7 月 28 日下午，我们与马玉昆老师进行了会议交流。马老师表示，我们认为读写锁要使用独立探测点的观点是正确的，并鼓励我们多动手查资料。针对我提出的问题“不同进程申请同一锁时，是否应该记录时间，以判断是否发生冲突”，他表示我们的工具更多地是一种静态的分析，没有必要判断出是否发生冲突，因此记录时间不是特别必要，但应该能分析出进程是否持有相同的锁。我发现似乎一分钟就有上千个不同的自旋锁的获取，模块的内存占用只会越来越大，是否需要定期清理、删除一部分信息，马老师表示，一个进程申请多次同一把锁的话，只记录第一次的调用栈信息就行了，这样最终我们模块的内存消耗应该会有一个上限值。针对我提出的锁的地址问题，马老师认为，同一锁在不同进程中的地址是相同的。
上周只使用 gdb 追踪了一部分有关自旋锁的情况，我们继续看一些其他情况。
(gdb) hbreak _raw_spin_lock_irq
Hardware assisted breakpoint 2 at 0xffffffff819ac03b: file kernel/locking/spinlock.c, line 166.
(gdb) c
Continuing.
Thread 1 hit Breakpoint 2, _raw_spin_lock_irq (
lock=lock@entry=0xffffffff830a9800 <ucounts_lock>)
at kernel/locking/spinlock.c:166
166 {
(gdb) s
167 __raw_spin_lock_irq(lock);
(gdb) s
__raw_spin_lock_irq (lock=0xffffffff830a9800 <ucounts_lock>)
at kernel/locking/spinlock.c:167
167 __raw_spin_lock_irq(lock);
(gdb) s
arch_irqs_disabled () at kernel/locking/spinlock.c:167
167 __raw_spin_lock_irq(lock);
(gdb) s
arch_local_save_flags () at ./arch/x86/include/asm/paravirt.h:789
789 return PVOP_CALLEE0(unsigned long, pv_irq_ops.save_fl);
(gdb) s
native_save_fl () at arch/x86/kernel/irqflags.S:11
11 pushf
(gdb) s
native_save_fl () at arch/x86/kernel/irqflags.S:12
12 pop %_ASM_AX
(gdb) s
native_save_fl () at arch/x86/kernel/irqflags.S:13
13 ret
(gdb) s
__raw_spin_lock_irq (lock=0xffffffff830a9800 <ucounts_lock>)
at ./include/linux/spinlock_api_smp.h:126
126 local_irq_disable();
(gdb) s
arch_local_irq_disable () at ./arch/x86/include/asm/paravirt.h:799
799 PVOP_VCALLEE0(pv_irq_ops.irq_disable);
(gdb) s
native_irq_disable () at ./arch/x86/include/asm/irqflags.h:46
46 asm volatile("cli": : :"memory");
(gdb) s
__raw_spin_lock_irq (lock=0xffffffff830a9800 <ucounts_lock>)
at ./include/linux/spinlock_api_smp.h:126
126 local_irq_disable();
(gdb) s
trace_hardirqs_off () at kernel/trace/trace_preemptirq.c:75
75 {
(gdb) s
78 if (!this_cpu_read(tracing_irq_cpu)) {
(gdb) s
79 this_cpu_write(tracing_irq_cpu, 1);
(gdb) s
81 if (!in_nmi())
(gdb) s
preempt_count () at ./arch/x86/include/asm/preempt.h:26
26 return raw_cpu_read_4(__preempt_count) & ~PREEMPT_NEED_RESCHED;
(gdb) s
82 trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
(gdb) s
trace_irq_disable_rcuidle (parent_ip=0, ip=18446744071588986995)
at kernel/trace/trace_preemptirq.c:82
82 trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
(gdb) s
static_key_false (key=0xffffffff829e8688 <__tracepoint_irq_disable+8>)
at kernel/trace/trace_preemptirq.c:82
82 trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
(gdb) s
arch_static_branch (branch=false,
key=0xffffffff829e8688 <__tracepoint_irq_disable+8>)
at ./arch/x86/include/asm/jump_label.h:38
38 asm_volatile_goto("1:"
(gdb) s
trace_hardirqs_off () at kernel/trace/trace_preemptirq.c:82
82 trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
(gdb) s
trace_irq_disable_rcuidle (parent_ip=0, ip=18446744071588986995)
at kernel/trace/trace_preemptirq.c:82
82 trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
(gdb) s
static_key_false (key=<optimized out>) at kernel/trace/trace_preemptirq.c:82
82 trace_irq_disable_rcuidle(CALLER_ADDR0, CALLER_ADDR1);
(gdb) s
arch_static_branch (branch=<optimized out>, key=<optimized out>)
at ./arch/x86/include/asm/jump_label.h:46
46 return false;
(gdb) s
__raw_spin_lock_irq (lock=0xffffffff830a9800 <ucounts_lock>)
at ./include/linux/spinlock_api_smp.h:129
129 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
do_raw_spin_lock (lock=0xffffffff830a9800 <ucounts_lock>)
at ./include/linux/spinlock_api_smp.h:129
129 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
queued_spin_lock (lock=0xffffffff830a9800 <ucounts_lock>)
at ./include/linux/spinlock_api_smp.h:129
129 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
atomic_cmpxchg (new=1, old=0, v=0xffffffff830a9800 <ucounts_lock>)
at ./include/linux/spinlock_api_smp.h:129
129 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
arch_atomic_cmpxchg (new=1, old=0, v=0xffffffff830a9800 <ucounts_lock>)
at ./arch/x86/include/asm/atomic.h:192
192 return arch_cmpxchg(&v->counter, old, new);
(gdb) s
queued_spin_lock (lock=0xffffffff830a9800 <ucounts_lock>)
at ./include/asm-generic/qspinlock.h:87
87 if (likely(val == 0))
(gdb) s
get_ucounts (ns=ns@entry=0xffffffff82850780 <init_user_ns>, uid=...)
at kernel/ucount.c:152
152 ucounts = find_ucounts(ns, uid, hashent);
(gdb) hbreak _raw_spin_lock_bh
Hardware ass isted breakpoint 3 at 0xffffffff819abcdb: file kernel/locking/spinlock.c, line 174.
(gdb) delete breakpoints 2
(gdb) c
Continuing.
Thread 1 hit Breakpoint 3, _raw_spin_lock_bh (
lock=lock@entry=0xffffffff831dc0bc <cgroup_idr_lock>)
at kernel/locking/spinlock.c:174
174 {
(gdb) s
175 __raw_spin_lock_bh(lock);
(gdb) s
__raw_spin_lock_bh (lock=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./include/linux/spinlock_api_smp.h:134
134 __local_bh_disable_ip(_RET_IP_, SOFTIRQ_LOCK_OFFSET);
(gdb) s
__local_bh_disable_ip (ip=18446744071580480599, cnt=cnt@entry=512)
at kernel/softirq.c:117
117 {
(gdb) s
120 WARN_ON_ONCE(in_irq());
(gdb) s
preempt_count () at ./arch/x86/include/asm/preempt.h:26
26 return raw_cpu_read_4(__preempt_count) & ~PREEMPT_NEED_RESCHED;
(gdb) s
__local_bh_disable_ip (ip=18446744071580480599, cnt=cnt@entry=512)
at kernel/softirq.c:120
120 WARN_ON_ONCE(in_irq());
(gdb) s
122 raw_local_irq_save(flags);
(gdb) s
arch_local_irq_save () at kernel/softirq.c:122
122 raw_local_irq_save(flags);
(gdb) s
arch_local_save_flags () at ./arch/x86/include/asm/paravirt.h:811
811 f = arch_local_save_flags();
(gdb) s
native_save_fl () at arch/x86/kernel/irqflags.S:11
11 pushf
(gdb) s
native_save_fl () at arch/x86/kernel/irqflags.S:12
12 pop %_ASM_AX
(gdb) s
native_save_fl () at arch/x86/kernel/irqflags.S:13
13 ret
(gdb) s
arch_local_irq_save () at ./arch/x86/include/asm/paravirt.h:812
812 arch_local_irq_disable();
(gdb) s
arch_local_irq_disable () at ./arch/x86/include/asm/paravirt.h:812
812 arch_local_irq_disable();
(gdb) s
native_irq_disable () at ./arch/x86/include/asm/irqflags.h:46
46 asm volatile("cli": : :"memory");
(gdb) s
__local_bh_disable_ip (ip=<optimized out>, cnt=cnt@entry=512)
at kernel/softirq.c:130
130 __preempt_count_add(cnt);
(gdb) s
__preempt_count_add (val=512) at ./arch/x86/include/asm/preempt.h:79
79 raw_cpu_add_4(__preempt_count, val);
(gdb) s
__local_bh_disable_ip (ip=<optimized out>, cnt=cnt@entry=512)
at kernel/softirq.c:136
136 raw_local_irq_restore(flags);
(gdb) s
arch_local_irq_restore (f=838) at ./arch/x86/include/asm/paravirt.h:794
794 PVOP_VCALLEE1(pv_irq_ops.restore_fl, f);
(gdb) s
native_restore_fl (flags=838) at ./arch/x86/include/asm/irqflags.h:38
38 asm volatile("push %0 ; popf"
(gdb) s
__local_bh_disable_ip (ip=<optimized out>, cnt=cnt@entry=512)
at kernel/softirq.c:142
142 trace_preempt_off(CALLER_ADDR0, get_lock_parent_ip());
(gdb) s
__raw_spin_lock_bh (lock=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./include/linux/spinlock_api_smp.h:136
136 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
do_raw_spin_lock (lock=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./include/linux/spinlock_api_smp.h:136
136 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
queued_spin_lock (lock=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./include/linux/spinlock_api_smp.h:136
136 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
atomic_cmpxchg (new=1, old=0, v=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./include/linux/spinlock_api_smp.h:136
136 LOCK_CONTENDED(lock, do_raw_spin_trylock, do_raw_spin_lock);
(gdb) s
arch_atomic_cmpxchg (new=1, old=0, v=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./arch/x86/include/asm/atomic.h:192
192 return arch_cmpxchg(&v->counter, old, new);
(gdb) s
queued_spin_lock (lock=0xffffffff831dc0bc <cgroup_idr_lock>)
at ./include/asm-generic/qspinlock.h:87
87 if (likely(val == 0))
(gdb) s
cgroup_idr_alloc (idr=idr@entry=0xffffffff828b7310 <cpuset_cgrp_subsys+176>,
ptr=ptr@entry=0xffffffff828b8700 <top_cpuset>, start=start@entry=1,
end=end@entry=2, gfp_mask=gfp_mask@entry=6291648)
at kernel/cgroup/cgroup.c:303
303 ret = idr_alloc(idr, ptr, start, end, gfp_mask & ~__GFP_DIRECT_RECLAIM);
(gdb) delete breakpoints 4
(gdb) hbreak _raw_spin_trylock
Hardware assisted breakpoint 5 at 0xffffffff819abc50: file kernel/locking/spinlock.c, line 134.
(gdb) c
Continuing.
Thread 1 hit Breakpoint 5, _raw_spin_trylock (
lock=lock@entry=0xffffffff83b37f40 <printing_lock>)
at kernel/locking/spinlock.c:134
134 {
(gdb) s
135 return __raw_spin_trylock(lock);
(gdb) s
__raw_spin_trylock (lock=lock@entry=0xffffffff83b37f40 <printing_lock>)
at kernel/locking/spinlock.c:135
135 return __raw_spin_trylock(lock);
(gdb) s
do_raw_spin_trylock (lock=0xffffffff83b37f40 <printing_lock>)
at kernel/locking/spinlock.c:135
135 return __raw_spin_trylock(lock);
(gdb) s
queued_spin_trylock (lock=0xffffffff83b37f40 <printing_lock>)
at kernel/locking/spinlock.c:135
135 return __raw_spin_trylock(lock);
(gdb) s
atomic_read (v=0xffffffff83b37f40 <printing_lock>)
at kernel/locking/spinlock.c:135
135 return __raw_spin_trylock(lock);
(gdb) s
arch_atomic_read (v=0xffffffff83b37f40 <printing_lock>)
at kernel/locking/spinlock.c:135
135 return __raw_spin_trylock(lock);
(gdb) s
__read_once_size (size=4, res=<synthetic pointer>,
p=0xffffffff83b37f40 <printing_lock>) at ./include/linux/compiler.h:276
276 __READ_ONCE_SIZE;
(gdb) s
queued_spin_trylock (lock=0xffffffff83b37f40 <printing_lock>)
at ./arch/x86/include/asm/atomic.h:29
29 return READ_ONCE((v)->counter);
(gdb) s
71 (atomic_cmpxchg_acquire(&lock->val, 0, _Q_LOCKED_VAL) == 0))
(gdb) s
atomic_cmpxchg (new=1, old=0, v=0xffffffff83b37f40 <printing_lock>)
at ./include/asm-generic/qspinlock.h:71
71 (atomic_cmpxchg_acquire(&lock->val, 0, _Q_LOCKED_VAL) == 0))
(gdb) s
arch_atomic_cmpxchg (new=1, old=0, v=0xffffffff83b37f40 <printing_lock>)
at ./arch/x86/include/asm/atomic.h:192
192 return arch_cmpxchg(&v->counter, old, new);
(gdb) s
vt_console_print (co=<optimized out>,
b=0xffffffff830afa20 <text> "[ 0.001000] x86/cpu: User Mode Instruction Prevention (UMIP) activated\nbytes, vmalloc)\n\n6)\n_idle_ns: 440795242287 ns\nss, 405916K reserved, 0K cma-reserved)\ninux 8 at https://catalog.redhat.com.\n=vg"..., count=74) at drivers/tty/vt/vt.c:2780
#define raw_spin_trylock_irq(lock) \
({ \
local_irq_disable(); \
raw_spin_trylock(lock) ? \
1 : ({ local_irq_enable(); 0; }); \
})#define raw_spin_trylock_irqsave(lock, flags) \
({ \
local_irq_save(flags); \
raw_spin_trylock(lock) ? \
1 : ({ local_irq_restore(flags); 0; }); \
})以上两个宏最后都会调用 raw_spin_trylock，但其他几种情况即 _raw_spin_lock_bh、_raw_spin_lock_irq、_raw_spin_lock_irqsave、_raw_spin_trylock 都需要追踪，其中 _raw_spin_trylock 情况似乎更加复杂（涉及到是否尝试加锁成功），我们先搁置它。
我曾经认为，KProbe 会自动关抢占和中断，但却发现在开发中常常遇到死锁。经过重新查找资料，我发现在 SMP 上 KProbe 不会像我想象的那样串行化，而是“多个探测点处理函数或同一处理函数的多个实例能够在不同的 CPU 上同时运行。”针对有关的问题，我希望用 Linux 的一些调试相关的编译选项来尝试。
我打开 KCSAN，看了一阵子，却没发现问题在哪。
编译时却出现：
include/linux/mm_types.h:576: seqcount_t write_protect_seq is larger than unsigned long rh_reserved3. Disable CONFIG_RH_KABI_SIZE_ALIGN_CHECKS if debugging.
把 CONFIG_RH_KABI_SIZE_ALIGN_CHECKS 选项关掉后，编译，使用新内核启动，终于发现有这样的输出：
[ 94.672481] ================================
[ 94.674330] WARNING: inconsistent lock state
[ 94.676280] 4.18.0-348.7.1.el8.qunchuwolao.x86_64 #1 Tainted: G OE ---------r- -
[ 94.681341] --------------------------------
[ 94.684226] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
[ 94.687553] swapper/3/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
[ 94.690554] ffffffffa0b5aed8 (&table_lock){?.-.}-{2:2}, at: pre_handler_spin_lock_irqsave+0x7a/0xc7 [lock_test]
[ 94.695917] {IN-HARDIRQ-W} state was registered at:
[ 94.699077] __lock_acquire+0x48a/0x80a
[ 94.701277] lock_acquire+0x1b1/0x30b
[ 94.703308] _raw_spin_lock+0x38/0x46
[ 94.705371] pre_handler_spin_lock_irqsave+0x7a/0xc7 [lock_test]
[ 94.708575] kprobe_ftrace_handler+0xbd/0x143
[ 94.710975] ftrace_ops_list_func+0xfa/0x182
[ 94.713366] ftrace_regs_call+0x5/0xd3
[ 94.715020] _raw_spin_lock_irqsave+0x5/0x68
[ 94.716880] hrtimer_interrupt+0x6e/0x34d
[ 94.718632] local_apic_timer_interrupt+0x3d/0x5c
[ 94.721887] smp_apic_timer_interrupt+0xe1/0x23d
[ 94.724428] apic_timer_interrupt+0xf/0x20
[ 94.726924] ftrace_get_addr_curr+0x15/0xac
[ 94.729685] add_breakpoints+0xe/0x4a
[ 94.731651] ftrace_replace_code+0x33/0x156
[ 94.734024] ftrace_modify_all_code+0x84/0xc8
[ 94.736374] arch_ftrace_update_code+0x22/0x3c
[ 94.738659] ftrace_run_update_code+0x13/0x30
[ 94.740919] ftrace_startup_enable+0x28/0x2f
[ 94.743322] ftrace_startup+0x7d/0xb3
[ 94.745275] register_ftrace_function+0x35/0x85
[ 94.748111] __arm_kprobe_ftrace+0x99/0xe0
[ 94.750509] arm_kprobe_ftrace+0x43/0x45
[ 94.752670] arm_kprobe+0x57/0x5b
[ 94.754508] register_kprobe+0x295/0x310
[ 94.756733] register_kprobes+0x2f/0x66
[ 94.758948] start_module+0xa9/0x1000 [lock_test]
[ 94.763396] do_one_initcall+0x103/0x337
[ 94.767621] do_init_module+0xa2/0x319
[ 94.771915] load_module+0x5af/0x607
[ 94.775982] __do_sys_finit_module+0x17b/0x193
[ 94.780675] __se_sys_finit_module+0xa/0xb
[ 94.784804] __x64_sys_finit_module+0x3e/0x43
[ 94.789191] do_syscall_64+0xa0/0x2b5
[ 94.793005] entry_SYSCALL_64_after_hwframe+0x6a/0xdf
[ 94.798510] irq event stamp: 389052
[ 94.802346] hardirqs last enabled at (389052): [<ffffffff81002d5a>] trace_hardirqs_on_thunk+0x1a/0x20
[ 94.809258] hardirqs last disabled at (389051): [<ffffffff81002d7a>] trace_hardirqs_off_thunk+0x1a/0x20
[ 94.816045] softirqs last enabled at (389048): [<ffffffff811358df>] _local_bh_enable+0x1d/0x22
[ 94.822653] softirqs last disabled at (389049): [<ffffffff81136211>] irq_exit+0xbb/0x118
[ 94.828817]
[ 94.828817] other info that might help us debug this:
[ 94.837872] Possible unsafe locking scenario:
[ 94.837872]
[ 94.844719] CPU0
[ 94.847774] ----
[ 94.850891] lock(&table_lock);
[ 94.854360] <Interrupt>
[ 94.857357] lock(&table_lock);
[ 94.860490]
[ 94.860490] *** DEADLOCK ***
[ 94.860490]
[ 94.868665] no locks held by swapper/3/0.
[ 94.872377]
[ 94.872377] stack backtrace:
[ 94.877853] CPU: 3 PID: 0 Comm: swapper/3 Tainted: G OE ---------r- - 4.18.0-348.7.1.el8.qunchuwolao.x86_64 #1
[ 94.885547] Hardware name: Red Hat KVM, BIOS 1.13.0-2.module_el8.5.0+746+bbd5d70c 04/01/2014
[ 94.891651] Call Trace:
[ 94.894598] <IRQ>
[ 94.897462] ? __dump_stack+0x1b/0x1c
[ 94.901493] ? dump_stack+0x84/0xc7
[ 94.905286] ? print_usage_bug+0x1c8/0x1d1
[ 94.909142] ? mark_lock_irq+0x18c/0x302
[ 94.912999] ? mark_lock+0x1f4/0x271
[ 94.916698] ? mark_usage+0x16b/0x1a9
[ 94.920482] ? __lock_acquire+0x48a/0x80a
[ 94.924485] ? _raw_spin_lock_irqsave+0x5/0x68
[ 94.928369] ? lock_acquire+0x1b1/0x30b
[ 94.932057] ? pre_handler_spin_lock_irqsave+0x7a/0xc7 [lock_test]
[ 94.937067] ? rcu_read_unlock+0x57/0x57
[ 94.940925] ? _raw_write_lock_irqsave+0x68/0x68
[ 94.945071] ? mark_held_locks+0x1f/0xa7
[ 94.948888] ? trace_hardirqs_on_thunk+0x1a/0x20
[ 94.954163] ? _raw_spin_lock_irqsave+0x5/0x68
[ 94.958385] ? _raw_spin_lock+0x38/0x46
[ 94.962234] ? pre_handler_spin_lock_irqsave+0x7a/0xc7 [lock_test]
[ 94.967147] ? pre_handler_spin_lock_irqsave+0x7a/0xc7 [lock_test]
[ 94.971952] ? _raw_spin_lock_irqsave+0x1/0x68
[ 94.976042] ? kprobe_ftrace_handler+0xbd/0x143
[ 94.980149] ? _raw_write_lock_irqsave+0x68/0x68
[ 94.984354] ? update_blocked_averages+0x90/0x27a
[ 94.988462] ? ftrace_ops_list_func+0xfa/0x182
[ 94.992892] ? ftrace_regs_call+0x5/0xd3
[ 94.996606] ? update_blocked_averages+0x80/0x27a
[ 95.000702] ? _raw_spin_lock_irqsave+0x1/0x68
[ 95.004765] ? _raw_spin_lock_irqsave+0x5/0x68
[ 95.008636] ? update_blocked_averages+0x90/0x27a
[ 95.012572] ? __update_blocked_fair+0x76e/0x76e
[ 95.017016] ? __kasan_check_read+0x10/0x11
[ 95.021592] ? run_rebalance_domains+0x59/0x65
[ 95.025460] ? __do_softirq+0x184/0x4fb
[ 95.028990] ? kvm_sched_clock_read+0x5/0xd
[ 95.032716] ? irq_exit+0xbb/0x118
[ 95.036143] ? smp_apic_timer_interrupt+0x1d1/0x23d
[ 95.040946] ? apic_timer_interrupt+0xf/0x20
[ 95.044797] </IRQ>
[ 95.047611] ? lockdep_hardirqs_on+0x116/0x198
[ 95.051603] ? native_safe_halt+0xe/0xf
[ 95.055039] ? default_idle+0xf/0x12
[ 95.058598] ? arch_cpu_idle+0x11/0x12
[ 95.062123] ? default_idle_call+0xd9/0x242
[ 95.065980] ? cpuidle_idle_call+0x159/0x1df
[ 95.069696] ? arch_cpu_idle_exit+0x28/0x28
[ 95.073465] ? __asan_loadN+0xe/0xf
[ 95.076717] ? tsc_verify_tsc_adjust+0x70/0x125
[ 95.080756] ? do_idle+0xdf/0x12c
[ 95.084953] ? cpu_startup_entry+0xc2/0xc4
[ 95.088580] ? cpu_in_idle+0x24/0x24
[ 95.091947] ? clockevents_config_and_register+0x4e/0x55
[ 95.096136] ? start_secondary+0x18f/0x1c0
[ 95.099829] ? smp_callin+0x1ab/0x1ab
[ 95.103278] ? start_cpu0+0xc/0xc
[ 95.106395] ? secondary_startup_64_no_verify+0xc2/0xcb
原来 KProbe 运行时并不会关中断？带着这样的疑问，我编写了一段程序加以测试，结果证明，确实有时 KProbe 运行时本地中断未被禁止：
