A tester is needed with dual P3 and USB

A tester is needed with dual P3 and USB

[Pete Zaitcev]

Hi, All:

I received a complaint that a UP kernel hangs on boot if USB is
enabled. SMP works. An SMP kernel started with "nosmp" hangs too.
The reporter is, umm, how shall I put it... is a power user.

I need someone to help me to track the problem down, because
I am curious. I heard of SMP hangs before, but a UP hang is
a novel idea.

The box is VA Linux 1000 (similar to IBM Netfinity 4000r).
Kernel is 2.4.8-ac10.

If you love me (or USB subsystem in kernel) enough to help
me out, and you your hardware reproduces the problem,
please reply privately.

Thanks in advance,
-- Pete

Re: A tester is needed with dual P3 and USB

[Pekka Pietikainen]

On Mon, Aug 27, 2001 at 06:22:04PM -0400, Pete Zaitcev wrote:
> Hi, All:
> 
> I received a complaint that a UP kernel hangs on boot if USB is
> enabled. SMP works. An SMP kernel started with "nosmp" hangs too.
> The reporter is, umm, how shall I put it... is a power user.
> I need someone to help me to track the problem down, because
> I am curious. I heard of SMP hangs before, but a UP hang is
> a novel idea.
> 
> The box is VA Linux 1000 (similar to IBM Netfinity 4000r).
> Kernel is 2.4.8-ac10.
Doesn't VA use one of those Intel boards which have the problem
with theis BIOS, which is seen as a hang with the adaptec driver?

Tried the same work-around? (enabling the APIC)

-- 
Pekka Pietikainen

Re: A tester is needed with dual P3 and USB

[Frank Schneider]

Pekka Pietikainen schrieb:
> 
> On Mon, Aug 27, 2001 at 06:22:04PM -0400, Pete Zaitcev wrote:
> > Hi, All:
> >
> > I received a complaint that a UP kernel hangs on boot if USB is
> > enabled. SMP works. An SMP kernel started with "nosmp" hangs too.
> > The reporter is, umm, how shall I put it... is a power user.
> > I need someone to help me to track the problem down, because
> > I am curious. I heard of SMP hangs before, but a UP hang is
> > a novel idea.
> >
> > The box is VA Linux 1000 (similar to IBM Netfinity 4000r).
> > Kernel is 2.4.8-ac10.
> Doesn't VA use one of those Intel boards which have the problem
> with theis BIOS, which is seen as a hang with the adaptec driver?

Thats the problem i can reproduce here (ASUS-P2B-DS with Dual PIII/850,
USB and AHA2940U2W-SCSI onboard, Kernel 2.4.2/3-SMP), SMP-Kernels with
option "nosmp" simply stop when they try to access the SCSI-System, the
aic7xxx-driver loads (no matter if build-in or module), but stucks at
initializing the first device.
No Ooop or crash, it simply does nothing any more.

But a "real" UP-Kernel (i used the one from RH7.1, 2.4.2-UP) has no
problems with booting here.

Solong..
Frank.

--
Frank Schneider, <SPATZ1@T-ONLINE.DE>.                           
... -.-

[PATCH] disableapic option

[Kurt Garloff]

[-- Attachment #1.1: Type: text/plain, Size: 1404 bytes --]

Hi Pekka, Linus, Alan,

On Tue, Aug 28, 2001 at 01:42:11AM +0300, Pekka Pietikainen wrote:
> Doesn't VA use one of those Intel boards which have the problem
> with theis BIOS, which is seen as a hang with the adaptec driver?
> 
> Tried the same work-around? (enabling the APIC)

Here's a patch that allows you to compile the kernel with APIC enabled
always. Some machines need it, but others barf, when they find out that
there is an MP table and an IOAPIC but no APIC on the CPU is found.
(Plug a K6 in a dual Pentium board, e.g.)

Then you may pass the disableapic boot parameter, so you bypass APIC setup
completely and the APIC compiled kernel works like compiled without.
(For people who compile their own kernels, this is irrelevant, except for
 testing. That's not the majority of Linux users anymore, though.)

Another possibility might be to make apic.c:221 fail gracefully instead of
BUG(), BTW.

Linus, Alan, please apply to your trees!

In theory, the compile time option could be removed now, but some people
probably want to be able to save a few bytes.

Regards,
-- 
Kurt Garloff                   <kurt@garloff.de>         [Eindhoven, NL]
Physics: Plasma simulations  <K.Garloff@Phys.TUE.NL>  [TU Eindhoven, NL]
Linux: SCSI, Security          <garloff@suse.de>    [SuSE Nuernberg, DE]
 (See mail header or public key servers for PGP2 and GPG public keys.)

[-- Attachment #1.2: disableapic-249.diff --]
[-- Type: text/plain, Size: 825 bytes --]

--- linux-244.compile/arch/i386/kernel/io_apic.c.orig	Mon Apr 30 17:12:43 2001
+++ linux-244.compile/arch/i386/kernel/io_apic.c	Fri May 11 17:44:37 2001
@@ -182,6 +182,7 @@
 int pirq_entries [MAX_PIRQS];
 int pirqs_enabled;
 int skip_ioapic_setup;
+int disable_apic;
 
 static int __init ioapic_setup(char *str)
 {
@@ -191,6 +192,15 @@
 
 __setup("noapic", ioapic_setup);
 
+static int __init disable_apic_setup(char *str)
+{
+	skip_ioapic_setup = 1;
+	disable_apic = 1;
+	return 1;
+}
+
+__setup("disableapic", disable_apic_setup);
+
 static int __init ioapic_pirq_setup(char *str)
 {
 	int i, max;
@@ -1605,7 +1615,7 @@
  */
 void IO_APIC_init_uniprocessor (void)
 {
-	if (!smp_found_config)
+	if (!smp_found_config || disable_apic)
 		return;
 	connect_bsp_APIC();
 	setup_local_APIC();

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]