| Message ID | 20240919081121.10784-2-ukleinek@debian.org (mailing list archive) |
|---|---|
| State | In Next |
| Delegated to: | Rafael Wysocki |
| Headers | show |
| Series | cpufreq: intel_pstate: Make hwp_notify_lock a raw spinlock | expand |
I apply this patch on Debian kernel 6.11-1~exp1 (2024-09-19) source code. uname -a Linux atzlinux-ce 6.11-rt-amd64 #1 SMP PREEMPT_RT atzlinux 6.11-1~exp1 (2024-09-19) x86_64 GNU/Linux I get the following errors in dmesg: [ 7.273690] ucsi_acpi USBC000:00: possible UCSI driver bug 2 [ 7.451822] ucsi_acpi USBC000:00: error -EINVAL: PPM init failed [ 7.537749] ------------[ cut here ]------------ [ 7.537753] WARNING: CPU: 12 PID: 96 at kernel/workqueue.c:2259 delayed_work_timer_fn+0x63/0x90 [ 7.537763] Modules linked in: rfcomm cmac algif_hash algif_skcipher af_alg snd_seq_dummy snd_hrtimer snd_seq snd_seq_device bnep btusb uvcvideo btrtl btintel videobuf2_vmalloc btbcm uvc btmtk videobuf2_memops videobuf2_v4l2 bluetooth videodev videobuf2_common mc snd_soc_skl_hda_dsp snd_soc_hdac_hdmi snd_sof_probes snd_soc_intel_hda_dsp_common binfmt_misc nls_ascii nls_cp437 vfat fat snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic snd_hda_scodec_component snd_soc_dmic snd_sof_pci_intel_tgl snd_sof_pci_intel_cnl snd_sof_intel_hda_generic soundwire_intel soundwire_generic_allocation soundwire_cadence snd_sof_intel_hda_common snd_sof_intel_hda_mlink snd_sof_intel_hda snd_sof_pci snd_sof_xtensa_dsp snd_sof iwlmvm snd_sof_utils intel_uncore_frequency snd_soc_hdac_hda intel_uncore_frequency_common snd_soc_acpi_intel_match x86_pkg_temp_thermal intel_powerclamp snd_soc_acpi soundwire_bus coretemp kvm_intel snd_ctl_led mac80211 snd_soc_avs kvm snd_soc_hda_codec snd_hda_ext_core snd_soc_core libarc4 [ 7.537811] crct10dif_pclmul snd_compress snd_pcm_dmaengine ghash_clmulni_intel sha512_ssse3 snd_hda_intel sha256_ssse3 sha1_ssse3 iwlwifi snd_intel_dspcfg snd_intel_sdw_acpi snd_hda_codec aesni_intel gf128mul snd_hda_core crypto_simd cryptd rapl mei_hdcp mei_pxp mei_wdt snd_hwdep intel_rapl_msr processor_thermal_device_pci intel_cstate cfg80211 processor_thermal_device processor_thermal_wt_hint intel_uncore snd_pcm thinkpad_acpi think_lmi iTCO_wdt processor_thermal_rfim mei_me firmware_attributes_class intel_pmc_bxt processor_thermal_rapl snd_timer ucsi_acpi mei iTCO_vendor_support intel_rapl_common nvram typec_ucsi wmi_bmof watchdog platform_profile processor_thermal_wt_req pcspkr processor_thermal_power_floor snd igen6_edac processor_thermal_mbox typec roles soundcore rfkill ac int3403_thermal joydev int340x_thermal_zone intel_pmc_core int3400_thermal intel_vsec intel_hid pmt_telemetry acpi_thermal_rel sparse_keymap acpi_tad pmt_class acpi_pad evdev serio_raw dm_mod loop configfs efi_pstore efivarfs ip_tables [ 7.537861] x_tables autofs4 ext4 crc16 mbcache jbd2 crc32c_generic i915 drm_buddy i2c_algo_bit drm_display_helper cec hid_multitouch rc_core hid_generic xhci_pci xhci_hcd ttm i2c_hid_acpi i2c_hid usbcore drm_kms_helper hid intel_lpss_pci video nvme intel_lpss crc32_pclmul i2c_i801 thunderbolt psmouse drm nvme_core e1000e crc32c_intel idma64 usb_common i2c_smbus fan button battery wmi [ 7.537887] CPU: 12 UID: 0 PID: 96 Comm: ktimers/12 Not tainted 6.11-rt-amd64 #1 atzlinux 6.11-1~exp1 [ 7.537891] Hardware name: LENOVO 21HD0078CD/21HD0078CD, BIOS N3QET47W (1.47 ) 07/18/2024 [ 7.537893] RIP: 0010:delayed_work_timer_fn+0x63/0x90 [ 7.537898] Code: 65 48 8b 3d 3f ad b4 77 65 8b 05 40 ad b4 77 8b 97 00 0d 00 00 25 00 00 ff 00 81 e2 00 01 00 00 09 d0 75 06 f6 47 2c 20 75 0b <0f> 0b 5b 5d 41 5c c3 cc cc cc cc e8 fd 76 00 00 48 85 c0 74 eb 48 [ 7.537900] RSP: 0018:ffffbb900046bc98 EFLAGS: 00010006 [ 7.537903] RAX: 0000000000000100 RBX: ffff9a6182053000 RCX: ffff9a64cf7a35c0 [ 7.537904] RDX: 0000000000000100 RSI: ffffffff884eab80 RDI: ffff9a6181130000 [ 7.537906] RBP: 0000000000002000 R08: ffff9a64cf7a3610 R09: ffff9a64cf7a3658 [ 7.537907] R10: 0000000000000002 R11: 0000000000000001 R12: ffff9a6183bfa900 [ 7.537908] R13: dead000000000122 R14: ffffffff884eab80 R15: ffff9a6183bfa920 [ 7.537910] FS: 0000000000000000(0000) GS:ffff9a64cf600000(0000) knlGS:0000000000000000 [ 7.537912] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 7.537913] CR2: 0000561c94ed4000 CR3: 000000041d62c000 CR4: 0000000000f50ef0 [ 7.537915] PKRU: 55555554 [ 7.537916] Call Trace: [ 7.537918] <TASK> [ 7.537919] ? delayed_work_timer_fn+0x63/0x90 [ 7.537923] ? __warn.cold+0x93/0xed [ 7.537926] ? delayed_work_timer_fn+0x63/0x90 [ 7.537933] ? report_bug+0xff/0x140 [ 7.537937] ? handle_bug+0x3c/0x80 [ 7.537942] ? exc_invalid_op+0x17/0x70 [ 7.537945] ? asm_exc_invalid_op+0x1a/0x20 [ 7.537947] ? __pfx_delayed_work_timer_fn+0x10/0x10 [ 7.537951] ? __pfx_delayed_work_timer_fn+0x10/0x10 [ 7.537955] ? delayed_work_timer_fn+0x63/0x90 [ 7.537958] ? __pfx_delayed_work_timer_fn+0x10/0x10 [ 7.537961] call_timer_fn+0x27/0x120 [ 7.537965] __run_timers+0x18d/0x300 [ 7.537969] timer_expire_remote+0x51/0x80 [ 7.537973] tmigr_handle_remote+0x3cc/0x470 [ 7.537979] handle_softirqs.isra.0+0xb5/0x230 [ 7.537982] ? __pfx_smpboot_thread_fn+0x10/0x10 [ 7.537986] run_timersd+0x3e/0x80 [ 7.537988] smpboot_thread_fn+0xda/0x1d0 [ 7.537992] kthread+0xcf/0x100 [ 7.537995] ? __pfx_kthread+0x10/0x10 [ 7.537997] ret_from_fork+0x31/0x50 [ 7.538000] ? __pfx_kthread+0x10/0x10 [ 7.538001] ret_from_fork_asm+0x1a/0x30 [ 7.538006] </TASK> [ 7.538007] ---[ end trace 0000000000000000 ]--- The whole dmesg is: https://linux-hardware.org/?probe=28057f9ed2&log=dmesg My notebook hardware info is: https://linux-hardware.org/?probe=28057f9ed2 If need do more test, please tell me. Thanks! 在 2024/9/19 16:11, Uwe Kleine-König 写道: > notify_hwp_interrupt() is called via sysvec_thermal() -> > smp_thermal_vector() -> intel_thermal_interrupt() in hard irq context. > For this reason it must not use a simple spin_lock that sleeps with > PREEMPT_RT enabled. So convert it to a raw spinlock. > > Reported-by: xiao sheng wen <atzlinux@sina.com> > Link: https://bugs.debian.org/1076483 > Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> > --- > Hello, > > this is deep in x86 land where I have *very* little experience and > knowledge. Is sysvec_thermal() (defined in arch/x86/kernel/irq.c) really > running in hard irq context? The stacktrace in the Debian bug report > suggests that. > > So please double check my reasoning before applying this patch. > > Thanks > Uwe > > drivers/cpufreq/intel_pstate.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) > > diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c > index aaea9a39eced..b0018f371ea3 100644 > --- a/drivers/cpufreq/intel_pstate.c > +++ b/drivers/cpufreq/intel_pstate.c > @@ -1845,7 +1845,7 @@ static void intel_pstate_notify_work(struct work_struct *work) > wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_STATUS, 0); > } > > -static DEFINE_SPINLOCK(hwp_notify_lock); > +static DEFINE_RAW_SPINLOCK(hwp_notify_lock); > static cpumask_t hwp_intr_enable_mask; > > #define HWP_GUARANTEED_PERF_CHANGE_STATUS BIT(0) > @@ -1868,7 +1868,7 @@ void notify_hwp_interrupt(void) > if (!(value & status_mask)) > return; > > - spin_lock_irqsave(&hwp_notify_lock, flags); > + raw_spin_lock_irqsave(&hwp_notify_lock, flags); > > if (!cpumask_test_cpu(this_cpu, &hwp_intr_enable_mask)) > goto ack_intr; > @@ -1876,13 +1876,13 @@ void notify_hwp_interrupt(void) > schedule_delayed_work(&all_cpu_data[this_cpu]->hwp_notify_work, > msecs_to_jiffies(10)); > > - spin_unlock_irqrestore(&hwp_notify_lock, flags); > + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); > > return; > > ack_intr: > wrmsrl_safe(MSR_HWP_STATUS, 0); > - spin_unlock_irqrestore(&hwp_notify_lock, flags); > + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); > } > > static void intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) > @@ -1895,9 +1895,9 @@ static void intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) > /* wrmsrl_on_cpu has to be outside spinlock as this can result in IPC */ > wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_INTERRUPT, 0x00); > > - spin_lock_irq(&hwp_notify_lock); > + raw_spin_lock_irq(&hwp_notify_lock); > cancel_work = cpumask_test_and_clear_cpu(cpudata->cpu, &hwp_intr_enable_mask); > - spin_unlock_irq(&hwp_notify_lock); > + raw_spin_unlock_irq(&hwp_notify_lock); > > if (cancel_work) > cancel_delayed_work_sync(&cpudata->hwp_notify_work); > @@ -1912,10 +1912,10 @@ static void intel_pstate_enable_hwp_interrupt(struct cpudata *cpudata) > if (boot_cpu_has(X86_FEATURE_HWP_NOTIFY)) { > u64 interrupt_mask = HWP_GUARANTEED_PERF_CHANGE_REQ; > > - spin_lock_irq(&hwp_notify_lock); > + raw_spin_lock_irq(&hwp_notify_lock); > INIT_DELAYED_WORK(&cpudata->hwp_notify_work, intel_pstate_notify_work); > cpumask_set_cpu(cpudata->cpu, &hwp_intr_enable_mask); > - spin_unlock_irq(&hwp_notify_lock); > + raw_spin_unlock_irq(&hwp_notify_lock); > > if (cpu_feature_enabled(X86_FEATURE_HWP_HIGHEST_PERF_CHANGE)) > interrupt_mask |= HWP_HIGHEST_PERF_CHANGE_REQ; > > base-commit: 3621a2c9142bd490af0666c0c02d52d60ce0d2a5
Hello, On 9/23/24 03:53, xiao sheng wen(肖盛文) wrote: > I apply this patch on Debian kernel 6.11-1~exp1 (2024-09-19) source code. > > uname -a > Linux atzlinux-ce 6.11-rt-amd64 #1 SMP PREEMPT_RT atzlinux 6.11-1~exp1 > (2024-09-19) x86_64 GNU/Linux > > I get the following errors in dmesg: > > [ 7.273690] ucsi_acpi USBC000:00: possible UCSI driver bug 2 > [ 7.451822] ucsi_acpi USBC000:00: error -EINVAL: PPM init failed > [ 7.537749] ------------[ cut here ]------------ > [ 7.537753] WARNING: CPU: 12 PID: 96 at kernel/workqueue.c:2259 > delayed_work_timer_fn+0x63/0x90 > [ 7.537763] Modules linked in: rfcomm cmac algif_hash algif_skcipher > af_alg snd_seq_dummy snd_hrtimer snd_seq snd_seq_device bnep btusb > uvcvideo btrtl btintel videobuf2_vmalloc btbcm uvc btmtk > videobuf2_memops videobuf2_v4l2 bluetooth videodev videobuf2_common mc > snd_soc_skl_hda_dsp snd_soc_hdac_hdmi snd_sof_probes > snd_soc_intel_hda_dsp_common binfmt_misc nls_ascii nls_cp437 vfat fat > snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic > snd_hda_scodec_component snd_soc_dmic snd_sof_pci_intel_tgl > snd_sof_pci_intel_cnl snd_sof_intel_hda_generic soundwire_intel > soundwire_generic_allocation soundwire_cadence snd_sof_intel_hda_common > snd_sof_intel_hda_mlink snd_sof_intel_hda snd_sof_pci snd_sof_xtensa_dsp > snd_sof iwlmvm snd_sof_utils intel_uncore_frequency snd_soc_hdac_hda > intel_uncore_frequency_common snd_soc_acpi_intel_match > x86_pkg_temp_thermal intel_powerclamp snd_soc_acpi soundwire_bus > coretemp kvm_intel snd_ctl_led mac80211 snd_soc_avs kvm > snd_soc_hda_codec snd_hda_ext_core snd_soc_core libarc4 > [ 7.537811] crct10dif_pclmul snd_compress snd_pcm_dmaengine > ghash_clmulni_intel sha512_ssse3 snd_hda_intel sha256_ssse3 sha1_ssse3 > iwlwifi snd_intel_dspcfg snd_intel_sdw_acpi snd_hda_codec aesni_intel > gf128mul snd_hda_core crypto_simd cryptd rapl mei_hdcp mei_pxp mei_wdt > snd_hwdep intel_rapl_msr processor_thermal_device_pci intel_cstate > cfg80211 processor_thermal_device processor_thermal_wt_hint intel_uncore > snd_pcm thinkpad_acpi think_lmi iTCO_wdt processor_thermal_rfim mei_me > firmware_attributes_class intel_pmc_bxt processor_thermal_rapl snd_timer > ucsi_acpi mei iTCO_vendor_support intel_rapl_common nvram typec_ucsi > wmi_bmof watchdog platform_profile processor_thermal_wt_req pcspkr > processor_thermal_power_floor snd igen6_edac processor_thermal_mbox > typec roles soundcore rfkill ac int3403_thermal joydev > int340x_thermal_zone intel_pmc_core int3400_thermal intel_vsec intel_hid > pmt_telemetry acpi_thermal_rel sparse_keymap acpi_tad pmt_class acpi_pad > evdev serio_raw dm_mod loop configfs efi_pstore efivarfs ip_tables > [ 7.537861] x_tables autofs4 ext4 crc16 mbcache jbd2 crc32c_generic > i915 drm_buddy i2c_algo_bit drm_display_helper cec hid_multitouch > rc_core hid_generic xhci_pci xhci_hcd ttm i2c_hid_acpi i2c_hid usbcore > drm_kms_helper hid intel_lpss_pci video nvme intel_lpss crc32_pclmul > i2c_i801 thunderbolt psmouse drm nvme_core e1000e crc32c_intel idma64 > usb_common i2c_smbus fan button battery wmi > [ 7.537887] CPU: 12 UID: 0 PID: 96 Comm: ktimers/12 Not tainted 6.11- > rt-amd64 #1 atzlinux 6.11-1~exp1 > [ 7.537891] Hardware name: LENOVO 21HD0078CD/21HD0078CD, BIOS > N3QET47W (1.47 ) 07/18/2024 > [ 7.537893] RIP: 0010:delayed_work_timer_fn+0x63/0x90 > [ 7.537898] Code: 65 48 8b 3d 3f ad b4 77 65 8b 05 40 ad b4 77 8b 97 > 00 0d 00 00 25 00 00 ff 00 81 e2 00 01 00 00 09 d0 75 06 f6 47 2c 20 75 > 0b <0f> 0b 5b 5d 41 5c c3 cc cc cc cc e8 fd 76 00 00 48 85 c0 74 eb 48 > [ 7.537900] RSP: 0018:ffffbb900046bc98 EFLAGS: 00010006 > [ 7.537903] RAX: 0000000000000100 RBX: ffff9a6182053000 RCX: > ffff9a64cf7a35c0 > [ 7.537904] RDX: 0000000000000100 RSI: ffffffff884eab80 RDI: > ffff9a6181130000 > [ 7.537906] RBP: 0000000000002000 R08: ffff9a64cf7a3610 R09: > ffff9a64cf7a3658 > [ 7.537907] R10: 0000000000000002 R11: 0000000000000001 R12: > ffff9a6183bfa900 > [ 7.537908] R13: dead000000000122 R14: ffffffff884eab80 R15: > ffff9a6183bfa920 > [ 7.537910] FS: 0000000000000000(0000) GS:ffff9a64cf600000(0000) > knlGS:0000000000000000 > [ 7.537912] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 7.537913] CR2: 0000561c94ed4000 CR3: 000000041d62c000 CR4: > 0000000000f50ef0 > [ 7.537915] PKRU: 55555554 > [ 7.537916] Call Trace: > [ 7.537918] <TASK> > [ 7.537919] ? delayed_work_timer_fn+0x63/0x90 > [ 7.537923] ? __warn.cold+0x93/0xed > [ 7.537926] ? delayed_work_timer_fn+0x63/0x90 > [ 7.537933] ? report_bug+0xff/0x140 > [ 7.537937] ? handle_bug+0x3c/0x80 > [ 7.537942] ? exc_invalid_op+0x17/0x70 > [ 7.537945] ? asm_exc_invalid_op+0x1a/0x20 > [ 7.537947] ? __pfx_delayed_work_timer_fn+0x10/0x10 > [ 7.537951] ? __pfx_delayed_work_timer_fn+0x10/0x10 > [ 7.537955] ? delayed_work_timer_fn+0x63/0x90 > [ 7.537958] ? __pfx_delayed_work_timer_fn+0x10/0x10 > [ 7.537961] call_timer_fn+0x27/0x120 > [ 7.537965] __run_timers+0x18d/0x300 > [ 7.537969] timer_expire_remote+0x51/0x80 > [ 7.537973] tmigr_handle_remote+0x3cc/0x470 > [ 7.537979] handle_softirqs.isra.0+0xb5/0x230 > [ 7.537982] ? __pfx_smpboot_thread_fn+0x10/0x10 > [ 7.537986] run_timersd+0x3e/0x80 > [ 7.537988] smpboot_thread_fn+0xda/0x1d0 > [ 7.537992] kthread+0xcf/0x100 > [ 7.537995] ? __pfx_kthread+0x10/0x10 > [ 7.537997] ret_from_fork+0x31/0x50 > [ 7.538000] ? __pfx_kthread+0x10/0x10 > [ 7.538001] ret_from_fork_asm+0x1a/0x30 > [ 7.538006] </TASK> > [ 7.538007] ---[ end trace 0000000000000000 ]--- > > The whole dmesg is: > https://linux-hardware.org/?probe=28057f9ed2&log=dmesg On a first glance this looks like an unrelated issue to me. Could you test with CONFIG_X86_INTEL_PSTATE=n? (If doesn't matter if you do this with my patch applied or not, without CONFIG_X86_INTEL_PSTATE the driver isn't build.) Best regards Uwe
Hello, 在 2024/9/26 15:48, Uwe Kleine-König 写道: > Hello, > > On 9/23/24 03:53, xiao sheng wen(肖盛文) wrote: >> I apply this patch on Debian kernel 6.11-1~exp1 (2024-09-19) source code. >> >> uname -a >> Linux atzlinux-ce 6.11-rt-amd64 #1 SMP PREEMPT_RT atzlinux 6.11-1~exp1 >> (2024-09-19) x86_64 GNU/Linux >> >> I get the following errors in dmesg: >> >> [ 7.273690] ucsi_acpi USBC000:00: possible UCSI driver bug 2 >> [ 7.451822] ucsi_acpi USBC000:00: error -EINVAL: PPM init failed >> [ 7.537749] ------------[ cut here ]------------ >> [ 7.537753] WARNING: CPU: 12 PID: 96 at kernel/workqueue.c:2259 >> delayed_work_timer_fn+0x63/0x90 >> [ 7.537763] Modules linked in: rfcomm cmac algif_hash >> algif_skcipher af_alg snd_seq_dummy snd_hrtimer snd_seq snd_seq_device >> bnep btusb uvcvideo btrtl btintel videobuf2_vmalloc btbcm uvc btmtk >> videobuf2_memops videobuf2_v4l2 bluetooth videodev videobuf2_common mc >> snd_soc_skl_hda_dsp snd_soc_hdac_hdmi snd_sof_probes >> snd_soc_intel_hda_dsp_common binfmt_misc nls_ascii nls_cp437 vfat fat >> snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic >> snd_hda_scodec_component snd_soc_dmic snd_sof_pci_intel_tgl >> snd_sof_pci_intel_cnl snd_sof_intel_hda_generic soundwire_intel >> soundwire_generic_allocation soundwire_cadence >> snd_sof_intel_hda_common snd_sof_intel_hda_mlink snd_sof_intel_hda >> snd_sof_pci snd_sof_xtensa_dsp snd_sof iwlmvm snd_sof_utils >> intel_uncore_frequency snd_soc_hdac_hda intel_uncore_frequency_common >> snd_soc_acpi_intel_match x86_pkg_temp_thermal intel_powerclamp >> snd_soc_acpi soundwire_bus coretemp kvm_intel snd_ctl_led mac80211 >> snd_soc_avs kvm snd_soc_hda_codec snd_hda_ext_core snd_soc_core libarc4 >> [ 7.537811] crct10dif_pclmul snd_compress snd_pcm_dmaengine >> ghash_clmulni_intel sha512_ssse3 snd_hda_intel sha256_ssse3 sha1_ssse3 >> iwlwifi snd_intel_dspcfg snd_intel_sdw_acpi snd_hda_codec aesni_intel >> gf128mul snd_hda_core crypto_simd cryptd rapl mei_hdcp mei_pxp mei_wdt >> snd_hwdep intel_rapl_msr processor_thermal_device_pci intel_cstate >> cfg80211 processor_thermal_device processor_thermal_wt_hint >> intel_uncore snd_pcm thinkpad_acpi think_lmi iTCO_wdt >> processor_thermal_rfim mei_me firmware_attributes_class intel_pmc_bxt >> processor_thermal_rapl snd_timer ucsi_acpi mei iTCO_vendor_support >> intel_rapl_common nvram typec_ucsi wmi_bmof watchdog platform_profile >> processor_thermal_wt_req pcspkr processor_thermal_power_floor snd >> igen6_edac processor_thermal_mbox typec roles soundcore rfkill ac >> int3403_thermal joydev int340x_thermal_zone intel_pmc_core >> int3400_thermal intel_vsec intel_hid pmt_telemetry acpi_thermal_rel >> sparse_keymap acpi_tad pmt_class acpi_pad evdev serio_raw dm_mod loop >> configfs efi_pstore efivarfs ip_tables >> [ 7.537861] x_tables autofs4 ext4 crc16 mbcache jbd2 >> crc32c_generic i915 drm_buddy i2c_algo_bit drm_display_helper cec >> hid_multitouch rc_core hid_generic xhci_pci xhci_hcd ttm i2c_hid_acpi >> i2c_hid usbcore drm_kms_helper hid intel_lpss_pci video nvme >> intel_lpss crc32_pclmul i2c_i801 thunderbolt psmouse drm nvme_core >> e1000e crc32c_intel idma64 usb_common i2c_smbus fan button battery wmi >> [ 7.537887] CPU: 12 UID: 0 PID: 96 Comm: ktimers/12 Not tainted >> 6.11- rt-amd64 #1 atzlinux 6.11-1~exp1 >> [ 7.537891] Hardware name: LENOVO 21HD0078CD/21HD0078CD, BIOS >> N3QET47W (1.47 ) 07/18/2024 >> [ 7.537893] RIP: 0010:delayed_work_timer_fn+0x63/0x90 >> [ 7.537898] Code: 65 48 8b 3d 3f ad b4 77 65 8b 05 40 ad b4 77 8b >> 97 00 0d 00 00 25 00 00 ff 00 81 e2 00 01 00 00 09 d0 75 06 f6 47 2c >> 20 75 0b <0f> 0b 5b 5d 41 5c c3 cc cc cc cc e8 fd 76 00 00 48 85 c0 74 >> eb 48 >> [ 7.537900] RSP: 0018:ffffbb900046bc98 EFLAGS: 00010006 >> [ 7.537903] RAX: 0000000000000100 RBX: ffff9a6182053000 RCX: >> ffff9a64cf7a35c0 >> [ 7.537904] RDX: 0000000000000100 RSI: ffffffff884eab80 RDI: >> ffff9a6181130000 >> [ 7.537906] RBP: 0000000000002000 R08: ffff9a64cf7a3610 R09: >> ffff9a64cf7a3658 >> [ 7.537907] R10: 0000000000000002 R11: 0000000000000001 R12: >> ffff9a6183bfa900 >> [ 7.537908] R13: dead000000000122 R14: ffffffff884eab80 R15: >> ffff9a6183bfa920 >> [ 7.537910] FS: 0000000000000000(0000) GS:ffff9a64cf600000(0000) >> knlGS:0000000000000000 >> [ 7.537912] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 >> [ 7.537913] CR2: 0000561c94ed4000 CR3: 000000041d62c000 CR4: >> 0000000000f50ef0 >> [ 7.537915] PKRU: 55555554 >> [ 7.537916] Call Trace: >> [ 7.537918] <TASK> >> [ 7.537919] ? delayed_work_timer_fn+0x63/0x90 >> [ 7.537923] ? __warn.cold+0x93/0xed >> [ 7.537926] ? delayed_work_timer_fn+0x63/0x90 >> [ 7.537933] ? report_bug+0xff/0x140 >> [ 7.537937] ? handle_bug+0x3c/0x80 >> [ 7.537942] ? exc_invalid_op+0x17/0x70 >> [ 7.537945] ? asm_exc_invalid_op+0x1a/0x20 >> [ 7.537947] ? __pfx_delayed_work_timer_fn+0x10/0x10 >> [ 7.537951] ? __pfx_delayed_work_timer_fn+0x10/0x10 >> [ 7.537955] ? delayed_work_timer_fn+0x63/0x90 >> [ 7.537958] ? __pfx_delayed_work_timer_fn+0x10/0x10 >> [ 7.537961] call_timer_fn+0x27/0x120 >> [ 7.537965] __run_timers+0x18d/0x300 >> [ 7.537969] timer_expire_remote+0x51/0x80 >> [ 7.537973] tmigr_handle_remote+0x3cc/0x470 >> [ 7.537979] handle_softirqs.isra.0+0xb5/0x230 >> [ 7.537982] ? __pfx_smpboot_thread_fn+0x10/0x10 >> [ 7.537986] run_timersd+0x3e/0x80 >> [ 7.537988] smpboot_thread_fn+0xda/0x1d0 >> [ 7.537992] kthread+0xcf/0x100 >> [ 7.537995] ? __pfx_kthread+0x10/0x10 >> [ 7.537997] ret_from_fork+0x31/0x50 >> [ 7.538000] ? __pfx_kthread+0x10/0x10 >> [ 7.538001] ret_from_fork_asm+0x1a/0x30 >> [ 7.538006] </TASK> >> [ 7.538007] ---[ end trace 0000000000000000 ]--- >> >> The whole dmesg is: >> https://linux-hardware.org/?probe=28057f9ed2&log=dmesg > > On a first glance this looks like an unrelated issue to me. Could you > test with CONFIG_X86_INTEL_PSTATE=n? (If doesn't matter if you do this > with my patch applied or not, without CONFIG_X86_INTEL_PSTATE the driver > isn't build.) > > Best regards > Uwe Yes, this WARNING is unrelated issue to your patch.It come from kernel [ktimers], so I don't need do test with CONFIG_X86_INTEL_PSTATE=n. I apply your patch on Debian kernel 6.10.11-1 (2024-09-22) source code. uname -a Linux atzlinux-ce 6.10.11-rt-amd64 #1 SMP PREEMPT_RT atzlinux 6.10.11-1 (2024-09-22) x86_64 GNU/Linux There is not any errors in dmesg now, work perfect! IMHO, this patch fix this bug in kernel 6.10.11-1. Thanks!
On Thu, 2024-09-19 at 10:11 +0200, Uwe Kleine-König wrote: > notify_hwp_interrupt() is called via sysvec_thermal() -> > smp_thermal_vector() -> intel_thermal_interrupt() in hard irq > context. > For this reason it must not use a simple spin_lock that sleeps with > PREEMPT_RT enabled. So convert it to a raw spinlock. > > Reported-by: xiao sheng wen <atzlinux@sina.com> > Link: https://bugs.debian.org/1076483 > Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> Missing Tested-by? Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> > --- > Hello, > > this is deep in x86 land where I have *very* little experience and > knowledge. Is sysvec_thermal() (defined in arch/x86/kernel/irq.c) > really > running in hard irq context? The stacktrace in the Debian bug report > suggests that. > > So please double check my reasoning before applying this patch. > This is thermal LVT and runs in hard IRQ context. > Thanks > Uwe > > drivers/cpufreq/intel_pstate.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) > > diff --git a/drivers/cpufreq/intel_pstate.c > b/drivers/cpufreq/intel_pstate.c > index aaea9a39eced..b0018f371ea3 100644 > --- a/drivers/cpufreq/intel_pstate.c > +++ b/drivers/cpufreq/intel_pstate.c > @@ -1845,7 +1845,7 @@ static void intel_pstate_notify_work(struct > work_struct *work) > wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_STATUS, 0); > } > > -static DEFINE_SPINLOCK(hwp_notify_lock); > +static DEFINE_RAW_SPINLOCK(hwp_notify_lock); > static cpumask_t hwp_intr_enable_mask; > > #define HWP_GUARANTEED_PERF_CHANGE_STATUS BIT(0) > @@ -1868,7 +1868,7 @@ void notify_hwp_interrupt(void) > if (!(value & status_mask)) > return; > > - spin_lock_irqsave(&hwp_notify_lock, flags); > + raw_spin_lock_irqsave(&hwp_notify_lock, flags); > > if (!cpumask_test_cpu(this_cpu, &hwp_intr_enable_mask)) > goto ack_intr; > @@ -1876,13 +1876,13 @@ void notify_hwp_interrupt(void) > schedule_delayed_work(&all_cpu_data[this_cpu]- > >hwp_notify_work, > msecs_to_jiffies(10)); > > - spin_unlock_irqrestore(&hwp_notify_lock, flags); > + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); > > return; > > ack_intr: > wrmsrl_safe(MSR_HWP_STATUS, 0); > - spin_unlock_irqrestore(&hwp_notify_lock, flags); > + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); > } > > static void intel_pstate_disable_hwp_interrupt(struct cpudata > *cpudata) > @@ -1895,9 +1895,9 @@ static void > intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) > /* wrmsrl_on_cpu has to be outside spinlock as this can > result in IPC */ > wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_INTERRUPT, 0x00); > > - spin_lock_irq(&hwp_notify_lock); > + raw_spin_lock_irq(&hwp_notify_lock); > cancel_work = cpumask_test_and_clear_cpu(cpudata->cpu, > &hwp_intr_enable_mask); > - spin_unlock_irq(&hwp_notify_lock); > + raw_spin_unlock_irq(&hwp_notify_lock); > > if (cancel_work) > cancel_delayed_work_sync(&cpudata->hwp_notify_work); > @@ -1912,10 +1912,10 @@ static void > intel_pstate_enable_hwp_interrupt(struct cpudata *cpudata) > if (boot_cpu_has(X86_FEATURE_HWP_NOTIFY)) { > u64 interrupt_mask = HWP_GUARANTEED_PERF_CHANGE_REQ; > > - spin_lock_irq(&hwp_notify_lock); > + raw_spin_lock_irq(&hwp_notify_lock); > INIT_DELAYED_WORK(&cpudata->hwp_notify_work, > intel_pstate_notify_work); > cpumask_set_cpu(cpudata->cpu, > &hwp_intr_enable_mask); > - spin_unlock_irq(&hwp_notify_lock); > + raw_spin_unlock_irq(&hwp_notify_lock); > > if > (cpu_feature_enabled(X86_FEATURE_HWP_HIGHEST_PERF_CHANGE)) > interrupt_mask |= > HWP_HIGHEST_PERF_CHANGE_REQ; > > base-commit: 3621a2c9142bd490af0666c0c02d52d60ce0d2a5
Hello, On Fri, Sep 27, 2024 at 03:47:08AM -0700, srinivas pandruvada wrote: > On Thu, 2024-09-19 at 10:11 +0200, Uwe Kleine-König wrote: > > notify_hwp_interrupt() is called via sysvec_thermal() -> > > smp_thermal_vector() -> intel_thermal_interrupt() in hard irq > > context. > > For this reason it must not use a simple spin_lock that sleeps with > > PREEMPT_RT enabled. So convert it to a raw spinlock. > > > > Reported-by: xiao sheng wen <atzlinux@sina.com> > > Link: https://bugs.debian.org/1076483 > > Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> > > Missing Tested-by? Neither I nor anybody else did test that patch before I sent it to the list (apart from a build test). I guess xiao sheng wen might have replied with a Tested-by tag. > Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Thanks Uwe
On 2024-09-19 10:11:21 [+0200], Uwe Kleine-König wrote: > notify_hwp_interrupt() is called via sysvec_thermal() -> > smp_thermal_vector() -> intel_thermal_interrupt() in hard irq context. > For this reason it must not use a simple spin_lock that sleeps with > PREEMPT_RT enabled. So convert it to a raw spinlock. > > Reported-by: xiao sheng wen <atzlinux@sina.com> > Link: https://bugs.debian.org/1076483 > Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> > --- > Hello, > > this is deep in x86 land where I have *very* little experience and > knowledge. Is sysvec_thermal() (defined in arch/x86/kernel/irq.c) really > running in hard irq context? The stacktrace in the Debian bug report > suggests that. > > So please double check my reasoning before applying this patch. Yes, these things are not threaded. Ideally you should have proper thermal setup so that you don't get any interrupts in regard to this and start throttling. Once that interrupt fires and starts throttling then your worst-case latency increases due to the interrupt and lower CPU frequency. Delaying the thermal interrupt to a thread might not be wise. Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> > Thanks > Uwe Sebastian
在 2024/9/28 19:51, Uwe Kleine-König 写道: > Hello, > > On Fri, Sep 27, 2024 at 03:47:08AM -0700, srinivas pandruvada wrote: >> On Thu, 2024-09-19 at 10:11 +0200, Uwe Kleine-König wrote: >>> notify_hwp_interrupt() is called via sysvec_thermal() -> >>> smp_thermal_vector() -> intel_thermal_interrupt() in hard irq >>> context. >>> For this reason it must not use a simple spin_lock that sleeps with >>> PREEMPT_RT enabled. So convert it to a raw spinlock. >>> >>> Reported-by: xiao sheng wen <atzlinux@sina.com> >>> Link: https://bugs.debian.org/1076483 >>> Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> >> >> Missing Tested-by? > > Neither I nor anybody else did test that patch before I sent it to the > list (apart from a build test). I guess xiao sheng wen might have > replied with a Tested-by tag. Yes, I had tested this patch on my notebook(base on Debian kernel 6.10.11-1 (2024-09-22) source code). Work perfect! This bug exist in 6.9.9-rt-amd64 #1 Debian 6.9.9-1: https://linux-hardware.org/?probe=d57f4d29a8&log=dmesg This bug fixed in PREEMPT_RT atzlinux 6.10.11-1 (2024-09-22) with this patch: https://linux-hardware.org/?probe=de5a44374a&log=dmesg Tested-by: xiao sheng wen <atzlinux@sina.com> > >> Acked-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> > > Thanks > Uwe Regards,
On Thu, Sep 19, 2024 at 10:12 AM Uwe Kleine-König <ukleinek@debian.org> wrote: > > notify_hwp_interrupt() is called via sysvec_thermal() -> > smp_thermal_vector() -> intel_thermal_interrupt() in hard irq context. > For this reason it must not use a simple spin_lock that sleeps with > PREEMPT_RT enabled. So convert it to a raw spinlock. > > Reported-by: xiao sheng wen <atzlinux@sina.com> > Link: https://bugs.debian.org/1076483 > Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> > --- > Hello, > > this is deep in x86 land where I have *very* little experience and > knowledge. Is sysvec_thermal() (defined in arch/x86/kernel/irq.c) really > running in hard irq context? The stacktrace in the Debian bug report > suggests that. > > So please double check my reasoning before applying this patch. > > Thanks > Uwe > > drivers/cpufreq/intel_pstate.c | 16 ++++++++-------- > 1 file changed, 8 insertions(+), 8 deletions(-) > > diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c > index aaea9a39eced..b0018f371ea3 100644 > --- a/drivers/cpufreq/intel_pstate.c > +++ b/drivers/cpufreq/intel_pstate.c > @@ -1845,7 +1845,7 @@ static void intel_pstate_notify_work(struct work_struct *work) > wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_STATUS, 0); > } > > -static DEFINE_SPINLOCK(hwp_notify_lock); > +static DEFINE_RAW_SPINLOCK(hwp_notify_lock); > static cpumask_t hwp_intr_enable_mask; > > #define HWP_GUARANTEED_PERF_CHANGE_STATUS BIT(0) > @@ -1868,7 +1868,7 @@ void notify_hwp_interrupt(void) > if (!(value & status_mask)) > return; > > - spin_lock_irqsave(&hwp_notify_lock, flags); > + raw_spin_lock_irqsave(&hwp_notify_lock, flags); > > if (!cpumask_test_cpu(this_cpu, &hwp_intr_enable_mask)) > goto ack_intr; > @@ -1876,13 +1876,13 @@ void notify_hwp_interrupt(void) > schedule_delayed_work(&all_cpu_data[this_cpu]->hwp_notify_work, > msecs_to_jiffies(10)); > > - spin_unlock_irqrestore(&hwp_notify_lock, flags); > + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); > > return; > > ack_intr: > wrmsrl_safe(MSR_HWP_STATUS, 0); > - spin_unlock_irqrestore(&hwp_notify_lock, flags); > + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); > } > > static void intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) > @@ -1895,9 +1895,9 @@ static void intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) > /* wrmsrl_on_cpu has to be outside spinlock as this can result in IPC */ > wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_INTERRUPT, 0x00); > > - spin_lock_irq(&hwp_notify_lock); > + raw_spin_lock_irq(&hwp_notify_lock); > cancel_work = cpumask_test_and_clear_cpu(cpudata->cpu, &hwp_intr_enable_mask); > - spin_unlock_irq(&hwp_notify_lock); > + raw_spin_unlock_irq(&hwp_notify_lock); > > if (cancel_work) > cancel_delayed_work_sync(&cpudata->hwp_notify_work); > @@ -1912,10 +1912,10 @@ static void intel_pstate_enable_hwp_interrupt(struct cpudata *cpudata) > if (boot_cpu_has(X86_FEATURE_HWP_NOTIFY)) { > u64 interrupt_mask = HWP_GUARANTEED_PERF_CHANGE_REQ; > > - spin_lock_irq(&hwp_notify_lock); > + raw_spin_lock_irq(&hwp_notify_lock); > INIT_DELAYED_WORK(&cpudata->hwp_notify_work, intel_pstate_notify_work); > cpumask_set_cpu(cpudata->cpu, &hwp_intr_enable_mask); > - spin_unlock_irq(&hwp_notify_lock); > + raw_spin_unlock_irq(&hwp_notify_lock); > > if (cpu_feature_enabled(X86_FEATURE_HWP_HIGHEST_PERF_CHANGE)) > interrupt_mask |= HWP_HIGHEST_PERF_CHANGE_REQ; > > base-commit: 3621a2c9142bd490af0666c0c02d52d60ce0d2a5 > -- Applied as 6.12-rc material, thanks!
Hello Rafael,
On 10/1/24 20:39, Rafael J. Wysocki wrote:
> Applied as 6.12-rc material, thanks!
Thanks. Triggered by this patch not being in today's next I looked at
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git and
didn't find it there. What am I missing?
Best regards
Uwe
Hi Uwe, On Wed, Oct 2, 2024 at 10:24 AM Uwe Kleine-König <ukleinek@debian.org> wrote: > > Hello Rafael, > > On 10/1/24 20:39, Rafael J. Wysocki wrote: > > Applied as 6.12-rc material, thanks! > > Thanks. Triggered by this patch not being in today's next I looked at > git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git and > didn't find it there. What am I missing? It's in my bleeding-edge branch where things go before hitting linux-next so they can receive some build testing coverage upfront. It'll show up in linux-next tomorrow.
diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c index aaea9a39eced..b0018f371ea3 100644 --- a/drivers/cpufreq/intel_pstate.c +++ b/drivers/cpufreq/intel_pstate.c @@ -1845,7 +1845,7 @@ static void intel_pstate_notify_work(struct work_struct *work) wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_STATUS, 0); } -static DEFINE_SPINLOCK(hwp_notify_lock); +static DEFINE_RAW_SPINLOCK(hwp_notify_lock); static cpumask_t hwp_intr_enable_mask; #define HWP_GUARANTEED_PERF_CHANGE_STATUS BIT(0) @@ -1868,7 +1868,7 @@ void notify_hwp_interrupt(void) if (!(value & status_mask)) return; - spin_lock_irqsave(&hwp_notify_lock, flags); + raw_spin_lock_irqsave(&hwp_notify_lock, flags); if (!cpumask_test_cpu(this_cpu, &hwp_intr_enable_mask)) goto ack_intr; @@ -1876,13 +1876,13 @@ void notify_hwp_interrupt(void) schedule_delayed_work(&all_cpu_data[this_cpu]->hwp_notify_work, msecs_to_jiffies(10)); - spin_unlock_irqrestore(&hwp_notify_lock, flags); + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); return; ack_intr: wrmsrl_safe(MSR_HWP_STATUS, 0); - spin_unlock_irqrestore(&hwp_notify_lock, flags); + raw_spin_unlock_irqrestore(&hwp_notify_lock, flags); } static void intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) @@ -1895,9 +1895,9 @@ static void intel_pstate_disable_hwp_interrupt(struct cpudata *cpudata) /* wrmsrl_on_cpu has to be outside spinlock as this can result in IPC */ wrmsrl_on_cpu(cpudata->cpu, MSR_HWP_INTERRUPT, 0x00); - spin_lock_irq(&hwp_notify_lock); + raw_spin_lock_irq(&hwp_notify_lock); cancel_work = cpumask_test_and_clear_cpu(cpudata->cpu, &hwp_intr_enable_mask); - spin_unlock_irq(&hwp_notify_lock); + raw_spin_unlock_irq(&hwp_notify_lock); if (cancel_work) cancel_delayed_work_sync(&cpudata->hwp_notify_work); @@ -1912,10 +1912,10 @@ static void intel_pstate_enable_hwp_interrupt(struct cpudata *cpudata) if (boot_cpu_has(X86_FEATURE_HWP_NOTIFY)) { u64 interrupt_mask = HWP_GUARANTEED_PERF_CHANGE_REQ; - spin_lock_irq(&hwp_notify_lock); + raw_spin_lock_irq(&hwp_notify_lock); INIT_DELAYED_WORK(&cpudata->hwp_notify_work, intel_pstate_notify_work); cpumask_set_cpu(cpudata->cpu, &hwp_intr_enable_mask); - spin_unlock_irq(&hwp_notify_lock); + raw_spin_unlock_irq(&hwp_notify_lock); if (cpu_feature_enabled(X86_FEATURE_HWP_HIGHEST_PERF_CHANGE)) interrupt_mask |= HWP_HIGHEST_PERF_CHANGE_REQ;
notify_hwp_interrupt() is called via sysvec_thermal() -> smp_thermal_vector() -> intel_thermal_interrupt() in hard irq context. For this reason it must not use a simple spin_lock that sleeps with PREEMPT_RT enabled. So convert it to a raw spinlock. Reported-by: xiao sheng wen <atzlinux@sina.com> Link: https://bugs.debian.org/1076483 Signed-off-by: Uwe Kleine-König <ukleinek@debian.org> --- Hello, this is deep in x86 land where I have *very* little experience and knowledge. Is sysvec_thermal() (defined in arch/x86/kernel/irq.c) really running in hard irq context? The stacktrace in the Debian bug report suggests that. So please double check my reasoning before applying this patch. Thanks Uwe drivers/cpufreq/intel_pstate.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) base-commit: 3621a2c9142bd490af0666c0c02d52d60ce0d2a5