Qemu 1:6.2+dfsg-2ubuntu6 deadlock bug
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
qemu (Ubuntu) |
Fix Released
|
High
|
Christian Ehrhardt | ||
Jammy |
Fix Released
|
Medium
|
Christian Ehrhardt | ||
Kinetic |
Fix Released
|
High
|
Christian Ehrhardt |
Bug Description
[Impact]
* A deadlock in the connect/disconnect code is present in qemu 6.2
* backport the fix which moves the unlock a bit to avoid racing with
a lock triggered by clipboard unregistration. For more see the
referenced upstream commit
[Test Plan]
* start a qemu wit hvnc
qemu-
* Connect and disconnect with VNC against it (TigerVNC is known
to work for this repro). You might wait a while in between
* Without the fix at some point a race will occur which deadlocks qemu -
no interaction is possible anymore.
[Where problems could occur]
* Lock changes are always scary, but this one isn't too complex.
It is not rewriting all of it, just moving it up a few lines so
that qemu_clipboard_
I can think of a problem if we'd find out later that inside of
that call we would have needed the lock to be held already.
It seems that isn't the case (it grabs the lock where needed),
but that is the issue I could think of. If that would happen it
would more likely then not be a deadlock but instead accessing
unsafe data which could lead to a variety of symptoms that are
hard to predict.
[Other Info]
* n/a
--- original bug ---
There is a known bug that will cause VM deadlock, the patch should be merged and released:
Related branches
- Sergio Durigan Junior (community): Approve
- Canonical Server: Pending requested
- git-ubuntu import: Pending requested
-
Diff: 101 lines (+79/-0)3 files modifieddebian/changelog (+7/-0)
debian/patches/series (+1/-0)
debian/patches/ubuntu/lp-1970563-ui-vnc.c-Fixed-a-deadlock-bug.patch (+71/-0)
- Sergio Durigan Junior (community): Approve
- Canonical Server: Pending requested
- git-ubuntu import: Pending requested
-
Diff: 101 lines (+79/-0)3 files modifieddebian/changelog (+7/-0)
debian/patches/series (+1/-0)
debian/patches/ubuntu/lp-1970563-ui-vnc.c-Fixed-a-deadlock-bug.patch (+71/-0)
That's clearly a fix for a bug, but I couldn't identify an upstream issue which describes the problem. The commit message has:
Fixes: 0bf41cab
but that's a reference to another commit, not to an issue. Finding original description of the bug would help identifying a test case for the SRU.
@xp are you able to point us to the upstream bug report, or to provide steps to reproduce the issue which we can use to verify the fix?
I'm marking this as Incomplete for now because the description of the problem is too vague, but I think this will become a valid SRU case.