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. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93 36 * $FreeBSD: src/sys/kern/kern_acct.c,v 1.23.2.1 2002/07/24 18:33:55 johan Exp $ 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/sysmsg.h> 42 #include <sys/proc.h> 43 #include <sys/priv.h> 44 #include <sys/mount.h> 45 #include <sys/vnode.h> 46 #include <sys/fcntl.h> 47 #include <sys/syslog.h> 48 #include <sys/kernel.h> 49 #include <sys/sysent.h> 50 #include <sys/sysctl.h> 51 #include <sys/nlookup.h> 52 #include <sys/acct.h> 53 #include <sys/resourcevar.h> 54 #include <sys/tty.h> 55 56 #include <vm/vm_zone.h> 57 58 static struct lock acct_lock = LOCK_INITIALIZER("acct_lock", 0, 0); 59 60 /* 61 * The routines implemented in this file are described in: 62 * Leffler, et al.: The Design and Implementation of the 4.3BSD 63 * UNIX Operating System (Addison Welley, 1989) 64 * on pages 62-63. 65 * 66 * Arguably, to simplify accounting operations, this mechanism should 67 * be replaced by one in which an accounting log file (similar to /dev/klog) 68 * is read by a user process, etc. However, that has its own problems. 69 */ 70 71 /* 72 * Internal accounting functions. 73 * The former's operation is described in Leffler, et al., and the latter 74 * was provided by UCB with the 4.4BSD-Lite release 75 */ 76 static comp_t encode_comp_t (u_long, u_long); 77 static void acctwatch (void *); 78 79 /* 80 * Accounting callout handle used for periodic scheduling of 81 * acctwatch. 82 */ 83 static struct callout acctwatch_handle; 84 85 /* 86 * Accounting vnode pointer, and saved vnode pointer. 87 */ 88 static struct vnode *acctp; 89 static struct vnode *savacctp; 90 91 /* 92 * Values associated with enabling and disabling accounting 93 */ 94 static int acctsuspend = 2; /* stop accounting when < 2% free space left */ 95 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW, 96 &acctsuspend, 0, "Percentage of free disk space below which accounting will be suspended"); 97 98 static int acctresume = 4; /* resume when free space risen to > 4% */ 99 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW, 100 &acctresume, 0, "Minimum percentage of free disk space needed to resume accounting"); 101 102 static int acctchkfreq = 15; /* frequency (in seconds) to check space */ 103 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW, 104 &acctchkfreq, 0, "Frequency (in seconds) of free disk space checking"); 105 106 static void 107 acct_init(void *arg __unused) 108 { 109 callout_init_lk(&acctwatch_handle, &acct_lock); 110 } 111 SYSINIT(acct, SI_SUB_DRIVERS, SI_ORDER_ANY, acct_init, NULL); 112 113 /* 114 * Accounting system call. Written based on the specification and 115 * previous implementation done by Mark Tinguely. 116 * 117 * acct(char *path) 118 * 119 * MPALMOSTSAFE 120 */ 121 int 122 sys_acct(struct sysmsg *sysmsg, const struct acct_args *uap) 123 { 124 struct thread *td = curthread; 125 struct nlookupdata nd; 126 struct vnode *vp; 127 int error; 128 129 error = priv_check(td, PRIV_ACCT); 130 if (error) 131 return (error); 132 133 lockmgr(&acct_lock, LK_EXCLUSIVE); 134 135 /* 136 * If accounting is to be started to a file, open that file for 137 * appending and make sure it's a 'normal'. 138 */ 139 if (uap->path != NULL) { 140 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 141 NLC_LOCKVP); 142 if (error == 0) 143 error = vn_open(&nd, NULL, FWRITE | O_APPEND, 0); 144 if (error == 0 && nd.nl_open_vp->v_type != VREG) 145 error = EACCES; 146 if (error) { 147 nlookup_done(&nd); 148 goto done; 149 } 150 vp = nd.nl_open_vp; 151 nd.nl_open_vp = NULL; 152 nlookup_done(&nd); 153 154 vn_unlock(vp); 155 } else { 156 vp = NULL; 157 } 158 159 /* 160 * If accounting was previously enabled, kill the old space-watcher, 161 * close the file. 162 */ 163 if (acctp != NULLVP || savacctp != NULLVP) { 164 callout_stop(&acctwatch_handle); 165 error = vn_close((acctp != NULLVP ? acctp : savacctp), 166 FWRITE | O_APPEND, NULL); 167 acctp = savacctp = NULLVP; 168 } 169 170 /* 171 * If no new file opened then leave. We never did an nlookup so 172 * don't try cleaning it up. 173 */ 174 if (uap->path == NULL) 175 goto done; 176 177 /* 178 * Save the new accounting file vnode, and schedule the new 179 * free space watcher. 180 */ 181 acctp = vp; 182 acctwatch(NULL); 183 done: 184 lockmgr(&acct_lock, LK_RELEASE); 185 186 return (error); 187 } 188 189 /* 190 * Write out process accounting information, on process exit. 191 * Data to be written out is specified in Leffler, et al. 192 * and are enumerated below. (They're also noted in the system 193 * "acct.h" header file.) 194 */ 195 196 int 197 acct_process(struct proc *p) 198 { 199 struct acct acct; 200 struct rusage *r; 201 struct rusage ru; 202 struct timeval tmp; 203 struct rlimit rlim; 204 int t; 205 int error; 206 struct vnode *vp; 207 208 /* 209 * If accounting isn't enabled, don't bother. Lock acct_lock 210 * make sure. 211 */ 212 if (acctp == NULLVP) 213 return 0; 214 lockmgr(&acct_lock, LK_SHARED); 215 vp = acctp; 216 if (vp == NULLVP) { 217 lockmgr(&acct_lock, LK_RELEASE); 218 return 0; 219 } 220 221 /* 222 * Get process accounting information. 223 */ 224 225 /* (1) The name of the command that ran */ 226 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 227 228 /* (2) The amount of user and system time that was used */ 229 calcru_proc(p, &ru); 230 acct.ac_utime = encode_comp_t(ru.ru_utime.tv_sec, ru.ru_utime.tv_usec); 231 acct.ac_stime = encode_comp_t(ru.ru_stime.tv_sec, ru.ru_stime.tv_usec); 232 233 /* (3) The elapsed time the commmand ran (and its starting time) */ 234 acct.ac_btime = p->p_start.tv_sec; 235 microtime(&tmp); 236 timevalsub(&tmp, &p->p_start); 237 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 238 239 /* (4) The average amount of memory used */ 240 r = &p->p_ru; 241 tmp = ru.ru_utime; 242 timevaladd(&tmp, &ru.ru_stime); 243 t = tmp.tv_sec * hz + tmp.tv_usec / ustick; 244 if (t) 245 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 246 else 247 acct.ac_mem = 0; 248 249 /* (5) The number of disk I/O operations done */ 250 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 251 252 /* (6) The UID and GID of the process */ 253 acct.ac_uid = p->p_ucred->cr_ruid; 254 acct.ac_gid = p->p_ucred->cr_rgid; 255 256 /* (7) The terminal from which the process was started */ 257 if ((p->p_flags & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) { 258 acct.ac_tty = 259 devid_from_dev(p->p_pgrp->pg_session->s_ttyp->t_dev); 260 } else { 261 acct.ac_tty = NOUDEV; 262 } 263 264 /* (8) The boolean flags that tell how the process terminated, etc. */ 265 acct.ac_flag = p->p_acflag; 266 267 /* 268 * Eliminate any file size rlimit. 269 */ 270 rlim.rlim_cur = RLIM_INFINITY; 271 rlim.rlim_max = RLIM_INFINITY; 272 plimit_modify(p, RLIMIT_FSIZE, &rlim); 273 274 /* 275 * Write the accounting information to the file. 276 */ 277 error = vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 278 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred, 279 NULL); 280 lockmgr(&acct_lock, LK_RELEASE); 281 282 return error; 283 } 284 285 /* 286 * Encode_comp_t converts from ticks in seconds and microseconds 287 * to ticks in 1/AHZ seconds. The encoding is described in 288 * Leffler, et al., on page 63. 289 */ 290 291 #define MANTSIZE 13 /* 13 bit mantissa. */ 292 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 293 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 294 295 static comp_t 296 encode_comp_t(u_long s, u_long us) 297 { 298 int exp, rnd; 299 300 exp = 0; 301 rnd = 0; 302 s *= AHZ; 303 s += us / (1000000 / AHZ); /* Maximize precision. */ 304 305 while (s > MAXFRACT) { 306 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 307 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 308 exp++; 309 } 310 311 /* If we need to round up, do it (and handle overflow correctly). */ 312 if (rnd && (++s > MAXFRACT)) { 313 s >>= EXPSIZE; 314 exp++; 315 } 316 317 /* Clean it up and polish it off. */ 318 exp <<= MANTSIZE; /* Shift the exponent into place */ 319 exp += s; /* and add on the mantissa. */ 320 return (exp); 321 } 322 323 /* 324 * Periodically check the file system to see if accounting 325 * should be turned on or off. Beware the case where the vnode 326 * has been vgone()'d out from underneath us, e.g. when the file 327 * system containing the accounting file has been forcibly unmounted. 328 */ 329 /* ARGSUSED */ 330 static void 331 acctwatch(void *a) 332 { 333 struct statfs sb; 334 335 if (savacctp != NULLVP) { 336 if (savacctp->v_type == VBAD) { 337 (void) vn_close(savacctp, FWRITE | O_APPEND, NULL); 338 savacctp = NULLVP; 339 return; 340 } 341 (void)VFS_STATFS(savacctp->v_mount, &sb, proc0.p_ucred); 342 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 343 acctp = savacctp; 344 savacctp = NULLVP; 345 log(LOG_NOTICE, "Accounting resumed\n"); 346 } 347 } else { 348 if (acctp == NULLVP) 349 return; 350 if (acctp->v_type == VBAD) { 351 (void) vn_close(acctp, FWRITE | O_APPEND, NULL); 352 acctp = NULLVP; 353 return; 354 } 355 (void)VFS_STATFS(acctp->v_mount, &sb, proc0.p_ucred); 356 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 357 savacctp = acctp; 358 acctp = NULLVP; 359 log(LOG_NOTICE, "Accounting suspended\n"); 360 } 361 } 362 callout_reset(&acctwatch_handle, acctchkfreq * hz, acctwatch, NULL); 363 } 364