Re: [Xenomai] sensoray 626 patch [Re: Xenomai Digest, Vol 27, Issue 12]

[Jorge Ramirez Ortiz <jro@xenomai.org>]

>>>  Thank you for committing the driver.
>>>  I have access to the Sensoray 626 board and I will test the remarks you have
>>>  proposed. It will take me sometime, however. First of all I would like to test my
>>>  driver with two Sensoray boards. For now, I see that it won't work with more than
>>>  one card. It requires some changes to the attach and probe function. After I
>>>  implement this functionality I will take a look into the interrupt handling 
>>> service.
>>>
>>>  Seizing the opportunity, I would like to ask you a question about handling PCI
>>>  device. When I'm calling pci_register_driver() and I have two exactly the same PCI
>>>  devices will this function call appropriate probe() function two times? If yes,
>>>  then is the probe() function an appropriate place to make list of available 
>>> devices?
>>>
>>>  Wojciech Domski
>>>
>>>  Domski.pl
>>>
>>>  Wojciech.Domski.pl
>>>
>>
>>  invoking pci_register_driver in dev_s626_attach causes the probe interface in the
>>  pci driver to be executed for all the devices of the kind described in the pci id
>>  table; this is probably the reason why you can only register one.
>>
>
> of course, in case it was not clear, you can only register the pci driver once
>
>>  It might be easier to follow the approach in analogy/national_instruments/mite.c


please have a look to the changes I pushed (untested). let me know if it helps with 
your issues.
http://git.xenomai.org/xenomai-jro.git/log/?h=sensoray

[note that this driver is still very much work in progress but I thought you'll need 
it sooner rather than later]

>>
>>  mite_init registers the driver which ends up calling mite_probe which will create a
>>  mite_struct and add it to the list of mite_devices.
>>  Incidentally if you were wondering, this is the reason why if the
>>  pci_register_driver call succeeded the driver can display all the devices in the
>>  system.
>>
>>  the same would need to be done for the s626
>>
>>  on a different note, we are adding a new interface to rtdm to support the busy 
>> loops.
>>  rtdm_busywait_safe(__condition, __busy_delay, __busy_wait)
>>
>>  this should allow you to do things like this
>>
>>  -       while (!MC_TEST(P_MC2, MC2_UPLD_DEBI))
>>  -               rtdm_task_sleep(2000);
>>  +       ret = rtdm_busywait_safe(!MC_TEST(P_MC2, MC2_UPLD_DEBI), BUSY_DELAY,
>>  BUSY_SLEEP);
>>  +       if (ret)
>>  +               __a4l_err( "error waiting");
>>
>>
>

-- 
jro