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
audit_worker_sync_vp(struct vnode * vp,struct mount * mp,const char * fmt,...)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
audit_record_write(struct vnode * vp,struct ucred * cred,void * data,size_t len)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
audit_worker_process_record(struct kaudit_record * ar)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
audit_worker(void * arg)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
audit_rotate_vnode(struct ucred * cred,struct vnode * vp)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
audit_worker_init(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