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 1.9 2004/03/01 06:33:17 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/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/namei.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 /* 63 * The routines implemented in this file are described in: 64 * Leffler, et al.: The Design and Implementation of the 4.3BSD 65 * UNIX Operating System (Addison Welley, 1989) 66 * on pages 62-63. 67 * 68 * Arguably, to simplify accounting operations, this mechanism should 69 * be replaced by one in which an accounting log file (similar to /dev/klog) 70 * is read by a user process, etc. However, that has its own problems. 71 */ 72 73 /* 74 * Internal accounting functions. 75 * The former's operation is described in Leffler, et al., and the latter 76 * was provided by UCB with the 4.4BSD-Lite release 77 */ 78 static comp_t encode_comp_t (u_long, u_long); 79 static void acctwatch (void *); 80 81 /* 82 * Accounting callout handle used for periodic scheduling of 83 * acctwatch. 84 */ 85 static struct callout_handle acctwatch_handle 86 = CALLOUT_HANDLE_INITIALIZER(&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, ""); 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, ""); 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, ""); 108 109 /* 110 * Accounting system call. Written based on the specification and 111 * previous implementation done by Mark Tinguely. 112 */ 113 int 114 acct(uap) 115 struct acct_args /* { 116 syscallarg(char *) path; 117 } */ *uap; 118 { 119 struct thread *td = curthread; 120 struct nameidata nd; 121 int error; 122 123 /* Make sure that the caller is root. */ 124 error = suser(td); 125 if (error) 126 return (error); 127 128 /* 129 * If accounting is to be started to a file, open that file for 130 * appending and make sure it's a 'normal'. 131 */ 132 if (SCARG(uap, path) != NULL) { 133 NDINIT(&nd, NAMEI_LOOKUP, 0, UIO_USERSPACE, 134 SCARG(uap, path), td); 135 error = vn_open(&nd, FWRITE | O_APPEND, 0); 136 if (error) 137 return (error); 138 NDFREE(&nd, NDF_ONLY_PNBUF); 139 VOP_UNLOCK(nd.ni_vp, NULL, 0, td); 140 if (nd.ni_vp->v_type != VREG) { 141 vn_close(nd.ni_vp, FWRITE | O_APPEND, td); 142 return (EACCES); 143 } 144 } 145 146 /* 147 * If accounting was previously enabled, kill the old space-watcher, 148 * close the file, and (if no new file was specified, leave). 149 */ 150 if (acctp != NULLVP || savacctp != NULLVP) { 151 untimeout(acctwatch, NULL, acctwatch_handle); 152 error = vn_close((acctp != NULLVP ? acctp : savacctp), 153 FWRITE | O_APPEND, td); 154 acctp = savacctp = NULLVP; 155 } 156 if (SCARG(uap, path) == NULL) 157 return (error); 158 159 /* 160 * Save the new accounting file vnode, and schedule the new 161 * free space watcher. 162 */ 163 acctp = nd.ni_vp; 164 acctwatch(NULL); 165 return (error); 166 } 167 168 /* 169 * Write out process accounting information, on process exit. 170 * Data to be written out is specified in Leffler, et al. 171 * and are enumerated below. (They're also noted in the system 172 * "acct.h" header file.) 173 */ 174 175 int 176 acct_process(struct proc *p) 177 { 178 struct acct acct; 179 struct rusage *r; 180 struct timeval ut, st, tmp; 181 int t; 182 struct vnode *vp; 183 struct thread *td = p->p_thread; 184 185 /* If accounting isn't enabled, don't bother */ 186 vp = acctp; 187 if (vp == NULLVP) 188 return (0); 189 190 /* 191 * Get process accounting information. 192 */ 193 194 /* (1) The name of the command that ran */ 195 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 196 197 /* (2) The amount of user and system time that was used */ 198 calcru(p, &ut, &st, NULL); 199 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 200 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 201 202 /* (3) The elapsed time the commmand ran (and its starting time) */ 203 acct.ac_btime = p->p_stats->p_start.tv_sec; 204 microtime(&tmp); 205 timevalsub(&tmp, &p->p_stats->p_start); 206 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 207 208 /* (4) The average amount of memory used */ 209 r = &p->p_stats->p_ru; 210 tmp = ut; 211 timevaladd(&tmp, &st); 212 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 213 if (t) 214 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 215 else 216 acct.ac_mem = 0; 217 218 /* (5) The number of disk I/O operations done */ 219 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 220 221 /* (6) The UID and GID of the process */ 222 acct.ac_uid = p->p_ucred->cr_ruid; 223 acct.ac_gid = p->p_ucred->cr_rgid; 224 225 /* (7) The terminal from which the process was started */ 226 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 227 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 228 else 229 acct.ac_tty = NOUDEV; 230 231 /* (8) The boolean flags that tell how the process terminated, etc. */ 232 acct.ac_flag = p->p_acflag; 233 234 /* 235 * Eliminate any file size rlimit. 236 */ 237 if (p->p_limit->p_refcnt > 1 && 238 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 239 p->p_limit->p_refcnt--; 240 p->p_limit = limcopy(p->p_limit); 241 } 242 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 243 244 /* 245 * Write the accounting information to the file. 246 */ 247 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 248 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 249 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred, 250 (int *)0, td)); 251 } 252 253 /* 254 * Encode_comp_t converts from ticks in seconds and microseconds 255 * to ticks in 1/AHZ seconds. The encoding is described in 256 * Leffler, et al., on page 63. 257 */ 258 259 #define MANTSIZE 13 /* 13 bit mantissa. */ 260 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 261 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 262 263 static comp_t 264 encode_comp_t(s, us) 265 u_long s, us; 266 { 267 int exp, rnd; 268 269 exp = 0; 270 rnd = 0; 271 s *= AHZ; 272 s += us / (1000000 / AHZ); /* Maximize precision. */ 273 274 while (s > MAXFRACT) { 275 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 276 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 277 exp++; 278 } 279 280 /* If we need to round up, do it (and handle overflow correctly). */ 281 if (rnd && (++s > MAXFRACT)) { 282 s >>= EXPSIZE; 283 exp++; 284 } 285 286 /* Clean it up and polish it off. */ 287 exp <<= MANTSIZE; /* Shift the exponent into place */ 288 exp += s; /* and add on the mantissa. */ 289 return (exp); 290 } 291 292 /* 293 * Periodically check the file system to see if accounting 294 * should be turned on or off. Beware the case where the vnode 295 * has been vgone()'d out from underneath us, e.g. when the file 296 * system containing the accounting file has been forcibly unmounted. 297 */ 298 /* ARGSUSED */ 299 static void 300 acctwatch(a) 301 void *a; 302 { 303 struct statfs sb; 304 305 if (savacctp != NULLVP) { 306 if (savacctp->v_type == VBAD) { 307 (void) vn_close(savacctp, FWRITE | O_APPEND, NULL); 308 savacctp = NULLVP; 309 return; 310 } 311 (void)VFS_STATFS(savacctp->v_mount, &sb, NULL); 312 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 313 acctp = savacctp; 314 savacctp = NULLVP; 315 log(LOG_NOTICE, "Accounting resumed\n"); 316 } 317 } else { 318 if (acctp == NULLVP) 319 return; 320 if (acctp->v_type == VBAD) { 321 (void) vn_close(acctp, FWRITE | O_APPEND, NULL); 322 acctp = NULLVP; 323 return; 324 } 325 (void)VFS_STATFS(acctp->v_mount, &sb, NULL); 326 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 327 savacctp = acctp; 328 acctp = NULLVP; 329 log(LOG_NOTICE, "Accounting suspended\n"); 330 } 331 } 332 acctwatch_handle = timeout(acctwatch, NULL, acctchkfreq * hz); 333 } 334