On Wed, Mar 20, 2024 at 03:56:29PM +0100, Heiner Kallweit wrote: > On 20.03.2024 15:15, Andy Shevchenko wrote: > > On Wed, Mar 20, 2024 at 03:37:34PM +0200, Jarkko Nikula wrote: > >> On 3/20/24 2:27 PM, Heiner Kallweit wrote: > >>> On 19.03.2024 22:11, Heiner Kallweit wrote: > >>>> On a N100-based mini pc I see the following. I found older reports with the same symptom, > >>>> but root cause seems to be different. > >>>> > >>>> - Interrupt 27 is not shared in my case > >>>> - I checked register values on entering the ISR. Interrupt mask and DW_IC_RAW_INTR_STAT > >>>> are both 0. > >>>> - After an interrupt storm of 100,000 interrupts the interrupt gets disabled > >>>> - The issue affects channel 0 only > >>>> > >>>> If not the I2C IP, then who else can touch the interrupt line? > >>> > >>> I noticed that after including INTEL_IDMA64 in my config the problem no longer occurs. > >>> So there seems to be a dependency. Should it be reflected in Kconfig, e.g. > >>> make MFD_INTEL_LPSS dependent on INTEL_IDMA64, or let it imply INTEL_IDMA64? > >>> > >> Hmm.. interesting. I'd say BIOS perhaps has left the IDMA active and is > >> generating interrupts until the idma64 driver acknowledges it. > >> > >> There should not be generic dependency since the i2c_designware is not using > >> the DMA and a quick test on one platform where idma64 and i2c_designware are > >> sharing the same interrupt without CONFIG_INTEL_IDMA64 not set didn't show > >> similar behavior. > >> > >> Andy: Do you have any additional ideas or debug hints to this? > > > > Can you share `cat /proc/interrupts` in non-working and working cases? > > > > Just to confirm: loading idma64 driver fixes the issue, correct? > > > I spoke too soon. Loading idma64 didn't actually fix the problem. > What happened: > Runtime PM kicked in and set the device to D3. Therefore reads to the idma64 > status register returned 0xffffffff, resulting the idma64 ISR returning > IRQ_HANDLED. Okay, bug in idma4 driver then :-) Good catch! > IRQ number still was rapidly growing. > > After disabling RPM the problem is back. > Who else could pull the level-triggered interrupt high? Need to understand a bit about RTE programming. You need to add 'apic=debug' to the kernel command line and look for IRQ 27 related bits, and also makes sense to collect this grep 0000001b /sys/kernel/debug/pinctrl/*/pins it will show the GPIO pins related to RTE 27 (and we can check if direct IRQ is enabled there). > [ 4.642978] irq 27: nobody cared (try booting with the "irqpoll" option) > [ 4.643067] CPU: 3 PID: 0 Comm: swapper/3 Not tainted 6.8.0-next-20240312+ #5 > [ 4.643133] Hardware name: Default string Default string/Default string, BIOS ADLN.M6.SODIMM.ZB.CY.015 08/08/2023 > [ 4.643215] Call Trace: > [ 4.643241] <IRQ> > [ 4.643266] dump_stack_lvl+0x81/0xe0 > [ 4.643319] dump_stack+0x10/0x20 > [ 4.643358] __report_bad_irq+0x2b/0xc0 > [ 4.643403] note_interrupt+0x28f/0x2d0 > [ 4.643442] ? __this_cpu_preempt_check+0x13/0x20 > [ 4.643496] handle_irq_event+0x70/0x80 > [ 4.643541] handle_fasteoi_irq+0x90/0x210 > [ 4.643586] __common_interrupt+0x6f/0x140 > [ 4.643633] common_interrupt+0xab/0xd0 > [ 4.643672] </IRQ> > [ 4.643694] <TASK> > [ 4.643720] asm_common_interrupt+0x27/0x40 > [ 4.643761] RIP: 0010:cpuidle_enter_state+0xff/0x4f0 > [ 4.643815] Code: 8b 00 48 0f a3 05 d1 5f 9d 00 0f 82 fd 02 00 00 31 ff e8 f4 e0 69 ff 80 7d cf 00 0f 85 cc 02 00 00 e8 65 bd 77 ff fb 45 85 ff <0f> 88 f4 01 00 00 49 63 f7 4c 89 f2 48 8d 04 76 48 8d 04 86 49 8d > [ 4.643957] RSP: 0018:ffffb21dc017fe58 EFLAGS: 00000202 > [ 4.644010] RAX: 000000000007987d RBX: ffffd21dbfdac690 RCX: 0000000000000000 > [ 4.644069] RDX: 0000000000000000 RSI: ffffffff8f724a02 RDI: ffffffff8f710113 > [ 4.644128] RBP: ffffb21dc017fe98 R08: 0000000000000001 R09: 0000000000000000 > [ 4.644186] R10: 0000000000000001 R11: ffff91cfb7bad824 R12: 0000000000000001 > [ 4.644244] R13: ffffffff8f9b69a0 R14: 0000000114be1938 R15: 0000000000000001 > [ 4.644315] ? cpuidle_enter_state+0xfb/0x4f0 > [ 4.644367] cpuidle_enter+0x29/0x40 > [ 4.644409] call_cpuidle+0x1d/0x40 > [ 4.644450] do_idle+0x1c2/0x220 > [ 4.644489] cpu_startup_entry+0x24/0x30 > [ 4.644529] start_secondary+0xf9/0x100 > [ 4.644571] common_startup_64+0x13e/0x148 > [ 4.644626] </TASK> > [ 4.644649] handlers: > [ 4.644674] [<000000008b061326>] idma64_irq [idma64] > [ 4.644852] [<000000008273d6d5>] i2c_dw_isr [i2c_designware_core] > [ 4.644920] Disabling IRQ #27 > > > CPU0 CPU1 CPU2 CPU3 > 8: 0 0 0 0 IR-IO-APIC 8-edge rtc0 > 9: 0 44 0 0 IR-IO-APIC 9-fasteoi acpi > 16: 0 3 0 0 IR-IO-APIC 16-fasteoi i801_smbus, idma64.6 > 27: 0 0 0 112018 IR-IO-APIC 27-fasteoi idma64.0, i2c_designware.0 > 29: 0 0 0 0 IR-IO-APIC 29-fasteoi idma64.2, i2c_designware.2 > 31: 0 0 0 0 IR-IO-APIC 31-fasteoi idma64.4, i2c_designware.4 > 32: 0 0 0 0 IR-IO-APIC 32-fasteoi idma64.5, i2c_designware.5 > 33: 0 0 0 0 IR-IO-APIC 33-fasteoi idma64.3, i2c_designware.3 > 37: 0 0 0 0 IR-IO-APIC 37-fasteoi idma64.7 > 40: 0 0 0 0 IR-IO-APIC 40-fasteoi idma64.1, i2c_designware.1 > -- With Best Regards, Andy Shevchenko
