x86: use the correct cpuid method to detect MWAIT support for C states
[linux-2.6/next.git] / kernel / auditsc.c
blobbce9ecdb771240885bd3d1c0c30f7565ebbd1eb5
1 /* auditsc.c -- System-call auditing support
2 * Handles all system-call specific auditing features.
4 * Copyright 2003-2004 Red Hat Inc., Durham, North Carolina.
5 * Copyright 2005 Hewlett-Packard Development Company, L.P.
6 * Copyright (C) 2005, 2006 IBM Corporation
7 * All Rights Reserved.
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 * Written by Rickard E. (Rik) Faith <faith@redhat.com>
25 * Many of the ideas implemented here are from Stephen C. Tweedie,
26 * especially the idea of avoiding a copy by using getname.
28 * The method for actual interception of syscall entry and exit (not in
29 * this file -- see entry.S) is based on a GPL'd patch written by
30 * okir@suse.de and Copyright 2003 SuSE Linux AG.
32 * POSIX message queue support added by George Wilson <ltcgcw@us.ibm.com>,
33 * 2006.
35 * The support of additional filter rules compares (>, <, >=, <=) was
36 * added by Dustin Kirkland <dustin.kirkland@us.ibm.com>, 2005.
38 * Modified by Amy Griffis <amy.griffis@hp.com> to collect additional
39 * filesystem information.
41 * Subject and object context labeling support added by <danjones@us.ibm.com>
42 * and <dustin.kirkland@us.ibm.com> for LSPP certification compliance.
45 #include <linux/init.h>
46 #include <asm/types.h>
47 #include <asm/atomic.h>
48 #include <linux/fs.h>
49 #include <linux/namei.h>
50 #include <linux/mm.h>
51 #include <linux/module.h>
52 #include <linux/mount.h>
53 #include <linux/socket.h>
54 #include <linux/mqueue.h>
55 #include <linux/audit.h>
56 #include <linux/personality.h>
57 #include <linux/time.h>
58 #include <linux/netlink.h>
59 #include <linux/compiler.h>
60 #include <asm/unistd.h>
61 #include <linux/security.h>
62 #include <linux/list.h>
63 #include <linux/tty.h>
64 #include <linux/selinux.h>
65 #include <linux/binfmts.h>
66 #include <linux/highmem.h>
67 #include <linux/syscalls.h>
68 #include <linux/inotify.h>
70 #include "audit.h"
72 extern struct list_head audit_filter_list[];
74 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
75 * for saving names from getname(). */
76 #define AUDIT_NAMES 20
78 /* Indicates that audit should log the full pathname. */
79 #define AUDIT_NAME_FULL -1
81 /* number of audit rules */
82 int audit_n_rules;
84 /* determines whether we collect data for signals sent */
85 int audit_signals;
87 /* When fs/namei.c:getname() is called, we store the pointer in name and
88 * we don't let putname() free it (instead we free all of the saved
89 * pointers at syscall exit time).
91 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
92 struct audit_names {
93 const char *name;
94 int name_len; /* number of name's characters to log */
95 unsigned name_put; /* call __putname() for this name */
96 unsigned long ino;
97 dev_t dev;
98 umode_t mode;
99 uid_t uid;
100 gid_t gid;
101 dev_t rdev;
102 u32 osid;
105 struct audit_aux_data {
106 struct audit_aux_data *next;
107 int type;
110 #define AUDIT_AUX_IPCPERM 0
112 /* Number of target pids per aux struct. */
113 #define AUDIT_AUX_PIDS 16
115 struct audit_aux_data_mq_open {
116 struct audit_aux_data d;
117 int oflag;
118 mode_t mode;
119 struct mq_attr attr;
122 struct audit_aux_data_mq_sendrecv {
123 struct audit_aux_data d;
124 mqd_t mqdes;
125 size_t msg_len;
126 unsigned int msg_prio;
127 struct timespec abs_timeout;
130 struct audit_aux_data_mq_notify {
131 struct audit_aux_data d;
132 mqd_t mqdes;
133 struct sigevent notification;
136 struct audit_aux_data_mq_getsetattr {
137 struct audit_aux_data d;
138 mqd_t mqdes;
139 struct mq_attr mqstat;
142 struct audit_aux_data_ipcctl {
143 struct audit_aux_data d;
144 struct ipc_perm p;
145 unsigned long qbytes;
146 uid_t uid;
147 gid_t gid;
148 mode_t mode;
149 u32 osid;
152 struct audit_aux_data_execve {
153 struct audit_aux_data d;
154 int argc;
155 int envc;
156 struct mm_struct *mm;
159 struct audit_aux_data_socketcall {
160 struct audit_aux_data d;
161 int nargs;
162 unsigned long args[0];
165 struct audit_aux_data_sockaddr {
166 struct audit_aux_data d;
167 int len;
168 char a[0];
171 struct audit_aux_data_fd_pair {
172 struct audit_aux_data d;
173 int fd[2];
176 struct audit_aux_data_pids {
177 struct audit_aux_data d;
178 pid_t target_pid[AUDIT_AUX_PIDS];
179 u32 target_sid[AUDIT_AUX_PIDS];
180 int pid_count;
183 struct audit_tree_refs {
184 struct audit_tree_refs *next;
185 struct audit_chunk *c[31];
188 /* The per-task audit context. */
189 struct audit_context {
190 int dummy; /* must be the first element */
191 int in_syscall; /* 1 if task is in a syscall */
192 enum audit_state state;
193 unsigned int serial; /* serial number for record */
194 struct timespec ctime; /* time of syscall entry */
195 uid_t loginuid; /* login uid (identity) */
196 int major; /* syscall number */
197 unsigned long argv[4]; /* syscall arguments */
198 int return_valid; /* return code is valid */
199 long return_code;/* syscall return code */
200 int auditable; /* 1 if record should be written */
201 int name_count;
202 struct audit_names names[AUDIT_NAMES];
203 char * filterkey; /* key for rule that triggered record */
204 struct dentry * pwd;
205 struct vfsmount * pwdmnt;
206 struct audit_context *previous; /* For nested syscalls */
207 struct audit_aux_data *aux;
208 struct audit_aux_data *aux_pids;
210 /* Save things to print about task_struct */
211 pid_t pid, ppid;
212 uid_t uid, euid, suid, fsuid;
213 gid_t gid, egid, sgid, fsgid;
214 unsigned long personality;
215 int arch;
217 pid_t target_pid;
218 u32 target_sid;
220 struct audit_tree_refs *trees, *first_trees;
221 int tree_count;
223 #if AUDIT_DEBUG
224 int put_count;
225 int ino_count;
226 #endif
229 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
230 static inline int open_arg(int flags, int mask)
232 int n = ACC_MODE(flags);
233 if (flags & (O_TRUNC | O_CREAT))
234 n |= AUDIT_PERM_WRITE;
235 return n & mask;
238 static int audit_match_perm(struct audit_context *ctx, int mask)
240 unsigned n = ctx->major;
241 switch (audit_classify_syscall(ctx->arch, n)) {
242 case 0: /* native */
243 if ((mask & AUDIT_PERM_WRITE) &&
244 audit_match_class(AUDIT_CLASS_WRITE, n))
245 return 1;
246 if ((mask & AUDIT_PERM_READ) &&
247 audit_match_class(AUDIT_CLASS_READ, n))
248 return 1;
249 if ((mask & AUDIT_PERM_ATTR) &&
250 audit_match_class(AUDIT_CLASS_CHATTR, n))
251 return 1;
252 return 0;
253 case 1: /* 32bit on biarch */
254 if ((mask & AUDIT_PERM_WRITE) &&
255 audit_match_class(AUDIT_CLASS_WRITE_32, n))
256 return 1;
257 if ((mask & AUDIT_PERM_READ) &&
258 audit_match_class(AUDIT_CLASS_READ_32, n))
259 return 1;
260 if ((mask & AUDIT_PERM_ATTR) &&
261 audit_match_class(AUDIT_CLASS_CHATTR_32, n))
262 return 1;
263 return 0;
264 case 2: /* open */
265 return mask & ACC_MODE(ctx->argv[1]);
266 case 3: /* openat */
267 return mask & ACC_MODE(ctx->argv[2]);
268 case 4: /* socketcall */
269 return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
270 case 5: /* execve */
271 return mask & AUDIT_PERM_EXEC;
272 default:
273 return 0;
278 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
279 * ->first_trees points to its beginning, ->trees - to the current end of data.
280 * ->tree_count is the number of free entries in array pointed to by ->trees.
281 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
282 * "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
283 * it's going to remain 1-element for almost any setup) until we free context itself.
284 * References in it _are_ dropped - at the same time we free/drop aux stuff.
287 #ifdef CONFIG_AUDIT_TREE
288 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
290 struct audit_tree_refs *p = ctx->trees;
291 int left = ctx->tree_count;
292 if (likely(left)) {
293 p->c[--left] = chunk;
294 ctx->tree_count = left;
295 return 1;
297 if (!p)
298 return 0;
299 p = p->next;
300 if (p) {
301 p->c[30] = chunk;
302 ctx->trees = p;
303 ctx->tree_count = 30;
304 return 1;
306 return 0;
309 static int grow_tree_refs(struct audit_context *ctx)
311 struct audit_tree_refs *p = ctx->trees;
312 ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
313 if (!ctx->trees) {
314 ctx->trees = p;
315 return 0;
317 if (p)
318 p->next = ctx->trees;
319 else
320 ctx->first_trees = ctx->trees;
321 ctx->tree_count = 31;
322 return 1;
324 #endif
326 static void unroll_tree_refs(struct audit_context *ctx,
327 struct audit_tree_refs *p, int count)
329 #ifdef CONFIG_AUDIT_TREE
330 struct audit_tree_refs *q;
331 int n;
332 if (!p) {
333 /* we started with empty chain */
334 p = ctx->first_trees;
335 count = 31;
336 /* if the very first allocation has failed, nothing to do */
337 if (!p)
338 return;
340 n = count;
341 for (q = p; q != ctx->trees; q = q->next, n = 31) {
342 while (n--) {
343 audit_put_chunk(q->c[n]);
344 q->c[n] = NULL;
347 while (n-- > ctx->tree_count) {
348 audit_put_chunk(q->c[n]);
349 q->c[n] = NULL;
351 ctx->trees = p;
352 ctx->tree_count = count;
353 #endif
356 static void free_tree_refs(struct audit_context *ctx)
358 struct audit_tree_refs *p, *q;
359 for (p = ctx->first_trees; p; p = q) {
360 q = p->next;
361 kfree(p);
365 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
367 #ifdef CONFIG_AUDIT_TREE
368 struct audit_tree_refs *p;
369 int n;
370 if (!tree)
371 return 0;
372 /* full ones */
373 for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
374 for (n = 0; n < 31; n++)
375 if (audit_tree_match(p->c[n], tree))
376 return 1;
378 /* partial */
379 if (p) {
380 for (n = ctx->tree_count; n < 31; n++)
381 if (audit_tree_match(p->c[n], tree))
382 return 1;
384 #endif
385 return 0;
388 /* Determine if any context name data matches a rule's watch data */
389 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
390 * otherwise. */
391 static int audit_filter_rules(struct task_struct *tsk,
392 struct audit_krule *rule,
393 struct audit_context *ctx,
394 struct audit_names *name,
395 enum audit_state *state)
397 int i, j, need_sid = 1;
398 u32 sid;
400 for (i = 0; i < rule->field_count; i++) {
401 struct audit_field *f = &rule->fields[i];
402 int result = 0;
404 switch (f->type) {
405 case AUDIT_PID:
406 result = audit_comparator(tsk->pid, f->op, f->val);
407 break;
408 case AUDIT_PPID:
409 if (ctx) {
410 if (!ctx->ppid)
411 ctx->ppid = sys_getppid();
412 result = audit_comparator(ctx->ppid, f->op, f->val);
414 break;
415 case AUDIT_UID:
416 result = audit_comparator(tsk->uid, f->op, f->val);
417 break;
418 case AUDIT_EUID:
419 result = audit_comparator(tsk->euid, f->op, f->val);
420 break;
421 case AUDIT_SUID:
422 result = audit_comparator(tsk->suid, f->op, f->val);
423 break;
424 case AUDIT_FSUID:
425 result = audit_comparator(tsk->fsuid, f->op, f->val);
426 break;
427 case AUDIT_GID:
428 result = audit_comparator(tsk->gid, f->op, f->val);
429 break;
430 case AUDIT_EGID:
431 result = audit_comparator(tsk->egid, f->op, f->val);
432 break;
433 case AUDIT_SGID:
434 result = audit_comparator(tsk->sgid, f->op, f->val);
435 break;
436 case AUDIT_FSGID:
437 result = audit_comparator(tsk->fsgid, f->op, f->val);
438 break;
439 case AUDIT_PERS:
440 result = audit_comparator(tsk->personality, f->op, f->val);
441 break;
442 case AUDIT_ARCH:
443 if (ctx)
444 result = audit_comparator(ctx->arch, f->op, f->val);
445 break;
447 case AUDIT_EXIT:
448 if (ctx && ctx->return_valid)
449 result = audit_comparator(ctx->return_code, f->op, f->val);
450 break;
451 case AUDIT_SUCCESS:
452 if (ctx && ctx->return_valid) {
453 if (f->val)
454 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
455 else
456 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
458 break;
459 case AUDIT_DEVMAJOR:
460 if (name)
461 result = audit_comparator(MAJOR(name->dev),
462 f->op, f->val);
463 else if (ctx) {
464 for (j = 0; j < ctx->name_count; j++) {
465 if (audit_comparator(MAJOR(ctx->names[j].dev), f->op, f->val)) {
466 ++result;
467 break;
471 break;
472 case AUDIT_DEVMINOR:
473 if (name)
474 result = audit_comparator(MINOR(name->dev),
475 f->op, f->val);
476 else if (ctx) {
477 for (j = 0; j < ctx->name_count; j++) {
478 if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
479 ++result;
480 break;
484 break;
485 case AUDIT_INODE:
486 if (name)
487 result = (name->ino == f->val);
488 else if (ctx) {
489 for (j = 0; j < ctx->name_count; j++) {
490 if (audit_comparator(ctx->names[j].ino, f->op, f->val)) {
491 ++result;
492 break;
496 break;
497 case AUDIT_WATCH:
498 if (name && rule->watch->ino != (unsigned long)-1)
499 result = (name->dev == rule->watch->dev &&
500 name->ino == rule->watch->ino);
501 break;
502 case AUDIT_DIR:
503 if (ctx)
504 result = match_tree_refs(ctx, rule->tree);
505 break;
506 case AUDIT_LOGINUID:
507 result = 0;
508 if (ctx)
509 result = audit_comparator(ctx->loginuid, f->op, f->val);
510 break;
511 case AUDIT_SUBJ_USER:
512 case AUDIT_SUBJ_ROLE:
513 case AUDIT_SUBJ_TYPE:
514 case AUDIT_SUBJ_SEN:
515 case AUDIT_SUBJ_CLR:
516 /* NOTE: this may return negative values indicating
517 a temporary error. We simply treat this as a
518 match for now to avoid losing information that
519 may be wanted. An error message will also be
520 logged upon error */
521 if (f->se_rule) {
522 if (need_sid) {
523 selinux_get_task_sid(tsk, &sid);
524 need_sid = 0;
526 result = selinux_audit_rule_match(sid, f->type,
527 f->op,
528 f->se_rule,
529 ctx);
531 break;
532 case AUDIT_OBJ_USER:
533 case AUDIT_OBJ_ROLE:
534 case AUDIT_OBJ_TYPE:
535 case AUDIT_OBJ_LEV_LOW:
536 case AUDIT_OBJ_LEV_HIGH:
537 /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
538 also applies here */
539 if (f->se_rule) {
540 /* Find files that match */
541 if (name) {
542 result = selinux_audit_rule_match(
543 name->osid, f->type, f->op,
544 f->se_rule, ctx);
545 } else if (ctx) {
546 for (j = 0; j < ctx->name_count; j++) {
547 if (selinux_audit_rule_match(
548 ctx->names[j].osid,
549 f->type, f->op,
550 f->se_rule, ctx)) {
551 ++result;
552 break;
556 /* Find ipc objects that match */
557 if (ctx) {
558 struct audit_aux_data *aux;
559 for (aux = ctx->aux; aux;
560 aux = aux->next) {
561 if (aux->type == AUDIT_IPC) {
562 struct audit_aux_data_ipcctl *axi = (void *)aux;
563 if (selinux_audit_rule_match(axi->osid, f->type, f->op, f->se_rule, ctx)) {
564 ++result;
565 break;
571 break;
572 case AUDIT_ARG0:
573 case AUDIT_ARG1:
574 case AUDIT_ARG2:
575 case AUDIT_ARG3:
576 if (ctx)
577 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
578 break;
579 case AUDIT_FILTERKEY:
580 /* ignore this field for filtering */
581 result = 1;
582 break;
583 case AUDIT_PERM:
584 result = audit_match_perm(ctx, f->val);
585 break;
588 if (!result)
589 return 0;
591 if (rule->filterkey)
592 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
593 switch (rule->action) {
594 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
595 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
597 return 1;
600 /* At process creation time, we can determine if system-call auditing is
601 * completely disabled for this task. Since we only have the task
602 * structure at this point, we can only check uid and gid.
604 static enum audit_state audit_filter_task(struct task_struct *tsk)
606 struct audit_entry *e;
607 enum audit_state state;
609 rcu_read_lock();
610 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
611 if (audit_filter_rules(tsk, &e->rule, NULL, NULL, &state)) {
612 rcu_read_unlock();
613 return state;
616 rcu_read_unlock();
617 return AUDIT_BUILD_CONTEXT;
620 /* At syscall entry and exit time, this filter is called if the
621 * audit_state is not low enough that auditing cannot take place, but is
622 * also not high enough that we already know we have to write an audit
623 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
625 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
626 struct audit_context *ctx,
627 struct list_head *list)
629 struct audit_entry *e;
630 enum audit_state state;
632 if (audit_pid && tsk->tgid == audit_pid)
633 return AUDIT_DISABLED;
635 rcu_read_lock();
636 if (!list_empty(list)) {
637 int word = AUDIT_WORD(ctx->major);
638 int bit = AUDIT_BIT(ctx->major);
640 list_for_each_entry_rcu(e, list, list) {
641 if ((e->rule.mask[word] & bit) == bit &&
642 audit_filter_rules(tsk, &e->rule, ctx, NULL,
643 &state)) {
644 rcu_read_unlock();
645 return state;
649 rcu_read_unlock();
650 return AUDIT_BUILD_CONTEXT;
653 /* At syscall exit time, this filter is called if any audit_names[] have been
654 * collected during syscall processing. We only check rules in sublists at hash
655 * buckets applicable to the inode numbers in audit_names[].
656 * Regarding audit_state, same rules apply as for audit_filter_syscall().
658 enum audit_state audit_filter_inodes(struct task_struct *tsk,
659 struct audit_context *ctx)
661 int i;
662 struct audit_entry *e;
663 enum audit_state state;
665 if (audit_pid && tsk->tgid == audit_pid)
666 return AUDIT_DISABLED;
668 rcu_read_lock();
669 for (i = 0; i < ctx->name_count; i++) {
670 int word = AUDIT_WORD(ctx->major);
671 int bit = AUDIT_BIT(ctx->major);
672 struct audit_names *n = &ctx->names[i];
673 int h = audit_hash_ino((u32)n->ino);
674 struct list_head *list = &audit_inode_hash[h];
676 if (list_empty(list))
677 continue;
679 list_for_each_entry_rcu(e, list, list) {
680 if ((e->rule.mask[word] & bit) == bit &&
681 audit_filter_rules(tsk, &e->rule, ctx, n, &state)) {
682 rcu_read_unlock();
683 return state;
687 rcu_read_unlock();
688 return AUDIT_BUILD_CONTEXT;
691 void audit_set_auditable(struct audit_context *ctx)
693 ctx->auditable = 1;
696 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
697 int return_valid,
698 int return_code)
700 struct audit_context *context = tsk->audit_context;
702 if (likely(!context))
703 return NULL;
704 context->return_valid = return_valid;
705 context->return_code = return_code;
707 if (context->in_syscall && !context->dummy && !context->auditable) {
708 enum audit_state state;
710 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
711 if (state == AUDIT_RECORD_CONTEXT) {
712 context->auditable = 1;
713 goto get_context;
716 state = audit_filter_inodes(tsk, context);
717 if (state == AUDIT_RECORD_CONTEXT)
718 context->auditable = 1;
722 get_context:
724 tsk->audit_context = NULL;
725 return context;
728 static inline void audit_free_names(struct audit_context *context)
730 int i;
732 #if AUDIT_DEBUG == 2
733 if (context->auditable
734 ||context->put_count + context->ino_count != context->name_count) {
735 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
736 " name_count=%d put_count=%d"
737 " ino_count=%d [NOT freeing]\n",
738 __FILE__, __LINE__,
739 context->serial, context->major, context->in_syscall,
740 context->name_count, context->put_count,
741 context->ino_count);
742 for (i = 0; i < context->name_count; i++) {
743 printk(KERN_ERR "names[%d] = %p = %s\n", i,
744 context->names[i].name,
745 context->names[i].name ?: "(null)");
747 dump_stack();
748 return;
750 #endif
751 #if AUDIT_DEBUG
752 context->put_count = 0;
753 context->ino_count = 0;
754 #endif
756 for (i = 0; i < context->name_count; i++) {
757 if (context->names[i].name && context->names[i].name_put)
758 __putname(context->names[i].name);
760 context->name_count = 0;
761 if (context->pwd)
762 dput(context->pwd);
763 if (context->pwdmnt)
764 mntput(context->pwdmnt);
765 context->pwd = NULL;
766 context->pwdmnt = NULL;
769 static inline void audit_free_aux(struct audit_context *context)
771 struct audit_aux_data *aux;
773 while ((aux = context->aux)) {
774 context->aux = aux->next;
775 kfree(aux);
777 while ((aux = context->aux_pids)) {
778 context->aux_pids = aux->next;
779 kfree(aux);
783 static inline void audit_zero_context(struct audit_context *context,
784 enum audit_state state)
786 uid_t loginuid = context->loginuid;
788 memset(context, 0, sizeof(*context));
789 context->state = state;
790 context->loginuid = loginuid;
793 static inline struct audit_context *audit_alloc_context(enum audit_state state)
795 struct audit_context *context;
797 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
798 return NULL;
799 audit_zero_context(context, state);
800 return context;
804 * audit_alloc - allocate an audit context block for a task
805 * @tsk: task
807 * Filter on the task information and allocate a per-task audit context
808 * if necessary. Doing so turns on system call auditing for the
809 * specified task. This is called from copy_process, so no lock is
810 * needed.
812 int audit_alloc(struct task_struct *tsk)
814 struct audit_context *context;
815 enum audit_state state;
817 if (likely(!audit_enabled))
818 return 0; /* Return if not auditing. */
820 state = audit_filter_task(tsk);
821 if (likely(state == AUDIT_DISABLED))
822 return 0;
824 if (!(context = audit_alloc_context(state))) {
825 audit_log_lost("out of memory in audit_alloc");
826 return -ENOMEM;
829 /* Preserve login uid */
830 context->loginuid = -1;
831 if (current->audit_context)
832 context->loginuid = current->audit_context->loginuid;
834 tsk->audit_context = context;
835 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
836 return 0;
839 static inline void audit_free_context(struct audit_context *context)
841 struct audit_context *previous;
842 int count = 0;
844 do {
845 previous = context->previous;
846 if (previous || (count && count < 10)) {
847 ++count;
848 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
849 " freeing multiple contexts (%d)\n",
850 context->serial, context->major,
851 context->name_count, count);
853 audit_free_names(context);
854 unroll_tree_refs(context, NULL, 0);
855 free_tree_refs(context);
856 audit_free_aux(context);
857 kfree(context->filterkey);
858 kfree(context);
859 context = previous;
860 } while (context);
861 if (count >= 10)
862 printk(KERN_ERR "audit: freed %d contexts\n", count);
865 void audit_log_task_context(struct audit_buffer *ab)
867 char *ctx = NULL;
868 unsigned len;
869 int error;
870 u32 sid;
872 selinux_get_task_sid(current, &sid);
873 if (!sid)
874 return;
876 error = selinux_sid_to_string(sid, &ctx, &len);
877 if (error) {
878 if (error != -EINVAL)
879 goto error_path;
880 return;
883 audit_log_format(ab, " subj=%s", ctx);
884 kfree(ctx);
885 return;
887 error_path:
888 audit_panic("error in audit_log_task_context");
889 return;
892 EXPORT_SYMBOL(audit_log_task_context);
894 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
896 char name[sizeof(tsk->comm)];
897 struct mm_struct *mm = tsk->mm;
898 struct vm_area_struct *vma;
900 /* tsk == current */
902 get_task_comm(name, tsk);
903 audit_log_format(ab, " comm=");
904 audit_log_untrustedstring(ab, name);
906 if (mm) {
907 down_read(&mm->mmap_sem);
908 vma = mm->mmap;
909 while (vma) {
910 if ((vma->vm_flags & VM_EXECUTABLE) &&
911 vma->vm_file) {
912 audit_log_d_path(ab, "exe=",
913 vma->vm_file->f_path.dentry,
914 vma->vm_file->f_path.mnt);
915 break;
917 vma = vma->vm_next;
919 up_read(&mm->mmap_sem);
921 audit_log_task_context(ab);
924 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
925 u32 sid)
927 struct audit_buffer *ab;
928 char *s = NULL;
929 u32 len;
930 int rc = 0;
932 ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
933 if (!ab)
934 return 1;
936 if (selinux_sid_to_string(sid, &s, &len)) {
937 audit_log_format(ab, "opid=%d obj=(none)", pid);
938 rc = 1;
939 } else
940 audit_log_format(ab, "opid=%d obj=%s", pid, s);
941 audit_log_end(ab);
942 kfree(s);
944 return rc;
947 static void audit_log_execve_info(struct audit_buffer *ab,
948 struct audit_aux_data_execve *axi)
950 int i;
951 long len, ret;
952 const char __user *p;
953 char *buf;
955 if (axi->mm != current->mm)
956 return; /* execve failed, no additional info */
958 p = (const char __user *)axi->mm->arg_start;
960 for (i = 0; i < axi->argc; i++, p += len) {
961 len = strnlen_user(p, MAX_ARG_STRLEN);
963 * We just created this mm, if we can't find the strings
964 * we just copied into it something is _very_ wrong. Similar
965 * for strings that are too long, we should not have created
966 * any.
968 if (!len || len > MAX_ARG_STRLEN) {
969 WARN_ON(1);
970 send_sig(SIGKILL, current, 0);
973 buf = kmalloc(len, GFP_KERNEL);
974 if (!buf) {
975 audit_panic("out of memory for argv string\n");
976 break;
979 ret = copy_from_user(buf, p, len);
981 * There is no reason for this copy to be short. We just
982 * copied them here, and the mm hasn't been exposed to user-
983 * space yet.
985 if (ret) {
986 WARN_ON(1);
987 send_sig(SIGKILL, current, 0);
990 audit_log_format(ab, "a%d=", i);
991 audit_log_untrustedstring(ab, buf);
992 audit_log_format(ab, "\n");
994 kfree(buf);
998 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1000 int i, call_panic = 0;
1001 struct audit_buffer *ab;
1002 struct audit_aux_data *aux;
1003 const char *tty;
1005 /* tsk == current */
1006 context->pid = tsk->pid;
1007 if (!context->ppid)
1008 context->ppid = sys_getppid();
1009 context->uid = tsk->uid;
1010 context->gid = tsk->gid;
1011 context->euid = tsk->euid;
1012 context->suid = tsk->suid;
1013 context->fsuid = tsk->fsuid;
1014 context->egid = tsk->egid;
1015 context->sgid = tsk->sgid;
1016 context->fsgid = tsk->fsgid;
1017 context->personality = tsk->personality;
1019 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1020 if (!ab)
1021 return; /* audit_panic has been called */
1022 audit_log_format(ab, "arch=%x syscall=%d",
1023 context->arch, context->major);
1024 if (context->personality != PER_LINUX)
1025 audit_log_format(ab, " per=%lx", context->personality);
1026 if (context->return_valid)
1027 audit_log_format(ab, " success=%s exit=%ld",
1028 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1029 context->return_code);
1031 mutex_lock(&tty_mutex);
1032 read_lock(&tasklist_lock);
1033 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1034 tty = tsk->signal->tty->name;
1035 else
1036 tty = "(none)";
1037 read_unlock(&tasklist_lock);
1038 audit_log_format(ab,
1039 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1040 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1041 " euid=%u suid=%u fsuid=%u"
1042 " egid=%u sgid=%u fsgid=%u tty=%s",
1043 context->argv[0],
1044 context->argv[1],
1045 context->argv[2],
1046 context->argv[3],
1047 context->name_count,
1048 context->ppid,
1049 context->pid,
1050 context->loginuid,
1051 context->uid,
1052 context->gid,
1053 context->euid, context->suid, context->fsuid,
1054 context->egid, context->sgid, context->fsgid, tty);
1056 mutex_unlock(&tty_mutex);
1058 audit_log_task_info(ab, tsk);
1059 if (context->filterkey) {
1060 audit_log_format(ab, " key=");
1061 audit_log_untrustedstring(ab, context->filterkey);
1062 } else
1063 audit_log_format(ab, " key=(null)");
1064 audit_log_end(ab);
1066 for (aux = context->aux; aux; aux = aux->next) {
1068 ab = audit_log_start(context, GFP_KERNEL, aux->type);
1069 if (!ab)
1070 continue; /* audit_panic has been called */
1072 switch (aux->type) {
1073 case AUDIT_MQ_OPEN: {
1074 struct audit_aux_data_mq_open *axi = (void *)aux;
1075 audit_log_format(ab,
1076 "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
1077 "mq_msgsize=%ld mq_curmsgs=%ld",
1078 axi->oflag, axi->mode, axi->attr.mq_flags,
1079 axi->attr.mq_maxmsg, axi->attr.mq_msgsize,
1080 axi->attr.mq_curmsgs);
1081 break; }
1083 case AUDIT_MQ_SENDRECV: {
1084 struct audit_aux_data_mq_sendrecv *axi = (void *)aux;
1085 audit_log_format(ab,
1086 "mqdes=%d msg_len=%zd msg_prio=%u "
1087 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1088 axi->mqdes, axi->msg_len, axi->msg_prio,
1089 axi->abs_timeout.tv_sec, axi->abs_timeout.tv_nsec);
1090 break; }
1092 case AUDIT_MQ_NOTIFY: {
1093 struct audit_aux_data_mq_notify *axi = (void *)aux;
1094 audit_log_format(ab,
1095 "mqdes=%d sigev_signo=%d",
1096 axi->mqdes,
1097 axi->notification.sigev_signo);
1098 break; }
1100 case AUDIT_MQ_GETSETATTR: {
1101 struct audit_aux_data_mq_getsetattr *axi = (void *)aux;
1102 audit_log_format(ab,
1103 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1104 "mq_curmsgs=%ld ",
1105 axi->mqdes,
1106 axi->mqstat.mq_flags, axi->mqstat.mq_maxmsg,
1107 axi->mqstat.mq_msgsize, axi->mqstat.mq_curmsgs);
1108 break; }
1110 case AUDIT_IPC: {
1111 struct audit_aux_data_ipcctl *axi = (void *)aux;
1112 audit_log_format(ab,
1113 "ouid=%u ogid=%u mode=%#o",
1114 axi->uid, axi->gid, axi->mode);
1115 if (axi->osid != 0) {
1116 char *ctx = NULL;
1117 u32 len;
1118 if (selinux_sid_to_string(
1119 axi->osid, &ctx, &len)) {
1120 audit_log_format(ab, " osid=%u",
1121 axi->osid);
1122 call_panic = 1;
1123 } else
1124 audit_log_format(ab, " obj=%s", ctx);
1125 kfree(ctx);
1127 break; }
1129 case AUDIT_IPC_SET_PERM: {
1130 struct audit_aux_data_ipcctl *axi = (void *)aux;
1131 audit_log_format(ab,
1132 "qbytes=%lx ouid=%u ogid=%u mode=%#o",
1133 axi->qbytes, axi->uid, axi->gid, axi->mode);
1134 break; }
1136 case AUDIT_EXECVE: {
1137 struct audit_aux_data_execve *axi = (void *)aux;
1138 audit_log_execve_info(ab, axi);
1139 break; }
1141 case AUDIT_SOCKETCALL: {
1142 int i;
1143 struct audit_aux_data_socketcall *axs = (void *)aux;
1144 audit_log_format(ab, "nargs=%d", axs->nargs);
1145 for (i=0; i<axs->nargs; i++)
1146 audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
1147 break; }
1149 case AUDIT_SOCKADDR: {
1150 struct audit_aux_data_sockaddr *axs = (void *)aux;
1152 audit_log_format(ab, "saddr=");
1153 audit_log_hex(ab, axs->a, axs->len);
1154 break; }
1156 case AUDIT_FD_PAIR: {
1157 struct audit_aux_data_fd_pair *axs = (void *)aux;
1158 audit_log_format(ab, "fd0=%d fd1=%d", axs->fd[0], axs->fd[1]);
1159 break; }
1162 audit_log_end(ab);
1165 for (aux = context->aux_pids; aux; aux = aux->next) {
1166 struct audit_aux_data_pids *axs = (void *)aux;
1167 int i;
1169 for (i = 0; i < axs->pid_count; i++)
1170 if (audit_log_pid_context(context, axs->target_pid[i],
1171 axs->target_sid[i]))
1172 call_panic = 1;
1175 if (context->target_pid &&
1176 audit_log_pid_context(context, context->target_pid,
1177 context->target_sid))
1178 call_panic = 1;
1180 if (context->pwd && context->pwdmnt) {
1181 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1182 if (ab) {
1183 audit_log_d_path(ab, "cwd=", context->pwd, context->pwdmnt);
1184 audit_log_end(ab);
1187 for (i = 0; i < context->name_count; i++) {
1188 struct audit_names *n = &context->names[i];
1190 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1191 if (!ab)
1192 continue; /* audit_panic has been called */
1194 audit_log_format(ab, "item=%d", i);
1196 if (n->name) {
1197 switch(n->name_len) {
1198 case AUDIT_NAME_FULL:
1199 /* log the full path */
1200 audit_log_format(ab, " name=");
1201 audit_log_untrustedstring(ab, n->name);
1202 break;
1203 case 0:
1204 /* name was specified as a relative path and the
1205 * directory component is the cwd */
1206 audit_log_d_path(ab, " name=", context->pwd,
1207 context->pwdmnt);
1208 break;
1209 default:
1210 /* log the name's directory component */
1211 audit_log_format(ab, " name=");
1212 audit_log_n_untrustedstring(ab, n->name_len,
1213 n->name);
1215 } else
1216 audit_log_format(ab, " name=(null)");
1218 if (n->ino != (unsigned long)-1) {
1219 audit_log_format(ab, " inode=%lu"
1220 " dev=%02x:%02x mode=%#o"
1221 " ouid=%u ogid=%u rdev=%02x:%02x",
1222 n->ino,
1223 MAJOR(n->dev),
1224 MINOR(n->dev),
1225 n->mode,
1226 n->uid,
1227 n->gid,
1228 MAJOR(n->rdev),
1229 MINOR(n->rdev));
1231 if (n->osid != 0) {
1232 char *ctx = NULL;
1233 u32 len;
1234 if (selinux_sid_to_string(
1235 n->osid, &ctx, &len)) {
1236 audit_log_format(ab, " osid=%u", n->osid);
1237 call_panic = 2;
1238 } else
1239 audit_log_format(ab, " obj=%s", ctx);
1240 kfree(ctx);
1243 audit_log_end(ab);
1245 if (call_panic)
1246 audit_panic("error converting sid to string");
1250 * audit_free - free a per-task audit context
1251 * @tsk: task whose audit context block to free
1253 * Called from copy_process and do_exit
1255 void audit_free(struct task_struct *tsk)
1257 struct audit_context *context;
1259 context = audit_get_context(tsk, 0, 0);
1260 if (likely(!context))
1261 return;
1263 /* Check for system calls that do not go through the exit
1264 * function (e.g., exit_group), then free context block.
1265 * We use GFP_ATOMIC here because we might be doing this
1266 * in the context of the idle thread */
1267 /* that can happen only if we are called from do_exit() */
1268 if (context->in_syscall && context->auditable)
1269 audit_log_exit(context, tsk);
1271 audit_free_context(context);
1275 * audit_syscall_entry - fill in an audit record at syscall entry
1276 * @tsk: task being audited
1277 * @arch: architecture type
1278 * @major: major syscall type (function)
1279 * @a1: additional syscall register 1
1280 * @a2: additional syscall register 2
1281 * @a3: additional syscall register 3
1282 * @a4: additional syscall register 4
1284 * Fill in audit context at syscall entry. This only happens if the
1285 * audit context was created when the task was created and the state or
1286 * filters demand the audit context be built. If the state from the
1287 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1288 * then the record will be written at syscall exit time (otherwise, it
1289 * will only be written if another part of the kernel requests that it
1290 * be written).
1292 void audit_syscall_entry(int arch, int major,
1293 unsigned long a1, unsigned long a2,
1294 unsigned long a3, unsigned long a4)
1296 struct task_struct *tsk = current;
1297 struct audit_context *context = tsk->audit_context;
1298 enum audit_state state;
1300 BUG_ON(!context);
1303 * This happens only on certain architectures that make system
1304 * calls in kernel_thread via the entry.S interface, instead of
1305 * with direct calls. (If you are porting to a new
1306 * architecture, hitting this condition can indicate that you
1307 * got the _exit/_leave calls backward in entry.S.)
1309 * i386 no
1310 * x86_64 no
1311 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1313 * This also happens with vm86 emulation in a non-nested manner
1314 * (entries without exits), so this case must be caught.
1316 if (context->in_syscall) {
1317 struct audit_context *newctx;
1319 #if AUDIT_DEBUG
1320 printk(KERN_ERR
1321 "audit(:%d) pid=%d in syscall=%d;"
1322 " entering syscall=%d\n",
1323 context->serial, tsk->pid, context->major, major);
1324 #endif
1325 newctx = audit_alloc_context(context->state);
1326 if (newctx) {
1327 newctx->previous = context;
1328 context = newctx;
1329 tsk->audit_context = newctx;
1330 } else {
1331 /* If we can't alloc a new context, the best we
1332 * can do is to leak memory (any pending putname
1333 * will be lost). The only other alternative is
1334 * to abandon auditing. */
1335 audit_zero_context(context, context->state);
1338 BUG_ON(context->in_syscall || context->name_count);
1340 if (!audit_enabled)
1341 return;
1343 context->arch = arch;
1344 context->major = major;
1345 context->argv[0] = a1;
1346 context->argv[1] = a2;
1347 context->argv[2] = a3;
1348 context->argv[3] = a4;
1350 state = context->state;
1351 context->dummy = !audit_n_rules;
1352 if (!context->dummy && (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT))
1353 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1354 if (likely(state == AUDIT_DISABLED))
1355 return;
1357 context->serial = 0;
1358 context->ctime = CURRENT_TIME;
1359 context->in_syscall = 1;
1360 context->auditable = !!(state == AUDIT_RECORD_CONTEXT);
1361 context->ppid = 0;
1365 * audit_syscall_exit - deallocate audit context after a system call
1366 * @tsk: task being audited
1367 * @valid: success/failure flag
1368 * @return_code: syscall return value
1370 * Tear down after system call. If the audit context has been marked as
1371 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1372 * filtering, or because some other part of the kernel write an audit
1373 * message), then write out the syscall information. In call cases,
1374 * free the names stored from getname().
1376 void audit_syscall_exit(int valid, long return_code)
1378 struct task_struct *tsk = current;
1379 struct audit_context *context;
1381 context = audit_get_context(tsk, valid, return_code);
1383 if (likely(!context))
1384 return;
1386 if (context->in_syscall && context->auditable)
1387 audit_log_exit(context, tsk);
1389 context->in_syscall = 0;
1390 context->auditable = 0;
1392 if (context->previous) {
1393 struct audit_context *new_context = context->previous;
1394 context->previous = NULL;
1395 audit_free_context(context);
1396 tsk->audit_context = new_context;
1397 } else {
1398 audit_free_names(context);
1399 unroll_tree_refs(context, NULL, 0);
1400 audit_free_aux(context);
1401 context->aux = NULL;
1402 context->aux_pids = NULL;
1403 context->target_pid = 0;
1404 context->target_sid = 0;
1405 kfree(context->filterkey);
1406 context->filterkey = NULL;
1407 tsk->audit_context = context;
1411 static inline void handle_one(const struct inode *inode)
1413 #ifdef CONFIG_AUDIT_TREE
1414 struct audit_context *context;
1415 struct audit_tree_refs *p;
1416 struct audit_chunk *chunk;
1417 int count;
1418 if (likely(list_empty(&inode->inotify_watches)))
1419 return;
1420 context = current->audit_context;
1421 p = context->trees;
1422 count = context->tree_count;
1423 rcu_read_lock();
1424 chunk = audit_tree_lookup(inode);
1425 rcu_read_unlock();
1426 if (!chunk)
1427 return;
1428 if (likely(put_tree_ref(context, chunk)))
1429 return;
1430 if (unlikely(!grow_tree_refs(context))) {
1431 printk(KERN_WARNING "out of memory, audit has lost a tree reference");
1432 audit_set_auditable(context);
1433 audit_put_chunk(chunk);
1434 unroll_tree_refs(context, p, count);
1435 return;
1437 put_tree_ref(context, chunk);
1438 #endif
1441 static void handle_path(const struct dentry *dentry)
1443 #ifdef CONFIG_AUDIT_TREE
1444 struct audit_context *context;
1445 struct audit_tree_refs *p;
1446 const struct dentry *d, *parent;
1447 struct audit_chunk *drop;
1448 unsigned long seq;
1449 int count;
1451 context = current->audit_context;
1452 p = context->trees;
1453 count = context->tree_count;
1454 retry:
1455 drop = NULL;
1456 d = dentry;
1457 rcu_read_lock();
1458 seq = read_seqbegin(&rename_lock);
1459 for(;;) {
1460 struct inode *inode = d->d_inode;
1461 if (inode && unlikely(!list_empty(&inode->inotify_watches))) {
1462 struct audit_chunk *chunk;
1463 chunk = audit_tree_lookup(inode);
1464 if (chunk) {
1465 if (unlikely(!put_tree_ref(context, chunk))) {
1466 drop = chunk;
1467 break;
1471 parent = d->d_parent;
1472 if (parent == d)
1473 break;
1474 d = parent;
1476 if (unlikely(read_seqretry(&rename_lock, seq) || drop)) { /* in this order */
1477 rcu_read_unlock();
1478 if (!drop) {
1479 /* just a race with rename */
1480 unroll_tree_refs(context, p, count);
1481 goto retry;
1483 audit_put_chunk(drop);
1484 if (grow_tree_refs(context)) {
1485 /* OK, got more space */
1486 unroll_tree_refs(context, p, count);
1487 goto retry;
1489 /* too bad */
1490 printk(KERN_WARNING
1491 "out of memory, audit has lost a tree reference");
1492 unroll_tree_refs(context, p, count);
1493 audit_set_auditable(context);
1494 return;
1496 rcu_read_unlock();
1497 #endif
1501 * audit_getname - add a name to the list
1502 * @name: name to add
1504 * Add a name to the list of audit names for this context.
1505 * Called from fs/namei.c:getname().
1507 void __audit_getname(const char *name)
1509 struct audit_context *context = current->audit_context;
1511 if (IS_ERR(name) || !name)
1512 return;
1514 if (!context->in_syscall) {
1515 #if AUDIT_DEBUG == 2
1516 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1517 __FILE__, __LINE__, context->serial, name);
1518 dump_stack();
1519 #endif
1520 return;
1522 BUG_ON(context->name_count >= AUDIT_NAMES);
1523 context->names[context->name_count].name = name;
1524 context->names[context->name_count].name_len = AUDIT_NAME_FULL;
1525 context->names[context->name_count].name_put = 1;
1526 context->names[context->name_count].ino = (unsigned long)-1;
1527 context->names[context->name_count].osid = 0;
1528 ++context->name_count;
1529 if (!context->pwd) {
1530 read_lock(&current->fs->lock);
1531 context->pwd = dget(current->fs->pwd);
1532 context->pwdmnt = mntget(current->fs->pwdmnt);
1533 read_unlock(&current->fs->lock);
1538 /* audit_putname - intercept a putname request
1539 * @name: name to intercept and delay for putname
1541 * If we have stored the name from getname in the audit context,
1542 * then we delay the putname until syscall exit.
1543 * Called from include/linux/fs.h:putname().
1545 void audit_putname(const char *name)
1547 struct audit_context *context = current->audit_context;
1549 BUG_ON(!context);
1550 if (!context->in_syscall) {
1551 #if AUDIT_DEBUG == 2
1552 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1553 __FILE__, __LINE__, context->serial, name);
1554 if (context->name_count) {
1555 int i;
1556 for (i = 0; i < context->name_count; i++)
1557 printk(KERN_ERR "name[%d] = %p = %s\n", i,
1558 context->names[i].name,
1559 context->names[i].name ?: "(null)");
1561 #endif
1562 __putname(name);
1564 #if AUDIT_DEBUG
1565 else {
1566 ++context->put_count;
1567 if (context->put_count > context->name_count) {
1568 printk(KERN_ERR "%s:%d(:%d): major=%d"
1569 " in_syscall=%d putname(%p) name_count=%d"
1570 " put_count=%d\n",
1571 __FILE__, __LINE__,
1572 context->serial, context->major,
1573 context->in_syscall, name, context->name_count,
1574 context->put_count);
1575 dump_stack();
1578 #endif
1581 static int audit_inc_name_count(struct audit_context *context,
1582 const struct inode *inode)
1584 if (context->name_count >= AUDIT_NAMES) {
1585 if (inode)
1586 printk(KERN_DEBUG "name_count maxed, losing inode data: "
1587 "dev=%02x:%02x, inode=%lu",
1588 MAJOR(inode->i_sb->s_dev),
1589 MINOR(inode->i_sb->s_dev),
1590 inode->i_ino);
1592 else
1593 printk(KERN_DEBUG "name_count maxed, losing inode data");
1594 return 1;
1596 context->name_count++;
1597 #if AUDIT_DEBUG
1598 context->ino_count++;
1599 #endif
1600 return 0;
1603 /* Copy inode data into an audit_names. */
1604 static void audit_copy_inode(struct audit_names *name, const struct inode *inode)
1606 name->ino = inode->i_ino;
1607 name->dev = inode->i_sb->s_dev;
1608 name->mode = inode->i_mode;
1609 name->uid = inode->i_uid;
1610 name->gid = inode->i_gid;
1611 name->rdev = inode->i_rdev;
1612 selinux_get_inode_sid(inode, &name->osid);
1616 * audit_inode - store the inode and device from a lookup
1617 * @name: name being audited
1618 * @dentry: dentry being audited
1620 * Called from fs/namei.c:path_lookup().
1622 void __audit_inode(const char *name, const struct dentry *dentry)
1624 int idx;
1625 struct audit_context *context = current->audit_context;
1626 const struct inode *inode = dentry->d_inode;
1628 if (!context->in_syscall)
1629 return;
1630 if (context->name_count
1631 && context->names[context->name_count-1].name
1632 && context->names[context->name_count-1].name == name)
1633 idx = context->name_count - 1;
1634 else if (context->name_count > 1
1635 && context->names[context->name_count-2].name
1636 && context->names[context->name_count-2].name == name)
1637 idx = context->name_count - 2;
1638 else {
1639 /* FIXME: how much do we care about inodes that have no
1640 * associated name? */
1641 if (audit_inc_name_count(context, inode))
1642 return;
1643 idx = context->name_count - 1;
1644 context->names[idx].name = NULL;
1646 handle_path(dentry);
1647 audit_copy_inode(&context->names[idx], inode);
1651 * audit_inode_child - collect inode info for created/removed objects
1652 * @dname: inode's dentry name
1653 * @dentry: dentry being audited
1654 * @parent: inode of dentry parent
1656 * For syscalls that create or remove filesystem objects, audit_inode
1657 * can only collect information for the filesystem object's parent.
1658 * This call updates the audit context with the child's information.
1659 * Syscalls that create a new filesystem object must be hooked after
1660 * the object is created. Syscalls that remove a filesystem object
1661 * must be hooked prior, in order to capture the target inode during
1662 * unsuccessful attempts.
1664 void __audit_inode_child(const char *dname, const struct dentry *dentry,
1665 const struct inode *parent)
1667 int idx;
1668 struct audit_context *context = current->audit_context;
1669 const char *found_parent = NULL, *found_child = NULL;
1670 const struct inode *inode = dentry->d_inode;
1671 int dirlen = 0;
1673 if (!context->in_syscall)
1674 return;
1676 if (inode)
1677 handle_one(inode);
1678 /* determine matching parent */
1679 if (!dname)
1680 goto add_names;
1682 /* parent is more likely, look for it first */
1683 for (idx = 0; idx < context->name_count; idx++) {
1684 struct audit_names *n = &context->names[idx];
1686 if (!n->name)
1687 continue;
1689 if (n->ino == parent->i_ino &&
1690 !audit_compare_dname_path(dname, n->name, &dirlen)) {
1691 n->name_len = dirlen; /* update parent data in place */
1692 found_parent = n->name;
1693 goto add_names;
1697 /* no matching parent, look for matching child */
1698 for (idx = 0; idx < context->name_count; idx++) {
1699 struct audit_names *n = &context->names[idx];
1701 if (!n->name)
1702 continue;
1704 /* strcmp() is the more likely scenario */
1705 if (!strcmp(dname, n->name) ||
1706 !audit_compare_dname_path(dname, n->name, &dirlen)) {
1707 if (inode)
1708 audit_copy_inode(n, inode);
1709 else
1710 n->ino = (unsigned long)-1;
1711 found_child = n->name;
1712 goto add_names;
1716 add_names:
1717 if (!found_parent) {
1718 if (audit_inc_name_count(context, parent))
1719 return;
1720 idx = context->name_count - 1;
1721 context->names[idx].name = NULL;
1722 audit_copy_inode(&context->names[idx], parent);
1725 if (!found_child) {
1726 if (audit_inc_name_count(context, inode))
1727 return;
1728 idx = context->name_count - 1;
1730 /* Re-use the name belonging to the slot for a matching parent
1731 * directory. All names for this context are relinquished in
1732 * audit_free_names() */
1733 if (found_parent) {
1734 context->names[idx].name = found_parent;
1735 context->names[idx].name_len = AUDIT_NAME_FULL;
1736 /* don't call __putname() */
1737 context->names[idx].name_put = 0;
1738 } else {
1739 context->names[idx].name = NULL;
1742 if (inode)
1743 audit_copy_inode(&context->names[idx], inode);
1744 else
1745 context->names[idx].ino = (unsigned long)-1;
1748 EXPORT_SYMBOL_GPL(__audit_inode_child);
1751 * auditsc_get_stamp - get local copies of audit_context values
1752 * @ctx: audit_context for the task
1753 * @t: timespec to store time recorded in the audit_context
1754 * @serial: serial value that is recorded in the audit_context
1756 * Also sets the context as auditable.
1758 void auditsc_get_stamp(struct audit_context *ctx,
1759 struct timespec *t, unsigned int *serial)
1761 if (!ctx->serial)
1762 ctx->serial = audit_serial();
1763 t->tv_sec = ctx->ctime.tv_sec;
1764 t->tv_nsec = ctx->ctime.tv_nsec;
1765 *serial = ctx->serial;
1766 ctx->auditable = 1;
1770 * audit_set_loginuid - set a task's audit_context loginuid
1771 * @task: task whose audit context is being modified
1772 * @loginuid: loginuid value
1774 * Returns 0.
1776 * Called (set) from fs/proc/base.c::proc_loginuid_write().
1778 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
1780 struct audit_context *context = task->audit_context;
1782 if (context) {
1783 /* Only log if audit is enabled */
1784 if (context->in_syscall) {
1785 struct audit_buffer *ab;
1787 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1788 if (ab) {
1789 audit_log_format(ab, "login pid=%d uid=%u "
1790 "old auid=%u new auid=%u",
1791 task->pid, task->uid,
1792 context->loginuid, loginuid);
1793 audit_log_end(ab);
1796 context->loginuid = loginuid;
1798 return 0;
1802 * audit_get_loginuid - get the loginuid for an audit_context
1803 * @ctx: the audit_context
1805 * Returns the context's loginuid or -1 if @ctx is NULL.
1807 uid_t audit_get_loginuid(struct audit_context *ctx)
1809 return ctx ? ctx->loginuid : -1;
1812 EXPORT_SYMBOL(audit_get_loginuid);
1815 * __audit_mq_open - record audit data for a POSIX MQ open
1816 * @oflag: open flag
1817 * @mode: mode bits
1818 * @u_attr: queue attributes
1820 * Returns 0 for success or NULL context or < 0 on error.
1822 int __audit_mq_open(int oflag, mode_t mode, struct mq_attr __user *u_attr)
1824 struct audit_aux_data_mq_open *ax;
1825 struct audit_context *context = current->audit_context;
1827 if (!audit_enabled)
1828 return 0;
1830 if (likely(!context))
1831 return 0;
1833 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1834 if (!ax)
1835 return -ENOMEM;
1837 if (u_attr != NULL) {
1838 if (copy_from_user(&ax->attr, u_attr, sizeof(ax->attr))) {
1839 kfree(ax);
1840 return -EFAULT;
1842 } else
1843 memset(&ax->attr, 0, sizeof(ax->attr));
1845 ax->oflag = oflag;
1846 ax->mode = mode;
1848 ax->d.type = AUDIT_MQ_OPEN;
1849 ax->d.next = context->aux;
1850 context->aux = (void *)ax;
1851 return 0;
1855 * __audit_mq_timedsend - record audit data for a POSIX MQ timed send
1856 * @mqdes: MQ descriptor
1857 * @msg_len: Message length
1858 * @msg_prio: Message priority
1859 * @u_abs_timeout: Message timeout in absolute time
1861 * Returns 0 for success or NULL context or < 0 on error.
1863 int __audit_mq_timedsend(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
1864 const struct timespec __user *u_abs_timeout)
1866 struct audit_aux_data_mq_sendrecv *ax;
1867 struct audit_context *context = current->audit_context;
1869 if (!audit_enabled)
1870 return 0;
1872 if (likely(!context))
1873 return 0;
1875 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1876 if (!ax)
1877 return -ENOMEM;
1879 if (u_abs_timeout != NULL) {
1880 if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
1881 kfree(ax);
1882 return -EFAULT;
1884 } else
1885 memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
1887 ax->mqdes = mqdes;
1888 ax->msg_len = msg_len;
1889 ax->msg_prio = msg_prio;
1891 ax->d.type = AUDIT_MQ_SENDRECV;
1892 ax->d.next = context->aux;
1893 context->aux = (void *)ax;
1894 return 0;
1898 * __audit_mq_timedreceive - record audit data for a POSIX MQ timed receive
1899 * @mqdes: MQ descriptor
1900 * @msg_len: Message length
1901 * @u_msg_prio: Message priority
1902 * @u_abs_timeout: Message timeout in absolute time
1904 * Returns 0 for success or NULL context or < 0 on error.
1906 int __audit_mq_timedreceive(mqd_t mqdes, size_t msg_len,
1907 unsigned int __user *u_msg_prio,
1908 const struct timespec __user *u_abs_timeout)
1910 struct audit_aux_data_mq_sendrecv *ax;
1911 struct audit_context *context = current->audit_context;
1913 if (!audit_enabled)
1914 return 0;
1916 if (likely(!context))
1917 return 0;
1919 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1920 if (!ax)
1921 return -ENOMEM;
1923 if (u_msg_prio != NULL) {
1924 if (get_user(ax->msg_prio, u_msg_prio)) {
1925 kfree(ax);
1926 return -EFAULT;
1928 } else
1929 ax->msg_prio = 0;
1931 if (u_abs_timeout != NULL) {
1932 if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
1933 kfree(ax);
1934 return -EFAULT;
1936 } else
1937 memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
1939 ax->mqdes = mqdes;
1940 ax->msg_len = msg_len;
1942 ax->d.type = AUDIT_MQ_SENDRECV;
1943 ax->d.next = context->aux;
1944 context->aux = (void *)ax;
1945 return 0;
1949 * __audit_mq_notify - record audit data for a POSIX MQ notify
1950 * @mqdes: MQ descriptor
1951 * @u_notification: Notification event
1953 * Returns 0 for success or NULL context or < 0 on error.
1956 int __audit_mq_notify(mqd_t mqdes, const struct sigevent __user *u_notification)
1958 struct audit_aux_data_mq_notify *ax;
1959 struct audit_context *context = current->audit_context;
1961 if (!audit_enabled)
1962 return 0;
1964 if (likely(!context))
1965 return 0;
1967 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1968 if (!ax)
1969 return -ENOMEM;
1971 if (u_notification != NULL) {
1972 if (copy_from_user(&ax->notification, u_notification, sizeof(ax->notification))) {
1973 kfree(ax);
1974 return -EFAULT;
1976 } else
1977 memset(&ax->notification, 0, sizeof(ax->notification));
1979 ax->mqdes = mqdes;
1981 ax->d.type = AUDIT_MQ_NOTIFY;
1982 ax->d.next = context->aux;
1983 context->aux = (void *)ax;
1984 return 0;
1988 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
1989 * @mqdes: MQ descriptor
1990 * @mqstat: MQ flags
1992 * Returns 0 for success or NULL context or < 0 on error.
1994 int __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
1996 struct audit_aux_data_mq_getsetattr *ax;
1997 struct audit_context *context = current->audit_context;
1999 if (!audit_enabled)
2000 return 0;
2002 if (likely(!context))
2003 return 0;
2005 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2006 if (!ax)
2007 return -ENOMEM;
2009 ax->mqdes = mqdes;
2010 ax->mqstat = *mqstat;
2012 ax->d.type = AUDIT_MQ_GETSETATTR;
2013 ax->d.next = context->aux;
2014 context->aux = (void *)ax;
2015 return 0;
2019 * audit_ipc_obj - record audit data for ipc object
2020 * @ipcp: ipc permissions
2022 * Returns 0 for success or NULL context or < 0 on error.
2024 int __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2026 struct audit_aux_data_ipcctl *ax;
2027 struct audit_context *context = current->audit_context;
2029 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2030 if (!ax)
2031 return -ENOMEM;
2033 ax->uid = ipcp->uid;
2034 ax->gid = ipcp->gid;
2035 ax->mode = ipcp->mode;
2036 selinux_get_ipc_sid(ipcp, &ax->osid);
2038 ax->d.type = AUDIT_IPC;
2039 ax->d.next = context->aux;
2040 context->aux = (void *)ax;
2041 return 0;
2045 * audit_ipc_set_perm - record audit data for new ipc permissions
2046 * @qbytes: msgq bytes
2047 * @uid: msgq user id
2048 * @gid: msgq group id
2049 * @mode: msgq mode (permissions)
2051 * Returns 0 for success or NULL context or < 0 on error.
2053 int __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
2055 struct audit_aux_data_ipcctl *ax;
2056 struct audit_context *context = current->audit_context;
2058 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2059 if (!ax)
2060 return -ENOMEM;
2062 ax->qbytes = qbytes;
2063 ax->uid = uid;
2064 ax->gid = gid;
2065 ax->mode = mode;
2067 ax->d.type = AUDIT_IPC_SET_PERM;
2068 ax->d.next = context->aux;
2069 context->aux = (void *)ax;
2070 return 0;
2073 int audit_argv_kb = 32;
2075 int audit_bprm(struct linux_binprm *bprm)
2077 struct audit_aux_data_execve *ax;
2078 struct audit_context *context = current->audit_context;
2080 if (likely(!audit_enabled || !context || context->dummy))
2081 return 0;
2084 * Even though the stack code doesn't limit the arg+env size any more,
2085 * the audit code requires that _all_ arguments be logged in a single
2086 * netlink skb. Hence cap it :-(
2088 if (bprm->argv_len > (audit_argv_kb << 10))
2089 return -E2BIG;
2091 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2092 if (!ax)
2093 return -ENOMEM;
2095 ax->argc = bprm->argc;
2096 ax->envc = bprm->envc;
2097 ax->mm = bprm->mm;
2098 ax->d.type = AUDIT_EXECVE;
2099 ax->d.next = context->aux;
2100 context->aux = (void *)ax;
2101 return 0;
2106 * audit_socketcall - record audit data for sys_socketcall
2107 * @nargs: number of args
2108 * @args: args array
2110 * Returns 0 for success or NULL context or < 0 on error.
2112 int audit_socketcall(int nargs, unsigned long *args)
2114 struct audit_aux_data_socketcall *ax;
2115 struct audit_context *context = current->audit_context;
2117 if (likely(!context || context->dummy))
2118 return 0;
2120 ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
2121 if (!ax)
2122 return -ENOMEM;
2124 ax->nargs = nargs;
2125 memcpy(ax->args, args, nargs * sizeof(unsigned long));
2127 ax->d.type = AUDIT_SOCKETCALL;
2128 ax->d.next = context->aux;
2129 context->aux = (void *)ax;
2130 return 0;
2134 * __audit_fd_pair - record audit data for pipe and socketpair
2135 * @fd1: the first file descriptor
2136 * @fd2: the second file descriptor
2138 * Returns 0 for success or NULL context or < 0 on error.
2140 int __audit_fd_pair(int fd1, int fd2)
2142 struct audit_context *context = current->audit_context;
2143 struct audit_aux_data_fd_pair *ax;
2145 if (likely(!context)) {
2146 return 0;
2149 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2150 if (!ax) {
2151 return -ENOMEM;
2154 ax->fd[0] = fd1;
2155 ax->fd[1] = fd2;
2157 ax->d.type = AUDIT_FD_PAIR;
2158 ax->d.next = context->aux;
2159 context->aux = (void *)ax;
2160 return 0;
2164 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2165 * @len: data length in user space
2166 * @a: data address in kernel space
2168 * Returns 0 for success or NULL context or < 0 on error.
2170 int audit_sockaddr(int len, void *a)
2172 struct audit_aux_data_sockaddr *ax;
2173 struct audit_context *context = current->audit_context;
2175 if (likely(!context || context->dummy))
2176 return 0;
2178 ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
2179 if (!ax)
2180 return -ENOMEM;
2182 ax->len = len;
2183 memcpy(ax->a, a, len);
2185 ax->d.type = AUDIT_SOCKADDR;
2186 ax->d.next = context->aux;
2187 context->aux = (void *)ax;
2188 return 0;
2191 void __audit_ptrace(struct task_struct *t)
2193 struct audit_context *context = current->audit_context;
2195 context->target_pid = t->pid;
2196 selinux_get_task_sid(t, &context->target_sid);
2200 * audit_signal_info - record signal info for shutting down audit subsystem
2201 * @sig: signal value
2202 * @t: task being signaled
2204 * If the audit subsystem is being terminated, record the task (pid)
2205 * and uid that is doing that.
2207 int __audit_signal_info(int sig, struct task_struct *t)
2209 struct audit_aux_data_pids *axp;
2210 struct task_struct *tsk = current;
2211 struct audit_context *ctx = tsk->audit_context;
2212 extern pid_t audit_sig_pid;
2213 extern uid_t audit_sig_uid;
2214 extern u32 audit_sig_sid;
2216 if (audit_pid && t->tgid == audit_pid) {
2217 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1) {
2218 audit_sig_pid = tsk->pid;
2219 if (ctx)
2220 audit_sig_uid = ctx->loginuid;
2221 else
2222 audit_sig_uid = tsk->uid;
2223 selinux_get_task_sid(tsk, &audit_sig_sid);
2225 if (!audit_signals || audit_dummy_context())
2226 return 0;
2229 /* optimize the common case by putting first signal recipient directly
2230 * in audit_context */
2231 if (!ctx->target_pid) {
2232 ctx->target_pid = t->tgid;
2233 selinux_get_task_sid(t, &ctx->target_sid);
2234 return 0;
2237 axp = (void *)ctx->aux_pids;
2238 if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2239 axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2240 if (!axp)
2241 return -ENOMEM;
2243 axp->d.type = AUDIT_OBJ_PID;
2244 axp->d.next = ctx->aux_pids;
2245 ctx->aux_pids = (void *)axp;
2247 BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2249 axp->target_pid[axp->pid_count] = t->tgid;
2250 selinux_get_task_sid(t, &axp->target_sid[axp->pid_count]);
2251 axp->pid_count++;
2253 return 0;
2257 * audit_core_dumps - record information about processes that end abnormally
2258 * @signr: signal value
2260 * If a process ends with a core dump, something fishy is going on and we
2261 * should record the event for investigation.
2263 void audit_core_dumps(long signr)
2265 struct audit_buffer *ab;
2266 u32 sid;
2268 if (!audit_enabled)
2269 return;
2271 if (signr == SIGQUIT) /* don't care for those */
2272 return;
2274 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2275 audit_log_format(ab, "auid=%u uid=%u gid=%u",
2276 audit_get_loginuid(current->audit_context),
2277 current->uid, current->gid);
2278 selinux_get_task_sid(current, &sid);
2279 if (sid) {
2280 char *ctx = NULL;
2281 u32 len;
2283 if (selinux_sid_to_string(sid, &ctx, &len))
2284 audit_log_format(ab, " ssid=%u", sid);
2285 else
2286 audit_log_format(ab, " subj=%s", ctx);
2287 kfree(ctx);
2289 audit_log_format(ab, " pid=%d comm=", current->pid);
2290 audit_log_untrustedstring(ab, current->comm);
2291 audit_log_format(ab, " sig=%ld", signr);
2292 audit_log_end(ab);