Re: Upgrade driver to 1.0.6, memory allocation?

[Takashi Iwai <tiwai@suse.de>]

At Wed, 18 Aug 2004 19:57:23 -0700,
Eliot Blennerhassett wrote:
> 
> Hello all,
> 
> with the upgrade to 1.0.6, my alsa driver broke. (asihpi)
> Now I'm trying to put it together again, and could do with some tips...
> Mostly it still contained lots of easy to fix snd_magic_cast() but
> Can anybody confirm that the following update is the 'correct' thing to do?
> 
> (My) old code:
> call snd_malloc_pages_fallback in the open callback.

Oh sorry, my last clean-up hits you...

> 
> New code:
> call snd_pcm_lib_malloc_pages() in the hw_params callback
> 
> and call snd_pcm_lib_preallocate_pages_for_all() in the sound card pcm init function
> (in my case snd_card_asihpi_pcm()
> ================================
> By default according to pcm_memory.c, only a maximum of 4 substreams are allowed.
> My cards have up to 16 substreams.
> 
> Chapter 10 of "Writing an alsa driver" says "snd_pcm_lib_malloc_pages(substream,
> size);
> Note that you have to pre-allocate to use this function. "
> 
> Does this only mean that you have to preallocate at least one buffer to use the
> function?

This doesn't matter.  snd_pcm_lib_malloc_pages() allocates the pages
automatically when no preallocated buffer exists or the request buffer
size exceeds the preallocated buffer.

Just call snd_pcm_preallocate_pages_for_all() for the all pcm
instances to use snd_pcm_lib_malloc_pages().


> If I read it correctly, snd_pcm_lib_malloc_pages will still try to allocate some
> memory for substreams> 4, this just becomes less likely as the system mem becomes
> fragmented.

Well but in your old code you allocated pages in the open callback,
right?  So it shouldn't be different.  The buffer allocation in
hw_params callback is usually also done only once per open.


> Is there a rationale for making the substreams limit 4 as opposed to any other number?
> I guess I can change the max_substreams by providing a module parameter to snd-pcm
> module.
> ============================
> snd_pcm_lib_preallocate_pages1()
> ...
> if (size> 0 && preallocate_dma && substream->number <maximum_substreams)
> preallocate_pcm_pages(substream, size);
> ...
> Maybe I'm missing something, but it seems the possible error return value from
> preallocate_pcm_pages is ignored - is this intentional?

Yes.  The error in preallocation isn't critical since
snd_pcm_lib_malloc_pages() itself allocates the buffer, too.


Takashi


-------------------------------------------------------
SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media
100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33
Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift.
http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285