xref: /openbmc/linux/kernel/audit.c (revision 4da722ca19f30f7db250db808d1ab1703607a932)
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  * Example user-space utilities: http://people.redhat.com/sgrubb/audit/
42  */
43 
44 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
45 
46 #include <linux/file.h>
47 #include <linux/init.h>
48 #include <linux/types.h>
49 #include <linux/atomic.h>
50 #include <linux/mm.h>
51 #include <linux/export.h>
52 #include <linux/slab.h>
53 #include <linux/err.h>
54 #include <linux/kthread.h>
55 #include <linux/kernel.h>
56 #include <linux/syscalls.h>
57 #include <linux/spinlock.h>
58 #include <linux/rcupdate.h>
59 #include <linux/mutex.h>
60 #include <linux/gfp.h>
61 #include <linux/pid.h>
62 #include <linux/slab.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 #define AUDIT_OFF	0
86 #define AUDIT_ON	1
87 #define AUDIT_LOCKED	2
88 u32		audit_enabled;
89 u32		audit_ever_enabled;
90 
91 EXPORT_SYMBOL_GPL(audit_enabled);
92 
93 /* Default state when kernel boots without any parameters. */
94 static u32	audit_default;
95 
96 /* If auditing cannot proceed, audit_failure selects what happens. */
97 static u32	audit_failure = AUDIT_FAIL_PRINTK;
98 
99 /* private audit network namespace index */
100 static unsigned int audit_net_id;
101 
102 /**
103  * struct audit_net - audit private network namespace data
104  * @sk: communication socket
105  */
106 struct audit_net {
107 	struct sock *sk;
108 };
109 
110 /**
111  * struct auditd_connection - kernel/auditd connection state
112  * @pid: auditd PID
113  * @portid: netlink portid
114  * @net: the associated network namespace
115  * @rcu: RCU head
116  *
117  * Description:
118  * This struct is RCU protected; you must either hold the RCU lock for reading
119  * or the associated spinlock for writing.
120  */
121 static struct auditd_connection {
122 	struct pid *pid;
123 	u32 portid;
124 	struct net *net;
125 	struct rcu_head rcu;
126 } *auditd_conn = NULL;
127 static DEFINE_SPINLOCK(auditd_conn_lock);
128 
129 /* If audit_rate_limit is non-zero, limit the rate of sending audit records
130  * to that number per second.  This prevents DoS attacks, but results in
131  * audit records being dropped. */
132 static u32	audit_rate_limit;
133 
134 /* Number of outstanding audit_buffers allowed.
135  * When set to zero, this means unlimited. */
136 static u32	audit_backlog_limit = 64;
137 #define AUDIT_BACKLOG_WAIT_TIME (60 * HZ)
138 static u32	audit_backlog_wait_time = AUDIT_BACKLOG_WAIT_TIME;
139 
140 /* The identity of the user shutting down the audit system. */
141 kuid_t		audit_sig_uid = INVALID_UID;
142 pid_t		audit_sig_pid = -1;
143 u32		audit_sig_sid = 0;
144 
145 /* Records can be lost in several ways:
146    0) [suppressed in audit_alloc]
147    1) out of memory in audit_log_start [kmalloc of struct audit_buffer]
148    2) out of memory in audit_log_move [alloc_skb]
149    3) suppressed due to audit_rate_limit
150    4) suppressed due to audit_backlog_limit
151 */
152 static atomic_t	audit_lost = ATOMIC_INIT(0);
153 
154 /* Hash for inode-based rules */
155 struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
156 
157 static struct kmem_cache *audit_buffer_cache;
158 
159 /* queue msgs to send via kauditd_task */
160 static struct sk_buff_head audit_queue;
161 /* queue msgs due to temporary unicast send problems */
162 static struct sk_buff_head audit_retry_queue;
163 /* queue msgs waiting for new auditd connection */
164 static struct sk_buff_head audit_hold_queue;
165 
166 /* queue servicing thread */
167 static struct task_struct *kauditd_task;
168 static DECLARE_WAIT_QUEUE_HEAD(kauditd_wait);
169 
170 /* waitqueue for callers who are blocked on the audit backlog */
171 static DECLARE_WAIT_QUEUE_HEAD(audit_backlog_wait);
172 
173 static struct audit_features af = {.vers = AUDIT_FEATURE_VERSION,
174 				   .mask = -1,
175 				   .features = 0,
176 				   .lock = 0,};
177 
178 static char *audit_feature_names[2] = {
179 	"only_unset_loginuid",
180 	"loginuid_immutable",
181 };
182 
183 
184 /* Serialize requests from userspace. */
185 DEFINE_MUTEX(audit_cmd_mutex);
186 
187 /* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
188  * audit records.  Since printk uses a 1024 byte buffer, this buffer
189  * should be at least that large. */
190 #define AUDIT_BUFSIZ 1024
191 
192 /* The audit_buffer is used when formatting an audit record.  The caller
193  * locks briefly to get the record off the freelist or to allocate the
194  * buffer, and locks briefly to send the buffer to the netlink layer or
195  * to place it on a transmit queue.  Multiple audit_buffers can be in
196  * use simultaneously. */
197 struct audit_buffer {
198 	struct sk_buff       *skb;	/* formatted skb ready to send */
199 	struct audit_context *ctx;	/* NULL or associated context */
200 	gfp_t		     gfp_mask;
201 };
202 
203 struct audit_reply {
204 	__u32 portid;
205 	struct net *net;
206 	struct sk_buff *skb;
207 };
208 
209 /**
210  * auditd_test_task - Check to see if a given task is an audit daemon
211  * @task: the task to check
212  *
213  * Description:
214  * Return 1 if the task is a registered audit daemon, 0 otherwise.
215  */
216 int auditd_test_task(struct task_struct *task)
217 {
218 	int rc;
219 	struct auditd_connection *ac;
220 
221 	rcu_read_lock();
222 	ac = rcu_dereference(auditd_conn);
223 	rc = (ac && ac->pid == task_tgid(task) ? 1 : 0);
224 	rcu_read_unlock();
225 
226 	return rc;
227 }
228 
229 /**
230  * auditd_pid_vnr - Return the auditd PID relative to the namespace
231  *
232  * Description:
233  * Returns the PID in relation to the namespace, 0 on failure.
234  */
235 static pid_t auditd_pid_vnr(void)
236 {
237 	pid_t pid;
238 	const struct auditd_connection *ac;
239 
240 	rcu_read_lock();
241 	ac = rcu_dereference(auditd_conn);
242 	if (!ac || !ac->pid)
243 		pid = 0;
244 	else
245 		pid = pid_vnr(ac->pid);
246 	rcu_read_unlock();
247 
248 	return pid;
249 }
250 
251 /**
252  * audit_get_sk - Return the audit socket for the given network namespace
253  * @net: the destination network namespace
254  *
255  * Description:
256  * Returns the sock pointer if valid, NULL otherwise.  The caller must ensure
257  * that a reference is held for the network namespace while the sock is in use.
258  */
259 static struct sock *audit_get_sk(const struct net *net)
260 {
261 	struct audit_net *aunet;
262 
263 	if (!net)
264 		return NULL;
265 
266 	aunet = net_generic(net, audit_net_id);
267 	return aunet->sk;
268 }
269 
270 void audit_panic(const char *message)
271 {
272 	switch (audit_failure) {
273 	case AUDIT_FAIL_SILENT:
274 		break;
275 	case AUDIT_FAIL_PRINTK:
276 		if (printk_ratelimit())
277 			pr_err("%s\n", message);
278 		break;
279 	case AUDIT_FAIL_PANIC:
280 		panic("audit: %s\n", message);
281 		break;
282 	}
283 }
284 
285 static inline int audit_rate_check(void)
286 {
287 	static unsigned long	last_check = 0;
288 	static int		messages   = 0;
289 	static DEFINE_SPINLOCK(lock);
290 	unsigned long		flags;
291 	unsigned long		now;
292 	unsigned long		elapsed;
293 	int			retval	   = 0;
294 
295 	if (!audit_rate_limit) return 1;
296 
297 	spin_lock_irqsave(&lock, flags);
298 	if (++messages < audit_rate_limit) {
299 		retval = 1;
300 	} else {
301 		now     = jiffies;
302 		elapsed = now - last_check;
303 		if (elapsed > HZ) {
304 			last_check = now;
305 			messages   = 0;
306 			retval     = 1;
307 		}
308 	}
309 	spin_unlock_irqrestore(&lock, flags);
310 
311 	return retval;
312 }
313 
314 /**
315  * audit_log_lost - conditionally log lost audit message event
316  * @message: the message stating reason for lost audit message
317  *
318  * Emit at least 1 message per second, even if audit_rate_check is
319  * throttling.
320  * Always increment the lost messages counter.
321 */
322 void audit_log_lost(const char *message)
323 {
324 	static unsigned long	last_msg = 0;
325 	static DEFINE_SPINLOCK(lock);
326 	unsigned long		flags;
327 	unsigned long		now;
328 	int			print;
329 
330 	atomic_inc(&audit_lost);
331 
332 	print = (audit_failure == AUDIT_FAIL_PANIC || !audit_rate_limit);
333 
334 	if (!print) {
335 		spin_lock_irqsave(&lock, flags);
336 		now = jiffies;
337 		if (now - last_msg > HZ) {
338 			print = 1;
339 			last_msg = now;
340 		}
341 		spin_unlock_irqrestore(&lock, flags);
342 	}
343 
344 	if (print) {
345 		if (printk_ratelimit())
346 			pr_warn("audit_lost=%u audit_rate_limit=%u audit_backlog_limit=%u\n",
347 				atomic_read(&audit_lost),
348 				audit_rate_limit,
349 				audit_backlog_limit);
350 		audit_panic(message);
351 	}
352 }
353 
354 static int audit_log_config_change(char *function_name, u32 new, u32 old,
355 				   int allow_changes)
356 {
357 	struct audit_buffer *ab;
358 	int rc = 0;
359 
360 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE);
361 	if (unlikely(!ab))
362 		return rc;
363 	audit_log_format(ab, "%s=%u old=%u", function_name, new, old);
364 	audit_log_session_info(ab);
365 	rc = audit_log_task_context(ab);
366 	if (rc)
367 		allow_changes = 0; /* Something weird, deny request */
368 	audit_log_format(ab, " res=%d", allow_changes);
369 	audit_log_end(ab);
370 	return rc;
371 }
372 
373 static int audit_do_config_change(char *function_name, u32 *to_change, u32 new)
374 {
375 	int allow_changes, rc = 0;
376 	u32 old = *to_change;
377 
378 	/* check if we are locked */
379 	if (audit_enabled == AUDIT_LOCKED)
380 		allow_changes = 0;
381 	else
382 		allow_changes = 1;
383 
384 	if (audit_enabled != AUDIT_OFF) {
385 		rc = audit_log_config_change(function_name, new, old, allow_changes);
386 		if (rc)
387 			allow_changes = 0;
388 	}
389 
390 	/* If we are allowed, make the change */
391 	if (allow_changes == 1)
392 		*to_change = new;
393 	/* Not allowed, update reason */
394 	else if (rc == 0)
395 		rc = -EPERM;
396 	return rc;
397 }
398 
399 static int audit_set_rate_limit(u32 limit)
400 {
401 	return audit_do_config_change("audit_rate_limit", &audit_rate_limit, limit);
402 }
403 
404 static int audit_set_backlog_limit(u32 limit)
405 {
406 	return audit_do_config_change("audit_backlog_limit", &audit_backlog_limit, limit);
407 }
408 
409 static int audit_set_backlog_wait_time(u32 timeout)
410 {
411 	return audit_do_config_change("audit_backlog_wait_time",
412 				      &audit_backlog_wait_time, timeout);
413 }
414 
415 static int audit_set_enabled(u32 state)
416 {
417 	int rc;
418 	if (state > AUDIT_LOCKED)
419 		return -EINVAL;
420 
421 	rc =  audit_do_config_change("audit_enabled", &audit_enabled, state);
422 	if (!rc)
423 		audit_ever_enabled |= !!state;
424 
425 	return rc;
426 }
427 
428 static int audit_set_failure(u32 state)
429 {
430 	if (state != AUDIT_FAIL_SILENT
431 	    && state != AUDIT_FAIL_PRINTK
432 	    && state != AUDIT_FAIL_PANIC)
433 		return -EINVAL;
434 
435 	return audit_do_config_change("audit_failure", &audit_failure, state);
436 }
437 
438 /**
439  * auditd_conn_free - RCU helper to release an auditd connection struct
440  * @rcu: RCU head
441  *
442  * Description:
443  * Drop any references inside the auditd connection tracking struct and free
444  * the memory.
445  */
446  static void auditd_conn_free(struct rcu_head *rcu)
447  {
448 	struct auditd_connection *ac;
449 
450 	ac = container_of(rcu, struct auditd_connection, rcu);
451 	put_pid(ac->pid);
452 	put_net(ac->net);
453 	kfree(ac);
454  }
455 
456 /**
457  * auditd_set - Set/Reset the auditd connection state
458  * @pid: auditd PID
459  * @portid: auditd netlink portid
460  * @net: auditd network namespace pointer
461  *
462  * Description:
463  * This function will obtain and drop network namespace references as
464  * necessary.  Returns zero on success, negative values on failure.
465  */
466 static int auditd_set(struct pid *pid, u32 portid, struct net *net)
467 {
468 	unsigned long flags;
469 	struct auditd_connection *ac_old, *ac_new;
470 
471 	if (!pid || !net)
472 		return -EINVAL;
473 
474 	ac_new = kzalloc(sizeof(*ac_new), GFP_KERNEL);
475 	if (!ac_new)
476 		return -ENOMEM;
477 	ac_new->pid = get_pid(pid);
478 	ac_new->portid = portid;
479 	ac_new->net = get_net(net);
480 
481 	spin_lock_irqsave(&auditd_conn_lock, flags);
482 	ac_old = rcu_dereference_protected(auditd_conn,
483 					   lockdep_is_held(&auditd_conn_lock));
484 	rcu_assign_pointer(auditd_conn, ac_new);
485 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
486 
487 	if (ac_old)
488 		call_rcu(&ac_old->rcu, auditd_conn_free);
489 
490 	return 0;
491 }
492 
493 /**
494  * kauditd_print_skb - Print the audit record to the ring buffer
495  * @skb: audit record
496  *
497  * Whatever the reason, this packet may not make it to the auditd connection
498  * so write it via printk so the information isn't completely lost.
499  */
500 static void kauditd_printk_skb(struct sk_buff *skb)
501 {
502 	struct nlmsghdr *nlh = nlmsg_hdr(skb);
503 	char *data = nlmsg_data(nlh);
504 
505 	if (nlh->nlmsg_type != AUDIT_EOE && printk_ratelimit())
506 		pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
507 }
508 
509 /**
510  * kauditd_rehold_skb - Handle a audit record send failure in the hold queue
511  * @skb: audit record
512  *
513  * Description:
514  * This should only be used by the kauditd_thread when it fails to flush the
515  * hold queue.
516  */
517 static void kauditd_rehold_skb(struct sk_buff *skb)
518 {
519 	/* put the record back in the queue at the same place */
520 	skb_queue_head(&audit_hold_queue, skb);
521 }
522 
523 /**
524  * kauditd_hold_skb - Queue an audit record, waiting for auditd
525  * @skb: audit record
526  *
527  * Description:
528  * Queue the audit record, waiting for an instance of auditd.  When this
529  * function is called we haven't given up yet on sending the record, but things
530  * are not looking good.  The first thing we want to do is try to write the
531  * record via printk and then see if we want to try and hold on to the record
532  * and queue it, if we have room.  If we want to hold on to the record, but we
533  * don't have room, record a record lost message.
534  */
535 static void kauditd_hold_skb(struct sk_buff *skb)
536 {
537 	/* at this point it is uncertain if we will ever send this to auditd so
538 	 * try to send the message via printk before we go any further */
539 	kauditd_printk_skb(skb);
540 
541 	/* can we just silently drop the message? */
542 	if (!audit_default) {
543 		kfree_skb(skb);
544 		return;
545 	}
546 
547 	/* if we have room, queue the message */
548 	if (!audit_backlog_limit ||
549 	    skb_queue_len(&audit_hold_queue) < audit_backlog_limit) {
550 		skb_queue_tail(&audit_hold_queue, skb);
551 		return;
552 	}
553 
554 	/* we have no other options - drop the message */
555 	audit_log_lost("kauditd hold queue overflow");
556 	kfree_skb(skb);
557 }
558 
559 /**
560  * kauditd_retry_skb - Queue an audit record, attempt to send again to auditd
561  * @skb: audit record
562  *
563  * Description:
564  * Not as serious as kauditd_hold_skb() as we still have a connected auditd,
565  * but for some reason we are having problems sending it audit records so
566  * queue the given record and attempt to resend.
567  */
568 static void kauditd_retry_skb(struct sk_buff *skb)
569 {
570 	/* NOTE: because records should only live in the retry queue for a
571 	 * short period of time, before either being sent or moved to the hold
572 	 * queue, we don't currently enforce a limit on this queue */
573 	skb_queue_tail(&audit_retry_queue, skb);
574 }
575 
576 /**
577  * auditd_reset - Disconnect the auditd connection
578  * @ac: auditd connection state
579  *
580  * Description:
581  * Break the auditd/kauditd connection and move all the queued records into the
582  * hold queue in case auditd reconnects.  It is important to note that the @ac
583  * pointer should never be dereferenced inside this function as it may be NULL
584  * or invalid, you can only compare the memory address!  If @ac is NULL then
585  * the connection will always be reset.
586  */
587 static void auditd_reset(const struct auditd_connection *ac)
588 {
589 	unsigned long flags;
590 	struct sk_buff *skb;
591 	struct auditd_connection *ac_old;
592 
593 	/* if it isn't already broken, break the connection */
594 	spin_lock_irqsave(&auditd_conn_lock, flags);
595 	ac_old = rcu_dereference_protected(auditd_conn,
596 					   lockdep_is_held(&auditd_conn_lock));
597 	if (ac && ac != ac_old) {
598 		/* someone already registered a new auditd connection */
599 		spin_unlock_irqrestore(&auditd_conn_lock, flags);
600 		return;
601 	}
602 	rcu_assign_pointer(auditd_conn, NULL);
603 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
604 
605 	if (ac_old)
606 		call_rcu(&ac_old->rcu, auditd_conn_free);
607 
608 	/* flush the retry queue to the hold queue, but don't touch the main
609 	 * queue since we need to process that normally for multicast */
610 	while ((skb = skb_dequeue(&audit_retry_queue)))
611 		kauditd_hold_skb(skb);
612 }
613 
614 /**
615  * auditd_send_unicast_skb - Send a record via unicast to auditd
616  * @skb: audit record
617  *
618  * Description:
619  * Send a skb to the audit daemon, returns positive/zero values on success and
620  * negative values on failure; in all cases the skb will be consumed by this
621  * function.  If the send results in -ECONNREFUSED the connection with auditd
622  * will be reset.  This function may sleep so callers should not hold any locks
623  * where this would cause a problem.
624  */
625 static int auditd_send_unicast_skb(struct sk_buff *skb)
626 {
627 	int rc;
628 	u32 portid;
629 	struct net *net;
630 	struct sock *sk;
631 	struct auditd_connection *ac;
632 
633 	/* NOTE: we can't call netlink_unicast while in the RCU section so
634 	 *       take a reference to the network namespace and grab local
635 	 *       copies of the namespace, the sock, and the portid; the
636 	 *       namespace and sock aren't going to go away while we hold a
637 	 *       reference and if the portid does become invalid after the RCU
638 	 *       section netlink_unicast() should safely return an error */
639 
640 	rcu_read_lock();
641 	ac = rcu_dereference(auditd_conn);
642 	if (!ac) {
643 		rcu_read_unlock();
644 		rc = -ECONNREFUSED;
645 		goto err;
646 	}
647 	net = get_net(ac->net);
648 	sk = audit_get_sk(net);
649 	portid = ac->portid;
650 	rcu_read_unlock();
651 
652 	rc = netlink_unicast(sk, skb, portid, 0);
653 	put_net(net);
654 	if (rc < 0)
655 		goto err;
656 
657 	return rc;
658 
659 err:
660 	if (ac && rc == -ECONNREFUSED)
661 		auditd_reset(ac);
662 	return rc;
663 }
664 
665 /**
666  * kauditd_send_queue - Helper for kauditd_thread to flush skb queues
667  * @sk: the sending sock
668  * @portid: the netlink destination
669  * @queue: the skb queue to process
670  * @retry_limit: limit on number of netlink unicast failures
671  * @skb_hook: per-skb hook for additional processing
672  * @err_hook: hook called if the skb fails the netlink unicast send
673  *
674  * Description:
675  * Run through the given queue and attempt to send the audit records to auditd,
676  * returns zero on success, negative values on failure.  It is up to the caller
677  * to ensure that the @sk is valid for the duration of this function.
678  *
679  */
680 static int kauditd_send_queue(struct sock *sk, u32 portid,
681 			      struct sk_buff_head *queue,
682 			      unsigned int retry_limit,
683 			      void (*skb_hook)(struct sk_buff *skb),
684 			      void (*err_hook)(struct sk_buff *skb))
685 {
686 	int rc = 0;
687 	struct sk_buff *skb;
688 	static unsigned int failed = 0;
689 
690 	/* NOTE: kauditd_thread takes care of all our locking, we just use
691 	 *       the netlink info passed to us (e.g. sk and portid) */
692 
693 	while ((skb = skb_dequeue(queue))) {
694 		/* call the skb_hook for each skb we touch */
695 		if (skb_hook)
696 			(*skb_hook)(skb);
697 
698 		/* can we send to anyone via unicast? */
699 		if (!sk) {
700 			if (err_hook)
701 				(*err_hook)(skb);
702 			continue;
703 		}
704 
705 		/* grab an extra skb reference in case of error */
706 		skb_get(skb);
707 		rc = netlink_unicast(sk, skb, portid, 0);
708 		if (rc < 0) {
709 			/* fatal failure for our queue flush attempt? */
710 			if (++failed >= retry_limit ||
711 			    rc == -ECONNREFUSED || rc == -EPERM) {
712 				/* yes - error processing for the queue */
713 				sk = NULL;
714 				if (err_hook)
715 					(*err_hook)(skb);
716 				if (!skb_hook)
717 					goto out;
718 				/* keep processing with the skb_hook */
719 				continue;
720 			} else
721 				/* no - requeue to preserve ordering */
722 				skb_queue_head(queue, skb);
723 		} else {
724 			/* it worked - drop the extra reference and continue */
725 			consume_skb(skb);
726 			failed = 0;
727 		}
728 	}
729 
730 out:
731 	return (rc >= 0 ? 0 : rc);
732 }
733 
734 /*
735  * kauditd_send_multicast_skb - Send a record to any multicast listeners
736  * @skb: audit record
737  *
738  * Description:
739  * Write a multicast message to anyone listening in the initial network
740  * namespace.  This function doesn't consume an skb as might be expected since
741  * it has to copy it anyways.
742  */
743 static void kauditd_send_multicast_skb(struct sk_buff *skb)
744 {
745 	struct sk_buff *copy;
746 	struct sock *sock = audit_get_sk(&init_net);
747 	struct nlmsghdr *nlh;
748 
749 	/* NOTE: we are not taking an additional reference for init_net since
750 	 *       we don't have to worry about it going away */
751 
752 	if (!netlink_has_listeners(sock, AUDIT_NLGRP_READLOG))
753 		return;
754 
755 	/*
756 	 * The seemingly wasteful skb_copy() rather than bumping the refcount
757 	 * using skb_get() is necessary because non-standard mods are made to
758 	 * the skb by the original kaudit unicast socket send routine.  The
759 	 * existing auditd daemon assumes this breakage.  Fixing this would
760 	 * require co-ordinating a change in the established protocol between
761 	 * the kaudit kernel subsystem and the auditd userspace code.  There is
762 	 * no reason for new multicast clients to continue with this
763 	 * non-compliance.
764 	 */
765 	copy = skb_copy(skb, GFP_KERNEL);
766 	if (!copy)
767 		return;
768 	nlh = nlmsg_hdr(copy);
769 	nlh->nlmsg_len = skb->len;
770 
771 	nlmsg_multicast(sock, copy, 0, AUDIT_NLGRP_READLOG, GFP_KERNEL);
772 }
773 
774 /**
775  * kauditd_thread - Worker thread to send audit records to userspace
776  * @dummy: unused
777  */
778 static int kauditd_thread(void *dummy)
779 {
780 	int rc;
781 	u32 portid = 0;
782 	struct net *net = NULL;
783 	struct sock *sk = NULL;
784 	struct auditd_connection *ac;
785 
786 #define UNICAST_RETRIES 5
787 
788 	set_freezable();
789 	while (!kthread_should_stop()) {
790 		/* NOTE: see the lock comments in auditd_send_unicast_skb() */
791 		rcu_read_lock();
792 		ac = rcu_dereference(auditd_conn);
793 		if (!ac) {
794 			rcu_read_unlock();
795 			goto main_queue;
796 		}
797 		net = get_net(ac->net);
798 		sk = audit_get_sk(net);
799 		portid = ac->portid;
800 		rcu_read_unlock();
801 
802 		/* attempt to flush the hold queue */
803 		rc = kauditd_send_queue(sk, portid,
804 					&audit_hold_queue, UNICAST_RETRIES,
805 					NULL, kauditd_rehold_skb);
806 		if (ac && rc < 0) {
807 			sk = NULL;
808 			auditd_reset(ac);
809 			goto main_queue;
810 		}
811 
812 		/* attempt to flush the retry queue */
813 		rc = kauditd_send_queue(sk, portid,
814 					&audit_retry_queue, UNICAST_RETRIES,
815 					NULL, kauditd_hold_skb);
816 		if (ac && rc < 0) {
817 			sk = NULL;
818 			auditd_reset(ac);
819 			goto main_queue;
820 		}
821 
822 main_queue:
823 		/* process the main queue - do the multicast send and attempt
824 		 * unicast, dump failed record sends to the retry queue; if
825 		 * sk == NULL due to previous failures we will just do the
826 		 * multicast send and move the record to the hold queue */
827 		rc = kauditd_send_queue(sk, portid, &audit_queue, 1,
828 					kauditd_send_multicast_skb,
829 					(sk ?
830 					 kauditd_retry_skb : kauditd_hold_skb));
831 		if (ac && rc < 0)
832 			auditd_reset(ac);
833 		sk = NULL;
834 
835 		/* drop our netns reference, no auditd sends past this line */
836 		if (net) {
837 			put_net(net);
838 			net = NULL;
839 		}
840 
841 		/* we have processed all the queues so wake everyone */
842 		wake_up(&audit_backlog_wait);
843 
844 		/* NOTE: we want to wake up if there is anything on the queue,
845 		 *       regardless of if an auditd is connected, as we need to
846 		 *       do the multicast send and rotate records from the
847 		 *       main queue to the retry/hold queues */
848 		wait_event_freezable(kauditd_wait,
849 				     (skb_queue_len(&audit_queue) ? 1 : 0));
850 	}
851 
852 	return 0;
853 }
854 
855 int audit_send_list(void *_dest)
856 {
857 	struct audit_netlink_list *dest = _dest;
858 	struct sk_buff *skb;
859 	struct sock *sk = audit_get_sk(dest->net);
860 
861 	/* wait for parent to finish and send an ACK */
862 	mutex_lock(&audit_cmd_mutex);
863 	mutex_unlock(&audit_cmd_mutex);
864 
865 	while ((skb = __skb_dequeue(&dest->q)) != NULL)
866 		netlink_unicast(sk, skb, dest->portid, 0);
867 
868 	put_net(dest->net);
869 	kfree(dest);
870 
871 	return 0;
872 }
873 
874 struct sk_buff *audit_make_reply(int seq, int type, int done,
875 				 int multi, const void *payload, int size)
876 {
877 	struct sk_buff	*skb;
878 	struct nlmsghdr	*nlh;
879 	void		*data;
880 	int		flags = multi ? NLM_F_MULTI : 0;
881 	int		t     = done  ? NLMSG_DONE  : type;
882 
883 	skb = nlmsg_new(size, GFP_KERNEL);
884 	if (!skb)
885 		return NULL;
886 
887 	nlh	= nlmsg_put(skb, 0, seq, t, size, flags);
888 	if (!nlh)
889 		goto out_kfree_skb;
890 	data = nlmsg_data(nlh);
891 	memcpy(data, payload, size);
892 	return skb;
893 
894 out_kfree_skb:
895 	kfree_skb(skb);
896 	return NULL;
897 }
898 
899 static int audit_send_reply_thread(void *arg)
900 {
901 	struct audit_reply *reply = (struct audit_reply *)arg;
902 	struct sock *sk = audit_get_sk(reply->net);
903 
904 	mutex_lock(&audit_cmd_mutex);
905 	mutex_unlock(&audit_cmd_mutex);
906 
907 	/* Ignore failure. It'll only happen if the sender goes away,
908 	   because our timeout is set to infinite. */
909 	netlink_unicast(sk, reply->skb, reply->portid, 0);
910 	put_net(reply->net);
911 	kfree(reply);
912 	return 0;
913 }
914 
915 /**
916  * audit_send_reply - send an audit reply message via netlink
917  * @request_skb: skb of request we are replying to (used to target the reply)
918  * @seq: sequence number
919  * @type: audit message type
920  * @done: done (last) flag
921  * @multi: multi-part message flag
922  * @payload: payload data
923  * @size: payload size
924  *
925  * Allocates an skb, builds the netlink message, and sends it to the port id.
926  * No failure notifications.
927  */
928 static void audit_send_reply(struct sk_buff *request_skb, int seq, int type, int done,
929 			     int multi, const void *payload, int size)
930 {
931 	struct net *net = sock_net(NETLINK_CB(request_skb).sk);
932 	struct sk_buff *skb;
933 	struct task_struct *tsk;
934 	struct audit_reply *reply = kmalloc(sizeof(struct audit_reply),
935 					    GFP_KERNEL);
936 
937 	if (!reply)
938 		return;
939 
940 	skb = audit_make_reply(seq, type, done, multi, payload, size);
941 	if (!skb)
942 		goto out;
943 
944 	reply->net = get_net(net);
945 	reply->portid = NETLINK_CB(request_skb).portid;
946 	reply->skb = skb;
947 
948 	tsk = kthread_run(audit_send_reply_thread, reply, "audit_send_reply");
949 	if (!IS_ERR(tsk))
950 		return;
951 	kfree_skb(skb);
952 out:
953 	kfree(reply);
954 }
955 
956 /*
957  * Check for appropriate CAP_AUDIT_ capabilities on incoming audit
958  * control messages.
959  */
960 static int audit_netlink_ok(struct sk_buff *skb, u16 msg_type)
961 {
962 	int err = 0;
963 
964 	/* Only support initial user namespace for now. */
965 	/*
966 	 * We return ECONNREFUSED because it tricks userspace into thinking
967 	 * that audit was not configured into the kernel.  Lots of users
968 	 * configure their PAM stack (because that's what the distro does)
969 	 * to reject login if unable to send messages to audit.  If we return
970 	 * ECONNREFUSED the PAM stack thinks the kernel does not have audit
971 	 * configured in and will let login proceed.  If we return EPERM
972 	 * userspace will reject all logins.  This should be removed when we
973 	 * support non init namespaces!!
974 	 */
975 	if (current_user_ns() != &init_user_ns)
976 		return -ECONNREFUSED;
977 
978 	switch (msg_type) {
979 	case AUDIT_LIST:
980 	case AUDIT_ADD:
981 	case AUDIT_DEL:
982 		return -EOPNOTSUPP;
983 	case AUDIT_GET:
984 	case AUDIT_SET:
985 	case AUDIT_GET_FEATURE:
986 	case AUDIT_SET_FEATURE:
987 	case AUDIT_LIST_RULES:
988 	case AUDIT_ADD_RULE:
989 	case AUDIT_DEL_RULE:
990 	case AUDIT_SIGNAL_INFO:
991 	case AUDIT_TTY_GET:
992 	case AUDIT_TTY_SET:
993 	case AUDIT_TRIM:
994 	case AUDIT_MAKE_EQUIV:
995 		/* Only support auditd and auditctl in initial pid namespace
996 		 * for now. */
997 		if (task_active_pid_ns(current) != &init_pid_ns)
998 			return -EPERM;
999 
1000 		if (!netlink_capable(skb, CAP_AUDIT_CONTROL))
1001 			err = -EPERM;
1002 		break;
1003 	case AUDIT_USER:
1004 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1005 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1006 		if (!netlink_capable(skb, CAP_AUDIT_WRITE))
1007 			err = -EPERM;
1008 		break;
1009 	default:  /* bad msg */
1010 		err = -EINVAL;
1011 	}
1012 
1013 	return err;
1014 }
1015 
1016 static void audit_log_common_recv_msg(struct audit_buffer **ab, u16 msg_type)
1017 {
1018 	uid_t uid = from_kuid(&init_user_ns, current_uid());
1019 	pid_t pid = task_tgid_nr(current);
1020 
1021 	if (!audit_enabled && msg_type != AUDIT_USER_AVC) {
1022 		*ab = NULL;
1023 		return;
1024 	}
1025 
1026 	*ab = audit_log_start(NULL, GFP_KERNEL, msg_type);
1027 	if (unlikely(!*ab))
1028 		return;
1029 	audit_log_format(*ab, "pid=%d uid=%u", pid, uid);
1030 	audit_log_session_info(*ab);
1031 	audit_log_task_context(*ab);
1032 }
1033 
1034 int is_audit_feature_set(int i)
1035 {
1036 	return af.features & AUDIT_FEATURE_TO_MASK(i);
1037 }
1038 
1039 
1040 static int audit_get_feature(struct sk_buff *skb)
1041 {
1042 	u32 seq;
1043 
1044 	seq = nlmsg_hdr(skb)->nlmsg_seq;
1045 
1046 	audit_send_reply(skb, seq, AUDIT_GET_FEATURE, 0, 0, &af, sizeof(af));
1047 
1048 	return 0;
1049 }
1050 
1051 static void audit_log_feature_change(int which, u32 old_feature, u32 new_feature,
1052 				     u32 old_lock, u32 new_lock, int res)
1053 {
1054 	struct audit_buffer *ab;
1055 
1056 	if (audit_enabled == AUDIT_OFF)
1057 		return;
1058 
1059 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_FEATURE_CHANGE);
1060 	audit_log_task_info(ab, current);
1061 	audit_log_format(ab, " feature=%s old=%u new=%u old_lock=%u new_lock=%u res=%d",
1062 			 audit_feature_names[which], !!old_feature, !!new_feature,
1063 			 !!old_lock, !!new_lock, res);
1064 	audit_log_end(ab);
1065 }
1066 
1067 static int audit_set_feature(struct sk_buff *skb)
1068 {
1069 	struct audit_features *uaf;
1070 	int i;
1071 
1072 	BUILD_BUG_ON(AUDIT_LAST_FEATURE + 1 > ARRAY_SIZE(audit_feature_names));
1073 	uaf = nlmsg_data(nlmsg_hdr(skb));
1074 
1075 	/* if there is ever a version 2 we should handle that here */
1076 
1077 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1078 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1079 		u32 old_feature, new_feature, old_lock, new_lock;
1080 
1081 		/* if we are not changing this feature, move along */
1082 		if (!(feature & uaf->mask))
1083 			continue;
1084 
1085 		old_feature = af.features & feature;
1086 		new_feature = uaf->features & feature;
1087 		new_lock = (uaf->lock | af.lock) & feature;
1088 		old_lock = af.lock & feature;
1089 
1090 		/* are we changing a locked feature? */
1091 		if (old_lock && (new_feature != old_feature)) {
1092 			audit_log_feature_change(i, old_feature, new_feature,
1093 						 old_lock, new_lock, 0);
1094 			return -EPERM;
1095 		}
1096 	}
1097 	/* nothing invalid, do the changes */
1098 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1099 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1100 		u32 old_feature, new_feature, old_lock, new_lock;
1101 
1102 		/* if we are not changing this feature, move along */
1103 		if (!(feature & uaf->mask))
1104 			continue;
1105 
1106 		old_feature = af.features & feature;
1107 		new_feature = uaf->features & feature;
1108 		old_lock = af.lock & feature;
1109 		new_lock = (uaf->lock | af.lock) & feature;
1110 
1111 		if (new_feature != old_feature)
1112 			audit_log_feature_change(i, old_feature, new_feature,
1113 						 old_lock, new_lock, 1);
1114 
1115 		if (new_feature)
1116 			af.features |= feature;
1117 		else
1118 			af.features &= ~feature;
1119 		af.lock |= new_lock;
1120 	}
1121 
1122 	return 0;
1123 }
1124 
1125 static int audit_replace(struct pid *pid)
1126 {
1127 	pid_t pvnr;
1128 	struct sk_buff *skb;
1129 
1130 	pvnr = pid_vnr(pid);
1131 	skb = audit_make_reply(0, AUDIT_REPLACE, 0, 0, &pvnr, sizeof(pvnr));
1132 	if (!skb)
1133 		return -ENOMEM;
1134 	return auditd_send_unicast_skb(skb);
1135 }
1136 
1137 static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1138 {
1139 	u32			seq;
1140 	void			*data;
1141 	int			err;
1142 	struct audit_buffer	*ab;
1143 	u16			msg_type = nlh->nlmsg_type;
1144 	struct audit_sig_info   *sig_data;
1145 	char			*ctx = NULL;
1146 	u32			len;
1147 
1148 	err = audit_netlink_ok(skb, msg_type);
1149 	if (err)
1150 		return err;
1151 
1152 	seq  = nlh->nlmsg_seq;
1153 	data = nlmsg_data(nlh);
1154 
1155 	switch (msg_type) {
1156 	case AUDIT_GET: {
1157 		struct audit_status	s;
1158 		memset(&s, 0, sizeof(s));
1159 		s.enabled		= audit_enabled;
1160 		s.failure		= audit_failure;
1161 		/* NOTE: use pid_vnr() so the PID is relative to the current
1162 		 *       namespace */
1163 		s.pid			= auditd_pid_vnr();
1164 		s.rate_limit		= audit_rate_limit;
1165 		s.backlog_limit		= audit_backlog_limit;
1166 		s.lost			= atomic_read(&audit_lost);
1167 		s.backlog		= skb_queue_len(&audit_queue);
1168 		s.feature_bitmap	= AUDIT_FEATURE_BITMAP_ALL;
1169 		s.backlog_wait_time	= audit_backlog_wait_time;
1170 		audit_send_reply(skb, seq, AUDIT_GET, 0, 0, &s, sizeof(s));
1171 		break;
1172 	}
1173 	case AUDIT_SET: {
1174 		struct audit_status	s;
1175 		memset(&s, 0, sizeof(s));
1176 		/* guard against past and future API changes */
1177 		memcpy(&s, data, min_t(size_t, sizeof(s), nlmsg_len(nlh)));
1178 		if (s.mask & AUDIT_STATUS_ENABLED) {
1179 			err = audit_set_enabled(s.enabled);
1180 			if (err < 0)
1181 				return err;
1182 		}
1183 		if (s.mask & AUDIT_STATUS_FAILURE) {
1184 			err = audit_set_failure(s.failure);
1185 			if (err < 0)
1186 				return err;
1187 		}
1188 		if (s.mask & AUDIT_STATUS_PID) {
1189 			/* NOTE: we are using the vnr PID functions below
1190 			 *       because the s.pid value is relative to the
1191 			 *       namespace of the caller; at present this
1192 			 *       doesn't matter much since you can really only
1193 			 *       run auditd from the initial pid namespace, but
1194 			 *       something to keep in mind if this changes */
1195 			pid_t new_pid = s.pid;
1196 			pid_t auditd_pid;
1197 			struct pid *req_pid = task_tgid(current);
1198 
1199 			/* sanity check - PID values must match */
1200 			if (new_pid != pid_vnr(req_pid))
1201 				return -EINVAL;
1202 
1203 			/* test the auditd connection */
1204 			audit_replace(req_pid);
1205 
1206 			auditd_pid = auditd_pid_vnr();
1207 			/* only the current auditd can unregister itself */
1208 			if ((!new_pid) && (new_pid != auditd_pid)) {
1209 				audit_log_config_change("audit_pid", new_pid,
1210 							auditd_pid, 0);
1211 				return -EACCES;
1212 			}
1213 			/* replacing a healthy auditd is not allowed */
1214 			if (auditd_pid && new_pid) {
1215 				audit_log_config_change("audit_pid", new_pid,
1216 							auditd_pid, 0);
1217 				return -EEXIST;
1218 			}
1219 
1220 			if (new_pid) {
1221 				/* register a new auditd connection */
1222 				err = auditd_set(req_pid,
1223 						 NETLINK_CB(skb).portid,
1224 						 sock_net(NETLINK_CB(skb).sk));
1225 				if (audit_enabled != AUDIT_OFF)
1226 					audit_log_config_change("audit_pid",
1227 								new_pid,
1228 								auditd_pid,
1229 								err ? 0 : 1);
1230 				if (err)
1231 					return err;
1232 
1233 				/* try to process any backlog */
1234 				wake_up_interruptible(&kauditd_wait);
1235 			} else {
1236 				if (audit_enabled != AUDIT_OFF)
1237 					audit_log_config_change("audit_pid",
1238 								new_pid,
1239 								auditd_pid, 1);
1240 
1241 				/* unregister the auditd connection */
1242 				auditd_reset(NULL);
1243 			}
1244 		}
1245 		if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
1246 			err = audit_set_rate_limit(s.rate_limit);
1247 			if (err < 0)
1248 				return err;
1249 		}
1250 		if (s.mask & AUDIT_STATUS_BACKLOG_LIMIT) {
1251 			err = audit_set_backlog_limit(s.backlog_limit);
1252 			if (err < 0)
1253 				return err;
1254 		}
1255 		if (s.mask & AUDIT_STATUS_BACKLOG_WAIT_TIME) {
1256 			if (sizeof(s) > (size_t)nlh->nlmsg_len)
1257 				return -EINVAL;
1258 			if (s.backlog_wait_time > 10*AUDIT_BACKLOG_WAIT_TIME)
1259 				return -EINVAL;
1260 			err = audit_set_backlog_wait_time(s.backlog_wait_time);
1261 			if (err < 0)
1262 				return err;
1263 		}
1264 		if (s.mask == AUDIT_STATUS_LOST) {
1265 			u32 lost = atomic_xchg(&audit_lost, 0);
1266 
1267 			audit_log_config_change("lost", 0, lost, 1);
1268 			return lost;
1269 		}
1270 		break;
1271 	}
1272 	case AUDIT_GET_FEATURE:
1273 		err = audit_get_feature(skb);
1274 		if (err)
1275 			return err;
1276 		break;
1277 	case AUDIT_SET_FEATURE:
1278 		err = audit_set_feature(skb);
1279 		if (err)
1280 			return err;
1281 		break;
1282 	case AUDIT_USER:
1283 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1284 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1285 		if (!audit_enabled && msg_type != AUDIT_USER_AVC)
1286 			return 0;
1287 
1288 		err = audit_filter(msg_type, AUDIT_FILTER_USER);
1289 		if (err == 1) { /* match or error */
1290 			err = 0;
1291 			if (msg_type == AUDIT_USER_TTY) {
1292 				err = tty_audit_push();
1293 				if (err)
1294 					break;
1295 			}
1296 			audit_log_common_recv_msg(&ab, msg_type);
1297 			if (msg_type != AUDIT_USER_TTY)
1298 				audit_log_format(ab, " msg='%.*s'",
1299 						 AUDIT_MESSAGE_TEXT_MAX,
1300 						 (char *)data);
1301 			else {
1302 				int size;
1303 
1304 				audit_log_format(ab, " data=");
1305 				size = nlmsg_len(nlh);
1306 				if (size > 0 &&
1307 				    ((unsigned char *)data)[size - 1] == '\0')
1308 					size--;
1309 				audit_log_n_untrustedstring(ab, data, size);
1310 			}
1311 			audit_log_end(ab);
1312 		}
1313 		break;
1314 	case AUDIT_ADD_RULE:
1315 	case AUDIT_DEL_RULE:
1316 		if (nlmsg_len(nlh) < sizeof(struct audit_rule_data))
1317 			return -EINVAL;
1318 		if (audit_enabled == AUDIT_LOCKED) {
1319 			audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1320 			audit_log_format(ab, " audit_enabled=%d res=0", audit_enabled);
1321 			audit_log_end(ab);
1322 			return -EPERM;
1323 		}
1324 		err = audit_rule_change(msg_type, seq, data, nlmsg_len(nlh));
1325 		break;
1326 	case AUDIT_LIST_RULES:
1327 		err = audit_list_rules_send(skb, seq);
1328 		break;
1329 	case AUDIT_TRIM:
1330 		audit_trim_trees();
1331 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1332 		audit_log_format(ab, " op=trim res=1");
1333 		audit_log_end(ab);
1334 		break;
1335 	case AUDIT_MAKE_EQUIV: {
1336 		void *bufp = data;
1337 		u32 sizes[2];
1338 		size_t msglen = nlmsg_len(nlh);
1339 		char *old, *new;
1340 
1341 		err = -EINVAL;
1342 		if (msglen < 2 * sizeof(u32))
1343 			break;
1344 		memcpy(sizes, bufp, 2 * sizeof(u32));
1345 		bufp += 2 * sizeof(u32);
1346 		msglen -= 2 * sizeof(u32);
1347 		old = audit_unpack_string(&bufp, &msglen, sizes[0]);
1348 		if (IS_ERR(old)) {
1349 			err = PTR_ERR(old);
1350 			break;
1351 		}
1352 		new = audit_unpack_string(&bufp, &msglen, sizes[1]);
1353 		if (IS_ERR(new)) {
1354 			err = PTR_ERR(new);
1355 			kfree(old);
1356 			break;
1357 		}
1358 		/* OK, here comes... */
1359 		err = audit_tag_tree(old, new);
1360 
1361 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1362 
1363 		audit_log_format(ab, " op=make_equiv old=");
1364 		audit_log_untrustedstring(ab, old);
1365 		audit_log_format(ab, " new=");
1366 		audit_log_untrustedstring(ab, new);
1367 		audit_log_format(ab, " res=%d", !err);
1368 		audit_log_end(ab);
1369 		kfree(old);
1370 		kfree(new);
1371 		break;
1372 	}
1373 	case AUDIT_SIGNAL_INFO:
1374 		len = 0;
1375 		if (audit_sig_sid) {
1376 			err = security_secid_to_secctx(audit_sig_sid, &ctx, &len);
1377 			if (err)
1378 				return err;
1379 		}
1380 		sig_data = kmalloc(sizeof(*sig_data) + len, GFP_KERNEL);
1381 		if (!sig_data) {
1382 			if (audit_sig_sid)
1383 				security_release_secctx(ctx, len);
1384 			return -ENOMEM;
1385 		}
1386 		sig_data->uid = from_kuid(&init_user_ns, audit_sig_uid);
1387 		sig_data->pid = audit_sig_pid;
1388 		if (audit_sig_sid) {
1389 			memcpy(sig_data->ctx, ctx, len);
1390 			security_release_secctx(ctx, len);
1391 		}
1392 		audit_send_reply(skb, seq, AUDIT_SIGNAL_INFO, 0, 0,
1393 				 sig_data, sizeof(*sig_data) + len);
1394 		kfree(sig_data);
1395 		break;
1396 	case AUDIT_TTY_GET: {
1397 		struct audit_tty_status s;
1398 		unsigned int t;
1399 
1400 		t = READ_ONCE(current->signal->audit_tty);
1401 		s.enabled = t & AUDIT_TTY_ENABLE;
1402 		s.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1403 
1404 		audit_send_reply(skb, seq, AUDIT_TTY_GET, 0, 0, &s, sizeof(s));
1405 		break;
1406 	}
1407 	case AUDIT_TTY_SET: {
1408 		struct audit_tty_status s, old;
1409 		struct audit_buffer	*ab;
1410 		unsigned int t;
1411 
1412 		memset(&s, 0, sizeof(s));
1413 		/* guard against past and future API changes */
1414 		memcpy(&s, data, min_t(size_t, sizeof(s), nlmsg_len(nlh)));
1415 		/* check if new data is valid */
1416 		if ((s.enabled != 0 && s.enabled != 1) ||
1417 		    (s.log_passwd != 0 && s.log_passwd != 1))
1418 			err = -EINVAL;
1419 
1420 		if (err)
1421 			t = READ_ONCE(current->signal->audit_tty);
1422 		else {
1423 			t = s.enabled | (-s.log_passwd & AUDIT_TTY_LOG_PASSWD);
1424 			t = xchg(&current->signal->audit_tty, t);
1425 		}
1426 		old.enabled = t & AUDIT_TTY_ENABLE;
1427 		old.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1428 
1429 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1430 		audit_log_format(ab, " op=tty_set old-enabled=%d new-enabled=%d"
1431 				 " old-log_passwd=%d new-log_passwd=%d res=%d",
1432 				 old.enabled, s.enabled, old.log_passwd,
1433 				 s.log_passwd, !err);
1434 		audit_log_end(ab);
1435 		break;
1436 	}
1437 	default:
1438 		err = -EINVAL;
1439 		break;
1440 	}
1441 
1442 	return err < 0 ? err : 0;
1443 }
1444 
1445 /**
1446  * audit_receive - receive messages from a netlink control socket
1447  * @skb: the message buffer
1448  *
1449  * Parse the provided skb and deal with any messages that may be present,
1450  * malformed skbs are discarded.
1451  */
1452 static void audit_receive(struct sk_buff  *skb)
1453 {
1454 	struct nlmsghdr *nlh;
1455 	/*
1456 	 * len MUST be signed for nlmsg_next to be able to dec it below 0
1457 	 * if the nlmsg_len was not aligned
1458 	 */
1459 	int len;
1460 	int err;
1461 
1462 	nlh = nlmsg_hdr(skb);
1463 	len = skb->len;
1464 
1465 	mutex_lock(&audit_cmd_mutex);
1466 	while (nlmsg_ok(nlh, len)) {
1467 		err = audit_receive_msg(skb, nlh);
1468 		/* if err or if this message says it wants a response */
1469 		if (err || (nlh->nlmsg_flags & NLM_F_ACK))
1470 			netlink_ack(skb, nlh, err, NULL);
1471 
1472 		nlh = nlmsg_next(nlh, &len);
1473 	}
1474 	mutex_unlock(&audit_cmd_mutex);
1475 }
1476 
1477 /* Run custom bind function on netlink socket group connect or bind requests. */
1478 static int audit_bind(struct net *net, int group)
1479 {
1480 	if (!capable(CAP_AUDIT_READ))
1481 		return -EPERM;
1482 
1483 	return 0;
1484 }
1485 
1486 static int __net_init audit_net_init(struct net *net)
1487 {
1488 	struct netlink_kernel_cfg cfg = {
1489 		.input	= audit_receive,
1490 		.bind	= audit_bind,
1491 		.flags	= NL_CFG_F_NONROOT_RECV,
1492 		.groups	= AUDIT_NLGRP_MAX,
1493 	};
1494 
1495 	struct audit_net *aunet = net_generic(net, audit_net_id);
1496 
1497 	aunet->sk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
1498 	if (aunet->sk == NULL) {
1499 		audit_panic("cannot initialize netlink socket in namespace");
1500 		return -ENOMEM;
1501 	}
1502 	aunet->sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1503 
1504 	return 0;
1505 }
1506 
1507 static void __net_exit audit_net_exit(struct net *net)
1508 {
1509 	struct audit_net *aunet = net_generic(net, audit_net_id);
1510 
1511 	/* NOTE: you would think that we would want to check the auditd
1512 	 * connection and potentially reset it here if it lives in this
1513 	 * namespace, but since the auditd connection tracking struct holds a
1514 	 * reference to this namespace (see auditd_set()) we are only ever
1515 	 * going to get here after that connection has been released */
1516 
1517 	netlink_kernel_release(aunet->sk);
1518 }
1519 
1520 static struct pernet_operations audit_net_ops __net_initdata = {
1521 	.init = audit_net_init,
1522 	.exit = audit_net_exit,
1523 	.id = &audit_net_id,
1524 	.size = sizeof(struct audit_net),
1525 };
1526 
1527 /* Initialize audit support at boot time. */
1528 static int __init audit_init(void)
1529 {
1530 	int i;
1531 
1532 	if (audit_initialized == AUDIT_DISABLED)
1533 		return 0;
1534 
1535 	audit_buffer_cache = kmem_cache_create("audit_buffer",
1536 					       sizeof(struct audit_buffer),
1537 					       0, SLAB_PANIC, NULL);
1538 
1539 	skb_queue_head_init(&audit_queue);
1540 	skb_queue_head_init(&audit_retry_queue);
1541 	skb_queue_head_init(&audit_hold_queue);
1542 
1543 	for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
1544 		INIT_LIST_HEAD(&audit_inode_hash[i]);
1545 
1546 	pr_info("initializing netlink subsys (%s)\n",
1547 		audit_default ? "enabled" : "disabled");
1548 	register_pernet_subsys(&audit_net_ops);
1549 
1550 	audit_initialized = AUDIT_INITIALIZED;
1551 	audit_enabled = audit_default;
1552 	audit_ever_enabled |= !!audit_default;
1553 
1554 	kauditd_task = kthread_run(kauditd_thread, NULL, "kauditd");
1555 	if (IS_ERR(kauditd_task)) {
1556 		int err = PTR_ERR(kauditd_task);
1557 		panic("audit: failed to start the kauditd thread (%d)\n", err);
1558 	}
1559 
1560 	audit_log(NULL, GFP_KERNEL, AUDIT_KERNEL,
1561 		"state=initialized audit_enabled=%u res=1",
1562 		 audit_enabled);
1563 
1564 	return 0;
1565 }
1566 __initcall(audit_init);
1567 
1568 /* Process kernel command-line parameter at boot time.  audit=0 or audit=1. */
1569 static int __init audit_enable(char *str)
1570 {
1571 	audit_default = !!simple_strtol(str, NULL, 0);
1572 	if (!audit_default)
1573 		audit_initialized = AUDIT_DISABLED;
1574 
1575 	pr_info("%s\n", audit_default ?
1576 		"enabled (after initialization)" : "disabled (until reboot)");
1577 
1578 	return 1;
1579 }
1580 __setup("audit=", audit_enable);
1581 
1582 /* Process kernel command-line parameter at boot time.
1583  * audit_backlog_limit=<n> */
1584 static int __init audit_backlog_limit_set(char *str)
1585 {
1586 	u32 audit_backlog_limit_arg;
1587 
1588 	pr_info("audit_backlog_limit: ");
1589 	if (kstrtouint(str, 0, &audit_backlog_limit_arg)) {
1590 		pr_cont("using default of %u, unable to parse %s\n",
1591 			audit_backlog_limit, str);
1592 		return 1;
1593 	}
1594 
1595 	audit_backlog_limit = audit_backlog_limit_arg;
1596 	pr_cont("%d\n", audit_backlog_limit);
1597 
1598 	return 1;
1599 }
1600 __setup("audit_backlog_limit=", audit_backlog_limit_set);
1601 
1602 static void audit_buffer_free(struct audit_buffer *ab)
1603 {
1604 	if (!ab)
1605 		return;
1606 
1607 	kfree_skb(ab->skb);
1608 	kmem_cache_free(audit_buffer_cache, ab);
1609 }
1610 
1611 static struct audit_buffer *audit_buffer_alloc(struct audit_context *ctx,
1612 					       gfp_t gfp_mask, int type)
1613 {
1614 	struct audit_buffer *ab;
1615 
1616 	ab = kmem_cache_alloc(audit_buffer_cache, gfp_mask);
1617 	if (!ab)
1618 		return NULL;
1619 
1620 	ab->skb = nlmsg_new(AUDIT_BUFSIZ, gfp_mask);
1621 	if (!ab->skb)
1622 		goto err;
1623 	if (!nlmsg_put(ab->skb, 0, 0, type, 0, 0))
1624 		goto err;
1625 
1626 	ab->ctx = ctx;
1627 	ab->gfp_mask = gfp_mask;
1628 
1629 	return ab;
1630 
1631 err:
1632 	audit_buffer_free(ab);
1633 	return NULL;
1634 }
1635 
1636 /**
1637  * audit_serial - compute a serial number for the audit record
1638  *
1639  * Compute a serial number for the audit record.  Audit records are
1640  * written to user-space as soon as they are generated, so a complete
1641  * audit record may be written in several pieces.  The timestamp of the
1642  * record and this serial number are used by the user-space tools to
1643  * determine which pieces belong to the same audit record.  The
1644  * (timestamp,serial) tuple is unique for each syscall and is live from
1645  * syscall entry to syscall exit.
1646  *
1647  * NOTE: Another possibility is to store the formatted records off the
1648  * audit context (for those records that have a context), and emit them
1649  * all at syscall exit.  However, this could delay the reporting of
1650  * significant errors until syscall exit (or never, if the system
1651  * halts).
1652  */
1653 unsigned int audit_serial(void)
1654 {
1655 	static atomic_t serial = ATOMIC_INIT(0);
1656 
1657 	return atomic_add_return(1, &serial);
1658 }
1659 
1660 static inline void audit_get_stamp(struct audit_context *ctx,
1661 				   struct timespec64 *t, unsigned int *serial)
1662 {
1663 	if (!ctx || !auditsc_get_stamp(ctx, t, serial)) {
1664 		ktime_get_real_ts64(t);
1665 		*serial = audit_serial();
1666 	}
1667 }
1668 
1669 /**
1670  * audit_log_start - obtain an audit buffer
1671  * @ctx: audit_context (may be NULL)
1672  * @gfp_mask: type of allocation
1673  * @type: audit message type
1674  *
1675  * Returns audit_buffer pointer on success or NULL on error.
1676  *
1677  * Obtain an audit buffer.  This routine does locking to obtain the
1678  * audit buffer, but then no locking is required for calls to
1679  * audit_log_*format.  If the task (ctx) is a task that is currently in a
1680  * syscall, then the syscall is marked as auditable and an audit record
1681  * will be written at syscall exit.  If there is no associated task, then
1682  * task context (ctx) should be NULL.
1683  */
1684 struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
1685 				     int type)
1686 {
1687 	struct audit_buffer *ab;
1688 	struct timespec64 t;
1689 	unsigned int uninitialized_var(serial);
1690 
1691 	if (audit_initialized != AUDIT_INITIALIZED)
1692 		return NULL;
1693 
1694 	if (unlikely(!audit_filter(type, AUDIT_FILTER_TYPE)))
1695 		return NULL;
1696 
1697 	/* NOTE: don't ever fail/sleep on these two conditions:
1698 	 * 1. auditd generated record - since we need auditd to drain the
1699 	 *    queue; also, when we are checking for auditd, compare PIDs using
1700 	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
1701 	 *    using a PID anchored in the caller's namespace
1702 	 * 2. generator holding the audit_cmd_mutex - we don't want to block
1703 	 *    while holding the mutex */
1704 	if (!(auditd_test_task(current) ||
1705 	      (current == __mutex_owner(&audit_cmd_mutex)))) {
1706 		long stime = audit_backlog_wait_time;
1707 
1708 		while (audit_backlog_limit &&
1709 		       (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
1710 			/* wake kauditd to try and flush the queue */
1711 			wake_up_interruptible(&kauditd_wait);
1712 
1713 			/* sleep if we are allowed and we haven't exhausted our
1714 			 * backlog wait limit */
1715 			if (gfpflags_allow_blocking(gfp_mask) && (stime > 0)) {
1716 				DECLARE_WAITQUEUE(wait, current);
1717 
1718 				add_wait_queue_exclusive(&audit_backlog_wait,
1719 							 &wait);
1720 				set_current_state(TASK_UNINTERRUPTIBLE);
1721 				stime = schedule_timeout(stime);
1722 				remove_wait_queue(&audit_backlog_wait, &wait);
1723 			} else {
1724 				if (audit_rate_check() && printk_ratelimit())
1725 					pr_warn("audit_backlog=%d > audit_backlog_limit=%d\n",
1726 						skb_queue_len(&audit_queue),
1727 						audit_backlog_limit);
1728 				audit_log_lost("backlog limit exceeded");
1729 				return NULL;
1730 			}
1731 		}
1732 	}
1733 
1734 	ab = audit_buffer_alloc(ctx, gfp_mask, type);
1735 	if (!ab) {
1736 		audit_log_lost("out of memory in audit_log_start");
1737 		return NULL;
1738 	}
1739 
1740 	audit_get_stamp(ab->ctx, &t, &serial);
1741 	audit_log_format(ab, "audit(%llu.%03lu:%u): ",
1742 			 (unsigned long long)t.tv_sec, t.tv_nsec/1000000, serial);
1743 
1744 	return ab;
1745 }
1746 
1747 /**
1748  * audit_expand - expand skb in the audit buffer
1749  * @ab: audit_buffer
1750  * @extra: space to add at tail of the skb
1751  *
1752  * Returns 0 (no space) on failed expansion, or available space if
1753  * successful.
1754  */
1755 static inline int audit_expand(struct audit_buffer *ab, int extra)
1756 {
1757 	struct sk_buff *skb = ab->skb;
1758 	int oldtail = skb_tailroom(skb);
1759 	int ret = pskb_expand_head(skb, 0, extra, ab->gfp_mask);
1760 	int newtail = skb_tailroom(skb);
1761 
1762 	if (ret < 0) {
1763 		audit_log_lost("out of memory in audit_expand");
1764 		return 0;
1765 	}
1766 
1767 	skb->truesize += newtail - oldtail;
1768 	return newtail;
1769 }
1770 
1771 /*
1772  * Format an audit message into the audit buffer.  If there isn't enough
1773  * room in the audit buffer, more room will be allocated and vsnprint
1774  * will be called a second time.  Currently, we assume that a printk
1775  * can't format message larger than 1024 bytes, so we don't either.
1776  */
1777 static void audit_log_vformat(struct audit_buffer *ab, const char *fmt,
1778 			      va_list args)
1779 {
1780 	int len, avail;
1781 	struct sk_buff *skb;
1782 	va_list args2;
1783 
1784 	if (!ab)
1785 		return;
1786 
1787 	BUG_ON(!ab->skb);
1788 	skb = ab->skb;
1789 	avail = skb_tailroom(skb);
1790 	if (avail == 0) {
1791 		avail = audit_expand(ab, AUDIT_BUFSIZ);
1792 		if (!avail)
1793 			goto out;
1794 	}
1795 	va_copy(args2, args);
1796 	len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args);
1797 	if (len >= avail) {
1798 		/* The printk buffer is 1024 bytes long, so if we get
1799 		 * here and AUDIT_BUFSIZ is at least 1024, then we can
1800 		 * log everything that printk could have logged. */
1801 		avail = audit_expand(ab,
1802 			max_t(unsigned, AUDIT_BUFSIZ, 1+len-avail));
1803 		if (!avail)
1804 			goto out_va_end;
1805 		len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args2);
1806 	}
1807 	if (len > 0)
1808 		skb_put(skb, len);
1809 out_va_end:
1810 	va_end(args2);
1811 out:
1812 	return;
1813 }
1814 
1815 /**
1816  * audit_log_format - format a message into the audit buffer.
1817  * @ab: audit_buffer
1818  * @fmt: format string
1819  * @...: optional parameters matching @fmt string
1820  *
1821  * All the work is done in audit_log_vformat.
1822  */
1823 void audit_log_format(struct audit_buffer *ab, const char *fmt, ...)
1824 {
1825 	va_list args;
1826 
1827 	if (!ab)
1828 		return;
1829 	va_start(args, fmt);
1830 	audit_log_vformat(ab, fmt, args);
1831 	va_end(args);
1832 }
1833 
1834 /**
1835  * audit_log_hex - convert a buffer to hex and append it to the audit skb
1836  * @ab: the audit_buffer
1837  * @buf: buffer to convert to hex
1838  * @len: length of @buf to be converted
1839  *
1840  * No return value; failure to expand is silently ignored.
1841  *
1842  * This function will take the passed buf and convert it into a string of
1843  * ascii hex digits. The new string is placed onto the skb.
1844  */
1845 void audit_log_n_hex(struct audit_buffer *ab, const unsigned char *buf,
1846 		size_t len)
1847 {
1848 	int i, avail, new_len;
1849 	unsigned char *ptr;
1850 	struct sk_buff *skb;
1851 
1852 	if (!ab)
1853 		return;
1854 
1855 	BUG_ON(!ab->skb);
1856 	skb = ab->skb;
1857 	avail = skb_tailroom(skb);
1858 	new_len = len<<1;
1859 	if (new_len >= avail) {
1860 		/* Round the buffer request up to the next multiple */
1861 		new_len = AUDIT_BUFSIZ*(((new_len-avail)/AUDIT_BUFSIZ) + 1);
1862 		avail = audit_expand(ab, new_len);
1863 		if (!avail)
1864 			return;
1865 	}
1866 
1867 	ptr = skb_tail_pointer(skb);
1868 	for (i = 0; i < len; i++)
1869 		ptr = hex_byte_pack_upper(ptr, buf[i]);
1870 	*ptr = 0;
1871 	skb_put(skb, len << 1); /* new string is twice the old string */
1872 }
1873 
1874 /*
1875  * Format a string of no more than slen characters into the audit buffer,
1876  * enclosed in quote marks.
1877  */
1878 void audit_log_n_string(struct audit_buffer *ab, const char *string,
1879 			size_t slen)
1880 {
1881 	int avail, new_len;
1882 	unsigned char *ptr;
1883 	struct sk_buff *skb;
1884 
1885 	if (!ab)
1886 		return;
1887 
1888 	BUG_ON(!ab->skb);
1889 	skb = ab->skb;
1890 	avail = skb_tailroom(skb);
1891 	new_len = slen + 3;	/* enclosing quotes + null terminator */
1892 	if (new_len > avail) {
1893 		avail = audit_expand(ab, new_len);
1894 		if (!avail)
1895 			return;
1896 	}
1897 	ptr = skb_tail_pointer(skb);
1898 	*ptr++ = '"';
1899 	memcpy(ptr, string, slen);
1900 	ptr += slen;
1901 	*ptr++ = '"';
1902 	*ptr = 0;
1903 	skb_put(skb, slen + 2);	/* don't include null terminator */
1904 }
1905 
1906 /**
1907  * audit_string_contains_control - does a string need to be logged in hex
1908  * @string: string to be checked
1909  * @len: max length of the string to check
1910  */
1911 bool audit_string_contains_control(const char *string, size_t len)
1912 {
1913 	const unsigned char *p;
1914 	for (p = string; p < (const unsigned char *)string + len; p++) {
1915 		if (*p == '"' || *p < 0x21 || *p > 0x7e)
1916 			return true;
1917 	}
1918 	return false;
1919 }
1920 
1921 /**
1922  * audit_log_n_untrustedstring - log a string that may contain random characters
1923  * @ab: audit_buffer
1924  * @len: length of string (not including trailing null)
1925  * @string: string to be logged
1926  *
1927  * This code will escape a string that is passed to it if the string
1928  * contains a control character, unprintable character, double quote mark,
1929  * or a space. Unescaped strings will start and end with a double quote mark.
1930  * Strings that are escaped are printed in hex (2 digits per char).
1931  *
1932  * The caller specifies the number of characters in the string to log, which may
1933  * or may not be the entire string.
1934  */
1935 void audit_log_n_untrustedstring(struct audit_buffer *ab, const char *string,
1936 				 size_t len)
1937 {
1938 	if (audit_string_contains_control(string, len))
1939 		audit_log_n_hex(ab, string, len);
1940 	else
1941 		audit_log_n_string(ab, string, len);
1942 }
1943 
1944 /**
1945  * audit_log_untrustedstring - log a string that may contain random characters
1946  * @ab: audit_buffer
1947  * @string: string to be logged
1948  *
1949  * Same as audit_log_n_untrustedstring(), except that strlen is used to
1950  * determine string length.
1951  */
1952 void audit_log_untrustedstring(struct audit_buffer *ab, const char *string)
1953 {
1954 	audit_log_n_untrustedstring(ab, string, strlen(string));
1955 }
1956 
1957 /* This is a helper-function to print the escaped d_path */
1958 void audit_log_d_path(struct audit_buffer *ab, const char *prefix,
1959 		      const struct path *path)
1960 {
1961 	char *p, *pathname;
1962 
1963 	if (prefix)
1964 		audit_log_format(ab, "%s", prefix);
1965 
1966 	/* We will allow 11 spaces for ' (deleted)' to be appended */
1967 	pathname = kmalloc(PATH_MAX+11, ab->gfp_mask);
1968 	if (!pathname) {
1969 		audit_log_string(ab, "<no_memory>");
1970 		return;
1971 	}
1972 	p = d_path(path, pathname, PATH_MAX+11);
1973 	if (IS_ERR(p)) { /* Should never happen since we send PATH_MAX */
1974 		/* FIXME: can we save some information here? */
1975 		audit_log_string(ab, "<too_long>");
1976 	} else
1977 		audit_log_untrustedstring(ab, p);
1978 	kfree(pathname);
1979 }
1980 
1981 void audit_log_session_info(struct audit_buffer *ab)
1982 {
1983 	unsigned int sessionid = audit_get_sessionid(current);
1984 	uid_t auid = from_kuid(&init_user_ns, audit_get_loginuid(current));
1985 
1986 	audit_log_format(ab, " auid=%u ses=%u", auid, sessionid);
1987 }
1988 
1989 void audit_log_key(struct audit_buffer *ab, char *key)
1990 {
1991 	audit_log_format(ab, " key=");
1992 	if (key)
1993 		audit_log_untrustedstring(ab, key);
1994 	else
1995 		audit_log_format(ab, "(null)");
1996 }
1997 
1998 void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1999 {
2000 	int i;
2001 
2002 	audit_log_format(ab, " %s=", prefix);
2003 	CAP_FOR_EACH_U32(i) {
2004 		audit_log_format(ab, "%08x",
2005 				 cap->cap[CAP_LAST_U32 - i]);
2006 	}
2007 }
2008 
2009 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
2010 {
2011 	audit_log_cap(ab, "cap_fp", &name->fcap.permitted);
2012 	audit_log_cap(ab, "cap_fi", &name->fcap.inheritable);
2013 	audit_log_format(ab, " cap_fe=%d cap_fver=%x",
2014 			 name->fcap.fE, name->fcap_ver);
2015 }
2016 
2017 static inline int audit_copy_fcaps(struct audit_names *name,
2018 				   const struct dentry *dentry)
2019 {
2020 	struct cpu_vfs_cap_data caps;
2021 	int rc;
2022 
2023 	if (!dentry)
2024 		return 0;
2025 
2026 	rc = get_vfs_caps_from_disk(dentry, &caps);
2027 	if (rc)
2028 		return rc;
2029 
2030 	name->fcap.permitted = caps.permitted;
2031 	name->fcap.inheritable = caps.inheritable;
2032 	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2033 	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >>
2034 				VFS_CAP_REVISION_SHIFT;
2035 
2036 	return 0;
2037 }
2038 
2039 /* Copy inode data into an audit_names. */
2040 void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
2041 		      struct inode *inode)
2042 {
2043 	name->ino   = inode->i_ino;
2044 	name->dev   = inode->i_sb->s_dev;
2045 	name->mode  = inode->i_mode;
2046 	name->uid   = inode->i_uid;
2047 	name->gid   = inode->i_gid;
2048 	name->rdev  = inode->i_rdev;
2049 	security_inode_getsecid(inode, &name->osid);
2050 	audit_copy_fcaps(name, dentry);
2051 }
2052 
2053 /**
2054  * audit_log_name - produce AUDIT_PATH record from struct audit_names
2055  * @context: audit_context for the task
2056  * @n: audit_names structure with reportable details
2057  * @path: optional path to report instead of audit_names->name
2058  * @record_num: record number to report when handling a list of names
2059  * @call_panic: optional pointer to int that will be updated if secid fails
2060  */
2061 void audit_log_name(struct audit_context *context, struct audit_names *n,
2062 		    const struct path *path, int record_num, int *call_panic)
2063 {
2064 	struct audit_buffer *ab;
2065 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
2066 	if (!ab)
2067 		return;
2068 
2069 	audit_log_format(ab, "item=%d", record_num);
2070 
2071 	if (path)
2072 		audit_log_d_path(ab, " name=", path);
2073 	else if (n->name) {
2074 		switch (n->name_len) {
2075 		case AUDIT_NAME_FULL:
2076 			/* log the full path */
2077 			audit_log_format(ab, " name=");
2078 			audit_log_untrustedstring(ab, n->name->name);
2079 			break;
2080 		case 0:
2081 			/* name was specified as a relative path and the
2082 			 * directory component is the cwd */
2083 			audit_log_d_path(ab, " name=", &context->pwd);
2084 			break;
2085 		default:
2086 			/* log the name's directory component */
2087 			audit_log_format(ab, " name=");
2088 			audit_log_n_untrustedstring(ab, n->name->name,
2089 						    n->name_len);
2090 		}
2091 	} else
2092 		audit_log_format(ab, " name=(null)");
2093 
2094 	if (n->ino != AUDIT_INO_UNSET)
2095 		audit_log_format(ab, " inode=%lu"
2096 				 " dev=%02x:%02x mode=%#ho"
2097 				 " ouid=%u ogid=%u rdev=%02x:%02x",
2098 				 n->ino,
2099 				 MAJOR(n->dev),
2100 				 MINOR(n->dev),
2101 				 n->mode,
2102 				 from_kuid(&init_user_ns, n->uid),
2103 				 from_kgid(&init_user_ns, n->gid),
2104 				 MAJOR(n->rdev),
2105 				 MINOR(n->rdev));
2106 	if (n->osid != 0) {
2107 		char *ctx = NULL;
2108 		u32 len;
2109 		if (security_secid_to_secctx(
2110 			n->osid, &ctx, &len)) {
2111 			audit_log_format(ab, " osid=%u", n->osid);
2112 			if (call_panic)
2113 				*call_panic = 2;
2114 		} else {
2115 			audit_log_format(ab, " obj=%s", ctx);
2116 			security_release_secctx(ctx, len);
2117 		}
2118 	}
2119 
2120 	/* log the audit_names record type */
2121 	audit_log_format(ab, " nametype=");
2122 	switch(n->type) {
2123 	case AUDIT_TYPE_NORMAL:
2124 		audit_log_format(ab, "NORMAL");
2125 		break;
2126 	case AUDIT_TYPE_PARENT:
2127 		audit_log_format(ab, "PARENT");
2128 		break;
2129 	case AUDIT_TYPE_CHILD_DELETE:
2130 		audit_log_format(ab, "DELETE");
2131 		break;
2132 	case AUDIT_TYPE_CHILD_CREATE:
2133 		audit_log_format(ab, "CREATE");
2134 		break;
2135 	default:
2136 		audit_log_format(ab, "UNKNOWN");
2137 		break;
2138 	}
2139 
2140 	audit_log_fcaps(ab, n);
2141 	audit_log_end(ab);
2142 }
2143 
2144 int audit_log_task_context(struct audit_buffer *ab)
2145 {
2146 	char *ctx = NULL;
2147 	unsigned len;
2148 	int error;
2149 	u32 sid;
2150 
2151 	security_task_getsecid(current, &sid);
2152 	if (!sid)
2153 		return 0;
2154 
2155 	error = security_secid_to_secctx(sid, &ctx, &len);
2156 	if (error) {
2157 		if (error != -EINVAL)
2158 			goto error_path;
2159 		return 0;
2160 	}
2161 
2162 	audit_log_format(ab, " subj=%s", ctx);
2163 	security_release_secctx(ctx, len);
2164 	return 0;
2165 
2166 error_path:
2167 	audit_panic("error in audit_log_task_context");
2168 	return error;
2169 }
2170 EXPORT_SYMBOL(audit_log_task_context);
2171 
2172 void audit_log_d_path_exe(struct audit_buffer *ab,
2173 			  struct mm_struct *mm)
2174 {
2175 	struct file *exe_file;
2176 
2177 	if (!mm)
2178 		goto out_null;
2179 
2180 	exe_file = get_mm_exe_file(mm);
2181 	if (!exe_file)
2182 		goto out_null;
2183 
2184 	audit_log_d_path(ab, " exe=", &exe_file->f_path);
2185 	fput(exe_file);
2186 	return;
2187 out_null:
2188 	audit_log_format(ab, " exe=(null)");
2189 }
2190 
2191 struct tty_struct *audit_get_tty(struct task_struct *tsk)
2192 {
2193 	struct tty_struct *tty = NULL;
2194 	unsigned long flags;
2195 
2196 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
2197 	if (tsk->signal)
2198 		tty = tty_kref_get(tsk->signal->tty);
2199 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
2200 	return tty;
2201 }
2202 
2203 void audit_put_tty(struct tty_struct *tty)
2204 {
2205 	tty_kref_put(tty);
2206 }
2207 
2208 void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
2209 {
2210 	const struct cred *cred;
2211 	char comm[sizeof(tsk->comm)];
2212 	struct tty_struct *tty;
2213 
2214 	if (!ab)
2215 		return;
2216 
2217 	/* tsk == current */
2218 	cred = current_cred();
2219 	tty = audit_get_tty(tsk);
2220 	audit_log_format(ab,
2221 			 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
2222 			 " euid=%u suid=%u fsuid=%u"
2223 			 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
2224 			 task_ppid_nr(tsk),
2225 			 task_tgid_nr(tsk),
2226 			 from_kuid(&init_user_ns, audit_get_loginuid(tsk)),
2227 			 from_kuid(&init_user_ns, cred->uid),
2228 			 from_kgid(&init_user_ns, cred->gid),
2229 			 from_kuid(&init_user_ns, cred->euid),
2230 			 from_kuid(&init_user_ns, cred->suid),
2231 			 from_kuid(&init_user_ns, cred->fsuid),
2232 			 from_kgid(&init_user_ns, cred->egid),
2233 			 from_kgid(&init_user_ns, cred->sgid),
2234 			 from_kgid(&init_user_ns, cred->fsgid),
2235 			 tty ? tty_name(tty) : "(none)",
2236 			 audit_get_sessionid(tsk));
2237 	audit_put_tty(tty);
2238 	audit_log_format(ab, " comm=");
2239 	audit_log_untrustedstring(ab, get_task_comm(comm, tsk));
2240 	audit_log_d_path_exe(ab, tsk->mm);
2241 	audit_log_task_context(ab);
2242 }
2243 EXPORT_SYMBOL(audit_log_task_info);
2244 
2245 /**
2246  * audit_log_link_denied - report a link restriction denial
2247  * @operation: specific link operation
2248  * @link: the path that triggered the restriction
2249  */
2250 void audit_log_link_denied(const char *operation, const struct path *link)
2251 {
2252 	struct audit_buffer *ab;
2253 	struct audit_names *name;
2254 
2255 	name = kzalloc(sizeof(*name), GFP_NOFS);
2256 	if (!name)
2257 		return;
2258 
2259 	/* Generate AUDIT_ANOM_LINK with subject, operation, outcome. */
2260 	ab = audit_log_start(current->audit_context, GFP_KERNEL,
2261 			     AUDIT_ANOM_LINK);
2262 	if (!ab)
2263 		goto out;
2264 	audit_log_format(ab, "op=%s", operation);
2265 	audit_log_task_info(ab, current);
2266 	audit_log_format(ab, " res=0");
2267 	audit_log_end(ab);
2268 
2269 	/* Generate AUDIT_PATH record with object. */
2270 	name->type = AUDIT_TYPE_NORMAL;
2271 	audit_copy_inode(name, link->dentry, d_backing_inode(link->dentry));
2272 	audit_log_name(current->audit_context, name, link, 0, NULL);
2273 out:
2274 	kfree(name);
2275 }
2276 
2277 /**
2278  * audit_log_end - end one audit record
2279  * @ab: the audit_buffer
2280  *
2281  * We can not do a netlink send inside an irq context because it blocks (last
2282  * arg, flags, is not set to MSG_DONTWAIT), so the audit buffer is placed on a
2283  * queue and a tasklet is scheduled to remove them from the queue outside the
2284  * irq context.  May be called in any context.
2285  */
2286 void audit_log_end(struct audit_buffer *ab)
2287 {
2288 	struct sk_buff *skb;
2289 	struct nlmsghdr *nlh;
2290 
2291 	if (!ab)
2292 		return;
2293 
2294 	if (audit_rate_check()) {
2295 		skb = ab->skb;
2296 		ab->skb = NULL;
2297 
2298 		/* setup the netlink header, see the comments in
2299 		 * kauditd_send_multicast_skb() for length quirks */
2300 		nlh = nlmsg_hdr(skb);
2301 		nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
2302 
2303 		/* queue the netlink packet and poke the kauditd thread */
2304 		skb_queue_tail(&audit_queue, skb);
2305 		wake_up_interruptible(&kauditd_wait);
2306 	} else
2307 		audit_log_lost("rate limit exceeded");
2308 
2309 	audit_buffer_free(ab);
2310 }
2311 
2312 /**
2313  * audit_log - Log an audit record
2314  * @ctx: audit context
2315  * @gfp_mask: type of allocation
2316  * @type: audit message type
2317  * @fmt: format string to use
2318  * @...: variable parameters matching the format string
2319  *
2320  * This is a convenience function that calls audit_log_start,
2321  * audit_log_vformat, and audit_log_end.  It may be called
2322  * in any context.
2323  */
2324 void audit_log(struct audit_context *ctx, gfp_t gfp_mask, int type,
2325 	       const char *fmt, ...)
2326 {
2327 	struct audit_buffer *ab;
2328 	va_list args;
2329 
2330 	ab = audit_log_start(ctx, gfp_mask, type);
2331 	if (ab) {
2332 		va_start(args, fmt);
2333 		audit_log_vformat(ab, fmt, args);
2334 		va_end(args);
2335 		audit_log_end(ab);
2336 	}
2337 }
2338 
2339 #ifdef CONFIG_SECURITY
2340 /**
2341  * audit_log_secctx - Converts and logs SELinux context
2342  * @ab: audit_buffer
2343  * @secid: security number
2344  *
2345  * This is a helper function that calls security_secid_to_secctx to convert
2346  * secid to secctx and then adds the (converted) SELinux context to the audit
2347  * log by calling audit_log_format, thus also preventing leak of internal secid
2348  * to userspace. If secid cannot be converted audit_panic is called.
2349  */
2350 void audit_log_secctx(struct audit_buffer *ab, u32 secid)
2351 {
2352 	u32 len;
2353 	char *secctx;
2354 
2355 	if (security_secid_to_secctx(secid, &secctx, &len)) {
2356 		audit_panic("Cannot convert secid to context");
2357 	} else {
2358 		audit_log_format(ab, " obj=%s", secctx);
2359 		security_release_secctx(secctx, len);
2360 	}
2361 }
2362 EXPORT_SYMBOL(audit_log_secctx);
2363 #endif
2364 
2365 EXPORT_SYMBOL(audit_log_start);
2366 EXPORT_SYMBOL(audit_log_end);
2367 EXPORT_SYMBOL(audit_log_format);
2368 EXPORT_SYMBOL(audit_log);
2369