[0/9] IOMMU: XSA-373 follow-on

Message

Jan Beulich June 9, 2021, 9:25 a.m. UTC

A number of further adjustments were left out of the XSA, for not
being a security concern (anymore in some of the cases, with the
changes put in place there). This is the collection, possibly
looking a little random in what it contains.

1: AMD/IOMMU: redo awaiting of command completion
2: AMD/IOMMU: re-work locking around sending of commands
3: VT-d: undo device mappings upon error
4: VT-d: adjust domid map updating when unmapping context
5: VT-d: clear_fault_bits() should clear all fault bits
6: VT-d: don't lose errors when flushing TLBs on multiple IOMMUs
7: VT-d: centralize mapping of QI entries
8: VT-d: drop/move a few QI related constants
9: IOMMU/PCI: don't let domain cleanup continue when device de-assignment failed

Jan

Comments

Jan Beulich June 23, 2021, 6:51 a.m. UTC | #1

On 09.06.2021 11:25, Jan Beulich wrote:
> A number of further adjustments were left out of the XSA, for not
> being a security concern (anymore in some of the cases, with the
> changes put in place there). This is the collection, possibly
> looking a little random in what it contains.
> 
> 1: AMD/IOMMU: redo awaiting of command completion
> 2: AMD/IOMMU: re-work locking around sending of commands

For these two I have v2 largely ready.

> 3: VT-d: undo device mappings upon error
> 4: VT-d: adjust domid map updating when unmapping context
> 5: VT-d: clear_fault_bits() should clear all fault bits
> 6: VT-d: don't lose errors when flushing TLBs on multiple IOMMUs
> 7: VT-d: centralize mapping of QI entries
> 8: VT-d: drop/move a few QI related constants

Kevin, any word on these?

> 9: IOMMU/PCI: don't let domain cleanup continue when device de-assignment failed

Paul, any feedback on this one?

Thanks, Jan

Tian, Kevin June 23, 2021, 6:58 a.m. UTC | #2

> From: Jan Beulich <[e-mail]>
> Sent: Wednesday, June 23, 2021 2:52 PM
> 
> On 09.06.2021 11:25, Jan Beulich wrote:
> > A number of further adjustments were left out of the XSA, for not
> > being a security concern (anymore in some of the cases, with the
> > changes put in place there). This is the collection, possibly
> > looking a little random in what it contains.
> >
> > 1: AMD/IOMMU: redo awaiting of command completion
> > 2: AMD/IOMMU: re-work locking around sending of commands
> 
> For these two I have v2 largely ready.
> 
> > 3: VT-d: undo device mappings upon error
> > 4: VT-d: adjust domid map updating when unmapping context
> > 5: VT-d: clear_fault_bits() should clear all fault bits
> > 6: VT-d: don't lose errors when flushing TLBs on multiple IOMMUs
> > 7: VT-d: centralize mapping of QI entries
> > 8: VT-d: drop/move a few QI related constants
> 
> Kevin, any word on these?

will take a look later today or tomorrow.

> 
> > 9: IOMMU/PCI: don't let domain cleanup continue when device de-
> assignment failed
> 
> Paul, any feedback on this one?
> 
> Thanks, Jan