nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-23 18:16:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: split vidmem_is_vidmem

Browse Source 
As the vidmem_is_vidmem flag has got two separate meanings in one bit,
split it in two bits into the enabled() API:

Add NVGPU_MM_HONORS_APERTURE bit, which is the same as vidmem_is_vidmem
with its original meaning, and use it to test which aperture bits to
write to hardware.

Add NVGPU_MM_UNIFIED_MEMORY bit, which has the opposite meaning: that
the GPU shares the SoC memory. When this flag is false, the GPU has its
own local video memory.

Jira NVGPU-86

Change-Id: I2d0bed3b1ede5a712be99323d3035b154bb23c3a
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: http://git-master/r/1496080
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
This commit is contained in:
Konsta Holtta
2017-06-05 17:40:44 +03:00
committed by mobile promotions
parent 26487b82df
commit 80197d2c9d
11 changed files with 33 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
drivers/gpu/nvgpu/common/linux/dma.c
View File

@@ -21,6 +21,7 @@
#include <nvgpu/lock.h>
#include <nvgpu/bug.h>
#include <nvgpu/gmmu.h>
#include <nvgpu/enabled.h>
#include <nvgpu/linux/dma.h>
@@ -69,7 +70,7 @@ int nvgpu_dma_alloc(struct gk20a *g, size_t size, struct nvgpu_mem *mem)
int nvgpu_dma_alloc_flags(struct gk20a *g, unsigned long flags, size_t size,
struct nvgpu_mem *mem)
{
if (g->mm.vidmem_is_vidmem) {
if (!nvgpu_is_enabled(g, NVGPU_MM_UNIFIED_MEMORY)) {
/*
* Force the no-kernel-mapping flag on because we don't support
* the lack of it for vidmem - the user should not care when
@@ -251,7 +252,7 @@ int nvgpu_dma_alloc_map(struct vm_gk20a *vm, size_t size,
int nvgpu_dma_alloc_map_flags(struct vm_gk20a *vm, unsigned long flags,
size_t size, struct nvgpu_mem *mem)
{
if (vm->mm->vidmem_is_vidmem) {
if (!nvgpu_is_enabled(gk20a_from_vm(vm), NVGPU_MM_UNIFIED_MEMORY)) {
/*
* Force the no-kernel-mapping flag on because we don't support
* the lack of it for vidmem - the user should not care when