Merge branch 'mesa_7_5_branch'

13 files changed, 362 insertions, 160 deletions

diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c
index 9747a55cbf..a0c8ed88f7 100644
--- a/src/gallium/auxiliary/util/u_tile.c
+++ b/src/gallium/auxiliary/util/u_tile.c
@@ -1202,6 +1202,19 @@ pipe_put_tile_z(struct pipe_transfer *pt,
       }
       break;
    case PIPE_FORMAT_S8Z24_UNORM:
+      {
+         uint *pDest = (uint *) (map + y * pt->stride + x*4);
+         assert(pt->usage == PIPE_TRANSFER_READ_WRITE);
+         for (i = 0; i < h; i++) {
+            for (j = 0; j < w; j++) {
+               /* convert 32-bit Z to 24-bit Z, preserve stencil */
+               pDest[j] = (pDest[j] & 0xff000000) | ptrc[j] >> 8;
+            }
+            pDest += pt->stride/4;
+            ptrc += srcStride;
+         }
+      }
+      break;
    case PIPE_FORMAT_X8Z24_UNORM:
       {
          uint *pDest = (uint *) (map + y * pt->stride + x*4);
@@ -1216,13 +1229,26 @@ pipe_put_tile_z(struct pipe_transfer *pt,
       }
       break;
    case PIPE_FORMAT_Z24S8_UNORM:
+      {
+         uint *pDest = (uint *) (map + y * pt->stride + x*4);
+         assert(pt->usage == PIPE_TRANSFER_READ_WRITE);
+         for (i = 0; i < h; i++) {
+            for (j = 0; j < w; j++) {
+               /* convert 32-bit Z to 24-bit Z, preserve stencil */
+               pDest[j] = (pDest[j] & 0xff) | (ptrc[j] & 0xffffff00);
+            }
+            pDest += pt->stride/4;
+            ptrc += srcStride;
+         }
+      }
+      break;
    case PIPE_FORMAT_Z24X8_UNORM:
       {
          uint *pDest = (uint *) (map + y * pt->stride + x*4);
          for (i = 0; i < h; i++) {
             for (j = 0; j < w; j++) {
                /* convert 32-bit Z to 24-bit Z (0 stencil) */
-               pDest[j] = ptrc[j] << 8;
+               pDest[j] = ptrc[j] & 0xffffff00;
             }
             pDest += pt->stride/4;
             ptrc += srcStride;
diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.c b/src/gallium/state_trackers/wgl/shared/stw_context.c
index 9df1ab7652..4968ecc692 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_context.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_context.c
@@ -60,8 +60,14 @@ stw_context(GLcontext *glctx)
 static INLINE struct stw_context *
 stw_current_context(void)
 {
-   GET_CURRENT_CONTEXT( glctx );
-   return stw_context(glctx);
+   /* We must check if multiple threads are being used or GET_CURRENT_CONTEXT 
+    * might return the current context of the thread first seen. */
+   _glapi_check_multithread();
+
+   {
+      GET_CURRENT_CONTEXT( glctx );
+      return stw_context(glctx);
+   }
 }
 
 BOOL
@@ -74,7 +80,7 @@ stw_copy_context(
    struct stw_context *dst;
    BOOL ret = FALSE;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->ctx_mutex );
    
    src = stw_lookup_context_locked( hglrcSrc );
    dst = stw_lookup_context_locked( hglrcDst );
@@ -87,7 +93,7 @@ stw_copy_context(
       (void) mask;
    }
 
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->ctx_mutex );
    
    return ret;
 }
@@ -101,7 +107,7 @@ stw_share_lists(
    struct stw_context *ctx2;
    BOOL ret = FALSE;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->ctx_mutex );
    
    ctx1 = stw_lookup_context_locked( hglrc1 );
    ctx2 = stw_lookup_context_locked( hglrc2 );
@@ -111,7 +117,7 @@ stw_share_lists(
       ret = _mesa_share_state(ctx2->st->ctx, ctx1->st->ctx);
    }
 
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->ctx_mutex );
    
    return ret;
 }
@@ -124,8 +130,10 @@ stw_viewport(GLcontext * glctx, GLint x, GLint y,
    struct stw_framebuffer *fb;
    
    fb = stw_framebuffer_from_hdc( ctx->hdc );
-   if(fb)
+   if(fb) {
       stw_framebuffer_update(fb);
+      stw_framebuffer_release(fb);
+   }
 }
 
 UINT_PTR
@@ -189,9 +197,9 @@ stw_create_layer_context(
    ctx->st->ctx->DriverCtx = ctx;
    ctx->st->ctx->Driver.Viewport = stw_viewport;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->ctx_mutex );
    ctx->hglrc = handle_table_add(stw_dev->ctx_table, ctx);
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->ctx_mutex );
    if (!ctx->hglrc)
       goto no_hglrc;
 
@@ -218,10 +226,10 @@ stw_delete_context(
    if (!stw_dev)
       return FALSE;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->ctx_mutex );
    ctx = stw_lookup_context_locked(hglrc);
    handle_table_remove(stw_dev->ctx_table, hglrc);
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->ctx_mutex );
 
    if (ctx) {
       struct stw_context *curctx = stw_current_context();
@@ -248,9 +256,9 @@ stw_release_context(
    if (!stw_dev)
       return FALSE;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->ctx_mutex );
    ctx = stw_lookup_context_locked( hglrc );
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->ctx_mutex );
 
    if (!ctx)
       return FALSE;
@@ -298,9 +306,9 @@ stw_make_current(
    HDC hdc,
    UINT_PTR hglrc )
 {
-   struct stw_context *curctx;
-   struct stw_context *ctx;
-   struct stw_framebuffer *fb;
+   struct stw_context *curctx = NULL;
+   struct stw_context *ctx = NULL;
+   struct stw_framebuffer *fb = NULL;
 
    if (!stw_dev)
       goto fail;
@@ -322,13 +330,13 @@ stw_make_current(
       return st_make_current( NULL, NULL, NULL );
    }
 
-   pipe_mutex_lock( stw_dev->mutex ); 
-
+   pipe_mutex_lock( stw_dev->ctx_mutex ); 
    ctx = stw_lookup_context_locked( hglrc );
+   pipe_mutex_unlock( stw_dev->ctx_mutex ); 
    if(!ctx)
       goto fail;
 
-   fb = stw_framebuffer_from_hdc_locked( hdc );
+   fb = stw_framebuffer_from_hdc( hdc );
    if(!fb) { 
       /* Applications should call SetPixelFormat before creating a context,
        * but not all do, and the opengl32 runtime seems to use a default pixel
@@ -336,13 +344,11 @@ stw_make_current(
        */
       int iPixelFormat = GetPixelFormat(hdc);
       if(iPixelFormat)
-         fb = stw_framebuffer_create_locked( hdc, iPixelFormat );
+         fb = stw_framebuffer_create( hdc, iPixelFormat );
       if(!fb) 
          goto fail;
    }
    
-   pipe_mutex_unlock( stw_dev->mutex );
-
    if(fb->iPixelFormat != ctx->iPixelFormat)
       goto fail;
 
@@ -361,12 +367,16 @@ stw_make_current(
 
 success:
    assert(fb);
-   if(fb)
+   if(fb) {
       stw_framebuffer_update(fb);
+      stw_framebuffer_release(fb);
+   }
    
    return TRUE;
 
 fail:
+   if(fb)
+      stw_framebuffer_release(fb);
    st_make_current( NULL, NULL, NULL );
    return FALSE;
 }
diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.c b/src/gallium/state_trackers/wgl/shared/stw_device.c
index ce46624146..0b6954915a 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_device.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_device.c
@@ -69,8 +69,6 @@ stw_flush_frontbuffer(struct pipe_screen *screen,
    fb = stw_framebuffer_from_hdc( hdc );
    /* fb can be NULL if window was destroyed already */
    if (fb) {
-      pipe_mutex_lock( fb->mutex );
-
 #if DEBUG
       {
          struct pipe_surface *surface2;
@@ -94,8 +92,7 @@ stw_flush_frontbuffer(struct pipe_screen *screen,
    
    if(fb) {
       stw_framebuffer_update(fb);
-      
-      pipe_mutex_unlock( fb->mutex );
+      stw_framebuffer_release(fb);
    }
 }
 
@@ -138,7 +135,8 @@ stw_init(const struct stw_winsys *stw_winsys)
    
    stw_dev->screen->flush_frontbuffer = &stw_flush_frontbuffer;
    
-   pipe_mutex_init( stw_dev->mutex );
+   pipe_mutex_init( stw_dev->ctx_mutex );
+   pipe_mutex_init( stw_dev->fb_mutex );
 
    stw_dev->ctx_table = handle_table_create();
    if (!stw_dev->ctx_table) {
@@ -179,7 +177,7 @@ stw_cleanup(void)
    if (!stw_dev)
       return;
    
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->ctx_mutex );
    {
       /* Ensure all contexts are destroyed */
       i = handle_table_get_first_handle(stw_dev->ctx_table);
@@ -189,11 +187,12 @@ stw_cleanup(void)
       }
       handle_table_destroy(stw_dev->ctx_table);
    }
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->ctx_mutex );
 
    stw_framebuffer_cleanup();
    
-   pipe_mutex_destroy( stw_dev->mutex );
+   pipe_mutex_destroy( stw_dev->fb_mutex );
+   pipe_mutex_destroy( stw_dev->ctx_mutex );
    
    stw_dev->screen->destroy(stw_dev->screen);
 
diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.h b/src/gallium/state_trackers/wgl/shared/stw_device.h
index e097f1f71e..e1bb9518dd 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_device.h
+++ b/src/gallium/state_trackers/wgl/shared/stw_device.h
@@ -57,10 +57,10 @@ struct stw_device
    unsigned pixelformat_count;
    unsigned pixelformat_extended_count;
 
-   pipe_mutex mutex;
-
+   pipe_mutex ctx_mutex;
    struct handle_table *ctx_table;
    
+   pipe_mutex fb_mutex;
    struct stw_framebuffer *fb_head;
    
 #ifdef DEBUG
diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
index 78a2dbc4d7..b8956bb550 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
+++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
@@ -45,6 +45,10 @@
 #include "stw_tls.h"
 
 
+/**
+ * Search the framebuffer with the matching HWND while holding the
+ * stw_dev::fb_mutex global lock.
+ */
 static INLINE struct stw_framebuffer *
 stw_framebuffer_from_hwnd_locked(
    HWND hwnd )
@@ -52,13 +56,20 @@ stw_framebuffer_from_hwnd_locked(
    struct stw_framebuffer *fb;
 
    for (fb = stw_dev->fb_head; fb != NULL; fb = fb->next)
-      if (fb->hWnd == hwnd)
+      if (fb->hWnd == hwnd) {
+         pipe_mutex_lock(fb->mutex);
          break;
+      }
 
    return fb;
 }
 
 
+/**
+ * Destroy this framebuffer. Both stw_dev::fb_mutex and stw_framebuffer::mutex
+ * must be held, by this order. Obviously no further access to fb can be done 
+ * after this.
+ */
 static INLINE void
 stw_framebuffer_destroy_locked(
    struct stw_framebuffer *fb )
@@ -74,12 +85,48 @@ stw_framebuffer_destroy_locked(
 
    st_unreference_framebuffer(fb->stfb);
    
+   pipe_mutex_unlock( fb->mutex );
+
    pipe_mutex_destroy( fb->mutex );
    
    FREE( fb );
 }
 
 
+void
+stw_framebuffer_release(
+   struct stw_framebuffer *fb)
+{
+   assert(fb);
+   pipe_mutex_unlock( fb->mutex );
+}
+
+
+static INLINE void
+stw_framebuffer_get_size( struct stw_framebuffer *fb )
+{
+   unsigned width, height;
+   RECT rect;
+
+   assert(fb->hWnd);
+   
+   GetClientRect( fb->hWnd, &rect );
+   width = rect.right - rect.left;
+   height = rect.bottom - rect.top;
+
+   if(width < 1)
+      width = 1;
+   if(height < 1)
+      height = 1;
+
+   if(width != fb->width || height != fb->height) {
+      fb->must_resize = TRUE;
+      fb->width = width; 
+      fb->height = height; 
+   }
+}
+
+
 /**
  * @sa http://msdn.microsoft.com/en-us/library/ms644975(VS.85).aspx
  * @sa http://msdn.microsoft.com/en-us/library/ms644960(VS.85).aspx
@@ -92,6 +139,7 @@ stw_call_window_proc(
 {
    struct stw_tls_data *tls_data;
    PCWPSTRUCT pParams = (PCWPSTRUCT)lParam;
+   struct stw_framebuffer *fb;
    
    tls_data = stw_tls_get_data();
    if(!tls_data)
@@ -100,71 +148,37 @@ stw_call_window_proc(
    if (nCode < 0)
        return CallNextHookEx(tls_data->hCallWndProcHook, nCode, wParam, lParam);
 
-   if (pParams->message == WM_SIZE && pParams->wParam != SIZE_MINIMIZED) {
-      struct stw_framebuffer *fb;
-
-      pipe_mutex_lock( stw_dev->mutex );
-      fb = stw_framebuffer_from_hwnd_locked( pParams->hwnd );
-      pipe_mutex_unlock( stw_dev->mutex );
-      
-      if(fb) {
-         unsigned width = LOWORD( pParams->lParam );
-         unsigned height = HIWORD( pParams->lParam );
-         
-         pipe_mutex_lock( fb->mutex );
-         fb->must_resize = TRUE;
-         fb->width = width;
-         fb->height = height;
-         pipe_mutex_unlock( fb->mutex );
+   if (pParams->message == WM_WINDOWPOSCHANGED) {
+      /* We handle WM_WINDOWPOSCHANGED instead of WM_SIZE because according to
+       * http://blogs.msdn.com/oldnewthing/archive/2008/01/15/7113860.aspx 
+       * WM_SIZE is generated from WM_WINDOWPOSCHANGED by DefWindowProc so it 
+       * can be masked out by the application. */
+      LPWINDOWPOS lpWindowPos = (LPWINDOWPOS)pParams->lParam;
+      if((lpWindowPos->flags & SWP_SHOWWINDOW) || 
+         !(lpWindowPos->flags & SWP_NOSIZE)) {
+         fb = stw_framebuffer_from_hwnd( pParams->hwnd );
+         if(fb) {
+            /* Size in WINDOWPOS includes the window frame, so get the size 
+             * of the client area via GetClientRect.  */
+            stw_framebuffer_get_size(fb);
+            stw_framebuffer_release(fb);
+         }
       }
    }
-
-   if (pParams->message == WM_DESTROY) {
-      struct stw_framebuffer *fb;
-
-      pipe_mutex_lock( stw_dev->mutex );
-      
+   else if (pParams->message == WM_DESTROY) {
+      pipe_mutex_lock( stw_dev->fb_mutex );
       fb = stw_framebuffer_from_hwnd_locked( pParams->hwnd );
       if(fb)
          stw_framebuffer_destroy_locked(fb);
-      
-      pipe_mutex_unlock( stw_dev->mutex );
+      pipe_mutex_unlock( stw_dev->fb_mutex );
    }
 
    return CallNextHookEx(tls_data->hCallWndProcHook, nCode, wParam, lParam);
 }
 
 
-static void
-stw_framebuffer_get_size( struct stw_framebuffer *fb )
-{
-   unsigned width, height;
-   RECT rect;
-
-   assert(fb->hWnd);
-   
-   GetClientRect( fb->hWnd, &rect );
-   width = rect.right - rect.left;
-   height = rect.bottom - rect.top;
-
-   if(width < 1)
-      width = 1;
-   if(height < 1)
-      height = 1;
-
-   if(width != fb->width || height != fb->height) {
-      fb->must_resize = TRUE;
-      fb->width = width; 
-      fb->height = height; 
-   }
-}
-
-
-/**
- * Create a new framebuffer object which will correspond to the given HDC.
- */
 struct stw_framebuffer *
-stw_framebuffer_create_locked(
+stw_framebuffer_create(
    HDC hdc,
    int iPixelFormat )
 {
@@ -193,8 +207,16 @@ stw_framebuffer_create_locked(
 
    pipe_mutex_init( fb->mutex );
 
+   /* This is the only case where we lock the stw_framebuffer::mutex before
+    * stw_dev::fb_mutex, since no other thread can know about this framebuffer
+    * and we must prevent any other thread from destroying it before we return.
+    */
+   pipe_mutex_lock( fb->mutex );
+
+   pipe_mutex_lock( stw_dev->fb_mutex );
    fb->next = stw_dev->fb_head;
    stw_dev->fb_head = fb;
+   pipe_mutex_unlock( stw_dev->fb_mutex );
 
    return fb;
 }
@@ -204,8 +226,8 @@ BOOL
 stw_framebuffer_allocate(
    struct stw_framebuffer *fb)
 {
-   pipe_mutex_lock( fb->mutex );
-
+   assert(fb);
+   
    if(!fb->stfb) {
       const struct stw_pixelformat_info *pfi = fb->pfi;
       enum pipe_format colorFormat, depthFormat, stencilFormat;
@@ -241,8 +263,6 @@ stw_framebuffer_allocate(
       fb->must_resize = TRUE;
    }
    
-   pipe_mutex_unlock( fb->mutex );
-
    return fb->stfb ? TRUE : FALSE;
 }
 
@@ -279,24 +299,27 @@ stw_framebuffer_cleanup( void )
    struct stw_framebuffer *fb;
    struct stw_framebuffer *next;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->fb_mutex );
 
    fb = stw_dev->fb_head;
    while (fb) {
       next = fb->next;
+      
+      pipe_mutex_lock(fb->mutex);
       stw_framebuffer_destroy_locked(fb);
+      
       fb = next;
    }
    stw_dev->fb_head = NULL;
    
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->fb_mutex );
 }
 
 
 /**
  * Given an hdc, return the corresponding stw_framebuffer.
  */
-struct stw_framebuffer *
+static INLINE struct stw_framebuffer *
 stw_framebuffer_from_hdc_locked(
    HDC hdc )
 {
@@ -313,8 +336,10 @@ stw_framebuffer_from_hdc_locked(
       return stw_framebuffer_from_hwnd_locked(hwnd);
    
    for (fb = stw_dev->fb_head; fb != NULL; fb = fb->next)
-      if (fb->hDC == hdc)
+      if (fb->hDC == hdc) {
+         pipe_mutex_lock(fb->mutex);
          break;
+      }
 
    return fb;
 }
@@ -329,9 +354,26 @@ stw_framebuffer_from_hdc(
 {
    struct stw_framebuffer *fb;
 
-   pipe_mutex_lock( stw_dev->mutex );
+   pipe_mutex_lock( stw_dev->fb_mutex );
    fb = stw_framebuffer_from_hdc_locked(hdc);
-   pipe_mutex_unlock( stw_dev->mutex );
+   pipe_mutex_unlock( stw_dev->fb_mutex );
+
+   return fb;
+}
+
+
+/**
+ * Given an hdc, return the corresponding stw_framebuffer.
+ */
+struct stw_framebuffer *
+stw_framebuffer_from_hwnd(
+   HWND hwnd )
+{
+   struct stw_framebuffer *fb;
+
+   pipe_mutex_lock( stw_dev->fb_mutex );
+   fb = stw_framebuffer_from_hwnd_locked(hwnd);
+   pipe_mutex_unlock( stw_dev->fb_mutex );
 
    return fb;
 }
@@ -351,22 +393,19 @@ stw_pixelformat_set(
    if (index >= count)
       return FALSE;
 
-   pipe_mutex_lock( stw_dev->mutex );
-   
    fb = stw_framebuffer_from_hdc_locked(hdc);
    if(fb) {
       /* SetPixelFormat must be called only once */
-      pipe_mutex_unlock( stw_dev->mutex );
+      stw_framebuffer_release( fb );
       return FALSE;
    }
 
-   fb = stw_framebuffer_create_locked(hdc, iPixelFormat);
+   fb = stw_framebuffer_create(hdc, iPixelFormat);
    if(!fb) {
-      pipe_mutex_unlock( stw_dev->mutex );
       return FALSE;
    }
       
-   pipe_mutex_unlock( stw_dev->mutex );
+   stw_framebuffer_release( fb );
 
    /* Some applications mistakenly use the undocumented wglSetPixelFormat 
     * function instead of SetPixelFormat, so we call SetPixelFormat here to 
@@ -383,13 +422,16 @@ int
 stw_pixelformat_get(
    HDC hdc )
 {
+   int iPixelFormat = 0;
    struct stw_framebuffer *fb;
 
    fb = stw_framebuffer_from_hdc(hdc);
-   if(!fb)
-      return 0;
+   if(fb) {
+      iPixelFormat = fb->iPixelFormat;
+      stw_framebuffer_release(fb);
+   }
    
-   return fb->iPixelFormat;
+   return iPixelFormat;
 }
 
 
@@ -405,10 +447,10 @@ stw_swap_buffers(
    if (fb == NULL)
       return FALSE;
 
-   if (!(fb->pfi->pfd.dwFlags & PFD_DOUBLEBUFFER))
+   if (!(fb->pfi->pfd.dwFlags & PFD_DOUBLEBUFFER)) {
+      stw_framebuffer_release(fb);
       return TRUE;
-
-   pipe_mutex_lock( fb->mutex );
+   }
 
    /* If we're swapping the buffer associated with the current context
     * we have to flush any pending rendering commands first.
@@ -419,7 +461,7 @@ stw_swap_buffers(
    
    if(!st_get_framebuffer_surface( fb->stfb, ST_SURFACE_BACK_LEFT, &surface )) {
       /* FIXME: this shouldn't happen, but does on glean */
-      pipe_mutex_unlock( fb->mutex );
+      stw_framebuffer_release(fb);
       return FALSE;
    }
 
@@ -433,8 +475,7 @@ stw_swap_buffers(
    stw_dev->stw_winsys->flush_frontbuffer( screen, surface, hdc );
    
    stw_framebuffer_update(fb);
-   
-   pipe_mutex_unlock( fb->mutex );
+   stw_framebuffer_release(fb);
    
    return TRUE;
 }
diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h
index 759e06b891..13d29f37e4 100644
--- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h
+++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h
@@ -41,6 +41,23 @@ struct stw_pixelformat_info;
  */
 struct stw_framebuffer
 {
+   /**
+    * This mutex has two purposes:
+    * - protect the access to the mutable data members below
+    * - prevent the the framebuffer from being deleted while being accessed.
+    * 
+    * It is OK to lock this mutex while holding the stw_device::fb_mutex lock, 
+    * but the opposite must never happen.
+    */
+   pipe_mutex mutex;
+   
+   /*
+    * Immutable members.
+    * 
+    * Note that even access to immutable members implies acquiring the mutex 
+    * above, to prevent the framebuffer from being destroyed.
+    */
+   
    HDC hDC;
    HWND hWnd;
 
@@ -48,7 +65,10 @@ struct stw_framebuffer
    const struct stw_pixelformat_info *pfi;
    GLvisual visual;
 
-   pipe_mutex mutex;
+   /*
+    * Mutable members. 
+    */
+
    struct st_framebuffer *stfb;
    
    /* FIXME: Make this work for multiple contexts bound to the same framebuffer */
@@ -56,15 +76,52 @@ struct stw_framebuffer
    unsigned width;
    unsigned height;
    
-   /** This is protected by stw_device::mutex, not the mutex above */
+   /** 
+    * This is protected by stw_device::fb_mutex, not the mutex above.
+    * 
+    * Deletions must be done by first acquiring stw_device::fb_mutex, and then
+    * acquiring the stw_framebuffer::mutex of the framebuffer to be deleted. 
+    * This ensures that nobody else is reading/writing to the.
+    * 
+    * It is not necessary to aquire the mutex above to navigate the linked list
+    * given that deletions are done with stw_device::fb_mutex held, so no other
+    * thread can delete.
+    */
    struct stw_framebuffer *next;
 };
 
+
+/**
+ * Create a new framebuffer object which will correspond to the given HDC.
+ * 
+ * This function will acquire stw_framebuffer::mutex. stw_framebuffer_release
+ * must be called when done 
+ */
 struct stw_framebuffer *
-stw_framebuffer_create_locked(
+stw_framebuffer_create(
    HDC hdc,
    int iPixelFormat );
 
+/**
+ * Search a framebuffer with a matching HWND.
+ * 
+ * This function will acquire stw_framebuffer::mutex. stw_framebuffer_release
+ * must be called when done 
+ */
+struct stw_framebuffer *
+stw_framebuffer_from_hwnd(
+   HWND hwnd );
+
+/**
+ * Search a framebuffer with a matching HDC.
+ * 
+ * This function will acquire stw_framebuffer::mutex. stw_framebuffer_release
+ * must be called when done 
+ */
+struct stw_framebuffer *
+stw_framebuffer_from_hdc(
+   HDC hdc );
+
 BOOL
 stw_framebuffer_allocate(
    struct stw_framebuffer *fb );
@@ -73,15 +130,19 @@ void
 stw_framebuffer_update(
    struct stw_framebuffer *fb);
 
+/**
+ * Release stw_framebuffer::mutex lock. This framebuffer must not be accessed
+ * after calling this function, as it may have been deleted by another thread
+ * in the meanwhile.
+ */
 void
-stw_framebuffer_cleanup(void);
-
-struct stw_framebuffer *
-stw_framebuffer_from_hdc_locked(
-   HDC hdc );
+stw_framebuffer_release(
+   struct stw_framebuffer *fb);
 
-struct stw_framebuffer *
-stw_framebuffer_from_hdc(
-   HDC hdc );
+/**
+ * Cleanup any existing framebuffers when exiting application.
+ */
+void
+stw_framebuffer_cleanup(void);
 
 #endif /* STW_FRAMEBUFFER_H */
diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c
index 3ce410d9be..ac2eb05341 100644
--- a/src/glx/x11/dri_glx.c
+++ b/src/glx/x11/dri_glx.c
@@ -305,6 +305,7 @@ CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc,
     drm_handle_t  hFB;
     int        junk;
     const __DRIconfig **driver_configs;
+    __GLcontextModes *visual;
 
     /* DRI protocol version. */
     dri_version.major = driDpy->driMajor;
@@ -417,6 +418,28 @@ CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc,
     psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs);
     psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs);
 
+    /* Visuals with depth != screen depth are subject to automatic compositing
+     * in the X server, so DRI1 can't render to them properly. Mark them as
+     * non-conformant to prevent apps from picking them up accidentally.
+     */
+    for (visual = psc->visuals; visual; visual = visual->next) {
+	XVisualInfo template;
+	XVisualInfo *visuals;
+	int num_visuals;
+	long mask;
+
+	template.visualid = visual->visualID;
+	mask = VisualIDMask;
+	visuals = XGetVisualInfo(dpy, mask, &template, &num_visuals);
+
+	if (visuals) {
+	    if (num_visuals > 0 && visuals->depth != DefaultDepth(dpy, scrn))
+		visual->visualRating = GLX_NON_CONFORMANT_CONFIG;
+
+	    XFree(visuals);
+	}
+    }
+
     return psp;
 
  handle_error:
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index c02ccc3528..299aa762c2 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -137,22 +137,15 @@ find_translated_vp(struct st_context *st,
 
       for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) {
          if (fragInputsRead & (1 << inAttr)) {
+            stfp->input_to_slot[inAttr] = numIn;
+            numIn++;
             if ((fragInputsRead & FRAG_BIT_FOGC)) {
-               if (stfp->Base.UsesPointCoord) {
-                  stfp->input_to_slot[inAttr] = numIn;
-                  numIn++;
+               /* leave placeholders for the
+                * extra registers we extract from fog */
+               if (stfp->Base.UsesFrontFacing ||
+                   stfp->Base.UsesPointCoord) {
+                  numIn += 2;
                }
-               if (stfp->Base.UsesFrontFacing) {
-                  stfp->input_to_slot[inAttr] = numIn;
-                  numIn++;
-               }
-               if (stfp->Base.UsesFogFragCoord) {
-                  stfp->input_to_slot[inAttr] = numIn;
-                  numIn++;
-               }
-            } else {
-               stfp->input_to_slot[inAttr] = numIn;
-               numIn++;
             }
          }
          else {
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 2027b713ce..8b5094a04f 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -1037,10 +1037,16 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
          st_cond_flush_get_tex_transfer(st, rbRead->texture, 0, 0, 0,
 					PIPE_TRANSFER_READ, srcx, srcy, width,
 					height);
+      struct pipe_transfer *ptTex;
+      enum pipe_transfer_usage transfer_usage;
 
-      struct pipe_transfer *ptTex =
-         st_cond_flush_get_tex_transfer(st, pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
-					0, 0, width, height);
+      if (type == GL_DEPTH && pf_is_depth_and_stencil(pt->format))
+         transfer_usage = PIPE_TRANSFER_READ_WRITE;
+      else
+         transfer_usage = PIPE_TRANSFER_WRITE;
+
+      ptTex = st_cond_flush_get_tex_transfer(st, pt, 0, 0, 0, transfer_usage,
+                                             0, 0, width, height);
 
       if (type == GL_COLOR) {
          /* alternate path using get/put_tile() */
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 0a72784ce0..909189f9d3 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -529,6 +529,7 @@ st_TexImage(GLcontext * ctx,
    GLint texelBytes, sizeInBytes;
    GLuint dstRowStride;
    struct gl_pixelstore_attrib unpackNB;
+   enum pipe_transfer_usage transfer_usage;
 
    DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
@@ -680,8 +681,14 @@ st_TexImage(GLcontext * ctx,
    }
 
    if (stImage->pt) {
+      if (format == GL_DEPTH_COMPONENT &&
+          pf_is_depth_and_stencil(stImage->pt->format))
+         transfer_usage = PIPE_TRANSFER_READ_WRITE;
+      else
+         transfer_usage = PIPE_TRANSFER_WRITE;
+
       texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
-                                            PIPE_TRANSFER_WRITE, 0, 0,
+                                            transfer_usage, 0, 0,
                                             stImage->base.Width,
                                             stImage->base.Height);
       if(stImage->transfer)
@@ -742,7 +749,7 @@ st_TexImage(GLcontext * ctx,
 	    st_texture_image_unmap(ctx->st, stImage);
             /* map next slice of 3D texture */
 	    texImage->Data = st_texture_image_map(ctx->st, stImage, i + 1,
-                                                  PIPE_TRANSFER_WRITE, 0, 0,
+                                                  transfer_usage, 0, 0,
                                                   stImage->base.Width,
                                                   stImage->base.Height);
 	    src += srcImageStride;
@@ -1041,6 +1048,7 @@ st_TexSubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level,
       _mesa_image_image_stride(packing, width, height, format, type);
    GLint i;
    const GLubyte *src;
+   enum pipe_transfer_usage transfer_usage;
 
    DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target),
@@ -1072,10 +1080,16 @@ st_TexSubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level,
     * from uploading the buffer under us.
     */
    if (stImage->pt) {
+      if (format == GL_DEPTH_COMPONENT &&
+          pf_is_depth_and_stencil(stImage->pt->format))
+         transfer_usage = PIPE_TRANSFER_READ_WRITE;
+      else
+         transfer_usage = PIPE_TRANSFER_WRITE;
+
       st_teximage_flush_before_map(ctx->st, stImage->pt, 0, level,
-				   PIPE_TRANSFER_WRITE);
+				   transfer_usage);
       texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, 
-                                            PIPE_TRANSFER_WRITE,
+                                            transfer_usage,
                                             xoffset, yoffset,
                                             width, height);
    }
@@ -1106,7 +1120,7 @@ st_TexSubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level,
          /* map next slice of 3D texture */
 	 texImage->Data = st_texture_image_map(ctx->st, stImage,
                                                zoffset + i + 1,
-                                               PIPE_TRANSFER_WRITE,
+                                               transfer_usage,
                                                xoffset, yoffset,
                                                width, height);
 	 src += srcImageStride;
@@ -1274,6 +1288,7 @@ fallback_copy_texsubimage(GLcontext *ctx, GLenum target, GLint level,
    struct pipe_screen *screen = pipe->screen;
    struct pipe_transfer *src_trans;
    GLvoid *texDest;
+   enum pipe_transfer_usage transfer_usage;
 
    assert(width <= MAX_WIDTH);
 
@@ -1288,10 +1303,16 @@ fallback_copy_texsubimage(GLcontext *ctx, GLenum target, GLint level,
 					       srcX, srcY,
 					       width, height);
 
+   if (baseFormat == GL_DEPTH_COMPONENT &&
+       pf_is_depth_and_stencil(stImage->pt->format))
+      transfer_usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      transfer_usage = PIPE_TRANSFER_WRITE;
+
    st_teximage_flush_before_map(ctx->st, stImage->pt, 0, 0,
-				PIPE_TRANSFER_WRITE);
+				transfer_usage);
 
-   texDest = st_texture_image_map(ctx->st, stImage, 0, PIPE_TRANSFER_WRITE,
+   texDest = st_texture_image_map(ctx->st, stImage, 0, transfer_usage,
                                   destX, destY, width, height);
 
    if (baseFormat == GL_DEPTH_COMPONENT ||
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index 13589151ba..ca32b2e573 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -280,7 +280,8 @@ st_notify_swapbuffers(struct st_framebuffer *stfb)
 		PIPE_FLUSH_SWAPBUFFERS |
 		PIPE_FLUSH_FRAME,
                 NULL );
-      ctx->st->frontbuffer_status = FRONT_STATUS_COPY_OF_BACK;
+      if (st_renderbuffer(stfb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer))
+         ctx->st->frontbuffer_status = FRONT_STATUS_COPY_OF_BACK;
    }
 }
 
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index 3140ebe04a..8aef3fc6dc 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -122,8 +122,15 @@ map_register_file_index(
                                      SWIZZLE_X,
                                      SWIZZLE_X,
                                      SWIZZLE_X);
+            /* register after fog */
+            return inputMapping[index] + 1;
          } else {
-            /* fixme: point coord */
+            *swizzle = MAKE_SWIZZLE4(SWIZZLE_Z,
+                                     SWIZZLE_W,
+                                     SWIZZLE_Z,
+                                     SWIZZLE_W);
+            /* register after frontface */
+            return inputMapping[index] + 2;
          }
       }
       /* inputs are mapped according to the user-defined map */
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 3365bc0b8b..e4d3bb33c7 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -453,20 +453,34 @@ st_translate_fragment_program(struct st_context *st,
             stfp->input_semantic_index[slot] = 1;
             interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
             break;
-         case FRAG_ATTRIB_FOGC:
-            if (stfp->Base.UsesPointCoord) {
-               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
-               stfp->input_semantic_index[slot] = num_generic++;
-               interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
-            } else if (stfp->Base.UsesFrontFacing) {
-               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FACE;
-               stfp->input_semantic_index[slot] = 0;
-               interpMode[slot] = TGSI_INTERPOLATE_CONSTANT;
-            } else {
+         case FRAG_ATTRIB_FOGC: {
+            int extra_decls = 0;
+            if (stfp->Base.UsesFogFragCoord) {
                stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
                stfp->input_semantic_index[slot] = 0;
                interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+               input_flags[slot] = stfp->Base.Base.InputFlags[attr];
+               ++extra_decls;
 	    }
+            if (stfp->Base.UsesFrontFacing) {
+               GLint idx = slot + extra_decls;
+               stfp->input_semantic_name[idx] = TGSI_SEMANTIC_FACE;
+               stfp->input_semantic_index[idx] = 0;
+               interpMode[idx] = TGSI_INTERPOLATE_CONSTANT;
+               input_flags[idx] = stfp->Base.Base.InputFlags[attr];
+               ++extra_decls;
+            }
+            if (stfp->Base.UsesPointCoord) {
+               GLint idx = slot + extra_decls;
+               stfp->input_semantic_name[idx] = TGSI_SEMANTIC_GENERIC;
+               stfp->input_semantic_index[idx] = num_generic++;
+               interpMode[idx] = TGSI_INTERPOLATE_PERSPECTIVE;
+               input_flags[idx] = stfp->Base.Base.InputFlags[attr];
+               ++extra_decls;
+            }
+            fs_num_inputs += extra_decls - 1;
+            continue;
+         }
             break;
          case FRAG_ATTRIB_TEX0:
          case FRAG_ATTRIB_TEX1: