o Add patches to improve OSS chanel driver.

Submited by:	luigi

o add mpg123 into runtime dependencies list.

1 files changed, 1167 insertions, 0 deletions

diff --git a/net/asterisk16/files/patch-channels::chan_oss.c b/net/asterisk16/files/patch-channels::chan_oss.c
new file mode 100644
index 0000000..ef8cfc1
--- /dev/null
+++ b/net/asterisk16/files/patch-channels::chan_oss.c
@@ -0,0 +1,1167 @@
+
+$FreeBSD$
+
+--- channels/chan_oss.c
++++ channels/chan_oss.c
+@@ -13,6 +13,8 @@
+  *
+  * This program is free software, distributed under the terms of
+  * the GNU General Public License
++ *
++ * FreeBSD changes by Luigi Rizzo, 2005.04.18
+  */
+ 
+ #include <asterisk/lock.h>
+@@ -54,21 +56,30 @@
+ #endif
+ 
+ /* Lets use 160 sample frames, just like GSM.  */
+-#define FRAME_SIZE 160
++/* this corresponds to 20ms of audio. */
++#define FRAME_SIZE 160 // was 160
+ 
+-/* When you set the frame size, you have to come up with
+-   the right buffer format as well. */
++/*
++ * When you set the frame size, you have to come up with
++ * the right buffer format as well.
++ * OSS lets you define a 'block' size (which should be a power of 2,
++ * which power is specified in the lower 16 bits) and the number of
++ * blocks allowed in the buffer (to avoid that the queue grows too large).
++ * The latter is specified in the top 16 bits.
++ * We use a block of 64 bytes (0x6), 5 blocks make a frame each sample
++ * being 2 bytes, and we make room to store two buffers.
++ * XXX the '10' is magic
++ */
++
++#define N_BLOCKS	(buffersize * 5 * 2)
+ /* 5 64-byte frames = one frame */
+-#define BUFFER_FMT ((buffersize * 10) << 16) | (0x0006);
++#define BUFFER_FMT (N_BLOCKS << 16) | (0x0006);
+ 
+ /* Don't switch between read/write modes faster than every 300 ms */
+-#define MIN_SWITCH_TIME 600
++#define MIN_SWITCH_TIME 300
+ 
+-static struct timeval lasttime;
+ 
+ static int usecnt;
+-static int silencesuppression = 0;
+-static int silencethreshold = 1000;
+ 
+ 
+ AST_MUTEX_DEFINE_STATIC(usecnt_lock);
+@@ -78,16 +89,15 @@
+ static char *tdesc = "OSS Console Channel Driver";
+ static char *config = "oss.conf";
+ 
+-static char context[AST_MAX_EXTENSION] = "default";
++static char default_context[AST_MAX_EXTENSION] = "default";
+ static char language[MAX_LANGUAGE] = "";
+-static char exten[AST_MAX_EXTENSION] = "s";
++static char oss_exten[AST_MAX_EXTENSION] = "s";
+ 
+-static int hookstate=0;
+ 
+-static short silence[FRAME_SIZE] = {0, };
+ 
+ struct sound {
+ 	int ind;
++	char *desc;
+ 	short *data;
+ 	int datalen;
+ 	int samplen;
+@@ -96,136 +106,178 @@
+ };
+ 
+ static struct sound sounds[] = {
+-	{ AST_CONTROL_RINGING, ringtone, sizeof(ringtone)/2, 16000, 32000, 1 },
+-	{ AST_CONTROL_BUSY, busy, sizeof(busy)/2, 4000, 4000, 1 },
+-	{ AST_CONTROL_CONGESTION, busy, sizeof(busy)/2, 2000, 2000, 1 },
+-	{ AST_CONTROL_RING, ring10, sizeof(ring10)/2, 16000, 32000, 1 },
+-	{ AST_CONTROL_ANSWER, answer, sizeof(answer)/2, 2200, 0, 0 },
++	{ AST_CONTROL_RINGING, "RINGING", ringtone, sizeof(ringtone)/2, 16000, 32000, 1 },
++	{ AST_CONTROL_BUSY, "BUSY", busy, sizeof(busy)/2, 4000, 4000, 1 },
++	{ AST_CONTROL_CONGESTION, "CONGESTION", busy, sizeof(busy)/2, 2000, 2000, 1 },
++	{ AST_CONTROL_RING, "RING10", ring10, sizeof(ring10)/2, 16000, 32000, 1 },
++	{ AST_CONTROL_ANSWER, "ANSWER", answer, sizeof(answer)/2, 2200, 0, 0 },
++	{ -1, NULL, 0, 0, 0, 0 },	/* end marker */
+ };
+ 
+-/* Sound command pipe */
+-static int sndcmd[2];
++
+ 
+ static struct chan_oss_pvt {
+ 	/* We only have one OSS structure -- near sighted perhaps, but it
+ 	   keeps this driver as simple as possible -- as it should be. */
++    /*
++     * cursound indicates which in struct sound we play. -1 means nothing,
++     * any other value is a valid sound, in which case sampsent indicates
++     * the next sample to send in [0..samplen + silencelen]
++     * nosound is set to disable the audio data from the channel
++     * (so we can play the tones etc.).
++     */
++    int sndcmd[2]; /* Sound command pipe */
++    int cursound;	/* index of sound to send */
++    int sampsent;	/* # of sound samples sent	*/
++    int nosound;
++
++    int total_blocks;	/* total blocks in the output device */
++    int sounddev;
++    enum { M_UNSET, M_FULL, M_READ, M_WRITE } duplex;
++    int autoanswer;
++    int autohangup;
++    int hookstate;
++    struct timeval lasttime;	/* last setformat */
++
++    int silencesuppression;
++    int silencethreshold;
++	char device[64];	/* device to open */
++
++    pthread_t sthread;
++
+ 	struct ast_channel *owner;
+ 	char exten[AST_MAX_EXTENSION];
+ 	char context[AST_MAX_EXTENSION];
+-} oss;
++} oss = {
++	.cursound = -1,
++	.sounddev = -1,
++	.duplex = M_UNSET, /* XXX check this */
++	.autoanswer = 1,
++	.autohangup = 1,
++	.silencethreshold = 1000,
++};
+ 
+-static int time_has_passed(void)
++/*
++ * returns true if too early to switch
++ */
++static int too_early(struct chan_oss_pvt *o)
+ {
+ 	struct timeval tv;
+ 	int ms;
+ 	gettimeofday(&tv, NULL);
+-	ms = (tv.tv_sec - lasttime.tv_sec) * 1000 +
+-			(tv.tv_usec - lasttime.tv_usec) / 1000;
+-	if (ms > MIN_SWITCH_TIME)
++	ms = (tv.tv_sec - o->lasttime.tv_sec) * 1000 +
++			(tv.tv_usec - o->lasttime.tv_usec) / 1000;
++	if (ms < MIN_SWITCH_TIME)
+ 		return -1;
+ 	return 0;
+ }
+ 
+-/* Number of buffers...  Each is FRAMESIZE/8 ms long.  For example
+-   with 160 sample frames, and a buffer size of 3, we have a 60ms buffer, 
+-   usually plenty. */
+-
+-static pthread_t sthread;
+-
+-#define MAX_BUFFER_SIZE 100
+-static int buffersize = 3;
+-
+-static int full_duplex = 0;
+-
+-/* Are we reading or writing (simulated full duplex) */
+-static int readmode = 1;
+-
+-/* File descriptor for sound device */
+-static int sounddev = -1;
+-
+-static int autoanswer = 1;
+- 
+-#if 0
+-static int calc_loudness(short *frame)
++/*
++ * Returns the number of blocks used in the audio output channel
++ */
++static int
++used_blocks(struct chan_oss_pvt *o)
+ {
+-	int sum = 0;
+-	int x;
+-	for (x=0;x<FRAME_SIZE;x++) {
+-		if (frame[x] < 0)
+-			sum -= frame[x];
+-		else
+-			sum += frame[x];
++    struct audio_buf_info info;
++    static int warned=0;
++    if (ioctl(o->sounddev, SNDCTL_DSP_GETOSPACE, &info)) {
++	if (!warned) {
++	    ast_log(LOG_WARNING, "Error reading output space\n");
++	    warned++;
+ 	}
+-	sum = sum/FRAME_SIZE;
+-	return sum;
++	return 1;
++    }
++    if (o->total_blocks == 0) {
++	ast_log(LOG_WARNING, "fragtotal %d size %d avail %d\n",
++		info.fragstotal,
++		info.fragsize,
++		info.fragments);
++	o->total_blocks = info.fragments;
++    }
++    return o->total_blocks - info.fragments;
+ }
+-#endif
+ 
+-static int cursound = -1;
+-static int sampsent = 0;
+-static int silencelen=0;
+-static int offset=0;
+-static int nosound=0;
++static int soundcard_writeframe(struct chan_oss_pvt *o, short *data)
++{	
++	/* Write an exactly FRAME_SIZE sized of frame */
++	int res;
++	static int errors = 0;
+ 
+-static int send_sound(void)
++	/*
++	 * nothing spectacular.
++	 * If the buffer is full just drop the extra, otherwise write
++	 */
++	res = used_blocks(o);
++	if (res > 10) {	/* no room to write a block */
++	    errors ++;
++	    if (errors == 0)
++	    	ast_log(LOG_WARNING, "write: used %d blocks (%d)\n", res, errors);
++	    return 0;
++	}
++	errors = 0;
++	res = write(o->sounddev, ((void *)data), FRAME_SIZE * 2);
++	return res;
++}
++
++/*
++ * handler for 'sound writable' events from the sound thread.
++ * Builds a frame from the high level description of the sounds,
++ * (tone+silence) and passes it to the audio device.
++ */
++static int send_sound(struct chan_oss_pvt *o)
+ {
+ 	short myframe[FRAME_SIZE];
+-	int total = FRAME_SIZE;
+-	short *frame = NULL;
+-	int amt=0;
+-	int res;
+-	int myoff;
+-	audio_buf_info abi;
+-	if (cursound > -1) {
+-		res = ioctl(sounddev, SNDCTL_DSP_GETOSPACE ,&abi);
+-		if (res) {
+-			ast_log(LOG_WARNING, "Unable to read output space\n");
+-			return -1;
+-		}
+-		/* Calculate how many samples we can send, max */
+-		if (total > (abi.fragments * abi.fragsize / 2)) 
+-			total = abi.fragments * abi.fragsize / 2;
+-		res = total;
+-		if (sampsent < sounds[cursound].samplen) {
+-			myoff=0;
+-			while(total) {
+-				amt = total;
+-				if (amt > (sounds[cursound].datalen - offset)) 
+-					amt = sounds[cursound].datalen - offset;
+-				memcpy(myframe + myoff, sounds[cursound].data + offset, amt * 2);
+-				total -= amt;
+-				offset += amt;
+-				sampsent += amt;
+-				myoff += amt;
+-				if (offset >= sounds[cursound].datalen)
+-					offset = 0;
+-			}
+-			/* Set it up for silence */
+-			if (sampsent >= sounds[cursound].samplen) 
+-				silencelen = sounds[cursound].silencelen;
+-			frame = myframe;
+-		} else {
+-			if (silencelen > 0) {
+-				frame = silence;
+-				silencelen -= res;
+-			} else {
+-				if (sounds[cursound].repeat) {
+-					/* Start over */
+-					sampsent = 0;
+-					offset = 0;
+-				} else {
+-					cursound = -1;
+-					nosound = 0;
+-				}
+-			}
++	int ofs = 0;
++	int l_sampsent = o->sampsent;
++	int l;
++	struct sound *s;
++
++	if (o->cursound < 0)	/* no sound to send */
++		return 0;
++	s = &sounds[o->cursound];
++	/*
++	 * prepare a frame
++	 */
++	
++	for (ofs = 0; ofs < FRAME_SIZE; ofs += l) {
++	    /* take chunks of sound and data until the buffer is full */
++	    l = s->samplen - l_sampsent;	/* sound available */
++	    if (l > 0) {
++		if (l > FRAME_SIZE - ofs)
++		    l = FRAME_SIZE - ofs;
++		if (0)
++		ast_log(LOG_WARNING, "send_sound sound %d/%d of %d into %d\n",
++			l_sampsent, l, s->samplen, ofs);
++		bcopy(s->data + l_sampsent, myframe + ofs, l*2);
++		l_sampsent += l;
++	    } else { /* no sound, maybe some silence */
++		static short silence[FRAME_SIZE] = {0, };
++
++	        l += s->silencelen;
++		if (l > 0) {
++		    if (l > FRAME_SIZE - ofs)
++			l = FRAME_SIZE - ofs;
++		    if (0)
++		    ast_log(LOG_WARNING, "send_sound silence %d/%d of %d into %d\n",
++			l_sampsent - s->samplen, l, s->silencelen, ofs);
++		    bcopy(silence, myframe + ofs, l*2);
++		    l_sampsent += l;
++		} else { /* silence is over, restart sound if loop */
++		    if (s->repeat == 0) {	/* last block */
++			ast_log(LOG_WARNING, "send_sound last block\n");
++			o->cursound = -1;
++			o->nosound = 0;	/* allow audio data */
++			if (ofs < FRAME_SIZE)	/* pad with silence */
++			    bcopy(silence, myframe + ofs, (FRAME_SIZE - ofs)*2);
++		    }
++		    l_sampsent = 0;
+ 		}
+-		if (frame)
+-			res = write(sounddev, frame, res * 2);
+-		if (res > 0)
+-			return 0;
+-		return res;
++	    }
+ 	}
+-	return 0;
++	l = soundcard_writeframe(o, myframe);
++	if (l > 0)
++	    o->sampsent = l_sampsent;	/* update status */
++	return 0;	/* fake success */
+ }
+ 
+ static void *sound_thread(void *unused)
+@@ -235,41 +287,53 @@
+ 	int max;
+ 	int res;
+ 	char ign[4096];
+-	if (read(sounddev, ign, sizeof(sounddev)) < 0)
++	if (read(oss.sounddev, ign, sizeof(ign)) < 0)
+ 		ast_log(LOG_WARNING, "Read error on sound device: %s\n", strerror(errno));
+ 	for(;;) {
+ 		FD_ZERO(&rfds);
+ 		FD_ZERO(&wfds);
+-		max = sndcmd[0];
+-		FD_SET(sndcmd[0], &rfds);
++		max = oss.sndcmd[0];
++		FD_SET(oss.sndcmd[0], &rfds);
+ 		if (!oss.owner) {
+-			FD_SET(sounddev, &rfds);
+-			if (sounddev > max)
+-				max = sounddev;
++			FD_SET(oss.sounddev, &rfds);
++			if (oss.sounddev > max)
++				max = oss.sounddev;
+ 		}
+-		if (cursound > -1) {
+-			FD_SET(sounddev, &wfds);
+-			if (sounddev > max)
+-				max = sounddev;
++		if (oss.cursound > -1) {
++			FD_SET(oss.sounddev, &wfds);
++			if (oss.sounddev > max)
++				max = oss.sounddev;
+ 		}
+ 		res = ast_select(max + 1, &rfds, &wfds, NULL, NULL);
+ 		if (res < 1) {
+ 			ast_log(LOG_WARNING, "select failed: %s\n", strerror(errno));
+ 			continue;
+ 		}
+-		if (FD_ISSET(sndcmd[0], &rfds)) {
+-			read(sndcmd[0], &cursound, sizeof(cursound));
+-			silencelen = 0;
+-			offset = 0;
+-			sampsent = 0;
++		if (FD_ISSET(oss.sndcmd[0], &rfds)) { /* read which sound to play from the pipe */
++			int i, what;
++
++			read(oss.sndcmd[0], &what, sizeof(what));
++			for (i = 0; sounds[i].ind != -1; i++)
++				if (sounds[i].ind == what) {
++					oss.cursound = i;
++					oss.sampsent = 0;
++					oss.nosound = 1; /* block other audio */
++					ast_log(LOG_WARNING, "play %s\n", sounds[i].desc);
++					break;
++				}
++			if (sounds[i].ind == -1)
++				oss.cursound = -1;
++			ast_log(LOG_WARNING, "cursound %d samplen %d silencelen %d\n",
++				oss.cursound, oss.cursound >=0 ? sounds[oss.cursound].samplen : -1,
++				oss.cursound >=0 ?  sounds[oss.cursound].silencelen : -1);
+ 		}
+-		if (FD_ISSET(sounddev, &rfds)) {
++		if (FD_ISSET(oss.sounddev, &rfds)) {
+ 			/* Ignore read */
+-			if (read(sounddev, ign, sizeof(ign)) < 0)
++			if (read(oss.sounddev, ign, sizeof(ign)) < 0)
+ 				ast_log(LOG_WARNING, "Read error on sound device: %s\n", strerror(errno));
+ 		}
+-		if (FD_ISSET(sounddev, &wfds))
+-			if (send_sound())
++		if (FD_ISSET(oss.sounddev, &wfds))
++			if (send_sound(&oss) < 0)
+ 				ast_log(LOG_WARNING, "Failed to write sound\n");
+ 	}
+ 	/* Never reached */
+@@ -277,6 +341,20 @@
+ }
+ 
+ #if 0
++static int calc_loudness(short *frame)
++{
++	int sum = 0;
++	int x;
++	for (x=0;x<FRAME_SIZE;x++) {
++		if (frame[x] < 0)
++			sum -= frame[x];
++		else
++			sum += frame[x];
++	}
++	sum = sum/FRAME_SIZE;
++	return sum;
++}
++
+ static int silence_suppress(short *buf)
+ {
+ #define SILBUF 3
+@@ -284,7 +362,7 @@
+ 	static int silentframes = 0;
+ 	static char silbuf[FRAME_SIZE * 2 * SILBUF];
+ 	static int silbufcnt=0;
+-	if (!silencesuppression)
++	if (!oss.silencesuppression)
+ 		return 0;
+ 	loudness = calc_loudness((short *)(buf));
+ 	if (option_debug)
+@@ -309,7 +387,7 @@
+ 		/* Write any buffered silence we have, it may have something
+ 		   important */
+ 		if (silbufcnt) {
+-			write(sounddev, silbuf, silbufcnt * FRAME_SIZE);
++			write(oss.sounddev, silbuf, silbufcnt * FRAME_SIZE);
+ 			silbufcnt = 0;
+ 		}
+ 	}
+@@ -317,27 +395,55 @@
+ }
+ #endif
+ 
+-static int setformat(void)
++/*
++ * reset and close the device if opened,
++ * then open and initialize it in the desired mode,
++ * trigger reads and writes so we can start using it.
++ */
++static int setformat(struct chan_oss_pvt *o, int mode)
+ {
+-	int fmt, desired, res, fd = sounddev;
++	int fmt, desired, res, fd;
+ 	static int warnedalready = 0;
+ 	static int warnedalready2 = 0;
++
++	if (o->sounddev >= 0) {
++		ioctl(o->sounddev, SNDCTL_DSP_RESET, 0);
++		close(o->sounddev);
++		o->duplex = M_UNSET;
++	}
++	fd = o->sounddev = open(o->device, mode |O_NONBLOCK);
++	if (o->sounddev < 0) {
++		ast_log(LOG_WARNING, "Unable to re-open DSP device: %s\n",
++			strerror(errno));
++		return -1;
++	}
++
++	gettimeofday(&o->lasttime, NULL);
+ 	fmt = AFMT_S16_LE;
+ 	res = ioctl(fd, SNDCTL_DSP_SETFMT, &fmt);
+ 	if (res < 0) {
+ 		ast_log(LOG_WARNING, "Unable to set format to 16-bit signed\n");
+ 		return -1;
+ 	}
+-	res = ioctl(fd, SNDCTL_DSP_SETDUPLEX, 0);
+-	
+-	/* Check to see if duplex set (FreeBSD Bug)*/
+-	res = ioctl(fd, SNDCTL_DSP_GETCAPS, &fmt);
+-	
+-	if ((fmt & DSP_CAP_DUPLEX) && !res) {
+-		if (option_verbose > 1) 
++	switch (mode) {
++	case O_RDWR:
++	    res = ioctl(fd, SNDCTL_DSP_SETDUPLEX, 0);
++	    /* Check to see if duplex set (FreeBSD Bug)*/
++	    res = ioctl(fd, SNDCTL_DSP_GETCAPS, &fmt);
++	    if (res == 0 && (fmt & DSP_CAP_DUPLEX)) {
++		    if (option_verbose > 1) 
+ 			ast_verbose(VERBOSE_PREFIX_2 "Console is full duplex\n");
+-		full_duplex = -1;
++		    o->duplex = M_FULL;
++	    };
++	    break;
++	case O_WRONLY:
++	    o->duplex = M_WRITE;
++	    break;
++	case O_RDONLY:
++	    o->duplex = M_READ;
++	    break;
+ 	}
++
+ 	fmt = 0;
+ 	res = ioctl(fd, SNDCTL_DSP_STEREO, &fmt);
+ 	if (res < 0) {
+@@ -348,6 +454,7 @@
+ 	desired = 8000;
+ 	fmt = desired;
+ 	res = ioctl(fd, SNDCTL_DSP_SPEED, &fmt);
++
+ 	if (res < 0) {
+ 		ast_log(LOG_WARNING, "Failed to set audio device to mono\n");
+ 		return -1;
+@@ -357,89 +464,54 @@
+ 			ast_log(LOG_WARNING, "Requested %d Hz, got %d Hz -- sound may be choppy\n", desired, fmt);
+ 	}
+ #if 1
+-	fmt = BUFFER_FMT;
++	/*
++	 * on freebsd, SETFRAGMENT does not work very well on some cards.
++	 * Better leave it out
++	 */
++	 
++	// fmt = BUFFER_FMT;
++	fmt = 0x8;	// 256-bytes fragment
+ 	res = ioctl(fd, SNDCTL_DSP_SETFRAGMENT, &fmt);
+ 	if (res < 0) {
+ 		if (!warnedalready2++)
+ 			ast_log(LOG_WARNING, "Unable to set fragment size -- sound may be choppy\n");
+ 	}
+ #endif
++	/* XXX on some cards, we need SNDCTL_DSP_SETTRIGGER to start outputting */
++	res = PCM_ENABLE_INPUT | PCM_ENABLE_OUTPUT;
++	res = ioctl(fd, SNDCTL_DSP_SETTRIGGER, &res);
++	/* it may fail if we are in half duplex, never mind */
+ 	return 0;
+ }
+ 
++/*
++ * make sure output mode is available. Returns 0 if done,
++ * 1 if too early to switch, -1 if error
++ */
+ static int soundcard_setoutput(int force)
+ {
+-	/* Make sure the soundcard is in output mode.  */
+-	int fd = sounddev;
+-	if (full_duplex || (!readmode && !force))
+-		return 0;
+-	readmode = 0;
+-	if (force || time_has_passed()) {
+-		ioctl(sounddev, SNDCTL_DSP_RESET, 0);
+-		/* Keep the same fd reserved by closing the sound device and copying stdin at the same
+-		   time. */
+-		/* dup2(0, sound); */ 
+-		close(sounddev);
+-		fd = open(DEV_DSP, O_WRONLY |O_NONBLOCK);
+-		if (fd < 0) {
+-			ast_log(LOG_WARNING, "Unable to re-open DSP device: %s\n", strerror(errno));
+-			return -1;
+-		}
+-		/* dup2 will close the original and make fd be sound */
+-		if (dup2(fd, sounddev) < 0) {
+-			ast_log(LOG_WARNING, "dup2() failed: %s\n", strerror(errno));
+-			return -1;
+-		}
+-		if (setformat()) {
+-			return -1;
+-		}
++	if (oss.duplex == M_FULL || (oss.duplex == M_WRITE && !force))
+ 		return 0;
+-	}
+-	return 1;
++	if (!force && too_early(&oss))
++		return 1;
++	if (setformat(&oss, O_WRONLY))
++		return -1;
++	return 0;
+ }
+ 
++/*
++ * make sure input mode is available. Returns 0 if done
++ * 1 if too early to switch, -1 if error
++ */
+ static int soundcard_setinput(int force)
+ {
+-	int fd = sounddev;
+-	if (full_duplex || (readmode && !force))
+-		return 0;
+-	readmode = -1;
+-	if (force || time_has_passed()) {
+-		ioctl(sounddev, SNDCTL_DSP_RESET, 0);
+-		close(sounddev);
+-		/* dup2(0, sound); */
+-		fd = open(DEV_DSP, O_RDONLY | O_NONBLOCK);
+-		if (fd < 0) {
+-			ast_log(LOG_WARNING, "Unable to re-open DSP device: %s\n", strerror(errno));
+-			return -1;
+-		}
+-		/* dup2 will close the original and make fd be sound */
+-		if (dup2(fd, sounddev) < 0) {
+-			ast_log(LOG_WARNING, "dup2() failed: %s\n", strerror(errno));
+-			return -1;
+-		}
+-		if (setformat()) {
+-			return -1;
+-		}
++	if (oss.duplex == M_FULL || (oss.duplex == M_READ && !force))
+ 		return 0;
+-	}
+-	return 1;
+-}
+-
+-static int soundcard_init(void)
+-{
+-	/* Assume it's full duplex for starters */
+-	int fd = open(DEV_DSP, 	O_RDWR | O_NONBLOCK);
+-	if (fd < 0) {
+-		ast_log(LOG_WARNING, "Unable to open %s: %s\n", DEV_DSP, strerror(errno));
+-		return fd;
+-	}
+-	gettimeofday(&lasttime, NULL);
+-	sounddev = fd;
+-	setformat();
+-	if (!full_duplex) 
+-		soundcard_setinput(1);
+-	return sounddev;
++	if (!force && too_early(&oss))
++		return 1;
++	if (setformat(&oss, O_RDONLY))
++		return -1;
++	return 0;
+ }
+ 
+ static int oss_digit(struct ast_channel *c, char digit)
+@@ -454,120 +526,81 @@
+ 	return 0;
+ }
+ 
++/* request to play a sound on the speaker */
++#define	RING(x) { int what = x; write(oss.sndcmd[1], &what, sizeof(what)); }
++
+ static int oss_call(struct ast_channel *c, char *dest, int timeout)
+ {
+-	int res = 3;
+ 	struct ast_frame f = { 0, };
+ 	ast_verbose( " << Call placed to '%s' on console >> \n", dest);
+-	if (autoanswer) {
++	if (oss.autoanswer) {
+ 		ast_verbose( " << Auto-answered >> \n" );
+ 		f.frametype = AST_FRAME_CONTROL;
+ 		f.subclass = AST_CONTROL_ANSWER;
+ 		ast_queue_frame(c, &f);
+ 	} else {
+-		nosound = 1;
+ 		ast_verbose( " << Type 'answer' to answer, or use 'autoanswer' for future calls >> \n");
+ 		f.frametype = AST_FRAME_CONTROL;
+ 		f.subclass = AST_CONTROL_RINGING;
+ 		ast_queue_frame(c, &f);
+-		write(sndcmd[1], &res, sizeof(res));
++		RING(AST_CONTROL_RING);
+ 	}
+ 	return 0;
+ }
+ 
+ static void answer_sound(void)
+ {
+-	int res;
+-	nosound = 1;
+-	res = 4;
+-	write(sndcmd[1], &res, sizeof(res));
+-	
++	RING(AST_CONTROL_ANSWER);
+ }
+ 
+ static int oss_answer(struct ast_channel *c)
+ {
+ 	ast_verbose( " << Console call has been answered >> \n");
+-	answer_sound();
++	answer_sound();	/* XXX do we really need it ? considering we shut down immediately... */
+ 	ast_setstate(c, AST_STATE_UP);
+-	cursound = -1;
+-	nosound=0;
++	oss.cursound = -1;
++	oss.nosound=0;
+ 	return 0;
+ }
+ 
+ static int oss_hangup(struct ast_channel *c)
+ {
+-	int res = 0;
+-	cursound = -1;
++	oss.cursound = -1;
+ 	c->pvt->pvt = NULL;
+ 	oss.owner = NULL;
+ 	ast_verbose( " << Hangup on console >> \n");
+ 	ast_mutex_lock(&usecnt_lock);
+ 	usecnt--;
+ 	ast_mutex_unlock(&usecnt_lock);
+-	if (hookstate) {
+-		if (autoanswer) {
++	if (oss.hookstate) {
++		if (oss.autoanswer || oss.autohangup) {
+ 			/* Assume auto-hangup too */
+-			hookstate = 0;
++			oss.hookstate = 0;
+ 		} else {
+ 			/* Make congestion noise */
+-			res = 2;
+-			write(sndcmd[1], &res, sizeof(res));
++			RING(AST_CONTROL_CONGESTION);
+ 		}
+ 	}
+ 	return 0;
+ }
+ 
+-static int soundcard_writeframe(short *data)
+-{	
+-	/* Write an exactly FRAME_SIZE sized of frame */
+-	static int bufcnt = 0;
+-	static short buffer[FRAME_SIZE * MAX_BUFFER_SIZE * 5];
+-	struct audio_buf_info info;
+-	int res;
+-	int fd = sounddev;
+-	static int warned=0;
+-	if (ioctl(fd, SNDCTL_DSP_GETOSPACE, &info)) {
+-		if (!warned)
+-			ast_log(LOG_WARNING, "Error reading output space\n");
+-		bufcnt = buffersize;
+-		warned++;
+-	}
+-	if ((info.fragments >= buffersize * 5) && (bufcnt == buffersize)) {
+-		/* We've run out of stuff, buffer again */
+-		bufcnt = 0;
+-	}
+-	if (bufcnt == buffersize) {
+-		/* Write sample immediately */
+-		res = write(fd, ((void *)data), FRAME_SIZE * 2);
+-	} else {
+-		/* Copy the data into our buffer */
+-		res = FRAME_SIZE * 2;
+-		memcpy(buffer + (bufcnt * FRAME_SIZE), data, FRAME_SIZE * 2);
+-		bufcnt++;
+-		if (bufcnt == buffersize) {
+-			res = write(fd, ((void *)buffer), FRAME_SIZE * 2 * buffersize);
+-		}
+-	}
+-	return res;
+-}
+-
+-
++/* used for data coming from the network */
+ static int oss_write(struct ast_channel *chan, struct ast_frame *f)
+ {
+ 	int res;
+-	static char sizbuf[8000];
+-	static int sizpos = 0;
+-	int len = sizpos;
+-	int pos;
++	int src;
++
++	// ast_log(LOG_WARNING, "oss_write size %d\n", f->datalen);
+ 	/* Immediately return if no sound is enabled */
+-	if (nosound)
++	if (oss.nosound)
+ 		return 0;
+ 	/* Stop any currently playing sound */
+-	cursound = -1;
+-	if (!full_duplex) {
++	oss.cursound = -1;
++	if (oss.duplex != M_FULL) {
++		/* XXX check this, looks weird! */
+ 		/* If we're half duplex, we have to switch to read mode
+ 		   to honor immediate needs if necessary */
+-		res = soundcard_setinput(1);
++		res = soundcard_setinput(1); /* force set if not full_duplex */
+ 		if (res < 0) {
+ 			ast_log(LOG_WARNING, "Unable to set device to input mode\n");
+ 			return -1;
+@@ -583,21 +616,30 @@
+ 		   so just pretend we wrote it */
+ 		return 0;
+ 	}
+-	/* We have to digest the frame in 160-byte portions */
+-	if (f->datalen > sizeof(sizbuf) - sizpos) {
+-		ast_log(LOG_WARNING, "Frame too large\n");
+-		return -1;
+-	}
+-	memcpy(sizbuf + sizpos, f->data, f->datalen);
+-	len += f->datalen;
+-	pos = 0;
+-	while(len - pos > FRAME_SIZE * 2) {
+-		soundcard_writeframe((short *)(sizbuf + pos));
+-		pos += FRAME_SIZE * 2;
++	/*
++	 * we could receive a sample which is not a multiple of our FRAME_SIZE,
++	 * so we buffer it locally and write to the device in FRAME_SIZE
++	 * chunks, keeping the residue stored for future use.
++	 */
++
++	src = 0; /* read position into f->data */
++	while ( src < f->datalen ) {
++	    static char buf[FRAME_SIZE*2];
++	    static int dst = 0;
++	    int l = sizeof(buf) - dst; /* how much room in the buffer */
++
++	    if (f->datalen - src >= l) {	/* enough to fill a frame */
++		memcpy(buf + dst, f->data + src, l);
++		soundcard_writeframe(&oss, (short *)buf);
++		src += l;
++		dst = 0;
++	    } else { /* copy residue */
++		l = f->datalen - src;
++		memcpy(buf + dst, f->data + src, l);
++		src += l;	/* but really, we are done */
++		dst += l;
++	    }
+ 	}
+-	if (len - pos) 
+-		memmove(sizbuf, sizbuf + pos, len - pos);
+-	sizpos = len - pos;
+ 	return 0;
+ }
+ 
+@@ -628,18 +670,15 @@
+ 		ast_log(LOG_WARNING, "Unable to set input mode\n");
+ 		return NULL;
+ 	}
+-	if (res > 0) {
++	if (res > 0) {	/* too early to switch ? */
+ 		/* Theoretically shouldn't happen, but anyway, return a NULL frame */
+ 		return &f;
+ 	}
+-	res = read(sounddev, buf + AST_FRIENDLY_OFFSET + readpos, FRAME_SIZE * 2 - readpos);
+-	if (res < 0) {
+-		ast_log(LOG_WARNING, "Error reading from sound device (If you're running 'artsd' then kill it): %s\n", strerror(errno));
+-#if 0
+-		CRASH;
+-#endif		
+-		return NULL;
+-	}
++
++	res = read(oss.sounddev, buf + AST_FRIENDLY_OFFSET + readpos, FRAME_SIZE * 2 - readpos);
++	// ast_log(LOG_WARNING, "oss_read() fd %d got %d\n", oss.sounddev, res);
++	if (res < 0) /* audio data not ready, return a NULL frame */
++		return &f;
+ 	readpos += res;
+ 	
+ 	if (readpos >= FRAME_SIZE * 2) {
+@@ -682,64 +721,66 @@
+ 	int res;
+ 	switch(cond) {
+ 	case AST_CONTROL_BUSY:
+-		res = 1;
+-		break;
+ 	case AST_CONTROL_CONGESTION:
+-		res = 2;
+-		break;
+ 	case AST_CONTROL_RINGING:
+-		res = 0;
++		res = cond;
+ 		break;
+ 	case -1:
+-		cursound = -1;
++		oss.cursound = -1;
+ 		return 0;
+ 	default:
+ 		ast_log(LOG_WARNING, "Don't know how to display condition %d on %s\n", cond, chan->name);
+ 		return -1;
+ 	}
+ 	if (res > -1) {
+-		write(sndcmd[1], &res, sizeof(res));
++		RING(res);
+ 	}
+ 	return 0;	
+ }
+ 
+-static struct ast_channel *oss_new(struct chan_oss_pvt *p, int state)
++static struct ast_channel *oss_new(struct chan_oss_pvt *oss, int state)
+ {
+ 	struct ast_channel *tmp;
++	struct ast_channel_pvt *pvt;
++
+ 	tmp = ast_channel_alloc(1);
+-	if (tmp) {
+-		snprintf(tmp->name, sizeof(tmp->name), "OSS/%s", DEV_DSP + 5);
+-		tmp->type = type;
+-		tmp->fds[0] = sounddev;
+-		tmp->nativeformats = AST_FORMAT_SLINEAR;
+-		tmp->pvt->pvt = p;
+-		tmp->pvt->send_digit = oss_digit;
+-		tmp->pvt->send_text = oss_text;
+-		tmp->pvt->hangup = oss_hangup;
+-		tmp->pvt->answer = oss_answer;
+-		tmp->pvt->read = oss_read;
+-		tmp->pvt->call = oss_call;
+-		tmp->pvt->write = oss_write;
+-		tmp->pvt->indicate = oss_indicate;
+-		tmp->pvt->fixup = oss_fixup;
+-		if (strlen(p->context))
+-			strncpy(tmp->context, p->context, sizeof(tmp->context)-1);
+-		if (strlen(p->exten))
+-			strncpy(tmp->exten, p->exten, sizeof(tmp->exten)-1);
+-		if (strlen(language))
+-			strncpy(tmp->language, language, sizeof(tmp->language)-1);
+-		p->owner = tmp;
+-		ast_setstate(tmp, state);
+-		ast_mutex_lock(&usecnt_lock);
+-		usecnt++;
+-		ast_mutex_unlock(&usecnt_lock);
+-		ast_update_use_count();
+-		if (state != AST_STATE_DOWN) {
+-			if (ast_pbx_start(tmp)) {
+-				ast_log(LOG_WARNING, "Unable to start PBX on %s\n", tmp->name);
+-				ast_hangup(tmp);
+-				tmp = NULL;
+-			}
++	if (tmp == NULL)
++		return NULL;
++	snprintf(tmp->name, sizeof(tmp->name), "OSS/%s", oss->device + 5);
++	tmp->type = type;
++	tmp->fds[0] = oss->sounddev;
++	tmp->nativeformats = AST_FORMAT_SLINEAR;
++	pvt = tmp->pvt;
++	pvt->pvt = oss;
++#if 1
++	pvt->send_digit = oss_digit;
++	pvt->send_text = oss_text;
++	pvt->hangup = oss_hangup;
++	pvt->answer = oss_answer;
++	pvt->read = oss_read;
++	pvt->call = oss_call;
++	pvt->write = oss_write;
++	pvt->indicate = oss_indicate;
++	pvt->fixup = oss_fixup;
++#endif
++	if (strlen(oss->context))
++		strncpy(tmp->context, oss->context, sizeof(tmp->context)-1);
++	if (strlen(oss->exten))
++		strncpy(tmp->exten, oss->exten, sizeof(tmp->exten)-1);
++	if (strlen(language))
++		strncpy(tmp->language, language, sizeof(tmp->language)-1);
++	oss->owner = tmp;
++	ast_setstate(tmp, state);
++	ast_mutex_lock(&usecnt_lock);
++	usecnt++;
++	ast_mutex_unlock(&usecnt_lock);
++	ast_update_use_count();
++	if (state != AST_STATE_DOWN) {
++		if (ast_pbx_start(tmp)) {
++			ast_log(LOG_WARNING, "Unable to start PBX on %s\n", tmp->name);
++			ast_hangup(tmp);
++			tmp = NULL;
++			/* XXX what about oss->owner and the channel itself ? */
+ 		}
+ 	}
+ 	return tmp;
+@@ -770,13 +811,13 @@
+ 	if ((argc != 1) && (argc != 2))
+ 		return RESULT_SHOWUSAGE;
+ 	if (argc == 1) {
+-		ast_cli(fd, "Auto answer is %s.\n", autoanswer ? "on" : "off");
++		ast_cli(fd, "Auto answer is %s.\n", oss.autoanswer ? "on" : "off");
+ 		return RESULT_SUCCESS;
+ 	} else {
+ 		if (!strcasecmp(argv[1], "on"))
+-			autoanswer = -1;
++			oss.autoanswer = -1;
+ 		else if (!strcasecmp(argv[1], "off"))
+-			autoanswer = 0;
++			oss.autoanswer = 0;
+ 		else
+ 			return RESULT_SHOWUSAGE;
+ 	}
+@@ -788,12 +829,14 @@
+ #ifndef MIN
+ #define MIN(a,b) ((a) < (b) ? (a) : (b))
+ #endif
++	int l = strlen(word);
++
+ 	switch(state) {
+ 	case 0:
+-		if (strlen(word) && !strncasecmp(word, "on", MIN(strlen(word), 2)))
++		if (l && !strncasecmp(word, "on", MIN(l, 2)))
+ 			return strdup("on");
+ 	case 1:
+-		if (strlen(word) && !strncasecmp(word, "off", MIN(strlen(word), 3)))
++		if (l && !strncasecmp(word, "off", MIN(l, 3)))
+ 			return strdup("off");
+ 	default:
+ 		return NULL;
+@@ -816,8 +859,8 @@
+ 		ast_cli(fd, "No one is calling us\n");
+ 		return RESULT_FAILURE;
+ 	}
+-	hookstate = 1;
+-	cursound = -1;
++	oss.hookstate = 1;
++	oss.cursound = -1;
+ 	ast_queue_frame(oss.owner, &f);
+ 	answer_sound();
+ 	return RESULT_SUCCESS;
+@@ -863,12 +906,12 @@
+ {
+ 	if (argc != 1)
+ 		return RESULT_SHOWUSAGE;
+-	cursound = -1;
+-	if (!oss.owner && !hookstate) {
++	oss.cursound = -1;
++	if (!oss.owner && !oss.hookstate) {
+ 		ast_cli(fd, "No call to hangup up\n");
+ 		return RESULT_FAILURE;
+ 	}
+-	hookstate = 0;
++	oss.hookstate = 0;
+ 	if (oss.owner) {
+ 		ast_queue_hangup(oss.owner);
+ 	}
+@@ -900,8 +943,8 @@
+ 		}
+ 		return RESULT_SUCCESS;
+ 	}
+-	mye = exten;
+-	myc = context;
++	mye = oss_exten;
++	myc = default_context;
+ 	if (argc == 2) {
+ 		char *stringp=NULL;
+ 		strncpy(tmp, argv[1], sizeof(tmp)-1);
+@@ -916,7 +959,7 @@
+ 	if (ast_exists_extension(NULL, myc, mye, 1, NULL)) {
+ 		strncpy(oss.exten, mye, sizeof(oss.exten)-1);
+ 		strncpy(oss.context, myc, sizeof(oss.context)-1);
+-		hookstate = 1;
++		oss.hookstate = 1;
+ 		oss_new(&oss, AST_STATE_RINGING);
+ 	} else
+ 		ast_cli(fd, "No such extension '%s' in context '%s'\n", mye, myc);
+@@ -974,21 +1017,47 @@
+ 	int res;
+ 	int x;
+ 	struct ast_config *cfg;
+-	struct ast_variable *v;
+-	res = pipe(sndcmd);
++
++	res = pipe(oss.sndcmd);
+ 	if (res) {
+ 		ast_log(LOG_ERROR, "Unable to create pipe\n");
+ 		return -1;
+ 	}
+-	res = soundcard_init();
+-	if (res < 0) {
++	/* load config file */
++	if ((cfg = ast_load(config))) {
++		struct ast_variable *v = ast_variable_browse(cfg, "general");
++		while(v) {
++			if (!strcasecmp(v->name, "autoanswer"))
++				oss.autoanswer = ast_true(v->value);
++			else if (!strcasecmp(v->name, "autohangup"))
++				oss.autohangup = ast_true(v->value);
++			else if (!strcasecmp(v->name, "oss.silencesuppression"))
++				oss.silencesuppression = ast_true(v->value);
++			else if (!strcasecmp(v->name, "silencethreshold"))
++				oss.silencethreshold = atoi(v->value);
++			else if (!strcasecmp(v->name, "device"))
++				strncpy(oss.device, v->value, sizeof(oss.device)-1);
++			else if (!strcasecmp(v->name, "context"))
++				strncpy(default_context, v->value, sizeof(default_context)-1);
++			else if (!strcasecmp(v->name, "language"))
++				strncpy(language, v->value, sizeof(language)-1);
++			else if (!strcasecmp(v->name, "extension"))
++				strncpy(oss_exten, v->value, sizeof(oss_exten)-1);
++			v=v->next;
++		}
++		ast_destroy(cfg);
++	}
++	if (!strlen(oss.device))
++		strncpy(oss.device, DEV_DSP, sizeof(oss.device)-1);
++	if (setformat(&oss, O_RDWR) < 0) {	/* open device */
+ 		if (option_verbose > 1) {
+ 			ast_verbose(VERBOSE_PREFIX_2 "No sound card detected -- console channel will be unavailable\n");
+ 			ast_verbose(VERBOSE_PREFIX_2 "Turn off OSS support by adding 'noload=chan_oss.so' in /etc/asterisk/modules.conf\n");
+ 		}
+ 		return 0;
+ 	}
+-	if (!full_duplex)
++	soundcard_setinput(1); /* force set if not full_duplex */
++	if (oss.duplex != M_FULL)
+ 		ast_log(LOG_WARNING, "XXX I don't work right with non-full duplex sound cards XXX\n");
+ 	res = ast_channel_register(type, tdesc, AST_FORMAT_SLINEAR, oss_request);
+ 	if (res < 0) {
+@@ -997,26 +1066,7 @@
+ 	}
+ 	for (x=0;x<sizeof(myclis)/sizeof(struct ast_cli_entry); x++)
+ 		ast_cli_register(myclis + x);
+-	if ((cfg = ast_load(config))) {
+-		v = ast_variable_browse(cfg, "general");
+-		while(v) {
+-			if (!strcasecmp(v->name, "autoanswer"))
+-				autoanswer = ast_true(v->value);
+-			else if (!strcasecmp(v->name, "silencesuppression"))
+-				silencesuppression = ast_true(v->value);
+-			else if (!strcasecmp(v->name, "silencethreshold"))
+-				silencethreshold = atoi(v->value);
+-			else if (!strcasecmp(v->name, "context"))
+-				strncpy(context, v->value, sizeof(context)-1);
+-			else if (!strcasecmp(v->name, "language"))
+-				strncpy(language, v->value, sizeof(language)-1);
+-			else if (!strcasecmp(v->name, "extension"))
+-				strncpy(exten, v->value, sizeof(exten)-1);
+-			v=v->next;
+-		}
+-		ast_destroy(cfg);
+-	}
+-	ast_pthread_create(&sthread, NULL, sound_thread, NULL);
++	ast_pthread_create(&oss.sthread, NULL, sound_thread, NULL);
+ 	return 0;
+ }
+ 
+@@ -1027,15 +1077,16 @@
+ 	int x;
+ 	for (x=0;x<sizeof(myclis)/sizeof(struct ast_cli_entry); x++)
+ 		ast_cli_unregister(myclis + x);
+-	close(sounddev);
+-	if (sndcmd[0] > 0) {
+-		close(sndcmd[0]);
+-		close(sndcmd[1]);
++	close(oss.sounddev);
++	if (oss.sndcmd[0] > 0) {
++		close(oss.sndcmd[0]);
++		close(oss.sndcmd[1]);
+ 	}
+ 	if (oss.owner)
+ 		ast_softhangup(oss.owner, AST_SOFTHANGUP_APPUNLOAD);
+ 	if (oss.owner)
+ 		return -1;
++	/* XXX what about the thread ? */
+ 	return 0;
+ }
+