xref: /openbmc/linux/kernel/audit.c (revision 5b4cb650)
1 /* audit.c -- Auditing support
2  * Gateway between the kernel (e.g., selinux) and the user-space audit daemon.
3  * System-call specific features have moved to auditsc.c
4  *
5  * Copyright 2003-2007 Red Hat Inc., Durham, North Carolina.
6  * All Rights Reserved.
7  *
8  * This program is free software; you can redistribute it and/or modify
9  * it under the terms of the GNU General Public License as published by
10  * the Free Software Foundation; either version 2 of the License, or
11  * (at your option) any later version.
12  *
13  * This program is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16  * GNU General Public License for more details.
17  *
18  * You should have received a copy of the GNU General Public License
19  * along with this program; if not, write to the Free Software
20  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
21  *
22  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
23  *
24  * Goals: 1) Integrate fully with Security Modules.
25  *	  2) Minimal run-time overhead:
26  *	     a) Minimal when syscall auditing is disabled (audit_enable=0).
27  *	     b) Small when syscall auditing is enabled and no audit record
28  *		is generated (defer as much work as possible to record
29  *		generation time):
30  *		i) context is allocated,
31  *		ii) names from getname are stored without a copy, and
32  *		iii) inode information stored from path_lookup.
33  *	  3) Ability to disable syscall auditing at boot time (audit=0).
34  *	  4) Usable by other parts of the kernel (if audit_log* is called,
35  *	     then a syscall record will be generated automatically for the
36  *	     current syscall).
37  *	  5) Netlink interface to user-space.
38  *	  6) Support low-overhead kernel-based filtering to minimize the
39  *	     information that must be passed to user-space.
40  *
41  * Audit userspace, documentation, tests, and bug/issue trackers:
42  * 	https://github.com/linux-audit
43  */
44 
45 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
46 
47 #include <linux/file.h>
48 #include <linux/init.h>
49 #include <linux/types.h>
50 #include <linux/atomic.h>
51 #include <linux/mm.h>
52 #include <linux/export.h>
53 #include <linux/slab.h>
54 #include <linux/err.h>
55 #include <linux/kthread.h>
56 #include <linux/kernel.h>
57 #include <linux/syscalls.h>
58 #include <linux/spinlock.h>
59 #include <linux/rcupdate.h>
60 #include <linux/mutex.h>
61 #include <linux/gfp.h>
62 #include <linux/pid.h>
63 
64 #include <linux/audit.h>
65 
66 #include <net/sock.h>
67 #include <net/netlink.h>
68 #include <linux/skbuff.h>
69 #ifdef CONFIG_SECURITY
70 #include <linux/security.h>
71 #endif
72 #include <linux/freezer.h>
73 #include <linux/pid_namespace.h>
74 #include <net/netns/generic.h>
75 
76 #include "audit.h"
77 
78 /* No auditing will take place until audit_initialized == AUDIT_INITIALIZED.
79  * (Initialization happens after skb_init is called.) */
80 #define AUDIT_DISABLED		-1
81 #define AUDIT_UNINITIALIZED	0
82 #define AUDIT_INITIALIZED	1
83 static int	audit_initialized;
84 
85 u32		audit_enabled = AUDIT_OFF;
86 bool		audit_ever_enabled = !!AUDIT_OFF;
87 
88 EXPORT_SYMBOL_GPL(audit_enabled);
89 
90 /* Default state when kernel boots without any parameters. */
91 static u32	audit_default = AUDIT_OFF;
92 
93 /* If auditing cannot proceed, audit_failure selects what happens. */
94 static u32	audit_failure = AUDIT_FAIL_PRINTK;
95 
96 /* private audit network namespace index */
97 static unsigned int audit_net_id;
98 
99 /**
100  * struct audit_net - audit private network namespace data
101  * @sk: communication socket
102  */
103 struct audit_net {
104 	struct sock *sk;
105 };
106 
107 /**
108  * struct auditd_connection - kernel/auditd connection state
109  * @pid: auditd PID
110  * @portid: netlink portid
111  * @net: the associated network namespace
112  * @rcu: RCU head
113  *
114  * Description:
115  * This struct is RCU protected; you must either hold the RCU lock for reading
116  * or the associated spinlock for writing.
117  */
118 static struct auditd_connection {
119 	struct pid *pid;
120 	u32 portid;
121 	struct net *net;
122 	struct rcu_head rcu;
123 } *auditd_conn = NULL;
124 static DEFINE_SPINLOCK(auditd_conn_lock);
125 
126 /* If audit_rate_limit is non-zero, limit the rate of sending audit records
127  * to that number per second.  This prevents DoS attacks, but results in
128  * audit records being dropped. */
129 static u32	audit_rate_limit;
130 
131 /* Number of outstanding audit_buffers allowed.
132  * When set to zero, this means unlimited. */
133 static u32	audit_backlog_limit = 64;
134 #define AUDIT_BACKLOG_WAIT_TIME (60 * HZ)
135 static u32	audit_backlog_wait_time = AUDIT_BACKLOG_WAIT_TIME;
136 
137 /* The identity of the user shutting down the audit system. */
138 kuid_t		audit_sig_uid = INVALID_UID;
139 pid_t		audit_sig_pid = -1;
140 u32		audit_sig_sid = 0;
141 
142 /* Records can be lost in several ways:
143    0) [suppressed in audit_alloc]
144    1) out of memory in audit_log_start [kmalloc of struct audit_buffer]
145    2) out of memory in audit_log_move [alloc_skb]
146    3) suppressed due to audit_rate_limit
147    4) suppressed due to audit_backlog_limit
148 */
149 static atomic_t	audit_lost = ATOMIC_INIT(0);
150 
151 /* Hash for inode-based rules */
152 struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
153 
154 static struct kmem_cache *audit_buffer_cache;
155 
156 /* queue msgs to send via kauditd_task */
157 static struct sk_buff_head audit_queue;
158 /* queue msgs due to temporary unicast send problems */
159 static struct sk_buff_head audit_retry_queue;
160 /* queue msgs waiting for new auditd connection */
161 static struct sk_buff_head audit_hold_queue;
162 
163 /* queue servicing thread */
164 static struct task_struct *kauditd_task;
165 static DECLARE_WAIT_QUEUE_HEAD(kauditd_wait);
166 
167 /* waitqueue for callers who are blocked on the audit backlog */
168 static DECLARE_WAIT_QUEUE_HEAD(audit_backlog_wait);
169 
170 static struct audit_features af = {.vers = AUDIT_FEATURE_VERSION,
171 				   .mask = -1,
172 				   .features = 0,
173 				   .lock = 0,};
174 
175 static char *audit_feature_names[2] = {
176 	"only_unset_loginuid",
177 	"loginuid_immutable",
178 };
179 
180 /**
181  * struct audit_ctl_mutex - serialize requests from userspace
182  * @lock: the mutex used for locking
183  * @owner: the task which owns the lock
184  *
185  * Description:
186  * This is the lock struct used to ensure we only process userspace requests
187  * in an orderly fashion.  We can't simply use a mutex/lock here because we
188  * need to track lock ownership so we don't end up blocking the lock owner in
189  * audit_log_start() or similar.
190  */
191 static struct audit_ctl_mutex {
192 	struct mutex lock;
193 	void *owner;
194 } audit_cmd_mutex;
195 
196 /* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
197  * audit records.  Since printk uses a 1024 byte buffer, this buffer
198  * should be at least that large. */
199 #define AUDIT_BUFSIZ 1024
200 
201 /* The audit_buffer is used when formatting an audit record.  The caller
202  * locks briefly to get the record off the freelist or to allocate the
203  * buffer, and locks briefly to send the buffer to the netlink layer or
204  * to place it on a transmit queue.  Multiple audit_buffers can be in
205  * use simultaneously. */
206 struct audit_buffer {
207 	struct sk_buff       *skb;	/* formatted skb ready to send */
208 	struct audit_context *ctx;	/* NULL or associated context */
209 	gfp_t		     gfp_mask;
210 };
211 
212 struct audit_reply {
213 	__u32 portid;
214 	struct net *net;
215 	struct sk_buff *skb;
216 };
217 
218 /**
219  * auditd_test_task - Check to see if a given task is an audit daemon
220  * @task: the task to check
221  *
222  * Description:
223  * Return 1 if the task is a registered audit daemon, 0 otherwise.
224  */
225 int auditd_test_task(struct task_struct *task)
226 {
227 	int rc;
228 	struct auditd_connection *ac;
229 
230 	rcu_read_lock();
231 	ac = rcu_dereference(auditd_conn);
232 	rc = (ac && ac->pid == task_tgid(task) ? 1 : 0);
233 	rcu_read_unlock();
234 
235 	return rc;
236 }
237 
238 /**
239  * audit_ctl_lock - Take the audit control lock
240  */
241 void audit_ctl_lock(void)
242 {
243 	mutex_lock(&audit_cmd_mutex.lock);
244 	audit_cmd_mutex.owner = current;
245 }
246 
247 /**
248  * audit_ctl_unlock - Drop the audit control lock
249  */
250 void audit_ctl_unlock(void)
251 {
252 	audit_cmd_mutex.owner = NULL;
253 	mutex_unlock(&audit_cmd_mutex.lock);
254 }
255 
256 /**
257  * audit_ctl_owner_current - Test to see if the current task owns the lock
258  *
259  * Description:
260  * Return true if the current task owns the audit control lock, false if it
261  * doesn't own the lock.
262  */
263 static bool audit_ctl_owner_current(void)
264 {
265 	return (current == audit_cmd_mutex.owner);
266 }
267 
268 /**
269  * auditd_pid_vnr - Return the auditd PID relative to the namespace
270  *
271  * Description:
272  * Returns the PID in relation to the namespace, 0 on failure.
273  */
274 static pid_t auditd_pid_vnr(void)
275 {
276 	pid_t pid;
277 	const struct auditd_connection *ac;
278 
279 	rcu_read_lock();
280 	ac = rcu_dereference(auditd_conn);
281 	if (!ac || !ac->pid)
282 		pid = 0;
283 	else
284 		pid = pid_vnr(ac->pid);
285 	rcu_read_unlock();
286 
287 	return pid;
288 }
289 
290 /**
291  * audit_get_sk - Return the audit socket for the given network namespace
292  * @net: the destination network namespace
293  *
294  * Description:
295  * Returns the sock pointer if valid, NULL otherwise.  The caller must ensure
296  * that a reference is held for the network namespace while the sock is in use.
297  */
298 static struct sock *audit_get_sk(const struct net *net)
299 {
300 	struct audit_net *aunet;
301 
302 	if (!net)
303 		return NULL;
304 
305 	aunet = net_generic(net, audit_net_id);
306 	return aunet->sk;
307 }
308 
309 void audit_panic(const char *message)
310 {
311 	switch (audit_failure) {
312 	case AUDIT_FAIL_SILENT:
313 		break;
314 	case AUDIT_FAIL_PRINTK:
315 		if (printk_ratelimit())
316 			pr_err("%s\n", message);
317 		break;
318 	case AUDIT_FAIL_PANIC:
319 		panic("audit: %s\n", message);
320 		break;
321 	}
322 }
323 
324 static inline int audit_rate_check(void)
325 {
326 	static unsigned long	last_check = 0;
327 	static int		messages   = 0;
328 	static DEFINE_SPINLOCK(lock);
329 	unsigned long		flags;
330 	unsigned long		now;
331 	unsigned long		elapsed;
332 	int			retval	   = 0;
333 
334 	if (!audit_rate_limit) return 1;
335 
336 	spin_lock_irqsave(&lock, flags);
337 	if (++messages < audit_rate_limit) {
338 		retval = 1;
339 	} else {
340 		now     = jiffies;
341 		elapsed = now - last_check;
342 		if (elapsed > HZ) {
343 			last_check = now;
344 			messages   = 0;
345 			retval     = 1;
346 		}
347 	}
348 	spin_unlock_irqrestore(&lock, flags);
349 
350 	return retval;
351 }
352 
353 /**
354  * audit_log_lost - conditionally log lost audit message event
355  * @message: the message stating reason for lost audit message
356  *
357  * Emit at least 1 message per second, even if audit_rate_check is
358  * throttling.
359  * Always increment the lost messages counter.
360 */
361 void audit_log_lost(const char *message)
362 {
363 	static unsigned long	last_msg = 0;
364 	static DEFINE_SPINLOCK(lock);
365 	unsigned long		flags;
366 	unsigned long		now;
367 	int			print;
368 
369 	atomic_inc(&audit_lost);
370 
371 	print = (audit_failure == AUDIT_FAIL_PANIC || !audit_rate_limit);
372 
373 	if (!print) {
374 		spin_lock_irqsave(&lock, flags);
375 		now = jiffies;
376 		if (now - last_msg > HZ) {
377 			print = 1;
378 			last_msg = now;
379 		}
380 		spin_unlock_irqrestore(&lock, flags);
381 	}
382 
383 	if (print) {
384 		if (printk_ratelimit())
385 			pr_warn("audit_lost=%u audit_rate_limit=%u audit_backlog_limit=%u\n",
386 				atomic_read(&audit_lost),
387 				audit_rate_limit,
388 				audit_backlog_limit);
389 		audit_panic(message);
390 	}
391 }
392 
393 static int audit_log_config_change(char *function_name, u32 new, u32 old,
394 				   int allow_changes)
395 {
396 	struct audit_buffer *ab;
397 	int rc = 0;
398 
399 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE);
400 	if (unlikely(!ab))
401 		return rc;
402 	audit_log_format(ab, "%s=%u old=%u ", function_name, new, old);
403 	audit_log_session_info(ab);
404 	rc = audit_log_task_context(ab);
405 	if (rc)
406 		allow_changes = 0; /* Something weird, deny request */
407 	audit_log_format(ab, " res=%d", allow_changes);
408 	audit_log_end(ab);
409 	return rc;
410 }
411 
412 static int audit_do_config_change(char *function_name, u32 *to_change, u32 new)
413 {
414 	int allow_changes, rc = 0;
415 	u32 old = *to_change;
416 
417 	/* check if we are locked */
418 	if (audit_enabled == AUDIT_LOCKED)
419 		allow_changes = 0;
420 	else
421 		allow_changes = 1;
422 
423 	if (audit_enabled != AUDIT_OFF) {
424 		rc = audit_log_config_change(function_name, new, old, allow_changes);
425 		if (rc)
426 			allow_changes = 0;
427 	}
428 
429 	/* If we are allowed, make the change */
430 	if (allow_changes == 1)
431 		*to_change = new;
432 	/* Not allowed, update reason */
433 	else if (rc == 0)
434 		rc = -EPERM;
435 	return rc;
436 }
437 
438 static int audit_set_rate_limit(u32 limit)
439 {
440 	return audit_do_config_change("audit_rate_limit", &audit_rate_limit, limit);
441 }
442 
443 static int audit_set_backlog_limit(u32 limit)
444 {
445 	return audit_do_config_change("audit_backlog_limit", &audit_backlog_limit, limit);
446 }
447 
448 static int audit_set_backlog_wait_time(u32 timeout)
449 {
450 	return audit_do_config_change("audit_backlog_wait_time",
451 				      &audit_backlog_wait_time, timeout);
452 }
453 
454 static int audit_set_enabled(u32 state)
455 {
456 	int rc;
457 	if (state > AUDIT_LOCKED)
458 		return -EINVAL;
459 
460 	rc =  audit_do_config_change("audit_enabled", &audit_enabled, state);
461 	if (!rc)
462 		audit_ever_enabled |= !!state;
463 
464 	return rc;
465 }
466 
467 static int audit_set_failure(u32 state)
468 {
469 	if (state != AUDIT_FAIL_SILENT
470 	    && state != AUDIT_FAIL_PRINTK
471 	    && state != AUDIT_FAIL_PANIC)
472 		return -EINVAL;
473 
474 	return audit_do_config_change("audit_failure", &audit_failure, state);
475 }
476 
477 /**
478  * auditd_conn_free - RCU helper to release an auditd connection struct
479  * @rcu: RCU head
480  *
481  * Description:
482  * Drop any references inside the auditd connection tracking struct and free
483  * the memory.
484  */
485 static void auditd_conn_free(struct rcu_head *rcu)
486 {
487 	struct auditd_connection *ac;
488 
489 	ac = container_of(rcu, struct auditd_connection, rcu);
490 	put_pid(ac->pid);
491 	put_net(ac->net);
492 	kfree(ac);
493 }
494 
495 /**
496  * auditd_set - Set/Reset the auditd connection state
497  * @pid: auditd PID
498  * @portid: auditd netlink portid
499  * @net: auditd network namespace pointer
500  *
501  * Description:
502  * This function will obtain and drop network namespace references as
503  * necessary.  Returns zero on success, negative values on failure.
504  */
505 static int auditd_set(struct pid *pid, u32 portid, struct net *net)
506 {
507 	unsigned long flags;
508 	struct auditd_connection *ac_old, *ac_new;
509 
510 	if (!pid || !net)
511 		return -EINVAL;
512 
513 	ac_new = kzalloc(sizeof(*ac_new), GFP_KERNEL);
514 	if (!ac_new)
515 		return -ENOMEM;
516 	ac_new->pid = get_pid(pid);
517 	ac_new->portid = portid;
518 	ac_new->net = get_net(net);
519 
520 	spin_lock_irqsave(&auditd_conn_lock, flags);
521 	ac_old = rcu_dereference_protected(auditd_conn,
522 					   lockdep_is_held(&auditd_conn_lock));
523 	rcu_assign_pointer(auditd_conn, ac_new);
524 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
525 
526 	if (ac_old)
527 		call_rcu(&ac_old->rcu, auditd_conn_free);
528 
529 	return 0;
530 }
531 
532 /**
533  * kauditd_print_skb - Print the audit record to the ring buffer
534  * @skb: audit record
535  *
536  * Whatever the reason, this packet may not make it to the auditd connection
537  * so write it via printk so the information isn't completely lost.
538  */
539 static void kauditd_printk_skb(struct sk_buff *skb)
540 {
541 	struct nlmsghdr *nlh = nlmsg_hdr(skb);
542 	char *data = nlmsg_data(nlh);
543 
544 	if (nlh->nlmsg_type != AUDIT_EOE && printk_ratelimit())
545 		pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
546 }
547 
548 /**
549  * kauditd_rehold_skb - Handle a audit record send failure in the hold queue
550  * @skb: audit record
551  *
552  * Description:
553  * This should only be used by the kauditd_thread when it fails to flush the
554  * hold queue.
555  */
556 static void kauditd_rehold_skb(struct sk_buff *skb)
557 {
558 	/* put the record back in the queue at the same place */
559 	skb_queue_head(&audit_hold_queue, skb);
560 }
561 
562 /**
563  * kauditd_hold_skb - Queue an audit record, waiting for auditd
564  * @skb: audit record
565  *
566  * Description:
567  * Queue the audit record, waiting for an instance of auditd.  When this
568  * function is called we haven't given up yet on sending the record, but things
569  * are not looking good.  The first thing we want to do is try to write the
570  * record via printk and then see if we want to try and hold on to the record
571  * and queue it, if we have room.  If we want to hold on to the record, but we
572  * don't have room, record a record lost message.
573  */
574 static void kauditd_hold_skb(struct sk_buff *skb)
575 {
576 	/* at this point it is uncertain if we will ever send this to auditd so
577 	 * try to send the message via printk before we go any further */
578 	kauditd_printk_skb(skb);
579 
580 	/* can we just silently drop the message? */
581 	if (!audit_default) {
582 		kfree_skb(skb);
583 		return;
584 	}
585 
586 	/* if we have room, queue the message */
587 	if (!audit_backlog_limit ||
588 	    skb_queue_len(&audit_hold_queue) < audit_backlog_limit) {
589 		skb_queue_tail(&audit_hold_queue, skb);
590 		return;
591 	}
592 
593 	/* we have no other options - drop the message */
594 	audit_log_lost("kauditd hold queue overflow");
595 	kfree_skb(skb);
596 }
597 
598 /**
599  * kauditd_retry_skb - Queue an audit record, attempt to send again to auditd
600  * @skb: audit record
601  *
602  * Description:
603  * Not as serious as kauditd_hold_skb() as we still have a connected auditd,
604  * but for some reason we are having problems sending it audit records so
605  * queue the given record and attempt to resend.
606  */
607 static void kauditd_retry_skb(struct sk_buff *skb)
608 {
609 	/* NOTE: because records should only live in the retry queue for a
610 	 * short period of time, before either being sent or moved to the hold
611 	 * queue, we don't currently enforce a limit on this queue */
612 	skb_queue_tail(&audit_retry_queue, skb);
613 }
614 
615 /**
616  * auditd_reset - Disconnect the auditd connection
617  * @ac: auditd connection state
618  *
619  * Description:
620  * Break the auditd/kauditd connection and move all the queued records into the
621  * hold queue in case auditd reconnects.  It is important to note that the @ac
622  * pointer should never be dereferenced inside this function as it may be NULL
623  * or invalid, you can only compare the memory address!  If @ac is NULL then
624  * the connection will always be reset.
625  */
626 static void auditd_reset(const struct auditd_connection *ac)
627 {
628 	unsigned long flags;
629 	struct sk_buff *skb;
630 	struct auditd_connection *ac_old;
631 
632 	/* if it isn't already broken, break the connection */
633 	spin_lock_irqsave(&auditd_conn_lock, flags);
634 	ac_old = rcu_dereference_protected(auditd_conn,
635 					   lockdep_is_held(&auditd_conn_lock));
636 	if (ac && ac != ac_old) {
637 		/* someone already registered a new auditd connection */
638 		spin_unlock_irqrestore(&auditd_conn_lock, flags);
639 		return;
640 	}
641 	rcu_assign_pointer(auditd_conn, NULL);
642 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
643 
644 	if (ac_old)
645 		call_rcu(&ac_old->rcu, auditd_conn_free);
646 
647 	/* flush the retry queue to the hold queue, but don't touch the main
648 	 * queue since we need to process that normally for multicast */
649 	while ((skb = skb_dequeue(&audit_retry_queue)))
650 		kauditd_hold_skb(skb);
651 }
652 
653 /**
654  * auditd_send_unicast_skb - Send a record via unicast to auditd
655  * @skb: audit record
656  *
657  * Description:
658  * Send a skb to the audit daemon, returns positive/zero values on success and
659  * negative values on failure; in all cases the skb will be consumed by this
660  * function.  If the send results in -ECONNREFUSED the connection with auditd
661  * will be reset.  This function may sleep so callers should not hold any locks
662  * where this would cause a problem.
663  */
664 static int auditd_send_unicast_skb(struct sk_buff *skb)
665 {
666 	int rc;
667 	u32 portid;
668 	struct net *net;
669 	struct sock *sk;
670 	struct auditd_connection *ac;
671 
672 	/* NOTE: we can't call netlink_unicast while in the RCU section so
673 	 *       take a reference to the network namespace and grab local
674 	 *       copies of the namespace, the sock, and the portid; the
675 	 *       namespace and sock aren't going to go away while we hold a
676 	 *       reference and if the portid does become invalid after the RCU
677 	 *       section netlink_unicast() should safely return an error */
678 
679 	rcu_read_lock();
680 	ac = rcu_dereference(auditd_conn);
681 	if (!ac) {
682 		rcu_read_unlock();
683 		kfree_skb(skb);
684 		rc = -ECONNREFUSED;
685 		goto err;
686 	}
687 	net = get_net(ac->net);
688 	sk = audit_get_sk(net);
689 	portid = ac->portid;
690 	rcu_read_unlock();
691 
692 	rc = netlink_unicast(sk, skb, portid, 0);
693 	put_net(net);
694 	if (rc < 0)
695 		goto err;
696 
697 	return rc;
698 
699 err:
700 	if (ac && rc == -ECONNREFUSED)
701 		auditd_reset(ac);
702 	return rc;
703 }
704 
705 /**
706  * kauditd_send_queue - Helper for kauditd_thread to flush skb queues
707  * @sk: the sending sock
708  * @portid: the netlink destination
709  * @queue: the skb queue to process
710  * @retry_limit: limit on number of netlink unicast failures
711  * @skb_hook: per-skb hook for additional processing
712  * @err_hook: hook called if the skb fails the netlink unicast send
713  *
714  * Description:
715  * Run through the given queue and attempt to send the audit records to auditd,
716  * returns zero on success, negative values on failure.  It is up to the caller
717  * to ensure that the @sk is valid for the duration of this function.
718  *
719  */
720 static int kauditd_send_queue(struct sock *sk, u32 portid,
721 			      struct sk_buff_head *queue,
722 			      unsigned int retry_limit,
723 			      void (*skb_hook)(struct sk_buff *skb),
724 			      void (*err_hook)(struct sk_buff *skb))
725 {
726 	int rc = 0;
727 	struct sk_buff *skb;
728 	static unsigned int failed = 0;
729 
730 	/* NOTE: kauditd_thread takes care of all our locking, we just use
731 	 *       the netlink info passed to us (e.g. sk and portid) */
732 
733 	while ((skb = skb_dequeue(queue))) {
734 		/* call the skb_hook for each skb we touch */
735 		if (skb_hook)
736 			(*skb_hook)(skb);
737 
738 		/* can we send to anyone via unicast? */
739 		if (!sk) {
740 			if (err_hook)
741 				(*err_hook)(skb);
742 			continue;
743 		}
744 
745 		/* grab an extra skb reference in case of error */
746 		skb_get(skb);
747 		rc = netlink_unicast(sk, skb, portid, 0);
748 		if (rc < 0) {
749 			/* fatal failure for our queue flush attempt? */
750 			if (++failed >= retry_limit ||
751 			    rc == -ECONNREFUSED || rc == -EPERM) {
752 				/* yes - error processing for the queue */
753 				sk = NULL;
754 				if (err_hook)
755 					(*err_hook)(skb);
756 				if (!skb_hook)
757 					goto out;
758 				/* keep processing with the skb_hook */
759 				continue;
760 			} else
761 				/* no - requeue to preserve ordering */
762 				skb_queue_head(queue, skb);
763 		} else {
764 			/* it worked - drop the extra reference and continue */
765 			consume_skb(skb);
766 			failed = 0;
767 		}
768 	}
769 
770 out:
771 	return (rc >= 0 ? 0 : rc);
772 }
773 
774 /*
775  * kauditd_send_multicast_skb - Send a record to any multicast listeners
776  * @skb: audit record
777  *
778  * Description:
779  * Write a multicast message to anyone listening in the initial network
780  * namespace.  This function doesn't consume an skb as might be expected since
781  * it has to copy it anyways.
782  */
783 static void kauditd_send_multicast_skb(struct sk_buff *skb)
784 {
785 	struct sk_buff *copy;
786 	struct sock *sock = audit_get_sk(&init_net);
787 	struct nlmsghdr *nlh;
788 
789 	/* NOTE: we are not taking an additional reference for init_net since
790 	 *       we don't have to worry about it going away */
791 
792 	if (!netlink_has_listeners(sock, AUDIT_NLGRP_READLOG))
793 		return;
794 
795 	/*
796 	 * The seemingly wasteful skb_copy() rather than bumping the refcount
797 	 * using skb_get() is necessary because non-standard mods are made to
798 	 * the skb by the original kaudit unicast socket send routine.  The
799 	 * existing auditd daemon assumes this breakage.  Fixing this would
800 	 * require co-ordinating a change in the established protocol between
801 	 * the kaudit kernel subsystem and the auditd userspace code.  There is
802 	 * no reason for new multicast clients to continue with this
803 	 * non-compliance.
804 	 */
805 	copy = skb_copy(skb, GFP_KERNEL);
806 	if (!copy)
807 		return;
808 	nlh = nlmsg_hdr(copy);
809 	nlh->nlmsg_len = skb->len;
810 
811 	nlmsg_multicast(sock, copy, 0, AUDIT_NLGRP_READLOG, GFP_KERNEL);
812 }
813 
814 /**
815  * kauditd_thread - Worker thread to send audit records to userspace
816  * @dummy: unused
817  */
818 static int kauditd_thread(void *dummy)
819 {
820 	int rc;
821 	u32 portid = 0;
822 	struct net *net = NULL;
823 	struct sock *sk = NULL;
824 	struct auditd_connection *ac;
825 
826 #define UNICAST_RETRIES 5
827 
828 	set_freezable();
829 	while (!kthread_should_stop()) {
830 		/* NOTE: see the lock comments in auditd_send_unicast_skb() */
831 		rcu_read_lock();
832 		ac = rcu_dereference(auditd_conn);
833 		if (!ac) {
834 			rcu_read_unlock();
835 			goto main_queue;
836 		}
837 		net = get_net(ac->net);
838 		sk = audit_get_sk(net);
839 		portid = ac->portid;
840 		rcu_read_unlock();
841 
842 		/* attempt to flush the hold queue */
843 		rc = kauditd_send_queue(sk, portid,
844 					&audit_hold_queue, UNICAST_RETRIES,
845 					NULL, kauditd_rehold_skb);
846 		if (ac && rc < 0) {
847 			sk = NULL;
848 			auditd_reset(ac);
849 			goto main_queue;
850 		}
851 
852 		/* attempt to flush the retry queue */
853 		rc = kauditd_send_queue(sk, portid,
854 					&audit_retry_queue, UNICAST_RETRIES,
855 					NULL, kauditd_hold_skb);
856 		if (ac && rc < 0) {
857 			sk = NULL;
858 			auditd_reset(ac);
859 			goto main_queue;
860 		}
861 
862 main_queue:
863 		/* process the main queue - do the multicast send and attempt
864 		 * unicast, dump failed record sends to the retry queue; if
865 		 * sk == NULL due to previous failures we will just do the
866 		 * multicast send and move the record to the hold queue */
867 		rc = kauditd_send_queue(sk, portid, &audit_queue, 1,
868 					kauditd_send_multicast_skb,
869 					(sk ?
870 					 kauditd_retry_skb : kauditd_hold_skb));
871 		if (ac && rc < 0)
872 			auditd_reset(ac);
873 		sk = NULL;
874 
875 		/* drop our netns reference, no auditd sends past this line */
876 		if (net) {
877 			put_net(net);
878 			net = NULL;
879 		}
880 
881 		/* we have processed all the queues so wake everyone */
882 		wake_up(&audit_backlog_wait);
883 
884 		/* NOTE: we want to wake up if there is anything on the queue,
885 		 *       regardless of if an auditd is connected, as we need to
886 		 *       do the multicast send and rotate records from the
887 		 *       main queue to the retry/hold queues */
888 		wait_event_freezable(kauditd_wait,
889 				     (skb_queue_len(&audit_queue) ? 1 : 0));
890 	}
891 
892 	return 0;
893 }
894 
895 int audit_send_list(void *_dest)
896 {
897 	struct audit_netlink_list *dest = _dest;
898 	struct sk_buff *skb;
899 	struct sock *sk = audit_get_sk(dest->net);
900 
901 	/* wait for parent to finish and send an ACK */
902 	audit_ctl_lock();
903 	audit_ctl_unlock();
904 
905 	while ((skb = __skb_dequeue(&dest->q)) != NULL)
906 		netlink_unicast(sk, skb, dest->portid, 0);
907 
908 	put_net(dest->net);
909 	kfree(dest);
910 
911 	return 0;
912 }
913 
914 struct sk_buff *audit_make_reply(int seq, int type, int done,
915 				 int multi, const void *payload, int size)
916 {
917 	struct sk_buff	*skb;
918 	struct nlmsghdr	*nlh;
919 	void		*data;
920 	int		flags = multi ? NLM_F_MULTI : 0;
921 	int		t     = done  ? NLMSG_DONE  : type;
922 
923 	skb = nlmsg_new(size, GFP_KERNEL);
924 	if (!skb)
925 		return NULL;
926 
927 	nlh	= nlmsg_put(skb, 0, seq, t, size, flags);
928 	if (!nlh)
929 		goto out_kfree_skb;
930 	data = nlmsg_data(nlh);
931 	memcpy(data, payload, size);
932 	return skb;
933 
934 out_kfree_skb:
935 	kfree_skb(skb);
936 	return NULL;
937 }
938 
939 static int audit_send_reply_thread(void *arg)
940 {
941 	struct audit_reply *reply = (struct audit_reply *)arg;
942 	struct sock *sk = audit_get_sk(reply->net);
943 
944 	audit_ctl_lock();
945 	audit_ctl_unlock();
946 
947 	/* Ignore failure. It'll only happen if the sender goes away,
948 	   because our timeout is set to infinite. */
949 	netlink_unicast(sk, reply->skb, reply->portid, 0);
950 	put_net(reply->net);
951 	kfree(reply);
952 	return 0;
953 }
954 
955 /**
956  * audit_send_reply - send an audit reply message via netlink
957  * @request_skb: skb of request we are replying to (used to target the reply)
958  * @seq: sequence number
959  * @type: audit message type
960  * @done: done (last) flag
961  * @multi: multi-part message flag
962  * @payload: payload data
963  * @size: payload size
964  *
965  * Allocates an skb, builds the netlink message, and sends it to the port id.
966  * No failure notifications.
967  */
968 static void audit_send_reply(struct sk_buff *request_skb, int seq, int type, int done,
969 			     int multi, const void *payload, int size)
970 {
971 	struct net *net = sock_net(NETLINK_CB(request_skb).sk);
972 	struct sk_buff *skb;
973 	struct task_struct *tsk;
974 	struct audit_reply *reply = kmalloc(sizeof(struct audit_reply),
975 					    GFP_KERNEL);
976 
977 	if (!reply)
978 		return;
979 
980 	skb = audit_make_reply(seq, type, done, multi, payload, size);
981 	if (!skb)
982 		goto out;
983 
984 	reply->net = get_net(net);
985 	reply->portid = NETLINK_CB(request_skb).portid;
986 	reply->skb = skb;
987 
988 	tsk = kthread_run(audit_send_reply_thread, reply, "audit_send_reply");
989 	if (!IS_ERR(tsk))
990 		return;
991 	kfree_skb(skb);
992 out:
993 	kfree(reply);
994 }
995 
996 /*
997  * Check for appropriate CAP_AUDIT_ capabilities on incoming audit
998  * control messages.
999  */
1000 static int audit_netlink_ok(struct sk_buff *skb, u16 msg_type)
1001 {
1002 	int err = 0;
1003 
1004 	/* Only support initial user namespace for now. */
1005 	/*
1006 	 * We return ECONNREFUSED because it tricks userspace into thinking
1007 	 * that audit was not configured into the kernel.  Lots of users
1008 	 * configure their PAM stack (because that's what the distro does)
1009 	 * to reject login if unable to send messages to audit.  If we return
1010 	 * ECONNREFUSED the PAM stack thinks the kernel does not have audit
1011 	 * configured in and will let login proceed.  If we return EPERM
1012 	 * userspace will reject all logins.  This should be removed when we
1013 	 * support non init namespaces!!
1014 	 */
1015 	if (current_user_ns() != &init_user_ns)
1016 		return -ECONNREFUSED;
1017 
1018 	switch (msg_type) {
1019 	case AUDIT_LIST:
1020 	case AUDIT_ADD:
1021 	case AUDIT_DEL:
1022 		return -EOPNOTSUPP;
1023 	case AUDIT_GET:
1024 	case AUDIT_SET:
1025 	case AUDIT_GET_FEATURE:
1026 	case AUDIT_SET_FEATURE:
1027 	case AUDIT_LIST_RULES:
1028 	case AUDIT_ADD_RULE:
1029 	case AUDIT_DEL_RULE:
1030 	case AUDIT_SIGNAL_INFO:
1031 	case AUDIT_TTY_GET:
1032 	case AUDIT_TTY_SET:
1033 	case AUDIT_TRIM:
1034 	case AUDIT_MAKE_EQUIV:
1035 		/* Only support auditd and auditctl in initial pid namespace
1036 		 * for now. */
1037 		if (task_active_pid_ns(current) != &init_pid_ns)
1038 			return -EPERM;
1039 
1040 		if (!netlink_capable(skb, CAP_AUDIT_CONTROL))
1041 			err = -EPERM;
1042 		break;
1043 	case AUDIT_USER:
1044 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1045 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1046 		if (!netlink_capable(skb, CAP_AUDIT_WRITE))
1047 			err = -EPERM;
1048 		break;
1049 	default:  /* bad msg */
1050 		err = -EINVAL;
1051 	}
1052 
1053 	return err;
1054 }
1055 
1056 static void audit_log_common_recv_msg(struct audit_buffer **ab, u16 msg_type)
1057 {
1058 	uid_t uid = from_kuid(&init_user_ns, current_uid());
1059 	pid_t pid = task_tgid_nr(current);
1060 
1061 	if (!audit_enabled && msg_type != AUDIT_USER_AVC) {
1062 		*ab = NULL;
1063 		return;
1064 	}
1065 
1066 	*ab = audit_log_start(NULL, GFP_KERNEL, msg_type);
1067 	if (unlikely(!*ab))
1068 		return;
1069 	audit_log_format(*ab, "pid=%d uid=%u ", pid, uid);
1070 	audit_log_session_info(*ab);
1071 	audit_log_task_context(*ab);
1072 }
1073 
1074 int is_audit_feature_set(int i)
1075 {
1076 	return af.features & AUDIT_FEATURE_TO_MASK(i);
1077 }
1078 
1079 
1080 static int audit_get_feature(struct sk_buff *skb)
1081 {
1082 	u32 seq;
1083 
1084 	seq = nlmsg_hdr(skb)->nlmsg_seq;
1085 
1086 	audit_send_reply(skb, seq, AUDIT_GET_FEATURE, 0, 0, &af, sizeof(af));
1087 
1088 	return 0;
1089 }
1090 
1091 static void audit_log_feature_change(int which, u32 old_feature, u32 new_feature,
1092 				     u32 old_lock, u32 new_lock, int res)
1093 {
1094 	struct audit_buffer *ab;
1095 
1096 	if (audit_enabled == AUDIT_OFF)
1097 		return;
1098 
1099 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_FEATURE_CHANGE);
1100 	if (!ab)
1101 		return;
1102 	audit_log_task_info(ab);
1103 	audit_log_format(ab, " feature=%s old=%u new=%u old_lock=%u new_lock=%u res=%d",
1104 			 audit_feature_names[which], !!old_feature, !!new_feature,
1105 			 !!old_lock, !!new_lock, res);
1106 	audit_log_end(ab);
1107 }
1108 
1109 static int audit_set_feature(struct sk_buff *skb)
1110 {
1111 	struct audit_features *uaf;
1112 	int i;
1113 
1114 	BUILD_BUG_ON(AUDIT_LAST_FEATURE + 1 > ARRAY_SIZE(audit_feature_names));
1115 	uaf = nlmsg_data(nlmsg_hdr(skb));
1116 
1117 	/* if there is ever a version 2 we should handle that here */
1118 
1119 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1120 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1121 		u32 old_feature, new_feature, old_lock, new_lock;
1122 
1123 		/* if we are not changing this feature, move along */
1124 		if (!(feature & uaf->mask))
1125 			continue;
1126 
1127 		old_feature = af.features & feature;
1128 		new_feature = uaf->features & feature;
1129 		new_lock = (uaf->lock | af.lock) & feature;
1130 		old_lock = af.lock & feature;
1131 
1132 		/* are we changing a locked feature? */
1133 		if (old_lock && (new_feature != old_feature)) {
1134 			audit_log_feature_change(i, old_feature, new_feature,
1135 						 old_lock, new_lock, 0);
1136 			return -EPERM;
1137 		}
1138 	}
1139 	/* nothing invalid, do the changes */
1140 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1141 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1142 		u32 old_feature, new_feature, old_lock, new_lock;
1143 
1144 		/* if we are not changing this feature, move along */
1145 		if (!(feature & uaf->mask))
1146 			continue;
1147 
1148 		old_feature = af.features & feature;
1149 		new_feature = uaf->features & feature;
1150 		old_lock = af.lock & feature;
1151 		new_lock = (uaf->lock | af.lock) & feature;
1152 
1153 		if (new_feature != old_feature)
1154 			audit_log_feature_change(i, old_feature, new_feature,
1155 						 old_lock, new_lock, 1);
1156 
1157 		if (new_feature)
1158 			af.features |= feature;
1159 		else
1160 			af.features &= ~feature;
1161 		af.lock |= new_lock;
1162 	}
1163 
1164 	return 0;
1165 }
1166 
1167 static int audit_replace(struct pid *pid)
1168 {
1169 	pid_t pvnr;
1170 	struct sk_buff *skb;
1171 
1172 	pvnr = pid_vnr(pid);
1173 	skb = audit_make_reply(0, AUDIT_REPLACE, 0, 0, &pvnr, sizeof(pvnr));
1174 	if (!skb)
1175 		return -ENOMEM;
1176 	return auditd_send_unicast_skb(skb);
1177 }
1178 
1179 static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1180 {
1181 	u32			seq;
1182 	void			*data;
1183 	int			err;
1184 	struct audit_buffer	*ab;
1185 	u16			msg_type = nlh->nlmsg_type;
1186 	struct audit_sig_info   *sig_data;
1187 	char			*ctx = NULL;
1188 	u32			len;
1189 
1190 	err = audit_netlink_ok(skb, msg_type);
1191 	if (err)
1192 		return err;
1193 
1194 	seq  = nlh->nlmsg_seq;
1195 	data = nlmsg_data(nlh);
1196 
1197 	switch (msg_type) {
1198 	case AUDIT_GET: {
1199 		struct audit_status	s;
1200 		memset(&s, 0, sizeof(s));
1201 		s.enabled		= audit_enabled;
1202 		s.failure		= audit_failure;
1203 		/* NOTE: use pid_vnr() so the PID is relative to the current
1204 		 *       namespace */
1205 		s.pid			= auditd_pid_vnr();
1206 		s.rate_limit		= audit_rate_limit;
1207 		s.backlog_limit		= audit_backlog_limit;
1208 		s.lost			= atomic_read(&audit_lost);
1209 		s.backlog		= skb_queue_len(&audit_queue);
1210 		s.feature_bitmap	= AUDIT_FEATURE_BITMAP_ALL;
1211 		s.backlog_wait_time	= audit_backlog_wait_time;
1212 		audit_send_reply(skb, seq, AUDIT_GET, 0, 0, &s, sizeof(s));
1213 		break;
1214 	}
1215 	case AUDIT_SET: {
1216 		struct audit_status	s;
1217 		memset(&s, 0, sizeof(s));
1218 		/* guard against past and future API changes */
1219 		memcpy(&s, data, min_t(size_t, sizeof(s), nlmsg_len(nlh)));
1220 		if (s.mask & AUDIT_STATUS_ENABLED) {
1221 			err = audit_set_enabled(s.enabled);
1222 			if (err < 0)
1223 				return err;
1224 		}
1225 		if (s.mask & AUDIT_STATUS_FAILURE) {
1226 			err = audit_set_failure(s.failure);
1227 			if (err < 0)
1228 				return err;
1229 		}
1230 		if (s.mask & AUDIT_STATUS_PID) {
1231 			/* NOTE: we are using the vnr PID functions below
1232 			 *       because the s.pid value is relative to the
1233 			 *       namespace of the caller; at present this
1234 			 *       doesn't matter much since you can really only
1235 			 *       run auditd from the initial pid namespace, but
1236 			 *       something to keep in mind if this changes */
1237 			pid_t new_pid = s.pid;
1238 			pid_t auditd_pid;
1239 			struct pid *req_pid = task_tgid(current);
1240 
1241 			/* Sanity check - PID values must match. Setting
1242 			 * pid to 0 is how auditd ends auditing. */
1243 			if (new_pid && (new_pid != pid_vnr(req_pid)))
1244 				return -EINVAL;
1245 
1246 			/* test the auditd connection */
1247 			audit_replace(req_pid);
1248 
1249 			auditd_pid = auditd_pid_vnr();
1250 			if (auditd_pid) {
1251 				/* replacing a healthy auditd is not allowed */
1252 				if (new_pid) {
1253 					audit_log_config_change("audit_pid",
1254 							new_pid, auditd_pid, 0);
1255 					return -EEXIST;
1256 				}
1257 				/* only current auditd can unregister itself */
1258 				if (pid_vnr(req_pid) != auditd_pid) {
1259 					audit_log_config_change("audit_pid",
1260 							new_pid, auditd_pid, 0);
1261 					return -EACCES;
1262 				}
1263 			}
1264 
1265 			if (new_pid) {
1266 				/* register a new auditd connection */
1267 				err = auditd_set(req_pid,
1268 						 NETLINK_CB(skb).portid,
1269 						 sock_net(NETLINK_CB(skb).sk));
1270 				if (audit_enabled != AUDIT_OFF)
1271 					audit_log_config_change("audit_pid",
1272 								new_pid,
1273 								auditd_pid,
1274 								err ? 0 : 1);
1275 				if (err)
1276 					return err;
1277 
1278 				/* try to process any backlog */
1279 				wake_up_interruptible(&kauditd_wait);
1280 			} else {
1281 				if (audit_enabled != AUDIT_OFF)
1282 					audit_log_config_change("audit_pid",
1283 								new_pid,
1284 								auditd_pid, 1);
1285 
1286 				/* unregister the auditd connection */
1287 				auditd_reset(NULL);
1288 			}
1289 		}
1290 		if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
1291 			err = audit_set_rate_limit(s.rate_limit);
1292 			if (err < 0)
1293 				return err;
1294 		}
1295 		if (s.mask & AUDIT_STATUS_BACKLOG_LIMIT) {
1296 			err = audit_set_backlog_limit(s.backlog_limit);
1297 			if (err < 0)
1298 				return err;
1299 		}
1300 		if (s.mask & AUDIT_STATUS_BACKLOG_WAIT_TIME) {
1301 			if (sizeof(s) > (size_t)nlh->nlmsg_len)
1302 				return -EINVAL;
1303 			if (s.backlog_wait_time > 10*AUDIT_BACKLOG_WAIT_TIME)
1304 				return -EINVAL;
1305 			err = audit_set_backlog_wait_time(s.backlog_wait_time);
1306 			if (err < 0)
1307 				return err;
1308 		}
1309 		if (s.mask == AUDIT_STATUS_LOST) {
1310 			u32 lost = atomic_xchg(&audit_lost, 0);
1311 
1312 			audit_log_config_change("lost", 0, lost, 1);
1313 			return lost;
1314 		}
1315 		break;
1316 	}
1317 	case AUDIT_GET_FEATURE:
1318 		err = audit_get_feature(skb);
1319 		if (err)
1320 			return err;
1321 		break;
1322 	case AUDIT_SET_FEATURE:
1323 		err = audit_set_feature(skb);
1324 		if (err)
1325 			return err;
1326 		break;
1327 	case AUDIT_USER:
1328 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1329 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1330 		if (!audit_enabled && msg_type != AUDIT_USER_AVC)
1331 			return 0;
1332 
1333 		err = audit_filter(msg_type, AUDIT_FILTER_USER);
1334 		if (err == 1) { /* match or error */
1335 			err = 0;
1336 			if (msg_type == AUDIT_USER_TTY) {
1337 				err = tty_audit_push();
1338 				if (err)
1339 					break;
1340 			}
1341 			audit_log_common_recv_msg(&ab, msg_type);
1342 			if (msg_type != AUDIT_USER_TTY)
1343 				audit_log_format(ab, " msg='%.*s'",
1344 						 AUDIT_MESSAGE_TEXT_MAX,
1345 						 (char *)data);
1346 			else {
1347 				int size;
1348 
1349 				audit_log_format(ab, " data=");
1350 				size = nlmsg_len(nlh);
1351 				if (size > 0 &&
1352 				    ((unsigned char *)data)[size - 1] == '\0')
1353 					size--;
1354 				audit_log_n_untrustedstring(ab, data, size);
1355 			}
1356 			audit_log_end(ab);
1357 		}
1358 		break;
1359 	case AUDIT_ADD_RULE:
1360 	case AUDIT_DEL_RULE:
1361 		if (nlmsg_len(nlh) < sizeof(struct audit_rule_data))
1362 			return -EINVAL;
1363 		if (audit_enabled == AUDIT_LOCKED) {
1364 			audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1365 			audit_log_format(ab, " audit_enabled=%d res=0", audit_enabled);
1366 			audit_log_end(ab);
1367 			return -EPERM;
1368 		}
1369 		err = audit_rule_change(msg_type, seq, data, nlmsg_len(nlh));
1370 		break;
1371 	case AUDIT_LIST_RULES:
1372 		err = audit_list_rules_send(skb, seq);
1373 		break;
1374 	case AUDIT_TRIM:
1375 		audit_trim_trees();
1376 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1377 		audit_log_format(ab, " op=trim res=1");
1378 		audit_log_end(ab);
1379 		break;
1380 	case AUDIT_MAKE_EQUIV: {
1381 		void *bufp = data;
1382 		u32 sizes[2];
1383 		size_t msglen = nlmsg_len(nlh);
1384 		char *old, *new;
1385 
1386 		err = -EINVAL;
1387 		if (msglen < 2 * sizeof(u32))
1388 			break;
1389 		memcpy(sizes, bufp, 2 * sizeof(u32));
1390 		bufp += 2 * sizeof(u32);
1391 		msglen -= 2 * sizeof(u32);
1392 		old = audit_unpack_string(&bufp, &msglen, sizes[0]);
1393 		if (IS_ERR(old)) {
1394 			err = PTR_ERR(old);
1395 			break;
1396 		}
1397 		new = audit_unpack_string(&bufp, &msglen, sizes[1]);
1398 		if (IS_ERR(new)) {
1399 			err = PTR_ERR(new);
1400 			kfree(old);
1401 			break;
1402 		}
1403 		/* OK, here comes... */
1404 		err = audit_tag_tree(old, new);
1405 
1406 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1407 
1408 		audit_log_format(ab, " op=make_equiv old=");
1409 		audit_log_untrustedstring(ab, old);
1410 		audit_log_format(ab, " new=");
1411 		audit_log_untrustedstring(ab, new);
1412 		audit_log_format(ab, " res=%d", !err);
1413 		audit_log_end(ab);
1414 		kfree(old);
1415 		kfree(new);
1416 		break;
1417 	}
1418 	case AUDIT_SIGNAL_INFO:
1419 		len = 0;
1420 		if (audit_sig_sid) {
1421 			err = security_secid_to_secctx(audit_sig_sid, &ctx, &len);
1422 			if (err)
1423 				return err;
1424 		}
1425 		sig_data = kmalloc(sizeof(*sig_data) + len, GFP_KERNEL);
1426 		if (!sig_data) {
1427 			if (audit_sig_sid)
1428 				security_release_secctx(ctx, len);
1429 			return -ENOMEM;
1430 		}
1431 		sig_data->uid = from_kuid(&init_user_ns, audit_sig_uid);
1432 		sig_data->pid = audit_sig_pid;
1433 		if (audit_sig_sid) {
1434 			memcpy(sig_data->ctx, ctx, len);
1435 			security_release_secctx(ctx, len);
1436 		}
1437 		audit_send_reply(skb, seq, AUDIT_SIGNAL_INFO, 0, 0,
1438 				 sig_data, sizeof(*sig_data) + len);
1439 		kfree(sig_data);
1440 		break;
1441 	case AUDIT_TTY_GET: {
1442 		struct audit_tty_status s;
1443 		unsigned int t;
1444 
1445 		t = READ_ONCE(current->signal->audit_tty);
1446 		s.enabled = t & AUDIT_TTY_ENABLE;
1447 		s.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1448 
1449 		audit_send_reply(skb, seq, AUDIT_TTY_GET, 0, 0, &s, sizeof(s));
1450 		break;
1451 	}
1452 	case AUDIT_TTY_SET: {
1453 		struct audit_tty_status s, old;
1454 		struct audit_buffer	*ab;
1455 		unsigned int t;
1456 
1457 		memset(&s, 0, sizeof(s));
1458 		/* guard against past and future API changes */
1459 		memcpy(&s, data, min_t(size_t, sizeof(s), nlmsg_len(nlh)));
1460 		/* check if new data is valid */
1461 		if ((s.enabled != 0 && s.enabled != 1) ||
1462 		    (s.log_passwd != 0 && s.log_passwd != 1))
1463 			err = -EINVAL;
1464 
1465 		if (err)
1466 			t = READ_ONCE(current->signal->audit_tty);
1467 		else {
1468 			t = s.enabled | (-s.log_passwd & AUDIT_TTY_LOG_PASSWD);
1469 			t = xchg(&current->signal->audit_tty, t);
1470 		}
1471 		old.enabled = t & AUDIT_TTY_ENABLE;
1472 		old.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1473 
1474 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1475 		audit_log_format(ab, " op=tty_set old-enabled=%d new-enabled=%d"
1476 				 " old-log_passwd=%d new-log_passwd=%d res=%d",
1477 				 old.enabled, s.enabled, old.log_passwd,
1478 				 s.log_passwd, !err);
1479 		audit_log_end(ab);
1480 		break;
1481 	}
1482 	default:
1483 		err = -EINVAL;
1484 		break;
1485 	}
1486 
1487 	return err < 0 ? err : 0;
1488 }
1489 
1490 /**
1491  * audit_receive - receive messages from a netlink control socket
1492  * @skb: the message buffer
1493  *
1494  * Parse the provided skb and deal with any messages that may be present,
1495  * malformed skbs are discarded.
1496  */
1497 static void audit_receive(struct sk_buff  *skb)
1498 {
1499 	struct nlmsghdr *nlh;
1500 	/*
1501 	 * len MUST be signed for nlmsg_next to be able to dec it below 0
1502 	 * if the nlmsg_len was not aligned
1503 	 */
1504 	int len;
1505 	int err;
1506 
1507 	nlh = nlmsg_hdr(skb);
1508 	len = skb->len;
1509 
1510 	audit_ctl_lock();
1511 	while (nlmsg_ok(nlh, len)) {
1512 		err = audit_receive_msg(skb, nlh);
1513 		/* if err or if this message says it wants a response */
1514 		if (err || (nlh->nlmsg_flags & NLM_F_ACK))
1515 			netlink_ack(skb, nlh, err, NULL);
1516 
1517 		nlh = nlmsg_next(nlh, &len);
1518 	}
1519 	audit_ctl_unlock();
1520 }
1521 
1522 /* Run custom bind function on netlink socket group connect or bind requests. */
1523 static int audit_bind(struct net *net, int group)
1524 {
1525 	if (!capable(CAP_AUDIT_READ))
1526 		return -EPERM;
1527 
1528 	return 0;
1529 }
1530 
1531 static int __net_init audit_net_init(struct net *net)
1532 {
1533 	struct netlink_kernel_cfg cfg = {
1534 		.input	= audit_receive,
1535 		.bind	= audit_bind,
1536 		.flags	= NL_CFG_F_NONROOT_RECV,
1537 		.groups	= AUDIT_NLGRP_MAX,
1538 	};
1539 
1540 	struct audit_net *aunet = net_generic(net, audit_net_id);
1541 
1542 	aunet->sk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
1543 	if (aunet->sk == NULL) {
1544 		audit_panic("cannot initialize netlink socket in namespace");
1545 		return -ENOMEM;
1546 	}
1547 	aunet->sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1548 
1549 	return 0;
1550 }
1551 
1552 static void __net_exit audit_net_exit(struct net *net)
1553 {
1554 	struct audit_net *aunet = net_generic(net, audit_net_id);
1555 
1556 	/* NOTE: you would think that we would want to check the auditd
1557 	 * connection and potentially reset it here if it lives in this
1558 	 * namespace, but since the auditd connection tracking struct holds a
1559 	 * reference to this namespace (see auditd_set()) we are only ever
1560 	 * going to get here after that connection has been released */
1561 
1562 	netlink_kernel_release(aunet->sk);
1563 }
1564 
1565 static struct pernet_operations audit_net_ops __net_initdata = {
1566 	.init = audit_net_init,
1567 	.exit = audit_net_exit,
1568 	.id = &audit_net_id,
1569 	.size = sizeof(struct audit_net),
1570 };
1571 
1572 /* Initialize audit support at boot time. */
1573 static int __init audit_init(void)
1574 {
1575 	int i;
1576 
1577 	if (audit_initialized == AUDIT_DISABLED)
1578 		return 0;
1579 
1580 	audit_buffer_cache = kmem_cache_create("audit_buffer",
1581 					       sizeof(struct audit_buffer),
1582 					       0, SLAB_PANIC, NULL);
1583 
1584 	skb_queue_head_init(&audit_queue);
1585 	skb_queue_head_init(&audit_retry_queue);
1586 	skb_queue_head_init(&audit_hold_queue);
1587 
1588 	for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
1589 		INIT_LIST_HEAD(&audit_inode_hash[i]);
1590 
1591 	mutex_init(&audit_cmd_mutex.lock);
1592 	audit_cmd_mutex.owner = NULL;
1593 
1594 	pr_info("initializing netlink subsys (%s)\n",
1595 		audit_default ? "enabled" : "disabled");
1596 	register_pernet_subsys(&audit_net_ops);
1597 
1598 	audit_initialized = AUDIT_INITIALIZED;
1599 
1600 	kauditd_task = kthread_run(kauditd_thread, NULL, "kauditd");
1601 	if (IS_ERR(kauditd_task)) {
1602 		int err = PTR_ERR(kauditd_task);
1603 		panic("audit: failed to start the kauditd thread (%d)\n", err);
1604 	}
1605 
1606 	audit_log(NULL, GFP_KERNEL, AUDIT_KERNEL,
1607 		"state=initialized audit_enabled=%u res=1",
1608 		 audit_enabled);
1609 
1610 	return 0;
1611 }
1612 postcore_initcall(audit_init);
1613 
1614 /*
1615  * Process kernel command-line parameter at boot time.
1616  * audit={0|off} or audit={1|on}.
1617  */
1618 static int __init audit_enable(char *str)
1619 {
1620 	if (!strcasecmp(str, "off") || !strcmp(str, "0"))
1621 		audit_default = AUDIT_OFF;
1622 	else if (!strcasecmp(str, "on") || !strcmp(str, "1"))
1623 		audit_default = AUDIT_ON;
1624 	else {
1625 		pr_err("audit: invalid 'audit' parameter value (%s)\n", str);
1626 		audit_default = AUDIT_ON;
1627 	}
1628 
1629 	if (audit_default == AUDIT_OFF)
1630 		audit_initialized = AUDIT_DISABLED;
1631 	if (audit_set_enabled(audit_default))
1632 		pr_err("audit: error setting audit state (%d)\n",
1633 		       audit_default);
1634 
1635 	pr_info("%s\n", audit_default ?
1636 		"enabled (after initialization)" : "disabled (until reboot)");
1637 
1638 	return 1;
1639 }
1640 __setup("audit=", audit_enable);
1641 
1642 /* Process kernel command-line parameter at boot time.
1643  * audit_backlog_limit=<n> */
1644 static int __init audit_backlog_limit_set(char *str)
1645 {
1646 	u32 audit_backlog_limit_arg;
1647 
1648 	pr_info("audit_backlog_limit: ");
1649 	if (kstrtouint(str, 0, &audit_backlog_limit_arg)) {
1650 		pr_cont("using default of %u, unable to parse %s\n",
1651 			audit_backlog_limit, str);
1652 		return 1;
1653 	}
1654 
1655 	audit_backlog_limit = audit_backlog_limit_arg;
1656 	pr_cont("%d\n", audit_backlog_limit);
1657 
1658 	return 1;
1659 }
1660 __setup("audit_backlog_limit=", audit_backlog_limit_set);
1661 
1662 static void audit_buffer_free(struct audit_buffer *ab)
1663 {
1664 	if (!ab)
1665 		return;
1666 
1667 	kfree_skb(ab->skb);
1668 	kmem_cache_free(audit_buffer_cache, ab);
1669 }
1670 
1671 static struct audit_buffer *audit_buffer_alloc(struct audit_context *ctx,
1672 					       gfp_t gfp_mask, int type)
1673 {
1674 	struct audit_buffer *ab;
1675 
1676 	ab = kmem_cache_alloc(audit_buffer_cache, gfp_mask);
1677 	if (!ab)
1678 		return NULL;
1679 
1680 	ab->skb = nlmsg_new(AUDIT_BUFSIZ, gfp_mask);
1681 	if (!ab->skb)
1682 		goto err;
1683 	if (!nlmsg_put(ab->skb, 0, 0, type, 0, 0))
1684 		goto err;
1685 
1686 	ab->ctx = ctx;
1687 	ab->gfp_mask = gfp_mask;
1688 
1689 	return ab;
1690 
1691 err:
1692 	audit_buffer_free(ab);
1693 	return NULL;
1694 }
1695 
1696 /**
1697  * audit_serial - compute a serial number for the audit record
1698  *
1699  * Compute a serial number for the audit record.  Audit records are
1700  * written to user-space as soon as they are generated, so a complete
1701  * audit record may be written in several pieces.  The timestamp of the
1702  * record and this serial number are used by the user-space tools to
1703  * determine which pieces belong to the same audit record.  The
1704  * (timestamp,serial) tuple is unique for each syscall and is live from
1705  * syscall entry to syscall exit.
1706  *
1707  * NOTE: Another possibility is to store the formatted records off the
1708  * audit context (for those records that have a context), and emit them
1709  * all at syscall exit.  However, this could delay the reporting of
1710  * significant errors until syscall exit (or never, if the system
1711  * halts).
1712  */
1713 unsigned int audit_serial(void)
1714 {
1715 	static atomic_t serial = ATOMIC_INIT(0);
1716 
1717 	return atomic_add_return(1, &serial);
1718 }
1719 
1720 static inline void audit_get_stamp(struct audit_context *ctx,
1721 				   struct timespec64 *t, unsigned int *serial)
1722 {
1723 	if (!ctx || !auditsc_get_stamp(ctx, t, serial)) {
1724 		ktime_get_coarse_real_ts64(t);
1725 		*serial = audit_serial();
1726 	}
1727 }
1728 
1729 /**
1730  * audit_log_start - obtain an audit buffer
1731  * @ctx: audit_context (may be NULL)
1732  * @gfp_mask: type of allocation
1733  * @type: audit message type
1734  *
1735  * Returns audit_buffer pointer on success or NULL on error.
1736  *
1737  * Obtain an audit buffer.  This routine does locking to obtain the
1738  * audit buffer, but then no locking is required for calls to
1739  * audit_log_*format.  If the task (ctx) is a task that is currently in a
1740  * syscall, then the syscall is marked as auditable and an audit record
1741  * will be written at syscall exit.  If there is no associated task, then
1742  * task context (ctx) should be NULL.
1743  */
1744 struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
1745 				     int type)
1746 {
1747 	struct audit_buffer *ab;
1748 	struct timespec64 t;
1749 	unsigned int uninitialized_var(serial);
1750 
1751 	if (audit_initialized != AUDIT_INITIALIZED)
1752 		return NULL;
1753 
1754 	if (unlikely(!audit_filter(type, AUDIT_FILTER_EXCLUDE)))
1755 		return NULL;
1756 
1757 	/* NOTE: don't ever fail/sleep on these two conditions:
1758 	 * 1. auditd generated record - since we need auditd to drain the
1759 	 *    queue; also, when we are checking for auditd, compare PIDs using
1760 	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
1761 	 *    using a PID anchored in the caller's namespace
1762 	 * 2. generator holding the audit_cmd_mutex - we don't want to block
1763 	 *    while holding the mutex */
1764 	if (!(auditd_test_task(current) || audit_ctl_owner_current())) {
1765 		long stime = audit_backlog_wait_time;
1766 
1767 		while (audit_backlog_limit &&
1768 		       (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
1769 			/* wake kauditd to try and flush the queue */
1770 			wake_up_interruptible(&kauditd_wait);
1771 
1772 			/* sleep if we are allowed and we haven't exhausted our
1773 			 * backlog wait limit */
1774 			if (gfpflags_allow_blocking(gfp_mask) && (stime > 0)) {
1775 				DECLARE_WAITQUEUE(wait, current);
1776 
1777 				add_wait_queue_exclusive(&audit_backlog_wait,
1778 							 &wait);
1779 				set_current_state(TASK_UNINTERRUPTIBLE);
1780 				stime = schedule_timeout(stime);
1781 				remove_wait_queue(&audit_backlog_wait, &wait);
1782 			} else {
1783 				if (audit_rate_check() && printk_ratelimit())
1784 					pr_warn("audit_backlog=%d > audit_backlog_limit=%d\n",
1785 						skb_queue_len(&audit_queue),
1786 						audit_backlog_limit);
1787 				audit_log_lost("backlog limit exceeded");
1788 				return NULL;
1789 			}
1790 		}
1791 	}
1792 
1793 	ab = audit_buffer_alloc(ctx, gfp_mask, type);
1794 	if (!ab) {
1795 		audit_log_lost("out of memory in audit_log_start");
1796 		return NULL;
1797 	}
1798 
1799 	audit_get_stamp(ab->ctx, &t, &serial);
1800 	audit_log_format(ab, "audit(%llu.%03lu:%u): ",
1801 			 (unsigned long long)t.tv_sec, t.tv_nsec/1000000, serial);
1802 
1803 	return ab;
1804 }
1805 
1806 /**
1807  * audit_expand - expand skb in the audit buffer
1808  * @ab: audit_buffer
1809  * @extra: space to add at tail of the skb
1810  *
1811  * Returns 0 (no space) on failed expansion, or available space if
1812  * successful.
1813  */
1814 static inline int audit_expand(struct audit_buffer *ab, int extra)
1815 {
1816 	struct sk_buff *skb = ab->skb;
1817 	int oldtail = skb_tailroom(skb);
1818 	int ret = pskb_expand_head(skb, 0, extra, ab->gfp_mask);
1819 	int newtail = skb_tailroom(skb);
1820 
1821 	if (ret < 0) {
1822 		audit_log_lost("out of memory in audit_expand");
1823 		return 0;
1824 	}
1825 
1826 	skb->truesize += newtail - oldtail;
1827 	return newtail;
1828 }
1829 
1830 /*
1831  * Format an audit message into the audit buffer.  If there isn't enough
1832  * room in the audit buffer, more room will be allocated and vsnprint
1833  * will be called a second time.  Currently, we assume that a printk
1834  * can't format message larger than 1024 bytes, so we don't either.
1835  */
1836 static void audit_log_vformat(struct audit_buffer *ab, const char *fmt,
1837 			      va_list args)
1838 {
1839 	int len, avail;
1840 	struct sk_buff *skb;
1841 	va_list args2;
1842 
1843 	if (!ab)
1844 		return;
1845 
1846 	BUG_ON(!ab->skb);
1847 	skb = ab->skb;
1848 	avail = skb_tailroom(skb);
1849 	if (avail == 0) {
1850 		avail = audit_expand(ab, AUDIT_BUFSIZ);
1851 		if (!avail)
1852 			goto out;
1853 	}
1854 	va_copy(args2, args);
1855 	len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args);
1856 	if (len >= avail) {
1857 		/* The printk buffer is 1024 bytes long, so if we get
1858 		 * here and AUDIT_BUFSIZ is at least 1024, then we can
1859 		 * log everything that printk could have logged. */
1860 		avail = audit_expand(ab,
1861 			max_t(unsigned, AUDIT_BUFSIZ, 1+len-avail));
1862 		if (!avail)
1863 			goto out_va_end;
1864 		len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args2);
1865 	}
1866 	if (len > 0)
1867 		skb_put(skb, len);
1868 out_va_end:
1869 	va_end(args2);
1870 out:
1871 	return;
1872 }
1873 
1874 /**
1875  * audit_log_format - format a message into the audit buffer.
1876  * @ab: audit_buffer
1877  * @fmt: format string
1878  * @...: optional parameters matching @fmt string
1879  *
1880  * All the work is done in audit_log_vformat.
1881  */
1882 void audit_log_format(struct audit_buffer *ab, const char *fmt, ...)
1883 {
1884 	va_list args;
1885 
1886 	if (!ab)
1887 		return;
1888 	va_start(args, fmt);
1889 	audit_log_vformat(ab, fmt, args);
1890 	va_end(args);
1891 }
1892 
1893 /**
1894  * audit_log_n_hex - convert a buffer to hex and append it to the audit skb
1895  * @ab: the audit_buffer
1896  * @buf: buffer to convert to hex
1897  * @len: length of @buf to be converted
1898  *
1899  * No return value; failure to expand is silently ignored.
1900  *
1901  * This function will take the passed buf and convert it into a string of
1902  * ascii hex digits. The new string is placed onto the skb.
1903  */
1904 void audit_log_n_hex(struct audit_buffer *ab, const unsigned char *buf,
1905 		size_t len)
1906 {
1907 	int i, avail, new_len;
1908 	unsigned char *ptr;
1909 	struct sk_buff *skb;
1910 
1911 	if (!ab)
1912 		return;
1913 
1914 	BUG_ON(!ab->skb);
1915 	skb = ab->skb;
1916 	avail = skb_tailroom(skb);
1917 	new_len = len<<1;
1918 	if (new_len >= avail) {
1919 		/* Round the buffer request up to the next multiple */
1920 		new_len = AUDIT_BUFSIZ*(((new_len-avail)/AUDIT_BUFSIZ) + 1);
1921 		avail = audit_expand(ab, new_len);
1922 		if (!avail)
1923 			return;
1924 	}
1925 
1926 	ptr = skb_tail_pointer(skb);
1927 	for (i = 0; i < len; i++)
1928 		ptr = hex_byte_pack_upper(ptr, buf[i]);
1929 	*ptr = 0;
1930 	skb_put(skb, len << 1); /* new string is twice the old string */
1931 }
1932 
1933 /*
1934  * Format a string of no more than slen characters into the audit buffer,
1935  * enclosed in quote marks.
1936  */
1937 void audit_log_n_string(struct audit_buffer *ab, const char *string,
1938 			size_t slen)
1939 {
1940 	int avail, new_len;
1941 	unsigned char *ptr;
1942 	struct sk_buff *skb;
1943 
1944 	if (!ab)
1945 		return;
1946 
1947 	BUG_ON(!ab->skb);
1948 	skb = ab->skb;
1949 	avail = skb_tailroom(skb);
1950 	new_len = slen + 3;	/* enclosing quotes + null terminator */
1951 	if (new_len > avail) {
1952 		avail = audit_expand(ab, new_len);
1953 		if (!avail)
1954 			return;
1955 	}
1956 	ptr = skb_tail_pointer(skb);
1957 	*ptr++ = '"';
1958 	memcpy(ptr, string, slen);
1959 	ptr += slen;
1960 	*ptr++ = '"';
1961 	*ptr = 0;
1962 	skb_put(skb, slen + 2);	/* don't include null terminator */
1963 }
1964 
1965 /**
1966  * audit_string_contains_control - does a string need to be logged in hex
1967  * @string: string to be checked
1968  * @len: max length of the string to check
1969  */
1970 bool audit_string_contains_control(const char *string, size_t len)
1971 {
1972 	const unsigned char *p;
1973 	for (p = string; p < (const unsigned char *)string + len; p++) {
1974 		if (*p == '"' || *p < 0x21 || *p > 0x7e)
1975 			return true;
1976 	}
1977 	return false;
1978 }
1979 
1980 /**
1981  * audit_log_n_untrustedstring - log a string that may contain random characters
1982  * @ab: audit_buffer
1983  * @len: length of string (not including trailing null)
1984  * @string: string to be logged
1985  *
1986  * This code will escape a string that is passed to it if the string
1987  * contains a control character, unprintable character, double quote mark,
1988  * or a space. Unescaped strings will start and end with a double quote mark.
1989  * Strings that are escaped are printed in hex (2 digits per char).
1990  *
1991  * The caller specifies the number of characters in the string to log, which may
1992  * or may not be the entire string.
1993  */
1994 void audit_log_n_untrustedstring(struct audit_buffer *ab, const char *string,
1995 				 size_t len)
1996 {
1997 	if (audit_string_contains_control(string, len))
1998 		audit_log_n_hex(ab, string, len);
1999 	else
2000 		audit_log_n_string(ab, string, len);
2001 }
2002 
2003 /**
2004  * audit_log_untrustedstring - log a string that may contain random characters
2005  * @ab: audit_buffer
2006  * @string: string to be logged
2007  *
2008  * Same as audit_log_n_untrustedstring(), except that strlen is used to
2009  * determine string length.
2010  */
2011 void audit_log_untrustedstring(struct audit_buffer *ab, const char *string)
2012 {
2013 	audit_log_n_untrustedstring(ab, string, strlen(string));
2014 }
2015 
2016 /* This is a helper-function to print the escaped d_path */
2017 void audit_log_d_path(struct audit_buffer *ab, const char *prefix,
2018 		      const struct path *path)
2019 {
2020 	char *p, *pathname;
2021 
2022 	if (prefix)
2023 		audit_log_format(ab, "%s", prefix);
2024 
2025 	/* We will allow 11 spaces for ' (deleted)' to be appended */
2026 	pathname = kmalloc(PATH_MAX+11, ab->gfp_mask);
2027 	if (!pathname) {
2028 		audit_log_string(ab, "<no_memory>");
2029 		return;
2030 	}
2031 	p = d_path(path, pathname, PATH_MAX+11);
2032 	if (IS_ERR(p)) { /* Should never happen since we send PATH_MAX */
2033 		/* FIXME: can we save some information here? */
2034 		audit_log_string(ab, "<too_long>");
2035 	} else
2036 		audit_log_untrustedstring(ab, p);
2037 	kfree(pathname);
2038 }
2039 
2040 void audit_log_session_info(struct audit_buffer *ab)
2041 {
2042 	unsigned int sessionid = audit_get_sessionid(current);
2043 	uid_t auid = from_kuid(&init_user_ns, audit_get_loginuid(current));
2044 
2045 	audit_log_format(ab, "auid=%u ses=%u", auid, sessionid);
2046 }
2047 
2048 void audit_log_key(struct audit_buffer *ab, char *key)
2049 {
2050 	audit_log_format(ab, " key=");
2051 	if (key)
2052 		audit_log_untrustedstring(ab, key);
2053 	else
2054 		audit_log_format(ab, "(null)");
2055 }
2056 
2057 void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
2058 {
2059 	int i;
2060 
2061 	if (cap_isclear(*cap)) {
2062 		audit_log_format(ab, " %s=0", prefix);
2063 		return;
2064 	}
2065 	audit_log_format(ab, " %s=", prefix);
2066 	CAP_FOR_EACH_U32(i)
2067 		audit_log_format(ab, "%08x", cap->cap[CAP_LAST_U32 - i]);
2068 }
2069 
2070 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
2071 {
2072 	audit_log_cap(ab, "cap_fp", &name->fcap.permitted);
2073 	audit_log_cap(ab, "cap_fi", &name->fcap.inheritable);
2074 	audit_log_format(ab, " cap_fe=%d cap_fver=%x",
2075 			 name->fcap.fE, name->fcap_ver);
2076 }
2077 
2078 static inline int audit_copy_fcaps(struct audit_names *name,
2079 				   const struct dentry *dentry)
2080 {
2081 	struct cpu_vfs_cap_data caps;
2082 	int rc;
2083 
2084 	if (!dentry)
2085 		return 0;
2086 
2087 	rc = get_vfs_caps_from_disk(dentry, &caps);
2088 	if (rc)
2089 		return rc;
2090 
2091 	name->fcap.permitted = caps.permitted;
2092 	name->fcap.inheritable = caps.inheritable;
2093 	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2094 	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >>
2095 				VFS_CAP_REVISION_SHIFT;
2096 
2097 	return 0;
2098 }
2099 
2100 /* Copy inode data into an audit_names. */
2101 void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
2102 		      struct inode *inode)
2103 {
2104 	name->ino   = inode->i_ino;
2105 	name->dev   = inode->i_sb->s_dev;
2106 	name->mode  = inode->i_mode;
2107 	name->uid   = inode->i_uid;
2108 	name->gid   = inode->i_gid;
2109 	name->rdev  = inode->i_rdev;
2110 	security_inode_getsecid(inode, &name->osid);
2111 	audit_copy_fcaps(name, dentry);
2112 }
2113 
2114 /**
2115  * audit_log_name - produce AUDIT_PATH record from struct audit_names
2116  * @context: audit_context for the task
2117  * @n: audit_names structure with reportable details
2118  * @path: optional path to report instead of audit_names->name
2119  * @record_num: record number to report when handling a list of names
2120  * @call_panic: optional pointer to int that will be updated if secid fails
2121  */
2122 void audit_log_name(struct audit_context *context, struct audit_names *n,
2123 		    const struct path *path, int record_num, int *call_panic)
2124 {
2125 	struct audit_buffer *ab;
2126 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
2127 	if (!ab)
2128 		return;
2129 
2130 	audit_log_format(ab, "item=%d", record_num);
2131 
2132 	if (path)
2133 		audit_log_d_path(ab, " name=", path);
2134 	else if (n->name) {
2135 		switch (n->name_len) {
2136 		case AUDIT_NAME_FULL:
2137 			/* log the full path */
2138 			audit_log_format(ab, " name=");
2139 			audit_log_untrustedstring(ab, n->name->name);
2140 			break;
2141 		case 0:
2142 			/* name was specified as a relative path and the
2143 			 * directory component is the cwd */
2144 			audit_log_d_path(ab, " name=", &context->pwd);
2145 			break;
2146 		default:
2147 			/* log the name's directory component */
2148 			audit_log_format(ab, " name=");
2149 			audit_log_n_untrustedstring(ab, n->name->name,
2150 						    n->name_len);
2151 		}
2152 	} else
2153 		audit_log_format(ab, " name=(null)");
2154 
2155 	if (n->ino != AUDIT_INO_UNSET)
2156 		audit_log_format(ab, " inode=%lu"
2157 				 " dev=%02x:%02x mode=%#ho"
2158 				 " ouid=%u ogid=%u rdev=%02x:%02x",
2159 				 n->ino,
2160 				 MAJOR(n->dev),
2161 				 MINOR(n->dev),
2162 				 n->mode,
2163 				 from_kuid(&init_user_ns, n->uid),
2164 				 from_kgid(&init_user_ns, n->gid),
2165 				 MAJOR(n->rdev),
2166 				 MINOR(n->rdev));
2167 	if (n->osid != 0) {
2168 		char *ctx = NULL;
2169 		u32 len;
2170 		if (security_secid_to_secctx(
2171 			n->osid, &ctx, &len)) {
2172 			audit_log_format(ab, " osid=%u", n->osid);
2173 			if (call_panic)
2174 				*call_panic = 2;
2175 		} else {
2176 			audit_log_format(ab, " obj=%s", ctx);
2177 			security_release_secctx(ctx, len);
2178 		}
2179 	}
2180 
2181 	/* log the audit_names record type */
2182 	switch(n->type) {
2183 	case AUDIT_TYPE_NORMAL:
2184 		audit_log_format(ab, " nametype=NORMAL");
2185 		break;
2186 	case AUDIT_TYPE_PARENT:
2187 		audit_log_format(ab, " nametype=PARENT");
2188 		break;
2189 	case AUDIT_TYPE_CHILD_DELETE:
2190 		audit_log_format(ab, " nametype=DELETE");
2191 		break;
2192 	case AUDIT_TYPE_CHILD_CREATE:
2193 		audit_log_format(ab, " nametype=CREATE");
2194 		break;
2195 	default:
2196 		audit_log_format(ab, " nametype=UNKNOWN");
2197 		break;
2198 	}
2199 
2200 	audit_log_fcaps(ab, n);
2201 	audit_log_end(ab);
2202 }
2203 
2204 int audit_log_task_context(struct audit_buffer *ab)
2205 {
2206 	char *ctx = NULL;
2207 	unsigned len;
2208 	int error;
2209 	u32 sid;
2210 
2211 	security_task_getsecid(current, &sid);
2212 	if (!sid)
2213 		return 0;
2214 
2215 	error = security_secid_to_secctx(sid, &ctx, &len);
2216 	if (error) {
2217 		if (error != -EINVAL)
2218 			goto error_path;
2219 		return 0;
2220 	}
2221 
2222 	audit_log_format(ab, " subj=%s", ctx);
2223 	security_release_secctx(ctx, len);
2224 	return 0;
2225 
2226 error_path:
2227 	audit_panic("error in audit_log_task_context");
2228 	return error;
2229 }
2230 EXPORT_SYMBOL(audit_log_task_context);
2231 
2232 void audit_log_d_path_exe(struct audit_buffer *ab,
2233 			  struct mm_struct *mm)
2234 {
2235 	struct file *exe_file;
2236 
2237 	if (!mm)
2238 		goto out_null;
2239 
2240 	exe_file = get_mm_exe_file(mm);
2241 	if (!exe_file)
2242 		goto out_null;
2243 
2244 	audit_log_d_path(ab, " exe=", &exe_file->f_path);
2245 	fput(exe_file);
2246 	return;
2247 out_null:
2248 	audit_log_format(ab, " exe=(null)");
2249 }
2250 
2251 struct tty_struct *audit_get_tty(void)
2252 {
2253 	struct tty_struct *tty = NULL;
2254 	unsigned long flags;
2255 
2256 	spin_lock_irqsave(&current->sighand->siglock, flags);
2257 	if (current->signal)
2258 		tty = tty_kref_get(current->signal->tty);
2259 	spin_unlock_irqrestore(&current->sighand->siglock, flags);
2260 	return tty;
2261 }
2262 
2263 void audit_put_tty(struct tty_struct *tty)
2264 {
2265 	tty_kref_put(tty);
2266 }
2267 
2268 void audit_log_task_info(struct audit_buffer *ab)
2269 {
2270 	const struct cred *cred;
2271 	char comm[sizeof(current->comm)];
2272 	struct tty_struct *tty;
2273 
2274 	if (!ab)
2275 		return;
2276 
2277 	cred = current_cred();
2278 	tty = audit_get_tty();
2279 	audit_log_format(ab,
2280 			 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
2281 			 " euid=%u suid=%u fsuid=%u"
2282 			 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
2283 			 task_ppid_nr(current),
2284 			 task_tgid_nr(current),
2285 			 from_kuid(&init_user_ns, audit_get_loginuid(current)),
2286 			 from_kuid(&init_user_ns, cred->uid),
2287 			 from_kgid(&init_user_ns, cred->gid),
2288 			 from_kuid(&init_user_ns, cred->euid),
2289 			 from_kuid(&init_user_ns, cred->suid),
2290 			 from_kuid(&init_user_ns, cred->fsuid),
2291 			 from_kgid(&init_user_ns, cred->egid),
2292 			 from_kgid(&init_user_ns, cred->sgid),
2293 			 from_kgid(&init_user_ns, cred->fsgid),
2294 			 tty ? tty_name(tty) : "(none)",
2295 			 audit_get_sessionid(current));
2296 	audit_put_tty(tty);
2297 	audit_log_format(ab, " comm=");
2298 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2299 	audit_log_d_path_exe(ab, current->mm);
2300 	audit_log_task_context(ab);
2301 }
2302 EXPORT_SYMBOL(audit_log_task_info);
2303 
2304 /**
2305  * audit_log_link_denied - report a link restriction denial
2306  * @operation: specific link operation
2307  */
2308 void audit_log_link_denied(const char *operation)
2309 {
2310 	struct audit_buffer *ab;
2311 
2312 	if (!audit_enabled || audit_dummy_context())
2313 		return;
2314 
2315 	/* Generate AUDIT_ANOM_LINK with subject, operation, outcome. */
2316 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_ANOM_LINK);
2317 	if (!ab)
2318 		return;
2319 	audit_log_format(ab, "op=%s", operation);
2320 	audit_log_task_info(ab);
2321 	audit_log_format(ab, " res=0");
2322 	audit_log_end(ab);
2323 }
2324 
2325 /**
2326  * audit_log_end - end one audit record
2327  * @ab: the audit_buffer
2328  *
2329  * We can not do a netlink send inside an irq context because it blocks (last
2330  * arg, flags, is not set to MSG_DONTWAIT), so the audit buffer is placed on a
2331  * queue and a tasklet is scheduled to remove them from the queue outside the
2332  * irq context.  May be called in any context.
2333  */
2334 void audit_log_end(struct audit_buffer *ab)
2335 {
2336 	struct sk_buff *skb;
2337 	struct nlmsghdr *nlh;
2338 
2339 	if (!ab)
2340 		return;
2341 
2342 	if (audit_rate_check()) {
2343 		skb = ab->skb;
2344 		ab->skb = NULL;
2345 
2346 		/* setup the netlink header, see the comments in
2347 		 * kauditd_send_multicast_skb() for length quirks */
2348 		nlh = nlmsg_hdr(skb);
2349 		nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
2350 
2351 		/* queue the netlink packet and poke the kauditd thread */
2352 		skb_queue_tail(&audit_queue, skb);
2353 		wake_up_interruptible(&kauditd_wait);
2354 	} else
2355 		audit_log_lost("rate limit exceeded");
2356 
2357 	audit_buffer_free(ab);
2358 }
2359 
2360 /**
2361  * audit_log - Log an audit record
2362  * @ctx: audit context
2363  * @gfp_mask: type of allocation
2364  * @type: audit message type
2365  * @fmt: format string to use
2366  * @...: variable parameters matching the format string
2367  *
2368  * This is a convenience function that calls audit_log_start,
2369  * audit_log_vformat, and audit_log_end.  It may be called
2370  * in any context.
2371  */
2372 void audit_log(struct audit_context *ctx, gfp_t gfp_mask, int type,
2373 	       const char *fmt, ...)
2374 {
2375 	struct audit_buffer *ab;
2376 	va_list args;
2377 
2378 	ab = audit_log_start(ctx, gfp_mask, type);
2379 	if (ab) {
2380 		va_start(args, fmt);
2381 		audit_log_vformat(ab, fmt, args);
2382 		va_end(args);
2383 		audit_log_end(ab);
2384 	}
2385 }
2386 
2387 EXPORT_SYMBOL(audit_log_start);
2388 EXPORT_SYMBOL(audit_log_end);
2389 EXPORT_SYMBOL(audit_log_format);
2390 EXPORT_SYMBOL(audit_log);
2391