LOR No 9 and strange other kernel messages

Don Lewis truckman at FreeBSD.org
Sat Jun 5 21:32:25 GMT 2004

On  5 Jun, Tai-hwa Liang wrote:
> On Fri, 4 Jun 2004, Don Lewis wrote:
>> On  4 Jun, Tai-hwa Liang wrote:
>> > On Fri, 4 Jun 2004, Daniel Lang wrote:
>> >> Hi,
>> >>
>> >> I just went through my syslog and stumbled across a LOR, which is
>> >> documented on the Zabbadoz LOR page (id 009).
>> >>
>> >> The reason for this mail is the following kernel messages:
>> >>
>> >> [..]
>> >> Jun  1 11:08:06 atrbg11 kernel: x: 2
>> >> Jun  1 11:08:06 atrbg11 kernel: x: 2
>> >> [..]
>> >>
>> >> Seems some leftovers from a developer who did not want to
>> >> bother others with the gory details. ;-))
>> >
>> > I guess that came from sys/dev/sound/pcm/sound.c:872. My Thinkpad T40 also
>> > ran into this when the disk/network(combination) load is high enough.
>> Looks likely, but I'd expect this would only get triggered when sysctl
>> is used to set or query the number of vchans.  In any case, this marks
>> another spot where locking is broken in the sound code.
> I use the default setting without tweaking any vchan related sysctl
> such like hw.snd.pcm0.vchans(leave it to default 0) or hw.snd.maxautovchans
> (also left to 0 on my T40).

Maybe your sound client software (or something else on the system) is
querying for the number of vchans.  Even "sysctl -a" will invoke the

> I'm wondering about whether this "bug" is ACPI related(interrupt storm?)
> since recently when I run "make buildworld buildkernel" on one terminal and
> do large file transmission(FTP ISO images on using em0) at the same time,
> the sound tends to become "broken." Meanwhile, the "x: 2" message popped
> on my console and later on my em0 doesn't work anymore(I have to do a reboot
> to get it back; however, the other part of the system still works while
> em0 was dead: sound continue playing, can compile/edit programs).

It is quite possible that locking collisions in the sound code might
happen more frequently when the system is busy.

Do your sound hardware and em0 share the same irq?  If they do, that
might be a clue that the problems are related.  Can you boot with ACPI
disabled, and if so, does it change the symptoms?

More information about the freebsd-current mailing list