Commit | Line | Data |
---|---|---|
4eacdf18 FW |
1 | /* |
2 | * Context tracking: Probe on high level context boundaries such as kernel | |
3 | * and userspace. This includes syscalls and exceptions entry/exit. | |
4 | * | |
5 | * This is used by RCU to remove its dependency on the timer tick while a CPU | |
6 | * runs in userspace. | |
7 | * | |
8 | * Started by Frederic Weisbecker: | |
9 | * | |
10 | * Copyright (C) 2012 Red Hat, Inc., Frederic Weisbecker <fweisbec@redhat.com> | |
11 | * | |
12 | * Many thanks to Gilad Ben-Yossef, Paul McKenney, Ingo Molnar, Andrew Morton, | |
13 | * Steven Rostedt, Peter Zijlstra for suggestions and improvements. | |
14 | * | |
15 | */ | |
16 | ||
91d1aa43 FW |
17 | #include <linux/context_tracking.h> |
18 | #include <linux/rcupdate.h> | |
19 | #include <linux/sched.h> | |
91d1aa43 | 20 | #include <linux/hardirq.h> |
6a61671b | 21 | #include <linux/export.h> |
4cdf77a8 | 22 | #include <linux/kprobes.h> |
91d1aa43 | 23 | |
1b6a259a FW |
24 | #define CREATE_TRACE_POINTS |
25 | #include <trace/events/context_tracking.h> | |
26 | ||
65f382fd | 27 | struct static_key context_tracking_enabled = STATIC_KEY_INIT_FALSE; |
48d6a816 | 28 | EXPORT_SYMBOL_GPL(context_tracking_enabled); |
65f382fd FW |
29 | |
30 | DEFINE_PER_CPU(struct context_tracking, context_tracking); | |
48d6a816 | 31 | EXPORT_SYMBOL_GPL(context_tracking); |
91d1aa43 | 32 | |
2e709338 FW |
33 | void context_tracking_cpu_set(int cpu) |
34 | { | |
65f382fd FW |
35 | if (!per_cpu(context_tracking.active, cpu)) { |
36 | per_cpu(context_tracking.active, cpu) = true; | |
37 | static_key_slow_inc(&context_tracking_enabled); | |
38 | } | |
2e709338 FW |
39 | } |
40 | ||
4eacdf18 | 41 | /** |
3aab4f50 RR |
42 | * context_tracking_enter - Inform the context tracking that the CPU is going |
43 | * enter user or guest space mode. | |
4eacdf18 FW |
44 | * |
45 | * This function must be called right before we switch from the kernel | |
3aab4f50 RR |
46 | * to user or guest space, when it's guaranteed the remaining kernel |
47 | * instructions to execute won't use any RCU read side critical section | |
48 | * because this function sets RCU in extended quiescent state. | |
4eacdf18 | 49 | */ |
3aab4f50 | 50 | void context_tracking_enter(enum ctx_state state) |
91d1aa43 FW |
51 | { |
52 | unsigned long flags; | |
53 | ||
0c06a5d4 FW |
54 | /* |
55 | * Repeat the user_enter() check here because some archs may be calling | |
56 | * this from asm and if no CPU needs context tracking, they shouldn't | |
58135f57 FW |
57 | * go further. Repeat the check here until they support the inline static |
58 | * key check. | |
0c06a5d4 | 59 | */ |
58135f57 | 60 | if (!context_tracking_is_enabled()) |
0c06a5d4 FW |
61 | return; |
62 | ||
91d1aa43 FW |
63 | /* |
64 | * Some contexts may involve an exception occuring in an irq, | |
65 | * leading to that nesting: | |
66 | * rcu_irq_enter() rcu_user_exit() rcu_user_exit() rcu_irq_exit() | |
67 | * This would mess up the dyntick_nesting count though. And rcu_irq_*() | |
68 | * helpers are enough to protect RCU uses inside the exception. So | |
69 | * just return immediately if we detect we are in an IRQ. | |
70 | */ | |
71 | if (in_interrupt()) | |
72 | return; | |
73 | ||
4eacdf18 | 74 | /* Kernel threads aren't supposed to go to userspace */ |
91d1aa43 FW |
75 | WARN_ON_ONCE(!current->mm); |
76 | ||
77 | local_irq_save(flags); | |
3aab4f50 | 78 | if ( __this_cpu_read(context_tracking.state) != state) { |
d65ec121 | 79 | if (__this_cpu_read(context_tracking.active)) { |
1b6a259a | 80 | trace_user_enter(0); |
d65ec121 FW |
81 | /* |
82 | * At this stage, only low level arch entry code remains and | |
83 | * then we'll run in userspace. We can assume there won't be | |
84 | * any RCU read-side critical section until the next call to | |
85 | * user_exit() or rcu_irq_enter(). Let's remove RCU's dependency | |
86 | * on the tick. | |
87 | */ | |
88 | vtime_user_enter(current); | |
89 | rcu_user_enter(); | |
90 | } | |
4eacdf18 | 91 | /* |
d65ec121 FW |
92 | * Even if context tracking is disabled on this CPU, because it's outside |
93 | * the full dynticks mask for example, we still have to keep track of the | |
94 | * context transitions and states to prevent inconsistency on those of | |
95 | * other CPUs. | |
96 | * If a task triggers an exception in userspace, sleep on the exception | |
97 | * handler and then migrate to another CPU, that new CPU must know where | |
98 | * the exception returns by the time we call exception_exit(). | |
99 | * This information can only be provided by the previous CPU when it called | |
100 | * exception_enter(). | |
101 | * OTOH we can spare the calls to vtime and RCU when context_tracking.active | |
102 | * is false because we know that CPU is not tickless. | |
4eacdf18 | 103 | */ |
3aab4f50 | 104 | __this_cpu_write(context_tracking.state, state); |
91d1aa43 FW |
105 | } |
106 | local_irq_restore(flags); | |
107 | } | |
3aab4f50 RR |
108 | NOKPROBE_SYMBOL(context_tracking_enter); |
109 | ||
110 | void context_tracking_user_enter(void) | |
111 | { | |
112 | context_tracking_enter(CONTEXT_USER); | |
113 | } | |
4cdf77a8 | 114 | NOKPROBE_SYMBOL(context_tracking_user_enter); |
91d1aa43 | 115 | |
4eacdf18 | 116 | /** |
3aab4f50 RR |
117 | * context_tracking_exit - Inform the context tracking that the CPU is |
118 | * exiting user or guest mode and entering the kernel. | |
4eacdf18 | 119 | * |
3aab4f50 RR |
120 | * This function must be called after we entered the kernel from user or |
121 | * guest space before any use of RCU read side critical section. This | |
122 | * potentially include any high level kernel code like syscalls, exceptions, | |
123 | * signal handling, etc... | |
4eacdf18 FW |
124 | * |
125 | * This call supports re-entrancy. This way it can be called from any exception | |
126 | * handler without needing to know if we came from userspace or not. | |
127 | */ | |
3aab4f50 | 128 | void context_tracking_exit(enum ctx_state state) |
91d1aa43 FW |
129 | { |
130 | unsigned long flags; | |
131 | ||
58135f57 | 132 | if (!context_tracking_is_enabled()) |
0c06a5d4 FW |
133 | return; |
134 | ||
91d1aa43 FW |
135 | if (in_interrupt()) |
136 | return; | |
137 | ||
138 | local_irq_save(flags); | |
3aab4f50 | 139 | if (__this_cpu_read(context_tracking.state) == state) { |
d65ec121 FW |
140 | if (__this_cpu_read(context_tracking.active)) { |
141 | /* | |
142 | * We are going to run code that may use RCU. Inform | |
143 | * RCU core about that (ie: we may need the tick again). | |
144 | */ | |
145 | rcu_user_exit(); | |
146 | vtime_user_exit(current); | |
1b6a259a | 147 | trace_user_exit(0); |
d65ec121 | 148 | } |
c467ea76 | 149 | __this_cpu_write(context_tracking.state, CONTEXT_KERNEL); |
91d1aa43 FW |
150 | } |
151 | local_irq_restore(flags); | |
152 | } | |
3aab4f50 RR |
153 | NOKPROBE_SYMBOL(context_tracking_exit); |
154 | ||
155 | void context_tracking_user_exit(void) | |
156 | { | |
157 | context_tracking_exit(CONTEXT_USER); | |
158 | } | |
4cdf77a8 | 159 | NOKPROBE_SYMBOL(context_tracking_user_exit); |
91d1aa43 | 160 | |
4eacdf18 | 161 | /** |
73d424f9 | 162 | * __context_tracking_task_switch - context switch the syscall callbacks |
4eacdf18 FW |
163 | * @prev: the task that is being switched out |
164 | * @next: the task that is being switched in | |
165 | * | |
166 | * The context tracking uses the syscall slow path to implement its user-kernel | |
167 | * boundaries probes on syscalls. This way it doesn't impact the syscall fast | |
168 | * path on CPUs that don't do context tracking. | |
169 | * | |
170 | * But we need to clear the flag on the previous task because it may later | |
171 | * migrate to some CPU that doesn't do the context tracking. As such the TIF | |
172 | * flag may not be desired there. | |
173 | */ | |
73d424f9 FW |
174 | void __context_tracking_task_switch(struct task_struct *prev, |
175 | struct task_struct *next) | |
91d1aa43 | 176 | { |
d65ec121 FW |
177 | clear_tsk_thread_flag(prev, TIF_NOHZ); |
178 | set_tsk_thread_flag(next, TIF_NOHZ); | |
91d1aa43 | 179 | } |
65f382fd FW |
180 | |
181 | #ifdef CONFIG_CONTEXT_TRACKING_FORCE | |
182 | void __init context_tracking_init(void) | |
183 | { | |
184 | int cpu; | |
185 | ||
186 | for_each_possible_cpu(cpu) | |
187 | context_tracking_cpu_set(cpu); | |
188 | } | |
189 | #endif |