9bcbdd9c58
Now that range timers and deferred timers are common, I found a
problem with these using the "perf timechart" tool. Frans Pop also
reported high scheduler latencies via LatencyTop, when using
iwlagn.
It turns out that on x86, these two 'opportunistic' timers only get
checked when another "real" timer happens. These opportunistic
timers have the objective to save power by hitchhiking on other
wakeups, as to avoid CPU wakeups by themselves as much as possible.
The change in this patch runs this check not only at timer
interrupts, but at all (device) interrupts. The effect is that:
1) the deferred timers/range timers get delayed less
2) the range timers cause less wakeups by themselves because
the percentage of hitchhiking on existing wakeup events goes up.
I've verified the working of the patch using "perf timechart", the
original exposed bug is gone with this patch. Frans also reported
success - the latencies are now down in the expected ~10 msec
range.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Tested-by: Frans Pop <elendil@planet.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091008064041.67219b13@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
241 lines
6.8 KiB
C
241 lines
6.8 KiB
C
/*
|
|
* Intel SMP support routines.
|
|
*
|
|
* (c) 1995 Alan Cox, Building #3 <alan@lxorguk.ukuu.org.uk>
|
|
* (c) 1998-99, 2000, 2009 Ingo Molnar <mingo@redhat.com>
|
|
* (c) 2002,2003 Andi Kleen, SuSE Labs.
|
|
*
|
|
* i386 and x86_64 integration by Glauber Costa <gcosta@redhat.com>
|
|
*
|
|
* This code is released under the GNU General Public License version 2 or
|
|
* later.
|
|
*/
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/mm.h>
|
|
#include <linux/delay.h>
|
|
#include <linux/spinlock.h>
|
|
#include <linux/kernel_stat.h>
|
|
#include <linux/mc146818rtc.h>
|
|
#include <linux/cache.h>
|
|
#include <linux/interrupt.h>
|
|
#include <linux/cpu.h>
|
|
|
|
#include <asm/mtrr.h>
|
|
#include <asm/tlbflush.h>
|
|
#include <asm/mmu_context.h>
|
|
#include <asm/proto.h>
|
|
#include <asm/apic.h>
|
|
/*
|
|
* Some notes on x86 processor bugs affecting SMP operation:
|
|
*
|
|
* Pentium, Pentium Pro, II, III (and all CPUs) have bugs.
|
|
* The Linux implications for SMP are handled as follows:
|
|
*
|
|
* Pentium III / [Xeon]
|
|
* None of the E1AP-E3AP errata are visible to the user.
|
|
*
|
|
* E1AP. see PII A1AP
|
|
* E2AP. see PII A2AP
|
|
* E3AP. see PII A3AP
|
|
*
|
|
* Pentium II / [Xeon]
|
|
* None of the A1AP-A3AP errata are visible to the user.
|
|
*
|
|
* A1AP. see PPro 1AP
|
|
* A2AP. see PPro 2AP
|
|
* A3AP. see PPro 7AP
|
|
*
|
|
* Pentium Pro
|
|
* None of 1AP-9AP errata are visible to the normal user,
|
|
* except occasional delivery of 'spurious interrupt' as trap #15.
|
|
* This is very rare and a non-problem.
|
|
*
|
|
* 1AP. Linux maps APIC as non-cacheable
|
|
* 2AP. worked around in hardware
|
|
* 3AP. fixed in C0 and above steppings microcode update.
|
|
* Linux does not use excessive STARTUP_IPIs.
|
|
* 4AP. worked around in hardware
|
|
* 5AP. symmetric IO mode (normal Linux operation) not affected.
|
|
* 'noapic' mode has vector 0xf filled out properly.
|
|
* 6AP. 'noapic' mode might be affected - fixed in later steppings
|
|
* 7AP. We do not assume writes to the LVT deassering IRQs
|
|
* 8AP. We do not enable low power mode (deep sleep) during MP bootup
|
|
* 9AP. We do not use mixed mode
|
|
*
|
|
* Pentium
|
|
* There is a marginal case where REP MOVS on 100MHz SMP
|
|
* machines with B stepping processors can fail. XXX should provide
|
|
* an L1cache=Writethrough or L1cache=off option.
|
|
*
|
|
* B stepping CPUs may hang. There are hardware work arounds
|
|
* for this. We warn about it in case your board doesn't have the work
|
|
* arounds. Basically that's so I can tell anyone with a B stepping
|
|
* CPU and SMP problems "tough".
|
|
*
|
|
* Specific items [From Pentium Processor Specification Update]
|
|
*
|
|
* 1AP. Linux doesn't use remote read
|
|
* 2AP. Linux doesn't trust APIC errors
|
|
* 3AP. We work around this
|
|
* 4AP. Linux never generated 3 interrupts of the same priority
|
|
* to cause a lost local interrupt.
|
|
* 5AP. Remote read is never used
|
|
* 6AP. not affected - worked around in hardware
|
|
* 7AP. not affected - worked around in hardware
|
|
* 8AP. worked around in hardware - we get explicit CS errors if not
|
|
* 9AP. only 'noapic' mode affected. Might generate spurious
|
|
* interrupts, we log only the first one and count the
|
|
* rest silently.
|
|
* 10AP. not affected - worked around in hardware
|
|
* 11AP. Linux reads the APIC between writes to avoid this, as per
|
|
* the documentation. Make sure you preserve this as it affects
|
|
* the C stepping chips too.
|
|
* 12AP. not affected - worked around in hardware
|
|
* 13AP. not affected - worked around in hardware
|
|
* 14AP. we always deassert INIT during bootup
|
|
* 15AP. not affected - worked around in hardware
|
|
* 16AP. not affected - worked around in hardware
|
|
* 17AP. not affected - worked around in hardware
|
|
* 18AP. not affected - worked around in hardware
|
|
* 19AP. not affected - worked around in BIOS
|
|
*
|
|
* If this sounds worrying believe me these bugs are either ___RARE___,
|
|
* or are signal timing bugs worked around in hardware and there's
|
|
* about nothing of note with C stepping upwards.
|
|
*/
|
|
|
|
/*
|
|
* this function sends a 'reschedule' IPI to another CPU.
|
|
* it goes straight through and wastes no time serializing
|
|
* anything. Worst case is that we lose a reschedule ...
|
|
*/
|
|
static void native_smp_send_reschedule(int cpu)
|
|
{
|
|
if (unlikely(cpu_is_offline(cpu))) {
|
|
WARN_ON(1);
|
|
return;
|
|
}
|
|
apic->send_IPI_mask(cpumask_of(cpu), RESCHEDULE_VECTOR);
|
|
}
|
|
|
|
void native_send_call_func_single_ipi(int cpu)
|
|
{
|
|
apic->send_IPI_mask(cpumask_of(cpu), CALL_FUNCTION_SINGLE_VECTOR);
|
|
}
|
|
|
|
void native_send_call_func_ipi(const struct cpumask *mask)
|
|
{
|
|
cpumask_var_t allbutself;
|
|
|
|
if (!alloc_cpumask_var(&allbutself, GFP_ATOMIC)) {
|
|
apic->send_IPI_mask(mask, CALL_FUNCTION_VECTOR);
|
|
return;
|
|
}
|
|
|
|
cpumask_copy(allbutself, cpu_online_mask);
|
|
cpumask_clear_cpu(smp_processor_id(), allbutself);
|
|
|
|
if (cpumask_equal(mask, allbutself) &&
|
|
cpumask_equal(cpu_online_mask, cpu_callout_mask))
|
|
apic->send_IPI_allbutself(CALL_FUNCTION_VECTOR);
|
|
else
|
|
apic->send_IPI_mask(mask, CALL_FUNCTION_VECTOR);
|
|
|
|
free_cpumask_var(allbutself);
|
|
}
|
|
|
|
/*
|
|
* this function calls the 'stop' function on all other CPUs in the system.
|
|
*/
|
|
|
|
asmlinkage void smp_reboot_interrupt(void)
|
|
{
|
|
ack_APIC_irq();
|
|
irq_enter();
|
|
stop_this_cpu(NULL);
|
|
irq_exit();
|
|
}
|
|
|
|
static void native_smp_send_stop(void)
|
|
{
|
|
unsigned long flags;
|
|
unsigned long wait;
|
|
|
|
if (reboot_force)
|
|
return;
|
|
|
|
/*
|
|
* Use an own vector here because smp_call_function
|
|
* does lots of things not suitable in a panic situation.
|
|
* On most systems we could also use an NMI here,
|
|
* but there are a few systems around where NMI
|
|
* is problematic so stay with an non NMI for now
|
|
* (this implies we cannot stop CPUs spinning with irq off
|
|
* currently)
|
|
*/
|
|
if (num_online_cpus() > 1) {
|
|
apic->send_IPI_allbutself(REBOOT_VECTOR);
|
|
|
|
/* Don't wait longer than a second */
|
|
wait = USEC_PER_SEC;
|
|
while (num_online_cpus() > 1 && wait--)
|
|
udelay(1);
|
|
}
|
|
|
|
local_irq_save(flags);
|
|
disable_local_APIC();
|
|
local_irq_restore(flags);
|
|
}
|
|
|
|
/*
|
|
* Reschedule call back. Nothing to do,
|
|
* all the work is done automatically when
|
|
* we return from the interrupt.
|
|
*/
|
|
void smp_reschedule_interrupt(struct pt_regs *regs)
|
|
{
|
|
ack_APIC_irq();
|
|
inc_irq_stat(irq_resched_count);
|
|
run_local_timers();
|
|
/*
|
|
* KVM uses this interrupt to force a cpu out of guest mode
|
|
*/
|
|
}
|
|
|
|
void smp_call_function_interrupt(struct pt_regs *regs)
|
|
{
|
|
ack_APIC_irq();
|
|
irq_enter();
|
|
generic_smp_call_function_interrupt();
|
|
inc_irq_stat(irq_call_count);
|
|
irq_exit();
|
|
}
|
|
|
|
void smp_call_function_single_interrupt(struct pt_regs *regs)
|
|
{
|
|
ack_APIC_irq();
|
|
irq_enter();
|
|
generic_smp_call_function_single_interrupt();
|
|
inc_irq_stat(irq_call_count);
|
|
irq_exit();
|
|
}
|
|
|
|
struct smp_ops smp_ops = {
|
|
.smp_prepare_boot_cpu = native_smp_prepare_boot_cpu,
|
|
.smp_prepare_cpus = native_smp_prepare_cpus,
|
|
.smp_cpus_done = native_smp_cpus_done,
|
|
|
|
.smp_send_stop = native_smp_send_stop,
|
|
.smp_send_reschedule = native_smp_send_reschedule,
|
|
|
|
.cpu_up = native_cpu_up,
|
|
.cpu_die = native_cpu_die,
|
|
.cpu_disable = native_cpu_disable,
|
|
.play_dead = native_play_dead,
|
|
|
|
.send_call_func_ipi = native_send_call_func_ipi,
|
|
.send_call_func_single_ipi = native_send_call_func_single_ipi,
|
|
};
|
|
EXPORT_SYMBOL_GPL(smp_ops);
|