OPENCORES.ORG

Hi!

Hi Robert,
i'm already aware of this problem when building linux 2.4 with gcc-3.4.2. at
the time i was looking at it i traced it to be related to gcc. kernel and
compiler should be of approx. the same version. i suggest you use gcc-3.2.3
with linux-2.4 (you shouldn't have any problems with it).

best regards,
p.

* Robert Millan (rmh@a...) wrote:
> Hi!
> 
> I got an Oops when booting Linux in or1ksim.  I was using linux-2.4 from CVS,
> compiled with gcc 3.4.2, running on or1ksim from HEAD branch.
> 
> Boot log in uart0.tx attached, simulator log (run with verbose = 1 and
> debug = 1) attached too.  I haven't used a higher debug level because lots of
> errors were printed already and most of them didn't look critical.
> 
> Linux boot log lists 0xc000b898 as the PC value when exception happens.  Here's
> a dasm output of this address:
> 
>         /usr/src/or1k/linux/linux-2.4/include/linux/list.h:83
>         c000b898:       d4 04 18 00     l.sw 0x0(r4),r3
> 
> Corresponding C code in list.h:
> 
>         static inline void __list_del(struct list_head *prev, struct list_head *next)
>         {       
>                 next->prev = prev;
>                 prev->next = next;  /* line 83 */
>         }
> 
> Linux boot log lists 0xc000b63c as a "Call Trace" address.  Since this address
> is in the stack trace, I guess it means this is the function that called
> __list_del (please correct me if i'm confused).  The info for 0xc000b63c is:
> 
>         /usr/src/or1k/linux/linux-2.4/kernel/sched.c:578
>         c000b63c:       84 61 00 2c     l.lwz r3,0x2c(r1)
>         c000b640:       84 83 00 28     l.lwz r4,0x28(r3)
>         c000b644:       bc 24 00 02     l.sfnei r4,0x2
>         c000b648:       0c 00 00 c1     l.bnf c000b94c <_schedule+0x390>
>         c000b64c:       15 00 00 00     l.nop 0x0
> 
> Corresponding C code in kernel/sched.c:
> 
>         /* move an exhausted RR process to be last.. */
>         if (unlikely(prev->policy == SCHED_RR))                 /* line 578 */
>                 if (!prev->counter) {
>                         prev->counter = NICE_TO_TICKS(prev->nice);
>                         move_last_runqueue(prev);
>                 }
> 
> Which once preprocessed reads:
> 
>         if (__builtin_expect ((prev->policy == 2), 0))
>                 if (!prev->counter) {
>                         prev->counter = (((20 - (prev->nice)) >> 2) + 1);
>                         move_last_runqueue (prev);
>                 }
> 
> (and I don't quite understand this, since both "unlikely" and
> "__builtin_expect" seem to be macros:
> 
> ./include/linux/compiler.h:#define __builtin_expect(x, expected_value) (x)
> ./include/linux/compiler.h:#define likely(x)    __builtin_expect((x),1)
> ./include/linux/compiler.h:#define unlikely(x)  __builtin_expect((x),0)
> )
> 
> Please let me know if you need more info.  I can send you a log with higher
> debug value, or even the vmlinux file.
> 
> -- 
> Robert Millan
> _______________________________________________
> http://www.opencores.org/mailman/listinfo/openrisc