Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756820AbZAOBTX (ORCPT ); Wed, 14 Jan 2009 20:19:23 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1754161AbZAOBTO (ORCPT ); Wed, 14 Jan 2009 20:19:14 -0500 Received: from TYO202.gate.nec.co.jp ([202.32.8.206]:45370 "EHLO tyo202.gate.nec.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751308AbZAOBTM (ORCPT ); Wed, 14 Jan 2009 20:19:12 -0500 Date: Thu, 15 Jan 2009 10:03:30 +0900 From: Daisuke Nishimura To: "KAMEZAWA Hiroyuki" Cc: nishimura@mxp.nes.nec.co.jp, "LKML" , "linux-mm" , "Andrew Morton" , "Balbir Singh" , "Pavel Emelyanov" , "Li Zefan" , "Paul Menage" Subject: Re: [RFC][PATCH 5/4] memcg: don't call res_counter_uncharge when obsolete Message-Id: <20090115100330.37d89d3d.nishimura@mxp.nes.nec.co.jp> In-Reply-To: <7602a77a9fc6b1e8757468048fde749a.squirrel@webmail-b.css.fujitsu.com> References: <20090113184533.6ffd2af9.nishimura@mxp.nes.nec.co.jp> <20090114175121.275ecd59.nishimura@mxp.nes.nec.co.jp> <7602a77a9fc6b1e8757468048fde749a.squirrel@webmail-b.css.fujitsu.com> Organization: NEC Soft, Ltd. X-Mailer: Sylpheed 2.4.8 (GTK+ 2.10.14; i686-pc-mingw32) Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-2022-JP Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 7918 Lines: 206 On Wed, 14 Jan 2009 22:43:05 +0900 (JST), "KAMEZAWA Hiroyuki" wrote: > Daisuke Nishimura さんは書きました： > > This is a new one. Please review. > > > > === > > From: Daisuke Nishimura > > > > mem_cgroup_get ensures that the memcg that has been got can be accessed > > even after the directory has been removed, but it doesn't ensure that > > parents > > of it can be accessed: parents might have been freed already by rmdir. > > > > This causes a bug in case of use_hierarchy==1, because > > res_counter_uncharge > > climb up the tree. > > > > Check if the memcg is obsolete, and don't call res_counter_uncharge when > > obsole. > > > Hmm, did you see panic ? I saw 2 types of bugs, A: spinlock lockup and B: general protection fault. (described below) Those bugs happened in case of (use_hierarchy && do_swap_account), and didn't happen (at leaset I haven't seen) in case of (!use_hierarchy && do_swap_account) nor (use_hierarchy && !do_swap_account). And, they didn't happen with this patch applied all through the last night. A: spinlock lockup === BUG: spinlock lockup on CPU#1, mmapstress10/27706, ffff880 3a41ef0a0 Pid: 27706, comm: mmapstress10 Not tainted 2.6.28-git12-7c 99bf20 #1 Call Trace: [] _raw_spin_lock+0xfb/0x122 [] _spin_lock+0x4e/0x5f [] res_counter_uncharge+0x2a/0x70 [] res_counter_uncharge+0x2a/0x70 [] swap_info_get+0x6a/0xa3 [] mem_cgroup_uncharge_swap+0x2a/0x35 [] swap_entry_free+0x8f/0x93 [] swap_free+0x19/0x28 [] delete_from_swap_cache+0x36/0x43 [] free_swap_and_cache+0xb1/0xeb [] unmap_vmas+0x57f/0x837 [] exit_mmap+0xa5/0x11c [] mmput+0x41/0x9f [] exit_mm+0x102/0x10d [] do_exit+0x1a2/0x73e [] __dequeue_signal+0x15/0x11c [] do_group_exit+0x73/0xa5 [] get_signal_to_deliver+0x34f/0x3a1 [] do_notify_resume+0x8c/0x7a5 [] lock_hrtimer_base+0x1b/0x3c [] getnstimeofday+0x57/0xb6 [] ktime_get_ts+0x22/0x4b [] ktime_get+0xc/0x41 [] hrtimer_nanosleep+0xa5/0xf1 [] hrtimer_wakeup+0x0/0x22 [] sysret_signal+0x7c/0xcb === This context has hold swap_lock already, so other contexts trying to hold swap_lock also get spinlock lockup bug. B: general protection fault === general protection fault: 0000 [#1] SMP last sysfs file: /sys/devices/system/cpu/cpu15/cache/index1/shared_cpu_map CPU 3 Modules linked in: ipt_REJECT xt_tcpudp iptable_filter ip_tables x_tables bridge stp ipv6 autofs4 hidp rfcomm l2cap bluetooth sunrpc dm_mirror dm_region_hash dm_log dm_multipath dm _mod sbs sbshc battery ac lp sg rtc_cmos rtc_core ide_cd_mod parport_pc rtc_lib parport se rio_raw cdrom acpi_memhotplug button e1000 i2c_i801 i2c_core shpchp pcspkr ata_piix libata megaraid_mbox megaraid_mm sd_mod scsi_mod ext3 jbd ehci_hcd ohci_hcd uhci_hcd [last unloa ded: microcode] Pid: 8051, comm: bash Not tainted 2.6.29-rc1-0ed85935 #1 RIP: 0010:[] [] _raw_spin_trylock+0x0/0x39 RSP: 0000:ffff8800bb995e00 EFLAGS: 00010092 RAX: ffff88010b54a620 RBX: 0097040900455377 RCX: 0000000000000000 RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0097040900455377 RBP: 009704090045538f R08: 0000000000000002 R09: 0000000000000001 R10: ffffe2000c861640 R11: ffffffff8026f9b5 R12: 0000000000000296 R13: 0000000000001000 R14: ffff8801003cf080 R15: 00007fa79a932028 FS: 00007fa79a9316f0(0000) GS:ffff8803af7d7a80(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 00007fa79a932028 CR3: 00000000cc8e0000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process bash (pid: 8051, threadinfo ffff8800bb994000, task ffff88010b54a620) Stack: ffffffff804d8f1e ffffffff8026f9b5 0097040900455377 0097040900455357 ffffffff8026f9b5 0000000000000282 ffff88010e1a2000 ffffe2000c861640 ffff8801094f9000 0000000000000000 ffffffff802b7371 00000001ed3e8025 Call Trace: [] ? _spin_lock+0x35/0x5f [] ? res_counter_uncharge+0x2a/0x70 [] ? res_counter_uncharge+0x2a/0x70 [] ? mem_cgroup_commit_charge_swapin+0x74/0x8a [] ? handle_mm_fault+0x5e3/0x750 [] ? do_page_fault+0x3b2/0x73e [] ? page_fault+0x1f/0x30 Code: 31 c0 e8 5f 4a ed ff 48 c7 c7 da df 5c 80 31 c0 e8 51 4a ed ff c7 05 cc d5 33 00 01 00 00 00 c7 05 62 c7 de 00 00 00 00 00 58 c3 <0f> b7 07 38 e0 8d 88 00 01 00 00 75 05 f0 6 6 0f b1 0f 0f 94 c1 RIP [] _raw_spin_trylock+0x0/0x39 RSP ---[ end trace 1ecf768aff114688 ]--- === > To handle the problem "parent may be obsolete", > > call mem_cgroup_get(parent) at create() > call mem_cgroup_put(parent) at freeing memcg. > (regardless of use_hierarchy.) > > is clearer way to go, I think. > > I wonder whether there is mis-accounting problem or not.. > > So, adding css_tryget() around problematic code can be a fix. > -- > mem = swap_cgroup_record(); > if (css_tryget(&mem->css)) { > res_counter_uncharge(&mem->memsw, PAZE_SIZE); > css_put(&mem->css) > } > -- > I like css_tryget() rather than mem_cgroup_obsolete(). I agree. The updated version is attached. Thanks, Daisuke nishimura. > To be honest, I'd like to remove memcg special stuff when I can. > > Thanks, > -Kame > === From: Daisuke Nishimura mem_cgroup_get ensures that the memcg that has been got can be accessed even after the directory has been removed, but it doesn't ensure that parents of it can be accessed: parents might have been freed already by rmdir. This causes a bug in case of use_hierarchy==1, because res_counter_uncharge climb up the tree. Check if the memcg is obsolete by css_tryget, and don't call res_counter_uncharge when obsole. Signed-off-by: Daisuke Nishimura --- mm/memcontrol.c | 15 ++++++++++++--- 1 files changed, 12 insertions(+), 3 deletions(-) diff --git a/mm/memcontrol.c b/mm/memcontrol.c index fb62b43..4e3b100 100644 --- a/mm/memcontrol.c +++ b/mm/memcontrol.c @@ -1182,7 +1182,10 @@ int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm, /* avoid double counting */ mem = swap_cgroup_record(ent, NULL); if (mem) { - res_counter_uncharge(&mem->memsw, PAGE_SIZE); + if (!css_tryget(&mem->css)) { + res_counter_uncharge(&mem->memsw, PAGE_SIZE); + css_put(&mem->css); + } mem_cgroup_put(mem); } } @@ -1252,7 +1255,10 @@ void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr) struct mem_cgroup *memcg; memcg = swap_cgroup_record(ent, NULL); if (memcg) { - res_counter_uncharge(&memcg->memsw, PAGE_SIZE); + if (!css_tryget(&memcg->css)) { + res_counter_uncharge(&memcg->memsw, PAGE_SIZE); + css_put(&memcg->css); + } mem_cgroup_put(memcg); } @@ -1397,7 +1403,10 @@ void mem_cgroup_uncharge_swap(swp_entry_t ent) memcg = swap_cgroup_record(ent, NULL); if (memcg) { - res_counter_uncharge(&memcg->memsw, PAGE_SIZE); + if (!css_tryget(&memcg->css)) { + res_counter_uncharge(&memcg->memsw, PAGE_SIZE); + css_put(&memcg->css); + } mem_cgroup_put(memcg); } } -- 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/