Re: [PATCH v11] ovl: Improving syncfs efficiency

[Jan Kara <jack@suse.cz>]

On Thu 16-04-20 10:00:13, Miklos Szeredi wrote:
> On Thu, Apr 16, 2020 at 9:39 AM Miklos Szeredi <miklos@szeredi.hu> wrote:
> >
> > On Thu, Apr 16, 2020 at 9:21 AM Miklos Szeredi <miklos@szeredi.hu> wrote:
> > >
> > > On Thu, Apr 16, 2020 at 8:09 AM Chengguang Xu <cgxu519@mykernel.net> wrote:
> > > >
> > > >  ---- 在 星期四, 2020-04-16 03:19:50 Miklos Szeredi <miklos@szeredi.hu> 撰写 ----
> > > >  > On Mon, Feb 10, 2020 at 4:10 AM Chengguang Xu <cgxu519@mykernel.net> wrote:
> > >
> > > >  > > +               if (current->flags & PF_MEMALLOC) {
> > > >  > > +                       spin_lock(&inode->i_lock);
> > > >  > > +                       ovl_set_flag(OVL_WRITE_INODE_PENDING, inode);
> > > >  > > +                       wqh = bit_waitqueue(&oi->flags,
> > > >  > > +                                       OVL_WRITE_INODE_PENDING);
> > > >  > > +                       prepare_to_wait(wqh, &wait.wq_entry,
> > > >  > > +                                       TASK_UNINTERRUPTIBLE);
> > > >  > > +                       spin_unlock(&inode->i_lock);
> > > >  > > +
> > > >  > > +                       ovl_wiw.inode = inode;
> > > >  > > +                       INIT_WORK(&ovl_wiw.work, ovl_write_inode_work_fn);
> > > >  > > +                       schedule_work(&ovl_wiw.work);
> > > >  > > +
> > > >  > > +                       schedule();
> > > >  > > +                       finish_wait(wqh, &wait.wq_entry);
> > > >  >
> > > >  > What is the reason to do this in another thread if this is a PF_MEMALLOC task?
> > > >
> > > > Some underlying filesystems(for example ext4) check the flag in ->write_inode()
> > > > and treate it as an abnormal case.(warn and return)
> > > >
> > > > ext4_write_inode():
> > > >         if (WARN_ON_ONCE(current->flags & PF_MEMALLOC) ||
> > > >                 sb_rdonly(inode->i_sb))
> > > >                         return 0;
> > > >
> > > > overlayfs inodes are always keeping clean even after wring/modifying  upperfile ,
> > > > so they are right target of kswapd  but in the point of lower layer, ext4 just thinks
> > > > kswapd is choosing a wrong dirty inode to reclam memory.
> > >
> > > I don't get it.  Why can't overlayfs just skip the writeback of upper
> > > inode in the reclaim case?  It will be written back through the normal
> > > relcaim channels.
> >
> > And how do we get reclaim on overlay inode at all?   Overlay inodes
> > will get evicted immediately after their refcount drops to zero, so
> > there's absolutely no chance that memory reclaim will encounter them,
> > no?
> 
> Spoke too soon.   Obviously this case is about dentry reclaim, not
> inode reclaim.
> 
> So how about temporarily clearing PF_MEMALLOC in this case?  Doing
> this is a kernel thread doesn't seem to add any advantages.

Clearing PF_MEMALLOC will not solve the deadlock I've described in the
reply to Chengguang. Ext4 really cannot safely handle data integrity
writeback (which is what write_inode_now(inode, 1) does) from direct
reclaim.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR