Re: [PATCH 2/3] mm, slub: count freed pages via rcu as this task'sreclaimed_slab

[Christoph Lameter]

On Wed, 10 Apr 2013, Joonsoo Kim wrote:

> Hello, Christoph.
>
> On Tue, Apr 09, 2013 at 02:28:06PM +0000, Christoph Lameter wrote:
> > On Tue, 9 Apr 2013, Joonsoo Kim wrote:
> >
> > > Currently, freed pages via rcu is not counted for reclaimed_slab, because
> > > it is freed in rcu context, not current task context. But, this free is
> > > initiated by this task, so counting this into this task's reclaimed_slab
> > > is meaningful to decide whether we continue reclaim, or not.
> > > So change code to count these pages for this task's reclaimed_slab.
> >
> > slab->reclaim_state guides the reclaim actions in vmscan.c. With this
> > patch slab->reclaim_state could get quite a high value without new pages being
> > available for allocation. slab->reclaim_state will only be updated
> > when the RCU period ends.
>
> Okay.
>
> In addition, there is a little place who use SLAB_DESTROY_BY_RCU.
> I will drop this patch[2/3] and [3/3] for next spin.

What you have discoverd is an issue that we have so far overlooked. Could
you add comments to both places explaining the situation? RCU is used for
some inode and the dentry cache. Failing to account for these frees could
pose a problem. One solution would be to ensure that we get through an RCU
quiescent period in the slabs reclaim. If we can ensure that then your
patch may be ok.


--
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/