Some bugs and patches

[Hans Petter Selasky]

Hi,

Do I have to post a PR or can I just post patches like this to get it 
committed?

1) I have some problems with an old soundblaster-8 (original)

*** sys/dev/sound/isa/sb8.c.ref Wed Oct 20 22:10:18 2004
--- sys/dev/sound/isa/sb8.c     Wed Oct 20 22:20:09 2004
***************
*** 565,571 ****
        if (sb->bd_flags & BD_F_HISPEED)
                sb_reset_dsp(sb);
        else
!               sb_cmd(sb, DSP_CMD_DMAEXIT_8);
  
        if (play)
                sb_cmd(sb, DSP_CMD_SPKOFF); /* speaker off */
--- 565,578 ----
        if (sb->bd_flags & BD_F_HISPEED)
                sb_reset_dsp(sb);
        else
!       {
!               /* sb_cmd(sb, DSP_CMD_DMAEXIT_8);
!                * NOTE: DSP_CMD_DMAEXIT_8
!                * does not work with old
!                * soundblaster 
!                */
!               sb_reset_dsp(sb);
!       }
  
        if (play)
                sb_cmd(sb, DSP_CMD_SPKOFF); /* speaker off */


2) there is panic with the vibra16x when trying to record:

*** sys/dev/sound/isa/sb16.c.ref        Wed Oct 20 22:10:18 2004
--- sys/dev/sound/isa/sb16.c    Mon Nov 22 17:56:05 2004
***************
*** 478,484 ****
  sb_intr(void *arg)
  {
        struct sb_info *sb = (struct sb_info *)arg;
!       int reason = 3, c;
  
        /*
         * The Vibra16X has separate flags for 8 and 16 bit transfers, but
--- 478,484 ----
  sb_intr(void *arg)
  {
        struct sb_info *sb = (struct sb_info *)arg;
!       int reason, c;
  
        /*
         * The Vibra16X has separate flags for 8 and 16 bit transfers, but
***************
*** 554,561 ****
        sb_reset_dsp(sb);
  
        if (sb->bd_flags & BD_F_SB16X) {
                pprio = sb->pch.run? 1 : 0;
!               sndbuf_dmasetup(sb->pch.buffer, pprio? sb->drq1 : NULL);
                sb->pch.dch = pprio? 1 : 0;
                sndbuf_dmasetup(sb->rch.buffer, pprio? sb->drq2 : sb->drq1);
                sb->rch.dch = pprio? 2 : 1;
--- 554,562 ----
        sb_reset_dsp(sb);
  
        if (sb->bd_flags & BD_F_SB16X) {
+               /* full-duplex doesn't work! */
                pprio = sb->pch.run? 1 : 0;
!               sndbuf_dmasetup(sb->pch.buffer, pprio? sb->drq1 : sb->drq2);
                sb->pch.dch = pprio? 1 : 0;
                sndbuf_dmasetup(sb->rch.buffer, pprio? sb->drq2 : sb->drq1);
                sb->rch.dch = pprio? 2 : 1;


The vibra16X supports full duplex. I traced the Windows driver, and what is 
does is that it programs one DMA channel 8-bit, and the other 16-bit. There 
might be some kind of auto detection logic here, because it always uses 8-bit 
for playback, even if I play 16-bit sound ...

3) Missing PCI-ID:

*** sbc.orig    Wed Dec 25 21:34:52 2002
--- sbc.c       Wed Dec 25 18:53:21 2002
***************
*** 251,256 ****
--- 251,257 ----
  
        {0x81167316, "ESS ES1681"},                     /* ESS1681 */
        {0x02017316, "ESS ES1688"},                     /* ESS1688 */
+       {0x68097316, "ESS ES1688"},                     /* ESS1688 */
        {0x68187316, "ESS ES1868"},                     /* ESS1868 */
        {0x03007316, "ESS ES1869"},                     /* ESS1869 */
        {0x69187316, "ESS ES1869"},                     /* ESS1869 */


4) There is a bug regarding the setting of the SIMPLEX flag (which is 
currently not working!)

Several device drivers call "pcm_setflags" before "pcm_register". The problem 
is that "pcm_register" clears these flags. Maybe the device drivers should be 
changed, or maybe "pcm_register" should be changed.

Here is my solution: In the file "pcm/sound.c" in the function 
"pcm_register()" "d->flags = 0;" should be changed into:

        /* d->flags = 0; should be cleared by allocator of softc;
         * cannot clear this field here, because several devices set
         * this flag before calling pcm_register()
         */

For example see sb8.c:
pcm_setflags(dev, pcm_getflags(dev) | SD_F_SIMPLEX);


5) I have a problem with one of my soundcards. When I try to 
"cat /dev/dsp0.0", the samplerate is always 48000Hz instead of 8000Hz. Is 
this a bug in the driver? How can I fix it?

pcm0: <Intel ICH4 (82801DB)> port 0x1c00-0x1cff,0x18c0-0x18ff mem 
0xe0100c00-0xe0100dff,0xe0100800-0xe01008ff irq 10 at device 31.5 on pci0
pcm0: [GIANT-LOCKED]
pcm0: <Unknown AC97 Codec (id = 0x43585430)>


6) I have modified the sound driver so that it can be opened from within the 
kernel. This enables me to make a sound-bridge that moves data from an 
ISDN/telephony device to a sound device (full duplex). This works very well 
and the delay is very little. Are there any plans to make such support 
generic or what do you think?


Yours
  HPS