2.2.2 processes stuck in 'D' state doing mean things to FS

David Mansfield (david@cobite.com)
Tue, 23 Feb 1999 00:22:50 -0500 (EST)

I was doing a 'load test' on the new kernel and managed to get three
processes stuck in 'D'. Here's output from ps -alx (applicable procs.):

FLAGS UID PID PPID PRI NI SIZE RSS WCHAN STA TTY TIME COMMAND
800 0 1285 1284 0 0 876 0 wait_on_sup DW p1 0:00 (tar)
0 0 1316 1288 0 0 772 0 wait_on_sup DW p2 0:00 (rm)
0 0 1319 1302 0 0 760 0 down_failed DW p3 0:00 (rm)

Kernel is SMP 2.2.2 + DAC960 Raid Beta4 compiled w/gcc2.7.2.3, RedHat 5.2.
256MB. There is only 1 cpu (450 MHZ Intel PII) installed (other goes in
tomorrow, maybe :-( but it is compiled SMP.

Here's what I did:

window #1: run 32 'memory hogs' each randomly accessng 10MB of ram (256
phys RAM in machine). Off to swap we go with molasses in our boots.

window #2:
cd /usr/src; while true; do tar cf - linux/ | (cd /tmp; tar -xvf -); done

window #3:
cd /tmp; while true; do rm -rvf linux/; done

window #4 ('doctor.. it hurts when I do this' 'so don't do it!')
cd /tmp/linux; rm -rf <directory being untarred to and at the same time
rm'ed by #2 and #3, in my case linux/include>

To summarize, window #2 is (was) currently tarring/untarring the
linux/include directory while #3 was in a recursive delete of same. I
changed directory into the copy of the linux tree (note: the CWD of #3 is
*above* the copy of the linux tree, but #4 is inside the copy of the tree)
and began a second rm -r of the same directory as was getting toasted and
written already.

Here is a 'snapshot' of the last living output of all three windows:

#2 the tar/untar (cwd /usr/src and /tmp):
linux/include/linux/coda_linux.h
linux/include/linux/openpic.h
linux/include/linux/prctl.h
linux/include/linux/ftape-header-segment.h
linux/include/linux/raid1.h

#3 the first rm -r (cwd /tmp):
linux/include/linux/elf.h
linux/include/linux/linkage.h
linux/include/linux/malloc.h
linux/include/linux/mc146818rtc.h
linux/include/linux/zorro.h
(note: zorro.h is the last file in an alphabetic listing... probably
doesn't matter)

#4 the second rm (cwd /tmp/linux):
[root@spike linux]# rm -rf include
(it didn't get anywhere)

Bang. All three are frozen.

Sorry for the bad news, if there anything more I can provide, let me know.

David

-- 
/==============================\
| David Mansfield              |
| david@cobite.com             |
\==============================/




-
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/