proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[dragonfly.git] / sys / kern / kern_acct.c
CommitLineData
984263bc
MD
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 $
dadab5e9 41 * $DragonFly: src/sys/kern/kern_acct.c,v 1.4 2003/06/25 03:55:57 dillon Exp $
984263bc
MD
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/mount.h>
49#include <sys/vnode.h>
50#include <sys/fcntl.h>
51#include <sys/syslog.h>
52#include <sys/kernel.h>
53#include <sys/sysent.h>
54#include <sys/sysctl.h>
55#include <sys/namei.h>
56#include <sys/acct.h>
57#include <sys/resourcevar.h>
58#include <sys/tty.h>
59
60#include <vm/vm_zone.h>
61
62/*
63 * The routines implemented in this file are described in:
64 * Leffler, et al.: The Design and Implementation of the 4.3BSD
65 * UNIX Operating System (Addison Welley, 1989)
66 * on pages 62-63.
67 *
68 * Arguably, to simplify accounting operations, this mechanism should
69 * be replaced by one in which an accounting log file (similar to /dev/klog)
70 * is read by a user process, etc. However, that has its own problems.
71 */
72
73/*
74 * Internal accounting functions.
75 * The former's operation is described in Leffler, et al., and the latter
76 * was provided by UCB with the 4.4BSD-Lite release
77 */
78static comp_t encode_comp_t __P((u_long, u_long));
79static void acctwatch __P((void *));
80
81/*
82 * Accounting callout handle used for periodic scheduling of
83 * acctwatch.
84 */
85static struct callout_handle acctwatch_handle
86 = CALLOUT_HANDLE_INITIALIZER(&acctwatch_handle);
87
88/*
89 * Accounting vnode pointer, and saved vnode pointer.
90 */
91static struct vnode *acctp;
92static struct vnode *savacctp;
93
94/*
95 * Values associated with enabling and disabling accounting
96 */
97static int acctsuspend = 2; /* stop accounting when < 2% free space left */
98SYSCTL_INT(_kern, OID_AUTO, acct_suspend, CTLFLAG_RW,
99 &acctsuspend, 0, "");
100
101static int acctresume = 4; /* resume when free space risen to > 4% */
102SYSCTL_INT(_kern, OID_AUTO, acct_resume, CTLFLAG_RW,
103 &acctresume, 0, "");
104
105static int acctchkfreq = 15; /* frequency (in seconds) to check space */
106SYSCTL_INT(_kern, OID_AUTO, acct_chkfreq, CTLFLAG_RW,
107 &acctchkfreq, 0, "");
108
109/*
110 * Accounting system call. Written based on the specification and
111 * previous implementation done by Mark Tinguely.
112 */
113int
41c20dac 114acct(uap)
984263bc
MD
115 struct acct_args /* {
116 syscallarg(char *) path;
117 } */ *uap;
118{
dadab5e9
MD
119 struct thread *td = curthread;
120 struct proc *p = td->td_proc; /* XXX */
984263bc
MD
121 struct nameidata nd;
122 int error;
123
124 /* Make sure that the caller is root. */
dadab5e9 125 error = suser(td);
984263bc
MD
126 if (error)
127 return (error);
128
129 /*
130 * If accounting is to be started to a file, open that file for
131 * appending and make sure it's a 'normal'.
132 */
133 if (SCARG(uap, path) != NULL) {
dadab5e9
MD
134 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE,
135 SCARG(uap, path), td);
984263bc
MD
136 error = vn_open(&nd, FWRITE | O_APPEND, 0);
137 if (error)
138 return (error);
139 NDFREE(&nd, NDF_ONLY_PNBUF);
dadab5e9 140 VOP_UNLOCK(nd.ni_vp, 0, td);
984263bc 141 if (nd.ni_vp->v_type != VREG) {
dadab5e9 142 vn_close(nd.ni_vp, FWRITE | O_APPEND, p->p_ucred, td);
984263bc
MD
143 return (EACCES);
144 }
145 }
146
147 /*
148 * If accounting was previously enabled, kill the old space-watcher,
149 * close the file, and (if no new file was specified, leave).
150 */
151 if (acctp != NULLVP || savacctp != NULLVP) {
152 untimeout(acctwatch, NULL, acctwatch_handle);
153 error = vn_close((acctp != NULLVP ? acctp : savacctp),
dadab5e9 154 FWRITE | O_APPEND, p->p_ucred, td);
984263bc
MD
155 acctp = savacctp = NULLVP;
156 }
157 if (SCARG(uap, path) == NULL)
158 return (error);
159
160 /*
161 * Save the new accounting file vnode, and schedule the new
162 * free space watcher.
163 */
164 acctp = nd.ni_vp;
165 acctwatch(NULL);
166 return (error);
167}
168
169/*
170 * Write out process accounting information, on process exit.
171 * Data to be written out is specified in Leffler, et al.
172 * and are enumerated below. (They're also noted in the system
173 * "acct.h" header file.)
174 */
175
176int
dadab5e9 177acct_process(struct proc *p)
984263bc
MD
178{
179 struct acct acct;
180 struct rusage *r;
181 struct timeval ut, st, tmp;
182 int t;
183 struct vnode *vp;
dadab5e9 184 struct thread *td = p->p_thread;
984263bc
MD
185
186 /* If accounting isn't enabled, don't bother */
187 vp = acctp;
188 if (vp == NULLVP)
189 return (0);
190
191 /*
192 * Get process accounting information.
193 */
194
195 /* (1) The name of the command that ran */
196 bcopy(p->p_comm, acct.ac_comm, sizeof acct.ac_comm);
197
198 /* (2) The amount of user and system time that was used */
199 calcru(p, &ut, &st, NULL);
200 acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
201 acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
202
203 /* (3) The elapsed time the commmand ran (and its starting time) */
204 acct.ac_btime = p->p_stats->p_start.tv_sec;
205 microtime(&tmp);
206 timevalsub(&tmp, &p->p_stats->p_start);
207 acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
208
209 /* (4) The average amount of memory used */
210 r = &p->p_stats->p_ru;
211 tmp = ut;
212 timevaladd(&tmp, &st);
213 t = tmp.tv_sec * hz + tmp.tv_usec / tick;
214 if (t)
215 acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
216 else
217 acct.ac_mem = 0;
218
219 /* (5) The number of disk I/O operations done */
220 acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
221
222 /* (6) The UID and GID of the process */
41c20dac
MD
223 acct.ac_uid = p->p_ucred->cr_ruid;
224 acct.ac_gid = p->p_ucred->cr_rgid;
984263bc
MD
225
226 /* (7) The terminal from which the process was started */
227 if ((p->p_flag & P_CONTROLT) && p->p_pgrp->pg_session->s_ttyp)
228 acct.ac_tty = dev2udev(p->p_pgrp->pg_session->s_ttyp->t_dev);
229 else
230 acct.ac_tty = NOUDEV;
231
232 /* (8) The boolean flags that tell how the process terminated, etc. */
233 acct.ac_flag = p->p_acflag;
234
235 /*
236 * Eliminate any file size rlimit.
237 */
238 if (p->p_limit->p_refcnt > 1 &&
239 (p->p_limit->p_lflags & PL_SHAREMOD) == 0) {
240 p->p_limit->p_refcnt--;
241 p->p_limit = limcopy(p->p_limit);
242 }
243 p->p_rlimit[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
244
245 /*
246 * Write the accounting information to the file.
247 */
dadab5e9 248 VOP_LEASE(vp, td, p->p_ucred, LEASE_WRITE);
984263bc
MD
249 return (vn_rdwr(UIO_WRITE, vp, (caddr_t)&acct, sizeof (acct),
250 (off_t)0, UIO_SYSSPACE, IO_APPEND|IO_UNIT, p->p_ucred,
dadab5e9 251 (int *)0, td));
984263bc
MD
252}
253
254/*
255 * Encode_comp_t converts from ticks in seconds and microseconds
256 * to ticks in 1/AHZ seconds. The encoding is described in
257 * Leffler, et al., on page 63.
258 */
259
260#define MANTSIZE 13 /* 13 bit mantissa. */
261#define EXPSIZE 3 /* Base 8 (3 bit) exponent. */
262#define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */
263
264static comp_t
265encode_comp_t(s, us)
266 u_long s, us;
267{
268 int exp, rnd;
269
270 exp = 0;
271 rnd = 0;
272 s *= AHZ;
273 s += us / (1000000 / AHZ); /* Maximize precision. */
274
275 while (s > MAXFRACT) {
276 rnd = s & (1 << (EXPSIZE - 1)); /* Round up? */
277 s >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */
278 exp++;
279 }
280
281 /* If we need to round up, do it (and handle overflow correctly). */
282 if (rnd && (++s > MAXFRACT)) {
283 s >>= EXPSIZE;
284 exp++;
285 }
286
287 /* Clean it up and polish it off. */
288 exp <<= MANTSIZE; /* Shift the exponent into place */
289 exp += s; /* and add on the mantissa. */
290 return (exp);
291}
292
293/*
294 * Periodically check the file system to see if accounting
295 * should be turned on or off. Beware the case where the vnode
296 * has been vgone()'d out from underneath us, e.g. when the file
297 * system containing the accounting file has been forcibly unmounted.
298 */
299/* ARGSUSED */
300static void
301acctwatch(a)
302 void *a;
303{
304 struct statfs sb;
305
306 if (savacctp != NULLVP) {
307 if (savacctp->v_type == VBAD) {
308 (void) vn_close(savacctp, FWRITE | O_APPEND, NOCRED,
309 NULL);
310 savacctp = NULLVP;
311 return;
312 }
dadab5e9 313 (void)VFS_STATFS(savacctp->v_mount, &sb, NULL);
984263bc
MD
314 if (sb.f_bavail > acctresume * sb.f_blocks / 100) {
315 acctp = savacctp;
316 savacctp = NULLVP;
317 log(LOG_NOTICE, "Accounting resumed\n");
318 }
319 } else {
320 if (acctp == NULLVP)
321 return;
322 if (acctp->v_type == VBAD) {
323 (void) vn_close(acctp, FWRITE | O_APPEND, NOCRED, NULL);
324 acctp = NULLVP;
325 return;
326 }
dadab5e9 327 (void)VFS_STATFS(acctp->v_mount, &sb, NULL);
984263bc
MD
328 if (sb.f_bavail <= acctsuspend * sb.f_blocks / 100) {
329 savacctp = acctp;
330 acctp = NULLVP;
331 log(LOG_NOTICE, "Accounting suspended\n");
332 }
333 }
334 acctwatch_handle = timeout(acctwatch, NULL, acctchkfreq * hz);
335}