All of lore.kernel.org
 help / color / mirror / Atom feed
* APIC problem report
@ 2009-03-19  8:12 Christopher Schramm
  2009-03-19 10:01 ` Ingo Molnar
  2009-03-21  0:43 ` Robert Hancock
  0 siblings, 2 replies; 7+ messages in thread
From: Christopher Schramm @ 2009-03-19  8:12 UTC (permalink / raw)
  To: linux-kernel

Hi,

I found a problem with APIC on a desktop PC equipped with Asus' M2N, 
running the manufacturer's AMI BIOS - version 0803.

I've tried booting Knoppix 5.3.1 (Kernel 2.6.24.4) and GParted live 
(Debian's 2.6.26-13). Both will only boot if noapic parameter is used.
Otherwise I get "MP-BIOS bug: 8254 timer not connected to IO-APIC" 
resulting in a kernel panic.

apic=debug gives "ACPI: Core revision 20080321", "Version 80050010" 
(twice), "ID: 0", "LVT0: 700", "LVT1:400" and "ESR value before enabling 
vector: 0x00000004  after: 0x00000000".

Hope it's of some use (and the best place to report it)

Christopher

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: APIC problem report
  2009-03-19  8:12 APIC problem report Christopher Schramm
@ 2009-03-19 10:01 ` Ingo Molnar
  2009-03-19 14:30   ` Christopher Schramm
  2009-03-21  0:43 ` Robert Hancock
  1 sibling, 1 reply; 7+ messages in thread
From: Ingo Molnar @ 2009-03-19 10:01 UTC (permalink / raw)
  To: Christopher Schramm, Yinghai Lu, Maciej W. Rozycki,
	Thomas Gleixner, H. Peter Anvin
  Cc: linux-kernel, Andrew Morton


* Christopher Schramm <debian@shakaweb.org> wrote:

> Hi,
>
> I found a problem with APIC on a desktop PC equipped with Asus' 
> M2N, running the manufacturer's AMI BIOS - version 0803.
>
> I've tried booting Knoppix 5.3.1 (Kernel 2.6.24.4) and GParted 
> live (Debian's 2.6.26-13). Both will only boot if noapic parameter 
> is used. Otherwise I get "MP-BIOS bug: 8254 timer not connected to 
> IO-APIC" resulting in a kernel panic.
>
> apic=debug gives "ACPI: Core revision 20080321", "Version 
> 80050010" (twice), "ID: 0", "LVT0: 700", "LVT1:400" and "ESR value 
> before enabling vector: 0x00000004 after: 0x00000000".
>
> Hope it's of some use (and the best place to report it)

Any chance to build and run a more recent kernel on it, to see 
whether this issue still occurs? The latest APIC code is at:

  http://people.redhat.com/mingo/tip.git/README

And i suspect you should pick up the Knoppix .config and build with 
that to reproduce this issue. A digital picture of the panic would 
be useful as well.

	Ingo

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: APIC problem report
  2009-03-19 10:01 ` Ingo Molnar
@ 2009-03-19 14:30   ` Christopher Schramm
  2009-03-19 15:49     ` Ingo Molnar
  0 siblings, 1 reply; 7+ messages in thread
From: Christopher Schramm @ 2009-03-19 14:30 UTC (permalink / raw)
  To: Ingo Molnar
  Cc: Yinghai Lu, Maciej W. Rozycki, Thomas Gleixner, H. Peter Anvin,
	linux-kernel, Andrew Morton

Thanks for the quick replay, Ingo!

Following your request, I installed Debian 5.0. Had to use noapic again 
for getting the boot media to work (think it utilizes 2.6.26).

The kernel panic is the same for all three systems: "not syncing: 
IO-APIC + timer doesn't work!".

The installed system had the same problem (2.6.26 again), then I 
compiled 2.6.28.8 (with default config). Booting that one with APIC 
enabled throws the known "MP-BIOS bug" message, but then continues 
booting - no kernel panic.

I will try Knoppix' kernel config and your git code, when I find some 
spare time.

Christopher

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: APIC problem report
  2009-03-19 14:30   ` Christopher Schramm
@ 2009-03-19 15:49     ` Ingo Molnar
  2009-03-19 17:44       ` Christopher Schramm
  0 siblings, 1 reply; 7+ messages in thread
From: Ingo Molnar @ 2009-03-19 15:49 UTC (permalink / raw)
  To: Christopher Schramm
  Cc: Yinghai Lu, Maciej W. Rozycki, Thomas Gleixner, H. Peter Anvin,
	linux-kernel, Andrew Morton


* Christopher Schramm <debian@shakaweb.org> wrote:

> Thanks for the quick replay, Ingo!
>
> Following your request, I installed Debian 5.0. Had to use noapic 
> again for getting the boot media to work (think it utilizes 
> 2.6.26).
>
> The kernel panic is the same for all three systems: "not syncing: 
> IO-APIC + timer doesn't work!".
>
> The installed system had the same problem (2.6.26 again), then I 
> compiled 2.6.28.8 (with default config). Booting that one with 
> APIC enabled throws the known "MP-BIOS bug" message, but then 
> continues booting - no kernel panic.
>
> I will try Knoppix' kernel config and your git code, when I find 
> some spare time.

By the looks of it, chances are that various APIC fixes and 
robustization patches might have fixed your problem:

 691874f: x86: I/O APIC: timer through 8259A second-chance
 03be750: x86: I/O APIC: keep the timer IRQ masked during set-up
 24742ec: x86: I/O APIC: unmask the second-chance timer interrupt
 d11d579: x86: I/O APIC: AEOI timer acknowledgement clean-ups

These first showed up in v2.6.27. Which would match your reported 
symptoms of v2.6.26 failig and 2.6.28 working. The bootup warning
about a BIOS problem is still there because there is an underlying 
problem - we just work it around and fall back more robustly.

	Ingo

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: APIC problem report
  2009-03-19 15:49     ` Ingo Molnar
@ 2009-03-19 17:44       ` Christopher Schramm
  0 siblings, 0 replies; 7+ messages in thread
From: Christopher Schramm @ 2009-03-19 17:44 UTC (permalink / raw)
  To: Ingo Molnar
  Cc: Yinghai Lu, Maciej W. Rozycki, Thomas Gleixner, H. Peter Anvin,
	linux-kernel, Andrew Morton

Ingo Molnar schrieb:
> By the looks of it, chances are that various APIC fixes and 
> robustization patches might have fixed your problem:
> 
>  691874f: x86: I/O APIC: timer through 8259A second-chance
>  03be750: x86: I/O APIC: keep the timer IRQ masked during set-up
>  24742ec: x86: I/O APIC: unmask the second-chance timer interrupt
>  d11d579: x86: I/O APIC: AEOI timer acknowledgement clean-ups
> 
> These first showed up in v2.6.27. Which would match your reported 
> symptoms of v2.6.26 failig and 2.6.28 working.

Yes, you're right. Looking at the detailed kernel messages, the "8259A 
second-chance" seems to do the trick. So it's already tackled. :-)

Christopher

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: APIC problem report
  2009-03-19  8:12 APIC problem report Christopher Schramm
  2009-03-19 10:01 ` Ingo Molnar
@ 2009-03-21  0:43 ` Robert Hancock
  2009-03-21 10:09   ` Ingo Molnar
  1 sibling, 1 reply; 7+ messages in thread
From: Robert Hancock @ 2009-03-21  0:43 UTC (permalink / raw)
  To: Christopher Schramm; +Cc: linux-kernel, mingo

Christopher Schramm wrote:
> Hi,
> 
> I found a problem with APIC on a desktop PC equipped with Asus' M2N, 
> running the manufacturer's AMI BIOS - version 0803.
> 
> I've tried booting Knoppix 5.3.1 (Kernel 2.6.24.4) and GParted live 
> (Debian's 2.6.26-13). Both will only boot if noapic parameter is used.
> Otherwise I get "MP-BIOS bug: 8254 timer not connected to IO-APIC" 
> resulting in a kernel panic.
> 
> apic=debug gives "ACPI: Core revision 20080321", "Version 80050010" 
> (twice), "ID: 0", "LVT0: 700", "LVT1:400" and "ESR value before enabling 
> vector: 0x00000004  after: 0x00000000".
> 
> Hope it's of some use (and the best place to report it)
> 
> Christopher

Try booting with acpi_use_timer_override kernel parameter. The problem 
is likely the same one that shows up on my Asus A8N-SLI Deluxe with the 
latest 1805 BIOS. The kernel ignores ACPI timer overrides on NVIDIA 
chipsets since they have historically been broken, however in later BIOS 
versions this has been fixed and the override is now apparently required 
for the 8254 timer to function properly through the IO-APIC.

That problem is still there in newer kernels, but it's not fatal since 
the kernel just routes the timer through the PIC instead..

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: APIC problem report
  2009-03-21  0:43 ` Robert Hancock
@ 2009-03-21 10:09   ` Ingo Molnar
  0 siblings, 0 replies; 7+ messages in thread
From: Ingo Molnar @ 2009-03-21 10:09 UTC (permalink / raw)
  To: Robert Hancock, Maciej W. Rozycki; +Cc: Christopher Schramm, linux-kernel


* Robert Hancock <hancockrwd@gmail.com> wrote:

> Christopher Schramm wrote:
>> Hi,
>>
>> I found a problem with APIC on a desktop PC equipped with Asus' M2N,  
>> running the manufacturer's AMI BIOS - version 0803.
>>
>> I've tried booting Knoppix 5.3.1 (Kernel 2.6.24.4) and GParted live  
>> (Debian's 2.6.26-13). Both will only boot if noapic parameter is used.
>> Otherwise I get "MP-BIOS bug: 8254 timer not connected to IO-APIC"  
>> resulting in a kernel panic.
>>
>> apic=debug gives "ACPI: Core revision 20080321", "Version 80050010"  
>> (twice), "ID: 0", "LVT0: 700", "LVT1:400" and "ESR value before 
>> enabling vector: 0x00000004  after: 0x00000000".
>>
>> Hope it's of some use (and the best place to report it)
>>
>> Christopher
>
> Try booting with acpi_use_timer_override kernel parameter. The 
> problem is likely the same one that shows up on my Asus A8N-SLI 
> Deluxe with the latest 1805 BIOS. The kernel ignores ACPI timer 
> overrides on NVIDIA chipsets since they have historically been 
> broken, however in later BIOS versions this has been fixed and the 
> override is now apparently required for the 8254 timer to function 
> properly through the IO-APIC.
>
> That problem is still there in newer kernels, but it's not fatal 
> since the kernel just routes the timer through the PIC instead..

Correct. Warning messages in this area are OK as long as the system 
ends up booting and working fine. What we would be interested in 
fixing is latest kernels not booting out of box at all - i.e. 
requiring noapic parameter or other quirks.

	Ingo

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2009-03-21 10:09 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-03-19  8:12 APIC problem report Christopher Schramm
2009-03-19 10:01 ` Ingo Molnar
2009-03-19 14:30   ` Christopher Schramm
2009-03-19 15:49     ` Ingo Molnar
2009-03-19 17:44       ` Christopher Schramm
2009-03-21  0:43 ` Robert Hancock
2009-03-21 10:09   ` Ingo Molnar

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.