1 .. _NMI_rcu_doc:
2
3 Using RCU to Protect Dynamic NMI Handlers
4 =========================================
5
6
7 Although RCU is usually used to protect read-mostly data structures,
8 it is possible to use RCU to provide dynamic non-maskable interrupt
9 handlers, as well as dynamic irq handlers. This document describes
10 how to do this, drawing loosely from Zwane Mwaikambo's NMI-timer
11 work in an old version of "arch/x86/kernel/traps.c".
12
13 The relevant pieces of code are listed below, each followed by a
14 brief explanation::
15
16 static int dummy_nmi_callback(struct pt_regs *regs, int cpu)
17 {
18 return 0;
19 }
20
21 The dummy_nmi_callback() function is a "dummy" NMI handler that does
22 nothing, but returns zero, thus saying that it did nothing, allowing
23 the NMI handler to take the default machine-specific action::
24
25 static nmi_callback_t nmi_callback = dummy_nmi_callback;
26
27 This nmi_callback variable is a global function pointer to the current
28 NMI handler::
29
30 void do_nmi(struct pt_regs * regs, long error_code)
31 {
32 int cpu;
33
34 nmi_enter();
35
36 cpu = smp_processor_id();
37 ++nmi_count(cpu);
38
39 if (!rcu_dereference_sched(nmi_callback)(regs, cpu))
40 default_do_nmi(regs);
41
42 nmi_exit();
43 }
44
45 The do_nmi() function processes each NMI. It first disables preemption
46 in the same way that a hardware irq would, then increments the per-CPU
47 count of NMIs. It then invokes the NMI handler stored in the nmi_callback
48 function pointer. If this handler returns zero, do_nmi() invokes the
49 default_do_nmi() function to handle a machine-specific NMI. Finally,
50 preemption is restored.
51
52 In theory, rcu_dereference_sched() is not needed, since this code runs
53 only on i386, which in theory does not need rcu_dereference_sched()
54 anyway. However, in practice it is a good documentation aid, particularly
55 for anyone attempting to do something similar on Alpha or on systems
56 with aggressive optimizing compilers.
57
58 Quick Quiz:
59 Why might the rcu_dereference_sched() be necessary on Alpha, given that the code referenced by the pointer is read-only?
60
61 :ref:`Answer to Quick Quiz <answer_quick_quiz_NMI>`
62
63 Back to the discussion of NMI and RCU::
64
65 void set_nmi_callback(nmi_callback_t callback)
66 {
67 rcu_assign_pointer(nmi_callback, callback);
68 }
69
70 The set_nmi_callback() function registers an NMI handler. Note that any
71 data that is to be used by the callback must be initialized up -before-
72 the call to set_nmi_callback(). On architectures that do not order
73 writes, the rcu_assign_pointer() ensures that the NMI handler sees the
74 initialized values::
75
76 void unset_nmi_callback(void)
77 {
78 rcu_assign_pointer(nmi_callback, dummy_nmi_callback);
79 }
80
81 This function unregisters an NMI handler, restoring the original
82 dummy_nmi_handler(). However, there may well be an NMI handler
83 currently executing on some other CPU. We therefore cannot free
84 up any data structures used by the old NMI handler until execution
85 of it completes on all other CPUs.
86
87 One way to accomplish this is via synchronize_rcu(), perhaps as
88 follows::
89
90 unset_nmi_callback();
91 synchronize_rcu();
92 kfree(my_nmi_data);
93
94 This works because (as of v4.20) synchronize_rcu() blocks until all
95 CPUs complete any preemption-disabled segments of code that they were
96 executing.
97 Since NMI handlers disable preemption, synchronize_rcu() is guaranteed
98 not to return until all ongoing NMI handlers exit. It is therefore safe
99 to free up the handler's data as soon as synchronize_rcu() returns.
100
101 Important note: for this to work, the architecture in question must
102 invoke nmi_enter() and nmi_exit() on NMI entry and exit, respectively.
103
104 .. _answer_quick_quiz_NMI:
105
106 Answer to Quick Quiz:
107 Why might the rcu_dereference_sched() be necessary on Alpha, given that the code referenced by the pointer is read-only?
108
109 The caller to set_nmi_callback() might well have
110 initialized some data that is to be used by the new NMI
111 handler. In this case, the rcu_dereference_sched() would
112 be needed, because otherwise a CPU that received an NMI
113 just after the new handler was set might see the pointer
114 to the new NMI handler, but the old pre-initialized
115 version of the handler's data.
116
117 This same sad story can happen on other CPUs when using
118 a compiler with aggressive pointer-value speculation
119 optimizations. (But please don't!)
120
121 More important, the rcu_dereference_sched() makes it
122 clear to someone reading the code that the pointer is
123 being protected by RCU-sched.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.