On Tue, Nov 11, 2014 at 02:53:10PM -0500, Johannes Weiner wrote:
> On Tue, Nov 11, 2014 at 05:34:19PM +1100, Dave Chinner wrote:
> > On Mon, Nov 10, 2014 at 09:44:05AM -0500, Johannes Weiner wrote:
> > > On Mon, Nov 10, 2014 at 05:46:40PM +1100, Dave Chinner wrote:
> > > > On Thu, Nov 06, 2014 at 0
On Tue, Nov 11, 2014 at 05:34:19PM +1100, Dave Chinner wrote:
> On Mon, Nov 10, 2014 at 09:44:05AM -0500, Johannes Weiner wrote:
> > On Mon, Nov 10, 2014 at 05:46:40PM +1100, Dave Chinner wrote:
> > > On Thu, Nov 06, 2014 at 06:50:28PM -0500, Johannes Weiner wrote:
> > > > The slab shrinkers curren
On Mon, Nov 10, 2014 at 09:44:05AM -0500, Johannes Weiner wrote:
> On Mon, Nov 10, 2014 at 05:46:40PM +1100, Dave Chinner wrote:
> > On Thu, Nov 06, 2014 at 06:50:28PM -0500, Johannes Weiner wrote:
> > > The slab shrinkers currently rely on the reclaim code providing an
> > > ad-hoc concept of NUMA
On Mon, Nov 10, 2014 at 05:46:40PM +1100, Dave Chinner wrote:
> On Thu, Nov 06, 2014 at 06:50:28PM -0500, Johannes Weiner wrote:
> > The slab shrinkers currently rely on the reclaim code providing an
> > ad-hoc concept of NUMA nodes that doesn't really exist: for all
> > scanned zones and lruvecs,
On Thu, Nov 06, 2014 at 06:50:28PM -0500, Johannes Weiner wrote:
> The slab shrinkers currently rely on the reclaim code providing an
> ad-hoc concept of NUMA nodes that doesn't really exist: for all
> scanned zones and lruvecs, the nodes and available LRU pages are
> summed up, only to have the sh
On Fri, Nov 07, 2014 at 12:18:11PM +0300, Vladimir Davydov wrote:
> On Thu, Nov 06, 2014 at 06:50:28PM -0500, Johannes Weiner wrote:
> [...]
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index a384339bf718..6a9ab5adf118 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> [...]
> > @@ -1876,7 +187
Hi Johannes,
The general idea sounds sane to me. A few comments inline.
On Thu, Nov 06, 2014 at 06:50:28PM -0500, Johannes Weiner wrote:
[...]
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index a384339bf718..6a9ab5adf118 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
[...]
> @@ -1876,7 +1872,8 @@ e
The slab shrinkers currently rely on the reclaim code providing an
ad-hoc concept of NUMA nodes that doesn't really exist: for all
scanned zones and lruvecs, the nodes and available LRU pages are
summed up, only to have the shrinkers then again walk that nodemask
when scanning slab caches. This du
8 matches
Mail list logo