aboutsummaryrefslogtreecommitdiffstats
path: root/kernel/trace/trace_kprobe.c
diff options
context:
space:
mode:
authorMasami Hiramatsu <mhiramat@redhat.com>2009-08-13 16:35:11 -0400
committerFrederic Weisbecker <fweisbec@gmail.com>2009-08-26 20:31:54 -0400
commit413d37d1eb69c1765b9ace0a612dac9b6c990e66 (patch)
tree6f483714046c388ab9f57b33f17f365cd4e57446 /kernel/trace/trace_kprobe.c
parentd93f12f3f417e49a175800da85c6fcb2a5096e03 (diff)
tracing: Add kprobe-based event tracer
Add kprobes-based event tracer on ftrace. This tracer is similar to the events tracer which is based on Tracepoint infrastructure. Instead of Tracepoint, this tracer is based on kprobes (kprobe and kretprobe). It probes anywhere where kprobes can probe(this means, all functions body except for __kprobes functions). Similar to the events tracer, this tracer doesn't need to be activated via current_tracer, instead of that, just set probe points via /sys/kernel/debug/tracing/kprobe_events. And you can set filters on each probe events via /sys/kernel/debug/tracing/events/kprobes/<EVENT>/filter. This tracer supports following probe arguments for each probe. %REG : Fetch register REG sN : Fetch Nth entry of stack (N >= 0) sa : Fetch stack address. @ADDR : Fetch memory at ADDR (ADDR should be in kernel) @SYM[+|-offs] : Fetch memory at SYM +|- offs (SYM should be a data symbol) aN : Fetch function argument. (N >= 0) rv : Fetch return value. ra : Fetch return address. +|-offs(FETCHARG) : fetch memory at FETCHARG +|- offs address. See Documentation/trace/kprobetrace.txt in the next patch for details. Changes from v13: - Support 'sa' for stack address. - Use call->data instead of container_of() macro. [fweisbec@gmail.com: Fixed conflict against latest tracing/core] Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com> Acked-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Avi Kivity <avi@redhat.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K.Prasad <prasad@linux.vnet.ibm.com> Cc: Lai Jiangshan <laijs@cn.fujitsu.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Cc: Przemysław Pawełczyk <przemyslaw@pawelczyk.it> Cc: Roland McGrath <roland@redhat.com> Cc: Sam Ravnborg <sam@ravnborg.org> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Vegard Nossum <vegard.nossum@gmail.com> LKML-Reference: <20090813203510.31965.29123.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Diffstat (limited to 'kernel/trace/trace_kprobe.c')
-rw-r--r--kernel/trace/trace_kprobe.c1202
1 files changed, 1202 insertions, 0 deletions
diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
new file mode 100644
index 000000000000..0c4f00aafb92
--- /dev/null
+++ b/kernel/trace/trace_kprobe.c
@@ -0,0 +1,1202 @@
1/*
2 * kprobe based kernel tracer
3 *
4 * Created by Masami Hiramatsu <mhiramat@redhat.com>
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18 */
19
20#include <linux/module.h>
21#include <linux/uaccess.h>
22#include <linux/kprobes.h>
23#include <linux/seq_file.h>
24#include <linux/slab.h>
25#include <linux/smp.h>
26#include <linux/debugfs.h>
27#include <linux/types.h>
28#include <linux/string.h>
29#include <linux/ctype.h>
30#include <linux/ptrace.h>
31
32#include "trace.h"
33#include "trace_output.h"
34
35#define TRACE_KPROBE_ARGS 6
36#define MAX_ARGSTR_LEN 63
37
38/* currently, trace_kprobe only supports X86. */
39
40struct fetch_func {
41 unsigned long (*func)(struct pt_regs *, void *);
42 void *data;
43};
44
45static __kprobes unsigned long call_fetch(struct fetch_func *f,
46 struct pt_regs *regs)
47{
48 return f->func(regs, f->data);
49}
50
51/* fetch handlers */
52static __kprobes unsigned long fetch_register(struct pt_regs *regs,
53 void *offset)
54{
55 return regs_get_register(regs, (unsigned int)((unsigned long)offset));
56}
57
58static __kprobes unsigned long fetch_stack(struct pt_regs *regs,
59 void *num)
60{
61 return regs_get_kernel_stack_nth(regs,
62 (unsigned int)((unsigned long)num));
63}
64
65static __kprobes unsigned long fetch_memory(struct pt_regs *regs, void *addr)
66{
67 unsigned long retval;
68
69 if (probe_kernel_address(addr, retval))
70 return 0;
71 return retval;
72}
73
74static __kprobes unsigned long fetch_argument(struct pt_regs *regs, void *num)
75{
76 return regs_get_argument_nth(regs, (unsigned int)((unsigned long)num));
77}
78
79static __kprobes unsigned long fetch_retvalue(struct pt_regs *regs,
80 void *dummy)
81{
82 return regs_return_value(regs);
83}
84
85static __kprobes unsigned long fetch_ip(struct pt_regs *regs, void *dummy)
86{
87 return instruction_pointer(regs);
88}
89
90static __kprobes unsigned long fetch_stack_address(struct pt_regs *regs,
91 void *dummy)
92{
93 return kernel_stack_pointer(regs);
94}
95
96/* Memory fetching by symbol */
97struct symbol_cache {
98 char *symbol;
99 long offset;
100 unsigned long addr;
101};
102
103static unsigned long update_symbol_cache(struct symbol_cache *sc)
104{
105 sc->addr = (unsigned long)kallsyms_lookup_name(sc->symbol);
106 if (sc->addr)
107 sc->addr += sc->offset;
108 return sc->addr;
109}
110
111static void free_symbol_cache(struct symbol_cache *sc)
112{
113 kfree(sc->symbol);
114 kfree(sc);
115}
116
117static struct symbol_cache *alloc_symbol_cache(const char *sym, long offset)
118{
119 struct symbol_cache *sc;
120
121 if (!sym || strlen(sym) == 0)
122 return NULL;
123 sc = kzalloc(sizeof(struct symbol_cache), GFP_KERNEL);
124 if (!sc)
125 return NULL;
126
127 sc->symbol = kstrdup(sym, GFP_KERNEL);
128 if (!sc->symbol) {
129 kfree(sc);
130 return NULL;
131 }
132 sc->offset = offset;
133
134 update_symbol_cache(sc);
135 return sc;
136}
137
138static __kprobes unsigned long fetch_symbol(struct pt_regs *regs, void *data)
139{
140 struct symbol_cache *sc = data;
141
142 if (sc->addr)
143 return fetch_memory(regs, (void *)sc->addr);
144 else
145 return 0;
146}
147
148/* Special indirect memory access interface */
149struct indirect_fetch_data {
150 struct fetch_func orig;
151 long offset;
152};
153
154static __kprobes unsigned long fetch_indirect(struct pt_regs *regs, void *data)
155{
156 struct indirect_fetch_data *ind = data;
157 unsigned long addr;
158
159 addr = call_fetch(&ind->orig, regs);
160 if (addr) {
161 addr += ind->offset;
162 return fetch_memory(regs, (void *)addr);
163 } else
164 return 0;
165}
166
167static __kprobes void free_indirect_fetch_data(struct indirect_fetch_data *data)
168{
169 if (data->orig.func == fetch_indirect)
170 free_indirect_fetch_data(data->orig.data);
171 else if (data->orig.func == fetch_symbol)
172 free_symbol_cache(data->orig.data);
173 kfree(data);
174}
175
176/**
177 * kprobe_trace_core
178 */
179
180struct trace_probe {
181 struct list_head list;
182 union {
183 struct kprobe kp;
184 struct kretprobe rp;
185 };
186 const char *symbol; /* symbol name */
187 unsigned int nr_args;
188 struct fetch_func args[TRACE_KPROBE_ARGS];
189 struct ftrace_event_call call;
190};
191
192static int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs);
193static int kretprobe_trace_func(struct kretprobe_instance *ri,
194 struct pt_regs *regs);
195
196static __kprobes int probe_is_return(struct trace_probe *tp)
197{
198 return (tp->rp.handler == kretprobe_trace_func);
199}
200
201static __kprobes const char *probe_symbol(struct trace_probe *tp)
202{
203 return tp->symbol ? tp->symbol : "unknown";
204}
205
206static __kprobes long probe_offset(struct trace_probe *tp)
207{
208 return (probe_is_return(tp)) ? tp->rp.kp.offset : tp->kp.offset;
209}
210
211static __kprobes void *probe_address(struct trace_probe *tp)
212{
213 return (probe_is_return(tp)) ? tp->rp.kp.addr : tp->kp.addr;
214}
215
216static int trace_arg_string(char *buf, size_t n, struct fetch_func *ff)
217{
218 int ret = -EINVAL;
219
220 if (ff->func == fetch_argument)
221 ret = snprintf(buf, n, "a%lu", (unsigned long)ff->data);
222 else if (ff->func == fetch_register) {
223 const char *name;
224 name = regs_query_register_name((unsigned int)((long)ff->data));
225 ret = snprintf(buf, n, "%%%s", name);
226 } else if (ff->func == fetch_stack)
227 ret = snprintf(buf, n, "s%lu", (unsigned long)ff->data);
228 else if (ff->func == fetch_memory)
229 ret = snprintf(buf, n, "@0x%p", ff->data);
230 else if (ff->func == fetch_symbol) {
231 struct symbol_cache *sc = ff->data;
232 ret = snprintf(buf, n, "@%s%+ld", sc->symbol, sc->offset);
233 } else if (ff->func == fetch_retvalue)
234 ret = snprintf(buf, n, "rv");
235 else if (ff->func == fetch_ip)
236 ret = snprintf(buf, n, "ra");
237 else if (ff->func == fetch_stack_address)
238 ret = snprintf(buf, n, "sa");
239 else if (ff->func == fetch_indirect) {
240 struct indirect_fetch_data *id = ff->data;
241 size_t l = 0;
242 ret = snprintf(buf, n, "%+ld(", id->offset);
243 if (ret >= n)
244 goto end;
245 l += ret;
246 ret = trace_arg_string(buf + l, n - l, &id->orig);
247 if (ret < 0)
248 goto end;
249 l += ret;
250 ret = snprintf(buf + l, n - l, ")");
251 ret += l;
252 }
253end:
254 if (ret >= n)
255 return -ENOSPC;
256 return ret;
257}
258
259static int register_probe_event(struct trace_probe *tp);
260static void unregister_probe_event(struct trace_probe *tp);
261
262static DEFINE_MUTEX(probe_lock);
263static LIST_HEAD(probe_list);
264
265static struct trace_probe *alloc_trace_probe(const char *symbol,
266 const char *event)
267{
268 struct trace_probe *tp;
269
270 tp = kzalloc(sizeof(struct trace_probe), GFP_KERNEL);
271 if (!tp)
272 return ERR_PTR(-ENOMEM);
273
274 if (symbol) {
275 tp->symbol = kstrdup(symbol, GFP_KERNEL);
276 if (!tp->symbol)
277 goto error;
278 }
279 if (event) {
280 tp->call.name = kstrdup(event, GFP_KERNEL);
281 if (!tp->call.name)
282 goto error;
283 }
284
285 INIT_LIST_HEAD(&tp->list);
286 return tp;
287error:
288 kfree(tp->symbol);
289 kfree(tp);
290 return ERR_PTR(-ENOMEM);
291}
292
293static void free_trace_probe(struct trace_probe *tp)
294{
295 int i;
296
297 for (i = 0; i < tp->nr_args; i++)
298 if (tp->args[i].func == fetch_symbol)
299 free_symbol_cache(tp->args[i].data);
300 else if (tp->args[i].func == fetch_indirect)
301 free_indirect_fetch_data(tp->args[i].data);
302
303 kfree(tp->call.name);
304 kfree(tp->symbol);
305 kfree(tp);
306}
307
308static struct trace_probe *find_probe_event(const char *event)
309{
310 struct trace_probe *tp;
311
312 list_for_each_entry(tp, &probe_list, list)
313 if (tp->call.name && !strcmp(tp->call.name, event))
314 return tp;
315 return NULL;
316}
317
318static void __unregister_trace_probe(struct trace_probe *tp)
319{
320 if (probe_is_return(tp))
321 unregister_kretprobe(&tp->rp);
322 else
323 unregister_kprobe(&tp->kp);
324}
325
326/* Unregister a trace_probe and probe_event: call with locking probe_lock */
327static void unregister_trace_probe(struct trace_probe *tp)
328{
329 if (tp->call.name)
330 unregister_probe_event(tp);
331 __unregister_trace_probe(tp);
332 list_del(&tp->list);
333}
334
335/* Register a trace_probe and probe_event */
336static int register_trace_probe(struct trace_probe *tp)
337{
338 struct trace_probe *old_tp;
339 int ret;
340
341 mutex_lock(&probe_lock);
342
343 if (probe_is_return(tp))
344 ret = register_kretprobe(&tp->rp);
345 else
346 ret = register_kprobe(&tp->kp);
347
348 if (ret) {
349 pr_warning("Could not insert probe(%d)\n", ret);
350 if (ret == -EILSEQ) {
351 pr_warning("Probing address(0x%p) is not an "
352 "instruction boundary.\n",
353 probe_address(tp));
354 ret = -EINVAL;
355 }
356 goto end;
357 }
358 /* register as an event */
359 if (tp->call.name) {
360 old_tp = find_probe_event(tp->call.name);
361 if (old_tp) {
362 /* delete old event */
363 unregister_trace_probe(old_tp);
364 free_trace_probe(old_tp);
365 }
366 ret = register_probe_event(tp);
367 if (ret) {
368 pr_warning("Faild to register probe event(%d)\n", ret);
369 __unregister_trace_probe(tp);
370 }
371 }
372 list_add_tail(&tp->list, &probe_list);
373end:
374 mutex_unlock(&probe_lock);
375 return ret;
376}
377
378/* Split symbol and offset. */
379static int split_symbol_offset(char *symbol, long *offset)
380{
381 char *tmp;
382 int ret;
383
384 if (!offset)
385 return -EINVAL;
386
387 tmp = strchr(symbol, '+');
388 if (!tmp)
389 tmp = strchr(symbol, '-');
390
391 if (tmp) {
392 /* skip sign because strict_strtol doesn't accept '+' */
393 ret = strict_strtol(tmp + 1, 0, offset);
394 if (ret)
395 return ret;
396 if (*tmp == '-')
397 *offset = -(*offset);
398 *tmp = '\0';
399 } else
400 *offset = 0;
401 return 0;
402}
403
404#define PARAM_MAX_ARGS 16
405#define PARAM_MAX_STACK (THREAD_SIZE / sizeof(unsigned long))
406
407static int parse_trace_arg(char *arg, struct fetch_func *ff, int is_return)
408{
409 int ret = 0;
410 unsigned long param;
411 long offset;
412 char *tmp;
413
414 switch (arg[0]) {
415 case 'a': /* argument */
416 ret = strict_strtoul(arg + 1, 10, &param);
417 if (ret || param > PARAM_MAX_ARGS)
418 ret = -EINVAL;
419 else {
420 ff->func = fetch_argument;
421 ff->data = (void *)param;
422 }
423 break;
424 case 'r': /* retval or retaddr */
425 if (is_return && arg[1] == 'v') {
426 ff->func = fetch_retvalue;
427 ff->data = NULL;
428 } else if (is_return && arg[1] == 'a') {
429 ff->func = fetch_ip;
430 ff->data = NULL;
431 } else
432 ret = -EINVAL;
433 break;
434 case '%': /* named register */
435 ret = regs_query_register_offset(arg + 1);
436 if (ret >= 0) {
437 ff->func = fetch_register;
438 ff->data = (void *)(unsigned long)ret;
439 ret = 0;
440 }
441 break;
442 case 's': /* stack */
443 if (arg[1] == 'a') {
444 ff->func = fetch_stack_address;
445 ff->data = NULL;
446 } else {
447 ret = strict_strtoul(arg + 1, 10, &param);
448 if (ret || param > PARAM_MAX_STACK)
449 ret = -EINVAL;
450 else {
451 ff->func = fetch_stack;
452 ff->data = (void *)param;
453 }
454 }
455 break;
456 case '@': /* memory or symbol */
457 if (isdigit(arg[1])) {
458 ret = strict_strtoul(arg + 1, 0, &param);
459 if (ret)
460 break;
461 ff->func = fetch_memory;
462 ff->data = (void *)param;
463 } else {
464 ret = split_symbol_offset(arg + 1, &offset);
465 if (ret)
466 break;
467 ff->data = alloc_symbol_cache(arg + 1,
468 offset);
469 if (ff->data)
470 ff->func = fetch_symbol;
471 else
472 ret = -EINVAL;
473 }
474 break;
475 case '+': /* indirect memory */
476 case '-':
477 tmp = strchr(arg, '(');
478 if (!tmp) {
479 ret = -EINVAL;
480 break;
481 }
482 *tmp = '\0';
483 ret = strict_strtol(arg + 1, 0, &offset);
484 if (ret)
485 break;
486 if (arg[0] == '-')
487 offset = -offset;
488 arg = tmp + 1;
489 tmp = strrchr(arg, ')');
490 if (tmp) {
491 struct indirect_fetch_data *id;
492 *tmp = '\0';
493 id = kzalloc(sizeof(struct indirect_fetch_data),
494 GFP_KERNEL);
495 if (!id)
496 return -ENOMEM;
497 id->offset = offset;
498 ret = parse_trace_arg(arg, &id->orig, is_return);
499 if (ret)
500 kfree(id);
501 else {
502 ff->func = fetch_indirect;
503 ff->data = (void *)id;
504 }
505 } else
506 ret = -EINVAL;
507 break;
508 default:
509 /* TODO: support custom handler */
510 ret = -EINVAL;
511 }
512 return ret;
513}
514
515static int create_trace_probe(int argc, char **argv)
516{
517 /*
518 * Argument syntax:
519 * - Add kprobe: p[:EVENT] SYMBOL[+OFFS|-OFFS]|ADDRESS [FETCHARGS]
520 * - Add kretprobe: r[:EVENT] SYMBOL[+0] [FETCHARGS]
521 * Fetch args:
522 * aN : fetch Nth of function argument. (N:0-)
523 * rv : fetch return value
524 * ra : fetch return address
525 * sa : fetch stack address
526 * sN : fetch Nth of stack (N:0-)
527 * @ADDR : fetch memory at ADDR (ADDR should be in kernel)
528 * @SYM[+|-offs] : fetch memory at SYM +|- offs (SYM is a data symbol)
529 * %REG : fetch register REG
530 * Indirect memory fetch:
531 * +|-offs(ARG) : fetch memory at ARG +|- offs address.
532 */
533 struct trace_probe *tp;
534 struct kprobe *kp;
535 int i, ret = 0;
536 int is_return = 0;
537 char *symbol = NULL, *event = NULL;
538 long offset = 0;
539 void *addr = NULL;
540
541 if (argc < 2)
542 return -EINVAL;
543
544 if (argv[0][0] == 'p')
545 is_return = 0;
546 else if (argv[0][0] == 'r')
547 is_return = 1;
548 else
549 return -EINVAL;
550
551 if (argv[0][1] == ':') {
552 event = &argv[0][2];
553 if (strlen(event) == 0) {
554 pr_info("Event name is not specifiled\n");
555 return -EINVAL;
556 }
557 }
558
559 if (isdigit(argv[1][0])) {
560 if (is_return)
561 return -EINVAL;
562 /* an address specified */
563 ret = strict_strtoul(&argv[0][2], 0, (unsigned long *)&addr);
564 if (ret)
565 return ret;
566 } else {
567 /* a symbol specified */
568 symbol = argv[1];
569 /* TODO: support .init module functions */
570 ret = split_symbol_offset(symbol, &offset);
571 if (ret)
572 return ret;
573 if (offset && is_return)
574 return -EINVAL;
575 }
576
577 /* setup a probe */
578 tp = alloc_trace_probe(symbol, event);
579 if (IS_ERR(tp))
580 return PTR_ERR(tp);
581
582 if (is_return) {
583 kp = &tp->rp.kp;
584 tp->rp.handler = kretprobe_trace_func;
585 } else {
586 kp = &tp->kp;
587 tp->kp.pre_handler = kprobe_trace_func;
588 }
589
590 if (tp->symbol) {
591 kp->symbol_name = tp->symbol;
592 kp->offset = offset;
593 } else
594 kp->addr = addr;
595
596 /* parse arguments */
597 argc -= 2; argv += 2; ret = 0;
598 for (i = 0; i < argc && i < TRACE_KPROBE_ARGS; i++) {
599 if (strlen(argv[i]) > MAX_ARGSTR_LEN) {
600 pr_info("Argument%d(%s) is too long.\n", i, argv[i]);
601 ret = -ENOSPC;
602 goto error;
603 }
604 ret = parse_trace_arg(argv[i], &tp->args[i], is_return);
605 if (ret)
606 goto error;
607 }
608 tp->nr_args = i;
609
610 ret = register_trace_probe(tp);
611 if (ret)
612 goto error;
613 return 0;
614
615error:
616 free_trace_probe(tp);
617 return ret;
618}
619
620static void cleanup_all_probes(void)
621{
622 struct trace_probe *tp;
623
624 mutex_lock(&probe_lock);
625 /* TODO: Use batch unregistration */
626 while (!list_empty(&probe_list)) {
627 tp = list_entry(probe_list.next, struct trace_probe, list);
628 unregister_trace_probe(tp);
629 free_trace_probe(tp);
630 }
631 mutex_unlock(&probe_lock);
632}
633
634
635/* Probes listing interfaces */
636static void *probes_seq_start(struct seq_file *m, loff_t *pos)
637{
638 mutex_lock(&probe_lock);
639 return seq_list_start(&probe_list, *pos);
640}
641
642static void *probes_seq_next(struct seq_file *m, void *v, loff_t *pos)
643{
644 return seq_list_next(v, &probe_list, pos);
645}
646
647static void probes_seq_stop(struct seq_file *m, void *v)
648{
649 mutex_unlock(&probe_lock);
650}
651
652static int probes_seq_show(struct seq_file *m, void *v)
653{
654 struct trace_probe *tp = v;
655 int i, ret;
656 char buf[MAX_ARGSTR_LEN + 1];
657
658 seq_printf(m, "%c", probe_is_return(tp) ? 'r' : 'p');
659 if (tp->call.name)
660 seq_printf(m, ":%s", tp->call.name);
661
662 if (tp->symbol)
663 seq_printf(m, " %s%+ld", probe_symbol(tp), probe_offset(tp));
664 else
665 seq_printf(m, " 0x%p", probe_address(tp));
666
667 for (i = 0; i < tp->nr_args; i++) {
668 ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]);
669 if (ret < 0) {
670 pr_warning("Argument%d decoding error(%d).\n", i, ret);
671 return ret;
672 }
673 seq_printf(m, " %s", buf);
674 }
675 seq_printf(m, "\n");
676 return 0;
677}
678
679static const struct seq_operations probes_seq_op = {
680 .start = probes_seq_start,
681 .next = probes_seq_next,
682 .stop = probes_seq_stop,
683 .show = probes_seq_show
684};
685
686static int probes_open(struct inode *inode, struct file *file)
687{
688 if ((file->f_mode & FMODE_WRITE) &&
689 (file->f_flags & O_TRUNC))
690 cleanup_all_probes();
691
692 return seq_open(file, &probes_seq_op);
693}
694
695static int command_trace_probe(const char *buf)
696{
697 char **argv;
698 int argc = 0, ret = 0;
699
700 argv = argv_split(GFP_KERNEL, buf, &argc);
701 if (!argv)
702 return -ENOMEM;
703
704 if (argc)
705 ret = create_trace_probe(argc, argv);
706
707 argv_free(argv);
708 return ret;
709}
710
711#define WRITE_BUFSIZE 128
712
713static ssize_t probes_write(struct file *file, const char __user *buffer,
714 size_t count, loff_t *ppos)
715{
716 char *kbuf, *tmp;
717 int ret;
718 size_t done;
719 size_t size;
720
721 kbuf = kmalloc(WRITE_BUFSIZE, GFP_KERNEL);
722 if (!kbuf)
723 return -ENOMEM;
724
725 ret = done = 0;
726 while (done < count) {
727 size = count - done;
728 if (size >= WRITE_BUFSIZE)
729 size = WRITE_BUFSIZE - 1;
730 if (copy_from_user(kbuf, buffer + done, size)) {
731 ret = -EFAULT;
732 goto out;
733 }
734 kbuf[size] = '\0';
735 tmp = strchr(kbuf, '\n');
736 if (tmp) {
737 *tmp = '\0';
738 size = tmp - kbuf + 1;
739 } else if (done + size < count) {
740 pr_warning("Line length is too long: "
741 "Should be less than %d.", WRITE_BUFSIZE);
742 ret = -EINVAL;
743 goto out;
744 }
745 done += size;
746 /* Remove comments */
747 tmp = strchr(kbuf, '#');
748 if (tmp)
749 *tmp = '\0';
750
751 ret = command_trace_probe(kbuf);
752 if (ret)
753 goto out;
754 }
755 ret = done;
756out:
757 kfree(kbuf);
758 return ret;
759}
760
761static const struct file_operations kprobe_events_ops = {
762 .owner = THIS_MODULE,
763 .open = probes_open,
764 .read = seq_read,
765 .llseek = seq_lseek,
766 .release = seq_release,
767 .write = probes_write,
768};
769
770/* Kprobe handler */
771static __kprobes int kprobe_trace_func(struct kprobe *kp, struct pt_regs *regs)
772{
773 struct trace_probe *tp = container_of(kp, struct trace_probe, kp);
774 struct kprobe_trace_entry *entry;
775 struct ring_buffer_event *event;
776 int size, i, pc;
777 unsigned long irq_flags;
778 struct ftrace_event_call *call = &event_kprobe;
779
780 if (&tp->call.name)
781 call = &tp->call;
782
783 local_save_flags(irq_flags);
784 pc = preempt_count();
785
786 size = SIZEOF_KPROBE_TRACE_ENTRY(tp->nr_args);
787
788 event = trace_current_buffer_lock_reserve(TRACE_KPROBE, size,
789 irq_flags, pc);
790 if (!event)
791 return 0;
792
793 entry = ring_buffer_event_data(event);
794 entry->nargs = tp->nr_args;
795 entry->ip = (unsigned long)kp->addr;
796 for (i = 0; i < tp->nr_args; i++)
797 entry->args[i] = call_fetch(&tp->args[i], regs);
798
799 if (!filter_current_check_discard(call, entry, event))
800 trace_nowake_buffer_unlock_commit(event, irq_flags, pc);
801 return 0;
802}
803
804/* Kretprobe handler */
805static __kprobes int kretprobe_trace_func(struct kretprobe_instance *ri,
806 struct pt_regs *regs)
807{
808 struct trace_probe *tp = container_of(ri->rp, struct trace_probe, rp);
809 struct kretprobe_trace_entry *entry;
810 struct ring_buffer_event *event;
811 int size, i, pc;
812 unsigned long irq_flags;
813 struct ftrace_event_call *call = &event_kretprobe;
814
815 if (&tp->call.name)
816 call = &tp->call;
817
818 local_save_flags(irq_flags);
819 pc = preempt_count();
820
821 size = SIZEOF_KRETPROBE_TRACE_ENTRY(tp->nr_args);
822
823 event = trace_current_buffer_lock_reserve(TRACE_KRETPROBE, size,
824 irq_flags, pc);
825 if (!event)
826 return 0;
827
828 entry = ring_buffer_event_data(event);
829 entry->nargs = tp->nr_args;
830 entry->func = (unsigned long)probe_address(tp);
831 entry->ret_ip = (unsigned long)ri->ret_addr;
832 for (i = 0; i < tp->nr_args; i++)
833 entry->args[i] = call_fetch(&tp->args[i], regs);
834
835 if (!filter_current_check_discard(call, entry, event))
836 trace_nowake_buffer_unlock_commit(event, irq_flags, pc);
837
838 return 0;
839}
840
841/* Event entry printers */
842enum print_line_t
843print_kprobe_event(struct trace_iterator *iter, int flags)
844{
845 struct kprobe_trace_entry *field;
846 struct trace_seq *s = &iter->seq;
847 int i;
848
849 trace_assign_type(field, iter->ent);
850
851 if (!seq_print_ip_sym(s, field->ip, flags | TRACE_ITER_SYM_OFFSET))
852 goto partial;
853
854 if (!trace_seq_puts(s, ":"))
855 goto partial;
856
857 for (i = 0; i < field->nargs; i++)
858 if (!trace_seq_printf(s, " 0x%lx", field->args[i]))
859 goto partial;
860
861 if (!trace_seq_puts(s, "\n"))
862 goto partial;
863
864 return TRACE_TYPE_HANDLED;
865partial:
866 return TRACE_TYPE_PARTIAL_LINE;
867}
868
869enum print_line_t
870print_kretprobe_event(struct trace_iterator *iter, int flags)
871{
872 struct kretprobe_trace_entry *field;
873 struct trace_seq *s = &iter->seq;
874 int i;
875
876 trace_assign_type(field, iter->ent);
877
878 if (!seq_print_ip_sym(s, field->ret_ip, flags | TRACE_ITER_SYM_OFFSET))
879 goto partial;
880
881 if (!trace_seq_puts(s, " <- "))
882 goto partial;
883
884 if (!seq_print_ip_sym(s, field->func, flags & ~TRACE_ITER_SYM_OFFSET))
885 goto partial;
886
887 if (!trace_seq_puts(s, ":"))
888 goto partial;
889
890 for (i = 0; i < field->nargs; i++)
891 if (!trace_seq_printf(s, " 0x%lx", field->args[i]))
892 goto partial;
893
894 if (!trace_seq_puts(s, "\n"))
895 goto partial;
896
897 return TRACE_TYPE_HANDLED;
898partial:
899 return TRACE_TYPE_PARTIAL_LINE;
900}
901
902static struct trace_event kprobe_trace_event = {
903 .type = TRACE_KPROBE,
904 .trace = print_kprobe_event,
905};
906
907static struct trace_event kretprobe_trace_event = {
908 .type = TRACE_KRETPROBE,
909 .trace = print_kretprobe_event,
910};
911
912static int probe_event_enable(struct ftrace_event_call *call)
913{
914 struct trace_probe *tp = (struct trace_probe *)call->data;
915
916 if (probe_is_return(tp))
917 return enable_kretprobe(&tp->rp);
918 else
919 return enable_kprobe(&tp->kp);
920}
921
922static void probe_event_disable(struct ftrace_event_call *call)
923{
924 struct trace_probe *tp = (struct trace_probe *)call->data;
925
926 if (probe_is_return(tp))
927 disable_kretprobe(&tp->rp);
928 else
929 disable_kprobe(&tp->kp);
930}
931
932static int probe_event_raw_init(struct ftrace_event_call *event_call)
933{
934 INIT_LIST_HEAD(&event_call->fields);
935 init_preds(event_call);
936 return 0;
937}
938
939#undef DEFINE_FIELD
940#define DEFINE_FIELD(type, item, name, is_signed) \
941 do { \
942 ret = trace_define_field(event_call, #type, name, \
943 offsetof(typeof(field), item), \
944 sizeof(field.item), is_signed, \
945 FILTER_OTHER); \
946 if (ret) \
947 return ret; \
948 } while (0)
949
950static int kprobe_event_define_fields(struct ftrace_event_call *event_call)
951{
952 int ret, i;
953 struct kprobe_trace_entry field;
954 char buf[MAX_ARGSTR_LEN + 1];
955 struct trace_probe *tp = (struct trace_probe *)event_call->data;
956
957 ret = trace_define_common_fields(event_call);
958 if (!ret)
959 return ret;
960
961 DEFINE_FIELD(unsigned long, ip, "ip", 0);
962 DEFINE_FIELD(int, nargs, "nargs", 1);
963 for (i = 0; i < tp->nr_args; i++) {
964 /* Set argN as a field */
965 sprintf(buf, "arg%d", i);
966 DEFINE_FIELD(unsigned long, args[i], buf, 0);
967 /* Set argument string as an alias field */
968 ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]);
969 if (ret < 0)
970 return ret;
971 DEFINE_FIELD(unsigned long, args[i], buf, 0);
972 }
973 return 0;
974}
975
976static int kretprobe_event_define_fields(struct ftrace_event_call *event_call)
977{
978 int ret, i;
979 struct kretprobe_trace_entry field;
980 char buf[MAX_ARGSTR_LEN + 1];
981 struct trace_probe *tp = (struct trace_probe *)event_call->data;
982
983 ret = trace_define_common_fields(event_call);
984 if (!ret)
985 return ret;
986
987 DEFINE_FIELD(unsigned long, func, "func", 0);
988 DEFINE_FIELD(unsigned long, ret_ip, "ret_ip", 0);
989 DEFINE_FIELD(int, nargs, "nargs", 1);
990 for (i = 0; i < tp->nr_args; i++) {
991 /* Set argN as a field */
992 sprintf(buf, "arg%d", i);
993 DEFINE_FIELD(unsigned long, args[i], buf, 0);
994 /* Set argument string as an alias field */
995 ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]);
996 if (ret < 0)
997 return ret;
998 DEFINE_FIELD(unsigned long, args[i], buf, 0);
999 }
1000 return 0;
1001}
1002
1003static int __probe_event_show_format(struct trace_seq *s,
1004 struct trace_probe *tp, const char *fmt,
1005 const char *arg)
1006{
1007 int i, ret;
1008 char buf[MAX_ARGSTR_LEN + 1];
1009
1010 /* Show aliases */
1011 for (i = 0; i < tp->nr_args; i++) {
1012 ret = trace_arg_string(buf, MAX_ARGSTR_LEN, &tp->args[i]);
1013 if (ret < 0)
1014 return ret;
1015 if (!trace_seq_printf(s, "\talias: %s;\toriginal: arg%d;\n",
1016 buf, i))
1017 return 0;
1018 }
1019 /* Show format */
1020 if (!trace_seq_printf(s, "\nprint fmt: \"%s", fmt))
1021 return 0;
1022
1023 for (i = 0; i < tp->nr_args; i++)
1024 if (!trace_seq_puts(s, " 0x%lx"))
1025 return 0;
1026
1027 if (!trace_seq_printf(s, "\", %s", arg))
1028 return 0;
1029
1030 for (i = 0; i < tp->nr_args; i++)
1031 if (!trace_seq_printf(s, ", arg%d", i))
1032 return 0;
1033
1034 return trace_seq_puts(s, "\n");
1035}
1036
1037#undef SHOW_FIELD
1038#define SHOW_FIELD(type, item, name) \
1039 do { \
1040 ret = trace_seq_printf(s, "\tfield: " #type " %s;\t" \
1041 "offset:%u;tsize:%u;\n", name, \
1042 (unsigned int)offsetof(typeof(field), item),\
1043 (unsigned int)sizeof(type)); \
1044 if (!ret) \
1045 return 0; \
1046 } while (0)
1047
1048static int kprobe_event_show_format(struct ftrace_event_call *call,
1049 struct trace_seq *s)
1050{
1051 struct kprobe_trace_entry field __attribute__((unused));
1052 int ret, i;
1053 char buf[8];
1054 struct trace_probe *tp = (struct trace_probe *)call->data;
1055
1056 SHOW_FIELD(unsigned long, ip, "ip");
1057 SHOW_FIELD(int, nargs, "nargs");
1058
1059 /* Show fields */
1060 for (i = 0; i < tp->nr_args; i++) {
1061 sprintf(buf, "arg%d", i);
1062 SHOW_FIELD(unsigned long, args[i], buf);
1063 }
1064 trace_seq_puts(s, "\n");
1065
1066 return __probe_event_show_format(s, tp, "%lx:", "ip");
1067}
1068
1069static int kretprobe_event_show_format(struct ftrace_event_call *call,
1070 struct trace_seq *s)
1071{
1072 struct kretprobe_trace_entry field __attribute__((unused));
1073 int ret, i;
1074 char buf[8];
1075 struct trace_probe *tp = (struct trace_probe *)call->data;
1076
1077 SHOW_FIELD(unsigned long, func, "func");
1078 SHOW_FIELD(unsigned long, ret_ip, "ret_ip");
1079 SHOW_FIELD(int, nargs, "nargs");
1080
1081 /* Show fields */
1082 for (i = 0; i < tp->nr_args; i++) {
1083 sprintf(buf, "arg%d", i);
1084 SHOW_FIELD(unsigned long, args[i], buf);
1085 }
1086 trace_seq_puts(s, "\n");
1087
1088 return __probe_event_show_format(s, tp, "%lx <- %lx:",
1089 "func, ret_ip");
1090}
1091
1092static int register_probe_event(struct trace_probe *tp)
1093{
1094 struct ftrace_event_call *call = &tp->call;
1095 int ret;
1096
1097 /* Initialize ftrace_event_call */
1098 call->system = "kprobes";
1099 if (probe_is_return(tp)) {
1100 call->event = &kretprobe_trace_event;
1101 call->id = TRACE_KRETPROBE;
1102 call->raw_init = probe_event_raw_init;
1103 call->show_format = kretprobe_event_show_format;
1104 call->define_fields = kretprobe_event_define_fields;
1105 } else {
1106 call->event = &kprobe_trace_event;
1107 call->id = TRACE_KPROBE;
1108 call->raw_init = probe_event_raw_init;
1109 call->show_format = kprobe_event_show_format;
1110 call->define_fields = kprobe_event_define_fields;
1111 }
1112 call->enabled = 1;
1113 call->regfunc = probe_event_enable;
1114 call->unregfunc = probe_event_disable;
1115 call->data = tp;
1116 ret = trace_add_event_call(call);
1117 if (ret)
1118 pr_info("Failed to register kprobe event: %s\n", call->name);
1119 return ret;
1120}
1121
1122static void unregister_probe_event(struct trace_probe *tp)
1123{
1124 /*
1125 * Prevent to unregister event itself because the event is shared
1126 * among other probes.
1127 */
1128 tp->call.event = NULL;
1129 trace_remove_event_call(&tp->call);
1130}
1131
1132/* Make a debugfs interface for controling probe points */
1133static __init int init_kprobe_trace(void)
1134{
1135 struct dentry *d_tracer;
1136 struct dentry *entry;
1137 int ret;
1138
1139 ret = register_ftrace_event(&kprobe_trace_event);
1140 if (!ret) {
1141 pr_warning("Could not register kprobe_trace_event type.\n");
1142 return 0;
1143 }
1144 ret = register_ftrace_event(&kretprobe_trace_event);
1145 if (!ret) {
1146 pr_warning("Could not register kretprobe_trace_event type.\n");
1147 return 0;
1148 }
1149
1150 d_tracer = tracing_init_dentry();
1151 if (!d_tracer)
1152 return 0;
1153
1154 entry = debugfs_create_file("kprobe_events", 0644, d_tracer,
1155 NULL, &kprobe_events_ops);
1156
1157 if (!entry)
1158 pr_warning("Could not create debugfs "
1159 "'kprobe_events' entry\n");
1160 return 0;
1161}
1162fs_initcall(init_kprobe_trace);
1163
1164
1165#ifdef CONFIG_FTRACE_STARTUP_TEST
1166
1167static int kprobe_trace_selftest_target(int a1, int a2, int a3,
1168 int a4, int a5, int a6)
1169{
1170 return a1 + a2 + a3 + a4 + a5 + a6;
1171}
1172
1173static __init int kprobe_trace_self_tests_init(void)
1174{
1175 int ret;
1176 int (*target)(int, int, int, int, int, int);
1177
1178 target = kprobe_trace_selftest_target;
1179
1180 pr_info("Testing kprobe tracing: ");
1181
1182 ret = command_trace_probe("p:testprobe kprobe_trace_selftest_target "
1183 "a1 a2 a3 a4 a5 a6");
1184 if (WARN_ON_ONCE(ret))
1185 pr_warning("error enabling function entry\n");
1186
1187 ret = command_trace_probe("r:testprobe2 kprobe_trace_selftest_target "
1188 "ra rv");
1189 if (WARN_ON_ONCE(ret))
1190 pr_warning("error enabling function return\n");
1191
1192 ret = target(1, 2, 3, 4, 5, 6);
1193
1194 cleanup_all_probes();
1195
1196 pr_cont("OK\n");
1197 return 0;
1198}
1199
1200late_initcall(kprobe_trace_self_tests_init);
1201
1202#endif