Hi all, This series arose from my attempt to remove the 'devices_lock' from the ->unmap() path. In actual fact, I think the current code in mainline gets this wrong, so I've fixed up the ordering and then removed the lock. Unfortunately, this relies on my deferred invalidation work so that the invalidation range is propagated through to the sync callback:
https://git.kernel.org/pub/scm/linux/kernel/git/will/linux.git/log/?h=iommu/cmdq At this point, if we decide to do anything for earier kernels, the easiest thing is probably to nobble the ATS feature at probe time. NOTE: this has not been tested, since I don't have access to any systems capable of ATS! Feedback welcome, Will Cc: Zhen Lei <[email protected]> Cc: Jean-Philippe Brucker <[email protected]> Cc: John Garry <[email protected]> Cc: Robin Murphy <[email protected]> --->8 Will Deacon (4): iommu/arm-smmu-v3: Document ordering guarantees of command insertion iommu/arm-smmu-v3: Rework enabling/disabling of ATS for PCI masters iommu/arm-smmu-v3: Fix ATC invalidation ordering wrt main TLBs iommu/arm-smmu-v3: Avoid locking on invalidation path when not using ATS drivers/iommu/arm-smmu-v3.c | 105 ++++++++++++++++++++++++++++++++------------ 1 file changed, 77 insertions(+), 28 deletions(-) -- 2.11.0 _______________________________________________ iommu mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/iommu
