On Mon, 2009-05-25 at 22:58 -0400, David Dillow wrote:
> On Mon, 2009-05-25 at 22:44 -0400, Jeremy Katz wrote:
> > On Monday, May 25 2009, David Dillow said:
> > > dash gets upset when it cannot open /dev/tty, and we lose stderr. Even
> > > if we make /dev/tty once we're in the initramfs, it will still be too
> > > late for the inital dash shell running the init script. So, make a small
> > > set of nodes in the dracut script while building the directory structure
> > > of the initramfs.
> >
> > Creating the nodes in the dracut script means that you can't create the
> > initramfs as non-root, though. :(
>
> Bummer, that. And while it makes the initial dash able to give us
> stderr, it doesn't fix the emergency shell. Of course, I noticed this
> after hitting send... so bad patch all around. :(
I'm a dope, initial dash has always had stderr.
This ugliness points /dev/tty at the actual device of /dev/console while
running the emergency shell, then puts it back when done so if we do
manage to mount things, /dev will have the right devices when
switch_root moves it.
I'm not sure if its worth it, but at least I have a prompt and stderr in
the emergency shell now, even after udev runs.
diff --git a/modules.d/99base/init b/modules.d/99base/init
index 5ee897e..4e2d73f 100755
--- a/modules.d/99base/init
+++ b/modules.d/99base/init
@@ -11,7 +11,13 @@ emergency_shell()
echo ; echo
echo "Bug in initramfs /init detected. Dropping to a shell. Good luck!"
echo
+ # Point /dev/tty at the device for /dev/console as we don't have a
+ # controlling shell here and dash won't give us stderr or a prompt
+ # w/o a working /dev/tty
+ rm -f /dev/tty; mknod /dev/tty c 5 1
sh -i
+ # Point it back to the real device
+ rm -f /dev/tty; mknod /dev/tty c 5 0
}
source_all() {
--
To unsubscribe from this list: send the line "unsubscribe initramfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
