On 2/11/20 1:17 PM, Christian König wrote:
Am 11.02.20 um 13:06 schrieb Nirmoy:
Hi Christian

On 2/11/20 12:50 PM, Christian König wrote:
Am 11.02.20 um 12:17 schrieb Nirmoy Das:
GPU address should belong to driver not in memory management.
This patch moves ttm bo.offset and gpu_offset calculation to amdgpu driver.

Actually for amdgpu we should be able to get completely rid of bo->offset. All we need to do is to change amdgpu_bo_gpu_offset() and do the correct offset calculation there.

If I am not wrong with this patch amdgpu completly got rid of ttm's bo->offset and ttm_mem_type_manager->gpu_offset.  Do you mean we don't even need amdgpu_bo->offset(added by this patch) as well ?

Yes exactly. IIRC we shouldn't have those many cases left using that and we can just use a helper for those which calculated the offset from the mem->start address.

Thanks,  I will get back with a V2


Regards,

Nirmoy


Christian.




Regards,

Nirmoy


Apart from that looks good to me.

Regards,
Christian.


Signed-off-by: Nirmoy Das <nirmoy....@amd.com>
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    |  2 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c |  4 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c        |  4 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_display.c   |  2 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c        |  2 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_object.c    | 22 ++++++++--
  drivers/gpu/drm/amd/amdgpu/amdgpu_object.h    |  3 ++
  drivers/gpu/drm/amd/amdgpu/amdgpu_test.c      |  2 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c       | 40 +++++++++----------
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h       |  2 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c   |  8 ++--
  .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c |  2 +-
  12 files changed, 55 insertions(+), 38 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 8609287620ea..613047ea9065 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -263,7 +263,7 @@ int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
          goto allocate_mem_pin_bo_failed;
      }
  -    r = amdgpu_ttm_alloc_gart(&bo->tbo);
+    r = amdgpu_ttm_alloc_gart(bo);
      if (r) {
          dev_err(adev->dev, "%p bind failed\n", bo);
          goto allocate_mem_kmap_bo_failed;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
index d9b35df33806..f29986231b27 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
@@ -98,7 +98,7 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size,
          amdgpu_bo_unreserve(sobj);
          goto out_cleanup;
      }
-    r = amdgpu_ttm_alloc_gart(&sobj->tbo);
+    r = amdgpu_ttm_alloc_gart(sobj);
      amdgpu_bo_unreserve(sobj);
      if (r) {
          goto out_cleanup;
@@ -117,7 +117,7 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size,
          amdgpu_bo_unreserve(sobj);
          goto out_cleanup;
      }
-    r = amdgpu_ttm_alloc_gart(&dobj->tbo);
+    r = amdgpu_ttm_alloc_gart(dobj);
      amdgpu_bo_unreserve(dobj);
      if (r) {
          goto out_cleanup;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 80ba6dfc54e2..921719d92ab1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -642,7 +642,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
      if (!r && p->uf_entry.tv.bo) {
          struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo);
  -        r = amdgpu_ttm_alloc_gart(&uf->tbo);
+        r = amdgpu_ttm_alloc_gart(uf);
          p->job->uf_addr += amdgpu_bo_gpu_offset(uf);
      }
  @@ -1676,5 +1676,5 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
              return r;
      }
  -    return amdgpu_ttm_alloc_gart(&(*bo)->tbo);
+    return amdgpu_ttm_alloc_gart(*bo);
  }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
index ca9056b0472a..3d0aa2b2926f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
@@ -199,7 +199,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc,
          }
      }
  -    r = amdgpu_ttm_alloc_gart(&new_abo->tbo);
+    r = amdgpu_ttm_alloc_gart(new_abo);
      if (unlikely(r != 0)) {
          DRM_ERROR("%p bind failed\n", new_abo);
          goto unpin;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c
index 143753d237e7..a02b1cbd006d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c
@@ -174,7 +174,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev,
          goto out_unref;
      }
  -    ret = amdgpu_ttm_alloc_gart(&abo->tbo);
+    ret = amdgpu_ttm_alloc_gart(abo);
      if (ret) {
          amdgpu_bo_unreserve(abo);
          dev_err(adev->dev, "%p bind failed\n", abo);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 6f60a581e3ba..650758a5b3aa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -274,7 +274,7 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev,
          goto error_unreserve;
      }
  -    r = amdgpu_ttm_alloc_gart(&(*bo_ptr)->tbo);
+    r = amdgpu_ttm_alloc_gart(*bo_ptr);
      if (r) {
          dev_err(adev->dev, "%p bind failed\n", *bo_ptr);
          goto error_unpin;
@@ -917,7 +917,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
          bo->pin_count++;
            if (max_offset != 0) {
-            u64 domain_start = bo->tbo.bdev->man[mem_type].gpu_offset;
+            u64 domain_start = amdgpu_bo_gpu_start(adev, mem_type);
              WARN_ON_ONCE(max_offset <
                       (amdgpu_bo_gpu_offset(bo) - domain_start));
          }
@@ -1248,6 +1248,18 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer,
      return 0;
  }
  +uint64_t amdgpu_bo_gpu_start(struct amdgpu_device *adev, uint32_t type)
+{
+    switch(type) {
+        case TTM_PL_TT:
+            return adev->gmc.gart_start;
+        case TTM_PL_VRAM:
+            return adev->gmc.vram_start;
+    }
+
+    return 0;
+}
+
  /**
   * amdgpu_bo_move_notify - notification about a memory move
   * @bo: pointer to a buffer object
@@ -1278,6 +1290,10 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
      if (evict)
          atomic64_inc(&adev->num_evictions);
  +    if (new_mem)
+        abo->offset = (new_mem->start << PAGE_SHIFT) +
+ amdgpu_bo_gpu_start(amdgpu_ttm_adev(bo->bdev), new_mem->mem_type);
+
      /* update statistics */
      if (!new_mem)
          return;
@@ -1467,7 +1483,7 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
      WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_VRAM &&
               !(bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS));
  -    return amdgpu_gmc_sign_extend(bo->tbo.offset);
+    return amdgpu_gmc_sign_extend(bo->offset);
  }
    /**
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index 96d805889e8d..c7e1028fd31f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
@@ -94,6 +94,8 @@ struct amdgpu_bo {
      unsigned            prime_shared_count;
      /* per VM structure for page tables and with virtual addresses */
      struct amdgpu_vm_bo_base    *vm_bo;
+     /* GPU address space is independent of CPU word size */
+    uint64_t            offset;
      /* Constant after initialization */
      struct amdgpu_bo        *parent;
      struct amdgpu_bo        *shadow;
@@ -294,6 +296,7 @@ int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow,   uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev,
                          uint32_t domain);
  +uint64_t amdgpu_bo_gpu_start(struct amdgpu_device *adev, uint32_t type);
  /*
   * sub allocation
   */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index 476f1f89aaad..dfab64bbb5fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -103,7 +103,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
              DRM_ERROR("Failed to pin GTT object %d\n", i);
              goto out_lclean_unres;
          }
-        r = amdgpu_ttm_alloc_gart(&gtt_obj[i]->tbo);
+        r = amdgpu_ttm_alloc_gart(gtt_obj[i]);
          if (r) {
              DRM_ERROR("%p bind failed\n", gtt_obj[i]);
              goto out_lclean_unpin;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 2c1d1eb1a7e1..15d9e8f807b9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -103,7 +103,6 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
      case TTM_PL_TT:
          /* GTT memory  */
          man->func = &amdgpu_gtt_mgr_func;
-        man->gpu_offset = adev->gmc.gart_start;
          man->available_caching = TTM_PL_MASK_CACHING;
          man->default_caching = TTM_PL_FLAG_CACHED;
          man->flags = TTM_MEMTYPE_FLAG_MAPPABLE | TTM_MEMTYPE_FLAG_CMA; @@ -111,7 +110,6 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
      case TTM_PL_VRAM:
          /* "On-card" video ram */
          man->func = &amdgpu_vram_mgr_func;
-        man->gpu_offset = adev->gmc.vram_start;
          man->flags = TTM_MEMTYPE_FLAG_FIXED |
                   TTM_MEMTYPE_FLAG_MAPPABLE;
          man->available_caching = TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_WC; @@ -122,7 +120,6 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
      case AMDGPU_PL_OA:
          /* On-chip GDS memory*/
          man->func = &ttm_bo_manager_func;
-        man->gpu_offset = 0;
          man->flags = TTM_MEMTYPE_FLAG_FIXED | TTM_MEMTYPE_FLAG_CMA;
          man->available_caching = TTM_PL_FLAG_UNCACHED;
          man->default_caching = TTM_PL_FLAG_UNCACHED;
@@ -270,7 +267,7 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo,
        if (mm_node->start != AMDGPU_BO_INVALID_OFFSET) {
          addr = mm_node->start << PAGE_SHIFT;
-        addr += bo->bdev->man[mem->mem_type].gpu_offset;
+        addr += amdgpu_bo_gpu_start(amdgpu_ttm_adev(bo->bdev), mem->mem_type);
      }
      return addr;
  }
@@ -1086,27 +1083,28 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm,
  /**
   * amdgpu_ttm_alloc_gart - Allocate GART memory for buffer object
   */
-int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
+int amdgpu_ttm_alloc_gart(struct amdgpu_bo *bo)
  {
-    struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
+    struct ttm_buffer_object *tbo = &bo->tbo;
+    struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev);
      struct ttm_operation_ctx ctx = { false, false };
-    struct amdgpu_ttm_tt *gtt = (void*)bo->ttm;
+    struct amdgpu_ttm_tt *gtt = (void*)tbo->ttm;
      struct ttm_mem_reg tmp;
      struct ttm_placement placement;
      struct ttm_place placements;
      uint64_t addr, flags;
      int r;
  -    if (bo->mem.start != AMDGPU_BO_INVALID_OFFSET)
+    if (tbo->mem.start != AMDGPU_BO_INVALID_OFFSET)
          return 0;
  -    addr = amdgpu_gmc_agp_addr(bo);
+    addr = amdgpu_gmc_agp_addr(tbo);
      if (addr != AMDGPU_BO_INVALID_OFFSET) {
-        bo->mem.start = addr >> PAGE_SHIFT;
+        tbo->mem.start = addr >> PAGE_SHIFT;
      } else {
            /* allocate GART space */
-        tmp = bo->mem;
+        tmp = tbo->mem;
          tmp.mm_node = NULL;
          placement.num_placement = 1;
          placement.placement = &placements;
@@ -1114,30 +1112,30 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
          placement.busy_placement = &placements;
          placements.fpfn = 0;
          placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
-        placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) |
+        placements.flags = (tbo->mem.placement & ~TTM_PL_MASK_MEM) |
              TTM_PL_FLAG_TT;
  -        r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx);
+        r = ttm_bo_mem_space(tbo, &placement, &tmp, &ctx);
          if (unlikely(r))
              return r;
            /* compute PTE flags for this buffer object */
-        flags = amdgpu_ttm_tt_pte_flags(adev, bo->ttm, &tmp);
+        flags = amdgpu_ttm_tt_pte_flags(adev, tbo->ttm, &tmp);
            /* Bind pages */
          gtt->offset = (u64)tmp.start << PAGE_SHIFT;
-        r = amdgpu_ttm_gart_bind(adev, bo, flags);
+        r = amdgpu_ttm_gart_bind(adev, tbo, flags);
          if (unlikely(r)) {
-            ttm_bo_mem_put(bo, &tmp);
+            ttm_bo_mem_put(tbo, &tmp);
              return r;
          }
  -        ttm_bo_mem_put(bo, &bo->mem);
-        bo->mem = tmp;
+        ttm_bo_mem_put(tbo, &tbo->mem);
+        tbo->mem = tmp;
      }
  -    bo->offset = (bo->mem.start << PAGE_SHIFT) +
- bo->bdev->man[bo->mem.mem_type].gpu_offset;
+    bo->offset = (tbo->mem.start << PAGE_SHIFT) +
+        amdgpu_bo_gpu_start(adev, tbo->mem.mem_type);
        return 0;
  }
@@ -2149,7 +2147,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
      }
        if (bo->tbo.mem.mem_type == TTM_PL_TT) {
-        r = amdgpu_ttm_alloc_gart(&bo->tbo);
+        r = amdgpu_ttm_alloc_gart(bo);
          if (r)
              return r;
      }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
index f1ebd424510c..4f6e7ef2b004 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
@@ -100,7 +100,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
              struct dma_fence **fence);
    int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma);
-int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
+int amdgpu_ttm_alloc_gart(struct amdgpu_bo *bo);
  int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
    #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
index 4cc7881f438c..c810cda7a06e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
@@ -37,12 +37,12 @@ static int amdgpu_vm_sdma_map_table(struct amdgpu_bo *table)
  {
      int r;
  -    r = amdgpu_ttm_alloc_gart(&table->tbo);
+    r = amdgpu_ttm_alloc_gart(table);
      if (r)
          return r;
        if (table->shadow)
-        r = amdgpu_ttm_alloc_gart(&table->shadow->tbo);
+        r = amdgpu_ttm_alloc_gart(table->shadow);
        return r;
  }
@@ -140,7 +140,7 @@ static void amdgpu_vm_sdma_copy_ptes(struct amdgpu_vm_update_params *p,
        src += p->num_dw_left * 4;
  -    pe += amdgpu_gmc_sign_extend(bo->tbo.offset);
+    pe += amdgpu_gmc_sign_extend(bo->offset);
      trace_amdgpu_vm_copy_ptes(pe, src, count, p->direct);
        amdgpu_vm_copy_pte(p->adev, ib, pe, src, count);
@@ -167,7 +167,7 @@ static void amdgpu_vm_sdma_set_ptes(struct amdgpu_vm_update_params *p,
  {
      struct amdgpu_ib *ib = p->job->ibs;
  -    pe += amdgpu_gmc_sign_extend(bo->tbo.offset);
+    pe += amdgpu_gmc_sign_extend(bo->offset);
      trace_amdgpu_vm_set_ptes(pe, addr, count, incr, flags, p->direct);
      if (count < 3) {
          amdgpu_vm_write_pte(p->adev, ib, pe, addr | flags,
diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index 6786d34f7d04..277dbc6ba8ec 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -5212,7 +5212,7 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane,
          return r;
      }
  -    r = amdgpu_ttm_alloc_gart(&rbo->tbo);
+    r = amdgpu_ttm_alloc_gart(rbo);
      if (unlikely(r != 0)) {
          amdgpu_bo_unpin(rbo);
          ttm_eu_backoff_reservation(&ticket, &list);


_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Reply via email to