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

TOMOYO Linux Cross Reference
Linux/kernel/trace/trace_uprobe.c

Version: ~ [ linux-5.12-rc7 ] ~ [ linux-5.11.13 ] ~ [ linux-5.10.29 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.111 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.186 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.230 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.266 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.266 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  * uprobes-based tracing events
  3  *
  4  * This program is free software; you can redistribute it and/or modify
  5  * it under the terms of the GNU General Public License version 2 as
  6  * published by the Free Software Foundation.
  7  *
  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.
 12  *
 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
 16  *
 17  * Copyright (C) IBM Corporation, 2010-2012
 18  * Author:      Srikar Dronamraju <srikar@linux.vnet.ibm.com>
 19  */
 20 
 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>
 26 
 27 #include "trace_probe.h"
 28 
 29 #define UPROBE_EVENT_SYSTEM     "uprobes"
 30 
 31 struct uprobe_trace_entry_head {
 32         struct trace_entry      ent;
 33         unsigned long           vaddr[];
 34 };
 35 
 36 #define SIZEOF_TRACE_ENTRY(is_return)                   \
 37         (sizeof(struct uprobe_trace_entry_head) +       \
 38          sizeof(unsigned long) * (is_return ? 2 : 1))
 39 
 40 #define DATAOF_TRACE_ENTRY(entry, is_return)            \
 41         ((void*)(entry) + SIZEOF_TRACE_ENTRY(is_return))
 42 
 43 struct trace_uprobe_filter {
 44         rwlock_t                rwlock;
 45         int                     nr_systemwide;
 46         struct list_head        perf_events;
 47 };
 48 
 49 /*
 50  * uprobe event core functions
 51  */
 52 struct trace_uprobe {
 53         struct list_head                list;
 54         struct trace_uprobe_filter      filter;
 55         struct uprobe_consumer          consumer;
 56         struct inode                    *inode;
 57         char                            *filename;
 58         unsigned long                   offset;
 59         unsigned long                   nhit;
 60         struct trace_probe              tp;
 61 };
 62 
 63 #define SIZEOF_TRACE_UPROBE(n)                          \
 64         (offsetof(struct trace_uprobe, tp.args) +       \
 65         (sizeof(struct probe_arg) * (n)))
 66 
 67 static int register_uprobe_event(struct trace_uprobe *tu);
 68 static int unregister_uprobe_event(struct trace_uprobe *tu);
 69 
 70 static DEFINE_MUTEX(uprobe_lock);
 71 static LIST_HEAD(uprobe_list);
 72 
 73 struct uprobe_dispatch_data {
 74         struct trace_uprobe     *tu;
 75         unsigned long           bp_addr;
 76 };
 77 
 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);
 81 
 82 #ifdef CONFIG_STACK_GROWSUP
 83 static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
 84 {
 85         return addr - (n * sizeof(long));
 86 }
 87 #else
 88 static unsigned long adjust_stack_addr(unsigned long addr, unsigned int n)
 89 {
 90         return addr + (n * sizeof(long));
 91 }
 92 #endif
 93 
 94 static unsigned long get_user_stack_nth(struct pt_regs *regs, unsigned int n)
 95 {
 96         unsigned long ret;
 97         unsigned long addr = user_stack_pointer(regs);
 98 
 99         addr = adjust_stack_addr(addr, n);
100 
101         if (copy_from_user(&ret, (void __force __user *) addr, sizeof(ret)))
102                 return 0;
103 
104         return ret;
105 }
106 
107 /*
108  * Uprobes-specific fetch functions
109  */
110 #define DEFINE_FETCH_stack(type)                                        \
111 static __kprobes void FETCH_FUNC_NAME(stack, type)(struct pt_regs *regs,\
112                                           void *offset, void *dest)     \
113 {                                                                       \
114         *(type *)dest = (type)get_user_stack_nth(regs,                  \
115                                               ((unsigned long)offset)); \
116 }
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
121 
122 #define DEFINE_FETCH_memory(type)                                       \
123 static __kprobes void FETCH_FUNC_NAME(memory, type)(struct pt_regs *regs,\
124                                                 void *addr, void *dest) \
125 {                                                                       \
126         type retval;                                                    \
127         void __user *vaddr = (void __force __user *) addr;              \
128                                                                         \
129         if (copy_from_user(&retval, vaddr, sizeof(type)))               \
130                 *(type *)dest = 0;                                      \
131         else                                                            \
132                 *(type *) dest = retval;                                \
133 }
134 DEFINE_BASIC_FETCH_FUNCS(memory)
135 /*
136  * Fetch a null-terminated string. Caller MUST set *(u32 *)dest with max
137  * length and relative data location.
138  */
139 static __kprobes void FETCH_FUNC_NAME(memory, string)(struct pt_regs *regs,
140                                                       void *addr, void *dest)
141 {
142         long ret;
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;
147 
148         if (!maxlen)
149                 return;
150 
151         ret = strncpy_from_user(dst, src, maxlen);
152 
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));
156         } else {
157                 *(u32 *)dest = make_data_rloc(ret, get_rloc_offs(rloc));
158         }
159 }
160 
161 static __kprobes void FETCH_FUNC_NAME(memory, string_size)(struct pt_regs *regs,
162                                                       void *addr, void *dest)
163 {
164         int len;
165         void __user *vaddr = (void __force __user *) addr;
166 
167         len = strnlen_user(vaddr, MAX_STRING_SIZE);
168 
169         if (len == 0 || len > MAX_STRING_SIZE)  /* Failed to check length */
170                 *(u32 *)dest = 0;
171         else
172                 *(u32 *)dest = len;
173 }
174 
175 static unsigned long translate_user_vaddr(void *file_offset)
176 {
177         unsigned long base_addr;
178         struct uprobe_dispatch_data *udd;
179 
180         udd = (void *) current->utask->vaddr;
181 
182         base_addr = udd->bp_addr - udd->tu->offset;
183         return base_addr + (unsigned long)file_offset;
184 }
185 
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)       \
189 {                                                                       \
190         void *vaddr = (void *)translate_user_vaddr(offset);             \
191                                                                         \
192         FETCH_FUNC_NAME(memory, type)(regs, vaddr, dest);               \
193 }
194 DEFINE_BASIC_FETCH_FUNCS(file_offset)
195 DEFINE_FETCH_file_offset(string)
196 DEFINE_FETCH_file_offset(string_size)
197 
198 /* Fetch type information table */
199 const struct fetch_type uprobes_fetch_type_table[] = {
200         /* Special types */
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"),
205         /* Basic types */
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),
214 
215         ASSIGN_FETCH_TYPE_END
216 };
217 
218 static inline void init_trace_uprobe_filter(struct trace_uprobe_filter *filter)
219 {
220         rwlock_init(&filter->rwlock);
221         filter->nr_systemwide = 0;
222         INIT_LIST_HEAD(&filter->perf_events);
223 }
224 
225 static inline bool uprobe_filter_is_empty(struct trace_uprobe_filter *filter)
226 {
227         return !filter->nr_systemwide && list_empty(&filter->perf_events);
228 }
229 
230 static inline bool is_ret_probe(struct trace_uprobe *tu)
231 {
232         return tu->consumer.ret_handler != NULL;
233 }
234 
235 /*
236  * Allocate new trace_uprobe and initialize it (including uprobes).
237  */
238 static struct trace_uprobe *
239 alloc_trace_uprobe(const char *group, const char *event, int nargs, bool is_ret)
240 {
241         struct trace_uprobe *tu;
242 
243         if (!event || !is_good_name(event))
244                 return ERR_PTR(-EINVAL);
245 
246         if (!group || !is_good_name(group))
247                 return ERR_PTR(-EINVAL);
248 
249         tu = kzalloc(SIZEOF_TRACE_UPROBE(nargs), GFP_KERNEL);
250         if (!tu)
251                 return ERR_PTR(-ENOMEM);
252 
253         tu->tp.call.class = &tu->tp.class;
254         tu->tp.call.name = kstrdup(event, GFP_KERNEL);
255         if (!tu->tp.call.name)
256                 goto error;
257 
258         tu->tp.class.system = kstrdup(group, GFP_KERNEL);
259         if (!tu->tp.class.system)
260                 goto error;
261 
262         INIT_LIST_HEAD(&tu->list);
263         tu->consumer.handler = uprobe_dispatcher;
264         if (is_ret)
265                 tu->consumer.ret_handler = uretprobe_dispatcher;
266         init_trace_uprobe_filter(&tu->filter);
267         tu->tp.call.flags |= TRACE_EVENT_FL_USE_CALL_FILTER;
268         return tu;
269 
270 error:
271         kfree(tu->tp.call.name);
272         kfree(tu);
273 
274         return ERR_PTR(-ENOMEM);
275 }
276 
277 static void free_trace_uprobe(struct trace_uprobe *tu)
278 {
279         int i;
280 
281         for (i = 0; i < tu->tp.nr_args; i++)
282                 traceprobe_free_probe_arg(&tu->tp.args[i]);
283 
284         iput(tu->inode);
285         kfree(tu->tp.call.class->system);
286         kfree(tu->tp.call.name);
287         kfree(tu->filename);
288         kfree(tu);
289 }
290 
291 static struct trace_uprobe *find_probe_event(const char *event, const char *group)
292 {
293         struct trace_uprobe *tu;
294 
295         list_for_each_entry(tu, &uprobe_list, list)
296                 if (strcmp(tu->tp.call.name, event) == 0 &&
297                     strcmp(tu->tp.call.class->system, group) == 0)
298                         return tu;
299 
300         return NULL;
301 }
302 
303 /* Unregister a trace_uprobe and probe_event: call with locking uprobe_lock */
304 static int unregister_trace_uprobe(struct trace_uprobe *tu)
305 {
306         int ret;
307 
308         ret = unregister_uprobe_event(tu);
309         if (ret)
310                 return ret;
311 
312         list_del(&tu->list);
313         free_trace_uprobe(tu);
314         return 0;
315 }
316 
317 /* Register a trace_uprobe and probe_event */
318 static int register_trace_uprobe(struct trace_uprobe *tu)
319 {
320         struct trace_uprobe *old_tu;
321         int ret;
322 
323         mutex_lock(&uprobe_lock);
324 
325         /* register as an event */
326         old_tu = find_probe_event(tu->tp.call.name, tu->tp.call.class->system);
327         if (old_tu) {
328                 /* delete old event */
329                 ret = unregister_trace_uprobe(old_tu);
330                 if (ret)
331                         goto end;
332         }
333 
334         ret = register_uprobe_event(tu);
335         if (ret) {
336                 pr_warning("Failed to register probe event(%d)\n", ret);
337                 goto end;
338         }
339 
340         list_add_tail(&tu->list, &uprobe_list);
341 
342 end:
343         mutex_unlock(&uprobe_lock);
344 
345         return ret;
346 }
347 
348 /*
349  * Argument syntax:
350  *  - Add uprobe: p|r[:[GRP/]EVENT] PATH:OFFSET [FETCHARGS]
351  *
352  *  - Remove uprobe: -:[GRP/]EVENT
353  */
354 static int create_trace_uprobe(int argc, char **argv)
355 {
356         struct trace_uprobe *tu;
357         struct inode *inode;
358         char *arg, *event, *group, *filename;
359         char buf[MAX_EVENT_NAME_LEN];
360         struct path path;
361         unsigned long offset;
362         bool is_delete, is_return;
363         int i, ret;
364 
365         inode = NULL;
366         ret = 0;
367         is_delete = false;
368         is_return = false;
369         event = NULL;
370         group = NULL;
371 
372         /* argc must be >= 1 */
373         if (argv[0][0] == '-')
374                 is_delete = true;
375         else if (argv[0][0] == 'r')
376                 is_return = true;
377         else if (argv[0][0] != 'p') {
378                 pr_info("Probe definition must be started with 'p', 'r' or '-'.\n");
379                 return -EINVAL;
380         }
381 
382         if (argv[0][1] == ':') {
383                 event = &argv[0][2];
384                 arg = strchr(event, '/');
385 
386                 if (arg) {
387                         group = event;
388                         event = arg + 1;
389                         event[-1] = '\0';
390 
391                         if (strlen(group) == 0) {
392                                 pr_info("Group name is not specified\n");
393                                 return -EINVAL;
394                         }
395                 }
396                 if (strlen(event) == 0) {
397                         pr_info("Event name is not specified\n");
398                         return -EINVAL;
399                 }
400         }
401         if (!group)
402                 group = UPROBE_EVENT_SYSTEM;
403 
404         if (is_delete) {
405                 int ret;
406 
407                 if (!event) {
408                         pr_info("Delete command needs an event name.\n");
409                         return -EINVAL;
410                 }
411                 mutex_lock(&uprobe_lock);
412                 tu = find_probe_event(event, group);
413 
414                 if (!tu) {
415                         mutex_unlock(&uprobe_lock);
416                         pr_info("Event %s/%s doesn't exist.\n", group, event);
417                         return -ENOENT;
418                 }
419                 /* delete an event */
420                 ret = unregister_trace_uprobe(tu);
421                 mutex_unlock(&uprobe_lock);
422                 return ret;
423         }
424 
425         if (argc < 2) {
426                 pr_info("Probe point is not specified.\n");
427                 return -EINVAL;
428         }
429         if (isdigit(argv[1][0])) {
430                 pr_info("probe point must be have a filename.\n");
431                 return -EINVAL;
432         }
433         arg = strchr(argv[1], ':');
434         if (!arg) {
435                 ret = -EINVAL;
436                 goto fail_address_parse;
437         }
438 
439         *arg++ = '\0';
440         filename = argv[1];
441         ret = kern_path(filename, LOOKUP_FOLLOW, &path);
442         if (ret)
443                 goto fail_address_parse;
444 
445         inode = igrab(path.dentry->d_inode);
446         path_put(&path);
447 
448         if (!inode || !S_ISREG(inode->i_mode)) {
449                 ret = -EINVAL;
450                 goto fail_address_parse;
451         }
452 
453         ret = kstrtoul(arg, 0, &offset);
454         if (ret)
455                 goto fail_address_parse;
456 
457         argc -= 2;
458         argv += 2;
459 
460         /* setup a probe */
461         if (!event) {
462                 char *tail;
463                 char *ptr;
464 
465                 tail = kstrdup(kbasename(filename), GFP_KERNEL);
466                 if (!tail) {
467                         ret = -ENOMEM;
468                         goto fail_address_parse;
469                 }
470 
471                 ptr = strpbrk(tail, ".-_");
472                 if (ptr)
473                         *ptr = '\0';
474 
475                 snprintf(buf, MAX_EVENT_NAME_LEN, "%c_%s_0x%lx", 'p', tail, offset);
476                 event = buf;
477                 kfree(tail);
478         }
479 
480         tu = alloc_trace_uprobe(group, event, argc, is_return);
481         if (IS_ERR(tu)) {
482                 pr_info("Failed to allocate trace_uprobe.(%d)\n", (int)PTR_ERR(tu));
483                 ret = PTR_ERR(tu);
484                 goto fail_address_parse;
485         }
486         tu->offset = offset;
487         tu->inode = inode;
488         tu->filename = kstrdup(filename, GFP_KERNEL);
489 
490         if (!tu->filename) {
491                 pr_info("Failed to allocate filename.\n");
492                 ret = -ENOMEM;
493                 goto error;
494         }
495 
496         /* parse arguments */
497         ret = 0;
498         for (i = 0; i < argc && i < MAX_TRACE_ARGS; i++) {
499                 struct probe_arg *parg = &tu->tp.args[i];
500 
501                 /* Increment count for freeing args in error case */
502                 tu->tp.nr_args++;
503 
504                 /* Parse argument name */
505                 arg = strchr(argv[i], '=');
506                 if (arg) {
507                         *arg++ = '\0';
508                         parg->name = kstrdup(argv[i], GFP_KERNEL);
509                 } else {
510                         arg = argv[i];
511                         /* If argument name is omitted, set "argN" */
512                         snprintf(buf, MAX_EVENT_NAME_LEN, "arg%d", i + 1);
513                         parg->name = kstrdup(buf, GFP_KERNEL);
514                 }
515 
516                 if (!parg->name) {
517                         pr_info("Failed to allocate argument[%d] name.\n", i);
518                         ret = -ENOMEM;
519                         goto error;
520                 }
521 
522                 if (!is_good_name(parg->name)) {
523                         pr_info("Invalid argument[%d] name: %s\n", i, parg->name);
524                         ret = -EINVAL;
525                         goto error;
526                 }
527 
528                 if (traceprobe_conflict_field_name(parg->name, tu->tp.args, i)) {
529                         pr_info("Argument[%d] name '%s' conflicts with "
530                                 "another field.\n", i, argv[i]);
531                         ret = -EINVAL;
532                         goto error;
533                 }
534 
535                 /* Parse fetch argument */
536                 ret = traceprobe_parse_probe_arg(arg, &tu->tp.size, parg,
537                                                  is_return, false);
538                 if (ret) {
539                         pr_info("Parse error at argument[%d]. (%d)\n", i, ret);
540                         goto error;
541                 }
542         }
543 
544         ret = register_trace_uprobe(tu);
545         if (ret)
546                 goto error;
547         return 0;
548 
549 error:
550         free_trace_uprobe(tu);
551         return ret;
552 
553 fail_address_parse:
554         if (inode)
555                 iput(inode);
556 
557         pr_info("Failed to parse address or file.\n");
558 
559         return ret;
560 }
561 
562 static int cleanup_all_probes(void)
563 {
564         struct trace_uprobe *tu;
565         int ret = 0;
566 
567         mutex_lock(&uprobe_lock);
568         while (!list_empty(&uprobe_list)) {
569                 tu = list_entry(uprobe_list.next, struct trace_uprobe, list);
570                 ret = unregister_trace_uprobe(tu);
571                 if (ret)
572                         break;
573         }
574         mutex_unlock(&uprobe_lock);
575         return ret;
576 }
577 
578 /* Probes listing interfaces */
579 static void *probes_seq_start(struct seq_file *m, loff_t *pos)
580 {
581         mutex_lock(&uprobe_lock);
582         return seq_list_start(&uprobe_list, *pos);
583 }
584 
585 static void *probes_seq_next(struct seq_file *m, void *v, loff_t *pos)
586 {
587         return seq_list_next(v, &uprobe_list, pos);
588 }
589 
590 static void probes_seq_stop(struct seq_file *m, void *v)
591 {
592         mutex_unlock(&uprobe_lock);
593 }
594 
595 static int probes_seq_show(struct seq_file *m, void *v)
596 {
597         struct trace_uprobe *tu = v;
598         char c = is_ret_probe(tu) ? 'r' : 'p';
599         int i;
600 
601         seq_printf(m, "%c:%s/%s", c, tu->tp.call.class->system, tu->tp.call.name);
602         seq_printf(m, " %s:0x%p", tu->filename, (void *)tu->offset);
603 
604         for (i = 0; i < tu->tp.nr_args; i++)
605                 seq_printf(m, " %s=%s", tu->tp.args[i].name, tu->tp.args[i].comm);
606 
607         seq_printf(m, "\n");
608         return 0;
609 }
610 
611 static const struct seq_operations probes_seq_op = {
612         .start  = probes_seq_start,
613         .next   = probes_seq_next,
614         .stop   = probes_seq_stop,
615         .show   = probes_seq_show
616 };
617 
618 static int probes_open(struct inode *inode, struct file *file)
619 {
620         int ret;
621 
622         if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
623                 ret = cleanup_all_probes();
624                 if (ret)
625                         return ret;
626         }
627 
628         return seq_open(file, &probes_seq_op);
629 }
630 
631 static ssize_t probes_write(struct file *file, const char __user *buffer,
632                             size_t count, loff_t *ppos)
633 {
634         return traceprobe_probes_write(file, buffer, count, ppos, create_trace_uprobe);
635 }
636 
637 static const struct file_operations uprobe_events_ops = {
638         .owner          = THIS_MODULE,
639         .open           = probes_open,
640         .read           = seq_read,
641         .llseek         = seq_lseek,
642         .release        = seq_release,
643         .write          = probes_write,
644 };
645 
646 /* Probes profiling interfaces */
647 static int probes_profile_seq_show(struct seq_file *m, void *v)
648 {
649         struct trace_uprobe *tu = v;
650 
651         seq_printf(m, "  %s %-44s %15lu\n", tu->filename, tu->tp.call.name, tu->nhit);
652         return 0;
653 }
654 
655 static const struct seq_operations profile_seq_op = {
656         .start  = probes_seq_start,
657         .next   = probes_seq_next,
658         .stop   = probes_seq_stop,
659         .show   = probes_profile_seq_show
660 };
661 
662 static int profile_open(struct inode *inode, struct file *file)
663 {
664         return seq_open(file, &profile_seq_op);
665 }
666 
667 static const struct file_operations uprobe_profile_ops = {
668         .owner          = THIS_MODULE,
669         .open           = profile_open,
670         .read           = seq_read,
671         .llseek         = seq_lseek,
672         .release        = seq_release,
673 };
674 
675 struct uprobe_cpu_buffer {
676         struct mutex mutex;
677         void *buf;
678 };
679 static struct uprobe_cpu_buffer __percpu *uprobe_cpu_buffer;
680 static int uprobe_buffer_refcnt;
681 
682 static int uprobe_buffer_init(void)
683 {
684         int cpu, err_cpu;
685 
686         uprobe_cpu_buffer = alloc_percpu(struct uprobe_cpu_buffer);
687         if (uprobe_cpu_buffer == NULL)
688                 return -ENOMEM;
689 
690         for_each_possible_cpu(cpu) {
691                 struct page *p = alloc_pages_node(cpu_to_node(cpu),
692                                                   GFP_KERNEL, 0);
693                 if (p == NULL) {
694                         err_cpu = cpu;
695                         goto err;
696                 }
697                 per_cpu_ptr(uprobe_cpu_buffer, cpu)->buf = page_address(p);
698                 mutex_init(&per_cpu_ptr(uprobe_cpu_buffer, cpu)->mutex);
699         }
700 
701         return 0;
702 
703 err:
704         for_each_possible_cpu(cpu) {
705                 if (cpu == err_cpu)
706                         break;
707                 free_page((unsigned long)per_cpu_ptr(uprobe_cpu_buffer, cpu)->buf);
708         }
709 
710         free_percpu(uprobe_cpu_buffer);
711         return -ENOMEM;
712 }
713 
714 static int uprobe_buffer_enable(void)
715 {
716         int ret = 0;
717 
718         BUG_ON(!mutex_is_locked(&event_mutex));
719 
720         if (uprobe_buffer_refcnt++ == 0) {
721                 ret = uprobe_buffer_init();
722                 if (ret < 0)
723                         uprobe_buffer_refcnt--;
724         }
725 
726         return ret;
727 }
728 
729 static void uprobe_buffer_disable(void)
730 {
731         int cpu;
732 
733         BUG_ON(!mutex_is_locked(&event_mutex));
734 
735         if (--uprobe_buffer_refcnt == 0) {
736                 for_each_possible_cpu(cpu)
737                         free_page((unsigned long)per_cpu_ptr(uprobe_cpu_buffer,
738                                                              cpu)->buf);
739 
740                 free_percpu(uprobe_cpu_buffer);
741                 uprobe_cpu_buffer = NULL;
742         }
743 }
744 
745 static struct uprobe_cpu_buffer *uprobe_buffer_get(void)
746 {
747         struct uprobe_cpu_buffer *ucb;
748         int cpu;
749 
750         cpu = raw_smp_processor_id();
751         ucb = per_cpu_ptr(uprobe_cpu_buffer, cpu);
752 
753         /*
754          * Use per-cpu buffers for fastest access, but we might migrate
755          * so the mutex makes sure we have sole access to it.
756          */
757         mutex_lock(&ucb->mutex);
758 
759         return ucb;
760 }
761 
762 static void uprobe_buffer_put(struct uprobe_cpu_buffer *ucb)
763 {
764         mutex_unlock(&ucb->mutex);
765 }
766 
767 static void uprobe_trace_print(struct trace_uprobe *tu,
768                                 unsigned long func, struct pt_regs *regs)
769 {
770         struct uprobe_trace_entry_head *entry;
771         struct ring_buffer_event *event;
772         struct ring_buffer *buffer;
773         struct uprobe_cpu_buffer *ucb;
774         void *data;
775         int size, dsize, esize;
776         struct ftrace_event_call *call = &tu->tp.call;
777 
778         dsize = __get_data_size(&tu->tp, regs);
779         esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
780 
781         if (WARN_ON_ONCE(!uprobe_cpu_buffer || tu->tp.size + dsize > PAGE_SIZE))
782                 return;
783 
784         ucb = uprobe_buffer_get();
785         store_trace_args(esize, &tu->tp, regs, ucb->buf, dsize);
786 
787         size = esize + tu->tp.size + dsize;
788         event = trace_current_buffer_lock_reserve(&buffer, call->event.type,
789                                                   size, 0, 0);
790         if (!event)
791                 goto out;
792 
793         entry = ring_buffer_event_data(event);
794         if (is_ret_probe(tu)) {
795                 entry->vaddr[0] = func;
796                 entry->vaddr[1] = instruction_pointer(regs);
797                 data = DATAOF_TRACE_ENTRY(entry, true);
798         } else {
799                 entry->vaddr[0] = instruction_pointer(regs);
800                 data = DATAOF_TRACE_ENTRY(entry, false);
801         }
802 
803         memcpy(data, ucb->buf, tu->tp.size + dsize);
804 
805         if (!call_filter_check_discard(call, entry, buffer, event))
806                 trace_buffer_unlock_commit(buffer, event, 0, 0);
807 
808 out:
809         uprobe_buffer_put(ucb);
810 }
811 
812 /* uprobe handler */
813 static int uprobe_trace_func(struct trace_uprobe *tu, struct pt_regs *regs)
814 {
815         if (!is_ret_probe(tu))
816                 uprobe_trace_print(tu, 0, regs);
817         return 0;
818 }
819 
820 static void uretprobe_trace_func(struct trace_uprobe *tu, unsigned long func,
821                                 struct pt_regs *regs)
822 {
823         uprobe_trace_print(tu, func, regs);
824 }
825 
826 /* Event entry printers */
827 static enum print_line_t
828 print_uprobe_event(struct trace_iterator *iter, int flags, struct trace_event *event)
829 {
830         struct uprobe_trace_entry_head *entry;
831         struct trace_seq *s = &iter->seq;
832         struct trace_uprobe *tu;
833         u8 *data;
834         int i;
835 
836         entry = (struct uprobe_trace_entry_head *)iter->ent;
837         tu = container_of(event, struct trace_uprobe, tp.call.event);
838 
839         if (is_ret_probe(tu)) {
840                 if (!trace_seq_printf(s, "%s: (0x%lx <- 0x%lx)", tu->tp.call.name,
841                                         entry->vaddr[1], entry->vaddr[0]))
842                         goto partial;
843                 data = DATAOF_TRACE_ENTRY(entry, true);
844         } else {
845                 if (!trace_seq_printf(s, "%s: (0x%lx)", tu->tp.call.name,
846                                         entry->vaddr[0]))
847                         goto partial;
848                 data = DATAOF_TRACE_ENTRY(entry, false);
849         }
850 
851         for (i = 0; i < tu->tp.nr_args; i++) {
852                 struct probe_arg *parg = &tu->tp.args[i];
853 
854                 if (!parg->type->print(s, parg->name, data + parg->offset, entry))
855                         goto partial;
856         }
857 
858         if (trace_seq_puts(s, "\n"))
859                 return TRACE_TYPE_HANDLED;
860 
861 partial:
862         return TRACE_TYPE_PARTIAL_LINE;
863 }
864 
865 typedef bool (*filter_func_t)(struct uprobe_consumer *self,
866                                 enum uprobe_filter_ctx ctx,
867                                 struct mm_struct *mm);
868 
869 static int
870 probe_event_enable(struct trace_uprobe *tu, int flag, filter_func_t filter)
871 {
872         int ret = 0;
873 
874         if (trace_probe_is_enabled(&tu->tp))
875                 return -EINTR;
876 
877         ret = uprobe_buffer_enable();
878         if (ret < 0)
879                 return ret;
880 
881         WARN_ON(!uprobe_filter_is_empty(&tu->filter));
882 
883         tu->tp.flags |= flag;
884         tu->consumer.filter = filter;
885         ret = uprobe_register(tu->inode, tu->offset, &tu->consumer);
886         if (ret)
887                 tu->tp.flags &= ~flag;
888 
889         return ret;
890 }
891 
892 static void probe_event_disable(struct trace_uprobe *tu, int flag)
893 {
894         if (!trace_probe_is_enabled(&tu->tp))
895                 return;
896 
897         WARN_ON(!uprobe_filter_is_empty(&tu->filter));
898 
899         uprobe_unregister(tu->inode, tu->offset, &tu->consumer);
900         tu->tp.flags &= ~flag;
901 
902         uprobe_buffer_disable();
903 }
904 
905 static int uprobe_event_define_fields(struct ftrace_event_call *event_call)
906 {
907         int ret, i, size;
908         struct uprobe_trace_entry_head field;
909         struct trace_uprobe *tu = event_call->data;
910 
911         if (is_ret_probe(tu)) {
912                 DEFINE_FIELD(unsigned long, vaddr[0], FIELD_STRING_FUNC, 0);
913                 DEFINE_FIELD(unsigned long, vaddr[1], FIELD_STRING_RETIP, 0);
914                 size = SIZEOF_TRACE_ENTRY(true);
915         } else {
916                 DEFINE_FIELD(unsigned long, vaddr[0], FIELD_STRING_IP, 0);
917                 size = SIZEOF_TRACE_ENTRY(false);
918         }
919         /* Set argument names as fields */
920         for (i = 0; i < tu->tp.nr_args; i++) {
921                 struct probe_arg *parg = &tu->tp.args[i];
922 
923                 ret = trace_define_field(event_call, parg->type->fmttype,
924                                          parg->name, size + parg->offset,
925                                          parg->type->size, parg->type->is_signed,
926                                          FILTER_OTHER);
927 
928                 if (ret)
929                         return ret;
930         }
931         return 0;
932 }
933 
934 #ifdef CONFIG_PERF_EVENTS
935 static bool
936 __uprobe_perf_filter(struct trace_uprobe_filter *filter, struct mm_struct *mm)
937 {
938         struct perf_event *event;
939 
940         if (filter->nr_systemwide)
941                 return true;
942 
943         list_for_each_entry(event, &filter->perf_events, hw.tp_list) {
944                 if (event->hw.tp_target->mm == mm)
945                         return true;
946         }
947 
948         return false;
949 }
950 
951 static inline bool
952 uprobe_filter_event(struct trace_uprobe *tu, struct perf_event *event)
953 {
954         return __uprobe_perf_filter(&tu->filter, event->hw.tp_target->mm);
955 }
956 
957 static int uprobe_perf_open(struct trace_uprobe *tu, struct perf_event *event)
958 {
959         bool done;
960 
961         write_lock(&tu->filter.rwlock);
962         if (event->hw.tp_target) {
963                 /*
964                  * event->parent != NULL means copy_process(), we can avoid
965                  * uprobe_apply(). current->mm must be probed and we can rely
966                  * on dup_mmap() which preserves the already installed bp's.
967                  *
968                  * attr.enable_on_exec means that exec/mmap will install the
969                  * breakpoints we need.
970                  */
971                 done = tu->filter.nr_systemwide ||
972                         event->parent || event->attr.enable_on_exec ||
973                         uprobe_filter_event(tu, event);
974                 list_add(&event->hw.tp_list, &tu->filter.perf_events);
975         } else {
976                 done = tu->filter.nr_systemwide;
977                 tu->filter.nr_systemwide++;
978         }
979         write_unlock(&tu->filter.rwlock);
980 
981         if (!done)
982                 uprobe_apply(tu->inode, tu->offset, &tu->consumer, true);
983 
984         return 0;
985 }
986 
987 static int uprobe_perf_close(struct trace_uprobe *tu, struct perf_event *event)
988 {
989         bool done;
990 
991         write_lock(&tu->filter.rwlock);
992         if (event->hw.tp_target) {
993                 list_del(&event->hw.tp_list);
994                 done = tu->filter.nr_systemwide ||
995                         (event->hw.tp_target->flags & PF_EXITING) ||
996                         uprobe_filter_event(tu, event);
997         } else {
998                 tu->filter.nr_systemwide--;
999                 done = tu->filter.nr_systemwide;
1000         }
1001         write_unlock(&tu->filter.rwlock);
1002 
1003         if (!done)
1004                 uprobe_apply(tu->inode, tu->offset, &tu->consumer, false);
1005 
1006         return 0;
1007 }
1008 
1009 static bool uprobe_perf_filter(struct uprobe_consumer *uc,
1010                                 enum uprobe_filter_ctx ctx, struct mm_struct *mm)
1011 {
1012         struct trace_uprobe *tu;
1013         int ret;
1014 
1015         tu = container_of(uc, struct trace_uprobe, consumer);
1016         read_lock(&tu->filter.rwlock);
1017         ret = __uprobe_perf_filter(&tu->filter, mm);
1018         read_unlock(&tu->filter.rwlock);
1019 
1020         return ret;
1021 }
1022 
1023 static void uprobe_perf_print(struct trace_uprobe *tu,
1024                                 unsigned long func, struct pt_regs *regs)
1025 {
1026         struct ftrace_event_call *call = &tu->tp.call;
1027         struct uprobe_trace_entry_head *entry;
1028         struct hlist_head *head;
1029         struct uprobe_cpu_buffer *ucb;
1030         void *data;
1031         int size, dsize, esize;
1032         int rctx;
1033 
1034         dsize = __get_data_size(&tu->tp, regs);
1035         esize = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
1036 
1037         if (WARN_ON_ONCE(!uprobe_cpu_buffer))
1038                 return;
1039 
1040         size = esize + tu->tp.size + dsize;
1041         size = ALIGN(size + sizeof(u32), sizeof(u64)) - sizeof(u32);
1042         if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE, "profile buffer not large enough"))
1043                 return;
1044 
1045         ucb = uprobe_buffer_get();
1046         store_trace_args(esize, &tu->tp, regs, ucb->buf, dsize);
1047 
1048         preempt_disable();
1049         head = this_cpu_ptr(call->perf_events);
1050         if (hlist_empty(head))
1051                 goto out;
1052 
1053         entry = perf_trace_buf_prepare(size, call->event.type, regs, &rctx);
1054         if (!entry)
1055                 goto out;
1056 
1057         if (is_ret_probe(tu)) {
1058                 entry->vaddr[0] = func;
1059                 entry->vaddr[1] = instruction_pointer(regs);
1060                 data = DATAOF_TRACE_ENTRY(entry, true);
1061         } else {
1062                 entry->vaddr[0] = instruction_pointer(regs);
1063                 data = DATAOF_TRACE_ENTRY(entry, false);
1064         }
1065 
1066         memcpy(data, ucb->buf, tu->tp.size + dsize);
1067 
1068         if (size - esize > tu->tp.size + dsize) {
1069                 int len = tu->tp.size + dsize;
1070 
1071                 memset(data + len, 0, size - esize - len);
1072         }
1073 
1074         perf_trace_buf_submit(entry, size, rctx, 0, 1, regs, head, NULL);
1075  out:
1076         preempt_enable();
1077         uprobe_buffer_put(ucb);
1078 }
1079 
1080 /* uprobe profile handler */
1081 static int uprobe_perf_func(struct trace_uprobe *tu, struct pt_regs *regs)
1082 {
1083         if (!uprobe_perf_filter(&tu->consumer, 0, current->mm))
1084                 return UPROBE_HANDLER_REMOVE;
1085 
1086         if (!is_ret_probe(tu))
1087                 uprobe_perf_print(tu, 0, regs);
1088         return 0;
1089 }
1090 
1091 static void uretprobe_perf_func(struct trace_uprobe *tu, unsigned long func,
1092                                 struct pt_regs *regs)
1093 {
1094         uprobe_perf_print(tu, func, regs);
1095 }
1096 #endif  /* CONFIG_PERF_EVENTS */
1097 
1098 static
1099 int trace_uprobe_register(struct ftrace_event_call *event, enum trace_reg type, void *data)
1100 {
1101         struct trace_uprobe *tu = event->data;
1102 
1103         switch (type) {
1104         case TRACE_REG_REGISTER:
1105                 return probe_event_enable(tu, TP_FLAG_TRACE, NULL);
1106 
1107         case TRACE_REG_UNREGISTER:
1108                 probe_event_disable(tu, TP_FLAG_TRACE);
1109                 return 0;
1110 
1111 #ifdef CONFIG_PERF_EVENTS
1112         case TRACE_REG_PERF_REGISTER:
1113                 return probe_event_enable(tu, TP_FLAG_PROFILE, uprobe_perf_filter);
1114 
1115         case TRACE_REG_PERF_UNREGISTER:
1116                 probe_event_disable(tu, TP_FLAG_PROFILE);
1117                 return 0;
1118 
1119         case TRACE_REG_PERF_OPEN:
1120                 return uprobe_perf_open(tu, data);
1121 
1122         case TRACE_REG_PERF_CLOSE:
1123                 return uprobe_perf_close(tu, data);
1124 
1125 #endif
1126         default:
1127                 return 0;
1128         }
1129         return 0;
1130 }
1131 
1132 static int uprobe_dispatcher(struct uprobe_consumer *con, struct pt_regs *regs)
1133 {
1134         struct trace_uprobe *tu;
1135         struct uprobe_dispatch_data udd;
1136         int ret = 0;
1137 
1138         tu = container_of(con, struct trace_uprobe, consumer);
1139         tu->nhit++;
1140 
1141         udd.tu = tu;
1142         udd.bp_addr = instruction_pointer(regs);
1143 
1144         current->utask->vaddr = (unsigned long) &udd;
1145 
1146         if (tu->tp.flags & TP_FLAG_TRACE)
1147                 ret |= uprobe_trace_func(tu, regs);
1148 
1149 #ifdef CONFIG_PERF_EVENTS
1150         if (tu->tp.flags & TP_FLAG_PROFILE)
1151                 ret |= uprobe_perf_func(tu, regs);
1152 #endif
1153         return ret;
1154 }
1155 
1156 static int uretprobe_dispatcher(struct uprobe_consumer *con,
1157                                 unsigned long func, struct pt_regs *regs)
1158 {
1159         struct trace_uprobe *tu;
1160         struct uprobe_dispatch_data udd;
1161 
1162         tu = container_of(con, struct trace_uprobe, consumer);
1163 
1164         udd.tu = tu;
1165         udd.bp_addr = func;
1166 
1167         current->utask->vaddr = (unsigned long) &udd;
1168 
1169         if (tu->tp.flags & TP_FLAG_TRACE)
1170                 uretprobe_trace_func(tu, func, regs);
1171 
1172 #ifdef CONFIG_PERF_EVENTS
1173         if (tu->tp.flags & TP_FLAG_PROFILE)
1174                 uretprobe_perf_func(tu, func, regs);
1175 #endif
1176         return 0;
1177 }
1178 
1179 static struct trace_event_functions uprobe_funcs = {
1180         .trace          = print_uprobe_event
1181 };
1182 
1183 static int register_uprobe_event(struct trace_uprobe *tu)
1184 {
1185         struct ftrace_event_call *call = &tu->tp.call;
1186         int ret;
1187 
1188         /* Initialize ftrace_event_call */
1189         INIT_LIST_HEAD(&call->class->fields);
1190         call->event.funcs = &uprobe_funcs;
1191         call->class->define_fields = uprobe_event_define_fields;
1192 
1193         if (set_print_fmt(&tu->tp, is_ret_probe(tu)) < 0)
1194                 return -ENOMEM;
1195 
1196         ret = register_ftrace_event(&call->event);
1197         if (!ret) {
1198                 kfree(call->print_fmt);
1199                 return -ENODEV;
1200         }
1201         call->flags = 0;
1202         call->class->reg = trace_uprobe_register;
1203         call->data = tu;
1204         ret = trace_add_event_call(call);
1205 
1206         if (ret) {
1207                 pr_info("Failed to register uprobe event: %s\n", call->name);
1208                 kfree(call->print_fmt);
1209                 unregister_ftrace_event(&call->event);
1210         }
1211 
1212         return ret;
1213 }
1214 
1215 static int unregister_uprobe_event(struct trace_uprobe *tu)
1216 {
1217         int ret;
1218 
1219         /* tu->event is unregistered in trace_remove_event_call() */
1220         ret = trace_remove_event_call(&tu->tp.call);
1221         if (ret)
1222                 return ret;
1223         kfree(tu->tp.call.print_fmt);
1224         tu->tp.call.print_fmt = NULL;
1225         return 0;
1226 }
1227 
1228 /* Make a trace interface for controling probe points */
1229 static __init int init_uprobe_trace(void)
1230 {
1231         struct dentry *d_tracer;
1232 
1233         d_tracer = tracing_init_dentry();
1234         if (!d_tracer)
1235                 return 0;
1236 
1237         trace_create_file("uprobe_events", 0644, d_tracer,
1238                                     NULL, &uprobe_events_ops);
1239         /* Profile interface */
1240         trace_create_file("uprobe_profile", 0444, d_tracer,
1241                                     NULL, &uprobe_profile_ops);
1242         return 0;
1243 }
1244 
1245 fs_initcall(init_uprobe_trace);
1246 

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

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

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

osdn.jp