]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
iommupt: Always add IOVA range to iotlb_gather in gather_range_pages()
authorYu Zhang <zhangyu1@linux.microsoft.com>
Tue, 3 Feb 2026 08:29:34 +0000 (16:29 +0800)
committerJoerg Roedel <joerg.roedel@amd.com>
Tue, 3 Feb 2026 13:36:21 +0000 (14:36 +0100)
commitb48ca920613858b477f75946907e72c74570af05
tree387bd80e1ecd5c6152e5b19b9c4b9f38ec634a82
parent12248a3862d50ef1a889153bb222d10d43e78c9d
iommupt: Always add IOVA range to iotlb_gather in gather_range_pages()

Add current (iova, len) to the iotlb gather, regardless of the setting
of PT_FEAT_FLUSH_RANGE or PT_FEAT_FLUSH_RANGE_NO_GAPS.

In gather_range_pages(), the current IOVA range is only added to
iotlb_gather when PT_FEAT_FLUSH_RANGE is set. Yet a virtual IOMMU with
NpCache uses only PT_FEAT_FLUSH_RANGE_NO_GAPS. In that case, iotlb_gather
will stay empty (start=ULONG_MAX, end=0) after initialization, and the
current (iova, len) will not be added to the iotlb_gather, causing
subsequent iommu_iotlb_sync() to perform IOTLB invalidation with wrong
parameters (e.g., amd_iommu_iotlb_sync() computes size from
gather->end - gather->start + 1, leading to an invalid range).

The disjoint check and sync for PT_FEAT_FLUSH_RANGE_NO_GAPS remain
unchanged: when the new range is disjoint from the existing gather,
we still sync first and then add the new range, so semantics for
NO_GAPS are preserved.

Fixes: 7c53f4238aa8 ("iommupt: Add unmap_pages op")
Cc: stable@vger.kernel.org
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Yu Zhang <zhangyu1@linux.microsoft.com>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
drivers/iommu/generic_pt/iommu_pt.h