Re: [PATCH] btrfs: use appropriate replacements for __sb_{start,end}_write calls

[Rakesh Pandit]

On Tue, Oct 10, 2017 at 02:08:11PM +0300, Nikolay Borisov wrote:
> 
> 
> On 10.10.2017 13:48, Rakesh Pandit wrote:
> > Commit a53f4f8e9c8eb ("btrfs: Don't call btrfs_start_transaction() on
> > frozen fs to avoid deadlock.") started using internal calls and we
> > replace them with more suitable ones.
> > 
> > Signed-off-by: Rakesh Pandit <rakesh@xxxxxxxxxx>
> > ---
> >  fs/btrfs/super.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> > index 35a128a..99c21ae 100644
> > --- a/fs/btrfs/super.c
> > +++ b/fs/btrfs/super.c
> > @@ -1205,8 +1205,8 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
> >  			 * happens. The pending operations are delayed to the
> >  			 * next commit after thawing.
> >  			 */
> > -			if (__sb_start_write(sb, SB_FREEZE_WRITE, false))
> > -				__sb_end_write(sb, SB_FREEZE_WRITE);
> > +			if (sb_start_write_trylock(sb))
> > +				sb_end_write(sb)
> >  			else
> >  				return 0;
> 
> On second thought, what's to prevent the filesystem to be frozen if
> sb_start_write/sb_end_write code executes? Or even after we are in the
> middle of btrfs_start_transaction?

sb->s_writers.frozen is protected by sb->s_umount and s_umount is
held when ->sync_fs is called.

> 
> 
> >  			trans = btrfs_start_transaction(root, 0);
> >