aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/radeon
diff options
context:
space:
mode:
authorDave Airlie <airlied@redhat.com>2009-09-23 02:56:27 -0400
committerDave Airlie <airlied@linux.ie>2009-09-24 23:08:18 -0400
commit513bcb4655e68706594e45dfa1d4b181500110ba (patch)
treeed457db4cfb202015866a131ad4e742503728fad /drivers/gpu/drm/radeon
parent35e4b7af21d77933abda3d41d1672589eb6c960c (diff)
drm/radeon/kms: don't require up to 64k allocations. (v2)
This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
Diffstat (limited to 'drivers/gpu/drm/radeon')
-rw-r--r--drivers/gpu/drm/radeon/r100.c188
-rw-r--r--drivers/gpu/drm/radeon/r100_track.h69
-rw-r--r--drivers/gpu/drm/radeon/r200.c79
-rw-r--r--drivers/gpu/drm/radeon/r300.c137
-rw-r--r--drivers/gpu/drm/radeon/r600_cs.c26
-rw-r--r--drivers/gpu/drm/radeon/radeon.h37
-rw-r--r--drivers/gpu/drm/radeon/radeon_cs.c105
7 files changed, 370 insertions, 271 deletions
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index 737970b43ae..9ab976d97e9 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -863,13 +863,11 @@ int r100_cs_parse_packet0(struct radeon_cs_parser *p,
863void r100_cs_dump_packet(struct radeon_cs_parser *p, 863void r100_cs_dump_packet(struct radeon_cs_parser *p,
864 struct radeon_cs_packet *pkt) 864 struct radeon_cs_packet *pkt)
865{ 865{
866 struct radeon_cs_chunk *ib_chunk;
867 volatile uint32_t *ib; 866 volatile uint32_t *ib;
868 unsigned i; 867 unsigned i;
869 unsigned idx; 868 unsigned idx;
870 869
871 ib = p->ib->ptr; 870 ib = p->ib->ptr;
872 ib_chunk = &p->chunks[p->chunk_ib_idx];
873 idx = pkt->idx; 871 idx = pkt->idx;
874 for (i = 0; i <= (pkt->count + 1); i++, idx++) { 872 for (i = 0; i <= (pkt->count + 1); i++, idx++) {
875 DRM_INFO("ib[%d]=0x%08X\n", idx, ib[idx]); 873 DRM_INFO("ib[%d]=0x%08X\n", idx, ib[idx]);
@@ -896,7 +894,7 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p,
896 idx, ib_chunk->length_dw); 894 idx, ib_chunk->length_dw);
897 return -EINVAL; 895 return -EINVAL;
898 } 896 }
899 header = ib_chunk->kdata[idx]; 897 header = radeon_get_ib_value(p, idx);
900 pkt->idx = idx; 898 pkt->idx = idx;
901 pkt->type = CP_PACKET_GET_TYPE(header); 899 pkt->type = CP_PACKET_GET_TYPE(header);
902 pkt->count = CP_PACKET_GET_COUNT(header); 900 pkt->count = CP_PACKET_GET_COUNT(header);
@@ -939,7 +937,6 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p,
939 */ 937 */
940int r100_cs_packet_parse_vline(struct radeon_cs_parser *p) 938int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
941{ 939{
942 struct radeon_cs_chunk *ib_chunk;
943 struct drm_mode_object *obj; 940 struct drm_mode_object *obj;
944 struct drm_crtc *crtc; 941 struct drm_crtc *crtc;
945 struct radeon_crtc *radeon_crtc; 942 struct radeon_crtc *radeon_crtc;
@@ -947,8 +944,9 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
947 int crtc_id; 944 int crtc_id;
948 int r; 945 int r;
949 uint32_t header, h_idx, reg; 946 uint32_t header, h_idx, reg;
947 volatile uint32_t *ib;
950 948
951 ib_chunk = &p->chunks[p->chunk_ib_idx]; 949 ib = p->ib->ptr;
952 950
953 /* parse the wait until */ 951 /* parse the wait until */
954 r = r100_cs_packet_parse(p, &waitreloc, p->idx); 952 r = r100_cs_packet_parse(p, &waitreloc, p->idx);
@@ -963,7 +961,7 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
963 return r; 961 return r;
964 } 962 }
965 963
966 if (ib_chunk->kdata[waitreloc.idx + 1] != RADEON_WAIT_CRTC_VLINE) { 964 if (radeon_get_ib_value(p, waitreloc.idx + 1) != RADEON_WAIT_CRTC_VLINE) {
967 DRM_ERROR("vline wait had illegal wait until\n"); 965 DRM_ERROR("vline wait had illegal wait until\n");
968 r = -EINVAL; 966 r = -EINVAL;
969 return r; 967 return r;
@@ -978,9 +976,9 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
978 p->idx += waitreloc.count; 976 p->idx += waitreloc.count;
979 p->idx += p3reloc.count; 977 p->idx += p3reloc.count;
980 978
981 header = ib_chunk->kdata[h_idx]; 979 header = radeon_get_ib_value(p, h_idx);
982 crtc_id = ib_chunk->kdata[h_idx + 5]; 980 crtc_id = radeon_get_ib_value(p, h_idx + 5);
983 reg = ib_chunk->kdata[h_idx] >> 2; 981 reg = header >> 2;
984 mutex_lock(&p->rdev->ddev->mode_config.mutex); 982 mutex_lock(&p->rdev->ddev->mode_config.mutex);
985 obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC); 983 obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC);
986 if (!obj) { 984 if (!obj) {
@@ -994,8 +992,9 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
994 992
995 if (!crtc->enabled) { 993 if (!crtc->enabled) {
996 /* if the CRTC isn't enabled - we need to nop out the wait until */ 994 /* if the CRTC isn't enabled - we need to nop out the wait until */
997 ib_chunk->kdata[h_idx + 2] = PACKET2(0); 995
998 ib_chunk->kdata[h_idx + 3] = PACKET2(0); 996 ib[h_idx + 2] = PACKET2(0);
997 ib[h_idx + 3] = PACKET2(0);
999 } else if (crtc_id == 1) { 998 } else if (crtc_id == 1) {
1000 switch (reg) { 999 switch (reg) {
1001 case AVIVO_D1MODE_VLINE_START_END: 1000 case AVIVO_D1MODE_VLINE_START_END:
@@ -1011,8 +1010,8 @@ int r100_cs_packet_parse_vline(struct radeon_cs_parser *p)
1011 r = -EINVAL; 1010 r = -EINVAL;
1012 goto out; 1011 goto out;
1013 } 1012 }
1014 ib_chunk->kdata[h_idx] = header; 1013 ib[h_idx] = header;
1015 ib_chunk->kdata[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1; 1014 ib[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1;
1016 } 1015 }
1017out: 1016out:
1018 mutex_unlock(&p->rdev->ddev->mode_config.mutex); 1017 mutex_unlock(&p->rdev->ddev->mode_config.mutex);
@@ -1033,7 +1032,6 @@ out:
1033int r100_cs_packet_next_reloc(struct radeon_cs_parser *p, 1032int r100_cs_packet_next_reloc(struct radeon_cs_parser *p,
1034 struct radeon_cs_reloc **cs_reloc) 1033 struct radeon_cs_reloc **cs_reloc)
1035{ 1034{
1036 struct radeon_cs_chunk *ib_chunk;
1037 struct radeon_cs_chunk *relocs_chunk; 1035 struct radeon_cs_chunk *relocs_chunk;
1038 struct radeon_cs_packet p3reloc; 1036 struct radeon_cs_packet p3reloc;
1039 unsigned idx; 1037 unsigned idx;
@@ -1044,7 +1042,6 @@ int r100_cs_packet_next_reloc(struct radeon_cs_parser *p,
1044 return -EINVAL; 1042 return -EINVAL;
1045 } 1043 }
1046 *cs_reloc = NULL; 1044 *cs_reloc = NULL;
1047 ib_chunk = &p->chunks[p->chunk_ib_idx];
1048 relocs_chunk = &p->chunks[p->chunk_relocs_idx]; 1045 relocs_chunk = &p->chunks[p->chunk_relocs_idx];
1049 r = r100_cs_packet_parse(p, &p3reloc, p->idx); 1046 r = r100_cs_packet_parse(p, &p3reloc, p->idx);
1050 if (r) { 1047 if (r) {
@@ -1057,7 +1054,7 @@ int r100_cs_packet_next_reloc(struct radeon_cs_parser *p,
1057 r100_cs_dump_packet(p, &p3reloc); 1054 r100_cs_dump_packet(p, &p3reloc);
1058 return -EINVAL; 1055 return -EINVAL;
1059 } 1056 }
1060 idx = ib_chunk->kdata[p3reloc.idx + 1]; 1057 idx = radeon_get_ib_value(p, p3reloc.idx + 1);
1061 if (idx >= relocs_chunk->length_dw) { 1058 if (idx >= relocs_chunk->length_dw) {
1062 DRM_ERROR("Relocs at %d after relocations chunk end %d !\n", 1059 DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
1063 idx, relocs_chunk->length_dw); 1060 idx, relocs_chunk->length_dw);
@@ -1126,7 +1123,6 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1126 struct radeon_cs_packet *pkt, 1123 struct radeon_cs_packet *pkt,
1127 unsigned idx, unsigned reg) 1124 unsigned idx, unsigned reg)
1128{ 1125{
1129 struct radeon_cs_chunk *ib_chunk;
1130 struct radeon_cs_reloc *reloc; 1126 struct radeon_cs_reloc *reloc;
1131 struct r100_cs_track *track; 1127 struct r100_cs_track *track;
1132 volatile uint32_t *ib; 1128 volatile uint32_t *ib;
@@ -1134,11 +1130,13 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1134 int r; 1130 int r;
1135 int i, face; 1131 int i, face;
1136 u32 tile_flags = 0; 1132 u32 tile_flags = 0;
1133 u32 idx_value;
1137 1134
1138 ib = p->ib->ptr; 1135 ib = p->ib->ptr;
1139 ib_chunk = &p->chunks[p->chunk_ib_idx];
1140 track = (struct r100_cs_track *)p->track; 1136 track = (struct r100_cs_track *)p->track;
1141 1137
1138 idx_value = radeon_get_ib_value(p, idx);
1139
1142 switch (reg) { 1140 switch (reg) {
1143 case RADEON_CRTC_GUI_TRIG_VLINE: 1141 case RADEON_CRTC_GUI_TRIG_VLINE:
1144 r = r100_cs_packet_parse_vline(p); 1142 r = r100_cs_packet_parse_vline(p);
@@ -1166,8 +1164,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1166 return r; 1164 return r;
1167 } 1165 }
1168 track->zb.robj = reloc->robj; 1166 track->zb.robj = reloc->robj;
1169 track->zb.offset = ib_chunk->kdata[idx]; 1167 track->zb.offset = idx_value;
1170 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1168 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1171 break; 1169 break;
1172 case RADEON_RB3D_COLOROFFSET: 1170 case RADEON_RB3D_COLOROFFSET:
1173 r = r100_cs_packet_next_reloc(p, &reloc); 1171 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1178,8 +1176,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1178 return r; 1176 return r;
1179 } 1177 }
1180 track->cb[0].robj = reloc->robj; 1178 track->cb[0].robj = reloc->robj;
1181 track->cb[0].offset = ib_chunk->kdata[idx]; 1179 track->cb[0].offset = idx_value;
1182 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1180 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1183 break; 1181 break;
1184 case RADEON_PP_TXOFFSET_0: 1182 case RADEON_PP_TXOFFSET_0:
1185 case RADEON_PP_TXOFFSET_1: 1183 case RADEON_PP_TXOFFSET_1:
@@ -1192,7 +1190,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1192 r100_cs_dump_packet(p, pkt); 1190 r100_cs_dump_packet(p, pkt);
1193 return r; 1191 return r;
1194 } 1192 }
1195 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1193 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1196 track->textures[i].robj = reloc->robj; 1194 track->textures[i].robj = reloc->robj;
1197 break; 1195 break;
1198 case RADEON_PP_CUBIC_OFFSET_T0_0: 1196 case RADEON_PP_CUBIC_OFFSET_T0_0:
@@ -1208,8 +1206,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1208 r100_cs_dump_packet(p, pkt); 1206 r100_cs_dump_packet(p, pkt);
1209 return r; 1207 return r;
1210 } 1208 }
1211 track->textures[0].cube_info[i].offset = ib_chunk->kdata[idx]; 1209 track->textures[0].cube_info[i].offset = idx_value;
1212 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1210 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1213 track->textures[0].cube_info[i].robj = reloc->robj; 1211 track->textures[0].cube_info[i].robj = reloc->robj;
1214 break; 1212 break;
1215 case RADEON_PP_CUBIC_OFFSET_T1_0: 1213 case RADEON_PP_CUBIC_OFFSET_T1_0:
@@ -1225,8 +1223,8 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1225 r100_cs_dump_packet(p, pkt); 1223 r100_cs_dump_packet(p, pkt);
1226 return r; 1224 return r;
1227 } 1225 }
1228 track->textures[1].cube_info[i].offset = ib_chunk->kdata[idx]; 1226 track->textures[1].cube_info[i].offset = idx_value;
1229 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1227 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1230 track->textures[1].cube_info[i].robj = reloc->robj; 1228 track->textures[1].cube_info[i].robj = reloc->robj;
1231 break; 1229 break;
1232 case RADEON_PP_CUBIC_OFFSET_T2_0: 1230 case RADEON_PP_CUBIC_OFFSET_T2_0:
@@ -1242,12 +1240,12 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1242 r100_cs_dump_packet(p, pkt); 1240 r100_cs_dump_packet(p, pkt);
1243 return r; 1241 return r;
1244 } 1242 }
1245 track->textures[2].cube_info[i].offset = ib_chunk->kdata[idx]; 1243 track->textures[2].cube_info[i].offset = idx_value;
1246 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1244 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1247 track->textures[2].cube_info[i].robj = reloc->robj; 1245 track->textures[2].cube_info[i].robj = reloc->robj;
1248 break; 1246 break;
1249 case RADEON_RE_WIDTH_HEIGHT: 1247 case RADEON_RE_WIDTH_HEIGHT:
1250 track->maxy = ((ib_chunk->kdata[idx] >> 16) & 0x7FF); 1248 track->maxy = ((idx_value >> 16) & 0x7FF);
1251 break; 1249 break;
1252 case RADEON_RB3D_COLORPITCH: 1250 case RADEON_RB3D_COLORPITCH:
1253 r = r100_cs_packet_next_reloc(p, &reloc); 1251 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1263,17 +1261,17 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1263 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 1261 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
1264 tile_flags |= RADEON_COLOR_MICROTILE_ENABLE; 1262 tile_flags |= RADEON_COLOR_MICROTILE_ENABLE;
1265 1263
1266 tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); 1264 tmp = idx_value & ~(0x7 << 16);
1267 tmp |= tile_flags; 1265 tmp |= tile_flags;
1268 ib[idx] = tmp; 1266 ib[idx] = tmp;
1269 1267
1270 track->cb[0].pitch = ib_chunk->kdata[idx] & RADEON_COLORPITCH_MASK; 1268 track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
1271 break; 1269 break;
1272 case RADEON_RB3D_DEPTHPITCH: 1270 case RADEON_RB3D_DEPTHPITCH:
1273 track->zb.pitch = ib_chunk->kdata[idx] & RADEON_DEPTHPITCH_MASK; 1271 track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
1274 break; 1272 break;
1275 case RADEON_RB3D_CNTL: 1273 case RADEON_RB3D_CNTL:
1276 switch ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) { 1274 switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
1277 case 7: 1275 case 7:
1278 case 8: 1276 case 8:
1279 case 9: 1277 case 9:
@@ -1291,13 +1289,13 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1291 break; 1289 break;
1292 default: 1290 default:
1293 DRM_ERROR("Invalid color buffer format (%d) !\n", 1291 DRM_ERROR("Invalid color buffer format (%d) !\n",
1294 ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f)); 1292 ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
1295 return -EINVAL; 1293 return -EINVAL;
1296 } 1294 }
1297 track->z_enabled = !!(ib_chunk->kdata[idx] & RADEON_Z_ENABLE); 1295 track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
1298 break; 1296 break;
1299 case RADEON_RB3D_ZSTENCILCNTL: 1297 case RADEON_RB3D_ZSTENCILCNTL:
1300 switch (ib_chunk->kdata[idx] & 0xf) { 1298 switch (idx_value & 0xf) {
1301 case 0: 1299 case 0:
1302 track->zb.cpp = 2; 1300 track->zb.cpp = 2;
1303 break; 1301 break;
@@ -1321,44 +1319,44 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1321 r100_cs_dump_packet(p, pkt); 1319 r100_cs_dump_packet(p, pkt);
1322 return r; 1320 return r;
1323 } 1321 }
1324 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1322 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1325 break; 1323 break;
1326 case RADEON_PP_CNTL: 1324 case RADEON_PP_CNTL:
1327 { 1325 {
1328 uint32_t temp = ib_chunk->kdata[idx] >> 4; 1326 uint32_t temp = idx_value >> 4;
1329 for (i = 0; i < track->num_texture; i++) 1327 for (i = 0; i < track->num_texture; i++)
1330 track->textures[i].enabled = !!(temp & (1 << i)); 1328 track->textures[i].enabled = !!(temp & (1 << i));
1331 } 1329 }
1332 break; 1330 break;
1333 case RADEON_SE_VF_CNTL: 1331 case RADEON_SE_VF_CNTL:
1334 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1332 track->vap_vf_cntl = idx_value;
1335 break; 1333 break;
1336 case RADEON_SE_VTX_FMT: 1334 case RADEON_SE_VTX_FMT:
1337 track->vtx_size = r100_get_vtx_size(ib_chunk->kdata[idx]); 1335 track->vtx_size = r100_get_vtx_size(idx_value);
1338 break; 1336 break;
1339 case RADEON_PP_TEX_SIZE_0: 1337 case RADEON_PP_TEX_SIZE_0:
1340 case RADEON_PP_TEX_SIZE_1: 1338 case RADEON_PP_TEX_SIZE_1:
1341 case RADEON_PP_TEX_SIZE_2: 1339 case RADEON_PP_TEX_SIZE_2:
1342 i = (reg - RADEON_PP_TEX_SIZE_0) / 8; 1340 i = (reg - RADEON_PP_TEX_SIZE_0) / 8;
1343 track->textures[i].width = (ib_chunk->kdata[idx] & RADEON_TEX_USIZE_MASK) + 1; 1341 track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
1344 track->textures[i].height = ((ib_chunk->kdata[idx] & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1; 1342 track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
1345 break; 1343 break;
1346 case RADEON_PP_TEX_PITCH_0: 1344 case RADEON_PP_TEX_PITCH_0:
1347 case RADEON_PP_TEX_PITCH_1: 1345 case RADEON_PP_TEX_PITCH_1:
1348 case RADEON_PP_TEX_PITCH_2: 1346 case RADEON_PP_TEX_PITCH_2:
1349 i = (reg - RADEON_PP_TEX_PITCH_0) / 8; 1347 i = (reg - RADEON_PP_TEX_PITCH_0) / 8;
1350 track->textures[i].pitch = ib_chunk->kdata[idx] + 32; 1348 track->textures[i].pitch = idx_value + 32;
1351 break; 1349 break;
1352 case RADEON_PP_TXFILTER_0: 1350 case RADEON_PP_TXFILTER_0:
1353 case RADEON_PP_TXFILTER_1: 1351 case RADEON_PP_TXFILTER_1:
1354 case RADEON_PP_TXFILTER_2: 1352 case RADEON_PP_TXFILTER_2:
1355 i = (reg - RADEON_PP_TXFILTER_0) / 24; 1353 i = (reg - RADEON_PP_TXFILTER_0) / 24;
1356 track->textures[i].num_levels = ((ib_chunk->kdata[idx] & RADEON_MAX_MIP_LEVEL_MASK) 1354 track->textures[i].num_levels = ((idx_value & RADEON_MAX_MIP_LEVEL_MASK)
1357 >> RADEON_MAX_MIP_LEVEL_SHIFT); 1355 >> RADEON_MAX_MIP_LEVEL_SHIFT);
1358 tmp = (ib_chunk->kdata[idx] >> 23) & 0x7; 1356 tmp = (idx_value >> 23) & 0x7;
1359 if (tmp == 2 || tmp == 6) 1357 if (tmp == 2 || tmp == 6)
1360 track->textures[i].roundup_w = false; 1358 track->textures[i].roundup_w = false;
1361 tmp = (ib_chunk->kdata[idx] >> 27) & 0x7; 1359 tmp = (idx_value >> 27) & 0x7;
1362 if (tmp == 2 || tmp == 6) 1360 if (tmp == 2 || tmp == 6)
1363 track->textures[i].roundup_h = false; 1361 track->textures[i].roundup_h = false;
1364 break; 1362 break;
@@ -1366,16 +1364,16 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1366 case RADEON_PP_TXFORMAT_1: 1364 case RADEON_PP_TXFORMAT_1:
1367 case RADEON_PP_TXFORMAT_2: 1365 case RADEON_PP_TXFORMAT_2:
1368 i = (reg - RADEON_PP_TXFORMAT_0) / 24; 1366 i = (reg - RADEON_PP_TXFORMAT_0) / 24;
1369 if (ib_chunk->kdata[idx] & RADEON_TXFORMAT_NON_POWER2) { 1367 if (idx_value & RADEON_TXFORMAT_NON_POWER2) {
1370 track->textures[i].use_pitch = 1; 1368 track->textures[i].use_pitch = 1;
1371 } else { 1369 } else {
1372 track->textures[i].use_pitch = 0; 1370 track->textures[i].use_pitch = 0;
1373 track->textures[i].width = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK); 1371 track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
1374 track->textures[i].height = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK); 1372 track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
1375 } 1373 }
1376 if (ib_chunk->kdata[idx] & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) 1374 if (idx_value & RADEON_TXFORMAT_CUBIC_MAP_ENABLE)
1377 track->textures[i].tex_coord_type = 2; 1375 track->textures[i].tex_coord_type = 2;
1378 switch ((ib_chunk->kdata[idx] & RADEON_TXFORMAT_FORMAT_MASK)) { 1376 switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) {
1379 case RADEON_TXFORMAT_I8: 1377 case RADEON_TXFORMAT_I8:
1380 case RADEON_TXFORMAT_RGB332: 1378 case RADEON_TXFORMAT_RGB332:
1381 case RADEON_TXFORMAT_Y8: 1379 case RADEON_TXFORMAT_Y8:
@@ -1402,13 +1400,13 @@ static int r100_packet0_check(struct radeon_cs_parser *p,
1402 track->textures[i].cpp = 4; 1400 track->textures[i].cpp = 4;
1403 break; 1401 break;
1404 } 1402 }
1405 track->textures[i].cube_info[4].width = 1 << ((ib_chunk->kdata[idx] >> 16) & 0xf); 1403 track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
1406 track->textures[i].cube_info[4].height = 1 << ((ib_chunk->kdata[idx] >> 20) & 0xf); 1404 track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
1407 break; 1405 break;
1408 case RADEON_PP_CUBIC_FACES_0: 1406 case RADEON_PP_CUBIC_FACES_0:
1409 case RADEON_PP_CUBIC_FACES_1: 1407 case RADEON_PP_CUBIC_FACES_1:
1410 case RADEON_PP_CUBIC_FACES_2: 1408 case RADEON_PP_CUBIC_FACES_2:
1411 tmp = ib_chunk->kdata[idx]; 1409 tmp = idx_value;
1412 i = (reg - RADEON_PP_CUBIC_FACES_0) / 4; 1410 i = (reg - RADEON_PP_CUBIC_FACES_0) / 4;
1413 for (face = 0; face < 4; face++) { 1411 for (face = 0; face < 4; face++) {
1414 track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf); 1412 track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
@@ -1427,15 +1425,14 @@ int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p,
1427 struct radeon_cs_packet *pkt, 1425 struct radeon_cs_packet *pkt,
1428 struct radeon_object *robj) 1426 struct radeon_object *robj)
1429{ 1427{
1430 struct radeon_cs_chunk *ib_chunk;
1431 unsigned idx; 1428 unsigned idx;
1432 1429 u32 value;
1433 ib_chunk = &p->chunks[p->chunk_ib_idx];
1434 idx = pkt->idx + 1; 1430 idx = pkt->idx + 1;
1435 if ((ib_chunk->kdata[idx+2] + 1) > radeon_object_size(robj)) { 1431 value = radeon_get_ib_value(p, idx + 2);
1432 if ((value + 1) > radeon_object_size(robj)) {
1436 DRM_ERROR("[drm] Buffer too small for PACKET3 INDX_BUFFER " 1433 DRM_ERROR("[drm] Buffer too small for PACKET3 INDX_BUFFER "
1437 "(need %u have %lu) !\n", 1434 "(need %u have %lu) !\n",
1438 ib_chunk->kdata[idx+2] + 1, 1435 value + 1,
1439 radeon_object_size(robj)); 1436 radeon_object_size(robj));
1440 return -EINVAL; 1437 return -EINVAL;
1441 } 1438 }
@@ -1445,59 +1442,20 @@ int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p,
1445static int r100_packet3_check(struct radeon_cs_parser *p, 1442static int r100_packet3_check(struct radeon_cs_parser *p,
1446 struct radeon_cs_packet *pkt) 1443 struct radeon_cs_packet *pkt)
1447{ 1444{
1448 struct radeon_cs_chunk *ib_chunk;
1449 struct radeon_cs_reloc *reloc; 1445 struct radeon_cs_reloc *reloc;
1450 struct r100_cs_track *track; 1446 struct r100_cs_track *track;
1451 unsigned idx; 1447 unsigned idx;
1452 unsigned i, c;
1453 volatile uint32_t *ib; 1448 volatile uint32_t *ib;
1454 int r; 1449 int r;
1455 1450
1456 ib = p->ib->ptr; 1451 ib = p->ib->ptr;
1457 ib_chunk = &p->chunks[p->chunk_ib_idx];
1458 idx = pkt->idx + 1; 1452 idx = pkt->idx + 1;
1459 track = (struct r100_cs_track *)p->track; 1453 track = (struct r100_cs_track *)p->track;
1460 switch (pkt->opcode) { 1454 switch (pkt->opcode) {
1461 case PACKET3_3D_LOAD_VBPNTR: 1455 case PACKET3_3D_LOAD_VBPNTR:
1462 c = ib_chunk->kdata[idx++]; 1456 r = r100_packet3_load_vbpntr(p, pkt, idx);
1463 track->num_arrays = c; 1457 if (r)
1464 for (i = 0; i < (c - 1); i += 2, idx += 3) { 1458 return r;
1465 r = r100_cs_packet_next_reloc(p, &reloc);
1466 if (r) {
1467 DRM_ERROR("No reloc for packet3 %d\n",
1468 pkt->opcode);
1469 r100_cs_dump_packet(p, pkt);
1470 return r;
1471 }
1472 ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
1473 track->arrays[i + 0].robj = reloc->robj;
1474 track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
1475 track->arrays[i + 0].esize &= 0x7F;
1476 r = r100_cs_packet_next_reloc(p, &reloc);
1477 if (r) {
1478 DRM_ERROR("No reloc for packet3 %d\n",
1479 pkt->opcode);
1480 r100_cs_dump_packet(p, pkt);
1481 return r;
1482 }
1483 ib[idx+2] = ib_chunk->kdata[idx+2] + ((u32)reloc->lobj.gpu_offset);
1484 track->arrays[i + 1].robj = reloc->robj;
1485 track->arrays[i + 1].esize = ib_chunk->kdata[idx] >> 24;
1486 track->arrays[i + 1].esize &= 0x7F;
1487 }
1488 if (c & 1) {
1489 r = r100_cs_packet_next_reloc(p, &reloc);
1490 if (r) {
1491 DRM_ERROR("No reloc for packet3 %d\n",
1492 pkt->opcode);
1493 r100_cs_dump_packet(p, pkt);
1494 return r;
1495 }
1496 ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
1497 track->arrays[i + 0].robj = reloc->robj;
1498 track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
1499 track->arrays[i + 0].esize &= 0x7F;
1500 }
1501 break; 1459 break;
1502 case PACKET3_INDX_BUFFER: 1460 case PACKET3_INDX_BUFFER:
1503 r = r100_cs_packet_next_reloc(p, &reloc); 1461 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1506,7 +1464,7 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
1506 r100_cs_dump_packet(p, pkt); 1464 r100_cs_dump_packet(p, pkt);
1507 return r; 1465 return r;
1508 } 1466 }
1509 ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); 1467 ib[idx+1] = radeon_get_ib_value(p, idx+1) + ((u32)reloc->lobj.gpu_offset);
1510 r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj); 1468 r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj);
1511 if (r) { 1469 if (r) {
1512 return r; 1470 return r;
@@ -1520,27 +1478,27 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
1520 r100_cs_dump_packet(p, pkt); 1478 r100_cs_dump_packet(p, pkt);
1521 return r; 1479 return r;
1522 } 1480 }
1523 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1481 ib[idx] = radeon_get_ib_value(p, idx) + ((u32)reloc->lobj.gpu_offset);
1524 track->num_arrays = 1; 1482 track->num_arrays = 1;
1525 track->vtx_size = r100_get_vtx_size(ib_chunk->kdata[idx+2]); 1483 track->vtx_size = r100_get_vtx_size(radeon_get_ib_value(p, idx + 2));
1526 1484
1527 track->arrays[0].robj = reloc->robj; 1485 track->arrays[0].robj = reloc->robj;
1528 track->arrays[0].esize = track->vtx_size; 1486 track->arrays[0].esize = track->vtx_size;
1529 1487
1530 track->max_indx = ib_chunk->kdata[idx+1]; 1488 track->max_indx = radeon_get_ib_value(p, idx+1);
1531 1489
1532 track->vap_vf_cntl = ib_chunk->kdata[idx+3]; 1490 track->vap_vf_cntl = radeon_get_ib_value(p, idx+3);
1533 track->immd_dwords = pkt->count - 1; 1491 track->immd_dwords = pkt->count - 1;
1534 r = r100_cs_track_check(p->rdev, track); 1492 r = r100_cs_track_check(p->rdev, track);
1535 if (r) 1493 if (r)
1536 return r; 1494 return r;
1537 break; 1495 break;
1538 case PACKET3_3D_DRAW_IMMD: 1496 case PACKET3_3D_DRAW_IMMD:
1539 if (((ib_chunk->kdata[idx+1] >> 4) & 0x3) != 3) { 1497 if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) {
1540 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); 1498 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1541 return -EINVAL; 1499 return -EINVAL;
1542 } 1500 }
1543 track->vap_vf_cntl = ib_chunk->kdata[idx+1]; 1501 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1544 track->immd_dwords = pkt->count - 1; 1502 track->immd_dwords = pkt->count - 1;
1545 r = r100_cs_track_check(p->rdev, track); 1503 r = r100_cs_track_check(p->rdev, track);
1546 if (r) 1504 if (r)
@@ -1548,11 +1506,11 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
1548 break; 1506 break;
1549 /* triggers drawing using in-packet vertex data */ 1507 /* triggers drawing using in-packet vertex data */
1550 case PACKET3_3D_DRAW_IMMD_2: 1508 case PACKET3_3D_DRAW_IMMD_2:
1551 if (((ib_chunk->kdata[idx] >> 4) & 0x3) != 3) { 1509 if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) {
1552 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); 1510 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1553 return -EINVAL; 1511 return -EINVAL;
1554 } 1512 }
1555 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1513 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1556 track->immd_dwords = pkt->count; 1514 track->immd_dwords = pkt->count;
1557 r = r100_cs_track_check(p->rdev, track); 1515 r = r100_cs_track_check(p->rdev, track);
1558 if (r) 1516 if (r)
@@ -1560,28 +1518,28 @@ static int r100_packet3_check(struct radeon_cs_parser *p,
1560 break; 1518 break;
1561 /* triggers drawing using in-packet vertex data */ 1519 /* triggers drawing using in-packet vertex data */
1562 case PACKET3_3D_DRAW_VBUF_2: 1520 case PACKET3_3D_DRAW_VBUF_2:
1563 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1521 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1564 r = r100_cs_track_check(p->rdev, track); 1522 r = r100_cs_track_check(p->rdev, track);
1565 if (r) 1523 if (r)
1566 return r; 1524 return r;
1567 break; 1525 break;
1568 /* triggers drawing of vertex buffers setup elsewhere */ 1526 /* triggers drawing of vertex buffers setup elsewhere */
1569 case PACKET3_3D_DRAW_INDX_2: 1527 case PACKET3_3D_DRAW_INDX_2:
1570 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1528 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1571 r = r100_cs_track_check(p->rdev, track); 1529 r = r100_cs_track_check(p->rdev, track);
1572 if (r) 1530 if (r)
1573 return r; 1531 return r;
1574 break; 1532 break;
1575 /* triggers drawing using indices to vertex buffer */ 1533 /* triggers drawing using indices to vertex buffer */
1576 case PACKET3_3D_DRAW_VBUF: 1534 case PACKET3_3D_DRAW_VBUF:
1577 track->vap_vf_cntl = ib_chunk->kdata[idx + 1]; 1535 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1578 r = r100_cs_track_check(p->rdev, track); 1536 r = r100_cs_track_check(p->rdev, track);
1579 if (r) 1537 if (r)
1580 return r; 1538 return r;
1581 break; 1539 break;
1582 /* triggers drawing of vertex buffers setup elsewhere */ 1540 /* triggers drawing of vertex buffers setup elsewhere */
1583 case PACKET3_3D_DRAW_INDX: 1541 case PACKET3_3D_DRAW_INDX:
1584 track->vap_vf_cntl = ib_chunk->kdata[idx + 1]; 1542 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1585 r = r100_cs_track_check(p->rdev, track); 1543 r = r100_cs_track_check(p->rdev, track);
1586 if (r) 1544 if (r)
1587 return r; 1545 return r;
diff --git a/drivers/gpu/drm/radeon/r100_track.h b/drivers/gpu/drm/radeon/r100_track.h
index 70a82eda394..0daf0d76a89 100644
--- a/drivers/gpu/drm/radeon/r100_track.h
+++ b/drivers/gpu/drm/radeon/r100_track.h
@@ -84,6 +84,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
84 struct radeon_cs_packet *pkt, 84 struct radeon_cs_packet *pkt,
85 unsigned idx, unsigned reg); 85 unsigned idx, unsigned reg);
86 86
87
88
87static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p, 89static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
88 struct radeon_cs_packet *pkt, 90 struct radeon_cs_packet *pkt,
89 unsigned idx, 91 unsigned idx,
@@ -93,9 +95,7 @@ static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
93 u32 tile_flags = 0; 95 u32 tile_flags = 0;
94 u32 tmp; 96 u32 tmp;
95 struct radeon_cs_reloc *reloc; 97 struct radeon_cs_reloc *reloc;
96 struct radeon_cs_chunk *ib_chunk; 98 u32 value;
97
98 ib_chunk = &p->chunks[p->chunk_ib_idx];
99 99
100 r = r100_cs_packet_next_reloc(p, &reloc); 100 r = r100_cs_packet_next_reloc(p, &reloc);
101 if (r) { 101 if (r) {
@@ -104,7 +104,8 @@ static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
104 r100_cs_dump_packet(p, pkt); 104 r100_cs_dump_packet(p, pkt);
105 return r; 105 return r;
106 } 106 }
107 tmp = ib_chunk->kdata[idx] & 0x003fffff; 107 value = radeon_get_ib_value(p, idx);
108 tmp = value & 0x003fffff;
108 tmp += (((u32)reloc->lobj.gpu_offset) >> 10); 109 tmp += (((u32)reloc->lobj.gpu_offset) >> 10);
109 110
110 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 111 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
@@ -119,6 +120,64 @@ static inline int r100_reloc_pitch_offset(struct radeon_cs_parser *p,
119 } 120 }
120 121
121 tmp |= tile_flags; 122 tmp |= tile_flags;
122 p->ib->ptr[idx] = (ib_chunk->kdata[idx] & 0x3fc00000) | tmp; 123 p->ib->ptr[idx] = (value & 0x3fc00000) | tmp;
123 return 0; 124 return 0;
124} 125}
126
127static inline int r100_packet3_load_vbpntr(struct radeon_cs_parser *p,
128 struct radeon_cs_packet *pkt,
129 int idx)
130{
131 unsigned c, i;
132 struct radeon_cs_reloc *reloc;
133 struct r100_cs_track *track;
134 int r = 0;
135 volatile uint32_t *ib;
136 u32 idx_value;
137
138 ib = p->ib->ptr;
139 track = (struct r100_cs_track *)p->track;
140 c = radeon_get_ib_value(p, idx++) & 0x1F;
141 track->num_arrays = c;
142 for (i = 0; i < (c - 1); i+=2, idx+=3) {
143 r = r100_cs_packet_next_reloc(p, &reloc);
144 if (r) {
145 DRM_ERROR("No reloc for packet3 %d\n",
146 pkt->opcode);
147 r100_cs_dump_packet(p, pkt);
148 return r;
149 }
150 idx_value = radeon_get_ib_value(p, idx);
151 ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
152
153 track->arrays[i + 0].esize = idx_value >> 8;
154 track->arrays[i + 0].robj = reloc->robj;
155 track->arrays[i + 0].esize &= 0x7F;
156 r = r100_cs_packet_next_reloc(p, &reloc);
157 if (r) {
158 DRM_ERROR("No reloc for packet3 %d\n",
159 pkt->opcode);
160 r100_cs_dump_packet(p, pkt);
161 return r;
162 }
163 ib[idx+2] = radeon_get_ib_value(p, idx + 2) + ((u32)reloc->lobj.gpu_offset);
164 track->arrays[i + 1].robj = reloc->robj;
165 track->arrays[i + 1].esize = idx_value >> 24;
166 track->arrays[i + 1].esize &= 0x7F;
167 }
168 if (c & 1) {
169 r = r100_cs_packet_next_reloc(p, &reloc);
170 if (r) {
171 DRM_ERROR("No reloc for packet3 %d\n",
172 pkt->opcode);
173 r100_cs_dump_packet(p, pkt);
174 return r;
175 }
176 idx_value = radeon_get_ib_value(p, idx);
177 ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
178 track->arrays[i + 0].robj = reloc->robj;
179 track->arrays[i + 0].esize = idx_value >> 8;
180 track->arrays[i + 0].esize &= 0x7F;
181 }
182 return r;
183}
diff --git a/drivers/gpu/drm/radeon/r200.c b/drivers/gpu/drm/radeon/r200.c
index 568c74bfba3..cf7fea5ff2e 100644
--- a/drivers/gpu/drm/radeon/r200.c
+++ b/drivers/gpu/drm/radeon/r200.c
@@ -96,7 +96,6 @@ int r200_packet0_check(struct radeon_cs_parser *p,
96 struct radeon_cs_packet *pkt, 96 struct radeon_cs_packet *pkt,
97 unsigned idx, unsigned reg) 97 unsigned idx, unsigned reg)
98{ 98{
99 struct radeon_cs_chunk *ib_chunk;
100 struct radeon_cs_reloc *reloc; 99 struct radeon_cs_reloc *reloc;
101 struct r100_cs_track *track; 100 struct r100_cs_track *track;
102 volatile uint32_t *ib; 101 volatile uint32_t *ib;
@@ -105,11 +104,11 @@ int r200_packet0_check(struct radeon_cs_parser *p,
105 int i; 104 int i;
106 int face; 105 int face;
107 u32 tile_flags = 0; 106 u32 tile_flags = 0;
107 u32 idx_value;
108 108
109 ib = p->ib->ptr; 109 ib = p->ib->ptr;
110 ib_chunk = &p->chunks[p->chunk_ib_idx];
111 track = (struct r100_cs_track *)p->track; 110 track = (struct r100_cs_track *)p->track;
112 111 idx_value = radeon_get_ib_value(p, idx);
113 switch (reg) { 112 switch (reg) {
114 case RADEON_CRTC_GUI_TRIG_VLINE: 113 case RADEON_CRTC_GUI_TRIG_VLINE:
115 r = r100_cs_packet_parse_vline(p); 114 r = r100_cs_packet_parse_vline(p);
@@ -137,8 +136,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
137 return r; 136 return r;
138 } 137 }
139 track->zb.robj = reloc->robj; 138 track->zb.robj = reloc->robj;
140 track->zb.offset = ib_chunk->kdata[idx]; 139 track->zb.offset = idx_value;
141 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 140 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
142 break; 141 break;
143 case RADEON_RB3D_COLOROFFSET: 142 case RADEON_RB3D_COLOROFFSET:
144 r = r100_cs_packet_next_reloc(p, &reloc); 143 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -149,8 +148,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
149 return r; 148 return r;
150 } 149 }
151 track->cb[0].robj = reloc->robj; 150 track->cb[0].robj = reloc->robj;
152 track->cb[0].offset = ib_chunk->kdata[idx]; 151 track->cb[0].offset = idx_value;
153 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 152 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
154 break; 153 break;
155 case R200_PP_TXOFFSET_0: 154 case R200_PP_TXOFFSET_0:
156 case R200_PP_TXOFFSET_1: 155 case R200_PP_TXOFFSET_1:
@@ -166,7 +165,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
166 r100_cs_dump_packet(p, pkt); 165 r100_cs_dump_packet(p, pkt);
167 return r; 166 return r;
168 } 167 }
169 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 168 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
170 track->textures[i].robj = reloc->robj; 169 track->textures[i].robj = reloc->robj;
171 break; 170 break;
172 case R200_PP_CUBIC_OFFSET_F1_0: 171 case R200_PP_CUBIC_OFFSET_F1_0:
@@ -208,12 +207,12 @@ int r200_packet0_check(struct radeon_cs_parser *p,
208 r100_cs_dump_packet(p, pkt); 207 r100_cs_dump_packet(p, pkt);
209 return r; 208 return r;
210 } 209 }
211 track->textures[i].cube_info[face - 1].offset = ib_chunk->kdata[idx]; 210 track->textures[i].cube_info[face - 1].offset = idx_value;
212 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 211 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
213 track->textures[i].cube_info[face - 1].robj = reloc->robj; 212 track->textures[i].cube_info[face - 1].robj = reloc->robj;
214 break; 213 break;
215 case RADEON_RE_WIDTH_HEIGHT: 214 case RADEON_RE_WIDTH_HEIGHT:
216 track->maxy = ((ib_chunk->kdata[idx] >> 16) & 0x7FF); 215 track->maxy = ((idx_value >> 16) & 0x7FF);
217 break; 216 break;
218 case RADEON_RB3D_COLORPITCH: 217 case RADEON_RB3D_COLORPITCH:
219 r = r100_cs_packet_next_reloc(p, &reloc); 218 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -229,17 +228,17 @@ int r200_packet0_check(struct radeon_cs_parser *p,
229 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 228 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
230 tile_flags |= RADEON_COLOR_MICROTILE_ENABLE; 229 tile_flags |= RADEON_COLOR_MICROTILE_ENABLE;
231 230
232 tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); 231 tmp = idx_value & ~(0x7 << 16);
233 tmp |= tile_flags; 232 tmp |= tile_flags;
234 ib[idx] = tmp; 233 ib[idx] = tmp;
235 234
236 track->cb[0].pitch = ib_chunk->kdata[idx] & RADEON_COLORPITCH_MASK; 235 track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
237 break; 236 break;
238 case RADEON_RB3D_DEPTHPITCH: 237 case RADEON_RB3D_DEPTHPITCH:
239 track->zb.pitch = ib_chunk->kdata[idx] & RADEON_DEPTHPITCH_MASK; 238 track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
240 break; 239 break;
241 case RADEON_RB3D_CNTL: 240 case RADEON_RB3D_CNTL:
242 switch ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) { 241 switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
243 case 7: 242 case 7:
244 case 8: 243 case 8:
245 case 9: 244 case 9:
@@ -257,18 +256,18 @@ int r200_packet0_check(struct radeon_cs_parser *p,
257 break; 256 break;
258 default: 257 default:
259 DRM_ERROR("Invalid color buffer format (%d) !\n", 258 DRM_ERROR("Invalid color buffer format (%d) !\n",
260 ((ib_chunk->kdata[idx] >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f)); 259 ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
261 return -EINVAL; 260 return -EINVAL;
262 } 261 }
263 if (ib_chunk->kdata[idx] & RADEON_DEPTHXY_OFFSET_ENABLE) { 262 if (idx_value & RADEON_DEPTHXY_OFFSET_ENABLE) {
264 DRM_ERROR("No support for depth xy offset in kms\n"); 263 DRM_ERROR("No support for depth xy offset in kms\n");
265 return -EINVAL; 264 return -EINVAL;
266 } 265 }
267 266
268 track->z_enabled = !!(ib_chunk->kdata[idx] & RADEON_Z_ENABLE); 267 track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
269 break; 268 break;
270 case RADEON_RB3D_ZSTENCILCNTL: 269 case RADEON_RB3D_ZSTENCILCNTL:
271 switch (ib_chunk->kdata[idx] & 0xf) { 270 switch (idx_value & 0xf) {
272 case 0: 271 case 0:
273 track->zb.cpp = 2; 272 track->zb.cpp = 2;
274 break; 273 break;
@@ -292,27 +291,27 @@ int r200_packet0_check(struct radeon_cs_parser *p,
292 r100_cs_dump_packet(p, pkt); 291 r100_cs_dump_packet(p, pkt);
293 return r; 292 return r;
294 } 293 }
295 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 294 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
296 break; 295 break;
297 case RADEON_PP_CNTL: 296 case RADEON_PP_CNTL:
298 { 297 {
299 uint32_t temp = ib_chunk->kdata[idx] >> 4; 298 uint32_t temp = idx_value >> 4;
300 for (i = 0; i < track->num_texture; i++) 299 for (i = 0; i < track->num_texture; i++)
301 track->textures[i].enabled = !!(temp & (1 << i)); 300 track->textures[i].enabled = !!(temp & (1 << i));
302 } 301 }
303 break; 302 break;
304 case RADEON_SE_VF_CNTL: 303 case RADEON_SE_VF_CNTL:
305 track->vap_vf_cntl = ib_chunk->kdata[idx]; 304 track->vap_vf_cntl = idx_value;
306 break; 305 break;
307 case 0x210c: 306 case 0x210c:
308 /* VAP_VF_MAX_VTX_INDX */ 307 /* VAP_VF_MAX_VTX_INDX */
309 track->max_indx = ib_chunk->kdata[idx] & 0x00FFFFFFUL; 308 track->max_indx = idx_value & 0x00FFFFFFUL;
310 break; 309 break;
311 case R200_SE_VTX_FMT_0: 310 case R200_SE_VTX_FMT_0:
312 track->vtx_size = r200_get_vtx_size_0(ib_chunk->kdata[idx]); 311 track->vtx_size = r200_get_vtx_size_0(idx_value);
313 break; 312 break;
314 case R200_SE_VTX_FMT_1: 313 case R200_SE_VTX_FMT_1:
315 track->vtx_size += r200_get_vtx_size_1(ib_chunk->kdata[idx]); 314 track->vtx_size += r200_get_vtx_size_1(idx_value);
316 break; 315 break;
317 case R200_PP_TXSIZE_0: 316 case R200_PP_TXSIZE_0:
318 case R200_PP_TXSIZE_1: 317 case R200_PP_TXSIZE_1:
@@ -321,8 +320,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
321 case R200_PP_TXSIZE_4: 320 case R200_PP_TXSIZE_4:
322 case R200_PP_TXSIZE_5: 321 case R200_PP_TXSIZE_5:
323 i = (reg - R200_PP_TXSIZE_0) / 32; 322 i = (reg - R200_PP_TXSIZE_0) / 32;
324 track->textures[i].width = (ib_chunk->kdata[idx] & RADEON_TEX_USIZE_MASK) + 1; 323 track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
325 track->textures[i].height = ((ib_chunk->kdata[idx] & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1; 324 track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
326 break; 325 break;
327 case R200_PP_TXPITCH_0: 326 case R200_PP_TXPITCH_0:
328 case R200_PP_TXPITCH_1: 327 case R200_PP_TXPITCH_1:
@@ -331,7 +330,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
331 case R200_PP_TXPITCH_4: 330 case R200_PP_TXPITCH_4:
332 case R200_PP_TXPITCH_5: 331 case R200_PP_TXPITCH_5:
333 i = (reg - R200_PP_TXPITCH_0) / 32; 332 i = (reg - R200_PP_TXPITCH_0) / 32;
334 track->textures[i].pitch = ib_chunk->kdata[idx] + 32; 333 track->textures[i].pitch = idx_value + 32;
335 break; 334 break;
336 case R200_PP_TXFILTER_0: 335 case R200_PP_TXFILTER_0:
337 case R200_PP_TXFILTER_1: 336 case R200_PP_TXFILTER_1:
@@ -340,12 +339,12 @@ int r200_packet0_check(struct radeon_cs_parser *p,
340 case R200_PP_TXFILTER_4: 339 case R200_PP_TXFILTER_4:
341 case R200_PP_TXFILTER_5: 340 case R200_PP_TXFILTER_5:
342 i = (reg - R200_PP_TXFILTER_0) / 32; 341 i = (reg - R200_PP_TXFILTER_0) / 32;
343 track->textures[i].num_levels = ((ib_chunk->kdata[idx] & R200_MAX_MIP_LEVEL_MASK) 342 track->textures[i].num_levels = ((idx_value & R200_MAX_MIP_LEVEL_MASK)
344 >> R200_MAX_MIP_LEVEL_SHIFT); 343 >> R200_MAX_MIP_LEVEL_SHIFT);
345 tmp = (ib_chunk->kdata[idx] >> 23) & 0x7; 344 tmp = (idx_value >> 23) & 0x7;
346 if (tmp == 2 || tmp == 6) 345 if (tmp == 2 || tmp == 6)
347 track->textures[i].roundup_w = false; 346 track->textures[i].roundup_w = false;
348 tmp = (ib_chunk->kdata[idx] >> 27) & 0x7; 347 tmp = (idx_value >> 27) & 0x7;
349 if (tmp == 2 || tmp == 6) 348 if (tmp == 2 || tmp == 6)
350 track->textures[i].roundup_h = false; 349 track->textures[i].roundup_h = false;
351 break; 350 break;
@@ -364,8 +363,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
364 case R200_PP_TXFORMAT_X_4: 363 case R200_PP_TXFORMAT_X_4:
365 case R200_PP_TXFORMAT_X_5: 364 case R200_PP_TXFORMAT_X_5:
366 i = (reg - R200_PP_TXFORMAT_X_0) / 32; 365 i = (reg - R200_PP_TXFORMAT_X_0) / 32;
367 track->textures[i].txdepth = ib_chunk->kdata[idx] & 0x7; 366 track->textures[i].txdepth = idx_value & 0x7;
368 tmp = (ib_chunk->kdata[idx] >> 16) & 0x3; 367 tmp = (idx_value >> 16) & 0x3;
369 /* 2D, 3D, CUBE */ 368 /* 2D, 3D, CUBE */
370 switch (tmp) { 369 switch (tmp) {
371 case 0: 370 case 0:
@@ -389,14 +388,14 @@ int r200_packet0_check(struct radeon_cs_parser *p,
389 case R200_PP_TXFORMAT_4: 388 case R200_PP_TXFORMAT_4:
390 case R200_PP_TXFORMAT_5: 389 case R200_PP_TXFORMAT_5:
391 i = (reg - R200_PP_TXFORMAT_0) / 32; 390 i = (reg - R200_PP_TXFORMAT_0) / 32;
392 if (ib_chunk->kdata[idx] & R200_TXFORMAT_NON_POWER2) { 391 if (idx_value & R200_TXFORMAT_NON_POWER2) {
393 track->textures[i].use_pitch = 1; 392 track->textures[i].use_pitch = 1;
394 } else { 393 } else {
395 track->textures[i].use_pitch = 0; 394 track->textures[i].use_pitch = 0;
396 track->textures[i].width = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK); 395 track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
397 track->textures[i].height = 1 << ((ib_chunk->kdata[idx] >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK); 396 track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
398 } 397 }
399 switch ((ib_chunk->kdata[idx] & RADEON_TXFORMAT_FORMAT_MASK)) { 398 switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) {
400 case R200_TXFORMAT_I8: 399 case R200_TXFORMAT_I8:
401 case R200_TXFORMAT_RGB332: 400 case R200_TXFORMAT_RGB332:
402 case R200_TXFORMAT_Y8: 401 case R200_TXFORMAT_Y8:
@@ -424,8 +423,8 @@ int r200_packet0_check(struct radeon_cs_parser *p,
424 track->textures[i].cpp = 4; 423 track->textures[i].cpp = 4;
425 break; 424 break;
426 } 425 }
427 track->textures[i].cube_info[4].width = 1 << ((ib_chunk->kdata[idx] >> 16) & 0xf); 426 track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
428 track->textures[i].cube_info[4].height = 1 << ((ib_chunk->kdata[idx] >> 20) & 0xf); 427 track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
429 break; 428 break;
430 case R200_PP_CUBIC_FACES_0: 429 case R200_PP_CUBIC_FACES_0:
431 case R200_PP_CUBIC_FACES_1: 430 case R200_PP_CUBIC_FACES_1:
@@ -433,7 +432,7 @@ int r200_packet0_check(struct radeon_cs_parser *p,
433 case R200_PP_CUBIC_FACES_3: 432 case R200_PP_CUBIC_FACES_3:
434 case R200_PP_CUBIC_FACES_4: 433 case R200_PP_CUBIC_FACES_4:
435 case R200_PP_CUBIC_FACES_5: 434 case R200_PP_CUBIC_FACES_5:
436 tmp = ib_chunk->kdata[idx]; 435 tmp = idx_value;
437 i = (reg - R200_PP_CUBIC_FACES_0) / 32; 436 i = (reg - R200_PP_CUBIC_FACES_0) / 32;
438 for (face = 0; face < 4; face++) { 437 for (face = 0; face < 4; face++) {
439 track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf); 438 track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index bb151ecdf8f..1ebea8cc8c9 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -697,17 +697,18 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
697 struct radeon_cs_packet *pkt, 697 struct radeon_cs_packet *pkt,
698 unsigned idx, unsigned reg) 698 unsigned idx, unsigned reg)
699{ 699{
700 struct radeon_cs_chunk *ib_chunk;
701 struct radeon_cs_reloc *reloc; 700 struct radeon_cs_reloc *reloc;
702 struct r100_cs_track *track; 701 struct r100_cs_track *track;
703 volatile uint32_t *ib; 702 volatile uint32_t *ib;
704 uint32_t tmp, tile_flags = 0; 703 uint32_t tmp, tile_flags = 0;
705 unsigned i; 704 unsigned i;
706 int r; 705 int r;
706 u32 idx_value;
707 707
708 ib = p->ib->ptr; 708 ib = p->ib->ptr;
709 ib_chunk = &p->chunks[p->chunk_ib_idx];
710 track = (struct r100_cs_track *)p->track; 709 track = (struct r100_cs_track *)p->track;
710 idx_value = radeon_get_ib_value(p, idx);
711
711 switch(reg) { 712 switch(reg) {
712 case AVIVO_D1MODE_VLINE_START_END: 713 case AVIVO_D1MODE_VLINE_START_END:
713 case RADEON_CRTC_GUI_TRIG_VLINE: 714 case RADEON_CRTC_GUI_TRIG_VLINE:
@@ -738,8 +739,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
738 return r; 739 return r;
739 } 740 }
740 track->cb[i].robj = reloc->robj; 741 track->cb[i].robj = reloc->robj;
741 track->cb[i].offset = ib_chunk->kdata[idx]; 742 track->cb[i].offset = idx_value;
742 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 743 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
743 break; 744 break;
744 case R300_ZB_DEPTHOFFSET: 745 case R300_ZB_DEPTHOFFSET:
745 r = r100_cs_packet_next_reloc(p, &reloc); 746 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -750,8 +751,8 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
750 return r; 751 return r;
751 } 752 }
752 track->zb.robj = reloc->robj; 753 track->zb.robj = reloc->robj;
753 track->zb.offset = ib_chunk->kdata[idx]; 754 track->zb.offset = idx_value;
754 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 755 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
755 break; 756 break;
756 case R300_TX_OFFSET_0: 757 case R300_TX_OFFSET_0:
757 case R300_TX_OFFSET_0+4: 758 case R300_TX_OFFSET_0+4:
@@ -777,32 +778,32 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
777 r100_cs_dump_packet(p, pkt); 778 r100_cs_dump_packet(p, pkt);
778 return r; 779 return r;
779 } 780 }
780 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 781 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
781 track->textures[i].robj = reloc->robj; 782 track->textures[i].robj = reloc->robj;
782 break; 783 break;
783 /* Tracked registers */ 784 /* Tracked registers */
784 case 0x2084: 785 case 0x2084:
785 /* VAP_VF_CNTL */ 786 /* VAP_VF_CNTL */
786 track->vap_vf_cntl = ib_chunk->kdata[idx]; 787 track->vap_vf_cntl = idx_value;
787 break; 788 break;
788 case 0x20B4: 789 case 0x20B4:
789 /* VAP_VTX_SIZE */ 790 /* VAP_VTX_SIZE */
790 track->vtx_size = ib_chunk->kdata[idx] & 0x7F; 791 track->vtx_size = idx_value & 0x7F;
791 break; 792 break;
792 case 0x2134: 793 case 0x2134:
793 /* VAP_VF_MAX_VTX_INDX */ 794 /* VAP_VF_MAX_VTX_INDX */
794 track->max_indx = ib_chunk->kdata[idx] & 0x00FFFFFFUL; 795 track->max_indx = idx_value & 0x00FFFFFFUL;
795 break; 796 break;
796 case 0x43E4: 797 case 0x43E4:
797 /* SC_SCISSOR1 */ 798 /* SC_SCISSOR1 */
798 track->maxy = ((ib_chunk->kdata[idx] >> 13) & 0x1FFF) + 1; 799 track->maxy = ((idx_value >> 13) & 0x1FFF) + 1;
799 if (p->rdev->family < CHIP_RV515) { 800 if (p->rdev->family < CHIP_RV515) {
800 track->maxy -= 1440; 801 track->maxy -= 1440;
801 } 802 }
802 break; 803 break;
803 case 0x4E00: 804 case 0x4E00:
804 /* RB3D_CCTL */ 805 /* RB3D_CCTL */
805 track->num_cb = ((ib_chunk->kdata[idx] >> 5) & 0x3) + 1; 806 track->num_cb = ((idx_value >> 5) & 0x3) + 1;
806 break; 807 break;
807 case 0x4E38: 808 case 0x4E38:
808 case 0x4E3C: 809 case 0x4E3C:
@@ -825,13 +826,13 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
825 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 826 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
826 tile_flags |= R300_COLOR_MICROTILE_ENABLE; 827 tile_flags |= R300_COLOR_MICROTILE_ENABLE;
827 828
828 tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); 829 tmp = idx_value & ~(0x7 << 16);
829 tmp |= tile_flags; 830 tmp |= tile_flags;
830 ib[idx] = tmp; 831 ib[idx] = tmp;
831 832
832 i = (reg - 0x4E38) >> 2; 833 i = (reg - 0x4E38) >> 2;
833 track->cb[i].pitch = ib_chunk->kdata[idx] & 0x3FFE; 834 track->cb[i].pitch = idx_value & 0x3FFE;
834 switch (((ib_chunk->kdata[idx] >> 21) & 0xF)) { 835 switch (((idx_value >> 21) & 0xF)) {
835 case 9: 836 case 9:
836 case 11: 837 case 11:
837 case 12: 838 case 12:
@@ -854,13 +855,13 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
854 break; 855 break;
855 default: 856 default:
856 DRM_ERROR("Invalid color buffer format (%d) !\n", 857 DRM_ERROR("Invalid color buffer format (%d) !\n",
857 ((ib_chunk->kdata[idx] >> 21) & 0xF)); 858 ((idx_value >> 21) & 0xF));
858 return -EINVAL; 859 return -EINVAL;
859 } 860 }
860 break; 861 break;
861 case 0x4F00: 862 case 0x4F00:
862 /* ZB_CNTL */ 863 /* ZB_CNTL */
863 if (ib_chunk->kdata[idx] & 2) { 864 if (idx_value & 2) {
864 track->z_enabled = true; 865 track->z_enabled = true;
865 } else { 866 } else {
866 track->z_enabled = false; 867 track->z_enabled = false;
@@ -868,7 +869,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
868 break; 869 break;
869 case 0x4F10: 870 case 0x4F10:
870 /* ZB_FORMAT */ 871 /* ZB_FORMAT */
871 switch ((ib_chunk->kdata[idx] & 0xF)) { 872 switch ((idx_value & 0xF)) {
872 case 0: 873 case 0:
873 case 1: 874 case 1:
874 track->zb.cpp = 2; 875 track->zb.cpp = 2;
@@ -878,7 +879,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
878 break; 879 break;
879 default: 880 default:
880 DRM_ERROR("Invalid z buffer format (%d) !\n", 881 DRM_ERROR("Invalid z buffer format (%d) !\n",
881 (ib_chunk->kdata[idx] & 0xF)); 882 (idx_value & 0xF));
882 return -EINVAL; 883 return -EINVAL;
883 } 884 }
884 break; 885 break;
@@ -897,17 +898,17 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
897 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 898 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
898 tile_flags |= R300_DEPTHMICROTILE_TILED;; 899 tile_flags |= R300_DEPTHMICROTILE_TILED;;
899 900
900 tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); 901 tmp = idx_value & ~(0x7 << 16);
901 tmp |= tile_flags; 902 tmp |= tile_flags;
902 ib[idx] = tmp; 903 ib[idx] = tmp;
903 904
904 track->zb.pitch = ib_chunk->kdata[idx] & 0x3FFC; 905 track->zb.pitch = idx_value & 0x3FFC;
905 break; 906 break;
906 case 0x4104: 907 case 0x4104:
907 for (i = 0; i < 16; i++) { 908 for (i = 0; i < 16; i++) {
908 bool enabled; 909 bool enabled;
909 910
910 enabled = !!(ib_chunk->kdata[idx] & (1 << i)); 911 enabled = !!(idx_value & (1 << i));
911 track->textures[i].enabled = enabled; 912 track->textures[i].enabled = enabled;
912 } 913 }
913 break; 914 break;
@@ -929,9 +930,9 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
929 case 0x44FC: 930 case 0x44FC:
930 /* TX_FORMAT1_[0-15] */ 931 /* TX_FORMAT1_[0-15] */
931 i = (reg - 0x44C0) >> 2; 932 i = (reg - 0x44C0) >> 2;
932 tmp = (ib_chunk->kdata[idx] >> 25) & 0x3; 933 tmp = (idx_value >> 25) & 0x3;
933 track->textures[i].tex_coord_type = tmp; 934 track->textures[i].tex_coord_type = tmp;
934 switch ((ib_chunk->kdata[idx] & 0x1F)) { 935 switch ((idx_value & 0x1F)) {
935 case R300_TX_FORMAT_X8: 936 case R300_TX_FORMAT_X8:
936 case R300_TX_FORMAT_Y4X4: 937 case R300_TX_FORMAT_Y4X4:
937 case R300_TX_FORMAT_Z3Y3X2: 938 case R300_TX_FORMAT_Z3Y3X2:
@@ -971,7 +972,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
971 break; 972 break;
972 default: 973 default:
973 DRM_ERROR("Invalid texture format %u\n", 974 DRM_ERROR("Invalid texture format %u\n",
974 (ib_chunk->kdata[idx] & 0x1F)); 975 (idx_value & 0x1F));
975 return -EINVAL; 976 return -EINVAL;
976 break; 977 break;
977 } 978 }
@@ -994,11 +995,11 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
994 case 0x443C: 995 case 0x443C:
995 /* TX_FILTER0_[0-15] */ 996 /* TX_FILTER0_[0-15] */
996 i = (reg - 0x4400) >> 2; 997 i = (reg - 0x4400) >> 2;
997 tmp = ib_chunk->kdata[idx] & 0x7; 998 tmp = idx_value & 0x7;
998 if (tmp == 2 || tmp == 4 || tmp == 6) { 999 if (tmp == 2 || tmp == 4 || tmp == 6) {
999 track->textures[i].roundup_w = false; 1000 track->textures[i].roundup_w = false;
1000 } 1001 }
1001 tmp = (ib_chunk->kdata[idx] >> 3) & 0x7; 1002 tmp = (idx_value >> 3) & 0x7;
1002 if (tmp == 2 || tmp == 4 || tmp == 6) { 1003 if (tmp == 2 || tmp == 4 || tmp == 6) {
1003 track->textures[i].roundup_h = false; 1004 track->textures[i].roundup_h = false;
1004 } 1005 }
@@ -1021,12 +1022,12 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
1021 case 0x453C: 1022 case 0x453C:
1022 /* TX_FORMAT2_[0-15] */ 1023 /* TX_FORMAT2_[0-15] */
1023 i = (reg - 0x4500) >> 2; 1024 i = (reg - 0x4500) >> 2;
1024 tmp = ib_chunk->kdata[idx] & 0x3FFF; 1025 tmp = idx_value & 0x3FFF;
1025 track->textures[i].pitch = tmp + 1; 1026 track->textures[i].pitch = tmp + 1;
1026 if (p->rdev->family >= CHIP_RV515) { 1027 if (p->rdev->family >= CHIP_RV515) {
1027 tmp = ((ib_chunk->kdata[idx] >> 15) & 1) << 11; 1028 tmp = ((idx_value >> 15) & 1) << 11;
1028 track->textures[i].width_11 = tmp; 1029 track->textures[i].width_11 = tmp;
1029 tmp = ((ib_chunk->kdata[idx] >> 16) & 1) << 11; 1030 tmp = ((idx_value >> 16) & 1) << 11;
1030 track->textures[i].height_11 = tmp; 1031 track->textures[i].height_11 = tmp;
1031 } 1032 }
1032 break; 1033 break;
@@ -1048,15 +1049,15 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
1048 case 0x44BC: 1049 case 0x44BC:
1049 /* TX_FORMAT0_[0-15] */ 1050 /* TX_FORMAT0_[0-15] */
1050 i = (reg - 0x4480) >> 2; 1051 i = (reg - 0x4480) >> 2;
1051 tmp = ib_chunk->kdata[idx] & 0x7FF; 1052 tmp = idx_value & 0x7FF;
1052 track->textures[i].width = tmp + 1; 1053 track->textures[i].width = tmp + 1;
1053 tmp = (ib_chunk->kdata[idx] >> 11) & 0x7FF; 1054 tmp = (idx_value >> 11) & 0x7FF;
1054 track->textures[i].height = tmp + 1; 1055 track->textures[i].height = tmp + 1;
1055 tmp = (ib_chunk->kdata[idx] >> 26) & 0xF; 1056 tmp = (idx_value >> 26) & 0xF;
1056 track->textures[i].num_levels = tmp; 1057 track->textures[i].num_levels = tmp;
1057 tmp = ib_chunk->kdata[idx] & (1 << 31); 1058 tmp = idx_value & (1 << 31);
1058 track->textures[i].use_pitch = !!tmp; 1059 track->textures[i].use_pitch = !!tmp;
1059 tmp = (ib_chunk->kdata[idx] >> 22) & 0xF; 1060 tmp = (idx_value >> 22) & 0xF;
1060 track->textures[i].txdepth = tmp; 1061 track->textures[i].txdepth = tmp;
1061 break; 1062 break;
1062 case R300_ZB_ZPASS_ADDR: 1063 case R300_ZB_ZPASS_ADDR:
@@ -1067,7 +1068,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
1067 r100_cs_dump_packet(p, pkt); 1068 r100_cs_dump_packet(p, pkt);
1068 return r; 1069 return r;
1069 } 1070 }
1070 ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); 1071 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
1071 break; 1072 break;
1072 case 0x4be8: 1073 case 0x4be8:
1073 /* valid register only on RV530 */ 1074 /* valid register only on RV530 */
@@ -1085,60 +1086,20 @@ static int r300_packet0_check(struct radeon_cs_parser *p,
1085static int r300_packet3_check(struct radeon_cs_parser *p, 1086static int r300_packet3_check(struct radeon_cs_parser *p,
1086 struct radeon_cs_packet *pkt) 1087 struct radeon_cs_packet *pkt)
1087{ 1088{
1088 struct radeon_cs_chunk *ib_chunk;
1089
1090 struct radeon_cs_reloc *reloc; 1089 struct radeon_cs_reloc *reloc;
1091 struct r100_cs_track *track; 1090 struct r100_cs_track *track;
1092 volatile uint32_t *ib; 1091 volatile uint32_t *ib;
1093 unsigned idx; 1092 unsigned idx;
1094 unsigned i, c;
1095 int r; 1093 int r;
1096 1094
1097 ib = p->ib->ptr; 1095 ib = p->ib->ptr;
1098 ib_chunk = &p->chunks[p->chunk_ib_idx];
1099 idx = pkt->idx + 1; 1096 idx = pkt->idx + 1;
1100 track = (struct r100_cs_track *)p->track; 1097 track = (struct r100_cs_track *)p->track;
1101 switch(pkt->opcode) { 1098 switch(pkt->opcode) {
1102 case PACKET3_3D_LOAD_VBPNTR: 1099 case PACKET3_3D_LOAD_VBPNTR:
1103 c = ib_chunk->kdata[idx++] & 0x1F; 1100 r = r100_packet3_load_vbpntr(p, pkt, idx);
1104 track->num_arrays = c; 1101 if (r)
1105 for (i = 0; i < (c - 1); i+=2, idx+=3) { 1102 return r;
1106 r = r100_cs_packet_next_reloc(p, &reloc);
1107 if (r) {
1108 DRM_ERROR("No reloc for packet3 %d\n",
1109 pkt->opcode);
1110 r100_cs_dump_packet(p, pkt);
1111 return r;
1112 }
1113 ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
1114 track->arrays[i + 0].robj = reloc->robj;
1115 track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
1116 track->arrays[i + 0].esize &= 0x7F;
1117 r = r100_cs_packet_next_reloc(p, &reloc);
1118 if (r) {
1119 DRM_ERROR("No reloc for packet3 %d\n",
1120 pkt->opcode);
1121 r100_cs_dump_packet(p, pkt);
1122 return r;
1123 }
1124 ib[idx+2] = ib_chunk->kdata[idx+2] + ((u32)reloc->lobj.gpu_offset);
1125 track->arrays[i + 1].robj = reloc->robj;
1126 track->arrays[i + 1].esize = ib_chunk->kdata[idx] >> 24;
1127 track->arrays[i + 1].esize &= 0x7F;
1128 }
1129 if (c & 1) {
1130 r = r100_cs_packet_next_reloc(p, &reloc);
1131 if (r) {
1132 DRM_ERROR("No reloc for packet3 %d\n",
1133 pkt->opcode);
1134 r100_cs_dump_packet(p, pkt);
1135 return r;
1136 }
1137 ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset);
1138 track->arrays[i + 0].robj = reloc->robj;
1139 track->arrays[i + 0].esize = ib_chunk->kdata[idx] >> 8;
1140 track->arrays[i + 0].esize &= 0x7F;
1141 }
1142 break; 1103 break;
1143 case PACKET3_INDX_BUFFER: 1104 case PACKET3_INDX_BUFFER:
1144 r = r100_cs_packet_next_reloc(p, &reloc); 1105 r = r100_cs_packet_next_reloc(p, &reloc);
@@ -1147,7 +1108,7 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
1147 r100_cs_dump_packet(p, pkt); 1108 r100_cs_dump_packet(p, pkt);
1148 return r; 1109 return r;
1149 } 1110 }
1150 ib[idx+1] = ib_chunk->kdata[idx+1] + ((u32)reloc->lobj.gpu_offset); 1111 ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
1151 r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj); 1112 r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj);
1152 if (r) { 1113 if (r) {
1153 return r; 1114 return r;
@@ -1158,11 +1119,11 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
1158 /* Number of dwords is vtx_size * (num_vertices - 1) 1119 /* Number of dwords is vtx_size * (num_vertices - 1)
1159 * PRIM_WALK must be equal to 3 vertex data in embedded 1120 * PRIM_WALK must be equal to 3 vertex data in embedded
1160 * in cmd stream */ 1121 * in cmd stream */
1161 if (((ib_chunk->kdata[idx+1] >> 4) & 0x3) != 3) { 1122 if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) {
1162 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); 1123 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1163 return -EINVAL; 1124 return -EINVAL;
1164 } 1125 }
1165 track->vap_vf_cntl = ib_chunk->kdata[idx+1]; 1126 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1166 track->immd_dwords = pkt->count - 1; 1127 track->immd_dwords = pkt->count - 1;
1167 r = r100_cs_track_check(p->rdev, track); 1128 r = r100_cs_track_check(p->rdev, track);
1168 if (r) { 1129 if (r) {
@@ -1173,11 +1134,11 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
1173 /* Number of dwords is vtx_size * (num_vertices - 1) 1134 /* Number of dwords is vtx_size * (num_vertices - 1)
1174 * PRIM_WALK must be equal to 3 vertex data in embedded 1135 * PRIM_WALK must be equal to 3 vertex data in embedded
1175 * in cmd stream */ 1136 * in cmd stream */
1176 if (((ib_chunk->kdata[idx] >> 4) & 0x3) != 3) { 1137 if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) {
1177 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); 1138 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1178 return -EINVAL; 1139 return -EINVAL;
1179 } 1140 }
1180 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1141 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1181 track->immd_dwords = pkt->count; 1142 track->immd_dwords = pkt->count;
1182 r = r100_cs_track_check(p->rdev, track); 1143 r = r100_cs_track_check(p->rdev, track);
1183 if (r) { 1144 if (r) {
@@ -1185,28 +1146,28 @@ static int r300_packet3_check(struct radeon_cs_parser *p,
1185 } 1146 }
1186 break; 1147 break;
1187 case PACKET3_3D_DRAW_VBUF: 1148 case PACKET3_3D_DRAW_VBUF:
1188 track->vap_vf_cntl = ib_chunk->kdata[idx + 1]; 1149 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1189 r = r100_cs_track_check(p->rdev, track); 1150 r = r100_cs_track_check(p->rdev, track);
1190 if (r) { 1151 if (r) {
1191 return r; 1152 return r;
1192 } 1153 }
1193 break; 1154 break;
1194 case PACKET3_3D_DRAW_VBUF_2: 1155 case PACKET3_3D_DRAW_VBUF_2:
1195 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1156 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1196 r = r100_cs_track_check(p->rdev, track); 1157 r = r100_cs_track_check(p->rdev, track);
1197 if (r) { 1158 if (r) {
1198 return r; 1159 return r;
1199 } 1160 }
1200 break; 1161 break;
1201 case PACKET3_3D_DRAW_INDX: 1162 case PACKET3_3D_DRAW_INDX:
1202 track->vap_vf_cntl = ib_chunk->kdata[idx + 1]; 1163 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1203 r = r100_cs_track_check(p->rdev, track); 1164 r = r100_cs_track_check(p->rdev, track);
1204 if (r) { 1165 if (r) {
1205 return r; 1166 return r;
1206 } 1167 }
1207 break; 1168 break;
1208 case PACKET3_3D_DRAW_INDX_2: 1169 case PACKET3_3D_DRAW_INDX_2:
1209 track->vap_vf_cntl = ib_chunk->kdata[idx]; 1170 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1210 r = r100_cs_track_check(p->rdev, track); 1171 r = r100_cs_track_check(p->rdev, track);
1211 if (r) { 1172 if (r) {
1212 return r; 1173 return r;
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index 33b89cd8743..c629b5aa4a3 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -57,7 +57,7 @@ int r600_cs_packet_parse(struct radeon_cs_parser *p,
57 idx, ib_chunk->length_dw); 57 idx, ib_chunk->length_dw);
58 return -EINVAL; 58 return -EINVAL;
59 } 59 }
60 header = ib_chunk->kdata[idx]; 60 header = radeon_get_ib_value(p, idx);
61 pkt->idx = idx; 61 pkt->idx = idx;
62 pkt->type = CP_PACKET_GET_TYPE(header); 62 pkt->type = CP_PACKET_GET_TYPE(header);
63 pkt->count = CP_PACKET_GET_COUNT(header); 63 pkt->count = CP_PACKET_GET_COUNT(header);
@@ -98,7 +98,6 @@ int r600_cs_packet_parse(struct radeon_cs_parser *p,
98static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p, 98static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
99 struct radeon_cs_reloc **cs_reloc) 99 struct radeon_cs_reloc **cs_reloc)
100{ 100{
101 struct radeon_cs_chunk *ib_chunk;
102 struct radeon_cs_chunk *relocs_chunk; 101 struct radeon_cs_chunk *relocs_chunk;
103 struct radeon_cs_packet p3reloc; 102 struct radeon_cs_packet p3reloc;
104 unsigned idx; 103 unsigned idx;
@@ -109,7 +108,6 @@ static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
109 return -EINVAL; 108 return -EINVAL;
110 } 109 }
111 *cs_reloc = NULL; 110 *cs_reloc = NULL;
112 ib_chunk = &p->chunks[p->chunk_ib_idx];
113 relocs_chunk = &p->chunks[p->chunk_relocs_idx]; 111 relocs_chunk = &p->chunks[p->chunk_relocs_idx];
114 r = r600_cs_packet_parse(p, &p3reloc, p->idx); 112 r = r600_cs_packet_parse(p, &p3reloc, p->idx);
115 if (r) { 113 if (r) {
@@ -121,7 +119,7 @@ static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
121 p3reloc.idx); 119 p3reloc.idx);
122 return -EINVAL; 120 return -EINVAL;
123 } 121 }
124 idx = ib_chunk->kdata[p3reloc.idx + 1]; 122 idx = radeon_get_ib_value(p, p3reloc.idx + 1);
125 if (idx >= relocs_chunk->length_dw) { 123 if (idx >= relocs_chunk->length_dw) {
126 DRM_ERROR("Relocs at %d after relocations chunk end %d !\n", 124 DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
127 idx, relocs_chunk->length_dw); 125 idx, relocs_chunk->length_dw);
@@ -146,7 +144,6 @@ static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
146static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p, 144static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
147 struct radeon_cs_reloc **cs_reloc) 145 struct radeon_cs_reloc **cs_reloc)
148{ 146{
149 struct radeon_cs_chunk *ib_chunk;
150 struct radeon_cs_chunk *relocs_chunk; 147 struct radeon_cs_chunk *relocs_chunk;
151 struct radeon_cs_packet p3reloc; 148 struct radeon_cs_packet p3reloc;
152 unsigned idx; 149 unsigned idx;
@@ -157,7 +154,6 @@ static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
157 return -EINVAL; 154 return -EINVAL;
158 } 155 }
159 *cs_reloc = NULL; 156 *cs_reloc = NULL;
160 ib_chunk = &p->chunks[p->chunk_ib_idx];
161 relocs_chunk = &p->chunks[p->chunk_relocs_idx]; 157 relocs_chunk = &p->chunks[p->chunk_relocs_idx];
162 r = r600_cs_packet_parse(p, &p3reloc, p->idx); 158 r = r600_cs_packet_parse(p, &p3reloc, p->idx);
163 if (r) { 159 if (r) {
@@ -169,7 +165,7 @@ static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
169 p3reloc.idx); 165 p3reloc.idx);
170 return -EINVAL; 166 return -EINVAL;
171 } 167 }
172 idx = ib_chunk->kdata[p3reloc.idx + 1]; 168 idx = radeon_get_ib_value(p, p3reloc.idx + 1);
173 if (idx >= relocs_chunk->length_dw) { 169 if (idx >= relocs_chunk->length_dw) {
174 DRM_ERROR("Relocs at %d after relocations chunk end %d !\n", 170 DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
175 idx, relocs_chunk->length_dw); 171 idx, relocs_chunk->length_dw);
@@ -218,7 +214,6 @@ static int r600_cs_parse_packet0(struct radeon_cs_parser *p,
218static int r600_packet3_check(struct radeon_cs_parser *p, 214static int r600_packet3_check(struct radeon_cs_parser *p,
219 struct radeon_cs_packet *pkt) 215 struct radeon_cs_packet *pkt)
220{ 216{
221 struct radeon_cs_chunk *ib_chunk;
222 struct radeon_cs_reloc *reloc; 217 struct radeon_cs_reloc *reloc;
223 volatile u32 *ib; 218 volatile u32 *ib;
224 unsigned idx; 219 unsigned idx;
@@ -227,8 +222,8 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
227 int r; 222 int r;
228 223
229 ib = p->ib->ptr; 224 ib = p->ib->ptr;
230 ib_chunk = &p->chunks[p->chunk_ib_idx];
231 idx = pkt->idx + 1; 225 idx = pkt->idx + 1;
226
232 switch (pkt->opcode) { 227 switch (pkt->opcode) {
233 case PACKET3_START_3D_CMDBUF: 228 case PACKET3_START_3D_CMDBUF:
234 if (p->family >= CHIP_RV770 || pkt->count) { 229 if (p->family >= CHIP_RV770 || pkt->count) {
@@ -281,7 +276,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
281 return -EINVAL; 276 return -EINVAL;
282 } 277 }
283 /* bit 4 is reg (0) or mem (1) */ 278 /* bit 4 is reg (0) or mem (1) */
284 if (ib_chunk->kdata[idx+0] & 0x10) { 279 if (radeon_get_ib_value(p, idx) & 0x10) {
285 r = r600_cs_packet_next_reloc(p, &reloc); 280 r = r600_cs_packet_next_reloc(p, &reloc);
286 if (r) { 281 if (r) {
287 DRM_ERROR("bad WAIT_REG_MEM\n"); 282 DRM_ERROR("bad WAIT_REG_MEM\n");
@@ -297,8 +292,8 @@ static int r600_packet3_check(struct radeon_cs_parser *p,
297 return -EINVAL; 292 return -EINVAL;
298 } 293 }
299 /* 0xffffffff/0x0 is flush all cache flag */ 294 /* 0xffffffff/0x0 is flush all cache flag */
300 if (ib_chunk->kdata[idx+1] != 0xffffffff || 295 if (radeon_get_ib_value(p, idx + 1) != 0xffffffff ||
301 ib_chunk->kdata[idx+2] != 0) { 296 radeon_get_ib_value(p, idx + 2) != 0) {
302 r = r600_cs_packet_next_reloc(p, &reloc); 297 r = r600_cs_packet_next_reloc(p, &reloc);
303 if (r) { 298 if (r) {
304 DRM_ERROR("bad SURFACE_SYNC\n"); 299 DRM_ERROR("bad SURFACE_SYNC\n");
@@ -639,7 +634,6 @@ int r600_cs_legacy(struct drm_device *dev, void *data, struct drm_file *filp,
639 * uncached). */ 634 * uncached). */
640 ib_chunk = &parser.chunks[parser.chunk_ib_idx]; 635 ib_chunk = &parser.chunks[parser.chunk_ib_idx];
641 parser.ib->length_dw = ib_chunk->length_dw; 636 parser.ib->length_dw = ib_chunk->length_dw;
642 memcpy((void *)parser.ib->ptr, ib_chunk->kdata, ib_chunk->length_dw*4);
643 *l = parser.ib->length_dw; 637 *l = parser.ib->length_dw;
644 r = r600_cs_parse(&parser); 638 r = r600_cs_parse(&parser);
645 if (r) { 639 if (r) {
@@ -647,6 +641,12 @@ int r600_cs_legacy(struct drm_device *dev, void *data, struct drm_file *filp,
647 r600_cs_parser_fini(&parser, r); 641 r600_cs_parser_fini(&parser, r);
648 return r; 642 return r;
649 } 643 }
644 r = radeon_cs_finish_pages(&parser);
645 if (r) {
646 DRM_ERROR("Invalid command stream !\n");
647 r600_cs_parser_fini(&parser, r);
648 return r;
649 }
650 r600_cs_parser_fini(&parser, r); 650 r600_cs_parser_fini(&parser, r);
651 return r; 651 return r;
652} 652}
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index d5de53e06ce..7e34e4376f9 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -342,7 +342,7 @@ struct radeon_ib {
342 unsigned long idx; 342 unsigned long idx;
343 uint64_t gpu_addr; 343 uint64_t gpu_addr;
344 struct radeon_fence *fence; 344 struct radeon_fence *fence;
345 volatile uint32_t *ptr; 345 uint32_t *ptr;
346 uint32_t length_dw; 346 uint32_t length_dw;
347}; 347};
348 348
@@ -415,7 +415,12 @@ struct radeon_cs_reloc {
415struct radeon_cs_chunk { 415struct radeon_cs_chunk {
416 uint32_t chunk_id; 416 uint32_t chunk_id;
417 uint32_t length_dw; 417 uint32_t length_dw;
418 int kpage_idx[2];
419 uint32_t *kpage[2];
418 uint32_t *kdata; 420 uint32_t *kdata;
421 void __user *user_ptr;
422 int last_copied_page;
423 int last_page_index;
419}; 424};
420 425
421struct radeon_cs_parser { 426struct radeon_cs_parser {
@@ -438,8 +443,38 @@ struct radeon_cs_parser {
438 struct radeon_ib *ib; 443 struct radeon_ib *ib;
439 void *track; 444 void *track;
440 unsigned family; 445 unsigned family;
446 int parser_error;
441}; 447};
442 448
449extern int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx);
450extern int radeon_cs_finish_pages(struct radeon_cs_parser *p);
451
452
453static inline u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx)
454{
455 struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
456 u32 pg_idx, pg_offset;
457 u32 idx_value = 0;
458 int new_page;
459
460 pg_idx = (idx * 4) / PAGE_SIZE;
461 pg_offset = (idx * 4) % PAGE_SIZE;
462
463 if (ibc->kpage_idx[0] == pg_idx)
464 return ibc->kpage[0][pg_offset/4];
465 if (ibc->kpage_idx[1] == pg_idx)
466 return ibc->kpage[1][pg_offset/4];
467
468 new_page = radeon_cs_update_pages(p, pg_idx);
469 if (new_page < 0) {
470 p->parser_error = new_page;
471 return 0;
472 }
473
474 idx_value = ibc->kpage[new_page][pg_offset/4];
475 return idx_value;
476}
477
443struct radeon_cs_packet { 478struct radeon_cs_packet {
444 unsigned idx; 479 unsigned idx;
445 unsigned type; 480 unsigned type;
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index 12f5990c2d2..dea8acf8886 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -142,15 +142,31 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
142 } 142 }
143 143
144 p->chunks[i].length_dw = user_chunk.length_dw; 144 p->chunks[i].length_dw = user_chunk.length_dw;
145 cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data; 145 p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data;
146 146
147 size = p->chunks[i].length_dw * sizeof(uint32_t); 147 cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data;
148 p->chunks[i].kdata = kmalloc(size, GFP_KERNEL); 148 if (p->chunks[i].chunk_id != RADEON_CHUNK_ID_IB) {
149 if (p->chunks[i].kdata == NULL) { 149 size = p->chunks[i].length_dw * sizeof(uint32_t);
150 return -ENOMEM; 150 p->chunks[i].kdata = kmalloc(size, GFP_KERNEL);
151 } 151 if (p->chunks[i].kdata == NULL) {
152 if (DRM_COPY_FROM_USER(p->chunks[i].kdata, cdata, size)) { 152 return -ENOMEM;
153 return -EFAULT; 153 }
154 if (DRM_COPY_FROM_USER(p->chunks[i].kdata,
155 p->chunks[i].user_ptr, size)) {
156 return -EFAULT;
157 }
158 } else {
159 p->chunks[i].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
160 p->chunks[i].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
161 if (p->chunks[i].kpage[0] == NULL || p->chunks[i].kpage[1] == NULL) {
162 kfree(p->chunks[i].kpage[0]);
163 kfree(p->chunks[i].kpage[1]);
164 return -ENOMEM;
165 }
166 p->chunks[i].kpage_idx[0] = -1;
167 p->chunks[i].kpage_idx[1] = -1;
168 p->chunks[i].last_copied_page = -1;
169 p->chunks[i].last_page_index = ((p->chunks[i].length_dw * 4) - 1) / PAGE_SIZE;
154 } 170 }
155 } 171 }
156 if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) { 172 if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) {
@@ -190,6 +206,8 @@ static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error)
190 kfree(parser->relocs_ptr); 206 kfree(parser->relocs_ptr);
191 for (i = 0; i < parser->nchunks; i++) { 207 for (i = 0; i < parser->nchunks; i++) {
192 kfree(parser->chunks[i].kdata); 208 kfree(parser->chunks[i].kdata);
209 kfree(parser->chunks[i].kpage[0]);
210 kfree(parser->chunks[i].kpage[1]);
193 } 211 }
194 kfree(parser->chunks); 212 kfree(parser->chunks);
195 kfree(parser->chunks_array); 213 kfree(parser->chunks_array);
@@ -238,8 +256,14 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
238 * uncached). */ 256 * uncached). */
239 ib_chunk = &parser.chunks[parser.chunk_ib_idx]; 257 ib_chunk = &parser.chunks[parser.chunk_ib_idx];
240 parser.ib->length_dw = ib_chunk->length_dw; 258 parser.ib->length_dw = ib_chunk->length_dw;
241 memcpy((void *)parser.ib->ptr, ib_chunk->kdata, ib_chunk->length_dw*4);
242 r = radeon_cs_parse(&parser); 259 r = radeon_cs_parse(&parser);
260 if (r || parser.parser_error) {
261 DRM_ERROR("Invalid command stream !\n");
262 radeon_cs_parser_fini(&parser, r);
263 mutex_unlock(&rdev->cs_mutex);
264 return r;
265 }
266 r = radeon_cs_finish_pages(&parser);
243 if (r) { 267 if (r) {
244 DRM_ERROR("Invalid command stream !\n"); 268 DRM_ERROR("Invalid command stream !\n");
245 radeon_cs_parser_fini(&parser, r); 269 radeon_cs_parser_fini(&parser, r);
@@ -254,3 +278,66 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
254 mutex_unlock(&rdev->cs_mutex); 278 mutex_unlock(&rdev->cs_mutex);
255 return r; 279 return r;
256} 280}
281
282int radeon_cs_finish_pages(struct radeon_cs_parser *p)
283{
284 struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
285 int i;
286 int size = PAGE_SIZE;
287
288 for (i = ibc->last_copied_page + 1; i <= ibc->last_page_index; i++) {
289 if (i == ibc->last_page_index) {
290 size = (ibc->length_dw * 4) % PAGE_SIZE;
291 if (size == 0)
292 size = PAGE_SIZE;
293 }
294
295 if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
296 ibc->user_ptr + (i * PAGE_SIZE),
297 size))
298 return -EFAULT;
299 }
300 return 0;
301}
302
303int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx)
304{
305 int new_page;
306 int num_extra_pages;
307 struct radeon_cs_chunk *ibc = &p->chunks[p->chunk_ib_idx];
308 int i;
309 int size = PAGE_SIZE;
310
311 num_extra_pages = (pg_idx - ibc->last_copied_page - 1);
312 for (i = ibc->last_copied_page + 1; i < ibc->last_copied_page + num_extra_pages; i++) {
313 if (DRM_COPY_FROM_USER(p->ib->ptr + (i * (PAGE_SIZE/4)),
314 ibc->user_ptr + (i * PAGE_SIZE),
315 PAGE_SIZE)) {
316 p->parser_error = -EFAULT;
317 return 0;
318 }
319 }
320
321 new_page = ibc->kpage_idx[0] < ibc->kpage_idx[1] ? 0 : 1;
322
323 if (pg_idx == ibc->last_page_index) {
324 size = (ibc->length_dw * 4) % PAGE_SIZE;
325 if (size == 0)
326 size = PAGE_SIZE;
327 }
328
329 if (DRM_COPY_FROM_USER(ibc->kpage[new_page],
330 ibc->user_ptr + (pg_idx * PAGE_SIZE),
331 size)) {
332 p->parser_error = -EFAULT;
333 return 0;
334 }
335
336 /* copy to IB here */
337 memcpy((void *)(p->ib->ptr+(pg_idx*(PAGE_SIZE/4))), ibc->kpage[new_page], size);
338
339 ibc->last_copied_page = pg_idx;
340 ibc->kpage_idx[new_page] = pg_idx;
341
342 return new_page;
343}
generated by cgit v1.2.2 (git 2.25.0) at 2024-10-19 15:59:23 -0400