xref: /linux/kernel/auditsc.c (revision 3e44c471a2dab210f7e9b1e5f7d4d54d52df59eb)
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
46 
47 #include <linux/init.h>
48 #include <asm/types.h>
49 #include <linux/atomic.h>
50 #include <linux/fs.h>
51 #include <linux/namei.h>
52 #include <linux/mm.h>
53 #include <linux/export.h>
54 #include <linux/slab.h>
55 #include <linux/mount.h>
56 #include <linux/socket.h>
57 #include <linux/mqueue.h>
58 #include <linux/audit.h>
59 #include <linux/personality.h>
60 #include <linux/time.h>
61 #include <linux/netlink.h>
62 #include <linux/compiler.h>
63 #include <asm/unistd.h>
64 #include <linux/security.h>
65 #include <linux/list.h>
66 #include <linux/tty.h>
67 #include <linux/binfmts.h>
68 #include <linux/highmem.h>
69 #include <linux/syscalls.h>
70 #include <asm/syscall.h>
71 #include <linux/capability.h>
72 #include <linux/fs_struct.h>
73 #include <linux/compat.h>
74 #include <linux/ctype.h>
75 #include <linux/string.h>
76 #include <uapi/linux/limits.h>
77 
78 #include "audit.h"
79 
80 /* flags stating the success for a syscall */
81 #define AUDITSC_INVALID 0
82 #define AUDITSC_SUCCESS 1
83 #define AUDITSC_FAILURE 2
84 
85 /* no execve audit message should be longer than this (userspace limits) */
86 #define MAX_EXECVE_AUDIT_LEN 7500
87 
88 /* max length to print of cmdline/proctitle value during audit */
89 #define MAX_PROCTITLE_AUDIT_LEN 128
90 
91 /* number of audit rules */
92 int audit_n_rules;
93 
94 /* determines whether we collect data for signals sent */
95 int audit_signals;
96 
97 struct audit_aux_data {
98 	struct audit_aux_data	*next;
99 	int			type;
100 };
101 
102 #define AUDIT_AUX_IPCPERM	0
103 
104 /* Number of target pids per aux struct. */
105 #define AUDIT_AUX_PIDS	16
106 
107 struct audit_aux_data_pids {
108 	struct audit_aux_data	d;
109 	pid_t			target_pid[AUDIT_AUX_PIDS];
110 	kuid_t			target_auid[AUDIT_AUX_PIDS];
111 	kuid_t			target_uid[AUDIT_AUX_PIDS];
112 	unsigned int		target_sessionid[AUDIT_AUX_PIDS];
113 	u32			target_sid[AUDIT_AUX_PIDS];
114 	char 			target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
115 	int			pid_count;
116 };
117 
118 struct audit_aux_data_bprm_fcaps {
119 	struct audit_aux_data	d;
120 	struct audit_cap_data	fcap;
121 	unsigned int		fcap_ver;
122 	struct audit_cap_data	old_pcap;
123 	struct audit_cap_data	new_pcap;
124 };
125 
126 struct audit_tree_refs {
127 	struct audit_tree_refs *next;
128 	struct audit_chunk *c[31];
129 };
130 
131 static int audit_match_perm(struct audit_context *ctx, int mask)
132 {
133 	unsigned n;
134 	if (unlikely(!ctx))
135 		return 0;
136 	n = ctx->major;
137 
138 	switch (audit_classify_syscall(ctx->arch, n)) {
139 	case 0:	/* native */
140 		if ((mask & AUDIT_PERM_WRITE) &&
141 		     audit_match_class(AUDIT_CLASS_WRITE, n))
142 			return 1;
143 		if ((mask & AUDIT_PERM_READ) &&
144 		     audit_match_class(AUDIT_CLASS_READ, n))
145 			return 1;
146 		if ((mask & AUDIT_PERM_ATTR) &&
147 		     audit_match_class(AUDIT_CLASS_CHATTR, n))
148 			return 1;
149 		return 0;
150 	case 1: /* 32bit on biarch */
151 		if ((mask & AUDIT_PERM_WRITE) &&
152 		     audit_match_class(AUDIT_CLASS_WRITE_32, n))
153 			return 1;
154 		if ((mask & AUDIT_PERM_READ) &&
155 		     audit_match_class(AUDIT_CLASS_READ_32, n))
156 			return 1;
157 		if ((mask & AUDIT_PERM_ATTR) &&
158 		     audit_match_class(AUDIT_CLASS_CHATTR_32, n))
159 			return 1;
160 		return 0;
161 	case 2: /* open */
162 		return mask & ACC_MODE(ctx->argv[1]);
163 	case 3: /* openat */
164 		return mask & ACC_MODE(ctx->argv[2]);
165 	case 4: /* socketcall */
166 		return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
167 	case 5: /* execve */
168 		return mask & AUDIT_PERM_EXEC;
169 	default:
170 		return 0;
171 	}
172 }
173 
174 static int audit_match_filetype(struct audit_context *ctx, int val)
175 {
176 	struct audit_names *n;
177 	umode_t mode = (umode_t)val;
178 
179 	if (unlikely(!ctx))
180 		return 0;
181 
182 	list_for_each_entry(n, &ctx->names_list, list) {
183 		if ((n->ino != -1) &&
184 		    ((n->mode & S_IFMT) == mode))
185 			return 1;
186 	}
187 
188 	return 0;
189 }
190 
191 /*
192  * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
193  * ->first_trees points to its beginning, ->trees - to the current end of data.
194  * ->tree_count is the number of free entries in array pointed to by ->trees.
195  * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
196  * "empty" becomes (p, p, 31) afterwards.  We don't shrink the list (and seriously,
197  * it's going to remain 1-element for almost any setup) until we free context itself.
198  * References in it _are_ dropped - at the same time we free/drop aux stuff.
199  */
200 
201 #ifdef CONFIG_AUDIT_TREE
202 static void audit_set_auditable(struct audit_context *ctx)
203 {
204 	if (!ctx->prio) {
205 		ctx->prio = 1;
206 		ctx->current_state = AUDIT_RECORD_CONTEXT;
207 	}
208 }
209 
210 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
211 {
212 	struct audit_tree_refs *p = ctx->trees;
213 	int left = ctx->tree_count;
214 	if (likely(left)) {
215 		p->c[--left] = chunk;
216 		ctx->tree_count = left;
217 		return 1;
218 	}
219 	if (!p)
220 		return 0;
221 	p = p->next;
222 	if (p) {
223 		p->c[30] = chunk;
224 		ctx->trees = p;
225 		ctx->tree_count = 30;
226 		return 1;
227 	}
228 	return 0;
229 }
230 
231 static int grow_tree_refs(struct audit_context *ctx)
232 {
233 	struct audit_tree_refs *p = ctx->trees;
234 	ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
235 	if (!ctx->trees) {
236 		ctx->trees = p;
237 		return 0;
238 	}
239 	if (p)
240 		p->next = ctx->trees;
241 	else
242 		ctx->first_trees = ctx->trees;
243 	ctx->tree_count = 31;
244 	return 1;
245 }
246 #endif
247 
248 static void unroll_tree_refs(struct audit_context *ctx,
249 		      struct audit_tree_refs *p, int count)
250 {
251 #ifdef CONFIG_AUDIT_TREE
252 	struct audit_tree_refs *q;
253 	int n;
254 	if (!p) {
255 		/* we started with empty chain */
256 		p = ctx->first_trees;
257 		count = 31;
258 		/* if the very first allocation has failed, nothing to do */
259 		if (!p)
260 			return;
261 	}
262 	n = count;
263 	for (q = p; q != ctx->trees; q = q->next, n = 31) {
264 		while (n--) {
265 			audit_put_chunk(q->c[n]);
266 			q->c[n] = NULL;
267 		}
268 	}
269 	while (n-- > ctx->tree_count) {
270 		audit_put_chunk(q->c[n]);
271 		q->c[n] = NULL;
272 	}
273 	ctx->trees = p;
274 	ctx->tree_count = count;
275 #endif
276 }
277 
278 static void free_tree_refs(struct audit_context *ctx)
279 {
280 	struct audit_tree_refs *p, *q;
281 	for (p = ctx->first_trees; p; p = q) {
282 		q = p->next;
283 		kfree(p);
284 	}
285 }
286 
287 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
288 {
289 #ifdef CONFIG_AUDIT_TREE
290 	struct audit_tree_refs *p;
291 	int n;
292 	if (!tree)
293 		return 0;
294 	/* full ones */
295 	for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
296 		for (n = 0; n < 31; n++)
297 			if (audit_tree_match(p->c[n], tree))
298 				return 1;
299 	}
300 	/* partial */
301 	if (p) {
302 		for (n = ctx->tree_count; n < 31; n++)
303 			if (audit_tree_match(p->c[n], tree))
304 				return 1;
305 	}
306 #endif
307 	return 0;
308 }
309 
310 static int audit_compare_uid(kuid_t uid,
311 			     struct audit_names *name,
312 			     struct audit_field *f,
313 			     struct audit_context *ctx)
314 {
315 	struct audit_names *n;
316 	int rc;
317 
318 	if (name) {
319 		rc = audit_uid_comparator(uid, f->op, name->uid);
320 		if (rc)
321 			return rc;
322 	}
323 
324 	if (ctx) {
325 		list_for_each_entry(n, &ctx->names_list, list) {
326 			rc = audit_uid_comparator(uid, f->op, n->uid);
327 			if (rc)
328 				return rc;
329 		}
330 	}
331 	return 0;
332 }
333 
334 static int audit_compare_gid(kgid_t gid,
335 			     struct audit_names *name,
336 			     struct audit_field *f,
337 			     struct audit_context *ctx)
338 {
339 	struct audit_names *n;
340 	int rc;
341 
342 	if (name) {
343 		rc = audit_gid_comparator(gid, f->op, name->gid);
344 		if (rc)
345 			return rc;
346 	}
347 
348 	if (ctx) {
349 		list_for_each_entry(n, &ctx->names_list, list) {
350 			rc = audit_gid_comparator(gid, f->op, n->gid);
351 			if (rc)
352 				return rc;
353 		}
354 	}
355 	return 0;
356 }
357 
358 static int audit_field_compare(struct task_struct *tsk,
359 			       const struct cred *cred,
360 			       struct audit_field *f,
361 			       struct audit_context *ctx,
362 			       struct audit_names *name)
363 {
364 	switch (f->val) {
365 	/* process to file object comparisons */
366 	case AUDIT_COMPARE_UID_TO_OBJ_UID:
367 		return audit_compare_uid(cred->uid, name, f, ctx);
368 	case AUDIT_COMPARE_GID_TO_OBJ_GID:
369 		return audit_compare_gid(cred->gid, name, f, ctx);
370 	case AUDIT_COMPARE_EUID_TO_OBJ_UID:
371 		return audit_compare_uid(cred->euid, name, f, ctx);
372 	case AUDIT_COMPARE_EGID_TO_OBJ_GID:
373 		return audit_compare_gid(cred->egid, name, f, ctx);
374 	case AUDIT_COMPARE_AUID_TO_OBJ_UID:
375 		return audit_compare_uid(tsk->loginuid, name, f, ctx);
376 	case AUDIT_COMPARE_SUID_TO_OBJ_UID:
377 		return audit_compare_uid(cred->suid, name, f, ctx);
378 	case AUDIT_COMPARE_SGID_TO_OBJ_GID:
379 		return audit_compare_gid(cred->sgid, name, f, ctx);
380 	case AUDIT_COMPARE_FSUID_TO_OBJ_UID:
381 		return audit_compare_uid(cred->fsuid, name, f, ctx);
382 	case AUDIT_COMPARE_FSGID_TO_OBJ_GID:
383 		return audit_compare_gid(cred->fsgid, name, f, ctx);
384 	/* uid comparisons */
385 	case AUDIT_COMPARE_UID_TO_AUID:
386 		return audit_uid_comparator(cred->uid, f->op, tsk->loginuid);
387 	case AUDIT_COMPARE_UID_TO_EUID:
388 		return audit_uid_comparator(cred->uid, f->op, cred->euid);
389 	case AUDIT_COMPARE_UID_TO_SUID:
390 		return audit_uid_comparator(cred->uid, f->op, cred->suid);
391 	case AUDIT_COMPARE_UID_TO_FSUID:
392 		return audit_uid_comparator(cred->uid, f->op, cred->fsuid);
393 	/* auid comparisons */
394 	case AUDIT_COMPARE_AUID_TO_EUID:
395 		return audit_uid_comparator(tsk->loginuid, f->op, cred->euid);
396 	case AUDIT_COMPARE_AUID_TO_SUID:
397 		return audit_uid_comparator(tsk->loginuid, f->op, cred->suid);
398 	case AUDIT_COMPARE_AUID_TO_FSUID:
399 		return audit_uid_comparator(tsk->loginuid, f->op, cred->fsuid);
400 	/* euid comparisons */
401 	case AUDIT_COMPARE_EUID_TO_SUID:
402 		return audit_uid_comparator(cred->euid, f->op, cred->suid);
403 	case AUDIT_COMPARE_EUID_TO_FSUID:
404 		return audit_uid_comparator(cred->euid, f->op, cred->fsuid);
405 	/* suid comparisons */
406 	case AUDIT_COMPARE_SUID_TO_FSUID:
407 		return audit_uid_comparator(cred->suid, f->op, cred->fsuid);
408 	/* gid comparisons */
409 	case AUDIT_COMPARE_GID_TO_EGID:
410 		return audit_gid_comparator(cred->gid, f->op, cred->egid);
411 	case AUDIT_COMPARE_GID_TO_SGID:
412 		return audit_gid_comparator(cred->gid, f->op, cred->sgid);
413 	case AUDIT_COMPARE_GID_TO_FSGID:
414 		return audit_gid_comparator(cred->gid, f->op, cred->fsgid);
415 	/* egid comparisons */
416 	case AUDIT_COMPARE_EGID_TO_SGID:
417 		return audit_gid_comparator(cred->egid, f->op, cred->sgid);
418 	case AUDIT_COMPARE_EGID_TO_FSGID:
419 		return audit_gid_comparator(cred->egid, f->op, cred->fsgid);
420 	/* sgid comparison */
421 	case AUDIT_COMPARE_SGID_TO_FSGID:
422 		return audit_gid_comparator(cred->sgid, f->op, cred->fsgid);
423 	default:
424 		WARN(1, "Missing AUDIT_COMPARE define.  Report as a bug\n");
425 		return 0;
426 	}
427 	return 0;
428 }
429 
430 /* Determine if any context name data matches a rule's watch data */
431 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
432  * otherwise.
433  *
434  * If task_creation is true, this is an explicit indication that we are
435  * filtering a task rule at task creation time.  This and tsk == current are
436  * the only situations where tsk->cred may be accessed without an rcu read lock.
437  */
438 static int audit_filter_rules(struct task_struct *tsk,
439 			      struct audit_krule *rule,
440 			      struct audit_context *ctx,
441 			      struct audit_names *name,
442 			      enum audit_state *state,
443 			      bool task_creation)
444 {
445 	const struct cred *cred;
446 	int i, need_sid = 1;
447 	u32 sid;
448 
449 	cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
450 
451 	for (i = 0; i < rule->field_count; i++) {
452 		struct audit_field *f = &rule->fields[i];
453 		struct audit_names *n;
454 		int result = 0;
455 		pid_t pid;
456 
457 		switch (f->type) {
458 		case AUDIT_PID:
459 			pid = task_pid_nr(tsk);
460 			result = audit_comparator(pid, f->op, f->val);
461 			break;
462 		case AUDIT_PPID:
463 			if (ctx) {
464 				if (!ctx->ppid)
465 					ctx->ppid = task_ppid_nr(tsk);
466 				result = audit_comparator(ctx->ppid, f->op, f->val);
467 			}
468 			break;
469 		case AUDIT_UID:
470 			result = audit_uid_comparator(cred->uid, f->op, f->uid);
471 			break;
472 		case AUDIT_EUID:
473 			result = audit_uid_comparator(cred->euid, f->op, f->uid);
474 			break;
475 		case AUDIT_SUID:
476 			result = audit_uid_comparator(cred->suid, f->op, f->uid);
477 			break;
478 		case AUDIT_FSUID:
479 			result = audit_uid_comparator(cred->fsuid, f->op, f->uid);
480 			break;
481 		case AUDIT_GID:
482 			result = audit_gid_comparator(cred->gid, f->op, f->gid);
483 			if (f->op == Audit_equal) {
484 				if (!result)
485 					result = in_group_p(f->gid);
486 			} else if (f->op == Audit_not_equal) {
487 				if (result)
488 					result = !in_group_p(f->gid);
489 			}
490 			break;
491 		case AUDIT_EGID:
492 			result = audit_gid_comparator(cred->egid, f->op, f->gid);
493 			if (f->op == Audit_equal) {
494 				if (!result)
495 					result = in_egroup_p(f->gid);
496 			} else if (f->op == Audit_not_equal) {
497 				if (result)
498 					result = !in_egroup_p(f->gid);
499 			}
500 			break;
501 		case AUDIT_SGID:
502 			result = audit_gid_comparator(cred->sgid, f->op, f->gid);
503 			break;
504 		case AUDIT_FSGID:
505 			result = audit_gid_comparator(cred->fsgid, f->op, f->gid);
506 			break;
507 		case AUDIT_PERS:
508 			result = audit_comparator(tsk->personality, f->op, f->val);
509 			break;
510 		case AUDIT_ARCH:
511 			if (ctx)
512 				result = audit_comparator(ctx->arch, f->op, f->val);
513 			break;
514 
515 		case AUDIT_EXIT:
516 			if (ctx && ctx->return_valid)
517 				result = audit_comparator(ctx->return_code, f->op, f->val);
518 			break;
519 		case AUDIT_SUCCESS:
520 			if (ctx && ctx->return_valid) {
521 				if (f->val)
522 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
523 				else
524 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
525 			}
526 			break;
527 		case AUDIT_DEVMAJOR:
528 			if (name) {
529 				if (audit_comparator(MAJOR(name->dev), f->op, f->val) ||
530 				    audit_comparator(MAJOR(name->rdev), f->op, f->val))
531 					++result;
532 			} else if (ctx) {
533 				list_for_each_entry(n, &ctx->names_list, list) {
534 					if (audit_comparator(MAJOR(n->dev), f->op, f->val) ||
535 					    audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
536 						++result;
537 						break;
538 					}
539 				}
540 			}
541 			break;
542 		case AUDIT_DEVMINOR:
543 			if (name) {
544 				if (audit_comparator(MINOR(name->dev), f->op, f->val) ||
545 				    audit_comparator(MINOR(name->rdev), f->op, f->val))
546 					++result;
547 			} else if (ctx) {
548 				list_for_each_entry(n, &ctx->names_list, list) {
549 					if (audit_comparator(MINOR(n->dev), f->op, f->val) ||
550 					    audit_comparator(MINOR(n->rdev), f->op, f->val)) {
551 						++result;
552 						break;
553 					}
554 				}
555 			}
556 			break;
557 		case AUDIT_INODE:
558 			if (name)
559 				result = audit_comparator(name->ino, f->op, f->val);
560 			else if (ctx) {
561 				list_for_each_entry(n, &ctx->names_list, list) {
562 					if (audit_comparator(n->ino, f->op, f->val)) {
563 						++result;
564 						break;
565 					}
566 				}
567 			}
568 			break;
569 		case AUDIT_OBJ_UID:
570 			if (name) {
571 				result = audit_uid_comparator(name->uid, f->op, f->uid);
572 			} else if (ctx) {
573 				list_for_each_entry(n, &ctx->names_list, list) {
574 					if (audit_uid_comparator(n->uid, f->op, f->uid)) {
575 						++result;
576 						break;
577 					}
578 				}
579 			}
580 			break;
581 		case AUDIT_OBJ_GID:
582 			if (name) {
583 				result = audit_gid_comparator(name->gid, f->op, f->gid);
584 			} else if (ctx) {
585 				list_for_each_entry(n, &ctx->names_list, list) {
586 					if (audit_gid_comparator(n->gid, f->op, f->gid)) {
587 						++result;
588 						break;
589 					}
590 				}
591 			}
592 			break;
593 		case AUDIT_WATCH:
594 			if (name)
595 				result = audit_watch_compare(rule->watch, name->ino, name->dev);
596 			break;
597 		case AUDIT_DIR:
598 			if (ctx)
599 				result = match_tree_refs(ctx, rule->tree);
600 			break;
601 		case AUDIT_LOGINUID:
602 			result = audit_uid_comparator(tsk->loginuid, f->op, f->uid);
603 			break;
604 		case AUDIT_LOGINUID_SET:
605 			result = audit_comparator(audit_loginuid_set(tsk), f->op, f->val);
606 			break;
607 		case AUDIT_SUBJ_USER:
608 		case AUDIT_SUBJ_ROLE:
609 		case AUDIT_SUBJ_TYPE:
610 		case AUDIT_SUBJ_SEN:
611 		case AUDIT_SUBJ_CLR:
612 			/* NOTE: this may return negative values indicating
613 			   a temporary error.  We simply treat this as a
614 			   match for now to avoid losing information that
615 			   may be wanted.   An error message will also be
616 			   logged upon error */
617 			if (f->lsm_rule) {
618 				if (need_sid) {
619 					security_task_getsecid(tsk, &sid);
620 					need_sid = 0;
621 				}
622 				result = security_audit_rule_match(sid, f->type,
623 				                                  f->op,
624 				                                  f->lsm_rule,
625 				                                  ctx);
626 			}
627 			break;
628 		case AUDIT_OBJ_USER:
629 		case AUDIT_OBJ_ROLE:
630 		case AUDIT_OBJ_TYPE:
631 		case AUDIT_OBJ_LEV_LOW:
632 		case AUDIT_OBJ_LEV_HIGH:
633 			/* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
634 			   also applies here */
635 			if (f->lsm_rule) {
636 				/* Find files that match */
637 				if (name) {
638 					result = security_audit_rule_match(
639 					           name->osid, f->type, f->op,
640 					           f->lsm_rule, ctx);
641 				} else if (ctx) {
642 					list_for_each_entry(n, &ctx->names_list, list) {
643 						if (security_audit_rule_match(n->osid, f->type,
644 									      f->op, f->lsm_rule,
645 									      ctx)) {
646 							++result;
647 							break;
648 						}
649 					}
650 				}
651 				/* Find ipc objects that match */
652 				if (!ctx || ctx->type != AUDIT_IPC)
653 					break;
654 				if (security_audit_rule_match(ctx->ipc.osid,
655 							      f->type, f->op,
656 							      f->lsm_rule, ctx))
657 					++result;
658 			}
659 			break;
660 		case AUDIT_ARG0:
661 		case AUDIT_ARG1:
662 		case AUDIT_ARG2:
663 		case AUDIT_ARG3:
664 			if (ctx)
665 				result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
666 			break;
667 		case AUDIT_FILTERKEY:
668 			/* ignore this field for filtering */
669 			result = 1;
670 			break;
671 		case AUDIT_PERM:
672 			result = audit_match_perm(ctx, f->val);
673 			break;
674 		case AUDIT_FILETYPE:
675 			result = audit_match_filetype(ctx, f->val);
676 			break;
677 		case AUDIT_FIELD_COMPARE:
678 			result = audit_field_compare(tsk, cred, f, ctx, name);
679 			break;
680 		}
681 		if (!result)
682 			return 0;
683 	}
684 
685 	if (ctx) {
686 		if (rule->prio <= ctx->prio)
687 			return 0;
688 		if (rule->filterkey) {
689 			kfree(ctx->filterkey);
690 			ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
691 		}
692 		ctx->prio = rule->prio;
693 	}
694 	switch (rule->action) {
695 	case AUDIT_NEVER:    *state = AUDIT_DISABLED;	    break;
696 	case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
697 	}
698 	return 1;
699 }
700 
701 /* At process creation time, we can determine if system-call auditing is
702  * completely disabled for this task.  Since we only have the task
703  * structure at this point, we can only check uid and gid.
704  */
705 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
706 {
707 	struct audit_entry *e;
708 	enum audit_state   state;
709 
710 	rcu_read_lock();
711 	list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
712 		if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
713 				       &state, true)) {
714 			if (state == AUDIT_RECORD_CONTEXT)
715 				*key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
716 			rcu_read_unlock();
717 			return state;
718 		}
719 	}
720 	rcu_read_unlock();
721 	return AUDIT_BUILD_CONTEXT;
722 }
723 
724 static int audit_in_mask(const struct audit_krule *rule, unsigned long val)
725 {
726 	int word, bit;
727 
728 	if (val > 0xffffffff)
729 		return false;
730 
731 	word = AUDIT_WORD(val);
732 	if (word >= AUDIT_BITMASK_SIZE)
733 		return false;
734 
735 	bit = AUDIT_BIT(val);
736 
737 	return rule->mask[word] & bit;
738 }
739 
740 /* At syscall entry and exit time, this filter is called if the
741  * audit_state is not low enough that auditing cannot take place, but is
742  * also not high enough that we already know we have to write an audit
743  * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
744  */
745 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
746 					     struct audit_context *ctx,
747 					     struct list_head *list)
748 {
749 	struct audit_entry *e;
750 	enum audit_state state;
751 
752 	if (audit_pid && tsk->tgid == audit_pid)
753 		return AUDIT_DISABLED;
754 
755 	rcu_read_lock();
756 	if (!list_empty(list)) {
757 		list_for_each_entry_rcu(e, list, list) {
758 			if (audit_in_mask(&e->rule, ctx->major) &&
759 			    audit_filter_rules(tsk, &e->rule, ctx, NULL,
760 					       &state, false)) {
761 				rcu_read_unlock();
762 				ctx->current_state = state;
763 				return state;
764 			}
765 		}
766 	}
767 	rcu_read_unlock();
768 	return AUDIT_BUILD_CONTEXT;
769 }
770 
771 /*
772  * Given an audit_name check the inode hash table to see if they match.
773  * Called holding the rcu read lock to protect the use of audit_inode_hash
774  */
775 static int audit_filter_inode_name(struct task_struct *tsk,
776 				   struct audit_names *n,
777 				   struct audit_context *ctx) {
778 	int h = audit_hash_ino((u32)n->ino);
779 	struct list_head *list = &audit_inode_hash[h];
780 	struct audit_entry *e;
781 	enum audit_state state;
782 
783 	if (list_empty(list))
784 		return 0;
785 
786 	list_for_each_entry_rcu(e, list, list) {
787 		if (audit_in_mask(&e->rule, ctx->major) &&
788 		    audit_filter_rules(tsk, &e->rule, ctx, n, &state, false)) {
789 			ctx->current_state = state;
790 			return 1;
791 		}
792 	}
793 
794 	return 0;
795 }
796 
797 /* At syscall exit time, this filter is called if any audit_names have been
798  * collected during syscall processing.  We only check rules in sublists at hash
799  * buckets applicable to the inode numbers in audit_names.
800  * Regarding audit_state, same rules apply as for audit_filter_syscall().
801  */
802 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
803 {
804 	struct audit_names *n;
805 
806 	if (audit_pid && tsk->tgid == audit_pid)
807 		return;
808 
809 	rcu_read_lock();
810 
811 	list_for_each_entry(n, &ctx->names_list, list) {
812 		if (audit_filter_inode_name(tsk, n, ctx))
813 			break;
814 	}
815 	rcu_read_unlock();
816 }
817 
818 /* Transfer the audit context pointer to the caller, clearing it in the tsk's struct */
819 static inline struct audit_context *audit_take_context(struct task_struct *tsk,
820 						      int return_valid,
821 						      long return_code)
822 {
823 	struct audit_context *context = tsk->audit_context;
824 
825 	if (!context)
826 		return NULL;
827 	context->return_valid = return_valid;
828 
829 	/*
830 	 * we need to fix up the return code in the audit logs if the actual
831 	 * return codes are later going to be fixed up by the arch specific
832 	 * signal handlers
833 	 *
834 	 * This is actually a test for:
835 	 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
836 	 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
837 	 *
838 	 * but is faster than a bunch of ||
839 	 */
840 	if (unlikely(return_code <= -ERESTARTSYS) &&
841 	    (return_code >= -ERESTART_RESTARTBLOCK) &&
842 	    (return_code != -ENOIOCTLCMD))
843 		context->return_code = -EINTR;
844 	else
845 		context->return_code  = return_code;
846 
847 	if (context->in_syscall && !context->dummy) {
848 		audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
849 		audit_filter_inodes(tsk, context);
850 	}
851 
852 	tsk->audit_context = NULL;
853 	return context;
854 }
855 
856 static inline void audit_proctitle_free(struct audit_context *context)
857 {
858 	kfree(context->proctitle.value);
859 	context->proctitle.value = NULL;
860 	context->proctitle.len = 0;
861 }
862 
863 static inline void audit_free_names(struct audit_context *context)
864 {
865 	struct audit_names *n, *next;
866 
867 	list_for_each_entry_safe(n, next, &context->names_list, list) {
868 		list_del(&n->list);
869 		if (n->name)
870 			putname(n->name);
871 		if (n->should_free)
872 			kfree(n);
873 	}
874 	context->name_count = 0;
875 	path_put(&context->pwd);
876 	context->pwd.dentry = NULL;
877 	context->pwd.mnt = NULL;
878 }
879 
880 static inline void audit_free_aux(struct audit_context *context)
881 {
882 	struct audit_aux_data *aux;
883 
884 	while ((aux = context->aux)) {
885 		context->aux = aux->next;
886 		kfree(aux);
887 	}
888 	while ((aux = context->aux_pids)) {
889 		context->aux_pids = aux->next;
890 		kfree(aux);
891 	}
892 }
893 
894 static inline struct audit_context *audit_alloc_context(enum audit_state state)
895 {
896 	struct audit_context *context;
897 
898 	context = kzalloc(sizeof(*context), GFP_KERNEL);
899 	if (!context)
900 		return NULL;
901 	context->state = state;
902 	context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
903 	INIT_LIST_HEAD(&context->killed_trees);
904 	INIT_LIST_HEAD(&context->names_list);
905 	return context;
906 }
907 
908 /**
909  * audit_alloc - allocate an audit context block for a task
910  * @tsk: task
911  *
912  * Filter on the task information and allocate a per-task audit context
913  * if necessary.  Doing so turns on system call auditing for the
914  * specified task.  This is called from copy_process, so no lock is
915  * needed.
916  */
917 int audit_alloc(struct task_struct *tsk)
918 {
919 	struct audit_context *context;
920 	enum audit_state     state;
921 	char *key = NULL;
922 
923 	if (likely(!audit_ever_enabled))
924 		return 0; /* Return if not auditing. */
925 
926 	state = audit_filter_task(tsk, &key);
927 	if (state == AUDIT_DISABLED) {
928 		clear_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
929 		return 0;
930 	}
931 
932 	if (!(context = audit_alloc_context(state))) {
933 		kfree(key);
934 		audit_log_lost("out of memory in audit_alloc");
935 		return -ENOMEM;
936 	}
937 	context->filterkey = key;
938 
939 	tsk->audit_context  = context;
940 	set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
941 	return 0;
942 }
943 
944 static inline void audit_free_context(struct audit_context *context)
945 {
946 	audit_free_names(context);
947 	unroll_tree_refs(context, NULL, 0);
948 	free_tree_refs(context);
949 	audit_free_aux(context);
950 	kfree(context->filterkey);
951 	kfree(context->sockaddr);
952 	audit_proctitle_free(context);
953 	kfree(context);
954 }
955 
956 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
957 				 kuid_t auid, kuid_t uid, unsigned int sessionid,
958 				 u32 sid, char *comm)
959 {
960 	struct audit_buffer *ab;
961 	char *ctx = NULL;
962 	u32 len;
963 	int rc = 0;
964 
965 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
966 	if (!ab)
967 		return rc;
968 
969 	audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid,
970 			 from_kuid(&init_user_ns, auid),
971 			 from_kuid(&init_user_ns, uid), sessionid);
972 	if (sid) {
973 		if (security_secid_to_secctx(sid, &ctx, &len)) {
974 			audit_log_format(ab, " obj=(none)");
975 			rc = 1;
976 		} else {
977 			audit_log_format(ab, " obj=%s", ctx);
978 			security_release_secctx(ctx, len);
979 		}
980 	}
981 	audit_log_format(ab, " ocomm=");
982 	audit_log_untrustedstring(ab, comm);
983 	audit_log_end(ab);
984 
985 	return rc;
986 }
987 
988 /*
989  * to_send and len_sent accounting are very loose estimates.  We aren't
990  * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
991  * within about 500 bytes (next page boundary)
992  *
993  * why snprintf?  an int is up to 12 digits long.  if we just assumed when
994  * logging that a[%d]= was going to be 16 characters long we would be wasting
995  * space in every audit message.  In one 7500 byte message we can log up to
996  * about 1000 min size arguments.  That comes down to about 50% waste of space
997  * if we didn't do the snprintf to find out how long arg_num_len was.
998  */
999 static int audit_log_single_execve_arg(struct audit_context *context,
1000 					struct audit_buffer **ab,
1001 					int arg_num,
1002 					size_t *len_sent,
1003 					const char __user *p,
1004 					char *buf)
1005 {
1006 	char arg_num_len_buf[12];
1007 	const char __user *tmp_p = p;
1008 	/* how many digits are in arg_num? 5 is the length of ' a=""' */
1009 	size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1010 	size_t len, len_left, to_send;
1011 	size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1012 	unsigned int i, has_cntl = 0, too_long = 0;
1013 	int ret;
1014 
1015 	/* strnlen_user includes the null we don't want to send */
1016 	len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1017 
1018 	/*
1019 	 * We just created this mm, if we can't find the strings
1020 	 * we just copied into it something is _very_ wrong. Similar
1021 	 * for strings that are too long, we should not have created
1022 	 * any.
1023 	 */
1024 	if (unlikely((len == 0) || len > MAX_ARG_STRLEN - 1)) {
1025 		WARN_ON(1);
1026 		send_sig(SIGKILL, current, 0);
1027 		return -1;
1028 	}
1029 
1030 	/* walk the whole argument looking for non-ascii chars */
1031 	do {
1032 		if (len_left > MAX_EXECVE_AUDIT_LEN)
1033 			to_send = MAX_EXECVE_AUDIT_LEN;
1034 		else
1035 			to_send = len_left;
1036 		ret = copy_from_user(buf, tmp_p, to_send);
1037 		/*
1038 		 * There is no reason for this copy to be short. We just
1039 		 * copied them here, and the mm hasn't been exposed to user-
1040 		 * space yet.
1041 		 */
1042 		if (ret) {
1043 			WARN_ON(1);
1044 			send_sig(SIGKILL, current, 0);
1045 			return -1;
1046 		}
1047 		buf[to_send] = '\0';
1048 		has_cntl = audit_string_contains_control(buf, to_send);
1049 		if (has_cntl) {
1050 			/*
1051 			 * hex messages get logged as 2 bytes, so we can only
1052 			 * send half as much in each message
1053 			 */
1054 			max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1055 			break;
1056 		}
1057 		len_left -= to_send;
1058 		tmp_p += to_send;
1059 	} while (len_left > 0);
1060 
1061 	len_left = len;
1062 
1063 	if (len > max_execve_audit_len)
1064 		too_long = 1;
1065 
1066 	/* rewalk the argument actually logging the message */
1067 	for (i = 0; len_left > 0; i++) {
1068 		int room_left;
1069 
1070 		if (len_left > max_execve_audit_len)
1071 			to_send = max_execve_audit_len;
1072 		else
1073 			to_send = len_left;
1074 
1075 		/* do we have space left to send this argument in this ab? */
1076 		room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1077 		if (has_cntl)
1078 			room_left -= (to_send * 2);
1079 		else
1080 			room_left -= to_send;
1081 		if (room_left < 0) {
1082 			*len_sent = 0;
1083 			audit_log_end(*ab);
1084 			*ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1085 			if (!*ab)
1086 				return 0;
1087 		}
1088 
1089 		/*
1090 		 * first record needs to say how long the original string was
1091 		 * so we can be sure nothing was lost.
1092 		 */
1093 		if ((i == 0) && (too_long))
1094 			audit_log_format(*ab, " a%d_len=%zu", arg_num,
1095 					 has_cntl ? 2*len : len);
1096 
1097 		/*
1098 		 * normally arguments are small enough to fit and we already
1099 		 * filled buf above when we checked for control characters
1100 		 * so don't bother with another copy_from_user
1101 		 */
1102 		if (len >= max_execve_audit_len)
1103 			ret = copy_from_user(buf, p, to_send);
1104 		else
1105 			ret = 0;
1106 		if (ret) {
1107 			WARN_ON(1);
1108 			send_sig(SIGKILL, current, 0);
1109 			return -1;
1110 		}
1111 		buf[to_send] = '\0';
1112 
1113 		/* actually log it */
1114 		audit_log_format(*ab, " a%d", arg_num);
1115 		if (too_long)
1116 			audit_log_format(*ab, "[%d]", i);
1117 		audit_log_format(*ab, "=");
1118 		if (has_cntl)
1119 			audit_log_n_hex(*ab, buf, to_send);
1120 		else
1121 			audit_log_string(*ab, buf);
1122 
1123 		p += to_send;
1124 		len_left -= to_send;
1125 		*len_sent += arg_num_len;
1126 		if (has_cntl)
1127 			*len_sent += to_send * 2;
1128 		else
1129 			*len_sent += to_send;
1130 	}
1131 	/* include the null we didn't log */
1132 	return len + 1;
1133 }
1134 
1135 static void audit_log_execve_info(struct audit_context *context,
1136 				  struct audit_buffer **ab)
1137 {
1138 	int i, len;
1139 	size_t len_sent = 0;
1140 	const char __user *p;
1141 	char *buf;
1142 
1143 	p = (const char __user *)current->mm->arg_start;
1144 
1145 	audit_log_format(*ab, "argc=%d", context->execve.argc);
1146 
1147 	/*
1148 	 * we need some kernel buffer to hold the userspace args.  Just
1149 	 * allocate one big one rather than allocating one of the right size
1150 	 * for every single argument inside audit_log_single_execve_arg()
1151 	 * should be <8k allocation so should be pretty safe.
1152 	 */
1153 	buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1154 	if (!buf) {
1155 		audit_panic("out of memory for argv string");
1156 		return;
1157 	}
1158 
1159 	for (i = 0; i < context->execve.argc; i++) {
1160 		len = audit_log_single_execve_arg(context, ab, i,
1161 						  &len_sent, p, buf);
1162 		if (len <= 0)
1163 			break;
1164 		p += len;
1165 	}
1166 	kfree(buf);
1167 }
1168 
1169 static void show_special(struct audit_context *context, int *call_panic)
1170 {
1171 	struct audit_buffer *ab;
1172 	int i;
1173 
1174 	ab = audit_log_start(context, GFP_KERNEL, context->type);
1175 	if (!ab)
1176 		return;
1177 
1178 	switch (context->type) {
1179 	case AUDIT_SOCKETCALL: {
1180 		int nargs = context->socketcall.nargs;
1181 		audit_log_format(ab, "nargs=%d", nargs);
1182 		for (i = 0; i < nargs; i++)
1183 			audit_log_format(ab, " a%d=%lx", i,
1184 				context->socketcall.args[i]);
1185 		break; }
1186 	case AUDIT_IPC: {
1187 		u32 osid = context->ipc.osid;
1188 
1189 		audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1190 				 from_kuid(&init_user_ns, context->ipc.uid),
1191 				 from_kgid(&init_user_ns, context->ipc.gid),
1192 				 context->ipc.mode);
1193 		if (osid) {
1194 			char *ctx = NULL;
1195 			u32 len;
1196 			if (security_secid_to_secctx(osid, &ctx, &len)) {
1197 				audit_log_format(ab, " osid=%u", osid);
1198 				*call_panic = 1;
1199 			} else {
1200 				audit_log_format(ab, " obj=%s", ctx);
1201 				security_release_secctx(ctx, len);
1202 			}
1203 		}
1204 		if (context->ipc.has_perm) {
1205 			audit_log_end(ab);
1206 			ab = audit_log_start(context, GFP_KERNEL,
1207 					     AUDIT_IPC_SET_PERM);
1208 			if (unlikely(!ab))
1209 				return;
1210 			audit_log_format(ab,
1211 				"qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1212 				context->ipc.qbytes,
1213 				context->ipc.perm_uid,
1214 				context->ipc.perm_gid,
1215 				context->ipc.perm_mode);
1216 		}
1217 		break; }
1218 	case AUDIT_MQ_OPEN: {
1219 		audit_log_format(ab,
1220 			"oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1221 			"mq_msgsize=%ld mq_curmsgs=%ld",
1222 			context->mq_open.oflag, context->mq_open.mode,
1223 			context->mq_open.attr.mq_flags,
1224 			context->mq_open.attr.mq_maxmsg,
1225 			context->mq_open.attr.mq_msgsize,
1226 			context->mq_open.attr.mq_curmsgs);
1227 		break; }
1228 	case AUDIT_MQ_SENDRECV: {
1229 		audit_log_format(ab,
1230 			"mqdes=%d msg_len=%zd msg_prio=%u "
1231 			"abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1232 			context->mq_sendrecv.mqdes,
1233 			context->mq_sendrecv.msg_len,
1234 			context->mq_sendrecv.msg_prio,
1235 			context->mq_sendrecv.abs_timeout.tv_sec,
1236 			context->mq_sendrecv.abs_timeout.tv_nsec);
1237 		break; }
1238 	case AUDIT_MQ_NOTIFY: {
1239 		audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1240 				context->mq_notify.mqdes,
1241 				context->mq_notify.sigev_signo);
1242 		break; }
1243 	case AUDIT_MQ_GETSETATTR: {
1244 		struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1245 		audit_log_format(ab,
1246 			"mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1247 			"mq_curmsgs=%ld ",
1248 			context->mq_getsetattr.mqdes,
1249 			attr->mq_flags, attr->mq_maxmsg,
1250 			attr->mq_msgsize, attr->mq_curmsgs);
1251 		break; }
1252 	case AUDIT_CAPSET: {
1253 		audit_log_format(ab, "pid=%d", context->capset.pid);
1254 		audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1255 		audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1256 		audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1257 		break; }
1258 	case AUDIT_MMAP: {
1259 		audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1260 				 context->mmap.flags);
1261 		break; }
1262 	case AUDIT_EXECVE: {
1263 		audit_log_execve_info(context, &ab);
1264 		break; }
1265 	}
1266 	audit_log_end(ab);
1267 }
1268 
1269 static inline int audit_proctitle_rtrim(char *proctitle, int len)
1270 {
1271 	char *end = proctitle + len - 1;
1272 	while (end > proctitle && !isprint(*end))
1273 		end--;
1274 
1275 	/* catch the case where proctitle is only 1 non-print character */
1276 	len = end - proctitle + 1;
1277 	len -= isprint(proctitle[len-1]) == 0;
1278 	return len;
1279 }
1280 
1281 static void audit_log_proctitle(struct task_struct *tsk,
1282 			 struct audit_context *context)
1283 {
1284 	int res;
1285 	char *buf;
1286 	char *msg = "(null)";
1287 	int len = strlen(msg);
1288 	struct audit_buffer *ab;
1289 
1290 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PROCTITLE);
1291 	if (!ab)
1292 		return;	/* audit_panic or being filtered */
1293 
1294 	audit_log_format(ab, "proctitle=");
1295 
1296 	/* Not  cached */
1297 	if (!context->proctitle.value) {
1298 		buf = kmalloc(MAX_PROCTITLE_AUDIT_LEN, GFP_KERNEL);
1299 		if (!buf)
1300 			goto out;
1301 		/* Historically called this from procfs naming */
1302 		res = get_cmdline(tsk, buf, MAX_PROCTITLE_AUDIT_LEN);
1303 		if (res == 0) {
1304 			kfree(buf);
1305 			goto out;
1306 		}
1307 		res = audit_proctitle_rtrim(buf, res);
1308 		if (res == 0) {
1309 			kfree(buf);
1310 			goto out;
1311 		}
1312 		context->proctitle.value = buf;
1313 		context->proctitle.len = res;
1314 	}
1315 	msg = context->proctitle.value;
1316 	len = context->proctitle.len;
1317 out:
1318 	audit_log_n_untrustedstring(ab, msg, len);
1319 	audit_log_end(ab);
1320 }
1321 
1322 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1323 {
1324 	int i, call_panic = 0;
1325 	struct audit_buffer *ab;
1326 	struct audit_aux_data *aux;
1327 	struct audit_names *n;
1328 
1329 	/* tsk == current */
1330 	context->personality = tsk->personality;
1331 
1332 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1333 	if (!ab)
1334 		return;		/* audit_panic has been called */
1335 	audit_log_format(ab, "arch=%x syscall=%d",
1336 			 context->arch, context->major);
1337 	if (context->personality != PER_LINUX)
1338 		audit_log_format(ab, " per=%lx", context->personality);
1339 	if (context->return_valid)
1340 		audit_log_format(ab, " success=%s exit=%ld",
1341 				 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1342 				 context->return_code);
1343 
1344 	audit_log_format(ab,
1345 			 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d",
1346 			 context->argv[0],
1347 			 context->argv[1],
1348 			 context->argv[2],
1349 			 context->argv[3],
1350 			 context->name_count);
1351 
1352 	audit_log_task_info(ab, tsk);
1353 	audit_log_key(ab, context->filterkey);
1354 	audit_log_end(ab);
1355 
1356 	for (aux = context->aux; aux; aux = aux->next) {
1357 
1358 		ab = audit_log_start(context, GFP_KERNEL, aux->type);
1359 		if (!ab)
1360 			continue; /* audit_panic has been called */
1361 
1362 		switch (aux->type) {
1363 
1364 		case AUDIT_BPRM_FCAPS: {
1365 			struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1366 			audit_log_format(ab, "fver=%x", axs->fcap_ver);
1367 			audit_log_cap(ab, "fp", &axs->fcap.permitted);
1368 			audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1369 			audit_log_format(ab, " fe=%d", axs->fcap.fE);
1370 			audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1371 			audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1372 			audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1373 			audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1374 			audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1375 			audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1376 			break; }
1377 
1378 		}
1379 		audit_log_end(ab);
1380 	}
1381 
1382 	if (context->type)
1383 		show_special(context, &call_panic);
1384 
1385 	if (context->fds[0] >= 0) {
1386 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1387 		if (ab) {
1388 			audit_log_format(ab, "fd0=%d fd1=%d",
1389 					context->fds[0], context->fds[1]);
1390 			audit_log_end(ab);
1391 		}
1392 	}
1393 
1394 	if (context->sockaddr_len) {
1395 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1396 		if (ab) {
1397 			audit_log_format(ab, "saddr=");
1398 			audit_log_n_hex(ab, (void *)context->sockaddr,
1399 					context->sockaddr_len);
1400 			audit_log_end(ab);
1401 		}
1402 	}
1403 
1404 	for (aux = context->aux_pids; aux; aux = aux->next) {
1405 		struct audit_aux_data_pids *axs = (void *)aux;
1406 
1407 		for (i = 0; i < axs->pid_count; i++)
1408 			if (audit_log_pid_context(context, axs->target_pid[i],
1409 						  axs->target_auid[i],
1410 						  axs->target_uid[i],
1411 						  axs->target_sessionid[i],
1412 						  axs->target_sid[i],
1413 						  axs->target_comm[i]))
1414 				call_panic = 1;
1415 	}
1416 
1417 	if (context->target_pid &&
1418 	    audit_log_pid_context(context, context->target_pid,
1419 				  context->target_auid, context->target_uid,
1420 				  context->target_sessionid,
1421 				  context->target_sid, context->target_comm))
1422 			call_panic = 1;
1423 
1424 	if (context->pwd.dentry && context->pwd.mnt) {
1425 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1426 		if (ab) {
1427 			audit_log_d_path(ab, " cwd=", &context->pwd);
1428 			audit_log_end(ab);
1429 		}
1430 	}
1431 
1432 	i = 0;
1433 	list_for_each_entry(n, &context->names_list, list) {
1434 		if (n->hidden)
1435 			continue;
1436 		audit_log_name(context, n, NULL, i++, &call_panic);
1437 	}
1438 
1439 	audit_log_proctitle(tsk, context);
1440 
1441 	/* Send end of event record to help user space know we are finished */
1442 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1443 	if (ab)
1444 		audit_log_end(ab);
1445 	if (call_panic)
1446 		audit_panic("error converting sid to string");
1447 }
1448 
1449 /**
1450  * audit_free - free a per-task audit context
1451  * @tsk: task whose audit context block to free
1452  *
1453  * Called from copy_process and do_exit
1454  */
1455 void __audit_free(struct task_struct *tsk)
1456 {
1457 	struct audit_context *context;
1458 
1459 	context = audit_take_context(tsk, 0, 0);
1460 	if (!context)
1461 		return;
1462 
1463 	/* Check for system calls that do not go through the exit
1464 	 * function (e.g., exit_group), then free context block.
1465 	 * We use GFP_ATOMIC here because we might be doing this
1466 	 * in the context of the idle thread */
1467 	/* that can happen only if we are called from do_exit() */
1468 	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1469 		audit_log_exit(context, tsk);
1470 	if (!list_empty(&context->killed_trees))
1471 		audit_kill_trees(&context->killed_trees);
1472 
1473 	audit_free_context(context);
1474 }
1475 
1476 /**
1477  * audit_syscall_entry - fill in an audit record at syscall entry
1478  * @major: major syscall type (function)
1479  * @a1: additional syscall register 1
1480  * @a2: additional syscall register 2
1481  * @a3: additional syscall register 3
1482  * @a4: additional syscall register 4
1483  *
1484  * Fill in audit context at syscall entry.  This only happens if the
1485  * audit context was created when the task was created and the state or
1486  * filters demand the audit context be built.  If the state from the
1487  * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1488  * then the record will be written at syscall exit time (otherwise, it
1489  * will only be written if another part of the kernel requests that it
1490  * be written).
1491  */
1492 void __audit_syscall_entry(int major, unsigned long a1, unsigned long a2,
1493 			   unsigned long a3, unsigned long a4)
1494 {
1495 	struct task_struct *tsk = current;
1496 	struct audit_context *context = tsk->audit_context;
1497 	enum audit_state     state;
1498 
1499 	if (!context)
1500 		return;
1501 
1502 	BUG_ON(context->in_syscall || context->name_count);
1503 
1504 	if (!audit_enabled)
1505 		return;
1506 
1507 	context->arch	    = syscall_get_arch();
1508 	context->major      = major;
1509 	context->argv[0]    = a1;
1510 	context->argv[1]    = a2;
1511 	context->argv[2]    = a3;
1512 	context->argv[3]    = a4;
1513 
1514 	state = context->state;
1515 	context->dummy = !audit_n_rules;
1516 	if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1517 		context->prio = 0;
1518 		state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1519 	}
1520 	if (state == AUDIT_DISABLED)
1521 		return;
1522 
1523 	context->serial     = 0;
1524 	context->ctime      = CURRENT_TIME;
1525 	context->in_syscall = 1;
1526 	context->current_state  = state;
1527 	context->ppid       = 0;
1528 }
1529 
1530 /**
1531  * audit_syscall_exit - deallocate audit context after a system call
1532  * @success: success value of the syscall
1533  * @return_code: return value of the syscall
1534  *
1535  * Tear down after system call.  If the audit context has been marked as
1536  * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1537  * filtering, or because some other part of the kernel wrote an audit
1538  * message), then write out the syscall information.  In call cases,
1539  * free the names stored from getname().
1540  */
1541 void __audit_syscall_exit(int success, long return_code)
1542 {
1543 	struct task_struct *tsk = current;
1544 	struct audit_context *context;
1545 
1546 	if (success)
1547 		success = AUDITSC_SUCCESS;
1548 	else
1549 		success = AUDITSC_FAILURE;
1550 
1551 	context = audit_take_context(tsk, success, return_code);
1552 	if (!context)
1553 		return;
1554 
1555 	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1556 		audit_log_exit(context, tsk);
1557 
1558 	context->in_syscall = 0;
1559 	context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1560 
1561 	if (!list_empty(&context->killed_trees))
1562 		audit_kill_trees(&context->killed_trees);
1563 
1564 	audit_free_names(context);
1565 	unroll_tree_refs(context, NULL, 0);
1566 	audit_free_aux(context);
1567 	context->aux = NULL;
1568 	context->aux_pids = NULL;
1569 	context->target_pid = 0;
1570 	context->target_sid = 0;
1571 	context->sockaddr_len = 0;
1572 	context->type = 0;
1573 	context->fds[0] = -1;
1574 	if (context->state != AUDIT_RECORD_CONTEXT) {
1575 		kfree(context->filterkey);
1576 		context->filterkey = NULL;
1577 	}
1578 	tsk->audit_context = context;
1579 }
1580 
1581 static inline void handle_one(const struct inode *inode)
1582 {
1583 #ifdef CONFIG_AUDIT_TREE
1584 	struct audit_context *context;
1585 	struct audit_tree_refs *p;
1586 	struct audit_chunk *chunk;
1587 	int count;
1588 	if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1589 		return;
1590 	context = current->audit_context;
1591 	p = context->trees;
1592 	count = context->tree_count;
1593 	rcu_read_lock();
1594 	chunk = audit_tree_lookup(inode);
1595 	rcu_read_unlock();
1596 	if (!chunk)
1597 		return;
1598 	if (likely(put_tree_ref(context, chunk)))
1599 		return;
1600 	if (unlikely(!grow_tree_refs(context))) {
1601 		pr_warn("out of memory, audit has lost a tree reference\n");
1602 		audit_set_auditable(context);
1603 		audit_put_chunk(chunk);
1604 		unroll_tree_refs(context, p, count);
1605 		return;
1606 	}
1607 	put_tree_ref(context, chunk);
1608 #endif
1609 }
1610 
1611 static void handle_path(const struct dentry *dentry)
1612 {
1613 #ifdef CONFIG_AUDIT_TREE
1614 	struct audit_context *context;
1615 	struct audit_tree_refs *p;
1616 	const struct dentry *d, *parent;
1617 	struct audit_chunk *drop;
1618 	unsigned long seq;
1619 	int count;
1620 
1621 	context = current->audit_context;
1622 	p = context->trees;
1623 	count = context->tree_count;
1624 retry:
1625 	drop = NULL;
1626 	d = dentry;
1627 	rcu_read_lock();
1628 	seq = read_seqbegin(&rename_lock);
1629 	for(;;) {
1630 		struct inode *inode = d_backing_inode(d);
1631 		if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1632 			struct audit_chunk *chunk;
1633 			chunk = audit_tree_lookup(inode);
1634 			if (chunk) {
1635 				if (unlikely(!put_tree_ref(context, chunk))) {
1636 					drop = chunk;
1637 					break;
1638 				}
1639 			}
1640 		}
1641 		parent = d->d_parent;
1642 		if (parent == d)
1643 			break;
1644 		d = parent;
1645 	}
1646 	if (unlikely(read_seqretry(&rename_lock, seq) || drop)) {  /* in this order */
1647 		rcu_read_unlock();
1648 		if (!drop) {
1649 			/* just a race with rename */
1650 			unroll_tree_refs(context, p, count);
1651 			goto retry;
1652 		}
1653 		audit_put_chunk(drop);
1654 		if (grow_tree_refs(context)) {
1655 			/* OK, got more space */
1656 			unroll_tree_refs(context, p, count);
1657 			goto retry;
1658 		}
1659 		/* too bad */
1660 		pr_warn("out of memory, audit has lost a tree reference\n");
1661 		unroll_tree_refs(context, p, count);
1662 		audit_set_auditable(context);
1663 		return;
1664 	}
1665 	rcu_read_unlock();
1666 #endif
1667 }
1668 
1669 static struct audit_names *audit_alloc_name(struct audit_context *context,
1670 						unsigned char type)
1671 {
1672 	struct audit_names *aname;
1673 
1674 	if (context->name_count < AUDIT_NAMES) {
1675 		aname = &context->preallocated_names[context->name_count];
1676 		memset(aname, 0, sizeof(*aname));
1677 	} else {
1678 		aname = kzalloc(sizeof(*aname), GFP_NOFS);
1679 		if (!aname)
1680 			return NULL;
1681 		aname->should_free = true;
1682 	}
1683 
1684 	aname->ino = (unsigned long)-1;
1685 	aname->type = type;
1686 	list_add_tail(&aname->list, &context->names_list);
1687 
1688 	context->name_count++;
1689 	return aname;
1690 }
1691 
1692 /**
1693  * audit_reusename - fill out filename with info from existing entry
1694  * @uptr: userland ptr to pathname
1695  *
1696  * Search the audit_names list for the current audit context. If there is an
1697  * existing entry with a matching "uptr" then return the filename
1698  * associated with that audit_name. If not, return NULL.
1699  */
1700 struct filename *
1701 __audit_reusename(const __user char *uptr)
1702 {
1703 	struct audit_context *context = current->audit_context;
1704 	struct audit_names *n;
1705 
1706 	list_for_each_entry(n, &context->names_list, list) {
1707 		if (!n->name)
1708 			continue;
1709 		if (n->name->uptr == uptr) {
1710 			n->name->refcnt++;
1711 			return n->name;
1712 		}
1713 	}
1714 	return NULL;
1715 }
1716 
1717 /**
1718  * audit_getname - add a name to the list
1719  * @name: name to add
1720  *
1721  * Add a name to the list of audit names for this context.
1722  * Called from fs/namei.c:getname().
1723  */
1724 void __audit_getname(struct filename *name)
1725 {
1726 	struct audit_context *context = current->audit_context;
1727 	struct audit_names *n;
1728 
1729 	if (!context->in_syscall)
1730 		return;
1731 
1732 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
1733 	if (!n)
1734 		return;
1735 
1736 	n->name = name;
1737 	n->name_len = AUDIT_NAME_FULL;
1738 	name->aname = n;
1739 	name->refcnt++;
1740 
1741 	if (!context->pwd.dentry)
1742 		get_fs_pwd(current->fs, &context->pwd);
1743 }
1744 
1745 /**
1746  * __audit_inode - store the inode and device from a lookup
1747  * @name: name being audited
1748  * @dentry: dentry being audited
1749  * @flags: attributes for this particular entry
1750  */
1751 void __audit_inode(struct filename *name, const struct dentry *dentry,
1752 		   unsigned int flags)
1753 {
1754 	struct audit_context *context = current->audit_context;
1755 	const struct inode *inode = d_backing_inode(dentry);
1756 	struct audit_names *n;
1757 	bool parent = flags & AUDIT_INODE_PARENT;
1758 
1759 	if (!context->in_syscall)
1760 		return;
1761 
1762 	if (!name)
1763 		goto out_alloc;
1764 
1765 	/*
1766 	 * If we have a pointer to an audit_names entry already, then we can
1767 	 * just use it directly if the type is correct.
1768 	 */
1769 	n = name->aname;
1770 	if (n) {
1771 		if (parent) {
1772 			if (n->type == AUDIT_TYPE_PARENT ||
1773 			    n->type == AUDIT_TYPE_UNKNOWN)
1774 				goto out;
1775 		} else {
1776 			if (n->type != AUDIT_TYPE_PARENT)
1777 				goto out;
1778 		}
1779 	}
1780 
1781 	list_for_each_entry_reverse(n, &context->names_list, list) {
1782 		if (n->ino) {
1783 			/* valid inode number, use that for the comparison */
1784 			if (n->ino != inode->i_ino ||
1785 			    n->dev != inode->i_sb->s_dev)
1786 				continue;
1787 		} else if (n->name) {
1788 			/* inode number has not been set, check the name */
1789 			if (strcmp(n->name->name, name->name))
1790 				continue;
1791 		} else
1792 			/* no inode and no name (?!) ... this is odd ... */
1793 			continue;
1794 
1795 		/* match the correct record type */
1796 		if (parent) {
1797 			if (n->type == AUDIT_TYPE_PARENT ||
1798 			    n->type == AUDIT_TYPE_UNKNOWN)
1799 				goto out;
1800 		} else {
1801 			if (n->type != AUDIT_TYPE_PARENT)
1802 				goto out;
1803 		}
1804 	}
1805 
1806 out_alloc:
1807 	/* unable to find an entry with both a matching name and type */
1808 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
1809 	if (!n)
1810 		return;
1811 	if (name) {
1812 		n->name = name;
1813 		name->refcnt++;
1814 	}
1815 
1816 out:
1817 	if (parent) {
1818 		n->name_len = n->name ? parent_len(n->name->name) : AUDIT_NAME_FULL;
1819 		n->type = AUDIT_TYPE_PARENT;
1820 		if (flags & AUDIT_INODE_HIDDEN)
1821 			n->hidden = true;
1822 	} else {
1823 		n->name_len = AUDIT_NAME_FULL;
1824 		n->type = AUDIT_TYPE_NORMAL;
1825 	}
1826 	handle_path(dentry);
1827 	audit_copy_inode(n, dentry, inode);
1828 }
1829 
1830 void __audit_file(const struct file *file)
1831 {
1832 	__audit_inode(NULL, file->f_path.dentry, 0);
1833 }
1834 
1835 /**
1836  * __audit_inode_child - collect inode info for created/removed objects
1837  * @parent: inode of dentry parent
1838  * @dentry: dentry being audited
1839  * @type:   AUDIT_TYPE_* value that we're looking for
1840  *
1841  * For syscalls that create or remove filesystem objects, audit_inode
1842  * can only collect information for the filesystem object's parent.
1843  * This call updates the audit context with the child's information.
1844  * Syscalls that create a new filesystem object must be hooked after
1845  * the object is created.  Syscalls that remove a filesystem object
1846  * must be hooked prior, in order to capture the target inode during
1847  * unsuccessful attempts.
1848  */
1849 void __audit_inode_child(const struct inode *parent,
1850 			 const struct dentry *dentry,
1851 			 const unsigned char type)
1852 {
1853 	struct audit_context *context = current->audit_context;
1854 	const struct inode *inode = d_backing_inode(dentry);
1855 	const char *dname = dentry->d_name.name;
1856 	struct audit_names *n, *found_parent = NULL, *found_child = NULL;
1857 
1858 	if (!context->in_syscall)
1859 		return;
1860 
1861 	if (inode)
1862 		handle_one(inode);
1863 
1864 	/* look for a parent entry first */
1865 	list_for_each_entry(n, &context->names_list, list) {
1866 		if (!n->name ||
1867 		    (n->type != AUDIT_TYPE_PARENT &&
1868 		     n->type != AUDIT_TYPE_UNKNOWN))
1869 			continue;
1870 
1871 		if (n->ino == parent->i_ino && n->dev == parent->i_sb->s_dev &&
1872 		    !audit_compare_dname_path(dname,
1873 					      n->name->name, n->name_len)) {
1874 			if (n->type == AUDIT_TYPE_UNKNOWN)
1875 				n->type = AUDIT_TYPE_PARENT;
1876 			found_parent = n;
1877 			break;
1878 		}
1879 	}
1880 
1881 	/* is there a matching child entry? */
1882 	list_for_each_entry(n, &context->names_list, list) {
1883 		/* can only match entries that have a name */
1884 		if (!n->name ||
1885 		    (n->type != type && n->type != AUDIT_TYPE_UNKNOWN))
1886 			continue;
1887 
1888 		if (!strcmp(dname, n->name->name) ||
1889 		    !audit_compare_dname_path(dname, n->name->name,
1890 						found_parent ?
1891 						found_parent->name_len :
1892 						AUDIT_NAME_FULL)) {
1893 			if (n->type == AUDIT_TYPE_UNKNOWN)
1894 				n->type = type;
1895 			found_child = n;
1896 			break;
1897 		}
1898 	}
1899 
1900 	if (!found_parent) {
1901 		/* create a new, "anonymous" parent record */
1902 		n = audit_alloc_name(context, AUDIT_TYPE_PARENT);
1903 		if (!n)
1904 			return;
1905 		audit_copy_inode(n, NULL, parent);
1906 	}
1907 
1908 	if (!found_child) {
1909 		found_child = audit_alloc_name(context, type);
1910 		if (!found_child)
1911 			return;
1912 
1913 		/* Re-use the name belonging to the slot for a matching parent
1914 		 * directory. All names for this context are relinquished in
1915 		 * audit_free_names() */
1916 		if (found_parent) {
1917 			found_child->name = found_parent->name;
1918 			found_child->name_len = AUDIT_NAME_FULL;
1919 			found_child->name->refcnt++;
1920 		}
1921 	}
1922 
1923 	if (inode)
1924 		audit_copy_inode(found_child, dentry, inode);
1925 	else
1926 		found_child->ino = (unsigned long)-1;
1927 }
1928 EXPORT_SYMBOL_GPL(__audit_inode_child);
1929 
1930 /**
1931  * auditsc_get_stamp - get local copies of audit_context values
1932  * @ctx: audit_context for the task
1933  * @t: timespec to store time recorded in the audit_context
1934  * @serial: serial value that is recorded in the audit_context
1935  *
1936  * Also sets the context as auditable.
1937  */
1938 int auditsc_get_stamp(struct audit_context *ctx,
1939 		       struct timespec *t, unsigned int *serial)
1940 {
1941 	if (!ctx->in_syscall)
1942 		return 0;
1943 	if (!ctx->serial)
1944 		ctx->serial = audit_serial();
1945 	t->tv_sec  = ctx->ctime.tv_sec;
1946 	t->tv_nsec = ctx->ctime.tv_nsec;
1947 	*serial    = ctx->serial;
1948 	if (!ctx->prio) {
1949 		ctx->prio = 1;
1950 		ctx->current_state = AUDIT_RECORD_CONTEXT;
1951 	}
1952 	return 1;
1953 }
1954 
1955 /* global counter which is incremented every time something logs in */
1956 static atomic_t session_id = ATOMIC_INIT(0);
1957 
1958 static int audit_set_loginuid_perm(kuid_t loginuid)
1959 {
1960 	/* if we are unset, we don't need privs */
1961 	if (!audit_loginuid_set(current))
1962 		return 0;
1963 	/* if AUDIT_FEATURE_LOGINUID_IMMUTABLE means never ever allow a change*/
1964 	if (is_audit_feature_set(AUDIT_FEATURE_LOGINUID_IMMUTABLE))
1965 		return -EPERM;
1966 	/* it is set, you need permission */
1967 	if (!capable(CAP_AUDIT_CONTROL))
1968 		return -EPERM;
1969 	/* reject if this is not an unset and we don't allow that */
1970 	if (is_audit_feature_set(AUDIT_FEATURE_ONLY_UNSET_LOGINUID) && uid_valid(loginuid))
1971 		return -EPERM;
1972 	return 0;
1973 }
1974 
1975 static void audit_log_set_loginuid(kuid_t koldloginuid, kuid_t kloginuid,
1976 				   unsigned int oldsessionid, unsigned int sessionid,
1977 				   int rc)
1978 {
1979 	struct audit_buffer *ab;
1980 	uid_t uid, oldloginuid, loginuid;
1981 
1982 	if (!audit_enabled)
1983 		return;
1984 
1985 	uid = from_kuid(&init_user_ns, task_uid(current));
1986 	oldloginuid = from_kuid(&init_user_ns, koldloginuid);
1987 	loginuid = from_kuid(&init_user_ns, kloginuid),
1988 
1989 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1990 	if (!ab)
1991 		return;
1992 	audit_log_format(ab, "pid=%d uid=%u", task_pid_nr(current), uid);
1993 	audit_log_task_context(ab);
1994 	audit_log_format(ab, " old-auid=%u auid=%u old-ses=%u ses=%u res=%d",
1995 			 oldloginuid, loginuid, oldsessionid, sessionid, !rc);
1996 	audit_log_end(ab);
1997 }
1998 
1999 /**
2000  * audit_set_loginuid - set current task's audit_context loginuid
2001  * @loginuid: loginuid value
2002  *
2003  * Returns 0.
2004  *
2005  * Called (set) from fs/proc/base.c::proc_loginuid_write().
2006  */
2007 int audit_set_loginuid(kuid_t loginuid)
2008 {
2009 	struct task_struct *task = current;
2010 	unsigned int oldsessionid, sessionid = (unsigned int)-1;
2011 	kuid_t oldloginuid;
2012 	int rc;
2013 
2014 	oldloginuid = audit_get_loginuid(current);
2015 	oldsessionid = audit_get_sessionid(current);
2016 
2017 	rc = audit_set_loginuid_perm(loginuid);
2018 	if (rc)
2019 		goto out;
2020 
2021 	/* are we setting or clearing? */
2022 	if (uid_valid(loginuid))
2023 		sessionid = (unsigned int)atomic_inc_return(&session_id);
2024 
2025 	task->sessionid = sessionid;
2026 	task->loginuid = loginuid;
2027 out:
2028 	audit_log_set_loginuid(oldloginuid, loginuid, oldsessionid, sessionid, rc);
2029 	return rc;
2030 }
2031 
2032 /**
2033  * __audit_mq_open - record audit data for a POSIX MQ open
2034  * @oflag: open flag
2035  * @mode: mode bits
2036  * @attr: queue attributes
2037  *
2038  */
2039 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2040 {
2041 	struct audit_context *context = current->audit_context;
2042 
2043 	if (attr)
2044 		memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2045 	else
2046 		memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2047 
2048 	context->mq_open.oflag = oflag;
2049 	context->mq_open.mode = mode;
2050 
2051 	context->type = AUDIT_MQ_OPEN;
2052 }
2053 
2054 /**
2055  * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2056  * @mqdes: MQ descriptor
2057  * @msg_len: Message length
2058  * @msg_prio: Message priority
2059  * @abs_timeout: Message timeout in absolute time
2060  *
2061  */
2062 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2063 			const struct timespec *abs_timeout)
2064 {
2065 	struct audit_context *context = current->audit_context;
2066 	struct timespec *p = &context->mq_sendrecv.abs_timeout;
2067 
2068 	if (abs_timeout)
2069 		memcpy(p, abs_timeout, sizeof(struct timespec));
2070 	else
2071 		memset(p, 0, sizeof(struct timespec));
2072 
2073 	context->mq_sendrecv.mqdes = mqdes;
2074 	context->mq_sendrecv.msg_len = msg_len;
2075 	context->mq_sendrecv.msg_prio = msg_prio;
2076 
2077 	context->type = AUDIT_MQ_SENDRECV;
2078 }
2079 
2080 /**
2081  * __audit_mq_notify - record audit data for a POSIX MQ notify
2082  * @mqdes: MQ descriptor
2083  * @notification: Notification event
2084  *
2085  */
2086 
2087 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2088 {
2089 	struct audit_context *context = current->audit_context;
2090 
2091 	if (notification)
2092 		context->mq_notify.sigev_signo = notification->sigev_signo;
2093 	else
2094 		context->mq_notify.sigev_signo = 0;
2095 
2096 	context->mq_notify.mqdes = mqdes;
2097 	context->type = AUDIT_MQ_NOTIFY;
2098 }
2099 
2100 /**
2101  * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2102  * @mqdes: MQ descriptor
2103  * @mqstat: MQ flags
2104  *
2105  */
2106 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2107 {
2108 	struct audit_context *context = current->audit_context;
2109 	context->mq_getsetattr.mqdes = mqdes;
2110 	context->mq_getsetattr.mqstat = *mqstat;
2111 	context->type = AUDIT_MQ_GETSETATTR;
2112 }
2113 
2114 /**
2115  * audit_ipc_obj - record audit data for ipc object
2116  * @ipcp: ipc permissions
2117  *
2118  */
2119 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2120 {
2121 	struct audit_context *context = current->audit_context;
2122 	context->ipc.uid = ipcp->uid;
2123 	context->ipc.gid = ipcp->gid;
2124 	context->ipc.mode = ipcp->mode;
2125 	context->ipc.has_perm = 0;
2126 	security_ipc_getsecid(ipcp, &context->ipc.osid);
2127 	context->type = AUDIT_IPC;
2128 }
2129 
2130 /**
2131  * audit_ipc_set_perm - record audit data for new ipc permissions
2132  * @qbytes: msgq bytes
2133  * @uid: msgq user id
2134  * @gid: msgq group id
2135  * @mode: msgq mode (permissions)
2136  *
2137  * Called only after audit_ipc_obj().
2138  */
2139 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2140 {
2141 	struct audit_context *context = current->audit_context;
2142 
2143 	context->ipc.qbytes = qbytes;
2144 	context->ipc.perm_uid = uid;
2145 	context->ipc.perm_gid = gid;
2146 	context->ipc.perm_mode = mode;
2147 	context->ipc.has_perm = 1;
2148 }
2149 
2150 void __audit_bprm(struct linux_binprm *bprm)
2151 {
2152 	struct audit_context *context = current->audit_context;
2153 
2154 	context->type = AUDIT_EXECVE;
2155 	context->execve.argc = bprm->argc;
2156 }
2157 
2158 
2159 /**
2160  * audit_socketcall - record audit data for sys_socketcall
2161  * @nargs: number of args, which should not be more than AUDITSC_ARGS.
2162  * @args: args array
2163  *
2164  */
2165 int __audit_socketcall(int nargs, unsigned long *args)
2166 {
2167 	struct audit_context *context = current->audit_context;
2168 
2169 	if (nargs <= 0 || nargs > AUDITSC_ARGS || !args)
2170 		return -EINVAL;
2171 	context->type = AUDIT_SOCKETCALL;
2172 	context->socketcall.nargs = nargs;
2173 	memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2174 	return 0;
2175 }
2176 
2177 /**
2178  * __audit_fd_pair - record audit data for pipe and socketpair
2179  * @fd1: the first file descriptor
2180  * @fd2: the second file descriptor
2181  *
2182  */
2183 void __audit_fd_pair(int fd1, int fd2)
2184 {
2185 	struct audit_context *context = current->audit_context;
2186 	context->fds[0] = fd1;
2187 	context->fds[1] = fd2;
2188 }
2189 
2190 /**
2191  * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2192  * @len: data length in user space
2193  * @a: data address in kernel space
2194  *
2195  * Returns 0 for success or NULL context or < 0 on error.
2196  */
2197 int __audit_sockaddr(int len, void *a)
2198 {
2199 	struct audit_context *context = current->audit_context;
2200 
2201 	if (!context->sockaddr) {
2202 		void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2203 		if (!p)
2204 			return -ENOMEM;
2205 		context->sockaddr = p;
2206 	}
2207 
2208 	context->sockaddr_len = len;
2209 	memcpy(context->sockaddr, a, len);
2210 	return 0;
2211 }
2212 
2213 void __audit_ptrace(struct task_struct *t)
2214 {
2215 	struct audit_context *context = current->audit_context;
2216 
2217 	context->target_pid = task_pid_nr(t);
2218 	context->target_auid = audit_get_loginuid(t);
2219 	context->target_uid = task_uid(t);
2220 	context->target_sessionid = audit_get_sessionid(t);
2221 	security_task_getsecid(t, &context->target_sid);
2222 	memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2223 }
2224 
2225 /**
2226  * audit_signal_info - record signal info for shutting down audit subsystem
2227  * @sig: signal value
2228  * @t: task being signaled
2229  *
2230  * If the audit subsystem is being terminated, record the task (pid)
2231  * and uid that is doing that.
2232  */
2233 int __audit_signal_info(int sig, struct task_struct *t)
2234 {
2235 	struct audit_aux_data_pids *axp;
2236 	struct task_struct *tsk = current;
2237 	struct audit_context *ctx = tsk->audit_context;
2238 	kuid_t uid = current_uid(), t_uid = task_uid(t);
2239 
2240 	if (audit_pid && t->tgid == audit_pid) {
2241 		if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2242 			audit_sig_pid = task_pid_nr(tsk);
2243 			if (uid_valid(tsk->loginuid))
2244 				audit_sig_uid = tsk->loginuid;
2245 			else
2246 				audit_sig_uid = uid;
2247 			security_task_getsecid(tsk, &audit_sig_sid);
2248 		}
2249 		if (!audit_signals || audit_dummy_context())
2250 			return 0;
2251 	}
2252 
2253 	/* optimize the common case by putting first signal recipient directly
2254 	 * in audit_context */
2255 	if (!ctx->target_pid) {
2256 		ctx->target_pid = task_tgid_nr(t);
2257 		ctx->target_auid = audit_get_loginuid(t);
2258 		ctx->target_uid = t_uid;
2259 		ctx->target_sessionid = audit_get_sessionid(t);
2260 		security_task_getsecid(t, &ctx->target_sid);
2261 		memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2262 		return 0;
2263 	}
2264 
2265 	axp = (void *)ctx->aux_pids;
2266 	if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2267 		axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2268 		if (!axp)
2269 			return -ENOMEM;
2270 
2271 		axp->d.type = AUDIT_OBJ_PID;
2272 		axp->d.next = ctx->aux_pids;
2273 		ctx->aux_pids = (void *)axp;
2274 	}
2275 	BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2276 
2277 	axp->target_pid[axp->pid_count] = task_tgid_nr(t);
2278 	axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2279 	axp->target_uid[axp->pid_count] = t_uid;
2280 	axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2281 	security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2282 	memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2283 	axp->pid_count++;
2284 
2285 	return 0;
2286 }
2287 
2288 /**
2289  * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2290  * @bprm: pointer to the bprm being processed
2291  * @new: the proposed new credentials
2292  * @old: the old credentials
2293  *
2294  * Simply check if the proc already has the caps given by the file and if not
2295  * store the priv escalation info for later auditing at the end of the syscall
2296  *
2297  * -Eric
2298  */
2299 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2300 			   const struct cred *new, const struct cred *old)
2301 {
2302 	struct audit_aux_data_bprm_fcaps *ax;
2303 	struct audit_context *context = current->audit_context;
2304 	struct cpu_vfs_cap_data vcaps;
2305 
2306 	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2307 	if (!ax)
2308 		return -ENOMEM;
2309 
2310 	ax->d.type = AUDIT_BPRM_FCAPS;
2311 	ax->d.next = context->aux;
2312 	context->aux = (void *)ax;
2313 
2314 	get_vfs_caps_from_disk(bprm->file->f_path.dentry, &vcaps);
2315 
2316 	ax->fcap.permitted = vcaps.permitted;
2317 	ax->fcap.inheritable = vcaps.inheritable;
2318 	ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2319 	ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2320 
2321 	ax->old_pcap.permitted   = old->cap_permitted;
2322 	ax->old_pcap.inheritable = old->cap_inheritable;
2323 	ax->old_pcap.effective   = old->cap_effective;
2324 
2325 	ax->new_pcap.permitted   = new->cap_permitted;
2326 	ax->new_pcap.inheritable = new->cap_inheritable;
2327 	ax->new_pcap.effective   = new->cap_effective;
2328 	return 0;
2329 }
2330 
2331 /**
2332  * __audit_log_capset - store information about the arguments to the capset syscall
2333  * @new: the new credentials
2334  * @old: the old (current) credentials
2335  *
2336  * Record the arguments userspace sent to sys_capset for later printing by the
2337  * audit system if applicable
2338  */
2339 void __audit_log_capset(const struct cred *new, const struct cred *old)
2340 {
2341 	struct audit_context *context = current->audit_context;
2342 	context->capset.pid = task_pid_nr(current);
2343 	context->capset.cap.effective   = new->cap_effective;
2344 	context->capset.cap.inheritable = new->cap_effective;
2345 	context->capset.cap.permitted   = new->cap_permitted;
2346 	context->type = AUDIT_CAPSET;
2347 }
2348 
2349 void __audit_mmap_fd(int fd, int flags)
2350 {
2351 	struct audit_context *context = current->audit_context;
2352 	context->mmap.fd = fd;
2353 	context->mmap.flags = flags;
2354 	context->type = AUDIT_MMAP;
2355 }
2356 
2357 static void audit_log_task(struct audit_buffer *ab)
2358 {
2359 	kuid_t auid, uid;
2360 	kgid_t gid;
2361 	unsigned int sessionid;
2362 	char comm[sizeof(current->comm)];
2363 
2364 	auid = audit_get_loginuid(current);
2365 	sessionid = audit_get_sessionid(current);
2366 	current_uid_gid(&uid, &gid);
2367 
2368 	audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2369 			 from_kuid(&init_user_ns, auid),
2370 			 from_kuid(&init_user_ns, uid),
2371 			 from_kgid(&init_user_ns, gid),
2372 			 sessionid);
2373 	audit_log_task_context(ab);
2374 	audit_log_format(ab, " pid=%d comm=", task_pid_nr(current));
2375 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2376 	audit_log_d_path_exe(ab, current->mm);
2377 }
2378 
2379 /**
2380  * audit_core_dumps - record information about processes that end abnormally
2381  * @signr: signal value
2382  *
2383  * If a process ends with a core dump, something fishy is going on and we
2384  * should record the event for investigation.
2385  */
2386 void audit_core_dumps(long signr)
2387 {
2388 	struct audit_buffer *ab;
2389 
2390 	if (!audit_enabled)
2391 		return;
2392 
2393 	if (signr == SIGQUIT)	/* don't care for those */
2394 		return;
2395 
2396 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2397 	if (unlikely(!ab))
2398 		return;
2399 	audit_log_task(ab);
2400 	audit_log_format(ab, " sig=%ld", signr);
2401 	audit_log_end(ab);
2402 }
2403 
2404 void __audit_seccomp(unsigned long syscall, long signr, int code)
2405 {
2406 	struct audit_buffer *ab;
2407 
2408 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_SECCOMP);
2409 	if (unlikely(!ab))
2410 		return;
2411 	audit_log_task(ab);
2412 	audit_log_format(ab, " sig=%ld arch=%x syscall=%ld compat=%d ip=0x%lx code=0x%x",
2413 			 signr, syscall_get_arch(), syscall, is_compat_task(),
2414 			 KSTK_EIP(current), code);
2415 	audit_log_end(ab);
2416 }
2417 
2418 struct list_head *audit_killed_trees(void)
2419 {
2420 	struct audit_context *ctx = current->audit_context;
2421 	if (likely(!ctx || !ctx->in_syscall))
2422 		return NULL;
2423 	return &ctx->killed_trees;
2424 }
2425