> > >>>>> "PM" == Pavel Machek <pavel@atrey.karlin.mff.cuni.cz> writes:
> >
> > PM> Hi! Do you want 2.0.33 to oops? Well, compile 2.1.X, take vmlinux,
> > PM> and try to execute it. ./vmlinux. It segfaults. Well, it *should*
> > PM> segfault. But it also produces kernel oops. Which is BAD. Oops is
> > PM> at iret, so it is pretty innocent.
> >
> > You forgot to tell whether you did it as an unpriviledged user or as
> > root. If you did it as root the kernel has every right to do as it
> > pleases. (Unless you can point to a relevant specification that says
> > otherwise, of course.)
>
> As an ordinary user. It happens since 2.0.0; I've said it
> many times, but apparently nobody cares.
This "works" for me too, under 2.0.33:
general protection: 0000
CPU: 0
EIP: 0010:[<0010a6bd>]
EFLAGS: 00010202
eax: 00000000 ebx: 080ab808 ecx: 080af5d8 edx: 00000000
esi: 401323a0 edi: 080ab808 ebp: bfff735c esp: 01e24fec
ds: 002b es: 002b fs: 002b gs: 002b ss: 0018
Process vmlinux (pid: 215, process nr: 5, stackpage=01e24000)
Stack: c0100000 00000023 00000202 bffffca0 0000002b
Call Trace:
Code: cf 89 f6 89 e1 51 f7 41 38 00 00 02 00 75 2c 53 e8 fa fb ff
Segmentation fault
Now ksymoops is segfaulting. Argh. Anyway:
0010a638 T ret_from_sys_call
___
Trevor Johnson