Re: initrd oops; still happens with 2.4.5ac2

[Masaru Kawashima]

On Mon, 28 May 2001 10:25:51 +0300
Ville Herva <[e-mail]> wrote:
> The oops call trace seems to be the same as in 
> 
> http://marc.theaimsgroup.com/?l=linux-kernel&m=99079948404775&w=2
> 
> Any ideas?

Did you try the patch posted by Go Taniguchi <[e-mail]>?
Following is the copy of his message and the patch itself.

--
Masaru Kawashima

=====================================================================
From: Go Taniguchi <[e-mail]>
To: LKML <[e-mail]>
Subject: [PATCH]initrd unmount problem
Date: 	Mon, 28 May 2001 13:26:20 +0900
Sender: [e-mail]
Organization: Turbolinx Inc.
X-Mailer: Mozilla 4.75 [ja] (X11; U; Linux 2.2.18-10 i686)

Hi,

It seems that ioctl_by_bdev() in fs/block_dev.c has a problem.
When initrd is unmounted it can cause OOPS. 
This problem occurs in recent ac patches.
May be vanilla too.

change_root() in fs/super.c calls ioctl_by_bdev() in
fs/block_dev.c which does not set inode_fake.i_bdev.

But ioctl of ramdisk (rd_ioctl() in rd.c) accesses to
i_bdev->bd_openers of the inode and which causes OOPS.

I attach the patch.

- GO!

--- linux/fs/block_dev.c.orig	Mon May 28 12:40:12 2001
+++ linux/fs/block_dev.c	Mon May 28 12:40:12 2001
@@ -602,6 +602,7 @@
 	if (!bdev->bd_op->ioctl)
 		return -EINVAL;
 	inode_fake.i_rdev=rdev;
+	inode_fake.i_bdev=bdev;
 	init_waitqueue_head(&inode_fake.i_wait);
 	set_fs(KERNEL_DS);
 	res = bdev->bd_op->ioctl(&inode_fake, NULL, cmd, arg);