[Xenomai-help] ata revalidation failed?

[Xenomai-help] ata revalidation failed?

[Rob Wheeler]

We have a robot platform with 4 identical computers installed in its
base.  Three of the computers are running an ubuntu 8.04 server
distribution.  The fourth machine is running Xenomai 2.4.5 with a
2.6.24.3 kernel.  The Xenomai machine is used to send controls to the
robot's motors at a 1KHz frequency.  Sporadically, the Xenomai machine
locks up and we can no longer ping it from the outside world and the
control program appears to no longer be running.  Unfortunately, this
particular HW platform does not have a serial connection.  In the
meantime, we've connected an LCD monitor to the Xenomai machine, and
when it locked up, we noticed the following messages:

ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen
ata3.00: cmd ca/00:08:89:f0:6b/00:00:00:00:00/e0 tag 0 dma 4096 out
ata3.00: status: {DRDY}
ata3.00: revalidation failed (errno=-5)

This usually repeats a small number of times on the monitor, but the
computer is otherwise non-responsive (cannot even use keyboard
connected directly to the computer).  This revalidation error seems
somewhat reoccurring, however, there are occasionally other messages
that popup, such as:

Buffer I/O error on device sda1, logical block 884241

We've been able to reproduce this problem on multiple instances of
this same make of computer running Xenomai, but never on the same HW
platform running a vanilla kernel.

Anyone have any ideas or suggestions on how to track this one down?

-R

Re: [Xenomai-help] ata revalidation failed?

[Gilles Chanteperdrix]

Rob Wheeler wrote:
> We have a robot platform with 4 identical computers installed in its
> base.  Three of the computers are running an ubuntu 8.04 server
> distribution.  The fourth machine is running Xenomai 2.4.5 with a
> 2.6.24.3 kernel.  The Xenomai machine is used to send controls to the

Hi Rob,

what happened to your tests of Xenomai 2.4.6.1. You could not get it
running ?

I had a similar bug on my laptop, and it turned out that it was because
I used the ACPI processor module, which should not be used with Xenomai
anyway.

> Anyone have any ideas or suggestions on how to track this one down?

Try and enable the I-pipe tracer, and trigger a trace the first time you
get the revalidation error. Send the logs to the network using netconsole.

Enable other I-pipe debugging options such as context checking. Enable
Xenomai debugging options too.

-- 
					    Gilles.

Re: [Xenomai-help] ata revalidation failed?

[Paul]

Hi Rob

On Monday 26 January 2009, Rob Wheeler wrote:
> We have a robot platform with 4 identical computers installed in its
> base.  Three of the computers are running an ubuntu 8.04 server
> distribution.  The fourth machine is running Xenomai 2.4.5 with a
> 2.6.24.3 kernel.  The Xenomai machine is used to send controls to the
> robot's motors at a 1KHz frequency.  Sporadically, the Xenomai machine
> locks up and we can no longer ping it from the outside world and the
> control program appears to no longer be running.  Unfortunately, this
> particular HW platform does not have a serial connection.  In the
> meantime, we've connected an LCD monitor to the Xenomai machine, and
> when it locked up, we noticed the following messages:
>
> ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen
> ata3.00: cmd ca/00:08:89:f0:6b/00:00:00:00:00/e0 tag 0 dma 4096 out
> ata3.00: status: {DRDY}
> ata3.00: revalidation failed (errno=-5)
>
> This usually repeats a small number of times on the monitor, but the
> computer is otherwise non-responsive (cannot even use keyboard
> connected directly to the computer).  This revalidation error seems
> somewhat reoccurring, however, there are occasionally other messages
> that popup, such as:
>
> Buffer I/O error on device sda1, logical block 884241
>
> We've been able to reproduce this problem on multiple instances of
> this same make of computer running Xenomai, but never on the same HW
> platform running a vanilla kernel.
>
> Anyone have any ideas or suggestions on how to track this one down?

Yes - I have eperienced similar issues here with an AMD64 box - By adding 
`pci=noacpi` to the boot line appears to have cured the problem.


Regards, Paul.