to get the current interrupt enable status. + Callers expect the status to be in X86_EFLAGS_IF, and other bits + may be set in the return value. We take advantage of this by + making sure that X86_EFLAGS_IF has the right value (and other bits + in that byte are 0), but other bits in the return value are + undefined. We need to toggle the state of the bit, because + Xen and x86 use opposite senses (mask vs enable). + */ +ENTRY(xen_save_fl_direct) + testb $0xff, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask + setz %ah + addb %ah,%ah +ENDPATCH(xen_save_fl_direct) + ret + ENDPROC(xen_save_fl_direct) + RELOC(xen_save_fl_direct, 0) + + +/* + In principle the caller should be passing us a value return + from xen_save_fl_direct, but for robustness sake we test only + the X86_EFLAGS_IF flag rather than the whole byte. After + setting the interrupt mask state, it checks for unmasked + pending events and enters the hypervisor to get them delivered + if so. + */ +ENTRY(xen_restore_fl_direct) + testb $X86_EFLAGS_IF>>8, %ah + setz PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_mask + /* Preempt here doesn't matter because that will deal with + any pending interrupts. The pending check may end up being + run on the wrong CPU, but that doesn't hurt. */ + + /* check for unmasked and pending */ + cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info)+XEN_vcpu_info_pending + jz 1f +2: call check_events +1: +ENDPATCH(xen_restore_fl_direct) + ret + ENDPROC(xen_restore_fl_direct) + RELOC(xen_restore_fl_direct, 2b+1) + +/* + We can't use sysexit directly, because we're not running in ring0. + But we can easily fake it up using iret. Assuming xen_sysexit + is jumped to with a standard stack frame, we can just strip it + back to a standard iret frame and use iret. + */ +ENTRY(xen_sysexit) + movl PT_EAX(%esp), %eax /* Shouldn't be necessary? */ + orl $X86_EFLAGS_IF, PT_EFLAGS(%esp) + lea PT_EIP(%esp), %esp + + jmp xen_iret +ENDPROC(xen_sysexit) + +/* + This is run where a normal iret would be run, with the same stack setup: + 8: eflags + 4: cs + esp-> 0: eip + + This attempts to make sure that any pending events are dealt + with on return to usermode, but there is a small window in + which an event can happen just before entering usermode. If + the nested interrupt ends up setting one of the TIF_WORK_MASK + pending work flags, they will not be tested again before + returning to usermode. This means that a process can end up + with pending work, which will be unprocessed until the process + enters and leaves the kernel again, which could be an + unbounded amount of time. This means that a pending signal or + reschedule event could be indefinitely delayed. + + The fix is to notice a nested interrupt in the critical + window, and if one occurs, then fold the nested interrupt into + the current interrupt stack frame, and re-process it + iteratively rather than recursively. This means that it will + exit via the normal path, and all pending work will be dealt + with appropriately. + + Because the nested interrupt handler needs to deal with the + current stack state in whatever form its in, we keep things + simple by only using a single register which is pushed/popped + on the stack. + */ +ENTRY(xen_iret) + /* test eflags for special cases */ + testl $(X86_EFLAGS_VM | XEN_EFLAGS_NMI), 8(%esp) + jnz hyper_iret + + push %eax + ESP_OFFSET=4 # bytes pushed onto stack + + /* Store vcpu_info pointer for easy access. Do it this + way to avoid having to reload %fs */ +#ifdef CONFIG_SMP + GET_THREAD_INFO(%eax) + movl TI_cpu(%eax),%eax + movl __per_cpu_offset(,%eax,4),%eax + mov per_cpu__xen_vcpu(%eax),%eax +#else + movl per_cpu__xen_vcpu, %eax +#endif + + /* check IF state we're restoring */ + testb $X86_EFLAGS_IF>>8, 8+1+ESP_OFFSET(%esp) + + /* Maybe enable events. Once this happens we could get a + recursive event, so the critical region starts immediately + afterwards. However, if that happens we don't end up + resuming the code, so we don't have to be worried about + being preempted to another CPU. */ + setz XEN_vcpu_info_mask(%eax) +xen_iret_start_crit: + + /* check for unmasked and pending */ + cmpw $0x0001, XEN_vcpu_info_pending(%eax) + + /* If there's something pending, mask events again so we + can jump back into xen_hypervisor_callback */ + sete XEN_vcpu_info_mask(%eax) + + popl %eax + + /* From this point on the registers are restored and the stack + updated, so we don't need to worry about it if we're preempted */ +iret_restore_end: + + /* Jump to hypervisor_callback after fixing up the stack. + Events are masked, so jumping out of the critical + region is OK. */ + je xen_hypervisor_callback + +1: iret +xen_iret_end_crit: +.section __ex_table,"a" + .align 4 + .long 1b,iret_exc +.previous + +hyper_iret: + /* put this out of line since its very rarely used */ + jmp hypercall_page + __HYPERVISOR_iret * 32 + + .globl xen_iret_start_crit, xen_iret_end_crit + +/* + This is called by xen_hypervisor_callback in entry.S when it sees + that the EIP at the time of interrupt was between xen_iret_start_crit + and xen_iret_end_crit. We're passed the EIP in %eax so we can do + a more refined determination of what to do. + + The stack format at this point is: + ---------------- + ss : (ss/esp may be present if we came from usermode) + esp : + eflags } outer exception info + cs } + eip } + ---------------- <- edi (copy dest) + eax : outer eax if it hasn't been restored + ---------------- + eflags } nested exception info + cs } (no ss/esp because we're nested + eip } from the same ring) + orig_eax }<- esi (copy src) + - - - - - - - - + fs } + es } + ds } SAVE_ALL state + eax } + : : + ebx }<- esp + ---------------- + + In order to deliver the nested exception properly, we need to shift + everything from the return addr up to the error code so it + sits just under the outer exception info. This means that when we + handle the exception, we do it in the context of the outer exception + rather than starting a new one. + + The only caveat is that if the outer eax hasn't been + restored yet (ie, it's still on stack), we need to insert + its value into the SAVE_ALL state before going on, since + it's usermode state which we eventually need to restore. + */ +ENTRY(xen_iret_crit_fixup) + /* + Paranoia: Make sure we're really coming from kernel space. + One could imagine a case where userspace jumps into the + critical range address, but just before the CPU delivers a GP, + it decides to deliver an interrupt instead. Unlikely? + Definitely. Easy to avoid? Yes. The Intel documents + explicitly say that the reported EIP for a bad jump is the + jump instruction itself, not the destination, but some virtual + environments get this wrong. + */ + movl PT_CS(%esp), %ecx + andl $SEGMENT_RPL_MASK, %ecx + cmpl $USER_RPL, %ecx + je 2f + + lea PT_ORIG_EAX(%esp), %esi + lea PT_EFLAGS(%esp), %edi + + /* If eip is before iret_restore_end then stack + hasn't been restored yet. */ + cmp $iret_restore_end, %eax + jae 1f + + movl 0+4(%edi),%eax /* copy EAX (just above top of frame) */ + movl %eax, PT_EAX(%esp) + + lea ESP_OFFSET(%edi),%edi /* move dest up over saved regs */ + + /* set up the copy */ +1: std + mov $PT_EIP / 4, %ecx /* saved regs up to orig_eax */ + rep movsl + cld + + lea 4(%edi),%esp /* point esp to new frame */ +2: jmp xen_do_upcall + + +/* + Force an event check by making a hypercall, + but preserve regs before making the call. + */ +check_events: + push %eax + push %ecx + push %edx + call force_evtchn_callback + pop %edx + pop %ecx + pop %eax + ret diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S new file mode 100644 index 00000000000..4ec10827370 --- /dev/null +++ b/arch/x86/xen/xen-asm_64.S @@ -0,0 +1,141 @@ +/* + Asm versions of Xen pv-ops, suitable for either direct use or inlining. + The inline versions are the same as the direct-use versions, with the + pre- and post-amble chopped off. + + This code is encoded for size rather than absolute efficiency, + with a view to being able to inline as much as possible. + + We only bother with direct forms (ie, vcpu in pda) of the operations + here; the indirect forms are better handled in C, since they're + generally too large to inline anyway. + */ + +#include + +#include +#include + +#include + +#define RELOC(x, v) .globl x##_reloc; x##_reloc=v +#define ENDPATCH(x) .globl x##_end; x##_end=. + +/* Pseudo-flag used for virtual NMI, which we don't implement yet */ +#define XEN_EFLAGS_NMI 0x80000000 + +#if 0 +#include + +/* + Enable events. This clears the event mask and tests the pending + event status with one and operation. If there are pending + events, then enter the hypervisor to get them handled. + */ +ENTRY(xen_irq_enable_direct) + /* Unmask events */ + movb $0, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + + /* Preempt here doesn't matter because that will deal with + any pending interrupts. The pending check may end up being + run on the wrong CPU, but that doesn't hurt. */ + + /* Test for pending */ + testb $0xff, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_pending) + jz 1f + +2: call check_events +1: +ENDPATCH(xen_irq_enable_direct) + ret + ENDPROC(xen_irq_enable_direct) + RELOC(xen_irq_enable_direct, 2b+1) + +/* + Disabling events is simply a matter of making the event mask + non-zero. + */ +ENTRY(xen_irq_disable_direct) + movb $1, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) +ENDPATCH(xen_irq_disable_direct) + ret + ENDPROC(xen_irq_disable_direct) + RELOC(xen_irq_disable_direct, 0) + +/* + (xen_)save_fl is used to get the current interrupt enable status. + Callers expect the status to be in X86_EFLAGS_IF, and other bits + may be set in the return value. We take advantage of this by + making sure that X86_EFLAGS_IF has the right value (and other bits + in that byte are 0), but other bits in the return value are + undefined. We need to toggle the state of the bit, because + Xen and x86 use opposite senses (mask vs enable). + */ +ENTRY(xen_save_fl_direct) + testb $0xff, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + setz %ah + addb %ah,%ah +ENDPATCH(xen_save_fl_direct) + ret + ENDPROC(xen_save_fl_direct) + RELOC(xen_save_fl_direct, 0) + +/* + In principle the caller should be passing us a value return + from xen_save_fl_direct, but for robustness sake we test only + the X86_EFLAGS_IF flag rather than the whole byte. After + setting the interrupt mask state, it checks for unmasked + pending events and enters the hypervisor to get them delivered + if so. + */ +ENTRY(xen_restore_fl_direct) + testb $X86_EFLAGS_IF>>8, %ah + setz PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_mask) + /* Preempt here doesn't matter because that will deal with + any pending interrupts. The pending check may end up being + run on the wrong CPU, but that doesn't hurt. */ + + /* check for unmasked and pending */ + cmpw $0x0001, PER_CPU_VAR(xen_vcpu_info, XEN_vcpu_info_pending) + jz 1f +2: call check_events +1: +ENDPATCH(xen_restore_fl_direct) + ret + ENDPROC(xen_restore_fl_direct) + RELOC(xen_restore_fl_direct, 2b+1) + + +/* + Force an event check by making a hypercall, + but preserve regs before making the call. + */ +check_events: + push %rax + push %rcx + push %rdx + push %rsi + push %rdi + push %r8 + push %r9 + push %r10 + push %r11 + call force_evtchn_callback + pop %r11 + pop %r10 + pop %r9 + pop %r8 + pop %rdi + pop %rsi + pop %rdx + pop %rcx + pop %rax + ret +#endif + +ENTRY(xen_iret) + pushq $0 + jmp hypercall_page + __HYPERVISOR_iret * 32 + +ENTRY(xen_sysexit) + ud2a -- cgit v1.2.3 From 15664f968a95d8fbf4a0d7b462fcc20f88906bb3 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:47 -0700 Subject: xen64: use set_fixmap for shared_info structure Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index f3f11acf785..dbe3549fad4 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -902,18 +902,11 @@ static __init void xen_pagetable_setup_start(pgd_t *base) void xen_setup_shared_info(void) { if (!xen_feature(XENFEAT_auto_translated_physmap)) { - unsigned long addr = fix_to_virt(FIX_PARAVIRT_BOOTMAP); - - /* - * Create a mapping for the shared info page. - * Should be set_fixmap(), but shared_info is a machine - * address with no corresponding pseudo-phys address. - */ - set_pte_mfn(addr, - PFN_DOWN(xen_start_info->shared_info), - PAGE_KERNEL); - - HYPERVISOR_shared_info = (struct shared_info *)addr; + set_fixmap(FIX_PARAVIRT_BOOTMAP, + xen_start_info->shared_info); + + HYPERVISOR_shared_info = + (struct shared_info *)fix_to_virt(FIX_PARAVIRT_BOOTMAP); } else HYPERVISOR_shared_info = (struct shared_info *)__va(xen_start_info->shared_info); @@ -1050,8 +1043,13 @@ static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot) #ifdef CONFIG_X86_F00F_BUG case FIX_F00F_IDT: #endif +#ifdef CONFIG_X86_32 case FIX_WP_TEST: case FIX_VDSO: + case FIX_KMAP_BEGIN ... FIX_KMAP_END: +#else + case VSYSCALL_LAST_PAGE ... VSYSCALL_FIRST_PAGE: +#endif #ifdef CONFIG_X86_LOCAL_APIC case FIX_APIC_BASE: /* maps dummy local APIC */ #endif -- cgit v1.2.3 From 7d087b68d6ddb2398fb7f6e45990b7248de640ef Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:48 -0700 Subject: xen: cpu_detect is 32-bit only Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index dbe3549fad4..2b7bea3bb6f 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1365,12 +1365,12 @@ asmlinkage void __init xen_start_kernel(void) /* set the limit of our address space */ xen_reserve_top(); +#ifdef CONFIG_X86_32 /* set up basic CPUID stuff */ cpu_detect(&new_cpu_data); -#ifdef CONFIG_X86_32 new_cpu_data.hard_math = 1; -#endif new_cpu_data.x86_capability[0] = cpuid_edx(1); +#endif /* Poke various useful things into boot_params */ boot_params.hdr.type_of_loader = (9 << 4) | 0; -- cgit v1.2.3 From 3d75e1b8ef1567348ceba93d4666a1c7c2333583 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:49 -0700 Subject: xen64: add hypervisor callbacks for events, etc Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/entry_64.S | 98 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 98 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index ae63e584c34..7cc2de79614 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -1312,3 +1312,101 @@ KPROBE_ENTRY(ignore_sysret) sysret CFI_ENDPROC ENDPROC(ignore_sysret) + +#ifdef CONFIG_XEN +ENTRY(xen_hypervisor_callback) + zeroentry xen_do_hypervisor_callback +END(xen_hypervisor_callback) + +/* +# A note on the "critical region" in our callback handler. +# We want to avoid stacking callback handlers due to events occurring +# during handling of the last event. To do this, we keep events disabled +# until we've done all processing. HOWEVER, we must enable events before +# popping the stack frame (can't be done atomically) and so it would still +# be possible to get enough handler activations to overflow the stack. +# Although unlikely, bugs of that kind are hard to track down, so we'd +# like to avoid the possibility. +# So, on entry to the handler we detect whether we interrupted an +# existing activation in its critical region -- if so, we pop the current +# activation and restart the handler using the previous one. +*/ +ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) + CFI_STARTPROC +/* Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will + see the correct pointer to the pt_regs */ + movq %rdi, %rsp # we don't return, adjust the stack frame + CFI_ENDPROC + CFI_DEFAULT_STACK +11: incl %gs:pda_irqcount + movq %rsp,%rbp + CFI_DEF_CFA_REGISTER rbp + cmovzq %gs:pda_irqstackptr,%rsp + pushq %rbp # backlink for old unwinder + call xen_evtchn_do_upcall + popq %rsp + CFI_DEF_CFA_REGISTER rsp + decl %gs:pda_irqcount + jmp error_exit + CFI_ENDPROC +END(do_hypervisor_callback) + +/* +# Hypervisor uses this for application faults while it executes. +# We get here for two reasons: +# 1. Fault while reloading DS, ES, FS or GS +# 2. Fault while executing IRET +# Category 1 we do not need to fix up as Xen has already reloaded all segment +# registers that could be reloaded and zeroed the others. +# Category 2 we fix up by killing the current process. We cannot use the +# normal Linux return path in this case because if we use the IRET hypercall +# to pop the stack frame we end up in an infinite loop of failsafe callbacks. +# We distinguish between categories by comparing each saved segment register +# with its current contents: any discrepancy means we in category 1. +*/ +ENTRY(xen_failsafe_callback) +#if 1 + ud2a +#else + _frame (RIP-0x30) + CFI_REL_OFFSET rcx, 0 + CFI_REL_OFFSET r11, 8 + movw %ds,%cx + cmpw %cx,0x10(%rsp) + CFI_REMEMBER_STATE + jne 1f + movw %es,%cx + cmpw %cx,0x18(%rsp) + jne 1f + movw %fs,%cx + cmpw %cx,0x20(%rsp) + jne 1f + movw %gs,%cx + cmpw %cx,0x28(%rsp) + jne 1f + /* All segments match their saved values => Category 2 (Bad IRET). */ + movq (%rsp),%rcx + CFI_RESTORE rcx + movq 8(%rsp),%r11 + CFI_RESTORE r11 + addq $0x30,%rsp + CFI_ADJUST_CFA_OFFSET -0x30 + movq $11,%rdi /* SIGSEGV */ + jmp do_exit + CFI_RESTORE_STATE +1: /* Segment mismatch => Category 1 (Bad segment). Retry the IRET. */ + movq (%rsp),%rcx + CFI_RESTORE rcx + movq 8(%rsp),%r11 + CFI_RESTORE r11 + addq $0x30,%rsp + CFI_ADJUST_CFA_OFFSET -0x30 + pushq $0 + CFI_ADJUST_CFA_OFFSET 8 + SAVE_ALL + jmp error_exit + CFI_ENDPROC +#endif +END(xen_failsafe_callback) + +#endif /* CONFIG_XEN */ -- cgit v1.2.3 From 084a2a4e7656209ea93aac9778defa03213ca31d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:50 -0700 Subject: xen64: early mapping setup Set up the initial pagetables to map the kernel mapping into the physical mapping space. This makes __va() usable, since it requires physical mappings. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 192 +++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 176 insertions(+), 16 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 2b7bea3bb6f..a991ee7ade9 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -33,6 +33,7 @@ #include #include #include +#include #include #include @@ -1294,6 +1295,157 @@ static void __init xen_reserve_top(void) #endif /* CONFIG_X86_32 */ } +#ifdef CONFIG_X86_64 +/* + * Like __va(), but returns address in the kernel mapping (which is + * all we have until the physical memory mapping has been set up. + */ +static void *__ka(phys_addr_t paddr) +{ + return (void *)(paddr + __START_KERNEL_map); +} + +/* Convert a machine address to physical address */ +static unsigned long m2p(phys_addr_t maddr) +{ + phys_addr_t paddr; + + maddr &= PTE_MASK; + paddr = mfn_to_pfn(maddr >> PAGE_SHIFT) << PAGE_SHIFT; + + return paddr; +} + +/* Convert a machine address to kernel virtual */ +static void *m2v(phys_addr_t maddr) +{ + return __ka(m2p(maddr)); +} + +static void walk(pgd_t *pgd, unsigned long addr) +{ + unsigned l4idx = pgd_index(addr); + unsigned l3idx = pud_index(addr); + unsigned l2idx = pmd_index(addr); + unsigned l1idx = pte_index(addr); + pgd_t l4; + pud_t l3; + pmd_t l2; + pte_t l1; + + xen_raw_printk("walk %p, %lx -> %d %d %d %d\n", + pgd, addr, l4idx, l3idx, l2idx, l1idx); + + l4 = pgd[l4idx]; + xen_raw_printk(" l4: %016lx\n", l4.pgd); + xen_raw_printk(" %016lx\n", pgd_val(l4)); + + l3 = ((pud_t *)(m2v(l4.pgd)))[l3idx]; + xen_raw_printk(" l3: %016lx\n", l3.pud); + xen_raw_printk(" %016lx\n", pud_val(l3)); + + l2 = ((pmd_t *)(m2v(l3.pud)))[l2idx]; + xen_raw_printk(" l2: %016lx\n", l2.pmd); + xen_raw_printk(" %016lx\n", pmd_val(l2)); + + l1 = ((pte_t *)(m2v(l2.pmd)))[l1idx]; + xen_raw_printk(" l1: %016lx\n", l1.pte); + xen_raw_printk(" %016lx\n", pte_val(l1)); +} + +static void set_page_prot(void *addr, pgprot_t prot) +{ + unsigned long pfn = __pa(addr) >> PAGE_SHIFT; + pte_t pte = pfn_pte(pfn, prot); + + xen_raw_printk("addr=%p pfn=%lx mfn=%lx prot=%016x pte=%016x\n", + addr, pfn, get_phys_to_machine(pfn), + pgprot_val(prot), pte.pte); + + if (HYPERVISOR_update_va_mapping((unsigned long)addr, pte, 0)) + BUG(); +} + +static void convert_pfn_mfn(void *v) +{ + pte_t *pte = v; + int i; + + /* All levels are converted the same way, so just treat them + as ptes. */ + for(i = 0; i < PTRS_PER_PTE; i++) + pte[i] = xen_make_pte(pte[i].pte); +} + +/* + * Set up the inital kernel pagetable. + * + * We can construct this by grafting the Xen provided pagetable into + * head_64.S's preconstructed pagetables. We copy the Xen L2's into + * level2_ident_pgt, level2_kernel_pgt and level2_fixmap_pgt. This + * means that only the kernel has a physical mapping to start with - + * but that's enough to get __va working. We need to fill in the rest + * of the physical mapping once some sort of allocator has been set + * up. + */ +static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) +{ + pud_t *l3; + pmd_t *l2; + + /* Zap identity mapping */ + init_level4_pgt[0] = __pgd(0); + + /* Pre-constructed entries are in pfn, so convert to mfn */ + convert_pfn_mfn(init_level4_pgt); + convert_pfn_mfn(level3_ident_pgt); + convert_pfn_mfn(level3_kernel_pgt); + + l3 = m2v(pgd[pgd_index(__START_KERNEL_map)].pgd); + l2 = m2v(l3[pud_index(__START_KERNEL_map)].pud); + + memcpy(level2_ident_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + memcpy(level2_kernel_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + + l3 = m2v(pgd[pgd_index(__START_KERNEL_map + PMD_SIZE)].pgd); + l2 = m2v(l3[pud_index(__START_KERNEL_map + PMD_SIZE)].pud); + memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + + /* Make pagetable pieces RO */ + set_page_prot(init_level4_pgt, PAGE_KERNEL_RO); + set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO); + set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(level2_ident_pgt, PAGE_KERNEL_RO); + set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO); + + /* Pin down new L4 */ + pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, PFN_DOWN(__pa(init_level4_pgt))); + + /* Unpin Xen-provided one */ + pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); + + /* Switch over */ + pgd = init_level4_pgt; + xen_write_cr3(__pa(pgd)); + + max_pfn_mapped = PFN_DOWN(__pa(pgd) + + xen_start_info->nr_pt_frames*PAGE_SIZE + + 512*1024); + + return pgd; +} +#else +static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) +{ + init_pg_tables_start = __pa(pgd); + init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; + max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024); + + return pgd; +} +#endif /* CONFIG_X86_64 */ + /* First C function to be called on Xen boot */ asmlinkage void __init xen_start_kernel(void) { @@ -1336,32 +1488,29 @@ asmlinkage void __init xen_start_kernel(void) pgd = (pgd_t *)xen_start_info->pt_base; -#ifdef CONFIG_X86_32 - init_pg_tables_start = __pa(pgd); - init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; - max_pfn_mapped = (init_pg_tables_end + 512*1024) >> PAGE_SHIFT; -#endif + /* Prevent unwanted bits from being set in PTEs. */ + __supported_pte_mask &= ~_PAGE_GLOBAL; + if (!is_initial_xendomain()) + __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD); + + /* Don't do the full vcpu_info placement stuff until we have a + possible map and a non-dummy shared_info. */ + per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0]; + + xen_raw_console_write("mapping kernel into physical memory\n"); + pgd = xen_setup_kernel_pagetable(pgd); - init_mm.pgd = pgd; /* use the Xen pagetables to start */ + init_mm.pgd = pgd; /* keep using Xen gdt for now; no urgent need to change it */ x86_write_percpu(xen_cr3, __pa(pgd)); x86_write_percpu(xen_current_cr3, __pa(pgd)); - /* Don't do the full vcpu_info placement stuff until we have a - possible map and a non-dummy shared_info. */ - per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0]; - pv_info.kernel_rpl = 1; if (xen_feature(XENFEAT_supervisor_mode_kernel)) pv_info.kernel_rpl = 0; - /* Prevent unwanted bits from being set in PTEs. */ - __supported_pte_mask &= ~_PAGE_GLOBAL; - if (!is_initial_xendomain()) - __supported_pte_mask &= ~(_PAGE_PWT | _PAGE_PCD); - /* set the limit of our address space */ xen_reserve_top(); @@ -1384,10 +1533,21 @@ asmlinkage void __init xen_start_kernel(void) add_preferred_console("hvc", 0, NULL); } + xen_raw_console_write("about to get started...\n"); + +#if 0 + xen_raw_printk("&boot_params=%p __pa(&boot_params)=%lx __va(__pa(&boot_params))=%lx\n", + &boot_params, __pa_symbol(&boot_params), + __va(__pa_symbol(&boot_params))); + + walk(pgd, &boot_params); + walk(pgd, __va(__pa(&boot_params))); +#endif + /* Start the world */ #ifdef CONFIG_X86_32 i386_start_kernel(); #else - x86_64_start_kernel((char *)&boot_params); + x86_64_start_reservations((char *)__pa_symbol(&boot_params)); #endif } -- cgit v1.2.3 From 22911b3f1cf5431058e56b1727e8ef77be5e0ac9 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:51 -0700 Subject: xen64: 64-bit starts using set_pte from very early It also doesn't need the 32-bit hack version of set_pte for initial pagetable construction, so just make it use the real thing. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index a991ee7ade9..392450787aa 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1194,7 +1194,11 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .kmap_atomic_pte = xen_kmap_atomic_pte, #endif +#ifdef CONFIG_X86_64 + .set_pte = xen_set_pte, +#else .set_pte = xen_set_pte_init, +#endif .set_pte_at = xen_set_pte_at, .set_pmd = xen_set_pmd_hyper, -- cgit v1.2.3 From d114e1981cc1a51131230993a082c27c79ab370a Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:52 -0700 Subject: xen64: map an initial chunk of physical memory Early in boot, map a chunk of extra physical memory for use later on. We need a pool of mapped pages to allocate further pages to construct pagetables mapping all physical memory. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 79 ++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 69 insertions(+), 10 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 392450787aa..e9e3bafe48c 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1381,6 +1381,61 @@ static void convert_pfn_mfn(void *v) pte[i] = xen_make_pte(pte[i].pte); } +/* + * Identity map, in addition to plain kernel map. This needs to be + * large enough to allocate page table pages to allocate the rest. + * Each page can map 2MB. + */ +static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss; + +static __init void xen_map_identity_early(unsigned long max_pfn) +{ + unsigned pmdidx, pteidx; + unsigned ident_pte; + unsigned long pfn; + + ident_pte = 0; + pfn = 0; + for(pmdidx = 0; pmdidx < PTRS_PER_PMD && pfn < max_pfn; pmdidx++) { + pte_t *pte_page; + + BUG_ON(level2_ident_pgt[pmdidx].pmd != level2_kernel_pgt[pmdidx].pmd); + + /* Reuse or allocate a page of ptes */ + if (pmd_present(level2_ident_pgt[pmdidx])) + pte_page = m2v(level2_ident_pgt[pmdidx].pmd); + else { + /* Check for free pte pages */ + if (ident_pte == ARRAY_SIZE(level1_ident_pgt)) + break; + + pte_page = &level1_ident_pgt[ident_pte]; + ident_pte += PTRS_PER_PTE; + + /* Install new l1 in l2(s) */ + level2_ident_pgt[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE); + level2_kernel_pgt[pmdidx] = level2_ident_pgt[pmdidx]; + } + + /* Install mappings */ + for(pteidx = 0; pteidx < PTRS_PER_PTE; pteidx++, pfn++) { + pte_t pte; + + if (pfn > max_pfn_mapped) + max_pfn_mapped = pfn; + + if (!pte_none(pte_page[pteidx])) + continue; + + pte = pfn_pte(pfn, PAGE_KERNEL_EXEC); + pte_page[pteidx] = pte; + } + } + + for(pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE) + set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO); +} + /* * Set up the inital kernel pagetable. * @@ -1392,7 +1447,7 @@ static void convert_pfn_mfn(void *v) * of the physical mapping once some sort of allocator has been set * up. */ -static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) +static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pfn) { pud_t *l3; pmd_t *l2; @@ -1415,6 +1470,9 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) l2 = m2v(l3[pud_index(__START_KERNEL_map + PMD_SIZE)].pud); memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); + /* Set up identity map */ + xen_map_identity_early(max_pfn); + /* Make pagetable pieces RO */ set_page_prot(init_level4_pgt, PAGE_KERNEL_RO); set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO); @@ -1424,7 +1482,7 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO); /* Pin down new L4 */ - pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, PFN_DOWN(__pa(init_level4_pgt))); + pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, PFN_DOWN(__pa_symbol(init_level4_pgt))); /* Unpin Xen-provided one */ pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); @@ -1433,19 +1491,23 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) pgd = init_level4_pgt; xen_write_cr3(__pa(pgd)); - max_pfn_mapped = PFN_DOWN(__pa(pgd) + - xen_start_info->nr_pt_frames*PAGE_SIZE + - 512*1024); + reserve_early(__pa(xen_start_info->pt_base), + __pa(xen_start_info->pt_base + + xen_start_info->nr_pt_frames * PAGE_SIZE), + "XEN PAGETABLES"); return pgd; } #else -static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd) +static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pfn) { init_pg_tables_start = __pa(pgd); init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024); + x86_write_percpu(xen_cr3, __pa(pgd)); + x86_write_percpu(xen_current_cr3, __pa(pgd)); + return pgd; } #endif /* CONFIG_X86_64 */ @@ -1502,15 +1564,12 @@ asmlinkage void __init xen_start_kernel(void) per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0]; xen_raw_console_write("mapping kernel into physical memory\n"); - pgd = xen_setup_kernel_pagetable(pgd); + pgd = xen_setup_kernel_pagetable(pgd, xen_start_info->nr_pages); init_mm.pgd = pgd; /* keep using Xen gdt for now; no urgent need to change it */ - x86_write_percpu(xen_cr3, __pa(pgd)); - x86_write_percpu(xen_current_cr3, __pa(pgd)); - pv_info.kernel_rpl = 1; if (xen_feature(XENFEAT_supervisor_mode_kernel)) pv_info.kernel_rpl = 0; -- cgit v1.2.3 From 39dbc5bd345ebf93e066dde7f8e29467eb61b42e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:53 -0700 Subject: xen32: create initial mappings like 64-bit Rearrange the pagetable initialization to share code with the 64-bit kernel. Rather than deferring anything to pagetable_setup_start, just set up an initial pagetable in swapper_pg_dir early at startup, and create an additional 8MB of physical memory mappings. This matches the native head_32.S mappings to a large degree, and allows the rest of the pagetable setup to continue without much Xen vs. native difference. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 130 +++++++++++++++++++---------------------------- 1 file changed, 52 insertions(+), 78 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index e9e3bafe48c..19c12a6c731 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -854,50 +854,6 @@ static __init void xen_set_pte_init(pte_t *ptep, pte_t pte) static __init void xen_pagetable_setup_start(pgd_t *base) { -#ifdef CONFIG_X86_32 - pgd_t *xen_pgd = (pgd_t *)xen_start_info->pt_base; - int i; - - init_mm.pgd = base; - /* - * copy top-level of Xen-supplied pagetable into place. This - * is a stand-in while we copy the pmd pages. - */ - memcpy(base, xen_pgd, PTRS_PER_PGD * sizeof(pgd_t)); - - /* - * For PAE, need to allocate new pmds, rather than - * share Xen's, since Xen doesn't like pmd's being - * shared between address spaces. - */ - for (i = 0; i < PTRS_PER_PGD; i++) { - if (pgd_val_ma(xen_pgd[i]) & _PAGE_PRESENT) { - pmd_t *pmd = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); - - memcpy(pmd, (void *)pgd_page_vaddr(xen_pgd[i]), - PAGE_SIZE); - - make_lowmem_page_readonly(pmd); - - set_pgd(&base[i], __pgd(1 + __pa(pmd))); - } else - pgd_clear(&base[i]); - } - - /* make sure zero_page is mapped RO so we can use it in pagetables */ - make_lowmem_page_readonly(empty_zero_page); - make_lowmem_page_readonly(base); - /* - * Switch to new pagetable. This is done before - * pagetable_init has done anything so that the new pages - * added to the table can be prepared properly for Xen. - */ - xen_write_cr3(__pa(base)); - - /* Unpin initial Xen pagetable */ - pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, - PFN_DOWN(__pa(xen_start_info->pt_base))); -#endif /* CONFIG_X86_32 */ } void xen_setup_shared_info(void) @@ -936,12 +892,6 @@ static __init void xen_pagetable_setup_done(pgd_t *base) pv_mmu_ops.set_pte = xen_set_pte; xen_setup_shared_info(); - -#ifdef CONFIG_X86_32 - /* Actually pin the pagetable down, but we can't set PG_pinned - yet because the page structures don't exist yet. */ - pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(base))); -#endif } static __init void xen_post_allocator_init(void) @@ -1299,14 +1249,17 @@ static void __init xen_reserve_top(void) #endif /* CONFIG_X86_32 */ } -#ifdef CONFIG_X86_64 /* * Like __va(), but returns address in the kernel mapping (which is * all we have until the physical memory mapping has been set up. */ static void *__ka(phys_addr_t paddr) { +#ifdef CONFIG_X86_64 return (void *)(paddr + __START_KERNEL_map); +#else + return __va(paddr); +#endif } /* Convert a machine address to physical address */ @@ -1326,6 +1279,7 @@ static void *m2v(phys_addr_t maddr) return __ka(m2p(maddr)); } +#ifdef CONFIG_X86_64 static void walk(pgd_t *pgd, unsigned long addr) { unsigned l4idx = pgd_index(addr); @@ -1356,13 +1310,14 @@ static void walk(pgd_t *pgd, unsigned long addr) xen_raw_printk(" l1: %016lx\n", l1.pte); xen_raw_printk(" %016lx\n", pte_val(l1)); } +#endif static void set_page_prot(void *addr, pgprot_t prot) { unsigned long pfn = __pa(addr) >> PAGE_SHIFT; pte_t pte = pfn_pte(pfn, prot); - xen_raw_printk("addr=%p pfn=%lx mfn=%lx prot=%016x pte=%016x\n", + xen_raw_printk("addr=%p pfn=%lx mfn=%lx prot=%016llx pte=%016llx\n", addr, pfn, get_phys_to_machine(pfn), pgprot_val(prot), pte.pte); @@ -1370,17 +1325,6 @@ static void set_page_prot(void *addr, pgprot_t prot) BUG(); } -static void convert_pfn_mfn(void *v) -{ - pte_t *pte = v; - int i; - - /* All levels are converted the same way, so just treat them - as ptes. */ - for(i = 0; i < PTRS_PER_PTE; i++) - pte[i] = xen_make_pte(pte[i].pte); -} - /* * Identity map, in addition to plain kernel map. This needs to be * large enough to allocate page table pages to allocate the rest. @@ -1388,7 +1332,7 @@ static void convert_pfn_mfn(void *v) */ static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss; -static __init void xen_map_identity_early(unsigned long max_pfn) +static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn) { unsigned pmdidx, pteidx; unsigned ident_pte; @@ -1399,11 +1343,9 @@ static __init void xen_map_identity_early(unsigned long max_pfn) for(pmdidx = 0; pmdidx < PTRS_PER_PMD && pfn < max_pfn; pmdidx++) { pte_t *pte_page; - BUG_ON(level2_ident_pgt[pmdidx].pmd != level2_kernel_pgt[pmdidx].pmd); - /* Reuse or allocate a page of ptes */ - if (pmd_present(level2_ident_pgt[pmdidx])) - pte_page = m2v(level2_ident_pgt[pmdidx].pmd); + if (pmd_present(pmd[pmdidx])) + pte_page = m2v(pmd[pmdidx].pmd); else { /* Check for free pte pages */ if (ident_pte == ARRAY_SIZE(level1_ident_pgt)) @@ -1412,9 +1354,7 @@ static __init void xen_map_identity_early(unsigned long max_pfn) pte_page = &level1_ident_pgt[ident_pte]; ident_pte += PTRS_PER_PTE; - /* Install new l1 in l2(s) */ - level2_ident_pgt[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE); - level2_kernel_pgt[pmdidx] = level2_ident_pgt[pmdidx]; + pmd[pmdidx] = __pmd(__pa(pte_page) | _PAGE_TABLE); } /* Install mappings */ @@ -1434,6 +1374,20 @@ static __init void xen_map_identity_early(unsigned long max_pfn) for(pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE) set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO); + + set_page_prot(pmd, PAGE_KERNEL_RO); +} + +#ifdef CONFIG_X86_64 +static void convert_pfn_mfn(void *v) +{ + pte_t *pte = v; + int i; + + /* All levels are converted the same way, so just treat them + as ptes. */ + for(i = 0; i < PTRS_PER_PTE; i++) + pte[i] = xen_make_pte(pte[i].pte); } /* @@ -1471,18 +1425,18 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pf memcpy(level2_fixmap_pgt, l2, sizeof(pmd_t) * PTRS_PER_PMD); /* Set up identity map */ - xen_map_identity_early(max_pfn); + xen_map_identity_early(level2_ident_pgt, max_pfn); /* Make pagetable pieces RO */ set_page_prot(init_level4_pgt, PAGE_KERNEL_RO); set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO); set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO); - set_page_prot(level2_ident_pgt, PAGE_KERNEL_RO); set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO); /* Pin down new L4 */ - pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, PFN_DOWN(__pa_symbol(init_level4_pgt))); + pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE, + PFN_DOWN(__pa_symbol(init_level4_pgt))); /* Unpin Xen-provided one */ pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); @@ -1498,17 +1452,37 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pf return pgd; } -#else +#else /* !CONFIG_X86_64 */ +static pmd_t level2_kernel_pgt[PTRS_PER_PMD] __page_aligned_bss; + static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pfn) { + pmd_t *kernel_pmd; + init_pg_tables_start = __pa(pgd); init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; max_pfn_mapped = PFN_DOWN(init_pg_tables_end + 512*1024); - x86_write_percpu(xen_cr3, __pa(pgd)); - x86_write_percpu(xen_current_cr3, __pa(pgd)); + kernel_pmd = m2v(pgd[KERNEL_PGD_BOUNDARY].pgd); + memcpy(level2_kernel_pgt, kernel_pmd, sizeof(pmd_t) * PTRS_PER_PMD); - return pgd; + xen_map_identity_early(level2_kernel_pgt, max_pfn); + + memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD); + set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY], + __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT)); + + set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO); + set_page_prot(empty_zero_page, PAGE_KERNEL_RO); + + pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); + + xen_write_cr3(__pa(swapper_pg_dir)); + + pin_pagetable_pfn(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(swapper_pg_dir))); + + return swapper_pg_dir; } #endif /* CONFIG_X86_64 */ -- cgit v1.2.3 From ebd879e397f6361727c36267a12d1650710e465a Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:54 -0700 Subject: xen: fix truncation of machine address arbitrary_virt_to_machine can truncate a machine address if its above 4G. Cast the problem away. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 2579e70cdd0..05d7392a7a4 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -186,7 +186,7 @@ xmaddr_t arbitrary_virt_to_machine(unsigned long address) BUG_ON(pte == NULL); - return XMADDR((pte_mfn(*pte) << PAGE_SHIFT) + offset); + return XMADDR(((phys_addr_t)pte_mfn(*pte) << PAGE_SHIFT) + offset); } void make_lowmem_page_readonly(void *vaddr) -- cgit v1.2.3 From ce803e705f1cbdd2703e83061622089b5b4a5417 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:55 -0700 Subject: xen64: use arbitrary_virt_to_machine for xen_set_pmd When building initial pagetables in 64-bit kernel the pud/pmd pointer may be in ioremap/fixmap space, so we need to walk the pagetable to look up the physical address. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 05d7392a7a4..a8f02327181 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -178,8 +178,9 @@ void set_phys_to_machine(unsigned long pfn, unsigned long mfn) p2m_top[topidx][idx] = mfn; } -xmaddr_t arbitrary_virt_to_machine(unsigned long address) +xmaddr_t arbitrary_virt_to_machine(void *vaddr) { + unsigned long address = (unsigned long)vaddr; unsigned int level; pte_t *pte = lookup_address(address, &level); unsigned offset = address & ~PAGE_MASK; @@ -253,7 +254,8 @@ void xen_set_pmd_hyper(pmd_t *ptr, pmd_t val) xen_mc_batch(); - u.ptr = virt_to_machine(ptr).maddr; + /* ptr may be ioremapped for 64-bit pagetable setup */ + u.ptr = arbitrary_virt_to_machine(ptr).maddr; u.val = pmd_val_ma(val); extend_mmu_update(&u); @@ -415,7 +417,8 @@ void xen_set_pud_hyper(pud_t *ptr, pud_t val) xen_mc_batch(); - u.ptr = virt_to_machine(ptr).maddr; + /* ptr may be ioremapped for 64-bit pagetable setup */ + u.ptr = arbitrary_virt_to_machine(ptr).maddr; u.val = pud_val_ma(val); extend_mmu_update(&u); -- cgit v1.2.3 From 4560a2947e32670fc6ede108c2b032c396180649 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:56 -0700 Subject: xen: set num_processors Someone's got to do it. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/smp.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 800bb2191e2..8310ca0ea37 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -155,8 +155,10 @@ static void __init xen_fill_possible_map(void) for (i = 0; i < NR_CPUS; i++) { rc = HYPERVISOR_vcpu_op(VCPUOP_is_up, i, NULL); - if (rc >= 0) + if (rc >= 0) { + num_processors++; cpu_set(i, cpu_possible_map); + } } } -- cgit v1.2.3 From 8745f8b0b914cf1d617ecc49726c24011858c74e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:57 -0700 Subject: xen64: defer setting pagetable alloc/release ops We need to wait until the page structure is available to use the proper pagetable page alloc/release operations, since they use struct page to determine if a pagetable is pinned. This happened to work in 32bit because nobody allocated new pagetable pages in the interim between xen_pagetable_setup_done and xen_post_allocator_init, but the 64-bit kenrel needs to allocate more pagetable levels. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 25 ++++++++++++------------- 1 file changed, 12 insertions(+), 13 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 19c12a6c731..da91404fc66 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -878,30 +878,29 @@ void xen_setup_shared_info(void) static __init void xen_pagetable_setup_done(pgd_t *base) { - /* This will work as long as patching hasn't happened yet - (which it hasn't) */ - pv_mmu_ops.alloc_pte = xen_alloc_pte; - pv_mmu_ops.alloc_pmd = xen_alloc_pmd; - pv_mmu_ops.release_pte = xen_release_pte; - pv_mmu_ops.release_pmd = xen_release_pmd; -#if PAGETABLE_LEVELS == 4 - pv_mmu_ops.alloc_pud = xen_alloc_pud; - pv_mmu_ops.release_pud = xen_release_pud; -#endif - - pv_mmu_ops.set_pte = xen_set_pte; - xen_setup_shared_info(); } static __init void xen_post_allocator_init(void) { + pv_mmu_ops.set_pte = xen_set_pte; pv_mmu_ops.set_pmd = xen_set_pmd; pv_mmu_ops.set_pud = xen_set_pud; #if PAGETABLE_LEVELS == 4 pv_mmu_ops.set_pgd = xen_set_pgd; #endif + /* This will work as long as patching hasn't happened yet + (which it hasn't) */ + pv_mmu_ops.alloc_pte = xen_alloc_pte; + pv_mmu_ops.alloc_pmd = xen_alloc_pmd; + pv_mmu_ops.release_pte = xen_release_pte; + pv_mmu_ops.release_pmd = xen_release_pmd; +#if PAGETABLE_LEVELS == 4 + pv_mmu_ops.alloc_pud = xen_alloc_pud; + pv_mmu_ops.release_pud = xen_release_pud; +#endif + xen_mark_init_mm_pinned(); } -- cgit v1.2.3 From 836fe2f291cb450a6193fa713878efe7d32bec6e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:06:58 -0700 Subject: xen: use set_pte_vaddr Make Xen's set_pte_mfn() use set_pte_vaddr rather than copying it. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Juan Quintela Signed-off-by: Mark McLoughlin Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 30 +----------------------------- 1 file changed, 1 insertion(+), 29 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index a8f02327181..eb31ed291b9 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -282,35 +282,7 @@ void xen_set_pmd(pmd_t *ptr, pmd_t val) */ void set_pte_mfn(unsigned long vaddr, unsigned long mfn, pgprot_t flags) { - pgd_t *pgd; - pud_t *pud; - pmd_t *pmd; - pte_t *pte; - - pgd = swapper_pg_dir + pgd_index(vaddr); - if (pgd_none(*pgd)) { - BUG(); - return; - } - pud = pud_offset(pgd, vaddr); - if (pud_none(*pud)) { - BUG(); - return; - } - pmd = pmd_offset(pud, vaddr); - if (pmd_none(*pmd)) { - BUG(); - return; - } - pte = pte_offset_kernel(pmd, vaddr); - /* stored as-is, to permit clearing entries */ - xen_set_pte(pte, mfn_pte(mfn, flags)); - - /* - * It's enough to flush this one mapping. - * (PGE mappings get flushed as well) - */ - __flush_tlb_one(vaddr); + set_pte_vaddr(vaddr, mfn_pte(mfn, flags)); } void xen_set_pte_at(struct mm_struct *mm, unsigned long addr, -- cgit v1.2.3 From e176d367d0cc8b8efd2e0960c9edf5d2fe7cd9f1 Mon Sep 17 00:00:00 2001 From: Eduardo Habkost Date: Tue, 8 Jul 2008 15:06:59 -0700 Subject: xen64: xen_write_idt_entry() and cvt_gate_to_trap() Changed to use the (to-be-)unified descriptor structs. Signed-off-by: Eduardo Habkost Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 26 ++++++++++---------------- 1 file changed, 10 insertions(+), 16 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index da91404fc66..f5e96f7a4c5 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -401,23 +401,18 @@ static void xen_write_ldt_entry(struct desc_struct *dt, int entrynum, preempt_enable(); } -static int cvt_gate_to_trap(int vector, u32 low, u32 high, +static int cvt_gate_to_trap(int vector, const gate_desc *val, struct trap_info *info) { - u8 type, dpl; - - type = (high >> 8) & 0x1f; - dpl = (high >> 13) & 3; - - if (type != 0xf && type != 0xe) + if (val->type != 0xf && val->type != 0xe) return 0; info->vector = vector; - info->address = (high & 0xffff0000) | (low & 0x0000ffff); - info->cs = low >> 16; - info->flags = dpl; + info->address = gate_offset(*val); + info->cs = gate_segment(*val); + info->flags = val->dpl; /* interrupt gates clear IF */ - if (type == 0xe) + if (val->type == 0xe) info->flags |= 4; return 1; @@ -444,11 +439,10 @@ static void xen_write_idt_entry(gate_desc *dt, int entrynum, const gate_desc *g) if (p >= start && (p + 8) <= end) { struct trap_info info[2]; - u32 *desc = (u32 *)g; info[1].address = 0; - if (cvt_gate_to_trap(entrynum, desc[0], desc[1], &info[0])) + if (cvt_gate_to_trap(entrynum, g, &info[0])) if (HYPERVISOR_set_trap_table(info)) BUG(); } @@ -461,13 +455,13 @@ static void xen_convert_trap_info(const struct desc_ptr *desc, { unsigned in, out, count; - count = (desc->size+1) / 8; + count = (desc->size+1) / sizeof(gate_desc); BUG_ON(count > 256); for (in = out = 0; in < count; in++) { - const u32 *entry = (u32 *)(desc->address + in * 8); + gate_desc *entry = (gate_desc*)(desc->address) + in; - if (cvt_gate_to_trap(in, entry[0], entry[1], &traps[out])) + if (cvt_gate_to_trap(in, entry, &traps[out])) out++; } traps[out].address = 0; -- cgit v1.2.3 From 997409d3d0bd6894f33e31ced251c0fdf523aa14 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:00 -0700 Subject: xen64: deal with extra words Xen pushes onto exception frames Xen pushes two extra words containing the values of rcx and r11. This pvop hook copies the words back into their appropriate registers, and cleans them off the stack. This leaves the stack in native form, so the normal handler can run unchanged. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 2 +- arch/x86/xen/xen-asm_64.S | 5 +++++ arch/x86/xen/xen-ops.h | 2 ++ 3 files changed, 8 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index f5e96f7a4c5..9d94483b3b5 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1091,7 +1091,7 @@ static const struct pv_irq_ops xen_irq_ops __initdata = { .safe_halt = xen_safe_halt, .halt = xen_halt, #ifdef CONFIG_X86_64 - .adjust_exception_frame = paravirt_nop, + .adjust_exception_frame = xen_adjust_exception_frame, #endif }; diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S index 4ec10827370..b147b495dae 100644 --- a/arch/x86/xen/xen-asm_64.S +++ b/arch/x86/xen/xen-asm_64.S @@ -133,6 +133,11 @@ check_events: ret #endif +ENTRY(xen_adjust_exception_frame) + mov 8+0(%rsp),%rcx + mov 8+8(%rsp),%r11 + ret $16 + ENTRY(xen_iret) pushq $0 jmp hypercall_page + __HYPERVISOR_iret * 32 diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index aca4a7803e2..c4800a2c5a4 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -67,7 +67,9 @@ DECL_ASM(void, xen_irq_disable_direct, void); DECL_ASM(unsigned long, xen_save_fl_direct, void); DECL_ASM(void, xen_restore_fl_direct, unsigned long); +/* These are not functions, and cannot be called normally */ void xen_iret(void); void xen_sysexit(void); +void xen_adjust_exception_frame(void); #endif /* XEN_OPS_H */ -- cgit v1.2.3 From 952d1d7055c8cbf95b4ad2f90be5ed37db8a48ee Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:01 -0700 Subject: xen64: add pvop for swapgs swapgs is a no-op under Xen, because the hypervisor makes sure the right version of %gs is current when switching between user and kernel modes. This means that the swapgs "implementation" can be inlined and used when the stack is unsafe (usermode). Unfortunately, it means that disabling patching will result in a non-booting kernel... Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 9d94483b3b5..8b60982e457 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1076,6 +1076,9 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { .set_iopl_mask = xen_set_iopl_mask, .io_delay = xen_io_delay, + /* Xen takes care of %gs when switching to usermode for us */ + .swapgs = paravirt_nop, + .lazy_mode = { .enter = paravirt_enter_lazy_cpu, .leave = xen_leave_lazy, -- cgit v1.2.3 From 88459d4c7eb68c4a15609e00e5d100e2a305f040 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:02 -0700 Subject: xen64: register callbacks in arch-independent way Use callback_op hypercall to register callbacks in a 32/64-bit independent way (64-bit doesn't need a code segment, but that detail is hidden in XEN_CALLBACK). Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 27 +++++++++++++++++---------- 1 file changed, 17 insertions(+), 10 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index f52f3855fb6..bea3d4f779d 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -91,19 +91,25 @@ static void __init fiddle_vdso(void) *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; } -void xen_enable_sysenter(void) +static __cpuinit int register_callback(unsigned type, const void *func) { - int cpu = smp_processor_id(); - extern void xen_sysenter_target(void); - /* Mask events on entry, even though they get enabled immediately */ - static struct callback_register sysenter = { - .type = CALLBACKTYPE_sysenter, - .address = XEN_CALLBACK(__KERNEL_CS, xen_sysenter_target), + struct callback_register callback = { + .type = type, + .address = XEN_CALLBACK(__KERNEL_CS, func), .flags = CALLBACKF_mask_events, }; + return HYPERVISOR_callback_op(CALLBACKOP_register, &callback); +} + +void __cpuinit xen_enable_sysenter(void) +{ + int cpu = smp_processor_id(); + extern void xen_sysenter_target(void); + if (!boot_cpu_has(X86_FEATURE_SEP) || - HYPERVISOR_callback_op(CALLBACKOP_register, &sysenter) != 0) { + register_callback(CALLBACKTYPE_sysenter, + xen_sysenter_target) != 0) { clear_cpu_cap(&cpu_data(cpu), X86_FEATURE_SEP); clear_cpu_cap(&boot_cpu_data, X86_FEATURE_SEP); } @@ -120,8 +126,9 @@ void __init xen_arch_setup(void) if (!xen_feature(XENFEAT_auto_translated_physmap)) HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_pae_extended_cr3); - HYPERVISOR_set_callbacks(__KERNEL_CS, (unsigned long)xen_hypervisor_callback, - __KERNEL_CS, (unsigned long)xen_failsafe_callback); + if (register_callback(CALLBACKTYPE_event, xen_hypervisor_callback) || + register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback)) + BUG(); xen_enable_sysenter(); -- cgit v1.2.3 From 0725cbb97793d4e65bf148e4872959cdbb8c6ddd Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:03 -0700 Subject: xen64: add identity irq->vector map The x86_64 interrupt subsystem is oriented towards vectors, as opposed to a flat irq space as it is in x86-32. This patch adds a simple identity irq->vector mapping so that we can continue to feed irqs into do_IRQ() and get a good result. Ideally x86_32 will unify with the 64-bit code and use vectors too. At that point we can move to mapping event channels to vectors, which will allow us to economise on irqs (so per-cpu event channels can share irqs, rather than having to allocte one per cpu, for example). Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 8b60982e457..52f2292672c 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1085,8 +1085,25 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { }, }; +static void __init __xen_init_IRQ(void) +{ +#ifdef CONFIG_X86_64 + int i; + + /* Create identity vector->irq map */ + for(i = 0; i < NR_VECTORS; i++) { + int cpu; + + for_each_possible_cpu(cpu) + per_cpu(vector_irq, cpu)[i] = i; + } +#endif /* CONFIG_X86_64 */ + + xen_init_IRQ(); +} + static const struct pv_irq_ops xen_irq_ops __initdata = { - .init_IRQ = xen_init_IRQ, + .init_IRQ = __xen_init_IRQ, .save_fl = xen_save_fl, .restore_fl = xen_restore_fl, .irq_disable = xen_irq_disable, -- cgit v1.2.3 From a8fc1089e49caa5dca346dfacb5c84abf9a22a0c Mon Sep 17 00:00:00 2001 From: Eduardo Habkost Date: Tue, 8 Jul 2008 15:07:05 -0700 Subject: xen64: implement xen_load_gs_index() xen-64: implement xen_load_gs_index() Signed-off-by: Eduardo Habkost Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 52f2292672c..3b6b7fcf5b5 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -385,6 +385,14 @@ static void xen_load_tls(struct thread_struct *t, unsigned int cpu) loadsegment(gs, 0); } +#ifdef CONFIG_X86_64 +static void xen_load_gs_index(unsigned int idx) +{ + if (HYPERVISOR_set_segment_base(SEGBASE_GS_USER_SEL, idx)) + BUG(); +} +#endif + static void xen_write_ldt_entry(struct desc_struct *dt, int entrynum, const void *ptr) { @@ -1063,6 +1071,9 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { .load_gdt = xen_load_gdt, .load_idt = xen_load_idt, .load_tls = xen_load_tls, +#ifdef CONFIG_X86_64 + .load_gs_index = xen_load_gs_index, +#endif .store_gdt = native_store_gdt, .store_idt = native_store_idt, -- cgit v1.2.3 From 5deb30d194d28b6bf7dacfb758267a51bf7c5b78 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:06 -0700 Subject: xen: rework pgd_walk to deal with 32/64 bit Rewrite pgd_walk to deal with 64-bit address spaces. There are two notible features of 64-bit workspaces: 1. The physical address is only 48 bits wide, with the upper 16 bits being sign extension; kernel addresses are negative, and userspace is positive. 2. The Xen hypervisor mapping is at the negative-most address, just above the sign-extension hole. 1. means that we can't easily use addresses when traversing the space, since we must deal with sign extension. This rewrite expresses everything in terms of pgd/pud/pmd indices, which means we don't need to worry about the exact configuration of the virtual memory space. This approach works equally well in 32-bit. To deal with 2, assume the hole is between the uppermost userspace address and PAGE_OFFSET. For 64-bit this skips the Xen mapping hole. For 32-bit, the hole is zero-sized. In all cases, the uppermost kernel address is FIXADDR_TOP. A side-effect of this patch is that the upper boundary is actually handled properly, exposing a long-standing bug in 32-bit, which failed to pin kernel pmd page. The kernel pmd is not shared, and so must be explicitly pinned, even though the kernel ptes are shared and don't need pinning. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/mmu.c | 115 ++++++++++++++++++++++++++++++++++------------------- 1 file changed, 75 insertions(+), 40 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index eb31ed291b9..046c1f23dd6 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -44,6 +44,7 @@ #include #include +#include #include #include #include @@ -491,77 +492,103 @@ void xen_set_pgd(pgd_t *ptr, pgd_t val) #endif /* PAGETABLE_LEVELS == 4 */ /* - (Yet another) pagetable walker. This one is intended for pinning a - pagetable. This means that it walks a pagetable and calls the - callback function on each page it finds making up the page table, - at every level. It walks the entire pagetable, but it only bothers - pinning pte pages which are below pte_limit. In the normal case - this will be TASK_SIZE, but at boot we need to pin up to - FIXADDR_TOP. But the important bit is that we don't pin beyond - there, because then we start getting into Xen's ptes. -*/ -static int pgd_walk(pgd_t *pgd_base, int (*func)(struct page *, enum pt_level), + * (Yet another) pagetable walker. This one is intended for pinning a + * pagetable. This means that it walks a pagetable and calls the + * callback function on each page it finds making up the page table, + * at every level. It walks the entire pagetable, but it only bothers + * pinning pte pages which are below limit. In the normal case this + * will be STACK_TOP_MAX, but at boot we need to pin up to + * FIXADDR_TOP. + * + * For 32-bit the important bit is that we don't pin beyond there, + * because then we start getting into Xen's ptes. + * + * For 64-bit, we must skip the Xen hole in the middle of the address + * space, just after the big x86-64 virtual hole. + */ +static int pgd_walk(pgd_t *pgd, int (*func)(struct page *, enum pt_level), unsigned long limit) { - pgd_t *pgd = pgd_base; int flush = 0; - unsigned long addr = 0; - unsigned long pgd_next; + unsigned hole_low, hole_high; + unsigned pgdidx_limit, pudidx_limit, pmdidx_limit; + unsigned pgdidx, pudidx, pmdidx; - BUG_ON(limit > FIXADDR_TOP); + /* The limit is the last byte to be touched */ + limit--; + BUG_ON(limit >= FIXADDR_TOP); if (xen_feature(XENFEAT_auto_translated_physmap)) return 0; - for (; addr != FIXADDR_TOP; pgd++, addr = pgd_next) { + /* + * 64-bit has a great big hole in the middle of the address + * space, which contains the Xen mappings. On 32-bit these + * will end up making a zero-sized hole and so is a no-op. + */ + hole_low = pgd_index(STACK_TOP_MAX + PGDIR_SIZE - 1); + hole_high = pgd_index(PAGE_OFFSET); + + pgdidx_limit = pgd_index(limit); +#if PTRS_PER_PUD > 1 + pudidx_limit = pud_index(limit); +#else + pudidx_limit = 0; +#endif +#if PTRS_PER_PMD > 1 + pmdidx_limit = pmd_index(limit); +#else + pmdidx_limit = 0; +#endif + + flush |= (*func)(virt_to_page(pgd), PT_PGD); + + for (pgdidx = 0; pgdidx <= pgdidx_limit; pgdidx++) { pud_t *pud; - unsigned long pud_limit, pud_next; - pgd_next = pud_limit = pgd_addr_end(addr, FIXADDR_TOP); + if (pgdidx >= hole_low && pgdidx < hole_high) + continue; - if (!pgd_val(*pgd)) + if (!pgd_val(pgd[pgdidx])) continue; - pud = pud_offset(pgd, 0); + pud = pud_offset(&pgd[pgdidx], 0); if (PTRS_PER_PUD > 1) /* not folded */ flush |= (*func)(virt_to_page(pud), PT_PUD); - for (; addr != pud_limit; pud++, addr = pud_next) { + for (pudidx = 0; pudidx < PTRS_PER_PUD; pudidx++) { pmd_t *pmd; - unsigned long pmd_limit; - pud_next = pud_addr_end(addr, pud_limit); - - if (pud_next < limit) - pmd_limit = pud_next; - else - pmd_limit = limit; + if (pgdidx == pgdidx_limit && + pudidx > pudidx_limit) + goto out; - if (pud_none(*pud)) + if (pud_none(pud[pudidx])) continue; - pmd = pmd_offset(pud, 0); + pmd = pmd_offset(&pud[pudidx], 0); if (PTRS_PER_PMD > 1) /* not folded */ flush |= (*func)(virt_to_page(pmd), PT_PMD); - for (; addr != pmd_limit; pmd++) { - addr += (PAGE_SIZE * PTRS_PER_PTE); - if ((pmd_limit-1) < (addr-1)) { - addr = pmd_limit; - break; - } + for (pmdidx = 0; pmdidx < PTRS_PER_PMD; pmdidx++) { + struct page *pte; + + if (pgdidx == pgdidx_limit && + pudidx == pudidx_limit && + pmdidx > pmdidx_limit) + goto out; - if (pmd_none(*pmd)) + if (pmd_none(pmd[pmdidx])) continue; - flush |= (*func)(pmd_page(*pmd), PT_PTE); + pte = pmd_page(pmd[pmdidx]); + flush |= (*func)(pte, PT_PTE); } } } - - flush |= (*func)(virt_to_page(pgd_base), PT_PGD); +out: return flush; } @@ -650,6 +677,11 @@ void xen_pgd_pin(pgd_t *pgd) xen_mc_batch(); } +#ifdef CONFIG_X86_PAE + /* Need to make sure unshared kernel PMD is pinnable */ + pin_page(virt_to_page(pgd_page(pgd[pgd_index(TASK_SIZE)])), PT_PMD); +#endif + xen_do_pin(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(pgd))); xen_mc_issue(0); } @@ -731,6 +763,10 @@ static void xen_pgd_unpin(pgd_t *pgd) xen_do_pin(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); +#ifdef CONFIG_X86_PAE + /* Need to make sure unshared kernel PMD is unpinned */ + pin_page(virt_to_page(pgd_page(pgd[pgd_index(TASK_SIZE)])), PT_PMD); +#endif pgd_walk(pgd, unpin_page, TASK_SIZE); xen_mc_issue(0); @@ -750,7 +786,6 @@ void xen_mm_unpin_all(void) list_for_each_entry(page, &pgd_list, lru) { if (PageSavePinned(page)) { BUG_ON(!PagePinned(page)); - printk("unpinning pinned %p\n", page_address(page)); xen_pgd_unpin((pgd_t *)page_address(page)); ClearPageSavePinned(page); } -- cgit v1.2.3 From b7c3c5c15936a40c79ef40af7b3bac801c7feb20 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:07 -0700 Subject: xen: make sure the kernel command line is right Point the boot params cmd_line_ptr to the domain-builder-provided command line. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 3b6b7fcf5b5..0172ba77452 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1587,6 +1587,7 @@ asmlinkage void __init xen_start_kernel(void) boot_params.hdr.ramdisk_image = xen_start_info->mod_start ? __pa(xen_start_info->mod_start) : 0; boot_params.hdr.ramdisk_size = xen_start_info->mod_len; + boot_params.hdr.cmd_line_ptr = __pa(xen_start_info->cmd_line); if (!is_initial_xendomain()) { add_preferred_console("xenboot", 0, NULL); -- cgit v1.2.3 From 4a5c3e77f70b3ea8b361d7fa9eb2e4dad18f70ae Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:09 -0700 Subject: xen64: implement failsafe callback Implement the failsafe callback, so that iret and segment register load exceptions are reported to the kernel. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/kernel/entry_64.S | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 7cc2de79614..6aa6932e21b 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -1365,10 +1365,8 @@ END(do_hypervisor_callback) # with its current contents: any discrepancy means we in category 1. */ ENTRY(xen_failsafe_callback) -#if 1 - ud2a -#else - _frame (RIP-0x30) + framesz = (RIP-0x30) /* workaround buggy gas */ + _frame framesz CFI_REL_OFFSET rcx, 0 CFI_REL_OFFSET r11, 8 movw %ds,%cx @@ -1391,8 +1389,13 @@ ENTRY(xen_failsafe_callback) CFI_RESTORE r11 addq $0x30,%rsp CFI_ADJUST_CFA_OFFSET -0x30 - movq $11,%rdi /* SIGSEGV */ - jmp do_exit + pushq $0 + CFI_ADJUST_CFA_OFFSET 8 + pushq %r11 + CFI_ADJUST_CFA_OFFSET 8 + pushq %rcx + CFI_ADJUST_CFA_OFFSET 8 + jmp general_protection CFI_RESTORE_STATE 1: /* Segment mismatch => Category 1 (Bad segment). Retry the IRET. */ movq (%rsp),%rcx @@ -1406,7 +1409,6 @@ ENTRY(xen_failsafe_callback) SAVE_ALL jmp error_exit CFI_ENDPROC -#endif END(xen_failsafe_callback) #endif /* CONFIG_XEN */ -- cgit v1.2.3 From 8a95408e183b3e4aaf3b6a66fa34bff4db53011b Mon Sep 17 00:00:00 2001 From: Eduardo Habkost Date: Tue, 8 Jul 2008 15:07:10 -0700 Subject: xen64: Clear %fs on xen_load_tls() We need to do this, otherwise we can get a GPF on hypercall return after TLS descriptor is cleared but %fs is still pointing to it. Signed-off-by: Eduardo Habkost Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 31 ++++++++++++++++++++++--------- 1 file changed, 22 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 0172ba77452..c13698faae5 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -364,14 +364,6 @@ static void load_TLS_descriptor(struct thread_struct *t, static void xen_load_tls(struct thread_struct *t, unsigned int cpu) { - xen_mc_batch(); - - load_TLS_descriptor(t, cpu, 0); - load_TLS_descriptor(t, cpu, 1); - load_TLS_descriptor(t, cpu, 2); - - xen_mc_issue(PARAVIRT_LAZY_CPU); - /* * XXX sleazy hack: If we're being called in a lazy-cpu zone, * it means we're in a context switch, and %gs has just been @@ -380,9 +372,30 @@ static void xen_load_tls(struct thread_struct *t, unsigned int cpu) * Either way, it has been saved, and the new value will get * loaded properly. This will go away as soon as Xen has been * modified to not save/restore %gs for normal hypercalls. + * + * On x86_64, this hack is not used for %gs, because gs points + * to KERNEL_GS_BASE (and uses it for PDA references), so we + * must not zero %gs on x86_64 + * + * For x86_64, we need to zero %fs, otherwise we may get an + * exception between the new %fs descriptor being loaded and + * %fs being effectively cleared at __switch_to(). */ - if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_CPU) + if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_CPU) { +#ifdef CONFIG_X86_32 loadsegment(gs, 0); +#else + loadsegment(fs, 0); +#endif + } + + xen_mc_batch(); + + load_TLS_descriptor(t, cpu, 0); + load_TLS_descriptor(t, cpu, 1); + load_TLS_descriptor(t, cpu, 2); + + xen_mc_issue(PARAVIRT_LAZY_CPU); } #ifdef CONFIG_X86_64 -- cgit v1.2.3 From d6182fbf04164016cb6540db02eef3d6bdc967c3 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:13 -0700 Subject: xen64: allocate and manage user pagetables Because the x86_64 architecture does not enforce segment limits, Xen cannot protect itself with them as it does in 32-bit mode. Therefore, to protect itself, it runs the guest kernel in ring 3. Since it also runs the guest userspace in ring3, the guest kernel must maintain a second pagetable for its userspace, which does not map kernel space. Naturally, the guest kernel pagetables map both kernel and userspace. The userspace pagetable is attached to the corresponding kernel pagetable via the pgd's page->private field. It is allocated and freed at the same time as the kernel pgd via the paravirt_pgd_alloc/free hooks. Fortunately, the user pagetable is almost entirely shared with the kernel pagetable; the only difference is the pgd page itself. set_pgd will populate all entries in the kernel pagetable, and also set the corresponding user pgd entry if the address is less than STACK_TOP_MAX. The user pagetable must be pinned and unpinned with the kernel one, but because the pagetables are aliased, pgd_walk() only needs to be called on the kernel pagetable. The user pgd page is then pinned/unpinned along with the kernel pgd page. xen_write_cr3 must write both the kernel and user cr3s. The init_mm.pgd pagetable never has a user pagetable allocated for it, because it can never be used while running usermode. One awkward area is that early in boot the page structures are not available. No user pagetable can exist at that point, but it complicates the logic to avoid looking at the page structure. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 99 ++++++++++++++++++++++++++++++++++++++++-------- arch/x86/xen/mmu.c | 91 +++++++++++++++++++++++++++++++++++++++----- arch/x86/xen/mmu.h | 2 + 3 files changed, 168 insertions(+), 24 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index c13698faae5..48f1a7eca8b 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -46,7 +46,6 @@ #include #include #include -#include #include "xen-ops.h" #include "mmu.h" @@ -711,29 +710,57 @@ static void set_current_cr3(void *v) x86_write_percpu(xen_current_cr3, (unsigned long)v); } -static void xen_write_cr3(unsigned long cr3) +static void __xen_write_cr3(bool kernel, unsigned long cr3) { struct mmuext_op *op; struct multicall_space mcs; - unsigned long mfn = pfn_to_mfn(PFN_DOWN(cr3)); + unsigned long mfn; - BUG_ON(preemptible()); + if (cr3) + mfn = pfn_to_mfn(PFN_DOWN(cr3)); + else + mfn = 0; - mcs = xen_mc_entry(sizeof(*op)); /* disables interrupts */ + WARN_ON(mfn == 0 && kernel); - /* Update while interrupts are disabled, so its atomic with - respect to ipis */ - x86_write_percpu(xen_cr3, cr3); + mcs = __xen_mc_entry(sizeof(*op)); op = mcs.args; - op->cmd = MMUEXT_NEW_BASEPTR; + op->cmd = kernel ? MMUEXT_NEW_BASEPTR : MMUEXT_NEW_USER_BASEPTR; op->arg1.mfn = mfn; MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); - /* Update xen_update_cr3 once the batch has actually - been submitted. */ - xen_mc_callback(set_current_cr3, (void *)cr3); + if (kernel) { + x86_write_percpu(xen_cr3, cr3); + + /* Update xen_current_cr3 once the batch has actually + been submitted. */ + xen_mc_callback(set_current_cr3, (void *)cr3); + } +} + +static void xen_write_cr3(unsigned long cr3) +{ + BUG_ON(preemptible()); + + xen_mc_batch(); /* disables interrupts */ + + /* Update while interrupts are disabled, so its atomic with + respect to ipis */ + x86_write_percpu(xen_cr3, cr3); + + __xen_write_cr3(true, cr3); + +#ifdef CONFIG_X86_64 + { + pgd_t *user_pgd = xen_get_user_pgd(__va(cr3)); + if (user_pgd) + __xen_write_cr3(false, __pa(user_pgd)); + else + __xen_write_cr3(false, 0); + } +#endif xen_mc_issue(PARAVIRT_LAZY_CPU); /* interrupts restored */ } @@ -794,6 +821,40 @@ static void xen_alloc_pmd(struct mm_struct *mm, u32 pfn) xen_alloc_ptpage(mm, pfn, PT_PMD); } +static int xen_pgd_alloc(struct mm_struct *mm) +{ + pgd_t *pgd = mm->pgd; + int ret = 0; + + BUG_ON(PagePinned(virt_to_page(pgd))); + +#ifdef CONFIG_X86_64 + { + struct page *page = virt_to_page(pgd); + + BUG_ON(page->private != 0); + + page->private = __get_free_page(GFP_KERNEL | __GFP_ZERO); + if (page->private == 0) + ret = -ENOMEM; + + BUG_ON(PagePinned(virt_to_page(xen_get_user_pgd(pgd)))); + } +#endif + + return ret; +} + +static void xen_pgd_free(struct mm_struct *mm, pgd_t *pgd) +{ +#ifdef CONFIG_X86_64 + pgd_t *user_pgd = xen_get_user_pgd(pgd); + + if (user_pgd) + free_page((unsigned long)user_pgd); +#endif +} + /* This should never happen until we're OK to use struct page */ static void xen_release_ptpage(u32 pfn, unsigned level) { @@ -1168,8 +1229,8 @@ static const struct pv_mmu_ops xen_mmu_ops __initdata = { .pte_update = paravirt_nop, .pte_update_defer = paravirt_nop, - .pgd_alloc = __paravirt_pgd_alloc, - .pgd_free = paravirt_nop, + .pgd_alloc = xen_pgd_alloc, + .pgd_free = xen_pgd_free, .alloc_pte = xen_alloc_pte_init, .release_pte = xen_release_pte_init, @@ -1480,7 +1541,15 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pf /* Switch over */ pgd = init_level4_pgt; - xen_write_cr3(__pa(pgd)); + + /* + * At this stage there can be no user pgd, and no page + * structure to attach it to, so make sure we just set kernel + * pgd. + */ + xen_mc_batch(); + __xen_write_cr3(true, __pa(pgd)); + xen_mc_issue(PARAVIRT_LAZY_CPU); reserve_early(__pa(xen_start_info->pt_base), __pa(xen_start_info->pt_base + diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 046c1f23dd6..a44d56e38bd 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -58,6 +58,13 @@ #include "multicalls.h" #include "mmu.h" +/* + * Just beyond the highest usermode address. STACK_TOP_MAX has a + * redzone above it, so round it up to a PGD boundary. + */ +#define USER_LIMIT ((STACK_TOP_MAX + PGDIR_SIZE - 1) & PGDIR_MASK) + + #define P2M_ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(unsigned long)) #define TOP_ENTRIES (MAX_DOMAIN_PAGES / P2M_ENTRIES_PER_PAGE) @@ -461,17 +468,45 @@ pud_t xen_make_pud(pudval_t pud) return native_make_pud(pud); } -void xen_set_pgd_hyper(pgd_t *ptr, pgd_t val) +pgd_t *xen_get_user_pgd(pgd_t *pgd) { - struct mmu_update u; + pgd_t *pgd_page = (pgd_t *)(((unsigned long)pgd) & PAGE_MASK); + unsigned offset = pgd - pgd_page; + pgd_t *user_ptr = NULL; - preempt_disable(); + if (offset < pgd_index(USER_LIMIT)) { + struct page *page = virt_to_page(pgd_page); + user_ptr = (pgd_t *)page->private; + if (user_ptr) + user_ptr += offset; + } - xen_mc_batch(); + return user_ptr; +} + +static void __xen_set_pgd_hyper(pgd_t *ptr, pgd_t val) +{ + struct mmu_update u; u.ptr = virt_to_machine(ptr).maddr; u.val = pgd_val_ma(val); extend_mmu_update(&u); +} + +/* + * Raw hypercall-based set_pgd, intended for in early boot before + * there's a page structure. This implies: + * 1. The only existing pagetable is the kernel's + * 2. It is always pinned + * 3. It has no user pagetable attached to it + */ +void __init xen_set_pgd_hyper(pgd_t *ptr, pgd_t val) +{ + preempt_disable(); + + xen_mc_batch(); + + __xen_set_pgd_hyper(ptr, val); xen_mc_issue(PARAVIRT_LAZY_MMU); @@ -480,14 +515,28 @@ void xen_set_pgd_hyper(pgd_t *ptr, pgd_t val) void xen_set_pgd(pgd_t *ptr, pgd_t val) { + pgd_t *user_ptr = xen_get_user_pgd(ptr); + /* If page is not pinned, we can just update the entry directly */ if (!page_pinned(ptr)) { *ptr = val; + if (user_ptr) { + WARN_ON(page_pinned(user_ptr)); + *user_ptr = val; + } return; } - xen_set_pgd_hyper(ptr, val); + /* If it's pinned, then we can at least batch the kernel and + user updates together. */ + xen_mc_batch(); + + __xen_set_pgd_hyper(ptr, val); + if (user_ptr) + __xen_set_pgd_hyper(user_ptr, val); + + xen_mc_issue(PARAVIRT_LAZY_MMU); } #endif /* PAGETABLE_LEVELS == 4 */ @@ -526,7 +575,7 @@ static int pgd_walk(pgd_t *pgd, int (*func)(struct page *, enum pt_level), * space, which contains the Xen mappings. On 32-bit these * will end up making a zero-sized hole and so is a no-op. */ - hole_low = pgd_index(STACK_TOP_MAX + PGDIR_SIZE - 1); + hole_low = pgd_index(USER_LIMIT); hole_high = pgd_index(PAGE_OFFSET); pgdidx_limit = pgd_index(limit); @@ -670,19 +719,31 @@ void xen_pgd_pin(pgd_t *pgd) { xen_mc_batch(); - if (pgd_walk(pgd, pin_page, TASK_SIZE)) { + if (pgd_walk(pgd, pin_page, USER_LIMIT)) { /* re-enable interrupts for kmap_flush_unused */ xen_mc_issue(0); kmap_flush_unused(); xen_mc_batch(); } +#ifdef CONFIG_X86_64 + { + pgd_t *user_pgd = xen_get_user_pgd(pgd); + + xen_do_pin(MMUEXT_PIN_L4_TABLE, PFN_DOWN(__pa(pgd))); + + if (user_pgd) { + pin_page(virt_to_page(user_pgd), PT_PGD); + xen_do_pin(MMUEXT_PIN_L4_TABLE, PFN_DOWN(__pa(user_pgd))); + } + } +#else /* CONFIG_X86_32 */ #ifdef CONFIG_X86_PAE /* Need to make sure unshared kernel PMD is pinnable */ pin_page(virt_to_page(pgd_page(pgd[pgd_index(TASK_SIZE)])), PT_PMD); #endif - xen_do_pin(MMUEXT_PIN_L3_TABLE, PFN_DOWN(__pa(pgd))); +#endif /* CONFIG_X86_64 */ xen_mc_issue(0); } @@ -763,11 +824,23 @@ static void xen_pgd_unpin(pgd_t *pgd) xen_do_pin(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd))); +#ifdef CONFIG_X86_64 + { + pgd_t *user_pgd = xen_get_user_pgd(pgd); + + if (user_pgd) { + xen_do_pin(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(user_pgd))); + unpin_page(virt_to_page(user_pgd), PT_PGD); + } + } +#endif + #ifdef CONFIG_X86_PAE /* Need to make sure unshared kernel PMD is unpinned */ pin_page(virt_to_page(pgd_page(pgd[pgd_index(TASK_SIZE)])), PT_PMD); #endif - pgd_walk(pgd, unpin_page, TASK_SIZE); + + pgd_walk(pgd, unpin_page, USER_LIMIT); xen_mc_issue(0); } diff --git a/arch/x86/xen/mmu.h b/arch/x86/xen/mmu.h index 19d544b0b6c..0f59bd03f9e 100644 --- a/arch/x86/xen/mmu.h +++ b/arch/x86/xen/mmu.h @@ -51,6 +51,8 @@ void xen_set_pgd(pgd_t *pgdp, pgd_t pgd); void xen_set_pgd_hyper(pgd_t *pgdp, pgd_t pgd); #endif +pgd_t *xen_get_user_pgd(pgd_t *pgd); + pte_t xen_ptep_modify_prot_start(struct mm_struct *mm, unsigned long addr, pte_t *ptep); void xen_ptep_modify_prot_commit(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte); -- cgit v1.2.3 From 6fcac6d305e8238939e169f4c52e8ec8a552a31f Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:14 -0700 Subject: xen64: set up syscall and sysenter entrypoints for 64-bit We set up entrypoints for syscall and sysenter. sysenter is only used for 32-bit compat processes, whereas syscall can be used in by both 32 and 64-bit processes. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 4 ++ arch/x86/xen/setup.c | 42 +++++++++++++-- arch/x86/xen/smp.c | 1 + arch/x86/xen/xen-asm_64.S | 129 +++++++++++++++++++++++++++++++++++++++++++++- arch/x86/xen/xen-ops.h | 3 ++ 5 files changed, 174 insertions(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 48f1a7eca8b..87d36044054 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1139,6 +1139,10 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { .iret = xen_iret, .irq_enable_sysexit = xen_sysexit, +#ifdef CONFIG_X86_64 + .usergs_sysret32 = xen_sysret32, + .usergs_sysret64 = xen_sysret64, +#endif .load_tr_desc = paravirt_nop, .set_ldt = xen_set_ldt, diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index bea3d4f779d..9d7a1440289 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -86,9 +86,11 @@ static void xen_idle(void) */ static void __init fiddle_vdso(void) { +#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION) extern const char vdso32_default_start; u32 *mask = VDSO32_SYMBOL(&vdso32_default_start, NOTE_MASK); *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; +#endif } static __cpuinit int register_callback(unsigned type, const void *func) @@ -106,15 +108,48 @@ void __cpuinit xen_enable_sysenter(void) { int cpu = smp_processor_id(); extern void xen_sysenter_target(void); + int ret; + +#ifdef CONFIG_X86_32 + if (!boot_cpu_has(X86_FEATURE_SEP)) { + return; + } +#else + if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL && + boot_cpu_data.x86_vendor != X86_VENDOR_CENTAUR) { + return; + } +#endif - if (!boot_cpu_has(X86_FEATURE_SEP) || - register_callback(CALLBACKTYPE_sysenter, - xen_sysenter_target) != 0) { + ret = register_callback(CALLBACKTYPE_sysenter, xen_sysenter_target); + if(ret != 0) { clear_cpu_cap(&cpu_data(cpu), X86_FEATURE_SEP); clear_cpu_cap(&boot_cpu_data, X86_FEATURE_SEP); } } +void __cpuinit xen_enable_syscall(void) +{ +#ifdef CONFIG_X86_64 + int cpu = smp_processor_id(); + int ret; + extern void xen_syscall_target(void); + extern void xen_syscall32_target(void); + + ret = register_callback(CALLBACKTYPE_syscall, xen_syscall_target); + if (ret != 0) { + printk("failed to set syscall: %d\n", ret); + clear_cpu_cap(&cpu_data(cpu), X86_FEATURE_SYSCALL); + clear_cpu_cap(&boot_cpu_data, X86_FEATURE_SYSCALL); + } else { + ret = register_callback(CALLBACKTYPE_syscall32, + xen_syscall32_target); + if (ret != 0) + printk("failed to set 32-bit syscall: %d\n", ret); + } +#endif /* CONFIG_X86_64 */ +} + void __init xen_arch_setup(void) { struct physdev_set_iopl set_iopl; @@ -131,6 +166,7 @@ void __init xen_arch_setup(void) BUG(); xen_enable_sysenter(); + xen_enable_syscall(); set_iopl.iopl = 1; rc = HYPERVISOR_physdev_op(PHYSDEVOP_set_iopl, &set_iopl); diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index 8310ca0ea37..f702199312a 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -69,6 +69,7 @@ static __cpuinit void cpu_bringup_and_idle(void) preempt_disable(); xen_enable_sysenter(); + xen_enable_syscall(); cpu = smp_processor_id(); smp_store_cpu_info(cpu); diff --git a/arch/x86/xen/xen-asm_64.S b/arch/x86/xen/xen-asm_64.S index b147b495dae..4038cbfe333 100644 --- a/arch/x86/xen/xen-asm_64.S +++ b/arch/x86/xen/xen-asm_64.S @@ -15,6 +15,8 @@ #include #include +#include +#include #include @@ -138,9 +140,132 @@ ENTRY(xen_adjust_exception_frame) mov 8+8(%rsp),%r11 ret $16 +hypercall_iret = hypercall_page + __HYPERVISOR_iret * 32 +/* + Xen64 iret frame: + + ss + rsp + rflags + cs + rip <-- standard iret frame + + flags + + rcx } + r11 }<-- pushed by hypercall page +rsp -> rax } + */ ENTRY(xen_iret) pushq $0 - jmp hypercall_page + __HYPERVISOR_iret * 32 +1: jmp hypercall_iret +ENDPATCH(xen_iret) +RELOC(xen_iret, 1b+1) +/* + sysexit is not used for 64-bit processes, so it's + only ever used to return to 32-bit compat userspace. + */ ENTRY(xen_sysexit) - ud2a + pushq $__USER32_DS + pushq %rcx + pushq $X86_EFLAGS_IF + pushq $__USER32_CS + pushq %rdx + + pushq $VGCF_in_syscall +1: jmp hypercall_iret +ENDPATCH(xen_sysexit) +RELOC(xen_sysexit, 1b+1) + +ENTRY(xen_sysret64) + /* We're already on the usermode stack at this point, but still + with the kernel gs, so we can easily switch back */ + movq %rsp, %gs:pda_oldrsp + movq %gs:pda_kernelstack,%rsp + + pushq $__USER_DS + pushq %gs:pda_oldrsp + pushq %r11 + pushq $__USER_CS + pushq %rcx + + pushq $VGCF_in_syscall +1: jmp hypercall_iret +ENDPATCH(xen_sysret64) +RELOC(xen_sysret64, 1b+1) + +ENTRY(xen_sysret32) + /* We're already on the usermode stack at this point, but still + with the kernel gs, so we can easily switch back */ + movq %rsp, %gs:pda_oldrsp + movq %gs:pda_kernelstack, %rsp + + pushq $__USER32_DS + pushq %gs:pda_oldrsp + pushq %r11 + pushq $__USER32_CS + pushq %rcx + + pushq $VGCF_in_syscall +1: jmp hypercall_iret +ENDPATCH(xen_sysret32) +RELOC(xen_sysret32, 1b+1) + +/* + Xen handles syscall callbacks much like ordinary exceptions, + which means we have: + - kernel gs + - kernel rsp + - an iret-like stack frame on the stack (including rcx and r11): + ss + rsp + rflags + cs + rip + r11 + rsp-> rcx + + In all the entrypoints, we undo all that to make it look + like a CPU-generated syscall/sysenter and jump to the normal + entrypoint. + */ + +.macro undo_xen_syscall + mov 0*8(%rsp),%rcx + mov 1*8(%rsp),%r11 + mov 5*8(%rsp),%rsp +.endm + +/* Normal 64-bit system call target */ +ENTRY(xen_syscall_target) + undo_xen_syscall + jmp system_call_after_swapgs +ENDPROC(xen_syscall_target) + +#ifdef CONFIG_IA32_EMULATION + +/* 32-bit compat syscall target */ +ENTRY(xen_syscall32_target) + undo_xen_syscall + jmp ia32_cstar_target +ENDPROC(xen_syscall32_target) + +/* 32-bit compat sysenter target */ +ENTRY(xen_sysenter_target) + undo_xen_syscall + jmp ia32_sysenter_target +ENDPROC(xen_sysenter_target) + +#else /* !CONFIG_IA32_EMULATION */ + +ENTRY(xen_syscall32_target) +ENTRY(xen_sysenter_target) + lea 16(%rsp), %rsp /* strip %rcx,%r11 */ + mov $-ENOSYS, %rax + pushq $VGCF_in_syscall + jmp hypercall_iret +ENDPROC(xen_syscall32_target) +ENDPROC(xen_sysenter_target) + +#endif /* CONFIG_IA32_EMULATION */ diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h index c4800a2c5a4..dd3c23152a2 100644 --- a/arch/x86/xen/xen-ops.h +++ b/arch/x86/xen/xen-ops.h @@ -26,6 +26,7 @@ char * __init xen_memory_setup(void); void __init xen_arch_setup(void); void __init xen_init_IRQ(void); void xen_enable_sysenter(void); +void xen_enable_syscall(void); void xen_vcpu_restore(void); void __init xen_build_dynamic_phys_to_machine(void); @@ -70,6 +71,8 @@ DECL_ASM(void, xen_restore_fl_direct, unsigned long); /* These are not functions, and cannot be called normally */ void xen_iret(void); void xen_sysexit(void); +void xen_sysret32(void); +void xen_sysret64(void); void xen_adjust_exception_frame(void); #endif /* XEN_OPS_H */ -- cgit v1.2.3 From bf18bf94dc72db998d0fbebc846c07c858a59c90 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:15 -0700 Subject: xen64: set up userspace syscall patch 64-bit userspace expects the vdso to be mapped at a specific fixed address, which happens to be in the middle of the kernel address space. Because we have split user and kernel pagetables, we need to make special arrangements for the vsyscall mapping to appear in the kernel part of the user pagetable. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 46 ++++++++++++++++++++++++++++++++++++---------- 1 file changed, 36 insertions(+), 10 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 87d36044054..f64b8729cd0 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -56,6 +56,18 @@ EXPORT_SYMBOL_GPL(hypercall_page); DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu); DEFINE_PER_CPU(struct vcpu_info, xen_vcpu_info); +/* + * Identity map, in addition to plain kernel map. This needs to be + * large enough to allocate page table pages to allocate the rest. + * Each page can map 2MB. + */ +static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss; + +#ifdef CONFIG_X86_64 +/* l3 pud for userspace vsyscall mapping */ +static pud_t level3_user_vsyscall[PTRS_PER_PUD] __page_aligned_bss; +#endif /* CONFIG_X86_64 */ + /* * Note about cr3 (pagetable base) values: * @@ -831,12 +843,20 @@ static int xen_pgd_alloc(struct mm_struct *mm) #ifdef CONFIG_X86_64 { struct page *page = virt_to_page(pgd); + pgd_t *user_pgd; BUG_ON(page->private != 0); - page->private = __get_free_page(GFP_KERNEL | __GFP_ZERO); - if (page->private == 0) - ret = -ENOMEM; + ret = -ENOMEM; + + user_pgd = (pgd_t *)__get_free_page(GFP_KERNEL | __GFP_ZERO); + page->private = (unsigned long)user_pgd; + + if (user_pgd != NULL) { + user_pgd[pgd_index(VSYSCALL_START)] = + __pgd(__pa(level3_user_vsyscall) | _PAGE_TABLE); + ret = 0; + } BUG_ON(PagePinned(virt_to_page(xen_get_user_pgd(pgd)))); } @@ -977,6 +997,9 @@ static __init void xen_post_allocator_init(void) pv_mmu_ops.release_pud = xen_release_pud; #endif +#ifdef CONFIG_X86_64 + SetPagePinned(virt_to_page(level3_user_vsyscall)); +#endif xen_mark_init_mm_pinned(); } @@ -1088,6 +1111,15 @@ static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot) } __native_set_fixmap(idx, pte); + +#ifdef CONFIG_X86_64 + /* Replicate changes to map the vsyscall page into the user + pagetable vsyscall mapping. */ + if (idx >= VSYSCALL_LAST_PAGE && idx <= VSYSCALL_FIRST_PAGE) { + unsigned long vaddr = __fix_to_virt(idx); + set_pte_vaddr_pud(level3_user_vsyscall, vaddr, pte); + } +#endif } static const struct pv_info xen_info __initdata = { @@ -1427,13 +1459,6 @@ static void set_page_prot(void *addr, pgprot_t prot) BUG(); } -/* - * Identity map, in addition to plain kernel map. This needs to be - * large enough to allocate page table pages to allocate the rest. - * Each page can map 2MB. - */ -static pte_t level1_ident_pgt[PTRS_PER_PTE * 4] __page_aligned_bss; - static __init void xen_map_identity_early(pmd_t *pmd, unsigned long max_pfn) { unsigned pmdidx, pteidx; @@ -1533,6 +1558,7 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pf set_page_prot(init_level4_pgt, PAGE_KERNEL_RO); set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO); set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO); + set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO); set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO); set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO); -- cgit v1.2.3 From 1153968a48e3ca3e2b7a437e8b82ec9e6f768e24 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:16 -0700 Subject: xen: implement Xen write_msr operation 64-bit uses MSRs for important things like the base for fs and gs-prefixed addresses. It's more efficient to use a hypercall to update these, rather than go via the trap and emulate path. Other MSR writes are just passed through; in an unprivileged domain they do nothing, but it might be useful later. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 31 ++++++++++++++++++++++++++++++- 1 file changed, 30 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index f64b8729cd0..776c0fb77d6 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -41,6 +41,7 @@ #include #include #include +#include #include #include #include @@ -777,6 +778,34 @@ static void xen_write_cr3(unsigned long cr3) xen_mc_issue(PARAVIRT_LAZY_CPU); /* interrupts restored */ } +static int xen_write_msr_safe(unsigned int msr, unsigned low, unsigned high) +{ + int ret; + + ret = 0; + + switch(msr) { +#ifdef CONFIG_X86_64 + unsigned which; + u64 base; + + case MSR_FS_BASE: which = SEGBASE_FS; goto set; + case MSR_KERNEL_GS_BASE: which = SEGBASE_GS_USER; goto set; + case MSR_GS_BASE: which = SEGBASE_GS_KERNEL; goto set; + + set: + base = ((u64)high << 32) | low; + if (HYPERVISOR_set_segment_base(which, base) != 0) + ret = -EFAULT; + break; +#endif + default: + ret = native_write_msr_safe(msr, low, high); + } + + return ret; +} + /* Early in boot, while setting up the initial pagetable, assume everything is pinned. */ static __init void xen_alloc_pte_init(struct mm_struct *mm, u32 pfn) @@ -1165,7 +1194,7 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = { .wbinvd = native_wbinvd, .read_msr = native_read_msr_safe, - .write_msr = native_write_msr_safe, + .write_msr = xen_write_msr_safe, .read_tsc = native_read_tsc, .read_pmc = native_read_pmc, -- cgit v1.2.3 From 51dd660a2cd6eab4d470cfe1009c7f473832b786 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 8 Jul 2008 15:07:17 -0700 Subject: xen: update Kconfig to allow 64-bit Xen Allow Xen to be enabled on 64-bit. Also extend domain size limit from 8 GB (on 32-bit) to 32 GB on 64-bit. Signed-off-by: Jeremy Fitzhardinge Cc: Stephen Tweedie Cc: Eduardo Habkost Cc: Mark McLoughlin Signed-off-by: Ingo Molnar --- arch/x86/xen/Kconfig | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig index c2cc9958087..20b49729bed 100644 --- a/arch/x86/xen/Kconfig +++ b/arch/x86/xen/Kconfig @@ -6,8 +6,8 @@ config XEN bool "Xen guest support" select PARAVIRT select PARAVIRT_CLOCK - depends on X86_32 - depends on X86_CMPXCHG && X86_TSC && X86_PAE && !(X86_VISWS || X86_VOYAGER) + depends on X86_64 || (X86_32 && X86_PAE && !(X86_VISWS || X86_VOYAGER)) + depends on X86_CMPXCHG && X86_TSC help This is the Linux Xen port. Enabling this will allow the kernel to boot in a paravirtualized environment under the @@ -15,10 +15,11 @@ config XEN config XEN_MAX_DOMAIN_MEMORY int "Maximum allowed size of a domain in gigabytes" - default 8 + default 8 if X86_32 + default 32 if X86_64 depends on XEN help The pseudo-physical to machine address array is sized according to the maximum possible memory size of a Xen domain. This array uses 1 page per gigabyte, so there's no - need to be too stingy here. \ No newline at end of file + need to be too stingy here. -- cgit v1.2.3 From b3fe124389f9dd97f0bbd954da2910e286648f0f Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 9 Jul 2008 13:45:33 +0200 Subject: xen64: fix build error on 32-bit + !HIGHMEM fix: arch/x86/xen/enlighten.c: In function 'xen_set_fixmap': arch/x86/xen/enlighten.c:1127: error: 'FIX_KMAP_BEGIN' undeclared (first use in this function) arch/x86/xen/enlighten.c:1127: error: (Each undeclared identifier is reported only once arch/x86/xen/enlighten.c:1127: error: for each function it appears in.) arch/x86/xen/enlighten.c:1127: error: 'FIX_KMAP_END' undeclared (first use in this function) make[1]: *** [arch/x86/xen/enlighten.o] Error 1 make: *** [arch/x86/xen/enlighten.o] Error 2 FIX_KMAP_BEGIN is only available on HIGHMEM. Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 776c0fb77d6..3da6acb7eaf 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1124,7 +1124,9 @@ static void xen_set_fixmap(unsigned idx, unsigned long phys, pgprot_t prot) #ifdef CONFIG_X86_32 case FIX_WP_TEST: case FIX_VDSO: +# ifdef CONFIG_HIGHMEM case FIX_KMAP_BEGIN ... FIX_KMAP_END: +# endif #else case VSYSCALL_LAST_PAGE ... VSYSCALL_FIRST_PAGE: #endif -- cgit v1.2.3 From 6596f2422306a05be2170efc114da49f26a047dd Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 9 Jul 2008 22:32:33 +0200 Subject: Revert "x86_64: there's no need to preallocate level1_fixmap_pgt" This reverts commit 033786969d1d1b5af12a32a19d3a760314d05329. Suresh Siddha reported that this broke booting on his 2GB testbox. Reported-by: Suresh Siddha Signed-off-by: Ingo Molnar --- arch/x86/kernel/head_64.S | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S index 2240f823676..db3280afe88 100644 --- a/arch/x86/kernel/head_64.S +++ b/arch/x86/kernel/head_64.S @@ -362,6 +362,12 @@ NEXT_PAGE(level3_kernel_pgt) .quad level2_fixmap_pgt - __START_KERNEL_map + _PAGE_TABLE NEXT_PAGE(level2_fixmap_pgt) + .fill 506,8,0 + .quad level1_fixmap_pgt - __START_KERNEL_map + _PAGE_TABLE + /* 8MB reserved for vsyscalls + a 2MB hole = 4 + 1 entries */ + .fill 5,8,0 + +NEXT_PAGE(level1_fixmap_pgt) .fill 512,8,0 NEXT_PAGE(level2_ident_pgt) -- cgit v1.2.3 From 62541c376668042e20122864a044360707b2fb82 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Thu, 10 Jul 2008 16:24:08 -0700 Subject: xen64: disable 32-bit syscall/sysenter if not supported. Old versions of Xen (3.1 and before) don't support sysenter or syscall from 32-bit compat userspaces. If we can't set the appropriate syscall callback, then disable the corresponding feature bit, which will cause the vdso32 setup to fall back appropriately. Linux assumes that syscall is always available to 32-bit userspace, and installs it by default if sysenter isn't available. In that case, we just disable vdso altogether, forcing userspace libc to fall back to int $0x80. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 38 +++++++++++++++++++------------------- 1 file changed, 19 insertions(+), 19 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index 9d7a1440289..9cce4a92aac 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -106,46 +106,46 @@ static __cpuinit int register_callback(unsigned type, const void *func) void __cpuinit xen_enable_sysenter(void) { - int cpu = smp_processor_id(); extern void xen_sysenter_target(void); int ret; + unsigned sysenter_feature; #ifdef CONFIG_X86_32 - if (!boot_cpu_has(X86_FEATURE_SEP)) { - return; - } + sysenter_feature = X86_FEATURE_SEP; #else - if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL && - boot_cpu_data.x86_vendor != X86_VENDOR_CENTAUR) { - return; - } + sysenter_feature = X86_FEATURE_SYSENTER32; #endif + if (!boot_cpu_has(sysenter_feature)) + return; + ret = register_callback(CALLBACKTYPE_sysenter, xen_sysenter_target); - if(ret != 0) { - clear_cpu_cap(&cpu_data(cpu), X86_FEATURE_SEP); - clear_cpu_cap(&boot_cpu_data, X86_FEATURE_SEP); - } + if(ret != 0) + setup_clear_cpu_cap(sysenter_feature); } void __cpuinit xen_enable_syscall(void) { #ifdef CONFIG_X86_64 - int cpu = smp_processor_id(); int ret; extern void xen_syscall_target(void); extern void xen_syscall32_target(void); ret = register_callback(CALLBACKTYPE_syscall, xen_syscall_target); if (ret != 0) { - printk("failed to set syscall: %d\n", ret); - clear_cpu_cap(&cpu_data(cpu), X86_FEATURE_SYSCALL); - clear_cpu_cap(&boot_cpu_data, X86_FEATURE_SYSCALL); - } else { + printk(KERN_ERR "Failed to set syscall: %d\n", ret); + /* Pretty fatal; 64-bit userspace has no other + mechanism for syscalls. */ + } + + if (boot_cpu_has(X86_FEATURE_SYSCALL32)) { ret = register_callback(CALLBACKTYPE_syscall32, xen_syscall32_target); - if (ret != 0) - printk("failed to set 32-bit syscall: %d\n", ret); + if (ret != 0) { + printk(KERN_INFO "Xen: 32-bit syscall not supported: disabling vdso\n"); + setup_clear_cpu_cap(X86_FEATURE_SYSCALL32); + sysctl_vsyscall32 = 0; + } } #endif /* CONFIG_X86_64 */ } -- cgit v1.2.3 From 71415c6a0877d5944d5dc3060f3b03513746158d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 11 Jul 2008 22:41:34 +0200 Subject: x86, xen, vdso: fix build error fix: arch/x86/xen/built-in.o: In function `xen_enable_syscall': (.cpuinit.text+0xdb): undefined reference to `sysctl_vsyscall32' Signed-off-by: Ingo Molnar --- arch/x86/xen/setup.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index 9cce4a92aac..3e11779755c 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -144,7 +144,9 @@ void __cpuinit xen_enable_syscall(void) if (ret != 0) { printk(KERN_INFO "Xen: 32-bit syscall not supported: disabling vdso\n"); setup_clear_cpu_cap(X86_FEATURE_SYSCALL32); +#ifdef CONFIG_COMPAT sysctl_vsyscall32 = 0; +#endif } } #endif /* CONFIG_X86_64 */ -- cgit v1.2.3 From 6a52e4b1cddd90fbfde8fb67021657936ee74b07 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 12 Jul 2008 02:22:00 -0700 Subject: x86_64: further cleanup of 32-bit compat syscall mechanisms AMD only supports "syscall" from 32-bit compat usermode. Intel and Centaur(?) only support "sysenter" from 32-bit compat usermode. Set the X86 feature bits accordingly, and set up the vdso in accordance with those bits. On the offchance we run on in a 64-bit environment which supports neither syscall nor sysenter from 32-bit mode, then fall back to the int $0x80 vdso. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/cpu/amd_64.c | 2 ++ arch/x86/kernel/cpu/common_64.c | 3 --- arch/x86/vdso/Makefile | 2 +- arch/x86/vdso/vdso32-setup.c | 19 +++++++++---------- arch/x86/vdso/vdso32.S | 13 ++++++++----- arch/x86/xen/setup.c | 10 +++++++--- 6 files changed, 27 insertions(+), 22 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/amd_64.c b/arch/x86/kernel/cpu/amd_64.c index 7c36fb8a28d..d1692b2a41f 100644 --- a/arch/x86/kernel/cpu/amd_64.c +++ b/arch/x86/kernel/cpu/amd_64.c @@ -115,6 +115,8 @@ static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) /* c->x86_power is 8000_0007 edx. Bit 8 is constant TSC */ if (c->x86_power & (1<<8)) set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC); + + set_cpu_cap(c, X86_FEATURE_SYSCALL32); } static void __cpuinit init_amd(struct cpuinfo_x86 *c) diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index 15419cd3c5a..736f50fa433 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -317,9 +317,6 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c) c->x86_phys_bits = eax & 0xff; } - /* Assume all 64-bit CPUs support 32-bit syscall */ - set_cpu_cap(c, X86_FEATURE_SYSCALL32); - if (c->x86_vendor != X86_VENDOR_UNKNOWN && cpu_devs[c->x86_vendor]->c_early_init) cpu_devs[c->x86_vendor]->c_early_init(c); diff --git a/arch/x86/vdso/Makefile b/arch/x86/vdso/Makefile index b7ad9f89d21..4d6ef0a336d 100644 --- a/arch/x86/vdso/Makefile +++ b/arch/x86/vdso/Makefile @@ -62,7 +62,7 @@ $(obj)/%-syms.lds: $(obj)/%.so.dbg FORCE # Build multiple 32-bit vDSO images to choose from at boot time. # obj-$(VDSO32-y) += vdso32-syms.lds -vdso32.so-$(CONFIG_X86_32) += int80 +vdso32.so-$(VDSO32-y) += int80 vdso32.so-$(CONFIG_COMPAT) += syscall vdso32.so-$(VDSO32-y) += sysenter diff --git a/arch/x86/vdso/vdso32-setup.c b/arch/x86/vdso/vdso32-setup.c index 0bce5429a51..513f330c583 100644 --- a/arch/x86/vdso/vdso32-setup.c +++ b/arch/x86/vdso/vdso32-setup.c @@ -193,17 +193,12 @@ static __init void relocate_vdso(Elf32_Ehdr *ehdr) } } -/* - * These symbols are defined by vdso32.S to mark the bounds - * of the ELF DSO images included therein. - */ -extern const char vdso32_default_start, vdso32_default_end; -extern const char vdso32_sysenter_start, vdso32_sysenter_end; static struct page *vdso32_pages[1]; #ifdef CONFIG_X86_64 #define vdso32_sysenter() (boot_cpu_has(X86_FEATURE_SYSENTER32)) +#define vdso32_syscall() (boot_cpu_has(X86_FEATURE_SYSCALL32)) /* May not be __init: called during resume */ void syscall32_cpu_init(void) @@ -226,6 +221,7 @@ static inline void map_compat_vdso(int map) #else /* CONFIG_X86_32 */ #define vdso32_sysenter() (boot_cpu_has(X86_FEATURE_SEP)) +#define vdso32_syscall() (0) void enable_sep_cpu(void) { @@ -296,12 +292,15 @@ int __init sysenter_setup(void) gate_vma_init(); #endif - if (!vdso32_sysenter()) { - vsyscall = &vdso32_default_start; - vsyscall_len = &vdso32_default_end - &vdso32_default_start; - } else { + if (vdso32_syscall()) { + vsyscall = &vdso32_syscall_start; + vsyscall_len = &vdso32_syscall_end - &vdso32_syscall_start; + } else if (vdso32_sysenter()){ vsyscall = &vdso32_sysenter_start; vsyscall_len = &vdso32_sysenter_end - &vdso32_sysenter_start; + } else { + vsyscall = &vdso32_int80_start; + vsyscall_len = &vdso32_int80_end - &vdso32_int80_start; } memcpy(syscall_page, vsyscall, vsyscall_len); diff --git a/arch/x86/vdso/vdso32.S b/arch/x86/vdso/vdso32.S index 1e36f72cab8..2ce5f82c333 100644 --- a/arch/x86/vdso/vdso32.S +++ b/arch/x86/vdso/vdso32.S @@ -2,14 +2,17 @@ __INITDATA - .globl vdso32_default_start, vdso32_default_end -vdso32_default_start: -#ifdef CONFIG_X86_32 + .globl vdso32_int80_start, vdso32_int80_end +vdso32_int80_start: .incbin "arch/x86/vdso/vdso32-int80.so" -#else +vdso32_int80_end: + + .globl vdso32_syscall_start, vdso32_syscall_end +vdso32_syscall_start: +#ifdef CONFIG_COMPAT .incbin "arch/x86/vdso/vdso32-syscall.so" #endif -vdso32_default_end: +vdso32_syscall_end: .globl vdso32_sysenter_start, vdso32_sysenter_end vdso32_sysenter_start: diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index 3e11779755c..e3648e64a63 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -83,12 +83,16 @@ static void xen_idle(void) /* * Set the bit indicating "nosegneg" library variants should be used. + * We only need to bother in pure 32-bit mode; compat 32-bit processes + * can have un-truncated segments, so wrapping around is allowed. */ static void __init fiddle_vdso(void) { -#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION) - extern const char vdso32_default_start; - u32 *mask = VDSO32_SYMBOL(&vdso32_default_start, NOTE_MASK); +#ifdef CONFIG_X86_32 + u32 *mask; + mask = VDSO32_SYMBOL(&vdso32_int80_start, NOTE_MASK); + *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; + mask = VDSO32_SYMBOL(&vdso32_sysenter_start, NOTE_MASK); *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; #endif } -- cgit v1.2.3 From d5303b811b9d6dad2e7396d545eb7db414d42a61 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 12 Jul 2008 02:22:06 -0700 Subject: x86: xen: no need to disable vdso32 Now that the vdso32 code can cope with both syscall and sysenter missing for 32-bit compat processes, just disable the features without disabling vdso altogether. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/xen/setup.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/setup.c b/arch/x86/xen/setup.c index e3648e64a63..b6acc3a0af4 100644 --- a/arch/x86/xen/setup.c +++ b/arch/x86/xen/setup.c @@ -137,7 +137,7 @@ void __cpuinit xen_enable_syscall(void) ret = register_callback(CALLBACKTYPE_syscall, xen_syscall_target); if (ret != 0) { - printk(KERN_ERR "Failed to set syscall: %d\n", ret); + printk(KERN_ERR "Failed to set syscall callback: %d\n", ret); /* Pretty fatal; 64-bit userspace has no other mechanism for syscalls. */ } @@ -145,13 +145,8 @@ void __cpuinit xen_enable_syscall(void) if (boot_cpu_has(X86_FEATURE_SYSCALL32)) { ret = register_callback(CALLBACKTYPE_syscall32, xen_syscall32_target); - if (ret != 0) { - printk(KERN_INFO "Xen: 32-bit syscall not supported: disabling vdso\n"); + if (ret != 0) setup_clear_cpu_cap(X86_FEATURE_SYSCALL32); -#ifdef CONFIG_COMPAT - sysctl_vsyscall32 = 0; -#endif - } } #endif /* CONFIG_X86_64 */ } -- cgit v1.2.3 From 094029479be8eb380447f42eff1b35362ef1a464 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Sat, 12 Jul 2008 02:22:12 -0700 Subject: x86_64: adjust exception frame on paranoid exceptions Exceptions using paranoidentry need to have their exception frames adjusted explicitly. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/entry_64.S | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index 6aa6932e21b..80d5663db3b 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -1189,6 +1189,7 @@ END(device_not_available) /* runs on exception stack */ KPROBE_ENTRY(debug) INTR_FRAME + PARAVIRT_ADJUST_EXCEPTION_FRAME pushq $0 CFI_ADJUST_CFA_OFFSET 8 paranoidentry do_debug, DEBUG_STACK @@ -1198,6 +1199,7 @@ KPROBE_END(debug) /* runs on exception stack */ KPROBE_ENTRY(nmi) INTR_FRAME + PARAVIRT_ADJUST_EXCEPTION_FRAME pushq $-1 CFI_ADJUST_CFA_OFFSET 8 paranoidentry do_nmi, 0, 0 @@ -1211,6 +1213,7 @@ KPROBE_END(nmi) KPROBE_ENTRY(int3) INTR_FRAME + PARAVIRT_ADJUST_EXCEPTION_FRAME pushq $0 CFI_ADJUST_CFA_OFFSET 8 paranoidentry do_int3, DEBUG_STACK @@ -1237,6 +1240,7 @@ END(coprocessor_segment_overrun) /* runs on exception stack */ ENTRY(double_fault) XCPT_FRAME + PARAVIRT_ADJUST_EXCEPTION_FRAME paranoidentry do_double_fault jmp paranoid_exit1 CFI_ENDPROC @@ -1253,6 +1257,7 @@ END(segment_not_present) /* runs on exception stack */ ENTRY(stack_segment) XCPT_FRAME + PARAVIRT_ADJUST_EXCEPTION_FRAME paranoidentry do_stack_segment jmp paranoid_exit1 CFI_ENDPROC @@ -1278,6 +1283,7 @@ END(spurious_interrupt_bug) /* runs on exception stack */ ENTRY(machine_check) INTR_FRAME + PARAVIRT_ADJUST_EXCEPTION_FRAME pushq $0 CFI_ADJUST_CFA_OFFSET 8 paranoidentry do_machine_check -- cgit v1.2.3 From 74d4affde8feb8d5bdebf7fba8e90e4eae3b7b1d Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 7 Jul 2008 12:07:50 -0700 Subject: x86/paravirt: add hooks for spinlock operations Ticket spinlocks have absolutely ghastly worst-case performance characteristics in a virtual environment. If there is any contention for physical CPUs (ie, there are more runnable vcpus than cpus), then ticket locks can cause the system to end up spending 90+% of its time spinning. The problem is that (v)cpus waiting on a ticket spinlock will be granted access to the lock in strict order they got their tickets. If the hypervisor scheduler doesn't give the vcpus time in that order, they will burn timeslices waiting for the scheduler to give the right vcpu some time. In the worst case it could take O(n^2) vcpu scheduler timeslices for everyone waiting on the lock to get it, not counting new cpus trying to take the lock while the log-jam is sorted out. These hooks allow a paravirt backend to replace the spinlock implementation. At the very least, this could revert the implementation back to the old lock algorithm, which allows the next scheduled vcpu to take the lock, and has basically fairly good performance. It also allows the spinlocks to take advantages of the hypervisor features to make locks more efficient (spin and block, for example). The cost to native execution is an extra direct call when using a spinlock function. There's no overhead if CONFIG_PARAVIRT is turned off. The lock structure is fixed at a single "unsigned int", initialized to zero, but the spinlock implementation can use it as it wishes. Thanks to Thomas Friebel's Xen Summit talk "Preventing Guests from Spinning Around" for pointing out this problem. Signed-off-by: Jeremy Fitzhardinge Cc: Jens Axboe Cc: Peter Zijlstra Cc: Christoph Lameter Cc: Petr Tesarik Cc: Virtualization Cc: Xen devel Cc: Thomas Friebel Cc: Nick Piggin Signed-off-by: Ingo Molnar --- arch/x86/kernel/paravirt.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index 2963ab5d91e..f3381686870 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -124,6 +124,7 @@ static void *get_call_destination(u8 type) .pv_irq_ops = pv_irq_ops, .pv_apic_ops = pv_apic_ops, .pv_mmu_ops = pv_mmu_ops, + .pv_lock_ops = pv_lock_ops, }; return *((void **)&tmpl + type); } @@ -450,6 +451,15 @@ struct pv_mmu_ops pv_mmu_ops = { .set_fixmap = native_set_fixmap, }; +struct pv_lock_ops pv_lock_ops = { + .spin_is_locked = __ticket_spin_is_locked, + .spin_is_contended = __ticket_spin_is_contended, + + .spin_lock = __ticket_spin_lock, + .spin_trylock = __ticket_spin_trylock, + .spin_unlock = __ticket_spin_unlock, +}; + EXPORT_SYMBOL_GPL(pv_time_ops); EXPORT_SYMBOL (pv_cpu_ops); EXPORT_SYMBOL (pv_mmu_ops); -- cgit v1.2.3 From 8efcbab674de2bee45a2e4cdf97de16b8e609ac8 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 7 Jul 2008 12:07:51 -0700 Subject: paravirt: introduce a "lock-byte" spinlock implementation Implement a version of the old spinlock algorithm, in which everyone spins waiting for a lock byte. In order to be compatible with the ticket-lock's use of a zero initializer, this uses the convention of '0' for unlocked and '1' for locked. This algorithm is much better than ticket locks in a virtual envionment, because it doesn't interact badly with the vcpu scheduler. If there are multiple vcpus spinning on a lock and the lock is released, the next vcpu to be scheduled will take the lock, rather than cycling around until the next ticketed vcpu gets it. To use this, you must call paravirt_use_bytelocks() very early, before any spinlocks have been taken. Signed-off-by: Jeremy Fitzhardinge Cc: Jens Axboe Cc: Peter Zijlstra Cc: Christoph Lameter Cc: Petr Tesarik Cc: Virtualization Cc: Xen devel Cc: Thomas Friebel Cc: Nick Piggin Signed-off-by: Ingo Molnar --- arch/x86/kernel/paravirt.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index f3381686870..bba4041bb7f 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -268,6 +268,15 @@ enum paravirt_lazy_mode paravirt_get_lazy_mode(void) return __get_cpu_var(paravirt_lazy_mode); } +void __init paravirt_use_bytelocks(void) +{ + pv_lock_ops.spin_is_locked = __byte_spin_is_locked; + pv_lock_ops.spin_is_contended = __byte_spin_is_contended; + pv_lock_ops.spin_lock = __byte_spin_lock; + pv_lock_ops.spin_trylock = __byte_spin_trylock; + pv_lock_ops.spin_unlock = __byte_spin_unlock; +} + struct pv_info pv_info = { .name = "bare hardware", .paravirt_enabled = 0, -- cgit v1.2.3 From 56397f8dadb40055479a8ffff23f21a890098a31 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 7 Jul 2008 12:07:52 -0700 Subject: xen: use lock-byte spinlock implementation Switch to using the lock-byte spinlock implementation, to avoid the worst of the performance hit from ticket locks. Signed-off-by: Jeremy Fitzhardinge Cc: Jens Axboe Cc: Peter Zijlstra Cc: Christoph Lameter Cc: Petr Tesarik Cc: Virtualization Cc: Xen devel Cc: Thomas Friebel Signed-off-by: Ingo Molnar --- arch/x86/xen/smp.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index f702199312a..a8ebafc09d4 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -430,4 +430,5 @@ void __init xen_smp_init(void) { smp_ops = xen_smp_ops; xen_fill_possible_map(); + paravirt_use_bytelocks(); } -- cgit v1.2.3 From 2d9e1e2f58b5612aa4eab0ab54c84308a29dbd79 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 7 Jul 2008 12:07:53 -0700 Subject: xen: implement Xen-specific spinlocks The standard ticket spinlocks are very expensive in a virtual environment, because their performance depends on Xen's scheduler giving vcpus time in the order that they're supposed to take the spinlock. This implements a Xen-specific spinlock, which should be much more efficient. The fast-path is essentially the old Linux-x86 locks, using a single lock byte. The locker decrements the byte; if the result is 0, then they have the lock. If the lock is negative, then locker must spin until the lock is positive again. When there's contention, the locker spin for 2^16[*] iterations waiting to get the lock. If it fails to get the lock in that time, it adds itself to the contention count in the lock and blocks on a per-cpu event channel. When unlocking the spinlock, the locker looks to see if there's anyone blocked waiting for the lock by checking for a non-zero waiter count. If there's a waiter, it traverses the per-cpu "lock_spinners" variable, which contains which lock each CPU is waiting on. It picks one CPU waiting on the lock and sends it an event to wake it up. This allows efficient fast-path spinlock operation, while allowing spinning vcpus to give up their processor time while waiting for a contended lock. [*] 2^16 iterations is threshold at which 98% locks have been taken according to Thomas Friebel's Xen Summit talk "Preventing Guests from Spinning Around". Therefore, we'd expect the lock and unlock slow paths will only be entered 2% of the time. Signed-off-by: Jeremy Fitzhardinge Cc: Jens Axboe Cc: Peter Zijlstra Cc: Christoph Lameter Cc: Petr Tesarik Cc: Virtualization Cc: Xen devel Cc: Thomas Friebel Cc: Nick Piggin Signed-off-by: Ingo Molnar --- arch/x86/xen/smp.c | 172 ++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 171 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c index a8ebafc09d4..e693812ac59 100644 --- a/arch/x86/xen/smp.c +++ b/arch/x86/xen/smp.c @@ -15,6 +15,7 @@ * This does not handle HOTPLUG_CPU yet. */ #include +#include #include #include @@ -35,6 +36,8 @@ #include "xen-ops.h" #include "mmu.h" +static void __cpuinit xen_init_lock_cpu(int cpu); + cpumask_t xen_cpu_initialized_map; static DEFINE_PER_CPU(int, resched_irq); @@ -179,6 +182,8 @@ static void __init xen_smp_prepare_cpus(unsigned int max_cpus) { unsigned cpu; + xen_init_lock_cpu(0); + smp_store_cpu_info(0); cpu_data(0).x86_max_cores = 1; set_cpu_sibling_map(0); @@ -301,6 +306,7 @@ static int __cpuinit xen_cpu_up(unsigned int cpu) clear_tsk_thread_flag(idle, TIF_FORK); #endif xen_setup_timer(cpu); + xen_init_lock_cpu(cpu); per_cpu(cpu_state, cpu) = CPU_UP_PREPARE; @@ -413,6 +419,170 @@ static irqreturn_t xen_call_function_single_interrupt(int irq, void *dev_id) return IRQ_HANDLED; } +struct xen_spinlock { + unsigned char lock; /* 0 -> free; 1 -> locked */ + unsigned short spinners; /* count of waiting cpus */ +}; + +static int xen_spin_is_locked(struct raw_spinlock *lock) +{ + struct xen_spinlock *xl = (struct xen_spinlock *)lock; + + return xl->lock != 0; +} + +static int xen_spin_is_contended(struct raw_spinlock *lock) +{ + struct xen_spinlock *xl = (struct xen_spinlock *)lock; + + /* Not strictly true; this is only the count of contended + lock-takers entering the slow path. */ + return xl->spinners != 0; +} + +static int xen_spin_trylock(struct raw_spinlock *lock) +{ + struct xen_spinlock *xl = (struct xen_spinlock *)lock; + u8 old = 1; + + asm("xchgb %b0,%1" + : "+q" (old), "+m" (xl->lock) : : "memory"); + + return old == 0; +} + +static DEFINE_PER_CPU(int, lock_kicker_irq) = -1; +static DEFINE_PER_CPU(struct xen_spinlock *, lock_spinners); + +static inline void spinning_lock(struct xen_spinlock *xl) +{ + __get_cpu_var(lock_spinners) = xl; + wmb(); /* set lock of interest before count */ + asm(LOCK_PREFIX " incw %0" + : "+m" (xl->spinners) : : "memory"); +} + +static inline void unspinning_lock(struct xen_spinlock *xl) +{ + asm(LOCK_PREFIX " decw %0" + : "+m" (xl->spinners) : : "memory"); + wmb(); /* decrement count before clearing lock */ + __get_cpu_var(lock_spinners) = NULL; +} + +static noinline int xen_spin_lock_slow(struct raw_spinlock *lock) +{ + struct xen_spinlock *xl = (struct xen_spinlock *)lock; + int irq = __get_cpu_var(lock_kicker_irq); + int ret; + + /* If kicker interrupts not initialized yet, just spin */ + if (irq == -1) + return 0; + + /* announce we're spinning */ + spinning_lock(xl); + + /* clear pending */ + xen_clear_irq_pending(irq); + + /* check again make sure it didn't become free while + we weren't looking */ + ret = xen_spin_trylock(lock); + if (ret) + goto out; + + /* block until irq becomes pending */ + xen_poll_irq(irq); + kstat_this_cpu.irqs[irq]++; + +out: + unspinning_lock(xl); + return ret; +} + +static void xen_spin_lock(struct raw_spinlock *lock) +{ + struct xen_spinlock *xl = (struct xen_spinlock *)lock; + int timeout; + u8 oldval; + + do { + timeout = 1 << 10; + + asm("1: xchgb %1,%0\n" + " testb %1,%1\n" + " jz 3f\n" + "2: rep;nop\n" + " cmpb $0,%0\n" + " je 1b\n" + " dec %2\n" + " jnz 2b\n" + "3:\n" + : "+m" (xl->lock), "=q" (oldval), "+r" (timeout) + : "1" (1) + : "memory"); + + } while (unlikely(oldval != 0 && !xen_spin_lock_slow(lock))); +} + +static noinline void xen_spin_unlock_slow(struct xen_spinlock *xl) +{ + int cpu; + + for_each_online_cpu(cpu) { + /* XXX should mix up next cpu selection */ + if (per_cpu(lock_spinners, cpu) == xl) { + xen_send_IPI_one(cpu, XEN_SPIN_UNLOCK_VECTOR); + break; + } + } +} + +static void xen_spin_unlock(struct raw_spinlock *lock) +{ + struct xen_spinlock *xl = (struct xen_spinlock *)lock; + + smp_wmb(); /* make sure no writes get moved after unlock */ + xl->lock = 0; /* release lock */ + + /* make sure unlock happens before kick */ + barrier(); + + if (unlikely(xl->spinners)) + xen_spin_unlock_slow(xl); +} + +static __cpuinit void xen_init_lock_cpu(int cpu) +{ + int irq; + const char *name; + + name = kasprintf(GFP_KERNEL, "spinlock%d", cpu); + irq = bind_ipi_to_irqhandler(XEN_SPIN_UNLOCK_VECTOR, + cpu, + xen_reschedule_interrupt, + IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING, + name, + NULL); + + if (irq >= 0) { + disable_irq(irq); /* make sure it's never delivered */ + per_cpu(lock_kicker_irq, cpu) = irq; + } + + printk("cpu %d spinlock event irq %d\n", cpu, irq); +} + +static void __init xen_init_spinlocks(void) +{ + pv_lock_ops.spin_is_locked = xen_spin_is_locked; + pv_lock_ops.spin_is_contended = xen_spin_is_contended; + pv_lock_ops.spin_lock = xen_spin_lock; + pv_lock_ops.spin_trylock = xen_spin_trylock; + pv_lock_ops.spin_unlock = xen_spin_unlock; +} + static const struct smp_ops xen_smp_ops __initdata = { .smp_prepare_boot_cpu = xen_smp_prepare_boot_cpu, .smp_prepare_cpus = xen_smp_prepare_cpus, @@ -430,5 +600,5 @@ void __init xen_smp_init(void) { smp_ops = xen_smp_ops; xen_fill_possible_map(); - paravirt_use_bytelocks(); + xen_init_spinlocks(); } -- cgit v1.2.3 From 4bb689eee12ceb6d669a0c9a519037c049a8af38 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 9 Jul 2008 14:33:33 +0200 Subject: x86: paravirt spinlocks, !CONFIG_SMP build fixes Signed-off-by: Ingo Molnar --- arch/x86/kernel/paravirt.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index bba4041bb7f..6aa8aed06d5 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -270,11 +270,13 @@ enum paravirt_lazy_mode paravirt_get_lazy_mode(void) void __init paravirt_use_bytelocks(void) { +#ifdef CONFIG_SMP pv_lock_ops.spin_is_locked = __byte_spin_is_locked; pv_lock_ops.spin_is_contended = __byte_spin_is_contended; pv_lock_ops.spin_lock = __byte_spin_lock; pv_lock_ops.spin_trylock = __byte_spin_trylock; pv_lock_ops.spin_unlock = __byte_spin_unlock; +#endif } struct pv_info pv_info = { @@ -461,12 +463,14 @@ struct pv_mmu_ops pv_mmu_ops = { }; struct pv_lock_ops pv_lock_ops = { +#ifdef CONFIG_SMP .spin_is_locked = __ticket_spin_is_locked, .spin_is_contended = __ticket_spin_is_contended, .spin_lock = __ticket_spin_lock, .spin_trylock = __ticket_spin_trylock, .spin_unlock = __ticket_spin_unlock, +#endif }; EXPORT_SYMBOL_GPL(pv_time_ops); -- cgit v1.2.3 From 9af98578d6af588f52d0dacd64fe42caa405a327 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 9 Jul 2008 14:39:15 +0200 Subject: x86: paravirt spinlocks, modular build fix fix: MODPOST 408 modules ERROR: "pv_lock_ops" [net/dccp/dccp.ko] undefined! ERROR: "pv_lock_ops" [fs/jbd2/jbd2.ko] undefined! ERROR: "pv_lock_ops" [drivers/media/common/saa7146_vv.ko] undefined! Signed-off-by: Ingo Molnar --- arch/x86/kernel/paravirt.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index 6aa8aed06d5..3edfd7af22a 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -472,6 +472,7 @@ struct pv_lock_ops pv_lock_ops = { .spin_unlock = __ticket_spin_unlock, #endif }; +EXPORT_SYMBOL_GPL(pv_lock_ops); EXPORT_SYMBOL_GPL(pv_time_ops); EXPORT_SYMBOL (pv_cpu_ops); -- cgit v1.2.3 From 34646bca474142e1424e5f6c4a33cb2ba0930ea1 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 9 Jul 2008 15:42:09 +0200 Subject: x86, paravirt-spinlocks: fix boot hang the paravirt-spinlock patches caused a boot hang with this config: http://redhat.com/~mingo/misc/config-Wed_Jul__9_14_47_04_CEST_2008.bad i have bisected it down to: | commit e17b58c2e85bc2ad2afc07fb8d898017c2b75ed1 | Author: Jeremy Fitzhardinge | Date: Mon Jul 7 12:07:53 2008 -0700 | | xen: implement Xen-specific spinlocks i.e. applying that patch alone causes the hang. The hang happens in the ftrace self-test: initcall utsname_sysctl_init+0x0/0x19 returned 0 after 0 msecs calling init_sched_switch_trace+0x0/0x4c Testing tracer sched_switch: PASSED initcall init_sched_switch_trace+0x0/0x4c returned 0 after 167 msecs calling init_function_trace+0x0/0x12 Testing tracer ftrace: [hard hang] it should have continued like this: Testing tracer ftrace: PASSED initcall init_function_trace+0x0/0x12 returned 0 after 198 msecs calling init_irqsoff_tracer+0x0/0x14 Testing tracer irqsoff: PASSED initcall init_irqsoff_tracer+0x0/0x14 returned 0 after 3 msecs calling init_mmio_trace+0x0/0x12 initcall init_mmio_trace+0x0/0x12 returned 0 after 0 msecs the problem is that such lowlevel primitives as spinlocks should never be built with -pg (which ftrace does). Marking paravirt.o as non-pg and marking all spinlock ops as always-inline solve the hang. Signed-off-by: Ingo Molnar --- arch/x86/kernel/Makefile | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index 5112c84f542..78d52171400 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -7,10 +7,11 @@ extra-y := head_$(BITS).o head$(BITS).o head.o init_task.o vmlinu CPPFLAGS_vmlinux.lds += -U$(UTS_MACHINE) ifdef CONFIG_FTRACE -# Do not profile debug utilities +# Do not profile debug and lowlevel utilities CFLAGS_REMOVE_tsc_64.o = -pg CFLAGS_REMOVE_tsc_32.o = -pg CFLAGS_REMOVE_rtc.o = -pg +CFLAGS_REMOVE_paravirt.o = -pg endif # -- cgit v1.2.3 From 6718d0d6da2749d3bff522e6057e97e6aa85e4d1 Mon Sep 17 00:00:00 2001 From: Roland McGrath Date: Wed, 9 Jul 2008 01:07:02 -0700 Subject: x86 ptrace: block-step fix The enable_single_step() logic bails out early if TF is already set. That skips some of the bookkeeping that keeps things straight. This makes PTRACE_SINGLEBLOCK break the behavior of a user task that was already setting TF itself in user mode. Fix the bookkeeping to notice the old TF setting as it should. Test case at: http://sources.redhat.com/cgi-bin/cvsweb.cgi/~checkout~/tests/ptrace-tests/tests/step-jump-cont-strict.c?cvsroot=systemtap Signed-off-by: Roland McGrath --- arch/x86/kernel/step.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c index 92c20fee678..0d2cb363ea7 100644 --- a/arch/x86/kernel/step.c +++ b/arch/x86/kernel/step.c @@ -105,6 +105,7 @@ static int is_setting_trap_flag(struct task_struct *child, struct pt_regs *regs) static int enable_single_step(struct task_struct *child) { struct pt_regs *regs = task_pt_regs(child); + unsigned long oflags; /* * Always set TIF_SINGLESTEP - this guarantees that @@ -113,11 +114,7 @@ static int enable_single_step(struct task_struct *child) */ set_tsk_thread_flag(child, TIF_SINGLESTEP); - /* - * If TF was already set, don't do anything else - */ - if (regs->flags & X86_EFLAGS_TF) - return 0; + oflags = regs->flags; /* Set TF on the kernel stack.. */ regs->flags |= X86_EFLAGS_TF; @@ -126,9 +123,22 @@ static int enable_single_step(struct task_struct *child) * ..but if TF is changed by the instruction we will trace, * don't mark it as being "us" that set it, so that we * won't clear it by hand later. + * + * Note that if we don't actually execute the popf because + * of a signal arriving right now or suchlike, we will lose + * track of the fact that it really was "us" that set it. */ - if (is_setting_trap_flag(child, regs)) + if (is_setting_trap_flag(child, regs)) { + clear_tsk_thread_flag(child, TIF_FORCED_TF); return 0; + } + + /* + * If TF was already set, check whether it was us who set it. + * If not, we should never attempt a block step. + */ + if (oflags & X86_EFLAGS_TF) + return test_tsk_thread_flag(child, TIF_FORCED_TF); set_tsk_thread_flag(child, TIF_FORCED_TF); -- cgit v1.2.3 From 64f097331928b01d704047c1dbc738bb6d2a9bf9 Mon Sep 17 00:00:00 2001 From: Roland McGrath Date: Wed, 9 Jul 2008 01:33:14 -0700 Subject: x86 ptrace: unify TIF_SINGLESTEP This unifies the treatment of TIF_SINGLESTEP on i386 and x86_64. The bit is now excluded from _TIF_WORK_MASK on i386 as it has been on x86_64. This means the do_notify_resume() path using it is never used, so TIF_SINGLESTEP is not cleared on returning to user mode. Both now leave TIF_SINGLESTEP set when returning to user, so that it's already set on an int $0x80 system call entry. This removes the need for testing TF on the system_call path. Doing it this way fixes the regression for PTRACE_SINGLESTEP into a sigreturn syscall, introduced by commit 1e2e99f0e4aa6363e8515ed17011c210c8f1b52a. The clear_TF_reenable case that sets TIF_SINGLESTEP can only happen on a non-exception kernel entry, i.e. sysenter/syscall instruction. That will always get to the syscall exit tracing path. Signed-off-by: Roland McGrath --- arch/x86/kernel/entry_32.S | 4 ---- arch/x86/kernel/signal_32.c | 6 ------ arch/x86/kernel/signal_64.c | 6 ------ 3 files changed, 16 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S index 6bc07f0f120..0ad987d02b7 100644 --- a/arch/x86/kernel/entry_32.S +++ b/arch/x86/kernel/entry_32.S @@ -383,10 +383,6 @@ syscall_exit: # setting need_resched or sigpending # between sampling and the iret TRACE_IRQS_OFF - testl $X86_EFLAGS_TF,PT_EFLAGS(%esp) # If tracing set singlestep flag on exit - jz no_singlestep - orl $_TIF_SINGLESTEP,TI_flags(%ebp) -no_singlestep: movl TI_flags(%ebp), %ecx testw $_TIF_ALLWORK_MASK, %cx # current->work jne syscall_exit_work diff --git a/arch/x86/kernel/signal_32.c b/arch/x86/kernel/signal_32.c index d9237363096..295b5f5c938 100644 --- a/arch/x86/kernel/signal_32.c +++ b/arch/x86/kernel/signal_32.c @@ -657,12 +657,6 @@ static void do_signal(struct pt_regs *regs) void do_notify_resume(struct pt_regs *regs, void *unused, __u32 thread_info_flags) { - /* Pending single-step? */ - if (thread_info_flags & _TIF_SINGLESTEP) { - regs->flags |= X86_EFLAGS_TF; - clear_thread_flag(TIF_SINGLESTEP); - } - /* deal with pending signal delivery */ if (thread_info_flags & _TIF_SIGPENDING) do_signal(regs); diff --git a/arch/x86/kernel/signal_64.c b/arch/x86/kernel/signal_64.c index e53b267662e..bf87684474f 100644 --- a/arch/x86/kernel/signal_64.c +++ b/arch/x86/kernel/signal_64.c @@ -487,12 +487,6 @@ static void do_signal(struct pt_regs *regs) void do_notify_resume(struct pt_regs *regs, void *unused, __u32 thread_info_flags) { - /* Pending single-step? */ - if (thread_info_flags & _TIF_SINGLESTEP) { - regs->flags |= X86_EFLAGS_TF; - clear_thread_flag(TIF_SINGLESTEP); - } - #ifdef CONFIG_X86_MCE /* notify userspace of pending MCEs */ if (thread_info_flags & _TIF_MCE_NOTIFY) -- cgit v1.2.3 From d4d67150165df8bf1cc05e532f6efca96f907cab Mon Sep 17 00:00:00 2001 From: Roland McGrath Date: Wed, 9 Jul 2008 02:38:07 -0700 Subject: x86 ptrace: unify syscall tracing This unifies and cleans up the syscall tracing code on i386 and x86_64. Using a single function for entry and exit tracing on 32-bit made the do_syscall_trace() into some terrible spaghetti. The logic is clear and simple using separate syscall_trace_enter() and syscall_trace_leave() functions as on 64-bit. The unification adds PTRACE_SYSEMU and PTRACE_SYSEMU_SINGLESTEP support on x86_64, for 32-bit ptrace() callers and for 64-bit ptrace() callers tracing either 32-bit or 64-bit tasks. It behaves just like 32-bit. Changing syscall_trace_enter() to return the syscall number shortens all the assembly paths, while adding the SYSEMU feature in a simple way. Signed-off-by: Roland McGrath --- arch/x86/ia32/ia32entry.S | 17 +++--- arch/x86/kernel/entry_32.S | 19 +++--- arch/x86/kernel/entry_64.S | 14 +++-- arch/x86/kernel/ptrace.c | 141 +++++++++++++++------------------------------ 4 files changed, 71 insertions(+), 120 deletions(-) (limited to 'arch') diff --git a/arch/x86/ia32/ia32entry.S b/arch/x86/ia32/ia32entry.S index 20371d0635e..8796d190525 100644 --- a/arch/x86/ia32/ia32entry.S +++ b/arch/x86/ia32/ia32entry.S @@ -37,6 +37,11 @@ movq %rax,R8(%rsp) .endm + /* + * Reload arg registers from stack in case ptrace changed them. + * We don't reload %eax because syscall_trace_enter() returned + * the value it wants us to use in the table lookup. + */ .macro LOAD_ARGS32 offset movl \offset(%rsp),%r11d movl \offset+8(%rsp),%r10d @@ -46,7 +51,6 @@ movl \offset+48(%rsp),%edx movl \offset+56(%rsp),%esi movl \offset+64(%rsp),%edi - movl \offset+72(%rsp),%eax .endm .macro CFI_STARTPROC32 simple @@ -137,13 +141,12 @@ ENTRY(ia32_sysenter_target) .previous GET_THREAD_INFO(%r10) orl $TS_COMPAT,TI_status(%r10) - testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP), \ - TI_flags(%r10) + testl $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%r10) CFI_REMEMBER_STATE jnz sysenter_tracesys -sysenter_do_call: cmpl $(IA32_NR_syscalls-1),%eax ja ia32_badsys +sysenter_do_call: IA32_ARG_FIXUP 1 call *ia32_sys_call_table(,%rax,8) movq %rax,RAX-ARGOFFSET(%rsp) @@ -242,8 +245,7 @@ ENTRY(ia32_cstar_target) .previous GET_THREAD_INFO(%r10) orl $TS_COMPAT,TI_status(%r10) - testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP), \ - TI_flags(%r10) + testl $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%r10) CFI_REMEMBER_STATE jnz cstar_tracesys cstar_do_call: @@ -336,8 +338,7 @@ ENTRY(ia32_syscall) SAVE_ARGS 0,0,1 GET_THREAD_INFO(%r10) orl $TS_COMPAT,TI_status(%r10) - testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP), \ - TI_flags(%r10) + testl $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%r10) jnz ia32_tracesys ia32_do_syscall: cmpl $(IA32_NR_syscalls-1),%eax diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S index 0ad987d02b7..cadf73f70d3 100644 --- a/arch/x86/kernel/entry_32.S +++ b/arch/x86/kernel/entry_32.S @@ -332,7 +332,7 @@ sysenter_past_esp: GET_THREAD_INFO(%ebp) /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */ - testw $(_TIF_SYSCALL_EMU|_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT),TI_flags(%ebp) + testw $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%ebp) jnz syscall_trace_entry cmpl $(nr_syscalls), %eax jae syscall_badsys @@ -370,7 +370,7 @@ ENTRY(system_call) GET_THREAD_INFO(%ebp) # system call tracing in operation / emulation /* Note, _TIF_SECCOMP is bit number 8, and so it needs testw and not testb */ - testw $(_TIF_SYSCALL_EMU|_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT),TI_flags(%ebp) + testw $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%ebp) jnz syscall_trace_entry cmpl $(nr_syscalls), %eax jae syscall_badsys @@ -510,12 +510,8 @@ END(work_pending) syscall_trace_entry: movl $-ENOSYS,PT_EAX(%esp) movl %esp, %eax - xorl %edx,%edx - call do_syscall_trace - cmpl $0, %eax - jne resume_userspace # ret != 0 -> running under PTRACE_SYSEMU, - # so must skip actual syscall - movl PT_ORIG_EAX(%esp), %eax + call syscall_trace_enter + /* What it returned is what we'll actually use. */ cmpl $(nr_syscalls), %eax jnae syscall_call jmp syscall_exit @@ -524,14 +520,13 @@ END(syscall_trace_entry) # perform syscall exit tracing ALIGN syscall_exit_work: - testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl + testb $_TIF_WORK_SYSCALL_EXIT, %cl jz work_pending TRACE_IRQS_ON - ENABLE_INTERRUPTS(CLBR_ANY) # could let do_syscall_trace() call + ENABLE_INTERRUPTS(CLBR_ANY) # could let syscall_trace_leave() call # schedule() instead movl %esp, %eax - movl $1, %edx - call do_syscall_trace + call syscall_trace_leave jmp resume_userspace END(syscall_exit_work) CFI_ENDPROC diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S index ae63e584c34..63001c6ecf6 100644 --- a/arch/x86/kernel/entry_64.S +++ b/arch/x86/kernel/entry_64.S @@ -349,8 +349,7 @@ ENTRY(system_call_after_swapgs) movq %rcx,RIP-ARGOFFSET(%rsp) CFI_REL_OFFSET rip,RIP-ARGOFFSET GET_THREAD_INFO(%rcx) - testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SECCOMP), \ - TI_flags(%rcx) + testl $_TIF_WORK_SYSCALL_ENTRY,TI_flags(%rcx) jnz tracesys cmpq $__NR_syscall_max,%rax ja badsys @@ -430,7 +429,12 @@ tracesys: FIXUP_TOP_OF_STACK %rdi movq %rsp,%rdi call syscall_trace_enter - LOAD_ARGS ARGOFFSET /* reload args from stack in case ptrace changed it */ + /* + * Reload arg registers from stack in case ptrace changed them. + * We don't reload %rax because syscall_trace_enter() returned + * the value it wants us to use in the table lookup. + */ + LOAD_ARGS ARGOFFSET, 1 RESTORE_REST cmpq $__NR_syscall_max,%rax ja int_ret_from_sys_call /* RAX(%rsp) set to -ENOSYS above */ @@ -483,7 +487,7 @@ int_very_careful: ENABLE_INTERRUPTS(CLBR_NONE) SAVE_REST /* Check for syscall exit trace */ - testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP),%edx + testl $_TIF_WORK_SYSCALL_EXIT,%edx jz int_signal pushq %rdi CFI_ADJUST_CFA_OFFSET 8 @@ -491,7 +495,7 @@ int_very_careful: call syscall_trace_leave popq %rdi CFI_ADJUST_CFA_OFFSET -8 - andl $~(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP),%edi + andl $~(_TIF_WORK_SYSCALL_EXIT|_TIF_SYSCALL_EMU),%edi jmp int_restore_rest int_signal: diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index 77040b6070e..34e77b16a42 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -1357,8 +1357,6 @@ const struct user_regset_view *task_user_regset_view(struct task_struct *task) #endif } -#ifdef CONFIG_X86_32 - void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, int error_code) { struct siginfo info; @@ -1377,89 +1375,10 @@ void send_sigtrap(struct task_struct *tsk, struct pt_regs *regs, int error_code) force_sig_info(SIGTRAP, &info, tsk); } -/* notification of system call entry/exit - * - triggered by current->work.syscall_trace - */ -int do_syscall_trace(struct pt_regs *regs, int entryexit) -{ - int is_sysemu = test_thread_flag(TIF_SYSCALL_EMU); - /* - * With TIF_SYSCALL_EMU set we want to ignore TIF_SINGLESTEP for syscall - * interception - */ - int is_singlestep = !is_sysemu && test_thread_flag(TIF_SINGLESTEP); - int ret = 0; - - /* do the secure computing check first */ - if (!entryexit) - secure_computing(regs->orig_ax); - - if (unlikely(current->audit_context)) { - if (entryexit) - audit_syscall_exit(AUDITSC_RESULT(regs->ax), - regs->ax); - /* Debug traps, when using PTRACE_SINGLESTEP, must be sent only - * on the syscall exit path. Normally, when TIF_SYSCALL_AUDIT is - * not used, entry.S will call us only on syscall exit, not - * entry; so when TIF_SYSCALL_AUDIT is used we must avoid - * calling send_sigtrap() on syscall entry. - * - * Note that when PTRACE_SYSEMU_SINGLESTEP is used, - * is_singlestep is false, despite his name, so we will still do - * the correct thing. - */ - else if (is_singlestep) - goto out; - } - - if (!(current->ptrace & PT_PTRACED)) - goto out; - - /* If a process stops on the 1st tracepoint with SYSCALL_TRACE - * and then is resumed with SYSEMU_SINGLESTEP, it will come in - * here. We have to check this and return */ - if (is_sysemu && entryexit) - return 0; - - /* Fake a debug trap */ - if (is_singlestep) - send_sigtrap(current, regs, 0); - - if (!test_thread_flag(TIF_SYSCALL_TRACE) && !is_sysemu) - goto out; - - /* the 0x80 provides a way for the tracing parent to distinguish - between a syscall stop and SIGTRAP delivery */ - /* Note that the debugger could change the result of test_thread_flag!*/ - ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD) ? 0x80:0)); - - /* - * this isn't the same as continuing with a signal, but it will do - * for normal use. strace only continues with a signal if the - * stopping signal is not SIGTRAP. -brl - */ - if (current->exit_code) { - send_sig(current->exit_code, current, 1); - current->exit_code = 0; - } - ret = is_sysemu; -out: - if (unlikely(current->audit_context) && !entryexit) - audit_syscall_entry(AUDIT_ARCH_I386, regs->orig_ax, - regs->bx, regs->cx, regs->dx, regs->si); - if (ret == 0) - return 0; - - regs->orig_ax = -1; /* force skip of syscall restarting */ - if (unlikely(current->audit_context)) - audit_syscall_exit(AUDITSC_RESULT(regs->ax), regs->ax); - return 1; -} - -#else /* CONFIG_X86_64 */ - static void syscall_trace(struct pt_regs *regs) { + if (!(current->ptrace & PT_PTRACED)) + return; #if 0 printk("trace %s ip %lx sp %lx ax %d origrax %d caller %lx tiflags %x ptrace %x\n", @@ -1481,39 +1400,71 @@ static void syscall_trace(struct pt_regs *regs) } } -asmlinkage void syscall_trace_enter(struct pt_regs *regs) +#ifdef CONFIG_X86_32 +# define IS_IA32 1 +#elif defined CONFIG_IA32_EMULATION +# define IS_IA32 test_thread_flag(TIF_IA32) +#else +# define IS_IA32 0 +#endif + +/* + * We must return the syscall number to actually look up in the table. + * This can be -1L to skip running any syscall at all. + */ +asmregparm long syscall_trace_enter(struct pt_regs *regs) { + long ret = 0; + /* do the secure computing check first */ secure_computing(regs->orig_ax); - if (test_thread_flag(TIF_SYSCALL_TRACE) - && (current->ptrace & PT_PTRACED)) + if (unlikely(test_thread_flag(TIF_SYSCALL_EMU))) + ret = -1L; + + if (ret || test_thread_flag(TIF_SYSCALL_TRACE)) syscall_trace(regs); if (unlikely(current->audit_context)) { - if (test_thread_flag(TIF_IA32)) { + if (IS_IA32) audit_syscall_entry(AUDIT_ARCH_I386, regs->orig_ax, regs->bx, regs->cx, regs->dx, regs->si); - } else { +#ifdef CONFIG_X86_64 + else audit_syscall_entry(AUDIT_ARCH_X86_64, regs->orig_ax, regs->di, regs->si, regs->dx, regs->r10); - } +#endif } + + return ret ?: regs->orig_ax; } -asmlinkage void syscall_trace_leave(struct pt_regs *regs) +asmregparm void syscall_trace_leave(struct pt_regs *regs) { if (unlikely(current->audit_context)) audit_syscall_exit(AUDITSC_RESULT(regs->ax), regs->ax); - if ((test_thread_flag(TIF_SYSCALL_TRACE) - || test_thread_flag(TIF_SINGLESTEP)) - && (current->ptrace & PT_PTRACED)) + if (test_thread_flag(TIF_SYSCALL_TRACE)) syscall_trace(regs); -} -#endif /* CONFIG_X86_32 */ + /* + * If TIF_SYSCALL_EMU is set, we only get here because of + * TIF_SINGLESTEP (i.e. this is PTRACE_SYSEMU_SINGLESTEP). + * We already reported this syscall instruction in + * syscall_trace_enter(), so don't do any more now. + */ + if (unlikely(test_thread_flag(TIF_SYSCALL_EMU))) + return; + + /* + * If we are single-stepping, synthesize a trap to follow the + * system call instruction. + */ + if (test_thread_flag(TIF_SINGLESTEP) && + (current->ptrace & PT_PTRACED)) + send_sigtrap(current, regs, 0); +} -- cgit v1.2.3 From 380fdd7585a4c2f41b48925eba85c0654b7b858b Mon Sep 17 00:00:00 2001 From: Roland McGrath Date: Wed, 9 Jul 2008 02:39:29 -0700 Subject: x86 ptrace: user-sets-TF nits This closes some arcane holes in single-step handling that can arise only when user programs set TF directly (via popf or sigreturn) and then use vDSO (syscall/sysenter) system call entry. In those entry paths, the clear_TF_reenable case hits and we must check TIF_SINGLESTEP to be sure our bookkeeping stays correct wrt the user's view of TF. Signed-off-by: Roland McGrath --- arch/x86/kernel/ptrace.c | 10 ++++++++++ arch/x86/kernel/step.c | 13 +++++++++++++ 2 files changed, 23 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c index 34e77b16a42..e37dccce85d 100644 --- a/arch/x86/kernel/ptrace.c +++ b/arch/x86/kernel/ptrace.c @@ -1416,6 +1416,16 @@ asmregparm long syscall_trace_enter(struct pt_regs *regs) { long ret = 0; + /* + * If we stepped into a sysenter/syscall insn, it trapped in + * kernel mode; do_debug() cleared TF and set TIF_SINGLESTEP. + * If user-mode had set TF itself, then it's still clear from + * do_debug() and we need to set it again to restore the user + * state. If we entered on the slow path, TF was already set. + */ + if (test_thread_flag(TIF_SINGLESTEP)) + regs->flags |= X86_EFLAGS_TF; + /* do the secure computing check first */ secure_computing(regs->orig_ax); diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c index 0d2cb363ea7..e8b9863ef8c 100644 --- a/arch/x86/kernel/step.c +++ b/arch/x86/kernel/step.c @@ -107,6 +107,19 @@ static int enable_single_step(struct task_struct *child) struct pt_regs *regs = task_pt_regs(child); unsigned long oflags; + /* + * If we stepped into a sysenter/syscall insn, it trapped in + * kernel mode; do_debug() cleared TF and set TIF_SINGLESTEP. + * If user-mode had set TF itself, then it's still clear from + * do_debug() and we need to set it again to restore the user + * state so we don't wrongly set TIF_FORCED_TF below. + * If enable_single_step() was used last and that is what + * set TIF_SINGLESTEP, then both TF and TIF_FORCED_TF are + * already set and our bookkeeping is fine. + */ + if (unlikely(test_tsk_thread_flag(child, TIF_SINGLESTEP))) + regs->flags |= X86_EFLAGS_TF; + /* * Always set TIF_SINGLESTEP - this guarantees that * we single-step system calls etc.. This will also -- cgit v1.2.3 From fab3b58d3b242b5903f78d60d86803a8aecdf6de Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 17 Jul 2008 13:50:15 +0200 Subject: x86 reboot quirks: add Dell Precision WorkStation T5400 as reported in: "reboot=bios is mandatory on Dell T5400 server." http://bugzilla.kernel.org/show_bug.cgi?id=11108 add a DMI reboot quirk. Signed-off-by: Ingo Molnar Cc: --- arch/x86/kernel/reboot.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c index f8a62160e15..9dcf39c0297 100644 --- a/arch/x86/kernel/reboot.c +++ b/arch/x86/kernel/reboot.c @@ -177,6 +177,14 @@ static struct dmi_system_id __initdata reboot_dmi_table[] = { DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"), }, }, + { /* Handle problems with rebooting on Dell T5400's */ + .callback = set_bios_reboot, + .ident = "Dell Precision T5400", + .matches = { + DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."), + DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"), + }, + }, { /* Handle problems with rebooting on HP laptops */ .callback = set_bios_reboot, .ident = "HP Compaq Laptop", -- cgit v1.2.3 From 93a0886e2368eafb9df5e2021fb185195cee88b2 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 15 Jul 2008 13:43:42 -0700 Subject: x86, xen, power: fix up config dependencies on PM Xen save/restore needs bits of code enabled by PM_SLEEP, and PM_SLEEP depends on PM. So make XEN_SAVE_RESTORE depend on PM and PM_SLEEP depend on XEN_SAVE_RESTORE. Signed-off-by: Jeremy Fitzhardinge Acked-by: Rafael J. Wysocki Signed-off-by: Ingo Molnar --- arch/x86/xen/Kconfig | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'arch') diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig index 20b49729bed..3815e425f47 100644 --- a/arch/x86/xen/Kconfig +++ b/arch/x86/xen/Kconfig @@ -23,3 +23,8 @@ config XEN_MAX_DOMAIN_MEMORY according to the maximum possible memory size of a Xen domain. This array uses 1 page per gigabyte, so there's no need to be too stingy here. + +config XEN_SAVE_RESTORE + bool + depends on PM + default y \ No newline at end of file -- cgit v1.2.3 From 4fdf08b5bf8d449cc9897395895157c6ff8ddc41 Mon Sep 17 00:00:00 2001 From: "H. Peter Anvin" Date: Thu, 17 Jul 2008 11:29:24 -0700 Subject: x86: unify and correct the GDT_ENTRY() macro Merge the GDT_ENTRY() macro between arch/x86/boot/pm.c and arch/x86/kernel/acpi/sleep.c and put the new one in . While we're at it, correct the bitmasks for the limit and flags. The new version relies on using ULL constants in order to cause type promotion rather than explicit casts; this avoids having to include in . Signed-off-by: H. Peter Anvin --- arch/x86/boot/pm.c | 6 ------ arch/x86/kernel/acpi/sleep.c | 10 +--------- 2 files changed, 1 insertion(+), 15 deletions(-) (limited to 'arch') diff --git a/arch/x86/boot/pm.c b/arch/x86/boot/pm.c index 328956fdb59..85a1cd8a8ff 100644 --- a/arch/x86/boot/pm.c +++ b/arch/x86/boot/pm.c @@ -98,12 +98,6 @@ static void reset_coprocessor(void) /* * Set up the GDT */ -#define GDT_ENTRY(flags, base, limit) \ - (((u64)(base & 0xff000000) << 32) | \ - ((u64)flags << 40) | \ - ((u64)(limit & 0x00ff0000) << 32) | \ - ((u64)(base & 0x00ffffff) << 16) | \ - ((u64)(limit & 0x0000ffff))) struct gdt_ptr { u16 len; diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c index 868de3d5c39..a3ddad18aaa 100644 --- a/arch/x86/kernel/acpi/sleep.c +++ b/arch/x86/kernel/acpi/sleep.c @@ -9,6 +9,7 @@ #include #include #include +#include #include "realmode/wakeup.h" #include "sleep.h" @@ -23,15 +24,6 @@ static unsigned long acpi_realmode; static char temp_stack[10240]; #endif -/* XXX: this macro should move to asm-x86/segment.h and be shared with the - boot code... */ -#define GDT_ENTRY(flags, base, limit) \ - (((u64)(base & 0xff000000) << 32) | \ - ((u64)flags << 40) | \ - ((u64)(limit & 0x00ff0000) << 32) | \ - ((u64)(base & 0x00ffffff) << 16) | \ - ((u64)(limit & 0x0000ffff))) - /** * acpi_save_state_mem - save kernel state * -- cgit v1.2.3 From 64d206d896ff70b828138577d5ff39deda5f1c4d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 18 Jul 2008 00:26:59 +0200 Subject: x86: rename CONFIG_NONPROMISC_DEVMEM to CONFIG_PROMISC_DEVMEM Linus observed: > The real bug is that we shouldn't have "double negatives", and > certainly not negative config options. Making that "promiscuous > /dev/mem" option a negated thing as a config option was bad. right ... lets rename this option. There should never be a negation in config options. [ that reminds me of CONFIG_SCHED_NO_NO_OMIT_FRAME_POINTER, but that is for another commit ;-) ] Signed-off-by: Ingo Molnar --- arch/x86/Kconfig.debug | 7 ++++--- arch/x86/mm/pat.c | 6 +++--- 2 files changed, 7 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug index ae36bfa814e..f0cf5d99079 100644 --- a/arch/x86/Kconfig.debug +++ b/arch/x86/Kconfig.debug @@ -5,10 +5,11 @@ config TRACE_IRQFLAGS_SUPPORT source "lib/Kconfig.debug" -config NONPROMISC_DEVMEM - bool "Filter access to /dev/mem" +config PROMISC_DEVMEM + bool "Allow unlimited access to /dev/mem" + default y help - If this option is left off, you allow userspace access to all + If this option is left on, you allow userspace (root) access to all of memory, including kernel and userspace memory. Accidental access to this is obviously disastrous, but specific access can be used by people debugging the kernel. diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index d4585077977..c34dc483839 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -373,8 +373,8 @@ pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, return vma_prot; } -#ifdef CONFIG_NONPROMISC_DEVMEM -/* This check is done in drivers/char/mem.c in case of NONPROMISC_DEVMEM*/ +#ifndef CONFIG_PROMISC_DEVMEM +/* This check is done in drivers/char/mem.c in case of !PROMISC_DEVMEM*/ static inline int range_is_allowed(unsigned long pfn, unsigned long size) { return 1; @@ -398,7 +398,7 @@ static inline int range_is_allowed(unsigned long pfn, unsigned long size) } return 1; } -#endif /* CONFIG_NONPROMISC_DEVMEM */ +#endif /* CONFIG_PROMISC_DEVMEM */ int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn, unsigned long size, pgprot_t *vma_prot) -- cgit v1.2.3 From 6879827f4e08da219c99b91e4e1d793a924103e3 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 18 Jul 2008 01:21:53 +0200 Subject: x86: remove arch/x86/kernel/smpcommon_32.c Yinghai Lu noticed that arch/x86/kernel/smpcommon_32.c got renamed to arch/x86/kernel/smpcommon.c but the old almost-empty file stayed around. Zap it. Reported-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/smpcommon_32.c | 1 - 1 file changed, 1 deletion(-) delete mode 100644 arch/x86/kernel/smpcommon_32.c (limited to 'arch') diff --git a/arch/x86/kernel/smpcommon_32.c b/arch/x86/kernel/smpcommon_32.c deleted file mode 100644 index 8b137891791..00000000000 --- a/arch/x86/kernel/smpcommon_32.c +++ /dev/null @@ -1 +0,0 @@ - -- cgit v1.2.3 From 29cbeb0e17d9d2ca824f62f71cfa7360b3157112 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Thu, 17 Jul 2008 21:50:23 -0700 Subject: x86: use cpu_clear in remove_cpu_from_maps Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/smpboot.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 687376ab07e..27456574f07 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -1311,7 +1311,7 @@ static void __ref remove_cpu_from_maps(int cpu) cpu_clear(cpu, cpu_callout_map); cpu_clear(cpu, cpu_callin_map); /* was set by cpu_init() */ - clear_bit(cpu, (unsigned long *)&cpu_initialized); + cpu_clear(cpu, cpu_initialized); numa_remove_cpu(cpu); } -- cgit v1.2.3 From 593f4a788e5d09e9f00182561437461b0b564de4 Mon Sep 17 00:00:00 2001 From: "Maciej W. Rozycki" Date: Wed, 16 Jul 2008 19:15:30 +0100 Subject: x86: APIC: remove apic_write_around(); use alternatives Use alternatives to select the workaround for the 11AP Pentium erratum for the affected steppings on the fly rather than build time. Remove the X86_GOOD_APIC configuration option and replace all the calls to apic_write_around() with plain apic_write(), protecting accesses to the ESR as appropriate due to the 3AP Pentium erratum. Remove apic_read_around() and all its invocations altogether as not needed. Remove apic_write_atomic() and all its implementing backends. The use of ASM_OUTPUT2() is not strictly needed for input constraints, but I have used it for readability's sake. I had the feeling no one else was brave enough to do it, so I went ahead and here it is. Verified by checking the generated assembly and tested with both a 32-bit and a 64-bit configuration, also with the 11AP "feature" forced on and verified with gdb on /proc/kcore to work as expected (as an 11AP machines are quite hard to get hands on these days). Some script complained about the use of "volatile", but apic_write() needs it for the same reason and is effectively a replacement for writel(), so I have disregarded it. I am not sure what the policy wrt defconfig files is, they are generated and there is risk of a conflict resulting from an unrelated change, so I have left changes to them out. The option will get removed from them at the next run. Some testing with machines other than mine will be needed to avoid some stupid mistake, but despite its volume, the change is not really that intrusive, so I am fairly confident that because it works for me, it will everywhere. Signed-off-by: Maciej W. Rozycki Signed-off-by: Ingo Molnar --- arch/x86/Kconfig.cpu | 4 --- arch/x86/kernel/apic_32.c | 75 ++++++++++++++++++++--------------------- arch/x86/kernel/cpu/bugs.c | 23 +------------ arch/x86/kernel/cpu/intel.c | 10 ++++++ arch/x86/kernel/cpu/mcheck/p4.c | 4 +-- arch/x86/kernel/io_apic_32.c | 14 ++++---- arch/x86/kernel/ipi.c | 6 ++-- arch/x86/kernel/nmi.c | 4 +-- arch/x86/kernel/paravirt.c | 1 - arch/x86/kernel/smpboot.c | 49 +++++++++++---------------- arch/x86/kernel/vmi_32.c | 1 - arch/x86/lguest/boot.c | 1 - arch/x86/xen/enlighten.c | 1 - 13 files changed, 80 insertions(+), 113 deletions(-) (limited to 'arch') diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu index abff1b84ed5..54b8c02c71e 100644 --- a/arch/x86/Kconfig.cpu +++ b/arch/x86/Kconfig.cpu @@ -362,10 +362,6 @@ config X86_ALIGNMENT_16 def_bool y depends on MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCYRIXIII || X86_ELAN || MK6 || M586MMX || M586TSC || M586 || M486 || MVIAC3_2 || MGEODEGX1 -config X86_GOOD_APIC - def_bool y - depends on MK7 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || MK8 || MEFFICEON || MCORE2 || MVIAC7 || X86_64 - config X86_INTEL_USERCOPY def_bool y depends on MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M586MMX || X86_GENERIC || MK8 || MK7 || MEFFICEON || MCORE2 diff --git a/arch/x86/kernel/apic_32.c b/arch/x86/kernel/apic_32.c index a437d027f20..2bc1186cc95 100644 --- a/arch/x86/kernel/apic_32.c +++ b/arch/x86/kernel/apic_32.c @@ -177,7 +177,7 @@ void __cpuinit enable_NMI_through_LVT0(void) /* Level triggered for 82489DX */ if (!lapic_is_integrated()) v |= APIC_LVT_LEVEL_TRIGGER; - apic_write_around(APIC_LVT0, v); + apic_write(APIC_LVT0, v); } /** @@ -212,9 +212,6 @@ int lapic_get_maxlvt(void) * this function twice on the boot CPU, once with a bogus timeout * value, second time for real. The other (noncalibrating) CPUs * call this function only once, with the real, calibrated value. - * - * We do reads before writes even if unnecessary, to get around the - * P5 APIC double write bug. */ static void __setup_APIC_LVTT(unsigned int clocks, int oneshot, int irqen) { @@ -229,18 +226,18 @@ static void __setup_APIC_LVTT(unsigned int clocks, int oneshot, int irqen) if (!irqen) lvtt_value |= APIC_LVT_MASKED; - apic_write_around(APIC_LVTT, lvtt_value); + apic_write(APIC_LVTT, lvtt_value); /* * Divide PICLK by 16 */ tmp_value = apic_read(APIC_TDCR); - apic_write_around(APIC_TDCR, (tmp_value - & ~(APIC_TDR_DIV_1 | APIC_TDR_DIV_TMBASE)) - | APIC_TDR_DIV_16); + apic_write(APIC_TDCR, + (tmp_value & ~(APIC_TDR_DIV_1 | APIC_TDR_DIV_TMBASE)) | + APIC_TDR_DIV_16); if (!oneshot) - apic_write_around(APIC_TMICT, clocks/APIC_DIVISOR); + apic_write(APIC_TMICT, clocks / APIC_DIVISOR); } /* @@ -249,7 +246,7 @@ static void __setup_APIC_LVTT(unsigned int clocks, int oneshot, int irqen) static int lapic_next_event(unsigned long delta, struct clock_event_device *evt) { - apic_write_around(APIC_TMICT, delta); + apic_write(APIC_TMICT, delta); return 0; } @@ -278,7 +275,7 @@ static void lapic_timer_setup(enum clock_event_mode mode, case CLOCK_EVT_MODE_SHUTDOWN: v = apic_read(APIC_LVTT); v |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR); - apic_write_around(APIC_LVTT, v); + apic_write(APIC_LVTT, v); break; case CLOCK_EVT_MODE_RESUME: /* Nothing to do here */ @@ -693,44 +690,44 @@ void clear_local_APIC(void) */ if (maxlvt >= 3) { v = ERROR_APIC_VECTOR; /* any non-zero vector will do */ - apic_write_around(APIC_LVTERR, v | APIC_LVT_MASKED); + apic_write(APIC_LVTERR, v | APIC_LVT_MASKED); } /* * Careful: we have to set masks only first to deassert * any level-triggered sources. */ v = apic_read(APIC_LVTT); - apic_write_around(APIC_LVTT, v | APIC_LVT_MASKED); + apic_write(APIC_LVTT, v | APIC_LVT_MASKED); v = apic_read(APIC_LVT0); - apic_write_around(APIC_LVT0, v | APIC_LVT_MASKED); + apic_write(APIC_LVT0, v | APIC_LVT_MASKED); v = apic_read(APIC_LVT1); - apic_write_around(APIC_LVT1, v | APIC_LVT_MASKED); + apic_write(APIC_LVT1, v | APIC_LVT_MASKED); if (maxlvt >= 4) { v = apic_read(APIC_LVTPC); - apic_write_around(APIC_LVTPC, v | APIC_LVT_MASKED); + apic_write(APIC_LVTPC, v | APIC_LVT_MASKED); } /* lets not touch this if we didn't frob it */ #ifdef CONFIG_X86_MCE_P4THERMAL if (maxlvt >= 5) { v = apic_read(APIC_LVTTHMR); - apic_write_around(APIC_LVTTHMR, v | APIC_LVT_MASKED); + apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED); } #endif /* * Clean APIC state for other OSs: */ - apic_write_around(APIC_LVTT, APIC_LVT_MASKED); - apic_write_around(APIC_LVT0, APIC_LVT_MASKED); - apic_write_around(APIC_LVT1, APIC_LVT_MASKED); + apic_write(APIC_LVTT, APIC_LVT_MASKED); + apic_write(APIC_LVT0, APIC_LVT_MASKED); + apic_write(APIC_LVT1, APIC_LVT_MASKED); if (maxlvt >= 3) - apic_write_around(APIC_LVTERR, APIC_LVT_MASKED); + apic_write(APIC_LVTERR, APIC_LVT_MASKED); if (maxlvt >= 4) - apic_write_around(APIC_LVTPC, APIC_LVT_MASKED); + apic_write(APIC_LVTPC, APIC_LVT_MASKED); #ifdef CONFIG_X86_MCE_P4THERMAL if (maxlvt >= 5) - apic_write_around(APIC_LVTTHMR, APIC_LVT_MASKED); + apic_write(APIC_LVTTHMR, APIC_LVT_MASKED); #endif /* Integrated APIC (!82489DX) ? */ if (lapic_is_integrated()) { @@ -756,7 +753,7 @@ void disable_local_APIC(void) */ value = apic_read(APIC_SPIV); value &= ~APIC_SPIV_APIC_ENABLED; - apic_write_around(APIC_SPIV, value); + apic_write(APIC_SPIV, value); /* * When LAPIC was disabled by the BIOS and enabled by the kernel, @@ -865,8 +862,8 @@ void __init sync_Arb_IDs(void) apic_wait_icr_idle(); apic_printk(APIC_DEBUG, "Synchronizing Arb IDs.\n"); - apic_write_around(APIC_ICR, APIC_DEST_ALLINC | APIC_INT_LEVELTRIG - | APIC_DM_INIT); + apic_write(APIC_ICR, + APIC_DEST_ALLINC | APIC_INT_LEVELTRIG | APIC_DM_INIT); } /* @@ -902,16 +899,16 @@ void __init init_bsp_APIC(void) else value |= APIC_SPIV_FOCUS_DISABLED; value |= SPURIOUS_APIC_VECTOR; - apic_write_around(APIC_SPIV, value); + apic_write(APIC_SPIV, value); /* * Set up the virtual wire mode. */ - apic_write_around(APIC_LVT0, APIC_DM_EXTINT); + apic_write(APIC_LVT0, APIC_DM_EXTINT); value = APIC_DM_NMI; if (!lapic_is_integrated()) /* 82489DX */ value |= APIC_LVT_LEVEL_TRIGGER; - apic_write_around(APIC_LVT1, value); + apic_write(APIC_LVT1, value); } static void __cpuinit lapic_setup_esr(void) @@ -926,7 +923,7 @@ static void __cpuinit lapic_setup_esr(void) /* enables sending errors */ value = ERROR_APIC_VECTOR; - apic_write_around(APIC_LVTERR, value); + apic_write(APIC_LVTERR, value); /* * spec says clear errors after enabling vector. */ @@ -989,7 +986,7 @@ void __cpuinit setup_local_APIC(void) */ value = apic_read(APIC_TASKPRI); value &= ~APIC_TPRI_MASK; - apic_write_around(APIC_TASKPRI, value); + apic_write(APIC_TASKPRI, value); /* * After a crash, we no longer service the interrupts and a pending @@ -1047,7 +1044,7 @@ void __cpuinit setup_local_APIC(void) * Set spurious IRQ vector */ value |= SPURIOUS_APIC_VECTOR; - apic_write_around(APIC_SPIV, value); + apic_write(APIC_SPIV, value); /* * Set up LVT0, LVT1: @@ -1069,7 +1066,7 @@ void __cpuinit setup_local_APIC(void) apic_printk(APIC_VERBOSE, "masked ExtINT on CPU#%d\n", smp_processor_id()); } - apic_write_around(APIC_LVT0, value); + apic_write(APIC_LVT0, value); /* * only the BP should see the LINT1 NMI signal, obviously. @@ -1080,7 +1077,7 @@ void __cpuinit setup_local_APIC(void) value = APIC_DM_NMI | APIC_LVT_MASKED; if (!integrated) /* 82489DX */ value |= APIC_LVT_LEVEL_TRIGGER; - apic_write_around(APIC_LVT1, value); + apic_write(APIC_LVT1, value); } void __cpuinit end_local_APIC_setup(void) @@ -1091,7 +1088,7 @@ void __cpuinit end_local_APIC_setup(void) /* Disable the local apic timer */ value = apic_read(APIC_LVTT); value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR); - apic_write_around(APIC_LVTT, value); + apic_write(APIC_LVTT, value); setup_apic_nmi_watchdog(NULL); apic_pm_activate(); @@ -1419,7 +1416,7 @@ void disconnect_bsp_APIC(int virt_wire_setup) value &= ~APIC_VECTOR_MASK; value |= APIC_SPIV_APIC_ENABLED; value |= 0xf; - apic_write_around(APIC_SPIV, value); + apic_write(APIC_SPIV, value); if (!virt_wire_setup) { /* @@ -1432,10 +1429,10 @@ void disconnect_bsp_APIC(int virt_wire_setup) APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED); value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING; value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_EXTINT); - apic_write_around(APIC_LVT0, value); + apic_write(APIC_LVT0, value); } else { /* Disable LVT0 */ - apic_write_around(APIC_LVT0, APIC_LVT_MASKED); + apic_write(APIC_LVT0, APIC_LVT_MASKED); } /* @@ -1449,7 +1446,7 @@ void disconnect_bsp_APIC(int virt_wire_setup) APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED); value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING; value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_NMI); - apic_write_around(APIC_LVT1, value); + apic_write(APIC_LVT1, value); } } diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c index 1b1c56bb338..c9b58a806e8 100644 --- a/arch/x86/kernel/cpu/bugs.c +++ b/arch/x86/kernel/cpu/bugs.c @@ -131,13 +131,7 @@ static void __init check_popad(void) * (for due to lack of "invlpg" and working WP on a i386) * - In order to run on anything without a TSC, we need to be * compiled for a i486. - * - In order to support the local APIC on a buggy Pentium machine, - * we need to be compiled with CONFIG_X86_GOOD_APIC disabled, - * which happens implicitly if compiled for a Pentium or lower - * (unless an advanced selection of CPU features is used) as an - * otherwise config implies a properly working local APIC without - * the need to do extra reads from the APIC. -*/ + */ static void __init check_config(void) { @@ -151,21 +145,6 @@ static void __init check_config(void) if (boot_cpu_data.x86 == 3) panic("Kernel requires i486+ for 'invlpg' and other features"); #endif - -/* - * If we were told we had a good local APIC, check for buggy Pentia, - * i.e. all B steppings and the C2 stepping of P54C when using their - * integrated APIC (see 11AP erratum in "Pentium Processor - * Specification Update"). - */ -#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86_GOOD_APIC) - if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL - && cpu_has_apic - && boot_cpu_data.x86 == 5 - && boot_cpu_data.x86_model == 2 - && (boot_cpu_data.x86_mask < 6 || boot_cpu_data.x86_mask == 11)) - panic("Kernel compiled for PMMX+, assumes a local APIC without the read-before-write bug!"); -#endif } diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c index 70609efdf1d..b75f2569b8f 100644 --- a/arch/x86/kernel/cpu/intel.c +++ b/arch/x86/kernel/cpu/intel.c @@ -227,6 +227,16 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c) if (cpu_has_bts) ds_init_intel(c); + /* + * See if we have a good local APIC by checking for buggy Pentia, + * i.e. all B steppings and the C2 stepping of P54C when using their + * integrated APIC (see 11AP erratum in "Pentium Processor + * Specification Update"). + */ + if (cpu_has_apic && (c->x86<<8 | c->x86_model<<4) == 0x520 && + (c->x86_mask < 0x6 || c->x86_mask == 0xb)) + set_cpu_cap(c, X86_FEATURE_11AP); + #ifdef CONFIG_X86_NUMAQ numaq_tsc_disable(); #endif diff --git a/arch/x86/kernel/cpu/mcheck/p4.c b/arch/x86/kernel/cpu/mcheck/p4.c index eef001ad3bd..9b60fce09f7 100644 --- a/arch/x86/kernel/cpu/mcheck/p4.c +++ b/arch/x86/kernel/cpu/mcheck/p4.c @@ -102,7 +102,7 @@ static void intel_init_thermal(struct cpuinfo_x86 *c) /* The temperature transition interrupt handler setup */ h = THERMAL_APIC_VECTOR; /* our delivery vector */ h |= (APIC_DM_FIXED | APIC_LVT_MASKED); /* we'll mask till we're ready */ - apic_write_around(APIC_LVTTHMR, h); + apic_write(APIC_LVTTHMR, h); rdmsr(MSR_IA32_THERM_INTERRUPT, l, h); wrmsr(MSR_IA32_THERM_INTERRUPT, l | 0x03 , h); @@ -114,7 +114,7 @@ static void intel_init_thermal(struct cpuinfo_x86 *c) wrmsr(MSR_IA32_MISC_ENABLE, l | (1<<3), h); l = apic_read(APIC_LVTTHMR); - apic_write_around(APIC_LVTTHMR, l & ~APIC_LVT_MASKED); + apic_write(APIC_LVTTHMR, l & ~APIC_LVT_MASKED); printk(KERN_INFO "CPU%d: Thermal monitoring enabled\n", cpu); /* enable thermal throttle processing */ diff --git a/arch/x86/kernel/io_apic_32.c b/arch/x86/kernel/io_apic_32.c index 558abf4c796..eabaf9244f5 100644 --- a/arch/x86/kernel/io_apic_32.c +++ b/arch/x86/kernel/io_apic_32.c @@ -756,7 +756,7 @@ void send_IPI_self(int vector) /* * Send the IPI. The write to APIC_ICR fires this off. */ - apic_write_around(APIC_ICR, cfg); + apic_write(APIC_ICR, cfg); } #endif /* !CONFIG_SMP */ @@ -2030,7 +2030,7 @@ static void mask_lapic_irq(unsigned int irq) unsigned long v; v = apic_read(APIC_LVT0); - apic_write_around(APIC_LVT0, v | APIC_LVT_MASKED); + apic_write(APIC_LVT0, v | APIC_LVT_MASKED); } static void unmask_lapic_irq(unsigned int irq) @@ -2038,7 +2038,7 @@ static void unmask_lapic_irq(unsigned int irq) unsigned long v; v = apic_read(APIC_LVT0); - apic_write_around(APIC_LVT0, v & ~APIC_LVT_MASKED); + apic_write(APIC_LVT0, v & ~APIC_LVT_MASKED); } static struct irq_chip lapic_chip __read_mostly = { @@ -2168,7 +2168,7 @@ static inline void __init check_timer(void) * The AEOI mode will finish them in the 8259A * automatically. */ - apic_write_around(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT); + apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT); init_8259A(1); timer_ack = (nmi_watchdog == NMI_IO_APIC && !APIC_INTEGRATED(ver)); @@ -2256,7 +2256,7 @@ static inline void __init check_timer(void) printk(KERN_INFO "...trying to set up timer as Virtual Wire IRQ..."); lapic_register_intr(0, vector); - apic_write_around(APIC_LVT0, APIC_DM_FIXED | vector); /* Fixed mode */ + apic_write(APIC_LVT0, APIC_DM_FIXED | vector); /* Fixed mode */ enable_8259A_irq(0); if (timer_irq_works()) { @@ -2264,14 +2264,14 @@ static inline void __init check_timer(void) goto out; } disable_8259A_irq(0); - apic_write_around(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_FIXED | vector); + apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_FIXED | vector); printk(" failed.\n"); printk(KERN_INFO "...trying to set up timer as ExtINT IRQ..."); init_8259A(0); make_8259A_irq(0); - apic_write_around(APIC_LVT0, APIC_DM_EXTINT); + apic_write(APIC_LVT0, APIC_DM_EXTINT); unlock_ExtINT_logic(); diff --git a/arch/x86/kernel/ipi.c b/arch/x86/kernel/ipi.c index 9d98cda39ad..3f7537b669d 100644 --- a/arch/x86/kernel/ipi.c +++ b/arch/x86/kernel/ipi.c @@ -70,7 +70,7 @@ void __send_IPI_shortcut(unsigned int shortcut, int vector) /* * Send the IPI. The write to APIC_ICR fires this off. */ - apic_write_around(APIC_ICR, cfg); + apic_write(APIC_ICR, cfg); } void send_IPI_self(int vector) @@ -98,7 +98,7 @@ static inline void __send_IPI_dest_field(unsigned long mask, int vector) * prepare target chip field */ cfg = __prepare_ICR2(mask); - apic_write_around(APIC_ICR2, cfg); + apic_write(APIC_ICR2, cfg); /* * program the ICR @@ -108,7 +108,7 @@ static inline void __send_IPI_dest_field(unsigned long mask, int vector) /* * Send the IPI. The write to APIC_ICR fires this off. */ - apic_write_around(APIC_ICR, cfg); + apic_write(APIC_ICR, cfg); } /* diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index ec024b3baad..384b49fed59 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -263,7 +263,7 @@ late_initcall(init_lapic_nmi_sysfs); static void __acpi_nmi_enable(void *__unused) { - apic_write_around(APIC_LVT0, APIC_DM_NMI); + apic_write(APIC_LVT0, APIC_DM_NMI); } /* @@ -277,7 +277,7 @@ void acpi_nmi_enable(void) static void __acpi_nmi_disable(void *__unused) { - apic_write_around(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); + apic_write(APIC_LVT0, APIC_DM_NMI | APIC_LVT_MASKED); } /* diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c index e0f571d58c1..5d7326a60b7 100644 --- a/arch/x86/kernel/paravirt.c +++ b/arch/x86/kernel/paravirt.c @@ -361,7 +361,6 @@ struct pv_cpu_ops pv_cpu_ops = { struct pv_apic_ops pv_apic_ops = { #ifdef CONFIG_X86_LOCAL_APIC .apic_write = native_apic_write, - .apic_write_atomic = native_apic_write_atomic, .apic_read = native_apic_read, .setup_boot_clock = setup_boot_APIC_clock, .setup_secondary_clock = setup_secondary_APIC_clock, diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c index 687376ab07e..f251f5c3882 100644 --- a/arch/x86/kernel/smpboot.c +++ b/arch/x86/kernel/smpboot.c @@ -546,8 +546,8 @@ static inline void __inquire_remote_apic(int apicid) printk(KERN_CONT "a previous APIC delivery may have failed\n"); - apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(apicid)); - apic_write_around(APIC_ICR, APIC_DM_REMRD | regs[i]); + apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(apicid)); + apic_write(APIC_ICR, APIC_DM_REMRD | regs[i]); timeout = 0; do { @@ -579,11 +579,11 @@ wakeup_secondary_cpu(int logical_apicid, unsigned long start_eip) int maxlvt; /* Target chip */ - apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(logical_apicid)); + apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(logical_apicid)); /* Boot on the stack */ /* Kick the second */ - apic_write_around(APIC_ICR, APIC_DM_NMI | APIC_DEST_LOGICAL); + apic_write(APIC_ICR, APIC_DM_NMI | APIC_DEST_LOGICAL); Dprintk("Waiting for send to finish...\n"); send_status = safe_apic_wait_icr_idle(); @@ -592,14 +592,9 @@ wakeup_secondary_cpu(int logical_apicid, unsigned long start_eip) * Give the other CPU some time to accept the IPI. */ udelay(200); - /* - * Due to the Pentium erratum 3AP. - */ maxlvt = lapic_get_maxlvt(); - if (maxlvt > 3) { - apic_read_around(APIC_SPIV); + if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */ apic_write(APIC_ESR, 0); - } accept_status = (apic_read(APIC_ESR) & 0xEF); Dprintk("NMI sent.\n"); @@ -625,12 +620,14 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip) return send_status; } + maxlvt = lapic_get_maxlvt(); + /* * Be paranoid about clearing APIC errors. */ if (APIC_INTEGRATED(apic_version[phys_apicid])) { - apic_read_around(APIC_SPIV); - apic_write(APIC_ESR, 0); + if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */ + apic_write(APIC_ESR, 0); apic_read(APIC_ESR); } @@ -639,13 +636,13 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip) /* * Turn INIT on target chip */ - apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid)); + apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid)); /* * Send IPI */ - apic_write_around(APIC_ICR, APIC_INT_LEVELTRIG | APIC_INT_ASSERT - | APIC_DM_INIT); + apic_write(APIC_ICR, + APIC_INT_LEVELTRIG | APIC_INT_ASSERT | APIC_DM_INIT); Dprintk("Waiting for send to finish...\n"); send_status = safe_apic_wait_icr_idle(); @@ -655,10 +652,10 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip) Dprintk("Deasserting INIT.\n"); /* Target chip */ - apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid)); + apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid)); /* Send IPI */ - apic_write_around(APIC_ICR, APIC_INT_LEVELTRIG | APIC_DM_INIT); + apic_write(APIC_ICR, APIC_INT_LEVELTRIG | APIC_DM_INIT); Dprintk("Waiting for send to finish...\n"); send_status = safe_apic_wait_icr_idle(); @@ -689,12 +686,10 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip) */ Dprintk("#startup loops: %d.\n", num_starts); - maxlvt = lapic_get_maxlvt(); - for (j = 1; j <= num_starts; j++) { Dprintk("Sending STARTUP #%d.\n", j); - apic_read_around(APIC_SPIV); - apic_write(APIC_ESR, 0); + if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */ + apic_write(APIC_ESR, 0); apic_read(APIC_ESR); Dprintk("After apic_write.\n"); @@ -703,12 +698,11 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip) */ /* Target chip */ - apic_write_around(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid)); + apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(phys_apicid)); /* Boot on the stack */ /* Kick the second */ - apic_write_around(APIC_ICR, APIC_DM_STARTUP - | (start_eip >> 12)); + apic_write(APIC_ICR, APIC_DM_STARTUP | (start_eip >> 12)); /* * Give the other CPU some time to accept the IPI. @@ -724,13 +718,8 @@ wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip) * Give the other CPU some time to accept the IPI. */ udelay(200); - /* - * Due to the Pentium erratum 3AP. - */ - if (maxlvt > 3) { - apic_read_around(APIC_SPIV); + if (maxlvt > 3) /* Due to the Pentium erratum 3AP. */ apic_write(APIC_ESR, 0); - } accept_status = (apic_read(APIC_ESR) & 0xEF); if (send_status || accept_status) break; diff --git a/arch/x86/kernel/vmi_32.c b/arch/x86/kernel/vmi_32.c index b15346092b7..0a1b1a9d922 100644 --- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -906,7 +906,6 @@ static inline int __init activate_vmi(void) #ifdef CONFIG_X86_LOCAL_APIC para_fill(pv_apic_ops.apic_read, APICRead); para_fill(pv_apic_ops.apic_write, APICWrite); - para_fill(pv_apic_ops.apic_write_atomic, APICWrite); #endif /* diff --git a/arch/x86/lguest/boot.c b/arch/x86/lguest/boot.c index 50dad44fb54..0313a5eec41 100644 --- a/arch/x86/lguest/boot.c +++ b/arch/x86/lguest/boot.c @@ -991,7 +991,6 @@ __init void lguest_init(void) #ifdef CONFIG_X86_LOCAL_APIC /* apic read/write intercepts */ pv_apic_ops.apic_write = lguest_apic_write; - pv_apic_ops.apic_write_atomic = lguest_apic_write; pv_apic_ops.apic_read = lguest_apic_read; #endif diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index bb508456ef5..7f26c371877 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1131,7 +1131,6 @@ static const struct pv_irq_ops xen_irq_ops __initdata = { static const struct pv_apic_ops xen_apic_ops __initdata = { #ifdef CONFIG_X86_LOCAL_APIC .apic_write = xen_apic_write, - .apic_write_atomic = xen_apic_write, .apic_read = xen_apic_read, .setup_boot_clock = paravirt_nop, .setup_secondary_clock = paravirt_nop, -- cgit v1.2.3 From 95c7c23b06bc92f1772b9c9460845f179ba8c39e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Tue, 15 Jul 2008 13:42:34 -0700 Subject: xen: report hypervisor version Various versions of the hypervisor have differences in what ABIs and features they support. Print some details into the boot log to help with remote debugging. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Ingo Molnar --- arch/x86/xen/enlighten.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index bb508456ef5..5328e46d9cf 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -167,10 +167,14 @@ void xen_vcpu_restore(void) static void __init xen_banner(void) { + unsigned version = HYPERVISOR_xen_version(XENVER_version, NULL); + struct xen_extraversion extra; + HYPERVISOR_xen_version(XENVER_extraversion, &extra); + printk(KERN_INFO "Booting paravirtualized kernel on %s\n", pv_info.name); - printk(KERN_INFO "Hypervisor signature: %s%s\n", - xen_start_info->magic, + printk(KERN_INFO "Xen version: %d.%d%s%s\n", + version >> 16, version & 0xffff, extra.extraversion, xen_feature(XENFEAT_mmu_pt_update_preserve_ad) ? " (preserve-AD)" : ""); } -- cgit v1.2.3 From fbdb7da91b0382d4b148d8b43c2eb4bab642bb5b Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Mon, 14 Jul 2008 15:34:09 -0700 Subject: x86_64: ia32_signal.c: use macro instead of immediate Make and use macro FIX_EFLAGS, instead of immediate value 0x40DD5 in ia32_restore_sigcontext(). Signed-off-by: Hiroshi Shimamoto Acked-by: "H. Peter Anvin" Signed-off-by: Ingo Molnar --- arch/x86/ia32/ia32_signal.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/ia32/ia32_signal.c b/arch/x86/ia32/ia32_signal.c index cb3856a18c8..dc9b9b9803f 100644 --- a/arch/x86/ia32/ia32_signal.c +++ b/arch/x86/ia32/ia32_signal.c @@ -36,6 +36,11 @@ #define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) +#define FIX_EFLAGS (X86_EFLAGS_AC | X86_EFLAGS_OF | \ + X86_EFLAGS_DF | X86_EFLAGS_TF | X86_EFLAGS_SF | \ + X86_EFLAGS_ZF | X86_EFLAGS_AF | X86_EFLAGS_PF | \ + X86_EFLAGS_CF) + asmlinkage int do_signal(struct pt_regs *regs, sigset_t *oldset); void signal_fault(struct pt_regs *regs, void __user *frame, char *where); @@ -248,7 +253,7 @@ static int ia32_restore_sigcontext(struct pt_regs *regs, regs->ss |= 3; err |= __get_user(tmpflags, &sc->flags); - regs->flags = (regs->flags & ~0x40DD5) | (tmpflags & 0x40DD5); + regs->flags = (regs->flags & ~FIX_EFLAGS) | (tmpflags & FIX_EFLAGS); /* disable syscall checks */ regs->orig_ax = -1; -- cgit v1.2.3 From 1f067167a83d1c7f80437fd1d32b55508aaca009 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Tue, 15 Jul 2008 00:02:28 -0700 Subject: x86: seperate memtest from init_64.c it's separate functionality that deserves its own file. This also prepares 32-bit memtest support. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/mm/Makefile | 1 + arch/x86/mm/init_64.c | 112 --------------------------------------------- arch/x86/mm/memtest.c | 123 ++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 124 insertions(+), 112 deletions(-) create mode 100644 arch/x86/mm/memtest.c (limited to 'arch') diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile index 9873716e9f7..1fbb844c3d7 100644 --- a/arch/x86/mm/Makefile +++ b/arch/x86/mm/Makefile @@ -21,3 +21,4 @@ obj-$(CONFIG_K8_NUMA) += k8topology_64.o endif obj-$(CONFIG_ACPI_NUMA) += srat_$(BITS).o +obj-$(CONFIG_MEMTEST) += memtest.o diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index 306049edd55..ec37121f670 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c @@ -517,118 +517,6 @@ static void __init init_gbpages(void) direct_gbpages = 0; } -#ifdef CONFIG_MEMTEST - -static void __init memtest(unsigned long start_phys, unsigned long size, - unsigned pattern) -{ - unsigned long i; - unsigned long *start; - unsigned long start_bad; - unsigned long last_bad; - unsigned long val; - unsigned long start_phys_aligned; - unsigned long count; - unsigned long incr; - - switch (pattern) { - case 0: - val = 0UL; - break; - case 1: - val = -1UL; - break; - case 2: - val = 0x5555555555555555UL; - break; - case 3: - val = 0xaaaaaaaaaaaaaaaaUL; - break; - default: - return; - } - - incr = sizeof(unsigned long); - start_phys_aligned = ALIGN(start_phys, incr); - count = (size - (start_phys_aligned - start_phys))/incr; - start = __va(start_phys_aligned); - start_bad = 0; - last_bad = 0; - - for (i = 0; i < count; i++) - start[i] = val; - for (i = 0; i < count; i++, start++, start_phys_aligned += incr) { - if (*start != val) { - if (start_phys_aligned == last_bad + incr) { - last_bad += incr; - } else { - if (start_bad) { - printk(KERN_CONT "\n %016lx bad mem addr %016lx - %016lx reserved", - val, start_bad, last_bad + incr); - reserve_early(start_bad, last_bad - start_bad, "BAD RAM"); - } - start_bad = last_bad = start_phys_aligned; - } - } - } - if (start_bad) { - printk(KERN_CONT "\n %016lx bad mem addr %016lx - %016lx reserved", - val, start_bad, last_bad + incr); - reserve_early(start_bad, last_bad - start_bad, "BAD RAM"); - } - -} - -/* default is disabled */ -static int memtest_pattern __initdata; - -static int __init parse_memtest(char *arg) -{ - if (arg) - memtest_pattern = simple_strtoul(arg, NULL, 0); - return 0; -} - -early_param("memtest", parse_memtest); - -static void __init early_memtest(unsigned long start, unsigned long end) -{ - u64 t_start, t_size; - unsigned pattern; - - if (!memtest_pattern) - return; - - printk(KERN_INFO "early_memtest: pattern num %d", memtest_pattern); - for (pattern = 0; pattern < memtest_pattern; pattern++) { - t_start = start; - t_size = 0; - while (t_start < end) { - t_start = find_e820_area_size(t_start, &t_size, 1); - - /* done ? */ - if (t_start >= end) - break; - if (t_start + t_size > end) - t_size = end - t_start; - - printk(KERN_CONT "\n %016llx - %016llx pattern %d", - (unsigned long long)t_start, - (unsigned long long)t_start + t_size, pattern); - - memtest(t_start, t_size, pattern); - - t_start += t_size; - } - } - printk(KERN_CONT "\n"); -} -#else -static void __init early_memtest(unsigned long start, unsigned long end) -{ -} -#endif - static unsigned long __init kernel_physical_mapping_init(unsigned long start, unsigned long end, unsigned long page_size_mask) diff --git a/arch/x86/mm/memtest.c b/arch/x86/mm/memtest.c new file mode 100644 index 00000000000..672e17f8262 --- /dev/null +++ b/arch/x86/mm/memtest.c @@ -0,0 +1,123 @@ +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +static void __init memtest(unsigned long start_phys, unsigned long size, + unsigned pattern) +{ + unsigned long i; + unsigned long *start; + unsigned long start_bad; + unsigned long last_bad; + unsigned long val; + unsigned long start_phys_aligned; + unsigned long count; + unsigned long incr; + + switch (pattern) { + case 0: + val = 0UL; + break; + case 1: + val = -1UL; + break; + case 2: +#ifdef CONFIG_X86_64 + val = 0x5555555555555555UL; +#else + val = 0x55555555UL; +#endif + break; + case 3: +#ifdef CONFIG_X86_64 + val = 0xaaaaaaaaaaaaaaaaUL; +#else + val = 0xaaaaaaaaUL; +#endif + break; + default: + return; + } + + incr = sizeof(unsigned long); + start_phys_aligned = ALIGN(start_phys, incr); + count = (size - (start_phys_aligned - start_phys))/incr; + start = __va(start_phys_aligned); + start_bad = 0; + last_bad = 0; + + for (i = 0; i < count; i++) + start[i] = val; + for (i = 0; i < count; i++, start++, start_phys_aligned += incr) { + if (*start != val) { + if (start_phys_aligned == last_bad + incr) { + last_bad += incr; + } else { + if (start_bad) { + printk(KERN_CONT "\n %010lx bad mem addr %010lx - %010lx reserved", + val, start_bad, last_bad + incr); + reserve_early(start_bad, last_bad - start_bad, "BAD RAM"); + } + start_bad = last_bad = start_phys_aligned; + } + } + } + if (start_bad) { + printk(KERN_CONT "\n %016lx bad mem addr %010lx - %010lx reserved", + val, start_bad, last_bad + incr); + reserve_early(start_bad, last_bad - start_bad, "BAD RAM"); + } + +} + +/* default is disabled */ +static int memtest_pattern __initdata; + +static int __init parse_memtest(char *arg) +{ + if (arg) + memtest_pattern = simple_strtoul(arg, NULL, 0); + return 0; +} + +early_param("memtest", parse_memtest); + +void __init early_memtest(unsigned long start, unsigned long end) +{ + u64 t_start, t_size; + unsigned pattern; + + if (!memtest_pattern) + return; + + printk(KERN_INFO "early_memtest: pattern num %d", memtest_pattern); + for (pattern = 0; pattern < memtest_pattern; pattern++) { + t_start = start; + t_size = 0; + while (t_start < end) { + t_start = find_e820_area_size(t_start, &t_size, 1); + + /* done ? */ + if (t_start >= end) + break; + if (t_start + t_size > end) + t_size = end - t_start; + + printk(KERN_CONT "\n %010llx - %010llx pattern %d", + (unsigned long long)t_start, + (unsigned long long)t_start + t_size, pattern); + + memtest(t_start, t_size, pattern); + + t_start += t_size; + } + } + printk(KERN_CONT "\n"); +} -- cgit v1.2.3 From caadbdce240c43e3e46c82fce6c00eb7f01e1beb Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Tue, 15 Jul 2008 00:03:44 -0700 Subject: x86: enable memory tester support on 32-bit only supports memory below max_low_pfn. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/Kconfig | 1 - arch/x86/mm/init_32.c | 3 +++ 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 96e0c2ebc38..03980cb0429 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -447,7 +447,6 @@ config PARAVIRT_DEBUG config MEMTEST bool "Memtest" - depends on X86_64 help This option adds a kernel parameter 'memtest', which allows memtest to be set. diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c index 9689a5138e6..3eeab6d0065 100644 --- a/arch/x86/mm/init_32.c +++ b/arch/x86/mm/init_32.c @@ -844,6 +844,9 @@ unsigned long __init_refok init_memory_mapping(unsigned long start, reserve_early(table_start << PAGE_SHIFT, table_end << PAGE_SHIFT, "PGTABLE"); + if (!after_init_bootmem) + early_memtest(start, end); + return end >> PAGE_SHIFT; } -- cgit v1.2.3 From 89b3b1f41bd94085da2f08dcb719bdbf7e8e9d57 Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Tue, 15 Jul 2008 21:02:54 +0400 Subject: x86: apic_64 - make calibrate_APIC_clock to return error code Make calibration_result to return error and check calibration_result to be sufficient inside calibrate_APIC_clock. Signed-off-by: Cyrill Gorcunov Cc: Cyrill Gorcunov Cc: macro@linux-mips.org Cc: yhlu.kernel@gmail.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic_64.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/apic_64.c b/arch/x86/kernel/apic_64.c index 1e3d32e27c1..ce294d623e5 100644 --- a/arch/x86/kernel/apic_64.c +++ b/arch/x86/kernel/apic_64.c @@ -314,7 +314,7 @@ static void setup_APIC_timer(void) #define TICK_COUNT 100000000 -static void __init calibrate_APIC_clock(void) +static int __init calibrate_APIC_clock(void) { unsigned apic, apic_start; unsigned long tsc, tsc_start; @@ -368,6 +368,17 @@ static void __init calibrate_APIC_clock(void) clockevent_delta2ns(0xF, &lapic_clockevent); calibration_result = result / HZ; + + /* + * Do a sanity check on the APIC calibration result + */ + if (calibration_result < (1000000 / HZ)) { + printk(KERN_WARNING + "APIC frequency too slow, disabling apic timer\n"); + return -1; + } + + return 0; } /* @@ -394,14 +405,7 @@ void __init setup_boot_APIC_clock(void) } printk(KERN_INFO "Using local APIC timer interrupts.\n"); - calibrate_APIC_clock(); - - /* - * Do a sanity check on the APIC calibration result - */ - if (calibration_result < (1000000 / HZ)) { - printk(KERN_WARNING - "APIC frequency too slow, disabling apic timer\n"); + if (calibrate_APIC_clock()) { /* No broadcast on UP ! */ if (num_possible_cpus() > 1) setup_APIC_timer(); -- cgit v1.2.3 From 836c129de971d526b6e85b8ad760bd635a00215e Mon Sep 17 00:00:00 2001 From: Cyrill Gorcunov Date: Tue, 15 Jul 2008 21:02:55 +0400 Subject: x86: apic_32 - introduce calibrate_APIC_clock Introduce calibrate_APIC_clock so it could help in further 32/64bit apic code merging. Signed-off-by: Cyrill Gorcunov Cc: Cyrill Gorcunov Cc: macro@linux-mips.org Cc: yhlu.kernel@gmail.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic_32.c | 93 +++++++++++++++++++++++++---------------------- 1 file changed, 50 insertions(+), 43 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/apic_32.c b/arch/x86/kernel/apic_32.c index 2bc1186cc95..d2a7eb511d6 100644 --- a/arch/x86/kernel/apic_32.c +++ b/arch/x86/kernel/apic_32.c @@ -369,12 +369,7 @@ static void __init lapic_cal_handler(struct clock_event_device *dev) } } -/* - * Setup the boot APIC - * - * Calibrate and verify the result. - */ -void __init setup_boot_APIC_clock(void) +static int __init calibrate_APIC_clock(void) { struct clock_event_device *levt = &__get_cpu_var(lapic_events); const long pm_100ms = PMTMR_TICKS_PER_SEC/10; @@ -384,24 +379,6 @@ void __init setup_boot_APIC_clock(void) long delta, deltapm; int pm_referenced = 0; - /* - * The local apic timer can be disabled via the kernel - * commandline or from the CPU detection code. Register the lapic - * timer as a dummy clock event source on SMP systems, so the - * broadcast mechanism is used. On UP systems simply ignore it. - */ - if (local_apic_timer_disabled) { - /* No broadcast on UP ! */ - if (num_possible_cpus() > 1) { - lapic_clockevent.mult = 1; - setup_APIC_timer(); - } - return; - } - - apic_printk(APIC_VERBOSE, "Using local APIC timer interrupts.\n" - "calibrating APIC timer ...\n"); - local_irq_disable(); /* Replace the global interrupt handler */ @@ -486,8 +463,6 @@ void __init setup_boot_APIC_clock(void) calibration_result / (1000000 / HZ), calibration_result % (1000000 / HZ)); - local_apic_timer_verify_ok = 1; - /* * Do a sanity check on the APIC calibration result */ @@ -495,12 +470,11 @@ void __init setup_boot_APIC_clock(void) local_irq_enable(); printk(KERN_WARNING "APIC frequency too slow, disabling apic timer\n"); - /* No broadcast on UP ! */ - if (num_possible_cpus() > 1) - setup_APIC_timer(); - return; + return -1; } + local_apic_timer_verify_ok = 1; + /* We trust the pm timer based calibration */ if (!pm_referenced) { apic_printk(APIC_VERBOSE, "... verify APIC timer\n"); @@ -540,22 +514,55 @@ void __init setup_boot_APIC_clock(void) if (!local_apic_timer_verify_ok) { printk(KERN_WARNING "APIC timer disabled due to verification failure.\n"); + return -1; + } + + return 0; +} + +/* + * Setup the boot APIC + * + * Calibrate and verify the result. + */ +void __init setup_boot_APIC_clock(void) +{ + /* + * The local apic timer can be disabled via the kernel + * commandline or from the CPU detection code. Register the lapic + * timer as a dummy clock event source on SMP systems, so the + * broadcast mechanism is used. On UP systems simply ignore it. + */ + if (local_apic_timer_disabled) { /* No broadcast on UP ! */ - if (num_possible_cpus() == 1) - return; - } else { - /* - * If nmi_watchdog is set to IO_APIC, we need the - * PIT/HPET going. Otherwise register lapic as a dummy - * device. - */ - if (nmi_watchdog != NMI_IO_APIC) - lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY; - else - printk(KERN_WARNING "APIC timer registered as dummy," - " due to nmi_watchdog=%d!\n", nmi_watchdog); + if (num_possible_cpus() > 1) { + lapic_clockevent.mult = 1; + setup_APIC_timer(); + } + return; + } + + apic_printk(APIC_VERBOSE, "Using local APIC timer interrupts.\n" + "calibrating APIC timer ...\n"); + + if (calibrate_APIC_clock()) { + /* No broadcast on UP ! */ + if (num_possible_cpus() > 1) + setup_APIC_timer(); + return; } + /* + * If nmi_watchdog is set to IO_APIC, we need the + * PIT/HPET going. Otherwise register lapic as a dummy + * device. + */ + if (nmi_watchdog != NMI_IO_APIC) + lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY; + else + printk(KERN_WARNING "APIC timer registered as dummy," + " due to nmi_watchdog=%d!\n", nmi_watchdog); + /* Setup the lapic or request the broadcast */ setup_APIC_timer(); } -- cgit v1.2.3 From 17c44697f293cf24cbbf51b4a5bd15f4fbc88b90 Mon Sep 17 00:00:00 2001 From: "Maciej W. Rozycki" Date: Mon, 14 Jul 2008 17:18:04 +0100 Subject: x86: I/O APIC: Include required by some code Include for i8259A_lock used in print_PIC() -- #if-0-ed out by default. The 32-bit version gets it right already. The plan is to enable this code with "apic=debug" eventually. This will aid with debugging strange problems without the need to ask people to apply patches. Signed-off-by: Maciej W. Rozycki Signed-off-by: Ingo Molnar --- arch/x86/kernel/io_apic_64.c | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/kernel/io_apic_64.c b/arch/x86/kernel/io_apic_64.c index 6510cde36b3..834b06afef4 100644 --- a/arch/x86/kernel/io_apic_64.c +++ b/arch/x86/kernel/io_apic_64.c @@ -45,6 +45,7 @@ #include #include #include +#include #include #include #include -- cgit v1.2.3 From baa1318841d4bc95d783e6c15219b264720002c8 Mon Sep 17 00:00:00 2001 From: "Maciej W. Rozycki" Date: Mon, 14 Jul 2008 18:44:51 +0100 Subject: x86: APIC: Make apic_verbosity unsigned As a microoptimisation, make apic_verbosity unsigned. This will make apic_printk(APIC_QUIET, ...) expand into just printk(...) with the surrounding condition and a reference to apic_verbosity removed. Signed-off-by: Maciej W. Rozycki Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic_32.c | 2 +- arch/x86/kernel/apic_64.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/apic_32.c b/arch/x86/kernel/apic_32.c index d2a7eb511d6..7f30c0f3dbe 100644 --- a/arch/x86/kernel/apic_32.c +++ b/arch/x86/kernel/apic_32.c @@ -75,7 +75,7 @@ char system_vectors[NR_VECTORS] = { [0 ... NR_VECTORS-1] = SYS_VECTOR_FREE}; /* * Debug level, exported for io_apic.c */ -int apic_verbosity; +unsigned int apic_verbosity; int pic_mode; diff --git a/arch/x86/kernel/apic_64.c b/arch/x86/kernel/apic_64.c index ce294d623e5..98c70f044e1 100644 --- a/arch/x86/kernel/apic_64.c +++ b/arch/x86/kernel/apic_64.c @@ -54,7 +54,7 @@ EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok); /* * Debug level, exported for io_apic.c */ -int apic_verbosity; +unsigned int apic_verbosity; /* Have we found an MP table */ int smp_found_config; -- cgit v1.2.3 From 49a66a0bcea8737e82035a31ad0d8cd64d027a5d Mon Sep 17 00:00:00 2001 From: "Maciej W. Rozycki" Date: Mon, 14 Jul 2008 19:08:13 +0100 Subject: x86: I/O APIC: Always report how the timer has been set up Following recent (and less so) issues with the 8254 timer when routed through the I/O or local APIC, always report which configurations have been tried and which one has been set up eventually. This is so that logs posted by people for some other reason can be used as a cross-reference when investigating any possible future problems. The change unifies messages printed on 32-bit and 64-bit platforms and adds trailing newlines (removes leading ones), so that proper log level annotation can be used and any possible interspersed output will not cause a mess. I have chosen to use apic_printk(APIC_QUIET, ...) rather than printk(...) so that the distinction of these messages is maintained making possible future decisions about changes in this area easier. A change posted separately making apic_verbosity unsigned removes any extra code that would otherwise be generated as a result of this design decision. Signed-off-by: Maciej W. Rozycki Signed-off-by: Ingo Molnar --- arch/x86/kernel/io_apic_32.c | 39 ++++++++++++++++++++++----------------- arch/x86/kernel/io_apic_64.c | 40 ++++++++++++++++++++++------------------ 2 files changed, 44 insertions(+), 35 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/io_apic_32.c b/arch/x86/kernel/io_apic_32.c index eabaf9244f5..de9aa0e3a9c 100644 --- a/arch/x86/kernel/io_apic_32.c +++ b/arch/x86/kernel/io_apic_32.c @@ -2177,8 +2177,9 @@ static inline void __init check_timer(void) pin2 = ioapic_i8259.pin; apic2 = ioapic_i8259.apic; - printk(KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n", - vector, apic1, pin1, apic2, pin2); + apic_printk(APIC_QUIET, KERN_INFO "..TIMER: vector=0x%02X " + "apic1=%d pin1=%d apic2=%d pin2=%d\n", + vector, apic1, pin1, apic2, pin2); /* * Some BIOS writers are clueless and report the ExtINTA @@ -2216,12 +2217,13 @@ static inline void __init check_timer(void) } clear_IO_APIC_pin(apic1, pin1); if (!no_pin1) - printk(KERN_ERR "..MP-BIOS bug: " - "8254 timer not connected to IO-APIC\n"); + apic_printk(APIC_QUIET, KERN_ERR "..MP-BIOS bug: " + "8254 timer not connected to IO-APIC\n"); - printk(KERN_INFO "...trying to set up timer (IRQ0) " - "through the 8259A ... "); - printk("\n..... (found pin %d) ...", pin2); + apic_printk(APIC_QUIET, KERN_INFO "...trying to set up timer " + "(IRQ0) through the 8259A ...\n"); + apic_printk(APIC_QUIET, KERN_INFO + "..... (found apic %d pin %d) ...\n", apic2, pin2); /* * legacy devices should be connected to IO APIC #0 */ @@ -2230,7 +2232,7 @@ static inline void __init check_timer(void) unmask_IO_APIC_irq(0); enable_8259A_irq(0); if (timer_irq_works()) { - printk("works.\n"); + apic_printk(APIC_QUIET, KERN_INFO "....... works.\n"); timer_through_8259 = 1; if (nmi_watchdog == NMI_IO_APIC) { disable_8259A_irq(0); @@ -2244,30 +2246,33 @@ static inline void __init check_timer(void) */ disable_8259A_irq(0); clear_IO_APIC_pin(apic2, pin2); - printk(" failed.\n"); + apic_printk(APIC_QUIET, KERN_INFO "....... failed.\n"); } if (nmi_watchdog == NMI_IO_APIC) { - printk(KERN_WARNING "timer doesn't work through the IO-APIC - disabling NMI Watchdog!\n"); + apic_printk(APIC_QUIET, KERN_WARNING "timer doesn't work " + "through the IO-APIC - disabling NMI Watchdog!\n"); nmi_watchdog = NMI_NONE; } timer_ack = 0; - printk(KERN_INFO "...trying to set up timer as Virtual Wire IRQ..."); + apic_printk(APIC_QUIET, KERN_INFO + "...trying to set up timer as Virtual Wire IRQ...\n"); lapic_register_intr(0, vector); apic_write(APIC_LVT0, APIC_DM_FIXED | vector); /* Fixed mode */ enable_8259A_irq(0); if (timer_irq_works()) { - printk(" works.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... works.\n"); goto out; } disable_8259A_irq(0); apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_FIXED | vector); - printk(" failed.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... failed.\n"); - printk(KERN_INFO "...trying to set up timer as ExtINT IRQ..."); + apic_printk(APIC_QUIET, KERN_INFO + "...trying to set up timer as ExtINT IRQ...\n"); init_8259A(0); make_8259A_irq(0); @@ -2276,12 +2281,12 @@ static inline void __init check_timer(void) unlock_ExtINT_logic(); if (timer_irq_works()) { - printk(" works.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... works.\n"); goto out; } - printk(" failed :(.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... failed :(.\n"); panic("IO-APIC + timer doesn't work! Boot with apic=debug and send a " - "report. Then try booting with the 'noapic' option"); + "report. Then try booting with the 'noapic' option.\n"); out: local_irq_restore(flags); } diff --git a/arch/x86/kernel/io_apic_64.c b/arch/x86/kernel/io_apic_64.c index 834b06afef4..64a46affd85 100644 --- a/arch/x86/kernel/io_apic_64.c +++ b/arch/x86/kernel/io_apic_64.c @@ -1697,8 +1697,9 @@ static inline void __init check_timer(void) pin2 = ioapic_i8259.pin; apic2 = ioapic_i8259.apic; - apic_printk(APIC_VERBOSE,KERN_INFO "..TIMER: vector=0x%02X apic1=%d pin1=%d apic2=%d pin2=%d\n", - cfg->vector, apic1, pin1, apic2, pin2); + apic_printk(APIC_QUIET, KERN_INFO "..TIMER: vector=0x%02X " + "apic1=%d pin1=%d apic2=%d pin2=%d\n", + cfg->vector, apic1, pin1, apic2, pin2); /* * Some BIOS writers are clueless and report the ExtINTA @@ -1736,14 +1737,13 @@ static inline void __init check_timer(void) } clear_IO_APIC_pin(apic1, pin1); if (!no_pin1) - apic_printk(APIC_QUIET,KERN_ERR "..MP-BIOS bug: " + apic_printk(APIC_QUIET, KERN_ERR "..MP-BIOS bug: " "8254 timer not connected to IO-APIC\n"); - apic_printk(APIC_VERBOSE,KERN_INFO - "...trying to set up timer (IRQ0) " - "through the 8259A ... "); - apic_printk(APIC_VERBOSE,"\n..... (found apic %d pin %d) ...", - apic2, pin2); + apic_printk(APIC_QUIET, KERN_INFO "...trying to set up timer " + "(IRQ0) through the 8259A ...\n"); + apic_printk(APIC_QUIET, KERN_INFO + "..... (found apic %d pin %d) ...\n", apic2, pin2); /* * legacy devices should be connected to IO APIC #0 */ @@ -1752,7 +1752,7 @@ static inline void __init check_timer(void) unmask_IO_APIC_irq(0); enable_8259A_irq(0); if (timer_irq_works()) { - apic_printk(APIC_VERBOSE," works.\n"); + apic_printk(APIC_QUIET, KERN_INFO "....... works.\n"); timer_through_8259 = 1; if (nmi_watchdog == NMI_IO_APIC) { disable_8259A_irq(0); @@ -1766,29 +1766,32 @@ static inline void __init check_timer(void) */ disable_8259A_irq(0); clear_IO_APIC_pin(apic2, pin2); - apic_printk(APIC_VERBOSE," failed.\n"); + apic_printk(APIC_QUIET, KERN_INFO "....... failed.\n"); } if (nmi_watchdog == NMI_IO_APIC) { - printk(KERN_WARNING "timer doesn't work through the IO-APIC - disabling NMI Watchdog!\n"); + apic_printk(APIC_QUIET, KERN_WARNING "timer doesn't work " + "through the IO-APIC - disabling NMI Watchdog!\n"); nmi_watchdog = NMI_NONE; } - apic_printk(APIC_VERBOSE, KERN_INFO "...trying to set up timer as Virtual Wire IRQ..."); + apic_printk(APIC_QUIET, KERN_INFO + "...trying to set up timer as Virtual Wire IRQ...\n"); lapic_register_intr(0); apic_write(APIC_LVT0, APIC_DM_FIXED | cfg->vector); /* Fixed mode */ enable_8259A_irq(0); if (timer_irq_works()) { - apic_printk(APIC_VERBOSE," works.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... works.\n"); goto out; } disable_8259A_irq(0); apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_FIXED | cfg->vector); - apic_printk(APIC_VERBOSE," failed.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... failed.\n"); - apic_printk(APIC_VERBOSE, KERN_INFO "...trying to set up timer as ExtINT IRQ..."); + apic_printk(APIC_QUIET, KERN_INFO + "...trying to set up timer as ExtINT IRQ...\n"); init_8259A(0); make_8259A_irq(0); @@ -1797,11 +1800,12 @@ static inline void __init check_timer(void) unlock_ExtINT_logic(); if (timer_irq_works()) { - apic_printk(APIC_VERBOSE," works.\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... works.\n"); goto out; } - apic_printk(APIC_VERBOSE," failed :(.\n"); - panic("IO-APIC + timer doesn't work! Try using the 'noapic' kernel parameter\n"); + apic_printk(APIC_QUIET, KERN_INFO "..... failed :(.\n"); + panic("IO-APIC + timer doesn't work! Boot with apic=debug and send a " + "report. Then try booting with the 'noapic' option.\n"); out: local_irq_restore(flags); } -- cgit v1.2.3 From 7019cc2dd6fafcdc6b104005482dc910dcdbb797 Mon Sep 17 00:00:00 2001 From: Russ Anderson Date: Wed, 9 Jul 2008 15:27:19 -0500 Subject: x86 BIOS interface for RTC on SGI UV Real-time code needs to know the number of cycles per second on SGI UV. The information is provided via a run time BIOS call. This patch provides the linux side of that interface. This is the first of several run time BIOS calls to be defined in uv/bios.h and bios_uv.c. Note that BIOS_CALL() is just a stub for now. The bios side is being worked on. Signed-off-by: Russ Anderson Cc: Jack Steiner Signed-off-by: Ingo Molnar --- arch/x86/kernel/Makefile | 1 + arch/x86/kernel/bios_uv.c | 48 ++++++++++++++++++++++++++++++++++++++++ arch/x86/kernel/genx2apic_uv_x.c | 23 +++++++++++++++++++ 3 files changed, 72 insertions(+) create mode 100644 arch/x86/kernel/bios_uv.c (limited to 'arch') diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile index da140611bb5..b78a17b1281 100644 --- a/arch/x86/kernel/Makefile +++ b/arch/x86/kernel/Makefile @@ -102,6 +102,7 @@ obj-$(CONFIG_OLPC) += olpc.o # 64 bit specific files ifeq ($(CONFIG_X86_64),y) obj-y += genapic_64.o genapic_flat_64.o genx2apic_uv_x.o tlb_uv.o + obj-y += bios_uv.o obj-$(CONFIG_X86_PM_TIMER) += pmtimer_64.o obj-$(CONFIG_AUDIT) += audit_64.o diff --git a/arch/x86/kernel/bios_uv.c b/arch/x86/kernel/bios_uv.c new file mode 100644 index 00000000000..c639bd55391 --- /dev/null +++ b/arch/x86/kernel/bios_uv.c @@ -0,0 +1,48 @@ +/* + * BIOS run time interface routines. + * + * Copyright (c) 2008 Silicon Graphics, Inc. All Rights Reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#include + +const char * +x86_bios_strerror(long status) +{ + const char *str; + switch (status) { + case 0: str = "Call completed without error"; break; + case -1: str = "Not implemented"; break; + case -2: str = "Invalid argument"; break; + case -3: str = "Call completed with error"; break; + default: str = "Unknown BIOS status code"; break; + } + return str; +} + +long +x86_bios_freq_base(unsigned long which, unsigned long *ticks_per_second, + unsigned long *drift_info) +{ + struct uv_bios_retval isrv; + + BIOS_CALL(isrv, BIOS_FREQ_BASE, which, 0, 0, 0, 0, 0, 0); + *ticks_per_second = isrv.v0; + *drift_info = isrv.v1; + return isrv.status; +} +EXPORT_SYMBOL_GPL(x86_bios_freq_base); diff --git a/arch/x86/kernel/genx2apic_uv_x.c b/arch/x86/kernel/genx2apic_uv_x.c index 711f11c30b0..3c392934069 100644 --- a/arch/x86/kernel/genx2apic_uv_x.c +++ b/arch/x86/kernel/genx2apic_uv_x.c @@ -24,6 +24,7 @@ #include #include #include +#include DEFINE_PER_CPU(struct uv_hub_info_s, __uv_hub_info); EXPORT_PER_CPU_SYMBOL_GPL(__uv_hub_info); @@ -40,6 +41,9 @@ EXPORT_SYMBOL_GPL(uv_cpu_to_blade); short uv_possible_blades; EXPORT_SYMBOL_GPL(uv_possible_blades); +unsigned long sn_rtc_cycles_per_second; +EXPORT_SYMBOL(sn_rtc_cycles_per_second); + /* Start with all IRQs pointing to boot CPU. IRQ balancing will shift them. */ static cpumask_t uv_target_cpus(void) @@ -272,6 +276,23 @@ static __init void map_mmioh_high(int max_pnode) map_high("MMIOH", mmioh.s.base, shift, map_uc); } +static __init void uv_rtc_init(void) +{ + long status, ticks_per_sec, drift; + + status = + x86_bios_freq_base(BIOS_FREQ_BASE_REALTIME_CLOCK, &ticks_per_sec, + &drift); + if (status != 0 || ticks_per_sec < 100000) { + printk(KERN_WARNING + "unable to determine platform RTC clock frequency, " + "guessing.\n"); + /* BIOS gives wrong value for clock freq. so guess */ + sn_rtc_cycles_per_second = 1000000000000UL / 30000UL; + } else + sn_rtc_cycles_per_second = ticks_per_sec; +} + static __init void uv_system_init(void) { union uvh_si_addr_map_config_u m_n_config; @@ -326,6 +347,8 @@ static __init void uv_system_init(void) gnode_upper = (((unsigned long)node_id.s.node_id) & ~((1 << n_val) - 1)) << m_val; + uv_rtc_init(); + for_each_present_cpu(cpu) { nid = cpu_to_node(cpu); pnode = uv_apicid_to_pnode(per_cpu(x86_cpu_to_apicid, cpu)); -- cgit v1.2.3 From 78cbac65fd77242f3e5d77f4d7a71e8bc869fe4d Mon Sep 17 00:00:00 2001 From: Alexander van Heukelum Date: Thu, 10 Jul 2008 21:14:52 +0200 Subject: x86: traps_xx: refactor die() like in x86_64 Make the diff between the traps_32.c and traps_64.c a bit smaller. Change traps_32.c to look more like traps_64.c: - move lock information to file scope - split out oops_begin() and oops_end() from die() - increment nest counter in oops_begin Only whitespace change in traps_64.c No functional changes intended. Signed-off-by: Alexander van Heukelum Acked-by: Cyrill Gorcunov Signed-off-by: Ingo Molnar --- arch/x86/kernel/traps_32.c | 95 +++++++++++++++++++++++++--------------------- arch/x86/kernel/traps_64.c | 2 +- 2 files changed, 52 insertions(+), 45 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/traps_32.c b/arch/x86/kernel/traps_32.c index 8a768973c4f..51cccde376a 100644 --- a/arch/x86/kernel/traps_32.c +++ b/arch/x86/kernel/traps_32.c @@ -383,6 +383,54 @@ int is_valid_bugaddr(unsigned long ip) return ud2 == 0x0b0f; } +static raw_spinlock_t die_lock = __RAW_SPIN_LOCK_UNLOCKED; +static int die_owner = -1; +static unsigned int die_nest_count; + +unsigned __kprobes long oops_begin(void) +{ + unsigned long flags; + + oops_enter(); + + if (die_owner != raw_smp_processor_id()) { + console_verbose(); + raw_local_irq_save(flags); + __raw_spin_lock(&die_lock); + die_owner = smp_processor_id(); + die_nest_count = 0; + bust_spinlocks(1); + } else { + raw_local_irq_save(flags); + } + die_nest_count++; + return flags; +} + +void __kprobes oops_end(unsigned long flags, struct pt_regs *regs, int signr) +{ + bust_spinlocks(0); + die_owner = -1; + add_taint(TAINT_DIE); + __raw_spin_unlock(&die_lock); + raw_local_irq_restore(flags); + + if (!regs) + return; + + if (kexec_should_crash(current)) + crash_kexec(regs); + + if (in_interrupt()) + panic("Fatal exception in interrupt"); + + if (panic_on_oops) + panic("Fatal exception"); + + oops_exit(); + do_exit(signr); +} + int __kprobes __die(const char *str, struct pt_regs *regs, long err) { unsigned short ss; @@ -423,31 +471,9 @@ int __kprobes __die(const char *str, struct pt_regs *regs, long err) */ void die(const char *str, struct pt_regs *regs, long err) { - static struct { - raw_spinlock_t lock; - u32 lock_owner; - int lock_owner_depth; - } die = { - .lock = __RAW_SPIN_LOCK_UNLOCKED, - .lock_owner = -1, - .lock_owner_depth = 0 - }; - unsigned long flags; - - oops_enter(); - - if (die.lock_owner != raw_smp_processor_id()) { - console_verbose(); - raw_local_irq_save(flags); - __raw_spin_lock(&die.lock); - die.lock_owner = smp_processor_id(); - die.lock_owner_depth = 0; - bust_spinlocks(1); - } else { - raw_local_irq_save(flags); - } + unsigned long flags = oops_begin(); - if (++die.lock_owner_depth < 3) { + if (die_nest_count < 3) { report_bug(regs->ip, regs); if (__die(str, regs, err)) @@ -456,26 +482,7 @@ void die(const char *str, struct pt_regs *regs, long err) printk(KERN_EMERG "Recursive die() failure, output suppressed\n"); } - bust_spinlocks(0); - die.lock_owner = -1; - add_taint(TAINT_DIE); - __raw_spin_unlock(&die.lock); - raw_local_irq_restore(flags); - - if (!regs) - return; - - if (kexec_should_crash(current)) - crash_kexec(regs); - - if (in_interrupt()) - panic("Fatal exception in interrupt"); - - if (panic_on_oops) - panic("Fatal exception"); - - oops_exit(); - do_exit(SIGSEGV); + oops_end(flags, regs, SIGSEGV); } static inline void diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c index 2696a683778..babdbe673b7 100644 --- a/arch/x86/kernel/traps_64.c +++ b/arch/x86/kernel/traps_64.c @@ -518,7 +518,7 @@ unsigned __kprobes long oops_begin(void) } void __kprobes oops_end(unsigned long flags, struct pt_regs *regs, int signr) -{ +{ die_owner = -1; bust_spinlocks(0); die_nest_count--; -- cgit v1.2.3 From 7dedcee394a3f61475d08002bd12e8068d044216 Mon Sep 17 00:00:00 2001 From: Alexander van Heukelum Date: Thu, 10 Jul 2008 21:16:39 +0200 Subject: x86: traps_xx: modify x86_64 to use _log_lvl variants i386 has show_trace_log_lvl and show_stack_log_lvl, allowing traces to be emitted with log-level annotations. This patch introduces them to x86_64, but log_lvl is only ever set to an empty string. Output of traces is unchanged. i386-chunk is whitespace-only. Signed-off-by: Alexander van Heukelum Cc: Cyrill Gorcunov Signed-off-by: Ingo Molnar --- arch/x86/kernel/traps_32.c | 2 +- arch/x86/kernel/traps_64.c | 24 ++++++++++++++++-------- 2 files changed, 17 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/traps_32.c b/arch/x86/kernel/traps_32.c index 51cccde376a..c971dce3847 100644 --- a/arch/x86/kernel/traps_32.c +++ b/arch/x86/kernel/traps_32.c @@ -256,7 +256,7 @@ static const struct stacktrace_ops print_trace_ops = { static void show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, - unsigned long *stack, unsigned long bp, char *log_lvl) + unsigned long *stack, unsigned long bp, char *log_lvl) { dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl); printk("%s =======================\n", log_lvl); diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c index babdbe673b7..c664e696200 100644 --- a/arch/x86/kernel/traps_64.c +++ b/arch/x86/kernel/traps_64.c @@ -355,17 +355,24 @@ static const struct stacktrace_ops print_trace_ops = { .address = print_trace_address, }; -void show_trace(struct task_struct *task, struct pt_regs *regs, - unsigned long *stack, unsigned long bp) +static void +show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, + unsigned long *stack, unsigned long bp, char *log_lvl) { printk("\nCall Trace:\n"); - dump_trace(task, regs, stack, bp, &print_trace_ops, NULL); + dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl); printk("\n"); } +void show_trace(struct task_struct *task, struct pt_regs *regs, + unsigned long *stack, unsigned long bp) +{ + show_trace_log_lvl(task, regs, stack, bp, ""); +} + static void -_show_stack(struct task_struct *task, struct pt_regs *regs, - unsigned long *sp, unsigned long bp) +show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, + unsigned long *sp, unsigned long bp, char *log_lvl) { unsigned long *stack; int i; @@ -399,12 +406,12 @@ _show_stack(struct task_struct *task, struct pt_regs *regs, printk(" %016lx", *stack++); touch_nmi_watchdog(); } - show_trace(task, regs, sp, bp); + show_trace_log_lvl(task, regs, sp, bp, log_lvl); } void show_stack(struct task_struct *task, unsigned long *sp) { - _show_stack(task, NULL, sp, 0); + show_stack_log_lvl(task, NULL, sp, 0, ""); } /* @@ -454,7 +461,8 @@ void show_registers(struct pt_regs *regs) u8 *ip; printk("Stack: "); - _show_stack(NULL, regs, (unsigned long *)sp, regs->bp); + show_stack_log_lvl(NULL, regs, (unsigned long *)sp, + regs->bp, ""); printk("\n"); printk(KERN_EMERG "Code: "); -- cgit v1.2.3 From 3f9b5cc018566ad9562df0648395649aebdbc5e0 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 18 Jul 2008 16:30:05 +0200 Subject: x86: re-enable OPTIMIZE_INLINING re-enable OPTIMIZE_INLINING more widely. Jeff Dike fixed the remaining outstanding issue in this commit: | commit 4f81c5350b44bcc501ab6f8a089b16d064b4d2f6 | Author: Jeff Dike | Date: Mon Jul 7 13:36:56 2008 -0400 | | [UML] fix gcc ICEs and unresolved externs [...] | This patch reintroduces unit-at-a-time for gcc >= 4.0, bringing back the | possibility of Uli's crash. If that happens, we'll debug it. it's still default-off and thus opt-in. Signed-off-by: Ingo Molnar --- arch/x86/Kconfig.debug | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug index ae36bfa814e..ffd5913b35d 100644 --- a/arch/x86/Kconfig.debug +++ b/arch/x86/Kconfig.debug @@ -287,7 +287,6 @@ config CPA_DEBUG config OPTIMIZE_INLINING bool "Allow gcc to uninline functions marked 'inline'" - depends on BROKEN help This option determines if the kernel forces gcc to inline the functions developers have marked 'inline'. Doing so takes away freedom from gcc to @@ -298,5 +297,7 @@ config OPTIMIZE_INLINING become the default in the future, until then this option is there to test gcc for this. + If unsure, say N. + endmenu -- cgit v1.2.3 From 8b2b9c1af065a45ef00c26964420489a53581779 Mon Sep 17 00:00:00 2001 From: Akinobu Mita Date: Tue, 15 Jul 2008 17:09:03 +0900 Subject: x86, intel_cacheinfo: fix use-after-free cache_kobject This avoids calling kobject_uevent() with cache_kobject that has already been deallocated in an error path. Signed-off-by: Akinobu Mita Cc: "H. Peter Anvin" Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/intel_cacheinfo.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/intel_cacheinfo.c b/arch/x86/kernel/cpu/intel_cacheinfo.c index 2c8afafa18e..ff517f0b8cc 100644 --- a/arch/x86/kernel/cpu/intel_cacheinfo.c +++ b/arch/x86/kernel/cpu/intel_cacheinfo.c @@ -780,15 +780,14 @@ static int __cpuinit cache_add_dev(struct sys_device * sys_dev) } kobject_put(per_cpu(cache_kobject, cpu)); cpuid4_cache_sysfs_exit(cpu); - break; + return retval; } kobject_uevent(&(this_object->kobj), KOBJ_ADD); } - if (!retval) - cpu_set(cpu, cache_dev_map); + cpu_set(cpu, cache_dev_map); kobject_uevent(per_cpu(cache_kobject, cpu), KOBJ_ADD); - return retval; + return 0; } static void __cpuinit cache_remove_dev(struct sys_device * sys_dev) -- cgit v1.2.3 From 47129654226b5bd418afe533ce4e11d6a0b6d6e4 Mon Sep 17 00:00:00 2001 From: Alexander Beregalov Date: Sun, 6 Jul 2008 20:13:49 +0400 Subject: x86 setup.c: cleanup includes x86: remove double includes in setup.c Signed-off-by: Alexander Beregalov Cc: yhlu.kernel@gmail.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/setup.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 531b55b8e81..4a2b8acc1d9 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -57,12 +57,8 @@ #include #include #include -#include #include -#include -#include -#include #include #include #include @@ -104,7 +100,6 @@ #include #include -#include #include #include #ifdef CONFIG_X86_64 -- cgit v1.2.3 From 9781f39fd209cd93ab98b669814191acc67f32fd Mon Sep 17 00:00:00 2001 From: Thomas Petazzoni Date: Thu, 10 Jul 2008 17:13:19 +0200 Subject: x86: consolidate the definition of the force_mwait variable The force_mwait variable iss defined either in arch/x86/kernel/cpu/amd.c or in arch/x86/kernel/setup_64.c, but it is only initialized and used in arch/x86/kernel/process.c. This patch moves the declaration to arch/x86/kernel/process.c. Signed-off-by: Thomas Petazzoni Cc: michael@free-electrons.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/amd.c | 2 -- arch/x86/kernel/process.c | 1 + 2 files changed, 1 insertion(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index 81a07ca65d4..cae9cabc303 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -24,8 +24,6 @@ extern void vide(void); __asm__(".align 4\nvide: ret"); -int force_mwait __cpuinitdata; - static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) { if (cpuid_eax(0x80000000) >= 0x80000007) { diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index 4d629c62f4f..74f2d196adb 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c @@ -15,6 +15,7 @@ unsigned long idle_nomwait; EXPORT_SYMBOL(idle_nomwait); struct kmem_cache *task_xstate_cachep; +static int force_mwait __cpuinitdata; int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) { -- cgit v1.2.3 From 5ff4789d045cdaec7629e027e4f8ff8e34308b81 Mon Sep 17 00:00:00 2001 From: Joerg Roedel Date: Mon, 14 Jul 2008 20:11:18 +0200 Subject: AMD IOMMU: set iommu for device from ACPI code too The device<->iommu relationship has to be set from the information in the ACPI table too. This patch adds this logic to the driver. Signed-off-by: Joerg Roedel Cc: iommu@lists.linux-foundation.org Cc: bhavna.sarathy@amd.com Cc: robert.richter@amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/amd_iommu_init.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c index 9bf1b8111b0..7661b02d720 100644 --- a/arch/x86/kernel/amd_iommu_init.c +++ b/arch/x86/kernel/amd_iommu_init.c @@ -426,11 +426,18 @@ static void set_dev_entry_bit(u16 devid, u8 bit) amd_iommu_dev_table[devid].data[i] |= (1 << _bit); } +/* Writes the specific IOMMU for a device into the rlookup table */ +static void __init set_iommu_for_device(struct amd_iommu *iommu, u16 devid) +{ + amd_iommu_rlookup_table[devid] = iommu; +} + /* * This function takes the device specific flags read from the ACPI * table and sets up the device table entry with that information */ -static void __init set_dev_entry_from_acpi(u16 devid, u32 flags, u32 ext_flags) +static void __init set_dev_entry_from_acpi(struct amd_iommu *iommu, + u16 devid, u32 flags, u32 ext_flags) { if (flags & ACPI_DEVFLAG_INITPASS) set_dev_entry_bit(devid, DEV_ENTRY_INIT_PASS); @@ -446,12 +453,8 @@ static void __init set_dev_entry_from_acpi(u16 devid, u32 flags, u32 ext_flags) set_dev_entry_bit(devid, DEV_ENTRY_LINT0_PASS); if (flags & ACPI_DEVFLAG_LINT1) set_dev_entry_bit(devid, DEV_ENTRY_LINT1_PASS); -} -/* Writes the specific IOMMU for a device into the rlookup table */ -static void __init set_iommu_for_device(struct amd_iommu *iommu, u16 devid) -{ - amd_iommu_rlookup_table[devid] = iommu; + set_iommu_for_device(iommu, devid); } /* @@ -550,11 +553,12 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, case IVHD_DEV_ALL: for (dev_i = iommu->first_device; dev_i <= iommu->last_device; ++dev_i) - set_dev_entry_from_acpi(dev_i, e->flags, 0); + set_dev_entry_from_acpi(iommu, dev_i, + e->flags, 0); break; case IVHD_DEV_SELECT: devid = e->devid; - set_dev_entry_from_acpi(devid, e->flags, 0); + set_dev_entry_from_acpi(iommu, devid, e->flags, 0); break; case IVHD_DEV_SELECT_RANGE_START: devid_start = e->devid; @@ -565,7 +569,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, case IVHD_DEV_ALIAS: devid = e->devid; devid_to = e->ext >> 8; - set_dev_entry_from_acpi(devid, e->flags, 0); + set_dev_entry_from_acpi(iommu, devid, e->flags, 0); amd_iommu_alias_table[devid] = devid_to; break; case IVHD_DEV_ALIAS_RANGE: @@ -577,7 +581,8 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, break; case IVHD_DEV_EXT_SELECT: devid = e->devid; - set_dev_entry_from_acpi(devid, e->flags, e->ext); + set_dev_entry_from_acpi(iommu, devid, e->flags, + e->ext); break; case IVHD_DEV_EXT_SELECT_RANGE: devid_start = e->devid; @@ -590,7 +595,7 @@ static void __init init_iommu_from_acpi(struct amd_iommu *iommu, for (dev_i = devid_start; dev_i <= devid; ++dev_i) { if (alias) amd_iommu_alias_table[dev_i] = devid_to; - set_dev_entry_from_acpi( + set_dev_entry_from_acpi(iommu, amd_iommu_alias_table[dev_i], flags, ext_flags); } -- cgit v1.2.3 From 6ac8d51f01d345af5ea4209004a9ea29b2f20891 Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Date: Tue, 15 Jul 2008 21:09:13 +0530 Subject: x86: introducing asm-x86/traps.h Declaring x86 traps under one hood. Declaring x86 do_traps before defining them. Signed-off-by: Jaswinder Singh Cc: Andi Kleen Cc: David Woodhouse Cc: Alexander van Heukelum Signed-off-by: Ingo Molnar --- arch/x86/kernel/traps_32.c | 21 +-------------------- arch/x86/kernel/traps_64.c | 22 +--------------------- 2 files changed, 2 insertions(+), 41 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/traps_32.c b/arch/x86/kernel/traps_32.c index c971dce3847..03df8e45e5a 100644 --- a/arch/x86/kernel/traps_32.c +++ b/arch/x86/kernel/traps_32.c @@ -58,6 +58,7 @@ #include #include #include +#include #include "mach_traps.h" @@ -77,26 +78,6 @@ char ignore_fpu_irq; gate_desc idt_table[256] __attribute__((__section__(".data.idt"))) = { { { { 0, 0 } } }, }; -asmlinkage void divide_error(void); -asmlinkage void debug(void); -asmlinkage void nmi(void); -asmlinkage void int3(void); -asmlinkage void overflow(void); -asmlinkage void bounds(void); -asmlinkage void invalid_op(void); -asmlinkage void device_not_available(void); -asmlinkage void coprocessor_segment_overrun(void); -asmlinkage void invalid_TSS(void); -asmlinkage void segment_not_present(void); -asmlinkage void stack_segment(void); -asmlinkage void general_protection(void); -asmlinkage void page_fault(void); -asmlinkage void coprocessor_error(void); -asmlinkage void simd_coprocessor_error(void); -asmlinkage void alignment_check(void); -asmlinkage void spurious_interrupt_bug(void); -asmlinkage void machine_check(void); - int panic_on_unrecovered_nmi; int kstack_depth_to_print = 24; static unsigned int code_bytes = 64; diff --git a/arch/x86/kernel/traps_64.c b/arch/x86/kernel/traps_64.c index c664e696200..3f18d73f420 100644 --- a/arch/x86/kernel/traps_64.c +++ b/arch/x86/kernel/traps_64.c @@ -51,30 +51,10 @@ #include #include #include +#include #include -asmlinkage void divide_error(void); -asmlinkage void debug(void); -asmlinkage void nmi(void); -asmlinkage void int3(void); -asmlinkage void overflow(void); -asmlinkage void bounds(void); -asmlinkage void invalid_op(void); -asmlinkage void device_not_available(void); -asmlinkage void double_fault(void); -asmlinkage void coprocessor_segment_overrun(void); -asmlinkage void invalid_TSS(void); -asmlinkage void segment_not_present(void); -asmlinkage void stack_segment(void); -asmlinkage void general_protection(void); -asmlinkage void page_fault(void); -asmlinkage void coprocessor_error(void); -asmlinkage void simd_coprocessor_error(void); -asmlinkage void alignment_check(void); -asmlinkage void spurious_interrupt_bug(void); -asmlinkage void machine_check(void); - int panic_on_unrecovered_nmi; int kstack_depth_to_print = 12; static unsigned int code_bytes = 64; -- cgit v1.2.3 From 1181f8b5f0302580af0958169ef4497c3eb57a61 Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Thu, 3 Jul 2008 13:12:13 -0700 Subject: x86_32: remove redundant KERN_INFO This printk has a KERN_ facility level in the format string. Signed-off-by: Hiroshi Shimamoto Signed-off-by: Ingo Molnar --- arch/x86/kernel/signal_32.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/signal_32.c b/arch/x86/kernel/signal_32.c index d9237363096..d633d801f85 100644 --- a/arch/x86/kernel/signal_32.c +++ b/arch/x86/kernel/signal_32.c @@ -212,7 +212,7 @@ asmlinkage unsigned long sys_sigreturn(unsigned long __unused) badframe: if (show_unhandled_signals && printk_ratelimit()) { - printk(KERN_INFO "%s%s[%d] bad frame in sigreturn frame:" + printk("%s%s[%d] bad frame in sigreturn frame:" "%p ip:%lx sp:%lx oeax:%lx", task_pid_nr(current) > 1 ? KERN_INFO : KERN_EMERG, current->comm, task_pid_nr(current), frame, regs->ip, -- cgit v1.2.3 From fa10c51a04a43ced5fd6033f19a74d2c82198b34 Mon Sep 17 00:00:00 2001 From: Alexander Beregalov Date: Wed, 9 Jul 2008 22:28:24 +0400 Subject: arch/x86/kernel/cpu/common_64.c: remove double inclusions x86: remove double inclusions in arch/x86/kernel/cpu/common_64.c Signed-off-by: Alexander Beregalov Cc: yhlu.kernel@gmail.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common_64.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index 7b8cc72feb4..2a4475beea4 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -7,12 +7,9 @@ #include #include #include -#include #include #include -#include #include -#include #include #include #include -- cgit v1.2.3 From 812b121d55316333a3480b294523d4e52f9dd366 Mon Sep 17 00:00:00 2001 From: Hiroshi Shimamoto Date: Wed, 16 Jul 2008 19:21:31 -0700 Subject: x86_64: ia32_signal.c: remove signal number conversion This was old code that was needed for iBCS and x86-64 never supported that. Signed-off-by: Hiroshi Shimamoto Signed-off-by: Ingo Molnar --- arch/x86/ia32/ia32_signal.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/ia32/ia32_signal.c b/arch/x86/ia32/ia32_signal.c index dc9b9b9803f..20af4c79579 100644 --- a/arch/x86/ia32/ia32_signal.c +++ b/arch/x86/ia32/ia32_signal.c @@ -520,7 +520,6 @@ int ia32_setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info, compat_sigset_t *set, struct pt_regs *regs) { struct rt_sigframe __user *frame; - struct exec_domain *ed = current_thread_info()->exec_domain; void __user *restorer; int err = 0; @@ -543,8 +542,7 @@ int ia32_setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info, if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame))) goto give_sigsegv; - err |= __put_user((ed && ed->signal_invmap && sig < 32 - ? ed->signal_invmap[sig] : sig), &frame->sig); + err |= __put_user(sig, &frame->sig); err |= __put_user(ptr_to_compat(&frame->info), &frame->pinfo); err |= __put_user(ptr_to_compat(&frame->uc), &frame->puc); err |= copy_siginfo_to_user32(&frame->info, info); -- cgit v1.2.3 From f2ba93929fdb91fd806be20e959a50f7db82790e Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Fri, 18 Jul 2008 13:35:37 +0100 Subject: x86: check function status in EDD boot code Without checking the return value of get_edd_info() and adding the entry only in the success case, 6 devices show up under /sys/firmware/edd/, no matter how many devices are actually present. Signed-off-by: Jan Beulich Signed-off-by: H. Peter Anvin --- arch/x86/boot/edd.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/boot/edd.c b/arch/x86/boot/edd.c index 03399d64013..d93cbc6464d 100644 --- a/arch/x86/boot/edd.c +++ b/arch/x86/boot/edd.c @@ -167,9 +167,8 @@ void query_edd(void) * Scan the BIOS-supported hard disks and query EDD * information... */ - get_edd_info(devno, &ei); - - if (boot_params.eddbuf_entries < EDDMAXNR) { + if (!get_edd_info(devno, &ei) + && boot_params.eddbuf_entries < EDDMAXNR) { memcpy(edp, &ei, sizeof ei); edp++; boot_params.eddbuf_entries++; -- cgit v1.2.3 From 369c99205f633d1e4038b15f5dc4a5500a4359c3 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Fri, 18 Jul 2008 13:37:53 +0100 Subject: x86: fix two modpost warnings Even though it's only the difference of the two __initdata symbols that's being calculated, modpost still doesn't like this. So rather calculate the size once in an __init function and store it for later use. Signed-off-by: Jan Beulich Signed-off-by: H. Peter Anvin --- arch/x86/vdso/vma.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/vdso/vma.c b/arch/x86/vdso/vma.c index 19a6cfaf5db..257ba4a10ab 100644 --- a/arch/x86/vdso/vma.c +++ b/arch/x86/vdso/vma.c @@ -21,7 +21,8 @@ unsigned int __read_mostly vdso_enabled = 1; extern char vdso_start[], vdso_end[]; extern unsigned short vdso_sync_cpuid; -struct page **vdso_pages; +static struct page **vdso_pages; +static unsigned vdso_size; static inline void *var_ref(void *p, char *name) { @@ -38,6 +39,7 @@ static int __init init_vdso_vars(void) int i; char *vbase; + vdso_size = npages << PAGE_SHIFT; vdso_pages = kmalloc(sizeof(struct page *) * npages, GFP_KERNEL); if (!vdso_pages) goto oom; @@ -101,20 +103,19 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack) struct mm_struct *mm = current->mm; unsigned long addr; int ret; - unsigned len = round_up(vdso_end - vdso_start, PAGE_SIZE); if (!vdso_enabled) return 0; down_write(&mm->mmap_sem); - addr = vdso_addr(mm->start_stack, len); - addr = get_unmapped_area(NULL, addr, len, 0, 0); + addr = vdso_addr(mm->start_stack, vdso_size); + addr = get_unmapped_area(NULL, addr, vdso_size, 0, 0); if (IS_ERR_VALUE(addr)) { ret = addr; goto up_fail; } - ret = install_special_mapping(mm, addr, len, + ret = install_special_mapping(mm, addr, vdso_size, VM_READ|VM_EXEC| VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC| VM_ALWAYSDUMP, -- cgit v1.2.3 From 08e1a13e7d14ba5d6a22bf4b8c6e11128d3bcdfe Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Fri, 18 Jul 2008 13:44:16 +0100 Subject: x86: reduce forbid_dac's visibility It's not used anywhere outside its declaring file. Signed-off-by: Jan Beulich Signed-off-by: H. Peter Anvin --- arch/x86/kernel/pci-dma.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index 8467ec2320f..702714bd151 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -9,8 +9,7 @@ #include #include -int forbid_dac __read_mostly; -EXPORT_SYMBOL(forbid_dac); +static int forbid_dac __read_mostly; const struct dma_mapping_ops *dma_ops; EXPORT_SYMBOL(dma_ops); -- cgit v1.2.3 From 08ad8afaa0f7343e9c64eec5dbbb178e390e03a2 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Fri, 18 Jul 2008 13:45:20 +0100 Subject: x86: reduce force_mwait visibility It's not used anywhere outside its single referencing file. Signed-off-by: Jan Beulich Signed-off-by: H. Peter Anvin --- arch/x86/kernel/cpu/amd.c | 2 -- arch/x86/kernel/process.c | 1 + 2 files changed, 1 insertion(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c index 81a07ca65d4..cae9cabc303 100644 --- a/arch/x86/kernel/cpu/amd.c +++ b/arch/x86/kernel/cpu/amd.c @@ -24,8 +24,6 @@ extern void vide(void); __asm__(".align 4\nvide: ret"); -int force_mwait __cpuinitdata; - static void __cpuinit early_init_amd(struct cpuinfo_x86 *c) { if (cpuid_eax(0x80000000) >= 0x80000007) { diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c index 158bd6a16f6..9f94bb1c811 100644 --- a/arch/x86/kernel/process.c +++ b/arch/x86/kernel/process.c @@ -199,6 +199,7 @@ static void poll_idle(void) * * idle=mwait overrides this decision and forces the usage of mwait. */ +static int __cpuinitdata force_mwait; #define MWAIT_INFO 0x05 #define MWAIT_ECX_EXTENDED_INFO 0x01 -- cgit v1.2.3 From 2ddf9b7b3e6660199269e34cfa27148440ddc3bf Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Fri, 18 Jul 2008 13:32:23 +0100 Subject: i386/xen: add proper unwind annotations to xen_sysenter_target Signed-off-by: Jan Beulich Cc: Jeremy Fitzhardinge Signed-off-by: H. Peter Anvin --- arch/x86/kernel/entry_32.S | 1 + 1 file changed, 1 insertion(+) (limited to 'arch') diff --git a/arch/x86/kernel/entry_32.S b/arch/x86/kernel/entry_32.S index 6bc07f0f120..ad5264c29e9 100644 --- a/arch/x86/kernel/entry_32.S +++ b/arch/x86/kernel/entry_32.S @@ -1024,6 +1024,7 @@ ENDPROC(kernel_thread_helper) ENTRY(xen_sysenter_target) RING0_INT_FRAME addl $5*4, %esp /* remove xen-provided frame */ + CFI_ADJUST_CFA_OFFSET -5*4 jmp sysenter_past_esp CFI_ENDPROC -- cgit v1.2.3 From ae79cdaacb5599781f8bb49f4bdd5723029669cf Mon Sep 17 00:00:00 2001 From: "venkatesh.pallipadi@intel.com" Date: Fri, 18 Jul 2008 16:08:13 -0700 Subject: x86: Add a arch directory for x86 under debugfs Add a directory for x86 arch under debugfs. Can be used to accumulate all x86 specific debugfs files. Signed-off-by: Venkatesh Pallipadi Signed-off-by: H. Peter Anvin --- arch/x86/kernel/kdebugfs.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/kdebugfs.c b/arch/x86/kernel/kdebugfs.c index c0320599171..f2d43bc7551 100644 --- a/arch/x86/kernel/kdebugfs.c +++ b/arch/x86/kernel/kdebugfs.c @@ -12,9 +12,13 @@ #include #include #include +#include #include +struct dentry *arch_debugfs_dir; +EXPORT_SYMBOL(arch_debugfs_dir); + #ifdef CONFIG_DEBUG_BOOT_PARAMS struct setup_data_node { u64 paddr; @@ -209,6 +213,10 @@ static int __init arch_kdebugfs_init(void) { int error = 0; + arch_debugfs_dir = debugfs_create_dir("x86", NULL); + if (!arch_debugfs_dir) + return -ENOMEM; + #ifdef CONFIG_DEBUG_BOOT_PARAMS error = boot_params_kdebugfs_init(); #endif -- cgit v1.2.3 From fec0962e0bed407927b9ff54bb0596a3ab7e4b61 Mon Sep 17 00:00:00 2001 From: "venkatesh.pallipadi@intel.com" Date: Fri, 18 Jul 2008 16:08:14 -0700 Subject: x86: Add a debugfs interface to dump PAT memtype Add a debugfs interface to list out all the PAT memtype reservations. Appears at debugfs x86/pat_memtype_list and output format is type @ - We do not hold the lock while printing the entire list. So, the list may not be a consistent copy in case where regions are getting added or deleted at the same time. Signed-off-by: Venkatesh Pallipadi Signed-off-by: H. Peter Anvin --- arch/x86/mm/pat.c | 88 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 88 insertions(+) (limited to 'arch') diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index d4585077977..0917a540a55 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -12,6 +12,8 @@ #include #include #include +#include +#include #include #include @@ -489,3 +491,89 @@ void unmap_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot) free_memtype(addr, addr + size); } + +#if defined(CONFIG_DEBUG_FS) + +/* get Nth element of the linked list */ +static struct memtype *memtype_get_idx(loff_t pos) +{ + struct memtype *list_node, *print_entry; + int i = 1; + + print_entry = kmalloc(sizeof(struct memtype), GFP_KERNEL); + if (!print_entry) + return NULL; + + spin_lock(&memtype_lock); + list_for_each_entry(list_node, &memtype_list, nd) { + if (pos == i) { + *print_entry = *list_node; + spin_unlock(&memtype_lock); + return print_entry; + } + ++i; + } + spin_unlock(&memtype_lock); + kfree(print_entry); + return NULL; +} + +static void *memtype_seq_start(struct seq_file *seq, loff_t *pos) +{ + if (*pos == 0) { + ++*pos; + seq_printf(seq, "PAT memtype list:\n"); + } + + return memtype_get_idx(*pos); +} + +static void *memtype_seq_next(struct seq_file *seq, void *v, loff_t *pos) +{ + ++*pos; + return memtype_get_idx(*pos); +} + +static void memtype_seq_stop(struct seq_file *seq, void *v) +{ +} + +static int memtype_seq_show(struct seq_file *seq, void *v) +{ + struct memtype *print_entry = (struct memtype *)v; + + seq_printf(seq, "%s @ 0x%Lx-0x%Lx\n", cattr_name(print_entry->type), + print_entry->start, print_entry->end); + kfree(print_entry); + return 0; +} + +static struct seq_operations memtype_seq_ops = { + .start = memtype_seq_start, + .next = memtype_seq_next, + .stop = memtype_seq_stop, + .show = memtype_seq_show, +}; + +static int memtype_seq_open(struct inode *inode, struct file *file) +{ + return seq_open(file, &memtype_seq_ops); +} + +static const struct file_operations memtype_fops = { + .open = memtype_seq_open, + .read = seq_read, + .llseek = seq_lseek, + .release = seq_release, +}; + +static int __init pat_memtype_list_init(void) +{ + debugfs_create_file("pat_memtype_list", S_IRUSR, arch_debugfs_dir, + NULL, &memtype_fops); + return 0; +} + +late_initcall(pat_memtype_list_init); + +#endif /* CONFIG_DEBUG_FS */ -- cgit v1.2.3 From e5849e71adcbb774ce40f09c1bcb48acca3b6da7 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Fri, 18 Jul 2008 17:28:40 -0700 Subject: x86: remove arch_get_ram_range no user now Signed-off-by: Yinghai Lu Signed-off-by: H. Peter Anvin --- arch/x86/kernel/e820.c | 21 --------------------- 1 file changed, 21 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c index 28c29180b38..df1b32fa88d 100644 --- a/arch/x86/kernel/e820.c +++ b/arch/x86/kernel/e820.c @@ -1367,24 +1367,3 @@ void __init setup_memory_map(void) printk(KERN_INFO "BIOS-provided physical RAM map:\n"); e820_print_map(who); } - -#ifdef CONFIG_X86_64 -int __init arch_get_ram_range(int slot, u64 *addr, u64 *size) -{ - int i; - - if (slot < 0 || slot >= e820.nr_map) - return -1; - for (i = slot; i < e820.nr_map; i++) { - if (e820.map[i].type != E820_RAM) - continue; - break; - } - if (i == e820.nr_map || e820.map[i].addr > (max_pfn << PAGE_SHIFT)) - return -1; - *addr = e820.map[i].addr; - *size = min_t(u64, e820.map[i].size + e820.map[i].addr, - max_pfn << PAGE_SHIFT) - *addr; - return i + 1; -} -#endif -- cgit v1.2.3 From d092633bff3b19faffc480fe9810805e7792a029 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 18 Jul 2008 00:26:59 +0200 Subject: Subject: devmem, x86: fix rename of CONFIG_NONPROMISC_DEVMEM From: Arjan van de Ven Date: Sat, 19 Jul 2008 15:47:17 -0700 CONFIG_NONPROMISC_DEVMEM was a rather confusing name - but renaming it to CONFIG_PROMISC_DEVMEM causes problems on architectures that do not support this feature; this patch renames it to CONFIG_STRICT_DEVMEM, so that architectures can opt-in into it. ( the polarity of the option is still the same as it was originally; it needs to be for now to not break architectures that don't have the infastructure yet to support this feature) Signed-off-by: Arjan van de Ven Cc: "V.Radhakrishnan" Signed-off-by: Ingo Molnar --- --- arch/x86/Kconfig.debug | 9 +++++---- arch/x86/configs/i386_defconfig | 2 +- arch/x86/configs/x86_64_defconfig | 2 +- arch/x86/mm/pat.c | 6 +++--- 4 files changed, 10 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug index f0cf5d99079..51c82147795 100644 --- a/arch/x86/Kconfig.debug +++ b/arch/x86/Kconfig.debug @@ -5,14 +5,15 @@ config TRACE_IRQFLAGS_SUPPORT source "lib/Kconfig.debug" -config PROMISC_DEVMEM - bool "Allow unlimited access to /dev/mem" - default y +config STRICT_DEVMEM + bool "Filter access to /dev/mem" help If this option is left on, you allow userspace (root) access to all of memory, including kernel and userspace memory. Accidental access to this is obviously disastrous, but specific access can - be used by people debugging the kernel. + be used by people debugging the kernel. Note that with PAT support + enabled, even in this case there are restrictions on /dev/mem + use due to the cache aliasing requirements. If this option is switched on, the /dev/mem file only allows userspace access to PCI space and the BIOS code and data regions. diff --git a/arch/x86/configs/i386_defconfig b/arch/x86/configs/i386_defconfig index 9bc34e2033e..4d73f53287b 100644 --- a/arch/x86/configs/i386_defconfig +++ b/arch/x86/configs/i386_defconfig @@ -2047,7 +2047,7 @@ CONFIG_PROVIDE_OHCI1394_DMA_INIT=y # CONFIG_SAMPLES is not set # CONFIG_KGDB is not set CONFIG_HAVE_ARCH_KGDB=y -# CONFIG_NONPROMISC_DEVMEM is not set +# CONFIG_STRICT_DEVMEM is not set CONFIG_EARLY_PRINTK=y CONFIG_DEBUG_STACKOVERFLOW=y CONFIG_DEBUG_STACK_USAGE=y diff --git a/arch/x86/configs/x86_64_defconfig b/arch/x86/configs/x86_64_defconfig index ae5124e064d..a4045242962 100644 --- a/arch/x86/configs/x86_64_defconfig +++ b/arch/x86/configs/x86_64_defconfig @@ -2012,7 +2012,7 @@ CONFIG_PROVIDE_OHCI1394_DMA_INIT=y # CONFIG_SAMPLES is not set # CONFIG_KGDB is not set CONFIG_HAVE_ARCH_KGDB=y -# CONFIG_NONPROMISC_DEVMEM is not set +# CONFIG_STRICT_DEVMEM is not set CONFIG_EARLY_PRINTK=y CONFIG_DEBUG_STACKOVERFLOW=y CONFIG_DEBUG_STACK_USAGE=y diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index c34dc483839..6bb597f4d70 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -373,8 +373,8 @@ pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, return vma_prot; } -#ifndef CONFIG_PROMISC_DEVMEM -/* This check is done in drivers/char/mem.c in case of !PROMISC_DEVMEM*/ +#ifdef CONFIG_STRICT_DEVMEM +/* This check is done in drivers/char/mem.c in case of STRICT_DEVMEM*/ static inline int range_is_allowed(unsigned long pfn, unsigned long size) { return 1; @@ -398,7 +398,7 @@ static inline int range_is_allowed(unsigned long pfn, unsigned long size) } return 1; } -#endif /* CONFIG_PROMISC_DEVMEM */ +#endif /* CONFIG_STRICT_DEVMEM */ int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn, unsigned long size, pgprot_t *vma_prot) -- cgit v1.2.3 From 5f1f2b3d9dbaee82cd532f28da459adcbf611499 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Fri, 18 Jul 2008 16:16:23 -0700 Subject: x86: improve debug printout: add target bootmem range in early_res_to_bootmem() Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/e820.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c index df1b32fa88d..6c60aeaac15 100644 --- a/arch/x86/kernel/e820.c +++ b/arch/x86/kernel/e820.c @@ -877,7 +877,8 @@ void __init early_res_to_bootmem(u64 start, u64 end) for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) count++; - printk(KERN_INFO "(%d early reservations) ==> bootmem\n", count); + printk(KERN_INFO "(%d early reservations) ==> bootmem [%010llx - %010llx]\n", + count, start, end); for (i = 0; i < count; i++) { struct early_res *r = &early_res[i]; printk(KERN_INFO " #%d [%010llx - %010llx] %16s", i, -- cgit v1.2.3 From 3c9cb6de1e5ad37d1558fdb0d9d2bed5a7bac0d9 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Sat, 19 Jul 2008 02:07:25 -0700 Subject: x86: introduce x86_quirks introduce x86_quirks array of boot-time quirk methods. No change in functionality intended. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/e820.c | 9 ++------- arch/x86/kernel/mpparse.c | 17 +++++------------ arch/x86/kernel/setup.c | 4 ++++ arch/x86/kernel/visws_quirks.c | 42 ++++++++++++++++++++---------------------- arch/x86/mach-default/setup.c | 24 ++++++++---------------- 5 files changed, 39 insertions(+), 57 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c index 6c60aeaac15..9af89078f7b 100644 --- a/arch/x86/kernel/e820.c +++ b/arch/x86/kernel/e820.c @@ -1299,11 +1299,6 @@ void __init e820_reserve_resources(void) } } -/* - * Non-standard memory setup can be specified via this quirk: - */ -char * (*arch_memory_setup_quirk)(void); - char *__init default_machine_specific_memory_setup(void) { char *who = "BIOS-e820"; @@ -1344,8 +1339,8 @@ char *__init default_machine_specific_memory_setup(void) char *__init __attribute__((weak)) machine_specific_memory_setup(void) { - if (arch_memory_setup_quirk) { - char *who = arch_memory_setup_quirk(); + if (x86_quirks->arch_memory_setup) { + char *who = x86_quirks->arch_memory_setup(); if (who) return who; diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c index 3b25e49380c..3cbd2df3abe 100644 --- a/arch/x86/kernel/mpparse.c +++ b/arch/x86/kernel/mpparse.c @@ -27,6 +27,7 @@ #include #include #include +#include #include #ifdef CONFIG_X86_32 @@ -725,12 +726,6 @@ static inline void __init construct_default_ISA_mptable(int mpc_default_type) static struct intel_mp_floating *mpf_found; -/* - * Machine specific quirk for finding the SMP config before other setup - * activities destroy the table: - */ -int (*mach_get_smp_config_quirk)(unsigned int early); - /* * Scan the memory blocks for an SMP configuration block. */ @@ -738,8 +733,8 @@ static void __init __get_smp_config(unsigned int early) { struct intel_mp_floating *mpf = mpf_found; - if (mach_get_smp_config_quirk) { - if (mach_get_smp_config_quirk(early)) + if (x86_quirks->mach_get_smp_config) { + if (x86_quirks->mach_get_smp_config(early)) return; } if (acpi_lapic && early) @@ -899,14 +894,12 @@ static int __init smp_scan_config(unsigned long base, unsigned long length, return 0; } -int (*mach_find_smp_config_quirk)(unsigned int reserve); - static void __init __find_smp_config(unsigned int reserve) { unsigned int address; - if (mach_find_smp_config_quirk) { - if (mach_find_smp_config_quirk(reserve)) + if (x86_quirks->mach_find_smp_config) { + if (x86_quirks->mach_find_smp_config(reserve)) return; } /* diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index 4a2b8acc1d9..bbcc13d0b56 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -574,6 +574,10 @@ static int __init setup_elfcorehdr(char *arg) early_param("elfcorehdr", setup_elfcorehdr); #endif +static struct x86_quirks default_x86_quirks __initdata; + +struct x86_quirks *x86_quirks __initdata = &default_x86_quirks; + /* * Determine if we were loaded by an EFI loader. If so, then we have also been * passed the efi memmap, systab, etc., so we should use these data structures diff --git a/arch/x86/kernel/visws_quirks.c b/arch/x86/kernel/visws_quirks.c index e94bdb6add1..41e01b145c4 100644 --- a/arch/x86/kernel/visws_quirks.c +++ b/arch/x86/kernel/visws_quirks.c @@ -73,7 +73,7 @@ int is_visws_box(void) return visws_board_type >= 0; } -static int __init visws_time_init_quirk(void) +static int __init visws_time_init(void) { printk(KERN_INFO "Starting Cobalt Timer system clock\n"); @@ -93,7 +93,7 @@ static int __init visws_time_init_quirk(void) return 0; } -static int __init visws_pre_intr_init_quirk(void) +static int __init visws_pre_intr_init(void) { init_VISWS_APIC_irqs(); @@ -114,7 +114,7 @@ EXPORT_SYMBOL(sgivwfb_mem_size); long long mem_size __initdata = 0; -static char * __init visws_memory_setup_quirk(void) +static char * __init visws_memory_setup(void) { long long gfx_mem_size = 8 * MB; @@ -176,7 +176,7 @@ static void visws_machine_power_off(void) outl(PIIX_SPECIAL_STOP, 0xCFC); } -static int __init visws_get_smp_config_quirk(unsigned int early) +static int __init visws_get_smp_config(unsigned int early) { /* * Prevent MP-table parsing by the generic code: @@ -192,7 +192,7 @@ extern unsigned int __cpuinitdata maxcpus; * No problem for Linux. */ -static void __init MP_processor_info (struct mpc_config_processor *m) +static void __init MP_processor_info(struct mpc_config_processor *m) { int ver, logical_apicid; physid_mask_t apic_cpus; @@ -232,7 +232,7 @@ static void __init MP_processor_info (struct mpc_config_processor *m) apic_version[m->mpc_apicid] = ver; } -int __init visws_find_smp_config_quirk(unsigned int reserve) +static int __init visws_find_smp_config(unsigned int reserve) { struct mpc_config_processor *mp = phys_to_virt(CO_CPU_TAB_PHYS); unsigned short ncpus = readw(phys_to_virt(CO_CPU_NUM_PHYS)); @@ -258,7 +258,17 @@ int __init visws_find_smp_config_quirk(unsigned int reserve) return 1; } -extern int visws_trap_init_quirk(void); +static int visws_trap_init(void); + +static struct x86_quirks visws_x86_quirks __initdata = { + .arch_time_init = visws_time_init, + .arch_pre_intr_init = visws_pre_intr_init, + .arch_memory_setup = visws_memory_setup, + .arch_intr_init = NULL, + .arch_trap_init = visws_trap_init, + .mach_get_smp_config = visws_get_smp_config, + .mach_find_smp_config = visws_find_smp_config, +}; void __init visws_early_detect(void) { @@ -272,16 +282,10 @@ void __init visws_early_detect(void) /* * Install special quirks for timer, interrupt and memory setup: - */ - arch_time_init_quirk = visws_time_init_quirk; - arch_pre_intr_init_quirk = visws_pre_intr_init_quirk; - arch_memory_setup_quirk = visws_memory_setup_quirk; - - /* * Fall back to generic behavior for traps: + * Override generic MP-table parsing: */ - arch_intr_init_quirk = NULL; - arch_trap_init_quirk = visws_trap_init_quirk; + x86_quirks = &visws_x86_quirks; /* * Install reboot quirks: @@ -294,12 +298,6 @@ void __init visws_early_detect(void) */ no_broadcast = 0; - /* - * Override generic MP-table parsing: - */ - mach_get_smp_config_quirk = visws_get_smp_config_quirk; - mach_find_smp_config_quirk = visws_find_smp_config_quirk; - #ifdef CONFIG_X86_IO_APIC /* * Turn off IO-APIC detection and initialization: @@ -426,7 +424,7 @@ static __init void cobalt_init(void) co_apic_read(CO_APIC_ID)); } -int __init visws_trap_init_quirk(void) +static int __init visws_trap_init(void) { lithium_init(); cobalt_init(); diff --git a/arch/x86/mach-default/setup.c b/arch/x86/mach-default/setup.c index 48278fa7d3d..631dbed9fb9 100644 --- a/arch/x86/mach-default/setup.c +++ b/arch/x86/mach-default/setup.c @@ -10,14 +10,6 @@ #include #include -/* - * Any quirks to be performed to initialize timers/irqs/etc? - */ -int (*arch_time_init_quirk)(void); -int (*arch_pre_intr_init_quirk)(void); -int (*arch_intr_init_quirk)(void); -int (*arch_trap_init_quirk)(void); - #ifdef CONFIG_HOTPLUG_CPU #define DEFAULT_SEND_IPI (1) #else @@ -37,8 +29,8 @@ int no_broadcast=DEFAULT_SEND_IPI; **/ void __init pre_intr_init_hook(void) { - if (arch_pre_intr_init_quirk) { - if (arch_pre_intr_init_quirk()) + if (x86_quirks->arch_pre_intr_init) { + if (x86_quirks->arch_pre_intr_init()) return; } init_ISA_irqs(); @@ -64,8 +56,8 @@ static struct irqaction irq2 = { **/ void __init intr_init_hook(void) { - if (arch_intr_init_quirk) { - if (arch_intr_init_quirk()) + if (x86_quirks->arch_intr_init) { + if (x86_quirks->arch_intr_init()) return; } #ifdef CONFIG_X86_LOCAL_APIC @@ -97,8 +89,8 @@ void __init pre_setup_arch_hook(void) **/ void __init trap_init_hook(void) { - if (arch_trap_init_quirk) { - if (arch_trap_init_quirk()) + if (x86_quirks->arch_trap_init) { + if (x86_quirks->arch_trap_init()) return; } } @@ -119,13 +111,13 @@ static struct irqaction irq0 = { **/ void __init time_init_hook(void) { - if (arch_time_init_quirk) { + if (x86_quirks->arch_time_init) { /* * A nonzero return code does not mean failure, it means * that the architecture quirk does not want any * generic (timer) setup to be performed after this: */ - if (arch_time_init_quirk()) + if (x86_quirks->arch_time_init()) return; } -- cgit v1.2.3 From 64898a8bad8c94ad7a4bd5cc86b66edfbb081f4a Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Sat, 19 Jul 2008 18:01:16 -0700 Subject: x86: extend and use x86_quirks to clean up NUMAQ code add these new x86_quirks methods: int *mpc_record; int (*mpc_apic_id)(struct mpc_config_processor *m); void (*mpc_oem_bus_info)(struct mpc_config_bus *m, char *name); void (*mpc_oem_pci_bus)(struct mpc_config_bus *m); void (*smp_read_mpc_oem)(struct mp_config_oemtable *oemtable, unsigned short oemsize); ... and move NUMAQ related mps table handling to numaq_32.c. also move the call to smp_read_mpc_oem() to smp_read_mpc() directly. Should not change functionality, albeit it would be nice to get it tested on real NUMAQ as well ... Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/mpparse.c | 191 +++++---------------------------------------- arch/x86/kernel/numaq_32.c | 190 +++++++++++++++++++++++++++++++++++++++++--- 2 files changed, 200 insertions(+), 181 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/mpparse.c b/arch/x86/kernel/mpparse.c index 3cbd2df3abe..6ae005ccaed 100644 --- a/arch/x86/kernel/mpparse.c +++ b/arch/x86/kernel/mpparse.c @@ -49,76 +49,6 @@ static int __init mpf_checksum(unsigned char *mp, int len) return sum & 0xFF; } -#ifdef CONFIG_X86_NUMAQ -int found_numaq; -/* - * Have to match translation table entries to main table entries by counter - * hence the mpc_record variable .... can't see a less disgusting way of - * doing this .... - */ -struct mpc_config_translation { - unsigned char mpc_type; - unsigned char trans_len; - unsigned char trans_type; - unsigned char trans_quad; - unsigned char trans_global; - unsigned char trans_local; - unsigned short trans_reserved; -}; - - -static int mpc_record; -static struct mpc_config_translation *translation_table[MAX_MPC_ENTRY] - __cpuinitdata; - -static inline int generate_logical_apicid(int quad, int phys_apicid) -{ - return (quad << 4) + (phys_apicid ? phys_apicid << 1 : 1); -} - - -static inline int mpc_apic_id(struct mpc_config_processor *m, - struct mpc_config_translation *translation_record) -{ - int quad = translation_record->trans_quad; - int logical_apicid = generate_logical_apicid(quad, m->mpc_apicid); - - printk(KERN_DEBUG "Processor #%d %u:%u APIC version %d (quad %d, apic %d)\n", - m->mpc_apicid, - (m->mpc_cpufeature & CPU_FAMILY_MASK) >> 8, - (m->mpc_cpufeature & CPU_MODEL_MASK) >> 4, - m->mpc_apicver, quad, logical_apicid); - return logical_apicid; -} - -int mp_bus_id_to_node[MAX_MP_BUSSES]; - -int mp_bus_id_to_local[MAX_MP_BUSSES]; - -static void mpc_oem_bus_info(struct mpc_config_bus *m, char *name, - struct mpc_config_translation *translation) -{ - int quad = translation->trans_quad; - int local = translation->trans_local; - - mp_bus_id_to_node[m->mpc_busid] = quad; - mp_bus_id_to_local[m->mpc_busid] = local; - printk(KERN_INFO "Bus #%d is %s (node %d)\n", - m->mpc_busid, name, quad); -} - -int quad_local_to_mp_bus_id [NR_CPUS/4][4]; -static void mpc_oem_pci_bus(struct mpc_config_bus *m, - struct mpc_config_translation *translation) -{ - int quad = translation->trans_quad; - int local = translation->trans_local; - - quad_local_to_mp_bus_id[quad][local] = m->mpc_busid; -} - -#endif - static void __cpuinit MP_processor_info(struct mpc_config_processor *m) { int apicid; @@ -128,14 +58,12 @@ static void __cpuinit MP_processor_info(struct mpc_config_processor *m) disabled_cpus++; return; } -#ifdef CONFIG_X86_NUMAQ - if (found_numaq) - apicid = mpc_apic_id(m, translation_table[mpc_record]); + + if (x86_quirks->mpc_apic_id) + apicid = x86_quirks->mpc_apic_id(m); else apicid = m->mpc_apicid; -#else - apicid = m->mpc_apicid; -#endif + if (m->mpc_cpuflag & CPU_BOOTPROCESSOR) { bootup_cpu = " (Bootup-CPU)"; boot_cpu_physical_apicid = m->mpc_apicid; @@ -152,12 +80,10 @@ static void __init MP_bus_info(struct mpc_config_bus *m) memcpy(str, m->mpc_bustype, 6); str[6] = 0; -#ifdef CONFIG_X86_NUMAQ - if (found_numaq) - mpc_oem_bus_info(m, str, translation_table[mpc_record]); -#else - printk(KERN_INFO "Bus #%d is %s\n", m->mpc_busid, str); -#endif + if (x86_quirks->mpc_oem_bus_info) + x86_quirks->mpc_oem_bus_info(m, str); + else + printk(KERN_INFO "Bus #%d is %s\n", m->mpc_busid, str); #if MAX_MP_BUSSES < 256 if (m->mpc_busid >= MAX_MP_BUSSES) { @@ -174,10 +100,9 @@ static void __init MP_bus_info(struct mpc_config_bus *m) mp_bus_id_to_type[m->mpc_busid] = MP_BUS_ISA; #endif } else if (strncmp(str, BUSTYPE_PCI, sizeof(BUSTYPE_PCI) - 1) == 0) { -#ifdef CONFIG_X86_NUMAQ - if (found_numaq) - mpc_oem_pci_bus(m, translation_table[mpc_record]); -#endif + if (x86_quirks->mpc_oem_pci_bus) + x86_quirks->mpc_oem_pci_bus(m); + clear_bit(m->mpc_busid, mp_bus_not_pci); #if defined(CONFIG_EISA) || defined (CONFIG_MCA) mp_bus_id_to_type[m->mpc_busid] = MP_BUS_PCI; @@ -317,83 +242,6 @@ static void __init MP_lintsrc_info(struct mpc_config_lintsrc *m) m->mpc_srcbusirq, m->mpc_destapic, m->mpc_destapiclint); } -#ifdef CONFIG_X86_NUMAQ -static void __init MP_translation_info(struct mpc_config_translation *m) -{ - printk(KERN_INFO - "Translation: record %d, type %d, quad %d, global %d, local %d\n", - mpc_record, m->trans_type, m->trans_quad, m->trans_global, - m->trans_local); - - if (mpc_record >= MAX_MPC_ENTRY) - printk(KERN_ERR "MAX_MPC_ENTRY exceeded!\n"); - else - translation_table[mpc_record] = m; /* stash this for later */ - if (m->trans_quad < MAX_NUMNODES && !node_online(m->trans_quad)) - node_set_online(m->trans_quad); -} - -/* - * Read/parse the MPC oem tables - */ - -static void __init smp_read_mpc_oem(struct mp_config_oemtable *oemtable, - unsigned short oemsize) -{ - int count = sizeof(*oemtable); /* the header size */ - unsigned char *oemptr = ((unsigned char *)oemtable) + count; - - mpc_record = 0; - printk(KERN_INFO "Found an OEM MPC table at %8p - parsing it ... \n", - oemtable); - if (memcmp(oemtable->oem_signature, MPC_OEM_SIGNATURE, 4)) { - printk(KERN_WARNING - "SMP mpc oemtable: bad signature [%c%c%c%c]!\n", - oemtable->oem_signature[0], oemtable->oem_signature[1], - oemtable->oem_signature[2], oemtable->oem_signature[3]); - return; - } - if (mpf_checksum((unsigned char *)oemtable, oemtable->oem_length)) { - printk(KERN_WARNING "SMP oem mptable: checksum error!\n"); - return; - } - while (count < oemtable->oem_length) { - switch (*oemptr) { - case MP_TRANSLATION: - { - struct mpc_config_translation *m = - (struct mpc_config_translation *)oemptr; - MP_translation_info(m); - oemptr += sizeof(*m); - count += sizeof(*m); - ++mpc_record; - break; - } - default: - { - printk(KERN_WARNING - "Unrecognised OEM table entry type! - %d\n", - (int)*oemptr); - return; - } - } - } -} - -void numaq_mps_oem_check(struct mp_config_table *mpc, char *oem, - char *productid) -{ - if (strncmp(oem, "IBM NUMA", 8)) - printk("Warning! Not a NUMA-Q system!\n"); - else - found_numaq = 1; - - if (mpc->mpc_oemptr) - smp_read_mpc_oem((struct mp_config_oemtable *)mpc->mpc_oemptr, - mpc->mpc_oemsize); -} -#endif /* CONFIG_X86_NUMAQ */ - /* * Read/parse the MPC */ @@ -458,7 +306,6 @@ static int __init smp_read_mpc(struct mp_config_table *mpc, unsigned early) } else mps_oem_check(mpc, oem, str); #endif - /* save the local APIC address, it might be non-default */ if (!acpi_lapic) mp_lapic_addr = mpc->mpc_lapic; @@ -466,12 +313,17 @@ static int __init smp_read_mpc(struct mp_config_table *mpc, unsigned early) if (early) return 1; + if (mpc->mpc_oemptr && x86_quirks->smp_read_mpc_oem) { + struct mp_config_oemtable *oem_table = (struct mp_config_oemtable *)(unsigned long)mpc->mpc_oemptr; + x86_quirks->smp_read_mpc_oem(oem_table, mpc->mpc_oemsize); + } + /* * Now process the configuration blocks. */ -#ifdef CONFIG_X86_NUMAQ - mpc_record = 0; -#endif + if (x86_quirks->mpc_record) + *x86_quirks->mpc_record = 0; + while (count < mpc->mpc_length) { switch (*mpt) { case MP_PROCESSOR: @@ -537,9 +389,8 @@ static int __init smp_read_mpc(struct mp_config_table *mpc, unsigned early) count = mpc->mpc_length; break; } -#ifdef CONFIG_X86_NUMAQ - ++mpc_record; -#endif + if (x86_quirks->mpc_record) + (*x86_quirks->mpc_record)++; } #ifdef CONFIG_X86_GENERICARCH diff --git a/arch/x86/kernel/numaq_32.c b/arch/x86/kernel/numaq_32.c index a23e8233b9a..7f4e00d1d89 100644 --- a/arch/x86/kernel/numaq_32.c +++ b/arch/x86/kernel/numaq_32.c @@ -33,6 +33,7 @@ #include #include #include +#include #define MB_TO_PAGES(addr) ((addr) << (20 - PAGE_SHIFT)) @@ -71,6 +72,181 @@ static void __init smp_dump_qct(void) } } + +void __init numaq_tsc_disable(void) +{ + if (!found_numaq) + return; + + if (num_online_nodes() > 1) { + printk(KERN_DEBUG "NUMAQ: disabling TSC\n"); + setup_clear_cpu_cap(X86_FEATURE_TSC); + } +} + +int found_numaq; +/* + * Have to match translation table entries to main table entries by counter + * hence the mpc_record variable .... can't see a less disgusting way of + * doing this .... + */ +struct mpc_config_translation { + unsigned char mpc_type; + unsigned char trans_len; + unsigned char trans_type; + unsigned char trans_quad; + unsigned char trans_global; + unsigned char trans_local; + unsigned short trans_reserved; +}; + +/* x86_quirks member */ +static int mpc_record; +static struct mpc_config_translation *translation_table[MAX_MPC_ENTRY] + __cpuinitdata; + +static inline int generate_logical_apicid(int quad, int phys_apicid) +{ + return (quad << 4) + (phys_apicid ? phys_apicid << 1 : 1); +} + +/* x86_quirks member */ +static int mpc_apic_id(struct mpc_config_processor *m) +{ + int quad = translation_table[mpc_record]->trans_quad; + int logical_apicid = generate_logical_apicid(quad, m->mpc_apicid); + + printk(KERN_DEBUG "Processor #%d %u:%u APIC version %d (quad %d, apic %d)\n", + m->mpc_apicid, + (m->mpc_cpufeature & CPU_FAMILY_MASK) >> 8, + (m->mpc_cpufeature & CPU_MODEL_MASK) >> 4, + m->mpc_apicver, quad, logical_apicid); + return logical_apicid; +} + +int mp_bus_id_to_node[MAX_MP_BUSSES]; + +int mp_bus_id_to_local[MAX_MP_BUSSES]; + +/* x86_quirks member */ +static void mpc_oem_bus_info(struct mpc_config_bus *m, char *name) +{ + int quad = translation_table[mpc_record]->trans_quad; + int local = translation_table[mpc_record]->trans_local; + + mp_bus_id_to_node[m->mpc_busid] = quad; + mp_bus_id_to_local[m->mpc_busid] = local; + printk(KERN_INFO "Bus #%d is %s (node %d)\n", + m->mpc_busid, name, quad); +} + +int quad_local_to_mp_bus_id [NR_CPUS/4][4]; + +/* x86_quirks member */ +static void mpc_oem_pci_bus(struct mpc_config_bus *m) +{ + int quad = translation_table[mpc_record]->trans_quad; + int local = translation_table[mpc_record]->trans_local; + + quad_local_to_mp_bus_id[quad][local] = m->mpc_busid; +} + +static void __init MP_translation_info(struct mpc_config_translation *m) +{ + printk(KERN_INFO + "Translation: record %d, type %d, quad %d, global %d, local %d\n", + mpc_record, m->trans_type, m->trans_quad, m->trans_global, + m->trans_local); + + if (mpc_record >= MAX_MPC_ENTRY) + printk(KERN_ERR "MAX_MPC_ENTRY exceeded!\n"); + else + translation_table[mpc_record] = m; /* stash this for later */ + if (m->trans_quad < MAX_NUMNODES && !node_online(m->trans_quad)) + node_set_online(m->trans_quad); +} + +static int __init mpf_checksum(unsigned char *mp, int len) +{ + int sum = 0; + + while (len--) + sum += *mp++; + + return sum & 0xFF; +} + +/* + * Read/parse the MPC oem tables + */ + +static void __init smp_read_mpc_oem(struct mp_config_oemtable *oemtable, + unsigned short oemsize) +{ + int count = sizeof(*oemtable); /* the header size */ + unsigned char *oemptr = ((unsigned char *)oemtable) + count; + + mpc_record = 0; + printk(KERN_INFO "Found an OEM MPC table at %8p - parsing it ... \n", + oemtable); + if (memcmp(oemtable->oem_signature, MPC_OEM_SIGNATURE, 4)) { + printk(KERN_WARNING + "SMP mpc oemtable: bad signature [%c%c%c%c]!\n", + oemtable->oem_signature[0], oemtable->oem_signature[1], + oemtable->oem_signature[2], oemtable->oem_signature[3]); + return; + } + if (mpf_checksum((unsigned char *)oemtable, oemtable->oem_length)) { + printk(KERN_WARNING "SMP oem mptable: checksum error!\n"); + return; + } + while (count < oemtable->oem_length) { + switch (*oemptr) { + case MP_TRANSLATION: + { + struct mpc_config_translation *m = + (struct mpc_config_translation *)oemptr; + MP_translation_info(m); + oemptr += sizeof(*m); + count += sizeof(*m); + ++mpc_record; + break; + } + default: + { + printk(KERN_WARNING + "Unrecognised OEM table entry type! - %d\n", + (int)*oemptr); + return; + } + } + } +} + +static struct x86_quirks numaq_x86_quirks __initdata = { + .arch_time_init = NULL, + .arch_pre_intr_init = NULL, + .arch_memory_setup = NULL, + .arch_intr_init = NULL, + .arch_trap_init = NULL, + .mach_get_smp_config = NULL, + .mach_find_smp_config = NULL, + .mpc_record = &mpc_record, + .mpc_apic_id = mpc_apic_id, + .mpc_oem_bus_info = mpc_oem_bus_info, + .mpc_oem_pci_bus = mpc_oem_pci_bus, + .smp_read_mpc_oem = smp_read_mpc_oem, +}; + +void numaq_mps_oem_check(struct mp_config_table *mpc, char *oem, + char *productid) +{ + if (strncmp(oem, "IBM NUMA", 8)) + printk("Warning! Not a NUMA-Q system!\n"); + else + found_numaq = 1; +} + static __init void early_check_numaq(void) { /* @@ -82,6 +258,9 @@ static __init void early_check_numaq(void) */ if (smp_found_config) early_get_smp_config(); + + if (found_numaq) + x86_quirks = &numaq_x86_quirks; } int __init get_memcfg_numaq(void) @@ -92,14 +271,3 @@ int __init get_memcfg_numaq(void) smp_dump_qct(); return 1; } - -void __init numaq_tsc_disable(void) -{ - if (!found_numaq) - return; - - if (num_online_nodes() > 1) { - printk(KERN_DEBUG "NUMAQ: disabling TSC\n"); - setup_clear_cpu_cap(X86_FEATURE_TSC); - } -} -- cgit v1.2.3 From 63b5d7af2556a7de6bf72c5dd0b85a32fb4c3767 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Sat, 19 Jul 2008 18:02:26 -0700 Subject: x86: add ->pre_time_init to x86_quirks so NUMAQ can use that to call numaq_pre_time_init() This allows us to remove a NUMAQ special from arch/x86/kernel/setup.c. (and paves the way to remove the NUMAQ subarch) Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/numaq_32.c | 7 +++++++ arch/x86/kernel/setup.c | 8 -------- arch/x86/kernel/time_32.c | 1 + arch/x86/mach-default/setup.c | 10 ++++++++++ 4 files changed, 18 insertions(+), 8 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/numaq_32.c b/arch/x86/kernel/numaq_32.c index 7f4e00d1d89..b8c45610b20 100644 --- a/arch/x86/kernel/numaq_32.c +++ b/arch/x86/kernel/numaq_32.c @@ -84,6 +84,12 @@ void __init numaq_tsc_disable(void) } } +static int __init numaq_pre_time_init(void) +{ + numaq_tsc_disable(); + return 0; +} + int found_numaq; /* * Have to match translation table entries to main table entries by counter @@ -224,6 +230,7 @@ static void __init smp_read_mpc_oem(struct mp_config_oemtable *oemtable, } static struct x86_quirks numaq_x86_quirks __initdata = { + .arch_pre_time_init = numaq_pre_time_init, .arch_time_init = NULL, .arch_pre_intr_init = NULL, .arch_memory_setup = NULL, diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c index bbcc13d0b56..4064616cfa8 100644 --- a/arch/x86/kernel/setup.c +++ b/arch/x86/kernel/setup.c @@ -853,14 +853,6 @@ void __init setup_arch(char **cmdline_p) init_cpu_to_node(); #endif -#ifdef CONFIG_X86_NUMAQ - /* - * need to check online nodes num, call it - * here before time_init/tsc_init - */ - numaq_tsc_disable(); -#endif - init_apic_mappings(); ioapic_init_mappings(); diff --git a/arch/x86/kernel/time_32.c b/arch/x86/kernel/time_32.c index 059ca6ee59b..ffe3c664afc 100644 --- a/arch/x86/kernel/time_32.c +++ b/arch/x86/kernel/time_32.c @@ -129,6 +129,7 @@ void __init hpet_time_init(void) */ void __init time_init(void) { + pre_time_init_hook(); tsc_init(); late_time_init = choose_time_init(); } diff --git a/arch/x86/mach-default/setup.c b/arch/x86/mach-default/setup.c index 631dbed9fb9..3d317836be9 100644 --- a/arch/x86/mach-default/setup.c +++ b/arch/x86/mach-default/setup.c @@ -102,6 +102,16 @@ static struct irqaction irq0 = { .name = "timer" }; +/** + * pre_time_init_hook - do any specific initialisations before. + * + **/ +void __init pre_time_init_hook(void) +{ + if (x86_quirks->arch_pre_time_init) + x86_quirks->arch_pre_time_init(); +} + /** * time_init_hook - do any specific initialisations for the system timer. * -- cgit v1.2.3 From e3a61b0a8c0e342e700a61cd554b01050f333a36 Mon Sep 17 00:00:00 2001 From: Simon Arlott Date: Sat, 19 Jul 2008 23:32:54 +0100 Subject: x86: add unknown_nmi_panic kernel parameter It's not possible to enable the unknown_nmi_panic sysctl option until init is run. It's useful to be able to panic the kernel during boot too, this adds a parameter to enable this option. Signed-off-by: Simon Arlott Signed-off-by: Ingo Molnar --- arch/x86/kernel/nmi.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index ec024b3baad..e0b44b7b717 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -448,6 +448,13 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) #ifdef CONFIG_SYSCTL +static int __init setup_unknown_nmi_panic(char *str) +{ + unknown_nmi_panic = 1; + return 1; +} +__setup("unknown_nmi_panic", setup_unknown_nmi_panic); + static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu) { unsigned char reason = get_nmi_reason(); -- cgit v1.2.3 From 9175fc06aee79c349790672178d3fd7507d75c86 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 21 Jul 2008 01:38:14 -0700 Subject: x86: use setup_clear_cpu_cap() when disabling the lapic ... so don't need to call clear_cpu_cap again in early_identify_cpu, and could use cleared_cpu_caps like other places. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/apic_32.c | 5 +---- arch/x86/kernel/apic_64.c | 2 +- arch/x86/kernel/cpu/common_64.c | 4 ---- 3 files changed, 2 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/apic_32.c b/arch/x86/kernel/apic_32.c index a437d027f20..e9a00e5074b 100644 --- a/arch/x86/kernel/apic_32.c +++ b/arch/x86/kernel/apic_32.c @@ -1214,9 +1214,6 @@ int apic_version[MAX_APICS]; int __init APIC_init_uniprocessor(void) { - if (disable_apic) - clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC); - if (!smp_found_config && !cpu_has_apic) return -1; @@ -1700,7 +1697,7 @@ early_param("lapic", parse_lapic); static int __init parse_nolapic(char *arg) { disable_apic = 1; - clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC); + setup_clear_cpu_cap(X86_FEATURE_APIC); return 0; } early_param("nolapic", parse_nolapic); diff --git a/arch/x86/kernel/apic_64.c b/arch/x86/kernel/apic_64.c index 1e3d32e27c1..16e586cacbd 100644 --- a/arch/x86/kernel/apic_64.c +++ b/arch/x86/kernel/apic_64.c @@ -1337,7 +1337,7 @@ early_param("apic", apic_set_verbosity); static __init int setup_disableapic(char *str) { disable_apic = 1; - clear_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC); + setup_clear_cpu_cap(X86_FEATURE_APIC); return 0; } early_param("disableapic", setup_disableapic); diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index 7b8cc72feb4..0485cf64452 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -324,10 +324,6 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c) cpu_devs[c->x86_vendor]->c_early_init(c); validate_pat_support(c); - - /* early_param could clear that, but recall get it set again */ - if (disable_apic) - clear_cpu_cap(c, X86_FEATURE_APIC); } /* -- cgit v1.2.3 From 7edf8891ad7aef5f4e97991fed6fb0e605e96ea0 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Mon, 21 Jul 2008 01:39:03 -0700 Subject: x86: remove extra calling to get ext cpuid level Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/common_64.c | 1 - 1 file changed, 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/common_64.c b/arch/x86/kernel/cpu/common_64.c index 0485cf64452..daee611f014 100644 --- a/arch/x86/kernel/cpu/common_64.c +++ b/arch/x86/kernel/cpu/common_64.c @@ -305,7 +305,6 @@ static void __cpuinit early_identify_cpu(struct cpuinfo_x86 *c) c->x86_capability[2] = cpuid_edx(0x80860001); } - c->extended_cpuid_level = cpuid_eax(0x80000000); if (c->extended_cpuid_level >= 0x80000007) c->x86_power = cpuid_edx(0x80000007); -- cgit v1.2.3