dhcpcd and multiple wm(4) on PC Engines, crashy

[Greg Troxel <gdt%lexort.com@localhost>]

I have 2 PC Engines apu2d4 boxes in service, and both have been rock
solid, with NetBSD 9 and now 10.  Probably earlier NetBSD on one of
them.

A 3rd was crashy/hangy, irregularly sort of week or so with 9, and more
like a few days with 10.  Mostly it was silent hang (there is no console
connected).

The 3rd had a different problem - a bad ethernet cable, moved from wm0
to wm1 and then replaced.

I realized that the configuration of dhcpcd was different

  solid1: dhcpcd $int in ifconfig.wm0, wm1 manual (router)

  solid2: dhcpcd=YES, with all of wm0/wm1/wm2 brought up, but only wm0
  having a cable to a switch

  crashy3: dhcpcd=YES, same as solid2, but wm1 not wm0.


Then, I changed the configuration of crashy3 to "allowinterfaces wm1"
and it has been up for 54 days.

So, I suspect there is something about wm(4) where whatever dhcpcd does
to it provokes a locking bug -- but that's way out on a limb.


Trying to probe further, on crashy3 I ran "dhcpcd wm2" and saw that wm2
was UP but still no carrier.   Survived several  days.

Today did "dhcpcd wm0" and wm0 did not come UP.

I do not find a way to ask the running dhcpcd for status (which
interfaces are you trying on, how are they), and `dhcpcd -U` does not
work:

  #> dhcpcd -U
  dhcpcd_readdumptimeout

yet

  _dhcpcd   687  0.0  0.1  19108   2284 ?      S    19Mar26   1:26.10 dhcpcd: [manager] [ip4] [ip6] 
  root      688  0.0  0.0  22840   1584 ?      S    19Mar26   1:14.57 dhcpcd: [privileged proxy] 
  _dhcpcd   692  0.0  0.0  21680   1376 ?      S    19Mar26   0:05.03 dhcpcd: [network proxy] 
  _dhcpcd   694  0.0  0.0  18820   1380 ?      S    19Mar26   0:02.47 dhcpcd: [control proxy] 


To be clear I am not suggesting there is a bug in dhcpcd that causes the
crash/hang.  I am also not 100% sure that this crashy3 box is 100%
hardware-healthy -- but so far this feels like software.

I wonder if anyone has wisdom or related experience.