Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754581Ab0IAA0l (ORCPT ); Tue, 31 Aug 2010 20:26:41 -0400 Received: from fgwmail7.fujitsu.co.jp ([192.51.44.37]:45677 "EHLO fgwmail7.fujitsu.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752842Ab0IAA0j (ORCPT ); Tue, 31 Aug 2010 20:26:39 -0400 X-SecurityPolicyCheck-FJ: OK by FujitsuOutboundMailChecker v1.3.1 Date: Wed, 1 Sep 2010 09:21:36 +0900 From: KAMEZAWA Hiroyuki To: balbir@linux.vnet.ibm.com Cc: linux-mm@kvack.org, "nishimura@mxp.nes.nec.co.jp" , gthelen@google.com, m-ikeda@ds.jp.nec.com, "akpm@linux-foundation.org" , "linux-kernel@vger.kernel.org" , "menage@google.com" , "lizf@cn.fujitsu.com" Subject: Re: [PATCH 2/5] memcg: quick memcg lookup array Message-Id: <20100901092136.6c801ac4.kamezawa.hiroyu@jp.fujitsu.com> In-Reply-To: <20100831071414.GN32680@balbir.in.ibm.com> References: <20100825170435.15f8eb73.kamezawa.hiroyu@jp.fujitsu.com> <20100825170741.f1f0a220.kamezawa.hiroyu@jp.fujitsu.com> <20100831071414.GN32680@balbir.in.ibm.com> Organization: FUJITSU Co. LTD. X-Mailer: Sylpheed 3.0.3 (GTK+ 2.10.14; i686-pc-mingw32) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 8734 Lines: 261 On Tue, 31 Aug 2010 12:44:14 +0530 Balbir Singh wrote: > * KAMEZAWA Hiroyuki [2010-08-25 17:07:41]: > > > From: KAMEZAWA Hiroyuki > > > > Now, memory cgroup has an ID per cgroup and make use of it at > > - hierarchy walk, > > - swap recording. > > > > This patch is for making more use of it. The final purpose is > > to replace page_cgroup->mem_cgroup's pointer to an unsigned short. > > > > This patch caches a pointer of memcg in an array. By this, we > > don't have to call css_lookup() which requires radix-hash walk. > > This saves some amount of memory footprint at lookup memcg via id. > > > > Changelog: 20100825 > > - applied comments. > > > > Changelog: 20100811 > > - adjusted onto mmotm-2010-08-11 > > - fixed RCU related parts. > > - use attach_id() callback. > > > > Changelog: 20100804 > > - fixed description in init/Kconfig > > > > Changelog: 20100730 > > - fixed rcu_read_unlock() placement. > > > > Signed-off-by: KAMEZAWA Hiroyuki > > --- > > init/Kconfig | 10 +++++++ > > mm/memcontrol.c | 75 +++++++++++++++++++++++++++++++++++++++++--------------- > > 2 files changed, 65 insertions(+), 20 deletions(-) > > > > Index: mmotm-0811/mm/memcontrol.c > > =================================================================== > > --- mmotm-0811.orig/mm/memcontrol.c > > +++ mmotm-0811/mm/memcontrol.c > > @@ -195,6 +195,7 @@ static void mem_cgroup_oom_notify(struct > > */ > > struct mem_cgroup { > > struct cgroup_subsys_state css; > > + int valid; /* for checking validness under RCU access.*/ > > /* > > * the counter to account for memory usage > > */ > > @@ -294,6 +295,29 @@ static bool move_file(void) > > &mc.to->move_charge_at_immigrate); > > } > > > > +/* 0 is unused */ > > +static atomic_t mem_cgroup_num; > > +#define NR_MEMCG_GROUPS (CONFIG_MEM_CGROUP_MAX_GROUPS + 1) > > +static struct mem_cgroup *mem_cgroups[NR_MEMCG_GROUPS] __read_mostly; > > + > > +/* Must be called under rcu_read_lock */ > > +static struct mem_cgroup *id_to_memcg(unsigned short id) > > +{ > > + struct mem_cgroup *mem; > > + /* see mem_cgroup_free() */ > > + mem = rcu_dereference_check(mem_cgroups[id], rcu_read_lock_held()); > > + if (likely(mem && mem->valid)) > > + return mem; > > + return NULL; > > +} > > + > > +static void register_memcg_id(struct mem_cgroup *mem) > > +{ > > + int id = css_id(&mem->css); > > + rcu_assign_pointer(mem_cgroups[id], mem); > > + VM_BUG_ON(!mem->valid); > > +} > > + > > /* > > * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft > > * limit reclaim to prevent infinite loops, if they ever occur. > > @@ -1847,18 +1871,7 @@ static void mem_cgroup_cancel_charge(str > > * it's concern. (dropping refcnt from swap can be called against removed > > * memcg.) > > */ > > -static struct mem_cgroup *mem_cgroup_lookup(unsigned short id) > > -{ > > - struct cgroup_subsys_state *css; > > > > - /* ID 0 is unused ID */ > > - if (!id) > > - return NULL; > > - css = css_lookup(&mem_cgroup_subsys, id); > > - if (!css) > > - return NULL; > > - return container_of(css, struct mem_cgroup, css); > > -} > > > > struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page) > > { > > @@ -1879,7 +1892,7 @@ struct mem_cgroup *try_get_mem_cgroup_fr > > ent.val = page_private(page); > > id = lookup_swap_cgroup(ent); > > rcu_read_lock(); > > - mem = mem_cgroup_lookup(id); > > + mem = id_to_memcg(id); > > if (mem && !css_tryget(&mem->css)) > > mem = NULL; > > rcu_read_unlock(); > > @@ -2231,7 +2244,7 @@ __mem_cgroup_commit_charge_swapin(struct > > > > id = swap_cgroup_record(ent, 0); > > rcu_read_lock(); > > - memcg = mem_cgroup_lookup(id); > > + memcg = id_to_memcg(id); > > if (memcg) { > > /* > > * This recorded memcg can be obsolete one. So, avoid > > @@ -2240,9 +2253,10 @@ __mem_cgroup_commit_charge_swapin(struct > > if (!mem_cgroup_is_root(memcg)) > > res_counter_uncharge(&memcg->memsw, PAGE_SIZE); > > mem_cgroup_swap_statistics(memcg, false); > > + rcu_read_unlock(); > > mem_cgroup_put(memcg); > > - } > > - rcu_read_unlock(); > > + } else > > + rcu_read_unlock(); > > } > > /* > > * At swapin, we may charge account against cgroup which has no tasks. > > @@ -2495,7 +2509,7 @@ void mem_cgroup_uncharge_swap(swp_entry_ > > > > id = swap_cgroup_record(ent, 0); > > rcu_read_lock(); > > - memcg = mem_cgroup_lookup(id); > > + memcg = id_to_memcg(id); > > if (memcg) { > > /* > > * We uncharge this because swap is freed. > > @@ -2504,9 +2518,10 @@ void mem_cgroup_uncharge_swap(swp_entry_ > > if (!mem_cgroup_is_root(memcg)) > > res_counter_uncharge(&memcg->memsw, PAGE_SIZE); > > mem_cgroup_swap_statistics(memcg, false); > > + rcu_read_unlock(); > > mem_cgroup_put(memcg); > > - } > > - rcu_read_unlock(); > > + } else > > + rcu_read_unlock(); > > } > > > > /** > > @@ -4010,6 +4025,9 @@ static struct mem_cgroup *mem_cgroup_all > > struct mem_cgroup *mem; > > int size = sizeof(struct mem_cgroup); > > > > + if (atomic_read(&mem_cgroup_num) == NR_MEMCG_GROUPS) > > + return NULL; > > + > > /* Can be very big if MAX_NUMNODES is very big */ > > if (size < PAGE_SIZE) > > mem = kmalloc(size, GFP_KERNEL); > > @@ -4020,6 +4038,7 @@ static struct mem_cgroup *mem_cgroup_all > > return NULL; > > > > memset(mem, 0, size); > > + mem->valid = 1; > > mem->stat = alloc_percpu(struct mem_cgroup_stat_cpu); > > if (!mem->stat) { > > if (size < PAGE_SIZE) > > @@ -4049,6 +4068,7 @@ static void __mem_cgroup_free(struct mem > > mem_cgroup_remove_from_trees(mem); > > free_css_id(&mem_cgroup_subsys, &mem->css); > > > > + atomic_dec(&mem_cgroup_num); > > for_each_node_state(node, N_POSSIBLE) > > free_mem_cgroup_per_zone_info(mem, node); > > > > @@ -4059,6 +4079,19 @@ static void __mem_cgroup_free(struct mem > > vfree(mem); > > } > > > > +static void mem_cgroup_free(struct mem_cgroup *mem) > > +{ > > + /* No more lookup */ > > + mem->valid = 0; > > + rcu_assign_pointer(mem_cgroups[css_id(&mem->css)], NULL); > > + /* > > + * Because we call vfree() etc...use synchronize_rcu() rather than > > + * call_rcu(); > > + */ > > + synchronize_rcu(); > > + __mem_cgroup_free(mem); > > +} > > + > > static void mem_cgroup_get(struct mem_cgroup *mem) > > { > > atomic_inc(&mem->refcnt); > > @@ -4068,7 +4101,7 @@ static void __mem_cgroup_put(struct mem_ > > { > > if (atomic_sub_and_test(count, &mem->refcnt)) { > > struct mem_cgroup *parent = parent_mem_cgroup(mem); > > - __mem_cgroup_free(mem); > > + mem_cgroup_free(mem); > > if (parent) > > mem_cgroup_put(parent); > > } > > @@ -4189,9 +4222,11 @@ mem_cgroup_create(struct cgroup_subsys * > > atomic_set(&mem->refcnt, 1); > > mem->move_charge_at_immigrate = 0; > > mutex_init(&mem->thresholds_lock); > > + atomic_inc(&mem_cgroup_num); > > + register_memcg_id(mem); > > return &mem->css; > > free_out: > > - __mem_cgroup_free(mem); > > + mem_cgroup_free(mem); > > root_mem_cgroup = NULL; > > return ERR_PTR(error); > > } > > Index: mmotm-0811/init/Kconfig > > =================================================================== > > --- mmotm-0811.orig/init/Kconfig > > +++ mmotm-0811/init/Kconfig > > @@ -594,6 +594,16 @@ config CGROUP_MEM_RES_CTLR_SWAP > > Now, memory usage of swap_cgroup is 2 bytes per entry. If swap page > > size is 4096bytes, 512k per 1Gbytes of swap. > > > > +config MEM_CGROUP_MAX_GROUPS > > + int "Maximum number of memory cgroups on a system" > > + range 1 65535 > > + default 8192 if 64BIT > > + default 2048 if 32BIT > > + help > > + Memory cgroup has limitation of the number of groups created. > > + Please select your favorite value. The more you allow, the more > > + memory(a pointer per group) will be consumed. > > + > > Looks good, quick thought - should we expost memcg->id to user space > through a config file? I don't see any reason at this point, unless we > do it for all controllers. > I wonder....showing whether 2 interfaces as "path name" and "id" to users is a good thing or not. Yes, it's convenient to developpers as me and you, others, but I don't think it's useful to users, novice people. I'd like to avoid showing show memcg->id to users as much as possible. I don't want to say but memcg is enough complicated. Thanks, -Kame -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/