kern_acct.c revision 1.62.10.3 1 /* $NetBSD: kern_acct.c,v 1.62.10.3 2006/05/06 23:31:30 christos 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 1.62.10.3 2006/05/06 23:31:30 christos 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/lock.h>
86 #include <sys/malloc.h>
87 #include <sys/namei.h>
88 #include <sys/errno.h>
89 #include <sys/acct.h>
90 #include <sys/resourcevar.h>
91 #include <sys/ioctl.h>
92 #include <sys/tty.h>
93 #include <sys/kauth.h>
94
95 #include <sys/sa.h>
96 #include <sys/syscallargs.h>
97
98 /*
99 * The routines implemented in this file are described in:
100 * Leffler, et al.: The Design and Implementation of the 4.3BSD
101 * UNIX Operating System (Addison Welley, 1989)
102 * on pages 62-63.
103 *
104 * Arguably, to simplify accounting operations, this mechanism should
105 * be replaced by one in which an accounting log file (similar to /dev/klog)
106 * is read by a user process, etc. However, that has its own problems.
107 */
108
109 /*
110 * Lock to serialize system calls and kernel threads.
111 */
112 static struct lock acct_lock;
113 #define ACCT_LOCK() \
114 do { \
115 (void) lockmgr(&acct_lock, LK_EXCLUSIVE, NULL); \
116 } while (/* CONSTCOND */0)
117 #define ACCT_UNLOCK() \
118 do { \
119 (void) lockmgr(&acct_lock, LK_RELEASE, NULL); \
120 } while (/* CONSTCOND */0)
121
122 /*
123 * The global accounting state and related data. Gain the lock before
124 * accessing these variables.
125 */
126 static enum {
127 ACCT_STOP,
128 ACCT_ACTIVE,
129 ACCT_SUSPENDED
130 } acct_state; /* The current accounting state. */
131 static struct vnode *acct_vp; /* Accounting vnode pointer. */
132 static kauth_cred_t acct_cred; /* Credential of accounting file
133 owner (i.e root). Used when
134 accounting file i/o. */
135 static struct proc *acct_dkwatcher; /* Free disk space checker. */
136
137 /*
138 * Values associated with enabling and disabling accounting
139 */
140 int acctsuspend = 2; /* stop accounting when < 2% free space left */
141 int acctresume = 4; /* resume when free space risen to > 4% */
142 int acctchkfreq = 15; /* frequency (in seconds) to check space */
143
144 /*
145 * Encode_comp_t converts from ticks in seconds and microseconds
146 * to ticks in 1/AHZ seconds. The encoding is described in
147 * Leffler, et al., on page 63.
148 */
149
150 #define MANTSIZE 13 /* 13 bit mantissa. */
151 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
152 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
153
154 static comp_t
155 encode_comp_t(u_long s, u_long us)
156 {
157 int exp, rnd;
158
159 exp = 0;
160 rnd = 0;
161 s *= AHZ;
162 s += us / (1000000 / AHZ); /* Maximize precision. */
163
164 while (s > MAXFRACT) {
165 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
166 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
167 exp++;
168 }
169
170 /* If we need to round up, do it (and handle overflow correctly). */
171 if (rnd && (++s > MAXFRACT)) {
172 s >>= EXPSIZE;
173 exp++;
174 }
175
176 /* Clean it up and polish it off. */
177 exp <<= MANTSIZE; /* Shift the exponent into place */
178 exp += s; /* and add on the mantissa. */
179 return (exp);
180 }
181
182 static int
183 acct_chkfree(void)
184 {
185 int error;
186 struct statvfs sb;
187 int64_t bavail;
188
189 error = VFS_STATVFS(acct_vp->v_mount, &sb, NULL);
190 if (error != 0)
191 return (error);
192
193 bavail = sb.f_bfree - sb.f_bresvd;
194
195 switch (acct_state) {
196 case ACCT_SUSPENDED:
197 if (bavail > acctresume * sb.f_blocks / 100) {
198 acct_state = ACCT_ACTIVE;
199 log(LOG_NOTICE, "Accounting resumed\n");
200 }
201 break;
202 case ACCT_ACTIVE:
203 if (bavail <= acctsuspend * sb.f_blocks / 100) {
204 acct_state = ACCT_SUSPENDED;
205 log(LOG_NOTICE, "Accounting suspended\n");
206 }
207 break;
208 case ACCT_STOP:
209 break;
210 }
211 return (0);
212 }
213
214 static void
215 acct_stop(void)
216 {
217 int error;
218
219 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) {
220 error = vn_close(acct_vp, FWRITE, acct_cred, NULL);
221 #ifdef DIAGNOSTIC
222 if (error != 0)
223 printf("acct_stop: failed to close, errno = %d\n",
224 error);
225 #endif
226 acct_vp = NULLVP;
227 }
228 if (acct_cred != NULL) {
229 kauth_cred_free(acct_cred);
230 acct_cred = NULL;
231 }
232 acct_state = ACCT_STOP;
233 }
234
235 /*
236 * Periodically check the file system to see if accounting
237 * should be turned on or off. Beware the case where the vnode
238 * has been vgone()'d out from underneath us, e.g. when the file
239 * system containing the accounting file has been forcibly unmounted.
240 */
241 static void
242 acctwatch(void *arg)
243 {
244 int error;
245
246 log(LOG_NOTICE, "Accounting started\n");
247 ACCT_LOCK();
248 while (acct_state != ACCT_STOP) {
249 if (acct_vp->v_type == VBAD) {
250 log(LOG_NOTICE, "Accounting terminated\n");
251 acct_stop();
252 continue;
253 }
254
255 error = acct_chkfree();
256 #ifdef DIAGNOSTIC
257 if (error != 0)
258 printf("acctwatch: failed to statvfs, error = %d\n",
259 error);
260 #endif
261
262 ACCT_UNLOCK();
263 error = tsleep(acctwatch, PSWP, "actwat", acctchkfreq * hz);
264 ACCT_LOCK();
265 #ifdef DIAGNOSTIC
266 if (error != 0 && error != EWOULDBLOCK)
267 printf("acctwatch: sleep error %d\n", error);
268 #endif
269 }
270 acct_dkwatcher = NULL;
271 ACCT_UNLOCK();
272
273 kthread_exit(0);
274 }
275
276 void
277 acct_init(void)
278 {
279
280 acct_state = ACCT_STOP;
281 acct_vp = NULLVP;
282 acct_cred = NULL;
283 lockinit(&acct_lock, PWAIT, "acctlk", 0, 0);
284 }
285
286 /*
287 * Accounting system call. Written based on the specification and
288 * previous implementation done by Mark Tinguely.
289 */
290 int
291 sys_acct(struct lwp *l, void *v, register_t *retval)
292 {
293 struct sys_acct_args /* {
294 syscallarg(const char *) path;
295 } */ *uap = v;
296 struct nameidata nd;
297 int error;
298 struct proc *p = l->l_proc;
299
300 /* Make sure that the caller is root. */
301 if ((error = kauth_authorize_generic(p->p_cred, KAUTH_GENERIC_ISSUSER,
302 &p->p_acflag)) != 0)
303 return (error);
304
305 /*
306 * If accounting is to be started to a file, open that file for
307 * writing and make sure it's a 'normal'.
308 */
309 if (SCARG(uap, path) != NULL) {
310 struct vattr va;
311 size_t pad;
312 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path),
313 l);
314 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0)
315 return (error);
316 if (nd.ni_vp->v_type != VREG) {
317 VOP_UNLOCK(nd.ni_vp, 0);
318 error = EACCES;
319 goto bad;
320 }
321 if ((error = VOP_GETATTR(nd.ni_vp, &va, p->p_cred, l)) != 0) {
322 VOP_UNLOCK(nd.ni_vp, 0);
323 goto bad;
324 }
325
326 if ((pad = (va.va_size % sizeof(struct acct))) != 0) {
327 u_quad_t size = va.va_size - pad;
328 #ifdef DIAGNOSTIC
329 printf("Size of accounting file not a multiple of "
330 "%lu - incomplete record truncated\n",
331 (unsigned long)sizeof(struct acct));
332 #endif
333 VATTR_NULL(&va);
334 va.va_size = size;
335 error = VOP_SETATTR(nd.ni_vp, &va, p->p_cred, l);
336 if (error != 0) {
337 VOP_UNLOCK(nd.ni_vp, 0);
338 goto bad;
339 }
340 }
341 VOP_UNLOCK(nd.ni_vp, 0);
342 }
343
344 ACCT_LOCK();
345
346 /*
347 * If accounting was previously enabled, kill the old space-watcher,
348 * free credential for accounting file i/o,
349 * ... (and, if no new file was specified, leave).
350 */
351 acct_stop();
352 if (SCARG(uap, path) == NULL)
353 goto out;
354
355 /*
356 * Save the new accounting file vnode and credential,
357 * and schedule the new free space watcher.
358 */
359 acct_state = ACCT_ACTIVE;
360 acct_vp = nd.ni_vp;
361 acct_cred = p->p_cred;
362 kauth_cred_hold(acct_cred);
363
364 error = acct_chkfree(); /* Initial guess. */
365 if (error != 0) {
366 acct_stop();
367 goto out;
368 }
369
370 if (acct_dkwatcher == NULL) {
371 error = kthread_create1(acctwatch, NULL, &acct_dkwatcher,
372 "acctwatch");
373 if (error != 0)
374 acct_stop();
375 }
376
377 out:
378 ACCT_UNLOCK();
379 return (error);
380 bad:
381 vn_close(nd.ni_vp, FWRITE, p->p_cred, l);
382 return error;
383 }
384
385 /*
386 * Write out process accounting information, on process exit.
387 * Data to be written out is specified in Leffler, et al.
388 * and are enumerated below. (They're also noted in the system
389 * "acct.h" header file.)
390 */
391 int
392 acct_process(struct lwp *l)
393 {
394 struct acct acct;
395 struct rusage *r;
396 struct timeval ut, st, tmp;
397 int s, t, error = 0;
398 struct plimit *oplim = NULL;
399 struct proc *p = l->l_proc;
400
401 ACCT_LOCK();
402
403 /* If accounting isn't enabled, don't bother */
404 if (acct_state != ACCT_ACTIVE)
405 goto out;
406
407 /*
408 * Raise the file limit so that accounting can't be stopped by
409 * the user.
410 *
411 * XXX We should think about the CPU limit, too.
412 */
413 if (p->p_limit->p_refcnt > 1) {
414 oplim = p->p_limit;
415 p->p_limit = limcopy(p->p_limit);
416 }
417 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
418
419 /*
420 * Get process accounting information.
421 */
422
423 /* (1) The name of the command that ran */
424 memcpy(acct.ac_comm, p->p_comm, sizeof(acct.ac_comm));
425
426 /* (2) The amount of user and system time that was used */
427 calcru(p, &ut, &st, NULL);
428 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
429 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
430
431 /* (3) The elapsed time the commmand ran (and its starting time) */
432 acct.ac_btime = p->p_stats->p_start.tv_sec;
433 s = splclock();
434 timersub(&time, &p->p_stats->p_start, &tmp);
435 splx(s);
436 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
437
438 /* (4) The average amount of memory used */
439 r = &p->p_stats->p_ru;
440 timeradd(&ut, &st, &tmp);
441 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
442 if (t)
443 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
444 else
445 acct.ac_mem = 0;
446
447 /* (5) The number of disk I/O operations done */
448 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
449
450 /* (6) The UID and GID of the process */
451 acct.ac_uid = kauth_cred_getuid(p->p_cred);
452 acct.ac_gid = kauth_cred_getgid(p->p_cred);
453
454 /* (7) The terminal from which the process was started */
455 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
456 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev;
457 else
458 acct.ac_tty = NODEV;
459
460 /* (8) The boolean flags that tell how the process terminated, etc. */
461 acct.ac_flag = p->p_acflag;
462
463 /*
464 * Now, just write the accounting information to the file.
465 */
466 VOP_LEASE(acct_vp, l, p->p_cred, LEASE_WRITE);
467 error = vn_rdwr(UIO_WRITE, acct_vp, (caddr_t)&acct,
468 sizeof(acct), (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT,
469 acct_cred, NULL, NULL);
470 if (error != 0)
471 log(LOG_ERR, "Accounting: write failed %d\n", error);
472
473 if (oplim) {
474 limfree(p->p_limit);
475 p->p_limit = oplim;
476 }
477
478 out:
479 ACCT_UNLOCK();
480 return (error);
481 }
482