linux/sound/oss/nm256_audio.c

1663 lines
42 KiB
C
Raw Normal View History

/*
* Audio driver for the NeoMagic 256AV and 256ZX chipsets in native
* mode, with AC97 mixer support.
*
* Overall design and parts of this code stolen from vidc_*.c and
* skeleton.c.
*
* Yeah, there are a lot of magic constants in here. You tell ME what
* they are. I just get this stuff psychically, remember?
*
* This driver was written by someone who wishes to remain anonymous.
* It is in the public domain, so share and enjoy. Try to make a profit
* off of it; go on, I dare you.
*
* Changes:
* 11-10-2000 Bartlomiej Zolnierkiewicz <bkz@linux-ide.org>
* Added some __init
* 19-04-2001 Marcus Meissner <mm@caldera.de>
* Ported to 2.4 PCI API.
*/
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/delay.h>
#include <linux/spinlock.h>
#include "sound_config.h"
static int nm256_debug;
static int force_load;
#include "nm256.h"
#include "nm256_coeff.h"
/*
* The size of the playback reserve. When the playback buffer has less
* than NM256_PLAY_WMARK_SIZE bytes to output, we request a new
* buffer.
*/
#define NM256_PLAY_WMARK_SIZE 512
static struct audio_driver nm256_audio_driver;
static int nm256_grabInterrupt (struct nm256_info *card);
static int nm256_releaseInterrupt (struct nm256_info *card);
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
static irqreturn_t nm256_interrupt (int irq, void *dev_id);
static irqreturn_t nm256_interrupt_zx (int irq, void *dev_id);
/* These belong in linux/pci.h. */
#define PCI_DEVICE_ID_NEOMAGIC_NM256AV_AUDIO 0x8005
#define PCI_DEVICE_ID_NEOMAGIC_NM256ZX_AUDIO 0x8006
#define PCI_DEVICE_ID_NEOMAGIC_NM256XL_PLUS_AUDIO 0x8016
/* List of cards. */
static struct nm256_info *nmcard_list;
/* Release the mapped-in memory for CARD. */
static void
nm256_release_ports (struct nm256_info *card)
{
int x;
for (x = 0; x < 2; x++) {
if (card->port[x].ptr != NULL) {
iounmap (card->port[x].ptr);
card->port[x].ptr = NULL;
}
}
}
/*
* Map in the memory ports for CARD, if they aren't already mapped in
* and have been configured. If successful, a zero value is returned;
* otherwise any previously mapped-in areas are released and a non-zero
* value is returned.
*
* This is invoked twice, once for each port. Ideally it would only be
* called once, but we now need to map in the second port in order to
* check how much memory the card has on the 256ZX.
*/
static int
nm256_remap_ports (struct nm256_info *card)
{
int x;
for (x = 0; x < 2; x++) {
if (card->port[x].ptr == NULL && card->port[x].end_offset > 0) {
u32 physaddr
= card->port[x].physaddr + card->port[x].start_offset;
u32 size
= card->port[x].end_offset - card->port[x].start_offset;
card->port[x].ptr = ioremap_nocache (physaddr, size);
if (card->port[x].ptr == NULL) {
printk (KERN_ERR "NM256: Unable to remap port %d\n", x + 1);
nm256_release_ports (card);
return -1;
}
}
}
return 0;
}
/* Locate the card in our list. */
static struct nm256_info *
nm256_find_card (int dev)
{
struct nm256_info *card;
for (card = nmcard_list; card != NULL; card = card->next_card)
if (card->dev[0] == dev || card->dev[1] == dev)
return card;
return NULL;
}
/*
* Ditto, but find the card struct corresponding to the mixer device DEV
* instead.
*/
static struct nm256_info *
nm256_find_card_for_mixer (int dev)
{
struct nm256_info *card;
for (card = nmcard_list; card != NULL; card = card->next_card)
if (card->mixer_oss_dev == dev)
return card;
return NULL;
}
static int usecache;
static int buffertop;
/* Check to see if we're using the bank of cached coefficients. */
static int
nm256_cachedCoefficients (struct nm256_info *card)
{
return usecache;
}
/* The actual rates supported by the card. */
static int samplerates[9] = {
8000, 11025, 16000, 22050, 24000, 32000, 44100, 48000, 99999999
};
/*
* Set the card samplerate, word size and stereo mode to correspond to
* the settings in the CARD struct for the specified device in DEV.
* We keep two separate sets of information, one for each device; the
* hardware is not actually configured until a read or write is
* attempted.
*/
static int
nm256_setInfo (int dev, struct nm256_info *card)
{
int x;
int w;
int targetrate;
if (card->dev[0] == dev)
w = 0;
else if (card->dev[1] == dev)
w = 1;
else
return -ENODEV;
targetrate = card->sinfo[w].samplerate;
if ((card->sinfo[w].bits != 8 && card->sinfo[w].bits != 16)
|| targetrate < samplerates[0]
|| targetrate > samplerates[7])
return -EINVAL;
for (x = 0; x < 8; x++)
if (targetrate < ((samplerates[x] + samplerates[x + 1]) / 2))
break;
if (x < 8) {
u8 ratebits = ((x << 4) & NM_RATE_MASK);
if (card->sinfo[w].bits == 16)
ratebits |= NM_RATE_BITS_16;
if (card->sinfo[w].stereo)
ratebits |= NM_RATE_STEREO;
card->sinfo[w].samplerate = samplerates[x];
if (card->dev_for_play == dev && card->playing) {
if (nm256_debug)
printk (KERN_DEBUG "Setting play ratebits to 0x%x\n",
ratebits);
nm256_loadCoefficient (card, 0, x);
nm256_writePort8 (card, 2,
NM_PLAYBACK_REG_OFFSET + NM_RATE_REG_OFFSET,
ratebits);
}
if (card->dev_for_record == dev && card->recording) {
if (nm256_debug)
printk (KERN_DEBUG "Setting record ratebits to 0x%x\n",
ratebits);
nm256_loadCoefficient (card, 1, x);
nm256_writePort8 (card, 2,
NM_RECORD_REG_OFFSET + NM_RATE_REG_OFFSET,
ratebits);
}
return 0;
}
else
return -EINVAL;
}
/* Start the play process going. */
static void
startPlay (struct nm256_info *card)
{
if (! card->playing) {
card->playing = 1;
if (nm256_grabInterrupt (card) == 0) {
nm256_setInfo (card->dev_for_play, card);
/* Enable playback engine and interrupts. */
nm256_writePort8 (card, 2, NM_PLAYBACK_ENABLE_REG,
NM_PLAYBACK_ENABLE_FLAG | NM_PLAYBACK_FREERUN);
/* Enable both channels. */
nm256_writePort16 (card, 2, NM_AUDIO_MUTE_REG, 0x0);
}
}
}
/*
* Request one chunk of AMT bytes from the recording device. When the
* operation is complete, the data will be copied into BUFFER and the
* function DMAbuf_inputintr will be invoked.
*/
static void
nm256_startRecording (struct nm256_info *card, char *buffer, u32 amt)
{
u32 endpos;
int enableEngine = 0;
u32 ringsize = card->recordBufferSize;
unsigned long flags;
if (amt > (ringsize / 2)) {
/*
* Of course this won't actually work right, because the
* caller is going to assume we will give what we got asked
* for.
*/
printk (KERN_ERR "NM256: Read request too large: %d\n", amt);
amt = ringsize / 2;
}
if (amt < 8) {
printk (KERN_ERR "NM256: Read request too small; %d\n", amt);
return;
}
spin_lock_irqsave(&card->lock,flags);
/*
* If we're not currently recording, set up the start and end registers
* for the recording engine.
*/
if (! card->recording) {
card->recording = 1;
if (nm256_grabInterrupt (card) == 0) {
card->curRecPos = 0;
nm256_setInfo (card->dev_for_record, card);
nm256_writePort32 (card, 2, NM_RBUFFER_START, card->abuf2);
nm256_writePort32 (card, 2, NM_RBUFFER_END,
card->abuf2 + ringsize);
nm256_writePort32 (card, 2, NM_RBUFFER_CURRP,
card->abuf2 + card->curRecPos);
enableEngine = 1;
}
else {
/* Not sure what else to do here. */
spin_unlock_irqrestore(&card->lock,flags);
return;
}
}
/*
* If we happen to go past the end of the buffer a bit (due to a
* delayed interrupt) it's OK. So might as well set the watermark
* right at the end of the data we want.
*/
endpos = card->abuf2 + ((card->curRecPos + amt) % ringsize);
card->recBuf = buffer;
card->requestedRecAmt = amt;
nm256_writePort32 (card, 2, NM_RBUFFER_WMARK, endpos);
/* Enable recording engine and interrupts. */
if (enableEngine)
nm256_writePort8 (card, 2, NM_RECORD_ENABLE_REG,
NM_RECORD_ENABLE_FLAG | NM_RECORD_FREERUN);
spin_unlock_irqrestore(&card->lock,flags);
}
/* Stop the play engine. */
static void
stopPlay (struct nm256_info *card)
{
/* Shut off sound from both channels. */
nm256_writePort16 (card, 2, NM_AUDIO_MUTE_REG,
NM_AUDIO_MUTE_LEFT | NM_AUDIO_MUTE_RIGHT);
/* Disable play engine. */
nm256_writePort8 (card, 2, NM_PLAYBACK_ENABLE_REG, 0);
if (card->playing) {
nm256_releaseInterrupt (card);
/* Reset the relevant state bits. */
card->playing = 0;
card->curPlayPos = 0;
}
}
/* Stop recording. */
static void
stopRecord (struct nm256_info *card)
{
/* Disable recording engine. */
nm256_writePort8 (card, 2, NM_RECORD_ENABLE_REG, 0);
if (card->recording) {
nm256_releaseInterrupt (card);
card->recording = 0;
card->curRecPos = 0;
}
}
/*
* Ring buffers, man. That's where the hip-hop, wild-n-wooly action's at.
* 1972? (Well, I suppose it was cheep-n-easy to implement.)
*
* Write AMT bytes of BUFFER to the playback ring buffer, and start the
* playback engine running. It will only accept up to 1/2 of the total
* size of the ring buffer. No check is made that we're about to overwrite
* the currently-playing sample.
*/
static void
nm256_write_block (struct nm256_info *card, char *buffer, u32 amt)
{
u32 ringsize = card->playbackBufferSize;
u32 endstop;
unsigned long flags;
if (amt > (ringsize / 2)) {
printk (KERN_ERR "NM256: Write request too large: %d\n", amt);
amt = (ringsize / 2);
}
if (amt < NM256_PLAY_WMARK_SIZE) {
printk (KERN_ERR "NM256: Write request too small: %d\n", amt);
return;
}
card->curPlayPos %= ringsize;
card->requested_amt = amt;
spin_lock_irqsave(&card->lock,flags);
if ((card->curPlayPos + amt) >= ringsize) {
u32 rem = ringsize - card->curPlayPos;
nm256_writeBuffer8 (card, buffer, 1,
card->abuf1 + card->curPlayPos,
rem);
if (amt > rem)
nm256_writeBuffer8 (card, buffer + rem, 1, card->abuf1,
amt - rem);
}
else
nm256_writeBuffer8 (card, buffer, 1,
card->abuf1 + card->curPlayPos,
amt);
/*
* Setup the start-n-stop-n-limit registers, and start that engine
* goin'.
*
* Normally we just let it wrap around to avoid the click-click
* action scene.
*/
if (! card->playing) {
/* The PBUFFER_END register in this case points to one sample
before the end of the buffer. */
int w = (card->dev_for_play == card->dev[0] ? 0 : 1);
int sampsize = (card->sinfo[w].bits == 16 ? 2 : 1);
if (card->sinfo[w].stereo)
sampsize *= 2;
/* Need to set the not-normally-changing-registers up. */
nm256_writePort32 (card, 2, NM_PBUFFER_START,
card->abuf1 + card->curPlayPos);
nm256_writePort32 (card, 2, NM_PBUFFER_END,
card->abuf1 + ringsize - sampsize);
nm256_writePort32 (card, 2, NM_PBUFFER_CURRP,
card->abuf1 + card->curPlayPos);
}
endstop = (card->curPlayPos + amt - NM256_PLAY_WMARK_SIZE) % ringsize;
nm256_writePort32 (card, 2, NM_PBUFFER_WMARK, card->abuf1 + endstop);
if (! card->playing)
startPlay (card);
spin_unlock_irqrestore(&card->lock,flags);
}
/* We just got a card playback interrupt; process it. */
static void
nm256_get_new_block (struct nm256_info *card)
{
/* Check to see how much got played so far. */
u32 amt = nm256_readPort32 (card, 2, NM_PBUFFER_CURRP) - card->abuf1;
if (amt >= card->playbackBufferSize) {
printk (KERN_ERR "NM256: Sound playback pointer invalid!\n");
amt = 0;
}
if (amt < card->curPlayPos)
amt = (card->playbackBufferSize - card->curPlayPos) + amt;
else
amt -= card->curPlayPos;
if (card->requested_amt > (amt + NM256_PLAY_WMARK_SIZE)) {
u32 endstop =
card->curPlayPos + card->requested_amt - NM256_PLAY_WMARK_SIZE;
nm256_writePort32 (card, 2, NM_PBUFFER_WMARK, card->abuf1 + endstop);
}
else {
card->curPlayPos += card->requested_amt;
/* Get a new block to write. This will eventually invoke
nm256_write_block () or stopPlay (). */
DMAbuf_outputintr (card->dev_for_play, 1);
}
}
/*
* Read the last-recorded block from the ring buffer, copy it into the
* saved buffer pointer, and invoke DMAuf_inputintr() with the recording
* device.
*/
static void
nm256_read_block (struct nm256_info *card)
{
/* Grab the current position of the recording pointer. */
u32 currptr = nm256_readPort32 (card, 2, NM_RBUFFER_CURRP) - card->abuf2;
u32 amtToRead = card->requestedRecAmt;
u32 ringsize = card->recordBufferSize;
if (currptr >= card->recordBufferSize) {
printk (KERN_ERR "NM256: Sound buffer record pointer invalid!\n");
currptr = 0;
}
/*
* This test is probably redundant; we shouldn't be here unless
* it's true.
*/
if (card->recording) {
/* If we wrapped around, copy everything from the start of our
recording buffer to the end of the buffer. */
if (currptr < card->curRecPos) {
u32 amt = min (ringsize - card->curRecPos, amtToRead);
nm256_readBuffer8 (card, card->recBuf, 1,
card->abuf2 + card->curRecPos,
amt);
amtToRead -= amt;
card->curRecPos += amt;
card->recBuf += amt;
if (card->curRecPos == ringsize)
card->curRecPos = 0;
}
if ((card->curRecPos < currptr) && (amtToRead > 0)) {
u32 amt = min (currptr - card->curRecPos, amtToRead);
nm256_readBuffer8 (card, card->recBuf, 1,
card->abuf2 + card->curRecPos, amt);
card->curRecPos = ((card->curRecPos + amt) % ringsize);
}
card->recBuf = NULL;
card->requestedRecAmt = 0;
DMAbuf_inputintr (card->dev_for_record);
}
}
/*
* Initialize the hardware.
*/
static void
nm256_initHw (struct nm256_info *card)
{
/* Reset everything. */
nm256_writePort8 (card, 2, 0x0, 0x11);
nm256_writePort16 (card, 2, 0x214, 0);
stopRecord (card);
stopPlay (card);
}
/*
* Handle a potential interrupt for the device referred to by DEV_ID.
*
* I don't like the cut-n-paste job here either between the two routines,
* but there are sufficient differences between the two interrupt handlers
* that parameterizing it isn't all that great either. (Could use a macro,
* I suppose...yucky bleah.)
*/
static irqreturn_t
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
nm256_interrupt (int irq, void *dev_id)
{
struct nm256_info *card = (struct nm256_info *)dev_id;
u16 status;
static int badintrcount;
int handled = 0;
if ((card == NULL) || (card->magsig != NM_MAGIC_SIG)) {
printk (KERN_ERR "NM256: Bad card pointer\n");
return IRQ_NONE;
}
status = nm256_readPort16 (card, 2, NM_INT_REG);
/* Not ours. */
if (status == 0) {
if (badintrcount++ > 1000) {
/*
* I'm not sure if the best thing is to stop the card from
* playing or just release the interrupt (after all, we're in
* a bad situation, so doing fancy stuff may not be such a good
* idea).
*
* I worry about the card engine continuing to play noise
* over and over, however--that could become a very
* obnoxious problem. And we know that when this usually
* happens things are fairly safe, it just means the user's
* inserted a PCMCIA card and someone's spamming us with IRQ 9s.
*/
handled = 1;
if (card->playing)
stopPlay (card);
if (card->recording)
stopRecord (card);
badintrcount = 0;
}
return IRQ_RETVAL(handled);
}
badintrcount = 0;
/* Rather boring; check for individual interrupts and process them. */
if (status & NM_PLAYBACK_INT) {
handled = 1;
status &= ~NM_PLAYBACK_INT;
NM_ACK_INT (card, NM_PLAYBACK_INT);
if (card->playing)
nm256_get_new_block (card);
}
if (status & NM_RECORD_INT) {
handled = 1;
status &= ~NM_RECORD_INT;
NM_ACK_INT (card, NM_RECORD_INT);
if (card->recording)
nm256_read_block (card);
}
if (status & NM_MISC_INT_1) {
u8 cbyte;
handled = 1;
status &= ~NM_MISC_INT_1;
printk (KERN_ERR "NM256: Got misc interrupt #1\n");
NM_ACK_INT (card, NM_MISC_INT_1);
nm256_writePort16 (card, 2, NM_INT_REG, 0x8000);
cbyte = nm256_readPort8 (card, 2, 0x400);
nm256_writePort8 (card, 2, 0x400, cbyte | 2);
}
if (status & NM_MISC_INT_2) {
u8 cbyte;
handled = 1;
status &= ~NM_MISC_INT_2;
printk (KERN_ERR "NM256: Got misc interrupt #2\n");
NM_ACK_INT (card, NM_MISC_INT_2);
cbyte = nm256_readPort8 (card, 2, 0x400);
nm256_writePort8 (card, 2, 0x400, cbyte & ~2);
}
/* Unknown interrupt. */
if (status) {
handled = 1;
printk (KERN_ERR "NM256: Fire in the hole! Unknown status 0x%x\n",
status);
/* Pray. */
NM_ACK_INT (card, status);
}
return IRQ_RETVAL(handled);
}
/*
* Handle a potential interrupt for the device referred to by DEV_ID.
* This handler is for the 256ZX, and is very similar to the non-ZX
* routine.
*/
static irqreturn_t
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
nm256_interrupt_zx (int irq, void *dev_id)
{
struct nm256_info *card = (struct nm256_info *)dev_id;
u32 status;
static int badintrcount;
int handled = 0;
if ((card == NULL) || (card->magsig != NM_MAGIC_SIG)) {
printk (KERN_ERR "NM256: Bad card pointer\n");
return IRQ_NONE;
}
status = nm256_readPort32 (card, 2, NM_INT_REG);
/* Not ours. */
if (status == 0) {
if (badintrcount++ > 1000) {
printk (KERN_ERR "NM256: Releasing interrupt, over 1000 invalid interrupts\n");
/*
* I'm not sure if the best thing is to stop the card from
* playing or just release the interrupt (after all, we're in
* a bad situation, so doing fancy stuff may not be such a good
* idea).
*
* I worry about the card engine continuing to play noise
* over and over, however--that could become a very
* obnoxious problem. And we know that when this usually
* happens things are fairly safe, it just means the user's
* inserted a PCMCIA card and someone's spamming us with
* IRQ 9s.
*/
handled = 1;
if (card->playing)
stopPlay (card);
if (card->recording)
stopRecord (card);
badintrcount = 0;
}
return IRQ_RETVAL(handled);
}
badintrcount = 0;
/* Rather boring; check for individual interrupts and process them. */
if (status & NM2_PLAYBACK_INT) {
handled = 1;
status &= ~NM2_PLAYBACK_INT;
NM2_ACK_INT (card, NM2_PLAYBACK_INT);
if (card->playing)
nm256_get_new_block (card);
}
if (status & NM2_RECORD_INT) {
handled = 1;
status &= ~NM2_RECORD_INT;
NM2_ACK_INT (card, NM2_RECORD_INT);
if (card->recording)
nm256_read_block (card);
}
if (status & NM2_MISC_INT_1) {
u8 cbyte;
handled = 1;
status &= ~NM2_MISC_INT_1;
printk (KERN_ERR "NM256: Got misc interrupt #1\n");
NM2_ACK_INT (card, NM2_MISC_INT_1);
cbyte = nm256_readPort8 (card, 2, 0x400);
nm256_writePort8 (card, 2, 0x400, cbyte | 2);
}
if (status & NM2_MISC_INT_2) {
u8 cbyte;
handled = 1;
status &= ~NM2_MISC_INT_2;
printk (KERN_ERR "NM256: Got misc interrupt #2\n");
NM2_ACK_INT (card, NM2_MISC_INT_2);
cbyte = nm256_readPort8 (card, 2, 0x400);
nm256_writePort8 (card, 2, 0x400, cbyte & ~2);
}
/* Unknown interrupt. */
if (status) {
handled = 1;
printk (KERN_ERR "NM256: Fire in the hole! Unknown status 0x%x\n",
status);
/* Pray. */
NM2_ACK_INT (card, status);
}
return IRQ_RETVAL(handled);
}
/*
* Request our interrupt.
*/
static int
nm256_grabInterrupt (struct nm256_info *card)
{
if (card->has_irq++ == 0) {
if (request_irq (card->irq, card->introutine, IRQF_SHARED,
"NM256_audio", card) < 0) {
printk (KERN_ERR "NM256: can't obtain IRQ %d\n", card->irq);
return -1;
}
}
return 0;
}
/*
* Release our interrupt.
*/
static int
nm256_releaseInterrupt (struct nm256_info *card)
{
if (card->has_irq <= 0) {
printk (KERN_ERR "nm256: too many calls to releaseInterrupt\n");
return -1;
}
card->has_irq--;
if (card->has_irq == 0) {
free_irq (card->irq, card);
}
return 0;
}
/*
* Waits for the mixer to become ready to be written; returns a zero value
* if it timed out.
*/
static int
nm256_isReady (struct ac97_hwint *dev)
{
struct nm256_info *card = (struct nm256_info *)dev->driver_private;
int t2 = 10;
u32 testaddr;
u16 testb;
int done = 0;
if (card->magsig != NM_MAGIC_SIG) {
printk (KERN_ERR "NM256: Bad magic signature in isReady!\n");
return 0;
}
testaddr = card->mixer_status_offset;
testb = card->mixer_status_mask;
/*
* Loop around waiting for the mixer to become ready.
*/
while (! done && t2-- > 0) {
if ((nm256_readPort16 (card, 2, testaddr) & testb) == 0)
done = 1;
else
udelay (100);
}
return done;
}
/*
* Return the contents of the AC97 mixer register REG. Returns a positive
* value if successful, or a negative error code.
*/
static int
nm256_readAC97Reg (struct ac97_hwint *dev, u8 reg)
{
struct nm256_info *card = (struct nm256_info *)dev->driver_private;
if (card->magsig != NM_MAGIC_SIG) {
printk (KERN_ERR "NM256: Bad magic signature in readAC97Reg!\n");
return -EINVAL;
}
if (reg < 128) {
int res;
nm256_isReady (dev);
res = nm256_readPort16 (card, 2, card->mixer + reg);
/* Magic delay. Bleah yucky. */
udelay (1000);
return res;
}
else
return -EINVAL;
}
/*
* Writes VALUE to AC97 mixer register REG. Returns 0 if successful, or
* a negative error code.
*/
static int
nm256_writeAC97Reg (struct ac97_hwint *dev, u8 reg, u16 value)
{
unsigned long flags;
int tries = 2;
int done = 0;
u32 base;
struct nm256_info *card = (struct nm256_info *)dev->driver_private;
if (card->magsig != NM_MAGIC_SIG) {
printk (KERN_ERR "NM256: Bad magic signature in writeAC97Reg!\n");
return -EINVAL;
}
base = card->mixer;
spin_lock_irqsave(&card->lock,flags);
nm256_isReady (dev);
/* Wait for the write to take, too. */
while ((tries-- > 0) && !done) {
nm256_writePort16 (card, 2, base + reg, value);
if (nm256_isReady (dev)) {
done = 1;
break;
}
}
spin_unlock_irqrestore(&card->lock,flags);
udelay (1000);
return ! done;
}
/*
* Initial register values to be written to the AC97 mixer.
* While most of these are identical to the reset values, we do this
* so that we have most of the register contents cached--this avoids
* reading from the mixer directly (which seems to be problematic,
* probably due to ignorance).
*/
struct initialValues
{
unsigned short port;
unsigned short value;
};
static struct initialValues nm256_ac97_initial_values[] =
{
{ AC97_MASTER_VOL_STEREO, 0x8000 },
{ AC97_HEADPHONE_VOL, 0x8000 },
{ AC97_MASTER_VOL_MONO, 0x0000 },
{ AC97_PCBEEP_VOL, 0x0000 },
{ AC97_PHONE_VOL, 0x0008 },
{ AC97_MIC_VOL, 0x8000 },
{ AC97_LINEIN_VOL, 0x8808 },
{ AC97_CD_VOL, 0x8808 },
{ AC97_VIDEO_VOL, 0x8808 },
{ AC97_AUX_VOL, 0x8808 },
{ AC97_PCMOUT_VOL, 0x0808 },
{ AC97_RECORD_SELECT, 0x0000 },
{ AC97_RECORD_GAIN, 0x0B0B },
{ AC97_GENERAL_PURPOSE, 0x0000 },
{ 0xffff, 0xffff }
};
/* Initialize the AC97 into a known state. */
static int
nm256_resetAC97 (struct ac97_hwint *dev)
{
struct nm256_info *card = (struct nm256_info *)dev->driver_private;
int x;
if (card->magsig != NM_MAGIC_SIG) {
printk (KERN_ERR "NM256: Bad magic signature in resetAC97!\n");
return -EINVAL;
}
/* Reset the mixer. 'Tis magic! */
nm256_writePort8 (card, 2, 0x6c0, 1);
// nm256_writePort8 (card, 2, 0x6cc, 0x87); /* This crashes Dell latitudes */
nm256_writePort8 (card, 2, 0x6cc, 0x80);
nm256_writePort8 (card, 2, 0x6cc, 0x0);
if (! card->mixer_values_init) {
for (x = 0; nm256_ac97_initial_values[x].port != 0xffff; x++) {
ac97_put_register (dev,
nm256_ac97_initial_values[x].port,
nm256_ac97_initial_values[x].value);
card->mixer_values_init = 1;
}
}
return 0;
}
/*
* We don't do anything particularly special here; it just passes the
* mixer ioctl to the AC97 driver.
*/
static int
nm256_default_mixer_ioctl (int dev, unsigned int cmd, void __user *arg)
{
struct nm256_info *card = nm256_find_card_for_mixer (dev);
if (card != NULL)
return ac97_mixer_ioctl (&(card->mdev), cmd, arg);
else
return -ENODEV;
}
static struct mixer_operations nm256_mixer_operations = {
.owner = THIS_MODULE,
.id = "NeoMagic",
.name = "NM256AC97Mixer",
.ioctl = nm256_default_mixer_ioctl
};
/*
* Default settings for the OSS mixer. These are set last, after the
* mixer is initialized.
*
* I "love" C sometimes. Got braces?
*/
static struct ac97_mixer_value_list mixer_defaults[] = {
{ SOUND_MIXER_VOLUME, { { 85, 85 } } },
{ SOUND_MIXER_SPEAKER, { { 100 } } },
{ SOUND_MIXER_PCM, { { 65, 65 } } },
{ SOUND_MIXER_CD, { { 65, 65 } } },
{ -1, { { 0, 0 } } }
};
/* Installs the AC97 mixer into CARD. */
static int __devinit
nm256_install_mixer (struct nm256_info *card)
{
int mixer;
card->mdev.reset_device = nm256_resetAC97;
card->mdev.read_reg = nm256_readAC97Reg;
card->mdev.write_reg = nm256_writeAC97Reg;
card->mdev.driver_private = (void *)card;
if (ac97_init (&(card->mdev)))
return -1;
mixer = sound_alloc_mixerdev();
if (num_mixers >= MAX_MIXER_DEV) {
printk ("NM256 mixer: Unable to alloc mixerdev\n");
return -1;
}
mixer_devs[mixer] = &nm256_mixer_operations;
card->mixer_oss_dev = mixer;
/* Some reasonable default values. */
ac97_set_values (&(card->mdev), mixer_defaults);
printk(KERN_INFO "Initialized AC97 mixer\n");
return 0;
}
/*
* See if the signature left by the NM256 BIOS is intact; if so, we use
* the associated address as the end of our audio buffer in the video
* RAM.
*/
static void __devinit
nm256_peek_for_sig (struct nm256_info *card)
{
u32 port1offset
= card->port[0].physaddr + card->port[0].end_offset - 0x0400;
/* The signature is located 1K below the end of video RAM. */
char __iomem *temp = ioremap_nocache (port1offset, 16);
/* Default buffer end is 5120 bytes below the top of RAM. */
u32 default_value = card->port[0].end_offset - 0x1400;
u32 sig;
/* Install the default value first, so we don't have to repeatedly
do it if there is a problem. */
card->port[0].end_offset = default_value;
if (temp == NULL) {
printk (KERN_ERR "NM256: Unable to scan for card signature in video RAM\n");
return;
}
sig = readl (temp);
if ((sig & NM_SIG_MASK) == NM_SIGNATURE) {
u32 pointer = readl (temp + 4);
/*
* If it's obviously invalid, don't use it (the port already has a
* suitable default value set).
*/
if (pointer != 0xffffffff)
card->port[0].end_offset = pointer;
printk (KERN_INFO "NM256: Found card signature in video RAM: 0x%x\n",
pointer);
}
iounmap (temp);
}
/*
* Install a driver for the PCI device referenced by PCIDEV.
* VERSTR is a human-readable version string.
*/
static int __devinit
nm256_install(struct pci_dev *pcidev, enum nm256rev rev, char *verstr)
{
struct nm256_info *card;
int x;
if (pci_enable_device(pcidev))
return 0;
card = kmalloc (sizeof (struct nm256_info), GFP_KERNEL);
if (card == NULL) {
printk (KERN_ERR "NM256: out of memory!\n");
return 0;
}
card->magsig = NM_MAGIC_SIG;
card->playing = 0;
card->recording = 0;
card->rev = rev;
spin_lock_init(&card->lock);
/* Init the memory port info. */
for (x = 0; x < 2; x++) {
card->port[x].physaddr = pci_resource_start (pcidev, x);
card->port[x].ptr = NULL;
card->port[x].start_offset = 0;
card->port[x].end_offset = 0;
}
/* Port 2 is easy. */
card->port[1].start_offset = 0;
card->port[1].end_offset = NM_PORT2_SIZE;
/* Yuck. But we have to map in port 2 so we can check how much RAM the
card has. */
if (nm256_remap_ports (card)) {
kfree (card);
return 0;
}
/*
* The NM256 has two memory ports. The first port is nothing
* more than a chunk of video RAM, which is used as the I/O ring
* buffer. The second port has the actual juicy stuff (like the
* mixer and the playback engine control registers).
*/
if (card->rev == REV_NM256AV) {
/* Ok, try to see if this is a non-AC97 version of the hardware. */
int pval = nm256_readPort16 (card, 2, NM_MIXER_PRESENCE);
if ((pval & NM_PRESENCE_MASK) != NM_PRESENCE_VALUE) {
if (! force_load) {
printk (KERN_ERR "NM256: This doesn't look to me like the AC97-compatible version.\n");
printk (KERN_ERR " You can force the driver to load by passing in the module\n");
printk (KERN_ERR " parameter:\n");
printk (KERN_ERR " force_load = 1\n");
printk (KERN_ERR "\n");
printk (KERN_ERR " More likely, you should be using the appropriate SB-16 or\n");
printk (KERN_ERR " CS4232 driver instead. (If your BIOS has settings for\n");
printk (KERN_ERR " IRQ and/or DMA for the sound card, this is *not* the correct\n");
printk (KERN_ERR " driver to use.)\n");
nm256_release_ports (card);
kfree (card);
return 0;
}
else {
printk (KERN_INFO "NM256: Forcing driver load as per user request.\n");
}
}
else {
/* printk (KERN_INFO "NM256: Congratulations. You're not running Eunice.\n")*/;
}
card->port[0].end_offset = 2560 * 1024;
card->introutine = nm256_interrupt;
card->mixer_status_offset = NM_MIXER_STATUS_OFFSET;
card->mixer_status_mask = NM_MIXER_READY_MASK;
}
else {
/* Not sure if there is any relevant detect for the ZX or not. */
if (nm256_readPort8 (card, 2, 0xa0b) != 0)
card->port[0].end_offset = 6144 * 1024;
else
card->port[0].end_offset = 4096 * 1024;
card->introutine = nm256_interrupt_zx;
card->mixer_status_offset = NM2_MIXER_STATUS_OFFSET;
card->mixer_status_mask = NM2_MIXER_READY_MASK;
}
if (buffertop >= 98304 && buffertop < card->port[0].end_offset)
card->port[0].end_offset = buffertop;
else
nm256_peek_for_sig (card);
card->port[0].start_offset = card->port[0].end_offset - 98304;
printk (KERN_INFO "NM256: Mapping port 1 from 0x%x - 0x%x\n",
card->port[0].start_offset, card->port[0].end_offset);
if (nm256_remap_ports (card)) {
kfree (card);
return 0;
}
/* See if we can get the interrupt. */
card->irq = pcidev->irq;
card->has_irq = 0;
if (nm256_grabInterrupt (card) != 0) {
nm256_release_ports (card);
kfree (card);
return 0;
}
nm256_releaseInterrupt (card);
/*
* Init the board.
*/
card->playbackBufferSize = 16384;
card->recordBufferSize = 16384;
card->coeffBuf = card->port[0].end_offset - NM_MAX_COEFFICIENT;
card->abuf2 = card->coeffBuf - card->recordBufferSize;
card->abuf1 = card->abuf2 - card->playbackBufferSize;
card->allCoeffBuf = card->abuf2 - (NM_TOTAL_COEFF_COUNT * 4);
/* Fixed setting. */
card->mixer = NM_MIXER_OFFSET;
card->mixer_values_init = 0;
card->is_open_play = 0;
card->is_open_record = 0;
card->coeffsCurrent = 0;
card->opencnt[0] = 0; card->opencnt[1] = 0;
/* Reasonable default settings, but largely unnecessary. */
for (x = 0; x < 2; x++) {
card->sinfo[x].bits = 8;
card->sinfo[x].stereo = 0;
card->sinfo[x].samplerate = 8000;
}
nm256_initHw (card);
for (x = 0; x < 2; x++) {
if ((card->dev[x] =
sound_install_audiodrv(AUDIO_DRIVER_VERSION,
"NM256", &nm256_audio_driver,
sizeof(struct audio_driver),
DMA_NODMA, AFMT_U8 | AFMT_S16_LE,
NULL, -1, -1)) >= 0) {
/* 1K minimum buffer size. */
audio_devs[card->dev[x]]->min_fragment = 10;
/* Maximum of 8K buffer size. */
audio_devs[card->dev[x]]->max_fragment = 13;
}
else {
printk(KERN_ERR "NM256: Too many PCM devices available\n");
nm256_release_ports (card);
kfree (card);
return 0;
}
}
pci_set_drvdata(pcidev,card);
/* Insert the card in the list. */
card->next_card = nmcard_list;
nmcard_list = card;
printk(KERN_INFO "Initialized NeoMagic %s audio in PCI native mode\n",
verstr);
/*
* And our mixer. (We should allow support for other mixers, maybe.)
*/
nm256_install_mixer (card);
return 1;
}
static int __devinit
nm256_probe(struct pci_dev *pcidev,const struct pci_device_id *pciid)
{
if (pcidev->device == PCI_DEVICE_ID_NEOMAGIC_NM256AV_AUDIO)
return nm256_install(pcidev, REV_NM256AV, "256AV");
if (pcidev->device == PCI_DEVICE_ID_NEOMAGIC_NM256ZX_AUDIO)
return nm256_install(pcidev, REV_NM256ZX, "256ZX");
if (pcidev->device == PCI_DEVICE_ID_NEOMAGIC_NM256XL_PLUS_AUDIO)
return nm256_install(pcidev, REV_NM256ZX, "256XL+");
return -1; /* should not come here ... */
}
static void __devinit
nm256_remove(struct pci_dev *pcidev) {
struct nm256_info *xcard = pci_get_drvdata(pcidev);
struct nm256_info *card,*next_card = NULL;
for (card = nmcard_list; card != NULL; card = next_card) {
next_card = card->next_card;
if (card == xcard) {
stopPlay (card);
stopRecord (card);
if (card->has_irq)
free_irq (card->irq, card);
nm256_release_ports (card);
sound_unload_mixerdev (card->mixer_oss_dev);
sound_unload_audiodev (card->dev[0]);
sound_unload_audiodev (card->dev[1]);
kfree (card);
break;
}
}
if (nmcard_list == card)
nmcard_list = next_card;
}
/*
* Open the device
*
* DEV - device
* MODE - mode to open device (logical OR of OPEN_READ and OPEN_WRITE)
*
* Called when opening the DMAbuf (dmabuf.c:259)
*/
static int
nm256_audio_open(int dev, int mode)
{
struct nm256_info *card = nm256_find_card (dev);
int w;
if (card == NULL)
return -ENODEV;
if (card->dev[0] == dev)
w = 0;
else if (card->dev[1] == dev)
w = 1;
else
return -ENODEV;
if (card->opencnt[w] > 0)
return -EBUSY;
/* No bits set? Huh? */
if (! ((mode & OPEN_READ) || (mode & OPEN_WRITE)))
return -EIO;
/*
* If it's open for both read and write, and the card's currently
* being read or written to, then do the opposite of what has
* already been done. Otherwise, don't specify any mode until the
* user actually tries to do I/O. (Some programs open the device
* for both read and write, but only actually do reading or writing.)
*/
if ((mode & OPEN_WRITE) && (mode & OPEN_READ)) {
if (card->is_open_play)
mode = OPEN_WRITE;
else if (card->is_open_record)
mode = OPEN_READ;
else mode = 0;
}
if (mode & OPEN_WRITE) {
if (card->is_open_play == 0) {
card->dev_for_play = dev;
card->is_open_play = 1;
}
else
return -EBUSY;
}
if (mode & OPEN_READ) {
if (card->is_open_record == 0) {
card->dev_for_record = dev;
card->is_open_record = 1;
}
else
return -EBUSY;
}
card->opencnt[w]++;
return 0;
}
/*
* Close the device
*
* DEV - device
*
* Called when closing the DMAbuf (dmabuf.c:477)
* after halt_xfer
*/
static void
nm256_audio_close(int dev)
{
struct nm256_info *card = nm256_find_card (dev);
if (card != NULL) {
int w;
if (card->dev[0] == dev)
w = 0;
else if (card->dev[1] == dev)
w = 1;
else
return;
card->opencnt[w]--;
if (card->opencnt[w] <= 0) {
card->opencnt[w] = 0;
if (card->dev_for_play == dev) {
stopPlay (card);
card->is_open_play = 0;
card->dev_for_play = -1;
}
if (card->dev_for_record == dev) {
stopRecord (card);
card->is_open_record = 0;
card->dev_for_record = -1;
}
}
}
}
/* Standard ioctl handler. */
static int
nm256_audio_ioctl(int dev, unsigned int cmd, void __user *arg)
{
int ret;
u32 oldinfo;
int w;
struct nm256_info *card = nm256_find_card (dev);
if (card == NULL)
return -ENODEV;
if (dev == card->dev[0])
w = 0;
else
w = 1;
/*
* The code here is messy. There are probably better ways to do
* it. (It should be possible to handle it the same way the AC97 mixer
* is done.)
*/
switch (cmd)
{
case SOUND_PCM_WRITE_RATE:
if (get_user(ret, (int __user *) arg))
return -EFAULT;
if (ret != 0) {
oldinfo = card->sinfo[w].samplerate;
card->sinfo[w].samplerate = ret;
ret = nm256_setInfo(dev, card);
if (ret != 0)
card->sinfo[w].samplerate = oldinfo;
}
if (ret == 0)
ret = card->sinfo[w].samplerate;
break;
case SOUND_PCM_READ_RATE:
ret = card->sinfo[w].samplerate;
break;
case SNDCTL_DSP_STEREO:
if (get_user(ret, (int __user *) arg))
return -EFAULT;
card->sinfo[w].stereo = ret ? 1 : 0;
ret = nm256_setInfo (dev, card);
if (ret == 0)
ret = card->sinfo[w].stereo;
break;
case SOUND_PCM_WRITE_CHANNELS:
if (get_user(ret, (int __user *) arg))
return -EFAULT;
if (ret < 1 || ret > 3)
ret = card->sinfo[w].stereo + 1;
else {
card->sinfo[w].stereo = ret - 1;
ret = nm256_setInfo (dev, card);
if (ret == 0)
ret = card->sinfo[w].stereo + 1;
}
break;
case SOUND_PCM_READ_CHANNELS:
ret = card->sinfo[w].stereo + 1;
break;
case SNDCTL_DSP_SETFMT:
if (get_user(ret, (int __user *) arg))
return -EFAULT;
if (ret != 0) {
oldinfo = card->sinfo[w].bits;
card->sinfo[w].bits = ret;
ret = nm256_setInfo (dev, card);
if (ret != 0)
card->sinfo[w].bits = oldinfo;
}
if (ret == 0)
ret = card->sinfo[w].bits;
break;
case SOUND_PCM_READ_BITS:
ret = card->sinfo[w].bits;
break;
default:
return -EINVAL;
}
return put_user(ret, (int __user *) arg);
}
/*
* Given the sound device DEV and an associated physical buffer PHYSBUF,
* return a pointer to the actual buffer in kernel space.
*
* This routine should exist as part of the soundcore routines.
*/
static char *
nm256_getDMAbuffer (int dev, unsigned long physbuf)
{
struct audio_operations *adev = audio_devs[dev];
struct dma_buffparms *dmap = adev->dmap_out;
char *dma_start =
(char *)(physbuf - (unsigned long)dmap->raw_buf_phys
+ (unsigned long)dmap->raw_buf);
return dma_start;
}
/*
* Output a block to sound device
*
* dev - device number
* buf - physical address of buffer
* total_count - total byte count in buffer
* intrflag - set if this has been called from an interrupt
* (via DMAbuf_outputintr)
* restart_dma - set if engine needs to be re-initialised
*
* Called when:
* 1. Starting output (dmabuf.c:1327)
* 2. (dmabuf.c:1504)
* 3. A new buffer needs to be sent to the device (dmabuf.c:1579)
*/
static void
nm256_audio_output_block(int dev, unsigned long physbuf,
int total_count, int intrflag)
{
struct nm256_info *card = nm256_find_card (dev);
if (card != NULL) {
char *dma_buf = nm256_getDMAbuffer (dev, physbuf);
card->is_open_play = 1;
card->dev_for_play = dev;
nm256_write_block (card, dma_buf, total_count);
}
}
/* Ditto, but do recording instead. */
static void
nm256_audio_start_input(int dev, unsigned long physbuf, int count,
int intrflag)
{
struct nm256_info *card = nm256_find_card (dev);
if (card != NULL) {
char *dma_buf = nm256_getDMAbuffer (dev, physbuf);
card->is_open_record = 1;
card->dev_for_record = dev;
nm256_startRecording (card, dma_buf, count);
}
}
/*
* Prepare for inputting samples to DEV.
* Each requested buffer will be BSIZE byes long, with a total of
* BCOUNT buffers.
*/
static int
nm256_audio_prepare_for_input(int dev, int bsize, int bcount)
{
struct nm256_info *card = nm256_find_card (dev);
if (card == NULL)
return -ENODEV;
if (card->is_open_record && card->dev_for_record != dev)
return -EBUSY;
audio_devs[dev]->dmap_in->flags |= DMA_NODMA;
return 0;
}
/*
* Prepare for outputting samples to `dev'
*
* Each buffer that will be passed will be `bsize' bytes long,
* with a total of `bcount' buffers.
*
* Called when:
* 1. A trigger enables audio output (dmabuf.c:978)
* 2. We get a write buffer without dma_mode setup (dmabuf.c:1152)
* 3. We restart a transfer (dmabuf.c:1324)
*/
static int
nm256_audio_prepare_for_output(int dev, int bsize, int bcount)
{
struct nm256_info *card = nm256_find_card (dev);
if (card == NULL)
return -ENODEV;
if (card->is_open_play && card->dev_for_play != dev)
return -EBUSY;
audio_devs[dev]->dmap_out->flags |= DMA_NODMA;
return 0;
}
/* Stop the current operations associated with DEV. */
static void
nm256_audio_reset(int dev)
{
struct nm256_info *card = nm256_find_card (dev);
if (card != NULL) {
if (card->dev_for_play == dev)
stopPlay (card);
if (card->dev_for_record == dev)
stopRecord (card);
}
}
static int
nm256_audio_local_qlen(int dev)
{
return 0;
}
static struct audio_driver nm256_audio_driver =
{
.owner = THIS_MODULE,
.open = nm256_audio_open,
.close = nm256_audio_close,
.output_block = nm256_audio_output_block,
.start_input = nm256_audio_start_input,
.ioctl = nm256_audio_ioctl,
.prepare_for_input = nm256_audio_prepare_for_input,
.prepare_for_output = nm256_audio_prepare_for_output,
.halt_io = nm256_audio_reset,
.local_qlen = nm256_audio_local_qlen,
};
static struct pci_device_id nm256_pci_tbl[] = {
{PCI_VENDOR_ID_NEOMAGIC, PCI_DEVICE_ID_NEOMAGIC_NM256AV_AUDIO,
PCI_ANY_ID, PCI_ANY_ID, 0, 0},
{PCI_VENDOR_ID_NEOMAGIC, PCI_DEVICE_ID_NEOMAGIC_NM256ZX_AUDIO,
PCI_ANY_ID, PCI_ANY_ID, 0, 0},
{PCI_VENDOR_ID_NEOMAGIC, PCI_DEVICE_ID_NEOMAGIC_NM256XL_PLUS_AUDIO,
PCI_ANY_ID, PCI_ANY_ID, 0, 0},
{0,}
};
MODULE_DEVICE_TABLE(pci, nm256_pci_tbl);
MODULE_LICENSE("GPL");
static struct pci_driver nm256_pci_driver = {
.name = "nm256_audio",
.id_table = nm256_pci_tbl,
.probe = nm256_probe,
.remove = nm256_remove,
};
module_param(usecache, bool, 0);
module_param(buffertop, int, 0);
module_param(nm256_debug, bool, 0644);
module_param(force_load, bool, 0);
static int __init do_init_nm256(void)
{
printk (KERN_INFO "NeoMagic 256AV/256ZX audio driver, version 1.1p\n");
return pci_register_driver(&nm256_pci_driver);
}
static void __exit cleanup_nm256 (void)
{
pci_unregister_driver(&nm256_pci_driver);
}
module_init(do_init_nm256);
module_exit(cleanup_nm256);
/*
* Local variables:
* c-basic-offset: 4
* End:
*/