xref: /freebsd/sys/security/audit/audit_worker.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1999-2008 Apple Inc.
5  * Copyright (c) 2006-2008, 2016, 2018 Robert N. M. Watson
6  * All rights reserved.
7  *
8  * Portions of this software were developed by BAE Systems, the University of
9  * Cambridge Computer Laboratory, and Memorial University under DARPA/AFRL
10  * contract FA8650-15-C-7558 ("CADETS"), as part of the DARPA Transparent
11  * Computing (TC) research program.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1.  Redistributions of source code must retain the above copyright
17  *     notice, this list of conditions and the following disclaimer.
18  * 2.  Redistributions in binary form must reproduce the above copyright
19  *     notice, this list of conditions and the following disclaimer in the
20  *     documentation and/or other materials provided with the distribution.
21  * 3.  Neither the name of Apple Inc. ("Apple") nor the names of
22  *     its contributors may be used to endorse or promote products derived
23  *     from this software without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY APPLE AND ITS CONTRIBUTORS "AS IS" AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED. IN NO EVENT SHALL APPLE OR ITS CONTRIBUTORS BE LIABLE FOR
29  * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
33  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
34  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  */
37 
38 #include <sys/param.h>
39 #include <sys/condvar.h>
40 #include <sys/conf.h>
41 #include <sys/file.h>
42 #include <sys/filedesc.h>
43 #include <sys/fcntl.h>
44 #include <sys/ipc.h>
45 #include <sys/kernel.h>
46 #include <sys/kthread.h>
47 #include <sys/malloc.h>
48 #include <sys/mount.h>
49 #include <sys/namei.h>
50 #include <sys/proc.h>
51 #include <sys/queue.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/protosw.h>
55 #include <sys/domain.h>
56 #include <sys/sx.h>
57 #include <sys/sysproto.h>
58 #include <sys/sysent.h>
59 #include <sys/systm.h>
60 #include <sys/ucred.h>
61 #include <sys/uio.h>
62 #include <sys/un.h>
63 #include <sys/unistd.h>
64 #include <sys/vnode.h>
65 
66 #include <bsm/audit.h>
67 #include <bsm/audit_internal.h>
68 #include <bsm/audit_kevents.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_pcb.h>
72 
73 #include <security/audit/audit.h>
74 #include <security/audit/audit_private.h>
75 
76 #include <vm/uma.h>
77 
78 #include <machine/stdarg.h>
79 
80 /*
81  * Worker thread that will schedule disk I/O, etc.
82  */
83 static struct proc		*audit_thread;
84 
85 /*
86  * audit_cred and audit_vp are the stored credential and vnode to use for
87  * active audit trail.  They are protected by the audit worker lock, which
88  * will be held across all I/O and all rotation to prevent them from being
89  * replaced (rotated) while in use.  The audit_file_rotate_wait flag is set
90  * when the kernel has delivered a trigger to auditd to rotate the trail, and
91  * is cleared when the next rotation takes place.  It is also protected by
92  * the audit worker lock.
93  */
94 static int		 audit_file_rotate_wait;
95 static struct ucred	*audit_cred;
96 static struct vnode	*audit_vp;
97 static off_t		 audit_size;
98 static struct sx	 audit_worker_lock;
99 
100 #define	AUDIT_WORKER_LOCK_INIT()	sx_init(&audit_worker_lock, \
101 					    "audit_worker_lock");
102 #define	AUDIT_WORKER_LOCK_ASSERT()	sx_assert(&audit_worker_lock, \
103 					    SA_XLOCKED)
104 #define	AUDIT_WORKER_LOCK()		sx_xlock(&audit_worker_lock)
105 #define	AUDIT_WORKER_UNLOCK()		sx_xunlock(&audit_worker_lock)
106 
107 static void
108 audit_worker_sync_vp(struct vnode *vp, struct mount *mp, const char *fmt, ...)
109 {
110 	struct mount *mp1;
111 	int error;
112 	va_list va;
113 
114 	va_start(va, fmt);
115 	error = vn_start_write(vp, &mp1, 0);
116 	if (error == 0) {
117 		VOP_LOCK(vp, LK_EXCLUSIVE | LK_RETRY);
118 		(void)VOP_FSYNC(vp, MNT_WAIT, curthread);
119 		VOP_UNLOCK(vp);
120 		vn_finished_write(mp1);
121 	}
122 	vfs_unbusy(mp);
123 	vpanic(fmt, va);
124 	va_end(va);
125 }
126 
127 /*
128  * Write an audit record to a file, performed as the last stage after both
129  * preselection and BSM conversion.  Both space management and write failures
130  * are handled in this function.
131  *
132  * No attempt is made to deal with possible failure to deliver a trigger to
133  * the audit daemon, since the message is asynchronous anyway.
134  */
135 static void
136 audit_record_write(struct vnode *vp, struct ucred *cred, void *data,
137     size_t len)
138 {
139 	static struct timeval last_lowspace_trigger;
140 	static struct timeval last_fail;
141 	static int cur_lowspace_trigger;
142 	struct statfs *mnt_stat;
143 	struct mount *mp;
144 	int error;
145 	static int cur_fail;
146 	long temp;
147 
148 	AUDIT_WORKER_LOCK_ASSERT();
149 
150 	if (vp == NULL)
151 		return;
152 
153 	mp = vp->v_mount;
154 	if (mp == NULL) {
155 		error = EINVAL;
156 		goto fail;
157 	}
158 	error = vfs_busy(mp, 0);
159 	if (error != 0) {
160 		mp = NULL;
161 		goto fail;
162 	}
163 	mnt_stat = &mp->mnt_stat;
164 
165 	/*
166 	 * First, gather statistics on the audit log file and file system so
167 	 * that we know how we're doing on space.  Consider failure of these
168 	 * operations to indicate a future inability to write to the file.
169 	 */
170 	error = VFS_STATFS(mp, mnt_stat);
171 	if (error != 0)
172 		goto fail;
173 
174 	/*
175 	 * We handle four different space-related limits:
176 	 *
177 	 * - A fixed (hard) limit on the minimum free blocks we require on
178 	 *   the file system, and results in record loss, a trigger, and
179 	 *   possible fail stop due to violating invariants.
180 	 *
181 	 * - An administrative (soft) limit, which when fallen below, results
182 	 *   in the kernel notifying the audit daemon of low space.
183 	 *
184 	 * - An audit trail size limit, which when gone above, results in the
185 	 *   kernel notifying the audit daemon that rotation is desired.
186 	 *
187 	 * - The total depth of the kernel audit record exceeding free space,
188 	 *   which can lead to possible fail stop (with drain), in order to
189 	 *   prevent violating invariants.  Failure here doesn't halt
190 	 *   immediately, but prevents new records from being generated.
191 	 *
192 	 * Possibly, the last of these should be handled differently, always
193 	 * allowing a full queue to be lost, rather than trying to prevent
194 	 * loss.
195 	 *
196 	 * First, handle the hard limit, which generates a trigger and may
197 	 * fail stop.  This is handled in the same manner as ENOSPC from
198 	 * VOP_WRITE, and results in record loss.
199 	 */
200 	if (mnt_stat->f_bfree < AUDIT_HARD_LIMIT_FREE_BLOCKS) {
201 		error = ENOSPC;
202 		goto fail_enospc;
203 	}
204 
205 	/*
206 	 * Second, handle falling below the soft limit, if defined; we send
207 	 * the daemon a trigger and continue processing the record.  Triggers
208 	 * are limited to 1/sec.
209 	 */
210 	if (audit_qctrl.aq_minfree != 0) {
211 		temp = mnt_stat->f_blocks / (100 / audit_qctrl.aq_minfree);
212 		if (mnt_stat->f_bfree < temp) {
213 			if (ppsratecheck(&last_lowspace_trigger,
214 			    &cur_lowspace_trigger, 1)) {
215 				(void)audit_send_trigger(
216 				    AUDIT_TRIGGER_LOW_SPACE);
217 				printf("Warning: disk space low (< %d%% free) "
218 				    "on audit log file-system\n",
219 				    audit_qctrl.aq_minfree);
220 			}
221 		}
222 	}
223 
224 	/*
225 	 * If the current file is getting full, generate a rotation trigger
226 	 * to the daemon.  This is only approximate, which is fine as more
227 	 * records may be generated before the daemon rotates the file.
228 	 */
229 	if (audit_fstat.af_filesz != 0 &&
230 	    audit_size >= audit_fstat.af_filesz * (audit_file_rotate_wait + 1)) {
231 		AUDIT_WORKER_LOCK_ASSERT();
232 
233 		audit_file_rotate_wait++;
234 		(void)audit_send_trigger(AUDIT_TRIGGER_ROTATE_KERNEL);
235 	}
236 
237 	/*
238 	 * If the estimated amount of audit data in the audit event queue
239 	 * (plus records allocated but not yet queued) has reached the amount
240 	 * of free space on the disk, then we need to go into an audit fail
241 	 * stop state, in which we do not permit the allocation/committing of
242 	 * any new audit records.  We continue to process records but don't
243 	 * allow any activities that might generate new records.  In the
244 	 * future, we might want to detect when space is available again and
245 	 * allow operation to continue, but this behavior is sufficient to
246 	 * meet fail stop requirements in CAPP.
247 	 */
248 	if (audit_fail_stop) {
249 		if ((unsigned long)((audit_q_len + audit_pre_q_len + 1) *
250 		    MAX_AUDIT_RECORD_SIZE) / mnt_stat->f_bsize >=
251 		    (unsigned long)(mnt_stat->f_bfree)) {
252 			if (ppsratecheck(&last_fail, &cur_fail, 1))
253 				printf("audit_record_write: free space "
254 				    "below size of audit queue, failing "
255 				    "stop\n");
256 			audit_in_failure = 1;
257 		} else if (audit_in_failure) {
258 			/*
259 			 * Note: if we want to handle recovery, this is the
260 			 * spot to do it: unset audit_in_failure, and issue a
261 			 * wakeup on the cv.
262 			 */
263 		}
264 	}
265 
266 	error = vn_rdwr(UIO_WRITE, vp, data, len, (off_t)0, UIO_SYSSPACE,
267 	    IO_APPEND|IO_UNIT, cred, NULL, NULL, curthread);
268 	if (error == ENOSPC)
269 		goto fail_enospc;
270 	else if (error)
271 		goto fail;
272 	AUDIT_WORKER_LOCK_ASSERT();
273 	audit_size += len;
274 
275 	/*
276 	 * Catch completion of a queue drain here; if we're draining and the
277 	 * queue is now empty, fail stop.  That audit_fail_stop is implicitly
278 	 * true, since audit_in_failure can only be set of audit_fail_stop is
279 	 * set.
280 	 *
281 	 * Note: if we handle recovery from audit_in_failure, then we need to
282 	 * make panic here conditional.
283 	 */
284 	if (audit_in_failure) {
285 		if (audit_q_len == 0 && audit_pre_q_len == 0) {
286 			audit_worker_sync_vp(vp, mp,
287 			    "Audit store overflow; record queue drained.");
288 		}
289 	}
290 
291 	vfs_unbusy(mp);
292 	return;
293 
294 fail_enospc:
295 	/*
296 	 * ENOSPC is considered a special case with respect to failures, as
297 	 * this can reflect either our preemptive detection of insufficient
298 	 * space, or ENOSPC returned by the vnode write call.
299 	 */
300 	if (audit_fail_stop) {
301 		audit_worker_sync_vp(vp, mp,
302 		    "Audit log space exhausted and fail-stop set.");
303 	}
304 	(void)audit_send_trigger(AUDIT_TRIGGER_NO_SPACE);
305 	audit_trail_suspended = 1;
306 	audit_syscalls_enabled_update();
307 
308 	/* FALLTHROUGH */
309 fail:
310 	/*
311 	 * We have failed to write to the file, so the current record is
312 	 * lost, which may require an immediate system halt.
313 	 */
314 	if (audit_panic_on_write_fail) {
315 		audit_worker_sync_vp(vp, mp,
316 		    "audit_worker: write error %d\n", error);
317 	} else if (ppsratecheck(&last_fail, &cur_fail, 1))
318 		printf("audit_worker: write error %d\n", error);
319 	if (mp != NULL)
320 		vfs_unbusy(mp);
321 }
322 
323 /*
324  * Given a kernel audit record, process as required.  Kernel audit records
325  * are converted to one, or possibly two, BSM records, depending on whether
326  * there is a user audit record present also.  Kernel records need be
327  * converted to BSM before they can be written out.  Both types will be
328  * written to disk, and audit pipes.
329  */
330 static void
331 audit_worker_process_record(struct kaudit_record *ar)
332 {
333 	struct au_record *bsm;
334 	au_class_t class;
335 	au_event_t event;
336 	au_id_t auid;
337 	int error, sorf;
338 	int locked;
339 
340 	/*
341 	 * We hold the audit worker lock over both writes, if there are two,
342 	 * so that the two records won't be split across a rotation and end
343 	 * up in two different trail files.
344 	 */
345 	if (((ar->k_ar_commit & AR_COMMIT_USER) &&
346 	    (ar->k_ar_commit & AR_PRESELECT_USER_TRAIL)) ||
347 	    (ar->k_ar_commit & AR_PRESELECT_TRAIL)) {
348 		AUDIT_WORKER_LOCK();
349 		locked = 1;
350 	} else
351 		locked = 0;
352 
353 	/*
354 	 * First, handle the user record, if any: commit to the system trail
355 	 * and audit pipes as selected.
356 	 */
357 	if ((ar->k_ar_commit & AR_COMMIT_USER) &&
358 	    (ar->k_ar_commit & AR_PRESELECT_USER_TRAIL)) {
359 		AUDIT_WORKER_LOCK_ASSERT();
360 		audit_record_write(audit_vp, audit_cred, ar->k_udata,
361 		    ar->k_ulen);
362 	}
363 
364 	if ((ar->k_ar_commit & AR_COMMIT_USER) &&
365 	    (ar->k_ar_commit & AR_PRESELECT_USER_PIPE))
366 		audit_pipe_submit_user(ar->k_udata, ar->k_ulen);
367 
368 	if (!(ar->k_ar_commit & AR_COMMIT_KERNEL) ||
369 	    ((ar->k_ar_commit & AR_PRESELECT_PIPE) == 0 &&
370 	    (ar->k_ar_commit & AR_PRESELECT_TRAIL) == 0 &&
371 	    (ar->k_ar_commit & AR_PRESELECT_DTRACE) == 0))
372 		goto out;
373 
374 	auid = ar->k_ar.ar_subj_auid;
375 	event = ar->k_ar.ar_event;
376 	class = au_event_class(event);
377 	if (ar->k_ar.ar_errno == 0)
378 		sorf = AU_PRS_SUCCESS;
379 	else
380 		sorf = AU_PRS_FAILURE;
381 
382 	error = kaudit_to_bsm(ar, &bsm);
383 	switch (error) {
384 	case BSM_NOAUDIT:
385 		goto out;
386 
387 	case BSM_FAILURE:
388 		printf("audit_worker_process_record: BSM_FAILURE\n");
389 		goto out;
390 
391 	case BSM_SUCCESS:
392 		break;
393 
394 	default:
395 		panic("kaudit_to_bsm returned %d", error);
396 	}
397 
398 	if (ar->k_ar_commit & AR_PRESELECT_TRAIL) {
399 		AUDIT_WORKER_LOCK_ASSERT();
400 		audit_record_write(audit_vp, audit_cred, bsm->data, bsm->len);
401 	}
402 
403 	if (ar->k_ar_commit & AR_PRESELECT_PIPE)
404 		audit_pipe_submit(auid, event, class, sorf,
405 		    ar->k_ar_commit & AR_PRESELECT_TRAIL, bsm->data,
406 		    bsm->len);
407 
408 #ifdef KDTRACE_HOOKS
409 	/*
410 	 * Version of the dtaudit commit hook that accepts BSM.
411 	 */
412 	if (ar->k_ar_commit & AR_PRESELECT_DTRACE) {
413 		if (dtaudit_hook_bsm != NULL)
414 			dtaudit_hook_bsm(ar, auid, event, class, sorf,
415 			    bsm->data, bsm->len);
416 	}
417 #endif
418 
419 	kau_free(bsm);
420 out:
421 	if (locked)
422 		AUDIT_WORKER_UNLOCK();
423 }
424 
425 /*
426  * The audit_worker thread is responsible for watching the event queue,
427  * dequeueing records, converting them to BSM format, and committing them to
428  * disk.  In order to minimize lock thrashing, records are dequeued in sets
429  * to a thread-local work queue.
430  *
431  * Note: this means that the effect bound on the size of the pending record
432  * queue is 2x the length of the global queue.
433  */
434 static void
435 audit_worker(void *arg)
436 {
437 	struct kaudit_queue ar_worklist;
438 	struct kaudit_record *ar;
439 	int lowater_signal;
440 
441 	TAILQ_INIT(&ar_worklist);
442 	mtx_lock(&audit_mtx);
443 	while (1) {
444 		mtx_assert(&audit_mtx, MA_OWNED);
445 
446 		/*
447 		 * Wait for a record.
448 		 */
449 		while (TAILQ_EMPTY(&audit_q))
450 			cv_wait(&audit_worker_cv, &audit_mtx);
451 
452 		/*
453 		 * If there are records in the global audit record queue,
454 		 * transfer them to a thread-local queue and process them
455 		 * one by one.  If we cross the low watermark threshold,
456 		 * signal any waiting processes that they may wake up and
457 		 * continue generating records.
458 		 */
459 		lowater_signal = 0;
460 		while ((ar = TAILQ_FIRST(&audit_q))) {
461 			TAILQ_REMOVE(&audit_q, ar, k_q);
462 			audit_q_len--;
463 			if (audit_q_len == audit_qctrl.aq_lowater)
464 				lowater_signal++;
465 			TAILQ_INSERT_TAIL(&ar_worklist, ar, k_q);
466 		}
467 		if (lowater_signal)
468 			cv_broadcast(&audit_watermark_cv);
469 
470 		mtx_unlock(&audit_mtx);
471 		while ((ar = TAILQ_FIRST(&ar_worklist))) {
472 			TAILQ_REMOVE(&ar_worklist, ar, k_q);
473 			audit_worker_process_record(ar);
474 			audit_free(ar);
475 		}
476 		mtx_lock(&audit_mtx);
477 	}
478 }
479 
480 /*
481  * audit_rotate_vnode() is called by a user or kernel thread to configure or
482  * de-configure auditing on a vnode.  The arguments are the replacement
483  * credential (referenced) and vnode (referenced and opened) to substitute
484  * for the current credential and vnode, if any.  If either is set to NULL,
485  * both should be NULL, and this is used to indicate that audit is being
486  * disabled.  Any previous cred/vnode will be closed and freed.  We re-enable
487  * generating rotation requests to auditd.
488  */
489 void
490 audit_rotate_vnode(struct ucred *cred, struct vnode *vp)
491 {
492 	struct ucred *old_audit_cred;
493 	struct vnode *old_audit_vp;
494 	struct vattr vattr;
495 
496 	KASSERT((cred != NULL && vp != NULL) || (cred == NULL && vp == NULL),
497 	    ("audit_rotate_vnode: cred %p vp %p", cred, vp));
498 
499 	if (vp != NULL) {
500 		vn_lock(vp, LK_SHARED | LK_RETRY);
501 		if (VOP_GETATTR(vp, &vattr, cred) != 0)
502 			vattr.va_size = 0;
503 		VOP_UNLOCK(vp);
504 	} else {
505 		vattr.va_size = 0;
506 	}
507 
508 	/*
509 	 * Rotate the vnode/cred, and clear the rotate flag so that we will
510 	 * send a rotate trigger if the new file fills.
511 	 */
512 	AUDIT_WORKER_LOCK();
513 	old_audit_cred = audit_cred;
514 	old_audit_vp = audit_vp;
515 	audit_cred = cred;
516 	audit_vp = vp;
517 	audit_size = vattr.va_size;
518 	audit_file_rotate_wait = 0;
519 	audit_trail_enabled = (audit_vp != NULL);
520 	audit_syscalls_enabled_update();
521 	AUDIT_WORKER_UNLOCK();
522 
523 	/*
524 	 * If there was an old vnode/credential, close and free.
525 	 */
526 	if (old_audit_vp != NULL) {
527 		vn_close(old_audit_vp, AUDIT_CLOSE_FLAGS, old_audit_cred,
528 		    curthread);
529 		crfree(old_audit_cred);
530 	}
531 }
532 
533 void
534 audit_worker_init(void)
535 {
536 	int error;
537 
538 	AUDIT_WORKER_LOCK_INIT();
539 	error = kproc_create(audit_worker, NULL, &audit_thread, RFHIGHPID,
540 	    0, "audit");
541 	if (error)
542 		panic("audit_worker_init: kproc_create returned %d", error);
543 }
544