1/*-
2 * Copyright (c) 2006 Roman Divacky
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer
10 *    in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote products
15 *    derived from this software without specific prior written permission
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD$");
31
32#include "opt_compat.h"
33#include "opt_kdtrace.h"
34
35#include <sys/param.h>
36#include <sys/systm.h>
37#include <sys/imgact.h>
38#include <sys/kernel.h>
39#include <sys/lock.h>
40#include <sys/malloc.h>
41#include <sys/mutex.h>
42#include <sys/sdt.h>
43#include <sys/sx.h>
44#include <sys/proc.h>
45#include <sys/syscallsubr.h>
46#include <sys/sysent.h>
47#include <sys/sysproto.h>
48#include <sys/unistd.h>
49
50#ifdef COMPAT_LINUX32
51#include <machine/../linux32/linux.h>
52#include <machine/../linux32/linux32_proto.h>
53#else
54#include <machine/../linux/linux.h>
55#include <machine/../linux/linux_proto.h>
56#endif
57
58#include <compat/linux/linux_dtrace.h>
59#include <compat/linux/linux_emul.h>
60#include <compat/linux/linux_futex.h>
61
62/**
63 * Special DTrace provider for the linuxulator.
64 *
65 * In this file we define the provider for the entire linuxulator. All
66 * modules (= files of the linuxulator) use it.
67 *
68 * We define a different name depending on the emulated bitsize, see
69 * ../../<ARCH>/linux{,32}/linux.h, e.g.:
70 *      native bitsize          = linuxulator
71 *      amd64, 32bit emulation  = linuxulator32
72 */
73LIN_SDT_PROVIDER_DEFINE(LINUX_DTRACE);
74
75/**
76 * Special DTrace module "locks", it covers some linuxulator internal
77 * locks.
78 */
79LIN_SDT_PROBE_DEFINE1(locks, emul_lock, locked, "struct mtx *");
80LIN_SDT_PROBE_DEFINE1(locks, emul_lock, unlock, "struct mtx *");
81LIN_SDT_PROBE_DEFINE1(locks, emul_shared_rlock, locked, "struct sx *");
82LIN_SDT_PROBE_DEFINE1(locks, emul_shared_rlock, unlock, "struct sx *");
83LIN_SDT_PROBE_DEFINE1(locks, emul_shared_wlock, locked, "struct sx *");
84LIN_SDT_PROBE_DEFINE1(locks, emul_shared_wlock, unlock, "struct sx *");
85
86/**
87 * DTrace probes in this module.
88 */
89LIN_SDT_PROBE_DEFINE2(emul, em_find, entry, "struct proc *", "int");
90LIN_SDT_PROBE_DEFINE0(emul, em_find, return);
91LIN_SDT_PROBE_DEFINE3(emul, proc_init, entry, "struct thread *", "pid_t",
92    "int");
93LIN_SDT_PROBE_DEFINE0(emul, proc_init, create_thread);
94LIN_SDT_PROBE_DEFINE0(emul, proc_init, fork);
95LIN_SDT_PROBE_DEFINE0(emul, proc_init, exec);
96LIN_SDT_PROBE_DEFINE0(emul, proc_init, return);
97LIN_SDT_PROBE_DEFINE1(emul, proc_exit, entry, "struct proc *");
98LIN_SDT_PROBE_DEFINE0(emul, proc_exit, futex_failed);
99LIN_SDT_PROBE_DEFINE3(emul, proc_exit, reparent, "pid_t", "pid_t",
100    "struct proc *");
101LIN_SDT_PROBE_DEFINE1(emul, proc_exit, child_clear_tid_error, "int");
102LIN_SDT_PROBE_DEFINE0(emul, proc_exit, return);
103LIN_SDT_PROBE_DEFINE2(emul, proc_exec, entry, "struct proc *",
104    "struct image_params *");
105LIN_SDT_PROBE_DEFINE0(emul, proc_exec, return);
106LIN_SDT_PROBE_DEFINE0(emul, linux_schedtail, entry);
107LIN_SDT_PROBE_DEFINE1(emul, linux_schedtail, copyout_error, "int");
108LIN_SDT_PROBE_DEFINE0(emul, linux_schedtail, return);
109LIN_SDT_PROBE_DEFINE1(emul, linux_set_tid_address, entry, "int *");
110LIN_SDT_PROBE_DEFINE0(emul, linux_set_tid_address, return);
111LIN_SDT_PROBE_DEFINE2(emul, linux_kill_threads, entry, "struct thread *",
112    "int");
113LIN_SDT_PROBE_DEFINE1(emul, linux_kill_threads, kill, "pid_t");
114LIN_SDT_PROBE_DEFINE0(emul, linux_kill_threads, return);
115
116struct sx	emul_shared_lock;
117struct mtx	emul_lock;
118
119/* this returns locked reference to the emuldata entry (if found) */
120struct linux_emuldata *
121em_find(struct proc *p, int locked)
122{
123	struct linux_emuldata *em;
124
125	LIN_SDT_PROBE2(emul, em_find, entry, p, locked);
126
127	if (locked == EMUL_DOLOCK)
128		EMUL_LOCK(&emul_lock);
129
130	em = p->p_emuldata;
131
132	if (em == NULL && locked == EMUL_DOLOCK)
133		EMUL_UNLOCK(&emul_lock);
134
135	LIN_SDT_PROBE1(emul, em_find, return, em);
136	return (em);
137}
138
139int
140linux_proc_init(struct thread *td, pid_t child, int flags)
141{
142	struct linux_emuldata *em, *p_em;
143	struct proc *p;
144
145	LIN_SDT_PROBE3(emul, proc_init, entry, td, child, flags);
146
147	if (child != 0) {
148		/* fork or create a thread */
149		em = malloc(sizeof *em, M_LINUX, M_WAITOK | M_ZERO);
150		em->pid = child;
151		em->pdeath_signal = 0;
152		em->flags = 0;
153		em->robust_futexes = NULL;
154		if (flags & LINUX_CLONE_THREAD) {
155			/* handled later in the code */
156			LIN_SDT_PROBE0(emul, proc_init, create_thread);
157		} else {
158			struct linux_emuldata_shared *s;
159
160			LIN_SDT_PROBE0(emul, proc_init, fork);
161
162			s = malloc(sizeof *s, M_LINUX, M_WAITOK | M_ZERO);
163			s->refs = 1;
164			s->group_pid = child;
165
166			LIST_INIT(&s->threads);
167			em->shared = s;
168		}
169	} else {
170		/* exec */
171		LIN_SDT_PROBE0(emul, proc_init, exec);
172
173		/* lookup the old one */
174		em = em_find(td->td_proc, EMUL_DOLOCK);
175		KASSERT(em != NULL, ("proc_init: emuldata not found in exec case.\n"));
176	}
177
178	em->child_clear_tid = NULL;
179	em->child_set_tid = NULL;
180
181	/*
182	 * allocate the shared struct only in clone()/fork cases in the case
183	 * of clone() td = calling proc and child = pid of the newly created
184	 * proc
185	 */
186	if (child != 0) {
187		if (flags & LINUX_CLONE_THREAD) {
188			/* lookup the parent */
189			/*
190			 * we dont have to lock the p_em because
191			 * its waiting for us in linux_clone so
192			 * there is no chance of it changing the
193			 * p_em->shared address
194			 */
195			p_em = em_find(td->td_proc, EMUL_DONTLOCK);
196			KASSERT(p_em != NULL, ("proc_init: parent emuldata not found for CLONE_THREAD\n"));
197			em->shared = p_em->shared;
198			EMUL_SHARED_WLOCK(&emul_shared_lock);
199			em->shared->refs++;
200			EMUL_SHARED_WUNLOCK(&emul_shared_lock);
201		} else {
202			/*
203			 * handled earlier to avoid malloc(M_WAITOK) with
204			 * rwlock held
205			 */
206		}
207
208		EMUL_SHARED_WLOCK(&emul_shared_lock);
209		LIST_INSERT_HEAD(&em->shared->threads, em, threads);
210		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
211
212		p = pfind(child);
213		KASSERT(p != NULL, ("process not found in proc_init\n"));
214		p->p_emuldata = em;
215		PROC_UNLOCK(p);
216	} else
217		EMUL_UNLOCK(&emul_lock);
218
219	LIN_SDT_PROBE0(emul, proc_init, return);
220	return (0);
221}
222
223void
224linux_proc_exit(void *arg __unused, struct proc *p)
225{
226	struct linux_emuldata *em;
227	int error, shared_flags, shared_xstat;
228	struct thread *td = FIRST_THREAD_IN_PROC(p);
229	int *child_clear_tid;
230	struct proc *q, *nq;
231
232	if (__predict_true(p->p_sysent != &elf_linux_sysvec))
233		return;
234
235	LIN_SDT_PROBE1(emul, proc_exit, entry, p);
236
237	release_futexes(p);
238
239	/* find the emuldata */
240	em = em_find(p, EMUL_DOLOCK);
241
242	KASSERT(em != NULL, ("proc_exit: emuldata not found.\n"));
243
244	/* reparent all procs that are not a thread leader to initproc */
245	if (em->shared->group_pid != p->p_pid) {
246		LIN_SDT_PROBE3(emul, proc_exit, reparent,
247		    em->shared->group_pid, p->p_pid, p);
248
249		child_clear_tid = em->child_clear_tid;
250		EMUL_UNLOCK(&emul_lock);
251		sx_xlock(&proctree_lock);
252		wakeup(initproc);
253		PROC_LOCK(p);
254		proc_reparent(p, initproc);
255		p->p_sigparent = SIGCHLD;
256		PROC_UNLOCK(p);
257		sx_xunlock(&proctree_lock);
258	} else {
259		child_clear_tid = em->child_clear_tid;
260		EMUL_UNLOCK(&emul_lock);
261	}
262
263	EMUL_SHARED_WLOCK(&emul_shared_lock);
264	shared_flags = em->shared->flags;
265	shared_xstat = em->shared->xstat;
266	LIST_REMOVE(em, threads);
267
268	em->shared->refs--;
269	if (em->shared->refs == 0) {
270		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
271		free(em->shared, M_LINUX);
272	} else
273		EMUL_SHARED_WUNLOCK(&emul_shared_lock);
274
275	if ((shared_flags & EMUL_SHARED_HASXSTAT) != 0)
276		p->p_xstat = shared_xstat;
277
278	if (child_clear_tid != NULL) {
279		struct linux_sys_futex_args cup;
280		int null = 0;
281
282		error = copyout(&null, child_clear_tid, sizeof(null));
283		if (error) {
284			LIN_SDT_PROBE1(emul, proc_exit,
285			    child_clear_tid_error, error);
286
287			free(em, M_LINUX);
288
289			LIN_SDT_PROBE0(emul, proc_exit, return);
290			return;
291		}
292
293		/* futexes stuff */
294		cup.uaddr = child_clear_tid;
295		cup.op = LINUX_FUTEX_WAKE;
296		cup.val = 0x7fffffff;	/* Awake everyone */
297		cup.timeout = NULL;
298		cup.uaddr2 = NULL;
299		cup.val3 = 0;
300		error = linux_sys_futex(FIRST_THREAD_IN_PROC(p), &cup);
301		/*
302		 * this cannot happen at the moment and if this happens it
303		 * probably means there is a user space bug
304		 */
305		if (error) {
306			LIN_SDT_PROBE0(emul, proc_exit, futex_failed);
307			printf(LMSG("futex stuff in proc_exit failed.\n"));
308		}
309	}
310
311	/* clean the stuff up */
312	free(em, M_LINUX);
313
314	/* this is a little weird but rewritten from exit1() */
315	sx_xlock(&proctree_lock);
316	q = LIST_FIRST(&p->p_children);
317	for (; q != NULL; q = nq) {
318		nq = LIST_NEXT(q, p_sibling);
319		if (q->p_flag & P_WEXIT)
320			continue;
321		if (__predict_false(q->p_sysent != &elf_linux_sysvec))
322			continue;
323		em = em_find(q, EMUL_DOLOCK);
324		KASSERT(em != NULL, ("linux_reparent: emuldata not found: %i\n", q->p_pid));
325		PROC_LOCK(q);
326		if ((q->p_flag & P_WEXIT) == 0 && em->pdeath_signal != 0) {
327			kern_psignal(q, em->pdeath_signal);
328		}
329		PROC_UNLOCK(q);
330		EMUL_UNLOCK(&emul_lock);
331	}
332	sx_xunlock(&proctree_lock);
333
334	LIN_SDT_PROBE0(emul, proc_exit, return);
335}
336
337/*
338 * This is used in a case of transition from FreeBSD binary execing to linux binary
339 * in this case we create linux emuldata proc entry with the pid of the currently running
340 * process.
341 */
342void
343linux_proc_exec(void *arg __unused, struct proc *p, struct image_params *imgp)
344{
345	if (__predict_false(imgp->sysent == &elf_linux_sysvec)) {
346		LIN_SDT_PROBE2(emul, proc_exec, entry, p, imgp);
347	}
348	if (__predict_false(imgp->sysent == &elf_linux_sysvec
349	    && p->p_sysent != &elf_linux_sysvec))
350		linux_proc_init(FIRST_THREAD_IN_PROC(p), p->p_pid, 0);
351	if (__predict_false((p->p_sysent->sv_flags & SV_ABI_MASK) ==
352	    SV_ABI_LINUX))
353		/* Kill threads regardless of imgp->sysent value */
354		linux_kill_threads(FIRST_THREAD_IN_PROC(p), SIGKILL);
355	if (__predict_false(imgp->sysent != &elf_linux_sysvec
356	    && p->p_sysent == &elf_linux_sysvec)) {
357		struct linux_emuldata *em;
358
359		/*
360		 * XXX:There's a race because here we assign p->p_emuldata NULL
361		 * but the process is still counted as linux one for a short
362 		 * time so some other process might reference it and try to
363 		 * access its p->p_emuldata and panicing on a NULL reference.
364		 */
365		em = em_find(p, EMUL_DONTLOCK);
366
367		KASSERT(em != NULL, ("proc_exec: emuldata not found.\n"));
368
369		EMUL_SHARED_WLOCK(&emul_shared_lock);
370		LIST_REMOVE(em, threads);
371
372		PROC_LOCK(p);
373		p->p_emuldata = NULL;
374		PROC_UNLOCK(p);
375
376		em->shared->refs--;
377		if (em->shared->refs == 0) {
378			EMUL_SHARED_WUNLOCK(&emul_shared_lock);
379			free(em->shared, M_LINUX);
380		} else
381			EMUL_SHARED_WUNLOCK(&emul_shared_lock);
382
383		free(em, M_LINUX);
384	}
385
386	if (__predict_false(imgp->sysent == &elf_linux_sysvec)) {
387		LIN_SDT_PROBE0(emul, proc_exec, return);
388	}
389}
390
391void
392linux_schedtail(struct thread *td)
393{
394	struct linux_emuldata *em;
395	struct proc *p;
396	int error = 0;
397	int *child_set_tid;
398
399	p = td->td_proc;
400
401	LIN_SDT_PROBE1(emul, linux_schedtail, entry, p);
402
403	/* find the emuldata */
404	em = em_find(p, EMUL_DOLOCK);
405
406	KASSERT(em != NULL, ("linux_schedtail: emuldata not found.\n"));
407	child_set_tid = em->child_set_tid;
408	EMUL_UNLOCK(&emul_lock);
409
410	if (child_set_tid != NULL) {
411		error = copyout(&p->p_pid, (int *)child_set_tid,
412		    sizeof(p->p_pid));
413
414		if (error != 0) {
415			LIN_SDT_PROBE1(emul, linux_schedtail, copyout_error,
416			    error);
417		}
418	}
419
420	LIN_SDT_PROBE0(emul, linux_schedtail, return);
421
422	return;
423}
424
425int
426linux_set_tid_address(struct thread *td, struct linux_set_tid_address_args *args)
427{
428	struct linux_emuldata *em;
429
430	LIN_SDT_PROBE1(emul, linux_set_tid_address, entry, args->tidptr);
431
432	/* find the emuldata */
433	em = em_find(td->td_proc, EMUL_DOLOCK);
434
435	KASSERT(em != NULL, ("set_tid_address: emuldata not found.\n"));
436
437	em->child_clear_tid = args->tidptr;
438	td->td_retval[0] = td->td_proc->p_pid;
439
440	EMUL_UNLOCK(&emul_lock);
441
442	LIN_SDT_PROBE0(emul, linux_set_tid_address, return);
443	return 0;
444}
445
446void
447linux_kill_threads(struct thread *td, int sig)
448{
449	struct linux_emuldata *em, *td_em, *tmp_em;
450	struct proc *sp;
451
452	LIN_SDT_PROBE2(emul, linux_kill_threads, entry, td, sig);
453
454	td_em = em_find(td->td_proc, EMUL_DONTLOCK);
455
456	KASSERT(td_em != NULL, ("linux_kill_threads: emuldata not found.\n"));
457
458	EMUL_SHARED_RLOCK(&emul_shared_lock);
459	LIST_FOREACH_SAFE(em, &td_em->shared->threads, threads, tmp_em) {
460		if (em->pid == td_em->pid)
461			continue;
462
463		sp = pfind(em->pid);
464		if ((sp->p_flag & P_WEXIT) == 0)
465			kern_psignal(sp, sig);
466		PROC_UNLOCK(sp);
467
468		LIN_SDT_PROBE1(emul, linux_kill_threads, kill, em->pid);
469	}
470	EMUL_SHARED_RUNLOCK(&emul_shared_lock);
471
472	LIN_SDT_PROBE0(emul, linux_kill_threads, return);
473}
474