Processes freezing in D state

Andrew Lewycky (amplewycky@undergrad.math.uwaterloo.ca)
Fri, 22 Jan 1999 11:04:04 -0500

Once under 2.2.0-pre8 and once under 2.2.0-final, I've had many of the
processes on my system simultaneously enter the D state. This kernel and
computer are UP x86, and nothing like this was happening in pre7 or
earlier. (But it is difficult to trigger. Today it occurred after
running
for about an hour, while yesterday my computer was on for ~16 hours,
doing roughly the same things, without any trouble.)

What happened was that I was working under X, when I noticed that both
Netscape and apt had frozen. ps ax reported that 11 processes were in
the D state, including update, syslogd, login, nmbd, communicator,
http (for apt), sync, bash and xterm. Alt-SysRq-T claimed that they all
had EIPs of C01B4000, which is init_task_union. (I guess that software
task switching broke thread_saved_pc.) Attempts to login on the console
or start another xterm froze. However I was able to determine the
following:

ls /tmp -> ok
touch /tmp/foo -> D
ls /tmp -> D
ls /tmp/foo -> no such file

cat /tmp/bar > /dev/null -> D (where /tmp/bar exists)
touch /tmp/bar -> D

here's what an strace of ls /tmp looked like:

open("/tmp", O_RDONLY | O_NONBLOCK) = 3
fcntl(3, F_SETFD, FD_CLOEXEC) = 0
lseek(3, 0, SEEK_CUR) = 0
getdents(3, <strace hangs here until CTRL-C>

here's what an strace of touch /tmp/foo looked like:

stat("/tmp/foo", 0xbffffc84) = -1 ENOENT
open("/tmp/foo", O_WRONLY | O_CREAT | O_TRUNC, 0666

That's all I've got so far. If there's a way that I can get the stack
trace for a frozen process, I'll try it next time this happens.

Thanks in advance.
Andrew Lewycky
amplewycky@undergrad.math.uwaterloo.ca

P.S. I'm not subscribed to linux-kernel, but I do read it on the web.
There's a long lag before messages show up there, so please CC me.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/