[27514] in Hotline Meeting
Sparc 5 audio bugs (was: portnoy)
daemon@ATHENA.MIT.EDU (John Hawkinson)
Wed Mar 22 12:51:06 1995
Date: Wed, 22 Mar 1995 12:50:38 -0500
To: bugs@MIT.EDU
Cc: nocturne@MIT.EDU, hotline@MIT.EDU, sipb-staff@MIT.EDU, solaris@MIT.EDU
In-Reply-To: <9503221534.AA04400@opus.MIT.EDU>
and <9503221702.AA01300@opus.MIT.EDU>
From: John Hawkinson <jhawk@MIT.EDU>
Hi.
Eric Mumpower, <nocturne@mit.edu>, sent mail to hotline about some
problems with portnoy (A sparc5 in the sipb office). I'd just like to
make some clarifications and comments. FYR, they are transactions
27511 and 27513 in menelaus:/usr/spool/discuss/hotline.
[While logged in in console mode...]
> ran the command "more README". The file was displayed, but this was
> immediately followed by an endless stream of errors messages saying
> "short write -1 XX", where XX is usually 400 but occasionally 392
> and less occasionally other smaller integers. The machine continues
> to behave as usual, except for this endless stream of messages being
> printed to the console.
[...]
> In case it is relevant, earlier, when I ran "more" on a file longer
> than the screen (such that the "More" status line was printed at the
> bottom of the screen), Portnoy's speaker would make a short
> popping/ripping noise.
[...]
> At some point at least half an hour after the errors started flooding, they
> stopped, followed immediately by the error message:
> "WARNING: audio_4321_initlist: dma_list not cleared"
Firstly, the ``short write'' problem has been observed on portnoy at
least once before. We're not sure what causes it (obviously). I believe
the machine was in console mode then as well...
As Eric later says, the ``dma_list not cleared'' errors have shown up
in the console window (under X) for the past few weeks.
Nevertheless, I suspect all of these problems stem from the sparc5
audio_cs driver under Solaris. Apparently this driver has notoriously
been buggy, and sun has been working on it, and finally gave up and
shipped Van Jacobson a sparc 5, and he wrote them a new audio
driver. So it's buggy; I'm not exactly how, but I believe that the
Solaris 2.3 stock version has been superceded by a number of patches,
and also the 2.4 version is better than the stock 2.3.
Second, at some point a week or two ago, portnoy's /dev/audio stopped
working. The previous time this had happened, I was able to get it
back by mucking about with loading and unloading the audio_cs driver
and rebooting the machine (I don't recally the exact procedure I
used). I haven't had a chance to muck with it again lately, hence its
current state of brokenness.
This problem, btw, is similar to one that some people have reported on
the cluster sparc 5s. Audio on some of them seemed to break, or not
work correctly, after the Jan/Feb patchrelease.
Anyway, there's something wrong, and I'm fairly sure it's not a
hardware problem (at least, not one we can fix :-)), but I think
there's a Sun patch that may address it. If someone would like me to
try and find the number, I'll be happy to go and look for it...
--jhawk
