2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1989, 1991, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/bitstring.h>
46 #include <sys/eventhandler.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
52 #include <sys/loginclass.h>
53 #include <sys/malloc.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
66 #include <sys/stack.h>
68 #include <sys/sysctl.h>
69 #include <sys/filedesc.h>
71 #include <sys/signalvar.h>
75 #include <sys/vnode.h>
83 #include <vm/vm_param.h>
84 #include <vm/vm_extern.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_object.h>
88 #include <vm/vm_page.h>
91 #ifdef COMPAT_FREEBSD32
92 #include <compat/freebsd32/freebsd32.h>
93 #include <compat/freebsd32/freebsd32_util.h>
96 SDT_PROVIDER_DEFINE(proc);
98 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
99 MALLOC_DEFINE(M_SESSION, "session", "session header");
100 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
101 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
103 static void doenterpgrp(struct proc *, struct pgrp *);
104 static void orphanpg(struct pgrp *pg);
105 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
106 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
107 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
109 static void pgadjustjobc(struct pgrp *pgrp, int entering);
110 static void pgdelete(struct pgrp *);
111 static int proc_ctor(void *mem, int size, void *arg, int flags);
112 static void proc_dtor(void *mem, int size, void *arg);
113 static int proc_init(void *mem, int size, int flags);
114 static void proc_fini(void *mem, int size);
115 static void pargs_free(struct pargs *pa);
118 * Other process lists
120 struct pidhashhead *pidhashtbl;
121 struct sx *pidhashtbl_lock;
124 struct pgrphashhead *pgrphashtbl;
126 struct proclist allproc;
127 struct sx __exclusive_cache_line allproc_lock;
128 struct sx __exclusive_cache_line proctree_lock;
129 struct mtx __exclusive_cache_line ppeers_lock;
130 struct mtx __exclusive_cache_line procid_lock;
131 uma_zone_t proc_zone;
134 * The offset of various fields in struct proc and struct thread.
135 * These are used by kernel debuggers to enumerate kernel threads and
138 const int proc_off_p_pid = offsetof(struct proc, p_pid);
139 const int proc_off_p_comm = offsetof(struct proc, p_comm);
140 const int proc_off_p_list = offsetof(struct proc, p_list);
141 const int proc_off_p_threads = offsetof(struct proc, p_threads);
142 const int thread_off_td_tid = offsetof(struct thread, td_tid);
143 const int thread_off_td_name = offsetof(struct thread, td_name);
144 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
145 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
146 const int thread_off_td_plist = offsetof(struct thread, td_plist);
148 EVENTHANDLER_LIST_DEFINE(process_ctor);
149 EVENTHANDLER_LIST_DEFINE(process_dtor);
150 EVENTHANDLER_LIST_DEFINE(process_init);
151 EVENTHANDLER_LIST_DEFINE(process_fini);
152 EVENTHANDLER_LIST_DEFINE(process_exit);
153 EVENTHANDLER_LIST_DEFINE(process_fork);
154 EVENTHANDLER_LIST_DEFINE(process_exec);
156 int kstack_pages = KSTACK_PAGES;
157 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
158 "Kernel stack size in pages");
159 static int vmmap_skip_res_cnt = 0;
160 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
161 &vmmap_skip_res_cnt, 0,
162 "Skip calculation of the pages resident count in kern.proc.vmmap");
164 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
165 #ifdef COMPAT_FREEBSD32
166 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
170 * Initialize global process hashing structures.
177 sx_init(&allproc_lock, "allproc");
178 sx_init(&proctree_lock, "proctree");
179 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
180 mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
182 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
183 pidhashlock = (pidhash + 1) / 64;
186 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
187 M_PROC, M_WAITOK | M_ZERO);
188 for (i = 0; i < pidhashlock + 1; i++)
189 sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
190 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
191 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
192 proc_ctor, proc_dtor, proc_init, proc_fini,
193 UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
198 * Prepare a proc for use.
201 proc_ctor(void *mem, int size, void *arg, int flags)
206 p = (struct proc *)mem;
207 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
208 td = FIRST_THREAD_IN_PROC(p);
210 /* Make sure all thread constructors are executed */
211 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
217 * Reclaim a proc after use.
220 proc_dtor(void *mem, int size, void *arg)
225 /* INVARIANTS checks go here */
226 p = (struct proc *)mem;
227 td = FIRST_THREAD_IN_PROC(p);
230 KASSERT((p->p_numthreads == 1),
231 ("bad number of threads in exiting process"));
232 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
234 /* Free all OSD associated to this thread. */
236 td_softdep_cleanup(td);
237 MPASS(td->td_su == NULL);
239 /* Make sure all thread destructors are executed */
240 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
242 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
243 if (p->p_ksi != NULL)
244 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
248 * Initialize type-stable parts of a proc (when newly created).
251 proc_init(void *mem, int size, int flags)
255 p = (struct proc *)mem;
256 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
257 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
258 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
259 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
260 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
261 cv_init(&p->p_pwait, "ppwait");
262 TAILQ_INIT(&p->p_threads); /* all threads in proc */
263 EVENTHANDLER_DIRECT_INVOKE(process_init, p);
264 p->p_stats = pstats_alloc();
270 * UMA should ensure that this function is never called.
271 * Freeing a proc structure would violate type stability.
274 proc_fini(void *mem, int size)
279 p = (struct proc *)mem;
280 EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
281 pstats_free(p->p_stats);
282 thread_free(FIRST_THREAD_IN_PROC(p));
283 mtx_destroy(&p->p_mtx);
284 if (p->p_ksi != NULL)
285 ksiginfo_free(p->p_ksi);
287 panic("proc reclaimed");
292 * PID space management.
294 * These bitmaps are used by fork_findpid.
296 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
297 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
298 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
299 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
301 static bitstr_t *proc_id_array[] = {
309 proc_id_set(int type, pid_t id)
312 KASSERT(type >= 0 && type < nitems(proc_id_array),
313 ("invalid type %d\n", type));
314 mtx_lock(&procid_lock);
315 KASSERT(bit_test(proc_id_array[type], id) == 0,
316 ("bit %d already set in %d\n", id, type));
317 bit_set(proc_id_array[type], id);
318 mtx_unlock(&procid_lock);
322 proc_id_set_cond(int type, pid_t id)
325 KASSERT(type >= 0 && type < nitems(proc_id_array),
326 ("invalid type %d\n", type));
327 if (bit_test(proc_id_array[type], id))
329 mtx_lock(&procid_lock);
330 bit_set(proc_id_array[type], id);
331 mtx_unlock(&procid_lock);
335 proc_id_clear(int type, pid_t id)
338 KASSERT(type >= 0 && type < nitems(proc_id_array),
339 ("invalid type %d\n", type));
340 mtx_lock(&procid_lock);
341 KASSERT(bit_test(proc_id_array[type], id) != 0,
342 ("bit %d not set in %d\n", id, type));
343 bit_clear(proc_id_array[type], id);
344 mtx_unlock(&procid_lock);
348 * Is p an inferior of the current process?
351 inferior(struct proc *p)
354 sx_assert(&proctree_lock, SX_LOCKED);
355 PROC_LOCK_ASSERT(p, MA_OWNED);
356 for (; p != curproc; p = proc_realparent(p)) {
364 * Shared lock all the pid hash lists.
367 pidhash_slockall(void)
371 for (i = 0; i < pidhashlock + 1; i++)
372 sx_slock(&pidhashtbl_lock[i]);
376 * Shared unlock all the pid hash lists.
379 pidhash_sunlockall(void)
383 for (i = 0; i < pidhashlock + 1; i++)
384 sx_sunlock(&pidhashtbl_lock[i]);
388 * Similar to pfind_any(), this function finds zombies.
391 pfind_any_locked(pid_t pid)
395 sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
396 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
397 if (p->p_pid == pid) {
399 if (p->p_state == PRS_NEW) {
410 * Locate a process by number.
412 * By not returning processes in the PRS_NEW state, we allow callers to avoid
413 * testing for that condition to avoid dereferencing p_ucred, et al.
415 static __always_inline struct proc *
416 _pfind(pid_t pid, bool zombie)
421 if (p->p_pid == pid) {
425 sx_slock(PIDHASHLOCK(pid));
426 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
427 if (p->p_pid == pid) {
429 if (p->p_state == PRS_NEW ||
430 (!zombie && p->p_state == PRS_ZOMBIE)) {
437 sx_sunlock(PIDHASHLOCK(pid));
445 return (_pfind(pid, false));
449 * Same as pfind but allow zombies.
455 return (_pfind(pid, true));
464 sx_slock(&allproc_lock);
465 FOREACH_PROC_IN_SYSTEM(p) {
467 if (p->p_state == PRS_NEW) {
471 FOREACH_THREAD_IN_PROC(p, td) {
472 if (td->td_tid == tid)
478 sx_sunlock(&allproc_lock);
483 * Locate a process group by number.
484 * The caller must hold proctree_lock.
491 sx_assert(&proctree_lock, SX_LOCKED);
493 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
494 if (pgrp->pg_id == pgid) {
503 * Locate process and do additional manipulations, depending on flags.
506 pget(pid_t pid, int flags, struct proc **pp)
512 if (p->p_pid == pid) {
516 if (pid <= PID_MAX) {
517 if ((flags & PGET_NOTWEXIT) == 0)
521 } else if ((flags & PGET_NOTID) == 0) {
526 if ((flags & PGET_CANSEE) != 0) {
527 error = p_cansee(curthread, p);
532 if ((flags & PGET_CANDEBUG) != 0) {
533 error = p_candebug(curthread, p);
537 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
541 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
545 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
547 * XXXRW: Not clear ESRCH is the right error during proc
553 if ((flags & PGET_HOLD) != 0) {
565 * Create a new process group.
566 * pgid must be equal to the pid of p.
567 * Begin a new session if required.
570 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
573 sx_assert(&proctree_lock, SX_XLOCKED);
575 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
576 KASSERT(p->p_pid == pgid,
577 ("enterpgrp: new pgrp and pid != pgid"));
578 KASSERT(pgfind(pgid) == NULL,
579 ("enterpgrp: pgrp with pgid exists"));
580 KASSERT(!SESS_LEADER(p),
581 ("enterpgrp: session leader attempted setpgrp"));
583 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
589 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
591 p->p_flag &= ~P_CONTROLT;
595 sess->s_sid = p->p_pid;
596 proc_id_set(PROC_ID_SESSION, p->p_pid);
597 refcount_init(&sess->s_count, 1);
598 sess->s_ttyvp = NULL;
599 sess->s_ttydp = NULL;
601 bcopy(p->p_session->s_login, sess->s_login,
602 sizeof(sess->s_login));
603 pgrp->pg_session = sess;
604 KASSERT(p == curproc,
605 ("enterpgrp: mksession and p != curproc"));
607 pgrp->pg_session = p->p_session;
608 sess_hold(pgrp->pg_session);
612 proc_id_set(PROC_ID_GROUP, p->p_pid);
613 LIST_INIT(&pgrp->pg_members);
616 * As we have an exclusive lock of proctree_lock,
617 * this should not deadlock.
619 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
621 SLIST_INIT(&pgrp->pg_sigiolst);
624 doenterpgrp(p, pgrp);
630 * Move p to an existing process group
633 enterthispgrp(struct proc *p, struct pgrp *pgrp)
636 sx_assert(&proctree_lock, SX_XLOCKED);
637 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
638 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
639 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
640 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
641 KASSERT(pgrp->pg_session == p->p_session,
642 ("%s: pgrp's session %p, p->p_session %p.\n",
646 KASSERT(pgrp != p->p_pgrp,
647 ("%s: p belongs to pgrp.", __func__));
649 doenterpgrp(p, pgrp);
655 * Move p to a process group
658 doenterpgrp(struct proc *p, struct pgrp *pgrp)
660 struct pgrp *savepgrp;
662 sx_assert(&proctree_lock, SX_XLOCKED);
663 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
664 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
665 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
666 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
668 savepgrp = p->p_pgrp;
671 * Adjust eligibility of affected pgrps to participate in job control.
672 * Increment eligibility counts before decrementing, otherwise we
673 * could reach 0 spuriously during the first call.
676 fixjobc(p, p->p_pgrp, 0);
681 LIST_REMOVE(p, p_pglist);
684 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
685 PGRP_UNLOCK(savepgrp);
687 if (LIST_EMPTY(&savepgrp->pg_members))
692 * remove process from process group
695 leavepgrp(struct proc *p)
697 struct pgrp *savepgrp;
699 sx_assert(&proctree_lock, SX_XLOCKED);
700 savepgrp = p->p_pgrp;
703 LIST_REMOVE(p, p_pglist);
706 PGRP_UNLOCK(savepgrp);
707 if (LIST_EMPTY(&savepgrp->pg_members))
713 * delete a process group
716 pgdelete(struct pgrp *pgrp)
718 struct session *savesess;
721 sx_assert(&proctree_lock, SX_XLOCKED);
722 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
723 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
726 * Reset any sigio structures pointing to us as a result of
727 * F_SETOWN with our pgid.
729 funsetownlst(&pgrp->pg_sigiolst);
732 tp = pgrp->pg_session->s_ttyp;
733 LIST_REMOVE(pgrp, pg_hash);
734 savesess = pgrp->pg_session;
737 /* Remove the reference to the pgrp before deallocating it. */
740 tty_rel_pgrp(tp, pgrp);
743 proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
744 mtx_destroy(&pgrp->pg_mtx);
746 sess_release(savesess);
750 pgadjustjobc(struct pgrp *pgrp, int entering)
758 if (pgrp->pg_jobc == 0)
765 * Adjust pgrp jobc counters when specified process changes process group.
766 * We count the number of processes in each process group that "qualify"
767 * the group for terminal job control (those with a parent in a different
768 * process group of the same session). If that count reaches zero, the
769 * process group becomes orphaned. Check both the specified process'
770 * process group and that of its children.
771 * entering == 0 => p is leaving specified group.
772 * entering == 1 => p is entering specified group.
775 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
777 struct pgrp *hispgrp;
778 struct session *mysession;
781 sx_assert(&proctree_lock, SX_LOCKED);
782 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
783 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
784 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
787 * Check p's parent to see whether p qualifies its own process
788 * group; if so, adjust count for p's process group.
790 mysession = pgrp->pg_session;
791 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
792 hispgrp->pg_session == mysession)
793 pgadjustjobc(pgrp, entering);
796 * Check this process' children to see whether they qualify
797 * their process groups; if so, adjust counts for children's
800 LIST_FOREACH(q, &p->p_children, p_sibling) {
802 if (hispgrp == pgrp ||
803 hispgrp->pg_session != mysession)
805 if (q->p_state == PRS_ZOMBIE)
807 pgadjustjobc(hispgrp, entering);
820 MPASS(p->p_flag & P_WEXIT);
822 * Do a quick check to see if there is anything to do with the
823 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
826 if (!SESS_LEADER(p) &&
827 (p->p_pgrp == p->p_pptr->p_pgrp) &&
828 LIST_EMPTY(&p->p_children)) {
834 sx_xlock(&proctree_lock);
835 if (SESS_LEADER(p)) {
839 * s_ttyp is not zero'd; we use this to indicate that
840 * the session once had a controlling terminal. (for
841 * logging and informational purposes)
852 * Signal foreground pgrp and revoke access to
853 * controlling terminal if it has not been revoked
856 * Because the TTY may have been revoked in the mean
857 * time and could already have a new session associated
858 * with it, make sure we don't send a SIGHUP to a
859 * foreground process group that does not belong to this
865 if (tp->t_session == sp)
866 tty_signal_pgrp(tp, SIGHUP);
871 sx_xunlock(&proctree_lock);
872 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
873 VOP_REVOKE(ttyvp, REVOKEALL);
877 sx_xlock(&proctree_lock);
880 fixjobc(p, p->p_pgrp, 0);
881 sx_xunlock(&proctree_lock);
885 * A process group has become orphaned;
886 * if there are any stopped processes in the group,
887 * hang-up all process in that group.
890 orphanpg(struct pgrp *pg)
894 PGRP_LOCK_ASSERT(pg, MA_OWNED);
896 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
898 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
900 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
902 kern_psignal(p, SIGHUP);
903 kern_psignal(p, SIGCONT);
913 sess_hold(struct session *s)
916 refcount_acquire(&s->s_count);
920 sess_release(struct session *s)
923 if (refcount_release(&s->s_count)) {
924 if (s->s_ttyp != NULL) {
926 tty_rel_sess(s->s_ttyp, s);
928 proc_id_clear(PROC_ID_SESSION, s->s_sid);
929 mtx_destroy(&s->s_mtx);
936 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
942 for (i = 0; i <= pgrphash; i++) {
943 if (!LIST_EMPTY(&pgrphashtbl[i])) {
944 printf("\tindx %d\n", i);
945 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
947 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
948 (void *)pgrp, (long)pgrp->pg_id,
949 (void *)pgrp->pg_session,
950 pgrp->pg_session->s_count,
951 (void *)LIST_FIRST(&pgrp->pg_members));
952 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
953 printf("\t\tpid %ld addr %p pgrp %p\n",
954 (long)p->p_pid, (void *)p,
964 * Calculate the kinfo_proc members which contain process-wide
966 * Must be called with the target process locked.
969 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
973 PROC_LOCK_ASSERT(p, MA_OWNED);
977 FOREACH_THREAD_IN_PROC(p, td) {
979 kp->ki_pctcpu += sched_pctcpu(td);
980 kp->ki_estcpu += sched_estcpu(td);
986 * Clear kinfo_proc and fill in any information that is common
987 * to all threads in the process.
988 * Must be called with the target process locked.
991 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
998 struct timeval boottime;
1000 PROC_LOCK_ASSERT(p, MA_OWNED);
1001 bzero(kp, sizeof(*kp));
1003 kp->ki_structsize = sizeof(*kp);
1005 kp->ki_addr =/* p->p_addr; */0; /* XXX */
1006 kp->ki_args = p->p_args;
1007 kp->ki_textvp = p->p_textvp;
1009 kp->ki_tracep = p->p_tracevp;
1010 kp->ki_traceflag = p->p_traceflag;
1012 kp->ki_fd = p->p_fd;
1013 kp->ki_vmspace = p->p_vmspace;
1014 kp->ki_flag = p->p_flag;
1015 kp->ki_flag2 = p->p_flag2;
1018 kp->ki_uid = cred->cr_uid;
1019 kp->ki_ruid = cred->cr_ruid;
1020 kp->ki_svuid = cred->cr_svuid;
1021 kp->ki_cr_flags = 0;
1022 if (cred->cr_flags & CRED_FLAG_CAPMODE)
1023 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1024 /* XXX bde doesn't like KI_NGROUPS */
1025 if (cred->cr_ngroups > KI_NGROUPS) {
1026 kp->ki_ngroups = KI_NGROUPS;
1027 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1029 kp->ki_ngroups = cred->cr_ngroups;
1030 bcopy(cred->cr_groups, kp->ki_groups,
1031 kp->ki_ngroups * sizeof(gid_t));
1032 kp->ki_rgid = cred->cr_rgid;
1033 kp->ki_svgid = cred->cr_svgid;
1034 /* If jailed(cred), emulate the old P_JAILED flag. */
1036 kp->ki_flag |= P_JAILED;
1037 /* If inside the jail, use 0 as a jail ID. */
1038 if (cred->cr_prison != curthread->td_ucred->cr_prison)
1039 kp->ki_jid = cred->cr_prison->pr_id;
1041 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1042 sizeof(kp->ki_loginclass));
1046 mtx_lock(&ps->ps_mtx);
1047 kp->ki_sigignore = ps->ps_sigignore;
1048 kp->ki_sigcatch = ps->ps_sigcatch;
1049 mtx_unlock(&ps->ps_mtx);
1051 if (p->p_state != PRS_NEW &&
1052 p->p_state != PRS_ZOMBIE &&
1053 p->p_vmspace != NULL) {
1054 struct vmspace *vm = p->p_vmspace;
1056 kp->ki_size = vm->vm_map.size;
1057 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1058 FOREACH_THREAD_IN_PROC(p, td0) {
1059 if (!TD_IS_SWAPPED(td0))
1060 kp->ki_rssize += td0->td_kstack_pages;
1062 kp->ki_swrss = vm->vm_swrss;
1063 kp->ki_tsize = vm->vm_tsize;
1064 kp->ki_dsize = vm->vm_dsize;
1065 kp->ki_ssize = vm->vm_ssize;
1066 } else if (p->p_state == PRS_ZOMBIE)
1067 kp->ki_stat = SZOMB;
1068 if (kp->ki_flag & P_INMEM)
1069 kp->ki_sflag = PS_INMEM;
1072 /* Calculate legacy swtime as seconds since 'swtick'. */
1073 kp->ki_swtime = (ticks - p->p_swtick) / hz;
1074 kp->ki_pid = p->p_pid;
1075 kp->ki_nice = p->p_nice;
1076 kp->ki_fibnum = p->p_fibnum;
1077 kp->ki_start = p->p_stats->p_start;
1078 getboottime(&boottime);
1079 timevaladd(&kp->ki_start, &boottime);
1081 rufetch(p, &kp->ki_rusage);
1082 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1083 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1085 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1086 /* Some callers want child times in a single value. */
1087 kp->ki_childtime = kp->ki_childstime;
1088 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1090 FOREACH_THREAD_IN_PROC(p, td0)
1091 kp->ki_cow += td0->td_cow;
1095 kp->ki_pgid = p->p_pgrp->pg_id;
1096 kp->ki_jobc = p->p_pgrp->pg_jobc;
1097 sp = p->p_pgrp->pg_session;
1100 kp->ki_sid = sp->s_sid;
1102 strlcpy(kp->ki_login, sp->s_login,
1103 sizeof(kp->ki_login));
1105 kp->ki_kiflag |= KI_CTTY;
1107 kp->ki_kiflag |= KI_SLEADER;
1108 /* XXX proctree_lock */
1113 if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1114 kp->ki_tdev = tty_udev(tp);
1115 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1116 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1118 kp->ki_tsid = tp->t_session->s_sid;
1120 kp->ki_tdev = NODEV;
1121 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1123 if (p->p_comm[0] != '\0')
1124 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1125 if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1126 p->p_sysent->sv_name[0] != '\0')
1127 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1128 kp->ki_siglist = p->p_siglist;
1129 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1130 kp->ki_acflag = p->p_acflag;
1131 kp->ki_lock = p->p_lock;
1133 kp->ki_ppid = p->p_oppid;
1134 if (p->p_flag & P_TRACED)
1135 kp->ki_tracer = p->p_pptr->p_pid;
1140 * Fill in information that is thread specific. Must be called with
1141 * target process locked. If 'preferthread' is set, overwrite certain
1142 * process-related fields that are maintained for both threads and
1146 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1152 PROC_LOCK_ASSERT(p, MA_OWNED);
1157 if (td->td_wmesg != NULL)
1158 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1160 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1161 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1162 sizeof(kp->ki_tdname)) {
1163 strlcpy(kp->ki_moretdname,
1164 td->td_name + sizeof(kp->ki_tdname) - 1,
1165 sizeof(kp->ki_moretdname));
1167 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1169 if (TD_ON_LOCK(td)) {
1170 kp->ki_kiflag |= KI_LOCKBLOCK;
1171 strlcpy(kp->ki_lockname, td->td_lockname,
1172 sizeof(kp->ki_lockname));
1174 kp->ki_kiflag &= ~KI_LOCKBLOCK;
1175 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1178 if (p->p_state == PRS_NORMAL) { /* approximate. */
1179 if (TD_ON_RUNQ(td) ||
1181 TD_IS_RUNNING(td)) {
1183 } else if (P_SHOULDSTOP(p)) {
1184 kp->ki_stat = SSTOP;
1185 } else if (TD_IS_SLEEPING(td)) {
1186 kp->ki_stat = SSLEEP;
1187 } else if (TD_ON_LOCK(td)) {
1188 kp->ki_stat = SLOCK;
1190 kp->ki_stat = SWAIT;
1192 } else if (p->p_state == PRS_ZOMBIE) {
1193 kp->ki_stat = SZOMB;
1198 /* Things in the thread */
1199 kp->ki_wchan = td->td_wchan;
1200 kp->ki_pri.pri_level = td->td_priority;
1201 kp->ki_pri.pri_native = td->td_base_pri;
1204 * Note: legacy fields; clamp at the old NOCPU value and/or
1205 * the maximum u_char CPU value.
1207 if (td->td_lastcpu == NOCPU)
1208 kp->ki_lastcpu_old = NOCPU_OLD;
1209 else if (td->td_lastcpu > MAXCPU_OLD)
1210 kp->ki_lastcpu_old = MAXCPU_OLD;
1212 kp->ki_lastcpu_old = td->td_lastcpu;
1214 if (td->td_oncpu == NOCPU)
1215 kp->ki_oncpu_old = NOCPU_OLD;
1216 else if (td->td_oncpu > MAXCPU_OLD)
1217 kp->ki_oncpu_old = MAXCPU_OLD;
1219 kp->ki_oncpu_old = td->td_oncpu;
1221 kp->ki_lastcpu = td->td_lastcpu;
1222 kp->ki_oncpu = td->td_oncpu;
1223 kp->ki_tdflags = td->td_flags;
1224 kp->ki_tid = td->td_tid;
1225 kp->ki_numthreads = p->p_numthreads;
1226 kp->ki_pcb = td->td_pcb;
1227 kp->ki_kstack = (void *)td->td_kstack;
1228 kp->ki_slptime = (ticks - td->td_slptick) / hz;
1229 kp->ki_pri.pri_class = td->td_pri_class;
1230 kp->ki_pri.pri_user = td->td_user_pri;
1233 rufetchtd(td, &kp->ki_rusage);
1234 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1235 kp->ki_pctcpu = sched_pctcpu(td);
1236 kp->ki_estcpu = sched_estcpu(td);
1237 kp->ki_cow = td->td_cow;
1240 /* We can't get this anymore but ps etc never used it anyway. */
1244 kp->ki_siglist = td->td_siglist;
1245 kp->ki_sigmask = td->td_sigmask;
1252 * Fill in a kinfo_proc structure for the specified process.
1253 * Must be called with the target process locked.
1256 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1259 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1261 fill_kinfo_proc_only(p, kp);
1262 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1263 fill_kinfo_aggregate(p, kp);
1270 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1274 * Copy parts of p_stats; zero the rest of p_stats (statistics).
1277 pstats_fork(struct pstats *src, struct pstats *dst)
1280 bzero(&dst->pstat_startzero,
1281 __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1282 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1283 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1287 pstats_free(struct pstats *ps)
1290 free(ps, M_SUBPROC);
1293 #ifdef COMPAT_FREEBSD32
1296 * This function is typically used to copy out the kernel address, so
1297 * it can be replaced by assignment of zero.
1299 static inline uint32_t
1300 ptr32_trim(const void *ptr)
1304 uptr = (uintptr_t)ptr;
1305 return ((uptr > UINT_MAX) ? 0 : uptr);
1308 #define PTRTRIM_CP(src,dst,fld) \
1309 do { (dst).fld = ptr32_trim((src).fld); } while (0)
1312 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1316 bzero(ki32, sizeof(struct kinfo_proc32));
1317 ki32->ki_structsize = sizeof(struct kinfo_proc32);
1318 CP(*ki, *ki32, ki_layout);
1319 PTRTRIM_CP(*ki, *ki32, ki_args);
1320 PTRTRIM_CP(*ki, *ki32, ki_paddr);
1321 PTRTRIM_CP(*ki, *ki32, ki_addr);
1322 PTRTRIM_CP(*ki, *ki32, ki_tracep);
1323 PTRTRIM_CP(*ki, *ki32, ki_textvp);
1324 PTRTRIM_CP(*ki, *ki32, ki_fd);
1325 PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1326 PTRTRIM_CP(*ki, *ki32, ki_wchan);
1327 CP(*ki, *ki32, ki_pid);
1328 CP(*ki, *ki32, ki_ppid);
1329 CP(*ki, *ki32, ki_pgid);
1330 CP(*ki, *ki32, ki_tpgid);
1331 CP(*ki, *ki32, ki_sid);
1332 CP(*ki, *ki32, ki_tsid);
1333 CP(*ki, *ki32, ki_jobc);
1334 CP(*ki, *ki32, ki_tdev);
1335 CP(*ki, *ki32, ki_tdev_freebsd11);
1336 CP(*ki, *ki32, ki_siglist);
1337 CP(*ki, *ki32, ki_sigmask);
1338 CP(*ki, *ki32, ki_sigignore);
1339 CP(*ki, *ki32, ki_sigcatch);
1340 CP(*ki, *ki32, ki_uid);
1341 CP(*ki, *ki32, ki_ruid);
1342 CP(*ki, *ki32, ki_svuid);
1343 CP(*ki, *ki32, ki_rgid);
1344 CP(*ki, *ki32, ki_svgid);
1345 CP(*ki, *ki32, ki_ngroups);
1346 for (i = 0; i < KI_NGROUPS; i++)
1347 CP(*ki, *ki32, ki_groups[i]);
1348 CP(*ki, *ki32, ki_size);
1349 CP(*ki, *ki32, ki_rssize);
1350 CP(*ki, *ki32, ki_swrss);
1351 CP(*ki, *ki32, ki_tsize);
1352 CP(*ki, *ki32, ki_dsize);
1353 CP(*ki, *ki32, ki_ssize);
1354 CP(*ki, *ki32, ki_xstat);
1355 CP(*ki, *ki32, ki_acflag);
1356 CP(*ki, *ki32, ki_pctcpu);
1357 CP(*ki, *ki32, ki_estcpu);
1358 CP(*ki, *ki32, ki_slptime);
1359 CP(*ki, *ki32, ki_swtime);
1360 CP(*ki, *ki32, ki_cow);
1361 CP(*ki, *ki32, ki_runtime);
1362 TV_CP(*ki, *ki32, ki_start);
1363 TV_CP(*ki, *ki32, ki_childtime);
1364 CP(*ki, *ki32, ki_flag);
1365 CP(*ki, *ki32, ki_kiflag);
1366 CP(*ki, *ki32, ki_traceflag);
1367 CP(*ki, *ki32, ki_stat);
1368 CP(*ki, *ki32, ki_nice);
1369 CP(*ki, *ki32, ki_lock);
1370 CP(*ki, *ki32, ki_rqindex);
1371 CP(*ki, *ki32, ki_oncpu);
1372 CP(*ki, *ki32, ki_lastcpu);
1374 /* XXX TODO: wrap cpu value as appropriate */
1375 CP(*ki, *ki32, ki_oncpu_old);
1376 CP(*ki, *ki32, ki_lastcpu_old);
1378 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1379 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1380 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1381 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1382 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1383 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1384 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1385 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1386 CP(*ki, *ki32, ki_tracer);
1387 CP(*ki, *ki32, ki_flag2);
1388 CP(*ki, *ki32, ki_fibnum);
1389 CP(*ki, *ki32, ki_cr_flags);
1390 CP(*ki, *ki32, ki_jid);
1391 CP(*ki, *ki32, ki_numthreads);
1392 CP(*ki, *ki32, ki_tid);
1393 CP(*ki, *ki32, ki_pri);
1394 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1395 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1396 PTRTRIM_CP(*ki, *ki32, ki_pcb);
1397 PTRTRIM_CP(*ki, *ki32, ki_kstack);
1398 PTRTRIM_CP(*ki, *ki32, ki_udata);
1399 PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1400 CP(*ki, *ki32, ki_sflag);
1401 CP(*ki, *ki32, ki_tdflags);
1406 kern_proc_out_size(struct proc *p, int flags)
1410 PROC_LOCK_ASSERT(p, MA_OWNED);
1412 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1413 #ifdef COMPAT_FREEBSD32
1414 if ((flags & KERN_PROC_MASK32) != 0) {
1415 size += sizeof(struct kinfo_proc32);
1418 size += sizeof(struct kinfo_proc);
1420 #ifdef COMPAT_FREEBSD32
1421 if ((flags & KERN_PROC_MASK32) != 0)
1422 size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1425 size += sizeof(struct kinfo_proc) * p->p_numthreads;
1432 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1435 struct kinfo_proc ki;
1436 #ifdef COMPAT_FREEBSD32
1437 struct kinfo_proc32 ki32;
1441 PROC_LOCK_ASSERT(p, MA_OWNED);
1442 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1445 fill_kinfo_proc(p, &ki);
1446 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1447 #ifdef COMPAT_FREEBSD32
1448 if ((flags & KERN_PROC_MASK32) != 0) {
1449 freebsd32_kinfo_proc_out(&ki, &ki32);
1450 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1454 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1457 FOREACH_THREAD_IN_PROC(p, td) {
1458 fill_kinfo_thread(td, &ki, 1);
1459 #ifdef COMPAT_FREEBSD32
1460 if ((flags & KERN_PROC_MASK32) != 0) {
1461 freebsd32_kinfo_proc_out(&ki, &ki32);
1462 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1466 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1477 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1480 struct kinfo_proc ki;
1483 if (req->oldptr == NULL)
1484 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1486 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1487 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1488 error = kern_proc_out(p, &sb, flags);
1489 error2 = sbuf_finish(&sb);
1493 else if (error2 != 0)
1499 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1504 for (i = 0; i < pidhashlock + 1; i++) {
1505 sx_slock(&pidhashtbl_lock[i]);
1506 for (j = i; j <= pidhash; j += pidhashlock + 1) {
1507 LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1508 if (p->p_state == PRS_NEW)
1510 error = cb(p, cbarg);
1511 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1513 sx_sunlock(&pidhashtbl_lock[i]);
1518 sx_sunlock(&pidhashtbl_lock[i]);
1523 struct kern_proc_out_args {
1524 struct sysctl_req *req;
1531 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1533 struct kern_proc_out_args *arg = origarg;
1534 int *name = arg->name;
1535 int oid_number = arg->oid_number;
1536 int flags = arg->flags;
1537 struct sysctl_req *req = arg->req;
1542 KASSERT(p->p_ucred != NULL,
1543 ("process credential is NULL for non-NEW proc"));
1545 * Show a user only appropriate processes.
1547 if (p_cansee(curthread, p))
1550 * TODO - make more efficient (see notes below).
1553 switch (oid_number) {
1556 if (p->p_ucred->cr_gid != (gid_t)name[0])
1560 case KERN_PROC_PGRP:
1561 /* could do this by traversing pgrp */
1562 if (p->p_pgrp == NULL ||
1563 p->p_pgrp->pg_id != (pid_t)name[0])
1567 case KERN_PROC_RGID:
1568 if (p->p_ucred->cr_rgid != (gid_t)name[0])
1572 case KERN_PROC_SESSION:
1573 if (p->p_session == NULL ||
1574 p->p_session->s_sid != (pid_t)name[0])
1579 if ((p->p_flag & P_CONTROLT) == 0 ||
1580 p->p_session == NULL)
1582 /* XXX proctree_lock */
1583 SESS_LOCK(p->p_session);
1584 if (p->p_session->s_ttyp == NULL ||
1585 tty_udev(p->p_session->s_ttyp) !=
1587 SESS_UNLOCK(p->p_session);
1590 SESS_UNLOCK(p->p_session);
1594 if (p->p_ucred->cr_uid != (uid_t)name[0])
1598 case KERN_PROC_RUID:
1599 if (p->p_ucred->cr_ruid != (uid_t)name[0])
1603 case KERN_PROC_PROC:
1610 error = sysctl_out_proc(p, req, flags);
1611 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1619 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1621 struct kern_proc_out_args iterarg;
1622 int *name = (int *)arg1;
1623 u_int namelen = arg2;
1625 int flags, oid_number;
1628 oid_number = oidp->oid_number;
1629 if (oid_number != KERN_PROC_ALL &&
1630 (oid_number & KERN_PROC_INC_THREAD) == 0)
1631 flags = KERN_PROC_NOTHREADS;
1634 oid_number &= ~KERN_PROC_INC_THREAD;
1636 #ifdef COMPAT_FREEBSD32
1637 if (req->flags & SCTL_MASK32)
1638 flags |= KERN_PROC_MASK32;
1640 if (oid_number == KERN_PROC_PID) {
1643 error = sysctl_wire_old_buffer(req, 0);
1646 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1648 error = sysctl_out_proc(p, req, flags);
1652 switch (oid_number) {
1657 case KERN_PROC_PROC:
1658 if (namelen != 0 && namelen != 1)
1667 if (req->oldptr == NULL) {
1668 /* overestimate by 5 procs */
1669 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1673 error = sysctl_wire_old_buffer(req, 0);
1677 iterarg.flags = flags;
1678 iterarg.oid_number = oid_number;
1680 iterarg.name = name;
1681 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1686 pargs_alloc(int len)
1690 pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1692 refcount_init(&pa->ar_ref, 1);
1693 pa->ar_length = len;
1698 pargs_free(struct pargs *pa)
1705 pargs_hold(struct pargs *pa)
1710 refcount_acquire(&pa->ar_ref);
1714 pargs_drop(struct pargs *pa)
1719 if (refcount_release(&pa->ar_ref))
1724 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1730 * This may return a short read if the string is shorter than the chunk
1731 * and is aligned at the end of the page, and the following page is not
1734 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1740 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */
1742 enum proc_vector_type {
1748 #ifdef COMPAT_FREEBSD32
1750 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1751 size_t *vsizep, enum proc_vector_type type)
1753 struct freebsd32_ps_strings pss;
1755 vm_offset_t vptr, ptr;
1756 uint32_t *proc_vector32;
1762 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1763 sizeof(pss)) != sizeof(pss))
1767 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1768 vsize = pss.ps_nargvstr;
1769 if (vsize > ARG_MAX)
1771 size = vsize * sizeof(int32_t);
1774 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1775 vsize = pss.ps_nenvstr;
1776 if (vsize > ARG_MAX)
1778 size = vsize * sizeof(int32_t);
1781 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1782 (pss.ps_nenvstr + 1) * sizeof(int32_t);
1785 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1786 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1789 if (aux.a_type == AT_NULL)
1793 if (aux.a_type != AT_NULL)
1796 size = vsize * sizeof(aux);
1799 KASSERT(0, ("Wrong proc vector type: %d", type));
1802 proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1803 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1807 if (type == PROC_AUX) {
1808 *proc_vectorp = (char **)proc_vector32;
1812 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1813 for (i = 0; i < (int)vsize; i++)
1814 proc_vector[i] = PTRIN(proc_vector32[i]);
1815 *proc_vectorp = proc_vector;
1818 free(proc_vector32, M_TEMP);
1824 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1825 size_t *vsizep, enum proc_vector_type type)
1827 struct ps_strings pss;
1829 vm_offset_t vptr, ptr;
1834 #ifdef COMPAT_FREEBSD32
1835 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1836 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1838 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1839 sizeof(pss)) != sizeof(pss))
1843 vptr = (vm_offset_t)pss.ps_argvstr;
1844 vsize = pss.ps_nargvstr;
1845 if (vsize > ARG_MAX)
1847 size = vsize * sizeof(char *);
1850 vptr = (vm_offset_t)pss.ps_envstr;
1851 vsize = pss.ps_nenvstr;
1852 if (vsize > ARG_MAX)
1854 size = vsize * sizeof(char *);
1858 * The aux array is just above env array on the stack. Check
1859 * that the address is naturally aligned.
1861 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1863 #if __ELF_WORD_SIZE == 64
1864 if (vptr % sizeof(uint64_t) != 0)
1866 if (vptr % sizeof(uint32_t) != 0)
1870 * We count the array size reading the aux vectors from the
1871 * stack until AT_NULL vector is returned. So (to keep the code
1872 * simple) we read the process stack twice: the first time here
1873 * to find the size and the second time when copying the vectors
1874 * to the allocated proc_vector.
1876 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1877 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1880 if (aux.a_type == AT_NULL)
1885 * If the PROC_AUXV_MAX entries are iterated over, and we have
1886 * not reached AT_NULL, it is most likely we are reading wrong
1887 * data: either the process doesn't have auxv array or data has
1888 * been modified. Return the error in this case.
1890 if (aux.a_type != AT_NULL)
1893 size = vsize * sizeof(aux);
1896 KASSERT(0, ("Wrong proc vector type: %d", type));
1897 return (EINVAL); /* In case we are built without INVARIANTS. */
1899 proc_vector = malloc(size, M_TEMP, M_WAITOK);
1900 if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1901 free(proc_vector, M_TEMP);
1904 *proc_vectorp = proc_vector;
1910 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */
1913 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1914 enum proc_vector_type type)
1916 size_t done, len, nchr, vsize;
1918 char **proc_vector, *sptr;
1919 char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1921 PROC_ASSERT_HELD(p);
1924 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1926 nchr = 2 * (PATH_MAX + ARG_MAX);
1928 error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1931 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1933 * The program may have scribbled into its argv array, e.g. to
1934 * remove some arguments. If that has happened, break out
1935 * before trying to read from NULL.
1937 if (proc_vector[i] == NULL)
1939 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1940 error = proc_read_string(td, p, sptr, pss_string,
1941 sizeof(pss_string));
1944 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1945 if (done + len >= nchr)
1946 len = nchr - done - 1;
1947 sbuf_bcat(sb, pss_string, len);
1948 if (len != GET_PS_STRINGS_CHUNK_SZ)
1950 done += GET_PS_STRINGS_CHUNK_SZ;
1952 sbuf_bcat(sb, "", 1);
1956 free(proc_vector, M_TEMP);
1961 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1964 return (get_ps_strings(curthread, p, sb, PROC_ARG));
1968 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1971 return (get_ps_strings(curthread, p, sb, PROC_ENV));
1975 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1981 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1983 #ifdef COMPAT_FREEBSD32
1984 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1985 size = vsize * sizeof(Elf32_Auxinfo);
1988 size = vsize * sizeof(Elf_Auxinfo);
1989 if (sbuf_bcat(sb, auxv, size) != 0)
1997 * This sysctl allows a process to retrieve the argument list or process
1998 * title for another process without groping around in the address space
1999 * of the other process. It also allow a process to set its own "process
2000 * title to a string of its own choice.
2003 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2005 int *name = (int *)arg1;
2006 u_int namelen = arg2;
2007 struct pargs *newpa, *pa;
2010 int flags, error = 0, error2;
2016 pid = (pid_t)name[0];
2018 * If the query is for this process and it is single-threaded, there
2019 * is nobody to modify pargs, thus we can just read.
2022 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2023 (pa = p->p_args) != NULL)
2024 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2026 flags = PGET_CANSEE;
2027 if (req->newptr != NULL)
2028 flags |= PGET_ISCURRENT;
2029 error = pget(pid, flags, &p);
2037 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2039 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2042 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2043 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2044 error = proc_getargv(curthread, p, &sb);
2045 error2 = sbuf_finish(&sb);
2048 if (error == 0 && error2 != 0)
2053 if (error != 0 || req->newptr == NULL)
2056 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2059 if (req->newlen == 0) {
2061 * Clear the argument pointer, so that we'll fetch arguments
2062 * with proc_getargv() until further notice.
2066 newpa = pargs_alloc(req->newlen);
2067 error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2082 * This sysctl allows a process to retrieve environment of another process.
2085 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2087 int *name = (int *)arg1;
2088 u_int namelen = arg2;
2096 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2099 if ((p->p_flag & P_SYSTEM) != 0) {
2104 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2105 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2106 error = proc_getenvv(curthread, p, &sb);
2107 error2 = sbuf_finish(&sb);
2110 return (error != 0 ? error : error2);
2114 * This sysctl allows a process to retrieve ELF auxiliary vector of
2118 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2120 int *name = (int *)arg1;
2121 u_int namelen = arg2;
2129 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2132 if ((p->p_flag & P_SYSTEM) != 0) {
2136 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2137 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2138 error = proc_getauxv(curthread, p, &sb);
2139 error2 = sbuf_finish(&sb);
2142 return (error != 0 ? error : error2);
2146 * This sysctl allows a process to retrieve the path of the executable for
2147 * itself or another process.
2150 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2152 pid_t *pidp = (pid_t *)arg1;
2153 unsigned int arglen = arg2;
2156 char *retbuf, *freebuf;
2161 if (*pidp == -1) { /* -1 means this process */
2162 p = req->td->td_proc;
2164 error = pget(*pidp, PGET_CANSEE, &p);
2178 error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2182 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2183 free(freebuf, M_TEMP);
2188 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2201 error = pget((pid_t)name[0], PGET_CANSEE, &p);
2204 sv_name = p->p_sysent->sv_name;
2206 return (sysctl_handle_string(oidp, sv_name, 0, req));
2209 #ifdef KINFO_OVMENTRY_SIZE
2210 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2213 #ifdef COMPAT_FREEBSD7
2215 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2217 vm_map_entry_t entry, tmp_entry;
2218 unsigned int last_timestamp;
2219 char *fullpath, *freepath;
2220 struct kinfo_ovmentry *kve;
2230 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2233 vm = vmspace_acquire_ref(p);
2238 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2241 vm_map_lock_read(map);
2242 VM_MAP_ENTRY_FOREACH(entry, map) {
2243 vm_object_t obj, tobj, lobj;
2246 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2249 bzero(kve, sizeof(*kve));
2250 kve->kve_structsize = sizeof(*kve);
2252 kve->kve_private_resident = 0;
2253 obj = entry->object.vm_object;
2255 VM_OBJECT_RLOCK(obj);
2256 if (obj->shadow_count == 1)
2257 kve->kve_private_resident =
2258 obj->resident_page_count;
2260 kve->kve_resident = 0;
2261 addr = entry->start;
2262 while (addr < entry->end) {
2263 if (pmap_extract(map->pmap, addr))
2264 kve->kve_resident++;
2268 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2270 VM_OBJECT_RLOCK(tobj);
2271 kve->kve_offset += tobj->backing_object_offset;
2274 VM_OBJECT_RUNLOCK(lobj);
2278 kve->kve_start = (void*)entry->start;
2279 kve->kve_end = (void*)entry->end;
2280 kve->kve_offset += (off_t)entry->offset;
2282 if (entry->protection & VM_PROT_READ)
2283 kve->kve_protection |= KVME_PROT_READ;
2284 if (entry->protection & VM_PROT_WRITE)
2285 kve->kve_protection |= KVME_PROT_WRITE;
2286 if (entry->protection & VM_PROT_EXECUTE)
2287 kve->kve_protection |= KVME_PROT_EXEC;
2289 if (entry->eflags & MAP_ENTRY_COW)
2290 kve->kve_flags |= KVME_FLAG_COW;
2291 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2292 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2293 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2294 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2296 last_timestamp = map->timestamp;
2297 vm_map_unlock_read(map);
2299 kve->kve_fileid = 0;
2304 kve->kve_type = vm_object_kvme_type(lobj, &vp);
2305 if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2306 kve->kve_type = KVME_TYPE_UNKNOWN;
2310 VM_OBJECT_RUNLOCK(lobj);
2312 kve->kve_ref_count = obj->ref_count;
2313 kve->kve_shadow_count = obj->shadow_count;
2314 VM_OBJECT_RUNLOCK(obj);
2316 vn_fullpath(curthread, vp, &fullpath,
2318 cred = curthread->td_ucred;
2319 vn_lock(vp, LK_SHARED | LK_RETRY);
2320 if (VOP_GETATTR(vp, &va, cred) == 0) {
2321 kve->kve_fileid = va.va_fileid;
2323 kve->kve_fsid = va.va_fsid;
2328 kve->kve_type = KVME_TYPE_NONE;
2329 kve->kve_ref_count = 0;
2330 kve->kve_shadow_count = 0;
2333 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2334 if (freepath != NULL)
2335 free(freepath, M_TEMP);
2337 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2338 vm_map_lock_read(map);
2341 if (last_timestamp != map->timestamp) {
2342 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2346 vm_map_unlock_read(map);
2352 #endif /* COMPAT_FREEBSD7 */
2354 #ifdef KINFO_VMENTRY_SIZE
2355 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2359 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2360 int *resident_count, bool *super)
2362 vm_object_t obj, tobj;
2366 vm_pindex_t pi, pi_adv, pindex;
2369 *resident_count = 0;
2370 if (vmmap_skip_res_cnt)
2374 obj = entry->object.vm_object;
2375 addr = entry->start;
2377 pi = OFF_TO_IDX(entry->offset);
2378 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2379 if (m_adv != NULL) {
2382 pi_adv = atop(entry->end - addr);
2384 for (tobj = obj;; tobj = tobj->backing_object) {
2385 m = vm_page_find_least(tobj, pindex);
2387 if (m->pindex == pindex)
2389 if (pi_adv > m->pindex - pindex) {
2390 pi_adv = m->pindex - pindex;
2394 if (tobj->backing_object == NULL)
2396 pindex += OFF_TO_IDX(tobj->
2397 backing_object_offset);
2401 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2402 (addr & (pagesizes[1] - 1)) == 0 &&
2403 (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2405 pi_adv = atop(pagesizes[1]);
2408 * We do not test the found page on validity.
2409 * Either the page is busy and being paged in,
2410 * or it was invalidated. The first case
2411 * should be counted as resident, the second
2412 * is not so clear; we do account both.
2416 *resident_count += pi_adv;
2422 * Must be called with the process locked and will return unlocked.
2425 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2427 vm_map_entry_t entry, tmp_entry;
2430 vm_object_t obj, tobj, lobj;
2431 char *fullpath, *freepath;
2432 struct kinfo_vmentry *kve;
2437 unsigned int last_timestamp;
2441 PROC_LOCK_ASSERT(p, MA_OWNED);
2445 vm = vmspace_acquire_ref(p);
2450 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2454 vm_map_lock_read(map);
2455 VM_MAP_ENTRY_FOREACH(entry, map) {
2456 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2460 bzero(kve, sizeof(*kve));
2461 obj = entry->object.vm_object;
2463 for (tobj = obj; tobj != NULL;
2464 tobj = tobj->backing_object) {
2465 VM_OBJECT_RLOCK(tobj);
2466 kve->kve_offset += tobj->backing_object_offset;
2469 if (obj->backing_object == NULL)
2470 kve->kve_private_resident =
2471 obj->resident_page_count;
2472 kern_proc_vmmap_resident(map, entry,
2473 &kve->kve_resident, &super);
2475 kve->kve_flags |= KVME_FLAG_SUPER;
2476 for (tobj = obj; tobj != NULL;
2477 tobj = tobj->backing_object) {
2478 if (tobj != obj && tobj != lobj)
2479 VM_OBJECT_RUNLOCK(tobj);
2485 kve->kve_start = entry->start;
2486 kve->kve_end = entry->end;
2487 kve->kve_offset += entry->offset;
2489 if (entry->protection & VM_PROT_READ)
2490 kve->kve_protection |= KVME_PROT_READ;
2491 if (entry->protection & VM_PROT_WRITE)
2492 kve->kve_protection |= KVME_PROT_WRITE;
2493 if (entry->protection & VM_PROT_EXECUTE)
2494 kve->kve_protection |= KVME_PROT_EXEC;
2496 if (entry->eflags & MAP_ENTRY_COW)
2497 kve->kve_flags |= KVME_FLAG_COW;
2498 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2499 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2500 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2501 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2502 if (entry->eflags & MAP_ENTRY_GROWS_UP)
2503 kve->kve_flags |= KVME_FLAG_GROWS_UP;
2504 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2505 kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2506 if (entry->eflags & MAP_ENTRY_USER_WIRED)
2507 kve->kve_flags |= KVME_FLAG_USER_WIRED;
2509 last_timestamp = map->timestamp;
2510 vm_map_unlock_read(map);
2515 kve->kve_type = vm_object_kvme_type(lobj, &vp);
2519 VM_OBJECT_RUNLOCK(lobj);
2521 kve->kve_ref_count = obj->ref_count;
2522 kve->kve_shadow_count = obj->shadow_count;
2523 VM_OBJECT_RUNLOCK(obj);
2525 vn_fullpath(curthread, vp, &fullpath,
2527 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2528 cred = curthread->td_ucred;
2529 vn_lock(vp, LK_SHARED | LK_RETRY);
2530 if (VOP_GETATTR(vp, &va, cred) == 0) {
2531 kve->kve_vn_fileid = va.va_fileid;
2532 kve->kve_vn_fsid = va.va_fsid;
2533 kve->kve_vn_fsid_freebsd11 =
2534 kve->kve_vn_fsid; /* truncate */
2536 MAKEIMODE(va.va_type, va.va_mode);
2537 kve->kve_vn_size = va.va_size;
2538 kve->kve_vn_rdev = va.va_rdev;
2539 kve->kve_vn_rdev_freebsd11 =
2540 kve->kve_vn_rdev; /* truncate */
2541 kve->kve_status = KF_ATTR_VALID;
2546 kve->kve_type = KVME_TYPE_NONE;
2547 kve->kve_ref_count = 0;
2548 kve->kve_shadow_count = 0;
2551 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2552 if (freepath != NULL)
2553 free(freepath, M_TEMP);
2555 /* Pack record size down */
2556 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2557 kve->kve_structsize =
2558 offsetof(struct kinfo_vmentry, kve_path) +
2559 strlen(kve->kve_path) + 1;
2561 kve->kve_structsize = sizeof(*kve);
2562 kve->kve_structsize = roundup(kve->kve_structsize,
2565 /* Halt filling and truncate rather than exceeding maxlen */
2566 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2568 vm_map_lock_read(map);
2570 } else if (maxlen != -1)
2571 maxlen -= kve->kve_structsize;
2573 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2575 vm_map_lock_read(map);
2578 if (last_timestamp != map->timestamp) {
2579 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2583 vm_map_unlock_read(map);
2591 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2595 int error, error2, *name;
2598 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2599 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2600 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2605 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2606 error2 = sbuf_finish(&sb);
2608 return (error != 0 ? error : error2);
2611 #if defined(STACK) || defined(DDB)
2613 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2615 struct kinfo_kstack *kkstp;
2616 int error, i, *name, numthreads;
2617 lwpid_t *lwpidarray;
2624 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2628 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2629 st = stack_create(M_WAITOK);
2634 if (lwpidarray != NULL) {
2635 free(lwpidarray, M_TEMP);
2638 numthreads = p->p_numthreads;
2640 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2643 } while (numthreads < p->p_numthreads);
2646 * XXXRW: During the below loop, execve(2) and countless other sorts
2647 * of changes could have taken place. Should we check to see if the
2648 * vmspace has been replaced, or the like, in order to prevent
2649 * giving a snapshot that spans, say, execve(2), with some threads
2650 * before and some after? Among other things, the credentials could
2651 * have changed, in which case the right to extract debug info might
2652 * no longer be assured.
2655 FOREACH_THREAD_IN_PROC(p, td) {
2656 KASSERT(i < numthreads,
2657 ("sysctl_kern_proc_kstack: numthreads"));
2658 lwpidarray[i] = td->td_tid;
2662 for (i = 0; i < numthreads; i++) {
2663 td = thread_find(p, lwpidarray[i]);
2667 bzero(kkstp, sizeof(*kkstp));
2668 (void)sbuf_new(&sb, kkstp->kkst_trace,
2669 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2671 kkstp->kkst_tid = td->td_tid;
2672 if (TD_IS_SWAPPED(td))
2673 kkstp->kkst_state = KKST_STATE_SWAPPED;
2674 else if (stack_save_td(st, td) == 0)
2675 kkstp->kkst_state = KKST_STATE_STACKOK;
2677 kkstp->kkst_state = KKST_STATE_RUNNING;
2680 stack_sbuf_print(&sb, st);
2683 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2690 if (lwpidarray != NULL)
2691 free(lwpidarray, M_TEMP);
2693 free(kkstp, M_TEMP);
2699 * This sysctl allows a process to retrieve the full list of groups from
2700 * itself or another process.
2703 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2705 pid_t *pidp = (pid_t *)arg1;
2706 unsigned int arglen = arg2;
2713 if (*pidp == -1) { /* -1 means this process */
2714 p = req->td->td_proc;
2717 error = pget(*pidp, PGET_CANSEE, &p);
2722 cred = crhold(p->p_ucred);
2725 error = SYSCTL_OUT(req, cred->cr_groups,
2726 cred->cr_ngroups * sizeof(gid_t));
2732 * This sysctl allows a process to retrieve or/and set the resource limit for
2736 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2738 int *name = (int *)arg1;
2739 u_int namelen = arg2;
2748 which = (u_int)name[1];
2749 if (which >= RLIM_NLIMITS)
2752 if (req->newptr != NULL && req->newlen != sizeof(rlim))
2755 flags = PGET_HOLD | PGET_NOTWEXIT;
2756 if (req->newptr != NULL)
2757 flags |= PGET_CANDEBUG;
2759 flags |= PGET_CANSEE;
2760 error = pget((pid_t)name[0], flags, &p);
2767 if (req->oldptr != NULL) {
2769 lim_rlimit_proc(p, which, &rlim);
2772 error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2779 if (req->newptr != NULL) {
2780 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2782 error = kern_proc_setrlimit(curthread, p, which, &rlim);
2791 * This sysctl allows a process to retrieve ps_strings structure location of
2795 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2797 int *name = (int *)arg1;
2798 u_int namelen = arg2;
2800 vm_offset_t ps_strings;
2802 #ifdef COMPAT_FREEBSD32
2803 uint32_t ps_strings32;
2809 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2812 #ifdef COMPAT_FREEBSD32
2813 if ((req->flags & SCTL_MASK32) != 0) {
2815 * We return 0 if the 32 bit emulation request is for a 64 bit
2818 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2819 PTROUT(p->p_sysent->sv_psstrings) : 0;
2821 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2825 ps_strings = p->p_sysent->sv_psstrings;
2827 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2832 * This sysctl allows a process to retrieve umask of another process.
2835 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2837 int *name = (int *)arg1;
2838 u_int namelen = arg2;
2847 pid = (pid_t)name[0];
2849 if (pid == p->p_pid || pid == 0) {
2850 fd_cmask = p->p_fd->fd_cmask;
2854 error = pget(pid, PGET_WANTREAD, &p);
2858 fd_cmask = p->p_fd->fd_cmask;
2861 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2866 * This sysctl allows a process to set and retrieve binary osreldate of
2870 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2872 int *name = (int *)arg1;
2873 u_int namelen = arg2;
2875 int flags, error, osrel;
2880 if (req->newptr != NULL && req->newlen != sizeof(osrel))
2883 flags = PGET_HOLD | PGET_NOTWEXIT;
2884 if (req->newptr != NULL)
2885 flags |= PGET_CANDEBUG;
2887 flags |= PGET_CANSEE;
2888 error = pget((pid_t)name[0], flags, &p);
2892 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2896 if (req->newptr != NULL) {
2897 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2912 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2914 int *name = (int *)arg1;
2915 u_int namelen = arg2;
2917 struct kinfo_sigtramp kst;
2918 const struct sysentvec *sv;
2920 #ifdef COMPAT_FREEBSD32
2921 struct kinfo_sigtramp32 kst32;
2927 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2931 #ifdef COMPAT_FREEBSD32
2932 if ((req->flags & SCTL_MASK32) != 0) {
2933 bzero(&kst32, sizeof(kst32));
2934 if (SV_PROC_FLAG(p, SV_ILP32)) {
2935 if (sv->sv_sigcode_base != 0) {
2936 kst32.ksigtramp_start = sv->sv_sigcode_base;
2937 kst32.ksigtramp_end = sv->sv_sigcode_base +
2940 kst32.ksigtramp_start = sv->sv_psstrings -
2942 kst32.ksigtramp_end = sv->sv_psstrings;
2946 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2950 bzero(&kst, sizeof(kst));
2951 if (sv->sv_sigcode_base != 0) {
2952 kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2953 kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2956 kst.ksigtramp_start = (char *)sv->sv_psstrings -
2958 kst.ksigtramp_end = (char *)sv->sv_psstrings;
2961 error = SYSCTL_OUT(req, &kst, sizeof(kst));
2966 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
2968 int *name = (int *)arg1;
2969 u_int namelen = arg2;
2974 #ifdef COMPAT_FREEBSD32
2979 if (namelen != 1 || req->newptr != NULL)
2982 pid = (pid_t)name[0];
2983 error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
2988 #ifdef COMPAT_FREEBSD32
2989 if (SV_CURPROC_FLAG(SV_ILP32)) {
2990 if (!SV_PROC_FLAG(p, SV_ILP32)) {
2996 if (pid <= PID_MAX) {
2997 td1 = FIRST_THREAD_IN_PROC(p);
2999 FOREACH_THREAD_IN_PROC(p, td1) {
3000 if (td1->td_tid == pid)
3009 * The access to the private thread flags. It is fine as far
3010 * as no out-of-thin-air values are read from td_pflags, and
3011 * usermode read of the td_sigblock_ptr is racy inherently,
3012 * since target process might have already changed it
3015 if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3016 addr = (uintptr_t)td1->td_sigblock_ptr;
3026 #ifdef COMPAT_FREEBSD32
3027 if (SV_CURPROC_FLAG(SV_ILP32)) {
3029 error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3032 error = SYSCTL_OUT(req, &addr, sizeof(addr));
3036 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
3039 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3040 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3041 "Return entire process table");
3043 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3044 sysctl_kern_proc, "Process table");
3046 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3047 sysctl_kern_proc, "Process table");
3049 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3050 sysctl_kern_proc, "Process table");
3052 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3053 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3055 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3056 sysctl_kern_proc, "Process table");
3058 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3059 sysctl_kern_proc, "Process table");
3061 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3062 sysctl_kern_proc, "Process table");
3064 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3065 sysctl_kern_proc, "Process table");
3067 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3068 sysctl_kern_proc, "Return process table, no threads");
3070 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3071 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3072 sysctl_kern_proc_args, "Process argument list");
3074 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3075 sysctl_kern_proc_env, "Process environment");
3077 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3078 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3080 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3081 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3083 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3084 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3085 "Process syscall vector name (ABI type)");
3087 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3088 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3090 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3091 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3093 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3094 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3096 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3097 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3099 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3100 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3102 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3103 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3105 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3106 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3108 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3109 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3111 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3112 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3113 "Return process table, no threads");
3115 #ifdef COMPAT_FREEBSD7
3116 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3117 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3120 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3121 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3123 #if defined(STACK) || defined(DDB)
3124 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3125 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3128 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3129 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3131 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3132 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3133 "Process resource limits");
3135 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3136 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3137 "Process ps_strings location");
3139 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3140 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3142 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3143 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3144 "Process binary osreldate");
3146 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3147 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3148 "Process signal trampoline location");
3150 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3151 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3152 "Thread sigfastblock address");
3157 * stop_all_proc() purpose is to stop all process which have usermode,
3158 * except current process for obvious reasons. This makes it somewhat
3159 * unreliable when invoked from multithreaded process. The service
3160 * must not be user-callable anyway.
3165 struct proc *cp, *p;
3167 bool restart, seen_stopped, seen_exiting, stopped_some;
3171 sx_xlock(&allproc_lock);
3173 seen_exiting = seen_stopped = stopped_some = restart = false;
3174 LIST_REMOVE(cp, p_list);
3175 LIST_INSERT_HEAD(&allproc, cp, p_list);
3177 p = LIST_NEXT(cp, p_list);
3180 LIST_REMOVE(cp, p_list);
3181 LIST_INSERT_AFTER(p, cp, p_list);
3183 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3187 if ((p->p_flag & P_WEXIT) != 0) {
3188 seen_exiting = true;
3192 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3194 * Stopped processes are tolerated when there
3195 * are no other processes which might continue
3196 * them. P_STOPPED_SINGLE but not
3197 * P_TOTAL_STOP process still has at least one
3200 seen_stopped = true;
3204 sx_xunlock(&allproc_lock);
3206 r = thread_single(p, SINGLE_ALLPROC);
3210 stopped_some = true;
3213 sx_xlock(&allproc_lock);
3215 /* Catch forked children we did not see in iteration. */
3216 if (gen != allproc_gen)
3218 sx_xunlock(&allproc_lock);
3219 if (restart || stopped_some || seen_exiting || seen_stopped) {
3220 kern_yield(PRI_USER);
3226 resume_all_proc(void)
3228 struct proc *cp, *p;
3231 sx_xlock(&allproc_lock);
3233 LIST_REMOVE(cp, p_list);
3234 LIST_INSERT_HEAD(&allproc, cp, p_list);
3236 p = LIST_NEXT(cp, p_list);
3239 LIST_REMOVE(cp, p_list);
3240 LIST_INSERT_AFTER(p, cp, p_list);
3242 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3243 sx_xunlock(&allproc_lock);
3245 thread_single_end(p, SINGLE_ALLPROC);
3248 sx_xlock(&allproc_lock);
3253 /* Did the loop above missed any stopped process ? */
3254 FOREACH_PROC_IN_SYSTEM(p) {
3255 /* No need for proc lock. */
3256 if ((p->p_flag & P_TOTAL_STOP) != 0)
3259 sx_xunlock(&allproc_lock);
3262 /* #define TOTAL_STOP_DEBUG 1 */
3263 #ifdef TOTAL_STOP_DEBUG
3264 volatile static int ap_resume;
3265 #include <sys/mount.h>
3268 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3274 error = sysctl_handle_int(oidp, &val, 0, req);
3275 if (error != 0 || req->newptr == NULL)
3280 while (ap_resume == 0)
3288 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3289 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3290 sysctl_debug_stop_all_proc, "I",