hi,

i've got a strange phenomenon and after some discussion on the
#kernelnewbies IRC channel, baldrick (duncan sands) suggested i reported
it here.

i got almost 1000 sendmail processes waiting for a read lock on
/etc/mail/aliases.db, but didn't find any process holding the write
lock. (the processes piled up in some 20 hours.)
the process reported in /proc/locks as holding the write lock doesn't
show up in ps(1) output or in /proc/*.
there was a sendmail process with this PID (normal delivery) an hour
before the pileup started, but at the time when it started the PIDs
hadn't reached the number range again.

i've got no idea what should have written the file (why else a write
lock?), my last change to it was a week ago...

anyway, i `fuser -k`'ed all processes and did a `cat /proc/locks` again.
this time it segfaulted! (it went fine before the `fuser -k`.)

the ksymoops(1) output is attached.

the kernel was built by me 9 days ago from the 2.4.21-4 debian kernel
source package.

it was the first time i experienced such a behaviour.

rj