xref: /freebsd/sys/kern/kern_tslog.c (revision fdafd315ad0d0f28a11b9fb4476a9ab059c62b92)
1 /*-
2  * Copyright (c) 2017 Colin Percival
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #include <sys/param.h>
28 #include <sys/kernel.h>
29 #include <sys/linker.h>
30 #include <sys/malloc.h>
31 #include <sys/proc.h>
32 #include <sys/sbuf.h>
33 #include <sys/sysctl.h>
34 #include <sys/systm.h>
35 #include <sys/tslog.h>
36 
37 #include <machine/atomic.h>
38 #include <machine/cpu.h>
39 
40 #ifndef TSLOGSIZE
41 #define TSLOGSIZE 262144
42 #endif
43 
44 static volatile long nrecs = 0;
45 static struct timestamp {
46 	void * td;
47 	int type;
48 	const char * f;
49 	const char * s;
50 	uint64_t tsc;
51 } timestamps[TSLOGSIZE];
52 
53 void
tslog(void * td,int type,const char * f,const char * s)54 tslog(void * td, int type, const char * f, const char * s)
55 {
56 	uint64_t tsc = get_cyclecount();
57 	long pos;
58 
59 	/* A NULL thread is thread0 before curthread is set. */
60 	if (td == NULL)
61 		td = &thread0;
62 
63 	/* Grab a slot. */
64 	pos = atomic_fetchadd_long(&nrecs, 1);
65 
66 	/* Store record. */
67 	if (pos < nitems(timestamps)) {
68 		timestamps[pos].td = td;
69 		timestamps[pos].type = type;
70 		timestamps[pos].f = f;
71 		timestamps[pos].s = s;
72 		timestamps[pos].tsc = tsc;
73 	}
74 }
75 
76 static int
sysctl_debug_tslog(SYSCTL_HANDLER_ARGS)77 sysctl_debug_tslog(SYSCTL_HANDLER_ARGS)
78 {
79 	int error;
80 	struct sbuf *sb;
81 	size_t i, limit;
82 	caddr_t loader_tslog;
83 	void * loader_tslog_buf;
84 	size_t loader_tslog_len;
85 
86 	/*
87 	 * This code can race against the code in tslog() which stores
88 	 * records: Theoretically we could end up reading a record after
89 	 * its slots have been reserved but before it has been written.
90 	 * Since this code takes orders of magnitude longer to run than
91 	 * tslog() takes to write a record, it is highly unlikely that
92 	 * anyone will ever experience this race.
93 	 */
94 	sb = sbuf_new_for_sysctl(NULL, NULL, 1024, req);
95 
96 	/* Get data from the boot loader, if it provided any. */
97 	loader_tslog = preload_search_by_type("TSLOG data");
98 	if (loader_tslog != NULL) {
99 		loader_tslog_buf = preload_fetch_addr(loader_tslog);
100 		loader_tslog_len = preload_fetch_size(loader_tslog);
101 		sbuf_bcat(sb, loader_tslog_buf, loader_tslog_len);
102 	}
103 
104 	/* Add data logged within the kernel. */
105 	limit = MIN(nrecs, nitems(timestamps));
106 	for (i = 0; i < limit; i++) {
107 		sbuf_printf(sb, "%p", timestamps[i].td);
108 		sbuf_printf(sb, " %llu",
109 		    (unsigned long long)timestamps[i].tsc);
110 		switch (timestamps[i].type) {
111 		case TS_ENTER:
112 			sbuf_cat(sb, " ENTER");
113 			break;
114 		case TS_EXIT:
115 			sbuf_cat(sb, " EXIT");
116 			break;
117 		case TS_THREAD:
118 			sbuf_cat(sb, " THREAD");
119 			break;
120 		case TS_EVENT:
121 			sbuf_cat(sb, " EVENT");
122 			break;
123 		}
124 		sbuf_printf(sb, " %s", timestamps[i].f ? timestamps[i].f : "(null)");
125 		if (timestamps[i].s)
126 			sbuf_printf(sb, " %s\n", timestamps[i].s);
127 		else
128 			sbuf_putc(sb, '\n');
129 	}
130 	error = sbuf_finish(sb);
131 	sbuf_delete(sb);
132 	return (error);
133 }
134 
135 SYSCTL_PROC(_debug, OID_AUTO, tslog,
136     CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_SKIP,
137     0, 0, sysctl_debug_tslog, "", "Dump recorded event timestamps");
138 
139 MALLOC_DEFINE(M_TSLOGUSER, "tsloguser", "Strings used by userland tslog");
140 static struct procdata {
141 	pid_t ppid;
142 	uint64_t tsc_forked;
143 	uint64_t tsc_exited;
144 	char * execname;
145 	char * namei;
146 	int reused;
147 } procs[PID_MAX + 1];
148 
149 void
tslog_user(pid_t pid,pid_t ppid,const char * execname,const char * namei)150 tslog_user(pid_t pid, pid_t ppid, const char * execname, const char * namei)
151 {
152 	uint64_t tsc = get_cyclecount();
153 
154 	/* If we wrapped, do nothing. */
155 	if (procs[pid].reused)
156 		return;
157 
158 	/* If we have a ppid, we're recording a fork. */
159 	if (ppid != (pid_t)(-1)) {
160 		/* If we have a ppid already, we wrapped. */
161 		if (procs[pid].ppid) {
162 			procs[pid].reused = 1;
163 			return;
164 		}
165 
166 		/* Fill in some fields. */
167 		procs[pid].ppid = ppid;
168 		procs[pid].tsc_forked = tsc;
169 		return;
170 	}
171 
172 	/* If we have an execname, record it. */
173 	if (execname != NULL) {
174 		if (procs[pid].execname != NULL)
175 			free(procs[pid].execname, M_TSLOGUSER);
176 		procs[pid].execname = strdup(execname, M_TSLOGUSER);
177 		return;
178 	}
179 
180 	/* Record the first namei for the process. */
181 	if (namei != NULL) {
182 		if (procs[pid].namei == NULL)
183 			procs[pid].namei = strdup(namei, M_TSLOGUSER);
184 		return;
185 	}
186 
187 	/* Otherwise we're recording an exit. */
188 	procs[pid].tsc_exited = tsc;
189 }
190 
191 static int
sysctl_debug_tslog_user(SYSCTL_HANDLER_ARGS)192 sysctl_debug_tslog_user(SYSCTL_HANDLER_ARGS)
193 {
194 	int error;
195 	struct sbuf *sb;
196 	pid_t pid;
197 
198 	sb = sbuf_new_for_sysctl(NULL, NULL, 1024, req);
199 
200 	/* Export the data we logged. */
201 	for (pid = 0; pid <= PID_MAX; pid++) {
202 		sbuf_printf(sb, "%zu", (size_t)pid);
203 		sbuf_printf(sb, " %zu", (size_t)procs[pid].ppid);
204 		sbuf_printf(sb, " %llu",
205 		    (unsigned long long)procs[pid].tsc_forked);
206 		sbuf_printf(sb, " %llu",
207 		    (unsigned long long)procs[pid].tsc_exited);
208 		sbuf_printf(sb, " \"%s\"", procs[pid].execname ?
209 		    procs[pid].execname : "");
210 		sbuf_printf(sb, " \"%s\"", procs[pid].namei ?
211 		    procs[pid].namei : "");
212 		sbuf_putc(sb, '\n');
213 	}
214 	error = sbuf_finish(sb);
215 	sbuf_delete(sb);
216 	return (error);
217 }
218 
219 SYSCTL_PROC(_debug, OID_AUTO, tslog_user,
220     CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_SKIP,
221     0, 0, sysctl_debug_tslog_user,
222     "", "Dump recorded userland event timestamps");
223