1 /*- 2 * Copyright (c) 1994 Christopher G. Demetriou 3 * Copyright (c) 1982, 1986, 1989, 1993 4 * The Regents of the University of California. All rights reserved. 5 * (c) UNIX System Laboratories, Inc. 6 * All or some portions of this file are derived from material licensed 7 * to the University of California by American Telephone and Telegraph 8 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 9 * the permission of UNIX System Laboratories, Inc. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 * 39 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93 40 */ 41 42 #include <sys/cdefs.h> 43 __FBSDID("$FreeBSD$"); 44 45 #include "opt_mac.h" 46 47 #include <sys/param.h> 48 #include <sys/systm.h> 49 #include <sys/lock.h> 50 #include <sys/mutex.h> 51 #include <sys/sysproto.h> 52 #include <sys/proc.h> 53 #include <sys/mac.h> 54 #include <sys/mount.h> 55 #include <sys/vnode.h> 56 #include <sys/fcntl.h> 57 #include <sys/syslog.h> 58 #include <sys/kernel.h> 59 #include <sys/sysent.h> 60 #include <sys/sysctl.h> 61 #include <sys/namei.h> 62 #include <sys/acct.h> 63 #include <sys/resourcevar.h> 64 #include <sys/tty.h> 65 66 /* 67 * The routines implemented in this file are described in: 68 * Leffler, et al.: The Design and Implementation of the 4.3BSD 69 * UNIX Operating System (Addison Welley, 1989) 70 * on pages 62-63. 71 * 72 * Arguably, to simplify accounting operations, this mechanism should 73 * be replaced by one in which an accounting log file (similar to /dev/klog) 74 * is read by a user process, etc. However, that has its own problems. 75 */ 76 77 /* 78 * Internal accounting functions. 79 * The former's operation is described in Leffler, et al., and the latter 80 * was provided by UCB with the 4.4BSD-Lite release 81 */ 82 static comp_t encode_comp_t(u_long, u_long); 83 static void acctwatch(void *); 84 85 /* 86 * Accounting callout used for periodic scheduling of acctwatch. 87 */ 88 static struct callout acctwatch_callout; 89 90 /* 91 * Accounting vnode pointer, saved vnode pointer, and flags for each. 92 */ 93 static struct vnode *acctp; 94 static struct ucred *acctcred; 95 static int acctflags; 96 static struct vnode *savacctp; 97 static struct ucred *savacctcred; 98 static int savacctflags; 99 100 static struct mtx acct_mtx; 101 MTX_SYSINIT(acct, &acct_mtx, "accounting", MTX_DEF); 102 103 /* 104 * Values associated with enabling and disabling accounting 105 */ 106 static int acctsuspend = 2; /* stop accounting when < 2% free space left */ 107 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW, 108 &acctsuspend, 0, "percentage of free disk space below which accounting stops"); 109 110 static int acctresume = 4; /* resume when free space risen to > 4% */ 111 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW, 112 &acctresume, 0, "percentage of free disk space above which accounting resumes"); 113 114 static int acctchkfreq = 15; /* frequency (in seconds) to check space */ 115 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW, 116 &acctchkfreq, 0, "frequency for checking the free space"); 117 118 /* 119 * Accounting system call. Written based on the specification and 120 * previous implementation done by Mark Tinguely. 121 * 122 * MPSAFE 123 */ 124 int 125 acct(td, uap) 126 struct thread *td; 127 struct acct_args /* { 128 char *path; 129 } */ *uap; 130 { 131 struct nameidata nd; 132 int error, flags; 133 134 /* Make sure that the caller is root. */ 135 error = suser(td); 136 if (error) 137 return (error); 138 139 mtx_lock(&Giant); 140 /* 141 * If accounting is to be started to a file, open that file for 142 * appending and make sure it's a 'normal'. 143 */ 144 if (uap->path != NULL) { 145 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td); 146 flags = FWRITE | O_APPEND; 147 error = vn_open(&nd, &flags, 0, -1); 148 if (error) 149 goto done2; 150 NDFREE(&nd, NDF_ONLY_PNBUF); 151 #ifdef MAC 152 error = mac_check_system_acct(td->td_ucred, nd.ni_vp); 153 if (error) { 154 vn_close(nd.ni_vp, flags, td->td_ucred, td); 155 goto done2; 156 } 157 #endif 158 VOP_UNLOCK(nd.ni_vp, 0, td); 159 if (nd.ni_vp->v_type != VREG) { 160 vn_close(nd.ni_vp, flags, td->td_ucred, td); 161 error = EACCES; 162 goto done2; 163 } 164 #ifdef MAC 165 } else { 166 error = mac_check_system_acct(td->td_ucred, NULL); 167 if (error) 168 goto done2; 169 #endif 170 } 171 172 /* 173 * If accounting was previously enabled, kill the old space-watcher, 174 * close the file, and (if no new file was specified, leave). 175 */ 176 177 /* 178 * XXX arr: Should not hold lock over vnode operation. 179 */ 180 181 mtx_lock(&acct_mtx); 182 if (acctp != NULLVP || savacctp != NULLVP) { 183 callout_stop(&acctwatch_callout); 184 error = vn_close((acctp != NULLVP ? acctp : savacctp), 185 (acctp != NULLVP ? acctflags : savacctflags), 186 (acctcred != NOCRED ? acctcred : savacctcred), td); 187 acctp = savacctp = NULLVP; 188 crfree(acctcred != NOCRED ? acctcred : savacctcred); 189 acctcred = savacctcred = NOCRED; 190 log(LOG_NOTICE, "Accounting disabled\n"); 191 } 192 if (uap->path == NULL) { 193 mtx_unlock(&acct_mtx); 194 goto done2; 195 } 196 197 /* 198 * Save the new accounting file vnode, and schedule the new 199 * free space watcher. 200 */ 201 acctp = nd.ni_vp; 202 acctcred = crhold(td->td_ucred); 203 acctflags = flags; 204 callout_init(&acctwatch_callout, 0); 205 mtx_unlock(&acct_mtx); 206 log(LOG_NOTICE, "Accounting enabled\n"); 207 acctwatch(NULL); 208 done2: 209 mtx_unlock(&Giant); 210 return (error); 211 } 212 213 /* 214 * Write out process accounting information, on process exit. 215 * Data to be written out is specified in Leffler, et al. 216 * and are enumerated below. (They're also noted in the system 217 * "acct.h" header file.) 218 */ 219 220 int 221 acct_process(td) 222 struct thread *td; 223 { 224 struct proc *p = td->td_proc; 225 struct acct acct; 226 struct rusage *r; 227 struct timeval ut, st, tmp; 228 int t, ret; 229 struct vnode *vp; 230 struct ucred *uc; 231 232 mtx_lock(&acct_mtx); 233 234 /* If accounting isn't enabled, don't bother */ 235 vp = acctp; 236 if (vp == NULLVP) { 237 mtx_unlock(&acct_mtx); 238 return (0); 239 } 240 241 /* 242 * Get process accounting information. 243 */ 244 245 PROC_LOCK(p); 246 /* (1) The name of the command that ran */ 247 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 248 249 /* (2) The amount of user and system time that was used */ 250 mtx_lock_spin(&sched_lock); 251 calcru(p, &ut, &st, NULL); 252 mtx_unlock_spin(&sched_lock); 253 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 254 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 255 256 /* (3) The elapsed time the command ran (and its starting time) */ 257 tmp = boottime; 258 timevaladd(&tmp, &p->p_stats->p_start); 259 acct.ac_btime = tmp.tv_sec; 260 microuptime(&tmp); 261 timevalsub(&tmp, &p->p_stats->p_start); 262 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 263 264 /* (4) The average amount of memory used */ 265 r = &p->p_stats->p_ru; 266 tmp = ut; 267 timevaladd(&tmp, &st); 268 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 269 if (t) 270 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 271 else 272 acct.ac_mem = 0; 273 274 /* (5) The number of disk I/O operations done */ 275 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 276 277 /* (6) The UID and GID of the process */ 278 acct.ac_uid = p->p_ucred->cr_ruid; 279 acct.ac_gid = p->p_ucred->cr_rgid; 280 281 /* (7) The terminal from which the process was started */ 282 SESS_LOCK(p->p_session); 283 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 284 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 285 else 286 acct.ac_tty = NOUDEV; 287 SESS_UNLOCK(p->p_session); 288 289 /* (8) The boolean flags that tell how the process terminated, etc. */ 290 acct.ac_flag = p->p_acflag; 291 PROC_UNLOCK(p); 292 293 /* 294 * Write the accounting information to the file. 295 */ 296 uc = crhold(acctcred); 297 vref(vp); 298 mtx_unlock(&acct_mtx); 299 300 /* 301 * Eliminate any file size rlimit. 302 */ 303 if (p->p_limit->p_refcnt > 1) { 304 p->p_limit->p_refcnt--; 305 p->p_limit = limcopy(p->p_limit); 306 } 307 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 308 309 VOP_LEASE(vp, td, uc, LEASE_WRITE); 310 ret = vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 311 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, uc, NOCRED, 312 (int *)0, td); 313 vrele(vp); 314 crfree(uc); 315 return (ret); 316 } 317 318 /* 319 * Encode_comp_t converts from ticks in seconds and microseconds 320 * to ticks in 1/AHZ seconds. The encoding is described in 321 * Leffler, et al., on page 63. 322 */ 323 324 #define MANTSIZE 13 /* 13 bit mantissa. */ 325 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 326 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 327 328 static comp_t 329 encode_comp_t(s, us) 330 u_long s, us; 331 { 332 int exp, rnd; 333 334 exp = 0; 335 rnd = 0; 336 s *= AHZ; 337 s += us / (1000000 / AHZ); /* Maximize precision. */ 338 339 while (s > MAXFRACT) { 340 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 341 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 342 exp++; 343 } 344 345 /* If we need to round up, do it (and handle overflow correctly). */ 346 if (rnd && (++s > MAXFRACT)) { 347 s >>= EXPSIZE; 348 exp++; 349 } 350 351 /* Clean it up and polish it off. */ 352 exp <<= MANTSIZE; /* Shift the exponent into place */ 353 exp += s; /* and add on the mantissa. */ 354 return (exp); 355 } 356 357 /* 358 * Periodically check the filesystem to see if accounting 359 * should be turned on or off. Beware the case where the vnode 360 * has been vgone()'d out from underneath us, e.g. when the file 361 * system containing the accounting file has been forcibly unmounted. 362 */ 363 /* ARGSUSED */ 364 static void 365 acctwatch(a) 366 void *a; 367 { 368 struct statfs sb; 369 370 mtx_lock(&acct_mtx); 371 372 /* 373 * XXX arr: Need to fix the issue of holding acct_mtx over 374 * the below vnode operations. 375 */ 376 377 if (savacctp != NULLVP) { 378 if (savacctp->v_type == VBAD) { 379 (void) vn_close(savacctp, savacctflags, savacctcred, 380 NULL); 381 savacctp = NULLVP; 382 savacctcred = NOCRED; 383 mtx_unlock(&acct_mtx); 384 return; 385 } 386 (void)VFS_STATFS(savacctp->v_mount, &sb, (struct thread *)0); 387 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 388 acctp = savacctp; 389 acctcred = savacctcred; 390 acctflags = savacctflags; 391 savacctp = NULLVP; 392 savacctcred = NOCRED; 393 log(LOG_NOTICE, "Accounting resumed\n"); 394 } 395 } else { 396 if (acctp == NULLVP) { 397 mtx_unlock(&acct_mtx); 398 return; 399 } 400 if (acctp->v_type == VBAD) { 401 (void) vn_close(acctp, acctflags, acctcred, NULL); 402 acctp = NULLVP; 403 crfree(acctcred); 404 acctcred = NOCRED; 405 mtx_unlock(&acct_mtx); 406 return; 407 } 408 (void)VFS_STATFS(acctp->v_mount, &sb, (struct thread *)0); 409 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 410 savacctp = acctp; 411 savacctflags = acctflags; 412 savacctcred = acctcred; 413 acctp = NULLVP; 414 acctcred = NOCRED; 415 log(LOG_NOTICE, "Accounting suspended\n"); 416 } 417 } 418 callout_reset(&acctwatch_callout, acctchkfreq * hz, acctwatch, NULL); 419 mtx_unlock(&acct_mtx); 420 } 421