xref: /linux/kernel/auditsc.c (revision c4ee0af3fa0dc65f690fc908f02b8355f9576ea0)
1 /* auditsc.c -- System-call auditing support
2  * Handles all system-call specific auditing features.
3  *
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.
8  *
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.
13  *
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.
18  *
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
22  *
23  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
24  *
25  * Many of the ideas implemented here are from Stephen C. Tweedie,
26  * especially the idea of avoiding a copy by using getname.
27  *
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.
31  *
32  * POSIX message queue support added by George Wilson <ltcgcw@us.ibm.com>,
33  * 2006.
34  *
35  * The support of additional filter rules compares (>, <, >=, <=) was
36  * added by Dustin Kirkland <dustin.kirkland@us.ibm.com>, 2005.
37  *
38  * Modified by Amy Griffis <amy.griffis@hp.com> to collect additional
39  * filesystem information.
40  *
41  * Subject and object context labeling support added by <danjones@us.ibm.com>
42  * and <dustin.kirkland@us.ibm.com> for LSPP certification compliance.
43  */
44 
45 #include <linux/init.h>
46 #include <asm/types.h>
47 #include <linux/atomic.h>
48 #include <linux/fs.h>
49 #include <linux/namei.h>
50 #include <linux/mm.h>
51 #include <linux/export.h>
52 #include <linux/slab.h>
53 #include <linux/mount.h>
54 #include <linux/socket.h>
55 #include <linux/mqueue.h>
56 #include <linux/audit.h>
57 #include <linux/personality.h>
58 #include <linux/time.h>
59 #include <linux/netlink.h>
60 #include <linux/compiler.h>
61 #include <asm/unistd.h>
62 #include <linux/security.h>
63 #include <linux/list.h>
64 #include <linux/tty.h>
65 #include <linux/binfmts.h>
66 #include <linux/highmem.h>
67 #include <linux/syscalls.h>
68 #include <linux/capability.h>
69 #include <linux/fs_struct.h>
70 #include <linux/compat.h>
71 
72 #include "audit.h"
73 
74 /* flags stating the success for a syscall */
75 #define AUDITSC_INVALID 0
76 #define AUDITSC_SUCCESS 1
77 #define AUDITSC_FAILURE 2
78 
79 /* no execve audit message should be longer than this (userspace limits) */
80 #define MAX_EXECVE_AUDIT_LEN 7500
81 
82 /* number of audit rules */
83 int audit_n_rules;
84 
85 /* determines whether we collect data for signals sent */
86 int audit_signals;
87 
88 struct audit_aux_data {
89 	struct audit_aux_data	*next;
90 	int			type;
91 };
92 
93 #define AUDIT_AUX_IPCPERM	0
94 
95 /* Number of target pids per aux struct. */
96 #define AUDIT_AUX_PIDS	16
97 
98 struct audit_aux_data_pids {
99 	struct audit_aux_data	d;
100 	pid_t			target_pid[AUDIT_AUX_PIDS];
101 	kuid_t			target_auid[AUDIT_AUX_PIDS];
102 	kuid_t			target_uid[AUDIT_AUX_PIDS];
103 	unsigned int		target_sessionid[AUDIT_AUX_PIDS];
104 	u32			target_sid[AUDIT_AUX_PIDS];
105 	char 			target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
106 	int			pid_count;
107 };
108 
109 struct audit_aux_data_bprm_fcaps {
110 	struct audit_aux_data	d;
111 	struct audit_cap_data	fcap;
112 	unsigned int		fcap_ver;
113 	struct audit_cap_data	old_pcap;
114 	struct audit_cap_data	new_pcap;
115 };
116 
117 struct audit_tree_refs {
118 	struct audit_tree_refs *next;
119 	struct audit_chunk *c[31];
120 };
121 
122 static inline int open_arg(int flags, int mask)
123 {
124 	int n = ACC_MODE(flags);
125 	if (flags & (O_TRUNC | O_CREAT))
126 		n |= AUDIT_PERM_WRITE;
127 	return n & mask;
128 }
129 
130 static int audit_match_perm(struct audit_context *ctx, int mask)
131 {
132 	unsigned n;
133 	if (unlikely(!ctx))
134 		return 0;
135 	n = ctx->major;
136 
137 	switch (audit_classify_syscall(ctx->arch, n)) {
138 	case 0:	/* native */
139 		if ((mask & AUDIT_PERM_WRITE) &&
140 		     audit_match_class(AUDIT_CLASS_WRITE, n))
141 			return 1;
142 		if ((mask & AUDIT_PERM_READ) &&
143 		     audit_match_class(AUDIT_CLASS_READ, n))
144 			return 1;
145 		if ((mask & AUDIT_PERM_ATTR) &&
146 		     audit_match_class(AUDIT_CLASS_CHATTR, n))
147 			return 1;
148 		return 0;
149 	case 1: /* 32bit on biarch */
150 		if ((mask & AUDIT_PERM_WRITE) &&
151 		     audit_match_class(AUDIT_CLASS_WRITE_32, n))
152 			return 1;
153 		if ((mask & AUDIT_PERM_READ) &&
154 		     audit_match_class(AUDIT_CLASS_READ_32, n))
155 			return 1;
156 		if ((mask & AUDIT_PERM_ATTR) &&
157 		     audit_match_class(AUDIT_CLASS_CHATTR_32, n))
158 			return 1;
159 		return 0;
160 	case 2: /* open */
161 		return mask & ACC_MODE(ctx->argv[1]);
162 	case 3: /* openat */
163 		return mask & ACC_MODE(ctx->argv[2]);
164 	case 4: /* socketcall */
165 		return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
166 	case 5: /* execve */
167 		return mask & AUDIT_PERM_EXEC;
168 	default:
169 		return 0;
170 	}
171 }
172 
173 static int audit_match_filetype(struct audit_context *ctx, int val)
174 {
175 	struct audit_names *n;
176 	umode_t mode = (umode_t)val;
177 
178 	if (unlikely(!ctx))
179 		return 0;
180 
181 	list_for_each_entry(n, &ctx->names_list, list) {
182 		if ((n->ino != -1) &&
183 		    ((n->mode & S_IFMT) == mode))
184 			return 1;
185 	}
186 
187 	return 0;
188 }
189 
190 /*
191  * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
192  * ->first_trees points to its beginning, ->trees - to the current end of data.
193  * ->tree_count is the number of free entries in array pointed to by ->trees.
194  * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
195  * "empty" becomes (p, p, 31) afterwards.  We don't shrink the list (and seriously,
196  * it's going to remain 1-element for almost any setup) until we free context itself.
197  * References in it _are_ dropped - at the same time we free/drop aux stuff.
198  */
199 
200 #ifdef CONFIG_AUDIT_TREE
201 static void audit_set_auditable(struct audit_context *ctx)
202 {
203 	if (!ctx->prio) {
204 		ctx->prio = 1;
205 		ctx->current_state = AUDIT_RECORD_CONTEXT;
206 	}
207 }
208 
209 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
210 {
211 	struct audit_tree_refs *p = ctx->trees;
212 	int left = ctx->tree_count;
213 	if (likely(left)) {
214 		p->c[--left] = chunk;
215 		ctx->tree_count = left;
216 		return 1;
217 	}
218 	if (!p)
219 		return 0;
220 	p = p->next;
221 	if (p) {
222 		p->c[30] = chunk;
223 		ctx->trees = p;
224 		ctx->tree_count = 30;
225 		return 1;
226 	}
227 	return 0;
228 }
229 
230 static int grow_tree_refs(struct audit_context *ctx)
231 {
232 	struct audit_tree_refs *p = ctx->trees;
233 	ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
234 	if (!ctx->trees) {
235 		ctx->trees = p;
236 		return 0;
237 	}
238 	if (p)
239 		p->next = ctx->trees;
240 	else
241 		ctx->first_trees = ctx->trees;
242 	ctx->tree_count = 31;
243 	return 1;
244 }
245 #endif
246 
247 static void unroll_tree_refs(struct audit_context *ctx,
248 		      struct audit_tree_refs *p, int count)
249 {
250 #ifdef CONFIG_AUDIT_TREE
251 	struct audit_tree_refs *q;
252 	int n;
253 	if (!p) {
254 		/* we started with empty chain */
255 		p = ctx->first_trees;
256 		count = 31;
257 		/* if the very first allocation has failed, nothing to do */
258 		if (!p)
259 			return;
260 	}
261 	n = count;
262 	for (q = p; q != ctx->trees; q = q->next, n = 31) {
263 		while (n--) {
264 			audit_put_chunk(q->c[n]);
265 			q->c[n] = NULL;
266 		}
267 	}
268 	while (n-- > ctx->tree_count) {
269 		audit_put_chunk(q->c[n]);
270 		q->c[n] = NULL;
271 	}
272 	ctx->trees = p;
273 	ctx->tree_count = count;
274 #endif
275 }
276 
277 static void free_tree_refs(struct audit_context *ctx)
278 {
279 	struct audit_tree_refs *p, *q;
280 	for (p = ctx->first_trees; p; p = q) {
281 		q = p->next;
282 		kfree(p);
283 	}
284 }
285 
286 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
287 {
288 #ifdef CONFIG_AUDIT_TREE
289 	struct audit_tree_refs *p;
290 	int n;
291 	if (!tree)
292 		return 0;
293 	/* full ones */
294 	for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
295 		for (n = 0; n < 31; n++)
296 			if (audit_tree_match(p->c[n], tree))
297 				return 1;
298 	}
299 	/* partial */
300 	if (p) {
301 		for (n = ctx->tree_count; n < 31; n++)
302 			if (audit_tree_match(p->c[n], tree))
303 				return 1;
304 	}
305 #endif
306 	return 0;
307 }
308 
309 static int audit_compare_uid(kuid_t uid,
310 			     struct audit_names *name,
311 			     struct audit_field *f,
312 			     struct audit_context *ctx)
313 {
314 	struct audit_names *n;
315 	int rc;
316 
317 	if (name) {
318 		rc = audit_uid_comparator(uid, f->op, name->uid);
319 		if (rc)
320 			return rc;
321 	}
322 
323 	if (ctx) {
324 		list_for_each_entry(n, &ctx->names_list, list) {
325 			rc = audit_uid_comparator(uid, f->op, n->uid);
326 			if (rc)
327 				return rc;
328 		}
329 	}
330 	return 0;
331 }
332 
333 static int audit_compare_gid(kgid_t gid,
334 			     struct audit_names *name,
335 			     struct audit_field *f,
336 			     struct audit_context *ctx)
337 {
338 	struct audit_names *n;
339 	int rc;
340 
341 	if (name) {
342 		rc = audit_gid_comparator(gid, f->op, name->gid);
343 		if (rc)
344 			return rc;
345 	}
346 
347 	if (ctx) {
348 		list_for_each_entry(n, &ctx->names_list, list) {
349 			rc = audit_gid_comparator(gid, f->op, n->gid);
350 			if (rc)
351 				return rc;
352 		}
353 	}
354 	return 0;
355 }
356 
357 static int audit_field_compare(struct task_struct *tsk,
358 			       const struct cred *cred,
359 			       struct audit_field *f,
360 			       struct audit_context *ctx,
361 			       struct audit_names *name)
362 {
363 	switch (f->val) {
364 	/* process to file object comparisons */
365 	case AUDIT_COMPARE_UID_TO_OBJ_UID:
366 		return audit_compare_uid(cred->uid, name, f, ctx);
367 	case AUDIT_COMPARE_GID_TO_OBJ_GID:
368 		return audit_compare_gid(cred->gid, name, f, ctx);
369 	case AUDIT_COMPARE_EUID_TO_OBJ_UID:
370 		return audit_compare_uid(cred->euid, name, f, ctx);
371 	case AUDIT_COMPARE_EGID_TO_OBJ_GID:
372 		return audit_compare_gid(cred->egid, name, f, ctx);
373 	case AUDIT_COMPARE_AUID_TO_OBJ_UID:
374 		return audit_compare_uid(tsk->loginuid, name, f, ctx);
375 	case AUDIT_COMPARE_SUID_TO_OBJ_UID:
376 		return audit_compare_uid(cred->suid, name, f, ctx);
377 	case AUDIT_COMPARE_SGID_TO_OBJ_GID:
378 		return audit_compare_gid(cred->sgid, name, f, ctx);
379 	case AUDIT_COMPARE_FSUID_TO_OBJ_UID:
380 		return audit_compare_uid(cred->fsuid, name, f, ctx);
381 	case AUDIT_COMPARE_FSGID_TO_OBJ_GID:
382 		return audit_compare_gid(cred->fsgid, name, f, ctx);
383 	/* uid comparisons */
384 	case AUDIT_COMPARE_UID_TO_AUID:
385 		return audit_uid_comparator(cred->uid, f->op, tsk->loginuid);
386 	case AUDIT_COMPARE_UID_TO_EUID:
387 		return audit_uid_comparator(cred->uid, f->op, cred->euid);
388 	case AUDIT_COMPARE_UID_TO_SUID:
389 		return audit_uid_comparator(cred->uid, f->op, cred->suid);
390 	case AUDIT_COMPARE_UID_TO_FSUID:
391 		return audit_uid_comparator(cred->uid, f->op, cred->fsuid);
392 	/* auid comparisons */
393 	case AUDIT_COMPARE_AUID_TO_EUID:
394 		return audit_uid_comparator(tsk->loginuid, f->op, cred->euid);
395 	case AUDIT_COMPARE_AUID_TO_SUID:
396 		return audit_uid_comparator(tsk->loginuid, f->op, cred->suid);
397 	case AUDIT_COMPARE_AUID_TO_FSUID:
398 		return audit_uid_comparator(tsk->loginuid, f->op, cred->fsuid);
399 	/* euid comparisons */
400 	case AUDIT_COMPARE_EUID_TO_SUID:
401 		return audit_uid_comparator(cred->euid, f->op, cred->suid);
402 	case AUDIT_COMPARE_EUID_TO_FSUID:
403 		return audit_uid_comparator(cred->euid, f->op, cred->fsuid);
404 	/* suid comparisons */
405 	case AUDIT_COMPARE_SUID_TO_FSUID:
406 		return audit_uid_comparator(cred->suid, f->op, cred->fsuid);
407 	/* gid comparisons */
408 	case AUDIT_COMPARE_GID_TO_EGID:
409 		return audit_gid_comparator(cred->gid, f->op, cred->egid);
410 	case AUDIT_COMPARE_GID_TO_SGID:
411 		return audit_gid_comparator(cred->gid, f->op, cred->sgid);
412 	case AUDIT_COMPARE_GID_TO_FSGID:
413 		return audit_gid_comparator(cred->gid, f->op, cred->fsgid);
414 	/* egid comparisons */
415 	case AUDIT_COMPARE_EGID_TO_SGID:
416 		return audit_gid_comparator(cred->egid, f->op, cred->sgid);
417 	case AUDIT_COMPARE_EGID_TO_FSGID:
418 		return audit_gid_comparator(cred->egid, f->op, cred->fsgid);
419 	/* sgid comparison */
420 	case AUDIT_COMPARE_SGID_TO_FSGID:
421 		return audit_gid_comparator(cred->sgid, f->op, cred->fsgid);
422 	default:
423 		WARN(1, "Missing AUDIT_COMPARE define.  Report as a bug\n");
424 		return 0;
425 	}
426 	return 0;
427 }
428 
429 /* Determine if any context name data matches a rule's watch data */
430 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
431  * otherwise.
432  *
433  * If task_creation is true, this is an explicit indication that we are
434  * filtering a task rule at task creation time.  This and tsk == current are
435  * the only situations where tsk->cred may be accessed without an rcu read lock.
436  */
437 static int audit_filter_rules(struct task_struct *tsk,
438 			      struct audit_krule *rule,
439 			      struct audit_context *ctx,
440 			      struct audit_names *name,
441 			      enum audit_state *state,
442 			      bool task_creation)
443 {
444 	const struct cred *cred;
445 	int i, need_sid = 1;
446 	u32 sid;
447 
448 	cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
449 
450 	for (i = 0; i < rule->field_count; i++) {
451 		struct audit_field *f = &rule->fields[i];
452 		struct audit_names *n;
453 		int result = 0;
454 
455 		switch (f->type) {
456 		case AUDIT_PID:
457 			result = audit_comparator(tsk->pid, f->op, f->val);
458 			break;
459 		case AUDIT_PPID:
460 			if (ctx) {
461 				if (!ctx->ppid)
462 					ctx->ppid = sys_getppid();
463 				result = audit_comparator(ctx->ppid, f->op, f->val);
464 			}
465 			break;
466 		case AUDIT_UID:
467 			result = audit_uid_comparator(cred->uid, f->op, f->uid);
468 			break;
469 		case AUDIT_EUID:
470 			result = audit_uid_comparator(cred->euid, f->op, f->uid);
471 			break;
472 		case AUDIT_SUID:
473 			result = audit_uid_comparator(cred->suid, f->op, f->uid);
474 			break;
475 		case AUDIT_FSUID:
476 			result = audit_uid_comparator(cred->fsuid, f->op, f->uid);
477 			break;
478 		case AUDIT_GID:
479 			result = audit_gid_comparator(cred->gid, f->op, f->gid);
480 			if (f->op == Audit_equal) {
481 				if (!result)
482 					result = in_group_p(f->gid);
483 			} else if (f->op == Audit_not_equal) {
484 				if (result)
485 					result = !in_group_p(f->gid);
486 			}
487 			break;
488 		case AUDIT_EGID:
489 			result = audit_gid_comparator(cred->egid, f->op, f->gid);
490 			if (f->op == Audit_equal) {
491 				if (!result)
492 					result = in_egroup_p(f->gid);
493 			} else if (f->op == Audit_not_equal) {
494 				if (result)
495 					result = !in_egroup_p(f->gid);
496 			}
497 			break;
498 		case AUDIT_SGID:
499 			result = audit_gid_comparator(cred->sgid, f->op, f->gid);
500 			break;
501 		case AUDIT_FSGID:
502 			result = audit_gid_comparator(cred->fsgid, f->op, f->gid);
503 			break;
504 		case AUDIT_PERS:
505 			result = audit_comparator(tsk->personality, f->op, f->val);
506 			break;
507 		case AUDIT_ARCH:
508 			if (ctx)
509 				result = audit_comparator(ctx->arch, f->op, f->val);
510 			break;
511 
512 		case AUDIT_EXIT:
513 			if (ctx && ctx->return_valid)
514 				result = audit_comparator(ctx->return_code, f->op, f->val);
515 			break;
516 		case AUDIT_SUCCESS:
517 			if (ctx && ctx->return_valid) {
518 				if (f->val)
519 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
520 				else
521 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
522 			}
523 			break;
524 		case AUDIT_DEVMAJOR:
525 			if (name) {
526 				if (audit_comparator(MAJOR(name->dev), f->op, f->val) ||
527 				    audit_comparator(MAJOR(name->rdev), f->op, f->val))
528 					++result;
529 			} else if (ctx) {
530 				list_for_each_entry(n, &ctx->names_list, list) {
531 					if (audit_comparator(MAJOR(n->dev), f->op, f->val) ||
532 					    audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
533 						++result;
534 						break;
535 					}
536 				}
537 			}
538 			break;
539 		case AUDIT_DEVMINOR:
540 			if (name) {
541 				if (audit_comparator(MINOR(name->dev), f->op, f->val) ||
542 				    audit_comparator(MINOR(name->rdev), f->op, f->val))
543 					++result;
544 			} else if (ctx) {
545 				list_for_each_entry(n, &ctx->names_list, list) {
546 					if (audit_comparator(MINOR(n->dev), f->op, f->val) ||
547 					    audit_comparator(MINOR(n->rdev), f->op, f->val)) {
548 						++result;
549 						break;
550 					}
551 				}
552 			}
553 			break;
554 		case AUDIT_INODE:
555 			if (name)
556 				result = audit_comparator(name->ino, f->op, f->val);
557 			else if (ctx) {
558 				list_for_each_entry(n, &ctx->names_list, list) {
559 					if (audit_comparator(n->ino, f->op, f->val)) {
560 						++result;
561 						break;
562 					}
563 				}
564 			}
565 			break;
566 		case AUDIT_OBJ_UID:
567 			if (name) {
568 				result = audit_uid_comparator(name->uid, f->op, f->uid);
569 			} else if (ctx) {
570 				list_for_each_entry(n, &ctx->names_list, list) {
571 					if (audit_uid_comparator(n->uid, f->op, f->uid)) {
572 						++result;
573 						break;
574 					}
575 				}
576 			}
577 			break;
578 		case AUDIT_OBJ_GID:
579 			if (name) {
580 				result = audit_gid_comparator(name->gid, f->op, f->gid);
581 			} else if (ctx) {
582 				list_for_each_entry(n, &ctx->names_list, list) {
583 					if (audit_gid_comparator(n->gid, f->op, f->gid)) {
584 						++result;
585 						break;
586 					}
587 				}
588 			}
589 			break;
590 		case AUDIT_WATCH:
591 			if (name)
592 				result = audit_watch_compare(rule->watch, name->ino, name->dev);
593 			break;
594 		case AUDIT_DIR:
595 			if (ctx)
596 				result = match_tree_refs(ctx, rule->tree);
597 			break;
598 		case AUDIT_LOGINUID:
599 			result = 0;
600 			if (ctx)
601 				result = audit_uid_comparator(tsk->loginuid, f->op, f->uid);
602 			break;
603 		case AUDIT_LOGINUID_SET:
604 			result = audit_comparator(audit_loginuid_set(tsk), f->op, f->val);
605 			break;
606 		case AUDIT_SUBJ_USER:
607 		case AUDIT_SUBJ_ROLE:
608 		case AUDIT_SUBJ_TYPE:
609 		case AUDIT_SUBJ_SEN:
610 		case AUDIT_SUBJ_CLR:
611 			/* NOTE: this may return negative values indicating
612 			   a temporary error.  We simply treat this as a
613 			   match for now to avoid losing information that
614 			   may be wanted.   An error message will also be
615 			   logged upon error */
616 			if (f->lsm_rule) {
617 				if (need_sid) {
618 					security_task_getsecid(tsk, &sid);
619 					need_sid = 0;
620 				}
621 				result = security_audit_rule_match(sid, f->type,
622 				                                  f->op,
623 				                                  f->lsm_rule,
624 				                                  ctx);
625 			}
626 			break;
627 		case AUDIT_OBJ_USER:
628 		case AUDIT_OBJ_ROLE:
629 		case AUDIT_OBJ_TYPE:
630 		case AUDIT_OBJ_LEV_LOW:
631 		case AUDIT_OBJ_LEV_HIGH:
632 			/* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
633 			   also applies here */
634 			if (f->lsm_rule) {
635 				/* Find files that match */
636 				if (name) {
637 					result = security_audit_rule_match(
638 					           name->osid, f->type, f->op,
639 					           f->lsm_rule, ctx);
640 				} else if (ctx) {
641 					list_for_each_entry(n, &ctx->names_list, list) {
642 						if (security_audit_rule_match(n->osid, f->type,
643 									      f->op, f->lsm_rule,
644 									      ctx)) {
645 							++result;
646 							break;
647 						}
648 					}
649 				}
650 				/* Find ipc objects that match */
651 				if (!ctx || ctx->type != AUDIT_IPC)
652 					break;
653 				if (security_audit_rule_match(ctx->ipc.osid,
654 							      f->type, f->op,
655 							      f->lsm_rule, ctx))
656 					++result;
657 			}
658 			break;
659 		case AUDIT_ARG0:
660 		case AUDIT_ARG1:
661 		case AUDIT_ARG2:
662 		case AUDIT_ARG3:
663 			if (ctx)
664 				result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
665 			break;
666 		case AUDIT_FILTERKEY:
667 			/* ignore this field for filtering */
668 			result = 1;
669 			break;
670 		case AUDIT_PERM:
671 			result = audit_match_perm(ctx, f->val);
672 			break;
673 		case AUDIT_FILETYPE:
674 			result = audit_match_filetype(ctx, f->val);
675 			break;
676 		case AUDIT_FIELD_COMPARE:
677 			result = audit_field_compare(tsk, cred, f, ctx, name);
678 			break;
679 		}
680 		if (!result)
681 			return 0;
682 	}
683 
684 	if (ctx) {
685 		if (rule->prio <= ctx->prio)
686 			return 0;
687 		if (rule->filterkey) {
688 			kfree(ctx->filterkey);
689 			ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
690 		}
691 		ctx->prio = rule->prio;
692 	}
693 	switch (rule->action) {
694 	case AUDIT_NEVER:    *state = AUDIT_DISABLED;	    break;
695 	case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
696 	}
697 	return 1;
698 }
699 
700 /* At process creation time, we can determine if system-call auditing is
701  * completely disabled for this task.  Since we only have the task
702  * structure at this point, we can only check uid and gid.
703  */
704 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
705 {
706 	struct audit_entry *e;
707 	enum audit_state   state;
708 
709 	rcu_read_lock();
710 	list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
711 		if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
712 				       &state, true)) {
713 			if (state == AUDIT_RECORD_CONTEXT)
714 				*key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
715 			rcu_read_unlock();
716 			return state;
717 		}
718 	}
719 	rcu_read_unlock();
720 	return AUDIT_BUILD_CONTEXT;
721 }
722 
723 /* At syscall entry and exit time, this filter is called if the
724  * audit_state is not low enough that auditing cannot take place, but is
725  * also not high enough that we already know we have to write an audit
726  * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
727  */
728 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
729 					     struct audit_context *ctx,
730 					     struct list_head *list)
731 {
732 	struct audit_entry *e;
733 	enum audit_state state;
734 
735 	if (audit_pid && tsk->tgid == audit_pid)
736 		return AUDIT_DISABLED;
737 
738 	rcu_read_lock();
739 	if (!list_empty(list)) {
740 		int word = AUDIT_WORD(ctx->major);
741 		int bit  = AUDIT_BIT(ctx->major);
742 
743 		list_for_each_entry_rcu(e, list, list) {
744 			if ((e->rule.mask[word] & bit) == bit &&
745 			    audit_filter_rules(tsk, &e->rule, ctx, NULL,
746 					       &state, false)) {
747 				rcu_read_unlock();
748 				ctx->current_state = state;
749 				return state;
750 			}
751 		}
752 	}
753 	rcu_read_unlock();
754 	return AUDIT_BUILD_CONTEXT;
755 }
756 
757 /*
758  * Given an audit_name check the inode hash table to see if they match.
759  * Called holding the rcu read lock to protect the use of audit_inode_hash
760  */
761 static int audit_filter_inode_name(struct task_struct *tsk,
762 				   struct audit_names *n,
763 				   struct audit_context *ctx) {
764 	int word, bit;
765 	int h = audit_hash_ino((u32)n->ino);
766 	struct list_head *list = &audit_inode_hash[h];
767 	struct audit_entry *e;
768 	enum audit_state state;
769 
770 	word = AUDIT_WORD(ctx->major);
771 	bit  = AUDIT_BIT(ctx->major);
772 
773 	if (list_empty(list))
774 		return 0;
775 
776 	list_for_each_entry_rcu(e, list, list) {
777 		if ((e->rule.mask[word] & bit) == bit &&
778 		    audit_filter_rules(tsk, &e->rule, ctx, n, &state, false)) {
779 			ctx->current_state = state;
780 			return 1;
781 		}
782 	}
783 
784 	return 0;
785 }
786 
787 /* At syscall exit time, this filter is called if any audit_names have been
788  * collected during syscall processing.  We only check rules in sublists at hash
789  * buckets applicable to the inode numbers in audit_names.
790  * Regarding audit_state, same rules apply as for audit_filter_syscall().
791  */
792 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
793 {
794 	struct audit_names *n;
795 
796 	if (audit_pid && tsk->tgid == audit_pid)
797 		return;
798 
799 	rcu_read_lock();
800 
801 	list_for_each_entry(n, &ctx->names_list, list) {
802 		if (audit_filter_inode_name(tsk, n, ctx))
803 			break;
804 	}
805 	rcu_read_unlock();
806 }
807 
808 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
809 						      int return_valid,
810 						      long return_code)
811 {
812 	struct audit_context *context = tsk->audit_context;
813 
814 	if (!context)
815 		return NULL;
816 	context->return_valid = return_valid;
817 
818 	/*
819 	 * we need to fix up the return code in the audit logs if the actual
820 	 * return codes are later going to be fixed up by the arch specific
821 	 * signal handlers
822 	 *
823 	 * This is actually a test for:
824 	 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
825 	 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
826 	 *
827 	 * but is faster than a bunch of ||
828 	 */
829 	if (unlikely(return_code <= -ERESTARTSYS) &&
830 	    (return_code >= -ERESTART_RESTARTBLOCK) &&
831 	    (return_code != -ENOIOCTLCMD))
832 		context->return_code = -EINTR;
833 	else
834 		context->return_code  = return_code;
835 
836 	if (context->in_syscall && !context->dummy) {
837 		audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
838 		audit_filter_inodes(tsk, context);
839 	}
840 
841 	tsk->audit_context = NULL;
842 	return context;
843 }
844 
845 static inline void audit_free_names(struct audit_context *context)
846 {
847 	struct audit_names *n, *next;
848 
849 #if AUDIT_DEBUG == 2
850 	if (context->put_count + context->ino_count != context->name_count) {
851 		int i = 0;
852 
853 		printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
854 		       " name_count=%d put_count=%d"
855 		       " ino_count=%d [NOT freeing]\n",
856 		       __FILE__, __LINE__,
857 		       context->serial, context->major, context->in_syscall,
858 		       context->name_count, context->put_count,
859 		       context->ino_count);
860 		list_for_each_entry(n, &context->names_list, list) {
861 			printk(KERN_ERR "names[%d] = %p = %s\n", i++,
862 			       n->name, n->name->name ?: "(null)");
863 		}
864 		dump_stack();
865 		return;
866 	}
867 #endif
868 #if AUDIT_DEBUG
869 	context->put_count  = 0;
870 	context->ino_count  = 0;
871 #endif
872 
873 	list_for_each_entry_safe(n, next, &context->names_list, list) {
874 		list_del(&n->list);
875 		if (n->name && n->name_put)
876 			final_putname(n->name);
877 		if (n->should_free)
878 			kfree(n);
879 	}
880 	context->name_count = 0;
881 	path_put(&context->pwd);
882 	context->pwd.dentry = NULL;
883 	context->pwd.mnt = NULL;
884 }
885 
886 static inline void audit_free_aux(struct audit_context *context)
887 {
888 	struct audit_aux_data *aux;
889 
890 	while ((aux = context->aux)) {
891 		context->aux = aux->next;
892 		kfree(aux);
893 	}
894 	while ((aux = context->aux_pids)) {
895 		context->aux_pids = aux->next;
896 		kfree(aux);
897 	}
898 }
899 
900 static inline struct audit_context *audit_alloc_context(enum audit_state state)
901 {
902 	struct audit_context *context;
903 
904 	context = kzalloc(sizeof(*context), GFP_KERNEL);
905 	if (!context)
906 		return NULL;
907 	context->state = state;
908 	context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
909 	INIT_LIST_HEAD(&context->killed_trees);
910 	INIT_LIST_HEAD(&context->names_list);
911 	return context;
912 }
913 
914 /**
915  * audit_alloc - allocate an audit context block for a task
916  * @tsk: task
917  *
918  * Filter on the task information and allocate a per-task audit context
919  * if necessary.  Doing so turns on system call auditing for the
920  * specified task.  This is called from copy_process, so no lock is
921  * needed.
922  */
923 int audit_alloc(struct task_struct *tsk)
924 {
925 	struct audit_context *context;
926 	enum audit_state     state;
927 	char *key = NULL;
928 
929 	if (likely(!audit_ever_enabled))
930 		return 0; /* Return if not auditing. */
931 
932 	state = audit_filter_task(tsk, &key);
933 	if (state == AUDIT_DISABLED) {
934 		clear_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
935 		return 0;
936 	}
937 
938 	if (!(context = audit_alloc_context(state))) {
939 		kfree(key);
940 		audit_log_lost("out of memory in audit_alloc");
941 		return -ENOMEM;
942 	}
943 	context->filterkey = key;
944 
945 	tsk->audit_context  = context;
946 	set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
947 	return 0;
948 }
949 
950 static inline void audit_free_context(struct audit_context *context)
951 {
952 	audit_free_names(context);
953 	unroll_tree_refs(context, NULL, 0);
954 	free_tree_refs(context);
955 	audit_free_aux(context);
956 	kfree(context->filterkey);
957 	kfree(context->sockaddr);
958 	kfree(context);
959 }
960 
961 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
962 				 kuid_t auid, kuid_t uid, unsigned int sessionid,
963 				 u32 sid, char *comm)
964 {
965 	struct audit_buffer *ab;
966 	char *ctx = NULL;
967 	u32 len;
968 	int rc = 0;
969 
970 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
971 	if (!ab)
972 		return rc;
973 
974 	audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid,
975 			 from_kuid(&init_user_ns, auid),
976 			 from_kuid(&init_user_ns, uid), sessionid);
977 	if (sid) {
978 		if (security_secid_to_secctx(sid, &ctx, &len)) {
979 			audit_log_format(ab, " obj=(none)");
980 			rc = 1;
981 		} else {
982 			audit_log_format(ab, " obj=%s", ctx);
983 			security_release_secctx(ctx, len);
984 		}
985 	}
986 	audit_log_format(ab, " ocomm=");
987 	audit_log_untrustedstring(ab, comm);
988 	audit_log_end(ab);
989 
990 	return rc;
991 }
992 
993 /*
994  * to_send and len_sent accounting are very loose estimates.  We aren't
995  * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
996  * within about 500 bytes (next page boundary)
997  *
998  * why snprintf?  an int is up to 12 digits long.  if we just assumed when
999  * logging that a[%d]= was going to be 16 characters long we would be wasting
1000  * space in every audit message.  In one 7500 byte message we can log up to
1001  * about 1000 min size arguments.  That comes down to about 50% waste of space
1002  * if we didn't do the snprintf to find out how long arg_num_len was.
1003  */
1004 static int audit_log_single_execve_arg(struct audit_context *context,
1005 					struct audit_buffer **ab,
1006 					int arg_num,
1007 					size_t *len_sent,
1008 					const char __user *p,
1009 					char *buf)
1010 {
1011 	char arg_num_len_buf[12];
1012 	const char __user *tmp_p = p;
1013 	/* how many digits are in arg_num? 5 is the length of ' a=""' */
1014 	size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1015 	size_t len, len_left, to_send;
1016 	size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1017 	unsigned int i, has_cntl = 0, too_long = 0;
1018 	int ret;
1019 
1020 	/* strnlen_user includes the null we don't want to send */
1021 	len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1022 
1023 	/*
1024 	 * We just created this mm, if we can't find the strings
1025 	 * we just copied into it something is _very_ wrong. Similar
1026 	 * for strings that are too long, we should not have created
1027 	 * any.
1028 	 */
1029 	if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1030 		WARN_ON(1);
1031 		send_sig(SIGKILL, current, 0);
1032 		return -1;
1033 	}
1034 
1035 	/* walk the whole argument looking for non-ascii chars */
1036 	do {
1037 		if (len_left > MAX_EXECVE_AUDIT_LEN)
1038 			to_send = MAX_EXECVE_AUDIT_LEN;
1039 		else
1040 			to_send = len_left;
1041 		ret = copy_from_user(buf, tmp_p, to_send);
1042 		/*
1043 		 * There is no reason for this copy to be short. We just
1044 		 * copied them here, and the mm hasn't been exposed to user-
1045 		 * space yet.
1046 		 */
1047 		if (ret) {
1048 			WARN_ON(1);
1049 			send_sig(SIGKILL, current, 0);
1050 			return -1;
1051 		}
1052 		buf[to_send] = '\0';
1053 		has_cntl = audit_string_contains_control(buf, to_send);
1054 		if (has_cntl) {
1055 			/*
1056 			 * hex messages get logged as 2 bytes, so we can only
1057 			 * send half as much in each message
1058 			 */
1059 			max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1060 			break;
1061 		}
1062 		len_left -= to_send;
1063 		tmp_p += to_send;
1064 	} while (len_left > 0);
1065 
1066 	len_left = len;
1067 
1068 	if (len > max_execve_audit_len)
1069 		too_long = 1;
1070 
1071 	/* rewalk the argument actually logging the message */
1072 	for (i = 0; len_left > 0; i++) {
1073 		int room_left;
1074 
1075 		if (len_left > max_execve_audit_len)
1076 			to_send = max_execve_audit_len;
1077 		else
1078 			to_send = len_left;
1079 
1080 		/* do we have space left to send this argument in this ab? */
1081 		room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1082 		if (has_cntl)
1083 			room_left -= (to_send * 2);
1084 		else
1085 			room_left -= to_send;
1086 		if (room_left < 0) {
1087 			*len_sent = 0;
1088 			audit_log_end(*ab);
1089 			*ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1090 			if (!*ab)
1091 				return 0;
1092 		}
1093 
1094 		/*
1095 		 * first record needs to say how long the original string was
1096 		 * so we can be sure nothing was lost.
1097 		 */
1098 		if ((i == 0) && (too_long))
1099 			audit_log_format(*ab, " a%d_len=%zu", arg_num,
1100 					 has_cntl ? 2*len : len);
1101 
1102 		/*
1103 		 * normally arguments are small enough to fit and we already
1104 		 * filled buf above when we checked for control characters
1105 		 * so don't bother with another copy_from_user
1106 		 */
1107 		if (len >= max_execve_audit_len)
1108 			ret = copy_from_user(buf, p, to_send);
1109 		else
1110 			ret = 0;
1111 		if (ret) {
1112 			WARN_ON(1);
1113 			send_sig(SIGKILL, current, 0);
1114 			return -1;
1115 		}
1116 		buf[to_send] = '\0';
1117 
1118 		/* actually log it */
1119 		audit_log_format(*ab, " a%d", arg_num);
1120 		if (too_long)
1121 			audit_log_format(*ab, "[%d]", i);
1122 		audit_log_format(*ab, "=");
1123 		if (has_cntl)
1124 			audit_log_n_hex(*ab, buf, to_send);
1125 		else
1126 			audit_log_string(*ab, buf);
1127 
1128 		p += to_send;
1129 		len_left -= to_send;
1130 		*len_sent += arg_num_len;
1131 		if (has_cntl)
1132 			*len_sent += to_send * 2;
1133 		else
1134 			*len_sent += to_send;
1135 	}
1136 	/* include the null we didn't log */
1137 	return len + 1;
1138 }
1139 
1140 static void audit_log_execve_info(struct audit_context *context,
1141 				  struct audit_buffer **ab)
1142 {
1143 	int i, len;
1144 	size_t len_sent = 0;
1145 	const char __user *p;
1146 	char *buf;
1147 
1148 	p = (const char __user *)current->mm->arg_start;
1149 
1150 	audit_log_format(*ab, "argc=%d", context->execve.argc);
1151 
1152 	/*
1153 	 * we need some kernel buffer to hold the userspace args.  Just
1154 	 * allocate one big one rather than allocating one of the right size
1155 	 * for every single argument inside audit_log_single_execve_arg()
1156 	 * should be <8k allocation so should be pretty safe.
1157 	 */
1158 	buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1159 	if (!buf) {
1160 		audit_panic("out of memory for argv string\n");
1161 		return;
1162 	}
1163 
1164 	for (i = 0; i < context->execve.argc; i++) {
1165 		len = audit_log_single_execve_arg(context, ab, i,
1166 						  &len_sent, p, buf);
1167 		if (len <= 0)
1168 			break;
1169 		p += len;
1170 	}
1171 	kfree(buf);
1172 }
1173 
1174 static void show_special(struct audit_context *context, int *call_panic)
1175 {
1176 	struct audit_buffer *ab;
1177 	int i;
1178 
1179 	ab = audit_log_start(context, GFP_KERNEL, context->type);
1180 	if (!ab)
1181 		return;
1182 
1183 	switch (context->type) {
1184 	case AUDIT_SOCKETCALL: {
1185 		int nargs = context->socketcall.nargs;
1186 		audit_log_format(ab, "nargs=%d", nargs);
1187 		for (i = 0; i < nargs; i++)
1188 			audit_log_format(ab, " a%d=%lx", i,
1189 				context->socketcall.args[i]);
1190 		break; }
1191 	case AUDIT_IPC: {
1192 		u32 osid = context->ipc.osid;
1193 
1194 		audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1195 				 from_kuid(&init_user_ns, context->ipc.uid),
1196 				 from_kgid(&init_user_ns, context->ipc.gid),
1197 				 context->ipc.mode);
1198 		if (osid) {
1199 			char *ctx = NULL;
1200 			u32 len;
1201 			if (security_secid_to_secctx(osid, &ctx, &len)) {
1202 				audit_log_format(ab, " osid=%u", osid);
1203 				*call_panic = 1;
1204 			} else {
1205 				audit_log_format(ab, " obj=%s", ctx);
1206 				security_release_secctx(ctx, len);
1207 			}
1208 		}
1209 		if (context->ipc.has_perm) {
1210 			audit_log_end(ab);
1211 			ab = audit_log_start(context, GFP_KERNEL,
1212 					     AUDIT_IPC_SET_PERM);
1213 			if (unlikely(!ab))
1214 				return;
1215 			audit_log_format(ab,
1216 				"qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1217 				context->ipc.qbytes,
1218 				context->ipc.perm_uid,
1219 				context->ipc.perm_gid,
1220 				context->ipc.perm_mode);
1221 		}
1222 		break; }
1223 	case AUDIT_MQ_OPEN: {
1224 		audit_log_format(ab,
1225 			"oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1226 			"mq_msgsize=%ld mq_curmsgs=%ld",
1227 			context->mq_open.oflag, context->mq_open.mode,
1228 			context->mq_open.attr.mq_flags,
1229 			context->mq_open.attr.mq_maxmsg,
1230 			context->mq_open.attr.mq_msgsize,
1231 			context->mq_open.attr.mq_curmsgs);
1232 		break; }
1233 	case AUDIT_MQ_SENDRECV: {
1234 		audit_log_format(ab,
1235 			"mqdes=%d msg_len=%zd msg_prio=%u "
1236 			"abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1237 			context->mq_sendrecv.mqdes,
1238 			context->mq_sendrecv.msg_len,
1239 			context->mq_sendrecv.msg_prio,
1240 			context->mq_sendrecv.abs_timeout.tv_sec,
1241 			context->mq_sendrecv.abs_timeout.tv_nsec);
1242 		break; }
1243 	case AUDIT_MQ_NOTIFY: {
1244 		audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1245 				context->mq_notify.mqdes,
1246 				context->mq_notify.sigev_signo);
1247 		break; }
1248 	case AUDIT_MQ_GETSETATTR: {
1249 		struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1250 		audit_log_format(ab,
1251 			"mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1252 			"mq_curmsgs=%ld ",
1253 			context->mq_getsetattr.mqdes,
1254 			attr->mq_flags, attr->mq_maxmsg,
1255 			attr->mq_msgsize, attr->mq_curmsgs);
1256 		break; }
1257 	case AUDIT_CAPSET: {
1258 		audit_log_format(ab, "pid=%d", context->capset.pid);
1259 		audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1260 		audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1261 		audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1262 		break; }
1263 	case AUDIT_MMAP: {
1264 		audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1265 				 context->mmap.flags);
1266 		break; }
1267 	case AUDIT_EXECVE: {
1268 		audit_log_execve_info(context, &ab);
1269 		break; }
1270 	}
1271 	audit_log_end(ab);
1272 }
1273 
1274 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1275 {
1276 	int i, call_panic = 0;
1277 	struct audit_buffer *ab;
1278 	struct audit_aux_data *aux;
1279 	struct audit_names *n;
1280 
1281 	/* tsk == current */
1282 	context->personality = tsk->personality;
1283 
1284 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1285 	if (!ab)
1286 		return;		/* audit_panic has been called */
1287 	audit_log_format(ab, "arch=%x syscall=%d",
1288 			 context->arch, context->major);
1289 	if (context->personality != PER_LINUX)
1290 		audit_log_format(ab, " per=%lx", context->personality);
1291 	if (context->return_valid)
1292 		audit_log_format(ab, " success=%s exit=%ld",
1293 				 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1294 				 context->return_code);
1295 
1296 	audit_log_format(ab,
1297 			 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d",
1298 			 context->argv[0],
1299 			 context->argv[1],
1300 			 context->argv[2],
1301 			 context->argv[3],
1302 			 context->name_count);
1303 
1304 	audit_log_task_info(ab, tsk);
1305 	audit_log_key(ab, context->filterkey);
1306 	audit_log_end(ab);
1307 
1308 	for (aux = context->aux; aux; aux = aux->next) {
1309 
1310 		ab = audit_log_start(context, GFP_KERNEL, aux->type);
1311 		if (!ab)
1312 			continue; /* audit_panic has been called */
1313 
1314 		switch (aux->type) {
1315 
1316 		case AUDIT_BPRM_FCAPS: {
1317 			struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1318 			audit_log_format(ab, "fver=%x", axs->fcap_ver);
1319 			audit_log_cap(ab, "fp", &axs->fcap.permitted);
1320 			audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1321 			audit_log_format(ab, " fe=%d", axs->fcap.fE);
1322 			audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1323 			audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1324 			audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1325 			audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1326 			audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1327 			audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1328 			break; }
1329 
1330 		}
1331 		audit_log_end(ab);
1332 	}
1333 
1334 	if (context->type)
1335 		show_special(context, &call_panic);
1336 
1337 	if (context->fds[0] >= 0) {
1338 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1339 		if (ab) {
1340 			audit_log_format(ab, "fd0=%d fd1=%d",
1341 					context->fds[0], context->fds[1]);
1342 			audit_log_end(ab);
1343 		}
1344 	}
1345 
1346 	if (context->sockaddr_len) {
1347 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1348 		if (ab) {
1349 			audit_log_format(ab, "saddr=");
1350 			audit_log_n_hex(ab, (void *)context->sockaddr,
1351 					context->sockaddr_len);
1352 			audit_log_end(ab);
1353 		}
1354 	}
1355 
1356 	for (aux = context->aux_pids; aux; aux = aux->next) {
1357 		struct audit_aux_data_pids *axs = (void *)aux;
1358 
1359 		for (i = 0; i < axs->pid_count; i++)
1360 			if (audit_log_pid_context(context, axs->target_pid[i],
1361 						  axs->target_auid[i],
1362 						  axs->target_uid[i],
1363 						  axs->target_sessionid[i],
1364 						  axs->target_sid[i],
1365 						  axs->target_comm[i]))
1366 				call_panic = 1;
1367 	}
1368 
1369 	if (context->target_pid &&
1370 	    audit_log_pid_context(context, context->target_pid,
1371 				  context->target_auid, context->target_uid,
1372 				  context->target_sessionid,
1373 				  context->target_sid, context->target_comm))
1374 			call_panic = 1;
1375 
1376 	if (context->pwd.dentry && context->pwd.mnt) {
1377 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1378 		if (ab) {
1379 			audit_log_d_path(ab, " cwd=", &context->pwd);
1380 			audit_log_end(ab);
1381 		}
1382 	}
1383 
1384 	i = 0;
1385 	list_for_each_entry(n, &context->names_list, list) {
1386 		if (n->hidden)
1387 			continue;
1388 		audit_log_name(context, n, NULL, i++, &call_panic);
1389 	}
1390 
1391 	/* Send end of event record to help user space know we are finished */
1392 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1393 	if (ab)
1394 		audit_log_end(ab);
1395 	if (call_panic)
1396 		audit_panic("error converting sid to string");
1397 }
1398 
1399 /**
1400  * audit_free - free a per-task audit context
1401  * @tsk: task whose audit context block to free
1402  *
1403  * Called from copy_process and do_exit
1404  */
1405 void __audit_free(struct task_struct *tsk)
1406 {
1407 	struct audit_context *context;
1408 
1409 	context = audit_get_context(tsk, 0, 0);
1410 	if (!context)
1411 		return;
1412 
1413 	/* Check for system calls that do not go through the exit
1414 	 * function (e.g., exit_group), then free context block.
1415 	 * We use GFP_ATOMIC here because we might be doing this
1416 	 * in the context of the idle thread */
1417 	/* that can happen only if we are called from do_exit() */
1418 	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1419 		audit_log_exit(context, tsk);
1420 	if (!list_empty(&context->killed_trees))
1421 		audit_kill_trees(&context->killed_trees);
1422 
1423 	audit_free_context(context);
1424 }
1425 
1426 /**
1427  * audit_syscall_entry - fill in an audit record at syscall entry
1428  * @arch: architecture type
1429  * @major: major syscall type (function)
1430  * @a1: additional syscall register 1
1431  * @a2: additional syscall register 2
1432  * @a3: additional syscall register 3
1433  * @a4: additional syscall register 4
1434  *
1435  * Fill in audit context at syscall entry.  This only happens if the
1436  * audit context was created when the task was created and the state or
1437  * filters demand the audit context be built.  If the state from the
1438  * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1439  * then the record will be written at syscall exit time (otherwise, it
1440  * will only be written if another part of the kernel requests that it
1441  * be written).
1442  */
1443 void __audit_syscall_entry(int arch, int major,
1444 			 unsigned long a1, unsigned long a2,
1445 			 unsigned long a3, unsigned long a4)
1446 {
1447 	struct task_struct *tsk = current;
1448 	struct audit_context *context = tsk->audit_context;
1449 	enum audit_state     state;
1450 
1451 	if (!context)
1452 		return;
1453 
1454 	BUG_ON(context->in_syscall || context->name_count);
1455 
1456 	if (!audit_enabled)
1457 		return;
1458 
1459 	context->arch	    = arch;
1460 	context->major      = major;
1461 	context->argv[0]    = a1;
1462 	context->argv[1]    = a2;
1463 	context->argv[2]    = a3;
1464 	context->argv[3]    = a4;
1465 
1466 	state = context->state;
1467 	context->dummy = !audit_n_rules;
1468 	if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1469 		context->prio = 0;
1470 		state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1471 	}
1472 	if (state == AUDIT_DISABLED)
1473 		return;
1474 
1475 	context->serial     = 0;
1476 	context->ctime      = CURRENT_TIME;
1477 	context->in_syscall = 1;
1478 	context->current_state  = state;
1479 	context->ppid       = 0;
1480 }
1481 
1482 /**
1483  * audit_syscall_exit - deallocate audit context after a system call
1484  * @success: success value of the syscall
1485  * @return_code: return value of the syscall
1486  *
1487  * Tear down after system call.  If the audit context has been marked as
1488  * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1489  * filtering, or because some other part of the kernel wrote an audit
1490  * message), then write out the syscall information.  In call cases,
1491  * free the names stored from getname().
1492  */
1493 void __audit_syscall_exit(int success, long return_code)
1494 {
1495 	struct task_struct *tsk = current;
1496 	struct audit_context *context;
1497 
1498 	if (success)
1499 		success = AUDITSC_SUCCESS;
1500 	else
1501 		success = AUDITSC_FAILURE;
1502 
1503 	context = audit_get_context(tsk, success, return_code);
1504 	if (!context)
1505 		return;
1506 
1507 	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1508 		audit_log_exit(context, tsk);
1509 
1510 	context->in_syscall = 0;
1511 	context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1512 
1513 	if (!list_empty(&context->killed_trees))
1514 		audit_kill_trees(&context->killed_trees);
1515 
1516 	audit_free_names(context);
1517 	unroll_tree_refs(context, NULL, 0);
1518 	audit_free_aux(context);
1519 	context->aux = NULL;
1520 	context->aux_pids = NULL;
1521 	context->target_pid = 0;
1522 	context->target_sid = 0;
1523 	context->sockaddr_len = 0;
1524 	context->type = 0;
1525 	context->fds[0] = -1;
1526 	if (context->state != AUDIT_RECORD_CONTEXT) {
1527 		kfree(context->filterkey);
1528 		context->filterkey = NULL;
1529 	}
1530 	tsk->audit_context = context;
1531 }
1532 
1533 static inline void handle_one(const struct inode *inode)
1534 {
1535 #ifdef CONFIG_AUDIT_TREE
1536 	struct audit_context *context;
1537 	struct audit_tree_refs *p;
1538 	struct audit_chunk *chunk;
1539 	int count;
1540 	if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1541 		return;
1542 	context = current->audit_context;
1543 	p = context->trees;
1544 	count = context->tree_count;
1545 	rcu_read_lock();
1546 	chunk = audit_tree_lookup(inode);
1547 	rcu_read_unlock();
1548 	if (!chunk)
1549 		return;
1550 	if (likely(put_tree_ref(context, chunk)))
1551 		return;
1552 	if (unlikely(!grow_tree_refs(context))) {
1553 		printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1554 		audit_set_auditable(context);
1555 		audit_put_chunk(chunk);
1556 		unroll_tree_refs(context, p, count);
1557 		return;
1558 	}
1559 	put_tree_ref(context, chunk);
1560 #endif
1561 }
1562 
1563 static void handle_path(const struct dentry *dentry)
1564 {
1565 #ifdef CONFIG_AUDIT_TREE
1566 	struct audit_context *context;
1567 	struct audit_tree_refs *p;
1568 	const struct dentry *d, *parent;
1569 	struct audit_chunk *drop;
1570 	unsigned long seq;
1571 	int count;
1572 
1573 	context = current->audit_context;
1574 	p = context->trees;
1575 	count = context->tree_count;
1576 retry:
1577 	drop = NULL;
1578 	d = dentry;
1579 	rcu_read_lock();
1580 	seq = read_seqbegin(&rename_lock);
1581 	for(;;) {
1582 		struct inode *inode = d->d_inode;
1583 		if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1584 			struct audit_chunk *chunk;
1585 			chunk = audit_tree_lookup(inode);
1586 			if (chunk) {
1587 				if (unlikely(!put_tree_ref(context, chunk))) {
1588 					drop = chunk;
1589 					break;
1590 				}
1591 			}
1592 		}
1593 		parent = d->d_parent;
1594 		if (parent == d)
1595 			break;
1596 		d = parent;
1597 	}
1598 	if (unlikely(read_seqretry(&rename_lock, seq) || drop)) {  /* in this order */
1599 		rcu_read_unlock();
1600 		if (!drop) {
1601 			/* just a race with rename */
1602 			unroll_tree_refs(context, p, count);
1603 			goto retry;
1604 		}
1605 		audit_put_chunk(drop);
1606 		if (grow_tree_refs(context)) {
1607 			/* OK, got more space */
1608 			unroll_tree_refs(context, p, count);
1609 			goto retry;
1610 		}
1611 		/* too bad */
1612 		printk(KERN_WARNING
1613 			"out of memory, audit has lost a tree reference\n");
1614 		unroll_tree_refs(context, p, count);
1615 		audit_set_auditable(context);
1616 		return;
1617 	}
1618 	rcu_read_unlock();
1619 #endif
1620 }
1621 
1622 static struct audit_names *audit_alloc_name(struct audit_context *context,
1623 						unsigned char type)
1624 {
1625 	struct audit_names *aname;
1626 
1627 	if (context->name_count < AUDIT_NAMES) {
1628 		aname = &context->preallocated_names[context->name_count];
1629 		memset(aname, 0, sizeof(*aname));
1630 	} else {
1631 		aname = kzalloc(sizeof(*aname), GFP_NOFS);
1632 		if (!aname)
1633 			return NULL;
1634 		aname->should_free = true;
1635 	}
1636 
1637 	aname->ino = (unsigned long)-1;
1638 	aname->type = type;
1639 	list_add_tail(&aname->list, &context->names_list);
1640 
1641 	context->name_count++;
1642 #if AUDIT_DEBUG
1643 	context->ino_count++;
1644 #endif
1645 	return aname;
1646 }
1647 
1648 /**
1649  * audit_reusename - fill out filename with info from existing entry
1650  * @uptr: userland ptr to pathname
1651  *
1652  * Search the audit_names list for the current audit context. If there is an
1653  * existing entry with a matching "uptr" then return the filename
1654  * associated with that audit_name. If not, return NULL.
1655  */
1656 struct filename *
1657 __audit_reusename(const __user char *uptr)
1658 {
1659 	struct audit_context *context = current->audit_context;
1660 	struct audit_names *n;
1661 
1662 	list_for_each_entry(n, &context->names_list, list) {
1663 		if (!n->name)
1664 			continue;
1665 		if (n->name->uptr == uptr)
1666 			return n->name;
1667 	}
1668 	return NULL;
1669 }
1670 
1671 /**
1672  * audit_getname - add a name to the list
1673  * @name: name to add
1674  *
1675  * Add a name to the list of audit names for this context.
1676  * Called from fs/namei.c:getname().
1677  */
1678 void __audit_getname(struct filename *name)
1679 {
1680 	struct audit_context *context = current->audit_context;
1681 	struct audit_names *n;
1682 
1683 	if (!context->in_syscall) {
1684 #if AUDIT_DEBUG == 2
1685 		printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1686 		       __FILE__, __LINE__, context->serial, name);
1687 		dump_stack();
1688 #endif
1689 		return;
1690 	}
1691 
1692 #if AUDIT_DEBUG
1693 	/* The filename _must_ have a populated ->name */
1694 	BUG_ON(!name->name);
1695 #endif
1696 
1697 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
1698 	if (!n)
1699 		return;
1700 
1701 	n->name = name;
1702 	n->name_len = AUDIT_NAME_FULL;
1703 	n->name_put = true;
1704 	name->aname = n;
1705 
1706 	if (!context->pwd.dentry)
1707 		get_fs_pwd(current->fs, &context->pwd);
1708 }
1709 
1710 /* audit_putname - intercept a putname request
1711  * @name: name to intercept and delay for putname
1712  *
1713  * If we have stored the name from getname in the audit context,
1714  * then we delay the putname until syscall exit.
1715  * Called from include/linux/fs.h:putname().
1716  */
1717 void audit_putname(struct filename *name)
1718 {
1719 	struct audit_context *context = current->audit_context;
1720 
1721 	BUG_ON(!context);
1722 	if (!context->in_syscall) {
1723 #if AUDIT_DEBUG == 2
1724 		printk(KERN_ERR "%s:%d(:%d): final_putname(%p)\n",
1725 		       __FILE__, __LINE__, context->serial, name);
1726 		if (context->name_count) {
1727 			struct audit_names *n;
1728 			int i = 0;
1729 
1730 			list_for_each_entry(n, &context->names_list, list)
1731 				printk(KERN_ERR "name[%d] = %p = %s\n", i++,
1732 				       n->name, n->name->name ?: "(null)");
1733 			}
1734 #endif
1735 		final_putname(name);
1736 	}
1737 #if AUDIT_DEBUG
1738 	else {
1739 		++context->put_count;
1740 		if (context->put_count > context->name_count) {
1741 			printk(KERN_ERR "%s:%d(:%d): major=%d"
1742 			       " in_syscall=%d putname(%p) name_count=%d"
1743 			       " put_count=%d\n",
1744 			       __FILE__, __LINE__,
1745 			       context->serial, context->major,
1746 			       context->in_syscall, name->name,
1747 			       context->name_count, context->put_count);
1748 			dump_stack();
1749 		}
1750 	}
1751 #endif
1752 }
1753 
1754 /**
1755  * __audit_inode - store the inode and device from a lookup
1756  * @name: name being audited
1757  * @dentry: dentry being audited
1758  * @flags: attributes for this particular entry
1759  */
1760 void __audit_inode(struct filename *name, const struct dentry *dentry,
1761 		   unsigned int flags)
1762 {
1763 	struct audit_context *context = current->audit_context;
1764 	const struct inode *inode = dentry->d_inode;
1765 	struct audit_names *n;
1766 	bool parent = flags & AUDIT_INODE_PARENT;
1767 
1768 	if (!context->in_syscall)
1769 		return;
1770 
1771 	if (!name)
1772 		goto out_alloc;
1773 
1774 #if AUDIT_DEBUG
1775 	/* The struct filename _must_ have a populated ->name */
1776 	BUG_ON(!name->name);
1777 #endif
1778 	/*
1779 	 * If we have a pointer to an audit_names entry already, then we can
1780 	 * just use it directly if the type is correct.
1781 	 */
1782 	n = name->aname;
1783 	if (n) {
1784 		if (parent) {
1785 			if (n->type == AUDIT_TYPE_PARENT ||
1786 			    n->type == AUDIT_TYPE_UNKNOWN)
1787 				goto out;
1788 		} else {
1789 			if (n->type != AUDIT_TYPE_PARENT)
1790 				goto out;
1791 		}
1792 	}
1793 
1794 	list_for_each_entry_reverse(n, &context->names_list, list) {
1795 		/* does the name pointer match? */
1796 		if (!n->name || n->name->name != name->name)
1797 			continue;
1798 
1799 		/* match the correct record type */
1800 		if (parent) {
1801 			if (n->type == AUDIT_TYPE_PARENT ||
1802 			    n->type == AUDIT_TYPE_UNKNOWN)
1803 				goto out;
1804 		} else {
1805 			if (n->type != AUDIT_TYPE_PARENT)
1806 				goto out;
1807 		}
1808 	}
1809 
1810 out_alloc:
1811 	/* unable to find the name from a previous getname(). Allocate a new
1812 	 * anonymous entry.
1813 	 */
1814 	n = audit_alloc_name(context, AUDIT_TYPE_NORMAL);
1815 	if (!n)
1816 		return;
1817 out:
1818 	if (parent) {
1819 		n->name_len = n->name ? parent_len(n->name->name) : AUDIT_NAME_FULL;
1820 		n->type = AUDIT_TYPE_PARENT;
1821 		if (flags & AUDIT_INODE_HIDDEN)
1822 			n->hidden = true;
1823 	} else {
1824 		n->name_len = AUDIT_NAME_FULL;
1825 		n->type = AUDIT_TYPE_NORMAL;
1826 	}
1827 	handle_path(dentry);
1828 	audit_copy_inode(n, dentry, inode);
1829 }
1830 
1831 /**
1832  * __audit_inode_child - collect inode info for created/removed objects
1833  * @parent: inode of dentry parent
1834  * @dentry: dentry being audited
1835  * @type:   AUDIT_TYPE_* value that we're looking for
1836  *
1837  * For syscalls that create or remove filesystem objects, audit_inode
1838  * can only collect information for the filesystem object's parent.
1839  * This call updates the audit context with the child's information.
1840  * Syscalls that create a new filesystem object must be hooked after
1841  * the object is created.  Syscalls that remove a filesystem object
1842  * must be hooked prior, in order to capture the target inode during
1843  * unsuccessful attempts.
1844  */
1845 void __audit_inode_child(const struct inode *parent,
1846 			 const struct dentry *dentry,
1847 			 const unsigned char type)
1848 {
1849 	struct audit_context *context = current->audit_context;
1850 	const struct inode *inode = dentry->d_inode;
1851 	const char *dname = dentry->d_name.name;
1852 	struct audit_names *n, *found_parent = NULL, *found_child = NULL;
1853 
1854 	if (!context->in_syscall)
1855 		return;
1856 
1857 	if (inode)
1858 		handle_one(inode);
1859 
1860 	/* look for a parent entry first */
1861 	list_for_each_entry(n, &context->names_list, list) {
1862 		if (!n->name || n->type != AUDIT_TYPE_PARENT)
1863 			continue;
1864 
1865 		if (n->ino == parent->i_ino &&
1866 		    !audit_compare_dname_path(dname, n->name->name, n->name_len)) {
1867 			found_parent = n;
1868 			break;
1869 		}
1870 	}
1871 
1872 	/* is there a matching child entry? */
1873 	list_for_each_entry(n, &context->names_list, list) {
1874 		/* can only match entries that have a name */
1875 		if (!n->name || n->type != type)
1876 			continue;
1877 
1878 		/* if we found a parent, make sure this one is a child of it */
1879 		if (found_parent && (n->name != found_parent->name))
1880 			continue;
1881 
1882 		if (!strcmp(dname, n->name->name) ||
1883 		    !audit_compare_dname_path(dname, n->name->name,
1884 						found_parent ?
1885 						found_parent->name_len :
1886 						AUDIT_NAME_FULL)) {
1887 			found_child = n;
1888 			break;
1889 		}
1890 	}
1891 
1892 	if (!found_parent) {
1893 		/* create a new, "anonymous" parent record */
1894 		n = audit_alloc_name(context, AUDIT_TYPE_PARENT);
1895 		if (!n)
1896 			return;
1897 		audit_copy_inode(n, NULL, parent);
1898 	}
1899 
1900 	if (!found_child) {
1901 		found_child = audit_alloc_name(context, type);
1902 		if (!found_child)
1903 			return;
1904 
1905 		/* Re-use the name belonging to the slot for a matching parent
1906 		 * directory. All names for this context are relinquished in
1907 		 * audit_free_names() */
1908 		if (found_parent) {
1909 			found_child->name = found_parent->name;
1910 			found_child->name_len = AUDIT_NAME_FULL;
1911 			/* don't call __putname() */
1912 			found_child->name_put = false;
1913 		}
1914 	}
1915 	if (inode)
1916 		audit_copy_inode(found_child, dentry, inode);
1917 	else
1918 		found_child->ino = (unsigned long)-1;
1919 }
1920 EXPORT_SYMBOL_GPL(__audit_inode_child);
1921 
1922 /**
1923  * auditsc_get_stamp - get local copies of audit_context values
1924  * @ctx: audit_context for the task
1925  * @t: timespec to store time recorded in the audit_context
1926  * @serial: serial value that is recorded in the audit_context
1927  *
1928  * Also sets the context as auditable.
1929  */
1930 int auditsc_get_stamp(struct audit_context *ctx,
1931 		       struct timespec *t, unsigned int *serial)
1932 {
1933 	if (!ctx->in_syscall)
1934 		return 0;
1935 	if (!ctx->serial)
1936 		ctx->serial = audit_serial();
1937 	t->tv_sec  = ctx->ctime.tv_sec;
1938 	t->tv_nsec = ctx->ctime.tv_nsec;
1939 	*serial    = ctx->serial;
1940 	if (!ctx->prio) {
1941 		ctx->prio = 1;
1942 		ctx->current_state = AUDIT_RECORD_CONTEXT;
1943 	}
1944 	return 1;
1945 }
1946 
1947 /* global counter which is incremented every time something logs in */
1948 static atomic_t session_id = ATOMIC_INIT(0);
1949 
1950 static int audit_set_loginuid_perm(kuid_t loginuid)
1951 {
1952 	/* if we are unset, we don't need privs */
1953 	if (!audit_loginuid_set(current))
1954 		return 0;
1955 	/* if AUDIT_FEATURE_LOGINUID_IMMUTABLE means never ever allow a change*/
1956 	if (is_audit_feature_set(AUDIT_FEATURE_LOGINUID_IMMUTABLE))
1957 		return -EPERM;
1958 	/* it is set, you need permission */
1959 	if (!capable(CAP_AUDIT_CONTROL))
1960 		return -EPERM;
1961 	/* reject if this is not an unset and we don't allow that */
1962 	if (is_audit_feature_set(AUDIT_FEATURE_ONLY_UNSET_LOGINUID) && uid_valid(loginuid))
1963 		return -EPERM;
1964 	return 0;
1965 }
1966 
1967 static void audit_log_set_loginuid(kuid_t koldloginuid, kuid_t kloginuid,
1968 				   unsigned int oldsessionid, unsigned int sessionid,
1969 				   int rc)
1970 {
1971 	struct audit_buffer *ab;
1972 	uid_t uid, ologinuid, nloginuid;
1973 
1974 	uid = from_kuid(&init_user_ns, task_uid(current));
1975 	ologinuid = from_kuid(&init_user_ns, koldloginuid);
1976 	nloginuid = from_kuid(&init_user_ns, kloginuid),
1977 
1978 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1979 	if (!ab)
1980 		return;
1981 	audit_log_format(ab, "pid=%d uid=%u old auid=%u new auid=%u old "
1982 			 "ses=%u new ses=%u res=%d", current->pid, uid, ologinuid,
1983 			 nloginuid, oldsessionid, sessionid, !rc);
1984 	audit_log_end(ab);
1985 }
1986 
1987 /**
1988  * audit_set_loginuid - set current task's audit_context loginuid
1989  * @loginuid: loginuid value
1990  *
1991  * Returns 0.
1992  *
1993  * Called (set) from fs/proc/base.c::proc_loginuid_write().
1994  */
1995 int audit_set_loginuid(kuid_t loginuid)
1996 {
1997 	struct task_struct *task = current;
1998 	unsigned int oldsessionid, sessionid = (unsigned int)-1;
1999 	kuid_t oldloginuid;
2000 	int rc;
2001 
2002 	oldloginuid = audit_get_loginuid(current);
2003 	oldsessionid = audit_get_sessionid(current);
2004 
2005 	rc = audit_set_loginuid_perm(loginuid);
2006 	if (rc)
2007 		goto out;
2008 
2009 	/* are we setting or clearing? */
2010 	if (uid_valid(loginuid))
2011 		sessionid = atomic_inc_return(&session_id);
2012 
2013 	task->sessionid = sessionid;
2014 	task->loginuid = loginuid;
2015 out:
2016 	audit_log_set_loginuid(oldloginuid, loginuid, oldsessionid, sessionid, rc);
2017 	return rc;
2018 }
2019 
2020 /**
2021  * __audit_mq_open - record audit data for a POSIX MQ open
2022  * @oflag: open flag
2023  * @mode: mode bits
2024  * @attr: queue attributes
2025  *
2026  */
2027 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2028 {
2029 	struct audit_context *context = current->audit_context;
2030 
2031 	if (attr)
2032 		memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2033 	else
2034 		memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2035 
2036 	context->mq_open.oflag = oflag;
2037 	context->mq_open.mode = mode;
2038 
2039 	context->type = AUDIT_MQ_OPEN;
2040 }
2041 
2042 /**
2043  * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2044  * @mqdes: MQ descriptor
2045  * @msg_len: Message length
2046  * @msg_prio: Message priority
2047  * @abs_timeout: Message timeout in absolute time
2048  *
2049  */
2050 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2051 			const struct timespec *abs_timeout)
2052 {
2053 	struct audit_context *context = current->audit_context;
2054 	struct timespec *p = &context->mq_sendrecv.abs_timeout;
2055 
2056 	if (abs_timeout)
2057 		memcpy(p, abs_timeout, sizeof(struct timespec));
2058 	else
2059 		memset(p, 0, sizeof(struct timespec));
2060 
2061 	context->mq_sendrecv.mqdes = mqdes;
2062 	context->mq_sendrecv.msg_len = msg_len;
2063 	context->mq_sendrecv.msg_prio = msg_prio;
2064 
2065 	context->type = AUDIT_MQ_SENDRECV;
2066 }
2067 
2068 /**
2069  * __audit_mq_notify - record audit data for a POSIX MQ notify
2070  * @mqdes: MQ descriptor
2071  * @notification: Notification event
2072  *
2073  */
2074 
2075 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2076 {
2077 	struct audit_context *context = current->audit_context;
2078 
2079 	if (notification)
2080 		context->mq_notify.sigev_signo = notification->sigev_signo;
2081 	else
2082 		context->mq_notify.sigev_signo = 0;
2083 
2084 	context->mq_notify.mqdes = mqdes;
2085 	context->type = AUDIT_MQ_NOTIFY;
2086 }
2087 
2088 /**
2089  * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2090  * @mqdes: MQ descriptor
2091  * @mqstat: MQ flags
2092  *
2093  */
2094 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2095 {
2096 	struct audit_context *context = current->audit_context;
2097 	context->mq_getsetattr.mqdes = mqdes;
2098 	context->mq_getsetattr.mqstat = *mqstat;
2099 	context->type = AUDIT_MQ_GETSETATTR;
2100 }
2101 
2102 /**
2103  * audit_ipc_obj - record audit data for ipc object
2104  * @ipcp: ipc permissions
2105  *
2106  */
2107 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2108 {
2109 	struct audit_context *context = current->audit_context;
2110 	context->ipc.uid = ipcp->uid;
2111 	context->ipc.gid = ipcp->gid;
2112 	context->ipc.mode = ipcp->mode;
2113 	context->ipc.has_perm = 0;
2114 	security_ipc_getsecid(ipcp, &context->ipc.osid);
2115 	context->type = AUDIT_IPC;
2116 }
2117 
2118 /**
2119  * audit_ipc_set_perm - record audit data for new ipc permissions
2120  * @qbytes: msgq bytes
2121  * @uid: msgq user id
2122  * @gid: msgq group id
2123  * @mode: msgq mode (permissions)
2124  *
2125  * Called only after audit_ipc_obj().
2126  */
2127 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2128 {
2129 	struct audit_context *context = current->audit_context;
2130 
2131 	context->ipc.qbytes = qbytes;
2132 	context->ipc.perm_uid = uid;
2133 	context->ipc.perm_gid = gid;
2134 	context->ipc.perm_mode = mode;
2135 	context->ipc.has_perm = 1;
2136 }
2137 
2138 void __audit_bprm(struct linux_binprm *bprm)
2139 {
2140 	struct audit_context *context = current->audit_context;
2141 
2142 	context->type = AUDIT_EXECVE;
2143 	context->execve.argc = bprm->argc;
2144 }
2145 
2146 
2147 /**
2148  * audit_socketcall - record audit data for sys_socketcall
2149  * @nargs: number of args, which should not be more than AUDITSC_ARGS.
2150  * @args: args array
2151  *
2152  */
2153 int __audit_socketcall(int nargs, unsigned long *args)
2154 {
2155 	struct audit_context *context = current->audit_context;
2156 
2157 	if (nargs <= 0 || nargs > AUDITSC_ARGS || !args)
2158 		return -EINVAL;
2159 	context->type = AUDIT_SOCKETCALL;
2160 	context->socketcall.nargs = nargs;
2161 	memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2162 	return 0;
2163 }
2164 
2165 /**
2166  * __audit_fd_pair - record audit data for pipe and socketpair
2167  * @fd1: the first file descriptor
2168  * @fd2: the second file descriptor
2169  *
2170  */
2171 void __audit_fd_pair(int fd1, int fd2)
2172 {
2173 	struct audit_context *context = current->audit_context;
2174 	context->fds[0] = fd1;
2175 	context->fds[1] = fd2;
2176 }
2177 
2178 /**
2179  * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2180  * @len: data length in user space
2181  * @a: data address in kernel space
2182  *
2183  * Returns 0 for success or NULL context or < 0 on error.
2184  */
2185 int __audit_sockaddr(int len, void *a)
2186 {
2187 	struct audit_context *context = current->audit_context;
2188 
2189 	if (!context->sockaddr) {
2190 		void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2191 		if (!p)
2192 			return -ENOMEM;
2193 		context->sockaddr = p;
2194 	}
2195 
2196 	context->sockaddr_len = len;
2197 	memcpy(context->sockaddr, a, len);
2198 	return 0;
2199 }
2200 
2201 void __audit_ptrace(struct task_struct *t)
2202 {
2203 	struct audit_context *context = current->audit_context;
2204 
2205 	context->target_pid = t->pid;
2206 	context->target_auid = audit_get_loginuid(t);
2207 	context->target_uid = task_uid(t);
2208 	context->target_sessionid = audit_get_sessionid(t);
2209 	security_task_getsecid(t, &context->target_sid);
2210 	memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2211 }
2212 
2213 /**
2214  * audit_signal_info - record signal info for shutting down audit subsystem
2215  * @sig: signal value
2216  * @t: task being signaled
2217  *
2218  * If the audit subsystem is being terminated, record the task (pid)
2219  * and uid that is doing that.
2220  */
2221 int __audit_signal_info(int sig, struct task_struct *t)
2222 {
2223 	struct audit_aux_data_pids *axp;
2224 	struct task_struct *tsk = current;
2225 	struct audit_context *ctx = tsk->audit_context;
2226 	kuid_t uid = current_uid(), t_uid = task_uid(t);
2227 
2228 	if (audit_pid && t->tgid == audit_pid) {
2229 		if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2230 			audit_sig_pid = tsk->pid;
2231 			if (uid_valid(tsk->loginuid))
2232 				audit_sig_uid = tsk->loginuid;
2233 			else
2234 				audit_sig_uid = uid;
2235 			security_task_getsecid(tsk, &audit_sig_sid);
2236 		}
2237 		if (!audit_signals || audit_dummy_context())
2238 			return 0;
2239 	}
2240 
2241 	/* optimize the common case by putting first signal recipient directly
2242 	 * in audit_context */
2243 	if (!ctx->target_pid) {
2244 		ctx->target_pid = t->tgid;
2245 		ctx->target_auid = audit_get_loginuid(t);
2246 		ctx->target_uid = t_uid;
2247 		ctx->target_sessionid = audit_get_sessionid(t);
2248 		security_task_getsecid(t, &ctx->target_sid);
2249 		memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2250 		return 0;
2251 	}
2252 
2253 	axp = (void *)ctx->aux_pids;
2254 	if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2255 		axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2256 		if (!axp)
2257 			return -ENOMEM;
2258 
2259 		axp->d.type = AUDIT_OBJ_PID;
2260 		axp->d.next = ctx->aux_pids;
2261 		ctx->aux_pids = (void *)axp;
2262 	}
2263 	BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2264 
2265 	axp->target_pid[axp->pid_count] = t->tgid;
2266 	axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2267 	axp->target_uid[axp->pid_count] = t_uid;
2268 	axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2269 	security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2270 	memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2271 	axp->pid_count++;
2272 
2273 	return 0;
2274 }
2275 
2276 /**
2277  * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2278  * @bprm: pointer to the bprm being processed
2279  * @new: the proposed new credentials
2280  * @old: the old credentials
2281  *
2282  * Simply check if the proc already has the caps given by the file and if not
2283  * store the priv escalation info for later auditing at the end of the syscall
2284  *
2285  * -Eric
2286  */
2287 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2288 			   const struct cred *new, const struct cred *old)
2289 {
2290 	struct audit_aux_data_bprm_fcaps *ax;
2291 	struct audit_context *context = current->audit_context;
2292 	struct cpu_vfs_cap_data vcaps;
2293 	struct dentry *dentry;
2294 
2295 	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2296 	if (!ax)
2297 		return -ENOMEM;
2298 
2299 	ax->d.type = AUDIT_BPRM_FCAPS;
2300 	ax->d.next = context->aux;
2301 	context->aux = (void *)ax;
2302 
2303 	dentry = dget(bprm->file->f_dentry);
2304 	get_vfs_caps_from_disk(dentry, &vcaps);
2305 	dput(dentry);
2306 
2307 	ax->fcap.permitted = vcaps.permitted;
2308 	ax->fcap.inheritable = vcaps.inheritable;
2309 	ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2310 	ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2311 
2312 	ax->old_pcap.permitted   = old->cap_permitted;
2313 	ax->old_pcap.inheritable = old->cap_inheritable;
2314 	ax->old_pcap.effective   = old->cap_effective;
2315 
2316 	ax->new_pcap.permitted   = new->cap_permitted;
2317 	ax->new_pcap.inheritable = new->cap_inheritable;
2318 	ax->new_pcap.effective   = new->cap_effective;
2319 	return 0;
2320 }
2321 
2322 /**
2323  * __audit_log_capset - store information about the arguments to the capset syscall
2324  * @pid: target pid of the capset call
2325  * @new: the new credentials
2326  * @old: the old (current) credentials
2327  *
2328  * Record the aguments userspace sent to sys_capset for later printing by the
2329  * audit system if applicable
2330  */
2331 void __audit_log_capset(pid_t pid,
2332 		       const struct cred *new, const struct cred *old)
2333 {
2334 	struct audit_context *context = current->audit_context;
2335 	context->capset.pid = pid;
2336 	context->capset.cap.effective   = new->cap_effective;
2337 	context->capset.cap.inheritable = new->cap_effective;
2338 	context->capset.cap.permitted   = new->cap_permitted;
2339 	context->type = AUDIT_CAPSET;
2340 }
2341 
2342 void __audit_mmap_fd(int fd, int flags)
2343 {
2344 	struct audit_context *context = current->audit_context;
2345 	context->mmap.fd = fd;
2346 	context->mmap.flags = flags;
2347 	context->type = AUDIT_MMAP;
2348 }
2349 
2350 static void audit_log_task(struct audit_buffer *ab)
2351 {
2352 	kuid_t auid, uid;
2353 	kgid_t gid;
2354 	unsigned int sessionid;
2355 
2356 	auid = audit_get_loginuid(current);
2357 	sessionid = audit_get_sessionid(current);
2358 	current_uid_gid(&uid, &gid);
2359 
2360 	audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2361 			 from_kuid(&init_user_ns, auid),
2362 			 from_kuid(&init_user_ns, uid),
2363 			 from_kgid(&init_user_ns, gid),
2364 			 sessionid);
2365 	audit_log_task_context(ab);
2366 	audit_log_format(ab, " pid=%d comm=", current->pid);
2367 	audit_log_untrustedstring(ab, current->comm);
2368 }
2369 
2370 static void audit_log_abend(struct audit_buffer *ab, char *reason, long signr)
2371 {
2372 	audit_log_task(ab);
2373 	audit_log_format(ab, " reason=");
2374 	audit_log_string(ab, reason);
2375 	audit_log_format(ab, " sig=%ld", signr);
2376 }
2377 /**
2378  * audit_core_dumps - record information about processes that end abnormally
2379  * @signr: signal value
2380  *
2381  * If a process ends with a core dump, something fishy is going on and we
2382  * should record the event for investigation.
2383  */
2384 void audit_core_dumps(long signr)
2385 {
2386 	struct audit_buffer *ab;
2387 
2388 	if (!audit_enabled)
2389 		return;
2390 
2391 	if (signr == SIGQUIT)	/* don't care for those */
2392 		return;
2393 
2394 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2395 	if (unlikely(!ab))
2396 		return;
2397 	audit_log_abend(ab, "memory violation", signr);
2398 	audit_log_end(ab);
2399 }
2400 
2401 void __audit_seccomp(unsigned long syscall, long signr, int code)
2402 {
2403 	struct audit_buffer *ab;
2404 
2405 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_SECCOMP);
2406 	if (unlikely(!ab))
2407 		return;
2408 	audit_log_task(ab);
2409 	audit_log_format(ab, " sig=%ld", signr);
2410 	audit_log_format(ab, " syscall=%ld", syscall);
2411 	audit_log_format(ab, " compat=%d", is_compat_task());
2412 	audit_log_format(ab, " ip=0x%lx", KSTK_EIP(current));
2413 	audit_log_format(ab, " code=0x%x", code);
2414 	audit_log_end(ab);
2415 }
2416 
2417 struct list_head *audit_killed_trees(void)
2418 {
2419 	struct audit_context *ctx = current->audit_context;
2420 	if (likely(!ctx || !ctx->in_syscall))
2421 		return NULL;
2422 	return &ctx->killed_trees;
2423 }
2424