1/*-
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * Copyright (c) 2005 Robert N. M. Watson
6 * All rights reserved.
7 *
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
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 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND 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 THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR 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, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * Copyright (c) 1994 Christopher G. Demetriou
38 *
39 * Redistribution and use in source and binary forms, with or without
40 * modification, are permitted provided that the following conditions
41 * are met:
42 * 1. Redistributions of source code must retain the above copyright
43 * notice, this list of conditions and the following disclaimer.
44 * 2. Redistributions in binary form must reproduce the above copyright
45 * notice, this list of conditions and the following disclaimer in the
46 * documentation and/or other materials provided with the distribution.
47 * 3. All advertising materials mentioning features or use of this software
48 * must display the following acknowledgement:
49 * This product includes software developed by the University of
50 * California, Berkeley and its contributors.
51 * 4. Neither the name of the University nor the names of its contributors
52 * may be used to endorse or promote products derived from this software
53 * without specific prior written permission.
54 *
55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * SUCH DAMAGE.
66 *
67 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93
68 */
69
70#include <sys/cdefs.h>
71__FBSDID("$FreeBSD: head/sys/kern/kern_acct.c 167389 2007-03-09 23:29:31Z jhb $");
72
73#include "opt_mac.h"
74
75#include <sys/param.h>
76#include <sys/systm.h>
77#include <sys/acct.h>
78#include <sys/fcntl.h>
79#include <sys/kernel.h>
80#include <sys/kthread.h>
81#include <sys/lock.h>
82#include <sys/mount.h>
83#include <sys/mutex.h>
84#include <sys/namei.h>
85#include <sys/priv.h>
86#include <sys/proc.h>
87#include <sys/resourcevar.h>
88#include <sys/sched.h>
89#include <sys/sx.h>
90#include <sys/sysctl.h>
91#include <sys/sysent.h>
92#include <sys/syslog.h>
93#include <sys/sysproto.h>
94#include <sys/tty.h>
95#include <sys/vnode.h>
96
97#include <security/mac/mac_framework.h>
98
99/*
100 * The routines implemented in this file are described in:
101 * Leffler, et al.: The Design and Implementation of the 4.3BSD
102 * UNIX Operating System (Addison Welley, 1989)
103 * on pages 62-63.
104 *
105 * Arguably, to simplify accounting operations, this mechanism should
106 * be replaced by one in which an accounting log file (similar to /dev/klog)
107 * is read by a user process, etc. However, that has its own problems.
108 */
109
110/*
111 * Internal accounting functions.
112 * The former's operation is described in Leffler, et al., and the latter
113 * was provided by UCB with the 4.4BSD-Lite release
114 */
115static comp_t encode_comp_t(u_long, u_long);
116static void acctwatch(void);
117static void acct_thread(void *);
118static int acct_disable(struct thread *);
119
120/*
121 * Accounting vnode pointer, saved vnode pointer, and flags for each.
122 * acct_sx protects against changes to the active vnode and credentials
123 * while accounting records are being committed to disk.
124 */
125static int acct_configured;
126static int acct_suspended;
127static struct vnode *acct_vp;
128static struct ucred *acct_cred;
129static int acct_flags;
130static struct sx acct_sx;
131
132SX_SYSINIT(acct, &acct_sx, "acct_sx");
133
134/*
135 * State of the accounting kthread.
136 */
137static int acct_state;
138
139#define ACCT_RUNNING 1 /* Accounting kthread is running. */
140#define ACCT_EXITREQ 2 /* Accounting kthread should exit. */
141
142/*
143 * Values associated with enabling and disabling accounting
144 */
145static int acctsuspend = 2; /* stop accounting when < 2% free space left */
146SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
147 &acctsuspend, 0, "percentage of free disk space below which accounting stops");
148
149static int acctresume = 4; /* resume when free space risen to > 4% */
150SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
151 &acctresume, 0, "percentage of free disk space above which accounting resumes");
152
153static int acctchkfreq = 15; /* frequency (in seconds) to check space */
154
155static int
156sysctl_acct_chkfreq(SYSCTL_HANDLER_ARGS)
157{
158 int error, value;
159
160 /* Write out the old value. */
161 error = SYSCTL_OUT(req, &acctchkfreq, sizeof(int));
162 if (error || req->newptr == NULL)
163 return (error);
164
165 /* Read in and verify the new value. */
166 error = SYSCTL_IN(req, &value, sizeof(int));
167 if (error)
168 return (error);
169 if (value <= 0)
170 return (EINVAL);
171 acctchkfreq = value;
172 return (0);
173}
174SYSCTL_PROC(_kern, OID_AUTO, acct_chkfreq, CTLTYPE_INT|CTLFLAG_RW,
175 &acctchkfreq, 0, sysctl_acct_chkfreq, "I",
176 "frequency for checking the free space");
177
178SYSCTL_INT(_kern, OID_AUTO, acct_configured, CTLFLAG_RD, &acct_configured, 0,
179 "Accounting configured or not");
180
181SYSCTL_INT(_kern, OID_AUTO, acct_suspended, CTLFLAG_RD, &acct_suspended, 0,
182 "Accounting suspended or not");
183
184/*
185 * Accounting system call. Written based on the specification and previous
186 * implementation done by Mark Tinguely.
187 */
188int
189acct(struct thread *td, struct acct_args *uap)
190{
191 struct nameidata nd;
192 int error, flags, vfslocked;
193
194 error = priv_check(td, PRIV_ACCT);
195 if (error)
196 return (error);
197
198 /*
199 * If accounting is to be started to a file, open that file for
200 * appending and make sure it's a 'normal'.
201 */
202 if (uap->path != NULL) {
203 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1,
204 UIO_USERSPACE, uap->path, td);
205 flags = FWRITE | O_APPEND;
206 error = vn_open(&nd, &flags, 0, -1);
207 if (error)
208 return (error);
209 vfslocked = NDHASGIANT(&nd);
210 NDFREE(&nd, NDF_ONLY_PNBUF);
211#ifdef MAC
212 error = mac_check_system_acct(td->td_ucred, nd.ni_vp);
213 if (error) {
214 VOP_UNLOCK(nd.ni_vp, 0, td);
215 vn_close(nd.ni_vp, flags, td->td_ucred, td);
216 VFS_UNLOCK_GIANT(vfslocked);
217 return (error);
218 }
219#endif
220 VOP_UNLOCK(nd.ni_vp, 0, td);
221 if (nd.ni_vp->v_type != VREG) {
222 vn_close(nd.ni_vp, flags, td->td_ucred, td);
223 VFS_UNLOCK_GIANT(vfslocked);
224 return (EACCES);
225 }
226 VFS_UNLOCK_GIANT(vfslocked);
227#ifdef MAC
228 } else {
229 error = mac_check_system_acct(td->td_ucred, NULL);
230 if (error)
231 return (error);
232#endif
233 }
234
235 /*
236 * Disallow concurrent access to the accounting vnode while we swap
237 * it out, in order to prevent access after close.
238 */
239 sx_xlock(&acct_sx);
240
241 /*
242 * If accounting was previously enabled, kill the old space-watcher,
243 * close the file, and (if no new file was specified, leave). Reset
244 * the suspended state regardless of whether accounting remains
245 * enabled.
246 */
247 acct_suspended = 0;
248 if (acct_vp != NULL) {
249 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
250 error = acct_disable(td);
251 VFS_UNLOCK_GIANT(vfslocked);
252 }
253 if (uap->path == NULL) {
254 if (acct_state & ACCT_RUNNING) {
255 acct_state |= ACCT_EXITREQ;
256 wakeup(&acct_state);
257 }
258 sx_xunlock(&acct_sx);
259 return (error);
260 }
261
262 /*
263 * Save the new accounting file vnode, and schedule the new
264 * free space watcher.
265 */
266 acct_vp = nd.ni_vp;
267 acct_cred = crhold(td->td_ucred);
268 acct_flags = flags;
269 if (acct_state & ACCT_RUNNING)
270 acct_state &= ~ACCT_EXITREQ;
271 else {
272 /*
273 * Try to start up an accounting kthread. We may start more
274 * than one, but if so the extras will commit suicide as
275 * soon as they start up.
276 */
277 error = kthread_create(acct_thread, NULL, NULL, 0, 0,
278 "accounting");
279 if (error) {
280 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
281 (void) vn_close(acct_vp, acct_flags, acct_cred, td);
282 VFS_UNLOCK_GIANT(vfslocked);
283 crfree(acct_cred);
284 acct_configured = 0;
285 acct_vp = NULL;
286 acct_cred = NULL;
287 acct_flags = 0;
288 sx_xunlock(&acct_sx);
289 log(LOG_NOTICE, "Unable to start accounting thread\n");
290 return (error);
291 }
292 }
293 acct_configured = 1;
294 sx_xunlock(&acct_sx);
295 log(LOG_NOTICE, "Accounting enabled\n");
296 return (error);
297}
298
299/*
300 * Disable currently in-progress accounting by closing the vnode, dropping
301 * our reference to the credential, and clearing the vnode's flags.
302 */
303static int
304acct_disable(struct thread *td)
305{
306 int error;
307
308 sx_assert(&acct_sx, SX_XLOCKED);
309 error = vn_close(acct_vp, acct_flags, acct_cred, td);
310 crfree(acct_cred);
311 acct_configured = 0;
312 acct_vp = NULL;
313 acct_cred = NULL;
314 acct_flags = 0;
315 log(LOG_NOTICE, "Accounting disabled\n");
316 return (error);
317}
318
319/*
320 * Write out process accounting information, on process exit.
321 * Data to be written out is specified in Leffler, et al.
322 * and are enumerated below. (They're also noted in the system
323 * "acct.h" header file.)
324 */
325int
326acct_process(struct thread *td)
327{
328 struct acct acct;
329 struct timeval ut, st, tmp;
330 struct plimit *newlim, *oldlim;
331 struct proc *p;
332 struct rusage *r;
333 int t, ret, vfslocked;
334
335 /*
336 * Lockless check of accounting condition before doing the hard
337 * work.
338 */
339 if (acct_vp == NULL || acct_suspended)
340 return (0);
341
342 sx_slock(&acct_sx);
343
344 /*
345 * If accounting isn't enabled, don't bother. Have to check again
346 * once we own the lock in case we raced with disabling of accounting
347 * by another thread.
348 */
349 if (acct_vp == NULL || acct_suspended) {
350 sx_sunlock(&acct_sx);
351 return (0);
352 }
353
354 p = td->td_proc;
355
356 /*
357 * Get process accounting information.
358 */
359
360 PROC_LOCK(p);
361 /* (1) The name of the command that ran */
362 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
363
364 /* (2) The amount of user and system time that was used */
365 calcru(p, &ut, &st);
366 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
367 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
368
369 /* (3) The elapsed time the command ran (and its starting time) */
370 tmp = boottime;
371 timevaladd(&tmp, &p->p_stats->p_start);
372 acct.ac_btime = tmp.tv_sec;
373 microuptime(&tmp);
374 timevalsub(&tmp, &p->p_stats->p_start);
375 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
376
377 /* (4) The average amount of memory used */
378 r = &p->p_stats->p_ru;
379 tmp = ut;
380 timevaladd(&tmp, &st);
381 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
382 if (t)
383 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
384 else
385 acct.ac_mem = 0;
386
387 /* (5) The number of disk I/O operations done */
388 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
389
390 /* (6) The UID and GID of the process */
391 acct.ac_uid = p->p_ucred->cr_ruid;
392 acct.ac_gid = p->p_ucred->cr_rgid;
393
394 /* (7) The terminal from which the process was started */
395 SESS_LOCK(p->p_session);
396 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
397 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
398 else
399 acct.ac_tty = NODEV;
400 SESS_UNLOCK(p->p_session);
401
402 /* (8) The boolean flags that tell how the process terminated, etc. */
403 acct.ac_flag = p->p_acflag;
404 PROC_UNLOCK(p);
405
406 /*
407 * Eliminate any file size rlimit.
408 */
409 newlim = lim_alloc();
410 PROC_LOCK(p);
411 oldlim = p->p_limit;
412 lim_copy(newlim, oldlim);
413 newlim->pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
414 p->p_limit = newlim;
415 PROC_UNLOCK(p);
416 lim_free(oldlim);
417
418 /*
419 * Write the accounting information to the file.
420 */
421 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
422 VOP_LEASE(acct_vp, td, acct_cred, LEASE_WRITE);
423 ret = vn_rdwr(UIO_WRITE, acct_vp, (caddr_t)&acct, sizeof (acct),
424 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, acct_cred, NOCRED,
425 (int *)0, td);
426 VFS_UNLOCK_GIANT(vfslocked);
427 sx_sunlock(&acct_sx);
428 return (ret);
429}
430
431/*
432 * Encode_comp_t converts from ticks in seconds and microseconds
433 * to ticks in 1/AHZ seconds. The encoding is described in
434 * Leffler, et al., on page 63.
435 */
436
437#define MANTSIZE 13 /* 13 bit mantissa. */
438#define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
439#define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
440
441static comp_t
442encode_comp_t(u_long s, u_long us)
443{
444 int exp, rnd;
445
446 exp = 0;
447 rnd = 0;
448 s *= AHZ;
449 s += us / (1000000 / AHZ); /* Maximize precision. */
450
451 while (s > MAXFRACT) {
452 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
453 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
454 exp++;
455 }
456
457 /* If we need to round up, do it (and handle overflow correctly). */
458 if (rnd && (++s > MAXFRACT)) {
459 s >>= EXPSIZE;
460 exp++;
461 }
462
463 /* Clean it up and polish it off. */
464 exp <<= MANTSIZE; /* Shift the exponent into place */
465 exp += s; /* and add on the mantissa. */
466 return (exp);
467}
468
469/*
470 * Periodically check the filesystem to see if accounting
471 * should be turned on or off. Beware the case where the vnode
472 * has been vgone()'d out from underneath us, e.g. when the file
473 * system containing the accounting file has been forcibly unmounted.
474 */
475/* ARGSUSED */
476static void
477acctwatch(void)
478{
479 struct statfs sb;
480 int vfslocked;
481
482 sx_assert(&acct_sx, SX_XLOCKED);
483
484 /*
485 * If accounting was disabled before our kthread was scheduled,
486 * then acct_vp might be NULL. If so, just ask our kthread to
487 * exit and return.
488 */
489 if (acct_vp == NULL) {
490 acct_state |= ACCT_EXITREQ;
491 return;
492 }
493
494 /*
495 * If our vnode is no longer valid, tear it down and signal the
496 * accounting thread to die.
497 */
498 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
499 if (acct_vp->v_type == VBAD) {
500 (void) acct_disable(NULL);
501 VFS_UNLOCK_GIANT(vfslocked);
502 acct_state |= ACCT_EXITREQ;
503 return;
504 }
505
506 /*
507 * Stopping here is better than continuing, maybe it will be VBAD
508 * next time around.
509 */
510 if (VFS_STATFS(acct_vp->v_mount, &sb, curthread) < 0) {
511 VFS_UNLOCK_GIANT(vfslocked);
512 return;
513 }
514 VFS_UNLOCK_GIANT(vfslocked);
515 if (acct_suspended) {
516 if (sb.f_bavail > (int64_t)(acctresume * sb.f_blocks /
517 100)) {
518 acct_suspended = 0;
519 log(LOG_NOTICE, "Accounting resumed\n");
520 }
521 } else {
522 if (sb.f_bavail <= (int64_t)(acctsuspend * sb.f_blocks /
523 100)) {
524 acct_suspended = 1;
525 log(LOG_NOTICE, "Accounting suspended\n");
526 }
527 }
528}
529
530/*
531 * The main loop for the dedicated kernel thread that periodically calls
532 * acctwatch().
533 */
534static void
535acct_thread(void *dummy)
536{
537 u_char pri;
538
539 /* This is a low-priority kernel thread. */
540 pri = PRI_MAX_KERN;
541 mtx_lock_spin(&sched_lock);
542 sched_prio(curthread, pri);
543 mtx_unlock_spin(&sched_lock);
544
545 /* If another accounting kthread is already running, just die. */
546 sx_xlock(&acct_sx);
547 if (acct_state & ACCT_RUNNING) {
548 sx_xunlock(&acct_sx);
549 kthread_exit(0);
550 }
551 acct_state |= ACCT_RUNNING;
552
553 /* Loop until we are asked to exit. */
554 while (!(acct_state & ACCT_EXITREQ)) {
555
556 /* Perform our periodic checks. */
557 acctwatch();
558
559 /*
560 * We check this flag again before sleeping since the
561 * acctwatch() might have shut down accounting and asked us
562 * to exit.
563 */
564 if (!(acct_state & ACCT_EXITREQ)) {
565 sx_sleep(&acct_state, &acct_sx, 0, "-",
566 acctchkfreq * hz);
567 }
568 }
569
570 /*
571 * Acknowledge the exit request and shutdown. We clear both the
572 * exit request and running flags.
573 */
574 acct_state = 0;
575 sx_xunlock(&acct_sx);
576 kthread_exit(0);
577}
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * Copyright (c) 2005 Robert N. M. Watson
6 * All rights reserved.
7 *
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
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 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND 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 THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR 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, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * Copyright (c) 1994 Christopher G. Demetriou
38 *
39 * Redistribution and use in source and binary forms, with or without
40 * modification, are permitted provided that the following conditions
41 * are met:
42 * 1. Redistributions of source code must retain the above copyright
43 * notice, this list of conditions and the following disclaimer.
44 * 2. Redistributions in binary form must reproduce the above copyright
45 * notice, this list of conditions and the following disclaimer in the
46 * documentation and/or other materials provided with the distribution.
47 * 3. All advertising materials mentioning features or use of this software
48 * must display the following acknowledgement:
49 * This product includes software developed by the University of
50 * California, Berkeley and its contributors.
51 * 4. Neither the name of the University nor the names of its contributors
52 * may be used to endorse or promote products derived from this software
53 * without specific prior written permission.
54 *
55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * SUCH DAMAGE.
66 *
67 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93
68 */
69
70#include <sys/cdefs.h>
71__FBSDID("$FreeBSD: head/sys/kern/kern_acct.c 167389 2007-03-09 23:29:31Z jhb $");
72
73#include "opt_mac.h"
74
75#include <sys/param.h>
76#include <sys/systm.h>
77#include <sys/acct.h>
78#include <sys/fcntl.h>
79#include <sys/kernel.h>
80#include <sys/kthread.h>
81#include <sys/lock.h>
82#include <sys/mount.h>
83#include <sys/mutex.h>
84#include <sys/namei.h>
85#include <sys/priv.h>
86#include <sys/proc.h>
87#include <sys/resourcevar.h>
88#include <sys/sched.h>
89#include <sys/sx.h>
90#include <sys/sysctl.h>
91#include <sys/sysent.h>
92#include <sys/syslog.h>
93#include <sys/sysproto.h>
94#include <sys/tty.h>
95#include <sys/vnode.h>
96
97#include <security/mac/mac_framework.h>
98
99/*
100 * The routines implemented in this file are described in:
101 * Leffler, et al.: The Design and Implementation of the 4.3BSD
102 * UNIX Operating System (Addison Welley, 1989)
103 * on pages 62-63.
104 *
105 * Arguably, to simplify accounting operations, this mechanism should
106 * be replaced by one in which an accounting log file (similar to /dev/klog)
107 * is read by a user process, etc. However, that has its own problems.
108 */
109
110/*
111 * Internal accounting functions.
112 * The former's operation is described in Leffler, et al., and the latter
113 * was provided by UCB with the 4.4BSD-Lite release
114 */
115static comp_t encode_comp_t(u_long, u_long);
116static void acctwatch(void);
117static void acct_thread(void *);
118static int acct_disable(struct thread *);
119
120/*
121 * Accounting vnode pointer, saved vnode pointer, and flags for each.
122 * acct_sx protects against changes to the active vnode and credentials
123 * while accounting records are being committed to disk.
124 */
125static int acct_configured;
126static int acct_suspended;
127static struct vnode *acct_vp;
128static struct ucred *acct_cred;
129static int acct_flags;
130static struct sx acct_sx;
131
132SX_SYSINIT(acct, &acct_sx, "acct_sx");
133
134/*
135 * State of the accounting kthread.
136 */
137static int acct_state;
138
139#define ACCT_RUNNING 1 /* Accounting kthread is running. */
140#define ACCT_EXITREQ 2 /* Accounting kthread should exit. */
141
142/*
143 * Values associated with enabling and disabling accounting
144 */
145static int acctsuspend = 2; /* stop accounting when < 2% free space left */
146SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
147 &acctsuspend, 0, "percentage of free disk space below which accounting stops");
148
149static int acctresume = 4; /* resume when free space risen to > 4% */
150SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
151 &acctresume, 0, "percentage of free disk space above which accounting resumes");
152
153static int acctchkfreq = 15; /* frequency (in seconds) to check space */
154
155static int
156sysctl_acct_chkfreq(SYSCTL_HANDLER_ARGS)
157{
158 int error, value;
159
160 /* Write out the old value. */
161 error = SYSCTL_OUT(req, &acctchkfreq, sizeof(int));
162 if (error || req->newptr == NULL)
163 return (error);
164
165 /* Read in and verify the new value. */
166 error = SYSCTL_IN(req, &value, sizeof(int));
167 if (error)
168 return (error);
169 if (value <= 0)
170 return (EINVAL);
171 acctchkfreq = value;
172 return (0);
173}
174SYSCTL_PROC(_kern, OID_AUTO, acct_chkfreq, CTLTYPE_INT|CTLFLAG_RW,
175 &acctchkfreq, 0, sysctl_acct_chkfreq, "I",
176 "frequency for checking the free space");
177
178SYSCTL_INT(_kern, OID_AUTO, acct_configured, CTLFLAG_RD, &acct_configured, 0,
179 "Accounting configured or not");
180
181SYSCTL_INT(_kern, OID_AUTO, acct_suspended, CTLFLAG_RD, &acct_suspended, 0,
182 "Accounting suspended or not");
183
184/*
185 * Accounting system call. Written based on the specification and previous
186 * implementation done by Mark Tinguely.
187 */
188int
189acct(struct thread *td, struct acct_args *uap)
190{
191 struct nameidata nd;
192 int error, flags, vfslocked;
193
194 error = priv_check(td, PRIV_ACCT);
195 if (error)
196 return (error);
197
198 /*
199 * If accounting is to be started to a file, open that file for
200 * appending and make sure it's a 'normal'.
201 */
202 if (uap->path != NULL) {
203 NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE | AUDITVNODE1,
204 UIO_USERSPACE, uap->path, td);
205 flags = FWRITE | O_APPEND;
206 error = vn_open(&nd, &flags, 0, -1);
207 if (error)
208 return (error);
209 vfslocked = NDHASGIANT(&nd);
210 NDFREE(&nd, NDF_ONLY_PNBUF);
211#ifdef MAC
212 error = mac_check_system_acct(td->td_ucred, nd.ni_vp);
213 if (error) {
214 VOP_UNLOCK(nd.ni_vp, 0, td);
215 vn_close(nd.ni_vp, flags, td->td_ucred, td);
216 VFS_UNLOCK_GIANT(vfslocked);
217 return (error);
218 }
219#endif
220 VOP_UNLOCK(nd.ni_vp, 0, td);
221 if (nd.ni_vp->v_type != VREG) {
222 vn_close(nd.ni_vp, flags, td->td_ucred, td);
223 VFS_UNLOCK_GIANT(vfslocked);
224 return (EACCES);
225 }
226 VFS_UNLOCK_GIANT(vfslocked);
227#ifdef MAC
228 } else {
229 error = mac_check_system_acct(td->td_ucred, NULL);
230 if (error)
231 return (error);
232#endif
233 }
234
235 /*
236 * Disallow concurrent access to the accounting vnode while we swap
237 * it out, in order to prevent access after close.
238 */
239 sx_xlock(&acct_sx);
240
241 /*
242 * If accounting was previously enabled, kill the old space-watcher,
243 * close the file, and (if no new file was specified, leave). Reset
244 * the suspended state regardless of whether accounting remains
245 * enabled.
246 */
247 acct_suspended = 0;
248 if (acct_vp != NULL) {
249 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
250 error = acct_disable(td);
251 VFS_UNLOCK_GIANT(vfslocked);
252 }
253 if (uap->path == NULL) {
254 if (acct_state & ACCT_RUNNING) {
255 acct_state |= ACCT_EXITREQ;
256 wakeup(&acct_state);
257 }
258 sx_xunlock(&acct_sx);
259 return (error);
260 }
261
262 /*
263 * Save the new accounting file vnode, and schedule the new
264 * free space watcher.
265 */
266 acct_vp = nd.ni_vp;
267 acct_cred = crhold(td->td_ucred);
268 acct_flags = flags;
269 if (acct_state & ACCT_RUNNING)
270 acct_state &= ~ACCT_EXITREQ;
271 else {
272 /*
273 * Try to start up an accounting kthread. We may start more
274 * than one, but if so the extras will commit suicide as
275 * soon as they start up.
276 */
277 error = kthread_create(acct_thread, NULL, NULL, 0, 0,
278 "accounting");
279 if (error) {
280 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
281 (void) vn_close(acct_vp, acct_flags, acct_cred, td);
282 VFS_UNLOCK_GIANT(vfslocked);
283 crfree(acct_cred);
284 acct_configured = 0;
285 acct_vp = NULL;
286 acct_cred = NULL;
287 acct_flags = 0;
288 sx_xunlock(&acct_sx);
289 log(LOG_NOTICE, "Unable to start accounting thread\n");
290 return (error);
291 }
292 }
293 acct_configured = 1;
294 sx_xunlock(&acct_sx);
295 log(LOG_NOTICE, "Accounting enabled\n");
296 return (error);
297}
298
299/*
300 * Disable currently in-progress accounting by closing the vnode, dropping
301 * our reference to the credential, and clearing the vnode's flags.
302 */
303static int
304acct_disable(struct thread *td)
305{
306 int error;
307
308 sx_assert(&acct_sx, SX_XLOCKED);
309 error = vn_close(acct_vp, acct_flags, acct_cred, td);
310 crfree(acct_cred);
311 acct_configured = 0;
312 acct_vp = NULL;
313 acct_cred = NULL;
314 acct_flags = 0;
315 log(LOG_NOTICE, "Accounting disabled\n");
316 return (error);
317}
318
319/*
320 * Write out process accounting information, on process exit.
321 * Data to be written out is specified in Leffler, et al.
322 * and are enumerated below. (They're also noted in the system
323 * "acct.h" header file.)
324 */
325int
326acct_process(struct thread *td)
327{
328 struct acct acct;
329 struct timeval ut, st, tmp;
330 struct plimit *newlim, *oldlim;
331 struct proc *p;
332 struct rusage *r;
333 int t, ret, vfslocked;
334
335 /*
336 * Lockless check of accounting condition before doing the hard
337 * work.
338 */
339 if (acct_vp == NULL || acct_suspended)
340 return (0);
341
342 sx_slock(&acct_sx);
343
344 /*
345 * If accounting isn't enabled, don't bother. Have to check again
346 * once we own the lock in case we raced with disabling of accounting
347 * by another thread.
348 */
349 if (acct_vp == NULL || acct_suspended) {
350 sx_sunlock(&acct_sx);
351 return (0);
352 }
353
354 p = td->td_proc;
355
356 /*
357 * Get process accounting information.
358 */
359
360 PROC_LOCK(p);
361 /* (1) The name of the command that ran */
362 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
363
364 /* (2) The amount of user and system time that was used */
365 calcru(p, &ut, &st);
366 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
367 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
368
369 /* (3) The elapsed time the command ran (and its starting time) */
370 tmp = boottime;
371 timevaladd(&tmp, &p->p_stats->p_start);
372 acct.ac_btime = tmp.tv_sec;
373 microuptime(&tmp);
374 timevalsub(&tmp, &p->p_stats->p_start);
375 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
376
377 /* (4) The average amount of memory used */
378 r = &p->p_stats->p_ru;
379 tmp = ut;
380 timevaladd(&tmp, &st);
381 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
382 if (t)
383 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
384 else
385 acct.ac_mem = 0;
386
387 /* (5) The number of disk I/O operations done */
388 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
389
390 /* (6) The UID and GID of the process */
391 acct.ac_uid = p->p_ucred->cr_ruid;
392 acct.ac_gid = p->p_ucred->cr_rgid;
393
394 /* (7) The terminal from which the process was started */
395 SESS_LOCK(p->p_session);
396 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
397 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
398 else
399 acct.ac_tty = NODEV;
400 SESS_UNLOCK(p->p_session);
401
402 /* (8) The boolean flags that tell how the process terminated, etc. */
403 acct.ac_flag = p->p_acflag;
404 PROC_UNLOCK(p);
405
406 /*
407 * Eliminate any file size rlimit.
408 */
409 newlim = lim_alloc();
410 PROC_LOCK(p);
411 oldlim = p->p_limit;
412 lim_copy(newlim, oldlim);
413 newlim->pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
414 p->p_limit = newlim;
415 PROC_UNLOCK(p);
416 lim_free(oldlim);
417
418 /*
419 * Write the accounting information to the file.
420 */
421 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
422 VOP_LEASE(acct_vp, td, acct_cred, LEASE_WRITE);
423 ret = vn_rdwr(UIO_WRITE, acct_vp, (caddr_t)&acct, sizeof (acct),
424 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, acct_cred, NOCRED,
425 (int *)0, td);
426 VFS_UNLOCK_GIANT(vfslocked);
427 sx_sunlock(&acct_sx);
428 return (ret);
429}
430
431/*
432 * Encode_comp_t converts from ticks in seconds and microseconds
433 * to ticks in 1/AHZ seconds. The encoding is described in
434 * Leffler, et al., on page 63.
435 */
436
437#define MANTSIZE 13 /* 13 bit mantissa. */
438#define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
439#define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
440
441static comp_t
442encode_comp_t(u_long s, u_long us)
443{
444 int exp, rnd;
445
446 exp = 0;
447 rnd = 0;
448 s *= AHZ;
449 s += us / (1000000 / AHZ); /* Maximize precision. */
450
451 while (s > MAXFRACT) {
452 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
453 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
454 exp++;
455 }
456
457 /* If we need to round up, do it (and handle overflow correctly). */
458 if (rnd && (++s > MAXFRACT)) {
459 s >>= EXPSIZE;
460 exp++;
461 }
462
463 /* Clean it up and polish it off. */
464 exp <<= MANTSIZE; /* Shift the exponent into place */
465 exp += s; /* and add on the mantissa. */
466 return (exp);
467}
468
469/*
470 * Periodically check the filesystem to see if accounting
471 * should be turned on or off. Beware the case where the vnode
472 * has been vgone()'d out from underneath us, e.g. when the file
473 * system containing the accounting file has been forcibly unmounted.
474 */
475/* ARGSUSED */
476static void
477acctwatch(void)
478{
479 struct statfs sb;
480 int vfslocked;
481
482 sx_assert(&acct_sx, SX_XLOCKED);
483
484 /*
485 * If accounting was disabled before our kthread was scheduled,
486 * then acct_vp might be NULL. If so, just ask our kthread to
487 * exit and return.
488 */
489 if (acct_vp == NULL) {
490 acct_state |= ACCT_EXITREQ;
491 return;
492 }
493
494 /*
495 * If our vnode is no longer valid, tear it down and signal the
496 * accounting thread to die.
497 */
498 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
499 if (acct_vp->v_type == VBAD) {
500 (void) acct_disable(NULL);
501 VFS_UNLOCK_GIANT(vfslocked);
502 acct_state |= ACCT_EXITREQ;
503 return;
504 }
505
506 /*
507 * Stopping here is better than continuing, maybe it will be VBAD
508 * next time around.
509 */
510 if (VFS_STATFS(acct_vp->v_mount, &sb, curthread) < 0) {
511 VFS_UNLOCK_GIANT(vfslocked);
512 return;
513 }
514 VFS_UNLOCK_GIANT(vfslocked);
515 if (acct_suspended) {
516 if (sb.f_bavail > (int64_t)(acctresume * sb.f_blocks /
517 100)) {
518 acct_suspended = 0;
519 log(LOG_NOTICE, "Accounting resumed\n");
520 }
521 } else {
522 if (sb.f_bavail <= (int64_t)(acctsuspend * sb.f_blocks /
523 100)) {
524 acct_suspended = 1;
525 log(LOG_NOTICE, "Accounting suspended\n");
526 }
527 }
528}
529
530/*
531 * The main loop for the dedicated kernel thread that periodically calls
532 * acctwatch().
533 */
534static void
535acct_thread(void *dummy)
536{
537 u_char pri;
538
539 /* This is a low-priority kernel thread. */
540 pri = PRI_MAX_KERN;
541 mtx_lock_spin(&sched_lock);
542 sched_prio(curthread, pri);
543 mtx_unlock_spin(&sched_lock);
544
545 /* If another accounting kthread is already running, just die. */
546 sx_xlock(&acct_sx);
547 if (acct_state & ACCT_RUNNING) {
548 sx_xunlock(&acct_sx);
549 kthread_exit(0);
550 }
551 acct_state |= ACCT_RUNNING;
552
553 /* Loop until we are asked to exit. */
554 while (!(acct_state & ACCT_EXITREQ)) {
555
556 /* Perform our periodic checks. */
557 acctwatch();
558
559 /*
560 * We check this flag again before sleeping since the
561 * acctwatch() might have shut down accounting and asked us
562 * to exit.
563 */
564 if (!(acct_state & ACCT_EXITREQ)) {
565 sx_sleep(&acct_state, &acct_sx, 0, "-",
566 acctchkfreq * hz);
567 }
568 }
569
570 /*
571 * Acknowledge the exit request and shutdown. We clear both the
572 * exit request and running flags.
573 */
574 acct_state = 0;
575 sx_xunlock(&acct_sx);
576 kthread_exit(0);
577}