bsd/kern/kern_acct.c

   1 /*
   2  * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
  29 /*-
  30  * Copyright (c) 1982, 1986, 1989, 1993
  31  *      The Regents of the University of California.  All rights reserved.
  32  * (c) UNIX System Laboratories, Inc.
  33  * All or some portions of this file are derived from material licensed
  34  * to the University of California by American Telephone and Telegraph
  35  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
  36  * the permission of UNIX System Laboratories, Inc.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. All advertising materials mentioning features or use of this software
  47  *    must display the following acknowledgement:
  48  *      This product includes software developed by the University of
  49  *      California, Berkeley and its contributors.
  50  * 4. Neither the name of the University nor the names of its contributors
  51  *    may be used to endorse or promote products derived from this software
  52  *    without specific prior written permission.
  53  *
  54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  64  * SUCH DAMAGE.
  65  *
  66  *      @(#)kern_acct.c 8.1 (Berkeley) 6/14/93
  67  */
  68 /* HISTORY
  69  * 08-May-95  Mac Gillon (mgillon) at NeXT
  70  *      Purged old history
  71  *      New version based on 4.4
  72  */
  73 /*
  74  * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
  75  * support for mandatory and extensible security protections.  This notice
  76  * is included in support of clause 2.2 (b) of the Apple Public License,
  77  * Version 2.0.
  78  */
  79
  80
  81 #include <sys/param.h>
  82 #include <sys/proc_internal.h>
  83 #include <sys/kauth.h>
  84 #include <sys/mount_internal.h>
  85 #include <sys/vnode_internal.h>
  86 #include <sys/file_internal.h>
  87 #include <sys/syslog.h>
  88 #include <sys/kernel.h>
  89 #include <sys/namei.h>
  90 #include <sys/errno.h>
  91 #include <sys/acct.h>
  92 #include <sys/resourcevar.h>
  93 #include <sys/ioctl.h>
  94 #include <sys/tty.h>
  95 #include <sys/sysproto.h>
  96 #if CONFIG_MACF
  97 #include <security/mac_framework.h>
  98 #endif
  99
 100 /*
 101  * The routines implemented in this file are described in:
 102  *      Leffler, et al.: The Design and Implementation of the 4.3BSD
 103  *          UNIX Operating System (Addison Welley, 1989)
 104  * on pages 62-63.
 105  *
 106  * Arguably, to simplify accounting operations, this mechanism should
 107  * be replaced by one in which an accounting log file (similar to /dev/klog)
 108  * is read by a user process, etc.  However, that has its own problems.
 109  */
 110
 111 /*
 112  * Internal accounting functions.
 113  * The former's operation is described in Leffler, et al., and the latter
 114  * was provided by UCB with the 4.4BSD-Lite release
 115  */
 116 comp_t  encode_comp_t(uint32_t, uint32_t);
 117 void    acctwatch(void *);
 118
 119 /*
 120  * Accounting vnode pointer, and suspended accounting vnode pointer.  States
 121  * are as follows:
 122  *
 123  *      acctp           suspend_acctp   state
 124  *      -------------   ------------    ------------------------------
 125  *      NULL            NULL            Accounting disabled
 126  *      !NULL           NULL            Accounting enabled
 127  *      NULL            !NULL           Accounting enabled, but suspended
 128  *      !NULL           !NULL           <not allowed>
 129  */
 130 struct  vnode *acctp;
 131 struct  vnode *suspend_acctp;
 132
 133 /*
 134  * Values associated with enabling and disabling accounting
 135  */
 136 int     acctsuspend = 2;        /* stop accounting when < 2% free space left */
 137 int     acctresume = 4;         /* resume when free space risen to > 4% */
 138 int     acctchkfreq = 15;       /* frequency (in seconds) to check space */
 139
 140
 141 static LCK_GRP_DECLARE(acct_subsys_lck_grp, "acct");
 142 static LCK_MTX_DECLARE(acct_subsys_mutex, &acct_subsys_lck_grp);
 143
 144 #define ACCT_SUBSYS_LOCK() lck_mtx_lock(&acct_subsys_mutex)
 145 #define ACCT_SUBSYS_UNLOCK() lck_mtx_unlock(&acct_subsys_mutex)
 146
 147
 148 /*
 149  * Accounting system call.  Written based on the specification and
 150  * previous implementation done by Mark Tinguely.
 151  */
 152 int
 153 acct(proc_t p, struct acct_args *uap, __unused int *retval)
 154 {
 155         struct nameidata nd;
 156         int error;
 157         struct vfs_context *ctx;
 158
 159         ctx = vfs_context_current();
 160
 161         /* Make sure that the caller is root. */
 162         if ((error = suser(vfs_context_ucred(ctx), &p->p_acflag))) {
 163                 return error;
 164         }
 165
 166         /*
 167          * If accounting is to be started to a file, open that file for
 168          * writing and make sure it's a 'normal'.
 169          */
 170         if (uap->path != USER_ADDR_NULL) {
 171                 NDINIT(&nd, LOOKUP, OP_OPEN, NOFOLLOW, UIO_USERSPACE, uap->path, ctx);
 172                 if ((error = vn_open(&nd, FWRITE, 0))) {
 173                         return error;
 174                 }
 175 #if CONFIG_MACF
 176                 error = mac_system_check_acct(vfs_context_ucred(ctx), nd.ni_vp);
 177                 if (error) {
 178                         vnode_put(nd.ni_vp);
 179                         vn_close(nd.ni_vp, FWRITE, ctx);
 180                         return error;
 181                 }
 182 #endif
 183                 vnode_put(nd.ni_vp);
 184
 185                 if (nd.ni_vp->v_type != VREG) {
 186                         vn_close(nd.ni_vp, FWRITE, ctx);
 187                         return EACCES;
 188                 }
 189         }
 190 #if CONFIG_MACF
 191         else {
 192                 error = mac_system_check_acct(vfs_context_ucred(ctx), NULL);
 193                 if (error) {
 194                         return error;
 195                 }
 196         }
 197 #endif
 198
 199         /*
 200          * If accounting was previously enabled, kill the old space-watcher,
 201          * close the file, and (if no new file was specified, leave).
 202          */
 203         ACCT_SUBSYS_LOCK();
 204         if (acctp != NULLVP || suspend_acctp != NULLVP) {
 205                 untimeout(acctwatch, NULL);
 206                 error = vn_close((acctp != NULLVP ? acctp : suspend_acctp),
 207                     FWRITE, vfs_context_current());
 208
 209                 acctp = suspend_acctp = NULLVP;
 210         }
 211         if (uap->path == USER_ADDR_NULL) {
 212                 ACCT_SUBSYS_UNLOCK();
 213                 return error;
 214         }
 215
 216         /*
 217          * Save the new accounting file vnode, and schedule the new
 218          * free space watcher.
 219          */
 220         acctp = nd.ni_vp;
 221         ACCT_SUBSYS_UNLOCK();
 222
 223         acctwatch(NULL);
 224         return error;
 225 }
 226
 227 /*
 228  * Write out process accounting information, on process exit.
 229  * Data to be written out is specified in Leffler, et al.
 230  * and are enumerated below.  (They're also noted in the system
 231  * "acct.h" header file.)
 232  */
 233 int
 234 acct_process(proc_t p)
 235 {
 236         struct acct an_acct;
 237         struct rusage rup, *r;
 238         struct timeval ut, st, tmp;
 239         int t;
 240         int error;
 241         struct vnode *vp;
 242         kauth_cred_t safecred;
 243         struct session * sessp;
 244         struct  tty *tp;
 245
 246         /* If accounting isn't enabled, don't bother */
 247         ACCT_SUBSYS_LOCK();
 248         vp = acctp;
 249         if (vp == NULLVP) {
 250                 ACCT_SUBSYS_UNLOCK();
 251                 return 0;
 252         }
 253
 254         /*
 255          * Get process accounting information.
 256          */
 257
 258         /* (1) The name of the command that ran */
 259         bcopy(p->p_comm, an_acct.ac_comm, sizeof an_acct.ac_comm);
 260
 261         /* (2) The amount of user and system time that was used */
 262         calcru(p, &ut, &st, NULL);
 263         an_acct.ac_utime = encode_comp_t((uint32_t)ut.tv_sec, ut.tv_usec);
 264         an_acct.ac_stime = encode_comp_t((uint32_t)st.tv_sec, st.tv_usec);
 265
 266         /* (3) The elapsed time the commmand ran (and its starting time) */
 267         an_acct.ac_btime = (u_int32_t)p->p_start.tv_sec;
 268         microtime(&tmp);
 269         timevalsub(&tmp, &p->p_start);
 270         an_acct.ac_etime = encode_comp_t((uint32_t)tmp.tv_sec, tmp.tv_usec);
 271
 272         /* (4) The average amount of memory used */
 273         proc_lock(p);
 274         rup = p->p_stats->p_ru;
 275         proc_unlock(p);
 276         r = &rup;
 277         tmp = ut;
 278         timevaladd(&tmp, &st);
 279         t = (int)(tmp.tv_sec * hz + tmp.tv_usec / tick);
 280         if (t) {
 281                 an_acct.ac_mem = (u_int16_t)((r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t);
 282         } else {
 283                 an_acct.ac_mem = 0;
 284         }
 285
 286         /* (5) The number of disk I/O operations done */
 287         an_acct.ac_io = encode_comp_t((uint32_t)(r->ru_inblock + r->ru_oublock), 0);
 288
 289         /* (6) The UID and GID of the process */
 290         safecred = kauth_cred_proc_ref(p);
 291
 292         an_acct.ac_uid = kauth_cred_getruid(safecred);
 293         an_acct.ac_gid = kauth_cred_getrgid(safecred);
 294
 295         /* (7) The terminal from which the process was started */
 296
 297         sessp = proc_session(p);
 298         if ((p->p_flag & P_CONTROLT) && (sessp != SESSION_NULL) && ((tp = SESSION_TP(sessp)) != TTY_NULL)) {
 299                 tty_lock(tp);
 300                 an_acct.ac_tty = tp->t_dev;
 301                 tty_unlock(tp);
 302         } else {
 303                 an_acct.ac_tty = NODEV;
 304         }
 305
 306         if (sessp != SESSION_NULL) {
 307                 session_rele(sessp);
 308         }
 309
 310         /* (8) The boolean flags that tell how the process terminated, etc. */
 311         an_acct.ac_flag = (u_int8_t)p->p_acflag;
 312
 313         /*
 314          * Now, just write the accounting information to the file.
 315          */
 316         if ((error = vnode_getwithref(vp)) == 0) {
 317                 error = vn_rdwr(UIO_WRITE, vp, (caddr_t)&an_acct, sizeof(an_acct),
 318                     (off_t)0, UIO_SYSSPACE, IO_APPEND | IO_UNIT, safecred,
 319                     (int *)0, p);
 320                 vnode_put(vp);
 321         }
 322
 323         kauth_cred_unref(&safecred);
 324         ACCT_SUBSYS_UNLOCK();
 325
 326         return error;
 327 }
 328
 329 /*
 330  * Encode_comp_t converts from ticks in seconds and microseconds
 331  * to ticks in 1/AHZ seconds.  The encoding is described in
 332  * Leffler, et al., on page 63.
 333  */
 334
 335 #define MANTSIZE        13                      /* 13 bit mantissa. */
 336 #define EXPSIZE         3                       /* Base 8 (3 bit) exponent. */
 337 #define MAXFRACT        ((1 << MANTSIZE) - 1)   /* Maximum fractional value. */
 338
 339 comp_t
 340 encode_comp_t(uint32_t s, uint32_t us)
 341 {
 342         int exp, rnd;
 343
 344         exp = 0;
 345         rnd = 0;
 346         s *= AHZ;
 347         s += us / (1000000 / AHZ);      /* Maximize precision. */
 348
 349         while (s > MAXFRACT) {
 350                 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
 351                 s >>= EXPSIZE;          /* Base 8 exponent == 3 bit shift. */
 352                 exp++;
 353         }
 354
 355         /* If we need to round up, do it (and handle overflow correctly). */
 356         if (rnd && (++s > MAXFRACT)) {
 357                 s >>= EXPSIZE;
 358                 exp++;
 359         }
 360
 361         /* Clean it up and polish it off. */
 362         exp <<= MANTSIZE;               /* Shift the exponent into place */
 363         exp += s;                       /* and add on the mantissa. */
 364         return (comp_t)exp;
 365 }
 366
 367 /*
 368  * Periodically check the file system to see if accounting
 369  * should be turned on or off.  Beware the case where the vnode
 370  * has been vgone()'d out from underneath us, e.g. when the file
 371  * system containing the accounting file has been forcibly unmounted.
 372  */
 373 /* ARGSUSED */
 374 void
 375 acctwatch(__unused void *a)
 376 {
 377         vfs_context_t ctx = vfs_context_current();
 378         struct vfs_attr va;
 379
 380         VFSATTR_INIT(&va);
 381         VFSATTR_WANTED(&va, f_blocks);
 382         VFSATTR_WANTED(&va, f_bavail);
 383
 384         ACCT_SUBSYS_LOCK();
 385         if (suspend_acctp != NULLVP) {
 386                 /*
 387                  * Resuming accounting when accounting is suspended, and the
 388                  * filesystem containing the suspended accounting file goes
 389                  * below a low watermark
 390                  */
 391                 if (suspend_acctp->v_type == VBAD) {
 392                         (void) vn_close(suspend_acctp, FWRITE, vfs_context_kernel());
 393                         suspend_acctp = NULLVP;
 394                         ACCT_SUBSYS_UNLOCK();
 395                         return;
 396                 }
 397                 (void)vfs_getattr(suspend_acctp->v_mount, &va, ctx);
 398                 if (va.f_bavail > acctresume * va.f_blocks / 100) {
 399                         acctp = suspend_acctp;
 400                         suspend_acctp = NULLVP;
 401                         log(LOG_NOTICE, "Accounting resumed\n");
 402                 }
 403         } else if (acctp != NULLVP) {
 404                 /*
 405                  * Suspending accounting when accounting is currently active,
 406                  * and the filesystem containing the active accounting file
 407                  * goes over a high watermark
 408                  */
 409                 if (acctp->v_type == VBAD) {
 410                         (void) vn_close(acctp, FWRITE, vfs_context_kernel());
 411                         acctp = NULLVP;
 412                         ACCT_SUBSYS_UNLOCK();
 413                         return;
 414                 }
 415                 (void)vfs_getattr(acctp->v_mount, &va, ctx);
 416                 if (va.f_bavail <= acctsuspend * va.f_blocks / 100) {
 417                         suspend_acctp = acctp;
 418                         acctp = NULLVP;
 419                         log(LOG_NOTICE, "Accounting suspended\n");
 420                 }
 421         } else {
 422                 ACCT_SUBSYS_UNLOCK();
 423                 return;
 424         }
 425         ACCT_SUBSYS_UNLOCK();
 426
 427         timeout(acctwatch, NULL, acctchkfreq * hz);
 428 }