MPlayer for Kindle 4 Touch

[elRicharD]

I found this http://www.mobileread.mobi/forums/sh...d.php?t=162743, I will take a look tomorrow maybe. Thanks!

[elRicharD]

OK, I got myself a nice strace, which I attach.

What I can see with a somehow limited Linux devel knowledge is that there are some missing alsa config files at the beginning, and then it looks for some device files that are not present. Then there is this:

open("/dev/snd/pcmC0D0p", O_RDWR|O_NONBLOCK|0x80000) = 4
fcntl64(4, F_SETFD, FD_CLOEXEC)         = 0
close(3)                                = 0
ioctl(4, AGPIOC_ACQUIRE or APM_IOC_STANDBY, 0xbefbe484) = 0
fcntl64(4, F_GETFL)                     = 0x80802 (flags O_RDWR|O_NONBLOCK|0x80000)
ioctl(4, AGPIOC_INFO, 0xbefbe5a8)       = 0
clock_gettime(CLOCK_MONOTONIC, {4268, 147286896}) = 0
ioctl(4, AGPIOC_SETUP, 0xbefbe478)      = 0
mmap2(NULL, 4096, PROT_READ, MAP_SHARED, 4, 0x80000) = -1 ENXIO (No such device or address)
ioctl(4, 0xc0844123, 0xbefbe3b8)        = 0
fcntl64(4, F_GETFL)                     = 0x80802 (flags O_RDWR|O_NONBLOCK|0x80000)
fcntl64(4, F_SETFL, O_RDWR|0x80000 /* O_??? */) = 0
ioctl(4, 0xc0844123, 0xdfb20)           = 0
close(4)

That looks suspicious, but I can't make much of it.

If you can look at it, I will appreciate any help. Thanks!

[morantis]

Nice, I think some developers forget how easy it is to accomplish some things.

[knc1]

Originally Posted by elRicharD

OK, I got myself a nice strace, which I attach.

What I can see with a somehow limited Linux devel knowledge is that there are some missing alsa config files at the beginning, and then it looks for some device files that are not present. Then there is this:

open("/dev/snd/pcmC0D0p", O_RDWR|O_NONBLOCK|0x80000) = 4
fcntl64(4, F_SETFD, FD_CLOEXEC)         = 0
close(3)                                = 0
ioctl(4, AGPIOC_ACQUIRE or APM_IOC_STANDBY, 0xbefbe484) = 0
fcntl64(4, F_GETFL)                     = 0x80802 (flags O_RDWR|O_NONBLOCK|0x80000)
ioctl(4, AGPIOC_INFO, 0xbefbe5a8)       = 0
clock_gettime(CLOCK_MONOTONIC, {4268, 147286896}) = 0
ioctl(4, AGPIOC_SETUP, 0xbefbe478)      = 0
mmap2(NULL, 4096, PROT_READ, MAP_SHARED, 4, 0x80000) = -1 ENXIO (No such device or address)
ioctl(4, 0xc0844123, 0xbefbe3b8)        = 0
fcntl64(4, F_GETFL)                     = 0x80802 (flags O_RDWR|O_NONBLOCK|0x80000)
fcntl64(4, F_SETFL, O_RDWR|0x80000 /* O_??? */) = 0
ioctl(4, 0xc0844123, 0xdfb20)           = 0
close(4)

That looks suspicious, but I can't make much of it.

If you can look at it, I will appreciate any help. Thanks!

I read it from the start and my eyes stop at the same point yours did.

It searched for control files for channels 0...31 and found only a control file for channel 0 ...
Then (attempted) to open("/dev/snd/pcmC0D0p" ...
(Under Linux, you can "open" a non-existent file.)
Grabbed the GPIO pin it needed
Grabbed the constant clock it needed
Tried to memory map the output device - which failed because it did not exist.

No place to write the output bits and bytes - you will not hear anything.

Next highly technical question: Why?

Recent Linux kernels do "dynamic device creation" -
Your kernel may not be new enough (or may not be optioned at build time) to do that.
Translation:
Get a version of this application that does not expect "dynamic device creation".

How?
You might be able to configure the application at build time not to expect "dynamic device creation" - check the build configuration options of the application.

If that feature is not made visible in the configuration options - check which version of the kernel headers you are building against.
You want to be certain you are not using NEWER kernel headers than the kernel version you are running. (OLDER is ok if the app builds)
I.E: The expectation of dynamic device creation may be controlled by the kernel header files.

Check the /dev/snd/* file system sub-tree -
What devices do you have?
Do you have a device but not a control configuation file?
I.E: Maybe you only have "/dev/snd/pcmC31D0p" and only a config file for "/dev/snd/pcmC0D0p".
Make config files for the device(s) that you actually have.

And/Or: manually make a device node for "/dev/snd/pcmC0D0p" which the software (and the config file) expects.

Duh...
Probably some other things, but those popped to mind first.

Tell us what you try and the results of your trials.

Duh 2:
Play some music with the Kindle's built-in player;
While the music is playing, SSH into the Kindle and inspect the /dev/snd/* sub-tree of the file system;
Discover what device names are present while the music plays.

Or, if your Kindle model doesn't have a music player, enable TTS and do the above while it is talking to you.

We KNOW the KT can make sounds - We just need to find the device name it is using (which may be dynamic) and then work forward with that knowledge.

Hint:
uname -a
Will tell us the details of the kernel version you are running.
Post that output when you get a chance.

Duh 3:
Why did it continue with the sign-on message after it knew the device did not exist?

That is a built-in delay for the dynamic device creation system to do its thing.
The message string output is slow and blocking (allowing multiple task schedulings) while being output.

After a serious of failures, the application re-tried to setup the output device (presuming it had how waited long enough for the dynamic device setup to have been completed by the kernel).

So it seems clear that this application is expecting dynamic device creation and even allows for very slow dynamic device creation. (maybe not slow enough?)

[knc1]

In addition to the info in the post above:

The project looks like it has been stalled for over a year:
http://gitorious.org/kindlebrew/mplayer

But of interest (in addition to the README file) is the comment at the top of the build.sh file:

## Prefix where alsa-lib 1.0.13 for arm is installed
## For some reason, I couldn't get alsa to make a dynamic version of
## libasound.so.2, so I copied it from the Kindle
## If you do that you'll need to create a symbolic link:
## ln -s libasound.so.2 libasound.so

ALSAPREFIX=/opt/kindle

WATCHOUT for those text files - when I cut and pasted that, it came out with CRLF line endings.
So be sure the line endings are correct!

Hmm...
Probably time for a re-build against current Kindle libraries, using the correct kernel headers (2.6.31?), if only just for the drill.

Why?
Because the SIP thread/project (and others) stalled at this same point - no output sound.

[elRicharD]

I checked what you say in your comment, knc1.

[root@kindle snd]# ls -ltr
crw-rw----    1 root     root      116,  16 Jun  2 14:27 pcmC0D0p
crw-rw----    1 root     root      116,  24 Jun  2 14:27 pcmC0D0c
crw-rw----    1 root     root      116,  33 Jun  2 14:27 timer
crw-rw----    1 root     root      116,   0 Jun  2 14:27 controlC0
drwxr-xr-x    2 root     root            60 Jun  2 14:27 by-path
[root@kindle snd]# uname -a
Linux kindle 2.6.31-rt11-lab126 #1 Wed Apr 4 20:41:38 PDT 2012 armv7l GNU/Linux

As you can see, the device exists, and it has permissions to be read and written by root. I am using the 3.0 kernel that comes with Ubuntu 11.10.

When I play music there are no additional files created. If I try to run mpg123 when the built-in player is running, I get

[root@kindle us]# ./mpg123 music/02.mp3 
[alsa.c:165] error: cannot open device default
[audio.c:630] error: failed to open audio device
[audio.c:180] error: Unable to find a working output module in this list: alsa
[audio.c:532] error: Failed to open audio output module
[mpg123.c:897] error: Failed to initialize output, goodbye.

So it seems the communication with alsa is successful. I think the problem might be in the alsa lib I used to compile. I will copy the libasound I have in the Kindle and retry.

Thanks a lot!

[elRicharD]

Originally Posted by elRicharD

So it seems the communication with alsa is successful. I think the problem might be in the alsa lib I used to compile. I will copy the libasound I have in the Kindle and retry.

Thanks a lot!

Copied the .so files from my kindle to the arm gcc library path and recompiled. It didn't work again, same error all along.

open("/dev/snd/pcmC0D0p", O_RDWR|O_NONBLOCK|0x80000) = 4                                         
fcntl64(4, F_SETFD, FD_CLOEXEC)         = 0                                                          
close(3)                                = 0                                                      
ioctl(4, AGPIOC_ACQUIRE or APM_IOC_STANDBY, 0xbeea63d8) = 0                                                                                                                                                                                    
fcntl64(4, F_GETFL)                     = 0x80802 (flags O_RDWR|O_NONBLOCK|0x80000)              
ioctl(4, AGPIOC_INFO, 0xbeea6590)       = 0                 
clock_gettime(CLOCK_MONOTONIC, {9324, 51067599}) = 0                                       
ioctl(4, AGPIOC_SETUP, 0xbeea6584)      = 0                                                
mmap2(NULL, 4096, PROT_READ, MAP_SHARED, 4, 0x80000) = -1 ENXIO (No such device or address)
ioctl(4, 0xc0844123, 0xbeea64f8)        = 0                                        
fcntl64(4, F_GETFL)                     = 0x80802 (flags O_RDWR|O_NONBLOCK|0x80000)                                   
fcntl64(4, F_SETFL, O_RDWR|0x80000 /* O_??? */) = 0

If I tweak the sliders on alsamixer I hear white noise when mpg123 starts and I don't when I stop it, so it is somehow communicating with the sound hw.

And lastly, if I run mpg123 and try to use the builtin player, it does not work, so probably the hardware is in use.

[knc1]

Originally Posted by elRicharD

I checked what you say in your comment, knc1.

[root@kindle snd]# ls -ltr
crw-rw----    1 root     root      116,  16 Jun  2 14:27 pcmC0D0p
crw-rw----    1 root     root      116,  24 Jun  2 14:27 pcmC0D0c
crw-rw----    1 root     root      116,  33 Jun  2 14:27 timer
crw-rw----    1 root     root      116,   0 Jun  2 14:27 controlC0
drwxr-xr-x    2 root     root            60 Jun  2 14:27 by-path
[root@kindle snd]# uname -a
Linux kindle 2.6.31-rt11-lab126 #1 Wed Apr 4 20:41:38 PDT 2012 armv7l GNU/Linux

As you can see, the device exists, and it has permissions to be read and written by root. I am using the 3.0 kernel that comes with Ubuntu 11.10.

Your running on the 2.6.31 kernel version, I don't think that devfs was working in that version (it is in the 3.0 Ubuntu kernel).

So the next thing to try is a rebuild against the older (v2.6.31) kernel headers, not your build system default (v3.0) headers.

Even better yet -
If you can pull the kernel headers out of an Amazon source code bundle ....
That *should* also pick up any Amazon/lab126 changes to the sound device hardware.

- - - -

What are you using for your "cross compile environment" ?

- - - -

On the subject of headers - -
The strace you are using was not built against the Amazon/ARM kernel headers, that is why all of the untranslated ioctl calls in the listing.

Not a show stopper, the one you have works "good enough" for now.

[elRicharD]

I straced mplayer and it seems to have the same issue mpg123 has. So we can rule that out. But yet that one does have sound I couldn't analyze it deep enough yet, but I upload it so that you can check it out.

Thanks a lot for your help! I am learning a lot, and I hope I can do something useful with this...

[elRicharD]

My env consists of the ubuntu package arm-linux-gnueabi How is the kernel headers thing done? Should I get them from kernel.org, unzip them and then use some parameter in ./configure to use those?

I have some experience in cross-compiling, I own a Dingoo A320 and compiled Dingux for it, but they have a readily available toolchain, I just got the compiler, not the libs. Where are those to be downloaded? I didn't think that could be an issue, and since libasound and mpg123 compiled just fine I didn't look further...

Thanks again!

[knc1]

Originally Posted by elRicharD

My env consists of the ubuntu package arm-linux-gnueabi How is the kernel headers thing done? Should I get them from kernel.org, unzip them and then use some parameter in ./configure to use those?

I have some experience in cross-compiling, I own a Dingoo A320 and compiled Dingux for it, but they have a readily available toolchain, I just got the compiler, not the libs. Where are those to be downloaded? I didn't think that could be an issue, and since libasound and mpg123 compiled just fine I didn't look further...

Thanks again!

Duh...
You got me.
Will have to do a bit of research on how to do this with only a cross toolchain.

Meanwhile, ever use scratchbox2?
Neither have I, but it seems to be the popular cross-compile engine these days, and:
apt-cache search scratchbox
Yup - it is there
sudo apt-get update
sudo apt-get install scratchbox2 (and maybe some of the related packages)

Although I would have to start off at ground zero researching how to use it.
Other devs here do use scratchbox2 - so we may get some help.

Your other question -
(after you either get a cross-environment or at least build up a chroot where things like the kernel headers for some other version can be installed)

There is a sticky here that lists links to the Amazon source bundles -
Grab yourself a copy matching your device.

We can worry about building it for the armv6l (and older kernel versions) after we get it working for your hardware.

[geekmaster]

You can start a kernel build, then interrupt it after the headers are built... No need to let it run to completion.

[knc1]

Originally Posted by geekmaster

You can start a kernel build, then interrupt it after the headers are built... No need to let it run to completion.

The kernel build system has a target to make just the headers.
"make prepare" ??? ah well. "make help" will tell you.

[NiLuJe]

Stupid question: Did you link against libasound statically or dynamically? In any case, which version of libasound did you build against?

Try to build (dynamically) against the same version as the one shipped on your target, the ALSA API/ABI can be tricky sometimes...

EDIT: The 'official' toolchain used on the Touch is (apparently) a specific CodeSourcery TC (can't remember the exact version right now). I'm personally using a self-built TC with a more recent GCC version using crosstool-ng, but, depending on what you want to do, that might not be the easiest solution, or the right one at all [ie. if you want to build a kernel/kernel module, you'll probably have to use the exact same TC as Amazon] .

[knc1]

Originally Posted by NiLuJe

Stupid question: Did you link against libasound statically or dynamically? In any case, which version of libasound did you build against?

Try to build (dynamically) against the same version as the one shipped on your target, the ALSA API/ABI can be tricky sometimes...

EDIT: The 'official' toolchain used on the Touch is (apparently) a specific CodeSourcery TC (can't remember the exact version right now). I'm personally using a self-built TC with a more recent GCC version using crosstool-ng, but, depending on what you want to do, that might not be the easiest solution, or the right one at all [ie. if you want to build a kernel/kernel module, you'll probably have to use the exact same TC as Amazon] .

The "series 4" GCC puts its identification into the binaries it builds.
If someone has access to Kindle Touch binaries (something unstripped or maybe better a *.so) and does a bit of symbol listing (its in an extra or info block of symbols), then it should be easy to id the compiler version (4.2.4 is my guess - but that is a guess).

Then, we should be able to pick out which toolchain it was from this list:
http://www.mentor.com/embedded-softw.../lite-edition/
Just follow the (ARM) "gnu/linux" link (the glibc toolchain), put in your name and there will be a list of both the CS and the MG toolchains.
Make your first guess by picking something with a reasonable release date.

Note: The "advanced archives" are just tar-balls, no installer, but those can be redistributed. (and I can post a copy on my server for folks).
It is only the "self installer" archives that are limited re-distribution (the installer is CS/MG I.P.).