FYI: I went through list of the kernel parameters and stumbled upon
this parameter 'pcie_port_pm=off'
It made the problem disappear, but it is probably still worth it to debug it.
If you have any kernel patches and/or tests,debug dumps,etc for me to
try/give I'll be happy to do so.
Best regards,
Maxim Levitsky
On Fri, Apr 24, 2020 at 11:49 AM Maxim Levitsky <maximlevitsky@xxxxxxxxx> wrote:
>
> Hi!
>
> I have the following problem:
> When I suspend my system, and then resume it using one of the wakeup sources
> (tested with two usb keyboards attached to different usb controllers
> and with a wifi card, using
> wake on wan wakeup) then on resume I get ACPI interrupt storm which
> leads to ACPI interrupt be disabled.
>
> Apr 24 11:18:06 starship kernel: irq 9: nobody cared (try booting with
> the "irqpoll" option)
> Apr 24 11:18:06 starship kernel: CPU: 0 PID: 0 Comm: swapper/0 Not
> tainted 5.3.7-301.fc31.x86_64 #1
> Apr 24 11:18:06 starship kernel: Hardware name: Gigabyte Technology
> Co., Ltd. TRX40 DESIGNARE/TRX40 DESIGNARE, BIOS F4c 03/05/2020
> Apr 24 11:18:06 starship kernel: Call Trace:
> Apr 24 11:18:06 starship kernel: <IRQ>
> Apr 24 11:18:06 starship kernel: dump_stack+0x5c/0x80
> Apr 24 11:18:06 starship kernel: __report_bad_irq+0x35/0xa7
> Apr 24 11:18:06 starship kernel: note_interrupt.cold+0xb/0x63
> Apr 24 11:18:06 starship kernel: handle_irq_event_percpu+0x6f/0x80
> Apr 24 11:18:06 starship kernel: handle_irq_event+0x28/0x48
> Apr 24 11:18:06 starship kernel: handle_fasteoi_irq+0x86/0x130
> Apr 24 11:18:06 starship kernel: handle_irq+0x1c/0x30
> Apr 24 11:18:06 starship kernel: do_IRQ+0x4b/0xd0
> Apr 24 11:18:06 starship kernel: common_interrupt+0xf/0xf
> Apr 24 11:18:06 starship kernel: </IRQ>
> Apr 24 11:18:06 starship kernel: RIP: 0010:poll_idle+0xa4/0x111
> Apr 24 11:18:06 starship kernel: Code: 04 40 48 c1 e0 05 45 8b 64 04
> 54 4d 69 e4 e8 03 00 00 65 4c 8b 34 25 c0 6b 01 00 b8 c9 00 00 00 49
> 8b 16 83 e2 08 75 >
> Apr 24 11:18:06 starship kernel: RSP: 0018:ffffffff8c403e28 EFLAGS:
> 00000246 ORIG_RAX: ffffffffffffffb1
> Apr 24 11:18:06 starship kernel: RAX: 0000000000000076 RBX:
> 0000001314a7d925 RCX: ffff98852a853050
> Apr 24 11:18:06 starship kernel: RDX: 0000000000000000 RSI:
> 0000000000000003 RDI: 0000001153236883
> Apr 24 11:18:06 starship kernel: RBP: ffff98852a853000 R08:
> 0000001314a7d925 R09: 000000000000016c
> Apr 24 11:18:06 starship kernel: R10: ffff98853ce276c4 R11:
> ffff98853ce276a4 R12: 00000000000007d0
> Apr 24 11:18:06 starship kernel: R13: 0000000000000000 R14:
> ffffffff8c413780 R15: ffffffff8c413780
> Apr 24 11:18:06 starship kernel: ? poll_idle+0x22/0x111
> Apr 24 11:18:06 starship kernel: cpuidle_enter_state+0x7c/0x420
> Apr 24 11:18:06 starship kernel: cpuidle_enter+0x29/0x40
> Apr 24 11:18:06 starship kernel: do_idle+0x1cf/0x250
> Apr 24 11:18:06 starship kernel: cpu_startup_entry+0x19/0x20
> Apr 24 11:18:06 starship kernel: start_kernel+0x54f/0x56c
> Apr 24 11:18:06 starship kernel: secondary_startup_64+0xa4/0xb0
> Apr 24 11:18:06 starship kernel: handlers:
> Apr 24 11:18:06 starship kernel: [<00000000ac140a6e>] acpi_irq
> Apr 24 11:18:06 starship kernel: Disabling IRQ #9
>
>
> If I wake the system with the power button this doesn't happen.
>
> I tried so far the following:
>
> I tried various combinations of acpi_sci=level/edge acpi_sci=low/high
> - this only hides the problem,
> in the way that power button stops working.
>
> I tried to make acpi interrupt handler lie about IRQ handled just to see
> when the interrupt storm ends (make acpi_irq return IRQ_HANDLED always)
>
> In this case the storm keep on going till I suspend the system again
> and wake it with the power button.
>
> I checked PME status on the chain of pcie bridges leading to the wifi
> device and looks like none have PME pending.
> (I can do the same with USB controller but I guess that would be pointless)
>
> I also checked /sys/firmware/acpi/interrupts/
>
> I have two invalid GPEs that became raised (STS is set) after resume,
> but that happens with power button as well (gpe04 and gpe05)
> I tried to remove the check for invalid gpes in 'counter_set' to try
> and ACK them, and that indeed cleared the STS bit but the storm kept
> on going.
>
> Anything else should I try?
>
> My motherboard is TRX40 Designare running latest F4C bios. I run the
> latest (git pullled yesterday) mainline kernel (5.7-rc2) with few my
> own
> patches that shouldn't be related. I tested this to happen on fedora
> kernel 5.3.7 as well.
>
> I use Fedora 31 if that makes any difference.
>
> ACPI tables attached.
>
> Best regards,
> Maxim Levitsky
