Messages in this thread | | | Date | Tue, 24 Aug 1999 14:48:02 +0200 | From | Werner Almesberger <> | Subject | Re: RPC: sendmsg returned error 101 |
| |
Bill Rugolsky Jr. wrote: > Was this issue (i.e., kernel RPC generating ENETUNREACH) ever resolved > satisfactorily? I can reproduce it consistently on two different machines.
I've seen this too, on RedHat 6.0. About the following sequence of actions occurred: - shutdown - tried to umount NFS. Failed because a process was running off it (cwd and executable on NFS, possibly also some open files) - network was shut down - something made the system want to use NFS (*) - a looong sequence of 101s - me timing out, hitting reset
(*) possible causes: page-in, process writes good-bye message, 2nd umount attempt, etc. Didn't look at what exactly caused it.
This is somewhat tricky to fix: the init scripts should have killed the process sitting on NFS before trying to umount. However, killing all processes before umount may kill a process that's keeping the network alive (e.g. pppd, atmarpd).
Finding out what can be killed and what can't, may be tricky, because there may be ordering constraints. Example: three NFS mounts: /ether, /ppp, /atm, each using the corresponding network. The demons are run as follows: /ether/pppd and /ppp/atmarpd. There be also an /ether/food which doesn't do anything network-related.
Now we could kill them in this order: food, atmarpd, pppd. But we may not be able to kill them in this order: food, pppd, atmarpd. (Note: this still kills the "well-known network demons" after the other things, so just maintaining a list of them doesn't help.)
A practical solution may be simply to forcibly unmount all stuck file systems. In any case it's better than forcing the user to press the reset button.
- Werner
-- _________________________________________________________________________ / Werner Almesberger, ICA, EPFL, CH werner.almesberger@ica.epfl.ch / /_IN_R_131__Tel_+41_21_693_6621__Fax_+41_21_693_6610_____________________/
- 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/
| |