I don't think it does, not on any specific size. I think it, that
version, just depends on its size and replacement policy being such
that that stale entries get replaced before they cause trouble.
[a 300-octet line - TLDR: "depends" may be too strong]
Possibly. "Depends on" in the sense of "this is the reason it works at
all in view of the missing invalidations". It _would_ also be accurate
to call it a masked bug; I wrote from the mindset of "why did this even
kinda-sorta-mostly work on a real KA630?".
Given how tiny the KA630's TB is, I suspect it would be quite hard to
find a realistic workload that failed for lack of invalidations,
probably somewhat difficult to construct a synthetic one either using
this kernel (it probably could be done without too much trouble if you
don't care about even attempting to be a UNIXy OS).
It does occur to me that I think I tried, years ago, to run NetBSD on a
more recent VAX - one of the desktop boxes, I think - and failed. At
the time I probably wrote it off to hardware issues, but I don't
remember the symptom enough, so it could have been this.
/~\ The ASCII Mouse
\ / Ribbon Campaign
X Against HTML mouse%rodents-montreal.org@localhost
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B