xref: /freebsd/sys/security/audit/audit_worker.c (revision f4b37ed0f8b307b1f3f0f630ca725d68f1dff30d)
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 /*
75  * Worker thread that will schedule disk I/O, etc.
76  */
77 static struct proc		*audit_thread;
78 
79 /*
80  * audit_cred and audit_vp are the stored credential and vnode to use for
81  * active audit trail.  They are protected by the audit worker lock, which
82  * will be held across all I/O and all rotation to prevent them from being
83  * replaced (rotated) while in use.  The audit_file_rotate_wait flag is set
84  * when the kernel has delivered a trigger to auditd to rotate the trail, and
85  * is cleared when the next rotation takes place.  It is also protected by
86  * the audit worker lock.
87  */
88 static int		 audit_file_rotate_wait;
89 static struct ucred	*audit_cred;
90 static struct vnode	*audit_vp;
91 static off_t		 audit_size;
92 static struct sx	 audit_worker_lock;
93 
94 #define	AUDIT_WORKER_LOCK_INIT()	sx_init(&audit_worker_lock, \
95 					    "audit_worker_lock");
96 #define	AUDIT_WORKER_LOCK_ASSERT()	sx_assert(&audit_worker_lock, \
97 					    SA_XLOCKED)
98 #define	AUDIT_WORKER_LOCK()		sx_xlock(&audit_worker_lock)
99 #define	AUDIT_WORKER_UNLOCK()		sx_xunlock(&audit_worker_lock)
100 
101 /*
102  * Write an audit record to a file, performed as the last stage after both
103  * preselection and BSM conversion.  Both space management and write failures
104  * are handled in this function.
105  *
106  * No attempt is made to deal with possible failure to deliver a trigger to
107  * the audit daemon, since the message is asynchronous anyway.
108  */
109 static void
110 audit_record_write(struct vnode *vp, struct ucred *cred, void *data,
111     size_t len)
112 {
113 	static struct timeval last_lowspace_trigger;
114 	static struct timeval last_fail;
115 	static int cur_lowspace_trigger;
116 	struct statfs *mnt_stat;
117 	int error;
118 	static int cur_fail;
119 	long temp;
120 
121 	AUDIT_WORKER_LOCK_ASSERT();
122 
123 	if (vp == NULL)
124 		return;
125 
126 	mnt_stat = &vp->v_mount->mnt_stat;
127 
128 	/*
129 	 * First, gather statistics on the audit log file and file system so
130 	 * that we know how we're doing on space.  Consider failure of these
131 	 * operations to indicate a future inability to write to the file.
132 	 */
133 	error = VFS_STATFS(vp->v_mount, mnt_stat);
134 	if (error)
135 		goto fail;
136 
137 	/*
138 	 * We handle four different space-related limits:
139 	 *
140 	 * - A fixed (hard) limit on the minimum free blocks we require on
141 	 *   the file system, and results in record loss, a trigger, and
142 	 *   possible fail stop due to violating invariants.
143 	 *
144 	 * - An administrative (soft) limit, which when fallen below, results
145 	 *   in the kernel notifying the audit daemon of low space.
146 	 *
147 	 * - An audit trail size limit, which when gone above, results in the
148 	 *   kernel notifying the audit daemon that rotation is desired.
149 	 *
150 	 * - The total depth of the kernel audit record exceeding free space,
151 	 *   which can lead to possible fail stop (with drain), in order to
152 	 *   prevent violating invariants.  Failure here doesn't halt
153 	 *   immediately, but prevents new records from being generated.
154 	 *
155 	 * Possibly, the last of these should be handled differently, always
156 	 * allowing a full queue to be lost, rather than trying to prevent
157 	 * loss.
158 	 *
159 	 * First, handle the hard limit, which generates a trigger and may
160 	 * fail stop.  This is handled in the same manner as ENOSPC from
161 	 * VOP_WRITE, and results in record loss.
162 	 */
163 	if (mnt_stat->f_bfree < AUDIT_HARD_LIMIT_FREE_BLOCKS) {
164 		error = ENOSPC;
165 		goto fail_enospc;
166 	}
167 
168 	/*
169 	 * Second, handle falling below the soft limit, if defined; we send
170 	 * the daemon a trigger and continue processing the record.  Triggers
171 	 * are limited to 1/sec.
172 	 */
173 	if (audit_qctrl.aq_minfree != 0) {
174 		temp = mnt_stat->f_blocks / (100 / audit_qctrl.aq_minfree);
175 		if (mnt_stat->f_bfree < temp) {
176 			if (ppsratecheck(&last_lowspace_trigger,
177 			    &cur_lowspace_trigger, 1)) {
178 				(void)audit_send_trigger(
179 				    AUDIT_TRIGGER_LOW_SPACE);
180 				printf("Warning: disk space low (< %d%% free) "
181 				    "on audit log file-system\n",
182 				    audit_qctrl.aq_minfree);
183 			}
184 		}
185 	}
186 
187 	/*
188 	 * If the current file is getting full, generate a rotation trigger
189 	 * to the daemon.  This is only approximate, which is fine as more
190 	 * records may be generated before the daemon rotates the file.
191 	 */
192 	if (audit_fstat.af_filesz != 0 &&
193 	    audit_size >= audit_fstat.af_filesz * (audit_file_rotate_wait + 1)) {
194 		AUDIT_WORKER_LOCK_ASSERT();
195 
196 		audit_file_rotate_wait++;
197 		(void)audit_send_trigger(AUDIT_TRIGGER_ROTATE_KERNEL);
198 	}
199 
200 	/*
201 	 * If the estimated amount of audit data in the audit event queue
202 	 * (plus records allocated but not yet queued) has reached the amount
203 	 * of free space on the disk, then we need to go into an audit fail
204 	 * stop state, in which we do not permit the allocation/committing of
205 	 * any new audit records.  We continue to process records but don't
206 	 * allow any activities that might generate new records.  In the
207 	 * future, we might want to detect when space is available again and
208 	 * allow operation to continue, but this behavior is sufficient to
209 	 * meet fail stop requirements in CAPP.
210 	 */
211 	if (audit_fail_stop) {
212 		if ((unsigned long)((audit_q_len + audit_pre_q_len + 1) *
213 		    MAX_AUDIT_RECORD_SIZE) / mnt_stat->f_bsize >=
214 		    (unsigned long)(mnt_stat->f_bfree)) {
215 			if (ppsratecheck(&last_fail, &cur_fail, 1))
216 				printf("audit_record_write: free space "
217 				    "below size of audit queue, failing "
218 				    "stop\n");
219 			audit_in_failure = 1;
220 		} else if (audit_in_failure) {
221 			/*
222 			 * Note: if we want to handle recovery, this is the
223 			 * spot to do it: unset audit_in_failure, and issue a
224 			 * wakeup on the cv.
225 			 */
226 		}
227 	}
228 
229 	error = vn_rdwr(UIO_WRITE, vp, data, len, (off_t)0, UIO_SYSSPACE,
230 	    IO_APPEND|IO_UNIT, cred, NULL, NULL, curthread);
231 	if (error == ENOSPC)
232 		goto fail_enospc;
233 	else if (error)
234 		goto fail;
235 	AUDIT_WORKER_LOCK_ASSERT();
236 	audit_size += len;
237 
238 	/*
239 	 * Catch completion of a queue drain here; if we're draining and the
240 	 * queue is now empty, fail stop.  That audit_fail_stop is implicitly
241 	 * true, since audit_in_failure can only be set of audit_fail_stop is
242 	 * set.
243 	 *
244 	 * Note: if we handle recovery from audit_in_failure, then we need to
245 	 * make panic here conditional.
246 	 */
247 	if (audit_in_failure) {
248 		if (audit_q_len == 0 && audit_pre_q_len == 0) {
249 			VOP_LOCK(vp, LK_EXCLUSIVE | LK_RETRY);
250 			(void)VOP_FSYNC(vp, MNT_WAIT, curthread);
251 			VOP_UNLOCK(vp, 0);
252 			panic("Audit store overflow; record queue drained.");
253 		}
254 	}
255 
256 	return;
257 
258 fail_enospc:
259 	/*
260 	 * ENOSPC is considered a special case with respect to failures, as
261 	 * this can reflect either our preemptive detection of insufficient
262 	 * space, or ENOSPC returned by the vnode write call.
263 	 */
264 	if (audit_fail_stop) {
265 		VOP_LOCK(vp, LK_EXCLUSIVE | LK_RETRY);
266 		(void)VOP_FSYNC(vp, MNT_WAIT, curthread);
267 		VOP_UNLOCK(vp, 0);
268 		panic("Audit log space exhausted and fail-stop set.");
269 	}
270 	(void)audit_send_trigger(AUDIT_TRIGGER_NO_SPACE);
271 	audit_suspended = 1;
272 
273 	/* FALLTHROUGH */
274 fail:
275 	/*
276 	 * We have failed to write to the file, so the current record is
277 	 * lost, which may require an immediate system halt.
278 	 */
279 	if (audit_panic_on_write_fail) {
280 		VOP_LOCK(vp, LK_EXCLUSIVE | LK_RETRY);
281 		(void)VOP_FSYNC(vp, MNT_WAIT, curthread);
282 		VOP_UNLOCK(vp, 0);
283 		panic("audit_worker: write error %d\n", error);
284 	} else if (ppsratecheck(&last_fail, &cur_fail, 1))
285 		printf("audit_worker: write error %d\n", error);
286 }
287 
288 /*
289  * Given a kernel audit record, process as required.  Kernel audit records
290  * are converted to one, or possibly two, BSM records, depending on whether
291  * there is a user audit record present also.  Kernel records need be
292  * converted to BSM before they can be written out.  Both types will be
293  * written to disk, and audit pipes.
294  */
295 static void
296 audit_worker_process_record(struct kaudit_record *ar)
297 {
298 	struct au_record *bsm;
299 	au_class_t class;
300 	au_event_t event;
301 	au_id_t auid;
302 	int error, sorf;
303 	int locked;
304 
305 	/*
306 	 * We hold the audit worker lock over both writes, if there are two,
307 	 * so that the two records won't be split across a rotation and end
308 	 * up in two different trail files.
309 	 */
310 	if (((ar->k_ar_commit & AR_COMMIT_USER) &&
311 	    (ar->k_ar_commit & AR_PRESELECT_USER_TRAIL)) ||
312 	    (ar->k_ar_commit & AR_PRESELECT_TRAIL)) {
313 		AUDIT_WORKER_LOCK();
314 		locked = 1;
315 	} else
316 		locked = 0;
317 
318 	/*
319 	 * First, handle the user record, if any: commit to the system trail
320 	 * and audit pipes as selected.
321 	 */
322 	if ((ar->k_ar_commit & AR_COMMIT_USER) &&
323 	    (ar->k_ar_commit & AR_PRESELECT_USER_TRAIL)) {
324 		AUDIT_WORKER_LOCK_ASSERT();
325 		audit_record_write(audit_vp, audit_cred, ar->k_udata,
326 		    ar->k_ulen);
327 	}
328 
329 	if ((ar->k_ar_commit & AR_COMMIT_USER) &&
330 	    (ar->k_ar_commit & AR_PRESELECT_USER_PIPE))
331 		audit_pipe_submit_user(ar->k_udata, ar->k_ulen);
332 
333 	if (!(ar->k_ar_commit & AR_COMMIT_KERNEL) ||
334 	    ((ar->k_ar_commit & AR_PRESELECT_PIPE) == 0 &&
335 	    (ar->k_ar_commit & AR_PRESELECT_TRAIL) == 0))
336 		goto out;
337 
338 	auid = ar->k_ar.ar_subj_auid;
339 	event = ar->k_ar.ar_event;
340 	class = au_event_class(event);
341 	if (ar->k_ar.ar_errno == 0)
342 		sorf = AU_PRS_SUCCESS;
343 	else
344 		sorf = AU_PRS_FAILURE;
345 
346 	error = kaudit_to_bsm(ar, &bsm);
347 	switch (error) {
348 	case BSM_NOAUDIT:
349 		goto out;
350 
351 	case BSM_FAILURE:
352 		printf("audit_worker_process_record: BSM_FAILURE\n");
353 		goto out;
354 
355 	case BSM_SUCCESS:
356 		break;
357 
358 	default:
359 		panic("kaudit_to_bsm returned %d", error);
360 	}
361 
362 	if (ar->k_ar_commit & AR_PRESELECT_TRAIL) {
363 		AUDIT_WORKER_LOCK_ASSERT();
364 		audit_record_write(audit_vp, audit_cred, bsm->data, bsm->len);
365 	}
366 
367 	if (ar->k_ar_commit & AR_PRESELECT_PIPE)
368 		audit_pipe_submit(auid, event, class, sorf,
369 		    ar->k_ar_commit & AR_PRESELECT_TRAIL, bsm->data,
370 		    bsm->len);
371 
372 	kau_free(bsm);
373 out:
374 	if (locked)
375 		AUDIT_WORKER_UNLOCK();
376 }
377 
378 /*
379  * The audit_worker thread is responsible for watching the event queue,
380  * dequeueing records, converting them to BSM format, and committing them to
381  * disk.  In order to minimize lock thrashing, records are dequeued in sets
382  * to a thread-local work queue.
383  *
384  * Note: this means that the effect bound on the size of the pending record
385  * queue is 2x the length of the global queue.
386  */
387 static void
388 audit_worker(void *arg)
389 {
390 	struct kaudit_queue ar_worklist;
391 	struct kaudit_record *ar;
392 	int lowater_signal;
393 
394 	TAILQ_INIT(&ar_worklist);
395 	mtx_lock(&audit_mtx);
396 	while (1) {
397 		mtx_assert(&audit_mtx, MA_OWNED);
398 
399 		/*
400 		 * Wait for a record.
401 		 */
402 		while (TAILQ_EMPTY(&audit_q))
403 			cv_wait(&audit_worker_cv, &audit_mtx);
404 
405 		/*
406 		 * If there are records in the global audit record queue,
407 		 * transfer them to a thread-local queue and process them
408 		 * one by one.  If we cross the low watermark threshold,
409 		 * signal any waiting processes that they may wake up and
410 		 * continue generating records.
411 		 */
412 		lowater_signal = 0;
413 		while ((ar = TAILQ_FIRST(&audit_q))) {
414 			TAILQ_REMOVE(&audit_q, ar, k_q);
415 			audit_q_len--;
416 			if (audit_q_len == audit_qctrl.aq_lowater)
417 				lowater_signal++;
418 			TAILQ_INSERT_TAIL(&ar_worklist, ar, k_q);
419 		}
420 		if (lowater_signal)
421 			cv_broadcast(&audit_watermark_cv);
422 
423 		mtx_unlock(&audit_mtx);
424 		while ((ar = TAILQ_FIRST(&ar_worklist))) {
425 			TAILQ_REMOVE(&ar_worklist, ar, k_q);
426 			audit_worker_process_record(ar);
427 			audit_free(ar);
428 		}
429 		mtx_lock(&audit_mtx);
430 	}
431 }
432 
433 /*
434  * audit_rotate_vnode() is called by a user or kernel thread to configure or
435  * de-configure auditing on a vnode.  The arguments are the replacement
436  * credential (referenced) and vnode (referenced and opened) to substitute
437  * for the current credential and vnode, if any.  If either is set to NULL,
438  * both should be NULL, and this is used to indicate that audit is being
439  * disabled.  Any previous cred/vnode will be closed and freed.  We re-enable
440  * generating rotation requests to auditd.
441  */
442 void
443 audit_rotate_vnode(struct ucred *cred, struct vnode *vp)
444 {
445 	struct ucred *old_audit_cred;
446 	struct vnode *old_audit_vp;
447 	struct vattr vattr;
448 
449 	KASSERT((cred != NULL && vp != NULL) || (cred == NULL && vp == NULL),
450 	    ("audit_rotate_vnode: cred %p vp %p", cred, vp));
451 
452 	if (vp != NULL) {
453 		vn_lock(vp, LK_SHARED | LK_RETRY);
454 		if (VOP_GETATTR(vp, &vattr, cred) != 0)
455 			vattr.va_size = 0;
456 		VOP_UNLOCK(vp, 0);
457 	} else {
458 		vattr.va_size = 0;
459 	}
460 
461 	/*
462 	 * Rotate the vnode/cred, and clear the rotate flag so that we will
463 	 * send a rotate trigger if the new file fills.
464 	 */
465 	AUDIT_WORKER_LOCK();
466 	old_audit_cred = audit_cred;
467 	old_audit_vp = audit_vp;
468 	audit_cred = cred;
469 	audit_vp = vp;
470 	audit_size = vattr.va_size;
471 	audit_file_rotate_wait = 0;
472 	audit_enabled = (audit_vp != NULL);
473 	AUDIT_WORKER_UNLOCK();
474 
475 	/*
476 	 * If there was an old vnode/credential, close and free.
477 	 */
478 	if (old_audit_vp != NULL) {
479 		vn_close(old_audit_vp, AUDIT_CLOSE_FLAGS, old_audit_cred,
480 		    curthread);
481 		crfree(old_audit_cred);
482 	}
483 }
484 
485 void
486 audit_worker_init(void)
487 {
488 	int error;
489 
490 	AUDIT_WORKER_LOCK_INIT();
491 	error = kproc_create(audit_worker, NULL, &audit_thread, RFHIGHPID,
492 	    0, "audit");
493 	if (error)
494 		panic("audit_worker_init: kproc_create returned %d", error);
495 }
496