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.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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.
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
39 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93
40 * $FreeBSD: src/sys/kern/kern_acct.c,v 1.23.2.1 2002/07/24 18:33:55 johan Exp $
41 * $DragonFly: src/sys/kern/kern_acct.c,v 1.28 2008/05/08 01:26:00 dillon Exp $
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.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/nlookup.h>
58 #include <sys/resourcevar.h>
61 #include <vm/vm_zone.h>
64 * The routines implemented in this file are described in:
65 * Leffler, et al.: The Design and Implementation of the 4.3BSD
66 * UNIX Operating System (Addison Welley, 1989)
69 * Arguably, to simplify accounting operations, this mechanism should
70 * be replaced by one in which an accounting log file (similar to /dev/klog)
71 * is read by a user process, etc. However, that has its own problems.
75 * Internal accounting functions.
76 * The former's operation is described in Leffler, et al., and the latter
77 * was provided by UCB with the 4.4BSD-Lite release
79 static comp_t
encode_comp_t (u_long
, u_long
);
80 static void acctwatch (void *);
83 * Accounting callout handle used for periodic scheduling of
86 static struct callout acctwatch_handle
;
89 * Accounting vnode pointer, and saved vnode pointer.
91 static struct vnode
*acctp
;
92 static struct vnode
*savacctp
;
95 * Values associated with enabling and disabling accounting
97 static int acctsuspend
= 2; /* stop accounting when < 2% free space left */
98 SYSCTL_INT(_kern
, OID_AUTO
, acct_suspend
, CTLFLAG_RW
,
99 &acctsuspend
, 0, "Percentage of free disk space below which accounting will be suspended");
101 static int acctresume
= 4; /* resume when free space risen to > 4% */
102 SYSCTL_INT(_kern
, OID_AUTO
, acct_resume
, CTLFLAG_RW
,
103 &acctresume
, 0, "Minimum percentage of free disk space needed to resume accounting");
105 static int acctchkfreq
= 15; /* frequency (in seconds) to check space */
106 SYSCTL_INT(_kern
, OID_AUTO
, acct_chkfreq
, CTLFLAG_RW
,
107 &acctchkfreq
, 0, "Frequency (in seconds) of free disk space checking");
110 acct_init(void *arg __unused
)
112 callout_init(&acctwatch_handle
);
114 SYSINIT(acct
, SI_SUB_DRIVERS
, SI_ORDER_ANY
, acct_init
, NULL
);
117 * Accounting system call. Written based on the specification and
118 * previous implementation done by Mark Tinguely.
123 sys_acct(struct acct_args
*uap
)
125 struct thread
*td
= curthread
;
126 struct nlookupdata nd
;
130 /* Make sure that the caller is root. */
131 error
= priv_check(td
, PRIV_ROOT
);
136 * If accounting is to be started to a file, open that file for
137 * appending and make sure it's a 'normal'.
139 if (uap
->path
!= NULL
) {
140 error
= nlookup_init(&nd
, uap
->path
, UIO_USERSPACE
,
143 error
= vn_open(&nd
, NULL
, FWRITE
| O_APPEND
, 0);
144 if (error
== 0 && nd
.nl_open_vp
->v_type
!= VREG
)
151 nd
.nl_open_vp
= NULL
;
160 * If accounting was previously enabled, kill the old space-watcher,
163 if (acctp
!= NULLVP
|| savacctp
!= NULLVP
) {
164 callout_stop(&acctwatch_handle
);
165 error
= vn_close((acctp
!= NULLVP
? acctp
: savacctp
),
167 acctp
= savacctp
= NULLVP
;
171 * If no new file opened then leave. We never did an nlookup so
172 * don't try cleaning it up.
174 if (uap
->path
== NULL
)
178 * Save the new accounting file vnode, and schedule the new
179 * free space watcher.
187 * Write out process accounting information, on process exit.
188 * Data to be written out is specified in Leffler, et al.
189 * and are enumerated below. (They're also noted in the system
190 * "acct.h" header file.)
194 acct_process(struct proc
*p
)
204 /* If accounting isn't enabled, don't bother */
210 * Get process accounting information.
213 /* (1) The name of the command that ran */
214 bcopy(p
->p_comm
, acct
.ac_comm
, sizeof acct
.ac_comm
);
216 /* (2) The amount of user and system time that was used */
218 acct
.ac_utime
= encode_comp_t(ru
.ru_utime
.tv_sec
, ru
.ru_utime
.tv_usec
);
219 acct
.ac_stime
= encode_comp_t(ru
.ru_stime
.tv_sec
, ru
.ru_stime
.tv_usec
);
221 /* (3) The elapsed time the commmand ran (and its starting time) */
222 acct
.ac_btime
= p
->p_start
.tv_sec
;
224 timevalsub(&tmp
, &p
->p_start
);
225 acct
.ac_etime
= encode_comp_t(tmp
.tv_sec
, tmp
.tv_usec
);
227 /* (4) The average amount of memory used */
230 timevaladd(&tmp
, &ru
.ru_stime
);
231 t
= tmp
.tv_sec
* hz
+ tmp
.tv_usec
/ tick
;
233 acct
.ac_mem
= (r
->ru_ixrss
+ r
->ru_idrss
+ r
->ru_isrss
) / t
;
237 /* (5) The number of disk I/O operations done */
238 acct
.ac_io
= encode_comp_t(r
->ru_inblock
+ r
->ru_oublock
, 0);
240 /* (6) The UID and GID of the process */
241 acct
.ac_uid
= p
->p_ucred
->cr_ruid
;
242 acct
.ac_gid
= p
->p_ucred
->cr_rgid
;
244 /* (7) The terminal from which the process was started */
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
);
248 acct
.ac_tty
= NOUDEV
;
250 /* (8) The boolean flags that tell how the process terminated, etc. */
251 acct
.ac_flag
= p
->p_acflag
;
254 * Eliminate any file size rlimit.
256 rlim
.rlim_cur
= RLIM_INFINITY
;
257 rlim
.rlim_max
= RLIM_INFINITY
;
258 plimit_modify(p
, RLIMIT_FSIZE
, &rlim
);
261 * Write the accounting information to the file.
263 return (vn_rdwr(UIO_WRITE
, vp
, (caddr_t
)&acct
, sizeof (acct
),
264 (off_t
)0, UIO_SYSSPACE
, IO_APPEND
|IO_UNIT
, p
->p_ucred
,
269 * Encode_comp_t converts from ticks in seconds and microseconds
270 * to ticks in 1/AHZ seconds. The encoding is described in
271 * Leffler, et al., on page 63.
274 #define MANTSIZE 13 /* 13 bit mantissa. */
275 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
276 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
279 encode_comp_t(u_long s
, u_long us
)
286 s
+= us
/ (1000000 / AHZ
); /* Maximize precision. */
288 while (s
> MAXFRACT
) {
289 rnd
= s
& (1 << (EXPSIZE
- 1)); /* Round up? */
290 s
>>= EXPSIZE
; /* Base 8 exponent == 3 bit shift. */
294 /* If we need to round up, do it (and handle overflow correctly). */
295 if (rnd
&& (++s
> MAXFRACT
)) {
300 /* Clean it up and polish it off. */
301 exp
<<= MANTSIZE
; /* Shift the exponent into place */
302 exp
+= s
; /* and add on the mantissa. */
307 * Periodically check the file system to see if accounting
308 * should be turned on or off. Beware the case where the vnode
309 * has been vgone()'d out from underneath us, e.g. when the file
310 * system containing the accounting file has been forcibly unmounted.
318 if (savacctp
!= NULLVP
) {
319 if (savacctp
->v_type
== VBAD
) {
320 (void) vn_close(savacctp
, FWRITE
| O_APPEND
);
324 (void)VFS_STATFS(savacctp
->v_mount
, &sb
, proc0
.p_ucred
);
325 if (sb
.f_bavail
> acctresume
* sb
.f_blocks
/ 100) {
328 log(LOG_NOTICE
, "Accounting resumed\n");
333 if (acctp
->v_type
== VBAD
) {
334 (void) vn_close(acctp
, FWRITE
| O_APPEND
);
338 (void)VFS_STATFS(acctp
->v_mount
, &sb
, proc0
.p_ucred
);
339 if (sb
.f_bavail
<= acctsuspend
* sb
.f_blocks
/ 100) {
342 log(LOG_NOTICE
, "Accounting suspended\n");
345 callout_reset(&acctwatch_handle
, acctchkfreq
* hz
, acctwatch
, NULL
);