Re: alsa-devel@lists.sourceforge.net

Re: alsa-devel@lists.sourceforge.net

[Mark Ziegler]

Hi Kai,

thank you. You are right: my ide-dma is turned off. I only activated it in the 
SUSE setup, but I've never check if it is really switched on.
Unfortunately I can not turn on the dma:

linux:/home/woody # hdparm -m 8 -d 1 -u 1 -c 1 /dev/hda
/dev/hda:
 setting 32-bit IO_support flag to 1
 setting multcount to 8
 setting unmaskirq to 1 (on)
 setting using_dma to 1 (on)
 HDIO_SET_DMA failed: Operation not permitted
 multcount    =  8 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  1 (on)
 using_dma    =  0 (off)

Ok, this is probably because I've bought a brandnew motherboard with a 
brandnew chipset (VIA KT400). So I will seek around for a dma solution. 
Anyway, I tried again ecasound:
woody@linux:~> ecasound  -z:db -z:intbuf -i alsa,ice_spdif -o test.wav
****************************************************************************
*               ecasound v2.0.4 (C) 1997-2001 Kai Vehmanen                 *
****************************************************************************
(eca-chainsetup) Using double-buffer of 100000 sample frames.
(eca-chainsetup) Enabling extra buffering on realtime devices.
- [ Chainsetup/Enabling audio inputs ] -------------------------------------
(eca-audio-objects) Audio object "alsa", mode "read".
(audio-io) Format: s16_le, channels 2, srate 44100, interleaved.
- [ Chainsetup/Enabling audio outputs ] ------------------------------------
(eca-audio-objects) Audio object "test.wav", mode "read/write".
(audio-io) Format: s16_le, channels 2, srate 44100, interleaved.
- [ Engine/Init - mixmode "simple" ] ---------------------------------------
(eca-main) Prefilling i/o buffers.
(audioio-alsa3) warning! capture overrun - samples lost!  Break was at least 
7.60 ms long.
(audioio-alsa3) warning! capture overrun - samples lost!  Break was at least 
215.45 ms long.
(audioio-alsa3) warning! capture overrun - samples lost!  Break was at least 
0.03 ms long.
- [ Engine/Exiting ] -------------------------------------------------------
(audioio-alsa3) WARNING! While reading from ALSA-pcm device C0D0, there were 3 
overruns.

That was only 1 Minute sampling.  Sometimes I get up to 500ms. That looks 
strange to me. Mmmh, I should really hurry up getting my dma to work. 
215ms!!!

But thank you very much

  Mark





-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: Re: alsa-devel@lists.sourceforge.net

[Kai Vehmanen]

Sorry for the late response..

On Thu, 21 Nov 2002, Mark Ziegler wrote:

> thank you. You are right: my ide-dma is turned off. I only activated it in the 
> SUSE setup, but I've never check if it is really switched on.
> Unfortunately I can not turn on the dma:
[...] 
> linux:/home/woody # hdparm -m 8 -d 1 -u 1 -c 1 /dev/hda
> /dev/hda:
>  setting 32-bit IO_support flag to 1
>  setting multcount to 8
>  setting unmaskirq to 1 (on)
>  setting using_dma to 1 (on)
>  HDIO_SET_DMA failed: Operation not permitted
>  multcount    =  8 (on)
>  IO_support   =  1 (32-bit)
>  unmaskirq    =  1 (on)
>  using_dma    =  0 (off)

This is bad. You should check the linux-kernel archives and other 
relevant mailing lists... hopefully some newer kernel provides dma-support 
for your ide-chipset. This is critical for low-latency audio.

> Anyway, I tried again ecasound:
> woody@linux:~> ecasound  -z:db -z:intbuf -i alsa,ice_spdif -o test.wav

Ok, this will give us quite a lot of info:

> (audioio-alsa3) warning! capture overrun - samples lost!  Break was at least 
> 7.60 ms long.
> (audioio-alsa3) warning! capture overrun - samples lost!  Break was at least 
> 215.45 ms long.
> (audioio-alsa3) warning! capture overrun - samples lost!  Break was at least 
> 0.03 ms long.

This means that the xruns are caused by process scheduling latencies, not
by disk i/o scheduling. The longs xruns are probably caused by disk i/o
events that reserve the cpu for a long period of time (dma not used -> all
transfers go through the cpu -> trouble). I'd suggest trying low-latency /
pre-emption kernel patches, and/or finding a kernel with dma support for 
your ide chipset.

-- 
 http://www.eca.cx
 Audio software for Linux!



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

Re: Re: alsa-devel@lists.sourceforge.net

[Eric Dantan Rzewnicki]

On Mon 02/12/2002 01:14:54, Kai Vehmanen wrote:
> Sorry for the late response..
> 
> On Thu, 21 Nov 2002, Mark Ziegler wrote:
> 
> > thank you. You are right: my ide-dma is turned off. I only activated it in the 
> > SUSE setup, but I've never check if it is really switched on.
> > Unfortunately I can not turn on the dma:
> [...] 
> > linux:/home/woody # hdparm -m 8 -d 1 -u 1 -c 1 /dev/hda
> > /dev/hda:
> >  setting 32-bit IO_support flag to 1
> >  setting multcount to 8
> >  setting unmaskirq to 1 (on)
> >  setting using_dma to 1 (on)
> >  HDIO_SET_DMA failed: Operation not permitted
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
FWIW a co-worker and I have both seen this problem recently with Intel's
D845EBG2 motherboard. 845E is the chipset on the board. The ICH4 IDE
controller (82801DB I/O Controller Hub 4) is supposed to be capable of
ATA 66/100. But, with 2.4.19 and 2.4.18 we were unable to set
using_dma = 1 
When we try we get the same "Operation not permitted" error noted 
above. My co-worker has worked around it by using the 2.2.20 kernel. For 
whatever reason the 2.2.20 kernel let's him turn dma on with this 
chipset. I've chosen to stay with my old box rather than downgrade to
the 2.2 series kernel.

I know this isn't a solution, but perhaps the extra info will help
someone.

> >  multcount    =  8 (on)
> >  IO_support   =  1 (32-bit)
> >  unmaskirq    =  1 (on)
> >  using_dma    =  0 (off)
> 
> This is bad. You should check the linux-kernel archives and other 
> relevant mailing lists... hopefully some newer kernel provides dma-support 
> for your ide-chipset. This is critical for low-latency audio.
> 

-Eric Rz.


-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T 
handheld. Power & Color in a compact size! 
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en