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"
31 struct uprobe_trace_entry_head {
32 struct trace_entry ent;
33 unsigned long vaddr[];
36 #define SIZEOF_TRACE_ENTRY(is_return) \
37 (sizeof(struct uprobe_trace_entry_head) + \
38 sizeof(unsigned long) * (is_return ? 2 : 1))
40 #define DATAOF_TRACE_ENTRY(entry, is_return) \
41 ((void*)(entry) + SIZEOF_TRACE_ENTRY(is_return))
43 struct trace_uprobe_filter {
46 struct list_head perf_events;
50 * uprobe event core functions
53 struct list_head list;
54 struct trace_uprobe_filter filter;
55 struct uprobe_consumer consumer;
60 struct trace_probe tp;
63 #define SIZEOF_TRACE_UPROBE(n) \
64 (offsetof(struct trace_uprobe, tp.args) + \
65 (sizeof(struct probe_arg) * (n)))
67 static int register_uprobe_event(struct trace_uprobe *tu);
68 static int unregister_uprobe_event(struct trace_uprobe *tu);
70 static DEFINE_MUTEX(uprobe_lock);
71 static LIST_HEAD(uprobe_list);
73 struct uprobe_dispatch_data {
74 struct trace_uprobe *tu;
75 unsigned long bp_addr;
78 static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs);
79 static int uretprobe_dispatcher(struct uprobe_consumer *con,
80 unsigned long func, struct pt_regs *regs);
82 #ifdef CONFIG_STACK_GROWSUP
83 static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
85 return addr - (n * sizeof(long));
88 static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
90 return addr + (n * sizeof(long));
94 static unsigned long get_user_stack_nth(struct pt_regs *regs, unsigned int n)
97 unsigned long addr = user_stack_pointer(regs);
99 addr = adjust_stack_addr(addr, n);
101 if (copy_from_user(&ret, (void __force __user *) addr, sizeof(ret)))
108 * Uprobes-specific fetch functions
110 #define DEFINE_FETCH_stack(type) \
111 static void FETCH_FUNC_NAME(stack, type)(struct pt_regs *regs, \
112 void *offset, void *dest) \
114 *(type *)dest = (type)get_user_stack_nth(regs, \
115 ((unsigned long)offset)); \
117 DEFINE_BASIC_FETCH_FUNCS(stack)
118 /* No string on the stack entry */
119 #define fetch_stack_string NULL
120 #define fetch_stack_string_size NULL
122 #define DEFINE_FETCH_memory(type) \
123 static void FETCH_FUNC_NAME(memory, type)(struct pt_regs *regs, \
124 void *addr, void *dest) \
127 void __user *vaddr = (void __force __user *) addr; \
129 if (copy_from_user(&retval, vaddr, sizeof(type))) \
132 *(type *) dest = retval; \
134 DEFINE_BASIC_FETCH_FUNCS(memory)
136 * Fetch a null-terminated string. Caller MUST set *(u32 *)dest with max
137 * length and relative data location.
139 static void FETCH_FUNC_NAME(memory, string)(struct pt_regs *regs,
140 void *addr, void *dest)
143 u32 rloc = *(u32 *)dest;
144 int maxlen = get_rloc_len(rloc);
145 u8 *dst = get_rloc_data(dest);
146 void __user *src = (void __force __user *) addr;
151 ret = strncpy_from_user(dst, src, maxlen);
156 * Include the terminating null byte. In this case it
157 * was copied by strncpy_from_user but not accounted
162 if (ret < 0) { /* Failed to fetch string */
163 ((u8 *)get_rloc_data(dest))[0] = '\0';
164 *(u32 *)dest = make_data_rloc(0, get_rloc_offs(rloc));
166 *(u32 *)dest = make_data_rloc(ret, get_rloc_offs(rloc));
170 static void FETCH_FUNC_NAME(memory, string_size)(struct pt_regs *regs,
171 void *addr, void *dest)
174 void __user *vaddr = (void __force __user *) addr;
176 len = strnlen_user(vaddr, MAX_STRING_SIZE);
178 if (len == 0 || len > MAX_STRING_SIZE) /* Failed to check length */
184 static unsigned long translate_user_vaddr(void *file_offset)
186 unsigned long base_addr;
187 struct uprobe_dispatch_data *udd;
189 udd = (void *) current->utask->vaddr;
191 base_addr = udd->bp_addr - udd->tu->offset;
192 return base_addr + (unsigned long)file_offset;
195 #define DEFINE_FETCH_file_offset(type) \
196 static void FETCH_FUNC_NAME(file_offset, type)(struct pt_regs *regs, \
197 void *offset, void *dest)\
199 void *vaddr = (void *)translate_user_vaddr(offset); \
201 FETCH_FUNC_NAME(memory, type)(regs, vaddr, dest); \
203 DEFINE_BASIC_FETCH_FUNCS(file_offset)
204 DEFINE_FETCH_file_offset(string)
205 DEFINE_FETCH_file_offset(string_size)
207 /* Fetch type information table */
208 static const struct fetch_type uprobes_fetch_type_table[] = {
210 [FETCH_TYPE_STRING] = __ASSIGN_FETCH_TYPE("string", string, string,
211 sizeof(u32), 1, "__data_loc char[]"),
212 [FETCH_TYPE_STRSIZE] = __ASSIGN_FETCH_TYPE("string_size", u32,
213 string_size, sizeof(u32), 0, "u32"),
215 ASSIGN_FETCH_TYPE(u8, u8, 0),
216 ASSIGN_FETCH_TYPE(u16, u16, 0),
217 ASSIGN_FETCH_TYPE(u32, u32, 0),
218 ASSIGN_FETCH_TYPE(u64, u64, 0),
219 ASSIGN_FETCH_TYPE(s8, u8, 1),
220 ASSIGN_FETCH_TYPE(s16, u16, 1),
221 ASSIGN_FETCH_TYPE(s32, u32, 1),
222 ASSIGN_FETCH_TYPE(s64, u64, 1),
224 ASSIGN_FETCH_TYPE_END
227 static inline void init_trace_uprobe_filter(struct trace_uprobe_filter *filter)
229 rwlock_init(&filter->rwlock);
230 filter->nr_systemwide = 0;
231 INIT_LIST_HEAD(&filter->perf_events);
234 static inline bool uprobe_filter_is_empty(struct trace_uprobe_filter *filter)
236 return !filter->nr_systemwide && list_empty(&filter->perf_events);
239 static inline bool is_ret_probe(struct trace_uprobe *tu)
241 return tu->consumer.ret_handler != NULL;
245 * Allocate new trace_uprobe and initialize it (including uprobes).
247 static struct trace_uprobe *
248 alloc_trace_uprobe(const char *group, const char *event, int nargs, bool is_ret)
250 struct trace_uprobe *tu;
252 if (!event || !is_good_name(event))
253 return ERR_PTR(-EINVAL);
255 if (!group || !is_good_name(group))
256 return ERR_PTR(-EINVAL);
258 tu = kzalloc(SIZEOF_TRACE_UPROBE(nargs), GFP_KERNEL);
260 return ERR_PTR(-ENOMEM);
262 tu->tp.call.class = &tu->tp.class;
263 tu->tp.call.name = kstrdup(event, GFP_KERNEL);
264 if (!tu->tp.call.name)
267 tu->tp.class.system = kstrdup(group, GFP_KERNEL);
268 if (!tu->tp.class.system)
271 INIT_LIST_HEAD(&tu->list);
272 INIT_LIST_HEAD(&tu->tp.files);
273 tu->consumer.handler = uprobe_dispatcher;
275 tu->consumer.ret_handler = uretprobe_dispatcher;
276 init_trace_uprobe_filter(&tu->filter);
280 kfree(tu->tp.call.name);
283 return ERR_PTR(-ENOMEM);
286 static void free_trace_uprobe(struct trace_uprobe *tu)
290 for (i = 0; i < tu->tp.nr_args; i++)
291 traceprobe_free_probe_arg(&tu->tp.args[i]);
294 kfree(tu->tp.call.class->system);
295 kfree(tu->tp.call.name);
300 static struct trace_uprobe *find_probe_event(const char *event, const char *group)
302 struct trace_uprobe *tu;
304 list_for_each_entry(tu, &uprobe_list, list)
305 if (strcmp(trace_event_name(&tu->tp.call), event) == 0 &&
306 strcmp(tu->tp.call.class->system, group) == 0)
312 /* Unregister a trace_uprobe and probe_event: call with locking uprobe_lock */
313 static int unregister_trace_uprobe(struct trace_uprobe *tu)
317 ret = unregister_uprobe_event(tu);
322 free_trace_uprobe(tu);
326 /* Register a trace_uprobe and probe_event */
327 static int register_trace_uprobe(struct trace_uprobe *tu)
329 struct trace_uprobe *old_tu;
332 mutex_lock(&uprobe_lock);
334 /* register as an event */
335 old_tu = find_probe_event(trace_event_name(&tu->tp.call),
336 tu->tp.call.class->system);
338 /* delete old event */
339 ret = unregister_trace_uprobe(old_tu);
344 ret = register_uprobe_event(tu);
346 pr_warning("Failed to register probe event(%d)\n", ret);
350 list_add_tail(&tu->list, &uprobe_list);
353 mutex_unlock(&uprobe_lock);
360 * - Add uprobe: p|r[:[GRP/]EVENT] PATH:OFFSET [FETCHARGS]
362 * - Remove uprobe: -:[GRP/]EVENT
364 static int create_trace_uprobe(int argc, char **argv)
366 struct trace_uprobe *tu;
368 char *arg, *event, *group, *filename;
369 char buf[MAX_EVENT_NAME_LEN];
371 unsigned long offset;
372 bool is_delete, is_return;
382 /* argc must be >= 1 */
383 if (argv[0][0] == '-')
385 else if (argv[0][0] == 'r')
387 else if (argv[0][0] != 'p') {
388 pr_info("Probe definition must be started with 'p', 'r' or '-'.\n");
392 if (argv[0][1] == ':') {
394 arg = strchr(event, '/');
401 if (strlen(group) == 0) {
402 pr_info("Group name is not specified\n");
406 if (strlen(event) == 0) {
407 pr_info("Event name is not specified\n");
412 group = UPROBE_EVENT_SYSTEM;
418 pr_info("Delete command needs an event name.\n");
421 mutex_lock(&uprobe_lock);
422 tu = find_probe_event(event, group);
425 mutex_unlock(&uprobe_lock);
426 pr_info("Event %s/%s doesn't exist.\n", group, event);
429 /* delete an event */
430 ret = unregister_trace_uprobe(tu);
431 mutex_unlock(&uprobe_lock);
436 pr_info("Probe point is not specified.\n");
439 if (isdigit(argv[1][0])) {
440 pr_info("probe point must be have a filename.\n");
443 arg = strchr(argv[1], ':');
446 goto fail_address_parse;
451 ret = kern_path(filename, LOOKUP_FOLLOW, &path);
453 goto fail_address_parse;
455 inode = igrab(d_inode(path.dentry));
458 if (!inode || !S_ISREG(inode->i_mode)) {
460 goto fail_address_parse;
463 ret = kstrtoul(arg, 0, &offset);
465 goto fail_address_parse;
475 tail = kstrdup(kbasename(filename), GFP_KERNEL);
478 goto fail_address_parse;
481 ptr = strpbrk(tail, ".-_");
485 snprintf(buf, MAX_EVENT_NAME_LEN, "%c_%s_0x%lx", 'p', tail, offset);
490 tu = alloc_trace_uprobe(group, event, argc, is_return);
492 pr_info("Failed to allocate trace_uprobe.(%d)\n", (int)PTR_ERR(tu));
494 goto fail_address_parse;
498 tu->filename = kstrdup(filename, GFP_KERNEL);
501 pr_info("Failed to allocate filename.\n");
506 /* parse arguments */
508 for (i = 0; i < argc && i < MAX_TRACE_ARGS; i++) {
509 struct probe_arg *parg = &tu->tp.args[i];
511 /* Increment count for freeing args in error case */
514 /* Parse argument name */
515 arg = strchr(argv[i], '=');
518 parg->name = kstrdup(argv[i], GFP_KERNEL);
521 /* If argument name is omitted, set "argN" */
522 snprintf(buf, MAX_EVENT_NAME_LEN, "arg%d", i + 1);
523 parg->name = kstrdup(buf, GFP_KERNEL);
527 pr_info("Failed to allocate argument[%d] name.\n", i);
532 if (!is_good_name(parg->name)) {
533 pr_info("Invalid argument[%d] name: %s\n", i, parg->name);
538 if (traceprobe_conflict_field_name(parg->name, tu->tp.args, i)) {
539 pr_info("Argument[%d] name '%s' conflicts with "
540 "another field.\n", i, argv[i]);
545 /* Parse fetch argument */
546 ret = traceprobe_parse_probe_arg(arg, &tu->tp.size, parg,
548 uprobes_fetch_type_table);
550 pr_info("Parse error at argument[%d]. (%d)\n", i, ret);
555 ret = register_trace_uprobe(tu);
561 free_trace_uprobe(tu);
567 pr_info("Failed to parse address or file.\n");
572 static int cleanup_all_probes(void)
574 struct trace_uprobe *tu;
577 mutex_lock(&uprobe_lock);
578 while (!list_empty(&uprobe_list)) {
579 tu = list_entry(uprobe_list.next, struct trace_uprobe, list);
580 ret = unregister_trace_uprobe(tu);
584 mutex_unlock(&uprobe_lock);
588 /* Probes listing interfaces */
589 static void *probes_seq_start(struct seq_file *m, loff_t *pos)
591 mutex_lock(&uprobe_lock);
592 return seq_list_start(&uprobe_list, *pos);
595 static void *probes_seq_next(struct seq_file *m, void *v, loff_t *pos)
597 return seq_list_next(v, &uprobe_list, pos);
600 static void probes_seq_stop(struct seq_file *m, void *v)
602 mutex_unlock(&uprobe_lock);
605 static int probes_seq_show(struct seq_file *m, void *v)
607 struct trace_uprobe *tu = v;
608 char c = is_ret_probe(tu) ? 'r' : 'p';
611 seq_printf(m, "%c:%s/%s", c, tu->tp.call.class->system,
612 trace_event_name(&tu->tp.call));
613 seq_printf(m, " %s:", tu->filename);
615 /* Don't print "0x (null)" when offset is 0 */
617 seq_printf(m, "0x%p", (void *)tu->offset);
619 switch (sizeof(void *)) {
621 seq_printf(m, "0x00000000");
625 seq_printf(m, "0x0000000000000000");
630 for (i = 0; i < tu->tp.nr_args; i++)
631 seq_printf(m, " %s=%s", tu->tp.args[i].name, tu->tp.args[i].comm);
637 static const struct seq_operations probes_seq_op = {
638 .start = probes_seq_start,
639 .next = probes_seq_next,
640 .stop = probes_seq_stop,
641 .show = probes_seq_show
644 static int probes_open(struct inode *inode, struct file *file)
648 if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
649 ret = cleanup_all_probes();
654 return seq_open(file, &probes_seq_op);
657 static ssize_t probes_write(struct file *file, const char __user *buffer,
658 size_t count, loff_t *ppos)
660 return traceprobe_probes_write(file, buffer, count, ppos, create_trace_uprobe);
663 static const struct file_operations uprobe_events_ops = {
664 .owner = THIS_MODULE,
668 .release = seq_release,
669 .write = probes_write,
672 /* Probes profiling interfaces */
673 static int probes_profile_seq_show(struct seq_file *m, void *v)
675 struct trace_uprobe *tu = v;
677 seq_printf(m, " %s %-44s %15lu\n", tu->filename,
678 trace_event_name(&tu->tp.call), tu->nhit);
682 static const struct seq_operations profile_seq_op = {
683 .start = probes_seq_start,
684 .next = probes_seq_next,
685 .stop = probes_seq_stop,
686 .show = probes_profile_seq_show
689 static int profile_open(struct inode *inode, struct file *file)
691 return seq_open(file, &profile_seq_op);
694 static const struct file_operations uprobe_profile_ops = {
695 .owner = THIS_MODULE,
696 .open = profile_open,
699 .release = seq_release,
702 struct uprobe_cpu_buffer {
706 static struct uprobe_cpu_buffer __percpu *uprobe_cpu_buffer;
707 static int uprobe_buffer_refcnt;
709 static int uprobe_buffer_init(void)
713 uprobe_cpu_buffer = alloc_percpu(struct uprobe_cpu_buffer);
714 if (uprobe_cpu_buffer == NULL)
717 for_each_possible_cpu(cpu) {
718 struct page *p = alloc_pages_node(cpu_to_node(cpu),
724 per_cpu_ptr(uprobe_cpu_buffer, cpu)->buf = page_address(p);
725 mutex_init(&per_cpu_ptr(uprobe_cpu_buffer, cpu)->mutex);
731 for_each_possible_cpu(cpu) {
734 free_page((unsigned long)per_cpu_ptr(uprobe_cpu_buffer, cpu)->buf);
737 free_percpu(uprobe_cpu_buffer);
741 static int uprobe_buffer_enable(void)
745 BUG_ON(!mutex_is_locked(&event_mutex));
747 if (uprobe_buffer_refcnt++ == 0) {
748 ret = uprobe_buffer_init();
750 uprobe_buffer_refcnt--;
756 static void uprobe_buffer_disable(void)
760 BUG_ON(!mutex_is_locked(&event_mutex));
762 if (--uprobe_buffer_refcnt == 0) {
763 for_each_possible_cpu(cpu)
764 free_page((unsigned long)per_cpu_ptr(uprobe_cpu_buffer,
767 free_percpu(uprobe_cpu_buffer);
768 uprobe_cpu_buffer = NULL;
772 static struct uprobe_cpu_buffer *uprobe_buffer_get(void)
774 struct uprobe_cpu_buffer *ucb;
777 cpu = raw_smp_processor_id();
778 ucb = per_cpu_ptr(uprobe_cpu_buffer, cpu);
781 * Use per-cpu buffers for fastest access, but we might migrate
782 * so the mutex makes sure we have sole access to it.
784 mutex_lock(&ucb->mutex);
789 static void uprobe_buffer_put(struct uprobe_cpu_buffer *ucb)
791 mutex_unlock(&ucb->mutex);
794 static void __uprobe_trace_func(struct trace_uprobe *tu,
795 unsigned long func, struct pt_regs *regs,
796 struct uprobe_cpu_buffer *ucb, int dsize,
797 struct trace_event_file *trace_file)
799 struct uprobe_trace_entry_head *entry;
800 struct ring_buffer_event *event;
801 struct ring_buffer *buffer;
804 struct trace_event_call *call = &tu->tp.call;
806 WARN_ON(call != trace_file->event_call);
808 if (WARN_ON_ONCE(tu->tp.size + dsize > PAGE_SIZE))
811 if (trace_trigger_soft_disabled(trace_file))
814 esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
815 size = esize + tu->tp.size + dsize;
816 event = trace_event_buffer_lock_reserve(&buffer, trace_file,
817 call->event.type, size, 0, 0);
821 entry = ring_buffer_event_data(event);
822 if (is_ret_probe(tu)) {
823 entry->vaddr[0] = func;
824 entry->vaddr[1] = instruction_pointer(regs);
825 data = DATAOF_TRACE_ENTRY(entry, true);
827 entry->vaddr[0] = instruction_pointer(regs);
828 data = DATAOF_TRACE_ENTRY(entry, false);
831 memcpy(data, ucb->buf, tu->tp.size + dsize);
833 event_trigger_unlock_commit(trace_file, buffer, event, entry, 0, 0);
837 static int uprobe_trace_func(struct trace_uprobe *tu, struct pt_regs *regs,
838 struct uprobe_cpu_buffer *ucb, int dsize)
840 struct event_file_link *link;
842 if (is_ret_probe(tu))
846 list_for_each_entry_rcu(link, &tu->tp.files, list)
847 __uprobe_trace_func(tu, 0, regs, ucb, dsize, link->file);
853 static void uretprobe_trace_func(struct trace_uprobe *tu, unsigned long func,
854 struct pt_regs *regs,
855 struct uprobe_cpu_buffer *ucb, int dsize)
857 struct event_file_link *link;
860 list_for_each_entry_rcu(link, &tu->tp.files, list)
861 __uprobe_trace_func(tu, func, regs, ucb, dsize, link->file);
865 /* Event entry printers */
866 static enum print_line_t
867 print_uprobe_event(struct trace_iterator *iter, int flags, struct trace_event *event)
869 struct uprobe_trace_entry_head *entry;
870 struct trace_seq *s = &iter->seq;
871 struct trace_uprobe *tu;
875 entry = (struct uprobe_trace_entry_head *)iter->ent;
876 tu = container_of(event, struct trace_uprobe, tp.call.event);
878 if (is_ret_probe(tu)) {
879 trace_seq_printf(s, "%s: (0x%lx <- 0x%lx)",
880 trace_event_name(&tu->tp.call),
881 entry->vaddr[1], entry->vaddr[0]);
882 data = DATAOF_TRACE_ENTRY(entry, true);
884 trace_seq_printf(s, "%s: (0x%lx)",
885 trace_event_name(&tu->tp.call),
887 data = DATAOF_TRACE_ENTRY(entry, false);
890 for (i = 0; i < tu->tp.nr_args; i++) {
891 struct probe_arg *parg = &tu->tp.args[i];
893 if (!parg->type->print(s, parg->name, data + parg->offset, entry))
897 trace_seq_putc(s, '\n');
900 return trace_handle_return(s);
903 typedef bool (*filter_func_t)(struct uprobe_consumer *self,
904 enum uprobe_filter_ctx ctx,
905 struct mm_struct *mm);
908 probe_event_enable(struct trace_uprobe *tu, struct trace_event_file *file,
909 filter_func_t filter)
911 bool enabled = trace_probe_is_enabled(&tu->tp);
912 struct event_file_link *link = NULL;
916 if (tu->tp.flags & TP_FLAG_PROFILE)
919 link = kmalloc(sizeof(*link), GFP_KERNEL);
924 list_add_tail_rcu(&link->list, &tu->tp.files);
926 tu->tp.flags |= TP_FLAG_TRACE;
928 if (tu->tp.flags & TP_FLAG_TRACE)
931 tu->tp.flags |= TP_FLAG_PROFILE;
934 WARN_ON(!uprobe_filter_is_empty(&tu->filter));
939 ret = uprobe_buffer_enable();
943 tu->consumer.filter = filter;
944 ret = uprobe_register(tu->inode, tu->offset, &tu->consumer);
951 uprobe_buffer_disable();
955 list_del(&link->list);
957 tu->tp.flags &= ~TP_FLAG_TRACE;
959 tu->tp.flags &= ~TP_FLAG_PROFILE;
965 probe_event_disable(struct trace_uprobe *tu, struct trace_event_file *file)
967 if (!trace_probe_is_enabled(&tu->tp))
971 struct event_file_link *link;
973 link = find_event_file_link(&tu->tp, file);
977 list_del_rcu(&link->list);
978 /* synchronize with u{,ret}probe_trace_func */
982 if (!list_empty(&tu->tp.files))
986 WARN_ON(!uprobe_filter_is_empty(&tu->filter));
988 uprobe_unregister(tu->inode, tu->offset, &tu->consumer);
989 tu->tp.flags &= file ? ~TP_FLAG_TRACE : ~TP_FLAG_PROFILE;
991 uprobe_buffer_disable();
994 static int uprobe_event_define_fields(struct trace_event_call *event_call)
997 struct uprobe_trace_entry_head field;
998 struct trace_uprobe *tu = event_call->data;
1000 if (is_ret_probe(tu)) {
1001 DEFINE_FIELD(unsigned long, vaddr[0], FIELD_STRING_FUNC, 0);
1002 DEFINE_FIELD(unsigned long, vaddr[1], FIELD_STRING_RETIP, 0);
1003 size = SIZEOF_TRACE_ENTRY(true);
1005 DEFINE_FIELD(unsigned long, vaddr[0], FIELD_STRING_IP, 0);
1006 size = SIZEOF_TRACE_ENTRY(false);
1008 /* Set argument names as fields */
1009 for (i = 0; i < tu->tp.nr_args; i++) {
1010 struct probe_arg *parg = &tu->tp.args[i];
1012 ret = trace_define_field(event_call, parg->type->fmttype,
1013 parg->name, size + parg->offset,
1014 parg->type->size, parg->type->is_signed,
1023 #ifdef CONFIG_PERF_EVENTS
1025 __uprobe_perf_filter(struct trace_uprobe_filter *filter, struct mm_struct *mm)
1027 struct perf_event *event;
1029 if (filter->nr_systemwide)
1032 list_for_each_entry(event, &filter->perf_events, hw.tp_list) {
1033 if (event->hw.target->mm == mm)
1041 uprobe_filter_event(struct trace_uprobe *tu, struct perf_event *event)
1043 return __uprobe_perf_filter(&tu->filter, event->hw.target->mm);
1046 static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
1050 write_lock(&tu->filter.rwlock);
1051 if (event->hw.target) {
1052 list_del(&event->hw.tp_list);
1053 done = tu->filter.nr_systemwide ||
1054 (event->hw.target->flags & PF_EXITING) ||
1055 uprobe_filter_event(tu, event);
1057 tu->filter.nr_systemwide--;
1058 done = tu->filter.nr_systemwide;
1060 write_unlock(&tu->filter.rwlock);
1063 return uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
1068 static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
1073 write_lock(&tu->filter.rwlock);
1074 if (event->hw.target) {
1076 * event->parent != NULL means copy_process(), we can avoid
1077 * uprobe_apply(). current->mm must be probed and we can rely
1078 * on dup_mmap() which preserves the already installed bp's.
1080 * attr.enable_on_exec means that exec/mmap will install the
1081 * breakpoints we need.
1083 done = tu->filter.nr_systemwide ||
1084 event->parent || event->attr.enable_on_exec ||
1085 uprobe_filter_event(tu, event);
1086 list_add(&event->hw.tp_list, &tu->filter.perf_events);
1088 done = tu->filter.nr_systemwide;
1089 tu->filter.nr_systemwide++;
1091 write_unlock(&tu->filter.rwlock);
1095 err = uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
1097 uprobe_perf_close(tu, event);
1102 static bool uprobe_perf_filter(struct uprobe_consumer *uc,
1103 enum uprobe_filter_ctx ctx, struct mm_struct *mm)
1105 struct trace_uprobe *tu;
1108 tu = container_of(uc, struct trace_uprobe, consumer);
1109 read_lock(&tu->filter.rwlock);
1110 ret = __uprobe_perf_filter(&tu->filter, mm);
1111 read_unlock(&tu->filter.rwlock);
1116 static void __uprobe_perf_func(struct trace_uprobe *tu,
1117 unsigned long func, struct pt_regs *regs,
1118 struct uprobe_cpu_buffer *ucb, int dsize)
1120 struct trace_event_call *call = &tu->tp.call;
1121 struct uprobe_trace_entry_head *entry;
1122 struct bpf_prog *prog = call->prog;
1123 struct hlist_head *head;
1128 if (prog && !trace_call_bpf(prog, regs))
1131 esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
1133 size = esize + tu->tp.size + dsize;
1134 size = ALIGN(size + sizeof(u32), sizeof(u64)) - sizeof(u32);
1135 if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE, "profile buffer not large enough"))
1139 head = this_cpu_ptr(call->perf_events);
1140 if (hlist_empty(head))
1143 entry = perf_trace_buf_prepare(size, call->event.type, NULL, &rctx);
1147 if (is_ret_probe(tu)) {
1148 entry->vaddr[0] = func;
1149 entry->vaddr[1] = instruction_pointer(regs);
1150 data = DATAOF_TRACE_ENTRY(entry, true);
1152 entry->vaddr[0] = instruction_pointer(regs);
1153 data = DATAOF_TRACE_ENTRY(entry, false);
1156 memcpy(data, ucb->buf, tu->tp.size + dsize);
1158 if (size - esize > tu->tp.size + dsize) {
1159 int len = tu->tp.size + dsize;
1161 memset(data + len, 0, size - esize - len);
1164 perf_trace_buf_submit(entry, size, rctx, 0, 1, regs, head, NULL);
1169 /* uprobe profile handler */
1170 static int uprobe_perf_func(struct trace_uprobe *tu, struct pt_regs *regs,
1171 struct uprobe_cpu_buffer *ucb, int dsize)
1173 if (!uprobe_perf_filter(&tu->consumer, 0, current->mm))
1174 return UPROBE_HANDLER_REMOVE;
1176 if (!is_ret_probe(tu))
1177 __uprobe_perf_func(tu, 0, regs, ucb, dsize);
1181 static void uretprobe_perf_func(struct trace_uprobe *tu, unsigned long func,
1182 struct pt_regs *regs,
1183 struct uprobe_cpu_buffer *ucb, int dsize)
1185 __uprobe_perf_func(tu, func, regs, ucb, dsize);
1187 #endif /* CONFIG_PERF_EVENTS */
1190 trace_uprobe_register(struct trace_event_call *event, enum trace_reg type,
1193 struct trace_uprobe *tu = event->data;
1194 struct trace_event_file *file = data;
1197 case TRACE_REG_REGISTER:
1198 return probe_event_enable(tu, file, NULL);
1200 case TRACE_REG_UNREGISTER:
1201 probe_event_disable(tu, file);
1204 #ifdef CONFIG_PERF_EVENTS
1205 case TRACE_REG_PERF_REGISTER:
1206 return probe_event_enable(tu, NULL, uprobe_perf_filter);
1208 case TRACE_REG_PERF_UNREGISTER:
1209 probe_event_disable(tu, NULL);
1212 case TRACE_REG_PERF_OPEN:
1213 return uprobe_perf_open(tu, data);
1215 case TRACE_REG_PERF_CLOSE:
1216 return uprobe_perf_close(tu, data);
1225 static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
1227 struct trace_uprobe *tu;
1228 struct uprobe_dispatch_data udd;
1229 struct uprobe_cpu_buffer *ucb;
1234 tu = container_of(con, struct trace_uprobe, consumer);
1238 udd.bp_addr = instruction_pointer(regs);
1240 current->utask->vaddr = (unsigned long) &udd;
1242 if (WARN_ON_ONCE(!uprobe_cpu_buffer))
1245 dsize = __get_data_size(&tu->tp, regs);
1246 esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
1248 ucb = uprobe_buffer_get();
1249 store_trace_args(esize, &tu->tp, regs, ucb->buf, dsize);
1251 if (tu->tp.flags & TP_FLAG_TRACE)
1252 ret |= uprobe_trace_func(tu, regs, ucb, dsize);
1254 #ifdef CONFIG_PERF_EVENTS
1255 if (tu->tp.flags & TP_FLAG_PROFILE)
1256 ret |= uprobe_perf_func(tu, regs, ucb, dsize);
1258 uprobe_buffer_put(ucb);
1262 static int uretprobe_dispatcher(struct uprobe_consumer *con,
1263 unsigned long func, struct pt_regs *regs)
1265 struct trace_uprobe *tu;
1266 struct uprobe_dispatch_data udd;
1267 struct uprobe_cpu_buffer *ucb;
1270 tu = container_of(con, struct trace_uprobe, consumer);
1275 current->utask->vaddr = (unsigned long) &udd;
1277 if (WARN_ON_ONCE(!uprobe_cpu_buffer))
1280 dsize = __get_data_size(&tu->tp, regs);
1281 esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
1283 ucb = uprobe_buffer_get();
1284 store_trace_args(esize, &tu->tp, regs, ucb->buf, dsize);
1286 if (tu->tp.flags & TP_FLAG_TRACE)
1287 uretprobe_trace_func(tu, func, regs, ucb, dsize);
1289 #ifdef CONFIG_PERF_EVENTS
1290 if (tu->tp.flags & TP_FLAG_PROFILE)
1291 uretprobe_perf_func(tu, func, regs, ucb, dsize);
1293 uprobe_buffer_put(ucb);
1297 static struct trace_event_functions uprobe_funcs = {
1298 .trace = print_uprobe_event
1301 static int register_uprobe_event(struct trace_uprobe *tu)
1303 struct trace_event_call *call = &tu->tp.call;
1306 /* Initialize trace_event_call */
1307 INIT_LIST_HEAD(&call->class->fields);
1308 call->event.funcs = &uprobe_funcs;
1309 call->class->define_fields = uprobe_event_define_fields;
1311 if (set_print_fmt(&tu->tp, is_ret_probe(tu)) < 0)
1314 ret = register_trace_event(&call->event);
1316 kfree(call->print_fmt);
1320 call->flags = TRACE_EVENT_FL_UPROBE;
1321 call->class->reg = trace_uprobe_register;
1323 ret = trace_add_event_call(call);
1326 pr_info("Failed to register uprobe event: %s\n",
1327 trace_event_name(call));
1328 kfree(call->print_fmt);
1329 unregister_trace_event(&call->event);
1335 static int unregister_uprobe_event(struct trace_uprobe *tu)
1339 /* tu->event is unregistered in trace_remove_event_call() */
1340 ret = trace_remove_event_call(&tu->tp.call);
1343 kfree(tu->tp.call.print_fmt);
1344 tu->tp.call.print_fmt = NULL;
1348 /* Make a trace interface for controling probe points */
1349 static __init int init_uprobe_trace(void)
1351 struct dentry *d_tracer;
1353 d_tracer = tracing_init_dentry();
1354 if (IS_ERR(d_tracer))
1357 trace_create_file("uprobe_events", 0644, d_tracer,
1358 NULL, &uprobe_events_ops);
1359 /* Profile interface */
1360 trace_create_file("uprobe_profile", 0444, d_tracer,
1361 NULL, &uprobe_profile_ops);
1365 fs_initcall(init_uprobe_trace);