1 /* $NetBSD: kern_acct.c,v 1.93 2011/09/03 14:09:03 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 1.93 2011/09/03 14:09:03 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/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 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 fsblkcnt_t bavail; 178 179 sb = kmem_alloc(sizeof(*sb), KM_SLEEP); 180 if (sb == NULL) 181 return (ENOMEM); 182 error = VFS_STATVFS(acct_vp->v_mount, sb); 183 if (error != 0) { 184 kmem_free(sb, sizeof(*sb)); 185 return (error); 186 } 187 188 if (sb->f_bfree < sb->f_bresvd) { 189 bavail = 0; 190 } else { 191 bavail = sb->f_bfree - sb->f_bresvd; 192 } 193 194 switch (acct_state) { 195 case ACCT_SUSPENDED: 196 if (bavail > acctresume * sb->f_blocks / 100) { 197 acct_state = ACCT_ACTIVE; 198 log(LOG_NOTICE, "Accounting resumed\n"); 199 } 200 break; 201 case ACCT_ACTIVE: 202 if (bavail <= acctsuspend * sb->f_blocks / 100) { 203 acct_state = ACCT_SUSPENDED; 204 log(LOG_NOTICE, "Accounting suspended\n"); 205 } 206 break; 207 case ACCT_STOP: 208 break; 209 } 210 kmem_free(sb, sizeof(*sb)); 211 return (0); 212 } 213 214 static void 215 acct_stop(void) 216 { 217 int error; 218 219 KASSERT(rw_write_held(&acct_lock)); 220 221 if (acct_vp != NULLVP && acct_vp->v_type != VBAD) { 222 error = vn_close(acct_vp, FWRITE, acct_cred); 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) 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 #endif 263 rw_exit(&acct_lock); 264 error = kpause("actwat", false, acctchkfreq * hz, NULL); 265 rw_enter(&acct_lock, RW_WRITER); 266 #ifdef DIAGNOSTIC 267 if (error != 0 && error != EWOULDBLOCK) 268 printf("acctwatch: sleep error %d\n", error); 269 #endif 270 } 271 acct_dkwatcher = NULL; 272 rw_exit(&acct_lock); 273 274 kthread_exit(0); 275 } 276 277 void 278 acct_init(void) 279 { 280 281 acct_state = ACCT_STOP; 282 acct_vp = NULLVP; 283 acct_cred = NULL; 284 rw_init(&acct_lock); 285 } 286 287 /* 288 * Accounting system call. Written based on the specification and 289 * previous implementation done by Mark Tinguely. 290 */ 291 int 292 sys_acct(struct lwp *l, const struct sys_acct_args *uap, register_t *retval) 293 { 294 /* { 295 syscallarg(const char *) path; 296 } */ 297 struct pathbuf *pb; 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; 313 314 error = pathbuf_copyin(SCARG(uap, path), &pb); 315 if (error) { 316 return error; 317 } 318 NDINIT(&nd, LOOKUP, FOLLOW | TRYEMULROOT, pb); 319 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0) { 320 pathbuf_destroy(pb); 321 return error; 322 } 323 if (nd.ni_vp->v_type != VREG) { 324 VOP_UNLOCK(nd.ni_vp); 325 error = EACCES; 326 goto bad; 327 } 328 if ((error = VOP_GETATTR(nd.ni_vp, &va, l->l_cred)) != 0) { 329 VOP_UNLOCK(nd.ni_vp); 330 goto bad; 331 } 332 333 if ((pad = (va.va_size % sizeof(struct acct))) != 0) { 334 u_quad_t size = va.va_size - pad; 335 #ifdef DIAGNOSTIC 336 printf("Size of accounting file not a multiple of " 337 "%lu - incomplete record truncated\n", 338 (unsigned long)sizeof(struct acct)); 339 #endif 340 vattr_null(&va); 341 va.va_size = size; 342 error = VOP_SETATTR(nd.ni_vp, &va, l->l_cred); 343 if (error != 0) { 344 VOP_UNLOCK(nd.ni_vp); 345 goto bad; 346 } 347 } 348 VOP_UNLOCK(nd.ni_vp); 349 } 350 351 rw_enter(&acct_lock, RW_WRITER); 352 353 /* 354 * If accounting was previously enabled, kill the old space-watcher, 355 * free credential for accounting file i/o, 356 * ... (and, if no new file was specified, leave). 357 */ 358 acct_stop(); 359 if (SCARG(uap, path) == NULL) 360 goto out; 361 362 /* 363 * Save the new accounting file vnode and credential, 364 * and schedule the new free space watcher. 365 */ 366 acct_state = ACCT_ACTIVE; 367 acct_vp = nd.ni_vp; 368 acct_cred = l->l_cred; 369 kauth_cred_hold(acct_cred); 370 371 pathbuf_destroy(pb); 372 373 error = acct_chkfree(); /* Initial guess. */ 374 if (error != 0) { 375 acct_stop(); 376 goto out; 377 } 378 379 if (acct_dkwatcher == NULL) { 380 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 381 acctwatch, NULL, &acct_dkwatcher, "acctwatch"); 382 if (error != 0) 383 acct_stop(); 384 } 385 386 out: 387 rw_exit(&acct_lock); 388 return (error); 389 bad: 390 vn_close(nd.ni_vp, FWRITE, l->l_cred); 391 pathbuf_destroy(pb); 392 return error; 393 } 394 395 /* 396 * Write out process accounting information, on process exit. 397 * Data to be written out is specified in Leffler, et al. 398 * and are enumerated below. (They're also noted in the system 399 * "acct.h" header file.) 400 */ 401 int 402 acct_process(struct lwp *l) 403 { 404 struct acct acct; 405 struct timeval ut, st, tmp; 406 struct rusage *r; 407 int t, error = 0; 408 struct rlimit orlim; 409 struct proc *p = l->l_proc; 410 411 if (acct_state != ACCT_ACTIVE) 412 return 0; 413 414 memset(&acct, 0, sizeof(acct)); /* to zerofill padded data */ 415 416 rw_enter(&acct_lock, RW_READER); 417 418 /* If accounting isn't enabled, don't bother */ 419 if (acct_state != ACCT_ACTIVE) 420 goto out; 421 422 /* 423 * Temporarily raise the file limit so that accounting can't 424 * be stopped by the user. 425 * 426 * XXX We should think about the CPU limit, too. 427 */ 428 lim_privatise(p); 429 orlim = p->p_rlimit[RLIMIT_FSIZE]; 430 /* Set current and max to avoid illegal values */ 431 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 432 p->p_rlimit[RLIMIT_FSIZE].rlim_max = RLIM_INFINITY; 433 434 /* 435 * Get process accounting information. 436 */ 437 438 /* (1) The name of the command that ran */ 439 strncpy(acct.ac_comm, p->p_comm, sizeof(acct.ac_comm)); 440 441 /* (2) The amount of user and system time that was used */ 442 mutex_enter(p->p_lock); 443 calcru(p, &ut, &st, NULL, NULL); 444 mutex_exit(p->p_lock); 445 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 446 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 447 448 /* (3) The elapsed time the commmand ran (and its starting time) */ 449 acct.ac_btime = p->p_stats->p_start.tv_sec; 450 getmicrotime(&tmp); 451 timersub(&tmp, &p->p_stats->p_start, &tmp); 452 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 453 454 /* (4) The average amount of memory used */ 455 r = &p->p_stats->p_ru; 456 timeradd(&ut, &st, &tmp); 457 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 458 if (t) 459 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 460 else 461 acct.ac_mem = 0; 462 463 /* (5) The number of disk I/O operations done */ 464 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 465 466 /* (6) The UID and GID of the process */ 467 acct.ac_uid = kauth_cred_getuid(l->l_cred); 468 acct.ac_gid = kauth_cred_getgid(l->l_cred); 469 470 /* (7) The terminal from which the process was started */ 471 mutex_enter(proc_lock); 472 if ((p->p_lflag & PL_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 473 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev; 474 else 475 acct.ac_tty = NODEV; 476 mutex_exit(proc_lock); 477 478 /* (8) The boolean flags that tell how the process terminated, etc. */ 479 acct.ac_flag = p->p_acflag; 480 481 /* 482 * Now, just write the accounting information to the file. 483 */ 484 error = vn_rdwr(UIO_WRITE, acct_vp, (void *)&acct, 485 sizeof(acct), (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, 486 acct_cred, NULL, NULL); 487 if (error != 0) 488 log(LOG_ERR, "Accounting: write failed %d\n", error); 489 490 /* Restore limit - rather pointless since process is about to exit */ 491 p->p_rlimit[RLIMIT_FSIZE] = orlim; 492 493 out: 494 rw_exit(&acct_lock); 495 return (error); 496 } 497