subr_prof.c revision 1.22.2.4 1 /* $NetBSD: subr_prof.c,v 1.22.2.4 2002/05/29 21:33:14 nathanw Exp $ */
2
3 /*-
4 * Copyright (c) 1982, 1986, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by the University of
18 * California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * @(#)subr_prof.c 8.4 (Berkeley) 2/14/95
36 */
37
38 #include <sys/cdefs.h>
39 __KERNEL_RCSID(0, "$NetBSD: subr_prof.c,v 1.22.2.4 2002/05/29 21:33:14 nathanw Exp $");
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/lwp.h>
45 #include <sys/proc.h>
46 #include <sys/user.h>
47 #include <sys/mount.h>
48 #include <sys/sa.h>
49 #include <sys/syscallargs.h>
50 #include <sys/sysctl.h>
51
52 #include <machine/cpu.h>
53
54 #ifdef GPROF
55 #include <sys/malloc.h>
56 #include <sys/gmon.h>
57
58 /*
59 * Froms is actually a bunch of unsigned shorts indexing tos
60 */
61 struct gmonparam _gmonparam = { GMON_PROF_OFF };
62
63 /* Actual start of the kernel text segment. */
64 extern char kernel_text[];
65
66 extern char etext[];
67
68
69 void
70 kmstartup()
71 {
72 char *cp;
73 struct gmonparam *p = &_gmonparam;
74 /*
75 * Round lowpc and highpc to multiples of the density we're using
76 * so the rest of the scaling (here and in gprof) stays in ints.
77 */
78 p->lowpc = ROUNDDOWN(((u_long)kernel_text),
79 HISTFRACTION * sizeof(HISTCOUNTER));
80 p->highpc = ROUNDUP((u_long)etext,
81 HISTFRACTION * sizeof(HISTCOUNTER));
82 p->textsize = p->highpc - p->lowpc;
83 printf("Profiling kernel, textsize=%ld [%lx..%lx]\n",
84 p->textsize, p->lowpc, p->highpc);
85 p->kcountsize = p->textsize / HISTFRACTION;
86 p->hashfraction = HASHFRACTION;
87 p->fromssize = p->textsize / HASHFRACTION;
88 p->tolimit = p->textsize * ARCDENSITY / 100;
89 if (p->tolimit < MINARCS)
90 p->tolimit = MINARCS;
91 else if (p->tolimit > MAXARCS)
92 p->tolimit = MAXARCS;
93 p->tossize = p->tolimit * sizeof(struct tostruct);
94 cp = (char *)malloc(p->kcountsize + p->fromssize + p->tossize,
95 M_GPROF, M_NOWAIT);
96 if (cp == 0) {
97 printf("No memory for profiling.\n");
98 return;
99 }
100 memset(cp, 0, p->kcountsize + p->tossize + p->fromssize);
101 p->tos = (struct tostruct *)cp;
102 cp += p->tossize;
103 p->kcount = (u_short *)cp;
104 cp += p->kcountsize;
105 p->froms = (u_short *)cp;
106 }
107
108 /*
109 * Return kernel profiling information.
110 */
111 int
112 sysctl_doprof(name, namelen, oldp, oldlenp, newp, newlen)
113 int *name;
114 u_int namelen;
115 void *oldp;
116 size_t *oldlenp;
117 void *newp;
118 size_t newlen;
119 {
120 struct gmonparam *gp = &_gmonparam;
121 int error;
122
123 /* all sysctl names at this level are terminal */
124 if (namelen != 1)
125 return (ENOTDIR); /* overloaded */
126
127 /* Check we got the necessary memory at startup. */
128 if (gp->kcount == NULL)
129 return (EOPNOTSUPP);
130
131 switch (name[0]) {
132 case GPROF_STATE:
133 error = sysctl_int(oldp, oldlenp, newp, newlen, &gp->state);
134 if (error)
135 return (error);
136 if (gp->state == GMON_PROF_OFF)
137 stopprofclock(&proc0);
138 else
139 startprofclock(&proc0);
140 return (0);
141 case GPROF_COUNT:
142 return (sysctl_struct(oldp, oldlenp, newp, newlen,
143 gp->kcount, gp->kcountsize));
144 case GPROF_FROMS:
145 return (sysctl_struct(oldp, oldlenp, newp, newlen,
146 gp->froms, gp->fromssize));
147 case GPROF_TOS:
148 return (sysctl_struct(oldp, oldlenp, newp, newlen,
149 gp->tos, gp->tossize));
150 case GPROF_GMONPARAM:
151 return (sysctl_rdstruct(oldp, oldlenp, newp, gp, sizeof(*gp)));
152 default:
153 return (EOPNOTSUPP);
154 }
155 /* NOTREACHED */
156 }
157 #endif /* GPROF */
158
159 /*
160 * Profiling system call.
161 *
162 * The scale factor is a fixed point number with 16 bits of fraction, so that
163 * 1.0 is represented as 0x10000. A scale factor of 0 turns off profiling.
164 */
165 /* ARGSUSED */
166 int
167 sys_profil(l, v, retval)
168 struct lwp *l;
169 void *v;
170 register_t *retval;
171 {
172 struct sys_profil_args /* {
173 syscallarg(caddr_t) samples;
174 syscallarg(u_int) size;
175 syscallarg(u_int) offset;
176 syscallarg(u_int) scale;
177 } */ *uap = v;
178 struct proc *p = l->l_proc;
179 struct uprof *upp;
180 int s;
181
182 if (SCARG(uap, scale) > (1 << 16))
183 return (EINVAL);
184 if (SCARG(uap, scale) == 0) {
185 stopprofclock(p);
186 return (0);
187 }
188 upp = &p->p_stats->p_prof;
189
190 /* Block profile interrupts while changing state. */
191 s = splstatclock();
192 upp->pr_off = SCARG(uap, offset);
193 upp->pr_scale = SCARG(uap, scale);
194 upp->pr_base = SCARG(uap, samples);
195 upp->pr_size = SCARG(uap, size);
196 startprofclock(p);
197 splx(s);
198
199 return (0);
200 }
201
202 /*
203 * Scale is a fixed-point number with the binary point 16 bits
204 * into the value, and is <= 1.0. pc is at most 32 bits, so the
205 * intermediate result is at most 48 bits.
206 */
207 #define PC_TO_INDEX(pc, prof) \
208 ((int)(((u_quad_t)((pc) - (prof)->pr_off) * \
209 (u_quad_t)((prof)->pr_scale)) >> 16) & ~1)
210
211 /*
212 * Collect user-level profiling statistics; called on a profiling tick,
213 * when a process is running in user-mode. This routine may be called
214 * from an interrupt context. We try to update the user profiling buffers
215 * cheaply with fuswintr() and suswintr(). If that fails, we revert to
216 * an AST that will vector us to trap() with a context in which copyin
217 * and copyout will work. Trap will then call addupc_task().
218 *
219 * Note that we may (rarely) not get around to the AST soon enough, and
220 * lose profile ticks when the next tick overwrites this one, but in this
221 * case the system is overloaded and the profile is probably already
222 * inaccurate.
223 */
224 void
225 addupc_intr(p, pc)
226 struct proc *p;
227 u_long pc;
228 {
229 struct uprof *prof;
230 caddr_t addr;
231 u_int i;
232 int v;
233
234 prof = &p->p_stats->p_prof;
235 if (pc < prof->pr_off ||
236 (i = PC_TO_INDEX(pc, prof)) >= prof->pr_size)
237 return; /* out of range; ignore */
238
239 addr = prof->pr_base + i;
240 if ((v = fuswintr(addr)) == -1 || suswintr(addr, v + 1) == -1) {
241 prof->pr_addr = pc;
242 prof->pr_ticks++;
243 need_proftick(p);
244 }
245 }
246
247 /*
248 * Much like before, but we can afford to take faults here. If the
249 * update fails, we simply turn off profiling.
250 */
251 void
252 addupc_task(p, pc, ticks)
253 struct proc *p;
254 u_long pc;
255 u_int ticks;
256 {
257 struct uprof *prof;
258 caddr_t addr;
259 u_int i;
260 u_short v;
261
262 /* Testing P_PROFIL may be unnecessary, but is certainly safe. */
263 if ((p->p_flag & P_PROFIL) == 0 || ticks == 0)
264 return;
265
266 prof = &p->p_stats->p_prof;
267 if (pc < prof->pr_off ||
268 (i = PC_TO_INDEX(pc, prof)) >= prof->pr_size)
269 return;
270
271 addr = prof->pr_base + i;
272 if (copyin(addr, (caddr_t)&v, sizeof(v)) == 0) {
273 v += ticks;
274 if (copyout((caddr_t)&v, addr, sizeof(v)) == 0)
275 return;
276 }
277 stopprofclock(p);
278 }
279