From: JoonSoo Kim <js1304@gmail.com> To: Glauber Costa <glommer@parallels.com> Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, cgroups@vger.kernel.org, Mel Gorman <mgorman@suse.de>, Tejun Heo <tj@kernel.org>, Andrew Morton <akpm@linux-foundation.org>, Michal Hocko <mhocko@suse.cz>, Johannes Weiner <hannes@cmpxchg.org>, kamezawa.hiroyu@jp.fujitsu.com, Christoph Lameter <cl@linux.com>, David Rientjes <rientjes@google.com>, Pekka Enberg <penberg@kernel.org>, devel@openvz.org, Pekka Enberg <penberg@cs.helsinki.fi>, Suleiman Souhlal <suleiman@google.com> Subject: Re: [PATCH v5 08/18] memcg: infrastructure to match an allocation to the right cache Date: Thu, 25 Oct 2012 03:10:29 +0900 [thread overview] Message-ID: <CAAmzW4N40MedsCfcj+eiM-i6cU65n3z7uy08YFyknXbBKj7Z-g@mail.gmail.com> (raw) In-Reply-To: <1350656442-1523-9-git-send-email-glommer@parallels.com> 2012/10/19 Glauber Costa <glommer@parallels.com>: > @@ -2930,9 +2937,188 @@ int memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s) > > void memcg_release_cache(struct kmem_cache *s) > { > + struct kmem_cache *root; > + int id = memcg_css_id(s->memcg_params->memcg); > + > + if (s->memcg_params->is_root_cache) > + goto out; > + > + root = s->memcg_params->root_cache; > + root->memcg_params->memcg_caches[id] = NULL; > + mem_cgroup_put(s->memcg_params->memcg); > +out: > kfree(s->memcg_params); > } memcg_css_id should be called after checking "s->memcg_params->is_root_cache". Because when is_root_cache == true, memcg_params has no memcg object. > +/* > + * This lock protects updaters, not readers. We want readers to be as fast as > + * they can, and they will either see NULL or a valid cache value. Our model > + * allow them to see NULL, in which case the root memcg will be selected. > + * > + * We need this lock because multiple allocations to the same cache from a non > + * GFP_WAIT area will span more than one worker. Only one of them can create > + * the cache. > + */ > +static DEFINE_MUTEX(memcg_cache_mutex); > +static struct kmem_cache *memcg_create_kmem_cache(struct mem_cgroup *memcg, > + struct kmem_cache *cachep) > +{ > + struct kmem_cache *new_cachep; > + int idx; > + > + BUG_ON(!memcg_can_account_kmem(memcg)); > + > + idx = memcg_css_id(memcg); > + > + mutex_lock(&memcg_cache_mutex); > + new_cachep = cachep->memcg_params->memcg_caches[idx]; > + if (new_cachep) > + goto out; > + > + new_cachep = kmem_cache_dup(memcg, cachep); > + > + if (new_cachep == NULL) { > + new_cachep = cachep; > + goto out; > + } > + > + mem_cgroup_get(memcg); > + cachep->memcg_params->memcg_caches[idx] = new_cachep; > + wmb(); /* the readers won't lock, make sure everybody sees it */ Is there any rmb() pair? As far as I know, without rmb(), wmb() doesn't guarantee anything. > + new_cachep->memcg_params->memcg = memcg; > + new_cachep->memcg_params->root_cache = cachep; It may be better these assignment before the statement "cachep->memcg_params->memcg_caches[idx] = new_cachep". Otherwise, it may produce race situation. And assigning value to memcg_params->memcg and root_cache is redundant, because it is already done in memcg_register_cache(). > +/* > + * Return the kmem_cache we're supposed to use for a slab allocation. > + * We try to use the current memcg's version of the cache. > + * > + * If the cache does not exist yet, if we are the first user of it, > + * we either create it immediately, if possible, or create it asynchronously > + * in a workqueue. > + * In the latter case, we will let the current allocation go through with > + * the original cache. > + * > + * Can't be called in interrupt context or from kernel threads. > + * This function needs to be called with rcu_read_lock() held. > + */ > +struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, > + gfp_t gfp) > +{ > + struct mem_cgroup *memcg; > + int idx; > + > + if (cachep->memcg_params && cachep->memcg_params->memcg) > + return cachep; In __memcg_kmem_get_cache, cachep may be always root cache. So checking "cachep->memcg_params->memcg" is somewhat strange. Is it right? Thanks.
WARNING: multiple messages have this Message-ID (diff)
From: JoonSoo Kim <js1304@gmail.com> To: Glauber Costa <glommer@parallels.com> Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, cgroups@vger.kernel.org, Mel Gorman <mgorman@suse.de>, Tejun Heo <tj@kernel.org>, Andrew Morton <akpm@linux-foundation.org>, Michal Hocko <mhocko@suse.cz>, Johannes Weiner <hannes@cmpxchg.org>, kamezawa.hiroyu@jp.fujitsu.com, Christoph Lameter <cl@linux.com>, David Rientjes <rientjes@google.com>, Pekka Enberg <penberg@kernel.org>, devel@openvz.org, Pekka Enberg <penberg@cs.helsinki.fi>, Suleiman Souhlal <suleiman@google.com> Subject: Re: [PATCH v5 08/18] memcg: infrastructure to match an allocation to the right cache Date: Thu, 25 Oct 2012 03:10:29 +0900 [thread overview] Message-ID: <CAAmzW4N40MedsCfcj+eiM-i6cU65n3z7uy08YFyknXbBKj7Z-g@mail.gmail.com> (raw) In-Reply-To: <1350656442-1523-9-git-send-email-glommer@parallels.com> 2012/10/19 Glauber Costa <glommer@parallels.com>: > @@ -2930,9 +2937,188 @@ int memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s) > > void memcg_release_cache(struct kmem_cache *s) > { > + struct kmem_cache *root; > + int id = memcg_css_id(s->memcg_params->memcg); > + > + if (s->memcg_params->is_root_cache) > + goto out; > + > + root = s->memcg_params->root_cache; > + root->memcg_params->memcg_caches[id] = NULL; > + mem_cgroup_put(s->memcg_params->memcg); > +out: > kfree(s->memcg_params); > } memcg_css_id should be called after checking "s->memcg_params->is_root_cache". Because when is_root_cache == true, memcg_params has no memcg object. > +/* > + * This lock protects updaters, not readers. We want readers to be as fast as > + * they can, and they will either see NULL or a valid cache value. Our model > + * allow them to see NULL, in which case the root memcg will be selected. > + * > + * We need this lock because multiple allocations to the same cache from a non > + * GFP_WAIT area will span more than one worker. Only one of them can create > + * the cache. > + */ > +static DEFINE_MUTEX(memcg_cache_mutex); > +static struct kmem_cache *memcg_create_kmem_cache(struct mem_cgroup *memcg, > + struct kmem_cache *cachep) > +{ > + struct kmem_cache *new_cachep; > + int idx; > + > + BUG_ON(!memcg_can_account_kmem(memcg)); > + > + idx = memcg_css_id(memcg); > + > + mutex_lock(&memcg_cache_mutex); > + new_cachep = cachep->memcg_params->memcg_caches[idx]; > + if (new_cachep) > + goto out; > + > + new_cachep = kmem_cache_dup(memcg, cachep); > + > + if (new_cachep == NULL) { > + new_cachep = cachep; > + goto out; > + } > + > + mem_cgroup_get(memcg); > + cachep->memcg_params->memcg_caches[idx] = new_cachep; > + wmb(); /* the readers won't lock, make sure everybody sees it */ Is there any rmb() pair? As far as I know, without rmb(), wmb() doesn't guarantee anything. > + new_cachep->memcg_params->memcg = memcg; > + new_cachep->memcg_params->root_cache = cachep; It may be better these assignment before the statement "cachep->memcg_params->memcg_caches[idx] = new_cachep". Otherwise, it may produce race situation. And assigning value to memcg_params->memcg and root_cache is redundant, because it is already done in memcg_register_cache(). > +/* > + * Return the kmem_cache we're supposed to use for a slab allocation. > + * We try to use the current memcg's version of the cache. > + * > + * If the cache does not exist yet, if we are the first user of it, > + * we either create it immediately, if possible, or create it asynchronously > + * in a workqueue. > + * In the latter case, we will let the current allocation go through with > + * the original cache. > + * > + * Can't be called in interrupt context or from kernel threads. > + * This function needs to be called with rcu_read_lock() held. > + */ > +struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, > + gfp_t gfp) > +{ > + struct mem_cgroup *memcg; > + int idx; > + > + if (cachep->memcg_params && cachep->memcg_params->memcg) > + return cachep; In __memcg_kmem_get_cache, cachep may be always root cache. So checking "cachep->memcg_params->memcg" is somewhat strange. Is it right? Thanks. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2012-10-24 18:10 UTC|newest] Thread overview: 132+ messages / expand[flat|nested] mbox.gz Atom feed top 2012-10-19 14:20 [PATCH v5 00/18] slab accounting for memcg Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 01/18] move slabinfo processing to slab_common.c Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-24 6:43 ` Pekka Enberg 2012-10-24 6:43 ` Pekka Enberg 2012-10-24 6:43 ` Pekka Enberg 2012-10-19 14:20 ` [PATCH v5 02/18] move print_slabinfo_header " Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 03/18] sl[au]b: process slabinfo_show in common code Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 04/18] slab: don't preemptively remove element from list in cache destroy Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 19:34 ` Christoph Lameter 2012-10-19 19:34 ` Christoph Lameter 2012-10-19 19:34 ` Christoph Lameter 2012-10-22 8:40 ` Glauber Costa 2012-10-22 8:40 ` Glauber Costa 2012-10-22 8:40 ` Glauber Costa 2012-10-24 6:54 ` Pekka Enberg 2012-10-24 6:54 ` Pekka Enberg 2012-10-24 6:54 ` Pekka Enberg 2012-10-24 16:19 ` Glauber Costa 2012-10-24 16:19 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 05/18] slab/slub: struct memcg_params Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-23 17:25 ` JoonSoo Kim 2012-10-23 17:25 ` JoonSoo Kim 2012-10-23 17:25 ` JoonSoo Kim 2012-10-24 8:42 ` Glauber Costa 2012-10-24 8:42 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 06/18] consider a memcg parameter in kmem_create_cache Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-23 17:50 ` JoonSoo Kim 2012-10-23 17:50 ` JoonSoo Kim 2012-10-23 17:50 ` JoonSoo Kim 2012-10-24 8:42 ` Glauber Costa 2012-10-24 8:42 ` Glauber Costa 2012-10-24 8:42 ` Glauber Costa 2012-10-25 13:42 ` Glauber Costa 2012-10-25 13:42 ` Glauber Costa 2012-10-25 13:42 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 07/18] Allocate memory for memcg caches whenever a new memcg appears Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 08/18] memcg: infrastructure to match an allocation to the right cache Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-24 18:10 ` JoonSoo Kim [this message] 2012-10-24 18:10 ` JoonSoo Kim 2012-10-25 11:05 ` Glauber Costa 2012-10-25 11:05 ` Glauber Costa 2012-10-25 11:05 ` Glauber Costa 2012-10-25 18:06 ` Tejun Heo 2012-10-25 18:06 ` Tejun Heo 2012-10-25 18:06 ` Tejun Heo 2012-10-25 18:08 ` Tejun Heo 2012-10-25 18:08 ` Tejun Heo 2012-10-19 14:20 ` [PATCH v5 09/18] memcg: skip memcg kmem allocations in specified code regions Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 10/18] sl[au]b: always get the cache from its page in kfree Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 19:44 ` Christoph Lameter 2012-10-19 19:44 ` Christoph Lameter 2012-10-19 19:44 ` Christoph Lameter 2012-10-22 10:13 ` Glauber Costa 2012-10-22 10:13 ` Glauber Costa 2012-10-22 10:13 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 11/18] sl[au]b: Allocate objects from memcg cache Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 19:46 ` Christoph Lameter 2012-10-19 19:46 ` Christoph Lameter 2012-10-19 19:46 ` Christoph Lameter 2012-10-29 15:14 ` JoonSoo Kim 2012-10-29 15:14 ` JoonSoo Kim 2012-10-29 15:14 ` JoonSoo Kim 2012-10-29 15:19 ` Glauber Costa 2012-10-29 15:19 ` Glauber Costa 2012-10-29 15:19 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 12/18] memcg: destroy memcg caches Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 13/18] memcg/sl[au]b Track all the memcg children of a kmem_cache Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-29 15:26 ` JoonSoo Kim 2012-10-29 15:26 ` JoonSoo Kim 2012-10-29 15:26 ` JoonSoo Kim 2012-10-30 11:31 ` Glauber Costa 2012-10-30 11:31 ` Glauber Costa 2012-10-30 11:31 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 14/18] memcg/sl[au]b: shrink dead caches Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 19:47 ` Christoph Lameter 2012-10-19 19:47 ` Christoph Lameter 2012-10-19 19:47 ` Christoph Lameter 2012-10-22 7:37 ` Glauber Costa 2012-10-22 7:37 ` Glauber Costa 2012-10-22 7:37 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 15/18] Aggregate memcg cache values in slabinfo Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 19:50 ` Christoph Lameter 2012-10-19 19:50 ` Christoph Lameter 2012-10-22 15:11 ` Glauber Costa 2012-10-22 15:11 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 16/18] slab: propagate tunables values Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 19:51 ` Christoph Lameter 2012-10-19 19:51 ` Christoph Lameter 2012-10-22 7:48 ` Glauber Costa 2012-10-22 7:48 ` Glauber Costa 2012-10-23 20:44 ` Christoph Lameter 2012-10-23 20:44 ` Christoph Lameter 2012-10-19 14:20 ` [PATCH v5 17/18] slub: slub-specific propagation changes Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` [PATCH v5 18/18] Add slab-specific documentation about the kmem controller Glauber Costa 2012-10-19 14:20 ` Glauber Costa 2012-10-19 14:20 ` Glauber Costa
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=CAAmzW4N40MedsCfcj+eiM-i6cU65n3z7uy08YFyknXbBKj7Z-g@mail.gmail.com \ --to=js1304@gmail.com \ --cc=akpm@linux-foundation.org \ --cc=cgroups@vger.kernel.org \ --cc=cl@linux.com \ --cc=devel@openvz.org \ --cc=glommer@parallels.com \ --cc=hannes@cmpxchg.org \ --cc=kamezawa.hiroyu@jp.fujitsu.com \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-mm@kvack.org \ --cc=mgorman@suse.de \ --cc=mhocko@suse.cz \ --cc=penberg@cs.helsinki.fi \ --cc=penberg@kernel.org \ --cc=rientjes@google.com \ --cc=suleiman@google.com \ --cc=tj@kernel.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.