1 /* $NetBSD: kern_acct.c,v 1.68 2006/10/12 01:32:14 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 1.68 2006/10/12 01:32:14 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 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 190 error = VFS_STATVFS(acct_vp->v_mount, sb, NULL); 191 if (error != 0) 192 return (error); 193 194 bavail = sb->f_bfree - sb->f_bresvd; 195 196 switch (acct_state) { 197 case ACCT_SUSPENDED: 198 if (bavail > acctresume * sb->f_blocks / 100) { 199 acct_state = ACCT_ACTIVE; 200 log(LOG_NOTICE, "Accounting resumed\n"); 201 } 202 break; 203 case ACCT_ACTIVE: 204 if (bavail <= acctsuspend * sb->f_blocks / 100) { 205 acct_state = ACCT_SUSPENDED; 206 log(LOG_NOTICE, "Accounting suspended\n"); 207 } 208 break; 209 case ACCT_STOP: 210 break; 211 } 212 free(sb, M_TEMP); 213 return (0); 214 } 215 216 static void 217 acct_stop(void) 218 { 219 int error; 220 221 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) { 222 error = vn_close(acct_vp, FWRITE, acct_cred, NULL); 223 #ifdef DIAGNOSTIC 224 if (error != 0) 225 printf("acct_stop: failed to close, errno = %d\n", 226 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 244 acctwatch(void *arg __unused) 245 { 246 int error; 247 248 log(LOG_NOTICE, "Accounting started\n"); 249 ACCT_LOCK(); 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 #endif 263 264 ACCT_UNLOCK(); 265 error = tsleep(acctwatch, PSWP, "actwat", acctchkfreq * hz); 266 ACCT_LOCK(); 267 #ifdef DIAGNOSTIC 268 if (error != 0 && error != EWOULDBLOCK) 269 printf("acctwatch: sleep error %d\n", error); 270 #endif 271 } 272 acct_dkwatcher = NULL; 273 ACCT_UNLOCK(); 274 275 kthread_exit(0); 276 } 277 278 void 279 acct_init(void) 280 { 281 282 acct_state = ACCT_STOP; 283 acct_vp = NULLVP; 284 acct_cred = NULL; 285 lockinit(&acct_lock, PWAIT, "acctlk", 0, 0); 286 } 287 288 /* 289 * Accounting system call. Written based on the specification and 290 * previous implementation done by Mark Tinguely. 291 */ 292 int 293 sys_acct(struct lwp *l, void *v, register_t *retval __unused) 294 { 295 struct sys_acct_args /* { 296 syscallarg(const char *) path; 297 } */ *uap = v; 298 struct nameidata nd; 299 int error; 300 301 /* Make sure that the caller is root. */ 302 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_ACCOUNTING, 303 0, NULL, NULL, NULL))) 304 return (error); 305 306 /* 307 * If accounting is to be started to a file, open that file for 308 * writing and make sure it's a 'normal'. 309 */ 310 if (SCARG(uap, path) != NULL) { 311 struct vattr va; 312 size_t pad; 313 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), 314 l); 315 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0) 316 return (error); 317 if (nd.ni_vp->v_type != VREG) { 318 VOP_UNLOCK(nd.ni_vp, 0); 319 error = EACCES; 320 goto bad; 321 } 322 if ((error = VOP_GETATTR(nd.ni_vp, &va, l->l_cred, l)) != 0) { 323 VOP_UNLOCK(nd.ni_vp, 0); 324 goto bad; 325 } 326 327 if ((pad = (va.va_size % sizeof(struct acct))) != 0) { 328 u_quad_t size = va.va_size - pad; 329 #ifdef DIAGNOSTIC 330 printf("Size of accounting file not a multiple of " 331 "%lu - incomplete record truncated\n", 332 (unsigned long)sizeof(struct acct)); 333 #endif 334 VATTR_NULL(&va); 335 va.va_size = size; 336 error = VOP_SETATTR(nd.ni_vp, &va, l->l_cred, l); 337 if (error != 0) { 338 VOP_UNLOCK(nd.ni_vp, 0); 339 goto bad; 340 } 341 } 342 VOP_UNLOCK(nd.ni_vp, 0); 343 } 344 345 ACCT_LOCK(); 346 347 /* 348 * If accounting was previously enabled, kill the old space-watcher, 349 * free credential for accounting file i/o, 350 * ... (and, if no new file was specified, leave). 351 */ 352 acct_stop(); 353 if (SCARG(uap, path) == NULL) 354 goto out; 355 356 /* 357 * Save the new accounting file vnode and credential, 358 * and schedule the new free space watcher. 359 */ 360 acct_state = ACCT_ACTIVE; 361 acct_vp = nd.ni_vp; 362 acct_cred = l->l_cred; 363 kauth_cred_hold(acct_cred); 364 365 error = acct_chkfree(); /* Initial guess. */ 366 if (error != 0) { 367 acct_stop(); 368 goto out; 369 } 370 371 if (acct_dkwatcher == NULL) { 372 error = kthread_create1(acctwatch, NULL, &acct_dkwatcher, 373 "acctwatch"); 374 if (error != 0) 375 acct_stop(); 376 } 377 378 out: 379 ACCT_UNLOCK(); 380 return (error); 381 bad: 382 vn_close(nd.ni_vp, FWRITE, l->l_cred, l); 383 return error; 384 } 385 386 /* 387 * Write out process accounting information, on process exit. 388 * Data to be written out is specified in Leffler, et al. 389 * and are enumerated below. (They're also noted in the system 390 * "acct.h" header file.) 391 */ 392 int 393 acct_process(struct lwp *l) 394 { 395 struct acct acct; 396 struct timeval ut, st, tmp; 397 struct rusage *r; 398 int t, error = 0; 399 struct plimit *oplim = NULL; 400 struct proc *p = l->l_proc; 401 402 ACCT_LOCK(); 403 404 /* If accounting isn't enabled, don't bother */ 405 if (acct_state != ACCT_ACTIVE) 406 goto out; 407 408 /* 409 * Raise the file limit so that accounting can't be stopped by 410 * the user. 411 * 412 * XXX We should think about the CPU limit, too. 413 */ 414 if (p->p_limit->p_refcnt > 1) { 415 oplim = p->p_limit; 416 p->p_limit = limcopy(p->p_limit); 417 } 418 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 419 420 /* 421 * Get process accounting information. 422 */ 423 424 /* (1) The name of the command that ran */ 425 memcpy(acct.ac_comm, p->p_comm, sizeof(acct.ac_comm)); 426 427 /* (2) The amount of user and system time that was used */ 428 calcru(p, &ut, &st, NULL); 429 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 430 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 431 432 /* (3) The elapsed time the commmand ran (and its starting time) */ 433 acct.ac_btime = p->p_stats->p_start.tv_sec; 434 getmicrotime(&tmp); 435 timersub(&tmp, &p->p_stats->p_start, &tmp); 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(l->l_cred); 452 acct.ac_gid = kauth_cred_getgid(l->l_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, l->l_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