Commit 88467db6 authored by Philip Yang's avatar Philip Yang Committed by Alex Deucher

drm/amdkfd: Fix partial migration bugs

Migration range from system memory to VRAM, if system page can not be
locked or unmapped, we do partial migration and leave some pages in
system memory. Several bugs found to copy pages and update GPU mapping
for this situation:

1. copy to vram should use migrate->npage which is total pages of range
as npages, not migrate->cpages which is number of pages can be migrated.

2. After partial copy, set VRAM res cursor as j + 1, j is number of
system pages copied plus 1 page to skip copy.

3. copy to ram, should collect all continuous VRAM pages and copy
together.

4. Call amdgpu_vm_update_range, should pass in offset as bytes, not
as number of pages.
Signed-off-by: default avatarPhilip Yang <Philip.Yang@amd.com>
Reviewed-by: default avatarFelix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
parent 4fac4fcf
...@@ -296,7 +296,7 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange, ...@@ -296,7 +296,7 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
struct migrate_vma *migrate, struct dma_fence **mfence, struct migrate_vma *migrate, struct dma_fence **mfence,
dma_addr_t *scratch) dma_addr_t *scratch)
{ {
uint64_t npages = migrate->cpages; uint64_t npages = migrate->npages;
struct device *dev = adev->dev; struct device *dev = adev->dev;
struct amdgpu_res_cursor cursor; struct amdgpu_res_cursor cursor;
dma_addr_t *src; dma_addr_t *src;
...@@ -344,7 +344,7 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange, ...@@ -344,7 +344,7 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
mfence); mfence);
if (r) if (r)
goto out_free_vram_pages; goto out_free_vram_pages;
amdgpu_res_next(&cursor, j << PAGE_SHIFT); amdgpu_res_next(&cursor, (j + 1) << PAGE_SHIFT);
j = 0; j = 0;
} else { } else {
amdgpu_res_next(&cursor, PAGE_SIZE); amdgpu_res_next(&cursor, PAGE_SIZE);
...@@ -590,7 +590,7 @@ svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange, ...@@ -590,7 +590,7 @@ svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
continue; continue;
} }
src[i] = svm_migrate_addr(adev, spage); src[i] = svm_migrate_addr(adev, spage);
if (i > 0 && src[i] != src[i - 1] + PAGE_SIZE) { if (j > 0 && src[i] != src[i - 1] + PAGE_SIZE) {
r = svm_migrate_copy_memory_gart(adev, dst + i - j, r = svm_migrate_copy_memory_gart(adev, dst + i - j,
src + i - j, j, src + i - j, j,
FROM_VRAM_TO_RAM, FROM_VRAM_TO_RAM,
......
...@@ -1295,7 +1295,7 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange, ...@@ -1295,7 +1295,7 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb, NULL, r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb, NULL,
last_start, prange->start + i, last_start, prange->start + i,
pte_flags, pte_flags,
last_start - prange->start, (last_start - prange->start) << PAGE_SHIFT,
bo_adev ? bo_adev->vm_manager.vram_base_offset : 0, bo_adev ? bo_adev->vm_manager.vram_base_offset : 0,
NULL, dma_addr, &vm->last_update); NULL, dma_addr, &vm->last_update);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment