lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [day] [month] [year] [list]
Message-ID: <CAD5jUk_E516CvKGUWcfZ7chRjv=8fgSBHppNX7YTpSLf0wk_4Q@mail.gmail.com>
Date:   Fri, 29 Nov 2019 17:52:22 -0500
From:   Radu Rendec <radu.rendec@...il.com>
To:     linux-kernel@...r.kernel.org
Subject: RCU stall on 4.9.191/CONFIG_PREEMPT, single CPU

Hi Everyone,

I'm experiencing what appears to be a complete system freeze on a single
CPU embedded system (PowerPC, MPC8378) running Linux 4.9.191 with
CONFIG_PREEMPT enabled.

Enabling all lockup detection in the kernel config revealed RCU stalls,
detected by rcu_preempt (sample included below). These occur every 30
seconds, give or take.

At that point the network stack is hosed (no replies to ARP requests).
Softirqs seem to be hosed as well - I added a test module that logs a
message every 0.5 seconds from a timer callback and, once the "freeze"
occurs, these messages are never printed.

I simply do not understand the RCU and scheduler implementation well
enough to figure this out on my own, so any idea or suggestion would be
helpful.

Looking at Documentation/RCU/stallwarn.txt, I found this among the
potential causes for an RCU stall:

    A CPU-bound real-time task in a CONFIG_PREEMPT kernel, which might
    happen to preempt a low-priority task in the middle of an RCU
    read-side critical section.   This is especially damaging if
    that low-priority task is not permitted to run on any other CPU,
    in which case the next RCU grace period can never complete

In my case all tasks are CPU-bound, simply because there is only one
CPU. And yes, I do have realtime tasks (many of them), so any of them
can potentially preempt a low-priority task in the middle of an RCU read
critical section. Does that mean it's a bad idea to run a fully
preemptible kernel on a single CPU? If that were the case, I would
expect to see this kind of problems much sooner/more often, but they are
only triggered by a very specific activity of the application that runs
on top.

Thanks,
Radu Rendec

[  902.232175] INFO: rcu_preempt detected stalls on CPUs/tasks:
[  902.237847]  (detected by 0, t=21002 jiffies, g=8316, c=8315, q=210)
[  902.244203] All QSes seen, last rcu_preempt kthread activity 21002
(602030-581028), jiffies_till_next_fqs=3, root ->qsmask 0x0
[  902.255586] backupRestore   R  running task        0  1525   1254 0x00000000
[  902.262633] Call Trace:
[  902.265093] [9ac0fc60] [8005b2f4] rcu_check_callbacks+0x848/0x860
(unreliable)
[  902.272319] [9ac0fcd0] [8005e878] update_process_times+0x30/0x60
[  902.278328] [9ac0fce0] [8007046c] tick_sched_timer+0x58/0xb4
[  902.283985] [9ac0fd10] [8005f844]
__hrtimer_run_queues.constprop.34+0x100/0x18c
[  902.291291] [9ac0fd50] [8005fae0] hrtimer_interrupt+0xc0/0x290
[  902.297122] [9ac0fda0] [8000a3c8] __timer_interrupt+0x15c/0x20c
[  902.303038] [9ac0fdb0] [8000a610] timer_interrupt+0x9c/0xc8
[  902.308614] [9ac0fdd0] [8000f678] ret_from_except+0x0/0x14
[  902.314106] --- interrupt: 901 at timespec64_add_safe+0x108/0x12c
[  902.314106]     LR = poll_select_set_timeout+0x94/0x110
[  902.325404] [9ac0fe90] [111294e8] 0x111294e8 (unreliable)
[  902.330804] [9ac0feb0] [800e8a1c] poll_select_set_timeout+0x50/0x110
[  902.337155] [9ac0ff10] [800e9d70] SyS_poll+0x68/0x114
[  902.342205] [9ac0ff40] [8000ef6c] ret_from_syscall+0x0/0x38
[  902.347773] --- interrupt: c01 at 0xe5855f8
[  902.347773]     LR = 0xe5855e0
[  902.354994] rcu_preempt kthread starved for 21002 jiffies! g8316
c8315 f0x2 RCU_GP_WAIT_FQS(3) ->state=0x0
[  902.364638] rcu_preempt     R  running task        0     7      2 0x00000800
[  902.371686] Call Trace:
[  902.374126] [9f043d00] [9ac08080] 0x9ac08080 (unreliable)
[  902.379526] [9f043dc0] [80415ca4] __schedule+0x278/0x490
[  902.384835] [9f043e10] [8041608c] schedule+0x3c/0xac
[  902.389799] [9f043e20] [80418e9c] schedule_timeout+0x128/0x240
[  902.395630] [9f043e60] [8005a0d4] rcu_gp_kthread+0x3dc/0xa6c
[  902.401289] [9f043ef0] [80038b7c] kthread+0xbc/0xd0
[  902.406164] [9f043f40] [8000f094] ret_from_kernel_thread+0x5c/0x64

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ