aha/drivers/macintosh/via-macii.c

552 lines
14 KiB
C
Raw Normal View History

/*
* Device driver for the via ADB on (many) Mac II-class machines
*
* Based on the original ADB keyboard handler Copyright (c) 1997 Alan Cox
* Also derived from code Copyright (C) 1996 Paul Mackerras.
*
* With various updates provided over the years by Michael Schmitz,
* Guideo Koerber and others.
*
* Rewrite for Unified ADB by Joshua M. Thompson (funaho@jurai.org)
*
* 1999-08-02 (jmt) - Initial rewrite for Unified ADB.
* 2000-03-29 Tony Mantler <tonym@mac.linux-m68k.org>
* - Big overhaul, should actually work now.
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
* 2006-12-31 Finn Thain <fthain@telegraphics.com.au> - Another overhaul.
*
* Suggested reading:
* Inside Macintosh, ch. 5 ADB Manager
* Guide to the Macinstosh Family Hardware, ch. 8 Apple Desktop Bus
* Rockwell R6522 VIA datasheet
*
* Apple's "ADB Analyzer" bus sniffer is invaluable:
* ftp://ftp.apple.com/developer/Tool_Chest/Devices_-_Hardware/Apple_Desktop_Bus/
*/
#include <stdarg.h>
#include <linux/types.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/delay.h>
#include <linux/adb.h>
#include <linux/interrupt.h>
#include <linux/init.h>
#include <asm/macintosh.h>
#include <asm/macints.h>
#include <asm/machw.h>
#include <asm/mac_via.h>
#include <asm/system.h>
static volatile unsigned char *via;
/* VIA registers - spaced 0x200 bytes apart */
#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 T1CL (4*RS) /* Timer 1 ctr/latch (low 8 bits) */
#define T1CH (5*RS) /* Timer 1 counter (high 8 bits) */
#define T1LL (6*RS) /* Timer 1 latch (low 8 bits) */
#define T1LH (7*RS) /* Timer 1 latch (high 8 bits) */
#define T2CL (8*RS) /* Timer 2 ctr/latch (low 8 bits) */
#define T2CH (9*RS) /* Timer 2 counter (high 8 bits) */
#define SR (10*RS) /* Shift register */
#define ACR (11*RS) /* Auxiliary control register */
#define PCR (12*RS) /* Peripheral control register */
#define IFR (13*RS) /* Interrupt flag register */
#define IER (14*RS) /* Interrupt enable register */
#define ANH (15*RS) /* A-side data, no handshake */
/* Bits in B data register: all active low */
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
#define CTLR_IRQ 0x08 /* Controller rcv status (input) */
#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 */
/* ADB transaction states according to GMHW */
#define ST_CMD 0x00 /* ADB state: command byte */
#define ST_EVEN 0x10 /* ADB state: even data byte */
#define ST_ODD 0x20 /* ADB state: odd data byte */
#define ST_IDLE 0x30 /* ADB state: idle, nothing to send */
static int macii_init_via(void);
static void macii_start(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 13:55:46 +00:00
static irqreturn_t macii_interrupt(int irq, void *arg);
static void macii_queue_poll(void);
static int macii_probe(void);
static int macii_init(void);
static int macii_send_request(struct adb_request *req, int sync);
static int macii_write(struct adb_request *req);
static int macii_autopoll(int devs);
static void macii_poll(void);
static int macii_reset_bus(void);
struct adb_driver via_macii_driver = {
"Mac II",
macii_probe,
macii_init,
macii_send_request,
macii_autopoll,
macii_poll,
macii_reset_bus
};
static enum macii_state {
idle,
sending,
reading,
read_done,
} macii_state;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
static struct adb_request *current_req; /* first request struct in the queue */
static struct adb_request *last_req; /* last request struct in the queue */
static unsigned char reply_buf[16]; /* storage for autopolled replies */
static unsigned char *reply_ptr; /* next byte in req->data or reply_buf */
static int reading_reply; /* store reply in reply_buf else req->reply */
static int data_index; /* index of the next byte to send from req->data */
static int reply_len; /* number of bytes received in reply_buf or req->reply */
static int status; /* VIA's ADB status bits captured upon interrupt */
static int last_status; /* status bits as at previous interrupt */
static int srq_asserted; /* have to poll for the device that asserted it */
static int command_byte; /* the most recent command byte transmitted */
static int autopoll_devs; /* bits set are device addresses to be polled */
/* Sanity check for request queue. Doesn't check for cycles. */
static int request_is_queued(struct adb_request *req) {
struct adb_request *cur;
unsigned long flags;
local_irq_save(flags);
cur = current_req;
while (cur) {
if (cur == req) {
local_irq_restore(flags);
return 1;
}
cur = cur->next;
}
local_irq_restore(flags);
return 0;
}
/* Check for MacII style ADB */
static int macii_probe(void)
{
if (macintosh_config->adb_type != MAC_ADB_II) return -ENODEV;
via = via1;
printk("adb: Mac II ADB Driver v1.0 for Unified ADB\n");
return 0;
}
/* Initialize the driver */
int macii_init(void)
{
unsigned long flags;
int err;
local_irq_save(flags);
err = macii_init_via();
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (err) goto out;
err = request_irq(IRQ_MAC_ADB, macii_interrupt, IRQ_FLG_LOCK, "ADB",
macii_interrupt);
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (err) goto out;
macii_state = idle;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
out:
local_irq_restore(flags);
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
return err;
}
/* initialize the hardware */
static int macii_init_via(void)
{
unsigned char x;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
/* We want CTLR_IRQ as input and ST_EVEN | ST_ODD as output lines. */
via[DIRB] = (via[DIRB] | ST_EVEN | ST_ODD) & ~CTLR_IRQ;
/* Set up state: idle */
via[B] |= ST_IDLE;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
last_status = via[B] & (ST_MASK|CTLR_IRQ);
/* Shift register on input */
via[ACR] = (via[ACR] & ~SR_CTRL) | SR_EXT;
/* Wipe any pending data and int */
x = via[SR];
return 0;
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
/* Send an ADB poll (Talk Register 0 command prepended to the request queue) */
static void macii_queue_poll(void)
{
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
/* No point polling the active device as it will never assert SRQ, so
* poll the next device in the autopoll list. This could leave us
* stuck in a polling loop if an unprobed device is asserting SRQ.
* In theory, that could only happen if a device was plugged in after
* probing started. Unplugging it again will break the cycle.
* (Simply polling the next higher device often ends up polling almost
* every device (after wrapping around), which takes too long.)
*/
int device_mask;
int next_device;
static struct adb_request req;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (!autopoll_devs) return;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
device_mask = (1 << (((command_byte & 0xF0) >> 4) + 1)) - 1;
if (autopoll_devs & ~device_mask)
next_device = ffs(autopoll_devs & ~device_mask) - 1;
else
next_device = ffs(autopoll_devs) - 1;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
BUG_ON(request_is_queued(&req));
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
adb_request(&req, NULL, ADBREQ_NOSEND, 1,
ADB_READREG(next_device, 0));
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
req.sent = 0;
req.complete = 0;
req.reply_len = 0;
req.next = current_req;
if (current_req != NULL) {
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
current_req = &req;
} else {
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
current_req = &req;
last_req = &req;
}
}
/* Send an ADB request; if sync, poll out the reply 'till it's done */
static int macii_send_request(struct adb_request *req, int sync)
{
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
int err;
unsigned long flags;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
BUG_ON(request_is_queued(req));
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
local_irq_save(flags);
err = macii_write(req);
local_irq_restore(flags);
if (!err && sync) {
while (!req->complete) {
macii_poll();
}
BUG_ON(request_is_queued(req));
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
return err;
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
/* Send an ADB request (append to request queue) */
static int macii_write(struct adb_request *req)
{
if (req->nbytes < 2 || req->data[0] != ADB_PACKET || req->nbytes > 15) {
req->complete = 1;
return -EINVAL;
}
req->next = NULL;
req->sent = 0;
req->complete = 0;
req->reply_len = 0;
if (current_req != NULL) {
last_req->next = req;
last_req = req;
} else {
current_req = req;
last_req = req;
if (macii_state == idle) macii_start();
}
return 0;
}
/* Start auto-polling */
static int macii_autopoll(int devs)
{
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
static struct adb_request req;
unsigned long flags;
int err = 0;
/* bit 1 == device 1, and so on. */
autopoll_devs = devs & 0xFFFE;
if (!autopoll_devs) return 0;
local_irq_save(flags);
if (current_req == NULL) {
/* Send a Talk Reg 0. The controller will repeatedly transmit
* this as long as it is idle.
*/
adb_request(&req, NULL, ADBREQ_NOSEND, 1,
ADB_READREG(ffs(autopoll_devs) - 1, 0));
err = macii_write(&req);
}
local_irq_restore(flags);
return err;
}
static inline int need_autopoll(void) {
/* Was the last command Talk Reg 0
* and is the target on the autopoll list?
*/
if ((command_byte & 0x0F) == 0x0C &&
((1 << ((command_byte & 0xF0) >> 4)) & autopoll_devs))
return 0;
return 1;
}
/* Prod the chip without interrupts */
static void macii_poll(void)
{
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
disable_irq(IRQ_MAC_ADB);
macii_interrupt(0, NULL);
enable_irq(IRQ_MAC_ADB);
}
/* Reset the bus */
static int macii_reset_bus(void)
{
static struct adb_request req;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (request_is_queued(&req))
return 0;
/* Command = 0, Address = ignored */
adb_request(&req, NULL, 0, 1, ADB_BUSRESET);
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
/* Don't want any more requests during the Global Reset low time. */
udelay(3000);
return 0;
}
/* Start sending ADB packet */
static void macii_start(void)
{
struct adb_request *req;
req = current_req;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
BUG_ON(req == NULL);
BUG_ON(macii_state != idle);
/* Now send it. Be careful though, that first byte of the request
* is actually ADB_PACKET; the real data begins at index 1!
* And req->nbytes is the number of bytes of real data plus one.
*/
/* store command byte */
command_byte = req->data[1];
/* Output mode */
via[ACR] |= SR_OUT;
/* Load data */
via[SR] = req->data[1];
/* set ADB state to 'command' */
via[B] = (via[B] & ~ST_MASK) | ST_CMD;
macii_state = sending;
data_index = 2;
}
/*
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
* The notorious ADB interrupt handler - does all of the protocol handling.
* Relies on the ADB controller sending and receiving data, thereby
* generating shift register interrupts (SR_INT) for us. This means there has
* to be activity on the ADB bus. The chip will poll to achieve this.
*
* The basic ADB state machine was left unchanged from the original MacII code
* by Alan Cox, which was based on the CUDA driver for PowerMac.
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
* The syntax of the ADB status lines is totally different on MacII,
* though. MacII uses the states Command -> Even -> Odd -> Even ->...-> Idle
* for sending and Idle -> Even -> Odd -> Even ->...-> Idle for receiving.
* Start and end of a receive packet are signalled by asserting /IRQ on the
* interrupt line (/IRQ means the CTLR_IRQ bit in port B; not to be confused
* with the VIA shift register interrupt. /IRQ never actually interrupts the
* processor, it's just an ordinary input.)
*/
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 13:55:46 +00:00
static irqreturn_t macii_interrupt(int irq, void *arg)
{
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
int x;
static int entered;
struct adb_request *req;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (!arg) {
/* Clear the SR IRQ flag when polling. */
if (via[IFR] & SR_INT)
via[IFR] = SR_INT;
else
return IRQ_NONE;
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
BUG_ON(entered++);
last_status = status;
status = via[B] & (ST_MASK|CTLR_IRQ);
switch (macii_state) {
case idle:
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (reading_reply) {
reply_ptr = current_req->reply;
} else {
BUG_ON(current_req != NULL);
reply_ptr = reply_buf;
}
x = via[SR];
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if ((status & CTLR_IRQ) && (x == 0xFF)) {
/* Bus timeout without SRQ sequence:
* data is "FF" while CTLR_IRQ is "H"
*/
reply_len = 0;
srq_asserted = 0;
macii_state = read_done;
} else {
macii_state = reading;
*reply_ptr = x;
reply_len = 1;
}
/* set ADB state = even for first data byte */
via[B] = (via[B] & ~ST_MASK) | ST_EVEN;
break;
case sending:
req = current_req;
if (data_index >= req->nbytes) {
req->sent = 1;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
macii_state = idle;
if (req->reply_expected) {
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
reading_reply = 1;
} else {
req->complete = 1;
current_req = req->next;
if (req->done) (*req->done)(req);
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (current_req)
macii_start();
else
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (need_autopoll())
macii_autopoll(autopoll_devs);
}
if (macii_state == idle) {
/* reset to shift in */
via[ACR] &= ~SR_OUT;
x = via[SR];
/* set ADB state idle - might get SRQ */
via[B] = (via[B] & ~ST_MASK) | ST_IDLE;
}
} else {
via[SR] = req->data[data_index++];
if ( (via[B] & ST_MASK) == ST_CMD ) {
/* just sent the command byte, set to EVEN */
via[B] = (via[B] & ~ST_MASK) | ST_EVEN;
} else {
/* invert state bits, toggle ODD/EVEN */
via[B] ^= ST_MASK;
}
}
break;
case reading:
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
x = via[SR];
BUG_ON((status & ST_MASK) == ST_CMD ||
(status & ST_MASK) == ST_IDLE);
/* Bus timeout with SRQ sequence:
* data is "XX FF" while CTLR_IRQ is "L L"
* End of packet without SRQ sequence:
* data is "XX...YY 00" while CTLR_IRQ is "L...H L"
* End of packet SRQ sequence:
* data is "XX...YY 00" while CTLR_IRQ is "L...L L"
* (where XX is the first response byte and
* YY is the last byte of valid response data.)
*/
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
srq_asserted = 0;
if (!(status & CTLR_IRQ)) {
if (x == 0xFF) {
if (!(last_status & CTLR_IRQ)) {
macii_state = read_done;
reply_len = 0;
srq_asserted = 1;
}
} else if (x == 0x00) {
macii_state = read_done;
if (!(last_status & CTLR_IRQ))
srq_asserted = 1;
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
}
if (macii_state == reading) {
BUG_ON(reply_len > 15);
reply_ptr++;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
*reply_ptr = x;
reply_len++;
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
/* invert state bits, toggle ODD/EVEN */
via[B] ^= ST_MASK;
break;
case read_done:
x = via[SR];
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (reading_reply) {
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
reading_reply = 0;
req = current_req;
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
req->reply_len = reply_len;
req->complete = 1;
current_req = req->next;
if (req->done) (*req->done)(req);
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
} else if (reply_len && autopoll_devs)
adb_input(reply_buf, reply_len, 0);
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
macii_state = idle;
/* SRQ seen before, initiate poll now */
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (srq_asserted)
macii_queue_poll();
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
if (current_req)
macii_start();
else
if (need_autopoll())
macii_autopoll(autopoll_devs);
if (macii_state == idle)
via[B] = (via[B] & ~ST_MASK) | ST_IDLE;
break;
default:
break;
}
m68k: Mac II ADB fixes Fix a crash caused by requests placed in the queue with the completed flag already set. This lead to some ADB_SYNC requests returning early and their request structs being popped off the stack while still queued. Stack corruption ensued or an invalid request callback pointer was invoked or both. Eliminate macii_retransmit() and its buggy implementation of macii_write(). Have macii_queue_poll() fully initialise the request queues. Fix a bug in macii_queue_poll() where the last_req pointer was not being set. This caused some requests to leave the queue before being completed (and would also corrupt the stack under certain conditions). Fix a race in macii_start that could set the state machine to "reading" while current_req was null. No longer send poll commands with the ADBREQ_REPLY flag -- doing that caused the replies to be stored in the request buffer where they were forgotten about. Don't autopoll by continuously sending new Talk commands. Get the controller to do that for us. This reduces the ADB interrupt rate on an idle bus to about 5 per second. Only autopoll the devices that were probed. Explicitly clear the interrupt flag when polling. Use disable_irq rather than local_irq_save when polling. Remove excess local_irq_save/restore pairs. Improve bus timeout and service request detection. Remove unused code (last_reply, adb_dir etc) and unneeded code (prefix_len, first_byte etc). Change TIP and TACK to their correct names on this ADB controller (ST_EVEN and ST_ODD). Add some commentry. Add a generous quantity of sanity checks (BUG_ONs). Let m68k macs use the adb_sync boot param too. Tested on Mac II, Mac IIci, Quadra 650, Quadra 700 etc. Signed-off-by: Finn Thain <fthain@telegraphics.com.au> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-01 20:32:59 +00:00
entered--;
return IRQ_HANDLED;
}