hdspmixer segfault (1.0.24+)

hdspmixer segfault (1.0.24+)

[Fernando Lopez-Lezcano]

Hi, I'm getting a segfault from hdspmixer using the current git 
alsa-tools (1.0.23 works, 1.0.24+ does not). See below for a backtrace. 
I'm running 2.6.33.x rt30, could this be the problem? (a dependency on 
newer kernels?)

-- Fernando



$ gdb ./hdspmixer/src/hdspmixer
GNU gdb (GDB) Fedora (7.2-41.fc14)
Copyright (C) 2010 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later 
<http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-redhat-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from 
/scratch/nando/alsa-tools/alsa-tools.git/hdspmixer/src/hdspmixer...done.
(gdb) run
Starting program: 
/scratch/nando/alsa-tools/alsa-tools.git/hdspmixer/src/hdspmixer
[Thread debugging using libthread_db enabled]

HDSPMixer 1.6 - Copyright (C) 2003 Thomas Charbonnel <thomas@undata.org>
This program comes with ABSOLUTELY NO WARRANTY
HDSPMixer is free software, see the file COPYING for details

Looking for RME cards:
Card 0: RME Hammerfall DSP + Multiface at 0xe9200000, irq 19
Multiface found!
Card 1: HDA Intel at 0xe9300000 irq 32
1 RME cards card found.
Initializing default presets

Program received signal SIGSEGV, Segmentation fault.
HDSPMixerPresets::restore_preset (this=0x81fe500, prst=1) at 
HDSPMixerPresets.cxx:176
176		    basew->inputs->strips[i]->pan->pos[z] = 
basew->inputs->strips[i]->data[card][speed][p]->pan_pos[z];
Missing separate debuginfos, use: debuginfo-install 
alsa-lib-1.0.24-1.fc14.i686 expat-2.0.1-10.fc13.i686 
fltk-1.1.10-2.fc14.i686 fontconfig-2.8.0-2.fc14.i686 
freetype-2.4.2-4.fc14.i686 glibc-2.13-1.i686 libX11-1.3.4-3.fc14.i686 
libXau-1.0.6-1.fc14.i686 libXext-1.1.2-2.fc14.i686 
libXft-2.1.14-1.fc13.i686 libXinerama-1.1-2.fc13.i686 
libXrender-0.9.6-1.fc14.i686 libgcc-4.5.1-4.fc14.i686 
libstdc++-4.5.1-4.fc14.i686 libxcb-1.7-1.fc14.i686
(gdb) bt
#0  HDSPMixerPresets::restore_preset (this=0x81fe500, prst=1) at 
HDSPMixerPresets.cxx:176
#1  0x080552de in HDSPMixerPresets::preset_change (this=0x81fe500, p=1) 
at HDSPMixerPresets.cxx:258
#2  0x0804b436 in readregisters_cb (arg=0x80a0710) at HDSPMixerWindow.cxx:67
#3  0x4dd311d3 in Fl::wait(double) () from /usr/lib/libfltk.so.1.1
#4  0x4dd313e4 in Fl::run() () from /usr/lib/libfltk.so.1.1
#5  0x0804af34 in main (argc=1, argv=0xbffff284) at hdspmixer.cxx:114
(gdb)

Re: hdspmixer segfault (1.0.24+)

[Adrian Knoth]

On Sat, Feb 26, 2011 at 04:55:24PM -0800, Fernando Lopez-Lezcano wrote:

> Hi, I'm getting a segfault from hdspmixer using the current git 
> alsa-tools (1.0.23 works, 1.0.24+ does not). See below for a backtrace. 
> I'm running 2.6.33.x rt30, could this be the problem? (a dependency on 
> newer kernels?)

I have fixed this bug a couple of days ago:

   http://thread.gmane.org/gmane.linux.alsa.devel/82452

They still need to be applied to the alsa-tools.git, but if you like,
you can grab the patches and verify they work for you.


Cheers

-- 
mail: adi@thur.de  	http://adi.thur.de	PGP/GPG: key via keyserver

Re: hdspmixer segfault (1.0.24+)

[Fernando Lopez-Lezcano]

On 02/27/2011 03:17 AM, Adrian Knoth wrote:
> On Sat, Feb 26, 2011 at 04:55:24PM -0800, Fernando Lopez-Lezcano wrote:
>
>> Hi, I'm getting a segfault from hdspmixer using the current git
>> alsa-tools (1.0.23 works, 1.0.24+ does not). See below for a backtrace.
>> I'm running 2.6.33.x rt30, could this be the problem? (a dependency on
>> newer kernels?)
>
> I have fixed this bug a couple of days ago:
>
>     http://thread.gmane.org/gmane.linux.alsa.devel/82452
>
> They still need to be applied to the alsa-tools.git, but if you like,
> you can grab the patches and verify they work for you.

They do work in a first (remote) test. Thanks very much!!
-- Fernando

Re: hdspmixer segfault (1.0.24+)

[Fernando Lopez-Lezcano]

On 02/27/2011 11:21 AM, Fernando Lopez-Lezcano wrote:
> On 02/27/2011 03:17 AM, Adrian Knoth wrote:
>> On Sat, Feb 26, 2011 at 04:55:24PM -0800, Fernando Lopez-Lezcano wrote:
>>
>>> Hi, I'm getting a segfault from hdspmixer using the current git
>>> alsa-tools (1.0.23 works, 1.0.24+ does not). See below for a backtrace.
>>> I'm running 2.6.33.x rt30, could this be the problem? (a dependency on
>>> newer kernels?)
>>
>> I have fixed this bug a couple of days ago:
>>
>>      http://thread.gmane.org/gmane.linux.alsa.devel/82452
>>
>> They still need to be applied to the alsa-tools.git, but if you like,
>> you can grab the patches and verify they work for you.
>
> They do work in a first (remote) test. Thanks very much!!

But one of my users (in Planet CCRMA) still has problems with his 9652, 
apparently the output labels for that card are missing and that is 
confirmed by looking at the source (his card works fine with the 1.0.23 
hdspmixer). I applied all your patches to current git except for the 
version number so ignore the 1.6 version printed. I'm waiting for a 
backtrace from his computer about the subsequent crash. See below for 
his message (and thanks for all the help!).

-- Fernando


-------- Original Message --------
Subject: 	[PlanetCCRMA] updated hdspmixer
Date: 	Sun, 27 Feb 2011 15:52:51 -0500
From: 	Craig Bourne <craigbourne@gmail.com>
To: 	planetccrma@ccrma.Stanford.EDU



I updated to alsa-tools 1.0.24.1-1.3.fc14.ccrma per today's note. This
does not yet fix hdspmixer as the following command-line invocation and
response shows.

-- 
Craig Bourne

*# hdspmixer*

HDSPMixer 1.6 - Copyright (C) 2003 Thomas Charbonnel <thomas@undata.org
<mailto:thomas@undata.org>>
This program comes with ABSOLUTELY NO WARRANTY
HDSPMixer is free software, see the file COPYING for details

Looking for RME cards:
Card 0: RME Hammerfall HDSP 9652 at 0xfdee0000, irq 17
HDSP 9652 found!
1 RME cards card found.
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Initializing default presets
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Unknown card type for output labels
Segmentation fault (core dumped)

Re: hdspmixer segfault (1.0.24+)

[Adrian Knoth]

On 02/28/11 00:54, Fernando Lopez-Lezcano wrote:

> But one of my users (in Planet CCRMA) still has problems with his 9652,
> apparently the output labels for that card are missing and that is
> confirmed by looking at the source (his card works fine with the 1.0.23
> hdspmixer). I applied all your patches to current git except for the

Indeed. There have never been labels, and now it's crashing, most likely
because of dereferencing the non-existing label array.

Do you mind to try this patch? Maybe it's enough. If Craig feels happy
to do some remote try&error, I can sort this out with him.

Right now, I'm simply assigning a stupid "1", "2" .. "64" channel
naming, but maybe Craig can provide us with the correct output labels? I
don't have access to a H9652, so I don't have the slightest clue how the
ports are named. ;)


HTH

diff --git a/hdspmixer/src/HDSPMixerOutput.cxx
b/hdspmixer/src/HDSPMixerOutput.cxx
index 05a126c..0d8ea5c 100644
--- a/hdspmixer/src/HDSPMixerOutput.cxx
+++ b/hdspmixer/src/HDSPMixerOutput.cxx
@@ -262,6 +262,7 @@ void HDSPMixerOutput::setLabels()
       }
     } else {
       fprintf(stderr, "Unknown card type for output labels\n");
+      labels_input = labels_playback = labels_madi_ss;
     }
 }

Re: hdspmixer segfault (1.0.24+)

[Fernando Lopez-Lezcano]

On 02/28/2011 09:23 AM, Craig Bourne wrote:
> I expect that I may have time Wednesday of this week to help with the 
> testing, etc.

(I released a new package of alsa-tools in the planetccrma-testing 
repository so you can test - you have to install 
planetccrma-repo-testing and then yum --enablerepo planetccrma-testing 
install alsa-tools to get it).

-- Fernando


>
> Though low level technical details are mostly absent from the 
> documentation that RME provided, I will try to locate and forward a 
> copy of their technical manual just in the event that that might be of 
> some help. [Ah! Here it is! 
> <http://media.zzounds.com/media/hdsp9652_e-2c1bba79532db8cd1c9b6f0beb31e855.pdf>]
>
> Meanwhile, please have a look at /RME9652.conf/ which reads, in part 
> (starting from about line 38):
>
> ...
>
>  slave.pcm {
>       type route
>       slave {
>          pcm {
>             type hw
>             card $CARD
>          }
> channels 26
>       }
>       ttable.0.24 1
>       ttable.1.25 1
>    }
> ...
>
> This number also reflects RME's characterization of the card (from the 
> title page of their manual)
>
> PCI Busmaster Digital I/O System
> 2 + 24 Channels SPDIF / ADAT Interface
> 24 Bit / 96 kHz Digital Audio
> ADAT Sync In
> MIDI I/O
> TotalMix™
> 24 Bit / 96 kHz ü
> SyncAlign® ZLM® SyncCheck®
>
...

_______________________________________________
Alsa-devel mailing list
Alsa-devel@alsa-project.org
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel

Re: hdspmixer segfault (1.0.24+)

[Adrian Knoth]

On 02/28/11 18:23, Craig Bourne wrote:

> PCI Busmaster Digital I/O System
> 2 + 24 Channels SPDIF / ADAT Interface

I've digged into the driver source. Your H9652 is more or less a
Digiface. The channel *count* was never a question, this is defined by
the driver anyway. The remaining question simply was how to label the
output destinations.

>From the manual, the answer is pretty obvious: ADAT-1.1 to ADAT-1.8,
followed by ADAT-2.1 to ADAT-2.8 and ADAT 3.1-to 3.8, last is S/PDIF.

In double-speed mode (everything above 48kHz), the ADAT channels are cut
in half, so ADAT 1.1 to ADAT 1.4, ADAT 2.1 to 2.4 and so on. S/PDIF
remains unchanged as it supports higher sample rates at static channel
counts (like AES/EBU).


> At various times in the past when one was required to specify (e.g.,
> Input/Output) channels to an ALSA tool (or a program that relied on
> the ALSA sound architecture) a mismatch in this number (presumably of
> pcm channels) had less than desirable consequences.

We're not talking driver here, the critical parts remain unchanged in
the kernel.


> All of this leads me to the conjecture that fudging an arbitrary
> number that stands in for "channels" is perhaps not the best course
> (not to forget that, at the end of our testing, I would hope that this
> sound card that costs as much as the computer that houses it will not
> have suffered any damage).

Of course, this is open source. It comes without any warranty and you're
free to avoid it, but to the best of my knowledge, you must not be
afraid.

Firstly, it's userspace, so nothing can really kill your card. Second,
the whole logic is an FPGA, so it can always be flashed to a working
state if something goes wrong (it cannot go wrong, as you cannot
accidentally flash your card). Last but not least: the channel count has
always been correct, 26 all the time, only the name array was missing.

So we have this "read 26 channel names from the naming array", and it's
now a question which names to write in there. I've given the answer
above, but want to stress again that we're talking pure C pure userland
here, so it's close to a hello world program:

   printf("Channel name is %s\n", name_array[channel_number]);

with name_array being something like { "ADAT-1.1", "ADAT-1.2", .. } and
so on. Right now, with the package prepared by Fernando, you'll simply
see "1", "2", .. "26" instead of "ADAT-1.1", but this will change in a
minute.


HTH

Re: hdspmixer segfault (1.0.24+)

[Adrian Knoth]

On Mon, Feb 28, 2011 at 04:35:36PM -0500, Craig Bourne wrote:

Hi!

> S/PDIF], this is all very reassuring. I will carry on with the testing on
> Wednesday and report back my results.

For all who care, foremost probably Fernando:

   $ git clone http://cluster.inf-ra.uni-jena.de/~adi/alsa-tools.git
   $ git checkout adi-fixes

(or simply pull this branch into your local alsa-tools git repo).

I've reworked the code in question. At least the label thing should be
fine now, but let's wait for Craig to confirm it.

Unfortunately, I won't be around on Wednesday. Maybe Fernando will
prepare another test package, if not, compiling from source isn't that
hard:

After these two git commands, go into hdspmixer an run ./gitcompile.
Then, if everythings builds correctly, run ./src/hdspmixer so see if
it's working.


Note that the git repo above might disappear once we have fixed all
remaining issues, so if you read this mail in a couple of months from
now, don't expect the URL to be still functional. ;)



Cheers

-- 
mail: adi@thur.de  	http://adi.thur.de	PGP/GPG: key via keyserver