]> git.karo-electronics.de Git - linux-beck.git/commitdiff
drm/amdgpu: change parameter passing in the VM code
authorChristian König <christian.koenig@amd.com>
Fri, 18 Mar 2016 20:00:35 +0000 (21:00 +0100)
committerAlex Deucher <alexander.deucher@amd.com>
Mon, 2 May 2016 19:26:56 +0000 (15:26 -0400)
Make it more flexible by passing src and page addresses
directly instead of the structures they contain.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

index 6630732ea1bd2e9ebabfad2022e726cd426844b2..9eec3e1be667977eefba92f3c98278de1e22dc77 100644 (file)
@@ -349,8 +349,8 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
  * amdgpu_vm_update_pages - helper to call the right asic function
  *
  * @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: GTT hw access flags
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
  * @ib: indirect buffer to fill with commands
  * @pe: addr of the page entry
  * @addr: dst addr to write into pe
@@ -362,8 +362,8 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
  * to setup the page table using the DMA.
  */
 static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
-                                  struct amdgpu_gart *gtt,
-                                  uint32_t gtt_flags,
+                                  uint64_t src,
+                                  dma_addr_t *pages_addr,
                                   struct amdgpu_ib *ib,
                                   uint64_t pe, uint64_t addr,
                                   unsigned count, uint32_t incr,
@@ -371,12 +371,11 @@ static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
 {
        trace_amdgpu_vm_set_page(pe, addr, count, incr, flags);
 
-       if ((gtt == &adev->gart) && (flags == gtt_flags)) {
-               uint64_t src = gtt->table_addr + (addr >> 12) * 8;
+       if (src) {
+               src += (addr >> 12) * 8;
                amdgpu_vm_copy_pte(adev, ib, pe, src, count);
 
-       } else if (gtt) {
-               dma_addr_t *pages_addr = gtt->pages_addr;
+       } else if (pages_addr) {
                amdgpu_vm_write_pte(adev, ib, pages_addr, pe, addr,
                                    count, incr, flags);
 
@@ -426,7 +425,7 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev,
        if (r)
                goto error;
 
-       amdgpu_vm_update_pages(adev, NULL, 0, &job->ibs[0], addr, 0, entries,
+       amdgpu_vm_update_pages(adev, 0, NULL, &job->ibs[0], addr, 0, entries,
                               0, 0);
        amdgpu_ring_pad_ib(ring, &job->ibs[0]);
 
@@ -536,7 +535,7 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
                    ((last_pt + incr * count) != pt)) {
 
                        if (count) {
-                               amdgpu_vm_update_pages(adev, NULL, 0, ib,
+                               amdgpu_vm_update_pages(adev, 0, NULL, ib,
                                                       last_pde, last_pt,
                                                       count, incr,
                                                       AMDGPU_PTE_VALID);
@@ -551,7 +550,7 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
        }
 
        if (count)
-               amdgpu_vm_update_pages(adev, NULL, 0, ib, last_pde, last_pt,
+               amdgpu_vm_update_pages(adev, 0, NULL, ib, last_pde, last_pt,
                                       count, incr, AMDGPU_PTE_VALID);
 
        if (ib->length_dw != 0) {
@@ -584,8 +583,8 @@ error_free:
  * amdgpu_vm_frag_ptes - add fragment information to PTEs
  *
  * @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: GTT hw mapping flags
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
  * @ib: IB for the update
  * @pe_start: first PTE to handle
  * @pe_end: last PTE to handle
@@ -593,8 +592,8 @@ error_free:
  * @flags: hw mapping flags
  */
 static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
-                               struct amdgpu_gart *gtt,
-                               uint32_t gtt_flags,
+                               uint64_t src,
+                               dma_addr_t *pages_addr,
                                struct amdgpu_ib *ib,
                                uint64_t pe_start, uint64_t pe_end,
                                uint64_t addr, uint32_t flags)
@@ -632,10 +631,11 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
                return;
 
        /* system pages are non continuously */
-       if (gtt || !(flags & AMDGPU_PTE_VALID) || (frag_start >= frag_end)) {
+       if (src || pages_addr || !(flags & AMDGPU_PTE_VALID) ||
+           (frag_start >= frag_end)) {
 
                count = (pe_end - pe_start) / 8;
-               amdgpu_vm_update_pages(adev, gtt, gtt_flags, ib, pe_start,
+               amdgpu_vm_update_pages(adev, src, pages_addr, ib, pe_start,
                                       addr, count, AMDGPU_GPU_PAGE_SIZE,
                                       flags);
                return;
@@ -644,21 +644,21 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
        /* handle the 4K area at the beginning */
        if (pe_start != frag_start) {
                count = (frag_start - pe_start) / 8;
-               amdgpu_vm_update_pages(adev, NULL, 0, ib, pe_start, addr,
+               amdgpu_vm_update_pages(adev, 0, NULL, ib, pe_start, addr,
                                       count, AMDGPU_GPU_PAGE_SIZE, flags);
                addr += AMDGPU_GPU_PAGE_SIZE * count;
        }
 
        /* handle the area in the middle */
        count = (frag_end - frag_start) / 8;
-       amdgpu_vm_update_pages(adev, NULL, 0, ib, frag_start, addr, count,
+       amdgpu_vm_update_pages(adev, 0, NULL, ib, frag_start, addr, count,
                               AMDGPU_GPU_PAGE_SIZE, flags | frag_flags);
 
        /* handle the 4K area at the end */
        if (frag_end != pe_end) {
                addr += AMDGPU_GPU_PAGE_SIZE * count;
                count = (pe_end - frag_end) / 8;
-               amdgpu_vm_update_pages(adev, NULL, 0, ib, frag_end, addr,
+               amdgpu_vm_update_pages(adev, 0, NULL, ib, frag_end, addr,
                                       count, AMDGPU_GPU_PAGE_SIZE, flags);
        }
 }
@@ -667,8 +667,8 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
  * amdgpu_vm_update_ptes - make sure that page tables are valid
  *
  * @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: GTT hw mapping flags
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
  * @vm: requested vm
  * @start: start of GPU address range
  * @end: end of GPU address range
@@ -678,8 +678,8 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
  * Update the page tables in the range @start - @end.
  */
 static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
-                                 struct amdgpu_gart *gtt,
-                                 uint32_t gtt_flags,
+                                 uint64_t src,
+                                 dma_addr_t *pages_addr,
                                  struct amdgpu_vm *vm,
                                  struct amdgpu_ib *ib,
                                  uint64_t start, uint64_t end,
@@ -707,7 +707,7 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
 
                if (last_pe_end != pe_start) {
 
-                       amdgpu_vm_frag_ptes(adev, gtt, gtt_flags, ib,
+                       amdgpu_vm_frag_ptes(adev, src, pages_addr, ib,
                                            last_pe_start, last_pe_end,
                                            last_dst, flags);
 
@@ -722,17 +722,16 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
                dst += nptes * AMDGPU_GPU_PAGE_SIZE;
        }
 
-       amdgpu_vm_frag_ptes(adev, gtt, gtt_flags, ib,
-                           last_pe_start, last_pe_end,
-                           last_dst, flags);
+       amdgpu_vm_frag_ptes(adev, src, pages_addr, ib, last_pe_start,
+                           last_pe_end, last_dst, flags);
 }
 
 /**
  * amdgpu_vm_bo_update_mapping - update a mapping in the vm page table
  *
  * @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: flags as they are used for GTT
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
  * @vm: requested vm
  * @start: start of mapped range
  * @last: last mapped entry
@@ -744,8 +743,8 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
  * Returns 0 for success, -EINVAL for failure.
  */
 static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
-                                      struct amdgpu_gart *gtt,
-                                      uint32_t gtt_flags,
+                                      uint64_t src,
+                                      dma_addr_t *pages_addr,
                                       struct amdgpu_vm *vm,
                                       uint64_t start, uint64_t last,
                                       uint32_t flags, uint64_t addr,
@@ -776,11 +775,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
        /* padding, etc. */
        ndw = 64;
 
-       if ((gtt == &adev->gart) && (flags == gtt_flags)) {
+       if (src) {
                /* only copy commands needed */
                ndw += ncmds * 7;
 
-       } else if (gtt) {
+       } else if (pages_addr) {
                /* header for write data commands */
                ndw += ncmds * 4;
 
@@ -810,8 +809,8 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
        if (r)
                goto error_free;
 
-       amdgpu_vm_update_ptes(adev, gtt, gtt_flags, vm, ib, start, last + 1,
-                             addr, flags);
+       amdgpu_vm_update_ptes(adev, src, pages_addr, vm, ib, start,
+                             last + 1, addr, flags);
 
        amdgpu_ring_pad_ib(ring, ib);
        WARN_ON(ib->length_dw > ndw);
@@ -853,12 +852,13 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
                                      uint32_t gtt_flags,
                                      struct amdgpu_vm *vm,
                                      struct amdgpu_bo_va_mapping *mapping,
-                                     uint64_t addr, struct fence **fence)
+                                     uint32_t flags, uint64_t addr,
+                                     struct fence **fence)
 {
        const uint64_t max_size = 64ULL * 1024ULL * 1024ULL / AMDGPU_GPU_PAGE_SIZE;
 
-       uint64_t start = mapping->it.start;
-       uint32_t flags = gtt_flags;
+       uint64_t src = 0, start = mapping->it.start;
+       dma_addr_t *pages_addr = NULL;
        int r;
 
        /* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here
@@ -871,10 +871,17 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
 
        trace_amdgpu_vm_bo_update(mapping);
 
+       if (gtt) {
+               if (flags == gtt_flags)
+                       src = adev->gart.table_addr + (addr >> 12) * 8;
+               else
+                       pages_addr = &gtt->pages_addr[addr >> 12];
+               addr = 0;
+       }
        addr += mapping->offset;
 
-       if (!gtt || ((gtt == &adev->gart) && (flags == gtt_flags)))
-               return amdgpu_vm_bo_update_mapping(adev, gtt, gtt_flags, vm,
+       if (!gtt || src)
+               return amdgpu_vm_bo_update_mapping(adev, src, pages_addr, vm,
                                                   start, mapping->it.last,
                                                   flags, addr, fence);
 
@@ -882,7 +889,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
                uint64_t last;
 
                last = min((uint64_t)mapping->it.last, start + max_size - 1);
-               r = amdgpu_vm_bo_update_mapping(adev, gtt, gtt_flags, vm,
+               r = amdgpu_vm_bo_update_mapping(adev, src, pages_addr, vm,
                                                start, last, flags, addr,
                                                fence);
                if (r)
@@ -914,7 +921,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
        struct amdgpu_vm *vm = bo_va->vm;
        struct amdgpu_bo_va_mapping *mapping;
        struct amdgpu_gart *gtt = NULL;
-       uint32_t flags;
+       uint32_t gtt_flags, flags;
        uint64_t addr;
        int r;
 
@@ -937,6 +944,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
        }
 
        flags = amdgpu_ttm_tt_pte_flags(adev, bo_va->bo->tbo.ttm, mem);
+       gtt_flags = (adev == bo_va->bo->adev) ? flags : 0;
 
        spin_lock(&vm->status_lock);
        if (!list_empty(&bo_va->vm_status))
@@ -944,8 +952,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
        spin_unlock(&vm->status_lock);
 
        list_for_each_entry(mapping, &bo_va->invalids, list) {
-               r = amdgpu_vm_bo_split_mapping(adev, gtt, flags, vm, mapping, addr,
-                                              &bo_va->last_pt_update);
+               r = amdgpu_vm_bo_split_mapping(adev, gtt, gtt_flags, vm, mapping,
+                                              flags, addr, &bo_va->last_pt_update);
                if (r)
                        return r;
        }
@@ -991,7 +999,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
                list_del(&mapping->list);
 
                r = amdgpu_vm_bo_split_mapping(adev, NULL, 0, vm, mapping,
-                                              0, NULL);
+                                              0, 0, NULL);
                kfree(mapping);
                if (r)
                        return r;