Skip to contentSkip to conversation

arm: davinci: fix edma dmaengine induced null pointer dereference on da830

PATCH New Matt Porter 2 participants 3 messages Mar 4, 2013 – Mar 5, 2013 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm: davinci: fix edma dmaengine induced null pointer dereference on da830

This adds additional error checking to the private edma api implementation to catch the case where the edma_alloc_slot() has an invalid controller parameter. The edma dmaengine wrapper driver relies on this condition being handled in order to avoid setting up a second edma dmaengine instance on DA830.

Verfied using a DA850 with the second EDMA controller platform instance removed to simulate a DA830 which only has a single EDMA controller.

Cc stable at kernel.org
---
 arch/arm/mach-davinci/dma.c |    2 ++
 1 file changed, 2 insertions(+)
arch/arm/mach-davinci/dma.c +2 -0
index a685e97..f9eb836 100644
@@ -747,6 +747,8 @@ int edma_alloc_slot(unsigned ctlr, int slot)
 		slot = EDMA_CHAN_SLOT(slot);
 
 	if (slot < 0) {
+		if (!edma_cc[ctlr])
+			return -EINVAL;
 		slot = edma_cc[ctlr]->num_channels;
 		for (;;) {
 			slot = find_next_zero_bit(edma_cc[ctlr]->edma_inuse,
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH 1/1 arm: davinci: fix edma dmaengine induced null pointer dereference on da830

Hi Matt,

I dropped the copy to stable folks since they dont want to be involved in patch reviews of this kind.

Quoted 6 lines · Matt Porter

This adds additional error checking to the private edma api implementation to catch the case where the edma_alloc_slot() has an invalid controller parameter. The edma dmaengine wrapper driver relies on this condition being handled in order to avoid setting up a second edma dmaengine instance on DA830.

Quoted 6 lines

Verfied using a DA850 with the second EDMA controller platform instance removed to simulate a DA830 which only has a single EDMA controller.

Cc stable at kernel.org

This should be stable at vger.kernel.org per Documentation/stable_kernel_rules.txt. Also, I think it is better to request the back port only from v3.7.x+ since the bug is important only after drivers/dma/edma.c was merged. So:

Cc stable at vger.kernel.org # v3.7.x+
Quoted 9 lines
---
 arch/arm/mach-davinci/dma.c |    2 ++
 1 file changed, 2 insertions(+)
arch/arm/mach-davinci/dma.c +0 -0
index a685e97..f9eb836 100644

arch/arm/mach-davinci/dma.c int edma_alloc_slot(unsigned ctlr, int slot) · line 751

Quoted 5 lines
arch/arm/mach-davinci/dma.c +2 -0
@@ -747,6 +747,8 @@ int edma_alloc_slot(unsigned ctlr, int slot)
 		slot = EDMA_CHAN_SLOT(slot);
 
 	if (slot < 0) {
+		if (!edma_cc[ctlr])
+			return -EINVAL;

Shouldn't such a check be done outside of the if() since there is an 'else if' later which also accesses edma_cc[ctlr]

	} else if (slot < edma_cc[ctlr]->num_channels ||
			slot >= edma_cc[ctlr]->num_slots) {
		return -EINVAL;

Thanks,
Sekhar

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw

PATCH 1/1 arm: davinci: fix edma dmaengine induced null pointer dereference on da830

Quoted 19 lines · Sekhar Nori

Hi Matt,

I dropped the copy to stable folks since they dont want to be involved in patch reviews of this kind.

Quoted 6 lines · Matt Porter

This adds additional error checking to the private edma api implementation to catch the case where the edma_alloc_slot() has an invalid controller parameter. The edma dmaengine wrapper driver relies on this condition being handled in order to avoid setting up a second edma dmaengine instance on DA830.

Quoted 6 lines

Verfied using a DA850 with the second EDMA controller platform instance removed to simulate a DA830 which only has a single EDMA controller.

Cc stable at kernel.org
Quoted 6 lines

This should be stable at vger.kernel.org per Documentation/stable_kernel_rules.txt. Also, I think it is better to request the back port only from v3.7.x+ since the bug is important only after drivers/dma/edma.c was merged. So:

Cc stable at vger.kernel.org # v3.7.x+

Agreed. Had the two year old email in my macros. Will update to reflect 3.7.x+ only.

Quoted 10 lines
Quoted 10 lines
 arch/arm/mach-davinci/dma.c |    2 ++
 1 file changed, 2 insertions(+)
arch/arm/mach-davinci/dma.c +0 -0
index a685e97..f9eb836 100644
@@ -747,6 +747,8 @@ int edma_alloc_slot(unsigned ctlr, int slot)
 		slot = EDMA_CHAN_SLOT(slot);
 
Quoted 6 lines

arch/arm/mach-davinci/dma.c int edma_alloc_slot(unsigned ctlr, int slot) · line 751

Quoted 3 lines
arch/arm/mach-davinci/dma.c +2 -0
 	if (slot < 0) {
+		if (!edma_cc[ctlr])
+			return -EINVAL;

Shouldn't such a check be done outside of the if() since there is an 'else if' later which also accesses edma_cc[ctlr]

Yeah, as it turns out I also separately added the correct version of this in the am33xx dmaengine series as it was an issue there. I will correct this in v2.

-Matt