my simh-vax crashed: panic: pmap_enter on PG_SREF page

[matthew green <mrg%eterna23.net@localhost>]

because we can't have only nice things this season, my simh crashed
while building netbsd.

this was a couple of days ago -current, built on amd64, building the
same source tree in simh-vax.

db> bt
panic: pmap_enter on PG_SREF page
Stack traceback :
0x94ba9c40: vpanic+0x179(0x802a0a35,0x94ba9cd8)
0x94ba9c60: vprintf+0x0(0x802a0a35)
0x94ba9c8c: pmap_enter+0x60a(0x8e8aec00,0x74a00000,0x2bdb000,0x3,0x21)
0x94ba9cf4: uvm_fault_lower_enter+0x57(0x94ba9df4,0x94ba9e10,0,0x838cc420,0x80a26d18)
0x94ba9d3c: uvm_fault_internal+0xd02(0x8cda5e00,0x74a00000,0x1,0)
0x94ba9ebc: trap+0x148(0x94ba9fb4)
db> ps
PID    LID S CPU     FLAGS       STRUCT LWP *               NAME WAIT
17504>17504 7   0     40000           8d400800            cc1plus

so it was a cc1plus processing a trap, and doing what appears to be
re-mapping a page previously mapped with pmap_kenter_pa() (the only
way PG_SREF should be set.)

this is a valid assert -- if it's been pmap_kenter_pa()'d it should
never be otherwise managed in kernel or userspace.

any one have an idea what is up?  i have ddb still for now.  this
system is a 256mb microvax3900 instance in open-simh latest pkgsrc.

thanks.


.mrg.