1 /* $OpenBSD: kern_acct.c,v 1.45 2021/12/13 16:37:37 deraadt Exp $ */ 2 /* $NetBSD: kern_acct.c,v 1.42 1996/02/04 02:15:12 christos Exp $ */ 3 4 /*- 5 * Copyright (c) 1994 Christopher G. Demetriou 6 * Copyright (c) 1982, 1986, 1989, 1993 7 * The Regents of the University of California. All rights reserved. 8 * (c) UNIX System Laboratories, Inc. 9 * All or some portions of this file are derived from material licensed 10 * to the University of California by American Telephone and Telegraph 11 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 12 * the permission of UNIX System Laboratories, Inc. 13 * 14 * Redistribution and use in source and binary forms, with or without 15 * modification, are permitted provided that the following conditions 16 * are met: 17 * 1. Redistributions of source code must retain the above copyright 18 * notice, this list of conditions and the following disclaimer. 19 * 2. Redistributions in binary form must reproduce the above copyright 20 * notice, this list of conditions and the following disclaimer in the 21 * documentation and/or other materials provided with the distribution. 22 * 3. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93 39 */ 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/proc.h> 44 #include <sys/mount.h> 45 #include <sys/vnode.h> 46 #include <sys/fcntl.h> 47 #include <sys/syslog.h> 48 #include <sys/kernel.h> 49 #include <sys/namei.h> 50 #include <sys/errno.h> 51 #include <sys/acct.h> 52 #include <sys/resourcevar.h> 53 #include <sys/ioctl.h> 54 #include <sys/tty.h> 55 #include <sys/kthread.h> 56 #include <sys/rwlock.h> 57 58 #include <sys/syscallargs.h> 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 */ 74 comp_t encode_comp_t(u_long, u_long); 75 int acct_start(void); 76 void acct_thread(void *); 77 void acct_shutdown(void); 78 79 /* 80 * Accounting vnode pointer, and saved vnode pointer. 81 */ 82 struct vnode *acctp; 83 struct vnode *savacctp; 84 85 /* 86 * Lock protecting acctp and savacctp. 87 */ 88 struct rwlock acct_lock = RWLOCK_INITIALIZER("acctlk"); 89 90 /* 91 * Values associated with enabling and disabling accounting 92 */ 93 int acctsuspend = 2; /* stop accounting when < 2% free space left */ 94 int acctresume = 4; /* resume when free space risen to > 4% */ 95 int acctrate = 15; /* delay (in seconds) between space checks */ 96 97 struct proc *acct_proc; 98 99 /* 100 * Accounting system call. Written based on the specification and 101 * previous implementation done by Mark Tinguely. 102 */ 103 int 104 sys_acct(struct proc *p, void *v, register_t *retval) 105 { 106 struct sys_acct_args /* { 107 syscallarg(const char *) path; 108 } */ *uap = v; 109 struct nameidata nd; 110 int error; 111 112 /* Make sure that the caller is root. */ 113 if ((error = suser(p)) != 0) 114 return (error); 115 116 /* 117 * If accounting is to be started to a file, open that file for 118 * writing and make sure it's 'normal'. 119 */ 120 if (SCARG(uap, path) != NULL) { 121 NDINIT(&nd, 0, 0, UIO_USERSPACE, SCARG(uap, path), p); 122 if ((error = vn_open(&nd, FWRITE|O_APPEND, 0)) != 0) 123 return (error); 124 VOP_UNLOCK(nd.ni_vp); 125 if (nd.ni_vp->v_type != VREG) { 126 vn_close(nd.ni_vp, FWRITE, p->p_ucred, p); 127 return (EACCES); 128 } 129 } 130 131 rw_enter_write(&acct_lock); 132 133 /* 134 * If accounting was previously enabled, kill the old space-watcher, 135 * close the file, and (if no new file was specified, leave). 136 */ 137 if (acctp != NULL || savacctp != NULL) { 138 wakeup(&acct_proc); 139 (void)vn_close((acctp != NULL ? acctp : savacctp), FWRITE, 140 p->p_ucred, p); 141 acctp = savacctp = NULL; 142 } 143 if (SCARG(uap, path) == NULL) 144 goto out; 145 146 /* 147 * Save the new accounting file vnode, and schedule the new 148 * free space watcher. 149 */ 150 acctp = nd.ni_vp; 151 if ((error = acct_start()) != 0) { 152 acctp = NULL; 153 (void)vn_close(nd.ni_vp, FWRITE, p->p_ucred, p); 154 } 155 156 out: 157 rw_exit_write(&acct_lock); 158 return (error); 159 } 160 161 /* 162 * Write out process accounting information, on process exit. 163 * Data to be written out is specified in Leffler, et al. 164 * and are enumerated below. (They're also noted in the system 165 * "acct.h" header file.) 166 */ 167 int 168 acct_process(struct proc *p) 169 { 170 struct acct acct; 171 struct process *pr = p->p_p; 172 struct rusage *r; 173 struct timespec booted, elapsed, realstart, st, tmp, uptime, ut; 174 int t; 175 struct vnode *vp; 176 int error = 0; 177 178 /* If accounting isn't enabled, don't bother */ 179 if (acctp == NULL) 180 return (0); 181 182 rw_enter_read(&acct_lock); 183 184 /* 185 * Check the vnode again in case accounting got disabled while waiting 186 * for the lock. 187 */ 188 vp = acctp; 189 if (vp == NULL) 190 goto out; 191 192 /* 193 * Get process accounting information. 194 */ 195 196 /* (1) The name of the command that ran */ 197 memcpy(acct.ac_comm, pr->ps_comm, sizeof acct.ac_comm); 198 199 /* (2) The amount of user and system time that was used */ 200 calctsru(&pr->ps_tu, &ut, &st, NULL); 201 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_nsec); 202 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_nsec); 203 204 /* (3) The elapsed time the command ran (and its starting time) */ 205 nanouptime(&uptime); 206 nanoboottime(&booted); 207 timespecadd(&booted, &pr->ps_start, &realstart); 208 acct.ac_btime = realstart.tv_sec; 209 timespecsub(&uptime, &pr->ps_start, &elapsed); 210 acct.ac_etime = encode_comp_t(elapsed.tv_sec, elapsed.tv_nsec); 211 212 /* (4) The average amount of memory used */ 213 r = &p->p_ru; 214 timespecadd(&ut, &st, &tmp); 215 t = tmp.tv_sec * hz + tmp.tv_nsec / (1000 * tick); 216 if (t) 217 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t; 218 else 219 acct.ac_mem = 0; 220 221 /* (5) The number of disk I/O operations done */ 222 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0); 223 224 /* (6) The UID and GID of the process */ 225 acct.ac_uid = pr->ps_ucred->cr_ruid; 226 acct.ac_gid = pr->ps_ucred->cr_rgid; 227 228 /* (7) The terminal from which the process was started */ 229 if ((pr->ps_flags & PS_CONTROLT) && 230 pr->ps_pgrp->pg_session->s_ttyp) 231 acct.ac_tty = pr->ps_pgrp->pg_session->s_ttyp->t_dev; 232 else 233 acct.ac_tty = -1; 234 235 /* (8) The boolean flags that tell how process terminated or misbehaved. */ 236 acct.ac_flag = pr->ps_acflag; 237 238 /* 239 * Now, just write the accounting information to the file. 240 */ 241 error = vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct), 242 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT|IO_NOLIMIT, 243 p->p_ucred, NULL, p); 244 245 out: 246 rw_exit_read(&acct_lock); 247 return (error); 248 } 249 250 /* 251 * Encode_comp_t converts from ticks in seconds and microseconds 252 * to ticks in 1/AHZ seconds. The encoding is described in 253 * Leffler, et al., on page 63. 254 */ 255 256 #define MANTSIZE 13 /* 13 bit mantissa. */ 257 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ 258 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ 259 260 comp_t 261 encode_comp_t(u_long s, u_long ns) 262 { 263 int exp, rnd; 264 265 exp = 0; 266 rnd = 0; 267 s *= AHZ; 268 s += ns / (1000000000 / AHZ); /* Maximize precision. */ 269 270 while (s > MAXFRACT) { 271 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */ 272 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ 273 exp++; 274 } 275 276 /* If we need to round up, do it (and handle overflow correctly). */ 277 if (rnd && (++s > MAXFRACT)) { 278 s >>= EXPSIZE; 279 exp++; 280 } 281 282 /* Clean it up and polish it off. */ 283 exp <<= MANTSIZE; /* Shift the exponent into place */ 284 exp += s; /* and add on the mantissa. */ 285 return (exp); 286 } 287 288 int 289 acct_start(void) 290 { 291 /* Already running. */ 292 if (acct_proc != NULL) 293 return (0); 294 295 return (kthread_create(acct_thread, NULL, &acct_proc, "acct")); 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 void 305 acct_thread(void *arg) 306 { 307 struct statfs sb; 308 struct proc *p = curproc; 309 310 rw_enter_write(&acct_lock); 311 for (;;) { 312 if (savacctp != NULL) { 313 if (savacctp->v_type == VBAD) { 314 (void) vn_close(savacctp, FWRITE, NOCRED, p); 315 savacctp = NULL; 316 break; 317 } 318 (void)VFS_STATFS(savacctp->v_mount, &sb, NULL); 319 if (sb.f_bavail > acctresume * sb.f_blocks / 100) { 320 acctp = savacctp; 321 savacctp = NULL; 322 log(LOG_NOTICE, "Accounting resumed\n"); 323 } 324 } else if (acctp != NULL) { 325 if (acctp->v_type == VBAD) { 326 (void) vn_close(acctp, FWRITE, NOCRED, p); 327 acctp = NULL; 328 break; 329 } 330 (void)VFS_STATFS(acctp->v_mount, &sb, NULL); 331 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) { 332 savacctp = acctp; 333 acctp = NULL; 334 log(LOG_NOTICE, "Accounting suspended\n"); 335 } 336 } else { 337 break; 338 } 339 rwsleep_nsec(&acct_proc, &acct_lock, PPAUSE, "acct", 340 SEC_TO_NSEC(acctrate)); 341 } 342 acct_proc = NULL; 343 rw_exit_write(&acct_lock); 344 kthread_exit(0); 345 } 346 347 void 348 acct_shutdown(void) 349 { 350 351 struct proc *p = curproc; 352 353 rw_enter_write(&acct_lock); 354 if (acctp != NULL || savacctp != NULL) { 355 vn_close((acctp != NULL ? acctp : savacctp), FWRITE, 356 NOCRED, p); 357 acctp = savacctp = NULL; 358 } 359 rw_exit_write(&acct_lock); 360 } 361