diff options
Diffstat (limited to 'arch/x86/kernel/apic/ipi.c')
-rw-r--r-- | arch/x86/kernel/apic/ipi.c | 176 |
1 files changed, 77 insertions, 99 deletions
diff --git a/arch/x86/kernel/apic/ipi.c b/arch/x86/kernel/apic/ipi.c index 9bfd6e397384..a44ba7209ef3 100644 --- a/arch/x86/kernel/apic/ipi.c +++ b/arch/x86/kernel/apic/ipi.c @@ -1,7 +1,9 @@ // SPDX-License-Identifier: GPL-2.0 #include <linux/cpumask.h> +#include <linux/delay.h> #include <linux/smp.h> + #include <asm/io_apic.h> #include "local.h" @@ -52,9 +54,9 @@ void apic_send_IPI_allbutself(unsigned int vector) return; if (static_branch_likely(&apic_use_ipi_shorthand)) - apic->send_IPI_allbutself(vector); + __apic_send_IPI_allbutself(vector); else - apic->send_IPI_mask_allbutself(cpu_online_mask, vector); + __apic_send_IPI_mask_allbutself(cpu_online_mask, vector); } /* @@ -68,12 +70,12 @@ void native_smp_send_reschedule(int cpu) WARN(1, "sched: Unexpected reschedule of offline CPU#%d!\n", cpu); return; } - apic->send_IPI(cpu, RESCHEDULE_VECTOR); + __apic_send_IPI(cpu, RESCHEDULE_VECTOR); } void native_send_call_func_single_ipi(int cpu) { - apic->send_IPI(cpu, CALL_FUNCTION_SINGLE_VECTOR); + __apic_send_IPI(cpu, CALL_FUNCTION_SINGLE_VECTOR); } void native_send_call_func_ipi(const struct cpumask *mask) @@ -85,14 +87,14 @@ void native_send_call_func_ipi(const struct cpumask *mask) goto sendmask; if (cpumask_test_cpu(cpu, mask)) - apic->send_IPI_all(CALL_FUNCTION_VECTOR); + __apic_send_IPI_all(CALL_FUNCTION_VECTOR); else if (num_online_cpus() > 1) - apic->send_IPI_allbutself(CALL_FUNCTION_VECTOR); + __apic_send_IPI_allbutself(CALL_FUNCTION_VECTOR); return; } sendmask: - apic->send_IPI_mask(mask, CALL_FUNCTION_VECTOR); + __apic_send_IPI_mask(mask, CALL_FUNCTION_VECTOR); } #endif /* CONFIG_SMP */ @@ -102,74 +104,77 @@ static inline int __prepare_ICR2(unsigned int mask) return SET_XAPIC_DEST_FIELD(mask); } -static inline void __xapic_wait_icr_idle(void) +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(); } -void __default_send_IPI_shortcut(unsigned int shortcut, int vector) +/* + * 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) { /* - * 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; - - /* - * 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(); - /* - * No need to touch the target chip field. Also the destination - * mode is ignored when a shorthand is used. - */ - cfg = __prepare_ICR(shortcut, vector, 0); - - /* - * Send the IPI. The write to APIC_ICR fires this off. - */ - native_apic_mem_write(APIC_ICR, cfg); + /* Destination field (ICR2) and the destination mode are ignored */ + native_apic_mem_write(APIC_ICR, __prepare_ICR(shortcut, vector, 0)); } /* * 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 mask, int vector, unsigned int dest) +void __default_send_IPI_dest_field(unsigned int dest_mask, int vector, + unsigned int dest_mode) { - unsigned long cfg; - - /* - * Wait for idle. - */ + /* See comment in __default_send_IPI_shortcut() */ 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); - - /* - * program the ICR - */ - cfg = __prepare_ICR(0, vector, dest); - - /* - * 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) @@ -184,18 +189,13 @@ 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); } @@ -203,18 +203,15 @@ 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); } @@ -224,7 +221,7 @@ 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); } void default_send_IPI_allbutself(int vector) @@ -243,50 +240,32 @@ void default_send_IPI_self(int vector) } #ifdef CONFIG_X86_32 - -void default_send_IPI_mask_sequence_logical(const struct cpumask *mask, - int vector) +void default_send_IPI_mask_sequence_logical(const struct cpumask *mask, int vector) { unsigned long flags; - unsigned int query_cpu; - - /* - * 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 - */ + 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]; @@ -302,7 +281,6 @@ void default_send_IPI_mask_logical(const struct cpumask *cpumask, int vector) } #ifdef CONFIG_SMP -/* must come after the send_IPI functions above for inlining */ static int convert_apicid_to_cpu(int apic_id) { int i; @@ -321,7 +299,7 @@ int safe_smp_processor_id(void) if (!boot_cpu_has(X86_FEATURE_APIC)) return 0; - apicid = hard_smp_processor_id(); + apicid = read_apic_id(); if (apicid == BAD_APICID) return 0; |