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>
45 #include <sys/eventhandler.h>
48 #include <sys/kernel.h>
49 #include <sys/limits.h>
51 #include <sys/loginclass.h>
52 #include <sys/malloc.h>
54 #include <sys/mount.h>
55 #include <sys/mutex.h>
57 #include <sys/ptrace.h>
58 #include <sys/refcount.h>
59 #include <sys/resourcevar.h>
60 #include <sys/rwlock.h>
62 #include <sys/sysent.h>
63 #include <sys/sched.h>
65 #include <sys/stack.h>
67 #include <sys/sysctl.h>
68 #include <sys/filedesc.h>
70 #include <sys/signalvar.h>
74 #include <sys/vnode.h>
82 #include <vm/vm_param.h>
83 #include <vm/vm_extern.h>
85 #include <vm/vm_map.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
90 #ifdef COMPAT_FREEBSD32
91 #include <compat/freebsd32/freebsd32.h>
92 #include <compat/freebsd32/freebsd32_util.h>
95 SDT_PROVIDER_DEFINE(proc);
97 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
98 MALLOC_DEFINE(M_SESSION, "session", "session header");
99 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
100 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
102 static void doenterpgrp(struct proc *, struct pgrp *);
103 static void orphanpg(struct pgrp *pg);
104 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
105 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
106 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
108 static void pgadjustjobc(struct pgrp *pgrp, int entering);
109 static void pgdelete(struct pgrp *);
110 static int proc_ctor(void *mem, int size, void *arg, int flags);
111 static void proc_dtor(void *mem, int size, void *arg);
112 static int proc_init(void *mem, int size, int flags);
113 static void proc_fini(void *mem, int size);
114 static void pargs_free(struct pargs *pa);
117 * Other process lists
119 struct pidhashhead *pidhashtbl;
120 struct sx *pidhashtbl_lock;
123 struct pgrphashhead *pgrphashtbl;
125 struct proclist allproc;
126 struct proclist zombproc;
127 struct sx __exclusive_cache_line allproc_lock;
128 struct sx __exclusive_cache_line zombproc_lock;
129 struct sx __exclusive_cache_line proctree_lock;
130 struct mtx __exclusive_cache_line ppeers_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 EVENTHANDLER_LIST_DECLARE(thread_ctor);
157 EVENTHANDLER_LIST_DECLARE(thread_dtor);
159 int kstack_pages = KSTACK_PAGES;
160 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
161 "Kernel stack size in pages");
162 static int vmmap_skip_res_cnt = 0;
163 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
164 &vmmap_skip_res_cnt, 0,
165 "Skip calculation of the pages resident count in kern.proc.vmmap");
167 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
168 #ifdef COMPAT_FREEBSD32
169 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
173 * Initialize global process hashing structures.
180 sx_init(&allproc_lock, "allproc");
181 sx_init(&zombproc_lock, "zombproc");
182 sx_init(&proctree_lock, "proctree");
183 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
185 LIST_INIT(&zombproc);
186 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
187 pidhashlock = (pidhash + 1) / 64;
190 pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
191 M_PROC, M_WAITOK | M_ZERO);
192 for (i = 0; i < pidhashlock + 1; i++)
193 sx_init(&pidhashtbl_lock[i], "pidhash");
194 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
195 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
196 proc_ctor, proc_dtor, proc_init, proc_fini,
197 UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
202 * Prepare a proc for use.
205 proc_ctor(void *mem, int size, void *arg, int flags)
210 p = (struct proc *)mem;
211 EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
212 td = FIRST_THREAD_IN_PROC(p);
214 /* Make sure all thread constructors are executed */
215 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
221 * Reclaim a proc after use.
224 proc_dtor(void *mem, int size, void *arg)
229 /* INVARIANTS checks go here */
230 p = (struct proc *)mem;
231 td = FIRST_THREAD_IN_PROC(p);
234 KASSERT((p->p_numthreads == 1),
235 ("bad number of threads in exiting process"));
236 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
238 /* Free all OSD associated to this thread. */
240 td_softdep_cleanup(td);
241 MPASS(td->td_su == NULL);
243 /* Make sure all thread destructors are executed */
244 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
246 EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
247 if (p->p_ksi != NULL)
248 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
252 * Initialize type-stable parts of a proc (when newly created).
255 proc_init(void *mem, int size, int flags)
259 p = (struct proc *)mem;
260 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
261 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
262 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
263 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
264 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
265 cv_init(&p->p_pwait, "ppwait");
266 TAILQ_INIT(&p->p_threads); /* all threads in proc */
267 EVENTHANDLER_DIRECT_INVOKE(process_init, p);
268 p->p_stats = pstats_alloc();
274 * UMA should ensure that this function is never called.
275 * Freeing a proc structure would violate type stability.
278 proc_fini(void *mem, int size)
283 p = (struct proc *)mem;
284 EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
285 pstats_free(p->p_stats);
286 thread_free(FIRST_THREAD_IN_PROC(p));
287 mtx_destroy(&p->p_mtx);
288 if (p->p_ksi != NULL)
289 ksiginfo_free(p->p_ksi);
291 panic("proc reclaimed");
296 * Is p an inferior of the current process?
299 inferior(struct proc *p)
302 sx_assert(&proctree_lock, SX_LOCKED);
303 PROC_LOCK_ASSERT(p, MA_OWNED);
304 for (; p != curproc; p = proc_realparent(p)) {
312 * Locate a process by number.
314 * By not returning processes in the PRS_NEW state, we allow callers to avoid
315 * testing for that condition to avoid dereferencing p_ucred, et al.
317 static __always_inline struct proc *
318 _pfind(pid_t pid, bool zombie)
323 if (p->p_pid == pid) {
327 sx_slock(PIDHASHLOCK(pid));
328 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
329 if (p->p_pid == pid) {
331 if (p->p_state == PRS_NEW ||
332 (zombie && p->p_state == PRS_ZOMBIE)) {
339 sx_sunlock(PIDHASHLOCK(pid));
347 return (_pfind(pid, false));
351 * Same as pfind but allow zombies.
357 return (_pfind(pid, true));
366 sx_slock(&allproc_lock);
367 FOREACH_PROC_IN_SYSTEM(p) {
369 if (p->p_state == PRS_NEW) {
373 FOREACH_THREAD_IN_PROC(p, td) {
374 if (td->td_tid == tid)
380 sx_sunlock(&allproc_lock);
385 * Locate a process group by number.
386 * The caller must hold proctree_lock.
393 sx_assert(&proctree_lock, SX_LOCKED);
395 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
396 if (pgrp->pg_id == pgid) {
405 * Locate process and do additional manipulations, depending on flags.
408 pget(pid_t pid, int flags, struct proc **pp)
414 if (p->p_pid == pid) {
418 if (pid <= PID_MAX) {
419 if ((flags & PGET_NOTWEXIT) == 0)
423 } else if ((flags & PGET_NOTID) == 0) {
428 if ((flags & PGET_CANSEE) != 0) {
429 error = p_cansee(curthread, p);
434 if ((flags & PGET_CANDEBUG) != 0) {
435 error = p_candebug(curthread, p);
439 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
443 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
447 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
449 * XXXRW: Not clear ESRCH is the right error during proc
455 if ((flags & PGET_HOLD) != 0) {
467 * Create a new process group.
468 * pgid must be equal to the pid of p.
469 * Begin a new session if required.
472 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
475 sx_assert(&proctree_lock, SX_XLOCKED);
477 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
478 KASSERT(p->p_pid == pgid,
479 ("enterpgrp: new pgrp and pid != pgid"));
480 KASSERT(pgfind(pgid) == NULL,
481 ("enterpgrp: pgrp with pgid exists"));
482 KASSERT(!SESS_LEADER(p),
483 ("enterpgrp: session leader attempted setpgrp"));
485 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
491 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
493 p->p_flag &= ~P_CONTROLT;
497 sess->s_sid = p->p_pid;
498 refcount_init(&sess->s_count, 1);
499 sess->s_ttyvp = NULL;
500 sess->s_ttydp = NULL;
502 bcopy(p->p_session->s_login, sess->s_login,
503 sizeof(sess->s_login));
504 pgrp->pg_session = sess;
505 KASSERT(p == curproc,
506 ("enterpgrp: mksession and p != curproc"));
508 pgrp->pg_session = p->p_session;
509 sess_hold(pgrp->pg_session);
513 LIST_INIT(&pgrp->pg_members);
516 * As we have an exclusive lock of proctree_lock,
517 * this should not deadlock.
519 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
521 SLIST_INIT(&pgrp->pg_sigiolst);
524 doenterpgrp(p, pgrp);
530 * Move p to an existing process group
533 enterthispgrp(struct proc *p, struct pgrp *pgrp)
536 sx_assert(&proctree_lock, SX_XLOCKED);
537 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
538 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
539 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
540 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
541 KASSERT(pgrp->pg_session == p->p_session,
542 ("%s: pgrp's session %p, p->p_session %p.\n",
546 KASSERT(pgrp != p->p_pgrp,
547 ("%s: p belongs to pgrp.", __func__));
549 doenterpgrp(p, pgrp);
555 * Move p to a process group
558 doenterpgrp(struct proc *p, struct pgrp *pgrp)
560 struct pgrp *savepgrp;
562 sx_assert(&proctree_lock, SX_XLOCKED);
563 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
564 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
565 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
566 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
568 savepgrp = p->p_pgrp;
571 * Adjust eligibility of affected pgrps to participate in job control.
572 * Increment eligibility counts before decrementing, otherwise we
573 * could reach 0 spuriously during the first call.
576 fixjobc(p, p->p_pgrp, 0);
581 LIST_REMOVE(p, p_pglist);
584 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
585 PGRP_UNLOCK(savepgrp);
587 if (LIST_EMPTY(&savepgrp->pg_members))
592 * remove process from process group
595 leavepgrp(struct proc *p)
597 struct pgrp *savepgrp;
599 sx_assert(&proctree_lock, SX_XLOCKED);
600 savepgrp = p->p_pgrp;
603 LIST_REMOVE(p, p_pglist);
606 PGRP_UNLOCK(savepgrp);
607 if (LIST_EMPTY(&savepgrp->pg_members))
613 * delete a process group
616 pgdelete(struct pgrp *pgrp)
618 struct session *savesess;
621 sx_assert(&proctree_lock, SX_XLOCKED);
622 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
623 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
626 * Reset any sigio structures pointing to us as a result of
627 * F_SETOWN with our pgid.
629 funsetownlst(&pgrp->pg_sigiolst);
632 tp = pgrp->pg_session->s_ttyp;
633 LIST_REMOVE(pgrp, pg_hash);
634 savesess = pgrp->pg_session;
637 /* Remove the reference to the pgrp before deallocating it. */
640 tty_rel_pgrp(tp, pgrp);
643 mtx_destroy(&pgrp->pg_mtx);
645 sess_release(savesess);
649 pgadjustjobc(struct pgrp *pgrp, int entering)
657 if (pgrp->pg_jobc == 0)
664 * Adjust pgrp jobc counters when specified process changes process group.
665 * We count the number of processes in each process group that "qualify"
666 * the group for terminal job control (those with a parent in a different
667 * process group of the same session). If that count reaches zero, the
668 * process group becomes orphaned. Check both the specified process'
669 * process group and that of its children.
670 * entering == 0 => p is leaving specified group.
671 * entering == 1 => p is entering specified group.
674 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
676 struct pgrp *hispgrp;
677 struct session *mysession;
680 sx_assert(&proctree_lock, SX_LOCKED);
681 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
682 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
683 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
686 * Check p's parent to see whether p qualifies its own process
687 * group; if so, adjust count for p's process group.
689 mysession = pgrp->pg_session;
690 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
691 hispgrp->pg_session == mysession)
692 pgadjustjobc(pgrp, entering);
695 * Check this process' children to see whether they qualify
696 * their process groups; if so, adjust counts for children's
699 LIST_FOREACH(q, &p->p_children, p_sibling) {
701 if (hispgrp == pgrp ||
702 hispgrp->pg_session != mysession)
704 if (q->p_state == PRS_ZOMBIE)
706 pgadjustjobc(hispgrp, entering);
719 MPASS(p->p_flag & P_WEXIT);
721 * Do a quick check to see if there is anything to do with the
722 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
725 if (!SESS_LEADER(p) &&
726 (p->p_pgrp == p->p_pptr->p_pgrp) &&
727 LIST_EMPTY(&p->p_children)) {
733 sx_xlock(&proctree_lock);
734 if (SESS_LEADER(p)) {
738 * s_ttyp is not zero'd; we use this to indicate that
739 * the session once had a controlling terminal. (for
740 * logging and informational purposes)
751 * Signal foreground pgrp and revoke access to
752 * controlling terminal if it has not been revoked
755 * Because the TTY may have been revoked in the mean
756 * time and could already have a new session associated
757 * with it, make sure we don't send a SIGHUP to a
758 * foreground process group that does not belong to this
764 if (tp->t_session == sp)
765 tty_signal_pgrp(tp, SIGHUP);
770 sx_xunlock(&proctree_lock);
771 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
772 VOP_REVOKE(ttyvp, REVOKEALL);
773 VOP_UNLOCK(ttyvp, 0);
776 sx_xlock(&proctree_lock);
779 fixjobc(p, p->p_pgrp, 0);
780 sx_xunlock(&proctree_lock);
784 * A process group has become orphaned;
785 * if there are any stopped processes in the group,
786 * hang-up all process in that group.
789 orphanpg(struct pgrp *pg)
793 PGRP_LOCK_ASSERT(pg, MA_OWNED);
795 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
797 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
799 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
801 kern_psignal(p, SIGHUP);
802 kern_psignal(p, SIGCONT);
812 sess_hold(struct session *s)
815 refcount_acquire(&s->s_count);
819 sess_release(struct session *s)
822 if (refcount_release(&s->s_count)) {
823 if (s->s_ttyp != NULL) {
825 tty_rel_sess(s->s_ttyp, s);
827 mtx_destroy(&s->s_mtx);
834 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
840 for (i = 0; i <= pgrphash; i++) {
841 if (!LIST_EMPTY(&pgrphashtbl[i])) {
842 printf("\tindx %d\n", i);
843 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
845 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
846 (void *)pgrp, (long)pgrp->pg_id,
847 (void *)pgrp->pg_session,
848 pgrp->pg_session->s_count,
849 (void *)LIST_FIRST(&pgrp->pg_members));
850 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
851 printf("\t\tpid %ld addr %p pgrp %p\n",
852 (long)p->p_pid, (void *)p,
862 * Calculate the kinfo_proc members which contain process-wide
864 * Must be called with the target process locked.
867 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
871 PROC_LOCK_ASSERT(p, MA_OWNED);
875 FOREACH_THREAD_IN_PROC(p, td) {
877 kp->ki_pctcpu += sched_pctcpu(td);
878 kp->ki_estcpu += sched_estcpu(td);
884 * Clear kinfo_proc and fill in any information that is common
885 * to all threads in the process.
886 * Must be called with the target process locked.
889 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
896 struct timeval boottime;
898 PROC_LOCK_ASSERT(p, MA_OWNED);
899 bzero(kp, sizeof(*kp));
901 kp->ki_structsize = sizeof(*kp);
903 kp->ki_addr =/* p->p_addr; */0; /* XXX */
904 kp->ki_args = p->p_args;
905 kp->ki_textvp = p->p_textvp;
907 kp->ki_tracep = p->p_tracevp;
908 kp->ki_traceflag = p->p_traceflag;
911 kp->ki_vmspace = p->p_vmspace;
912 kp->ki_flag = p->p_flag;
913 kp->ki_flag2 = p->p_flag2;
916 kp->ki_uid = cred->cr_uid;
917 kp->ki_ruid = cred->cr_ruid;
918 kp->ki_svuid = cred->cr_svuid;
920 if (cred->cr_flags & CRED_FLAG_CAPMODE)
921 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
922 /* XXX bde doesn't like KI_NGROUPS */
923 if (cred->cr_ngroups > KI_NGROUPS) {
924 kp->ki_ngroups = KI_NGROUPS;
925 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
927 kp->ki_ngroups = cred->cr_ngroups;
928 bcopy(cred->cr_groups, kp->ki_groups,
929 kp->ki_ngroups * sizeof(gid_t));
930 kp->ki_rgid = cred->cr_rgid;
931 kp->ki_svgid = cred->cr_svgid;
932 /* If jailed(cred), emulate the old P_JAILED flag. */
934 kp->ki_flag |= P_JAILED;
935 /* If inside the jail, use 0 as a jail ID. */
936 if (cred->cr_prison != curthread->td_ucred->cr_prison)
937 kp->ki_jid = cred->cr_prison->pr_id;
939 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
940 sizeof(kp->ki_loginclass));
944 mtx_lock(&ps->ps_mtx);
945 kp->ki_sigignore = ps->ps_sigignore;
946 kp->ki_sigcatch = ps->ps_sigcatch;
947 mtx_unlock(&ps->ps_mtx);
949 if (p->p_state != PRS_NEW &&
950 p->p_state != PRS_ZOMBIE &&
951 p->p_vmspace != NULL) {
952 struct vmspace *vm = p->p_vmspace;
954 kp->ki_size = vm->vm_map.size;
955 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
956 FOREACH_THREAD_IN_PROC(p, td0) {
957 if (!TD_IS_SWAPPED(td0))
958 kp->ki_rssize += td0->td_kstack_pages;
960 kp->ki_swrss = vm->vm_swrss;
961 kp->ki_tsize = vm->vm_tsize;
962 kp->ki_dsize = vm->vm_dsize;
963 kp->ki_ssize = vm->vm_ssize;
964 } else if (p->p_state == PRS_ZOMBIE)
966 if (kp->ki_flag & P_INMEM)
967 kp->ki_sflag = PS_INMEM;
970 /* Calculate legacy swtime as seconds since 'swtick'. */
971 kp->ki_swtime = (ticks - p->p_swtick) / hz;
972 kp->ki_pid = p->p_pid;
973 kp->ki_nice = p->p_nice;
974 kp->ki_fibnum = p->p_fibnum;
975 kp->ki_start = p->p_stats->p_start;
976 getboottime(&boottime);
977 timevaladd(&kp->ki_start, &boottime);
979 rufetch(p, &kp->ki_rusage);
980 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
981 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
983 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
984 /* Some callers want child times in a single value. */
985 kp->ki_childtime = kp->ki_childstime;
986 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
988 FOREACH_THREAD_IN_PROC(p, td0)
989 kp->ki_cow += td0->td_cow;
993 kp->ki_pgid = p->p_pgrp->pg_id;
994 kp->ki_jobc = p->p_pgrp->pg_jobc;
995 sp = p->p_pgrp->pg_session;
998 kp->ki_sid = sp->s_sid;
1000 strlcpy(kp->ki_login, sp->s_login,
1001 sizeof(kp->ki_login));
1003 kp->ki_kiflag |= KI_CTTY;
1005 kp->ki_kiflag |= KI_SLEADER;
1006 /* XXX proctree_lock */
1011 if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1012 kp->ki_tdev = tty_udev(tp);
1013 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1014 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1016 kp->ki_tsid = tp->t_session->s_sid;
1018 kp->ki_tdev = NODEV;
1019 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1021 if (p->p_comm[0] != '\0')
1022 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1023 if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1024 p->p_sysent->sv_name[0] != '\0')
1025 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1026 kp->ki_siglist = p->p_siglist;
1027 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1028 kp->ki_acflag = p->p_acflag;
1029 kp->ki_lock = p->p_lock;
1031 kp->ki_ppid = p->p_oppid;
1032 if (p->p_flag & P_TRACED)
1033 kp->ki_tracer = p->p_pptr->p_pid;
1038 * Fill in information that is thread specific. Must be called with
1039 * target process locked. If 'preferthread' is set, overwrite certain
1040 * process-related fields that are maintained for both threads and
1044 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1050 PROC_LOCK_ASSERT(p, MA_OWNED);
1055 if (td->td_wmesg != NULL)
1056 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1058 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1059 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1060 sizeof(kp->ki_tdname)) {
1061 strlcpy(kp->ki_moretdname,
1062 td->td_name + sizeof(kp->ki_tdname) - 1,
1063 sizeof(kp->ki_moretdname));
1065 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1067 if (TD_ON_LOCK(td)) {
1068 kp->ki_kiflag |= KI_LOCKBLOCK;
1069 strlcpy(kp->ki_lockname, td->td_lockname,
1070 sizeof(kp->ki_lockname));
1072 kp->ki_kiflag &= ~KI_LOCKBLOCK;
1073 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1076 if (p->p_state == PRS_NORMAL) { /* approximate. */
1077 if (TD_ON_RUNQ(td) ||
1079 TD_IS_RUNNING(td)) {
1081 } else if (P_SHOULDSTOP(p)) {
1082 kp->ki_stat = SSTOP;
1083 } else if (TD_IS_SLEEPING(td)) {
1084 kp->ki_stat = SSLEEP;
1085 } else if (TD_ON_LOCK(td)) {
1086 kp->ki_stat = SLOCK;
1088 kp->ki_stat = SWAIT;
1090 } else if (p->p_state == PRS_ZOMBIE) {
1091 kp->ki_stat = SZOMB;
1096 /* Things in the thread */
1097 kp->ki_wchan = td->td_wchan;
1098 kp->ki_pri.pri_level = td->td_priority;
1099 kp->ki_pri.pri_native = td->td_base_pri;
1102 * Note: legacy fields; clamp at the old NOCPU value and/or
1103 * the maximum u_char CPU value.
1105 if (td->td_lastcpu == NOCPU)
1106 kp->ki_lastcpu_old = NOCPU_OLD;
1107 else if (td->td_lastcpu > MAXCPU_OLD)
1108 kp->ki_lastcpu_old = MAXCPU_OLD;
1110 kp->ki_lastcpu_old = td->td_lastcpu;
1112 if (td->td_oncpu == NOCPU)
1113 kp->ki_oncpu_old = NOCPU_OLD;
1114 else if (td->td_oncpu > MAXCPU_OLD)
1115 kp->ki_oncpu_old = MAXCPU_OLD;
1117 kp->ki_oncpu_old = td->td_oncpu;
1119 kp->ki_lastcpu = td->td_lastcpu;
1120 kp->ki_oncpu = td->td_oncpu;
1121 kp->ki_tdflags = td->td_flags;
1122 kp->ki_tid = td->td_tid;
1123 kp->ki_numthreads = p->p_numthreads;
1124 kp->ki_pcb = td->td_pcb;
1125 kp->ki_kstack = (void *)td->td_kstack;
1126 kp->ki_slptime = (ticks - td->td_slptick) / hz;
1127 kp->ki_pri.pri_class = td->td_pri_class;
1128 kp->ki_pri.pri_user = td->td_user_pri;
1131 rufetchtd(td, &kp->ki_rusage);
1132 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1133 kp->ki_pctcpu = sched_pctcpu(td);
1134 kp->ki_estcpu = sched_estcpu(td);
1135 kp->ki_cow = td->td_cow;
1138 /* We can't get this anymore but ps etc never used it anyway. */
1142 kp->ki_siglist = td->td_siglist;
1143 kp->ki_sigmask = td->td_sigmask;
1150 * Fill in a kinfo_proc structure for the specified process.
1151 * Must be called with the target process locked.
1154 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1157 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1159 fill_kinfo_proc_only(p, kp);
1160 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1161 fill_kinfo_aggregate(p, kp);
1168 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1172 * Copy parts of p_stats; zero the rest of p_stats (statistics).
1175 pstats_fork(struct pstats *src, struct pstats *dst)
1178 bzero(&dst->pstat_startzero,
1179 __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1180 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1181 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1185 pstats_free(struct pstats *ps)
1188 free(ps, M_SUBPROC);
1192 * Locate a zombie process by number
1199 sx_slock(&zombproc_lock);
1200 LIST_FOREACH(p, &zombproc, p_list) {
1201 if (p->p_pid == pid) {
1206 sx_sunlock(&zombproc_lock);
1210 #ifdef COMPAT_FREEBSD32
1213 * This function is typically used to copy out the kernel address, so
1214 * it can be replaced by assignment of zero.
1216 static inline uint32_t
1217 ptr32_trim(void *ptr)
1221 uptr = (uintptr_t)ptr;
1222 return ((uptr > UINT_MAX) ? 0 : uptr);
1225 #define PTRTRIM_CP(src,dst,fld) \
1226 do { (dst).fld = ptr32_trim((src).fld); } while (0)
1229 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1233 bzero(ki32, sizeof(struct kinfo_proc32));
1234 ki32->ki_structsize = sizeof(struct kinfo_proc32);
1235 CP(*ki, *ki32, ki_layout);
1236 PTRTRIM_CP(*ki, *ki32, ki_args);
1237 PTRTRIM_CP(*ki, *ki32, ki_paddr);
1238 PTRTRIM_CP(*ki, *ki32, ki_addr);
1239 PTRTRIM_CP(*ki, *ki32, ki_tracep);
1240 PTRTRIM_CP(*ki, *ki32, ki_textvp);
1241 PTRTRIM_CP(*ki, *ki32, ki_fd);
1242 PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1243 PTRTRIM_CP(*ki, *ki32, ki_wchan);
1244 CP(*ki, *ki32, ki_pid);
1245 CP(*ki, *ki32, ki_ppid);
1246 CP(*ki, *ki32, ki_pgid);
1247 CP(*ki, *ki32, ki_tpgid);
1248 CP(*ki, *ki32, ki_sid);
1249 CP(*ki, *ki32, ki_tsid);
1250 CP(*ki, *ki32, ki_jobc);
1251 CP(*ki, *ki32, ki_tdev);
1252 CP(*ki, *ki32, ki_tdev_freebsd11);
1253 CP(*ki, *ki32, ki_siglist);
1254 CP(*ki, *ki32, ki_sigmask);
1255 CP(*ki, *ki32, ki_sigignore);
1256 CP(*ki, *ki32, ki_sigcatch);
1257 CP(*ki, *ki32, ki_uid);
1258 CP(*ki, *ki32, ki_ruid);
1259 CP(*ki, *ki32, ki_svuid);
1260 CP(*ki, *ki32, ki_rgid);
1261 CP(*ki, *ki32, ki_svgid);
1262 CP(*ki, *ki32, ki_ngroups);
1263 for (i = 0; i < KI_NGROUPS; i++)
1264 CP(*ki, *ki32, ki_groups[i]);
1265 CP(*ki, *ki32, ki_size);
1266 CP(*ki, *ki32, ki_rssize);
1267 CP(*ki, *ki32, ki_swrss);
1268 CP(*ki, *ki32, ki_tsize);
1269 CP(*ki, *ki32, ki_dsize);
1270 CP(*ki, *ki32, ki_ssize);
1271 CP(*ki, *ki32, ki_xstat);
1272 CP(*ki, *ki32, ki_acflag);
1273 CP(*ki, *ki32, ki_pctcpu);
1274 CP(*ki, *ki32, ki_estcpu);
1275 CP(*ki, *ki32, ki_slptime);
1276 CP(*ki, *ki32, ki_swtime);
1277 CP(*ki, *ki32, ki_cow);
1278 CP(*ki, *ki32, ki_runtime);
1279 TV_CP(*ki, *ki32, ki_start);
1280 TV_CP(*ki, *ki32, ki_childtime);
1281 CP(*ki, *ki32, ki_flag);
1282 CP(*ki, *ki32, ki_kiflag);
1283 CP(*ki, *ki32, ki_traceflag);
1284 CP(*ki, *ki32, ki_stat);
1285 CP(*ki, *ki32, ki_nice);
1286 CP(*ki, *ki32, ki_lock);
1287 CP(*ki, *ki32, ki_rqindex);
1288 CP(*ki, *ki32, ki_oncpu);
1289 CP(*ki, *ki32, ki_lastcpu);
1291 /* XXX TODO: wrap cpu value as appropriate */
1292 CP(*ki, *ki32, ki_oncpu_old);
1293 CP(*ki, *ki32, ki_lastcpu_old);
1295 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1296 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1297 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1298 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1299 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1300 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1301 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1302 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1303 CP(*ki, *ki32, ki_tracer);
1304 CP(*ki, *ki32, ki_flag2);
1305 CP(*ki, *ki32, ki_fibnum);
1306 CP(*ki, *ki32, ki_cr_flags);
1307 CP(*ki, *ki32, ki_jid);
1308 CP(*ki, *ki32, ki_numthreads);
1309 CP(*ki, *ki32, ki_tid);
1310 CP(*ki, *ki32, ki_pri);
1311 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1312 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1313 PTRTRIM_CP(*ki, *ki32, ki_pcb);
1314 PTRTRIM_CP(*ki, *ki32, ki_kstack);
1315 PTRTRIM_CP(*ki, *ki32, ki_udata);
1316 PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1317 CP(*ki, *ki32, ki_sflag);
1318 CP(*ki, *ki32, ki_tdflags);
1323 kern_proc_out_size(struct proc *p, int flags)
1327 PROC_LOCK_ASSERT(p, MA_OWNED);
1329 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1330 #ifdef COMPAT_FREEBSD32
1331 if ((flags & KERN_PROC_MASK32) != 0) {
1332 size += sizeof(struct kinfo_proc32);
1335 size += sizeof(struct kinfo_proc);
1337 #ifdef COMPAT_FREEBSD32
1338 if ((flags & KERN_PROC_MASK32) != 0)
1339 size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1342 size += sizeof(struct kinfo_proc) * p->p_numthreads;
1349 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1352 struct kinfo_proc ki;
1353 #ifdef COMPAT_FREEBSD32
1354 struct kinfo_proc32 ki32;
1358 PROC_LOCK_ASSERT(p, MA_OWNED);
1359 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1362 fill_kinfo_proc(p, &ki);
1363 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1364 #ifdef COMPAT_FREEBSD32
1365 if ((flags & KERN_PROC_MASK32) != 0) {
1366 freebsd32_kinfo_proc_out(&ki, &ki32);
1367 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1371 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1374 FOREACH_THREAD_IN_PROC(p, td) {
1375 fill_kinfo_thread(td, &ki, 1);
1376 #ifdef COMPAT_FREEBSD32
1377 if ((flags & KERN_PROC_MASK32) != 0) {
1378 freebsd32_kinfo_proc_out(&ki, &ki32);
1379 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1383 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1394 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1397 struct kinfo_proc ki;
1400 if (req->oldptr == NULL)
1401 return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1403 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1404 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1405 error = kern_proc_out(p, &sb, flags);
1406 error2 = sbuf_finish(&sb);
1410 else if (error2 != 0)
1416 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1421 for (i = 0; i < pidhashlock + 1; i++) {
1422 sx_slock(&pidhashtbl_lock[i]);
1423 for (j = i; j <= pidhash; j += pidhashlock + 1) {
1424 LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1425 if (p->p_state == PRS_NEW)
1427 error = cb(p, cbarg);
1428 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1430 sx_sunlock(&pidhashtbl_lock[i]);
1435 sx_sunlock(&pidhashtbl_lock[i]);
1440 struct kern_proc_out_args {
1441 struct sysctl_req *req;
1448 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1450 struct kern_proc_out_args *arg = origarg;
1451 int *name = arg->name;
1452 int oid_number = arg->oid_number;
1453 int flags = arg->flags;
1454 struct sysctl_req *req = arg->req;
1459 KASSERT(p->p_ucred != NULL,
1460 ("process credential is NULL for non-NEW proc"));
1462 * Show a user only appropriate processes.
1464 if (p_cansee(curthread, p))
1467 * TODO - make more efficient (see notes below).
1470 switch (oid_number) {
1473 if (p->p_ucred->cr_gid != (gid_t)name[0])
1477 case KERN_PROC_PGRP:
1478 /* could do this by traversing pgrp */
1479 if (p->p_pgrp == NULL ||
1480 p->p_pgrp->pg_id != (pid_t)name[0])
1484 case KERN_PROC_RGID:
1485 if (p->p_ucred->cr_rgid != (gid_t)name[0])
1489 case KERN_PROC_SESSION:
1490 if (p->p_session == NULL ||
1491 p->p_session->s_sid != (pid_t)name[0])
1496 if ((p->p_flag & P_CONTROLT) == 0 ||
1497 p->p_session == NULL)
1499 /* XXX proctree_lock */
1500 SESS_LOCK(p->p_session);
1501 if (p->p_session->s_ttyp == NULL ||
1502 tty_udev(p->p_session->s_ttyp) !=
1504 SESS_UNLOCK(p->p_session);
1507 SESS_UNLOCK(p->p_session);
1511 if (p->p_ucred->cr_uid != (uid_t)name[0])
1515 case KERN_PROC_RUID:
1516 if (p->p_ucred->cr_ruid != (uid_t)name[0])
1520 case KERN_PROC_PROC:
1527 error = sysctl_out_proc(p, req, flags);
1528 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1536 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1538 struct kern_proc_out_args iterarg;
1539 int *name = (int *)arg1;
1540 u_int namelen = arg2;
1542 int flags, oid_number;
1545 oid_number = oidp->oid_number;
1546 if (oid_number != KERN_PROC_ALL &&
1547 (oid_number & KERN_PROC_INC_THREAD) == 0)
1548 flags = KERN_PROC_NOTHREADS;
1551 oid_number &= ~KERN_PROC_INC_THREAD;
1553 #ifdef COMPAT_FREEBSD32
1554 if (req->flags & SCTL_MASK32)
1555 flags |= KERN_PROC_MASK32;
1557 if (oid_number == KERN_PROC_PID) {
1560 error = sysctl_wire_old_buffer(req, 0);
1563 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1565 error = sysctl_out_proc(p, req, flags);
1569 switch (oid_number) {
1574 case KERN_PROC_PROC:
1575 if (namelen != 0 && namelen != 1)
1584 if (req->oldptr == NULL) {
1585 /* overestimate by 5 procs */
1586 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1590 error = sysctl_wire_old_buffer(req, 0);
1594 iterarg.flags = flags;
1595 iterarg.oid_number = oid_number;
1597 iterarg.name = name;
1598 error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1603 pargs_alloc(int len)
1607 pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1609 refcount_init(&pa->ar_ref, 1);
1610 pa->ar_length = len;
1615 pargs_free(struct pargs *pa)
1622 pargs_hold(struct pargs *pa)
1627 refcount_acquire(&pa->ar_ref);
1631 pargs_drop(struct pargs *pa)
1636 if (refcount_release(&pa->ar_ref))
1641 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1647 * This may return a short read if the string is shorter than the chunk
1648 * and is aligned at the end of the page, and the following page is not
1651 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1657 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */
1659 enum proc_vector_type {
1665 #ifdef COMPAT_FREEBSD32
1667 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1668 size_t *vsizep, enum proc_vector_type type)
1670 struct freebsd32_ps_strings pss;
1672 vm_offset_t vptr, ptr;
1673 uint32_t *proc_vector32;
1679 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1680 sizeof(pss)) != sizeof(pss))
1684 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1685 vsize = pss.ps_nargvstr;
1686 if (vsize > ARG_MAX)
1688 size = vsize * sizeof(int32_t);
1691 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1692 vsize = pss.ps_nenvstr;
1693 if (vsize > ARG_MAX)
1695 size = vsize * sizeof(int32_t);
1698 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1699 (pss.ps_nenvstr + 1) * sizeof(int32_t);
1702 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1703 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1706 if (aux.a_type == AT_NULL)
1710 if (aux.a_type != AT_NULL)
1713 size = vsize * sizeof(aux);
1716 KASSERT(0, ("Wrong proc vector type: %d", type));
1719 proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1720 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1724 if (type == PROC_AUX) {
1725 *proc_vectorp = (char **)proc_vector32;
1729 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1730 for (i = 0; i < (int)vsize; i++)
1731 proc_vector[i] = PTRIN(proc_vector32[i]);
1732 *proc_vectorp = proc_vector;
1735 free(proc_vector32, M_TEMP);
1741 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1742 size_t *vsizep, enum proc_vector_type type)
1744 struct ps_strings pss;
1746 vm_offset_t vptr, ptr;
1751 #ifdef COMPAT_FREEBSD32
1752 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1753 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1755 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1756 sizeof(pss)) != sizeof(pss))
1760 vptr = (vm_offset_t)pss.ps_argvstr;
1761 vsize = pss.ps_nargvstr;
1762 if (vsize > ARG_MAX)
1764 size = vsize * sizeof(char *);
1767 vptr = (vm_offset_t)pss.ps_envstr;
1768 vsize = pss.ps_nenvstr;
1769 if (vsize > ARG_MAX)
1771 size = vsize * sizeof(char *);
1775 * The aux array is just above env array on the stack. Check
1776 * that the address is naturally aligned.
1778 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1780 #if __ELF_WORD_SIZE == 64
1781 if (vptr % sizeof(uint64_t) != 0)
1783 if (vptr % sizeof(uint32_t) != 0)
1787 * We count the array size reading the aux vectors from the
1788 * stack until AT_NULL vector is returned. So (to keep the code
1789 * simple) we read the process stack twice: the first time here
1790 * to find the size and the second time when copying the vectors
1791 * to the allocated proc_vector.
1793 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1794 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1797 if (aux.a_type == AT_NULL)
1802 * If the PROC_AUXV_MAX entries are iterated over, and we have
1803 * not reached AT_NULL, it is most likely we are reading wrong
1804 * data: either the process doesn't have auxv array or data has
1805 * been modified. Return the error in this case.
1807 if (aux.a_type != AT_NULL)
1810 size = vsize * sizeof(aux);
1813 KASSERT(0, ("Wrong proc vector type: %d", type));
1814 return (EINVAL); /* In case we are built without INVARIANTS. */
1816 proc_vector = malloc(size, M_TEMP, M_WAITOK);
1817 if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1818 free(proc_vector, M_TEMP);
1821 *proc_vectorp = proc_vector;
1827 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */
1830 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1831 enum proc_vector_type type)
1833 size_t done, len, nchr, vsize;
1835 char **proc_vector, *sptr;
1836 char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1838 PROC_ASSERT_HELD(p);
1841 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1843 nchr = 2 * (PATH_MAX + ARG_MAX);
1845 error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1848 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1850 * The program may have scribbled into its argv array, e.g. to
1851 * remove some arguments. If that has happened, break out
1852 * before trying to read from NULL.
1854 if (proc_vector[i] == NULL)
1856 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1857 error = proc_read_string(td, p, sptr, pss_string,
1858 sizeof(pss_string));
1861 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1862 if (done + len >= nchr)
1863 len = nchr - done - 1;
1864 sbuf_bcat(sb, pss_string, len);
1865 if (len != GET_PS_STRINGS_CHUNK_SZ)
1867 done += GET_PS_STRINGS_CHUNK_SZ;
1869 sbuf_bcat(sb, "", 1);
1873 free(proc_vector, M_TEMP);
1878 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1881 return (get_ps_strings(curthread, p, sb, PROC_ARG));
1885 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1888 return (get_ps_strings(curthread, p, sb, PROC_ENV));
1892 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1898 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1900 #ifdef COMPAT_FREEBSD32
1901 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1902 size = vsize * sizeof(Elf32_Auxinfo);
1905 size = vsize * sizeof(Elf_Auxinfo);
1906 if (sbuf_bcat(sb, auxv, size) != 0)
1914 * This sysctl allows a process to retrieve the argument list or process
1915 * title for another process without groping around in the address space
1916 * of the other process. It also allow a process to set its own "process
1917 * title to a string of its own choice.
1920 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1922 int *name = (int *)arg1;
1923 u_int namelen = arg2;
1924 struct pargs *newpa, *pa;
1927 int flags, error = 0, error2;
1933 pid = (pid_t)name[0];
1935 * If the query is for this process and it is single-threaded, there
1936 * is nobody to modify pargs, thus we can just read.
1939 if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
1940 (pa = p->p_args) != NULL)
1941 return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
1943 flags = PGET_CANSEE;
1944 if (req->newptr != NULL)
1945 flags |= PGET_ISCURRENT;
1946 error = pget(pid, flags, &p);
1954 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1956 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1959 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1960 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1961 error = proc_getargv(curthread, p, &sb);
1962 error2 = sbuf_finish(&sb);
1965 if (error == 0 && error2 != 0)
1970 if (error != 0 || req->newptr == NULL)
1973 if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
1976 if (req->newlen == 0) {
1978 * Clear the argument pointer, so that we'll fetch arguments
1979 * with proc_getargv() until further notice.
1983 newpa = pargs_alloc(req->newlen);
1984 error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1999 * This sysctl allows a process to retrieve environment of another process.
2002 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2004 int *name = (int *)arg1;
2005 u_int namelen = arg2;
2013 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2016 if ((p->p_flag & P_SYSTEM) != 0) {
2021 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2022 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2023 error = proc_getenvv(curthread, p, &sb);
2024 error2 = sbuf_finish(&sb);
2027 return (error != 0 ? error : error2);
2031 * This sysctl allows a process to retrieve ELF auxiliary vector of
2035 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2037 int *name = (int *)arg1;
2038 u_int namelen = arg2;
2046 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2049 if ((p->p_flag & P_SYSTEM) != 0) {
2053 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2054 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2055 error = proc_getauxv(curthread, p, &sb);
2056 error2 = sbuf_finish(&sb);
2059 return (error != 0 ? error : error2);
2063 * This sysctl allows a process to retrieve the path of the executable for
2064 * itself or another process.
2067 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2069 pid_t *pidp = (pid_t *)arg1;
2070 unsigned int arglen = arg2;
2073 char *retbuf, *freebuf;
2078 if (*pidp == -1) { /* -1 means this process */
2079 p = req->td->td_proc;
2081 error = pget(*pidp, PGET_CANSEE, &p);
2095 error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2099 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2100 free(freebuf, M_TEMP);
2105 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2118 error = pget((pid_t)name[0], PGET_CANSEE, &p);
2121 sv_name = p->p_sysent->sv_name;
2123 return (sysctl_handle_string(oidp, sv_name, 0, req));
2126 #ifdef KINFO_OVMENTRY_SIZE
2127 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2130 #ifdef COMPAT_FREEBSD7
2132 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2134 vm_map_entry_t entry, tmp_entry;
2135 unsigned int last_timestamp;
2136 char *fullpath, *freepath;
2137 struct kinfo_ovmentry *kve;
2147 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2150 vm = vmspace_acquire_ref(p);
2155 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2158 vm_map_lock_read(map);
2159 for (entry = map->header.next; entry != &map->header;
2160 entry = entry->next) {
2161 vm_object_t obj, tobj, lobj;
2164 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2167 bzero(kve, sizeof(*kve));
2168 kve->kve_structsize = sizeof(*kve);
2170 kve->kve_private_resident = 0;
2171 obj = entry->object.vm_object;
2173 VM_OBJECT_RLOCK(obj);
2174 if (obj->shadow_count == 1)
2175 kve->kve_private_resident =
2176 obj->resident_page_count;
2178 kve->kve_resident = 0;
2179 addr = entry->start;
2180 while (addr < entry->end) {
2181 if (pmap_extract(map->pmap, addr))
2182 kve->kve_resident++;
2186 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2188 VM_OBJECT_RLOCK(tobj);
2189 kve->kve_offset += tobj->backing_object_offset;
2192 VM_OBJECT_RUNLOCK(lobj);
2196 kve->kve_start = (void*)entry->start;
2197 kve->kve_end = (void*)entry->end;
2198 kve->kve_offset += (off_t)entry->offset;
2200 if (entry->protection & VM_PROT_READ)
2201 kve->kve_protection |= KVME_PROT_READ;
2202 if (entry->protection & VM_PROT_WRITE)
2203 kve->kve_protection |= KVME_PROT_WRITE;
2204 if (entry->protection & VM_PROT_EXECUTE)
2205 kve->kve_protection |= KVME_PROT_EXEC;
2207 if (entry->eflags & MAP_ENTRY_COW)
2208 kve->kve_flags |= KVME_FLAG_COW;
2209 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2210 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2211 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2212 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2214 last_timestamp = map->timestamp;
2215 vm_map_unlock_read(map);
2217 kve->kve_fileid = 0;
2223 switch (lobj->type) {
2225 kve->kve_type = KVME_TYPE_DEFAULT;
2228 kve->kve_type = KVME_TYPE_VNODE;
2233 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2234 kve->kve_type = KVME_TYPE_VNODE;
2235 if ((lobj->flags & OBJ_TMPFS) != 0) {
2236 vp = lobj->un_pager.swp.swp_tmpfs;
2240 kve->kve_type = KVME_TYPE_SWAP;
2244 kve->kve_type = KVME_TYPE_DEVICE;
2247 kve->kve_type = KVME_TYPE_PHYS;
2250 kve->kve_type = KVME_TYPE_DEAD;
2253 kve->kve_type = KVME_TYPE_SG;
2256 kve->kve_type = KVME_TYPE_UNKNOWN;
2260 VM_OBJECT_RUNLOCK(lobj);
2262 kve->kve_ref_count = obj->ref_count;
2263 kve->kve_shadow_count = obj->shadow_count;
2264 VM_OBJECT_RUNLOCK(obj);
2266 vn_fullpath(curthread, vp, &fullpath,
2268 cred = curthread->td_ucred;
2269 vn_lock(vp, LK_SHARED | LK_RETRY);
2270 if (VOP_GETATTR(vp, &va, cred) == 0) {
2271 kve->kve_fileid = va.va_fileid;
2273 kve->kve_fsid = va.va_fsid;
2278 kve->kve_type = KVME_TYPE_NONE;
2279 kve->kve_ref_count = 0;
2280 kve->kve_shadow_count = 0;
2283 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2284 if (freepath != NULL)
2285 free(freepath, M_TEMP);
2287 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2288 vm_map_lock_read(map);
2291 if (last_timestamp != map->timestamp) {
2292 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2296 vm_map_unlock_read(map);
2302 #endif /* COMPAT_FREEBSD7 */
2304 #ifdef KINFO_VMENTRY_SIZE
2305 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2309 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2310 int *resident_count, bool *super)
2312 vm_object_t obj, tobj;
2315 vm_paddr_t locked_pa;
2316 vm_pindex_t pi, pi_adv, pindex;
2319 *resident_count = 0;
2320 if (vmmap_skip_res_cnt)
2324 obj = entry->object.vm_object;
2325 addr = entry->start;
2327 pi = OFF_TO_IDX(entry->offset);
2328 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2329 if (m_adv != NULL) {
2332 pi_adv = atop(entry->end - addr);
2334 for (tobj = obj;; tobj = tobj->backing_object) {
2335 m = vm_page_find_least(tobj, pindex);
2337 if (m->pindex == pindex)
2339 if (pi_adv > m->pindex - pindex) {
2340 pi_adv = m->pindex - pindex;
2344 if (tobj->backing_object == NULL)
2346 pindex += OFF_TO_IDX(tobj->
2347 backing_object_offset);
2351 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2352 (addr & (pagesizes[1] - 1)) == 0 &&
2353 (pmap_mincore(map->pmap, addr, &locked_pa) &
2354 MINCORE_SUPER) != 0) {
2356 pi_adv = atop(pagesizes[1]);
2359 * We do not test the found page on validity.
2360 * Either the page is busy and being paged in,
2361 * or it was invalidated. The first case
2362 * should be counted as resident, the second
2363 * is not so clear; we do account both.
2367 *resident_count += pi_adv;
2370 PA_UNLOCK_COND(locked_pa);
2374 * Must be called with the process locked and will return unlocked.
2377 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2379 vm_map_entry_t entry, tmp_entry;
2382 vm_object_t obj, tobj, lobj;
2383 char *fullpath, *freepath;
2384 struct kinfo_vmentry *kve;
2389 unsigned int last_timestamp;
2393 PROC_LOCK_ASSERT(p, MA_OWNED);
2397 vm = vmspace_acquire_ref(p);
2402 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2406 vm_map_lock_read(map);
2407 for (entry = map->header.next; entry != &map->header;
2408 entry = entry->next) {
2409 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2413 bzero(kve, sizeof(*kve));
2414 obj = entry->object.vm_object;
2416 for (tobj = obj; tobj != NULL;
2417 tobj = tobj->backing_object) {
2418 VM_OBJECT_RLOCK(tobj);
2419 kve->kve_offset += tobj->backing_object_offset;
2422 if (obj->backing_object == NULL)
2423 kve->kve_private_resident =
2424 obj->resident_page_count;
2425 kern_proc_vmmap_resident(map, entry,
2426 &kve->kve_resident, &super);
2428 kve->kve_flags |= KVME_FLAG_SUPER;
2429 for (tobj = obj; tobj != NULL;
2430 tobj = tobj->backing_object) {
2431 if (tobj != obj && tobj != lobj)
2432 VM_OBJECT_RUNLOCK(tobj);
2438 kve->kve_start = entry->start;
2439 kve->kve_end = entry->end;
2440 kve->kve_offset += entry->offset;
2442 if (entry->protection & VM_PROT_READ)
2443 kve->kve_protection |= KVME_PROT_READ;
2444 if (entry->protection & VM_PROT_WRITE)
2445 kve->kve_protection |= KVME_PROT_WRITE;
2446 if (entry->protection & VM_PROT_EXECUTE)
2447 kve->kve_protection |= KVME_PROT_EXEC;
2449 if (entry->eflags & MAP_ENTRY_COW)
2450 kve->kve_flags |= KVME_FLAG_COW;
2451 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2452 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2453 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2454 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2455 if (entry->eflags & MAP_ENTRY_GROWS_UP)
2456 kve->kve_flags |= KVME_FLAG_GROWS_UP;
2457 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2458 kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2460 last_timestamp = map->timestamp;
2461 vm_map_unlock_read(map);
2467 switch (lobj->type) {
2469 kve->kve_type = KVME_TYPE_DEFAULT;
2472 kve->kve_type = KVME_TYPE_VNODE;
2477 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2478 kve->kve_type = KVME_TYPE_VNODE;
2479 if ((lobj->flags & OBJ_TMPFS) != 0) {
2480 vp = lobj->un_pager.swp.swp_tmpfs;
2484 kve->kve_type = KVME_TYPE_SWAP;
2488 kve->kve_type = KVME_TYPE_DEVICE;
2491 kve->kve_type = KVME_TYPE_PHYS;
2494 kve->kve_type = KVME_TYPE_DEAD;
2497 kve->kve_type = KVME_TYPE_SG;
2499 case OBJT_MGTDEVICE:
2500 kve->kve_type = KVME_TYPE_MGTDEVICE;
2503 kve->kve_type = KVME_TYPE_UNKNOWN;
2507 VM_OBJECT_RUNLOCK(lobj);
2509 kve->kve_ref_count = obj->ref_count;
2510 kve->kve_shadow_count = obj->shadow_count;
2511 VM_OBJECT_RUNLOCK(obj);
2513 vn_fullpath(curthread, vp, &fullpath,
2515 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2516 cred = curthread->td_ucred;
2517 vn_lock(vp, LK_SHARED | LK_RETRY);
2518 if (VOP_GETATTR(vp, &va, cred) == 0) {
2519 kve->kve_vn_fileid = va.va_fileid;
2520 kve->kve_vn_fsid = va.va_fsid;
2521 kve->kve_vn_fsid_freebsd11 =
2522 kve->kve_vn_fsid; /* truncate */
2524 MAKEIMODE(va.va_type, va.va_mode);
2525 kve->kve_vn_size = va.va_size;
2526 kve->kve_vn_rdev = va.va_rdev;
2527 kve->kve_vn_rdev_freebsd11 =
2528 kve->kve_vn_rdev; /* truncate */
2529 kve->kve_status = KF_ATTR_VALID;
2534 kve->kve_type = KVME_TYPE_NONE;
2535 kve->kve_ref_count = 0;
2536 kve->kve_shadow_count = 0;
2539 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2540 if (freepath != NULL)
2541 free(freepath, M_TEMP);
2543 /* Pack record size down */
2544 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2545 kve->kve_structsize =
2546 offsetof(struct kinfo_vmentry, kve_path) +
2547 strlen(kve->kve_path) + 1;
2549 kve->kve_structsize = sizeof(*kve);
2550 kve->kve_structsize = roundup(kve->kve_structsize,
2553 /* Halt filling and truncate rather than exceeding maxlen */
2554 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2556 vm_map_lock_read(map);
2558 } else if (maxlen != -1)
2559 maxlen -= kve->kve_structsize;
2561 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2563 vm_map_lock_read(map);
2566 if (last_timestamp != map->timestamp) {
2567 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2571 vm_map_unlock_read(map);
2579 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2583 int error, error2, *name;
2586 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2587 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2588 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2593 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2594 error2 = sbuf_finish(&sb);
2596 return (error != 0 ? error : error2);
2599 #if defined(STACK) || defined(DDB)
2601 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2603 struct kinfo_kstack *kkstp;
2604 int error, i, *name, numthreads;
2605 lwpid_t *lwpidarray;
2612 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2616 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2617 st = stack_create(M_WAITOK);
2622 if (lwpidarray != NULL) {
2623 free(lwpidarray, M_TEMP);
2626 numthreads = p->p_numthreads;
2628 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2631 } while (numthreads < p->p_numthreads);
2634 * XXXRW: During the below loop, execve(2) and countless other sorts
2635 * of changes could have taken place. Should we check to see if the
2636 * vmspace has been replaced, or the like, in order to prevent
2637 * giving a snapshot that spans, say, execve(2), with some threads
2638 * before and some after? Among other things, the credentials could
2639 * have changed, in which case the right to extract debug info might
2640 * no longer be assured.
2643 FOREACH_THREAD_IN_PROC(p, td) {
2644 KASSERT(i < numthreads,
2645 ("sysctl_kern_proc_kstack: numthreads"));
2646 lwpidarray[i] = td->td_tid;
2650 for (i = 0; i < numthreads; i++) {
2651 td = thread_find(p, lwpidarray[i]);
2655 bzero(kkstp, sizeof(*kkstp));
2656 (void)sbuf_new(&sb, kkstp->kkst_trace,
2657 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2659 kkstp->kkst_tid = td->td_tid;
2660 if (TD_IS_SWAPPED(td)) {
2661 kkstp->kkst_state = KKST_STATE_SWAPPED;
2662 } else if (TD_IS_RUNNING(td)) {
2663 if (stack_save_td_running(st, td) == 0)
2664 kkstp->kkst_state = KKST_STATE_STACKOK;
2666 kkstp->kkst_state = KKST_STATE_RUNNING;
2668 kkstp->kkst_state = KKST_STATE_STACKOK;
2669 stack_save_td(st, td);
2673 stack_sbuf_print(&sb, st);
2676 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2683 if (lwpidarray != NULL)
2684 free(lwpidarray, M_TEMP);
2686 free(kkstp, M_TEMP);
2692 * This sysctl allows a process to retrieve the full list of groups from
2693 * itself or another process.
2696 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2698 pid_t *pidp = (pid_t *)arg1;
2699 unsigned int arglen = arg2;
2706 if (*pidp == -1) { /* -1 means this process */
2707 p = req->td->td_proc;
2710 error = pget(*pidp, PGET_CANSEE, &p);
2715 cred = crhold(p->p_ucred);
2718 error = SYSCTL_OUT(req, cred->cr_groups,
2719 cred->cr_ngroups * sizeof(gid_t));
2725 * This sysctl allows a process to retrieve or/and set the resource limit for
2729 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2731 int *name = (int *)arg1;
2732 u_int namelen = arg2;
2741 which = (u_int)name[1];
2742 if (which >= RLIM_NLIMITS)
2745 if (req->newptr != NULL && req->newlen != sizeof(rlim))
2748 flags = PGET_HOLD | PGET_NOTWEXIT;
2749 if (req->newptr != NULL)
2750 flags |= PGET_CANDEBUG;
2752 flags |= PGET_CANSEE;
2753 error = pget((pid_t)name[0], flags, &p);
2760 if (req->oldptr != NULL) {
2762 lim_rlimit_proc(p, which, &rlim);
2765 error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2772 if (req->newptr != NULL) {
2773 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2775 error = kern_proc_setrlimit(curthread, p, which, &rlim);
2784 * This sysctl allows a process to retrieve ps_strings structure location of
2788 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2790 int *name = (int *)arg1;
2791 u_int namelen = arg2;
2793 vm_offset_t ps_strings;
2795 #ifdef COMPAT_FREEBSD32
2796 uint32_t ps_strings32;
2802 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2805 #ifdef COMPAT_FREEBSD32
2806 if ((req->flags & SCTL_MASK32) != 0) {
2808 * We return 0 if the 32 bit emulation request is for a 64 bit
2811 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2812 PTROUT(p->p_sysent->sv_psstrings) : 0;
2814 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2818 ps_strings = p->p_sysent->sv_psstrings;
2820 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2825 * This sysctl allows a process to retrieve umask of another process.
2828 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2830 int *name = (int *)arg1;
2831 u_int namelen = arg2;
2840 pid = (pid_t)name[0];
2842 if (pid == p->p_pid || pid == 0) {
2843 fd_cmask = p->p_fd->fd_cmask;
2847 error = pget(pid, PGET_WANTREAD, &p);
2851 fd_cmask = p->p_fd->fd_cmask;
2854 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2859 * This sysctl allows a process to set and retrieve binary osreldate of
2863 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2865 int *name = (int *)arg1;
2866 u_int namelen = arg2;
2868 int flags, error, osrel;
2873 if (req->newptr != NULL && req->newlen != sizeof(osrel))
2876 flags = PGET_HOLD | PGET_NOTWEXIT;
2877 if (req->newptr != NULL)
2878 flags |= PGET_CANDEBUG;
2880 flags |= PGET_CANSEE;
2881 error = pget((pid_t)name[0], flags, &p);
2885 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2889 if (req->newptr != NULL) {
2890 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2905 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2907 int *name = (int *)arg1;
2908 u_int namelen = arg2;
2910 struct kinfo_sigtramp kst;
2911 const struct sysentvec *sv;
2913 #ifdef COMPAT_FREEBSD32
2914 struct kinfo_sigtramp32 kst32;
2920 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2924 #ifdef COMPAT_FREEBSD32
2925 if ((req->flags & SCTL_MASK32) != 0) {
2926 bzero(&kst32, sizeof(kst32));
2927 if (SV_PROC_FLAG(p, SV_ILP32)) {
2928 if (sv->sv_sigcode_base != 0) {
2929 kst32.ksigtramp_start = sv->sv_sigcode_base;
2930 kst32.ksigtramp_end = sv->sv_sigcode_base +
2933 kst32.ksigtramp_start = sv->sv_psstrings -
2935 kst32.ksigtramp_end = sv->sv_psstrings;
2939 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2943 bzero(&kst, sizeof(kst));
2944 if (sv->sv_sigcode_base != 0) {
2945 kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2946 kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2949 kst.ksigtramp_start = (char *)sv->sv_psstrings -
2951 kst.ksigtramp_end = (char *)sv->sv_psstrings;
2954 error = SYSCTL_OUT(req, &kst, sizeof(kst));
2958 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table");
2960 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2961 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2962 "Return entire process table");
2964 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2965 sysctl_kern_proc, "Process table");
2967 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2968 sysctl_kern_proc, "Process table");
2970 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2971 sysctl_kern_proc, "Process table");
2973 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2974 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2976 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2977 sysctl_kern_proc, "Process table");
2979 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2980 sysctl_kern_proc, "Process table");
2982 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2983 sysctl_kern_proc, "Process table");
2985 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2986 sysctl_kern_proc, "Process table");
2988 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2989 sysctl_kern_proc, "Return process table, no threads");
2991 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2992 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2993 sysctl_kern_proc_args, "Process argument list");
2995 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2996 sysctl_kern_proc_env, "Process environment");
2998 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2999 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3001 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3002 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3004 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3005 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3006 "Process syscall vector name (ABI type)");
3008 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3009 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3011 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3012 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3014 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3015 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3017 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3018 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3020 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3021 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3023 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3024 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3026 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3027 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3029 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3030 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3032 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3033 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3034 "Return process table, no threads");
3036 #ifdef COMPAT_FREEBSD7
3037 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3038 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3041 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3042 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3044 #if defined(STACK) || defined(DDB)
3045 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3046 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3049 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3050 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3052 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3053 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3054 "Process resource limits");
3056 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3057 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3058 "Process ps_strings location");
3060 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3061 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3063 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3064 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3065 "Process binary osreldate");
3067 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3068 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3069 "Process signal trampoline location");
3074 * stop_all_proc() purpose is to stop all process which have usermode,
3075 * except current process for obvious reasons. This makes it somewhat
3076 * unreliable when invoked from multithreaded process. The service
3077 * must not be user-callable anyway.
3082 struct proc *cp, *p;
3084 bool restart, seen_stopped, seen_exiting, stopped_some;
3088 sx_xlock(&allproc_lock);
3090 seen_exiting = seen_stopped = stopped_some = restart = false;
3091 LIST_REMOVE(cp, p_list);
3092 LIST_INSERT_HEAD(&allproc, cp, p_list);
3094 p = LIST_NEXT(cp, p_list);
3097 LIST_REMOVE(cp, p_list);
3098 LIST_INSERT_AFTER(p, cp, p_list);
3100 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3104 if ((p->p_flag & P_WEXIT) != 0) {
3105 seen_exiting = true;
3109 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3111 * Stopped processes are tolerated when there
3112 * are no other processes which might continue
3113 * them. P_STOPPED_SINGLE but not
3114 * P_TOTAL_STOP process still has at least one
3117 seen_stopped = true;
3122 sx_xunlock(&allproc_lock);
3123 r = thread_single(p, SINGLE_ALLPROC);
3127 stopped_some = true;
3130 sx_xlock(&allproc_lock);
3132 /* Catch forked children we did not see in iteration. */
3133 if (gen != allproc_gen)
3135 sx_xunlock(&allproc_lock);
3136 if (restart || stopped_some || seen_exiting || seen_stopped) {
3137 kern_yield(PRI_USER);
3143 resume_all_proc(void)
3145 struct proc *cp, *p;
3148 sx_xlock(&allproc_lock);
3150 LIST_REMOVE(cp, p_list);
3151 LIST_INSERT_HEAD(&allproc, cp, p_list);
3153 p = LIST_NEXT(cp, p_list);
3156 LIST_REMOVE(cp, p_list);
3157 LIST_INSERT_AFTER(p, cp, p_list);
3159 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3160 sx_xunlock(&allproc_lock);
3162 thread_single_end(p, SINGLE_ALLPROC);
3165 sx_xlock(&allproc_lock);
3170 /* Did the loop above missed any stopped process ? */
3171 FOREACH_PROC_IN_SYSTEM(p) {
3172 /* No need for proc lock. */
3173 if ((p->p_flag & P_TOTAL_STOP) != 0)
3176 sx_xunlock(&allproc_lock);
3179 /* #define TOTAL_STOP_DEBUG 1 */
3180 #ifdef TOTAL_STOP_DEBUG
3181 volatile static int ap_resume;
3182 #include <sys/mount.h>
3185 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3191 error = sysctl_handle_int(oidp, &val, 0, req);
3192 if (error != 0 || req->newptr == NULL)
3197 while (ap_resume == 0)
3205 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3206 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3207 sysctl_debug_stop_all_proc, "I",