~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/kernel/tracepoint.c

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.11 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.84 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.154 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.201 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.201 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.77 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * Copyright (C) 2008-2014 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/signal.h>
 28 #include <linux/sched/task.h>
 29 #include <linux/static_key.h>
 30 
 31 extern tracepoint_ptr_t __start___tracepoints_ptrs[];
 32 extern tracepoint_ptr_t __stop___tracepoints_ptrs[];
 33 
 34 DEFINE_SRCU(tracepoint_srcu);
 35 EXPORT_SYMBOL_GPL(tracepoint_srcu);
 36 
 37 /* Set to 1 to enable tracepoint debug output */
 38 static const int tracepoint_debug;
 39 
 40 #ifdef CONFIG_MODULES
 41 /*
 42  * Tracepoint module list mutex protects the local module list.
 43  */
 44 static DEFINE_MUTEX(tracepoint_module_list_mutex);
 45 
 46 /* Local list of struct tp_module */
 47 static LIST_HEAD(tracepoint_module_list);
 48 #endif /* CONFIG_MODULES */
 49 
 50 /*
 51  * tracepoints_mutex protects the builtin and module tracepoints.
 52  * tracepoints_mutex nests inside tracepoint_module_list_mutex.
 53  */
 54 static DEFINE_MUTEX(tracepoints_mutex);
 55 
 56 static struct rcu_head *early_probes;
 57 static bool ok_to_free_tracepoints;
 58 
 59 /*
 60  * Note about RCU :
 61  * It is used to delay the free of multiple probes array until a quiescent
 62  * state is reached.
 63  */
 64 struct tp_probes {
 65         struct rcu_head rcu;
 66         struct tracepoint_func probes[0];
 67 };
 68 
 69 static inline void *allocate_probes(int count)
 70 {
 71         struct tp_probes *p  = kmalloc(count * sizeof(struct tracepoint_func)
 72                         + sizeof(struct tp_probes), GFP_KERNEL);
 73         return p == NULL ? NULL : p->probes;
 74 }
 75 
 76 static void srcu_free_old_probes(struct rcu_head *head)
 77 {
 78         kfree(container_of(head, struct tp_probes, rcu));
 79 }
 80 
 81 static void rcu_free_old_probes(struct rcu_head *head)
 82 {
 83         call_srcu(&tracepoint_srcu, head, srcu_free_old_probes);
 84 }
 85 
 86 static __init int release_early_probes(void)
 87 {
 88         struct rcu_head *tmp;
 89 
 90         ok_to_free_tracepoints = true;
 91 
 92         while (early_probes) {
 93                 tmp = early_probes;
 94                 early_probes = tmp->next;
 95                 call_rcu(tmp, rcu_free_old_probes);
 96         }
 97 
 98         return 0;
 99 }
100 
101 /* SRCU is initialized at core_initcall */
102 postcore_initcall(release_early_probes);
103 
104 static inline void release_probes(struct tracepoint_func *old)
105 {
106         if (old) {
107                 struct tp_probes *tp_probes = container_of(old,
108                         struct tp_probes, probes[0]);
109 
110                 /*
111                  * We can't free probes if SRCU is not initialized yet.
112                  * Postpone the freeing till after SRCU is initialized.
113                  */
114                 if (unlikely(!ok_to_free_tracepoints)) {
115                         tp_probes->rcu.next = early_probes;
116                         early_probes = &tp_probes->rcu;
117                         return;
118                 }
119 
120                 /*
121                  * Tracepoint probes are protected by both sched RCU and SRCU,
122                  * by calling the SRCU callback in the sched RCU callback we
123                  * cover both cases. So let us chain the SRCU and sched RCU
124                  * callbacks to wait for both grace periods.
125                  */
126                 call_rcu(&tp_probes->rcu, rcu_free_old_probes);
127         }
128 }
129 
130 static void debug_print_probes(struct tracepoint_func *funcs)
131 {
132         int i;
133 
134         if (!tracepoint_debug || !funcs)
135                 return;
136 
137         for (i = 0; funcs[i].func; i++)
138                 printk(KERN_DEBUG "Probe %d : %p\n", i, funcs[i].func);
139 }
140 
141 static struct tracepoint_func *
142 func_add(struct tracepoint_func **funcs, struct tracepoint_func *tp_func,
143          int prio)
144 {
145         struct tracepoint_func *old, *new;
146         int nr_probes = 0;
147         int pos = -1;
148 
149         if (WARN_ON(!tp_func->func))
150                 return ERR_PTR(-EINVAL);
151 
152         debug_print_probes(*funcs);
153         old = *funcs;
154         if (old) {
155                 /* (N -> N+1), (N != 0, 1) probes */
156                 for (nr_probes = 0; old[nr_probes].func; nr_probes++) {
157                         /* Insert before probes of lower priority */
158                         if (pos < 0 && old[nr_probes].prio < prio)
159                                 pos = nr_probes;
160                         if (old[nr_probes].func == tp_func->func &&
161                             old[nr_probes].data == tp_func->data)
162                                 return ERR_PTR(-EEXIST);
163                 }
164         }
165         /* + 2 : one for new probe, one for NULL func */
166         new = allocate_probes(nr_probes + 2);
167         if (new == NULL)
168                 return ERR_PTR(-ENOMEM);
169         if (old) {
170                 if (pos < 0) {
171                         pos = nr_probes;
172                         memcpy(new, old, nr_probes * sizeof(struct tracepoint_func));
173                 } else {
174                         /* Copy higher priority probes ahead of the new probe */
175                         memcpy(new, old, pos * sizeof(struct tracepoint_func));
176                         /* Copy the rest after it. */
177                         memcpy(new + pos + 1, old + pos,
178                                (nr_probes - pos) * sizeof(struct tracepoint_func));
179                 }
180         } else
181                 pos = 0;
182         new[pos] = *tp_func;
183         new[nr_probes + 1].func = NULL;
184         *funcs = new;
185         debug_print_probes(*funcs);
186         return old;
187 }
188 
189 static void *func_remove(struct tracepoint_func **funcs,
190                 struct tracepoint_func *tp_func)
191 {
192         int nr_probes = 0, nr_del = 0, i;
193         struct tracepoint_func *old, *new;
194 
195         old = *funcs;
196 
197         if (!old)
198                 return ERR_PTR(-ENOENT);
199 
200         debug_print_probes(*funcs);
201         /* (N -> M), (N > 1, M >= 0) probes */
202         if (tp_func->func) {
203                 for (nr_probes = 0; old[nr_probes].func; nr_probes++) {
204                         if (old[nr_probes].func == tp_func->func &&
205                              old[nr_probes].data == tp_func->data)
206                                 nr_del++;
207                 }
208         }
209 
210         /*
211          * If probe is NULL, then nr_probes = nr_del = 0, and then the
212          * entire entry will be removed.
213          */
214         if (nr_probes - nr_del == 0) {
215                 /* N -> 0, (N > 1) */
216                 *funcs = NULL;
217                 debug_print_probes(*funcs);
218                 return old;
219         } else {
220                 int j = 0;
221                 /* N -> M, (N > 1, M > 0) */
222                 /* + 1 for NULL */
223                 new = allocate_probes(nr_probes - nr_del + 1);
224                 if (new == NULL)
225                         return ERR_PTR(-ENOMEM);
226                 for (i = 0; old[i].func; i++)
227                         if (old[i].func != tp_func->func
228                                         || old[i].data != tp_func->data)
229                                 new[j++] = old[i];
230                 new[nr_probes - nr_del].func = NULL;
231                 *funcs = new;
232         }
233         debug_print_probes(*funcs);
234         return old;
235 }
236 
237 /*
238  * Add the probe function to a tracepoint.
239  */
240 static int tracepoint_add_func(struct tracepoint *tp,
241                                struct tracepoint_func *func, int prio)
242 {
243         struct tracepoint_func *old, *tp_funcs;
244         int ret;
245 
246         if (tp->regfunc && !static_key_enabled(&tp->key)) {
247                 ret = tp->regfunc();
248                 if (ret < 0)
249                         return ret;
250         }
251 
252         tp_funcs = rcu_dereference_protected(tp->funcs,
253                         lockdep_is_held(&tracepoints_mutex));
254         old = func_add(&tp_funcs, func, prio);
255         if (IS_ERR(old)) {
256                 WARN_ON_ONCE(PTR_ERR(old) != -ENOMEM);
257                 return PTR_ERR(old);
258         }
259 
260         /*
261          * rcu_assign_pointer has as smp_store_release() which makes sure
262          * that the new probe callbacks array is consistent before setting
263          * a pointer to it.  This array is referenced by __DO_TRACE from
264          * include/linux/tracepoint.h using rcu_dereference_sched().
265          */
266         rcu_assign_pointer(tp->funcs, tp_funcs);
267         if (!static_key_enabled(&tp->key))
268                 static_key_slow_inc(&tp->key);
269         release_probes(old);
270         return 0;
271 }
272 
273 /*
274  * Remove a probe function from a tracepoint.
275  * Note: only waiting an RCU period after setting elem->call to the empty
276  * function insures that the original callback is not used anymore. This insured
277  * by preempt_disable around the call site.
278  */
279 static int tracepoint_remove_func(struct tracepoint *tp,
280                 struct tracepoint_func *func)
281 {
282         struct tracepoint_func *old, *tp_funcs;
283 
284         tp_funcs = rcu_dereference_protected(tp->funcs,
285                         lockdep_is_held(&tracepoints_mutex));
286         old = func_remove(&tp_funcs, func);
287         if (IS_ERR(old)) {
288                 WARN_ON_ONCE(PTR_ERR(old) != -ENOMEM);
289                 return PTR_ERR(old);
290         }
291 
292         if (!tp_funcs) {
293                 /* Removed last function */
294                 if (tp->unregfunc && static_key_enabled(&tp->key))
295                         tp->unregfunc();
296 
297                 if (static_key_enabled(&tp->key))
298                         static_key_slow_dec(&tp->key);
299         }
300         rcu_assign_pointer(tp->funcs, tp_funcs);
301         release_probes(old);
302         return 0;
303 }
304 
305 /**
306  * tracepoint_probe_register_prio -  Connect a probe to a tracepoint with priority
307  * @tp: tracepoint
308  * @probe: probe handler
309  * @data: tracepoint data
310  * @prio: priority of this function over other registered functions
311  *
312  * Returns 0 if ok, error value on error.
313  * Note: if @tp is within a module, the caller is responsible for
314  * unregistering the probe before the module is gone. This can be
315  * performed either with a tracepoint module going notifier, or from
316  * within module exit functions.
317  */
318 int tracepoint_probe_register_prio(struct tracepoint *tp, void *probe,
319                                    void *data, int prio)
320 {
321         struct tracepoint_func tp_func;
322         int ret;
323 
324         mutex_lock(&tracepoints_mutex);
325         tp_func.func = probe;
326         tp_func.data = data;
327         tp_func.prio = prio;
328         ret = tracepoint_add_func(tp, &tp_func, prio);
329         mutex_unlock(&tracepoints_mutex);
330         return ret;
331 }
332 EXPORT_SYMBOL_GPL(tracepoint_probe_register_prio);
333 
334 /**
335  * tracepoint_probe_register -  Connect a probe to a tracepoint
336  * @tp: tracepoint
337  * @probe: probe handler
338  * @data: tracepoint data
339  *
340  * Returns 0 if ok, error value on error.
341  * Note: if @tp is within a module, the caller is responsible for
342  * unregistering the probe before the module is gone. This can be
343  * performed either with a tracepoint module going notifier, or from
344  * within module exit functions.
345  */
346 int tracepoint_probe_register(struct tracepoint *tp, void *probe, void *data)
347 {
348         return tracepoint_probe_register_prio(tp, probe, data, TRACEPOINT_DEFAULT_PRIO);
349 }
350 EXPORT_SYMBOL_GPL(tracepoint_probe_register);
351 
352 /**
353  * tracepoint_probe_unregister -  Disconnect a probe from a tracepoint
354  * @tp: tracepoint
355  * @probe: probe function pointer
356  * @data: tracepoint data
357  *
358  * Returns 0 if ok, error value on error.
359  */
360 int tracepoint_probe_unregister(struct tracepoint *tp, void *probe, void *data)
361 {
362         struct tracepoint_func tp_func;
363         int ret;
364 
365         mutex_lock(&tracepoints_mutex);
366         tp_func.func = probe;
367         tp_func.data = data;
368         ret = tracepoint_remove_func(tp, &tp_func);
369         mutex_unlock(&tracepoints_mutex);
370         return ret;
371 }
372 EXPORT_SYMBOL_GPL(tracepoint_probe_unregister);
373 
374 static void for_each_tracepoint_range(
375                 tracepoint_ptr_t *begin, tracepoint_ptr_t *end,
376                 void (*fct)(struct tracepoint *tp, void *priv),
377                 void *priv)
378 {
379         tracepoint_ptr_t *iter;
380 
381         if (!begin)
382                 return;
383         for (iter = begin; iter < end; iter++)
384                 fct(tracepoint_ptr_deref(iter), priv);
385 }
386 
387 #ifdef CONFIG_MODULES
388 bool trace_module_has_bad_taint(struct module *mod)
389 {
390         return mod->taints & ~((1 << TAINT_OOT_MODULE) | (1 << TAINT_CRAP) |
391                                (1 << TAINT_UNSIGNED_MODULE));
392 }
393 
394 static BLOCKING_NOTIFIER_HEAD(tracepoint_notify_list);
395 
396 /**
397  * register_tracepoint_notifier - register tracepoint coming/going notifier
398  * @nb: notifier block
399  *
400  * Notifiers registered with this function are called on module
401  * coming/going with the tracepoint_module_list_mutex held.
402  * The notifier block callback should expect a "struct tp_module" data
403  * pointer.
404  */
405 int register_tracepoint_module_notifier(struct notifier_block *nb)
406 {
407         struct tp_module *tp_mod;
408         int ret;
409 
410         mutex_lock(&tracepoint_module_list_mutex);
411         ret = blocking_notifier_chain_register(&tracepoint_notify_list, nb);
412         if (ret)
413                 goto end;
414         list_for_each_entry(tp_mod, &tracepoint_module_list, list)
415                 (void) nb->notifier_call(nb, MODULE_STATE_COMING, tp_mod);
416 end:
417         mutex_unlock(&tracepoint_module_list_mutex);
418         return ret;
419 }
420 EXPORT_SYMBOL_GPL(register_tracepoint_module_notifier);
421 
422 /**
423  * unregister_tracepoint_notifier - unregister tracepoint coming/going notifier
424  * @nb: notifier block
425  *
426  * The notifier block callback should expect a "struct tp_module" data
427  * pointer.
428  */
429 int unregister_tracepoint_module_notifier(struct notifier_block *nb)
430 {
431         struct tp_module *tp_mod;
432         int ret;
433 
434         mutex_lock(&tracepoint_module_list_mutex);
435         ret = blocking_notifier_chain_unregister(&tracepoint_notify_list, nb);
436         if (ret)
437                 goto end;
438         list_for_each_entry(tp_mod, &tracepoint_module_list, list)
439                 (void) nb->notifier_call(nb, MODULE_STATE_GOING, tp_mod);
440 end:
441         mutex_unlock(&tracepoint_module_list_mutex);
442         return ret;
443 
444 }
445 EXPORT_SYMBOL_GPL(unregister_tracepoint_module_notifier);
446 
447 /*
448  * Ensure the tracer unregistered the module's probes before the module
449  * teardown is performed. Prevents leaks of probe and data pointers.
450  */
451 static void tp_module_going_check_quiescent(struct tracepoint *tp, void *priv)
452 {
453         WARN_ON_ONCE(tp->funcs);
454 }
455 
456 static int tracepoint_module_coming(struct module *mod)
457 {
458         struct tp_module *tp_mod;
459         int ret = 0;
460 
461         if (!mod->num_tracepoints)
462                 return 0;
463 
464         /*
465          * We skip modules that taint the kernel, especially those with different
466          * module headers (for forced load), to make sure we don't cause a crash.
467          * Staging, out-of-tree, and unsigned GPL modules are fine.
468          */
469         if (trace_module_has_bad_taint(mod))
470                 return 0;
471         mutex_lock(&tracepoint_module_list_mutex);
472         tp_mod = kmalloc(sizeof(struct tp_module), GFP_KERNEL);
473         if (!tp_mod) {
474                 ret = -ENOMEM;
475                 goto end;
476         }
477         tp_mod->mod = mod;
478         list_add_tail(&tp_mod->list, &tracepoint_module_list);
479         blocking_notifier_call_chain(&tracepoint_notify_list,
480                         MODULE_STATE_COMING, tp_mod);
481 end:
482         mutex_unlock(&tracepoint_module_list_mutex);
483         return ret;
484 }
485 
486 static void tracepoint_module_going(struct module *mod)
487 {
488         struct tp_module *tp_mod;
489 
490         if (!mod->num_tracepoints)
491                 return;
492 
493         mutex_lock(&tracepoint_module_list_mutex);
494         list_for_each_entry(tp_mod, &tracepoint_module_list, list) {
495                 if (tp_mod->mod == mod) {
496                         blocking_notifier_call_chain(&tracepoint_notify_list,
497                                         MODULE_STATE_GOING, tp_mod);
498                         list_del(&tp_mod->list);
499                         kfree(tp_mod);
500                         /*
501                          * Called the going notifier before checking for
502                          * quiescence.
503                          */
504                         for_each_tracepoint_range(mod->tracepoints_ptrs,
505                                 mod->tracepoints_ptrs + mod->num_tracepoints,
506                                 tp_module_going_check_quiescent, NULL);
507                         break;
508                 }
509         }
510         /*
511          * In the case of modules that were tainted at "coming", we'll simply
512          * walk through the list without finding it. We cannot use the "tainted"
513          * flag on "going", in case a module taints the kernel only after being
514          * loaded.
515          */
516         mutex_unlock(&tracepoint_module_list_mutex);
517 }
518 
519 static int tracepoint_module_notify(struct notifier_block *self,
520                 unsigned long val, void *data)
521 {
522         struct module *mod = data;
523         int ret = 0;
524 
525         switch (val) {
526         case MODULE_STATE_COMING:
527                 ret = tracepoint_module_coming(mod);
528                 break;
529         case MODULE_STATE_LIVE:
530                 break;
531         case MODULE_STATE_GOING:
532                 tracepoint_module_going(mod);
533                 break;
534         case MODULE_STATE_UNFORMED:
535                 break;
536         }
537         return ret;
538 }
539 
540 static struct notifier_block tracepoint_module_nb = {
541         .notifier_call = tracepoint_module_notify,
542         .priority = 0,
543 };
544 
545 static __init int init_tracepoints(void)
546 {
547         int ret;
548 
549         ret = register_module_notifier(&tracepoint_module_nb);
550         if (ret)
551                 pr_warn("Failed to register tracepoint module enter notifier\n");
552 
553         return ret;
554 }
555 __initcall(init_tracepoints);
556 #endif /* CONFIG_MODULES */
557 
558 /**
559  * for_each_kernel_tracepoint - iteration on all kernel tracepoints
560  * @fct: callback
561  * @priv: private data
562  */
563 void for_each_kernel_tracepoint(void (*fct)(struct tracepoint *tp, void *priv),
564                 void *priv)
565 {
566         for_each_tracepoint_range(__start___tracepoints_ptrs,
567                 __stop___tracepoints_ptrs, fct, priv);
568 }
569 EXPORT_SYMBOL_GPL(for_each_kernel_tracepoint);
570 
571 #ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
572 
573 /* NB: reg/unreg are called while guarded with the tracepoints_mutex */
574 static int sys_tracepoint_refcount;
575 
576 int syscall_regfunc(void)
577 {
578         struct task_struct *p, *t;
579 
580         if (!sys_tracepoint_refcount) {
581                 read_lock(&tasklist_lock);
582                 for_each_process_thread(p, t) {
583                         set_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
584                 }
585                 read_unlock(&tasklist_lock);
586         }
587         sys_tracepoint_refcount++;
588 
589         return 0;
590 }
591 
592 void syscall_unregfunc(void)
593 {
594         struct task_struct *p, *t;
595 
596         sys_tracepoint_refcount--;
597         if (!sys_tracepoint_refcount) {
598                 read_lock(&tasklist_lock);
599                 for_each_process_thread(p, t) {
600                         clear_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
601                 }
602                 read_unlock(&tasklist_lock);
603         }
604 }
605 #endif
606 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp