Re: [Openvortex-dev] Driver updates to come.

[Jeff Muizelaar]

On 9 Jul 2003, Manuel Jander wrote:

> Hello,
>
> I rewrote the entire vortex_adb_waveroute() stuff, because frankly its
> all shit. I wrote a small and lean resource manager to allocate mixer
> and SRC usage dynamically without wasting any of them.
>
> Doing all this i noticed that doing the "adbctrl" stuff after routing
> dimished the chanced of chip and dale effect on DMA channel 0 from
> "always" to "only sometimes". The other DMA channels just behave the
> same (no chip and dale ever).
>
> This is a rather big update and could evetually break things. So decided
> to make a patch rather than updating the CVS.
> Mono recording doesnt work yet, but any feedback, suggestions, opinions
> etc mat be useful. Can anyone tell me if it works ?

I am reviewing the patch right now, without actually testing yet. I'll
list stuff as I come across it.

1. I think it would be better to use something like 'struct vortex_stream'
instead of 'stream_t'. stream_t gives the impression that it is a global
type and not something specific to the drivers. Also, typedefs, although
used readily through out alsa, are frounded upon in kernel style.

2. How about seperating vortex_adb_checkinout into vortex_adb_checkin and
vortex_adb_checkout, so that the each function only does one thing.

3. locking: I am not sure if using a spinlock is the best thing to do
here. Afaik we don't ever adjust the routing in the interrupt context so
sleeping during route allocation should be ok. I was thinking of just
having a adb semaphore.

4. I merged the spin_unlock stuff in read/write_codec

Other than that, from my brief look, things look pretty good.

-Jeff