Re: md /tmp and async mounts

On Fri, Jun 02, 2006 at 12:35:27AM +0300, Giorgos Keramidas wrote:
> On 2006-06-01 17:06, Kris Kennaway <kris_at_obsecurity.org> wrote:
> > On Thu, Jun 01, 2006 at 03:20:24AM +0300, Giorgos Keramidas wrote:
> >
> > [diff snipped]
> >
> > Thanks for the patch.
> >
> > > This doesn't make 'async' the default for all mdmfs mounts.
> > >
> > > Making async the default for mdmfs-based filesystems would
> > > require a fair bit of changes to mdmfs, which would have to grow
> > > mount-option parsing capabilities for the `optarg' of the -o
> > > option and make sure that async is always added, unless mdmfs is
> > > called explicitly with `noasync' in its -o option.
> > >
> > > Having said this, if this is considered a change we should make,
> > > I can prepare a patch for mdmfs too.
> >
> > I think it should be the default.  I think we should also avoid using
> > -M (malloc backing for md) in favour of swap backing because of the
> > panic problems - this is too easy to trigger by accident and it makes
> > FreeBSD look bad since it looks like a bug.
> 
> Ok, I'll prepare a patch that enables async and disables -M.  We should
> also document the fact that tmpmfs="YES" and varmfs="YES" in rc.conf may
> require the presence of at least one swap device by default, and point
> the users to -M with a warning if they run FreeBSD without a swap device
> but still want to use tmpmfs or varmfs :)

Actually I thought about this and don't think it's the case.  swap
backing doesn't use swap unless you have memory pressure, and if you
have memory pressure and no swap your system is going to explode
regardless.

Kris