2 * Copyright (C) 2008-2014 Mathieu Desnoyers
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
18 #include <linux/module.h>
19 #include <linux/mutex.h>
20 #include <linux/types.h>
21 #include <linux/jhash.h>
22 #include <linux/list.h>
23 #include <linux/rcupdate.h>
24 #include <linux/tracepoint.h>
25 #include <linux/err.h>
26 #include <linux/slab.h>
27 #include <linux/sched.h>
28 #include <linux/static_key.h>
30 extern struct tracepoint * const __start___tracepoints_ptrs[];
31 extern struct tracepoint * const __stop___tracepoints_ptrs[];
33 /* Set to 1 to enable tracepoint debug output */
34 static const int tracepoint_debug;
38 * Tracepoint module list mutex protects the local module list.
40 static DEFINE_MUTEX(tracepoint_module_list_mutex);
42 /* Local list of struct tp_module */
43 static LIST_HEAD(tracepoint_module_list);
44 #endif /* CONFIG_MODULES */
47 * tracepoints_mutex protects the builtin and module tracepoints.
48 * tracepoints_mutex nests inside tracepoint_module_list_mutex.
50 static DEFINE_MUTEX(tracepoints_mutex);
54 * It is used to delay the free of multiple probes array until a quiescent
59 struct tracepoint_func probes[0];
62 static inline void *allocate_probes(int count)
64 struct tp_probes *p = kmalloc(count * sizeof(struct tracepoint_func)
65 + sizeof(struct tp_probes), GFP_KERNEL);
66 return p == NULL ? NULL : p->probes;
69 static void rcu_free_old_probes(struct rcu_head *head)
71 kfree(container_of(head, struct tp_probes, rcu));
74 static inline void release_probes(struct tracepoint_func *old)
77 struct tp_probes *tp_probes = container_of(old,
78 struct tp_probes, probes[0]);
79 call_rcu_sched(&tp_probes->rcu, rcu_free_old_probes);
83 static void debug_print_probes(struct tracepoint_func *funcs)
87 if (!tracepoint_debug || !funcs)
90 for (i = 0; funcs[i].func; i++)
91 printk(KERN_DEBUG "Probe %d : %p\n", i, funcs[i].func);
94 static struct tracepoint_func *func_add(struct tracepoint_func **funcs,
95 struct tracepoint_func *tp_func)
98 struct tracepoint_func *old, *new;
100 if (WARN_ON(!tp_func->func))
101 return ERR_PTR(-EINVAL);
103 debug_print_probes(*funcs);
106 /* (N -> N+1), (N != 0, 1) probes */
107 for (nr_probes = 0; old[nr_probes].func; nr_probes++)
108 if (old[nr_probes].func == tp_func->func &&
109 old[nr_probes].data == tp_func->data)
110 return ERR_PTR(-EEXIST);
112 /* + 2 : one for new probe, one for NULL func */
113 new = allocate_probes(nr_probes + 2);
115 return ERR_PTR(-ENOMEM);
117 memcpy(new, old, nr_probes * sizeof(struct tracepoint_func));
118 new[nr_probes] = *tp_func;
119 new[nr_probes + 1].func = NULL;
121 debug_print_probes(*funcs);
125 static void *func_remove(struct tracepoint_func **funcs,
126 struct tracepoint_func *tp_func)
128 int nr_probes = 0, nr_del = 0, i;
129 struct tracepoint_func *old, *new;
134 return ERR_PTR(-ENOENT);
136 debug_print_probes(*funcs);
137 /* (N -> M), (N > 1, M >= 0) probes */
139 for (nr_probes = 0; old[nr_probes].func; nr_probes++) {
140 if (old[nr_probes].func == tp_func->func &&
141 old[nr_probes].data == tp_func->data)
147 * If probe is NULL, then nr_probes = nr_del = 0, and then the
148 * entire entry will be removed.
150 if (nr_probes - nr_del == 0) {
151 /* N -> 0, (N > 1) */
153 debug_print_probes(*funcs);
157 /* N -> M, (N > 1, M > 0) */
159 new = allocate_probes(nr_probes - nr_del + 1);
161 return ERR_PTR(-ENOMEM);
162 for (i = 0; old[i].func; i++)
163 if (old[i].func != tp_func->func
164 || old[i].data != tp_func->data)
166 new[nr_probes - nr_del].func = NULL;
169 debug_print_probes(*funcs);
174 * Add the probe function to a tracepoint.
176 static int tracepoint_add_func(struct tracepoint *tp,
177 struct tracepoint_func *func)
179 struct tracepoint_func *old, *tp_funcs;
181 if (tp->regfunc && !static_key_enabled(&tp->key))
184 tp_funcs = rcu_dereference_protected(tp->funcs,
185 lockdep_is_held(&tracepoints_mutex));
186 old = func_add(&tp_funcs, func);
193 * rcu_assign_pointer has a smp_wmb() which makes sure that the new
194 * probe callbacks array is consistent before setting a pointer to it.
195 * This array is referenced by __DO_TRACE from
196 * include/linux/tracepoints.h. A matching smp_read_barrier_depends()
199 rcu_assign_pointer(tp->funcs, tp_funcs);
200 if (!static_key_enabled(&tp->key))
201 static_key_slow_inc(&tp->key);
207 * Remove a probe function from a tracepoint.
208 * Note: only waiting an RCU period after setting elem->call to the empty
209 * function insures that the original callback is not used anymore. This insured
210 * by preempt_disable around the call site.
212 static int tracepoint_remove_func(struct tracepoint *tp,
213 struct tracepoint_func *func)
215 struct tracepoint_func *old, *tp_funcs;
217 tp_funcs = rcu_dereference_protected(tp->funcs,
218 lockdep_is_held(&tracepoints_mutex));
219 old = func_remove(&tp_funcs, func);
226 /* Removed last function */
227 if (tp->unregfunc && static_key_enabled(&tp->key))
230 if (static_key_enabled(&tp->key))
231 static_key_slow_dec(&tp->key);
233 rcu_assign_pointer(tp->funcs, tp_funcs);
239 * tracepoint_probe_register - Connect a probe to a tracepoint
241 * @probe: probe handler
242 * @data: tracepoint data
244 * Returns 0 if ok, error value on error.
245 * Note: if @tp is within a module, the caller is responsible for
246 * unregistering the probe before the module is gone. This can be
247 * performed either with a tracepoint module going notifier, or from
248 * within module exit functions.
250 int tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data)
252 struct tracepoint_func tp_func;
255 mutex_lock(&tracepoints_mutex);
256 tp_func.func = probe;
258 ret = tracepoint_add_func(tp, &tp_func);
259 mutex_unlock(&tracepoints_mutex);
262 EXPORT_SYMBOL_GPL(tracepoint_probe_register);
265 * tracepoint_probe_unregister - Disconnect a probe from a tracepoint
267 * @probe: probe function pointer
268 * @data: tracepoint data
270 * Returns 0 if ok, error value on error.
272 int tracepoint_probe_unregister(struct tracepoint *tp, void *probe, void *data)
274 struct tracepoint_func tp_func;
277 mutex_lock(&tracepoints_mutex);
278 tp_func.func = probe;
280 ret = tracepoint_remove_func(tp, &tp_func);
281 mutex_unlock(&tracepoints_mutex);
284 EXPORT_SYMBOL_GPL(tracepoint_probe_unregister);
286 #ifdef CONFIG_MODULES
287 bool trace_module_has_bad_taint(struct module *mod)
289 return mod->taints & ~((1 << TAINT_OOT_MODULE) | (1 << TAINT_CRAP) |
290 (1 << TAINT_UNSIGNED_MODULE));
293 static BLOCKING_NOTIFIER_HEAD(tracepoint_notify_list);
296 * register_tracepoint_notifier - register tracepoint coming/going notifier
297 * @nb: notifier block
299 * Notifiers registered with this function are called on module
300 * coming/going with the tracepoint_module_list_mutex held.
301 * The notifier block callback should expect a "struct tp_module" data
304 int register_tracepoint_module_notifier(struct notifier_block *nb)
306 struct tp_module *tp_mod;
309 mutex_lock(&tracepoint_module_list_mutex);
310 ret = blocking_notifier_chain_register(&tracepoint_notify_list, nb);
313 list_for_each_entry(tp_mod, &tracepoint_module_list, list)
314 (void) nb->notifier_call(nb, MODULE_STATE_COMING, tp_mod);
316 mutex_unlock(&tracepoint_module_list_mutex);
319 EXPORT_SYMBOL_GPL(register_tracepoint_module_notifier);
322 * unregister_tracepoint_notifier - unregister tracepoint coming/going notifier
323 * @nb: notifier block
325 * The notifier block callback should expect a "struct tp_module" data
328 int unregister_tracepoint_module_notifier(struct notifier_block *nb)
330 struct tp_module *tp_mod;
333 mutex_lock(&tracepoint_module_list_mutex);
334 ret = blocking_notifier_chain_unregister(&tracepoint_notify_list, nb);
337 list_for_each_entry(tp_mod, &tracepoint_module_list, list)
338 (void) nb->notifier_call(nb, MODULE_STATE_GOING, tp_mod);
340 mutex_unlock(&tracepoint_module_list_mutex);
344 EXPORT_SYMBOL_GPL(unregister_tracepoint_module_notifier);
347 * Ensure the tracer unregistered the module's probes before the module
348 * teardown is performed. Prevents leaks of probe and data pointers.
350 static void tp_module_going_check_quiescent(struct tracepoint * const *begin,
351 struct tracepoint * const *end)
353 struct tracepoint * const *iter;
357 for (iter = begin; iter < end; iter++)
358 WARN_ON_ONCE((*iter)->funcs);
361 static int tracepoint_module_coming(struct module *mod)
363 struct tp_module *tp_mod;
366 if (!mod->num_tracepoints)
370 * We skip modules that taint the kernel, especially those with different
371 * module headers (for forced load), to make sure we don't cause a crash.
372 * Staging, out-of-tree, and unsigned GPL modules are fine.
374 if (trace_module_has_bad_taint(mod))
376 mutex_lock(&tracepoint_module_list_mutex);
377 tp_mod = kmalloc(sizeof(struct tp_module), GFP_KERNEL);
383 list_add_tail(&tp_mod->list, &tracepoint_module_list);
384 blocking_notifier_call_chain(&tracepoint_notify_list,
385 MODULE_STATE_COMING, tp_mod);
387 mutex_unlock(&tracepoint_module_list_mutex);
391 static void tracepoint_module_going(struct module *mod)
393 struct tp_module *tp_mod;
395 if (!mod->num_tracepoints)
398 mutex_lock(&tracepoint_module_list_mutex);
399 list_for_each_entry(tp_mod, &tracepoint_module_list, list) {
400 if (tp_mod->mod == mod) {
401 blocking_notifier_call_chain(&tracepoint_notify_list,
402 MODULE_STATE_GOING, tp_mod);
403 list_del(&tp_mod->list);
406 * Called the going notifier before checking for
409 tp_module_going_check_quiescent(mod->tracepoints_ptrs,
410 mod->tracepoints_ptrs + mod->num_tracepoints);
415 * In the case of modules that were tainted at "coming", we'll simply
416 * walk through the list without finding it. We cannot use the "tainted"
417 * flag on "going", in case a module taints the kernel only after being
420 mutex_unlock(&tracepoint_module_list_mutex);
423 static int tracepoint_module_notify(struct notifier_block *self,
424 unsigned long val, void *data)
426 struct module *mod = data;
430 case MODULE_STATE_COMING:
431 ret = tracepoint_module_coming(mod);
433 case MODULE_STATE_LIVE:
435 case MODULE_STATE_GOING:
436 tracepoint_module_going(mod);
438 case MODULE_STATE_UNFORMED:
444 static struct notifier_block tracepoint_module_nb = {
445 .notifier_call = tracepoint_module_notify,
449 static __init int init_tracepoints(void)
453 ret = register_module_notifier(&tracepoint_module_nb);
455 pr_warning("Failed to register tracepoint module enter notifier\n");
459 __initcall(init_tracepoints);
460 #endif /* CONFIG_MODULES */
462 static void for_each_tracepoint_range(struct tracepoint * const *begin,
463 struct tracepoint * const *end,
464 void (*fct)(struct tracepoint *tp, void *priv),
467 struct tracepoint * const *iter;
471 for (iter = begin; iter < end; iter++)
476 * for_each_kernel_tracepoint - iteration on all kernel tracepoints
478 * @priv: private data
480 void for_each_kernel_tracepoint(void (*fct)(struct tracepoint *tp, void *priv),
483 for_each_tracepoint_range(__start___tracepoints_ptrs,
484 __stop___tracepoints_ptrs, fct, priv);
486 EXPORT_SYMBOL_GPL(for_each_kernel_tracepoint);
488 #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
490 /* NB: reg/unreg are called while guarded with the tracepoints_mutex */
491 static int sys_tracepoint_refcount;
493 void syscall_regfunc(void)
495 struct task_struct *p, *t;
497 if (!sys_tracepoint_refcount) {
498 read_lock(&tasklist_lock);
499 for_each_process_thread(p, t) {
500 set_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
502 read_unlock(&tasklist_lock);
504 sys_tracepoint_refcount++;
507 void syscall_unregfunc(void)
509 struct task_struct *p, *t;
511 sys_tracepoint_refcount--;
512 if (!sys_tracepoint_refcount) {
513 read_lock(&tasklist_lock);
514 for_each_process_thread(p, t) {
515 clear_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
517 read_unlock(&tasklist_lock);