Re: [PATCH] slub: do not put a slab to cpu partial list when cpu_partial is 0

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Jun 20, 2013 at 08:26:03AM +0800, Wanpeng Li wrote:
> On Wed, Jun 19, 2013 at 05:52:50PM +0900, Joonsoo Kim wrote:
> >On Wed, Jun 19, 2013 at 04:00:32PM +0800, Wanpeng Li wrote:
> >> On Wed, Jun 19, 2013 at 03:33:55PM +0900, Joonsoo Kim wrote:
> >> >In free path, we don't check number of cpu_partial, so one slab can
> >> >be linked in cpu partial list even if cpu_partial is 0. To prevent this,
> >> >we should check number of cpu_partial in put_cpu_partial().
> >> >
> >> 
> >> How about skip get_partial entirely? put_cpu_partial is called 
> >> in two paths, one is during refill cpu partial lists in alloc 
> >> slow path, the other is in free slow path. And cpu_partial is 0 
> >> just in debug mode. 
> >> 
> >> - alloc slow path, there is unnecessary to call get_partial 
> >>   since cpu partial lists won't be used in debug mode. 
> >> - free slow patch, new.inuse won't be true in debug mode 
> >>   which lead to put_cpu_partial won't be called.
> >> 
> >
> >In debug mode, put_cpu_partial() can't be called already on both path.
> >But, if we assign 0 to cpu_partial via sysfs, put_cpu_partial() will be called
> >on free slow path. On alloc slow path, it can't be called, because following
> >test in get_partial_node() is always failed.
> >
> >available > s->cpu_partial / 2
> 
> Is it always true? We can freeze slab from partial list, and 
> s->cpu_partial is 0. 

Do you mean node partial list?

At first, acquire_slab() is called for a cpu slab(not cpu partial list)
in get_partial_node(), and then, check above test. In this time, available
is always higher than 0, so, if we assign 0 to s->cpu_partial, we break
the loop and we don't try to get a slab for cpu partial list.

Thanks.

> 
> Regards,
> Wanpeng Li 
> 
> >
> >Thanks.
> >
> >> Regards,
> >> Wanpeng Li 
> >> 
> >> >Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> >> >
> >> >diff --git a/mm/slub.c b/mm/slub.c
> >> >index 57707f0..7033b4f 100644
> >> >--- a/mm/slub.c
> >> >+++ b/mm/slub.c
> >> >@@ -1955,6 +1955,9 @@ static void put_cpu_partial(struct kmem_cache *s, struct page *page, int drain)
> >> > 	int pages;
> >> > 	int pobjects;
> >> >
> >> >+	if (!s->cpu_partial)
> >> >+		return;
> >> >+
> >> > 	do {
> >> > 		pages = 0;
> >> > 		pobjects = 0;
> >> >-- 
> >> >1.7.9.5
> >> >
> >> >--
> >> >To unsubscribe, send a message with 'unsubscribe linux-mm' in
> >> >the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
> >> >see: http://www.linux-mm.org/ .
> >> >Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
> >> 
> >> --
> >> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> >> the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
> >> see: http://www.linux-mm.org/ .
> >> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]