CURRENT, devbuf memory allocation and hald

Pascal Hofstee caelian at
Thu Apr 17 11:30:44 UTC 2008


After updating my CURRENT system i noticed that my system started to
completely break down after a while (ususally after coming home from a
day at college) with the console spewing loads of messages about
ata-subsystem(s) being unable to allocate any memory, existing shells
no longer being able to fork new subprocesses because of memory
problems etc. Initially i had assumed it might bave been a rare side
effect of the recent commits to the ATA-system, at least one of which
Soren mentioned fixed a couple of bogons.

Now after some additionial investigation i have come to the
realisation that the most likely source of the problem was because
according to vmstat -m the "InUse" statistics for the memory type
"devbuf" were consistently increasing by about 140k over a 10 second
time peroid. I decided to shoot down any daemons originating from
port-installed software to ensure the problem was indeed caused by
some part of the base system and not some 3rd party influence and i
managed to trace the problem down to hald.

Whenever i halted hald the increase in devbuf-usage would halt (not
decline simply stay at the same value), and as soon as i started hald
again the increase would resume with the 140k/10s ratio until hald was
stopped again.

If anyone in the FreeBSD-GNOME team might have any idea with regards
to what may be causing this problem i am more than happy to cooperate
in trying to resolve this issue.

The system in question is an AMD Athlon64 3800+, 2GB RAM using an MSI
K9N-Neo (nVidia MCP55-based mainboard) running FreeBSD/i386
8.0-CURRENT as of April 16th late afternoon CET.

With kind regards,
  Pascal Hofstee

More information about the freebsd-gnome mailing list