linux-kernel - [BUG] 4.4.8 hard-lockup after ~12h

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [day] [month] [year] [list]

Date:	Thu, 21 Apr 2016 20:32:41 -0300
From:	Henrique de Moraes Holschuh <hmh@....eng.br>
To:	LKML <linux-kernel@...r.kernel.org>
Subject: [BUG] 4.4.8 hard-lockup after ~12h

I just experienced a weird hang on 4.4.8.  Unfortunately, the logs are
sparse.  They do show an issue with either the GPU, or whatever is required
for it to talk to the driver (PCIe MSI ?).  Refer to the attached kernel log
for a large number of >10ms GPU ring 0 stalls/GPU lockups back-to-back,
right before it hang.

The screen went crazy when the GPU started misbehaving, but the box did not
hang immediately: it took a few seconds or so (enough for systemd to flush
some of the journal to disk), but then it eventually hang something
important in userspace/kernel for more than one minute, triggering the
watchdog.

The box has a Radeon 4770 (RV740) connected to two DVI monitors.  This box
runs Debian stable, and has "irqbalance" running.  The kernel that crashed
is a vanilla 4.4.8 kernel, without any changes from upstream.

This box has been well behaved so far, and did not crash on either 4.4.6 or
4.4.7.  It also works perfectly on 3.18.y (tested up to 3.18.29, when I
switched to 4.4.y).  IMHO, it looks like chances are high that the hang was
caused by a 4.4.8 regression.

Unfortunately, I really can't risk many hard-resets on this box: it messes
up the SSDs, so currently I can't bissect hangs.

/proc/interrupts, lspci -vv, and abridged kernel boot logs attached.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

View attachment "proc.interrupts.txt" of type "text/plain" (5585 bytes)

View attachment "4.4.8-crash.txt" of type "text/plain" (76832 bytes)

View attachment "lspci.txt" of type "text/plain" (63928 bytes)