New: e1000e: Sometimes causes resume from suspend to RAM to fail

http://bugzilla.kernel.org/show_bug.cgi?id=11545

           Summary: e1000e: Sometimes causes resume from suspend to RAM to
                    fail
           Product: Drivers
           Version: 2.5
     KernelVersion: 2.6.27-rc6
          Platform: All
        OS/Version: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: Network
        AssignedTo: [email protected]
        ReportedBy: [email protected]
                CC: [email protected]

Latest working kernel version: Unknown
Earliest failing kernel version: Unknown; first seen with 2.6.27-rc4
Distribution: Debian
Hardware Environment: HP Compaq 2510p laptop
Software Environment: Debian unstable

Problem Description:
Most of the time the laptop resumes perfectly from suspend, but sometimes (about 1 in 5-10 times) it fails to resume fairly early (with the display still off). Only solution at that point is a hard power off.

I've used pm_trace to try to find out the cause of the failure. That gave:
Magic number: 0:983:221
hash matches drivers/base/power/main.c:350
e1000e 0000:00:19.0: hash matches
rtc_cmos 00:06: setting system clock to 2020-12-11 09:12:26 UTC (1607677946)

So it looks like the e1000e device or driver is the cause. (Line 350 in power/main.c has 'TRACE_RESUME(0);'.)

This was with e1000e compiled into the kernel (as I wanted to use netconsole for another issue), but I've also seen the failure with e1000e modular. I will now make e1000e modular again and unload it before suspending to see if that makes resume stable.

Note that the eth0 NIC is unused (no cable connected). I use the wireless NIC instead.

Suggestions how to do further tracing or instrumentation would be very welcome.