xref: /freebsd/sys/kern/kern_acct.c (revision f0adf7f5cdd241db2f2c817683191a6ef64a4e95)
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  */
41 
42 #include <sys/cdefs.h>
43 __FBSDID("$FreeBSD$");
44 
45 #include "opt_mac.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/lock.h>
50 #include <sys/mutex.h>
51 #include <sys/sysproto.h>
52 #include <sys/proc.h>
53 #include <sys/mac.h>
54 #include <sys/mount.h>
55 #include <sys/vnode.h>
56 #include <sys/fcntl.h>
57 #include <sys/syslog.h>
58 #include <sys/kernel.h>
59 #include <sys/sysent.h>
60 #include <sys/sysctl.h>
61 #include <sys/namei.h>
62 #include <sys/acct.h>
63 #include <sys/resourcevar.h>
64 #include <sys/tty.h>
65 
66 /*
67  * The routines implemented in this file are described in:
68  *      Leffler, et al.: The Design and Implementation of the 4.3BSD
69  *	    UNIX Operating System (Addison Welley, 1989)
70  * on pages 62-63.
71  *
72  * Arguably, to simplify accounting operations, this mechanism should
73  * be replaced by one in which an accounting log file (similar to /dev/klog)
74  * is read by a user process, etc.  However, that has its own problems.
75  */
76 
77 /*
78  * Internal accounting functions.
79  * The former's operation is described in Leffler, et al., and the latter
80  * was provided by UCB with the 4.4BSD-Lite release
81  */
82 static comp_t	encode_comp_t(u_long, u_long);
83 static void	acctwatch(void *);
84 
85 /*
86  * Accounting callout used for periodic scheduling of acctwatch.
87  */
88 static struct	callout acctwatch_callout;
89 
90 /*
91  * Accounting vnode pointer, saved vnode pointer, and flags for each.
92  */
93 static struct	vnode *acctp;
94 static struct	ucred *acctcred;
95 static int	acctflags;
96 static struct	vnode *savacctp;
97 static struct	ucred *savacctcred;
98 static int	savacctflags;
99 
100 static struct mtx	acct_mtx;
101 MTX_SYSINIT(acct, &acct_mtx, "accounting", MTX_DEF);
102 
103 /*
104  * Values associated with enabling and disabling accounting
105  */
106 static int acctsuspend = 2;	/* stop accounting when < 2% free space left */
107 SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
108 	&acctsuspend, 0, "percentage of free disk space below which accounting stops");
109 
110 static int acctresume = 4;	/* resume when free space risen to > 4% */
111 SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
112 	&acctresume, 0, "percentage of free disk space above which accounting resumes");
113 
114 static int acctchkfreq = 15;	/* frequency (in seconds) to check space */
115 SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW,
116 	&acctchkfreq, 0, "frequency for checking the free space");
117 
118 /*
119  * Accounting system call.  Written based on the specification and
120  * previous implementation done by Mark Tinguely.
121  *
122  * MPSAFE
123  */
124 int
125 acct(td, uap)
126 	struct thread *td;
127 	struct acct_args /* {
128 		char *path;
129 	} */ *uap;
130 {
131 	struct nameidata nd;
132 	int error, flags;
133 
134 	/* Make sure that the caller is root. */
135 	error = suser(td);
136 	if (error)
137 		return (error);
138 
139 	mtx_lock(&Giant);
140 
141 	/*
142 	 * If accounting is to be started to a file, open that file for
143 	 * appending and make sure it's a 'normal'.
144 	 */
145 	if (uap->path != NULL) {
146 		NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->path, td);
147 		flags = FWRITE | O_APPEND;
148 		error = vn_open(&nd, &flags, 0, -1);
149 		if (error)
150 			goto done2;
151 		NDFREE(&nd, NDF_ONLY_PNBUF);
152 #ifdef MAC
153 		error = mac_check_system_acct(td->td_ucred, nd.ni_vp);
154 		if (error) {
155 			vn_close(nd.ni_vp, flags, td->td_ucred, td);
156 			goto done2;
157 		}
158 #endif
159 		VOP_UNLOCK(nd.ni_vp, 0, td);
160 		if (nd.ni_vp->v_type != VREG) {
161 			vn_close(nd.ni_vp, flags, td->td_ucred, td);
162 			error = EACCES;
163 			goto done2;
164 		}
165 #ifdef MAC
166 	} else {
167 		error = mac_check_system_acct(td->td_ucred, NULL);
168 		if (error)
169 			goto done2;
170 #endif
171 	}
172 
173 	mtx_lock(&acct_mtx);
174 
175 	/*
176 	 * If accounting was previously enabled, kill the old space-watcher,
177 	 * close the file, and (if no new file was specified, leave).
178 	 *
179 	 * XXX arr: should not hold lock over vnode operation.
180 	 */
181 	if (acctp != NULLVP || savacctp != NULLVP) {
182 		callout_stop(&acctwatch_callout);
183 		error = vn_close((acctp != NULLVP ? acctp : savacctp),
184 		    (acctp != NULLVP ? acctflags : savacctflags),
185 		    (acctcred != NOCRED ? acctcred : savacctcred), td);
186 		acctp = savacctp = NULLVP;
187 		crfree(acctcred != NOCRED ? acctcred : savacctcred);
188 		acctcred = savacctcred = NOCRED;
189 		log(LOG_NOTICE, "Accounting disabled\n");
190 	}
191 	if (uap->path == NULL) {
192 		mtx_unlock(&acct_mtx);
193 		goto done2;
194 	}
195 
196 	/*
197 	 * Save the new accounting file vnode, and schedule the new
198 	 * free space watcher.
199 	 */
200 	acctp = nd.ni_vp;
201 	acctcred = crhold(td->td_ucred);
202 	acctflags = flags;
203 	callout_init(&acctwatch_callout, 0);
204 	mtx_unlock(&acct_mtx);
205 	log(LOG_NOTICE, "Accounting enabled\n");
206 	acctwatch(NULL);
207 
208 done2:
209 	mtx_unlock(&Giant);
210 	return (error);
211 }
212 
213 /*
214  * Write out process accounting information, on process exit.
215  * Data to be written out is specified in Leffler, et al.
216  * and are enumerated below.  (They're also noted in the system
217  * "acct.h" header file.)
218  */
219 int
220 acct_process(td)
221 	struct thread *td;
222 {
223 	struct acct acct;
224 	struct timeval ut, st, tmp;
225 	struct plimit *newlim, *oldlim;
226 	struct proc *p;
227 	struct rusage *r;
228 	struct ucred *uc;
229 	struct vnode *vp;
230 	int t, ret;
231 
232 	mtx_lock(&acct_mtx);
233 
234 	/* If accounting isn't enabled, don't bother */
235 	vp = acctp;
236 	if (vp == NULLVP) {
237 		mtx_unlock(&acct_mtx);
238 		return (0);
239 	}
240 
241 	p = td->td_proc;
242 
243 	/*
244 	 * Get process accounting information.
245 	 */
246 
247 	PROC_LOCK(p);
248 	/* (1) The name of the command that ran */
249 	bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
250 
251 	/* (2) The amount of user and system time that was used */
252 	mtx_lock_spin(&sched_lock);
253 	calcru(p, &ut, &st, NULL);
254 	mtx_unlock_spin(&sched_lock);
255 	acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
256 	acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
257 
258 	/* (3) The elapsed time the command ran (and its starting time) */
259 	tmp = boottime;
260 	timevaladd(&tmp, &p->p_stats->p_start);
261 	acct.ac_btime = tmp.tv_sec;
262 	microuptime(&tmp);
263 	timevalsub(&tmp, &p->p_stats->p_start);
264 	acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
265 
266 	/* (4) The average amount of memory used */
267 	r = &p->p_stats->p_ru;
268 	tmp = ut;
269 	timevaladd(&tmp, &st);
270 	t = tmp.tv_sec * hz + tmp.tv_usec / tick;
271 	if (t)
272 		acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
273 	else
274 		acct.ac_mem = 0;
275 
276 	/* (5) The number of disk I/O operations done */
277 	acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
278 
279 	/* (6) The UID and GID of the process */
280 	acct.ac_uid = p->p_ucred->cr_ruid;
281 	acct.ac_gid = p->p_ucred->cr_rgid;
282 
283 	/* (7) The terminal from which the process was started */
284 	SESS_LOCK(p->p_session);
285 	if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
286 		acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
287 	else
288 		acct.ac_tty = NODEV;
289 	SESS_UNLOCK(p->p_session);
290 
291 	/* (8) The boolean flags that tell how the process terminated, etc. */
292 	acct.ac_flag = p->p_acflag;
293 	PROC_UNLOCK(p);
294 
295 	/*
296 	 * Finish doing things that require acct_mtx, and release acct_mtx.
297 	 */
298 	uc = crhold(acctcred);
299 	vref(vp);
300 	mtx_unlock(&acct_mtx);
301 
302 	/*
303 	 * Eliminate any file size rlimit.
304 	 */
305 	newlim = lim_alloc();
306 	PROC_LOCK(p);
307 	oldlim = p->p_limit;
308 	lim_copy(newlim, oldlim);
309 	newlim->pl_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
310 	p->p_limit = newlim;
311 	PROC_UNLOCK(p);
312 	lim_free(oldlim);
313 
314 	/*
315 	 * Write the accounting information to the file.
316 	 */
317 	VOP_LEASE(vp, td, uc, LEASE_WRITE);
318 	ret = vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
319 	    (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, uc, NOCRED,
320 	    (int *)0, td);
321 	vrele(vp);
322 	crfree(uc);
323 	return (ret);
324 }
325 
326 /*
327  * Encode_comp_t converts from ticks in seconds and microseconds
328  * to ticks in 1/AHZ seconds.  The encoding is described in
329  * Leffler, et al., on page 63.
330  */
331 
332 #define	MANTSIZE	13			/* 13 bit mantissa. */
333 #define	EXPSIZE		3			/* Base 8 (3 bit) exponent. */
334 #define	MAXFRACT	((1 << MANTSIZE) - 1)	/* Maximum fractional value. */
335 
336 static comp_t
337 encode_comp_t(s, us)
338 	u_long s, us;
339 {
340 	int exp, rnd;
341 
342 	exp = 0;
343 	rnd = 0;
344 	s *= AHZ;
345 	s += us / (1000000 / AHZ);	/* Maximize precision. */
346 
347 	while (s > MAXFRACT) {
348 	rnd = s & (1 << (EXPSIZE - 1));	/* Round up? */
349 		s >>= EXPSIZE;		/* Base 8 exponent == 3 bit shift. */
350 		exp++;
351 	}
352 
353 	/* If we need to round up, do it (and handle overflow correctly). */
354 	if (rnd && (++s > MAXFRACT)) {
355 		s >>= EXPSIZE;
356 		exp++;
357 	}
358 
359 	/* Clean it up and polish it off. */
360 	exp <<= MANTSIZE;		/* Shift the exponent into place */
361 	exp += s;			/* and add on the mantissa. */
362 	return (exp);
363 }
364 
365 /*
366  * Periodically check the filesystem to see if accounting
367  * should be turned on or off.  Beware the case where the vnode
368  * has been vgone()'d out from underneath us, e.g. when the file
369  * system containing the accounting file has been forcibly unmounted.
370  */
371 /* ARGSUSED */
372 static void
373 acctwatch(a)
374 	void *a;
375 {
376 	struct statfs sb;
377 
378 	mtx_lock(&acct_mtx);
379 
380 	/*
381 	 * XXX arr: need to fix the issue of holding acct_mtx over
382 	 * the below vnode operations.
383 	 */
384 	if (savacctp != NULLVP) {
385 		if (savacctp->v_type == VBAD) {
386 			(void) vn_close(savacctp, savacctflags, savacctcred,
387 			    NULL);
388 			savacctp = NULLVP;
389 			savacctcred = NOCRED;
390 			mtx_unlock(&acct_mtx);
391 			return;
392 		}
393 		(void)VFS_STATFS(savacctp->v_mount, &sb, (struct thread *)0);
394 		if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
395 			acctp = savacctp;
396 			acctcred = savacctcred;
397 			acctflags = savacctflags;
398 			savacctp = NULLVP;
399 			savacctcred = NOCRED;
400 			log(LOG_NOTICE, "Accounting resumed\n");
401 		}
402 	} else {
403 		if (acctp == NULLVP) {
404 			mtx_unlock(&acct_mtx);
405 			return;
406 		}
407 		if (acctp->v_type == VBAD) {
408 			(void) vn_close(acctp, acctflags, acctcred, NULL);
409 			acctp = NULLVP;
410 			crfree(acctcred);
411 			acctcred = NOCRED;
412 			mtx_unlock(&acct_mtx);
413 			return;
414 		}
415 		(void)VFS_STATFS(acctp->v_mount, &sb, (struct thread *)0);
416 		if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
417 			savacctp = acctp;
418 			savacctflags = acctflags;
419 			savacctcred = acctcred;
420 			acctp = NULLVP;
421 			acctcred = NOCRED;
422 			log(LOG_NOTICE, "Accounting suspended\n");
423 		}
424 	}
425 	callout_reset(&acctwatch_callout, acctchkfreq * hz, acctwatch, NULL);
426 	mtx_unlock(&acct_mtx);
427 }
428