ACK: [SRU][O][PATCH 0/1] [Ubuntu 25.04] s390/pci: Fix PCHID per port multi-function detection (LP: #2104120)
Tim Whisonant
tim.whisonant at canonical.com
Thu Mar 27 01:37:36 UTC 2025
On Tue, Mar 25, 2025 at 05:51:51PM +0100, Massimiliano Pellizzer wrote:
> BugLink: https://bugs.launchpad.net/bugs/2104120
>
> [Impact]
>
> Description: s390/pci: Fix PCHID per port multi-function detection
>
> Symptom: There are two cases in which PCI functions are incorrectly
> placed, or not placed, in a shared PCI topology. In the first
> case a PCI physical function (PF) which is initially in standby
> was incorrectly not creating a new PCI topology when hotplugged,
> but would if configured during boot. In the second case a PCI
> virtual function (VF) would be placed in the shared PCI topology
> of a PF if that PF is from the same card even if the parent PF
> is not in the configuration which should lead the VF to be
> isolated in its own single function topology.
>
> Problem: Both problems are caused by logic bugs in the determination of
> which PCI functions create topologies and when a PCI function
> may be added to an existing topology.
>
> [Fix]
>
> Cherry pick from mainline of:
> - dc287e4c9149 s390/pci: Fix SR-IOV for PFs initially in standby
> - 05793884a1f3 s390/pci: Pull search for parent PF out of pci_iov_setup_virtfn()
> - 2844ddbd540f s390/pci: Fix handling of isolated VFs
>
> [Test Plan]
>
> Fix tested by IBM.
>
> [Where problems could occur]
>
> The fix affects SR-IOV support in the s390x PCI subsystem.
> A user might experience issues like missing or misconfigured VFs,
> failed VF passthrough to guests, sysfs inconsistencies, or degraded I/O performance.
>
>
> [Other Info]
>
> The three commits mentioned above have been already included in Noble.
>
Acked-by: Tim Whisonant <tim.whisonant at canonical.com>
More information about the kernel-team
mailing list