Re: [PATCH] leds: heartbeat: fix bug on panic

[Alexander Holler <holler@ahsoftware.de>]

Am 04.07.2012 09:05, schrieb Bryan Wu:
> On Tue, Jul 3, 2012 at 2:35 PM, Alexander Holler <holler@ahsoftware.de> wrote:
>> With commit 49dca5aebfdeadd4bf27b6cb4c60392147dc35a4 I introduced
>> a bug (visible if CONFIG_PROVE_RCU is enabled) which occures when a panic
>> has happened:
>>
>> [ 1526.520230] ===============================
>> [ 1526.520230] [ INFO: suspicious RCU usage. ]
>> [ 1526.520230] 3.5.0-rc1+ #12 Not tainted
>> [ 1526.520230] -------------------------------
>> [ 1526.520230] /c/kernel-tests/mm/include/linux/rcupdate.h:436 Illegal context switch in RCU read-side critical section!
>> [ 1526.520230]
>> [ 1526.520230] other info that might help us debug this:
>> [ 1526.520230]
>> [ 1526.520230]
>> [ 1526.520230] rcu_scheduler_active = 1, debug_locks = 0
>> [ 1526.520230] 3 locks held by net.agent/3279:
>> [ 1526.520230]  #0:  (&mm->mmap_sem){++++++}, at: [<ffffffff82f85962>] do_page_fault+0x193/0x390
>> [ 1526.520230]  #1:  (panic_lock){+.+...}, at: [<ffffffff82ed2830>] panic+0x37/0x1d3
>> [ 1526.520230]  #2:  (rcu_read_lock){.+.+..}, at: [<ffffffff810b9b28>] rcu_lock_acquire+0x0/0x29
>> [ 1526.520230]
>> [ 1526.520230] stack backtrace:
>> [ 1526.520230] Pid: 3279, comm: net.agent Not tainted 3.5.0-rc1+ #12
>> [ 1526.520230] Call Trace:
>> [ 1526.520230]  [<ffffffff810e1570>] lockdep_rcu_suspicious+0x109/0x112
>> [ 1526.520230]  [<ffffffff810bfe3a>] rcu_preempt_sleep_check+0x45/0x47
>> [ 1526.520230]  [<ffffffff810bfe5a>] __might_sleep+0x1e/0x19a
>> [ 1526.520230]  [<ffffffff82f8010e>] down_write+0x26/0x81
>> [ 1526.520230]  [<ffffffff8276a966>] led_trigger_unregister+0x1f/0x9c
>> [ 1526.520230]  [<ffffffff8276def5>] heartbeat_reboot_notifier+0x15/0x19
>> [ 1526.520230]  [<ffffffff82f85bf5>] notifier_call_chain+0x96/0xcd
>> [ 1526.520230]  [<ffffffff82f85cba>] __atomic_notifier_call_chain+0x8e/0xff
>> [ 1526.520230]  [<ffffffff81094b7c>] ? kmsg_dump+0x37/0x1eb
>> [ 1526.520230]  [<ffffffff82f85d3f>] atomic_notifier_call_chain+0x14/0x16
>> [ 1526.520230]  [<ffffffff82ed28e1>] panic+0xe8/0x1d3
>> [ 1526.520230]  [<ffffffff811473e2>] out_of_memory+0x15d/0x1d3
>>
>> So in case of a panic, now just turn of the LED. Other approaches like
>> scheduling a work to unregister the trigger aren't working because there
>> isn't much which still runs after a panic occured (except timers).
>>
>> Signed-off-by: Alexander Holler <holler@ahsoftware.de>
>> ---
>>   drivers/leds/ledtrig-heartbeat.c |   16 +++++++++++++++-
>>   1 files changed, 15 insertions(+), 1 deletions(-)
>>
>> diff --git a/drivers/leds/ledtrig-heartbeat.c b/drivers/leds/ledtrig-heartbeat.c
>> index 41dc76d..a019fbb 100644
>> --- a/drivers/leds/ledtrig-heartbeat.c
>> +++ b/drivers/leds/ledtrig-heartbeat.c
>> @@ -21,6 +21,8 @@
>>   #include <linux/reboot.h>
>>   #include "leds.h"
>>
>> +static int panic_heartbeats;
>> +
>>   struct heartbeat_trig_data {
>>          unsigned int phase;
>>          unsigned int period;
>> @@ -34,6 +36,11 @@ static void led_heartbeat_function(unsigned long data)
>>          unsigned long brightness = LED_OFF;
>>          unsigned long delay = 0;
>>
>> +       if (unlikely(panic_heartbeats)) {
>> +               led_set_brightness(led_cdev, LED_OFF);
>> +               return;
>> +       }
>> +
>>          /* acts like an actual heart beat -- ie thump-thump-pause... */
>>          switch (heartbeat_data->phase) {
>>          case 0:
>> @@ -111,12 +118,19 @@ static int heartbeat_reboot_notifier(struct notifier_block *nb,
>>          return NOTIFY_DONE;
>>   }
>>
>> +static int heartbeat_panic_notifier(struct notifier_block *nb,
>> +                                    unsigned long code, void *unused)
>> +{
>> +       panic_heartbeats = 1;
>
> Can we just set LED as OFF and delete the timer here? because timer is
> also useless after a kernel panic.
> So we don't need this global static variable here.

No, the necessary information (heartbeat_trig_data) isn't available here.

Regards,

Alexander