2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
10 * Copyright (c) 1994 Christopher G. Demetriou
11 * Copyright (c) 2005 Robert N. M. Watson
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. All advertising materials mentioning features or use of this software
22 * must display the following acknowledgement:
23 * This product includes software developed by the University of
24 * California, Berkeley and its contributors.
25 * 4. Neither the name of the University nor the names of its contributors
26 * may be used to endorse or promote products derived from this software
27 * without specific prior written permission.
29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93
44 #include <sys/cdefs.h>
45 __FBSDID("$FreeBSD$");
49 #include <sys/param.h>
50 #include <sys/systm.h>
52 #include <sys/mutex.h>
53 #include <sys/sysproto.h>
56 #include <sys/mount.h>
57 #include <sys/vnode.h>
58 #include <sys/fcntl.h>
59 #include <sys/syslog.h>
60 #include <sys/kernel.h>
62 #include <sys/sysent.h>
63 #include <sys/sysctl.h>
64 #include <sys/namei.h>
66 #include <sys/resourcevar.h>
70 * The routines implemented in this file are described in:
71 * Leffler, et al.: The Design and Implementation of the 4.3BSD
72 * UNIX Operating System (Addison Welley, 1989)
75 * Arguably, to simplify accounting operations, this mechanism should
76 * be replaced by one in which an accounting log file (similar to /dev/klog)
77 * is read by a user process, etc. However, that has its own problems.
81 * Internal accounting functions.
82 * The former's operation is described in Leffler, et al., and the latter
83 * was provided by UCB with the 4.4BSD-Lite release
85 static comp_t encode_comp_t(u_long, u_long);
86 static void acctwatch(void *);
89 * Accounting callout used for periodic scheduling of acctwatch.
91 static struct callout acctwatch_callout;
94 * Accounting vnode pointer, saved vnode pointer, and flags for each.
95 * acct_sx protects against changes to the active vnode and credentials
96 * while accounting records are being committed to disk.
98 static int acct_suspended;
99 static struct vnode *acct_vp;
100 static struct ucred *acct_cred;
101 static int acct_flags;
102 static struct sx acct_sx;
104 SX_SYSINIT(acct, &acct_sx, "acct_sx");
107 * Values associated with enabling and disabling accounting
109 static int acctsuspend = 2; /* stop accounting when < 2% free space left */
110 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
111 &acctsuspend, 0, "percentage of free disk space below which accounting stops");
113 static int acctresume = 4; /* resume when free space risen to > 4% */
114 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
115 &acctresume, 0, "percentage of free disk space above which accounting resumes");
117 static int acctchkfreq = 15; /* frequency (in seconds) to check space */
118 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW,
119 &acctchkfreq, 0, "frequency for checking the free space");
121 SYSCTL_INT(_kern, OID_AUTO, acct_suspended, CTLFLAG_RD, &acct_suspended, 0,
122 "Accounting suspended or not");
125 * Accounting system call. Written based on the specification and
126 * previous implementation done by Mark Tinguely.
131 acct(struct thread *td, struct acct_args *uap)
136 /* Make sure that the caller is root. */
142 * If accounting is to be started to a file, open that file for
143 * appending and make sure it's a 'normal'. While we could
144 * conditionally acquire Giant here, we're actually interacting with
145 * vnodes from possibly two file systems, making the logic a bit
146 * complicated. For now, use Giant unconditionally.
149 if (uap->path != NULL) {
150 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
151 flags = FWRITE | O_APPEND;
152 error = vn_open(&nd, &flags, 0, -1);
155 NDFREE(&nd, NDF_ONLY_PNBUF);
157 error = mac_check_system_acct(td->td_ucred, nd.ni_vp);
159 VOP_UNLOCK(nd.ni_vp, 0, td);
160 vn_close(nd.ni_vp, flags, td->td_ucred, td);
164 VOP_UNLOCK(nd.ni_vp, 0, td);
165 if (nd.ni_vp->v_type != VREG) {
166 vn_close(nd.ni_vp, flags, td->td_ucred, td);
172 error = mac_check_system_acct(td->td_ucred, NULL);
179 * Disallow concurrent access to the accounting vnode while we swap
180 * it out, in order to prevent access after close.
185 * If accounting was previously enabled, kill the old space-watcher,
186 * close the file, and (if no new file was specified, leave). Reset
187 * the suspended state regardless of whether accounting remains
191 if (acct_vp != NULL) {
192 callout_stop(&acctwatch_callout);
193 error = vn_close(acct_vp, acct_flags, acct_cred, td);
198 log(LOG_NOTICE, "Accounting disabled\n");
200 if (uap->path == NULL) {
201 sx_xunlock(&acct_sx);
206 * Save the new accounting file vnode, and schedule the new
207 * free space watcher.
210 acct_cred = crhold(td->td_ucred);
212 callout_init(&acctwatch_callout, CALLOUT_MPSAFE);
213 sx_xunlock(&acct_sx);
214 log(LOG_NOTICE, "Accounting enabled\n");
222 * Write out process accounting information, on process exit.
223 * Data to be written out is specified in Leffler, et al.
224 * and are enumerated below. (They're also noted in the system
225 * "acct.h" header file.)
228 acct_process(struct thread *td)
231 struct timeval ut, st, tmp;
232 struct plimit *newlim, *oldlim;
235 int t, ret, vfslocked;
238 * Lockless check of accounting condition before doing the hard
241 if (acct_vp == NULL || acct_suspended)
247 * If accounting isn't enabled, don't bother. Have to check again
248 * once we own the lock in case we raced with disabling of accounting
251 if (acct_vp == NULL || acct_suspended) {
252 sx_sunlock(&acct_sx);
259 * Get process accounting information.
263 /* (1) The name of the command that ran */
264 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
266 /* (2) The amount of user and system time that was used */
268 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
269 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
271 /* (3) The elapsed time the command ran (and its starting time) */
273 timevaladd(&tmp, &p->p_stats->p_start);
274 acct.ac_btime = tmp.tv_sec;
276 timevalsub(&tmp, &p->p_stats->p_start);
277 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
279 /* (4) The average amount of memory used */
280 r = &p->p_stats->p_ru;
282 timevaladd(&tmp, &st);
283 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
285 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
289 /* (5) The number of disk I/O operations done */
290 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
292 /* (6) The UID and GID of the process */
293 acct.ac_uid = p->p_ucred->cr_ruid;
294 acct.ac_gid = p->p_ucred->cr_rgid;
296 /* (7) The terminal from which the process was started */
297 SESS_LOCK(p->p_session);
298 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
299 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
302 SESS_UNLOCK(p->p_session);
304 /* (8) The boolean flags that tell how the process terminated, etc. */
305 acct.ac_flag = p->p_acflag;
309 * Eliminate any file size rlimit.
311 newlim = lim_alloc();
314 lim_copy(newlim, oldlim);
315 newlim->pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
321 * Write the accounting information to the file.
323 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
324 VOP_LEASE(acct_vp, td, acct_cred, LEASE_WRITE);
325 ret = vn_rdwr(UIO_WRITE, acct_vp, (caddr_t)&acct, sizeof (acct),
326 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, acct_cred, NOCRED,
328 VFS_UNLOCK_GIANT(vfslocked);
329 sx_sunlock(&acct_sx);
334 * Encode_comp_t converts from ticks in seconds and microseconds
335 * to ticks in 1/AHZ seconds. The encoding is described in
336 * Leffler, et al., on page 63.
339 #define MANTSIZE 13 /* 13 bit mantissa. */
340 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
341 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
344 encode_comp_t(u_long s, u_long us)
351 s += us / (1000000 / AHZ); /* Maximize precision. */
353 while (s > MAXFRACT) {
354 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
355 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
359 /* If we need to round up, do it (and handle overflow correctly). */
360 if (rnd && (++s > MAXFRACT)) {
365 /* Clean it up and polish it off. */
366 exp <<= MANTSIZE; /* Shift the exponent into place */
367 exp += s; /* and add on the mantissa. */
372 * Periodically check the filesystem to see if accounting
373 * should be turned on or off. Beware the case where the vnode
374 * has been vgone()'d out from underneath us, e.g. when the file
375 * system containing the accounting file has been forcibly unmounted.
385 vfslocked = VFS_LOCK_GIANT(acct_vp->v_mount);
386 if (acct_vp->v_type == VBAD) {
387 (void) vn_close(acct_vp, acct_flags, acct_cred, NULL);
388 VFS_UNLOCK_GIANT(vfslocked);
393 sx_xunlock(&acct_sx);
394 log(LOG_NOTICE, "Accounting disabled\n");
398 * Stopping here is better than continuing, maybe it will be VBAD
401 if (VFS_STATFS(acct_vp->v_mount, &sb, curthread) < 0) {
402 VFS_UNLOCK_GIANT(vfslocked);
403 sx_xunlock(&acct_sx);
406 VFS_UNLOCK_GIANT(vfslocked);
407 if (acct_suspended) {
408 if (sb.f_bavail > (int64_t)(acctresume * sb.f_blocks /
411 log(LOG_NOTICE, "Accounting resumed\n");
414 if (sb.f_bavail <= (int64_t)(acctsuspend * sb.f_blocks /
417 log(LOG_NOTICE, "Accounting suspended\n");
420 callout_reset(&acctwatch_callout, acctchkfreq * hz, acctwatch, NULL);
421 sx_xunlock(&acct_sx);