Re: tmpfs sizing problems (i386, 4GB mem)

[Antti Kantee <pooka%NetBSD.org@localhost>]

On Mon, Jun 28, 2010 at 04:17:45PM +0100, Mindaugas Rasiukevicius wrote:
> Chris Ross <cross+netbsd%distal.com@localhost> wrote:
> > 
> >    On a machine running 5.99.31, I'm having a problem with my tmpfs 
> > /tmp.  I have it configured to be 600m in size, but it's only about 
> > 200m.  After trying to figure out if it was an argument parsing issue, I 
> > have verified that if I just call "mount_tmpfs tmpfs /mnt" I get the 
> > following (df -k output)
> 
> Investigating this problem.  It is due to "dynamic" calculation of the
> limit from free memory.

On a tangent, I recently wanted to use large sparse files on /tmp
(which is 99.9% of the time one of {ffs,mfs,tmpfs}).  Since my host
doesn't have 8TB of memory, I ended up requiring a local mount of
mfs...

From a quick look it seemed the only reason tmpfs doesn't support
large sparse files is that the accounting for the pages is done
already when the file is extended instead of when the pages are
faulted into.  I don't know if it makes sense to stuff a callback
like GOP_ALLOC into aobj only for this reason, but if you're working
in the area, maybe you can have a look.

thanks,
  antti