xref: /openbmc/linux/kernel/audit.c (revision 4cff79e9)
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 #include <linux/slab.h>
64 
65 #include <linux/audit.h>
66 
67 #include <net/sock.h>
68 #include <net/netlink.h>
69 #include <linux/skbuff.h>
70 #ifdef CONFIG_SECURITY
71 #include <linux/security.h>
72 #endif
73 #include <linux/freezer.h>
74 #include <linux/pid_namespace.h>
75 #include <net/netns/generic.h>
76 
77 #include "audit.h"
78 
79 /* No auditing will take place until audit_initialized == AUDIT_INITIALIZED.
80  * (Initialization happens after skb_init is called.) */
81 #define AUDIT_DISABLED		-1
82 #define AUDIT_UNINITIALIZED	0
83 #define AUDIT_INITIALIZED	1
84 static int	audit_initialized;
85 
86 #define AUDIT_OFF	0
87 #define AUDIT_ON	1
88 #define AUDIT_LOCKED	2
89 u32		audit_enabled = AUDIT_OFF;
90 bool		audit_ever_enabled = !!AUDIT_OFF;
91 
92 EXPORT_SYMBOL_GPL(audit_enabled);
93 
94 /* Default state when kernel boots without any parameters. */
95 static u32	audit_default = AUDIT_OFF;
96 
97 /* If auditing cannot proceed, audit_failure selects what happens. */
98 static u32	audit_failure = AUDIT_FAIL_PRINTK;
99 
100 /* private audit network namespace index */
101 static unsigned int audit_net_id;
102 
103 /**
104  * struct audit_net - audit private network namespace data
105  * @sk: communication socket
106  */
107 struct audit_net {
108 	struct sock *sk;
109 };
110 
111 /**
112  * struct auditd_connection - kernel/auditd connection state
113  * @pid: auditd PID
114  * @portid: netlink portid
115  * @net: the associated network namespace
116  * @rcu: RCU head
117  *
118  * Description:
119  * This struct is RCU protected; you must either hold the RCU lock for reading
120  * or the associated spinlock for writing.
121  */
122 static struct auditd_connection {
123 	struct pid *pid;
124 	u32 portid;
125 	struct net *net;
126 	struct rcu_head rcu;
127 } *auditd_conn = NULL;
128 static DEFINE_SPINLOCK(auditd_conn_lock);
129 
130 /* If audit_rate_limit is non-zero, limit the rate of sending audit records
131  * to that number per second.  This prevents DoS attacks, but results in
132  * audit records being dropped. */
133 static u32	audit_rate_limit;
134 
135 /* Number of outstanding audit_buffers allowed.
136  * When set to zero, this means unlimited. */
137 static u32	audit_backlog_limit = 64;
138 #define AUDIT_BACKLOG_WAIT_TIME (60 * HZ)
139 static u32	audit_backlog_wait_time = AUDIT_BACKLOG_WAIT_TIME;
140 
141 /* The identity of the user shutting down the audit system. */
142 kuid_t		audit_sig_uid = INVALID_UID;
143 pid_t		audit_sig_pid = -1;
144 u32		audit_sig_sid = 0;
145 
146 /* Records can be lost in several ways:
147    0) [suppressed in audit_alloc]
148    1) out of memory in audit_log_start [kmalloc of struct audit_buffer]
149    2) out of memory in audit_log_move [alloc_skb]
150    3) suppressed due to audit_rate_limit
151    4) suppressed due to audit_backlog_limit
152 */
153 static atomic_t	audit_lost = ATOMIC_INIT(0);
154 
155 /* Hash for inode-based rules */
156 struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
157 
158 static struct kmem_cache *audit_buffer_cache;
159 
160 /* queue msgs to send via kauditd_task */
161 static struct sk_buff_head audit_queue;
162 /* queue msgs due to temporary unicast send problems */
163 static struct sk_buff_head audit_retry_queue;
164 /* queue msgs waiting for new auditd connection */
165 static struct sk_buff_head audit_hold_queue;
166 
167 /* queue servicing thread */
168 static struct task_struct *kauditd_task;
169 static DECLARE_WAIT_QUEUE_HEAD(kauditd_wait);
170 
171 /* waitqueue for callers who are blocked on the audit backlog */
172 static DECLARE_WAIT_QUEUE_HEAD(audit_backlog_wait);
173 
174 static struct audit_features af = {.vers = AUDIT_FEATURE_VERSION,
175 				   .mask = -1,
176 				   .features = 0,
177 				   .lock = 0,};
178 
179 static char *audit_feature_names[2] = {
180 	"only_unset_loginuid",
181 	"loginuid_immutable",
182 };
183 
184 /**
185  * struct audit_ctl_mutex - serialize requests from userspace
186  * @lock: the mutex used for locking
187  * @owner: the task which owns the lock
188  *
189  * Description:
190  * This is the lock struct used to ensure we only process userspace requests
191  * in an orderly fashion.  We can't simply use a mutex/lock here because we
192  * need to track lock ownership so we don't end up blocking the lock owner in
193  * audit_log_start() or similar.
194  */
195 static struct audit_ctl_mutex {
196 	struct mutex lock;
197 	void *owner;
198 } audit_cmd_mutex;
199 
200 /* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
201  * audit records.  Since printk uses a 1024 byte buffer, this buffer
202  * should be at least that large. */
203 #define AUDIT_BUFSIZ 1024
204 
205 /* The audit_buffer is used when formatting an audit record.  The caller
206  * locks briefly to get the record off the freelist or to allocate the
207  * buffer, and locks briefly to send the buffer to the netlink layer or
208  * to place it on a transmit queue.  Multiple audit_buffers can be in
209  * use simultaneously. */
210 struct audit_buffer {
211 	struct sk_buff       *skb;	/* formatted skb ready to send */
212 	struct audit_context *ctx;	/* NULL or associated context */
213 	gfp_t		     gfp_mask;
214 };
215 
216 struct audit_reply {
217 	__u32 portid;
218 	struct net *net;
219 	struct sk_buff *skb;
220 };
221 
222 /**
223  * auditd_test_task - Check to see if a given task is an audit daemon
224  * @task: the task to check
225  *
226  * Description:
227  * Return 1 if the task is a registered audit daemon, 0 otherwise.
228  */
229 int auditd_test_task(struct task_struct *task)
230 {
231 	int rc;
232 	struct auditd_connection *ac;
233 
234 	rcu_read_lock();
235 	ac = rcu_dereference(auditd_conn);
236 	rc = (ac && ac->pid == task_tgid(task) ? 1 : 0);
237 	rcu_read_unlock();
238 
239 	return rc;
240 }
241 
242 /**
243  * audit_ctl_lock - Take the audit control lock
244  */
245 void audit_ctl_lock(void)
246 {
247 	mutex_lock(&audit_cmd_mutex.lock);
248 	audit_cmd_mutex.owner = current;
249 }
250 
251 /**
252  * audit_ctl_unlock - Drop the audit control lock
253  */
254 void audit_ctl_unlock(void)
255 {
256 	audit_cmd_mutex.owner = NULL;
257 	mutex_unlock(&audit_cmd_mutex.lock);
258 }
259 
260 /**
261  * audit_ctl_owner_current - Test to see if the current task owns the lock
262  *
263  * Description:
264  * Return true if the current task owns the audit control lock, false if it
265  * doesn't own the lock.
266  */
267 static bool audit_ctl_owner_current(void)
268 {
269 	return (current == audit_cmd_mutex.owner);
270 }
271 
272 /**
273  * auditd_pid_vnr - Return the auditd PID relative to the namespace
274  *
275  * Description:
276  * Returns the PID in relation to the namespace, 0 on failure.
277  */
278 static pid_t auditd_pid_vnr(void)
279 {
280 	pid_t pid;
281 	const struct auditd_connection *ac;
282 
283 	rcu_read_lock();
284 	ac = rcu_dereference(auditd_conn);
285 	if (!ac || !ac->pid)
286 		pid = 0;
287 	else
288 		pid = pid_vnr(ac->pid);
289 	rcu_read_unlock();
290 
291 	return pid;
292 }
293 
294 /**
295  * audit_get_sk - Return the audit socket for the given network namespace
296  * @net: the destination network namespace
297  *
298  * Description:
299  * Returns the sock pointer if valid, NULL otherwise.  The caller must ensure
300  * that a reference is held for the network namespace while the sock is in use.
301  */
302 static struct sock *audit_get_sk(const struct net *net)
303 {
304 	struct audit_net *aunet;
305 
306 	if (!net)
307 		return NULL;
308 
309 	aunet = net_generic(net, audit_net_id);
310 	return aunet->sk;
311 }
312 
313 void audit_panic(const char *message)
314 {
315 	switch (audit_failure) {
316 	case AUDIT_FAIL_SILENT:
317 		break;
318 	case AUDIT_FAIL_PRINTK:
319 		if (printk_ratelimit())
320 			pr_err("%s\n", message);
321 		break;
322 	case AUDIT_FAIL_PANIC:
323 		panic("audit: %s\n", message);
324 		break;
325 	}
326 }
327 
328 static inline int audit_rate_check(void)
329 {
330 	static unsigned long	last_check = 0;
331 	static int		messages   = 0;
332 	static DEFINE_SPINLOCK(lock);
333 	unsigned long		flags;
334 	unsigned long		now;
335 	unsigned long		elapsed;
336 	int			retval	   = 0;
337 
338 	if (!audit_rate_limit) return 1;
339 
340 	spin_lock_irqsave(&lock, flags);
341 	if (++messages < audit_rate_limit) {
342 		retval = 1;
343 	} else {
344 		now     = jiffies;
345 		elapsed = now - last_check;
346 		if (elapsed > HZ) {
347 			last_check = now;
348 			messages   = 0;
349 			retval     = 1;
350 		}
351 	}
352 	spin_unlock_irqrestore(&lock, flags);
353 
354 	return retval;
355 }
356 
357 /**
358  * audit_log_lost - conditionally log lost audit message event
359  * @message: the message stating reason for lost audit message
360  *
361  * Emit at least 1 message per second, even if audit_rate_check is
362  * throttling.
363  * Always increment the lost messages counter.
364 */
365 void audit_log_lost(const char *message)
366 {
367 	static unsigned long	last_msg = 0;
368 	static DEFINE_SPINLOCK(lock);
369 	unsigned long		flags;
370 	unsigned long		now;
371 	int			print;
372 
373 	atomic_inc(&audit_lost);
374 
375 	print = (audit_failure == AUDIT_FAIL_PANIC || !audit_rate_limit);
376 
377 	if (!print) {
378 		spin_lock_irqsave(&lock, flags);
379 		now = jiffies;
380 		if (now - last_msg > HZ) {
381 			print = 1;
382 			last_msg = now;
383 		}
384 		spin_unlock_irqrestore(&lock, flags);
385 	}
386 
387 	if (print) {
388 		if (printk_ratelimit())
389 			pr_warn("audit_lost=%u audit_rate_limit=%u audit_backlog_limit=%u\n",
390 				atomic_read(&audit_lost),
391 				audit_rate_limit,
392 				audit_backlog_limit);
393 		audit_panic(message);
394 	}
395 }
396 
397 static int audit_log_config_change(char *function_name, u32 new, u32 old,
398 				   int allow_changes)
399 {
400 	struct audit_buffer *ab;
401 	int rc = 0;
402 
403 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE);
404 	if (unlikely(!ab))
405 		return rc;
406 	audit_log_format(ab, "%s=%u old=%u", function_name, new, old);
407 	audit_log_session_info(ab);
408 	rc = audit_log_task_context(ab);
409 	if (rc)
410 		allow_changes = 0; /* Something weird, deny request */
411 	audit_log_format(ab, " res=%d", allow_changes);
412 	audit_log_end(ab);
413 	return rc;
414 }
415 
416 static int audit_do_config_change(char *function_name, u32 *to_change, u32 new)
417 {
418 	int allow_changes, rc = 0;
419 	u32 old = *to_change;
420 
421 	/* check if we are locked */
422 	if (audit_enabled == AUDIT_LOCKED)
423 		allow_changes = 0;
424 	else
425 		allow_changes = 1;
426 
427 	if (audit_enabled != AUDIT_OFF) {
428 		rc = audit_log_config_change(function_name, new, old, allow_changes);
429 		if (rc)
430 			allow_changes = 0;
431 	}
432 
433 	/* If we are allowed, make the change */
434 	if (allow_changes == 1)
435 		*to_change = new;
436 	/* Not allowed, update reason */
437 	else if (rc == 0)
438 		rc = -EPERM;
439 	return rc;
440 }
441 
442 static int audit_set_rate_limit(u32 limit)
443 {
444 	return audit_do_config_change("audit_rate_limit", &audit_rate_limit, limit);
445 }
446 
447 static int audit_set_backlog_limit(u32 limit)
448 {
449 	return audit_do_config_change("audit_backlog_limit", &audit_backlog_limit, limit);
450 }
451 
452 static int audit_set_backlog_wait_time(u32 timeout)
453 {
454 	return audit_do_config_change("audit_backlog_wait_time",
455 				      &audit_backlog_wait_time, timeout);
456 }
457 
458 static int audit_set_enabled(u32 state)
459 {
460 	int rc;
461 	if (state > AUDIT_LOCKED)
462 		return -EINVAL;
463 
464 	rc =  audit_do_config_change("audit_enabled", &audit_enabled, state);
465 	if (!rc)
466 		audit_ever_enabled |= !!state;
467 
468 	return rc;
469 }
470 
471 static int audit_set_failure(u32 state)
472 {
473 	if (state != AUDIT_FAIL_SILENT
474 	    && state != AUDIT_FAIL_PRINTK
475 	    && state != AUDIT_FAIL_PANIC)
476 		return -EINVAL;
477 
478 	return audit_do_config_change("audit_failure", &audit_failure, state);
479 }
480 
481 /**
482  * auditd_conn_free - RCU helper to release an auditd connection struct
483  * @rcu: RCU head
484  *
485  * Description:
486  * Drop any references inside the auditd connection tracking struct and free
487  * the memory.
488  */
489 static void auditd_conn_free(struct rcu_head *rcu)
490 {
491 	struct auditd_connection *ac;
492 
493 	ac = container_of(rcu, struct auditd_connection, rcu);
494 	put_pid(ac->pid);
495 	put_net(ac->net);
496 	kfree(ac);
497 }
498 
499 /**
500  * auditd_set - Set/Reset the auditd connection state
501  * @pid: auditd PID
502  * @portid: auditd netlink portid
503  * @net: auditd network namespace pointer
504  *
505  * Description:
506  * This function will obtain and drop network namespace references as
507  * necessary.  Returns zero on success, negative values on failure.
508  */
509 static int auditd_set(struct pid *pid, u32 portid, struct net *net)
510 {
511 	unsigned long flags;
512 	struct auditd_connection *ac_old, *ac_new;
513 
514 	if (!pid || !net)
515 		return -EINVAL;
516 
517 	ac_new = kzalloc(sizeof(*ac_new), GFP_KERNEL);
518 	if (!ac_new)
519 		return -ENOMEM;
520 	ac_new->pid = get_pid(pid);
521 	ac_new->portid = portid;
522 	ac_new->net = get_net(net);
523 
524 	spin_lock_irqsave(&auditd_conn_lock, flags);
525 	ac_old = rcu_dereference_protected(auditd_conn,
526 					   lockdep_is_held(&auditd_conn_lock));
527 	rcu_assign_pointer(auditd_conn, ac_new);
528 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
529 
530 	if (ac_old)
531 		call_rcu(&ac_old->rcu, auditd_conn_free);
532 
533 	return 0;
534 }
535 
536 /**
537  * kauditd_print_skb - Print the audit record to the ring buffer
538  * @skb: audit record
539  *
540  * Whatever the reason, this packet may not make it to the auditd connection
541  * so write it via printk so the information isn't completely lost.
542  */
543 static void kauditd_printk_skb(struct sk_buff *skb)
544 {
545 	struct nlmsghdr *nlh = nlmsg_hdr(skb);
546 	char *data = nlmsg_data(nlh);
547 
548 	if (nlh->nlmsg_type != AUDIT_EOE && printk_ratelimit())
549 		pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
550 }
551 
552 /**
553  * kauditd_rehold_skb - Handle a audit record send failure in the hold queue
554  * @skb: audit record
555  *
556  * Description:
557  * This should only be used by the kauditd_thread when it fails to flush the
558  * hold queue.
559  */
560 static void kauditd_rehold_skb(struct sk_buff *skb)
561 {
562 	/* put the record back in the queue at the same place */
563 	skb_queue_head(&audit_hold_queue, skb);
564 }
565 
566 /**
567  * kauditd_hold_skb - Queue an audit record, waiting for auditd
568  * @skb: audit record
569  *
570  * Description:
571  * Queue the audit record, waiting for an instance of auditd.  When this
572  * function is called we haven't given up yet on sending the record, but things
573  * are not looking good.  The first thing we want to do is try to write the
574  * record via printk and then see if we want to try and hold on to the record
575  * and queue it, if we have room.  If we want to hold on to the record, but we
576  * don't have room, record a record lost message.
577  */
578 static void kauditd_hold_skb(struct sk_buff *skb)
579 {
580 	/* at this point it is uncertain if we will ever send this to auditd so
581 	 * try to send the message via printk before we go any further */
582 	kauditd_printk_skb(skb);
583 
584 	/* can we just silently drop the message? */
585 	if (!audit_default) {
586 		kfree_skb(skb);
587 		return;
588 	}
589 
590 	/* if we have room, queue the message */
591 	if (!audit_backlog_limit ||
592 	    skb_queue_len(&audit_hold_queue) < audit_backlog_limit) {
593 		skb_queue_tail(&audit_hold_queue, skb);
594 		return;
595 	}
596 
597 	/* we have no other options - drop the message */
598 	audit_log_lost("kauditd hold queue overflow");
599 	kfree_skb(skb);
600 }
601 
602 /**
603  * kauditd_retry_skb - Queue an audit record, attempt to send again to auditd
604  * @skb: audit record
605  *
606  * Description:
607  * Not as serious as kauditd_hold_skb() as we still have a connected auditd,
608  * but for some reason we are having problems sending it audit records so
609  * queue the given record and attempt to resend.
610  */
611 static void kauditd_retry_skb(struct sk_buff *skb)
612 {
613 	/* NOTE: because records should only live in the retry queue for a
614 	 * short period of time, before either being sent or moved to the hold
615 	 * queue, we don't currently enforce a limit on this queue */
616 	skb_queue_tail(&audit_retry_queue, skb);
617 }
618 
619 /**
620  * auditd_reset - Disconnect the auditd connection
621  * @ac: auditd connection state
622  *
623  * Description:
624  * Break the auditd/kauditd connection and move all the queued records into the
625  * hold queue in case auditd reconnects.  It is important to note that the @ac
626  * pointer should never be dereferenced inside this function as it may be NULL
627  * or invalid, you can only compare the memory address!  If @ac is NULL then
628  * the connection will always be reset.
629  */
630 static void auditd_reset(const struct auditd_connection *ac)
631 {
632 	unsigned long flags;
633 	struct sk_buff *skb;
634 	struct auditd_connection *ac_old;
635 
636 	/* if it isn't already broken, break the connection */
637 	spin_lock_irqsave(&auditd_conn_lock, flags);
638 	ac_old = rcu_dereference_protected(auditd_conn,
639 					   lockdep_is_held(&auditd_conn_lock));
640 	if (ac && ac != ac_old) {
641 		/* someone already registered a new auditd connection */
642 		spin_unlock_irqrestore(&auditd_conn_lock, flags);
643 		return;
644 	}
645 	rcu_assign_pointer(auditd_conn, NULL);
646 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
647 
648 	if (ac_old)
649 		call_rcu(&ac_old->rcu, auditd_conn_free);
650 
651 	/* flush the retry queue to the hold queue, but don't touch the main
652 	 * queue since we need to process that normally for multicast */
653 	while ((skb = skb_dequeue(&audit_retry_queue)))
654 		kauditd_hold_skb(skb);
655 }
656 
657 /**
658  * auditd_send_unicast_skb - Send a record via unicast to auditd
659  * @skb: audit record
660  *
661  * Description:
662  * Send a skb to the audit daemon, returns positive/zero values on success and
663  * negative values on failure; in all cases the skb will be consumed by this
664  * function.  If the send results in -ECONNREFUSED the connection with auditd
665  * will be reset.  This function may sleep so callers should not hold any locks
666  * where this would cause a problem.
667  */
668 static int auditd_send_unicast_skb(struct sk_buff *skb)
669 {
670 	int rc;
671 	u32 portid;
672 	struct net *net;
673 	struct sock *sk;
674 	struct auditd_connection *ac;
675 
676 	/* NOTE: we can't call netlink_unicast while in the RCU section so
677 	 *       take a reference to the network namespace and grab local
678 	 *       copies of the namespace, the sock, and the portid; the
679 	 *       namespace and sock aren't going to go away while we hold a
680 	 *       reference and if the portid does become invalid after the RCU
681 	 *       section netlink_unicast() should safely return an error */
682 
683 	rcu_read_lock();
684 	ac = rcu_dereference(auditd_conn);
685 	if (!ac) {
686 		rcu_read_unlock();
687 		kfree_skb(skb);
688 		rc = -ECONNREFUSED;
689 		goto err;
690 	}
691 	net = get_net(ac->net);
692 	sk = audit_get_sk(net);
693 	portid = ac->portid;
694 	rcu_read_unlock();
695 
696 	rc = netlink_unicast(sk, skb, portid, 0);
697 	put_net(net);
698 	if (rc < 0)
699 		goto err;
700 
701 	return rc;
702 
703 err:
704 	if (ac && rc == -ECONNREFUSED)
705 		auditd_reset(ac);
706 	return rc;
707 }
708 
709 /**
710  * kauditd_send_queue - Helper for kauditd_thread to flush skb queues
711  * @sk: the sending sock
712  * @portid: the netlink destination
713  * @queue: the skb queue to process
714  * @retry_limit: limit on number of netlink unicast failures
715  * @skb_hook: per-skb hook for additional processing
716  * @err_hook: hook called if the skb fails the netlink unicast send
717  *
718  * Description:
719  * Run through the given queue and attempt to send the audit records to auditd,
720  * returns zero on success, negative values on failure.  It is up to the caller
721  * to ensure that the @sk is valid for the duration of this function.
722  *
723  */
724 static int kauditd_send_queue(struct sock *sk, u32 portid,
725 			      struct sk_buff_head *queue,
726 			      unsigned int retry_limit,
727 			      void (*skb_hook)(struct sk_buff *skb),
728 			      void (*err_hook)(struct sk_buff *skb))
729 {
730 	int rc = 0;
731 	struct sk_buff *skb;
732 	static unsigned int failed = 0;
733 
734 	/* NOTE: kauditd_thread takes care of all our locking, we just use
735 	 *       the netlink info passed to us (e.g. sk and portid) */
736 
737 	while ((skb = skb_dequeue(queue))) {
738 		/* call the skb_hook for each skb we touch */
739 		if (skb_hook)
740 			(*skb_hook)(skb);
741 
742 		/* can we send to anyone via unicast? */
743 		if (!sk) {
744 			if (err_hook)
745 				(*err_hook)(skb);
746 			continue;
747 		}
748 
749 		/* grab an extra skb reference in case of error */
750 		skb_get(skb);
751 		rc = netlink_unicast(sk, skb, portid, 0);
752 		if (rc < 0) {
753 			/* fatal failure for our queue flush attempt? */
754 			if (++failed >= retry_limit ||
755 			    rc == -ECONNREFUSED || rc == -EPERM) {
756 				/* yes - error processing for the queue */
757 				sk = NULL;
758 				if (err_hook)
759 					(*err_hook)(skb);
760 				if (!skb_hook)
761 					goto out;
762 				/* keep processing with the skb_hook */
763 				continue;
764 			} else
765 				/* no - requeue to preserve ordering */
766 				skb_queue_head(queue, skb);
767 		} else {
768 			/* it worked - drop the extra reference and continue */
769 			consume_skb(skb);
770 			failed = 0;
771 		}
772 	}
773 
774 out:
775 	return (rc >= 0 ? 0 : rc);
776 }
777 
778 /*
779  * kauditd_send_multicast_skb - Send a record to any multicast listeners
780  * @skb: audit record
781  *
782  * Description:
783  * Write a multicast message to anyone listening in the initial network
784  * namespace.  This function doesn't consume an skb as might be expected since
785  * it has to copy it anyways.
786  */
787 static void kauditd_send_multicast_skb(struct sk_buff *skb)
788 {
789 	struct sk_buff *copy;
790 	struct sock *sock = audit_get_sk(&init_net);
791 	struct nlmsghdr *nlh;
792 
793 	/* NOTE: we are not taking an additional reference for init_net since
794 	 *       we don't have to worry about it going away */
795 
796 	if (!netlink_has_listeners(sock, AUDIT_NLGRP_READLOG))
797 		return;
798 
799 	/*
800 	 * The seemingly wasteful skb_copy() rather than bumping the refcount
801 	 * using skb_get() is necessary because non-standard mods are made to
802 	 * the skb by the original kaudit unicast socket send routine.  The
803 	 * existing auditd daemon assumes this breakage.  Fixing this would
804 	 * require co-ordinating a change in the established protocol between
805 	 * the kaudit kernel subsystem and the auditd userspace code.  There is
806 	 * no reason for new multicast clients to continue with this
807 	 * non-compliance.
808 	 */
809 	copy = skb_copy(skb, GFP_KERNEL);
810 	if (!copy)
811 		return;
812 	nlh = nlmsg_hdr(copy);
813 	nlh->nlmsg_len = skb->len;
814 
815 	nlmsg_multicast(sock, copy, 0, AUDIT_NLGRP_READLOG, GFP_KERNEL);
816 }
817 
818 /**
819  * kauditd_thread - Worker thread to send audit records to userspace
820  * @dummy: unused
821  */
822 static int kauditd_thread(void *dummy)
823 {
824 	int rc;
825 	u32 portid = 0;
826 	struct net *net = NULL;
827 	struct sock *sk = NULL;
828 	struct auditd_connection *ac;
829 
830 #define UNICAST_RETRIES 5
831 
832 	set_freezable();
833 	while (!kthread_should_stop()) {
834 		/* NOTE: see the lock comments in auditd_send_unicast_skb() */
835 		rcu_read_lock();
836 		ac = rcu_dereference(auditd_conn);
837 		if (!ac) {
838 			rcu_read_unlock();
839 			goto main_queue;
840 		}
841 		net = get_net(ac->net);
842 		sk = audit_get_sk(net);
843 		portid = ac->portid;
844 		rcu_read_unlock();
845 
846 		/* attempt to flush the hold queue */
847 		rc = kauditd_send_queue(sk, portid,
848 					&audit_hold_queue, UNICAST_RETRIES,
849 					NULL, kauditd_rehold_skb);
850 		if (ac && rc < 0) {
851 			sk = NULL;
852 			auditd_reset(ac);
853 			goto main_queue;
854 		}
855 
856 		/* attempt to flush the retry queue */
857 		rc = kauditd_send_queue(sk, portid,
858 					&audit_retry_queue, UNICAST_RETRIES,
859 					NULL, kauditd_hold_skb);
860 		if (ac && rc < 0) {
861 			sk = NULL;
862 			auditd_reset(ac);
863 			goto main_queue;
864 		}
865 
866 main_queue:
867 		/* process the main queue - do the multicast send and attempt
868 		 * unicast, dump failed record sends to the retry queue; if
869 		 * sk == NULL due to previous failures we will just do the
870 		 * multicast send and move the record to the hold queue */
871 		rc = kauditd_send_queue(sk, portid, &audit_queue, 1,
872 					kauditd_send_multicast_skb,
873 					(sk ?
874 					 kauditd_retry_skb : kauditd_hold_skb));
875 		if (ac && rc < 0)
876 			auditd_reset(ac);
877 		sk = NULL;
878 
879 		/* drop our netns reference, no auditd sends past this line */
880 		if (net) {
881 			put_net(net);
882 			net = NULL;
883 		}
884 
885 		/* we have processed all the queues so wake everyone */
886 		wake_up(&audit_backlog_wait);
887 
888 		/* NOTE: we want to wake up if there is anything on the queue,
889 		 *       regardless of if an auditd is connected, as we need to
890 		 *       do the multicast send and rotate records from the
891 		 *       main queue to the retry/hold queues */
892 		wait_event_freezable(kauditd_wait,
893 				     (skb_queue_len(&audit_queue) ? 1 : 0));
894 	}
895 
896 	return 0;
897 }
898 
899 int audit_send_list(void *_dest)
900 {
901 	struct audit_netlink_list *dest = _dest;
902 	struct sk_buff *skb;
903 	struct sock *sk = audit_get_sk(dest->net);
904 
905 	/* wait for parent to finish and send an ACK */
906 	audit_ctl_lock();
907 	audit_ctl_unlock();
908 
909 	while ((skb = __skb_dequeue(&dest->q)) != NULL)
910 		netlink_unicast(sk, skb, dest->portid, 0);
911 
912 	put_net(dest->net);
913 	kfree(dest);
914 
915 	return 0;
916 }
917 
918 struct sk_buff *audit_make_reply(int seq, int type, int done,
919 				 int multi, const void *payload, int size)
920 {
921 	struct sk_buff	*skb;
922 	struct nlmsghdr	*nlh;
923 	void		*data;
924 	int		flags = multi ? NLM_F_MULTI : 0;
925 	int		t     = done  ? NLMSG_DONE  : type;
926 
927 	skb = nlmsg_new(size, GFP_KERNEL);
928 	if (!skb)
929 		return NULL;
930 
931 	nlh	= nlmsg_put(skb, 0, seq, t, size, flags);
932 	if (!nlh)
933 		goto out_kfree_skb;
934 	data = nlmsg_data(nlh);
935 	memcpy(data, payload, size);
936 	return skb;
937 
938 out_kfree_skb:
939 	kfree_skb(skb);
940 	return NULL;
941 }
942 
943 static int audit_send_reply_thread(void *arg)
944 {
945 	struct audit_reply *reply = (struct audit_reply *)arg;
946 	struct sock *sk = audit_get_sk(reply->net);
947 
948 	audit_ctl_lock();
949 	audit_ctl_unlock();
950 
951 	/* Ignore failure. It'll only happen if the sender goes away,
952 	   because our timeout is set to infinite. */
953 	netlink_unicast(sk, reply->skb, reply->portid, 0);
954 	put_net(reply->net);
955 	kfree(reply);
956 	return 0;
957 }
958 
959 /**
960  * audit_send_reply - send an audit reply message via netlink
961  * @request_skb: skb of request we are replying to (used to target the reply)
962  * @seq: sequence number
963  * @type: audit message type
964  * @done: done (last) flag
965  * @multi: multi-part message flag
966  * @payload: payload data
967  * @size: payload size
968  *
969  * Allocates an skb, builds the netlink message, and sends it to the port id.
970  * No failure notifications.
971  */
972 static void audit_send_reply(struct sk_buff *request_skb, int seq, int type, int done,
973 			     int multi, const void *payload, int size)
974 {
975 	struct net *net = sock_net(NETLINK_CB(request_skb).sk);
976 	struct sk_buff *skb;
977 	struct task_struct *tsk;
978 	struct audit_reply *reply = kmalloc(sizeof(struct audit_reply),
979 					    GFP_KERNEL);
980 
981 	if (!reply)
982 		return;
983 
984 	skb = audit_make_reply(seq, type, done, multi, payload, size);
985 	if (!skb)
986 		goto out;
987 
988 	reply->net = get_net(net);
989 	reply->portid = NETLINK_CB(request_skb).portid;
990 	reply->skb = skb;
991 
992 	tsk = kthread_run(audit_send_reply_thread, reply, "audit_send_reply");
993 	if (!IS_ERR(tsk))
994 		return;
995 	kfree_skb(skb);
996 out:
997 	kfree(reply);
998 }
999 
1000 /*
1001  * Check for appropriate CAP_AUDIT_ capabilities on incoming audit
1002  * control messages.
1003  */
1004 static int audit_netlink_ok(struct sk_buff *skb, u16 msg_type)
1005 {
1006 	int err = 0;
1007 
1008 	/* Only support initial user namespace for now. */
1009 	/*
1010 	 * We return ECONNREFUSED because it tricks userspace into thinking
1011 	 * that audit was not configured into the kernel.  Lots of users
1012 	 * configure their PAM stack (because that's what the distro does)
1013 	 * to reject login if unable to send messages to audit.  If we return
1014 	 * ECONNREFUSED the PAM stack thinks the kernel does not have audit
1015 	 * configured in and will let login proceed.  If we return EPERM
1016 	 * userspace will reject all logins.  This should be removed when we
1017 	 * support non init namespaces!!
1018 	 */
1019 	if (current_user_ns() != &init_user_ns)
1020 		return -ECONNREFUSED;
1021 
1022 	switch (msg_type) {
1023 	case AUDIT_LIST:
1024 	case AUDIT_ADD:
1025 	case AUDIT_DEL:
1026 		return -EOPNOTSUPP;
1027 	case AUDIT_GET:
1028 	case AUDIT_SET:
1029 	case AUDIT_GET_FEATURE:
1030 	case AUDIT_SET_FEATURE:
1031 	case AUDIT_LIST_RULES:
1032 	case AUDIT_ADD_RULE:
1033 	case AUDIT_DEL_RULE:
1034 	case AUDIT_SIGNAL_INFO:
1035 	case AUDIT_TTY_GET:
1036 	case AUDIT_TTY_SET:
1037 	case AUDIT_TRIM:
1038 	case AUDIT_MAKE_EQUIV:
1039 		/* Only support auditd and auditctl in initial pid namespace
1040 		 * for now. */
1041 		if (task_active_pid_ns(current) != &init_pid_ns)
1042 			return -EPERM;
1043 
1044 		if (!netlink_capable(skb, CAP_AUDIT_CONTROL))
1045 			err = -EPERM;
1046 		break;
1047 	case AUDIT_USER:
1048 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1049 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1050 		if (!netlink_capable(skb, CAP_AUDIT_WRITE))
1051 			err = -EPERM;
1052 		break;
1053 	default:  /* bad msg */
1054 		err = -EINVAL;
1055 	}
1056 
1057 	return err;
1058 }
1059 
1060 static void audit_log_common_recv_msg(struct audit_buffer **ab, u16 msg_type)
1061 {
1062 	uid_t uid = from_kuid(&init_user_ns, current_uid());
1063 	pid_t pid = task_tgid_nr(current);
1064 
1065 	if (!audit_enabled && msg_type != AUDIT_USER_AVC) {
1066 		*ab = NULL;
1067 		return;
1068 	}
1069 
1070 	*ab = audit_log_start(NULL, GFP_KERNEL, msg_type);
1071 	if (unlikely(!*ab))
1072 		return;
1073 	audit_log_format(*ab, "pid=%d uid=%u", pid, uid);
1074 	audit_log_session_info(*ab);
1075 	audit_log_task_context(*ab);
1076 }
1077 
1078 int is_audit_feature_set(int i)
1079 {
1080 	return af.features & AUDIT_FEATURE_TO_MASK(i);
1081 }
1082 
1083 
1084 static int audit_get_feature(struct sk_buff *skb)
1085 {
1086 	u32 seq;
1087 
1088 	seq = nlmsg_hdr(skb)->nlmsg_seq;
1089 
1090 	audit_send_reply(skb, seq, AUDIT_GET_FEATURE, 0, 0, &af, sizeof(af));
1091 
1092 	return 0;
1093 }
1094 
1095 static void audit_log_feature_change(int which, u32 old_feature, u32 new_feature,
1096 				     u32 old_lock, u32 new_lock, int res)
1097 {
1098 	struct audit_buffer *ab;
1099 
1100 	if (audit_enabled == AUDIT_OFF)
1101 		return;
1102 
1103 	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_FEATURE_CHANGE);
1104 	if (!ab)
1105 		return;
1106 	audit_log_task_info(ab, current);
1107 	audit_log_format(ab, " feature=%s old=%u new=%u old_lock=%u new_lock=%u res=%d",
1108 			 audit_feature_names[which], !!old_feature, !!new_feature,
1109 			 !!old_lock, !!new_lock, res);
1110 	audit_log_end(ab);
1111 }
1112 
1113 static int audit_set_feature(struct sk_buff *skb)
1114 {
1115 	struct audit_features *uaf;
1116 	int i;
1117 
1118 	BUILD_BUG_ON(AUDIT_LAST_FEATURE + 1 > ARRAY_SIZE(audit_feature_names));
1119 	uaf = nlmsg_data(nlmsg_hdr(skb));
1120 
1121 	/* if there is ever a version 2 we should handle that here */
1122 
1123 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1124 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1125 		u32 old_feature, new_feature, old_lock, new_lock;
1126 
1127 		/* if we are not changing this feature, move along */
1128 		if (!(feature & uaf->mask))
1129 			continue;
1130 
1131 		old_feature = af.features & feature;
1132 		new_feature = uaf->features & feature;
1133 		new_lock = (uaf->lock | af.lock) & feature;
1134 		old_lock = af.lock & feature;
1135 
1136 		/* are we changing a locked feature? */
1137 		if (old_lock && (new_feature != old_feature)) {
1138 			audit_log_feature_change(i, old_feature, new_feature,
1139 						 old_lock, new_lock, 0);
1140 			return -EPERM;
1141 		}
1142 	}
1143 	/* nothing invalid, do the changes */
1144 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1145 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1146 		u32 old_feature, new_feature, old_lock, new_lock;
1147 
1148 		/* if we are not changing this feature, move along */
1149 		if (!(feature & uaf->mask))
1150 			continue;
1151 
1152 		old_feature = af.features & feature;
1153 		new_feature = uaf->features & feature;
1154 		old_lock = af.lock & feature;
1155 		new_lock = (uaf->lock | af.lock) & feature;
1156 
1157 		if (new_feature != old_feature)
1158 			audit_log_feature_change(i, old_feature, new_feature,
1159 						 old_lock, new_lock, 1);
1160 
1161 		if (new_feature)
1162 			af.features |= feature;
1163 		else
1164 			af.features &= ~feature;
1165 		af.lock |= new_lock;
1166 	}
1167 
1168 	return 0;
1169 }
1170 
1171 static int audit_replace(struct pid *pid)
1172 {
1173 	pid_t pvnr;
1174 	struct sk_buff *skb;
1175 
1176 	pvnr = pid_vnr(pid);
1177 	skb = audit_make_reply(0, AUDIT_REPLACE, 0, 0, &pvnr, sizeof(pvnr));
1178 	if (!skb)
1179 		return -ENOMEM;
1180 	return auditd_send_unicast_skb(skb);
1181 }
1182 
1183 static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1184 {
1185 	u32			seq;
1186 	void			*data;
1187 	int			err;
1188 	struct audit_buffer	*ab;
1189 	u16			msg_type = nlh->nlmsg_type;
1190 	struct audit_sig_info   *sig_data;
1191 	char			*ctx = NULL;
1192 	u32			len;
1193 
1194 	err = audit_netlink_ok(skb, msg_type);
1195 	if (err)
1196 		return err;
1197 
1198 	seq  = nlh->nlmsg_seq;
1199 	data = nlmsg_data(nlh);
1200 
1201 	switch (msg_type) {
1202 	case AUDIT_GET: {
1203 		struct audit_status	s;
1204 		memset(&s, 0, sizeof(s));
1205 		s.enabled		= audit_enabled;
1206 		s.failure		= audit_failure;
1207 		/* NOTE: use pid_vnr() so the PID is relative to the current
1208 		 *       namespace */
1209 		s.pid			= auditd_pid_vnr();
1210 		s.rate_limit		= audit_rate_limit;
1211 		s.backlog_limit		= audit_backlog_limit;
1212 		s.lost			= atomic_read(&audit_lost);
1213 		s.backlog		= skb_queue_len(&audit_queue);
1214 		s.feature_bitmap	= AUDIT_FEATURE_BITMAP_ALL;
1215 		s.backlog_wait_time	= audit_backlog_wait_time;
1216 		audit_send_reply(skb, seq, AUDIT_GET, 0, 0, &s, sizeof(s));
1217 		break;
1218 	}
1219 	case AUDIT_SET: {
1220 		struct audit_status	s;
1221 		memset(&s, 0, sizeof(s));
1222 		/* guard against past and future API changes */
1223 		memcpy(&s, data, min_t(size_t, sizeof(s), nlmsg_len(nlh)));
1224 		if (s.mask & AUDIT_STATUS_ENABLED) {
1225 			err = audit_set_enabled(s.enabled);
1226 			if (err < 0)
1227 				return err;
1228 		}
1229 		if (s.mask & AUDIT_STATUS_FAILURE) {
1230 			err = audit_set_failure(s.failure);
1231 			if (err < 0)
1232 				return err;
1233 		}
1234 		if (s.mask & AUDIT_STATUS_PID) {
1235 			/* NOTE: we are using the vnr PID functions below
1236 			 *       because the s.pid value is relative to the
1237 			 *       namespace of the caller; at present this
1238 			 *       doesn't matter much since you can really only
1239 			 *       run auditd from the initial pid namespace, but
1240 			 *       something to keep in mind if this changes */
1241 			pid_t new_pid = s.pid;
1242 			pid_t auditd_pid;
1243 			struct pid *req_pid = task_tgid(current);
1244 
1245 			/* Sanity check - PID values must match. Setting
1246 			 * pid to 0 is how auditd ends auditing. */
1247 			if (new_pid && (new_pid != pid_vnr(req_pid)))
1248 				return -EINVAL;
1249 
1250 			/* test the auditd connection */
1251 			audit_replace(req_pid);
1252 
1253 			auditd_pid = auditd_pid_vnr();
1254 			if (auditd_pid) {
1255 				/* replacing a healthy auditd is not allowed */
1256 				if (new_pid) {
1257 					audit_log_config_change("audit_pid",
1258 							new_pid, auditd_pid, 0);
1259 					return -EEXIST;
1260 				}
1261 				/* only current auditd can unregister itself */
1262 				if (pid_vnr(req_pid) != auditd_pid) {
1263 					audit_log_config_change("audit_pid",
1264 							new_pid, auditd_pid, 0);
1265 					return -EACCES;
1266 				}
1267 			}
1268 
1269 			if (new_pid) {
1270 				/* register a new auditd connection */
1271 				err = auditd_set(req_pid,
1272 						 NETLINK_CB(skb).portid,
1273 						 sock_net(NETLINK_CB(skb).sk));
1274 				if (audit_enabled != AUDIT_OFF)
1275 					audit_log_config_change("audit_pid",
1276 								new_pid,
1277 								auditd_pid,
1278 								err ? 0 : 1);
1279 				if (err)
1280 					return err;
1281 
1282 				/* try to process any backlog */
1283 				wake_up_interruptible(&kauditd_wait);
1284 			} else {
1285 				if (audit_enabled != AUDIT_OFF)
1286 					audit_log_config_change("audit_pid",
1287 								new_pid,
1288 								auditd_pid, 1);
1289 
1290 				/* unregister the auditd connection */
1291 				auditd_reset(NULL);
1292 			}
1293 		}
1294 		if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
1295 			err = audit_set_rate_limit(s.rate_limit);
1296 			if (err < 0)
1297 				return err;
1298 		}
1299 		if (s.mask & AUDIT_STATUS_BACKLOG_LIMIT) {
1300 			err = audit_set_backlog_limit(s.backlog_limit);
1301 			if (err < 0)
1302 				return err;
1303 		}
1304 		if (s.mask & AUDIT_STATUS_BACKLOG_WAIT_TIME) {
1305 			if (sizeof(s) > (size_t)nlh->nlmsg_len)
1306 				return -EINVAL;
1307 			if (s.backlog_wait_time > 10*AUDIT_BACKLOG_WAIT_TIME)
1308 				return -EINVAL;
1309 			err = audit_set_backlog_wait_time(s.backlog_wait_time);
1310 			if (err < 0)
1311 				return err;
1312 		}
1313 		if (s.mask == AUDIT_STATUS_LOST) {
1314 			u32 lost = atomic_xchg(&audit_lost, 0);
1315 
1316 			audit_log_config_change("lost", 0, lost, 1);
1317 			return lost;
1318 		}
1319 		break;
1320 	}
1321 	case AUDIT_GET_FEATURE:
1322 		err = audit_get_feature(skb);
1323 		if (err)
1324 			return err;
1325 		break;
1326 	case AUDIT_SET_FEATURE:
1327 		err = audit_set_feature(skb);
1328 		if (err)
1329 			return err;
1330 		break;
1331 	case AUDIT_USER:
1332 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1333 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1334 		if (!audit_enabled && msg_type != AUDIT_USER_AVC)
1335 			return 0;
1336 
1337 		err = audit_filter(msg_type, AUDIT_FILTER_USER);
1338 		if (err == 1) { /* match or error */
1339 			err = 0;
1340 			if (msg_type == AUDIT_USER_TTY) {
1341 				err = tty_audit_push();
1342 				if (err)
1343 					break;
1344 			}
1345 			audit_log_common_recv_msg(&ab, msg_type);
1346 			if (msg_type != AUDIT_USER_TTY)
1347 				audit_log_format(ab, " msg='%.*s'",
1348 						 AUDIT_MESSAGE_TEXT_MAX,
1349 						 (char *)data);
1350 			else {
1351 				int size;
1352 
1353 				audit_log_format(ab, " data=");
1354 				size = nlmsg_len(nlh);
1355 				if (size > 0 &&
1356 				    ((unsigned char *)data)[size - 1] == '\0')
1357 					size--;
1358 				audit_log_n_untrustedstring(ab, data, size);
1359 			}
1360 			audit_log_end(ab);
1361 		}
1362 		break;
1363 	case AUDIT_ADD_RULE:
1364 	case AUDIT_DEL_RULE:
1365 		if (nlmsg_len(nlh) < sizeof(struct audit_rule_data))
1366 			return -EINVAL;
1367 		if (audit_enabled == AUDIT_LOCKED) {
1368 			audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1369 			audit_log_format(ab, " audit_enabled=%d res=0", audit_enabled);
1370 			audit_log_end(ab);
1371 			return -EPERM;
1372 		}
1373 		err = audit_rule_change(msg_type, seq, data, nlmsg_len(nlh));
1374 		break;
1375 	case AUDIT_LIST_RULES:
1376 		err = audit_list_rules_send(skb, seq);
1377 		break;
1378 	case AUDIT_TRIM:
1379 		audit_trim_trees();
1380 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1381 		audit_log_format(ab, " op=trim res=1");
1382 		audit_log_end(ab);
1383 		break;
1384 	case AUDIT_MAKE_EQUIV: {
1385 		void *bufp = data;
1386 		u32 sizes[2];
1387 		size_t msglen = nlmsg_len(nlh);
1388 		char *old, *new;
1389 
1390 		err = -EINVAL;
1391 		if (msglen < 2 * sizeof(u32))
1392 			break;
1393 		memcpy(sizes, bufp, 2 * sizeof(u32));
1394 		bufp += 2 * sizeof(u32);
1395 		msglen -= 2 * sizeof(u32);
1396 		old = audit_unpack_string(&bufp, &msglen, sizes[0]);
1397 		if (IS_ERR(old)) {
1398 			err = PTR_ERR(old);
1399 			break;
1400 		}
1401 		new = audit_unpack_string(&bufp, &msglen, sizes[1]);
1402 		if (IS_ERR(new)) {
1403 			err = PTR_ERR(new);
1404 			kfree(old);
1405 			break;
1406 		}
1407 		/* OK, here comes... */
1408 		err = audit_tag_tree(old, new);
1409 
1410 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1411 
1412 		audit_log_format(ab, " op=make_equiv old=");
1413 		audit_log_untrustedstring(ab, old);
1414 		audit_log_format(ab, " new=");
1415 		audit_log_untrustedstring(ab, new);
1416 		audit_log_format(ab, " res=%d", !err);
1417 		audit_log_end(ab);
1418 		kfree(old);
1419 		kfree(new);
1420 		break;
1421 	}
1422 	case AUDIT_SIGNAL_INFO:
1423 		len = 0;
1424 		if (audit_sig_sid) {
1425 			err = security_secid_to_secctx(audit_sig_sid, &ctx, &len);
1426 			if (err)
1427 				return err;
1428 		}
1429 		sig_data = kmalloc(sizeof(*sig_data) + len, GFP_KERNEL);
1430 		if (!sig_data) {
1431 			if (audit_sig_sid)
1432 				security_release_secctx(ctx, len);
1433 			return -ENOMEM;
1434 		}
1435 		sig_data->uid = from_kuid(&init_user_ns, audit_sig_uid);
1436 		sig_data->pid = audit_sig_pid;
1437 		if (audit_sig_sid) {
1438 			memcpy(sig_data->ctx, ctx, len);
1439 			security_release_secctx(ctx, len);
1440 		}
1441 		audit_send_reply(skb, seq, AUDIT_SIGNAL_INFO, 0, 0,
1442 				 sig_data, sizeof(*sig_data) + len);
1443 		kfree(sig_data);
1444 		break;
1445 	case AUDIT_TTY_GET: {
1446 		struct audit_tty_status s;
1447 		unsigned int t;
1448 
1449 		t = READ_ONCE(current->signal->audit_tty);
1450 		s.enabled = t & AUDIT_TTY_ENABLE;
1451 		s.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1452 
1453 		audit_send_reply(skb, seq, AUDIT_TTY_GET, 0, 0, &s, sizeof(s));
1454 		break;
1455 	}
1456 	case AUDIT_TTY_SET: {
1457 		struct audit_tty_status s, old;
1458 		struct audit_buffer	*ab;
1459 		unsigned int t;
1460 
1461 		memset(&s, 0, sizeof(s));
1462 		/* guard against past and future API changes */
1463 		memcpy(&s, data, min_t(size_t, sizeof(s), nlmsg_len(nlh)));
1464 		/* check if new data is valid */
1465 		if ((s.enabled != 0 && s.enabled != 1) ||
1466 		    (s.log_passwd != 0 && s.log_passwd != 1))
1467 			err = -EINVAL;
1468 
1469 		if (err)
1470 			t = READ_ONCE(current->signal->audit_tty);
1471 		else {
1472 			t = s.enabled | (-s.log_passwd & AUDIT_TTY_LOG_PASSWD);
1473 			t = xchg(&current->signal->audit_tty, t);
1474 		}
1475 		old.enabled = t & AUDIT_TTY_ENABLE;
1476 		old.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1477 
1478 		audit_log_common_recv_msg(&ab, AUDIT_CONFIG_CHANGE);
1479 		audit_log_format(ab, " op=tty_set old-enabled=%d new-enabled=%d"
1480 				 " old-log_passwd=%d new-log_passwd=%d res=%d",
1481 				 old.enabled, s.enabled, old.log_passwd,
1482 				 s.log_passwd, !err);
1483 		audit_log_end(ab);
1484 		break;
1485 	}
1486 	default:
1487 		err = -EINVAL;
1488 		break;
1489 	}
1490 
1491 	return err < 0 ? err : 0;
1492 }
1493 
1494 /**
1495  * audit_receive - receive messages from a netlink control socket
1496  * @skb: the message buffer
1497  *
1498  * Parse the provided skb and deal with any messages that may be present,
1499  * malformed skbs are discarded.
1500  */
1501 static void audit_receive(struct sk_buff  *skb)
1502 {
1503 	struct nlmsghdr *nlh;
1504 	/*
1505 	 * len MUST be signed for nlmsg_next to be able to dec it below 0
1506 	 * if the nlmsg_len was not aligned
1507 	 */
1508 	int len;
1509 	int err;
1510 
1511 	nlh = nlmsg_hdr(skb);
1512 	len = skb->len;
1513 
1514 	audit_ctl_lock();
1515 	while (nlmsg_ok(nlh, len)) {
1516 		err = audit_receive_msg(skb, nlh);
1517 		/* if err or if this message says it wants a response */
1518 		if (err || (nlh->nlmsg_flags & NLM_F_ACK))
1519 			netlink_ack(skb, nlh, err, NULL);
1520 
1521 		nlh = nlmsg_next(nlh, &len);
1522 	}
1523 	audit_ctl_unlock();
1524 }
1525 
1526 /* Run custom bind function on netlink socket group connect or bind requests. */
1527 static int audit_bind(struct net *net, int group)
1528 {
1529 	if (!capable(CAP_AUDIT_READ))
1530 		return -EPERM;
1531 
1532 	return 0;
1533 }
1534 
1535 static int __net_init audit_net_init(struct net *net)
1536 {
1537 	struct netlink_kernel_cfg cfg = {
1538 		.input	= audit_receive,
1539 		.bind	= audit_bind,
1540 		.flags	= NL_CFG_F_NONROOT_RECV,
1541 		.groups	= AUDIT_NLGRP_MAX,
1542 	};
1543 
1544 	struct audit_net *aunet = net_generic(net, audit_net_id);
1545 
1546 	aunet->sk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
1547 	if (aunet->sk == NULL) {
1548 		audit_panic("cannot initialize netlink socket in namespace");
1549 		return -ENOMEM;
1550 	}
1551 	aunet->sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1552 
1553 	return 0;
1554 }
1555 
1556 static void __net_exit audit_net_exit(struct net *net)
1557 {
1558 	struct audit_net *aunet = net_generic(net, audit_net_id);
1559 
1560 	/* NOTE: you would think that we would want to check the auditd
1561 	 * connection and potentially reset it here if it lives in this
1562 	 * namespace, but since the auditd connection tracking struct holds a
1563 	 * reference to this namespace (see auditd_set()) we are only ever
1564 	 * going to get here after that connection has been released */
1565 
1566 	netlink_kernel_release(aunet->sk);
1567 }
1568 
1569 static struct pernet_operations audit_net_ops __net_initdata = {
1570 	.init = audit_net_init,
1571 	.exit = audit_net_exit,
1572 	.id = &audit_net_id,
1573 	.size = sizeof(struct audit_net),
1574 };
1575 
1576 /* Initialize audit support at boot time. */
1577 static int __init audit_init(void)
1578 {
1579 	int i;
1580 
1581 	if (audit_initialized == AUDIT_DISABLED)
1582 		return 0;
1583 
1584 	audit_buffer_cache = kmem_cache_create("audit_buffer",
1585 					       sizeof(struct audit_buffer),
1586 					       0, SLAB_PANIC, NULL);
1587 
1588 	skb_queue_head_init(&audit_queue);
1589 	skb_queue_head_init(&audit_retry_queue);
1590 	skb_queue_head_init(&audit_hold_queue);
1591 
1592 	for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
1593 		INIT_LIST_HEAD(&audit_inode_hash[i]);
1594 
1595 	mutex_init(&audit_cmd_mutex.lock);
1596 	audit_cmd_mutex.owner = NULL;
1597 
1598 	pr_info("initializing netlink subsys (%s)\n",
1599 		audit_default ? "enabled" : "disabled");
1600 	register_pernet_subsys(&audit_net_ops);
1601 
1602 	audit_initialized = AUDIT_INITIALIZED;
1603 
1604 	kauditd_task = kthread_run(kauditd_thread, NULL, "kauditd");
1605 	if (IS_ERR(kauditd_task)) {
1606 		int err = PTR_ERR(kauditd_task);
1607 		panic("audit: failed to start the kauditd thread (%d)\n", err);
1608 	}
1609 
1610 	audit_log(NULL, GFP_KERNEL, AUDIT_KERNEL,
1611 		"state=initialized audit_enabled=%u res=1",
1612 		 audit_enabled);
1613 
1614 	return 0;
1615 }
1616 postcore_initcall(audit_init);
1617 
1618 /*
1619  * Process kernel command-line parameter at boot time.
1620  * audit={0|off} or audit={1|on}.
1621  */
1622 static int __init audit_enable(char *str)
1623 {
1624 	if (!strcasecmp(str, "off") || !strcmp(str, "0"))
1625 		audit_default = AUDIT_OFF;
1626 	else if (!strcasecmp(str, "on") || !strcmp(str, "1"))
1627 		audit_default = AUDIT_ON;
1628 	else {
1629 		pr_err("audit: invalid 'audit' parameter value (%s)\n", str);
1630 		audit_default = AUDIT_ON;
1631 	}
1632 
1633 	if (audit_default == AUDIT_OFF)
1634 		audit_initialized = AUDIT_DISABLED;
1635 	if (audit_set_enabled(audit_default))
1636 		pr_err("audit: error setting audit state (%d)\n",
1637 		       audit_default);
1638 
1639 	pr_info("%s\n", audit_default ?
1640 		"enabled (after initialization)" : "disabled (until reboot)");
1641 
1642 	return 1;
1643 }
1644 __setup("audit=", audit_enable);
1645 
1646 /* Process kernel command-line parameter at boot time.
1647  * audit_backlog_limit=<n> */
1648 static int __init audit_backlog_limit_set(char *str)
1649 {
1650 	u32 audit_backlog_limit_arg;
1651 
1652 	pr_info("audit_backlog_limit: ");
1653 	if (kstrtouint(str, 0, &audit_backlog_limit_arg)) {
1654 		pr_cont("using default of %u, unable to parse %s\n",
1655 			audit_backlog_limit, str);
1656 		return 1;
1657 	}
1658 
1659 	audit_backlog_limit = audit_backlog_limit_arg;
1660 	pr_cont("%d\n", audit_backlog_limit);
1661 
1662 	return 1;
1663 }
1664 __setup("audit_backlog_limit=", audit_backlog_limit_set);
1665 
1666 static void audit_buffer_free(struct audit_buffer *ab)
1667 {
1668 	if (!ab)
1669 		return;
1670 
1671 	kfree_skb(ab->skb);
1672 	kmem_cache_free(audit_buffer_cache, ab);
1673 }
1674 
1675 static struct audit_buffer *audit_buffer_alloc(struct audit_context *ctx,
1676 					       gfp_t gfp_mask, int type)
1677 {
1678 	struct audit_buffer *ab;
1679 
1680 	ab = kmem_cache_alloc(audit_buffer_cache, gfp_mask);
1681 	if (!ab)
1682 		return NULL;
1683 
1684 	ab->skb = nlmsg_new(AUDIT_BUFSIZ, gfp_mask);
1685 	if (!ab->skb)
1686 		goto err;
1687 	if (!nlmsg_put(ab->skb, 0, 0, type, 0, 0))
1688 		goto err;
1689 
1690 	ab->ctx = ctx;
1691 	ab->gfp_mask = gfp_mask;
1692 
1693 	return ab;
1694 
1695 err:
1696 	audit_buffer_free(ab);
1697 	return NULL;
1698 }
1699 
1700 /**
1701  * audit_serial - compute a serial number for the audit record
1702  *
1703  * Compute a serial number for the audit record.  Audit records are
1704  * written to user-space as soon as they are generated, so a complete
1705  * audit record may be written in several pieces.  The timestamp of the
1706  * record and this serial number are used by the user-space tools to
1707  * determine which pieces belong to the same audit record.  The
1708  * (timestamp,serial) tuple is unique for each syscall and is live from
1709  * syscall entry to syscall exit.
1710  *
1711  * NOTE: Another possibility is to store the formatted records off the
1712  * audit context (for those records that have a context), and emit them
1713  * all at syscall exit.  However, this could delay the reporting of
1714  * significant errors until syscall exit (or never, if the system
1715  * halts).
1716  */
1717 unsigned int audit_serial(void)
1718 {
1719 	static atomic_t serial = ATOMIC_INIT(0);
1720 
1721 	return atomic_add_return(1, &serial);
1722 }
1723 
1724 static inline void audit_get_stamp(struct audit_context *ctx,
1725 				   struct timespec64 *t, unsigned int *serial)
1726 {
1727 	if (!ctx || !auditsc_get_stamp(ctx, t, serial)) {
1728 		*t = current_kernel_time64();
1729 		*serial = audit_serial();
1730 	}
1731 }
1732 
1733 /**
1734  * audit_log_start - obtain an audit buffer
1735  * @ctx: audit_context (may be NULL)
1736  * @gfp_mask: type of allocation
1737  * @type: audit message type
1738  *
1739  * Returns audit_buffer pointer on success or NULL on error.
1740  *
1741  * Obtain an audit buffer.  This routine does locking to obtain the
1742  * audit buffer, but then no locking is required for calls to
1743  * audit_log_*format.  If the task (ctx) is a task that is currently in a
1744  * syscall, then the syscall is marked as auditable and an audit record
1745  * will be written at syscall exit.  If there is no associated task, then
1746  * task context (ctx) should be NULL.
1747  */
1748 struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
1749 				     int type)
1750 {
1751 	struct audit_buffer *ab;
1752 	struct timespec64 t;
1753 	unsigned int uninitialized_var(serial);
1754 
1755 	if (audit_initialized != AUDIT_INITIALIZED)
1756 		return NULL;
1757 
1758 	if (unlikely(!audit_filter(type, AUDIT_FILTER_TYPE)))
1759 		return NULL;
1760 
1761 	/* NOTE: don't ever fail/sleep on these two conditions:
1762 	 * 1. auditd generated record - since we need auditd to drain the
1763 	 *    queue; also, when we are checking for auditd, compare PIDs using
1764 	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
1765 	 *    using a PID anchored in the caller's namespace
1766 	 * 2. generator holding the audit_cmd_mutex - we don't want to block
1767 	 *    while holding the mutex */
1768 	if (!(auditd_test_task(current) || audit_ctl_owner_current())) {
1769 		long stime = audit_backlog_wait_time;
1770 
1771 		while (audit_backlog_limit &&
1772 		       (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
1773 			/* wake kauditd to try and flush the queue */
1774 			wake_up_interruptible(&kauditd_wait);
1775 
1776 			/* sleep if we are allowed and we haven't exhausted our
1777 			 * backlog wait limit */
1778 			if (gfpflags_allow_blocking(gfp_mask) && (stime > 0)) {
1779 				DECLARE_WAITQUEUE(wait, current);
1780 
1781 				add_wait_queue_exclusive(&audit_backlog_wait,
1782 							 &wait);
1783 				set_current_state(TASK_UNINTERRUPTIBLE);
1784 				stime = schedule_timeout(stime);
1785 				remove_wait_queue(&audit_backlog_wait, &wait);
1786 			} else {
1787 				if (audit_rate_check() && printk_ratelimit())
1788 					pr_warn("audit_backlog=%d > audit_backlog_limit=%d\n",
1789 						skb_queue_len(&audit_queue),
1790 						audit_backlog_limit);
1791 				audit_log_lost("backlog limit exceeded");
1792 				return NULL;
1793 			}
1794 		}
1795 	}
1796 
1797 	ab = audit_buffer_alloc(ctx, gfp_mask, type);
1798 	if (!ab) {
1799 		audit_log_lost("out of memory in audit_log_start");
1800 		return NULL;
1801 	}
1802 
1803 	audit_get_stamp(ab->ctx, &t, &serial);
1804 	audit_log_format(ab, "audit(%llu.%03lu:%u): ",
1805 			 (unsigned long long)t.tv_sec, t.tv_nsec/1000000, serial);
1806 
1807 	return ab;
1808 }
1809 
1810 /**
1811  * audit_expand - expand skb in the audit buffer
1812  * @ab: audit_buffer
1813  * @extra: space to add at tail of the skb
1814  *
1815  * Returns 0 (no space) on failed expansion, or available space if
1816  * successful.
1817  */
1818 static inline int audit_expand(struct audit_buffer *ab, int extra)
1819 {
1820 	struct sk_buff *skb = ab->skb;
1821 	int oldtail = skb_tailroom(skb);
1822 	int ret = pskb_expand_head(skb, 0, extra, ab->gfp_mask);
1823 	int newtail = skb_tailroom(skb);
1824 
1825 	if (ret < 0) {
1826 		audit_log_lost("out of memory in audit_expand");
1827 		return 0;
1828 	}
1829 
1830 	skb->truesize += newtail - oldtail;
1831 	return newtail;
1832 }
1833 
1834 /*
1835  * Format an audit message into the audit buffer.  If there isn't enough
1836  * room in the audit buffer, more room will be allocated and vsnprint
1837  * will be called a second time.  Currently, we assume that a printk
1838  * can't format message larger than 1024 bytes, so we don't either.
1839  */
1840 static void audit_log_vformat(struct audit_buffer *ab, const char *fmt,
1841 			      va_list args)
1842 {
1843 	int len, avail;
1844 	struct sk_buff *skb;
1845 	va_list args2;
1846 
1847 	if (!ab)
1848 		return;
1849 
1850 	BUG_ON(!ab->skb);
1851 	skb = ab->skb;
1852 	avail = skb_tailroom(skb);
1853 	if (avail == 0) {
1854 		avail = audit_expand(ab, AUDIT_BUFSIZ);
1855 		if (!avail)
1856 			goto out;
1857 	}
1858 	va_copy(args2, args);
1859 	len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args);
1860 	if (len >= avail) {
1861 		/* The printk buffer is 1024 bytes long, so if we get
1862 		 * here and AUDIT_BUFSIZ is at least 1024, then we can
1863 		 * log everything that printk could have logged. */
1864 		avail = audit_expand(ab,
1865 			max_t(unsigned, AUDIT_BUFSIZ, 1+len-avail));
1866 		if (!avail)
1867 			goto out_va_end;
1868 		len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args2);
1869 	}
1870 	if (len > 0)
1871 		skb_put(skb, len);
1872 out_va_end:
1873 	va_end(args2);
1874 out:
1875 	return;
1876 }
1877 
1878 /**
1879  * audit_log_format - format a message into the audit buffer.
1880  * @ab: audit_buffer
1881  * @fmt: format string
1882  * @...: optional parameters matching @fmt string
1883  *
1884  * All the work is done in audit_log_vformat.
1885  */
1886 void audit_log_format(struct audit_buffer *ab, const char *fmt, ...)
1887 {
1888 	va_list args;
1889 
1890 	if (!ab)
1891 		return;
1892 	va_start(args, fmt);
1893 	audit_log_vformat(ab, fmt, args);
1894 	va_end(args);
1895 }
1896 
1897 /**
1898  * audit_log_n_hex - convert a buffer to hex and append it to the audit skb
1899  * @ab: the audit_buffer
1900  * @buf: buffer to convert to hex
1901  * @len: length of @buf to be converted
1902  *
1903  * No return value; failure to expand is silently ignored.
1904  *
1905  * This function will take the passed buf and convert it into a string of
1906  * ascii hex digits. The new string is placed onto the skb.
1907  */
1908 void audit_log_n_hex(struct audit_buffer *ab, const unsigned char *buf,
1909 		size_t len)
1910 {
1911 	int i, avail, new_len;
1912 	unsigned char *ptr;
1913 	struct sk_buff *skb;
1914 
1915 	if (!ab)
1916 		return;
1917 
1918 	BUG_ON(!ab->skb);
1919 	skb = ab->skb;
1920 	avail = skb_tailroom(skb);
1921 	new_len = len<<1;
1922 	if (new_len >= avail) {
1923 		/* Round the buffer request up to the next multiple */
1924 		new_len = AUDIT_BUFSIZ*(((new_len-avail)/AUDIT_BUFSIZ) + 1);
1925 		avail = audit_expand(ab, new_len);
1926 		if (!avail)
1927 			return;
1928 	}
1929 
1930 	ptr = skb_tail_pointer(skb);
1931 	for (i = 0; i < len; i++)
1932 		ptr = hex_byte_pack_upper(ptr, buf[i]);
1933 	*ptr = 0;
1934 	skb_put(skb, len << 1); /* new string is twice the old string */
1935 }
1936 
1937 /*
1938  * Format a string of no more than slen characters into the audit buffer,
1939  * enclosed in quote marks.
1940  */
1941 void audit_log_n_string(struct audit_buffer *ab, const char *string,
1942 			size_t slen)
1943 {
1944 	int avail, new_len;
1945 	unsigned char *ptr;
1946 	struct sk_buff *skb;
1947 
1948 	if (!ab)
1949 		return;
1950 
1951 	BUG_ON(!ab->skb);
1952 	skb = ab->skb;
1953 	avail = skb_tailroom(skb);
1954 	new_len = slen + 3;	/* enclosing quotes + null terminator */
1955 	if (new_len > avail) {
1956 		avail = audit_expand(ab, new_len);
1957 		if (!avail)
1958 			return;
1959 	}
1960 	ptr = skb_tail_pointer(skb);
1961 	*ptr++ = '"';
1962 	memcpy(ptr, string, slen);
1963 	ptr += slen;
1964 	*ptr++ = '"';
1965 	*ptr = 0;
1966 	skb_put(skb, slen + 2);	/* don't include null terminator */
1967 }
1968 
1969 /**
1970  * audit_string_contains_control - does a string need to be logged in hex
1971  * @string: string to be checked
1972  * @len: max length of the string to check
1973  */
1974 bool audit_string_contains_control(const char *string, size_t len)
1975 {
1976 	const unsigned char *p;
1977 	for (p = string; p < (const unsigned char *)string + len; p++) {
1978 		if (*p == '"' || *p < 0x21 || *p > 0x7e)
1979 			return true;
1980 	}
1981 	return false;
1982 }
1983 
1984 /**
1985  * audit_log_n_untrustedstring - log a string that may contain random characters
1986  * @ab: audit_buffer
1987  * @len: length of string (not including trailing null)
1988  * @string: string to be logged
1989  *
1990  * This code will escape a string that is passed to it if the string
1991  * contains a control character, unprintable character, double quote mark,
1992  * or a space. Unescaped strings will start and end with a double quote mark.
1993  * Strings that are escaped are printed in hex (2 digits per char).
1994  *
1995  * The caller specifies the number of characters in the string to log, which may
1996  * or may not be the entire string.
1997  */
1998 void audit_log_n_untrustedstring(struct audit_buffer *ab, const char *string,
1999 				 size_t len)
2000 {
2001 	if (audit_string_contains_control(string, len))
2002 		audit_log_n_hex(ab, string, len);
2003 	else
2004 		audit_log_n_string(ab, string, len);
2005 }
2006 
2007 /**
2008  * audit_log_untrustedstring - log a string that may contain random characters
2009  * @ab: audit_buffer
2010  * @string: string to be logged
2011  *
2012  * Same as audit_log_n_untrustedstring(), except that strlen is used to
2013  * determine string length.
2014  */
2015 void audit_log_untrustedstring(struct audit_buffer *ab, const char *string)
2016 {
2017 	audit_log_n_untrustedstring(ab, string, strlen(string));
2018 }
2019 
2020 /* This is a helper-function to print the escaped d_path */
2021 void audit_log_d_path(struct audit_buffer *ab, const char *prefix,
2022 		      const struct path *path)
2023 {
2024 	char *p, *pathname;
2025 
2026 	if (prefix)
2027 		audit_log_format(ab, "%s", prefix);
2028 
2029 	/* We will allow 11 spaces for ' (deleted)' to be appended */
2030 	pathname = kmalloc(PATH_MAX+11, ab->gfp_mask);
2031 	if (!pathname) {
2032 		audit_log_string(ab, "<no_memory>");
2033 		return;
2034 	}
2035 	p = d_path(path, pathname, PATH_MAX+11);
2036 	if (IS_ERR(p)) { /* Should never happen since we send PATH_MAX */
2037 		/* FIXME: can we save some information here? */
2038 		audit_log_string(ab, "<too_long>");
2039 	} else
2040 		audit_log_untrustedstring(ab, p);
2041 	kfree(pathname);
2042 }
2043 
2044 void audit_log_session_info(struct audit_buffer *ab)
2045 {
2046 	unsigned int sessionid = audit_get_sessionid(current);
2047 	uid_t auid = from_kuid(&init_user_ns, audit_get_loginuid(current));
2048 
2049 	audit_log_format(ab, " auid=%u ses=%u", auid, sessionid);
2050 }
2051 
2052 void audit_log_key(struct audit_buffer *ab, char *key)
2053 {
2054 	audit_log_format(ab, " key=");
2055 	if (key)
2056 		audit_log_untrustedstring(ab, key);
2057 	else
2058 		audit_log_format(ab, "(null)");
2059 }
2060 
2061 void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
2062 {
2063 	int i;
2064 
2065 	audit_log_format(ab, " %s=", prefix);
2066 	CAP_FOR_EACH_U32(i) {
2067 		audit_log_format(ab, "%08x",
2068 				 cap->cap[CAP_LAST_U32 - i]);
2069 	}
2070 }
2071 
2072 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
2073 {
2074 	audit_log_cap(ab, "cap_fp", &name->fcap.permitted);
2075 	audit_log_cap(ab, "cap_fi", &name->fcap.inheritable);
2076 	audit_log_format(ab, " cap_fe=%d cap_fver=%x",
2077 			 name->fcap.fE, name->fcap_ver);
2078 }
2079 
2080 static inline int audit_copy_fcaps(struct audit_names *name,
2081 				   const struct dentry *dentry)
2082 {
2083 	struct cpu_vfs_cap_data caps;
2084 	int rc;
2085 
2086 	if (!dentry)
2087 		return 0;
2088 
2089 	rc = get_vfs_caps_from_disk(dentry, &caps);
2090 	if (rc)
2091 		return rc;
2092 
2093 	name->fcap.permitted = caps.permitted;
2094 	name->fcap.inheritable = caps.inheritable;
2095 	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2096 	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >>
2097 				VFS_CAP_REVISION_SHIFT;
2098 
2099 	return 0;
2100 }
2101 
2102 /* Copy inode data into an audit_names. */
2103 void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
2104 		      struct inode *inode)
2105 {
2106 	name->ino   = inode->i_ino;
2107 	name->dev   = inode->i_sb->s_dev;
2108 	name->mode  = inode->i_mode;
2109 	name->uid   = inode->i_uid;
2110 	name->gid   = inode->i_gid;
2111 	name->rdev  = inode->i_rdev;
2112 	security_inode_getsecid(inode, &name->osid);
2113 	audit_copy_fcaps(name, dentry);
2114 }
2115 
2116 /**
2117  * audit_log_name - produce AUDIT_PATH record from struct audit_names
2118  * @context: audit_context for the task
2119  * @n: audit_names structure with reportable details
2120  * @path: optional path to report instead of audit_names->name
2121  * @record_num: record number to report when handling a list of names
2122  * @call_panic: optional pointer to int that will be updated if secid fails
2123  */
2124 void audit_log_name(struct audit_context *context, struct audit_names *n,
2125 		    const struct path *path, int record_num, int *call_panic)
2126 {
2127 	struct audit_buffer *ab;
2128 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
2129 	if (!ab)
2130 		return;
2131 
2132 	audit_log_format(ab, "item=%d", record_num);
2133 
2134 	if (path)
2135 		audit_log_d_path(ab, " name=", path);
2136 	else if (n->name) {
2137 		switch (n->name_len) {
2138 		case AUDIT_NAME_FULL:
2139 			/* log the full path */
2140 			audit_log_format(ab, " name=");
2141 			audit_log_untrustedstring(ab, n->name->name);
2142 			break;
2143 		case 0:
2144 			/* name was specified as a relative path and the
2145 			 * directory component is the cwd */
2146 			audit_log_d_path(ab, " name=", &context->pwd);
2147 			break;
2148 		default:
2149 			/* log the name's directory component */
2150 			audit_log_format(ab, " name=");
2151 			audit_log_n_untrustedstring(ab, n->name->name,
2152 						    n->name_len);
2153 		}
2154 	} else
2155 		audit_log_format(ab, " name=(null)");
2156 
2157 	if (n->ino != AUDIT_INO_UNSET)
2158 		audit_log_format(ab, " inode=%lu"
2159 				 " dev=%02x:%02x mode=%#ho"
2160 				 " ouid=%u ogid=%u rdev=%02x:%02x",
2161 				 n->ino,
2162 				 MAJOR(n->dev),
2163 				 MINOR(n->dev),
2164 				 n->mode,
2165 				 from_kuid(&init_user_ns, n->uid),
2166 				 from_kgid(&init_user_ns, n->gid),
2167 				 MAJOR(n->rdev),
2168 				 MINOR(n->rdev));
2169 	if (n->osid != 0) {
2170 		char *ctx = NULL;
2171 		u32 len;
2172 		if (security_secid_to_secctx(
2173 			n->osid, &ctx, &len)) {
2174 			audit_log_format(ab, " osid=%u", n->osid);
2175 			if (call_panic)
2176 				*call_panic = 2;
2177 		} else {
2178 			audit_log_format(ab, " obj=%s", ctx);
2179 			security_release_secctx(ctx, len);
2180 		}
2181 	}
2182 
2183 	/* log the audit_names record type */
2184 	audit_log_format(ab, " nametype=");
2185 	switch(n->type) {
2186 	case AUDIT_TYPE_NORMAL:
2187 		audit_log_format(ab, "NORMAL");
2188 		break;
2189 	case AUDIT_TYPE_PARENT:
2190 		audit_log_format(ab, "PARENT");
2191 		break;
2192 	case AUDIT_TYPE_CHILD_DELETE:
2193 		audit_log_format(ab, "DELETE");
2194 		break;
2195 	case AUDIT_TYPE_CHILD_CREATE:
2196 		audit_log_format(ab, "CREATE");
2197 		break;
2198 	default:
2199 		audit_log_format(ab, "UNKNOWN");
2200 		break;
2201 	}
2202 
2203 	audit_log_fcaps(ab, n);
2204 	audit_log_end(ab);
2205 }
2206 
2207 int audit_log_task_context(struct audit_buffer *ab)
2208 {
2209 	char *ctx = NULL;
2210 	unsigned len;
2211 	int error;
2212 	u32 sid;
2213 
2214 	security_task_getsecid(current, &sid);
2215 	if (!sid)
2216 		return 0;
2217 
2218 	error = security_secid_to_secctx(sid, &ctx, &len);
2219 	if (error) {
2220 		if (error != -EINVAL)
2221 			goto error_path;
2222 		return 0;
2223 	}
2224 
2225 	audit_log_format(ab, " subj=%s", ctx);
2226 	security_release_secctx(ctx, len);
2227 	return 0;
2228 
2229 error_path:
2230 	audit_panic("error in audit_log_task_context");
2231 	return error;
2232 }
2233 EXPORT_SYMBOL(audit_log_task_context);
2234 
2235 void audit_log_d_path_exe(struct audit_buffer *ab,
2236 			  struct mm_struct *mm)
2237 {
2238 	struct file *exe_file;
2239 
2240 	if (!mm)
2241 		goto out_null;
2242 
2243 	exe_file = get_mm_exe_file(mm);
2244 	if (!exe_file)
2245 		goto out_null;
2246 
2247 	audit_log_d_path(ab, " exe=", &exe_file->f_path);
2248 	fput(exe_file);
2249 	return;
2250 out_null:
2251 	audit_log_format(ab, " exe=(null)");
2252 }
2253 
2254 struct tty_struct *audit_get_tty(struct task_struct *tsk)
2255 {
2256 	struct tty_struct *tty = NULL;
2257 	unsigned long flags;
2258 
2259 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
2260 	if (tsk->signal)
2261 		tty = tty_kref_get(tsk->signal->tty);
2262 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
2263 	return tty;
2264 }
2265 
2266 void audit_put_tty(struct tty_struct *tty)
2267 {
2268 	tty_kref_put(tty);
2269 }
2270 
2271 void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
2272 {
2273 	const struct cred *cred;
2274 	char comm[sizeof(tsk->comm)];
2275 	struct tty_struct *tty;
2276 
2277 	if (!ab)
2278 		return;
2279 
2280 	/* tsk == current */
2281 	cred = current_cred();
2282 	tty = audit_get_tty(tsk);
2283 	audit_log_format(ab,
2284 			 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
2285 			 " euid=%u suid=%u fsuid=%u"
2286 			 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
2287 			 task_ppid_nr(tsk),
2288 			 task_tgid_nr(tsk),
2289 			 from_kuid(&init_user_ns, audit_get_loginuid(tsk)),
2290 			 from_kuid(&init_user_ns, cred->uid),
2291 			 from_kgid(&init_user_ns, cred->gid),
2292 			 from_kuid(&init_user_ns, cred->euid),
2293 			 from_kuid(&init_user_ns, cred->suid),
2294 			 from_kuid(&init_user_ns, cred->fsuid),
2295 			 from_kgid(&init_user_ns, cred->egid),
2296 			 from_kgid(&init_user_ns, cred->sgid),
2297 			 from_kgid(&init_user_ns, cred->fsgid),
2298 			 tty ? tty_name(tty) : "(none)",
2299 			 audit_get_sessionid(tsk));
2300 	audit_put_tty(tty);
2301 	audit_log_format(ab, " comm=");
2302 	audit_log_untrustedstring(ab, get_task_comm(comm, tsk));
2303 	audit_log_d_path_exe(ab, tsk->mm);
2304 	audit_log_task_context(ab);
2305 }
2306 EXPORT_SYMBOL(audit_log_task_info);
2307 
2308 /**
2309  * audit_log_link_denied - report a link restriction denial
2310  * @operation: specific link operation
2311  */
2312 void audit_log_link_denied(const char *operation)
2313 {
2314 	struct audit_buffer *ab;
2315 
2316 	if (!audit_enabled || audit_dummy_context())
2317 		return;
2318 
2319 	/* Generate AUDIT_ANOM_LINK with subject, operation, outcome. */
2320 	ab = audit_log_start(current->audit_context, GFP_KERNEL,
2321 			     AUDIT_ANOM_LINK);
2322 	if (!ab)
2323 		return;
2324 	audit_log_format(ab, "op=%s", operation);
2325 	audit_log_task_info(ab, current);
2326 	audit_log_format(ab, " res=0");
2327 	audit_log_end(ab);
2328 }
2329 
2330 /**
2331  * audit_log_end - end one audit record
2332  * @ab: the audit_buffer
2333  *
2334  * We can not do a netlink send inside an irq context because it blocks (last
2335  * arg, flags, is not set to MSG_DONTWAIT), so the audit buffer is placed on a
2336  * queue and a tasklet is scheduled to remove them from the queue outside the
2337  * irq context.  May be called in any context.
2338  */
2339 void audit_log_end(struct audit_buffer *ab)
2340 {
2341 	struct sk_buff *skb;
2342 	struct nlmsghdr *nlh;
2343 
2344 	if (!ab)
2345 		return;
2346 
2347 	if (audit_rate_check()) {
2348 		skb = ab->skb;
2349 		ab->skb = NULL;
2350 
2351 		/* setup the netlink header, see the comments in
2352 		 * kauditd_send_multicast_skb() for length quirks */
2353 		nlh = nlmsg_hdr(skb);
2354 		nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
2355 
2356 		/* queue the netlink packet and poke the kauditd thread */
2357 		skb_queue_tail(&audit_queue, skb);
2358 		wake_up_interruptible(&kauditd_wait);
2359 	} else
2360 		audit_log_lost("rate limit exceeded");
2361 
2362 	audit_buffer_free(ab);
2363 }
2364 
2365 /**
2366  * audit_log - Log an audit record
2367  * @ctx: audit context
2368  * @gfp_mask: type of allocation
2369  * @type: audit message type
2370  * @fmt: format string to use
2371  * @...: variable parameters matching the format string
2372  *
2373  * This is a convenience function that calls audit_log_start,
2374  * audit_log_vformat, and audit_log_end.  It may be called
2375  * in any context.
2376  */
2377 void audit_log(struct audit_context *ctx, gfp_t gfp_mask, int type,
2378 	       const char *fmt, ...)
2379 {
2380 	struct audit_buffer *ab;
2381 	va_list args;
2382 
2383 	ab = audit_log_start(ctx, gfp_mask, type);
2384 	if (ab) {
2385 		va_start(args, fmt);
2386 		audit_log_vformat(ab, fmt, args);
2387 		va_end(args);
2388 		audit_log_end(ab);
2389 	}
2390 }
2391 
2392 EXPORT_SYMBOL(audit_log_start);
2393 EXPORT_SYMBOL(audit_log_end);
2394 EXPORT_SYMBOL(audit_log_format);
2395 EXPORT_SYMBOL(audit_log);
2396