kern_acct.c revision 1.74
1 /* $NetBSD: kern_acct.c,v 1.74 2007/04/22 08:30:00 dsl 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 1.74 2007/04/22 08:30:00 dsl 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/malloc.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 * Mutex to serialize system calls and kernel threads. 109 */ 110 kmutex_t acct_mutex; 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 proc *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 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 173 acct_chkfree(void) 174 { 175 int error; 176 struct statvfs *sb; 177 int64_t bavail; 178 179 sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK); 180 error = VFS_STATVFS(acct_vp->v_mount, sb, NULL); 181 if (error != 0) 182 return (error); 183 184 bavail = sb->f_bfree - sb->f_bresvd; 185 186 switch (acct_state) { 187 case ACCT_SUSPENDED: 188 if (bavail > acctresume * sb->f_blocks / 100) { 189 acct_state = ACCT_ACTIVE; 190 log(LOG_NOTICE, "Accounting resumed\n"); 191 } 192 break; 193 case ACCT_ACTIVE: 194 if (bavail <= acctsuspend * sb->f_blocks / 100) { 195 acct_state = ACCT_SUSPENDED; 196 log(LOG_NOTICE, "Accounting suspended\n"); 197 } 198 break; 199 case ACCT_STOP: 200 break; 201 } 202 free(sb, M_TEMP); 203 return (0); 204 } 205 206 static void 207 acct_stop(void) 208 { 209 int error; 210 211 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) { 212 error = vn_close(acct_vp, FWRITE, acct_cred, NULL); 213 #ifdef DIAGNOSTIC 214 if (error != 0) 215 printf("acct_stop: failed to close, errno = %d\n", 216 error); 217 #endif 218 acct_vp = NULLVP; 219 } 220 if (acct_cred != NULL) { 221 kauth_cred_free(acct_cred); 222 acct_cred = NULL; 223 } 224 acct_state = ACCT_STOP; 225 } 226 227 /* 228 * Periodically check the file system to see if accounting 229 * should be turned on or off. Beware the case where the vnode 230 * has been vgone()'d out from underneath us, e.g. when the file 231 * system containing the accounting file has been forcibly unmounted. 232 */ 233 static void 234 acctwatch(void *arg) 235 { 236 int error; 237 238 log(LOG_NOTICE, "Accounting started\n"); 239 mutex_enter(&acct_mutex); 240 while (acct_state != ACCT_STOP) { 241 if (acct_vp->v_type == VBAD) { 242 log(LOG_NOTICE, "Accounting terminated\n"); 243 acct_stop(); 244 continue; 245 } 246 247 error = acct_chkfree(); 248 #ifdef DIAGNOSTIC 249 if (error != 0) 250 printf("acctwatch: failed to statvfs, error = %d\n", 251 error); 252 #endif 253 error = kpause("actwat", false, acctchkfreq * hz, &acct_mutex); 254 #ifdef DIAGNOSTIC 255 if (error != 0 && error != EWOULDBLOCK) 256 printf("acctwatch: sleep error %d\n", error); 257 #endif 258 } 259 acct_dkwatcher = NULL; 260 mutex_exit(&acct_mutex); 261 262 kthread_exit(0); 263 } 264 265 void 266 acct_init(void) 267 { 268 269 acct_state = ACCT_STOP; 270 acct_vp = NULLVP; 271 acct_cred = NULL; 272 mutex_init(&acct_mutex, MUTEX_DEFAULT, IPL_NONE); 273 } 274 275 /* 276 * Accounting system call. Written based on the specification and 277 * previous implementation done by Mark Tinguely. 278 */ 279 int 280 sys_acct(struct lwp *l, void *v, register_t *retval) 281 { 282 struct sys_acct_args /* { 283 syscallarg(const char *) path; 284 } */ *uap = v; 285 struct nameidata nd; 286 int error; 287 288 /* Make sure that the caller is root. */ 289 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_ACCOUNTING, 290 0, NULL, NULL, NULL))) 291 return (error); 292 293 /* 294 * If accounting is to be started to a file, open that file for 295 * writing and make sure it's a 'normal'. 296 */ 297 if (SCARG(uap, path) != NULL) { 298 struct vattr va; 299 size_t pad; 300 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, SCARG(uap, path), 301 l); 302 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0) 303 return (error); 304 if (nd.ni_vp->v_type != VREG) { 305 VOP_UNLOCK(nd.ni_vp, 0); 306 error = EACCES; 307 goto bad; 308 } 309 if ((error = VOP_GETATTR(nd.ni_vp, &va, l->l_cred, l)) != 0) { 310 VOP_UNLOCK(nd.ni_vp, 0); 311 goto bad; 312 } 313 314 if ((pad = (va.va_size % sizeof(struct acct))) != 0) { 315 u_quad_t size = va.va_size - pad; 316 #ifdef DIAGNOSTIC 317 printf("Size of accounting file not a multiple of " 318 "%lu - incomplete record truncated\n", 319 (unsigned long)sizeof(struct acct)); 320 #endif 321 VATTR_NULL(&va); 322 va.va_size = size; 323 error = VOP_SETATTR(nd.ni_vp, &va, l->l_cred, l); 324 if (error != 0) { 325 VOP_UNLOCK(nd.ni_vp, 0); 326 goto bad; 327 } 328 } 329 VOP_UNLOCK(nd.ni_vp, 0); 330 } 331 332 mutex_enter(&acct_mutex); 333 334 /* 335 * If accounting was previously enabled, kill the old space-watcher, 336 * free credential for accounting file i/o, 337 * ... (and, if no new file was specified, leave). 338 */ 339 acct_stop(); 340 if (SCARG(uap, path) == NULL) 341 goto out; 342 343 /* 344 * Save the new accounting file vnode and credential, 345 * and schedule the new free space watcher. 346 */ 347 acct_state = ACCT_ACTIVE; 348 acct_vp = nd.ni_vp; 349 acct_cred = l->l_cred; 350 kauth_cred_hold(acct_cred); 351 352 error = acct_chkfree(); /* Initial guess. */ 353 if (error != 0) { 354 acct_stop(); 355 goto out; 356 } 357 358 if (acct_dkwatcher == NULL) { 359 error = kthread_create1(acctwatch, NULL, &acct_dkwatcher, 360 "acctwatch"); 361 if (error != 0) 362 acct_stop(); 363 } 364 365 out: 366 mutex_exit(&acct_mutex); 367 return (error); 368 bad: 369 vn_close(nd.ni_vp, FWRITE, l->l_cred, l); 370 return error; 371 } 372 373 /* 374 * Write out process accounting information, on process exit. 375 * Data to be written out is specified in Leffler, et al. 376 * and are enumerated below. (They're also noted in the system 377 * "acct.h" header file.) 378 */ 379 int 380 acct_process(struct lwp *l) 381 { 382 struct acct acct; 383 struct timeval ut, st, tmp; 384 struct rusage *r; 385 int t, error = 0; 386 struct plimit *oplim = NULL; 387 struct proc *p = l->l_proc; 388 389 mutex_enter(&acct_mutex); 390 391 /* If accounting isn't enabled, don't bother */ 392 if (acct_state != ACCT_ACTIVE) 393 goto out; 394 395 /* 396 * Raise the file limit so that accounting can't be stopped by 397 * the user. 398 * 399 * XXX We should think about the CPU limit, too. 400 */ 401 mutex_enter(&p->p_mutex); 402 if (p->p_limit->p_refcnt > 1) { 403 oplim = p->p_limit; 404 p->p_limit = limcopy(p); 405 } 406 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 407 mutex_exit(&p->p_mutex); 408 409 /* 410 * Get process accounting information. 411 */ 412 413 /* (1) The name of the command that ran */ 414 memcpy(acct.ac_comm, p->p_comm, sizeof(acct.ac_comm)); 415 416 /* (2) The amount of user and system time that was used */ 417 mutex_enter(&p->p_smutex); 418 calcru(p, &ut, &st, NULL, NULL); 419 mutex_exit(&p->p_smutex); 420 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 421 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 422 423 /* (3) The elapsed time the commmand ran (and its starting time) */ 424 acct.ac_btime = p->p_stats->p_start.tv_sec; 425 getmicrotime(&tmp); 426 timersub(&tmp, &p->p_stats->p_start, &tmp); 427 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 428 429 /* (4) The average amount of memory used */ 430 r = &p->p_stats->p_ru; 431 timeradd(&ut, &st, &tmp); 432 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 433 if (t) 434 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 435 else 436 acct.ac_mem = 0; 437 438 /* (5) The number of disk I/O operations done */ 439 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 440 441 /* (6) The UID and GID of the process */ 442 acct.ac_uid = kauth_cred_getuid(l->l_cred); 443 acct.ac_gid = kauth_cred_getgid(l->l_cred); 444 445 /* (7) The terminal from which the process was started */ 446 mutex_enter(&proclist_lock); 447 if ((p->p_lflag & PL_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 448 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev; 449 else 450 acct.ac_tty = NODEV; 451 mutex_exit(&proclist_lock); 452 453 /* (8) The boolean flags that tell how the process terminated, etc. */ 454 acct.ac_flag = p->p_acflag; 455 456 /* 457 * Now, just write the accounting information to the file. 458 */ 459 VOP_LEASE(acct_vp, l, l->l_cred, LEASE_WRITE); 460 error = vn_rdwr(UIO_WRITE, acct_vp, (void *)&acct, 461 sizeof(acct), (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, 462 acct_cred, NULL, NULL); 463 if (error != 0) 464 log(LOG_ERR, "Accounting: write failed %d\n", error); 465 466 if (oplim) { 467 mutex_enter(&p->p_mutex); 468 limfree(p->p_limit); 469 p->p_limit = oplim; 470 mutex_exit(&p->p_mutex); 471 } 472 473 out: 474 mutex_exit(&acct_mutex); 475 return (error); 476 } 477
Indexes created Sat Sep 20 22:09:52 GMT 2025