Hello,
I got this machine working booting it with "irqpoll" parameter however
it drops the performance a lot and I'd like to try to fix the root
problem of it.
Without 'irqpoll' the disk and network adaptor doesn't work.
Thanks in advance for any help,
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br
On Thu, 19 Nov 2009 19:11:49 -0200
Otavio Salvador <[email protected]> wrote:
> Hello,
>
> I got this machine working booting it with "irqpoll" parameter however
> it drops the performance a lot and I'd like to try to fix the root
> problem of it.
>
> Without 'irqpoll' the disk and network adaptor doesn't work.
That implies the IRQ routing is failing which usually but not always
means the tables in the firmware are wrong. The first step would probably
be to dump all the relevant PCI register space and the ACPI tables and
compare them with the IRQ routing documentation for the chipset.
Actually no - thats step 2, step 1 is asking the vendor if there is
updated firmware to fix IRQ delivery problems
On Thu, 19 Nov 2009 19:11:49 -0200
Otavio Salvador <[email protected]> wrote:
>
> I got this machine working booting it with "irqpoll" parameter however
> it drops the performance a lot and I'd like to try to fix the root
> problem of it.
>
> Without 'irqpoll' the disk and network adaptor doesn't work.
>
> ...
>
> [ 0.000000] ACPI Error: A valid RSDP was not found (20090903/tbxfroot-219)
No ACPI.
Please raise a report against [email protected] at
bugzilla.kernel.org, thanks.
On 11/23/2009 03:40 PM, Andrew Morton wrote:
> On Thu, 19 Nov 2009 19:11:49 -0200
> Otavio Salvador<[email protected]> wrote:
>
>>
>> I got this machine working booting it with "irqpoll" parameter however
>> it drops the performance a lot and I'd like to try to fix the root
>> problem of it.
>>
>> Without 'irqpoll' the disk and network adaptor doesn't work.
>>
>> ...
>>
>> [ 0.000000] ACPI Error: A valid RSDP was not found (20090903/tbxfroot-219)
>
> No ACPI.
>
> Please raise a report against [email protected] at
> bugzilla.kernel.org, thanks.
>
I believe that just means the machine doesn't support ACPI at all..
Hello,
On Mon, Nov 23, 2009 at 9:31 PM, Robert Hancock <[email protected]> wrote:
> On 11/23/2009 03:40 PM, Andrew Morton wrote:
>>> [ ? ?0.000000] ACPI Error: A valid RSDP was not found
>>> (20090903/tbxfroot-219)
>>
>> No ACPI.
>>
>> Please raise a report against [email protected] at
>> bugzilla.kernel.org, thanks.
>
> I believe that just means the machine doesn't support ACPI at all..
Yes; same as me.
Andrew should I report it anyway?
>From Alan message I understand that he believes it is a firmware
issue. I tried (without success) to contact DM&P about it asking for a
newer BIOS but nothing yet.
If I could workaround it, it would be nice since it could avoid issues
for users of those devices. I did research on Internet and found a lot
of users having the network adaptor not work and like and those
problems mostly looks like same issue I'm having.
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br
On 11/23/2009 06:01 PM, Otavio Salvador wrote:
> Hello,
>
> On Mon, Nov 23, 2009 at 9:31 PM, Robert Hancock<[email protected]> wrote:
>> On 11/23/2009 03:40 PM, Andrew Morton wrote:
>>>> [ 0.000000] ACPI Error: A valid RSDP was not found
>>>> (20090903/tbxfroot-219)
>>>
>>> No ACPI.
>>>
>>> Please raise a report against [email protected] at
>>> bugzilla.kernel.org, thanks.
>>
>> I believe that just means the machine doesn't support ACPI at all..
>
> Yes; same as me.
>
> Andrew should I report it anyway?
>
> From Alan message I understand that he believes it is a firmware
> issue. I tried (without success) to contact DM&P about it asking for a
> newer BIOS but nothing yet.
>
> If I could workaround it, it would be nice since it could avoid issues
> for users of those devices. I did research on Internet and found a lot
> of users having the network adaptor not work and like and those
> problems mostly looks like same issue I'm having.
Looking at the BIOS manual on their web site:
ftp://[email protected]/vortex86sx/Vortex86SX_BIOS_Manual.pdf
you can try toggling a few of the options: Try switching OnBoard IDE
Operate Mode between Legacy Mode and Native Mode. If the kernel and the
actual device disagree here, that could cause IRQ delivery problems. You
can also try playing with the "Plug & Play OS" setting.
> >>> [ ? ?0.000000] ACPI Error: A valid RSDP was not found
Are there any BIOS SETUP options for enabling/disabling ACPI?
Does any other version of Linux, or any other OS find ACPI support?
does "acpidump" run as root output anything?
I couldn't follow your link to the BIOS manual, but if there
are some docs on this system that could tell us what the
hardware is and if it is supposed to suport ACPI or not,
that would be a clue.
thanks,
-Len Brown, Intel Open Source Technology Center
On Donnerstag, 19. November 2009, Otavio Salvador wrote:
> I got this machine working booting it with "irqpoll" parameter however
> it drops the performance a lot and I'd like to try to fix the root
> problem of it.
>
> Without 'irqpoll' the disk and network adaptor doesn't work.
What kernel revision do you are using?
We are using a Vortex86DX here and it works as expected. But I do not know the
differences in the SoCs of Vortex86SX and Vortex86DX (okay, SX lacks the
floating point unit). Find our dmesg output attached. Maybe it could help.
jbe
--
Pengutronix e.K. | Juergen Beisert |
Linux Solutions for Science and Industry | Phone: +49-8766-939 228 |
Vertretung Sued/Muenchen, Germany | Fax: +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686 | http://www.pengutronix.de/ |
Hello Robert,
On Mon, Nov 23, 2009 at 11:29 PM, Robert Hancock <[email protected]> wrote:
> Looking at the BIOS manual on their web site:
>
> ftp://[email protected]/vortex86sx/Vortex86SX_BIOS_Manual.pdf
>
> you can try toggling a few of the options: Try switching OnBoard IDE Operate
> Mode between Legacy Mode and Native Mode. If the kernel and the actual
> device disagree here, that could cause IRQ delivery problems. You can also
> try playing with the "Plug & Play OS" setting.
The machine with me has no "OnBoard IDE Operate Mode" option. "Plug &
Play OS" has been tried with "Yes" and "No".
If the kernel and the actual device disagree, could it also "break"
the ethernet adaptor? This also doesn't work.
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br
Hello Juergen,
On Tue, Nov 24, 2009 at 7:32 AM, Juergen Beisert <[email protected]> wrote:
> On Donnerstag, 19. November 2009, Otavio Salvador wrote:
>> I got this machine working booting it with "irqpoll" parameter however
>> it drops the performance a lot and I'd like to try to fix the root
>> problem of it.
>>
>> Without 'irqpoll' the disk and network adaptor doesn't work.
>
> What kernel revision do you are using?
I tested it with 2.6.31 and 2.6.32-rc7.
> We are using a Vortex86DX here and it works as expected. But I do not know the
> differences in the SoCs of Vortex86SX and Vortex86DX (okay, SX lacks the
> floating point unit). Find our dmesg output attached. Maybe it could help.
I'll take a look on the dmesg output and see if I see something interesting :-)
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br
Hello Juergen,
On Tue, Nov 24, 2009 at 9:31 AM, Otavio Salvador
<[email protected]> wrote:
> I'll take a look on the dmesg output and see if I see something interesting :-)
In fact I did find some interesting issues. First:
PCI: Probing PCI hardware
pci 0000:00:07.0: default IRQ router [17f3:6031]
pci 0000:00:0a.0: BAR 0: address space collision on of device
[0x0e2000-0x0e2fff]
pci 0000:00:0a.0: BAR 0: can't allocate resource
pci 0000:00:0a.1: BAR 0: address space collision on of device
[0x0e3000-0x0e30ff]
pci 0000:00:0a.1: BAR 0: can't allocate resource
pci 0000:00:0b.0: BAR 0: address space collision on of device
[0x0e0000-0x0e0fff]
pci 0000:00:0b.0: BAR 0: can't allocate resource
pci 0000:00:0b.1: BAR 0: address space collision on of device
[0x0e1000-0x0e10ff]
pci 0000:00:0b.1: BAR 0: can't allocate resource
NET: Registered protocol family 2
This is the same collision I have. Another interesting issue is that
your disk is using the USB bus while mine uses IT821x.
Is your network working? mine does not (inside of Linux but does for
tftp booting).
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br
On Dienstag, 24. November 2009, Otavio Salvador wrote:
> Hello Juergen,
>
> On Tue, Nov 24, 2009 at 9:31 AM, Otavio Salvador
>
> <[email protected]> wrote:
> > I'll take a look on the dmesg output and see if I see something
> > interesting :-)
>
> In fact I did find some interesting issues. First:
>
> PCI: Probing PCI hardware
> pci 0000:00:07.0: default IRQ router [17f3:6031]
> pci 0000:00:0a.0: BAR 0: address space collision on of device
> [0x0e2000-0x0e2fff]
> pci 0000:00:0a.0: BAR 0: can't allocate resource
> pci 0000:00:0a.1: BAR 0: address space collision on of device
> [0x0e3000-0x0e30ff]
> pci 0000:00:0a.1: BAR 0: can't allocate resource
> pci 0000:00:0b.0: BAR 0: address space collision on of device
> [0x0e0000-0x0e0fff]
> pci 0000:00:0b.0: BAR 0: can't allocate resource
> pci 0000:00:0b.1: BAR 0: address space collision on of device
> [0x0e1000-0x0e10ff]
> pci 0000:00:0b.1: BAR 0: can't allocate resource
> NET: Registered protocol family 2
>
> This is the same collision I have. Another interesting issue is that
> your disk is using the USB bus while mine uses IT821x.
Yes, our system starts from a fixed soldered USB based disk.
> Is your network working? mine does not (inside of Linux but does for
> tftp booting).
Based on a "RDC R6040 Fast Ethernet Adapter" and the kernel states:
[...]
r6040: RDC R6040 NAPI net driver,version 0.24 (08Jul2009)
[...]
Runs with NFS and without any 'irqpoll' trick. (here the modul:
http://www.ssv-comm.de/en/products/esom2586.php)
jbe
--
Pengutronix e.K. | Juergen Beisert |
Linux Solutions for Science and Industry | Phone: +49-8766-939 228 |
Vertretung Sued/Muenchen, Germany | Fax: +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686 | http://www.pengutronix.de/ |
Otavio Salvador wrote:
> On Tue, Nov 24, 2009 at 9:31 AM, Otavio Salvador
>
> <[email protected]> wrote:
> > I'll take a look on the dmesg output and see if I see something
> > interesting :-)
>
> In fact I did find some interesting issues. First:
> [...]
Our kernel states:
CPU: Vortex86 SoC 05/02 stepping 02
Your kernel states:
CPU: Cyrix Cx486SLC
BTW: Try to enable the DEBUG macro in "arch/x86/include/asm/pci_x86.h" and
boot again this kernel. On earlier kernels this increases the verbosity of
the kernel when it tries to detect the IRQ routing.
jbe
--
Pengutronix e.K. | Juergen Beisert |
Linux Solutions for Science and Industry | Phone: +49-8766-939 228 |
Vertretung Sued/Muenchen, Germany | Fax: +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686 | http://www.pengutronix.de/ |
On 11/24/2009 05:29 AM, Otavio Salvador wrote:
> Hello Robert,
>
> On Mon, Nov 23, 2009 at 11:29 PM, Robert Hancock<[email protected]> wrote:
>> Looking at the BIOS manual on their web site:
>>
>> ftp://[email protected]/vortex86sx/Vortex86SX_BIOS_Manual.pdf
>>
>> you can try toggling a few of the options: Try switching OnBoard IDE Operate
>> Mode between Legacy Mode and Native Mode. If the kernel and the actual
>> device disagree here, that could cause IRQ delivery problems. You can also
>> try playing with the "Plug& Play OS" setting.
>
> The machine with me has no "OnBoard IDE Operate Mode" option. "Plug&
> Play OS" has been tried with "Yes" and "No".
>
> If the kernel and the actual device disagree, could it also "break"
> the ethernet adaptor? This also doesn't work.
Well, the ethernet adapter would likely be a different problem. Do you
get any interrupts showing up in /proc/interrupts for either the disk or
the network adapter?
Hello,
On Tue, Nov 24, 2009 at 9:42 PM, Robert Hancock <[email protected]> wrote:
> On 11/24/2009 05:29 AM, Otavio Salvador wrote:
> Well, the ethernet adapter would likely be a different problem. Do you get
> any interrupts showing up in /proc/interrupts for either the disk or the
> network adapter?
I guess I did the test with a mangled configuration file and then the
network didn't work. Now I tested it again and it has work fine.
What still fails is the disk. It still doesn't work et all.
I'm sending the /proc/interrupts. It got IRQ 14 for the disk adaptor.
Cheers,
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br
Hello,
On Tue, Nov 24, 2009 at 10:51 AM, Juergen Beisert <[email protected]> wrote:
> Otavio Salvador wrote:
>> On Tue, Nov 24, 2009 at 9:31 AM, Otavio Salvador
>>
>> <[email protected]> wrote:
>> > I'll take a look on the dmesg output and see if I see something
>> > interesting :-)
>>
>> In fact I did find some interesting issues. First:
>> [...]
>
> Our kernel states:
> CPU: Vortex86 SoC 05/02 stepping 02
>
> Your kernel states:
> CPU: Cyrix Cx486SLC
>
> BTW: Try to enable the DEBUG macro in "arch/x86/include/asm/pci_x86.h" and
> boot again this kernel. On earlier kernels this increases the verbosity of
> the kernel when it tries to detect the IRQ routing.
So I did. Here goes the dmesg with the debug enabled.
--
Otavio Salvador O.S. Systems
E-mail: [email protected] http://www.ossystems.com.br
Mobile: +55 53 9981-7854 http://projetos.ossystems.com.br