kern_acct.c revision 1.38 1 /* $NetBSD: kern_acct.c,v 1.38 1994/12/24 15:07:22 cgd Exp $ */
2
3 /*-
4 * Copyright (c) 1994 Christopher G. Demetriou
5 * Copyright (c) 1982, 1986, 1989, 1993
6 * The Regents of the University of California. All rights reserved.
7 * (c) UNIX System Laboratories, Inc.
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. All advertising materials mentioning features or use of this software
22 * must display the following acknowledgement:
23 * This product includes software developed by the University of
24 * California, Berkeley and its contributors.
25 * 4. Neither the name of the University nor the names of its contributors
26 * may be used to endorse or promote products derived from this software
27 * without specific prior written permission.
28 *
29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
39 * SUCH DAMAGE.
40 *
41 * @(#)kern_acct.c 8.1 (Berkeley) 6/14/93
42 */
43
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/proc.h>
47 #include <sys/mount.h>
48 #include <sys/vnode.h>
49 #include <sys/file.h>
50 #include <sys/syslog.h>
51 #include <sys/kernel.h>
52 #include <sys/namei.h>
53 #include <sys/errno.h>
54 #include <sys/acct.h>
55 #include <sys/resourcevar.h>
56 #include <sys/ioctl.h>
57 #include <sys/tty.h>
58
59 #include <sys/syscallargs.h>
60
61 /*
62 * The routines implemented in this file are described in:
63 * Leffler, et al.: The Design and Implementation of the 4.3BSD
64 * UNIX Operating System (Addison Welley, 1989)
65 * on pages 62-63.
66 *
67 * Arguably, to simplify accounting operations, this mechanism should
68 * be replaced by one in which an accounting log file (similar to /dev/klog)
69 * is read by a user process, etc. However, that has its own problems.
70 */
71
72 /*
73 * Internal accounting functions.
74 * The former's operation is described in Leffler, et al., and the latter
75 * was provided by UCB with the 4.4BSD-Lite release
76 */
77 comp_t encode_comp_t __P((u_long, u_long));
78 void acctwatch __P((void *));
79
80 /*
81 * Accounting vnode pointer, and saved vnode pointer.
82 */
83 struct vnode *acctp;
84 struct vnode *savacctp;
85
86 /*
87 * Values associated with enabling and disabling accounting
88 */
89 int acctsuspend = 2; /* stop accounting when < 2% free space left */
90 int acctresume = 4; /* resume when free space risen to > 4% */
91 int acctchkfreq = 15; /* frequency (in seconds) to check space */
92
93 /*
94 * Accounting system call. Written based on the specification and
95 * previous implementation done by Mark Tinguely.
96 */
97 int
98 acct(p, uap, retval)
99 struct proc *p;
100 struct acct_args /* {
101 syscallarg(char *) path;
102 } */ *uap;
103 register_t *retval;
104 {
105 struct nameidata nd;
106 int error;
107
108 /* Make sure that the caller is root. */
109 if (error = suser(p->p_ucred, &p->p_acflag))
110 return (error);
111
112 /*
113 * If accounting is to be started to a file, open that file for
114 * writing and make sure it's a 'normal'.
115 */
116 if (SCARG(uap, path) != NULL) {
117 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path),
118 p);
119 if (error = vn_open(&nd, FWRITE, 0))
120 return (error);
121 VOP_UNLOCK(nd.ni_vp);
122 if (nd.ni_vp->v_type != VREG) {
123 vn_close(nd.ni_vp, FWRITE, p->p_ucred, p);
124 return (EACCES);
125 }
126 }
127
128 /*
129 * If accounting was previously enabled, kill the old space-watcher,
130 * close the file, and (if no new file was specified, leave).
131 */
132 if (acctp != NULLVP || savacctp != NULLVP) {
133 untimeout(acctwatch, NULL);
134 error = vn_close((acctp != NULLVP ? acctp : savacctp), FWRITE,
135 p->p_ucred, p);
136 acctp = savacctp = NULLVP;
137 }
138 if (SCARG(uap, path) == NULL)
139 return (error);
140
141 /*
142 * Save the new accounting file vnode, and schedule the new
143 * free space watcher.
144 */
145 acctp = nd.ni_vp;
146 acctwatch(NULL);
147 return (error);
148 }
149
150 /*
151 * Write out process accounting information, on process exit.
152 * Data to be written out is specified in Leffler, et al.
153 * and are enumerated below. (They're also noted in the system
154 * "acct.h" header file.)
155 */
156 int
157 acct_process(p)
158 struct proc *p;
159 {
160 struct acct acct;
161 struct rusage *r;
162 struct timeval ut, st, tmp;
163 int s, t;
164 struct vnode *vp;
165
166 /* If accounting isn't enabled, don't bother */
167 vp = acctp;
168 if (vp == NULLVP)
169 return (0);
170
171 /*
172 * Get process accounting information.
173 */
174
175 /* (1) The name of the command that ran */
176 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
177
178 /* (2) The amount of user and system time that was used */
179 calcru(p, &ut, &st, NULL);
180 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
181 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
182
183 /* (3) The elapsed time the commmand ran (and its starting time) */
184 acct.ac_btime = p->p_stats->p_start.tv_sec;
185 s = splclock();
186 tmp = time;
187 splx(s);
188 __timersub(&tmp, &p->p_stats->p_start);
189 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
190
191 /* (4) The average amount of memory used */
192 r = &p->p_stats->p_ru;
193 tmp = ut;
194 __timeradd(&tmp, &st);
195 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
196 if (t)
197 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
198 else
199 acct.ac_mem = 0;
200
201 /* (5) The number of disk I/O operations done */
202 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
203
204 /* (6) The UID and GID of the process */
205 acct.ac_uid = p->p_cred->p_ruid;
206 acct.ac_gid = p->p_cred->p_rgid;
207
208 /* (7) The terminal from which the process was started */
209 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
210 acct.ac_tty = p->p_pgrp->pg_session->s_ttyp->t_dev;
211 else
212 acct.ac_tty = NODEV;
213
214 /* (8) The boolean flags that tell how the process terminated, etc. */
215 acct.ac_flag = p->p_acflag;
216
217 /*
218 * Now, just write the accounting information to the file.
219 */
220 VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
221 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
222 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred,
223 (int *)0, p));
224 }
225
226 /*
227 * Encode_comp_t converts from ticks in seconds and microseconds
228 * to ticks in 1/AHZ seconds. The encoding is described in
229 * Leffler, et al., on page 63.
230 */
231
232 #define MANTSIZE 13 /* 13 bit mantissa. */
233 #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
234 #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
235
236 comp_t
237 encode_comp_t(s, us)
238 u_long s, us;
239 {
240 int exp, rnd;
241
242 exp = 0;
243 rnd = 0;
244 s *= AHZ;
245 s += us / (1000000 / AHZ); /* Maximize precision. */
246
247 while (s > MAXFRACT) {
248 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
249 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
250 exp++;
251 }
252
253 /* If we need to round up, do it (and handle overflow correctly). */
254 if (rnd && (++s > MAXFRACT)) {
255 s >>= EXPSIZE;
256 exp++;
257 }
258
259 /* Clean it up and polish it off. */
260 exp <<= MANTSIZE; /* Shift the exponent into place */
261 exp += s; /* and add on the mantissa. */
262 return (exp);
263 }
264
265 /*
266 * Periodically check the file system to see if accounting
267 * should be turned on or off. Beware the case where the vnode
268 * has been vgone()'d out from underneath us, e.g. when the file
269 * system containing the accounting file has been forcibly unmounted.
270 */
271 /* ARGSUSED */
272 void
273 acctwatch(a)
274 void *a;
275 {
276 struct statfs sb;
277
278 if (savacctp != NULLVP) {
279 if (savacctp->v_type == VBAD) {
280 (void) vn_close(savacctp, FWRITE, NOCRED, NULL);
281 savacctp = NULLVP;
282 return;
283 }
284 (void)VFS_STATFS(savacctp->v_mount, &sb, (struct proc *)0);
285 if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
286 acctp = savacctp;
287 savacctp = NULLVP;
288 log(LOG_NOTICE, "Accounting resumed\n");
289 }
290 } else if (acctp != NULLVP) {
291 if (acctp->v_type == VBAD) {
292 (void) vn_close(acctp, FWRITE, NOCRED, NULL);
293 acctp = NULLVP;
294 return;
295 }
296 (void)VFS_STATFS(acctp->v_mount, &sb, (struct proc *)0);
297 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
298 savacctp = acctp;
299 acctp = NULLVP;
300 log(LOG_NOTICE, "Accounting suspended\n");
301 }
302 } else
303 return;
304 timeout(acctwatch, NULL, acctchkfreq * hz);
305 }
306