USB audio device on Raspberry Pi - link_elf: symbol isa_dmastatus undefined

[Matthias Meyser]

Am 05.05.2014 18:08, schrieb Bernd Walter:
> On Mon, May 05, 2014 at 09:34:48AM -0600, Ian Lepore wrote:
>> On Mon, 2014-05-05 at 16:47 +0200, Bernd Walter wrote:
>>> On Mon, May 05, 2014 at 04:33:56PM +0200, Hans Petter Selasky wrote:
>>>> On 05/05/14 15:50, Matthias Meyser wrote:
>>>>>
>>>>> Am 02.05.2014 18:00, schrieb Hans Petter Selasky:
>>>>>> On 05/01/14 17:08, Hans Petter Selasky wrote:
>>>>>>> On 05/01/14 01:34, Johny Mattsson wrote:
>>>>>>>> On 1 May 2014 00:14, Ian Lepore <ian at freebsd.org> wrote:
>>>>>>>>
>>>>>>>>> I was doing some testing on a wandboard (about twice as fast an an
>>>>>>>>> rpi)
>>>>>>>>> with
>>>>>>>>> more than 20k int/sec without having any problems.
>>>>>>>>>
>>>>>>>>
>>>>>>>> On a similar note, I've pushed an i.MX 283 (400MHz) board to above 300k
>>>>>>>> int/sec, on Linux. Admittedly at that point my shell wasn't what you'd
>>>>>>>> call
>>>>>>>> "responsive" however =) The ISR in that scenario was the GPIO
>>>>>>>> handler, so
>>>>>>>> probably a bit more light-weight than an audio ISR.
>>>>>>>
>>>>>>> I'll have a look and see if I can fix it.
>>>>>> Here is a patch (work in progress) which you can try:
>>>>>> http://home.selasky.org:8192/dwc_otg_isoc_support_wip.diff
>>>>>>
>>>>>> Still not working 100% reliable. Trying to figure out the last bits
>>>>>> and pieces.
>>>>>
>>>>> For testing it would be very helful if someone could add
>>>>>
>>>>> device sound
>>>>> device snd_uaudio
>>>>>
>>>>> to RPI-B kernel.
>>>>>
>>>>> Having this in BEAGLEBONE would be nice to.
>>>>> Perhaps this schould go in every config that supports usb.
>>>>>
>>>>
>>>> Hi,
>>>>
>>>> The following patch should make "devd" load sound.ko and snd_uaudio.ko
>>>> automatically:
>>>>
>>>> http://svnweb.freebsd.org/changeset/base/265359
>>>
>>> This won't work unless sound.ko especting ISA in ARM kernel is fixed
>>> as well.
>>>
>>
>> This shouldn't be a problem.  The code that references isa stuff is
>> wrapped in #ifdef DEV_ISA, and an arm kernel build won't have that
>> defined.  The problem was that the makefile for building a sound module
>> created an opt_isa.h containing #define DEV_ISA when building for arm so
>> the loadable sound modules had references to isa but the driver in the
>> kernel didn't.
>
> Didn't test myself, but this thread started with a problem like this.

For clarification.

Modules would be fine. But no *sound* and *snd* Modules are
build during normal buildworld/kernel or are not installed during 
installworld on arm/armv6.