drm/amdgpu: move more logic into amdgpu_vm_map_gart v3

No need to duplicate that code over and over again. Also stop using the
flags to determine if we need to map the addresses.

v2: constify the pages_addr
v3: rebased, fix typo in commit message

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>

5 files changed, 33 insertions, 42 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 532e0a17dd9a..1529e0aecb83 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -283,7 +283,7 @@ struct amdgpu_vm_pte_funcs {
                          unsigned count);
         /* write pte one entry at a time with addr mapping */
         void (*write_pte)(struct amdgpu_ib *ib,
-                          uint64_t pe,
+                          const dma_addr_t *pages_addr, uint64_t pe,
                           uint64_t addr, unsigned count,
                           uint32_t incr, uint32_t flags);
         /* for linear pte/pde updates without addr mapping */
@@ -962,7 +962,7 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 void amdgpu_vm_flush(struct amdgpu_ring *ring,
                      struct amdgpu_vm *vm,
                      struct fence *updates);
-uint64_t amdgpu_vm_map_gart(struct amdgpu_device *adev, uint64_t addr);
+uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr);
 int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
                                     struct amdgpu_vm *vm);
 int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
@@ -2198,7 +2198,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid))
 #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags))
 #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count)))
-#define amdgpu_vm_write_pte(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (addr), (count), (incr), (flags)))
+#define amdgpu_vm_write_pte(adev, ib, pa, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pa), (pe), (addr), (count), (incr), (flags)))
 #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))
 #define amdgpu_vm_pad_ib(adev, ib) ((adev)->vm_manager.vm_pte_funcs->pad_ib((ib)))
 #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib)))
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index dfbcc64ef9fa..ae3b275f2a38 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -299,9 +299,14 @@ static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
                 uint64_t src = adev->gart.table_addr + (addr >> 12) * 8;
                 amdgpu_vm_copy_pte(adev, ib, pe, src, count);
 
-        } else if ((flags & AMDGPU_PTE_SYSTEM) || (count < 3)) {
-                amdgpu_vm_write_pte(adev, ib, pe, addr,
-                                      count, incr, flags);
+        } else if (flags & AMDGPU_PTE_SYSTEM) {
+                dma_addr_t *pages_addr = adev->gart.pages_addr;
+                amdgpu_vm_write_pte(adev, ib, pages_addr, pe, addr,
+                                    count, incr, flags);
+
+        } else if (count < 3) {
+                amdgpu_vm_write_pte(adev, ib, NULL, pe, addr,
+                                    count, incr, flags);
 
         } else {
                 amdgpu_vm_set_pte_pde(adev, ib, pe, addr,
@@ -378,24 +383,31 @@ error:
 }
 
 /**
- * amdgpu_vm_map_gart - get the physical address of a gart page
+ * amdgpu_vm_map_gart - Resolve gart mapping of addr
  *
- * @adev: amdgpu_device pointer
+ * @pages_addr: optional DMA address to use for lookup
  * @addr: the unmapped addr
  *
  * Look up the physical address of the page that the pte resolves
- * to (cayman+).
- * Returns the physical address of the page.
+ * to and return the pointer for the page table entry.
  */
-uint64_t amdgpu_vm_map_gart(struct amdgpu_device *adev, uint64_t addr)
+uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr)
 {
         uint64_t result;
 
-        /* page table offset */
-        result = adev->gart.pages_addr[addr >> PAGE_SHIFT];
+        if (pages_addr) {
+                /* page table offset */
+                result = pages_addr[addr >> PAGE_SHIFT];
+
+                /* in case cpu page size != gpu page size*/
+                result |= addr & (~PAGE_MASK);
+
+        } else {
+                /* No mapping required */
+                result = addr;
+        }
 
-        /* in case cpu page size != gpu page size*/
-        result |= addr & (~PAGE_MASK);
+        result &= 0xFFFFFFFFFFFFF000ULL;
 
         return result;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index e8a48ae8d360..abfa3f2f15fa 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -714,7 +714,7 @@ static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib,
  * Update PTEs by writing them manually using sDMA (CIK).
  */
 static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib,
-                                  uint64_t pe,
+                                  const dma_addr_t *pages_addr, uint64_t pe,
                                   uint64_t addr, unsigned count,
                                   uint32_t incr, uint32_t flags)
 {
@@ -733,14 +733,7 @@ static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib,
                 ib->ptr[ib->length_dw++] = upper_32_bits(pe);
                 ib->ptr[ib->length_dw++] = ndw;
                 for (; ndw > 0; ndw -= 2, --count, pe += 8) {
-                        if (flags & AMDGPU_PTE_SYSTEM) {
-                                value = amdgpu_vm_map_gart(ib->ring->adev, addr);
-                                value &= 0xFFFFFFFFFFFFF000ULL;
-                        } else if (flags & AMDGPU_PTE_VALID) {
-                                value = addr;
-                        } else {
-                                value = 0;
-                        }
+                        value = amdgpu_vm_map_gart(pages_addr, addr);
                         addr += incr;
                         value |= flags;
                         ib->ptr[ib->length_dw++] = value;
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index 9fae4bf1a6c6..750231d06aa0 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -772,7 +772,7 @@ static void sdma_v2_4_vm_copy_pte(struct amdgpu_ib *ib,
  * Update PTEs by writing them manually using sDMA (CIK).
  */
 static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib,
-                                   uint64_t pe,
+                                   const dma_addr_t *pages_addr, uint64_t pe,
                                    uint64_t addr, unsigned count,
                                    uint32_t incr, uint32_t flags)
 {
@@ -791,14 +791,7 @@ static void sdma_v2_4_vm_write_pte(struct amdgpu_ib *ib,
                 ib->ptr[ib->length_dw++] = upper_32_bits(pe);
                 ib->ptr[ib->length_dw++] = ndw;
                 for (; ndw > 0; ndw -= 2, --count, pe += 8) {
-                        if (flags & AMDGPU_PTE_SYSTEM) {
-                                value = amdgpu_vm_map_gart(ib->ring->adev, addr);
-                                value &= 0xFFFFFFFFFFFFF000ULL;
-                        } else if (flags & AMDGPU_PTE_VALID) {
-                                value = addr;
-                        } else {
-                                value = 0;
-                        }
+                        value = amdgpu_vm_map_gart(pages_addr, addr);
                         addr += incr;
                         value |= flags;
                         ib->ptr[ib->length_dw++] = value;
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index b2fbf96dad7a..7af4b57ad3d4 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -922,7 +922,7 @@ static void sdma_v3_0_vm_copy_pte(struct amdgpu_ib *ib,
  * Update PTEs by writing them manually using sDMA (CIK).
  */
 static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib,
-                                   uint64_t pe,
+                                   const dma_addr_t *pages_addr, uint64_t pe,
                                    uint64_t addr, unsigned count,
                                    uint32_t incr, uint32_t flags)
 {
@@ -941,14 +941,7 @@ static void sdma_v3_0_vm_write_pte(struct amdgpu_ib *ib,
                 ib->ptr[ib->length_dw++] = upper_32_bits(pe);
                 ib->ptr[ib->length_dw++] = ndw;
                 for (; ndw > 0; ndw -= 2, --count, pe += 8) {
-                        if (flags & AMDGPU_PTE_SYSTEM) {
-                                value = amdgpu_vm_map_gart(ib->ring->adev, addr);
-                                value &= 0xFFFFFFFFFFFFF000ULL;
-                        } else if (flags & AMDGPU_PTE_VALID) {
-                                value = addr;
-                        } else {
-                                value = 0;
-                        }
+                        value = amdgpu_vm_map_gart(pages_addr, addr);
                         addr += incr;
                         value |= flags;
                         ib->ptr[ib->length_dw++] = value;