Re: [PATCH] mm/filemap.c: unconditionally call mark_page_accessed

[Nick Piggin]

On Thu, Mar 15, 2007 at 05:44:01PM +0000, Hugh Dickins wrote:
> On Thu, 15 Mar 2007, Nick Piggin wrote:
> > On Thu, Mar 15, 2007 at 11:56:59AM -0400, Chuck Ebbert wrote:
> > > Ashif Harji wrote:
> > > > 
> > > > This patch unconditionally calls mark_page_accessed to prevent pages,
> > > > especially for small files, from being evicted from the page cache
> > > > despite frequent access.
> > > > 
> > > > Signed-off-by: Ashif Harji <asharji@xxxxxxxxxxxxxxxxx>
> 
> Yeah, yeah, I'm not a real mman, I don't have my own patch and
> website for this ;) but I'm old, let me mumble some history...
> 
> Ashif's patch would take us back to 2.4.10 when mark_page_accessed
> was introduced: in 2.4.11 someone (probably Andrea) immediately
> added a !offset || !filp->f_reada condition on it there, which
> remains in 2.4 to this day.  That _probably_ means that Ashif's
> patch is suboptimal, and that your !offset patch is good.
> 
> f_reada went away in 2.5.8, and the !offset condition remained
> until 2.6.11, when Miquel (CC'ed) replaced it by today's prev_index
> condition.  His changelog entry appended below.  Since it's Miquel
> who removed the !offset condition, he should be consulted on its
> reintroduction.

Yeah I did go back and check up on that changelog, because I knew
we had a !offset check there at one stage, which is immune to this
problem (or at least can handle it a little better).

I suspect that Miquel was probably more interested in _increasing_
mark_page_accessed coverage with his new condition than restricting
it from the !offset cases.

Thanks for digging it up and posting here, though.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/