probably/hopefully fixed auich driver for many systems: note that the codec_access_semaphore will NOT be reset on certain systems if an illegal/non existing register was accessed before. We need to do the access anyway to get such systems working. I hope someone can create a more extensive entry in syslog here to identify the offending access as that can be done AFAICT. This will help improve the driver later on because the non-existing-register accesses could be removed. Anyhow: I think the change I just did should remain, as it makes the driver more fault-tolerant to faults in itself (or hardware). Lowered the wait time to 1mS btw as that should suffice big_time. My Dell Inspiron C610 laptop with ICH3-M and Crystal CS4205 now works perfectly. No sound at all before.. Fix found in Linux-2.6.17/sound/pci/intel8x0m.c. Hope no-one is offended by my 'intrusion' here. Bye.

git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@25852 a95241bf-73f2-0310-859d-f6bbb57e9c96

src/add-ons/kernel/drivers/audio/ac97/auich/io.c

@@ -108,13 +108,19 @@ static int
 auich_codec_wait(device_config *config)
 {
 	int i;
-	for (i = 0; i < 1100; i++) {
+	/* Anyone holding a semaphore for 1 msec should be 'shot'... */
+	for (i = 0; i < 200; i++) {
 		if ((auich_reg_read_8(config, AUICH_REG_ACC_SEMA) & 0x01) == 0)
 			return B_OK;
 		if (i > 100)
 			snooze(10);
 	}
-	return B_TIMED_OUT;
+	/* access to some forbidden (non existant) ac97 registers will not
+	 * reset the semaphore. So even if you don't get the semaphore, still
+	 * continue the access. We don't really need the semaphore anyway. */ 
+	PRINT(("codec semaphore timed out!\n"));
+
+	return B_OK;
 }
 
 uint16