xref: /openbmc/linux/kernel/audit.c (revision 83b975b5)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* audit.c -- Auditing support
3  * Gateway between the kernel (e.g., selinux) and the user-space audit daemon.
4  * System-call specific features have moved to auditsc.c
5  *
6  * Copyright 2003-2007 Red Hat Inc., Durham, North Carolina.
7  * All Rights Reserved.
8  *
9  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
10  *
11  * Goals: 1) Integrate fully with Security Modules.
12  *	  2) Minimal run-time overhead:
13  *	     a) Minimal when syscall auditing is disabled (audit_enable=0).
14  *	     b) Small when syscall auditing is enabled and no audit record
15  *		is generated (defer as much work as possible to record
16  *		generation time):
17  *		i) context is allocated,
18  *		ii) names from getname are stored without a copy, and
19  *		iii) inode information stored from path_lookup.
20  *	  3) Ability to disable syscall auditing at boot time (audit=0).
21  *	  4) Usable by other parts of the kernel (if audit_log* is called,
22  *	     then a syscall record will be generated automatically for the
23  *	     current syscall).
24  *	  5) Netlink interface to user-space.
25  *	  6) Support low-overhead kernel-based filtering to minimize the
26  *	     information that must be passed to user-space.
27  *
28  * Audit userspace, documentation, tests, and bug/issue trackers:
29  * 	https://github.com/linux-audit
30  */
31 
32 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
33 
34 #include <linux/file.h>
35 #include <linux/init.h>
36 #include <linux/types.h>
37 #include <linux/atomic.h>
38 #include <linux/mm.h>
39 #include <linux/export.h>
40 #include <linux/slab.h>
41 #include <linux/err.h>
42 #include <linux/kthread.h>
43 #include <linux/kernel.h>
44 #include <linux/syscalls.h>
45 #include <linux/spinlock.h>
46 #include <linux/rcupdate.h>
47 #include <linux/mutex.h>
48 #include <linux/gfp.h>
49 #include <linux/pid.h>
50 
51 #include <linux/audit.h>
52 
53 #include <net/sock.h>
54 #include <net/netlink.h>
55 #include <linux/skbuff.h>
56 #ifdef CONFIG_SECURITY
57 #include <linux/security.h>
58 #endif
59 #include <linux/freezer.h>
60 #include <linux/pid_namespace.h>
61 #include <net/netns/generic.h>
62 
63 #include "audit.h"
64 
65 /* No auditing will take place until audit_initialized == AUDIT_INITIALIZED.
66  * (Initialization happens after skb_init is called.) */
67 #define AUDIT_DISABLED		-1
68 #define AUDIT_UNINITIALIZED	0
69 #define AUDIT_INITIALIZED	1
70 static int	audit_initialized = AUDIT_UNINITIALIZED;
71 
72 u32		audit_enabled = AUDIT_OFF;
73 bool		audit_ever_enabled = !!AUDIT_OFF;
74 
75 EXPORT_SYMBOL_GPL(audit_enabled);
76 
77 /* Default state when kernel boots without any parameters. */
78 static u32	audit_default = AUDIT_OFF;
79 
80 /* If auditing cannot proceed, audit_failure selects what happens. */
81 static u32	audit_failure = AUDIT_FAIL_PRINTK;
82 
83 /* private audit network namespace index */
84 static unsigned int audit_net_id;
85 
86 /**
87  * struct audit_net - audit private network namespace data
88  * @sk: communication socket
89  */
90 struct audit_net {
91 	struct sock *sk;
92 };
93 
94 /**
95  * struct auditd_connection - kernel/auditd connection state
96  * @pid: auditd PID
97  * @portid: netlink portid
98  * @net: the associated network namespace
99  * @rcu: RCU head
100  *
101  * Description:
102  * This struct is RCU protected; you must either hold the RCU lock for reading
103  * or the associated spinlock for writing.
104  */
105 struct auditd_connection {
106 	struct pid *pid;
107 	u32 portid;
108 	struct net *net;
109 	struct rcu_head rcu;
110 };
111 static struct auditd_connection __rcu *auditd_conn;
112 static DEFINE_SPINLOCK(auditd_conn_lock);
113 
114 /* If audit_rate_limit is non-zero, limit the rate of sending audit records
115  * to that number per second.  This prevents DoS attacks, but results in
116  * audit records being dropped. */
117 static u32	audit_rate_limit;
118 
119 /* Number of outstanding audit_buffers allowed.
120  * When set to zero, this means unlimited. */
121 static u32	audit_backlog_limit = 64;
122 #define AUDIT_BACKLOG_WAIT_TIME (60 * HZ)
123 static u32	audit_backlog_wait_time = AUDIT_BACKLOG_WAIT_TIME;
124 
125 /* The identity of the user shutting down the audit system. */
126 static kuid_t		audit_sig_uid = INVALID_UID;
127 static pid_t		audit_sig_pid = -1;
128 static u32		audit_sig_sid;
129 
130 /* Records can be lost in several ways:
131    0) [suppressed in audit_alloc]
132    1) out of memory in audit_log_start [kmalloc of struct audit_buffer]
133    2) out of memory in audit_log_move [alloc_skb]
134    3) suppressed due to audit_rate_limit
135    4) suppressed due to audit_backlog_limit
136 */
137 static atomic_t	audit_lost = ATOMIC_INIT(0);
138 
139 /* Monotonically increasing sum of time the kernel has spent
140  * waiting while the backlog limit is exceeded.
141  */
142 static atomic_t audit_backlog_wait_time_actual = ATOMIC_INIT(0);
143 
144 /* Hash for inode-based rules */
145 struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
146 
147 static struct kmem_cache *audit_buffer_cache;
148 
149 /* queue msgs to send via kauditd_task */
150 static struct sk_buff_head audit_queue;
151 /* queue msgs due to temporary unicast send problems */
152 static struct sk_buff_head audit_retry_queue;
153 /* queue msgs waiting for new auditd connection */
154 static struct sk_buff_head audit_hold_queue;
155 
156 /* queue servicing thread */
157 static struct task_struct *kauditd_task;
158 static DECLARE_WAIT_QUEUE_HEAD(kauditd_wait);
159 
160 /* waitqueue for callers who are blocked on the audit backlog */
161 static DECLARE_WAIT_QUEUE_HEAD(audit_backlog_wait);
162 
163 static struct audit_features af = {.vers = AUDIT_FEATURE_VERSION,
164 				   .mask = -1,
165 				   .features = 0,
166 				   .lock = 0,};
167 
168 static char *audit_feature_names[2] = {
169 	"only_unset_loginuid",
170 	"loginuid_immutable",
171 };
172 
173 /**
174  * struct audit_ctl_mutex - serialize requests from userspace
175  * @lock: the mutex used for locking
176  * @owner: the task which owns the lock
177  *
178  * Description:
179  * This is the lock struct used to ensure we only process userspace requests
180  * in an orderly fashion.  We can't simply use a mutex/lock here because we
181  * need to track lock ownership so we don't end up blocking the lock owner in
182  * audit_log_start() or similar.
183  */
184 static struct audit_ctl_mutex {
185 	struct mutex lock;
186 	void *owner;
187 } audit_cmd_mutex;
188 
189 /* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
190  * audit records.  Since printk uses a 1024 byte buffer, this buffer
191  * should be at least that large. */
192 #define AUDIT_BUFSIZ 1024
193 
194 /* The audit_buffer is used when formatting an audit record.  The caller
195  * locks briefly to get the record off the freelist or to allocate the
196  * buffer, and locks briefly to send the buffer to the netlink layer or
197  * to place it on a transmit queue.  Multiple audit_buffers can be in
198  * use simultaneously. */
199 struct audit_buffer {
200 	struct sk_buff       *skb;	/* formatted skb ready to send */
201 	struct audit_context *ctx;	/* NULL or associated context */
202 	gfp_t		     gfp_mask;
203 };
204 
205 struct audit_reply {
206 	__u32 portid;
207 	struct net *net;
208 	struct sk_buff *skb;
209 };
210 
211 /**
212  * auditd_test_task - Check to see if a given task is an audit daemon
213  * @task: the task to check
214  *
215  * Description:
216  * Return 1 if the task is a registered audit daemon, 0 otherwise.
217  */
218 int auditd_test_task(struct task_struct *task)
219 {
220 	int rc;
221 	struct auditd_connection *ac;
222 
223 	rcu_read_lock();
224 	ac = rcu_dereference(auditd_conn);
225 	rc = (ac && ac->pid == task_tgid(task) ? 1 : 0);
226 	rcu_read_unlock();
227 
228 	return rc;
229 }
230 
231 /**
232  * audit_ctl_lock - Take the audit control lock
233  */
234 void audit_ctl_lock(void)
235 {
236 	mutex_lock(&audit_cmd_mutex.lock);
237 	audit_cmd_mutex.owner = current;
238 }
239 
240 /**
241  * audit_ctl_unlock - Drop the audit control lock
242  */
243 void audit_ctl_unlock(void)
244 {
245 	audit_cmd_mutex.owner = NULL;
246 	mutex_unlock(&audit_cmd_mutex.lock);
247 }
248 
249 /**
250  * audit_ctl_owner_current - Test to see if the current task owns the lock
251  *
252  * Description:
253  * Return true if the current task owns the audit control lock, false if it
254  * doesn't own the lock.
255  */
256 static bool audit_ctl_owner_current(void)
257 {
258 	return (current == audit_cmd_mutex.owner);
259 }
260 
261 /**
262  * auditd_pid_vnr - Return the auditd PID relative to the namespace
263  *
264  * Description:
265  * Returns the PID in relation to the namespace, 0 on failure.
266  */
267 static pid_t auditd_pid_vnr(void)
268 {
269 	pid_t pid;
270 	const struct auditd_connection *ac;
271 
272 	rcu_read_lock();
273 	ac = rcu_dereference(auditd_conn);
274 	if (!ac || !ac->pid)
275 		pid = 0;
276 	else
277 		pid = pid_vnr(ac->pid);
278 	rcu_read_unlock();
279 
280 	return pid;
281 }
282 
283 /**
284  * audit_get_sk - Return the audit socket for the given network namespace
285  * @net: the destination network namespace
286  *
287  * Description:
288  * Returns the sock pointer if valid, NULL otherwise.  The caller must ensure
289  * that a reference is held for the network namespace while the sock is in use.
290  */
291 static struct sock *audit_get_sk(const struct net *net)
292 {
293 	struct audit_net *aunet;
294 
295 	if (!net)
296 		return NULL;
297 
298 	aunet = net_generic(net, audit_net_id);
299 	return aunet->sk;
300 }
301 
302 void audit_panic(const char *message)
303 {
304 	switch (audit_failure) {
305 	case AUDIT_FAIL_SILENT:
306 		break;
307 	case AUDIT_FAIL_PRINTK:
308 		if (printk_ratelimit())
309 			pr_err("%s\n", message);
310 		break;
311 	case AUDIT_FAIL_PANIC:
312 		panic("audit: %s\n", message);
313 		break;
314 	}
315 }
316 
317 static inline int audit_rate_check(void)
318 {
319 	static unsigned long	last_check = 0;
320 	static int		messages   = 0;
321 	static DEFINE_SPINLOCK(lock);
322 	unsigned long		flags;
323 	unsigned long		now;
324 	int			retval	   = 0;
325 
326 	if (!audit_rate_limit) return 1;
327 
328 	spin_lock_irqsave(&lock, flags);
329 	if (++messages < audit_rate_limit) {
330 		retval = 1;
331 	} else {
332 		now = jiffies;
333 		if (time_after(now, last_check + HZ)) {
334 			last_check = now;
335 			messages   = 0;
336 			retval     = 1;
337 		}
338 	}
339 	spin_unlock_irqrestore(&lock, flags);
340 
341 	return retval;
342 }
343 
344 /**
345  * audit_log_lost - conditionally log lost audit message event
346  * @message: the message stating reason for lost audit message
347  *
348  * Emit at least 1 message per second, even if audit_rate_check is
349  * throttling.
350  * Always increment the lost messages counter.
351 */
352 void audit_log_lost(const char *message)
353 {
354 	static unsigned long	last_msg = 0;
355 	static DEFINE_SPINLOCK(lock);
356 	unsigned long		flags;
357 	unsigned long		now;
358 	int			print;
359 
360 	atomic_inc(&audit_lost);
361 
362 	print = (audit_failure == AUDIT_FAIL_PANIC || !audit_rate_limit);
363 
364 	if (!print) {
365 		spin_lock_irqsave(&lock, flags);
366 		now = jiffies;
367 		if (time_after(now, last_msg + HZ)) {
368 			print = 1;
369 			last_msg = now;
370 		}
371 		spin_unlock_irqrestore(&lock, flags);
372 	}
373 
374 	if (print) {
375 		if (printk_ratelimit())
376 			pr_warn("audit_lost=%u audit_rate_limit=%u audit_backlog_limit=%u\n",
377 				atomic_read(&audit_lost),
378 				audit_rate_limit,
379 				audit_backlog_limit);
380 		audit_panic(message);
381 	}
382 }
383 
384 static int audit_log_config_change(char *function_name, u32 new, u32 old,
385 				   int allow_changes)
386 {
387 	struct audit_buffer *ab;
388 	int rc = 0;
389 
390 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_CONFIG_CHANGE);
391 	if (unlikely(!ab))
392 		return rc;
393 	audit_log_format(ab, "op=set %s=%u old=%u ", function_name, new, old);
394 	audit_log_session_info(ab);
395 	rc = audit_log_task_context(ab);
396 	if (rc)
397 		allow_changes = 0; /* Something weird, deny request */
398 	audit_log_format(ab, " res=%d", allow_changes);
399 	audit_log_end(ab);
400 	return rc;
401 }
402 
403 static int audit_do_config_change(char *function_name, u32 *to_change, u32 new)
404 {
405 	int allow_changes, rc = 0;
406 	u32 old = *to_change;
407 
408 	/* check if we are locked */
409 	if (audit_enabled == AUDIT_LOCKED)
410 		allow_changes = 0;
411 	else
412 		allow_changes = 1;
413 
414 	if (audit_enabled != AUDIT_OFF) {
415 		rc = audit_log_config_change(function_name, new, old, allow_changes);
416 		if (rc)
417 			allow_changes = 0;
418 	}
419 
420 	/* If we are allowed, make the change */
421 	if (allow_changes == 1)
422 		*to_change = new;
423 	/* Not allowed, update reason */
424 	else if (rc == 0)
425 		rc = -EPERM;
426 	return rc;
427 }
428 
429 static int audit_set_rate_limit(u32 limit)
430 {
431 	return audit_do_config_change("audit_rate_limit", &audit_rate_limit, limit);
432 }
433 
434 static int audit_set_backlog_limit(u32 limit)
435 {
436 	return audit_do_config_change("audit_backlog_limit", &audit_backlog_limit, limit);
437 }
438 
439 static int audit_set_backlog_wait_time(u32 timeout)
440 {
441 	return audit_do_config_change("audit_backlog_wait_time",
442 				      &audit_backlog_wait_time, timeout);
443 }
444 
445 static int audit_set_enabled(u32 state)
446 {
447 	int rc;
448 	if (state > AUDIT_LOCKED)
449 		return -EINVAL;
450 
451 	rc =  audit_do_config_change("audit_enabled", &audit_enabled, state);
452 	if (!rc)
453 		audit_ever_enabled |= !!state;
454 
455 	return rc;
456 }
457 
458 static int audit_set_failure(u32 state)
459 {
460 	if (state != AUDIT_FAIL_SILENT
461 	    && state != AUDIT_FAIL_PRINTK
462 	    && state != AUDIT_FAIL_PANIC)
463 		return -EINVAL;
464 
465 	return audit_do_config_change("audit_failure", &audit_failure, state);
466 }
467 
468 /**
469  * auditd_conn_free - RCU helper to release an auditd connection struct
470  * @rcu: RCU head
471  *
472  * Description:
473  * Drop any references inside the auditd connection tracking struct and free
474  * the memory.
475  */
476 static void auditd_conn_free(struct rcu_head *rcu)
477 {
478 	struct auditd_connection *ac;
479 
480 	ac = container_of(rcu, struct auditd_connection, rcu);
481 	put_pid(ac->pid);
482 	put_net(ac->net);
483 	kfree(ac);
484 }
485 
486 /**
487  * auditd_set - Set/Reset the auditd connection state
488  * @pid: auditd PID
489  * @portid: auditd netlink portid
490  * @net: auditd network namespace pointer
491  *
492  * Description:
493  * This function will obtain and drop network namespace references as
494  * necessary.  Returns zero on success, negative values on failure.
495  */
496 static int auditd_set(struct pid *pid, u32 portid, struct net *net)
497 {
498 	unsigned long flags;
499 	struct auditd_connection *ac_old, *ac_new;
500 
501 	if (!pid || !net)
502 		return -EINVAL;
503 
504 	ac_new = kzalloc(sizeof(*ac_new), GFP_KERNEL);
505 	if (!ac_new)
506 		return -ENOMEM;
507 	ac_new->pid = get_pid(pid);
508 	ac_new->portid = portid;
509 	ac_new->net = get_net(net);
510 
511 	spin_lock_irqsave(&auditd_conn_lock, flags);
512 	ac_old = rcu_dereference_protected(auditd_conn,
513 					   lockdep_is_held(&auditd_conn_lock));
514 	rcu_assign_pointer(auditd_conn, ac_new);
515 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
516 
517 	if (ac_old)
518 		call_rcu(&ac_old->rcu, auditd_conn_free);
519 
520 	return 0;
521 }
522 
523 /**
524  * kauditd_printk_skb - Print the audit record to the ring buffer
525  * @skb: audit record
526  *
527  * Whatever the reason, this packet may not make it to the auditd connection
528  * so write it via printk so the information isn't completely lost.
529  */
530 static void kauditd_printk_skb(struct sk_buff *skb)
531 {
532 	struct nlmsghdr *nlh = nlmsg_hdr(skb);
533 	char *data = nlmsg_data(nlh);
534 
535 	if (nlh->nlmsg_type != AUDIT_EOE && printk_ratelimit())
536 		pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
537 }
538 
539 /**
540  * kauditd_rehold_skb - Handle a audit record send failure in the hold queue
541  * @skb: audit record
542  * @error: error code (unused)
543  *
544  * Description:
545  * This should only be used by the kauditd_thread when it fails to flush the
546  * hold queue.
547  */
548 static void kauditd_rehold_skb(struct sk_buff *skb, __always_unused int error)
549 {
550 	/* put the record back in the queue */
551 	skb_queue_tail(&audit_hold_queue, skb);
552 }
553 
554 /**
555  * kauditd_hold_skb - Queue an audit record, waiting for auditd
556  * @skb: audit record
557  * @error: error code
558  *
559  * Description:
560  * Queue the audit record, waiting for an instance of auditd.  When this
561  * function is called we haven't given up yet on sending the record, but things
562  * are not looking good.  The first thing we want to do is try to write the
563  * record via printk and then see if we want to try and hold on to the record
564  * and queue it, if we have room.  If we want to hold on to the record, but we
565  * don't have room, record a record lost message.
566  */
567 static void kauditd_hold_skb(struct sk_buff *skb, int error)
568 {
569 	/* at this point it is uncertain if we will ever send this to auditd so
570 	 * try to send the message via printk before we go any further */
571 	kauditd_printk_skb(skb);
572 
573 	/* can we just silently drop the message? */
574 	if (!audit_default)
575 		goto drop;
576 
577 	/* the hold queue is only for when the daemon goes away completely,
578 	 * not -EAGAIN failures; if we are in a -EAGAIN state requeue the
579 	 * record on the retry queue unless it's full, in which case drop it
580 	 */
581 	if (error == -EAGAIN) {
582 		if (!audit_backlog_limit ||
583 		    skb_queue_len(&audit_retry_queue) < audit_backlog_limit) {
584 			skb_queue_tail(&audit_retry_queue, skb);
585 			return;
586 		}
587 		audit_log_lost("kauditd retry queue overflow");
588 		goto drop;
589 	}
590 
591 	/* if we have room in the hold queue, queue the message */
592 	if (!audit_backlog_limit ||
593 	    skb_queue_len(&audit_hold_queue) < audit_backlog_limit) {
594 		skb_queue_tail(&audit_hold_queue, skb);
595 		return;
596 	}
597 
598 	/* we have no other options - drop the message */
599 	audit_log_lost("kauditd hold queue overflow");
600 drop:
601 	kfree_skb(skb);
602 }
603 
604 /**
605  * kauditd_retry_skb - Queue an audit record, attempt to send again to auditd
606  * @skb: audit record
607  * @error: error code (unused)
608  *
609  * Description:
610  * Not as serious as kauditd_hold_skb() as we still have a connected auditd,
611  * but for some reason we are having problems sending it audit records so
612  * queue the given record and attempt to resend.
613  */
614 static void kauditd_retry_skb(struct sk_buff *skb, __always_unused int error)
615 {
616 	if (!audit_backlog_limit ||
617 	    skb_queue_len(&audit_retry_queue) < audit_backlog_limit) {
618 		skb_queue_tail(&audit_retry_queue, skb);
619 		return;
620 	}
621 
622 	/* we have to drop the record, send it via printk as a last effort */
623 	kauditd_printk_skb(skb);
624 	audit_log_lost("kauditd retry queue overflow");
625 	kfree_skb(skb);
626 }
627 
628 /**
629  * auditd_reset - Disconnect the auditd connection
630  * @ac: auditd connection state
631  *
632  * Description:
633  * Break the auditd/kauditd connection and move all the queued records into the
634  * hold queue in case auditd reconnects.  It is important to note that the @ac
635  * pointer should never be dereferenced inside this function as it may be NULL
636  * or invalid, you can only compare the memory address!  If @ac is NULL then
637  * the connection will always be reset.
638  */
639 static void auditd_reset(const struct auditd_connection *ac)
640 {
641 	unsigned long flags;
642 	struct sk_buff *skb;
643 	struct auditd_connection *ac_old;
644 
645 	/* if it isn't already broken, break the connection */
646 	spin_lock_irqsave(&auditd_conn_lock, flags);
647 	ac_old = rcu_dereference_protected(auditd_conn,
648 					   lockdep_is_held(&auditd_conn_lock));
649 	if (ac && ac != ac_old) {
650 		/* someone already registered a new auditd connection */
651 		spin_unlock_irqrestore(&auditd_conn_lock, flags);
652 		return;
653 	}
654 	rcu_assign_pointer(auditd_conn, NULL);
655 	spin_unlock_irqrestore(&auditd_conn_lock, flags);
656 
657 	if (ac_old)
658 		call_rcu(&ac_old->rcu, auditd_conn_free);
659 
660 	/* flush the retry queue to the hold queue, but don't touch the main
661 	 * queue since we need to process that normally for multicast */
662 	while ((skb = skb_dequeue(&audit_retry_queue)))
663 		kauditd_hold_skb(skb, -ECONNREFUSED);
664 }
665 
666 /**
667  * auditd_send_unicast_skb - Send a record via unicast to auditd
668  * @skb: audit record
669  *
670  * Description:
671  * Send a skb to the audit daemon, returns positive/zero values on success and
672  * negative values on failure; in all cases the skb will be consumed by this
673  * function.  If the send results in -ECONNREFUSED the connection with auditd
674  * will be reset.  This function may sleep so callers should not hold any locks
675  * where this would cause a problem.
676  */
677 static int auditd_send_unicast_skb(struct sk_buff *skb)
678 {
679 	int rc;
680 	u32 portid;
681 	struct net *net;
682 	struct sock *sk;
683 	struct auditd_connection *ac;
684 
685 	/* NOTE: we can't call netlink_unicast while in the RCU section so
686 	 *       take a reference to the network namespace and grab local
687 	 *       copies of the namespace, the sock, and the portid; the
688 	 *       namespace and sock aren't going to go away while we hold a
689 	 *       reference and if the portid does become invalid after the RCU
690 	 *       section netlink_unicast() should safely return an error */
691 
692 	rcu_read_lock();
693 	ac = rcu_dereference(auditd_conn);
694 	if (!ac) {
695 		rcu_read_unlock();
696 		kfree_skb(skb);
697 		rc = -ECONNREFUSED;
698 		goto err;
699 	}
700 	net = get_net(ac->net);
701 	sk = audit_get_sk(net);
702 	portid = ac->portid;
703 	rcu_read_unlock();
704 
705 	rc = netlink_unicast(sk, skb, portid, 0);
706 	put_net(net);
707 	if (rc < 0)
708 		goto err;
709 
710 	return rc;
711 
712 err:
713 	if (ac && rc == -ECONNREFUSED)
714 		auditd_reset(ac);
715 	return rc;
716 }
717 
718 /**
719  * kauditd_send_queue - Helper for kauditd_thread to flush skb queues
720  * @sk: the sending sock
721  * @portid: the netlink destination
722  * @queue: the skb queue to process
723  * @retry_limit: limit on number of netlink unicast failures
724  * @skb_hook: per-skb hook for additional processing
725  * @err_hook: hook called if the skb fails the netlink unicast send
726  *
727  * Description:
728  * Run through the given queue and attempt to send the audit records to auditd,
729  * returns zero on success, negative values on failure.  It is up to the caller
730  * to ensure that the @sk is valid for the duration of this function.
731  *
732  */
733 static int kauditd_send_queue(struct sock *sk, u32 portid,
734 			      struct sk_buff_head *queue,
735 			      unsigned int retry_limit,
736 			      void (*skb_hook)(struct sk_buff *skb),
737 			      void (*err_hook)(struct sk_buff *skb, int error))
738 {
739 	int rc = 0;
740 	struct sk_buff *skb = NULL;
741 	struct sk_buff *skb_tail;
742 	unsigned int failed = 0;
743 
744 	/* NOTE: kauditd_thread takes care of all our locking, we just use
745 	 *       the netlink info passed to us (e.g. sk and portid) */
746 
747 	skb_tail = skb_peek_tail(queue);
748 	while ((skb != skb_tail) && (skb = skb_dequeue(queue))) {
749 		/* call the skb_hook for each skb we touch */
750 		if (skb_hook)
751 			(*skb_hook)(skb);
752 
753 		/* can we send to anyone via unicast? */
754 		if (!sk) {
755 			if (err_hook)
756 				(*err_hook)(skb, -ECONNREFUSED);
757 			continue;
758 		}
759 
760 retry:
761 		/* grab an extra skb reference in case of error */
762 		skb_get(skb);
763 		rc = netlink_unicast(sk, skb, portid, 0);
764 		if (rc < 0) {
765 			/* send failed - try a few times unless fatal error */
766 			if (++failed >= retry_limit ||
767 			    rc == -ECONNREFUSED || rc == -EPERM) {
768 				sk = NULL;
769 				if (err_hook)
770 					(*err_hook)(skb, rc);
771 				if (rc == -EAGAIN)
772 					rc = 0;
773 				/* continue to drain the queue */
774 				continue;
775 			} else
776 				goto retry;
777 		} else {
778 			/* skb sent - drop the extra reference and continue */
779 			consume_skb(skb);
780 			failed = 0;
781 		}
782 	}
783 
784 	return (rc >= 0 ? 0 : rc);
785 }
786 
787 /*
788  * kauditd_send_multicast_skb - Send a record to any multicast listeners
789  * @skb: audit record
790  *
791  * Description:
792  * Write a multicast message to anyone listening in the initial network
793  * namespace.  This function doesn't consume an skb as might be expected since
794  * it has to copy it anyways.
795  */
796 static void kauditd_send_multicast_skb(struct sk_buff *skb)
797 {
798 	struct sk_buff *copy;
799 	struct sock *sock = audit_get_sk(&init_net);
800 	struct nlmsghdr *nlh;
801 
802 	/* NOTE: we are not taking an additional reference for init_net since
803 	 *       we don't have to worry about it going away */
804 
805 	if (!netlink_has_listeners(sock, AUDIT_NLGRP_READLOG))
806 		return;
807 
808 	/*
809 	 * The seemingly wasteful skb_copy() rather than bumping the refcount
810 	 * using skb_get() is necessary because non-standard mods are made to
811 	 * the skb by the original kaudit unicast socket send routine.  The
812 	 * existing auditd daemon assumes this breakage.  Fixing this would
813 	 * require co-ordinating a change in the established protocol between
814 	 * the kaudit kernel subsystem and the auditd userspace code.  There is
815 	 * no reason for new multicast clients to continue with this
816 	 * non-compliance.
817 	 */
818 	copy = skb_copy(skb, GFP_KERNEL);
819 	if (!copy)
820 		return;
821 	nlh = nlmsg_hdr(copy);
822 	nlh->nlmsg_len = skb->len;
823 
824 	nlmsg_multicast(sock, copy, 0, AUDIT_NLGRP_READLOG, GFP_KERNEL);
825 }
826 
827 /**
828  * kauditd_thread - Worker thread to send audit records to userspace
829  * @dummy: unused
830  */
831 static int kauditd_thread(void *dummy)
832 {
833 	int rc;
834 	u32 portid = 0;
835 	struct net *net = NULL;
836 	struct sock *sk = NULL;
837 	struct auditd_connection *ac;
838 
839 #define UNICAST_RETRIES 5
840 
841 	set_freezable();
842 	while (!kthread_should_stop()) {
843 		/* NOTE: see the lock comments in auditd_send_unicast_skb() */
844 		rcu_read_lock();
845 		ac = rcu_dereference(auditd_conn);
846 		if (!ac) {
847 			rcu_read_unlock();
848 			goto main_queue;
849 		}
850 		net = get_net(ac->net);
851 		sk = audit_get_sk(net);
852 		portid = ac->portid;
853 		rcu_read_unlock();
854 
855 		/* attempt to flush the hold queue */
856 		rc = kauditd_send_queue(sk, portid,
857 					&audit_hold_queue, UNICAST_RETRIES,
858 					NULL, kauditd_rehold_skb);
859 		if (rc < 0) {
860 			sk = NULL;
861 			auditd_reset(ac);
862 			goto main_queue;
863 		}
864 
865 		/* attempt to flush the retry queue */
866 		rc = kauditd_send_queue(sk, portid,
867 					&audit_retry_queue, UNICAST_RETRIES,
868 					NULL, kauditd_hold_skb);
869 		if (rc < 0) {
870 			sk = NULL;
871 			auditd_reset(ac);
872 			goto main_queue;
873 		}
874 
875 main_queue:
876 		/* process the main queue - do the multicast send and attempt
877 		 * unicast, dump failed record sends to the retry queue; if
878 		 * sk == NULL due to previous failures we will just do the
879 		 * multicast send and move the record to the hold queue */
880 		rc = kauditd_send_queue(sk, portid, &audit_queue, 1,
881 					kauditd_send_multicast_skb,
882 					(sk ?
883 					 kauditd_retry_skb : kauditd_hold_skb));
884 		if (ac && rc < 0)
885 			auditd_reset(ac);
886 		sk = NULL;
887 
888 		/* drop our netns reference, no auditd sends past this line */
889 		if (net) {
890 			put_net(net);
891 			net = NULL;
892 		}
893 
894 		/* we have processed all the queues so wake everyone */
895 		wake_up(&audit_backlog_wait);
896 
897 		/* NOTE: we want to wake up if there is anything on the queue,
898 		 *       regardless of if an auditd is connected, as we need to
899 		 *       do the multicast send and rotate records from the
900 		 *       main queue to the retry/hold queues */
901 		wait_event_freezable(kauditd_wait,
902 				     (skb_queue_len(&audit_queue) ? 1 : 0));
903 	}
904 
905 	return 0;
906 }
907 
908 int audit_send_list_thread(void *_dest)
909 {
910 	struct audit_netlink_list *dest = _dest;
911 	struct sk_buff *skb;
912 	struct sock *sk = audit_get_sk(dest->net);
913 
914 	/* wait for parent to finish and send an ACK */
915 	audit_ctl_lock();
916 	audit_ctl_unlock();
917 
918 	while ((skb = __skb_dequeue(&dest->q)) != NULL)
919 		netlink_unicast(sk, skb, dest->portid, 0);
920 
921 	put_net(dest->net);
922 	kfree(dest);
923 
924 	return 0;
925 }
926 
927 struct sk_buff *audit_make_reply(int seq, int type, int done,
928 				 int multi, const void *payload, int size)
929 {
930 	struct sk_buff	*skb;
931 	struct nlmsghdr	*nlh;
932 	void		*data;
933 	int		flags = multi ? NLM_F_MULTI : 0;
934 	int		t     = done  ? NLMSG_DONE  : type;
935 
936 	skb = nlmsg_new(size, GFP_KERNEL);
937 	if (!skb)
938 		return NULL;
939 
940 	nlh	= nlmsg_put(skb, 0, seq, t, size, flags);
941 	if (!nlh)
942 		goto out_kfree_skb;
943 	data = nlmsg_data(nlh);
944 	memcpy(data, payload, size);
945 	return skb;
946 
947 out_kfree_skb:
948 	kfree_skb(skb);
949 	return NULL;
950 }
951 
952 static void audit_free_reply(struct audit_reply *reply)
953 {
954 	if (!reply)
955 		return;
956 
957 	kfree_skb(reply->skb);
958 	if (reply->net)
959 		put_net(reply->net);
960 	kfree(reply);
961 }
962 
963 static int audit_send_reply_thread(void *arg)
964 {
965 	struct audit_reply *reply = (struct audit_reply *)arg;
966 
967 	audit_ctl_lock();
968 	audit_ctl_unlock();
969 
970 	/* Ignore failure. It'll only happen if the sender goes away,
971 	   because our timeout is set to infinite. */
972 	netlink_unicast(audit_get_sk(reply->net), reply->skb, reply->portid, 0);
973 	reply->skb = NULL;
974 	audit_free_reply(reply);
975 	return 0;
976 }
977 
978 /**
979  * audit_send_reply - send an audit reply message via netlink
980  * @request_skb: skb of request we are replying to (used to target the reply)
981  * @seq: sequence number
982  * @type: audit message type
983  * @done: done (last) flag
984  * @multi: multi-part message flag
985  * @payload: payload data
986  * @size: payload size
987  *
988  * Allocates a skb, builds the netlink message, and sends it to the port id.
989  */
990 static void audit_send_reply(struct sk_buff *request_skb, int seq, int type, int done,
991 			     int multi, const void *payload, int size)
992 {
993 	struct task_struct *tsk;
994 	struct audit_reply *reply;
995 
996 	reply = kzalloc(sizeof(*reply), GFP_KERNEL);
997 	if (!reply)
998 		return;
999 
1000 	reply->skb = audit_make_reply(seq, type, done, multi, payload, size);
1001 	if (!reply->skb)
1002 		goto err;
1003 	reply->net = get_net(sock_net(NETLINK_CB(request_skb).sk));
1004 	reply->portid = NETLINK_CB(request_skb).portid;
1005 
1006 	tsk = kthread_run(audit_send_reply_thread, reply, "audit_send_reply");
1007 	if (IS_ERR(tsk))
1008 		goto err;
1009 
1010 	return;
1011 
1012 err:
1013 	audit_free_reply(reply);
1014 }
1015 
1016 /*
1017  * Check for appropriate CAP_AUDIT_ capabilities on incoming audit
1018  * control messages.
1019  */
1020 static int audit_netlink_ok(struct sk_buff *skb, u16 msg_type)
1021 {
1022 	int err = 0;
1023 
1024 	/* Only support initial user namespace for now. */
1025 	/*
1026 	 * We return ECONNREFUSED because it tricks userspace into thinking
1027 	 * that audit was not configured into the kernel.  Lots of users
1028 	 * configure their PAM stack (because that's what the distro does)
1029 	 * to reject login if unable to send messages to audit.  If we return
1030 	 * ECONNREFUSED the PAM stack thinks the kernel does not have audit
1031 	 * configured in and will let login proceed.  If we return EPERM
1032 	 * userspace will reject all logins.  This should be removed when we
1033 	 * support non init namespaces!!
1034 	 */
1035 	if (current_user_ns() != &init_user_ns)
1036 		return -ECONNREFUSED;
1037 
1038 	switch (msg_type) {
1039 	case AUDIT_LIST:
1040 	case AUDIT_ADD:
1041 	case AUDIT_DEL:
1042 		return -EOPNOTSUPP;
1043 	case AUDIT_GET:
1044 	case AUDIT_SET:
1045 	case AUDIT_GET_FEATURE:
1046 	case AUDIT_SET_FEATURE:
1047 	case AUDIT_LIST_RULES:
1048 	case AUDIT_ADD_RULE:
1049 	case AUDIT_DEL_RULE:
1050 	case AUDIT_SIGNAL_INFO:
1051 	case AUDIT_TTY_GET:
1052 	case AUDIT_TTY_SET:
1053 	case AUDIT_TRIM:
1054 	case AUDIT_MAKE_EQUIV:
1055 		/* Only support auditd and auditctl in initial pid namespace
1056 		 * for now. */
1057 		if (task_active_pid_ns(current) != &init_pid_ns)
1058 			return -EPERM;
1059 
1060 		if (!netlink_capable(skb, CAP_AUDIT_CONTROL))
1061 			err = -EPERM;
1062 		break;
1063 	case AUDIT_USER:
1064 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1065 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1066 		if (!netlink_capable(skb, CAP_AUDIT_WRITE))
1067 			err = -EPERM;
1068 		break;
1069 	default:  /* bad msg */
1070 		err = -EINVAL;
1071 	}
1072 
1073 	return err;
1074 }
1075 
1076 static void audit_log_common_recv_msg(struct audit_context *context,
1077 					struct audit_buffer **ab, u16 msg_type)
1078 {
1079 	uid_t uid = from_kuid(&init_user_ns, current_uid());
1080 	pid_t pid = task_tgid_nr(current);
1081 
1082 	if (!audit_enabled && msg_type != AUDIT_USER_AVC) {
1083 		*ab = NULL;
1084 		return;
1085 	}
1086 
1087 	*ab = audit_log_start(context, GFP_KERNEL, msg_type);
1088 	if (unlikely(!*ab))
1089 		return;
1090 	audit_log_format(*ab, "pid=%d uid=%u ", pid, uid);
1091 	audit_log_session_info(*ab);
1092 	audit_log_task_context(*ab);
1093 }
1094 
1095 static inline void audit_log_user_recv_msg(struct audit_buffer **ab,
1096 					   u16 msg_type)
1097 {
1098 	audit_log_common_recv_msg(NULL, ab, msg_type);
1099 }
1100 
1101 static int is_audit_feature_set(int i)
1102 {
1103 	return af.features & AUDIT_FEATURE_TO_MASK(i);
1104 }
1105 
1106 
1107 static int audit_get_feature(struct sk_buff *skb)
1108 {
1109 	u32 seq;
1110 
1111 	seq = nlmsg_hdr(skb)->nlmsg_seq;
1112 
1113 	audit_send_reply(skb, seq, AUDIT_GET_FEATURE, 0, 0, &af, sizeof(af));
1114 
1115 	return 0;
1116 }
1117 
1118 static void audit_log_feature_change(int which, u32 old_feature, u32 new_feature,
1119 				     u32 old_lock, u32 new_lock, int res)
1120 {
1121 	struct audit_buffer *ab;
1122 
1123 	if (audit_enabled == AUDIT_OFF)
1124 		return;
1125 
1126 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_FEATURE_CHANGE);
1127 	if (!ab)
1128 		return;
1129 	audit_log_task_info(ab);
1130 	audit_log_format(ab, " feature=%s old=%u new=%u old_lock=%u new_lock=%u res=%d",
1131 			 audit_feature_names[which], !!old_feature, !!new_feature,
1132 			 !!old_lock, !!new_lock, res);
1133 	audit_log_end(ab);
1134 }
1135 
1136 static int audit_set_feature(struct audit_features *uaf)
1137 {
1138 	int i;
1139 
1140 	BUILD_BUG_ON(AUDIT_LAST_FEATURE + 1 > ARRAY_SIZE(audit_feature_names));
1141 
1142 	/* if there is ever a version 2 we should handle that here */
1143 
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 		new_lock = (uaf->lock | af.lock) & feature;
1155 		old_lock = af.lock & feature;
1156 
1157 		/* are we changing a locked feature? */
1158 		if (old_lock && (new_feature != old_feature)) {
1159 			audit_log_feature_change(i, old_feature, new_feature,
1160 						 old_lock, new_lock, 0);
1161 			return -EPERM;
1162 		}
1163 	}
1164 	/* nothing invalid, do the changes */
1165 	for (i = 0; i <= AUDIT_LAST_FEATURE; i++) {
1166 		u32 feature = AUDIT_FEATURE_TO_MASK(i);
1167 		u32 old_feature, new_feature, old_lock, new_lock;
1168 
1169 		/* if we are not changing this feature, move along */
1170 		if (!(feature & uaf->mask))
1171 			continue;
1172 
1173 		old_feature = af.features & feature;
1174 		new_feature = uaf->features & feature;
1175 		old_lock = af.lock & feature;
1176 		new_lock = (uaf->lock | af.lock) & feature;
1177 
1178 		if (new_feature != old_feature)
1179 			audit_log_feature_change(i, old_feature, new_feature,
1180 						 old_lock, new_lock, 1);
1181 
1182 		if (new_feature)
1183 			af.features |= feature;
1184 		else
1185 			af.features &= ~feature;
1186 		af.lock |= new_lock;
1187 	}
1188 
1189 	return 0;
1190 }
1191 
1192 static int audit_replace(struct pid *pid)
1193 {
1194 	pid_t pvnr;
1195 	struct sk_buff *skb;
1196 
1197 	pvnr = pid_vnr(pid);
1198 	skb = audit_make_reply(0, AUDIT_REPLACE, 0, 0, &pvnr, sizeof(pvnr));
1199 	if (!skb)
1200 		return -ENOMEM;
1201 	return auditd_send_unicast_skb(skb);
1202 }
1203 
1204 static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1205 {
1206 	u32			seq;
1207 	void			*data;
1208 	int			data_len;
1209 	int			err;
1210 	struct audit_buffer	*ab;
1211 	u16			msg_type = nlh->nlmsg_type;
1212 	struct audit_sig_info   *sig_data;
1213 	char			*ctx = NULL;
1214 	u32			len;
1215 
1216 	err = audit_netlink_ok(skb, msg_type);
1217 	if (err)
1218 		return err;
1219 
1220 	seq  = nlh->nlmsg_seq;
1221 	data = nlmsg_data(nlh);
1222 	data_len = nlmsg_len(nlh);
1223 
1224 	switch (msg_type) {
1225 	case AUDIT_GET: {
1226 		struct audit_status	s;
1227 		memset(&s, 0, sizeof(s));
1228 		s.enabled		   = audit_enabled;
1229 		s.failure		   = audit_failure;
1230 		/* NOTE: use pid_vnr() so the PID is relative to the current
1231 		 *       namespace */
1232 		s.pid			   = auditd_pid_vnr();
1233 		s.rate_limit		   = audit_rate_limit;
1234 		s.backlog_limit		   = audit_backlog_limit;
1235 		s.lost			   = atomic_read(&audit_lost);
1236 		s.backlog		   = skb_queue_len(&audit_queue);
1237 		s.feature_bitmap	   = AUDIT_FEATURE_BITMAP_ALL;
1238 		s.backlog_wait_time	   = audit_backlog_wait_time;
1239 		s.backlog_wait_time_actual = atomic_read(&audit_backlog_wait_time_actual);
1240 		audit_send_reply(skb, seq, AUDIT_GET, 0, 0, &s, sizeof(s));
1241 		break;
1242 	}
1243 	case AUDIT_SET: {
1244 		struct audit_status	s;
1245 		memset(&s, 0, sizeof(s));
1246 		/* guard against past and future API changes */
1247 		memcpy(&s, data, min_t(size_t, sizeof(s), data_len));
1248 		if (s.mask & AUDIT_STATUS_ENABLED) {
1249 			err = audit_set_enabled(s.enabled);
1250 			if (err < 0)
1251 				return err;
1252 		}
1253 		if (s.mask & AUDIT_STATUS_FAILURE) {
1254 			err = audit_set_failure(s.failure);
1255 			if (err < 0)
1256 				return err;
1257 		}
1258 		if (s.mask & AUDIT_STATUS_PID) {
1259 			/* NOTE: we are using the vnr PID functions below
1260 			 *       because the s.pid value is relative to the
1261 			 *       namespace of the caller; at present this
1262 			 *       doesn't matter much since you can really only
1263 			 *       run auditd from the initial pid namespace, but
1264 			 *       something to keep in mind if this changes */
1265 			pid_t new_pid = s.pid;
1266 			pid_t auditd_pid;
1267 			struct pid *req_pid = task_tgid(current);
1268 
1269 			/* Sanity check - PID values must match. Setting
1270 			 * pid to 0 is how auditd ends auditing. */
1271 			if (new_pid && (new_pid != pid_vnr(req_pid)))
1272 				return -EINVAL;
1273 
1274 			/* test the auditd connection */
1275 			audit_replace(req_pid);
1276 
1277 			auditd_pid = auditd_pid_vnr();
1278 			if (auditd_pid) {
1279 				/* replacing a healthy auditd is not allowed */
1280 				if (new_pid) {
1281 					audit_log_config_change("audit_pid",
1282 							new_pid, auditd_pid, 0);
1283 					return -EEXIST;
1284 				}
1285 				/* only current auditd can unregister itself */
1286 				if (pid_vnr(req_pid) != auditd_pid) {
1287 					audit_log_config_change("audit_pid",
1288 							new_pid, auditd_pid, 0);
1289 					return -EACCES;
1290 				}
1291 			}
1292 
1293 			if (new_pid) {
1294 				/* register a new auditd connection */
1295 				err = auditd_set(req_pid,
1296 						 NETLINK_CB(skb).portid,
1297 						 sock_net(NETLINK_CB(skb).sk));
1298 				if (audit_enabled != AUDIT_OFF)
1299 					audit_log_config_change("audit_pid",
1300 								new_pid,
1301 								auditd_pid,
1302 								err ? 0 : 1);
1303 				if (err)
1304 					return err;
1305 
1306 				/* try to process any backlog */
1307 				wake_up_interruptible(&kauditd_wait);
1308 			} else {
1309 				if (audit_enabled != AUDIT_OFF)
1310 					audit_log_config_change("audit_pid",
1311 								new_pid,
1312 								auditd_pid, 1);
1313 
1314 				/* unregister the auditd connection */
1315 				auditd_reset(NULL);
1316 			}
1317 		}
1318 		if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
1319 			err = audit_set_rate_limit(s.rate_limit);
1320 			if (err < 0)
1321 				return err;
1322 		}
1323 		if (s.mask & AUDIT_STATUS_BACKLOG_LIMIT) {
1324 			err = audit_set_backlog_limit(s.backlog_limit);
1325 			if (err < 0)
1326 				return err;
1327 		}
1328 		if (s.mask & AUDIT_STATUS_BACKLOG_WAIT_TIME) {
1329 			if (sizeof(s) > (size_t)nlh->nlmsg_len)
1330 				return -EINVAL;
1331 			if (s.backlog_wait_time > 10*AUDIT_BACKLOG_WAIT_TIME)
1332 				return -EINVAL;
1333 			err = audit_set_backlog_wait_time(s.backlog_wait_time);
1334 			if (err < 0)
1335 				return err;
1336 		}
1337 		if (s.mask == AUDIT_STATUS_LOST) {
1338 			u32 lost = atomic_xchg(&audit_lost, 0);
1339 
1340 			audit_log_config_change("lost", 0, lost, 1);
1341 			return lost;
1342 		}
1343 		if (s.mask == AUDIT_STATUS_BACKLOG_WAIT_TIME_ACTUAL) {
1344 			u32 actual = atomic_xchg(&audit_backlog_wait_time_actual, 0);
1345 
1346 			audit_log_config_change("backlog_wait_time_actual", 0, actual, 1);
1347 			return actual;
1348 		}
1349 		break;
1350 	}
1351 	case AUDIT_GET_FEATURE:
1352 		err = audit_get_feature(skb);
1353 		if (err)
1354 			return err;
1355 		break;
1356 	case AUDIT_SET_FEATURE:
1357 		if (data_len < sizeof(struct audit_features))
1358 			return -EINVAL;
1359 		err = audit_set_feature(data);
1360 		if (err)
1361 			return err;
1362 		break;
1363 	case AUDIT_USER:
1364 	case AUDIT_FIRST_USER_MSG ... AUDIT_LAST_USER_MSG:
1365 	case AUDIT_FIRST_USER_MSG2 ... AUDIT_LAST_USER_MSG2:
1366 		if (!audit_enabled && msg_type != AUDIT_USER_AVC)
1367 			return 0;
1368 		/* exit early if there isn't at least one character to print */
1369 		if (data_len < 2)
1370 			return -EINVAL;
1371 
1372 		err = audit_filter(msg_type, AUDIT_FILTER_USER);
1373 		if (err == 1) { /* match or error */
1374 			char *str = data;
1375 
1376 			err = 0;
1377 			if (msg_type == AUDIT_USER_TTY) {
1378 				err = tty_audit_push();
1379 				if (err)
1380 					break;
1381 			}
1382 			audit_log_user_recv_msg(&ab, msg_type);
1383 			if (msg_type != AUDIT_USER_TTY) {
1384 				/* ensure NULL termination */
1385 				str[data_len - 1] = '\0';
1386 				audit_log_format(ab, " msg='%.*s'",
1387 						 AUDIT_MESSAGE_TEXT_MAX,
1388 						 str);
1389 			} else {
1390 				audit_log_format(ab, " data=");
1391 				if (str[data_len - 1] == '\0')
1392 					data_len--;
1393 				audit_log_n_untrustedstring(ab, str, data_len);
1394 			}
1395 			audit_log_end(ab);
1396 		}
1397 		break;
1398 	case AUDIT_ADD_RULE:
1399 	case AUDIT_DEL_RULE:
1400 		if (data_len < sizeof(struct audit_rule_data))
1401 			return -EINVAL;
1402 		if (audit_enabled == AUDIT_LOCKED) {
1403 			audit_log_common_recv_msg(audit_context(), &ab,
1404 						  AUDIT_CONFIG_CHANGE);
1405 			audit_log_format(ab, " op=%s audit_enabled=%d res=0",
1406 					 msg_type == AUDIT_ADD_RULE ?
1407 						"add_rule" : "remove_rule",
1408 					 audit_enabled);
1409 			audit_log_end(ab);
1410 			return -EPERM;
1411 		}
1412 		err = audit_rule_change(msg_type, seq, data, data_len);
1413 		break;
1414 	case AUDIT_LIST_RULES:
1415 		err = audit_list_rules_send(skb, seq);
1416 		break;
1417 	case AUDIT_TRIM:
1418 		audit_trim_trees();
1419 		audit_log_common_recv_msg(audit_context(), &ab,
1420 					  AUDIT_CONFIG_CHANGE);
1421 		audit_log_format(ab, " op=trim res=1");
1422 		audit_log_end(ab);
1423 		break;
1424 	case AUDIT_MAKE_EQUIV: {
1425 		void *bufp = data;
1426 		u32 sizes[2];
1427 		size_t msglen = data_len;
1428 		char *old, *new;
1429 
1430 		err = -EINVAL;
1431 		if (msglen < 2 * sizeof(u32))
1432 			break;
1433 		memcpy(sizes, bufp, 2 * sizeof(u32));
1434 		bufp += 2 * sizeof(u32);
1435 		msglen -= 2 * sizeof(u32);
1436 		old = audit_unpack_string(&bufp, &msglen, sizes[0]);
1437 		if (IS_ERR(old)) {
1438 			err = PTR_ERR(old);
1439 			break;
1440 		}
1441 		new = audit_unpack_string(&bufp, &msglen, sizes[1]);
1442 		if (IS_ERR(new)) {
1443 			err = PTR_ERR(new);
1444 			kfree(old);
1445 			break;
1446 		}
1447 		/* OK, here comes... */
1448 		err = audit_tag_tree(old, new);
1449 
1450 		audit_log_common_recv_msg(audit_context(), &ab,
1451 					  AUDIT_CONFIG_CHANGE);
1452 		audit_log_format(ab, " op=make_equiv old=");
1453 		audit_log_untrustedstring(ab, old);
1454 		audit_log_format(ab, " new=");
1455 		audit_log_untrustedstring(ab, new);
1456 		audit_log_format(ab, " res=%d", !err);
1457 		audit_log_end(ab);
1458 		kfree(old);
1459 		kfree(new);
1460 		break;
1461 	}
1462 	case AUDIT_SIGNAL_INFO:
1463 		len = 0;
1464 		if (audit_sig_sid) {
1465 			err = security_secid_to_secctx(audit_sig_sid, &ctx, &len);
1466 			if (err)
1467 				return err;
1468 		}
1469 		sig_data = kmalloc(struct_size(sig_data, ctx, len), GFP_KERNEL);
1470 		if (!sig_data) {
1471 			if (audit_sig_sid)
1472 				security_release_secctx(ctx, len);
1473 			return -ENOMEM;
1474 		}
1475 		sig_data->uid = from_kuid(&init_user_ns, audit_sig_uid);
1476 		sig_data->pid = audit_sig_pid;
1477 		if (audit_sig_sid) {
1478 			memcpy(sig_data->ctx, ctx, len);
1479 			security_release_secctx(ctx, len);
1480 		}
1481 		audit_send_reply(skb, seq, AUDIT_SIGNAL_INFO, 0, 0,
1482 				 sig_data, struct_size(sig_data, ctx, len));
1483 		kfree(sig_data);
1484 		break;
1485 	case AUDIT_TTY_GET: {
1486 		struct audit_tty_status s;
1487 		unsigned int t;
1488 
1489 		t = READ_ONCE(current->signal->audit_tty);
1490 		s.enabled = t & AUDIT_TTY_ENABLE;
1491 		s.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1492 
1493 		audit_send_reply(skb, seq, AUDIT_TTY_GET, 0, 0, &s, sizeof(s));
1494 		break;
1495 	}
1496 	case AUDIT_TTY_SET: {
1497 		struct audit_tty_status s, old;
1498 		struct audit_buffer	*ab;
1499 		unsigned int t;
1500 
1501 		memset(&s, 0, sizeof(s));
1502 		/* guard against past and future API changes */
1503 		memcpy(&s, data, min_t(size_t, sizeof(s), data_len));
1504 		/* check if new data is valid */
1505 		if ((s.enabled != 0 && s.enabled != 1) ||
1506 		    (s.log_passwd != 0 && s.log_passwd != 1))
1507 			err = -EINVAL;
1508 
1509 		if (err)
1510 			t = READ_ONCE(current->signal->audit_tty);
1511 		else {
1512 			t = s.enabled | (-s.log_passwd & AUDIT_TTY_LOG_PASSWD);
1513 			t = xchg(&current->signal->audit_tty, t);
1514 		}
1515 		old.enabled = t & AUDIT_TTY_ENABLE;
1516 		old.log_passwd = !!(t & AUDIT_TTY_LOG_PASSWD);
1517 
1518 		audit_log_common_recv_msg(audit_context(), &ab,
1519 					  AUDIT_CONFIG_CHANGE);
1520 		audit_log_format(ab, " op=tty_set old-enabled=%d new-enabled=%d"
1521 				 " old-log_passwd=%d new-log_passwd=%d res=%d",
1522 				 old.enabled, s.enabled, old.log_passwd,
1523 				 s.log_passwd, !err);
1524 		audit_log_end(ab);
1525 		break;
1526 	}
1527 	default:
1528 		err = -EINVAL;
1529 		break;
1530 	}
1531 
1532 	return err < 0 ? err : 0;
1533 }
1534 
1535 /**
1536  * audit_receive - receive messages from a netlink control socket
1537  * @skb: the message buffer
1538  *
1539  * Parse the provided skb and deal with any messages that may be present,
1540  * malformed skbs are discarded.
1541  */
1542 static void audit_receive(struct sk_buff  *skb)
1543 {
1544 	struct nlmsghdr *nlh;
1545 	/*
1546 	 * len MUST be signed for nlmsg_next to be able to dec it below 0
1547 	 * if the nlmsg_len was not aligned
1548 	 */
1549 	int len;
1550 	int err;
1551 
1552 	nlh = nlmsg_hdr(skb);
1553 	len = skb->len;
1554 
1555 	audit_ctl_lock();
1556 	while (nlmsg_ok(nlh, len)) {
1557 		err = audit_receive_msg(skb, nlh);
1558 		/* if err or if this message says it wants a response */
1559 		if (err || (nlh->nlmsg_flags & NLM_F_ACK))
1560 			netlink_ack(skb, nlh, err, NULL);
1561 
1562 		nlh = nlmsg_next(nlh, &len);
1563 	}
1564 	audit_ctl_unlock();
1565 
1566 	/* can't block with the ctrl lock, so penalize the sender now */
1567 	if (audit_backlog_limit &&
1568 	    (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
1569 		DECLARE_WAITQUEUE(wait, current);
1570 
1571 		/* wake kauditd to try and flush the queue */
1572 		wake_up_interruptible(&kauditd_wait);
1573 
1574 		add_wait_queue_exclusive(&audit_backlog_wait, &wait);
1575 		set_current_state(TASK_UNINTERRUPTIBLE);
1576 		schedule_timeout(audit_backlog_wait_time);
1577 		remove_wait_queue(&audit_backlog_wait, &wait);
1578 	}
1579 }
1580 
1581 /* Log information about who is connecting to the audit multicast socket */
1582 static void audit_log_multicast(int group, const char *op, int err)
1583 {
1584 	const struct cred *cred;
1585 	struct tty_struct *tty;
1586 	char comm[sizeof(current->comm)];
1587 	struct audit_buffer *ab;
1588 
1589 	if (!audit_enabled)
1590 		return;
1591 
1592 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_EVENT_LISTENER);
1593 	if (!ab)
1594 		return;
1595 
1596 	cred = current_cred();
1597 	tty = audit_get_tty();
1598 	audit_log_format(ab, "pid=%u uid=%u auid=%u tty=%s ses=%u",
1599 			 task_pid_nr(current),
1600 			 from_kuid(&init_user_ns, cred->uid),
1601 			 from_kuid(&init_user_ns, audit_get_loginuid(current)),
1602 			 tty ? tty_name(tty) : "(none)",
1603 			 audit_get_sessionid(current));
1604 	audit_put_tty(tty);
1605 	audit_log_task_context(ab); /* subj= */
1606 	audit_log_format(ab, " comm=");
1607 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
1608 	audit_log_d_path_exe(ab, current->mm); /* exe= */
1609 	audit_log_format(ab, " nl-mcgrp=%d op=%s res=%d", group, op, !err);
1610 	audit_log_end(ab);
1611 }
1612 
1613 /* Run custom bind function on netlink socket group connect or bind requests. */
1614 static int audit_multicast_bind(struct net *net, int group)
1615 {
1616 	int err = 0;
1617 
1618 	if (!capable(CAP_AUDIT_READ))
1619 		err = -EPERM;
1620 	audit_log_multicast(group, "connect", err);
1621 	return err;
1622 }
1623 
1624 static void audit_multicast_unbind(struct net *net, int group)
1625 {
1626 	audit_log_multicast(group, "disconnect", 0);
1627 }
1628 
1629 static int __net_init audit_net_init(struct net *net)
1630 {
1631 	struct netlink_kernel_cfg cfg = {
1632 		.input	= audit_receive,
1633 		.bind	= audit_multicast_bind,
1634 		.unbind	= audit_multicast_unbind,
1635 		.flags	= NL_CFG_F_NONROOT_RECV,
1636 		.groups	= AUDIT_NLGRP_MAX,
1637 	};
1638 
1639 	struct audit_net *aunet = net_generic(net, audit_net_id);
1640 
1641 	aunet->sk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
1642 	if (aunet->sk == NULL) {
1643 		audit_panic("cannot initialize netlink socket in namespace");
1644 		return -ENOMEM;
1645 	}
1646 	/* limit the timeout in case auditd is blocked/stopped */
1647 	aunet->sk->sk_sndtimeo = HZ / 10;
1648 
1649 	return 0;
1650 }
1651 
1652 static void __net_exit audit_net_exit(struct net *net)
1653 {
1654 	struct audit_net *aunet = net_generic(net, audit_net_id);
1655 
1656 	/* NOTE: you would think that we would want to check the auditd
1657 	 * connection and potentially reset it here if it lives in this
1658 	 * namespace, but since the auditd connection tracking struct holds a
1659 	 * reference to this namespace (see auditd_set()) we are only ever
1660 	 * going to get here after that connection has been released */
1661 
1662 	netlink_kernel_release(aunet->sk);
1663 }
1664 
1665 static struct pernet_operations audit_net_ops __net_initdata = {
1666 	.init = audit_net_init,
1667 	.exit = audit_net_exit,
1668 	.id = &audit_net_id,
1669 	.size = sizeof(struct audit_net),
1670 };
1671 
1672 /* Initialize audit support at boot time. */
1673 static int __init audit_init(void)
1674 {
1675 	int i;
1676 
1677 	if (audit_initialized == AUDIT_DISABLED)
1678 		return 0;
1679 
1680 	audit_buffer_cache = kmem_cache_create("audit_buffer",
1681 					       sizeof(struct audit_buffer),
1682 					       0, SLAB_PANIC, NULL);
1683 
1684 	skb_queue_head_init(&audit_queue);
1685 	skb_queue_head_init(&audit_retry_queue);
1686 	skb_queue_head_init(&audit_hold_queue);
1687 
1688 	for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
1689 		INIT_LIST_HEAD(&audit_inode_hash[i]);
1690 
1691 	mutex_init(&audit_cmd_mutex.lock);
1692 	audit_cmd_mutex.owner = NULL;
1693 
1694 	pr_info("initializing netlink subsys (%s)\n",
1695 		audit_default ? "enabled" : "disabled");
1696 	register_pernet_subsys(&audit_net_ops);
1697 
1698 	audit_initialized = AUDIT_INITIALIZED;
1699 
1700 	kauditd_task = kthread_run(kauditd_thread, NULL, "kauditd");
1701 	if (IS_ERR(kauditd_task)) {
1702 		int err = PTR_ERR(kauditd_task);
1703 		panic("audit: failed to start the kauditd thread (%d)\n", err);
1704 	}
1705 
1706 	audit_log(NULL, GFP_KERNEL, AUDIT_KERNEL,
1707 		"state=initialized audit_enabled=%u res=1",
1708 		 audit_enabled);
1709 
1710 	return 0;
1711 }
1712 postcore_initcall(audit_init);
1713 
1714 /*
1715  * Process kernel command-line parameter at boot time.
1716  * audit={0|off} or audit={1|on}.
1717  */
1718 static int __init audit_enable(char *str)
1719 {
1720 	if (!strcasecmp(str, "off") || !strcmp(str, "0"))
1721 		audit_default = AUDIT_OFF;
1722 	else if (!strcasecmp(str, "on") || !strcmp(str, "1"))
1723 		audit_default = AUDIT_ON;
1724 	else {
1725 		pr_err("audit: invalid 'audit' parameter value (%s)\n", str);
1726 		audit_default = AUDIT_ON;
1727 	}
1728 
1729 	if (audit_default == AUDIT_OFF)
1730 		audit_initialized = AUDIT_DISABLED;
1731 	if (audit_set_enabled(audit_default))
1732 		pr_err("audit: error setting audit state (%d)\n",
1733 		       audit_default);
1734 
1735 	pr_info("%s\n", audit_default ?
1736 		"enabled (after initialization)" : "disabled (until reboot)");
1737 
1738 	return 1;
1739 }
1740 __setup("audit=", audit_enable);
1741 
1742 /* Process kernel command-line parameter at boot time.
1743  * audit_backlog_limit=<n> */
1744 static int __init audit_backlog_limit_set(char *str)
1745 {
1746 	u32 audit_backlog_limit_arg;
1747 
1748 	pr_info("audit_backlog_limit: ");
1749 	if (kstrtouint(str, 0, &audit_backlog_limit_arg)) {
1750 		pr_cont("using default of %u, unable to parse %s\n",
1751 			audit_backlog_limit, str);
1752 		return 1;
1753 	}
1754 
1755 	audit_backlog_limit = audit_backlog_limit_arg;
1756 	pr_cont("%d\n", audit_backlog_limit);
1757 
1758 	return 1;
1759 }
1760 __setup("audit_backlog_limit=", audit_backlog_limit_set);
1761 
1762 static void audit_buffer_free(struct audit_buffer *ab)
1763 {
1764 	if (!ab)
1765 		return;
1766 
1767 	kfree_skb(ab->skb);
1768 	kmem_cache_free(audit_buffer_cache, ab);
1769 }
1770 
1771 static struct audit_buffer *audit_buffer_alloc(struct audit_context *ctx,
1772 					       gfp_t gfp_mask, int type)
1773 {
1774 	struct audit_buffer *ab;
1775 
1776 	ab = kmem_cache_alloc(audit_buffer_cache, gfp_mask);
1777 	if (!ab)
1778 		return NULL;
1779 
1780 	ab->skb = nlmsg_new(AUDIT_BUFSIZ, gfp_mask);
1781 	if (!ab->skb)
1782 		goto err;
1783 	if (!nlmsg_put(ab->skb, 0, 0, type, 0, 0))
1784 		goto err;
1785 
1786 	ab->ctx = ctx;
1787 	ab->gfp_mask = gfp_mask;
1788 
1789 	return ab;
1790 
1791 err:
1792 	audit_buffer_free(ab);
1793 	return NULL;
1794 }
1795 
1796 /**
1797  * audit_serial - compute a serial number for the audit record
1798  *
1799  * Compute a serial number for the audit record.  Audit records are
1800  * written to user-space as soon as they are generated, so a complete
1801  * audit record may be written in several pieces.  The timestamp of the
1802  * record and this serial number are used by the user-space tools to
1803  * determine which pieces belong to the same audit record.  The
1804  * (timestamp,serial) tuple is unique for each syscall and is live from
1805  * syscall entry to syscall exit.
1806  *
1807  * NOTE: Another possibility is to store the formatted records off the
1808  * audit context (for those records that have a context), and emit them
1809  * all at syscall exit.  However, this could delay the reporting of
1810  * significant errors until syscall exit (or never, if the system
1811  * halts).
1812  */
1813 unsigned int audit_serial(void)
1814 {
1815 	static atomic_t serial = ATOMIC_INIT(0);
1816 
1817 	return atomic_inc_return(&serial);
1818 }
1819 
1820 static inline void audit_get_stamp(struct audit_context *ctx,
1821 				   struct timespec64 *t, unsigned int *serial)
1822 {
1823 	if (!ctx || !auditsc_get_stamp(ctx, t, serial)) {
1824 		ktime_get_coarse_real_ts64(t);
1825 		*serial = audit_serial();
1826 	}
1827 }
1828 
1829 /**
1830  * audit_log_start - obtain an audit buffer
1831  * @ctx: audit_context (may be NULL)
1832  * @gfp_mask: type of allocation
1833  * @type: audit message type
1834  *
1835  * Returns audit_buffer pointer on success or NULL on error.
1836  *
1837  * Obtain an audit buffer.  This routine does locking to obtain the
1838  * audit buffer, but then no locking is required for calls to
1839  * audit_log_*format.  If the task (ctx) is a task that is currently in a
1840  * syscall, then the syscall is marked as auditable and an audit record
1841  * will be written at syscall exit.  If there is no associated task, then
1842  * task context (ctx) should be NULL.
1843  */
1844 struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
1845 				     int type)
1846 {
1847 	struct audit_buffer *ab;
1848 	struct timespec64 t;
1849 	unsigned int serial;
1850 
1851 	if (audit_initialized != AUDIT_INITIALIZED)
1852 		return NULL;
1853 
1854 	if (unlikely(!audit_filter(type, AUDIT_FILTER_EXCLUDE)))
1855 		return NULL;
1856 
1857 	/* NOTE: don't ever fail/sleep on these two conditions:
1858 	 * 1. auditd generated record - since we need auditd to drain the
1859 	 *    queue; also, when we are checking for auditd, compare PIDs using
1860 	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
1861 	 *    using a PID anchored in the caller's namespace
1862 	 * 2. generator holding the audit_cmd_mutex - we don't want to block
1863 	 *    while holding the mutex, although we do penalize the sender
1864 	 *    later in audit_receive() when it is safe to block
1865 	 */
1866 	if (!(auditd_test_task(current) || audit_ctl_owner_current())) {
1867 		long stime = audit_backlog_wait_time;
1868 
1869 		while (audit_backlog_limit &&
1870 		       (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
1871 			/* wake kauditd to try and flush the queue */
1872 			wake_up_interruptible(&kauditd_wait);
1873 
1874 			/* sleep if we are allowed and we haven't exhausted our
1875 			 * backlog wait limit */
1876 			if (gfpflags_allow_blocking(gfp_mask) && (stime > 0)) {
1877 				long rtime = stime;
1878 
1879 				DECLARE_WAITQUEUE(wait, current);
1880 
1881 				add_wait_queue_exclusive(&audit_backlog_wait,
1882 							 &wait);
1883 				set_current_state(TASK_UNINTERRUPTIBLE);
1884 				stime = schedule_timeout(rtime);
1885 				atomic_add(rtime - stime, &audit_backlog_wait_time_actual);
1886 				remove_wait_queue(&audit_backlog_wait, &wait);
1887 			} else {
1888 				if (audit_rate_check() && printk_ratelimit())
1889 					pr_warn("audit_backlog=%d > audit_backlog_limit=%d\n",
1890 						skb_queue_len(&audit_queue),
1891 						audit_backlog_limit);
1892 				audit_log_lost("backlog limit exceeded");
1893 				return NULL;
1894 			}
1895 		}
1896 	}
1897 
1898 	ab = audit_buffer_alloc(ctx, gfp_mask, type);
1899 	if (!ab) {
1900 		audit_log_lost("out of memory in audit_log_start");
1901 		return NULL;
1902 	}
1903 
1904 	audit_get_stamp(ab->ctx, &t, &serial);
1905 	/* cancel dummy context to enable supporting records */
1906 	if (ctx)
1907 		ctx->dummy = 0;
1908 	audit_log_format(ab, "audit(%llu.%03lu:%u): ",
1909 			 (unsigned long long)t.tv_sec, t.tv_nsec/1000000, serial);
1910 
1911 	return ab;
1912 }
1913 
1914 /**
1915  * audit_expand - expand skb in the audit buffer
1916  * @ab: audit_buffer
1917  * @extra: space to add at tail of the skb
1918  *
1919  * Returns 0 (no space) on failed expansion, or available space if
1920  * successful.
1921  */
1922 static inline int audit_expand(struct audit_buffer *ab, int extra)
1923 {
1924 	struct sk_buff *skb = ab->skb;
1925 	int oldtail = skb_tailroom(skb);
1926 	int ret = pskb_expand_head(skb, 0, extra, ab->gfp_mask);
1927 	int newtail = skb_tailroom(skb);
1928 
1929 	if (ret < 0) {
1930 		audit_log_lost("out of memory in audit_expand");
1931 		return 0;
1932 	}
1933 
1934 	skb->truesize += newtail - oldtail;
1935 	return newtail;
1936 }
1937 
1938 /*
1939  * Format an audit message into the audit buffer.  If there isn't enough
1940  * room in the audit buffer, more room will be allocated and vsnprint
1941  * will be called a second time.  Currently, we assume that a printk
1942  * can't format message larger than 1024 bytes, so we don't either.
1943  */
1944 static void audit_log_vformat(struct audit_buffer *ab, const char *fmt,
1945 			      va_list args)
1946 {
1947 	int len, avail;
1948 	struct sk_buff *skb;
1949 	va_list args2;
1950 
1951 	if (!ab)
1952 		return;
1953 
1954 	BUG_ON(!ab->skb);
1955 	skb = ab->skb;
1956 	avail = skb_tailroom(skb);
1957 	if (avail == 0) {
1958 		avail = audit_expand(ab, AUDIT_BUFSIZ);
1959 		if (!avail)
1960 			goto out;
1961 	}
1962 	va_copy(args2, args);
1963 	len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args);
1964 	if (len >= avail) {
1965 		/* The printk buffer is 1024 bytes long, so if we get
1966 		 * here and AUDIT_BUFSIZ is at least 1024, then we can
1967 		 * log everything that printk could have logged. */
1968 		avail = audit_expand(ab,
1969 			max_t(unsigned, AUDIT_BUFSIZ, 1+len-avail));
1970 		if (!avail)
1971 			goto out_va_end;
1972 		len = vsnprintf(skb_tail_pointer(skb), avail, fmt, args2);
1973 	}
1974 	if (len > 0)
1975 		skb_put(skb, len);
1976 out_va_end:
1977 	va_end(args2);
1978 out:
1979 	return;
1980 }
1981 
1982 /**
1983  * audit_log_format - format a message into the audit buffer.
1984  * @ab: audit_buffer
1985  * @fmt: format string
1986  * @...: optional parameters matching @fmt string
1987  *
1988  * All the work is done in audit_log_vformat.
1989  */
1990 void audit_log_format(struct audit_buffer *ab, const char *fmt, ...)
1991 {
1992 	va_list args;
1993 
1994 	if (!ab)
1995 		return;
1996 	va_start(args, fmt);
1997 	audit_log_vformat(ab, fmt, args);
1998 	va_end(args);
1999 }
2000 
2001 /**
2002  * audit_log_n_hex - convert a buffer to hex and append it to the audit skb
2003  * @ab: the audit_buffer
2004  * @buf: buffer to convert to hex
2005  * @len: length of @buf to be converted
2006  *
2007  * No return value; failure to expand is silently ignored.
2008  *
2009  * This function will take the passed buf and convert it into a string of
2010  * ascii hex digits. The new string is placed onto the skb.
2011  */
2012 void audit_log_n_hex(struct audit_buffer *ab, const unsigned char *buf,
2013 		size_t len)
2014 {
2015 	int i, avail, new_len;
2016 	unsigned char *ptr;
2017 	struct sk_buff *skb;
2018 
2019 	if (!ab)
2020 		return;
2021 
2022 	BUG_ON(!ab->skb);
2023 	skb = ab->skb;
2024 	avail = skb_tailroom(skb);
2025 	new_len = len<<1;
2026 	if (new_len >= avail) {
2027 		/* Round the buffer request up to the next multiple */
2028 		new_len = AUDIT_BUFSIZ*(((new_len-avail)/AUDIT_BUFSIZ) + 1);
2029 		avail = audit_expand(ab, new_len);
2030 		if (!avail)
2031 			return;
2032 	}
2033 
2034 	ptr = skb_tail_pointer(skb);
2035 	for (i = 0; i < len; i++)
2036 		ptr = hex_byte_pack_upper(ptr, buf[i]);
2037 	*ptr = 0;
2038 	skb_put(skb, len << 1); /* new string is twice the old string */
2039 }
2040 
2041 /*
2042  * Format a string of no more than slen characters into the audit buffer,
2043  * enclosed in quote marks.
2044  */
2045 void audit_log_n_string(struct audit_buffer *ab, const char *string,
2046 			size_t slen)
2047 {
2048 	int avail, new_len;
2049 	unsigned char *ptr;
2050 	struct sk_buff *skb;
2051 
2052 	if (!ab)
2053 		return;
2054 
2055 	BUG_ON(!ab->skb);
2056 	skb = ab->skb;
2057 	avail = skb_tailroom(skb);
2058 	new_len = slen + 3;	/* enclosing quotes + null terminator */
2059 	if (new_len > avail) {
2060 		avail = audit_expand(ab, new_len);
2061 		if (!avail)
2062 			return;
2063 	}
2064 	ptr = skb_tail_pointer(skb);
2065 	*ptr++ = '"';
2066 	memcpy(ptr, string, slen);
2067 	ptr += slen;
2068 	*ptr++ = '"';
2069 	*ptr = 0;
2070 	skb_put(skb, slen + 2);	/* don't include null terminator */
2071 }
2072 
2073 /**
2074  * audit_string_contains_control - does a string need to be logged in hex
2075  * @string: string to be checked
2076  * @len: max length of the string to check
2077  */
2078 bool audit_string_contains_control(const char *string, size_t len)
2079 {
2080 	const unsigned char *p;
2081 	for (p = string; p < (const unsigned char *)string + len; p++) {
2082 		if (*p == '"' || *p < 0x21 || *p > 0x7e)
2083 			return true;
2084 	}
2085 	return false;
2086 }
2087 
2088 /**
2089  * audit_log_n_untrustedstring - log a string that may contain random characters
2090  * @ab: audit_buffer
2091  * @len: length of string (not including trailing null)
2092  * @string: string to be logged
2093  *
2094  * This code will escape a string that is passed to it if the string
2095  * contains a control character, unprintable character, double quote mark,
2096  * or a space. Unescaped strings will start and end with a double quote mark.
2097  * Strings that are escaped are printed in hex (2 digits per char).
2098  *
2099  * The caller specifies the number of characters in the string to log, which may
2100  * or may not be the entire string.
2101  */
2102 void audit_log_n_untrustedstring(struct audit_buffer *ab, const char *string,
2103 				 size_t len)
2104 {
2105 	if (audit_string_contains_control(string, len))
2106 		audit_log_n_hex(ab, string, len);
2107 	else
2108 		audit_log_n_string(ab, string, len);
2109 }
2110 
2111 /**
2112  * audit_log_untrustedstring - log a string that may contain random characters
2113  * @ab: audit_buffer
2114  * @string: string to be logged
2115  *
2116  * Same as audit_log_n_untrustedstring(), except that strlen is used to
2117  * determine string length.
2118  */
2119 void audit_log_untrustedstring(struct audit_buffer *ab, const char *string)
2120 {
2121 	audit_log_n_untrustedstring(ab, string, strlen(string));
2122 }
2123 
2124 /* This is a helper-function to print the escaped d_path */
2125 void audit_log_d_path(struct audit_buffer *ab, const char *prefix,
2126 		      const struct path *path)
2127 {
2128 	char *p, *pathname;
2129 
2130 	if (prefix)
2131 		audit_log_format(ab, "%s", prefix);
2132 
2133 	/* We will allow 11 spaces for ' (deleted)' to be appended */
2134 	pathname = kmalloc(PATH_MAX+11, ab->gfp_mask);
2135 	if (!pathname) {
2136 		audit_log_format(ab, "\"<no_memory>\"");
2137 		return;
2138 	}
2139 	p = d_path(path, pathname, PATH_MAX+11);
2140 	if (IS_ERR(p)) { /* Should never happen since we send PATH_MAX */
2141 		/* FIXME: can we save some information here? */
2142 		audit_log_format(ab, "\"<too_long>\"");
2143 	} else
2144 		audit_log_untrustedstring(ab, p);
2145 	kfree(pathname);
2146 }
2147 
2148 void audit_log_session_info(struct audit_buffer *ab)
2149 {
2150 	unsigned int sessionid = audit_get_sessionid(current);
2151 	uid_t auid = from_kuid(&init_user_ns, audit_get_loginuid(current));
2152 
2153 	audit_log_format(ab, "auid=%u ses=%u", auid, sessionid);
2154 }
2155 
2156 void audit_log_key(struct audit_buffer *ab, char *key)
2157 {
2158 	audit_log_format(ab, " key=");
2159 	if (key)
2160 		audit_log_untrustedstring(ab, key);
2161 	else
2162 		audit_log_format(ab, "(null)");
2163 }
2164 
2165 int audit_log_task_context(struct audit_buffer *ab)
2166 {
2167 	char *ctx = NULL;
2168 	unsigned len;
2169 	int error;
2170 	u32 sid;
2171 
2172 	security_current_getsecid_subj(&sid);
2173 	if (!sid)
2174 		return 0;
2175 
2176 	error = security_secid_to_secctx(sid, &ctx, &len);
2177 	if (error) {
2178 		if (error != -EINVAL)
2179 			goto error_path;
2180 		return 0;
2181 	}
2182 
2183 	audit_log_format(ab, " subj=%s", ctx);
2184 	security_release_secctx(ctx, len);
2185 	return 0;
2186 
2187 error_path:
2188 	audit_panic("error in audit_log_task_context");
2189 	return error;
2190 }
2191 EXPORT_SYMBOL(audit_log_task_context);
2192 
2193 void audit_log_d_path_exe(struct audit_buffer *ab,
2194 			  struct mm_struct *mm)
2195 {
2196 	struct file *exe_file;
2197 
2198 	if (!mm)
2199 		goto out_null;
2200 
2201 	exe_file = get_mm_exe_file(mm);
2202 	if (!exe_file)
2203 		goto out_null;
2204 
2205 	audit_log_d_path(ab, " exe=", &exe_file->f_path);
2206 	fput(exe_file);
2207 	return;
2208 out_null:
2209 	audit_log_format(ab, " exe=(null)");
2210 }
2211 
2212 struct tty_struct *audit_get_tty(void)
2213 {
2214 	struct tty_struct *tty = NULL;
2215 	unsigned long flags;
2216 
2217 	spin_lock_irqsave(&current->sighand->siglock, flags);
2218 	if (current->signal)
2219 		tty = tty_kref_get(current->signal->tty);
2220 	spin_unlock_irqrestore(&current->sighand->siglock, flags);
2221 	return tty;
2222 }
2223 
2224 void audit_put_tty(struct tty_struct *tty)
2225 {
2226 	tty_kref_put(tty);
2227 }
2228 
2229 void audit_log_task_info(struct audit_buffer *ab)
2230 {
2231 	const struct cred *cred;
2232 	char comm[sizeof(current->comm)];
2233 	struct tty_struct *tty;
2234 
2235 	if (!ab)
2236 		return;
2237 
2238 	cred = current_cred();
2239 	tty = audit_get_tty();
2240 	audit_log_format(ab,
2241 			 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
2242 			 " euid=%u suid=%u fsuid=%u"
2243 			 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
2244 			 task_ppid_nr(current),
2245 			 task_tgid_nr(current),
2246 			 from_kuid(&init_user_ns, audit_get_loginuid(current)),
2247 			 from_kuid(&init_user_ns, cred->uid),
2248 			 from_kgid(&init_user_ns, cred->gid),
2249 			 from_kuid(&init_user_ns, cred->euid),
2250 			 from_kuid(&init_user_ns, cred->suid),
2251 			 from_kuid(&init_user_ns, cred->fsuid),
2252 			 from_kgid(&init_user_ns, cred->egid),
2253 			 from_kgid(&init_user_ns, cred->sgid),
2254 			 from_kgid(&init_user_ns, cred->fsgid),
2255 			 tty ? tty_name(tty) : "(none)",
2256 			 audit_get_sessionid(current));
2257 	audit_put_tty(tty);
2258 	audit_log_format(ab, " comm=");
2259 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2260 	audit_log_d_path_exe(ab, current->mm);
2261 	audit_log_task_context(ab);
2262 }
2263 EXPORT_SYMBOL(audit_log_task_info);
2264 
2265 /**
2266  * audit_log_path_denied - report a path restriction denial
2267  * @type: audit message type (AUDIT_ANOM_LINK, AUDIT_ANOM_CREAT, etc)
2268  * @operation: specific operation name
2269  */
2270 void audit_log_path_denied(int type, const char *operation)
2271 {
2272 	struct audit_buffer *ab;
2273 
2274 	if (!audit_enabled || audit_dummy_context())
2275 		return;
2276 
2277 	/* Generate log with subject, operation, outcome. */
2278 	ab = audit_log_start(audit_context(), GFP_KERNEL, type);
2279 	if (!ab)
2280 		return;
2281 	audit_log_format(ab, "op=%s", operation);
2282 	audit_log_task_info(ab);
2283 	audit_log_format(ab, " res=0");
2284 	audit_log_end(ab);
2285 }
2286 
2287 /* global counter which is incremented every time something logs in */
2288 static atomic_t session_id = ATOMIC_INIT(0);
2289 
2290 static int audit_set_loginuid_perm(kuid_t loginuid)
2291 {
2292 	/* if we are unset, we don't need privs */
2293 	if (!audit_loginuid_set(current))
2294 		return 0;
2295 	/* if AUDIT_FEATURE_LOGINUID_IMMUTABLE means never ever allow a change*/
2296 	if (is_audit_feature_set(AUDIT_FEATURE_LOGINUID_IMMUTABLE))
2297 		return -EPERM;
2298 	/* it is set, you need permission */
2299 	if (!capable(CAP_AUDIT_CONTROL))
2300 		return -EPERM;
2301 	/* reject if this is not an unset and we don't allow that */
2302 	if (is_audit_feature_set(AUDIT_FEATURE_ONLY_UNSET_LOGINUID)
2303 				 && uid_valid(loginuid))
2304 		return -EPERM;
2305 	return 0;
2306 }
2307 
2308 static void audit_log_set_loginuid(kuid_t koldloginuid, kuid_t kloginuid,
2309 				   unsigned int oldsessionid,
2310 				   unsigned int sessionid, int rc)
2311 {
2312 	struct audit_buffer *ab;
2313 	uid_t uid, oldloginuid, loginuid;
2314 	struct tty_struct *tty;
2315 
2316 	if (!audit_enabled)
2317 		return;
2318 
2319 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_LOGIN);
2320 	if (!ab)
2321 		return;
2322 
2323 	uid = from_kuid(&init_user_ns, task_uid(current));
2324 	oldloginuid = from_kuid(&init_user_ns, koldloginuid);
2325 	loginuid = from_kuid(&init_user_ns, kloginuid);
2326 	tty = audit_get_tty();
2327 
2328 	audit_log_format(ab, "pid=%d uid=%u", task_tgid_nr(current), uid);
2329 	audit_log_task_context(ab);
2330 	audit_log_format(ab, " old-auid=%u auid=%u tty=%s old-ses=%u ses=%u res=%d",
2331 			 oldloginuid, loginuid, tty ? tty_name(tty) : "(none)",
2332 			 oldsessionid, sessionid, !rc);
2333 	audit_put_tty(tty);
2334 	audit_log_end(ab);
2335 }
2336 
2337 /**
2338  * audit_set_loginuid - set current task's loginuid
2339  * @loginuid: loginuid value
2340  *
2341  * Returns 0.
2342  *
2343  * Called (set) from fs/proc/base.c::proc_loginuid_write().
2344  */
2345 int audit_set_loginuid(kuid_t loginuid)
2346 {
2347 	unsigned int oldsessionid, sessionid = AUDIT_SID_UNSET;
2348 	kuid_t oldloginuid;
2349 	int rc;
2350 
2351 	oldloginuid = audit_get_loginuid(current);
2352 	oldsessionid = audit_get_sessionid(current);
2353 
2354 	rc = audit_set_loginuid_perm(loginuid);
2355 	if (rc)
2356 		goto out;
2357 
2358 	/* are we setting or clearing? */
2359 	if (uid_valid(loginuid)) {
2360 		sessionid = (unsigned int)atomic_inc_return(&session_id);
2361 		if (unlikely(sessionid == AUDIT_SID_UNSET))
2362 			sessionid = (unsigned int)atomic_inc_return(&session_id);
2363 	}
2364 
2365 	current->sessionid = sessionid;
2366 	current->loginuid = loginuid;
2367 out:
2368 	audit_log_set_loginuid(oldloginuid, loginuid, oldsessionid, sessionid, rc);
2369 	return rc;
2370 }
2371 
2372 /**
2373  * audit_signal_info - record signal info for shutting down audit subsystem
2374  * @sig: signal value
2375  * @t: task being signaled
2376  *
2377  * If the audit subsystem is being terminated, record the task (pid)
2378  * and uid that is doing that.
2379  */
2380 int audit_signal_info(int sig, struct task_struct *t)
2381 {
2382 	kuid_t uid = current_uid(), auid;
2383 
2384 	if (auditd_test_task(t) &&
2385 	    (sig == SIGTERM || sig == SIGHUP ||
2386 	     sig == SIGUSR1 || sig == SIGUSR2)) {
2387 		audit_sig_pid = task_tgid_nr(current);
2388 		auid = audit_get_loginuid(current);
2389 		if (uid_valid(auid))
2390 			audit_sig_uid = auid;
2391 		else
2392 			audit_sig_uid = uid;
2393 		security_current_getsecid_subj(&audit_sig_sid);
2394 	}
2395 
2396 	return audit_signal_info_syscall(t);
2397 }
2398 
2399 /**
2400  * audit_log_end - end one audit record
2401  * @ab: the audit_buffer
2402  *
2403  * We can not do a netlink send inside an irq context because it blocks (last
2404  * arg, flags, is not set to MSG_DONTWAIT), so the audit buffer is placed on a
2405  * queue and a kthread is scheduled to remove them from the queue outside the
2406  * irq context.  May be called in any context.
2407  */
2408 void audit_log_end(struct audit_buffer *ab)
2409 {
2410 	struct sk_buff *skb;
2411 	struct nlmsghdr *nlh;
2412 
2413 	if (!ab)
2414 		return;
2415 
2416 	if (audit_rate_check()) {
2417 		skb = ab->skb;
2418 		ab->skb = NULL;
2419 
2420 		/* setup the netlink header, see the comments in
2421 		 * kauditd_send_multicast_skb() for length quirks */
2422 		nlh = nlmsg_hdr(skb);
2423 		nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
2424 
2425 		/* queue the netlink packet and poke the kauditd thread */
2426 		skb_queue_tail(&audit_queue, skb);
2427 		wake_up_interruptible(&kauditd_wait);
2428 	} else
2429 		audit_log_lost("rate limit exceeded");
2430 
2431 	audit_buffer_free(ab);
2432 }
2433 
2434 /**
2435  * audit_log - Log an audit record
2436  * @ctx: audit context
2437  * @gfp_mask: type of allocation
2438  * @type: audit message type
2439  * @fmt: format string to use
2440  * @...: variable parameters matching the format string
2441  *
2442  * This is a convenience function that calls audit_log_start,
2443  * audit_log_vformat, and audit_log_end.  It may be called
2444  * in any context.
2445  */
2446 void audit_log(struct audit_context *ctx, gfp_t gfp_mask, int type,
2447 	       const char *fmt, ...)
2448 {
2449 	struct audit_buffer *ab;
2450 	va_list args;
2451 
2452 	ab = audit_log_start(ctx, gfp_mask, type);
2453 	if (ab) {
2454 		va_start(args, fmt);
2455 		audit_log_vformat(ab, fmt, args);
2456 		va_end(args);
2457 		audit_log_end(ab);
2458 	}
2459 }
2460 
2461 EXPORT_SYMBOL(audit_log_start);
2462 EXPORT_SYMBOL(audit_log_end);
2463 EXPORT_SYMBOL(audit_log_format);
2464 EXPORT_SYMBOL(audit_log);
2465