xref: /freebsd/sys/kern/kern_acct.c (revision 4cf49a43559ed9fdad601bdcccd2c55963008675)
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/sysproto.h>
46 #include <sys/proc.h>
47 #include <sys/mount.h>
48 #include <sys/vnode.h>
49 #include <sys/fcntl.h>
50 #include <sys/syslog.h>
51 #include <sys/kernel.h>
52 #include <sys/sysent.h>
53 #include <sys/sysctl.h>
54 #include <sys/namei.h>
55 #include <sys/acct.h>
56 #include <sys/resourcevar.h>
57 #include <sys/tty.h>
58 
59 /*
60  * The routines implemented in this file are described in:
61  *      Leffler, et al.: The Design and Implementation of the 4.3BSD
62  *	    UNIX Operating System (Addison Welley, 1989)
63  * on pages 62-63.
64  *
65  * Arguably, to simplify accounting operations, this mechanism should
66  * be replaced by one in which an accounting log file (similar to /dev/klog)
67  * is read by a user process, etc.  However, that has its own problems.
68  */
69 
70 /*
71  * Internal accounting functions.
72  * The former's operation is described in Leffler, et al., and the latter
73  * was provided by UCB with the 4.4BSD-Lite release
74  */
75 static comp_t	encode_comp_t __P((u_long, u_long));
76 static void	acctwatch __P((void *));
77 
78 /*
79  * Accounting callout handle used for periodic scheduling of
80  * acctwatch.
81  */
82 static struct	callout_handle acctwatch_handle
83     = CALLOUT_HANDLE_INITIALIZER(&acctwatch_handle);
84 
85 /*
86  * Accounting vnode pointer, and saved vnode pointer.
87  */
88 static struct	vnode *acctp;
89 static struct	vnode *savacctp;
90 
91 /*
92  * Values associated with enabling and disabling accounting
93  */
94 static int acctsuspend = 2;	/* stop accounting when < 2% free space left */
95 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
96 	&acctsuspend, 0, "");
97 
98 static int acctresume = 4;	/* resume when free space risen to > 4% */
99 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
100 	&acctresume, 0, "");
101 
102 static int acctchkfreq = 15;	/* frequency (in seconds) to check space */
103 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW,
104 	&acctchkfreq, 0, "");
105 
106 /*
107  * Accounting system call.  Written based on the specification and
108  * previous implementation done by Mark Tinguely.
109  */
110 int
111 acct(a1, uap)
112 	struct proc *a1;
113 	struct acct_args /* {
114 		syscallarg(char *) path;
115 	} */ *uap;
116 {
117 	struct proc *p = curproc;	/* XXX */
118 	struct nameidata nd;
119 	int error;
120 
121 	/* Make sure that the caller is root. */
122 	error = suser(p);
123 	if (error)
124 		return (error);
125 
126 	/*
127 	 * If accounting is to be started to a file, open that file for
128 	 * writing and make sure it's a 'normal'.
129 	 */
130 	if (SCARG(uap, path) != NULL) {
131 		NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path),
132 		       p);
133 		error = vn_open(&nd, FWRITE, 0);
134 		if (error)
135 			return (error);
136 		VOP_UNLOCK(nd.ni_vp, 0, p);
137 		if (nd.ni_vp->v_type != VREG) {
138 			vn_close(nd.ni_vp, FWRITE, p->p_ucred, p);
139 			return (EACCES);
140 		}
141 	}
142 
143 	/*
144 	 * If accounting was previously enabled, kill the old space-watcher,
145 	 * close the file, and (if no new file was specified, leave).
146 	 */
147 	if (acctp != NULLVP || savacctp != NULLVP) {
148 		untimeout(acctwatch, NULL, acctwatch_handle);
149 		error = vn_close((acctp != NULLVP ? acctp : savacctp), FWRITE,
150 		    p->p_ucred, p);
151 		acctp = savacctp = NULLVP;
152 	}
153 	if (SCARG(uap, path) == NULL)
154 		return (error);
155 
156 	/*
157 	 * Save the new accounting file vnode, and schedule the new
158 	 * free space watcher.
159 	 */
160 	acctp = nd.ni_vp;
161 	acctwatch(NULL);
162 	return (error);
163 }
164 
165 /*
166  * Write out process accounting information, on process exit.
167  * Data to be written out is specified in Leffler, et al.
168  * and are enumerated below.  (They're also noted in the system
169  * "acct.h" header file.)
170  */
171 
172 int
173 acct_process(p)
174 	struct proc *p;
175 {
176 	struct acct acct;
177 	struct rusage *r;
178 	struct timeval ut, st, tmp;
179 	int t;
180 	struct vnode *vp;
181 
182 	/* If accounting isn't enabled, don't bother */
183 	vp = acctp;
184 	if (vp == NULLVP)
185 		return (0);
186 
187 	/*
188 	 * Get process accounting information.
189 	 */
190 
191 	/* (1) The name of the command that ran */
192 	bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
193 
194 	/* (2) The amount of user and system time that was used */
195 	calcru(p, &ut, &st, NULL);
196 	acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
197 	acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
198 
199 	/* (3) The elapsed time the commmand ran (and its starting time) */
200 	acct.ac_btime = p->p_stats->p_start.tv_sec;
201 	microtime(&tmp);
202 	timevalsub(&tmp, &p->p_stats->p_start);
203 	acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
204 
205 	/* (4) The average amount of memory used */
206 	r = &p->p_stats->p_ru;
207 	tmp = ut;
208 	timevaladd(&tmp, &st);
209 	t = tmp.tv_sec * hz + tmp.tv_usec / tick;
210 	if (t)
211 		acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
212 	else
213 		acct.ac_mem = 0;
214 
215 	/* (5) The number of disk I/O operations done */
216 	acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
217 
218 	/* (6) The UID and GID of the process */
219 	acct.ac_uid = p->p_cred->p_ruid;
220 	acct.ac_gid = p->p_cred->p_rgid;
221 
222 	/* (7) The terminal from which the process was started */
223 	if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
224 		acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
225 	else
226 		acct.ac_tty = NOUDEV;
227 
228 	/* (8) The boolean flags that tell how the process terminated, etc. */
229 	acct.ac_flag = p->p_acflag;
230 
231 	/*
232 	 * Eliminate any file size rlimit.
233 	 */
234 	if (p->p_limit->p_refcnt > 1 &&
235 	    (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
236 		p->p_limit->p_refcnt--;
237 		p->p_limit = limcopy(p->p_limit);
238 	}
239 	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
240 
241 	/*
242 	 * Write the accounting information to the file.
243 	 */
244 	VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
245 	return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
246 	    (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred,
247 	    (int *)0, p));
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 static comp_t
261 encode_comp_t(s, us)
262 	u_long s, us;
263 {
264 	int exp, rnd;
265 
266 	exp = 0;
267 	rnd = 0;
268 	s *= AHZ;
269 	s += us / (1000000 / AHZ);	/* Maximize precision. */
270 
271 	while (s > MAXFRACT) {
272 	rnd = s & (1 << (EXPSIZE - 1));	/* Round up? */
273 		s >>= EXPSIZE;		/* Base 8 exponent == 3 bit shift. */
274 		exp++;
275 	}
276 
277 	/* If we need to round up, do it (and handle overflow correctly). */
278 	if (rnd && (++s > MAXFRACT)) {
279 		s >>= EXPSIZE;
280 		exp++;
281 	}
282 
283 	/* Clean it up and polish it off. */
284 	exp <<= MANTSIZE;		/* Shift the exponent into place */
285 	exp += s;			/* and add on the mantissa. */
286 	return (exp);
287 }
288 
289 /*
290  * Periodically check the file system to see if accounting
291  * should be turned on or off.  Beware the case where the vnode
292  * has been vgone()'d out from underneath us, e.g. when the file
293  * system containing the accounting file has been forcibly unmounted.
294  */
295 /* ARGSUSED */
296 static void
297 acctwatch(a)
298 	void *a;
299 {
300 	struct statfs sb;
301 
302 	if (savacctp != NULLVP) {
303 		if (savacctp->v_type == VBAD) {
304 			(void) vn_close(savacctp, FWRITE, NOCRED, NULL);
305 			savacctp = NULLVP;
306 			return;
307 		}
308 		(void)VFS_STATFS(savacctp->v_mount, &sb, (struct proc *)0);
309 		if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
310 			acctp = savacctp;
311 			savacctp = NULLVP;
312 			log(LOG_NOTICE, "Accounting resumed\n");
313 		}
314 	} else {
315 		if (acctp == NULLVP)
316 			return;
317 		if (acctp->v_type == VBAD) {
318 			(void) vn_close(acctp, FWRITE, NOCRED, NULL);
319 			acctp = NULLVP;
320 			return;
321 		}
322 		(void)VFS_STATFS(acctp->v_mount, &sb, (struct proc *)0);
323 		if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
324 			savacctp = acctp;
325 			acctp = NULLVP;
326 			log(LOG_NOTICE, "Accounting suspended\n");
327 		}
328 	}
329 	acctwatch_handle = timeout(acctwatch, NULL, acctchkfreq * hz);
330 }
331