linprocfs and linux_base port upgrade

[Boris Samorodov]

On Wed, 22 Feb 2006 11:16:58 +0100 Alexander Leidinger wrote:
> Kris Kennaway <kris at obsecurity.org> wrote:

> Regarding the question of the OP: changing "ro" to "ro,noauto" has a similar
> effect in the race-case.

> > On Wed, Feb 22, 2006 at 11:01:57AM +0800, Mars G. Miro wrote:
> >
> >> > No, it's because it tried and failed to umount linprocfs.  Presumably
> >> > you didnt have linprocfs mounted in your jail, but some packages
> >>
> >> Well mounting/unmounting stuff inside the jail is a pain. I now recall
> >> I had to mount linprocfs from the host to the jail, thus I was able to
> >> build them. Hrm, perhaps its time for jail_<jailname>_linprocfs_enable
> >> (as with devfs inside a jail) ...
> >
> > I think you're missing my point: if you don't have linprocfs mounted,
> > ports like jdk will fail.  If you do have it mounted, ports like
> > linux_base will fail [because they need to umount it and remount it].
> >
> > The latter should be fixed so that you can consistently set up a jail
> > and have it work in both cases.

> Perhaps (completely untested):
> @exec mkdir -p %D/proc || true

Imo "mkdir -p" doesn't generate errors if the target directory exists.
Do we need "|| true"?

> @dirrmtry proc
> @unexec [ ! -d %D/proc ] || echo "+++ Please unmount linprocfs and remove
> %D/proc by hand!"
> and not touching linprocfs at all.

> Kris, would this work on pointyhat?


WBR
-- 
Boris B. Samorodov, Research Engineer
InPharmTech Co,     http://www.ipt.ru
Telephone & Internet Service Provider