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