xref: /netbsd-src/sys/compat/linux/common/linux_exec.c (revision 8ac07aec990b9d2e483062509d0a9fa5b4f57cf2)
1 /*	$NetBSD: linux_exec.c,v 1.104 2008/04/24 15:35:27 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 1994, 1995, 1998, 2000, 2007, 2008 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 1.104 2008/04/24 15:35:27 ad 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(struct proc *, struct exec_package *);
85 static void linux_e_proc_fork(struct proc *, struct proc *, int);
86 static void linux_e_proc_exit(struct proc *);
87 static void linux_e_proc_init(struct proc *, struct proc *, int);
88 
89 #ifdef LINUX_NPTL
90 void linux_userret(void);
91 #endif
92 
93 /*
94  * Emulation switch.
95  */
96 
97 struct uvm_object *emul_linux_object;
98 
99 const struct emul emul_linux = {
100 	"linux",
101 	"/emul/linux",
102 #ifndef __HAVE_MINIMAL_EMUL
103 	0,
104 	(const int *)native_to_linux_errno,
105 	LINUX_SYS_syscall,
106 	LINUX_SYS_NSYSENT,
107 #endif
108 	linux_sysent,
109 	linux_syscallnames,
110 	linux_sendsig,
111 	linux_trapsignal,
112 	NULL,
113 	linux_sigcode,
114 	linux_esigcode,
115 	&emul_linux_object,
116 	linux_setregs,
117 	linux_e_proc_exec,
118 	linux_e_proc_fork,
119 	linux_e_proc_exit,
120 	NULL,
121 	NULL,
122 #ifdef __HAVE_SYSCALL_INTERN
123 	linux_syscall_intern,
124 #else
125 #error Implement __HAVE_SYSCALL_INTERN for this platform
126 #endif
127 	NULL,
128 	NULL,
129 
130 	uvm_default_mapaddr,
131 
132 	linux_usertrap,
133 	0,
134 	NULL,		/* e_startlwp */
135 };
136 
137 static void
138 linux_e_proc_init(p, parent, forkflags)
139 	struct proc *p, *parent;
140 	int forkflags;
141 {
142 	struct linux_emuldata *e = p->p_emuldata;
143 	struct linux_emuldata_shared *s;
144 	struct linux_emuldata *ep = NULL;
145 
146 	if (!e) {
147 		/* allocate new Linux emuldata */
148 		MALLOC(e, void *, sizeof(struct linux_emuldata),
149 			M_EMULDATA, M_WAITOK);
150 	} else  {
151 		mutex_enter(proc_lock);
152 		e->s->refs--;
153 		if (e->s->refs == 0)
154 			FREE(e->s, M_EMULDATA);
155 		mutex_exit(proc_lock);
156 	}
157 
158 	memset(e, '\0', sizeof(struct linux_emuldata));
159 
160 	e->proc = p;
161 
162 	if (parent)
163 		ep = parent->p_emuldata;
164 
165 	if (forkflags & FORK_SHAREVM) {
166 		mutex_enter(proc_lock);
167 #ifdef DIAGNOSTIC
168 		if (ep == NULL) {
169 			killproc(p, "FORK_SHAREVM while emuldata is NULL\n");
170 			mutex_exit(proc_lock);
171 			FREE(e, M_EMULDATA);
172 			return;
173 		}
174 #endif
175 		s = ep->s;
176 		s->refs++;
177 	} else {
178 		struct vmspace *vm;
179 
180 		MALLOC(s, void *, sizeof(struct linux_emuldata_shared),
181 			M_EMULDATA, M_WAITOK);
182 		s->refs = 1;
183 
184 		/*
185 		 * Set the process idea of the break to the real value.
186 		 * For fork, we use parent's vmspace since our's
187 		 * is not setup at the time of this call and is going
188 		 * to be copy of parent's anyway. For exec, just
189 		 * use our own vmspace.
190 		 */
191 		vm = (parent) ? parent->p_vmspace : p->p_vmspace;
192 		s->p_break = (char *)vm->vm_daddr + ctob(vm->vm_dsize);
193 
194 		/*
195 		 * Linux threads are emulated as NetBSD processes (not lwp)
196 		 * We use native PID for Linux TID. The Linux TID is the
197 		 * PID of the first process in the group. It is stored
198 		 * here
199 		 */
200 		s->group_pid = p->p_pid;
201 
202 		/*
203 		 * Initialize the list of threads in the group
204 		 */
205 		LIST_INIT(&s->threads);
206 
207 		s->xstat = 0;
208 		s->flags = 0;
209 		mutex_enter(proc_lock);
210 	}
211 
212 	e->s = s;
213 
214 	/*
215 	 * Add this thread in the group thread list
216 	 */
217 	LIST_INSERT_HEAD(&s->threads, e, threads);
218 
219 #ifdef LINUX_NPTL
220 	if (ep != NULL) {
221 		e->parent_tidptr = ep->parent_tidptr;
222 		e->child_tidptr = ep->child_tidptr;
223 		e->clone_flags = ep->clone_flags;
224 	}
225 #endif /* LINUX_NPTL */
226 
227 	mutex_exit(proc_lock);
228 	p->p_emuldata = e;
229 }
230 
231 /*
232  * Allocate new per-process structures. Called when executing Linux
233  * process. We can reuse the old emuldata - if it's not null,
234  * the executed process is of same emulation as original forked one.
235  */
236 static void
237 linux_e_proc_exec(struct proc *p, struct exec_package *epp)
238 {
239 
240 	/* exec, use our vmspace */
241 	linux_e_proc_init(p, NULL, 0);
242 }
243 
244 /*
245  * Emulation per-process exit hook.
246  */
247 static void
248 linux_e_proc_exit(struct proc *p)
249 {
250 	struct linux_emuldata *e = p->p_emuldata;
251 
252 #ifdef LINUX_NPTL
253 	linux_nptl_proc_exit(p);
254 #endif
255 
256 	/* Remove the thread for the group thread list */
257 	mutex_enter(proc_lock);
258 	LIST_REMOVE(e, threads);
259 
260 	/* free Linux emuldata and set the pointer to null */
261 	e->s->refs--;
262 	if (e->s->refs == 0)
263 		FREE(e->s, M_EMULDATA);
264 	p->p_emuldata = NULL;
265 	mutex_exit(proc_lock);
266 	FREE(e, M_EMULDATA);
267 }
268 
269 /*
270  * Emulation fork hook.
271  */
272 static void
273 linux_e_proc_fork(p, parent, forkflags)
274 	struct proc *p, *parent;
275 	int forkflags;
276 {
277 
278 	/*
279 	 * The new process might share some vmspace-related stuff
280 	 * with parent, depending on fork flags (CLONE_VM et.al).
281 	 * Force allocation of new base emuldata, and share the
282 	 * VM-related parts only if necessary.
283 	 */
284 	p->p_emuldata = NULL;
285 	linux_e_proc_init(p, parent, forkflags);
286 
287 #ifdef LINUX_NPTL
288 	linux_nptl_proc_fork(p, parent, linux_userret);
289 #endif
290 
291 	return;
292 }
293 
294 #ifdef LINUX_NPTL
295 void
296 linux_userret(void)
297 {
298 	struct lwp *l = curlwp;
299 	struct proc *p = l->l_proc;
300 	struct linux_emuldata *led = p->p_emuldata;
301 	int error;
302 
303 	/* LINUX_CLONE_CHILD_SETTID: copy child's TID to child's memory  */
304 	if (led->clone_flags & LINUX_CLONE_CHILD_SETTID) {
305 		if ((error = copyout(&l->l_proc->p_pid,
306 		    led->child_tidptr,  sizeof(l->l_proc->p_pid))) != 0)
307 			printf("linux_userret: LINUX_CLONE_CHILD_SETTID "
308 			    "failed (led->child_tidptr = %p, p->p_pid = %d)\n",
309 			    led->child_tidptr, p->p_pid);
310 	}
311 
312 	/* LINUX_CLONE_SETTLS: allocate a new TLS */
313 	if (led->clone_flags & LINUX_CLONE_SETTLS) {
314 		if (linux_set_newtls(l, linux_get_newtls(l)) != 0)
315 			printf("linux_userret: linux_set_tls failed");
316 	}
317 
318 	return;
319 }
320 
321 void
322 linux_nptl_proc_exit(struct proc *p)
323 {
324 	struct linux_emuldata *e = p->p_emuldata;
325 
326 	mutex_enter(proc_lock);
327 
328 	/*
329 	 * Check if we are a thread group leader victim of another
330 	 * thread doing exit_group(). If we are, change the exit code.
331 	 */
332 	if ((e->s->group_pid == p->p_pid) &&
333 	    (e->s->flags & LINUX_LES_INEXITGROUP)) {
334 		p->p_xstat = e->s->xstat;
335 	}
336 
337 	/*
338 	 * Members of the thread groups others than the leader should
339 	 * exit quietely: no zombie stage, no signal. We do that by
340 	 * reparenting to init. init will collect us and nobody will
341 	 * notice what happened.
342 	 */
343 #ifdef DEBUG_LINUX
344 	printf("%s:%d e->s->group_pid = %d, p->p_pid = %d, flags = 0x%x\n",
345 	    __func__, __LINE__, e->s->group_pid, p->p_pid, e->s->flags);
346 #endif
347 	if ((e->s->group_pid != p->p_pid) &&
348 	    (e->clone_flags & LINUX_CLONE_THREAD)) {
349 		proc_reparent(p, initproc);
350 		cv_broadcast(&initproc->p_waitcv);
351 	}
352 
353 	mutex_exit(proc_lock);
354 
355 	/* Emulate LINUX_CLONE_CHILD_CLEARTID */
356 	if (e->clear_tid != NULL) {
357 		int error;
358 		int null = 0;
359 		struct linux_sys_futex_args cup;
360 		register_t retval;
361 
362 		error = copyout(&null, e->clear_tid, sizeof(null));
363 #ifdef DEBUG_LINUX
364 		if (error != 0)
365 			printf("%s: cannot clear TID\n", __func__);
366 #endif
367 
368 		SCARG(&cup, uaddr) = e->clear_tid;
369 		SCARG(&cup, op) = LINUX_FUTEX_WAKE;
370 		SCARG(&cup, val) = 0x7fffffff; /* Awake everyone */
371 		SCARG(&cup, timeout) = NULL;
372 		SCARG(&cup, uaddr2) = NULL;
373 		SCARG(&cup, val3) = 0;
374 		if ((error = linux_sys_futex(curlwp, &cup, &retval)) != 0)
375 			printf("%s: linux_sys_futex failed\n", __func__);
376 	}
377 
378 	return;
379 }
380 
381 void
382 linux_nptl_proc_fork(p, parent, luserret)
383 	struct proc *p;
384 	struct proc *parent;
385 	void (*luserret)(void);
386 {
387 	struct linux_emuldata *e = p->p_emuldata;
388 
389 	/* LINUX_CLONE_CHILD_CLEARTID: clear TID in child's memory on exit() */
390 	if (e->clone_flags & LINUX_CLONE_CHILD_CLEARTID)
391 		e->clear_tid = e->child_tidptr;
392 
393 	/* LINUX_CLONE_PARENT_SETTID: set child's TID in parent's memory */
394 	if (e->clone_flags & LINUX_CLONE_PARENT_SETTID) {
395 		if (copyout_proc(parent, &p->p_pid,
396 		    e->parent_tidptr,  sizeof(p->p_pid)) != 0)
397 			printf("%s: LINUX_CLONE_PARENT_SETTID "
398 			    "failed (e->parent_tidptr = %p, "
399 			    "parent->p_pid = %d, p->p_pid = %d)\n",
400 			    __func__, e->parent_tidptr,
401 			    parent->p_pid, p->p_pid);
402 	}
403 
404 	/*
405 	 * CLONE_CHILD_SETTID and LINUX_CLONE_SETTLS require child's VM
406 	 * setup to be completed, we postpone them until userret time.
407 	 */
408 	if (e->clone_flags &
409 	    (LINUX_CLONE_CHILD_CLEARTID | LINUX_CLONE_SETTLS))
410 		p->p_userret = luserret;
411 
412 	return;
413 }
414 
415 void
416 linux_nptl_proc_init(struct proc *p, struct proc *parent)
417 {
418 	struct linux_emuldata *e = p->p_emuldata;
419 	struct linux_emuldata *ep;
420 
421 	if ((parent != NULL) && (parent->p_emuldata != NULL)) {
422 		ep = parent->p_emuldata;
423 
424 		e->parent_tidptr = ep->parent_tidptr;
425 		e->child_tidptr = ep->child_tidptr;
426 		e->clone_flags = ep->clone_flags;
427 	}
428 
429 	return;
430 }
431 #endif /* LINUX_NPTL */
432