[PATCH] iommu: Restore iommu_flush_iotlb_all()

[PATCH] iommu: Restore iommu_flush_iotlb_all()

[Joerg Roedel]

From: Joerg Roedel <jroedel@suse.de>

This patch restores the iommu_flush_iotlb_all() function.
Commit

	69e5a17511f6 ("iommu: Remove useless flush from iommu_create_device_direct_mappings()")

claims it removed the last call-site, except it did not. There is still
at least one caller in

	drivers/gpu/drm/msm/msm_iommu.c

so keep the function around until all call-sites are updated.

Cc: Jason Gunthorpe <jgg@ziepe.ca>
Fixes: 69e5a17511f6 ("iommu: Remove useless flush from iommu_create_device_direct_mappings()")
Signed-off-by: Joerg Roedel <jroedel@suse.de>
---
 include/linux/iommu.h | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/include/linux/iommu.h b/include/linux/iommu.h
index 522efdc7d815..8cce372a33f1 100644
--- a/include/linux/iommu.h
+++ b/include/linux/iommu.h
@@ -853,6 +853,12 @@ void iommu_set_dma_strict(void);
 extern int report_iommu_fault(struct iommu_domain *domain, struct device *dev,
 			      unsigned long iova, int flags);
 
+static inline void iommu_flush_iotlb_all(struct iommu_domain *domain)
+{
+	if (domain->ops->flush_iotlb_all)
+		domain->ops->flush_iotlb_all(domain);
+}
+
 static inline void iommu_iotlb_sync(struct iommu_domain *domain,
 				  struct iommu_iotlb_gather *iotlb_gather)
 {
@@ -1137,6 +1143,10 @@ static inline ssize_t iommu_map_sg(struct iommu_domain *domain,
 	return -ENODEV;
 }
 
+static inline void iommu_flush_iotlb_all(struct iommu_domain *domain)
+{
+}
+
 static inline void iommu_iotlb_sync(struct iommu_domain *domain,
 				  struct iommu_iotlb_gather *iotlb_gather)
 {
-- 
2.47.0

Re: [PATCH] iommu: Restore iommu_flush_iotlb_all()

[Will Deacon]

On Tue, Oct 29, 2024 at 12:23:02PM +0100, Joerg Roedel wrote:
> From: Joerg Roedel <jroedel@suse.de>
> 
> This patch restores the iommu_flush_iotlb_all() function.
> Commit
> 
> 	69e5a17511f6 ("iommu: Remove useless flush from iommu_create_device_direct_mappings()")
> 
> claims it removed the last call-site, except it did not. There is still
> at least one caller in
> 
> 	drivers/gpu/drm/msm/msm_iommu.c
> 
> so keep the function around until all call-sites are updated.
> 
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Fixes: 69e5a17511f6 ("iommu: Remove useless flush from iommu_create_device_direct_mappings()")
> Signed-off-by: Joerg Roedel <jroedel@suse.de>
> ---
>  include/linux/iommu.h | 10 ++++++++++
>  1 file changed, 10 insertions(+)

Acked-by: Will Deacon <will@kernel.org>

Will

Re: [PATCH] iommu: Restore iommu_flush_iotlb_all()

[Jason Gunthorpe]

On Tue, Oct 29, 2024 at 12:23:02PM +0100, Joerg Roedel wrote:
> From: Joerg Roedel <jroedel@suse.de>
> 
> This patch restores the iommu_flush_iotlb_all() function.
> Commit
> 
> 	69e5a17511f6 ("iommu: Remove useless flush from iommu_create_device_direct_mappings()")
> 
> claims it removed the last call-site, except it did not. There is still
> at least one caller in
> 
> 	drivers/gpu/drm/msm/msm_iommu.c

Yikes, that iss some scary stuff

> so keep the function around until all call-sites are updated.
> 
> Cc: Jason Gunthorpe <jgg@ziepe.ca>
> Fixes: 69e5a17511f6 ("iommu: Remove useless flush from iommu_create_device_direct_mappings()")
> Signed-off-by: Joerg Roedel <jroedel@suse.de>
> ---
>  include/linux/iommu.h | 10 ++++++++++
>  1 file changed, 10 insertions(+)

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Strange that 0-day didn't flag this in my trees. We really need to get
0-day running on the iommu list too :\

Jason

Re: [PATCH] iommu: Restore iommu_flush_iotlb_all()

[Joerg Roedel]

On Tue, Oct 29, 2024 at 08:42:01AM -0300, Jason Gunthorpe wrote:
> Strange that 0-day didn't flag this in my trees. We really need to get
> 0-day running on the iommu list too :\

In general I agree, but these issues where found with
allnoconfig/allmodconfig builds. Or in other words, compile tests I
expect patch submitters to do before sending changes my way.

Regards,

	Joerg

Re: [PATCH] iommu: Restore iommu_flush_iotlb_all()

[Jason Gunthorpe]

On Tue, Oct 29, 2024 at 01:03:42PM +0100, Joerg Roedel wrote:
> On Tue, Oct 29, 2024 at 08:42:01AM -0300, Jason Gunthorpe wrote:
> > Strange that 0-day didn't flag this in my trees. We really need to get
> > 0-day running on the iommu list too :\
> 
> In general I agree, but these issues where found with
> allnoconfig/allmodconfig builds. Or in other words, compile tests I
> expect patch submitters to do before sending changes my way.

I don't necessarily expect that. We have 0-day robots to do this, and
they are more effective..

Jason