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 * $FreeBSD$ 41 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/lock.h> 46 #include <sys/mutex.h> 47 #include <sys/sysproto.h> 48 #include <sys/proc.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/namei.h> 57 #include <sys/acct.h> 58 #include <sys/resourcevar.h> 59 #include <sys/tty.h> 60 61 /* 62 * The routines implemented in this file are described in: 63 * Leffler, et al.: The Design and Implementation of the 4.3BSD 64 * UNIX Operating System (Addison Welley, 1989) 65 * on pages 62-63. 66 * 67 * Arguably, to simplify accounting operations, this mechanism should 68 * be replaced by one in which an accounting log file (similar to /dev/klog) 69 * is read by a user process, etc. However, that has its own problems. 70 */ 71 72 /* 73 * Internal accounting functions. 74 * The former's operation is described in Leffler, et al., and the latter 75 * was provided by UCB with the 4.4BSD-Lite release 76 */ 77 static comp_t encode_comp_t(u_long, u_long); 78 static void acctwatch(void *); 79 80 /* 81 * Accounting callout used for periodic scheduling of acctwatch. 82 */ 83 static struct callout acctwatch_callout; 84 85 /* 86 * Accounting vnode pointer, saved vnode pointer, and flags for each. 87 */ 88 static struct vnode *acctp; 89 static struct ucred *acctcred; 90 static int acctflags; 91 static struct vnode *savacctp; 92 static struct ucred *savacctcred; 93 static int savacctflags; 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 stops"); 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, "percentage of free disk space above which accounting resumes"); 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 for checking the free space"); 109 110 /* 111 * Accounting system call. Written based on the specification and 112 * previous implementation done by Mark Tinguely. 113 * 114 * MPSAFE 115 */ 116 int 117 acct(td, uap) 118 struct thread *td; 119 struct acct_args /* { 120 syscallarg(char *) path; 121 } */ *uap; 122 { 123 struct nameidata nd; 124 int error, flags; 125 126 /* Make sure that the caller is root. */ 127 error = suser(td); 128 if (error) 129 return (error); 130 131 mtx_lock(&Giant); 132 /* 133 * If accounting is to be started to a file, open that file for 134 * appending and make sure it's a 'normal'. 135 */ 136 if (SCARG(uap, path) != NULL) { 137 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), 138 td); 139 flags = FWRITE | O_APPEND; 140 error = vn_open(&nd, &flags, 0); 141 if (error) 142 goto done2; 143 NDFREE(&nd, NDF_ONLY_PNBUF); 144 VOP_UNLOCK(nd.ni_vp, 0, td); 145 if (nd.ni_vp->v_type != VREG) { 146 vn_close(nd.ni_vp, flags, td->td_ucred, td); 147 error = EACCES; 148 goto done2; 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_callout); 158 error = vn_close((acctp != NULLVP ? acctp : savacctp), 159 (acctp != NULLVP ? acctflags : savacctflags), 160 (acctcred != NOCRED ? acctcred : savacctcred), td); 161 acctp = savacctp = NULLVP; 162 crfree(acctcred != NOCRED ? acctcred : savacctcred); 163 acctcred = savacctcred = NOCRED; 164 } 165 if (SCARG(uap, path) == NULL) 166 goto done2; 167 168 /* 169 * Save the new accounting file vnode, and schedule the new 170 * free space watcher. 171 */ 172 acctp = nd.ni_vp; 173 acctcred = crhold(td->td_ucred); 174 acctflags = flags; 175 callout_init(&acctwatch_callout, 0); 176 acctwatch(NULL); 177 done2: 178 mtx_unlock(&Giant); 179 return (error); 180 } 181 182 /* 183 * Write out process accounting information, on process exit. 184 * Data to be written out is specified in Leffler, et al. 185 * and are enumerated below. (They're also noted in the system 186 * "acct.h" header file.) 187 */ 188 189 int 190 acct_process(td) 191 struct thread *td; 192 { 193 struct proc *p = td->td_proc; 194 struct acct acct; 195 struct rusage *r; 196 struct timeval ut, st, tmp; 197 int t; 198 struct vnode *vp; 199 200 /* If accounting isn't enabled, don't bother */ 201 vp = acctp; 202 if (vp == NULLVP) 203 return (0); 204 205 /* 206 * Get process accounting information. 207 */ 208 209 /* (1) The name of the command that ran */ 210 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm); 211 212 /* (2) The amount of user and system time that was used */ 213 mtx_lock_spin(&sched_lock); 214 calcru(p, &ut, &st, NULL); 215 mtx_unlock_spin(&sched_lock); 216 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec); 217 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec); 218 219 /* (3) The elapsed time the commmand ran (and its starting time) */ 220 acct.ac_btime = p->p_stats->p_start.tv_sec; 221 microtime(&tmp); 222 timevalsub(&tmp, &p->p_stats->p_start); 223 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec); 224 225 /* (4) The average amount of memory used */ 226 r = &p->p_stats->p_ru; 227 tmp = ut; 228 timevaladd(&tmp, &st); 229 t = tmp.tv_sec * hz + tmp.tv_usec / tick; 230 if (t) 231 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 232 else 233 acct.ac_mem = 0; 234 235 /* (5) The number of disk I/O operations done */ 236 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 237 238 /* (6) The UID and GID of the process */ 239 acct.ac_uid = p->p_ucred->cr_ruid; 240 acct.ac_gid = p->p_ucred->cr_rgid; 241 242 /* (7) The terminal from which the process was started */ 243 PROC_LOCK(p); 244 SESS_LOCK(p->p_session); 245 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp) 246 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev); 247 else 248 acct.ac_tty = NOUDEV; 249 SESS_UNLOCK(p->p_session); 250 PROC_UNLOCK(p); 251 252 /* (8) The boolean flags that tell how the process terminated, etc. */ 253 acct.ac_flag = p->p_acflag; 254 255 /* 256 * Eliminate any file size rlimit. 257 */ 258 if (p->p_limit->p_refcnt > 1 && 259 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) { 260 p->p_limit->p_refcnt--; 261 p->p_limit = limcopy(p->p_limit); 262 } 263 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; 264 265 /* 266 * Write the accounting information to the file. 267 */ 268 VOP_LEASE(vp, td, acctcred, LEASE_WRITE); 269 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 270 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, acctcred, NOCRED, 271 (int *)0, td)); 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(s, us) 286 u_long s, us; 287 { 288 int exp, rnd; 289 290 exp = 0; 291 rnd = 0; 292 s *= AHZ; 293 s += us / (1000000 / AHZ); /* Maximize precision. */ 294 295 while (s > MAXFRACT) { 296 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 297 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 298 exp++; 299 } 300 301 /* If we need to round up, do it (and handle overflow correctly). */ 302 if (rnd && (++s > MAXFRACT)) { 303 s >>= EXPSIZE; 304 exp++; 305 } 306 307 /* Clean it up and polish it off. */ 308 exp <<= MANTSIZE; /* Shift the exponent into place */ 309 exp += s; /* and add on the mantissa. */ 310 return (exp); 311 } 312 313 /* 314 * Periodically check the filesystem to see if accounting 315 * should be turned on or off. Beware the case where the vnode 316 * has been vgone()'d out from underneath us, e.g. when the file 317 * system containing the accounting file has been forcibly unmounted. 318 */ 319 /* ARGSUSED */ 320 static void 321 acctwatch(a) 322 void *a; 323 { 324 struct statfs sb; 325 326 if (savacctp != NULLVP) { 327 if (savacctp->v_type == VBAD) { 328 (void) vn_close(savacctp, savacctflags, savacctcred, 329 NULL); 330 savacctp = NULLVP; 331 savacctcred = NOCRED; 332 return; 333 } 334 (void)VFS_STATFS(savacctp->v_mount, &sb, (struct thread *)0); 335 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 336 acctp = savacctp; 337 acctcred = savacctcred; 338 acctflags = savacctflags; 339 savacctp = NULLVP; 340 savacctcred = NOCRED; 341 log(LOG_NOTICE, "Accounting resumed\n"); 342 } 343 } else { 344 if (acctp == NULLVP) 345 return; 346 if (acctp->v_type == VBAD) { 347 (void) vn_close(acctp, acctflags, acctcred, NULL); 348 acctp = NULLVP; 349 crfree(acctcred); 350 acctcred = NOCRED; 351 return; 352 } 353 (void)VFS_STATFS(acctp->v_mount, &sb, (struct thread *)0); 354 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 355 savacctp = acctp; 356 savacctflags = acctflags; 357 acctp = NULLVP; 358 acctcred = NOCRED; 359 log(LOG_NOTICE, "Accounting suspended\n"); 360 } 361 } 362 callout_reset(&acctwatch_callout, acctchkfreq * hz, acctwatch, NULL); 363 } 364