kernel_optimize_test/arch/parisc/kernel/irq.c
Grant Grundler 462b529f91 [PARISC] remove global_ack_eiem
Kudos to Thibaut Varene for spotting the (mis)use of appropriately named
global_ack_eiem. This took a long time to figure out and both insight
from myself, Kyle McMartin, and James Bottomley were required to narrow
down which bit of code could have this race condition.

The symptom was interrupts stopped getting delivered while some workload
was generating IO interrupts on two different CPUs. One of the interrupt
sources would get masked off and stay unmasked. Problem was global_ack_eiem
was accessed with read/modified/write sequence and not protected by
a spinlock.

PA-RISC doesn't need a global ack flag though. External Interrupts
are _always_ delivered to a single CPU (except for "global broadcast
interrupt" which AFAIK currently is not used.) So we don't have to worry
about any given IRQ vector getting delivered to more than one CPU.

Tested on a500 and rp34xx boxen. rsync to/from gsyprf11 (a500)
would lock up the box since NIC (tg3) interrupt and SCSI (sym2)
were on "opposite" CPUs (2 CPU system). Put them on the same CPU
or apply this patch and 10GB of data would rsync completely.

Please apply the following critical patch.

thanks,
grant

Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Acked-by: Thibaut VARENE <T-Bone@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
2007-06-12 01:23:30 -04:00

426 lines
10 KiB
C

/*
* Code to handle x86 style IRQs plus some generic interrupt stuff.
*
* Copyright (C) 1992 Linus Torvalds
* Copyright (C) 1994, 1995, 1996, 1997, 1998 Ralf Baechle
* Copyright (C) 1999 SuSE GmbH (Philipp Rumpf, prumpf@tux.org)
* Copyright (C) 1999-2000 Grant Grundler
* Copyright (c) 2005 Matthew Wilcox
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <linux/bitops.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/kernel_stat.h>
#include <linux/seq_file.h>
#include <linux/spinlock.h>
#include <linux/types.h>
#include <asm/io.h>
#include <asm/smp.h>
#undef PARISC_IRQ_CR16_COUNTS
extern irqreturn_t timer_interrupt(int, void *);
extern irqreturn_t ipi_interrupt(int, void *);
#define EIEM_MASK(irq) (1UL<<(CPU_IRQ_MAX - irq))
/* Bits in EIEM correlate with cpu_irq_action[].
** Numbered *Big Endian*! (ie bit 0 is MSB)
*/
static volatile unsigned long cpu_eiem = 0;
/*
** local ACK bitmap ... habitually set to 1, but reset to zero
** between ->ack() and ->end() of the interrupt to prevent
** re-interruption of a processing interrupt.
*/
static DEFINE_PER_CPU(unsigned long, local_ack_eiem) = ~0UL;
static void cpu_disable_irq(unsigned int irq)
{
unsigned long eirr_bit = EIEM_MASK(irq);
cpu_eiem &= ~eirr_bit;
/* Do nothing on the other CPUs. If they get this interrupt,
* The & cpu_eiem in the do_cpu_irq_mask() ensures they won't
* handle it, and the set_eiem() at the bottom will ensure it
* then gets disabled */
}
static void cpu_enable_irq(unsigned int irq)
{
unsigned long eirr_bit = EIEM_MASK(irq);
cpu_eiem |= eirr_bit;
/* This is just a simple NOP IPI. But what it does is cause
* all the other CPUs to do a set_eiem(cpu_eiem) at the end
* of the interrupt handler */
smp_send_all_nop();
}
static unsigned int cpu_startup_irq(unsigned int irq)
{
cpu_enable_irq(irq);
return 0;
}
void no_ack_irq(unsigned int irq) { }
void no_end_irq(unsigned int irq) { }
void cpu_ack_irq(unsigned int irq)
{
unsigned long mask = EIEM_MASK(irq);
int cpu = smp_processor_id();
/* Clear in EIEM so we can no longer process */
per_cpu(local_ack_eiem, cpu) &= ~mask;
/* disable the interrupt */
set_eiem(cpu_eiem & per_cpu(local_ack_eiem, cpu));
/* and now ack it */
mtctl(mask, 23);
}
void cpu_end_irq(unsigned int irq)
{
unsigned long mask = EIEM_MASK(irq);
int cpu = smp_processor_id();
/* set it in the eiems---it's no longer in process */
per_cpu(local_ack_eiem, cpu) |= mask;
/* enable the interrupt */
set_eiem(cpu_eiem & per_cpu(local_ack_eiem, cpu));
}
#ifdef CONFIG_SMP
int cpu_check_affinity(unsigned int irq, cpumask_t *dest)
{
int cpu_dest;
/* timer and ipi have to always be received on all CPUs */
if (CHECK_IRQ_PER_CPU(irq)) {
/* Bad linux design decision. The mask has already
* been set; we must reset it */
irq_desc[irq].affinity = CPU_MASK_ALL;
return -EINVAL;
}
/* whatever mask they set, we just allow one CPU */
cpu_dest = first_cpu(*dest);
*dest = cpumask_of_cpu(cpu_dest);
return 0;
}
static void cpu_set_affinity_irq(unsigned int irq, cpumask_t dest)
{
if (cpu_check_affinity(irq, &dest))
return;
irq_desc[irq].affinity = dest;
}
#endif
static struct hw_interrupt_type cpu_interrupt_type = {
.typename = "CPU",
.startup = cpu_startup_irq,
.shutdown = cpu_disable_irq,
.enable = cpu_enable_irq,
.disable = cpu_disable_irq,
.ack = cpu_ack_irq,
.end = cpu_end_irq,
#ifdef CONFIG_SMP
.set_affinity = cpu_set_affinity_irq,
#endif
/* XXX: Needs to be written. We managed without it so far, but
* we really ought to write it.
*/
.retrigger = NULL,
};
int show_interrupts(struct seq_file *p, void *v)
{
int i = *(loff_t *) v, j;
unsigned long flags;
if (i == 0) {
seq_puts(p, " ");
for_each_online_cpu(j)
seq_printf(p, " CPU%d", j);
#ifdef PARISC_IRQ_CR16_COUNTS
seq_printf(p, " [min/avg/max] (CPU cycle counts)");
#endif
seq_putc(p, '\n');
}
if (i < NR_IRQS) {
struct irqaction *action;
spin_lock_irqsave(&irq_desc[i].lock, flags);
action = irq_desc[i].action;
if (!action)
goto skip;
seq_printf(p, "%3d: ", i);
#ifdef CONFIG_SMP
for_each_online_cpu(j)
seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
#else
seq_printf(p, "%10u ", kstat_irqs(i));
#endif
seq_printf(p, " %14s", irq_desc[i].chip->typename);
#ifndef PARISC_IRQ_CR16_COUNTS
seq_printf(p, " %s", action->name);
while ((action = action->next))
seq_printf(p, ", %s", action->name);
#else
for ( ;action; action = action->next) {
unsigned int k, avg, min, max;
min = max = action->cr16_hist[0];
for (avg = k = 0; k < PARISC_CR16_HIST_SIZE; k++) {
int hist = action->cr16_hist[k];
if (hist) {
avg += hist;
} else
break;
if (hist > max) max = hist;
if (hist < min) min = hist;
}
avg /= k;
seq_printf(p, " %s[%d/%d/%d]", action->name,
min,avg,max);
}
#endif
seq_putc(p, '\n');
skip:
spin_unlock_irqrestore(&irq_desc[i].lock, flags);
}
return 0;
}
/*
** The following form a "set": Virtual IRQ, Transaction Address, Trans Data.
** Respectively, these map to IRQ region+EIRR, Processor HPA, EIRR bit.
**
** To use txn_XXX() interfaces, get a Virtual IRQ first.
** Then use that to get the Transaction address and data.
*/
int cpu_claim_irq(unsigned int irq, struct irq_chip *type, void *data)
{
if (irq_desc[irq].action)
return -EBUSY;
if (irq_desc[irq].chip != &cpu_interrupt_type)
return -EBUSY;
if (type) {
irq_desc[irq].chip = type;
irq_desc[irq].chip_data = data;
cpu_interrupt_type.enable(irq);
}
return 0;
}
int txn_claim_irq(int irq)
{
return cpu_claim_irq(irq, NULL, NULL) ? -1 : irq;
}
/*
* The bits_wide parameter accommodates the limitations of the HW/SW which
* use these bits:
* Legacy PA I/O (GSC/NIO): 5 bits (architected EIM register)
* V-class (EPIC): 6 bits
* N/L/A-class (iosapic): 8 bits
* PCI 2.2 MSI: 16 bits
* Some PCI devices: 32 bits (Symbios SCSI/ATM/HyperFabric)
*
* On the service provider side:
* o PA 1.1 (and PA2.0 narrow mode) 5-bits (width of EIR register)
* o PA 2.0 wide mode 6-bits (per processor)
* o IA64 8-bits (0-256 total)
*
* So a Legacy PA I/O device on a PA 2.0 box can't use all the bits supported
* by the processor...and the N/L-class I/O subsystem supports more bits than
* PA2.0 has. The first case is the problem.
*/
int txn_alloc_irq(unsigned int bits_wide)
{
int irq;
/* never return irq 0 cause that's the interval timer */
for (irq = CPU_IRQ_BASE + 1; irq <= CPU_IRQ_MAX; irq++) {
if (cpu_claim_irq(irq, NULL, NULL) < 0)
continue;
if ((irq - CPU_IRQ_BASE) >= (1 << bits_wide))
continue;
return irq;
}
/* unlikely, but be prepared */
return -1;
}
unsigned long txn_affinity_addr(unsigned int irq, int cpu)
{
#ifdef CONFIG_SMP
irq_desc[irq].affinity = cpumask_of_cpu(cpu);
#endif
return cpu_data[cpu].txn_addr;
}
unsigned long txn_alloc_addr(unsigned int virt_irq)
{
static int next_cpu = -1;
next_cpu++; /* assign to "next" CPU we want this bugger on */
/* validate entry */
while ((next_cpu < NR_CPUS) && (!cpu_data[next_cpu].txn_addr ||
!cpu_online(next_cpu)))
next_cpu++;
if (next_cpu >= NR_CPUS)
next_cpu = 0; /* nothing else, assign monarch */
return txn_affinity_addr(virt_irq, next_cpu);
}
unsigned int txn_alloc_data(unsigned int virt_irq)
{
return virt_irq - CPU_IRQ_BASE;
}
static inline int eirr_to_irq(unsigned long eirr)
{
int bit = fls_long(eirr);
return (BITS_PER_LONG - bit) + TIMER_IRQ;
}
/* ONLY called from entry.S:intr_extint() */
void do_cpu_irq_mask(struct pt_regs *regs)
{
struct pt_regs *old_regs;
unsigned long eirr_val;
int irq, cpu = smp_processor_id();
#ifdef CONFIG_SMP
cpumask_t dest;
#endif
old_regs = set_irq_regs(regs);
local_irq_disable();
irq_enter();
eirr_val = mfctl(23) & cpu_eiem & per_cpu(local_ack_eiem, cpu);
if (!eirr_val)
goto set_out;
irq = eirr_to_irq(eirr_val);
#ifdef CONFIG_SMP
dest = irq_desc[irq].affinity;
if (CHECK_IRQ_PER_CPU(irq_desc[irq].status) &&
!cpu_isset(smp_processor_id(), dest)) {
int cpu = first_cpu(dest);
printk(KERN_DEBUG "redirecting irq %d from CPU %d to %d\n",
irq, smp_processor_id(), cpu);
gsc_writel(irq + CPU_IRQ_BASE,
cpu_data[cpu].hpa);
goto set_out;
}
#endif
__do_IRQ(irq);
out:
irq_exit();
set_irq_regs(old_regs);
return;
set_out:
set_eiem(cpu_eiem & per_cpu(local_ack_eiem, cpu));
goto out;
}
static struct irqaction timer_action = {
.handler = timer_interrupt,
.name = "timer",
.flags = IRQF_DISABLED | IRQF_TIMER | IRQF_PERCPU | IRQF_IRQPOLL,
};
#ifdef CONFIG_SMP
static struct irqaction ipi_action = {
.handler = ipi_interrupt,
.name = "IPI",
.flags = IRQF_DISABLED | IRQF_PERCPU,
};
#endif
static void claim_cpu_irqs(void)
{
int i;
for (i = CPU_IRQ_BASE; i <= CPU_IRQ_MAX; i++) {
irq_desc[i].chip = &cpu_interrupt_type;
}
irq_desc[TIMER_IRQ].action = &timer_action;
irq_desc[TIMER_IRQ].status |= IRQ_PER_CPU;
#ifdef CONFIG_SMP
irq_desc[IPI_IRQ].action = &ipi_action;
irq_desc[IPI_IRQ].status = IRQ_PER_CPU;
#endif
}
void __init init_IRQ(void)
{
local_irq_disable(); /* PARANOID - should already be disabled */
mtctl(~0UL, 23); /* EIRR : clear all pending external intr */
claim_cpu_irqs();
#ifdef CONFIG_SMP
if (!cpu_eiem)
cpu_eiem = EIEM_MASK(IPI_IRQ) | EIEM_MASK(TIMER_IRQ);
#else
cpu_eiem = EIEM_MASK(TIMER_IRQ);
#endif
set_eiem(cpu_eiem); /* EIEM : enable all external intr */
}
void ack_bad_irq(unsigned int irq)
{
printk("unexpected IRQ %d\n", irq);
}