Ocfs2 devel - Apr 2004 - About 2.6 porting bug 58 "system halt when unmount a volumeimmediately after read/write"

Yes, seems if umount only, it will print call trace:=20
"kernel BUG at /usr/src/linux-2.6.3/fs/buffer.c:2660!"
and the system is not halt,If then follows rmmod at once, there will be kernel
panic and system halt.
So problem has ready happened when unmounting.
> -----Original Message-----
> From: ocfs2-devel-bounces@oss.oracle.com=20
> [mailto:ocfs2-devel-bounces@oss.oracle.com] On Behalf Of Zhang, Sonic
> Sent: 2004=C4=EA4=D4=C222=C8=D5 15:59
> To: Ocfs2-Devel
> Subject: [Ocfs2-devel] About 2.6 porting bug 58 "system halt=20
> when unmount a volumeimmediately after read/write"
>=20
>=20
> Hi,
>=20
> 	We find this bug since svn version 847, in which a=20
> patch against kernel 2.4 is merged in.=20
>=20
> 	This 2.6 porting bug occurs only when unmount a volume=20
> immediately after read/write and before the=20
> journal_commit_cache() is called in journal_commit_thread.=20
>=20
> 	Kernel reports:
> 	VFS: Busy inodes after unmount. Self-destruct in 5=20
> seconds. Have a nice day...
>=20
> 	It has nothing to do with remod.
>=20
> 	Any suggestion?
>=20
>=20
> *********************************************
> Sonic Zhang
> Software Engineer
> Intel China Software Lab
> Tel: (086)021-52574545-1667
> iNet: 752-1667
> *********************************************=20
>=20
> _______________________________________________
> Ocfs2-devel mailing list
> Ocfs2-devel@oss.oracle.com=20
> http://oss.oracle.com/mailman/listinfo/ocfs2-devel
>=20

Hi,

	I think I find some clues of this bug.

	Since svn version 847, function ocfs_release_oin() is removed from the file
oin.c. In function ocfs_file_release() of the former version, the inode usage
count i_count is decreased in ocfs_release_oin() each time dentry->d_count is
0. But, in the latest version, it won't be decreased until the journal lock
is released in journal_commit_cache().

	I remember that this change is to avoid access empty inode pointer reserved in
journal lock. Am I right?

	In kernel 2.6 VFS routine sys_umount()->...->generic_shutdown_super(), if
the inode usage count i_count is not 0, error information is reported. Then, the
system will halt in routine kill_bdev()->truncate_inode_pages().

	I am looking for a good solution for this bug. Do you have some advice?

	Thank you.


*********************************************
Sonic Zhang
Software Engineer
Intel China Software Lab
Tel: (086)021-52574545-1667
iNet: 752-1667
*********************************************=20

-----Original Message-----
From: ocfs2-devel-bounces@oss.oracle.com
[mailto:ocfs2-devel-bounces@oss.oracle.com] On Behalf Of Zhang, Sonic
Sent: 2004=C4=EA4=D4=C222=C8=D5 15:59
To: Ocfs2-Devel
Subject: [Ocfs2-devel] About 2.6 porting bug 58 "system halt when unmount a
volumeimmediately after read/write"

Hi,

	We find this bug since svn version 847, in which a patch against
kernel 2.4 is merged in.=20

	This 2.6 porting bug occurs only when unmount a volume
immediately after read/write and before the journal_commit_cache() is
called in journal_commit_thread.=20

	Kernel reports:
	VFS: Busy inodes after unmount. Self-destruct in 5 seconds. Have
a nice day...

	It has nothing to do with remod.

	Any suggestion?


*********************************************
Sonic Zhang
Software Engineer
Intel China Software Lab
Tel: (086)021-52574545-1667
iNet: 752-1667
*********************************************=20

_______________________________________________
Ocfs2-devel mailing list
Ocfs2-devel@oss.oracle.com
http://oss.oracle.com/mailman/listinfo/ocfs2-devel