Bug#647770: FW: Re: linux-image-2.6.32-5-amd64: Random system freeze after "ohci_hcd 0000:00:12.0: HC died; cleaning up"
Eugen Sarychev wrote:
> As it happens, the picture on the screen gets frozen like a
> screenshot. Also, the last ~500ms of sound are looped, I think it is
> the sound chip buffer. Looping is going on for some minutes, then it
> is quiet.
Thanks. This means it might be possible to get some information by
keeping a terminal open running "tail -f /var/log/syslog".
> It is the completely fresh installation, made with unofficial Debian
> 6.0.3 AMD64 netinstall with firmware, downloaded from
> cdimage.debian.org and updated. The hardware was working well for 1
> year long with Ubuntu. But I can remember, that ~2 years ago on 2
> ubuntu versions (xx.04 AND xx.10) I had the same symptoms.
That's useful. I assume this means that 9.04 and 9.10 had occasional
freezes and 10.04 did not?
> Do I understand right: I should write the squeeze-backports source
> to the apt-list and install 3 packages, linux-image, linux-base and
> initframs-tools in version 3.x? Do I need something extra to do,
> aside from uninstalling of the 2.6.x-kernel packages?
I don't think linux-image 3.x is in backports. So it should be enough
to add something like
deb http://ftp.debian.org/debian experimental main
to sources.list and then to run
apt-get -t experimental install linux-base initramfs-tools
apt-get -t experimental install linux-image-3.1.0-1-amd64
and reboot into v3.1 to see whether it happens there.
> The big problem what we here have, there is no trigger for
> reproducing. I am suspicious about sound, WiFi, video, adobe-flash
> too. Because I am often using this things and logicaly it crashes
> when I'm watching something on youtube, for example. Yesterday as I
> was trying to catch the kernel log, it was frozen 2 times. The first
> time I had no log, because the LAN link was unstable for no reason
> and it happened only 2-3 minutes after booting. The second time as I
> got the log, it took several hours. This means, I will of course
> make a test with 3.x kernel, but cann not guarantee after 2 days of
> testing that the bug is gone.
> My question is, if there is a way to
> force the kernel to give more debugging info.
Adding the "debug" parameter to the kernel command line makes more
messages go to the console, but that wouldn't be too useful here,
given that you already retrieved everything you could with netconsole.
I think the interval when it was fixed in Ubuntu is our best lead at
the moment. Thanks for the quick feedback.