[ext4:dev 47/50] fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)

[ext4:dev 47/50] fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)

[Fengguang Wu]

Hi Zheng,

FYI, there are new smatch warnings show up in

tree:   git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
head:   b2f1a8e5612e2488a3995ec06b422b44651d088b
commit: 3e1b02687146ddefa8e53c2a5e13d3bd87b16b09 [47/50] ext4: add support for hole punching non-extent-mapped files

+ fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)

vim +1799 fs/ext4/indirect.c

3e1b0268 (Zheng Liu      2012-09-28  1783) 	down_write(&EXT4_I(inode)->i_data_sem);
3e1b0268 (Zheng Liu      2012-09-28  1784) 	ext4_discard_preallocations(inode);
3e1b0268 (Zheng Liu      2012-09-28  1785) 
3e1b0268 (Zheng Liu      2012-09-28  1786) 	err = ext4_free_hole_blocks(handle, inode, offsets1, n1, offsets2, n2);
3e1b0268 (Zheng Liu      2012-09-28  1787) 
3e1b0268 (Zheng Liu      2012-09-28  1788) 	if (IS_SYNC(inode))
3e1b0268 (Zheng Liu      2012-09-28  1789) 		ext4_handle_sync(handle);
3e1b0268 (Zheng Liu      2012-09-28  1790) 
3e1b0268 (Zheng Liu      2012-09-28  1791) 	up_write(&EXT4_I(inode)->i_data_sem);
3e1b0268 (Zheng Liu      2012-09-28  1792) 
3e1b0268 (Zheng Liu      2012-09-28  1793) out:
3e1b0268 (Zheng Liu      2012-09-28  1794) 	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
3e1b0268 (Zheng Liu      2012-09-28  1795) 	ext4_mark_inode_dirty(handle, inode);
3e1b0268 (Zheng Liu      2012-09-28  1796) 	ext4_journal_stop(handle);
3e1b0268 (Zheng Liu      2012-09-28  1797) error:
3e1b0268 (Zheng Liu      2012-09-28  1798) 	mutex_unlock(&inode->i_mutex);
3e1b0268 (Zheng Liu      2012-09-28 @1799) 	return err;
3e1b0268 (Zheng Liu      2012-09-28  1800) }

---
0-DAY kernel build testing backend         Open Source Technology Centre
Fengguang Wu, Yuanhan Liu                              Intel Corporation

Re: [ext4:dev 47/50] fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)

[Zheng Liu]

On Fri, Sep 28, 2012 at 02:07:19PM +0800, Fengguang Wu wrote:
> Hi Zheng,
> 
> FYI, there are new smatch warnings show up in

Hi Fengguang,

It is my fault that it will return directly when it starts a transaction
and gets an error.  I will fix it in next version.  Thanks.

Regards,
Zheng

> 
> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
> head:   b2f1a8e5612e2488a3995ec06b422b44651d088b
> commit: 3e1b02687146ddefa8e53c2a5e13d3bd87b16b09 [47/50] ext4: add support for hole punching non-extent-mapped files
> 
> + fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)
> 
> vim +1799 fs/ext4/indirect.c
> 
> 3e1b0268 (Zheng Liu      2012-09-28  1783) 	down_write(&EXT4_I(inode)->i_data_sem);
> 3e1b0268 (Zheng Liu      2012-09-28  1784) 	ext4_discard_preallocations(inode);
> 3e1b0268 (Zheng Liu      2012-09-28  1785) 
> 3e1b0268 (Zheng Liu      2012-09-28  1786) 	err = ext4_free_hole_blocks(handle, inode, offsets1, n1, offsets2, n2);
> 3e1b0268 (Zheng Liu      2012-09-28  1787) 
> 3e1b0268 (Zheng Liu      2012-09-28  1788) 	if (IS_SYNC(inode))
> 3e1b0268 (Zheng Liu      2012-09-28  1789) 		ext4_handle_sync(handle);
> 3e1b0268 (Zheng Liu      2012-09-28  1790) 
> 3e1b0268 (Zheng Liu      2012-09-28  1791) 	up_write(&EXT4_I(inode)->i_data_sem);
> 3e1b0268 (Zheng Liu      2012-09-28  1792) 
> 3e1b0268 (Zheng Liu      2012-09-28  1793) out:
> 3e1b0268 (Zheng Liu      2012-09-28  1794) 	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
> 3e1b0268 (Zheng Liu      2012-09-28  1795) 	ext4_mark_inode_dirty(handle, inode);
> 3e1b0268 (Zheng Liu      2012-09-28  1796) 	ext4_journal_stop(handle);
> 3e1b0268 (Zheng Liu      2012-09-28  1797) error:
> 3e1b0268 (Zheng Liu      2012-09-28  1798) 	mutex_unlock(&inode->i_mutex);
> 3e1b0268 (Zheng Liu      2012-09-28 @1799) 	return err;
> 3e1b0268 (Zheng Liu      2012-09-28  1800) }

Re: [ext4:dev 47/50] fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)

[Theodore Ts'o]

On Fri, Sep 28, 2012 at 02:07:19PM +0800, Fengguang Wu wrote:
> Hi Zheng,
> 
> FYI, there are new smatch warnings show up in
> 
> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
> head:   b2f1a8e5612e2488a3995ec06b422b44651d088b
> commit: 3e1b02687146ddefa8e53c2a5e13d3bd87b16b09 [47/50] ext4: add support for hole punching non-extent-mapped files

I'm going to be dropping the indirect punch hole patches from the ext4
dev tree, based on the number of xfstests failures that I found in my
overnight testing.

I had hoped the changes were small and localized, and wouldn't cause
problems if the punch hole feature wasn't used, but at this point,
given that many distributions are starting to use the ext4 file system
for legacy ext3 file systems, my judgement is that the risks are too
great for the upcoming merge window; so this will probably need to
wait for the next development cycle.

     	     	  	  				- Ted

Re: [ext4:dev 47/50] fs/ext4/indirect.c:1799 ext4_ind_punch_hole() warn: inconsistent returns mutex:&inode->i_mutex: locked (1710) unlocked (1799)

[Zheng Liu]

On Fri, Sep 28, 2012 at 09:20:55AM -0400, Theodore Ts'o wrote:
> On Fri, Sep 28, 2012 at 02:07:19PM +0800, Fengguang Wu wrote:
> > Hi Zheng,
> > 
> > FYI, there are new smatch warnings show up in
> > 
> > tree:   git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
> > head:   b2f1a8e5612e2488a3995ec06b422b44651d088b
> > commit: 3e1b02687146ddefa8e53c2a5e13d3bd87b16b09 [47/50] ext4: add support for hole punching non-extent-mapped files
> 
> I'm going to be dropping the indirect punch hole patches from the ext4
> dev tree, based on the number of xfstests failures that I found in my
> overnight testing.
> 
> I had hoped the changes were small and localized, and wouldn't cause
> problems if the punch hole feature wasn't used, but at this point,
> given that many distributions are starting to use the ext4 file system
> for legacy ext3 file systems, my judgement is that the risks are too
> great for the upcoming merge window; so this will probably need to
> wait for the next development cycle.

Hi Ted,

These patches would need to do more tests, and I will look at them in
next development cycle.  I am not sure whether indirect punch hole is
really useful for us or not.  What do you think about this feature?
Thanks

Regards,
Zheng