2.4.22 init/do_mounts.c Bug when booting a devfs CD

2.4.22 init/do_mounts.c Bug when booting a devfs CD

[Chris Lingard]

[-- Attachment #1: Type: text/plain, Size: 974 bytes --]

Hope I have the right person;  you may be interested in this.

I boot a CD using initrd; the RAMDISK uses devfs within the linuxrc script
to pivot_root/chroot as in the documentation.  In the documentation it
is suggested that root=/dev/rd/0  is used, if using devfs.

A boot using root=/dev/ram0 works fine, root=/dev/rd/0 shows a bug
in init/do_mounts.c, in void prepare_namespace(void)

The test:
if (initrd_load() && ROOT_DEV != MKDEV(RAMDISK_MAJOR, 0))
is true, and handle_initrd() is wrongly called.  This is a bug, (I assume
that this code is to relinquish the original boot).  The call that is used
is at the end of main.c where linuxrc is called instead of /sbin/init.

This is because rd/0 is not in root_dev_names; a patch is attached to fix
this.

Strangly enough, it still boots, despite linuxrc being called twice :-)

An alternative would be to change the documentation of initrd, so
that only root=/dev/ram0 is mentioned, despite using devfs.

Chris Lingard

[-- Attachment #2: linux-2.4.22-rd0.patch --]
[-- Type: text/x-diff, Size: 531 bytes --]

diff -Naur linux-2.4.22.old/init/do_mounts.c linux-2.4.22/init/do_mounts.c
--- linux-2.4.22.old/init/do_mounts.c	2003-09-10 12:18:55.000000000 +0100
+++ linux-2.4.22/init/do_mounts.c	2003-09-23 12:46:36.000000000 +0100
@@ -227,6 +227,11 @@
 	{ "ataraid/d13p",0x72D0 },
 	{ "ataraid/d14p",0x72E0 },
 	{ "ataraid/d15p",0x72F0 },
+#ifdef CONFIG_BLK_DEV_RAM
+#ifdef CONFIG_BLK_DEV_INITRD
+        { "rd/0",     0x0100 },
+#endif
+#endif
         { "rd/c0d0p",0x3000 },
         { "rd/c0d0p1",0x3001 },
         { "rd/c0d0p2",0x3002 },