Skip to contentSkip to conversation

iommu/amd: Fix domain flush size when syncing iotlb

PATCH Superseded Jon Pan-Doh 5 participants 11 messages Apr 26, 2023 – May 22, 2023

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

PATCH 1/1 iommu/amd: Fix domain flush size when syncing iotlb

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")
---
 drivers/iommu/amd/iommu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/amd/iommu.c +1 -1
index 5a505ba5467e..da45b1ab042d 100644
@@ -2378,7 +2378,7 @@ static void amd_iommu_iotlb_sync(struct iommu_domain *domain,
 	unsigned long flags;
 
 	spin_lock_irqsave(&dom->lock, flags);
-	domain_flush_pages(dom, gather->start, gather->end - gather->start, 1);
+	domain_flush_pages(dom, gather->start, gather->end - gather->start + 1, 1);
 	amd_iommu_domain_flush_complete(dom);
 	spin_unlock_irqrestore(&dom->lock, flags);
 }
Signature · 1 line
2.40.0.634.g4ca3ef3211-goog
Acked-by on v1 1/1
#1 lore raw
Quoted 11 lines · Jon Pan-Doh

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

Quoted 8 lines

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")

Sorry for that.

#2 lore raw
Quoted 24 lines · Nadav Amit
Quoted 12 lines · Jon Pan-Doh

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

Quoted 8 lines

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")
Quoted 3 lines

Sorry for that.

P.S. I think it is stable-worthy.

maintainer Reviewed-by on v1 1/1
#3 lore raw
In reply to Jon Pan-Doh
Quoted 7 lines · Jon Pan-Doh

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

Quoted 7 lines

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")

Thanks!

-Vasant

Quoted 17 lines
---
 drivers/iommu/amd/iommu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/amd/iommu.c +1 -1
index 5a505ba5467e..da45b1ab042d 100644
@@ -2378,7 +2378,7 @@ static void amd_iommu_iotlb_sync(struct iommu_domain *domain,
 	unsigned long flags;
 
 	spin_lock_irqsave(&dom->lock, flags);
-	domain_flush_pages(dom, gather->start, gather->end - gather->start, 1);
+	domain_flush_pages(dom, gather->start, gather->end - gather->start + 1, 1);
 	amd_iommu_domain_flush_complete(dom);
 	spin_unlock_irqrestore(&dom->lock, flags);
 }
#4 lore raw
In reply to Jon Pan-Doh

Hi Joerg,

[Cc'ing [email protected] per Nadav's suggestion]

This bug fix seems to have gotten the necessary reviews (AMD and previous commit author). Is it eligible to be applied?

Thanks,
Jon

Quoted 35 lines · Jon Pan-Doh

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")
---
 drivers/iommu/amd/iommu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/iommu/amd/iommu.c +1 -2
index 5a505ba5467e..da45b1ab042d 100644
@@ -2378,7 +2378,7 @@ static void amd_iommu_iotlb_sync(struct iommu_domain *domain,
        unsigned long flags;
 
        spin_lock_irqsave(&dom->lock, flags);
-       domain_flush_pages(dom, gather->start, gather->end - gather->start, 1);
+       domain_flush_pages(dom, gather->start, gather->end - gather->start + 1, 1);
        amd_iommu_domain_flush_complete(dom);
        spin_unlock_irqrestore(&dom->lock, flags);
 }
--

2.40.0.634.g4ca3ef3211-goog

#7 lore raw
In reply to Jon Pan-Doh
Quoted 8 lines · Jon Pan-Doh

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

Quoted 7 lines

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")
#6 lore raw
In reply to Greg KH
Quoted 6 lines · Greg KH
Quoted 3 lines · Jon Pan-Doh

Hi Joerg,

[Cc'ing [email protected] per Nadav's suggestion]

Quoted 8 lines

<formletter>

This is not the correct way to submit patches for inclusion in the stable kernel tree. Please read:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
for how to do this properly.

</formletter>

Thanks for the link. Added the Cc tag as a reply to the patch. Apologies for top-posting previously.

#8 lore raw
In reply to Jon Pan-Doh
Quoted 17 lines · Jon Pan-Doh

!! External Email

Quoted 11 lines · Jon Pan-Doh

When running on an AMD vIOMMU, we observed multiple invalidations (of decreasing power of 2 aligned sizes) when unmapping a single page.

Domain flush takes gather bounds (end-start) as size param. However, gather->end is defined as the last inclusive address (start + size - 1). This leads to an off by 1 error.

With this patch, verified that 1 invalidation occurs when unmapping a single page.

Quoted 6 lines
Quoted 4 lines
Fixes a270be1b3fdf ("iommu/amd: Use only natural aligned flushes in a VM")

Assuming the patch was still not applied, I think that Greg KH would prefer that you send v2 that includes the “Cc: <[email protected]>” (notice the “<“ and “>" signs) and ideally the version in which I added the buggy code (sorry for that again) after a “#” sign.

I think it is version 5.15.

IOW: If the patch was still not applied by Joerg, please send v2 with:

Cc <[email protected] <mailto:[email protected]>> # 5.15.x

In the commit log.

#9 lore raw
Quoted 6 lines · Nadav Amit

IOW: If the patch was still not applied by Joerg, please send v2 with:

Cc <[email protected] <mailto:[email protected]>> # 5.15.x

In the commit log.

Thanks for the clarification/guidance. I've resent the patch for review with added tags (Acked-by, Reviewed-by). Will re-read the patch submission doc and try to avoid formatting issues in the future.