MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/4151
# Merge Request Required Information
JIRA: https://issues.redhat.com/browse/RHEL-28780
JIRA: https://issues.redhat.com/browse/RHEL-12083
JIRA: https://issues.redhat.com/browse/RHEL-12322
JIRA: https://issues.redhat.com/browse/RHEL-29105
JIRA: https://issues.redhat.com/browse/RHEL-29357
JIRA: https://issues.redhat.com/browse/RHEL-29359
Omitted-fix: ed8b94f6e0ac ("powerpc/pseries/iommu: Fix iommu initialisation during DLPAR add")
- Reverted by 1fba2bf8e9d5 ("Revert "powerpc/pseries/iommu: Fix iommu initialisation during DLPAR add"")
Upstream Status: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Upstream Status: git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu.git
branch: next
Tested: In progress
- general cki coverage
- Nvidia testing arm-smmu-v3 and iommufd related changes they have requested.
- Multiple rounds testing of amd_iommu, intel_iommu, and arm-smmu-v3 with
various iommu configurations with disk i/o using fio,
covering lazy iotlb invalidation, strict iotlb invalidation,
and passthrough. Also tested with forcedac set. Intel
Scalable Mode capable systems tested with the iotlb invalidation
policies, and passthrough with scalable mode enabled, and disabled.
AMD systems tested tested with v1 pages tables and v2.
- Tested booting with various iommu configurations, and verifying system
in correct state on AMD, Intel, and ARM.
- Limited test on ppc64le. The system I had access to was
setting up a 64-bit bypass window, and using dma_direct
calls. It ran, but since I don't normally touch ppc64le
iommu code, I need to investigate more or get IBM assistance
to more thoroughly test it.
- Working on getting testing assistance from IBM for the s390x changes.
## Summary of Changes
This brings iommu, iommufd, and dma mapping api up to 6.9 with some additions from Joerg's
next branch minus some commits changes in a 6.9 SEV-SNP pull for AMD. Some hightlights:
- The removal of the amd_iommu_v2 code, and the addition of it's replacement based on the
iommu core SVA api, along with a re-org of the amd_iommu code.
- The migration of s390 to the iommu core dma-iommu dma ops implementation, joining Intel,
AMD, and ARM as users of the same code base.
- The beginnings of a re-work of the arm-smmu-v3 driver by Jason, and others.
- A number of changes to iommufd as it continues to get fleshed out.
- IOPT memory usage observability (code that was basis for talk at LPC last year)
Example output in vmstat files:
```
# grep iommu /sys/devices/system/node/node*/vmstat
/sys/devices/system/node/node0/vmstat:nr_iommu_pages 342
/sys/devices/system/node/node1/vmstat:nr_iommu_pages 0
```
- Continued work on shared virtual addressing and io page faulting (PRI).
- Dynamic swiotlb memory pools. This is not enabled yet, as they still seem to be
shaking out issues upstream, but the code is in place now.
- Re-working of iommu core domain allocation.
Note: iommufd selftest is being enabled in separate work that has been delegated to
another engineer starting to help with iommu. So that will be enabled in the
next few weeks to add more coverage for iommufd.
Conflicts wise, they should be noted in the individual commits, but
not too bad overall. 13/30 were dropping unsupported bits, and another
8 were context diffs. A couple caused by out of order backports due
to fixes, and couple upstream conflicts from colliding patchsets that
had to be resolved in the merge commits.
Signed-off-by: Jerry Snitselaar <jsnitsel@redhat.com>
Approved-by: Jan Stancek <jstancek@redhat.com>
Approved-by: Donald Dutile <ddutile@redhat.com>
Approved-by: Phil Auld <pauld@redhat.com>
Approved-by: David Airlie <airlied@redhat.com>
Approved-by: Lenny Szubowicz <lszubowi@redhat.com>
Approved-by: Steve Best <sbest@redhat.com>
Approved-by: John W. Linville <linville@redhat.com>
Approved-by: Mark Langsdorf <mlangsdo@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>
Merged-by: Lucas Zampieri <lzampier@redhat.com>
304e2a4e29