1/* $NetBSD: kern_acct.c,v 1.94 2013/10/19 21:01:39 mrg Exp $ */
2
3/*-
4 * Copyright (c) 1982, 1986, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)kern_acct.c 8.8 (Berkeley) 5/14/95
37 */
38
39/*-
40 * Copyright (c) 1994 Christopher G. Demetriou
41 *
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
44 * are met:
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Berkeley and its contributors.
54 * 4. Neither the name of the University nor the names of its contributors
55 * may be used to endorse or promote products derived from this software
56 * without specific prior written permission.
57 *
58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * SUCH DAMAGE.
69 *
70 * @(#)kern_acct.c 8.8 (Berkeley) 5/14/95
71 */
72
73#include <sys/cdefs.h>
74__KERNEL_RCSID(0, "$NetBSD: kern_acct.c,v 1.94 2013/10/19 21:01:39 mrg Exp $");
75
76#include <sys/param.h>
77#include <sys/systm.h>
78#include <sys/proc.h>
79#include <sys/mount.h>
80#include <sys/vnode.h>
81#include <sys/file.h>
82#include <sys/syslog.h>
83#include <sys/kernel.h>
84#include <sys/kthread.h>
85#include <sys/kmem.h>
86#include <sys/namei.h>
87#include <sys/errno.h>
88#include <sys/acct.h>
89#include <sys/resourcevar.h>
90#include <sys/ioctl.h>
91#include <sys/tty.h>
92#include <sys/kauth.h>
93
94#include <sys/syscallargs.h>
95
96/*
97 * The routines implemented in this file are described in:
98 * Leffler, et al.: The Design and Implementation of the 4.3BSD
99 * UNIX Operating System (Addison Welley, 1989)
100 * on pages 62-63.
101 *
102 * Arguably, to simplify accounting operations, this mechanism should
103 * be replaced by one in which an accounting log file (similar to /dev/klog)
104 * is read by a user process, etc. However, that has its own problems.
105 */
106
107/*
108 * Lock to serialize system calls and kernel threads.
109 */
110krwlock_t acct_lock;
111
112/*
113 * The global accounting state and related data. Gain the mutex before
114 * accessing these variables.
115 */
116static enum {
117 ACCT_STOP,
118 ACCT_ACTIVE,
119 ACCT_SUSPENDED
120} acct_state; /* The current accounting state. */
121static struct vnode *acct_vp; /* Accounting vnode pointer. */
122static kauth_cred_t acct_cred; /* Credential of accounting file
123 owner (i.e root). Used when
124 accounting file i/o. */
125static struct lwp *acct_dkwatcher; /* Free disk space checker. */
126
127/*
128 * Values associated with enabling and disabling accounting
129 */
130int acctsuspend = 2; /* stop accounting when < 2% free space left */
131int acctresume = 4; /* resume when free space risen to > 4% */
132int acctchkfreq = 15; /* frequency (in seconds) to check space */
133
134/*
135 * Encode_comp_t converts from ticks in seconds and microseconds
136 * to ticks in 1/AHZ seconds. The encoding is described in
137 * Leffler, et al., on page 63.
138 */
139
140#define MANTSIZE 13 /* 13 bit mantissa. */
141#define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
142#define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
143
144static comp_t
145encode_comp_t(u_long s, u_long us)
146{
147 int exp, rnd;
148
149 exp = 0;
150 rnd = 0;
151 s *= AHZ;
152 s += us / (1000000 / AHZ); /* Maximize precision. */
153
154 while (s > MAXFRACT) {
155 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
156 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
157 exp++;
158 }
159
160 /* If we need to round up, do it (and handle overflow correctly). */
161 if (rnd && (++s > MAXFRACT)) {
162 s >>= EXPSIZE;
163 exp++;
164 }
165
166 /* Clean it up and polish it off. */
167 exp <<= MANTSIZE; /* Shift the exponent into place */
168 exp += s; /* and add on the mantissa. */
169 return (exp);
170}
171
172static int
173acct_chkfree(void)
174{
175 int error;
176 struct statvfs *sb;
177 fsblkcnt_t bavail;
178
179 sb = kmem_alloc(sizeof(*sb), KM_SLEEP);
180 if (sb == NULL)
181 return (ENOMEM);
182 error = VFS_STATVFS(acct_vp->v_mount, sb);
183 if (error != 0) {
184 kmem_free(sb, sizeof(*sb));
185 return (error);
186 }
187
188 if (sb->f_bfree < sb->f_bresvd) {
189 bavail = 0;
190 } else {
191 bavail = sb->f_bfree - sb->f_bresvd;
192 }
193
194 switch (acct_state) {
195 case ACCT_SUSPENDED:
196 if (bavail > acctresume * sb->f_blocks / 100) {
197 acct_state = ACCT_ACTIVE;
198 log(LOG_NOTICE, "Accounting resumed\n");
199 }
200 break;
201 case ACCT_ACTIVE:
202 if (bavail <= acctsuspend * sb->f_blocks / 100) {
203 acct_state = ACCT_SUSPENDED;
204 log(LOG_NOTICE, "Accounting suspended\n");
205 }
206 break;
207 case ACCT_STOP:
208 break;
209 }
210 kmem_free(sb, sizeof(*sb));
211 return (0);
212}
213
214static void
215acct_stop(void)
216{
217 int error;
218
219 KASSERT(rw_write_held(&acct_lock));
220
221 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) {
222 error = vn_close(acct_vp, FWRITE, acct_cred);
223#ifdef DIAGNOSTIC
224 if (error != 0)
225 printf("acct_stop: failed to close, errno = %d\n",
226 error);
227#else
228 __USE(error);
229#endif
230 acct_vp = NULLVP;
231 }
232 if (acct_cred != NULL) {
233 kauth_cred_free(acct_cred);
234 acct_cred = NULL;
235 }
236 acct_state = ACCT_STOP;
237}
238
239/*
240 * Periodically check the file system to see if accounting
241 * should be turned on or off. Beware the case where the vnode
242 * has been vgone()'d out from underneath us, e.g. when the file
243 * system containing the accounting file has been forcibly unmounted.
244 */
245static void
246acctwatch(void *arg)
247{
248 int error;
249
250 log(LOG_NOTICE, "Accounting started\n");
251 rw_enter(&acct_lock, RW_WRITER);
252 while (acct_state != ACCT_STOP) {
253 if (acct_vp->v_type == VBAD) {
254 log(LOG_NOTICE, "Accounting terminated\n");
255 acct_stop();
256 continue;
257 }
258
259 error = acct_chkfree();
260#ifdef DIAGNOSTIC
261 if (error != 0)
262 printf("acctwatch: failed to statvfs, error = %d\n",
263 error);
264#else
265 __USE(error);
266#endif
267 rw_exit(&acct_lock);
268 error = kpause("actwat", false, acctchkfreq * hz, NULL);
269 rw_enter(&acct_lock, RW_WRITER);
270#ifdef DIAGNOSTIC
271 if (error != 0 && error != EWOULDBLOCK)
272 printf("acctwatch: sleep error %d\n", error);
273#endif
274 }
275 acct_dkwatcher = NULL;
276 rw_exit(&acct_lock);
277
278 kthread_exit(0);
279}
280
281void
282acct_init(void)
283{
284
285 acct_state = ACCT_STOP;
286 acct_vp = NULLVP;
287 acct_cred = NULL;
288 rw_init(&acct_lock);
289}
290
291/*
292 * Accounting system call. Written based on the specification and
293 * previous implementation done by Mark Tinguely.
294 */
295int
296sys_acct(struct lwp *l, const struct sys_acct_args *uap, register_t *retval)
297{
298 /* {
299 syscallarg(const char *) path;
300 } */
301 struct pathbuf *pb;
302 struct nameidata nd;
303 int error;
304
305 /* Make sure that the caller is root. */
306 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_ACCOUNTING,
307 0, NULL, NULL, NULL)))
308 return (error);
309
310 /*
311 * If accounting is to be started to a file, open that file for
312 * writing and make sure it's a 'normal'.
313 */
314 if (SCARG(uap, path) != NULL) {
315 struct vattr va;
316 size_t pad;
317
318 error = pathbuf_copyin(SCARG(uap, path), &pb);
319 if (error) {
320 return error;
321 }
322 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb);
323 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0) {
324 pathbuf_destroy(pb);
325 return error;
326 }
327 if (nd.ni_vp->v_type != VREG) {
328 VOP_UNLOCK(nd.ni_vp);
329 error = EACCES;
330 goto bad;
331 }
332 if ((error = VOP_GETATTR(nd.ni_vp, &va, l->l_cred)) != 0) {
333 VOP_UNLOCK(nd.ni_vp);
334 goto bad;
335 }
336
337 if ((pad = (va.va_size % sizeof(struct acct))) != 0) {
338 u_quad_t size = va.va_size - pad;
339#ifdef DIAGNOSTIC
340 printf("Size of accounting file not a multiple of "
341 "%lu - incomplete record truncated\n",
342 (unsigned long)sizeof(struct acct));
343#endif
344 vattr_null(&va);
345 va.va_size = size;
346 error = VOP_SETATTR(nd.ni_vp, &va, l->l_cred);
347 if (error != 0) {
348 VOP_UNLOCK(nd.ni_vp);
349 goto bad;
350 }
351 }
352 VOP_UNLOCK(nd.ni_vp);
353 }
354
355 rw_enter(&acct_lock, RW_WRITER);
356
357 /*
358 * If accounting was previously enabled, kill the old space-watcher,
359 * free credential for accounting file i/o,
360 * ... (and, if no new file was specified, leave).
361 */
362 acct_stop();
363 if (SCARG(uap, path) == NULL)
364 goto out;
365
366 /*
367 * Save the new accounting file vnode and credential,
368 * and schedule the new free space watcher.
369 */
370 acct_state = ACCT_ACTIVE;
371 acct_vp = nd.ni_vp;
372 acct_cred = l->l_cred;
373 kauth_cred_hold(acct_cred);
374
375 pathbuf_destroy(pb);
376
377 error = acct_chkfree(); /* Initial guess. */
378 if (error != 0) {
379 acct_stop();
380 goto out;
381 }
382
383 if (acct_dkwatcher == NULL) {
384 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL,
385 acctwatch, NULL, &acct_dkwatcher, "acctwatch");
386 if (error != 0)
387 acct_stop();
388 }
389
390 out:
391 rw_exit(&acct_lock);
392 return (error);
393 bad:
394 vn_close(nd.ni_vp, FWRITE, l->l_cred);
395 pathbuf_destroy(pb);
396 return error;
397}
398
399/*
400 * Write out process accounting information, on process exit.
401 * Data to be written out is specified in Leffler, et al.
402 * and are enumerated below. (They're also noted in the system
403 * "acct.h" header file.)
404 */
405int
406acct_process(struct lwp *l)
407{
408 struct acct acct;
409 struct timeval ut, st, tmp;
410 struct rusage *r;
411 int t, error = 0;
412 struct rlimit orlim;
413 struct proc *p = l->l_proc;
414
415 if (acct_state != ACCT_ACTIVE)
416 return 0;
417
418 memset(&acct, 0, sizeof(acct)); /* to zerofill padded data */
419
420 rw_enter(&acct_lock, RW_READER);
421
422 /* If accounting isn't enabled, don't bother */
423 if (acct_state != ACCT_ACTIVE)
424 goto out;
425
426 /*
427 * Temporarily raise the file limit so that accounting can't
428 * be stopped by the user.
429 *
430 * XXX We should think about the CPU limit, too.
431 */
432 lim_privatise(p);
433 orlim = p->p_rlimit[RLIMIT_FSIZE];
434 /* Set current and max to avoid illegal values */
435 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
436 p->p_rlimit[RLIMIT_FSIZE].rlim_max = RLIM_INFINITY;
437
438 /*
439 * Get process accounting information.
440 */
441
442 /* (1) The name of the command that ran */
443 strncpy(acct.ac_comm, p->p_comm, sizeof(acct.ac_comm));
444
445 /* (2) The amount of user and system time that was used */
446 mutex_enter(p->p_lock);
447 calcru(p, &ut, &st, NULL, NULL);
448 mutex_exit(p->p_lock);
449 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
450 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
451
452 /* (3) The elapsed time the commmand ran (and its starting time) */
453 acct.ac_btime = p->p_stats->p_start.tv_sec;
454 getmicrotime(&tmp);
455 timersub(&tmp, &p->p_stats->p_start, &tmp);
456 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
457
458 /* (4) The average amount of memory used */
459 r = &p->p_stats->p_ru;
460 timeradd(&ut, &st, &tmp);
461 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
462 if (t)
463 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
464 else
465 acct.ac_mem = 0;
466
467 /* (5) The number of disk I/O operations done */
468 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
469
470 /* (6) The UID and GID of the process */
471 acct.ac_uid = kauth_cred_getuid(l->l_cred);
472 acct.ac_gid = kauth_cred_getgid(l->l_cred);
473
474 /* (7) The terminal from which the process was started */
475 mutex_enter(proc_lock);
476 if ((p->p_lflag & PL_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
477 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev;
478 else
479 acct.ac_tty = NODEV;
480 mutex_exit(proc_lock);
481
482 /* (8) The boolean flags that tell how the process terminated, etc. */
483 acct.ac_flag = p->p_acflag;
484
485 /*
486 * Now, just write the accounting information to the file.
487 */
488 error = vn_rdwr(UIO_WRITE, acct_vp, (void *)&acct,
489 sizeof(acct), (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT,
490 acct_cred, NULL, NULL);
491 if (error != 0)
492 log(LOG_ERR, "Accounting: write failed %d\n", error);
493
494 /* Restore limit - rather pointless since process is about to exit */
495 p->p_rlimit[RLIMIT_FSIZE] = orlim;
496
497 out:
498 rw_exit(&acct_lock);
499 return (error);
500}
501