On Tue, Jun 11, 2002 at 02:59:45PM -0400, Matt Zimmerman wrote: > On Tue, Jun 11, 2002 at 01:34:00PM -0500, Steve Langasek wrote: > > On Tue, Jun 11, 2002 at 02:31:32PM -0400, Matt Zimmerman wrote: > > > If a chroot() will always (or almost always) be required after > > > pivot_root(), it should simply be done in pivot_root.c, and the 'chroot' > > > program can stay where it is. > > How do you make the pivot_root command chroot its parent process? And > > even if you can, does it follow that this is always correct? > chroot(8) doesn't chroot its parent process either. If pivot_root were > doing that job in this instance, it would presumably work similarly to > chroot(8). So you're suggesting a syntax such as exec pivot_root /newroot /newroot/mnt [ /sbin/init | <command> ] ? That seems more awkward to me than the current behavior; in particular, there are a few commands I'm running here after pivotting and before exec'ing init. I guess I don't see much advantage to making pivot_root a more complex tool just for this reason. Steve Langasek postmodern programmer
Attachment:
pgp4gzwZpgjsu.pgp
Description: PGP signature