Re: [PATCH v2 16/17] mm/slab: introduce new slab management type, OBJFREELIST_SLAB
From: Joonsoo Kim
Date: Fri Feb 26 2016 - 12:06:42 EST
2016-02-27 1:21 GMT+09:00 Christoph Lameter <cl@xxxxxxxxx>:
> On Fri, 26 Feb 2016, js1304@xxxxxxxxx wrote:
>> Although this idea can apply to all caches whose size is larger than
>> management array size, it isn't applied to caches which have a
>> constructor. If such cache's object is used for management array,
>> constructor should be called for it before that object is returned to
>> user. I guess that overhead overwhelm benefit in that case so this idea
>> doesn't applied to them at least now.
> Caches which have a constructor (or are used with SLAB_RCU_FREE) have a
> defined content even when they are free. Therefore they cannot be used
> for the freelist.
Yes, I know. I already handled it. I attach related hunk.
+static bool set_objfreelist_slab_cache(struct kmem_cache *cachep,
+ size_t size, unsigned long flags)
+ size_t left;
+ cachep->num = 0;
+ if (cachep->ctor || flags & SLAB_DESTROY_BY_RCU)
+ return false;
So, if there is ctor or RCU slabs, objfreelist will not be used.
>> For summary, from now on, slab management type is determined by
>> following logic.
>> 1) if management array size is smaller than object size and no ctor, it
>> becomes OBJFREELIST_SLAB.
> Also do not do this for RCU slabs.
>> 2) if management array size is smaller than leftover, it becomes
>> NORMAL_SLAB which uses leftover as a array.
>> 3) if OFF_SLAB help to save memory than way 4), it becomes OFF_SLAB.
>> It allocate a management array from the other cache so memory waste
> Wonder how many of these ugly off slabs are left after what you did here.
See below result.
>> TOTAL = OBJFREELIST + NORMAL(leftover) + NORMAL + OFF
>> 126 = 0 + 60 + 25 + 41
>> 126 = 97 + 12 + 15 + 2
97 is the number of 1) type caches.
12 is the number of 2) type caches.
and so on...
>> Result shows that number of caches that doesn't waste memory increase
>> from 60 to 109.
> Great results.
>> v2: fix SLAB_DESTROTY_BY_RCU cache type handling
> Ok how are they handled now? Do not see that dealt with in the patch.