lrmd crashes
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
pacemaker (Ubuntu) |
Fix Released
|
Medium
|
Unassigned | ||
Xenial |
Fix Released
|
Medium
|
Dan Streetman |
Bug Description
[impact]
lrmd crashes and dumps core.
[test case]
I can not reproduce, but it is reproducable in the specific setup of the person reporting the bug to me.
[regression potential]
this patches the cancel/cleanup part of the code, so regressions would likely involve possible memory leaks (instead of use-after-free segfaults), failure to correctly cancel or cleanup operations, or other failure during cancel action.
[scope]
this is fixed by commits:
933d46ef2059175
94a4c58f675d163
dc36d4375c04902
deabcc5a6aa93da
b85037b75255061
which are all included starting with version 1.1.17, and Bionic includes version 1.1.18, so this is fixed already in Bionic and later.
This is needed only for Xenial.
[other info]
As mentioned in the test case section, I do not have a setup where I'm able to reproduce this, but I can ask the initial reporter to test and verify the fix, and they have verified a test build fixed the problem for them.
Also, the upstream commits removed two symbols, which I elided from the backported patches; those symbols are still available and, while it is unlikely there were any users of those symbols outside pacemaker itself, this change should not break any possible external users. See patch 0002 header in the upload for more detail.
description: | updated |
Changed in pacemaker (Ubuntu Xenial): | |
status: | New → In Progress |
assignee: | nobody → Dan Streetman (ddstreet) |
importance: | Undecided → Medium |
Changed in pacemaker (Ubuntu): | |
status: | New → Fix Released |
description: | updated |
Changed in pacemaker (Ubuntu): | |
importance: | Undecided → Medium |
This looks good, though this is quite a lot of code changes (and refactoring) for a bug without a clear reproduction scenario. Would it be possible to, along with the verification to be done by the reporting person, perform some sanity runs to make sure the cancel/cleanup parts of the code did not regress? Thanks!