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 1.14 2004/09/17 17:06:59 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 acctwatch_handle; 86 87 /* 88 * Accounting vnode pointer, and saved vnode pointer. 89 */ 90 static struct vnode *acctp; 91 static struct vnode *savacctp; 92 93 /* 94 * Values associated with enabling and disabling accounting 95 */ 96 static int acctsuspend = 2; /* stop accounting when < 2% free space left */ 97 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW, 98 &acctsuspend, 0, "Percentage of free disk space below which accounting will be suspended"); 99 100 static int acctresume = 4; /* resume when free space risen to > 4% */ 101 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW, 102 &acctresume, 0, "Minimum percentage of free disk space needed to resume accounting"); 103 104 static int acctchkfreq = 15; /* frequency (in seconds) to check space */ 105 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW, 106 &acctchkfreq, 0, "Frequency (in seconds) of free disk space checking"); 107 108 static void 109 acct_init(void *arg __unused) 110 { 111 callout_init(&acctwatch_handle); 112 } 113 SYSINIT(acct, SI_SUB_DRIVERS, SI_ORDER_ANY, acct_init, NULL); 114 115 /* 116 * Accounting system call. Written based on the specification and 117 * previous implementation done by Mark Tinguely. 118 */ 119 int 120 acct(uap) 121 struct acct_args /* { 122 syscallarg(char *) path; 123 } */ *uap; 124 { 125 struct thread *td = curthread; 126 struct nameidata nd; 127 int error; 128 129 /* Make sure that the caller is root. */ 130 error = suser(td); 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 (SCARG(uap, path) != NULL) { 139 NDINIT(&nd, NAMEI_LOOKUP, 0, UIO_USERSPACE, 140 SCARG(uap, path), td); 141 error = vn_open(&nd, FWRITE | O_APPEND, 0); 142 if (error) 143 return (error); 144 NDFREE(&nd, NDF_ONLY_PNBUF); 145 VOP_UNLOCK(nd.ni_vp, NULL, 0, td); 146 if (nd.ni_vp->v_type != VREG) { 147 vn_close(nd.ni_vp, FWRITE | O_APPEND, td); 148 return (EACCES); 149 } 150 } 151 152 /* 153 * If accounting was previously enabled, kill the old space-watcher, 154 * close the file, and (if no new file was specified, leave). 155 */ 156 if (acctp != NULLVP || savacctp != NULLVP) { 157 callout_stop(&acctwatch_handle); 158 error = vn_close((acctp != NULLVP ? acctp : savacctp), 159 FWRITE | O_APPEND, td); 160 acctp = savacctp = NULLVP; 161 } 162 if (SCARG(uap, path) == NULL) 163 return (error); 164 165 /* 166 * Save the new accounting file vnode, and schedule the new 167 * free space watcher. 168 */ 169 acctp = nd.ni_vp; 170 acctwatch(NULL); 171 return (error); 172 } 173 174 /* 175 * Write out process accounting information, on process exit. 176 * Data to be written out is specified in Leffler, et al. 177 * and are enumerated below. (They're also noted in the system 178 * "acct.h" header file.) 179 */ 180 181 int 182 acct_process(struct proc *p) 183 { 184 struct acct acct; 185 struct rusage *r; 186 struct timeval ut, st, tmp; 187 int t; 188 struct vnode *vp; 189 struct thread *td = p->p_thread; 190 191 /* If accounting isn't enabled, don't bother */ 192 vp = acctp; 193 if (vp == NULLVP) 194 return (0); 195 196 /* 197 * Get process accounting information. 198 */ 199 200 /* (1) The name of the command that ran */ 201 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 202 203 /* (2) The amount of user and system time that was used */ 204 calcru(p, &ut, &st, NULL); 205 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 206 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 207 208 /* (3) The elapsed time the commmand ran (and its starting time) */ 209 acct.ac_btime = p->p_thread->td_start.tv_sec; 210 microtime(&tmp); 211 timevalsub(&tmp, &p->p_thread->td_start); 212 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 213 214 /* (4) The average amount of memory used */ 215 r = &p->p_stats->p_ru; 216 tmp = ut; 217 timevaladd(&tmp, &st); 218 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 219 if (t) 220 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 221 else 222 acct.ac_mem = 0; 223 224 /* (5) The number of disk I/O operations done */ 225 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 226 227 /* (6) The UID and GID of the process */ 228 acct.ac_uid = p->p_ucred->cr_ruid; 229 acct.ac_gid = p->p_ucred->cr_rgid; 230 231 /* (7) The terminal from which the process was started */ 232 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 233 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 234 else 235 acct.ac_tty = NOUDEV; 236 237 /* (8) The boolean flags that tell how the process terminated, etc. */ 238 acct.ac_flag = p->p_acflag; 239 240 /* 241 * Eliminate any file size rlimit. 242 */ 243 if (p->p_limit->p_refcnt > 1 && 244 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 245 p->p_limit->p_refcnt--; 246 p->p_limit = limcopy(p->p_limit); 247 } 248 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 249 250 /* 251 * Write the accounting information to the file. 252 */ 253 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE); 254 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 255 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred, 256 (int *)0, td)); 257 } 258 259 /* 260 * Encode_comp_t converts from ticks in seconds and microseconds 261 * to ticks in 1/AHZ seconds. The encoding is described in 262 * Leffler, et al., on page 63. 263 */ 264 265 #define MANTSIZE 13 /* 13 bit mantissa. */ 266 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 267 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 268 269 static comp_t 270 encode_comp_t(s, us) 271 u_long s, us; 272 { 273 int exp, rnd; 274 275 exp = 0; 276 rnd = 0; 277 s *= AHZ; 278 s += us / (1000000 / AHZ); /* Maximize precision. */ 279 280 while (s > MAXFRACT) { 281 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 282 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 283 exp++; 284 } 285 286 /* If we need to round up, do it (and handle overflow correctly). */ 287 if (rnd && (++s > MAXFRACT)) { 288 s >>= EXPSIZE; 289 exp++; 290 } 291 292 /* Clean it up and polish it off. */ 293 exp <<= MANTSIZE; /* Shift the exponent into place */ 294 exp += s; /* and add on the mantissa. */ 295 return (exp); 296 } 297 298 /* 299 * Periodically check the file system to see if accounting 300 * should be turned on or off. Beware the case where the vnode 301 * has been vgone()'d out from underneath us, e.g. when the file 302 * system containing the accounting file has been forcibly unmounted. 303 */ 304 /* ARGSUSED */ 305 static void 306 acctwatch(a) 307 void *a; 308 { 309 struct statfs sb; 310 311 if (savacctp != NULLVP) { 312 if (savacctp->v_type == VBAD) { 313 (void) vn_close(savacctp, FWRITE | O_APPEND, NULL); 314 savacctp = NULLVP; 315 return; 316 } 317 (void)VFS_STATFS(savacctp->v_mount, &sb, NULL); 318 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 319 acctp = savacctp; 320 savacctp = NULLVP; 321 log(LOG_NOTICE, "Accounting resumed\n"); 322 } 323 } else { 324 if (acctp == NULLVP) 325 return; 326 if (acctp->v_type == VBAD) { 327 (void) vn_close(acctp, FWRITE | O_APPEND, NULL); 328 acctp = NULLVP; 329 return; 330 } 331 (void)VFS_STATFS(acctp->v_mount, &sb, NULL); 332 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 333 savacctp = acctp; 334 acctp = NULLVP; 335 log(LOG_NOTICE, "Accounting suspended\n"); 336 } 337 } 338 callout_reset(&acctwatch_handle, acctchkfreq * hz, acctwatch, NULL); 339 } 340