I've reproduced this on a mainline linux kernel, and it does appear to
just nvme.poll_queues being poorly documented. A check was added in
upstream commit 9c9e76d5792b1 to prevent the value from being set too
large at runtime, but it also changed the behaviour when loaded at boot
(incidentally, I think) from a silent capping to a boot failure. That
seems silly to me - it's a boot-necessary module for most devices these
days, it shouldn't cause a boot failure just because a boot arg was
large. If anything it should cap and produce a warning. At the very
least this behaviour should be documented somewhere.
As this is an upstream issue I would recommend opening a bug report
there. The fix would be just breaking apart the if statement in
io_queue_count_set() in drivers/nvme/host/pci.c into two parts, one that
checks for validity like it does now, and the other caps the value if
necessary.
I'll mark this bug report as 'opinion' since I don't think it's worth
diverging from upstream by making an Ubuntu-specific fix, but this is
still a useful reference for future upstream discussions. Let me know if
you submit upstream, I can add a link to the upstream bug here too.
** Changed in: xen (Ubuntu)
Status: New => Invalid
** Changed in: linux
Status: New => Confirmed
** Changed in: linux (Ubuntu)
Status: New => Opinion
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2170270
Title:
Xen Dom0 fails to boot with "nvme.poll_queues=24" and
"dom0_max_vcpus=6 dom0_vcpus_pin"
To manage notifications about this bug go to:
https://bugs.launchpad.net/linux/+bug/2170270/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
