On 2024/3/27 21:05, Jason Gunthorpe wrote:
On Wed, Mar 27, 2024 at 05:54:33AM -0700, Yi Liu wrote:There is no error handling now in __iommu_set_group_pasid(), it relies on its caller to loop all the devices to undo the pasid attachment. This is not self-contained and has unnecessary remove_dev_pasid() calls on the devices that have not changed in the __iommu_set_group_pasid() call. this results in unnecessary warnings by the underlying iommu drivers. Like the Intel iommu driver, it would warn when there is no pasid attachment to destroy in the remove_dev_pasid() callback. The ideal way is to handle the error within __iommu_set_group_pasid(). This not only makes __iommu_set_group_pasid() self-contained, but also avoids unnecessary warnings. Fixes: 16603704559c7a68 ("iommu: Add attach/detach_dev_pasid iommu interfaces") Suggested-by: Jason Gunthorpe <jgg@xxxxxxxxxx> Signed-off-by: Yi Liu <yi.l.liu@xxxxxxxxx> --- drivers/iommu/iommu.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-)Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx> This will need revising when we get to PASID replace, but good enough for now
yes, the rollback part would be revised a bit when coming to support PASID domain replacement. Already in the pending list, will send it out soon. :) -- Regards, Yi Liu
