Merge git://proxy01.pd.intel.com:9419/git/mesa/mesa into crestline

9 files changed, 164 insertions, 197 deletions

diff --git a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
index 32da40661b4..3bc84d862d3 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
@@ -315,6 +315,9 @@ static inline void nv10_render_line(GLcontext *ctx,GLuint v1,GLuint v2)
 	GLuint vertsize = nmesa->vertex_size;
 	GLuint size_dword = vertsize*(2)/4;
 
+	/* OUT_RINGp wants size in DWORDS */
+	vertsize >>= 2;
+
 	nv10ExtendPrimitive(nmesa, size_dword);
 	nv10StartPrimitive(nmesa,GL_LINES+1,size_dword);
 	OUT_RINGp((nouveauVertex*)(vertptr+(v1*vertsize)),vertsize);
@@ -329,6 +332,9 @@ static inline void nv10_render_triangle(GLcontext *ctx,GLuint v1,GLuint v2,GLuin
 	GLuint vertsize = nmesa->vertex_size;
 	GLuint size_dword = vertsize*(3)/4;
 
+	/* OUT_RINGp wants size in DWORDS */
+	vertsize >>= 2;
+
 	nv10ExtendPrimitive(nmesa, size_dword);
 	nv10StartPrimitive(nmesa,GL_TRIANGLES+1,size_dword);
 	OUT_RINGp((nouveauVertex*)(vertptr+(v1*vertsize)),vertsize);
@@ -344,6 +350,9 @@ static inline void nv10_render_quad(GLcontext *ctx,GLuint v1,GLuint v2,GLuint v3
 	GLuint vertsize = nmesa->vertex_size;
 	GLuint size_dword = vertsize*(4)/4;
 
+	/* OUT_RINGp wants size in DWORDS */
+	vertsize >>= 2;
+
 	nv10ExtendPrimitive(nmesa, size_dword);
 	nv10StartPrimitive(nmesa,GL_QUADS+1,size_dword);
 	OUT_RINGp((nouveauVertex*)(vertptr+(v1*vertsize)),vertsize);
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state.c b/src/mesa/drivers/dri/nouveau/nv20_state.c
index 030713c0dbb..3d8d83a865a 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state.c
@@ -78,7 +78,18 @@ static void nv20BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, GLenum dfac
 
 static void nv20Clear(GLcontext *ctx, GLbitfield mask)
 {
-	/* TODO */
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLuint hw_bufs = 0;
+
+	if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT))
+		hw_bufs |= 0xf0;
+	if (mask & (BUFFER_BIT_DEPTH))
+		hw_bufs |= 0x03;
+
+	if (hw_bufs) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLEAR_WHICH_BUFFERS, 1);
+		OUT_RING_CACHE(hw_bufs);
+	}
 }
 
 static void nv20ClearColor(GLcontext *ctx, const GLfloat color[4])
@@ -557,9 +568,11 @@ static void nv20Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
 	   y += nmesa->drawY;
 	}
 
-        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_X2_X1, 2);
-        OUT_RING_CACHE(((x+w-1) << 16) | x);
-        OUT_RING_CACHE(((y+h-1) << 16) | y);
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_X2_X1, 1);
+        OUT_RING_CACHE((w << 16) | x );
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_Y2_Y1, 1);
+        OUT_RING_CACHE((h << 16) | y );
+
 }
 
 /** Select flat or smooth shading */
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index d10a9d87d31..68c8f5b87d0 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -133,7 +133,6 @@ const struct dri_extension stencil_two_side[] = {
 
 extern struct tnl_pipeline_stage _r300_render_stage;
 extern const struct tnl_pipeline_stage _r300_tcl_stage;
-extern const struct tnl_pipeline_stage _r300_texrect_stage;
 
 static const struct tnl_pipeline_stage *r300_pipeline[] = {
 
@@ -164,8 +163,6 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 
 	/* Else do them here.
 	 */
-	/* scale texture rectangle to 0..1. */
-	&_r300_texrect_stage,
 	&_r300_render_stage,
 	&_tnl_render_stage,	/* FALLBACK  */
 	0,
@@ -204,7 +201,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 						     "def_max_anisotropy");
 
 	//r300->texmicrotile = GL_TRUE;
-	
+
 	/* Init default driver functions then plug in our R300-specific functions
 	 * (the texture functions are especially important)
 	 */
@@ -213,7 +210,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	r300InitStateFuncs(&functions);
 	r300InitTextureFuncs(&functions);
 	r300InitShaderFuncs(&functions);
-	
+
 #ifdef USER_BUFFERS
 	radeon_mm_init(r300);
 #endif
@@ -221,7 +218,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	if (hw_tcl_on) {
 		r300_init_vbo_funcs(&functions);
 	}
-#endif	
+#endif
 	if (!radeonInitContext(&r300->radeon, &functions,
 			       glVisual, driContextPriv, sharedContextPrivate)) {
 		FREE(r300);
@@ -264,8 +261,8 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	 * texturable memory at once.
 	 */
 
-	ctx = r300->radeon.glCtx; 
-	
+	ctx = r300->radeon.glCtx;
+
 	ctx->Const.MaxTextureImageUnits = driQueryOptioni(&r300->radeon.optionCache,
 						     "texture_image_units");
 	ctx->Const.MaxTextureCoordUnits = driQueryOptioni(&r300->radeon.optionCache,
@@ -283,7 +280,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	ctx->Const.MinLineWidthAA = 1.0;
 	ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
 	ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
-	
+
 #ifdef USER_BUFFERS
 	/* Needs further modifications */
 #if 0
@@ -337,10 +334,10 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	ctx->_MaintainTexEnvProgram = GL_TRUE;
 
 	driInitExtensions(ctx, card_extensions, GL_TRUE);
-	
+
 	if (driQueryOptionb(&r300->radeon.optionCache, "disable_stencil_two_side") == 0)
 		driInitSingleExtension(ctx, stencil_two_side);
-	
+
 	if (r300->radeon.glCtx->Mesa_DXTn && !driQueryOptionb (&r300->radeon.optionCache, "disable_s3tc")) {
 	  _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
 	  _mesa_enable_extension( ctx, "GL_S3_s3tc" );
@@ -354,7 +351,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	radeonInitSpanFuncs(ctx);
 	r300InitCmdBuf(r300);
 	r300InitState(r300);
-	
+
 #ifdef RADEON_VTXFMT_A
 	radeon_init_vtxfmt_a(r300);
 #endif
@@ -407,9 +404,9 @@ static void r300FreeGartAllocations(r300ContextPtr r300)
 	/* Cannot flush/lock if no context exists. */
 	if (in_use)
 		r300FlushCmdBuf(r300, __FUNCTION__);
-	
+
 	done_age = radeonGetAge((radeonContextPtr)r300);
-	
+
 	for (i = r300->rmm->u_last; i > 0; i--) {
 		if (r300->rmm->u_list[i].ptr == NULL) {
 			continue;
@@ -421,7 +418,7 @@ static void r300FreeGartAllocations(r300ContextPtr r300)
 		}
 
 		assert(r300->rmm->u_list[i].h_pending == 0);
-		
+
 		tries = 0;
 		while(r300->rmm->u_list[i].age > done_age && tries++ < 1000) {
 			usleep(10);
@@ -430,10 +427,10 @@ static void r300FreeGartAllocations(r300ContextPtr r300)
 		if (tries >= 1000) {
 			WARN_ONCE("Failed to idle region!");
 		}
-		
+
 		memfree.region_offset = (char *)r300->rmm->u_list[i].ptr -
 			(char *)r300->radeon.radeonScreen->gartTextures.map;
-		
+
 		ret = drmCommandWrite(r300->radeon.radeonScreen->driScreen->fd,
 				DRM_RADEON_FREE, &memfree, sizeof(memfree));
 		if (ret) {
@@ -442,7 +439,7 @@ static void r300FreeGartAllocations(r300ContextPtr r300)
 		} else {
 			if (i == r300->rmm->u_last)
 				r300->rmm->u_last--;
-			
+
 			r300->rmm->u_list[i].pending = 0;
 			r300->rmm->u_list[i].ptr = NULL;
 			if (r300->rmm->u_list[i].fb) {
@@ -490,12 +487,12 @@ void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
 		_tnl_DestroyContext(r300->radeon.glCtx);
 		_vbo_DestroyContext(r300->radeon.glCtx);
 		_swrast_DestroyContext(r300->radeon.glCtx);
-		
+
 		if (r300->dma.current.buf) {
 			r300ReleaseDmaRegion(r300, &r300->dma.current, __FUNCTION__ );
 #ifndef USER_BUFFERS
 			r300FlushCmdBuf(r300, __FUNCTION__);
-#endif  
+#endif
 		}
 		r300FreeGartAllocations(r300);
 		r300DestroyCmdBuf(r300);
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index fe261dbbc6e..c8b81d98497 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -553,6 +553,7 @@ struct r300_stencilbuffer_state {
 #define VSF_MAX_FRAGMENT_TEMPS (14)
 
 #define STATE_R300_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
+#define STATE_R300_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1)
 
 struct r300_vertex_shader_fragment {
 	int length;
@@ -602,12 +603,6 @@ extern int hw_tcl_on;
 /* Should but doesnt work */
 //#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->curr_vp)
 
-//#define TMU_ENABLED(ctx, unit) (hw_tcl_on ? ctx->Texture.Unit[unit]._ReallyEnabled && (OutputsWritten & (1<<(VERT_RESULT_TEX0+(unit)))) :
-//	(r300->state.render_inputs & (_TNL_BIT_TEX0<<(unit))))
-//#define TMU_ENABLED(ctx, unit) (hw_tcl_on ? ctx->Texture.Unit[unit]._ReallyEnabled && OutputsWritten & (1<<(VERT_RESULT_TEX0+(unit))) :
-//	ctx->Texture.Unit[unit]._ReallyEnabled && r300->state.render_inputs & (_TNL_BIT_TEX0<<(unit)))
-
-#define TMU_ENABLED(ctx, unit) (ctx->Texture.Unit[unit]._ReallyEnabled)
 
 /* r300_vertex_shader_state and r300_vertex_program should probably be merged together someday.
  * Keeping them them seperate for now should ensure fixed pipeline keeps functioning properly.
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c
index 0d7d1f1af29..68a75ec7f00 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.c
@@ -929,13 +929,40 @@ static void emit_tex(struct r300_fragment_program *rp,
 	COMPILE_STATE;
 	GLuint coord = t_src(rp, fpi->SrcReg[0]);
 	GLuint dest = undef, rdest = undef;
-	GLuint din = cs->dest_in_node, uin = cs->used_in_node;
+	GLuint din, uin;
 	int unit = fpi->TexSrcUnit;
 	int hwsrc, hwdest;
+	GLuint tempreg = 0;
 
 	/* Resolve source/dest to hardware registers */
-	hwsrc = t_hw_src(rp, coord, GL_TRUE);
 	if (opcode != R300_FPITX_OP_KIL) {
+		if (fpi->TexSrcTarget == TEXTURE_RECT_INDEX) {
+			/**
+			 * Hardware uses [0..1]x[0..1] range for rectangle textures
+			 * instead of [0..Width]x[0..Height].
+			 * Add a scaling instruction.
+			 *
+			 * \todo Refactor this once we have proper rewriting/optimization
+			 * support for programs.
+			 */
+			GLint tokens[6] = { STATE_INTERNAL, STATE_R300_TEXRECT_FACTOR, 0, 0, 0, 0 };
+			int factor_index;
+			GLuint factorreg;
+
+			tokens[2] = unit;
+			factor_index = _mesa_add_state_reference(rp->mesa_program.Base.Parameters, tokens);
+			factorreg = emit_const4fv(rp,
+					rp->mesa_program.Base.Parameters->ParameterValues[factor_index]);
+			tempreg = keep(get_temp_reg(rp));
+
+			emit_arith(rp, PFS_OP_MAD, tempreg, WRITEMASK_XYZW,
+			           coord, factorreg, pfs_zero, 0);
+
+			hwsrc = t_hw_src(rp, tempreg, GL_TRUE);
+		} else {
+			hwsrc = t_hw_src(rp, coord, GL_TRUE);
+		}
+
 		dest = t_dst(rp, fpi->DstReg);
 
 		/* r300 doesn't seem to be able to do TEX->output reg */
@@ -956,8 +983,12 @@ static void emit_tex(struct r300_fragment_program *rp,
 	} else {
 		hwdest = 0;
 		unit = 0;
+		hwsrc = t_hw_src(rp, coord, GL_TRUE);
 	}
 
+	din = cs->dest_in_node;
+	uin = cs->used_in_node;
+
 	/* Indirection if source has been written in this node, or if the
 	 * dest has been read/written in this node
 	 */
@@ -1009,6 +1040,10 @@ static void emit_tex(struct r300_fragment_program *rp,
 			   pfs_one, pfs_zero, 0);
 		free_temp(rp, dest);
 	}
+
+	/* Free temp register */
+	if (tempreg != 0)
+		free_temp(rp, tempreg);
 }
 
 
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index 1f4a2d2e647..1f65f9a7e7e 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -325,7 +325,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * Most likely this is used to ignore rest of the program in cases
  * where group of verts arent visible. For some reason this "section"
  * is sometimes accepted other instruction that have no relationship with
- *position calculations. 
+ *position calculations.
  */
 #define R300_VAP_PVS_CNTL_1                 0x22D0
 #       define R300_PVS_CNTL_1_PROGRAM_START_SHIFT   0
@@ -566,8 +566,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_RE_FOG_SCALE                     0x4294
 #define R300_RE_FOG_START                     0x4298
 
-/* Not sure why there are duplicate of factor and constant values. 
- * My best guess so far is that there are seperate zbiases for test and write. 
+/* Not sure why there are duplicate of factor and constant values.
+ * My best guess so far is that there are seperate zbiases for test and write.
  * Ordering might be wrong.
  * Some of the tests indicate that fgl has a fallback implementation of zbias
  * via pixel shaders.
@@ -909,7 +909,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* 32 bit chroma key */
 #define R300_TX_CHROMA_KEY_0                      0x4580
 /* ff00ff00 == { 0, 1.0, 0, 1.0 } */
-#define R300_TX_BORDER_COLOR_0              0x45C0 
+#define R300_TX_BORDER_COLOR_0              0x45C0
 
 /* END: Texture specification */
 
@@ -999,6 +999,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #		define R300_FPITX_OP_KIL	2
 #		define R300_FPITX_OP_TXP	3
 #		define R300_FPITX_OP_TXB	4
+#	define R300_FPITX_OPCODE_MASK           (7 << 15)
 
 /* ALU
  * The ALU instructions register blocks are enumerated according to the order
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 0864558e8dc..63b21b9379a 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -202,23 +202,23 @@ static void inline fire_EB(r300ContextPtr rmesa, unsigned long addr, int vertex_
 	unsigned long t_addr;
 	unsigned long magic_1, magic_2;
 	GLcontext *ctx;
-	ctx = rmesa->radeon.glCtx; 
-	
+	ctx = rmesa->radeon.glCtx;
+
 	assert(elt_size == 2 || elt_size == 4);
-	
+
 	if(addr & (elt_size-1)){
 		WARN_ONCE("Badly aligned buffer\n");
 		return ;
 	}
 #ifdef OPTIMIZE_ELTS
 	addr_a = 0;
-	
+
 	magic_1 = (addr % 32) / 4;
 	t_addr = addr & (~0x1d);
 	magic_2 = (vertex_count + 1 + (t_addr & 0x2)) / 2 + magic_1;
-	
+
 	check_space(6);
-	
+
 	start_packet3(RADEON_CP_PACKET3_3D_DRAW_INDX_2, 0);
 	if(elt_size == 4){
 		e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count<<16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
@@ -234,7 +234,7 @@ static void inline fire_EB(r300ContextPtr rmesa, unsigned long addr, int vertex_
 		e32(R300_EB_UNK1 | (magic_1 << 16) | R300_EB_UNK2);
 		e32(t_addr);
 	}
-	
+
 	if(elt_size == 4){
 		e32(vertex_count /*+ addr_a/4*/); /* Total number of dwords needed? */
 	} else {
@@ -249,11 +249,11 @@ static void inline fire_EB(r300ContextPtr rmesa, unsigned long addr, int vertex_
 #endif
 #else
 	(void)magic_2, (void)magic_1, (void)t_addr;
-	
+
 	addr_a = 0;
-	
+
 	check_space(6);
-	
+
 	start_packet3(RADEON_CP_PACKET3_3D_DRAW_INDX_2, 0);
 	if(elt_size == 4){
 		e32(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (vertex_count<<16) | type | R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
@@ -264,14 +264,14 @@ static void inline fire_EB(r300ContextPtr rmesa, unsigned long addr, int vertex_
 	start_packet3(RADEON_CP_PACKET3_INDX_BUFFER, 2);
 	e32(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
 	e32(addr /*& 0xffffffe3*/);
-	
+
 	if(elt_size == 4){
 		e32(vertex_count /*+ addr_a/4*/); /* Total number of dwords needed? */
 	} else {
 		e32((vertex_count+1)/2 /*+ addr_a/4*/); /* Total number of dwords needed? */
 	}
 	//cp_delay(rmesa, 1);
-#endif	
+#endif
 }
 
 static void r300_render_vb_primitive(r300ContextPtr rmesa,
@@ -303,12 +303,12 @@ static void r300_render_vb_primitive(r300ContextPtr rmesa,
 		//e32(rmesa->state.Elts[start]);
 		return;
 	}
-	
+
 	if(num_verts > 65535){ /* not implemented yet */
 		WARN_ONCE("Too many elts\n");
 		return;
 	}
-	
+
 	r300EmitElts(ctx, rmesa->state.VB.Elts, num_verts, rmesa->state.VB.elt_size);
 	fire_EB(rmesa, rmesa->state.elt_dma.aos_offset, num_verts, type, rmesa->state.VB.elt_size);
 #endif
@@ -328,7 +328,7 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
 	int cmd_written = 0;
 	drm_radeon_cmd_header_t *cmd = NULL;
 
-   
+
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
@@ -336,26 +336,26 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
  		TNLcontext *tnl = TNL_CONTEXT(ctx);
 		radeon_vb_to_rvb(rmesa, VB, &tnl->vb);
 	}
-	
+
 	r300UpdateShaders(rmesa);
 	if (r300EmitArrays(ctx))
 		return GL_TRUE;
 
 	r300UpdateShaderStates(rmesa);
-	
+
 	reg_start(R300_RB3D_DSTCACHE_CTLSTAT,0);
 	e32(R300_RB3D_DSTCACHE_UNKNOWN_0A);
 
 	reg_start(R300_RB3D_ZCACHE_CTLSTAT,0);
 	e32(R300_RB3D_ZCACHE_UNKNOWN_03);
-	
+
 	r300EmitState(rmesa);
-	
+
 	for(i=0; i < VB->PrimitiveCount; i++){
 		GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 		GLuint start = VB->Primitive[i].start;
 		GLuint length = VB->Primitive[i].count;
-		
+
 		r300_render_vb_primitive(rmesa, ctx, start, start + length, prim);
 	}
 
@@ -453,11 +453,6 @@ int r300Fallback(GLcontext *ctx)
 		/* GL_POINT_SPRITE_NV */
 		FALLBACK_IF(ctx->Point.PointSprite);
 
-	/* Fallback for rectangular texture */
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-		if (ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_RECT_BIT)
-			return R300_FALLBACK_TCL;
-
 	return R300_FALLBACK_NONE;
 }
 
@@ -494,19 +489,19 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program *vp;
-   
+
    	hw_tcl_on=future_hw_tcl_on;
-   
+
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 	if(hw_tcl_on == GL_FALSE)
 		return GL_TRUE;
-	
+
 	if (r300Fallback(ctx) >= R300_FALLBACK_TCL) {
 		hw_tcl_on = GL_FALSE;
 		return GL_TRUE;
 	}
-	
+
 	r300UpdateShaders(rmesa);
 
 	vp = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
@@ -520,13 +515,13 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	struct tnl_cache *cache;
 	struct tnl_cache_item *c;
-	
+
 	cache = tnl->vp_cache;
 	c = cache->items[0xc000cc0e % cache->size];
-	
+
 	if(c && c->data == vp)
 		vp->native = GL_FALSE;
-	
+
 #endif
 #if 0
 	vp->native = GL_FALSE;
@@ -536,7 +531,7 @@ static GLboolean r300_run_tcl_render(GLcontext *ctx,
 		return GL_TRUE;
 	}
 	//r300UpdateShaderStates(rmesa);
-	
+
 	return r300_run_vb_render(ctx, stage);
 }
 
@@ -549,107 +544,3 @@ const struct tnl_pipeline_stage _r300_tcl_stage = {
 	r300_run_tcl_render	/* run */
 };
 
-/* R300 texture rectangle expects coords in 0..1 range, not 0..dimension
- * as in the extension spec.  Need to translate here.
- *
- * Note that swrast expects 0..dimension, so if a fallback is active,
- * don't do anything.  (Maybe need to configure swrast to match hw)
- */
-struct texrect_stage_data {
-   GLvector4f texcoord[MAX_TEXTURE_UNITS];
-};
-
-#define TEXRECT_STAGE_DATA(stage) ((struct texrect_stage_data *)stage->privatePtr)
-
-
-static GLboolean run_texrect_stage( GLcontext *ctx,
-				    struct tnl_pipeline_stage *stage )
-{
-   struct texrect_stage_data *store = TEXRECT_STAGE_DATA(stage);
-   r300ContextPtr rmesa = R300_CONTEXT(ctx);
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   GLuint i;
-
-   if (rmesa->radeon.Fallback)
-      return GL_TRUE;
-
-   for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++) {
-      if (ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_RECT_BIT) {
-	 struct gl_texture_object *texObj = ctx->Texture.Unit[i].CurrentRect;
-	 struct gl_texture_image *texImage = texObj->Image[0][texObj->BaseLevel];
-	 const GLfloat iw = 1.0/texImage->Width;
-	 const GLfloat ih = 1.0/texImage->Height;
-	 GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data;
-	 GLint instride = VB->TexCoordPtr[i]->stride;
-	 GLfloat (*out)[4] = store->texcoord[i].data;
-	 GLint j;
-
-	 store->texcoord[i].size = VB->TexCoordPtr[i]->size;
-	 for (j = 0 ; j < VB->Count ; j++) {
-	    switch (VB->TexCoordPtr[i]->size) {
-	    case 4:
-	       out[j][3] = in[3];
-	    /* fallthrough */
-	    case 3:
-	       out[j][2] = in[2];
-	    /* fallthrough */
-	    default:
-	       out[j][0] = in[0] * iw;
-	       out[j][1] = in[1] * ih;
-	    }
-	    in = (GLfloat *)((GLubyte *)in + instride);
-	 }
-
-	 VB->AttribPtr[VERT_ATTRIB_TEX0+i] = VB->TexCoordPtr[i] = &store->texcoord[i];
-      }
-   }
-
-   return GL_TRUE;
-}
-
-
-/* Called the first time stage->run() is invoked.
- */
-static GLboolean alloc_texrect_data( GLcontext *ctx,
-				     struct tnl_pipeline_stage *stage )
-{
-   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
-   struct texrect_stage_data *store;
-   GLuint i;
-
-   stage->privatePtr = CALLOC(sizeof(*store));
-   store = TEXRECT_STAGE_DATA(stage);
-   if (!store)
-      return GL_FALSE;
-
-   for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++)
-      _mesa_vector4f_alloc( &store->texcoord[i], 0, VB->Size, 32 );
-
-   return GL_TRUE;
-}
-
-static void free_texrect_data( struct tnl_pipeline_stage *stage )
-{
-   struct texrect_stage_data *store = TEXRECT_STAGE_DATA(stage);
-   GLuint i;
-
-   if (store) {
-      for (i = 0 ; i < MAX_TEXTURE_UNITS ; i++)
-	 if (store->texcoord[i].data)
-	    _mesa_vector4f_free( &store->texcoord[i] );
-      FREE( store );
-      stage->privatePtr = NULL;
-   }
-}
-
-const struct tnl_pipeline_stage _r300_texrect_stage =
-{
-   "r300 texrect stage",			/* name */
-   NULL,
-   alloc_texrect_data,
-   free_texrect_data,
-   NULL,
-   run_texrect_stage
-};
-
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 47f1546c07e..17658efdb2f 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -1058,23 +1058,41 @@ r300UpdateDrawBuffer(GLcontext *ctx)
 static void r300FetchStateParameter(GLcontext *ctx, const enum state_index state[],
                   GLfloat *value)
 {
-    r300ContextPtr r300 = R300_CONTEXT(ctx);
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
-    switch(state[0])
-    {
-    case STATE_INTERNAL:
-    	switch(state[1])
-	{
-	case STATE_R300_WINDOW_DIMENSION:
-	    value[0] = r300->radeon.dri.drawable->w*0.5f;/* width*0.5 */
-    	    value[1] = r300->radeon.dri.drawable->h*0.5f;/* height*0.5 */
-	    value[2] = 0.5F; 				/* for moving range [-1 1] -> [0 1] */
-    	    value[3] = 1.0F; 				/* not used */
-	    break;
-	default:;
+	switch(state[0]) {
+	case STATE_INTERNAL:
+		switch(state[1]) {
+		case STATE_R300_WINDOW_DIMENSION:
+			value[0] = r300->radeon.dri.drawable->w*0.5f;/* width*0.5 */
+			value[1] = r300->radeon.dri.drawable->h*0.5f;/* height*0.5 */
+			value[2] = 0.5F; 				/* for moving range [-1 1] -> [0 1] */
+			value[3] = 1.0F; 				/* not used */
+			break;
+
+		case STATE_R300_TEXRECT_FACTOR: {
+			struct gl_texture_object* t = ctx->Texture.Unit[state[2]].CurrentRect;
+
+			if (t && t->Image[0][t->BaseLevel]) {
+				struct gl_texture_image* image = t->Image[0][t->BaseLevel];
+				value[0] = 1.0 / image->Width2;
+				value[1] = 1.0 / image->Height2;
+			} else {
+				value[0] = 1.0;
+				value[1] = 1.0;
+			}
+			value[2] = 1.0;
+			value[3] = 1.0;
+			break; }
+
+		default:
+			break;
+		}
+		break;
+
+	default:
+		break;
 	}
-    default:;
-    }
 }
 
 /**
@@ -1210,7 +1228,7 @@ void r300_setup_textures(GLcontext *ctx)
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int hw_tmu=0;
 	int last_hw_tmu=-1; /* -1 translates into no setup costs for fields */
-	int tmu_mappings[R300_MAX_TEXTURE_UNITS] = { -1 };
+	int tmu_mappings[R300_MAX_TEXTURE_UNITS] = { -1, };
 	struct r300_fragment_program *rp =
 		(struct r300_fragment_program *)
 		(char *)ctx->FragmentProgram._Current;
@@ -1239,7 +1257,7 @@ void r300_setup_textures(GLcontext *ctx)
 
 	/* We cannot let disabled tmu offsets pass DRM */
 	for(i=0; i < mtu; i++) {
-		if(TMU_ENABLED(ctx, i)) {
+		if (ctx->Texture.Unit[i]._ReallyEnabled) {
 
 #if 0 /* Enables old behaviour */
 			hw_tmu = i;
@@ -1299,6 +1317,7 @@ void r300_setup_textures(GLcontext *ctx)
 
 	for(i = 0; i < rp->tex.length; i++){
 		int unit;
+		int opcode;
 		unsigned long val;
 
 		unit = rp->tex.inst[i] >> R300_FPITX_IMAGE_SHIFT;
@@ -1307,10 +1326,19 @@ void r300_setup_textures(GLcontext *ctx)
 		val = rp->tex.inst[i];
 		val &= ~R300_FPITX_IMAGE_MASK;
 
-		assert(tmu_mappings[unit] >= 0);
-
-		val |= tmu_mappings[unit] << R300_FPITX_IMAGE_SHIFT;
-		r300->hw.fpt.cmd[R300_FPT_INSTR_0+i] = val;
+		opcode = (val & R300_FPITX_OPCODE_MASK) >> R300_FPITX_OPCODE_SHIFT;
+		if (opcode == R300_FPITX_OP_KIL) {
+			r300->hw.fpt.cmd[R300_FPT_INSTR_0+i] = val;
+		} else {
+			if (tmu_mappings[unit] >= 0) {
+				val |= tmu_mappings[unit] << R300_FPITX_IMAGE_SHIFT;
+				r300->hw.fpt.cmd[R300_FPT_INSTR_0+i] = val;
+			} else {
+				// We get here when the corresponding texture image is incomplete
+				// (e.g. incomplete mipmaps etc.)
+				r300->hw.fpt.cmd[R300_FPT_INSTR_0+i] = val;
+			}
+		}
 	}
 
 	r300->hw.fpt.cmd[R300_FPT_CMD_0] = cmdpacket0(R300_PFS_TEXI_0, rp->tex.length);
diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c
index 1c9a098a2d5..f5f04b05f5a 100644
--- a/src/mesa/swrast/s_context.c
+++ b/src/mesa/swrast/s_context.c
@@ -505,9 +505,7 @@ _swrast_update_texture_samplers(GLcontext *ctx)
 
    for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {
       const struct gl_texture_object *tObj = ctx->Texture.Unit[u]._Current;
-      if (tObj)
-         swrast->TextureSample[u] =
-            _swrast_choose_texture_sample_func(ctx, tObj);
+      swrast->TextureSample[u] = _swrast_choose_texture_sample_func(ctx, tObj);
    }
 }