kern_acct.c revision 69286
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: head/sys/kern/kern_acct.c 69286 2000-11-27 22:52:31Z jake $ 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/mount.h> 48#include <sys/vnode.h> 49#include <sys/fcntl.h> 50#include <sys/syslog.h> 51#include <sys/kernel.h> 52#include <sys/sysent.h> 53#include <sys/sysctl.h> 54#include <sys/namei.h> 55#include <sys/acct.h> 56#include <sys/resourcevar.h> 57#include <sys/tty.h> 58 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 */ 76static comp_t encode_comp_t __P((u_long, u_long)); 77static void acctwatch __P((void *)); 78 79/* 80 * Accounting callout used for periodic scheduling of acctwatch. 81 */ 82static struct callout acctwatch_callout; 83 84/* 85 * Accounting vnode pointer, and saved vnode pointer. 86 */ 87static struct vnode *acctp; 88static struct vnode *savacctp; 89 90/* 91 * Values associated with enabling and disabling accounting 92 */ 93static int acctsuspend = 2; /* stop accounting when < 2% free space left */ 94SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW, 95 &acctsuspend, 0, "percentage of free disk space below which accounting stops"); 96 97static int acctresume = 4; /* resume when free space risen to > 4% */ 98SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW, 99 &acctresume, 0, "percentage of free disk space above which accounting resumes"); 100 101static int acctchkfreq = 15; /* frequency (in seconds) to check space */ 102SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW, 103 &acctchkfreq, 0, "frequency for checking the free space"); 104 105/* 106 * Accounting system call. Written based on the specification and 107 * previous implementation done by Mark Tinguely. 108 */ 109int 110acct(a1, uap) 111 struct proc *a1; 112 struct acct_args /* { 113 syscallarg(char *) path; 114 } */ *uap; 115{ 116 struct proc *p = curproc; /* XXX */ 117 struct nameidata nd; 118 int error, flags; 119 120 /* Make sure that the caller is root. */ 121 error = suser(p); 122 if (error) 123 return (error); 124 125 /* 126 * If accounting is to be started to a file, open that file for 127 * writing and make sure it's a 'normal'. 128 */ 129 if (SCARG(uap, path) != NULL) { 130 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), 131 p); 132 flags = FWRITE; 133 error = vn_open(&nd, &flags, 0); 134 if (error) 135 return (error); 136 NDFREE(&nd, NDF_ONLY_PNBUF); 137 VOP_UNLOCK(nd.ni_vp, 0, p); 138 if (nd.ni_vp->v_type != VREG) { 139 vn_close(nd.ni_vp, FWRITE, p->p_ucred, p); 140 return (EACCES); 141 } 142 } 143 144 /* 145 * If accounting was previously enabled, kill the old space-watcher, 146 * close the file, and (if no new file was specified, leave). 147 */ 148 if (acctp != NULLVP || savacctp != NULLVP) { 149 callout_stop(&acctwatch_callout); 150 error = vn_close((acctp != NULLVP ? acctp : savacctp), FWRITE, 151 p->p_ucred, p); 152 acctp = savacctp = NULLVP; 153 } 154 if (SCARG(uap, path) == NULL) 155 return (error); 156 157 /* 158 * Save the new accounting file vnode, and schedule the new 159 * free space watcher. 160 */ 161 acctp = nd.ni_vp; 162 callout_init(&acctwatch_callout, 0); 163 acctwatch(NULL); 164 return (error); 165} 166 167/* 168 * Write out process accounting information, on process exit. 169 * Data to be written out is specified in Leffler, et al. 170 * and are enumerated below. (They're also noted in the system 171 * "acct.h" header file.) 172 */ 173 174int 175acct_process(p) 176 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; 183 184 /* If accounting isn't enabled, don't bother */ 185 vp = acctp; 186 if (vp == NULLVP) 187 return (0); 188 189 /* 190 * Get process accounting information. 191 */ 192 193 /* (1) The name of the command that ran */ 194 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 195 196 /* (2) The amount of user and system time that was used */ 197 calcru(p, &ut, &st, NULL); 198 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 199 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 200 201 /* (3) The elapsed time the commmand ran (and its starting time) */ 202 acct.ac_btime = p->p_stats->p_start.tv_sec; 203 microtime(&tmp); 204 timevalsub(&tmp, &p->p_stats->p_start); 205 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 206 207 /* (4) The average amount of memory used */ 208 r = &p->p_stats->p_ru; 209 tmp = ut; 210 timevaladd(&tmp, &st); 211 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 212 if (t) 213 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 214 else 215 acct.ac_mem = 0; 216 217 /* (5) The number of disk I/O operations done */ 218 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 219 220 /* (6) The UID and GID of the process */ 221 acct.ac_uid = p->p_cred->p_ruid; 222 acct.ac_gid = p->p_cred->p_rgid; 223 224 /* (7) The terminal from which the process was started */ 225 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 226 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 227 else 228 acct.ac_tty = NOUDEV; 229 230 /* (8) The boolean flags that tell how the process terminated, etc. */ 231 acct.ac_flag = p->p_acflag; 232 233 /* 234 * Eliminate any file size rlimit. 235 */ 236 if (p->p_limit->p_refcnt > 1 && 237 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 238 p->p_limit->p_refcnt--; 239 p->p_limit = limcopy(p->p_limit); 240 } 241 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 242 243 /* 244 * Write the accounting information to the file. 245 */ 246 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE); 247 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 248 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred, 249 (int *)0, p)); 250} 251 252/* 253 * Encode_comp_t converts from ticks in seconds and microseconds 254 * to ticks in 1/AHZ seconds. The encoding is described in 255 * Leffler, et al., on page 63. 256 */ 257 258#define MANTSIZE 13 /* 13 bit mantissa. */ 259#define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 260#define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 261 262static comp_t 263encode_comp_t(s, us) 264 u_long s, us; 265{ 266 int exp, rnd; 267 268 exp = 0; 269 rnd = 0; 270 s *= AHZ; 271 s += us / (1000000 / AHZ); /* Maximize precision. */ 272 273 while (s > MAXFRACT) { 274 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 275 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 276 exp++; 277 } 278 279 /* If we need to round up, do it (and handle overflow correctly). */ 280 if (rnd && (++s > MAXFRACT)) { 281 s >>= EXPSIZE; 282 exp++; 283 } 284 285 /* Clean it up and polish it off. */ 286 exp <<= MANTSIZE; /* Shift the exponent into place */ 287 exp += s; /* and add on the mantissa. */ 288 return (exp); 289} 290 291/* 292 * Periodically check the file system to see if accounting 293 * should be turned on or off. Beware the case where the vnode 294 * has been vgone()'d out from underneath us, e.g. when the file 295 * system containing the accounting file has been forcibly unmounted. 296 */ 297/* ARGSUSED */ 298static void 299acctwatch(a) 300 void *a; 301{ 302 struct statfs sb; 303 304 if (savacctp != NULLVP) { 305 if (savacctp->v_type == VBAD) { 306 (void) vn_close(savacctp, FWRITE, NOCRED, NULL); 307 savacctp = NULLVP; 308 return; 309 } 310 (void)VFS_STATFS(savacctp->v_mount, &sb, (struct proc *)0); 311 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 312 acctp = savacctp; 313 savacctp = NULLVP; 314 log(LOG_NOTICE, "Accounting resumed\n"); 315 } 316 } else { 317 if (acctp == NULLVP) 318 return; 319 if (acctp->v_type == VBAD) { 320 (void) vn_close(acctp, FWRITE, NOCRED, NULL); 321 acctp = NULLVP; 322 return; 323 } 324 (void)VFS_STATFS(acctp->v_mount, &sb, (struct proc *)0); 325 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 326 savacctp = acctp; 327 acctp = NULLVP; 328 log(LOG_NOTICE, "Accounting suspended\n"); 329 } 330 } 331 callout_reset(&acctwatch_callout, acctchkfreq * hz, acctwatch, NULL); 332} 333