Date:   Fri, 30 Sep 2022 17:57:17 +0800
From:   "Zhang Xincheng" <zhangxincheng@...ontech.com>
To:     "maz" <maz@...nel.org>
Cc:     "tglx" <tglx@...utronix.de>,
        "linux-kernel" <linux-kernel@...r.kernel.org>,
        "oleksandr" <oleksandr@...alenko.name>,
        "Hans de Goede" <hdegoede@...hat.com>,
        "bigeasy" <bigeasy@...utronix.de>,
        "mark.rutland" <mark.rutland@....com>,
        "michael" <michael@...le.cc>
Subject: Re: [PATCH] interrupt: discover and disable very frequent interrupts

> Irrespective of the patch itself, I would really like to understand
> why you consider that it is a better course of action to kill a device
> (and potentially the whole machine) than to let the storm eventually
> calm down? A frequent interrupt is not necessarily the sign of
> something going wrong. It is the sign of a busy system. I prefer my
> systems busy rather than dead.

Because I found that some peripherals will send interrupts to the CPU very frequently 
in some cases, and the interrupts will be handled correctly, which will cause the CPU 
to do nothing but handle the interrupts. At the same time, the RCU system will report 
the following logs:

[  838.131628] rcu: INFO: rcu_sched self-detected stall on CPU
[  838.137189] rcu:     0-....: (194839 ticks this GP) idle=f02/1/0x4000000000000004 softirq=9993/9993 fqs=97428 
[  838.146912] rcu:      (t=195015 jiffies g=6773 q=0)
[  838.151516] Task dump for CPU 0:
[  838.154730] systemd-sleep   R  running task        0  3445      1 0x0000000a
[  838.161764] Call trace:
[  838.164198]  dump_backtrace+0x0/0x190
[  838.167846]  show_stack+0x14/0x20
[  838.171148]  sched_show_task+0x134/0x160
[  838.175057]  dump_cpu_task+0x40/0x4c
[  838.178618]  rcu_dump_cpu_stacks+0xc4/0x108
[  838.182788]  rcu_check_callbacks+0x6e4/0x898
[  838.187044]  update_process_times+0x2c/0x88
[  838.191214]  tick_sched_handle.isra.5+0x3c/0x50
[  838.195730]  tick_sched_timer+0x48/0x98
[  838.199552]  __hrtimer_run_queues+0xec/0x2f8
[  838.203808]  hrtimer_interrupt+0x10c/0x298
[  838.207891]  arch_timer_handler_phys+0x2c/0x38
[  838.212321]  handle_percpu_devid_irq+0x88/0x228
[  838.216837]  generic_handle_irq+0x2c/0x40
[  838.220833]  __handle_domain_irq+0x60/0xb8
[  838.224915]  gic_handle_irq+0x7c/0x178
[  838.228650]  el1_irq+0xb0/0x140
[  838.231778]  __do_softirq+0x84/0x2e8
[  838.235340]  irq_exit+0x9c/0xb8
[  838.238468]  __handle_domain_irq+0x64/0xb8
[  838.242550]  gic_handle_irq+0x7c/0x178
[  838.246285]  el1_irq+0xb0/0x140
[  838.249413]  resume_irqs+0xfc/0x148
[  838.252888]  resume_device_irqs+0x10/0x18
[  838.256883]  dpm_resume_noirq+0x10/0x20
[  838.260706]  suspend_devices_and_enter+0x170/0x788
[  838.265483]  pm_suspend+0x41c/0x4cc
[  838.268958]  state_store+0xbc/0x160
[  838.272433]  kobj_attr_store+0x14/0x28
[  838.276168]  sysfs_kf_write+0x40/0x50
[  838.279817]  kernfs_fop_write+0xcc/0x1e0
[  838.283726]  __vfs_write+0x18/0x140
[  838.287201]  vfs_write+0xa4/0x1b0
[  838.290503]  ksys_write+0x4c/0xb8
[  838.293804]  __arm64_sys_write+0x18/0x20
[  838.297713]  el0_svc_common+0x90/0x178
[  838.301449]  el0_svc_handler+0x9c/0xa8
[  838.305184]  el0_svc+0x8/0xc

The log is from the process of waking up a sleeping machine, 
I left the machine in this state for a night and it successfully woke up,
 and then I saw from /proc/interrupts that a GPIO interrupt triggered 
more than 13 billion times.

29: 1368200001  0  0  0  0  0  0  0     phytium_gpio6   Edge ACPI:Event

> Something like this should be limited to a debug feature. It would
> also be a lot more useful if it was built as an interrupt *limiting*
> feature, rather then killing the interrupt forever (which is IMHO a
> ludicrous thing to do).

It's a good idea to have it as a debugging feature.


Thanks,

             Zhang Xincheng

Powered by blists - more mailing lists