Re: [patch] mm: remove sysctl to manually rescue unevictable pages

[Johannes Weiner]

On Mon, Sep 26, 2011 at 05:59:39PM +0530, kautuk.c @samsung.com wrote:
> On Mon, Sep 26, 2011 at 5:40 PM, kautuk.c @samsung.com
> <consul.kautuk@xxxxxxxxx> wrote:
> > On Mon, Sep 26, 2011 at 4:59 PM, Johannes Weiner <jweiner@xxxxxxxxxx> wrote:
> >> On Sun, Sep 25, 2011 at 04:29:40PM +0530, Kautuk Consul wrote:
> >>> write_scan_unavictable_node checks the value req returned by
> >>> strict_strtoul and returns 1 if req is 0.
> >>>
> >>> However, when strict_strtoul returns 0, it means successful conversion
> >>> of buf to unsigned long.
> >>>
> >>> Due to this, the function was not proceeding to scan the zones for
> >>> unevictable pages even though we write a valid value to the
> >>> scan_unevictable_pages sys file.
> >>
> >> Given that there is not a real reason for this knob (anymore) and that
> >> it apparently never really worked since the day it was introduced, how
> >> about we just drop all that code instead?
> >>
> >>        Hannes
> >>
> >> ---
> >> From: Johannes Weiner <jweiner@xxxxxxxxxx>
> >> Subject: mm: remove sysctl to manually rescue unevictable pages
> >>
> >> At one point, anonymous pages were supposed to go on the unevictable
> >> list when no swap space was configured, and the idea was to manually
> >> rescue those pages after adding swap and making them evictable again.
> >> But nowadays, swap-backed pages on the anon LRU list are not scanned
> >> without available swap space anyway, so there is no point in moving
> >> them to a separate list anymore.
> >
> > Is this code only for anonymous pages ?
> > It seems to look at all pages in the zone both file as well as anon.
> >
> >>
> >> The manual rescue could also be used in case pages were stranded on
> >> the unevictable list due to race conditions.  But the code has been
> >> around for a while now and newly discovered bugs should be properly
> >> reported and dealt with instead of relying on such a manual fixup.
> >
> > What you say seems to be all right for anon pages, but what about file
> > pages ?
> > I'm not sure about how this could happen, but what if some file-system caused
> > a file cache page to be set to evictable or reclaimable without
> > actually removing
> > that page from the unevictable list ?
> 
> What I would like to also add is that while the transition of an anon
> page from and
> to the unevictable lists is straight-forward, should we make the same assumption
> about file cache pages ?

We should make no assumptions if our code base is open source :-)

> I am not sure about this, but could a file-system cause this kind of a problem
> independent of the mlocking behaviour of a user-mode app ?

Currently, I only see shmem and ramfs meddling with unevictability
outside of mlock and they both look correct to me.

I'd say that if a filesystem required this knob and user-intervention
for the VM to behave correctly, it needs fixing.
--
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/