New issue
Advanced search Search tips

Issue 896773 link

Starred by 4 users

Issue metadata

Status: Available
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 3
Type: Bug



Sign in to add a comment

hung task on host kernel when running crostini tast tests

Project Member Reported by smbar...@chromium.org, Oct 18

Issue description

R71-11151.6.0 - nami - vm.CrostiniStartEverything
https://stainless.corp.google.com/browse/chromeos-autotest-results/249720163-chromeos-test/

Oddly this is only showing up on the 71 branch as far as I can tell.
 
ramoops
127 KB View Download
Owner: dverkamp@chromium.org
Status: Assigned (was: Untriaged)
Daniel, can you take a look at this?
I can take a look, but at first glance, this looks like it is a (host) kernel bug of some kind, so I'm not sure I am the best equipped to diagnose it. :)

The crash in the ramoops is in the host 4.4 kernel. At this level, crosvm is just another user-mode process doing I/O operations through the normal read()/write()/fsync() interfaces. If we can hang the system using these syscalls, there's definitely something wrong in kernel-land. The kernel stacks in the oops seem to have several processors waiting on ext4 and jbd2 calls; maybe we are somehow hitting a deadlock?

If this has reproduced more than once, it would be good to get some more data points (e.g. are they all on the same host kernel version?) and try to narrow it down some more.
Owner: ----
Status: Available (was: Assigned)
I'm unassigning myself from this, since I'm not actively looking at it.

Have we seen any more of these in recent tests?

Sign in to add a comment