Hello!
I thought I report it here – although unless it rings a bell I won´t
probably be doing much about it like a git-bisect since that machine is
slow and it could take ages…
With Debian kernel 3.3.0-trunk-686-pae the machine suspends and resume for
lots of times for 30-40 days uptime and more if I let it.
With Debian kernel 3.4-trunk-686-pae as well as 3.5-trunk-686-pae tried
out today often not always the kernel locks up hard after resume. That is,
the image is read back but then when switching to graphics or shortly
afterwards the mouse pointer is frozen and the machine does not respond to
ping anymore.
Graphics is ancient:
deepdance:~> lspci -nn | grep VGA
01:00.0 VGA compatible controller [0300]: S3 Inc. SuperSavage IX/C SDR
[5333:8c2e] (rev 05)
CPU is Mobile Intel(R) Pentium(R) III CPU - M 1133MHz, the machine has
768 MiB of RAM.
Filesystems are (aged) BTRFS (which contributes to the extreme slowness of
the machine it seems – apt-get (dist-)upgrade without eatmydata takes
long).
Kernel parameters are:
linux /vmlinuz-3.3.0-trunk-686-pae root=/dev/mapper/deepdance-debian ro
vga=791 threadirqs init=/bin/systemd resume=/dev/mapper/deepdance-swap
I didn´t report this with 3.4 cause I hoped that difficult to debug bug
would just go away with 3.5, but it seems it didn´t. If it could just dump
out something useful before it crashes… I am using threadirqs since quite
some time on various machines, but I could try disabling it if it seems to
be suspicious.
I am back on 3.3 right now. It doesn´t matter that much on this machine.
Unless any brilliant ideas I might just keep it at this report. Maybe it
rings a bell for someone who has a suggestion to try out.
I have 3.4 on ThinkPad T42 with some radeon chip and 3.4 and since some
while 3.5 on ThinkPad T520 with Intel Sandybridge graphics – no issues
there. So it only seems to affect this machine. T520 uses threadirqs, I
think the T42 as well.
Thanks,
--
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA B82F 991B EAAC A599 84C7