Re: [PATCH] rcu: ftrace: avoid tracing a few functions executed in multi_cpu_stop()

[Steven Rostedt <rostedt@goodmis.org>]

On Mon, 18 Apr 2022 12:37:35 +0800
Patrick Wang <patrick.wang.shcn@gmail.com> wrote:

> A few functions are in the call chain of rcu_momentary_dyntick_idle()
> which is executed in multi_cpu_stop() and marked notrace. They are running
> in traced when ftrace modify code. This may cause non-ftrace_modify_code
> CPUs stall:

I'm confused by this. How is traced functions causing this exactly? Is this
on RISC-V?

> 
> [   72.686113] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks:
> [   72.687344] rcu: 	1-...!: (0 ticks this GP) idle=14f/1/0x4000000000000000 softirq=3397/3397 fqs=0
> [   72.687800] rcu: 	3-...!: (0 ticks this GP) idle=ee9/1/0x4000000000000000 softirq=5168/5168 fqs=0
> [   72.688280] 	(detected by 0, t=8137 jiffies, g=5889, q=2 ncpus=4)
> [   72.688739] Task dump for CPU 1:
> [   72.688991] task:migration/1     state:R  running task     stack:    0 pid:   19 ppid:     2 flags:0x00000000
> [   72.689594] Stopper: multi_cpu_stop+0x0/0x18c <- stop_machine_cpuslocked+0x128/0x174
> [   72.690242] Call Trace:
> [   72.690603] Task dump for CPU 3:
> [   72.690761] task:migration/3     state:R  running task     stack:    0 pid:   29 ppid:     2 flags:0x00000000
> [   72.691135] Stopper: multi_cpu_stop+0x0/0x18c <- stop_machine_cpuslocked+0x128/0x174
> [   72.691474] Call Trace:
> [   72.691733] rcu: rcu_preempt kthread timer wakeup didn't happen for 8136 jiffies! g5889 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402
> [   72.692180] rcu: 	Possible timer handling issue on cpu=2 timer-softirq=594
> [   72.692485] rcu: rcu_preempt kthread starved for 8137 jiffies! g5889 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x402 ->cpu=2
> [   72.692876] rcu: 	Unless rcu_preempt kthread gets sufficient CPU time, OOM is now expected behavior.
> [   72.693232] rcu: RCU grace-period kthread stack dump:
> [   72.693433] task:rcu_preempt     state:I stack:    0 pid:   14 ppid:     2 flags:0x00000000
> [   72.693788] Call Trace:
> [   72.694018] [<ffffffff807f3740>] schedule+0x56/0xc2
> [   72.694306] [<ffffffff807f9cd8>] schedule_timeout+0x82/0x184
> [   72.694539] [<ffffffff8007c456>] rcu_gp_fqs_loop+0x19a/0x318
> [   72.694809] [<ffffffff8007e408>] rcu_gp_kthread+0x11a/0x140
> [   72.695325] [<ffffffff800324d6>] kthread+0xee/0x118
> [   72.695657] [<ffffffff8000398a>] ret_from_exception+0x0/0x14
> [   72.696089] rcu: Stack dump where RCU GP kthread last ran:
> [   72.696383] Task dump for CPU 2:
> [   72.696562] task:migration/2     state:R  running task     stack:    0 pid:   24 ppid:     2 flags:0x00000000
> [   72.697059] Stopper: multi_cpu_stop+0x0/0x18c <- stop_machine_cpuslocked+0x128/0x174
> [   72.697471] Call Trace:
> 
> Mark rcu_preempt_deferred_qs(), rcu_preempt_need_deferred_qs() and
> rcu_preempt_deferred_qs_irqrestore() notrace to avoid this.
> 

The rcu_momentary_dyntick_idle() was marked notrace because of RISC-V not
being able to call functions from within stop machine. If that's what is
being prevented, then I'm fine with this (although I'm thinking we need
different kinds of "notrace" for different architectures as one arch's
limitation should not be cause for another's).

But before I ack this patch, I want to understand the real issues here.

-- Steve