FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_proc.c
1 /*-
2 * Copyright (c) 1982, 1986, 1989, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD: releng/10.3/sys/kern/kern_proc.c 295454 2016-02-10 00:08:51Z jhb $");
34
35 #include "opt_compat.h"
36 #include "opt_ddb.h"
37 #include "opt_kdtrace.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/elf.h>
45 #include <sys/exec.h>
46 #include <sys/kernel.h>
47 #include <sys/limits.h>
48 #include <sys/lock.h>
49 #include <sys/loginclass.h>
50 #include <sys/malloc.h>
51 #include <sys/mman.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/proc.h>
55 #include <sys/ptrace.h>
56 #include <sys/refcount.h>
57 #include <sys/resourcevar.h>
58 #include <sys/rwlock.h>
59 #include <sys/sbuf.h>
60 #include <sys/sysent.h>
61 #include <sys/sched.h>
62 #include <sys/smp.h>
63 #include <sys/stack.h>
64 #include <sys/stat.h>
65 #include <sys/sysctl.h>
66 #include <sys/filedesc.h>
67 #include <sys/tty.h>
68 #include <sys/signalvar.h>
69 #include <sys/sdt.h>
70 #include <sys/sx.h>
71 #include <sys/user.h>
72 #include <sys/jail.h>
73 #include <sys/vnode.h>
74 #include <sys/eventhandler.h>
75
76 #ifdef DDB
77 #include <ddb/ddb.h>
78 #endif
79
80 #include <vm/vm.h>
81 #include <vm/vm_param.h>
82 #include <vm/vm_extern.h>
83 #include <vm/pmap.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88
89 #ifdef COMPAT_FREEBSD32
90 #include <compat/freebsd32/freebsd32.h>
91 #include <compat/freebsd32/freebsd32_util.h>
92 #endif
93
94 SDT_PROVIDER_DEFINE(proc);
95 SDT_PROBE_DEFINE4(proc, kernel, ctor, entry, "struct proc *", "int",
96 "void *", "int");
97 SDT_PROBE_DEFINE4(proc, kernel, ctor, return, "struct proc *", "int",
98 "void *", "int");
99 SDT_PROBE_DEFINE4(proc, kernel, dtor, entry, "struct proc *", "int",
100 "void *", "struct thread *");
101 SDT_PROBE_DEFINE3(proc, kernel, dtor, return, "struct proc *", "int",
102 "void *");
103 SDT_PROBE_DEFINE3(proc, kernel, init, entry, "struct proc *", "int",
104 "int");
105 SDT_PROBE_DEFINE3(proc, kernel, init, return, "struct proc *", "int",
106 "int");
107
108 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
109 MALLOC_DEFINE(M_SESSION, "session", "session header");
110 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
111 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
112
113 static void doenterpgrp(struct proc *, struct pgrp *);
114 static void orphanpg(struct pgrp *pg);
115 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
116 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
117 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
118 int preferthread);
119 static void pgadjustjobc(struct pgrp *pgrp, int entering);
120 static void pgdelete(struct pgrp *);
121 static int proc_ctor(void *mem, int size, void *arg, int flags);
122 static void proc_dtor(void *mem, int size, void *arg);
123 static int proc_init(void *mem, int size, int flags);
124 static void proc_fini(void *mem, int size);
125 static void pargs_free(struct pargs *pa);
126 static struct proc *zpfind_locked(pid_t pid);
127
128 /*
129 * Other process lists
130 */
131 struct pidhashhead *pidhashtbl;
132 u_long pidhash;
133 struct pgrphashhead *pgrphashtbl;
134 u_long pgrphash;
135 struct proclist allproc;
136 struct proclist zombproc;
137 struct sx allproc_lock;
138 struct sx proctree_lock;
139 struct mtx ppeers_lock;
140 uma_zone_t proc_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_threads = offsetof(struct proc, p_threads);
151 const int thread_off_td_tid = offsetof(struct thread, td_tid);
152 const int thread_off_td_name = offsetof(struct thread, td_name);
153 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
154 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
155 const int thread_off_td_plist = offsetof(struct thread, td_plist);
156
157 int kstack_pages = KSTACK_PAGES;
158 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
159 "Kernel stack size in pages");
160 static int vmmap_skip_res_cnt = 0;
161 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
162 &vmmap_skip_res_cnt, 0,
163 "Skip calculation of the pages resident count in kern.proc.vmmap");
164
165 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
166 #ifdef COMPAT_FREEBSD32
167 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
168 #endif
169
170 /*
171 * Initialize global process hashing structures.
172 */
173 void
174 procinit()
175 {
176
177 sx_init(&allproc_lock, "allproc");
178 sx_init(&proctree_lock, "proctree");
179 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
180 LIST_INIT(&allproc);
181 LIST_INIT(&zombproc);
182 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
183 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
184 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
185 proc_ctor, proc_dtor, proc_init, proc_fini,
186 UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
187 uihashinit();
188 }
189
190 /*
191 * Prepare a proc for use.
192 */
193 static int
194 proc_ctor(void *mem, int size, void *arg, int flags)
195 {
196 struct proc *p;
197
198 p = (struct proc *)mem;
199 SDT_PROBE4(proc, kernel, ctor , entry, p, size, arg, flags);
200 EVENTHANDLER_INVOKE(process_ctor, p);
201 SDT_PROBE4(proc, kernel, ctor , return, p, size, arg, flags);
202 return (0);
203 }
204
205 /*
206 * Reclaim a proc after use.
207 */
208 static void
209 proc_dtor(void *mem, int size, void *arg)
210 {
211 struct proc *p;
212 struct thread *td;
213
214 /* INVARIANTS checks go here */
215 p = (struct proc *)mem;
216 td = FIRST_THREAD_IN_PROC(p);
217 SDT_PROBE4(proc, kernel, dtor, entry, p, size, arg, td);
218 if (td != NULL) {
219 #ifdef INVARIANTS
220 KASSERT((p->p_numthreads == 1),
221 ("bad number of threads in exiting process"));
222 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
223 #endif
224 /* Free all OSD associated to this thread. */
225 osd_thread_exit(td);
226 }
227 EVENTHANDLER_INVOKE(process_dtor, p);
228 if (p->p_ksi != NULL)
229 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
230 SDT_PROBE3(proc, kernel, dtor, return, p, size, arg);
231 }
232
233 /*
234 * Initialize type-stable parts of a proc (when newly created).
235 */
236 static int
237 proc_init(void *mem, int size, int flags)
238 {
239 struct proc *p;
240
241 p = (struct proc *)mem;
242 SDT_PROBE3(proc, kernel, init, entry, p, size, flags);
243 p->p_sched = (struct p_sched *)&p[1];
244 bzero(&p->p_mtx, sizeof(struct mtx));
245 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
246 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
247 cv_init(&p->p_pwait, "ppwait");
248 cv_init(&p->p_dbgwait, "dbgwait");
249 TAILQ_INIT(&p->p_threads); /* all threads in proc */
250 EVENTHANDLER_INVOKE(process_init, p);
251 p->p_stats = pstats_alloc();
252 p->p_pgrp = NULL;
253 SDT_PROBE3(proc, kernel, init, return, p, size, flags);
254 return (0);
255 }
256
257 /*
258 * UMA should ensure that this function is never called.
259 * Freeing a proc structure would violate type stability.
260 */
261 static void
262 proc_fini(void *mem, int size)
263 {
264 #ifdef notnow
265 struct proc *p;
266
267 p = (struct proc *)mem;
268 EVENTHANDLER_INVOKE(process_fini, p);
269 pstats_free(p->p_stats);
270 thread_free(FIRST_THREAD_IN_PROC(p));
271 mtx_destroy(&p->p_mtx);
272 if (p->p_ksi != NULL)
273 ksiginfo_free(p->p_ksi);
274 #else
275 panic("proc reclaimed");
276 #endif
277 }
278
279 /*
280 * Is p an inferior of the current process?
281 */
282 int
283 inferior(struct proc *p)
284 {
285
286 sx_assert(&proctree_lock, SX_LOCKED);
287 PROC_LOCK_ASSERT(p, MA_OWNED);
288 for (; p != curproc; p = proc_realparent(p)) {
289 if (p->p_pid == 0)
290 return (0);
291 }
292 return (1);
293 }
294
295 struct proc *
296 pfind_locked(pid_t pid)
297 {
298 struct proc *p;
299
300 sx_assert(&allproc_lock, SX_LOCKED);
301 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
302 if (p->p_pid == pid) {
303 PROC_LOCK(p);
304 if (p->p_state == PRS_NEW) {
305 PROC_UNLOCK(p);
306 p = NULL;
307 }
308 break;
309 }
310 }
311 return (p);
312 }
313
314 /*
315 * Locate a process by number; return only "live" processes -- i.e., neither
316 * zombies nor newly born but incompletely initialized processes. By not
317 * returning processes in the PRS_NEW state, we allow callers to avoid
318 * testing for that condition to avoid dereferencing p_ucred, et al.
319 */
320 struct proc *
321 pfind(pid_t pid)
322 {
323 struct proc *p;
324
325 sx_slock(&allproc_lock);
326 p = pfind_locked(pid);
327 sx_sunlock(&allproc_lock);
328 return (p);
329 }
330
331 static struct proc *
332 pfind_tid_locked(pid_t tid)
333 {
334 struct proc *p;
335 struct thread *td;
336
337 sx_assert(&allproc_lock, SX_LOCKED);
338 FOREACH_PROC_IN_SYSTEM(p) {
339 PROC_LOCK(p);
340 if (p->p_state == PRS_NEW) {
341 PROC_UNLOCK(p);
342 continue;
343 }
344 FOREACH_THREAD_IN_PROC(p, td) {
345 if (td->td_tid == tid)
346 goto found;
347 }
348 PROC_UNLOCK(p);
349 }
350 found:
351 return (p);
352 }
353
354 /*
355 * Locate a process group by number.
356 * The caller must hold proctree_lock.
357 */
358 struct pgrp *
359 pgfind(pgid)
360 register pid_t pgid;
361 {
362 register struct pgrp *pgrp;
363
364 sx_assert(&proctree_lock, SX_LOCKED);
365
366 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
367 if (pgrp->pg_id == pgid) {
368 PGRP_LOCK(pgrp);
369 return (pgrp);
370 }
371 }
372 return (NULL);
373 }
374
375 /*
376 * Locate process and do additional manipulations, depending on flags.
377 */
378 int
379 pget(pid_t pid, int flags, struct proc **pp)
380 {
381 struct proc *p;
382 int error;
383
384 sx_slock(&allproc_lock);
385 if (pid <= PID_MAX) {
386 p = pfind_locked(pid);
387 if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
388 p = zpfind_locked(pid);
389 } else if ((flags & PGET_NOTID) == 0) {
390 p = pfind_tid_locked(pid);
391 } else {
392 p = NULL;
393 }
394 sx_sunlock(&allproc_lock);
395 if (p == NULL)
396 return (ESRCH);
397 if ((flags & PGET_CANSEE) != 0) {
398 error = p_cansee(curthread, p);
399 if (error != 0)
400 goto errout;
401 }
402 if ((flags & PGET_CANDEBUG) != 0) {
403 error = p_candebug(curthread, p);
404 if (error != 0)
405 goto errout;
406 }
407 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
408 error = EPERM;
409 goto errout;
410 }
411 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
412 error = ESRCH;
413 goto errout;
414 }
415 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
416 /*
417 * XXXRW: Not clear ESRCH is the right error during proc
418 * execve().
419 */
420 error = ESRCH;
421 goto errout;
422 }
423 if ((flags & PGET_HOLD) != 0) {
424 _PHOLD(p);
425 PROC_UNLOCK(p);
426 }
427 *pp = p;
428 return (0);
429 errout:
430 PROC_UNLOCK(p);
431 return (error);
432 }
433
434 /*
435 * Create a new process group.
436 * pgid must be equal to the pid of p.
437 * Begin a new session if required.
438 */
439 int
440 enterpgrp(p, pgid, pgrp, sess)
441 register struct proc *p;
442 pid_t pgid;
443 struct pgrp *pgrp;
444 struct session *sess;
445 {
446
447 sx_assert(&proctree_lock, SX_XLOCKED);
448
449 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
450 KASSERT(p->p_pid == pgid,
451 ("enterpgrp: new pgrp and pid != pgid"));
452 KASSERT(pgfind(pgid) == NULL,
453 ("enterpgrp: pgrp with pgid exists"));
454 KASSERT(!SESS_LEADER(p),
455 ("enterpgrp: session leader attempted setpgrp"));
456
457 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
458
459 if (sess != NULL) {
460 /*
461 * new session
462 */
463 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
464 PROC_LOCK(p);
465 p->p_flag &= ~P_CONTROLT;
466 PROC_UNLOCK(p);
467 PGRP_LOCK(pgrp);
468 sess->s_leader = p;
469 sess->s_sid = p->p_pid;
470 refcount_init(&sess->s_count, 1);
471 sess->s_ttyvp = NULL;
472 sess->s_ttydp = NULL;
473 sess->s_ttyp = NULL;
474 bcopy(p->p_session->s_login, sess->s_login,
475 sizeof(sess->s_login));
476 pgrp->pg_session = sess;
477 KASSERT(p == curproc,
478 ("enterpgrp: mksession and p != curproc"));
479 } else {
480 pgrp->pg_session = p->p_session;
481 sess_hold(pgrp->pg_session);
482 PGRP_LOCK(pgrp);
483 }
484 pgrp->pg_id = pgid;
485 LIST_INIT(&pgrp->pg_members);
486
487 /*
488 * As we have an exclusive lock of proctree_lock,
489 * this should not deadlock.
490 */
491 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
492 pgrp->pg_jobc = 0;
493 SLIST_INIT(&pgrp->pg_sigiolst);
494 PGRP_UNLOCK(pgrp);
495
496 doenterpgrp(p, pgrp);
497
498 return (0);
499 }
500
501 /*
502 * Move p to an existing process group
503 */
504 int
505 enterthispgrp(p, pgrp)
506 register struct proc *p;
507 struct pgrp *pgrp;
508 {
509
510 sx_assert(&proctree_lock, SX_XLOCKED);
511 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
512 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
513 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
514 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
515 KASSERT(pgrp->pg_session == p->p_session,
516 ("%s: pgrp's session %p, p->p_session %p.\n",
517 __func__,
518 pgrp->pg_session,
519 p->p_session));
520 KASSERT(pgrp != p->p_pgrp,
521 ("%s: p belongs to pgrp.", __func__));
522
523 doenterpgrp(p, pgrp);
524
525 return (0);
526 }
527
528 /*
529 * Move p to a process group
530 */
531 static void
532 doenterpgrp(p, pgrp)
533 struct proc *p;
534 struct pgrp *pgrp;
535 {
536 struct pgrp *savepgrp;
537
538 sx_assert(&proctree_lock, SX_XLOCKED);
539 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
540 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
541 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
542 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
543
544 savepgrp = p->p_pgrp;
545
546 /*
547 * Adjust eligibility of affected pgrps to participate in job control.
548 * Increment eligibility counts before decrementing, otherwise we
549 * could reach 0 spuriously during the first call.
550 */
551 fixjobc(p, pgrp, 1);
552 fixjobc(p, p->p_pgrp, 0);
553
554 PGRP_LOCK(pgrp);
555 PGRP_LOCK(savepgrp);
556 PROC_LOCK(p);
557 LIST_REMOVE(p, p_pglist);
558 p->p_pgrp = pgrp;
559 PROC_UNLOCK(p);
560 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
561 PGRP_UNLOCK(savepgrp);
562 PGRP_UNLOCK(pgrp);
563 if (LIST_EMPTY(&savepgrp->pg_members))
564 pgdelete(savepgrp);
565 }
566
567 /*
568 * remove process from process group
569 */
570 int
571 leavepgrp(p)
572 register struct proc *p;
573 {
574 struct pgrp *savepgrp;
575
576 sx_assert(&proctree_lock, SX_XLOCKED);
577 savepgrp = p->p_pgrp;
578 PGRP_LOCK(savepgrp);
579 PROC_LOCK(p);
580 LIST_REMOVE(p, p_pglist);
581 p->p_pgrp = NULL;
582 PROC_UNLOCK(p);
583 PGRP_UNLOCK(savepgrp);
584 if (LIST_EMPTY(&savepgrp->pg_members))
585 pgdelete(savepgrp);
586 return (0);
587 }
588
589 /*
590 * delete a process group
591 */
592 static void
593 pgdelete(pgrp)
594 register struct pgrp *pgrp;
595 {
596 struct session *savesess;
597 struct tty *tp;
598
599 sx_assert(&proctree_lock, SX_XLOCKED);
600 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
601 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
602
603 /*
604 * Reset any sigio structures pointing to us as a result of
605 * F_SETOWN with our pgid.
606 */
607 funsetownlst(&pgrp->pg_sigiolst);
608
609 PGRP_LOCK(pgrp);
610 tp = pgrp->pg_session->s_ttyp;
611 LIST_REMOVE(pgrp, pg_hash);
612 savesess = pgrp->pg_session;
613 PGRP_UNLOCK(pgrp);
614
615 /* Remove the reference to the pgrp before deallocating it. */
616 if (tp != NULL) {
617 tty_lock(tp);
618 tty_rel_pgrp(tp, pgrp);
619 }
620
621 mtx_destroy(&pgrp->pg_mtx);
622 free(pgrp, M_PGRP);
623 sess_release(savesess);
624 }
625
626 static void
627 pgadjustjobc(pgrp, entering)
628 struct pgrp *pgrp;
629 int entering;
630 {
631
632 PGRP_LOCK(pgrp);
633 if (entering)
634 pgrp->pg_jobc++;
635 else {
636 --pgrp->pg_jobc;
637 if (pgrp->pg_jobc == 0)
638 orphanpg(pgrp);
639 }
640 PGRP_UNLOCK(pgrp);
641 }
642
643 /*
644 * Adjust pgrp jobc counters when specified process changes process group.
645 * We count the number of processes in each process group that "qualify"
646 * the group for terminal job control (those with a parent in a different
647 * process group of the same session). If that count reaches zero, the
648 * process group becomes orphaned. Check both the specified process'
649 * process group and that of its children.
650 * entering == 0 => p is leaving specified group.
651 * entering == 1 => p is entering specified group.
652 */
653 void
654 fixjobc(p, pgrp, entering)
655 register struct proc *p;
656 register struct pgrp *pgrp;
657 int entering;
658 {
659 register struct pgrp *hispgrp;
660 register struct session *mysession;
661
662 sx_assert(&proctree_lock, SX_LOCKED);
663 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
664 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
665 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
666
667 /*
668 * Check p's parent to see whether p qualifies its own process
669 * group; if so, adjust count for p's process group.
670 */
671 mysession = pgrp->pg_session;
672 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
673 hispgrp->pg_session == mysession)
674 pgadjustjobc(pgrp, entering);
675
676 /*
677 * Check this process' children to see whether they qualify
678 * their process groups; if so, adjust counts for children's
679 * process groups.
680 */
681 LIST_FOREACH(p, &p->p_children, p_sibling) {
682 hispgrp = p->p_pgrp;
683 if (hispgrp == pgrp ||
684 hispgrp->pg_session != mysession)
685 continue;
686 PROC_LOCK(p);
687 if (p->p_state == PRS_ZOMBIE) {
688 PROC_UNLOCK(p);
689 continue;
690 }
691 PROC_UNLOCK(p);
692 pgadjustjobc(hispgrp, entering);
693 }
694 }
695
696 /*
697 * A process group has become orphaned;
698 * if there are any stopped processes in the group,
699 * hang-up all process in that group.
700 */
701 static void
702 orphanpg(pg)
703 struct pgrp *pg;
704 {
705 register struct proc *p;
706
707 PGRP_LOCK_ASSERT(pg, MA_OWNED);
708
709 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
710 PROC_LOCK(p);
711 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
712 PROC_UNLOCK(p);
713 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
714 PROC_LOCK(p);
715 kern_psignal(p, SIGHUP);
716 kern_psignal(p, SIGCONT);
717 PROC_UNLOCK(p);
718 }
719 return;
720 }
721 PROC_UNLOCK(p);
722 }
723 }
724
725 void
726 sess_hold(struct session *s)
727 {
728
729 refcount_acquire(&s->s_count);
730 }
731
732 void
733 sess_release(struct session *s)
734 {
735
736 if (refcount_release(&s->s_count)) {
737 if (s->s_ttyp != NULL) {
738 tty_lock(s->s_ttyp);
739 tty_rel_sess(s->s_ttyp, s);
740 }
741 mtx_destroy(&s->s_mtx);
742 free(s, M_SESSION);
743 }
744 }
745
746 #ifdef DDB
747
748 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
749 {
750 register struct pgrp *pgrp;
751 register struct proc *p;
752 register int i;
753
754 for (i = 0; i <= pgrphash; i++) {
755 if (!LIST_EMPTY(&pgrphashtbl[i])) {
756 printf("\tindx %d\n", i);
757 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
758 printf(
759 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
760 (void *)pgrp, (long)pgrp->pg_id,
761 (void *)pgrp->pg_session,
762 pgrp->pg_session->s_count,
763 (void *)LIST_FIRST(&pgrp->pg_members));
764 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
765 printf("\t\tpid %ld addr %p pgrp %p\n",
766 (long)p->p_pid, (void *)p,
767 (void *)p->p_pgrp);
768 }
769 }
770 }
771 }
772 }
773 #endif /* DDB */
774
775 /*
776 * Calculate the kinfo_proc members which contain process-wide
777 * informations.
778 * Must be called with the target process locked.
779 */
780 static void
781 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
782 {
783 struct thread *td;
784
785 PROC_LOCK_ASSERT(p, MA_OWNED);
786
787 kp->ki_estcpu = 0;
788 kp->ki_pctcpu = 0;
789 FOREACH_THREAD_IN_PROC(p, td) {
790 thread_lock(td);
791 kp->ki_pctcpu += sched_pctcpu(td);
792 kp->ki_estcpu += td->td_estcpu;
793 thread_unlock(td);
794 }
795 }
796
797 /*
798 * Clear kinfo_proc and fill in any information that is common
799 * to all threads in the process.
800 * Must be called with the target process locked.
801 */
802 static void
803 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
804 {
805 struct thread *td0;
806 struct tty *tp;
807 struct session *sp;
808 struct ucred *cred;
809 struct sigacts *ps;
810
811 PROC_LOCK_ASSERT(p, MA_OWNED);
812 bzero(kp, sizeof(*kp));
813
814 kp->ki_structsize = sizeof(*kp);
815 kp->ki_paddr = p;
816 kp->ki_addr =/* p->p_addr; */0; /* XXX */
817 kp->ki_args = p->p_args;
818 kp->ki_textvp = p->p_textvp;
819 #ifdef KTRACE
820 kp->ki_tracep = p->p_tracevp;
821 kp->ki_traceflag = p->p_traceflag;
822 #endif
823 kp->ki_fd = p->p_fd;
824 kp->ki_vmspace = p->p_vmspace;
825 kp->ki_flag = p->p_flag;
826 kp->ki_flag2 = p->p_flag2;
827 cred = p->p_ucred;
828 if (cred) {
829 kp->ki_uid = cred->cr_uid;
830 kp->ki_ruid = cred->cr_ruid;
831 kp->ki_svuid = cred->cr_svuid;
832 kp->ki_cr_flags = 0;
833 if (cred->cr_flags & CRED_FLAG_CAPMODE)
834 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
835 /* XXX bde doesn't like KI_NGROUPS */
836 if (cred->cr_ngroups > KI_NGROUPS) {
837 kp->ki_ngroups = KI_NGROUPS;
838 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
839 } else
840 kp->ki_ngroups = cred->cr_ngroups;
841 bcopy(cred->cr_groups, kp->ki_groups,
842 kp->ki_ngroups * sizeof(gid_t));
843 kp->ki_rgid = cred->cr_rgid;
844 kp->ki_svgid = cred->cr_svgid;
845 /* If jailed(cred), emulate the old P_JAILED flag. */
846 if (jailed(cred)) {
847 kp->ki_flag |= P_JAILED;
848 /* If inside the jail, use 0 as a jail ID. */
849 if (cred->cr_prison != curthread->td_ucred->cr_prison)
850 kp->ki_jid = cred->cr_prison->pr_id;
851 }
852 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
853 sizeof(kp->ki_loginclass));
854 }
855 ps = p->p_sigacts;
856 if (ps) {
857 mtx_lock(&ps->ps_mtx);
858 kp->ki_sigignore = ps->ps_sigignore;
859 kp->ki_sigcatch = ps->ps_sigcatch;
860 mtx_unlock(&ps->ps_mtx);
861 }
862 if (p->p_state != PRS_NEW &&
863 p->p_state != PRS_ZOMBIE &&
864 p->p_vmspace != NULL) {
865 struct vmspace *vm = p->p_vmspace;
866
867 kp->ki_size = vm->vm_map.size;
868 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
869 FOREACH_THREAD_IN_PROC(p, td0) {
870 if (!TD_IS_SWAPPED(td0))
871 kp->ki_rssize += td0->td_kstack_pages;
872 }
873 kp->ki_swrss = vm->vm_swrss;
874 kp->ki_tsize = vm->vm_tsize;
875 kp->ki_dsize = vm->vm_dsize;
876 kp->ki_ssize = vm->vm_ssize;
877 } else if (p->p_state == PRS_ZOMBIE)
878 kp->ki_stat = SZOMB;
879 if (kp->ki_flag & P_INMEM)
880 kp->ki_sflag = PS_INMEM;
881 else
882 kp->ki_sflag = 0;
883 /* Calculate legacy swtime as seconds since 'swtick'. */
884 kp->ki_swtime = (ticks - p->p_swtick) / hz;
885 kp->ki_pid = p->p_pid;
886 kp->ki_nice = p->p_nice;
887 kp->ki_fibnum = p->p_fibnum;
888 kp->ki_start = p->p_stats->p_start;
889 timevaladd(&kp->ki_start, &boottime);
890 PROC_STATLOCK(p);
891 rufetch(p, &kp->ki_rusage);
892 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
893 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
894 PROC_STATUNLOCK(p);
895 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
896 /* Some callers want child times in a single value. */
897 kp->ki_childtime = kp->ki_childstime;
898 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
899
900 FOREACH_THREAD_IN_PROC(p, td0)
901 kp->ki_cow += td0->td_cow;
902
903 tp = NULL;
904 if (p->p_pgrp) {
905 kp->ki_pgid = p->p_pgrp->pg_id;
906 kp->ki_jobc = p->p_pgrp->pg_jobc;
907 sp = p->p_pgrp->pg_session;
908
909 if (sp != NULL) {
910 kp->ki_sid = sp->s_sid;
911 SESS_LOCK(sp);
912 strlcpy(kp->ki_login, sp->s_login,
913 sizeof(kp->ki_login));
914 if (sp->s_ttyvp)
915 kp->ki_kiflag |= KI_CTTY;
916 if (SESS_LEADER(p))
917 kp->ki_kiflag |= KI_SLEADER;
918 /* XXX proctree_lock */
919 tp = sp->s_ttyp;
920 SESS_UNLOCK(sp);
921 }
922 }
923 if ((p->p_flag & P_CONTROLT) && tp != NULL) {
924 kp->ki_tdev = tty_udev(tp);
925 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
926 if (tp->t_session)
927 kp->ki_tsid = tp->t_session->s_sid;
928 } else
929 kp->ki_tdev = NODEV;
930 if (p->p_comm[0] != '\0')
931 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
932 if (p->p_sysent && p->p_sysent->sv_name != NULL &&
933 p->p_sysent->sv_name[0] != '\0')
934 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
935 kp->ki_siglist = p->p_siglist;
936 kp->ki_xstat = p->p_xstat;
937 kp->ki_acflag = p->p_acflag;
938 kp->ki_lock = p->p_lock;
939 if (p->p_pptr)
940 kp->ki_ppid = p->p_pptr->p_pid;
941 }
942
943 /*
944 * Fill in information that is thread specific. Must be called with
945 * target process locked. If 'preferthread' is set, overwrite certain
946 * process-related fields that are maintained for both threads and
947 * processes.
948 */
949 static void
950 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
951 {
952 struct proc *p;
953
954 p = td->td_proc;
955 kp->ki_tdaddr = td;
956 PROC_LOCK_ASSERT(p, MA_OWNED);
957
958 if (preferthread)
959 PROC_STATLOCK(p);
960 thread_lock(td);
961 if (td->td_wmesg != NULL)
962 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
963 else
964 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
965 strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
966 if (TD_ON_LOCK(td)) {
967 kp->ki_kiflag |= KI_LOCKBLOCK;
968 strlcpy(kp->ki_lockname, td->td_lockname,
969 sizeof(kp->ki_lockname));
970 } else {
971 kp->ki_kiflag &= ~KI_LOCKBLOCK;
972 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
973 }
974
975 if (p->p_state == PRS_NORMAL) { /* approximate. */
976 if (TD_ON_RUNQ(td) ||
977 TD_CAN_RUN(td) ||
978 TD_IS_RUNNING(td)) {
979 kp->ki_stat = SRUN;
980 } else if (P_SHOULDSTOP(p)) {
981 kp->ki_stat = SSTOP;
982 } else if (TD_IS_SLEEPING(td)) {
983 kp->ki_stat = SSLEEP;
984 } else if (TD_ON_LOCK(td)) {
985 kp->ki_stat = SLOCK;
986 } else {
987 kp->ki_stat = SWAIT;
988 }
989 } else if (p->p_state == PRS_ZOMBIE) {
990 kp->ki_stat = SZOMB;
991 } else {
992 kp->ki_stat = SIDL;
993 }
994
995 /* Things in the thread */
996 kp->ki_wchan = td->td_wchan;
997 kp->ki_pri.pri_level = td->td_priority;
998 kp->ki_pri.pri_native = td->td_base_pri;
999 kp->ki_lastcpu = td->td_lastcpu;
1000 kp->ki_oncpu = td->td_oncpu;
1001 kp->ki_tdflags = td->td_flags;
1002 kp->ki_tid = td->td_tid;
1003 kp->ki_numthreads = p->p_numthreads;
1004 kp->ki_pcb = td->td_pcb;
1005 kp->ki_kstack = (void *)td->td_kstack;
1006 kp->ki_slptime = (ticks - td->td_slptick) / hz;
1007 kp->ki_pri.pri_class = td->td_pri_class;
1008 kp->ki_pri.pri_user = td->td_user_pri;
1009
1010 if (preferthread) {
1011 rufetchtd(td, &kp->ki_rusage);
1012 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1013 kp->ki_pctcpu = sched_pctcpu(td);
1014 kp->ki_estcpu = td->td_estcpu;
1015 kp->ki_cow = td->td_cow;
1016 }
1017
1018 /* We can't get this anymore but ps etc never used it anyway. */
1019 kp->ki_rqindex = 0;
1020
1021 if (preferthread)
1022 kp->ki_siglist = td->td_siglist;
1023 kp->ki_sigmask = td->td_sigmask;
1024 thread_unlock(td);
1025 if (preferthread)
1026 PROC_STATUNLOCK(p);
1027 }
1028
1029 /*
1030 * Fill in a kinfo_proc structure for the specified process.
1031 * Must be called with the target process locked.
1032 */
1033 void
1034 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1035 {
1036
1037 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1038
1039 fill_kinfo_proc_only(p, kp);
1040 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1041 fill_kinfo_aggregate(p, kp);
1042 }
1043
1044 struct pstats *
1045 pstats_alloc(void)
1046 {
1047
1048 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1049 }
1050
1051 /*
1052 * Copy parts of p_stats; zero the rest of p_stats (statistics).
1053 */
1054 void
1055 pstats_fork(struct pstats *src, struct pstats *dst)
1056 {
1057
1058 bzero(&dst->pstat_startzero,
1059 __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1060 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1061 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1062 }
1063
1064 void
1065 pstats_free(struct pstats *ps)
1066 {
1067
1068 free(ps, M_SUBPROC);
1069 }
1070
1071 static struct proc *
1072 zpfind_locked(pid_t pid)
1073 {
1074 struct proc *p;
1075
1076 sx_assert(&allproc_lock, SX_LOCKED);
1077 LIST_FOREACH(p, &zombproc, p_list) {
1078 if (p->p_pid == pid) {
1079 PROC_LOCK(p);
1080 break;
1081 }
1082 }
1083 return (p);
1084 }
1085
1086 /*
1087 * Locate a zombie process by number
1088 */
1089 struct proc *
1090 zpfind(pid_t pid)
1091 {
1092 struct proc *p;
1093
1094 sx_slock(&allproc_lock);
1095 p = zpfind_locked(pid);
1096 sx_sunlock(&allproc_lock);
1097 return (p);
1098 }
1099
1100 #ifdef COMPAT_FREEBSD32
1101
1102 /*
1103 * This function is typically used to copy out the kernel address, so
1104 * it can be replaced by assignment of zero.
1105 */
1106 static inline uint32_t
1107 ptr32_trim(void *ptr)
1108 {
1109 uintptr_t uptr;
1110
1111 uptr = (uintptr_t)ptr;
1112 return ((uptr > UINT_MAX) ? 0 : uptr);
1113 }
1114
1115 #define PTRTRIM_CP(src,dst,fld) \
1116 do { (dst).fld = ptr32_trim((src).fld); } while (0)
1117
1118 static void
1119 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1120 {
1121 int i;
1122
1123 bzero(ki32, sizeof(struct kinfo_proc32));
1124 ki32->ki_structsize = sizeof(struct kinfo_proc32);
1125 CP(*ki, *ki32, ki_layout);
1126 PTRTRIM_CP(*ki, *ki32, ki_args);
1127 PTRTRIM_CP(*ki, *ki32, ki_paddr);
1128 PTRTRIM_CP(*ki, *ki32, ki_addr);
1129 PTRTRIM_CP(*ki, *ki32, ki_tracep);
1130 PTRTRIM_CP(*ki, *ki32, ki_textvp);
1131 PTRTRIM_CP(*ki, *ki32, ki_fd);
1132 PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1133 PTRTRIM_CP(*ki, *ki32, ki_wchan);
1134 CP(*ki, *ki32, ki_pid);
1135 CP(*ki, *ki32, ki_ppid);
1136 CP(*ki, *ki32, ki_pgid);
1137 CP(*ki, *ki32, ki_tpgid);
1138 CP(*ki, *ki32, ki_sid);
1139 CP(*ki, *ki32, ki_tsid);
1140 CP(*ki, *ki32, ki_jobc);
1141 CP(*ki, *ki32, ki_tdev);
1142 CP(*ki, *ki32, ki_siglist);
1143 CP(*ki, *ki32, ki_sigmask);
1144 CP(*ki, *ki32, ki_sigignore);
1145 CP(*ki, *ki32, ki_sigcatch);
1146 CP(*ki, *ki32, ki_uid);
1147 CP(*ki, *ki32, ki_ruid);
1148 CP(*ki, *ki32, ki_svuid);
1149 CP(*ki, *ki32, ki_rgid);
1150 CP(*ki, *ki32, ki_svgid);
1151 CP(*ki, *ki32, ki_ngroups);
1152 for (i = 0; i < KI_NGROUPS; i++)
1153 CP(*ki, *ki32, ki_groups[i]);
1154 CP(*ki, *ki32, ki_size);
1155 CP(*ki, *ki32, ki_rssize);
1156 CP(*ki, *ki32, ki_swrss);
1157 CP(*ki, *ki32, ki_tsize);
1158 CP(*ki, *ki32, ki_dsize);
1159 CP(*ki, *ki32, ki_ssize);
1160 CP(*ki, *ki32, ki_xstat);
1161 CP(*ki, *ki32, ki_acflag);
1162 CP(*ki, *ki32, ki_pctcpu);
1163 CP(*ki, *ki32, ki_estcpu);
1164 CP(*ki, *ki32, ki_slptime);
1165 CP(*ki, *ki32, ki_swtime);
1166 CP(*ki, *ki32, ki_cow);
1167 CP(*ki, *ki32, ki_runtime);
1168 TV_CP(*ki, *ki32, ki_start);
1169 TV_CP(*ki, *ki32, ki_childtime);
1170 CP(*ki, *ki32, ki_flag);
1171 CP(*ki, *ki32, ki_kiflag);
1172 CP(*ki, *ki32, ki_traceflag);
1173 CP(*ki, *ki32, ki_stat);
1174 CP(*ki, *ki32, ki_nice);
1175 CP(*ki, *ki32, ki_lock);
1176 CP(*ki, *ki32, ki_rqindex);
1177 CP(*ki, *ki32, ki_oncpu);
1178 CP(*ki, *ki32, ki_lastcpu);
1179 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1180 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1181 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1182 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1183 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1184 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1185 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1186 CP(*ki, *ki32, ki_flag2);
1187 CP(*ki, *ki32, ki_fibnum);
1188 CP(*ki, *ki32, ki_cr_flags);
1189 CP(*ki, *ki32, ki_jid);
1190 CP(*ki, *ki32, ki_numthreads);
1191 CP(*ki, *ki32, ki_tid);
1192 CP(*ki, *ki32, ki_pri);
1193 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1194 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1195 PTRTRIM_CP(*ki, *ki32, ki_pcb);
1196 PTRTRIM_CP(*ki, *ki32, ki_kstack);
1197 PTRTRIM_CP(*ki, *ki32, ki_udata);
1198 CP(*ki, *ki32, ki_sflag);
1199 CP(*ki, *ki32, ki_tdflags);
1200 }
1201 #endif
1202
1203 int
1204 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1205 {
1206 struct thread *td;
1207 struct kinfo_proc ki;
1208 #ifdef COMPAT_FREEBSD32
1209 struct kinfo_proc32 ki32;
1210 #endif
1211 int error;
1212
1213 PROC_LOCK_ASSERT(p, MA_OWNED);
1214 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1215
1216 error = 0;
1217 fill_kinfo_proc(p, &ki);
1218 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1219 #ifdef COMPAT_FREEBSD32
1220 if ((flags & KERN_PROC_MASK32) != 0) {
1221 freebsd32_kinfo_proc_out(&ki, &ki32);
1222 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1223 error = ENOMEM;
1224 } else
1225 #endif
1226 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1227 error = ENOMEM;
1228 } else {
1229 FOREACH_THREAD_IN_PROC(p, td) {
1230 fill_kinfo_thread(td, &ki, 1);
1231 #ifdef COMPAT_FREEBSD32
1232 if ((flags & KERN_PROC_MASK32) != 0) {
1233 freebsd32_kinfo_proc_out(&ki, &ki32);
1234 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1235 error = ENOMEM;
1236 } else
1237 #endif
1238 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1239 error = ENOMEM;
1240 if (error != 0)
1241 break;
1242 }
1243 }
1244 PROC_UNLOCK(p);
1245 return (error);
1246 }
1247
1248 static int
1249 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1250 int doingzomb)
1251 {
1252 struct sbuf sb;
1253 struct kinfo_proc ki;
1254 struct proc *np;
1255 int error, error2;
1256 pid_t pid;
1257
1258 pid = p->p_pid;
1259 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1260 error = kern_proc_out(p, &sb, flags);
1261 error2 = sbuf_finish(&sb);
1262 sbuf_delete(&sb);
1263 if (error != 0)
1264 return (error);
1265 else if (error2 != 0)
1266 return (error2);
1267 if (doingzomb)
1268 np = zpfind(pid);
1269 else {
1270 if (pid == 0)
1271 return (0);
1272 np = pfind(pid);
1273 }
1274 if (np == NULL)
1275 return (ESRCH);
1276 if (np != p) {
1277 PROC_UNLOCK(np);
1278 return (ESRCH);
1279 }
1280 PROC_UNLOCK(np);
1281 return (0);
1282 }
1283
1284 static int
1285 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1286 {
1287 int *name = (int *)arg1;
1288 u_int namelen = arg2;
1289 struct proc *p;
1290 int flags, doingzomb, oid_number;
1291 int error = 0;
1292
1293 oid_number = oidp->oid_number;
1294 if (oid_number != KERN_PROC_ALL &&
1295 (oid_number & KERN_PROC_INC_THREAD) == 0)
1296 flags = KERN_PROC_NOTHREADS;
1297 else {
1298 flags = 0;
1299 oid_number &= ~KERN_PROC_INC_THREAD;
1300 }
1301 #ifdef COMPAT_FREEBSD32
1302 if (req->flags & SCTL_MASK32)
1303 flags |= KERN_PROC_MASK32;
1304 #endif
1305 if (oid_number == KERN_PROC_PID) {
1306 if (namelen != 1)
1307 return (EINVAL);
1308 error = sysctl_wire_old_buffer(req, 0);
1309 if (error)
1310 return (error);
1311 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1312 if (error != 0)
1313 return (error);
1314 error = sysctl_out_proc(p, req, flags, 0);
1315 return (error);
1316 }
1317
1318 switch (oid_number) {
1319 case KERN_PROC_ALL:
1320 if (namelen != 0)
1321 return (EINVAL);
1322 break;
1323 case KERN_PROC_PROC:
1324 if (namelen != 0 && namelen != 1)
1325 return (EINVAL);
1326 break;
1327 default:
1328 if (namelen != 1)
1329 return (EINVAL);
1330 break;
1331 }
1332
1333 if (!req->oldptr) {
1334 /* overestimate by 5 procs */
1335 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1336 if (error)
1337 return (error);
1338 }
1339 error = sysctl_wire_old_buffer(req, 0);
1340 if (error != 0)
1341 return (error);
1342 sx_slock(&allproc_lock);
1343 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1344 if (!doingzomb)
1345 p = LIST_FIRST(&allproc);
1346 else
1347 p = LIST_FIRST(&zombproc);
1348 for (; p != 0; p = LIST_NEXT(p, p_list)) {
1349 /*
1350 * Skip embryonic processes.
1351 */
1352 PROC_LOCK(p);
1353 if (p->p_state == PRS_NEW) {
1354 PROC_UNLOCK(p);
1355 continue;
1356 }
1357 KASSERT(p->p_ucred != NULL,
1358 ("process credential is NULL for non-NEW proc"));
1359 /*
1360 * Show a user only appropriate processes.
1361 */
1362 if (p_cansee(curthread, p)) {
1363 PROC_UNLOCK(p);
1364 continue;
1365 }
1366 /*
1367 * TODO - make more efficient (see notes below).
1368 * do by session.
1369 */
1370 switch (oid_number) {
1371
1372 case KERN_PROC_GID:
1373 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1374 PROC_UNLOCK(p);
1375 continue;
1376 }
1377 break;
1378
1379 case KERN_PROC_PGRP:
1380 /* could do this by traversing pgrp */
1381 if (p->p_pgrp == NULL ||
1382 p->p_pgrp->pg_id != (pid_t)name[0]) {
1383 PROC_UNLOCK(p);
1384 continue;
1385 }
1386 break;
1387
1388 case KERN_PROC_RGID:
1389 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1390 PROC_UNLOCK(p);
1391 continue;
1392 }
1393 break;
1394
1395 case KERN_PROC_SESSION:
1396 if (p->p_session == NULL ||
1397 p->p_session->s_sid != (pid_t)name[0]) {
1398 PROC_UNLOCK(p);
1399 continue;
1400 }
1401 break;
1402
1403 case KERN_PROC_TTY:
1404 if ((p->p_flag & P_CONTROLT) == 0 ||
1405 p->p_session == NULL) {
1406 PROC_UNLOCK(p);
1407 continue;
1408 }
1409 /* XXX proctree_lock */
1410 SESS_LOCK(p->p_session);
1411 if (p->p_session->s_ttyp == NULL ||
1412 tty_udev(p->p_session->s_ttyp) !=
1413 (dev_t)name[0]) {
1414 SESS_UNLOCK(p->p_session);
1415 PROC_UNLOCK(p);
1416 continue;
1417 }
1418 SESS_UNLOCK(p->p_session);
1419 break;
1420
1421 case KERN_PROC_UID:
1422 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1423 PROC_UNLOCK(p);
1424 continue;
1425 }
1426 break;
1427
1428 case KERN_PROC_RUID:
1429 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1430 PROC_UNLOCK(p);
1431 continue;
1432 }
1433 break;
1434
1435 case KERN_PROC_PROC:
1436 break;
1437
1438 default:
1439 break;
1440
1441 }
1442
1443 error = sysctl_out_proc(p, req, flags, doingzomb);
1444 if (error) {
1445 sx_sunlock(&allproc_lock);
1446 return (error);
1447 }
1448 }
1449 }
1450 sx_sunlock(&allproc_lock);
1451 return (0);
1452 }
1453
1454 struct pargs *
1455 pargs_alloc(int len)
1456 {
1457 struct pargs *pa;
1458
1459 pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1460 M_WAITOK);
1461 refcount_init(&pa->ar_ref, 1);
1462 pa->ar_length = len;
1463 return (pa);
1464 }
1465
1466 static void
1467 pargs_free(struct pargs *pa)
1468 {
1469
1470 free(pa, M_PARGS);
1471 }
1472
1473 void
1474 pargs_hold(struct pargs *pa)
1475 {
1476
1477 if (pa == NULL)
1478 return;
1479 refcount_acquire(&pa->ar_ref);
1480 }
1481
1482 void
1483 pargs_drop(struct pargs *pa)
1484 {
1485
1486 if (pa == NULL)
1487 return;
1488 if (refcount_release(&pa->ar_ref))
1489 pargs_free(pa);
1490 }
1491
1492 static int
1493 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1494 size_t len)
1495 {
1496 struct iovec iov;
1497 struct uio uio;
1498
1499 iov.iov_base = (caddr_t)buf;
1500 iov.iov_len = len;
1501 uio.uio_iov = &iov;
1502 uio.uio_iovcnt = 1;
1503 uio.uio_offset = offset;
1504 uio.uio_resid = (ssize_t)len;
1505 uio.uio_segflg = UIO_SYSSPACE;
1506 uio.uio_rw = UIO_READ;
1507 uio.uio_td = td;
1508
1509 return (proc_rwmem(p, &uio));
1510 }
1511
1512 static int
1513 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1514 size_t len)
1515 {
1516 size_t i;
1517 int error;
1518
1519 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1520 /*
1521 * Reading the chunk may validly return EFAULT if the string is shorter
1522 * than the chunk and is aligned at the end of the page, assuming the
1523 * next page is not mapped. So if EFAULT is returned do a fallback to
1524 * one byte read loop.
1525 */
1526 if (error == EFAULT) {
1527 for (i = 0; i < len; i++, buf++, sptr++) {
1528 error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1529 if (error != 0)
1530 return (error);
1531 if (*buf == '\0')
1532 break;
1533 }
1534 error = 0;
1535 }
1536 return (error);
1537 }
1538
1539 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */
1540
1541 enum proc_vector_type {
1542 PROC_ARG,
1543 PROC_ENV,
1544 PROC_AUX,
1545 };
1546
1547 #ifdef COMPAT_FREEBSD32
1548 static int
1549 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1550 size_t *vsizep, enum proc_vector_type type)
1551 {
1552 struct freebsd32_ps_strings pss;
1553 Elf32_Auxinfo aux;
1554 vm_offset_t vptr, ptr;
1555 uint32_t *proc_vector32;
1556 char **proc_vector;
1557 size_t vsize, size;
1558 int i, error;
1559
1560 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1561 &pss, sizeof(pss));
1562 if (error != 0)
1563 return (error);
1564 switch (type) {
1565 case PROC_ARG:
1566 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1567 vsize = pss.ps_nargvstr;
1568 if (vsize > ARG_MAX)
1569 return (ENOEXEC);
1570 size = vsize * sizeof(int32_t);
1571 break;
1572 case PROC_ENV:
1573 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1574 vsize = pss.ps_nenvstr;
1575 if (vsize > ARG_MAX)
1576 return (ENOEXEC);
1577 size = vsize * sizeof(int32_t);
1578 break;
1579 case PROC_AUX:
1580 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1581 (pss.ps_nenvstr + 1) * sizeof(int32_t);
1582 if (vptr % 4 != 0)
1583 return (ENOEXEC);
1584 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1585 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1586 if (error != 0)
1587 return (error);
1588 if (aux.a_type == AT_NULL)
1589 break;
1590 ptr += sizeof(aux);
1591 }
1592 if (aux.a_type != AT_NULL)
1593 return (ENOEXEC);
1594 vsize = i + 1;
1595 size = vsize * sizeof(aux);
1596 break;
1597 default:
1598 KASSERT(0, ("Wrong proc vector type: %d", type));
1599 return (EINVAL);
1600 }
1601 proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1602 error = proc_read_mem(td, p, vptr, proc_vector32, size);
1603 if (error != 0)
1604 goto done;
1605 if (type == PROC_AUX) {
1606 *proc_vectorp = (char **)proc_vector32;
1607 *vsizep = vsize;
1608 return (0);
1609 }
1610 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1611 for (i = 0; i < (int)vsize; i++)
1612 proc_vector[i] = PTRIN(proc_vector32[i]);
1613 *proc_vectorp = proc_vector;
1614 *vsizep = vsize;
1615 done:
1616 free(proc_vector32, M_TEMP);
1617 return (error);
1618 }
1619 #endif
1620
1621 static int
1622 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1623 size_t *vsizep, enum proc_vector_type type)
1624 {
1625 struct ps_strings pss;
1626 Elf_Auxinfo aux;
1627 vm_offset_t vptr, ptr;
1628 char **proc_vector;
1629 size_t vsize, size;
1630 int error, i;
1631
1632 #ifdef COMPAT_FREEBSD32
1633 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1634 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1635 #endif
1636 error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1637 &pss, sizeof(pss));
1638 if (error != 0)
1639 return (error);
1640 switch (type) {
1641 case PROC_ARG:
1642 vptr = (vm_offset_t)pss.ps_argvstr;
1643 vsize = pss.ps_nargvstr;
1644 if (vsize > ARG_MAX)
1645 return (ENOEXEC);
1646 size = vsize * sizeof(char *);
1647 break;
1648 case PROC_ENV:
1649 vptr = (vm_offset_t)pss.ps_envstr;
1650 vsize = pss.ps_nenvstr;
1651 if (vsize > ARG_MAX)
1652 return (ENOEXEC);
1653 size = vsize * sizeof(char *);
1654 break;
1655 case PROC_AUX:
1656 /*
1657 * The aux array is just above env array on the stack. Check
1658 * that the address is naturally aligned.
1659 */
1660 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1661 * sizeof(char *);
1662 #if __ELF_WORD_SIZE == 64
1663 if (vptr % sizeof(uint64_t) != 0)
1664 #else
1665 if (vptr % sizeof(uint32_t) != 0)
1666 #endif
1667 return (ENOEXEC);
1668 /*
1669 * We count the array size reading the aux vectors from the
1670 * stack until AT_NULL vector is returned. So (to keep the code
1671 * simple) we read the process stack twice: the first time here
1672 * to find the size and the second time when copying the vectors
1673 * to the allocated proc_vector.
1674 */
1675 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1676 error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1677 if (error != 0)
1678 return (error);
1679 if (aux.a_type == AT_NULL)
1680 break;
1681 ptr += sizeof(aux);
1682 }
1683 /*
1684 * If the PROC_AUXV_MAX entries are iterated over, and we have
1685 * not reached AT_NULL, it is most likely we are reading wrong
1686 * data: either the process doesn't have auxv array or data has
1687 * been modified. Return the error in this case.
1688 */
1689 if (aux.a_type != AT_NULL)
1690 return (ENOEXEC);
1691 vsize = i + 1;
1692 size = vsize * sizeof(aux);
1693 break;
1694 default:
1695 KASSERT(0, ("Wrong proc vector type: %d", type));
1696 return (EINVAL); /* In case we are built without INVARIANTS. */
1697 }
1698 proc_vector = malloc(size, M_TEMP, M_WAITOK);
1699 if (proc_vector == NULL)
1700 return (ENOMEM);
1701 error = proc_read_mem(td, p, vptr, proc_vector, size);
1702 if (error != 0) {
1703 free(proc_vector, M_TEMP);
1704 return (error);
1705 }
1706 *proc_vectorp = proc_vector;
1707 *vsizep = vsize;
1708
1709 return (0);
1710 }
1711
1712 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */
1713
1714 static int
1715 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1716 enum proc_vector_type type)
1717 {
1718 size_t done, len, nchr, vsize;
1719 int error, i;
1720 char **proc_vector, *sptr;
1721 char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1722
1723 PROC_ASSERT_HELD(p);
1724
1725 /*
1726 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1727 */
1728 nchr = 2 * (PATH_MAX + ARG_MAX);
1729
1730 error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1731 if (error != 0)
1732 return (error);
1733 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1734 /*
1735 * The program may have scribbled into its argv array, e.g. to
1736 * remove some arguments. If that has happened, break out
1737 * before trying to read from NULL.
1738 */
1739 if (proc_vector[i] == NULL)
1740 break;
1741 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1742 error = proc_read_string(td, p, sptr, pss_string,
1743 sizeof(pss_string));
1744 if (error != 0)
1745 goto done;
1746 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1747 if (done + len >= nchr)
1748 len = nchr - done - 1;
1749 sbuf_bcat(sb, pss_string, len);
1750 if (len != GET_PS_STRINGS_CHUNK_SZ)
1751 break;
1752 done += GET_PS_STRINGS_CHUNK_SZ;
1753 }
1754 sbuf_bcat(sb, "", 1);
1755 done += len + 1;
1756 }
1757 done:
1758 free(proc_vector, M_TEMP);
1759 return (error);
1760 }
1761
1762 int
1763 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1764 {
1765
1766 return (get_ps_strings(curthread, p, sb, PROC_ARG));
1767 }
1768
1769 int
1770 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1771 {
1772
1773 return (get_ps_strings(curthread, p, sb, PROC_ENV));
1774 }
1775
1776 int
1777 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1778 {
1779 size_t vsize, size;
1780 char **auxv;
1781 int error;
1782
1783 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1784 if (error == 0) {
1785 #ifdef COMPAT_FREEBSD32
1786 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1787 size = vsize * sizeof(Elf32_Auxinfo);
1788 else
1789 #endif
1790 size = vsize * sizeof(Elf_Auxinfo);
1791 if (sbuf_bcat(sb, auxv, size) != 0)
1792 error = ENOMEM;
1793 free(auxv, M_TEMP);
1794 }
1795 return (error);
1796 }
1797
1798 /*
1799 * This sysctl allows a process to retrieve the argument list or process
1800 * title for another process without groping around in the address space
1801 * of the other process. It also allow a process to set its own "process
1802 * title to a string of its own choice.
1803 */
1804 static int
1805 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1806 {
1807 int *name = (int *)arg1;
1808 u_int namelen = arg2;
1809 struct pargs *newpa, *pa;
1810 struct proc *p;
1811 struct sbuf sb;
1812 int flags, error = 0, error2;
1813
1814 if (namelen != 1)
1815 return (EINVAL);
1816
1817 flags = PGET_CANSEE;
1818 if (req->newptr != NULL)
1819 flags |= PGET_ISCURRENT;
1820 error = pget((pid_t)name[0], flags, &p);
1821 if (error)
1822 return (error);
1823
1824 pa = p->p_args;
1825 if (pa != NULL) {
1826 pargs_hold(pa);
1827 PROC_UNLOCK(p);
1828 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1829 pargs_drop(pa);
1830 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1831 _PHOLD(p);
1832 PROC_UNLOCK(p);
1833 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1834 error = proc_getargv(curthread, p, &sb);
1835 error2 = sbuf_finish(&sb);
1836 PRELE(p);
1837 sbuf_delete(&sb);
1838 if (error == 0 && error2 != 0)
1839 error = error2;
1840 } else {
1841 PROC_UNLOCK(p);
1842 }
1843 if (error != 0 || req->newptr == NULL)
1844 return (error);
1845
1846 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1847 return (ENOMEM);
1848 newpa = pargs_alloc(req->newlen);
1849 error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1850 if (error != 0) {
1851 pargs_free(newpa);
1852 return (error);
1853 }
1854 PROC_LOCK(p);
1855 pa = p->p_args;
1856 p->p_args = newpa;
1857 PROC_UNLOCK(p);
1858 pargs_drop(pa);
1859 return (0);
1860 }
1861
1862 /*
1863 * This sysctl allows a process to retrieve environment of another process.
1864 */
1865 static int
1866 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1867 {
1868 int *name = (int *)arg1;
1869 u_int namelen = arg2;
1870 struct proc *p;
1871 struct sbuf sb;
1872 int error, error2;
1873
1874 if (namelen != 1)
1875 return (EINVAL);
1876
1877 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1878 if (error != 0)
1879 return (error);
1880 if ((p->p_flag & P_SYSTEM) != 0) {
1881 PRELE(p);
1882 return (0);
1883 }
1884
1885 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1886 error = proc_getenvv(curthread, p, &sb);
1887 error2 = sbuf_finish(&sb);
1888 PRELE(p);
1889 sbuf_delete(&sb);
1890 return (error != 0 ? error : error2);
1891 }
1892
1893 /*
1894 * This sysctl allows a process to retrieve ELF auxiliary vector of
1895 * another process.
1896 */
1897 static int
1898 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1899 {
1900 int *name = (int *)arg1;
1901 u_int namelen = arg2;
1902 struct proc *p;
1903 struct sbuf sb;
1904 int error, error2;
1905
1906 if (namelen != 1)
1907 return (EINVAL);
1908
1909 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1910 if (error != 0)
1911 return (error);
1912 if ((p->p_flag & P_SYSTEM) != 0) {
1913 PRELE(p);
1914 return (0);
1915 }
1916 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1917 error = proc_getauxv(curthread, p, &sb);
1918 error2 = sbuf_finish(&sb);
1919 PRELE(p);
1920 sbuf_delete(&sb);
1921 return (error != 0 ? error : error2);
1922 }
1923
1924 /*
1925 * This sysctl allows a process to retrieve the path of the executable for
1926 * itself or another process.
1927 */
1928 static int
1929 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1930 {
1931 pid_t *pidp = (pid_t *)arg1;
1932 unsigned int arglen = arg2;
1933 struct proc *p;
1934 struct vnode *vp;
1935 char *retbuf, *freebuf;
1936 int error;
1937
1938 if (arglen != 1)
1939 return (EINVAL);
1940 if (*pidp == -1) { /* -1 means this process */
1941 p = req->td->td_proc;
1942 } else {
1943 error = pget(*pidp, PGET_CANSEE, &p);
1944 if (error != 0)
1945 return (error);
1946 }
1947
1948 vp = p->p_textvp;
1949 if (vp == NULL) {
1950 if (*pidp != -1)
1951 PROC_UNLOCK(p);
1952 return (0);
1953 }
1954 vref(vp);
1955 if (*pidp != -1)
1956 PROC_UNLOCK(p);
1957 error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1958 vrele(vp);
1959 if (error)
1960 return (error);
1961 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1962 free(freebuf, M_TEMP);
1963 return (error);
1964 }
1965
1966 static int
1967 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1968 {
1969 struct proc *p;
1970 char *sv_name;
1971 int *name;
1972 int namelen;
1973 int error;
1974
1975 namelen = arg2;
1976 if (namelen != 1)
1977 return (EINVAL);
1978
1979 name = (int *)arg1;
1980 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1981 if (error != 0)
1982 return (error);
1983 sv_name = p->p_sysent->sv_name;
1984 PROC_UNLOCK(p);
1985 return (sysctl_handle_string(oidp, sv_name, 0, req));
1986 }
1987
1988 #ifdef KINFO_OVMENTRY_SIZE
1989 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1990 #endif
1991
1992 #ifdef COMPAT_FREEBSD7
1993 static int
1994 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1995 {
1996 vm_map_entry_t entry, tmp_entry;
1997 unsigned int last_timestamp;
1998 char *fullpath, *freepath;
1999 struct kinfo_ovmentry *kve;
2000 struct vattr va;
2001 struct ucred *cred;
2002 int error, *name;
2003 struct vnode *vp;
2004 struct proc *p;
2005 vm_map_t map;
2006 struct vmspace *vm;
2007
2008 name = (int *)arg1;
2009 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2010 if (error != 0)
2011 return (error);
2012 vm = vmspace_acquire_ref(p);
2013 if (vm == NULL) {
2014 PRELE(p);
2015 return (ESRCH);
2016 }
2017 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2018
2019 map = &vm->vm_map;
2020 vm_map_lock_read(map);
2021 for (entry = map->header.next; entry != &map->header;
2022 entry = entry->next) {
2023 vm_object_t obj, tobj, lobj;
2024 vm_offset_t addr;
2025
2026 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2027 continue;
2028
2029 bzero(kve, sizeof(*kve));
2030 kve->kve_structsize = sizeof(*kve);
2031
2032 kve->kve_private_resident = 0;
2033 obj = entry->object.vm_object;
2034 if (obj != NULL) {
2035 VM_OBJECT_RLOCK(obj);
2036 if (obj->shadow_count == 1)
2037 kve->kve_private_resident =
2038 obj->resident_page_count;
2039 }
2040 kve->kve_resident = 0;
2041 addr = entry->start;
2042 while (addr < entry->end) {
2043 if (pmap_extract(map->pmap, addr))
2044 kve->kve_resident++;
2045 addr += PAGE_SIZE;
2046 }
2047
2048 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2049 if (tobj != obj)
2050 VM_OBJECT_RLOCK(tobj);
2051 if (lobj != obj)
2052 VM_OBJECT_RUNLOCK(lobj);
2053 lobj = tobj;
2054 }
2055
2056 kve->kve_start = (void*)entry->start;
2057 kve->kve_end = (void*)entry->end;
2058 kve->kve_offset = (off_t)entry->offset;
2059
2060 if (entry->protection & VM_PROT_READ)
2061 kve->kve_protection |= KVME_PROT_READ;
2062 if (entry->protection & VM_PROT_WRITE)
2063 kve->kve_protection |= KVME_PROT_WRITE;
2064 if (entry->protection & VM_PROT_EXECUTE)
2065 kve->kve_protection |= KVME_PROT_EXEC;
2066
2067 if (entry->eflags & MAP_ENTRY_COW)
2068 kve->kve_flags |= KVME_FLAG_COW;
2069 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2070 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2071 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2072 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2073
2074 last_timestamp = map->timestamp;
2075 vm_map_unlock_read(map);
2076
2077 kve->kve_fileid = 0;
2078 kve->kve_fsid = 0;
2079 freepath = NULL;
2080 fullpath = "";
2081 if (lobj) {
2082 vp = NULL;
2083 switch (lobj->type) {
2084 case OBJT_DEFAULT:
2085 kve->kve_type = KVME_TYPE_DEFAULT;
2086 break;
2087 case OBJT_VNODE:
2088 kve->kve_type = KVME_TYPE_VNODE;
2089 vp = lobj->handle;
2090 vref(vp);
2091 break;
2092 case OBJT_SWAP:
2093 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2094 kve->kve_type = KVME_TYPE_VNODE;
2095 if ((lobj->flags & OBJ_TMPFS) != 0) {
2096 vp = lobj->un_pager.swp.swp_tmpfs;
2097 vref(vp);
2098 }
2099 } else {
2100 kve->kve_type = KVME_TYPE_SWAP;
2101 }
2102 break;
2103 case OBJT_DEVICE:
2104 kve->kve_type = KVME_TYPE_DEVICE;
2105 break;
2106 case OBJT_PHYS:
2107 kve->kve_type = KVME_TYPE_PHYS;
2108 break;
2109 case OBJT_DEAD:
2110 kve->kve_type = KVME_TYPE_DEAD;
2111 break;
2112 case OBJT_SG:
2113 kve->kve_type = KVME_TYPE_SG;
2114 break;
2115 default:
2116 kve->kve_type = KVME_TYPE_UNKNOWN;
2117 break;
2118 }
2119 if (lobj != obj)
2120 VM_OBJECT_RUNLOCK(lobj);
2121
2122 kve->kve_ref_count = obj->ref_count;
2123 kve->kve_shadow_count = obj->shadow_count;
2124 VM_OBJECT_RUNLOCK(obj);
2125 if (vp != NULL) {
2126 vn_fullpath(curthread, vp, &fullpath,
2127 &freepath);
2128 cred = curthread->td_ucred;
2129 vn_lock(vp, LK_SHARED | LK_RETRY);
2130 if (VOP_GETATTR(vp, &va, cred) == 0) {
2131 kve->kve_fileid = va.va_fileid;
2132 kve->kve_fsid = va.va_fsid;
2133 }
2134 vput(vp);
2135 }
2136 } else {
2137 kve->kve_type = KVME_TYPE_NONE;
2138 kve->kve_ref_count = 0;
2139 kve->kve_shadow_count = 0;
2140 }
2141
2142 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2143 if (freepath != NULL)
2144 free(freepath, M_TEMP);
2145
2146 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2147 vm_map_lock_read(map);
2148 if (error)
2149 break;
2150 if (last_timestamp != map->timestamp) {
2151 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2152 entry = tmp_entry;
2153 }
2154 }
2155 vm_map_unlock_read(map);
2156 vmspace_free(vm);
2157 PRELE(p);
2158 free(kve, M_TEMP);
2159 return (error);
2160 }
2161 #endif /* COMPAT_FREEBSD7 */
2162
2163 #ifdef KINFO_VMENTRY_SIZE
2164 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2165 #endif
2166
2167 static void
2168 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2169 struct kinfo_vmentry *kve)
2170 {
2171 vm_object_t obj, tobj;
2172 vm_page_t m, m_adv;
2173 vm_offset_t addr;
2174 vm_paddr_t locked_pa;
2175 vm_pindex_t pi, pi_adv, pindex;
2176
2177 locked_pa = 0;
2178 obj = entry->object.vm_object;
2179 addr = entry->start;
2180 m_adv = NULL;
2181 pi = OFF_TO_IDX(entry->offset);
2182 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2183 if (m_adv != NULL) {
2184 m = m_adv;
2185 } else {
2186 pi_adv = OFF_TO_IDX(entry->end - addr);
2187 pindex = pi;
2188 for (tobj = obj;; tobj = tobj->backing_object) {
2189 m = vm_page_find_least(tobj, pindex);
2190 if (m != NULL) {
2191 if (m->pindex == pindex)
2192 break;
2193 if (pi_adv > m->pindex - pindex) {
2194 pi_adv = m->pindex - pindex;
2195 m_adv = m;
2196 }
2197 }
2198 if (tobj->backing_object == NULL)
2199 goto next;
2200 pindex += OFF_TO_IDX(tobj->
2201 backing_object_offset);
2202 }
2203 }
2204 m_adv = NULL;
2205 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2206 (addr & (pagesizes[1] - 1)) == 0 &&
2207 (pmap_mincore(map->pmap, addr, &locked_pa) &
2208 MINCORE_SUPER) != 0) {
2209 kve->kve_flags |= KVME_FLAG_SUPER;
2210 pi_adv = OFF_TO_IDX(pagesizes[1]);
2211 } else {
2212 /*
2213 * We do not test the found page on validity.
2214 * Either the page is busy and being paged in,
2215 * or it was invalidated. The first case
2216 * should be counted as resident, the second
2217 * is not so clear; we do account both.
2218 */
2219 pi_adv = 1;
2220 }
2221 kve->kve_resident += pi_adv;
2222 next:;
2223 }
2224 PA_UNLOCK_COND(locked_pa);
2225 }
2226
2227 /*
2228 * Must be called with the process locked and will return unlocked.
2229 */
2230 int
2231 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2232 {
2233 vm_map_entry_t entry, tmp_entry;
2234 struct vattr va;
2235 vm_map_t map;
2236 vm_object_t obj, tobj, lobj;
2237 char *fullpath, *freepath;
2238 struct kinfo_vmentry *kve;
2239 struct ucred *cred;
2240 struct vnode *vp;
2241 struct vmspace *vm;
2242 vm_offset_t addr;
2243 unsigned int last_timestamp;
2244 int error;
2245
2246 PROC_LOCK_ASSERT(p, MA_OWNED);
2247
2248 _PHOLD(p);
2249 PROC_UNLOCK(p);
2250 vm = vmspace_acquire_ref(p);
2251 if (vm == NULL) {
2252 PRELE(p);
2253 return (ESRCH);
2254 }
2255 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2256
2257 error = 0;
2258 map = &vm->vm_map;
2259 vm_map_lock_read(map);
2260 for (entry = map->header.next; entry != &map->header;
2261 entry = entry->next) {
2262 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2263 continue;
2264
2265 addr = entry->end;
2266 bzero(kve, sizeof(*kve));
2267 obj = entry->object.vm_object;
2268 if (obj != NULL) {
2269 for (tobj = obj; tobj != NULL;
2270 tobj = tobj->backing_object) {
2271 VM_OBJECT_RLOCK(tobj);
2272 lobj = tobj;
2273 }
2274 if (obj->backing_object == NULL)
2275 kve->kve_private_resident =
2276 obj->resident_page_count;
2277 if (!vmmap_skip_res_cnt)
2278 kern_proc_vmmap_resident(map, entry, kve);
2279 for (tobj = obj; tobj != NULL;
2280 tobj = tobj->backing_object) {
2281 if (tobj != obj && tobj != lobj)
2282 VM_OBJECT_RUNLOCK(tobj);
2283 }
2284 } else {
2285 lobj = NULL;
2286 }
2287
2288 kve->kve_start = entry->start;
2289 kve->kve_end = entry->end;
2290 kve->kve_offset = entry->offset;
2291
2292 if (entry->protection & VM_PROT_READ)
2293 kve->kve_protection |= KVME_PROT_READ;
2294 if (entry->protection & VM_PROT_WRITE)
2295 kve->kve_protection |= KVME_PROT_WRITE;
2296 if (entry->protection & VM_PROT_EXECUTE)
2297 kve->kve_protection |= KVME_PROT_EXEC;
2298
2299 if (entry->eflags & MAP_ENTRY_COW)
2300 kve->kve_flags |= KVME_FLAG_COW;
2301 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2302 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2303 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2304 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2305 if (entry->eflags & MAP_ENTRY_GROWS_UP)
2306 kve->kve_flags |= KVME_FLAG_GROWS_UP;
2307 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2308 kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2309
2310 last_timestamp = map->timestamp;
2311 vm_map_unlock_read(map);
2312
2313 freepath = NULL;
2314 fullpath = "";
2315 if (lobj != NULL) {
2316 vp = NULL;
2317 switch (lobj->type) {
2318 case OBJT_DEFAULT:
2319 kve->kve_type = KVME_TYPE_DEFAULT;
2320 break;
2321 case OBJT_VNODE:
2322 kve->kve_type = KVME_TYPE_VNODE;
2323 vp = lobj->handle;
2324 vref(vp);
2325 break;
2326 case OBJT_SWAP:
2327 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2328 kve->kve_type = KVME_TYPE_VNODE;
2329 if ((lobj->flags & OBJ_TMPFS) != 0) {
2330 vp = lobj->un_pager.swp.swp_tmpfs;
2331 vref(vp);
2332 }
2333 } else {
2334 kve->kve_type = KVME_TYPE_SWAP;
2335 }
2336 break;
2337 case OBJT_DEVICE:
2338 kve->kve_type = KVME_TYPE_DEVICE;
2339 break;
2340 case OBJT_PHYS:
2341 kve->kve_type = KVME_TYPE_PHYS;
2342 break;
2343 case OBJT_DEAD:
2344 kve->kve_type = KVME_TYPE_DEAD;
2345 break;
2346 case OBJT_SG:
2347 kve->kve_type = KVME_TYPE_SG;
2348 break;
2349 case OBJT_MGTDEVICE:
2350 kve->kve_type = KVME_TYPE_MGTDEVICE;
2351 break;
2352 default:
2353 kve->kve_type = KVME_TYPE_UNKNOWN;
2354 break;
2355 }
2356 if (lobj != obj)
2357 VM_OBJECT_RUNLOCK(lobj);
2358
2359 kve->kve_ref_count = obj->ref_count;
2360 kve->kve_shadow_count = obj->shadow_count;
2361 VM_OBJECT_RUNLOCK(obj);
2362 if (vp != NULL) {
2363 vn_fullpath(curthread, vp, &fullpath,
2364 &freepath);
2365 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2366 cred = curthread->td_ucred;
2367 vn_lock(vp, LK_SHARED | LK_RETRY);
2368 if (VOP_GETATTR(vp, &va, cred) == 0) {
2369 kve->kve_vn_fileid = va.va_fileid;
2370 kve->kve_vn_fsid = va.va_fsid;
2371 kve->kve_vn_mode =
2372 MAKEIMODE(va.va_type, va.va_mode);
2373 kve->kve_vn_size = va.va_size;
2374 kve->kve_vn_rdev = va.va_rdev;
2375 kve->kve_status = KF_ATTR_VALID;
2376 }
2377 vput(vp);
2378 }
2379 } else {
2380 kve->kve_type = KVME_TYPE_NONE;
2381 kve->kve_ref_count = 0;
2382 kve->kve_shadow_count = 0;
2383 }
2384
2385 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2386 if (freepath != NULL)
2387 free(freepath, M_TEMP);
2388
2389 /* Pack record size down */
2390 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2391 kve->kve_structsize =
2392 offsetof(struct kinfo_vmentry, kve_path) +
2393 strlen(kve->kve_path) + 1;
2394 else
2395 kve->kve_structsize = sizeof(*kve);
2396 kve->kve_structsize = roundup(kve->kve_structsize,
2397 sizeof(uint64_t));
2398
2399 /* Halt filling and truncate rather than exceeding maxlen */
2400 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2401 error = 0;
2402 vm_map_lock_read(map);
2403 break;
2404 } else if (maxlen != -1)
2405 maxlen -= kve->kve_structsize;
2406
2407 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2408 error = ENOMEM;
2409 vm_map_lock_read(map);
2410 if (error != 0)
2411 break;
2412 if (last_timestamp != map->timestamp) {
2413 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2414 entry = tmp_entry;
2415 }
2416 }
2417 vm_map_unlock_read(map);
2418 vmspace_free(vm);
2419 PRELE(p);
2420 free(kve, M_TEMP);
2421 return (error);
2422 }
2423
2424 static int
2425 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2426 {
2427 struct proc *p;
2428 struct sbuf sb;
2429 int error, error2, *name;
2430
2431 name = (int *)arg1;
2432 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2433 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2434 if (error != 0) {
2435 sbuf_delete(&sb);
2436 return (error);
2437 }
2438 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2439 error2 = sbuf_finish(&sb);
2440 sbuf_delete(&sb);
2441 return (error != 0 ? error : error2);
2442 }
2443
2444 #if defined(STACK) || defined(DDB)
2445 static int
2446 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2447 {
2448 struct kinfo_kstack *kkstp;
2449 int error, i, *name, numthreads;
2450 lwpid_t *lwpidarray;
2451 struct thread *td;
2452 struct stack *st;
2453 struct sbuf sb;
2454 struct proc *p;
2455
2456 name = (int *)arg1;
2457 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2458 if (error != 0)
2459 return (error);
2460
2461 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2462 st = stack_create();
2463
2464 lwpidarray = NULL;
2465 numthreads = 0;
2466 PROC_LOCK(p);
2467 repeat:
2468 if (numthreads < p->p_numthreads) {
2469 if (lwpidarray != NULL) {
2470 free(lwpidarray, M_TEMP);
2471 lwpidarray = NULL;
2472 }
2473 numthreads = p->p_numthreads;
2474 PROC_UNLOCK(p);
2475 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2476 M_WAITOK | M_ZERO);
2477 PROC_LOCK(p);
2478 goto repeat;
2479 }
2480 i = 0;
2481
2482 /*
2483 * XXXRW: During the below loop, execve(2) and countless other sorts
2484 * of changes could have taken place. Should we check to see if the
2485 * vmspace has been replaced, or the like, in order to prevent
2486 * giving a snapshot that spans, say, execve(2), with some threads
2487 * before and some after? Among other things, the credentials could
2488 * have changed, in which case the right to extract debug info might
2489 * no longer be assured.
2490 */
2491 FOREACH_THREAD_IN_PROC(p, td) {
2492 KASSERT(i < numthreads,
2493 ("sysctl_kern_proc_kstack: numthreads"));
2494 lwpidarray[i] = td->td_tid;
2495 i++;
2496 }
2497 numthreads = i;
2498 for (i = 0; i < numthreads; i++) {
2499 td = thread_find(p, lwpidarray[i]);
2500 if (td == NULL) {
2501 continue;
2502 }
2503 bzero(kkstp, sizeof(*kkstp));
2504 (void)sbuf_new(&sb, kkstp->kkst_trace,
2505 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2506 thread_lock(td);
2507 kkstp->kkst_tid = td->td_tid;
2508 if (TD_IS_SWAPPED(td))
2509 kkstp->kkst_state = KKST_STATE_SWAPPED;
2510 else if (TD_IS_RUNNING(td))
2511 kkstp->kkst_state = KKST_STATE_RUNNING;
2512 else {
2513 kkstp->kkst_state = KKST_STATE_STACKOK;
2514 stack_save_td(st, td);
2515 }
2516 thread_unlock(td);
2517 PROC_UNLOCK(p);
2518 stack_sbuf_print(&sb, st);
2519 sbuf_finish(&sb);
2520 sbuf_delete(&sb);
2521 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2522 PROC_LOCK(p);
2523 if (error)
2524 break;
2525 }
2526 _PRELE(p);
2527 PROC_UNLOCK(p);
2528 if (lwpidarray != NULL)
2529 free(lwpidarray, M_TEMP);
2530 stack_destroy(st);
2531 free(kkstp, M_TEMP);
2532 return (error);
2533 }
2534 #endif
2535
2536 /*
2537 * This sysctl allows a process to retrieve the full list of groups from
2538 * itself or another process.
2539 */
2540 static int
2541 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2542 {
2543 pid_t *pidp = (pid_t *)arg1;
2544 unsigned int arglen = arg2;
2545 struct proc *p;
2546 struct ucred *cred;
2547 int error;
2548
2549 if (arglen != 1)
2550 return (EINVAL);
2551 if (*pidp == -1) { /* -1 means this process */
2552 p = req->td->td_proc;
2553 } else {
2554 error = pget(*pidp, PGET_CANSEE, &p);
2555 if (error != 0)
2556 return (error);
2557 }
2558
2559 cred = crhold(p->p_ucred);
2560 if (*pidp != -1)
2561 PROC_UNLOCK(p);
2562
2563 error = SYSCTL_OUT(req, cred->cr_groups,
2564 cred->cr_ngroups * sizeof(gid_t));
2565 crfree(cred);
2566 return (error);
2567 }
2568
2569 /*
2570 * This sysctl allows a process to retrieve or/and set the resource limit for
2571 * another process.
2572 */
2573 static int
2574 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2575 {
2576 int *name = (int *)arg1;
2577 u_int namelen = arg2;
2578 struct rlimit rlim;
2579 struct proc *p;
2580 u_int which;
2581 int flags, error;
2582
2583 if (namelen != 2)
2584 return (EINVAL);
2585
2586 which = (u_int)name[1];
2587 if (which >= RLIM_NLIMITS)
2588 return (EINVAL);
2589
2590 if (req->newptr != NULL && req->newlen != sizeof(rlim))
2591 return (EINVAL);
2592
2593 flags = PGET_HOLD | PGET_NOTWEXIT;
2594 if (req->newptr != NULL)
2595 flags |= PGET_CANDEBUG;
2596 else
2597 flags |= PGET_CANSEE;
2598 error = pget((pid_t)name[0], flags, &p);
2599 if (error != 0)
2600 return (error);
2601
2602 /*
2603 * Retrieve limit.
2604 */
2605 if (req->oldptr != NULL) {
2606 PROC_LOCK(p);
2607 lim_rlimit(p, which, &rlim);
2608 PROC_UNLOCK(p);
2609 }
2610 error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2611 if (error != 0)
2612 goto errout;
2613
2614 /*
2615 * Set limit.
2616 */
2617 if (req->newptr != NULL) {
2618 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2619 if (error == 0)
2620 error = kern_proc_setrlimit(curthread, p, which, &rlim);
2621 }
2622
2623 errout:
2624 PRELE(p);
2625 return (error);
2626 }
2627
2628 /*
2629 * This sysctl allows a process to retrieve ps_strings structure location of
2630 * another process.
2631 */
2632 static int
2633 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2634 {
2635 int *name = (int *)arg1;
2636 u_int namelen = arg2;
2637 struct proc *p;
2638 vm_offset_t ps_strings;
2639 int error;
2640 #ifdef COMPAT_FREEBSD32
2641 uint32_t ps_strings32;
2642 #endif
2643
2644 if (namelen != 1)
2645 return (EINVAL);
2646
2647 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2648 if (error != 0)
2649 return (error);
2650 #ifdef COMPAT_FREEBSD32
2651 if ((req->flags & SCTL_MASK32) != 0) {
2652 /*
2653 * We return 0 if the 32 bit emulation request is for a 64 bit
2654 * process.
2655 */
2656 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2657 PTROUT(p->p_sysent->sv_psstrings) : 0;
2658 PROC_UNLOCK(p);
2659 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2660 return (error);
2661 }
2662 #endif
2663 ps_strings = p->p_sysent->sv_psstrings;
2664 PROC_UNLOCK(p);
2665 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2666 return (error);
2667 }
2668
2669 /*
2670 * This sysctl allows a process to retrieve umask of another process.
2671 */
2672 static int
2673 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2674 {
2675 int *name = (int *)arg1;
2676 u_int namelen = arg2;
2677 struct proc *p;
2678 int error;
2679 u_short fd_cmask;
2680
2681 if (namelen != 1)
2682 return (EINVAL);
2683
2684 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2685 if (error != 0)
2686 return (error);
2687
2688 FILEDESC_SLOCK(p->p_fd);
2689 fd_cmask = p->p_fd->fd_cmask;
2690 FILEDESC_SUNLOCK(p->p_fd);
2691 PRELE(p);
2692 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2693 return (error);
2694 }
2695
2696 /*
2697 * This sysctl allows a process to set and retrieve binary osreldate of
2698 * another process.
2699 */
2700 static int
2701 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2702 {
2703 int *name = (int *)arg1;
2704 u_int namelen = arg2;
2705 struct proc *p;
2706 int flags, error, osrel;
2707
2708 if (namelen != 1)
2709 return (EINVAL);
2710
2711 if (req->newptr != NULL && req->newlen != sizeof(osrel))
2712 return (EINVAL);
2713
2714 flags = PGET_HOLD | PGET_NOTWEXIT;
2715 if (req->newptr != NULL)
2716 flags |= PGET_CANDEBUG;
2717 else
2718 flags |= PGET_CANSEE;
2719 error = pget((pid_t)name[0], flags, &p);
2720 if (error != 0)
2721 return (error);
2722
2723 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2724 if (error != 0)
2725 goto errout;
2726
2727 if (req->newptr != NULL) {
2728 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2729 if (error != 0)
2730 goto errout;
2731 if (osrel < 0) {
2732 error = EINVAL;
2733 goto errout;
2734 }
2735 p->p_osrel = osrel;
2736 }
2737 errout:
2738 PRELE(p);
2739 return (error);
2740 }
2741
2742 static int
2743 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2744 {
2745 int *name = (int *)arg1;
2746 u_int namelen = arg2;
2747 struct proc *p;
2748 struct kinfo_sigtramp kst;
2749 const struct sysentvec *sv;
2750 int error;
2751 #ifdef COMPAT_FREEBSD32
2752 struct kinfo_sigtramp32 kst32;
2753 #endif
2754
2755 if (namelen != 1)
2756 return (EINVAL);
2757
2758 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2759 if (error != 0)
2760 return (error);
2761 sv = p->p_sysent;
2762 #ifdef COMPAT_FREEBSD32
2763 if ((req->flags & SCTL_MASK32) != 0) {
2764 bzero(&kst32, sizeof(kst32));
2765 if (SV_PROC_FLAG(p, SV_ILP32)) {
2766 if (sv->sv_sigcode_base != 0) {
2767 kst32.ksigtramp_start = sv->sv_sigcode_base;
2768 kst32.ksigtramp_end = sv->sv_sigcode_base +
2769 *sv->sv_szsigcode;
2770 } else {
2771 kst32.ksigtramp_start = sv->sv_psstrings -
2772 *sv->sv_szsigcode;
2773 kst32.ksigtramp_end = sv->sv_psstrings;
2774 }
2775 }
2776 PROC_UNLOCK(p);
2777 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2778 return (error);
2779 }
2780 #endif
2781 bzero(&kst, sizeof(kst));
2782 if (sv->sv_sigcode_base != 0) {
2783 kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2784 kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2785 *sv->sv_szsigcode;
2786 } else {
2787 kst.ksigtramp_start = (char *)sv->sv_psstrings -
2788 *sv->sv_szsigcode;
2789 kst.ksigtramp_end = (char *)sv->sv_psstrings;
2790 }
2791 PROC_UNLOCK(p);
2792 error = SYSCTL_OUT(req, &kst, sizeof(kst));
2793 return (error);
2794 }
2795
2796 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table");
2797
2798 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2799 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2800 "Return entire process table");
2801
2802 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2803 sysctl_kern_proc, "Process table");
2804
2805 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2806 sysctl_kern_proc, "Process table");
2807
2808 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2809 sysctl_kern_proc, "Process table");
2810
2811 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2812 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2813
2814 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2815 sysctl_kern_proc, "Process table");
2816
2817 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2818 sysctl_kern_proc, "Process table");
2819
2820 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2821 sysctl_kern_proc, "Process table");
2822
2823 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2824 sysctl_kern_proc, "Process table");
2825
2826 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2827 sysctl_kern_proc, "Return process table, no threads");
2828
2829 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2830 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2831 sysctl_kern_proc_args, "Process argument list");
2832
2833 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2834 sysctl_kern_proc_env, "Process environment");
2835
2836 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2837 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2838
2839 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2840 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2841
2842 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2843 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2844 "Process syscall vector name (ABI type)");
2845
2846 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2847 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2848
2849 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2850 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2851
2852 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2853 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2854
2855 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2856 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2857
2858 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2859 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2860
2861 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2862 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2863
2864 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2865 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2866
2867 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2868 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2869
2870 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2871 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2872 "Return process table, no threads");
2873
2874 #ifdef COMPAT_FREEBSD7
2875 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2876 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2877 #endif
2878
2879 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2880 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2881
2882 #if defined(STACK) || defined(DDB)
2883 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2884 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2885 #endif
2886
2887 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2888 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2889
2890 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2891 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2892 "Process resource limits");
2893
2894 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2895 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2896 "Process ps_strings location");
2897
2898 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2899 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2900
2901 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2902 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2903 "Process binary osreldate");
2904
2905 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2906 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2907 "Process signal trampoline location");
2908
2909 int allproc_gen;
2910
2911 void
2912 stop_all_proc(void)
2913 {
2914 struct proc *cp, *p;
2915 int r, gen;
2916 bool restart, seen_stopped, seen_exiting, stopped_some;
2917
2918 cp = curproc;
2919 /*
2920 * stop_all_proc() assumes that all process which have
2921 * usermode must be stopped, except current process, for
2922 * obvious reasons. Since other threads in the process
2923 * establishing global stop could unstop something, disable
2924 * calls from multithreaded processes as precaution. The
2925 * service must not be user-callable anyway.
2926 */
2927 KASSERT((cp->p_flag & P_HADTHREADS) == 0 ||
2928 (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc"));
2929
2930 allproc_loop:
2931 sx_xlock(&allproc_lock);
2932 gen = allproc_gen;
2933 seen_exiting = seen_stopped = stopped_some = restart = false;
2934 LIST_REMOVE(cp, p_list);
2935 LIST_INSERT_HEAD(&allproc, cp, p_list);
2936 for (;;) {
2937 p = LIST_NEXT(cp, p_list);
2938 if (p == NULL)
2939 break;
2940 LIST_REMOVE(cp, p_list);
2941 LIST_INSERT_AFTER(p, cp, p_list);
2942 PROC_LOCK(p);
2943 if ((p->p_flag & (P_KTHREAD | P_SYSTEM |
2944 P_TOTAL_STOP)) != 0) {
2945 PROC_UNLOCK(p);
2946 continue;
2947 }
2948 if ((p->p_flag & P_WEXIT) != 0) {
2949 seen_exiting = true;
2950 PROC_UNLOCK(p);
2951 continue;
2952 }
2953 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
2954 /*
2955 * Stopped processes are tolerated when there
2956 * are no other processes which might continue
2957 * them. P_STOPPED_SINGLE but not
2958 * P_TOTAL_STOP process still has at least one
2959 * thread running.
2960 */
2961 seen_stopped = true;
2962 PROC_UNLOCK(p);
2963 continue;
2964 }
2965 _PHOLD(p);
2966 sx_xunlock(&allproc_lock);
2967 r = thread_single(p, SINGLE_ALLPROC);
2968 if (r != 0)
2969 restart = true;
2970 else
2971 stopped_some = true;
2972 _PRELE(p);
2973 PROC_UNLOCK(p);
2974 sx_xlock(&allproc_lock);
2975 }
2976 /* Catch forked children we did not see in iteration. */
2977 if (gen != allproc_gen)
2978 restart = true;
2979 sx_xunlock(&allproc_lock);
2980 if (restart || stopped_some || seen_exiting || seen_stopped) {
2981 kern_yield(PRI_USER);
2982 goto allproc_loop;
2983 }
2984 }
2985
2986 void
2987 resume_all_proc(void)
2988 {
2989 struct proc *cp, *p;
2990
2991 cp = curproc;
2992 sx_xlock(&allproc_lock);
2993 LIST_REMOVE(cp, p_list);
2994 LIST_INSERT_HEAD(&allproc, cp, p_list);
2995 for (;;) {
2996 p = LIST_NEXT(cp, p_list);
2997 if (p == NULL)
2998 break;
2999 LIST_REMOVE(cp, p_list);
3000 LIST_INSERT_AFTER(p, cp, p_list);
3001 PROC_LOCK(p);
3002 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3003 sx_xunlock(&allproc_lock);
3004 _PHOLD(p);
3005 thread_single_end(p, SINGLE_ALLPROC);
3006 _PRELE(p);
3007 PROC_UNLOCK(p);
3008 sx_xlock(&allproc_lock);
3009 } else {
3010 PROC_UNLOCK(p);
3011 }
3012 }
3013 sx_xunlock(&allproc_lock);
3014 }
3015
3016 #define TOTAL_STOP_DEBUG 1
3017 #ifdef TOTAL_STOP_DEBUG
3018 volatile static int ap_resume;
3019 #include <sys/mount.h>
3020
3021 static int
3022 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3023 {
3024 int error, val;
3025
3026 val = 0;
3027 ap_resume = 0;
3028 error = sysctl_handle_int(oidp, &val, 0, req);
3029 if (error != 0 || req->newptr == NULL)
3030 return (error);
3031 if (val != 0) {
3032 stop_all_proc();
3033 syncer_suspend();
3034 while (ap_resume == 0)
3035 ;
3036 syncer_resume();
3037 resume_all_proc();
3038 }
3039 return (0);
3040 }
3041
3042 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3043 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3044 sysctl_debug_stop_all_proc, "I",
3045 "");
3046 #endif
Cache object: acf239933f8b1daeb8a2faad38b3b606
|