Re: [PATCH v2 6/6] dma-debug: Ensure mappings are created and released with matching attributes

On Fri, May 01, 2026 at 09:35:10AM +0300, Leon Romanovsky wrote:
>From: Leon Romanovsky <[email protected]>
>
>The DMA API expects that callers use the same attributes when mapping
>and unmapping. Add tracking to verify this and catch mismatches.
>
>Signed-off-by: Leon Romanovsky <[email protected]>
>---
> kernel/dma/debug.c | 23 +++++++++++++++++++++++
> 1 file changed, 23 insertions(+)
>
>diff --git a/kernel/dma/debug.c b/kernel/dma/debug.c
>index 3dfed51c3d9aa..c38efc1ac8d6c 100644
>--- a/kernel/dma/debug.c
>+++ b/kernel/dma/debug.c
>@@ -1074,6 +1074,29 @@ static void check_unmap(struct dma_debug_entry *ref)
> 			   type2name[entry->type]);
> 	}
>
>+	/*
>+	 * This may be no bug in reality - but DMA API still expects
>+	 * that entry is unmapped with same attributes as it was mapped.
>+	 *
>+	 * DMA_ATTR_UNMAP_VALID lists the attributes that must be identical
>+	 * between map and unmap. Any attribute outside this set (e.g.
>+	 * DMA_ATTR_NO_WARN, DMA_ATTR_SKIP_CPU_SYNC) is allowed to differ.
>+	 */
>+#define DMA_ATTR_UNMAP_VALID                                               \
>+	(DMA_ATTR_NO_KERNEL_MAPPING | DMA_ATTR_FORCE_CONTIGUOUS |          \
>+	 DMA_ATTR_MMIO | DMA_ATTR_REQUIRE_COHERENT | DMA_ATTR_PRIVILEGED | \
>+	 DMA_ATTR_CC_SHARED)
>+	if ((ref->attrs & DMA_ATTR_UNMAP_VALID) !=
>+	    (entry->attrs & DMA_ATTR_UNMAP_VALID)) {
>+		err_printk(ref->dev, entry,
>+			   "device driver frees "
>+			   "DMA memory with different attributes "
>+			   "[device address=0x%016llx] [size=%llu bytes] "
>+			   "[mapped with 0x%lx] [unmapped with 0x%lx]\n",
>+			   ref->dev_addr, ref->size, entry->attrs, ref->attrs);
>+	}
>+#undef DMA_ATTR_UNMAP_VALID
>+
> 	hash_bucket_del(entry);
> 	put_hash_bucket(bucket, flags);
>
>
>-- 
>2.53.0
>
>

Reviewed-by: Samiullah Khawaja <[email protected]>