[Qemu-devel] 2.6.21 kernel on emulated/real Malta board

[Qemu-devel] 2.6.21 kernel on emulated/real Malta board

[Aurelien Jarno]

Hi all,

Since I switch to 2.6.21 kernel on my emulated Malta board (QEMU), I
have noticed something strange. The kernel starts to boot up to the
timer calibration, and then it restart the boot again. A small example:

  ...

  Primary data cache 0kB, 4-way, linesize 0 bytes.
  Synthesized TLB refill handler (36 instructions).
  Synthesized TLB load handler fastpath (48 instructions).
  Synthesized TLB store handler fastpath (48 instructions).
  Synthesized TLB modify handler fastpath (47 instructions).
  Enable cache parity protection for MIPS 20KC/25KF CPUs.
  PID hash table entries: 512 (order: 9, 4096 bytes)
  CPU frequency 100.00 MHz
  Using 100.003 MHz high precision timer.
  Linux version 2.6.21.1 (aurel32@i386) (gcc version 4.1.1 ()) #1 Tue May 15 12:22:07 CEST 2007

  LINUX started...
  CPU revision is: 000182a0
  FPU revision is: 000f8200

  ...

I have traced the problem down to the CONFIG_EARLY_PRINTK option. 
Disabling it in the .config file (be aware that running make *config 
will reenable this function), removes the problem.

The problem occurs for kernel 2.6.21 from both linux-mips.org and
kernel.org, on both endianness, and for both 32- and 64-bit kernels.

It would be nice if somebody could tell me if the real board (I don't
own one) suffers from the same problem or not, so that I can look for a
possible bug in QEMU.

Thanks,
Aurelien

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] 2.6.21 kernel on emulated/real Malta board

[Stefan Weil]

Hi,

2.6.21 with patches from OpenWrt works for AR7 with
CONFIG_EARLY_PRINTK and real hardware or QEMU emulation.
Maybe I can compare it to Malta tomorrow.

Perhaps some environment settings are missing for
early printk and result in a crash...

Stefan


Aurelien Jarno schrieb:
> Hi all,
>
> Since I switch to 2.6.21 kernel on my emulated Malta board (QEMU), I
> have noticed something strange. The kernel starts to boot up to the
> timer calibration, and then it restart the boot again. A small example:
>
> ...
>
> I have traced the problem down to the CONFIG_EARLY_PRINTK option.
> Disabling it in the .config file (be aware that running make *config
> will reenable this function), removes the problem.
>
> The problem occurs for kernel 2.6.21 from both linux-mips.org and
> kernel.org, on both endianness, and for both 32- and 64-bit kernels.
>
> It would be nice if somebody could tell me if the real board (I don't
> own one) suffers from the same problem or not, so that I can look for a
> possible bug in QEMU.
>
> Thanks,
> Aurelien
>

[Qemu-devel] Re: 2.6.21 kernel on emulated/real Malta board

[Geert Uytterhoeven]

On Sat, 16 Jun 2007, Aurelien Jarno wrote:
> Since I switch to 2.6.21 kernel on my emulated Malta board (QEMU), I
> have noticed something strange. The kernel starts to boot up to the
> timer calibration, and then it restart the boot again. A small example:
> 
>   ...
> 
>   Primary data cache 0kB, 4-way, linesize 0 bytes.
>   Synthesized TLB refill handler (36 instructions).
>   Synthesized TLB load handler fastpath (48 instructions).
>   Synthesized TLB store handler fastpath (48 instructions).
>   Synthesized TLB modify handler fastpath (47 instructions).
>   Enable cache parity protection for MIPS 20KC/25KF CPUs.
>   PID hash table entries: 512 (order: 9, 4096 bytes)
>   CPU frequency 100.00 MHz
>   Using 100.003 MHz high precision timer.
>   Linux version 2.6.21.1 (aurel32@i386) (gcc version 4.1.1 ()) #1 Tue May 15 12:22:07 CEST 2007
> 
>   LINUX started...
>   CPU revision is: 000182a0
>   FPU revision is: 000f8200
> 
>   ...
> 
> I have traced the problem down to the CONFIG_EARLY_PRINTK option. 
> Disabling it in the .config file (be aware that running make *config 
> will reenable this function), removes the problem.

I guess it's just the printk buffer that's being output again to the new
console, when the console subsystem switches from early console to real
console.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: [Qemu-devel] Re: 2.6.21 kernel on emulated/real Malta board

[Stefan Weil]

I just tried Linux 2.6.21.5 with Malta BE in QEMU.
I get the same console messages like Aurelien but
then Malta continues to boot normally.

So Geert is right: all output from early printk is
output again when the real console is activated.

Stefan

Geert Uytterhoeven schrieb:
> On Sat, 16 Jun 2007, Aurelien Jarno wrote:
>> Since I switch to 2.6.21 kernel on my emulated Malta board (QEMU), I
>> have noticed something strange. The kernel starts to boot up to the
>> timer calibration, and then it restart the boot again. A small example:
>>
>> ...
>>
>> Primary data cache 0kB, 4-way, linesize 0 bytes.
>> Synthesized TLB refill handler (36 instructions).
>> Synthesized TLB load handler fastpath (48 instructions).
>> Synthesized TLB store handler fastpath (48 instructions).
>> Synthesized TLB modify handler fastpath (47 instructions).
>> Enable cache parity protection for MIPS 20KC/25KF CPUs.
>> PID hash table entries: 512 (order: 9, 4096 bytes)
>> CPU frequency 100.00 MHz
>> Using 100.003 MHz high precision timer.
>> Linux version 2.6.21.1 (aurel32@i386) (gcc version 4.1.1 ()) #1 Tue
>> May 15 12:22:07 CEST 2007
>>
>> LINUX started...
>> CPU revision is: 000182a0
>> FPU revision is: 000f8200
>>
>> ...
>>
>> I have traced the problem down to the CONFIG_EARLY_PRINTK option.
>> Disabling it in the .config file (be aware that running make *config
>> will reenable this function), removes the problem.
>
> I guess it's just the printk buffer that's being output again to the new
> console, when the console subsystem switches from early console to real
> console.
>
> Gr{oetje,eeting}s,
>
> Geert

[Qemu-devel] Re: 2.6.21 kernel on emulated/real Malta board

[Ralf Baechle]

On Sun, Jun 17, 2007 at 10:06:55AM +0200, Geert Uytterhoeven wrote:

> I guess it's just the printk buffer that's being output again to the new
> console, when the console subsystem switches from early console to real
> console.

Correct.

  Ralf