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 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/sysproto.h> 46 #include <sys/proc.h> 47 #include <sys/priv.h> 48 #include <sys/mount.h> 49 #include <sys/vnode.h> 50 #include <sys/fcntl.h> 51 #include <sys/syslog.h> 52 #include <sys/kernel.h> 53 #include <sys/sysent.h> 54 #include <sys/sysctl.h> 55 #include <sys/nlookup.h> 56 #include <sys/acct.h> 57 #include <sys/resourcevar.h> 58 #include <sys/tty.h> 59 60 #include <vm/vm_zone.h> 61 62 #include <sys/mplock2.h> 63 64 /* 65 * The routines implemented in this file are described in: 66 * Leffler, et al.: The Design and Implementation of the 4.3BSD 67 * UNIX Operating System (Addison Welley, 1989) 68 * on pages 62-63. 69 * 70 * Arguably, to simplify accounting operations, this mechanism should 71 * be replaced by one in which an accounting log file (similar to /dev/klog) 72 * is read by a user process, etc. However, that has its own problems. 73 */ 74 75 /* 76 * Internal accounting functions. 77 * The former's operation is described in Leffler, et al., and the latter 78 * was provided by UCB with the 4.4BSD-Lite release 79 */ 80 static comp_t encode_comp_t (u_long, u_long); 81 static void acctwatch (void *); 82 83 /* 84 * Accounting callout handle used for periodic scheduling of 85 * acctwatch. 86 */ 87 static struct callout acctwatch_handle; 88 89 /* 90 * Accounting vnode pointer, and saved vnode pointer. 91 */ 92 static struct vnode *acctp; 93 static struct vnode *savacctp; 94 95 /* 96 * Values associated with enabling and disabling accounting 97 */ 98 static int acctsuspend = 2; /* stop accounting when < 2% free space left */ 99 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW, 100 &acctsuspend, 0, "Percentage of free disk space below which accounting will be suspended"); 101 102 static int acctresume = 4; /* resume when free space risen to > 4% */ 103 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW, 104 &acctresume, 0, "Minimum percentage of free disk space needed to resume accounting"); 105 106 static int acctchkfreq = 15; /* frequency (in seconds) to check space */ 107 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW, 108 &acctchkfreq, 0, "Frequency (in seconds) of free disk space checking"); 109 110 static void 111 acct_init(void *arg __unused) 112 { 113 callout_init(&acctwatch_handle); 114 } 115 SYSINIT(acct, SI_SUB_DRIVERS, SI_ORDER_ANY, acct_init, NULL); 116 117 /* 118 * Accounting system call. Written based on the specification and 119 * previous implementation done by Mark Tinguely. 120 * 121 * acct(char *path) 122 * 123 * MPALMOSTSAFE 124 */ 125 int 126 sys_acct(struct acct_args *uap) 127 { 128 struct thread *td = curthread; 129 struct nlookupdata nd; 130 struct vnode *vp; 131 int error; 132 133 error = priv_check(td, PRIV_ACCT); 134 if (error) 135 return (error); 136 137 get_mplock(); 138 139 /* 140 * If accounting is to be started to a file, open that file for 141 * appending and make sure it's a 'normal'. 142 */ 143 if (uap->path != NULL) { 144 error = nlookup_init(&nd, uap->path, UIO_USERSPACE, 145 NLC_LOCKVP); 146 if (error == 0) 147 error = vn_open(&nd, NULL, FWRITE | O_APPEND, 0); 148 if (error == 0 && nd.nl_open_vp->v_type != VREG) 149 error = EACCES; 150 if (error) { 151 nlookup_done(&nd); 152 goto done; 153 } 154 vp = nd.nl_open_vp; 155 nd.nl_open_vp = NULL; 156 nlookup_done(&nd); 157 158 vn_unlock(vp); 159 } else { 160 vp = NULL; 161 } 162 163 /* 164 * If accounting was previously enabled, kill the old space-watcher, 165 * close the file. 166 */ 167 if (acctp != NULLVP || savacctp != NULLVP) { 168 callout_stop(&acctwatch_handle); 169 error = vn_close((acctp != NULLVP ? acctp : savacctp), 170 FWRITE | O_APPEND); 171 acctp = savacctp = NULLVP; 172 } 173 174 /* 175 * If no new file opened then leave. We never did an nlookup so 176 * don't try cleaning it up. 177 */ 178 if (uap->path == NULL) 179 goto done; 180 181 /* 182 * Save the new accounting file vnode, and schedule the new 183 * free space watcher. 184 */ 185 acctp = vp; 186 acctwatch(NULL); 187 done: 188 rel_mplock(); 189 return (error); 190 } 191 192 /* 193 * Write out process accounting information, on process exit. 194 * Data to be written out is specified in Leffler, et al. 195 * and are enumerated below. (They're also noted in the system 196 * "acct.h" header file.) 197 */ 198 199 int 200 acct_process(struct proc *p) 201 { 202 struct acct acct; 203 struct rusage *r; 204 struct rusage ru; 205 struct timeval tmp; 206 struct rlimit rlim; 207 int t; 208 struct vnode *vp; 209 210 /* If accounting isn't enabled, don't bother */ 211 vp = acctp; 212 if (vp == NULLVP) 213 return (0); 214 215 /* 216 * Get process accounting information. 217 */ 218 219 /* (1) The name of the command that ran */ 220 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 221 222 /* (2) The amount of user and system time that was used */ 223 calcru_proc(p, &ru); 224 acct.ac_utime = encode_comp_t(ru.ru_utime.tv_sec, ru.ru_utime.tv_usec); 225 acct.ac_stime = encode_comp_t(ru.ru_stime.tv_sec, ru.ru_stime.tv_usec); 226 227 /* (3) The elapsed time the commmand ran (and its starting time) */ 228 acct.ac_btime = p->p_start.tv_sec; 229 microtime(&tmp); 230 timevalsub(&tmp, &p->p_start); 231 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 232 233 /* (4) The average amount of memory used */ 234 r = &p->p_ru; 235 tmp = ru.ru_utime; 236 timevaladd(&tmp, &ru.ru_stime); 237 t = tmp.tv_sec * hz + tmp.tv_usec / ustick; 238 if (t) 239 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 240 else 241 acct.ac_mem = 0; 242 243 /* (5) The number of disk I/O operations done */ 244 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 245 246 /* (6) The UID and GID of the process */ 247 acct.ac_uid = p->p_ucred->cr_ruid; 248 acct.ac_gid = p->p_ucred->cr_rgid; 249 250 /* (7) The terminal from which the process was started */ 251 if ((p->p_flags & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 252 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 253 else 254 acct.ac_tty = NOUDEV; 255 256 /* (8) The boolean flags that tell how the process terminated, etc. */ 257 acct.ac_flag = p->p_acflag; 258 259 /* 260 * Eliminate any file size rlimit. 261 */ 262 rlim.rlim_cur = RLIM_INFINITY; 263 rlim.rlim_max = RLIM_INFINITY; 264 plimit_modify(p, RLIMIT_FSIZE, &rlim); 265 266 /* 267 * Write the accounting information to the file. 268 */ 269 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 270 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred, 271 NULL)); 272 } 273 274 /* 275 * Encode_comp_t converts from ticks in seconds and microseconds 276 * to ticks in 1/AHZ seconds. The encoding is described in 277 * Leffler, et al., on page 63. 278 */ 279 280 #define MANTSIZE 13 /* 13 bit mantissa. */ 281 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 282 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 283 284 static comp_t 285 encode_comp_t(u_long s, u_long us) 286 { 287 int exp, rnd; 288 289 exp = 0; 290 rnd = 0; 291 s *= AHZ; 292 s += us / (1000000 / AHZ); /* Maximize precision. */ 293 294 while (s > MAXFRACT) { 295 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 296 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 297 exp++; 298 } 299 300 /* If we need to round up, do it (and handle overflow correctly). */ 301 if (rnd && (++s > MAXFRACT)) { 302 s >>= EXPSIZE; 303 exp++; 304 } 305 306 /* Clean it up and polish it off. */ 307 exp <<= MANTSIZE; /* Shift the exponent into place */ 308 exp += s; /* and add on the mantissa. */ 309 return (exp); 310 } 311 312 /* 313 * Periodically check the file system to see if accounting 314 * should be turned on or off. Beware the case where the vnode 315 * has been vgone()'d out from underneath us, e.g. when the file 316 * system containing the accounting file has been forcibly unmounted. 317 */ 318 /* ARGSUSED */ 319 static void 320 acctwatch(void *a) 321 { 322 struct statfs sb; 323 324 if (savacctp != NULLVP) { 325 if (savacctp->v_type == VBAD) { 326 (void) vn_close(savacctp, FWRITE | O_APPEND); 327 savacctp = NULLVP; 328 return; 329 } 330 (void)VFS_STATFS(savacctp->v_mount, &sb, proc0.p_ucred); 331 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 332 acctp = savacctp; 333 savacctp = NULLVP; 334 log(LOG_NOTICE, "Accounting resumed\n"); 335 } 336 } else { 337 if (acctp == NULLVP) 338 return; 339 if (acctp->v_type == VBAD) { 340 (void) vn_close(acctp, FWRITE | O_APPEND); 341 acctp = NULLVP; 342 return; 343 } 344 (void)VFS_STATFS(acctp->v_mount, &sb, proc0.p_ucred); 345 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 346 savacctp = acctp; 347 acctp = NULLVP; 348 log(LOG_NOTICE, "Accounting suspended\n"); 349 } 350 } 351 callout_reset(&acctwatch_handle, acctchkfreq * hz, acctwatch, NULL); 352 } 353