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/vm_radix.h>
96 #include <vm/uma.h>
97
98 #include <fs/devfs/devfs.h>
99
100 #ifdef COMPAT_FREEBSD32
101 #include <compat/freebsd32/freebsd32.h>
102 #include <compat/freebsd32/freebsd32_util.h>
103 #endif
104
105 SDT_PROVIDER_DEFINE(proc);
106
107 MALLOC_DEFINE(M_SESSION, "session", "session header");
108 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
109 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
110
111 static void doenterpgrp(struct proc *, struct pgrp *);
112 static void orphanpg(struct pgrp *pg);
113 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
116 int preferthread);
117 static void pgdelete(struct pgrp *);
118 static int pgrp_init(void *mem, int size, int flags);
119 static int proc_ctor(void *mem, int size, void *arg, int flags);
120 static void proc_dtor(void *mem, int size, void *arg);
121 static int proc_init(void *mem, int size, int flags);
122 static void proc_fini(void *mem, int size);
123 static void pargs_free(struct pargs *pa);
124
125 /*
126 * Other process lists
127 */
128 struct pidhashhead *pidhashtbl = NULL;
129 struct sx *pidhashtbl_lock;
130 u_long pidhash;
131 u_long pidhashlock;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
135 struct sx __exclusive_cache_line allproc_lock;
136 struct sx __exclusive_cache_line proctree_lock;
137 struct mtx __exclusive_cache_line ppeers_lock;
138 struct mtx __exclusive_cache_line procid_lock;
139 uma_zone_t proc_zone;
140 uma_zone_t pgrp_zone;
141
142 /*
143 * The offset of various fields in struct proc and struct thread.
144 * These are used by kernel debuggers to enumerate kernel threads and
145 * processes.
146 */
147 const int proc_off_p_pid = offsetof(struct proc, p_pid);
148 const int proc_off_p_comm = offsetof(struct proc, p_comm);
149 const int proc_off_p_list = offsetof(struct proc, p_list);
150 const int proc_off_p_hash = offsetof(struct proc, p_hash);
151 const int proc_off_p_threads = offsetof(struct proc, p_threads);
152 const int thread_off_td_tid = offsetof(struct thread, td_tid);
153 const int thread_off_td_name = offsetof(struct thread, td_name);
154 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
155 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
156 const int thread_off_td_plist = offsetof(struct thread, td_plist);
157
158 EVENTHANDLER_LIST_DEFINE(process_ctor);
159 EVENTHANDLER_LIST_DEFINE(process_dtor);
160 EVENTHANDLER_LIST_DEFINE(process_init);
161 EVENTHANDLER_LIST_DEFINE(process_fini);
162 EVENTHANDLER_LIST_DEFINE(process_exit);
163 EVENTHANDLER_LIST_DEFINE(process_fork);
164 EVENTHANDLER_LIST_DEFINE(process_exec);
165
166 int kstack_pages = KSTACK_PAGES;
167 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
168 &kstack_pages, 0,
169 "Kernel stack size in pages");
170 static int vmmap_skip_res_cnt = 0;
171 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
172 &vmmap_skip_res_cnt, 0,
173 "Skip calculation of the pages resident count in kern.proc.vmmap");
174
175 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
176 #ifdef COMPAT_FREEBSD32
177 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
178 #endif
179
180 /*
181 * Initialize global process hashing structures.
182 */
183 void
procinit(void)184 procinit(void)
185 {
186 u_long i;
187
188 sx_init(&allproc_lock, "allproc");
189 sx_init(&proctree_lock, "proctree");
190 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
191 mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
192 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
193 pidhashlock = (pidhash + 1) / 64;
194 if (pidhashlock > 0)
195 pidhashlock--;
196 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
197 M_PROC, M_WAITOK | M_ZERO);
198 for (i = 0; i < pidhashlock + 1; i++)
199 sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
200 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
201 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
202 proc_ctor, proc_dtor, proc_init, proc_fini,
203 UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
204 pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
205 pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
206 uihashinit();
207 }
208
209 /*
210 * Prepare a proc for use.
211 */
212 static int
proc_ctor(void * mem,int size,void * arg,int flags)213 proc_ctor(void *mem, int size, void *arg, int flags)
214 {
215 struct proc *p;
216 struct thread *td;
217
218 p = (struct proc *)mem;
219 #ifdef KDTRACE_HOOKS
220 kdtrace_proc_ctor(p);
221 #endif
222 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
223 td = FIRST_THREAD_IN_PROC(p);
224 if (td != NULL) {
225 /* Make sure all thread constructors are executed */
226 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
227 }
228 return (0);
229 }
230
231 /*
232 * Reclaim a proc after use.
233 */
234 static void
proc_dtor(void * mem,int size,void * arg)235 proc_dtor(void *mem, int size, void *arg)
236 {
237 struct proc *p;
238 struct thread *td;
239
240 /* INVARIANTS checks go here */
241 p = (struct proc *)mem;
242 td = FIRST_THREAD_IN_PROC(p);
243 if (td != NULL) {
244 #ifdef INVARIANTS
245 KASSERT((p->p_numthreads == 1),
246 ("bad number of threads in exiting process"));
247 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
248 #endif
249 /* Free all OSD associated to this thread. */
250 osd_thread_exit(td);
251 ast_kclear(td);
252
253 /* Make sure all thread destructors are executed */
254 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
255 }
256 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
257 #ifdef KDTRACE_HOOKS
258 kdtrace_proc_dtor(p);
259 #endif
260 if (p->p_ksi != NULL)
261 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
262 }
263
264 /*
265 * Initialize type-stable parts of a proc (when newly created).
266 */
267 static int
proc_init(void * mem,int size,int flags)268 proc_init(void *mem, int size, int flags)
269 {
270 struct proc *p;
271
272 p = (struct proc *)mem;
273 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
274 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
275 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
276 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
277 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
278 cv_init(&p->p_pwait, "ppwait");
279 TAILQ_INIT(&p->p_threads); /* all threads in proc */
280 EVENTHANDLER_DIRECT_INVOKE(process_init, p);
281 p->p_stats = pstats_alloc();
282 p->p_pgrp = NULL;
283 TAILQ_INIT(&p->p_kqtim_stop);
284 return (0);
285 }
286
287 /*
288 * UMA should ensure that this function is never called.
289 * Freeing a proc structure would violate type stability.
290 */
291 static void
proc_fini(void * mem,int size)292 proc_fini(void *mem, int size)
293 {
294 #ifdef notnow
295 struct proc *p;
296
297 p = (struct proc *)mem;
298 EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
299 pstats_free(p->p_stats);
300 thread_free(FIRST_THREAD_IN_PROC(p));
301 mtx_destroy(&p->p_mtx);
302 if (p->p_ksi != NULL)
303 ksiginfo_free(p->p_ksi);
304 #else
305 panic("proc reclaimed");
306 #endif
307 }
308
309 static int
pgrp_init(void * mem,int size,int flags)310 pgrp_init(void *mem, int size, int flags)
311 {
312 struct pgrp *pg;
313
314 pg = mem;
315 mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
316 sx_init(&pg->pg_killsx, "killpg racer");
317 return (0);
318 }
319
320 /*
321 * PID space management.
322 *
323 * These bitmaps are used by fork_findpid.
324 */
325 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
326 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
327 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
328 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
329
330 static bitstr_t *proc_id_array[] = {
331 proc_id_pidmap,
332 proc_id_grpidmap,
333 proc_id_sessidmap,
334 proc_id_reapmap,
335 };
336
337 void
proc_id_set(int type,pid_t id)338 proc_id_set(int type, pid_t id)
339 {
340
341 KASSERT(type >= 0 && type < nitems(proc_id_array),
342 ("invalid type %d\n", type));
343 mtx_lock(&procid_lock);
344 KASSERT(bit_test(proc_id_array[type], id) == 0,
345 ("bit %d already set in %d\n", id, type));
346 bit_set(proc_id_array[type], id);
347 mtx_unlock(&procid_lock);
348 }
349
350 void
proc_id_set_cond(int type,pid_t id)351 proc_id_set_cond(int type, pid_t id)
352 {
353
354 KASSERT(type >= 0 && type < nitems(proc_id_array),
355 ("invalid type %d\n", type));
356 if (bit_test(proc_id_array[type], id))
357 return;
358 mtx_lock(&procid_lock);
359 bit_set(proc_id_array[type], id);
360 mtx_unlock(&procid_lock);
361 }
362
363 void
proc_id_clear(int type,pid_t id)364 proc_id_clear(int type, pid_t id)
365 {
366
367 KASSERT(type >= 0 && type < nitems(proc_id_array),
368 ("invalid type %d\n", type));
369 mtx_lock(&procid_lock);
370 KASSERT(bit_test(proc_id_array[type], id) != 0,
371 ("bit %d not set in %d\n", id, type));
372 bit_clear(proc_id_array[type], id);
373 mtx_unlock(&procid_lock);
374 }
375
376 /*
377 * Is p an inferior of the current process?
378 */
379 int
inferior(struct proc * p)380 inferior(struct proc *p)
381 {
382
383 sx_assert(&proctree_lock, SX_LOCKED);
384 PROC_LOCK_ASSERT(p, MA_OWNED);
385 for (; p != curproc; p = proc_realparent(p)) {
386 if (p->p_pid == 0)
387 return (0);
388 }
389 return (1);
390 }
391
392 /*
393 * Shared lock all the pid hash lists.
394 */
395 void
pidhash_slockall(void)396 pidhash_slockall(void)
397 {
398 u_long i;
399
400 for (i = 0; i < pidhashlock + 1; i++)
401 sx_slock(&pidhashtbl_lock[i]);
402 }
403
404 /*
405 * Shared unlock all the pid hash lists.
406 */
407 void
pidhash_sunlockall(void)408 pidhash_sunlockall(void)
409 {
410 u_long i;
411
412 for (i = 0; i < pidhashlock + 1; i++)
413 sx_sunlock(&pidhashtbl_lock[i]);
414 }
415
416 /*
417 * Similar to pfind(), this function locate a process by number.
418 */
419 struct proc *
pfind_any_locked(pid_t pid)420 pfind_any_locked(pid_t pid)
421 {
422 struct proc *p;
423
424 sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
425 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
426 if (p->p_pid == pid) {
427 PROC_LOCK(p);
428 if (p->p_state == PRS_NEW) {
429 PROC_UNLOCK(p);
430 p = NULL;
431 }
432 break;
433 }
434 }
435 return (p);
436 }
437
438 /*
439 * Locate a process by number.
440 *
441 * By not returning processes in the PRS_NEW state, we allow callers to avoid
442 * testing for that condition to avoid dereferencing p_ucred, et al.
443 */
444 static __always_inline struct proc *
_pfind(pid_t pid,bool zombie)445 _pfind(pid_t pid, bool zombie)
446 {
447 struct proc *p;
448
449 p = curproc;
450 if (p->p_pid == pid) {
451 PROC_LOCK(p);
452 return (p);
453 }
454 sx_slock(PIDHASHLOCK(pid));
455 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
456 if (p->p_pid == pid) {
457 PROC_LOCK(p);
458 if (p->p_state == PRS_NEW ||
459 (!zombie && p->p_state == PRS_ZOMBIE)) {
460 PROC_UNLOCK(p);
461 p = NULL;
462 }
463 break;
464 }
465 }
466 sx_sunlock(PIDHASHLOCK(pid));
467 return (p);
468 }
469
470 struct proc *
pfind(pid_t pid)471 pfind(pid_t pid)
472 {
473
474 return (_pfind(pid, false));
475 }
476
477 /*
478 * Same as pfind but allow zombies.
479 */
480 struct proc *
pfind_any(pid_t pid)481 pfind_any(pid_t pid)
482 {
483
484 return (_pfind(pid, true));
485 }
486
487 /*
488 * Locate a process group by number.
489 * The caller must hold proctree_lock.
490 */
491 struct pgrp *
pgfind(pid_t pgid)492 pgfind(pid_t pgid)
493 {
494 struct pgrp *pgrp;
495
496 sx_assert(&proctree_lock, SX_LOCKED);
497
498 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
499 if (pgrp->pg_id == pgid) {
500 PGRP_LOCK(pgrp);
501 return (pgrp);
502 }
503 }
504 return (NULL);
505 }
506
507 /*
508 * Locate process and do additional manipulations, depending on flags.
509 */
510 int
pget(pid_t pid,int flags,struct proc ** pp)511 pget(pid_t pid, int flags, struct proc **pp)
512 {
513 struct proc *p;
514 struct thread *td1;
515 int error;
516
517 p = curproc;
518 if (p->p_pid == pid) {
519 PROC_LOCK(p);
520 } else {
521 p = NULL;
522 if (pid <= PID_MAX) {
523 if ((flags & PGET_NOTWEXIT) == 0)
524 p = pfind_any(pid);
525 else
526 p = pfind(pid);
527 } else if ((flags & PGET_NOTID) == 0) {
528 td1 = tdfind(pid, -1);
529 if (td1 != NULL)
530 p = td1->td_proc;
531 }
532 if (p == NULL)
533 return (ESRCH);
534 if ((flags & PGET_CANSEE) != 0) {
535 error = p_cansee(curthread, p);
536 if (error != 0)
537 goto errout;
538 }
539 }
540 if ((flags & PGET_CANDEBUG) != 0) {
541 error = p_candebug(curthread, p);
542 if (error != 0)
543 goto errout;
544 }
545 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
546 error = EPERM;
547 goto errout;
548 }
549 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
550 error = ESRCH;
551 goto errout;
552 }
553 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
554 /*
555 * XXXRW: Not clear ESRCH is the right error during proc
556 * execve().
557 */
558 error = ESRCH;
559 goto errout;
560 }
561 if ((flags & PGET_HOLD) != 0) {
562 _PHOLD(p);
563 PROC_UNLOCK(p);
564 }
565 *pp = p;
566 return (0);
567 errout:
568 PROC_UNLOCK(p);
569 return (error);
570 }
571
572 /*
573 * Create a new process group.
574 * pgid must be equal to the pid of p.
575 * Begin a new session if required.
576 */
577 int
enterpgrp(struct proc * p,pid_t pgid,struct pgrp * pgrp,struct session * sess)578 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
579 {
580 struct pgrp *old_pgrp;
581
582 sx_assert(&proctree_lock, SX_XLOCKED);
583
584 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
585 KASSERT(p->p_pid == pgid,
586 ("enterpgrp: new pgrp and pid != pgid"));
587 KASSERT(pgfind(pgid) == NULL,
588 ("enterpgrp: pgrp with pgid exists"));
589 KASSERT(!SESS_LEADER(p),
590 ("enterpgrp: session leader attempted setpgrp"));
591
592 old_pgrp = p->p_pgrp;
593 if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
594 sx_xunlock(&proctree_lock);
595 sx_xlock(&old_pgrp->pg_killsx);
596 sx_xunlock(&old_pgrp->pg_killsx);
597 return (ERESTART);
598 }
599 MPASS(old_pgrp == p->p_pgrp);
600
601 if (sess != NULL) {
602 /*
603 * new session
604 */
605 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
606 PROC_LOCK(p);
607 p->p_flag &= ~P_CONTROLT;
608 PROC_UNLOCK(p);
609 PGRP_LOCK(pgrp);
610 sess->s_leader = p;
611 sess->s_sid = p->p_pid;
612 proc_id_set(PROC_ID_SESSION, p->p_pid);
613 refcount_init(&sess->s_count, 1);
614 sess->s_ttyvp = NULL;
615 sess->s_ttydp = NULL;
616 sess->s_ttyp = NULL;
617 bcopy(p->p_session->s_login, sess->s_login,
618 sizeof(sess->s_login));
619 pgrp->pg_session = sess;
620 KASSERT(p == curproc,
621 ("enterpgrp: mksession and p != curproc"));
622 } else {
623 pgrp->pg_session = p->p_session;
624 sess_hold(pgrp->pg_session);
625 PGRP_LOCK(pgrp);
626 }
627 pgrp->pg_id = pgid;
628 proc_id_set(PROC_ID_GROUP, p->p_pid);
629 LIST_INIT(&pgrp->pg_members);
630 pgrp->pg_flags = 0;
631
632 /*
633 * As we have an exclusive lock of proctree_lock,
634 * this should not deadlock.
635 */
636 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
637 SLIST_INIT(&pgrp->pg_sigiolst);
638 PGRP_UNLOCK(pgrp);
639
640 doenterpgrp(p, pgrp);
641
642 sx_xunlock(&old_pgrp->pg_killsx);
643 return (0);
644 }
645
646 /*
647 * Move p to an existing process group
648 */
649 int
enterthispgrp(struct proc * p,struct pgrp * pgrp)650 enterthispgrp(struct proc *p, struct pgrp *pgrp)
651 {
652 struct pgrp *old_pgrp;
653
654 sx_assert(&proctree_lock, SX_XLOCKED);
655 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
656 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
657 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
658 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
659 KASSERT(pgrp->pg_session == p->p_session,
660 ("%s: pgrp's session %p, p->p_session %p proc %p\n",
661 __func__, pgrp->pg_session, p->p_session, p));
662 KASSERT(pgrp != p->p_pgrp,
663 ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
664
665 old_pgrp = p->p_pgrp;
666 if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
667 sx_xunlock(&proctree_lock);
668 sx_xlock(&old_pgrp->pg_killsx);
669 sx_xunlock(&old_pgrp->pg_killsx);
670 return (ERESTART);
671 }
672 MPASS(old_pgrp == p->p_pgrp);
673 if (!sx_try_xlock(&pgrp->pg_killsx)) {
674 sx_xunlock(&old_pgrp->pg_killsx);
675 sx_xunlock(&proctree_lock);
676 sx_xlock(&pgrp->pg_killsx);
677 sx_xunlock(&pgrp->pg_killsx);
678 return (ERESTART);
679 }
680
681 doenterpgrp(p, pgrp);
682
683 sx_xunlock(&pgrp->pg_killsx);
684 sx_xunlock(&old_pgrp->pg_killsx);
685 return (0);
686 }
687
688 /*
689 * If true, any child of q which belongs to group pgrp, qualifies the
690 * process group pgrp as not orphaned.
691 */
692 static bool
isjobproc(struct proc * q,struct pgrp * pgrp)693 isjobproc(struct proc *q, struct pgrp *pgrp)
694 {
695 sx_assert(&proctree_lock, SX_LOCKED);
696
697 return (q->p_pgrp != pgrp &&
698 q->p_pgrp->pg_session == pgrp->pg_session);
699 }
700
701 static struct proc *
jobc_reaper(struct proc * p)702 jobc_reaper(struct proc *p)
703 {
704 struct proc *pp;
705
706 sx_assert(&proctree_lock, SA_LOCKED);
707
708 for (pp = p;;) {
709 pp = pp->p_reaper;
710 if (pp->p_reaper == pp ||
711 (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
712 return (pp);
713 }
714 }
715
716 static struct proc *
jobc_parent(struct proc * p,struct proc * p_exiting)717 jobc_parent(struct proc *p, struct proc *p_exiting)
718 {
719 struct proc *pp;
720
721 sx_assert(&proctree_lock, SA_LOCKED);
722
723 pp = proc_realparent(p);
724 if (pp->p_pptr == NULL || pp == p_exiting ||
725 (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
726 return (pp);
727 return (jobc_reaper(pp));
728 }
729
730 int
pgrp_calc_jobc(struct pgrp * pgrp)731 pgrp_calc_jobc(struct pgrp *pgrp)
732 {
733 struct proc *q;
734 int cnt;
735
736 #ifdef INVARIANTS
737 if (!mtx_owned(&pgrp->pg_mtx))
738 sx_assert(&proctree_lock, SA_LOCKED);
739 #endif
740
741 cnt = 0;
742 LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
743 if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
744 q->p_pptr == NULL)
745 continue;
746 if (isjobproc(jobc_parent(q, NULL), pgrp))
747 cnt++;
748 }
749 return (cnt);
750 }
751
752 /*
753 * Move p to a process group
754 */
755 static void
doenterpgrp(struct proc * p,struct pgrp * pgrp)756 doenterpgrp(struct proc *p, struct pgrp *pgrp)
757 {
758 struct pgrp *savepgrp;
759 struct proc *pp;
760
761 sx_assert(&proctree_lock, SX_XLOCKED);
762 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
763 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
764 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
765 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
766
767 savepgrp = p->p_pgrp;
768 pp = jobc_parent(p, NULL);
769
770 PGRP_LOCK(pgrp);
771 PGRP_LOCK(savepgrp);
772 if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
773 orphanpg(savepgrp);
774 PROC_LOCK(p);
775 LIST_REMOVE(p, p_pglist);
776 p->p_pgrp = pgrp;
777 PROC_UNLOCK(p);
778 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
779 if (isjobproc(pp, pgrp))
780 pgrp->pg_flags &= ~PGRP_ORPHANED;
781 PGRP_UNLOCK(savepgrp);
782 PGRP_UNLOCK(pgrp);
783 if (LIST_EMPTY(&savepgrp->pg_members))
784 pgdelete(savepgrp);
785 }
786
787 /*
788 * remove process from process group
789 */
790 int
leavepgrp(struct proc * p)791 leavepgrp(struct proc *p)
792 {
793 struct pgrp *savepgrp;
794
795 sx_assert(&proctree_lock, SX_XLOCKED);
796 savepgrp = p->p_pgrp;
797 PGRP_LOCK(savepgrp);
798 PROC_LOCK(p);
799 LIST_REMOVE(p, p_pglist);
800 p->p_pgrp = NULL;
801 PROC_UNLOCK(p);
802 PGRP_UNLOCK(savepgrp);
803 if (LIST_EMPTY(&savepgrp->pg_members))
804 pgdelete(savepgrp);
805 return (0);
806 }
807
808 /*
809 * delete a process group
810 */
811 static void
pgdelete(struct pgrp * pgrp)812 pgdelete(struct pgrp *pgrp)
813 {
814 struct session *savesess;
815 struct tty *tp;
816
817 sx_assert(&proctree_lock, SX_XLOCKED);
818 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
819 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
820
821 /*
822 * Reset any sigio structures pointing to us as a result of
823 * F_SETOWN with our pgid. The proctree lock ensures that
824 * new sigio structures will not be added after this point.
825 */
826 funsetownlst(&pgrp->pg_sigiolst);
827
828 PGRP_LOCK(pgrp);
829 tp = pgrp->pg_session->s_ttyp;
830 LIST_REMOVE(pgrp, pg_hash);
831 savesess = pgrp->pg_session;
832 PGRP_UNLOCK(pgrp);
833
834 /* Remove the reference to the pgrp before deallocating it. */
835 if (tp != NULL) {
836 tty_lock(tp);
837 tty_rel_pgrp(tp, pgrp);
838 }
839
840 proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
841 uma_zfree(pgrp_zone, pgrp);
842 sess_release(savesess);
843 }
844
845
846 static void
fixjobc_kill(struct proc * p)847 fixjobc_kill(struct proc *p)
848 {
849 struct proc *q;
850 struct pgrp *pgrp;
851
852 sx_assert(&proctree_lock, SX_LOCKED);
853 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
854 pgrp = p->p_pgrp;
855 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
856 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
857
858 /*
859 * p no longer affects process group orphanage for children.
860 * It is marked by the flag because p is only physically
861 * removed from its process group on wait(2).
862 */
863 MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
864 p->p_treeflag |= P_TREE_GRPEXITED;
865
866 /*
867 * Check if exiting p orphans its own group.
868 */
869 pgrp = p->p_pgrp;
870 if (isjobproc(jobc_parent(p, NULL), pgrp)) {
871 PGRP_LOCK(pgrp);
872 if (pgrp_calc_jobc(pgrp) == 0)
873 orphanpg(pgrp);
874 PGRP_UNLOCK(pgrp);
875 }
876
877 /*
878 * Check this process' children to see whether they qualify
879 * their process groups after reparenting to reaper.
880 */
881 LIST_FOREACH(q, &p->p_children, p_sibling) {
882 pgrp = q->p_pgrp;
883 PGRP_LOCK(pgrp);
884 if (pgrp_calc_jobc(pgrp) == 0) {
885 /*
886 * We want to handle exactly the children that
887 * has p as realparent. Then, when calculating
888 * jobc_parent for children, we should ignore
889 * P_TREE_GRPEXITED flag already set on p.
890 */
891 if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
892 orphanpg(pgrp);
893 } else
894 pgrp->pg_flags &= ~PGRP_ORPHANED;
895 PGRP_UNLOCK(pgrp);
896 }
897 LIST_FOREACH(q, &p->p_orphans, p_orphan) {
898 pgrp = q->p_pgrp;
899 PGRP_LOCK(pgrp);
900 if (pgrp_calc_jobc(pgrp) == 0) {
901 if (isjobproc(p, pgrp))
902 orphanpg(pgrp);
903 } else
904 pgrp->pg_flags &= ~PGRP_ORPHANED;
905 PGRP_UNLOCK(pgrp);
906 }
907 }
908
909 void
killjobc(void)910 killjobc(void)
911 {
912 struct session *sp;
913 struct tty *tp;
914 struct proc *p;
915 struct vnode *ttyvp;
916
917 p = curproc;
918 MPASS(p->p_flag & P_WEXIT);
919 sx_assert(&proctree_lock, SX_LOCKED);
920
921 if (SESS_LEADER(p)) {
922 sp = p->p_session;
923
924 /*
925 * s_ttyp is not zero'd; we use this to indicate that
926 * the session once had a controlling terminal. (for
927 * logging and informational purposes)
928 */
929 SESS_LOCK(sp);
930 ttyvp = sp->s_ttyvp;
931 tp = sp->s_ttyp;
932 sp->s_ttyvp = NULL;
933 sp->s_ttydp = NULL;
934 sp->s_leader = NULL;
935 SESS_UNLOCK(sp);
936
937 /*
938 * Signal foreground pgrp and revoke access to
939 * controlling terminal if it has not been revoked
940 * already.
941 *
942 * Because the TTY may have been revoked in the mean
943 * time and could already have a new session associated
944 * with it, make sure we don't send a SIGHUP to a
945 * foreground process group that does not belong to this
946 * session.
947 */
948
949 if (tp != NULL) {
950 tty_lock(tp);
951 if (tp->t_session == sp)
952 tty_signal_pgrp(tp, SIGHUP);
953 tty_unlock(tp);
954 }
955
956 if (ttyvp != NULL) {
957 sx_xunlock(&proctree_lock);
958 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
959 VOP_REVOKE(ttyvp, REVOKEALL);
960 VOP_UNLOCK(ttyvp);
961 }
962 devfs_ctty_unref(ttyvp);
963 sx_xlock(&proctree_lock);
964 }
965 }
966 fixjobc_kill(p);
967 }
968
969 /*
970 * A process group has become orphaned, mark it as such for signal
971 * delivery code. If there are any stopped processes in the group,
972 * hang-up all process in that group.
973 */
974 static void
orphanpg(struct pgrp * pg)975 orphanpg(struct pgrp *pg)
976 {
977 struct proc *p;
978
979 PGRP_LOCK_ASSERT(pg, MA_OWNED);
980
981 pg->pg_flags |= PGRP_ORPHANED;
982
983 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
984 PROC_LOCK(p);
985 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
986 PROC_UNLOCK(p);
987 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
988 PROC_LOCK(p);
989 kern_psignal(p, SIGHUP);
990 kern_psignal(p, SIGCONT);
991 PROC_UNLOCK(p);
992 }
993 return;
994 }
995 PROC_UNLOCK(p);
996 }
997 }
998
999 void
sess_hold(struct session * s)1000 sess_hold(struct session *s)
1001 {
1002
1003 refcount_acquire(&s->s_count);
1004 }
1005
1006 void
sess_release(struct session * s)1007 sess_release(struct session *s)
1008 {
1009
1010 if (refcount_release(&s->s_count)) {
1011 if (s->s_ttyp != NULL) {
1012 tty_lock(s->s_ttyp);
1013 tty_rel_sess(s->s_ttyp, s);
1014 }
1015 proc_id_clear(PROC_ID_SESSION, s->s_sid);
1016 mtx_destroy(&s->s_mtx);
1017 free(s, M_SESSION);
1018 }
1019 }
1020
1021 #ifdef DDB
1022
1023 static void
db_print_pgrp_one(struct pgrp * pgrp,struct proc * p)1024 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1025 {
1026 db_printf(
1027 " pid %d at %p pr %d pgrp %p e %d jc %d\n",
1028 p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1029 p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1030 p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1031 }
1032
DB_SHOW_COMMAND_FLAGS(pgrpdump,pgrpdump,DB_CMD_MEMSAFE)1033 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1034 {
1035 struct pgrp *pgrp;
1036 struct proc *p;
1037 int i;
1038
1039 for (i = 0; i <= pgrphash; i++) {
1040 if (!LIST_EMPTY(&pgrphashtbl[i])) {
1041 db_printf("indx %d\n", i);
1042 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1043 db_printf(
1044 " pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1045 pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1046 pgrp->pg_session->s_count,
1047 LIST_FIRST(&pgrp->pg_members));
1048 LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1049 db_print_pgrp_one(pgrp, p);
1050 }
1051 }
1052 }
1053 }
1054 #endif /* DDB */
1055
1056 /*
1057 * Calculate the kinfo_proc members which contain process-wide
1058 * informations.
1059 * Must be called with the target process locked.
1060 */
1061 static void
fill_kinfo_aggregate(struct proc * p,struct kinfo_proc * kp)1062 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1063 {
1064 struct thread *td;
1065
1066 PROC_LOCK_ASSERT(p, MA_OWNED);
1067
1068 kp->ki_estcpu = 0;
1069 kp->ki_pctcpu = 0;
1070 FOREACH_THREAD_IN_PROC(p, td) {
1071 thread_lock(td);
1072 kp->ki_pctcpu += sched_pctcpu(td);
1073 kp->ki_estcpu += sched_estcpu(td);
1074 thread_unlock(td);
1075 }
1076 }
1077
1078 /*
1079 * Fill in any information that is common to all threads in the process.
1080 * Must be called with the target process locked.
1081 */
1082 static void
fill_kinfo_proc_only(struct proc * p,struct kinfo_proc * kp)1083 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1084 {
1085 struct thread *td0;
1086 struct ucred *cred;
1087 struct sigacts *ps;
1088 struct timeval boottime;
1089
1090 PROC_LOCK_ASSERT(p, MA_OWNED);
1091
1092 kp->ki_structsize = sizeof(*kp);
1093 kp->ki_paddr = p;
1094 kp->ki_addr =/* p->p_addr; */0; /* XXX */
1095 kp->ki_args = p->p_args;
1096 kp->ki_textvp = p->p_textvp;
1097 #ifdef KTRACE
1098 kp->ki_tracep = ktr_get_tracevp(p, false);
1099 kp->ki_traceflag = p->p_traceflag;
1100 #endif
1101 kp->ki_fd = p->p_fd;
1102 kp->ki_pd = p->p_pd;
1103 kp->ki_vmspace = p->p_vmspace;
1104 kp->ki_flag = p->p_flag;
1105 kp->ki_flag2 = p->p_flag2;
1106 cred = p->p_ucred;
1107 if (cred) {
1108 kp->ki_uid = cred->cr_uid;
1109 kp->ki_ruid = cred->cr_ruid;
1110 kp->ki_svuid = cred->cr_svuid;
1111 kp->ki_cr_flags = 0;
1112 if (cred->cr_flags & CRED_FLAG_CAPMODE)
1113 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1114 /* XXX bde doesn't like KI_NGROUPS */
1115 if (1 + cred->cr_ngroups > KI_NGROUPS) {
1116 kp->ki_ngroups = KI_NGROUPS;
1117 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1118 } else
1119 kp->ki_ngroups = 1 + cred->cr_ngroups;
1120 kp->ki_groups[0] = cred->cr_gid;
1121 bcopy(cred->cr_groups, kp->ki_groups + 1,
1122 (kp->ki_ngroups - 1) * sizeof(gid_t));
1123 kp->ki_rgid = cred->cr_rgid;
1124 kp->ki_svgid = cred->cr_svgid;
1125 /* If jailed(cred), emulate the old P_JAILED flag. */
1126 if (jailed(cred)) {
1127 kp->ki_flag |= P_JAILED;
1128 /* If inside the jail, use 0 as a jail ID. */
1129 if (cred->cr_prison != curthread->td_ucred->cr_prison)
1130 kp->ki_jid = cred->cr_prison->pr_id;
1131 }
1132 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1133 sizeof(kp->ki_loginclass));
1134 }
1135 ps = p->p_sigacts;
1136 if (ps) {
1137 mtx_lock(&ps->ps_mtx);
1138 kp->ki_sigignore = ps->ps_sigignore;
1139 kp->ki_sigcatch = ps->ps_sigcatch;
1140 mtx_unlock(&ps->ps_mtx);
1141 }
1142 if (p->p_state != PRS_NEW &&
1143 p->p_state != PRS_ZOMBIE &&
1144 p->p_vmspace != NULL) {
1145 struct vmspace *vm = p->p_vmspace;
1146
1147 kp->ki_size = vm->vm_map.size;
1148 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1149 FOREACH_THREAD_IN_PROC(p, td0)
1150 kp->ki_rssize += td0->td_kstack_pages;
1151 kp->ki_swrss = vm->vm_swrss;
1152 kp->ki_tsize = vm->vm_tsize;
1153 kp->ki_dsize = vm->vm_dsize;
1154 kp->ki_ssize = vm->vm_ssize;
1155 } else if (p->p_state == PRS_ZOMBIE)
1156 kp->ki_stat = SZOMB;
1157 kp->ki_sflag = PS_INMEM;
1158 /* Calculate legacy swtime as seconds since 'swtick'. */
1159 kp->ki_swtime = (ticks - p->p_swtick) / hz;
1160 kp->ki_pid = p->p_pid;
1161 kp->ki_nice = p->p_nice;
1162 kp->ki_fibnum = p->p_fibnum;
1163 kp->ki_start = p->p_stats->p_start;
1164 getboottime(&boottime);
1165 timevaladd(&kp->ki_start, &boottime);
1166 PROC_STATLOCK(p);
1167 rufetch(p, &kp->ki_rusage);
1168 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1169 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1170 PROC_STATUNLOCK(p);
1171 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1172 /* Some callers want child times in a single value. */
1173 kp->ki_childtime = kp->ki_childstime;
1174 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1175
1176 FOREACH_THREAD_IN_PROC(p, td0)
1177 kp->ki_cow += td0->td_cow;
1178
1179 if (p->p_comm[0] != '\0')
1180 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1181 if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1182 p->p_sysent->sv_name[0] != '\0')
1183 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1184 kp->ki_siglist = p->p_siglist;
1185 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1186 kp->ki_acflag = p->p_acflag;
1187 kp->ki_lock = p->p_lock;
1188 if (p->p_pptr) {
1189 kp->ki_ppid = p->p_oppid;
1190 if (p->p_flag & P_TRACED)
1191 kp->ki_tracer = p->p_pptr->p_pid;
1192 }
1193 }
1194
1195 /*
1196 * Fill job-related process information.
1197 */
1198 static void
fill_kinfo_proc_pgrp(struct proc * p,struct kinfo_proc * kp)1199 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1200 {
1201 struct tty *tp;
1202 struct session *sp;
1203 struct pgrp *pgrp;
1204
1205 sx_assert(&proctree_lock, SA_LOCKED);
1206 PROC_LOCK_ASSERT(p, MA_OWNED);
1207
1208 pgrp = p->p_pgrp;
1209 if (pgrp == NULL)
1210 return;
1211
1212 kp->ki_pgid = pgrp->pg_id;
1213 kp->ki_jobc = pgrp_calc_jobc(pgrp);
1214
1215 sp = pgrp->pg_session;
1216 tp = NULL;
1217
1218 if (sp != NULL) {
1219 kp->ki_sid = sp->s_sid;
1220 SESS_LOCK(sp);
1221 strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1222 if (sp->s_ttyvp)
1223 kp->ki_kiflag |= KI_CTTY;
1224 if (SESS_LEADER(p))
1225 kp->ki_kiflag |= KI_SLEADER;
1226 tp = sp->s_ttyp;
1227 SESS_UNLOCK(sp);
1228 }
1229
1230 if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1231 kp->ki_tdev = tty_udev(tp);
1232 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1233 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1234 if (tp->t_session)
1235 kp->ki_tsid = tp->t_session->s_sid;
1236 } else {
1237 kp->ki_tdev = NODEV;
1238 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1239 }
1240 }
1241
1242 /*
1243 * Fill in information that is thread specific. Must be called with
1244 * target process locked. If 'preferthread' is set, overwrite certain
1245 * process-related fields that are maintained for both threads and
1246 * processes.
1247 */
1248 static void
fill_kinfo_thread(struct thread * td,struct kinfo_proc * kp,int preferthread)1249 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1250 {
1251 struct proc *p;
1252
1253 p = td->td_proc;
1254 kp->ki_tdaddr = td;
1255 PROC_LOCK_ASSERT(p, MA_OWNED);
1256
1257 if (preferthread)
1258 PROC_STATLOCK(p);
1259 thread_lock(td);
1260 if (td->td_wmesg != NULL)
1261 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1262 else
1263 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1264 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1265 sizeof(kp->ki_tdname)) {
1266 strlcpy(kp->ki_moretdname,
1267 td->td_name + sizeof(kp->ki_tdname) - 1,
1268 sizeof(kp->ki_moretdname));
1269 } else {
1270 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1271 }
1272 if (TD_ON_LOCK(td)) {
1273 kp->ki_kiflag |= KI_LOCKBLOCK;
1274 strlcpy(kp->ki_lockname, td->td_lockname,
1275 sizeof(kp->ki_lockname));
1276 } else {
1277 kp->ki_kiflag &= ~KI_LOCKBLOCK;
1278 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1279 }
1280
1281 if (p->p_state == PRS_NORMAL) { /* approximate. */
1282 if (TD_ON_RUNQ(td) ||
1283 TD_CAN_RUN(td) ||
1284 TD_IS_RUNNING(td)) {
1285 kp->ki_stat = SRUN;
1286 } else if (P_SHOULDSTOP(p)) {
1287 kp->ki_stat = SSTOP;
1288 } else if (TD_IS_SLEEPING(td)) {
1289 kp->ki_stat = SSLEEP;
1290 } else if (TD_ON_LOCK(td)) {
1291 kp->ki_stat = SLOCK;
1292 } else {
1293 kp->ki_stat = SWAIT;
1294 }
1295 } else if (p->p_state == PRS_ZOMBIE) {
1296 kp->ki_stat = SZOMB;
1297 } else {
1298 kp->ki_stat = SIDL;
1299 }
1300
1301 /* Things in the thread */
1302 kp->ki_wchan = td->td_wchan;
1303 kp->ki_pri.pri_level = td->td_priority;
1304 kp->ki_pri.pri_native = td->td_base_pri;
1305
1306 /*
1307 * Note: legacy fields; clamp at the old NOCPU value and/or
1308 * the maximum u_char CPU value.
1309 */
1310 if (td->td_lastcpu == NOCPU)
1311 kp->ki_lastcpu_old = NOCPU_OLD;
1312 else if (td->td_lastcpu > MAXCPU_OLD)
1313 kp->ki_lastcpu_old = MAXCPU_OLD;
1314 else
1315 kp->ki_lastcpu_old = td->td_lastcpu;
1316
1317 if (td->td_oncpu == NOCPU)
1318 kp->ki_oncpu_old = NOCPU_OLD;
1319 else if (td->td_oncpu > MAXCPU_OLD)
1320 kp->ki_oncpu_old = MAXCPU_OLD;
1321 else
1322 kp->ki_oncpu_old = td->td_oncpu;
1323
1324 kp->ki_lastcpu = td->td_lastcpu;
1325 kp->ki_oncpu = td->td_oncpu;
1326 kp->ki_tdflags = td->td_flags;
1327 kp->ki_tid = td->td_tid;
1328 kp->ki_numthreads = p->p_numthreads;
1329 kp->ki_pcb = td->td_pcb;
1330 kp->ki_kstack = (void *)td->td_kstack;
1331 kp->ki_slptime = (ticks - td->td_slptick) / hz;
1332 kp->ki_pri.pri_class = td->td_pri_class;
1333 kp->ki_pri.pri_user = td->td_user_pri;
1334
1335 if (preferthread) {
1336 rufetchtd(td, &kp->ki_rusage);
1337 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1338 kp->ki_pctcpu = sched_pctcpu(td);
1339 kp->ki_estcpu = sched_estcpu(td);
1340 kp->ki_cow = td->td_cow;
1341 }
1342
1343 /* We can't get this anymore but ps etc never used it anyway. */
1344 kp->ki_rqindex = 0;
1345
1346 if (preferthread)
1347 kp->ki_siglist = td->td_siglist;
1348 kp->ki_sigmask = td->td_sigmask;
1349 thread_unlock(td);
1350 if (preferthread)
1351 PROC_STATUNLOCK(p);
1352
1353 if ((td->td_pflags & TDP2_UEXTERR) != 0)
1354 kp->ki_uerrmsg = td->td_exterr_ptr;
1355 }
1356
1357 /*
1358 * Fill in a kinfo_proc structure for the specified process.
1359 * Must be called with the target process locked.
1360 */
1361 void
fill_kinfo_proc(struct proc * p,struct kinfo_proc * kp)1362 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1363 {
1364 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1365
1366 bzero(kp, sizeof(*kp));
1367
1368 fill_kinfo_proc_pgrp(p,kp);
1369 fill_kinfo_proc_only(p, kp);
1370 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1371 fill_kinfo_aggregate(p, kp);
1372 }
1373
1374 struct pstats *
pstats_alloc(void)1375 pstats_alloc(void)
1376 {
1377
1378 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1379 }
1380
1381 /*
1382 * Copy parts of p_stats; zero the rest of p_stats (statistics).
1383 */
1384 void
pstats_fork(struct pstats * src,struct pstats * dst)1385 pstats_fork(struct pstats *src, struct pstats *dst)
1386 {
1387
1388 bzero(&dst->pstat_startzero,
1389 __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1390 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1391 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1392 }
1393
1394 void
pstats_free(struct pstats * ps)1395 pstats_free(struct pstats *ps)
1396 {
1397
1398 free(ps, M_SUBPROC);
1399 }
1400
1401 #ifdef COMPAT_FREEBSD32
1402
1403 /*
1404 * This function is typically used to copy out the kernel address, so
1405 * it can be replaced by assignment of zero.
1406 */
1407 static inline uint32_t
ptr32_trim(const void * ptr)1408 ptr32_trim(const void *ptr)
1409 {
1410 uintptr_t uptr;
1411
1412 uptr = (uintptr_t)ptr;
1413 return ((uptr > UINT_MAX) ? 0 : uptr);
1414 }
1415
1416 #define PTRTRIM_CP(src,dst,fld) \
1417 do { (dst).fld = ptr32_trim((src).fld); } while (0)
1418
1419 static void
freebsd32_kinfo_proc_out(const struct kinfo_proc * ki,struct kinfo_proc32 * ki32)1420 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1421 {
1422 int i;
1423
1424 bzero(ki32, sizeof(struct kinfo_proc32));
1425 ki32->ki_structsize = sizeof(struct kinfo_proc32);
1426 CP(*ki, *ki32, ki_layout);
1427 PTRTRIM_CP(*ki, *ki32, ki_args);
1428 PTRTRIM_CP(*ki, *ki32, ki_paddr);
1429 PTRTRIM_CP(*ki, *ki32, ki_addr);
1430 PTRTRIM_CP(*ki, *ki32, ki_tracep);
1431 PTRTRIM_CP(*ki, *ki32, ki_textvp);
1432 PTRTRIM_CP(*ki, *ki32, ki_fd);
1433 PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1434 PTRTRIM_CP(*ki, *ki32, ki_wchan);
1435 CP(*ki, *ki32, ki_pid);
1436 CP(*ki, *ki32, ki_ppid);
1437 CP(*ki, *ki32, ki_pgid);
1438 CP(*ki, *ki32, ki_tpgid);
1439 CP(*ki, *ki32, ki_sid);
1440 CP(*ki, *ki32, ki_tsid);
1441 CP(*ki, *ki32, ki_jobc);
1442 CP(*ki, *ki32, ki_tdev);
1443 CP(*ki, *ki32, ki_tdev_freebsd11);
1444 CP(*ki, *ki32, ki_siglist);
1445 CP(*ki, *ki32, ki_sigmask);
1446 CP(*ki, *ki32, ki_sigignore);
1447 CP(*ki, *ki32, ki_sigcatch);
1448 CP(*ki, *ki32, ki_uid);
1449 CP(*ki, *ki32, ki_ruid);
1450 CP(*ki, *ki32, ki_svuid);
1451 CP(*ki, *ki32, ki_rgid);
1452 CP(*ki, *ki32, ki_svgid);
1453 CP(*ki, *ki32, ki_ngroups);
1454 for (i = 0; i < KI_NGROUPS; i++)
1455 CP(*ki, *ki32, ki_groups[i]);
1456 CP(*ki, *ki32, ki_size);
1457 CP(*ki, *ki32, ki_rssize);
1458 CP(*ki, *ki32, ki_swrss);
1459 CP(*ki, *ki32, ki_tsize);
1460 CP(*ki, *ki32, ki_dsize);
1461 CP(*ki, *ki32, ki_ssize);
1462 CP(*ki, *ki32, ki_xstat);
1463 CP(*ki, *ki32, ki_acflag);
1464 CP(*ki, *ki32, ki_pctcpu);
1465 CP(*ki, *ki32, ki_estcpu);
1466 CP(*ki, *ki32, ki_slptime);
1467 CP(*ki, *ki32, ki_swtime);
1468 CP(*ki, *ki32, ki_cow);
1469 CP(*ki, *ki32, ki_runtime);
1470 TV_CP(*ki, *ki32, ki_start);
1471 TV_CP(*ki, *ki32, ki_childtime);
1472 CP(*ki, *ki32, ki_flag);
1473 CP(*ki, *ki32, ki_kiflag);
1474 CP(*ki, *ki32, ki_traceflag);
1475 CP(*ki, *ki32, ki_stat);
1476 CP(*ki, *ki32, ki_nice);
1477 CP(*ki, *ki32, ki_lock);
1478 CP(*ki, *ki32, ki_rqindex);
1479 CP(*ki, *ki32, ki_oncpu);
1480 CP(*ki, *ki32, ki_lastcpu);
1481
1482 /* XXX TODO: wrap cpu value as appropriate */
1483 CP(*ki, *ki32, ki_oncpu_old);
1484 CP(*ki, *ki32, ki_lastcpu_old);
1485
1486 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1487 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1488 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1489 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1490 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1491 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1492 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1493 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1494 CP(*ki, *ki32, ki_tracer);
1495 CP(*ki, *ki32, ki_flag2);
1496 CP(*ki, *ki32, ki_fibnum);
1497 CP(*ki, *ki32, ki_cr_flags);
1498 CP(*ki, *ki32, ki_jid);
1499 CP(*ki, *ki32, ki_numthreads);
1500 CP(*ki, *ki32, ki_tid);
1501 CP(*ki, *ki32, ki_pri);
1502 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1503 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1504 PTRTRIM_CP(*ki, *ki32, ki_pcb);
1505 PTRTRIM_CP(*ki, *ki32, ki_kstack);
1506 PTRTRIM_CP(*ki, *ki32, ki_udata);
1507 PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1508 CP(*ki, *ki32, ki_sflag);
1509 CP(*ki, *ki32, ki_tdflags);
1510 PTRTRIM_CP(*ki, *ki32, ki_uerrmsg);
1511 }
1512 #endif
1513
1514 static ssize_t
kern_proc_out_size(struct proc * p,int flags)1515 kern_proc_out_size(struct proc *p, int flags)
1516 {
1517 ssize_t size = 0;
1518
1519 PROC_LOCK_ASSERT(p, MA_OWNED);
1520
1521 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1522 #ifdef COMPAT_FREEBSD32
1523 if ((flags & KERN_PROC_MASK32) != 0) {
1524 size += sizeof(struct kinfo_proc32);
1525 } else
1526 #endif
1527 size += sizeof(struct kinfo_proc);
1528 } else {
1529 #ifdef COMPAT_FREEBSD32
1530 if ((flags & KERN_PROC_MASK32) != 0)
1531 size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1532 else
1533 #endif
1534 size += sizeof(struct kinfo_proc) * p->p_numthreads;
1535 }
1536 PROC_UNLOCK(p);
1537 return (size);
1538 }
1539
1540 int
kern_proc_out(struct proc * p,struct sbuf * sb,int flags)1541 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1542 {
1543 struct thread *td;
1544 struct kinfo_proc ki;
1545 #ifdef COMPAT_FREEBSD32
1546 struct kinfo_proc32 ki32;
1547 #endif
1548 int error;
1549
1550 PROC_LOCK_ASSERT(p, MA_OWNED);
1551 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1552
1553 error = 0;
1554 fill_kinfo_proc(p, &ki);
1555 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1556 #ifdef COMPAT_FREEBSD32
1557 if ((flags & KERN_PROC_MASK32) != 0) {
1558 freebsd32_kinfo_proc_out(&ki, &ki32);
1559 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1560 error = ENOMEM;
1561 } else
1562 #endif
1563 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1564 error = ENOMEM;
1565 } else {
1566 FOREACH_THREAD_IN_PROC(p, td) {
1567 fill_kinfo_thread(td, &ki, 1);
1568 #ifdef COMPAT_FREEBSD32
1569 if ((flags & KERN_PROC_MASK32) != 0) {
1570 freebsd32_kinfo_proc_out(&ki, &ki32);
1571 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1572 error = ENOMEM;
1573 } else
1574 #endif
1575 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1576 error = ENOMEM;
1577 if (error != 0)
1578 break;
1579 }
1580 }
1581 PROC_UNLOCK(p);
1582 return (error);
1583 }
1584
1585 static int
sysctl_out_proc(struct proc * p,struct sysctl_req * req,int flags)1586 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1587 {
1588 struct sbuf sb;
1589 struct kinfo_proc ki;
1590 int error, error2;
1591
1592 if (req->oldptr == NULL)
1593 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1594
1595 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1596 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1597 error = kern_proc_out(p, &sb, flags);
1598 error2 = sbuf_finish(&sb);
1599 sbuf_delete(&sb);
1600 if (error != 0)
1601 return (error);
1602 else if (error2 != 0)
1603 return (error2);
1604 return (0);
1605 }
1606
1607 int
proc_iterate(int (* cb)(struct proc *,void *),void * cbarg)1608 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1609 {
1610 struct proc *p;
1611 int error, i, j;
1612
1613 for (i = 0; i < pidhashlock + 1; i++) {
1614 sx_slock(&proctree_lock);
1615 sx_slock(&pidhashtbl_lock[i]);
1616 for (j = i; j <= pidhash; j += pidhashlock + 1) {
1617 LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1618 if (p->p_state == PRS_NEW)
1619 continue;
1620 error = cb(p, cbarg);
1621 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1622 if (error != 0) {
1623 sx_sunlock(&pidhashtbl_lock[i]);
1624 sx_sunlock(&proctree_lock);
1625 return (error);
1626 }
1627 }
1628 }
1629 sx_sunlock(&pidhashtbl_lock[i]);
1630 sx_sunlock(&proctree_lock);
1631 }
1632 return (0);
1633 }
1634
1635 struct kern_proc_out_args {
1636 struct sysctl_req *req;
1637 int flags;
1638 int oid_number;
1639 int *name;
1640 };
1641
1642 static int
sysctl_kern_proc_iterate(struct proc * p,void * origarg)1643 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1644 {
1645 struct kern_proc_out_args *arg = origarg;
1646 int *name = arg->name;
1647 int oid_number = arg->oid_number;
1648 int flags = arg->flags;
1649 struct sysctl_req *req = arg->req;
1650 int error = 0;
1651
1652 PROC_LOCK(p);
1653
1654 KASSERT(p->p_ucred != NULL,
1655 ("process credential is NULL for non-NEW proc"));
1656 /*
1657 * Show a user only appropriate processes.
1658 */
1659 if (p_cansee(curthread, p))
1660 goto skip;
1661 /*
1662 * TODO - make more efficient (see notes below).
1663 * do by session.
1664 */
1665 switch (oid_number) {
1666 case KERN_PROC_GID:
1667 if (p->p_ucred->cr_gid != (gid_t)name[0])
1668 goto skip;
1669 break;
1670
1671 case KERN_PROC_PGRP:
1672 /* could do this by traversing pgrp */
1673 if (p->p_pgrp == NULL ||
1674 p->p_pgrp->pg_id != (pid_t)name[0])
1675 goto skip;
1676 break;
1677
1678 case KERN_PROC_RGID:
1679 if (p->p_ucred->cr_rgid != (gid_t)name[0])
1680 goto skip;
1681 break;
1682
1683 case KERN_PROC_SESSION:
1684 if (p->p_session == NULL ||
1685 p->p_session->s_sid != (pid_t)name[0])
1686 goto skip;
1687 break;
1688
1689 case KERN_PROC_TTY:
1690 if ((p->p_flag & P_CONTROLT) == 0 ||
1691 p->p_session == NULL)
1692 goto skip;
1693 /* XXX proctree_lock */
1694 SESS_LOCK(p->p_session);
1695 if (p->p_session->s_ttyp == NULL ||
1696 tty_udev(p->p_session->s_ttyp) !=
1697 (dev_t)name[0]) {
1698 SESS_UNLOCK(p->p_session);
1699 goto skip;
1700 }
1701 SESS_UNLOCK(p->p_session);
1702 break;
1703
1704 case KERN_PROC_UID:
1705 if (p->p_ucred->cr_uid != (uid_t)name[0])
1706 goto skip;
1707 break;
1708
1709 case KERN_PROC_RUID:
1710 if (p->p_ucred->cr_ruid != (uid_t)name[0])
1711 goto skip;
1712 break;
1713
1714 case KERN_PROC_PROC:
1715 break;
1716
1717 default:
1718 break;
1719 }
1720 error = sysctl_out_proc(p, req, flags);
1721 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1722 return (error);
1723 skip:
1724 PROC_UNLOCK(p);
1725 return (0);
1726 }
1727
1728 static int
sysctl_kern_proc(SYSCTL_HANDLER_ARGS)1729 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1730 {
1731 struct kern_proc_out_args iterarg;
1732 int *name = (int *)arg1;
1733 u_int namelen = arg2;
1734 struct proc *p;
1735 int flags, oid_number;
1736 int error = 0;
1737
1738 oid_number = oidp->oid_number;
1739 if (oid_number != KERN_PROC_ALL &&
1740 (oid_number & KERN_PROC_INC_THREAD) == 0)
1741 flags = KERN_PROC_NOTHREADS;
1742 else {
1743 flags = 0;
1744 oid_number &= ~KERN_PROC_INC_THREAD;
1745 }
1746 #ifdef COMPAT_FREEBSD32
1747 if (req->flags & SCTL_MASK32)
1748 flags |= KERN_PROC_MASK32;
1749 #endif
1750 if (oid_number == KERN_PROC_PID) {
1751 if (namelen != 1)
1752 return (EINVAL);
1753 error = sysctl_wire_old_buffer(req, 0);
1754 if (error)
1755 return (error);
1756 sx_slock(&proctree_lock);
1757 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1758 if (error == 0)
1759 error = sysctl_out_proc(p, req, flags);
1760 sx_sunlock(&proctree_lock);
1761 return (error);
1762 }
1763
1764 switch (oid_number) {
1765 case KERN_PROC_ALL:
1766 if (namelen != 0)
1767 return (EINVAL);
1768 break;
1769 case KERN_PROC_PROC:
1770 if (namelen != 0 && namelen != 1)
1771 return (EINVAL);
1772 break;
1773 default:
1774 if (namelen != 1)
1775 return (EINVAL);
1776 break;
1777 }
1778
1779 if (req->oldptr == NULL) {
1780 /* overestimate by 5 procs */
1781 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1782 if (error)
1783 return (error);
1784 } else {
1785 error = sysctl_wire_old_buffer(req, 0);
1786 if (error != 0)
1787 return (error);
1788 }
1789 iterarg.flags = flags;
1790 iterarg.oid_number = oid_number;
1791 iterarg.req = req;
1792 iterarg.name = name;
1793 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1794 return (error);
1795 }
1796
1797 struct pargs *
pargs_alloc(int len)1798 pargs_alloc(int len)
1799 {
1800 struct pargs *pa;
1801
1802 pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1803 M_WAITOK);
1804 refcount_init(&pa->ar_ref, 1);
1805 pa->ar_length = len;
1806 return (pa);
1807 }
1808
1809 static void
pargs_free(struct pargs * pa)1810 pargs_free(struct pargs *pa)
1811 {
1812
1813 free(pa, M_PARGS);
1814 }
1815
1816 void
pargs_hold(struct pargs * pa)1817 pargs_hold(struct pargs *pa)
1818 {
1819
1820 if (pa == NULL)
1821 return;
1822 refcount_acquire(&pa->ar_ref);
1823 }
1824
1825 void
pargs_drop(struct pargs * pa)1826 pargs_drop(struct pargs *pa)
1827 {
1828
1829 if (pa == NULL)
1830 return;
1831 if (refcount_release(&pa->ar_ref))
1832 pargs_free(pa);
1833 }
1834
1835 static int
proc_read_string(struct thread * td,struct proc * p,const char * sptr,char * buf,size_t len)1836 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1837 size_t len)
1838 {
1839 ssize_t n;
1840
1841 /*
1842 * This may return a short read if the string is shorter than the chunk
1843 * and is aligned at the end of the page, and the following page is not
1844 * mapped.
1845 */
1846 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1847 if (n <= 0)
1848 return (ENOMEM);
1849 return (0);
1850 }
1851
1852 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */
1853
1854 enum proc_vector_type {
1855 PROC_ARG,
1856 PROC_ENV,
1857 PROC_AUX,
1858 };
1859
1860 #ifdef COMPAT_FREEBSD32
1861 static int
get_proc_vector32(struct thread * td,struct proc * p,char *** proc_vectorp,size_t * vsizep,enum proc_vector_type type)1862 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1863 size_t *vsizep, enum proc_vector_type type)
1864 {
1865 struct freebsd32_ps_strings pss;
1866 Elf32_Auxinfo aux;
1867 vm_offset_t vptr, ptr;
1868 uint32_t *proc_vector32;
1869 char **proc_vector;
1870 size_t vsize, size;
1871 int i, error;
1872
1873 error = 0;
1874 if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1875 sizeof(pss))
1876 return (ENOMEM);
1877 switch (type) {
1878 case PROC_ARG:
1879 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1880 vsize = pss.ps_nargvstr;
1881 if (vsize > ARG_MAX)
1882 return (ENOEXEC);
1883 size = vsize * sizeof(int32_t);
1884 break;
1885 case PROC_ENV:
1886 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1887 vsize = pss.ps_nenvstr;
1888 if (vsize > ARG_MAX)
1889 return (ENOEXEC);
1890 size = vsize * sizeof(int32_t);
1891 break;
1892 case PROC_AUX:
1893 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1894 (pss.ps_nenvstr + 1) * sizeof(int32_t);
1895 if (vptr % 4 != 0)
1896 return (ENOEXEC);
1897 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1898 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1899 sizeof(aux))
1900 return (ENOMEM);
1901 if (aux.a_type == AT_NULL)
1902 break;
1903 ptr += sizeof(aux);
1904 }
1905 if (aux.a_type != AT_NULL)
1906 return (ENOEXEC);
1907 vsize = i + 1;
1908 size = vsize * sizeof(aux);
1909 break;
1910 default:
1911 KASSERT(0, ("Wrong proc vector type: %d", type));
1912 return (EINVAL);
1913 }
1914 proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1915 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1916 error = ENOMEM;
1917 goto done;
1918 }
1919 if (type == PROC_AUX) {
1920 *proc_vectorp = (char **)proc_vector32;
1921 *vsizep = vsize;
1922 return (0);
1923 }
1924 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1925 for (i = 0; i < (int)vsize; i++)
1926 proc_vector[i] = PTRIN(proc_vector32[i]);
1927 *proc_vectorp = proc_vector;
1928 *vsizep = vsize;
1929 done:
1930 free(proc_vector32, M_TEMP);
1931 return (error);
1932 }
1933 #endif
1934
1935 static int
get_proc_vector(struct thread * td,struct proc * p,char *** proc_vectorp,size_t * vsizep,enum proc_vector_type type)1936 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1937 size_t *vsizep, enum proc_vector_type type)
1938 {
1939 struct ps_strings pss;
1940 Elf_Auxinfo aux;
1941 vm_offset_t vptr, ptr;
1942 char **proc_vector;
1943 size_t vsize, size;
1944 int i;
1945
1946 #ifdef COMPAT_FREEBSD32
1947 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1948 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1949 #endif
1950 if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1951 sizeof(pss))
1952 return (ENOMEM);
1953 switch (type) {
1954 case PROC_ARG:
1955 vptr = (vm_offset_t)pss.ps_argvstr;
1956 vsize = pss.ps_nargvstr;
1957 if (vsize > ARG_MAX)
1958 return (ENOEXEC);
1959 size = vsize * sizeof(char *);
1960 break;
1961 case PROC_ENV:
1962 vptr = (vm_offset_t)pss.ps_envstr;
1963 vsize = pss.ps_nenvstr;
1964 if (vsize > ARG_MAX)
1965 return (ENOEXEC);
1966 size = vsize * sizeof(char *);
1967 break;
1968 case PROC_AUX:
1969 /*
1970 * The aux array is just above env array on the stack. Check
1971 * that the address is naturally aligned.
1972 */
1973 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1974 * sizeof(char *);
1975 #if __ELF_WORD_SIZE == 64
1976 if (vptr % sizeof(uint64_t) != 0)
1977 #else
1978 if (vptr % sizeof(uint32_t) != 0)
1979 #endif
1980 return (ENOEXEC);
1981 /*
1982 * We count the array size reading the aux vectors from the
1983 * stack until AT_NULL vector is returned. So (to keep the code
1984 * simple) we read the process stack twice: the first time here
1985 * to find the size and the second time when copying the vectors
1986 * to the allocated proc_vector.
1987 */
1988 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1989 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1990 sizeof(aux))
1991 return (ENOMEM);
1992 if (aux.a_type == AT_NULL)
1993 break;
1994 ptr += sizeof(aux);
1995 }
1996 /*
1997 * If the PROC_AUXV_MAX entries are iterated over, and we have
1998 * not reached AT_NULL, it is most likely we are reading wrong
1999 * data: either the process doesn't have auxv array or data has
2000 * been modified. Return the error in this case.
2001 */
2002 if (aux.a_type != AT_NULL)
2003 return (ENOEXEC);
2004 vsize = i + 1;
2005 size = vsize * sizeof(aux);
2006 break;
2007 default:
2008 KASSERT(0, ("Wrong proc vector type: %d", type));
2009 return (EINVAL); /* In case we are built without INVARIANTS. */
2010 }
2011 proc_vector = malloc(size, M_TEMP, M_WAITOK);
2012 if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2013 free(proc_vector, M_TEMP);
2014 return (ENOMEM);
2015 }
2016 *proc_vectorp = proc_vector;
2017 *vsizep = vsize;
2018
2019 return (0);
2020 }
2021
2022 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */
2023
2024 static int
get_ps_strings(struct thread * td,struct proc * p,struct sbuf * sb,enum proc_vector_type type)2025 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2026 enum proc_vector_type type)
2027 {
2028 size_t done, len, nchr, vsize;
2029 int error, i;
2030 char **proc_vector, *sptr;
2031 char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2032
2033 PROC_ASSERT_HELD(p);
2034
2035 /*
2036 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2037 */
2038 nchr = 2 * (PATH_MAX + ARG_MAX);
2039
2040 error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2041 if (error != 0)
2042 return (error);
2043 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2044 /*
2045 * The program may have scribbled into its argv array, e.g. to
2046 * remove some arguments. If that has happened, break out
2047 * before trying to read from NULL.
2048 */
2049 if (proc_vector[i] == NULL)
2050 break;
2051 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2052 error = proc_read_string(td, p, sptr, pss_string,
2053 sizeof(pss_string));
2054 if (error != 0)
2055 goto done;
2056 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2057 if (done + len >= nchr)
2058 len = nchr - done - 1;
2059 sbuf_bcat(sb, pss_string, len);
2060 if (len != GET_PS_STRINGS_CHUNK_SZ)
2061 break;
2062 done += GET_PS_STRINGS_CHUNK_SZ;
2063 }
2064 sbuf_bcat(sb, "", 1);
2065 done += len + 1;
2066 }
2067 done:
2068 free(proc_vector, M_TEMP);
2069 return (error);
2070 }
2071
2072 int
proc_getargv(struct thread * td,struct proc * p,struct sbuf * sb)2073 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2074 {
2075
2076 return (get_ps_strings(curthread, p, sb, PROC_ARG));
2077 }
2078
2079 int
proc_getenvv(struct thread * td,struct proc * p,struct sbuf * sb)2080 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2081 {
2082
2083 return (get_ps_strings(curthread, p, sb, PROC_ENV));
2084 }
2085
2086 int
proc_getauxv(struct thread * td,struct proc * p,struct sbuf * sb)2087 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2088 {
2089 size_t vsize, size;
2090 char **auxv;
2091 int error;
2092
2093 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2094 if (error == 0) {
2095 #ifdef COMPAT_FREEBSD32
2096 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2097 size = vsize * sizeof(Elf32_Auxinfo);
2098 else
2099 #endif
2100 size = vsize * sizeof(Elf_Auxinfo);
2101 if (sbuf_bcat(sb, auxv, size) != 0)
2102 error = ENOMEM;
2103 free(auxv, M_TEMP);
2104 }
2105 return (error);
2106 }
2107
2108 /*
2109 * This sysctl allows a process to retrieve the argument list or process
2110 * title for another process without groping around in the address space
2111 * of the other process. It also allow a process to set its own "process
2112 * title to a string of its own choice.
2113 */
2114 static int
sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)2115 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2116 {
2117 int *name = (int *)arg1;
2118 u_int namelen = arg2;
2119 struct pargs *newpa, *pa;
2120 struct proc *p;
2121 struct sbuf sb;
2122 int flags, error = 0, error2;
2123 pid_t pid;
2124
2125 if (namelen != 1)
2126 return (EINVAL);
2127
2128 p = curproc;
2129 pid = (pid_t)name[0];
2130 if (pid == -1) {
2131 pid = p->p_pid;
2132 }
2133
2134 /*
2135 * If the query is for this process and it is single-threaded, there
2136 * is nobody to modify pargs, thus we can just read.
2137 */
2138 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2139 (pa = p->p_args) != NULL)
2140 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2141
2142 flags = PGET_CANSEE;
2143 if (req->newptr != NULL)
2144 flags |= PGET_ISCURRENT;
2145 error = pget(pid, flags, &p);
2146 if (error)
2147 return (error);
2148
2149 pa = p->p_args;
2150 if (pa != NULL) {
2151 pargs_hold(pa);
2152 PROC_UNLOCK(p);
2153 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2154 pargs_drop(pa);
2155 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2156 _PHOLD(p);
2157 PROC_UNLOCK(p);
2158 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2159 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2160 error = proc_getargv(curthread, p, &sb);
2161 error2 = sbuf_finish(&sb);
2162 PRELE(p);
2163 sbuf_delete(&sb);
2164 if (error == 0 && error2 != 0)
2165 error = error2;
2166 } else {
2167 PROC_UNLOCK(p);
2168 }
2169 if (error != 0 || req->newptr == NULL)
2170 return (error);
2171
2172 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2173 return (ENOMEM);
2174
2175 if (req->newlen == 0) {
2176 /*
2177 * Clear the argument pointer, so that we'll fetch arguments
2178 * with proc_getargv() until further notice.
2179 */
2180 newpa = NULL;
2181 } else {
2182 newpa = pargs_alloc(req->newlen);
2183 error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2184 if (error != 0) {
2185 pargs_free(newpa);
2186 return (error);
2187 }
2188 }
2189 PROC_LOCK(p);
2190 pa = p->p_args;
2191 p->p_args = newpa;
2192 PROC_UNLOCK(p);
2193 pargs_drop(pa);
2194 return (0);
2195 }
2196
2197 /*
2198 * This sysctl allows a process to retrieve environment of another process.
2199 */
2200 static int
sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)2201 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2202 {
2203 int *name = (int *)arg1;
2204 u_int namelen = arg2;
2205 struct proc *p;
2206 struct sbuf sb;
2207 int error, error2;
2208
2209 if (namelen != 1)
2210 return (EINVAL);
2211
2212 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2213 if (error != 0)
2214 return (error);
2215 if ((p->p_flag & P_SYSTEM) != 0) {
2216 PRELE(p);
2217 return (0);
2218 }
2219
2220 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2221 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2222 error = proc_getenvv(curthread, p, &sb);
2223 error2 = sbuf_finish(&sb);
2224 PRELE(p);
2225 sbuf_delete(&sb);
2226 return (error != 0 ? error : error2);
2227 }
2228
2229 /*
2230 * This sysctl allows a process to retrieve ELF auxiliary vector of
2231 * another process.
2232 */
2233 static int
sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)2234 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2235 {
2236 int *name = (int *)arg1;
2237 u_int namelen = arg2;
2238 struct proc *p;
2239 struct sbuf sb;
2240 int error, error2;
2241
2242 if (namelen != 1)
2243 return (EINVAL);
2244
2245 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2246 if (error != 0)
2247 return (error);
2248 if ((p->p_flag & P_SYSTEM) != 0) {
2249 PRELE(p);
2250 return (0);
2251 }
2252 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2253 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2254 error = proc_getauxv(curthread, p, &sb);
2255 error2 = sbuf_finish(&sb);
2256 PRELE(p);
2257 sbuf_delete(&sb);
2258 return (error != 0 ? error : error2);
2259 }
2260
2261 /*
2262 * Look up the canonical executable path running in the specified process.
2263 * It tries to return the same hardlink name as was used for execve(2).
2264 * This allows the programs that modify their behavior based on their progname,
2265 * to operate correctly.
2266 *
2267 * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2268 * calling conventions.
2269 * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2270 * allocated and freed by caller.
2271 * freebuf should be freed by caller, from the M_TEMP malloc type.
2272 */
2273 int
proc_get_binpath(struct proc * p,char * binname,char ** retbuf,char ** freebuf)2274 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2275 char **freebuf)
2276 {
2277 struct nameidata nd;
2278 struct vnode *vp, *dvp;
2279 size_t freepath_size;
2280 int error;
2281 bool do_fullpath;
2282
2283 PROC_LOCK_ASSERT(p, MA_OWNED);
2284
2285 vp = p->p_textvp;
2286 if (vp == NULL) {
2287 PROC_UNLOCK(p);
2288 *retbuf = "";
2289 *freebuf = NULL;
2290 return (0);
2291 }
2292 vref(vp);
2293 dvp = p->p_textdvp;
2294 if (dvp != NULL)
2295 vref(dvp);
2296 if (p->p_binname != NULL)
2297 strlcpy(binname, p->p_binname, MAXPATHLEN);
2298 PROC_UNLOCK(p);
2299
2300 do_fullpath = true;
2301 *freebuf = NULL;
2302 if (dvp != NULL && binname[0] != '\0') {
2303 freepath_size = MAXPATHLEN;
2304 if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2305 retbuf, freebuf, &freepath_size) == 0) {
2306 /*
2307 * Recheck the looked up path. The binary
2308 * might have been renamed or replaced, in
2309 * which case we should not report old name.
2310 */
2311 NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2312 error = namei(&nd);
2313 if (error == 0) {
2314 if (nd.ni_vp == vp)
2315 do_fullpath = false;
2316 vrele(nd.ni_vp);
2317 NDFREE_PNBUF(&nd);
2318 }
2319 }
2320 }
2321 if (do_fullpath) {
2322 free(*freebuf, M_TEMP);
2323 *freebuf = NULL;
2324 error = vn_fullpath(vp, retbuf, freebuf);
2325 }
2326 vrele(vp);
2327 if (dvp != NULL)
2328 vrele(dvp);
2329 return (error);
2330 }
2331
2332 /*
2333 * This sysctl allows a process to retrieve the path of the executable for
2334 * itself or another process.
2335 */
2336 static int
sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)2337 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2338 {
2339 pid_t *pidp = (pid_t *)arg1;
2340 unsigned int arglen = arg2;
2341 struct proc *p;
2342 char *retbuf, *freebuf, *binname;
2343 int error;
2344
2345 if (arglen != 1)
2346 return (EINVAL);
2347 binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2348 binname[0] = '\0';
2349 if (*pidp == -1) { /* -1 means this process */
2350 error = 0;
2351 p = req->td->td_proc;
2352 PROC_LOCK(p);
2353 } else {
2354 error = pget(*pidp, PGET_CANSEE, &p);
2355 }
2356
2357 if (error == 0)
2358 error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2359 free(binname, M_TEMP);
2360 if (error != 0)
2361 return (error);
2362 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2363 free(freebuf, M_TEMP);
2364 return (error);
2365 }
2366
2367 static int
sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)2368 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2369 {
2370 struct proc *p;
2371 char *sv_name;
2372 int *name;
2373 int namelen;
2374 int error;
2375
2376 namelen = arg2;
2377 if (namelen != 1)
2378 return (EINVAL);
2379
2380 name = (int *)arg1;
2381 error = pget((pid_t)name[0], PGET_CANSEE, &p);
2382 if (error != 0)
2383 return (error);
2384 sv_name = p->p_sysent->sv_name;
2385 PROC_UNLOCK(p);
2386 return (sysctl_handle_string(oidp, sv_name, 0, req));
2387 }
2388
2389 #ifdef KINFO_OVMENTRY_SIZE
2390 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2391 #endif
2392
2393 #ifdef COMPAT_FREEBSD7
2394 static int
sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)2395 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2396 {
2397 vm_map_entry_t entry, tmp_entry;
2398 unsigned int last_timestamp, namelen;
2399 char *fullpath, *freepath;
2400 struct kinfo_ovmentry *kve;
2401 struct vattr va;
2402 struct ucred *cred;
2403 int error, *name;
2404 struct vnode *vp;
2405 struct proc *p;
2406 vm_map_t map;
2407 struct vmspace *vm;
2408
2409 namelen = arg2;
2410 if (namelen != 1)
2411 return (EINVAL);
2412
2413 name = (int *)arg1;
2414 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2415 if (error != 0)
2416 return (error);
2417 vm = vmspace_acquire_ref(p);
2418 if (vm == NULL) {
2419 PRELE(p);
2420 return (ESRCH);
2421 }
2422 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2423
2424 map = &vm->vm_map;
2425 vm_map_lock_read(map);
2426 VM_MAP_ENTRY_FOREACH(entry, map) {
2427 vm_object_t obj, tobj, lobj;
2428 vm_offset_t addr;
2429
2430 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2431 continue;
2432
2433 bzero(kve, sizeof(*kve));
2434 kve->kve_structsize = sizeof(*kve);
2435
2436 kve->kve_private_resident = 0;
2437 obj = entry->object.vm_object;
2438 if (obj != NULL) {
2439 VM_OBJECT_RLOCK(obj);
2440 if (obj->shadow_count == 1)
2441 kve->kve_private_resident =
2442 obj->resident_page_count;
2443 }
2444 kve->kve_resident = 0;
2445 addr = entry->start;
2446 while (addr < entry->end) {
2447 if (pmap_extract(map->pmap, addr))
2448 kve->kve_resident++;
2449 addr += PAGE_SIZE;
2450 }
2451
2452 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2453 if (tobj != obj) {
2454 VM_OBJECT_RLOCK(tobj);
2455 kve->kve_offset += tobj->backing_object_offset;
2456 }
2457 if (lobj != obj)
2458 VM_OBJECT_RUNLOCK(lobj);
2459 lobj = tobj;
2460 }
2461
2462 kve->kve_start = (void*)entry->start;
2463 kve->kve_end = (void*)entry->end;
2464 kve->kve_offset += (off_t)entry->offset;
2465
2466 if (entry->protection & VM_PROT_READ)
2467 kve->kve_protection |= KVME_PROT_READ;
2468 if (entry->protection & VM_PROT_WRITE)
2469 kve->kve_protection |= KVME_PROT_WRITE;
2470 if (entry->protection & VM_PROT_EXECUTE)
2471 kve->kve_protection |= KVME_PROT_EXEC;
2472
2473 if (entry->eflags & MAP_ENTRY_COW)
2474 kve->kve_flags |= KVME_FLAG_COW;
2475 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2476 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2477 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2478 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2479
2480 last_timestamp = map->timestamp;
2481 vm_map_unlock_read(map);
2482
2483 kve->kve_fileid = 0;
2484 kve->kve_fsid = 0;
2485 freepath = NULL;
2486 fullpath = "";
2487 if (lobj) {
2488 kve->kve_type = vm_object_kvme_type(lobj, &vp);
2489 if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2490 kve->kve_type = KVME_TYPE_UNKNOWN;
2491 if (vp != NULL)
2492 vref(vp);
2493 if (lobj != obj)
2494 VM_OBJECT_RUNLOCK(lobj);
2495
2496 kve->kve_ref_count = obj->ref_count;
2497 kve->kve_shadow_count = obj->shadow_count;
2498 VM_OBJECT_RUNLOCK(obj);
2499 if (vp != NULL) {
2500 vn_fullpath(vp, &fullpath, &freepath);
2501 cred = curthread->td_ucred;
2502 vn_lock(vp, LK_SHARED | LK_RETRY);
2503 if (VOP_GETATTR(vp, &va, cred) == 0) {
2504 kve->kve_fileid = va.va_fileid;
2505 /* truncate */
2506 kve->kve_fsid = va.va_fsid;
2507 }
2508 vput(vp);
2509 }
2510 } else {
2511 kve->kve_type = KVME_TYPE_NONE;
2512 kve->kve_ref_count = 0;
2513 kve->kve_shadow_count = 0;
2514 }
2515
2516 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2517 if (freepath != NULL)
2518 free(freepath, M_TEMP);
2519
2520 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2521 vm_map_lock_read(map);
2522 if (error)
2523 break;
2524 if (last_timestamp != map->timestamp) {
2525 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2526 entry = tmp_entry;
2527 }
2528 }
2529 vm_map_unlock_read(map);
2530 vmspace_free(vm);
2531 PRELE(p);
2532 free(kve, M_TEMP);
2533 return (error);
2534 }
2535 #endif /* COMPAT_FREEBSD7 */
2536
2537 #ifdef KINFO_VMENTRY_SIZE
2538 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2539 #endif
2540
2541 void
kern_proc_vmmap_resident(vm_map_t map,vm_map_entry_t entry,int * resident_count,bool * super)2542 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2543 int *resident_count, bool *super)
2544 {
2545 vm_object_t obj, tobj;
2546 vm_page_t m, m_adv;
2547 vm_offset_t addr;
2548 vm_paddr_t pa;
2549 vm_pindex_t pi, pi_adv, pindex;
2550 int incore;
2551
2552 *super = false;
2553 *resident_count = 0;
2554 if (vmmap_skip_res_cnt)
2555 return;
2556
2557 pa = 0;
2558 obj = entry->object.vm_object;
2559 addr = entry->start;
2560 m_adv = NULL;
2561 pi = OFF_TO_IDX(entry->offset);
2562 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2563 if (m_adv != NULL) {
2564 m = m_adv;
2565 } else {
2566 pi_adv = atop(entry->end - addr);
2567 pindex = pi;
2568 for (tobj = obj;; tobj = tobj->backing_object) {
2569 m = vm_radix_lookup_ge(&tobj->rtree, pindex);
2570 if (m != NULL) {
2571 if (m->pindex == pindex)
2572 break;
2573 if (pi_adv > m->pindex - pindex) {
2574 pi_adv = m->pindex - pindex;
2575 m_adv = m;
2576 }
2577 }
2578 if (tobj->backing_object == NULL)
2579 goto next;
2580 pindex += OFF_TO_IDX(tobj->
2581 backing_object_offset);
2582 }
2583 }
2584 m_adv = NULL;
2585 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2586 (addr & (pagesizes[1] - 1)) == 0 && (incore =
2587 pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2588 *super = true;
2589 /*
2590 * The virtual page might be smaller than the physical
2591 * page, so we use the page size reported by the pmap
2592 * rather than m->psind.
2593 */
2594 pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2595 } else {
2596 /*
2597 * We do not test the found page on validity.
2598 * Either the page is busy and being paged in,
2599 * or it was invalidated. The first case
2600 * should be counted as resident, the second
2601 * is not so clear; we do account both.
2602 */
2603 pi_adv = 1;
2604 }
2605 *resident_count += pi_adv;
2606 next:;
2607 }
2608 }
2609
2610 /*
2611 * Must be called with the process locked and will return unlocked.
2612 */
2613 int
kern_proc_vmmap_out(struct proc * p,struct sbuf * sb,ssize_t maxlen,int flags)2614 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2615 {
2616 vm_map_entry_t entry, tmp_entry;
2617 struct vattr va;
2618 vm_map_t map;
2619 vm_object_t lobj, nobj, obj, tobj;
2620 char *fullpath, *freepath;
2621 struct kinfo_vmentry *kve;
2622 struct ucred *cred;
2623 struct vnode *vp;
2624 struct vmspace *vm;
2625 vm_offset_t addr;
2626 unsigned int last_timestamp;
2627 int error;
2628 key_t key;
2629 unsigned short seq;
2630 bool guard, super;
2631
2632 PROC_LOCK_ASSERT(p, MA_OWNED);
2633
2634 _PHOLD(p);
2635 PROC_UNLOCK(p);
2636 vm = vmspace_acquire_ref(p);
2637 if (vm == NULL) {
2638 PRELE(p);
2639 return (ESRCH);
2640 }
2641 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2642
2643 error = 0;
2644 map = &vm->vm_map;
2645 vm_map_lock_read(map);
2646 VM_MAP_ENTRY_FOREACH(entry, map) {
2647 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2648 continue;
2649
2650 addr = entry->end;
2651 bzero(kve, sizeof(*kve));
2652 obj = entry->object.vm_object;
2653 if (obj != NULL) {
2654 if ((obj->flags & OBJ_ANON) != 0)
2655 kve->kve_obj = (uintptr_t)obj;
2656
2657 for (tobj = obj; tobj != NULL;
2658 tobj = tobj->backing_object) {
2659 VM_OBJECT_RLOCK(tobj);
2660 kve->kve_offset += tobj->backing_object_offset;
2661 lobj = tobj;
2662 }
2663 if (obj->backing_object == NULL)
2664 kve->kve_private_resident =
2665 obj->resident_page_count;
2666 kern_proc_vmmap_resident(map, entry,
2667 &kve->kve_resident, &super);
2668 if (super)
2669 kve->kve_flags |= KVME_FLAG_SUPER;
2670 for (tobj = obj; tobj != NULL; tobj = nobj) {
2671 nobj = tobj->backing_object;
2672 if (tobj != obj && tobj != lobj)
2673 VM_OBJECT_RUNLOCK(tobj);
2674 }
2675 } else {
2676 lobj = NULL;
2677 }
2678
2679 kve->kve_start = entry->start;
2680 kve->kve_end = entry->end;
2681 kve->kve_offset += entry->offset;
2682
2683 if (entry->protection & VM_PROT_READ)
2684 kve->kve_protection |= KVME_PROT_READ;
2685 if (entry->protection & VM_PROT_WRITE)
2686 kve->kve_protection |= KVME_PROT_WRITE;
2687 if (entry->protection & VM_PROT_EXECUTE)
2688 kve->kve_protection |= KVME_PROT_EXEC;
2689 if (entry->max_protection & VM_PROT_READ)
2690 kve->kve_protection |= KVME_MAX_PROT_READ;
2691 if (entry->max_protection & VM_PROT_WRITE)
2692 kve->kve_protection |= KVME_MAX_PROT_WRITE;
2693 if (entry->max_protection & VM_PROT_EXECUTE)
2694 kve->kve_protection |= KVME_MAX_PROT_EXEC;
2695
2696 if (entry->eflags & MAP_ENTRY_COW)
2697 kve->kve_flags |= KVME_FLAG_COW;
2698 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2699 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2700 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2701 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2702 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2703 kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2704 if (entry->eflags & MAP_ENTRY_USER_WIRED)
2705 kve->kve_flags |= KVME_FLAG_USER_WIRED;
2706
2707 guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2708
2709 last_timestamp = map->timestamp;
2710 vm_map_unlock_read(map);
2711
2712 freepath = NULL;
2713 fullpath = "";
2714 if (lobj != NULL) {
2715 kve->kve_type = vm_object_kvme_type(lobj, &vp);
2716 if (vp != NULL)
2717 vref(vp);
2718 if (lobj != obj)
2719 VM_OBJECT_RUNLOCK(lobj);
2720
2721 kve->kve_ref_count = obj->ref_count;
2722 kve->kve_shadow_count = obj->shadow_count;
2723 if (obj->type == OBJT_DEVICE ||
2724 obj->type == OBJT_MGTDEVICE) {
2725 cdev_pager_get_path(obj, kve->kve_path,
2726 sizeof(kve->kve_path));
2727 }
2728 VM_OBJECT_RUNLOCK(obj);
2729 if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2730 kve->kve_flags |= KVME_FLAG_SYSVSHM;
2731 shmobjinfo(lobj, &key, &seq);
2732 kve->kve_vn_fileid = key;
2733 kve->kve_vn_fsid_freebsd11 = seq;
2734 }
2735 if ((lobj->flags & OBJ_POSIXSHM) != 0) {
2736 kve->kve_flags |= KVME_FLAG_POSIXSHM;
2737 shm_get_path(lobj, kve->kve_path,
2738 sizeof(kve->kve_path));
2739 }
2740 if (vp != NULL) {
2741 vn_fullpath(vp, &fullpath, &freepath);
2742 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2743 cred = curthread->td_ucred;
2744 vn_lock(vp, LK_SHARED | LK_RETRY);
2745 if (VOP_GETATTR(vp, &va, cred) == 0) {
2746 kve->kve_vn_fileid = va.va_fileid;
2747 kve->kve_vn_fsid = va.va_fsid;
2748 kve->kve_vn_fsid_freebsd11 =
2749 kve->kve_vn_fsid; /* truncate */
2750 kve->kve_vn_mode =
2751 MAKEIMODE(va.va_type, va.va_mode);
2752 kve->kve_vn_size = va.va_size;
2753 kve->kve_vn_rdev = va.va_rdev;
2754 kve->kve_vn_rdev_freebsd11 =
2755 kve->kve_vn_rdev; /* truncate */
2756 kve->kve_status = KF_ATTR_VALID;
2757 }
2758 vput(vp);
2759 strlcpy(kve->kve_path, fullpath, sizeof(
2760 kve->kve_path));
2761 free(freepath, M_TEMP);
2762 }
2763 } else {
2764 kve->kve_type = guard ? KVME_TYPE_GUARD :
2765 KVME_TYPE_NONE;
2766 kve->kve_ref_count = 0;
2767 kve->kve_shadow_count = 0;
2768 }
2769
2770 /* Pack record size down */
2771 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2772 kve->kve_structsize =
2773 offsetof(struct kinfo_vmentry, kve_path) +
2774 strlen(kve->kve_path) + 1;
2775 else
2776 kve->kve_structsize = sizeof(*kve);
2777 kve->kve_structsize = roundup(kve->kve_structsize,
2778 sizeof(uint64_t));
2779
2780 /* Halt filling and truncate rather than exceeding maxlen */
2781 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2782 error = 0;
2783 vm_map_lock_read(map);
2784 break;
2785 } else if (maxlen != -1)
2786 maxlen -= kve->kve_structsize;
2787
2788 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2789 error = ENOMEM;
2790 vm_map_lock_read(map);
2791 if (error != 0)
2792 break;
2793 if (last_timestamp != map->timestamp) {
2794 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2795 entry = tmp_entry;
2796 }
2797 }
2798 vm_map_unlock_read(map);
2799 vmspace_free(vm);
2800 PRELE(p);
2801 free(kve, M_TEMP);
2802 return (error);
2803 }
2804
2805 static int
sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)2806 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2807 {
2808 struct proc *p;
2809 struct sbuf sb;
2810 u_int namelen;
2811 int error, error2, *name;
2812
2813 namelen = arg2;
2814 if (namelen != 1)
2815 return (EINVAL);
2816
2817 name = (int *)arg1;
2818 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2819 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2820 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2821 if (error != 0) {
2822 sbuf_delete(&sb);
2823 return (error);
2824 }
2825 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2826 error2 = sbuf_finish(&sb);
2827 sbuf_delete(&sb);
2828 return (error != 0 ? error : error2);
2829 }
2830
2831 #if defined(STACK) || defined(DDB)
2832 static int
sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)2833 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2834 {
2835 struct kinfo_kstack *kkstp;
2836 int error, i, *name, numthreads;
2837 lwpid_t *lwpidarray;
2838 struct thread *td;
2839 struct stack *st;
2840 struct sbuf sb;
2841 struct proc *p;
2842 u_int namelen;
2843
2844 namelen = arg2;
2845 if (namelen != 1)
2846 return (EINVAL);
2847
2848 name = (int *)arg1;
2849 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2850 if (error != 0)
2851 return (error);
2852
2853 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2854 st = stack_create(M_WAITOK);
2855
2856 lwpidarray = NULL;
2857 PROC_LOCK(p);
2858 do {
2859 if (lwpidarray != NULL) {
2860 free(lwpidarray, M_TEMP);
2861 lwpidarray = NULL;
2862 }
2863 numthreads = p->p_numthreads;
2864 PROC_UNLOCK(p);
2865 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2866 M_WAITOK | M_ZERO);
2867 PROC_LOCK(p);
2868 } while (numthreads < p->p_numthreads);
2869
2870 /*
2871 * XXXRW: During the below loop, execve(2) and countless other sorts
2872 * of changes could have taken place. Should we check to see if the
2873 * vmspace has been replaced, or the like, in order to prevent
2874 * giving a snapshot that spans, say, execve(2), with some threads
2875 * before and some after? Among other things, the credentials could
2876 * have changed, in which case the right to extract debug info might
2877 * no longer be assured.
2878 */
2879 i = 0;
2880 FOREACH_THREAD_IN_PROC(p, td) {
2881 KASSERT(i < numthreads,
2882 ("sysctl_kern_proc_kstack: numthreads"));
2883 lwpidarray[i] = td->td_tid;
2884 i++;
2885 }
2886 PROC_UNLOCK(p);
2887 numthreads = i;
2888 for (i = 0; i < numthreads; i++) {
2889 td = tdfind(lwpidarray[i], p->p_pid);
2890 if (td == NULL) {
2891 continue;
2892 }
2893 bzero(kkstp, sizeof(*kkstp));
2894 (void)sbuf_new(&sb, kkstp->kkst_trace,
2895 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2896 thread_lock(td);
2897 kkstp->kkst_tid = td->td_tid;
2898 if (stack_save_td(st, td) == 0)
2899 kkstp->kkst_state = KKST_STATE_STACKOK;
2900 else
2901 kkstp->kkst_state = KKST_STATE_RUNNING;
2902 thread_unlock(td);
2903 PROC_UNLOCK(p);
2904 stack_sbuf_print(&sb, st);
2905 sbuf_finish(&sb);
2906 sbuf_delete(&sb);
2907 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2908 if (error)
2909 break;
2910 }
2911 PRELE(p);
2912 if (lwpidarray != NULL)
2913 free(lwpidarray, M_TEMP);
2914 stack_destroy(st);
2915 free(kkstp, M_TEMP);
2916 return (error);
2917 }
2918 #endif
2919
2920 /*
2921 * This sysctl allows a process to retrieve the full list of groups from
2922 * itself or another process.
2923 */
2924 static int
sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)2925 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2926 {
2927 pid_t *pidp = (pid_t *)arg1;
2928 unsigned int arglen = arg2;
2929 struct proc *p;
2930 struct ucred *cred;
2931 int error;
2932
2933 if (arglen != 1)
2934 return (EINVAL);
2935 if (*pidp == -1) { /* -1 means this process */
2936 p = req->td->td_proc;
2937 PROC_LOCK(p);
2938 } else {
2939 error = pget(*pidp, PGET_CANSEE, &p);
2940 if (error != 0)
2941 return (error);
2942 }
2943
2944 cred = crhold(p->p_ucred);
2945 PROC_UNLOCK(p);
2946
2947 error = SYSCTL_OUT(req, &cred->cr_gid, sizeof(gid_t));
2948 if (error == 0)
2949 error = SYSCTL_OUT(req, cred->cr_groups,
2950 cred->cr_ngroups * sizeof(gid_t));
2951
2952 crfree(cred);
2953 return (error);
2954 }
2955
2956 /*
2957 * This sysctl allows a process to retrieve or/and set the resource limit for
2958 * another process.
2959 */
2960 static int
sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)2961 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2962 {
2963 int *name = (int *)arg1;
2964 u_int namelen = arg2;
2965 struct rlimit rlim;
2966 struct proc *p;
2967 u_int which;
2968 int flags, error;
2969
2970 if (namelen != 2)
2971 return (EINVAL);
2972
2973 which = (u_int)name[1];
2974 if (which >= RLIM_NLIMITS)
2975 return (EINVAL);
2976
2977 if (req->newptr != NULL && req->newlen != sizeof(rlim))
2978 return (EINVAL);
2979
2980 flags = PGET_HOLD | PGET_NOTWEXIT;
2981 if (req->newptr != NULL)
2982 flags |= PGET_CANDEBUG;
2983 else
2984 flags |= PGET_CANSEE;
2985 error = pget((pid_t)name[0], flags, &p);
2986 if (error != 0)
2987 return (error);
2988
2989 /*
2990 * Retrieve limit.
2991 */
2992 if (req->oldptr != NULL) {
2993 PROC_LOCK(p);
2994 lim_rlimit_proc(p, which, &rlim);
2995 PROC_UNLOCK(p);
2996 }
2997 error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2998 if (error != 0)
2999 goto errout;
3000
3001 /*
3002 * Set limit.
3003 */
3004 if (req->newptr != NULL) {
3005 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
3006 if (error == 0)
3007 error = kern_proc_setrlimit(curthread, p, which, &rlim);
3008 }
3009
3010 errout:
3011 PRELE(p);
3012 return (error);
3013 }
3014
3015 /*
3016 * This sysctl allows a process to retrieve ps_strings structure location of
3017 * another process.
3018 */
3019 static int
sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)3020 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
3021 {
3022 int *name = (int *)arg1;
3023 u_int namelen = arg2;
3024 struct proc *p;
3025 vm_offset_t ps_strings;
3026 int error;
3027 #ifdef COMPAT_FREEBSD32
3028 uint32_t ps_strings32;
3029 #endif
3030
3031 if (namelen != 1)
3032 return (EINVAL);
3033
3034 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3035 if (error != 0)
3036 return (error);
3037 #ifdef COMPAT_FREEBSD32
3038 if ((req->flags & SCTL_MASK32) != 0) {
3039 /*
3040 * We return 0 if the 32 bit emulation request is for a 64 bit
3041 * process.
3042 */
3043 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3044 PTROUT(PROC_PS_STRINGS(p)) : 0;
3045 PROC_UNLOCK(p);
3046 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3047 return (error);
3048 }
3049 #endif
3050 ps_strings = PROC_PS_STRINGS(p);
3051 PROC_UNLOCK(p);
3052 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3053 return (error);
3054 }
3055
3056 /*
3057 * This sysctl allows a process to retrieve umask of another process.
3058 */
3059 static int
sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)3060 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3061 {
3062 int *name = (int *)arg1;
3063 u_int namelen = arg2;
3064 struct proc *p;
3065 int error;
3066 u_short cmask;
3067 pid_t pid;
3068
3069 if (namelen != 1)
3070 return (EINVAL);
3071
3072 pid = (pid_t)name[0];
3073 p = curproc;
3074 if (pid == p->p_pid || pid == 0) {
3075 cmask = p->p_pd->pd_cmask;
3076 goto out;
3077 }
3078
3079 error = pget(pid, PGET_WANTREAD, &p);
3080 if (error != 0)
3081 return (error);
3082
3083 cmask = p->p_pd->pd_cmask;
3084 PRELE(p);
3085 out:
3086 error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3087 return (error);
3088 }
3089
3090 /*
3091 * This sysctl allows a process to set and retrieve binary osreldate of
3092 * another process.
3093 */
3094 static int
sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)3095 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3096 {
3097 int *name = (int *)arg1;
3098 u_int namelen = arg2;
3099 struct proc *p;
3100 int flags, error, osrel;
3101
3102 if (namelen != 1)
3103 return (EINVAL);
3104
3105 if (req->newptr != NULL && req->newlen != sizeof(osrel))
3106 return (EINVAL);
3107
3108 flags = PGET_HOLD | PGET_NOTWEXIT;
3109 if (req->newptr != NULL)
3110 flags |= PGET_CANDEBUG;
3111 else
3112 flags |= PGET_CANSEE;
3113 error = pget((pid_t)name[0], flags, &p);
3114 if (error != 0)
3115 return (error);
3116
3117 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3118 if (error != 0)
3119 goto errout;
3120
3121 if (req->newptr != NULL) {
3122 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3123 if (error != 0)
3124 goto errout;
3125 if (osrel < 0) {
3126 error = EINVAL;
3127 goto errout;
3128 }
3129 p->p_osrel = osrel;
3130 }
3131 errout:
3132 PRELE(p);
3133 return (error);
3134 }
3135
3136 static int
sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)3137 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3138 {
3139 int *name = (int *)arg1;
3140 u_int namelen = arg2;
3141 struct proc *p;
3142 struct kinfo_sigtramp kst;
3143 const struct sysentvec *sv;
3144 int error;
3145 #ifdef COMPAT_FREEBSD32
3146 struct kinfo_sigtramp32 kst32;
3147 #endif
3148
3149 if (namelen != 1)
3150 return (EINVAL);
3151
3152 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3153 if (error != 0)
3154 return (error);
3155 sv = p->p_sysent;
3156 #ifdef COMPAT_FREEBSD32
3157 if ((req->flags & SCTL_MASK32) != 0) {
3158 bzero(&kst32, sizeof(kst32));
3159 if (SV_PROC_FLAG(p, SV_ILP32)) {
3160 if (PROC_HAS_SHP(p)) {
3161 kst32.ksigtramp_start = PROC_SIGCODE(p);
3162 kst32.ksigtramp_end = kst32.ksigtramp_start +
3163 ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3164 *sv->sv_szsigcode :
3165 (uintptr_t)sv->sv_szsigcode);
3166 } else {
3167 kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3168 *sv->sv_szsigcode;
3169 kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3170 }
3171 }
3172 PROC_UNLOCK(p);
3173 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3174 return (error);
3175 }
3176 #endif
3177 bzero(&kst, sizeof(kst));
3178 if (PROC_HAS_SHP(p)) {
3179 kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3180 kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3181 ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3182 (uintptr_t)sv->sv_szsigcode);
3183 } else {
3184 kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3185 *sv->sv_szsigcode;
3186 kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3187 }
3188 PROC_UNLOCK(p);
3189 error = SYSCTL_OUT(req, &kst, sizeof(kst));
3190 return (error);
3191 }
3192
3193 static int
sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)3194 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3195 {
3196 int *name = (int *)arg1;
3197 u_int namelen = arg2;
3198 pid_t pid;
3199 struct proc *p;
3200 struct thread *td1;
3201 uintptr_t addr;
3202 #ifdef COMPAT_FREEBSD32
3203 uint32_t addr32;
3204 #endif
3205 int error;
3206
3207 if (namelen != 1 || req->newptr != NULL)
3208 return (EINVAL);
3209
3210 pid = (pid_t)name[0];
3211 error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3212 if (error != 0)
3213 return (error);
3214
3215 PROC_LOCK(p);
3216 #ifdef COMPAT_FREEBSD32
3217 if (SV_CURPROC_FLAG(SV_ILP32)) {
3218 if (!SV_PROC_FLAG(p, SV_ILP32)) {
3219 error = EINVAL;
3220 goto errlocked;
3221 }
3222 }
3223 #endif
3224 if (pid <= PID_MAX) {
3225 td1 = FIRST_THREAD_IN_PROC(p);
3226 } else {
3227 FOREACH_THREAD_IN_PROC(p, td1) {
3228 if (td1->td_tid == pid)
3229 break;
3230 }
3231 }
3232 if (td1 == NULL) {
3233 error = ESRCH;
3234 goto errlocked;
3235 }
3236 /*
3237 * The access to the private thread flags. It is fine as far
3238 * as no out-of-thin-air values are read from td_pflags, and
3239 * usermode read of the td_sigblock_ptr is racy inherently,
3240 * since target process might have already changed it
3241 * meantime.
3242 */
3243 if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3244 addr = (uintptr_t)td1->td_sigblock_ptr;
3245 else
3246 error = ENOTTY;
3247
3248 errlocked:
3249 _PRELE(p);
3250 PROC_UNLOCK(p);
3251 if (error != 0)
3252 return (error);
3253
3254 #ifdef COMPAT_FREEBSD32
3255 if (SV_CURPROC_FLAG(SV_ILP32)) {
3256 addr32 = addr;
3257 error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3258 } else
3259 #endif
3260 error = SYSCTL_OUT(req, &addr, sizeof(addr));
3261 return (error);
3262 }
3263
3264 static int
sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)3265 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3266 {
3267 struct kinfo_vm_layout kvm;
3268 struct proc *p;
3269 struct vmspace *vmspace;
3270 int error, *name;
3271
3272 name = (int *)arg1;
3273 if ((u_int)arg2 != 1)
3274 return (EINVAL);
3275
3276 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3277 if (error != 0)
3278 return (error);
3279 #ifdef COMPAT_FREEBSD32
3280 if (SV_CURPROC_FLAG(SV_ILP32)) {
3281 if (!SV_PROC_FLAG(p, SV_ILP32)) {
3282 PROC_UNLOCK(p);
3283 return (EINVAL);
3284 }
3285 }
3286 #endif
3287 vmspace = vmspace_acquire_ref(p);
3288 PROC_UNLOCK(p);
3289
3290 memset(&kvm, 0, sizeof(kvm));
3291 kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3292 kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3293 kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3294 kvm.kvm_text_size = vmspace->vm_tsize;
3295 kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3296 kvm.kvm_data_size = vmspace->vm_dsize;
3297 kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3298 kvm.kvm_stack_size = vmspace->vm_ssize;
3299 kvm.kvm_shp_addr = vmspace->vm_shp_base;
3300 kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3301 if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3302 kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3303 if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3304 kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3305 if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3306 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3307 if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3308 kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3309 if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3310 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3311 if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3312 PROC_HAS_SHP(p))
3313 kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3314
3315 #ifdef COMPAT_FREEBSD32
3316 if (SV_CURPROC_FLAG(SV_ILP32)) {
3317 struct kinfo_vm_layout32 kvm32;
3318
3319 memset(&kvm32, 0, sizeof(kvm32));
3320 kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3321 kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3322 kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3323 kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3324 kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3325 kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3326 kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3327 kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3328 kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3329 kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3330 kvm32.kvm_map_flags = kvm.kvm_map_flags;
3331 error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3332 goto out;
3333 }
3334 #endif
3335
3336 error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3337 #ifdef COMPAT_FREEBSD32
3338 out:
3339 #endif
3340 vmspace_free(vmspace);
3341 return (error);
3342 }
3343
3344 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
3345 "Process table");
3346
3347 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3348 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3349 "Return entire process table");
3350
3351 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3352 sysctl_kern_proc, "Process table");
3353
3354 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3355 sysctl_kern_proc, "Process table");
3356
3357 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3358 sysctl_kern_proc, "Process table");
3359
3360 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3361 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3362
3363 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3364 sysctl_kern_proc, "Process table");
3365
3366 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3367 sysctl_kern_proc, "Process table");
3368
3369 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3370 sysctl_kern_proc, "Process table");
3371
3372 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3373 sysctl_kern_proc, "Process table");
3374
3375 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3376 sysctl_kern_proc, "Return process table, no threads");
3377
3378 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3379 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3380 sysctl_kern_proc_args, "Process argument list");
3381
3382 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3383 sysctl_kern_proc_env, "Process environment");
3384
3385 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3386 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3387
3388 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3389 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3390
3391 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3392 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3393 "Process syscall vector name (ABI type)");
3394
3395 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3396 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3397
3398 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3399 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3400
3401 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3402 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3403
3404 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3405 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3406
3407 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3408 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3409
3410 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3411 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3412
3413 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3414 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3415
3416 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3417 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3418
3419 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3420 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3421 "Return process table, including threads");
3422
3423 #ifdef COMPAT_FREEBSD7
3424 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3425 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3426 #endif
3427
3428 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3429 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3430
3431 #if defined(STACK) || defined(DDB)
3432 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3433 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3434 #endif
3435
3436 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3437 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3438
3439 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3440 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3441 "Process resource limits");
3442
3443 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3444 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3445 "Process ps_strings location");
3446
3447 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3448 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3449
3450 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3451 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3452 "Process binary osreldate");
3453
3454 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3455 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3456 "Process signal trampoline location");
3457
3458 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3459 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3460 "Thread sigfastblock address");
3461
3462 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3463 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3464 "Process virtual address space layout info");
3465
3466 static struct sx stop_all_proc_blocker;
3467 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3468
3469 bool
stop_all_proc_block(void)3470 stop_all_proc_block(void)
3471 {
3472 return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3473 }
3474
3475 void
stop_all_proc_unblock(void)3476 stop_all_proc_unblock(void)
3477 {
3478 sx_xunlock(&stop_all_proc_blocker);
3479 }
3480
3481 int allproc_gen;
3482
3483 /*
3484 * stop_all_proc() purpose is to stop all process which have usermode,
3485 * except current process for obvious reasons. This makes it somewhat
3486 * unreliable when invoked from multithreaded process. The service
3487 * must not be user-callable anyway.
3488 */
3489 void
stop_all_proc(void)3490 stop_all_proc(void)
3491 {
3492 struct proc *cp, *p;
3493 int r, gen;
3494 bool restart, seen_stopped, seen_exiting, stopped_some;
3495
3496 if (!stop_all_proc_block())
3497 return;
3498
3499 cp = curproc;
3500 allproc_loop:
3501 sx_xlock(&allproc_lock);
3502 gen = allproc_gen;
3503 seen_exiting = seen_stopped = stopped_some = restart = false;
3504 LIST_REMOVE(cp, p_list);
3505 LIST_INSERT_HEAD(&allproc, cp, p_list);
3506 for (;;) {
3507 p = LIST_NEXT(cp, p_list);
3508 if (p == NULL)
3509 break;
3510 LIST_REMOVE(cp, p_list);
3511 LIST_INSERT_AFTER(p, cp, p_list);
3512 PROC_LOCK(p);
3513 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3514 P_STOPPED_SIG)) != 0) {
3515 PROC_UNLOCK(p);
3516 continue;
3517 }
3518 if ((p->p_flag2 & P2_WEXIT) != 0) {
3519 seen_exiting = true;
3520 PROC_UNLOCK(p);
3521 continue;
3522 }
3523 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3524 /*
3525 * Stopped processes are tolerated when there
3526 * are no other processes which might continue
3527 * them. P_STOPPED_SINGLE but not
3528 * P_TOTAL_STOP process still has at least one
3529 * thread running.
3530 */
3531 seen_stopped = true;
3532 PROC_UNLOCK(p);
3533 continue;
3534 }
3535 if ((p->p_flag & P_TRACED) != 0) {
3536 /*
3537 * thread_single() below cannot stop traced p,
3538 * so skip it. OTOH, we cannot require
3539 * restart because debugger might be either
3540 * already stopped or traced as well.
3541 */
3542 PROC_UNLOCK(p);
3543 continue;
3544 }
3545 sx_xunlock(&allproc_lock);
3546 _PHOLD(p);
3547 r = thread_single(p, SINGLE_ALLPROC);
3548 if (r != 0)
3549 restart = true;
3550 else
3551 stopped_some = true;
3552 _PRELE(p);
3553 PROC_UNLOCK(p);
3554 sx_xlock(&allproc_lock);
3555 }
3556 /* Catch forked children we did not see in iteration. */
3557 if (gen != allproc_gen)
3558 restart = true;
3559 sx_xunlock(&allproc_lock);
3560 if (restart || stopped_some || seen_exiting || seen_stopped) {
3561 kern_yield(PRI_USER);
3562 goto allproc_loop;
3563 }
3564 }
3565
3566 void
resume_all_proc(void)3567 resume_all_proc(void)
3568 {
3569 struct proc *cp, *p;
3570
3571 cp = curproc;
3572 sx_xlock(&allproc_lock);
3573 again:
3574 LIST_REMOVE(cp, p_list);
3575 LIST_INSERT_HEAD(&allproc, cp, p_list);
3576 for (;;) {
3577 p = LIST_NEXT(cp, p_list);
3578 if (p == NULL)
3579 break;
3580 LIST_REMOVE(cp, p_list);
3581 LIST_INSERT_AFTER(p, cp, p_list);
3582 PROC_LOCK(p);
3583 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3584 sx_xunlock(&allproc_lock);
3585 _PHOLD(p);
3586 thread_single_end(p, SINGLE_ALLPROC);
3587 _PRELE(p);
3588 PROC_UNLOCK(p);
3589 sx_xlock(&allproc_lock);
3590 } else {
3591 PROC_UNLOCK(p);
3592 }
3593 }
3594 /* Did the loop above missed any stopped process ? */
3595 FOREACH_PROC_IN_SYSTEM(p) {
3596 /* No need for proc lock. */
3597 if ((p->p_flag & P_TOTAL_STOP) != 0)
3598 goto again;
3599 }
3600 sx_xunlock(&allproc_lock);
3601
3602 stop_all_proc_unblock();
3603 }
3604
3605 /* #define TOTAL_STOP_DEBUG 1 */
3606 #ifdef TOTAL_STOP_DEBUG
3607 volatile static int ap_resume;
3608 #include <sys/mount.h>
3609
3610 static int
sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)3611 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3612 {
3613 int error, val;
3614
3615 val = 0;
3616 ap_resume = 0;
3617 error = sysctl_handle_int(oidp, &val, 0, req);
3618 if (error != 0 || req->newptr == NULL)
3619 return (error);
3620 if (val != 0) {
3621 stop_all_proc();
3622 syncer_suspend();
3623 while (ap_resume == 0)
3624 ;
3625 syncer_resume();
3626 resume_all_proc();
3627 }
3628 return (0);
3629 }
3630
3631 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3632 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3633 sysctl_debug_stop_all_proc, "I",
3634 "");
3635 #endif
3636