Re: [PATCH] x86/smp: Add irq_enter/exit() in smp_reschedule_interrupt()

From: Wanpeng Li
Date: Mon Oct 17 2016 - 00:20:03 EST


2016-10-16 21:39 GMT+08:00 Peter Zijlstra <peterz@xxxxxxxxxxxxx>:
> On Fri, Oct 14, 2016 at 09:48:53AM +0800, Wanpeng Li wrote:
>> ===============================
>> [ INFO: suspicious RCU usage. ]
>> 4.8.0+ #24 Not tainted
>> -------------------------------
>> ./arch/x86/include/asm/msr-trace.h:47 suspicious rcu_dereference_check() usage!
>>
>> other info that might help us debug this:
>>
>>
>> RCU used illegally from idle CPU!
>> rcu_scheduler_active = 1, debug_locks = 0
>> RCU used illegally from extended quiescent state!
>> no locks held by swapper/1/0.
>>
>> stack backtrace:
>> CPU: 1 PID: 0 Comm: swapper/1 Not tainted 4.8.0+ #24
>> Hardware name: Dell Inc. OptiPlex 7020/0F5C5X, BIOS A03 01/08/2015
>> ffff90285de03f58 ffffffff9d44a0c9 ffff90285ca5d100 0000000000000001
>> ffff90285de03f88 ffffffff9d0ebd67 ffff902845165410 000000000000080b
>> 0000000000000000 0000000000000000 ffff90285de03fb8 ffffffff9d492b95
>> Call Trace:
>> <IRQ> [<ffffffff9d44a0c9>] dump_stack+0x99/0xd0
>> [<ffffffff9d0ebd67>] lockdep_rcu_suspicious+0xe7/0x120
>> [<ffffffff9d492b95>] do_trace_write_msr+0x135/0x140
>> [<ffffffff9d06f860>] native_write_msr+0x20/0x30
>> [<ffffffff9d065fad>] native_apic_msr_eoi_write+0x1d/0x30
>> [<ffffffff9d05bd1d>] smp_reschedule_interrupt+0x1d/0x30
>> [<ffffffff9d8daec6>] reschedule_interrupt+0x96/0xa0
>> <EOI> [<ffffffff9d732634>] ? cpuidle_enter_state+0xe4/0x360
>> [<ffffffff9d73261f>] ? cpuidle_enter_state+0xcf/0x360
>> [<ffffffff9d7328e7>] cpuidle_enter+0x17/0x20
>> [<ffffffff9d0e1a73>] call_cpuidle+0x23/0x50
>> [<ffffffff9d0e1d0c>] cpu_startup_entry+0x15c/0x280
>> [<ffffffff9d05ce64>] start_secondary+0x154/0x180
>>
>> Reschedule interrupt may be called in cpu idle state. This causes lockdep
>> check warning above.
>>
>> Add irq_enter/exit() in smp_reschedule_interrupt(), irq_enter() tells the RCU
>> subsystems to end the extended quiescent state, so the following trace call in
>> ack_APIC_irq() works correctly.
>>
>> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
>> Cc: Mike Galbraith <efault@xxxxxx>
>> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
>> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
>> Signed-off-by: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>> ---
>> arch/x86/kernel/smp.c | 2 ++
>> 1 file changed, 2 insertions(+)
>>
>> diff --git a/arch/x86/kernel/smp.c b/arch/x86/kernel/smp.c
>> index 658777c..ac2ee87 100644
>> --- a/arch/x86/kernel/smp.c
>> +++ b/arch/x86/kernel/smp.c
>> @@ -259,8 +259,10 @@ static inline void __smp_reschedule_interrupt(void)
>>
>> __visible void smp_reschedule_interrupt(struct pt_regs *regs)
>> {
>> + irq_enter();
>> ack_APIC_irq();
>> __smp_reschedule_interrupt();
>> + irq_exit();
>
> Urgh, I really hate this...
>
> So now we're making a very frequent interrupt slower because of debug
> code :/

Do you have a better idea? :)

Regards,
Wanpeng Li