[PATCH v3 2/3] drm/amdgpu: Reset the clear flag in buddy during resume

[Christian König]

On 08.07.25 08:54, Arunpravin Paneer Selvam wrote:
> - Added a handler in DRM buddy manager to reset the cleared
>   flag for the blocks in the freelist.
> 
> - This is necessary because, upon resuming, the VRAM becomes
>   cluttered with BIOS data, yet the VRAM backend manager
>   believes that everything has been cleared.
> 
> v2:
>   - Add lock before accessing drm_buddy_clear_reset_blocks()(Matthew Auld)
>   - Force merge the two dirty blocks.(Matthew Auld)
>   - Add a new unit test case for this issue.(Matthew Auld)
>   - Having this function being able to flip the state either way would be
>     good. (Matthew Brost)
> 
> v3(Matthew Auld):
>   - Do merge step first to avoid the use of extra reset flag.
> 
> Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam at amd.com>
> Suggested-by: Christian König <christian.koenig at amd.com>
> Cc: stable at vger.kernel.org
> Fixes: a68c7eaa7a8f ("drm/amdgpu: Enable clear page functionality")

Acked-by: Christian König <christian.koenig at amd.com>

> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3812

I'm not 100% sure if that really fully closes this issue. Keep an eye open if the warning we added in patch #1 ever triggers.

Regards,
Christian.

> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c   |  2 +
>  drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h      |  1 +
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c | 17 ++++++++
>  drivers/gpu/drm/drm_buddy.c                  | 43 ++++++++++++++++++++
>  include/drm/drm_buddy.h                      |  2 +
>  5 files changed, 65 insertions(+)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> index a59f194e3360..b89e46f29b51 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
> @@ -5193,6 +5193,8 @@ int amdgpu_device_resume(struct drm_device *dev, bool notify_clients)
>  		dev->dev->power.disable_depth--;
>  #endif
>  	}
> +
> +	amdgpu_vram_mgr_clear_reset_blocks(adev);
>  	adev->in_suspend = false;
>  
>  	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D0))
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
> index 208b7d1d8a27..450e4bf093b7 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
> @@ -154,6 +154,7 @@ int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr,
>  				  uint64_t start, uint64_t size);
>  int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr,
>  				      uint64_t start);
> +void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev);
>  
>  bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
>  			    struct ttm_resource *res);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
> index abdc52b0895a..07c936e90d8e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
> @@ -782,6 +782,23 @@ uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr)
>  	return atomic64_read(&mgr->vis_usage);
>  }
>  
> +/**
> + * amdgpu_vram_mgr_clear_reset_blocks - reset clear blocks
> + *
> + * @adev: amdgpu device pointer
> + *
> + * Reset the cleared drm buddy blocks.
> + */
> +void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev)
> +{
> +	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
> +	struct drm_buddy *mm = &mgr->mm;
> +
> +	mutex_lock(&mgr->lock);
> +	drm_buddy_reset_clear(mm, false);
> +	mutex_unlock(&mgr->lock);
> +}
> +
>  /**
>   * amdgpu_vram_mgr_intersects - test each drm buddy block for intersection
>   *
> diff --git a/drivers/gpu/drm/drm_buddy.c b/drivers/gpu/drm/drm_buddy.c
> index a1e652b7631d..a94061f373de 100644
> --- a/drivers/gpu/drm/drm_buddy.c
> +++ b/drivers/gpu/drm/drm_buddy.c
> @@ -405,6 +405,49 @@ drm_get_buddy(struct drm_buddy_block *block)
>  }
>  EXPORT_SYMBOL(drm_get_buddy);
>  
> +/**
> + * drm_buddy_reset_clear - reset blocks clear state
> + *
> + * @mm: DRM buddy manager
> + * @is_clear: blocks clear state
> + *
> + * Reset the clear state based on @is_clear value for each block
> + * in the freelist.
> + */
> +void drm_buddy_reset_clear(struct drm_buddy *mm, bool is_clear)
> +{
> +	u64 root_size, size, start;
> +	unsigned int order;
> +	int i;
> +
> +	size = mm->size;
> +	for (i = 0; i < mm->n_roots; ++i) {
> +		order = ilog2(size) - ilog2(mm->chunk_size);
> +		start = drm_buddy_block_offset(mm->roots[i]);
> +		__force_merge(mm, start, start + size, order);
> +
> +		root_size = mm->chunk_size << order;
> +		size -= root_size;
> +	}
> +
> +	for (i = 0; i <= mm->max_order; ++i) {
> +		struct drm_buddy_block *block;
> +
> +		list_for_each_entry_reverse(block, &mm->free_list[i], link) {
> +			if (is_clear != drm_buddy_block_is_clear(block)) {
> +				if (is_clear) {
> +					mark_cleared(block);
> +					mm->clear_avail += drm_buddy_block_size(mm, block);
> +				} else {
> +					clear_reset(block);
> +					mm->clear_avail -= drm_buddy_block_size(mm, block);
> +				}
> +			}
> +		}
> +	}
> +}
> +EXPORT_SYMBOL(drm_buddy_reset_clear);
> +
>  /**
>   * drm_buddy_free_block - free a block
>   *
> diff --git a/include/drm/drm_buddy.h b/include/drm/drm_buddy.h
> index 9689a7c5dd36..513837632b7d 100644
> --- a/include/drm/drm_buddy.h
> +++ b/include/drm/drm_buddy.h
> @@ -160,6 +160,8 @@ int drm_buddy_block_trim(struct drm_buddy *mm,
>  			 u64 new_size,
>  			 struct list_head *blocks);
>  
> +void drm_buddy_reset_clear(struct drm_buddy *mm, bool is_clear);
> +
>  void drm_buddy_free_block(struct drm_buddy *mm, struct drm_buddy_block *block);
>  
>  void drm_buddy_free_list(struct drm_buddy *mm,