Bump to latest mesa in Debian

15 files changed, 469 insertions, 134 deletions

diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 3158ac8412b..5abb91d9876 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -5,7 +5,6 @@ TOP = ../../../../..
 include $(TOP)/configs/current
 
 LIBNAME = r300_dri.so
-DEFINES += -DCOMPILE_R300 -DGLX_DIRECT_RENDERING -DR200_MERGED=0
 
 MINIGLX_SOURCES = server/radeon_dri.c
 
@@ -70,7 +69,8 @@ C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 X86_SOURCES = 
 #r200_vtxtmp_x86.S
 
-DEFINES += -DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R300
+DRIVER_DEFINES = -DCOMPILE_R300 -DGLX_DIRECT_RENDERING -DR200_MERGED=0 \
+	-DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R300
 
 SYMLINKS = \
 	server/radeon_dri.c \
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index 09f7669bd71..f39f71584ce 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -337,8 +337,10 @@ void r300InitCmdBuf(r300ContextPtr r300)
 		r300->hw.unk4260.cmd[0] = cmdpacket0(0x4260, 3);
 	ALLOC_STATE( unk4274, always, 5, "unk4274", 0 );
 		r300->hw.unk4274.cmd[0] = cmdpacket0(0x4274, 4);
-	ALLOC_STATE( unk4288, always, 6, "unk4288", 0 );
-		r300->hw.unk4288.cmd[0] = cmdpacket0(0x4288, 5);
+	ALLOC_STATE( unk4288, always, 4, "unk4288", 0 );
+		r300->hw.unk4288.cmd[0] = cmdpacket0(0x4288, 3);
+	ALLOC_STATE( fogp, always, 3, "fogp", 0 );
+		r300->hw.fogp.cmd[0] = cmdpacket0(R300_RE_FOG_SCALE, 2);
 	ALLOC_STATE( unk42A0, always, 2, "unk42A0", 0 );
 		r300->hw.unk42A0.cmd[0] = cmdpacket0(0x42A0, 1);
 	ALLOC_STATE( zbs, always, R300_ZBS_CMDSIZE, "zbs", 0 );
@@ -374,10 +376,10 @@ void r300InitCmdBuf(r300ContextPtr r300)
 		r300->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_PFS_INSTR2_0, 1);
 	ALLOC_STATE( fpi[3], variable, R300_FPI_CMDSIZE, "fpi/3", 3 );
 		r300->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_PFS_INSTR3_0, 1);
-	ALLOC_STATE( unk4BC0, always, 2, "unk4BC0", 0 );
-		r300->hw.unk4BC0.cmd[0] = cmdpacket0(0x4BC0, 1);
-	ALLOC_STATE( unk4BC8, always, 4, "unk4BC8", 0 );
-		r300->hw.unk4BC8.cmd[0] = cmdpacket0(0x4BC8, 3);
+	ALLOC_STATE( fogs, always, R300_FOGS_CMDSIZE, "fogs", 0 );
+		r300->hw.fogs.cmd[R300_FOGS_CMD_0] = cmdpacket0(R300_RE_FOG_STATE, 1);
+	ALLOC_STATE( fogc, always, R300_FOGC_CMDSIZE, "fogc", 0 );
+		r300->hw.fogc.cmd[R300_FOGC_CMD_0] = cmdpacket0(R300_FOG_COLOR_R, 3);
 	ALLOC_STATE( at, always, R300_AT_CMDSIZE, "at", 0 );
 		r300->hw.at.cmd[R300_AT_CMD_0] = cmdpacket0(R300_PP_ALPHA_TEST, 2);
 	ALLOC_STATE( unk4BD8, always, 2, "unk4BD8", 0 );
@@ -478,6 +480,7 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4260);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4274);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4288);
+	insert_at_tail(&r300->hw.atomlist, &r300->hw.fogp);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42A0);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.zbs);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42B4);
@@ -495,8 +498,8 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[1]);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[2]);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[3]);
-	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BC0);
-	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BC8);
+	insert_at_tail(&r300->hw.atomlist, &r300->hw.fogs);
+	insert_at_tail(&r300->hw.atomlist, &r300->hw.fogc);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.at);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BD8);
 	insert_at_tail(&r300->hw.atomlist, &r300->hw.fpp);
@@ -541,9 +544,9 @@ void r300InitCmdBuf(r300ContextPtr r300)
 		size = 64*256;
 
 	if (RADEON_DEBUG & (DEBUG_IOCTL|DEBUG_DMA)) {
-		fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%d\n",
+		fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%ld\n",
 			sizeof(drm_r300_cmd_header_t));
-		fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%d\n",
+		fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%ld\n",
 			sizeof(drm_radeon_cmd_buffer_t));
 		fprintf(stderr,
 			"Allocating %d bytes command buffer (max state is %d bytes)\n",
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 4943d549973..cadb27ba8b9 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -372,6 +372,61 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	return GL_TRUE;
 }
 
+static void r300FreeGartAllocations(r300ContextPtr r300)
+{
+	int i, ret, tries=0, done_age;
+	drm_radeon_mem_free_t memfree;
+
+	memfree.region = RADEON_MEM_REGION_GART;
+
+#ifdef USER_BUFFERS
+	done_age = radeonGetAge((radeonContextPtr)r300);
+	
+	for (i = r300->rmm->u_last; i > 0; i--) {
+		if (r300->rmm->u_list[i].ptr == NULL) {
+			continue;
+		}
+		
+		assert(r300->rmm->u_list[i].pending);
+		assert(r300->rmm->u_list[i].h_pending == 0);
+		
+		tries = 0;
+		while(r300->rmm->u_list[i].age > done_age && tries++ < 1000) {
+			usleep(10);
+			done_age = radeonGetAge((radeonContextPtr)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) {
+			fprintf(stderr, "Failed to free at %p\nret = %s\n",
+				r300->rmm->u_list[i].ptr, strerror(-ret));
+		} 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) {
+				LOCK_HARDWARE(&(r300->radeon));
+				ret = mmFreeMem(r300->rmm->u_list[i].fb);
+				UNLOCK_HARDWARE(&(r300->radeon));
+				if (ret) fprintf(stderr, "failed to free!\n");
+				r300->rmm->u_list[i].fb = NULL;
+			}
+			r300->rmm->u_list[i].ref_count = 0;
+		}
+	}
+    r300->rmm->u_head = i;
+#endif /* USER_BUFFERS */
+}
+
 /* Destroy the device specific context.
  */
 void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
@@ -403,12 +458,11 @@ void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
 		_ac_DestroyContext(r300->radeon.glCtx);
 		_swrast_DestroyContext(r300->radeon.glCtx);
 		
-		r300ReleaseArrays(r300->radeon.glCtx);
 		if (r300->dma.current.buf) {
 			r300ReleaseDmaRegion(r300, &r300->dma.current, __FUNCTION__ );
-			r300FlushCmdBuf(r300, __FUNCTION__ );
 		}
-				
+		r300FlushCmdBuf(r300, __FUNCTION__);
+		r300FreeGartAllocations(r300);
 		r300DestroyCmdBuf(r300);
 
 		if (radeon->state.scissor.pClipRects) {
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 176341d7715..7ff805fd5d3 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -47,11 +47,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "colormac.h"
 #include "radeon_context.h"
 
-/* PPC doesnt support 16 bit elts ... */
-#ifndef MESA_BIG_ENDIAN
 #define USER_BUFFERS
 #define RADEON_VTXFMT_A
 #define HW_VBOS
+
+/* We don't handle 16 bits elts swapping yet */
+#ifdef MESA_BIG_ENDIAN
+#define FORCE_32BITS_ELTS
 #endif
 
 //#define OPTIMIZE_ELTS
@@ -346,6 +348,21 @@ struct r300_state_atom {
 #define R300_FPP_PARAM_0	1
 #define R300_FPP_CMDSIZE	(32*4+1)
 
+#define R300_FOGS_CMD_0		0
+#define R300_FOGS_STATE		1
+#define R300_FOGS_CMDSIZE	2
+
+#define R300_FOGC_CMD_0		0
+#define R300_FOGC_R		1
+#define R300_FOGC_G		2
+#define R300_FOGC_B		3
+#define R300_FOGC_CMDSIZE	4
+
+#define R300_FOGP_CMD_0		0
+#define R300_FOGP_SCALE		1
+#define R300_FOGP_START		2
+#define R300_FOGP_CMDSIZE	3
+
 #define R300_AT_CMD_0		0
 #define R300_AT_ALPHA_TEST	1
 #define R300_AT_UNKNOWN		2
@@ -432,6 +449,8 @@ struct r300_hw_state {
 	struct r300_state_atom unk4260; /* (4260) */
 	struct r300_state_atom unk4274; /* (4274) */
 	struct r300_state_atom unk4288; /* (4288) */
+	struct r300_state_atom fogp;	/* fog parameters (4294) */
+	struct r300_state_atom unk429C; /* (429C) */
 	struct r300_state_atom unk42A0;	/* (42A0) */
 	struct r300_state_atom zbs;	/* zbias (42A4) */
 	struct r300_state_atom unk42B4; /* (42B4) */
@@ -446,8 +465,8 @@ struct r300_hw_state {
 	struct r300_state_atom fpt;     /* texi - (4620) */
 	struct r300_state_atom unk46A4;	/* (46A4) */
 	struct r300_state_atom fpi[4];	/* fp instructions (46C0/47C0/48C0/49C0) */
-	struct r300_state_atom unk4BC0;	/* (4BC0) */
-	struct r300_state_atom unk4BC8;	/* (4BC8) */
+	struct r300_state_atom fogs;	/* fog state (4BC0) */
+	struct r300_state_atom fogc;	/* fog color (4BC8) */
 	struct r300_state_atom at;	/* alpha test (4BD4) */
 	struct r300_state_atom unk4BD8;	/* (4BD8) */
 	struct r300_state_atom fpp;     /* 0x4C00 and following */
@@ -532,7 +551,8 @@ struct r300_vap_reg_state {
 /* Vertex shader state */
 
 /* Perhaps more if we store programs in vmem? */
-#define VSF_MAX_FRAGMENT_LENGTH (256*4)
+/* drm_r300_cmd_header_t->vpu->count is unsigned char */
+#define VSF_MAX_FRAGMENT_LENGTH (255*4)
 	
 /* Can be tested with colormat currently. */
 #define VSF_MAX_FRAGMENT_TEMPS (14)
@@ -759,7 +779,7 @@ struct r300_state {
 	GLuint *Elts;
 	struct r300_dma_region elt_dma;
 	
-	GLuint render_inputs; /* actual render inputs that R300 was configured for. 
+	DECLARE_RENDERINPUTS(render_inputs_bitset); /* actual render inputs that R300 was configured for. 
 				 They are the same as tnl->render_inputs for fixed pipeline */	
 	
 	struct {
@@ -810,6 +830,7 @@ struct r300_context {
 #endif
 
 	GLboolean texmicrotile;
+	GLboolean span_dlocking;
 };
 
 struct r300_buffer_object {
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c
index db509059dd7..e045f0c6a1e 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.c
@@ -1039,7 +1039,7 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 	const struct prog_instruction *inst = mp->Base.Instructions;
 	struct prog_instruction *fpi;
 	pfs_reg_t src[3], dest, temp;
-	int flags, mask;
+	int flags, mask = 0;
 
 	if (!inst || inst[0].Opcode == OPCODE_END) {
 		ERROR("empty program?\n");
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index 0115d622a99..159285962d2 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -380,10 +380,16 @@ static void r300EmitClearState(GLcontext * ctx)
 	int i;
 	LOCAL_VARS;
 	
+	
 	R300_STATECHANGE(r300, vir[0]);
 	reg_start(R300_VAP_INPUT_ROUTE_0_0, 0);
 	e32(0x21030003);
 	
+	/* disable fog */
+	R300_STATECHANGE(r300, fogs);
+	reg_start(R300_RE_FOG_STATE, 0);
+	e32(0x0);
+	
 	R300_STATECHANGE(r300, vir[1]);
 	reg_start(R300_VAP_INPUT_ROUTE_1_0, 0);
 	e32(0xF688F688);
@@ -553,7 +559,7 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask, GLboolean all,
 
 #ifdef CB_DPATH
 	/* Make sure it fits there. */
-	r300EnsureCmdBufSpace(r300, 419*3, __FUNCTION__);
+	r300EnsureCmdBufSpace(r300, 421*3, __FUNCTION__);
 	if(flags || bits)
 		r300EmitClearState(ctx);
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_maos.c b/src/mesa/drivers/dri/r300/r300_maos.c
index 1aa005c7207..6b8365e6d9e 100644
--- a/src/mesa/drivers/dri/r300/r300_maos.c
+++ b/src/mesa/drivers/dri/r300/r300_maos.c
@@ -252,6 +252,43 @@ void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts, int elt_siz
 	memcpy(out, elts, n_elts * elt_size);
 }
 
+	/* Mesa assumes that all missing components are from (0, 0, 0, 1) */
+#define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<<R300_INPUT_ROUTE_X_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_Y<<R300_INPUT_ROUTE_Y_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_Z<<R300_INPUT_ROUTE_Z_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_W<<R300_INPUT_ROUTE_W_SHIFT))
+
+#define ALL_DEFAULT ((R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_X_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Y_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Z_SHIFT) \
+		| (R300_INPUT_ROUTE_SELECT_ONE<<R300_INPUT_ROUTE_W_SHIFT))
+
+
+static GLuint t_comps(GLuint aos_size)
+{
+	GLuint mask;
+	mask = (1 << (aos_size*3)) - 1;
+	return (ALL_COMPONENTS & mask) | (ALL_DEFAULT & ~mask);
+}
+
+static GLuint fix_comps(GLuint dw, int fmt)
+{	
+#ifdef MESA_BIG_ENDIAN
+	if (fmt == 2) {
+		GLuint dw_temp = 0;
+
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_X_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_W_SHIFT;
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_Y_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_Z_SHIFT;
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_Z_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_Y_SHIFT;
+		dw_temp |= ((dw >> R300_INPUT_ROUTE_W_SHIFT) & R300_INPUT_ROUTE_SELECT_MASK) << R300_INPUT_ROUTE_X_SHIFT;
+		
+		return dw_temp;
+	}
+#endif /* MESA_BIG_ENDIAN */
+	return dw;
+		
+}
+
 /* Emit vertex data to GART memory (unless immediate mode)
  * Route inputs to the vertex processor
  */
@@ -264,12 +301,14 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 	//struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
 	GLuint nr = 0;
 	GLuint count = VB->Count;
-	GLuint dw,mask;
+	GLuint dw;
 	GLuint vic_1 = 0;	/* R300_VAP_INPUT_CNTL_1 */
 	GLuint aa_vap_reg = 0; /* VAP register assignment */
 	GLuint i;
-	GLuint inputs = 0;
-	
+	DECLARE_RENDERINPUTS(inputs_bitset);
+
+	RENDERINPUTS_ZERO( inputs_bitset );
+
 #define CONFIGURE_AOS(r, f, v, sz, cn) { \
 		if (RADEON_DEBUG & DEBUG_STATE) \
 			fprintf(stderr, "Enabling "#v "\n"); \
@@ -300,23 +339,23 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 		GLuint InputsRead = CURRENT_VERTEX_SHADER(ctx)->Base.InputsRead;
 		struct r300_vertex_program *prog=(struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
 		if (InputsRead & (1<<VERT_ATTRIB_POS)) {
-			inputs |= _TNL_BIT_POS;
+			RENDERINPUTS_SET( inputs_bitset, _TNL_ATTRIB_POS );
 			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_POS];
 		}
 		if (InputsRead & (1<<VERT_ATTRIB_NORMAL)) {
-			inputs |= _TNL_BIT_NORMAL;
+			RENDERINPUTS_SET( inputs_bitset, _TNL_ATTRIB_NORMAL );
 			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_NORMAL];
 		}
 		if (InputsRead & (1<<VERT_ATTRIB_COLOR0)) {
-			inputs |= _TNL_BIT_COLOR0;
+			RENDERINPUTS_SET( inputs_bitset, _TNL_ATTRIB_COLOR0 );
 			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_COLOR0];
 		}
 		if (InputsRead & (1<<VERT_ATTRIB_COLOR1)) {
-			inputs |= _TNL_BIT_COLOR1;
+			RENDERINPUTS_SET( inputs_bitset, _TNL_ATTRIB_COLOR1 );
 			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_COLOR1];
 		}
 		if (InputsRead & (1<<VERT_ATTRIB_FOG)) {
-			inputs |= _TNL_BIT_FOG;
+			RENDERINPUTS_SET( inputs_bitset, _TNL_ATTRIB_FOG );
 			rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_FOG];
 		}
 		if(ctx->Const.MaxTextureUnits > 8) { /* Not sure if this can even happen... */
@@ -325,17 +364,17 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 		}
 		for (i=0;i<ctx->Const.MaxTextureUnits;i++) {
 			if (InputsRead & (1<<(VERT_ATTRIB_TEX0+i))) {
-				inputs |= _TNL_BIT_TEX0<<i;
+				RENDERINPUTS_SET( inputs_bitset, _TNL_ATTRIB_TEX(i) );
 				rmesa->state.aos[nr++].aos_reg = prog->inputs[VERT_ATTRIB_TEX0+i];
 			}
 		}
 		nr = 0;
 	} else {
-		inputs = TNL_CONTEXT(ctx)->render_inputs;
+		RENDERINPUTS_COPY( inputs_bitset, TNL_CONTEXT(ctx)->render_inputs_bitset );
 	}
-	rmesa->state.render_inputs = inputs;
+	RENDERINPUTS_COPY( rmesa->state.render_inputs_bitset, inputs_bitset );
 
-	if (inputs & _TNL_BIT_POS) {
+	if (RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_POS )) {
 		CONFIGURE_AOS(i_coords,	AOS_FORMAT_FLOAT,
 						VB->AttribPtr[VERT_ATTRIB_POS],
 						immd ? 4 : VB->AttribPtr[VERT_ATTRIB_POS].size,
@@ -344,7 +383,7 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 		vic_1 |= R300_INPUT_CNTL_POS;
 	}
 
-	if (inputs & _TNL_BIT_NORMAL) {
+	if (RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_NORMAL )) {
 		CONFIGURE_AOS(i_normal,	AOS_FORMAT_FLOAT,
 						VB->AttribPtr[VERT_ATTRIB_NORMAL],
 						immd ? 4 : VB->AttribPtr[VERT_ATTRIB_NORMAL].size,
@@ -353,7 +392,7 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 		vic_1 |= R300_INPUT_CNTL_NORMAL;
 	}
 
-	if (inputs & _TNL_BIT_COLOR0) {
+	if (RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_COLOR0 )) {
 		int emitsize=4;
 
 		if (!immd) {
@@ -376,7 +415,7 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 		vic_1 |= R300_INPUT_CNTL_COLOR;
 	}
 
-	if (inputs & _TNL_BIT_COLOR1) {
+	if (RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_COLOR1 )) {
 		int emitsize=4;
 
 		if (!immd) {
@@ -398,7 +437,7 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 	}
 
 #if 0
-	if (inputs & _TNL_BIT_FOG) {
+	if (RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_FOG )) {
 		CONFIGURE_AOS(	AOS_FORMAT_FLOAT,
 						VB->FogCoordPtr,
 						immd ? 4 : VB->FogCoordPtr->size,
@@ -408,7 +447,7 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 
 	r300->state.texture.tc_count = 0;
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (inputs & (_TNL_BIT_TEX0 << i)) {
+		if (RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_TEX(i) )) {
 			CONFIGURE_AOS(i_tex[i], AOS_FORMAT_FLOAT,
 							VB->AttribPtr[VERT_ATTRIB_TEX0+i],
 							immd ? 4 : VB->AttribPtr[VERT_ATTRIB_TEX0+i].size,
@@ -467,17 +506,6 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 	((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->packet0.count = (nr+1)>>1;
 
 
-	/* Mesa assumes that all missing components are from (0, 0, 0, 1) */
-#define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<<R300_INPUT_ROUTE_X_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_Y<<R300_INPUT_ROUTE_Y_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_Z<<R300_INPUT_ROUTE_Z_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_W<<R300_INPUT_ROUTE_W_SHIFT))
-
-#define ALL_DEFAULT ((R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_X_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Y_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Z_SHIFT) \
-		| (R300_INPUT_ROUTE_SELECT_ONE<<R300_INPUT_ROUTE_W_SHIFT))
-
 	R300_STATECHANGE(r300, vir[1]);
 
 	for(i=0; i < nr; i++)
@@ -486,33 +514,22 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 			r300->state.aos[i].aos_size=/*3*/4; /* XXX */
 		}
 		
-		
-	for(i=0;i+1<nr;i+=2){
+	for (i=0;i+1<nr;i+=2) {
 		/* do i first.. */
-		mask=(1<<(r300->state.aos[i].aos_size*3))-1;
-		dw=(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE;
-
+		dw = fix_comps(t_comps(r300->state.aos[i].aos_size), r300->state.aos[i].aos_format) | R300_INPUT_ROUTE_ENABLE;
 		/* i+1 */
-		mask=(1<<(r300->state.aos[i+1].aos_size*3))-1;
-		dw|=(
-		(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE
-		)<<16;
-
+		dw |= (fix_comps(t_comps(r300->state.aos[i+1].aos_size), r300->state.aos[i+1].aos_format) | R300_INPUT_ROUTE_ENABLE) << 16;
+		
 		//fprintf(stderr, "vir1 dw=%08x\n", dw);
 		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(i>>1)]=dw;
-		}
-	if(nr & 1){
-		mask=(1<<(r300->state.aos[nr-1].aos_size*3))-1;
-		dw=(ALL_COMPONENTS & mask)
-		| (ALL_DEFAULT & ~mask)
-		| R300_INPUT_ROUTE_ENABLE;
-		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(nr>>1)]=dw;
+	}
+	if (nr & 1) {
+		dw = fix_comps(t_comps(r300->state.aos[nr-1].aos_size), r300->state.aos[nr-1].aos_format) | R300_INPUT_ROUTE_ENABLE;
+		
 		//fprintf(stderr, "vir1 dw=%08x\n", dw);
-		}
+		r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(nr>>1)]=dw;
+	}
+
 	/* Set the rest of INPUT_ROUTE_1 to 0 */
 	//for(i=((count+1)>>1); i<8; i++)r300->hw.vir[1].cmd[R300_VIR_CNTL_0+i]=0x0;
 	((drm_r300_cmd_header_t*)r300->hw.vir[1].cmd)->packet0.count = (nr+1)>>1;
@@ -531,17 +548,17 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 #if 0
 	r300->hw.vic.cmd[R300_VIC_CNTL_1]=0;
 
-	if(r300->state.render_inputs & _TNL_BIT_POS)
+	if(RENDERINPUTS_TEST( r300->state.render_inputs_bitset, _TNL_ATTRIB_POS ))
 		r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_POS;
 
-	if(r300->state.render_inputs & _TNL_BIT_NORMAL)
+	if(RENDERINPUTS_TEST( r300->state.render_inputs_bitset, _TNL_ATTRIB_NORMAL ))
 		r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_NORMAL;
 
-	if(r300->state.render_inputs & _TNL_BIT_COLOR0)
+	if(RENDERINPUTS_TEST( r300->state.render_inputs_bitset, _TNL_ATTRIB_COLOR0 ))
 		r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_COLOR;
 
 	for(i=0;i < ctx->Const.MaxTextureUnits;i++)
-		if(r300->state.render_inputs & (_TNL_BIT_TEX0<<i))
+		if(RENDERINPUTS_TEST( r300->state.render_inputs_bitset, _TNL_ATTRIB_TEX(i) ))
 			r300->hw.vic.cmd[R300_VIC_CNTL_1]|=(R300_INPUT_CNTL_TC0<<i);
 #endif
 
@@ -573,15 +590,15 @@ void r300EmitArrays(GLcontext * ctx, GLboolean immd)
 			if(OutputsWritten & (1<<(VERT_RESULT_TEX0+i)))
 				r300->hw.vof.cmd[R300_VOF_CNTL_1] |= (4<<(3*i));
 	} else {
-		if(inputs & _TNL_BIT_POS)
+		if(RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_POS ))
 			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
-		if(inputs & _TNL_BIT_COLOR0)
+		if(RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_COLOR0 ))
 			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT;
-		if(inputs & _TNL_BIT_COLOR1)
+		if(RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_COLOR1 ))
 			r300->hw.vof.cmd[R300_VOF_CNTL_0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
 
 		for(i=0;i < ctx->Const.MaxTextureUnits;i++)
-			if(inputs & (_TNL_BIT_TEX0<<i))
+			if(RENDERINPUTS_TEST( inputs_bitset, _TNL_ATTRIB_TEX(i) ))
 				r300->hw.vof.cmd[R300_VOF_CNTL_1]|=(4<<(3*i));
 	}
 
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index c78462af2ed..4afe4f239c2 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -517,6 +517,10 @@ I am fairly certain that they are correct unless stated otherwise in comments.
 #	define R300_PM_BACK_LINE              (1 << 7)
 #	define R300_PM_BACK_FILL              (1 << 8)
 
+/* Fog parameters */
+#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. 
    Ordering might be wrong.
@@ -1113,7 +1117,17 @@ I am fairly certain that they are correct unless stated otherwise in comments.
 #       define R300_FPI2_UNKNOWN_31              (1 << 31)
 /* END */
 
-/* gap */
+/* Fog state and color */
+#define R300_RE_FOG_STATE                   0x4BC0
+#       define R300_FOG_ENABLE                   (1 << 0)
+#	define R300_FOG_MODE_LINEAR              (0 << 1)
+#	define R300_FOG_MODE_EXP                 (1 << 1)
+#	define R300_FOG_MODE_EXP2                (2 << 1)
+#	define R300_FOG_MODE_MASK                (3 << 1)
+#define R300_FOG_COLOR_R                    0x4BC8
+#define R300_FOG_COLOR_G                    0x4BCC
+#define R300_FOG_COLOR_B                    0x4BD0
+
 #define R300_PP_ALPHA_TEST                  0x4BD4
 #       define R300_REF_ALPHA_MASK               0x000000ff
 #       define R300_ALPHA_TEST_FAIL              (0 << 8)
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index a0375f90227..f86a439bc29 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -382,7 +382,6 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
 	
 	r300UpdateShaders(rmesa);
 	if (rmesa->state.VB.LockCount == 0 || 1) {
- 	  	r300ReleaseArrays(ctx);
 		r300EmitArrays(ctx, GL_FALSE);
 
 		r300UpdateShaderStates(rmesa);
@@ -446,6 +445,7 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
 #ifdef USER_BUFFERS
 	r300UseArrays(ctx);
 #endif
+	r300ReleaseArrays(ctx);
 	return GL_FALSE;
 }
 
@@ -462,19 +462,19 @@ int r300Fallback(GLcontext *ctx)
 {
 	int i;
 
-	//FALLBACK_IF(ctx->RenderMode != GL_RENDER);  // We do not do SELECT or FEEDBACK (yet ?)
+	FALLBACK_IF(ctx->RenderMode != GL_RENDER);  // We do not do SELECT or FEEDBACK (yet ?)
 	
 #if 0 /* These should work now.. */
 	FALLBACK_IF(ctx->Color.DitherFlag);
 	FALLBACK_IF(ctx->Color.AlphaEnabled); // GL_ALPHA_TEST
 	FALLBACK_IF(ctx->Color.BlendEnabled); // GL_BLEND
 	FALLBACK_IF(ctx->Polygon.OffsetFill); // GL_POLYGON_OFFSET_FILL
+	FALLBACK_IF(ctx->Fog.Enabled);
 #endif
 	FALLBACK_IF(ctx->Polygon.OffsetPoint); // GL_POLYGON_OFFSET_POINT
 	FALLBACK_IF(ctx->Polygon.OffsetLine); // GL_POLYGON_OFFSET_LINE
 	//FALLBACK_IF(ctx->Stencil.Enabled); // GL_STENCIL_TEST
 	
-	//FALLBACK_IF(ctx->Fog.Enabled); // GL_FOG disable as swtcl doesnt seem to support this
 	//FALLBACK_IF(ctx->Polygon.SmoothFlag); // GL_POLYGON_SMOOTH disabling to get blender going
 	FALLBACK_IF(ctx->Polygon.StippleFlag); // GL_POLYGON_STIPPLE
 	FALLBACK_IF(ctx->Multisample.Enabled); // GL_MULTISAMPLE_ARB
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index d4dafe62a73..70d42c2ea3b 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -417,6 +417,24 @@ static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state)
 	case GL_TEXTURE_3D:
 		break;
 
+	case GL_FOG:
+		R300_STATECHANGE(r300, fogs);
+		if (state) {
+			r300->hw.fogs.cmd[R300_FOGS_STATE] |=
+			    R300_FOG_ENABLE;
+			
+			ctx->Driver.Fogfv( ctx, GL_FOG_MODE, NULL );
+			ctx->Driver.Fogfv( ctx, GL_FOG_DENSITY, &ctx->Fog.Density );
+			ctx->Driver.Fogfv( ctx, GL_FOG_START, &ctx->Fog.Start );
+			ctx->Driver.Fogfv( ctx, GL_FOG_END, &ctx->Fog.End );
+			ctx->Driver.Fogfv( ctx, GL_FOG_COLOR, ctx->Fog.Color );
+		} else {
+			r300->hw.fogs.cmd[R300_FOGS_STATE] &=
+			    ~R300_FOG_ENABLE;
+		}
+		
+		break;
+
 	case GL_ALPHA_TEST:
 		R300_STATECHANGE(r300, at);
 		if (state) {
@@ -640,6 +658,101 @@ static void r300ColorMask(GLcontext* ctx,
 }
 
 /* =============================================================
+ * Fog
+ */
+static void r300Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param )
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	union { int i; float f; } fogScale, fogStart;
+	
+	(void) param;
+	
+	fogScale.i = r300->hw.fogp.cmd[R300_FOGP_SCALE];
+	fogStart.i = r300->hw.fogp.cmd[R300_FOGP_START];
+
+	switch (pname) {
+	case GL_FOG_MODE:
+		if (!ctx->Fog.Enabled)
+			return;
+		switch (ctx->Fog.Mode) {
+		case GL_LINEAR:
+			R300_STATECHANGE(r300, fogs);
+			r300->hw.fogs.cmd[R300_FOGS_STATE] =
+				(r300->hw.fogs.cmd[R300_FOGS_STATE] & ~R300_FOG_MODE_MASK) | R300_FOG_MODE_LINEAR;
+
+			if (ctx->Fog.Start == ctx->Fog.End) {
+				fogScale.f = -1.0;
+				fogStart.f = 1.0;
+			}
+			else {
+				fogScale.f = 1.0 / (ctx->Fog.End-ctx->Fog.Start);
+				fogStart.f = -ctx->Fog.Start / (ctx->Fog.End-ctx->Fog.Start);
+			}
+			break;
+		case GL_EXP:
+			R300_STATECHANGE(r300, fogs);
+			r300->hw.fogs.cmd[R300_FOGS_STATE] =
+				(r300->hw.fogs.cmd[R300_FOGS_STATE] & ~R300_FOG_MODE_MASK) | R300_FOG_MODE_EXP;
+			fogScale.f = 0.0933*ctx->Fog.Density;
+			fogStart.f = 0.0;
+			break;
+		case GL_EXP2:
+			R300_STATECHANGE(r300, fogs);
+			r300->hw.fogs.cmd[R300_FOGS_STATE] =
+				(r300->hw.fogs.cmd[R300_FOGS_STATE] & ~R300_FOG_MODE_MASK) | R300_FOG_MODE_EXP2;
+			fogScale.f = 0.3*ctx->Fog.Density;
+			fogStart.f = 0.0;
+		default:
+			return;
+		}
+		break;
+	case GL_FOG_DENSITY:
+		switch (ctx->Fog.Mode) {
+		case GL_EXP:
+			fogScale.f = 0.0933*ctx->Fog.Density;
+			fogStart.f = 0.0;
+			break;
+		case GL_EXP2:
+			fogScale.f = 0.3*ctx->Fog.Density;
+			fogStart.f = 0.0;
+		default:
+			break;
+		}
+		break;
+	case GL_FOG_START:
+	case GL_FOG_END:
+		if (ctx->Fog.Mode == GL_LINEAR) {
+			if (ctx->Fog.Start == ctx->Fog.End) {
+				fogScale.f = -1.0;
+				fogStart.f = 1.0;
+			}
+			else {
+				fogScale.f = 1.0 / (ctx->Fog.End-ctx->Fog.Start);
+				fogStart.f = -ctx->Fog.Start / (ctx->Fog.End-ctx->Fog.Start);
+			}
+		}
+		break;
+	case GL_FOG_COLOR:
+		R300_STATECHANGE(r300, fogc);
+		r300->hw.fogc.cmd[R300_FOGC_R] = (GLuint) (ctx->Fog.Color[0]*1023.0F) & 0x3FF;
+		r300->hw.fogc.cmd[R300_FOGC_G] = (GLuint) (ctx->Fog.Color[1]*1023.0F) & 0x3FF;
+		r300->hw.fogc.cmd[R300_FOGC_B] = (GLuint) (ctx->Fog.Color[2]*1023.0F) & 0x3FF;
+		break;
+	case GL_FOG_COORD_SRC:
+		break;
+	default:
+		return;
+	}
+
+	if (fogScale.i != r300->hw.fogp.cmd[R300_FOGP_SCALE] ||
+	    fogStart.i != r300->hw.fogp.cmd[R300_FOGP_START]) {
+		R300_STATECHANGE(r300, fogp);
+		r300->hw.fogp.cmd[R300_FOGP_SCALE] = fogScale.i;
+		r300->hw.fogp.cmd[R300_FOGP_START] = fogStart.i;
+	}
+}
+
+/* =============================================================
  * Point state
  */
 static void r300PointSize(GLcontext * ctx, GLfloat size)
@@ -750,8 +863,8 @@ static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
                                     GLenum func, GLint ref, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint refmask = ((ctx->Stencil.Ref[0] << R300_RB3D_ZS2_STENCIL_REF_SHIFT) |
-			  (ctx->Stencil.ValueMask[0] << R300_RB3D_ZS2_STENCIL_MASK_SHIFT));
+	GLuint refmask = (((ctx->Stencil.Ref[0] & 0xff) << R300_RB3D_ZS2_STENCIL_REF_SHIFT) |
+			  ((ctx->Stencil.ValueMask[0] & 0xff) << R300_RB3D_ZS2_STENCIL_MASK_SHIFT));
 			  
 	GLuint flag;
 
@@ -777,7 +890,7 @@ static void r300StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask)
 
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_2]  &= ~(R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT);
-	rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= ctx->Stencil.WriteMask[0] << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT;
+	rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= (ctx->Stencil.WriteMask[0] & 0xff) << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT;
 }
 
 
@@ -807,9 +920,9 @@ static void r300ClearStencil(GLcontext * ctx, GLint s)
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
 	rmesa->state.stencil.clear =
-	    ((GLuint) ctx->Stencil.Clear |
+	    ((GLuint) (ctx->Stencil.Clear & 0xff) |
 	     (R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_MASK_SHIFT) |
-	     (ctx->Stencil.WriteMask[0] << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT));
+	     ((ctx->Stencil.WriteMask[0] & 0xff) << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT));
 }
 
 /* =============================================================
@@ -1157,6 +1270,15 @@ void r300_setup_textures(GLcontext *ctx)
 		fprintf(stderr, "TX_ENABLE: %08x  last_hw_tmu=%d\n", r300->hw.txe.cmd[R300_TXE_ENABLE], last_hw_tmu);
 }
 
+union r300_outputs_written {
+	GLuint vp_outputs;                       /* hw_tcl_on */
+	DECLARE_RENDERINPUTS(index_bitset);      /* !hw_tcl_on */
+};
+
+#define R300_OUTPUTS_WRITTEN_TEST(ow, vp_result, tnl_attrib) \
+	((hw_tcl_on) ? (ow).vp_outputs & (1 << (vp_result)) : \
+	RENDERINPUTS_TEST( (ow.index_bitset), (tnl_attrib) ))
+
 void r300_setup_rs_unit(GLcontext *ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
@@ -1171,16 +1293,16 @@ void r300_setup_rs_unit(GLcontext *ctx)
 		0x00,
 		0x00
 	};
-	GLuint OutputsWritten;
+	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int in_texcoords, col_interp_nr;
 	int i;
 
 	if(hw_tcl_on)
-		OutputsWritten = CURRENT_VERTEX_SHADER(ctx)->Base.OutputsWritten;
+		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->Base.OutputsWritten;
 	else
-		OutputsWritten = r300->state.render_inputs;
+		RENDERINPUTS_COPY( OutputsWritten.index_bitset, r300->state.render_inputs_bitset );
 
 	if (ctx->FragmentProgram._Current)
 		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
@@ -1211,7 +1333,7 @@ void r300_setup_rs_unit(GLcontext *ctx)
 					| (fp_reg << R300_RS_ROUTE_DEST_SHIFT);
 			high_rr = fp_reg;
 
-			if (!(OutputsWritten & (hw_tcl_on ? (1 << (VERT_RESULT_TEX0+i)) : (_TNL_BIT_TEX0<<i)))) {
+			if (!R300_OUTPUTS_WRITTEN_TEST( OutputsWritten, VERT_RESULT_TEX0+i, _TNL_ATTRIB_TEX(i) )) {
 				/* Passing invalid data here can lock the GPU. */
 				WARN_ONCE("fragprog wants coords for tex%d, vp doesn't provide them!\n", i);
 				//_mesa_print_program(&CURRENT_VERTEX_SHADER(ctx)->Base);
@@ -1221,12 +1343,12 @@ void r300_setup_rs_unit(GLcontext *ctx)
 			fp_reg++;
 		} 
 		/* Need to count all coords enabled at vof */
-		if (OutputsWritten & (hw_tcl_on ? (1 << (VERT_RESULT_TEX0+i)) : (_TNL_BIT_TEX0<<i)))
+		if (R300_OUTPUTS_WRITTEN_TEST( OutputsWritten, VERT_RESULT_TEX0+i, _TNL_ATTRIB_TEX(i) ))
 			in_texcoords++;
 	}
 
 	if (InputsRead & FRAG_BIT_COL0) {
-		if (!(OutputsWritten & (hw_tcl_on ? (1<<VERT_RESULT_COL0) : _TNL_BIT_COLOR0))) {
+		if (!R300_OUTPUTS_WRITTEN_TEST( OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0 )) {
 			WARN_ONCE("fragprog wants col0, vp doesn't provide it\n");
 			goto out; /* FIXME */
 			//_mesa_print_program(&CURRENT_VERTEX_SHADER(ctx)->Base);
@@ -1242,7 +1364,7 @@ void r300_setup_rs_unit(GLcontext *ctx)
 	out:
 	
 	if (InputsRead & FRAG_BIT_COL1) {
-		if (!(OutputsWritten & (hw_tcl_on ? (1<<VERT_RESULT_COL1) : _TNL_BIT_COLOR1))) {
+		if (!R300_OUTPUTS_WRITTEN_TEST( OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1 )) {
 			WARN_ONCE("fragprog wants col1, vp doesn't provide it\n");
 			//exit(-1);
 		}
@@ -1281,6 +1403,7 @@ void r300_setup_rs_unit(GLcontext *ctx)
 #define bump_vpu_count(ptr, new_count)   do{\
 	drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr));\
 	int _nc=(new_count)/4; \
+	assert(_nc < 256); \
 	if(_nc>_p->vpu.count)_p->vpu.count=_nc;\
 	}while(0)
 
@@ -1402,7 +1525,7 @@ static void r300GenerateSimpleVertexShader(r300ContextPtr r300)
 		)
 	o_reg += 2;
 	
-	if (r300->state.render_inputs & _TNL_BIT_COLOR1) {
+	if (RENDERINPUTS_TEST( r300->state.render_inputs_bitset, _TNL_ATTRIB_COLOR1 )) {
 		WRITE_OP(
 			EASY_VSF_OP(MUL, o_reg++, ALL, RESULT),
 			VSF_REG(r300->state.vap_reg.i_color[1]),
@@ -1413,7 +1536,7 @@ static void r300GenerateSimpleVertexShader(r300ContextPtr r300)
 	
 	/* Pass through texture coordinates, if any */
 	for(i=0;i < r300->radeon.glCtx->Const.MaxTextureUnits;i++)
-		if(r300->state.render_inputs & (_TNL_BIT_TEX0<<i)){
+		if (RENDERINPUTS_TEST( r300->state.render_inputs_bitset, _TNL_ATTRIB_TEX(i) )){
 			// fprintf(stderr, "i_tex[%d]=%d\n", i, r300->state.vap_reg.i_tex[i]);
 			WRITE_OP(
 				EASY_VSF_OP(MUL, o_reg++ /* 2+i */, ALL, RESULT),
@@ -1870,7 +1993,8 @@ void r300ResetHwState(r300ContextPtr r300)
 		r300->hw.gb_misc.cmd[R300_GB_MISC_TILE_CONFIG] = R300_GB_TILE_ENABLE
 							| R300_GB_TILE_PIPE_COUNT_RV300
 							| R300_GB_TILE_SIZE_16;
-	r300->hw.gb_misc.cmd[R300_GB_MISC_SELECT] = 0x00000000;
+	/* set to 0 when fog is disabled? */
+	r300->hw.gb_misc.cmd[R300_GB_MISC_SELECT] = R300_GB_FOG_SELECT_1_1_W;
 	r300->hw.gb_misc.cmd[R300_GB_MISC_AA_CONFIG] = 0x00000000; /* No antialiasing */
 
 	//r300->hw.txe.cmd[R300_TXE_ENABLE] = 0;
@@ -1907,9 +2031,6 @@ void r300ResetHwState(r300ContextPtr r300)
 	r300PolygonMode(ctx, GL_BACK, ctx->Polygon.BackMode);
 	r300->hw.unk4288.cmd[2] = 0x00000001;
 	r300->hw.unk4288.cmd[3] = 0x00000000;
-	r300->hw.unk4288.cmd[4] = 0x00000000;
-	r300->hw.unk4288.cmd[5] = 0x00000000;
-
 	r300->hw.unk42A0.cmd[1] = 0x00000000;
 
 	r300PolygonOffset(ctx, ctx->Polygon.OffsetFactor, ctx->Polygon.OffsetUnits);
@@ -1949,13 +2070,13 @@ void r300ResetHwState(r300ContextPtr r300)
 		r300->hw.fpi[3].cmd[i] = FP_SELA(0,W,NO,FP_TMP(0),0,0);
 	}
 #endif
-
-	r300->hw.unk4BC0.cmd[1] = 0;
-
-	r300->hw.unk4BC8.cmd[1] = 0;
-	r300->hw.unk4BC8.cmd[2] = 0;
-	r300->hw.unk4BC8.cmd[3] = 0;
-
+	r300Enable(ctx, GL_FOG, ctx->Fog.Enabled);
+	ctx->Driver.Fogfv( ctx, GL_FOG_MODE, NULL );
+	ctx->Driver.Fogfv( ctx, GL_FOG_DENSITY, &ctx->Fog.Density );
+	ctx->Driver.Fogfv( ctx, GL_FOG_START, &ctx->Fog.Start );
+	ctx->Driver.Fogfv( ctx, GL_FOG_END, &ctx->Fog.End );
+	ctx->Driver.Fogfv( ctx, GL_FOG_COLOR, ctx->Fog.Color );
+	ctx->Driver.Fogfv( ctx, GL_FOG_COORDINATE_SOURCE_EXT, NULL );
 
 	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
 	r300->hw.unk4BD8.cmd[1] = 0;
@@ -2128,6 +2249,7 @@ void r300InitStateFuncs(struct dd_function_table* functions)
 	functions->DepthFunc = r300DepthFunc;
 	functions->DepthMask = r300DepthMask;
 	functions->CullFace = r300CullFace;
+	functions->Fogfv = r300Fogfv;
 	functions->FrontFace = r300FrontFace;
 	functions->ShadeModel = r300ShadeModel;
 
diff --git a/src/mesa/drivers/dri/r300/r300_texmem.c b/src/mesa/drivers/dri/r300/r300_texmem.c
index a9f64b47c4c..c1d01020cdb 100644
--- a/src/mesa/drivers/dri/r300/r300_texmem.c
+++ b/src/mesa/drivers/dri/r300/r300_texmem.c
@@ -130,6 +130,12 @@ static void r300UploadGARTClientSubImage(r300ContextPtr rmesa,
 		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
 		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
 		break;
+	case 8:
+	case 16:
+		blit_format = R200_CP_COLOR_FORMAT_CI8;
+		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
+		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
+		break;
 	default:
 		return;
 	}
@@ -144,6 +150,10 @@ static void r300UploadGARTClientSubImage(r300ContextPtr rmesa,
 	width = texImage->Width;
 	height = texImage->Height;
 
+	if (texFormat->TexelBytes > 4) {
+		width *= texFormat->TexelBytes;
+	}
+
 	r300EmitWait(rmesa, R300_WAIT_3D);
 
 	r300EmitBlit(rmesa, blit_format,
@@ -177,6 +187,10 @@ static void r300UploadRectSubImage(r300ContextPtr rmesa,
 	case 4:
 		blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
 		break;
+	case 8:
+	case 16:
+		blit_format = R200_CP_COLOR_FORMAT_CI8;
+		break;
 	default:
 		return;
 	}
@@ -189,6 +203,10 @@ static void r300UploadRectSubImage(r300ContextPtr rmesa,
 	height = texImage->Height;
 	dstPitch = t->pitch;
 
+	if (texFormat->TexelBytes > 4) {
+		width *= texFormat->TexelBytes;
+	}
+
 	if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) {
 		/* In this case, could also use GART texturing.  This is
 		 * currently disabled, but has been tested & works.
diff --git a/src/mesa/drivers/dri/r300/r300_vertexprog.c b/src/mesa/drivers/dri/r300/r300_vertexprog.c
index f794c9aa67a..9e32cfcfef9 100644
--- a/src/mesa/drivers/dri/r300/r300_vertexprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertexprog.c
@@ -916,6 +916,16 @@ void r300_translate_vertex_shader(struct r300_vertex_program *vp)
 			fprintf(stderr, "Dont know how to handle op %d yet\n", vpi->Opcode);
 			exit(-1);
 		break;
+		case OPCODE_SWZ:
+			hw_op=(src[0].File == PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 : R300_VPI_OUT_OP_MAD;
+			
+			o_inst->op=MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+				t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File));
+			o_inst->src1=t_src(vp, &src[0]);
+			o_inst->src2=ONE_SRC_0;
+			o_inst->src3=ZERO_SRC_0;
+
+			goto next;
 		case OPCODE_END:
 			break;
 		default:
diff --git a/src/mesa/drivers/dri/r300/radeon_context.c b/src/mesa/drivers/dri/r300/radeon_context.c
index 4e351dd66b0..0147c6492e5 100644
--- a/src/mesa/drivers/dri/r300/radeon_context.c
+++ b/src/mesa/drivers/dri/r300/radeon_context.c
@@ -76,7 +76,7 @@ static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name)
 	case GL_RENDERER:
 	{
 		unsigned offset;
-		GLuint agp_mode = radeon->radeonScreen->IsPCI ? 0 :
+		GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 :
 			radeon->radeonScreen->AGPMode;
 		const char* chipname;
 
diff --git a/src/mesa/drivers/dri/r300/radeon_span.c b/src/mesa/drivers/dri/r300/radeon_span.c
index 1f000b97666..1b74f6779b4 100644
--- a/src/mesa/drivers/dri/r300/radeon_span.c
+++ b/src/mesa/drivers/dri/r300/radeon_span.c
@@ -40,7 +40,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Keith Whitwell <[email protected]>
  *
  */
-
+#include <unistd.h>
 #include "glheader.h"
 #include "imports.h"
 #include "swrast/swrast.h"
@@ -252,6 +252,29 @@ do {									\
 static void radeonSpanRenderStart( GLcontext *ctx )
 {
    radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   {
+	static int first = 1;
+	r300ContextPtr r300 = (r300ContextPtr)rmesa;
+	
+	if (first) {
+		r300->span_dlocking = getenv("R300_SPAN_DISABLE_LOCKING") ? 1 : 0;
+		if (r300->span_dlocking == 0) {
+			fprintf(stderr, "Try R300_SPAN_DISABLE_LOCKING env var if this hangs.\n");
+			fflush(stderr);
+			sleep(1);
+		}
+		first = 0;
+	}
+	
+	if (r300->span_dlocking) {
+		r300Flush(ctx);
+		LOCK_HARDWARE( rmesa );
+		radeonWaitForIdleLocked( rmesa );
+		UNLOCK_HARDWARE( rmesa );
+		
+		return;
+	}
+   }
    //   R300_FIREVERTICES( rmesa );
    // old code has flush
    r300Flush(ctx);
@@ -262,8 +285,10 @@ static void radeonSpanRenderStart( GLcontext *ctx )
 static void radeonSpanRenderFinish( GLcontext *ctx )
 {
    radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
+   r300ContextPtr r300 = (r300ContextPtr)rmesa;
    _swrast_flush( ctx );
-   UNLOCK_HARDWARE( rmesa );
+   if (r300->span_dlocking == 0)
+	UNLOCK_HARDWARE( rmesa );
 }
 
 void radeonInitSpanFuncs( GLcontext *ctx )
diff --git a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
index 5ea9bb4eb45..1044973b001 100644
--- a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
+++ b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
@@ -71,7 +71,7 @@ void radeon_vb_to_rvb(r300ContextPtr rmesa, struct radeon_vertex_buffer *rvb, st
 	CONV_VB(VERT_ATTRIB_COLOR1, SecondaryColorPtr[0]);
 	CONV_VB(VERT_ATTRIB_FOG, FogCoordPtr);
 	
-	for (i=0; i < MAX_TEXTURE_COORD_UNITS; i++)
+	for (i=0; i < ctx->Const.MaxTextureCoordUnits; i++)
 		CONV_VB(VERT_ATTRIB_TEX0 + i, TexCoordPtr[i]);
 	
 	rvb->Primitive = vb->Primitive;
@@ -182,7 +182,7 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 	unsigned int min = ~0, max = 0;
 	struct tnl_prim prim;
 	static void *ptr = NULL;
-	static struct r300_dma_region rvb;
+	struct r300_dma_region rvb;
 	const GLvoid *indices = c_indices;
 	
 	if (count > 65535) {
@@ -205,8 +205,8 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 		return;
 	
 	FLUSH_CURRENT( ctx, 0 );
-	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 	
+	memset(&rvb, 0, sizeof(rvb));
 	switch (type) {
 	case GL_UNSIGNED_BYTE:
 		for (i=0; i < count; i++) {
@@ -216,14 +216,22 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 				max = ((unsigned char *)indices)[i];
 		}
 		
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
-		
+#endif		
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 			
+#ifdef FORCE_32BITS_ELTS
+		for (i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#else
 		for (i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#endif
 	break;
 		
 	case GL_UNSIGNED_SHORT:
@@ -234,14 +242,23 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 				max = ((unsigned short int *)indices)[i];
 		}
 		
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
+#endif
 		
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
+#ifdef FORCE_32BITS_ELTS
+		for (i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#else
 		for (i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#endif
 	break;
 	
 	case GL_UNSIGNED_INT:
@@ -252,17 +269,20 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 				max = ((unsigned int *)indices)[i];
 		}
 		
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		if (max - min <= 65535)
 			elt_size = 2;
 		else 
 			elt_size = 4;
-		
+#endif
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
 		
-		if (max - min <= 65535)
+		if (elt_size == 2)
 			for (i=0; i < count; i++)
 				((unsigned short int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
 		else
@@ -278,12 +298,12 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 	if (ctx->NewState) 
 		_mesa_update_state( ctx );
 	
-	for (i=_TNL_ATTRIB_MAT_FRONT_AMBIENT; i < _TNL_ATTRIB_INDEX; i++) {
+	for (i=_TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 		rmesa->temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
 		TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &rmesa->dummy_attrib[i];
 	}
 	r300UpdateShaders(rmesa);
-	for (i=_TNL_ATTRIB_MAT_FRONT_AMBIENT; i < _TNL_ATTRIB_INDEX; i++) {
+	for (i=_TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 		TNL_CONTEXT(ctx)->vb.AttribPtr[i] = rmesa->temp_attrib[i];
 	}
 	
@@ -311,8 +331,10 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 			return;
 		}
 	} else {
-		if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL)
+		if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL) {
+			r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 			goto fallback;
+		}
 		
 		rmesa->state.VB.Count = max - min + 1;
 	}
@@ -337,6 +359,7 @@ static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const G
 	if(rvb.buf)
 		radeon_mm_use(rmesa, rvb.buf->id);
 	
+	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 	return;
 	
 	fallback:
@@ -355,7 +378,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	int elt_size;
 	int i;
 	void *ptr = NULL;
-	static struct r300_dma_region rvb;
+	struct r300_dma_region rvb;
 	const GLvoid *indices = c_indices;
 	
 	if (count > 65535) {
@@ -381,23 +404,34 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 #ifdef OPTIMIZE_ELTS
 	min = 0;
 #endif
-	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 	
-	switch(type){
+	memset(&rvb, 0, sizeof(rvb));
+	switch (type){
 	case GL_UNSIGNED_BYTE:
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
-		
+#endif	
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
+#ifdef FORCE_32BITS_ELTS
+		for(i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#else
 		for(i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
+#endif
 	break;
 	
 	case GL_UNSIGNED_SHORT:
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		elt_size = 2;
-		
+#endif	
 #ifdef OPTIMIZE_ELTS
 		if (min == 0 && ctx->Array.ElementArrayBufferObj->Name){
 			ptr = indices;
@@ -408,21 +442,29 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 
+#ifdef FORCE_32BITS_ELTS
+		for(i=0; i < count; i++)
+			((unsigned int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#else
 		for(i=0; i < count; i++)
 			((unsigned short int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
+#endif
 	break;
 	
 	case GL_UNSIGNED_INT:
+#ifdef FORCE_32BITS_ELTS
+		elt_size = 4;
+#else
 		if (max - min <= 65535)
 			elt_size = 2;
 		else 
 			elt_size = 4;
-		
+#endif	
 		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
 		rvb.aos_offset = GET_START(&rvb);
 		ptr = rvb.address + rvb.start;
 		
-		if (max - min <= 65535)
+		if (elt_size == 2)
 			for (i=0; i < count; i++)
 				((unsigned short int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
 		else
@@ -440,12 +482,12 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	if (ctx->NewState) 
 		_mesa_update_state( ctx );
 	
-	for (i=_TNL_ATTRIB_MAT_FRONT_AMBIENT; i < _TNL_ATTRIB_INDEX; i++) {
+	for (i=_TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 		rmesa->temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
 		TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &rmesa->dummy_attrib[i];
 	}
 	r300UpdateShaders(rmesa);
-	for (i=_TNL_ATTRIB_MAT_FRONT_AMBIENT; i < _TNL_ATTRIB_INDEX; i++) {
+	for (i=_TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 		TNL_CONTEXT(ctx)->vb.AttribPtr[i] = rmesa->temp_attrib[i];
 	}
 
@@ -473,8 +515,10 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 			return;
 		}*/
 	} else {
-		if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL)
+		if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL) {
+			r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 			goto fallback;
+		}
 		
 		rmesa->state.VB.Count = max - min + 1;
 	}
@@ -501,6 +545,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	if(rvb.buf)
 		radeon_mm_use(rmesa, rvb.buf->id);
 	
+	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
 	return ;
 	
 	fallback:
@@ -533,12 +578,12 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 	
 	/* XXX: setup_arrays before state update? */
 	
-	for (i=_TNL_ATTRIB_MAT_FRONT_AMBIENT; i < _TNL_ATTRIB_INDEX; i++) {
+	for (i=_TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 		rmesa->temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
 		TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &rmesa->dummy_attrib[i];
 	}
 	r300UpdateShaders(rmesa);
-	for (i=_TNL_ATTRIB_MAT_FRONT_AMBIENT; i < _TNL_ATTRIB_INDEX; i++) {
+	for (i=_TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 		TNL_CONTEXT(ctx)->vb.AttribPtr[i] = rmesa->temp_attrib[i];
 	}