1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 1994-1996 Søren Schmidt
5 * Copyright (c) 2006 Roman Divacky
6 * All rights reserved.
7 * Copyright (c) 2013 Dmitry Chagin <dchagin@FreeBSD.org>
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/param.h>
32 #include <sys/fcntl.h>
33 #include <sys/imgact.h>
34 #include <sys/ktr.h>
35 #include <sys/lock.h>
36 #include <sys/malloc.h>
37 #include <sys/mutex.h>
38 #include <sys/proc.h>
39 #include <sys/resourcevar.h>
40 #include <sys/sx.h>
41 #include <sys/syscallsubr.h>
42 #include <sys/sysent.h>
43
44 #include <compat/linux/linux_emul.h>
45 #include <compat/linux/linux_mib.h>
46 #include <compat/linux/linux_misc.h>
47 #include <compat/linux/linux_persona.h>
48 #include <compat/linux/linux_util.h>
49
50 #if BYTE_ORDER == LITTLE_ENDIAN
51 #define SHELLMAGIC 0x2123 /* #! */
52 #else
53 #define SHELLMAGIC 0x2321
54 #endif
55
56 /*
57 * This returns reference to the thread emuldata entry (if found)
58 *
59 * Hold PROC_LOCK when referencing emuldata from other threads.
60 */
61 struct linux_emuldata *
em_find(struct thread * td)62 em_find(struct thread *td)
63 {
64 struct linux_emuldata *em;
65
66 em = td->td_emuldata;
67
68 return (em);
69 }
70
71 /*
72 * This returns reference to the proc pemuldata entry (if found)
73 *
74 * Hold PROC_LOCK when referencing proc pemuldata from other threads.
75 * Hold LINUX_PEM_LOCK wher referencing pemuldata members.
76 */
77 struct linux_pemuldata *
pem_find(struct proc * p)78 pem_find(struct proc *p)
79 {
80 struct linux_pemuldata *pem;
81
82 pem = p->p_emuldata;
83
84 return (pem);
85 }
86
87 /*
88 * Linux apps generally expect the soft open file limit to be set
89 * to 1024, often iterating over all the file descriptors up to that
90 * limit instead of using closefrom(2). Give them what they want,
91 * unless there already is a resource limit in place.
92 */
93 static void
linux_set_default_openfiles(struct thread * td,struct proc * p)94 linux_set_default_openfiles(struct thread *td, struct proc *p)
95 {
96 struct rlimit rlim;
97 int error __diagused;
98
99 if (linux_default_openfiles < 0)
100 return;
101
102 PROC_LOCK(p);
103 lim_rlimit_proc(p, RLIMIT_NOFILE, &rlim);
104 PROC_UNLOCK(p);
105 if (rlim.rlim_cur != rlim.rlim_max ||
106 rlim.rlim_cur <= linux_default_openfiles)
107 return;
108 rlim.rlim_cur = linux_default_openfiles;
109 error = kern_proc_setrlimit(td, p, RLIMIT_NOFILE, &rlim);
110 KASSERT(error == 0, ("kern_proc_setrlimit failed"));
111 }
112
113 /*
114 * The default stack size limit in Linux is 8MB.
115 */
116 static void
linux_set_default_stacksize(struct thread * td,struct proc * p)117 linux_set_default_stacksize(struct thread *td, struct proc *p)
118 {
119 struct rlimit rlim;
120 int error __diagused;
121
122 if (linux_default_stacksize < 0)
123 return;
124
125 PROC_LOCK(p);
126 lim_rlimit_proc(p, RLIMIT_STACK, &rlim);
127 PROC_UNLOCK(p);
128 if (rlim.rlim_cur != rlim.rlim_max ||
129 rlim.rlim_cur <= linux_default_stacksize)
130 return;
131 rlim.rlim_cur = linux_default_stacksize;
132 error = kern_proc_setrlimit(td, p, RLIMIT_STACK, &rlim);
133 KASSERT(error == 0, ("kern_proc_setrlimit failed"));
134 }
135
136 void
linux_proc_init(struct thread * td,struct thread * newtd,bool init_thread)137 linux_proc_init(struct thread *td, struct thread *newtd, bool init_thread)
138 {
139 struct linux_emuldata *em;
140 struct linux_pemuldata *pem;
141 struct proc *p;
142
143 if (newtd != NULL) {
144 p = newtd->td_proc;
145
146 /* non-exec call */
147 em = malloc(sizeof(*em), M_LINUX, M_WAITOK | M_ZERO);
148 if (init_thread) {
149 LINUX_CTR1(proc_init, "thread newtd(%d)",
150 newtd->td_tid);
151
152 em->em_tid = newtd->td_tid;
153 } else {
154 LINUX_CTR1(proc_init, "fork newtd(%d)", p->p_pid);
155
156 em->em_tid = p->p_pid;
157
158 pem = malloc(sizeof(*pem), M_LINUX, M_WAITOK | M_ZERO);
159 sx_init(&pem->pem_sx, "lpemlk");
160 p->p_emuldata = pem;
161 }
162 newtd->td_emuldata = em;
163
164 linux_set_default_openfiles(td, p);
165 linux_set_default_stacksize(td, p);
166 } else {
167 p = td->td_proc;
168
169 /* exec */
170 LINUX_CTR1(proc_init, "exec newtd(%d)", p->p_pid);
171
172 /* lookup the old one */
173 em = em_find(td);
174 KASSERT(em != NULL, ("proc_init: thread emuldata not found.\n"));
175
176 em->em_tid = p->p_pid;
177 em->flags = 0;
178 em->robust_futexes = NULL;
179 em->child_clear_tid = NULL;
180 em->child_set_tid = NULL;
181
182 pem = pem_find(p);
183 KASSERT(pem != NULL, ("proc_init: proc emuldata not found.\n"));
184 pem->persona = 0;
185 pem->oom_score_adj = 0;
186 }
187 }
188
189 void
linux_on_exit(struct proc * p)190 linux_on_exit(struct proc *p)
191 {
192 struct linux_pemuldata *pem;
193 struct thread *td = curthread;
194
195 MPASS(SV_CURPROC_ABI() == SV_ABI_LINUX);
196
197 LINUX_CTR3(proc_exit, "thread(%d) proc(%d) p %p",
198 td->td_tid, p->p_pid, p);
199
200 pem = pem_find(p);
201 if (pem == NULL)
202 return;
203 (p->p_sysent->sv_thread_detach)(td);
204
205 p->p_emuldata = NULL;
206
207 sx_destroy(&pem->pem_sx);
208 free(pem, M_LINUX);
209 }
210
211 int
linux_common_execve(struct thread * td,struct image_args * eargs)212 linux_common_execve(struct thread *td, struct image_args *eargs)
213 {
214 struct linux_pemuldata *pem;
215 struct vmspace *oldvmspace;
216 struct linux_emuldata *em;
217 struct proc *p;
218 int error;
219
220 p = td->td_proc;
221
222 error = pre_execve(td, &oldvmspace);
223 if (error != 0)
224 return (error);
225
226 error = kern_execve(td, eargs, NULL, oldvmspace);
227 post_execve(td, error, oldvmspace);
228 if (error != EJUSTRETURN)
229 return (error);
230
231 /*
232 * In a case of transition from Linux binary execing to
233 * FreeBSD binary we destroy Linux emuldata thread & proc entries.
234 */
235 if (SV_CURPROC_ABI() != SV_ABI_LINUX) {
236
237 /* Clear ABI root directory if set. */
238 linux_pwd_onexec_native(td);
239
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 free(em, M_LINUX);
251 free(pem, M_LINUX);
252 }
253 return (EJUSTRETURN);
254 }
255
256 int
linux_on_exec(struct proc * p,struct image_params * imgp)257 linux_on_exec(struct proc *p, struct image_params *imgp)
258 {
259 struct thread *td;
260 struct thread *othertd;
261 #if defined(__amd64__)
262 struct linux_pemuldata *pem;
263 #endif
264 int error;
265
266 td = curthread;
267 MPASS((imgp->sysent->sv_flags & SV_ABI_MASK) == SV_ABI_LINUX);
268
269 /*
270 * When execing to Linux binary, we create Linux emuldata
271 * thread entry.
272 */
273 if (SV_PROC_ABI(p) == SV_ABI_LINUX) {
274 /*
275 * Process already was under Linuxolator
276 * before exec. Update emuldata to reflect
277 * single-threaded cleaned state after exec.
278 */
279 linux_proc_init(td, NULL, false);
280 } else {
281 /*
282 * We are switching the process to Linux emulator.
283 */
284 linux_proc_init(td, td, false);
285
286 /*
287 * Create a transient td_emuldata for all suspended
288 * threads, so that p->p_sysent->sv_thread_detach() ==
289 * linux_thread_detach() can find expected but unused
290 * emuldata.
291 */
292 FOREACH_THREAD_IN_PROC(td->td_proc, othertd) {
293 if (othertd == td)
294 continue;
295 linux_proc_init(td, othertd, true);
296 }
297
298 /* Set ABI root directory. */
299 if ((error = linux_pwd_onexec(td)) != 0)
300 return (error);
301 }
302 #if defined(__amd64__)
303 /*
304 * An IA32 executable which has executable stack will have the
305 * READ_IMPLIES_EXEC personality flag set automatically.
306 */
307 if (SV_PROC_FLAG(td->td_proc, SV_ILP32) &&
308 imgp->stack_prot & VM_PROT_EXECUTE) {
309 pem = pem_find(p);
310 pem->persona |= LINUX_READ_IMPLIES_EXEC;
311 }
312 #endif
313 return (0);
314 }
315
316 void
linux_thread_dtor(struct thread * td)317 linux_thread_dtor(struct thread *td)
318 {
319 struct linux_emuldata *em;
320
321 em = em_find(td);
322 if (em == NULL)
323 return;
324 td->td_emuldata = NULL;
325
326 LINUX_CTR1(thread_dtor, "thread(%d)", em->em_tid);
327
328 free(em, M_LINUX);
329 }
330
331 void
linux_schedtail(struct thread * td)332 linux_schedtail(struct thread *td)
333 {
334 struct linux_emuldata *em;
335 #ifdef KTR
336 int error;
337 #else
338 int error __unused;
339 #endif
340 int *child_set_tid;
341
342 em = em_find(td);
343 KASSERT(em != NULL, ("linux_schedtail: thread emuldata not found.\n"));
344 child_set_tid = em->child_set_tid;
345
346 if (child_set_tid != NULL) {
347 error = copyout(&em->em_tid, child_set_tid,
348 sizeof(em->em_tid));
349 LINUX_CTR4(schedtail, "thread(%d) %p stored %d error %d",
350 td->td_tid, child_set_tid, em->em_tid, error);
351 } else
352 LINUX_CTR1(schedtail, "thread(%d)", em->em_tid);
353 }
354