[1299] in Athena Bugs
progress on RT's garbage-packet bug
daemon@ATHENA.MIT.EDU (don@ATHENA.MIT.EDU)
Wed Nov 2 22:38:02 1988
From: <don@ATHENA.MIT.EDU>
Date: Wed, 2 Nov 88 22:37:16 EST
To: rt-testers@ATHENA.MIT.EDU
i have a simpler patch for this bug: by introducing a delay at the end
of the ungermann-bass initializing routine unzap(), i've made the bug go away.
my rationale for introducing this klooge in this routine, is as follows:
ifconfig causes the garbage packet to go out during an ioctl() call:
ioctl ->...-> unioctl->arpwhoha->unoutput->unstart, which sends garbage.
in the course of debugging with the rt kernel debugger, i found that
simply resuming execution after a certain breakpoint caused the arpwhoha()
call to send a good packet, instead of garbage.
so, we have here a timing-bug, presumably a symptom of some mis-coding.
i did a laborious search of the code leading to the above unstart() call,
before finding that resuming exec'n from various breakpts in unzap() can
cause the unstart() to send either garbage, the correct packet, or nothing
at all!
i'm treating this patch's success as a hint, and am continuing to work on the
bug. -don
