Hi Akshata, FYI, the error/warning still remains. tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master head: bb1b0acdcd66e0d8eedee3570d249e076b89ab32 commit: 8d79acd567db183e675cccc6cc737d2959e2a2d9 [400/5508] drm/xe/migrate: Add helper function to program identity map config: i386-randconfig-001-20240820 (https://download.01.org/0day-ci/archive/20240821/202408210141.HdA7oipm-lkp@xxxxxxxxx/config) compiler: clang version 18.1.5 (https://github.com/llvm/llvm-project 617a15a9eac96088ae5e9134248d8236e34b91b1) reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240821/202408210141.HdA7oipm-lkp@xxxxxxxxx/reproduce) If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp@xxxxxxxxx> | Closes: https://lore.kernel.org/oe-kbuild-all/202408210141.HdA7oipm-lkp@xxxxxxxxx/ All errors (new ones prefixed by >>): >> drivers/gpu/drm/xe/xe_migrate.c:304:52: error: result of comparison of constant 274877906944 with expression of type 'resource_size_t' (aka 'unsigned int') is always true [-Werror,-Wtautological-constant-out-of-range-compare] 304 | xe_assert(xe, (xe->mem.vram.actual_physical_size <= SZ_256G)); | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~ drivers/gpu/drm/xe/xe_assert.h:108:54: note: expanded from macro 'xe_assert' 108 | #define xe_assert(xe, condition) xe_assert_msg((xe), condition, "") | ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~ drivers/gpu/drm/xe/xe_assert.h:111:24: note: expanded from macro 'xe_assert_msg' 111 | __xe_assert_msg(__xe, condition, \ | ~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 112 | "platform: %s subplatform: %d\n" \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 113 | "graphics: %s %u.%02u step %s\n" \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 114 | "media: %s %u.%02u step %s\n" \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 115 | msg, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 116 | __xe->info.platform_name, __xe->info.subplatform, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 117 | __xe->info.graphics_name, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 118 | __xe->info.graphics_verx100 / 100, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 119 | __xe->info.graphics_verx100 % 100, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 120 | xe_step_name(__xe->info.step.graphics), \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 121 | __xe->info.media_name, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 122 | __xe->info.media_verx100 / 100, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 123 | __xe->info.media_verx100 % 100, \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 124 | xe_step_name(__xe->info.step.media), \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 125 | ## arg); \ | ~~~~~~~ drivers/gpu/drm/xe/xe_assert.h:84:31: note: expanded from macro '__xe_assert_msg' 84 | (void)drm_WARN(&(xe)->drm, !(condition), "[" DRM_NAME "] Assertion `%s` failed!\n" msg, \ | ~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 85 | __stringify(condition), ## arg); \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ include/drm/drm_print.h:635:7: note: expanded from macro 'drm_WARN' 635 | WARN(condition, "%s %s: [drm] " format, \ | ~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 636 | dev_driver_string((drm)->dev), \ | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 637 | dev_name((drm)->dev), ## arg) | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ include/asm-generic/bug.h:132:25: note: expanded from macro 'WARN' 132 | int __ret_warn_on = !!(condition); \ | ^~~~~~~~~ In file included from drivers/gpu/drm/xe/xe_migrate.c:1494: drivers/gpu/drm/xe/tests/xe_migrate.c:404:12: error: too many arguments to function call, expected 10, have 11 402 | batch_size += pte_update_size(m, src_is_vram, src_is_vram, src, &src_it, &src_L0, | ~~~~~~~~~~~~~~~ 403 | &src_L0_ofs, &src_L0_pt, 0, 0, 404 | avail_pts); | ^~~~~~~~~ drivers/gpu/drm/xe/xe_migrate.c:490:12: note: 'pte_update_size' declared here 490 | static u32 pte_update_size(struct xe_migrate *m, | ^ ~~~~~~~~~~~~~~~~~~~~~ 491 | bool is_vram, | ~~~~~~~~~~~~~ 492 | struct ttm_resource *res, | ~~~~~~~~~~~~~~~~~~~~~~~~~ 493 | struct xe_res_cursor *cur, | ~~~~~~~~~~~~~~~~~~~~~~~~~~ 494 | u64 *L0, u64 *L0_ofs, u32 *L0_pt, | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 495 | u32 cmd_size, u32 pt_ofs, u32 avail_pts) | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from drivers/gpu/drm/xe/xe_migrate.c:1494: drivers/gpu/drm/xe/tests/xe_migrate.c:408:23: error: too many arguments to function call, expected 10, have 11 406 | batch_size += pte_update_size(m, dst_is_vram, dst_is_vram, dst, &dst_it, &src_L0, | ~~~~~~~~~~~~~~~ 407 | &dst_L0_ofs, &dst_L0_pt, 0, 408 | avail_pts, avail_pts); | ^~~~~~~~~ drivers/gpu/drm/xe/xe_migrate.c:490:12: note: 'pte_update_size' declared here 490 | static u32 pte_update_size(struct xe_migrate *m, | ^ ~~~~~~~~~~~~~~~~~~~~~ 491 | bool is_vram, | ~~~~~~~~~~~~~ 492 | struct ttm_resource *res, | ~~~~~~~~~~~~~~~~~~~~~~~~~ 493 | struct xe_res_cursor *cur, | ~~~~~~~~~~~~~~~~~~~~~~~~~~ 494 | u64 *L0, u64 *L0_ofs, u32 *L0_pt, | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 495 | u32 cmd_size, u32 pt_ofs, u32 avail_pts) | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3 errors generated. vim +304 drivers/gpu/drm/xe/xe_migrate.c 177 178 static int xe_migrate_prepare_vm(struct xe_tile *tile, struct xe_migrate *m, 179 struct xe_vm *vm) 180 { 181 struct xe_device *xe = tile_to_xe(tile); 182 u16 pat_index = xe->pat.idx[XE_CACHE_WB]; 183 u8 id = tile->id; 184 u32 num_entries = NUM_PT_SLOTS, num_level = vm->pt_root[id]->level, 185 num_setup = num_level + 1; 186 u32 map_ofs, level, i; 187 struct xe_bo *bo, *batch = tile->mem.kernel_bb_pool->bo; 188 u64 entry, pt30_ofs; 189 190 /* Can't bump NUM_PT_SLOTS too high */ 191 BUILD_BUG_ON(NUM_PT_SLOTS > SZ_2M/XE_PAGE_SIZE); 192 /* Must be a multiple of 64K to support all platforms */ 193 BUILD_BUG_ON(NUM_PT_SLOTS * XE_PAGE_SIZE % SZ_64K); 194 /* And one slot reserved for the 4KiB page table updates */ 195 BUILD_BUG_ON(!(NUM_KERNEL_PDE & 1)); 196 197 /* Need to be sure everything fits in the first PT, or create more */ 198 xe_tile_assert(tile, m->batch_base_ofs + batch->size < SZ_2M); 199 200 bo = xe_bo_create_pin_map(vm->xe, tile, vm, 201 num_entries * XE_PAGE_SIZE, 202 ttm_bo_type_kernel, 203 XE_BO_FLAG_VRAM_IF_DGFX(tile) | 204 XE_BO_FLAG_PINNED); 205 if (IS_ERR(bo)) 206 return PTR_ERR(bo); 207 208 /* PT31 reserved for 2M identity map */ 209 pt30_ofs = bo->size - 2 * XE_PAGE_SIZE; 210 entry = vm->pt_ops->pde_encode_bo(bo, pt30_ofs, pat_index); 211 xe_pt_write(xe, &vm->pt_root[id]->bo->vmap, 0, entry); 212 213 map_ofs = (num_entries - num_setup) * XE_PAGE_SIZE; 214 215 /* Map the entire BO in our level 0 pt */ 216 for (i = 0, level = 0; i < num_entries; level++) { 217 entry = vm->pt_ops->pte_encode_bo(bo, i * XE_PAGE_SIZE, 218 pat_index, 0); 219 220 xe_map_wr(xe, &bo->vmap, map_ofs + level * 8, u64, entry); 221 222 if (vm->flags & XE_VM_FLAG_64K) 223 i += 16; 224 else 225 i += 1; 226 } 227 228 if (!IS_DGFX(xe)) { 229 /* Write out batch too */ 230 m->batch_base_ofs = NUM_PT_SLOTS * XE_PAGE_SIZE; 231 for (i = 0; i < batch->size; 232 i += vm->flags & XE_VM_FLAG_64K ? XE_64K_PAGE_SIZE : 233 XE_PAGE_SIZE) { 234 entry = vm->pt_ops->pte_encode_bo(batch, i, 235 pat_index, 0); 236 237 xe_map_wr(xe, &bo->vmap, map_ofs + level * 8, u64, 238 entry); 239 level++; 240 } 241 if (xe->info.has_usm) { 242 xe_tile_assert(tile, batch->size == SZ_1M); 243 244 batch = tile->primary_gt->usm.bb_pool->bo; 245 m->usm_batch_base_ofs = m->batch_base_ofs + SZ_1M; 246 xe_tile_assert(tile, batch->size == SZ_512K); 247 248 for (i = 0; i < batch->size; 249 i += vm->flags & XE_VM_FLAG_64K ? XE_64K_PAGE_SIZE : 250 XE_PAGE_SIZE) { 251 entry = vm->pt_ops->pte_encode_bo(batch, i, 252 pat_index, 0); 253 254 xe_map_wr(xe, &bo->vmap, map_ofs + level * 8, u64, 255 entry); 256 level++; 257 } 258 } 259 } else { 260 u64 batch_addr = xe_bo_addr(batch, 0, XE_PAGE_SIZE); 261 262 m->batch_base_ofs = xe_migrate_vram_ofs(xe, batch_addr); 263 264 if (xe->info.has_usm) { 265 batch = tile->primary_gt->usm.bb_pool->bo; 266 batch_addr = xe_bo_addr(batch, 0, XE_PAGE_SIZE); 267 m->usm_batch_base_ofs = xe_migrate_vram_ofs(xe, batch_addr); 268 } 269 } 270 271 for (level = 1; level < num_level; level++) { 272 u32 flags = 0; 273 274 if (vm->flags & XE_VM_FLAG_64K && level == 1) 275 flags = XE_PDE_64K; 276 277 entry = vm->pt_ops->pde_encode_bo(bo, map_ofs + (u64)(level - 1) * 278 XE_PAGE_SIZE, pat_index); 279 xe_map_wr(xe, &bo->vmap, map_ofs + XE_PAGE_SIZE * level, u64, 280 entry | flags); 281 } 282 283 /* Write PDE's that point to our BO. */ 284 for (i = 0; i < map_ofs / PAGE_SIZE; i++) { 285 entry = vm->pt_ops->pde_encode_bo(bo, (u64)i * XE_PAGE_SIZE, 286 pat_index); 287 288 xe_map_wr(xe, &bo->vmap, map_ofs + XE_PAGE_SIZE + 289 (i + 1) * 8, u64, entry); 290 } 291 292 /* Set up a 1GiB NULL mapping at 255GiB offset. */ 293 level = 2; 294 xe_map_wr(xe, &bo->vmap, map_ofs + XE_PAGE_SIZE * level + 255 * 8, u64, 295 vm->pt_ops->pte_encode_addr(xe, 0, pat_index, level, IS_DGFX(xe), 0) 296 | XE_PTE_NULL); 297 m->cleared_mem_ofs = (255ULL << xe_pt_shift(level)); 298 299 /* Identity map the entire vram at 256GiB offset */ 300 if (IS_DGFX(xe)) { 301 u64 pt31_ofs = bo->size - XE_PAGE_SIZE; 302 303 xe_migrate_program_identity(xe, vm, bo, map_ofs, 256, pat_index, pt31_ofs); > 304 xe_assert(xe, (xe->mem.vram.actual_physical_size <= SZ_256G)); 305 } 306 307 /* 308 * Example layout created above, with root level = 3: 309 * [PT0...PT7]: kernel PT's for copy/clear; 64 or 4KiB PTE's 310 * [PT8]: Kernel PT for VM_BIND, 4 KiB PTE's 311 * [PT9...PT27]: Userspace PT's for VM_BIND, 4 KiB PTE's 312 * [PT28 = PDE 0] [PT29 = PDE 1] [PT30 = PDE 2] [PT31 = 2M vram identity map] 313 * 314 * This makes the lowest part of the VM point to the pagetables. 315 * Hence the lowest 2M in the vm should point to itself, with a few writes 316 * and flushes, other parts of the VM can be used either for copying and 317 * clearing. 318 * 319 * For performance, the kernel reserves PDE's, so about 20 are left 320 * for async VM updates. 321 * 322 * To make it easier to work, each scratch PT is put in slot (1 + PT #) 323 * everywhere, this allows lockless updates to scratch pages by using 324 * the different addresses in VM. 325 */ 326 #define NUM_VMUSA_UNIT_PER_PAGE 32 327 #define VM_SA_UPDATE_UNIT_SIZE (XE_PAGE_SIZE / NUM_VMUSA_UNIT_PER_PAGE) 328 #define NUM_VMUSA_WRITES_PER_UNIT (VM_SA_UPDATE_UNIT_SIZE / sizeof(u64)) 329 drm_suballoc_manager_init(&m->vm_update_sa, 330 (size_t)(map_ofs / XE_PAGE_SIZE - NUM_KERNEL_PDE) * 331 NUM_VMUSA_UNIT_PER_PAGE, 0); 332 333 m->pt_bo = bo; 334 return 0; 335 } 336 -- 0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
