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