Re: [Intel-wired-lan] igc: Network failure, reboot required: igc: Failed to read reg 0xc030!

[Jesper Juhl]

On Fri, 13 Sept 2024 at 09:02, Lifshits, Vitaly
<vitaly.lifshits@xxxxxxxxx> wrote:
>
> On 9/12/2024 10:45 PM, Jesper Juhl wrote:
> >> Would you be able to decode the stack trace? It may be helpful
> >> to figure out which line of code this is:
> >>
> >>   igc_update_stats+0x8a/0x6d0 [igc
> >> 22e0a697bfd5a86bd5c20d279bfffd
> >> 131de6bb32]
> >
> > Of course. Just tell me what to do.
> >
> > - Jesper
> >
> > On Thu, 12 Sept 2024 at 17:37, Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
> >>
> >> On Thu, 12 Sep 2024 15:03:14 +0200 Jesper Juhl wrote:
> >>> It just happened again.
> >>> Same error message, but different stacktrace:
> >>
> >> Hm, I wonder if it's power management related or the device just goes
> >> sideways for other reasons. The crashes are in accessing statistics
> >> and the relevant function doesn't resume the device. But then again,
> >> it could just be that stats reading is the most common control path
> >> operation.
> >>

I doubt it's related to power management since the machine is not idle
when this happens.

> >> Hopefully the Intel team can help.
> >>
> >> Would you be able to decode the stack trace? It may be helpful
> >> to figure out which line of code this is:
> >>
> >>    igc_update_stats+0x8a/0x6d0 [igc
> >> 22e0a697bfd5a86bd5c20d279bfffd131de6bb32]
>
I didn't manage to decode it with the distro kernel. I'll build a
custom kernel straight from the git repo and wait for the problem to
happen again, then I'll report back with a decoded trace.

> Hi Jasper,
>
> I agree with Kuba that it might be related to power management, and I
> wonder if it can be related to PTM.
> Anyway, can you please share the following information?
>
> 1. Is runtime D3 enabled? (you can check the value in
> /sys/devices/pci:(pci SBDF)/power/control)

$ cat /sys/devices/pci0000\:00/power/control
auto

> 2. What is the NVM version that your NIC has? (ethtool -i eno1)

$ sudo ethtool -i eno1
driver: igc
version: 6.10.9-arch1-2
firmware-version: 1082:8770
expansion-rom-version:
bus-info: 0000:0c:00.0
supports-statistics: yes
supports-test: yes
supports-eeprom-access: yes
supports-register-dump: yes
supports-priv-flags: yes

> 3. Can you please elaborate on you bug?
> Does it happen while the system is in idle state?

I don't know. It might, but I've only ever observed it while actively
using the machine. I usually notice the problem when watching a
youtube video or playing an online game and suddenly the network
connection dies.

> Does it run any
> traffic?

Yes, there's usually always network traffic when the problem occurs.

> What is the system's link partner (switch? other NIC?)

It's a "tp-link" switch: TL-SG105-M2 5-Port 2.5G Multi-Gigabit Desktop Switch

Kind regards
 Jesper Juhl