Fix channel 1, sort of. recording now generates IRQs and the output

appears to be the correct length, but quality of output has not yet
been tested. Also, full duplex audio (that is, playback on channel 1)
does not yet work. Two constants and I am there!

Obtained from:	major hints from ALSA

1 files changed, 8 insertions, 2 deletions

diff --git a/sys/dev/sound/pci/solo.c b/sys/dev/sound/pci/solo.c
index ccc5116..1353647 100644
--- a/sys/dev/sound/pci/solo.c
+++ b/sys/dev/sound/pci/solo.c
@@ -709,12 +709,18 @@ essmix_setrecsrc(snd_mixer *m, u_int32_t src)
 static int
 ess_dmasetup(struct ess_info *sc, int ch, u_int32_t base, u_int16_t cnt, int dir)
 {
+
+/*
+ * XXX -- the constants written to register 8 and b in the playback case
+ * are certainly not 0x00. But I don't know what they really are
+ */
 	KASSERT(ch == 1 || ch == 2, ("bad ch"));
 	sc->dmasz[ch - 1] = cnt;
 	if (ch == 1) {
-		port_wr(sc->vc, 0x8, 0x50 | (dir == PCMDIR_PLAY)? 0x04 : 0x02, 1); /* command */
+		port_wr(sc->vc, 0x8, dir == PCMDIR_PLAY? 0x00 : 0xc4, 1); /* command */
 		port_wr(sc->vc, 0xd, 0xff, 1); /* reset */
 		port_wr(sc->vc, 0xf, 0x01, 1); /* mask */
+		port_wr(sc->vc, 0xb, dir == PCMDIR_PLAY? 0x00 : 0x54, 1); /* mode */
 		port_wr(sc->vc, 0x0, base, 4);
 		port_wr(sc->vc, 0x4, cnt, 2);
 
@@ -744,7 +750,7 @@ ess_dmatrigger(struct ess_info *sc, int ch, int go)
 {
 	KASSERT(ch == 1 || ch == 2, ("bad ch"));
 	if (ch == 1)
-		port_wr(sc->vc, 0xf, go? 0x01 : 0x00, 1); /* mask */
+		port_wr(sc->vc, 0xf, go? 0x00 : 0x01, 1); /* mask */
 	else if (ch == 2)
 		port_wr(sc->io, 0x6, 0x08 | (go? 0x02 : 0x00), 1); /* autoinit */
 	return 0;