Re: [PATCH] mm: avoid blocking lock_page() in kcompactd

[Michal Hocko <mhocko@kernel.org>]

On Thu 13-02-20 20:27:24, Matthew Wilcox wrote:
> On Thu, Feb 13, 2020 at 06:08:24PM +0100, Michal Hocko wrote:
> > On Thu 13-02-20 08:46:07, Matthew Wilcox wrote:
> > > On Thu, Feb 13, 2020 at 08:48:47AM +0100, Michal Hocko wrote:
> > > > Can we pursue on this please? An explicit NOFS scope annotation with a
> > > > reference to compaction potentially locking up on pages in the readahead
> > > > would be a great start.
> > > 
> > > How about this (on top of the current readahead series):
> > > 
> > > diff --git a/mm/readahead.c b/mm/readahead.c
> > > index 29ca25c8f01e..32fd32b913da 100644
> > > --- a/mm/readahead.c
> > > +++ b/mm/readahead.c
> > > @@ -160,6 +160,16 @@ unsigned long page_cache_readahead_limit(struct address_space *mapping,
> > >  		.nr_pages = 0,
> > >  	};
> > >  
> > > +	/*
> > > +	 * Partway through the readahead operation, we will have added
> > > +	 * locked pages to the page cache, but will not yet have submitted
> > > +	 * them for I/O.  Adding another page may need to allocate
> > > +	 * memory, which can trigger memory migration.	Telling the VM
> > 
> > I would go with s@migration@compaction@ because it would make it more
> > obvious that this is about high order allocations.
> 
> Perhaps even just 'reclaim' -- it's about compaction today, but tomorrow's
> VM might try to reclaim these pages too.  They are on the LRU, after all.
> 
> So I currently have:
> 
>         /*
>          * Partway through the readahead operation, we will have added
>          * locked pages to the page cache, but will not yet have submitted
>          * them for I/O.  Adding another page may need to allocate memory,
>          * which can trigger memory reclaim.  Telling the VM we're in
>          * the middle of a filesystem operation will cause it to not
>          * touch file-backed pages, preventing a deadlock.  Most (all?)
>          * filesystems already specify __GFP_NOFS in their mapping's
>          * gfp_mask, but let's be explicit here.
>          */

OK, Thanks!
-- 
Michal Hocko
SUSE Labs