Re: [9fans] usb/usbaudio and output speed?

Re: [9fans] usb/usbaudio and output speed?

[Russ Cox]

	term% usb/usbd
	term% usb/usbaudio
	term% usb/usbaudio	#  Note, invoking usb/usbaudio twice.
	term% kill usbaudio | rc

This is a problem in the usb driver.

It still doesn't the explain the other
ps hangs I've seen.  In general, ps hangs
because some proc's p->debug gets locked
and never unlocked.  Then opening that
proc's status file hangs.

In this case, postnote got confused because
a rendez structure got freed while a process
was still sleeping on it.  The result was that
postnote just keeps looping staring at this
freed memory, which will never again look like
an available lock.  Postnote is sitting on p->debug
while doing this.  Usually the looping doesn't
happen or loops once just to deal with transient
lock ordering problems.  It can't really do much
when it's been pointed at bad memory.

If anyone has any other reliable recipes for
making ps hang, I'm still all ears.  The
hangs I've seen had nothing to do with usb.

Russ

Re: [9fans] usb/usbaudio and output speed?

[nigel]

> Van Halen?  Oh man; you've got to get back to Bondi and hang with Bruce
> some more.
>
> 	- Dan C.

Asbury Park surely?

Re: [9fans] usb/usbaudio and output speed?

[Dan Cross]

(Wrong Bruce, Nigel.)

	- Dan C.

:-)

Re: [9fans] usb/usbaudio and output speed?

[Russ Cox]

> at home today (hey, that thing's heavy!).  I'll see if I can get
> it to crash again, and figure out where it's dying.  btw- I
> accidentally started two copies of usbaudio last night, and
> afterwards, `ps' started hanging.  It was a bummer.

If you can reliably make ps hang, please tell me how.
I've seen it happen only three times in the last three
months, and there's never enough information to figure
out exactly how it got into the state.

Russ

Re: [9fans] usb/usbaudio and output speed?

[Dan Cross]

> If you can reliably make ps hang, please tell me how.
> I've seen it happen only three times in the last three
> months, and there's never enough information to figure
> out exactly how it got into the state.

I've only seen it one other time in the last year, and suspected
it was a subtle race condition that only got tickled occasionally.
I'll see if I can get it to happen again, and let you know.

	- Dan C.

Re: [9fans] usb/usbaudio and output speed?

[Dan Cross]

> If you can reliably make ps hang, please tell me how.
> I've seen it happen only three times in the last three
> months, and there's never enough information to figure
> out exactly how it got into the state.

Okay, I can reliably (sic) make ps et al hang.  This is on an IBM
Thinkpad T21 aquired circa December, 2000.  The only USB device
connected to it at the time was a Xitel HiFi-LINK (USB audio).  It was
plugged into City power, and there was an audio CD (Rachmaninoff piano
concertos, I believe; I doubt that's relevant, but you never know) in
the CD-ROM drive at the time.  Here's what I did:

I booted up and logged in as none to minimize potential problems due to
my local environment; hopefully this should make this repeatable for
others.  My termrc is more or less what's in the distrubtion, modulo
some minor changes setting environment variables and the like, and one
``if ...'' that detects what network I'm using (and invokes ipconfig
accordingly) based on plan9.ini, so that should add minimal noise to
the mix.  I did an ``aux/vga -l $vgasize'' and started rio.  I swept
out a window and executed the following:

	term% usb/usbd
	term% usb/usbaudio
	term% usb/usbaudio	#  Note, invoking usb/usbaudio twice.
	term% kill usbaudio | rc

...And the last command hung.  I then swept out another window, tried
to run ps, but it also hung.  I then swept a third window, halted KFS,
and rebooted.  I repeated the experiment a couple of times to make sure
it wasn't a fluke, but each time got the same result.  I didn't have
time last night to investigate further (neither do I have time at the
moment, unfortunately).

Anyway, it seems that ps hangs pretty consistently when I do this.  Russ,
can you duplicate it?

	- Dan C.

Re: [9fans] usb/usbaudio and output speed?

[Boyd Roberts]

Dan Cross wrote:

> .... and there was an audio CD (Rachmaninoff piano concertos,
> I believe; I doubt that's relevant, but you never know)

I'd say it was the Rachmaninoff.  Haven't I told you to stick to Van Halen,
when the coding gets serious?

Re: [9fans] usb/usbaudio and output speed?

[Aharon Robbins]

In article <3DB80361.9070403@strakt.com>,
Boyd Roberts <9fans@cse.psu.edu> wrote:
>Dan Cross wrote:
>
>> .... and there was an audio CD (Rachmaninoff piano concertos,
>> I believe; I doubt that's relevant, but you never know)
>
>I'd say it was the Rachmaninoff.  Haven't I told you to stick to Van Halen,
>when the coding gets serious?

Nah, Fleetwood Mac is better.
--
Aharon (Arnold) Robbins --- Pioneer Consulting Ltd.	arnold@skeeve.com
P.O. Box 354		Home Phone: +972  8 979-0381	Fax: +1 928 569 9018
Nof Ayalon		Cell Phone: +972 51  297-545
D.N. Shimshon 99785	ISRAEL

Re: [9fans] usb/usbaudio and output speed?

[Jim Choate]

On Thu, 24 Oct 2002, Aharon Robbins wrote:

> In article <3DB80361.9070403@strakt.com>,
> Boyd Roberts <9fans@cse.psu.edu> wrote:
> >Dan Cross wrote:
> >
> >> .... and there was an audio CD (Rachmaninoff piano concertos,
> >> I believe; I doubt that's relevant, but you never know)
> >
> >I'd say it was the Rachmaninoff.  Haven't I told you to stick to Van Halen,
> >when the coding gets serious?
>
> Nah, Fleetwood Mac is better.

Skinny Puppy (!!!!)
The Cult (The Cure suck)
Social D
Ramones
T. Rex
Hanoi Rocks
Kidney Thieves
Rasputina
FLA
Front 242
Shonen Knife
Scraping Foetus Off The Wheel
Throbbing Gristle
Einsterzende Neubauten
Chemlab
NIN
Ministry
Nirvana
STP
AC/DC
Led Zep
Black Sabbath
Blue Cheer
Cream
Yardbirds
Doors
Red Hot Chili Peppers
PIL
Gorillaz
The White Stripes
The Vines
Lords of the New Church
Tool
Electric Frankenstein
GWAR
The Muffs
Man or Astroman (!!!!)
...


 --
    ____________________________________________________________________

    We don't see things as they are,                      ravage@ssz.com
    we see them as we are.                                   www.ssz.com
                                                  jchoate@open-forge.org
    Anais Nin                                         www.open-forge.org

    --------------------------------------------------------------------

Re: [9fans] usb/usbaudio and output speed?

[Dan Cross]

Van Halen?  Oh man; you've got to get back to Bondi and hang with Bruce
some more.

	- Dan C.

Re: [9fans] usb/usbaudio and output speed?

[Anthony Mandic]

Boyd Roberts wrote:
>
> Dan Cross wrote:
>
> > .... and there was an audio CD (Rachmaninoff piano concertos,
> > I believe; I doubt that's relevant, but you never know)
>
> I'd say it was the Rachmaninoff.  Haven't I told you to stick to Van Halen,
> when the coding gets serious?

	I was just listening to Iron Butterfly. Amazing what you can
	get done during 17 minutes of In-A-Gadda-Da-Vida (Vita?).

-am	� 2002

Re: [9fans] usb/usbaudio and output speed?

[Russ Cox]

usb/usbaudio is just simulating /dev/volume for you.
the real settings are accesssed via /dev/audioctl.
it's entirely likely that usbaudio gets the speed simulation
in /dev/volume wrong, since we tend to use only 44100.
the volume file is really a crutch.  it doesn't work well.

Re: [9fans] usb/usbaudio and output speed?

[Dan Cross]

> usb/usbaudio is just simulating /dev/volume for you.
> the real settings are accesssed via /dev/audioctl.
> it's entirely likely that usbaudio gets the speed simulation
> in /dev/volume wrong, since we tend to use only 44100.
> the volume file is really a crutch.  it doesn't work well.

Good to know.  However, when I tried to write things into
/dev/audioctl, the machine paniced.  I'm afraid I left my laptop
at home today (hey, that thing's heavy!).  I'll see if I can get
it to crash again, and figure out where it's dying.  btw- I
accidentally started two copies of usbaudio last night, and
afterwards, `ps' started hanging.  It was a bummer.

	- Dan C.

Re: [9fans] usb/usbaudio and output speed?

[Russ Cox]

you use what gets displayed.  /dev/volume
shows speed in hz so you use hz.   the volume
levels are in fact percentages.

/dev/volume speaks the language it understands.
echo -n speed 22050 >/dev/volume does
what you were trying to do.

the behavior you are seeing are very weird,
but probably just bugs in untested code paths.

Re: [9fans] usb/usbaudio and output speed?

[Dan Cross]

> you use what gets displayed.  /dev/volume
> shows speed in hz so you use hz.   the volume
> levels are in fact percentages.
>
> /dev/volume speaks the language it understands.
> echo -n speed 22050 >/dev/volume does
> what you were trying to do.
>
> the behavior you are seeing are very weird,
> but probably just bugs in untested code paths.

Hmm.  That was what I tried at first, but it didn't work.  I got
to thinking about this percentage business later.  Anyway, here's
what happens when I try to specify the speed in Hz:

term% grep speed /dev/volume
speed out 44100
term% echo -n speed out 22050 > /dev/volume
term% grep speed /dev/volume
speed out 55010
term% kill usbaudio | rc

After that, I decided it might be good to try to run usbaudio in
debugging mode, but that didn't work, either; it tries to dereference
a nil pointer somewhere:

term% usb/usbaudio -d
usbaudio 1034: suicide: sys: trap: fault read addr=0x0 pc=0x00008f5b
term%

Okay, a short acid trip showed me that I have to specify an integer to
along with the -d argument.  I guess I could have seen that by looking
at the code and seeing the ARGF() macro, but hey.  Anyway, running with
``-d9'' shows me:

term% echo -n speed 22050 > /dev/volume
line: speed 22050
rwrite: speed playback '11034400'
controlproc: setcontrol play speed 11034400
Configuring play endpoint for 55010 Hz
sent `ep 4 1 w 4 55010' to /dev/usb0/1/ctl
term%

11034400 is pretty dang close to like 22050 times 500, but I don't know
if that means anything.  Anyway:

term% echo -n speed out 34 > /dev/volume
line: speed out 34
rwrite: speed playback '21996'
controlproc: setcontrol play speed 21996
Configuring play endpoint for 21996 Hz
sent `ep 4 1 w 4 21996' to /dev/usb0/1/ctl
term%

Looks closer to reality, but I'm still kind of confused as to what's
going on (and I'm afraid I've got too much to do for tomorrow to really
dig much more tonight).

	- Dan C.

[9fans] usb/usbaudio and output speed?

[Dan Cross]

Hmm, this is weird.

term% grep speed /dev/volume
speed out 44100
term% echo -n speed out 50 > /dev/volume
term% grep speed /dev/volume
speed out 30000
term% echo -n speed out 100 > /dev/volume
term% grep speed /dev/volume
speed out 55010
term%

I was under the impression that, when setting things like output speed,
one should use percentages of values.  However, 50% of 44100 in real
space certainly isn't 30000.  Is this a bug, or am I just not
understanding how the USB audio device works?  Thanks!

	- Dan C.