Browse Source

[PATCH] ppc64: iommu vmerge fix

This fixes a bug in the PPC64 iommu vmerge code which results in the
potential for iommu_unmap_sg to go off unmapping more than it should.

This was found on a test system which resulted in PCI bus errors due to
PCI memory being unmapped while DMAs were still in progress.

Signed-off-by: Brian King <brking@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Brian King 20 years ago
parent
commit
ac9af7cba9
1 changed files with 4 additions and 3 deletions
  1. 4 3
      arch/ppc64/kernel/iommu.c

+ 4 - 3
arch/ppc64/kernel/iommu.c

@@ -242,7 +242,7 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
 	dma_addr_t dma_next = 0, dma_addr;
 	unsigned long flags;
 	struct scatterlist *s, *outs, *segstart;
-	int outcount;
+	int outcount, incount;
 	unsigned long handle;
 
 	BUG_ON(direction == DMA_NONE);
@@ -252,6 +252,7 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
 
 	outs = s = segstart = &sglist[0];
 	outcount = 1;
+	incount = nelems;
 	handle = 0;
 
 	/* Init first segment length for backout at failure */
@@ -338,10 +339,10 @@ int iommu_map_sg(struct device *dev, struct iommu_table *tbl,
 
 	DBG("mapped %d elements:\n", outcount);
 
-	/* For the sake of iommu_free_sg, we clear out the length in the
+	/* For the sake of iommu_unmap_sg, we clear out the length in the
 	 * next entry of the sglist if we didn't fill the list completely
 	 */
-	if (outcount < nelems) {
+	if (outcount < incount) {
 		outs++;
 		outs->dma_address = DMA_ERROR_CODE;
 		outs->dma_length = 0;