[2212] in Athena Bugs
next house rt/ethernet msgs
daemon@ATHENA.MIT.EDU (don@ATHENA.MIT.EDU)
Tue May 9 00:31:46 1989
From: <don@ATHENA.MIT.EDU>
Date: Tue, 9 May 89 00:31:19 EDT
To: bugs@ATHENA.MIT.EDU
Cc: geer@ATHENA.MIT.EDU, mk@ATHENA.MIT.EDU
i've fixed and tested the bug. i added jeff's suggested ifconfig-control
of the message, so that 'iconfig un0 debug' makes the machine act as it
does now. i also added a subtler fix, which allows the kernel to report
truly disconnected cable: it counts the cable-errors that come between
packet-receipts, and only reports when the tally goes over 50.
fortuitously, mike shanzer has an rt which shows the message frequently:
apparently, the rt's drop was previously occupied by a vax, which didn't
complain about the drop's being broken. my new kernel (which is based on 6.2)
silences mike's rt's complaints, unless we unplug the ethernet cable.
further, i've tested the ifconfig debug option on both failure-modes;
it turns the spurious messages on & off again, as desired, while it only
delays the correct messages.
unhappily, one could argue that all of the messages that mike's machine sends
are correct, since its cable is not too long. if the next house machines'
ethernet cards are not complaining on every packet's transmission, there's
nothing more i can do; it's too hard/subtle to distinguish between high & low
failure-rates. if, however, the next house rt's cards are consistently
reporting these spurious errors, then i can add a simple algorithm that will
reject the spurious errors, while still reporting intermittent errors &
true disconnects.
the modified kernel is /mit/don/vmunix; the modified if_un.c is privately held.
-don
