[PATCH v2 1/3] drm/amdkfd: ratelimited svm debug messages

Felix Kuehling felix.kuehling at amd.com
Tue Oct 12 22:49:35 UTC 2021


Am 2021-10-12 um 9:55 a.m. schrieb Philip Yang:
> No function change, use pr_debug_ratelimited to avoid per page debug
> message overflowing dmesg buf and console log.
>
> use dev_err to show error message from unexpected situation, to provide
> clue to help debug without enabling dynamic debug log. Define dev_fmt to
> output function name in error message.
>
> Signed-off-by: Philip Yang <Philip.Yang at amd.com>

Reviewed-by: Felix Kuehling <Felix.Kuehling at amd.com>


> ---
>  drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 34 +++++++++++++-----------
>  drivers/gpu/drm/amd/amdkfd/kfd_svm.c     | 17 +++++++-----
>  2 files changed, 30 insertions(+), 21 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> index f53e17a94ad8..b05c0579d0b9 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> @@ -20,7 +20,6 @@
>   * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
>   * OTHER DEALINGS IN THE SOFTWARE.
>   */
> -
>  #include <linux/types.h>
>  #include <linux/hmm.h>
>  #include <linux/dma-direction.h>
> @@ -34,6 +33,11 @@
>  #include "kfd_svm.h"
>  #include "kfd_migrate.h"
>  
> +#ifdef dev_fmt
> +#undef dev_fmt
> +#endif
> +#define dev_fmt(fmt) "kfd_migrate: %s: " fmt, __func__
> +
>  static uint64_t
>  svm_migrate_direct_mapping_addr(struct amdgpu_device *adev, uint64_t addr)
>  {
> @@ -151,14 +155,14 @@ svm_migrate_copy_memory_gart(struct amdgpu_device *adev, dma_addr_t *sys,
>  			gart_d = svm_migrate_direct_mapping_addr(adev, *vram);
>  		}
>  		if (r) {
> -			pr_debug("failed %d to create gart mapping\n", r);
> +			dev_err(adev->dev, "fail %d create gart mapping\n", r);
>  			goto out_unlock;
>  		}
>  
>  		r = amdgpu_copy_buffer(ring, gart_s, gart_d, size * PAGE_SIZE,
>  				       NULL, &next, false, true, false);
>  		if (r) {
> -			pr_debug("failed %d to copy memory\n", r);
> +			dev_err(adev->dev, "fail %d to copy memory\n", r);
>  			goto out_unlock;
>  		}
>  
> @@ -285,7 +289,7 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
>  
>  	r = svm_range_vram_node_new(adev, prange, true);
>  	if (r) {
> -		pr_debug("failed %d get 0x%llx pages from vram\n", r, npages);
> +		dev_err(adev->dev, "fail %d to alloc vram\n", r);
>  		goto out;
>  	}
>  
> @@ -305,7 +309,7 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
>  					      DMA_TO_DEVICE);
>  			r = dma_mapping_error(dev, src[i]);
>  			if (r) {
> -				pr_debug("failed %d dma_map_page\n", r);
> +				dev_err(adev->dev, "fail %d dma_map_page\n", r);
>  				goto out_free_vram_pages;
>  			}
>  		} else {
> @@ -325,8 +329,8 @@ svm_migrate_copy_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
>  			continue;
>  		}
>  
> -		pr_debug("dma mapping src to 0x%llx, page_to_pfn 0x%lx\n",
> -			 src[i] >> PAGE_SHIFT, page_to_pfn(spage));
> +		pr_debug_ratelimited("dma mapping src to 0x%llx, pfn 0x%lx\n",
> +				     src[i] >> PAGE_SHIFT, page_to_pfn(spage));
>  
>  		if (j >= (cursor.size >> PAGE_SHIFT) - 1 && i < npages - 1) {
>  			r = svm_migrate_copy_memory_gart(adev, src + i - j,
> @@ -405,8 +409,8 @@ svm_migrate_vma_to_vram(struct amdgpu_device *adev, struct svm_range *prange,
>  
>  	r = migrate_vma_setup(&migrate);
>  	if (r) {
> -		pr_debug("failed %d prepare migrate svms 0x%p [0x%lx 0x%lx]\n",
> -			 r, prange->svms, prange->start, prange->last);
> +		dev_err(adev->dev, "vma setup fail %d range [0x%lx 0x%lx]\n", r,
> +			prange->start, prange->last);
>  		goto out_free;
>  	}
>  	if (migrate.cpages != npages) {
> @@ -506,7 +510,7 @@ static void svm_migrate_page_free(struct page *page)
>  	struct svm_range_bo *svm_bo = page->zone_device_data;
>  
>  	if (svm_bo) {
> -		pr_debug("svm_bo ref left: %d\n", kref_read(&svm_bo->kref));
> +		pr_debug_ratelimited("ref: %d\n", kref_read(&svm_bo->kref));
>  		svm_range_bo_unref(svm_bo);
>  	}
>  }
> @@ -572,12 +576,12 @@ svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
>  		dst[i] = dma_map_page(dev, dpage, 0, PAGE_SIZE, DMA_FROM_DEVICE);
>  		r = dma_mapping_error(dev, dst[i]);
>  		if (r) {
> -			pr_debug("failed %d dma_map_page\n", r);
> +			dev_err(adev->dev, "fail %d dma_map_page\n", r);
>  			goto out_oom;
>  		}
>  
> -		pr_debug("dma mapping dst to 0x%llx, page_to_pfn 0x%lx\n",
> -			      dst[i] >> PAGE_SHIFT, page_to_pfn(dpage));
> +		pr_debug_ratelimited("dma mapping dst to 0x%llx, pfn 0x%lx\n",
> +				     dst[i] >> PAGE_SHIFT, page_to_pfn(dpage));
>  
>  		migrate->dst[i] = migrate_pfn(page_to_pfn(dpage));
>  		migrate->dst[i] |= MIGRATE_PFN_LOCKED;
> @@ -631,8 +635,8 @@ svm_migrate_vma_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
>  
>  	r = migrate_vma_setup(&migrate);
>  	if (r) {
> -		pr_debug("failed %d prepare migrate svms 0x%p [0x%lx 0x%lx]\n",
> -			 r, prange->svms, prange->start, prange->last);
> +		dev_err(adev->dev, "vma setup fail %d range [0x%lx 0x%lx]\n", r,
> +			prange->start, prange->last);
>  		goto out_free;
>  	}
>  
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> index 425d55deca10..49c92713c2ad 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
> @@ -33,6 +33,11 @@
>  #include "kfd_svm.h"
>  #include "kfd_migrate.h"
>  
> +#ifdef dev_fmt
> +#undef dev_fmt
> +#endif
> +#define dev_fmt(fmt) "kfd_svm: %s: " fmt, __func__
> +
>  #define AMDGPU_SVM_RANGE_RESTORE_DELAY_MS 1
>  
>  /* Long enough to ensure no retry fault comes after svm range is restored and
> @@ -158,17 +163,17 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
>  				   bo_adev->vm_manager.vram_base_offset -
>  				   bo_adev->kfd.dev->pgmap.range.start;
>  			addr[i] |= SVM_RANGE_VRAM_DOMAIN;
> -			pr_debug("vram address detected: 0x%llx\n", addr[i]);
> +			pr_debug_ratelimited("vram address: 0x%llx\n", addr[i]);
>  			continue;
>  		}
>  		addr[i] = dma_map_page(dev, page, 0, PAGE_SIZE, dir);
>  		r = dma_mapping_error(dev, addr[i]);
>  		if (r) {
> -			pr_debug("failed %d dma_map_page\n", r);
> +			dev_err(dev, "failed %d dma_map_page\n", r);
>  			return r;
>  		}
> -		pr_debug("dma mapping 0x%llx for page addr 0x%lx\n",
> -			 addr[i] >> PAGE_SHIFT, page_to_pfn(page));
> +		pr_debug_ratelimited("dma mapping 0x%llx for page addr 0x%lx\n",
> +				     addr[i] >> PAGE_SHIFT, page_to_pfn(page));
>  	}
>  	return 0;
>  }
> @@ -217,7 +222,7 @@ void svm_range_dma_unmap(struct device *dev, dma_addr_t *dma_addr,
>  	for (i = offset; i < offset + npages; i++) {
>  		if (!svm_is_valid_dma_mapping_addr(dev, dma_addr[i]))
>  			continue;
> -		pr_debug("dma unmapping 0x%llx\n", dma_addr[i] >> PAGE_SHIFT);
> +		pr_debug_ratelimited("unmap 0x%llx\n", dma_addr[i] >> PAGE_SHIFT);
>  		dma_unmap_page(dev, dma_addr[i], PAGE_SIZE, dir);
>  		dma_addr[i] = 0;
>  	}
> @@ -1454,7 +1459,7 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
>  		/* This should never happen. actual_loc gets set by
>  		 * svm_migrate_ram_to_vram after allocating a BO.
>  		 */
> -		WARN(1, "VRAM BO missing during validation\n");
> +		WARN_ONCE(1, "VRAM BO missing during validation\n");
>  		return -EINVAL;
>  	}
>  


More information about the amd-gfx mailing list