[Linux-ia64] 2.5.50+ patch for the impatient...

[Linux-ia64] 2.5.50+ patch for the impatient...

[David Mosberger]

For those of you who're impatiently waiting for a 2.5.xx update, I put
a preliminary patch at:

  ftp://ftp.hpl.hp.com/pub/linux-ia64/patch-2.5.50-1.968-lia64.diff.gz

The patch is relative to revision 1.968 of Linus' bitkeeper tree
(which is 2.5.50 with a bunch of additional patches).  Only the Ski
simulator has been tested.  Kernel modules are definitely broken at
the moment, since the ia64-part of the in-kernel module-loader is
missing (I don't use modules myself and in any case it may be better
to wait until things have settled down a bit; it looks like Rich
Henderson has convinced Rusty to make modules effectively shared
objects).

I hope to have a more complete update by the time Linus gets ready to
relese 2.5.51.  So if you're not urgently waiting for 2.5.50, you
might want to wait till then.

Enjoy,

	--david

RE: [Linux-ia64] 2.5.50+ patch for the impatient...

[Matt_Domsch]

> For those of you who're impatiently waiting for a 2.5.xx update, I put
> a preliminary patch at:

I'm still unable to create include/asm-ia64/offsets.h properly.

include/linux/sched.h has:
union thread_union {
        struct thread_info thread_info;
        unsigned long stack[INIT_THREAD_SIZE/sizeof(long)];
};

(INIT_THREAD_SIZE needs to be number)

include/asm-ia64/thread_info.h has:
#define INIT_THREAD_SIZE                /* tell sched.h not to declare the
thread_union */

(INIT_THREAD_SIZE isn't a number)


Deleting the declaration of INIT_THREAD_SIZE lets it continue and builds
offsets.h.  Is that safe?

Thanks,
Matt

--
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com

Re: [Linux-ia64] 2.5.50+ patch for the impatient...

[Jesse Barnes]

Thanks David.  Does this mean that your 'to-linus' tree will compile?
If so, I could just use that to send out an SN update (one that
compiles, btw), otherwise I'll generate a patch against this version and
hopefully have it to you next week.

Thanks,
Jesse

On Thu, Dec 05, 2002 at 12:13:59AM -0800, David Mosberger wrote:
> For those of you who're impatiently waiting for a 2.5.xx update, I put
> a preliminary patch at:
> 
>   ftp://ftp.hpl.hp.com/pub/linux-ia64/patch-2.5.50-1.968-lia64.diff.gz
> 
> The patch is relative to revision 1.968 of Linus' bitkeeper tree
> (which is 2.5.50 with a bunch of additional patches).  Only the Ski
> simulator has been tested.  Kernel modules are definitely broken at
> the moment, since the ia64-part of the in-kernel module-loader is
> missing (I don't use modules myself and in any case it may be better
> to wait until things have settled down a bit; it looks like Rich
> Henderson has convinced Rusty to make modules effectively shared
> objects).
> 
> I hope to have a more complete update by the time Linus gets ready to
> relese 2.5.51.  So if you're not urgently waiting for 2.5.50, you
> might want to wait till then.
> 
> Enjoy,
> 
> 	--david
> 
> _______________________________________________
> Linux-IA64 mailing list
> Linux-IA64@linuxia64.org
> http://lists.linuxia64.org/lists/listinfo/linux-ia64

RE: [Linux-ia64] 2.5.50+ patch for the impatient...

[David Mosberger]

>>>>> On Thu, 5 Dec 2002 10:19:02 -0600, Matt_Domsch@Dell.com said:

  Matt> I'm still unable to create include/asm-ia64/offsets.h
  Matt> properly.

  Matt> include/linux/sched.h has: union thread_union { struct
  Matt> thread_info thread_info; unsigned long
  Matt> stack[INIT_THREAD_SIZE/sizeof(long)]; };

  Matt> (INIT_THREAD_SIZE needs to be number)

  Matt> include/asm-ia64/thread_info.h has: #define INIT_THREAD_SIZE
  Matt> /* tell sched.h not to declare the thread_union */

  Matt> (INIT_THREAD_SIZE isn't a number)

Something must have gone wrong with applying the ia64 patch.  In sched.h,
it should say:

 #ifndef INIT_THREAD_SIZE

right before the thread_union declaration.

	--david

RE: [Linux-ia64] 2.5.50+ patch for the impatient...

[Matt Domsch]

> Something must have gone wrong with applying the ia64 patch.

Yep.  Fixed here, thanks.

Without CONFIG_SMP, cpu_possible() isn't defined.  I think it needs this:

=== include/asm-ia64/smp.h 1.8 vs edited ==--- 1.8/include/asm-ia64/smp.h	Mon Aug 12 06:06:34 2002
+++ edited/include/asm-ia64/smp.h	Thu Dec  5 16:49:20 2002
@@ -138,6 +138,8 @@
 extern void smp_send_reschedule (int cpu);
 extern void smp_send_reschedule_all (void);
 
+#else /* !CONFIG_SMP */
+#define cpu_possible(cpu)	(!cpu)
 
 #endif /* CONFIG_SMP */
 #endif /* _ASM_IA64_SMP_H */


Thanks,
Matt

-- 
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com

Re: [Linux-ia64] 2.5.50+ patch for the impatient...

[Keith Owens]

On Thu, 5 Dec 2002 00:13:59 -0800, 
David Mosberger <davidm@napali.hpl.hp.com> wrote:
>For those of you who're impatiently waiting for a 2.5.xx update, I put
>a preliminary patch at:
>
>  ftp://ftp.hpl.hp.com/pub/linux-ia64/patch-2.5.50-1.968-lia64.diff.gz
>
>The patch is relative to revision 1.968 of Linus' bitkeeper tree
>(which is 2.5.50 with a bunch of additional patches).  Only the Ski
>simulator has been tested.  Kernel modules are definitely broken at
>the moment, since the ia64-part of the in-kernel module-loader is
>missing (I don't use modules myself and in any case it may be better
>to wait until things have settled down a bit; it looks like Rich
>Henderson has convinced Rusty to make modules effectively shared
>objects).

kdb for 2.5 kernels is also completely broken.  It needs the section
data in kallsyms and Rusty has deleted that information.  See
http://www.kerneltrap.org/node.php?idP0.  Do not expect kdb for any
2.5 kernel until Rusty puts the debugging data back.  And BTW, do not
expect ksymoops to work either.

RE: [Linux-ia64] 2.5.50+ patch for the impatient...

[David Mosberger]

>>>>> On Thu, 5 Dec 2002 16:50:46 -0600 (CST), Matt Domsch <Matt_Domsch@Dell.com> said:

  >> Something must have gone wrong with applying the ia64 patch.
  Matt> Yep.  Fixed here, thanks.

  Matt> Without CONFIG_SMP, cpu_possible() isn't defined.  I think it
  Matt> needs this:

  Matt> === include/asm-ia64/smp.h 1.8 vs edited === ---
  Matt> 1.8/include/asm-ia64/smp.h Mon Aug 12 06:06:34 2002 +++
  Matt> edited/include/asm-ia64/smp.h Thu Dec 5 16:49:20 2002 @@
  Matt> -138,6 +138,8 @@ extern void smp_send_reschedule (int cpu);
  Matt> extern void smp_send_reschedule_all (void);
 
  Matt> +#else /* !CONFIG_SMP */ +#define cpu_possible(cpu) (!cpu)
 
  Matt>  #endif /* CONFIG_SMP */ #endif /* _ASM_IA64_SMP_H */

Actually, I think this needs to be defined in linux/smp.h.  This
appears to be fixed already in Linus's bk tree
(see http://linux.bkbits.net:8080/linux-2.5/diffs/include/linux/smp.h@1.18).

	--david