diff options
Diffstat (limited to 'arch/x86/kernel/apic/ipi.c')
| -rw-r--r-- | arch/x86/kernel/apic/ipi.c | 347 |
1 files changed, 197 insertions, 150 deletions
diff --git a/arch/x86/kernel/apic/ipi.c b/arch/x86/kernel/apic/ipi.c index 82f9244fe61f..98a57cb4aa86 100644 --- a/arch/x86/kernel/apic/ipi.c +++ b/arch/x86/kernel/apic/ipi.c @@ -1,81 +1,189 @@ // SPDX-License-Identifier: GPL-2.0 -#include <linux/cpumask.h> -#include <linux/interrupt.h> -#include <linux/mm.h> +#include <linux/cpumask.h> #include <linux/delay.h> -#include <linux/spinlock.h> -#include <linux/kernel_stat.h> -#include <linux/mc146818rtc.h> -#include <linux/cache.h> -#include <linux/cpu.h> - -#include <asm/smp.h> -#include <asm/mtrr.h> -#include <asm/tlbflush.h> -#include <asm/mmu_context.h> -#include <asm/apic.h> -#include <asm/proto.h> -#include <asm/ipi.h> - -void __default_send_IPI_shortcut(unsigned int shortcut, int vector, unsigned int dest) +#include <linux/smp.h> +#include <linux/string_choices.h> + +#include <asm/io_apic.h> + +#include "local.h" + +DEFINE_STATIC_KEY_FALSE(apic_use_ipi_shorthand); + +#ifdef CONFIG_SMP +static int apic_ipi_shorthand_off __ro_after_init; + +static __init int apic_ipi_shorthand(char *str) { - /* - * Subtle. In the case of the 'never do double writes' workaround - * we have to lock out interrupts to be safe. As we don't care - * of the value read we use an atomic rmw access to avoid costly - * cli/sti. Otherwise we use an even cheaper single atomic write - * to the APIC. - */ - unsigned int cfg; + get_option(&str, &apic_ipi_shorthand_off); + return 1; +} +__setup("no_ipi_broadcast=", apic_ipi_shorthand); - /* - * Wait for idle. - */ - __xapic_wait_icr_idle(); +static int __init print_ipi_mode(void) +{ + pr_info("IPI shorthand broadcast: %s\n", + str_disabled_enabled(apic_ipi_shorthand_off)); + return 0; +} +late_initcall(print_ipi_mode); +void apic_smt_update(void) +{ /* - * No need to touch the target chip field + * Do not switch to broadcast mode if: + * - Disabled on the command line + * - Only a single CPU is online + * - Not all present CPUs have been at least booted once + * + * The latter is important as the local APIC might be in some + * random state and a broadcast might cause havoc. That's + * especially true for NMI broadcasting. */ - cfg = __prepare_ICR(shortcut, vector, dest); + if (apic_ipi_shorthand_off || num_online_cpus() == 1 || + !cpumask_equal(cpu_present_mask, &cpus_booted_once_mask)) { + static_branch_disable(&apic_use_ipi_shorthand); + } else { + static_branch_enable(&apic_use_ipi_shorthand); + } +} - /* - * Send the IPI. The write to APIC_ICR fires this off. - */ - native_apic_mem_write(APIC_ICR, cfg); +void apic_send_IPI_allbutself(unsigned int vector) +{ + if (num_online_cpus() < 2) + return; + + if (static_branch_likely(&apic_use_ipi_shorthand)) + __apic_send_IPI_allbutself(vector); + else + __apic_send_IPI_mask_allbutself(cpu_online_mask, vector); } /* - * This is used to send an IPI with no shorthand notation (the destination is - * specified in bits 56 to 63 of the ICR). + * Send a 'reschedule' IPI to another CPU. It goes straight through and + * wastes no time serializing anything. Worst case is that we lose a + * reschedule ... */ -void __default_send_IPI_dest_field(unsigned int mask, int vector, unsigned int dest) +void native_smp_send_reschedule(int cpu) +{ + if (unlikely(cpu_is_offline(cpu))) { + WARN(1, "sched: Unexpected reschedule of offline CPU#%d!\n", cpu); + return; + } + __apic_send_IPI(cpu, RESCHEDULE_VECTOR); +} + +void native_send_call_func_single_ipi(int cpu) +{ + __apic_send_IPI(cpu, CALL_FUNCTION_SINGLE_VECTOR); +} + +void native_send_call_func_ipi(const struct cpumask *mask) +{ + if (static_branch_likely(&apic_use_ipi_shorthand)) { + unsigned int cpu = smp_processor_id(); + + if (!cpumask_or_equal(mask, cpumask_of(cpu), cpu_online_mask)) + goto sendmask; + + if (cpumask_test_cpu(cpu, mask)) + __apic_send_IPI_all(CALL_FUNCTION_VECTOR); + else if (num_online_cpus() > 1) + __apic_send_IPI_allbutself(CALL_FUNCTION_VECTOR); + return; + } + +sendmask: + __apic_send_IPI_mask(mask, CALL_FUNCTION_VECTOR); +} + +void apic_send_nmi_to_offline_cpu(unsigned int cpu) +{ + if (WARN_ON_ONCE(!apic->nmi_to_offline_cpu)) + return; + if (WARN_ON_ONCE(!cpumask_test_cpu(cpu, &cpus_booted_once_mask))) + return; + apic->send_IPI(cpu, NMI_VECTOR); +} +#endif /* CONFIG_SMP */ + +static inline int __prepare_ICR2(unsigned int mask) { - unsigned long cfg; + return SET_XAPIC_DEST_FIELD(mask); +} +u32 apic_mem_wait_icr_idle_timeout(void) +{ + int cnt; + + for (cnt = 0; cnt < 1000; cnt++) { + if (!(apic_read(APIC_ICR) & APIC_ICR_BUSY)) + return 0; + inc_irq_stat(icr_read_retry_count); + udelay(100); + } + return APIC_ICR_BUSY; +} + +void apic_mem_wait_icr_idle(void) +{ + while (native_apic_mem_read(APIC_ICR) & APIC_ICR_BUSY) + cpu_relax(); +} + +/* + * This is safe against interruption because it only writes the lower 32 + * bits of the APIC_ICR register. The destination field is ignored for + * short hand IPIs. + * + * wait_icr_idle() + * write(ICR2, dest) + * NMI + * wait_icr_idle() + * write(ICR) + * wait_icr_idle() + * write(ICR) + * + * This function does not need to disable interrupts as there is no ICR2 + * interaction. The memory write is direct except when the machine is + * affected by the 11AP Pentium erratum, which turns the plain write into + * an XCHG operation. + */ +static void __default_send_IPI_shortcut(unsigned int shortcut, int vector) +{ /* - * Wait for idle. + * Wait for the previous ICR command to complete. Use + * safe_apic_wait_icr_idle() for the NMI vector as there have been + * issues where otherwise the system hangs when the panic CPU tries + * to stop the others before launching the kdump kernel. */ if (unlikely(vector == NMI_VECTOR)) - safe_apic_wait_icr_idle(); + apic_mem_wait_icr_idle_timeout(); else - __xapic_wait_icr_idle(); + apic_mem_wait_icr_idle(); - /* - * prepare target chip field - */ - cfg = __prepare_ICR2(mask); - native_apic_mem_write(APIC_ICR2, cfg); + /* Destination field (ICR2) and the destination mode are ignored */ + native_apic_mem_write(APIC_ICR, __prepare_ICR(shortcut, vector, 0)); +} - /* - * program the ICR - */ - cfg = __prepare_ICR(0, vector, dest); +/* + * This is used to send an IPI with no shorthand notation (the destination is + * specified in bits 56 to 63 of the ICR). + */ +void __default_send_IPI_dest_field(unsigned int dest_mask, int vector, + unsigned int dest_mode) +{ + /* See comment in __default_send_IPI_shortcut() */ + if (unlikely(vector == NMI_VECTOR)) + apic_mem_wait_icr_idle_timeout(); + else + apic_mem_wait_icr_idle(); - /* - * Send the IPI. The write to APIC_ICR fires this off. - */ - native_apic_mem_write(APIC_ICR, cfg); + /* Set the IPI destination field in the ICR */ + native_apic_mem_write(APIC_ICR2, __prepare_ICR2(dest_mask)); + /* Send it with the proper destination mode */ + native_apic_mem_write(APIC_ICR, __prepare_ICR(0, vector, dest_mode)); } void default_send_IPI_single_phys(int cpu, int vector) @@ -90,18 +198,13 @@ void default_send_IPI_single_phys(int cpu, int vector) void default_send_IPI_mask_sequence_phys(const struct cpumask *mask, int vector) { - unsigned long query_cpu; unsigned long flags; + unsigned long cpu; - /* - * Hack. The clustered APIC addressing mode doesn't allow us to send - * to an arbitrary mask, so I do a unicast to each CPU instead. - * - mbligh - */ local_irq_save(flags); - for_each_cpu(query_cpu, mask) { + for_each_cpu(cpu, mask) { __default_send_IPI_dest_field(per_cpu(x86_cpu_to_apicid, - query_cpu), vector, APIC_DEST_PHYSICAL); + cpu), vector, APIC_DEST_PHYSICAL); } local_irq_restore(flags); } @@ -109,18 +212,15 @@ void default_send_IPI_mask_sequence_phys(const struct cpumask *mask, int vector) void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask, int vector) { - unsigned int this_cpu = smp_processor_id(); - unsigned int query_cpu; + unsigned int cpu, this_cpu = smp_processor_id(); unsigned long flags; - /* See Hack comment above */ - local_irq_save(flags); - for_each_cpu(query_cpu, mask) { - if (query_cpu == this_cpu) + for_each_cpu(cpu, mask) { + if (cpu == this_cpu) continue; __default_send_IPI_dest_field(per_cpu(x86_cpu_to_apicid, - query_cpu), vector, APIC_DEST_PHYSICAL); + cpu), vector, APIC_DEST_PHYSICAL); } local_irq_restore(flags); } @@ -130,54 +230,51 @@ void default_send_IPI_mask_allbutself_phys(const struct cpumask *mask, */ void default_send_IPI_single(int cpu, int vector) { - apic->send_IPI_mask(cpumask_of(cpu), vector); + __apic_send_IPI_mask(cpumask_of(cpu), vector); } -#ifdef CONFIG_X86_32 +void default_send_IPI_allbutself(int vector) +{ + __default_send_IPI_shortcut(APIC_DEST_ALLBUT, vector); +} -void default_send_IPI_mask_sequence_logical(const struct cpumask *mask, - int vector) +void default_send_IPI_all(int vector) { - unsigned long flags; - unsigned int query_cpu; + __default_send_IPI_shortcut(APIC_DEST_ALLINC, vector); +} - /* - * Hack. The clustered APIC addressing mode doesn't allow us to send - * to an arbitrary mask, so I do a unicasts to each CPU instead. This - * should be modified to do 1 message per cluster ID - mbligh - */ +void default_send_IPI_self(int vector) +{ + __default_send_IPI_shortcut(APIC_DEST_SELF, vector); +} + +#ifdef CONFIG_X86_32 +void default_send_IPI_mask_sequence_logical(const struct cpumask *mask, int vector) +{ + unsigned long flags; + unsigned int cpu; local_irq_save(flags); - for_each_cpu(query_cpu, mask) - __default_send_IPI_dest_field( - early_per_cpu(x86_cpu_to_logical_apicid, query_cpu), - vector, apic->dest_logical); + for_each_cpu(cpu, mask) + __default_send_IPI_dest_field(1U << cpu, vector, APIC_DEST_LOGICAL); local_irq_restore(flags); } void default_send_IPI_mask_allbutself_logical(const struct cpumask *mask, int vector) { + unsigned int cpu, this_cpu = smp_processor_id(); unsigned long flags; - unsigned int query_cpu; - unsigned int this_cpu = smp_processor_id(); - - /* See Hack comment above */ local_irq_save(flags); - for_each_cpu(query_cpu, mask) { - if (query_cpu == this_cpu) + for_each_cpu(cpu, mask) { + if (cpu == this_cpu) continue; - __default_send_IPI_dest_field( - early_per_cpu(x86_cpu_to_logical_apicid, query_cpu), - vector, apic->dest_logical); - } + __default_send_IPI_dest_field(1U << cpu, vector, APIC_DEST_LOGICAL); + } local_irq_restore(flags); } -/* - * This is only used on smaller machines. - */ void default_send_IPI_mask_logical(const struct cpumask *cpumask, int vector) { unsigned long mask = cpumask_bits(cpumask)[0]; @@ -188,57 +285,7 @@ void default_send_IPI_mask_logical(const struct cpumask *cpumask, int vector) local_irq_save(flags); WARN_ON(mask & ~cpumask_bits(cpu_online_mask)[0]); - __default_send_IPI_dest_field(mask, vector, apic->dest_logical); + __default_send_IPI_dest_field(mask, vector, APIC_DEST_LOGICAL); local_irq_restore(flags); } - -void default_send_IPI_allbutself(int vector) -{ - /* - * if there are no other CPUs in the system then we get an APIC send - * error if we try to broadcast, thus avoid sending IPIs in this case. - */ - if (!(num_online_cpus() > 1)) - return; - - __default_local_send_IPI_allbutself(vector); -} - -void default_send_IPI_all(int vector) -{ - __default_local_send_IPI_all(vector); -} - -void default_send_IPI_self(int vector) -{ - __default_send_IPI_shortcut(APIC_DEST_SELF, vector, apic->dest_logical); -} - -/* must come after the send_IPI functions above for inlining */ -static int convert_apicid_to_cpu(int apic_id) -{ - int i; - - for_each_possible_cpu(i) { - if (per_cpu(x86_cpu_to_apicid, i) == apic_id) - return i; - } - return -1; -} - -int safe_smp_processor_id(void) -{ - int apicid, cpuid; - - if (!boot_cpu_has(X86_FEATURE_APIC)) - return 0; - - apicid = hard_smp_processor_id(); - if (apicid == BAD_APICID) - return 0; - - cpuid = convert_apicid_to_cpu(apicid); - - return cpuid >= 0 ? cpuid : 0; -} #endif |