Re: [lkp] [sched/fair] 41e0d37f7a: divide error: 0000 [#1] SMP
From: Rafael J. Wysocki
Date: Tue May 03 2016 - 11:07:49 EST
On Tuesday, May 03, 2016 03:53:12 PM Rafael J. Wysocki wrote:
> On Tuesday, May 03, 2016 03:22:24 PM Rafael J. Wysocki wrote:
> > On Tue, May 3, 2016 at 2:58 PM, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> > > On Tue, May 3, 2016 at 2:54 PM, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> > >> On Tue, May 3, 2016 at 2:15 PM, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> > >>> On Tue, May 3, 2016 at 10:32 AM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > >>>> On Tue, May 03, 2016 at 09:10:51AM +0800, kernel test robot wrote:
> > >>>>> FYI, we noticed the following commit:
> > >>>>>
> > >>>>> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched/core
> > >>>>> commit 41e0d37f7ac81297c07ba311e4ad39465b8c8295 ("sched/fair: Do not call cpufreq hook unless util changed")
> > >>>>
> > >>>>
> > >>>>> [ 14.860950] Freeing unused kernel memory: 260K (ffff88103edbf000 - ffff88103ee00000)
> > >>>>> [ 14.873013] systemd[1]: RTC configured in localtime, applying delta of 480 minutes to system time.
> > >>>>> [ 14.884474] random: systemd urandom read with 5 bits of entropy available
> > >>>>> [ 14.903975] divide error: 0000 [#1] SMP
> > >>>>> [ 14.908375] Modules linked in:
> > >>>>> [ 14.911793] CPU: 39 PID: 1 Comm: systemd Not tainted 4.6.0-rc4-00016-g41e0d37 #1
> > >>>>> [ 14.920051] Hardware name: Intel Corporation S2600WP/S2600WP, BIOS SE5C600.86B.02.02.0002.122320131210 12/23/2013
> > >>>>> [ 14.931509] task: ffff8810101d8000 ti: ffff88081ab20000 task.ti: ffff88081ab20000
> > >>>>> [ 14.939862] RIP: 0010:[<ffffffff8176ad32>] [<ffffffff8176ad32>] intel_pstate_get+0x32/0x40
> > >>>>> [ 14.949202] RSP: 0018:ffff88081ab23d70 EFLAGS: 00010006
> > >>>>> [ 14.955129] RAX: 0000000000000000 RBX: 0000000000000024 RCX: ffff8808091e0300
> > >>>>> [ 14.963094] RDX: 0000000000000000 RSI: 0000000000000100 RDI: 0000000000000024
> > >>>>> [ 14.971057] RBP: ffff88081ab23d88 R08: 0000000000001000 R09: 00000000096a1000
> > >>>>> [ 14.979022] R10: 0000000000ffff10 R11: 000000000000000f R12: 0000000000000202
> > >>>>> [ 14.986984] R13: ffff88101390a040 R14: ffff88100e48e180 R15: ffff88101390a040
> > >>>>> [ 14.994950] FS: 00007f66fe117880(0000) GS:ffff8810139c0000(0000) knlGS:0000000000000000
> > >>>>> [ 15.003982] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > >>>>> [ 15.010393] CR2: 000055f78b760098 CR3: 000000103d759000 CR4: 00000000001406e0
> > >>>>> [ 15.018359] Stack:
> > >>>>> [ 15.020602] ffffffff81764dad 0000000000000024 ffff88100e48e180 ffff88081ab23dc8
> > >>>>> [ 15.028899] ffffffff81040267 ffff88101390a0ac 0000000000000340 ffff88081ab23f20
> > >>>>> [ 15.037197] ffff88103cd7c400 ffff88100e48e180 ffff88101390a040 ffff88081ab23e30
> > >>>>> [ 15.045493] Call Trace:
> > >>>>> [ 15.048223] [<ffffffff81764dad>] ? cpufreq_quick_get+0x3d/0x90
> > >>>>> [ 15.054832] [<ffffffff81040267>] show_cpuinfo+0x3c7/0x410
> > >>>>> [ 15.060956] [<ffffffff8121f5c4>] seq_read+0x2c4/0x3a0
> > >>>>> [ 15.066685] [<ffffffff81266ea8>] proc_reg_read+0x48/0x70
> > >>>>> [ 15.072713] [<ffffffff811f9d58>] __vfs_read+0x28/0xd0
> > >>>>> [ 15.078451] [<ffffffff813bab63>] ? security_file_permission+0xa3/0xc0
> > >>>>> [ 15.085737] [<ffffffff811faa97>] ? rw_verify_area+0x57/0xd0
> > >>>>> [ 15.092054] [<ffffffff811fab96>] vfs_read+0x86/0x130
> > >>>>> [ 15.097691] [<ffffffff811fbf96>] SyS_read+0x46/0xa0
> > >>>>> [ 15.103234] [<ffffffff818f71b2>] entry_SYSCALL_64_fastpath+0x1a/0xa4
> > >>>>> [ 15.110421] Code: 05 dc 1b c3 00 89 ff 55 48 89 e5 48 8b 0c f8 48 85 c9 74 1f 48 63 51 1c 48 63 41 20 5d 48 0f af c2 31 d2 48 0f af 81 88 00 00 00 <48> f7 b1 90 00 00 00 c3 31 c0 5d c3 66 90 0f 1f 44 00 00 8b 77
> > >>>>> [ 15.132161] RIP [<ffffffff8176ad32>] intel_pstate_get+0x32/0x40
> > >>>>> [ 15.138875] RSP <ffff88081ab23d70>
> > >>>>> [ 15.142770] ---[ end trace e5d5a8bedf5502e1 ]---
> > >>>>> [ 15.149323] Kernel panic - not syncing: Fatal exception
> > >>>>>
> > >>>>
> > >>>> That's intel_pstate.c:get_avg_frequency(), which assumes mperf != 0. It
> > >>>> being 0 seems to suggest intel_pstate_sample() hasn't been called yet or
> > >>>> so.
> > >>>
> > >>> Well, what's the tree based on?
> > >>>
> > >>> The mainline does this:
> > >>>
> > >>> bool sample_taken = intel_pstate_sample(cpu, time);
> > >>>
> > >>> if (sample_taken && !hwp_active)
> > >>> intel_pstate_adjust_busy_pstate(cpu);
> > >>>
> > >>> and (the mainline version of) intel_pstate_sample() returns false when
> > >>> it is called for the first time after setting the update_util hook.
> > >>
> > >> If that helps, I can expose my pm-cpufreq-fixes branch to pull from.
> > >> It contains all cpufreq material that went into the Linus' tree to
> > >> date and is based on 4.5-rc3.
> > >
> > > In fact, it is exposed already:
> > >
> > > git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git \
> > > pm-cpufreq-fixes
> > >
> > > and the top-most commit is 1becf03545a0859ceaaf9e8c2d9861882a71cb01
> > > (cpufreq: intel_pstate: Fix processing for turbo activation ratio).
> >
> > Ah, that will fail as well.
> >
> > The problem is that intel_pstate_get() can be called before we take
> > the first sample.
> >
> > I need to think about how to fix that.
>
> Maybe something like the below (untested, but builds).
>
> It will make intel_pstate_get() return 0 until avg_frequency gets populated
> which is actually OK.
The previous one would break the HWP case, so below is a new one (still
untested).
---
drivers/cpufreq/intel_pstate.c | 19 +++++++++----------
1 file changed, 9 insertions(+), 10 deletions(-)
Index: linux-pm/drivers/cpufreq/intel_pstate.c
===================================================================
--- linux-pm.orig/drivers/cpufreq/intel_pstate.c
+++ linux-pm/drivers/cpufreq/intel_pstate.c
@@ -114,6 +114,7 @@ struct cpudata {
u64 prev_mperf;
u64 prev_tsc;
u64 prev_cummulative_iowait;
+ int avg_frequency;
struct sample sample;
};
@@ -1044,7 +1045,7 @@ static inline void intel_pstate_adjust_b
sample->mperf,
sample->aperf,
sample->tsc,
- get_avg_frequency(cpu));
+ cpu->avg_frequency);
}
static void intel_pstate_update_util(struct update_util_data *data, u64 time,
@@ -1056,8 +1057,11 @@ static void intel_pstate_update_util(str
if ((s64)delta_ns >= pid_params.sample_rate_ns) {
bool sample_taken = intel_pstate_sample(cpu, time);
- if (sample_taken && !hwp_active)
- intel_pstate_adjust_busy_pstate(cpu);
+ if (sample_taken) {
+ cpu->avg_frequency = get_avg_frequency(cpu);
+ if (!hwp_active)
+ intel_pstate_adjust_busy_pstate(cpu);
+ }
}
}
@@ -1123,14 +1127,9 @@ static int intel_pstate_init_cpu(unsigne
static unsigned int intel_pstate_get(unsigned int cpu_num)
{
- struct sample *sample;
- struct cpudata *cpu;
+ struct cpudata *cpu = all_cpu_data[cpu_num];
- cpu = all_cpu_data[cpu_num];
- if (!cpu)
- return 0;
- sample = &cpu->sample;
- return get_avg_frequency(cpu);
+ return cpu ? cpu->avg_frequency : 0;
}
static void intel_pstate_set_update_util_hook(unsigned int cpu_num)