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 __kprobes 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 __kprobes 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 __kprobes
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
);
153 if (ret
< 0) { /* Failed to fetch string */
154 ((u8
*)get_rloc_data(dest
))[0] = '\0';
155 *(u32
*)dest
= make_data_rloc(0, get_rloc_offs(rloc
));
157 *(u32
*)dest
= make_data_rloc(ret
, get_rloc_offs(rloc
));
161 static __kprobes
void FETCH_FUNC_NAME(memory
, string_size
)(struct pt_regs
*regs
,
162 void *addr
, void *dest
)
165 void __user
*vaddr
= (void __force __user
*) addr
;
167 len
= strnlen_user(vaddr
, MAX_STRING_SIZE
);
169 if (len
== 0 || len
> MAX_STRING_SIZE
) /* Failed to check length */
175 static unsigned long translate_user_vaddr(void *file_offset
)
177 unsigned long base_addr
;
178 struct uprobe_dispatch_data
*udd
;
180 udd
= (void *) current
->utask
->vaddr
;
182 base_addr
= udd
->bp_addr
- udd
->tu
->offset
;
183 return base_addr
+ (unsigned long)file_offset
;
186 #define DEFINE_FETCH_file_offset(type) \
187 static __kprobes void FETCH_FUNC_NAME(file_offset, type)(struct pt_regs *regs,\
188 void *offset, void *dest) \
190 void *vaddr = (void *)translate_user_vaddr(offset); \
192 FETCH_FUNC_NAME(memory, type)(regs, vaddr, dest); \
194 DEFINE_BASIC_FETCH_FUNCS(file_offset
)
195 DEFINE_FETCH_file_offset(string
)
196 DEFINE_FETCH_file_offset(string_size
)
198 /* Fetch type information table */
199 const struct fetch_type uprobes_fetch_type_table
[] = {
201 [FETCH_TYPE_STRING
] = __ASSIGN_FETCH_TYPE("string", string
, string
,
202 sizeof(u32
), 1, "__data_loc char[]"),
203 [FETCH_TYPE_STRSIZE
] = __ASSIGN_FETCH_TYPE("string_size", u32
,
204 string_size
, sizeof(u32
), 0, "u32"),
206 ASSIGN_FETCH_TYPE(u8
, u8
, 0),
207 ASSIGN_FETCH_TYPE(u16
, u16
, 0),
208 ASSIGN_FETCH_TYPE(u32
, u32
, 0),
209 ASSIGN_FETCH_TYPE(u64
, u64
, 0),
210 ASSIGN_FETCH_TYPE(s8
, u8
, 1),
211 ASSIGN_FETCH_TYPE(s16
, u16
, 1),
212 ASSIGN_FETCH_TYPE(s32
, u32
, 1),
213 ASSIGN_FETCH_TYPE(s64
, u64
, 1),
215 ASSIGN_FETCH_TYPE_END
218 static inline void init_trace_uprobe_filter(struct trace_uprobe_filter
*filter
)
220 rwlock_init(&filter
->rwlock
);
221 filter
->nr_systemwide
= 0;
222 INIT_LIST_HEAD(&filter
->perf_events
);
225 static inline bool uprobe_filter_is_empty(struct trace_uprobe_filter
*filter
)
227 return !filter
->nr_systemwide
&& list_empty(&filter
->perf_events
);
230 static inline bool is_ret_probe(struct trace_uprobe
*tu
)
232 return tu
->consumer
.ret_handler
!= NULL
;
236 * Allocate new trace_uprobe and initialize it (including uprobes).
238 static struct trace_uprobe
*
239 alloc_trace_uprobe(const char *group
, const char *event
, int nargs
, bool is_ret
)
241 struct trace_uprobe
*tu
;
243 if (!event
|| !is_good_name(event
))
244 return ERR_PTR(-EINVAL
);
246 if (!group
|| !is_good_name(group
))
247 return ERR_PTR(-EINVAL
);
249 tu
= kzalloc(SIZEOF_TRACE_UPROBE(nargs
), GFP_KERNEL
);
251 return ERR_PTR(-ENOMEM
);
253 tu
->tp
.call
.class = &tu
->tp
.class;
254 tu
->tp
.call
.name
= kstrdup(event
, GFP_KERNEL
);
255 if (!tu
->tp
.call
.name
)
258 tu
->tp
.class.system
= kstrdup(group
, GFP_KERNEL
);
259 if (!tu
->tp
.class.system
)
262 INIT_LIST_HEAD(&tu
->list
);
263 INIT_LIST_HEAD(&tu
->tp
.files
);
264 tu
->consumer
.handler
= uprobe_dispatcher
;
266 tu
->consumer
.ret_handler
= uretprobe_dispatcher
;
267 init_trace_uprobe_filter(&tu
->filter
);
268 tu
->tp
.call
.flags
|= TRACE_EVENT_FL_USE_CALL_FILTER
;
272 kfree(tu
->tp
.call
.name
);
275 return ERR_PTR(-ENOMEM
);
278 static void free_trace_uprobe(struct trace_uprobe
*tu
)
282 for (i
= 0; i
< tu
->tp
.nr_args
; i
++)
283 traceprobe_free_probe_arg(&tu
->tp
.args
[i
]);
286 kfree(tu
->tp
.call
.class->system
);
287 kfree(tu
->tp
.call
.name
);
292 static struct trace_uprobe
*find_probe_event(const char *event
, const char *group
)
294 struct trace_uprobe
*tu
;
296 list_for_each_entry(tu
, &uprobe_list
, list
)
297 if (strcmp(ftrace_event_name(&tu
->tp
.call
), event
) == 0 &&
298 strcmp(tu
->tp
.call
.class->system
, group
) == 0)
304 /* Unregister a trace_uprobe and probe_event: call with locking uprobe_lock */
305 static int unregister_trace_uprobe(struct trace_uprobe
*tu
)
309 ret
= unregister_uprobe_event(tu
);
314 free_trace_uprobe(tu
);
318 /* Register a trace_uprobe and probe_event */
319 static int register_trace_uprobe(struct trace_uprobe
*tu
)
321 struct trace_uprobe
*old_tu
;
324 mutex_lock(&uprobe_lock
);
326 /* register as an event */
327 old_tu
= find_probe_event(ftrace_event_name(&tu
->tp
.call
),
328 tu
->tp
.call
.class->system
);
330 /* delete old event */
331 ret
= unregister_trace_uprobe(old_tu
);
336 ret
= register_uprobe_event(tu
);
338 pr_warning("Failed to register probe event(%d)\n", ret
);
342 list_add_tail(&tu
->list
, &uprobe_list
);
345 mutex_unlock(&uprobe_lock
);
352 * - Add uprobe: p|r[:[GRP/]EVENT] PATH:OFFSET [FETCHARGS]
354 * - Remove uprobe: -:[GRP/]EVENT
356 static int create_trace_uprobe(int argc
, char **argv
)
358 struct trace_uprobe
*tu
;
360 char *arg
, *event
, *group
, *filename
;
361 char buf
[MAX_EVENT_NAME_LEN
];
363 unsigned long offset
;
364 bool is_delete
, is_return
;
374 /* argc must be >= 1 */
375 if (argv
[0][0] == '-')
377 else if (argv
[0][0] == 'r')
379 else if (argv
[0][0] != 'p') {
380 pr_info("Probe definition must be started with 'p', 'r' or '-'.\n");
384 if (argv
[0][1] == ':') {
386 arg
= strchr(event
, '/');
393 if (strlen(group
) == 0) {
394 pr_info("Group name is not specified\n");
398 if (strlen(event
) == 0) {
399 pr_info("Event name is not specified\n");
404 group
= UPROBE_EVENT_SYSTEM
;
410 pr_info("Delete command needs an event name.\n");
413 mutex_lock(&uprobe_lock
);
414 tu
= find_probe_event(event
, group
);
417 mutex_unlock(&uprobe_lock
);
418 pr_info("Event %s/%s doesn't exist.\n", group
, event
);
421 /* delete an event */
422 ret
= unregister_trace_uprobe(tu
);
423 mutex_unlock(&uprobe_lock
);
428 pr_info("Probe point is not specified.\n");
431 if (isdigit(argv
[1][0])) {
432 pr_info("probe point must be have a filename.\n");
435 arg
= strchr(argv
[1], ':');
438 goto fail_address_parse
;
443 ret
= kern_path(filename
, LOOKUP_FOLLOW
, &path
);
445 goto fail_address_parse
;
447 inode
= igrab(path
.dentry
->d_inode
);
450 if (!inode
|| !S_ISREG(inode
->i_mode
)) {
452 goto fail_address_parse
;
455 ret
= kstrtoul(arg
, 0, &offset
);
457 goto fail_address_parse
;
467 tail
= kstrdup(kbasename(filename
), GFP_KERNEL
);
470 goto fail_address_parse
;
473 ptr
= strpbrk(tail
, ".-_");
477 snprintf(buf
, MAX_EVENT_NAME_LEN
, "%c_%s_0x%lx", 'p', tail
, offset
);
482 tu
= alloc_trace_uprobe(group
, event
, argc
, is_return
);
484 pr_info("Failed to allocate trace_uprobe.(%d)\n", (int)PTR_ERR(tu
));
486 goto fail_address_parse
;
490 tu
->filename
= kstrdup(filename
, GFP_KERNEL
);
493 pr_info("Failed to allocate filename.\n");
498 /* parse arguments */
500 for (i
= 0; i
< argc
&& i
< MAX_TRACE_ARGS
; i
++) {
501 struct probe_arg
*parg
= &tu
->tp
.args
[i
];
503 /* Increment count for freeing args in error case */
506 /* Parse argument name */
507 arg
= strchr(argv
[i
], '=');
510 parg
->name
= kstrdup(argv
[i
], GFP_KERNEL
);
513 /* If argument name is omitted, set "argN" */
514 snprintf(buf
, MAX_EVENT_NAME_LEN
, "arg%d", i
+ 1);
515 parg
->name
= kstrdup(buf
, GFP_KERNEL
);
519 pr_info("Failed to allocate argument[%d] name.\n", i
);
524 if (!is_good_name(parg
->name
)) {
525 pr_info("Invalid argument[%d] name: %s\n", i
, parg
->name
);
530 if (traceprobe_conflict_field_name(parg
->name
, tu
->tp
.args
, i
)) {
531 pr_info("Argument[%d] name '%s' conflicts with "
532 "another field.\n", i
, argv
[i
]);
537 /* Parse fetch argument */
538 ret
= traceprobe_parse_probe_arg(arg
, &tu
->tp
.size
, parg
,
541 pr_info("Parse error at argument[%d]. (%d)\n", i
, ret
);
546 ret
= register_trace_uprobe(tu
);
552 free_trace_uprobe(tu
);
559 pr_info("Failed to parse address or file.\n");
564 static int cleanup_all_probes(void)
566 struct trace_uprobe
*tu
;
569 mutex_lock(&uprobe_lock
);
570 while (!list_empty(&uprobe_list
)) {
571 tu
= list_entry(uprobe_list
.next
, struct trace_uprobe
, list
);
572 ret
= unregister_trace_uprobe(tu
);
576 mutex_unlock(&uprobe_lock
);
580 /* Probes listing interfaces */
581 static void *probes_seq_start(struct seq_file
*m
, loff_t
*pos
)
583 mutex_lock(&uprobe_lock
);
584 return seq_list_start(&uprobe_list
, *pos
);
587 static void *probes_seq_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
589 return seq_list_next(v
, &uprobe_list
, pos
);
592 static void probes_seq_stop(struct seq_file
*m
, void *v
)
594 mutex_unlock(&uprobe_lock
);
597 static int probes_seq_show(struct seq_file
*m
, void *v
)
599 struct trace_uprobe
*tu
= v
;
600 char c
= is_ret_probe(tu
) ? 'r' : 'p';
603 seq_printf(m
, "%c:%s/%s", c
, tu
->tp
.call
.class->system
,
604 ftrace_event_name(&tu
->tp
.call
));
605 seq_printf(m
, " %s:0x%p", tu
->filename
, (void *)tu
->offset
);
607 for (i
= 0; i
< tu
->tp
.nr_args
; i
++)
608 seq_printf(m
, " %s=%s", tu
->tp
.args
[i
].name
, tu
->tp
.args
[i
].comm
);
614 static const struct seq_operations probes_seq_op
= {
615 .start
= probes_seq_start
,
616 .next
= probes_seq_next
,
617 .stop
= probes_seq_stop
,
618 .show
= probes_seq_show
621 static int probes_open(struct inode
*inode
, struct file
*file
)
625 if ((file
->f_mode
& FMODE_WRITE
) && (file
->f_flags
& O_TRUNC
)) {
626 ret
= cleanup_all_probes();
631 return seq_open(file
, &probes_seq_op
);
634 static ssize_t
probes_write(struct file
*file
, const char __user
*buffer
,
635 size_t count
, loff_t
*ppos
)
637 return traceprobe_probes_write(file
, buffer
, count
, ppos
, create_trace_uprobe
);
640 static const struct file_operations uprobe_events_ops
= {
641 .owner
= THIS_MODULE
,
645 .release
= seq_release
,
646 .write
= probes_write
,
649 /* Probes profiling interfaces */
650 static int probes_profile_seq_show(struct seq_file
*m
, void *v
)
652 struct trace_uprobe
*tu
= v
;
654 seq_printf(m
, " %s %-44s %15lu\n", tu
->filename
,
655 ftrace_event_name(&tu
->tp
.call
), tu
->nhit
);
659 static const struct seq_operations profile_seq_op
= {
660 .start
= probes_seq_start
,
661 .next
= probes_seq_next
,
662 .stop
= probes_seq_stop
,
663 .show
= probes_profile_seq_show
666 static int profile_open(struct inode
*inode
, struct file
*file
)
668 return seq_open(file
, &profile_seq_op
);
671 static const struct file_operations uprobe_profile_ops
= {
672 .owner
= THIS_MODULE
,
673 .open
= profile_open
,
676 .release
= seq_release
,
679 struct uprobe_cpu_buffer
{
683 static struct uprobe_cpu_buffer __percpu
*uprobe_cpu_buffer
;
684 static int uprobe_buffer_refcnt
;
686 static int uprobe_buffer_init(void)
690 uprobe_cpu_buffer
= alloc_percpu(struct uprobe_cpu_buffer
);
691 if (uprobe_cpu_buffer
== NULL
)
694 for_each_possible_cpu(cpu
) {
695 struct page
*p
= alloc_pages_node(cpu_to_node(cpu
),
701 per_cpu_ptr(uprobe_cpu_buffer
, cpu
)->buf
= page_address(p
);
702 mutex_init(&per_cpu_ptr(uprobe_cpu_buffer
, cpu
)->mutex
);
708 for_each_possible_cpu(cpu
) {
711 free_page((unsigned long)per_cpu_ptr(uprobe_cpu_buffer
, cpu
)->buf
);
714 free_percpu(uprobe_cpu_buffer
);
718 static int uprobe_buffer_enable(void)
722 BUG_ON(!mutex_is_locked(&event_mutex
));
724 if (uprobe_buffer_refcnt
++ == 0) {
725 ret
= uprobe_buffer_init();
727 uprobe_buffer_refcnt
--;
733 static void uprobe_buffer_disable(void)
735 BUG_ON(!mutex_is_locked(&event_mutex
));
737 if (--uprobe_buffer_refcnt
== 0) {
738 free_percpu(uprobe_cpu_buffer
);
739 uprobe_cpu_buffer
= NULL
;
743 static struct uprobe_cpu_buffer
*uprobe_buffer_get(void)
745 struct uprobe_cpu_buffer
*ucb
;
748 cpu
= raw_smp_processor_id();
749 ucb
= per_cpu_ptr(uprobe_cpu_buffer
, cpu
);
752 * Use per-cpu buffers for fastest access, but we might migrate
753 * so the mutex makes sure we have sole access to it.
755 mutex_lock(&ucb
->mutex
);
760 static void uprobe_buffer_put(struct uprobe_cpu_buffer
*ucb
)
762 mutex_unlock(&ucb
->mutex
);
765 static void __uprobe_trace_func(struct trace_uprobe
*tu
,
766 unsigned long func
, struct pt_regs
*regs
,
767 struct uprobe_cpu_buffer
*ucb
, int dsize
,
768 struct ftrace_event_file
*ftrace_file
)
770 struct uprobe_trace_entry_head
*entry
;
771 struct ring_buffer_event
*event
;
772 struct ring_buffer
*buffer
;
775 struct ftrace_event_call
*call
= &tu
->tp
.call
;
777 WARN_ON(call
!= ftrace_file
->event_call
);
779 if (WARN_ON_ONCE(tu
->tp
.size
+ dsize
> PAGE_SIZE
))
782 if (ftrace_trigger_soft_disabled(ftrace_file
))
785 esize
= SIZEOF_TRACE_ENTRY(is_ret_probe(tu
));
786 size
= esize
+ tu
->tp
.size
+ dsize
;
787 event
= trace_event_buffer_lock_reserve(&buffer
, ftrace_file
,
788 call
->event
.type
, size
, 0, 0);
792 entry
= ring_buffer_event_data(event
);
793 if (is_ret_probe(tu
)) {
794 entry
->vaddr
[0] = func
;
795 entry
->vaddr
[1] = instruction_pointer(regs
);
796 data
= DATAOF_TRACE_ENTRY(entry
, true);
798 entry
->vaddr
[0] = instruction_pointer(regs
);
799 data
= DATAOF_TRACE_ENTRY(entry
, false);
802 memcpy(data
, ucb
->buf
, tu
->tp
.size
+ dsize
);
804 event_trigger_unlock_commit(ftrace_file
, buffer
, event
, entry
, 0, 0);
808 static int uprobe_trace_func(struct trace_uprobe
*tu
, struct pt_regs
*regs
,
809 struct uprobe_cpu_buffer
*ucb
, int dsize
)
811 struct event_file_link
*link
;
813 if (is_ret_probe(tu
))
817 list_for_each_entry_rcu(link
, &tu
->tp
.files
, list
)
818 __uprobe_trace_func(tu
, 0, regs
, ucb
, dsize
, link
->file
);
824 static void uretprobe_trace_func(struct trace_uprobe
*tu
, unsigned long func
,
825 struct pt_regs
*regs
,
826 struct uprobe_cpu_buffer
*ucb
, int dsize
)
828 struct event_file_link
*link
;
831 list_for_each_entry_rcu(link
, &tu
->tp
.files
, list
)
832 __uprobe_trace_func(tu
, func
, regs
, ucb
, dsize
, link
->file
);
836 /* Event entry printers */
837 static enum print_line_t
838 print_uprobe_event(struct trace_iterator
*iter
, int flags
, struct trace_event
*event
)
840 struct uprobe_trace_entry_head
*entry
;
841 struct trace_seq
*s
= &iter
->seq
;
842 struct trace_uprobe
*tu
;
846 entry
= (struct uprobe_trace_entry_head
*)iter
->ent
;
847 tu
= container_of(event
, struct trace_uprobe
, tp
.call
.event
);
849 if (is_ret_probe(tu
)) {
850 if (!trace_seq_printf(s
, "%s: (0x%lx <- 0x%lx)",
851 ftrace_event_name(&tu
->tp
.call
),
852 entry
->vaddr
[1], entry
->vaddr
[0]))
854 data
= DATAOF_TRACE_ENTRY(entry
, true);
856 if (!trace_seq_printf(s
, "%s: (0x%lx)",
857 ftrace_event_name(&tu
->tp
.call
),
860 data
= DATAOF_TRACE_ENTRY(entry
, false);
863 for (i
= 0; i
< tu
->tp
.nr_args
; i
++) {
864 struct probe_arg
*parg
= &tu
->tp
.args
[i
];
866 if (!parg
->type
->print(s
, parg
->name
, data
+ parg
->offset
, entry
))
870 if (trace_seq_puts(s
, "\n"))
871 return TRACE_TYPE_HANDLED
;
874 return TRACE_TYPE_PARTIAL_LINE
;
877 typedef bool (*filter_func_t
)(struct uprobe_consumer
*self
,
878 enum uprobe_filter_ctx ctx
,
879 struct mm_struct
*mm
);
882 probe_event_enable(struct trace_uprobe
*tu
, struct ftrace_event_file
*file
,
883 filter_func_t filter
)
885 bool enabled
= trace_probe_is_enabled(&tu
->tp
);
886 struct event_file_link
*link
= NULL
;
890 link
= kmalloc(sizeof(*link
), GFP_KERNEL
);
895 list_add_tail_rcu(&link
->list
, &tu
->tp
.files
);
897 tu
->tp
.flags
|= TP_FLAG_TRACE
;
899 tu
->tp
.flags
|= TP_FLAG_PROFILE
;
901 ret
= uprobe_buffer_enable();
905 WARN_ON(!uprobe_filter_is_empty(&tu
->filter
));
910 tu
->consumer
.filter
= filter
;
911 ret
= uprobe_register(tu
->inode
, tu
->offset
, &tu
->consumer
);
914 list_del(&link
->list
);
916 tu
->tp
.flags
&= ~TP_FLAG_TRACE
;
918 tu
->tp
.flags
&= ~TP_FLAG_PROFILE
;
925 probe_event_disable(struct trace_uprobe
*tu
, struct ftrace_event_file
*file
)
927 if (!trace_probe_is_enabled(&tu
->tp
))
931 struct event_file_link
*link
;
933 link
= find_event_file_link(&tu
->tp
, file
);
937 list_del_rcu(&link
->list
);
938 /* synchronize with u{,ret}probe_trace_func */
942 if (!list_empty(&tu
->tp
.files
))
946 WARN_ON(!uprobe_filter_is_empty(&tu
->filter
));
948 uprobe_unregister(tu
->inode
, tu
->offset
, &tu
->consumer
);
949 tu
->tp
.flags
&= file
? ~TP_FLAG_TRACE
: ~TP_FLAG_PROFILE
;
951 uprobe_buffer_disable();
954 static int uprobe_event_define_fields(struct ftrace_event_call
*event_call
)
957 struct uprobe_trace_entry_head field
;
958 struct trace_uprobe
*tu
= event_call
->data
;
960 if (is_ret_probe(tu
)) {
961 DEFINE_FIELD(unsigned long, vaddr
[0], FIELD_STRING_FUNC
, 0);
962 DEFINE_FIELD(unsigned long, vaddr
[1], FIELD_STRING_RETIP
, 0);
963 size
= SIZEOF_TRACE_ENTRY(true);
965 DEFINE_FIELD(unsigned long, vaddr
[0], FIELD_STRING_IP
, 0);
966 size
= SIZEOF_TRACE_ENTRY(false);
968 /* Set argument names as fields */
969 for (i
= 0; i
< tu
->tp
.nr_args
; i
++) {
970 struct probe_arg
*parg
= &tu
->tp
.args
[i
];
972 ret
= trace_define_field(event_call
, parg
->type
->fmttype
,
973 parg
->name
, size
+ parg
->offset
,
974 parg
->type
->size
, parg
->type
->is_signed
,
983 #ifdef CONFIG_PERF_EVENTS
985 __uprobe_perf_filter(struct trace_uprobe_filter
*filter
, struct mm_struct
*mm
)
987 struct perf_event
*event
;
989 if (filter
->nr_systemwide
)
992 list_for_each_entry(event
, &filter
->perf_events
, hw
.tp_list
) {
993 if (event
->hw
.tp_target
->mm
== mm
)
1001 uprobe_filter_event(struct trace_uprobe
*tu
, struct perf_event
*event
)
1003 return __uprobe_perf_filter(&tu
->filter
, event
->hw
.tp_target
->mm
);
1006 static int uprobe_perf_open(struct trace_uprobe
*tu
, struct perf_event
*event
)
1010 write_lock(&tu
->filter
.rwlock
);
1011 if (event
->hw
.tp_target
) {
1013 * event->parent != NULL means copy_process(), we can avoid
1014 * uprobe_apply(). current->mm must be probed and we can rely
1015 * on dup_mmap() which preserves the already installed bp's.
1017 * attr.enable_on_exec means that exec/mmap will install the
1018 * breakpoints we need.
1020 done
= tu
->filter
.nr_systemwide
||
1021 event
->parent
|| event
->attr
.enable_on_exec
||
1022 uprobe_filter_event(tu
, event
);
1023 list_add(&event
->hw
.tp_list
, &tu
->filter
.perf_events
);
1025 done
= tu
->filter
.nr_systemwide
;
1026 tu
->filter
.nr_systemwide
++;
1028 write_unlock(&tu
->filter
.rwlock
);
1031 uprobe_apply(tu
->inode
, tu
->offset
, &tu
->consumer
, true);
1036 static int uprobe_perf_close(struct trace_uprobe
*tu
, struct perf_event
*event
)
1040 write_lock(&tu
->filter
.rwlock
);
1041 if (event
->hw
.tp_target
) {
1042 list_del(&event
->hw
.tp_list
);
1043 done
= tu
->filter
.nr_systemwide
||
1044 (event
->hw
.tp_target
->flags
& PF_EXITING
) ||
1045 uprobe_filter_event(tu
, event
);
1047 tu
->filter
.nr_systemwide
--;
1048 done
= tu
->filter
.nr_systemwide
;
1050 write_unlock(&tu
->filter
.rwlock
);
1053 uprobe_apply(tu
->inode
, tu
->offset
, &tu
->consumer
, false);
1058 static bool uprobe_perf_filter(struct uprobe_consumer
*uc
,
1059 enum uprobe_filter_ctx ctx
, struct mm_struct
*mm
)
1061 struct trace_uprobe
*tu
;
1064 tu
= container_of(uc
, struct trace_uprobe
, consumer
);
1065 read_lock(&tu
->filter
.rwlock
);
1066 ret
= __uprobe_perf_filter(&tu
->filter
, mm
);
1067 read_unlock(&tu
->filter
.rwlock
);
1072 static void __uprobe_perf_func(struct trace_uprobe
*tu
,
1073 unsigned long func
, struct pt_regs
*regs
,
1074 struct uprobe_cpu_buffer
*ucb
, int dsize
)
1076 struct ftrace_event_call
*call
= &tu
->tp
.call
;
1077 struct uprobe_trace_entry_head
*entry
;
1078 struct hlist_head
*head
;
1083 esize
= SIZEOF_TRACE_ENTRY(is_ret_probe(tu
));
1085 size
= esize
+ tu
->tp
.size
+ dsize
;
1086 size
= ALIGN(size
+ sizeof(u32
), sizeof(u64
)) - sizeof(u32
);
1087 if (WARN_ONCE(size
> PERF_MAX_TRACE_SIZE
, "profile buffer not large enough"))
1091 head
= this_cpu_ptr(call
->perf_events
);
1092 if (hlist_empty(head
))
1095 entry
= perf_trace_buf_prepare(size
, call
->event
.type
, regs
, &rctx
);
1099 if (is_ret_probe(tu
)) {
1100 entry
->vaddr
[0] = func
;
1101 entry
->vaddr
[1] = instruction_pointer(regs
);
1102 data
= DATAOF_TRACE_ENTRY(entry
, true);
1104 entry
->vaddr
[0] = instruction_pointer(regs
);
1105 data
= DATAOF_TRACE_ENTRY(entry
, false);
1108 memcpy(data
, ucb
->buf
, tu
->tp
.size
+ dsize
);
1110 if (size
- esize
> tu
->tp
.size
+ dsize
) {
1111 int len
= tu
->tp
.size
+ dsize
;
1113 memset(data
+ len
, 0, size
- esize
- len
);
1116 perf_trace_buf_submit(entry
, size
, rctx
, 0, 1, regs
, head
, NULL
);
1121 /* uprobe profile handler */
1122 static int uprobe_perf_func(struct trace_uprobe
*tu
, struct pt_regs
*regs
,
1123 struct uprobe_cpu_buffer
*ucb
, int dsize
)
1125 if (!uprobe_perf_filter(&tu
->consumer
, 0, current
->mm
))
1126 return UPROBE_HANDLER_REMOVE
;
1128 if (!is_ret_probe(tu
))
1129 __uprobe_perf_func(tu
, 0, regs
, ucb
, dsize
);
1133 static void uretprobe_perf_func(struct trace_uprobe
*tu
, unsigned long func
,
1134 struct pt_regs
*regs
,
1135 struct uprobe_cpu_buffer
*ucb
, int dsize
)
1137 __uprobe_perf_func(tu
, func
, regs
, ucb
, dsize
);
1139 #endif /* CONFIG_PERF_EVENTS */
1142 trace_uprobe_register(struct ftrace_event_call
*event
, enum trace_reg type
,
1145 struct trace_uprobe
*tu
= event
->data
;
1146 struct ftrace_event_file
*file
= data
;
1149 case TRACE_REG_REGISTER
:
1150 return probe_event_enable(tu
, file
, NULL
);
1152 case TRACE_REG_UNREGISTER
:
1153 probe_event_disable(tu
, file
);
1156 #ifdef CONFIG_PERF_EVENTS
1157 case TRACE_REG_PERF_REGISTER
:
1158 return probe_event_enable(tu
, NULL
, uprobe_perf_filter
);
1160 case TRACE_REG_PERF_UNREGISTER
:
1161 probe_event_disable(tu
, NULL
);
1164 case TRACE_REG_PERF_OPEN
:
1165 return uprobe_perf_open(tu
, data
);
1167 case TRACE_REG_PERF_CLOSE
:
1168 return uprobe_perf_close(tu
, data
);
1177 static int uprobe_dispatcher(struct uprobe_consumer
*con
, struct pt_regs
*regs
)
1179 struct trace_uprobe
*tu
;
1180 struct uprobe_dispatch_data udd
;
1181 struct uprobe_cpu_buffer
*ucb
;
1186 tu
= container_of(con
, struct trace_uprobe
, consumer
);
1190 udd
.bp_addr
= instruction_pointer(regs
);
1192 current
->utask
->vaddr
= (unsigned long) &udd
;
1194 #ifdef CONFIG_PERF_EVENTS
1195 if ((tu
->tp
.flags
& TP_FLAG_TRACE
) == 0 &&
1196 !uprobe_perf_filter(&tu
->consumer
, 0, current
->mm
))
1197 return UPROBE_HANDLER_REMOVE
;
1200 if (WARN_ON_ONCE(!uprobe_cpu_buffer
))
1203 dsize
= __get_data_size(&tu
->tp
, regs
);
1204 esize
= SIZEOF_TRACE_ENTRY(is_ret_probe(tu
));
1206 ucb
= uprobe_buffer_get();
1207 store_trace_args(esize
, &tu
->tp
, regs
, ucb
->buf
, dsize
);
1209 if (tu
->tp
.flags
& TP_FLAG_TRACE
)
1210 ret
|= uprobe_trace_func(tu
, regs
, ucb
, dsize
);
1212 #ifdef CONFIG_PERF_EVENTS
1213 if (tu
->tp
.flags
& TP_FLAG_PROFILE
)
1214 ret
|= uprobe_perf_func(tu
, regs
, ucb
, dsize
);
1216 uprobe_buffer_put(ucb
);
1220 static int uretprobe_dispatcher(struct uprobe_consumer
*con
,
1221 unsigned long func
, struct pt_regs
*regs
)
1223 struct trace_uprobe
*tu
;
1224 struct uprobe_dispatch_data udd
;
1225 struct uprobe_cpu_buffer
*ucb
;
1228 tu
= container_of(con
, struct trace_uprobe
, consumer
);
1233 current
->utask
->vaddr
= (unsigned long) &udd
;
1235 if (WARN_ON_ONCE(!uprobe_cpu_buffer
))
1238 dsize
= __get_data_size(&tu
->tp
, regs
);
1239 esize
= SIZEOF_TRACE_ENTRY(is_ret_probe(tu
));
1241 ucb
= uprobe_buffer_get();
1242 store_trace_args(esize
, &tu
->tp
, regs
, ucb
->buf
, dsize
);
1244 if (tu
->tp
.flags
& TP_FLAG_TRACE
)
1245 uretprobe_trace_func(tu
, func
, regs
, ucb
, dsize
);
1247 #ifdef CONFIG_PERF_EVENTS
1248 if (tu
->tp
.flags
& TP_FLAG_PROFILE
)
1249 uretprobe_perf_func(tu
, func
, regs
, ucb
, dsize
);
1251 uprobe_buffer_put(ucb
);
1255 static struct trace_event_functions uprobe_funcs
= {
1256 .trace
= print_uprobe_event
1259 static int register_uprobe_event(struct trace_uprobe
*tu
)
1261 struct ftrace_event_call
*call
= &tu
->tp
.call
;
1264 /* Initialize ftrace_event_call */
1265 INIT_LIST_HEAD(&call
->class->fields
);
1266 call
->event
.funcs
= &uprobe_funcs
;
1267 call
->class->define_fields
= uprobe_event_define_fields
;
1269 if (set_print_fmt(&tu
->tp
, is_ret_probe(tu
)) < 0)
1272 ret
= register_ftrace_event(&call
->event
);
1274 kfree(call
->print_fmt
);
1278 call
->class->reg
= trace_uprobe_register
;
1280 ret
= trace_add_event_call(call
);
1283 pr_info("Failed to register uprobe event: %s\n",
1284 ftrace_event_name(call
));
1285 kfree(call
->print_fmt
);
1286 unregister_ftrace_event(&call
->event
);
1292 static int unregister_uprobe_event(struct trace_uprobe
*tu
)
1296 /* tu->event is unregistered in trace_remove_event_call() */
1297 ret
= trace_remove_event_call(&tu
->tp
.call
);
1300 kfree(tu
->tp
.call
.print_fmt
);
1301 tu
->tp
.call
.print_fmt
= NULL
;
1305 /* Make a trace interface for controling probe points */
1306 static __init
int init_uprobe_trace(void)
1308 struct dentry
*d_tracer
;
1310 d_tracer
= tracing_init_dentry();
1314 trace_create_file("uprobe_events", 0644, d_tracer
,
1315 NULL
, &uprobe_events_ops
);
1316 /* Profile interface */
1317 trace_create_file("uprobe_profile", 0444, d_tracer
,
1318 NULL
, &uprobe_profile_ops
);
1322 fs_initcall(init_uprobe_trace
);