Oops in schedule in 2.5.24

[Petr Vandrovec]

Hi,
  I decided to give 2.5.24 a try now when IDE94 and IDE95 were out.
So I applied IDE94/95 and built usual kernel (PIII, SMP) for my UP P4
host. And it died with

Unable to handle kernel NULL reference to 0x0000002A
Call trace: schedule + 491/1102
            sys_sched_yield + 444/455
            syscall

866) was 0x00000002, so next happened to be 0xFFFFFFE2, and then it
died when context_switch() tried to access next->mm :-( No idea how 0x000002
happened to be in queue->next.

Another explanation is that rq->idle was 0xFFFFFFE2, and rq->nr_running
was 0. But it looks impossible too.

  Does anybody have any idea what's wrong? Crash happened few minutes
after boot. For now I reverted back to 2.5.21-changeset1.493, which 
works flawlessly.

  Box is running slapd (which invoked yield()) and grabs TV teletext
using bttv. Network card driver is e100. There should be always one
process ready to run: dnetc.
                                        Thanks,
                                            Petr Vandrovec
                                            [e-mail]