1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * NMI backtrace support 4 * 5 * Gratuitously copied from arch/x86/kernel/apic/hw_nmi.c by Russell King, 6 * with the following header: 7 * 8 * HW NMI watchdog support 9 * 10 * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc. 11 * 12 * Arch specific calls to support NMI watchdog 13 * 14 * Bits copied from original nmi.c file 15 */ 16 #include <linux/cpumask.h> 17 #include <linux/delay.h> 18 #include <linux/kprobes.h> 19 #include <linux/stringify.h> 20 #include <linux/nmi.h> 21 #include <linux/cpu.h> 22 #include <linux/export.h> 23 #include <linux/sched/debug.h> 24 25 #ifdef arch_trigger_cpumask_backtrace 26 /* For reliability, we're prepared to waste bits here. */ 27 static DECLARE_BITMAP(backtrace_mask, NR_CPUS) __read_mostly; 28 29 /* "in progress" flag of arch_trigger_cpumask_backtrace */ 30 static unsigned long backtrace_flag; 31 32 #define NMI_BT_TIMEOUT_SEC 10 33 34 /* 35 * When raise() is called it will be passed a pointer to the 36 * backtrace_mask. Architectures that call nmi_cpu_backtrace() 37 * directly from their raise() functions may rely on the mask 38 * they are passed being updated as a side effect of this call. 39 */ 40 void nmi_trigger_cpumask_backtrace(const cpumask_t *mask, 41 int exclude_cpu, 42 void (*raise)(cpumask_t *mask)) 43 { 44 int i, this_cpu = get_cpu(); 45 46 if (test_and_set_bit(0, &backtrace_flag)) { 47 /* 48 * If there is already a trigger_all_cpu_backtrace() in progress 49 * (backtrace_flag == 1), don't output double cpu dump infos. 50 */ 51 put_cpu(); 52 return; 53 } 54 55 cpumask_copy(to_cpumask(backtrace_mask), mask); 56 if (exclude_cpu != -1) 57 cpumask_clear_cpu(exclude_cpu, to_cpumask(backtrace_mask)); 58 59 /* 60 * Don't try to send an NMI to this cpu; it may work on some 61 * architectures, but on others it may not, and we'll get 62 * information at least as useful just by doing a dump_stack() here. 63 * Note that nmi_cpu_backtrace(NULL) will clear the cpu bit. 64 */ 65 if (cpumask_test_cpu(this_cpu, to_cpumask(backtrace_mask))) 66 nmi_cpu_backtrace(NULL); 67 68 if (!cpumask_empty(to_cpumask(backtrace_mask))) { 69 pr_info("Sending NMI from CPU %d to CPUs %*pbl:\n", 70 this_cpu, nr_cpumask_bits, to_cpumask(backtrace_mask)); 71 nmi_backtrace_stall_snap(to_cpumask(backtrace_mask)); 72 raise(to_cpumask(backtrace_mask)); 73 } 74 75 /* Wait for up to NMI_BT_TIMEOUT_SEC seconds for all CPUs to do the backtrace */ 76 for (i = 0; i < NMI_BT_TIMEOUT_SEC * 1000; i++) { 77 if (cpumask_empty(to_cpumask(backtrace_mask))) 78 break; 79 mdelay(1); 80 touch_softlockup_watchdog(); 81 } 82 83 if (!cpumask_empty(to_cpumask(backtrace_mask))) { 84 pr_warn("After " __stringify(NMI_BT_TIMEOUT_SEC) " seconds, these CPUS still haven't responded to the NMI: %*pbl\n", 85 cpumask_pr_args(to_cpumask(backtrace_mask))); 86 87 nmi_backtrace_stall_check(to_cpumask(backtrace_mask)); 88 } 89 90 /* 91 * Force flush any remote buffers that might be stuck in IRQ context 92 * and therefore could not run their irq_work. 93 */ 94 printk_trigger_flush(); 95 96 clear_bit_unlock(0, &backtrace_flag); 97 put_cpu(); 98 } 99 100 // Dump stacks even for idle CPUs. 101 static bool backtrace_idle; 102 module_param(backtrace_idle, bool, 0644); 103 104 bool nmi_cpu_backtrace(struct pt_regs *regs) 105 { 106 int cpu = smp_processor_id(); 107 unsigned long flags; 108 109 if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { 110 /* 111 * Allow nested NMI backtraces while serializing 112 * against other CPUs. 113 */ 114 printk_cpu_sync_get_irqsave(flags); 115 if (!READ_ONCE(backtrace_idle) && regs && cpu_in_idle(instruction_pointer(regs))) { 116 pr_warn("NMI backtrace for cpu %d skipped: idling at %pS\n", 117 cpu, (void *)instruction_pointer(regs)); 118 } else { 119 pr_warn("NMI backtrace for cpu %d\n", cpu); 120 if (regs) 121 show_regs(regs); 122 else 123 dump_stack(); 124 } 125 printk_cpu_sync_put_irqrestore(flags); 126 cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); 127 return true; 128 } 129 130 return false; 131 } 132 NOKPROBE_SYMBOL(nmi_cpu_backtrace); 133 134 void cpumask_backtrace(const cpumask_t *mask, int exclude_cpu) 135 { 136 arch_trigger_cpumask_backtrace(mask, exclude_cpu); 137 } 138 EXPORT_SYMBOL_GPL(cpumask_backtrace); 139 #endif 140