> Tom Butts wrote:
>
> >User process fault: interruption code 0x7
> >CPU 0
> >Process insmod (pid:24 stackpage=008AB000)
> >User PSW: 070de000 80442a60    Not tainted
> >Followed by register sets [...]
>
> Well, don't leave out the interesting part ;-)
>
> This is seriously weird, I'd really like to see
> the register contents and the User Code dump.
>
> Also interesting would be the results of
>   gdb /sbin/insmod
>   disass *0x442a60
>
> Alan Cox wrote:
>
> >Thats a kernel crash. Are you sure you have matching oco and kernel ?
>
> Actually, this is a crash in user space (we dump a message
> looking similar to a kernel crash when the debugging flag
> is on ...).  Notice the 'User process fault' ...
>
> Now why insmod would crash with a data exception is
> really weird.   This can either be a floating point
> exception, but those are normally masked off unless
> specifically requested by the application, or else
> illegal operands of one of the 'decimal-operand'
> instructions, but those are never generated by the
> compiler ...


I'm sure they ARE generated, albeit incidentally.


Consider this sequence of bytes
47f0 f846 0700 0600

The first byte is a four-byte branch instruction and might be generated
by the C compiler
The third byte is the start of a six-byte decimal instruction (ZAP)
which is prone data exceptions.

In Linux it could arise from buffer overflows, string format errrors -
any of the bugs phreaks delight in exploiting.

Not that I suggest that this arose from a deliberate action. Just don't
be too surprised that you get decimal errors.


--
Cheers
John Summerfield

Microsoft's most solid OS: http://www.geocities.com/rcwoolley/

Note: mail delivered to me is deemed to be intended for me, for my
disposition.

Reply via email to