Re: [PATCH v3 0/2] PCI: Wait after FLR for devices that don't honor Immediate Readiness

[Bjorn Helgaas]

On Mon, Oct 12, 2026 at 05:23:31PM -0500, Bjorn Helgaas wrote:
> On Sun, Oct 11, 2026 at 09:01:43PM +0200, Alexander Gruhlke wrote:
> > Since v7.2, VFIO passthrough of a Samsung 990 PRO NVMe [144d:a80c] fails
> > on the second VM start, and the device is inaccessible until reboot.  The
> > drive advertises Immediate Readiness but returns ~0 for up to 7 ms after
> > FLR, so since 10baa9b4df40 ("PCI: Drop unnecessary retries when restoring
> > BARs") the config restore after FLR is lost.
> > 
> > Patch 1 ignores Immediate Readiness on that drive.  Firmware 8B2QJXD7
> > fixes it, but the Revision ID and Subsystem ID don't change, so the
> > quirk can't tell firmware versions apart.  A 990 EVO Plus [144d:a80d]
> > reports the same errors [1], but it isn't confirmed to advertise
> > Immediate Readiness, so it's not included.
> > 
> > Patch 2 is independent and keeps pci_dev_wait() from treating ~0 as
> > ready with RRS Software Visibility, which also affects the Intel
> > [8086:0a54] in the open d591f6804e7e regression.  Hui, Linux-Fan: testing
> > patch 2 on that drive would be appreciated.
> > 
> > Tested on v7.2.9 (Ryzen 9 7950X, X670E): with patch 1 and drive firmware
> > 1B2QJXD7, three VM starts in a row work, and every FLR reports "ready
> > 0ms after FLR".  With firmware 8B2QJXD7, plain v7.2.9 works.
> > 
> > Patch 2 is only tested for regressions, as the drive no longer shows the
> > problem: below a Root Port with RRS SV, three VM starts work and every
> > FLR reports "ready 0ms after FLR".  The VF path is untested with the
> > patch, but on another machine, an X550 VF below a Root Port with RRS SV
> > reads ~0 from the Vendor ID and valid data from the Command register
> > 100 ms after FLR.  v1's variant, which read the Command register, was
> > tested against the old firmware, and the drive's Vendor ID read ~0
> > while it wasn't ready.  Applies to pci/next.
> > 
> > Changes since v2:
> > - Patch 1: add Lukas's Reviewed-by
> > - Patch 2: read the Command register on VFs whose Vendor ID reads ~0
> >   instead of treating them as ready (Lukas, sashiko)
> > 
> > v2: https://lore.kernel.org/r/20261011163249.18479-1-gruhlke@xxxxxxxxxxx
> > v1: https://lore.kernel.org/r/20261010160529.27974-1-gruhlke@xxxxxxxxxxx
> > 
> > [1] https://bbs.archlinux.org/viewtopic.php?id=314833
> > 
> > Alexander Gruhlke (2):
> >   PCI: Ignore Immediate Readiness on Samsung 990 PRO NVMe
> >   PCI: Don't treat ~0 as ready in pci_dev_wait() with RRS SV
> > 
> >  drivers/pci/pci.c    | 14 ++++++++++----
> >  drivers/pci/quirks.c | 11 +++++++++++
> >  2 files changed, 21 insertions(+), 4 deletions(-)
> 
> Applied to pci/reset for v7.4, thank you!

On second thought, I think at least the first should be in v7.3
because it fixes a v7.2 regression.  The other regression dates to
v6.12, but it seems worth keeping the fixes together.

I moved them both to pci/for-linus for v7.3.