Re: [RFC PATCH 2/3] drm/amdgpu/ttm: replace duplicate code with exiting function

2021-07-15 Thread Christian König 

Am 15.07.21 um 11:29 schrieb Lazar, Lijo:



On 7/15/2021 2:56 PM, Christian König wrote:

Am 15.07.21 um 10:24 schrieb Kevin Wang:

using exiting function to replace duplicate code blocks in
amdgpu_ttm_vram_write().


NAK, this should intentionally only use the MM path and not the aper 
path.




What about platform configs which don't support HDP?


Good question, I have no idea.

As far as I know the write interface here intentionally uses the 
MM_INDEX/MM_DATA to bypass the HDP cache.


Regards,
Christian.



Thanks,
Lijo


But you could use amdgpu_device_mm_access() here now.

Christian.



Signed-off-by: Kevin Wang 
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 7 +--
  1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

index 2aa2eb5de37a..2c98e4345ad3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2207,7 +2207,6 @@ static ssize_t amdgpu_ttm_vram_write(struct 
file *f, const char __user *buf,

  return -ENXIO;
  while (size) {
-    unsigned long flags;
  uint32_t value;
  if (*pos >= adev->gmc.mc_vram_size)
@@ -2217,11 +2216,7 @@ static ssize_t amdgpu_ttm_vram_write(struct 
file *f, const char __user *buf,

  if (r)
  return r;
-    spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-    WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x8000);
-    WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
-    WREG32_NO_KIQ(mmMM_DATA, value);
-    spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
+    amdgpu_device_vram_access(adev, *pos, &value, 4, true);
  result += 4;
  buf += 4;




___
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 2/3] drm/amdgpu/ttm: replace duplicate code with exiting function

2021-07-15 Thread Lazar, Lijo 



On 7/15/2021 2:56 PM, Christian König wrote:

Am 15.07.21 um 10:24 schrieb Kevin Wang:

using exiting function to replace duplicate code blocks in
amdgpu_ttm_vram_write().


NAK, this should intentionally only use the MM path and not the aper path.



What about platform configs which don't support HDP?

Thanks,
Lijo


But you could use amdgpu_device_mm_access() here now.

Christian.



Signed-off-by: Kevin Wang 
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 7 +--
  1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

index 2aa2eb5de37a..2c98e4345ad3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2207,7 +2207,6 @@ static ssize_t amdgpu_ttm_vram_write(struct file 
*f, const char __user *buf,

  return -ENXIO;
  while (size) {
-    unsigned long flags;
  uint32_t value;
  if (*pos >= adev->gmc.mc_vram_size)
@@ -2217,11 +2216,7 @@ static ssize_t amdgpu_ttm_vram_write(struct 
file *f, const char __user *buf,

  if (r)
  return r;
-    spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-    WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x8000);
-    WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
-    WREG32_NO_KIQ(mmMM_DATA, value);
-    spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
+    amdgpu_device_vram_access(adev, *pos, &value, 4, true);
  result += 4;
  buf += 4;



___
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 2/3] drm/amdgpu/ttm: replace duplicate code with exiting function

2021-07-15 Thread Christian König 

Am 15.07.21 um 10:24 schrieb Kevin Wang:

using exiting function to replace duplicate code blocks in
amdgpu_ttm_vram_write().


NAK, this should intentionally only use the MM path and not the aper path.

But you could use amdgpu_device_mm_access() here now.

Christian.



Signed-off-by: Kevin Wang 
---
  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 7 +--
  1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 2aa2eb5de37a..2c98e4345ad3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2207,7 +2207,6 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, 
const char __user *buf,
return -ENXIO;
  
  	while (size) {

-   unsigned long flags;
uint32_t value;
  
  		if (*pos >= adev->gmc.mc_vram_size)

@@ -2217,11 +2216,7 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, 
const char __user *buf,
if (r)
return r;
  
-		spin_lock_irqsave(&adev->mmio_idx_lock, flags);

-   WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x8000);
-   WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
-   WREG32_NO_KIQ(mmMM_DATA, value);
-   spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
+   amdgpu_device_vram_access(adev, *pos, &value, 4, true);
  
  		result += 4;

buf += 4;


___
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

[RFC PATCH 2/3] drm/amdgpu/ttm: replace duplicate code with exiting function

2021-07-15 Thread Kevin Wang 
using exiting function to replace duplicate code blocks in
amdgpu_ttm_vram_write().

Signed-off-by: Kevin Wang 
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 7 +--
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 2aa2eb5de37a..2c98e4345ad3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -2207,7 +2207,6 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, 
const char __user *buf,
return -ENXIO;
 
while (size) {
-   unsigned long flags;
uint32_t value;
 
if (*pos >= adev->gmc.mc_vram_size)
@@ -2217,11 +2216,7 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, 
const char __user *buf,
if (r)
return r;
 
-   spin_lock_irqsave(&adev->mmio_idx_lock, flags);
-   WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x8000);
-   WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31);
-   WREG32_NO_KIQ(mmMM_DATA, value);
-   spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
+   amdgpu_device_vram_access(adev, *pos, &value, 4, true);
 
result += 4;
buf += 4;
-- 
2.25.1

___
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx