aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorMarek Olšák <maraeo@gmail.com>2011-10-24 19:38:45 -0400
committerDave Airlie <airlied@redhat.com>2011-11-20 02:53:13 -0500
commite70f224c1938af208b64b02c5cec27889fefcaec (patch)
treea247c81526e3b169158693c151efda3c2bc073e9
parent6991b8f2a3193397461104a27be417addb8d032b (diff)
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS
This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
-rw-r--r--drivers/gpu/drm/radeon/evergreen_cs.c92
-rw-r--r--drivers/gpu/drm/radeon/r300.c94
-rw-r--r--drivers/gpu/drm/radeon/r600_cs.c26
-rw-r--r--drivers/gpu/drm/radeon/radeon.h3
-rw-r--r--drivers/gpu/drm/radeon/radeon_cs.c11
-rw-r--r--drivers/gpu/drm/radeon/radeon_drv.c3
-rw-r--r--include/drm/radeon_drm.h4
7 files changed, 135 insertions, 98 deletions
diff --git a/drivers/gpu/drm/radeon/evergreen_cs.c b/drivers/gpu/drm/radeon/evergreen_cs.c
index 7fdfa8ea7570..38e1bda73d33 100644
--- a/drivers/gpu/drm/radeon/evergreen_cs.c
+++ b/drivers/gpu/drm/radeon/evergreen_cs.c
@@ -480,21 +480,23 @@ static int evergreen_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
480 } 480 }
481 break; 481 break;
482 case DB_Z_INFO: 482 case DB_Z_INFO:
483 r = evergreen_cs_packet_next_reloc(p, &reloc);
484 if (r) {
485 dev_warn(p->dev, "bad SET_CONTEXT_REG "
486 "0x%04X\n", reg);
487 return -EINVAL;
488 }
489 track->db_z_info = radeon_get_ib_value(p, idx); 483 track->db_z_info = radeon_get_ib_value(p, idx);
490 ib[idx] &= ~Z_ARRAY_MODE(0xf); 484 if (!p->keep_tiling_flags) {
491 track->db_z_info &= ~Z_ARRAY_MODE(0xf); 485 r = evergreen_cs_packet_next_reloc(p, &reloc);
492 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) { 486 if (r) {
493 ib[idx] |= Z_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 487 dev_warn(p->dev, "bad SET_CONTEXT_REG "
494 track->db_z_info |= Z_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 488 "0x%04X\n", reg);
495 } else { 489 return -EINVAL;
496 ib[idx] |= Z_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 490 }
497 track->db_z_info |= Z_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 491 ib[idx] &= ~Z_ARRAY_MODE(0xf);
492 track->db_z_info &= ~Z_ARRAY_MODE(0xf);
493 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
494 ib[idx] |= Z_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
495 track->db_z_info |= Z_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
496 } else {
497 ib[idx] |= Z_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
498 track->db_z_info |= Z_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
499 }
498 } 500 }
499 break; 501 break;
500 case DB_STENCIL_INFO: 502 case DB_STENCIL_INFO:
@@ -607,40 +609,44 @@ static int evergreen_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
607 case CB_COLOR5_INFO: 609 case CB_COLOR5_INFO:
608 case CB_COLOR6_INFO: 610 case CB_COLOR6_INFO:
609 case CB_COLOR7_INFO: 611 case CB_COLOR7_INFO:
610 r = evergreen_cs_packet_next_reloc(p, &reloc);
611 if (r) {
612 dev_warn(p->dev, "bad SET_CONTEXT_REG "
613 "0x%04X\n", reg);
614 return -EINVAL;
615 }
616 tmp = (reg - CB_COLOR0_INFO) / 0x3c; 612 tmp = (reg - CB_COLOR0_INFO) / 0x3c;
617 track->cb_color_info[tmp] = radeon_get_ib_value(p, idx); 613 track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
618 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) { 614 if (!p->keep_tiling_flags) {
619 ib[idx] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 615 r = evergreen_cs_packet_next_reloc(p, &reloc);
620 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 616 if (r) {
621 } else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) { 617 dev_warn(p->dev, "bad SET_CONTEXT_REG "
622 ib[idx] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 618 "0x%04X\n", reg);
623 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 619 return -EINVAL;
620 }
621 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
622 ib[idx] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
623 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
624 } else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) {
625 ib[idx] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
626 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
627 }
624 } 628 }
625 break; 629 break;
626 case CB_COLOR8_INFO: 630 case CB_COLOR8_INFO:
627 case CB_COLOR9_INFO: 631 case CB_COLOR9_INFO:
628 case CB_COLOR10_INFO: 632 case CB_COLOR10_INFO:
629 case CB_COLOR11_INFO: 633 case CB_COLOR11_INFO:
630 r = evergreen_cs_packet_next_reloc(p, &reloc);
631 if (r) {
632 dev_warn(p->dev, "bad SET_CONTEXT_REG "
633 "0x%04X\n", reg);
634 return -EINVAL;
635 }
636 tmp = ((reg - CB_COLOR8_INFO) / 0x1c) + 8; 634 tmp = ((reg - CB_COLOR8_INFO) / 0x1c) + 8;
637 track->cb_color_info[tmp] = radeon_get_ib_value(p, idx); 635 track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
638 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) { 636 if (!p->keep_tiling_flags) {
639 ib[idx] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 637 r = evergreen_cs_packet_next_reloc(p, &reloc);
640 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 638 if (r) {
641 } else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) { 639 dev_warn(p->dev, "bad SET_CONTEXT_REG "
642 ib[idx] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 640 "0x%04X\n", reg);
643 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 641 return -EINVAL;
642 }
643 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
644 ib[idx] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
645 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
646 } else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) {
647 ib[idx] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
648 track->cb_color_info[tmp] |= CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
649 }
644 } 650 }
645 break; 651 break;
646 case CB_COLOR0_PITCH: 652 case CB_COLOR0_PITCH:
@@ -1311,10 +1317,12 @@ static int evergreen_packet3_check(struct radeon_cs_parser *p,
1311 return -EINVAL; 1317 return -EINVAL;
1312 } 1318 }
1313 ib[idx+1+(i*8)+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); 1319 ib[idx+1+(i*8)+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
1314 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 1320 if (!p->keep_tiling_flags) {
1315 ib[idx+1+(i*8)+1] |= TEX_ARRAY_MODE(ARRAY_2D_TILED_THIN1); 1321 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
1316 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 1322 ib[idx+1+(i*8)+1] |= TEX_ARRAY_MODE(ARRAY_2D_TILED_THIN1);
1317 ib[idx+1+(i*8)+1] |= TEX_ARRAY_MODE(ARRAY_1D_TILED_THIN1); 1323 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
1324 ib[idx+1+(i*8)+1] |= TEX_ARRAY_MODE(ARRAY_1D_TILED_THIN1);
1325 }
1318 texture = reloc->robj; 1326 texture = reloc->robj;
1319 /* tex mip base */ 1327 /* tex mip base */
1320 r = evergreen_cs_packet_next_reloc(p, &reloc); 1328 r = evergreen_cs_packet_next_reloc(p, &reloc);
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index 400b26df652a..c93bc64707e1 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -701,16 +701,21 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
701 return r; 701 return r;
702 } 702 }
703 703
704 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 704 if (p->keep_tiling_flags) {
705 tile_flags |= R300_TXO_MACRO_TILE; 705 ib[idx] = (idx_value & 31) | /* keep the 1st 5 bits */
706 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 706 ((idx_value & ~31) + (u32)reloc->lobj.gpu_offset);
707 tile_flags |= R300_TXO_MICRO_TILE; 707 } else {
708 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE) 708 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
709 tile_flags |= R300_TXO_MICRO_TILE_SQUARE; 709 tile_flags |= R300_TXO_MACRO_TILE;
710 710 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
711 tmp = idx_value + ((u32)reloc->lobj.gpu_offset); 711 tile_flags |= R300_TXO_MICRO_TILE;
712 tmp |= tile_flags; 712 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
713 ib[idx] = tmp; 713 tile_flags |= R300_TXO_MICRO_TILE_SQUARE;
714
715 tmp = idx_value + ((u32)reloc->lobj.gpu_offset);
716 tmp |= tile_flags;
717 ib[idx] = tmp;
718 }
714 track->textures[i].robj = reloc->robj; 719 track->textures[i].robj = reloc->robj;
715 track->tex_dirty = true; 720 track->tex_dirty = true;
716 break; 721 break;
@@ -760,24 +765,26 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
760 /* RB3D_COLORPITCH1 */ 765 /* RB3D_COLORPITCH1 */
761 /* RB3D_COLORPITCH2 */ 766 /* RB3D_COLORPITCH2 */
762 /* RB3D_COLORPITCH3 */ 767 /* RB3D_COLORPITCH3 */
763 r = r100_cs_packet_next_reloc(p, &reloc); 768 if (!p->keep_tiling_flags) {
764 if (r) { 769 r = r100_cs_packet_next_reloc(p, &reloc);
765 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 770 if (r) {
766 idx, reg); 771 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
767 r100_cs_dump_packet(p, pkt); 772 idx, reg);
768 return r; 773 r100_cs_dump_packet(p, pkt);
769 } 774 return r;
775 }
770 776
771 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 777 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
772 tile_flags |= R300_COLOR_TILE_ENABLE; 778 tile_flags |= R300_COLOR_TILE_ENABLE;
773 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 779 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
774 tile_flags |= R300_COLOR_MICROTILE_ENABLE; 780 tile_flags |= R300_COLOR_MICROTILE_ENABLE;
775 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE) 781 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
776 tile_flags |= R300_COLOR_MICROTILE_SQUARE_ENABLE; 782 tile_flags |= R300_COLOR_MICROTILE_SQUARE_ENABLE;
777 783
778 tmp = idx_value & ~(0x7 << 16); 784 tmp = idx_value & ~(0x7 << 16);
779 tmp |= tile_flags; 785 tmp |= tile_flags;
780 ib[idx] = tmp; 786 ib[idx] = tmp;
787 }
781 i = (reg - 0x4E38) >> 2; 788 i = (reg - 0x4E38) >> 2;
782 track->cb[i].pitch = idx_value & 0x3FFE; 789 track->cb[i].pitch = idx_value & 0x3FFE;
783 switch (((idx_value >> 21) & 0xF)) { 790 switch (((idx_value >> 21) & 0xF)) {
@@ -843,25 +850,26 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
843 break; 850 break;
844 case 0x4F24: 851 case 0x4F24:
845 /* ZB_DEPTHPITCH */ 852 /* ZB_DEPTHPITCH */
846 r = r100_cs_packet_next_reloc(p, &reloc); 853 if (!p->keep_tiling_flags) {
847 if (r) { 854 r = r100_cs_packet_next_reloc(p, &reloc);
848 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 855 if (r) {
849 idx, reg); 856 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
850 r100_cs_dump_packet(p, pkt); 857 idx, reg);
851 return r; 858 r100_cs_dump_packet(p, pkt);
852 } 859 return r;
853 860 }
854 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
855 tile_flags |= R300_DEPTHMACROTILE_ENABLE;
856 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
857 tile_flags |= R300_DEPTHMICROTILE_TILED;
858 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
859 tile_flags |= R300_DEPTHMICROTILE_TILED_SQUARE;
860 861
861 tmp = idx_value & ~(0x7 << 16); 862 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
862 tmp |= tile_flags; 863 tile_flags |= R300_DEPTHMACROTILE_ENABLE;
863 ib[idx] = tmp; 864 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
865 tile_flags |= R300_DEPTHMICROTILE_TILED;
866 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO_SQUARE)
867 tile_flags |= R300_DEPTHMICROTILE_TILED_SQUARE;
864 868
869 tmp = idx_value & ~(0x7 << 16);
870 tmp |= tile_flags;
871 ib[idx] = tmp;
872 }
865 track->zb.pitch = idx_value & 0x3FFC; 873 track->zb.pitch = idx_value & 0x3FFC;
866 track->zb_dirty = true; 874 track->zb_dirty = true;
867 break; 875 break;
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index 0a2e023c1557..cb1acffd2430 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -941,7 +941,8 @@ static int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
941 track->db_depth_control = radeon_get_ib_value(p, idx); 941 track->db_depth_control = radeon_get_ib_value(p, idx);
942 break; 942 break;
943 case R_028010_DB_DEPTH_INFO: 943 case R_028010_DB_DEPTH_INFO:
944 if (r600_cs_packet_next_is_pkt3_nop(p)) { 944 if (!p->keep_tiling_flags &&
945 r600_cs_packet_next_is_pkt3_nop(p)) {
945 r = r600_cs_packet_next_reloc(p, &reloc); 946 r = r600_cs_packet_next_reloc(p, &reloc);
946 if (r) { 947 if (r) {
947 dev_warn(p->dev, "bad SET_CONTEXT_REG " 948 dev_warn(p->dev, "bad SET_CONTEXT_REG "
@@ -992,7 +993,8 @@ static int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
992 case R_0280B4_CB_COLOR5_INFO: 993 case R_0280B4_CB_COLOR5_INFO:
993 case R_0280B8_CB_COLOR6_INFO: 994 case R_0280B8_CB_COLOR6_INFO:
994 case R_0280BC_CB_COLOR7_INFO: 995 case R_0280BC_CB_COLOR7_INFO:
995 if (r600_cs_packet_next_is_pkt3_nop(p)) { 996 if (!p->keep_tiling_flags &&
997 r600_cs_packet_next_is_pkt3_nop(p)) {
996 r = r600_cs_packet_next_reloc(p, &reloc); 998 r = r600_cs_packet_next_reloc(p, &reloc);
997 if (r) { 999 if (r) {
998 dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg); 1000 dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
@@ -1291,10 +1293,12 @@ static int r600_check_texture_resource(struct radeon_cs_parser *p, u32 idx,
1291 mip_offset <<= 8; 1293 mip_offset <<= 8;
1292 1294
1293 word0 = radeon_get_ib_value(p, idx + 0); 1295 word0 = radeon_get_ib_value(p, idx + 0);
1294 if (tiling_flags & RADEON_TILING_MACRO) 1296 if (!p->keep_tiling_flags) {
1295 word0 |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1); 1297 if (tiling_flags & RADEON_TILING_MACRO)
1296 else if (tiling_flags & RADEON_TILING_MICRO) 1298 word0 |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1);
1297 word0 |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1); 1299 else if (tiling_flags & RADEON_TILING_MICRO)
1300 word0 |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1);
1301 }
1298 word1 = radeon_get_ib_value(p, idx + 1); 1302 word1 = radeon_get_ib_value(p, idx + 1);
1299 w0 = G_038000_TEX_WIDTH(word0) + 1; 1303 w0 = G_038000_TEX_WIDTH(word0) + 1;
1300 h0 = G_038004_TEX_HEIGHT(word1) + 1; 1304 h0 = G_038004_TEX_HEIGHT(word1) + 1;
@@ -1621,10 +1625,12 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
1621 return -EINVAL; 1625 return -EINVAL;
1622 } 1626 }
1623 base_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); 1627 base_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
1624 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 1628 if (!p->keep_tiling_flags) {
1625 ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1); 1629 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
1626 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 1630 ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1);
1627 ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1); 1631 else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
1632 ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1);
1633 }
1628 texture = reloc->robj; 1634 texture = reloc->robj;
1629 /* tex mip base */ 1635 /* tex mip base */
1630 r = r600_cs_packet_next_reloc(p, &reloc); 1636 r = r600_cs_packet_next_reloc(p, &reloc);
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index fc5a1d642cb5..8227e76b5c70 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -611,7 +611,8 @@ struct radeon_cs_parser {
611 struct radeon_ib *ib; 611 struct radeon_ib *ib;
612 void *track; 612 void *track;
613 unsigned family; 613 unsigned family;
614 int parser_error; 614 int parser_error;
615 bool keep_tiling_flags;
615}; 616};
616 617
617extern int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx); 618extern int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx);
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index ccaa243c1442..29afd71e0840 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -93,7 +93,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
93{ 93{
94 struct drm_radeon_cs *cs = data; 94 struct drm_radeon_cs *cs = data;
95 uint64_t *chunk_array_ptr; 95 uint64_t *chunk_array_ptr;
96 unsigned size, i; 96 unsigned size, i, flags = 0;
97 97
98 if (!cs->num_chunks) { 98 if (!cs->num_chunks) {
99 return 0; 99 return 0;
@@ -140,6 +140,10 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
140 if (p->chunks[i].length_dw == 0) 140 if (p->chunks[i].length_dw == 0)
141 return -EINVAL; 141 return -EINVAL;
142 } 142 }
143 if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS &&
144 !p->chunks[i].length_dw) {
145 return -EINVAL;
146 }
143 147
144 p->chunks[i].length_dw = user_chunk.length_dw; 148 p->chunks[i].length_dw = user_chunk.length_dw;
145 p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data; 149 p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data;
@@ -155,6 +159,9 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
155 p->chunks[i].user_ptr, size)) { 159 p->chunks[i].user_ptr, size)) {
156 return -EFAULT; 160 return -EFAULT;
157 } 161 }
162 if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
163 flags = p->chunks[i].kdata[0];
164 }
158 } else { 165 } else {
159 p->chunks[i].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL); 166 p->chunks[i].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
160 p->chunks[i].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL); 167 p->chunks[i].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
@@ -174,6 +181,8 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
174 p->chunks[p->chunk_ib_idx].length_dw); 181 p->chunks[p->chunk_ib_idx].length_dw);
175 return -EINVAL; 182 return -EINVAL;
176 } 183 }
184
185 p->keep_tiling_flags = (flags & RADEON_CS_KEEP_TILING_FLAGS) != 0;
177 return 0; 186 return 0;
178} 187}
179 188
diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
index a0b35e909489..71499fc3daf5 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.c
+++ b/drivers/gpu/drm/radeon/radeon_drv.c
@@ -53,9 +53,10 @@
53 * 2.9.0 - r600 tiling (s3tc,rgtc) working, SET_PREDICATION packet 3 on r600 + eg, backend query 53 * 2.9.0 - r600 tiling (s3tc,rgtc) working, SET_PREDICATION packet 3 on r600 + eg, backend query
54 * 2.10.0 - fusion 2D tiling 54 * 2.10.0 - fusion 2D tiling
55 * 2.11.0 - backend map, initial compute support for the CS checker 55 * 2.11.0 - backend map, initial compute support for the CS checker
56 * 2.12.0 - RADEON_CS_KEEP_TILING_FLAGS
56 */ 57 */
57#define KMS_DRIVER_MAJOR 2 58#define KMS_DRIVER_MAJOR 2
58#define KMS_DRIVER_MINOR 11 59#define KMS_DRIVER_MINOR 12
59#define KMS_DRIVER_PATCHLEVEL 0 60#define KMS_DRIVER_PATCHLEVEL 0
60int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags); 61int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags);
61int radeon_driver_unload_kms(struct drm_device *dev); 62int radeon_driver_unload_kms(struct drm_device *dev);
diff --git a/include/drm/radeon_drm.h b/include/drm/radeon_drm.h
index b65be6054a18..be94be6d6f17 100644
--- a/include/drm/radeon_drm.h
+++ b/include/drm/radeon_drm.h
@@ -874,6 +874,10 @@ struct drm_radeon_gem_pwrite {
874 874
875#define RADEON_CHUNK_ID_RELOCS 0x01 875#define RADEON_CHUNK_ID_RELOCS 0x01
876#define RADEON_CHUNK_ID_IB 0x02 876#define RADEON_CHUNK_ID_IB 0x02
877#define RADEON_CHUNK_ID_FLAGS 0x03
878
879/* The first dword of RADEON_CHUNK_ID_FLAGS is a uint32 of these flags: */
880#define RADEON_CS_KEEP_TILING_FLAGS 0x01
877 881
878struct drm_radeon_cs_chunk { 882struct drm_radeon_cs_chunk {
879 uint32_t chunk_id; 883 uint32_t chunk_id;