Hi Viktor, Krystian,
Two updates before drawing conclusions:
1. srv06 running noble HWE 7.0.0-31 crashed after ~21d 21h (Oct 13 20:00:31
UTC).
Only two lines reached the journal before the host froze:
BUG: kernel NULL pointer dereference, address: 0000000000000000
#PF: supervisor write access in kernel mode
The visible console tail was a qemu ioctl path (kvm_on_user_return), RIP not
captured.
So 7.0 is not crash-free on these hosts yet, it's just less frequent
(6.8: every 3-9 days, 7.0.0-31: ~22 days). srv06 also had two earlier crashes
on mainline 6.14 with AMD-Vi IO_PAGE_FAULT, so a host-specific factor is
possible.
(Also, for the record: this is our TEST environment, not production.)
2. Viktor, regarding the commit: "KVM: x86/mmu: Zap _all_ roots when unmapping
gfn
range in TDP MMU" is CVE-2021-47639, fixed in 5.15.33 / 5.16.19 / 5.17.2
(2022),
so it should already be present in Ubuntu 6.8. Could you share the actual
commit
hashes of the 6.9/6.10 change you mean? Without them it's hard for Krystian
to
evaluate anything.
What I changed:
- srv04 and srv06 on 7.0.0-38-generic
- kernel.panic_on_oops=1, kernel.panic=30 on AMD hosts (auto-reboot instead of
hours-long freezes)
- kdump enabled ("ready to kdump" on srv06; srv04 after next reboot)
Question:
- Are you aware of KVM/SVM fixes between 7.0.0-31 and 7.0.0-38 that could be
relevant?
The next crash should give us a full vmcore with a reliable stack, which
I'll attach.
Thanks,
Radovan
** CVE added: https://cve.org/CVERecord?id=CVE-2021-47639
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2163642
Title:
Kernel panic "Attempted to kill the idle task" on AMD Opteron multi-
node NUMA under KVM (6.8.0-137)
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2163642/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
