FS#55645 - [linux] 4.13.2 kworker deadlock in mac80211; breaks poweroff/reboot

Attached to Project: Arch Linux
Opened by Kevin Dodd (Jesin) - Friday, 15 September 2017, 22:04 GMT
Last edited by Doug Newgard (Scimmia) - Friday, 06 October 2017, 22:30 GMT
Task Type Bug Report
Category Kernel
Status Closed
Assigned To Tobias Powalowski (tpowa)
Jan Alexander Steffens (heftig)
Architecture All
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 5
Private No

Details

Linux 4.13.0 through 4.13.2 deadlock in mac80211 on some systems. This fills the system logs with complaints, but more importantly prevents the system from fully powering itself off and rebooting automatically. Rebooting such a system requires the user to be physically present and able to hold down the power button or unplug the power supply or battery. (If you need more detailed information about the symptoms of the problem, including system logs, my upstream bug report at https://bugzilla.kernel.org/show_bug.cgi?id=196927 has those.)

The bug is reportedly fixed by this upstream commit:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=bde59c475e0883e4c4294bcd9b9c7e08ae18c828
This task depends upon

Closed by  Doug Newgard (Scimmia)
Friday, 06 October 2017, 22:30 GMT
Reason for closing:  Fixed
Additional comments about closing:  linux 4.13.5-1
Comment by Kevin Dodd (Jesin) - Saturday, 16 September 2017, 17:22 GMT
I believe this means the relevant patch can be retrieved from here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/patch?id=bde59c475e0883e4c4294bcd9b9c7e08ae18c828
Comment by Kevin Dodd (Jesin) - Saturday, 16 September 2017, 19:58 GMT
It's possible that this patch could be needed as well (and should be applied first):
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/patch?id=98e93e968e4947cd71c2eb69e323682daa453ee7
Comment by Kevin Dodd (Jesin) - Saturday, 16 September 2017, 20:09 GMT
I am attaching a patch to the PKGBUILD that applies both of these patches (without updating the pkgrel, so it can be applied to a later pkgrel if necessary)
   PKGBUILD.patch (1.7 KiB)
Comment by Kevin Dodd (Jesin) - Saturday, 16 September 2017, 22:58 GMT
I am now using a kernel built from the PKGBUILD obtained by applying the patch I attached in my last comment to the PKGBUILD for linux-4.13.2-1, and the bug is gone. As far as I'm concerned, if you accept that patch, you can resolve this bug as "fixed".
Comment by Kevin Dodd (Jesin) - Monday, 18 September 2017, 23:19 GMT
I suppose the upstream patches should all be grouped together, instead of putting some of them after the local config files. Here is a replacement patch that does this. (This produces the same package as the previous patch, so it also counts as a fix.)
   PKGBUILD.patch (1.8 KiB)
   PKGBUILD (8.6 KiB)
Comment by Kevin Dodd (Jesin) - Thursday, 21 September 2017, 15:40 GMT
The fixes were not merged into 4.13.3 (though I do not know why), so they still need to be applied separately. Here is a PKGBUILD for 4.13.3 that includes these fixes.
   PKGBUILD (8.6 KiB)
Comment by Kevin Dodd (Jesin) - Thursday, 28 September 2017, 19:54 GMT
This is still needed in 4.13.4, so here's a new PKGBUILD
   PKGBUILD (8.6 KiB)
Comment by loqs (loqs) - Monday, 02 October 2017, 20:45 GMT
https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/commit/queue-4.13/mac80211-fix-deadlock-in-driver-managed-rx-ba-session-start.patch?id=07da3bce8d789380945aba0a21541b6893da5b7f
Should be in 4.13.5 unless the patch is found to cause an issue.
Comment by Kevin Dodd (Jesin) - Thursday, 05 October 2017, 21:17 GMT
Linux 4.13.5 includes the fix, so as soon as that version hits [testing] you can close this bug as resolved.
Comment by loqs (loqs) - Friday, 06 October 2017, 18:25 GMT
@Jesin if you could just verify 4.13.5-1 now in testing does resolve the issue then request that this FS is closed.