2 * uprobes-based tracing events
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 * Copyright (C) IBM Corporation, 2010-2012
18 * Author: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
21 #include <linux/module.h>
22 #include <linux/uaccess.h>
23 #include <linux/uprobes.h>
24 #include <linux/namei.h>
25 #include <linux/string.h>
27 #include "trace_probe.h"
29 #define UPROBE_EVENT_SYSTEM "uprobes"
32 * uprobe event core functions
35 struct uprobe_trace_consumer {
36 struct uprobe_consumer cons;
37 struct trace_uprobe *tu;
41 struct list_head list;
42 struct ftrace_event_class class;
43 struct ftrace_event_call call;
44 struct uprobe_trace_consumer *consumer;
49 unsigned int flags; /* For TP_FLAG_* */
50 ssize_t size; /* trace entry size */
52 struct probe_arg args[];
55 #define SIZEOF_TRACE_UPROBE(n) \
56 (offsetof(struct trace_uprobe, args) + \
57 (sizeof(struct probe_arg) * (n)))
59 static int register_uprobe_event(struct trace_uprobe *tu);
60 static void unregister_uprobe_event(struct trace_uprobe *tu);
62 static DEFINE_MUTEX(uprobe_lock);
63 static LIST_HEAD(uprobe_list);
65 static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs);
68 * Allocate new trace_uprobe and initialize it (including uprobes).
70 static struct trace_uprobe *
71 alloc_trace_uprobe(const char *group, const char *event, int nargs)
73 struct trace_uprobe *tu;
75 if (!event || !is_good_name(event))
76 return ERR_PTR(-EINVAL);
78 if (!group || !is_good_name(group))
79 return ERR_PTR(-EINVAL);
81 tu = kzalloc(SIZEOF_TRACE_UPROBE(nargs), GFP_KERNEL);
83 return ERR_PTR(-ENOMEM);
85 tu->call.class = &tu->class;
86 tu->call.name = kstrdup(event, GFP_KERNEL);
90 tu->class.system = kstrdup(group, GFP_KERNEL);
91 if (!tu->class.system)
94 INIT_LIST_HEAD(&tu->list);
101 return ERR_PTR(-ENOMEM);
104 static void free_trace_uprobe(struct trace_uprobe *tu)
108 for (i = 0; i < tu->nr_args; i++)
109 traceprobe_free_probe_arg(&tu->args[i]);
112 kfree(tu->call.class->system);
113 kfree(tu->call.name);
118 static struct trace_uprobe *find_probe_event(const char *event, const char *group)
120 struct trace_uprobe *tu;
122 list_for_each_entry(tu, &uprobe_list, list)
123 if (strcmp(tu->call.name, event) == 0 &&
124 strcmp(tu->call.class->system, group) == 0)
130 /* Unregister a trace_uprobe and probe_event: call with locking uprobe_lock */
131 static void unregister_trace_uprobe(struct trace_uprobe *tu)
134 unregister_uprobe_event(tu);
135 free_trace_uprobe(tu);
138 /* Register a trace_uprobe and probe_event */
139 static int register_trace_uprobe(struct trace_uprobe *tu)
141 struct trace_uprobe *old_tp;
144 mutex_lock(&uprobe_lock);
146 /* register as an event */
147 old_tp = find_probe_event(tu->call.name, tu->call.class->system);
149 /* delete old event */
150 unregister_trace_uprobe(old_tp);
152 ret = register_uprobe_event(tu);
154 pr_warning("Failed to register probe event(%d)\n", ret);
158 list_add_tail(&tu->list, &uprobe_list);
161 mutex_unlock(&uprobe_lock);
168 * - Add uprobe: p[:[GRP/]EVENT] PATH:SYMBOL[+offs] [FETCHARGS]
170 * - Remove uprobe: -:[GRP/]EVENT
172 static int create_trace_uprobe(int argc, char **argv)
174 struct trace_uprobe *tu;
176 char *arg, *event, *group, *filename;
177 char buf[MAX_EVENT_NAME_LEN];
179 unsigned long offset;
189 /* argc must be >= 1 */
190 if (argv[0][0] == '-')
192 else if (argv[0][0] != 'p') {
193 pr_info("Probe definition must be started with 'p' or '-'.\n");
197 if (argv[0][1] == ':') {
199 arg = strchr(event, '/');
206 if (strlen(group) == 0) {
207 pr_info("Group name is not specified\n");
211 if (strlen(event) == 0) {
212 pr_info("Event name is not specified\n");
217 group = UPROBE_EVENT_SYSTEM;
221 pr_info("Delete command needs an event name.\n");
224 mutex_lock(&uprobe_lock);
225 tu = find_probe_event(event, group);
228 mutex_unlock(&uprobe_lock);
229 pr_info("Event %s/%s doesn't exist.\n", group, event);
232 /* delete an event */
233 unregister_trace_uprobe(tu);
234 mutex_unlock(&uprobe_lock);
239 pr_info("Probe point is not specified.\n");
242 if (isdigit(argv[1][0])) {
243 pr_info("probe point must be have a filename.\n");
246 arg = strchr(argv[1], ':');
248 goto fail_address_parse;
252 ret = kern_path(filename, LOOKUP_FOLLOW, &path);
254 goto fail_address_parse;
256 ret = kstrtoul(arg, 0, &offset);
258 goto fail_address_parse;
260 inode = igrab(path.dentry->d_inode);
261 if (!S_ISREG(inode->i_mode)) {
263 goto fail_address_parse;
274 tail = kstrdup(kbasename(filename), GFP_KERNEL);
277 goto fail_address_parse;
280 ptr = strpbrk(tail, ".-_");
284 snprintf(buf, MAX_EVENT_NAME_LEN, "%c_%s_0x%lx", 'p', tail, offset);
289 tu = alloc_trace_uprobe(group, event, argc);
291 pr_info("Failed to allocate trace_uprobe.(%d)\n", (int)PTR_ERR(tu));
293 goto fail_address_parse;
297 tu->filename = kstrdup(filename, GFP_KERNEL);
300 pr_info("Failed to allocate filename.\n");
305 /* parse arguments */
307 for (i = 0; i < argc && i < MAX_TRACE_ARGS; i++) {
308 /* Increment count for freeing args in error case */
311 /* Parse argument name */
312 arg = strchr(argv[i], '=');
315 tu->args[i].name = kstrdup(argv[i], GFP_KERNEL);
318 /* If argument name is omitted, set "argN" */
319 snprintf(buf, MAX_EVENT_NAME_LEN, "arg%d", i + 1);
320 tu->args[i].name = kstrdup(buf, GFP_KERNEL);
323 if (!tu->args[i].name) {
324 pr_info("Failed to allocate argument[%d] name.\n", i);
329 if (!is_good_name(tu->args[i].name)) {
330 pr_info("Invalid argument[%d] name: %s\n", i, tu->args[i].name);
335 if (traceprobe_conflict_field_name(tu->args[i].name, tu->args, i)) {
336 pr_info("Argument[%d] name '%s' conflicts with "
337 "another field.\n", i, argv[i]);
342 /* Parse fetch argument */
343 ret = traceprobe_parse_probe_arg(arg, &tu->size, &tu->args[i], false, false);
345 pr_info("Parse error at argument[%d]. (%d)\n", i, ret);
350 ret = register_trace_uprobe(tu);
356 free_trace_uprobe(tu);
363 pr_info("Failed to parse address or file.\n");
368 static void cleanup_all_probes(void)
370 struct trace_uprobe *tu;
372 mutex_lock(&uprobe_lock);
373 while (!list_empty(&uprobe_list)) {
374 tu = list_entry(uprobe_list.next, struct trace_uprobe, list);
375 unregister_trace_uprobe(tu);
377 mutex_unlock(&uprobe_lock);
380 /* Probes listing interfaces */
381 static void *probes_seq_start(struct seq_file *m, loff_t *pos)
383 mutex_lock(&uprobe_lock);
384 return seq_list_start(&uprobe_list, *pos);
387 static void *probes_seq_next(struct seq_file *m, void *v, loff_t *pos)
389 return seq_list_next(v, &uprobe_list, pos);
392 static void probes_seq_stop(struct seq_file *m, void *v)
394 mutex_unlock(&uprobe_lock);
397 static int probes_seq_show(struct seq_file *m, void *v)
399 struct trace_uprobe *tu = v;
402 seq_printf(m, "p:%s/%s", tu->call.class->system, tu->call.name);
403 seq_printf(m, " %s:0x%p", tu->filename, (void *)tu->offset);
405 for (i = 0; i < tu->nr_args; i++)
406 seq_printf(m, " %s=%s", tu->args[i].name, tu->args[i].comm);
412 static const struct seq_operations probes_seq_op = {
413 .start = probes_seq_start,
414 .next = probes_seq_next,
415 .stop = probes_seq_stop,
416 .show = probes_seq_show
419 static int probes_open(struct inode *inode, struct file *file)
421 if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC))
422 cleanup_all_probes();
424 return seq_open(file, &probes_seq_op);
427 static ssize_t probes_write(struct file *file, const char __user *buffer,
428 size_t count, loff_t *ppos)
430 return traceprobe_probes_write(file, buffer, count, ppos, create_trace_uprobe);
433 static const struct file_operations uprobe_events_ops = {
434 .owner = THIS_MODULE,
438 .release = seq_release,
439 .write = probes_write,
442 /* Probes profiling interfaces */
443 static int probes_profile_seq_show(struct seq_file *m, void *v)
445 struct trace_uprobe *tu = v;
447 seq_printf(m, " %s %-44s %15lu\n", tu->filename, tu->call.name, tu->nhit);
451 static const struct seq_operations profile_seq_op = {
452 .start = probes_seq_start,
453 .next = probes_seq_next,
454 .stop = probes_seq_stop,
455 .show = probes_profile_seq_show
458 static int profile_open(struct inode *inode, struct file *file)
460 return seq_open(file, &profile_seq_op);
463 static const struct file_operations uprobe_profile_ops = {
464 .owner = THIS_MODULE,
465 .open = profile_open,
468 .release = seq_release,
472 static void uprobe_trace_func(struct trace_uprobe *tu, struct pt_regs *regs)
474 struct uprobe_trace_entry_head *entry;
475 struct ring_buffer_event *event;
476 struct ring_buffer *buffer;
479 unsigned long irq_flags;
480 struct ftrace_event_call *call = &tu->call;
484 local_save_flags(irq_flags);
485 pc = preempt_count();
487 size = sizeof(*entry) + tu->size;
489 event = trace_current_buffer_lock_reserve(&buffer, call->event.type,
490 size, irq_flags, pc);
494 entry = ring_buffer_event_data(event);
495 entry->ip = uprobe_get_swbp_addr(task_pt_regs(current));
496 data = (u8 *)&entry[1];
497 for (i = 0; i < tu->nr_args; i++)
498 call_fetch(&tu->args[i].fetch, regs, data + tu->args[i].offset);
500 if (!filter_current_check_discard(buffer, call, entry, event))
501 trace_buffer_unlock_commit(buffer, event, irq_flags, pc);
504 /* Event entry printers */
505 static enum print_line_t
506 print_uprobe_event(struct trace_iterator *iter, int flags, struct trace_event *event)
508 struct uprobe_trace_entry_head *field;
509 struct trace_seq *s = &iter->seq;
510 struct trace_uprobe *tu;
514 field = (struct uprobe_trace_entry_head *)iter->ent;
515 tu = container_of(event, struct trace_uprobe, call.event);
517 if (!trace_seq_printf(s, "%s: (", tu->call.name))
520 if (!seq_print_ip_sym(s, field->ip, flags | TRACE_ITER_SYM_OFFSET))
523 if (!trace_seq_puts(s, ")"))
526 data = (u8 *)&field[1];
527 for (i = 0; i < tu->nr_args; i++) {
528 if (!tu->args[i].type->print(s, tu->args[i].name,
529 data + tu->args[i].offset, field))
533 if (trace_seq_puts(s, "\n"))
534 return TRACE_TYPE_HANDLED;
537 return TRACE_TYPE_PARTIAL_LINE;
540 static int probe_event_enable(struct trace_uprobe *tu, int flag)
542 struct uprobe_trace_consumer *utc;
545 if (!tu->inode || tu->consumer)
548 utc = kzalloc(sizeof(struct uprobe_trace_consumer), GFP_KERNEL);
552 utc->cons.handler = uprobe_dispatcher;
553 ret = uprobe_register(tu->inode, tu->offset, &utc->cons);
566 static void probe_event_disable(struct trace_uprobe *tu, int flag)
568 if (!tu->inode || !tu->consumer)
571 uprobe_unregister(tu->inode, tu->offset, &tu->consumer->cons);
577 static int uprobe_event_define_fields(struct ftrace_event_call *event_call)
580 struct uprobe_trace_entry_head field;
581 struct trace_uprobe *tu = (struct trace_uprobe *)event_call->data;
583 DEFINE_FIELD(unsigned long, ip, FIELD_STRING_IP, 0);
584 /* Set argument names as fields */
585 for (i = 0; i < tu->nr_args; i++) {
586 ret = trace_define_field(event_call, tu->args[i].type->fmttype,
588 sizeof(field) + tu->args[i].offset,
589 tu->args[i].type->size,
590 tu->args[i].type->is_signed,
599 #define LEN_OR_ZERO (len ? len - pos : 0)
600 static int __set_print_fmt(struct trace_uprobe *tu, char *buf, int len)
602 const char *fmt, *arg;
607 arg = "REC->" FIELD_STRING_IP;
609 /* When len=0, we just calculate the needed length */
611 pos += snprintf(buf + pos, LEN_OR_ZERO, "\"%s", fmt);
613 for (i = 0; i < tu->nr_args; i++) {
614 pos += snprintf(buf + pos, LEN_OR_ZERO, " %s=%s",
615 tu->args[i].name, tu->args[i].type->fmt);
618 pos += snprintf(buf + pos, LEN_OR_ZERO, "\", %s", arg);
620 for (i = 0; i < tu->nr_args; i++) {
621 pos += snprintf(buf + pos, LEN_OR_ZERO, ", REC->%s",
625 return pos; /* return the length of print_fmt */
629 static int set_print_fmt(struct trace_uprobe *tu)
634 /* First: called with 0 length to calculate the needed length */
635 len = __set_print_fmt(tu, NULL, 0);
636 print_fmt = kmalloc(len + 1, GFP_KERNEL);
640 /* Second: actually write the @print_fmt */
641 __set_print_fmt(tu, print_fmt, len + 1);
642 tu->call.print_fmt = print_fmt;
647 #ifdef CONFIG_PERF_EVENTS
648 /* uprobe profile handler */
649 static void uprobe_perf_func(struct trace_uprobe *tu, struct pt_regs *regs)
651 struct ftrace_event_call *call = &tu->call;
652 struct uprobe_trace_entry_head *entry;
653 struct hlist_head *head;
658 __size = sizeof(*entry) + tu->size;
659 size = ALIGN(__size + sizeof(u32), sizeof(u64));
661 if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE, "profile buffer not large enough"))
666 entry = perf_trace_buf_prepare(size, call->event.type, regs, &rctx);
670 entry->ip = uprobe_get_swbp_addr(task_pt_regs(current));
671 data = (u8 *)&entry[1];
672 for (i = 0; i < tu->nr_args; i++)
673 call_fetch(&tu->args[i].fetch, regs, data + tu->args[i].offset);
675 head = this_cpu_ptr(call->perf_events);
676 perf_trace_buf_submit(entry, size, rctx, entry->ip, 1, regs, head, NULL);
681 #endif /* CONFIG_PERF_EVENTS */
684 int trace_uprobe_register(struct ftrace_event_call *event, enum trace_reg type, void *data)
686 struct trace_uprobe *tu = (struct trace_uprobe *)event->data;
689 case TRACE_REG_REGISTER:
690 return probe_event_enable(tu, TP_FLAG_TRACE);
692 case TRACE_REG_UNREGISTER:
693 probe_event_disable(tu, TP_FLAG_TRACE);
696 #ifdef CONFIG_PERF_EVENTS
697 case TRACE_REG_PERF_REGISTER:
698 return probe_event_enable(tu, TP_FLAG_PROFILE);
700 case TRACE_REG_PERF_UNREGISTER:
701 probe_event_disable(tu, TP_FLAG_PROFILE);
710 static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
712 struct uprobe_trace_consumer *utc;
713 struct trace_uprobe *tu;
715 utc = container_of(con, struct uprobe_trace_consumer, cons);
717 if (!tu || tu->consumer != utc)
720 if (tu->flags & TP_FLAG_TRACE)
721 uprobe_trace_func(tu, regs);
723 #ifdef CONFIG_PERF_EVENTS
724 if (tu->flags & TP_FLAG_PROFILE)
725 uprobe_perf_func(tu, regs);
730 static struct trace_event_functions uprobe_funcs = {
731 .trace = print_uprobe_event
734 static int register_uprobe_event(struct trace_uprobe *tu)
736 struct ftrace_event_call *call = &tu->call;
739 /* Initialize ftrace_event_call */
740 INIT_LIST_HEAD(&call->class->fields);
741 call->event.funcs = &uprobe_funcs;
742 call->class->define_fields = uprobe_event_define_fields;
744 if (set_print_fmt(tu) < 0)
747 ret = register_ftrace_event(&call->event);
749 kfree(call->print_fmt);
753 call->class->reg = trace_uprobe_register;
755 ret = trace_add_event_call(call);
758 pr_info("Failed to register uprobe event: %s\n", call->name);
759 kfree(call->print_fmt);
760 unregister_ftrace_event(&call->event);
766 static void unregister_uprobe_event(struct trace_uprobe *tu)
768 /* tu->event is unregistered in trace_remove_event_call() */
769 trace_remove_event_call(&tu->call);
770 kfree(tu->call.print_fmt);
771 tu->call.print_fmt = NULL;
774 /* Make a trace interface for controling probe points */
775 static __init int init_uprobe_trace(void)
777 struct dentry *d_tracer;
779 d_tracer = tracing_init_dentry();
783 trace_create_file("uprobe_events", 0644, d_tracer,
784 NULL, &uprobe_events_ops);
785 /* Profile interface */
786 trace_create_file("uprobe_profile", 0444, d_tracer,
787 NULL, &uprobe_profile_ops);
791 fs_initcall(init_uprobe_trace);