Skip to contentSkip to conversation

firewire-ohci: work around oversized DMA reads on JMicron controllers

PATCH Applied Hector Martin 3 participants 5 messages Nov 3, 2017 – Jan 13, 2018 Firewire

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 firewire-ohci: work around oversized DMA reads on JMicron controllers

At least some JMicron controllers issue buggy oversized DMA reads when fetching context descriptors, always fetching 0x20 bytes at once for descriptors which are only 0x10 bytes long. This is often harmless, but can cause page faults on modern systems with IOMMUs:

DMAR: [DMA Read] Request device [05:00.0] fault addr fff56000 [fault reason 06] PTE Read access is not set firewire_ohci 0000:05:00.0: DMA context IT0 has stopped, error code: evt_descriptor_read

This works around the problem by always leaving 0x10 padding bytes at the end of descriptor buffer pages, which should be harmless to do unconditionally for controllers in case others have the same behavior.

---
 drivers/firewire/ohci.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/firewire/ohci.c +7 -1
index 8bf89267dc25..d731b413cb2c 100644
@@ -1130,7 +1130,13 @@ static int context_add_buffer(struct context *ctx)
 		return -ENOMEM;
 
 	offset = (void *)&desc->buffer - (void *)desc;
-	desc->buffer_size = PAGE_SIZE - offset;
+	/*
+	 * Some controllers, like JMicron ones, always issue 0x20-byte DMA reads
+	 * for descriptors, even 0x10-byte ones. This can cause page faults when
+	 * an IOMMU is in use and the oversized read crosses a page boundary.
+	 * Work around this by always leaving at least 0x10 bytes of padding.
+	 */
+	desc->buffer_size = PAGE_SIZE - offset - 0x10;
 	desc->buffer_bus = bus_addr + offset;
 	desc->used = 0;
 
Signature · 1 line
2.14.3
Reviewed-by on v1 1/1
#1 lore raw
Quoted 5 lines · Hector Martin

At least some JMicron controllers issue buggy oversized DMA reads when fetching context descriptors, always fetching 0x20 bytes at once for descriptors which are only 0x10 bytes long. This is often harmless, but can cause page faults on modern systems with IOMMUs:

Quoted 8 lines

DMAR: [DMA Read] Request device [05:00.0] fault addr fff56000 [fault reason 06] PTE Read access is not set firewire_ohci 0000:05:00.0: DMA context IT0 has stopped, error code: evt_descriptor_read

This works around the problem by always leaving 0x10 padding bytes at the end of descriptor buffer pages, which should be harmless to do unconditionally for controllers in case others have the same behavior.

Quoted 23 lines
---
 drivers/firewire/ohci.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/firewire/ohci.c +7 -1
index 8bf89267dc25..d731b413cb2c 100644
@@ -1130,7 +1130,13 @@ static int context_add_buffer(struct context *ctx)
 		return -ENOMEM;
 
 	offset = (void *)&desc->buffer - (void *)desc;
-	desc->buffer_size = PAGE_SIZE - offset;
+	/*
+	 * Some controllers, like JMicron ones, always issue 0x20-byte DMA reads
+	 * for descriptors, even 0x10-byte ones. This can cause page faults when
+	 * an IOMMU is in use and the oversized read crosses a page boundary.
+	 * Work around this by always leaving at least 0x10 bytes of padding.
+	 */
+	desc->buffer_size = PAGE_SIZE - offset - 0x10;
 	desc->buffer_bus = bus_addr + offset;
 	desc->used = 0;
 
#2 lore raw
Quoted 9 lines · On Nov 03 Clemens Ladisch wrote:
Quoted 8 lines · Hector Martin

At least some JMicron controllers issue buggy oversized DMA reads when fetching context descriptors, always fetching 0x20 bytes at once for descriptors which are only 0x10 bytes long. This is often harmless, but can cause page faults on modern systems with IOMMUs:

DMAR: [DMA Read] Request device [05:00.0] fault addr fff56000 [fault reason 06] PTE Read access is not set firewire_ohci 0000:05:00.0: DMA context IT0 has stopped, error code: evt_descriptor_read

Quoted 7 lines
Quoted 5 lines

This works around the problem by always leaving 0x10 padding bytes at the end of descriptor buffer pages, which should be harmless to do unconditionally for controllers in case others have the same behavior.

Thanks Hector for the troubleshooting and for the patch. Thanks Clemens for the review.

It's been a while since I last reviewed and tested kernel patches, and also my main FireWire equipped PC is currently tied up in work for which reboots aren't desirable. But I am updating a long unused secondary FireWire'd PC right now and give the patch some testing this week. (This one even has a JMicron controller, but not an IOMMU.)

Quoted 23 lines
Quoted 23 lines
---
 drivers/firewire/ohci.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/firewire/ohci.c +7 -1
index 8bf89267dc25..d731b413cb2c 100644
@@ -1130,7 +1130,13 @@ static int context_add_buffer(struct context *ctx)
 		return -ENOMEM;
 
 	offset = (void *)&desc->buffer - (void *)desc;
-	desc->buffer_size = PAGE_SIZE - offset;
+	/*
+	 * Some controllers, like JMicron ones, always issue 0x20-byte DMA reads
+	 * for descriptors, even 0x10-byte ones. This can cause page faults when
+	 * an IOMMU is in use and the oversized read crosses a page boundary.
+	 * Work around this by always leaving at least 0x10 bytes of padding.
+	 */
+	desc->buffer_size = PAGE_SIZE - offset - 0x10;
 	desc->buffer_bus = bus_addr + offset;
 	desc->used = 0;
 
Signature · 3 lines
Stefan Richter
-======----= =-== -==--
http://arcgraph.de/sr/
#3 lore raw
Quoted 9 lines · Stefan Richter

Thanks Hector for the troubleshooting and for the patch. Thanks Clemens for the review.

It's been a while since I last reviewed and tested kernel patches, and also my main FireWire equipped PC is currently tied up in work for which reboots aren't desirable. But I am updating a long unused secondary FireWire'd PC right now and give the patch some testing this week. (This one even has a JMicron controller, but not an IOMMU.)

Hi Stefan, did you ever get around to testing the patch? It doesn't seem to have made it into any trees or percolated up to mainline yet.

Signature · 2 lines
Hector Martin "marcan" ([email protected])
Public Key: https://mrcn.st/pub
#4 lore raw
Quoted 4 lines · On Jan 11 Hector Martin 'marcan' wrote:
Quoted 3 lines · Stefan Richter

Thanks Hector for the troubleshooting and for the patch. Thanks Clemens for the review.

Quoted 8 lines
Quoted 6 lines

It's been a while since I last reviewed and tested kernel patches, and also my main FireWire equipped PC is currently tied up in work for which reboots aren't desirable. But I am updating a long unused secondary FireWire'd PC right now and give the patch some testing this week. (This one even has a JMicron controller, but not an IOMMU.)

Hi Stefan, did you ever get around to testing the patch? It doesn't seem to have made it into any trees or percolated up to mainline yet.

Pushed to linux1394.git:for-next now. Sorry that I took so long. Thanks once more for your effort.

(Getting the test computer up to date was a PITA; no fun when there is almost no spare time...)

Signature · 3 lines
Stefan Richter
-======---=- ---= -==-=
http://arcgraph.de/sr/