SEO
vBulletin Search Engine Optimization
| |||||||
| Register | FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
02-22-2008, 11:46 AM
| ||||
| ||||
 Re: mountd(8) diff: request for testing Otto Moerbeek wrote: > It is not about mouning and unmounting the NFS exported file > system. I got a reading disorder apparently. Nice way to completely hang my system BTW. Not even alt-ctrl-del or alt-ctrl-esc worked anymore. > The kernel maintains a list of exported directories. When a > (any) filesystem is mounted mountd gets klicked and does > currently something like: > > tell kernel to unexport all filesystems > process mount list and /etc/exports to compute a new list of exported > tell kernel the new exportlist > > If certain NFS requests arrives on the server after the > unexport, but before the new exportlist is activated, you loose. Fascinating... > A typical test scenario would be > > export /usr/src on server > mount /usr/src on 1 or more clients > start make obj build on clients > unmount and mount filesystems on server. OK, first lets give it a try with the latest snapshot. Unmodified. My exports looks like this: /home/public/han -maproot=han:nfs marsupilami /mnt/mp3 -maproot=han:nfs marsupilami And I started building something in /home/public/han on the client and then I ran. % mount /mnt/mp3 ; umount /mnt/mp3 Most of the time this went OK, one time I got: mount_ffs: /dev/wd1a on /mnt/mp3: Device busy The build in /home/public/han almost always fails with something like: Assembler messages: FATAL: can't create display_usage.o: Permission denied > Additionally, you want to check the output of showmount -e to > check if the list of exported filesystems is still OK on the > server. Looks fine. > You can also run mountd in debug mode to see what is going on, > and compare old behaviour with the new behaviour. Also looks OK. Lets try again with you patch in the next message. # Han      |
Linear Mode
