Public bug reported:
This patch fixes a race condition that was reintroduced in the 4.8
kernel, as part of the P9 changes, after having been originally fixed in
3.19. The effect of the race condition is that a secondary thread can
start trying to execute code from the guest while the thread is still in
hypervisor mode, so it can cause many different symptoms, one of which
seems to be the timebase corruption that leads to the lockup in
ktime_get_ts64. It could cause other problems such as CPU cores locking
up hard or memory corruption.
This patch is only needed on the host. It should be applied to any 4.8
kernel being used as a host, including the Ubuntu 16.10 kernel.
Hi Paul
I built a kernel with your patch and put at the host and in the guest. I can
still see some ktime_get_ts64 at the host. The guest dmesg is clean.
This patch fixes another race condition I found in the fastsleep code.
Please apply this patch as well and test.
Sure will do and provide feedback.
** Affects: linux (Ubuntu)
Importance: Undecided
Assignee: Taco Screen team (taco-screen-team)
Status: New
** Tags: architecture-ppc64 bugnameltc-132390 severity-critical
targetmilestone-inin1610
** Tags added: architecture-ppc64 bugnameltc-132390 severity-critical
targetmilestone-inin1610
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1636330
Title:
guest experiencing Transmit Timeouts on CX4
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1636330/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
