After my first summary, I received a very interesting answer from
"William H. Magill" <magill_at_isc.upenn.edu>:
> I just ran this exact problem through dial-a-prayer (Digital Support).
> Basically, the performance of vquota on "broke" in 3.2c, and was fixed in
> 3.2g... sort of.
>
> In 3.2 before c, vquoton had to be explicity run to start the quotas.
> Under 3.2g the code was "broken" so that it was not necessary to do
> anything more than to add the "userquota,groupquota" entries (not rq alone)
> to fstab.
>
> Under 3.2G they determined that UFS and AdvFS quotas were "different" and
> made AdvFS work like UFS again - ie rquire the explicit use of "vquotaon"
> in startup.... however, they neglected to include such a startup script in
> /sbin/initd.d - there was only a script for the UFS quota starupt.
> The "fix" under 3.2g is to add the startup script (simply take the
> /sbin/init.d/quota script; change all references from "quota" to "vquota"
> and create a new file called /sbin/init.d/vquota, and link it back to
> /sbin/rc3.d/S01vquota.
>
> Under 4.x the entire problem has been fixed. There is only one quota
> script and it detects which kind of quotas are running.
....
I could not believe that, at every boot, user and group quotas on AdvFs
were off, unless I manually do "vquotaon -a" !
I thought the problem appeared only with a new kernel or so...
So beware, if you need AdvFs user/group quotas (at least with DUnix3.2C),
you have to duplicate or modify /sbin/init.d/quota to reflect the use of
AdvFs.
Why doesn't Dec give some patch for this bug ? I'm sure I was not alone
with this problem.
Many thanks, William.
--
_______________________________________________________________________
Magali BERNARD (magali_at_univ-st-etienne.fr)
CRITeR - 23 rue du Dr Paul Michelon - 42023 St-Etienne Cedex 2 - FRANCE
Tel: 04.77.48.50.62
Received on Tue Apr 29 1997 - 14:59:04 NZST