On Thursday 24 of June 2004 10:46, Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.7/2.6.7-m
>m2/

Well, I have alarmingly many problems with this patch (my system is a dual
Opteron - full config log attached):

1) There is an Oops at early boot time, caused probably by earlyprintk, which
makes the serial console stop working (the call trace goes too fast to read
and of course it does not go to the serial console ...).

2) There is an Oops when trying to unload the sound driver (snd-intel8x0). At
present I'm unable to grab it. because of 1).

3) This breaks fb console on my system quite a bit:

> +core-fbcon-fixes.patch

- there is an ugly grey background area behind the penguin logos at startup.
The (screen behind the) logos are (is) ok without it. ;-)

There were no such problems on 2.6.7-mm1, and I've already reversed the
bk-acpi.patch. I'm currently trying to figure out which patches should be
reversed to cure 1) and 2) and grab the Oops traces.

The kernel config is attached and I'll send you more info when there's more
info.

Yours,
rjw

----------------------------
For a successful technology, reality must take precedence over public
relations, for nature cannot be fooled.
-- Richard P. Feynman

Attachments:

(No filename) (1.25 kB)
hardware.log (24.85 kB)
2.6.7-mm2-config (27.70 kB)
Download all attachments

2004-06-24 20:59:42

by Adrian Bunk

[permalink] [raw]

Subject: 2.6.7-mm2: compile error SCHED_SMT + NUMA + gcc 2.95

I'm getting the following compile error in 2.6.7-mm2 with SCHED_SMT=y
and NUMA=y when using gcc 2.95 (it doesn't seem to be specific to -mm):

<-- snip -->

...
CC arch/i386/kernel/smpboot.o
arch/i386/kernel/smpboot.c: In function `arch_init_sched_domains':
arch/i386/kernel/smpboot.c:1195: invalid lvalue in unary `&'
arch/i386/kernel/smpboot.c:1231: invalid lvalue in unary `&'
make[1]: *** [arch/i386/kernel/smpboot.o] Error 1

<-- snip -->

cu
Adrian

--

"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed

2004-06-24 21:21:54

by Andrew Morton

[permalink] [raw]

Subject: Re: 2.6.7-mm2: compile error SCHED_SMT + NUMA + gcc 2.95

Adrian Bunk <[email protected]> wrote:
>
> CC arch/i386/kernel/smpboot.o
> arch/i386/kernel/smpboot.c: In function `arch_init_sched_domains':
> arch/i386/kernel/smpboot.c:1195: invalid lvalue in unary `&'
> arch/i386/kernel/smpboot.c:1231: invalid lvalue in unary `&'

--- 25/arch/i386/kernel/smpboot.c~smpboot-build-fix Thu Jun 24 14:22:55 2004
+++ 25-akpm/arch/i386/kernel/smpboot.c Thu Jun 24 14:22:55 2004
@@ -1192,7 +1192,9 @@ __init void arch_init_sched_domains(void
int j;
cpumask_t nodemask;
struct sched_group *node = &sched_group_nodes[i];
- cpus_and(nodemask, node_to_cpumask(i), cpu_possible_map);
+ cpumask_t node_cpumask = node_to_cpumask(i);
+
+ cpus_and(nodemask, node_cpumask, cpu_possible_map);

if (cpus_empty(nodemask))
continue;
@@ -1228,7 +1230,9 @@ __init void arch_init_sched_domains(void
for (i = 0; i < MAX_NUMNODES; i++) {
struct sched_group *cpu = &sched_group_nodes[i];
cpumask_t nodemask;
- cpus_and(nodemask, node_to_cpumask(i), cpu_possible_map);
+ cpumask_t node_cpumask = node_to_cpumask(i);
+
+ cpus_and(nodemask, node_cpumask, cpu_possible_map);

if (cpus_empty(nodemask))
continue;
_

2004-06-24 21:32:57

by Rafał J. Wysocki

[permalink] [raw]

Subject: Re: 2.6.7-mm2

On Thursday 24 of June 2004 22:57, R. J. Wysocki wrote:
> On Thursday 24 of June 2004 10:46, Andrew Morton wrote:
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.7/2.6.7
> >-m m2/
>
> Well, I have alarmingly many problems with this patch (my system is a dual
> Opteron - full config log attached):
>
> 1) There is an Oops at early boot time, caused probably by earlyprintk,
> which makes the serial console stop working (the call trace goes too fast
> to read and of course it does not go to the serial console ...).

I have reversed:

+reduce-tlb-flushing-during-process-migration-2.patch

and it cured this Oops, though it has nothing in common with the console,
obviously.

The serial console just does not work for me at all (earlyprintk apparently
does). Attached is what I get from the earlyprintk. Ther rest of kernel
messages goes to the local console (tty0), although it is not even present in
the kernel command line (sigh).

Yours,
rjw

----------------------------
For a successful technology, reality must take precedence over public
relations, for nature cannot be fooled.
-- Richard P. Feynman

Attachments:

(No filename) (1.12 kB)
2.6.7-mm2-earlyprintk.log (9.19 kB)
Download all attachments

2004-06-24 21:45:35

by Rafał J. Wysocki

[permalink] [raw]

On Mon, 28 Jun 2004, Andrew Morton wrote:

> Peter Lundkvist <[email protected]> wrote:
> >
> > I have problems running cyrus lmtpd; it fails randomly after
> > fetching 30-50 mails. I have only seen this problem in
> > 2.6.7-mm2 (-mm1 was ok, have not tested -mm3 yet).
>
> The "flexible-mmap" patch was rather broken. It was dropped for -mm3 and a
> fixed version will be in -mm4.

Ahh, that could perhaps explain why I was getting strange errors from
ldconfig about being unable to mmap certain libraries when I was using
-mm2 (-mm3 works). I'll be sure to watch out for that behaviour with -mm4

--
Jesper Juhl <[email protected]>