Per-User QUOTA's vs blocksize

[Scott Mitchell]

On Tue, Mar 09, 2004 at 02:05:54PM +0000, Matthew Seaman wrote:
> On Mon, Mar 08, 2004 at 06:48:51PM +0000, Stacey Roberts wrote:
> 
> > "You may limit allocations based on disk space (block quotas)"
> > 
> > What exactly is the size of a block?
> 
> As the quota system uses the term, a block is 512b -- this unit is
> also called a sector in some situations, but it's basically the same
> thing.  It reflects the size of the underlying sector structures
> within the filesystem.

True, but the term is used inappropriately by the quota tools and their
documentation - these appear to work exclusively in KB units for their
(ahem) 'block' quotas, at least on recent versions on FreeBSD.

See, for example, rev. 1.18 of edquota.c and the associated PR:
http://www.freebsd.org/cgi/cvsweb.cgi/src/usr.sbin/edquota/edquota.c?rev=1.18&content-type=text/x-cvsweb-markup
http://www.freebsd.org/cgi/query-pr.cgi?pr=41936

There's a lot of explicit [some number of bytes]/1024 calculations in the
various quota tools.

I have empirical evidence of this too - I set up 'block' quotas for my
users many moons ago, assuming 512 byte blocks.  I was then quite surprised
to run out of disk space last weekend, with du(1) showing several users
massively over quota, but repquota(8) saying all was well.  Halving all the
quota settings keeps everything in agreement.

The documentation is certainly misleading on this point.  I'll file a PR,
unless anyone cares to beat me to it...

	Scott

-- 
===========================================================================
Scott Mitchell           | PGP Key ID | "Eagles may soar, but weasels
Cambridge, England       | 0x54B171B9 |  don't get sucked into jet engines"
scott at fishballoon.org | 0xAA775B8B |      -- Anon