kernel_optimize_test/drivers/char/snsc_event.c
David Howells 7d12e780e0 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 15:10:12 +01:00

310 lines
7.4 KiB
C

/*
* SN Platform system controller communication support
*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*
* Copyright (C) 2004-2006 Silicon Graphics, Inc. All rights reserved.
*/
/*
* System controller event handler
*
* These routines deal with environmental events arriving from the
* system controllers.
*/
#include <linux/interrupt.h>
#include <linux/sched.h>
#include <linux/byteorder/generic.h>
#include <asm/sn/sn_sal.h>
#include <asm/unaligned.h>
#include "snsc.h"
static struct subch_data_s *event_sd;
void scdrv_event(unsigned long);
DECLARE_TASKLET(sn_sysctl_event, scdrv_event, 0);
/*
* scdrv_event_interrupt
*
* Pull incoming environmental events off the physical link to the
* system controller and put them in a temporary holding area in SAL.
* Schedule scdrv_event() to move them along to their ultimate
* destination.
*/
static irqreturn_t
scdrv_event_interrupt(int irq, void *subch_data)
{
struct subch_data_s *sd = subch_data;
unsigned long flags;
int status;
spin_lock_irqsave(&sd->sd_rlock, flags);
status = ia64_sn_irtr_intr(sd->sd_nasid, sd->sd_subch);
if ((status > 0) && (status & SAL_IROUTER_INTR_RECV)) {
tasklet_schedule(&sn_sysctl_event);
}
spin_unlock_irqrestore(&sd->sd_rlock, flags);
return IRQ_HANDLED;
}
/*
* scdrv_parse_event
*
* Break an event (as read from SAL) into useful pieces so we can decide
* what to do with it.
*/
static int
scdrv_parse_event(char *event, int *src, int *code, int *esp_code, char *desc)
{
char *desc_end;
__be32 from_buf;
/* record event source address */
from_buf = get_unaligned((__be32 *)event);
*src = be32_to_cpup(&from_buf);
event += 4; /* move on to event code */
/* record the system controller's event code */
from_buf = get_unaligned((__be32 *)event);
*code = be32_to_cpup(&from_buf);
event += 4; /* move on to event arguments */
/* how many arguments are in the packet? */
if (*event++ != 2) {
/* if not 2, give up */
return -1;
}
/* parse out the ESP code */
if (*event++ != IR_ARG_INT) {
/* not an integer argument, so give up */
return -1;
}
from_buf = get_unaligned((__be32 *)event);
*esp_code = be32_to_cpup(&from_buf);
event += 4;
/* parse out the event description */
if (*event++ != IR_ARG_ASCII) {
/* not an ASCII string, so give up */
return -1;
}
event[CHUNKSIZE-1] = '\0'; /* ensure this string ends! */
event += 2; /* skip leading CR/LF */
desc_end = desc + sprintf(desc, "%s", event);
/* strip trailing CR/LF (if any) */
for (desc_end--;
(desc_end != desc) && ((*desc_end == 0xd) || (*desc_end == 0xa));
desc_end--) {
*desc_end = '\0';
}
return 0;
}
/*
* scdrv_event_severity
*
* Figure out how urgent a message we should write to the console/syslog
* via printk.
*/
static char *
scdrv_event_severity(int code)
{
int ev_class = (code & EV_CLASS_MASK);
int ev_severity = (code & EV_SEVERITY_MASK);
char *pk_severity = KERN_NOTICE;
switch (ev_class) {
case EV_CLASS_POWER:
switch (ev_severity) {
case EV_SEVERITY_POWER_LOW_WARNING:
case EV_SEVERITY_POWER_HIGH_WARNING:
pk_severity = KERN_WARNING;
break;
case EV_SEVERITY_POWER_HIGH_FAULT:
case EV_SEVERITY_POWER_LOW_FAULT:
pk_severity = KERN_ALERT;
break;
}
break;
case EV_CLASS_FAN:
switch (ev_severity) {
case EV_SEVERITY_FAN_WARNING:
pk_severity = KERN_WARNING;
break;
case EV_SEVERITY_FAN_FAULT:
pk_severity = KERN_CRIT;
break;
}
break;
case EV_CLASS_TEMP:
switch (ev_severity) {
case EV_SEVERITY_TEMP_ADVISORY:
pk_severity = KERN_WARNING;
break;
case EV_SEVERITY_TEMP_CRITICAL:
pk_severity = KERN_CRIT;
break;
case EV_SEVERITY_TEMP_FAULT:
pk_severity = KERN_ALERT;
break;
}
break;
case EV_CLASS_ENV:
pk_severity = KERN_ALERT;
break;
case EV_CLASS_TEST_FAULT:
pk_severity = KERN_ALERT;
break;
case EV_CLASS_TEST_WARNING:
pk_severity = KERN_WARNING;
break;
case EV_CLASS_PWRD_NOTIFY:
pk_severity = KERN_ALERT;
break;
}
return pk_severity;
}
/*
* scdrv_dispatch_event
*
* Do the right thing with an incoming event. That's often nothing
* more than printing it to the system log. For power-down notifications
* we start a graceful shutdown.
*/
static void
scdrv_dispatch_event(char *event, int len)
{
static int snsc_shutting_down = 0;
int code, esp_code, src, class;
char desc[CHUNKSIZE];
char *severity;
if (scdrv_parse_event(event, &src, &code, &esp_code, desc) < 0) {
/* ignore uninterpretible event */
return;
}
/* how urgent is the message? */
severity = scdrv_event_severity(code);
class = (code & EV_CLASS_MASK);
if (class == EV_CLASS_PWRD_NOTIFY || code == ENV_PWRDN_PEND) {
struct task_struct *p;
if (snsc_shutting_down)
return;
snsc_shutting_down = 1;
/* give a message for each type of event */
if (class == EV_CLASS_PWRD_NOTIFY)
printk(KERN_NOTICE "Power off indication received."
" Sending SIGPWR to init...\n");
else if (code == ENV_PWRDN_PEND)
printk(KERN_CRIT "WARNING: Shutting down the system"
" due to a critical environmental condition."
" Sending SIGPWR to init...\n");
/* give a SIGPWR signal to init proc */
kill_cad_pid(SIGPWR, 0);
} else {
/* print to system log */
printk("%s|$(0x%x)%s\n", severity, esp_code, desc);
}
}
/*
* scdrv_event
*
* Called as a tasklet when an event arrives from the L1. Read the event
* from where it's temporarily stored in SAL and call scdrv_dispatch_event()
* to send it on its way. Keep trying to read events until SAL indicates
* that there are no more immediately available.
*/
void
scdrv_event(unsigned long dummy)
{
int status;
int len;
unsigned long flags;
struct subch_data_s *sd = event_sd;
/* anything to read? */
len = CHUNKSIZE;
spin_lock_irqsave(&sd->sd_rlock, flags);
status = ia64_sn_irtr_recv(sd->sd_nasid, sd->sd_subch,
sd->sd_rb, &len);
while (!(status < 0)) {
spin_unlock_irqrestore(&sd->sd_rlock, flags);
scdrv_dispatch_event(sd->sd_rb, len);
len = CHUNKSIZE;
spin_lock_irqsave(&sd->sd_rlock, flags);
status = ia64_sn_irtr_recv(sd->sd_nasid, sd->sd_subch,
sd->sd_rb, &len);
}
spin_unlock_irqrestore(&sd->sd_rlock, flags);
}
/*
* scdrv_event_init
*
* Sets up a system controller subchannel to begin receiving event
* messages. This is sort of a specialized version of scdrv_open()
* in drivers/char/sn_sysctl.c.
*/
void
scdrv_event_init(struct sysctl_data_s *scd)
{
int rv;
event_sd = kzalloc(sizeof (struct subch_data_s), GFP_KERNEL);
if (event_sd == NULL) {
printk(KERN_WARNING "%s: couldn't allocate subchannel info"
" for event monitoring\n", __FUNCTION__);
return;
}
/* initialize subch_data_s fields */
event_sd->sd_nasid = scd->scd_nasid;
spin_lock_init(&event_sd->sd_rlock);
/* ask the system controllers to send events to this node */
event_sd->sd_subch = ia64_sn_sysctl_event_init(scd->scd_nasid);
if (event_sd->sd_subch < 0) {
kfree(event_sd);
printk(KERN_WARNING "%s: couldn't open event subchannel\n",
__FUNCTION__);
return;
}
/* hook event subchannel up to the system controller interrupt */
rv = request_irq(SGI_UART_VECTOR, scdrv_event_interrupt,
IRQF_SHARED | IRQF_DISABLED,
"system controller events", event_sd);
if (rv) {
printk(KERN_WARNING "%s: irq request failed (%d)\n",
__FUNCTION__, rv);
ia64_sn_irtr_close(event_sd->sd_nasid, event_sd->sd_subch);
kfree(event_sd);
return;
}
}