Re: "Exception at ..." with 2.1.9 and 2.1.10 kernels

Richard Henderson (richard@atheist.tamu.edu)
Tue, 19 Nov 1996 19:48:03 -0600 (CST)


> It was a null pointer being passed. This does bring up a minor question of
> whether a null pointer should be handled here or higher up.

Well, that's sort of the whole point of the exceptions in 2.1.

In 2.0, we spent a lot of time higher up verifying pointers and it was
somewhat error prone not to miss one (esp with things like ioctl in which
sometimes somethin is a pointer and sometimes now).

In 2.1, we want to rid ourselves of the time spent verifying pointers,
esp since they are almost always valid anyway. So we let the hardware
do it for us in the form of an exception. Random exceptions are bad,
and are seen in the form of kernel oopses. But when we are prepared
for one and know what to do if we get one, they can be very friendly.

Anyway, the message that pops up is for debugging the mechanism that
was put in place for this wonderful new feature. And since it is also
useful as a clue that userland software needs to be fixed, it may very
well stay around for a while.

r~