xref: /freebsd/sys/kern/kern_proc.c (revision 1123986db5f0b0b3bfc68840a6d8e12f93d6ca3e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 #include "opt_ddb.h"
34 #include "opt_ktrace.h"
35 #include "opt_kstack_pages.h"
36 #include "opt_stack.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/bitstring.h>
41 #include <sys/conf.h>
42 #include <sys/elf.h>
43 #include <sys/eventhandler.h>
44 #include <sys/exec.h>
45 #include <sys/fcntl.h>
46 #include <sys/ipc.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/limits.h>
50 #include <sys/lock.h>
51 #include <sys/loginclass.h>
52 #include <sys/malloc.h>
53 #include <sys/mman.h>
54 #include <sys/mount.h>
55 #include <sys/mutex.h>
56 #include <sys/namei.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/shm.h>
66 #include <sys/smp.h>
67 #include <sys/stack.h>
68 #include <sys/stat.h>
69 #include <sys/dtrace_bsd.h>
70 #include <sys/sysctl.h>
71 #include <sys/filedesc.h>
72 #include <sys/tty.h>
73 #include <sys/signalvar.h>
74 #include <sys/sdt.h>
75 #include <sys/sx.h>
76 #include <sys/user.h>
77 #include <sys/vnode.h>
78 #include <sys/wait.h>
79 #ifdef KTRACE
80 #include <sys/ktrace.h>
81 #endif
82 
83 #ifdef DDB
84 #include <ddb/ddb.h>
85 #endif
86 
87 #include <vm/vm.h>
88 #include <vm/vm_param.h>
89 #include <vm/vm_extern.h>
90 #include <vm/pmap.h>
91 #include <vm/vm_map.h>
92 #include <vm/vm_object.h>
93 #include <vm/vm_page.h>
94 #include <vm/vm_pager.h>
95 #include <vm/uma.h>
96 
97 #include <fs/devfs/devfs.h>
98 
99 #ifdef COMPAT_FREEBSD32
100 #include <compat/freebsd32/freebsd32.h>
101 #include <compat/freebsd32/freebsd32_util.h>
102 #endif
103 
104 SDT_PROVIDER_DEFINE(proc);
105 
106 MALLOC_DEFINE(M_SESSION, "session", "session header");
107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
109 
110 static void doenterpgrp(struct proc *, struct pgrp *);
111 static void orphanpg(struct pgrp *pg);
112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
115     int preferthread);
116 static void pgdelete(struct pgrp *);
117 static int pgrp_init(void *mem, int size, int flags);
118 static int proc_ctor(void *mem, int size, void *arg, int flags);
119 static void proc_dtor(void *mem, int size, void *arg);
120 static int proc_init(void *mem, int size, int flags);
121 static void proc_fini(void *mem, int size);
122 static void pargs_free(struct pargs *pa);
123 
124 /*
125  * Other process lists
126  */
127 struct pidhashhead *pidhashtbl = NULL;
128 struct sx *pidhashtbl_lock;
129 u_long pidhash;
130 u_long pidhashlock;
131 struct pgrphashhead *pgrphashtbl;
132 u_long pgrphash;
133 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
134 struct sx __exclusive_cache_line allproc_lock;
135 struct sx __exclusive_cache_line proctree_lock;
136 struct mtx __exclusive_cache_line ppeers_lock;
137 struct mtx __exclusive_cache_line procid_lock;
138 uma_zone_t proc_zone;
139 uma_zone_t pgrp_zone;
140 
141 /*
142  * The offset of various fields in struct proc and struct thread.
143  * These are used by kernel debuggers to enumerate kernel threads and
144  * processes.
145  */
146 const int proc_off_p_pid = offsetof(struct proc, p_pid);
147 const int proc_off_p_comm = offsetof(struct proc, p_comm);
148 const int proc_off_p_list = offsetof(struct proc, p_list);
149 const int proc_off_p_hash = offsetof(struct proc, p_hash);
150 const int proc_off_p_threads = offsetof(struct proc, p_threads);
151 const int thread_off_td_tid = offsetof(struct thread, td_tid);
152 const int thread_off_td_name = offsetof(struct thread, td_name);
153 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
154 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
155 const int thread_off_td_plist = offsetof(struct thread, td_plist);
156 
157 EVENTHANDLER_LIST_DEFINE(process_ctor);
158 EVENTHANDLER_LIST_DEFINE(process_dtor);
159 EVENTHANDLER_LIST_DEFINE(process_init);
160 EVENTHANDLER_LIST_DEFINE(process_fini);
161 EVENTHANDLER_LIST_DEFINE(process_exit);
162 EVENTHANDLER_LIST_DEFINE(process_fork);
163 EVENTHANDLER_LIST_DEFINE(process_exec);
164 
165 int kstack_pages = KSTACK_PAGES;
166 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
167     &kstack_pages, 0,
168     "Kernel stack size in pages");
169 static int vmmap_skip_res_cnt = 0;
170 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
171     &vmmap_skip_res_cnt, 0,
172     "Skip calculation of the pages resident count in kern.proc.vmmap");
173 
174 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
175 #ifdef COMPAT_FREEBSD32
176 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
177 #endif
178 
179 /*
180  * Initialize global process hashing structures.
181  */
182 void
procinit(void)183 procinit(void)
184 {
185 	u_long i;
186 
187 	sx_init(&allproc_lock, "allproc");
188 	sx_init(&proctree_lock, "proctree");
189 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
190 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
191 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
192 	pidhashlock = (pidhash + 1) / 64;
193 	if (pidhashlock > 0)
194 		pidhashlock--;
195 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
196 	    M_PROC, M_WAITOK | M_ZERO);
197 	for (i = 0; i < pidhashlock + 1; i++)
198 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
199 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
200 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
201 	    proc_ctor, proc_dtor, proc_init, proc_fini,
202 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
203 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
204 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
205 	uihashinit();
206 }
207 
208 /*
209  * Prepare a proc for use.
210  */
211 static int
proc_ctor(void * mem,int size,void * arg,int flags)212 proc_ctor(void *mem, int size, void *arg, int flags)
213 {
214 	struct proc *p;
215 	struct thread *td;
216 
217 	p = (struct proc *)mem;
218 #ifdef KDTRACE_HOOKS
219 	kdtrace_proc_ctor(p);
220 #endif
221 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
222 	td = FIRST_THREAD_IN_PROC(p);
223 	if (td != NULL) {
224 		/* Make sure all thread constructors are executed */
225 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
226 	}
227 	return (0);
228 }
229 
230 /*
231  * Reclaim a proc after use.
232  */
233 static void
proc_dtor(void * mem,int size,void * arg)234 proc_dtor(void *mem, int size, void *arg)
235 {
236 	struct proc *p;
237 	struct thread *td;
238 
239 	/* INVARIANTS checks go here */
240 	p = (struct proc *)mem;
241 	td = FIRST_THREAD_IN_PROC(p);
242 	if (td != NULL) {
243 #ifdef INVARIANTS
244 		KASSERT((p->p_numthreads == 1),
245 		    ("bad number of threads in exiting process"));
246 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
247 #endif
248 		/* Free all OSD associated to this thread. */
249 		osd_thread_exit(td);
250 		ast_kclear(td);
251 
252 		/* Make sure all thread destructors are executed */
253 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
254 	}
255 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
256 #ifdef KDTRACE_HOOKS
257 	kdtrace_proc_dtor(p);
258 #endif
259 	if (p->p_ksi != NULL)
260 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
261 }
262 
263 /*
264  * Initialize type-stable parts of a proc (when newly created).
265  */
266 static int
proc_init(void * mem,int size,int flags)267 proc_init(void *mem, int size, int flags)
268 {
269 	struct proc *p;
270 
271 	p = (struct proc *)mem;
272 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
273 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
274 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
275 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
276 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
277 	cv_init(&p->p_pwait, "ppwait");
278 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
279 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
280 	p->p_stats = pstats_alloc();
281 	p->p_pgrp = NULL;
282 	TAILQ_INIT(&p->p_kqtim_stop);
283 	return (0);
284 }
285 
286 /*
287  * UMA should ensure that this function is never called.
288  * Freeing a proc structure would violate type stability.
289  */
290 static void
proc_fini(void * mem,int size)291 proc_fini(void *mem, int size)
292 {
293 #ifdef notnow
294 	struct proc *p;
295 
296 	p = (struct proc *)mem;
297 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
298 	pstats_free(p->p_stats);
299 	thread_free(FIRST_THREAD_IN_PROC(p));
300 	mtx_destroy(&p->p_mtx);
301 	if (p->p_ksi != NULL)
302 		ksiginfo_free(p->p_ksi);
303 #else
304 	panic("proc reclaimed");
305 #endif
306 }
307 
308 static int
pgrp_init(void * mem,int size,int flags)309 pgrp_init(void *mem, int size, int flags)
310 {
311 	struct pgrp *pg;
312 
313 	pg = mem;
314 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
315 	sx_init(&pg->pg_killsx, "killpg racer");
316 	return (0);
317 }
318 
319 /*
320  * PID space management.
321  *
322  * These bitmaps are used by fork_findpid.
323  */
324 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
325 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
326 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
327 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
328 
329 static bitstr_t *proc_id_array[] = {
330 	proc_id_pidmap,
331 	proc_id_grpidmap,
332 	proc_id_sessidmap,
333 	proc_id_reapmap,
334 };
335 
336 void
proc_id_set(int type,pid_t id)337 proc_id_set(int type, pid_t id)
338 {
339 
340 	KASSERT(type >= 0 && type < nitems(proc_id_array),
341 	    ("invalid type %d\n", type));
342 	mtx_lock(&procid_lock);
343 	KASSERT(bit_test(proc_id_array[type], id) == 0,
344 	    ("bit %d already set in %d\n", id, type));
345 	bit_set(proc_id_array[type], id);
346 	mtx_unlock(&procid_lock);
347 }
348 
349 void
proc_id_set_cond(int type,pid_t id)350 proc_id_set_cond(int type, pid_t id)
351 {
352 
353 	KASSERT(type >= 0 && type < nitems(proc_id_array),
354 	    ("invalid type %d\n", type));
355 	if (bit_test(proc_id_array[type], id))
356 		return;
357 	mtx_lock(&procid_lock);
358 	bit_set(proc_id_array[type], id);
359 	mtx_unlock(&procid_lock);
360 }
361 
362 void
proc_id_clear(int type,pid_t id)363 proc_id_clear(int type, pid_t id)
364 {
365 
366 	KASSERT(type >= 0 && type < nitems(proc_id_array),
367 	    ("invalid type %d\n", type));
368 	mtx_lock(&procid_lock);
369 	KASSERT(bit_test(proc_id_array[type], id) != 0,
370 	    ("bit %d not set in %d\n", id, type));
371 	bit_clear(proc_id_array[type], id);
372 	mtx_unlock(&procid_lock);
373 }
374 
375 /*
376  * Is p an inferior of the current process?
377  */
378 int
inferior(struct proc * p)379 inferior(struct proc *p)
380 {
381 
382 	sx_assert(&proctree_lock, SX_LOCKED);
383 	PROC_LOCK_ASSERT(p, MA_OWNED);
384 	for (; p != curproc; p = proc_realparent(p)) {
385 		if (p->p_pid == 0)
386 			return (0);
387 	}
388 	return (1);
389 }
390 
391 /*
392  * Shared lock all the pid hash lists.
393  */
394 void
pidhash_slockall(void)395 pidhash_slockall(void)
396 {
397 	u_long i;
398 
399 	for (i = 0; i < pidhashlock + 1; i++)
400 		sx_slock(&pidhashtbl_lock[i]);
401 }
402 
403 /*
404  * Shared unlock all the pid hash lists.
405  */
406 void
pidhash_sunlockall(void)407 pidhash_sunlockall(void)
408 {
409 	u_long i;
410 
411 	for (i = 0; i < pidhashlock + 1; i++)
412 		sx_sunlock(&pidhashtbl_lock[i]);
413 }
414 
415 /*
416  * Similar to pfind(), this function locate a process by number.
417  */
418 struct proc *
pfind_any_locked(pid_t pid)419 pfind_any_locked(pid_t pid)
420 {
421 	struct proc *p;
422 
423 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
424 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
425 		if (p->p_pid == pid) {
426 			PROC_LOCK(p);
427 			if (p->p_state == PRS_NEW) {
428 				PROC_UNLOCK(p);
429 				p = NULL;
430 			}
431 			break;
432 		}
433 	}
434 	return (p);
435 }
436 
437 /*
438  * Locate a process by number.
439  *
440  * By not returning processes in the PRS_NEW state, we allow callers to avoid
441  * testing for that condition to avoid dereferencing p_ucred, et al.
442  */
443 static __always_inline struct proc *
_pfind(pid_t pid,bool zombie)444 _pfind(pid_t pid, bool zombie)
445 {
446 	struct proc *p;
447 
448 	p = curproc;
449 	if (p->p_pid == pid) {
450 		PROC_LOCK(p);
451 		return (p);
452 	}
453 	sx_slock(PIDHASHLOCK(pid));
454 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
455 		if (p->p_pid == pid) {
456 			PROC_LOCK(p);
457 			if (p->p_state == PRS_NEW ||
458 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
459 				PROC_UNLOCK(p);
460 				p = NULL;
461 			}
462 			break;
463 		}
464 	}
465 	sx_sunlock(PIDHASHLOCK(pid));
466 	return (p);
467 }
468 
469 struct proc *
pfind(pid_t pid)470 pfind(pid_t pid)
471 {
472 
473 	return (_pfind(pid, false));
474 }
475 
476 /*
477  * Same as pfind but allow zombies.
478  */
479 struct proc *
pfind_any(pid_t pid)480 pfind_any(pid_t pid)
481 {
482 
483 	return (_pfind(pid, true));
484 }
485 
486 /*
487  * Locate a process group by number.
488  * The caller must hold proctree_lock.
489  */
490 struct pgrp *
pgfind(pid_t pgid)491 pgfind(pid_t pgid)
492 {
493 	struct pgrp *pgrp;
494 
495 	sx_assert(&proctree_lock, SX_LOCKED);
496 
497 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
498 		if (pgrp->pg_id == pgid) {
499 			PGRP_LOCK(pgrp);
500 			return (pgrp);
501 		}
502 	}
503 	return (NULL);
504 }
505 
506 /*
507  * Locate process and do additional manipulations, depending on flags.
508  */
509 int
pget(pid_t pid,int flags,struct proc ** pp)510 pget(pid_t pid, int flags, struct proc **pp)
511 {
512 	struct proc *p;
513 	struct thread *td1;
514 	int error;
515 
516 	p = curproc;
517 	if (p->p_pid == pid) {
518 		PROC_LOCK(p);
519 	} else {
520 		p = NULL;
521 		if (pid <= PID_MAX) {
522 			if ((flags & PGET_NOTWEXIT) == 0)
523 				p = pfind_any(pid);
524 			else
525 				p = pfind(pid);
526 		} else if ((flags & PGET_NOTID) == 0) {
527 			td1 = tdfind(pid, -1);
528 			if (td1 != NULL)
529 				p = td1->td_proc;
530 		}
531 		if (p == NULL)
532 			return (ESRCH);
533 		if ((flags & PGET_CANSEE) != 0) {
534 			error = p_cansee(curthread, p);
535 			if (error != 0)
536 				goto errout;
537 		}
538 	}
539 	if ((flags & PGET_CANDEBUG) != 0) {
540 		error = p_candebug(curthread, p);
541 		if (error != 0)
542 			goto errout;
543 	}
544 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
545 		error = EPERM;
546 		goto errout;
547 	}
548 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
549 		error = ESRCH;
550 		goto errout;
551 	}
552 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
553 		/*
554 		 * XXXRW: Not clear ESRCH is the right error during proc
555 		 * execve().
556 		 */
557 		error = ESRCH;
558 		goto errout;
559 	}
560 	if ((flags & PGET_HOLD) != 0) {
561 		_PHOLD(p);
562 		PROC_UNLOCK(p);
563 	}
564 	*pp = p;
565 	return (0);
566 errout:
567 	PROC_UNLOCK(p);
568 	return (error);
569 }
570 
571 /*
572  * Create a new process group.
573  * pgid must be equal to the pid of p.
574  * Begin a new session if required.
575  */
576 int
enterpgrp(struct proc * p,pid_t pgid,struct pgrp * pgrp,struct session * sess)577 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
578 {
579 	struct pgrp *old_pgrp;
580 
581 	sx_assert(&proctree_lock, SX_XLOCKED);
582 
583 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
584 	KASSERT(p->p_pid == pgid,
585 	    ("enterpgrp: new pgrp and pid != pgid"));
586 	KASSERT(pgfind(pgid) == NULL,
587 	    ("enterpgrp: pgrp with pgid exists"));
588 	KASSERT(!SESS_LEADER(p),
589 	    ("enterpgrp: session leader attempted setpgrp"));
590 
591 	old_pgrp = p->p_pgrp;
592 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
593 		sx_xunlock(&proctree_lock);
594 		sx_xlock(&old_pgrp->pg_killsx);
595 		sx_xunlock(&old_pgrp->pg_killsx);
596 		return (ERESTART);
597 	}
598 	MPASS(old_pgrp == p->p_pgrp);
599 
600 	if (sess != NULL) {
601 		/*
602 		 * new session
603 		 */
604 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
605 		PROC_LOCK(p);
606 		p->p_flag &= ~P_CONTROLT;
607 		PROC_UNLOCK(p);
608 		PGRP_LOCK(pgrp);
609 		sess->s_leader = p;
610 		sess->s_sid = p->p_pid;
611 		proc_id_set(PROC_ID_SESSION, p->p_pid);
612 		refcount_init(&sess->s_count, 1);
613 		sess->s_ttyvp = NULL;
614 		sess->s_ttydp = NULL;
615 		sess->s_ttyp = NULL;
616 		bcopy(p->p_session->s_login, sess->s_login,
617 			    sizeof(sess->s_login));
618 		pgrp->pg_session = sess;
619 		KASSERT(p == curproc,
620 		    ("enterpgrp: mksession and p != curproc"));
621 	} else {
622 		pgrp->pg_session = p->p_session;
623 		sess_hold(pgrp->pg_session);
624 		PGRP_LOCK(pgrp);
625 	}
626 	pgrp->pg_id = pgid;
627 	proc_id_set(PROC_ID_GROUP, p->p_pid);
628 	LIST_INIT(&pgrp->pg_members);
629 	pgrp->pg_flags = 0;
630 
631 	/*
632 	 * As we have an exclusive lock of proctree_lock,
633 	 * this should not deadlock.
634 	 */
635 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
636 	SLIST_INIT(&pgrp->pg_sigiolst);
637 	PGRP_UNLOCK(pgrp);
638 
639 	doenterpgrp(p, pgrp);
640 
641 	sx_xunlock(&old_pgrp->pg_killsx);
642 	return (0);
643 }
644 
645 /*
646  * Move p to an existing process group
647  */
648 int
enterthispgrp(struct proc * p,struct pgrp * pgrp)649 enterthispgrp(struct proc *p, struct pgrp *pgrp)
650 {
651 	struct pgrp *old_pgrp;
652 
653 	sx_assert(&proctree_lock, SX_XLOCKED);
654 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
655 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
656 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
657 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
658 	KASSERT(pgrp->pg_session == p->p_session,
659 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
660 	    __func__, pgrp->pg_session, p->p_session, p));
661 	KASSERT(pgrp != p->p_pgrp,
662 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
663 
664 	old_pgrp = p->p_pgrp;
665 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
666 		sx_xunlock(&proctree_lock);
667 		sx_xlock(&old_pgrp->pg_killsx);
668 		sx_xunlock(&old_pgrp->pg_killsx);
669 		return (ERESTART);
670 	}
671 	MPASS(old_pgrp == p->p_pgrp);
672 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
673 		sx_xunlock(&old_pgrp->pg_killsx);
674 		sx_xunlock(&proctree_lock);
675 		sx_xlock(&pgrp->pg_killsx);
676 		sx_xunlock(&pgrp->pg_killsx);
677 		return (ERESTART);
678 	}
679 
680 	doenterpgrp(p, pgrp);
681 
682 	sx_xunlock(&pgrp->pg_killsx);
683 	sx_xunlock(&old_pgrp->pg_killsx);
684 	return (0);
685 }
686 
687 /*
688  * If true, any child of q which belongs to group pgrp, qualifies the
689  * process group pgrp as not orphaned.
690  */
691 static bool
isjobproc(struct proc * q,struct pgrp * pgrp)692 isjobproc(struct proc *q, struct pgrp *pgrp)
693 {
694 	sx_assert(&proctree_lock, SX_LOCKED);
695 
696 	return (q->p_pgrp != pgrp &&
697 	    q->p_pgrp->pg_session == pgrp->pg_session);
698 }
699 
700 static struct proc *
jobc_reaper(struct proc * p)701 jobc_reaper(struct proc *p)
702 {
703 	struct proc *pp;
704 
705 	sx_assert(&proctree_lock, SA_LOCKED);
706 
707 	for (pp = p;;) {
708 		pp = pp->p_reaper;
709 		if (pp->p_reaper == pp ||
710 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
711 			return (pp);
712 	}
713 }
714 
715 static struct proc *
jobc_parent(struct proc * p,struct proc * p_exiting)716 jobc_parent(struct proc *p, struct proc *p_exiting)
717 {
718 	struct proc *pp;
719 
720 	sx_assert(&proctree_lock, SA_LOCKED);
721 
722 	pp = proc_realparent(p);
723 	if (pp->p_pptr == NULL || pp == p_exiting ||
724 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
725 		return (pp);
726 	return (jobc_reaper(pp));
727 }
728 
729 int
pgrp_calc_jobc(struct pgrp * pgrp)730 pgrp_calc_jobc(struct pgrp *pgrp)
731 {
732 	struct proc *q;
733 	int cnt;
734 
735 #ifdef INVARIANTS
736 	if (!mtx_owned(&pgrp->pg_mtx))
737 		sx_assert(&proctree_lock, SA_LOCKED);
738 #endif
739 
740 	cnt = 0;
741 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
742 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
743 		    q->p_pptr == NULL)
744 			continue;
745 		if (isjobproc(jobc_parent(q, NULL), pgrp))
746 			cnt++;
747 	}
748 	return (cnt);
749 }
750 
751 /*
752  * Move p to a process group
753  */
754 static void
doenterpgrp(struct proc * p,struct pgrp * pgrp)755 doenterpgrp(struct proc *p, struct pgrp *pgrp)
756 {
757 	struct pgrp *savepgrp;
758 	struct proc *pp;
759 
760 	sx_assert(&proctree_lock, SX_XLOCKED);
761 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
762 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
763 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
764 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
765 
766 	savepgrp = p->p_pgrp;
767 	pp = jobc_parent(p, NULL);
768 
769 	PGRP_LOCK(pgrp);
770 	PGRP_LOCK(savepgrp);
771 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
772 		orphanpg(savepgrp);
773 	PROC_LOCK(p);
774 	LIST_REMOVE(p, p_pglist);
775 	p->p_pgrp = pgrp;
776 	PROC_UNLOCK(p);
777 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
778 	if (isjobproc(pp, pgrp))
779 		pgrp->pg_flags &= ~PGRP_ORPHANED;
780 	PGRP_UNLOCK(savepgrp);
781 	PGRP_UNLOCK(pgrp);
782 	if (LIST_EMPTY(&savepgrp->pg_members))
783 		pgdelete(savepgrp);
784 }
785 
786 /*
787  * remove process from process group
788  */
789 int
leavepgrp(struct proc * p)790 leavepgrp(struct proc *p)
791 {
792 	struct pgrp *savepgrp;
793 
794 	sx_assert(&proctree_lock, SX_XLOCKED);
795 	savepgrp = p->p_pgrp;
796 	PGRP_LOCK(savepgrp);
797 	PROC_LOCK(p);
798 	LIST_REMOVE(p, p_pglist);
799 	p->p_pgrp = NULL;
800 	PROC_UNLOCK(p);
801 	PGRP_UNLOCK(savepgrp);
802 	if (LIST_EMPTY(&savepgrp->pg_members))
803 		pgdelete(savepgrp);
804 	return (0);
805 }
806 
807 /*
808  * delete a process group
809  */
810 static void
pgdelete(struct pgrp * pgrp)811 pgdelete(struct pgrp *pgrp)
812 {
813 	struct session *savesess;
814 	struct tty *tp;
815 
816 	sx_assert(&proctree_lock, SX_XLOCKED);
817 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
818 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
819 
820 	/*
821 	 * Reset any sigio structures pointing to us as a result of
822 	 * F_SETOWN with our pgid.  The proctree lock ensures that
823 	 * new sigio structures will not be added after this point.
824 	 */
825 	funsetownlst(&pgrp->pg_sigiolst);
826 
827 	PGRP_LOCK(pgrp);
828 	tp = pgrp->pg_session->s_ttyp;
829 	LIST_REMOVE(pgrp, pg_hash);
830 	savesess = pgrp->pg_session;
831 	PGRP_UNLOCK(pgrp);
832 
833 	/* Remove the reference to the pgrp before deallocating it. */
834 	if (tp != NULL) {
835 		tty_lock(tp);
836 		tty_rel_pgrp(tp, pgrp);
837 	}
838 
839 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
840 	uma_zfree(pgrp_zone, pgrp);
841 	sess_release(savesess);
842 }
843 
844 
845 static void
fixjobc_kill(struct proc * p)846 fixjobc_kill(struct proc *p)
847 {
848 	struct proc *q;
849 	struct pgrp *pgrp;
850 
851 	sx_assert(&proctree_lock, SX_LOCKED);
852 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
853 	pgrp = p->p_pgrp;
854 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
855 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
856 
857 	/*
858 	 * p no longer affects process group orphanage for children.
859 	 * It is marked by the flag because p is only physically
860 	 * removed from its process group on wait(2).
861 	 */
862 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
863 	p->p_treeflag |= P_TREE_GRPEXITED;
864 
865 	/*
866 	 * Check if exiting p orphans its own group.
867 	 */
868 	pgrp = p->p_pgrp;
869 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
870 		PGRP_LOCK(pgrp);
871 		if (pgrp_calc_jobc(pgrp) == 0)
872 			orphanpg(pgrp);
873 		PGRP_UNLOCK(pgrp);
874 	}
875 
876 	/*
877 	 * Check this process' children to see whether they qualify
878 	 * their process groups after reparenting to reaper.
879 	 */
880 	LIST_FOREACH(q, &p->p_children, p_sibling) {
881 		pgrp = q->p_pgrp;
882 		PGRP_LOCK(pgrp);
883 		if (pgrp_calc_jobc(pgrp) == 0) {
884 			/*
885 			 * We want to handle exactly the children that
886 			 * has p as realparent.  Then, when calculating
887 			 * jobc_parent for children, we should ignore
888 			 * P_TREE_GRPEXITED flag already set on p.
889 			 */
890 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
891 				orphanpg(pgrp);
892 		} else
893 			pgrp->pg_flags &= ~PGRP_ORPHANED;
894 		PGRP_UNLOCK(pgrp);
895 	}
896 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
897 		pgrp = q->p_pgrp;
898 		PGRP_LOCK(pgrp);
899 		if (pgrp_calc_jobc(pgrp) == 0) {
900 			if (isjobproc(p, pgrp))
901 				orphanpg(pgrp);
902 		} else
903 			pgrp->pg_flags &= ~PGRP_ORPHANED;
904 		PGRP_UNLOCK(pgrp);
905 	}
906 }
907 
908 void
killjobc(void)909 killjobc(void)
910 {
911 	struct session *sp;
912 	struct tty *tp;
913 	struct proc *p;
914 	struct vnode *ttyvp;
915 
916 	p = curproc;
917 	MPASS(p->p_flag & P_WEXIT);
918 	sx_assert(&proctree_lock, SX_LOCKED);
919 
920 	if (SESS_LEADER(p)) {
921 		sp = p->p_session;
922 
923 		/*
924 		 * s_ttyp is not zero'd; we use this to indicate that
925 		 * the session once had a controlling terminal. (for
926 		 * logging and informational purposes)
927 		 */
928 		SESS_LOCK(sp);
929 		ttyvp = sp->s_ttyvp;
930 		tp = sp->s_ttyp;
931 		sp->s_ttyvp = NULL;
932 		sp->s_ttydp = NULL;
933 		sp->s_leader = NULL;
934 		SESS_UNLOCK(sp);
935 
936 		/*
937 		 * Signal foreground pgrp and revoke access to
938 		 * controlling terminal if it has not been revoked
939 		 * already.
940 		 *
941 		 * Because the TTY may have been revoked in the mean
942 		 * time and could already have a new session associated
943 		 * with it, make sure we don't send a SIGHUP to a
944 		 * foreground process group that does not belong to this
945 		 * session.
946 		 */
947 
948 		if (tp != NULL) {
949 			tty_lock(tp);
950 			if (tp->t_session == sp)
951 				tty_signal_pgrp(tp, SIGHUP);
952 			tty_unlock(tp);
953 		}
954 
955 		if (ttyvp != NULL) {
956 			sx_xunlock(&proctree_lock);
957 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
958 				VOP_REVOKE(ttyvp, REVOKEALL);
959 				VOP_UNLOCK(ttyvp);
960 			}
961 			devfs_ctty_unref(ttyvp);
962 			sx_xlock(&proctree_lock);
963 		}
964 	}
965 	fixjobc_kill(p);
966 }
967 
968 /*
969  * A process group has become orphaned, mark it as such for signal
970  * delivery code.  If there are any stopped processes in the group,
971  * hang-up all process in that group.
972  */
973 static void
orphanpg(struct pgrp * pg)974 orphanpg(struct pgrp *pg)
975 {
976 	struct proc *p;
977 
978 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
979 
980 	pg->pg_flags |= PGRP_ORPHANED;
981 
982 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
983 		PROC_LOCK(p);
984 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
985 			PROC_UNLOCK(p);
986 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
987 				PROC_LOCK(p);
988 				kern_psignal(p, SIGHUP);
989 				kern_psignal(p, SIGCONT);
990 				PROC_UNLOCK(p);
991 			}
992 			return;
993 		}
994 		PROC_UNLOCK(p);
995 	}
996 }
997 
998 void
sess_hold(struct session * s)999 sess_hold(struct session *s)
1000 {
1001 
1002 	refcount_acquire(&s->s_count);
1003 }
1004 
1005 void
sess_release(struct session * s)1006 sess_release(struct session *s)
1007 {
1008 
1009 	if (refcount_release(&s->s_count)) {
1010 		if (s->s_ttyp != NULL) {
1011 			tty_lock(s->s_ttyp);
1012 			tty_rel_sess(s->s_ttyp, s);
1013 		}
1014 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1015 		mtx_destroy(&s->s_mtx);
1016 		free(s, M_SESSION);
1017 	}
1018 }
1019 
1020 #ifdef DDB
1021 
1022 static void
db_print_pgrp_one(struct pgrp * pgrp,struct proc * p)1023 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1024 {
1025 	db_printf(
1026 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1027 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1028 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1029 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1030 }
1031 
DB_SHOW_COMMAND_FLAGS(pgrpdump,pgrpdump,DB_CMD_MEMSAFE)1032 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1033 {
1034 	struct pgrp *pgrp;
1035 	struct proc *p;
1036 	int i;
1037 
1038 	for (i = 0; i <= pgrphash; i++) {
1039 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1040 			db_printf("indx %d\n", i);
1041 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1042 				db_printf(
1043 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1044 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1045 				    pgrp->pg_session->s_count,
1046 				    LIST_FIRST(&pgrp->pg_members));
1047 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1048 					db_print_pgrp_one(pgrp, p);
1049 			}
1050 		}
1051 	}
1052 }
1053 #endif /* DDB */
1054 
1055 /*
1056  * Calculate the kinfo_proc members which contain process-wide
1057  * informations.
1058  * Must be called with the target process locked.
1059  */
1060 static void
fill_kinfo_aggregate(struct proc * p,struct kinfo_proc * kp)1061 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1062 {
1063 	struct thread *td;
1064 
1065 	PROC_LOCK_ASSERT(p, MA_OWNED);
1066 
1067 	kp->ki_estcpu = 0;
1068 	kp->ki_pctcpu = 0;
1069 	FOREACH_THREAD_IN_PROC(p, td) {
1070 		thread_lock(td);
1071 		kp->ki_pctcpu += sched_pctcpu(td);
1072 		kp->ki_estcpu += sched_estcpu(td);
1073 		thread_unlock(td);
1074 	}
1075 }
1076 
1077 /*
1078  * Fill in any information that is common to all threads in the process.
1079  * Must be called with the target process locked.
1080  */
1081 static void
fill_kinfo_proc_only(struct proc * p,struct kinfo_proc * kp)1082 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1083 {
1084 	struct thread *td0;
1085 	struct ucred *cred;
1086 	struct sigacts *ps;
1087 	struct timeval boottime;
1088 
1089 	PROC_LOCK_ASSERT(p, MA_OWNED);
1090 
1091 	kp->ki_structsize = sizeof(*kp);
1092 	kp->ki_paddr = p;
1093 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1094 	kp->ki_args = p->p_args;
1095 	kp->ki_textvp = p->p_textvp;
1096 #ifdef KTRACE
1097 	kp->ki_tracep = ktr_get_tracevp(p, false);
1098 	kp->ki_traceflag = p->p_traceflag;
1099 #endif
1100 	kp->ki_fd = p->p_fd;
1101 	kp->ki_pd = p->p_pd;
1102 	kp->ki_vmspace = p->p_vmspace;
1103 	kp->ki_flag = p->p_flag;
1104 	kp->ki_flag2 = p->p_flag2;
1105 	cred = p->p_ucred;
1106 	if (cred) {
1107 		kp->ki_uid = cred->cr_uid;
1108 		kp->ki_ruid = cred->cr_ruid;
1109 		kp->ki_svuid = cred->cr_svuid;
1110 		kp->ki_cr_flags = 0;
1111 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1112 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1113 		/* XXX bde doesn't like KI_NGROUPS */
1114 		if (cred->cr_ngroups > KI_NGROUPS) {
1115 			kp->ki_ngroups = KI_NGROUPS;
1116 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1117 		} else
1118 			kp->ki_ngroups = cred->cr_ngroups;
1119 		bcopy(cred->cr_groups, kp->ki_groups,
1120 		    kp->ki_ngroups * sizeof(gid_t));
1121 		kp->ki_rgid = cred->cr_rgid;
1122 		kp->ki_svgid = cred->cr_svgid;
1123 		/* If jailed(cred), emulate the old P_JAILED flag. */
1124 		if (jailed(cred)) {
1125 			kp->ki_flag |= P_JAILED;
1126 			/* If inside the jail, use 0 as a jail ID. */
1127 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1128 				kp->ki_jid = cred->cr_prison->pr_id;
1129 		}
1130 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1131 		    sizeof(kp->ki_loginclass));
1132 	}
1133 	ps = p->p_sigacts;
1134 	if (ps) {
1135 		mtx_lock(&ps->ps_mtx);
1136 		kp->ki_sigignore = ps->ps_sigignore;
1137 		kp->ki_sigcatch = ps->ps_sigcatch;
1138 		mtx_unlock(&ps->ps_mtx);
1139 	}
1140 	if (p->p_state != PRS_NEW &&
1141 	    p->p_state != PRS_ZOMBIE &&
1142 	    p->p_vmspace != NULL) {
1143 		struct vmspace *vm = p->p_vmspace;
1144 
1145 		kp->ki_size = vm->vm_map.size;
1146 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1147 		FOREACH_THREAD_IN_PROC(p, td0)
1148 			kp->ki_rssize += td0->td_kstack_pages;
1149 		kp->ki_swrss = vm->vm_swrss;
1150 		kp->ki_tsize = vm->vm_tsize;
1151 		kp->ki_dsize = vm->vm_dsize;
1152 		kp->ki_ssize = vm->vm_ssize;
1153 	} else if (p->p_state == PRS_ZOMBIE)
1154 		kp->ki_stat = SZOMB;
1155 	kp->ki_sflag = PS_INMEM;
1156 	/* Calculate legacy swtime as seconds since 'swtick'. */
1157 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1158 	kp->ki_pid = p->p_pid;
1159 	kp->ki_nice = p->p_nice;
1160 	kp->ki_fibnum = p->p_fibnum;
1161 	kp->ki_start = p->p_stats->p_start;
1162 	getboottime(&boottime);
1163 	timevaladd(&kp->ki_start, &boottime);
1164 	PROC_STATLOCK(p);
1165 	rufetch(p, &kp->ki_rusage);
1166 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1167 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1168 	PROC_STATUNLOCK(p);
1169 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1170 	/* Some callers want child times in a single value. */
1171 	kp->ki_childtime = kp->ki_childstime;
1172 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1173 
1174 	FOREACH_THREAD_IN_PROC(p, td0)
1175 		kp->ki_cow += td0->td_cow;
1176 
1177 	if (p->p_comm[0] != '\0')
1178 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1179 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1180 	    p->p_sysent->sv_name[0] != '\0')
1181 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1182 	kp->ki_siglist = p->p_siglist;
1183 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1184 	kp->ki_acflag = p->p_acflag;
1185 	kp->ki_lock = p->p_lock;
1186 	if (p->p_pptr) {
1187 		kp->ki_ppid = p->p_oppid;
1188 		if (p->p_flag & P_TRACED)
1189 			kp->ki_tracer = p->p_pptr->p_pid;
1190 	}
1191 }
1192 
1193 /*
1194  * Fill job-related process information.
1195  */
1196 static void
fill_kinfo_proc_pgrp(struct proc * p,struct kinfo_proc * kp)1197 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1198 {
1199 	struct tty *tp;
1200 	struct session *sp;
1201 	struct pgrp *pgrp;
1202 
1203 	sx_assert(&proctree_lock, SA_LOCKED);
1204 	PROC_LOCK_ASSERT(p, MA_OWNED);
1205 
1206 	pgrp = p->p_pgrp;
1207 	if (pgrp == NULL)
1208 		return;
1209 
1210 	kp->ki_pgid = pgrp->pg_id;
1211 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1212 
1213 	sp = pgrp->pg_session;
1214 	tp = NULL;
1215 
1216 	if (sp != NULL) {
1217 		kp->ki_sid = sp->s_sid;
1218 		SESS_LOCK(sp);
1219 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1220 		if (sp->s_ttyvp)
1221 			kp->ki_kiflag |= KI_CTTY;
1222 		if (SESS_LEADER(p))
1223 			kp->ki_kiflag |= KI_SLEADER;
1224 		tp = sp->s_ttyp;
1225 		SESS_UNLOCK(sp);
1226 	}
1227 
1228 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1229 		kp->ki_tdev = tty_udev(tp);
1230 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1231 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1232 		if (tp->t_session)
1233 			kp->ki_tsid = tp->t_session->s_sid;
1234 	} else {
1235 		kp->ki_tdev = NODEV;
1236 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1237 	}
1238 }
1239 
1240 /*
1241  * Fill in information that is thread specific.  Must be called with
1242  * target process locked.  If 'preferthread' is set, overwrite certain
1243  * process-related fields that are maintained for both threads and
1244  * processes.
1245  */
1246 static void
fill_kinfo_thread(struct thread * td,struct kinfo_proc * kp,int preferthread)1247 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1248 {
1249 	struct proc *p;
1250 
1251 	p = td->td_proc;
1252 	kp->ki_tdaddr = td;
1253 	PROC_LOCK_ASSERT(p, MA_OWNED);
1254 
1255 	if (preferthread)
1256 		PROC_STATLOCK(p);
1257 	thread_lock(td);
1258 	if (td->td_wmesg != NULL)
1259 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1260 	else
1261 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1262 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1263 	    sizeof(kp->ki_tdname)) {
1264 		strlcpy(kp->ki_moretdname,
1265 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1266 		    sizeof(kp->ki_moretdname));
1267 	} else {
1268 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1269 	}
1270 	if (TD_ON_LOCK(td)) {
1271 		kp->ki_kiflag |= KI_LOCKBLOCK;
1272 		strlcpy(kp->ki_lockname, td->td_lockname,
1273 		    sizeof(kp->ki_lockname));
1274 	} else {
1275 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1276 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1277 	}
1278 
1279 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1280 		if (TD_ON_RUNQ(td) ||
1281 		    TD_CAN_RUN(td) ||
1282 		    TD_IS_RUNNING(td)) {
1283 			kp->ki_stat = SRUN;
1284 		} else if (P_SHOULDSTOP(p)) {
1285 			kp->ki_stat = SSTOP;
1286 		} else if (TD_IS_SLEEPING(td)) {
1287 			kp->ki_stat = SSLEEP;
1288 		} else if (TD_ON_LOCK(td)) {
1289 			kp->ki_stat = SLOCK;
1290 		} else {
1291 			kp->ki_stat = SWAIT;
1292 		}
1293 	} else if (p->p_state == PRS_ZOMBIE) {
1294 		kp->ki_stat = SZOMB;
1295 	} else {
1296 		kp->ki_stat = SIDL;
1297 	}
1298 
1299 	/* Things in the thread */
1300 	kp->ki_wchan = td->td_wchan;
1301 	kp->ki_pri.pri_level = td->td_priority;
1302 	kp->ki_pri.pri_native = td->td_base_pri;
1303 
1304 	/*
1305 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1306 	 * the maximum u_char CPU value.
1307 	 */
1308 	if (td->td_lastcpu == NOCPU)
1309 		kp->ki_lastcpu_old = NOCPU_OLD;
1310 	else if (td->td_lastcpu > MAXCPU_OLD)
1311 		kp->ki_lastcpu_old = MAXCPU_OLD;
1312 	else
1313 		kp->ki_lastcpu_old = td->td_lastcpu;
1314 
1315 	if (td->td_oncpu == NOCPU)
1316 		kp->ki_oncpu_old = NOCPU_OLD;
1317 	else if (td->td_oncpu > MAXCPU_OLD)
1318 		kp->ki_oncpu_old = MAXCPU_OLD;
1319 	else
1320 		kp->ki_oncpu_old = td->td_oncpu;
1321 
1322 	kp->ki_lastcpu = td->td_lastcpu;
1323 	kp->ki_oncpu = td->td_oncpu;
1324 	kp->ki_tdflags = td->td_flags;
1325 	kp->ki_tid = td->td_tid;
1326 	kp->ki_numthreads = p->p_numthreads;
1327 	kp->ki_pcb = td->td_pcb;
1328 	kp->ki_kstack = (void *)td->td_kstack;
1329 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1330 	kp->ki_pri.pri_class = td->td_pri_class;
1331 	kp->ki_pri.pri_user = td->td_user_pri;
1332 
1333 	if (preferthread) {
1334 		rufetchtd(td, &kp->ki_rusage);
1335 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1336 		kp->ki_pctcpu = sched_pctcpu(td);
1337 		kp->ki_estcpu = sched_estcpu(td);
1338 		kp->ki_cow = td->td_cow;
1339 	}
1340 
1341 	/* We can't get this anymore but ps etc never used it anyway. */
1342 	kp->ki_rqindex = 0;
1343 
1344 	if (preferthread)
1345 		kp->ki_siglist = td->td_siglist;
1346 	kp->ki_sigmask = td->td_sigmask;
1347 	thread_unlock(td);
1348 	if (preferthread)
1349 		PROC_STATUNLOCK(p);
1350 }
1351 
1352 /*
1353  * Fill in a kinfo_proc structure for the specified process.
1354  * Must be called with the target process locked.
1355  */
1356 void
fill_kinfo_proc(struct proc * p,struct kinfo_proc * kp)1357 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1358 {
1359 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1360 
1361 	bzero(kp, sizeof(*kp));
1362 
1363 	fill_kinfo_proc_pgrp(p,kp);
1364 	fill_kinfo_proc_only(p, kp);
1365 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1366 	fill_kinfo_aggregate(p, kp);
1367 }
1368 
1369 struct pstats *
pstats_alloc(void)1370 pstats_alloc(void)
1371 {
1372 
1373 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1374 }
1375 
1376 /*
1377  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1378  */
1379 void
pstats_fork(struct pstats * src,struct pstats * dst)1380 pstats_fork(struct pstats *src, struct pstats *dst)
1381 {
1382 
1383 	bzero(&dst->pstat_startzero,
1384 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1385 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1386 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1387 }
1388 
1389 void
pstats_free(struct pstats * ps)1390 pstats_free(struct pstats *ps)
1391 {
1392 
1393 	free(ps, M_SUBPROC);
1394 }
1395 
1396 #ifdef COMPAT_FREEBSD32
1397 
1398 /*
1399  * This function is typically used to copy out the kernel address, so
1400  * it can be replaced by assignment of zero.
1401  */
1402 static inline uint32_t
ptr32_trim(const void * ptr)1403 ptr32_trim(const void *ptr)
1404 {
1405 	uintptr_t uptr;
1406 
1407 	uptr = (uintptr_t)ptr;
1408 	return ((uptr > UINT_MAX) ? 0 : uptr);
1409 }
1410 
1411 #define PTRTRIM_CP(src,dst,fld) \
1412 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1413 
1414 static void
freebsd32_kinfo_proc_out(const struct kinfo_proc * ki,struct kinfo_proc32 * ki32)1415 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1416 {
1417 	int i;
1418 
1419 	bzero(ki32, sizeof(struct kinfo_proc32));
1420 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1421 	CP(*ki, *ki32, ki_layout);
1422 	PTRTRIM_CP(*ki, *ki32, ki_args);
1423 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1424 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1425 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1426 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1427 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1428 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1429 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1430 	CP(*ki, *ki32, ki_pid);
1431 	CP(*ki, *ki32, ki_ppid);
1432 	CP(*ki, *ki32, ki_pgid);
1433 	CP(*ki, *ki32, ki_tpgid);
1434 	CP(*ki, *ki32, ki_sid);
1435 	CP(*ki, *ki32, ki_tsid);
1436 	CP(*ki, *ki32, ki_jobc);
1437 	CP(*ki, *ki32, ki_tdev);
1438 	CP(*ki, *ki32, ki_tdev_freebsd11);
1439 	CP(*ki, *ki32, ki_siglist);
1440 	CP(*ki, *ki32, ki_sigmask);
1441 	CP(*ki, *ki32, ki_sigignore);
1442 	CP(*ki, *ki32, ki_sigcatch);
1443 	CP(*ki, *ki32, ki_uid);
1444 	CP(*ki, *ki32, ki_ruid);
1445 	CP(*ki, *ki32, ki_svuid);
1446 	CP(*ki, *ki32, ki_rgid);
1447 	CP(*ki, *ki32, ki_svgid);
1448 	CP(*ki, *ki32, ki_ngroups);
1449 	for (i = 0; i < KI_NGROUPS; i++)
1450 		CP(*ki, *ki32, ki_groups[i]);
1451 	CP(*ki, *ki32, ki_size);
1452 	CP(*ki, *ki32, ki_rssize);
1453 	CP(*ki, *ki32, ki_swrss);
1454 	CP(*ki, *ki32, ki_tsize);
1455 	CP(*ki, *ki32, ki_dsize);
1456 	CP(*ki, *ki32, ki_ssize);
1457 	CP(*ki, *ki32, ki_xstat);
1458 	CP(*ki, *ki32, ki_acflag);
1459 	CP(*ki, *ki32, ki_pctcpu);
1460 	CP(*ki, *ki32, ki_estcpu);
1461 	CP(*ki, *ki32, ki_slptime);
1462 	CP(*ki, *ki32, ki_swtime);
1463 	CP(*ki, *ki32, ki_cow);
1464 	CP(*ki, *ki32, ki_runtime);
1465 	TV_CP(*ki, *ki32, ki_start);
1466 	TV_CP(*ki, *ki32, ki_childtime);
1467 	CP(*ki, *ki32, ki_flag);
1468 	CP(*ki, *ki32, ki_kiflag);
1469 	CP(*ki, *ki32, ki_traceflag);
1470 	CP(*ki, *ki32, ki_stat);
1471 	CP(*ki, *ki32, ki_nice);
1472 	CP(*ki, *ki32, ki_lock);
1473 	CP(*ki, *ki32, ki_rqindex);
1474 	CP(*ki, *ki32, ki_oncpu);
1475 	CP(*ki, *ki32, ki_lastcpu);
1476 
1477 	/* XXX TODO: wrap cpu value as appropriate */
1478 	CP(*ki, *ki32, ki_oncpu_old);
1479 	CP(*ki, *ki32, ki_lastcpu_old);
1480 
1481 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1482 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1483 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1484 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1485 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1486 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1487 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1488 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1489 	CP(*ki, *ki32, ki_tracer);
1490 	CP(*ki, *ki32, ki_flag2);
1491 	CP(*ki, *ki32, ki_fibnum);
1492 	CP(*ki, *ki32, ki_cr_flags);
1493 	CP(*ki, *ki32, ki_jid);
1494 	CP(*ki, *ki32, ki_numthreads);
1495 	CP(*ki, *ki32, ki_tid);
1496 	CP(*ki, *ki32, ki_pri);
1497 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1498 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1499 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1500 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1501 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1502 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1503 	CP(*ki, *ki32, ki_sflag);
1504 	CP(*ki, *ki32, ki_tdflags);
1505 }
1506 #endif
1507 
1508 static ssize_t
kern_proc_out_size(struct proc * p,int flags)1509 kern_proc_out_size(struct proc *p, int flags)
1510 {
1511 	ssize_t size = 0;
1512 
1513 	PROC_LOCK_ASSERT(p, MA_OWNED);
1514 
1515 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1516 #ifdef COMPAT_FREEBSD32
1517 		if ((flags & KERN_PROC_MASK32) != 0) {
1518 			size += sizeof(struct kinfo_proc32);
1519 		} else
1520 #endif
1521 			size += sizeof(struct kinfo_proc);
1522 	} else {
1523 #ifdef COMPAT_FREEBSD32
1524 		if ((flags & KERN_PROC_MASK32) != 0)
1525 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1526 		else
1527 #endif
1528 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1529 	}
1530 	PROC_UNLOCK(p);
1531 	return (size);
1532 }
1533 
1534 int
kern_proc_out(struct proc * p,struct sbuf * sb,int flags)1535 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1536 {
1537 	struct thread *td;
1538 	struct kinfo_proc ki;
1539 #ifdef COMPAT_FREEBSD32
1540 	struct kinfo_proc32 ki32;
1541 #endif
1542 	int error;
1543 
1544 	PROC_LOCK_ASSERT(p, MA_OWNED);
1545 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1546 
1547 	error = 0;
1548 	fill_kinfo_proc(p, &ki);
1549 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1550 #ifdef COMPAT_FREEBSD32
1551 		if ((flags & KERN_PROC_MASK32) != 0) {
1552 			freebsd32_kinfo_proc_out(&ki, &ki32);
1553 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1554 				error = ENOMEM;
1555 		} else
1556 #endif
1557 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1558 				error = ENOMEM;
1559 	} else {
1560 		FOREACH_THREAD_IN_PROC(p, td) {
1561 			fill_kinfo_thread(td, &ki, 1);
1562 #ifdef COMPAT_FREEBSD32
1563 			if ((flags & KERN_PROC_MASK32) != 0) {
1564 				freebsd32_kinfo_proc_out(&ki, &ki32);
1565 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1566 					error = ENOMEM;
1567 			} else
1568 #endif
1569 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1570 					error = ENOMEM;
1571 			if (error != 0)
1572 				break;
1573 		}
1574 	}
1575 	PROC_UNLOCK(p);
1576 	return (error);
1577 }
1578 
1579 static int
sysctl_out_proc(struct proc * p,struct sysctl_req * req,int flags)1580 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1581 {
1582 	struct sbuf sb;
1583 	struct kinfo_proc ki;
1584 	int error, error2;
1585 
1586 	if (req->oldptr == NULL)
1587 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1588 
1589 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1590 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1591 	error = kern_proc_out(p, &sb, flags);
1592 	error2 = sbuf_finish(&sb);
1593 	sbuf_delete(&sb);
1594 	if (error != 0)
1595 		return (error);
1596 	else if (error2 != 0)
1597 		return (error2);
1598 	return (0);
1599 }
1600 
1601 int
proc_iterate(int (* cb)(struct proc *,void *),void * cbarg)1602 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1603 {
1604 	struct proc *p;
1605 	int error, i, j;
1606 
1607 	for (i = 0; i < pidhashlock + 1; i++) {
1608 		sx_slock(&proctree_lock);
1609 		sx_slock(&pidhashtbl_lock[i]);
1610 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1611 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1612 				if (p->p_state == PRS_NEW)
1613 					continue;
1614 				error = cb(p, cbarg);
1615 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1616 				if (error != 0) {
1617 					sx_sunlock(&pidhashtbl_lock[i]);
1618 					sx_sunlock(&proctree_lock);
1619 					return (error);
1620 				}
1621 			}
1622 		}
1623 		sx_sunlock(&pidhashtbl_lock[i]);
1624 		sx_sunlock(&proctree_lock);
1625 	}
1626 	return (0);
1627 }
1628 
1629 struct kern_proc_out_args {
1630 	struct sysctl_req *req;
1631 	int flags;
1632 	int oid_number;
1633 	int *name;
1634 };
1635 
1636 static int
sysctl_kern_proc_iterate(struct proc * p,void * origarg)1637 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1638 {
1639 	struct kern_proc_out_args *arg = origarg;
1640 	int *name = arg->name;
1641 	int oid_number = arg->oid_number;
1642 	int flags = arg->flags;
1643 	struct sysctl_req *req = arg->req;
1644 	int error = 0;
1645 
1646 	PROC_LOCK(p);
1647 
1648 	KASSERT(p->p_ucred != NULL,
1649 	    ("process credential is NULL for non-NEW proc"));
1650 	/*
1651 	 * Show a user only appropriate processes.
1652 	 */
1653 	if (p_cansee(curthread, p))
1654 		goto skip;
1655 	/*
1656 	 * TODO - make more efficient (see notes below).
1657 	 * do by session.
1658 	 */
1659 	switch (oid_number) {
1660 	case KERN_PROC_GID:
1661 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1662 			goto skip;
1663 		break;
1664 
1665 	case KERN_PROC_PGRP:
1666 		/* could do this by traversing pgrp */
1667 		if (p->p_pgrp == NULL ||
1668 		    p->p_pgrp->pg_id != (pid_t)name[0])
1669 			goto skip;
1670 		break;
1671 
1672 	case KERN_PROC_RGID:
1673 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1674 			goto skip;
1675 		break;
1676 
1677 	case KERN_PROC_SESSION:
1678 		if (p->p_session == NULL ||
1679 		    p->p_session->s_sid != (pid_t)name[0])
1680 			goto skip;
1681 		break;
1682 
1683 	case KERN_PROC_TTY:
1684 		if ((p->p_flag & P_CONTROLT) == 0 ||
1685 		    p->p_session == NULL)
1686 			goto skip;
1687 		/* XXX proctree_lock */
1688 		SESS_LOCK(p->p_session);
1689 		if (p->p_session->s_ttyp == NULL ||
1690 		    tty_udev(p->p_session->s_ttyp) !=
1691 		    (dev_t)name[0]) {
1692 			SESS_UNLOCK(p->p_session);
1693 			goto skip;
1694 		}
1695 		SESS_UNLOCK(p->p_session);
1696 		break;
1697 
1698 	case KERN_PROC_UID:
1699 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1700 			goto skip;
1701 		break;
1702 
1703 	case KERN_PROC_RUID:
1704 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1705 			goto skip;
1706 		break;
1707 
1708 	case KERN_PROC_PROC:
1709 		break;
1710 
1711 	default:
1712 		break;
1713 	}
1714 	error = sysctl_out_proc(p, req, flags);
1715 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1716 	return (error);
1717 skip:
1718 	PROC_UNLOCK(p);
1719 	return (0);
1720 }
1721 
1722 static int
sysctl_kern_proc(SYSCTL_HANDLER_ARGS)1723 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1724 {
1725 	struct kern_proc_out_args iterarg;
1726 	int *name = (int *)arg1;
1727 	u_int namelen = arg2;
1728 	struct proc *p;
1729 	int flags, oid_number;
1730 	int error = 0;
1731 
1732 	oid_number = oidp->oid_number;
1733 	if (oid_number != KERN_PROC_ALL &&
1734 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1735 		flags = KERN_PROC_NOTHREADS;
1736 	else {
1737 		flags = 0;
1738 		oid_number &= ~KERN_PROC_INC_THREAD;
1739 	}
1740 #ifdef COMPAT_FREEBSD32
1741 	if (req->flags & SCTL_MASK32)
1742 		flags |= KERN_PROC_MASK32;
1743 #endif
1744 	if (oid_number == KERN_PROC_PID) {
1745 		if (namelen != 1)
1746 			return (EINVAL);
1747 		error = sysctl_wire_old_buffer(req, 0);
1748 		if (error)
1749 			return (error);
1750 		sx_slock(&proctree_lock);
1751 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1752 		if (error == 0)
1753 			error = sysctl_out_proc(p, req, flags);
1754 		sx_sunlock(&proctree_lock);
1755 		return (error);
1756 	}
1757 
1758 	switch (oid_number) {
1759 	case KERN_PROC_ALL:
1760 		if (namelen != 0)
1761 			return (EINVAL);
1762 		break;
1763 	case KERN_PROC_PROC:
1764 		if (namelen != 0 && namelen != 1)
1765 			return (EINVAL);
1766 		break;
1767 	default:
1768 		if (namelen != 1)
1769 			return (EINVAL);
1770 		break;
1771 	}
1772 
1773 	if (req->oldptr == NULL) {
1774 		/* overestimate by 5 procs */
1775 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1776 		if (error)
1777 			return (error);
1778 	} else {
1779 		error = sysctl_wire_old_buffer(req, 0);
1780 		if (error != 0)
1781 			return (error);
1782 	}
1783 	iterarg.flags = flags;
1784 	iterarg.oid_number = oid_number;
1785 	iterarg.req = req;
1786 	iterarg.name = name;
1787 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1788 	return (error);
1789 }
1790 
1791 struct pargs *
pargs_alloc(int len)1792 pargs_alloc(int len)
1793 {
1794 	struct pargs *pa;
1795 
1796 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1797 		M_WAITOK);
1798 	refcount_init(&pa->ar_ref, 1);
1799 	pa->ar_length = len;
1800 	return (pa);
1801 }
1802 
1803 static void
pargs_free(struct pargs * pa)1804 pargs_free(struct pargs *pa)
1805 {
1806 
1807 	free(pa, M_PARGS);
1808 }
1809 
1810 void
pargs_hold(struct pargs * pa)1811 pargs_hold(struct pargs *pa)
1812 {
1813 
1814 	if (pa == NULL)
1815 		return;
1816 	refcount_acquire(&pa->ar_ref);
1817 }
1818 
1819 void
pargs_drop(struct pargs * pa)1820 pargs_drop(struct pargs *pa)
1821 {
1822 
1823 	if (pa == NULL)
1824 		return;
1825 	if (refcount_release(&pa->ar_ref))
1826 		pargs_free(pa);
1827 }
1828 
1829 static int
proc_read_string(struct thread * td,struct proc * p,const char * sptr,char * buf,size_t len)1830 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1831     size_t len)
1832 {
1833 	ssize_t n;
1834 
1835 	/*
1836 	 * This may return a short read if the string is shorter than the chunk
1837 	 * and is aligned at the end of the page, and the following page is not
1838 	 * mapped.
1839 	 */
1840 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1841 	if (n <= 0)
1842 		return (ENOMEM);
1843 	return (0);
1844 }
1845 
1846 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1847 
1848 enum proc_vector_type {
1849 	PROC_ARG,
1850 	PROC_ENV,
1851 	PROC_AUX,
1852 };
1853 
1854 #ifdef COMPAT_FREEBSD32
1855 static int
get_proc_vector32(struct thread * td,struct proc * p,char *** proc_vectorp,size_t * vsizep,enum proc_vector_type type)1856 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1857     size_t *vsizep, enum proc_vector_type type)
1858 {
1859 	struct freebsd32_ps_strings pss;
1860 	Elf32_Auxinfo aux;
1861 	vm_offset_t vptr, ptr;
1862 	uint32_t *proc_vector32;
1863 	char **proc_vector;
1864 	size_t vsize, size;
1865 	int i, error;
1866 
1867 	error = 0;
1868 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1869 	    sizeof(pss))
1870 		return (ENOMEM);
1871 	switch (type) {
1872 	case PROC_ARG:
1873 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1874 		vsize = pss.ps_nargvstr;
1875 		if (vsize > ARG_MAX)
1876 			return (ENOEXEC);
1877 		size = vsize * sizeof(int32_t);
1878 		break;
1879 	case PROC_ENV:
1880 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1881 		vsize = pss.ps_nenvstr;
1882 		if (vsize > ARG_MAX)
1883 			return (ENOEXEC);
1884 		size = vsize * sizeof(int32_t);
1885 		break;
1886 	case PROC_AUX:
1887 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1888 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1889 		if (vptr % 4 != 0)
1890 			return (ENOEXEC);
1891 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1892 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1893 			    sizeof(aux))
1894 				return (ENOMEM);
1895 			if (aux.a_type == AT_NULL)
1896 				break;
1897 			ptr += sizeof(aux);
1898 		}
1899 		if (aux.a_type != AT_NULL)
1900 			return (ENOEXEC);
1901 		vsize = i + 1;
1902 		size = vsize * sizeof(aux);
1903 		break;
1904 	default:
1905 		KASSERT(0, ("Wrong proc vector type: %d", type));
1906 		return (EINVAL);
1907 	}
1908 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1909 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1910 		error = ENOMEM;
1911 		goto done;
1912 	}
1913 	if (type == PROC_AUX) {
1914 		*proc_vectorp = (char **)proc_vector32;
1915 		*vsizep = vsize;
1916 		return (0);
1917 	}
1918 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1919 	for (i = 0; i < (int)vsize; i++)
1920 		proc_vector[i] = PTRIN(proc_vector32[i]);
1921 	*proc_vectorp = proc_vector;
1922 	*vsizep = vsize;
1923 done:
1924 	free(proc_vector32, M_TEMP);
1925 	return (error);
1926 }
1927 #endif
1928 
1929 static int
get_proc_vector(struct thread * td,struct proc * p,char *** proc_vectorp,size_t * vsizep,enum proc_vector_type type)1930 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1931     size_t *vsizep, enum proc_vector_type type)
1932 {
1933 	struct ps_strings pss;
1934 	Elf_Auxinfo aux;
1935 	vm_offset_t vptr, ptr;
1936 	char **proc_vector;
1937 	size_t vsize, size;
1938 	int i;
1939 
1940 #ifdef COMPAT_FREEBSD32
1941 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1942 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1943 #endif
1944 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1945 	    sizeof(pss))
1946 		return (ENOMEM);
1947 	switch (type) {
1948 	case PROC_ARG:
1949 		vptr = (vm_offset_t)pss.ps_argvstr;
1950 		vsize = pss.ps_nargvstr;
1951 		if (vsize > ARG_MAX)
1952 			return (ENOEXEC);
1953 		size = vsize * sizeof(char *);
1954 		break;
1955 	case PROC_ENV:
1956 		vptr = (vm_offset_t)pss.ps_envstr;
1957 		vsize = pss.ps_nenvstr;
1958 		if (vsize > ARG_MAX)
1959 			return (ENOEXEC);
1960 		size = vsize * sizeof(char *);
1961 		break;
1962 	case PROC_AUX:
1963 		/*
1964 		 * The aux array is just above env array on the stack. Check
1965 		 * that the address is naturally aligned.
1966 		 */
1967 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1968 		    * sizeof(char *);
1969 #if __ELF_WORD_SIZE == 64
1970 		if (vptr % sizeof(uint64_t) != 0)
1971 #else
1972 		if (vptr % sizeof(uint32_t) != 0)
1973 #endif
1974 			return (ENOEXEC);
1975 		/*
1976 		 * We count the array size reading the aux vectors from the
1977 		 * stack until AT_NULL vector is returned.  So (to keep the code
1978 		 * simple) we read the process stack twice: the first time here
1979 		 * to find the size and the second time when copying the vectors
1980 		 * to the allocated proc_vector.
1981 		 */
1982 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1983 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1984 			    sizeof(aux))
1985 				return (ENOMEM);
1986 			if (aux.a_type == AT_NULL)
1987 				break;
1988 			ptr += sizeof(aux);
1989 		}
1990 		/*
1991 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1992 		 * not reached AT_NULL, it is most likely we are reading wrong
1993 		 * data: either the process doesn't have auxv array or data has
1994 		 * been modified. Return the error in this case.
1995 		 */
1996 		if (aux.a_type != AT_NULL)
1997 			return (ENOEXEC);
1998 		vsize = i + 1;
1999 		size = vsize * sizeof(aux);
2000 		break;
2001 	default:
2002 		KASSERT(0, ("Wrong proc vector type: %d", type));
2003 		return (EINVAL); /* In case we are built without INVARIANTS. */
2004 	}
2005 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2006 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2007 		free(proc_vector, M_TEMP);
2008 		return (ENOMEM);
2009 	}
2010 	*proc_vectorp = proc_vector;
2011 	*vsizep = vsize;
2012 
2013 	return (0);
2014 }
2015 
2016 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2017 
2018 static int
get_ps_strings(struct thread * td,struct proc * p,struct sbuf * sb,enum proc_vector_type type)2019 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2020     enum proc_vector_type type)
2021 {
2022 	size_t done, len, nchr, vsize;
2023 	int error, i;
2024 	char **proc_vector, *sptr;
2025 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2026 
2027 	PROC_ASSERT_HELD(p);
2028 
2029 	/*
2030 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2031 	 */
2032 	nchr = 2 * (PATH_MAX + ARG_MAX);
2033 
2034 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2035 	if (error != 0)
2036 		return (error);
2037 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2038 		/*
2039 		 * The program may have scribbled into its argv array, e.g. to
2040 		 * remove some arguments.  If that has happened, break out
2041 		 * before trying to read from NULL.
2042 		 */
2043 		if (proc_vector[i] == NULL)
2044 			break;
2045 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2046 			error = proc_read_string(td, p, sptr, pss_string,
2047 			    sizeof(pss_string));
2048 			if (error != 0)
2049 				goto done;
2050 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2051 			if (done + len >= nchr)
2052 				len = nchr - done - 1;
2053 			sbuf_bcat(sb, pss_string, len);
2054 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2055 				break;
2056 			done += GET_PS_STRINGS_CHUNK_SZ;
2057 		}
2058 		sbuf_bcat(sb, "", 1);
2059 		done += len + 1;
2060 	}
2061 done:
2062 	free(proc_vector, M_TEMP);
2063 	return (error);
2064 }
2065 
2066 int
proc_getargv(struct thread * td,struct proc * p,struct sbuf * sb)2067 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2068 {
2069 
2070 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2071 }
2072 
2073 int
proc_getenvv(struct thread * td,struct proc * p,struct sbuf * sb)2074 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2075 {
2076 
2077 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2078 }
2079 
2080 int
proc_getauxv(struct thread * td,struct proc * p,struct sbuf * sb)2081 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2082 {
2083 	size_t vsize, size;
2084 	char **auxv;
2085 	int error;
2086 
2087 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2088 	if (error == 0) {
2089 #ifdef COMPAT_FREEBSD32
2090 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2091 			size = vsize * sizeof(Elf32_Auxinfo);
2092 		else
2093 #endif
2094 			size = vsize * sizeof(Elf_Auxinfo);
2095 		if (sbuf_bcat(sb, auxv, size) != 0)
2096 			error = ENOMEM;
2097 		free(auxv, M_TEMP);
2098 	}
2099 	return (error);
2100 }
2101 
2102 /*
2103  * This sysctl allows a process to retrieve the argument list or process
2104  * title for another process without groping around in the address space
2105  * of the other process.  It also allow a process to set its own "process
2106  * title to a string of its own choice.
2107  */
2108 static int
sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)2109 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2110 {
2111 	int *name = (int *)arg1;
2112 	u_int namelen = arg2;
2113 	struct pargs *newpa, *pa;
2114 	struct proc *p;
2115 	struct sbuf sb;
2116 	int flags, error = 0, error2;
2117 	pid_t pid;
2118 
2119 	if (namelen != 1)
2120 		return (EINVAL);
2121 
2122 	p = curproc;
2123 	pid = (pid_t)name[0];
2124 	if (pid == -1) {
2125 		pid = p->p_pid;
2126 	}
2127 
2128 	/*
2129 	 * If the query is for this process and it is single-threaded, there
2130 	 * is nobody to modify pargs, thus we can just read.
2131 	 */
2132 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2133 	    (pa = p->p_args) != NULL)
2134 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2135 
2136 	flags = PGET_CANSEE;
2137 	if (req->newptr != NULL)
2138 		flags |= PGET_ISCURRENT;
2139 	error = pget(pid, flags, &p);
2140 	if (error)
2141 		return (error);
2142 
2143 	pa = p->p_args;
2144 	if (pa != NULL) {
2145 		pargs_hold(pa);
2146 		PROC_UNLOCK(p);
2147 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2148 		pargs_drop(pa);
2149 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2150 		_PHOLD(p);
2151 		PROC_UNLOCK(p);
2152 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2153 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2154 		error = proc_getargv(curthread, p, &sb);
2155 		error2 = sbuf_finish(&sb);
2156 		PRELE(p);
2157 		sbuf_delete(&sb);
2158 		if (error == 0 && error2 != 0)
2159 			error = error2;
2160 	} else {
2161 		PROC_UNLOCK(p);
2162 	}
2163 	if (error != 0 || req->newptr == NULL)
2164 		return (error);
2165 
2166 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2167 		return (ENOMEM);
2168 
2169 	if (req->newlen == 0) {
2170 		/*
2171 		 * Clear the argument pointer, so that we'll fetch arguments
2172 		 * with proc_getargv() until further notice.
2173 		 */
2174 		newpa = NULL;
2175 	} else {
2176 		newpa = pargs_alloc(req->newlen);
2177 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2178 		if (error != 0) {
2179 			pargs_free(newpa);
2180 			return (error);
2181 		}
2182 	}
2183 	PROC_LOCK(p);
2184 	pa = p->p_args;
2185 	p->p_args = newpa;
2186 	PROC_UNLOCK(p);
2187 	pargs_drop(pa);
2188 	return (0);
2189 }
2190 
2191 /*
2192  * This sysctl allows a process to retrieve environment of another process.
2193  */
2194 static int
sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)2195 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2196 {
2197 	int *name = (int *)arg1;
2198 	u_int namelen = arg2;
2199 	struct proc *p;
2200 	struct sbuf sb;
2201 	int error, error2;
2202 
2203 	if (namelen != 1)
2204 		return (EINVAL);
2205 
2206 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2207 	if (error != 0)
2208 		return (error);
2209 	if ((p->p_flag & P_SYSTEM) != 0) {
2210 		PRELE(p);
2211 		return (0);
2212 	}
2213 
2214 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2215 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2216 	error = proc_getenvv(curthread, p, &sb);
2217 	error2 = sbuf_finish(&sb);
2218 	PRELE(p);
2219 	sbuf_delete(&sb);
2220 	return (error != 0 ? error : error2);
2221 }
2222 
2223 /*
2224  * This sysctl allows a process to retrieve ELF auxiliary vector of
2225  * another process.
2226  */
2227 static int
sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)2228 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2229 {
2230 	int *name = (int *)arg1;
2231 	u_int namelen = arg2;
2232 	struct proc *p;
2233 	struct sbuf sb;
2234 	int error, error2;
2235 
2236 	if (namelen != 1)
2237 		return (EINVAL);
2238 
2239 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2240 	if (error != 0)
2241 		return (error);
2242 	if ((p->p_flag & P_SYSTEM) != 0) {
2243 		PRELE(p);
2244 		return (0);
2245 	}
2246 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2247 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2248 	error = proc_getauxv(curthread, p, &sb);
2249 	error2 = sbuf_finish(&sb);
2250 	PRELE(p);
2251 	sbuf_delete(&sb);
2252 	return (error != 0 ? error : error2);
2253 }
2254 
2255 /*
2256  * Look up the canonical executable path running in the specified process.
2257  * It tries to return the same hardlink name as was used for execve(2).
2258  * This allows the programs that modify their behavior based on their progname,
2259  * to operate correctly.
2260  *
2261  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2262  *   calling conventions.
2263  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2264  *   allocated and freed by caller.
2265  * freebuf should be freed by caller, from the M_TEMP malloc type.
2266  */
2267 int
proc_get_binpath(struct proc * p,char * binname,char ** retbuf,char ** freebuf)2268 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2269     char **freebuf)
2270 {
2271 	struct nameidata nd;
2272 	struct vnode *vp, *dvp;
2273 	size_t freepath_size;
2274 	int error;
2275 	bool do_fullpath;
2276 
2277 	PROC_LOCK_ASSERT(p, MA_OWNED);
2278 
2279 	vp = p->p_textvp;
2280 	if (vp == NULL) {
2281 		PROC_UNLOCK(p);
2282 		*retbuf = "";
2283 		*freebuf = NULL;
2284 		return (0);
2285 	}
2286 	vref(vp);
2287 	dvp = p->p_textdvp;
2288 	if (dvp != NULL)
2289 		vref(dvp);
2290 	if (p->p_binname != NULL)
2291 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2292 	PROC_UNLOCK(p);
2293 
2294 	do_fullpath = true;
2295 	*freebuf = NULL;
2296 	if (dvp != NULL && binname[0] != '\0') {
2297 		freepath_size = MAXPATHLEN;
2298 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2299 		    retbuf, freebuf, &freepath_size) == 0) {
2300 			/*
2301 			 * Recheck the looked up path.  The binary
2302 			 * might have been renamed or replaced, in
2303 			 * which case we should not report old name.
2304 			 */
2305 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2306 			error = namei(&nd);
2307 			if (error == 0) {
2308 				if (nd.ni_vp == vp)
2309 					do_fullpath = false;
2310 				vrele(nd.ni_vp);
2311 				NDFREE_PNBUF(&nd);
2312 			}
2313 		}
2314 	}
2315 	if (do_fullpath) {
2316 		free(*freebuf, M_TEMP);
2317 		*freebuf = NULL;
2318 		error = vn_fullpath(vp, retbuf, freebuf);
2319 	}
2320 	vrele(vp);
2321 	if (dvp != NULL)
2322 		vrele(dvp);
2323 	return (error);
2324 }
2325 
2326 /*
2327  * This sysctl allows a process to retrieve the path of the executable for
2328  * itself or another process.
2329  */
2330 static int
sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)2331 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2332 {
2333 	pid_t *pidp = (pid_t *)arg1;
2334 	unsigned int arglen = arg2;
2335 	struct proc *p;
2336 	char *retbuf, *freebuf, *binname;
2337 	int error;
2338 
2339 	if (arglen != 1)
2340 		return (EINVAL);
2341 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2342 	binname[0] = '\0';
2343 	if (*pidp == -1) {	/* -1 means this process */
2344 		error = 0;
2345 		p = req->td->td_proc;
2346 		PROC_LOCK(p);
2347 	} else {
2348 		error = pget(*pidp, PGET_CANSEE, &p);
2349 	}
2350 
2351 	if (error == 0)
2352 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2353 	free(binname, M_TEMP);
2354 	if (error != 0)
2355 		return (error);
2356 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2357 	free(freebuf, M_TEMP);
2358 	return (error);
2359 }
2360 
2361 static int
sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)2362 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2363 {
2364 	struct proc *p;
2365 	char *sv_name;
2366 	int *name;
2367 	int namelen;
2368 	int error;
2369 
2370 	namelen = arg2;
2371 	if (namelen != 1)
2372 		return (EINVAL);
2373 
2374 	name = (int *)arg1;
2375 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2376 	if (error != 0)
2377 		return (error);
2378 	sv_name = p->p_sysent->sv_name;
2379 	PROC_UNLOCK(p);
2380 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2381 }
2382 
2383 #ifdef KINFO_OVMENTRY_SIZE
2384 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2385 #endif
2386 
2387 #ifdef COMPAT_FREEBSD7
2388 static int
sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)2389 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2390 {
2391 	vm_map_entry_t entry, tmp_entry;
2392 	unsigned int last_timestamp, namelen;
2393 	char *fullpath, *freepath;
2394 	struct kinfo_ovmentry *kve;
2395 	struct vattr va;
2396 	struct ucred *cred;
2397 	int error, *name;
2398 	struct vnode *vp;
2399 	struct proc *p;
2400 	vm_map_t map;
2401 	struct vmspace *vm;
2402 
2403 	namelen = arg2;
2404 	if (namelen != 1)
2405 		return (EINVAL);
2406 
2407 	name = (int *)arg1;
2408 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2409 	if (error != 0)
2410 		return (error);
2411 	vm = vmspace_acquire_ref(p);
2412 	if (vm == NULL) {
2413 		PRELE(p);
2414 		return (ESRCH);
2415 	}
2416 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2417 
2418 	map = &vm->vm_map;
2419 	vm_map_lock_read(map);
2420 	VM_MAP_ENTRY_FOREACH(entry, map) {
2421 		vm_object_t obj, tobj, lobj;
2422 		vm_offset_t addr;
2423 
2424 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2425 			continue;
2426 
2427 		bzero(kve, sizeof(*kve));
2428 		kve->kve_structsize = sizeof(*kve);
2429 
2430 		kve->kve_private_resident = 0;
2431 		obj = entry->object.vm_object;
2432 		if (obj != NULL) {
2433 			VM_OBJECT_RLOCK(obj);
2434 			if (obj->shadow_count == 1)
2435 				kve->kve_private_resident =
2436 				    obj->resident_page_count;
2437 		}
2438 		kve->kve_resident = 0;
2439 		addr = entry->start;
2440 		while (addr < entry->end) {
2441 			if (pmap_extract(map->pmap, addr))
2442 				kve->kve_resident++;
2443 			addr += PAGE_SIZE;
2444 		}
2445 
2446 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2447 			if (tobj != obj) {
2448 				VM_OBJECT_RLOCK(tobj);
2449 				kve->kve_offset += tobj->backing_object_offset;
2450 			}
2451 			if (lobj != obj)
2452 				VM_OBJECT_RUNLOCK(lobj);
2453 			lobj = tobj;
2454 		}
2455 
2456 		kve->kve_start = (void*)entry->start;
2457 		kve->kve_end = (void*)entry->end;
2458 		kve->kve_offset += (off_t)entry->offset;
2459 
2460 		if (entry->protection & VM_PROT_READ)
2461 			kve->kve_protection |= KVME_PROT_READ;
2462 		if (entry->protection & VM_PROT_WRITE)
2463 			kve->kve_protection |= KVME_PROT_WRITE;
2464 		if (entry->protection & VM_PROT_EXECUTE)
2465 			kve->kve_protection |= KVME_PROT_EXEC;
2466 
2467 		if (entry->eflags & MAP_ENTRY_COW)
2468 			kve->kve_flags |= KVME_FLAG_COW;
2469 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2470 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2471 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2472 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2473 
2474 		last_timestamp = map->timestamp;
2475 		vm_map_unlock_read(map);
2476 
2477 		kve->kve_fileid = 0;
2478 		kve->kve_fsid = 0;
2479 		freepath = NULL;
2480 		fullpath = "";
2481 		if (lobj) {
2482 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2483 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2484 				kve->kve_type = KVME_TYPE_UNKNOWN;
2485 			if (vp != NULL)
2486 				vref(vp);
2487 			if (lobj != obj)
2488 				VM_OBJECT_RUNLOCK(lobj);
2489 
2490 			kve->kve_ref_count = obj->ref_count;
2491 			kve->kve_shadow_count = obj->shadow_count;
2492 			VM_OBJECT_RUNLOCK(obj);
2493 			if (vp != NULL) {
2494 				vn_fullpath(vp, &fullpath, &freepath);
2495 				cred = curthread->td_ucred;
2496 				vn_lock(vp, LK_SHARED | LK_RETRY);
2497 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2498 					kve->kve_fileid = va.va_fileid;
2499 					/* truncate */
2500 					kve->kve_fsid = va.va_fsid;
2501 				}
2502 				vput(vp);
2503 			}
2504 		} else {
2505 			kve->kve_type = KVME_TYPE_NONE;
2506 			kve->kve_ref_count = 0;
2507 			kve->kve_shadow_count = 0;
2508 		}
2509 
2510 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2511 		if (freepath != NULL)
2512 			free(freepath, M_TEMP);
2513 
2514 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2515 		vm_map_lock_read(map);
2516 		if (error)
2517 			break;
2518 		if (last_timestamp != map->timestamp) {
2519 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2520 			entry = tmp_entry;
2521 		}
2522 	}
2523 	vm_map_unlock_read(map);
2524 	vmspace_free(vm);
2525 	PRELE(p);
2526 	free(kve, M_TEMP);
2527 	return (error);
2528 }
2529 #endif	/* COMPAT_FREEBSD7 */
2530 
2531 #ifdef KINFO_VMENTRY_SIZE
2532 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2533 #endif
2534 
2535 void
kern_proc_vmmap_resident(vm_map_t map,vm_map_entry_t entry,int * resident_count,bool * super)2536 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2537     int *resident_count, bool *super)
2538 {
2539 	vm_object_t obj, tobj;
2540 	vm_page_t m, m_adv;
2541 	vm_offset_t addr;
2542 	vm_paddr_t pa;
2543 	vm_pindex_t pi, pi_adv, pindex;
2544 	int incore;
2545 
2546 	*super = false;
2547 	*resident_count = 0;
2548 	if (vmmap_skip_res_cnt)
2549 		return;
2550 
2551 	pa = 0;
2552 	obj = entry->object.vm_object;
2553 	addr = entry->start;
2554 	m_adv = NULL;
2555 	pi = OFF_TO_IDX(entry->offset);
2556 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2557 		if (m_adv != NULL) {
2558 			m = m_adv;
2559 		} else {
2560 			pi_adv = atop(entry->end - addr);
2561 			pindex = pi;
2562 			for (tobj = obj;; tobj = tobj->backing_object) {
2563 				m = vm_page_find_least(tobj, pindex);
2564 				if (m != NULL) {
2565 					if (m->pindex == pindex)
2566 						break;
2567 					if (pi_adv > m->pindex - pindex) {
2568 						pi_adv = m->pindex - pindex;
2569 						m_adv = m;
2570 					}
2571 				}
2572 				if (tobj->backing_object == NULL)
2573 					goto next;
2574 				pindex += OFF_TO_IDX(tobj->
2575 				    backing_object_offset);
2576 			}
2577 		}
2578 		m_adv = NULL;
2579 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2580 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2581 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2582 			*super = true;
2583 			/*
2584 			 * The virtual page might be smaller than the physical
2585 			 * page, so we use the page size reported by the pmap
2586 			 * rather than m->psind.
2587 			 */
2588 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2589 		} else {
2590 			/*
2591 			 * We do not test the found page on validity.
2592 			 * Either the page is busy and being paged in,
2593 			 * or it was invalidated.  The first case
2594 			 * should be counted as resident, the second
2595 			 * is not so clear; we do account both.
2596 			 */
2597 			pi_adv = 1;
2598 		}
2599 		*resident_count += pi_adv;
2600 next:;
2601 	}
2602 }
2603 
2604 /*
2605  * Must be called with the process locked and will return unlocked.
2606  */
2607 int
kern_proc_vmmap_out(struct proc * p,struct sbuf * sb,ssize_t maxlen,int flags)2608 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2609 {
2610 	vm_map_entry_t entry, tmp_entry;
2611 	struct vattr va;
2612 	vm_map_t map;
2613 	vm_object_t lobj, nobj, obj, tobj;
2614 	char *fullpath, *freepath;
2615 	struct kinfo_vmentry *kve;
2616 	struct ucred *cred;
2617 	struct vnode *vp;
2618 	struct vmspace *vm;
2619 	vm_offset_t addr;
2620 	unsigned int last_timestamp;
2621 	int error;
2622 	key_t key;
2623 	unsigned short seq;
2624 	bool guard, super;
2625 
2626 	PROC_LOCK_ASSERT(p, MA_OWNED);
2627 
2628 	_PHOLD(p);
2629 	PROC_UNLOCK(p);
2630 	vm = vmspace_acquire_ref(p);
2631 	if (vm == NULL) {
2632 		PRELE(p);
2633 		return (ESRCH);
2634 	}
2635 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2636 
2637 	error = 0;
2638 	map = &vm->vm_map;
2639 	vm_map_lock_read(map);
2640 	VM_MAP_ENTRY_FOREACH(entry, map) {
2641 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2642 			continue;
2643 
2644 		addr = entry->end;
2645 		bzero(kve, sizeof(*kve));
2646 		obj = entry->object.vm_object;
2647 		if (obj != NULL) {
2648 			if ((obj->flags & OBJ_ANON) != 0)
2649 				kve->kve_obj = (uintptr_t)obj;
2650 
2651 			for (tobj = obj; tobj != NULL;
2652 			    tobj = tobj->backing_object) {
2653 				VM_OBJECT_RLOCK(tobj);
2654 				kve->kve_offset += tobj->backing_object_offset;
2655 				lobj = tobj;
2656 			}
2657 			if (obj->backing_object == NULL)
2658 				kve->kve_private_resident =
2659 				    obj->resident_page_count;
2660 			kern_proc_vmmap_resident(map, entry,
2661 			    &kve->kve_resident, &super);
2662 			if (super)
2663 				kve->kve_flags |= KVME_FLAG_SUPER;
2664 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2665 				nobj = tobj->backing_object;
2666 				if (tobj != obj && tobj != lobj)
2667 					VM_OBJECT_RUNLOCK(tobj);
2668 			}
2669 		} else {
2670 			lobj = NULL;
2671 		}
2672 
2673 		kve->kve_start = entry->start;
2674 		kve->kve_end = entry->end;
2675 		kve->kve_offset += entry->offset;
2676 
2677 		if (entry->protection & VM_PROT_READ)
2678 			kve->kve_protection |= KVME_PROT_READ;
2679 		if (entry->protection & VM_PROT_WRITE)
2680 			kve->kve_protection |= KVME_PROT_WRITE;
2681 		if (entry->protection & VM_PROT_EXECUTE)
2682 			kve->kve_protection |= KVME_PROT_EXEC;
2683 		if (entry->max_protection & VM_PROT_READ)
2684 			kve->kve_protection |= KVME_MAX_PROT_READ;
2685 		if (entry->max_protection & VM_PROT_WRITE)
2686 			kve->kve_protection |= KVME_MAX_PROT_WRITE;
2687 		if (entry->max_protection & VM_PROT_EXECUTE)
2688 			kve->kve_protection |= KVME_MAX_PROT_EXEC;
2689 
2690 		if (entry->eflags & MAP_ENTRY_COW)
2691 			kve->kve_flags |= KVME_FLAG_COW;
2692 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2693 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2694 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2695 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2696 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2697 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2698 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2699 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2700 
2701 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2702 
2703 		last_timestamp = map->timestamp;
2704 		vm_map_unlock_read(map);
2705 
2706 		freepath = NULL;
2707 		fullpath = "";
2708 		if (lobj != NULL) {
2709 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2710 			if (vp != NULL)
2711 				vref(vp);
2712 			if (lobj != obj)
2713 				VM_OBJECT_RUNLOCK(lobj);
2714 
2715 			kve->kve_ref_count = obj->ref_count;
2716 			kve->kve_shadow_count = obj->shadow_count;
2717 			if (obj->type == OBJT_DEVICE ||
2718 			    obj->type == OBJT_MGTDEVICE) {
2719 				cdev_pager_get_path(obj, kve->kve_path,
2720 				    sizeof(kve->kve_path));
2721 			}
2722 			VM_OBJECT_RUNLOCK(obj);
2723 			if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2724 				kve->kve_flags |= KVME_FLAG_SYSVSHM;
2725 				shmobjinfo(lobj, &key, &seq);
2726 				kve->kve_vn_fileid = key;
2727 				kve->kve_vn_fsid_freebsd11 = seq;
2728 			}
2729 			if ((lobj->flags & OBJ_POSIXSHM) != 0) {
2730 				kve->kve_flags |= KVME_FLAG_POSIXSHM;
2731 				shm_get_path(lobj, kve->kve_path,
2732 				    sizeof(kve->kve_path));
2733 			}
2734 			if (vp != NULL) {
2735 				vn_fullpath(vp, &fullpath, &freepath);
2736 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2737 				cred = curthread->td_ucred;
2738 				vn_lock(vp, LK_SHARED | LK_RETRY);
2739 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2740 					kve->kve_vn_fileid = va.va_fileid;
2741 					kve->kve_vn_fsid = va.va_fsid;
2742 					kve->kve_vn_fsid_freebsd11 =
2743 					    kve->kve_vn_fsid; /* truncate */
2744 					kve->kve_vn_mode =
2745 					    MAKEIMODE(va.va_type, va.va_mode);
2746 					kve->kve_vn_size = va.va_size;
2747 					kve->kve_vn_rdev = va.va_rdev;
2748 					kve->kve_vn_rdev_freebsd11 =
2749 					    kve->kve_vn_rdev; /* truncate */
2750 					kve->kve_status = KF_ATTR_VALID;
2751 				}
2752 				vput(vp);
2753 				strlcpy(kve->kve_path, fullpath, sizeof(
2754 				    kve->kve_path));
2755 				free(freepath, M_TEMP);
2756 			}
2757 		} else {
2758 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2759 			    KVME_TYPE_NONE;
2760 			kve->kve_ref_count = 0;
2761 			kve->kve_shadow_count = 0;
2762 		}
2763 
2764 		/* Pack record size down */
2765 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2766 			kve->kve_structsize =
2767 			    offsetof(struct kinfo_vmentry, kve_path) +
2768 			    strlen(kve->kve_path) + 1;
2769 		else
2770 			kve->kve_structsize = sizeof(*kve);
2771 		kve->kve_structsize = roundup(kve->kve_structsize,
2772 		    sizeof(uint64_t));
2773 
2774 		/* Halt filling and truncate rather than exceeding maxlen */
2775 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2776 			error = 0;
2777 			vm_map_lock_read(map);
2778 			break;
2779 		} else if (maxlen != -1)
2780 			maxlen -= kve->kve_structsize;
2781 
2782 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2783 			error = ENOMEM;
2784 		vm_map_lock_read(map);
2785 		if (error != 0)
2786 			break;
2787 		if (last_timestamp != map->timestamp) {
2788 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2789 			entry = tmp_entry;
2790 		}
2791 	}
2792 	vm_map_unlock_read(map);
2793 	vmspace_free(vm);
2794 	PRELE(p);
2795 	free(kve, M_TEMP);
2796 	return (error);
2797 }
2798 
2799 static int
sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)2800 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2801 {
2802 	struct proc *p;
2803 	struct sbuf sb;
2804 	u_int namelen;
2805 	int error, error2, *name;
2806 
2807 	namelen = arg2;
2808 	if (namelen != 1)
2809 		return (EINVAL);
2810 
2811 	name = (int *)arg1;
2812 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2813 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2814 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2815 	if (error != 0) {
2816 		sbuf_delete(&sb);
2817 		return (error);
2818 	}
2819 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2820 	error2 = sbuf_finish(&sb);
2821 	sbuf_delete(&sb);
2822 	return (error != 0 ? error : error2);
2823 }
2824 
2825 #if defined(STACK) || defined(DDB)
2826 static int
sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)2827 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2828 {
2829 	struct kinfo_kstack *kkstp;
2830 	int error, i, *name, numthreads;
2831 	lwpid_t *lwpidarray;
2832 	struct thread *td;
2833 	struct stack *st;
2834 	struct sbuf sb;
2835 	struct proc *p;
2836 	u_int namelen;
2837 
2838 	namelen = arg2;
2839 	if (namelen != 1)
2840 		return (EINVAL);
2841 
2842 	name = (int *)arg1;
2843 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2844 	if (error != 0)
2845 		return (error);
2846 
2847 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2848 	st = stack_create(M_WAITOK);
2849 
2850 	lwpidarray = NULL;
2851 	PROC_LOCK(p);
2852 	do {
2853 		if (lwpidarray != NULL) {
2854 			free(lwpidarray, M_TEMP);
2855 			lwpidarray = NULL;
2856 		}
2857 		numthreads = p->p_numthreads;
2858 		PROC_UNLOCK(p);
2859 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2860 		    M_WAITOK | M_ZERO);
2861 		PROC_LOCK(p);
2862 	} while (numthreads < p->p_numthreads);
2863 
2864 	/*
2865 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2866 	 * of changes could have taken place.  Should we check to see if the
2867 	 * vmspace has been replaced, or the like, in order to prevent
2868 	 * giving a snapshot that spans, say, execve(2), with some threads
2869 	 * before and some after?  Among other things, the credentials could
2870 	 * have changed, in which case the right to extract debug info might
2871 	 * no longer be assured.
2872 	 */
2873 	i = 0;
2874 	FOREACH_THREAD_IN_PROC(p, td) {
2875 		KASSERT(i < numthreads,
2876 		    ("sysctl_kern_proc_kstack: numthreads"));
2877 		lwpidarray[i] = td->td_tid;
2878 		i++;
2879 	}
2880 	PROC_UNLOCK(p);
2881 	numthreads = i;
2882 	for (i = 0; i < numthreads; i++) {
2883 		td = tdfind(lwpidarray[i], p->p_pid);
2884 		if (td == NULL) {
2885 			continue;
2886 		}
2887 		bzero(kkstp, sizeof(*kkstp));
2888 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2889 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2890 		thread_lock(td);
2891 		kkstp->kkst_tid = td->td_tid;
2892 		if (stack_save_td(st, td) == 0)
2893 			kkstp->kkst_state = KKST_STATE_STACKOK;
2894 		else
2895 			kkstp->kkst_state = KKST_STATE_RUNNING;
2896 		thread_unlock(td);
2897 		PROC_UNLOCK(p);
2898 		stack_sbuf_print(&sb, st);
2899 		sbuf_finish(&sb);
2900 		sbuf_delete(&sb);
2901 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2902 		if (error)
2903 			break;
2904 	}
2905 	PRELE(p);
2906 	if (lwpidarray != NULL)
2907 		free(lwpidarray, M_TEMP);
2908 	stack_destroy(st);
2909 	free(kkstp, M_TEMP);
2910 	return (error);
2911 }
2912 #endif
2913 
2914 /*
2915  * This sysctl allows a process to retrieve the full list of groups from
2916  * itself or another process.
2917  */
2918 static int
sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)2919 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2920 {
2921 	pid_t *pidp = (pid_t *)arg1;
2922 	unsigned int arglen = arg2;
2923 	struct proc *p;
2924 	struct ucred *cred;
2925 	int error;
2926 
2927 	if (arglen != 1)
2928 		return (EINVAL);
2929 	if (*pidp == -1) {	/* -1 means this process */
2930 		p = req->td->td_proc;
2931 		PROC_LOCK(p);
2932 	} else {
2933 		error = pget(*pidp, PGET_CANSEE, &p);
2934 		if (error != 0)
2935 			return (error);
2936 	}
2937 
2938 	cred = crhold(p->p_ucred);
2939 	PROC_UNLOCK(p);
2940 
2941 	error = SYSCTL_OUT(req, cred->cr_groups,
2942 	    cred->cr_ngroups * sizeof(gid_t));
2943 	crfree(cred);
2944 	return (error);
2945 }
2946 
2947 /*
2948  * This sysctl allows a process to retrieve or/and set the resource limit for
2949  * another process.
2950  */
2951 static int
sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)2952 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2953 {
2954 	int *name = (int *)arg1;
2955 	u_int namelen = arg2;
2956 	struct rlimit rlim;
2957 	struct proc *p;
2958 	u_int which;
2959 	int flags, error;
2960 
2961 	if (namelen != 2)
2962 		return (EINVAL);
2963 
2964 	which = (u_int)name[1];
2965 	if (which >= RLIM_NLIMITS)
2966 		return (EINVAL);
2967 
2968 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2969 		return (EINVAL);
2970 
2971 	flags = PGET_HOLD | PGET_NOTWEXIT;
2972 	if (req->newptr != NULL)
2973 		flags |= PGET_CANDEBUG;
2974 	else
2975 		flags |= PGET_CANSEE;
2976 	error = pget((pid_t)name[0], flags, &p);
2977 	if (error != 0)
2978 		return (error);
2979 
2980 	/*
2981 	 * Retrieve limit.
2982 	 */
2983 	if (req->oldptr != NULL) {
2984 		PROC_LOCK(p);
2985 		lim_rlimit_proc(p, which, &rlim);
2986 		PROC_UNLOCK(p);
2987 	}
2988 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2989 	if (error != 0)
2990 		goto errout;
2991 
2992 	/*
2993 	 * Set limit.
2994 	 */
2995 	if (req->newptr != NULL) {
2996 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2997 		if (error == 0)
2998 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2999 	}
3000 
3001 errout:
3002 	PRELE(p);
3003 	return (error);
3004 }
3005 
3006 /*
3007  * This sysctl allows a process to retrieve ps_strings structure location of
3008  * another process.
3009  */
3010 static int
sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)3011 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
3012 {
3013 	int *name = (int *)arg1;
3014 	u_int namelen = arg2;
3015 	struct proc *p;
3016 	vm_offset_t ps_strings;
3017 	int error;
3018 #ifdef COMPAT_FREEBSD32
3019 	uint32_t ps_strings32;
3020 #endif
3021 
3022 	if (namelen != 1)
3023 		return (EINVAL);
3024 
3025 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3026 	if (error != 0)
3027 		return (error);
3028 #ifdef COMPAT_FREEBSD32
3029 	if ((req->flags & SCTL_MASK32) != 0) {
3030 		/*
3031 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3032 		 * process.
3033 		 */
3034 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3035 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3036 		PROC_UNLOCK(p);
3037 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3038 		return (error);
3039 	}
3040 #endif
3041 	ps_strings = PROC_PS_STRINGS(p);
3042 	PROC_UNLOCK(p);
3043 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3044 	return (error);
3045 }
3046 
3047 /*
3048  * This sysctl allows a process to retrieve umask of another process.
3049  */
3050 static int
sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)3051 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3052 {
3053 	int *name = (int *)arg1;
3054 	u_int namelen = arg2;
3055 	struct proc *p;
3056 	int error;
3057 	u_short cmask;
3058 	pid_t pid;
3059 
3060 	if (namelen != 1)
3061 		return (EINVAL);
3062 
3063 	pid = (pid_t)name[0];
3064 	p = curproc;
3065 	if (pid == p->p_pid || pid == 0) {
3066 		cmask = p->p_pd->pd_cmask;
3067 		goto out;
3068 	}
3069 
3070 	error = pget(pid, PGET_WANTREAD, &p);
3071 	if (error != 0)
3072 		return (error);
3073 
3074 	cmask = p->p_pd->pd_cmask;
3075 	PRELE(p);
3076 out:
3077 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3078 	return (error);
3079 }
3080 
3081 /*
3082  * This sysctl allows a process to set and retrieve binary osreldate of
3083  * another process.
3084  */
3085 static int
sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)3086 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3087 {
3088 	int *name = (int *)arg1;
3089 	u_int namelen = arg2;
3090 	struct proc *p;
3091 	int flags, error, osrel;
3092 
3093 	if (namelen != 1)
3094 		return (EINVAL);
3095 
3096 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3097 		return (EINVAL);
3098 
3099 	flags = PGET_HOLD | PGET_NOTWEXIT;
3100 	if (req->newptr != NULL)
3101 		flags |= PGET_CANDEBUG;
3102 	else
3103 		flags |= PGET_CANSEE;
3104 	error = pget((pid_t)name[0], flags, &p);
3105 	if (error != 0)
3106 		return (error);
3107 
3108 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3109 	if (error != 0)
3110 		goto errout;
3111 
3112 	if (req->newptr != NULL) {
3113 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3114 		if (error != 0)
3115 			goto errout;
3116 		if (osrel < 0) {
3117 			error = EINVAL;
3118 			goto errout;
3119 		}
3120 		p->p_osrel = osrel;
3121 	}
3122 errout:
3123 	PRELE(p);
3124 	return (error);
3125 }
3126 
3127 static int
sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)3128 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3129 {
3130 	int *name = (int *)arg1;
3131 	u_int namelen = arg2;
3132 	struct proc *p;
3133 	struct kinfo_sigtramp kst;
3134 	const struct sysentvec *sv;
3135 	int error;
3136 #ifdef COMPAT_FREEBSD32
3137 	struct kinfo_sigtramp32 kst32;
3138 #endif
3139 
3140 	if (namelen != 1)
3141 		return (EINVAL);
3142 
3143 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3144 	if (error != 0)
3145 		return (error);
3146 	sv = p->p_sysent;
3147 #ifdef COMPAT_FREEBSD32
3148 	if ((req->flags & SCTL_MASK32) != 0) {
3149 		bzero(&kst32, sizeof(kst32));
3150 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3151 			if (PROC_HAS_SHP(p)) {
3152 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3153 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3154 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3155 				    *sv->sv_szsigcode :
3156 				    (uintptr_t)sv->sv_szsigcode);
3157 			} else {
3158 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3159 				    *sv->sv_szsigcode;
3160 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3161 			}
3162 		}
3163 		PROC_UNLOCK(p);
3164 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3165 		return (error);
3166 	}
3167 #endif
3168 	bzero(&kst, sizeof(kst));
3169 	if (PROC_HAS_SHP(p)) {
3170 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3171 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3172 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3173 		    (uintptr_t)sv->sv_szsigcode);
3174 	} else {
3175 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3176 		    *sv->sv_szsigcode;
3177 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3178 	}
3179 	PROC_UNLOCK(p);
3180 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3181 	return (error);
3182 }
3183 
3184 static int
sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)3185 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3186 {
3187 	int *name = (int *)arg1;
3188 	u_int namelen = arg2;
3189 	pid_t pid;
3190 	struct proc *p;
3191 	struct thread *td1;
3192 	uintptr_t addr;
3193 #ifdef COMPAT_FREEBSD32
3194 	uint32_t addr32;
3195 #endif
3196 	int error;
3197 
3198 	if (namelen != 1 || req->newptr != NULL)
3199 		return (EINVAL);
3200 
3201 	pid = (pid_t)name[0];
3202 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3203 	if (error != 0)
3204 		return (error);
3205 
3206 	PROC_LOCK(p);
3207 #ifdef COMPAT_FREEBSD32
3208 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3209 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3210 			error = EINVAL;
3211 			goto errlocked;
3212 		}
3213 	}
3214 #endif
3215 	if (pid <= PID_MAX) {
3216 		td1 = FIRST_THREAD_IN_PROC(p);
3217 	} else {
3218 		FOREACH_THREAD_IN_PROC(p, td1) {
3219 			if (td1->td_tid == pid)
3220 				break;
3221 		}
3222 	}
3223 	if (td1 == NULL) {
3224 		error = ESRCH;
3225 		goto errlocked;
3226 	}
3227 	/*
3228 	 * The access to the private thread flags.  It is fine as far
3229 	 * as no out-of-thin-air values are read from td_pflags, and
3230 	 * usermode read of the td_sigblock_ptr is racy inherently,
3231 	 * since target process might have already changed it
3232 	 * meantime.
3233 	 */
3234 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3235 		addr = (uintptr_t)td1->td_sigblock_ptr;
3236 	else
3237 		error = ENOTTY;
3238 
3239 errlocked:
3240 	_PRELE(p);
3241 	PROC_UNLOCK(p);
3242 	if (error != 0)
3243 		return (error);
3244 
3245 #ifdef COMPAT_FREEBSD32
3246 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3247 		addr32 = addr;
3248 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3249 	} else
3250 #endif
3251 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3252 	return (error);
3253 }
3254 
3255 static int
sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)3256 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3257 {
3258 	struct kinfo_vm_layout kvm;
3259 	struct proc *p;
3260 	struct vmspace *vmspace;
3261 	int error, *name;
3262 
3263 	name = (int *)arg1;
3264 	if ((u_int)arg2 != 1)
3265 		return (EINVAL);
3266 
3267 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3268 	if (error != 0)
3269 		return (error);
3270 #ifdef COMPAT_FREEBSD32
3271 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3272 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3273 			PROC_UNLOCK(p);
3274 			return (EINVAL);
3275 		}
3276 	}
3277 #endif
3278 	vmspace = vmspace_acquire_ref(p);
3279 	PROC_UNLOCK(p);
3280 
3281 	memset(&kvm, 0, sizeof(kvm));
3282 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3283 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3284 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3285 	kvm.kvm_text_size = vmspace->vm_tsize;
3286 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3287 	kvm.kvm_data_size = vmspace->vm_dsize;
3288 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3289 	kvm.kvm_stack_size = vmspace->vm_ssize;
3290 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3291 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3292 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3293 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3294 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3295 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3296 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3297 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3298 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3299 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3300 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3301 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3302 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3303 	    PROC_HAS_SHP(p))
3304 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3305 
3306 #ifdef COMPAT_FREEBSD32
3307 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3308 		struct kinfo_vm_layout32 kvm32;
3309 
3310 		memset(&kvm32, 0, sizeof(kvm32));
3311 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3312 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3313 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3314 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3315 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3316 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3317 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3318 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3319 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3320 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3321 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3322 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3323 		goto out;
3324 	}
3325 #endif
3326 
3327 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3328 #ifdef COMPAT_FREEBSD32
3329 out:
3330 #endif
3331 	vmspace_free(vmspace);
3332 	return (error);
3333 }
3334 
3335 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3336     "Process table");
3337 
3338 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3339 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3340 	"Return entire process table");
3341 
3342 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3343 	sysctl_kern_proc, "Process table");
3344 
3345 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3346 	sysctl_kern_proc, "Process table");
3347 
3348 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3349 	sysctl_kern_proc, "Process table");
3350 
3351 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3352 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3353 
3354 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3355 	sysctl_kern_proc, "Process table");
3356 
3357 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3358 	sysctl_kern_proc, "Process table");
3359 
3360 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3361 	sysctl_kern_proc, "Process table");
3362 
3363 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3364 	sysctl_kern_proc, "Process table");
3365 
3366 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3367 	sysctl_kern_proc, "Return process table, no threads");
3368 
3369 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3370 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3371 	sysctl_kern_proc_args, "Process argument list");
3372 
3373 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3374 	sysctl_kern_proc_env, "Process environment");
3375 
3376 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3377 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3378 
3379 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3380 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3381 
3382 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3383 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3384 	"Process syscall vector name (ABI type)");
3385 
3386 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3387 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3388 
3389 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3390 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3391 
3392 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3393 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3394 
3395 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3396 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3397 
3398 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3399 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3400 
3401 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3402 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3403 
3404 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3405 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3406 
3407 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3408 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3409 
3410 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3411 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3412 	"Return process table, including threads");
3413 
3414 #ifdef COMPAT_FREEBSD7
3415 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3416 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3417 #endif
3418 
3419 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3420 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3421 
3422 #if defined(STACK) || defined(DDB)
3423 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3424 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3425 #endif
3426 
3427 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3428 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3429 
3430 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3431 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3432 	"Process resource limits");
3433 
3434 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3435 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3436 	"Process ps_strings location");
3437 
3438 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3439 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3440 
3441 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3442 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3443 	"Process binary osreldate");
3444 
3445 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3446 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3447 	"Process signal trampoline location");
3448 
3449 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3450 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3451 	"Thread sigfastblock address");
3452 
3453 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3454 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3455 	"Process virtual address space layout info");
3456 
3457 static struct sx stop_all_proc_blocker;
3458 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3459 
3460 bool
stop_all_proc_block(void)3461 stop_all_proc_block(void)
3462 {
3463 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3464 }
3465 
3466 void
stop_all_proc_unblock(void)3467 stop_all_proc_unblock(void)
3468 {
3469 	sx_xunlock(&stop_all_proc_blocker);
3470 }
3471 
3472 int allproc_gen;
3473 
3474 /*
3475  * stop_all_proc() purpose is to stop all process which have usermode,
3476  * except current process for obvious reasons.  This makes it somewhat
3477  * unreliable when invoked from multithreaded process.  The service
3478  * must not be user-callable anyway.
3479  */
3480 void
stop_all_proc(void)3481 stop_all_proc(void)
3482 {
3483 	struct proc *cp, *p;
3484 	int r, gen;
3485 	bool restart, seen_stopped, seen_exiting, stopped_some;
3486 
3487 	if (!stop_all_proc_block())
3488 		return;
3489 
3490 	cp = curproc;
3491 allproc_loop:
3492 	sx_xlock(&allproc_lock);
3493 	gen = allproc_gen;
3494 	seen_exiting = seen_stopped = stopped_some = restart = false;
3495 	LIST_REMOVE(cp, p_list);
3496 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3497 	for (;;) {
3498 		p = LIST_NEXT(cp, p_list);
3499 		if (p == NULL)
3500 			break;
3501 		LIST_REMOVE(cp, p_list);
3502 		LIST_INSERT_AFTER(p, cp, p_list);
3503 		PROC_LOCK(p);
3504 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3505 		    P_STOPPED_SIG)) != 0) {
3506 			PROC_UNLOCK(p);
3507 			continue;
3508 		}
3509 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3510 			seen_exiting = true;
3511 			PROC_UNLOCK(p);
3512 			continue;
3513 		}
3514 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3515 			/*
3516 			 * Stopped processes are tolerated when there
3517 			 * are no other processes which might continue
3518 			 * them.  P_STOPPED_SINGLE but not
3519 			 * P_TOTAL_STOP process still has at least one
3520 			 * thread running.
3521 			 */
3522 			seen_stopped = true;
3523 			PROC_UNLOCK(p);
3524 			continue;
3525 		}
3526 		if ((p->p_flag & P_TRACED) != 0) {
3527 			/*
3528 			 * thread_single() below cannot stop traced p,
3529 			 * so skip it.  OTOH, we cannot require
3530 			 * restart because debugger might be either
3531 			 * already stopped or traced as well.
3532 			 */
3533 			PROC_UNLOCK(p);
3534 			continue;
3535 		}
3536 		sx_xunlock(&allproc_lock);
3537 		_PHOLD(p);
3538 		r = thread_single(p, SINGLE_ALLPROC);
3539 		if (r != 0)
3540 			restart = true;
3541 		else
3542 			stopped_some = true;
3543 		_PRELE(p);
3544 		PROC_UNLOCK(p);
3545 		sx_xlock(&allproc_lock);
3546 	}
3547 	/* Catch forked children we did not see in iteration. */
3548 	if (gen != allproc_gen)
3549 		restart = true;
3550 	sx_xunlock(&allproc_lock);
3551 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3552 		kern_yield(PRI_USER);
3553 		goto allproc_loop;
3554 	}
3555 }
3556 
3557 void
resume_all_proc(void)3558 resume_all_proc(void)
3559 {
3560 	struct proc *cp, *p;
3561 
3562 	cp = curproc;
3563 	sx_xlock(&allproc_lock);
3564 again:
3565 	LIST_REMOVE(cp, p_list);
3566 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3567 	for (;;) {
3568 		p = LIST_NEXT(cp, p_list);
3569 		if (p == NULL)
3570 			break;
3571 		LIST_REMOVE(cp, p_list);
3572 		LIST_INSERT_AFTER(p, cp, p_list);
3573 		PROC_LOCK(p);
3574 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3575 			sx_xunlock(&allproc_lock);
3576 			_PHOLD(p);
3577 			thread_single_end(p, SINGLE_ALLPROC);
3578 			_PRELE(p);
3579 			PROC_UNLOCK(p);
3580 			sx_xlock(&allproc_lock);
3581 		} else {
3582 			PROC_UNLOCK(p);
3583 		}
3584 	}
3585 	/*  Did the loop above missed any stopped process ? */
3586 	FOREACH_PROC_IN_SYSTEM(p) {
3587 		/* No need for proc lock. */
3588 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3589 			goto again;
3590 	}
3591 	sx_xunlock(&allproc_lock);
3592 
3593 	stop_all_proc_unblock();
3594 }
3595 
3596 /* #define	TOTAL_STOP_DEBUG	1 */
3597 #ifdef TOTAL_STOP_DEBUG
3598 volatile static int ap_resume;
3599 #include <sys/mount.h>
3600 
3601 static int
sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)3602 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3603 {
3604 	int error, val;
3605 
3606 	val = 0;
3607 	ap_resume = 0;
3608 	error = sysctl_handle_int(oidp, &val, 0, req);
3609 	if (error != 0 || req->newptr == NULL)
3610 		return (error);
3611 	if (val != 0) {
3612 		stop_all_proc();
3613 		syncer_suspend();
3614 		while (ap_resume == 0)
3615 			;
3616 		syncer_resume();
3617 		resume_all_proc();
3618 	}
3619 	return (0);
3620 }
3621 
3622 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3623     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3624     sysctl_debug_stop_all_proc, "I",
3625     "");
3626 #endif
3627