Re: very bad behavior on overquota writes

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Nov 21, 2012 at 05:54:30PM +0000, Roger Brooks wrote:
> SunOS 4 used to have some strange behaviour with overquota writes to
> an NFS filesystem.  It's a long time ago, but as far as I remember,
> the overquota writes would appear to succeed and then the close
> would fail with EDQUOT.  There's an awful lot of code which doesn't
> bother check for errors on close (and even if you did, it was
> difficult to know how to handle the error)!

Yes, the linux client seems to behave the same. I guess it's because it caches
everything in memory, and does the write asynchronously, or at close()/fflush()
time. But it's broken anyway: close gets and error, but it's EIO not EDQUOT.

A NetBSD client reports the EDQUOT at write() time, I guess writes happens
more synchrounously here.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--