Re: [PATCH v3] drm/amdgpu: fix ftrace event amdgpu_bo_move always move on same heap

[Christian König]

Am 13.12.23 um 08:27 schrieb Wang, Beyond:


[AMD Official Use Only - General]


Issue: during evict or validate happened on amdgpu_bo, the 'from' and

'to' is always same in ftrace event of amdgpu_bo_move

where calling the 'trace_amdgpu_bo_move', the comment says move_notify

is called before move happens, but actually it is called after move

happens, here the new_mem is same as bo->resource

Fix: move trace_amdgpu_bo_move from move_notify to amdgpu_bo_move

Signed-off-by: Wang, Beyond wang.bey...@amd.com



Yeah, that makes much more sense. Reviewed-by: Christian König 



Regards,
Christian.


---

drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 13 +

drivers/gpu/drm/amd/amdgpu/amdgpu_object.h |  4 +---

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c |  5 +++--

3 files changed, 5 insertions(+), 17 deletions(-)

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


index 7416799750c1..1870775d582c 100644

--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c

+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c

@@ -1282,19 +1282,15 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo 
*bo, void *buffer,


  * amdgpu_bo_move_notify - notification about a memory move

  * @bo: pointer to a buffer object

  * @evict: if this move is evicting the buffer from the graphics 
address space


- * @new_mem: new information of the bufer object

  *

  * Marks the corresponding _bo buffer object as invalid, also 
performs


  * bookkeeping.

  * TTM driver callback which is called when ttm moves a buffer.

  */

-void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,

-  bool evict,

-  struct ttm_resource *new_mem)

+void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict)

{

    struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);

    struct amdgpu_bo *abo;

-   struct ttm_resource *old_mem = bo->resource;

    if (!amdgpu_bo_is_amdgpu_bo(bo))

    return;

@@ -1313,13 +1309,6 @@ void amdgpu_bo_move_notify(struct 
ttm_buffer_object *bo,


    /* remember the eviction */

    if (evict)

atomic64_inc(>num_evictions);

-

-   /* update statistics */

-   if (!new_mem)

-   return;

-

-   /* move_notify is called before move happens */

-   trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type);

}

void amdgpu_bo_get_memory(struct amdgpu_bo *bo,

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h


index 876acde6b10a..dee2c577427e 100644

--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h

+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h

@@ -360,9 +360,7 @@ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, 
void *metadata,


int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer,

   size_t buffer_size, uint32_t *metadata_size,

   uint64_t *flags);

-void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,

-  bool evict,

-  struct ttm_resource *new_mem);

+void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict);

void amdgpu_bo_release_notify(struct ttm_buffer_object *bo);

vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo);

void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence,

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


index 41ed6a3e5a06..f0fffbf2bdd5 100644

--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -576,10 +576,11 @@ static int amdgpu_bo_move(struct 
ttm_buffer_object *bo, bool evict,


    return r;

    }

+   trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type);

out:

    /* update statistics */

    atomic64_add(bo->base.size, >num_bytes_moved);

-   amdgpu_bo_move_notify(bo, evict, new_mem);

+   amdgpu_bo_move_notify(bo, evict);

    return 0;

}

@@ -1852,7 +1853,7 @@ static int amdgpu_ttm_access_memory(struct 
ttm_buffer_object *bo,


static void

amdgpu_bo_delete_mem_notify(struct ttm_buffer_object *bo)

{

-   amdgpu_bo_move_notify(bo, false, NULL);

+   amdgpu_bo_move_notify(bo, false);

}

static struct ttm_device_funcs amdgpu_bo_driver = {

--

2.34.1

RE: [PATCH v3] drm/amdgpu: fix ftrace event amdgpu_bo_move always move on same heap

[Wang, Beyond]

[AMD Official Use Only - General]



Issue: during evict or validate happened on amdgpu_bo, the 'from' and
'to' is always same in ftrace event of amdgpu_bo_move

where calling the 'trace_amdgpu_bo_move', the comment says move_notify
is called before move happens, but actually it is called after move
happens, here the new_mem is same as bo->resource

Fix: move trace_amdgpu_bo_move from move_notify to amdgpu_bo_move

Signed-off-by: Wang, Beyond wang.bey...@amd.com
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 13 +
drivers/gpu/drm/amd/amdgpu/amdgpu_object.h |  4 +---
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c|  5 +++--
3 files changed, 5 insertions(+), 17 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 7416799750c1..1870775d582c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1282,19 +1282,15 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void 
*buffer,
  * amdgpu_bo_move_notify - notification about a memory move
  * @bo: pointer to a buffer object
  * @evict: if this move is evicting the buffer from the graphics address space
- * @new_mem: new information of the bufer object
  *
  * Marks the corresponding _bo buffer object as invalid, also performs
  * bookkeeping.
  * TTM driver callback which is called when ttm moves a buffer.
  */
-void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
-  bool evict,
-  struct ttm_resource *new_mem)
+void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict)
{
struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
struct amdgpu_bo *abo;
-   struct ttm_resource *old_mem = bo->resource;

if (!amdgpu_bo_is_amdgpu_bo(bo))
return;
@@ -1313,13 +1309,6 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
/* remember the eviction */
if (evict)
atomic64_inc(>num_evictions);
-
-   /* update statistics */
-   if (!new_mem)
-   return;
-
-   /* move_notify is called before move happens */
-   trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type);
}

void amdgpu_bo_get_memory(struct amdgpu_bo *bo,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index 876acde6b10a..dee2c577427e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
@@ -360,9 +360,7 @@ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void 
*metadata,
int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer,
   size_t buffer_size, uint32_t *metadata_size,
   uint64_t *flags);
-void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
-  bool evict,
-  struct ttm_resource *new_mem);
+void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict);
void amdgpu_bo_release_notify(struct ttm_buffer_object *bo);
vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo);
void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c 
b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 41ed6a3e5a06..f0fffbf2bdd5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -576,10 +576,11 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, 
bool evict,
return r;
}

+   trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type);
out:
/* update statistics */
atomic64_add(bo->base.size, >num_bytes_moved);
-   amdgpu_bo_move_notify(bo, evict, new_mem);
+   amdgpu_bo_move_notify(bo, evict);
return 0;
}

@@ -1852,7 +1853,7 @@ static int amdgpu_ttm_access_memory(struct 
ttm_buffer_object *bo,
static void
amdgpu_bo_delete_mem_notify(struct ttm_buffer_object *bo)
{
-   amdgpu_bo_move_notify(bo, false, NULL);
+   amdgpu_bo_move_notify(bo, false);
}

static struct ttm_device_funcs amdgpu_bo_driver = {
--
2.34.1