[Bug 1959] New: cs46xx driver mmap_valid 0-->1 in kernel 2.6.x?

[Bug 1959] New: cs46xx driver mmap_valid 0-->1 in kernel 2.6.x?

[Martin J. Bligh]

http://bugme.osdl.org/show_bug.cgi?id=1959

           Summary: cs46xx driver mmap_valid 0-->1 in kernel 2.6.x?
    Kernel Version: 2.6.x
            Status: NEW
          Severity: normal
             Owner: drivers_sound@kernel-bugs.osdl.org
         Submitter: jaba@mikrobitti.fi


Distribution: Gentoo Linux
Hardware Environment: Asus A7V133, AMD Athlon
Software Environment: Gentoo Linux 1.4, default-x86-1.4, gcc-3.3.2,
glibc-2.3.3_pre20040117-r0, 2.6.2-rc1-mm2
Problem Description: This problem has been bothering me all over the kernel
2.6.x. I don't know if any other applications are affected by this bug, but a
game called Enemy Territory surely is.

With the default kernel 2.6 cs46xx driver (with "new DSP support" enabled) the
sound in Enemy Territory doesn't work at all, complaining "unable to mmap
/dev/dsp" during the ET startup. But if I do the patch below to cs46xx.c, sound
starts to work.

The patch I've used:
---
--- cs46xx_old.c        2004-01-27 14:09:46.625277768 +0200
+++ cs46xx.c    2004-01-27 12:32:28.000000000 +0200
@@ -54 +54 @@
-static int mmap_valid[SNDRV_CARDS] = {[0 ... (SNDRV_CARDS - 1)] = 0};
+static int mmap_valid[SNDRV_CARDS] = {[0 ... (SNDRV_CARDS - 1)] = 1};
---

Is there a reason to keep mmap_valid=0 in cs46xx driver or is this a real bug? I
haven't seen any new bugs with this patch applied, all the multimedia
applications (mplayer, xine, xmms, noatun) keeps on working.

Re: [Bug 1959] New: cs46xx driver mmap_valid 0-->1 in kernel 2.6.x?

[Takashi Iwai]

At Tue, 27 Jan 2004 07:14:40 -0800,
Martin J. Bligh wrote:
> 
> Is there a reason to keep mmap_valid=0 in cs46xx driver or is this a real bug? I
> haven't seen any new bugs with this patch applied, all the multimedia
> applications (mplayer, xine, xmms, noatun) keeps on working.

cs46xx doesn't always support the direct hardware buffer.
Only in some cases with a proper period (fragment) numbers, it can
support the direct buffer access.  mmap_valid option forces the driver
to allow OSS apps the direct accessing via mmap.  This might not work
always, depending on the parameter the app uses.  Use at your own
risk.

In short: it's not a real bug as long as mmap_valid=1 option works.
You had luck that your OSS apps (using mmap) seem working :)

--
Takashi Iwai <tiwai@suse.de>		ALSA Developer - www.alsa-project.org