blob: 54daca4f57b4321aa29b541641f0763c3a7d50df [file] [log] [blame]
/*
**********************************************************************
* cardwo.c - PCM output HAL for emu10k1 driver
* Copyright 1999, 2000 Creative Labs, Inc.
*
**********************************************************************
*
* Date Author Summary of changes
* ---- ------ ------------------
* October 20, 1999 Bertrand Lee base code release
*
**********************************************************************
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 of
* the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the Free
* Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139,
* USA.
*
**********************************************************************
*/
#include <linux/poll.h>
#include "hwaccess.h"
#include "8010.h"
#include "voicemgr.h"
#include "cardwo.h"
#include "audio.h"
static u32 samplerate_to_linearpitch(u32 samplingrate)
{
samplingrate = (samplingrate << 8) / 375;
return (samplingrate >> 1) + (samplingrate & 1);
}
static void query_format(struct emu10k1_wavedevice *wave_dev, struct wave_format *wave_fmt)
{
int i, j, do_passthrough = 0, is_ac3 = 0;
struct emu10k1_card *card = wave_dev->card;
struct woinst *woinst = wave_dev->woinst;
if ((wave_fmt->channels > 2) && (wave_fmt->id != AFMT_S16_LE) && (wave_fmt->id != AFMT_U8))
wave_fmt->channels = 2;
if ((wave_fmt->channels < 1) || (wave_fmt->channels > WAVEOUT_MAXVOICES))
wave_fmt->channels = 2;
if (wave_fmt->channels == 2)
woinst->num_voices = 1;
else
woinst->num_voices = wave_fmt->channels;
if (wave_fmt->samplingrate >= 0x2ee00)
wave_fmt->samplingrate = 0x2ee00;
wave_fmt->passthrough = 0;
do_passthrough = is_ac3 = 0;
if (card->pt.selected)
do_passthrough = 1;
switch (wave_fmt->id) {
case AFMT_S16_LE:
wave_fmt->bitsperchannel = 16;
break;
case AFMT_U8:
wave_fmt->bitsperchannel = 8;
break;
case AFMT_AC3:
do_passthrough = 1;
is_ac3 = 1;
break;
default:
wave_fmt->id = AFMT_S16_LE;
wave_fmt->bitsperchannel = 16;
break;
}
if (do_passthrough) {
/* currently only one waveout instance may use pass-through */
if (woinst->state != WAVE_STATE_CLOSED ||
card->pt.state != PT_STATE_INACTIVE ||
(wave_fmt->samplingrate != 48000 && !is_ac3)) {
DPF(2, "unable to set pass-through mode\n");
} else if (USE_PT_METHOD1) {
i = emu10k1_find_control_gpr(&card->mgr, card->pt.patch_name, card->pt.intr_gpr_name);
j = emu10k1_find_control_gpr(&card->mgr, card->pt.patch_name, card->pt.enable_gpr_name);
if (i < 0 || j < 0)
DPF(2, "unable to set pass-through mode\n");
else {
wave_fmt->samplingrate = 48000;
wave_fmt->channels = 2;
card->pt.pos_gpr = emu10k1_find_control_gpr(&card->mgr, card->pt.patch_name,
card->pt.pos_gpr_name);
wave_fmt->passthrough = 1;
card->pt.intr_gpr = i;
card->pt.enable_gpr = j;
card->pt.state = PT_STATE_INACTIVE;
DPD(2, "is_ac3 is %d\n", is_ac3);
card->pt.ac3data = is_ac3;
wave_fmt->bitsperchannel = 16;
}
}else{
DPF(2, "Using Passthrough Method 2\n");
card->pt.enable_gpr = emu10k1_find_control_gpr(&card->mgr, card->pt.patch_name,
card->pt.enable_gpr_name);
wave_fmt->passthrough = 2;
wave_fmt->bitsperchannel = 16;
}
}
wave_fmt->bytesperchannel = wave_fmt->bitsperchannel >> 3;
wave_fmt->bytespersample = wave_fmt->channels * wave_fmt->bytesperchannel;
wave_fmt->bytespersec = wave_fmt->bytespersample * wave_fmt->samplingrate;
if (wave_fmt->channels == 2)
wave_fmt->bytespervoicesample = wave_fmt->channels * wave_fmt->bytesperchannel;
else
wave_fmt->bytespervoicesample = wave_fmt->bytesperchannel;
}
static int get_voice(struct emu10k1_card *card, struct woinst *woinst, unsigned int voicenum)
{
struct emu_voice *voice = &woinst->voice[voicenum];
/* Allocate voices here, if no voices available, return error. */
voice->usage = VOICE_USAGE_PLAYBACK;
voice->flags = 0;
if (woinst->format.channels == 2)
voice->flags |= VOICE_FLAGS_STEREO;
if (woinst->format.bitsperchannel == 16)
voice->flags |= VOICE_FLAGS_16BIT;
if (emu10k1_voice_alloc(card, voice) < 0) {
voice->usage = VOICE_USAGE_FREE;
return -1;
}
/* Calculate pitch */
voice->initial_pitch = (u16) (srToPitch(woinst->format.samplingrate) >> 8);
voice->pitch_target = samplerate_to_linearpitch(woinst->format.samplingrate);
DPD(2, "Initial pitch --> %#x\n", voice->initial_pitch);
voice->startloop = (voice->mem.emupageindex << 12) /
woinst->format.bytespervoicesample;
voice->endloop = voice->startloop + woinst->buffer.size / woinst->format.bytespervoicesample;
voice->start = voice->startloop;
voice->params[0].volume_target = 0xffff;
voice->params[0].initial_fc = 0xff;
voice->params[0].initial_attn = 0x00;
voice->params[0].byampl_env_sustain = 0x7f;
voice->params[0].byampl_env_decay = 0x7f;
if (voice->flags & VOICE_FLAGS_STEREO) {
if (woinst->format.passthrough == 2) {
voice->params[0].send_routing = voice->params[1].send_routing = card->waveout.send_routing[ROUTE_PT];
voice->params[0].send_routing2 = voice->params[1].send_routing2 = card->waveout.send_routing2[ROUTE_PT];
voice->params[0].send_dcba = 0xff;
voice->params[1].send_dcba = 0xff00;
voice->params[0].send_hgfe = voice->params[1].send_hgfe=0;
} else {
voice->params[0].send_dcba = card->waveout.send_dcba[SEND_LEFT];
voice->params[0].send_hgfe = card->waveout.send_hgfe[SEND_LEFT];
voice->params[1].send_dcba = card->waveout.send_dcba[SEND_RIGHT];
voice->params[1].send_hgfe = card->waveout.send_hgfe[SEND_RIGHT];
if (woinst->device) {
// /dev/dps1
voice->params[0].send_routing = voice->params[1].send_routing = card->waveout.send_routing[ROUTE_PCM1];
voice->params[0].send_routing2 = voice->params[1].send_routing2 = card->waveout.send_routing2[ROUTE_PCM1];
} else {
voice->params[0].send_routing = voice->params[1].send_routing = card->waveout.send_routing[ROUTE_PCM];
voice->params[0].send_routing2 = voice->params[1].send_routing2 = card->waveout.send_routing2[ROUTE_PCM];
}
}
voice->params[1].volume_target = 0xffff;
voice->params[1].initial_fc = 0xff;
voice->params[1].initial_attn = 0x00;
voice->params[1].byampl_env_sustain = 0x7f;
voice->params[1].byampl_env_decay = 0x7f;
} else {
if (woinst->num_voices > 1) {
// Multichannel pcm
voice->params[0].send_dcba=0xff;
voice->params[0].send_hgfe=0;
if (card->is_audigy) {
voice->params[0].send_routing = 0x3f3f3f00 + card->mchannel_fx + voicenum;
voice->params[0].send_routing2 = 0x3f3f3f3f;
} else {
voice->params[0].send_routing = 0xfff0 + card->mchannel_fx + voicenum;
}
} else {
voice->params[0].send_dcba = card->waveout.send_dcba[SEND_MONO];
voice->params[0].send_hgfe = card->waveout.send_hgfe[SEND_MONO];
if (woinst->device) {
voice->params[0].send_routing = card->waveout.send_routing[ROUTE_PCM1];
voice->params[0].send_routing2 = card->waveout.send_routing2[ROUTE_PCM1];
} else {
voice->params[0].send_routing = card->waveout.send_routing[ROUTE_PCM];
voice->params[0].send_routing2 = card->waveout.send_routing2[ROUTE_PCM];
}
}
}
DPD(2, "voice: startloop=%#x, endloop=%#x\n", voice->startloop, voice->endloop);
emu10k1_voice_playback_setup(voice);
return 0;
}
int emu10k1_waveout_open(struct emu10k1_wavedevice *wave_dev)
{
struct emu10k1_card *card = wave_dev->card;
struct woinst *woinst = wave_dev->woinst;
struct waveout_buffer *buffer = &woinst->buffer;
unsigned int voicenum;
u16 delay;
DPF(2, "emu10k1_waveout_open()\n");
for (voicenum = 0; voicenum < woinst->num_voices; voicenum++) {
if (emu10k1_voice_alloc_buffer(card, &woinst->voice[voicenum].mem, woinst->buffer.pages) < 0) {
ERROR();
emu10k1_waveout_close(wave_dev);
return -1;
}
if (get_voice(card, woinst, voicenum) < 0) {
ERROR();
emu10k1_waveout_close(wave_dev);
return -1;
}
}
buffer->fill_silence = 0;
buffer->silence_bytes = 0;
buffer->silence_pos = 0;
buffer->hw_pos = 0;
buffer->free_bytes = woinst->buffer.size;
delay = (48000 * woinst->buffer.fragment_size) /
(woinst->format.samplingrate * woinst->format.bytespervoicesample);
emu10k1_timer_install(card, &woinst->timer, delay);
woinst->state = WAVE_STATE_OPEN;
return 0;
}
void emu10k1_waveout_close(struct emu10k1_wavedevice *wave_dev)
{
struct emu10k1_card *card = wave_dev->card;
struct woinst *woinst = wave_dev->woinst;
unsigned int voicenum;
DPF(2, "emu10k1_waveout_close()\n");
emu10k1_waveout_stop(wave_dev);
emu10k1_timer_uninstall(card, &woinst->timer);
for (voicenum = 0; voicenum < woinst->num_voices; voicenum++) {
emu10k1_voice_free(&woinst->voice[voicenum]);
emu10k1_voice_free_buffer(card, &woinst->voice[voicenum].mem);
}
woinst->state = WAVE_STATE_CLOSED;
}
void emu10k1_waveout_start(struct emu10k1_wavedevice *wave_dev)
{
struct emu10k1_card *card = wave_dev->card;
struct woinst *woinst = wave_dev->woinst;
struct pt_data *pt = &card->pt;
DPF(2, "emu10k1_waveout_start()\n");
if (woinst->format.passthrough == 2) {
emu10k1_pt_setup(wave_dev);
sblive_writeptr(card, (card->is_audigy ? A_GPR_BASE : GPR_BASE) + pt->enable_gpr, 0, 1);
pt->state = PT_STATE_PLAYING;
}
/* Actual start */
emu10k1_voices_start(woinst->voice, woinst->num_voices, woinst->total_played);
emu10k1_timer_enable(card, &woinst->timer);
woinst->state |= WAVE_STATE_STARTED;
}
int emu10k1_waveout_setformat(struct emu10k1_wavedevice *wave_dev, struct wave_format *format)
{
struct emu10k1_card *card = wave_dev->card;
struct woinst *woinst = wave_dev->woinst;
unsigned int voicenum;
u16 delay;
DPF(2, "emu10k1_waveout_setformat()\n");
if (woinst->state & WAVE_STATE_STARTED)
return -1;
query_format(wave_dev, format);
if (woinst->format.samplingrate != format->samplingrate ||
woinst->format.channels != format->channels ||
woinst->format.bitsperchannel != format->bitsperchannel) {
woinst->format = *format;
if (woinst->state == WAVE_STATE_CLOSED)
return 0;
emu10k1_timer_uninstall(card, &woinst->timer);
for (voicenum = 0; voicenum < woinst->num_voices; voicenum++) {
emu10k1_voice_free(&woinst->voice[voicenum]);
if (get_voice(card, woinst, voicenum) < 0) {
ERROR();
emu10k1_waveout_close(wave_dev);
return -1;
}
}
delay = (48000 * woinst->buffer.fragment_size) /
(woinst->format.samplingrate * woinst->format.bytespervoicesample);
emu10k1_timer_install(card, &woinst->timer, delay);
}
return 0;
}
void emu10k1_waveout_stop(struct emu10k1_wavedevice *wave_dev)
{
struct emu10k1_card *card = wave_dev->card;
struct woinst *woinst = wave_dev->woinst;
DPF(2, "emu10k1_waveout_stop()\n");
if (!(woinst->state & WAVE_STATE_STARTED))
return;
emu10k1_timer_disable(card, &woinst->timer);
/* Stop actual voices */
emu10k1_voices_stop(woinst->voice, woinst->num_voices);
emu10k1_waveout_update(woinst);
woinst->state &= ~WAVE_STATE_STARTED;
}
/**
* emu10k1_waveout_getxfersize -
*
* gives the total free bytes on the voice buffer, including silence bytes
* (basically: total_free_bytes = free_bytes + silence_bytes).
*
*/
void emu10k1_waveout_getxfersize(struct woinst *woinst, u32 *total_free_bytes)
{
struct waveout_buffer *buffer = &woinst->buffer;
int pending_bytes;
if (woinst->mmapped) {
*total_free_bytes = buffer->free_bytes;
return;
}
pending_bytes = buffer->size - buffer->free_bytes;
buffer->fill_silence = (pending_bytes < (signed) buffer->fragment_size * 2) ? 1 : 0;
if (pending_bytes > (signed) buffer->silence_bytes) {
*total_free_bytes = (buffer->free_bytes + buffer->silence_bytes);
} else {
*total_free_bytes = buffer->size;
buffer->silence_bytes = pending_bytes;
if (pending_bytes < 0) {
buffer->silence_pos = buffer->hw_pos;
buffer->silence_bytes = 0;
buffer->free_bytes = buffer->size;
DPF(1, "buffer underrun\n");
}
}
}
/**
* copy_block -
*
* copies a block of pcm data to a voice buffer.
* Notice that the voice buffer is actually a set of disjointed memory pages.
*
*/
static void copy_block(void **dst, u32 str, u8 __user *src, u32 len)
{
unsigned int pg;
unsigned int pgoff;
unsigned int k;
pg = str / PAGE_SIZE;
pgoff = str % PAGE_SIZE;
if (len > PAGE_SIZE - pgoff) {
k = PAGE_SIZE - pgoff;
if (__copy_from_user((u8 *)dst[pg] + pgoff, src, k))
return;
len -= k;
while (len > PAGE_SIZE) {
if (__copy_from_user(dst[++pg], src + k, PAGE_SIZE))
return;
k += PAGE_SIZE;
len -= PAGE_SIZE;
}
if (__copy_from_user(dst[++pg], src + k, len))
return;
} else
__copy_from_user((u8 *)dst[pg] + pgoff, src, len);
}
/**
* copy_ilv_block -
*
* copies a block of pcm data containing n interleaved channels to n mono voice buffers.
* Notice that the voice buffer is actually a set of disjointed memory pages.
*
*/
static void copy_ilv_block(struct woinst *woinst, u32 str, u8 __user *src, u32 len)
{
unsigned int pg;
unsigned int pgoff;
unsigned int voice_num;
struct emu_voice *voice = woinst->voice;
pg = str / PAGE_SIZE;
pgoff = str % PAGE_SIZE;
while (len) {
for (voice_num = 0; voice_num < woinst->num_voices; voice_num++) {
if (__copy_from_user((u8 *)(voice[voice_num].mem.addr[pg]) + pgoff, src, woinst->format.bytespervoicesample))
return;
src += woinst->format.bytespervoicesample;
}
len -= woinst->format.bytespervoicesample;
pgoff += woinst->format.bytespervoicesample;
if (pgoff >= PAGE_SIZE) {
pgoff = 0;
pg++;
}
}
}
/**
* fill_block -
*
* fills a set voice buffers with a block of a given sample.
*
*/
static void fill_block(struct woinst *woinst, u32 str, u8 data, u32 len)
{
unsigned int pg;
unsigned int pgoff;
unsigned int voice_num;
struct emu_voice *voice = woinst->voice;
unsigned int k;
pg = str / PAGE_SIZE;
pgoff = str % PAGE_SIZE;
if (len > PAGE_SIZE - pgoff) {
k = PAGE_SIZE - pgoff;
for (voice_num = 0; voice_num < woinst->num_voices; voice_num++)
memset((u8 *)voice[voice_num].mem.addr[pg] + pgoff, data, k);
len -= k;
while (len > PAGE_SIZE) {
pg++;
for (voice_num = 0; voice_num < woinst->num_voices; voice_num++)
memset(voice[voice_num].mem.addr[pg], data, PAGE_SIZE);
len -= PAGE_SIZE;
}
pg++;
for (voice_num = 0; voice_num < woinst->num_voices; voice_num++)
memset(voice[voice_num].mem.addr[pg], data, len);
} else {
for (voice_num = 0; voice_num < woinst->num_voices; voice_num++)
memset((u8 *)voice[voice_num].mem.addr[pg] + pgoff, data, len);
}
}
/**
* emu10k1_waveout_xferdata -
*
* copies pcm data to the voice buffer. Silence samples
* previously added to the buffer are overwritten.
*
*/
void emu10k1_waveout_xferdata(struct woinst *woinst, u8 __user *data, u32 *size)
{
struct waveout_buffer *buffer = &woinst->buffer;
struct voice_mem *mem = &woinst->voice[0].mem;
u32 sizetocopy, sizetocopy_now, start;
unsigned long flags;
sizetocopy = min_t(u32, buffer->size, *size);
*size = sizetocopy;
if (!sizetocopy)
return;
spin_lock_irqsave(&woinst->lock, flags);
start = (buffer->size + buffer->silence_pos - buffer->silence_bytes) % buffer->size;
if (sizetocopy > buffer->silence_bytes) {
buffer->silence_pos += sizetocopy - buffer->silence_bytes;
buffer->free_bytes -= sizetocopy - buffer->silence_bytes;
buffer->silence_bytes = 0;
} else
buffer->silence_bytes -= sizetocopy;
spin_unlock_irqrestore(&woinst->lock, flags);
sizetocopy_now = buffer->size - start;
if (sizetocopy > sizetocopy_now) {
sizetocopy -= sizetocopy_now;
if (woinst->num_voices > 1) {
copy_ilv_block(woinst, start, data, sizetocopy_now);
copy_ilv_block(woinst, 0, data + sizetocopy_now * woinst->num_voices, sizetocopy);
} else {
copy_block(mem->addr, start, data, sizetocopy_now);
copy_block(mem->addr, 0, data + sizetocopy_now, sizetocopy);
}
} else {
if (woinst->num_voices > 1)
copy_ilv_block(woinst, start, data, sizetocopy);
else
copy_block(mem->addr, start, data, sizetocopy);
}
}
/**
* emu10k1_waveout_fillsilence -
*
* adds samples of silence to the voice buffer so that we
* don't loop over stale pcm data.
*
*/
void emu10k1_waveout_fillsilence(struct woinst *woinst)
{
struct waveout_buffer *buffer = &woinst->buffer;
u32 sizetocopy, sizetocopy_now, start;
u8 filldata;
unsigned long flags;
sizetocopy = buffer->fragment_size;
if (woinst->format.bitsperchannel == 16)
filldata = 0x00;
else
filldata = 0x80;
spin_lock_irqsave(&woinst->lock, flags);
buffer->silence_bytes += sizetocopy;
buffer->free_bytes -= sizetocopy;
buffer->silence_pos %= buffer->size;
start = buffer->silence_pos;
buffer->silence_pos += sizetocopy;
spin_unlock_irqrestore(&woinst->lock, flags);
sizetocopy_now = buffer->size - start;
if (sizetocopy > sizetocopy_now) {
sizetocopy -= sizetocopy_now;
fill_block(woinst, start, filldata, sizetocopy_now);
fill_block(woinst, 0, filldata, sizetocopy);
} else {
fill_block(woinst, start, filldata, sizetocopy);
}
}
/**
* emu10k1_waveout_update -
*
* updates the position of the voice buffer hardware pointer (hw_pos)
* and the number of free bytes on the buffer (free_bytes).
* The free bytes _don't_ include silence bytes that may have been
* added to the buffer.
*
*/
void emu10k1_waveout_update(struct woinst *woinst)
{
u32 hw_pos;
u32 diff;
/* There is no actual start yet */
if (!(woinst->state & WAVE_STATE_STARTED)) {
hw_pos = woinst->buffer.hw_pos;
} else {
/* hw_pos in sample units */
hw_pos = sblive_readptr(woinst->voice[0].card, CCCA_CURRADDR, woinst->voice[0].num);
if(hw_pos < woinst->voice[0].start)
hw_pos += woinst->buffer.size / woinst->format.bytespervoicesample - woinst->voice[0].start;
else
hw_pos -= woinst->voice[0].start;
hw_pos *= woinst->format.bytespervoicesample;
}
diff = (woinst->buffer.size + hw_pos - woinst->buffer.hw_pos) % woinst->buffer.size;
woinst->total_played += diff;
woinst->buffer.free_bytes += diff;
woinst->buffer.hw_pos = hw_pos;
}