linux_dsm_epyc7002/arch/powerpc/kernel/smp.c

833 lines
18 KiB
C
Raw Normal View History

/*
* SMP support for ppc.
*
* Written by Cort Dougan (cort@cs.nmt.edu) borrowing a great
* deal of code from the sparc and intel versions.
*
* Copyright (C) 1999 Cort Dougan <cort@cs.nmt.edu>
*
* PowerPC-64 Support added by Dave Engebretsen, Peter Bergner, and
* Mike Corrigan {engebret|bergner|mikec}@us.ibm.com
*
* 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 of the License, or (at your option) any later version.
*/
#undef DEBUG
#include <linux/kernel.h>
#include <linux/export.h>
#include <linux/sched.h>
#include <linux/smp.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/cache.h>
#include <linux/err.h>
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 05:29:42 +07:00
#include <linux/device.h>
#include <linux/cpu.h>
#include <linux/notifier.h>
#include <linux/topology.h>
#include <asm/ptrace.h>
#include <linux/atomic.h>
#include <asm/irq.h>
#include <asm/hw_irq.h>
#include <asm/kvm_ppc.h>
#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/prom.h>
#include <asm/smp.h>
#include <asm/time.h>
#include <asm/machdep.h>
#include <asm/cputhreads.h>
#include <asm/cputable.h>
#include <asm/mpic.h>
#include <asm/vdso_datapage.h>
#ifdef CONFIG_PPC64
#include <asm/paca.h>
#endif
#include <asm/vdso.h>
#include <asm/debug.h>
#include <asm/kexec.h>
#ifdef DEBUG
#include <asm/udbg.h>
#define DBG(fmt...) udbg_printf(fmt)
#else
#define DBG(fmt...)
#endif
#ifdef CONFIG_HOTPLUG_CPU
/* State of each CPU during hotplug phases */
static DEFINE_PER_CPU(int, cpu_state) = { 0 };
#endif
struct thread_info *secondary_ti;
DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map);
DEFINE_PER_CPU(cpumask_var_t, cpu_core_map);
EXPORT_PER_CPU_SYMBOL(cpu_sibling_map);
EXPORT_PER_CPU_SYMBOL(cpu_core_map);
/* SMP operations for this machine */
struct smp_ops_t *smp_ops;
/* Can't be static due to PowerMac hackery */
volatile unsigned int cpu_callin_map[NR_CPUS];
int smt_enabled_at_boot = 1;
static void (*crash_ipi_function_ptr)(struct pt_regs *) = NULL;
/*
* Returns 1 if the specified cpu should be brought up during boot.
* Used to inhibit booting threads if they've been disabled or
* limited on the command line
*/
int smp_generic_cpu_bootable(unsigned int nr)
{
/* Special case - we inhibit secondary thread startup
* during boot if the user requests it.
*/
if (system_state == SYSTEM_BOOTING && cpu_has_feature(CPU_FTR_SMT)) {
if (!smt_enabled_at_boot && cpu_thread_in_core(nr) != 0)
return 0;
if (smt_enabled_at_boot
&& cpu_thread_in_core(nr) >= smt_enabled_at_boot)
return 0;
}
return 1;
}
#ifdef CONFIG_PPC64
int smp_generic_kick_cpu(int nr)
{
BUG_ON(nr < 0 || nr >= NR_CPUS);
/*
* The processor is currently spinning, waiting for the
* cpu_start field to become non-zero After we set cpu_start,
* the processor will continue on to secondary_start
*/
if (!paca[nr].cpu_start) {
paca[nr].cpu_start = 1;
smp_mb();
return 0;
}
#ifdef CONFIG_HOTPLUG_CPU
/*
* Ok it's not there, so it might be soft-unplugged, let's
* try to bring it back
*/
generic_set_cpu_up(nr);
smp_wmb();
smp_send_reschedule(nr);
#endif /* CONFIG_HOTPLUG_CPU */
return 0;
}
#endif /* CONFIG_PPC64 */
static irqreturn_t call_function_action(int irq, void *data)
{
generic_smp_call_function_interrupt();
return IRQ_HANDLED;
}
static irqreturn_t reschedule_action(int irq, void *data)
{
scheduler_ipi();
return IRQ_HANDLED;
}
static irqreturn_t tick_broadcast_ipi_action(int irq, void *data)
{
tick_broadcast_ipi_handler();
return IRQ_HANDLED;
}
static irqreturn_t debug_ipi_action(int irq, void *data)
{
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
if (crash_ipi_function_ptr) {
crash_ipi_function_ptr(get_irq_regs());
return IRQ_HANDLED;
}
#ifdef CONFIG_DEBUGGER
debugger_ipi(get_irq_regs());
#endif /* CONFIG_DEBUGGER */
return IRQ_HANDLED;
}
static irq_handler_t smp_ipi_action[] = {
[PPC_MSG_CALL_FUNCTION] = call_function_action,
[PPC_MSG_RESCHEDULE] = reschedule_action,
[PPC_MSG_TICK_BROADCAST] = tick_broadcast_ipi_action,
[PPC_MSG_DEBUGGER_BREAK] = debug_ipi_action,
};
const char *smp_ipi_name[] = {
[PPC_MSG_CALL_FUNCTION] = "ipi call function",
[PPC_MSG_RESCHEDULE] = "ipi reschedule",
[PPC_MSG_TICK_BROADCAST] = "ipi tick-broadcast",
[PPC_MSG_DEBUGGER_BREAK] = "ipi debugger",
};
/* optional function to request ipi, for controllers with >= 4 ipis */
int smp_request_message_ipi(int virq, int msg)
{
int err;
if (msg < 0 || msg > PPC_MSG_DEBUGGER_BREAK) {
return -EINVAL;
}
#if !defined(CONFIG_DEBUGGER) && !defined(CONFIG_KEXEC)
if (msg == PPC_MSG_DEBUGGER_BREAK) {
return 1;
}
#endif
err = request_irq(virq, smp_ipi_action[msg],
IRQF_PERCPU | IRQF_NO_THREAD | IRQF_NO_SUSPEND,
smp_ipi_name[msg], NULL);
WARN(err < 0, "unable to request_irq %d for %s (rc %d)\n",
virq, smp_ipi_name[msg], err);
return err;
}
#ifdef CONFIG_PPC_SMP_MUXED_IPI
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
struct cpu_messages {
int messages; /* current messages */
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
unsigned long data; /* data for cause ipi */
};
static DEFINE_PER_CPU_SHARED_ALIGNED(struct cpu_messages, ipi_message);
void smp_muxed_ipi_set_data(int cpu, unsigned long data)
{
struct cpu_messages *info = &per_cpu(ipi_message, cpu);
info->data = data;
}
void smp_muxed_ipi_message_pass(int cpu, int msg)
{
struct cpu_messages *info = &per_cpu(ipi_message, cpu);
char *message = (char *)&info->messages;
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
powerpc: Make sure IPI handlers see data written by IPI senders We have been observing hangs, both of KVM guest vcpu tasks and more generally, where a process that is woken doesn't properly wake up and continue to run, but instead sticks in TASK_WAKING state. This happens because the update of rq->wake_list in ttwu_queue_remote() is not ordered with the update of ipi_message in smp_muxed_ipi_message_pass(), and the reading of rq->wake_list in scheduler_ipi() is not ordered with the reading of ipi_message in smp_ipi_demux(). Thus it is possible for the IPI receiver not to see the updated rq->wake_list and therefore conclude that there is nothing for it to do. In order to make sure that anything done before smp_send_reschedule() is ordered before anything done in the resulting call to scheduler_ipi(), this adds barriers in smp_muxed_message_pass() and smp_ipi_demux(). The barrier in smp_muxed_message_pass() is a full barrier to ensure that there is a full ordering between the smp_send_reschedule() caller and scheduler_ipi(). In smp_ipi_demux(), we use xchg() rather than xchg_local() because xchg() includes release and acquire barriers. Using xchg() rather than xchg_local() makes sense given that ipi_message is not just accessed locally. This moves the barrier between setting the message and calling the cause_ipi() function into the individual cause_ipi implementations. Most of them -- those that used outb, out_8 or similar -- already had a full barrier because out_8 etc. include a sync before the MMIO store. This adds an explicit barrier in the two remaining cases. These changes made no measurable difference to the speed of IPIs as measured using a simple ping-pong latency test across two CPUs on different cores of a POWER7 machine. The analysis of the reason why processes were not waking up properly is due to Milton Miller. Cc: stable@vger.kernel.org # v3.0+ Reported-by: Milton Miller <miltonm@bga.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-09-05 01:33:08 +07:00
/*
* Order previous accesses before accesses in the IPI handler.
*/
smp_mb();
message[msg] = 1;
powerpc: Make sure IPI handlers see data written by IPI senders We have been observing hangs, both of KVM guest vcpu tasks and more generally, where a process that is woken doesn't properly wake up and continue to run, but instead sticks in TASK_WAKING state. This happens because the update of rq->wake_list in ttwu_queue_remote() is not ordered with the update of ipi_message in smp_muxed_ipi_message_pass(), and the reading of rq->wake_list in scheduler_ipi() is not ordered with the reading of ipi_message in smp_ipi_demux(). Thus it is possible for the IPI receiver not to see the updated rq->wake_list and therefore conclude that there is nothing for it to do. In order to make sure that anything done before smp_send_reschedule() is ordered before anything done in the resulting call to scheduler_ipi(), this adds barriers in smp_muxed_message_pass() and smp_ipi_demux(). The barrier in smp_muxed_message_pass() is a full barrier to ensure that there is a full ordering between the smp_send_reschedule() caller and scheduler_ipi(). In smp_ipi_demux(), we use xchg() rather than xchg_local() because xchg() includes release and acquire barriers. Using xchg() rather than xchg_local() makes sense given that ipi_message is not just accessed locally. This moves the barrier between setting the message and calling the cause_ipi() function into the individual cause_ipi implementations. Most of them -- those that used outb, out_8 or similar -- already had a full barrier because out_8 etc. include a sync before the MMIO store. This adds an explicit barrier in the two remaining cases. These changes made no measurable difference to the speed of IPIs as measured using a simple ping-pong latency test across two CPUs on different cores of a POWER7 machine. The analysis of the reason why processes were not waking up properly is due to Milton Miller. Cc: stable@vger.kernel.org # v3.0+ Reported-by: Milton Miller <miltonm@bga.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-09-05 01:33:08 +07:00
/*
* cause_ipi functions are required to include a full barrier
* before doing whatever causes the IPI.
*/
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
smp_ops->cause_ipi(cpu, info->data);
}
#ifdef __BIG_ENDIAN__
#define IPI_MESSAGE(A) (1 << (24 - 8 * (A)))
#else
#define IPI_MESSAGE(A) (1 << (8 * (A)))
#endif
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
irqreturn_t smp_ipi_demux(void)
{
powerpc: Replace __get_cpu_var uses This still has not been merged and now powerpc is the only arch that does not have this change. Sorry about missing linuxppc-dev before. V2->V2 - Fix up to work against 3.18-rc1 __get_cpu_var() is used for multiple purposes in the kernel source. One of them is address calculation via the form &__get_cpu_var(x). This calculates the address for the instance of the percpu variable of the current processor based on an offset. Other use cases are for storing and retrieving data from the current processors percpu area. __get_cpu_var() can be used as an lvalue when writing data or on the right side of an assignment. __get_cpu_var() is defined as : __get_cpu_var() always only does an address determination. However, store and retrieve operations could use a segment prefix (or global register on other platforms) to avoid the address calculation. this_cpu_write() and this_cpu_read() can directly take an offset into a percpu area and use optimized assembly code to read and write per cpu variables. This patch converts __get_cpu_var into either an explicit address calculation using this_cpu_ptr() or into a use of this_cpu operations that use the offset. Thereby address calculations are avoided and less registers are used when code is generated. At the end of the patch set all uses of __get_cpu_var have been removed so the macro is removed too. The patch set includes passes over all arches as well. Once these operations are used throughout then specialized macros can be defined in non -x86 arches as well in order to optimize per cpu access by f.e. using a global register that may be set to the per cpu base. Transformations done to __get_cpu_var() 1. Determine the address of the percpu instance of the current processor. DEFINE_PER_CPU(int, y); int *x = &__get_cpu_var(y); Converts to int *x = this_cpu_ptr(&y); 2. Same as #1 but this time an array structure is involved. DEFINE_PER_CPU(int, y[20]); int *x = __get_cpu_var(y); Converts to int *x = this_cpu_ptr(y); 3. Retrieve the content of the current processors instance of a per cpu variable. DEFINE_PER_CPU(int, y); int x = __get_cpu_var(y) Converts to int x = __this_cpu_read(y); 4. Retrieve the content of a percpu struct DEFINE_PER_CPU(struct mystruct, y); struct mystruct x = __get_cpu_var(y); Converts to memcpy(&x, this_cpu_ptr(&y), sizeof(x)); 5. Assignment to a per cpu variable DEFINE_PER_CPU(int, y) __get_cpu_var(y) = x; Converts to __this_cpu_write(y, x); 6. Increment/Decrement etc of a per cpu variable DEFINE_PER_CPU(int, y); __get_cpu_var(y)++ Converts to __this_cpu_inc(y) Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> CC: Paul Mackerras <paulus@samba.org> Signed-off-by: Christoph Lameter <cl@linux.com> [mpe: Fix build errors caused by set/or_softirq_pending(), and rework assignment in __set_breakpoint() to use memcpy().] Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2014-10-22 03:23:25 +07:00
struct cpu_messages *info = this_cpu_ptr(&ipi_message);
unsigned int all;
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
mb(); /* order any irq clear */
do {
powerpc: Make sure IPI handlers see data written by IPI senders We have been observing hangs, both of KVM guest vcpu tasks and more generally, where a process that is woken doesn't properly wake up and continue to run, but instead sticks in TASK_WAKING state. This happens because the update of rq->wake_list in ttwu_queue_remote() is not ordered with the update of ipi_message in smp_muxed_ipi_message_pass(), and the reading of rq->wake_list in scheduler_ipi() is not ordered with the reading of ipi_message in smp_ipi_demux(). Thus it is possible for the IPI receiver not to see the updated rq->wake_list and therefore conclude that there is nothing for it to do. In order to make sure that anything done before smp_send_reschedule() is ordered before anything done in the resulting call to scheduler_ipi(), this adds barriers in smp_muxed_message_pass() and smp_ipi_demux(). The barrier in smp_muxed_message_pass() is a full barrier to ensure that there is a full ordering between the smp_send_reschedule() caller and scheduler_ipi(). In smp_ipi_demux(), we use xchg() rather than xchg_local() because xchg() includes release and acquire barriers. Using xchg() rather than xchg_local() makes sense given that ipi_message is not just accessed locally. This moves the barrier between setting the message and calling the cause_ipi() function into the individual cause_ipi implementations. Most of them -- those that used outb, out_8 or similar -- already had a full barrier because out_8 etc. include a sync before the MMIO store. This adds an explicit barrier in the two remaining cases. These changes made no measurable difference to the speed of IPIs as measured using a simple ping-pong latency test across two CPUs on different cores of a POWER7 machine. The analysis of the reason why processes were not waking up properly is due to Milton Miller. Cc: stable@vger.kernel.org # v3.0+ Reported-by: Milton Miller <miltonm@bga.com> Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-09-05 01:33:08 +07:00
all = xchg(&info->messages, 0);
if (all & IPI_MESSAGE(PPC_MSG_CALL_FUNCTION))
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
generic_smp_call_function_interrupt();
if (all & IPI_MESSAGE(PPC_MSG_RESCHEDULE))
scheduler_ipi();
if (all & IPI_MESSAGE(PPC_MSG_TICK_BROADCAST))
tick_broadcast_ipi_handler();
if (all & IPI_MESSAGE(PPC_MSG_DEBUGGER_BREAK))
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
debug_ipi_action(0, NULL);
} while (info->messages);
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
return IRQ_HANDLED;
}
#endif /* CONFIG_PPC_SMP_MUXED_IPI */
powerpc: Consolidate ipi message mux and demux Consolidate the mux and demux of ipi messages into smp.c and call a new smp_ops callback to actually trigger the ipi. The powerpc architecture code is optimised for having 4 distinct ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi single, scheduler ipi, and enter debugger). However, several interrupt controllers only provide a single software triggered interrupt that can be delivered to each cpu. To resolve this limitation, each smp_ops implementation created a per-cpu variable that is manipulated with atomic bitops. Since these lines will be contended they are optimialy marked as shared_aligned and take a full cache line for each cpu. Distro kernels may have 2 or 3 of these in their config, each taking per-cpu space even though at most one will be in use. This consolidation removes smp_message_recv and replaces the single call actions cases with direct calls from the common message recognition loop. The complicated debugger ipi case with its muxed crash handling code is moved to debug_ipi_action which is now called from the demux code (instead of the multi-message action calling smp_message_recv). I put a call to reschedule_action to increase the likelyhood of correctly merging the anticipated scheduler_ipi() hook coming from the scheduler tree; that single required call can be inlined later. The actual message decode is a copy of the old pseries xics code with its memory barriers and cache line spacing, augmented with a per-cpu unsigned long based on the book-e doorbell code. The optional data is set via a callback from the implementation and is passed to the new cause-ipi hook along with the logical cpu number. While currently only the doorbell implemntation uses this data it should be almost zero cost to retrieve and pass it -- it adds a single register load for the argument from the same cache line to which we just completed a store and the register is dead on return from the call. I extended the data element from unsigned int to unsigned long in case some other code wanted to associate a pointer. The doorbell check_self is replaced by a call to smp_muxed_ipi_resend, conditioned on the CPU_DBELL feature. The ifdef guard could be relaxed to CONFIG_SMP but I left it with BOOKE for now. Also, the doorbell interrupt vector for book-e was not calling irq_enter and irq_exit, which throws off cpu accounting and causes code to not realize it is running in interrupt context. Add the missing calls. Signed-off-by: Milton Miller <miltonm@bga.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2011-05-11 02:29:39 +07:00
static inline void do_message_pass(int cpu, int msg)
{
if (smp_ops->message_pass)
smp_ops->message_pass(cpu, msg);
#ifdef CONFIG_PPC_SMP_MUXED_IPI
else
smp_muxed_ipi_message_pass(cpu, msg);
#endif
}
void smp_send_reschedule(int cpu)
{
if (likely(smp_ops))
do_message_pass(cpu, PPC_MSG_RESCHEDULE);
}
KVM: PPC: Add support for Book3S processors in hypervisor mode This adds support for KVM running on 64-bit Book 3S processors, specifically POWER7, in hypervisor mode. Using hypervisor mode means that the guest can use the processor's supervisor mode. That means that the guest can execute privileged instructions and access privileged registers itself without trapping to the host. This gives excellent performance, but does mean that KVM cannot emulate a processor architecture other than the one that the hardware implements. This code assumes that the guest is running paravirtualized using the PAPR (Power Architecture Platform Requirements) interface, which is the interface that IBM's PowerVM hypervisor uses. That means that existing Linux distributions that run on IBM pSeries machines will also run under KVM without modification. In order to communicate the PAPR hypercalls to qemu, this adds a new KVM_EXIT_PAPR_HCALL exit code to include/linux/kvm.h. Currently the choice between book3s_hv support and book3s_pr support (i.e. the existing code, which runs the guest in user mode) has to be made at kernel configuration time, so a given kernel binary can only do one or the other. This new book3s_hv code doesn't support MMIO emulation at present. Since we are running paravirtualized guests, this isn't a serious restriction. With the guest running in supervisor mode, most exceptions go straight to the guest. We will never get data or instruction storage or segment interrupts, alignment interrupts, decrementer interrupts, program interrupts, single-step interrupts, etc., coming to the hypervisor from the guest. Therefore this introduces a new KVMTEST_NONHV macro for the exception entry path so that we don't have to do the KVM test on entry to those exception handlers. We do however get hypervisor decrementer, hypervisor data storage, hypervisor instruction storage, and hypervisor emulation assist interrupts, so we have to handle those. In hypervisor mode, real-mode accesses can access all of RAM, not just a limited amount. Therefore we put all the guest state in the vcpu.arch and use the shadow_vcpu in the PACA only for temporary scratch space. We allocate the vcpu with kzalloc rather than vzalloc, and we don't use anything in the kvmppc_vcpu_book3s struct, so we don't allocate it. We don't have a shared page with the guest, but we still need a kvm_vcpu_arch_shared struct to store the values of various registers, so we include one in the vcpu_arch struct. The POWER7 processor has a restriction that all threads in a core have to be in the same partition. MMU-on kernel code counts as a partition (partition 0), so we have to do a partition switch on every entry to and exit from the guest. At present we require the host and guest to run in single-thread mode because of this hardware restriction. This code allocates a hashed page table for the guest and initializes it with HPTEs for the guest's Virtual Real Memory Area (VRMA). We require that the guest memory is allocated using 16MB huge pages, in order to simplify the low-level memory management. This also means that we can get away without tracking paging activity in the host for now, since huge pages can't be paged or swapped. This also adds a few new exports needed by the book3s_hv code. Signed-off-by: Paul Mackerras <paulus@samba.org> Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 07:21:34 +07:00
EXPORT_SYMBOL_GPL(smp_send_reschedule);
void arch_send_call_function_single_ipi(int cpu)
{
do_message_pass(cpu, PPC_MSG_CALL_FUNCTION);
}
void arch_send_call_function_ipi_mask(const struct cpumask *mask)
{
unsigned int cpu;
for_each_cpu(cpu, mask)
do_message_pass(cpu, PPC_MSG_CALL_FUNCTION);
}
#ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST
void tick_broadcast(const struct cpumask *mask)
{
unsigned int cpu;
for_each_cpu(cpu, mask)
do_message_pass(cpu, PPC_MSG_TICK_BROADCAST);
}
#endif
#if defined(CONFIG_DEBUGGER) || defined(CONFIG_KEXEC)
void smp_send_debugger_break(void)
{
int cpu;
int me = raw_smp_processor_id();
if (unlikely(!smp_ops))
return;
for_each_online_cpu(cpu)
if (cpu != me)
do_message_pass(cpu, PPC_MSG_DEBUGGER_BREAK);
}
#endif
#ifdef CONFIG_KEXEC
void crash_send_ipi(void (*crash_ipi_callback)(struct pt_regs *))
{
crash_ipi_function_ptr = crash_ipi_callback;
if (crash_ipi_callback) {
mb();
smp_send_debugger_break();
}
}
#endif
static void stop_this_cpu(void *dummy)
{
/* Remove this CPU */
set_cpu_online(smp_processor_id(), false);
local_irq_disable();
while (1)
;
}
void smp_send_stop(void)
{
smp_call_function(stop_this_cpu, NULL, 0);
}
struct thread_info *current_set[NR_CPUS];
static void smp_store_cpu_info(int id)
{
per_cpu(cpu_pvr, id) = mfspr(SPRN_PVR);
#ifdef CONFIG_PPC_FSL_BOOK3E
per_cpu(next_tlbcam_idx, id)
= (mfspr(SPRN_TLB1CFG) & TLBnCFG_N_ENTRY) - 1;
#endif
}
void __init smp_prepare_cpus(unsigned int max_cpus)
{
unsigned int cpu;
DBG("smp_prepare_cpus\n");
/*
* setup_cpu may need to be called on the boot cpu. We havent
* spun any cpus up but lets be paranoid.
*/
BUG_ON(boot_cpuid != smp_processor_id());
/* Fixup boot cpu */
smp_store_cpu_info(boot_cpuid);
cpu_callin_map[boot_cpuid] = 1;
for_each_possible_cpu(cpu) {
zalloc_cpumask_var_node(&per_cpu(cpu_sibling_map, cpu),
GFP_KERNEL, cpu_to_node(cpu));
zalloc_cpumask_var_node(&per_cpu(cpu_core_map, cpu),
GFP_KERNEL, cpu_to_node(cpu));
powerpc: reorder per-cpu NUMA information's initialization There is an issue currently where NUMA information is used on powerpc (and possibly ia64) before it has been read from the device-tree, which leads to large slab consumption with CONFIG_SLUB and memoryless nodes. NUMA powerpc non-boot CPU's cpu_to_node/cpu_to_mem is only accurate after start_secondary(), similar to ia64, which is invoked via smp_init(). Commit 6ee0578b4daae ("workqueue: mark init_workqueues() as early_initcall()") made init_workqueues() be invoked via do_pre_smp_initcalls(), which is obviously before the secondary processors are online. Additionally, the following commits changed init_workqueues() to use cpu_to_node to determine the node to use for kthread_create_on_node: bce903809ab3f ("workqueue: add wq_numa_tbl_len and wq_numa_possible_cpumask[]") f3f90ad469342 ("workqueue: determine NUMA node of workers accourding to the allowed cpumask") Therefore, when init_workqueues() runs, it sees all CPUs as being on Node 0. On LPARs or KVM guests where Node 0 is memoryless, this leads to a high number of slab deactivations (http://www.spinics.net/lists/linux-mm/msg67489.html). Fix this by initializing the powerpc-specific CPU<->node/local memory node mapping as early as possible, which on powerpc is do_init_bootmem(). Currently that function initializes the mapping for the boot CPU, but we extend it to setup the mapping for all possible CPUs. Then, in smp_prepare_cpus(), we can correspondingly set the per-cpu values for all possible CPUs. That ensures that before the early_initcalls run (and really as early as possible), the per-cpu NUMA mapping is accurate. While testing memoryless nodes on PowerKVM guests with a fix to the workqueue logic to use cpu_to_mem() instead of cpu_to_node(), with a guest topology of: available: 2 nodes (0-1) node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 node 0 size: 0 MB node 0 free: 0 MB node 1 cpus: 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 node 1 size: 16336 MB node 1 free: 15329 MB node distances: node 0 1 0: 10 40 1: 40 10 the slab consumption decreases from Slab: 932416 kB SUnreclaim: 902336 kB to Slab: 395264 kB SUnreclaim: 359424 kB And we a corresponding increase in the slab efficiency from slab mem objs slabs used active active ------------------------------------------------------------ kmalloc-16384 337 MB 11.28% 100.00% task_struct 288 MB 9.93% 100.00% to slab mem objs slabs used active active ------------------------------------------------------------ kmalloc-16384 37 MB 100.00% 100.00% task_struct 31 MB 100.00% 100.00% Powerpc didn't support memoryless nodes until recently (64bb80d87f01 "powerpc/numa: Enable CONFIG_HAVE_MEMORYLESS_NODES" and 8c272261194d "powerpc/numa: Enable USE_PERCPU_NUMA_NODE_ID"). Those commits also helped improve memory consumption with these kind of environments. Signed-off-by: Nishanth Aravamudan <nacc@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2014-07-18 06:15:12 +07:00
/*
* numa_node_id() works after this.
*/
if (cpu_present(cpu)) {
set_cpu_numa_node(cpu, numa_cpu_lookup_table[cpu]);
set_cpu_numa_mem(cpu,
local_memory_node(numa_cpu_lookup_table[cpu]));
}
}
cpumask_set_cpu(boot_cpuid, cpu_sibling_mask(boot_cpuid));
cpumask_set_cpu(boot_cpuid, cpu_core_mask(boot_cpuid));
if (smp_ops && smp_ops->probe)
smp_ops->probe();
}
void smp_prepare_boot_cpu(void)
{
BUG_ON(smp_processor_id() != boot_cpuid);
#ifdef CONFIG_PPC64
paca[boot_cpuid].__current = current;
#endif
set_numa_node(numa_cpu_lookup_table[boot_cpuid]);
current_set[boot_cpuid] = task_thread_info(current);
}
#ifdef CONFIG_HOTPLUG_CPU
int generic_cpu_disable(void)
{
unsigned int cpu = smp_processor_id();
if (cpu == boot_cpuid)
return -EBUSY;
set_cpu_online(cpu, false);
#ifdef CONFIG_PPC64
vdso_data->processorCount--;
#endif
migrate_irqs();
return 0;
}
void generic_cpu_die(unsigned int cpu)
{
int i;
for (i = 0; i < 100; i++) {
smp_rmb();
if (per_cpu(cpu_state, cpu) == CPU_DEAD)
return;
msleep(100);
}
printk(KERN_ERR "CPU%d didn't die...\n", cpu);
}
void generic_set_cpu_dead(unsigned int cpu)
{
per_cpu(cpu_state, cpu) = CPU_DEAD;
}
/*
* The cpu_state should be set to CPU_UP_PREPARE in kick_cpu(), otherwise
* the cpu_state is always CPU_DEAD after calling generic_set_cpu_dead(),
* which makes the delay in generic_cpu_die() not happen.
*/
void generic_set_cpu_up(unsigned int cpu)
{
per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
}
int generic_check_cpu_restart(unsigned int cpu)
{
return per_cpu(cpu_state, cpu) == CPU_UP_PREPARE;
}
static bool secondaries_inhibited(void)
{
return kvm_hv_mode_active();
}
#else /* HOTPLUG_CPU */
#define secondaries_inhibited() 0
#endif
static void cpu_idle_thread_init(unsigned int cpu, struct task_struct *idle)
{
struct thread_info *ti = task_thread_info(idle);
#ifdef CONFIG_PPC64
paca[cpu].__current = idle;
paca[cpu].kstack = (unsigned long)ti + THREAD_SIZE - STACK_FRAME_OVERHEAD;
#endif
ti->cpu = cpu;
secondary_ti = current_set[cpu] = ti;
}
int __cpu_up(unsigned int cpu, struct task_struct *tidle)
{
int rc, c;
/*
* Don't allow secondary threads to come online if inhibited
*/
if (threads_per_core > 1 && secondaries_inhibited() &&
cpu_thread_in_subcore(cpu))
return -EBUSY;
if (smp_ops == NULL ||
(smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu)))
return -EINVAL;
cpu_idle_thread_init(cpu, tidle);
/* Make sure callin-map entry is 0 (can be leftover a CPU
* hotplug
*/
cpu_callin_map[cpu] = 0;
/* The information for processor bringup must
* be written out to main store before we release
* the processor.
*/
smp_mb();
/* wake up cpus */
DBG("smp: kicking cpu %d\n", cpu);
rc = smp_ops->kick_cpu(cpu);
if (rc) {
pr_err("smp: failed starting cpu %d (rc %d)\n", cpu, rc);
return rc;
}
/*
* wait to see if the cpu made a callin (is actually up).
* use this value that I found through experimentation.
* -- Cort
*/
if (system_state < SYSTEM_RUNNING)
for (c = 50000; c && !cpu_callin_map[cpu]; c--)
udelay(100);
#ifdef CONFIG_HOTPLUG_CPU
else
/*
* CPUs can take much longer to come up in the
* hotplug case. Wait five seconds.
*/
for (c = 5000; c && !cpu_callin_map[cpu]; c--)
msleep(1);
#endif
if (!cpu_callin_map[cpu]) {
printk(KERN_ERR "Processor %u is stuck.\n", cpu);
return -ENOENT;
}
DBG("Processor %u found.\n", cpu);
if (smp_ops->give_timebase)
smp_ops->give_timebase();
powerpc/smp: Wait until secondaries are active & online Anton has a busy ppc64le KVM box where guests sometimes hit the infamous "kernel BUG at kernel/smpboot.c:134!" issue during boot: BUG_ON(td->cpu != smp_processor_id()); Basically a per CPU hotplug thread scheduled on the wrong CPU. The oops output confirms it: CPU: 0 Comm: watchdog/130 The problem is that we aren't ensuring the CPU active bit is set for the secondary before allowing the master to continue on. The master unparks the secondary CPU's kthreads and the scheduler looks for a CPU to run on. It calls select_task_rq() and realises the suggested CPU is not in the cpus_allowed mask. It then ends up in select_fallback_rq(), and since the active bit isnt't set we choose some other CPU to run on. This seems to have been introduced by 6acbfb96976f "sched: Fix hotplug vs. set_cpus_allowed_ptr()", which changed from setting active before online to setting active after online. However that was in turn fixing a bug where other code assumed an active CPU was also online, so we can't just revert that fix. The simplest fix is just to spin waiting for both active & online to be set. We already have a barrier prior to set_cpu_online() (which also sets active), to ensure all other setup is completed before online & active are set. Fixes: 6acbfb96976f ("sched: Fix hotplug vs. set_cpus_allowed_ptr()") Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Signed-off-by: Anton Blanchard <anton@samba.org> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2015-02-24 13:58:02 +07:00
/* Wait until cpu puts itself in the online & active maps */
while (!cpu_online(cpu) || !cpu_active(cpu))
cpu_relax();
return 0;
}
/* Return the value of the reg property corresponding to the given
* logical cpu.
*/
int cpu_to_core_id(int cpu)
{
struct device_node *np;
const __be32 *reg;
int id = -1;
np = of_get_cpu_node(cpu, NULL);
if (!np)
goto out;
reg = of_get_property(np, "reg", NULL);
if (!reg)
goto out;
id = be32_to_cpup(reg);
out:
of_node_put(np);
return id;
}
powerpc: Cleanup APIs for cpu/thread/core mappings These APIs take logical cpu number as input Change cpu_first_thread_in_core() to cpu_first_thread_sibling() Change cpu_last_thread_in_core() to cpu_last_thread_sibling() These APIs convert core number (index) to logical cpu/thread numbers Add cpu_first_thread_of_core(int core) Changed cpu_thread_to_core() to cpu_core_index_of_thread(int cpu) The goal is to make 'threads_per_core' accessible to the pseries_energy module. Instead of making an API to read threads_per_core, this is a higher level wrapper function to convert from logical cpu number to core number. The current APIs cpu_first_thread_in_core() and cpu_last_thread_in_core() returns logical CPU number while cpu_thread_to_core() returns core number or index which is not a logical CPU number. The new APIs are now clearly named to distinguish 'core number' versus first and last 'logical cpu number' in that core. The new APIs cpu_{first,last}_thread_sibling() work on logical cpu numbers. While cpu_first_thread_of_core() and cpu_core_index_of_thread() work on core index. Example usage: (4 threads per core system) cpu_first_thread_sibling(5) = 4 cpu_last_thread_sibling(5) = 7 cpu_core_index_of_thread(5) = 1 cpu_first_thread_of_core(1) = 4 cpu_core_index_of_thread() is used in cpu_to_drc_index() in the module and cpu_first_thread_of_core() is used in drc_index_to_cpu() in the module. Make API changes to few callers. Export symbols for use in modules. Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-10-06 15:36:59 +07:00
/* Helper routines for cpu to core mapping */
int cpu_core_index_of_thread(int cpu)
{
return cpu >> threads_shift;
}
EXPORT_SYMBOL_GPL(cpu_core_index_of_thread);
int cpu_first_thread_of_core(int core)
{
return core << threads_shift;
}
EXPORT_SYMBOL_GPL(cpu_first_thread_of_core);
static void traverse_siblings_chip_id(int cpu, bool add, int chipid)
{
const struct cpumask *mask;
struct device_node *np;
int i, plen;
const __be32 *prop;
mask = add ? cpu_online_mask : cpu_present_mask;
for_each_cpu(i, mask) {
np = of_get_cpu_node(i, NULL);
if (!np)
continue;
prop = of_get_property(np, "ibm,chip-id", &plen);
if (prop && plen == sizeof(int) &&
of_read_number(prop, 1) == chipid) {
if (add) {
cpumask_set_cpu(cpu, cpu_core_mask(i));
cpumask_set_cpu(i, cpu_core_mask(cpu));
} else {
cpumask_clear_cpu(cpu, cpu_core_mask(i));
cpumask_clear_cpu(i, cpu_core_mask(cpu));
}
}
of_node_put(np);
}
}
/* Must be called when no change can occur to cpu_present_mask,
* i.e. during cpu online or offline.
*/
static struct device_node *cpu_to_l2cache(int cpu)
{
struct device_node *np;
struct device_node *cache;
if (!cpu_present(cpu))
return NULL;
np = of_get_cpu_node(cpu, NULL);
if (np == NULL)
return NULL;
cache = of_find_next_cache_node(np);
of_node_put(np);
return cache;
}
static void traverse_core_siblings(int cpu, bool add)
{
struct device_node *l2_cache, *np;
const struct cpumask *mask;
int i, chip, plen;
const __be32 *prop;
/* First see if we have ibm,chip-id properties in cpu nodes */
np = of_get_cpu_node(cpu, NULL);
if (np) {
chip = -1;
prop = of_get_property(np, "ibm,chip-id", &plen);
if (prop && plen == sizeof(int))
chip = of_read_number(prop, 1);
of_node_put(np);
if (chip >= 0) {
traverse_siblings_chip_id(cpu, add, chip);
return;
}
}
l2_cache = cpu_to_l2cache(cpu);
mask = add ? cpu_online_mask : cpu_present_mask;
for_each_cpu(i, mask) {
np = cpu_to_l2cache(i);
if (!np)
continue;
if (np == l2_cache) {
if (add) {
cpumask_set_cpu(cpu, cpu_core_mask(i));
cpumask_set_cpu(i, cpu_core_mask(cpu));
} else {
cpumask_clear_cpu(cpu, cpu_core_mask(i));
cpumask_clear_cpu(i, cpu_core_mask(cpu));
}
}
of_node_put(np);
}
of_node_put(l2_cache);
}
/* Activate a secondary processor. */
void start_secondary(void *unused)
{
unsigned int cpu = smp_processor_id();
int i, base;
atomic_inc(&init_mm.mm_count);
current->active_mm = &init_mm;
smp_store_cpu_info(cpu);
set_dec(tb_ticks_per_jiffy);
preempt_disable();
cpu_callin_map[cpu] = 1;
if (smp_ops->setup_cpu)
smp_ops->setup_cpu(cpu);
if (smp_ops->take_timebase)
smp_ops->take_timebase();
secondary_cpu_time_init();
#ifdef CONFIG_PPC64
if (system_state == SYSTEM_RUNNING)
vdso_data->processorCount++;
vdso_getcpu_init();
#endif
/* Update sibling maps */
powerpc: Cleanup APIs for cpu/thread/core mappings These APIs take logical cpu number as input Change cpu_first_thread_in_core() to cpu_first_thread_sibling() Change cpu_last_thread_in_core() to cpu_last_thread_sibling() These APIs convert core number (index) to logical cpu/thread numbers Add cpu_first_thread_of_core(int core) Changed cpu_thread_to_core() to cpu_core_index_of_thread(int cpu) The goal is to make 'threads_per_core' accessible to the pseries_energy module. Instead of making an API to read threads_per_core, this is a higher level wrapper function to convert from logical cpu number to core number. The current APIs cpu_first_thread_in_core() and cpu_last_thread_in_core() returns logical CPU number while cpu_thread_to_core() returns core number or index which is not a logical CPU number. The new APIs are now clearly named to distinguish 'core number' versus first and last 'logical cpu number' in that core. The new APIs cpu_{first,last}_thread_sibling() work on logical cpu numbers. While cpu_first_thread_of_core() and cpu_core_index_of_thread() work on core index. Example usage: (4 threads per core system) cpu_first_thread_sibling(5) = 4 cpu_last_thread_sibling(5) = 7 cpu_core_index_of_thread(5) = 1 cpu_first_thread_of_core(1) = 4 cpu_core_index_of_thread() is used in cpu_to_drc_index() in the module and cpu_first_thread_of_core() is used in drc_index_to_cpu() in the module. Make API changes to few callers. Export symbols for use in modules. Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-10-06 15:36:59 +07:00
base = cpu_first_thread_sibling(cpu);
for (i = 0; i < threads_per_core; i++) {
powerpc: Set cpu sibling mask before online cpu It seems following race is possible: cpu0 cpux smp_init->cpu_up->_cpu_up __cpu_up kick_cpu(1) ------------------------------------------------------------------------- waiting online ... ... notify CPU_STARTING set cpux active set cpux online ------------------------------------------------------------------------- finish waiting online ... sched_init_smp init_sched_domains(cpu_active_mask) build_sched_domains set cpux sibling info ------------------------------------------------------------------------- Execution of cpu0 and cpux could be concurrent between two separator lines. So if the cpux sibling information was set too late (normally impossible, but could be triggered by adding some delay in start_secondary, after setting cpu online), build_sched_domains() running on cpu0 might see cpux active, with an empty sibling mask, then cause some bad address accessing like following: [ 0.099855] Unable to handle kernel paging request for data at address 0xc00000038518078f [ 0.099868] Faulting instruction address: 0xc0000000000b7a64 [ 0.099883] Oops: Kernel access of bad area, sig: 11 [#1] [ 0.099895] PREEMPT SMP NR_CPUS=16 DEBUG_PAGEALLOC NUMA pSeries [ 0.099922] Modules linked in: [ 0.099940] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.10.0-rc1-00120-gb973425-dirty #16 [ 0.099956] task: c0000001fed80000 ti: c0000001fed7c000 task.ti: c0000001fed7c000 [ 0.099971] NIP: c0000000000b7a64 LR: c0000000000b7a40 CTR: c0000000000b4934 [ 0.099985] REGS: c0000001fed7f760 TRAP: 0300 Not tainted (3.10.0-rc1-00120-gb973425-dirty) [ 0.099997] MSR: 8000000000009032 <SF,EE,ME,IR,DR,RI> CR: 24272828 XER: 20000003 [ 0.100045] SOFTE: 1 [ 0.100053] CFAR: c000000000445ee8 [ 0.100064] DAR: c00000038518078f, DSISR: 40000000 [ 0.100073] GPR00: 0000000000000080 c0000001fed7f9e0 c000000000c84d48 0000000000000010 GPR04: 0000000000000010 0000000000000000 c0000001fc55e090 0000000000000000 GPR08: ffffffffffffffff c000000000b80b30 c000000000c962d8 00000003845ffc5f GPR12: 0000000000000000 c00000000f33d000 c00000000000b9e4 0000000000000000 GPR16: 0000000000000000 0000000000000000 0000000000000001 0000000000000000 GPR20: c000000000ccf750 0000000000000000 c000000000c94d48 c0000001fc504000 GPR24: c0000001fc504000 c0000001fecef848 c000000000c94d48 c000000000ccf000 GPR28: c0000001fc522090 0000000000000010 c0000001fecef848 c0000001fed7fae0 [ 0.100293] NIP [c0000000000b7a64] .get_group+0x84/0xc4 [ 0.100307] LR [c0000000000b7a40] .get_group+0x60/0xc4 [ 0.100318] Call Trace: [ 0.100332] [c0000001fed7f9e0] [c0000000000dbce4] .lock_is_held+0xa8/0xd0 (unreliable) [ 0.100354] [c0000001fed7fa70] [c0000000000bf62c] .build_sched_domains+0x728/0xd14 [ 0.100375] [c0000001fed7fbe0] [c000000000af67bc] .sched_init_smp+0x4fc/0x654 [ 0.100394] [c0000001fed7fce0] [c000000000adce24] .kernel_init_freeable+0x17c/0x30c [ 0.100413] [c0000001fed7fdb0] [c00000000000ba08] .kernel_init+0x24/0x12c [ 0.100431] [c0000001fed7fe30] [c000000000009f74] .ret_from_kernel_thread+0x5c/0x68 [ 0.100445] Instruction dump: [ 0.100456] 38800010 38a00000 4838e3f5 60000000 7c6307b4 2fbf0000 419e0040 3d220001 [ 0.100496] 78601f24 39491590 e93e0008 7d6a002a <7d69582a> f97f0000 7d4a002a e93e0010 [ 0.100559] ---[ end trace 31fd0ba7d8756001 ]--- This patch tries to move the sibling maps updating before notify_cpu_starting() and cpu online, and a write barrier there to make sure sibling maps are updated before active and online mask. Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2013-05-16 17:20:26 +07:00
if (cpu_is_offline(base + i) && (cpu != base + i))
continue;
cpumask_set_cpu(cpu, cpu_sibling_mask(base + i));
cpumask_set_cpu(base + i, cpu_sibling_mask(cpu));
/* cpu_core_map should be a superset of
* cpu_sibling_map even if we don't have cache
* information, so update the former here, too.
*/
cpumask_set_cpu(cpu, cpu_core_mask(base + i));
cpumask_set_cpu(base + i, cpu_core_mask(cpu));
}
traverse_core_siblings(cpu, true);
set_numa_node(numa_cpu_lookup_table[cpu]);
set_numa_mem(local_memory_node(numa_cpu_lookup_table[cpu]));
powerpc: Set cpu sibling mask before online cpu It seems following race is possible: cpu0 cpux smp_init->cpu_up->_cpu_up __cpu_up kick_cpu(1) ------------------------------------------------------------------------- waiting online ... ... notify CPU_STARTING set cpux active set cpux online ------------------------------------------------------------------------- finish waiting online ... sched_init_smp init_sched_domains(cpu_active_mask) build_sched_domains set cpux sibling info ------------------------------------------------------------------------- Execution of cpu0 and cpux could be concurrent between two separator lines. So if the cpux sibling information was set too late (normally impossible, but could be triggered by adding some delay in start_secondary, after setting cpu online), build_sched_domains() running on cpu0 might see cpux active, with an empty sibling mask, then cause some bad address accessing like following: [ 0.099855] Unable to handle kernel paging request for data at address 0xc00000038518078f [ 0.099868] Faulting instruction address: 0xc0000000000b7a64 [ 0.099883] Oops: Kernel access of bad area, sig: 11 [#1] [ 0.099895] PREEMPT SMP NR_CPUS=16 DEBUG_PAGEALLOC NUMA pSeries [ 0.099922] Modules linked in: [ 0.099940] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.10.0-rc1-00120-gb973425-dirty #16 [ 0.099956] task: c0000001fed80000 ti: c0000001fed7c000 task.ti: c0000001fed7c000 [ 0.099971] NIP: c0000000000b7a64 LR: c0000000000b7a40 CTR: c0000000000b4934 [ 0.099985] REGS: c0000001fed7f760 TRAP: 0300 Not tainted (3.10.0-rc1-00120-gb973425-dirty) [ 0.099997] MSR: 8000000000009032 <SF,EE,ME,IR,DR,RI> CR: 24272828 XER: 20000003 [ 0.100045] SOFTE: 1 [ 0.100053] CFAR: c000000000445ee8 [ 0.100064] DAR: c00000038518078f, DSISR: 40000000 [ 0.100073] GPR00: 0000000000000080 c0000001fed7f9e0 c000000000c84d48 0000000000000010 GPR04: 0000000000000010 0000000000000000 c0000001fc55e090 0000000000000000 GPR08: ffffffffffffffff c000000000b80b30 c000000000c962d8 00000003845ffc5f GPR12: 0000000000000000 c00000000f33d000 c00000000000b9e4 0000000000000000 GPR16: 0000000000000000 0000000000000000 0000000000000001 0000000000000000 GPR20: c000000000ccf750 0000000000000000 c000000000c94d48 c0000001fc504000 GPR24: c0000001fc504000 c0000001fecef848 c000000000c94d48 c000000000ccf000 GPR28: c0000001fc522090 0000000000000010 c0000001fecef848 c0000001fed7fae0 [ 0.100293] NIP [c0000000000b7a64] .get_group+0x84/0xc4 [ 0.100307] LR [c0000000000b7a40] .get_group+0x60/0xc4 [ 0.100318] Call Trace: [ 0.100332] [c0000001fed7f9e0] [c0000000000dbce4] .lock_is_held+0xa8/0xd0 (unreliable) [ 0.100354] [c0000001fed7fa70] [c0000000000bf62c] .build_sched_domains+0x728/0xd14 [ 0.100375] [c0000001fed7fbe0] [c000000000af67bc] .sched_init_smp+0x4fc/0x654 [ 0.100394] [c0000001fed7fce0] [c000000000adce24] .kernel_init_freeable+0x17c/0x30c [ 0.100413] [c0000001fed7fdb0] [c00000000000ba08] .kernel_init+0x24/0x12c [ 0.100431] [c0000001fed7fe30] [c000000000009f74] .ret_from_kernel_thread+0x5c/0x68 [ 0.100445] Instruction dump: [ 0.100456] 38800010 38a00000 4838e3f5 60000000 7c6307b4 2fbf0000 419e0040 3d220001 [ 0.100496] 78601f24 39491590 e93e0008 7d6a002a <7d69582a> f97f0000 7d4a002a e93e0010 [ 0.100559] ---[ end trace 31fd0ba7d8756001 ]--- This patch tries to move the sibling maps updating before notify_cpu_starting() and cpu online, and a write barrier there to make sure sibling maps are updated before active and online mask. Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2013-05-16 17:20:26 +07:00
smp_wmb();
notify_cpu_starting(cpu);
set_cpu_online(cpu, true);
local_irq_enable();
cpu_startup_entry(CPUHP_ONLINE);
BUG();
}
int setup_profiling_timer(unsigned int multiplier)
{
return 0;
}
#ifdef CONFIG_SCHED_SMT
/* cpumask of CPUs with asymetric SMT dependancy */
sched: Fix compiler warnings Commit 143e1e28cb (sched: Rework sched_domain topology definition) introduced a number of functions with a return value of 'const int'. gcc doesn't know what to do with that and, if the kernel is compiled with W=1, complains with the following warnings whenever sched.h is included. include/linux/sched.h:875:25: warning: type qualifiers ignored on function return type include/linux/sched.h:882:25: warning: type qualifiers ignored on function return type include/linux/sched.h:889:25: warning: type qualifiers ignored on function return type include/linux/sched.h:1002:21: warning: type qualifiers ignored on function return type Commits fb2aa855 (sched, ARM: Create a dedicated scheduler topology table) and 607b45e9a (sched, powerpc: Create a dedicated topology table) introduce the same warning in the arm and powerpc code. Drop 'const' from the function declarations to fix the problem. The fix for all three patches has to be applied together to avoid compilation failures for the affected architectures. Acked-by: Vincent Guittot <vincent.guittot@linaro.org> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Cc: Russell King <linux@arm.linux.org.uk> Cc: Paul Mackerras <paulus@samba.org> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com> Signed-off-by: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1403658329-13196-1-git-send-email-linux@roeck-us.net Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-06-25 08:05:29 +07:00
static int powerpc_smt_flags(void)
{
sched: Rename capacity related flags It is better not to think about compute capacity as being equivalent to "CPU power". The upcoming "power aware" scheduler work may create confusion with the notion of energy consumption if "power" is used too liberally. Let's rename the following feature flags since they do relate to capacity: SD_SHARE_CPUPOWER -> SD_SHARE_CPUCAPACITY ARCH_POWER -> ARCH_CAPACITY NONTASK_POWER -> NONTASK_CAPACITY Signed-off-by: Nicolas Pitre <nico@linaro.org> Signed-off-by: Peter Zijlstra <peterz@infradead.org> Cc: Vincent Guittot <vincent.guittot@linaro.org> Cc: Daniel Lezcano <daniel.lezcano@linaro.org> Cc: Morten Rasmussen <morten.rasmussen@arm.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linaro-kernel@lists.linaro.org Cc: Andy Fleming <afleming@freescale.com> Cc: Anton Blanchard <anton@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Grant Likely <grant.likely@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: Rob Herring <robh+dt@kernel.org> Cc: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Cc: Toshi Kani <toshi.kani@hp.com> Cc: Vasant Hegde <hegdevasant@linux.vnet.ibm.com> Cc: Vincent Guittot <vincent.guittot@linaro.org> Cc: devicetree@vger.kernel.org Cc: linux-kernel@vger.kernel.org Cc: linuxppc-dev@lists.ozlabs.org Link: http://lkml.kernel.org/n/tip-e93lpnxb87owfievqatey6b5@git.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-05-28 00:50:41 +07:00
int flags = SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
if (cpu_has_feature(CPU_FTR_ASYM_SMT)) {
printk_once(KERN_INFO "Enabling Asymmetric SMT scheduling\n");
flags |= SD_ASYM_PACKING;
}
return flags;
}
#endif
static struct sched_domain_topology_level powerpc_topology[] = {
#ifdef CONFIG_SCHED_SMT
{ cpu_smt_mask, powerpc_smt_flags, SD_INIT_NAME(SMT) },
#endif
{ cpu_cpu_mask, SD_INIT_NAME(DIE) },
{ NULL, },
};
void __init smp_cpus_done(unsigned int max_cpus)
{
cpumask_var_t old_mask;
/* We want the setup_cpu() here to be called from CPU 0, but our
* init thread may have been "borrowed" by another CPU in the meantime
* se we pin us down to CPU 0 for a short while
*/
alloc_cpumask_var(&old_mask, GFP_NOWAIT);
cpumask_copy(old_mask, tsk_cpus_allowed(current));
set_cpus_allowed_ptr(current, cpumask_of(boot_cpuid));
if (smp_ops && smp_ops->setup_cpu)
smp_ops->setup_cpu(boot_cpuid);
set_cpus_allowed_ptr(current, old_mask);
free_cpumask_var(old_mask);
if (smp_ops && smp_ops->bringup_done)
smp_ops->bringup_done();
dump_numa_cpu_topology();
set_sched_topology(powerpc_topology);
}
#ifdef CONFIG_HOTPLUG_CPU
int __cpu_disable(void)
{
int cpu = smp_processor_id();
int base, i;
int err;
if (!smp_ops->cpu_disable)
return -ENOSYS;
err = smp_ops->cpu_disable();
if (err)
return err;
/* Update sibling maps */
powerpc: Cleanup APIs for cpu/thread/core mappings These APIs take logical cpu number as input Change cpu_first_thread_in_core() to cpu_first_thread_sibling() Change cpu_last_thread_in_core() to cpu_last_thread_sibling() These APIs convert core number (index) to logical cpu/thread numbers Add cpu_first_thread_of_core(int core) Changed cpu_thread_to_core() to cpu_core_index_of_thread(int cpu) The goal is to make 'threads_per_core' accessible to the pseries_energy module. Instead of making an API to read threads_per_core, this is a higher level wrapper function to convert from logical cpu number to core number. The current APIs cpu_first_thread_in_core() and cpu_last_thread_in_core() returns logical CPU number while cpu_thread_to_core() returns core number or index which is not a logical CPU number. The new APIs are now clearly named to distinguish 'core number' versus first and last 'logical cpu number' in that core. The new APIs cpu_{first,last}_thread_sibling() work on logical cpu numbers. While cpu_first_thread_of_core() and cpu_core_index_of_thread() work on core index. Example usage: (4 threads per core system) cpu_first_thread_sibling(5) = 4 cpu_last_thread_sibling(5) = 7 cpu_core_index_of_thread(5) = 1 cpu_first_thread_of_core(1) = 4 cpu_core_index_of_thread() is used in cpu_to_drc_index() in the module and cpu_first_thread_of_core() is used in drc_index_to_cpu() in the module. Make API changes to few callers. Export symbols for use in modules. Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2010-10-06 15:36:59 +07:00
base = cpu_first_thread_sibling(cpu);
for (i = 0; i < threads_per_core; i++) {
cpumask_clear_cpu(cpu, cpu_sibling_mask(base + i));
cpumask_clear_cpu(base + i, cpu_sibling_mask(cpu));
cpumask_clear_cpu(cpu, cpu_core_mask(base + i));
cpumask_clear_cpu(base + i, cpu_core_mask(cpu));
}
traverse_core_siblings(cpu, false);
return 0;
}
void __cpu_die(unsigned int cpu)
{
if (smp_ops->cpu_die)
smp_ops->cpu_die(cpu);
}
void cpu_die(void)
{
if (ppc_md.cpu_die)
ppc_md.cpu_die();
/* If we return, we re-enter start_secondary */
start_secondary_resume();
}
#endif