xref: /freebsd/sys/kern/kern_acct.c (revision 729362425c09cf6b362366aabc6fb547eee8035a)
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 "opt_mac.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/lock.h>
48 #include <sys/mutex.h>
49 #include <sys/sysproto.h>
50 #include <sys/proc.h>
51 #include <sys/mac.h>
52 #include <sys/mount.h>
53 #include <sys/vnode.h>
54 #include <sys/fcntl.h>
55 #include <sys/syslog.h>
56 #include <sys/kernel.h>
57 #include <sys/sysent.h>
58 #include <sys/sysctl.h>
59 #include <sys/namei.h>
60 #include <sys/acct.h>
61 #include <sys/resourcevar.h>
62 #include <sys/tty.h>
63 
64 /*
65  * The routines implemented in this file are described in:
66  *      Leffler, et al.: The Design and Implementation of the 4.3BSD
67  *	    UNIX Operating System (Addison Welley, 1989)
68  * on pages 62-63.
69  *
70  * Arguably, to simplify accounting operations, this mechanism should
71  * be replaced by one in which an accounting log file (similar to /dev/klog)
72  * is read by a user process, etc.  However, that has its own problems.
73  */
74 
75 /*
76  * Internal accounting functions.
77  * The former's operation is described in Leffler, et al., and the latter
78  * was provided by UCB with the 4.4BSD-Lite release
79  */
80 static comp_t	encode_comp_t(u_long, u_long);
81 static void	acctwatch(void *);
82 
83 /*
84  * Accounting callout used for periodic scheduling of acctwatch.
85  */
86 static struct	callout acctwatch_callout;
87 
88 /*
89  * Accounting vnode pointer, saved vnode pointer, and flags for each.
90  */
91 static struct	vnode *acctp;
92 static struct	ucred *acctcred;
93 static int	acctflags;
94 static struct	vnode *savacctp;
95 static struct	ucred *savacctcred;
96 static int	savacctflags;
97 
98 static struct mtx	acct_mtx;
99 MTX_SYSINIT(acct, &acct_mtx, "accounting", MTX_DEF);
100 
101 /*
102  * Values associated with enabling and disabling accounting
103  */
104 static int acctsuspend = 2;	/* stop accounting when < 2% free space left */
105 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
106 	&acctsuspend, 0, "percentage of free disk space below which accounting stops");
107 
108 static int acctresume = 4;	/* resume when free space risen to > 4% */
109 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
110 	&acctresume, 0, "percentage of free disk space above which accounting resumes");
111 
112 static int acctchkfreq = 15;	/* frequency (in seconds) to check space */
113 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW,
114 	&acctchkfreq, 0, "frequency for checking the free space");
115 
116 /*
117  * Accounting system call.  Written based on the specification and
118  * previous implementation done by Mark Tinguely.
119  *
120  * MPSAFE
121  */
122 int
123 acct(td, uap)
124 	struct thread *td;
125 	struct acct_args /* {
126 		char *path;
127 	} */ *uap;
128 {
129 	struct nameidata nd;
130 	int error, flags;
131 
132 	/* Make sure that the caller is root. */
133 	error = suser(td);
134 	if (error)
135 		return (error);
136 
137 	mtx_lock(&Giant);
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 		NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
144 		flags = FWRITE | O_APPEND;
145 		error = vn_open(&nd, &flags, 0);
146 		if (error)
147 			goto done2;
148 		NDFREE(&nd, NDF_ONLY_PNBUF);
149 #ifdef MAC
150 		error = mac_check_system_acct(td->td_ucred, nd.ni_vp);
151 		if (error) {
152 			vn_close(nd.ni_vp, flags, td->td_ucred, td);
153 			goto done2;
154 		}
155 #endif
156 		VOP_UNLOCK(nd.ni_vp, 0, td);
157 		if (nd.ni_vp->v_type != VREG) {
158 			vn_close(nd.ni_vp, flags, td->td_ucred, td);
159 			error = EACCES;
160 			goto done2;
161 		}
162 #ifdef MAC
163 	} else {
164 		error = mac_check_system_acct(td->td_ucred, NULL);
165 		if (error)
166 			goto done2;
167 #endif
168 	}
169 
170 	/*
171 	 * If accounting was previously enabled, kill the old space-watcher,
172 	 * close the file, and (if no new file was specified, leave).
173 	 */
174 
175 	/*
176 	 * XXX arr: Should not hold lock over vnode operation.
177 	 */
178 
179 	mtx_lock(&acct_mtx);
180 	if (acctp != NULLVP || savacctp != NULLVP) {
181 		callout_stop(&acctwatch_callout);
182 		error = vn_close((acctp != NULLVP ? acctp : savacctp),
183 		    (acctp != NULLVP ? acctflags : savacctflags),
184 		    (acctcred != NOCRED ? acctcred : savacctcred), td);
185 		acctp = savacctp = NULLVP;
186 		crfree(acctcred != NOCRED ? acctcred : savacctcred);
187 		acctcred = savacctcred = NOCRED;
188 	}
189 	if (uap->path == NULL) {
190 		mtx_unlock(&acct_mtx);
191 		goto done2;
192 	}
193 
194 	/*
195 	 * Save the new accounting file vnode, and schedule the new
196 	 * free space watcher.
197 	 */
198 	acctp = nd.ni_vp;
199 	acctcred = crhold(td->td_ucred);
200 	acctflags = flags;
201 	callout_init(&acctwatch_callout, 0);
202 	mtx_unlock(&acct_mtx);
203 	acctwatch(NULL);
204 done2:
205 	mtx_unlock(&Giant);
206 	return (error);
207 }
208 
209 /*
210  * Write out process accounting information, on process exit.
211  * Data to be written out is specified in Leffler, et al.
212  * and are enumerated below.  (They're also noted in the system
213  * "acct.h" header file.)
214  */
215 
216 int
217 acct_process(td)
218 	struct thread *td;
219 {
220 	struct proc *p = td->td_proc;
221 	struct acct acct;
222 	struct rusage *r;
223 	struct timeval ut, st, tmp;
224 	int t, ret;
225 	struct vnode *vp;
226 	struct ucred *uc;
227 
228 	mtx_lock(&acct_mtx);
229 
230 	/* If accounting isn't enabled, don't bother */
231 	vp = acctp;
232 	if (vp == NULLVP) {
233 		mtx_unlock(&acct_mtx);
234 		return (0);
235 	}
236 
237 	/*
238 	 * Get process accounting information.
239 	 */
240 
241 	/* (1) The name of the command that ran */
242 	bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
243 
244 	/* (2) The amount of user and system time that was used */
245 	mtx_lock_spin(&sched_lock);
246 	calcru(p, &ut, &st, NULL);
247 	mtx_unlock_spin(&sched_lock);
248 	acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
249 	acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
250 
251 	/* (3) The elapsed time the command ran (and its starting time) */
252 	acct.ac_btime = p->p_stats->p_start.tv_sec;
253 	microtime(&tmp);
254 	timevalsub(&tmp, &p->p_stats->p_start);
255 	acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
256 
257 	/* (4) The average amount of memory used */
258 	r = &p->p_stats->p_ru;
259 	tmp = ut;
260 	timevaladd(&tmp, &st);
261 	t = tmp.tv_sec * hz + tmp.tv_usec / tick;
262 	if (t)
263 		acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
264 	else
265 		acct.ac_mem = 0;
266 
267 	/* (5) The number of disk I/O operations done */
268 	acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
269 
270 	/* (6) The UID and GID of the process */
271 	acct.ac_uid = p->p_ucred->cr_ruid;
272 	acct.ac_gid = p->p_ucred->cr_rgid;
273 
274 	/* (7) The terminal from which the process was started */
275 	PROC_LOCK(p);
276 	SESS_LOCK(p->p_session);
277 	if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
278 		acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
279 	else
280 		acct.ac_tty = NOUDEV;
281 	SESS_UNLOCK(p->p_session);
282 	PROC_UNLOCK(p);
283 
284 	/* (8) The boolean flags that tell how the process terminated, etc. */
285 	acct.ac_flag = p->p_acflag;
286 
287 	/*
288 	 * Write the accounting information to the file.
289 	 */
290 	uc = crhold(acctcred);
291 	vref(vp);
292 	mtx_unlock(&acct_mtx);
293 
294 	/*
295 	 * Eliminate any file size rlimit.
296 	 */
297 	if (p->p_limit->p_refcnt > 1) {
298 		p->p_limit->p_refcnt--;
299 		p->p_limit = limcopy(p->p_limit);
300 	}
301 	p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
302 
303 	VOP_LEASE(vp, td, uc, LEASE_WRITE);
304 	ret = vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
305 	    (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, uc, NOCRED,
306 	    (int *)0, td);
307 	vrele(vp);
308 	crfree(uc);
309 	return (ret);
310 }
311 
312 /*
313  * Encode_comp_t converts from ticks in seconds and microseconds
314  * to ticks in 1/AHZ seconds.  The encoding is described in
315  * Leffler, et al., on page 63.
316  */
317 
318 #define	MANTSIZE	13			/* 13 bit mantissa. */
319 #define	EXPSIZE		3			/* Base 8 (3 bit) exponent. */
320 #define	MAXFRACT	((1 << MANTSIZE) - 1)	/* Maximum fractional value. */
321 
322 static comp_t
323 encode_comp_t(s, us)
324 	u_long s, us;
325 {
326 	int exp, rnd;
327 
328 	exp = 0;
329 	rnd = 0;
330 	s *= AHZ;
331 	s += us / (1000000 / AHZ);	/* Maximize precision. */
332 
333 	while (s > MAXFRACT) {
334 	rnd = s & (1 << (EXPSIZE - 1));	/* Round up? */
335 		s >>= EXPSIZE;		/* Base 8 exponent == 3 bit shift. */
336 		exp++;
337 	}
338 
339 	/* If we need to round up, do it (and handle overflow correctly). */
340 	if (rnd && (++s > MAXFRACT)) {
341 		s >>= EXPSIZE;
342 		exp++;
343 	}
344 
345 	/* Clean it up and polish it off. */
346 	exp <<= MANTSIZE;		/* Shift the exponent into place */
347 	exp += s;			/* and add on the mantissa. */
348 	return (exp);
349 }
350 
351 /*
352  * Periodically check the filesystem to see if accounting
353  * should be turned on or off.  Beware the case where the vnode
354  * has been vgone()'d out from underneath us, e.g. when the file
355  * system containing the accounting file has been forcibly unmounted.
356  */
357 /* ARGSUSED */
358 static void
359 acctwatch(a)
360 	void *a;
361 {
362 	struct statfs sb;
363 
364 	mtx_lock(&acct_mtx);
365 
366 	/*
367 	 * XXX arr: Need to fix the issue of holding acct_mtx over
368 	 * the below vnode operations.
369 	 */
370 
371 	if (savacctp != NULLVP) {
372 		if (savacctp->v_type == VBAD) {
373 			(void) vn_close(savacctp, savacctflags, savacctcred,
374 			    NULL);
375 			savacctp = NULLVP;
376 			savacctcred = NOCRED;
377 			mtx_unlock(&acct_mtx);
378 			return;
379 		}
380 		(void)VFS_STATFS(savacctp->v_mount, &sb, (struct thread *)0);
381 		if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
382 			acctp = savacctp;
383 			acctcred = savacctcred;
384 			acctflags = savacctflags;
385 			savacctp = NULLVP;
386 			savacctcred = NOCRED;
387 			log(LOG_NOTICE, "Accounting resumed\n");
388 		}
389 	} else {
390 		if (acctp == NULLVP) {
391 			mtx_unlock(&acct_mtx);
392 			return;
393 		}
394 		if (acctp->v_type == VBAD) {
395 			(void) vn_close(acctp, acctflags, acctcred, NULL);
396 			acctp = NULLVP;
397 			crfree(acctcred);
398 			acctcred = NOCRED;
399 			mtx_unlock(&acct_mtx);
400 			return;
401 		}
402 		(void)VFS_STATFS(acctp->v_mount, &sb, (struct thread *)0);
403 		if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
404 			savacctp = acctp;
405 			savacctflags = acctflags;
406 			savacctcred = acctcred;
407 			acctp = NULLVP;
408 			acctcred = NOCRED;
409 			log(LOG_NOTICE, "Accounting suspended\n");
410 		}
411 	}
412 	callout_reset(&acctwatch_callout, acctchkfreq * hz, acctwatch, NULL);
413 	mtx_unlock(&acct_mtx);
414 }
415