linux/drivers/macintosh/via-maciisi.c

680 lines
16 KiB
C
Raw Normal View History

/*
* Device driver for the IIsi-style ADB on some Mac LC and II-class machines
*
* Based on via-cuda.c and via-macii.c, as well as the original
* adb-bus.c, which in turn is somewhat influenced by (but uses no
* code from) the NetBSD HWDIRECT ADB code. Original IIsi driver work
* was done by Robert Thompson and integrated into the old style
* driver by Michael Schmitz.
*
* Original sources (c) Alan Cox, Paul Mackerras, and others.
*
* Rewritten for Unified ADB by David Huggins-Daines <dhd@debian.org>
*
* 7/13/2000- extensive changes by Andrew McPherson <andrew@macduff.dhs.org>
* Works about 30% of the time now.
*/
#include <linux/types.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/adb.h>
#include <linux/cuda.h>
#include <linux/delay.h>
#include <linux/interrupt.h>
#include <asm/macintosh.h>
#include <asm/macints.h>
#include <asm/machw.h>
#include <asm/mac_via.h>
static volatile unsigned char *via;
/* VIA registers - spaced 0x200 bytes apart - only the ones we actually use */
#define RS 0x200 /* skip between registers */
#define B 0 /* B-side data */
#define A RS /* A-side data */
#define DIRB (2*RS) /* B-side direction (1=output) */
#define DIRA (3*RS) /* A-side direction (1=output) */
#define SR (10*RS) /* Shift register */
#define ACR (11*RS) /* Auxiliary control register */
#define IFR (13*RS) /* Interrupt flag register */
#define IER (14*RS) /* Interrupt enable register */
/* Bits in B data register: all active low */
#define TREQ 0x08 /* Transfer request (input) */
#define TACK 0x10 /* Transfer acknowledge (output) */
#define TIP 0x20 /* Transfer in progress (output) */
#define ST_MASK 0x30 /* mask for selecting ADB state bits */
/* Bits in ACR */
#define SR_CTRL 0x1c /* Shift register control bits */
#define SR_EXT 0x0c /* Shift on external clock */
#define SR_OUT 0x10 /* Shift out if 1 */
/* Bits in IFR and IER */
#define IER_SET 0x80 /* set bits in IER */
#define IER_CLR 0 /* clear bits in IER */
#define SR_INT 0x04 /* Shift register full/empty */
#define SR_DATA 0x08 /* Shift register data */
#define SR_CLOCK 0x10 /* Shift register clock */
#define ADB_DELAY 150
#undef DEBUG_MACIISI_ADB
static struct adb_request* current_req = NULL;
static struct adb_request* last_req = NULL;
static unsigned char maciisi_rbuf[16];
static unsigned char *reply_ptr = NULL;
static int data_index;
static int reading_reply;
static int reply_len;
static int tmp;
static int need_sync;
static enum maciisi_state {
idle,
sending,
reading,
} maciisi_state;
static int maciisi_probe(void);
static int maciisi_init(void);
static int maciisi_send_request(struct adb_request* req, int sync);
static void maciisi_sync(struct adb_request *req);
static int maciisi_write(struct adb_request* req);
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 maciisi_interrupt(int irq, void* arg);
static void maciisi_input(unsigned char *buf, int nb);
static int maciisi_init_via(void);
static void maciisi_poll(void);
static int maciisi_start(void);
struct adb_driver via_maciisi_driver = {
"Mac IIsi",
maciisi_probe,
maciisi_init,
maciisi_send_request,
NULL, /* maciisi_adb_autopoll, */
maciisi_poll,
NULL /* maciisi_reset_adb_bus */
};
static int
maciisi_probe(void)
{
if (macintosh_config->adb_type != MAC_ADB_IISI)
return -ENODEV;
via = via1;
return 0;
}
static int
maciisi_init(void)
{
int err;
if (via == NULL)
return -ENODEV;
if ((err = maciisi_init_via())) {
printk(KERN_ERR "maciisi_init: maciisi_init_via() failed, code %d\n", err);
via = NULL;
return err;
}
if (request_irq(IRQ_MAC_ADB, maciisi_interrupt, IRQ_FLG_LOCK | IRQ_FLG_FAST,
"ADB", maciisi_interrupt)) {
printk(KERN_ERR "maciisi_init: can't get irq %d\n", IRQ_MAC_ADB);
return -EAGAIN;
}
printk("adb: Mac IIsi driver v0.2 for Unified ADB.\n");
return 0;
}
/* Flush data from the ADB controller */
static void
maciisi_stfu(void)
{
int status = via[B] & (TIP|TREQ);
if (status & TREQ) {
#ifdef DEBUG_MACIISI_ADB
printk (KERN_DEBUG "maciisi_stfu called with TREQ high!\n");
#endif
return;
}
udelay(ADB_DELAY);
via[ACR] &= ~SR_OUT;
via[IER] = IER_CLR | SR_INT;
udelay(ADB_DELAY);
status = via[B] & (TIP|TREQ);
if (!(status & TREQ))
{
via[B] |= TIP;
while(1)
{
int poll_timeout = ADB_DELAY * 5;
/* Poll for SR interrupt */
while (!(via[IFR] & SR_INT) && poll_timeout-- > 0)
status = via[B] & (TIP|TREQ);
tmp = via[SR]; /* Clear shift register */
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "maciisi_stfu: status %x timeout %d data %x\n",
status, poll_timeout, tmp);
#endif
if(via[B] & TREQ)
break;
/* ACK on-off */
via[B] |= TACK;
udelay(ADB_DELAY);
via[B] &= ~TACK;
}
/* end frame */
via[B] &= ~TIP;
udelay(ADB_DELAY);
}
via[IER] = IER_SET | SR_INT;
}
/* All specifically VIA-related initialization goes here */
static int
maciisi_init_via(void)
{
int i;
/* Set the lines up. We want TREQ as input TACK|TIP as output */
via[DIRB] = (via[DIRB] | TACK | TIP) & ~TREQ;
/* Shift register on input */
via[ACR] = (via[ACR] & ~SR_CTRL) | SR_EXT;
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "maciisi_init_via: initial status %x\n", via[B] & (TIP|TREQ));
#endif
/* Wipe any pending data and int */
tmp = via[SR];
/* Enable keyboard interrupts */
via[IER] = IER_SET | SR_INT;
/* Set initial state: idle */
via[B] &= ~(TACK|TIP);
/* Clear interrupt bit */
via[IFR] = SR_INT;
for(i = 0; i < 60; i++) {
udelay(ADB_DELAY);
maciisi_stfu();
udelay(ADB_DELAY);
if(via[B] & TREQ)
break;
}
if (i == 60)
printk(KERN_ERR "maciisi_init_via: bus jam?\n");
maciisi_state = idle;
need_sync = 0;
return 0;
}
/* Send a request, possibly waiting for a reply */
static int
maciisi_send_request(struct adb_request* req, int sync)
{
int i;
#ifdef DEBUG_MACIISI_ADB
static int dump_packet = 0;
#endif
if (via == NULL) {
req->complete = 1;
return -ENXIO;
}
#ifdef DEBUG_MACIISI_ADB
if (dump_packet) {
printk(KERN_DEBUG "maciisi_send_request:");
for (i = 0; i < req->nbytes; i++) {
printk(" %.2x", req->data[i]);
}
printk(" sync %d\n", sync);
}
#endif
req->reply_expected = 1;
i = maciisi_write(req);
if (i)
{
/* Normally, if a packet requires syncing, that happens at the end of
* maciisi_send_request. But if the transfer fails, it will be restarted
* by maciisi_interrupt(). We use need_sync to tell maciisi_interrupt
* when to sync a packet that it sends out.
*
* Suggestions on a better way to do this are welcome.
*/
if(i == -EBUSY && sync)
need_sync = 1;
else
need_sync = 0;
return i;
}
if(sync)
maciisi_sync(req);
return 0;
}
/* Poll the ADB chip until the request completes */
static void maciisi_sync(struct adb_request *req)
{
int count = 0;
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "maciisi_sync called\n");
#endif
/* If for some reason the ADB chip shuts up on us, we want to avoid an endless loop. */
while (!req->complete && count++ < 50) {
maciisi_poll();
}
/* This could be BAD... when the ADB controller doesn't respond
* for this long, it's probably not coming back :-( */
if(count >= 50) /* Hopefully shouldn't happen */
printk(KERN_ERR "maciisi_send_request: poll timed out!\n");
}
int
maciisi_request(struct adb_request *req, void (*done)(struct adb_request *),
int nbytes, ...)
{
va_list list;
int i;
req->nbytes = nbytes;
req->done = done;
req->reply_expected = 0;
va_start(list, nbytes);
for (i = 0; i < nbytes; i++)
req->data[i++] = va_arg(list, int);
va_end(list);
return maciisi_send_request(req, 1);
}
/* Enqueue a request, and run the queue if possible */
static int
maciisi_write(struct adb_request* req)
{
unsigned long flags;
int i;
/* We will accept CUDA packets - the VIA sends them to us, so
it figures that we should be able to send them to it */
if (req->nbytes < 2 || req->data[0] > CUDA_PACKET) {
printk(KERN_ERR "maciisi_write: packet too small or not an ADB or CUDA packet\n");
req->complete = 1;
return -EINVAL;
}
req->next = NULL;
req->sent = 0;
req->complete = 0;
req->reply_len = 0;
local_irq_save(flags);
if (current_req) {
last_req->next = req;
last_req = req;
} else {
current_req = req;
last_req = req;
}
if (maciisi_state == idle)
{
i = maciisi_start();
if(i != 0)
{
local_irq_restore(flags);
return i;
}
}
else
{
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "maciisi_write: would start, but state is %d\n", maciisi_state);
#endif
local_irq_restore(flags);
return -EBUSY;
}
local_irq_restore(flags);
return 0;
}
static int
maciisi_start(void)
{
struct adb_request* req;
int status;
#ifdef DEBUG_MACIISI_ADB
status = via[B] & (TIP | TREQ);
printk(KERN_DEBUG "maciisi_start called, state=%d, status=%x, ifr=%x\n", maciisi_state, status, via[IFR]);
#endif
if (maciisi_state != idle) {
/* shouldn't happen */
printk(KERN_ERR "maciisi_start: maciisi_start called when driver busy!\n");
return -EBUSY;
}
req = current_req;
if (req == NULL)
return -EINVAL;
status = via[B] & (TIP|TREQ);
if (!(status & TREQ)) {
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "maciisi_start: bus busy - aborting\n");
#endif
return -EBUSY;
}
/* Okay, send */
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "maciisi_start: sending\n");
#endif
/* Set state to active */
via[B] |= TIP;
/* ACK off */
via[B] &= ~TACK;
/* Delay */
udelay(ADB_DELAY);
/* Shift out and send */
via[ACR] |= SR_OUT;
via[SR] = req->data[0];
data_index = 1;
/* ACK on */
via[B] |= TACK;
maciisi_state = sending;
return 0;
}
void
maciisi_poll(void)
{
unsigned long flags;
local_irq_save(flags);
if (via[IFR] & SR_INT) {
maciisi_interrupt(0, NULL);
}
else /* avoid calling this function too quickly in a loop */
udelay(ADB_DELAY);
local_irq_restore(flags);
}
/* Shift register interrupt - this is *supposed* to mean that the
register is either full or empty. In practice, I have no idea what
it means :( */
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
maciisi_interrupt(int irq, void* arg)
{
int status;
struct adb_request *req;
#ifdef DEBUG_MACIISI_ADB
static int dump_reply = 0;
#endif
int i;
unsigned long flags;
local_irq_save(flags);
status = via[B] & (TIP|TREQ);
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "state %d status %x ifr %x\n", maciisi_state, status, via[IFR]);
#endif
if (!(via[IFR] & SR_INT)) {
/* Shouldn't happen, we hope */
printk(KERN_ERR "maciisi_interrupt: called without interrupt flag set\n");
local_irq_restore(flags);
return IRQ_NONE;
}
/* Clear the interrupt */
/* via[IFR] = SR_INT; */
switch_start:
switch (maciisi_state) {
case idle:
if (status & TIP)
printk(KERN_ERR "maciisi_interrupt: state is idle but TIP asserted!\n");
if(!reading_reply)
udelay(ADB_DELAY);
/* Shift in */
via[ACR] &= ~SR_OUT;
/* Signal start of frame */
via[B] |= TIP;
/* Clear the interrupt (throw this value on the floor, it's useless) */
tmp = via[SR];
/* ACK adb chip, high-low */
via[B] |= TACK;
udelay(ADB_DELAY);
via[B] &= ~TACK;
reply_len = 0;
maciisi_state = reading;
if (reading_reply) {
reply_ptr = current_req->reply;
} else {
reply_ptr = maciisi_rbuf;
}
break;
case sending:
/* via[SR]; */
/* Set ACK off */
via[B] &= ~TACK;
req = current_req;
if (!(status & TREQ)) {
/* collision */
printk(KERN_ERR "maciisi_interrupt: send collision\n");
/* Set idle and input */
via[ACR] &= ~SR_OUT;
tmp = via[SR];
via[B] &= ~TIP;
/* Must re-send */
reading_reply = 0;
reply_len = 0;
maciisi_state = idle;
udelay(ADB_DELAY);
/* process this now, because the IFR has been cleared */
goto switch_start;
}
udelay(ADB_DELAY);
if (data_index >= req->nbytes) {
/* Sent the whole packet, put the bus back in idle state */
/* Shift in, we are about to read a reply (hopefully) */
via[ACR] &= ~SR_OUT;
tmp = via[SR];
/* End of frame */
via[B] &= ~TIP;
req->sent = 1;
maciisi_state = idle;
if (req->reply_expected) {
/* Note: only set this once we've
successfully sent the packet */
reading_reply = 1;
} else {
current_req = req->next;
if (req->done)
(*req->done)(req);
/* Do any queued requests now */
i = maciisi_start();
if(i == 0 && need_sync) {
/* Packet needs to be synced */
maciisi_sync(current_req);
}
if(i != -EBUSY)
need_sync = 0;
}
} else {
/* Sending more stuff */
/* Shift out */
via[ACR] |= SR_OUT;
/* Write */
via[SR] = req->data[data_index++];
/* Signal 'byte ready' */
via[B] |= TACK;
}
break;
case reading:
/* Shift in */
/* via[ACR] &= ~SR_OUT; */ /* Not in 2.2 */
if (reply_len++ > 16) {
printk(KERN_ERR "maciisi_interrupt: reply too long, aborting read\n");
via[B] |= TACK;
udelay(ADB_DELAY);
via[B] &= ~(TACK|TIP);
maciisi_state = idle;
i = maciisi_start();
if(i == 0 && need_sync) {
/* Packet needs to be synced */
maciisi_sync(current_req);
}
if(i != -EBUSY)
need_sync = 0;
break;
}
/* Read data */
*reply_ptr++ = via[SR];
status = via[B] & (TIP|TREQ);
/* ACK on/off */
via[B] |= TACK;
udelay(ADB_DELAY);
via[B] &= ~TACK;
if (!(status & TREQ))
break; /* more stuff to deal with */
/* end of frame */
via[B] &= ~TIP;
tmp = via[SR]; /* That's what happens in 2.2 */
udelay(ADB_DELAY); /* Give controller time to recover */
/* end of packet, deal with it */
if (reading_reply) {
req = current_req;
req->reply_len = reply_ptr - req->reply;
if (req->data[0] == ADB_PACKET) {
/* Have to adjust the reply from ADB commands */
if (req->reply_len <= 2 || (req->reply[1] & 2) != 0) {
/* the 0x2 bit indicates no response */
req->reply_len = 0;
} else {
/* leave just the command and result bytes in the reply */
req->reply_len -= 2;
memmove(req->reply, req->reply + 2, req->reply_len);
}
}
#ifdef DEBUG_MACIISI_ADB
if (dump_reply) {
int i;
printk(KERN_DEBUG "maciisi_interrupt: reply is ");
for (i = 0; i < req->reply_len; ++i)
printk(" %.2x", req->reply[i]);
printk("\n");
}
#endif
req->complete = 1;
current_req = req->next;
if (req->done)
(*req->done)(req);
/* Obviously, we got it */
reading_reply = 0;
} else {
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
maciisi_input(maciisi_rbuf, reply_ptr - maciisi_rbuf);
}
maciisi_state = idle;
status = via[B] & (TIP|TREQ);
if (!(status & TREQ)) {
/* Timeout?! More likely, another packet coming in already */
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "extra data after packet: status %x ifr %x\n",
status, via[IFR]);
#endif
#if 0
udelay(ADB_DELAY);
via[B] |= TIP;
maciisi_state = reading;
reading_reply = 0;
reply_ptr = maciisi_rbuf;
#else
/* Process the packet now */
reading_reply = 0;
goto switch_start;
#endif
/* We used to do this... but the controller might actually have data for us */
/* maciisi_stfu(); */
}
else {
/* Do any queued requests now if possible */
i = maciisi_start();
if(i == 0 && need_sync) {
/* Packet needs to be synced */
maciisi_sync(current_req);
}
if(i != -EBUSY)
need_sync = 0;
}
break;
default:
printk("maciisi_interrupt: unknown maciisi_state %d?\n", maciisi_state);
}
local_irq_restore(flags);
return IRQ_HANDLED;
}
static void
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
maciisi_input(unsigned char *buf, int nb)
{
#ifdef DEBUG_MACIISI_ADB
int i;
#endif
switch (buf[0]) {
case ADB_PACKET:
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
adb_input(buf+2, nb-2, buf[1] & 0x40);
break;
default:
#ifdef DEBUG_MACIISI_ADB
printk(KERN_DEBUG "data from IIsi ADB (%d bytes):", nb);
for (i = 0; i < nb; ++i)
printk(" %.2x", buf[i]);
printk("\n");
#endif
break;
}
}