Correction and new findings.
Ignore my earlier PME-Enable theory: the kernel DOES arm the NIC
correctly. Reading lspci immediately after a keyboard-wake shows
"PME-Enable+" on the NIC, which is then cleared on resume. The
"PME-Enable-" seen while the system is up is normal.
The actual finding: the NIC generates a PME on the magic packet while
in s2idle, the root port receives it, but the system does not wake.
before suspend: IRQ 125 count = 4
(suspend, send magic packet, no wake, wake via USB keyboard instead)
after resume: IRQ 125 count = 5
125: ... IR-PCI-MSI-0000:00:1c.0 0-edge PCIe PME, aerdrv, PCIe bwctrl
lspci -s 00:1c.0 -vvv
RootCtl: ... PMEIntEna+ CRSVisible-
RootSta: PME ReqID 5600, PMEStatus- PMEPending-
5600 = 0000:56:00.0, the igc NIC. So the wake event is generated and
counted, but it does not bring the system out of s2idle - the interrupt
appears to be serviced only once the system is awake for another reason.
This looks like the PCIe PME interrupt not acting as a wakeup source in
s2idle on this platform, rather than an igc problem.
Root port 00:1c.0 is enabled as a wakeup source
(/sys/bus/pci/devices/0000:00:1c.0/power/wakeup = enabled), as is the
NIC. Neither pcie_pme=nomsi nor pcie_ports=compat changes anything.
Hardware: ASUS NUC 15 (NUC15CR, latest BIOS), Intel I226-V / igc,
kernel 7.0.0-38-generic.
State | USB keyboard wake | magic packet wake
----------|-------------------|-------------------
S5 | works | works
s2idle | works | fails (PME generated, no wake)
deep (S3) | broken | broken
Separate issue: mem_sleep=deep is broken independently of WoL - any
wake attempt powers the machine off instead of resuming.
Also at boot: "_OSC: OS now controls [... PME ...]" and
"pcieport 0000:00:1c.0: PME: Spurious native interrupt!".
Hope it helps
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2171282
Title:
Wake-on-LAN from s2idle does not work on ASUS NUC 15 (Intel I226-V,
igc); wake from S5 and USB wake from s2idle both work
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2171282/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
