gpu: nvgpu: add placeholder for IPA to PA

Add __nvgpu_sgl_phys function that can be used to implement IPA
to PA translation in a subsequent change.
Adapt existing function prototypes to add pointer to gpu context,
as we will need to check if IPA to PA translation is needed.

JIRA EVLR-2442
Bug 200392719

Change-Id: I5a734c958c8277d1bf673c020dafb31263f142d6
Signed-off-by: Thomas Fleury <tfleury@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1673142
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

9 files changed, 37 insertions, 24 deletions

diff --git a/drivers/gpu/nvgpu/common/linux/nvgpu_mem.c b/drivers/gpu/nvgpu/common/linux/nvgpu_mem.c
index d4549e1b..63a14f40 100644
--- a/drivers/gpu/nvgpu/common/linux/nvgpu_mem.c
+++ b/drivers/gpu/nvgpu/common/linux/nvgpu_mem.c
@@ -34,6 +34,11 @@
 #include "gk20a/gk20a.h"
 #include "gk20a/mm_gk20a.h"
 
+static inline u64 __nvgpu_sgl_phys(struct gk20a *g, struct nvgpu_sgl *sgl)
+{
+        return sg_phys((struct scatterlist *)sgl);
+}
+
 int nvgpu_mem_begin(struct gk20a *g, struct nvgpu_mem *mem)
 {
         void *cpu_va;
@@ -309,10 +314,12 @@ u64 nvgpu_mem_get_addr_sgl(struct gk20a *g, struct scatterlist *sgl)
 {
         if (nvgpu_is_enabled(g, NVGPU_MM_USE_PHYSICAL_SG) ||
             !nvgpu_iommuable(g))
-                return g->ops.mm.gpu_phys_addr(g, NULL, sg_phys(sgl));
+                return g->ops.mm.gpu_phys_addr(g, NULL,
+                        __nvgpu_sgl_phys(g, (struct nvgpu_sgl *)sgl));
 
         if (sg_dma_address(sgl) == 0)
-                return g->ops.mm.gpu_phys_addr(g, NULL, sg_phys(sgl));
+                return g->ops.mm.gpu_phys_addr(g, NULL,
+                        __nvgpu_sgl_phys(g, (struct nvgpu_sgl *)sgl));
 
         if (sg_dma_address(sgl) == DMA_ERROR_CODE)
                 return 0;
@@ -376,7 +383,7 @@ u64 nvgpu_mem_get_phys_addr(struct gk20a *g, struct nvgpu_mem *mem)
         if (mem->aperture == APERTURE_VIDMEM)
                 return nvgpu_mem_get_addr(g, mem);
 
-        return sg_phys(mem->priv.sgt->sgl);
+        return __nvgpu_sgl_phys(g, (struct nvgpu_sgl *)mem->priv.sgt->sgl);
 }
 
 /*
@@ -501,9 +508,9 @@ static struct nvgpu_sgl *nvgpu_mem_linux_sgl_next(struct nvgpu_sgl *sgl)
         return (struct nvgpu_sgl *)sg_next((struct scatterlist *)sgl);
 }
 
-static u64 nvgpu_mem_linux_sgl_phys(struct nvgpu_sgl *sgl)
+static u64 nvgpu_mem_linux_sgl_phys(struct gk20a *g, struct nvgpu_sgl *sgl)
 {
-        return (u64)sg_phys((struct scatterlist *)sgl);
+        return (u64)__nvgpu_sgl_phys(g, sgl);
 }
 
 static u64 nvgpu_mem_linux_sgl_dma(struct nvgpu_sgl *sgl)
@@ -522,7 +529,7 @@ static u64 nvgpu_mem_linux_sgl_gpu_addr(struct gk20a *g,
 {
         if (sg_dma_address((struct scatterlist *)sgl) == 0)
                 return g->ops.mm.gpu_phys_addr(g, attrs,
-                        sg_phys((struct scatterlist *)sgl));
+                                __nvgpu_sgl_phys(g, sgl));
 
         if (sg_dma_address((struct scatterlist *)sgl) == DMA_ERROR_CODE)
                 return 0;
diff --git a/drivers/gpu/nvgpu/common/mm/gmmu.c b/drivers/gpu/nvgpu/common/mm/gmmu.c
index 3b57e781..44e540dc 100644
--- a/drivers/gpu/nvgpu/common/mm/gmmu.c
+++ b/drivers/gpu/nvgpu/common/mm/gmmu.c
@@ -543,7 +543,7 @@ static int __nvgpu_gmmu_do_update_page_table(struct vm_gk20a *vm,
                 }
 
                 phys_addr = g->ops.mm.gpu_phys_addr(g, attrs,
-                            nvgpu_sgt_get_phys(sgt, sgl)) + space_to_skip;
+                            nvgpu_sgt_get_phys(g, sgt, sgl)) + space_to_skip;
                 chunk_length = min(length,
                         nvgpu_sgt_get_length(sgt, sgl) - space_to_skip);
 
@@ -629,7 +629,7 @@ static int __nvgpu_gmmu_update_page_table(struct vm_gk20a *vm,
                    sgt ? "MAP" : "UNMAP",
                    virt_addr,
                    length,
-                   sgt ? nvgpu_sgt_get_phys(sgt, sgt->sgl) : 0,
+                   sgt ? nvgpu_sgt_get_phys(g, sgt, sgt->sgl) : 0,
                    space_to_skip,
                    page_size >> 10,
                    nvgpu_gmmu_perm_str(attrs->rw_flag),
diff --git a/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c b/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
index 2b32d869..4def4938 100644
--- a/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
+++ b/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
@@ -81,9 +81,10 @@ struct nvgpu_sgl *nvgpu_sgt_get_next(struct nvgpu_sgt *sgt,
         return sgt->ops->sgl_next(sgl);
 }
 
-u64 nvgpu_sgt_get_phys(struct nvgpu_sgt *sgt, struct nvgpu_sgl *sgl)
+u64 nvgpu_sgt_get_phys(struct gk20a *g, struct nvgpu_sgt *sgt,
+                       struct nvgpu_sgl *sgl)
 {
-        return sgt->ops->sgl_phys(sgl);
+        return sgt->ops->sgl_phys(g, sgl);
 }
 
 u64 nvgpu_sgt_get_dma(struct nvgpu_sgt *sgt, struct nvgpu_sgl *sgl)
@@ -156,7 +157,7 @@ u64 nvgpu_sgt_alignment(struct gk20a *g, struct nvgpu_sgt *sgt)
          * of the SGT.
          */
         nvgpu_sgt_for_each_sgl(sgl, sgt) {
-                chunk_align = 1ULL << __ffs(nvgpu_sgt_get_phys(sgt, sgl) |
+                chunk_align = 1ULL << __ffs(nvgpu_sgt_get_phys(g, sgt, sgl) |
                                             nvgpu_sgt_get_length(sgt, sgl));
 
                 if (align)
diff --git a/drivers/gpu/nvgpu/common/mm/page_allocator.c b/drivers/gpu/nvgpu/common/mm/page_allocator.c
index 13ccc48b..773d33ef 100644
--- a/drivers/gpu/nvgpu/common/mm/page_allocator.c
+++ b/drivers/gpu/nvgpu/common/mm/page_allocator.c
@@ -160,7 +160,7 @@ static struct nvgpu_sgl *nvgpu_page_alloc_sgl_next(struct nvgpu_sgl *sgl)
         return (struct nvgpu_sgl *)sgl_impl->next;
 }
 
-static u64 nvgpu_page_alloc_sgl_phys(struct nvgpu_sgl *sgl)
+static u64 nvgpu_page_alloc_sgl_phys(struct gk20a *g, struct nvgpu_sgl *sgl)
 {
         struct nvgpu_mem_sgl *sgl_impl = (struct nvgpu_mem_sgl *)sgl;
 
@@ -231,11 +231,12 @@ static void __nvgpu_free_pages(struct nvgpu_page_allocator *a,
                                bool free_buddy_alloc)
 {
         struct nvgpu_sgl *sgl = alloc->sgt.sgl;
+        struct gk20a *g = a->owner->g;
 
         if (free_buddy_alloc) {
                 while (sgl) {
                         nvgpu_free(&a->source_allocator,
-                                   nvgpu_sgt_get_phys(&alloc->sgt, sgl));
+                                   nvgpu_sgt_get_phys(g, &alloc->sgt, sgl));
                         sgl = nvgpu_sgt_get_next(&alloc->sgt, sgl);
                 }
         }
@@ -615,6 +616,7 @@ fail:
 static struct nvgpu_page_alloc *__nvgpu_alloc_pages(
         struct nvgpu_page_allocator *a, u64 len)
 {
+        struct gk20a *g = a->owner->g;
         struct nvgpu_page_alloc *alloc = NULL;
         struct nvgpu_sgl *sgl;
         u64 pages;
@@ -635,7 +637,7 @@ static struct nvgpu_page_alloc *__nvgpu_alloc_pages(
         while (sgl) {
                 palloc_dbg(a, "  Chunk %2d: 0x%010llx + 0x%llx",
                            i++,
-                           nvgpu_sgt_get_phys(&alloc->sgt, sgl),
+                           nvgpu_sgt_get_phys(g, &alloc->sgt, sgl),
                            nvgpu_sgt_get_length(&alloc->sgt, sgl));
                 sgl = nvgpu_sgt_get_next(&alloc->sgt, sgl);
         }
@@ -779,6 +781,7 @@ static u64 nvgpu_page_alloc_fixed(struct nvgpu_allocator *__a,
         struct nvgpu_page_allocator *a = page_allocator(__a);
         struct nvgpu_page_alloc *alloc = NULL;
         struct nvgpu_sgl *sgl;
+        struct gk20a *g = a->owner->g;
         u64 aligned_len, pages;
         int i = 0;
 
@@ -802,7 +805,7 @@ static u64 nvgpu_page_alloc_fixed(struct nvgpu_allocator *__a,
         while (sgl) {
                 palloc_dbg(a, "  Chunk %2d: 0x%010llx + 0x%llx",
                            i++,
-                           nvgpu_sgt_get_phys(&alloc->sgt, sgl),
+                           nvgpu_sgt_get_phys(g, &alloc->sgt, sgl),
                            nvgpu_sgt_get_length(&alloc->sgt, sgl));
                 sgl = nvgpu_sgt_get_next(&alloc->sgt, sgl);
         }
diff --git a/drivers/gpu/nvgpu/common/mm/vidmem.c b/drivers/gpu/nvgpu/common/mm/vidmem.c
index e4137090..f534f45c 100644
--- a/drivers/gpu/nvgpu/common/mm/vidmem.c
+++ b/drivers/gpu/nvgpu/common/mm/vidmem.c
@@ -430,7 +430,7 @@ int nvgpu_vidmem_clear(struct gk20a *g, struct nvgpu_mem *mem)
                 err = gk20a_ce_execute_ops(g,
                         g->mm.vidmem.ce_ctx_id,
                         0,
-                        nvgpu_sgt_get_phys(&alloc->sgt, sgl),
+                        nvgpu_sgt_get_phys(g, &alloc->sgt, sgl),
                         nvgpu_sgt_get_length(&alloc->sgt, sgl),
                         0x00000000,
                         NVGPU_CE_DST_LOCATION_LOCAL_FB,
@@ -445,7 +445,7 @@ int nvgpu_vidmem_clear(struct gk20a *g, struct nvgpu_mem *mem)
                 }
 
                 vidmem_dbg(g, "  > [0x%llx  +0x%llx]",
-                           nvgpu_sgt_get_phys(&alloc->sgt, sgl),
+                           nvgpu_sgt_get_phys(g, &alloc->sgt, sgl),
                            nvgpu_sgt_get_length(&alloc->sgt, sgl));
 
                 gk20a_last_fence = gk20a_fence_out;
diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index 0def724d..5bd4dc57 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -867,6 +867,7 @@ static int gk20a_init_fifo_setup_sw(struct gk20a *g)
         struct fifo_gk20a *f = &g->fifo;
         unsigned int chid, i;
         int err = 0;
+        u64 userd_base;
 
         gk20a_dbg_fn("");
 
@@ -929,9 +930,9 @@ static int gk20a_init_fifo_setup_sw(struct gk20a *g)
         }
         gk20a_dbg(gpu_dbg_map, "userd gpu va = 0x%llx", f->userd.gpu_va);
 
+        userd_base = nvgpu_mem_get_addr(g, &f->userd);
         for (chid = 0; chid < f->num_channels; chid++) {
-                f->channel[chid].userd_iova =
-                        nvgpu_mem_get_addr(g, &f->userd) +
+                f->channel[chid].userd_iova = userd_base +
                         chid * f->userd_entry_size;
                 f->channel[chid].userd_gpu_va =
                         f->userd.gpu_va + chid * f->userd_entry_size;
diff --git a/drivers/gpu/nvgpu/gk20a/pramin_gk20a.c b/drivers/gpu/nvgpu/gk20a/pramin_gk20a.c
index 67fd2480..a76e2580 100644
--- a/drivers/gpu/nvgpu/gk20a/pramin_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/pramin_gk20a.c
@@ -34,7 +34,7 @@
 u32 gk20a_pramin_enter(struct gk20a *g, struct nvgpu_mem *mem,
                        struct nvgpu_sgt *sgt, struct nvgpu_sgl *sgl, u32 w)
 {
-        u64 bufbase = nvgpu_sgt_get_phys(sgt, sgl);
+        u64 bufbase = nvgpu_sgt_get_phys(g, sgt, sgl);
         u64 addr = bufbase + w * sizeof(u32);
         u32 hi = (u32)((addr & ~(u64)0xfffff)
                 >> bus_bar0_window_target_bar0_window_base_shift_v());
@@ -48,7 +48,7 @@ u32 gk20a_pramin_enter(struct gk20a *g, struct nvgpu_mem *mem,
         gk20a_dbg(gpu_dbg_mem,
                         "0x%08x:%08x begin for %p,%p at [%llx,%llx] (sz %llx)",
                         hi, lo, mem, sgl, bufbase,
-                        bufbase + nvgpu_sgt_get_phys(sgt, sgl),
+                        bufbase + nvgpu_sgt_get_phys(g, sgt, sgl),
                         nvgpu_sgt_get_length(sgt, sgl));
 
         WARN_ON(!bufbase);
diff --git a/drivers/gpu/nvgpu/include/nvgpu/nvgpu_mem.h b/drivers/gpu/nvgpu/include/nvgpu/nvgpu_mem.h
index 373c3eef..be0fa6bf 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/nvgpu_mem.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/nvgpu_mem.h
@@ -68,7 +68,7 @@ struct nvgpu_sgl;
 
 struct nvgpu_sgt_ops {
         struct nvgpu_sgl *(*sgl_next)(struct nvgpu_sgl *sgl);
-        u64   (*sgl_phys)(struct nvgpu_sgl *sgl);
+        u64   (*sgl_phys)(struct gk20a *g, struct nvgpu_sgl *sgl);
         u64   (*sgl_dma)(struct nvgpu_sgl *sgl);
         u64   (*sgl_length)(struct nvgpu_sgl *sgl);
         u64   (*sgl_gpu_addr)(struct gk20a *g, struct nvgpu_sgl *sgl,
@@ -254,7 +254,8 @@ struct nvgpu_sgt *nvgpu_sgt_create_from_mem(struct gk20a *g,
 
 struct nvgpu_sgl *nvgpu_sgt_get_next(struct nvgpu_sgt *sgt,
                                           struct nvgpu_sgl *sgl);
-u64 nvgpu_sgt_get_phys(struct nvgpu_sgt *sgt, struct nvgpu_sgl *sgl);
+u64 nvgpu_sgt_get_phys(struct gk20a *g, struct nvgpu_sgt *sgt,
+                       struct nvgpu_sgl *sgl);
 u64 nvgpu_sgt_get_dma(struct nvgpu_sgt *sgt, struct nvgpu_sgl *sgl);
 u64 nvgpu_sgt_get_length(struct nvgpu_sgt *sgt, struct nvgpu_sgl *sgl);
 u64 nvgpu_sgt_get_gpu_addr(struct gk20a *g, struct nvgpu_sgt *sgt,
diff --git a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_mm_gp10b.c b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_mm_gp10b.c
index ca517975..e615c486 100644
--- a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_mm_gp10b.c
+++ b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_mm_gp10b.c
@@ -122,7 +122,7 @@ u64 vgpu_gp10b_locked_gmmu_map(struct vm_gk20a *vm,
                         continue;
                 }
 
-                phys_addr = nvgpu_sgt_get_phys(sgt, sgl) + space_to_skip;
+                phys_addr = nvgpu_sgt_get_phys(g, sgt, sgl) + space_to_skip;
                 chunk_length = min(size,
                            nvgpu_sgt_get_length(sgt, sgl) - space_to_skip);