Problem with mm1 patch for 2.5.51

Problem with mm1 patch for 2.5.51

[Aniruddha M Marathe]

Hi,
I applied mm1 patch to kernel 2.5.51 and I ran LM bench to test its performance.
Here are the errors that I obtained.

EXT3-fs error (device ide0(3,6)) in start_transaction: Journal has aborted

ext3_free_blocks: aborting transaction: Journal has aborted in __ext3_journal_get_undo_access<2>

EXT3-fs error (device ide0(3,6)) in ext3_free_blocks: Journal has aborted

ext3_free_blocks: aborting transaction: Journal has aborted in __ext3_journal_get_undo_access<2>

EXT3-fs error (device ide0(3,6)) in ext3_free_blocks: Journal has aborted

ext3_reserve_inode_write: aborting transaction: Journal has aborted in __ext3_journal_get_write_access<2>

EXT3-fs error (device ide0(3,6)) in ext3_reserve_inode_write: Journal has aborted

EXT3-fs error (device ide0(3,6)) in ext3_truncate: Journal has aborted

ext3_reserve_inode_write: aborting transaction: Journal has aborted in __ext3_journal_get_write_access<2>

EXT3-fs error (device ide0(3,6)) in ext3_reserve_inode_write: Journal has aborted

EXT3-fs error (device ide0(3,6)) in ext3_orphan_del: Journal has aborted

ext3_reserve_inode_write: aborting transaction: Journal has aborted in __ext3_journal_get_write_access<2>

EXT3-fs error (device ide0(3,6)) in ext3_reserve_inode_write: Journal has aborted

EXT3-fs error (device ide0(3,6)) in ext3_delete_inode: Journal has aborted

I had observered similar errors for mm2 patch for kernel 2.5.50. that patch was later removed from the site. It seems that the problem still persists. are changes in fs/ext3/balloc.c and fs/ext3/inode.c responsible for this ?

Regards,
--------------------------------------------------------------
Aniruddha Marathe
Systems Engineer,
4th floor, WIPRO technologies,
53/1, Hosur road,
Madivala,
Bangalore - 560068
Karnataka, India
Phone: +91-80-5502001 extension 5092
E-mail: aniruddha.marathe@wipro.com
---------------------------------------------------------------

Re: Problem with mm1 patch for 2.5.51

[Andrew Morton]

Aniruddha M Marathe wrote:
> 
> Hi,
> I applied mm1 patch to kernel 2.5.51 and I ran LM bench to test its performance.
> Here are the errors that I obtained.
> 
> EXT3-fs error (device ide0(3,6)) in start_transaction: Journal has aborted

An off-by-one was gratuitously added to ext3_free_blocks


--- 25/fs/ext3/balloc.c~dud-patch	Tue Dec 10 21:07:20 2002
+++ 25-akpm/fs/ext3/balloc.c	Tue Dec 10 21:07:27 2002
@@ -122,7 +122,7 @@ void ext3_free_blocks (handle_t *handle,
 	es = EXT3_SB(sb)->s_es;
 	if (block < le32_to_cpu(es->s_first_data_block) ||
 	    block + count < block ||
-	    block + count >= le32_to_cpu(es->s_blocks_count)) {
+	    block + count > le32_to_cpu(es->s_blocks_count)) {
 		ext3_error (sb, "ext3_free_blocks",
 			    "Freeing blocks not in datazone - "
 			    "block = %lu, count = %lu", block, count);

Re: Problem with mm1 patch for 2.5.51

[Con Kolivas]

Aniruddha M Marathe said:
>EXT3-fs error (device ide0(3,6)) in ext3_delete_inode: Journal has aborted
etc..


I suspect I suffered a similar fate with the osdl test box. While I was away
letting it run a benchmark in smp mode the filesystem had remounted read only. I
tried rebooting to make some sense of what had happened but was unable to start
the machine with any kernel. I've asked the osdl people to have a look at the
box for me.

Previously a run in uniprocessor mode ran flawlessly.

Con

Re: Problem with mm1 patch for 2.5.51

[Robert Love]

On Wed, 2002-12-11 at 00:08, Con Kolivas wrote:

> I suspect I suffered a similar fate with the osdl test box. While I was away
> letting it run a benchmark in smp mode the filesystem had remounted read only. I
> tried rebooting to make some sense of what had happened but was unable to start
> the machine with any kernel. I've asked the osdl people to have a look at the
> box for me.
> 
> Previously a run in uniprocessor mode ran flawlessly.

This looks like a fix to the ext3 bug I posted last week, which was on
UP.

It comes and goes... I am in 2.5-mm with no problems now, but it was
killing me last week.  *shrug*

Here's to hoping this fixes it...

	Robert Love