linux-kernel - Re: RCU stall/SOFT-Lockup on 4.11.3/4.13.11 after multiple days uptime

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>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20171112120915.3072b927@neptune.home>
Date:   Sun, 12 Nov 2017 12:09:28 +0100
From:   Bruno Prémont <bonbons@...ophe.eu>
To:     "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
Cc:     Josh Triplett <josh@...htriplett.org>,
        Steven Rostedt <rostedt@...dmis.org>,
        Mathieu Desnoyers <mathieu.desnoyers@...icios.com>,
        Lai Jiangshan <jiangshanlai@...il.com>,
        linux-kernel@...r.kernel.org
Subject: Re: RCU stall/SOFT-Lockup on 4.11.3/4.13.11 after multiple days
 uptime

On Sat, 11 November 2017 "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com> wrote:
> On Sat, Nov 11, 2017 at 08:38:32PM +0100, Bruno Prémont wrote:
> > Hi,
> > 
> > On a single-CPU KVM-based virtual machine I'm suffering from RCU stall
> > and soft-lockup. 4.10.x kernels run fine (4.10.12) but starting with
> > 4.11.x (4.11.3, 4.13.11) I'm getting system freezes for no apparent
> > reason.
> > 
> > All info I have is following console dump (from 4.13.11):
> > [526415.290012] INFO: rcu_sched self-detected stall on CPU
> > [526415.290012] o0-...: (745847 ticks this GP) idle=ba2/2/0 softirq=37393463/37393463 fqs=0
> > [526415.290012] o (t=745854 jiffies g=23779976 c=23779975 q=32)
> > [526415.290012] rcu_sched kthread starved for 745854 jiffies! g23779976 c23779975 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x0  
> 
> The above line says that the rcu_sched kthread asked to sleep for three
> jiffies, but ended up sleeping for more than 745,854 jiffies.
> 
> If your system does not let the RCU's kernel threads run, RCU cannot
> help you much.
> 
> The ->state of 0x0 indicates that the kthread is in fact runnable, but
> did not get a chance to run.  Was the system heavily loaded to the
> point where you would expect a kthread to remain preempted for many
> minutes?
> 
> I am guessing that the answer is no, given that CPU 0 is actually idle
> (idle=ba2/2/0).  Seems unlikely, but I have to ask:  Did you bind the
> kthread to a specific CPU?

The system should be lightly loaded (about 5-10% CPU usage on average), so
plenty of time for RCU to do its work.

I didn't bind processes (be it userspace process or kthread) to a specific
CPU, thus it's all auto-configured.

I guess the question then is what is the system busy with or waiting for
that prevents RCU to get its work done...
Shouldn't the watchdog print a trace of where CPU#0 is stuck? If so I might need
to check at which log level and make sure that loglevel reaches console.
Nothing did hit the disk though.

Thanks,
Bruno

> 							Thanx, Paul
> 
> > [526440.020015] watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [swapper/0:0]
> > [526468.020005] watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [swapper/0:0]
> > [526478.320009] INFO: rcu_sched self-detected stall on CPU
> > [526478.320009] o0-...: (752143 ticks this GP) idle=ba2/2/0 softirq=37393463/37393463 fqs=0
> > [526478.320009] o (t=752157 jiffies g=23779976 c=23779975 q=32)
> > [526478.320009] rcu_sched kthread starved for 752157 jiffies! g23779976 c23779975 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x0
> > [526504.020016] watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [swapper/0:0]
> > [526532.020007] watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [swapper/0:0]
> > ...
> > 
> > Attached is kernel config (4.13.11).
> > 
> > 
> > The output obtained with 4.11.3 was:
> > [  280.680010] INFO: rcu_sched self-detected stall on CPU
> > [  280.680021] o0-...: (27312 ticks this GP) dile=b11/2/0 softirq=6119/6119 fqs=0
> > [  280.680021] o (t=27312 jiffies g=441 c=440 q=0)
> > [  280.680021] rcu_sched_kthread starved for 27312 jiffies! g441 c440 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x0
> > ...
> > 
> > 
> > As it's a remote VM for which I don't have access to the host I have little
> > options for further digging (can't trigger sysrq's).
> > 
> > 
> > Same kernel (4.13.11) seems to be running just fine on another KVM-base VM that
> > has two CPUs.
> > 
> > 
> > Does it ring a bell or is there some info that might be of any use,
> > assuming I can obtain it?
> > 
> > Bruno  
> 
>