i386-linux.c revision 286963
1/*
2 * Copyright 1997 Sean Eric Fagan
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 * 1. Redistributions of source code must retain the above copyright
8 *    notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 *    notice, this list of conditions and the following disclaimer in the
11 *    documentation and/or other materials provided with the distribution.
12 * 3. All advertising materials mentioning features or use of this software
13 *    must display the following acknowledgement:
14 *	This product includes software developed by Sean Eric Fagan
15 * 4. Neither the name of the author may be used to endorse or promote
16 *    products derived from this software without specific prior written
17 *    permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32#ifndef lint
33static const char rcsid[] =
34  "$FreeBSD: head/usr.bin/truss/i386-linux.c 286963 2015-08-20 14:51:11Z jhb $";
35#endif /* not lint */
36
37/*
38 * Linux/i386-specific system call handling.  Given how much of this code
39 * is taken from the freebsd equivalent, I can probably put even more of
40 * it in support routines that can be used by any personality support.
41 */
42
43#include <sys/types.h>
44#include <sys/ptrace.h>
45
46#include <machine/reg.h>
47#include <machine/psl.h>
48
49#include <errno.h>
50#include <fcntl.h>
51#include <signal.h>
52#include <stdio.h>
53#include <stdlib.h>
54#include <string.h>
55#include <time.h>
56#include <unistd.h>
57
58#include "truss.h"
59#include "syscall.h"
60#include "extern.h"
61
62#include "linux_syscalls.h"
63
64static int nsyscalls = nitems(linux_syscallnames);
65
66/*
67 * This is what this particular file uses to keep track of a system call.
68 * It is probably not quite sufficient -- I can probably use the same
69 * structure for the various syscall personalities, and I also probably
70 * need to nest system calls (for signal handlers).
71 *
72 * 'struct syscall' describes the system call; it may be NULL, however,
73 * if we don't know about this particular system call yet.
74 */
75struct linux_syscall {
76	struct syscall *sc;
77	const char *name;
78	int number;
79	unsigned long args[5];
80	int nargs;	/* number of arguments -- *not* number of words! */
81	char **s_args;	/* the printable arguments */
82};
83
84static struct linux_syscall *
85alloc_fsc(void)
86{
87
88	return (malloc(sizeof(struct linux_syscall)));
89}
90
91/* Clear up and free parts of the fsc structure. */
92static void
93free_fsc(struct linux_syscall *fsc)
94{
95	int i;
96
97	if (fsc->s_args) {
98		for (i = 0; i < fsc->nargs; i++)
99			free(fsc->s_args[i]);
100		free(fsc->s_args);
101	}
102	free(fsc);
103}
104
105/*
106 * Called when a process has entered a system call.  nargs is the
107 * number of words, not number of arguments (a necessary distinction
108 * in some cases).  Note that if the STOPEVENT() code in i386/i386/trap.c
109 * is ever changed these functions need to keep up.
110 */
111
112void
113i386_linux_syscall_entry(struct trussinfo *trussinfo, int nargs)
114{
115	struct reg regs;
116	struct linux_syscall *fsc;
117	struct syscall *sc;
118	lwpid_t tid;
119	int i, syscall_num;
120
121	tid = trussinfo->curthread->tid;
122
123	if (ptrace(PT_GETREGS, tid, (caddr_t)&regs, 0) < 0) {
124		fprintf(trussinfo->outfile, "-- CANNOT READ REGISTERS --\n");
125		return;
126	}
127
128	syscall_num = regs.r_eax;
129
130	fsc = alloc_fsc();
131	if (fsc == NULL)
132		return;
133	fsc->number = syscall_num;
134	fsc->name = (syscall_num < 0 || syscall_num >= nsyscalls) ?
135	    NULL : linux_syscallnames[syscall_num];
136	if (!fsc->name) {
137		fprintf(trussinfo->outfile, "-- UNKNOWN SYSCALL %d --\n",
138		    syscall_num);
139	}
140
141	if (fsc->name && (trussinfo->flags & FOLLOWFORKS) &&
142	    (strcmp(fsc->name, "linux_fork") == 0 ||
143	    strcmp(fsc->name, "linux_vfork") == 0))
144		trussinfo->curthread->in_fork = 1;
145
146	if (nargs == 0)
147		return;
148
149	/*
150	 * Linux passes syscall arguments in registers, not
151	 * on the stack.  Fortunately, we've got access to the
152	 * register set.  Note that we don't bother checking the
153	 * number of arguments.	And what does linux do for syscalls
154	 * that have more than five arguments?
155	 */
156
157	fsc->args[0] = regs.r_ebx;
158	fsc->args[1] = regs.r_ecx;
159	fsc->args[2] = regs.r_edx;
160	fsc->args[3] = regs.r_esi;
161	fsc->args[4] = regs.r_edi;
162
163	sc = get_syscall(fsc->name);
164	if (sc)
165		fsc->nargs = sc->nargs;
166	else {
167#if DEBUG
168		fprintf(trussinfo->outfile, "unknown syscall %s -- setting "
169		    "args to %d\n", fsc->name, nargs);
170#endif
171		fsc->nargs = nargs;
172	}
173
174	fsc->s_args = calloc(1, (1 + fsc->nargs) * sizeof(char *));
175	fsc->sc = sc;
176
177	/*
178	 * At this point, we set up the system call arguments.
179	 * We ignore any OUT ones, however -- those are arguments that
180	 * are set by the system call, and so are probably meaningless
181	 * now.	This doesn't currently support arguments that are
182	 * passed in *and* out, however.
183	 */
184
185	if (fsc->name) {
186#if DEBUG
187		fprintf(stderr, "syscall %s(", fsc->name);
188#endif
189		for (i = 0; i < fsc->nargs; i++) {
190#if DEBUG
191			fprintf(stderr, "0x%x%s", sc ?
192			    fsc->args[sc->args[i].offset] : fsc->args[i],
193			    i < (fsc->nargs - 1) ? "," : "");
194#endif
195			if (sc && !(sc->args[i].type & OUT)) {
196				fsc->s_args[i] = print_arg(&sc->args[i],
197				    fsc->args, 0, trussinfo);
198			}
199		}
200#if DEBUG
201		fprintf(stderr, ")\n");
202#endif
203	}
204
205#if DEBUG
206	fprintf(trussinfo->outfile, "\n");
207#endif
208
209	trussinfo->curthread->fsc = fsc;
210}
211
212/*
213 * Linux syscalls return negative errno's, we do positive and map them
214 */
215static const int bsd_to_linux_errno[] = {
216	-0,  -1,  -2,  -3,  -4,  -5,  -6,  -7,  -8,  -9,
217	-10, -35, -12, -13, -14, -15, -16, -17, -18, -19,
218	-20, -21, -22, -23, -24, -25, -26, -27, -28, -29,
219	-30, -31, -32, -33, -34, -11,-115,-114, -88, -89,
220	-90, -91, -92, -93, -94, -95, -96, -97, -98, -99,
221	-100,-101,-102,-103,-104,-105,-106,-107,-108,-109,
222	-110,-111, -40, -36,-112,-113, -39, -11, -87,-122,
223	-116, -66,  -6,  -6,  -6,  -6,  -6, -37, -38,  -9,
224	-6,
225};
226
227long
228i386_linux_syscall_exit(struct trussinfo *trussinfo, int syscall_num __unused)
229{
230	struct reg regs;
231	struct linux_syscall *fsc;
232	struct syscall *sc;
233	lwpid_t tid;
234	long retval;
235	int errorp, i;
236
237	if (trussinfo->curthread->fsc == NULL)
238		return (-1);
239
240	tid = trussinfo->curthread->tid;
241
242	if (ptrace(PT_GETREGS, tid, (caddr_t)&regs, 0) < 0) {
243		fprintf(trussinfo->outfile, "-- CANNOT READ REGISTERS --\n");
244		return (-1);
245	}
246
247	retval = regs.r_eax;
248	errorp = !!(regs.r_eflags & PSL_C);
249
250	/*
251	 * This code, while simpler than the initial versions I used, could
252	 * stand some significant cleaning.
253	 */
254
255	fsc = trussinfo->curthread->fsc;
256	sc = fsc->sc;
257	if (!sc) {
258		for (i = 0; i < fsc->nargs; i++)
259			asprintf(&fsc->s_args[i], "0x%lx", fsc->args[i]);
260	} else {
261		/*
262		 * Here, we only look for arguments that have OUT masked in --
263		 * otherwise, they were handled in the syscall_entry function.
264		 */
265		for (i = 0; i < sc->nargs; i++) {
266			char *temp;
267
268			if (sc->args[i].type & OUT) {
269				/*
270				 * If an error occurred, then don't bother
271				 * getting the data; it may not be valid.
272				 */
273				if (errorp) {
274					asprintf(&temp, "0x%lx",
275					    fsc->args[sc->args[i].offset]);
276				} else {
277					temp = print_arg(&sc->args[i],
278					    fsc->args, retval, trussinfo);
279				}
280				fsc->s_args[i] = temp;
281			}
282		}
283	}
284
285	/*
286	 * It would probably be a good idea to merge the error handling,
287	 * but that complicates things considerably.
288	 */
289	if (errorp) {
290		for (i = 0; (size_t)i < nitems(bsd_to_linux_errno); i++) {
291			if (retval == bsd_to_linux_errno[i])
292				break;
293		}
294	}
295
296	if (fsc->name != NULL && (strcmp(fsc->name, "linux_execve") == 0 ||
297	    strcmp(fsc->name, "exit") == 0))
298		trussinfo->curthread->in_syscall = 1;
299
300	print_syscall_ret(trussinfo, fsc->name, fsc->nargs, fsc->s_args, errorp,
301	    errorp ? i : retval, fsc->sc);
302	free_fsc(fsc);
303
304	return (retval);
305}
306