APPLIED: [SRU][PATCH v4 0/13][f/gcp] Fix nvme errors w/ swiotlb in confidential VMs

Khaled Elmously khalid.elmously at canonical.com
Thu Oct 14 02:37:51 UTC 2021


Applied to f/gcp - thanks for the reviews!


On 2021-10-13 02:49:54 , Khalid Elmously wrote:
> BugLink: https://bugs.launchpad.net/bugs/1943902
> 
> 
> 
> [Impact]
> Using nvme with swiotlb in confidential VMs can encounter hardware read/write errors.
> 
> [Fix]
> The following upstream patches from v5.12 address this:
> 
> 3d2d861eb03e nvme-pci: set min_align_mask
> 1f221a0d0dbf swiotlb: respect min_align_mask
> 16fc3cef33a0 swiotlb: don't modify orig_addr in swiotlb_tbl_sync_single
> 26a7e094783d swiotlb: refactor swiotlb_tbl_map_single
> ca10d0f8e530 swiotlb: clean up swiotlb_tbl_unmap_single
> c32a77fd1878 swiotlb: factor out a nr_slots helper
> c7fbeca757fe swiotlb: factor out an io_tlb_offset helper
> b5d7ccb7aac3 swiotlb: add a IO_TLB_SIZE define
> 
> [Test]
> Using a confidential VM, with 'swiotlb=force' set on the kernel command line, and an additional nvme device attached:
> 
> $ sudo mkfs.xfs -f /dev/nvme2n1
> meta-data=/dev/nvme2n1 isize=512 agcount=4, agsize=131072 blks
>          = sectsz=512 attr=2, projid32bit=1
>          = crc=1 finobt=1, sparse=0, rmapbt=0, refl
> ink=0
> data = bsize=4096 blocks=524288, imaxpct=25
>          = sunit=0 swidth=0 blks
> naming =version 2 bsize=4096 ascii-ci=0 ftype=1
> log =internal log bsize=4096 blocks=2560, version=2
>          = sectsz=512 sunit=0 blks, lazy-count=1
> realtime =none extsz=4096 blocks=0, rtextents=0
> mkfs.xfs: pwrite failed: Input/output error
> 
> 
> Note the input/output error
> 
> The error no longer happens with the fixes applied.
> 
> [Regression Potential]
> Potential regression in swiotlb could cause hardware read/write errors
> 
> 
> 
> 
> 
> 
> 
> v2:
>   Merged the NVME_CTRL_PAGE_[SIZE|SHIFT] definitions into last patch
> 
> v3:
>   Included further follow-up fixes.
> 
> v4:
>   Re-ordered the commits - no code changes from v3
> 
> 
> Bumyong Lee (1):
>   swiotlb: manipulate orig_addr when tlb_addr has offset
> 
> Christoph Hellwig (9):
>   swiotlb: add a IO_TLB_SIZE define
>   swiotlb: factor out an io_tlb_offset helper
>   swiotlb: factor out a nr_slots helper
>   swiotlb: clean up swiotlb_tbl_unmap_single
>   swiotlb: refactor swiotlb_tbl_map_single
>   swiotlb: don't modify orig_addr in swiotlb_tbl_sync_single
>   swiotlb: respect min_align_mask
>   swiotlb: remove the tbl_dma_addr argument to swiotlb_tbl_map_single
>   swiotlb: move orig addr and size validation into swiotlb_bounce
> 
> Claire Chang (1):
>   swiotlb: Fix the type of index
> 
> Jianxiong Gao (2):
>   driver core: add a min_align_mask field to struct
>     device_dma_parameters
>   nvme-pci: set min_align_mask
> 
>  drivers/iommu/intel-iommu.c |   1 -
>  drivers/nvme/host/pci.c     |   9 ++
>  drivers/xen/swiotlb-xen.c   |   3 +-
>  include/linux/device.h      |   1 +
>  include/linux/dma-mapping.h |  16 ++
>  include/linux/swiotlb.h     |  11 +-
>  kernel/dma/swiotlb.c        | 306 ++++++++++++++++++++----------------
>  7 files changed, 199 insertions(+), 148 deletions(-)
> 
> -- 
> 2.17.1
> 



More information about the kernel-team mailing list