Un-killable user processes in Linux?

Question

I use sshfs (fusermount) to connect remote mountpoints.

When I forget to unmount them using...

fusermount -u /REMOTE

...all processes which used the mountpoints hang.
I try to kill them with killall, kill, killproc, nothing works!

How is that possible?
Being unable to kill a userspace process is normally a Microsoft Windows feature ;)

rd@acid:/home/me > pgrep mc
11749
11827
12068

rd@acid:/home/me > su
Password:

<netgod>@acid:/home/me # kill 12068

<netgod>@acid:/home/me # killall mc

<netgod>@acid:/home/me # pgrep mc
11749
11827
12068

Answer 1

The filesystem is indeed usermode, but the system call from the client process which is accessing the filesystem via the kernel, so really they're hung in kernel land.

Have you tried a forced unmount of the filesystem (umount -f /REMOTE)?

Answer 2

kill -9 PID isn't working?

Answer 3

I think that what you might be seeing are those processes hung performing some type of I/O that cannot now be completed. If you look at the the 'state' field given by the ps command, I'm pretty sure you'll see these processes in the 'D' state. If a process is in this state, it cannot be killed in the normal means. The processes will hang around until a reboot is performed.