xref: /freebsd/sys/security/audit/audit_worker.c (revision 1f4bcc459a76b7aa664f3fd557684cd0ba6da352)
1 /*-
2  * Copyright (c) 1999-2008 Apple Inc.
3  * Copyright (c) 2006-2008 Robert N. M. Watson
4  * All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1.  Redistributions of source code must retain the above copyright
10  *     notice, this list of conditions and the following disclaimer.
11  * 2.  Redistributions in binary form must reproduce the above copyright
12  *     notice, this list of conditions and the following disclaimer in the
13  *     documentation and/or other materials provided with the distribution.
14  * 3.  Neither the name of Apple Inc. ("Apple") nor the names of
15  *     its contributors may be used to endorse or promote products derived
16  *     from this software without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY APPLE AND ITS CONTRIBUTORS "AS IS" AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED. IN NO EVENT SHALL APPLE OR ITS CONTRIBUTORS BE LIABLE FOR
22  * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
26  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
27  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28  * POSSIBILITY OF SUCH DAMAGE.
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include <sys/param.h>
35 #include <sys/condvar.h>
36 #include <sys/conf.h>
37 #include <sys/file.h>
38 #include <sys/filedesc.h>
39 #include <sys/fcntl.h>
40 #include <sys/ipc.h>
41 #include <sys/kernel.h>
42 #include <sys/kthread.h>
43 #include <sys/malloc.h>
44 #include <sys/mount.h>
45 #include <sys/namei.h>
46 #include <sys/proc.h>
47 #include <sys/queue.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/domain.h>
52 #include <sys/sx.h>
53 #include <sys/sysproto.h>
54 #include <sys/sysent.h>
55 #include <sys/systm.h>
56 #include <sys/ucred.h>
57 #include <sys/uio.h>
58 #include <sys/un.h>
59 #include <sys/unistd.h>
60 #include <sys/vnode.h>
61 
62 #include <bsm/audit.h>
63 #include <bsm/audit_internal.h>
64 #include <bsm/audit_kevents.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_pcb.h>
68 
69 #include <security/audit/audit.h>
70 #include <security/audit/audit_private.h>
71 
72 #include <vm/uma.h>
73 
74 #include <machine/stdarg.h>
75 
76 /*
77  * Worker thread that will schedule disk I/O, etc.
78  */
79 static struct proc		*audit_thread;
80 
81 /*
82  * audit_cred and audit_vp are the stored credential and vnode to use for
83  * active audit trail.  They are protected by the audit worker lock, which
84  * will be held across all I/O and all rotation to prevent them from being
85  * replaced (rotated) while in use.  The audit_file_rotate_wait flag is set
86  * when the kernel has delivered a trigger to auditd to rotate the trail, and
87  * is cleared when the next rotation takes place.  It is also protected by
88  * the audit worker lock.
89  */
90 static int		 audit_file_rotate_wait;
91 static struct ucred	*audit_cred;
92 static struct vnode	*audit_vp;
93 static off_t		 audit_size;
94 static struct sx	 audit_worker_lock;
95 
96 #define	AUDIT_WORKER_LOCK_INIT()	sx_init(&audit_worker_lock, \
97 					    "audit_worker_lock");
98 #define	AUDIT_WORKER_LOCK_ASSERT()	sx_assert(&audit_worker_lock, \
99 					    SA_XLOCKED)
100 #define	AUDIT_WORKER_LOCK()		sx_xlock(&audit_worker_lock)
101 #define	AUDIT_WORKER_UNLOCK()		sx_xunlock(&audit_worker_lock)
102 
103 static void
104 audit_worker_sync_vp(struct vnode *vp, struct mount *mp, const char *fmt, ...)
105 {
106 	struct mount *mp1;
107 	int error;
108 	va_list va;
109 
110 	va_start(va, fmt);
111 	error = vn_start_write(vp, &mp1, 0);
112 	if (error == 0) {
113 		VOP_LOCK(vp, LK_EXCLUSIVE | LK_RETRY);
114 		(void)VOP_FSYNC(vp, MNT_WAIT, curthread);
115 		VOP_UNLOCK(vp, 0);
116 		vn_finished_write(mp1);
117 	}
118 	vfs_unbusy(mp);
119 	vpanic(fmt, va);
120 	va_end(va);
121 }
122 
123 /*
124  * Write an audit record to a file, performed as the last stage after both
125  * preselection and BSM conversion.  Both space management and write failures
126  * are handled in this function.
127  *
128  * No attempt is made to deal with possible failure to deliver a trigger to
129  * the audit daemon, since the message is asynchronous anyway.
130  */
131 static void
132 audit_record_write(struct vnode *vp, struct ucred *cred, void *data,
133     size_t len)
134 {
135 	static struct timeval last_lowspace_trigger;
136 	static struct timeval last_fail;
137 	static int cur_lowspace_trigger;
138 	struct statfs *mnt_stat;
139 	struct mount *mp;
140 	int error;
141 	static int cur_fail;
142 	long temp;
143 
144 	AUDIT_WORKER_LOCK_ASSERT();
145 
146 	if (vp == NULL)
147 		return;
148 
149 	mp = vp->v_mount;
150 	if (mp == NULL) {
151 		error = EINVAL;
152 		goto fail;
153 	}
154 	error = vfs_busy(mp, 0);
155 	if (error != 0) {
156 		mp = NULL;
157 		goto fail;
158 	}
159 	mnt_stat = &mp->mnt_stat;
160 
161 	/*
162 	 * First, gather statistics on the audit log file and file system so
163 	 * that we know how we're doing on space.  Consider failure of these
164 	 * operations to indicate a future inability to write to the file.
165 	 */
166 	error = VFS_STATFS(mp, mnt_stat);
167 	if (error != 0)
168 		goto fail;
169 
170 	/*
171 	 * We handle four different space-related limits:
172 	 *
173 	 * - A fixed (hard) limit on the minimum free blocks we require on
174 	 *   the file system, and results in record loss, a trigger, and
175 	 *   possible fail stop due to violating invariants.
176 	 *
177 	 * - An administrative (soft) limit, which when fallen below, results
178 	 *   in the kernel notifying the audit daemon of low space.
179 	 *
180 	 * - An audit trail size limit, which when gone above, results in the
181 	 *   kernel notifying the audit daemon that rotation is desired.
182 	 *
183 	 * - The total depth of the kernel audit record exceeding free space,
184 	 *   which can lead to possible fail stop (with drain), in order to
185 	 *   prevent violating invariants.  Failure here doesn't halt
186 	 *   immediately, but prevents new records from being generated.
187 	 *
188 	 * Possibly, the last of these should be handled differently, always
189 	 * allowing a full queue to be lost, rather than trying to prevent
190 	 * loss.
191 	 *
192 	 * First, handle the hard limit, which generates a trigger and may
193 	 * fail stop.  This is handled in the same manner as ENOSPC from
194 	 * VOP_WRITE, and results in record loss.
195 	 */
196 	if (mnt_stat->f_bfree < AUDIT_HARD_LIMIT_FREE_BLOCKS) {
197 		error = ENOSPC;
198 		goto fail_enospc;
199 	}
200 
201 	/*
202 	 * Second, handle falling below the soft limit, if defined; we send
203 	 * the daemon a trigger and continue processing the record.  Triggers
204 	 * are limited to 1/sec.
205 	 */
206 	if (audit_qctrl.aq_minfree != 0) {
207 		temp = mnt_stat->f_blocks / (100 / audit_qctrl.aq_minfree);
208 		if (mnt_stat->f_bfree < temp) {
209 			if (ppsratecheck(&last_lowspace_trigger,
210 			    &cur_lowspace_trigger, 1)) {
211 				(void)audit_send_trigger(
212 				    AUDIT_TRIGGER_LOW_SPACE);
213 				printf("Warning: disk space low (< %d%% free) "
214 				    "on audit log file-system\n",
215 				    audit_qctrl.aq_minfree);
216 			}
217 		}
218 	}
219 
220 	/*
221 	 * If the current file is getting full, generate a rotation trigger
222 	 * to the daemon.  This is only approximate, which is fine as more
223 	 * records may be generated before the daemon rotates the file.
224 	 */
225 	if (audit_fstat.af_filesz != 0 &&
226 	    audit_size >= audit_fstat.af_filesz * (audit_file_rotate_wait + 1)) {
227 		AUDIT_WORKER_LOCK_ASSERT();
228 
229 		audit_file_rotate_wait++;
230 		(void)audit_send_trigger(AUDIT_TRIGGER_ROTATE_KERNEL);
231 	}
232 
233 	/*
234 	 * If the estimated amount of audit data in the audit event queue
235 	 * (plus records allocated but not yet queued) has reached the amount
236 	 * of free space on the disk, then we need to go into an audit fail
237 	 * stop state, in which we do not permit the allocation/committing of
238 	 * any new audit records.  We continue to process records but don't
239 	 * allow any activities that might generate new records.  In the
240 	 * future, we might want to detect when space is available again and
241 	 * allow operation to continue, but this behavior is sufficient to
242 	 * meet fail stop requirements in CAPP.
243 	 */
244 	if (audit_fail_stop) {
245 		if ((unsigned long)((audit_q_len + audit_pre_q_len + 1) *
246 		    MAX_AUDIT_RECORD_SIZE) / mnt_stat->f_bsize >=
247 		    (unsigned long)(mnt_stat->f_bfree)) {
248 			if (ppsratecheck(&last_fail, &cur_fail, 1))
249 				printf("audit_record_write: free space "
250 				    "below size of audit queue, failing "
251 				    "stop\n");
252 			audit_in_failure = 1;
253 		} else if (audit_in_failure) {
254 			/*
255 			 * Note: if we want to handle recovery, this is the
256 			 * spot to do it: unset audit_in_failure, and issue a
257 			 * wakeup on the cv.
258 			 */
259 		}
260 	}
261 
262 	error = vn_rdwr(UIO_WRITE, vp, data, len, (off_t)0, UIO_SYSSPACE,
263 	    IO_APPEND|IO_UNIT, cred, NULL, NULL, curthread);
264 	if (error == ENOSPC)
265 		goto fail_enospc;
266 	else if (error)
267 		goto fail;
268 	AUDIT_WORKER_LOCK_ASSERT();
269 	audit_size += len;
270 
271 	/*
272 	 * Catch completion of a queue drain here; if we're draining and the
273 	 * queue is now empty, fail stop.  That audit_fail_stop is implicitly
274 	 * true, since audit_in_failure can only be set of audit_fail_stop is
275 	 * set.
276 	 *
277 	 * Note: if we handle recovery from audit_in_failure, then we need to
278 	 * make panic here conditional.
279 	 */
280 	if (audit_in_failure) {
281 		if (audit_q_len == 0 && audit_pre_q_len == 0) {
282 			audit_worker_sync_vp(vp, mp,
283 			    "Audit store overflow; record queue drained.");
284 		}
285 	}
286 
287 	vfs_unbusy(mp);
288 	return;
289 
290 fail_enospc:
291 	/*
292 	 * ENOSPC is considered a special case with respect to failures, as
293 	 * this can reflect either our preemptive detection of insufficient
294 	 * space, or ENOSPC returned by the vnode write call.
295 	 */
296 	if (audit_fail_stop) {
297 		audit_worker_sync_vp(vp, mp,
298 		    "Audit log space exhausted and fail-stop set.");
299 	}
300 	(void)audit_send_trigger(AUDIT_TRIGGER_NO_SPACE);
301 	audit_suspended = 1;
302 
303 	/* FALLTHROUGH */
304 fail:
305 	/*
306 	 * We have failed to write to the file, so the current record is
307 	 * lost, which may require an immediate system halt.
308 	 */
309 	if (audit_panic_on_write_fail) {
310 		audit_worker_sync_vp(vp, mp,
311 		    "audit_worker: write error %d\n", error);
312 	} else if (ppsratecheck(&last_fail, &cur_fail, 1))
313 		printf("audit_worker: write error %d\n", error);
314 	if (mp != NULL)
315 		vfs_unbusy(mp);
316 }
317 
318 /*
319  * Given a kernel audit record, process as required.  Kernel audit records
320  * are converted to one, or possibly two, BSM records, depending on whether
321  * there is a user audit record present also.  Kernel records need be
322  * converted to BSM before they can be written out.  Both types will be
323  * written to disk, and audit pipes.
324  */
325 static void
326 audit_worker_process_record(struct kaudit_record *ar)
327 {
328 	struct au_record *bsm;
329 	au_class_t class;
330 	au_event_t event;
331 	au_id_t auid;
332 	int error, sorf;
333 	int locked;
334 
335 	/*
336 	 * We hold the audit worker lock over both writes, if there are two,
337 	 * so that the two records won't be split across a rotation and end
338 	 * up in two different trail files.
339 	 */
340 	if (((ar->k_ar_commit & AR_COMMIT_USER) &&
341 	    (ar->k_ar_commit & AR_PRESELECT_USER_TRAIL)) ||
342 	    (ar->k_ar_commit & AR_PRESELECT_TRAIL)) {
343 		AUDIT_WORKER_LOCK();
344 		locked = 1;
345 	} else
346 		locked = 0;
347 
348 	/*
349 	 * First, handle the user record, if any: commit to the system trail
350 	 * and audit pipes as selected.
351 	 */
352 	if ((ar->k_ar_commit & AR_COMMIT_USER) &&
353 	    (ar->k_ar_commit & AR_PRESELECT_USER_TRAIL)) {
354 		AUDIT_WORKER_LOCK_ASSERT();
355 		audit_record_write(audit_vp, audit_cred, ar->k_udata,
356 		    ar->k_ulen);
357 	}
358 
359 	if ((ar->k_ar_commit & AR_COMMIT_USER) &&
360 	    (ar->k_ar_commit & AR_PRESELECT_USER_PIPE))
361 		audit_pipe_submit_user(ar->k_udata, ar->k_ulen);
362 
363 	if (!(ar->k_ar_commit & AR_COMMIT_KERNEL) ||
364 	    ((ar->k_ar_commit & AR_PRESELECT_PIPE) == 0 &&
365 	    (ar->k_ar_commit & AR_PRESELECT_TRAIL) == 0))
366 		goto out;
367 
368 	auid = ar->k_ar.ar_subj_auid;
369 	event = ar->k_ar.ar_event;
370 	class = au_event_class(event);
371 	if (ar->k_ar.ar_errno == 0)
372 		sorf = AU_PRS_SUCCESS;
373 	else
374 		sorf = AU_PRS_FAILURE;
375 
376 	error = kaudit_to_bsm(ar, &bsm);
377 	switch (error) {
378 	case BSM_NOAUDIT:
379 		goto out;
380 
381 	case BSM_FAILURE:
382 		printf("audit_worker_process_record: BSM_FAILURE\n");
383 		goto out;
384 
385 	case BSM_SUCCESS:
386 		break;
387 
388 	default:
389 		panic("kaudit_to_bsm returned %d", error);
390 	}
391 
392 	if (ar->k_ar_commit & AR_PRESELECT_TRAIL) {
393 		AUDIT_WORKER_LOCK_ASSERT();
394 		audit_record_write(audit_vp, audit_cred, bsm->data, bsm->len);
395 	}
396 
397 	if (ar->k_ar_commit & AR_PRESELECT_PIPE)
398 		audit_pipe_submit(auid, event, class, sorf,
399 		    ar->k_ar_commit & AR_PRESELECT_TRAIL, bsm->data,
400 		    bsm->len);
401 
402 	kau_free(bsm);
403 out:
404 	if (locked)
405 		AUDIT_WORKER_UNLOCK();
406 }
407 
408 /*
409  * The audit_worker thread is responsible for watching the event queue,
410  * dequeueing records, converting them to BSM format, and committing them to
411  * disk.  In order to minimize lock thrashing, records are dequeued in sets
412  * to a thread-local work queue.
413  *
414  * Note: this means that the effect bound on the size of the pending record
415  * queue is 2x the length of the global queue.
416  */
417 static void
418 audit_worker(void *arg)
419 {
420 	struct kaudit_queue ar_worklist;
421 	struct kaudit_record *ar;
422 	int lowater_signal;
423 
424 	TAILQ_INIT(&ar_worklist);
425 	mtx_lock(&audit_mtx);
426 	while (1) {
427 		mtx_assert(&audit_mtx, MA_OWNED);
428 
429 		/*
430 		 * Wait for a record.
431 		 */
432 		while (TAILQ_EMPTY(&audit_q))
433 			cv_wait(&audit_worker_cv, &audit_mtx);
434 
435 		/*
436 		 * If there are records in the global audit record queue,
437 		 * transfer them to a thread-local queue and process them
438 		 * one by one.  If we cross the low watermark threshold,
439 		 * signal any waiting processes that they may wake up and
440 		 * continue generating records.
441 		 */
442 		lowater_signal = 0;
443 		while ((ar = TAILQ_FIRST(&audit_q))) {
444 			TAILQ_REMOVE(&audit_q, ar, k_q);
445 			audit_q_len--;
446 			if (audit_q_len == audit_qctrl.aq_lowater)
447 				lowater_signal++;
448 			TAILQ_INSERT_TAIL(&ar_worklist, ar, k_q);
449 		}
450 		if (lowater_signal)
451 			cv_broadcast(&audit_watermark_cv);
452 
453 		mtx_unlock(&audit_mtx);
454 		while ((ar = TAILQ_FIRST(&ar_worklist))) {
455 			TAILQ_REMOVE(&ar_worklist, ar, k_q);
456 			audit_worker_process_record(ar);
457 			audit_free(ar);
458 		}
459 		mtx_lock(&audit_mtx);
460 	}
461 }
462 
463 /*
464  * audit_rotate_vnode() is called by a user or kernel thread to configure or
465  * de-configure auditing on a vnode.  The arguments are the replacement
466  * credential (referenced) and vnode (referenced and opened) to substitute
467  * for the current credential and vnode, if any.  If either is set to NULL,
468  * both should be NULL, and this is used to indicate that audit is being
469  * disabled.  Any previous cred/vnode will be closed and freed.  We re-enable
470  * generating rotation requests to auditd.
471  */
472 void
473 audit_rotate_vnode(struct ucred *cred, struct vnode *vp)
474 {
475 	struct ucred *old_audit_cred;
476 	struct vnode *old_audit_vp;
477 	struct vattr vattr;
478 
479 	KASSERT((cred != NULL && vp != NULL) || (cred == NULL && vp == NULL),
480 	    ("audit_rotate_vnode: cred %p vp %p", cred, vp));
481 
482 	if (vp != NULL) {
483 		vn_lock(vp, LK_SHARED | LK_RETRY);
484 		if (VOP_GETATTR(vp, &vattr, cred) != 0)
485 			vattr.va_size = 0;
486 		VOP_UNLOCK(vp, 0);
487 	} else {
488 		vattr.va_size = 0;
489 	}
490 
491 	/*
492 	 * Rotate the vnode/cred, and clear the rotate flag so that we will
493 	 * send a rotate trigger if the new file fills.
494 	 */
495 	AUDIT_WORKER_LOCK();
496 	old_audit_cred = audit_cred;
497 	old_audit_vp = audit_vp;
498 	audit_cred = cred;
499 	audit_vp = vp;
500 	audit_size = vattr.va_size;
501 	audit_file_rotate_wait = 0;
502 	audit_enabled = (audit_vp != NULL);
503 	AUDIT_WORKER_UNLOCK();
504 
505 	/*
506 	 * If there was an old vnode/credential, close and free.
507 	 */
508 	if (old_audit_vp != NULL) {
509 		vn_close(old_audit_vp, AUDIT_CLOSE_FLAGS, old_audit_cred,
510 		    curthread);
511 		crfree(old_audit_cred);
512 	}
513 }
514 
515 void
516 audit_worker_init(void)
517 {
518 	int error;
519 
520 	AUDIT_WORKER_LOCK_INIT();
521 	error = kproc_create(audit_worker, NULL, &audit_thread, RFHIGHPID,
522 	    0, "audit");
523 	if (error)
524 		panic("audit_worker_init: kproc_create returned %d", error);
525 }
526