On 05/02/2018 09:11, Chris Wilson wrote:
The pmu_event_read() callchain is in an atomic context which then
Call-chain.
forbids us sleeping, such as required to wake the device up for rpm.
In this case, we can only read our registers iff the device is already
s/iff/if/
awake. In the case of rc6 this means that we cannot report the latest
values when the whole system is asleep.
<3>[ 101.309358] BUG: sleeping function called from invalid context at
drivers/base/power/runtime.c:1034
<3>[ 101.309502] in_atomic(): 1, irqs_disabled(): 1, pid: 1513, name: perf_pmu
<4>[ 101.309558] 3 locks held by perf_pmu/1513:
<4>[ 101.309563] #0: (&cpuctx_mutex){+.+.}, at: [<00000000c3ec2708>]
perf_event_ctx_lock_nested+0xbc/0x1d0
<4>[ 101.309598] #1: (&event->child_mutex){+.+.}, at: [<000000004cd02bc9>]
__perf_event_read_value+0x32/0xf0
<4>[ 101.309624] #2: (&cpuctx_lock){....}, at: [<000000002343d2dc>]
__perf_event_read+0x64/0x2e0
<4>[ 101.309653] irq event stamp: 25750
<4>[ 101.309662] hardirqs last enabled at (25749): [<00000000c4700020>]
entry_SYSCALL_64_fastpath+0x5/0x8f
<4>[ 101.309670] hardirqs last disabled at (25750): [<00000000e34beea3>]
generic_exec_single+0x90/0xe0
<4>[ 101.309678] softirqs last enabled at (25466): [<00000000bebe2149>]
__do_softirq+0x3a1/0x4aa
<4>[ 101.309687] softirqs last disabled at (25443): [<00000000062fd6b7>]
irq_exit+0xa4/0xb0
<3>[ 101.309692] Preemption disabled at:
<4>[ 101.309699] [<000000005125cef8>] perf_event_read+0x98/0x2e0
<4>[ 101.309741] CPU: 0 PID: 1513 Comm: perf_pmu Tainted: G U
4.15.0-CI-CI_DRM_3717+ #1
<4>[ 101.309747] Hardware name: /NUC6CAYB, BIOS
AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017
<4>[ 101.309752] Call Trace:
<4>[ 101.309764] dump_stack+0x5f/0x86
<4>[ 101.309774] ___might_sleep+0x1d9/0x240
<4>[ 101.309788] __pm_runtime_resume+0x75/0x80
<4>[ 101.309870] intel_runtime_pm_get+0x1c/0xa0 [i915]
<4>[ 101.309938] __i915_pmu_event_read+0x54/0x1c0 [i915]
<4>[ 101.310008] i915_pmu_event_read+0x1e/0x40 [i915]
<4>[ 101.310019] __perf_event_read+0x2b8/0x2e0
<4>[ 101.310031] ? __perf_event_output_stop+0xa0/0xa0
<4>[ 101.310040] generic_exec_single+0x96/0xe0
<4>[ 101.310050] ? __perf_event_output_stop+0xa0/0xa0
<4>[ 101.310058] smp_call_function_single+0x88/0x120
<4>[ 101.310076] ? perf_event_read+0xed/0x2e0
<4>[ 101.310083] ? preempt_count_add+0x44/0x90
<4>[ 101.310090] perf_event_read+0xed/0x2e0
<4>[ 101.310106] __perf_event_read_value+0x3c/0xf0
<4>[ 101.310119] perf_read+0x15e/0x290
<4>[ 101.310130] ? lock_acquire+0xaf/0x200
<4>[ 101.310143] ? free_object+0x6e/0x90
<4>[ 101.310156] __vfs_read+0x1e/0x120
<4>[ 101.310166] ? _raw_spin_unlock_irqrestore+0x39/0x60
<4>[ 101.310174] ? free_object+0x6e/0x90
<4>[ 101.310184] ? entry_SYSCALL_64_fastpath+0x5/0x8f
<4>[ 101.310195] vfs_read+0x9e/0x150
<4>[ 101.310206] SyS_read+0x40/0xa0
<4>[ 101.310219] entry_SYSCALL_64_fastpath+0x22/0x8f
<4>[ 101.310226] RIP: 0033:0x7f361c4276d0
<4>[ 101.310232] RSP: 002b:00007ffcce301ee8 EFLAGS: 00000246
<3>[ 101.311022] BUG: scheduling while atomic: perf_pmu/1513/0x00000004
<4>[ 101.311082] 3 locks held by perf_pmu/1513:
<4>[ 101.311088] #0: (&cpuctx_mutex){+.+.}, at: [<00000000c3ec2708>]
perf_event_ctx_lock_nested+0xbc/0x1d0
<4>[ 101.311124] #1: (&event->child_mutex){+.+.}, at: [<000000004cd02bc9>]
__perf_event_read_value+0x32/0xf0
<4>[ 101.311152] #2: (&cpuctx_lock){+.+.}, at: [<000000002343d2dc>]
__perf_event_read+0x64/0x2e0
<4>[ 101.311180] Modules linked in: snd_hda_codec_hdmi snd_hda_codec_realtek
snd_hda_codec_generic x86_pkg_temp_thermal coretemp crct10dif_pclmul crc32_pclmul
ghash_clmulni_intel i915 lpc_ich r8169 mii snd_hda_intel snd_hda_codec snd_hwdep
snd_hda_core snd_pcm pinctrl_broxton pinctrl_intel mei_me prime_numbers mei
<3>[ 101.311314] Preemption disabled at:
<4>[ 101.311321] [<000000005125cef8>] perf_event_read+0x98/0x2e0
<4>[ 101.311362] CPU: 0 PID: 1513 Comm: perf_pmu Tainted: G U W
4.15.0-CI-CI_DRM_3717+ #1
<4>[ 101.311368] Hardware name: /NUC6CAYB, BIOS
AYAPLCEL.86A.0040.2017.0619.1722 06/19/2017
<4>[ 101.311374] Call Trace:
<4>[ 101.311385] dump_stack+0x5f/0x86
<4>[ 101.311394] ? perf_event_read+0x98/0x2e0
<4>[ 101.311402] __schedule_bug+0x7a/0xd0
<4>[ 101.311413] __schedule+0x84c/0xaf0
<4>[ 101.311424] ? mark_held_locks+0x64/0x90
<4>[ 101.311432] ? _raw_spin_unlock_irqrestore+0x4c/0x60
<4>[ 101.311446] schedule+0x37/0x90
<4>[ 101.311456] schedule_timeout+0x1da/0x520
<4>[ 101.311471] ? collect_expired_timers+0xa0/0xa0
<4>[ 101.311485] ? msleep+0x1d/0x40
<4>[ 101.311497] ? pci_restore_standard_config+0x40/0x40
<4>[ 101.311504] msleep+0x34/0x40
<4>[ 101.311514] pci_raw_set_power_state+0x18f/0x220
<4>[ 101.311526] pci_set_power_state+0x94/0x110
<4>[ 101.311537] pci_restore_standard_config+0x22/0x40
<4>[ 101.311545] pci_pm_runtime_resume+0x3d/0x90
<4>[ 101.311556] __rpm_callback+0xb1/0x1e0
<4>[ 101.311566] ? pci_restore_standard_config+0x40/0x40
<4>[ 101.311577] rpm_callback+0x1a/0x70
<4>[ 101.311586] ? pci_restore_standard_config+0x40/0x40
<4>[ 101.311594] rpm_resume+0x4d7/0x770
<4>[ 101.311615] __pm_runtime_resume+0x42/0x80
<4>[ 101.311700] intel_runtime_pm_get+0x1c/0xa0 [i915]
<4>[ 101.311769] __i915_pmu_event_read+0x54/0x1c0 [i915]
<4>[ 101.311840] i915_pmu_event_read+0x1e/0x40 [i915]
<4>[ 101.311852] __perf_event_read+0x2b8/0x2e0
<4>[ 101.311864] ? __perf_event_output_stop+0xa0/0xa0
<4>[ 101.311874] generic_exec_single+0x96/0xe0
<4>[ 101.311883] ? __perf_event_output_stop+0xa0/0xa0
<4>[ 101.311891] smp_call_function_single+0x88/0x120
<4>[ 101.311910] ? perf_event_read+0xed/0x2e0
<4>[ 101.311916] ? preempt_count_add+0x44/0x90
<4>[ 101.311924] perf_event_read+0xed/0x2e0
<4>[ 101.311940] __perf_event_read_value+0x3c/0xf0
<4>[ 101.311954] perf_read+0x15e/0x290
<4>[ 101.311963] ? lock_acquire+0xaf/0x200
<4>[ 101.311975] ? free_object+0x6e/0x90
<4>[ 101.311989] __vfs_read+0x1e/0x120
<4>[ 101.311999] ? _raw_spin_unlock_irqrestore+0x39/0x60
<4>[ 101.312007] ? free_object+0x6e/0x90
<4>[ 101.312018] ? entry_SYSCALL_64_fastpath+0x5/0x8f
<4>[ 101.312028] vfs_read+0x9e/0x150
<4>[ 101.312040] SyS_read+0x40/0xa0
<4>[ 101.312054] entry_SYSCALL_64_fastpath+0x22/0x8f
<4>[ 101.312060] RIP: 0033:0x7f361c4276d0
<4>[ 101.312066] RSP: 002b:00007ffcce301ee8 EFLAGS: 00000246
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104943
Signed-off-by: Chris Wilson <[email protected]>
Cc: Tvrtko Ursulin <[email protected]>
---
drivers/gpu/drm/i915/i915_pmu.c | 29 +++++++++++++++++------------
drivers/gpu/drm/i915/i915_pmu.h | 1 +
2 files changed, 18 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c
index ecb0198bfb7a..9c50b173e9a1 100644
--- a/drivers/gpu/drm/i915/i915_pmu.c
+++ b/drivers/gpu/drm/i915/i915_pmu.c
@@ -423,18 +423,23 @@ static u64 __i915_pmu_event_read(struct perf_event *event)
val = count_interrupts(i915);
break;
case I915_PMU_RC6_RESIDENCY:
- intel_runtime_pm_get(i915);
- val = intel_rc6_residency_ns(i915,
- IS_VALLEYVIEW(i915) ?
- VLV_GT_RENDER_RC6 :
- GEN6_GT_GFX_RC6);
- if (HAS_RC6p(i915))
- val += intel_rc6_residency_ns(i915,
- GEN6_GT_GFX_RC6p);
- if (HAS_RC6pp(i915))
- val += intel_rc6_residency_ns(i915,
-
GEN6_GT_GFX_RC6pp);
- intel_runtime_pm_put(i915);
+ if (intel_runtime_pm_get_if_in_use(i915)) {
+ val = intel_rc6_residency_ns(i915,
+ IS_VALLEYVIEW(i915) ?
+ VLV_GT_RENDER_RC6 :
+ GEN6_GT_GFX_RC6);
+ if (HAS_RC6p(i915))
+ val += intel_rc6_residency_ns(i915,
+ GEN6_GT_GFX_RC6p);
+ if (HAS_RC6pp(i915))
+ val += intel_rc6_residency_ns(i915,
+ GEN6_GT_GFX_RC6pp);
+ intel_runtime_pm_put(i915);
+
+ i915->pmu.sample[__I915_SAMPLE_RC6].cur = val;
+ } else {
+ val = i915->pmu.sample[__I915_SAMPLE_RC6].cur;
+ }
Downside is that idle system, when auto-suspended, will show 0% RC6. But
I have no alternative ideas. :(
break;
}
}
diff --git a/drivers/gpu/drm/i915/i915_pmu.h b/drivers/gpu/drm/i915/i915_pmu.h
index 5a2e013a56bb..249983ed3f08 100644
--- a/drivers/gpu/drm/i915/i915_pmu.h
+++ b/drivers/gpu/drm/i915/i915_pmu.h
@@ -27,6 +27,7 @@
enum {
__I915_SAMPLE_FREQ_ACT = 0,
__I915_SAMPLE_FREQ_REQ,
+ __I915_SAMPLE_RC6,
__I915_NUM_PMU_SAMPLERS
};
Reviewed-by: Tvrtko Ursulin <[email protected]>
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
