xref: /netbsd-src/sys/compat/linux/common/linux_exec.c (revision 7fa608457b817eca6e0977b37f758ae064f3c99c)
1 /*	$NetBSD: linux_exec.c,v 1.97 2007/10/19 18:52:10 njoly Exp $	*/
2 
3 /*-
4  * Copyright (c) 1994, 1995, 1998, 2000, 2007 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Christos Zoulas, Frank van der Linden, Eric Haszlakiewicz and
9  * Thor Lancelot Simon.
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 NetBSD
22  *        Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 #include <sys/cdefs.h>
41 __KERNEL_RCSID(0, "$NetBSD: linux_exec.c,v 1.97 2007/10/19 18:52:10 njoly Exp $");
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/proc.h>
47 #include <sys/malloc.h>
48 #include <sys/namei.h>
49 #include <sys/vnode.h>
50 #include <sys/mount.h>
51 #include <sys/exec.h>
52 #include <sys/exec_elf.h>
53 
54 #include <sys/mman.h>
55 #include <sys/syscallargs.h>
56 
57 #include <sys/ptrace.h>	/* For proc_reparent() */
58 
59 #include <uvm/uvm_extern.h>
60 
61 #include <sys/cpu.h>
62 #include <machine/reg.h>
63 
64 #include <compat/linux/common/linux_types.h>
65 #include <compat/linux/common/linux_signal.h>
66 #include <compat/linux/common/linux_util.h>
67 #include <compat/linux/common/linux_sched.h>
68 #include <compat/linux/common/linux_machdep.h>
69 #include <compat/linux/common/linux_exec.h>
70 #include <compat/linux/common/linux_futex.h>
71 #include <compat/linux/common/linux_ipc.h>
72 #include <compat/linux/common/linux_sem.h>
73 
74 #include <compat/linux/linux_syscallargs.h>
75 #include <compat/linux/linux_syscall.h>
76 #include <compat/linux/common/linux_misc.h>
77 #include <compat/linux/common/linux_errno.h>
78 #include <compat/linux/common/linux_emuldata.h>
79 
80 extern struct sysent linux_sysent[];
81 extern const char * const linux_syscallnames[];
82 extern char linux_sigcode[], linux_esigcode[];
83 
84 static void linux_e_proc_exec __P((struct proc *, struct exec_package *));
85 static void linux_e_proc_fork __P((struct proc *, struct proc *, int));
86 static void linux_e_proc_exit __P((struct proc *));
87 static void linux_e_proc_init __P((struct proc *, struct proc *, int));
88 
89 #ifdef LINUX_NPTL
90 void linux_userret(void);
91 #endif
92 
93 /*
94  * Execve(2). Just check the alternate emulation path, and pass it on
95  * to the NetBSD execve().
96  */
97 int
98 linux_sys_execve(l, v, retval)
99 	struct lwp *l;
100 	void *v;
101 	register_t *retval;
102 {
103 	struct linux_sys_execve_args /* {
104 		syscallarg(const char *) path;
105 		syscallarg(char **) argv;
106 		syscallarg(char **) envp;
107 	} */ *uap = v;
108 	struct sys_execve_args ap;
109 
110 	SCARG(&ap, path) = SCARG(uap, path);
111 	SCARG(&ap, argp) = SCARG(uap, argp);
112 	SCARG(&ap, envp) = SCARG(uap, envp);
113 
114 	return sys_execve(l, &ap, retval);
115 }
116 
117 /*
118  * Emulation switch.
119  */
120 
121 struct uvm_object *emul_linux_object;
122 
123 const struct emul emul_linux = {
124 	"linux",
125 	"/emul/linux",
126 #ifndef __HAVE_MINIMAL_EMUL
127 	0,
128 	(const int *)native_to_linux_errno,
129 	LINUX_SYS_syscall,
130 	LINUX_SYS_NSYSENT,
131 #endif
132 	linux_sysent,
133 	linux_syscallnames,
134 	linux_sendsig,
135 	linux_trapsignal,
136 	NULL,
137 	linux_sigcode,
138 	linux_esigcode,
139 	&emul_linux_object,
140 	linux_setregs,
141 	linux_e_proc_exec,
142 	linux_e_proc_fork,
143 	linux_e_proc_exit,
144 	NULL,
145 	NULL,
146 #ifdef __HAVE_SYSCALL_INTERN
147 	linux_syscall_intern,
148 #else
149 #error Implement __HAVE_SYSCALL_INTERN for this platform
150 #endif
151 	NULL,
152 	NULL,
153 
154 	uvm_default_mapaddr,
155 
156 	linux_usertrap,
157 	0,
158 	NULL,		/* e_startlwp */
159 };
160 
161 static void
162 linux_e_proc_init(p, parent, forkflags)
163 	struct proc *p, *parent;
164 	int forkflags;
165 {
166 	struct linux_emuldata *e = p->p_emuldata;
167 	struct linux_emuldata_shared *s;
168 	struct linux_emuldata *ep = NULL;
169 
170 	if (!e) {
171 		/* allocate new Linux emuldata */
172 		MALLOC(e, void *, sizeof(struct linux_emuldata),
173 			M_EMULDATA, M_WAITOK);
174 	} else  {
175 		e->s->refs--;
176 		if (e->s->refs == 0)
177 			FREE(e->s, M_EMULDATA);
178 	}
179 
180 	memset(e, '\0', sizeof(struct linux_emuldata));
181 
182 	e->proc = p;
183 
184 	if (parent)
185 		ep = parent->p_emuldata;
186 
187 	if (forkflags & FORK_SHAREVM) {
188 #ifdef DIAGNOSTIC
189 		if (ep == NULL) {
190 			killproc(p, "FORK_SHAREVM while emuldata is NULL\n");
191 			FREE(e, M_EMULDATA);
192 			return;
193 		}
194 #endif
195 		s = ep->s;
196 		s->refs++;
197 	} else {
198 		struct vmspace *vm;
199 
200 		MALLOC(s, void *, sizeof(struct linux_emuldata_shared),
201 			M_EMULDATA, M_WAITOK);
202 		s->refs = 1;
203 
204 		/*
205 		 * Set the process idea of the break to the real value.
206 		 * For fork, we use parent's vmspace since our's
207 		 * is not setup at the time of this call and is going
208 		 * to be copy of parent's anyway. For exec, just
209 		 * use our own vmspace.
210 		 */
211 		vm = (parent) ? parent->p_vmspace : p->p_vmspace;
212 		s->p_break = (char *)vm->vm_daddr + ctob(vm->vm_dsize);
213 
214 		/*
215 		 * Linux threads are emulated as NetBSD processes (not lwp)
216 		 * We use native PID for Linux TID. The Linux TID is the
217 		 * PID of the first process in the group. It is stored
218 		 * here
219 		 */
220 		s->group_pid = p->p_pid;
221 
222 		/*
223 		 * Initialize the list of threads in the group
224 		 */
225 		LIST_INIT(&s->threads);
226 
227 		s->xstat = 0;
228 		s->flags = 0;
229 	}
230 
231 	e->s = s;
232 
233 	/*
234 	 * Add this thread in the group thread list
235 	 */
236 	LIST_INSERT_HEAD(&s->threads, e, threads);
237 
238 #ifdef LINUX_NPTL
239 	if (ep != NULL) {
240 		e->parent_tidptr = ep->parent_tidptr;
241 		e->child_tidptr = ep->child_tidptr;
242 		e->clone_flags = ep->clone_flags;
243 	}
244 #endif /* LINUX_NPTL */
245 
246 	p->p_emuldata = e;
247 }
248 
249 /*
250  * Allocate new per-process structures. Called when executing Linux
251  * process. We can reuse the old emuldata - if it's not null,
252  * the executed process is of same emulation as original forked one.
253  */
254 static void
255 linux_e_proc_exec(struct proc *p, struct exec_package *epp)
256 {
257 	/* exec, use our vmspace */
258 	linux_e_proc_init(p, NULL, 0);
259 }
260 
261 /*
262  * Emulation per-process exit hook.
263  */
264 static void
265 linux_e_proc_exit(p)
266 	struct proc *p;
267 {
268 	struct linux_emuldata *e = p->p_emuldata;
269 
270 #ifdef LINUX_NPTL
271 	linux_nptl_proc_exit(p);
272 #endif
273 	/* Remove the thread for the group thread list */
274 	LIST_REMOVE(e, threads);
275 
276 	/* free Linux emuldata and set the pointer to null */
277 	e->s->refs--;
278 	if (e->s->refs == 0)
279 		FREE(e->s, M_EMULDATA);
280 	FREE(e, M_EMULDATA);
281 	p->p_emuldata = NULL;
282 }
283 
284 /*
285  * Emulation fork hook.
286  */
287 static void
288 linux_e_proc_fork(p, parent, forkflags)
289 	struct proc *p, *parent;
290 	int forkflags;
291 {
292 	/*
293 	 * The new process might share some vmspace-related stuff
294 	 * with parent, depending on fork flags (CLONE_VM et.al).
295 	 * Force allocation of new base emuldata, and share the
296 	 * VM-related parts only if necessary.
297 	 */
298 	p->p_emuldata = NULL;
299 	linux_e_proc_init(p, parent, forkflags);
300 
301 #ifdef LINUX_NPTL
302 	linux_nptl_proc_fork(p, parent, linux_userret);
303 #endif
304 
305 	return;
306 }
307 
308 #ifdef LINUX_NPTL
309 void
310 linux_userret(void)
311 {
312 	struct lwp *l = curlwp;
313 	struct proc *p = l->l_proc;
314 	struct linux_emuldata *led = p->p_emuldata;
315 	int error;
316 
317 	/* LINUX_CLONE_CHILD_SETTID: copy child's TID to child's memory  */
318 	if (led->clone_flags & LINUX_CLONE_CHILD_SETTID) {
319 		if ((error = copyout(&l->l_proc->p_pid,
320 		    led->child_tidptr,  sizeof(l->l_proc->p_pid))) != 0)
321 			printf("linux_userret: LINUX_CLONE_CHILD_SETTID "
322 			    "failed (led->child_tidptr = %p, p->p_pid = %d)\n",
323 			    led->child_tidptr, p->p_pid);
324 	}
325 
326 	/* LINUX_CLONE_SETTLS: allocate a new TLS */
327 	if (led->clone_flags & LINUX_CLONE_SETTLS) {
328 		if (linux_set_newtls(l, linux_get_newtls(l)) != 0)
329 			printf("linux_userret: linux_set_tls failed");
330 	}
331 
332 	return;
333 }
334 
335 void
336 linux_nptl_proc_exit(p)
337 	struct proc *p;
338 {
339 	struct linux_emuldata *e = p->p_emuldata;
340 
341 	mutex_enter(&proclist_lock);
342 
343 	/*
344 	 * Check if we are a thread group leader victim of another
345 	 * thread doing exit_group(). If we are, change the exit code.
346 	 */
347 	if ((e->s->group_pid == p->p_pid) &&
348 	    (e->s->flags & LINUX_LES_INEXITGROUP)) {
349 		p->p_xstat = e->s->xstat;
350 	}
351 
352 	/*
353 	 * Members of the thread groups others than the leader should
354 	 * exit quietely: no zombie stage, no signal. We do that by
355 	 * reparenting to init. init will collect us and nobody will
356 	 * notice what happened.
357 	 */
358 #ifdef DEBUG_LINUX
359 	printf("%s:%d e->s->group_pid = %d, p->p_pid = %d, flags = 0x%x\n",
360 	    __func__, __LINE__, e->s->group_pid, p->p_pid, e->s->flags);
361 #endif
362 	if (e->s->group_pid != p->p_pid) {
363 		proc_reparent(p, initproc);
364 		cv_broadcast(&initproc->p_waitcv);
365 	}
366 
367 	mutex_exit(&proclist_lock);
368 
369 	/* Emulate LINUX_CLONE_CHILD_CLEARTID */
370 	if (e->clear_tid != NULL) {
371 		int error;
372 		int null = 0;
373 		struct linux_sys_futex_args cup;
374 		register_t retval;
375 
376 		error = copyout(&null, e->clear_tid, sizeof(null));
377 #ifdef DEBUG_LINUX
378 		if (error != 0)
379 			printf("%s: cannot clear TID\n", __func__);
380 #endif
381 
382 		SCARG(&cup, uaddr) = e->clear_tid;
383 		SCARG(&cup, op) = LINUX_FUTEX_WAKE;
384 		SCARG(&cup, val) = 0x7fffffff; /* Awake everyone */
385 		SCARG(&cup, timeout) = NULL;
386 		SCARG(&cup, uaddr2) = NULL;
387 		SCARG(&cup, val3) = 0;
388 		if ((error = linux_sys_futex(curlwp, &cup, &retval)) != 0)
389 			printf("%s: linux_sys_futex failed\n", __func__);
390 	}
391 
392 	return;
393 }
394 
395 void
396 linux_nptl_proc_fork(p, parent, luserret)
397 	struct proc *p;
398 	struct proc *parent;
399 	void (*luserret)(void);
400 {
401 #ifdef LINUX_NPTL
402 	struct linux_emuldata *e;
403 #endif
404 
405 	e = p->p_emuldata;
406 
407 	/* LINUX_CLONE_CHILD_CLEARTID: clear TID in child's memory on exit() */
408 	if (e->clone_flags & LINUX_CLONE_CHILD_CLEARTID)
409 		e->clear_tid = e->child_tidptr;
410 
411 	/* LINUX_CLONE_PARENT_SETTID: set child's TID in parent's memory */
412 	if (e->clone_flags & LINUX_CLONE_PARENT_SETTID) {
413 		if (copyout_proc(parent, &p->p_pid,
414 		    e->parent_tidptr,  sizeof(p->p_pid)) != 0)
415 			printf("%s: LINUX_CLONE_PARENT_SETTID "
416 			    "failed (e->parent_tidptr = %p, "
417 			    "parent->p_pid = %d, p->p_pid = %d)\n",
418 			    __func__, e->parent_tidptr,
419 			    parent->p_pid, p->p_pid);
420 	}
421 
422 	/*
423 	 * CLONE_CHILD_SETTID and LINUX_CLONE_SETTLS require child's VM
424 	 * setup to be completed, we postpone them until userret time.
425 	 */
426 	if (e->clone_flags &
427 	    (LINUX_CLONE_CHILD_CLEARTID | LINUX_CLONE_SETTLS))
428 		p->p_userret = luserret;
429 
430 	return;
431 }
432 
433 void
434 linux_nptl_proc_init(p, parent)
435 	struct proc *p;
436 	struct proc *parent;
437 {
438 	struct linux_emuldata *e = p->p_emuldata;
439 	struct linux_emuldata *ep;
440 
441 	if ((parent != NULL) && (parent->p_emuldata != NULL)) {
442 		ep = parent->p_emuldata;
443 
444 		e->parent_tidptr = ep->parent_tidptr;
445 		e->child_tidptr = ep->child_tidptr;
446 		e->clone_flags = ep->clone_flags;
447 	}
448 
449 	return;
450 }
451 
452 
453 #endif /* LINUX_NPTL */
454