ALSA can be backported to 2.4 kernels without much problems. Here are
my experiences with ALSA backporting into my linux-2.4-jp kernel
from 2.5.4 and 2.5.6, but it should work with other 2.4 kernels, too.
I use a patch sequence method. That means, a sequence of patch files is
created to remove the old sound drivers and insert ALSA, so keeping
track of the changes in 2.5 kernels will be easier.
There are three steps to be taken:
1. Set up a patch sequence to remove the 2.4 kernel sound.
The directory 'linux-2.4/drivers/sound' must be removed completely, the
Makefile must be changed back to remove the sound driver dependencies.
Additionally, changes related to sound drivers in 'pre' patches can
occure, they must be reverted, too. Also, it is useful to change the
defconfig's and config.in's in the architecture directories in order
to clean up. The sequence I use consists of a three step rollback:
1918 Mär 12 12:26 01_kernel-sound-remove-0-2.4.19pre3
4839575 Mär 12 11:41 01_kernel-sound-remove-1-2.4.19pre2
21675 Mär 12 11:41 01_kernel-sound-remove-2-core
The 2.4 kernel has no sound capabilities at this point and is
ready to get ALSA inserted.
2. Extract ALSA from 2.5 kernel. The ALSA directories are located
in 'linux-2.5/sound' and 'linux-2.5/include/linux/sound'. They can be
diff'ed against a 'linux-2.4-nosound' tree and patched to get a
'linux-2.4-alsa' tree. To complete the patch, the MAINTAINERS, Makefile,
and architecture dependent config.in's must be adapted from the 2.5 kernel.
linux-2.4-alsa/MAINTAINERS
linux-2.4-alsa/Makefile
linux-2.4-alsa/arch/i386/config.in
linux-2.4-alsa/arch/ppc/config.in
As you see, i386 and ppc are supported. If you want to support a different
architecture, the config.in's must be added respectively.
A convenient method is to keep a patch sequence according to 2.5 kernels
as they are still under development. Backporting ALSA will be work in progress
as long as development is going on. The ALSA versions used in 2.5 are
2.5.4 - alsa 0.9.0beta10
2.5.5 - alsa 0.9.0beta11
2.5.6 - alsa 0.9.0beta12
Using incremental patches is recommended to keep track with the
2.5 development. The patch sequence I use contains a 0.9.0beta10
ALSA patch of kernel 2.5.4, and two incremental patches to update to the
status of 2.5.6:
10419626 Mär 13 12:06 02_alsa-0.9.0beta10
30029 Mär 13 11:50 02_alsa-0.9.0beta10-beta12-include
750533 Mär 14 02:12 02_alsa-0.9.0beta10-beta12-sound
In these patches, there are some minor changes:
- the remap_page_range() first argument is dropped in several drivers
for compatibility
- three times a 'need_resched()' is changed to 'current->need_resched'
- in include/linux/proc_fs.h, some definitions for using 'PDE()' are added
3. Some extra changes are required for 2.4 kernels.
I collect them in a separate patch. Fortunately, there are no dramatic
additions:
1456 Mär 14 00:37 02_alsa-0.9.0beta12-addon
In this patch
- ISA I/O bus memory addresses are defined by some macros in
linux/include/asm-i386/io.h
- minor fixes in kernel config for making them work properly with 'make
xconfig'
At this point, the 2.4 ALSA kernel can be configured and will compile.
Maybe you have to increase the default tkparse.c variable space from
2048 to a higher value, like 4096.
Also take care of /etc/modules.conf to adjust some names before rebooting.
The char-major-14 alias can be safely removed. Here is a configuration for a
single Maestro3 card.
alias char-major-116 snd
alias snd-card-0 snd-maestro3
#alias char-major-14 soundcore
alias sound-slot-0 snd-card-0
alias sound-service-0-0 snd-mixer-oss
alias sound-service-0-1 snd-seq-oss
alias sound-service-0-3 snd-pcm-oss
alias sound-service-0-8 snd-seq-oss
alias sound-service-0-12 snd-pcm-oss
options snd snd_cards_limit=1 snd_major=116
options snd-maestro3 snd_index=0
After rebooting with the new kernel and ALSA had been configured as modules,
you can check the status with lsmod and you should get something like this:
snd-pcm-oss 42608 1 (autoclean)
snd-mixer-oss 10800 1 (autoclean) [snd-pcm-oss]
snd-maestro3 12944 2
snd-pcm 58144 0 [snd-pcm-oss snd-maestro3]
snd-timer 12144 0 [snd-pcm]
snd-ac97-codec 23664 0 [snd-maestro3]
snd 32480 0 [snd-pcm-oss snd-mixer-oss snd-maestro3
snd-pcm snd-timer snd-ac97-codec]
(The snd-seq-oss modules is not installed due to lack of hardware sequencer)
Now the ALSA library and utilities be compiled. Take care of the
correct version number.
Problems
There are some issues, but no showstoppers. With xmms playing MP3 files, the
following messages are written into /var/log/messages:
[...]
Mar 14 13:28:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -984, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:29:16 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -576, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:32:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -424, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:33:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -864, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:35:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -832, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:36:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -960, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:45:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -696, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:47:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -952, max jitter = 8192): wrong interrupt acknowledge?
Mar 14 13:52:18 jungle kernel: Unexpected hw_pointer value (stream = 0,
delta: -760, max jitter = 8192): wrong interrupt acknowledge?
[...]
The sound still works with these messages. Notice that ALSA generates
the messages at full minute intervals. i don't know what is going on.
While shutting, down, ALSA is not able to free all resources, and writes
messages like the following to /var/log/messages:
Mar 14 13:15:44 jungle kernel: Not freed snd_alloc_pages = 48
Mar 14 13:15:44 jungle kernel: Not freed snd_alloc_kmalloc = 255
Mar 14 13:15:44 jungle kernel: kmalloc(9) from d093dac1 not freed
Mar 14 13:15:44 jungle kernel: kmalloc(76) from d093daa5 not freed
Mar 14 13:15:44 jungle kernel: kmalloc(9) from d093dac1 not freed
Mar 14 13:15:44 jungle kernel: kmalloc(76) from d093daa5 not freed
Mar 14 13:15:44 jungle kernel: kmalloc(9) from d093dac1 not freed
Mar 14 13:15:44 jungle kernel: kmalloc(76) from d093daa5 not freed
This might have the same reason as the 'unexpected hw_pointer value' stated
above.
I hope I could give some useful information to everybody who likes to
use ALSA in a 2.4 kernel. If you like to help me how to solve the problems
stated above, please email me.
Cheers,
Jörg Prante <joerg@infolinux.de>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/