Hi Intel team, I run a home lab server on CWWK CW-AD4L-N V1 and recently updated my kernel to 6.12.38 with an upgrade from Debian Bookworm to Trixie.
I've observed that using the new kernel, the network connection drops every couple of hours, and in the worst case scenario, the connections would completely disappear and the machine would have to be rebooted. Going back to 6.12.31 fixes the issue. From https://launchpad.net/debian/+source/linux/+changelog it seems that only this change was between .37+ and .31 - igc: disable L1.2 PCI-E link substate to avoid performance issue This led me to this thread https://lore.kernel.org/intel-wired-lan/[email protected]/ and it seems that the fix is for i226V with speed cap. But in my case, it wasn't capped to start with. I can easily saturate my internet pipe using the 6.12.31 kernel without the fix. Idle Latency: 1.15 ms (jitter: 0.12ms, low: 1.04ms, high: 1.38ms) Download: 519.37 Mbps (data used: 397.8 MB) 7.27 ms (jitter: 0.55ms, low: 1.28ms, high: 11.43ms) Upload: 522.60 Mbps (data used: 235.1 MB) 4.96 ms (jitter: 0.37ms, low: 2.77ms, high: 5.75ms) Packet Loss: Not available. Result URL: https://www.speedtest.net/result/c/bcacca17-8666-4bcf-8300-10c6920ab420 Output of 'lspci -vv -s <x>' w.r.t L1SubCtl1 is: Capabilities: [1e0 v1] L1 PM Substates L1SubCap: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2+ ASPM_L1.1+ L1_PM_Substates+ PortCommonModeRestoreTime=55us PortTPowerOnTime=70us L1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2- ASPM_L1.1- T_CommonMode=0us LTR1.2_Threshold=138240ns L1SubCtl2: T_PwrOn=70us Kernel driver in use: igc Can we hide this behind a module option so that systems that aren't speed capped won't be affected by the patch? Happy to run any test or debug if needed. Thank you, Chia-Wei Yeh -- "If it looks like a duck, walks like a duck, and quacks like a duck, then to the end user it's a duck, and end users have made it pretty clear they want a duck; whether the duck drinks hot chocolate or coffee is irrelevant."
