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 * $FreeBSD: src/sys/kern/kern_acct.c,v 1.23.2.1 2002/07/24 18:33:55 johan Exp $ 41 * $DragonFly: src/sys/kern/kern_acct.c,v 1.28 2008/05/08 01:26:00 dillon Exp $ 42 */ 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/sysproto.h> 47 #include <sys/proc.h> 48 #include <sys/priv.h> 49 #include <sys/mount.h> 50 #include <sys/vnode.h> 51 #include <sys/fcntl.h> 52 #include <sys/syslog.h> 53 #include <sys/kernel.h> 54 #include <sys/sysent.h> 55 #include <sys/sysctl.h> 56 #include <sys/nlookup.h> 57 #include <sys/acct.h> 58 #include <sys/resourcevar.h> 59 #include <sys/tty.h> 60 61 #include <vm/vm_zone.h> 62 63 /* 64 * The routines implemented in this file are described in: 65 * Leffler, et al.: The Design and Implementation of the 4.3BSD 66 * UNIX Operating System (Addison Welley, 1989) 67 * on pages 62-63. 68 * 69 * Arguably, to simplify accounting operations, this mechanism should 70 * be replaced by one in which an accounting log file (similar to /dev/klog) 71 * is read by a user process, etc. However, that has its own problems. 72 */ 73 74 /* 75 * Internal accounting functions. 76 * The former's operation is described in Leffler, et al., and the latter 77 * was provided by UCB with the 4.4BSD-Lite release 78 */ 79 static comp_t encode_comp_t (u_long, u_long); 80 static void acctwatch (void *); 81 82 /* 83 * Accounting callout handle used for periodic scheduling of 84 * acctwatch. 85 */ 86 static struct callout acctwatch_handle; 87 88 /* 89 * Accounting vnode pointer, and saved vnode pointer. 90 */ 91 static struct vnode *acctp; 92 static struct vnode *savacctp; 93 94 /* 95 * Values associated with enabling and disabling accounting 96 */ 97 static int acctsuspend = 2; /* stop accounting when < 2% free space left */ 98 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW, 99 &acctsuspend, 0, "Percentage of free disk space below which accounting will be suspended"); 100 101 static int acctresume = 4; /* resume when free space risen to > 4% */ 102 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW, 103 &acctresume, 0, "Minimum percentage of free disk space needed to resume accounting"); 104 105 static int acctchkfreq = 15; /* frequency (in seconds) to check space */ 106 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW, 107 &acctchkfreq, 0, "Frequency (in seconds) of free disk space checking"); 108 109 static void 110 acct_init(void *arg __unused) 111 { 112 callout_init(&acctwatch_handle); 113 } 114 SYSINIT(acct, SI_SUB_DRIVERS, SI_ORDER_ANY, acct_init, NULL); 115 116 /* 117 * Accounting system call. Written based on the specification and 118 * previous implementation done by Mark Tinguely. 119 * 120 * acct(char *path) 121 */ 122 int 123 sys_acct(struct acct_args *uap) 124 { 125 struct thread *td = curthread; 126 struct nlookupdata nd; 127 struct vnode *vp; 128 int error; 129 130 error = priv_check(td, PRIV_ACCT); 131 if (error) 132 return (error); 133 134 /* 135 * If accounting is to be started to a file, open that file for 136 * appending and make sure it's a 'normal'. 137 */ 138 if (uap->path != NULL) { 139 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 140 NLC_LOCKVP); 141 if (error == 0) 142 error = vn_open(&nd, NULL, FWRITE | O_APPEND, 0); 143 if (error == 0 && nd.nl_open_vp->v_type != VREG) 144 error = EACCES; 145 if (error) { 146 nlookup_done(&nd); 147 return (error); 148 } 149 vp = nd.nl_open_vp; 150 nd.nl_open_vp = NULL; 151 nlookup_done(&nd); 152 153 vn_unlock(vp); 154 } else { 155 vp = NULL; 156 } 157 158 /* 159 * If accounting was previously enabled, kill the old space-watcher, 160 * close the file. 161 */ 162 if (acctp != NULLVP || savacctp != NULLVP) { 163 callout_stop(&acctwatch_handle); 164 error = vn_close((acctp != NULLVP ? acctp : savacctp), 165 FWRITE | O_APPEND); 166 acctp = savacctp = NULLVP; 167 } 168 169 /* 170 * If no new file opened then leave. We never did an nlookup so 171 * don't try cleaning it up. 172 */ 173 if (uap->path == NULL) 174 return (error); 175 176 /* 177 * Save the new accounting file vnode, and schedule the new 178 * free space watcher. 179 */ 180 acctp = vp; 181 acctwatch(NULL); 182 return (error); 183 } 184 185 /* 186 * Write out process accounting information, on process exit. 187 * Data to be written out is specified in Leffler, et al. 188 * and are enumerated below. (They're also noted in the system 189 * "acct.h" header file.) 190 */ 191 192 int 193 acct_process(struct proc *p) 194 { 195 struct acct acct; 196 struct rusage *r; 197 struct rusage ru; 198 struct timeval tmp; 199 struct rlimit rlim; 200 int t; 201 struct vnode *vp; 202 203 /* If accounting isn't enabled, don't bother */ 204 vp = acctp; 205 if (vp == NULLVP) 206 return (0); 207 208 /* 209 * Get process accounting information. 210 */ 211 212 /* (1) The name of the command that ran */ 213 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 214 215 /* (2) The amount of user and system time that was used */ 216 calcru_proc(p, &ru); 217 acct.ac_utime = encode_comp_t(ru.ru_utime.tv_sec, ru.ru_utime.tv_usec); 218 acct.ac_stime = encode_comp_t(ru.ru_stime.tv_sec, ru.ru_stime.tv_usec); 219 220 /* (3) The elapsed time the commmand ran (and its starting time) */ 221 acct.ac_btime = p->p_start.tv_sec; 222 microtime(&tmp); 223 timevalsub(&tmp, &p->p_start); 224 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 225 226 /* (4) The average amount of memory used */ 227 r = &p->p_ru; 228 tmp = ru.ru_utime;; 229 timevaladd(&tmp, &ru.ru_stime); 230 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 231 if (t) 232 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 233 else 234 acct.ac_mem = 0; 235 236 /* (5) The number of disk I/O operations done */ 237 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 238 239 /* (6) The UID and GID of the process */ 240 acct.ac_uid = p->p_ucred->cr_ruid; 241 acct.ac_gid = p->p_ucred->cr_rgid; 242 243 /* (7) The terminal from which the process was started */ 244 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 245 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 246 else 247 acct.ac_tty = NOUDEV; 248 249 /* (8) The boolean flags that tell how the process terminated, etc. */ 250 acct.ac_flag = p->p_acflag; 251 252 /* 253 * Eliminate any file size rlimit. 254 */ 255 rlim.rlim_cur = RLIM_INFINITY; 256 rlim.rlim_max = RLIM_INFINITY; 257 plimit_modify(p, RLIMIT_FSIZE, &rlim); 258 259 /* 260 * Write the accounting information to the file. 261 */ 262 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 263 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred, 264 NULL)); 265 } 266 267 /* 268 * Encode_comp_t converts from ticks in seconds and microseconds 269 * to ticks in 1/AHZ seconds. The encoding is described in 270 * Leffler, et al., on page 63. 271 */ 272 273 #define MANTSIZE 13 /* 13 bit mantissa. */ 274 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 275 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 276 277 static comp_t 278 encode_comp_t(u_long s, u_long us) 279 { 280 int exp, rnd; 281 282 exp = 0; 283 rnd = 0; 284 s *= AHZ; 285 s += us / (1000000 / AHZ); /* Maximize precision. */ 286 287 while (s > MAXFRACT) { 288 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 289 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 290 exp++; 291 } 292 293 /* If we need to round up, do it (and handle overflow correctly). */ 294 if (rnd && (++s > MAXFRACT)) { 295 s >>= EXPSIZE; 296 exp++; 297 } 298 299 /* Clean it up and polish it off. */ 300 exp <<= MANTSIZE; /* Shift the exponent into place */ 301 exp += s; /* and add on the mantissa. */ 302 return (exp); 303 } 304 305 /* 306 * Periodically check the file system to see if accounting 307 * should be turned on or off. Beware the case where the vnode 308 * has been vgone()'d out from underneath us, e.g. when the file 309 * system containing the accounting file has been forcibly unmounted. 310 */ 311 /* ARGSUSED */ 312 static void 313 acctwatch(void *a) 314 { 315 struct statfs sb; 316 317 if (savacctp != NULLVP) { 318 if (savacctp->v_type == VBAD) { 319 (void) vn_close(savacctp, FWRITE | O_APPEND); 320 savacctp = NULLVP; 321 return; 322 } 323 (void)VFS_STATFS(savacctp->v_mount, &sb, proc0.p_ucred); 324 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 325 acctp = savacctp; 326 savacctp = NULLVP; 327 log(LOG_NOTICE, "Accounting resumed\n"); 328 } 329 } else { 330 if (acctp == NULLVP) 331 return; 332 if (acctp->v_type == VBAD) { 333 (void) vn_close(acctp, FWRITE | O_APPEND); 334 acctp = NULLVP; 335 return; 336 } 337 (void)VFS_STATFS(acctp->v_mount, &sb, proc0.p_ucred); 338 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 339 savacctp = acctp; 340 acctp = NULLVP; 341 log(LOG_NOTICE, "Accounting suspended\n"); 342 } 343 } 344 callout_reset(&acctwatch_handle, acctchkfreq * hz, acctwatch, NULL); 345 } 346