x86, reboot: Fix a warning message triggered by stop_other_cpus()
PATCH 1/1 x86, reboot: Fix a warning message triggered by stop_other_cpus()
Quoted 1 line
From
96214a16a0e0da523ae770c30a8641dc8bec89ceMon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Tue, 22 May 2012 09:50:40 +0800
Subject: [PATCH] x86, reboot: Fix a warning message triggered by
stop_other_cpus()
When rebooting our 24 CPU Westmere servers with 3.4-rc6, we can always see such warning msg
Restarting system.
machine restart
------------[ cut here ]------------ WARNING: at arch/x86/kernel/smp.c:125 native_smp_send_reschedule+0x74/0xa7() Hardware name: X8DTN Modules linked in: igb [last unloaded: scsi_wait_scan] Pid: 1, comm: systemd-shutdow Not tainted 3.4.0-rc6+ #22 Call Trace: <IRQ> [<ffffffff8102a41f>] warn_slowpath_common+0x7e/0x96 [<ffffffff8102a44c>] warn_slowpath_null+0x15/0x17 [<ffffffff81018cf7>] native_smp_send_reschedule+0x74/0xa7 [<ffffffff810561c1>] trigger_load_balance+0x279/0x2a6 [<ffffffff81050112>] scheduler_tick+0xe0/0xe9 [<ffffffff81036768>] update_process_times+0x60/0x70 [<ffffffff81062f2f>] tick_sched_timer+0x68/0x92 [<ffffffff81046e33>] __run_hrtimer+0xb3/0x13c [<ffffffff81062ec7>] ? tick_nohz_handler+0xd0/0xd0 [<ffffffff810474f2>] hrtimer_interrupt+0xdb/0x198 [<ffffffff81019a35>] smp_apic_timer_interrupt+0x81/0x94 [<ffffffff81655187>] apic_timer_interrupt+0x67/0x70 <EOI> [<ffffffff8101a3c4>] ? default_send_IPI_mask_allbutself_phys+0xb4/0xc4 [<ffffffff8101c680>] physflat_send_IPI_allbutself+0x12/0x14 [<ffffffff81018db4>] native_nmi_stop_other_cpus+0x8a/0xd6 [<ffffffff810188ba>] native_machine_shutdown+0x50/0x67 [<ffffffff81018926>] machine_shutdown+0xa/0xc [<ffffffff8101897e>] native_machine_restart+0x20/0x32 [<ffffffff810189b0>] machine_restart+0xa/0xc [<ffffffff8103b196>] kernel_restart+0x47/0x4c [<ffffffff8103b2e6>] sys_reboot+0x13e/0x17c [<ffffffff8164e436>] ? _raw_spin_unlock_bh+0x10/0x12 [<ffffffff810fcac9>] ? bdi_queue_work+0xcf/0xd8 [<ffffffff810fe82f>] ? __bdi_start_writeback+0xae/0xb7 [<ffffffff810e0d64>] ? iterate_supers+0xa3/0xb7 [<ffffffff816547a2>] system_call_fastpath+0x16/0x1b ---[ end trace 320af5cb1cb60c5b ]---
The root cause seems to be the default_send_IPI_mask_allbutself_phys() takes quiet some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, the time is long enough for a timer interrupt to happen, which will in turn trigger to kick load balance to a stopped CPU and cause this warning in native_smp_send_reschedule().
So disabling the local irq before stop_other_cpu() can fix this problem (tested 25 times reboot ok), and it is fine as there should be nobody caring the timer interrupt in such reboot stage.
The patch is against v3.4-rc6
--- arch/x86/kernel/reboot.c | 7 +++++-- 1 files changed, 5 insertions(+), 2 deletions(-)
arch/x86/kernel/reboot.c +5 -2
@@ -659,9 +659,12 @@ void native_machine_shutdown(void)/* Make certain I only run on the appropriate processor */ set_cpus_allowed_ptr (current,cpumask_of (reboot_cpu_id)); -/* O.K Now that I'm on the appropriate processor, -* stop all of the others. +/* +* O.K Now that I'm on the appropriate processor, stop all of the +* others. Also disable the local irq to not receive the per-cpu +* timer interrupt which may trigger scheduler's load balance. */ +local_irq_disable ();stop_other_cpus ();#endif
Signature · 1 line
1.7.5.1
Quoted 4 lines · Feng Tang
The root cause seems to be the
default_send_IPI_mask_allbutself_phys()takes quiet some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, the
I sent Ingo a patch (which he took in), that reverts this path to interrupts again (and NMIs as a fallback if it fails). The problem will still exist I assume though I don't know if it changes the timing characteristics.
Quoted 7 lines
time is long enough for a timer interrupt to happen, which will in turn trigger to kick load balance to a stopped CPU and cause this warning in
native_smp_send_reschedule().So disabling the local irq before
stop_other_cpu()can fix this problem (tested 25 times reboot ok), and it is fine as there should be nobody caring the timer interrupt in such reboot stage.
It seems to be fine. I just do not know enough about the load balancing the timer interrupt does to know if this is the right way to go or not.
So from the shutdown cpu perspective, you are just blocking the interrupt from occurring while we shutdown the other cpus, which prevents the load balancer from running and seeing cpus disappear beneath it. Sounds reasonable.
Though on the other hand the same interrupt on another cpu might spit out the same warning as cpus disappear without notification if it tries to reschedule. Probably rare and the window is probably small as the cpus would probably all 'stop' rather quickly.
I don't know. I guess it is ok.
Cheers,
Don
Hi Don,
Quoted 9 lines · On Tue, 22 May 2012 09:33:56 -0400 Don Zickus <[email protected]> wrote:
Quoted 4 lines · Feng Tang
The root cause seems to be the
default_send_IPI_mask_allbutself_phys()takes quiet some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, theI sent Ingo a patch (which he took in), that reverts this path to interrupts again (and NMIs as a fallback if it fails). The problem will still exist I assume though I don't know if it changes the timing characteristics.
Thanks for your review and letting me know the revert which I just saw in the latest Linus' tree. And I think this new mechanism will greatly reduce the possibility, but it still exist as the old nmi_stop_other_cpus's implementation still exist. So
Ingo,
Could you review this patch? thanks,
-Feng
Quoted 15 lines · Feng Tang
Hi Don,
Quoted 9 lines · On Tue, 22 May 2012 09:33:56 -0400 Don Zickus <[email protected]> wrote:
Quoted 4 lines · Feng Tang
The root cause seems to be the
default_send_IPI_mask_allbutself_phys()takes quiet some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, theI sent Ingo a patch (which he took in), that reverts this path to interrupts again (and NMIs as a fallback if it fails). The problem will still exist I assume though I don't know if it changes the timing characteristics.
Quoted 8 lines
Thanks for your review and letting me know the revert which I just saw in the latest Linus' tree. And I think this new mechanism will greatly reduce the possibility, but it still exist as the old nmi_stop_other_cpus's implementation still exist. So
Ingo,
Could you review this patch? thanks,
It doesn't apply anymore, so please merge and add Don's Acked-by.
Thanks,
Ingo
Feng Tang
Acked-by Don ZickusQuoted 24 lines · On Wed, 30 May 2012 10:51:10 +0200 Ingo Molnar <[email protected]> wrote:
Quoted 21 lines · Feng Tang
Hi Don,
Quoted 10 lines · On Tue, 22 May 2012 09:33:56 -0400 Don Zickus <[email protected]> wrote:
Quoted 4 lines · Feng Tang
The root cause seems to be the
default_send_IPI_mask_allbutself_phys()takes quiet some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, theI sent Ingo a patch (which he took in), that reverts this path to interrupts again (and NMIs as a fallback if it fails). The problem will still exist I assume though I don't know if it changes the timing characteristics.
Thanks for your review and letting me know the revert which I just saw in the latest Linus' tree. And I think this new mechanism will greatly reduce the possibility, but it still exist as the old nmi_stop_other_cpus's implementation still exist. So
Quoted 6 lines
Quoted 3 lines
Ingo,
Could you review this patch? thanks,
It doesn't apply anymore, so please merge and add Don's Acked-by.
Hi Ingo,
Following is the updated patch, which could be applied to Linus's tree and tip tree's master branch.
Thanks,
Feng
-------------------------
Quoted 1 line
From
9c4fc1b95da9cf35e1b1e8a71a67fc594adfdd78Mon Sep 17 00:00:00 2001
From: Feng Tang <[email protected]>
Date: Wed, 30 May 2012 21:01:49 +0800
Subject: [PATCH] x86, reboot: Fix a warning message triggered by
stop_other_cpus()
When rebooting our 24 CPU Westmere servers with 3.4-rc6, we can always see such warning msg:
Restarting system.
machine restart
------------[ cut here ]------------ WARNING: at arch/x86/kernel/smp.c:125 native_smp_send_reschedule+0x74/0xa7() Hardware name: X8DTN Modules linked in: igb [last unloaded: scsi_wait_scan] Pid: 1, comm: systemd-shutdow Not tainted 3.4.0-rc6+ #22 Call Trace: <IRQ> [<ffffffff8102a41f>] warn_slowpath_common+0x7e/0x96 [<ffffffff8102a44c>] warn_slowpath_null+0x15/0x17 [<ffffffff81018cf7>] native_smp_send_reschedule+0x74/0xa7 [<ffffffff810561c1>] trigger_load_balance+0x279/0x2a6 [<ffffffff81050112>] scheduler_tick+0xe0/0xe9 [<ffffffff81036768>] update_process_times+0x60/0x70 [<ffffffff81062f2f>] tick_sched_timer+0x68/0x92 [<ffffffff81046e33>] __run_hrtimer+0xb3/0x13c [<ffffffff81062ec7>] ? tick_nohz_handler+0xd0/0xd0 [<ffffffff810474f2>] hrtimer_interrupt+0xdb/0x198 [<ffffffff81019a35>] smp_apic_timer_interrupt+0x81/0x94 [<ffffffff81655187>] apic_timer_interrupt+0x67/0x70 <EOI> [<ffffffff8101a3c4>] ? default_send_IPI_mask_allbutself_phys+0xb4/0xc4 [<ffffffff8101c680>] physflat_send_IPI_allbutself+0x12/0x14 [<ffffffff81018db4>] native_nmi_stop_other_cpus+0x8a/0xd6 [<ffffffff810188ba>] native_machine_shutdown+0x50/0x67 [<ffffffff81018926>] machine_shutdown+0xa/0xc [<ffffffff8101897e>] native_machine_restart+0x20/0x32 [<ffffffff810189b0>] machine_restart+0xa/0xc [<ffffffff8103b196>] kernel_restart+0x47/0x4c [<ffffffff8103b2e6>] sys_reboot+0x13e/0x17c [<ffffffff8164e436>] ? _raw_spin_unlock_bh+0x10/0x12 [<ffffffff810fcac9>] ? bdi_queue_work+0xcf/0xd8 [<ffffffff810fe82f>] ? __bdi_start_writeback+0xae/0xb7 [<ffffffff810e0d64>] ? iterate_supers+0xa3/0xb7 [<ffffffff816547a2>] system_call_fastpath+0x16/0x1b ---[ end trace 320af5cb1cb60c5b ]---
The root cause seems to be the default_send_IPI_mask_allbutself_phys() takes quiet some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, the time is long enough for a timer interrupt to happen, which will in turn trigger to kick load balance to a stopped CPU and cause this warning in native_smp_send_reschedule().
So disabling the local irq before stop_other_cpu() can fix this problem (tested 25 times reboot ok), and it is fine as there should be nobody caring the timer interrupt in such reboot stage.
The latest 3.4 kernel slightly change this behavior by sending REBOOT_VECTOR first and only send NMI_VECTOR if the REBOOT_VCTOR fails, and this patch is still needed to prevent the problem.
--- arch/x86/kernel/reboot.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-)
arch/x86/kernel/reboot.c +4 -2
@@ -658,9 +658,11 @@ void native_machine_shutdown(void)set_cpus_allowed_ptr (current,cpumask_of (reboot_cpu_id));/* -* O.K Now that I'm on the appropriate processor, -* stop all of the others. +* O.K Now that I'm on the appropriate processor, stop all of the +* others. Also disable the local irq to not receive the per-cpu +* timer interrupt which may trigger scheduler's load balance. */ +local_irq_disable ();stop_other_cpus ();#endif
Signature · 1 line
1.7.9
[tip:x86/urgent] x86/reboot: Fix a warning message triggered by stop_other_cpus()
Commit-ID: 55c844a4dd16a4d1fdc0cf2a283ec631a02ec448 Gitweb: http://git.kernel.org/tip/55c844a4dd16a4d1fdc0cf2a283ec631a02ec448 Author: Feng Tang <[email protected]> AuthorDate: Wed, 30 May 2012 23:15:41 +0800 Committer: Ingo Molnar <[email protected]> CommitDate: Wed, 6 Jun 2012 12:03:23 +0200
x86/reboot: Fix a warning message triggered by stop_other_cpus()
When rebooting our 24 CPU Westmere servers with 3.4-rc6, we always see this warning msg:
Restarting system.
machine restart
------------[ cut here ]------------ WARNING: at arch/x86/kernel/smp.c:125 native_smp_send_reschedule+0x74/0xa7() Hardware name: X8DTN Modules linked in: igb [last unloaded: scsi_wait_scan] Pid: 1, comm: systemd-shutdow Not tainted 3.4.0-rc6+ #22 Call Trace: <IRQ> [<ffffffff8102a41f>] warn_slowpath_common+0x7e/0x96 [<ffffffff8102a44c>] warn_slowpath_null+0x15/0x17 [<ffffffff81018cf7>] native_smp_send_reschedule+0x74/0xa7 [<ffffffff810561c1>] trigger_load_balance+0x279/0x2a6 [<ffffffff81050112>] scheduler_tick+0xe0/0xe9 [<ffffffff81036768>] update_process_times+0x60/0x70 [<ffffffff81062f2f>] tick_sched_timer+0x68/0x92 [<ffffffff81046e33>] __run_hrtimer+0xb3/0x13c [<ffffffff81062ec7>] ? tick_nohz_handler+0xd0/0xd0 [<ffffffff810474f2>] hrtimer_interrupt+0xdb/0x198 [<ffffffff81019a35>] smp_apic_timer_interrupt+0x81/0x94 [<ffffffff81655187>] apic_timer_interrupt+0x67/0x70 <EOI> [<ffffffff8101a3c4>] ? default_send_IPI_mask_allbutself_phys+0xb4/0xc4 [<ffffffff8101c680>] physflat_send_IPI_allbutself+0x12/0x14 [<ffffffff81018db4>] native_nmi_stop_other_cpus+0x8a/0xd6 [<ffffffff810188ba>] native_machine_shutdown+0x50/0x67 [<ffffffff81018926>] machine_shutdown+0xa/0xc [<ffffffff8101897e>] native_machine_restart+0x20/0x32 [<ffffffff810189b0>] machine_restart+0xa/0xc [<ffffffff8103b196>] kernel_restart+0x47/0x4c [<ffffffff8103b2e6>] sys_reboot+0x13e/0x17c [<ffffffff8164e436>] ? _raw_spin_unlock_bh+0x10/0x12 [<ffffffff810fcac9>] ? bdi_queue_work+0xcf/0xd8 [<ffffffff810fe82f>] ? __bdi_start_writeback+0xae/0xb7 [<ffffffff810e0d64>] ? iterate_supers+0xa3/0xb7 [<ffffffff816547a2>] system_call_fastpath+0x16/0x1b ---[ end trace 320af5cb1cb60c5b ]---
The root cause seems to be the default_send_IPI_mask_allbutself_phys() takes quite some time (I measured it could be several ms) to complete sending NMIs to all the other 23 CPUs, and for HZ=250/1000 system, the time is long enough for a timer interrupt to happen, which will in turn trigger to kick load balance to a stopped CPU and cause this warning in native_smp_send_reschedule().
So disabling the local irq before stop_other_cpu() can fix this problem (tested 25 times reboot ok), and it is fine as there should be nobody caring the timer interrupt in such reboot stage.
The latest 3.4 kernel slightly changes this behavior by sending REBOOT_VECTOR first and only send NMI_VECTOR if the REBOOT_VCTOR fails, and this patch is still needed to prevent the problem.
--- arch/x86/kernel/reboot.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-)
arch/x86/kernel/reboot.c +4 -2
@@ -639,9 +639,11 @@ void native_machine_shutdown(void)set_cpus_allowed_ptr (current,cpumask_of (reboot_cpu_id));/* -* O.K Now that I'm on the appropriate processor, -* stop all of the others. +* O.K Now that I'm on the appropriate processor, stop all of the +* others. Also disable the local irq to not receive the per-cpu +* timer interrupt which may trigger scheduler's load balance. */ +local_irq_disable ();stop_other_cpus ();#endif