drm/amdgpu: remove dead code after hmm_range_fault()
authorJason Gunthorpe <jgg@mellanox.com>
Fri, 1 May 2020 18:20:46 +0000 (15:20 -0300)
committerJason Gunthorpe <jgg@mellanox.com>
Mon, 11 May 2020 13:47:29 +0000 (10:47 -0300)
Since amdgpu does not use the snapshot mode of hmm_range_fault() a
successful return already proves that all entries in the pfns are
HMM_PFN_VALID, there is no need to check the return result of
hmm_device_entry_to_page().

Link: https://lore.kernel.org/r/3-v2-b4e84f444c7d+24f57-hmm_no_flags_jgg@mellanox.com
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

index 7eb745b8acce0f0da3842033a7d8c760092157a7..41ae7f96f4819482569baa3d0271f8686984599b 100644 (file)
@@ -862,17 +862,13 @@ retry:
                goto out_free_pfns;
        }
 
-       for (i = 0; i < ttm->num_pages; i++) {
-               /* FIXME: The pages cannot be touched outside the notifier_lock */
+       /*
+        * Due to default_flags, all pages are HMM_PFN_VALID or
+        * hmm_range_fault() fails. FIXME: The pages cannot be touched outside
+        * the notifier_lock, and mmu_interval_read_retry() must be done first.
+        */
+       for (i = 0; i < ttm->num_pages; i++)
                pages[i] = hmm_device_entry_to_page(range, range->pfns[i]);
-               if (unlikely(!pages[i])) {
-                       pr_err("Page fault failed for pfn[%lu] = 0x%llx\n",
-                              i, range->pfns[i]);
-                       r = -ENOMEM;
-
-                       goto out_free_pfns;
-               }
-       }
 
        gtt->range = range;
        mmput(mm);