From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757265AbbAZXrA (ORCPT ); Mon, 26 Jan 2015 18:47:00 -0500 Received: from userp1040.oracle.com ([156.151.31.81]:30126 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757218AbbAZXq4 (ORCPT ); Mon, 26 Jan 2015 18:46:56 -0500 Message-ID: <54C6D1C4.4070802@oracle.com> Date: Mon, 26 Jan 2015 18:46:12 -0500 From: Sasha Levin User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.3.0 MIME-Version: 1.0 To: Kirill Tkhai , linux-kernel@vger.kernel.org CC: Peter Zijlstra , Oleg Nesterov , Ingo Molnar , Burke Libbey , Vladimir Davydov , Kirill Tkhai Subject: Re: [PATCH] sched: Fix race between task_group and sched_task_group References: <1414405105.19914.169.camel@tkhai> In-Reply-To: <1414405105.19914.169.camel@tkhai> Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit X-Source-IP: ucsinet21.oracle.com [156.151.31.93] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 10/27/2014 06:18 AM, Kirill Tkhai wrote: > The race may happen when somebody is changing task_group of a forking task. > Child's cgroup is the same as parent's after dup_task_struct() (there just > memory copying). Also, cfs_rq and rt_rq are the same as parent's. > > But if parent changes its task_group before it's called cgroup_post_fork(), > we do not reflect this situation on child. Child's cfs_rq and rt_rq remain > the same, while child's task_group changes in cgroup_post_fork(). > > To fix this we introduce fork() method, which calls sched_move_task() directly. > This function changes sched_task_group on appropriate (also its logic has > no problem with freshly created tasks, so we shouldn't introduce something > special; we are able just to use it). > > Possibly, this decides the Burke Libbey's problem: https://lkml.org/lkml/2014/10/24/456 Hi, This seems to cause the following lockdep warning: [ 3517.958378] ====================================================== [ 3517.959661] [ INFO: possible circular locking dependency detected ] [ 3517.960172] 3.19.0-rc5-next-20150123-sasha-00063-gf82b1d7 #1824 Not tainted [ 3517.960172] ------------------------------------------------------- [ 3517.960172] trinity-c7/29839 is trying to acquire lock: [ 3517.960172] (&(&base->lock)->rlock){-.-.-.}, at: lock_timer_base.isra.19 (kernel/time/timer.c:751) [ 3517.960172] [ 3517.960172] but task is already holding lock: [ 3517.960172] (&ctx->lock){-.-.-.}, at: perf_event_context_sched_in (kernel/events/core.c:2600) [ 3517.960172] [ 3517.960172] which lock already depends on the new lock. [ 3517.960172] [ 3517.960172] [ 3517.960172] the existing dependency chain (in reverse order) is: [ 3517.960172] -> #5 (&ctx->lock){-.-.-.}: [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] _raw_spin_lock (include/linux/spinlock_api_smp.h:145 kernel/locking/spinlock.c:151) [ 3517.960172] __perf_event_task_sched_out (kernel/events/core.c:2434 kernel/events/core.c:2460) [ 3517.960172] __schedule (include/linux/perf_event.h:730 kernel/sched/core.c:2209 kernel/sched/core.c:2333 kernel/sched/core.c:2823) [ 3517.960172] schedule (kernel/sched/core.c:2853) [ 3517.960172] p9_client_rpc (net/9p/client.c:756 (discriminator 13)) [ 3517.960172] p9_client_read (net/9p/client.c:1582) [ 3517.960172] v9fs_fid_readn (fs/9p/vfs_file.c:386) [ 3517.960172] v9fs_fid_readpage (fs/9p/vfs_addr.c:71) [ 3517.960172] v9fs_vfs_readpage (fs/9p/vfs_addr.c:105) [ 3517.960172] filemap_fault (mm/filemap.c:1763 mm/filemap.c:1944) [ 3517.960172] __do_fault (mm/memory.c:2654) [ 3517.960172] handle_mm_fault (mm/memory.c:2842 mm/memory.c:3143 mm/memory.c:3267 mm/memory.c:3296) [ 3517.960172] __do_page_fault (arch/x86/mm/fault.c:1233) [ 3517.960172] trace_do_page_fault (arch/x86/mm/fault.c:1325 include/linux/jump_label.h:114 include/linux/context_tracking_state.h:27 include/linux/context_tracking.h:45 arch/x86/mm/fault.c:1326) [ 3517.960172] do_async_page_fault (arch/x86/kernel/kvm.c:280) [ 3517.960172] async_page_fault (arch/x86/kernel/entry_64.S:1286) [ 3517.960172] -> #4 (&rq->lock){-.-.-.}: [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] _raw_spin_lock (include/linux/spinlock_api_smp.h:145 kernel/locking/spinlock.c:151) [ 3517.960172] task_rq_lock (kernel/sched/core.c:344) [ 3517.960172] sched_move_task (kernel/sched/core.c:7556) [ 3517.960172] cpu_cgroup_fork (kernel/sched/core.c:8003) [ 3517.960172] cgroup_post_fork (kernel/cgroup.c:5239 (discriminator 2)) [ 3517.960172] copy_process (kernel/fork.c:1544) [ 3517.960172] do_fork (kernel/fork.c:1653) [ 3517.960172] kernel_thread (kernel/fork.c:1702) [ 3517.960172] rest_init (init/main.c:406) [ 3517.960172] start_kernel (init/main.c:503) [ 3517.960172] x86_64_start_reservations (arch/x86/kernel/head64.c:199) [ 3517.960172] x86_64_start_kernel (arch/x86/kernel/head64.c:188) [ 3517.960172] -> #3 (&p->pi_lock){-.-.-.}: [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] _raw_spin_lock_irqsave (include/linux/spinlock_api_smp.h:119 kernel/locking/spinlock.c:159) [ 3517.960172] try_to_wake_up (kernel/sched/core.c:1704) [ 3517.960172] default_wake_function (kernel/sched/core.c:2995) [ 3517.960172] autoremove_wake_function (kernel/sched/wait.c:295) [ 3517.960172] __wake_up_common (kernel/sched/wait.c:73) [ 3517.960172] __wake_up (include/linux/spinlock.h:372 kernel/sched/wait.c:96) [ 3517.960172] wakeup_kswapd (mm/vmscan.c:3519) [ 3517.960172] __alloc_pages_nodemask (mm/page_alloc.c:2530 mm/page_alloc.c:2628 mm/page_alloc.c:2849) [ 3517.960172] new_page_node (mm/migrate.c:1180) [ 3517.960172] migrate_pages (mm/migrate.c:913 mm/migrate.c:1112) [ 3517.960172] SYSC_move_pages (mm/migrate.c:1265 mm/migrate.c:1340 mm/migrate.c:1495) [ 3517.960172] SyS_move_pages (mm/migrate.c:1440) [ 3517.960172] tracesys_phase2 (arch/x86/kernel/entry_64.S:530) [ 3517.960172] -> #2 (&pgdat->kswapd_wait){..-.-.}: [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] _raw_spin_lock_irqsave (include/linux/spinlock_api_smp.h:119 kernel/locking/spinlock.c:159) [ 3517.960172] __wake_up (kernel/sched/wait.c:95) [ 3517.960172] wakeup_kswapd (mm/vmscan.c:3519) [ 3517.960172] __alloc_pages_nodemask (mm/page_alloc.c:2530 mm/page_alloc.c:2628 mm/page_alloc.c:2849) [ 3517.960172] alloc_pages_current (mm/mempolicy.c:2147) [ 3517.960172] __get_free_pages (mm/page_alloc.c:2885) [ 3517.960172] alloc_loc_track (mm/slub.c:3963) [ 3517.960172] process_slab (mm/slub.c:4029 mm/slub.c:4063) [ 3517.960172] list_locations (mm/slub.c:4095 (discriminator 3)) [ 3517.960172] alloc_calls_show (mm/slub.c:4700) [ 3517.960172] slab_attr_show (mm/slub.c:4950) [ 3517.960172] sysfs_kf_seq_show (fs/sysfs/file.c:64) [ 3517.960172] kernfs_seq_show (fs/kernfs/file.c:169) [ 3517.960172] seq_read (fs/seq_file.c:228) [ 3517.960172] kernfs_fop_read (fs/kernfs/file.c:251) [ 3517.960172] __vfs_read (fs/read_write.c:430) [ 3517.960172] vfs_read (fs/read_write.c:446) [ 3517.960172] kernel_read (fs/exec.c:819) [ 3517.960172] copy_module_from_fd.isra.28 (kernel/module.c:2547) [ 3517.960172] SyS_finit_module (kernel/module.c:3429 kernel/module.c:3413) [ 3517.960172] tracesys_phase2 (arch/x86/kernel/entry_64.S:530) [ 3517.960172] -> #1 (&(&n->list_lock)->rlock){-.-.-.}: [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] _raw_spin_lock (include/linux/spinlock_api_smp.h:145 kernel/locking/spinlock.c:151) [ 3517.960172] get_partial_node.isra.35 (mm/slub.c:1630) [ 3517.960172] __slab_alloc (mm/slub.c:1737 mm/slub.c:2207 mm/slub.c:2378) [ 3517.960172] kmem_cache_alloc (mm/slub.c:2459 mm/slub.c:2506) [ 3517.960172] __debug_object_init (include/linux/slab.h:572 lib/debugobjects.c:99 lib/debugobjects.c:312) [ 3517.960172] debug_object_init (lib/debugobjects.c:365) [ 3517.960172] timer_fixup_activate (kernel/time/timer.c:512) [ 3517.960172] debug_object_activate (lib/debugobjects.c:280 lib/debugobjects.c:439) [ 3517.960172] mod_timer (kernel/time/timer.c:589 include/linux/jump_label.h:114 include/trace/events/timer.h:44 kernel/time/timer.c:641 kernel/time/timer.c:778 kernel/time/timer.c:897) [ 3517.960172] isicom_init (drivers/tty/isicom.c:1708) [ 3517.960172] do_one_initcall (init/main.c:798) [ 3517.960172] kernel_init_freeable (init/main.c:863 init/main.c:871 init/main.c:890 init/main.c:1011) [ 3517.960172] kernel_init (init/main.c:943) [ 3517.960172] ret_from_fork (arch/x86/kernel/entry_64.S:349) [ 3517.960172] -> #0 (&(&base->lock)->rlock){-.-.-.}: [ 3517.960172] __lock_acquire (kernel/locking/lockdep.c:1842 kernel/locking/lockdep.c:1947 kernel/locking/lockdep.c:2133 kernel/locking/lockdep.c:3184) [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] _raw_spin_lock_irqsave (include/linux/spinlock_api_smp.h:119 kernel/locking/spinlock.c:159) [ 3517.960172] lock_timer_base.isra.19 (kernel/time/timer.c:751) [ 3517.960172] mod_timer (kernel/time/timer.c:774 kernel/time/timer.c:897) [ 3517.960172] add_timer (kernel/time/timer.c:947) [ 3517.960172] __queue_delayed_work (kernel/workqueue.c:1452) [ 3517.960172] queue_delayed_work_on (kernel/workqueue.c:1480) [ 3517.960172] schedule_orphans_remove (kernel/events/core.c:1416) [ 3517.960172] event_sched_in.isra.47 (kernel/events/core.c:1793) [ 3517.960172] group_sched_in (kernel/events/core.c:1816) [ 3517.960172] ctx_sched_in (kernel/events/core.c:2547 kernel/events/core.c:2578) [ 3517.960172] perf_event_sched_in (kernel/events/core.c:1930) [ 3517.960172] perf_event_context_sched_in (kernel/events/core.c:2613) [ 3517.960172] __perf_event_task_sched_in (kernel/events/core.c:2703) [ 3517.960172] finish_task_switch (include/linux/perf_event.h:721 kernel/sched/core.c:2257) [ 3517.960172] __schedule (kernel/sched/core.c:2368 kernel/sched/core.c:2823) [ 3517.960172] schedule_user (kernel/sched/core.c:2852 include/linux/jump_label.h:114 include/linux/context_tracking_state.h:27 include/linux/context_tracking.h:45 kernel/sched/core.c:2871) [ 3517.960172] retint_careful (arch/x86/kernel/entry_64.S:905) [ 3517.960172] [ 3517.960172] other info that might help us debug this: [ 3517.960172] [ 3517.960172] Chain exists of: &(&base->lock)->rlock --> &rq->lock --> &ctx->lock [ 3517.960172] Possible unsafe locking scenario: [ 3517.960172] [ 3517.960172] CPU0 CPU1 [ 3517.960172] ---- ---- [ 3517.960172] lock(&ctx->lock); [ 3517.960172] lock(&rq->lock); [ 3517.960172] lock(&ctx->lock); [ 3517.960172] lock(&(&base->lock)->rlock); [ 3517.960172] [ 3517.960172] *** DEADLOCK *** [ 3517.960172] [ 3517.960172] 2 locks held by trinity-c7/29839: [ 3517.960172] #0: (&cpuctx_lock){-.-.-.}, at: perf_event_context_sched_in (kernel/events/core.c:340 kernel/events/core.c:2599) [ 3517.960172] #1: (&ctx->lock){-.-.-.}, at: perf_event_context_sched_in (kernel/events/core.c:2600) [ 3517.960172] [ 3517.960172] stack backtrace: [ 3517.960172] CPU: 7 PID: 29839 Comm: trinity-c7 Not tainted 3.19.0-rc5-next-20150123-sasha-00063-gf82b1d7 #1824 [ 3517.960172] ffffffffa67fa820 00000000167dc919 ffff880353a8f700 ffffffffa0a859d2 [ 3517.960172] 0000000000000000 ffffffffa67fcef0 ffff880353a8f760 ffffffff96407cd9 [ 3517.960172] ffff88017c680138 ffffffffa6879cf0 0000000000000000 ffff880353a90000 [ 3517.960172] Call Trace: [ 3517.960172] dump_stack (lib/dump_stack.c:52) [ 3517.960172] print_circular_bug (kernel/locking/lockdep.c:1217) [ 3517.960172] __lock_acquire (kernel/locking/lockdep.c:1842 kernel/locking/lockdep.c:1947 kernel/locking/lockdep.c:2133 kernel/locking/lockdep.c:3184) [ 3517.960172] ? debug_check_no_locks_freed (kernel/locking/lockdep.c:3051) [ 3517.960172] ? __lock_acquire (kernel/locking/lockdep.c:3144) [ 3517.960172] lock_acquire (kernel/locking/lockdep.c:3604) [ 3517.960172] ? lock_timer_base.isra.19 (kernel/time/timer.c:751) [ 3517.960172] _raw_spin_lock_irqsave (include/linux/spinlock_api_smp.h:119 kernel/locking/spinlock.c:159) [ 3517.960172] ? lock_timer_base.isra.19 (kernel/time/timer.c:751) [ 3517.960172] lock_timer_base.isra.19 (kernel/time/timer.c:751) [ 3517.960172] mod_timer (kernel/time/timer.c:774 kernel/time/timer.c:897) [ 3517.960172] ? x86_pmu_start_txn (arch/x86/kernel/cpu/perf_event.c:1024) [ 3517.960172] ? msleep (kernel/time/timer.c:886) [ 3517.960172] add_timer (kernel/time/timer.c:947) [ 3517.960172] __queue_delayed_work (kernel/workqueue.c:1452) [ 3517.960172] queue_delayed_work_on (kernel/workqueue.c:1480) [ 3517.960172] schedule_orphans_remove (kernel/events/core.c:1416) [ 3517.960172] event_sched_in.isra.47 (kernel/events/core.c:1793) [ 3517.960172] ? sched_clock (./arch/x86/include/asm/paravirt.h:192 arch/x86/kernel/tsc.c:304) [ 3517.960172] group_sched_in (kernel/events/core.c:1816) [ 3517.960172] ? sched_clock_cpu (kernel/sched/clock.c:311) [ 3517.960172] ctx_sched_in (kernel/events/core.c:2547 kernel/events/core.c:2578) [ 3517.960172] ? perf_event_context_sched_in (kernel/events/core.c:2600) [ 3517.960172] perf_event_sched_in (kernel/events/core.c:1930) [ 3517.960172] perf_event_context_sched_in (kernel/events/core.c:2613) [ 3517.960172] __perf_event_task_sched_in (kernel/events/core.c:2703) [ 3517.960172] ? perf_pmu_enable (kernel/events/core.c:2694) [ 3517.960172] finish_task_switch (include/linux/perf_event.h:721 kernel/sched/core.c:2257) [ 3517.960172] __schedule (kernel/sched/core.c:2368 kernel/sched/core.c:2823) [ 3517.960172] ? trace_hardirqs_on_caller (kernel/locking/lockdep.c:2554 kernel/locking/lockdep.c:2601) [ 3517.960172] schedule_user (kernel/sched/core.c:2852 include/linux/jump_label.h:114 include/linux/context_tracking_state.h:27 include/linux/context_tracking.h:45 kernel/sched/core.c:2871) [ 3517.960172] retint_careful (arch/x86/kernel/entry_64.S:905) Thanks, Sasha