On Wednesday, November 14, 2012 12:30:58 PM Joe Kowalski wrote: > On Saturday, November 10, 2012 04:35:33 PM Philip Guenther wrote: > > Hmm, it works for me, using an exports of > > > > /usr/src -mapall=guenther 127.0.0.1 > > /usr/obj -mapall=guenther 127.0.0.1 > > > > and switching the -mapall on /usr/src from "guenther" to "otheruser", > > pkill -HUP mountd, then testing by creating a file in a mode 777 > > directory on the NFS mount from 127.0.0.1. Various other switches and > > tests show that it doesn't seem to be related to the sorting of the > > directory names. > > > > So: need more data. Run mountd yourself with the -d option and snag > > the *complete* output from start, making changes to /etc/exports and > > HUP'ing it, etc. > > I've been working with Tim on this. >
I looked at this some more today. Previously it seemed that the only way to change the user that mapall was configured to was to reboot the system. However today I found that by unmounting the physical partition on the server on which the export resides, and then remounting it, restarting the nfs related services, and mounting the nfs exports on the client, then the new user that mapall points to goes into effect. It seems to me that for some reason, when mountd initially applies the mapall function, the uid it applies gets stuck somewhere (the vfs layer?) and isn't cleared until the physical filesystem is unmounted and remounted. Joe Kowalski
