gpu: nvgpu: Implement common nvgpu_mem_rd* functions

nvgpu_mem_rd*() functions were implemented per OS. They also used
nvgpu_pramin_access_batched() and implemented a big portion of logic
for using PRAMIN in OS specific code.

Make the implementation for the functions generic. Move all PRAMIN
logic to PRAMIN and simplify the interface provided by PRAMIN.

Change-Id: I1acb9e8d7d424325dc73314d5738cb2c9ebf7692
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1753708
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

4 files changed, 185 insertions, 60 deletions

diff --git a/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c b/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
index 855d455d..9f3b6cfa 100644
--- a/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
+++ b/drivers/gpu/nvgpu/common/mm/nvgpu_mem.c
@@ -177,3 +177,112 @@ u64 nvgpu_sgt_alignment(struct gk20a *g, struct nvgpu_sgt *sgt)
 
         return align;
 }
+
+u32 nvgpu_mem_rd32(struct gk20a *g, struct nvgpu_mem *mem, u32 w)
+{
+        u32 data = 0;
+
+        if (mem->aperture == APERTURE_SYSMEM) {
+                u32 *ptr = mem->cpu_va;
+
+                WARN_ON(!ptr);
+                data = ptr[w];
+        } else if (mem->aperture == APERTURE_VIDMEM) {
+                nvgpu_pramin_rd_n(g, mem, w * sizeof(u32), sizeof(u32), &data);
+        } else {
+                WARN_ON("Accessing unallocated nvgpu_mem");
+        }
+
+        return data;
+}
+
+u32 nvgpu_mem_rd(struct gk20a *g, struct nvgpu_mem *mem, u32 offset)
+{
+        WARN_ON(offset & 3);
+        return nvgpu_mem_rd32(g, mem, offset / sizeof(u32));
+}
+
+void nvgpu_mem_rd_n(struct gk20a *g, struct nvgpu_mem *mem,
+                u32 offset, void *dest, u32 size)
+{
+        WARN_ON(offset & 3);
+        WARN_ON(size & 3);
+
+        if (mem->aperture == APERTURE_SYSMEM) {
+                u8 *src = (u8 *)mem->cpu_va + offset;
+
+                WARN_ON(!mem->cpu_va);
+                memcpy(dest, src, size);
+        } else if (mem->aperture == APERTURE_VIDMEM) {
+                nvgpu_pramin_rd_n(g, mem, offset, size, dest);
+        } else {
+                WARN_ON("Accessing unallocated nvgpu_mem");
+        }
+}
+
+void nvgpu_mem_wr32(struct gk20a *g, struct nvgpu_mem *mem, u32 w, u32 data)
+{
+        if (mem->aperture == APERTURE_SYSMEM) {
+                u32 *ptr = mem->cpu_va;
+
+                WARN_ON(!ptr);
+                ptr[w] = data;
+        } else if (mem->aperture == APERTURE_VIDMEM) {
+                nvgpu_pramin_wr_n(g, mem, w * sizeof(u32), sizeof(u32), &data);
+                if (!mem->skip_wmb)
+                        nvgpu_wmb();
+        } else {
+                WARN_ON("Accessing unallocated nvgpu_mem");
+        }
+}
+
+void nvgpu_mem_wr(struct gk20a *g, struct nvgpu_mem *mem, u32 offset, u32 data)
+{
+        WARN_ON(offset & 3);
+        nvgpu_mem_wr32(g, mem, offset / sizeof(u32), data);
+}
+
+void nvgpu_mem_wr_n(struct gk20a *g, struct nvgpu_mem *mem, u32 offset,
+                void *src, u32 size)
+{
+        WARN_ON(offset & 3);
+        WARN_ON(size & 3);
+
+        if (mem->aperture == APERTURE_SYSMEM) {
+                u8 *dest = (u8 *)mem->cpu_va + offset;
+
+                WARN_ON(!mem->cpu_va);
+                memcpy(dest, src, size);
+        } else if (mem->aperture == APERTURE_VIDMEM) {
+                nvgpu_pramin_wr_n(g, mem, offset, size, src);
+                if (!mem->skip_wmb)
+                        nvgpu_wmb();
+        } else {
+                WARN_ON("Accessing unallocated nvgpu_mem");
+        }
+}
+
+void nvgpu_memset(struct gk20a *g, struct nvgpu_mem *mem, u32 offset,
+                u32 c, u32 size)
+{
+        WARN_ON(offset & 3);
+        WARN_ON(size & 3);
+        WARN_ON(c & ~0xff);
+
+        c &= 0xff;
+
+        if (mem->aperture == APERTURE_SYSMEM) {
+                u8 *dest = (u8 *)mem->cpu_va + offset;
+
+                WARN_ON(!mem->cpu_va);
+                memset(dest, c, size);
+        } else if (mem->aperture == APERTURE_VIDMEM) {
+                u32 repeat_value = c | (c << 8) | (c << 16) | (c << 24);
+
+                nvgpu_pramin_memset(g, mem, offset, size, repeat_value);
+                if (!mem->skip_wmb)
+                        nvgpu_wmb();
+        } else {
+                WARN_ON("Accessing unallocated nvgpu_mem");
+        }
+}
diff --git a/drivers/gpu/nvgpu/common/posix/io.c b/drivers/gpu/nvgpu/common/posix/io.c
index ce018940..dc32c20e 100644
--- a/drivers/gpu/nvgpu/common/posix/io.c
+++ b/drivers/gpu/nvgpu/common/posix/io.c
@@ -35,6 +35,11 @@ void nvgpu_writel(struct gk20a *g, u32 r, u32 v)
         BUG();
 }
 
+void nvgpu_writel_relaxed(struct gk20a *g, u32 r, u32 v)
+{
+        BUG();
+}
+
 u32 nvgpu_readl(struct gk20a *g, u32 r)
 {
         BUG();
diff --git a/drivers/gpu/nvgpu/common/posix/posix-nvgpu_mem.c b/drivers/gpu/nvgpu/common/posix/posix-nvgpu_mem.c
index 7f3bf9f1..fa92a7c6 100644
--- a/drivers/gpu/nvgpu/common/posix/posix-nvgpu_mem.c
+++ b/drivers/gpu/nvgpu/common/posix/posix-nvgpu_mem.c
@@ -27,65 +27,6 @@
 #include <nvgpu/nvgpu_mem.h>
 
 /*
- * DMA memory buffers - obviously we don't really have DMA in userspace but we
- * can emulate a lot of the DMA mem functionality for unit testing purposes.
- */
-
-u32 nvgpu_mem_rd32(struct gk20a *g, struct nvgpu_mem *mem, u32 w)
-{
-        u32 *mem_ptr = (u32 *)mem->cpu_va;
-
-        return mem_ptr[w];
-}
-
-u32 nvgpu_mem_rd(struct gk20a *g, struct nvgpu_mem *mem, u32 offset)
-{
-        if (offset & 0x3)
-                BUG();
-
-        return nvgpu_mem_rd32(g, mem, offset >> 2);
-}
-
-void nvgpu_mem_rd_n(struct gk20a *g, struct nvgpu_mem *mem, u32 offset,
-                    void *dest, u32 size)
-{
-        if (offset & 0x3 || size & 0x3)
-                BUG();
-
-        memcpy(dest, ((char *)mem->cpu_va) + offset, size);
-}
-
-void nvgpu_mem_wr32(struct gk20a *g, struct nvgpu_mem *mem, u32 w, u32 data)
-{
-        u32 *mem_ptr = (u32 *)mem->cpu_va;
-
-        mem_ptr[w] = data;
-}
-
-void nvgpu_mem_wr(struct gk20a *g, struct nvgpu_mem *mem, u32 offset, u32 data)
-{
-        if (offset & 0x3)
-                BUG();
-
-        nvgpu_mem_wr32(g, mem, offset >> 2, data);
-}
-
-void nvgpu_mem_wr_n(struct gk20a *g, struct nvgpu_mem *mem, u32 offset,
-                    void *src, u32 size)
-{
-        if (offset & 0x3 || size & 0x3)
-                BUG();
-
-        memcpy(((char *)mem->cpu_va) + offset, src, size);
-}
-
-void nvgpu_memset(struct gk20a *g, struct nvgpu_mem *mem, u32 offset,
-                  u32 c, u32 size)
-{
-        memset(((char *)mem->cpu_va) + offset, c, size);
-}
-
-/*
  * These functions are somewhat meaningless.
  */
 u64 nvgpu_mem_get_addr(struct gk20a *g, struct nvgpu_mem *mem)
diff --git a/drivers/gpu/nvgpu/common/pramin.c b/drivers/gpu/nvgpu/common/pramin.c
index 4c6a4a0d..99d588aa 100644
--- a/drivers/gpu/nvgpu/common/pramin.c
+++ b/drivers/gpu/nvgpu/common/pramin.c
@@ -28,11 +28,18 @@
 #include "gk20a/gk20a.h"
 
 /*
+ * This typedef is for functions that get called during the access_batched()
+ * operation.
+ */
+typedef void (*pramin_access_batch_fn)(struct gk20a *g, u32 start, u32 words,
+                                       u32 **arg);
+
+/*
  * The PRAMIN range is 1 MB, must change base addr if a buffer crosses that.
  * This same loop is used for read/write/memset. Offset and size in bytes.
  * One call to "loop" is done per range, with "arg" supplied.
  */
-void nvgpu_pramin_access_batched(struct gk20a *g, struct nvgpu_mem *mem,
+static void nvgpu_pramin_access_batched(struct gk20a *g, struct nvgpu_mem *mem,
                 u32 offset, u32 size, pramin_access_batch_fn loop, u32 **arg)
 {
         struct nvgpu_page_alloc *alloc = NULL;
@@ -87,6 +94,69 @@ void nvgpu_pramin_access_batched(struct gk20a *g, struct nvgpu_mem *mem,
         }
 }
 
+static void nvgpu_pramin_access_batch_rd_n(struct gk20a *g,
+                                           u32 start, u32 words, u32 **arg)
+{
+        u32 r = start, *dest_u32 = *arg;
+
+        while (words--) {
+                *dest_u32++ = nvgpu_readl(g, r);
+                r += sizeof(u32);
+        }
+
+        *arg = dest_u32;
+}
+
+void nvgpu_pramin_rd_n(struct gk20a *g, struct nvgpu_mem *mem,
+                                           u32 start, u32 size, void *dest)
+{
+        u32 *dest_u32 = dest;
+
+        return nvgpu_pramin_access_batched(g, mem, start, size,
+                        nvgpu_pramin_access_batch_rd_n, &dest_u32);
+}
+
+static void nvgpu_pramin_access_batch_wr_n(struct gk20a *g,
+                                           u32 start, u32 words, u32 **arg)
+{
+        u32 r = start, *src_u32 = *arg;
+
+        while (words--) {
+                nvgpu_writel_relaxed(g, r, *src_u32++);
+                r += sizeof(u32);
+        }
+
+        *arg = src_u32;
+}
+
+void nvgpu_pramin_wr_n(struct gk20a *g, struct nvgpu_mem *mem,
+                                           u32 start, u32 size, void *src)
+{
+        u32 *src_u32 = src;
+
+        return nvgpu_pramin_access_batched(g, mem, start, size,
+                        nvgpu_pramin_access_batch_wr_n, &src_u32);
+}
+
+static void nvgpu_pramin_access_batch_set(struct gk20a *g,
+                                          u32 start, u32 words, u32 **arg)
+{
+        u32 r = start, repeat = **arg;
+
+        while (words--) {
+                nvgpu_writel_relaxed(g, r, repeat);
+                r += sizeof(u32);
+        }
+}
+
+void nvgpu_pramin_memset(struct gk20a *g, struct nvgpu_mem *mem,
+                         u32 start, u32 size, u32 w)
+{
+        u32 *p = &w;
+
+        return nvgpu_pramin_access_batched(g, mem, start, size,
+                        nvgpu_pramin_access_batch_set, &p);
+}
 void nvgpu_init_pramin(struct mm_gk20a *mm)
 {
         mm->pramin_window = 0;