diff options
Diffstat (limited to 'arch/x86/kernel/nmi.c')
| -rw-r--r-- | arch/x86/kernel/nmi.c | 424 |
1 files changed, 311 insertions, 113 deletions
diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c index 18bc9b51ac9b..3d239ed12744 100644 --- a/arch/x86/kernel/nmi.c +++ b/arch/x86/kernel/nmi.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: GPL-2.0-only /* * Copyright (C) 1991, 1992 Linus Torvalds * Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs @@ -21,54 +22,62 @@ #include <linux/ratelimit.h> #include <linux/slab.h> #include <linux/export.h> +#include <linux/atomic.h> #include <linux/sched/clock.h> +#include <linux/kvm_types.h> -#if defined(CONFIG_EDAC) -#include <linux/edac.h> -#endif - -#include <linux/atomic.h> +#include <asm/cpu_entry_area.h> #include <asm/traps.h> #include <asm/mach_traps.h> #include <asm/nmi.h> #include <asm/x86_init.h> #include <asm/reboot.h> #include <asm/cache.h> +#include <asm/nospec-branch.h> +#include <asm/microcode.h> +#include <asm/sev.h> +#include <asm/fred.h> #define CREATE_TRACE_POINTS #include <trace/events/nmi.h> +/* + * An emergency handler can be set in any context including NMI + */ struct nmi_desc { raw_spinlock_t lock; + nmi_handler_t emerg_handler; struct list_head head; }; -static struct nmi_desc nmi_desc[NMI_MAX] = -{ - { - .lock = __RAW_SPIN_LOCK_UNLOCKED(&nmi_desc[0].lock), - .head = LIST_HEAD_INIT(nmi_desc[0].head), - }, - { - .lock = __RAW_SPIN_LOCK_UNLOCKED(&nmi_desc[1].lock), - .head = LIST_HEAD_INIT(nmi_desc[1].head), - }, - { - .lock = __RAW_SPIN_LOCK_UNLOCKED(&nmi_desc[2].lock), - .head = LIST_HEAD_INIT(nmi_desc[2].head), - }, - { - .lock = __RAW_SPIN_LOCK_UNLOCKED(&nmi_desc[3].lock), - .head = LIST_HEAD_INIT(nmi_desc[3].head), - }, +#define NMI_DESC_INIT(type) { \ + .lock = __RAW_SPIN_LOCK_UNLOCKED(&nmi_desc[type].lock), \ + .head = LIST_HEAD_INIT(nmi_desc[type].head), \ +} +static struct nmi_desc nmi_desc[NMI_MAX] = { + NMI_DESC_INIT(NMI_LOCAL), + NMI_DESC_INIT(NMI_UNKNOWN), + NMI_DESC_INIT(NMI_SERR), + NMI_DESC_INIT(NMI_IO_CHECK), }; +#define nmi_to_desc(type) (&nmi_desc[type]) + struct nmi_stats { unsigned int normal; unsigned int unknown; unsigned int external; unsigned int swallow; + unsigned long recv_jiffies; + unsigned long idt_seq; + unsigned long idt_nmi_seq; + unsigned long idt_ignored; + atomic_long_t idt_calls; + unsigned long idt_seq_snap; + unsigned long idt_nmi_seq_snap; + unsigned long idt_ignored_snap; + long idt_calls_snap; }; static DEFINE_PER_CPU(struct nmi_stats, nmi_stats); @@ -76,6 +85,9 @@ static DEFINE_PER_CPU(struct nmi_stats, nmi_stats); static int ignore_nmis __read_mostly; int unknown_nmi_panic; +int panic_on_unrecovered_nmi; +int panic_on_io_nmi; + /* * Prevent NMI reason port (0x61) being accessed simultaneously, can * only be used in NMI handler. @@ -89,8 +101,6 @@ static int __init setup_unknown_nmi_panic(char *str) } __setup("unknown_nmi_panic", setup_unknown_nmi_panic); -#define nmi_to_desc(type) (&nmi_desc[type]) - static u64 nmi_longest_ns = 1 * NSEC_PER_MSEC; static int __init nmi_warning_debugfs(void) @@ -101,26 +111,42 @@ static int __init nmi_warning_debugfs(void) } fs_initcall(nmi_warning_debugfs); -static void nmi_max_handler(struct irq_work *w) +static void nmi_check_duration(struct nmiaction *action, u64 duration) { - struct nmiaction *a = container_of(w, struct nmiaction, irq_work); int remainder_ns, decimal_msecs; - u64 whole_msecs = READ_ONCE(a->max_duration); - remainder_ns = do_div(whole_msecs, (1000 * 1000)); - decimal_msecs = remainder_ns / 1000; + if (duration < nmi_longest_ns || duration < action->max_duration) + return; + + action->max_duration = duration; - printk_ratelimited(KERN_INFO - "INFO: NMI handler (%ps) took too long to run: %lld.%03d msecs\n", - a->handler, whole_msecs, decimal_msecs); + /* Convert duration from nsec to msec */ + remainder_ns = do_div(duration, NSEC_PER_MSEC); + decimal_msecs = remainder_ns / NSEC_PER_USEC; + + pr_info_ratelimited("INFO: NMI handler (%ps) took too long to run: %lld.%03d msecs\n", + action->handler, duration, decimal_msecs); } static int nmi_handle(unsigned int type, struct pt_regs *regs) { struct nmi_desc *desc = nmi_to_desc(type); + nmi_handler_t ehandler; struct nmiaction *a; int handled=0; + /* + * Call the emergency handler, if set + * + * In the case of crash_nmi_callback() emergency handler, it will + * return in the case of the crashing CPU to enable it to complete + * other necessary crashing actions ASAP. Other handlers in the + * linked list won't need to be run. + */ + ehandler = desc->emerg_handler; + if (ehandler) + return ehandler(type, regs); + rcu_read_lock(); /* @@ -139,11 +165,7 @@ static int nmi_handle(unsigned int type, struct pt_regs *regs) delta = sched_clock() - delta; trace_nmi_handler(a->handler, (int)delta, thishandled); - if (delta < nmi_longest_ns || delta < a->max_duration) - continue; - - a->max_duration = delta; - irq_work_queue(&a->irq_work); + nmi_check_duration(a, delta); } rcu_read_unlock(); @@ -158,11 +180,9 @@ int __register_nmi_handler(unsigned int type, struct nmiaction *action) struct nmi_desc *desc = nmi_to_desc(type); unsigned long flags; - if (!action->handler) + if (WARN_ON_ONCE(!action->handler || !list_empty(&action->list))) return -EINVAL; - init_irq_work(&action->irq_work, nmi_max_handler); - raw_spin_lock_irqsave(&desc->lock, flags); /* @@ -180,7 +200,7 @@ int __register_nmi_handler(unsigned int type, struct nmiaction *action) list_add_rcu(&action->list, &desc->head); else list_add_tail_rcu(&action->list, &desc->head); - + raw_spin_unlock_irqrestore(&desc->lock, flags); return 0; } @@ -189,7 +209,7 @@ EXPORT_SYMBOL(__register_nmi_handler); void unregister_nmi_handler(unsigned int type, const char *name) { struct nmi_desc *desc = nmi_to_desc(type); - struct nmiaction *n; + struct nmiaction *n, *found = NULL; unsigned long flags; raw_spin_lock_irqsave(&desc->lock, flags); @@ -203,15 +223,44 @@ void unregister_nmi_handler(unsigned int type, const char *name) WARN(in_nmi(), "Trying to free NMI (%s) from NMI context!\n", n->name); list_del_rcu(&n->list); + found = n; break; } } raw_spin_unlock_irqrestore(&desc->lock, flags); - synchronize_rcu(); + if (found) { + synchronize_rcu(); + INIT_LIST_HEAD(&found->list); + } } EXPORT_SYMBOL_GPL(unregister_nmi_handler); +/** + * set_emergency_nmi_handler - Set emergency handler + * @type: NMI type + * @handler: the emergency handler to be stored + * + * Set an emergency NMI handler which, if set, will preempt all the other + * handlers in the linked list. If a NULL handler is passed in, it will clear + * it. It is expected that concurrent calls to this function will not happen + * or the system is screwed beyond repair. + */ +void set_emergency_nmi_handler(unsigned int type, nmi_handler_t handler) +{ + struct nmi_desc *desc = nmi_to_desc(type); + + if (WARN_ON_ONCE(desc->emerg_handler == handler)) + return; + desc->emerg_handler = handler; + + /* + * Ensure the emergency handler is visible to other CPUs before + * function return + */ + smp_wmb(); +} + static void pci_serr_error(unsigned char reason, struct pt_regs *regs) { @@ -279,10 +328,9 @@ unknown_nmi_error(unsigned char reason, struct pt_regs *regs) int handled; /* - * Use 'false' as back-to-back NMIs are dealt with one level up. - * Of course this makes having multiple 'unknown' handlers useless - * as only the first one is ever run (unless it can actually determine - * if it caused the NMI) + * As a last resort, let the "unknown" handlers make a + * best-effort attempt to figure out if they can claim + * responsibility for this Unknown NMI. */ handled = nmi_handle(NMI_UNKNOWN, regs); if (handled) { @@ -292,38 +340,38 @@ unknown_nmi_error(unsigned char reason, struct pt_regs *regs) __this_cpu_add(nmi_stats.unknown, 1); - pr_emerg("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n", - reason, smp_processor_id()); + pr_emerg_ratelimited("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n", + reason, smp_processor_id()); - pr_emerg("Do you have a strange power saving mode enabled?\n"); if (unknown_nmi_panic || panic_on_unrecovered_nmi) nmi_panic(regs, "NMI: Not continuing"); - pr_emerg("Dazed and confused, but trying to continue\n"); + pr_emerg_ratelimited("Dazed and confused, but trying to continue\n"); } NOKPROBE_SYMBOL(unknown_nmi_error); static DEFINE_PER_CPU(bool, swallow_nmi); static DEFINE_PER_CPU(unsigned long, last_nmi_rip); -static void default_do_nmi(struct pt_regs *regs) +static noinstr void default_do_nmi(struct pt_regs *regs) { unsigned char reason = 0; int handled; bool b2b = false; /* - * CPU-specific NMI must be processed before non-CPU-specific - * NMI, otherwise we may lose it, because the CPU-specific - * NMI can not be detected/processed on other CPUs. - */ - - /* - * Back-to-back NMIs are interesting because they can either - * be two NMI or more than two NMIs (any thing over two is dropped - * due to NMI being edge-triggered). If this is the second half - * of the back-to-back NMI, assume we dropped things and process - * more handlers. Otherwise reset the 'swallow' NMI behaviour + * Back-to-back NMIs are detected by comparing the RIP of the + * current NMI with that of the previous NMI. If it is the same, + * it is assumed that the CPU did not have a chance to jump back + * into a non-NMI context and execute code in between the two + * NMIs. + * + * They are interesting because even if there are more than two, + * only a maximum of two can be detected (anything over two is + * dropped due to NMI being edge-triggered). If this is the + * second half of the back-to-back NMI, assume we dropped things + * and process more handlers. Otherwise, reset the 'swallow' NMI + * behavior. */ if (regs->ip == __this_cpu_read(last_nmi_rip)) b2b = true; @@ -332,6 +380,16 @@ static void default_do_nmi(struct pt_regs *regs) __this_cpu_write(last_nmi_rip, regs->ip); + instrumentation_begin(); + + if (microcode_nmi_handler_enabled() && microcode_nmi_handler()) + goto out; + + /* + * CPU-specific NMI must be processed before non-CPU-specific + * NMI, otherwise we may lose it, because the CPU-specific + * NMI can not be detected/processed on other CPUs. + */ handled = nmi_handle(NMI_LOCAL, regs); __this_cpu_add(nmi_stats.normal, handled); if (handled) { @@ -345,7 +403,7 @@ static void default_do_nmi(struct pt_regs *regs) */ if (handled > 1) __this_cpu_write(swallow_nmi, true); - return; + goto out; } /* @@ -368,16 +426,17 @@ static void default_do_nmi(struct pt_regs *regs) pci_serr_error(reason, regs); else if (reason & NMI_REASON_IOCHK) io_check_error(reason, regs); -#ifdef CONFIG_X86_32 + /* * Reassert NMI in case it became active * meanwhile as it's edge-triggered: */ - reassert_nmi(); -#endif + if (IS_ENABLED(CONFIG_X86_32)) + reassert_nmi(); + __this_cpu_add(nmi_stats.external, 1); raw_spin_unlock(&nmi_reason_lock); - return; + goto out; } raw_spin_unlock(&nmi_reason_lock); @@ -402,9 +461,9 @@ static void default_do_nmi(struct pt_regs *regs) * a 'real' unknown NMI. For example, while processing * a perf NMI another perf NMI comes in along with a * 'real' unknown NMI. These two NMIs get combined into - * one (as descibed above). When the next NMI gets + * one (as described above). When the next NMI gets * processed, it will be flagged by perf as handled, but - * noone will know that there was a 'real' unknown NMI sent + * no one will know that there was a 'real' unknown NMI sent * also. As a result it gets swallowed. Or if the first * perf NMI returns two events handled then the second * NMI will get eaten by the logic below, again losing a @@ -415,8 +474,10 @@ static void default_do_nmi(struct pt_regs *regs) __this_cpu_add(nmi_stats.swallow, 1); else unknown_nmi_error(reason, regs); + +out: + instrumentation_end(); } -NOKPROBE_SYMBOL(default_do_nmi); /* * NMIs can page fault or hit breakpoints which will cause it to lose @@ -470,71 +531,209 @@ enum nmi_states { }; static DEFINE_PER_CPU(enum nmi_states, nmi_state); static DEFINE_PER_CPU(unsigned long, nmi_cr2); +static DEFINE_PER_CPU(unsigned long, nmi_dr7); -#ifdef CONFIG_X86_64 -/* - * In x86_64, we need to handle breakpoint -> NMI -> breakpoint. Without - * some care, the inner breakpoint will clobber the outer breakpoint's - * stack. - * - * If a breakpoint is being processed, and the debug stack is being - * used, if an NMI comes in and also hits a breakpoint, the stack - * pointer will be set to the same fixed address as the breakpoint that - * was interrupted, causing that stack to be corrupted. To handle this - * case, check if the stack that was interrupted is the debug stack, and - * if so, change the IDT so that new breakpoints will use the current - * stack and not switch to the fixed address. On return of the NMI, - * switch back to the original IDT. - */ -static DEFINE_PER_CPU(int, update_debug_stack); -#endif - -dotraplinkage notrace void -do_nmi(struct pt_regs *regs, long error_code) +DEFINE_IDTENTRY_RAW(exc_nmi) { + irqentry_state_t irq_state; + struct nmi_stats *nsp = this_cpu_ptr(&nmi_stats); + + /* + * Re-enable NMIs right here when running as an SEV-ES guest. This might + * cause nested NMIs, but those can be handled safely. + */ + sev_es_nmi_complete(); + if (IS_ENABLED(CONFIG_NMI_CHECK_CPU)) + raw_atomic_long_inc(&nsp->idt_calls); + + if (arch_cpu_is_offline(smp_processor_id())) { + if (microcode_nmi_handler_enabled()) + microcode_offline_nmi_handler(); + return; + } + if (this_cpu_read(nmi_state) != NMI_NOT_RUNNING) { this_cpu_write(nmi_state, NMI_LATCHED); return; } this_cpu_write(nmi_state, NMI_EXECUTING); this_cpu_write(nmi_cr2, read_cr2()); + nmi_restart: + if (IS_ENABLED(CONFIG_NMI_CHECK_CPU)) { + WRITE_ONCE(nsp->idt_seq, nsp->idt_seq + 1); + WARN_ON_ONCE(!(nsp->idt_seq & 0x1)); + WRITE_ONCE(nsp->recv_jiffies, jiffies); + } -#ifdef CONFIG_X86_64 /* - * If we interrupted a breakpoint, it is possible that - * the nmi handler will have breakpoints too. We need to - * change the IDT such that breakpoints that happen here - * continue to use the NMI stack. + * Needs to happen before DR7 is accessed, because the hypervisor can + * intercept DR7 reads/writes, turning those into #VC exceptions. */ - if (unlikely(is_debug_stack(regs->sp))) { - debug_stack_set_zero(); - this_cpu_write(update_debug_stack, 1); - } -#endif + sev_es_ist_enter(regs); + + this_cpu_write(nmi_dr7, local_db_save()); - nmi_enter(); + irq_state = irqentry_nmi_enter(regs); inc_irq_stat(__nmi_count); - if (!ignore_nmis) + if (IS_ENABLED(CONFIG_NMI_CHECK_CPU) && ignore_nmis) { + WRITE_ONCE(nsp->idt_ignored, nsp->idt_ignored + 1); + } else if (!ignore_nmis) { + if (IS_ENABLED(CONFIG_NMI_CHECK_CPU)) { + WRITE_ONCE(nsp->idt_nmi_seq, nsp->idt_nmi_seq + 1); + WARN_ON_ONCE(!(nsp->idt_nmi_seq & 0x1)); + } default_do_nmi(regs); + if (IS_ENABLED(CONFIG_NMI_CHECK_CPU)) { + WRITE_ONCE(nsp->idt_nmi_seq, nsp->idt_nmi_seq + 1); + WARN_ON_ONCE(nsp->idt_nmi_seq & 0x1); + } + } - nmi_exit(); + irqentry_nmi_exit(regs, irq_state); -#ifdef CONFIG_X86_64 - if (unlikely(this_cpu_read(update_debug_stack))) { - debug_stack_reset(); - this_cpu_write(update_debug_stack, 0); - } -#endif + local_db_restore(this_cpu_read(nmi_dr7)); + + sev_es_ist_exit(); if (unlikely(this_cpu_read(nmi_cr2) != read_cr2())) write_cr2(this_cpu_read(nmi_cr2)); + if (IS_ENABLED(CONFIG_NMI_CHECK_CPU)) { + WRITE_ONCE(nsp->idt_seq, nsp->idt_seq + 1); + WARN_ON_ONCE(nsp->idt_seq & 0x1); + WRITE_ONCE(nsp->recv_jiffies, jiffies); + } if (this_cpu_dec_return(nmi_state)) goto nmi_restart; } -NOKPROBE_SYMBOL(do_nmi); + +#if IS_ENABLED(CONFIG_KVM_INTEL) +DEFINE_IDTENTRY_RAW(exc_nmi_kvm_vmx) +{ + exc_nmi(regs); +} +EXPORT_SYMBOL_FOR_KVM(asm_exc_nmi_kvm_vmx); +#endif + +#ifdef CONFIG_NMI_CHECK_CPU + +static char *nmi_check_stall_msg[] = { +/* */ +/* +--------- nmi_seq & 0x1: CPU is currently in NMI handler. */ +/* | +------ cpu_is_offline(cpu) */ +/* | | +--- nsp->idt_calls_snap != atomic_long_read(&nsp->idt_calls): */ +/* | | | NMI handler has been invoked. */ +/* | | | */ +/* V V V */ +/* 0 0 0 */ "NMIs are not reaching exc_nmi() handler", +/* 0 0 1 */ "exc_nmi() handler is ignoring NMIs", +/* 0 1 0 */ "CPU is offline and NMIs are not reaching exc_nmi() handler", +/* 0 1 1 */ "CPU is offline and exc_nmi() handler is legitimately ignoring NMIs", +/* 1 0 0 */ "CPU is in exc_nmi() handler and no further NMIs are reaching handler", +/* 1 0 1 */ "CPU is in exc_nmi() handler which is legitimately ignoring NMIs", +/* 1 1 0 */ "CPU is offline in exc_nmi() handler and no more NMIs are reaching exc_nmi() handler", +/* 1 1 1 */ "CPU is offline in exc_nmi() handler which is legitimately ignoring NMIs", +}; + +void nmi_backtrace_stall_snap(const struct cpumask *btp) +{ + int cpu; + struct nmi_stats *nsp; + + for_each_cpu(cpu, btp) { + nsp = per_cpu_ptr(&nmi_stats, cpu); + nsp->idt_seq_snap = READ_ONCE(nsp->idt_seq); + nsp->idt_nmi_seq_snap = READ_ONCE(nsp->idt_nmi_seq); + nsp->idt_ignored_snap = READ_ONCE(nsp->idt_ignored); + nsp->idt_calls_snap = atomic_long_read(&nsp->idt_calls); + } +} + +void nmi_backtrace_stall_check(const struct cpumask *btp) +{ + int cpu; + int idx; + unsigned long nmi_seq; + unsigned long j = jiffies; + char *modp; + char *msgp; + char *msghp; + struct nmi_stats *nsp; + + for_each_cpu(cpu, btp) { + nsp = per_cpu_ptr(&nmi_stats, cpu); + modp = ""; + msghp = ""; + nmi_seq = READ_ONCE(nsp->idt_nmi_seq); + if (nsp->idt_nmi_seq_snap + 1 == nmi_seq && (nmi_seq & 0x1)) { + msgp = "CPU entered NMI handler function, but has not exited"; + } else if (nsp->idt_nmi_seq_snap == nmi_seq || + nsp->idt_nmi_seq_snap + 1 == nmi_seq) { + idx = ((nmi_seq & 0x1) << 2) | + (cpu_is_offline(cpu) << 1) | + (nsp->idt_calls_snap != atomic_long_read(&nsp->idt_calls)); + msgp = nmi_check_stall_msg[idx]; + if (nsp->idt_ignored_snap != READ_ONCE(nsp->idt_ignored) && (idx & 0x1)) + modp = ", but OK because ignore_nmis was set"; + if (nsp->idt_nmi_seq_snap + 1 == nmi_seq) + msghp = " (CPU exited one NMI handler function)"; + else if (nmi_seq & 0x1) + msghp = " (CPU currently in NMI handler function)"; + else + msghp = " (CPU was never in an NMI handler function)"; + } else { + msgp = "CPU is handling NMIs"; + } + pr_alert("%s: CPU %d: %s%s%s\n", __func__, cpu, msgp, modp, msghp); + pr_alert("%s: last activity: %lu jiffies ago.\n", + __func__, j - READ_ONCE(nsp->recv_jiffies)); + } +} + +#endif + +#ifdef CONFIG_X86_FRED +/* + * With FRED, CR2/DR6 is pushed to #PF/#DB stack frame during FRED + * event delivery, i.e., there is no problem of transient states. + * And NMI unblocking only happens when the stack frame indicates + * that so should happen. + * + * Thus, the NMI entry stub for FRED is really straightforward and + * as simple as most exception handlers. As such, #DB is allowed + * during NMI handling. + */ +DEFINE_FREDENTRY_NMI(exc_nmi) +{ + irqentry_state_t irq_state; + + if (arch_cpu_is_offline(smp_processor_id())) { + if (microcode_nmi_handler_enabled()) + microcode_offline_nmi_handler(); + return; + } + + /* + * Save CR2 for eventual restore to cover the case where the NMI + * hits the VMENTER/VMEXIT region where guest CR2 is life. This + * prevents guest state corruption in case that the NMI handler + * takes a page fault. + */ + this_cpu_write(nmi_cr2, read_cr2()); + + irq_state = irqentry_nmi_enter(regs); + + inc_irq_stat(__nmi_count); + default_do_nmi(regs); + + irqentry_nmi_exit(regs, irq_state); + + if (unlikely(this_cpu_read(nmi_cr2) != read_cr2())) + write_cr2(this_cpu_read(nmi_cr2)); +} +#endif void stop_nmi(void) { @@ -551,4 +750,3 @@ void local_touch_nmi(void) { __this_cpu_write(last_nmi_rip, 0); } -EXPORT_SYMBOL_GPL(local_touch_nmi); |
