RE: Regression caused by using node_to_bdi()

[Zhao Lei <zhaolei@cn.fujitsu.com>]

Hi, Boaz

> -----Original Message-----
> From: Zhao Lei [mailto:zhaolei@cn.fujitsu.com]
> Sent: Monday, April 13, 2015 3:00 PM
> To: 'Boaz Harrosh'; 'Christoph Hellwig'
> Cc: 'linux-fsdevel@vger.kernel.org'; 'Jan Kara'; 'Jens Axboe'; 'LKML'
> Subject: RE: Regression caused by using node_to_bdi()
> 
> Hi, Boaz
> 
> > -----Original Message-----
> > From: Boaz Harrosh [mailto:boaz@plexistor.com]
> > Sent: Sunday, April 12, 2015 10:39 PM
> > To: Boaz Harrosh; Zhao Lei; 'Christoph Hellwig'
> > Cc: linux-fsdevel@vger.kernel.org; 'Jan Kara'; 'Jens Axboe'; 'LKML'
> > Subject: Re: Regression caused by using node_to_bdi()
> >
> > On 04/12/2015 02:33 PM, Boaz Harrosh wrote:
> > > On 04/10/2015 02:25 PM, Zhao Lei wrote:
> > >> Hi, Christoph Hellwig
> > >>
> > <>
> > >>
> > >> Is there some way to speed up it(inline, or some access some
> > >> variant in struct directly, ...)?
> > >>
> > >
> > > Christoph hi
> > >
> > > Both node_to_bdi() and sb_is_blkdev_sb()  (and I_BDEV() &&
> > > blk_get_backing_dev_info()) Are an exported function calls.
> > >
> > > Can we not make blockdev_superblock->s_bdi == NULL, and then
> > > optimize-out the call to sb_is_blkdev_sb() to only that case.
> > > Something like:
> > >
> > > ---
> > >
> > > diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c index
> > > 32a8bbd..e0375e1 100644
> > > --- a/fs/fs-writeback.c
> > > +++ b/fs/fs-writeback.c
> > > @@ -78,7 +78,7 @@ int writeback_in_progress(struct backing_dev_info
> > > *bdi)  }  EXPORT_SYMBOL(writeback_in_progress);
> > >
> > > -struct backing_dev_info *inode_to_bdi(struct inode *inode)
> > > +struct backing_dev_info *__inode_to_bdi(struct inode *inode)
> > >  {
> > >  	struct super_block *sb;
> > >
> > > @@ -92,7 +92,7 @@ struct backing_dev_info *inode_to_bdi(struct inode
> > > *inode)  #endif
> > >  	return sb->s_bdi;
> > >  }
> > > -EXPORT_SYMBOL_GPL(inode_to_bdi);
> > > +EXPORT_SYMBOL_GPL(__inode_to_bdi);
> > >
> > >  static inline struct inode *wb_inode(struct list_head *head)  {
> > > diff --git a/include/linux/backing-dev.h
> > > b/include/linux/backing-dev.h index aff923a..7d172f5 100644
> > > --- a/include/linux/backing-dev.h
> > > +++ b/include/linux/backing-dev.h
> > > @@ -107,7 +107,16 @@ struct backing_dev_info {  #endif  };
> > >
> > > -struct backing_dev_info *inode_to_bdi(struct inode *inode);
> > > +struct backing_dev_info *__inode_to_bdi(struct inode *inode);
> > > +
> > > +static inline
> > > +struct backing_dev_info *inode_to_bdi(struct inode *inode) {
> > > +	if (!inode || !inode->i_sb)
> > > +		return __inode_to_bdi(inode);
> > > +
> > > +	return inode->i_sb->s_bdi;
> > > +}
> > >
> >
> > This patch actually boots. Lei could you please test to see if it
> > fixes your slowness?
> >
> The good news is this patch passed compile and 10-time tests.
> The bad news is it have more performance down(strange)...
> 
> v3.19-rc1              : io_speed: valcnt=10 avg=214.688
> range=[211.460,216.190] diff=  2.24% stdev=1.417 cv=0.66%
> v4.0-rc1               : io_speed: valcnt=10 avg=204.917
> range=[203.370,205.890] diff=  1.24% stdev=0.663 cv=0.32%
> v4.0-rc1_00001_82ad06  : io_speed: valcnt=10 avg=189.337
> range=[186.280,192.060] diff=  3.10% stdev=2.305 cv=1.22% *<- this patch
> 
> I applied this patch on top of v4.0-rc1.
> 
A new bad news:
This patch make filesystem unstable.

My env entered to following command line in booting after
apply this patch to v4.0-rc1:

Welcome to emergency mode! After logging in, type "journalctl -xb: to view
System logs, ...
Give root password for maintenance
(or press Control-D to continue)

I confirmed this error message for more than 3 times.
(and confirmed no-problem without this patch)

In previous performance test(which get result in my last mail), I hadn't 
pay attention to that message, and just type Ctrl-D and begin test.

Thanks
Zhaolei

> > Thanks
> > Boaz
> >
> > >  int __must_check bdi_init(struct backing_dev_info *bdi);  void
> > > bdi_destroy(struct backing_dev_info *bdi);
> > >