xref: /freebsd/sys/compat/linux/linux_emul.c (revision 4e462178745853ecc014c13f82f89cfe39b83e9c)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1994-1996 Søren Schmidt
5  * Copyright (c) 2006 Roman Divacky
6  * Copyright (c) 2013 Dmitry Chagin
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  */
30 
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
33 
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/fcntl.h>
37 #include <sys/imgact.h>
38 #include <sys/kernel.h>
39 #include <sys/ktr.h>
40 #include <sys/lock.h>
41 #include <sys/malloc.h>
42 #include <sys/mutex.h>
43 #include <sys/sx.h>
44 #include <sys/proc.h>
45 #include <sys/syscallsubr.h>
46 #include <sys/sysent.h>
47 
48 #include <compat/linux/linux_emul.h>
49 #include <compat/linux/linux_misc.h>
50 #include <compat/linux/linux_persona.h>
51 #include <compat/linux/linux_util.h>
52 
53 #if BYTE_ORDER == LITTLE_ENDIAN
54 #define SHELLMAGIC	0x2123 /* #! */
55 #else
56 #define SHELLMAGIC	0x2321
57 #endif
58 
59 /*
60  * This returns reference to the thread emuldata entry (if found)
61  *
62  * Hold PROC_LOCK when referencing emuldata from other threads.
63  */
64 struct linux_emuldata *
65 em_find(struct thread *td)
66 {
67 	struct linux_emuldata *em;
68 
69 	em = td->td_emuldata;
70 
71 	return (em);
72 }
73 
74 /*
75  * This returns reference to the proc pemuldata entry (if found)
76  *
77  * Hold PROC_LOCK when referencing proc pemuldata from other threads.
78  * Hold LINUX_PEM_LOCK wher referencing pemuldata members.
79  */
80 struct linux_pemuldata *
81 pem_find(struct proc *p)
82 {
83 	struct linux_pemuldata *pem;
84 
85 	pem = p->p_emuldata;
86 
87 	return (pem);
88 }
89 
90 void
91 linux_proc_init(struct thread *td, struct thread *newtd, int flags)
92 {
93 	struct linux_emuldata *em;
94 	struct linux_pemuldata *pem;
95 	struct epoll_emuldata *emd;
96 	struct proc *p;
97 
98 	if (newtd != NULL) {
99 		p = newtd->td_proc;
100 
101 		/* non-exec call */
102 		em = malloc(sizeof(*em), M_TEMP, M_WAITOK | M_ZERO);
103 		if (flags & LINUX_CLONE_THREAD) {
104 			LINUX_CTR1(proc_init, "thread newtd(%d)",
105 			    newtd->td_tid);
106 
107 			em->em_tid = newtd->td_tid;
108 		} else {
109 			LINUX_CTR1(proc_init, "fork newtd(%d)", p->p_pid);
110 
111 			em->em_tid = p->p_pid;
112 
113 			pem = malloc(sizeof(*pem), M_LINUX, M_WAITOK | M_ZERO);
114 			sx_init(&pem->pem_sx, "lpemlk");
115 			p->p_emuldata = pem;
116 		}
117 		newtd->td_emuldata = em;
118 	} else {
119 		p = td->td_proc;
120 
121 		/* exec */
122 		LINUX_CTR1(proc_init, "exec newtd(%d)", p->p_pid);
123 
124 		/* lookup the old one */
125 		em = em_find(td);
126 		KASSERT(em != NULL, ("proc_init: emuldata not found in exec case.\n"));
127 
128 		em->em_tid = p->p_pid;
129 		em->flags = 0;
130 		em->pdeath_signal = 0;
131 		em->robust_futexes = NULL;
132 		em->child_clear_tid = NULL;
133 		em->child_set_tid = NULL;
134 
135 		 /* epoll should be destroyed in a case of exec. */
136 		pem = pem_find(p);
137 		KASSERT(pem != NULL, ("proc_exit: proc emuldata not found.\n"));
138 		pem->persona = 0;
139 		if (pem->epoll != NULL) {
140 			emd = pem->epoll;
141 			pem->epoll = NULL;
142 			free(emd, M_EPOLL);
143 		}
144 	}
145 
146 }
147 
148 void
149 linux_proc_exit(void *arg __unused, struct proc *p)
150 {
151 	struct linux_pemuldata *pem;
152 	struct epoll_emuldata *emd;
153 	struct thread *td = curthread;
154 
155 	if (__predict_false(SV_CURPROC_ABI() != SV_ABI_LINUX))
156 		return;
157 
158 	LINUX_CTR3(proc_exit, "thread(%d) proc(%d) p %p",
159 	    td->td_tid, p->p_pid, p);
160 
161 	pem = pem_find(p);
162 	if (pem == NULL)
163 		return;
164 	(p->p_sysent->sv_thread_detach)(td);
165 
166 	p->p_emuldata = NULL;
167 
168 	if (pem->epoll != NULL) {
169 		emd = pem->epoll;
170 		pem->epoll = NULL;
171 		free(emd, M_EPOLL);
172 	}
173 
174 	sx_destroy(&pem->pem_sx);
175 	free(pem, M_LINUX);
176 }
177 
178 /*
179  * If a Linux binary is exec'ing something, try this image activator
180  * first.  We override standard shell script execution in order to
181  * be able to modify the interpreter path.  We only do this if a Linux
182  * binary is doing the exec, so we do not create an EXEC module for it.
183  */
184 int
185 linux_exec_imgact_try(struct image_params *imgp)
186 {
187 	const char *head = (const char *)imgp->image_header;
188 	char *rpath;
189 	int error = -1;
190 
191 	/*
192 	 * The interpreter for shell scripts run from a Linux binary needs
193 	 * to be located in /compat/linux if possible in order to recursively
194 	 * maintain Linux path emulation.
195 	 */
196 	if (((const short *)head)[0] == SHELLMAGIC) {
197 		/*
198 		 * Run our normal shell image activator.  If it succeeds attempt
199 		 * to use the alternate path for the interpreter.  If an
200 		 * alternate path is found, use our stringspace to store it.
201 		 */
202 		if ((error = exec_shell_imgact(imgp)) == 0) {
203 			linux_emul_convpath(FIRST_THREAD_IN_PROC(imgp->proc),
204 			    imgp->interpreter_name, UIO_SYSSPACE, &rpath, 0,
205 			    AT_FDCWD);
206 			if (rpath != NULL)
207 				imgp->args->fname_buf =
208 				    imgp->interpreter_name = rpath;
209 		}
210 	}
211 	return (error);
212 }
213 
214 int
215 linux_common_execve(struct thread *td, struct image_args *eargs)
216 {
217 	struct linux_pemuldata *pem;
218 	struct epoll_emuldata *emd;
219 	struct vmspace *oldvmspace;
220 	struct linux_emuldata *em;
221 	struct proc *p;
222 	int error;
223 
224 	p = td->td_proc;
225 
226 	error = pre_execve(td, &oldvmspace);
227 	if (error != 0)
228 		return (error);
229 
230 	error = kern_execve(td, eargs, NULL);
231 	post_execve(td, error, oldvmspace);
232 	if (error != EJUSTRETURN)
233 		return (error);
234 
235 	/*
236 	 * In a case of transition from Linux binary execing to
237 	 * FreeBSD binary we destroy Linux emuldata thread & proc entries.
238 	 */
239 	if (SV_CURPROC_ABI() != SV_ABI_LINUX) {
240 		PROC_LOCK(p);
241 		em = em_find(td);
242 		KASSERT(em != NULL, ("proc_exec: thread emuldata not found.\n"));
243 		td->td_emuldata = NULL;
244 
245 		pem = pem_find(p);
246 		KASSERT(pem != NULL, ("proc_exec: proc pemuldata not found.\n"));
247 		p->p_emuldata = NULL;
248 		PROC_UNLOCK(p);
249 
250 		if (pem->epoll != NULL) {
251 			emd = pem->epoll;
252 			pem->epoll = NULL;
253 			free(emd, M_EPOLL);
254 		}
255 
256 		free(em, M_TEMP);
257 		free(pem, M_LINUX);
258 	}
259 	return (EJUSTRETURN);
260 }
261 
262 void
263 linux_proc_exec(void *arg __unused, struct proc *p, struct image_params *imgp)
264 {
265 	struct thread *td = curthread;
266 	struct thread *othertd;
267 #if defined(__amd64__)
268 	struct linux_pemuldata *pem;
269 #endif
270 
271 	/*
272 	 * In a case of execing from Linux binary properly detach
273 	 * other threads from the user space.
274 	 */
275 	if (__predict_false(SV_PROC_ABI(p) == SV_ABI_LINUX)) {
276 		FOREACH_THREAD_IN_PROC(p, othertd) {
277 			if (td != othertd)
278 				(p->p_sysent->sv_thread_detach)(othertd);
279 		}
280 	}
281 
282 	/*
283 	 * In a case of execing to Linux binary we create Linux
284 	 * emuldata thread entry.
285 	 */
286 	if (__predict_false((imgp->sysent->sv_flags & SV_ABI_MASK) ==
287 	    SV_ABI_LINUX)) {
288 
289 		if (SV_PROC_ABI(p) == SV_ABI_LINUX)
290 			linux_proc_init(td, NULL, 0);
291 		else
292 			linux_proc_init(td, td, 0);
293 #if defined(__amd64__)
294 		/*
295 		 * An IA32 executable which has executable stack will have the
296 		 * READ_IMPLIES_EXEC personality flag set automatically.
297 		 */
298 		if (SV_PROC_FLAG(td->td_proc, SV_ILP32) &&
299 		    imgp->stack_prot & VM_PROT_EXECUTE) {
300 			pem = pem_find(p);
301 			pem->persona |= LINUX_READ_IMPLIES_EXEC;
302 		}
303 #endif
304 	}
305 }
306 
307 void
308 linux_thread_dtor(void *arg __unused, struct thread *td)
309 {
310 	struct linux_emuldata *em;
311 
312 	em = em_find(td);
313 	if (em == NULL)
314 		return;
315 	td->td_emuldata = NULL;
316 
317 	LINUX_CTR1(thread_dtor, "thread(%d)", em->em_tid);
318 
319 	free(em, M_TEMP);
320 }
321 
322 void
323 linux_schedtail(struct thread *td)
324 {
325 	struct linux_emuldata *em;
326 	struct proc *p;
327 	int error = 0;
328 	int *child_set_tid;
329 
330 	p = td->td_proc;
331 
332 	em = em_find(td);
333 	KASSERT(em != NULL, ("linux_schedtail: thread emuldata not found.\n"));
334 	child_set_tid = em->child_set_tid;
335 
336 	if (child_set_tid != NULL) {
337 		error = copyout(&em->em_tid, child_set_tid,
338 		    sizeof(em->em_tid));
339 		LINUX_CTR4(schedtail, "thread(%d) %p stored %d error %d",
340 		    td->td_tid, child_set_tid, em->em_tid, error);
341 	} else
342 		LINUX_CTR1(schedtail, "thread(%d)", em->em_tid);
343 }
344