diff options
Diffstat (limited to 'arch/mips/kernel/smp.c')
| -rw-r--r-- | arch/mips/kernel/smp.c | 270 |
1 files changed, 170 insertions, 100 deletions
diff --git a/arch/mips/kernel/smp.c b/arch/mips/kernel/smp.c index 770d4d1516cb..4868e79f3b30 100644 --- a/arch/mips/kernel/smp.c +++ b/arch/mips/kernel/smp.c @@ -1,17 +1,5 @@ +// SPDX-License-Identifier: GPL-2.0-or-later /* - * 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. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. * * Copyright (C) 2000, 2001 Kanoj Sarcar * Copyright (C) 2000, 2001 Ralf Baechle @@ -22,6 +10,7 @@ #include <linux/delay.h> #include <linux/init.h> #include <linux/interrupt.h> +#include <linux/profile.h> #include <linux/smp.h> #include <linux/spinlock.h> #include <linux/threads.h> @@ -39,16 +28,17 @@ #include <linux/atomic.h> #include <asm/cpu.h> +#include <asm/ginvt.h> #include <asm/processor.h> #include <asm/idle.h> #include <asm/r4k-timer.h> -#include <asm/mips-cpc.h> +#include <asm/mips-cps.h> #include <asm/mmu_context.h> #include <asm/time.h> #include <asm/setup.h> #include <asm/maar.h> -int __cpu_number_map[NR_CPUS]; /* Map physical to logical */ +int __cpu_number_map[CONFIG_MIPS_NR_CPU_NR_MAP]; /* Map physical to logical */ EXPORT_SYMBOL(__cpu_number_map); int __cpu_logical_map[NR_CPUS]; /* Map logical to physical */ @@ -66,10 +56,13 @@ EXPORT_SYMBOL(cpu_sibling_map); cpumask_t cpu_core_map[NR_CPUS] __read_mostly; EXPORT_SYMBOL(cpu_core_map); +#ifndef CONFIG_HOTPLUG_PARALLEL +static DECLARE_COMPLETION(cpu_starting); static DECLARE_COMPLETION(cpu_running); +#endif /* - * A logcal cpu mask containing only one VPE per core to + * A logical cpu mask containing only one VPE per core to * reduce the number of IPIs on large MT systems. */ cpumask_t cpu_foreign_map[NR_CPUS] __read_mostly; @@ -83,6 +76,26 @@ static cpumask_t cpu_core_setup_map; cpumask_t cpu_coherent_mask; +struct cpumask __cpu_primary_thread_mask __read_mostly; + +unsigned int smp_max_threads __initdata = UINT_MAX; + +static int __init early_nosmt(char *s) +{ + smp_max_threads = 1; + return 0; +} +early_param("nosmt", early_nosmt); + +static int __init early_smt(char *s) +{ + get_option(&s, &smp_max_threads); + /* Ensure at least one thread is available */ + smp_max_threads = clamp_val(smp_max_threads, 1U, UINT_MAX); + return 0; +} +early_param("smt", early_smt); + #ifdef CONFIG_GENERIC_IRQ_IPI static struct irq_desc *call_desc; static struct irq_desc *sched_desc; @@ -96,8 +109,7 @@ static inline void set_cpu_sibling_map(int cpu) if (smp_num_siblings > 1) { for_each_cpu(i, &cpu_sibling_setup_map) { - if (cpu_data[cpu].package == cpu_data[i].package && - cpu_data[cpu].core == cpu_data[i].core) { + if (cpus_are_siblings(cpu, i)) { cpumask_set_cpu(i, &cpu_sibling_map[cpu]); cpumask_set_cpu(cpu, &cpu_sibling_map[i]); } @@ -134,8 +146,7 @@ void calculate_cpu_foreign_map(void) for_each_online_cpu(i) { core_present = 0; for_each_cpu(k, &temp_foreign_map) - if (cpu_data[i].package == cpu_data[k].package && - cpu_data[i].core == cpu_data[k].core) + if (cpus_are_siblings(i, k)) core_present = 1; if (!core_present) cpumask_set_cpu(i, &temp_foreign_map); @@ -146,10 +157,10 @@ void calculate_cpu_foreign_map(void) &temp_foreign_map, &cpu_sibling_map[i]); } -struct plat_smp_ops *mp_ops; +const struct plat_smp_ops *mp_ops; EXPORT_SYMBOL(mp_ops); -void register_smp_ops(struct plat_smp_ops *ops) +void register_smp_ops(const struct plat_smp_ops *ops) { if (mp_ops) printk(KERN_WARNING "Overriding previously set SMP ops\n"); @@ -186,13 +197,13 @@ void mips_smp_send_ipi_mask(const struct cpumask *mask, unsigned int action) if (mips_cpc_present()) { for_each_cpu(cpu, mask) { - core = cpu_data[cpu].core; - - if (core == current_cpu_data.core) + if (cpus_are_siblings(cpu, smp_processor_id())) continue; + core = cpu_core(&cpu_data[cpu]); + while (!cpumask_test_cpu(cpu, &cpu_coherent_mask)) { - mips_cm_lock_other(core, 0); + mips_cm_lock_other_cpu(cpu, CM_GCR_Cx_OTHER_BLOCK_LOCAL); mips_cpc_lock_other(core); write_cpc_co_cmd(CPC_Cx_CMD_PWRUP); mips_cpc_unlock_other(); @@ -219,25 +230,13 @@ static irqreturn_t ipi_call_interrupt(int irq, void *dev_id) return IRQ_HANDLED; } -static struct irqaction irq_resched = { - .handler = ipi_resched_interrupt, - .flags = IRQF_PERCPU, - .name = "IPI resched" -}; - -static struct irqaction irq_call = { - .handler = ipi_call_interrupt, - .flags = IRQF_PERCPU, - .name = "IPI call" -}; - -static void smp_ipi_init_one(unsigned int virq, - struct irqaction *action) +static void smp_ipi_init_one(unsigned int virq, const char *name, + irq_handler_t handler) { int ret; irq_set_handler(virq, handle_percpu_irq); - ret = setup_irq(virq, action); + ret = request_irq(virq, handler, IRQF_PERCPU, name, NULL); BUG_ON(ret); } @@ -290,12 +289,15 @@ int mips_smp_ipi_allocate(const struct cpumask *mask) int cpu; for_each_cpu(cpu, mask) { - smp_ipi_init_one(call_virq + cpu, &irq_call); - smp_ipi_init_one(sched_virq + cpu, &irq_resched); + smp_ipi_init_one(call_virq + cpu, "IPI call", + ipi_call_interrupt); + smp_ipi_init_one(sched_virq + cpu, "IPI resched", + ipi_resched_interrupt); } } else { - smp_ipi_init_one(call_virq, &irq_call); - smp_ipi_init_one(sched_virq, &irq_resched); + smp_ipi_init_one(call_virq, "IPI call", ipi_call_interrupt); + smp_ipi_init_one(sched_virq, "IPI resched", + ipi_resched_interrupt); } return 0; @@ -323,8 +325,8 @@ int mips_smp_ipi_free(const struct cpumask *mask) int cpu; for_each_cpu(cpu, mask) { - remove_irq(call_virq + cpu, &irq_call); - remove_irq(sched_virq + cpu, &irq_resched); + free_irq(call_virq + cpu, NULL); + free_irq(sched_virq + cpu, NULL); } } irq_destroy_ipi(call_virq, mask); @@ -354,10 +356,11 @@ early_initcall(mips_smp_ipi_init); */ asmlinkage void start_secondary(void) { - unsigned int cpu; + unsigned int cpu = raw_smp_processor_id(); cpu_probe(); per_cpu_trap_init(false); + rcutree_report_cpu_starting(cpu); mips_clockevent_init(); mp_ops->init_secondary(); cpu_report(); @@ -368,24 +371,38 @@ asmlinkage void start_secondary(void) * to an option instead of something based on .cputype */ +#ifdef CONFIG_HOTPLUG_PARALLEL + cpuhp_ap_sync_alive(); +#endif calibrate_delay(); - preempt_disable(); - cpu = smp_processor_id(); cpu_data[cpu].udelay_val = loops_per_jiffy; + set_cpu_sibling_map(cpu); + set_cpu_core_map(cpu); + cpumask_set_cpu(cpu, &cpu_coherent_mask); notify_cpu_starting(cpu); - complete(&cpu_running); +#ifndef CONFIG_HOTPLUG_PARALLEL + /* Notify boot CPU that we're starting & ready to sync counters */ + complete(&cpu_starting); +#endif + synchronise_count_slave(cpu); + /* The CPU is running and counters synchronised, now mark it online */ set_cpu_online(cpu, true); - set_cpu_sibling_map(cpu); - set_cpu_core_map(cpu); - calculate_cpu_foreign_map(); +#ifndef CONFIG_HOTPLUG_PARALLEL + /* + * Notify boot CPU that we're up & online and it can safely return + * from __cpu_up + */ + complete(&cpu_running); +#endif + /* * irq will be enabled in ->smp_finish(), enabling it too early * is dangerous. @@ -433,35 +450,48 @@ void __init smp_prepare_cpus(unsigned int max_cpus) } /* preload SMP state for boot cpu */ -void smp_prepare_boot_cpu(void) +void __init smp_prepare_boot_cpu(void) { + if (mp_ops->prepare_boot_cpu) + mp_ops->prepare_boot_cpu(); set_cpu_possible(0, true); set_cpu_online(0, true); } +#ifdef CONFIG_HOTPLUG_PARALLEL +int arch_cpuhp_kick_ap_alive(unsigned int cpu, struct task_struct *tidle) +{ + return mp_ops->boot_secondary(cpu, tidle); +} +#else int __cpu_up(unsigned int cpu, struct task_struct *tidle) { - mp_ops->boot_secondary(cpu, tidle); + int err; - /* - * We must check for timeout here, as the CPU will not be marked - * online until the counters are synchronised. - */ - if (!wait_for_completion_timeout(&cpu_running, + err = mp_ops->boot_secondary(cpu, tidle); + if (err) + return err; + + /* Wait for CPU to start and be ready to sync counters */ + if (!wait_for_completion_timeout(&cpu_starting, msecs_to_jiffies(1000))) { pr_crit("CPU%u: failed to start\n", cpu); return -EIO; } - synchronise_count_master(cpu); + /* Wait for CPU to finish startup & mark itself online before return */ + wait_for_completion(&cpu_running); return 0; } +#endif +#ifdef CONFIG_PROFILING /* Not really SMP stuff ... */ int setup_profiling_timer(unsigned int multiplier) { return 0; } +#endif static void flush_tlb_all_ipi(void *info) { @@ -470,12 +500,21 @@ static void flush_tlb_all_ipi(void *info) void flush_tlb_all(void) { + if (cpu_has_mmid) { + htw_stop(); + ginvt_full(); + sync_ginv(); + instruction_hazard(); + htw_start(); + return; + } + on_each_cpu(flush_tlb_all_ipi, NULL, 1); } static void flush_tlb_mm_ipi(void *mm) { - local_flush_tlb_mm((struct mm_struct *)mm); + drop_mmu_context((struct mm_struct *)mm); } /* @@ -507,8 +546,8 @@ static inline void smp_on_each_tlb(void (*func) (void *info), void *info) * address spaces, a new context is obtained on the current cpu, and tlb * context on other cpus are invalidated to force a new context allocation * at switch_mm time, should the mm ever be used on other cpus. For - * multithreaded address spaces, intercpu interrupts have to be sent. - * Another case where intercpu interrupts are required is when the target + * multithreaded address spaces, inter-CPU interrupts have to be sent. + * Another case where inter-CPU interrupts are required is when the target * mm might be active on another cpu (eg debuggers doing the flushes on * behalf of debugees, kswapd stealing pages from another process etc). * Kanoj 07/00. @@ -516,19 +555,30 @@ static inline void smp_on_each_tlb(void (*func) (void *info), void *info) void flush_tlb_mm(struct mm_struct *mm) { + if (!mm) + return; + + if (atomic_read(&mm->mm_users) == 0) + return; /* happens as a result of exit_mmap() */ + preempt_disable(); - if ((atomic_read(&mm->mm_users) != 1) || (current->mm != mm)) { + if (cpu_has_mmid) { + /* + * No need to worry about other CPUs - the ginvt in + * drop_mmu_context() will be globalized. + */ + } else if ((atomic_read(&mm->mm_users) != 1) || (current->mm != mm)) { smp_on_other_tlbs(flush_tlb_mm_ipi, mm); } else { unsigned int cpu; for_each_online_cpu(cpu) { if (cpu != smp_processor_id() && cpu_context(cpu, mm)) - cpu_context(cpu, mm) = 0; + set_cpu_context(cpu, mm, 0); } } - local_flush_tlb_mm(mm); + drop_mmu_context(mm); preempt_enable(); } @@ -549,9 +599,26 @@ static void flush_tlb_range_ipi(void *info) void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned long end) { struct mm_struct *mm = vma->vm_mm; + unsigned long addr; + u32 old_mmid; preempt_disable(); - if ((atomic_read(&mm->mm_users) != 1) || (current->mm != mm)) { + if (cpu_has_mmid) { + htw_stop(); + old_mmid = read_c0_memorymapid(); + write_c0_memorymapid(cpu_asid(0, mm)); + mtc0_tlbw_hazard(); + addr = round_down(start, PAGE_SIZE * 2); + end = round_up(end, PAGE_SIZE * 2); + do { + ginvt_va_mmid(addr); + sync_ginv(); + addr += PAGE_SIZE * 2; + } while (addr < end); + write_c0_memorymapid(old_mmid); + instruction_hazard(); + htw_start(); + } else if ((atomic_read(&mm->mm_users) != 1) || (current->mm != mm)) { struct flush_tlb_data fd = { .vma = vma, .addr1 = start, @@ -559,6 +626,7 @@ void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned l }; smp_on_other_tlbs(flush_tlb_range_ipi, &fd); + local_flush_tlb_range(vma, start, end); } else { unsigned int cpu; int exec = vma->vm_flags & VM_EXEC; @@ -571,10 +639,10 @@ void flush_tlb_range(struct vm_area_struct *vma, unsigned long start, unsigned l * mm has been completely unused by that CPU. */ if (cpu != smp_processor_id() && cpu_context(cpu, mm)) - cpu_context(cpu, mm) = !exec; + set_cpu_context(cpu, mm, !exec); } + local_flush_tlb_range(vma, start, end); } - local_flush_tlb_range(vma, start, end); preempt_enable(); } @@ -604,14 +672,28 @@ static void flush_tlb_page_ipi(void *info) void flush_tlb_page(struct vm_area_struct *vma, unsigned long page) { + u32 old_mmid; + preempt_disable(); - if ((atomic_read(&vma->vm_mm->mm_users) != 1) || (current->mm != vma->vm_mm)) { + if (cpu_has_mmid) { + htw_stop(); + old_mmid = read_c0_memorymapid(); + write_c0_memorymapid(cpu_asid(0, vma->vm_mm)); + mtc0_tlbw_hazard(); + ginvt_va_mmid(page); + sync_ginv(); + write_c0_memorymapid(old_mmid); + instruction_hazard(); + htw_start(); + } else if ((atomic_read(&vma->vm_mm->mm_users) != 1) || + (current->mm != vma->vm_mm)) { struct flush_tlb_data fd = { .vma = vma, .addr1 = page, }; smp_on_other_tlbs(flush_tlb_page_ipi, &fd); + local_flush_tlb_page(vma, page); } else { unsigned int cpu; @@ -623,10 +705,10 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long page) * by that CPU. */ if (cpu != smp_processor_id() && cpu_context(cpu, vma->vm_mm)) - cpu_context(cpu, vma->vm_mm) = 1; + set_cpu_context(cpu, vma->vm_mm, 1); } + local_flush_tlb_page(vma, page); } - local_flush_tlb_page(vma, page); preempt_enable(); } @@ -645,45 +727,33 @@ void flush_tlb_one(unsigned long vaddr) EXPORT_SYMBOL(flush_tlb_page); EXPORT_SYMBOL(flush_tlb_one); -#ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST - -static DEFINE_PER_CPU(atomic_t, tick_broadcast_count); -static DEFINE_PER_CPU(struct call_single_data, tick_broadcast_csd); - -void tick_broadcast(const struct cpumask *mask) +#ifdef CONFIG_HOTPLUG_CORE_SYNC_DEAD +void arch_cpuhp_cleanup_dead_cpu(unsigned int cpu) { - atomic_t *count; - struct call_single_data *csd; - int cpu; - - for_each_cpu(cpu, mask) { - count = &per_cpu(tick_broadcast_count, cpu); - csd = &per_cpu(tick_broadcast_csd, cpu); - - if (atomic_inc_return(count) == 1) - smp_call_function_single_async(cpu, csd); - } + if (mp_ops->cleanup_dead_cpu) + mp_ops->cleanup_dead_cpu(cpu); } +#endif + +#ifdef CONFIG_GENERIC_CLOCKEVENTS_BROADCAST static void tick_broadcast_callee(void *info) { - int cpu = smp_processor_id(); tick_receive_broadcast(); - atomic_set(&per_cpu(tick_broadcast_count, cpu), 0); } -static int __init tick_broadcast_init(void) +static DEFINE_PER_CPU(call_single_data_t, tick_broadcast_csd) = + CSD_INIT(tick_broadcast_callee, NULL); + +void tick_broadcast(const struct cpumask *mask) { - struct call_single_data *csd; + call_single_data_t *csd; int cpu; - for (cpu = 0; cpu < NR_CPUS; cpu++) { + for_each_cpu(cpu, mask) { csd = &per_cpu(tick_broadcast_csd, cpu); - csd->func = tick_broadcast_callee; + smp_call_function_single_async(cpu, csd); } - - return 0; } -early_initcall(tick_broadcast_init); #endif /* CONFIG_GENERIC_CLOCKEVENTS_BROADCAST */ |
