Patchwork ext4: fix ext4_flush_completed_IO wait semantics

login
register
mail settings
Submitter Dmitri Monakho
Date Oct. 5, 2012, 1:53 p.m.
Message ID <877gr5auu7.fsf@openvz.org>
Download mbox | patch
Permalink /patch/189482/
State Superseded
Headers show

Comments

Dmitri Monakho - Oct. 5, 2012, 1:53 p.m.
On Fri, 5 Oct 2012 09:28:53 -0400, Theodore Ts'o <tytso@mit.edu> wrote:
> On Fri, Oct 05, 2012 at 05:01:30PM +0400, Dmitry Monakhov wrote:
> > > This WARN_ON is triggering on the truncate path...
> > Yeap, this is false positive one. We skip i_mutex on ext4_evict_inode
> > This is strange xfsstress 269'th should caught that for me.
> > I'll try to prepare workaround ASAP.
> 
> This is the patch which I'm currently testing.  If it passes I'll fold
> it into your patch.  Anyone see any problems with it?
> 
>    	     	     	       	    	     - Ted
> 
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index f18e786..cd171dd 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -238,8 +238,10 @@ void ext4_evict_inode(struct inode *inode)
>  	 * protection against it
>  	 */
>  	sb_start_intwrite(inode->i_sb);
> +	mutex_lock(&inode->i_mutex);
>  	handle = ext4_journal_start(inode, ext4_blocks_for_truncate(inode)+3);
>  	if (IS_ERR(handle)) {
> +		mutex_unlock(&inode->i_mutex);
I mutex is not needed for journal start.
>  		ext4_std_error(inode->i_sb, PTR_ERR(handle));
>  		/*
>  		 * If we're going to skip the normal cleanup, we still need to
> @@ -256,12 +258,14 @@ void ext4_evict_inode(struct inode *inode)
>  	inode->i_size = 0;
>  	err = ext4_mark_inode_dirty(handle, inode);
>  	if (err) {
> +		mutex_unlock(&inode->i_mutex);
>  		ext4_warning(inode->i_sb,
>  			     "couldn't mark inode dirty (err %d)", err);
>  		goto stop_handle;
>  	}
>  	if (inode->i_blocks)
>  		ext4_truncate(inode);
> +	mutex_unlock(&inode->i_mutex);
Jiaying Zhang removed i_mutex from ext4_evict_inode here 8c0bec2151
because of false positive lockdep complains.
Let's just learn warning about evict_inode
Theodore Ts'o - Oct. 5, 2012, 2:20 p.m.
On Fri, Oct 05, 2012 at 05:53:36PM +0400, Dmitry Monakhov wrote:
> Jiaying Zhang removed i_mutex from ext4_evict_inode here 8c0bec2151
> because of false positive lockdep complains.
> Let's just learn warning about evict_inode 

Yes, good point.  I'm starting a testing cycle with your patch.

     	  	      	       	 	 - Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

From edcd1dd6d8c4c17af09d429cc89e370bdc1e737a Mon Sep 17 00:00:00 2001
From: Dmitry Monakhov <dmonakhov@openvz.org>
Date: Fri, 5 Oct 2012 17:40:19 +0400
Subject: [PATCH] ext4: fix false positive warning ext4_evict_inode


Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
---
 fs/ext4/page-io.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index 1f5df21..68e896e 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c
@@ -229,7 +229,8 @@  static void ext4_end_io_work(struct work_struct *work)
 int ext4_flush_unwritten_io(struct inode *inode)
 {
 	int ret;
-	WARN_ON_ONCE(!mutex_is_locked(&inode->i_mutex));
+	WARN_ON_ONCE(!mutex_is_locked(&inode->i_mutex) &&
+		     !(inode->i_state & I_FREEING));
 	ret = ext4_do_flush_completed_IO(inode, NULL);
 	ext4_unwritten_wait(inode);
 	return ret;
-- 
1.7.7.6