Subject: Things leading to the death of serial and init buggedness
To: None <port-mac68k@NetBSD.ORG>
From: Valtteri Vuorikoski <vuori@sci.fi>
List: port-mac68k
Date: 10/19/1995 14:27:58
After waving various dead chickens over things related and
unrelated to the SCC and serial in general, I got annoyed with
cu and wrote a small comm program to help me with this thing,
and found out the following things about what leads to serial dying:
- with old driver, cu. Run it once and after that serial is dead
until next reboot. With new, it either has no effect or hangs the
machine totally when exiting
- (at least) beeps (or maybe it's some vt control characters. Some
of them work anyway) kill it with my own program. It doesn't seem
to have problems working until a beep comes over the line. I noticed
this when in zsh I typed 'to' instead of 'top' and it asked me whether
to correct it to 'top', and beeped. The beep never got as far as the
speakers, when serial dropped dead. Exiting the program caused it to
get stuck in close(), but C-z'ing it and killing it got it out, and
on subsequent starts it exits nicely. top's 'Enter number of seconds
between updates'-prompt also seems to cause it to get stuck, though
rts/cts stays up, so I'm not sure whether that's caused by my program
or what
- the new serial driver doesn't appear to be working very much, I don't
know whether that's caused my just my machine. When I use my own program,
some characters from me to the remote seem to get through (enter), since
it appears to be sending a new prompt, which however doesn't appear here,
but instead some character sequences. Pressing enter many times, it appears
that the sequences repeat. They almost always end in (. Some common ones are:
ub( ur( eb( [Ksb( Ksb( KCr(
(spaces are just separators here)
A thing which seems more generic is that init gets a segmentation fault
if you hup it in single-user:
init: kernel security level changed from 0 to 1
init: fatal signal: Segmentation fault
NetBSD/mac68k (Amnesiac) (ttye0)
login:
After a few keypresses at that point, it panics:
panic: init died (signal 0, exit 11)
ps in ddb shows init still in the process list. Could someone see if this
happens on other machines too and send-pr if appropriate. My init is fairly
-current.
--
'Good-bye and hello, as always'