1 /* $NetBSD: kern_acct.c,v 1.99 2021/12/05 04:35:38 msaitoh 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 1.99 2021/12/05 04:35:38 msaitoh 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 */
110 krwlock_t acct_lock;
111
112 /*
113 * The global accounting state and related data. Gain the mutex before
114 * accessing these variables.
115 */
116 static enum {
117 ACCT_STOP,
118 ACCT_ACTIVE,
119 ACCT_SUSPENDED
120 } acct_state; /* The current accounting state. */
121 static struct vnode *acct_vp; /* Accounting vnode pointer. */
122 static kauth_cred_t acct_cred; /* Credential of accounting file
123 owner (i.e root). Used when
124 accounting file i/o. */
125 static struct lwp *acct_dkwatcher; /* Free disk space checker. */
126
127 /*
128 * Values associated with enabling and disabling accounting
129 */
130 int acctsuspend = 2; /* stop accounting when < 2% free space left */
131 int acctresume = 4; /* resume when free space risen to > 4% */
132 int 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
144 static comp_t
encode_comp_t(u_long s,u_long us)145 encode_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
172 static int
acct_chkfree(void)173 acct_chkfree(void)
174 {
175 int error;
176 struct statvfs *sb;
177 fsblkcnt_t bavail;
178
179 sb = kmem_alloc(sizeof(*sb), KM_SLEEP);
180 error = VFS_STATVFS(acct_vp->v_mount, sb);
181 if (error != 0) {
182 kmem_free(sb, sizeof(*sb));
183 return (error);
184 }
185
186 if (sb->f_bfree < sb->f_bresvd) {
187 bavail = 0;
188 } else {
189 bavail = sb->f_bfree - sb->f_bresvd;
190 }
191
192 switch (acct_state) {
193 case ACCT_SUSPENDED:
194 if (bavail > acctresume * sb->f_blocks / 100) {
195 acct_state = ACCT_ACTIVE;
196 log(LOG_NOTICE, "Accounting resumed\n");
197 }
198 break;
199 case ACCT_ACTIVE:
200 if (bavail <= acctsuspend * sb->f_blocks / 100) {
201 acct_state = ACCT_SUSPENDED;
202 log(LOG_NOTICE, "Accounting suspended\n");
203 }
204 break;
205 case ACCT_STOP:
206 break;
207 }
208 kmem_free(sb, sizeof(*sb));
209 return (0);
210 }
211
212 static void
acct_stop(void)213 acct_stop(void)
214 {
215 int error;
216
217 KASSERT(rw_write_held(&acct_lock));
218
219 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) {
220 error = vn_close(acct_vp, FWRITE, acct_cred);
221 #ifdef DIAGNOSTIC
222 if (error != 0)
223 printf("acct_stop: failed to close, errno = %d\n",
224 error);
225 #else
226 __USE(error);
227 #endif
228 acct_vp = NULLVP;
229 }
230 if (acct_cred != NULL) {
231 kauth_cred_free(acct_cred);
232 acct_cred = NULL;
233 }
234 acct_state = ACCT_STOP;
235 }
236
237 /*
238 * Periodically check the file system to see if accounting
239 * should be turned on or off. Beware the case where the vnode
240 * has been vgone()'d out from underneath us, e.g. when the file
241 * system containing the accounting file has been forcibly unmounted.
242 */
243 static void
acctwatch(void * arg)244 acctwatch(void *arg)
245 {
246 int error;
247
248 log(LOG_NOTICE, "Accounting started\n");
249 rw_enter(&acct_lock, RW_WRITER);
250 while (acct_state != ACCT_STOP) {
251 if (acct_vp->v_type == VBAD) {
252 log(LOG_NOTICE, "Accounting terminated\n");
253 acct_stop();
254 continue;
255 }
256
257 error = acct_chkfree();
258 #ifdef DIAGNOSTIC
259 if (error != 0)
260 printf("acctwatch: failed to statvfs, error = %d\n",
261 error);
262 #else
263 __USE(error);
264 #endif
265 rw_exit(&acct_lock);
266 error = kpause("actwat", false, acctchkfreq * hz, NULL);
267 rw_enter(&acct_lock, RW_WRITER);
268 #ifdef DIAGNOSTIC
269 if (error != 0 && error != EWOULDBLOCK)
270 printf("acctwatch: sleep error %d\n", error);
271 #endif
272 }
273 acct_dkwatcher = NULL;
274 rw_exit(&acct_lock);
275
276 kthread_exit(0);
277 }
278
279 void
acct_init(void)280 acct_init(void)
281 {
282
283 acct_state = ACCT_STOP;
284 acct_vp = NULLVP;
285 acct_cred = NULL;
286 rw_init(&acct_lock);
287 }
288
289 /*
290 * Accounting system call. Written based on the specification and
291 * previous implementation done by Mark Tinguely.
292 */
293 int
sys_acct(struct lwp * l,const struct sys_acct_args * uap,register_t * retval)294 sys_acct(struct lwp *l, const struct sys_acct_args *uap, register_t *retval)
295 {
296 /* {
297 syscallarg(const char *) path;
298 } */
299 struct pathbuf *pb;
300 struct vnode *vp;
301 int error;
302
303 /* Make sure that the caller is root. */
304 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_ACCOUNTING,
305 0, NULL, NULL, NULL)))
306 return (error);
307
308 /*
309 * If accounting is to be started to a file, open that file for
310 * writing and make sure it's a 'normal'.
311 */
312 if (SCARG(uap, path) != NULL) {
313 struct vattr va;
314 size_t pad;
315
316 error = pathbuf_copyin(SCARG(uap, path), &pb);
317 if (error) {
318 return error;
319 }
320 error = vn_open(NULL, pb, TRYEMULROOT, FWRITE|O_APPEND, 0,
321 &vp, NULL, NULL);
322 if (error != 0) {
323 pathbuf_destroy(pb);
324 return error;
325 }
326 if (vp->v_type != VREG) {
327 VOP_UNLOCK(vp);
328 error = EACCES;
329 goto bad;
330 }
331 if ((error = VOP_GETATTR(vp, &va, l->l_cred)) != 0) {
332 VOP_UNLOCK(vp);
333 goto bad;
334 }
335
336 if ((pad = (va.va_size % sizeof(struct acct))) != 0) {
337 u_quad_t size = va.va_size - pad;
338 #ifdef DIAGNOSTIC
339 printf("Size of accounting file not a multiple of "
340 "%lu - incomplete record truncated\n",
341 (unsigned long)sizeof(struct acct));
342 #endif
343 vattr_null(&va);
344 va.va_size = size;
345 error = VOP_SETATTR(vp, &va, l->l_cred);
346 if (error != 0) {
347 VOP_UNLOCK(vp);
348 goto bad;
349 }
350 }
351 VOP_UNLOCK(vp);
352 }
353
354 rw_enter(&acct_lock, RW_WRITER);
355
356 /*
357 * If accounting was previously enabled, kill the old space-watcher,
358 * free credential for accounting file i/o,
359 * ... (and, if no new file was specified, leave).
360 */
361 acct_stop();
362 if (SCARG(uap, path) == NULL)
363 goto out;
364
365 /*
366 * Save the new accounting file vnode and credential,
367 * and schedule the new free space watcher.
368 */
369 acct_state = ACCT_ACTIVE;
370 acct_vp = vp;
371 acct_cred = l->l_cred;
372 kauth_cred_hold(acct_cred);
373
374 pathbuf_destroy(pb);
375
376 error = acct_chkfree(); /* Initial guess. */
377 if (error != 0) {
378 acct_stop();
379 goto out;
380 }
381
382 if (acct_dkwatcher == NULL) {
383 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL,
384 acctwatch, NULL, &acct_dkwatcher, "acctwatch");
385 if (error != 0)
386 acct_stop();
387 }
388
389 out:
390 rw_exit(&acct_lock);
391 return (error);
392 bad:
393 vn_close(vp, FWRITE, l->l_cred);
394 pathbuf_destroy(pb);
395 return error;
396 }
397
398 /*
399 * Write out process accounting information, on process exit.
400 * Data to be written out is specified in Leffler, et al.
401 * and are enumerated below. (They're also noted in the system
402 * "acct.h" header file.)
403 */
404 int
acct_process(struct lwp * l)405 acct_process(struct lwp *l)
406 {
407 struct acct acct;
408 struct timeval ut, st, tmp;
409 struct rusage *r;
410 int t, error = 0;
411 struct rlimit orlim;
412 struct proc *p = l->l_proc;
413
414 if (acct_state != ACCT_ACTIVE)
415 return 0;
416
417 memset(&acct, 0, sizeof(acct)); /* to zerofill padded data */
418
419 rw_enter(&acct_lock, RW_READER);
420
421 /* If accounting isn't enabled, don't bother */
422 if (acct_state != ACCT_ACTIVE)
423 goto out;
424
425 /*
426 * Temporarily raise the file limit so that accounting can't
427 * be stopped by the user.
428 *
429 * XXX We should think about the CPU limit, too.
430 */
431 lim_privatise(p);
432 orlim = p->p_rlimit[RLIMIT_FSIZE];
433 /* Set current and max to avoid illegal values */
434 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
435 p->p_rlimit[RLIMIT_FSIZE].rlim_max = RLIM_INFINITY;
436
437 /*
438 * Get process accounting information.
439 */
440
441 /* (1) The name of the command that ran */
442 strncpy(acct.ac_comm, p->p_comm, sizeof(acct.ac_comm));
443
444 /* (2) The amount of user and system time that was used */
445 mutex_enter(p->p_lock);
446 calcru(p, &ut, &st, NULL, NULL);
447 mutex_exit(p->p_lock);
448 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
449 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
450
451 /* (3) The elapsed time the command ran (and its starting time) */
452 acct.ac_btime = p->p_stats->p_start.tv_sec;
453 getmicrotime(&tmp);
454 timersub(&tmp, &p->p_stats->p_start, &tmp);
455 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
456
457 /* (4) The average amount of memory used */
458 r = &p->p_stats->p_ru;
459 timeradd(&ut, &st, &tmp);
460 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
461 if (t)
462 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
463 else
464 acct.ac_mem = 0;
465
466 /* (5) The number of disk I/O operations done */
467 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
468
469 /* (6) The UID and GID of the process */
470 acct.ac_uid = kauth_cred_getuid(l->l_cred);
471 acct.ac_gid = kauth_cred_getgid(l->l_cred);
472
473 /* (7) The terminal from which the process was started */
474 mutex_enter(&proc_lock);
475 if ((p->p_lflag & PL_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
476 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev;
477 else
478 acct.ac_tty = NODEV;
479 mutex_exit(&proc_lock);
480
481 /* (8) The boolean flags that tell how the process terminated, etc. */
482 acct.ac_flag = p->p_acflag;
483
484 /*
485 * Now, just write the accounting information to the file.
486 */
487 error = vn_rdwr(UIO_WRITE, acct_vp, (void *)&acct,
488 sizeof(acct), (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT,
489 acct_cred, NULL, NULL);
490 if (error != 0)
491 log(LOG_ERR, "Accounting: write failed %d\n", error);
492
493 /* Restore limit - rather pointless since process is about to exit */
494 p->p_rlimit[RLIMIT_FSIZE] = orlim;
495
496 out:
497 rw_exit(&acct_lock);
498 return (error);
499 }
500