> and another one - i have here a machine (dual p6 333, 128m, ide)
> with an rootfs on ext2 and the same on xfs plus my homedir nfs
> mounted from a fileserver - then i run the xfs root fs and will
> start netscape i get a core dump with the following backtrace
>
> (gdb) bt
> #0 0x4024dc55 in __getdents (fd=16, buf=0x90ea030 "", nbytes=65536)
> at ../sysdeps/unix/sysv/linux/getdents.c:84
> #1 0x4024d82c in __readdir (dirp=0x90ea000) at ../sysdeps/unix/readdir.c:5
7
> ...
Rather depends on what fd 16 is in netscape at this point. It is probably
a directory in xfs given the nbytes value. Is it possible to find this
out, an strace output of netscape could be huge depending on how far it
has got before it crashes. I suppose another way would be to start netscape
in gdb, and when it crashes go look at the /proc directory for the process,
doing an ls -l in the /proc/pid/fd subdirectory will give us what is at this
descriptor.
For example mine looks like this right now:
lr-x------ 1 lord network 64 Sep 8 10:27 0 -> /dev/null
l-wx------ 1 lord network 64 Sep 8 10:27 1 -> pipe:[2732573]
l-wx------ 1 lord network 64 Sep 8 10:27 10 -> pipe:[2732425]
lr-x------ 1 lord network 64 Sep 8 10:27 11 -> pipe:[2732460]
l-wx------ 1 lord network 64 Sep 8 10:27 12 -> pipe:[2732460]
lrwx------ 1 lord network 64 Sep 8 10:27 13 ->
/home/tulip15/lord/.netscape/cert7.db
lrwx------ 1 lord network 64 Sep 8 10:27 14 ->
/home/tulip15/lord/.netscape/key3.db
lrwx------ 1 lord network 64 Sep 8 10:27 15 ->
/var/tmp/lord-cache/index.db
lrwx------ 1 lord network 64 Sep 8 10:27 16 ->
/home/tulip15/lord/.netscape/history.dat
lr-x------ 1 lord network 64 Sep 8 10:27 17 -> pipe:[2732573]
l-wx------ 1 lord network 64 Sep 8 10:27 18 -> pipe:[2732573]
lrwx------ 1 lord network 64 Sep 8 10:27 19 -> /dev/tty1
l-wx------ 1 lord network 64 Sep 8 10:27 2 -> pipe:[2732573]
lrwx------ 1 lord network 64 Sep 8 10:27 20 -> /dev/tty1
lrwx------ 1 lord network 64 Sep 8 10:27 21 ->
/home/tulip15/lord/.netscape/xover-cache/host-fido.engr.sgi.com/hostinfo.dat
lr-x------ 1 lord network 64 Sep 8 10:27 22 -> pipe:[2732703]
lr-x------ 1 lord network 64 Sep 8 10:27 23 -> pipe:[2868891]
l-wx------ 1 lord network 64 Sep 8 10:27 24 -> pipe:[2732703]
lrwx------ 1 lord network 64 Sep 8 10:27 25 -> socket:[3123735]
l-wx------ 1 lord network 64 Sep 8 10:27 26 -> pipe:[2868891]
lrwx------ 1 lord network 64 Sep 8 10:27 27 -> socket:[3284401]
lr-x------ 1 lord network 64 Sep 8 10:27 29 ->
/usr/lib/netscape/java/classes/java40.jar
lrwx------ 1 lord network 64 Sep 8 10:27 3 -> /dev/zero
lr-x------ 1 lord network 64 Sep 8 10:27 30 -> pipe:[2864534]
l-wx------ 1 lord network 64 Sep 8 10:27 31 -> pipe:[2864534]
lrwx------ 1 lord network 64 Sep 8 10:27 33 -> socket:[3284399]
lrwx------ 1 lord network 64 Sep 8 10:27 34 -> socket:[3284400]
l-wx------ 1 lord network 64 Sep 8 10:27 4 -> pipe:[2732417]
lr-x------ 1 lord network 64 Sep 8 10:27 5 -> pipe:[2732418]
lr-x------ 1 lord network 64 Sep 8 10:27 6 -> pipe:[2732419]
l-wx------ 1 lord network 64 Sep 8 10:27 7 -> pipe:[2732419]
lrwx------ 1 lord network 64 Sep 8 10:27 8 -> socket:[2732423]
lr-x------ 1 lord network 64 Sep 8 10:27 9 -> pipe:[2732425]
>From your description, I do not think XFS and NFS are stepping on each other,
since they are never both involved in the same filesystem.
Steve
>
> if i do it with the ext2 root fs (and the _same_ kernel) it
> works fine ... maybe the xfs and the nfs code are walking on
> eachother foot anywhere here?
>
> some more side infos: nfs server is 2.2.17pre17 with dhiggens/trond
> patches, the ext2 system is ext2 only and the xfs one xfs only,
> also it is not the netscape bin - running the bin from the
> ext2 fs on the xfs root system crashes too, ah - and the
> xfs code is fresh from now but the problem is here for
> at least some weeks (as long as i know) and is
> absolutely reproducable ...
>
> any ideas?
>
> t
>
> --
> thomas.graichen@xxxxxxxxxxxxx
> technical director innominate AG
> clustering & security networking people
> tel: +49.30.308806-13 fax: -77 http://innominate.de
|