ACK: [SRU][Bionic][PATCH 0/2] Revert PCI/MSI changes to fix boot issues (LP#1946149)
Thadeu Lima de Souza Cascardo
cascardo at canonical.com
Fri Oct 15 11:17:46 UTC 2021
On Fri, Oct 15, 2021 at 12:54:27PM +0200, Kleber Sacilotto de Souza wrote:
> BugLink: https://bugs.launchpad.net/bugs/194614
> [ Impact ]
> The bionic 4.15 kernels are failing to boot on r5.metal instances on AWS. The
> default kernel is bionic/linux-aws-5.4(5.4.0-1056-aws), when changing to
> bionic/linux-aws(4.15.0-1113-aws) or bionic/linux (4.15.0-160.168) the machine
> fails to boot the 4.15 kernel.
> This problem only appears on metal instances, which uses NVME instead of XVDA
> [ Fix ]
> It was discovered that after reverting the following two commits from upstream
> stable the 4.15 kernels can be booted again on the affected AWS metal instance:
> PCI/MSI: Enforce that MSI-X table entry is masked for update
> PCI/MSI: Enforce MSI[X] entry updates to be visible
Both commits seem to "fix" theoretical or specification issues that have been
"broken" since forever. We should just let them be on our 4.15 kernels. But I
wonder if this is going to break when applied on later series, or if there is
some interaction that we are not aware of. Does it work when applied to 5.4?
> [ Test Case ]
> Deploy a r5.metal instance on AWS with a bionic image, which should boot
> initially with bionic/linux-aws-5.4. Install bionic/linux or bionic/linux-aws
> (4.15 based) and reboot the system.
> [ Where problems could occur ]
> These two commits are part of a larger patchset fixing PCI/MSI issues which were
> backported to some upstream stable releases. By reverting only part of the set
> we might end up with MSI issues that were not present before the whole set was
> applied. Regression potential can be minimized by testing the kernels with these
> two reverted patches on all the platforms available.
> Kleber Sacilotto de Souza (2):
> UBUNTU: SAUCE: Revert "PCI/MSI: Enforce MSI[X] entry updates to be
> UBUNTU: SAUCE: Revert "PCI/MSI: Enforce that MSI-X table entry is
> masked for update"
> drivers/pci/msi.c | 20 --------------------
> 1 file changed, 20 deletions(-)
More information about the kernel-team