linux_dsm_epyc7002/arch/mips/sgi-ip27/ip27-irq.c
Alex Smith 4ace6139bf MIPS: SMP: Don't increment irq_count multiple times for call function IPIs
The majority of SMP platforms handle their IPIs through do_IRQ()
which calls irq_{enter/exit}(). When a call function IPI is received,
smp_call_function_interrupt() is called which also calls
irq_{enter,exit}(), meaning irq_count is raised twice.

When tick broadcasting is used (which is implemented via a call
function IPI), this incorrectly causes all CPU idle time on the core
receiving broadcast ticks to be accounted as time spent servicing
IRQs, as account_process_tick() will account as such if irq_count is
greater than 1. This results in 100% CPU usage being reported on a
core which receives its ticks via broadcast.

This patch removes the SMP smp_call_function_interrupt() wrapper which
calls irq_{enter,exit}(). Platforms which handle their IPIs through
do_IRQ() now call generic_smp_call_function_interrupt() directly to
avoid incrementing irq_count a second time. Platforms which don't
(loongson, sgi-ip27, sibyte) call generic_smp_call_function_interrupt()
wrapped in irq_{enter,exit}().

Signed-off-by: Alex Smith <alex.smith@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10770/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2015-08-03 09:25:12 +02:00

213 lines
5.5 KiB
C

/*
* ip27-irq.c: Highlevel interrupt handling for IP27 architecture.
*
* Copyright (C) 1999, 2000 Ralf Baechle (ralf@gnu.org)
* Copyright (C) 1999, 2000 Silicon Graphics, Inc.
* Copyright (C) 1999 - 2001 Kanoj Sarcar
*/
#undef DEBUG
#include <linux/init.h>
#include <linux/irq.h>
#include <linux/errno.h>
#include <linux/signal.h>
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/interrupt.h>
#include <linux/ioport.h>
#include <linux/timex.h>
#include <linux/smp.h>
#include <linux/random.h>
#include <linux/kernel.h>
#include <linux/kernel_stat.h>
#include <linux/delay.h>
#include <linux/bitops.h>
#include <asm/bootinfo.h>
#include <asm/io.h>
#include <asm/mipsregs.h>
#include <asm/processor.h>
#include <asm/sn/addrs.h>
#include <asm/sn/agent.h>
#include <asm/sn/arch.h>
#include <asm/sn/hub.h>
#include <asm/sn/intr.h>
/*
* Linux has a controller-independent x86 interrupt architecture.
* every controller has a 'controller-template', that is used
* by the main code to do the right thing. Each driver-visible
* interrupt source is transparently wired to the appropriate
* controller. Thus drivers need not be aware of the
* interrupt-controller.
*
* Various interrupt controllers we handle: 8259 PIC, SMP IO-APIC,
* PIIX4's internal 8259 PIC and SGI's Visual Workstation Cobalt (IO-)APIC.
* (IO-APICs assumed to be messaging to Pentium local-APICs)
*
* the code is designed to be easily extended with new/different
* interrupt controllers, without having to do assembly magic.
*/
extern asmlinkage void ip27_irq(void);
/*
* Find first bit set
*/
static int ms1bit(unsigned long x)
{
int b = 0, s;
s = 16; if (x >> 16 == 0) s = 0; b += s; x >>= s;
s = 8; if (x >> 8 == 0) s = 0; b += s; x >>= s;
s = 4; if (x >> 4 == 0) s = 0; b += s; x >>= s;
s = 2; if (x >> 2 == 0) s = 0; b += s; x >>= s;
s = 1; if (x >> 1 == 0) s = 0; b += s;
return b;
}
/*
* This code is unnecessarily complex, because we do
* intr enabling. Basically, once we grab the set of intrs we need
* to service, we must mask _all_ these interrupts; firstly, to make
* sure the same intr does not intr again, causing recursion that
* can lead to stack overflow. Secondly, we can not just mask the
* one intr we are do_IRQing, because the non-masked intrs in the
* first set might intr again, causing multiple servicings of the
* same intr. This effect is mostly seen for intercpu intrs.
* Kanoj 05.13.00
*/
static void ip27_do_irq_mask0(void)
{
int irq, swlevel;
hubreg_t pend0, mask0;
cpuid_t cpu = smp_processor_id();
int pi_int_mask0 =
(cputoslice(cpu) == 0) ? PI_INT_MASK0_A : PI_INT_MASK0_B;
/* copied from Irix intpend0() */
pend0 = LOCAL_HUB_L(PI_INT_PEND0);
mask0 = LOCAL_HUB_L(pi_int_mask0);
pend0 &= mask0; /* Pick intrs we should look at */
if (!pend0)
return;
swlevel = ms1bit(pend0);
#ifdef CONFIG_SMP
if (pend0 & (1UL << CPU_RESCHED_A_IRQ)) {
LOCAL_HUB_CLR_INTR(CPU_RESCHED_A_IRQ);
scheduler_ipi();
} else if (pend0 & (1UL << CPU_RESCHED_B_IRQ)) {
LOCAL_HUB_CLR_INTR(CPU_RESCHED_B_IRQ);
scheduler_ipi();
} else if (pend0 & (1UL << CPU_CALL_A_IRQ)) {
LOCAL_HUB_CLR_INTR(CPU_CALL_A_IRQ);
irq_enter();
generic_smp_call_function_interrupt();
irq_exit();
} else if (pend0 & (1UL << CPU_CALL_B_IRQ)) {
LOCAL_HUB_CLR_INTR(CPU_CALL_B_IRQ);
irq_enter();
generic_smp_call_function_interrupt();
irq_exit();
} else
#endif
{
/* "map" swlevel to irq */
struct slice_data *si = cpu_data[cpu].data;
irq = si->level_to_irq[swlevel];
do_IRQ(irq);
}
LOCAL_HUB_L(PI_INT_PEND0);
}
static void ip27_do_irq_mask1(void)
{
int irq, swlevel;
hubreg_t pend1, mask1;
cpuid_t cpu = smp_processor_id();
int pi_int_mask1 = (cputoslice(cpu) == 0) ? PI_INT_MASK1_A : PI_INT_MASK1_B;
struct slice_data *si = cpu_data[cpu].data;
/* copied from Irix intpend0() */
pend1 = LOCAL_HUB_L(PI_INT_PEND1);
mask1 = LOCAL_HUB_L(pi_int_mask1);
pend1 &= mask1; /* Pick intrs we should look at */
if (!pend1)
return;
swlevel = ms1bit(pend1);
/* "map" swlevel to irq */
irq = si->level_to_irq[swlevel];
LOCAL_HUB_CLR_INTR(swlevel);
do_IRQ(irq);
LOCAL_HUB_L(PI_INT_PEND1);
}
static void ip27_prof_timer(void)
{
panic("CPU %d got a profiling interrupt", smp_processor_id());
}
static void ip27_hub_error(void)
{
panic("CPU %d got a hub error interrupt", smp_processor_id());
}
asmlinkage void plat_irq_dispatch(void)
{
unsigned long pending = read_c0_cause() & read_c0_status();
extern unsigned int rt_timer_irq;
if (pending & CAUSEF_IP4)
do_IRQ(rt_timer_irq);
else if (pending & CAUSEF_IP2) /* PI_INT_PEND_0 or CC_PEND_{A|B} */
ip27_do_irq_mask0();
else if (pending & CAUSEF_IP3) /* PI_INT_PEND_1 */
ip27_do_irq_mask1();
else if (pending & CAUSEF_IP5)
ip27_prof_timer();
else if (pending & CAUSEF_IP6)
ip27_hub_error();
}
void __init arch_init_irq(void)
{
}
void install_ipi(void)
{
int slice = LOCAL_HUB_L(PI_CPU_NUM);
int cpu = smp_processor_id();
struct slice_data *si = cpu_data[cpu].data;
struct hub_data *hub = hub_data(cpu_to_node(cpu));
int resched, call;
resched = CPU_RESCHED_A_IRQ + slice;
__set_bit(resched, hub->irq_alloc_mask);
__set_bit(resched, si->irq_enable_mask);
LOCAL_HUB_CLR_INTR(resched);
call = CPU_CALL_A_IRQ + slice;
__set_bit(call, hub->irq_alloc_mask);
__set_bit(call, si->irq_enable_mask);
LOCAL_HUB_CLR_INTR(call);
if (slice == 0) {
LOCAL_HUB_S(PI_INT_MASK0_A, si->irq_enable_mask[0]);
LOCAL_HUB_S(PI_INT_MASK1_A, si->irq_enable_mask[1]);
} else {
LOCAL_HUB_S(PI_INT_MASK0_B, si->irq_enable_mask[0]);
LOCAL_HUB_S(PI_INT_MASK1_B, si->irq_enable_mask[1]);
}
}