Re: [PATCH 01/19] tree wide: get rid of __GFP_REPEAT for order-0 allocations part I

[Michal Hocko]

On Thu 14-04-16 12:56:28, David Rientjes wrote:
> On Mon, 11 Apr 2016, Michal Hocko wrote:
> 
> > From: Michal Hocko <mhocko@xxxxxxxx>
> > 
> > __GFP_REPEAT has a rather weak semantic but since it has been introduced
> > around 2.6.12 it has been ignored for low order allocations. Yet we have
> > the full kernel tree with its usage for apparently order-0 allocations.
> > This is really confusing because __GFP_REPEAT is explicitly documented
> > to allow allocation failures which is a weaker semantic than the current
> > order-0 has (basically nofail).
> > 
> > Let's simply drop __GFP_REPEAT from those places. This would allow
> > to identify place which really need allocator to retry harder and
> > formulate a more specific semantic for what the flag is supposed to do
> > actually.
> > 
> > Cc: linux-arch@xxxxxxxxxxxxxxx
> > Signed-off-by: Michal Hocko <mhocko@xxxxxxxx>
> 
> I did exactly this before, and Andrew objected saying that __GFP_REPEAT 
> may not be needed for the current page allocator's implementation but 
> could with others and that setting __GFP_REPEAT for an allocation 
> provided useful information with regards to intent.

>From what I've seen it was more a copy&paste of the arch code which
spread out this flag and there was also a misleading usage.

> At the time, I attempted to eliminate __GFP_REPEAT entirely.

This is not my plan. I actually want to provide a useful semantic for
something like this flag - aka try really hard but eventually fail
for all orders and stop being special only for those that are costly. I
will call it __GFP_BEST_EFFORT. But I have to clean up the current usage
first. Costly orders will keep __GFP_REPEAT because the intent is clear
there. All others will lose the flag and then we can start adding
__GFP_BEST_EFFORT where it matters also for lower orders.
-- 
Michal Hocko
SUSE Labs