[Linux-ia64] More on B1 BigSur and test8 kernel

[Linux-ia64] More on B1 BigSur and test8 kernel

[H . J . Lu]

On Sat, Sep 09, 2000 at 05:17:12PM -0700, David Mosberger wrote:
> >>>>> On Sat, 9 Sep 2000 14:49:52 -0700, "H . J . Lu" <hjl@valinux.com> said:
> 
>   HJ> Yes, I did "make dep". What should include/asm-ia64/offsets.h
>   HJ> look like? That may explain many things.
> 
> Well, it depends on your kernel configuration.  If you did a "make
> dep", you should have the right values in there.
> 
> I'm not sure why the new kernel isn't working for you.  Nothing has
> changed in the kernel that would explain an infinite stream of irq 0.
> 

More info on my UP BigSur with B1 stepping and BIOS Build 59:

1.  The kernel from TurboLinux 0828 boots and prints out

ACPI: Intel    RSDT   0.0
Acpi cfg:bind to Boot time Acpi OSD
Acpi cfg:acpi initialize pass
Acpi cfg:acpi load firmware tables pass
Acpi cfg:acpi load namespace pass
      CPU 0 (0000:0000): Available.
IOSAPIC Version 2.1: address 0xfec00000 IRQs 0x0 - 0x3f
Acpi cfg:get pci vectors 
Acpi cfg:_STA not found: pci bus 0 exist 
Acpi cfg:_STA not found: pci bus 1 exist 
Acpi cfg:_STA not found: pci bus 2 exist 
Acpi cfg:_STA not found: pci bus 3 exist 
      1 CPUs available, 1 CPUs total
ACPI:         -0550: <7>ACPI: *** Success: Entire namespace and objects deleted
Acpi cfg:acpi terminate pass
Acpi cfg:bind to Run time Acpi OSD
ia64_mca_init : begin
ia64_mca_init : registered mca rendezvous spinloop and wakeup mech.
ia64_mca_init : correctable mca vector setup done
ia64_mca_init : registered os mca handler with SAL
ia64_mca_init : os init handler at 5414b0
ia64_mca_init : registered os init handler with SAL
ia64_mca_init : platform-specific mca handling setup done
Mca related initialization done
On node 0 totalpages: 64672
zone(0): 64672 pages.
zone(1): 0 pages.
zone(2): 0 pages.
Placing software IO TLB between 0xe000000000100000 - 0xe000000000300000
Kernel command line: root=/dev/sda2 init=/bin/bash
fpswa interface at 3f197010
timer: CPU 0 base freq\x133.344MHz, ITC ratio\x10/2, ITC freqf6.722MHz
Console: colour VGA+ 80x25
Unexpected irq vector 0x0 on CPU 0!
Calibrating delay loop... 545.26 BogoMIPS


2. The test8 kernel + David's ia64 patch + my 2 patches + comment out

Unexpected irq vector 0x0 on CPU 0!

compiled on my BigSur with 0825 RedHat ia64 running TurboLinux 0828
kernel. I get

ACPI: Intel    RSDT   0.0
Acpi cfg:bind to Boot time Acpi OSD
Acpi cfg:acpi initialize pass
Acpi cfg:acpi load firmware tables pass
Acpi cfg:acpi load namespace pass
      CPU 0 (0000:0000): Available.
IOSAPIC Version 2.1: address 0xfec00000 IRQs 0x0 - 0x3f
Acpi cfg:get pci vectors 
Acpi cfg:_STA not found: pci bus 0 exist 
Acpi cfg:_STA not found: pci bus 1 exist 
Acpi cfg:_STA not found: pci bus 2 exist 

If I leave

Unexpected irq vector 0x0 on CPU 0!

in, I get a stream of 

Unexpected irq vector 0x0 on CPU 0!

and machin locks up.

3. The test8 kernel + David's ia64 patch +  my 2 patches + comment out

Unexpected irq vector 0x0 on CPU 0!

cross compiled on ia32 with the 0828 snapshot, I get

ACPI: Intel    RSDT   0.0
Acpi cfg:bind to Boot time Acpi OSD
Acpi cfg:acpi initialize pass
Acpi cfg:acpi load firmware tables pass
ACPI:	-0542 <7>ACPI: *** Warning: Type misatch
swapper[0]: General Exception: ......
.....

and machine locks up.

4. The test8 kernel + David's ia64 patch +  my 2 patches cross compiled
 on ia32 with the 0828 snapshot, I get

ACPI: Intel    RSDT   0.0
Acpi cfg:bind to Boot time Acpi OSD
Acpi cfg:acpi initialize pass
Unexpected irq vector 0x0 on CPU 0!pass
Calibrating delay loop... 545.26 BogoMIPS
.....

Then later I get SCSI timeout since QLA1280 gets IRQ 3.

I have no idea if it is a hardware, compiler or kernel ACPI bug. Is
it possible that

Unexpected irq vector 0x0 on CPU 0!

comes at the wrong time which screws up the ACPI parsing. That may
explain why only TurnboLinux's kernel works on my B1 BigSur.

Thanks.

-- 
H.J. Lu (hjl@gnu.org)

Re: [Linux-ia64] More on B1 BigSur and test8 kernel

[David Mosberger]

>>>>> On Sun, 10 Sep 2000 12:25:24 -0700, "H . J . Lu" <hjl@valinux.com> said:

  HJ> compiled on my BigSur with 0825 RedHat ia64 running TurboLinux
  HJ> 0828 kernel. I get         ^^^^

That is your problem.  The new toolchain is not ready for the
compiler.  I stopped tracking the lastest patches and decided to wait
for a new snapshot.  If you build the kernel with the new toolchain,
you're on your own.

	--david

Re: [Linux-ia64] More on B1 BigSur and test8 kernel

[Uros Prestor]

David Mosberger <davidm@hpl.hp.com> said:

> >>>>> On Sun, 10 Sep 2000 12:25:24 -0700, "H . J . Lu" <hjl@valinux.com> 
said:
> 
>   HJ> compiled on my BigSur with 0825 RedHat ia64 running TurboLinux
>   HJ> 0828 kernel. I get         ^^^^
> 
> That is your problem.  The new toolchain is not ready for the
> compiler.  I stopped tracking the lastest patches and decided to wait
> for a new snapshot.  If you build the kernel with the new toolchain,
> you're on your own.

David,

I was compiling test8 with the May compiler snapshot and I observed
pretty much the same behaviour as HJ.  I didn't try out his patches, though.

Uros

-- 
Uros Prestor
uros@turbolinux.com

Re: [Linux-ia64] More on B1 BigSur and test8 kernel

[David Mosberger]

>>>>> On Sun, 10 Sep 2000 19:58:52 -0000, Uros Prestor <uros@turbolinux.com> said:

  Uros> observed pretty much the same behaviour as HJ.  I didn't try
  Uros> out his patches, though.

I doubt those patches will help.  Have you tried with the ACPI
interpreter turned off?  Perhaps that will create irq problems later
on, but it should at least tell you if it's the source of the irq 0
stream.

	--david

Re: [Linux-ia64] More on B1 BigSur and test8 kernel

[H . J . Lu]

On Sun, Sep 10, 2000 at 12:30:01PM -0700, David Mosberger wrote:
> >>>>> On Sun, 10 Sep 2000 12:25:24 -0700, "H . J . Lu" <hjl@valinux.com> said:
> 
>   HJ> compiled on my BigSur with 0825 RedHat ia64 running TurboLinux
>   HJ> 0828 kernel. I get         ^^^^
> 
> That is your problem.  The new toolchain is not ready for the
> compiler.  I stopped tracking the lastest patches and decided to wait
> for a new snapshot.  If you build the kernel with the new toolchain,
> you're on your own.
> 

I got the similar results with the 20000501 toolchain come with the
0825 RedHat ia64 beta distribution. I don't think it is a purely
compiler issue since the same kernel, which works on A2 stepping,
failed to boot on B1 stepping. I suspect some interaction between
the ACPI stuff and the interrupts. It seems on B1 stepping, the
kernel will get interrupts very early on in the ACPI code.

-- 
H.J. Lu (hjl@gnu.org)