Re: [Xenomai-core] kernel panic

Re: [Xenomai-core] kernel panic

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1850 bytes --]

Wael BACHTA wrote:
> Hi,
> 
> I use an Intel Pentium 4 processor which can be used as 2 virtual
> processors when setting the SMP option during the kernel compiling (I can
> see 2 processors with "cat /proc/cpuinfo").

Hyperthreading? No good idea for RT if you are after low latencies. This
virtualisation of a second CPU is incomplete and can cause congestions
so that you stall on low-prio work at arbitrary points in your code. I
would suggest to measure the effect to be sure if it is acceptable
(latency test under high non-RT load like kernel builds, with SMP vs. !SMP).

> 
> I patched the kernel with adeos in order to install xenomai. I discovered
> that :
> - when completely disabling the ACPI and setting the SMP, my system boots
> but I have only one processor when I look into /proc/cpuinfo
> - Everything goes well when I don't disable completely the ACPI.
> 
> I also use a PCI framegrabber. its driver is a real time one developed

[If it is for common hardware, publishing it would be nice. I guess that
there are others interested in such support.]

> under xenomai. I noticed that this driver works fine only if the PCI_MSI
> is set. Otherwise, I don't have any interrupts.
> 
> I must use my real time driver and the 2 virtual processors.
> But, when I enable the ACPI, the SMP and the PCI_MSI, I have a kernel
> panic. My PC boots when I disable the ACPI but I loose 1 processor.

What about ACPI on (/wo ACPI_PROCESSOR) and SMP off?

> 
> I don't see what's wrong.

Likely Adeos/I-pipe, as the amount of MSI-related bug reports also
indicates. Maybe this workaround already helps you:

https://mail.gna.org/public/adeos-main/2006-09/msg00032.html

If not, try to catch the oops (e.g. via serial or netconsole) for
further debugging. In any case, please post your findings.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

[Xenomai-core] kernel panic

[Wael BACHTA]

Hi,

I use an Intel Pentium 4 processor which can be used as 2 virtual
processors when setting the SMP option during the kernel compiling (I can
see 2 processors with "cat /proc/cpuinfo").

I patched the kernel with adeos in order to install xenomai. I discovered
that :
- when completely disabling the ACPI and setting the SMP, my system boots
but I have only one processor when I look into /proc/cpuinfo
- Everything goes well when I don't disable completely the ACPI.

I also use a PCI framegrabber. its driver is a real time one developed
under xenomai. I noticed that this driver works fine only if the PCI_MSI
is set. Otherwise, I don't have any interrupts.

I must use my real time driver and the 2 virtual processors.
But, when I enable the ACPI, the SMP and the PCI_MSI, I have a kernel
panic. My PC boots when I disable the ACPI but I loose 1 processor.

I don't see what's wrong.

Thanks for your help,

Wael.

-------------------------------------
Wael BACHTA
Strasbourg I University
LSIIT Laboratory, AVR group
Bd S. Brant
BP 10413, 67412 ILLKIRCH cedex, FRANCE
Tel : +33 (0)3 90 24 44 78
Fax : +33 (0)3 90 24 44 80
http://eavr.u-strasbg.fr

Re: [Xenomai-core] kernel panic

[Philippe Gerum]

On Fri, 2006-10-06 at 18:56 +0000, Wael BACHTA wrote:
> Hi,
> 
> I use an Intel Pentium 4 processor which can be used as 2 virtual
> processors when setting the SMP option during the kernel compiling (I can
> see 2 processors with "cat /proc/cpuinfo").
> 
> I patched the kernel with adeos in order to install xenomai. I discovered
> that :
> - when completely disabling the ACPI and setting the SMP, my system boots
> but I have only one processor when I look into /proc/cpuinfo
> - Everything goes well when I don't disable completely the ACPI.
> 
> I also use a PCI framegrabber. its driver is a real time one developed
> under xenomai. I noticed that this driver works fine only if the PCI_MSI
> is set. Otherwise, I don't have any interrupts.
> 
> I must use my real time driver and the 2 virtual processors.
> But, when I enable the ACPI, the SMP and the PCI_MSI, I have a kernel
> panic. My PC boots when I disable the ACPI but I loose 1 processor.
> 
> I don't see what's wrong.
> 

Try this patch:
http://download.gna.org/adeos/patches/v2.6/i386/adeos-ipipe-2.6.17-i386-1.5-00.patch


> Thanks for your help,
> 
> Wael.
> 
> -------------------------------------
> Wael BACHTA
> Strasbourg I University
> LSIIT Laboratory, AVR group
> Bd S. Brant
> BP 10413, 67412 ILLKIRCH cedex, FRANCE
> Tel : +33 (0)3 90 24 44 78
> Fax : +33 (0)3 90 24 44 80
> http://eavr.u-strasbg.fr
> 
> 
> _______________________________________________
> Xenomai-core mailing list
> Xenomai-core@domain.hid
> https://mail.gna.org/listinfo/xenomai-core
-- 
Philippe.

Re: [Xenomai-core] kernel panic

[Philippe Gerum]

On Thu, 2006-10-19 at 16:48 +0000, Wael BACHTA wrote:
> Hi,
> 
> Thank you for your hints.
> 
> I patched my kernel the new adeos patch
> (adeos-ipipe-2.6.17-i386-1.5-00.patch). and I did this tests :
> 
> - SMP on and PCI_MSI on : my PC boots. I installed xenomai 2.2.0. I can see
> my 2 virtual processors with "cat /proc/cpuinfo" BUT, I don't have
> interrupts.
> 
> - SMP on and PCI_MSI off : The PC boots and I see my 2 virtual processors.
> 
> - SMP off and PCI_MSI on : The PC boots, I have my interrupts.
> 
> It seems to me that the new patch disables the PCI_MSI when the SMP is on.
> The new patch doesn't seem to fix my problem.
> 

Clearly not, it doesn't disable anything (in case it's not obvious, I'm
looking for a real fix, not for a work-around). I managed to put my
hands on a previously failing SMP box with PCI_MSI enabled, and 1.5-00
does fix the issue there. So the current issue we are discussing must be
different.

Could you redo test #1 disabling Xenomai, and dump /proc/interrupts on
your box? TIA,

> Thanks for your help.
> 
> Wael.
> 
> 
> 
> > On Fri, 2006-10-06 at 18:56 +0000, Wael BACHTA wrote:
> >> Hi,
> >>
> >> I use an Intel Pentium 4 processor which can be used as 2 virtual
> >> processors when setting the SMP option during the kernel compiling (I
> >> can
> >> see 2 processors with "cat /proc/cpuinfo").
> >>
> >> I patched the kernel with adeos in order to install xenomai. I
> >> discovered
> >> that :
> >> - when completely disabling the ACPI and setting the SMP, my system
> >> boots
> >> but I have only one processor when I look into /proc/cpuinfo
> >> - Everything goes well when I don't disable completely the ACPI.
> >>
> >> I also use a PCI framegrabber. its driver is a real time one developed
> >> under xenomai. I noticed that this driver works fine only if the PCI_MSI
> >> is set. Otherwise, I don't have any interrupts.
> >>
> >> I must use my real time driver and the 2 virtual processors.
> >> But, when I enable the ACPI, the SMP and the PCI_MSI, I have a kernel
> >> panic. My PC boots when I disable the ACPI but I loose 1 processor.
> >>
> >> I don't see what's wrong.
> >>
> >
> > Try this patch:
> > http://download.gna.org/adeos/patches/v2.6/i386/adeos-ipipe-2.6.17-i386-1.5-00.patch
> >
> >
> >> Thanks for your help,
> >>
> >> Wael.
> >>
> >> -------------------------------------
> >> Wael BACHTA
> >> Strasbourg I University
> >> LSIIT Laboratory, AVR group
> >> Bd S. Brant
> >> BP 10413, 67412 ILLKIRCH cedex, FRANCE
> >> Tel : +33 (0)3 90 24 44 78
> >> Fax : +33 (0)3 90 24 44 80
> >> http://eavr.u-strasbg.fr
> >>
> >>
> >> _______________________________________________
> >> Xenomai-core mailing list
> >> Xenomai-core@domain.hid
> >> https://mail.gna.org/listinfo/xenomai-core
> > --
> > Philippe.
> >
> >
> -------------------------------------
> Wael BACHTA
> Strasbourg I University
> LSIIT Laboratory, AVR group
> Bd S. Brant
> BP 10413, 67412 ILLKIRCH cedex, FRANCE
> Tel : +33 (0)3 90 24 44 78
> Fax : +33 (0)3 90 24 44 80
> http://eavr.u-strasbg.fr
> 
-- 
Philippe.

Re: [Xenomai-core] kernel panic

[Wael BACHTA]

Hi,

Thank you for your hints.

I patched my kernel the new adeos patch
(adeos-ipipe-2.6.17-i386-1.5-00.patch). and I did this tests :

- SMP on and PCI_MSI on : my PC boots. I installed xenomai 2.2.0. I can see
my 2 virtual processors with "cat /proc/cpuinfo" BUT, I don't have
interrupts.

- SMP on and PCI_MSI off : The PC boots and I see my 2 virtual processors.

- SMP off and PCI_MSI on : The PC boots, I have my interrupts.

It seems to me that the new patch disables the PCI_MSI when the SMP is on.
The new patch doesn't seem to fix my problem.

Thanks for your help.

Wael.



> On Fri, 2006-10-06 at 18:56 +0000, Wael BACHTA wrote:
>> Hi,
>>
>> I use an Intel Pentium 4 processor which can be used as 2 virtual
>> processors when setting the SMP option during the kernel compiling (I
>> can
>> see 2 processors with "cat /proc/cpuinfo").
>>
>> I patched the kernel with adeos in order to install xenomai. I
>> discovered
>> that :
>> - when completely disabling the ACPI and setting the SMP, my system
>> boots
>> but I have only one processor when I look into /proc/cpuinfo
>> - Everything goes well when I don't disable completely the ACPI.
>>
>> I also use a PCI framegrabber. its driver is a real time one developed
>> under xenomai. I noticed that this driver works fine only if the PCI_MSI
>> is set. Otherwise, I don't have any interrupts.
>>
>> I must use my real time driver and the 2 virtual processors.
>> But, when I enable the ACPI, the SMP and the PCI_MSI, I have a kernel
>> panic. My PC boots when I disable the ACPI but I loose 1 processor.
>>
>> I don't see what's wrong.
>>
>
> Try this patch:
> http://download.gna.org/adeos/patches/v2.6/i386/adeos-ipipe-2.6.17-i386-1.5-00.patch
>
>
>> Thanks for your help,
>>
>> Wael.
>>
>> -------------------------------------
>> Wael BACHTA
>> Strasbourg I University
>> LSIIT Laboratory, AVR group
>> Bd S. Brant
>> BP 10413, 67412 ILLKIRCH cedex, FRANCE
>> Tel : +33 (0)3 90 24 44 78
>> Fax : +33 (0)3 90 24 44 80
>> http://eavr.u-strasbg.fr
>>
>>
>> _______________________________________________
>> Xenomai-core mailing list
>> Xenomai-core@domain.hid
>> https://mail.gna.org/listinfo/xenomai-core
> --
> Philippe.
>
>
-------------------------------------
Wael BACHTA
Strasbourg I University
LSIIT Laboratory, AVR group
Bd S. Brant
BP 10413, 67412 ILLKIRCH cedex, FRANCE
Tel : +33 (0)3 90 24 44 78
Fax : +33 (0)3 90 24 44 80
http://eavr.u-strasbg.fr

Re: [Xenomai-core] kernel panic

[Wael BACHTA]

I did the following tests :

- patch 1.5, SMP on, PCI_MSI on, xenomai off : the /proc/interrupts is the
following :
           CPU0       CPU1
  0:      74130          0    IO-APIC-edge  timer
  1:         67          0    IO-APIC-edge  i8042
  7:          0          0    IO-APIC-edge  parport0
  8:          1          0    IO-APIC-edge  rtc
  9:          0          0   IO-APIC-level  acpi
 12:       3542          0    IO-APIC-edge  i8042
 14:        426          0    IO-APIC-edge  ide0
 82:       6454          0   IO-APIC-level  libata
NMI:          0          0
LOC:      73930      73929
ERR:          0
MIS:          0

- patch 1.5, SMP on, PCI_MSI on, xenomai on : the /proc/interrupts is the
following :
           CPU0       CPU1
  0:     149096          0    IO-APIC-edge  timer, rthal_broadcast_timer
  1:        261          0    IO-APIC-edge  i8042
  7:          0          0    IO-APIC-edge  parport0
  8:          1          0    IO-APIC-edge  rtc
  9:          0          0   IO-APIC-level  acpi
 12:       1754          0    IO-APIC-edge  i8042
 14:       3684          0    IO-APIC-edge  ide0
 82:      10624          0   IO-APIC-level  libata
NMI:          0          0
LOC:     148927     148927
ERR:          0
MIS:          0

- patch 1.3-07, SMP off, PCI_MSI on, xenomai on (The config for which my
driver works well)
           CPU0
  0:     445409    IO-APIC-edge  timer, rthal_broadcast_timer
  1:        437    IO-APIC-edge  i8042
  2:          0          XT-PIC  cascade
  8:          1    IO-APIC-edge  rtc
 12:       6694    IO-APIC-edge  i8042
 14:       4746    IO-APIC-edge  ide0
193:      12218   IO-APIC-level  libata
NMI:          0
LOC:     445321
ERR:          0
MIS:          0









> On Thu, 2006-10-19 at 16:48 +0000, Wael BACHTA wrote:
>> Hi,
>>
>> Thank you for your hints.
>>
>> I patched my kernel the new adeos patch
>> (adeos-ipipe-2.6.17-i386-1.5-00.patch). and I did this tests :
>>
>> - SMP on and PCI_MSI on : my PC boots. I installed xenomai 2.2.0. I can
>> see
>> my 2 virtual processors with "cat /proc/cpuinfo" BUT, I don't have
>> interrupts.
>>
>> - SMP on and PCI_MSI off : The PC boots and I see my 2 virtual
>> processors.
>>
>> - SMP off and PCI_MSI on : The PC boots, I have my interrupts.
>>
>> It seems to me that the new patch disables the PCI_MSI when the SMP is
>> on.
>> The new patch doesn't seem to fix my problem.
>>
>
> Clearly not, it doesn't disable anything (in case it's not obvious, I'm
> looking for a real fix, not for a work-around). I managed to put my
> hands on a previously failing SMP box with PCI_MSI enabled, and 1.5-00
> does fix the issue there. So the current issue we are discussing must be
> different.
>
> Could you redo test #1 disabling Xenomai, and dump /proc/interrupts on
> your box? TIA,
>
>> Thanks for your help.
>>
>> Wael.
>>
>>
>>
>> > On Fri, 2006-10-06 at 18:56 +0000, Wael BACHTA wrote:
>> >> Hi,
>> >>
>> >> I use an Intel Pentium 4 processor which can be used as 2 virtual
>> >> processors when setting the SMP option during the kernel compiling (I
>> >> can
>> >> see 2 processors with "cat /proc/cpuinfo").
>> >>
>> >> I patched the kernel with adeos in order to install xenomai. I
>> >> discovered
>> >> that :
>> >> - when completely disabling the ACPI and setting the SMP, my system
>> >> boots
>> >> but I have only one processor when I look into /proc/cpuinfo
>> >> - Everything goes well when I don't disable completely the ACPI.
>> >>
>> >> I also use a PCI framegrabber. its driver is a real time one
>> developed
>> >> under xenomai. I noticed that this driver works fine only if the
>> PCI_MSI
>> >> is set. Otherwise, I don't have any interrupts.
>> >>
>> >> I must use my real time driver and the 2 virtual processors.
>> >> But, when I enable the ACPI, the SMP and the PCI_MSI, I have a kernel
>> >> panic. My PC boots when I disable the ACPI but I loose 1 processor.
>> >>
>> >> I don't see what's wrong.
>> >>
>> >
>> > Try this patch:
>> > http://download.gna.org/adeos/patches/v2.6/i386/adeos-ipipe-2.6.17-i386-1.5-00.patch
>> >
>> >
>> >> Thanks for your help,
>> >>
>> >> Wael.
>> >>
>> >> -------------------------------------
>> >> Wael BACHTA
>> >> Strasbourg I University
>> >> LSIIT Laboratory, AVR group
>> >> Bd S. Brant
>> >> BP 10413, 67412 ILLKIRCH cedex, FRANCE
>> >> Tel : +33 (0)3 90 24 44 78
>> >> Fax : +33 (0)3 90 24 44 80
>> >> http://eavr.u-strasbg.fr
>> >>
>> >>
>> >> _______________________________________________
>> >> Xenomai-core mailing list
>> >> Xenomai-core@domain.hid
>> >> https://mail.gna.org/listinfo/xenomai-core
>> > --
>> > Philippe.
>> >
>> >
>> -------------------------------------
>> Wael BACHTA
>> Strasbourg I University
>> LSIIT Laboratory, AVR group
>> Bd S. Brant
>> BP 10413, 67412 ILLKIRCH cedex, FRANCE
>> Tel : +33 (0)3 90 24 44 78
>> Fax : +33 (0)3 90 24 44 80
>> http://eavr.u-strasbg.fr
>>
> --
> Philippe.
>
>