xref: /dflybsd-src/sys/kern/kern_acct.c (revision 92b817bb6b0d8e22d601698e9d3864697def6575)
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: 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 $
42  */
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/sysproto.h>
47 #include <sys/proc.h>
48 #include <sys/priv.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>
57 #include <sys/acct.h>
58 #include <sys/resourcevar.h>
59 #include <sys/tty.h>
60 
61 #include <vm/vm_zone.h>
62 
63 /*
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)
67  * on pages 62-63.
68  *
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.
72  */
73 
74 /*
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
78  */
79 static comp_t	encode_comp_t (u_long, u_long);
80 static void	acctwatch (void *);
81 
82 /*
83  * Accounting callout handle used for periodic scheduling of
84  * acctwatch.
85  */
86 static struct	callout acctwatch_handle;
87 
88 /*
89  * Accounting vnode pointer, and saved vnode pointer.
90  */
91 static struct	vnode *acctp;
92 static struct	vnode *savacctp;
93 
94 /*
95  * Values associated with enabling and disabling accounting
96  */
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");
100 
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");
104 
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");
108 
109 static void
110 acct_init(void *arg __unused)
111 {
112 	callout_init(&acctwatch_handle);
113 }
114 SYSINIT(acct, SI_SUB_DRIVERS, SI_ORDER_ANY, acct_init, NULL);
115 
116 /*
117  * Accounting system call.  Written based on the specification and
118  * previous implementation done by Mark Tinguely.
119  *
120  * acct(char *path)
121  *
122  * MPALMOSTSAFE
123  */
124 int
125 sys_acct(struct acct_args *uap)
126 {
127 	struct thread *td = curthread;
128 	struct nlookupdata nd;
129 	struct vnode *vp;
130 	int error;
131 
132 	error = priv_check(td, PRIV_ACCT);
133 	if (error)
134 		return (error);
135 
136 	get_mplock();
137 
138 	/*
139 	 * If accounting is to be started to a file, open that file for
140 	 * appending and make sure it's a 'normal'.
141 	 */
142 	if (uap->path != NULL) {
143 		error = nlookup_init(&nd, uap->path, UIO_USERSPACE,
144 					NLC_LOCKVP);
145 		if (error == 0)
146 			error = vn_open(&nd, NULL, FWRITE | O_APPEND, 0);
147 		if (error == 0 && nd.nl_open_vp->v_type != VREG)
148 			error = EACCES;
149 		if (error) {
150 			nlookup_done(&nd);
151 			goto done;
152 		}
153 		vp = nd.nl_open_vp;
154 		nd.nl_open_vp = NULL;
155 		nlookup_done(&nd);
156 
157 		vn_unlock(vp);
158 	} else {
159 		vp = NULL;
160 	}
161 
162 	/*
163 	 * If accounting was previously enabled, kill the old space-watcher,
164 	 * close the file.
165 	 */
166 	if (acctp != NULLVP || savacctp != NULLVP) {
167 		callout_stop(&acctwatch_handle);
168 		error = vn_close((acctp != NULLVP ? acctp : savacctp),
169 				FWRITE | O_APPEND);
170 		acctp = savacctp = NULLVP;
171 	}
172 
173 	/*
174 	 * If no new file opened then leave.  We never did an nlookup so
175 	 * don't try cleaning it up.
176 	 */
177 	if (uap->path == NULL)
178 		goto done;
179 
180 	/*
181 	 * Save the new accounting file vnode, and schedule the new
182 	 * free space watcher.
183 	 */
184 	acctp = vp;
185 	acctwatch(NULL);
186 done:
187 	rel_mplock();
188 	return (error);
189 }
190 
191 /*
192  * Write out process accounting information, on process exit.
193  * Data to be written out is specified in Leffler, et al.
194  * and are enumerated below.  (They're also noted in the system
195  * "acct.h" header file.)
196  */
197 
198 int
199 acct_process(struct proc *p)
200 {
201 	struct acct acct;
202 	struct rusage *r;
203 	struct rusage ru;
204 	struct timeval tmp;
205 	struct rlimit rlim;
206 	int t;
207 	struct vnode *vp;
208 
209 	/* If accounting isn't enabled, don't bother */
210 	vp = acctp;
211 	if (vp == NULLVP)
212 		return (0);
213 
214 	/*
215 	 * Get process accounting information.
216 	 */
217 
218 	/* (1) The name of the command that ran */
219 	bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
220 
221 	/* (2) The amount of user and system time that was used */
222 	calcru_proc(p, &ru);
223 	acct.ac_utime = encode_comp_t(ru.ru_utime.tv_sec, ru.ru_utime.tv_usec);
224 	acct.ac_stime = encode_comp_t(ru.ru_stime.tv_sec, ru.ru_stime.tv_usec);
225 
226 	/* (3) The elapsed time the commmand ran (and its starting time) */
227 	acct.ac_btime = p->p_start.tv_sec;
228 	microtime(&tmp);
229 	timevalsub(&tmp, &p->p_start);
230 	acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
231 
232 	/* (4) The average amount of memory used */
233 	r = &p->p_ru;
234 	tmp = ru.ru_utime;;
235 	timevaladd(&tmp, &ru.ru_stime);
236 	t = tmp.tv_sec * hz + tmp.tv_usec / tick;
237 	if (t)
238 		acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
239 	else
240 		acct.ac_mem = 0;
241 
242 	/* (5) The number of disk I/O operations done */
243 	acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
244 
245 	/* (6) The UID and GID of the process */
246 	acct.ac_uid = p->p_ucred->cr_ruid;
247 	acct.ac_gid = p->p_ucred->cr_rgid;
248 
249 	/* (7) The terminal from which the process was started */
250 	if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
251 		acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
252 	else
253 		acct.ac_tty = NOUDEV;
254 
255 	/* (8) The boolean flags that tell how the process terminated, etc. */
256 	acct.ac_flag = p->p_acflag;
257 
258 	/*
259 	 * Eliminate any file size rlimit.
260 	 */
261 	rlim.rlim_cur = RLIM_INFINITY;
262 	rlim.rlim_max = RLIM_INFINITY;
263 	plimit_modify(p, RLIMIT_FSIZE, &rlim);
264 
265 	/*
266 	 * Write the accounting information to the file.
267 	 */
268 	return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
269 	    (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred,
270 	    NULL));
271 }
272 
273 /*
274  * Encode_comp_t converts from ticks in seconds and microseconds
275  * to ticks in 1/AHZ seconds.  The encoding is described in
276  * Leffler, et al., on page 63.
277  */
278 
279 #define	MANTSIZE	13			/* 13 bit mantissa. */
280 #define	EXPSIZE		3			/* Base 8 (3 bit) exponent. */
281 #define	MAXFRACT	((1 << MANTSIZE) - 1)	/* Maximum fractional value. */
282 
283 static comp_t
284 encode_comp_t(u_long s, u_long us)
285 {
286 	int exp, rnd;
287 
288 	exp = 0;
289 	rnd = 0;
290 	s *= AHZ;
291 	s += us / (1000000 / AHZ);	/* Maximize precision. */
292 
293 	while (s > MAXFRACT) {
294 	rnd = s & (1 << (EXPSIZE - 1));	/* Round up? */
295 		s >>= EXPSIZE;		/* Base 8 exponent == 3 bit shift. */
296 		exp++;
297 	}
298 
299 	/* If we need to round up, do it (and handle overflow correctly). */
300 	if (rnd && (++s > MAXFRACT)) {
301 		s >>= EXPSIZE;
302 		exp++;
303 	}
304 
305 	/* Clean it up and polish it off. */
306 	exp <<= MANTSIZE;		/* Shift the exponent into place */
307 	exp += s;			/* and add on the mantissa. */
308 	return (exp);
309 }
310 
311 /*
312  * Periodically check the file system to see if accounting
313  * should be turned on or off.  Beware the case where the vnode
314  * has been vgone()'d out from underneath us, e.g. when the file
315  * system containing the accounting file has been forcibly unmounted.
316  */
317 /* ARGSUSED */
318 static void
319 acctwatch(void *a)
320 {
321 	struct statfs sb;
322 
323 	if (savacctp != NULLVP) {
324 		if (savacctp->v_type == VBAD) {
325 			(void) vn_close(savacctp, FWRITE | O_APPEND);
326 			savacctp = NULLVP;
327 			return;
328 		}
329 		(void)VFS_STATFS(savacctp->v_mount, &sb, proc0.p_ucred);
330 		if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
331 			acctp = savacctp;
332 			savacctp = NULLVP;
333 			log(LOG_NOTICE, "Accounting resumed\n");
334 		}
335 	} else {
336 		if (acctp == NULLVP)
337 			return;
338 		if (acctp->v_type == VBAD) {
339 			(void) vn_close(acctp, FWRITE | O_APPEND);
340 			acctp = NULLVP;
341 			return;
342 		}
343 		(void)VFS_STATFS(acctp->v_mount, &sb, proc0.p_ucred);
344 		if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
345 			savacctp = acctp;
346 			acctp = NULLVP;
347 			log(LOG_NOTICE, "Accounting suspended\n");
348 		}
349 	}
350 	callout_reset(&acctwatch_handle, acctchkfreq * hz, acctwatch, NULL);
351 }
352