Report on NetBSD-CURRENT and `hpcmips` on MobilePro 780

["Alex Maestas" <port-mips%se30.xyz@localhost>]

Hi,

Recently I've come across an `hpcmips`-compatible WinCE device, an 
NEC MobilePro 780, and loaded up NetBSD on it.

This post confirms previous reports that things broke on the 
physical hardware sometime between NetBSD-7.1 and 7.2; 7.2 did not 
boot, hanging after detecting apmdev, apparently while detecting 
`pcic0`, logging messages about deprecated powerhook_establish:
https://infosec.exchange/@atax1a/110885572993394050

However, I have now built NetBSD-CURRENT as of the 16th, while 
still using the 7.1 userland, and the machine now completes its 
boot:

NetBSD tochtli.home.se30.xyz 10.99.7 NetBSD 10.99.7 (GENERIC) #0: 
Wed Aug 16 07:25:40 GMT 2023 
atax1a@tecpatl.local:/Users/atax1a/package/netbsd-src/obj/sys/arch/hpcmips/compile/GENERIC 
hpcmips

This configuration IS producing some strange behavior, though. 
Pipes in the shell seem to gain garbage data, but only sometimes, 
and occasionally things just drop core inexplicably:

tochtli# yes | head -n4000 | tail | od -h
Bus error (core dumped)
tochtli# yes | head | tail | od -h
0000000     0a79    0a79    0a79    0a79    0a79    0a79    0a79 
0a79
0000020     0a79    0a79
0000024
tochtli# yes | head | tail | od -h
0000000     0000    0000    0000    0000    0000    0000    0000 
0000
0000020     1c9c    0041    0000    0000    0032    0000    ffff 
ffff
0000040     ffff    ffff    ffff    0003    0000    0000    0000 
0000
0000060     0000    0000    0000    0000    0000    0000    0000 
0000
*
0004000     0a79    0a79    0a79    0a79    0a79    0a79    0a79 
0a79
0004020     0a79    0a79
0004024
tochtli# yes | head -n4000 | tail | od -h
0000000     0a79    0a79    0a79    0a79    0a79    0a79    0a79 
0a79
0000020     0a79    0a79
0000024
tochtli# yes | head -n4000 | tail | od -h
0000000     0a79    0a79    0a79    0a79    0a79    0a79    0a79 
0a79
0000020     0a79    0a79
0000024
Memory fault (core dumped)

My next step is to load up the -CURRENT userland and see if this 
breaks things further, or if it improves the situation.

Regards (and I hope my mail client is sufficiently configured for 
mailing list posts),
--
Alex Maestas