Get batchbuffers working natively again. This code still

relies over-heavily on the DRI lock to protect offsets in the
command stream from changing before being queued on the ring,
and should be viewed as being a pretty temporary mechanism
before a more robust alternative is implemented, most likely
based on a batchbuffer relocation list that will be used
to patch prebuilt batchbuffers after validation takes place.

19 files changed, 173 insertions, 123 deletions

diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index 132794d798c..d113619e9f9 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -29,7 +29,6 @@ DRIVER_SOURCES = \
 	intel_context.c \
 	intel_ioctl.c \
 	intel_pixel.c \
-	intel_render.c \
 	intel_screen.c \
 	intel_span.c \
 	intel_state.c \
@@ -37,6 +36,7 @@ DRIVER_SOURCES = \
 
 
 DISABLED = \
+	intel_render.c \
 	i830_context.c \
 	i830_metaops.c \
 	i830_state.c \
diff --git a/src/mesa/drivers/dri/i915/bufmgr.h b/src/mesa/drivers/dri/i915/bufmgr.h
index e435b9d975f..3761a82b598 100644
--- a/src/mesa/drivers/dri/i915/bufmgr.h
+++ b/src/mesa/drivers/dri/i915/bufmgr.h
@@ -137,7 +137,7 @@ int bmValidateBufferList( struct bufmgr *,
 			  struct bm_buffer_list *,
 			  unsigned flags );
 
-void bmFenceBufferList( struct bufmgr *,
+unsigned bmFenceBufferList( struct bufmgr *,
 			struct bm_buffer_list * );
 
 void bmFreeBufferList( struct bm_buffer_list * );
diff --git a/src/mesa/drivers/dri/i915/bufmgr_fake.c b/src/mesa/drivers/dri/i915/bufmgr_fake.c
index 7a2fec8bf42..d08a642a545 100644
--- a/src/mesa/drivers/dri/i915/bufmgr_fake.c
+++ b/src/mesa/drivers/dri/i915/bufmgr_fake.c
@@ -166,6 +166,7 @@ static int bmAllocMem( struct bufmgr *bm,
    if (buf->block)
       buf->block->buf = buf;
 
+   assert(buf->block);
    return buf->block != NULL;
 }
 
@@ -255,6 +256,8 @@ static int move_buffers( struct bufmgr *bm,
 	 if (flags & BM_NO_UPLOAD)
 	    goto cleanup;
 
+	 assert(!buffers[i]->mapped);
+
 	 DBG("try to move buffer %d size 0x%x to pools 0x%x\n", 
 		      buffers[i]->id, buffers[i]->size, flags & BM_MEM_MASK);
 
@@ -308,7 +311,7 @@ static int move_buffers( struct bufmgr *bm,
 	 free_block(bm, newMem[i]);
    }
    
-   DBG("%s - fail\n", __FUNCTION__);
+   _mesa_printf("%s - fail\n", __FUNCTION__);
    return 0;   
 }
 
@@ -538,6 +541,8 @@ void bmBufferData(struct bufmgr *bm,
 
    DBG("bmBufferData %d sz 0x%x data: %p\n", buffer, size, data);
 
+   assert(!buf->mapped);
+
    if (buf->block) {
       if ((buf->block->mem_type != BM_MEM_LOCAL && !bmTestFence(bm, buf->block->fence)) ||
 	  (buf->size && buf->size != size) ||
@@ -699,6 +704,7 @@ int bmValidateBufferList( struct bufmgr *bm,
       if (!delayed_free(bm) &&
 	  !evict_lru(bm, flags))
 	 return 0;
+      exit(1);
    }
 
 
@@ -706,7 +712,8 @@ int bmValidateBufferList( struct bufmgr *bm,
       DBG("%d: buf %d ofs 0x%x\n",
 		   i, bufs[i]->id, bufs[i]->block->mem->ofs);
 
-      list->elem[i].offset_return[0] = bufs[i]->block->mem->ofs;
+      if (list->elem[i].offset_return)
+	 list->elem[i].offset_return[0] = bufs[i]->block->mem->ofs;
       
       if (list->elem[i].memtype_return)
 	 list->elem[i].memtype_return[0] = bufs[i]->block->mem_type;
@@ -724,7 +731,7 @@ int bmValidateBufferList( struct bufmgr *bm,
  * The buffer manager knows how to emit and test fences directly
  * through the drm and without callbacks or whatever into the driver.
  */
-void bmFenceBufferList( struct bufmgr *bm, struct bm_buffer_list *list )
+unsigned bmFenceBufferList( struct bufmgr *bm, struct bm_buffer_list *list )
 {
 
    DBG("%s (%d bufs)\n", __FUNCTION__, list->nr);
@@ -741,7 +748,11 @@ void bmFenceBufferList( struct bufmgr *bm, struct bm_buffer_list *list )
 	 move_to_head(&buf->block->pool->lru, buf->block);
 	 buf->block->fence = fence;
       }
+
+      return fence;
    }
+   else
+      return 0;
 }
 
 
@@ -761,6 +772,9 @@ unsigned bmSetFence( struct bufmgr *bm )
 
 int bmTestFence( struct bufmgr *bm, unsigned fence )
 {
+/*    if (fence % 1024 == 0) */
+/*       _mesa_printf("%d %d\n", fence, bm->intel->sarea->last_dispatch); */
+
    return fence <= bm->intel->sarea->last_dispatch;
 }
 
diff --git a/src/mesa/drivers/dri/i915/i830_metaops.c b/src/mesa/drivers/dri/i915/i830_metaops.c
index 50511bfa7fd..8984b6c6ecd 100644
--- a/src/mesa/drivers/dri/i915/i830_metaops.c
+++ b/src/mesa/drivers/dri/i915/i830_metaops.c
@@ -442,7 +442,7 @@ i830TryTextureReadPixels( GLcontext *ctx,
    int textureFormat;
    GLenum glTextureFormat;
    int src_offset = i830->meta.Buffer[I830_DESTREG_CBUFADDR2];
-   int destOffset = intelAgpOffsetFromVirtual( &i830->intel, pixels);
+   int destOffset = 0;
    int destFormat, depthFormat, destPitch;
    drm_clip_rect_t tmp;
 
@@ -596,7 +596,7 @@ i830TryTextureDrawPixels( GLcontext *ctx,
    int textureFormat;
    GLenum glTextureFormat;
    int dst_offset = i830->meta.Buffer[I830_DESTREG_CBUFADDR2];
-   int src_offset = intelAgpOffsetFromVirtual( intel, pixels );
+   int src_offset = 0;
 
    if (INTEL_DEBUG & DEBUG_PIXEL)
       fprintf(stderr, "%s\n", __FUNCTION__);
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index 1d1e922e73e..ea54b673d2d 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -109,7 +109,7 @@ GLboolean i915CreateContext( const __GLcontextModes *mesaVis,
 
    if (!i915) return GL_FALSE;
 
-   _mesa_printf( "\ntexmem branch\n\n");
+   _mesa_printf( "\ntexmem branch (i915)\n\n");
    
    i915InitVtbl( i915 );
 
@@ -167,6 +167,7 @@ GLboolean i915CreateContext( const __GLcontextModes *mesaVis,
 				 intelScreen->depth.pitch,
 				 intelScreen->height);
 
+   intelInitBatchBuffer(intel);
 
    /* Advertise the full hardware capabilities.  The new memory
     * manager should cope much better with overload situations:
diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h
index 6d9c4ea6ce8..0db119cd413 100644
--- a/src/mesa/drivers/dri/i915/i915_context.h
+++ b/src/mesa/drivers/dri/i915/i915_context.h
@@ -43,6 +43,7 @@
 #define I915_UPLOAD_PROGRAM          0x8
 #define I915_UPLOAD_CONSTANTS        0x10
 #define I915_UPLOAD_FOG              0x20
+#define I915_UPLOAD_INVARIENT        0x40
 #define I915_UPLOAD_TEX(i)           (0x00010000<<(i))
 #define I915_UPLOAD_TEX_ALL          (0x00ff0000)
 #define I915_UPLOAD_TEX_0_SHIFT      16
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index 39156122974..4f4792c0eed 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -237,6 +237,11 @@ static void i915_emit_state( intelContextPtr intel )
    if (VERBOSE) 
       fprintf(stderr, "%s dirty: %x\n", __FUNCTION__, dirty);
 
+   if (dirty & I915_UPLOAD_INVARIENT) {
+      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_INVARIENT:\n"); 
+      i915_emit_invarient_state( intel );
+   }
+
    if (dirty & I915_UPLOAD_CTX) {
       if (VERBOSE) fprintf(stderr, "I915_UPLOAD_CTX:\n"); 
       emit( i915, state->Ctx, sizeof(state->Ctx) );
diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.c b/src/mesa/drivers/dri/i915/intel_batchbuffer.c
index e8e61f49c13..6f54fdb1c5a 100644
--- a/src/mesa/drivers/dri/i915/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.c
@@ -37,6 +37,7 @@
 #include "intel_batchbuffer.h"
 #include "intel_context.h"
 
+#include "bufmgr.h"
 
 
 
@@ -59,6 +60,9 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv )
 
    intelFlush( &intel->ctx );
    LOCK_HARDWARE( intel );
+   intelInstallBatchBuffer(intel);   
+   intelValidateBuffers( intel );
+
    {
       intelScreenPrivate *intelScreen = intel->intelScreen;
       __DRIdrawablePrivate *dPriv = intel->driDrawable;
@@ -117,6 +121,8 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv )
       }
    }
    intelFlushBatchLocked( intel, GL_TRUE, GL_TRUE, GL_TRUE );
+   assert(intel->buffer_list == NULL);
+
    UNLOCK_HARDWARE( intel );
 }
 
@@ -273,6 +279,8 @@ void intelClearWithBlit(GLcontext *ctx, GLbitfield flags, GLboolean all,
 
    intelFlush( &intel->ctx );
    LOCK_HARDWARE( intel );
+   intelInstallBatchBuffer(intel);   
+   intelValidateBuffers( intel );
    {
       /* flip top to bottom */
       cy = intel->driDrawable->h-cy1-ch;
@@ -362,60 +370,72 @@ void intelClearWithBlit(GLcontext *ctx, GLbitfield flags, GLboolean all,
 
 
 
-void intelDestroyBatchBuffer( GLcontext *ctx )
+void intelDestroyBatchBuffer( struct intel_context *intel )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-
-   if (intel->alloc.ptr) {
-      intelFreeAGP( intel, intel->alloc.ptr );
-      intel->alloc.ptr = 0;
-   }
 }
 
 
-void intelInitBatchBuffer( GLcontext *ctx )
+void intelInstallBatchBuffer( struct intel_context *intel )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   assert(!intel->batch.ptr);
+
+   intel->alloc.current++;
+   intel->alloc.current %= INTEL_ALLOC_NR;
+
+   DBG("%s: %d\n", __FUNCTION__, intel->alloc.current);
+
+   intel->batch.size = INTEL_ALLOC_SIZE;
+   intel->batch.space = intel->batch.size;
+   intel->batch.start_offset = 0;
 
-   /* AGP allocation won't work: 
+   intel->batch.ptr = bmMapBuffer( intel->bm, 
+				   intel->alloc.buffer[intel->alloc.current],
+				   BM_WRITE | BM_MEM_AGP );
+
+
+   assert(!intel->buffer_list);
+   intel->buffer_list = bmNewBufferList();
+      
+   /* Add the batchbuffer 
     */
-   if (1 || !intel->intelScreen->allow_batchbuffer || getenv("INTEL_NO_BATCH")) {
-      intel->alloc.size = 8 * 1024;
-      intel->alloc.ptr = malloc( intel->alloc.size );
-      intel->alloc.offset = 0;
-   }
-   else {
-      switch (intel->intelScreen->deviceID) {
-      case PCI_CHIP_I865_G:
-	 /* HW bug?  Seems to crash if batchbuffer crosses 4k boundary.
-	  */
-	 intel->alloc.size = 8 * 1024; 
-	 break;
-      default:
-	 /* This is the smallest amount of memory the kernel deals with.
-	  * We'd ideally like to make this smaller.
-	  */
-	 intel->alloc.size = 1 << intel->intelScreen->logTextureGranularity;
-	 break;
+   bmAddBuffer(intel->buffer_list,
+	       intel->alloc.buffer[intel->alloc.current],
+	       BM_READ,
+	       NULL,
+	       &intel->batch.start_offset);
+
+
+   if (0) {
+      static int foo;
+      if (foo++ > 10) {
+	 _mesa_printf("foo\n");
+	 exit(1);
       }
-
-      intel->alloc.ptr = intelAllocateAGP( intel, intel->alloc.size );
-      if (intel->alloc.ptr)
-	 intel->alloc.offset = 
-	    intelAgpOffsetFromVirtual( intel, intel->alloc.ptr );
    }
+}
 
-   if (!intel->alloc.ptr) {
-      FALLBACK(intel, INTEL_FALLBACK_NO_BATCHBUFFER, 1);
-   }
-   else {
-      intel->prim.flush = 0;
-      intel->vtbl.emit_invarient_state( intel );
-
-      /* Make sure this gets to the hardware, even if we have no cliprects:
-       */
-      LOCK_HARDWARE( intel );
-      intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_TRUE );
-      UNLOCK_HARDWARE( intel );
-   }
+void intelInitBatchBuffer( struct intel_context *intel )
+{
+   GLint i;
+
+   _mesa_printf("%s: %d\n", __FUNCTION__, intel->alloc.current);
+   bmGenBuffers(intel->bm,
+		INTEL_ALLOC_NR,
+		intel->alloc.buffer);
+
+   for (i = 0; i < INTEL_ALLOC_NR; i++)
+      bmBufferData(intel->bm, 
+		   intel->alloc.buffer[i],
+		   INTEL_ALLOC_SIZE,
+		   NULL,
+		   BM_MEM_AGP);
+		
+
+}
+
+
+void intelValidateBuffers( struct intel_context *intel )
+{
+   if (!bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP))
+      assert(0);
 }
diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.h b/src/mesa/drivers/dri/i915/intel_batchbuffer.h
index 0d6aae93ae2..6e43fcb5f0e 100644
--- a/src/mesa/drivers/dri/i915/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.h
@@ -42,6 +42,7 @@ do {									\
    if (VERBOSE) fprintf(stderr, 					\
 			"BEGIN_BATCH(%d) in %s, %d dwords free\n",	\
 			(n), __FUNCTION__, intel->batch.space/4);	\
+   assert(intel->locked); \
    if (intel->batch.space < (n)*4)					\
       intelFlushBatch(intel, GL_TRUE);					\
    batch_ptr = intel->batch.ptr;					\
@@ -62,8 +63,11 @@ do {								\
    assert(intel->batch.space >= 0);				\
 } while(0)
 
-extern void intelInitBatchBuffer( GLcontext *ctx );
-extern void intelDestroyBatchBuffer( GLcontext *ctx );
+extern void intelInitBatchBuffer( struct intel_context *intel );
+extern void intelDestroyBatchBuffer( struct intel_context *intel );
+
+void intelInstallBatchBuffer( struct intel_context *intel );
+
 
 extern void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim );
 extern void intelWrapInlinePrimitive( intelContextPtr intel );
@@ -97,5 +101,6 @@ extern void intelEmitFillBlitLocked( intelContextPtr intel,
 
 GLuint *intelExtendInlinePrimitive( intelContextPtr intel, 
 				    GLuint dwords );
+void intelValidateBuffers( struct intel_context *intel );
 
 #endif
diff --git a/src/mesa/drivers/dri/i915/intel_blit.c b/src/mesa/drivers/dri/i915/intel_blit.c
new file mode 100644
index 00000000000..e69de29bb2d
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_blit.c
diff --git a/src/mesa/drivers/dri/i915/intel_context.c b/src/mesa/drivers/dri/i915/intel_context.c
index b9a7b8622d6..ff5a6e43f23 100644
--- a/src/mesa/drivers/dri/i915/intel_context.c
+++ b/src/mesa/drivers/dri/i915/intel_context.c
@@ -389,7 +389,7 @@ GLboolean intelInitContext( intelContextPtr intel,
 /* 			  DRI_TEXMGR_DO_TEXTURE_RECT ); */
 
 
-   intel->prim.flush = intelInitBatchBuffer;
+   intel->prim.flush = NULL;
    intel->prim.primitive = ~0;
 
 
@@ -433,7 +433,7 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
       _swrast_DestroyContext (&intel->ctx);
       intel->Fallback = 0;	/* don't call _swrast_Flush later */
 
-      intelDestroyBatchBuffer(&intel->ctx);
+      intelDestroyBatchBuffer(intel);
       
 
       if ( release_texture_heaps ) {
@@ -581,6 +581,13 @@ void intelGetLock( intelContextPtr intel, GLuint flags )
    __DRIscreenPrivate *sPriv = intel->driScreen;
    drmI830Sarea * sarea = intel->sarea;
    int me = intel->hHWContext;
+   static int foo = 0;
+
+/*    _mesa_printf("%s\n", __FUNCTION__); */
+/*    if (foo++ > 1) { */
+/*       _mesa_printf("%s - foo\n", __FUNCTION__); */
+/*       abort(); */
+/*    } */
 
    drmGetLock(intel->driFd, intel->hHWContext, flags);
 
diff --git a/src/mesa/drivers/dri/i915/intel_context.h b/src/mesa/drivers/dri/i915/intel_context.h
index 7f7f0f39088..1234a8241dc 100644
--- a/src/mesa/drivers/dri/i915/intel_context.h
+++ b/src/mesa/drivers/dri/i915/intel_context.h
@@ -146,6 +146,8 @@ struct intel_context
    GLuint Fallback;
    GLuint NewGLState;
    
+   GLuint last_fence;
+
    struct {
       GLuint start_offset;
       GLint size;
@@ -153,15 +155,16 @@ struct intel_context
       GLubyte *ptr;
    } batch;
       
+#define INTEL_ALLOC_NR 64
+#define INTEL_ALLOC_SIZE 4096
+
    struct {
-      void *ptr;
-      GLint size;
-      GLuint offset;
-      GLuint active_buf;
-      GLuint irq_emitted;
+      GLuint buffer[INTEL_ALLOC_NR];
+      GLuint current;
    } alloc;
 
    struct {
+      GLuint id;
       GLuint primitive;
       GLubyte *start_ptr;      
       void (*flush)( GLcontext * );
@@ -211,11 +214,12 @@ struct intel_context
    GLenum render_primitive;
    GLenum reduced_primitive;
    GLuint vertex_size;
-   char *verts;			/* points to tnl->clipspace.vertex_buf */
+   GLubyte *verts;			/* points to tnl->clipspace.vertex_buf */
 
 
    struct intel_region *front_region;
    struct intel_region *back_region;
+   struct intel_region *draw_region;
    struct intel_region *depth_region;
 
 
diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.h b/src/mesa/drivers/dri/i915/intel_ioctl.h
index d16dae8e870..89d2be06323 100644
--- a/src/mesa/drivers/dri/i915/intel_ioctl.h
+++ b/src/mesa/drivers/dri/i915/intel_ioctl.h
@@ -46,11 +46,6 @@ extern void intelRefillBatchLocked( intelContextPtr intel, GLboolean allow_unloc
 extern void intelFinish( GLcontext *ctx );
 extern void intelFlush( GLcontext *ctx );
 
-extern void *intelAllocateAGP( intelContextPtr intel, GLsizei size );
-extern void intelFreeAGP( intelContextPtr intel, void *pointer );
-
-
-extern GLuint intelAgpOffsetFromVirtual( intelContextPtr intel, const GLvoid *p );
 
 void intelWaitIrq( intelContextPtr intel, int seq );
 int intelEmitIrqLocked( intelContextPtr intel );
diff --git a/src/mesa/drivers/dri/i915/intel_pixel.c b/src/mesa/drivers/dri/i915/intel_pixel.c
index ee1e2f94074..cbd898f1f4c 100644
--- a/src/mesa/drivers/dri/i915/intel_pixel.c
+++ b/src/mesa/drivers/dri/i915/intel_pixel.c
@@ -180,7 +180,7 @@ intelTryReadPixels( GLcontext *ctx,
       int nbox = dPriv->numClipRects;
       int src_offset = intel->drawOffset;
       int src_pitch = intel->intelScreen->front.pitch;
-      int dst_offset = intelAgpOffsetFromVirtual( intel, pixels);
+      int dst_offset = 0;
       drm_clip_rect_t *box = dPriv->pClipRects;
       int i;
 
@@ -261,7 +261,7 @@ static void do_draw_pix( GLcontext *ctx,
    drm_clip_rect_t *box = dPriv->pClipRects;
    int nbox = dPriv->numClipRects;
    int i;
-   int src_offset = intelAgpOffsetFromVirtual( intel, pixels);
+   int src_offset = 0;
    int src_pitch = pitch;
 
    if (INTEL_DEBUG & DEBUG_PIXEL)
@@ -504,6 +504,7 @@ static GLboolean intelTryCopyPixels( GLcontext *ctx,
 
    intelFlush( &intel->ctx );
    LOCK_HARDWARE( intel );
+   intelInstallBatchBuffer( intel );
    {
       __DRIdrawablePrivate *dPriv = intel->driDrawable;
       drm_clip_rect_t *box = dPriv->pClipRects;
@@ -513,7 +514,6 @@ static GLboolean intelTryCopyPixels( GLcontext *ctx,
       GLuint dst_offset = 0;
       GLuint src_offset = 0;
       GLuint i;
-      struct bm_buffer_list *list = bmNewBufferList();
 
 #if 0
       dsty -= height;			/* cope with pixel zoom */
@@ -547,9 +547,12 @@ static GLboolean intelTryCopyPixels( GLcontext *ctx,
       dsty += dPriv->y;
 
 
-      bmAddBuffer(list, dst->buffer, BM_WRITE, NULL, &dst_offset);
-      bmAddBuffer(list, src->buffer, BM_READ, NULL, &src_offset);
-      if (!bmValidateBufferList(intel->bm, list, BM_NO_EVICT|BM_NO_UPLOAD|BM_MEM_AGP)) 
+      bmAddBuffer(intel->buffer_list, dst->buffer, BM_NO_EVICT|BM_NO_UPLOAD|BM_WRITE, 
+		  NULL, &dst_offset);
+      bmAddBuffer(intel->buffer_list, src->buffer, BM_NO_EVICT|BM_NO_UPLOAD|BM_READ, 
+		  NULL, &src_offset);
+
+      if (!bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP)) 
 	 goto out;
       
       /* Could do slightly more clipping: Eg, take the intersection of
@@ -583,15 +586,9 @@ static GLboolean intelTryCopyPixels( GLcontext *ctx,
 				  bx, by, /* dstx, dsty */
 				  bw, bh );
       }
-
-      intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_FALSE);
-      bmFenceBufferList(intel->bm, list);
-
- out:
-      bmFreeBufferList(list);
    }
-
-
+ out:
+   intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_FALSE);
    UNLOCK_HARDWARE( intel );
    return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/i915/intel_screen.c b/src/mesa/drivers/dri/i915/intel_screen.c
index 97d2b801f1e..bd127182e4b 100644
--- a/src/mesa/drivers/dri/i915/intel_screen.c
+++ b/src/mesa/drivers/dri/i915/intel_screen.c
@@ -166,7 +166,7 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv)
 			 
    intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset;
    
-   if (0) intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv);
+   if (1) intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv);
 
    intelScreen->drmMinor = sPriv->drmMinor;
 
diff --git a/src/mesa/drivers/dri/i915/intel_state.c b/src/mesa/drivers/dri/i915/intel_state.c
index a2f2b207388..25a7553fad2 100644
--- a/src/mesa/drivers/dri/i915/intel_state.c
+++ b/src/mesa/drivers/dri/i915/intel_state.c
@@ -192,10 +192,17 @@ static void intelDrawBuffer(GLcontext *ctx, GLenum mode )
    
    intelSetFrontClipRects( intel );
 
+/*    if (intel->draw_region) */
+/*       intel_region_release(intel, intel->draw_region); */
+
    if (front) {
       intel->drawOffset = screen->front.offset;
+/*       intel->draw_region = intel_region_reference(intel->front_region); */
+      intel->draw_region = intel->front_region;
    } else {
       intel->drawOffset = screen->back.offset;
+/*       intel->draw_region = intel_region_reference(intel->back_region); */
+      intel->draw_region = intel->back_region;
    }
 
    intel->vtbl.set_draw_offset( intel, intel->drawOffset );
diff --git a/src/mesa/drivers/dri/i915/intel_tex.h b/src/mesa/drivers/dri/i915/intel_tex.h
index 33d2e279c95..be64cf03934 100644
--- a/src/mesa/drivers/dri/i915/intel_tex.h
+++ b/src/mesa/drivers/dri/i915/intel_tex.h
@@ -111,4 +111,12 @@ GLboolean intel_prevalidate_buffers( struct intel_context *intel );
 GLboolean intel_validate_buffers( struct intel_context *intel );
 void intel_fence_buffers( struct intel_context *intel );
 
+
+void intel_tex_map_images( struct intel_context *intel,
+			   struct intel_texture_object *intelObj );
+
+void intel_tex_unmap_images( struct intel_context *intel,
+			     struct intel_texture_object *intelObj );
+
+
 #endif
diff --git a/src/mesa/drivers/dri/i915/intel_tex_copy.c b/src/mesa/drivers/dri/i915/intel_tex_copy.c
index 127a029d58c..0f202fdc878 100644
--- a/src/mesa/drivers/dri/i915/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/i915/intel_tex_copy.c
@@ -105,9 +105,9 @@ static GLboolean do_copy_texsubimage( struct intel_context *intel,
  
 
    LOCK_HARDWARE(intel);
+   intelInstallBatchBuffer(intel);
    {
       __DRIdrawablePrivate *dPriv = intel->driDrawable;
-      struct bm_buffer_list *list = bmNewBufferList();
       GLuint image_offset = intel_miptree_image_offset(intelImage->mt, 
 						       intelImage->face,
 						       intelImage->level);
@@ -131,9 +131,15 @@ static GLboolean do_copy_texsubimage( struct intel_context *intel,
       y += dPriv->y;
 
 
-      bmAddBuffer(list, intelImage->mt->region->buffer, BM_WRITE, NULL, &dst_offset);
-      bmAddBuffer(list, src->buffer, BM_READ, NULL, &src_offset);
-      if (!bmValidateBufferList(intel->bm, list, BM_MEM_AGP)) {
+      bmAddBuffer(intel->buffer_list, 
+		  intelImage->mt->region->buffer, 
+		  BM_WRITE, NULL, &dst_offset);
+
+      bmAddBuffer(intel->buffer_list, 
+		  src->buffer, 
+		  BM_READ, NULL, &src_offset);
+
+      if (!bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP)) {
 	 ret = GL_FALSE;
 	 goto out;
       }
@@ -146,12 +152,8 @@ static GLboolean do_copy_texsubimage( struct intel_context *intel,
 			       x, y, 
 			       dstx, dsty,
 			       width, height );
-
-      intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_FALSE);
-      bmFenceBufferList(intel->bm, list);
-
    out:
-      bmFreeBufferList(list);
+      intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_FALSE);
    }
    
 
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index fe43a8e3d7d..e92aa3e0c41 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -87,12 +87,6 @@ static void intel_flush_inline_primitive( GLcontext *ctx )
    intel->prim.primitive = ~0;
    intel->prim.start_ptr = 0;
    intel->prim.flush = 0;
-
-   intelFlushBatch(intel, GL_TRUE); 
-   intel_fence_buffers(intel);
-   UNLOCK_HARDWARE(intel);
-
-
 }
 
 
@@ -101,30 +95,7 @@ static void intel_flush_inline_primitive( GLcontext *ctx )
 void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim )
 {
    BATCH_LOCALS;
-
-   if (0)
-      fprintf(stderr, "%s %x\n", __FUNCTION__, prim);
-
-
-   /* Finish any in-progress primitive:
-    */
-   INTEL_FIREVERTICES( intel );
-
-   LOCK_HARDWARE(intel);
-   intel_validate_buffers( intel );
-   intel->vtbl.emit_state( intel );
    
-   /* Make sure there is some space in this buffer:
-    */
-   if (intel->vertex_size * 10 * sizeof(GLuint) >= intel->batch.space)
-      intelFlushBatch(intel, GL_TRUE); 
-
-   if (((int)intel->batch.ptr) & 0x4) {
-      BEGIN_BATCH(1);
-      OUT_BATCH(0);
-      ADVANCE_BATCH();
-   }
-
    /* Emit a slot which will be filled with the inline primitive
     * command later.
     */
@@ -146,6 +117,9 @@ void intelWrapInlinePrimitive( intelContextPtr intel )
 
    intel_flush_inline_primitive( &intel->ctx );
    intelFlushBatch(intel, GL_TRUE);
+   intelInstallBatchBuffer( intel );    
+   intel_validate_buffers( intel );
+   intel->vtbl.emit_state( intel );
    intelStartInlinePrimitive( intel, prim );
 }
 
@@ -859,7 +833,14 @@ static void intelRunPipeline( GLcontext *ctx )
 
 static void intelRenderStart( GLcontext *ctx )
 {
-   INTEL_CONTEXT(ctx)->vtbl.render_start( INTEL_CONTEXT(ctx) );
+   struct intel_context *intel = intel_context(ctx);
+
+   intel->vtbl.render_start( INTEL_CONTEXT(ctx) );
+
+   LOCK_HARDWARE(intel);
+   intelInstallBatchBuffer( intel );    
+   intel_validate_buffers( intel );
+   intel->vtbl.emit_state( intel );
 }
 
 static void intelRenderFinish( GLcontext *ctx )
@@ -871,6 +852,9 @@ static void intelRenderFinish( GLcontext *ctx )
 
    if (intel->prim.flush)
       intel->prim.flush(ctx);
+
+   intelFlushBatch(intel, GL_TRUE); 
+   UNLOCK_HARDWARE(intel);
 }