• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2008 Mathieu Desnoyers
3  *
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.
8  *
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.
13  *
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.
17  */
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>
29 
30 extern struct tracepoint * const __start___tracepoints_ptrs[];
31 extern struct tracepoint * const __stop___tracepoints_ptrs[];
32 
33 /* Set to 1 to enable tracepoint debug output */
34 static const int tracepoint_debug;
35 
36 /*
37  * Tracepoints mutex protects the builtin and module tracepoints and the hash
38  * table, as well as the local module list.
39  */
40 static DEFINE_MUTEX(tracepoints_mutex);
41 
42 #ifdef CONFIG_MODULES
43 /* Local list of struct module */
44 static LIST_HEAD(tracepoint_module_list);
45 #endif /* CONFIG_MODULES */
46 
47 /*
48  * Tracepoint hash table, containing the active tracepoints.
49  * Protected by tracepoints_mutex.
50  */
51 #define TRACEPOINT_HASH_BITS 6
52 #define TRACEPOINT_TABLE_SIZE (1 << TRACEPOINT_HASH_BITS)
53 static struct hlist_head tracepoint_table[TRACEPOINT_TABLE_SIZE];
54 
55 /*
56  * Note about RCU :
57  * It is used to delay the free of multiple probes array until a quiescent
58  * state is reached.
59  * Tracepoint entries modifications are protected by the tracepoints_mutex.
60  */
61 struct tracepoint_entry {
62 	struct hlist_node hlist;
63 	struct tracepoint_func *funcs;
64 	int refcount;	/* Number of times armed. 0 if disarmed. */
65 	char name[0];
66 };
67 
68 struct tp_probes {
69 	union {
70 		struct rcu_head rcu;
71 		struct list_head list;
72 	} u;
73 	struct tracepoint_func probes[0];
74 };
75 
allocate_probes(int count)76 static inline void *allocate_probes(int count)
77 {
78 	struct tp_probes *p  = kmalloc(count * sizeof(struct tracepoint_func)
79 			+ sizeof(struct tp_probes), GFP_KERNEL);
80 	return p == NULL ? NULL : p->probes;
81 }
82 
rcu_free_old_probes(struct rcu_head * head)83 static void rcu_free_old_probes(struct rcu_head *head)
84 {
85 	kfree(container_of(head, struct tp_probes, u.rcu));
86 }
87 
release_probes(struct tracepoint_func * old)88 static inline void release_probes(struct tracepoint_func *old)
89 {
90 	if (old) {
91 		struct tp_probes *tp_probes = container_of(old,
92 			struct tp_probes, probes[0]);
93 		call_rcu_sched(&tp_probes->u.rcu, rcu_free_old_probes);
94 	}
95 }
96 
debug_print_probes(struct tracepoint_entry * entry)97 static void debug_print_probes(struct tracepoint_entry *entry)
98 {
99 	int i;
100 
101 	if (!tracepoint_debug || !entry->funcs)
102 		return;
103 
104 	for (i = 0; entry->funcs[i].func; i++)
105 		printk(KERN_DEBUG "Probe %d : %p\n", i, entry->funcs[i].func);
106 }
107 
108 static struct tracepoint_func *
tracepoint_entry_add_probe(struct tracepoint_entry * entry,void * probe,void * data)109 tracepoint_entry_add_probe(struct tracepoint_entry *entry,
110 			   void *probe, void *data)
111 {
112 	int nr_probes = 0;
113 	struct tracepoint_func *old, *new;
114 
115 	if (WARN_ON(!probe))
116 		return ERR_PTR(-EINVAL);
117 
118 	debug_print_probes(entry);
119 	old = entry->funcs;
120 	if (old) {
121 		/* (N -> N+1), (N != 0, 1) probes */
122 		for (nr_probes = 0; old[nr_probes].func; nr_probes++)
123 			if (old[nr_probes].func == probe &&
124 			    old[nr_probes].data == data)
125 				return ERR_PTR(-EEXIST);
126 	}
127 	/* + 2 : one for new probe, one for NULL func */
128 	new = allocate_probes(nr_probes + 2);
129 	if (new == NULL)
130 		return ERR_PTR(-ENOMEM);
131 	if (old)
132 		memcpy(new, old, nr_probes * sizeof(struct tracepoint_func));
133 	new[nr_probes].func = probe;
134 	new[nr_probes].data = data;
135 	new[nr_probes + 1].func = NULL;
136 	entry->refcount = nr_probes + 1;
137 	entry->funcs = new;
138 	debug_print_probes(entry);
139 	return old;
140 }
141 
142 static void *
tracepoint_entry_remove_probe(struct tracepoint_entry * entry,void * probe,void * data)143 tracepoint_entry_remove_probe(struct tracepoint_entry *entry,
144 			      void *probe, void *data)
145 {
146 	int nr_probes = 0, nr_del = 0, i;
147 	struct tracepoint_func *old, *new;
148 
149 	old = entry->funcs;
150 
151 	if (!old)
152 		return ERR_PTR(-ENOENT);
153 
154 	debug_print_probes(entry);
155 	/* (N -> M), (N > 1, M >= 0) probes */
156 	if (probe) {
157 		for (nr_probes = 0; old[nr_probes].func; nr_probes++) {
158 			if (old[nr_probes].func == probe &&
159 			     old[nr_probes].data == data)
160 				nr_del++;
161 		}
162 	}
163 
164 	/*
165 	 * If probe is NULL, then nr_probes = nr_del = 0, and then the
166 	 * entire entry will be removed.
167 	 */
168 	if (nr_probes - nr_del == 0) {
169 		/* N -> 0, (N > 1) */
170 		entry->funcs = NULL;
171 		entry->refcount = 0;
172 		debug_print_probes(entry);
173 		return old;
174 	} else {
175 		int j = 0;
176 		/* N -> M, (N > 1, M > 0) */
177 		/* + 1 for NULL */
178 		new = allocate_probes(nr_probes - nr_del + 1);
179 		if (new == NULL)
180 			return ERR_PTR(-ENOMEM);
181 		for (i = 0; old[i].func; i++)
182 			if (old[i].func != probe || old[i].data != data)
183 				new[j++] = old[i];
184 		new[nr_probes - nr_del].func = NULL;
185 		entry->refcount = nr_probes - nr_del;
186 		entry->funcs = new;
187 	}
188 	debug_print_probes(entry);
189 	return old;
190 }
191 
192 /*
193  * Get tracepoint if the tracepoint is present in the tracepoint hash table.
194  * Must be called with tracepoints_mutex held.
195  * Returns NULL if not present.
196  */
get_tracepoint(const char * name)197 static struct tracepoint_entry *get_tracepoint(const char *name)
198 {
199 	struct hlist_head *head;
200 	struct tracepoint_entry *e;
201 	u32 hash = jhash(name, strlen(name), 0);
202 
203 	head = &tracepoint_table[hash & (TRACEPOINT_TABLE_SIZE - 1)];
204 	hlist_for_each_entry(e, head, hlist) {
205 		if (!strcmp(name, e->name))
206 			return e;
207 	}
208 	return NULL;
209 }
210 
211 /*
212  * Add the tracepoint to the tracepoint hash table. Must be called with
213  * tracepoints_mutex held.
214  */
add_tracepoint(const char * name)215 static struct tracepoint_entry *add_tracepoint(const char *name)
216 {
217 	struct hlist_head *head;
218 	struct tracepoint_entry *e;
219 	size_t name_len = strlen(name) + 1;
220 	u32 hash = jhash(name, name_len-1, 0);
221 
222 	head = &tracepoint_table[hash & (TRACEPOINT_TABLE_SIZE - 1)];
223 	hlist_for_each_entry(e, head, hlist) {
224 		if (!strcmp(name, e->name)) {
225 			printk(KERN_NOTICE
226 				"tracepoint %s busy\n", name);
227 			return ERR_PTR(-EEXIST);	/* Already there */
228 		}
229 	}
230 	/*
231 	 * Using kmalloc here to allocate a variable length element. Could
232 	 * cause some memory fragmentation if overused.
233 	 */
234 	e = kmalloc(sizeof(struct tracepoint_entry) + name_len, GFP_KERNEL);
235 	if (!e)
236 		return ERR_PTR(-ENOMEM);
237 	memcpy(&e->name[0], name, name_len);
238 	e->funcs = NULL;
239 	e->refcount = 0;
240 	hlist_add_head(&e->hlist, head);
241 	return e;
242 }
243 
244 /*
245  * Remove the tracepoint from the tracepoint hash table. Must be called with
246  * mutex_lock held.
247  */
remove_tracepoint(struct tracepoint_entry * e)248 static inline void remove_tracepoint(struct tracepoint_entry *e)
249 {
250 	hlist_del(&e->hlist);
251 	kfree(e);
252 }
253 
254 /*
255  * Sets the probe callback corresponding to one tracepoint.
256  */
set_tracepoint(struct tracepoint_entry ** entry,struct tracepoint * elem,int active)257 static void set_tracepoint(struct tracepoint_entry **entry,
258 	struct tracepoint *elem, int active)
259 {
260 	WARN_ON(strcmp((*entry)->name, elem->name) != 0);
261 
262 	if (elem->regfunc && !static_key_enabled(&elem->key) && active)
263 		elem->regfunc();
264 	else if (elem->unregfunc && static_key_enabled(&elem->key) && !active)
265 		elem->unregfunc();
266 
267 	/*
268 	 * rcu_assign_pointer has a smp_wmb() which makes sure that the new
269 	 * probe callbacks array is consistent before setting a pointer to it.
270 	 * This array is referenced by __DO_TRACE from
271 	 * include/linux/tracepoints.h. A matching smp_read_barrier_depends()
272 	 * is used.
273 	 */
274 	rcu_assign_pointer(elem->funcs, (*entry)->funcs);
275 	if (active && !static_key_enabled(&elem->key))
276 		static_key_slow_inc(&elem->key);
277 	else if (!active && static_key_enabled(&elem->key))
278 		static_key_slow_dec(&elem->key);
279 }
280 
281 /*
282  * Disable a tracepoint and its probe callback.
283  * Note: only waiting an RCU period after setting elem->call to the empty
284  * function insures that the original callback is not used anymore. This insured
285  * by preempt_disable around the call site.
286  */
disable_tracepoint(struct tracepoint * elem)287 static void disable_tracepoint(struct tracepoint *elem)
288 {
289 	if (elem->unregfunc && static_key_enabled(&elem->key))
290 		elem->unregfunc();
291 
292 	if (static_key_enabled(&elem->key))
293 		static_key_slow_dec(&elem->key);
294 	rcu_assign_pointer(elem->funcs, NULL);
295 }
296 
297 /**
298  * tracepoint_update_probe_range - Update a probe range
299  * @begin: beginning of the range
300  * @end: end of the range
301  *
302  * Updates the probe callback corresponding to a range of tracepoints.
303  * Called with tracepoints_mutex held.
304  */
tracepoint_update_probe_range(struct tracepoint * const * begin,struct tracepoint * const * end)305 static void tracepoint_update_probe_range(struct tracepoint * const *begin,
306 					  struct tracepoint * const *end)
307 {
308 	struct tracepoint * const *iter;
309 	struct tracepoint_entry *mark_entry;
310 
311 	if (!begin)
312 		return;
313 
314 	for (iter = begin; iter < end; iter++) {
315 		mark_entry = get_tracepoint((*iter)->name);
316 		if (mark_entry) {
317 			set_tracepoint(&mark_entry, *iter,
318 					!!mark_entry->refcount);
319 		} else {
320 			disable_tracepoint(*iter);
321 		}
322 	}
323 }
324 
325 #ifdef CONFIG_MODULES
module_update_tracepoints(void)326 void module_update_tracepoints(void)
327 {
328 	struct tp_module *tp_mod;
329 
330 	list_for_each_entry(tp_mod, &tracepoint_module_list, list)
331 		tracepoint_update_probe_range(tp_mod->tracepoints_ptrs,
332 			tp_mod->tracepoints_ptrs + tp_mod->num_tracepoints);
333 }
334 #else /* CONFIG_MODULES */
module_update_tracepoints(void)335 void module_update_tracepoints(void)
336 {
337 }
338 #endif /* CONFIG_MODULES */
339 
340 
341 /*
342  * Update probes, removing the faulty probes.
343  * Called with tracepoints_mutex held.
344  */
tracepoint_update_probes(void)345 static void tracepoint_update_probes(void)
346 {
347 	/* Core kernel tracepoints */
348 	tracepoint_update_probe_range(__start___tracepoints_ptrs,
349 		__stop___tracepoints_ptrs);
350 	/* tracepoints in modules. */
351 	module_update_tracepoints();
352 }
353 
354 static struct tracepoint_func *
tracepoint_add_probe(const char * name,void * probe,void * data)355 tracepoint_add_probe(const char *name, void *probe, void *data)
356 {
357 	struct tracepoint_entry *entry;
358 	struct tracepoint_func *old;
359 
360 	entry = get_tracepoint(name);
361 	if (!entry) {
362 		entry = add_tracepoint(name);
363 		if (IS_ERR(entry))
364 			return (struct tracepoint_func *)entry;
365 	}
366 	old = tracepoint_entry_add_probe(entry, probe, data);
367 	if (IS_ERR(old) && !entry->refcount)
368 		remove_tracepoint(entry);
369 	return old;
370 }
371 
372 /**
373  * tracepoint_probe_register -  Connect a probe to a tracepoint
374  * @name: tracepoint name
375  * @probe: probe handler
376  *
377  * Returns 0 if ok, error value on error.
378  * The probe address must at least be aligned on the architecture pointer size.
379  */
tracepoint_probe_register(const char * name,void * probe,void * data)380 int tracepoint_probe_register(const char *name, void *probe, void *data)
381 {
382 	struct tracepoint_func *old;
383 
384 	mutex_lock(&tracepoints_mutex);
385 	old = tracepoint_add_probe(name, probe, data);
386 	if (IS_ERR(old)) {
387 		mutex_unlock(&tracepoints_mutex);
388 		return PTR_ERR(old);
389 	}
390 	tracepoint_update_probes();		/* may update entry */
391 	mutex_unlock(&tracepoints_mutex);
392 	release_probes(old);
393 	return 0;
394 }
395 EXPORT_SYMBOL_GPL(tracepoint_probe_register);
396 
397 static struct tracepoint_func *
tracepoint_remove_probe(const char * name,void * probe,void * data)398 tracepoint_remove_probe(const char *name, void *probe, void *data)
399 {
400 	struct tracepoint_entry *entry;
401 	struct tracepoint_func *old;
402 
403 	entry = get_tracepoint(name);
404 	if (!entry)
405 		return ERR_PTR(-ENOENT);
406 	old = tracepoint_entry_remove_probe(entry, probe, data);
407 	if (IS_ERR(old))
408 		return old;
409 	if (!entry->refcount)
410 		remove_tracepoint(entry);
411 	return old;
412 }
413 
414 /**
415  * tracepoint_probe_unregister -  Disconnect a probe from a tracepoint
416  * @name: tracepoint name
417  * @probe: probe function pointer
418  *
419  * We do not need to call a synchronize_sched to make sure the probes have
420  * finished running before doing a module unload, because the module unload
421  * itself uses stop_machine(), which insures that every preempt disabled section
422  * have finished.
423  */
tracepoint_probe_unregister(const char * name,void * probe,void * data)424 int tracepoint_probe_unregister(const char *name, void *probe, void *data)
425 {
426 	struct tracepoint_func *old;
427 
428 	mutex_lock(&tracepoints_mutex);
429 	old = tracepoint_remove_probe(name, probe, data);
430 	if (IS_ERR(old)) {
431 		mutex_unlock(&tracepoints_mutex);
432 		return PTR_ERR(old);
433 	}
434 	tracepoint_update_probes();		/* may update entry */
435 	mutex_unlock(&tracepoints_mutex);
436 	release_probes(old);
437 	return 0;
438 }
439 EXPORT_SYMBOL_GPL(tracepoint_probe_unregister);
440 
441 static LIST_HEAD(old_probes);
442 static int need_update;
443 
tracepoint_add_old_probes(void * old)444 static void tracepoint_add_old_probes(void *old)
445 {
446 	need_update = 1;
447 	if (old) {
448 		struct tp_probes *tp_probes = container_of(old,
449 			struct tp_probes, probes[0]);
450 		list_add(&tp_probes->u.list, &old_probes);
451 	}
452 }
453 
454 /**
455  * tracepoint_probe_register_noupdate -  register a probe but not connect
456  * @name: tracepoint name
457  * @probe: probe handler
458  *
459  * caller must call tracepoint_probe_update_all()
460  */
tracepoint_probe_register_noupdate(const char * name,void * probe,void * data)461 int tracepoint_probe_register_noupdate(const char *name, void *probe,
462 				       void *data)
463 {
464 	struct tracepoint_func *old;
465 
466 	mutex_lock(&tracepoints_mutex);
467 	old = tracepoint_add_probe(name, probe, data);
468 	if (IS_ERR(old)) {
469 		mutex_unlock(&tracepoints_mutex);
470 		return PTR_ERR(old);
471 	}
472 	tracepoint_add_old_probes(old);
473 	mutex_unlock(&tracepoints_mutex);
474 	return 0;
475 }
476 EXPORT_SYMBOL_GPL(tracepoint_probe_register_noupdate);
477 
478 /**
479  * tracepoint_probe_unregister_noupdate -  remove a probe but not disconnect
480  * @name: tracepoint name
481  * @probe: probe function pointer
482  *
483  * caller must call tracepoint_probe_update_all()
484  */
tracepoint_probe_unregister_noupdate(const char * name,void * probe,void * data)485 int tracepoint_probe_unregister_noupdate(const char *name, void *probe,
486 					 void *data)
487 {
488 	struct tracepoint_func *old;
489 
490 	mutex_lock(&tracepoints_mutex);
491 	old = tracepoint_remove_probe(name, probe, data);
492 	if (IS_ERR(old)) {
493 		mutex_unlock(&tracepoints_mutex);
494 		return PTR_ERR(old);
495 	}
496 	tracepoint_add_old_probes(old);
497 	mutex_unlock(&tracepoints_mutex);
498 	return 0;
499 }
500 EXPORT_SYMBOL_GPL(tracepoint_probe_unregister_noupdate);
501 
502 /**
503  * tracepoint_probe_update_all -  update tracepoints
504  */
tracepoint_probe_update_all(void)505 void tracepoint_probe_update_all(void)
506 {
507 	LIST_HEAD(release_probes);
508 	struct tp_probes *pos, *next;
509 
510 	mutex_lock(&tracepoints_mutex);
511 	if (!need_update) {
512 		mutex_unlock(&tracepoints_mutex);
513 		return;
514 	}
515 	if (!list_empty(&old_probes))
516 		list_replace_init(&old_probes, &release_probes);
517 	need_update = 0;
518 	tracepoint_update_probes();
519 	mutex_unlock(&tracepoints_mutex);
520 	list_for_each_entry_safe(pos, next, &release_probes, u.list) {
521 		list_del(&pos->u.list);
522 		call_rcu_sched(&pos->u.rcu, rcu_free_old_probes);
523 	}
524 }
525 EXPORT_SYMBOL_GPL(tracepoint_probe_update_all);
526 
527 /**
528  * tracepoint_get_iter_range - Get a next tracepoint iterator given a range.
529  * @tracepoint: current tracepoints (in), next tracepoint (out)
530  * @begin: beginning of the range
531  * @end: end of the range
532  *
533  * Returns whether a next tracepoint has been found (1) or not (0).
534  * Will return the first tracepoint in the range if the input tracepoint is
535  * NULL.
536  */
tracepoint_get_iter_range(struct tracepoint * const ** tracepoint,struct tracepoint * const * begin,struct tracepoint * const * end)537 static int tracepoint_get_iter_range(struct tracepoint * const **tracepoint,
538 	struct tracepoint * const *begin, struct tracepoint * const *end)
539 {
540 	if (!*tracepoint && begin != end) {
541 		*tracepoint = begin;
542 		return 1;
543 	}
544 	if (*tracepoint >= begin && *tracepoint < end)
545 		return 1;
546 	return 0;
547 }
548 
549 #ifdef CONFIG_MODULES
tracepoint_get_iter(struct tracepoint_iter * iter)550 static void tracepoint_get_iter(struct tracepoint_iter *iter)
551 {
552 	int found = 0;
553 	struct tp_module *iter_mod;
554 
555 	/* Core kernel tracepoints */
556 	if (!iter->module) {
557 		found = tracepoint_get_iter_range(&iter->tracepoint,
558 				__start___tracepoints_ptrs,
559 				__stop___tracepoints_ptrs);
560 		if (found)
561 			goto end;
562 	}
563 	/* Tracepoints in modules */
564 	mutex_lock(&tracepoints_mutex);
565 	list_for_each_entry(iter_mod, &tracepoint_module_list, list) {
566 		/*
567 		 * Sorted module list
568 		 */
569 		if (iter_mod < iter->module)
570 			continue;
571 		else if (iter_mod > iter->module)
572 			iter->tracepoint = NULL;
573 		found = tracepoint_get_iter_range(&iter->tracepoint,
574 			iter_mod->tracepoints_ptrs,
575 			iter_mod->tracepoints_ptrs
576 				+ iter_mod->num_tracepoints);
577 		if (found) {
578 			iter->module = iter_mod;
579 			break;
580 		}
581 	}
582 	mutex_unlock(&tracepoints_mutex);
583 end:
584 	if (!found)
585 		tracepoint_iter_reset(iter);
586 }
587 #else /* CONFIG_MODULES */
tracepoint_get_iter(struct tracepoint_iter * iter)588 static void tracepoint_get_iter(struct tracepoint_iter *iter)
589 {
590 	int found = 0;
591 
592 	/* Core kernel tracepoints */
593 	found = tracepoint_get_iter_range(&iter->tracepoint,
594 			__start___tracepoints_ptrs,
595 			__stop___tracepoints_ptrs);
596 	if (!found)
597 		tracepoint_iter_reset(iter);
598 }
599 #endif /* CONFIG_MODULES */
600 
tracepoint_iter_start(struct tracepoint_iter * iter)601 void tracepoint_iter_start(struct tracepoint_iter *iter)
602 {
603 	tracepoint_get_iter(iter);
604 }
605 EXPORT_SYMBOL_GPL(tracepoint_iter_start);
606 
tracepoint_iter_next(struct tracepoint_iter * iter)607 void tracepoint_iter_next(struct tracepoint_iter *iter)
608 {
609 	iter->tracepoint++;
610 	/*
611 	 * iter->tracepoint may be invalid because we blindly incremented it.
612 	 * Make sure it is valid by marshalling on the tracepoints, getting the
613 	 * tracepoints from following modules if necessary.
614 	 */
615 	tracepoint_get_iter(iter);
616 }
617 EXPORT_SYMBOL_GPL(tracepoint_iter_next);
618 
tracepoint_iter_stop(struct tracepoint_iter * iter)619 void tracepoint_iter_stop(struct tracepoint_iter *iter)
620 {
621 }
622 EXPORT_SYMBOL_GPL(tracepoint_iter_stop);
623 
tracepoint_iter_reset(struct tracepoint_iter * iter)624 void tracepoint_iter_reset(struct tracepoint_iter *iter)
625 {
626 #ifdef CONFIG_MODULES
627 	iter->module = NULL;
628 #endif /* CONFIG_MODULES */
629 	iter->tracepoint = NULL;
630 }
631 EXPORT_SYMBOL_GPL(tracepoint_iter_reset);
632 
633 #ifdef CONFIG_MODULES
tracepoint_module_coming(struct module * mod)634 static int tracepoint_module_coming(struct module *mod)
635 {
636 	struct tp_module *tp_mod, *iter;
637 	int ret = 0;
638 
639 	/*
640 	 * We skip modules that taint the kernel, especially those with different
641 	 * module headers (for forced load), to make sure we don't cause a crash.
642 	 * Staging and out-of-tree GPL modules are fine.
643 	 */
644 	if (mod->taints & ~((1 << TAINT_OOT_MODULE) | (1 << TAINT_CRAP)))
645 		return 0;
646 	mutex_lock(&tracepoints_mutex);
647 	tp_mod = kmalloc(sizeof(struct tp_module), GFP_KERNEL);
648 	if (!tp_mod) {
649 		ret = -ENOMEM;
650 		goto end;
651 	}
652 	tp_mod->num_tracepoints = mod->num_tracepoints;
653 	tp_mod->tracepoints_ptrs = mod->tracepoints_ptrs;
654 
655 	/*
656 	 * tracepoint_module_list is kept sorted by struct module pointer
657 	 * address for iteration on tracepoints from a seq_file that can release
658 	 * the mutex between calls.
659 	 */
660 	list_for_each_entry_reverse(iter, &tracepoint_module_list, list) {
661 		BUG_ON(iter == tp_mod);	/* Should never be in the list twice */
662 		if (iter < tp_mod) {
663 			/* We belong to the location right after iter. */
664 			list_add(&tp_mod->list, &iter->list);
665 			goto module_added;
666 		}
667 	}
668 	/* We belong to the beginning of the list */
669 	list_add(&tp_mod->list, &tracepoint_module_list);
670 module_added:
671 	tracepoint_update_probe_range(mod->tracepoints_ptrs,
672 		mod->tracepoints_ptrs + mod->num_tracepoints);
673 end:
674 	mutex_unlock(&tracepoints_mutex);
675 	return ret;
676 }
677 
tracepoint_module_going(struct module * mod)678 static int tracepoint_module_going(struct module *mod)
679 {
680 	struct tp_module *pos;
681 
682 	mutex_lock(&tracepoints_mutex);
683 	tracepoint_update_probe_range(mod->tracepoints_ptrs,
684 		mod->tracepoints_ptrs + mod->num_tracepoints);
685 	list_for_each_entry(pos, &tracepoint_module_list, list) {
686 		if (pos->tracepoints_ptrs == mod->tracepoints_ptrs) {
687 			list_del(&pos->list);
688 			kfree(pos);
689 			break;
690 		}
691 	}
692 	/*
693 	 * In the case of modules that were tainted at "coming", we'll simply
694 	 * walk through the list without finding it. We cannot use the "tainted"
695 	 * flag on "going", in case a module taints the kernel only after being
696 	 * loaded.
697 	 */
698 	mutex_unlock(&tracepoints_mutex);
699 	return 0;
700 }
701 
tracepoint_module_notify(struct notifier_block * self,unsigned long val,void * data)702 int tracepoint_module_notify(struct notifier_block *self,
703 			     unsigned long val, void *data)
704 {
705 	struct module *mod = data;
706 	int ret = 0;
707 
708 	switch (val) {
709 	case MODULE_STATE_COMING:
710 		ret = tracepoint_module_coming(mod);
711 		break;
712 	case MODULE_STATE_LIVE:
713 		break;
714 	case MODULE_STATE_GOING:
715 		ret = tracepoint_module_going(mod);
716 		break;
717 	}
718 	return ret;
719 }
720 
721 struct notifier_block tracepoint_module_nb = {
722 	.notifier_call = tracepoint_module_notify,
723 	.priority = 0,
724 };
725 
init_tracepoints(void)726 static int init_tracepoints(void)
727 {
728 	return register_module_notifier(&tracepoint_module_nb);
729 }
730 __initcall(init_tracepoints);
731 #endif /* CONFIG_MODULES */
732 
733 #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
734 
735 /* NB: reg/unreg are called while guarded with the tracepoints_mutex */
736 static int sys_tracepoint_refcount;
737 
syscall_regfunc(void)738 void syscall_regfunc(void)
739 {
740 	unsigned long flags;
741 	struct task_struct *g, *t;
742 
743 	if (!sys_tracepoint_refcount) {
744 		read_lock_irqsave(&tasklist_lock, flags);
745 		do_each_thread(g, t) {
746 			/* Skip kernel threads. */
747 			if (t->mm)
748 				set_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
749 		} while_each_thread(g, t);
750 		read_unlock_irqrestore(&tasklist_lock, flags);
751 	}
752 	sys_tracepoint_refcount++;
753 }
754 
syscall_unregfunc(void)755 void syscall_unregfunc(void)
756 {
757 	unsigned long flags;
758 	struct task_struct *g, *t;
759 
760 	sys_tracepoint_refcount--;
761 	if (!sys_tracepoint_refcount) {
762 		read_lock_irqsave(&tasklist_lock, flags);
763 		do_each_thread(g, t) {
764 			clear_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
765 		} while_each_thread(g, t);
766 		read_unlock_irqrestore(&tasklist_lock, flags);
767 	}
768 }
769 #endif
770