kern_acct.c revision 1.86.4.1
1 /* $NetBSD: kern_acct.c,v 1.86.4.1 2008/05/10 23:49:03 wrstuden 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 1.86.4.1 2008/05/10 23:49:03 wrstuden 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/sa.h> 95 #include <sys/syscallargs.h> 96 97 /* 98 * The routines implemented in this file are described in: 99 * Leffler, et al.: The Design and Implementation of the 4.3BSD 100 * UNIX Operating System (Addison Welley, 1989) 101 * on pages 62-63. 102 * 103 * Arguably, to simplify accounting operations, this mechanism should 104 * be replaced by one in which an accounting log file (similar to /dev/klog) 105 * is read by a user process, etc. However, that has its own problems. 106 */ 107 108 /* 109 * Lock to serialize system calls and kernel threads. 110 */ 111 krwlock_t acct_lock; 112 113 /* 114 * The global accounting state and related data. Gain the mutex before 115 * accessing these variables. 116 */ 117 static enum { 118 ACCT_STOP, 119 ACCT_ACTIVE, 120 ACCT_SUSPENDED 121 } acct_state; /* The current accounting state. */ 122 static struct vnode *acct_vp; /* Accounting vnode pointer. */ 123 static kauth_cred_t acct_cred; /* Credential of accounting file 124 owner (i.e root). Used when 125 accounting file i/o. */ 126 static struct lwp *acct_dkwatcher; /* Free disk space checker. */ 127 128 /* 129 * Values associated with enabling and disabling accounting 130 */ 131 int acctsuspend = 2; /* stop accounting when < 2% free space left */ 132 int acctresume = 4; /* resume when free space risen to > 4% */ 133 int acctchkfreq = 15; /* frequency (in seconds) to check space */ 134 135 /* 136 * Encode_comp_t converts from ticks in seconds and microseconds 137 * to ticks in 1/AHZ seconds. The encoding is described in 138 * Leffler, et al., on page 63. 139 */ 140 141 #define MANTSIZE 13 /* 13 bit mantissa. */ 142 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 143 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 144 145 static comp_t 146 encode_comp_t(u_long s, u_long us) 147 { 148 int exp, rnd; 149 150 exp = 0; 151 rnd = 0; 152 s *= AHZ; 153 s += us / (1000000 / AHZ); /* Maximize precision. */ 154 155 while (s > MAXFRACT) { 156 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 157 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 158 exp++; 159 } 160 161 /* If we need to round up, do it (and handle overflow correctly). */ 162 if (rnd && (++s > MAXFRACT)) { 163 s >>= EXPSIZE; 164 exp++; 165 } 166 167 /* Clean it up and polish it off. */ 168 exp <<= MANTSIZE; /* Shift the exponent into place */ 169 exp += s; /* and add on the mantissa. */ 170 return (exp); 171 } 172 173 static int 174 acct_chkfree(void) 175 { 176 int error; 177 struct statvfs *sb; 178 int64_t bavail; 179 180 sb = kmem_alloc(sizeof(*sb), KM_SLEEP); 181 if (sb == NULL) 182 return (ENOMEM); 183 error = VFS_STATVFS(acct_vp->v_mount, sb); 184 if (error != 0) { 185 kmem_free(sb, sizeof(*sb)); 186 return (error); 187 } 188 189 bavail = sb->f_bfree - sb->f_bresvd; 190 191 switch (acct_state) { 192 case ACCT_SUSPENDED: 193 if (bavail > acctresume * sb->f_blocks / 100) { 194 acct_state = ACCT_ACTIVE; 195 log(LOG_NOTICE, "Accounting resumed\n"); 196 } 197 break; 198 case ACCT_ACTIVE: 199 if (bavail <= acctsuspend * sb->f_blocks / 100) { 200 acct_state = ACCT_SUSPENDED; 201 log(LOG_NOTICE, "Accounting suspended\n"); 202 } 203 break; 204 case ACCT_STOP: 205 break; 206 } 207 kmem_free(sb, sizeof(*sb)); 208 return (0); 209 } 210 211 static void 212 acct_stop(void) 213 { 214 int error; 215 216 KASSERT(rw_write_held(&acct_lock)); 217 218 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) { 219 error = vn_close(acct_vp, FWRITE, acct_cred); 220 #ifdef DIAGNOSTIC 221 if (error != 0) 222 printf("acct_stop: failed to close, errno = %d\n", 223 error); 224 #endif 225 acct_vp = NULLVP; 226 } 227 if (acct_cred != NULL) { 228 kauth_cred_free(acct_cred); 229 acct_cred = NULL; 230 } 231 acct_state = ACCT_STOP; 232 } 233 234 /* 235 * Periodically check the file system to see if accounting 236 * should be turned on or off. Beware the case where the vnode 237 * has been vgone()'d out from underneath us, e.g. when the file 238 * system containing the accounting file has been forcibly unmounted. 239 */ 240 static void 241 acctwatch(void *arg) 242 { 243 int error; 244 245 log(LOG_NOTICE, "Accounting started\n"); 246 rw_enter(&acct_lock, RW_WRITER); 247 while (acct_state != ACCT_STOP) { 248 if (acct_vp->v_type == VBAD) { 249 log(LOG_NOTICE, "Accounting terminated\n"); 250 acct_stop(); 251 continue; 252 } 253 254 error = acct_chkfree(); 255 #ifdef DIAGNOSTIC 256 if (error != 0) 257 printf("acctwatch: failed to statvfs, error = %d\n", 258 error); 259 #endif 260 rw_exit(&acct_lock); 261 error = kpause("actwat", false, acctchkfreq * hz, NULL); 262 rw_enter(&acct_lock, RW_WRITER); 263 #ifdef DIAGNOSTIC 264 if (error != 0 && error != EWOULDBLOCK) 265 printf("acctwatch: sleep error %d\n", error); 266 #endif 267 } 268 acct_dkwatcher = NULL; 269 rw_exit(&acct_lock); 270 271 kthread_exit(0); 272 } 273 274 void 275 acct_init(void) 276 { 277 278 acct_state = ACCT_STOP; 279 acct_vp = NULLVP; 280 acct_cred = NULL; 281 rw_init(&acct_lock); 282 } 283 284 /* 285 * Accounting system call. Written based on the specification and 286 * previous implementation done by Mark Tinguely. 287 */ 288 int 289 sys_acct(struct lwp *l, const struct sys_acct_args *uap, register_t *retval) 290 { 291 /* { 292 syscallarg(const char *) path; 293 } */ 294 struct nameidata nd; 295 int error; 296 297 /* Make sure that the caller is root. */ 298 if ((error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_ACCOUNTING, 299 0, NULL, NULL, NULL))) 300 return (error); 301 302 /* 303 * If accounting is to be started to a file, open that file for 304 * writing and make sure it's a 'normal'. 305 */ 306 if (SCARG(uap, path) != NULL) { 307 struct vattr va; 308 size_t pad; 309 NDINIT(&nd, LOOKUP, NOFOLLOW | TRYEMULROOT, UIO_USERSPACE, 310 SCARG(uap, path)); 311 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0) 312 return (error); 313 if (nd.ni_vp->v_type != VREG) { 314 VOP_UNLOCK(nd.ni_vp, 0); 315 error = EACCES; 316 goto bad; 317 } 318 if ((error = VOP_GETATTR(nd.ni_vp, &va, l->l_cred)) != 0) { 319 VOP_UNLOCK(nd.ni_vp, 0); 320 goto bad; 321 } 322 323 if ((pad = (va.va_size % sizeof(struct acct))) != 0) { 324 u_quad_t size = va.va_size - pad; 325 #ifdef DIAGNOSTIC 326 printf("Size of accounting file not a multiple of " 327 "%lu - incomplete record truncated\n", 328 (unsigned long)sizeof(struct acct)); 329 #endif 330 VATTR_NULL(&va); 331 va.va_size = size; 332 error = VOP_SETATTR(nd.ni_vp, &va, l->l_cred); 333 if (error != 0) { 334 VOP_UNLOCK(nd.ni_vp, 0); 335 goto bad; 336 } 337 } 338 VOP_UNLOCK(nd.ni_vp, 0); 339 } 340 341 rw_enter(&acct_lock, RW_WRITER); 342 343 /* 344 * If accounting was previously enabled, kill the old space-watcher, 345 * free credential for accounting file i/o, 346 * ... (and, if no new file was specified, leave). 347 */ 348 acct_stop(); 349 if (SCARG(uap, path) == NULL) 350 goto out; 351 352 /* 353 * Save the new accounting file vnode and credential, 354 * and schedule the new free space watcher. 355 */ 356 acct_state = ACCT_ACTIVE; 357 acct_vp = nd.ni_vp; 358 acct_cred = l->l_cred; 359 kauth_cred_hold(acct_cred); 360 361 error = acct_chkfree(); /* Initial guess. */ 362 if (error != 0) { 363 acct_stop(); 364 goto out; 365 } 366 367 if (acct_dkwatcher == NULL) { 368 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 369 acctwatch, NULL, &acct_dkwatcher, "acctwatch"); 370 if (error != 0) 371 acct_stop(); 372 } 373 374 out: 375 rw_exit(&acct_lock); 376 return (error); 377 bad: 378 vn_close(nd.ni_vp, FWRITE, l->l_cred); 379 return error; 380 } 381 382 /* 383 * Write out process accounting information, on process exit. 384 * Data to be written out is specified in Leffler, et al. 385 * and are enumerated below. (They're also noted in the system 386 * "acct.h" header file.) 387 */ 388 int 389 acct_process(struct lwp *l) 390 { 391 struct acct acct; 392 struct timeval ut, st, tmp; 393 struct rusage *r; 394 int t, error = 0; 395 struct rlimit orlim; 396 struct proc *p = l->l_proc; 397 398 if (acct_state != ACCT_ACTIVE) 399 return 0; 400 401 rw_enter(&acct_lock, RW_READER); 402 403 /* If accounting isn't enabled, don't bother */ 404 if (acct_state != ACCT_ACTIVE) 405 goto out; 406 407 /* 408 * Temporarily raise the file limit so that accounting can't 409 * be stopped by the user. 410 * 411 * XXX We should think about the CPU limit, too. 412 */ 413 lim_privatise(p, false); 414 orlim = p->p_rlimit[RLIMIT_FSIZE]; 415 /* Set current and max to avoid illegal values */ 416 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 417 p->p_rlimit[RLIMIT_FSIZE].rlim_max = 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 mutex_enter(p->p_lock); 428 calcru(p, &ut, &st, NULL, NULL); 429 mutex_exit(p->p_lock); 430 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 431 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 432 433 /* (3) The elapsed time the commmand ran (and its starting time) */ 434 acct.ac_btime = p->p_stats->p_start.tv_sec; 435 getmicrotime(&tmp); 436 timersub(&tmp, &p->p_stats->p_start, &tmp); 437 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 438 439 /* (4) The average amount of memory used */ 440 r = &p->p_stats->p_ru; 441 timeradd(&ut, &st, &tmp); 442 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 443 if (t) 444 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 445 else 446 acct.ac_mem = 0; 447 448 /* (5) The number of disk I/O operations done */ 449 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 450 451 /* (6) The UID and GID of the process */ 452 acct.ac_uid = kauth_cred_getuid(l->l_cred); 453 acct.ac_gid = kauth_cred_getgid(l->l_cred); 454 455 /* (7) The terminal from which the process was started */ 456 mutex_enter(proc_lock); 457 if ((p->p_lflag & PL_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 458 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev; 459 else 460 acct.ac_tty = NODEV; 461 mutex_exit(proc_lock); 462 463 /* (8) The boolean flags that tell how the process terminated, etc. */ 464 acct.ac_flag = p->p_acflag; 465 466 /* 467 * Now, just write the accounting information to the file. 468 */ 469 error = vn_rdwr(UIO_WRITE, acct_vp, (void *)&acct, 470 sizeof(acct), (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, 471 acct_cred, NULL, NULL); 472 if (error != 0) 473 log(LOG_ERR, "Accounting: write failed %d\n", error); 474 475 /* Restore limit - rather pointless since process is about to exit */ 476 p->p_rlimit[RLIMIT_FSIZE] = orlim; 477 478 out: 479 rw_exit(&acct_lock); 480 return (error); 481 } 482
Indexes created Sun Sep 21 18:09:52 GMT 2025