[<prev] [next>] [day] [month] [year] [list]
Message-ID: <22f11b9d-a33e-4ff4-8cb4-01b865069ce7@huaweicloud.com>
Date: Fri, 27 Sep 2024 17:50:49 +0800
From: Chen Ridong <chenridong@...weicloud.com>
To: Vishal Chourasia <vishalc@...ux.ibm.com>,
"tj@...nel.org >> Tejun Heo" <tj@...nel.org>,
Zefan Li <lizefan.x@...edance.com>, Waiman Long <longman@...hat.com>,
Michal Koutný <mkoutny@...e.com>
Cc: cgroups@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v5 1/3] cgroup/bpf: use a dedicated workqueue for cgroup
bpf destruction
On 2024/9/27 12:22, Vishal Chourasia wrote:
> On Mon, Sep 23, 2024 at 11:43:50AM +0000, Chen Ridong wrote:
>> I found a hung_task problem as shown below:
>>
>> INFO: task kworker/0:0:8 blocked for more than 327 seconds.
>> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
>> Workqueue: events cgroup_bpf_release
>> Call Trace:
>> <TASK>
>> __schedule+0x5a2/0x2050
>> ? find_held_lock+0x33/0x100
>> ? wq_worker_sleeping+0x9e/0xe0
>> schedule+0x9f/0x180
>> schedule_preempt_disabled+0x25/0x50
>> __mutex_lock+0x512/0x740
>> ? cgroup_bpf_release+0x1e/0x4d0
>> ? cgroup_bpf_release+0xcf/0x4d0
>> ? process_scheduled_works+0x161/0x8a0
>> ? cgroup_bpf_release+0x1e/0x4d0
>> ? mutex_lock_nested+0x2b/0x40
>> ? __pfx_delay_tsc+0x10/0x10
>> mutex_lock_nested+0x2b/0x40
>> cgroup_bpf_release+0xcf/0x4d0
>> ? process_scheduled_works+0x161/0x8a0
>> ? trace_event_raw_event_workqueue_execute_start+0x64/0xd0
>> ? process_scheduled_works+0x161/0x8a0
>> process_scheduled_works+0x23a/0x8a0
>> worker_thread+0x231/0x5b0
>> ? __pfx_worker_thread+0x10/0x10
>> kthread+0x14d/0x1c0
>> ? __pfx_kthread+0x10/0x10
>> ret_from_fork+0x59/0x70
>> ? __pfx_kthread+0x10/0x10
>> ret_from_fork_asm+0x1b/0x30
>> </TASK>
>>
>> This issue can be reproduced by the following pressuse test:
>> 1. A large number of cpuset cgroups are deleted.
>> 2. Set cpu on and off repeatly.
>> 3. Set watchdog_thresh repeatly.
>> The scripts can be obtained at LINK mentioned above the signature.
>>
>> The reason for this issue is cgroup_mutex and cpu_hotplug_lock are
>> acquired in different tasks, which may lead to deadlock.
>> It can lead to a deadlock through the following steps:
>> 1. A large number of cpusets are deleted asynchronously, which puts a
>> large number of cgroup_bpf_release works into system_wq. The max_active
>> of system_wq is WQ_DFL_ACTIVE(256). Consequently, all active works are
>> cgroup_bpf_release works, and many cgroup_bpf_release works will be put
>> into inactive queue. As illustrated in the diagram, there are 256 (in
>> the acvtive queue) + n (in the inactive queue) works.
>> 2. Setting watchdog_thresh will hold cpu_hotplug_lock.read and put
>> smp_call_on_cpu work into system_wq. However step 1 has already filled
>> system_wq, 'sscs.work' is put into inactive queue. 'sscs.work' has
>> to wait until the works that were put into the inacvtive queue earlier
>> have executed (n cgroup_bpf_release), so it will be blocked for a while.
>> 3. Cpu offline requires cpu_hotplug_lock.write, which is blocked by step 2.
>> 4. Cpusets that were deleted at step 1 put cgroup_release works into
>> cgroup_destroy_wq. They are competing to get cgroup_mutex all the time.
>> When cgroup_metux is acqured by work at css_killed_work_fn, it will
>> call cpuset_css_offline, which needs to acqure cpu_hotplug_lock.read.
>> However, cpuset_css_offline will be blocked for step 3.
>> 5. At this moment, there are 256 works in active queue that are
>> cgroup_bpf_release, they are attempting to acquire cgroup_mutex, and as
>> a result, all of them are blocked. Consequently, sscs.work can not be
>> executed. Ultimately, this situation leads to four processes being
>> blocked, forming a deadlock.
>>
>> system_wq(step1) WatchDog(step2) cpu offline(step3) cgroup_destroy_wq(step4)
>> ...
>> 2000+ cgroups deleted asyn
>> 256 actives + n inactives
>> __lockup_detector_reconfigure
>> P(cpu_hotplug_lock.read)
>> put sscs.work into system_wq
>> 256 + n + 1(sscs.work)
>> sscs.work wait to be executed
>> warting sscs.work finish
>> percpu_down_write
>> P(cpu_hotplug_lock.write)
>> ...blocking...
>> css_killed_work_fn
>> P(cgroup_mutex)
>> cpuset_css_offline
>> P(cpu_hotplug_lock.read)
>> ...blocking...
>> 256 cgroup_bpf_release
>> mutex_lock(&cgroup_mutex);
>> ..blocking...
>>
>> To fix the problem, place cgroup_bpf_release works on a dedicated
>> workqueue which can break the loop and solve the problem. System wqs are
>> for misc things which shouldn't create a large number of concurrent work
>> items. If something is going to generate >WQ_DFL_ACTIVE(256) concurrent
>> work items, it should use its own dedicated workqueue.
>>
>> Fixes: 4bfc0bb2c60e ("bpf: decouple the lifetime of cgroup_bpf from cgroup itself")
>> Link: https://lore.kernel.org/cgroups/e90c32d2-2a85-4f28-9154-09c7d320cb60@huawei.com/T/#t
>> Signed-off-by: Chen Ridong <chenridong@...wei.com>
> Tested-by: Vishal Chourasia <vishalc@...ux.ibm.com>
>
> Thank you Chen, for sharing the details on how to reproduce, and for the
> patchset.
>
> Steps I followed to reproduce:
> 1) run cgroup-make.sh
> 2) run hotplug.sh
> 3) run watchdog.sh
>
> # cat cgroup-make.sh
> #!/bin/bash
>
> echo 30 > /proc/sys/kernel/hung_task_timeout_secs
> cat /proc/sys/kernel/hung_task_timeout_secs
>
> echo +cpuset > /sys/fs/cgroup/cgroup.subtree_control
> mkdir /sys/fs/cgroup/memory
> # echo +memory > /sys/fs/cgroup/memory/cgroup.subtree_control
> mkdir /sys/fs/cgroup/cpuset
> echo +cpuset > /sys/fs/cgroup/cpuset/cgroup.subtree_control
> echo +cpu > /sys/fs/cgroup/cpuset/cgroup.subtree_control
>
>
> timestamp=$(date +%s)
> echo $timestamp
> while true; do
> for i in {0..2000}; do
> mkdir /sys/fs/cgroup/cpuset/test${timestamp}_${i} &
> mkdir /sys/fs/cgroup/memory/test${timestamp}_${i} &
> done
>
> for i in {0..2000}; do
> rmdir /sys/fs/cgroup/cpuset/test${timestamp}_${i} &
> rmdir /sys/fs/cgroup/memory/test${timestamp}_${i} &
> done
> done
>
> # cat hotplug.sh
> #!/bin/bash
>
> while true
> do
> echo 1 > /sys/devices/system/cpu/cpu2/online
> echo 1 > /sys/devices/system/cpu/cpu3/online
> echo 0 > /sys/devices/system/cpu/cpu2/online
> echo 0 > /sys/devices/system/cpu/cpu3/online
> done
>
> # cat watchdog.sh
> #!/bin/bash
>
> while true
> do
> echo 12 > /proc/sys/kernel/watchdog_thresh
> echo 11 > /proc/sys/kernel/watchdog_thresh
> echo 10 > /proc/sys/kernel/watchdog_thresh
> done
>
> With these steps I able to get the hung_task timeout log messages
> INFO: task kworker/7:1:84 blocked for more than 30 seconds.
> Not tainted 6.11.0-chenridong_base-10547-g684a64bf32b6-dirty #59
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> task:kworker/7:1 state:D stack:0 pid:84 tgid:84 ppid:2 flags:0x00000000
> Workqueue: events cgroup_bpf_release
> Call Trace:
> [c00000000ee779a0] [c00000000ee779e0] 0xc00000000ee779e0 (unreliable)
> [c00000000ee77b50] [c00000000001f79c] __switch_to+0x14c/0x220
> [c00000000ee77bb0] [c0000000010e8cd0] __schedule+0x2c0/0x840
> [c00000000ee77c90] [c0000000010e9290] schedule+0x40/0x110
> [c00000000ee77d00] [c0000000010e95b0] schedule_preempt_disabled+0x20/0x30
> [c00000000ee77d20] [c0000000010ec408] __mutex_lock.constprop.0+0x5e8/0xbe0
> [c00000000ee77db0] [c000000000472f58] cgroup_bpf_release+0x98/0x3d0
> [c00000000ee77e40] [c0000000001886a8] process_one_work+0x1f8/0x520
> [c00000000ee77ef0] [c00000000018a01c] worker_thread+0x33c/0x4f0
> [c00000000ee77f90] [c0000000001970c8] kthread+0x138/0x140
> [c00000000ee77fe0] [c00000000000dd58] start_kernel_thread+0x14/0x18
> INFO: task kworker/4:1:98 blocked for more than 30 seconds.
> Not tainted 6.11.0-chenridong_base-10547-g684a64bf32b6-dirty #59
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> task:kworker/4:1 state:D stack:0 pid:98 tgid:98 ppid:2 flags:0x00000000
> Workqueue: events cgroup_bpf_release
> Call Trace:
> [c00000000ee1f9a0] [c00000000ee1f9e0] 0xc00000000ee1f9e0 (unreliable)
> [c00000000ee1fb50] [c00000000001f79c] __switch_to+0x14c/0x220
> [c00000000ee1fbb0] [c0000000010e8cd0] __schedule+0x2c0/0x840
> [c00000000ee1fc90] [c0000000010e9290] schedule+0x40/0x110
> [c00000000ee1fd00] [c0000000010e95b0] schedule_preempt_disabled+0x20/0x30
> [c00000000ee1fd20] [c0000000010ec408] __mutex_lock.constprop.0+0x5e8/0xbe0
> [c00000000ee1fdb0] [c000000000472f58] cgroup_bpf_release+0x98/0x3d0
> [c00000000ee1fe40] [c0000000001886a8] process_one_work+0x1f8/0x520
> [c00000000ee1fef0] [c00000000018a01c] worker_thread+0x33c/0x4f0
> [c00000000ee1ff90] [c0000000001970c8] kthread+0x138/0x140
> [c00000000ee1ffe0] [c00000000000dd58] start_kernel_thread+0x14/0x18
> INFO: task kworker/5:1:110 blocked for more than 30 seconds.
> Not tainted 6.11.0-chenridong_base-10547-g684a64bf32b6-dirty #59
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> task:kworker/5:1 state:D stack:0 pid:110 tgid:110 ppid:2 flags:0x00000000
> Workqueue: events cgroup_bpf_release
> Call Trace:
> [c0000000608bf9a0] [c0000000608bf9e0] 0xc0000000608bf9e0 (unreliable)
> [c0000000608bfb50] [c00000000001f79c] __switch_to+0x14c/0x220
> [c0000000608bfbb0] [c0000000010e8cd0] __schedule+0x2c0/0x840
> [c0000000608bfc90] [c0000000010e9290] schedule+0x40/0x110
> [c0000000608bfd00] [c0000000010e95b0] schedule_preempt_disabled+0x20/0x30
> [c0000000608bfd20] [c0000000010ec408] __mutex_lock.constprop.0+0x5e8/0xbe0
> [c0000000608bfdb0] [c000000000472f58] cgroup_bpf_release+0x98/0x3d0
> [c0000000608bfe40] [c0000000001886a8] process_one_work+0x1f8/0x520
> [c0000000608bfef0] [c00000000018a01c] worker_thread+0x33c/0x4f0
> [c0000000608bff90] [c0000000001970c8] kthread+0x138/0x140
> [c0000000608bffe0] [c00000000000dd58] start_kernel_thread+0x14/0x18
> Future hung task reports are suppressed, see sysctl kernel.hung_task_warnings
>
> After applying this patchset, I didn't see any log messages being
> printed in dmesg.
>
> State of the git repo:
> $git log --oneline
> a40aebb33934 (HEAD -> patches/v5_20240923_chenridong_add_dedicated_wq_for_cgroup_bpf_and_adjust_wq_max_active) workqueue: Adjust WQ_MAX_ACTIVE from 512 to 2048
> 08a2979a9e59 workqueue: doc: Add a note saturating the system_wq is not permitted
> 0e6f5ea2955f cgroup/bpf: use a dedicated workqueue for cgroup bpf destruction
> 684a64bf32b6 Merge tag 'nfs-for-6.12-1' of git://git.linux-nfs.org/projects/anna/linux-nfs
>
>
>
>> ---
>> kernel/bpf/cgroup.c | 18 +++++++++++++++++-
>> 1 file changed, 17 insertions(+), 1 deletion(-)
>>
>> diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
>> index e7113d700b87..1a7609f61d44 100644
>> --- a/kernel/bpf/cgroup.c
>> +++ b/kernel/bpf/cgroup.c
>> @@ -24,6 +24,22 @@
>> DEFINE_STATIC_KEY_ARRAY_FALSE(cgroup_bpf_enabled_key, MAX_CGROUP_BPF_ATTACH_TYPE);
>> EXPORT_SYMBOL(cgroup_bpf_enabled_key);
>>
>> +/*
>> + * cgroup bpf destruction makes heavy use of work items and there can be a lot
>> + * of concurrent destructions. Use a separate workqueue so that cgroup bpf
>> + * destruction work items don't end up filling up max_active of system_wq
>> + * which may lead to deadlock.
>> + */
>> +static struct workqueue_struct *cgroup_bpf_destroy_wq;
>> +
>> +static int __init cgroup_bpf_wq_init(void)
>> +{
>> + cgroup_bpf_destroy_wq = alloc_workqueue("cgroup_bpf_destroy", 0, 1);
>> + WARN_ON_ONCE(!cgroup_bpf_destroy_wq);
>> + return 0;
>> +}
>> +core_initcall(cgroup_bpf_wq_init);
>> +
>> /* __always_inline is necessary to prevent indirect call through run_prog
>> * function pointer.
>> */
>> @@ -334,7 +350,7 @@ static void cgroup_bpf_release_fn(struct percpu_ref *ref)
>> struct cgroup *cgrp = container_of(ref, struct cgroup, bpf.refcnt);
>>
>> INIT_WORK(&cgrp->bpf.release_work, cgroup_bpf_release);
>> - queue_work(system_wq, &cgrp->bpf.release_work);
>> + queue_work(cgroup_bpf_destroy_wq, &cgrp->bpf.release_work);
>> }
>>
>> /* Get underlying bpf_prog of bpf_prog_list entry, regardless if it's through
>> --
>> 2.34.1
Thank you for doing that.
Best regards,
Ridong
Powered by blists - more mailing lists