]>
Commit | Line | Data |
---|---|---|
85c8721f | 1 | /* auditsc.c -- System-call auditing support |
1da177e4 LT |
2 | * Handles all system-call specific auditing features. |
3 | * | |
4 | * Copyright 2003-2004 Red Hat Inc., Durham, North Carolina. | |
5 | * All Rights Reserved. | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Written by Rickard E. (Rik) Faith <faith@redhat.com> | |
22 | * | |
23 | * Many of the ideas implemented here are from Stephen C. Tweedie, | |
24 | * especially the idea of avoiding a copy by using getname. | |
25 | * | |
26 | * The method for actual interception of syscall entry and exit (not in | |
27 | * this file -- see entry.S) is based on a GPL'd patch written by | |
28 | * okir@suse.de and Copyright 2003 SuSE Linux AG. | |
29 | * | |
30 | */ | |
31 | ||
32 | #include <linux/init.h> | |
33 | #include <asm/atomic.h> | |
34 | #include <asm/types.h> | |
35 | #include <linux/mm.h> | |
36 | #include <linux/module.h> | |
3ec3b2fb | 37 | #include <linux/socket.h> |
1da177e4 LT |
38 | #include <linux/audit.h> |
39 | #include <linux/personality.h> | |
40 | #include <linux/time.h> | |
41 | #include <asm/unistd.h> | |
42 | ||
43 | /* 0 = no checking | |
44 | 1 = put_count checking | |
45 | 2 = verbose put_count checking | |
46 | */ | |
47 | #define AUDIT_DEBUG 0 | |
48 | ||
49 | /* No syscall auditing will take place unless audit_enabled != 0. */ | |
50 | extern int audit_enabled; | |
51 | ||
52 | /* AUDIT_NAMES is the number of slots we reserve in the audit_context | |
53 | * for saving names from getname(). */ | |
54 | #define AUDIT_NAMES 20 | |
55 | ||
56 | /* AUDIT_NAMES_RESERVED is the number of slots we reserve in the | |
57 | * audit_context from being used for nameless inodes from | |
58 | * path_lookup. */ | |
59 | #define AUDIT_NAMES_RESERVED 7 | |
60 | ||
61 | /* At task start time, the audit_state is set in the audit_context using | |
62 | a per-task filter. At syscall entry, the audit_state is augmented by | |
63 | the syscall filter. */ | |
64 | enum audit_state { | |
65 | AUDIT_DISABLED, /* Do not create per-task audit_context. | |
66 | * No syscall-specific audit records can | |
67 | * be generated. */ | |
68 | AUDIT_SETUP_CONTEXT, /* Create the per-task audit_context, | |
69 | * but don't necessarily fill it in at | |
70 | * syscall entry time (i.e., filter | |
71 | * instead). */ | |
72 | AUDIT_BUILD_CONTEXT, /* Create the per-task audit_context, | |
73 | * and always fill it in at syscall | |
74 | * entry time. This makes a full | |
75 | * syscall record available if some | |
76 | * other part of the kernel decides it | |
77 | * should be recorded. */ | |
78 | AUDIT_RECORD_CONTEXT /* Create the per-task audit_context, | |
79 | * always fill it in at syscall entry | |
80 | * time, and always write out the audit | |
81 | * record at syscall exit time. */ | |
82 | }; | |
83 | ||
84 | /* When fs/namei.c:getname() is called, we store the pointer in name and | |
85 | * we don't let putname() free it (instead we free all of the saved | |
86 | * pointers at syscall exit time). | |
87 | * | |
88 | * Further, in fs/namei.c:path_lookup() we store the inode and device. */ | |
89 | struct audit_names { | |
90 | const char *name; | |
91 | unsigned long ino; | |
92 | dev_t dev; | |
93 | umode_t mode; | |
94 | uid_t uid; | |
95 | gid_t gid; | |
96 | dev_t rdev; | |
97 | }; | |
98 | ||
99 | struct audit_aux_data { | |
100 | struct audit_aux_data *next; | |
101 | int type; | |
102 | }; | |
103 | ||
104 | #define AUDIT_AUX_IPCPERM 0 | |
105 | ||
106 | struct audit_aux_data_ipcctl { | |
107 | struct audit_aux_data d; | |
108 | struct ipc_perm p; | |
109 | unsigned long qbytes; | |
110 | uid_t uid; | |
111 | gid_t gid; | |
112 | mode_t mode; | |
113 | }; | |
114 | ||
3ec3b2fb DW |
115 | struct audit_aux_data_socketcall { |
116 | struct audit_aux_data d; | |
117 | int nargs; | |
118 | unsigned long args[0]; | |
119 | }; | |
120 | ||
121 | struct audit_aux_data_sockaddr { | |
122 | struct audit_aux_data d; | |
123 | int len; | |
124 | char a[0]; | |
125 | }; | |
126 | ||
1da177e4 LT |
127 | |
128 | /* The per-task audit context. */ | |
129 | struct audit_context { | |
130 | int in_syscall; /* 1 if task is in a syscall */ | |
131 | enum audit_state state; | |
132 | unsigned int serial; /* serial number for record */ | |
133 | struct timespec ctime; /* time of syscall entry */ | |
134 | uid_t loginuid; /* login uid (identity) */ | |
135 | int major; /* syscall number */ | |
136 | unsigned long argv[4]; /* syscall arguments */ | |
137 | int return_valid; /* return code is valid */ | |
2fd6f58b | 138 | long return_code;/* syscall return code */ |
1da177e4 LT |
139 | int auditable; /* 1 if record should be written */ |
140 | int name_count; | |
141 | struct audit_names names[AUDIT_NAMES]; | |
142 | struct audit_context *previous; /* For nested syscalls */ | |
143 | struct audit_aux_data *aux; | |
144 | ||
145 | /* Save things to print about task_struct */ | |
146 | pid_t pid; | |
147 | uid_t uid, euid, suid, fsuid; | |
148 | gid_t gid, egid, sgid, fsgid; | |
149 | unsigned long personality; | |
2fd6f58b | 150 | int arch; |
1da177e4 LT |
151 | |
152 | #if AUDIT_DEBUG | |
153 | int put_count; | |
154 | int ino_count; | |
155 | #endif | |
156 | }; | |
157 | ||
158 | /* Public API */ | |
159 | /* There are three lists of rules -- one to search at task creation | |
160 | * time, one to search at syscall entry time, and another to search at | |
161 | * syscall exit time. */ | |
162 | static LIST_HEAD(audit_tsklist); | |
163 | static LIST_HEAD(audit_entlist); | |
164 | static LIST_HEAD(audit_extlist); | |
165 | ||
166 | struct audit_entry { | |
167 | struct list_head list; | |
168 | struct rcu_head rcu; | |
169 | struct audit_rule rule; | |
170 | }; | |
171 | ||
172 | /* Check to see if two rules are identical. It is called from | |
173 | * audit_del_rule during AUDIT_DEL. */ | |
174 | static int audit_compare_rule(struct audit_rule *a, struct audit_rule *b) | |
175 | { | |
176 | int i; | |
177 | ||
178 | if (a->flags != b->flags) | |
179 | return 1; | |
180 | ||
181 | if (a->action != b->action) | |
182 | return 1; | |
183 | ||
184 | if (a->field_count != b->field_count) | |
185 | return 1; | |
186 | ||
187 | for (i = 0; i < a->field_count; i++) { | |
188 | if (a->fields[i] != b->fields[i] | |
189 | || a->values[i] != b->values[i]) | |
190 | return 1; | |
191 | } | |
192 | ||
193 | for (i = 0; i < AUDIT_BITMASK_SIZE; i++) | |
194 | if (a->mask[i] != b->mask[i]) | |
195 | return 1; | |
196 | ||
197 | return 0; | |
198 | } | |
199 | ||
200 | /* Note that audit_add_rule and audit_del_rule are called via | |
201 | * audit_receive() in audit.c, and are protected by | |
202 | * audit_netlink_sem. */ | |
203 | static inline int audit_add_rule(struct audit_entry *entry, | |
204 | struct list_head *list) | |
205 | { | |
206 | if (entry->rule.flags & AUDIT_PREPEND) { | |
207 | entry->rule.flags &= ~AUDIT_PREPEND; | |
208 | list_add_rcu(&entry->list, list); | |
209 | } else { | |
210 | list_add_tail_rcu(&entry->list, list); | |
211 | } | |
212 | return 0; | |
213 | } | |
214 | ||
215 | static void audit_free_rule(struct rcu_head *head) | |
216 | { | |
217 | struct audit_entry *e = container_of(head, struct audit_entry, rcu); | |
218 | kfree(e); | |
219 | } | |
220 | ||
221 | /* Note that audit_add_rule and audit_del_rule are called via | |
222 | * audit_receive() in audit.c, and are protected by | |
223 | * audit_netlink_sem. */ | |
224 | static inline int audit_del_rule(struct audit_rule *rule, | |
225 | struct list_head *list) | |
226 | { | |
227 | struct audit_entry *e; | |
228 | ||
229 | /* Do not use the _rcu iterator here, since this is the only | |
230 | * deletion routine. */ | |
231 | list_for_each_entry(e, list, list) { | |
232 | if (!audit_compare_rule(rule, &e->rule)) { | |
233 | list_del_rcu(&e->list); | |
234 | call_rcu(&e->rcu, audit_free_rule); | |
235 | return 0; | |
236 | } | |
237 | } | |
238 | return -EFAULT; /* No matching rule */ | |
239 | } | |
240 | ||
1da177e4 LT |
241 | /* Copy rule from user-space to kernel-space. Called during |
242 | * AUDIT_ADD. */ | |
243 | static int audit_copy_rule(struct audit_rule *d, struct audit_rule *s) | |
244 | { | |
245 | int i; | |
246 | ||
247 | if (s->action != AUDIT_NEVER | |
248 | && s->action != AUDIT_POSSIBLE | |
249 | && s->action != AUDIT_ALWAYS) | |
250 | return -1; | |
251 | if (s->field_count < 0 || s->field_count > AUDIT_MAX_FIELDS) | |
252 | return -1; | |
253 | ||
254 | d->flags = s->flags; | |
255 | d->action = s->action; | |
256 | d->field_count = s->field_count; | |
257 | for (i = 0; i < d->field_count; i++) { | |
258 | d->fields[i] = s->fields[i]; | |
259 | d->values[i] = s->values[i]; | |
260 | } | |
261 | for (i = 0; i < AUDIT_BITMASK_SIZE; i++) d->mask[i] = s->mask[i]; | |
262 | return 0; | |
263 | } | |
264 | ||
c94c257c SH |
265 | int audit_receive_filter(int type, int pid, int uid, int seq, void *data, |
266 | uid_t loginuid) | |
1da177e4 LT |
267 | { |
268 | u32 flags; | |
269 | struct audit_entry *entry; | |
270 | int err = 0; | |
271 | ||
272 | switch (type) { | |
273 | case AUDIT_LIST: | |
274 | /* The *_rcu iterators not needed here because we are | |
275 | always called with audit_netlink_sem held. */ | |
276 | list_for_each_entry(entry, &audit_tsklist, list) | |
277 | audit_send_reply(pid, seq, AUDIT_LIST, 0, 1, | |
278 | &entry->rule, sizeof(entry->rule)); | |
279 | list_for_each_entry(entry, &audit_entlist, list) | |
280 | audit_send_reply(pid, seq, AUDIT_LIST, 0, 1, | |
281 | &entry->rule, sizeof(entry->rule)); | |
282 | list_for_each_entry(entry, &audit_extlist, list) | |
283 | audit_send_reply(pid, seq, AUDIT_LIST, 0, 1, | |
284 | &entry->rule, sizeof(entry->rule)); | |
285 | audit_send_reply(pid, seq, AUDIT_LIST, 1, 1, NULL, 0); | |
286 | break; | |
287 | case AUDIT_ADD: | |
288 | if (!(entry = kmalloc(sizeof(*entry), GFP_KERNEL))) | |
289 | return -ENOMEM; | |
290 | if (audit_copy_rule(&entry->rule, data)) { | |
291 | kfree(entry); | |
292 | return -EINVAL; | |
293 | } | |
294 | flags = entry->rule.flags; | |
295 | if (!err && (flags & AUDIT_PER_TASK)) | |
296 | err = audit_add_rule(entry, &audit_tsklist); | |
297 | if (!err && (flags & AUDIT_AT_ENTRY)) | |
298 | err = audit_add_rule(entry, &audit_entlist); | |
299 | if (!err && (flags & AUDIT_AT_EXIT)) | |
300 | err = audit_add_rule(entry, &audit_extlist); | |
c0404993 SG |
301 | audit_log(NULL, AUDIT_CONFIG_CHANGE, |
302 | "auid %u added an audit rule\n", loginuid); | |
1da177e4 LT |
303 | break; |
304 | case AUDIT_DEL: | |
305 | flags =((struct audit_rule *)data)->flags; | |
306 | if (!err && (flags & AUDIT_PER_TASK)) | |
307 | err = audit_del_rule(data, &audit_tsklist); | |
308 | if (!err && (flags & AUDIT_AT_ENTRY)) | |
309 | err = audit_del_rule(data, &audit_entlist); | |
310 | if (!err && (flags & AUDIT_AT_EXIT)) | |
311 | err = audit_del_rule(data, &audit_extlist); | |
c0404993 SG |
312 | audit_log(NULL, AUDIT_CONFIG_CHANGE, |
313 | "auid %u removed an audit rule\n", loginuid); | |
1da177e4 LT |
314 | break; |
315 | default: | |
316 | return -EINVAL; | |
317 | } | |
318 | ||
319 | return err; | |
320 | } | |
1da177e4 LT |
321 | |
322 | /* Compare a task_struct with an audit_rule. Return 1 on match, 0 | |
323 | * otherwise. */ | |
324 | static int audit_filter_rules(struct task_struct *tsk, | |
325 | struct audit_rule *rule, | |
326 | struct audit_context *ctx, | |
327 | enum audit_state *state) | |
328 | { | |
329 | int i, j; | |
330 | ||
331 | for (i = 0; i < rule->field_count; i++) { | |
332 | u32 field = rule->fields[i] & ~AUDIT_NEGATE; | |
333 | u32 value = rule->values[i]; | |
334 | int result = 0; | |
335 | ||
336 | switch (field) { | |
337 | case AUDIT_PID: | |
338 | result = (tsk->pid == value); | |
339 | break; | |
340 | case AUDIT_UID: | |
341 | result = (tsk->uid == value); | |
342 | break; | |
343 | case AUDIT_EUID: | |
344 | result = (tsk->euid == value); | |
345 | break; | |
346 | case AUDIT_SUID: | |
347 | result = (tsk->suid == value); | |
348 | break; | |
349 | case AUDIT_FSUID: | |
350 | result = (tsk->fsuid == value); | |
351 | break; | |
352 | case AUDIT_GID: | |
353 | result = (tsk->gid == value); | |
354 | break; | |
355 | case AUDIT_EGID: | |
356 | result = (tsk->egid == value); | |
357 | break; | |
358 | case AUDIT_SGID: | |
359 | result = (tsk->sgid == value); | |
360 | break; | |
361 | case AUDIT_FSGID: | |
362 | result = (tsk->fsgid == value); | |
363 | break; | |
364 | case AUDIT_PERS: | |
365 | result = (tsk->personality == value); | |
366 | break; | |
2fd6f58b DW |
367 | case AUDIT_ARCH: |
368 | if (ctx) | |
369 | result = (ctx->arch == value); | |
370 | break; | |
1da177e4 LT |
371 | |
372 | case AUDIT_EXIT: | |
373 | if (ctx && ctx->return_valid) | |
374 | result = (ctx->return_code == value); | |
375 | break; | |
376 | case AUDIT_SUCCESS: | |
377 | if (ctx && ctx->return_valid) | |
2fd6f58b | 378 | result = (ctx->return_valid == AUDITSC_SUCCESS); |
1da177e4 LT |
379 | break; |
380 | case AUDIT_DEVMAJOR: | |
381 | if (ctx) { | |
382 | for (j = 0; j < ctx->name_count; j++) { | |
383 | if (MAJOR(ctx->names[j].dev)==value) { | |
384 | ++result; | |
385 | break; | |
386 | } | |
387 | } | |
388 | } | |
389 | break; | |
390 | case AUDIT_DEVMINOR: | |
391 | if (ctx) { | |
392 | for (j = 0; j < ctx->name_count; j++) { | |
393 | if (MINOR(ctx->names[j].dev)==value) { | |
394 | ++result; | |
395 | break; | |
396 | } | |
397 | } | |
398 | } | |
399 | break; | |
400 | case AUDIT_INODE: | |
401 | if (ctx) { | |
402 | for (j = 0; j < ctx->name_count; j++) { | |
403 | if (ctx->names[j].ino == value) { | |
404 | ++result; | |
405 | break; | |
406 | } | |
407 | } | |
408 | } | |
409 | break; | |
410 | case AUDIT_LOGINUID: | |
411 | result = 0; | |
412 | if (ctx) | |
413 | result = (ctx->loginuid == value); | |
414 | break; | |
415 | case AUDIT_ARG0: | |
416 | case AUDIT_ARG1: | |
417 | case AUDIT_ARG2: | |
418 | case AUDIT_ARG3: | |
419 | if (ctx) | |
420 | result = (ctx->argv[field-AUDIT_ARG0]==value); | |
421 | break; | |
422 | } | |
423 | ||
424 | if (rule->fields[i] & AUDIT_NEGATE) | |
425 | result = !result; | |
426 | if (!result) | |
427 | return 0; | |
428 | } | |
429 | switch (rule->action) { | |
430 | case AUDIT_NEVER: *state = AUDIT_DISABLED; break; | |
431 | case AUDIT_POSSIBLE: *state = AUDIT_BUILD_CONTEXT; break; | |
432 | case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break; | |
433 | } | |
434 | return 1; | |
435 | } | |
436 | ||
437 | /* At process creation time, we can determine if system-call auditing is | |
438 | * completely disabled for this task. Since we only have the task | |
439 | * structure at this point, we can only check uid and gid. | |
440 | */ | |
441 | static enum audit_state audit_filter_task(struct task_struct *tsk) | |
442 | { | |
443 | struct audit_entry *e; | |
444 | enum audit_state state; | |
445 | ||
446 | rcu_read_lock(); | |
447 | list_for_each_entry_rcu(e, &audit_tsklist, list) { | |
448 | if (audit_filter_rules(tsk, &e->rule, NULL, &state)) { | |
449 | rcu_read_unlock(); | |
450 | return state; | |
451 | } | |
452 | } | |
453 | rcu_read_unlock(); | |
454 | return AUDIT_BUILD_CONTEXT; | |
455 | } | |
456 | ||
457 | /* At syscall entry and exit time, this filter is called if the | |
458 | * audit_state is not low enough that auditing cannot take place, but is | |
23f32d18 | 459 | * also not high enough that we already know we have to write an audit |
1da177e4 LT |
460 | * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT). |
461 | */ | |
462 | static enum audit_state audit_filter_syscall(struct task_struct *tsk, | |
463 | struct audit_context *ctx, | |
464 | struct list_head *list) | |
465 | { | |
466 | struct audit_entry *e; | |
467 | enum audit_state state; | |
468 | int word = AUDIT_WORD(ctx->major); | |
469 | int bit = AUDIT_BIT(ctx->major); | |
470 | ||
471 | rcu_read_lock(); | |
472 | list_for_each_entry_rcu(e, list, list) { | |
473 | if ((e->rule.mask[word] & bit) == bit | |
474 | && audit_filter_rules(tsk, &e->rule, ctx, &state)) { | |
475 | rcu_read_unlock(); | |
476 | return state; | |
477 | } | |
478 | } | |
479 | rcu_read_unlock(); | |
480 | return AUDIT_BUILD_CONTEXT; | |
481 | } | |
482 | ||
483 | /* This should be called with task_lock() held. */ | |
484 | static inline struct audit_context *audit_get_context(struct task_struct *tsk, | |
485 | int return_valid, | |
486 | int return_code) | |
487 | { | |
488 | struct audit_context *context = tsk->audit_context; | |
489 | ||
490 | if (likely(!context)) | |
491 | return NULL; | |
492 | context->return_valid = return_valid; | |
493 | context->return_code = return_code; | |
494 | ||
495 | if (context->in_syscall && !context->auditable) { | |
496 | enum audit_state state; | |
497 | state = audit_filter_syscall(tsk, context, &audit_extlist); | |
498 | if (state == AUDIT_RECORD_CONTEXT) | |
499 | context->auditable = 1; | |
500 | } | |
501 | ||
502 | context->pid = tsk->pid; | |
503 | context->uid = tsk->uid; | |
504 | context->gid = tsk->gid; | |
505 | context->euid = tsk->euid; | |
506 | context->suid = tsk->suid; | |
507 | context->fsuid = tsk->fsuid; | |
508 | context->egid = tsk->egid; | |
509 | context->sgid = tsk->sgid; | |
510 | context->fsgid = tsk->fsgid; | |
511 | context->personality = tsk->personality; | |
512 | tsk->audit_context = NULL; | |
513 | return context; | |
514 | } | |
515 | ||
516 | static inline void audit_free_names(struct audit_context *context) | |
517 | { | |
518 | int i; | |
519 | ||
520 | #if AUDIT_DEBUG == 2 | |
521 | if (context->auditable | |
522 | ||context->put_count + context->ino_count != context->name_count) { | |
523 | printk(KERN_ERR "audit.c:%d(:%d): major=%d in_syscall=%d" | |
524 | " name_count=%d put_count=%d" | |
525 | " ino_count=%d [NOT freeing]\n", | |
526 | __LINE__, | |
527 | context->serial, context->major, context->in_syscall, | |
528 | context->name_count, context->put_count, | |
529 | context->ino_count); | |
530 | for (i = 0; i < context->name_count; i++) | |
531 | printk(KERN_ERR "names[%d] = %p = %s\n", i, | |
532 | context->names[i].name, | |
533 | context->names[i].name); | |
534 | dump_stack(); | |
535 | return; | |
536 | } | |
537 | #endif | |
538 | #if AUDIT_DEBUG | |
539 | context->put_count = 0; | |
540 | context->ino_count = 0; | |
541 | #endif | |
542 | ||
543 | for (i = 0; i < context->name_count; i++) | |
544 | if (context->names[i].name) | |
545 | __putname(context->names[i].name); | |
546 | context->name_count = 0; | |
547 | } | |
548 | ||
549 | static inline void audit_free_aux(struct audit_context *context) | |
550 | { | |
551 | struct audit_aux_data *aux; | |
552 | ||
553 | while ((aux = context->aux)) { | |
554 | context->aux = aux->next; | |
555 | kfree(aux); | |
556 | } | |
557 | } | |
558 | ||
559 | static inline void audit_zero_context(struct audit_context *context, | |
560 | enum audit_state state) | |
561 | { | |
562 | uid_t loginuid = context->loginuid; | |
563 | ||
564 | memset(context, 0, sizeof(*context)); | |
565 | context->state = state; | |
566 | context->loginuid = loginuid; | |
567 | } | |
568 | ||
569 | static inline struct audit_context *audit_alloc_context(enum audit_state state) | |
570 | { | |
571 | struct audit_context *context; | |
572 | ||
573 | if (!(context = kmalloc(sizeof(*context), GFP_KERNEL))) | |
574 | return NULL; | |
575 | audit_zero_context(context, state); | |
576 | return context; | |
577 | } | |
578 | ||
579 | /* Filter on the task information and allocate a per-task audit context | |
580 | * if necessary. Doing so turns on system call auditing for the | |
581 | * specified task. This is called from copy_process, so no lock is | |
582 | * needed. */ | |
583 | int audit_alloc(struct task_struct *tsk) | |
584 | { | |
585 | struct audit_context *context; | |
586 | enum audit_state state; | |
587 | ||
588 | if (likely(!audit_enabled)) | |
589 | return 0; /* Return if not auditing. */ | |
590 | ||
591 | state = audit_filter_task(tsk); | |
592 | if (likely(state == AUDIT_DISABLED)) | |
593 | return 0; | |
594 | ||
595 | if (!(context = audit_alloc_context(state))) { | |
596 | audit_log_lost("out of memory in audit_alloc"); | |
597 | return -ENOMEM; | |
598 | } | |
599 | ||
600 | /* Preserve login uid */ | |
601 | context->loginuid = -1; | |
602 | if (current->audit_context) | |
603 | context->loginuid = current->audit_context->loginuid; | |
604 | ||
605 | tsk->audit_context = context; | |
606 | set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT); | |
607 | return 0; | |
608 | } | |
609 | ||
610 | static inline void audit_free_context(struct audit_context *context) | |
611 | { | |
612 | struct audit_context *previous; | |
613 | int count = 0; | |
614 | ||
615 | do { | |
616 | previous = context->previous; | |
617 | if (previous || (count && count < 10)) { | |
618 | ++count; | |
619 | printk(KERN_ERR "audit(:%d): major=%d name_count=%d:" | |
620 | " freeing multiple contexts (%d)\n", | |
621 | context->serial, context->major, | |
622 | context->name_count, count); | |
623 | } | |
624 | audit_free_names(context); | |
625 | audit_free_aux(context); | |
626 | kfree(context); | |
627 | context = previous; | |
628 | } while (context); | |
629 | if (count >= 10) | |
630 | printk(KERN_ERR "audit: freed %d contexts\n", count); | |
631 | } | |
632 | ||
219f0817 SS |
633 | static void audit_log_task_info(struct audit_buffer *ab) |
634 | { | |
635 | char name[sizeof(current->comm)]; | |
636 | struct mm_struct *mm = current->mm; | |
637 | struct vm_area_struct *vma; | |
638 | ||
639 | get_task_comm(name, current); | |
640 | audit_log_format(ab, " comm=%s", name); | |
641 | ||
642 | if (!mm) | |
643 | return; | |
644 | ||
645 | down_read(&mm->mmap_sem); | |
646 | vma = mm->mmap; | |
647 | while (vma) { | |
648 | if ((vma->vm_flags & VM_EXECUTABLE) && | |
649 | vma->vm_file) { | |
650 | audit_log_d_path(ab, "exe=", | |
651 | vma->vm_file->f_dentry, | |
652 | vma->vm_file->f_vfsmnt); | |
653 | break; | |
654 | } | |
655 | vma = vma->vm_next; | |
656 | } | |
657 | up_read(&mm->mmap_sem); | |
658 | } | |
659 | ||
1da177e4 LT |
660 | static void audit_log_exit(struct audit_context *context) |
661 | { | |
662 | int i; | |
663 | struct audit_buffer *ab; | |
664 | ||
c0404993 | 665 | ab = audit_log_start(context, AUDIT_SYSCALL); |
1da177e4 LT |
666 | if (!ab) |
667 | return; /* audit_panic has been called */ | |
668 | audit_log_format(ab, "syscall=%d", context->major); | |
669 | if (context->personality != PER_LINUX) | |
670 | audit_log_format(ab, " per=%lx", context->personality); | |
2fd6f58b | 671 | audit_log_format(ab, " arch=%x", context->arch); |
1da177e4 | 672 | if (context->return_valid) |
2fd6f58b DW |
673 | audit_log_format(ab, " success=%s exit=%ld", |
674 | (context->return_valid==AUDITSC_SUCCESS)?"yes":"no", | |
675 | context->return_code); | |
1da177e4 LT |
676 | audit_log_format(ab, |
677 | " a0=%lx a1=%lx a2=%lx a3=%lx items=%d" | |
678 | " pid=%d loginuid=%d uid=%d gid=%d" | |
679 | " euid=%d suid=%d fsuid=%d" | |
680 | " egid=%d sgid=%d fsgid=%d", | |
681 | context->argv[0], | |
682 | context->argv[1], | |
683 | context->argv[2], | |
684 | context->argv[3], | |
685 | context->name_count, | |
686 | context->pid, | |
687 | context->loginuid, | |
688 | context->uid, | |
689 | context->gid, | |
690 | context->euid, context->suid, context->fsuid, | |
691 | context->egid, context->sgid, context->fsgid); | |
219f0817 | 692 | audit_log_task_info(ab); |
1da177e4 LT |
693 | audit_log_end(ab); |
694 | while (context->aux) { | |
695 | struct audit_aux_data *aux; | |
696 | ||
c0404993 SG |
697 | aux = context->aux; |
698 | ||
699 | ab = audit_log_start(context, aux->type); | |
1da177e4 LT |
700 | if (!ab) |
701 | continue; /* audit_panic has been called */ | |
702 | ||
1da177e4 | 703 | switch (aux->type) { |
c0404993 | 704 | case AUDIT_IPC: { |
1da177e4 LT |
705 | struct audit_aux_data_ipcctl *axi = (void *)aux; |
706 | audit_log_format(ab, | |
c0404993 | 707 | " qbytes=%lx iuid=%d igid=%d mode=%x", |
1da177e4 | 708 | axi->qbytes, axi->uid, axi->gid, axi->mode); |
3ec3b2fb DW |
709 | break; } |
710 | ||
711 | case AUDIT_SOCKETCALL: { | |
712 | int i; | |
713 | struct audit_aux_data_socketcall *axs = (void *)aux; | |
714 | audit_log_format(ab, "nargs=%d", axs->nargs); | |
715 | for (i=0; i<axs->nargs; i++) | |
716 | audit_log_format(ab, " a%d=%lx", i, axs->args[i]); | |
717 | break; } | |
718 | ||
719 | case AUDIT_SOCKADDR: { | |
720 | struct audit_aux_data_sockaddr *axs = (void *)aux; | |
721 | ||
722 | audit_log_format(ab, "saddr="); | |
723 | audit_log_hex(ab, axs->a, axs->len); | |
724 | break; } | |
1da177e4 LT |
725 | } |
726 | audit_log_end(ab); | |
c0404993 SG |
727 | |
728 | context->aux = aux->next; | |
1da177e4 LT |
729 | kfree(aux); |
730 | } | |
731 | ||
732 | for (i = 0; i < context->name_count; i++) { | |
c0404993 | 733 | ab = audit_log_start(context, AUDIT_PATH); |
1da177e4 LT |
734 | if (!ab) |
735 | continue; /* audit_panic has been called */ | |
736 | audit_log_format(ab, "item=%d", i); | |
83c7d091 DW |
737 | if (context->names[i].name) { |
738 | audit_log_format(ab, " name="); | |
739 | audit_log_untrustedstring(ab, context->names[i].name); | |
740 | } | |
1da177e4 LT |
741 | if (context->names[i].ino != (unsigned long)-1) |
742 | audit_log_format(ab, " inode=%lu dev=%02x:%02x mode=%#o" | |
c0404993 | 743 | " ouid=%d ogid=%d rdev=%02x:%02x", |
1da177e4 LT |
744 | context->names[i].ino, |
745 | MAJOR(context->names[i].dev), | |
746 | MINOR(context->names[i].dev), | |
747 | context->names[i].mode, | |
748 | context->names[i].uid, | |
749 | context->names[i].gid, | |
750 | MAJOR(context->names[i].rdev), | |
751 | MINOR(context->names[i].rdev)); | |
752 | audit_log_end(ab); | |
753 | } | |
754 | } | |
755 | ||
756 | /* Free a per-task audit context. Called from copy_process and | |
757 | * __put_task_struct. */ | |
758 | void audit_free(struct task_struct *tsk) | |
759 | { | |
760 | struct audit_context *context; | |
761 | ||
762 | task_lock(tsk); | |
763 | context = audit_get_context(tsk, 0, 0); | |
764 | task_unlock(tsk); | |
765 | ||
766 | if (likely(!context)) | |
767 | return; | |
768 | ||
769 | /* Check for system calls that do not go through the exit | |
770 | * function (e.g., exit_group), then free context block. */ | |
771 | if (context->in_syscall && context->auditable) | |
772 | audit_log_exit(context); | |
773 | ||
774 | audit_free_context(context); | |
775 | } | |
776 | ||
777 | /* Compute a serial number for the audit record. Audit records are | |
778 | * written to user-space as soon as they are generated, so a complete | |
779 | * audit record may be written in several pieces. The timestamp of the | |
23f32d18 | 780 | * record and this serial number are used by the user-space tools to |
1da177e4 LT |
781 | * determine which pieces belong to the same audit record. The |
782 | * (timestamp,serial) tuple is unique for each syscall and is live from | |
783 | * syscall entry to syscall exit. | |
784 | * | |
785 | * Atomic values are only guaranteed to be 24-bit, so we count down. | |
786 | * | |
787 | * NOTE: Another possibility is to store the formatted records off the | |
788 | * audit context (for those records that have a context), and emit them | |
789 | * all at syscall exit. However, this could delay the reporting of | |
790 | * significant errors until syscall exit (or never, if the system | |
791 | * halts). */ | |
792 | static inline unsigned int audit_serial(void) | |
793 | { | |
794 | static atomic_t serial = ATOMIC_INIT(0xffffff); | |
795 | unsigned int a, b; | |
796 | ||
797 | do { | |
798 | a = atomic_read(&serial); | |
799 | if (atomic_dec_and_test(&serial)) | |
800 | atomic_set(&serial, 0xffffff); | |
801 | b = atomic_read(&serial); | |
802 | } while (b != a - 1); | |
803 | ||
804 | return 0xffffff - b; | |
805 | } | |
806 | ||
807 | /* Fill in audit context at syscall entry. This only happens if the | |
808 | * audit context was created when the task was created and the state or | |
809 | * filters demand the audit context be built. If the state from the | |
810 | * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT, | |
811 | * then the record will be written at syscall exit time (otherwise, it | |
812 | * will only be written if another part of the kernel requests that it | |
813 | * be written). */ | |
2fd6f58b | 814 | void audit_syscall_entry(struct task_struct *tsk, int arch, int major, |
1da177e4 LT |
815 | unsigned long a1, unsigned long a2, |
816 | unsigned long a3, unsigned long a4) | |
817 | { | |
818 | struct audit_context *context = tsk->audit_context; | |
819 | enum audit_state state; | |
820 | ||
821 | BUG_ON(!context); | |
822 | ||
823 | /* This happens only on certain architectures that make system | |
824 | * calls in kernel_thread via the entry.S interface, instead of | |
825 | * with direct calls. (If you are porting to a new | |
826 | * architecture, hitting this condition can indicate that you | |
827 | * got the _exit/_leave calls backward in entry.S.) | |
828 | * | |
829 | * i386 no | |
830 | * x86_64 no | |
831 | * ppc64 yes (see arch/ppc64/kernel/misc.S) | |
832 | * | |
833 | * This also happens with vm86 emulation in a non-nested manner | |
834 | * (entries without exits), so this case must be caught. | |
835 | */ | |
836 | if (context->in_syscall) { | |
837 | struct audit_context *newctx; | |
838 | ||
839 | #if defined(__NR_vm86) && defined(__NR_vm86old) | |
840 | /* vm86 mode should only be entered once */ | |
841 | if (major == __NR_vm86 || major == __NR_vm86old) | |
842 | return; | |
843 | #endif | |
844 | #if AUDIT_DEBUG | |
845 | printk(KERN_ERR | |
846 | "audit(:%d) pid=%d in syscall=%d;" | |
847 | " entering syscall=%d\n", | |
848 | context->serial, tsk->pid, context->major, major); | |
849 | #endif | |
850 | newctx = audit_alloc_context(context->state); | |
851 | if (newctx) { | |
852 | newctx->previous = context; | |
853 | context = newctx; | |
854 | tsk->audit_context = newctx; | |
855 | } else { | |
856 | /* If we can't alloc a new context, the best we | |
857 | * can do is to leak memory (any pending putname | |
858 | * will be lost). The only other alternative is | |
859 | * to abandon auditing. */ | |
860 | audit_zero_context(context, context->state); | |
861 | } | |
862 | } | |
863 | BUG_ON(context->in_syscall || context->name_count); | |
864 | ||
865 | if (!audit_enabled) | |
866 | return; | |
867 | ||
2fd6f58b | 868 | context->arch = arch; |
1da177e4 LT |
869 | context->major = major; |
870 | context->argv[0] = a1; | |
871 | context->argv[1] = a2; | |
872 | context->argv[2] = a3; | |
873 | context->argv[3] = a4; | |
874 | ||
875 | state = context->state; | |
876 | if (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT) | |
877 | state = audit_filter_syscall(tsk, context, &audit_entlist); | |
878 | if (likely(state == AUDIT_DISABLED)) | |
879 | return; | |
880 | ||
881 | context->serial = audit_serial(); | |
882 | context->ctime = CURRENT_TIME; | |
883 | context->in_syscall = 1; | |
884 | context->auditable = !!(state == AUDIT_RECORD_CONTEXT); | |
885 | } | |
886 | ||
887 | /* Tear down after system call. If the audit context has been marked as | |
888 | * auditable (either because of the AUDIT_RECORD_CONTEXT state from | |
889 | * filtering, or because some other part of the kernel write an audit | |
890 | * message), then write out the syscall information. In call cases, | |
891 | * free the names stored from getname(). */ | |
2fd6f58b | 892 | void audit_syscall_exit(struct task_struct *tsk, int valid, long return_code) |
1da177e4 LT |
893 | { |
894 | struct audit_context *context; | |
895 | ||
896 | get_task_struct(tsk); | |
897 | task_lock(tsk); | |
2fd6f58b | 898 | context = audit_get_context(tsk, valid, return_code); |
1da177e4 LT |
899 | task_unlock(tsk); |
900 | ||
901 | /* Not having a context here is ok, since the parent may have | |
902 | * called __put_task_struct. */ | |
903 | if (likely(!context)) | |
904 | return; | |
905 | ||
906 | if (context->in_syscall && context->auditable) | |
907 | audit_log_exit(context); | |
908 | ||
909 | context->in_syscall = 0; | |
910 | context->auditable = 0; | |
2fd6f58b | 911 | |
1da177e4 LT |
912 | if (context->previous) { |
913 | struct audit_context *new_context = context->previous; | |
914 | context->previous = NULL; | |
915 | audit_free_context(context); | |
916 | tsk->audit_context = new_context; | |
917 | } else { | |
918 | audit_free_names(context); | |
919 | audit_free_aux(context); | |
920 | audit_zero_context(context, context->state); | |
921 | tsk->audit_context = context; | |
922 | } | |
923 | put_task_struct(tsk); | |
924 | } | |
925 | ||
926 | /* Add a name to the list. Called from fs/namei.c:getname(). */ | |
927 | void audit_getname(const char *name) | |
928 | { | |
929 | struct audit_context *context = current->audit_context; | |
930 | ||
931 | if (!context || IS_ERR(name) || !name) | |
932 | return; | |
933 | ||
934 | if (!context->in_syscall) { | |
935 | #if AUDIT_DEBUG == 2 | |
936 | printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n", | |
937 | __FILE__, __LINE__, context->serial, name); | |
938 | dump_stack(); | |
939 | #endif | |
940 | return; | |
941 | } | |
942 | BUG_ON(context->name_count >= AUDIT_NAMES); | |
943 | context->names[context->name_count].name = name; | |
944 | context->names[context->name_count].ino = (unsigned long)-1; | |
945 | ++context->name_count; | |
946 | } | |
947 | ||
948 | /* Intercept a putname request. Called from | |
949 | * include/linux/fs.h:putname(). If we have stored the name from | |
950 | * getname in the audit context, then we delay the putname until syscall | |
951 | * exit. */ | |
952 | void audit_putname(const char *name) | |
953 | { | |
954 | struct audit_context *context = current->audit_context; | |
955 | ||
956 | BUG_ON(!context); | |
957 | if (!context->in_syscall) { | |
958 | #if AUDIT_DEBUG == 2 | |
959 | printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n", | |
960 | __FILE__, __LINE__, context->serial, name); | |
961 | if (context->name_count) { | |
962 | int i; | |
963 | for (i = 0; i < context->name_count; i++) | |
964 | printk(KERN_ERR "name[%d] = %p = %s\n", i, | |
965 | context->names[i].name, | |
966 | context->names[i].name); | |
967 | } | |
968 | #endif | |
969 | __putname(name); | |
970 | } | |
971 | #if AUDIT_DEBUG | |
972 | else { | |
973 | ++context->put_count; | |
974 | if (context->put_count > context->name_count) { | |
975 | printk(KERN_ERR "%s:%d(:%d): major=%d" | |
976 | " in_syscall=%d putname(%p) name_count=%d" | |
977 | " put_count=%d\n", | |
978 | __FILE__, __LINE__, | |
979 | context->serial, context->major, | |
980 | context->in_syscall, name, context->name_count, | |
981 | context->put_count); | |
982 | dump_stack(); | |
983 | } | |
984 | } | |
985 | #endif | |
986 | } | |
987 | ||
988 | /* Store the inode and device from a lookup. Called from | |
989 | * fs/namei.c:path_lookup(). */ | |
990 | void audit_inode(const char *name, const struct inode *inode) | |
991 | { | |
992 | int idx; | |
993 | struct audit_context *context = current->audit_context; | |
994 | ||
995 | if (!context->in_syscall) | |
996 | return; | |
997 | if (context->name_count | |
998 | && context->names[context->name_count-1].name | |
999 | && context->names[context->name_count-1].name == name) | |
1000 | idx = context->name_count - 1; | |
1001 | else if (context->name_count > 1 | |
1002 | && context->names[context->name_count-2].name | |
1003 | && context->names[context->name_count-2].name == name) | |
1004 | idx = context->name_count - 2; | |
1005 | else { | |
1006 | /* FIXME: how much do we care about inodes that have no | |
1007 | * associated name? */ | |
1008 | if (context->name_count >= AUDIT_NAMES - AUDIT_NAMES_RESERVED) | |
1009 | return; | |
1010 | idx = context->name_count++; | |
1011 | context->names[idx].name = NULL; | |
1012 | #if AUDIT_DEBUG | |
1013 | ++context->ino_count; | |
1014 | #endif | |
1015 | } | |
1016 | context->names[idx].ino = inode->i_ino; | |
1017 | context->names[idx].dev = inode->i_sb->s_dev; | |
1018 | context->names[idx].mode = inode->i_mode; | |
1019 | context->names[idx].uid = inode->i_uid; | |
1020 | context->names[idx].gid = inode->i_gid; | |
1021 | context->names[idx].rdev = inode->i_rdev; | |
1022 | } | |
1023 | ||
197c69c6 | 1024 | int audit_get_stamp(struct audit_context *ctx, |
d812ddbb | 1025 | struct timespec *t, unsigned int *serial) |
1da177e4 LT |
1026 | { |
1027 | if (ctx) { | |
1028 | t->tv_sec = ctx->ctime.tv_sec; | |
1029 | t->tv_nsec = ctx->ctime.tv_nsec; | |
1030 | *serial = ctx->serial; | |
1031 | ctx->auditable = 1; | |
197c69c6 | 1032 | return 1; |
1da177e4 | 1033 | } |
197c69c6 | 1034 | return 0; |
1da177e4 LT |
1035 | } |
1036 | ||
456be6cd | 1037 | int audit_set_loginuid(struct task_struct *task, uid_t loginuid) |
1da177e4 | 1038 | { |
456be6cd | 1039 | if (task->audit_context) { |
c0404993 SG |
1040 | struct audit_buffer *ab; |
1041 | ||
1042 | ab = audit_log_start(NULL, AUDIT_LOGIN); | |
1043 | if (ab) { | |
1044 | audit_log_format(ab, "login pid=%d uid=%u " | |
1045 | "old loginuid=%u new loginuid=%u", | |
1046 | task->pid, task->uid, | |
1047 | task->audit_context->loginuid, loginuid); | |
1048 | audit_log_end(ab); | |
1049 | } | |
456be6cd | 1050 | task->audit_context->loginuid = loginuid; |
1da177e4 LT |
1051 | } |
1052 | return 0; | |
1053 | } | |
1054 | ||
1055 | uid_t audit_get_loginuid(struct audit_context *ctx) | |
1056 | { | |
1057 | return ctx ? ctx->loginuid : -1; | |
1058 | } | |
1059 | ||
1060 | int audit_ipc_perms(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode) | |
1061 | { | |
1062 | struct audit_aux_data_ipcctl *ax; | |
1063 | struct audit_context *context = current->audit_context; | |
1064 | ||
1065 | if (likely(!context)) | |
1066 | return 0; | |
1067 | ||
1068 | ax = kmalloc(sizeof(*ax), GFP_KERNEL); | |
1069 | if (!ax) | |
1070 | return -ENOMEM; | |
1071 | ||
1072 | ax->qbytes = qbytes; | |
1073 | ax->uid = uid; | |
1074 | ax->gid = gid; | |
1075 | ax->mode = mode; | |
1076 | ||
c0404993 | 1077 | ax->d.type = AUDIT_IPC; |
1da177e4 LT |
1078 | ax->d.next = context->aux; |
1079 | context->aux = (void *)ax; | |
1080 | return 0; | |
1081 | } | |
c2f0c7c3 | 1082 | |
3ec3b2fb DW |
1083 | int audit_socketcall(int nargs, unsigned long *args) |
1084 | { | |
1085 | struct audit_aux_data_socketcall *ax; | |
1086 | struct audit_context *context = current->audit_context; | |
1087 | ||
1088 | if (likely(!context)) | |
1089 | return 0; | |
1090 | ||
1091 | ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL); | |
1092 | if (!ax) | |
1093 | return -ENOMEM; | |
1094 | ||
1095 | ax->nargs = nargs; | |
1096 | memcpy(ax->args, args, nargs * sizeof(unsigned long)); | |
1097 | ||
1098 | ax->d.type = AUDIT_SOCKETCALL; | |
1099 | ax->d.next = context->aux; | |
1100 | context->aux = (void *)ax; | |
1101 | return 0; | |
1102 | } | |
1103 | ||
1104 | int audit_sockaddr(int len, void *a) | |
1105 | { | |
1106 | struct audit_aux_data_sockaddr *ax; | |
1107 | struct audit_context *context = current->audit_context; | |
1108 | ||
1109 | if (likely(!context)) | |
1110 | return 0; | |
1111 | ||
1112 | ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL); | |
1113 | if (!ax) | |
1114 | return -ENOMEM; | |
1115 | ||
1116 | ax->len = len; | |
1117 | memcpy(ax->a, a, len); | |
1118 | ||
1119 | ax->d.type = AUDIT_SOCKADDR; | |
1120 | ax->d.next = context->aux; | |
1121 | context->aux = (void *)ax; | |
1122 | return 0; | |
1123 | } | |
1124 | ||
c2f0c7c3 SG |
1125 | void audit_signal_info(int sig, struct task_struct *t) |
1126 | { | |
1127 | extern pid_t audit_sig_pid; | |
1128 | extern uid_t audit_sig_uid; | |
1129 | extern int audit_pid; | |
1130 | ||
1131 | if (unlikely(audit_pid && t->pid == audit_pid)) { | |
1132 | if (sig == SIGTERM || sig == SIGHUP) { | |
1133 | struct audit_context *ctx = current->audit_context; | |
1134 | audit_sig_pid = current->pid; | |
1135 | if (ctx) | |
1136 | audit_sig_uid = ctx->loginuid; | |
1137 | else | |
1138 | audit_sig_uid = current->uid; | |
1139 | } | |
1140 | } | |
1141 | } | |
1142 |