[Openvortex-dev] Re: [Alsa-devel] au8830+via & alsa 0.9.7c

[Manuel Jander]

Hallo Wilfied,

On Mon, 2003-10-20 at 16:37, Wilfried Weissmann wrote:
> hi,
> 
> last weekend i grabbed my old aureal vortex 2 card and replaced my 
> current (lowcost) card with it. and after installing the brand new 
> alsa-0.9.7c i got a fully functional soundcard with full duplex and 
> hardware mixing. whoooho! i was in heaven... for about 30 seconds. then 
> my via kt133 chipset got me an instant ticket to the ground level. 
> "welcome aboard ladies and gentlemen to flight kt133. our current 
> crusing altitute is 0 feet above ground level." :/
> but the goods news is that only the playback stopped. no kernel crash, 
> no freezes, only inconveniet "Vortex: vortex_fifo_setadbctrl fail" in my 
>   /var/log/messages. but as an old internet surfer i knew that the truth 
> is somewhere out there. at google! so i looked up the pci register fix 
> put the line "install snd-au8830 /sbin/setpci -d '12eb:*' 40.B=ff ; 
> insmod snd-au8830" in my modules.conf and everything worked fine. at the 
> openvortex list i read that there is also another fix that tunes the via 
> agp-bridge. i have not tried this one yet, but i guess that would work 
> too (although in the example from the mail they set a register in the 
> au88x0 instead of the bus-bridge). but i thought that the fix was 
> already in the driver. afterwards i had a look at the sources and could 
> not find code that would do this. is there another workaround in the 
> driver (apart from the loop with a counter around the 
> vortex_fifo_setadbctrl() that stopps the kernel from freezing) or is it 
> just not in there?

Well, you just confirmed me that there was really a bug in the old
binary driver. The old binary driver keeps looping with some locks held
for a hardware flag that never changes, freezing your computer. Jeff
Muizelaar added a "lifeboat" to that loop, but very later we noticed
that this semes to help stability. Now we have the proove :D

> there is also somewhere a bug in recording (unless this is caused by 
> still running on the 0.9.6 libs. i will upgrade that soon.). i 
> discovered it when i used speakfreely with alsa support. in one second 
> intervals there was some loud noise that sounded like garbage data. when 
> i compiled speakfreely with oss support and linear audio then the 
> distortion was gone. with "arecord -f MU_LAW -r 16000 -t au x.au" i can 
> reproduce something similar. this time there is 1 second of normal 
> audio. for few milliseconds there appears some short sample from the 
> previous second again. then 1 second of normal playback and a sample 
> from that is also played again....
> it feels like that the buffer does not have the right size. if you 
> choose higher sampling rates, 16-bit samples and stereo then the samples 
> that are played twice are getting shorter and finally the problem 
> disappears. any ideas?

This could be a period size bug. I didn't test the Alaw support myself,
but know i see that it could be useful for VoIP.

> i am using kernel 2.4.22 tainted by NVidia.o(tm) on a athlon k7 800MHz 
> box with a via kt133 chipset. the distribution is debian-sarge 
> (testing). no alsa plugins are used.

Great to here that story :D

Best Regards

Manuel Jander