When accessing remote filesystems, it sometimes happens that the reading
process gets locked. I think that it is locked inside the kernel, because
it is impossible to kill -9 it.
It happened to me twice when I used the update command in the debian
deselect package. I did tons of other accesses on this volume, but none of
them triggered this bug....
The system is still running, but this console is locked and I cannot
unmount this partition because of the open file. Here is what I can see in
another console :
8:31 [root:2] Studio:~# lsof /mnt/debian
COMMAND PID USER FD TYPE DEVICE SIZE/OFF INODE NAME
zcat 2685 root 0r REG 0, 2 158525 556423252
/mnt/debian (bouddha:/debian)
[the inode number seems much too big.... is this the reason for the bug
or the consequence of it ?]
This nfs partition is mounted at boot-time with this entry in my fstab :
bouddha:/debian /mnt/debian nfs defaults,nodev,noexec,nosuid,rsize=8192,ro
0 2
8:55 [root:2] Studio:~# kill -9 2685
8:55 [root:2] Studio:~#
the zcat process is still running, and in my locked console, I still see
the dselect message :
Uncompressing /mnt/debian/stable/binary-i386/Packages.gz ...
I feel this is strange, because all the accesses I made to this partition
with the usual shell tools never locked anything, but I got the lock
twice with dselect and always at the same place. Still I believe (correct
me if I'm wrong) that the problem cannot be in user space, because then
the kill -9 should work.
I am running the 2.0.27 kernel and bouddha is running Linux 2.0.22
If you want any more precisions, mail me. The process is still running.
Michel "Walken" LESPINASSE - Student at Ecole Centrale Paris (France)
www Email : walken@via.ecp.fr
(o o) VideoLan project : http://videolan.via.ecp.fr/
------oOO--(_)--OOo-------------------------------------------------------
Yow ! 1135 KB/s remote host TCP bandwidth over 10Mb/s ethernet. Beat that!