Also pass "-ppax" in an attempt to get improved performance,
but this doesn't really work in the typical case, because the
install ramdisk image contains a version of pax(1) that doesn't
support the -M flag. (MAKEDEV detects the failure and falls back
to using mknod(8) as usual.)
Removes a load of code that is easier done as a script.
Whenever init exits, use a different error number. Since it is difficult
for init to output messages when /dev/console is absent, having the exit
code reported by the kernel panic gives a chance of identifying the fault.
Also exit if MAKEDEV fails to create /dev/console.
MAKEDEV, MAKEDEV.subr and MAKEDEV.local in the set of files that are
copied if they exist.
Now booting an INSTALL kernel with an embedded ramdisk correctly creates
/dev again, fixing a problem introduced when MAKEDEV.subr was created a
few days ago.
1. don't print the program name in the message [syslog does it]
2. no extraneous \n [syslog takes care of it]
3. quote strings as `%s' [we are using syslog]
4. use %m instead of error = %d
How it works:
- after successful execution of /etc/rc, check the value of "init.root"
sysctl node, if it's different than "/", chroot() into its value and run
/etc/rc inside the chroot(),
- in single-user, return back to the original / file system.
Allows running with / file system on e.g., cgd(4), vnd(4) or ccd(4) volumes.
Idea first discussed with Matt Thomas, implemented by Jachym Holecek <freza
(at) liberouter.org> with some nitpicks by me. Successfully used by me for
almost a year with / on a cgd(4) volume (for more information about the setup
check ftp://ftp.NetBSD.org/pub/NetBSD/misc/salo/init-chroot/ ).
If /dev/MAKEDEV doesn't exist then try to run /etc/MAKEDEV instead.
Use the (new) '-n inodes' option to mount_mfs and base the size on the number
of inodes and the size of the MAKEDEV script.
number of the console from sysctl("machdep.console_device"), rather
than making in optional on the (unused) #ifdef DEBUG.
Certain platforms do not have the console at makedev(0,0) (including
many arm32 and some sh3 platforms), so the "mfs MAKEDEV" hack would
have failed on those.
Noted in private discussion with Quentin Barnes.
SIGKILL, go into an infinite loop (like in the runcom SIGTERM case), on the
assumption that reboot(8) is in the middle of taking the system down.
This fixes `panic: init died' from sysinst.
- space problems in casts
- don't cast 0, use NULL
- don't have long lines
- don't cast malloc
- use warn/warnx instead of fprintf
- fix bug where a failed mount_mfs would cause the program to continue.
- pass lint
