xref: /dragonfly/sys/kern/kern_acct.c (revision 52f9f0d9)
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 #include <sys/mplock2.h>
64 
65 /*
66  * The routines implemented in this file are described in:
67  *      Leffler, et al.: The Design and Implementation of the 4.3BSD
68  *	    UNIX Operating System (Addison Welley, 1989)
69  * on pages 62-63.
70  *
71  * Arguably, to simplify accounting operations, this mechanism should
72  * be replaced by one in which an accounting log file (similar to /dev/klog)
73  * is read by a user process, etc.  However, that has its own problems.
74  */
75 
76 /*
77  * Internal accounting functions.
78  * The former's operation is described in Leffler, et al., and the latter
79  * was provided by UCB with the 4.4BSD-Lite release
80  */
81 static comp_t	encode_comp_t (u_long, u_long);
82 static void	acctwatch (void *);
83 
84 /*
85  * Accounting callout handle used for periodic scheduling of
86  * acctwatch.
87  */
88 static struct	callout acctwatch_handle;
89 
90 /*
91  * Accounting vnode pointer, and saved vnode pointer.
92  */
93 static struct	vnode *acctp;
94 static struct	vnode *savacctp;
95 
96 /*
97  * Values associated with enabling and disabling accounting
98  */
99 static int acctsuspend = 2;	/* stop accounting when < 2% free space left */
100 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
101 	&acctsuspend, 0, "Percentage of free disk space below which accounting will be suspended");
102 
103 static int acctresume = 4;	/* resume when free space risen to > 4% */
104 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
105 	&acctresume, 0, "Minimum percentage of free disk space needed to resume accounting");
106 
107 static int acctchkfreq = 15;	/* frequency (in seconds) to check space */
108 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW,
109 	&acctchkfreq, 0, "Frequency (in seconds) of free disk space checking");
110 
111 static void
112 acct_init(void *arg __unused)
113 {
114 	callout_init(&acctwatch_handle);
115 }
116 SYSINIT(acct, SI_SUB_DRIVERS, SI_ORDER_ANY, acct_init, NULL);
117 
118 /*
119  * Accounting system call.  Written based on the specification and
120  * previous implementation done by Mark Tinguely.
121  *
122  * acct(char *path)
123  *
124  * MPALMOSTSAFE
125  */
126 int
127 sys_acct(struct acct_args *uap)
128 {
129 	struct thread *td = curthread;
130 	struct nlookupdata nd;
131 	struct vnode *vp;
132 	int error;
133 
134 	error = priv_check(td, PRIV_ACCT);
135 	if (error)
136 		return (error);
137 
138 	get_mplock();
139 
140 	/*
141 	 * If accounting is to be started to a file, open that file for
142 	 * appending and make sure it's a 'normal'.
143 	 */
144 	if (uap->path != NULL) {
145 		error = nlookup_init(&nd, uap->path, UIO_USERSPACE,
146 					NLC_LOCKVP);
147 		if (error == 0)
148 			error = vn_open(&nd, NULL, FWRITE | O_APPEND, 0);
149 		if (error == 0 && nd.nl_open_vp->v_type != VREG)
150 			error = EACCES;
151 		if (error) {
152 			nlookup_done(&nd);
153 			goto done;
154 		}
155 		vp = nd.nl_open_vp;
156 		nd.nl_open_vp = NULL;
157 		nlookup_done(&nd);
158 
159 		vn_unlock(vp);
160 	} else {
161 		vp = NULL;
162 	}
163 
164 	/*
165 	 * If accounting was previously enabled, kill the old space-watcher,
166 	 * close the file.
167 	 */
168 	if (acctp != NULLVP || savacctp != NULLVP) {
169 		callout_stop(&acctwatch_handle);
170 		error = vn_close((acctp != NULLVP ? acctp : savacctp),
171 				FWRITE | O_APPEND);
172 		acctp = savacctp = NULLVP;
173 	}
174 
175 	/*
176 	 * If no new file opened then leave.  We never did an nlookup so
177 	 * don't try cleaning it up.
178 	 */
179 	if (uap->path == NULL)
180 		goto done;
181 
182 	/*
183 	 * Save the new accounting file vnode, and schedule the new
184 	 * free space watcher.
185 	 */
186 	acctp = vp;
187 	acctwatch(NULL);
188 done:
189 	rel_mplock();
190 	return (error);
191 }
192 
193 /*
194  * Write out process accounting information, on process exit.
195  * Data to be written out is specified in Leffler, et al.
196  * and are enumerated below.  (They're also noted in the system
197  * "acct.h" header file.)
198  */
199 
200 int
201 acct_process(struct proc *p)
202 {
203 	struct acct acct;
204 	struct rusage *r;
205 	struct rusage ru;
206 	struct timeval tmp;
207 	struct rlimit rlim;
208 	int t;
209 	struct vnode *vp;
210 
211 	/* If accounting isn't enabled, don't bother */
212 	vp = acctp;
213 	if (vp == NULLVP)
214 		return (0);
215 
216 	/*
217 	 * Get process accounting information.
218 	 */
219 
220 	/* (1) The name of the command that ran */
221 	bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
222 
223 	/* (2) The amount of user and system time that was used */
224 	calcru_proc(p, &ru);
225 	acct.ac_utime = encode_comp_t(ru.ru_utime.tv_sec, ru.ru_utime.tv_usec);
226 	acct.ac_stime = encode_comp_t(ru.ru_stime.tv_sec, ru.ru_stime.tv_usec);
227 
228 	/* (3) The elapsed time the commmand ran (and its starting time) */
229 	acct.ac_btime = p->p_start.tv_sec;
230 	microtime(&tmp);
231 	timevalsub(&tmp, &p->p_start);
232 	acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
233 
234 	/* (4) The average amount of memory used */
235 	r = &p->p_ru;
236 	tmp = ru.ru_utime;;
237 	timevaladd(&tmp, &ru.ru_stime);
238 	t = tmp.tv_sec * hz + tmp.tv_usec / ustick;
239 	if (t)
240 		acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
241 	else
242 		acct.ac_mem = 0;
243 
244 	/* (5) The number of disk I/O operations done */
245 	acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
246 
247 	/* (6) The UID and GID of the process */
248 	acct.ac_uid = p->p_ucred->cr_ruid;
249 	acct.ac_gid = p->p_ucred->cr_rgid;
250 
251 	/* (7) The terminal from which the process was started */
252 	if ((p->p_flags & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
253 		acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
254 	else
255 		acct.ac_tty = NOUDEV;
256 
257 	/* (8) The boolean flags that tell how the process terminated, etc. */
258 	acct.ac_flag = p->p_acflag;
259 
260 	/*
261 	 * Eliminate any file size rlimit.
262 	 */
263 	rlim.rlim_cur = RLIM_INFINITY;
264 	rlim.rlim_max = RLIM_INFINITY;
265 	plimit_modify(p, RLIMIT_FSIZE, &rlim);
266 
267 	/*
268 	 * Write the accounting information to the file.
269 	 */
270 	return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
271 	    (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred,
272 	    NULL));
273 }
274 
275 /*
276  * Encode_comp_t converts from ticks in seconds and microseconds
277  * to ticks in 1/AHZ seconds.  The encoding is described in
278  * Leffler, et al., on page 63.
279  */
280 
281 #define	MANTSIZE	13			/* 13 bit mantissa. */
282 #define	EXPSIZE		3			/* Base 8 (3 bit) exponent. */
283 #define	MAXFRACT	((1 << MANTSIZE) - 1)	/* Maximum fractional value. */
284 
285 static comp_t
286 encode_comp_t(u_long s, u_long us)
287 {
288 	int exp, rnd;
289 
290 	exp = 0;
291 	rnd = 0;
292 	s *= AHZ;
293 	s += us / (1000000 / AHZ);	/* Maximize precision. */
294 
295 	while (s > MAXFRACT) {
296 	rnd = s & (1 << (EXPSIZE - 1));	/* Round up? */
297 		s >>= EXPSIZE;		/* Base 8 exponent == 3 bit shift. */
298 		exp++;
299 	}
300 
301 	/* If we need to round up, do it (and handle overflow correctly). */
302 	if (rnd && (++s > MAXFRACT)) {
303 		s >>= EXPSIZE;
304 		exp++;
305 	}
306 
307 	/* Clean it up and polish it off. */
308 	exp <<= MANTSIZE;		/* Shift the exponent into place */
309 	exp += s;			/* and add on the mantissa. */
310 	return (exp);
311 }
312 
313 /*
314  * Periodically check the file system to see if accounting
315  * should be turned on or off.  Beware the case where the vnode
316  * has been vgone()'d out from underneath us, e.g. when the file
317  * system containing the accounting file has been forcibly unmounted.
318  */
319 /* ARGSUSED */
320 static void
321 acctwatch(void *a)
322 {
323 	struct statfs sb;
324 
325 	if (savacctp != NULLVP) {
326 		if (savacctp->v_type == VBAD) {
327 			(void) vn_close(savacctp, FWRITE | O_APPEND);
328 			savacctp = NULLVP;
329 			return;
330 		}
331 		(void)VFS_STATFS(savacctp->v_mount, &sb, proc0.p_ucred);
332 		if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
333 			acctp = savacctp;
334 			savacctp = NULLVP;
335 			log(LOG_NOTICE, "Accounting resumed\n");
336 		}
337 	} else {
338 		if (acctp == NULLVP)
339 			return;
340 		if (acctp->v_type == VBAD) {
341 			(void) vn_close(acctp, FWRITE | O_APPEND);
342 			acctp = NULLVP;
343 			return;
344 		}
345 		(void)VFS_STATFS(acctp->v_mount, &sb, proc0.p_ucred);
346 		if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
347 			savacctp = acctp;
348 			acctp = NULLVP;
349 			log(LOG_NOTICE, "Accounting suspended\n");
350 		}
351 	}
352 	callout_reset(&acctwatch_handle, acctchkfreq * hz, acctwatch, NULL);
353 }
354