linux_exec.c revision 1.85
1/*	$NetBSD: linux_exec.c,v 1.85 2006/08/23 19:49:09 manu Exp $	*/
2
3/*-
4 * Copyright (c) 1994, 1995, 1998, 2000 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 1.85 2006/08/23 19:49:09 manu 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/sa.h>
56#include <sys/syscallargs.h>
57
58#include <sys/ptrace.h>	/* For proc_reparent() */
59
60#include <uvm/uvm_extern.h>
61
62#include <machine/cpu.h>
63#include <machine/reg.h>
64
65#include <compat/linux/common/linux_types.h>
66#include <compat/linux/common/linux_signal.h>
67#include <compat/linux/common/linux_util.h>
68#include <compat/linux/common/linux_sched.h>
69#include <compat/linux/common/linux_machdep.h>
70#include <compat/linux/common/linux_exec.h>
71#include <compat/linux/common/linux_futex.h>
72
73#include <compat/linux/linux_syscallargs.h>
74#include <compat/linux/linux_syscall.h>
75#include <compat/linux/common/linux_misc.h>
76#include <compat/linux/common/linux_errno.h>
77#include <compat/linux/common/linux_emuldata.h>
78
79extern struct sysent linux_sysent[];
80extern const char * const linux_syscallnames[];
81extern char linux_sigcode[], linux_esigcode[];
82
83static void linux_e_proc_exec __P((struct proc *, struct exec_package *));
84static void linux_e_proc_fork __P((struct proc *, struct proc *, int));
85static void linux_e_proc_exit __P((struct proc *));
86static void linux_e_proc_init __P((struct proc *, struct proc *, int));
87
88#ifdef LINUX_NPTL
89void linux_userret __P((struct lwp *, void *));
90#endif
91
92/*
93 * Execve(2). Just check the alternate emulation path, and pass it on
94 * to the NetBSD execve().
95 */
96int
97linux_sys_execve(l, v, retval)
98	struct lwp *l;
99	void *v;
100	register_t *retval;
101{
102	struct linux_sys_execve_args /* {
103		syscallarg(const char *) path;
104		syscallarg(char **) argv;
105		syscallarg(char **) envp;
106	} */ *uap = v;
107	struct proc *p = l->l_proc;
108	struct sys_execve_args ap;
109	caddr_t sg;
110
111	sg = stackgap_init(p, 0);
112	CHECK_ALT_EXIST(l, &sg, SCARG(uap, path));
113
114	SCARG(&ap, path) = SCARG(uap, path);
115	SCARG(&ap, argp) = SCARG(uap, argp);
116	SCARG(&ap, envp) = SCARG(uap, envp);
117
118	return sys_execve(l, &ap, retval);
119}
120
121/*
122 * Emulation switch.
123 */
124
125struct uvm_object *emul_linux_object;
126
127const struct emul emul_linux = {
128	"linux",
129	"/emul/linux",
130#ifndef __HAVE_MINIMAL_EMUL
131	0,
132	(const int *)native_to_linux_errno,
133	LINUX_SYS_syscall,
134	LINUX_SYS_NSYSENT,
135#endif
136	linux_sysent,
137	linux_syscallnames,
138	linux_sendsig,
139	linux_trapsignal,
140	NULL,
141	linux_sigcode,
142	linux_esigcode,
143	&emul_linux_object,
144	linux_setregs,
145	linux_e_proc_exec,
146	linux_e_proc_fork,
147	linux_e_proc_exit,
148	NULL,
149	NULL,
150#ifdef __HAVE_SYSCALL_INTERN
151	linux_syscall_intern,
152#else
153#error Implement __HAVE_SYSCALL_INTERN for this platform
154#endif
155	NULL,
156	NULL,
157
158	uvm_default_mapaddr,
159
160	linux_usertrap,
161};
162
163static void
164linux_e_proc_init(p, parent, forkflags)
165	struct proc *p, *parent;
166	int forkflags;
167{
168	struct linux_emuldata *e = p->p_emuldata;
169	struct linux_emuldata_shared *s;
170	struct linux_emuldata *ep = NULL;
171
172	if (!e) {
173		/* allocate new Linux emuldata */
174		MALLOC(e, void *, sizeof(struct linux_emuldata),
175			M_EMULDATA, M_WAITOK);
176	} else  {
177		e->s->refs--;
178		if (e->s->refs == 0)
179			FREE(e->s, M_EMULDATA);
180	}
181
182	memset(e, '\0', sizeof(struct linux_emuldata));
183
184	e->proc = p;
185
186	if (parent)
187		ep = parent->p_emuldata;
188
189	if (forkflags & FORK_SHAREVM) {
190#ifdef DIAGNOSTIC
191		if (ep == NULL) {
192			killproc(p, "FORK_SHAREVM while emuldata is NULL\n");
193			return;
194		}
195#endif
196		s = ep->s;
197		s->refs++;
198	} else {
199		struct vmspace *vm;
200
201		MALLOC(s, void *, sizeof(struct linux_emuldata_shared),
202			M_EMULDATA, M_WAITOK);
203		s->refs = 1;
204
205		/*
206		 * Set the process idea of the break to the real value.
207		 * For fork, we use parent's vmspace since our's
208		 * is not setup at the time of this call and is going
209		 * to be copy of parent's anyway. For exec, just
210		 * use our own vmspace.
211		 */
212		vm = (parent) ? parent->p_vmspace : p->p_vmspace;
213		s->p_break = vm->vm_daddr + ctob(vm->vm_dsize);
214
215		/*
216		 * Linux threads are emulated as NetBSD processes (not lwp)
217		 * We use native PID for Linux TID. The Linux TID is the
218		 * PID of the first process in the group. It is stored
219		 * here
220		 */
221		s->group_pid = p->p_pid;
222
223		/*
224		 * Initialize the list of threads in the group
225		 */
226		LIST_INIT(&s->threads);
227
228		s->xstat = 0;
229		s->flags = 0;
230	}
231
232	e->s = s;
233
234	/*
235	 * Add this thread in the group thread list
236	 */
237	LIST_INSERT_HEAD(&s->threads, e, threads);
238
239#ifdef LINUX_NPTL
240	if (ep != NULL) {
241		e->parent_tidptr = ep->parent_tidptr;
242		e->child_tidptr = ep->child_tidptr;
243		e->clone_flags = ep->clone_flags;
244	}
245#endif /* LINUX_NPTL */
246
247	p->p_emuldata = e;
248}
249
250/*
251 * Allocate new per-process structures. Called when executing Linux
252 * process. We can reuse the old emuldata - if it's not null,
253 * the executed process is of same emulation as original forked one.
254 */
255static void
256linux_e_proc_exec(p, epp)
257	struct proc *p;
258	struct exec_package *epp;
259{
260	/* exec, use our vmspace */
261	linux_e_proc_init(p, NULL, 0);
262}
263
264/*
265 * Emulation per-process exit hook.
266 */
267static void
268linux_e_proc_exit(p)
269	struct proc *p;
270{
271	struct linux_emuldata *e = p->p_emuldata;
272
273#ifdef LINUX_NPTL
274	linux_nptl_proc_exit(p);
275#endif
276	/* Remove the thread for the group thread list */
277	LIST_REMOVE(e, threads);
278
279	/* free Linux emuldata and set the pointer to null */
280	e->s->refs--;
281	if (e->s->refs == 0)
282		FREE(e->s, M_EMULDATA);
283	FREE(e, M_EMULDATA);
284	p->p_emuldata = NULL;
285}
286
287/*
288 * Emulation fork hook.
289 */
290static void
291linux_e_proc_fork(p, parent, forkflags)
292	struct proc *p, *parent;
293	int forkflags;
294{
295	/*
296	 * The new process might share some vmspace-related stuff
297	 * with parent, depending on fork flags (CLONE_VM et.al).
298	 * Force allocation of new base emuldata, and share the
299	 * VM-related parts only if necessary.
300	 */
301	p->p_emuldata = NULL;
302	linux_e_proc_init(p, parent, forkflags);
303
304#ifdef LINUX_NPTL
305	linux_nptl_proc_fork(p, parent, (*linux_userret));
306#endif
307
308	return;
309}
310
311#ifdef LINUX_NPTL
312void
313linux_userret(l, arg)
314	struct lwp *l;
315	void *arg;
316{
317	struct proc *p = l->l_proc;
318	struct linux_emuldata *led = p->p_emuldata;
319	int error;
320
321	p->p_userret = NULL;
322
323	/* LINUX_CLONE_CHILD_SETTID: copy child's TID to child's memory  */
324	if (led->clone_flags & LINUX_CLONE_CHILD_SETTID) {
325		if ((error = copyout(&l->l_proc->p_pid,
326		    led->child_tidptr,  sizeof(l->l_proc->p_pid))) != 0)
327			printf("linux_userret: LINUX_CLONE_CHILD_SETTID "
328			    "failed (led->child_tidptr = %p, p->p_pid = %d)\n",
329			    led->child_tidptr, p->p_pid);
330	}
331
332	/* LINUX_CLONE_SETTLS: allocate a new TLS */
333	if (led->clone_flags & LINUX_CLONE_SETTLS) {
334		if (linux_set_newtls(l, linux_get_newtls(l)) != 0)
335			printf("linux_userret: linux_set_tls failed");
336	}
337
338	return;
339}
340
341void
342linux_nptl_proc_exit(p)
343	struct proc *p;
344{
345	struct linux_emuldata *e = p->p_emuldata;
346	int s;
347
348	/*
349	 * Check if we are a thread group leader victim of another
350	 * thread doing exit_group(). If we are, change the exit code.
351	 */
352	if ((e->s->group_pid == p->p_pid) &&
353	    (e->s->flags & LINUX_LES_INEXITGROUP)) {
354		p->p_xstat = e->s->xstat;
355	}
356
357	/*
358	 * Members of the thread groups others than the leader should
359	 * exit quietely: no zombie stage, no signal. We do that by
360	 * reparenting to init. init will collect us and nobody will
361	 * notice what happened.
362	 */
363#ifdef DEBUG_LINUX
364	printf("%s:%d e->s->group_pid = %d, p->p_pid = %d, flags = 0x%x\n",
365	    __func__, __LINE__, e->s->group_pid, p->p_pid, e->s->flags);
366#endif
367	if (e->s->group_pid != p->p_pid) {
368		wakeup(initproc);
369		SCHED_LOCK(s);
370		proc_reparent(p, initproc);
371		SCHED_UNLOCK(s);
372	}
373
374	/* Emulate LINUX_CLONE_CHILD_CLEARTID */
375	if (e->clear_tid != NULL) {
376		int error;
377		int null = 0;
378		struct linux_sys_futex_args cup;
379		register_t retval;
380		struct lwp *l;
381
382		error = copyout(&null, e->clear_tid, sizeof(null));
383#ifdef DEBUG_LINUX
384		if (error != 0)
385			printf("%s: cannot clear TID\n", __func__);
386#endif
387
388		l = proc_representative_lwp(p);
389		SCARG(&cup, uaddr) = e->clear_tid;
390		SCARG(&cup, op) = LINUX_FUTEX_WAKE;
391		SCARG(&cup, val) = 0x7fffffff; /* Awake everyone */
392		SCARG(&cup, timeout) = NULL;
393		SCARG(&cup, uaddr2) = NULL;
394		SCARG(&cup, val3) = 0;
395		if ((error = linux_sys_futex(l, &cup, &retval)) != 0)
396			printf("%s: linux_sys_futex failed\n", __func__);
397	}
398
399	return;
400}
401
402void
403linux_nptl_proc_fork(p, parent, luserret)
404	struct proc *p;
405	struct proc *parent;
406	void (luserret)(struct lwp *, void *);
407{
408#ifdef LINUX_NPTL
409	struct linux_emuldata *e;
410#endif
411	e = p->p_emuldata;
412
413	/* LINUX_CLONE_CHILD_CLEARTID: clear TID in child's memory on exit() */
414	if (e->clone_flags & LINUX_CLONE_CHILD_CLEARTID)
415		e->clear_tid = e->child_tidptr;
416
417	/* LINUX_CLONE_PARENT_SETTID: set child's TID in parent's memory */
418	if (e->clone_flags & LINUX_CLONE_PARENT_SETTID) {
419		if (copyout_proc(parent, &p->p_pid,
420		    e->parent_tidptr,  sizeof(p->p_pid)) != 0)
421			printf("%s: LINUX_CLONE_PARENT_SETTID "
422			    "failed (e->parent_tidptr = %p, "
423			    "parent->p_pid = %d, p->p_pid = %d)\n",
424			    __func__, e->parent_tidptr,
425			    parent->p_pid, p->p_pid);
426	}
427
428	/*
429	 * CLONE_CHILD_SETTID and LINUX_CLONE_SETTLS require child's VM
430	 * setup to be completed, we postpone them until userret time.
431	 */
432	if (e->clone_flags &
433	    (LINUX_CLONE_CHILD_CLEARTID | LINUX_CLONE_SETTLS))
434		p->p_userret = (*luserret);
435
436	return;
437}
438
439void
440linux_nptl_proc_init(p, parent)
441	struct proc *p;
442	struct proc *parent;
443{
444	struct linux_emuldata *e = p->p_emuldata;
445	struct linux_emuldata *ep;
446
447	if ((parent != NULL) && (parent->p_emuldata != NULL)) {
448		ep = parent->p_emuldata;
449
450		e->parent_tidptr = ep->parent_tidptr;
451		e->child_tidptr = ep->child_tidptr;
452		e->clone_flags = ep->clone_flags;
453	}
454
455	return;
456}
457
458
459#endif /* LINUX_NPTL */
460