retitle 307524 mount failures after /etc/init.d/nfs-kernel-server reload thanks
I've been trying to trace what's going on in the code here. It looks to me like the EPERM is being printed at line 336 of mountd.c, and is generated in the nfsctl system call. In particular, this should invoke either write_getfd() or write_getfs() in fs/nfsd/nfsctl.c (in the kernel). My kernel-fu is weak, but I believe the only way for this to die with -EPERM is for auth_unix_lookup() (whose function is a bit obscure to me; it appears to be some sort of curried authentication function that finds its result in a cache?) to fail or for exp_rootfh to return -EPERM (if it can). I don't see why either should suddenly start failing on the N-plus-1st mount when it worked for the first N mounts. *some time later* Aha: I just found the trigger. If I call the init script with "reload", I start getting the symptoms I described with 100% reproducibility. Doing a restart instead works. I guess I'll stick to restarting after changes for now, since it should be safe on any NFS server, but IMO reload ought to be fixed too. Daniel -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
