2 * Copyright (c) 1982, 1986, 1989, 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_compat.h"
37 #include "opt_ktrace.h"
38 #include "opt_kstack_pages.h"
39 #include "opt_stack.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
44 #include <sys/eventhandler.h>
47 #include <sys/kernel.h>
48 #include <sys/limits.h>
50 #include <sys/loginclass.h>
51 #include <sys/malloc.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
56 #include <sys/ptrace.h>
57 #include <sys/refcount.h>
58 #include <sys/resourcevar.h>
59 #include <sys/rwlock.h>
61 #include <sys/sysent.h>
62 #include <sys/sched.h>
64 #include <sys/stack.h>
66 #include <sys/sysctl.h>
67 #include <sys/filedesc.h>
69 #include <sys/signalvar.h>
73 #include <sys/vnode.h>
81 #include <vm/vm_param.h>
82 #include <vm/vm_extern.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
89 #ifdef COMPAT_FREEBSD32
90 #include <compat/freebsd32/freebsd32.h>
91 #include <compat/freebsd32/freebsd32_util.h>
94 SDT_PROVIDER_DEFINE(proc);
95 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
97 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
99 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
101 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
102 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
103 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
105 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
106 MALLOC_DEFINE(M_SESSION, "session", "session header");
107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
110 static void doenterpgrp(struct proc *, struct pgrp *);
111 static void orphanpg(struct pgrp *pg);
112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
116 static void pgadjustjobc(struct pgrp *pgrp, int entering);
117 static void pgdelete(struct pgrp *);
118 static int proc_ctor(void *mem, int size, void *arg, int flags);
119 static void proc_dtor(void *mem, int size, void *arg);
120 static int proc_init(void *mem, int size, int flags);
121 static void proc_fini(void *mem, int size);
122 static void pargs_free(struct pargs *pa);
123 static struct proc *zpfind_locked(pid_t pid);
126 * Other process lists
128 struct pidhashhead *pidhashtbl;
130 struct pgrphashhead *pgrphashtbl;
132 struct proclist allproc;
133 struct proclist zombproc;
134 struct sx allproc_lock;
135 struct sx proctree_lock;
136 struct mtx ppeers_lock;
137 uma_zone_t proc_zone;
140 * The offset of various fields in struct proc and struct thread.
141 * These are used by kernel debuggers to enumerate kernel threads and
144 const int proc_off_p_pid = offsetof(struct proc, p_pid);
145 const int proc_off_p_comm = offsetof(struct proc, p_comm);
146 const int proc_off_p_list = offsetof(struct proc, p_list);
147 const int proc_off_p_threads = offsetof(struct proc, p_threads);
148 const int thread_off_td_tid = offsetof(struct thread, td_tid);
149 const int thread_off_td_name = offsetof(struct thread, td_name);
150 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
151 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
152 const int thread_off_td_plist = offsetof(struct thread, td_plist);
154 int kstack_pages = KSTACK_PAGES;
155 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
156 "Kernel stack size in pages");
157 static int vmmap_skip_res_cnt = 0;
158 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
159 &vmmap_skip_res_cnt, 0,
160 "Skip calculation of the pages resident count in kern.proc.vmmap");
162 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
163 #ifdef COMPAT_FREEBSD32
164 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
168 * Initialize global process hashing structures.
174 sx_init(&allproc_lock, "allproc");
175 sx_init(&proctree_lock, "proctree");
176 mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
178 LIST_INIT(&zombproc);
179 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
180 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
181 proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
182 proc_ctor, proc_dtor, proc_init, proc_fini,
183 UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
188 * Prepare a proc for use.
191 proc_ctor(void *mem, int size, void *arg, int flags)
196 p = (struct proc *)mem;
197 SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
198 EVENTHANDLER_INVOKE(process_ctor, p);
199 SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
200 td = FIRST_THREAD_IN_PROC(p);
202 /* Make sure all thread constructors are executed */
203 EVENTHANDLER_INVOKE(thread_ctor, td);
209 * Reclaim a proc after use.
212 proc_dtor(void *mem, int size, void *arg)
217 /* INVARIANTS checks go here */
218 p = (struct proc *)mem;
219 td = FIRST_THREAD_IN_PROC(p);
220 SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
223 KASSERT((p->p_numthreads == 1),
224 ("bad number of threads in exiting process"));
225 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
227 /* Free all OSD associated to this thread. */
230 /* Make sure all thread destructors are executed */
231 EVENTHANDLER_INVOKE(thread_dtor, td);
233 EVENTHANDLER_INVOKE(process_dtor, p);
234 if (p->p_ksi != NULL)
235 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
236 SDT_PROBE3(proc, , dtor, return, p, size, arg);
240 * Initialize type-stable parts of a proc (when newly created).
243 proc_init(void *mem, int size, int flags)
247 p = (struct proc *)mem;
248 SDT_PROBE3(proc, , init, entry, p, size, flags);
249 mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
250 mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
251 mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
252 mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
253 mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
254 cv_init(&p->p_pwait, "ppwait");
255 cv_init(&p->p_dbgwait, "dbgwait");
256 TAILQ_INIT(&p->p_threads); /* all threads in proc */
257 EVENTHANDLER_INVOKE(process_init, p);
258 p->p_stats = pstats_alloc();
260 SDT_PROBE3(proc, , init, return, p, size, flags);
265 * UMA should ensure that this function is never called.
266 * Freeing a proc structure would violate type stability.
269 proc_fini(void *mem, int size)
274 p = (struct proc *)mem;
275 EVENTHANDLER_INVOKE(process_fini, p);
276 pstats_free(p->p_stats);
277 thread_free(FIRST_THREAD_IN_PROC(p));
278 mtx_destroy(&p->p_mtx);
279 if (p->p_ksi != NULL)
280 ksiginfo_free(p->p_ksi);
282 panic("proc reclaimed");
287 * Is p an inferior of the current process?
290 inferior(struct proc *p)
293 sx_assert(&proctree_lock, SX_LOCKED);
294 PROC_LOCK_ASSERT(p, MA_OWNED);
295 for (; p != curproc; p = proc_realparent(p)) {
303 pfind_locked(pid_t pid)
307 sx_assert(&allproc_lock, SX_LOCKED);
308 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
309 if (p->p_pid == pid) {
311 if (p->p_state == PRS_NEW) {
322 * Locate a process by number; return only "live" processes -- i.e., neither
323 * zombies nor newly born but incompletely initialized processes. By not
324 * returning processes in the PRS_NEW state, we allow callers to avoid
325 * testing for that condition to avoid dereferencing p_ucred, et al.
332 sx_slock(&allproc_lock);
333 p = pfind_locked(pid);
334 sx_sunlock(&allproc_lock);
339 pfind_tid_locked(pid_t tid)
344 sx_assert(&allproc_lock, SX_LOCKED);
345 FOREACH_PROC_IN_SYSTEM(p) {
347 if (p->p_state == PRS_NEW) {
351 FOREACH_THREAD_IN_PROC(p, td) {
352 if (td->td_tid == tid)
362 * Locate a process group by number.
363 * The caller must hold proctree_lock.
370 sx_assert(&proctree_lock, SX_LOCKED);
372 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
373 if (pgrp->pg_id == pgid) {
382 * Locate process and do additional manipulations, depending on flags.
385 pget(pid_t pid, int flags, struct proc **pp)
390 sx_slock(&allproc_lock);
391 if (pid <= PID_MAX) {
392 p = pfind_locked(pid);
393 if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
394 p = zpfind_locked(pid);
395 } else if ((flags & PGET_NOTID) == 0) {
396 p = pfind_tid_locked(pid);
400 sx_sunlock(&allproc_lock);
403 if ((flags & PGET_CANSEE) != 0) {
404 error = p_cansee(curthread, p);
408 if ((flags & PGET_CANDEBUG) != 0) {
409 error = p_candebug(curthread, p);
413 if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
417 if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
421 if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
423 * XXXRW: Not clear ESRCH is the right error during proc
429 if ((flags & PGET_HOLD) != 0) {
441 * Create a new process group.
442 * pgid must be equal to the pid of p.
443 * Begin a new session if required.
446 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
449 sx_assert(&proctree_lock, SX_XLOCKED);
451 KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
452 KASSERT(p->p_pid == pgid,
453 ("enterpgrp: new pgrp and pid != pgid"));
454 KASSERT(pgfind(pgid) == NULL,
455 ("enterpgrp: pgrp with pgid exists"));
456 KASSERT(!SESS_LEADER(p),
457 ("enterpgrp: session leader attempted setpgrp"));
459 mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
465 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
467 p->p_flag &= ~P_CONTROLT;
471 sess->s_sid = p->p_pid;
472 refcount_init(&sess->s_count, 1);
473 sess->s_ttyvp = NULL;
474 sess->s_ttydp = NULL;
476 bcopy(p->p_session->s_login, sess->s_login,
477 sizeof(sess->s_login));
478 pgrp->pg_session = sess;
479 KASSERT(p == curproc,
480 ("enterpgrp: mksession and p != curproc"));
482 pgrp->pg_session = p->p_session;
483 sess_hold(pgrp->pg_session);
487 LIST_INIT(&pgrp->pg_members);
490 * As we have an exclusive lock of proctree_lock,
491 * this should not deadlock.
493 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
495 SLIST_INIT(&pgrp->pg_sigiolst);
498 doenterpgrp(p, pgrp);
504 * Move p to an existing process group
507 enterthispgrp(struct proc *p, struct pgrp *pgrp)
510 sx_assert(&proctree_lock, SX_XLOCKED);
511 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
512 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
513 PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
514 SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
515 KASSERT(pgrp->pg_session == p->p_session,
516 ("%s: pgrp's session %p, p->p_session %p.\n",
520 KASSERT(pgrp != p->p_pgrp,
521 ("%s: p belongs to pgrp.", __func__));
523 doenterpgrp(p, pgrp);
529 * Move p to a process group
532 doenterpgrp(struct proc *p, struct pgrp *pgrp)
534 struct pgrp *savepgrp;
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);
542 savepgrp = p->p_pgrp;
545 * Adjust eligibility of affected pgrps to participate in job control.
546 * Increment eligibility counts before decrementing, otherwise we
547 * could reach 0 spuriously during the first call.
550 fixjobc(p, p->p_pgrp, 0);
555 LIST_REMOVE(p, p_pglist);
558 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
559 PGRP_UNLOCK(savepgrp);
561 if (LIST_EMPTY(&savepgrp->pg_members))
566 * remove process from process group
569 leavepgrp(struct proc *p)
571 struct pgrp *savepgrp;
573 sx_assert(&proctree_lock, SX_XLOCKED);
574 savepgrp = p->p_pgrp;
577 LIST_REMOVE(p, p_pglist);
580 PGRP_UNLOCK(savepgrp);
581 if (LIST_EMPTY(&savepgrp->pg_members))
587 * delete a process group
590 pgdelete(struct pgrp *pgrp)
592 struct session *savesess;
595 sx_assert(&proctree_lock, SX_XLOCKED);
596 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
597 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
600 * Reset any sigio structures pointing to us as a result of
601 * F_SETOWN with our pgid.
603 funsetownlst(&pgrp->pg_sigiolst);
606 tp = pgrp->pg_session->s_ttyp;
607 LIST_REMOVE(pgrp, pg_hash);
608 savesess = pgrp->pg_session;
611 /* Remove the reference to the pgrp before deallocating it. */
614 tty_rel_pgrp(tp, pgrp);
617 mtx_destroy(&pgrp->pg_mtx);
619 sess_release(savesess);
623 pgadjustjobc(struct pgrp *pgrp, int entering)
631 if (pgrp->pg_jobc == 0)
638 * Adjust pgrp jobc counters when specified process changes process group.
639 * We count the number of processes in each process group that "qualify"
640 * the group for terminal job control (those with a parent in a different
641 * process group of the same session). If that count reaches zero, the
642 * process group becomes orphaned. Check both the specified process'
643 * process group and that of its children.
644 * entering == 0 => p is leaving specified group.
645 * entering == 1 => p is entering specified group.
648 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
650 struct pgrp *hispgrp;
651 struct session *mysession;
654 sx_assert(&proctree_lock, SX_LOCKED);
655 PROC_LOCK_ASSERT(p, MA_NOTOWNED);
656 PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
657 SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
660 * Check p's parent to see whether p qualifies its own process
661 * group; if so, adjust count for p's process group.
663 mysession = pgrp->pg_session;
664 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
665 hispgrp->pg_session == mysession)
666 pgadjustjobc(pgrp, entering);
669 * Check this process' children to see whether they qualify
670 * their process groups; if so, adjust counts for children's
673 LIST_FOREACH(q, &p->p_children, p_sibling) {
675 if (hispgrp == pgrp ||
676 hispgrp->pg_session != mysession)
678 if (q->p_state == PRS_ZOMBIE)
680 pgadjustjobc(hispgrp, entering);
693 MPASS(p->p_flag & P_WEXIT);
695 * Do a quick check to see if there is anything to do with the
696 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
699 if (!SESS_LEADER(p) &&
700 (p->p_pgrp == p->p_pptr->p_pgrp) &&
701 LIST_EMPTY(&p->p_children)) {
707 sx_xlock(&proctree_lock);
708 if (SESS_LEADER(p)) {
712 * s_ttyp is not zero'd; we use this to indicate that
713 * the session once had a controlling terminal. (for
714 * logging and informational purposes)
725 * Signal foreground pgrp and revoke access to
726 * controlling terminal if it has not been revoked
729 * Because the TTY may have been revoked in the mean
730 * time and could already have a new session associated
731 * with it, make sure we don't send a SIGHUP to a
732 * foreground process group that does not belong to this
738 if (tp->t_session == sp)
739 tty_signal_pgrp(tp, SIGHUP);
744 sx_xunlock(&proctree_lock);
745 if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
746 VOP_REVOKE(ttyvp, REVOKEALL);
747 VOP_UNLOCK(ttyvp, 0);
750 sx_xlock(&proctree_lock);
753 fixjobc(p, p->p_pgrp, 0);
754 sx_xunlock(&proctree_lock);
758 * A process group has become orphaned;
759 * if there are any stopped processes in the group,
760 * hang-up all process in that group.
763 orphanpg(struct pgrp *pg)
767 PGRP_LOCK_ASSERT(pg, MA_OWNED);
769 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
771 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
773 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
775 kern_psignal(p, SIGHUP);
776 kern_psignal(p, SIGCONT);
786 sess_hold(struct session *s)
789 refcount_acquire(&s->s_count);
793 sess_release(struct session *s)
796 if (refcount_release(&s->s_count)) {
797 if (s->s_ttyp != NULL) {
799 tty_rel_sess(s->s_ttyp, s);
801 mtx_destroy(&s->s_mtx);
808 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
814 for (i = 0; i <= pgrphash; i++) {
815 if (!LIST_EMPTY(&pgrphashtbl[i])) {
816 printf("\tindx %d\n", i);
817 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
819 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
820 (void *)pgrp, (long)pgrp->pg_id,
821 (void *)pgrp->pg_session,
822 pgrp->pg_session->s_count,
823 (void *)LIST_FIRST(&pgrp->pg_members));
824 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
825 printf("\t\tpid %ld addr %p pgrp %p\n",
826 (long)p->p_pid, (void *)p,
836 * Calculate the kinfo_proc members which contain process-wide
838 * Must be called with the target process locked.
841 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
845 PROC_LOCK_ASSERT(p, MA_OWNED);
849 FOREACH_THREAD_IN_PROC(p, td) {
851 kp->ki_pctcpu += sched_pctcpu(td);
852 kp->ki_estcpu += sched_estcpu(td);
858 * Clear kinfo_proc and fill in any information that is common
859 * to all threads in the process.
860 * Must be called with the target process locked.
863 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
870 struct timeval boottime;
872 /* For proc_realparent. */
873 sx_assert(&proctree_lock, SX_LOCKED);
874 PROC_LOCK_ASSERT(p, MA_OWNED);
875 bzero(kp, sizeof(*kp));
877 kp->ki_structsize = sizeof(*kp);
879 kp->ki_addr =/* p->p_addr; */0; /* XXX */
880 kp->ki_args = p->p_args;
881 kp->ki_textvp = p->p_textvp;
883 kp->ki_tracep = p->p_tracevp;
884 kp->ki_traceflag = p->p_traceflag;
887 kp->ki_vmspace = p->p_vmspace;
888 kp->ki_flag = p->p_flag;
889 kp->ki_flag2 = p->p_flag2;
892 kp->ki_uid = cred->cr_uid;
893 kp->ki_ruid = cred->cr_ruid;
894 kp->ki_svuid = cred->cr_svuid;
896 if (cred->cr_flags & CRED_FLAG_CAPMODE)
897 kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
898 /* XXX bde doesn't like KI_NGROUPS */
899 if (cred->cr_ngroups > KI_NGROUPS) {
900 kp->ki_ngroups = KI_NGROUPS;
901 kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
903 kp->ki_ngroups = cred->cr_ngroups;
904 bcopy(cred->cr_groups, kp->ki_groups,
905 kp->ki_ngroups * sizeof(gid_t));
906 kp->ki_rgid = cred->cr_rgid;
907 kp->ki_svgid = cred->cr_svgid;
908 /* If jailed(cred), emulate the old P_JAILED flag. */
910 kp->ki_flag |= P_JAILED;
911 /* If inside the jail, use 0 as a jail ID. */
912 if (cred->cr_prison != curthread->td_ucred->cr_prison)
913 kp->ki_jid = cred->cr_prison->pr_id;
915 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
916 sizeof(kp->ki_loginclass));
920 mtx_lock(&ps->ps_mtx);
921 kp->ki_sigignore = ps->ps_sigignore;
922 kp->ki_sigcatch = ps->ps_sigcatch;
923 mtx_unlock(&ps->ps_mtx);
925 if (p->p_state != PRS_NEW &&
926 p->p_state != PRS_ZOMBIE &&
927 p->p_vmspace != NULL) {
928 struct vmspace *vm = p->p_vmspace;
930 kp->ki_size = vm->vm_map.size;
931 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
932 FOREACH_THREAD_IN_PROC(p, td0) {
933 if (!TD_IS_SWAPPED(td0))
934 kp->ki_rssize += td0->td_kstack_pages;
936 kp->ki_swrss = vm->vm_swrss;
937 kp->ki_tsize = vm->vm_tsize;
938 kp->ki_dsize = vm->vm_dsize;
939 kp->ki_ssize = vm->vm_ssize;
940 } else if (p->p_state == PRS_ZOMBIE)
942 if (kp->ki_flag & P_INMEM)
943 kp->ki_sflag = PS_INMEM;
946 /* Calculate legacy swtime as seconds since 'swtick'. */
947 kp->ki_swtime = (ticks - p->p_swtick) / hz;
948 kp->ki_pid = p->p_pid;
949 kp->ki_nice = p->p_nice;
950 kp->ki_fibnum = p->p_fibnum;
951 kp->ki_start = p->p_stats->p_start;
952 getboottime(&boottime);
953 timevaladd(&kp->ki_start, &boottime);
955 rufetch(p, &kp->ki_rusage);
956 kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
957 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
959 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
960 /* Some callers want child times in a single value. */
961 kp->ki_childtime = kp->ki_childstime;
962 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
964 FOREACH_THREAD_IN_PROC(p, td0)
965 kp->ki_cow += td0->td_cow;
969 kp->ki_pgid = p->p_pgrp->pg_id;
970 kp->ki_jobc = p->p_pgrp->pg_jobc;
971 sp = p->p_pgrp->pg_session;
974 kp->ki_sid = sp->s_sid;
976 strlcpy(kp->ki_login, sp->s_login,
977 sizeof(kp->ki_login));
979 kp->ki_kiflag |= KI_CTTY;
981 kp->ki_kiflag |= KI_SLEADER;
982 /* XXX proctree_lock */
987 if ((p->p_flag & P_CONTROLT) && tp != NULL) {
988 kp->ki_tdev = tty_udev(tp);
989 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
990 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
992 kp->ki_tsid = tp->t_session->s_sid;
995 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
997 if (p->p_comm[0] != '\0')
998 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
999 if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1000 p->p_sysent->sv_name[0] != '\0')
1001 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1002 kp->ki_siglist = p->p_siglist;
1003 kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1004 kp->ki_acflag = p->p_acflag;
1005 kp->ki_lock = p->p_lock;
1007 kp->ki_ppid = proc_realparent(p)->p_pid;
1008 if (p->p_flag & P_TRACED)
1009 kp->ki_tracer = p->p_pptr->p_pid;
1014 * Fill in information that is thread specific. Must be called with
1015 * target process locked. If 'preferthread' is set, overwrite certain
1016 * process-related fields that are maintained for both threads and
1020 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1026 PROC_LOCK_ASSERT(p, MA_OWNED);
1031 if (td->td_wmesg != NULL)
1032 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1034 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1035 if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1036 sizeof(kp->ki_tdname)) {
1037 strlcpy(kp->ki_moretdname,
1038 td->td_name + sizeof(kp->ki_tdname) - 1,
1039 sizeof(kp->ki_moretdname));
1041 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1043 if (TD_ON_LOCK(td)) {
1044 kp->ki_kiflag |= KI_LOCKBLOCK;
1045 strlcpy(kp->ki_lockname, td->td_lockname,
1046 sizeof(kp->ki_lockname));
1048 kp->ki_kiflag &= ~KI_LOCKBLOCK;
1049 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1052 if (p->p_state == PRS_NORMAL) { /* approximate. */
1053 if (TD_ON_RUNQ(td) ||
1055 TD_IS_RUNNING(td)) {
1057 } else if (P_SHOULDSTOP(p)) {
1058 kp->ki_stat = SSTOP;
1059 } else if (TD_IS_SLEEPING(td)) {
1060 kp->ki_stat = SSLEEP;
1061 } else if (TD_ON_LOCK(td)) {
1062 kp->ki_stat = SLOCK;
1064 kp->ki_stat = SWAIT;
1066 } else if (p->p_state == PRS_ZOMBIE) {
1067 kp->ki_stat = SZOMB;
1072 /* Things in the thread */
1073 kp->ki_wchan = td->td_wchan;
1074 kp->ki_pri.pri_level = td->td_priority;
1075 kp->ki_pri.pri_native = td->td_base_pri;
1078 * Note: legacy fields; clamp at the old NOCPU value and/or
1079 * the maximum u_char CPU value.
1081 if (td->td_lastcpu == NOCPU)
1082 kp->ki_lastcpu_old = NOCPU_OLD;
1083 else if (td->td_lastcpu > MAXCPU_OLD)
1084 kp->ki_lastcpu_old = MAXCPU_OLD;
1086 kp->ki_lastcpu_old = td->td_lastcpu;
1088 if (td->td_oncpu == NOCPU)
1089 kp->ki_oncpu_old = NOCPU_OLD;
1090 else if (td->td_oncpu > MAXCPU_OLD)
1091 kp->ki_oncpu_old = MAXCPU_OLD;
1093 kp->ki_oncpu_old = td->td_oncpu;
1095 kp->ki_lastcpu = td->td_lastcpu;
1096 kp->ki_oncpu = td->td_oncpu;
1097 kp->ki_tdflags = td->td_flags;
1098 kp->ki_tid = td->td_tid;
1099 kp->ki_numthreads = p->p_numthreads;
1100 kp->ki_pcb = td->td_pcb;
1101 kp->ki_kstack = (void *)td->td_kstack;
1102 kp->ki_slptime = (ticks - td->td_slptick) / hz;
1103 kp->ki_pri.pri_class = td->td_pri_class;
1104 kp->ki_pri.pri_user = td->td_user_pri;
1107 rufetchtd(td, &kp->ki_rusage);
1108 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1109 kp->ki_pctcpu = sched_pctcpu(td);
1110 kp->ki_estcpu = sched_estcpu(td);
1111 kp->ki_cow = td->td_cow;
1114 /* We can't get this anymore but ps etc never used it anyway. */
1118 kp->ki_siglist = td->td_siglist;
1119 kp->ki_sigmask = td->td_sigmask;
1126 * Fill in a kinfo_proc structure for the specified process.
1127 * Must be called with the target process locked.
1130 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1133 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1135 fill_kinfo_proc_only(p, kp);
1136 fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1137 fill_kinfo_aggregate(p, kp);
1144 return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1148 * Copy parts of p_stats; zero the rest of p_stats (statistics).
1151 pstats_fork(struct pstats *src, struct pstats *dst)
1154 bzero(&dst->pstat_startzero,
1155 __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1156 bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1157 __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1161 pstats_free(struct pstats *ps)
1164 free(ps, M_SUBPROC);
1167 static struct proc *
1168 zpfind_locked(pid_t pid)
1172 sx_assert(&allproc_lock, SX_LOCKED);
1173 LIST_FOREACH(p, &zombproc, p_list) {
1174 if (p->p_pid == pid) {
1183 * Locate a zombie process by number
1190 sx_slock(&allproc_lock);
1191 p = zpfind_locked(pid);
1192 sx_sunlock(&allproc_lock);
1196 #ifdef COMPAT_FREEBSD32
1199 * This function is typically used to copy out the kernel address, so
1200 * it can be replaced by assignment of zero.
1202 static inline uint32_t
1203 ptr32_trim(void *ptr)
1207 uptr = (uintptr_t)ptr;
1208 return ((uptr > UINT_MAX) ? 0 : uptr);
1211 #define PTRTRIM_CP(src,dst,fld) \
1212 do { (dst).fld = ptr32_trim((src).fld); } while (0)
1215 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1219 bzero(ki32, sizeof(struct kinfo_proc32));
1220 ki32->ki_structsize = sizeof(struct kinfo_proc32);
1221 CP(*ki, *ki32, ki_layout);
1222 PTRTRIM_CP(*ki, *ki32, ki_args);
1223 PTRTRIM_CP(*ki, *ki32, ki_paddr);
1224 PTRTRIM_CP(*ki, *ki32, ki_addr);
1225 PTRTRIM_CP(*ki, *ki32, ki_tracep);
1226 PTRTRIM_CP(*ki, *ki32, ki_textvp);
1227 PTRTRIM_CP(*ki, *ki32, ki_fd);
1228 PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1229 PTRTRIM_CP(*ki, *ki32, ki_wchan);
1230 CP(*ki, *ki32, ki_pid);
1231 CP(*ki, *ki32, ki_ppid);
1232 CP(*ki, *ki32, ki_pgid);
1233 CP(*ki, *ki32, ki_tpgid);
1234 CP(*ki, *ki32, ki_sid);
1235 CP(*ki, *ki32, ki_tsid);
1236 CP(*ki, *ki32, ki_jobc);
1237 CP(*ki, *ki32, ki_tdev);
1238 CP(*ki, *ki32, ki_tdev_freebsd11);
1239 CP(*ki, *ki32, ki_siglist);
1240 CP(*ki, *ki32, ki_sigmask);
1241 CP(*ki, *ki32, ki_sigignore);
1242 CP(*ki, *ki32, ki_sigcatch);
1243 CP(*ki, *ki32, ki_uid);
1244 CP(*ki, *ki32, ki_ruid);
1245 CP(*ki, *ki32, ki_svuid);
1246 CP(*ki, *ki32, ki_rgid);
1247 CP(*ki, *ki32, ki_svgid);
1248 CP(*ki, *ki32, ki_ngroups);
1249 for (i = 0; i < KI_NGROUPS; i++)
1250 CP(*ki, *ki32, ki_groups[i]);
1251 CP(*ki, *ki32, ki_size);
1252 CP(*ki, *ki32, ki_rssize);
1253 CP(*ki, *ki32, ki_swrss);
1254 CP(*ki, *ki32, ki_tsize);
1255 CP(*ki, *ki32, ki_dsize);
1256 CP(*ki, *ki32, ki_ssize);
1257 CP(*ki, *ki32, ki_xstat);
1258 CP(*ki, *ki32, ki_acflag);
1259 CP(*ki, *ki32, ki_pctcpu);
1260 CP(*ki, *ki32, ki_estcpu);
1261 CP(*ki, *ki32, ki_slptime);
1262 CP(*ki, *ki32, ki_swtime);
1263 CP(*ki, *ki32, ki_cow);
1264 CP(*ki, *ki32, ki_runtime);
1265 TV_CP(*ki, *ki32, ki_start);
1266 TV_CP(*ki, *ki32, ki_childtime);
1267 CP(*ki, *ki32, ki_flag);
1268 CP(*ki, *ki32, ki_kiflag);
1269 CP(*ki, *ki32, ki_traceflag);
1270 CP(*ki, *ki32, ki_stat);
1271 CP(*ki, *ki32, ki_nice);
1272 CP(*ki, *ki32, ki_lock);
1273 CP(*ki, *ki32, ki_rqindex);
1274 CP(*ki, *ki32, ki_oncpu);
1275 CP(*ki, *ki32, ki_lastcpu);
1277 /* XXX TODO: wrap cpu value as appropriate */
1278 CP(*ki, *ki32, ki_oncpu_old);
1279 CP(*ki, *ki32, ki_lastcpu_old);
1281 bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1282 bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1283 bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1284 bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1285 bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1286 bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1287 bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1288 bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1289 CP(*ki, *ki32, ki_tracer);
1290 CP(*ki, *ki32, ki_flag2);
1291 CP(*ki, *ki32, ki_fibnum);
1292 CP(*ki, *ki32, ki_cr_flags);
1293 CP(*ki, *ki32, ki_jid);
1294 CP(*ki, *ki32, ki_numthreads);
1295 CP(*ki, *ki32, ki_tid);
1296 CP(*ki, *ki32, ki_pri);
1297 freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1298 freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1299 PTRTRIM_CP(*ki, *ki32, ki_pcb);
1300 PTRTRIM_CP(*ki, *ki32, ki_kstack);
1301 PTRTRIM_CP(*ki, *ki32, ki_udata);
1302 CP(*ki, *ki32, ki_sflag);
1303 CP(*ki, *ki32, ki_tdflags);
1308 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1311 struct kinfo_proc ki;
1312 #ifdef COMPAT_FREEBSD32
1313 struct kinfo_proc32 ki32;
1317 PROC_LOCK_ASSERT(p, MA_OWNED);
1318 MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1321 fill_kinfo_proc(p, &ki);
1322 if ((flags & KERN_PROC_NOTHREADS) != 0) {
1323 #ifdef COMPAT_FREEBSD32
1324 if ((flags & KERN_PROC_MASK32) != 0) {
1325 freebsd32_kinfo_proc_out(&ki, &ki32);
1326 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1330 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1333 FOREACH_THREAD_IN_PROC(p, td) {
1334 fill_kinfo_thread(td, &ki, 1);
1335 #ifdef COMPAT_FREEBSD32
1336 if ((flags & KERN_PROC_MASK32) != 0) {
1337 freebsd32_kinfo_proc_out(&ki, &ki32);
1338 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1342 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1353 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1357 struct kinfo_proc ki;
1363 sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1364 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1365 error = kern_proc_out(p, &sb, flags);
1366 error2 = sbuf_finish(&sb);
1370 else if (error2 != 0)
1390 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1392 int *name = (int *)arg1;
1393 u_int namelen = arg2;
1395 int flags, doingzomb, oid_number;
1398 oid_number = oidp->oid_number;
1399 if (oid_number != KERN_PROC_ALL &&
1400 (oid_number & KERN_PROC_INC_THREAD) == 0)
1401 flags = KERN_PROC_NOTHREADS;
1404 oid_number &= ~KERN_PROC_INC_THREAD;
1406 #ifdef COMPAT_FREEBSD32
1407 if (req->flags & SCTL_MASK32)
1408 flags |= KERN_PROC_MASK32;
1410 if (oid_number == KERN_PROC_PID) {
1413 error = sysctl_wire_old_buffer(req, 0);
1416 sx_slock(&proctree_lock);
1417 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1419 error = sysctl_out_proc(p, req, flags, 0);
1420 sx_sunlock(&proctree_lock);
1424 switch (oid_number) {
1429 case KERN_PROC_PROC:
1430 if (namelen != 0 && namelen != 1)
1440 /* overestimate by 5 procs */
1441 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1445 error = sysctl_wire_old_buffer(req, 0);
1448 sx_slock(&proctree_lock);
1449 sx_slock(&allproc_lock);
1450 for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1452 p = LIST_FIRST(&allproc);
1454 p = LIST_FIRST(&zombproc);
1455 for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1457 * Skip embryonic processes.
1460 if (p->p_state == PRS_NEW) {
1464 KASSERT(p->p_ucred != NULL,
1465 ("process credential is NULL for non-NEW proc"));
1467 * Show a user only appropriate processes.
1469 if (p_cansee(curthread, p)) {
1474 * TODO - make more efficient (see notes below).
1477 switch (oid_number) {
1480 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1486 case KERN_PROC_PGRP:
1487 /* could do this by traversing pgrp */
1488 if (p->p_pgrp == NULL ||
1489 p->p_pgrp->pg_id != (pid_t)name[0]) {
1495 case KERN_PROC_RGID:
1496 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1502 case KERN_PROC_SESSION:
1503 if (p->p_session == NULL ||
1504 p->p_session->s_sid != (pid_t)name[0]) {
1511 if ((p->p_flag & P_CONTROLT) == 0 ||
1512 p->p_session == NULL) {
1516 /* XXX proctree_lock */
1517 SESS_LOCK(p->p_session);
1518 if (p->p_session->s_ttyp == NULL ||
1519 tty_udev(p->p_session->s_ttyp) !=
1521 SESS_UNLOCK(p->p_session);
1525 SESS_UNLOCK(p->p_session);
1529 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1535 case KERN_PROC_RUID:
1536 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1542 case KERN_PROC_PROC:
1550 error = sysctl_out_proc(p, req, flags, doingzomb);
1552 sx_sunlock(&allproc_lock);
1553 sx_sunlock(&proctree_lock);
1558 sx_sunlock(&allproc_lock);
1559 sx_sunlock(&proctree_lock);
1564 pargs_alloc(int len)
1568 pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1570 refcount_init(&pa->ar_ref, 1);
1571 pa->ar_length = len;
1576 pargs_free(struct pargs *pa)
1583 pargs_hold(struct pargs *pa)
1588 refcount_acquire(&pa->ar_ref);
1592 pargs_drop(struct pargs *pa)
1597 if (refcount_release(&pa->ar_ref))
1602 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1608 * This may return a short read if the string is shorter than the chunk
1609 * and is aligned at the end of the page, and the following page is not
1612 n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1618 #define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */
1620 enum proc_vector_type {
1626 #ifdef COMPAT_FREEBSD32
1628 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1629 size_t *vsizep, enum proc_vector_type type)
1631 struct freebsd32_ps_strings pss;
1633 vm_offset_t vptr, ptr;
1634 uint32_t *proc_vector32;
1640 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1641 sizeof(pss)) != sizeof(pss))
1645 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1646 vsize = pss.ps_nargvstr;
1647 if (vsize > ARG_MAX)
1649 size = vsize * sizeof(int32_t);
1652 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1653 vsize = pss.ps_nenvstr;
1654 if (vsize > ARG_MAX)
1656 size = vsize * sizeof(int32_t);
1659 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1660 (pss.ps_nenvstr + 1) * sizeof(int32_t);
1663 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1664 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1667 if (aux.a_type == AT_NULL)
1671 if (aux.a_type != AT_NULL)
1674 size = vsize * sizeof(aux);
1677 KASSERT(0, ("Wrong proc vector type: %d", type));
1680 proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1681 if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1685 if (type == PROC_AUX) {
1686 *proc_vectorp = (char **)proc_vector32;
1690 proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1691 for (i = 0; i < (int)vsize; i++)
1692 proc_vector[i] = PTRIN(proc_vector32[i]);
1693 *proc_vectorp = proc_vector;
1696 free(proc_vector32, M_TEMP);
1702 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1703 size_t *vsizep, enum proc_vector_type type)
1705 struct ps_strings pss;
1707 vm_offset_t vptr, ptr;
1712 #ifdef COMPAT_FREEBSD32
1713 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1714 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1716 if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1717 sizeof(pss)) != sizeof(pss))
1721 vptr = (vm_offset_t)pss.ps_argvstr;
1722 vsize = pss.ps_nargvstr;
1723 if (vsize > ARG_MAX)
1725 size = vsize * sizeof(char *);
1728 vptr = (vm_offset_t)pss.ps_envstr;
1729 vsize = pss.ps_nenvstr;
1730 if (vsize > ARG_MAX)
1732 size = vsize * sizeof(char *);
1736 * The aux array is just above env array on the stack. Check
1737 * that the address is naturally aligned.
1739 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1741 #if __ELF_WORD_SIZE == 64
1742 if (vptr % sizeof(uint64_t) != 0)
1744 if (vptr % sizeof(uint32_t) != 0)
1748 * We count the array size reading the aux vectors from the
1749 * stack until AT_NULL vector is returned. So (to keep the code
1750 * simple) we read the process stack twice: the first time here
1751 * to find the size and the second time when copying the vectors
1752 * to the allocated proc_vector.
1754 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1755 if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1758 if (aux.a_type == AT_NULL)
1763 * If the PROC_AUXV_MAX entries are iterated over, and we have
1764 * not reached AT_NULL, it is most likely we are reading wrong
1765 * data: either the process doesn't have auxv array or data has
1766 * been modified. Return the error in this case.
1768 if (aux.a_type != AT_NULL)
1771 size = vsize * sizeof(aux);
1774 KASSERT(0, ("Wrong proc vector type: %d", type));
1775 return (EINVAL); /* In case we are built without INVARIANTS. */
1777 proc_vector = malloc(size, M_TEMP, M_WAITOK);
1778 if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1779 free(proc_vector, M_TEMP);
1782 *proc_vectorp = proc_vector;
1788 #define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */
1791 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1792 enum proc_vector_type type)
1794 size_t done, len, nchr, vsize;
1796 char **proc_vector, *sptr;
1797 char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1799 PROC_ASSERT_HELD(p);
1802 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1804 nchr = 2 * (PATH_MAX + ARG_MAX);
1806 error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1809 for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1811 * The program may have scribbled into its argv array, e.g. to
1812 * remove some arguments. If that has happened, break out
1813 * before trying to read from NULL.
1815 if (proc_vector[i] == NULL)
1817 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1818 error = proc_read_string(td, p, sptr, pss_string,
1819 sizeof(pss_string));
1822 len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1823 if (done + len >= nchr)
1824 len = nchr - done - 1;
1825 sbuf_bcat(sb, pss_string, len);
1826 if (len != GET_PS_STRINGS_CHUNK_SZ)
1828 done += GET_PS_STRINGS_CHUNK_SZ;
1830 sbuf_bcat(sb, "", 1);
1834 free(proc_vector, M_TEMP);
1839 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1842 return (get_ps_strings(curthread, p, sb, PROC_ARG));
1846 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1849 return (get_ps_strings(curthread, p, sb, PROC_ENV));
1853 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1859 error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1861 #ifdef COMPAT_FREEBSD32
1862 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1863 size = vsize * sizeof(Elf32_Auxinfo);
1866 size = vsize * sizeof(Elf_Auxinfo);
1867 if (sbuf_bcat(sb, auxv, size) != 0)
1875 * This sysctl allows a process to retrieve the argument list or process
1876 * title for another process without groping around in the address space
1877 * of the other process. It also allow a process to set its own "process
1878 * title to a string of its own choice.
1881 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1883 int *name = (int *)arg1;
1884 u_int namelen = arg2;
1885 struct pargs *newpa, *pa;
1888 int flags, error = 0, error2;
1893 flags = PGET_CANSEE;
1894 if (req->newptr != NULL)
1895 flags |= PGET_ISCURRENT;
1896 error = pget((pid_t)name[0], flags, &p);
1904 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1906 } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1909 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1910 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1911 error = proc_getargv(curthread, p, &sb);
1912 error2 = sbuf_finish(&sb);
1915 if (error == 0 && error2 != 0)
1920 if (error != 0 || req->newptr == NULL)
1923 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1925 newpa = pargs_alloc(req->newlen);
1926 error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1940 * This sysctl allows a process to retrieve environment of another process.
1943 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1945 int *name = (int *)arg1;
1946 u_int namelen = arg2;
1954 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1957 if ((p->p_flag & P_SYSTEM) != 0) {
1962 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1963 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1964 error = proc_getenvv(curthread, p, &sb);
1965 error2 = sbuf_finish(&sb);
1968 return (error != 0 ? error : error2);
1972 * This sysctl allows a process to retrieve ELF auxiliary vector of
1976 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1978 int *name = (int *)arg1;
1979 u_int namelen = arg2;
1987 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1990 if ((p->p_flag & P_SYSTEM) != 0) {
1994 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1995 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1996 error = proc_getauxv(curthread, p, &sb);
1997 error2 = sbuf_finish(&sb);
2000 return (error != 0 ? error : error2);
2004 * This sysctl allows a process to retrieve the path of the executable for
2005 * itself or another process.
2008 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2010 pid_t *pidp = (pid_t *)arg1;
2011 unsigned int arglen = arg2;
2014 char *retbuf, *freebuf;
2019 if (*pidp == -1) { /* -1 means this process */
2020 p = req->td->td_proc;
2022 error = pget(*pidp, PGET_CANSEE, &p);
2036 error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2040 error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2041 free(freebuf, M_TEMP);
2046 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2059 error = pget((pid_t)name[0], PGET_CANSEE, &p);
2062 sv_name = p->p_sysent->sv_name;
2064 return (sysctl_handle_string(oidp, sv_name, 0, req));
2067 #ifdef KINFO_OVMENTRY_SIZE
2068 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2071 #ifdef COMPAT_FREEBSD7
2073 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2075 vm_map_entry_t entry, tmp_entry;
2076 unsigned int last_timestamp;
2077 char *fullpath, *freepath;
2078 struct kinfo_ovmentry *kve;
2088 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2091 vm = vmspace_acquire_ref(p);
2096 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2099 vm_map_lock_read(map);
2100 for (entry = map->header.next; entry != &map->header;
2101 entry = entry->next) {
2102 vm_object_t obj, tobj, lobj;
2105 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2108 bzero(kve, sizeof(*kve));
2109 kve->kve_structsize = sizeof(*kve);
2111 kve->kve_private_resident = 0;
2112 obj = entry->object.vm_object;
2114 VM_OBJECT_RLOCK(obj);
2115 if (obj->shadow_count == 1)
2116 kve->kve_private_resident =
2117 obj->resident_page_count;
2119 kve->kve_resident = 0;
2120 addr = entry->start;
2121 while (addr < entry->end) {
2122 if (pmap_extract(map->pmap, addr))
2123 kve->kve_resident++;
2127 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2129 VM_OBJECT_RLOCK(tobj);
2131 VM_OBJECT_RUNLOCK(lobj);
2135 kve->kve_start = (void*)entry->start;
2136 kve->kve_end = (void*)entry->end;
2137 kve->kve_offset = (off_t)entry->offset;
2139 if (entry->protection & VM_PROT_READ)
2140 kve->kve_protection |= KVME_PROT_READ;
2141 if (entry->protection & VM_PROT_WRITE)
2142 kve->kve_protection |= KVME_PROT_WRITE;
2143 if (entry->protection & VM_PROT_EXECUTE)
2144 kve->kve_protection |= KVME_PROT_EXEC;
2146 if (entry->eflags & MAP_ENTRY_COW)
2147 kve->kve_flags |= KVME_FLAG_COW;
2148 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2149 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2150 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2151 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2153 last_timestamp = map->timestamp;
2154 vm_map_unlock_read(map);
2156 kve->kve_fileid = 0;
2162 switch (lobj->type) {
2164 kve->kve_type = KVME_TYPE_DEFAULT;
2167 kve->kve_type = KVME_TYPE_VNODE;
2172 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2173 kve->kve_type = KVME_TYPE_VNODE;
2174 if ((lobj->flags & OBJ_TMPFS) != 0) {
2175 vp = lobj->un_pager.swp.swp_tmpfs;
2179 kve->kve_type = KVME_TYPE_SWAP;
2183 kve->kve_type = KVME_TYPE_DEVICE;
2186 kve->kve_type = KVME_TYPE_PHYS;
2189 kve->kve_type = KVME_TYPE_DEAD;
2192 kve->kve_type = KVME_TYPE_SG;
2195 kve->kve_type = KVME_TYPE_UNKNOWN;
2199 VM_OBJECT_RUNLOCK(lobj);
2201 kve->kve_ref_count = obj->ref_count;
2202 kve->kve_shadow_count = obj->shadow_count;
2203 VM_OBJECT_RUNLOCK(obj);
2205 vn_fullpath(curthread, vp, &fullpath,
2207 cred = curthread->td_ucred;
2208 vn_lock(vp, LK_SHARED | LK_RETRY);
2209 if (VOP_GETATTR(vp, &va, cred) == 0) {
2210 kve->kve_fileid = va.va_fileid;
2212 kve->kve_fsid = va.va_fsid;
2217 kve->kve_type = KVME_TYPE_NONE;
2218 kve->kve_ref_count = 0;
2219 kve->kve_shadow_count = 0;
2222 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2223 if (freepath != NULL)
2224 free(freepath, M_TEMP);
2226 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2227 vm_map_lock_read(map);
2230 if (last_timestamp != map->timestamp) {
2231 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2235 vm_map_unlock_read(map);
2241 #endif /* COMPAT_FREEBSD7 */
2243 #ifdef KINFO_VMENTRY_SIZE
2244 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2248 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2249 struct kinfo_vmentry *kve)
2251 vm_object_t obj, tobj;
2254 vm_paddr_t locked_pa;
2255 vm_pindex_t pi, pi_adv, pindex;
2258 obj = entry->object.vm_object;
2259 addr = entry->start;
2261 pi = OFF_TO_IDX(entry->offset);
2262 for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2263 if (m_adv != NULL) {
2266 pi_adv = atop(entry->end - addr);
2268 for (tobj = obj;; tobj = tobj->backing_object) {
2269 m = vm_page_find_least(tobj, pindex);
2271 if (m->pindex == pindex)
2273 if (pi_adv > m->pindex - pindex) {
2274 pi_adv = m->pindex - pindex;
2278 if (tobj->backing_object == NULL)
2280 pindex += OFF_TO_IDX(tobj->
2281 backing_object_offset);
2285 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2286 (addr & (pagesizes[1] - 1)) == 0 &&
2287 (pmap_mincore(map->pmap, addr, &locked_pa) &
2288 MINCORE_SUPER) != 0) {
2289 kve->kve_flags |= KVME_FLAG_SUPER;
2290 pi_adv = atop(pagesizes[1]);
2293 * We do not test the found page on validity.
2294 * Either the page is busy and being paged in,
2295 * or it was invalidated. The first case
2296 * should be counted as resident, the second
2297 * is not so clear; we do account both.
2301 kve->kve_resident += pi_adv;
2304 PA_UNLOCK_COND(locked_pa);
2308 * Must be called with the process locked and will return unlocked.
2311 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2313 vm_map_entry_t entry, tmp_entry;
2316 vm_object_t obj, tobj, lobj;
2317 char *fullpath, *freepath;
2318 struct kinfo_vmentry *kve;
2323 unsigned int last_timestamp;
2326 PROC_LOCK_ASSERT(p, MA_OWNED);
2330 vm = vmspace_acquire_ref(p);
2335 kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2339 vm_map_lock_read(map);
2340 for (entry = map->header.next; entry != &map->header;
2341 entry = entry->next) {
2342 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2346 bzero(kve, sizeof(*kve));
2347 obj = entry->object.vm_object;
2349 for (tobj = obj; tobj != NULL;
2350 tobj = tobj->backing_object) {
2351 VM_OBJECT_RLOCK(tobj);
2354 if (obj->backing_object == NULL)
2355 kve->kve_private_resident =
2356 obj->resident_page_count;
2357 if (!vmmap_skip_res_cnt)
2358 kern_proc_vmmap_resident(map, entry, kve);
2359 for (tobj = obj; tobj != NULL;
2360 tobj = tobj->backing_object) {
2361 if (tobj != obj && tobj != lobj)
2362 VM_OBJECT_RUNLOCK(tobj);
2368 kve->kve_start = entry->start;
2369 kve->kve_end = entry->end;
2370 kve->kve_offset = entry->offset;
2372 if (entry->protection & VM_PROT_READ)
2373 kve->kve_protection |= KVME_PROT_READ;
2374 if (entry->protection & VM_PROT_WRITE)
2375 kve->kve_protection |= KVME_PROT_WRITE;
2376 if (entry->protection & VM_PROT_EXECUTE)
2377 kve->kve_protection |= KVME_PROT_EXEC;
2379 if (entry->eflags & MAP_ENTRY_COW)
2380 kve->kve_flags |= KVME_FLAG_COW;
2381 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2382 kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2383 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2384 kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2385 if (entry->eflags & MAP_ENTRY_GROWS_UP)
2386 kve->kve_flags |= KVME_FLAG_GROWS_UP;
2387 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2388 kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2390 last_timestamp = map->timestamp;
2391 vm_map_unlock_read(map);
2397 switch (lobj->type) {
2399 kve->kve_type = KVME_TYPE_DEFAULT;
2402 kve->kve_type = KVME_TYPE_VNODE;
2407 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2408 kve->kve_type = KVME_TYPE_VNODE;
2409 if ((lobj->flags & OBJ_TMPFS) != 0) {
2410 vp = lobj->un_pager.swp.swp_tmpfs;
2414 kve->kve_type = KVME_TYPE_SWAP;
2418 kve->kve_type = KVME_TYPE_DEVICE;
2421 kve->kve_type = KVME_TYPE_PHYS;
2424 kve->kve_type = KVME_TYPE_DEAD;
2427 kve->kve_type = KVME_TYPE_SG;
2429 case OBJT_MGTDEVICE:
2430 kve->kve_type = KVME_TYPE_MGTDEVICE;
2433 kve->kve_type = KVME_TYPE_UNKNOWN;
2437 VM_OBJECT_RUNLOCK(lobj);
2439 kve->kve_ref_count = obj->ref_count;
2440 kve->kve_shadow_count = obj->shadow_count;
2441 VM_OBJECT_RUNLOCK(obj);
2443 vn_fullpath(curthread, vp, &fullpath,
2445 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2446 cred = curthread->td_ucred;
2447 vn_lock(vp, LK_SHARED | LK_RETRY);
2448 if (VOP_GETATTR(vp, &va, cred) == 0) {
2449 kve->kve_vn_fileid = va.va_fileid;
2450 kve->kve_vn_fsid = va.va_fsid;
2451 kve->kve_vn_fsid_freebsd11 =
2452 kve->kve_vn_fsid; /* truncate */
2454 MAKEIMODE(va.va_type, va.va_mode);
2455 kve->kve_vn_size = va.va_size;
2456 kve->kve_vn_rdev = va.va_rdev;
2457 kve->kve_vn_rdev_freebsd11 =
2458 kve->kve_vn_rdev; /* truncate */
2459 kve->kve_status = KF_ATTR_VALID;
2464 kve->kve_type = KVME_TYPE_NONE;
2465 kve->kve_ref_count = 0;
2466 kve->kve_shadow_count = 0;
2469 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2470 if (freepath != NULL)
2471 free(freepath, M_TEMP);
2473 /* Pack record size down */
2474 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2475 kve->kve_structsize =
2476 offsetof(struct kinfo_vmentry, kve_path) +
2477 strlen(kve->kve_path) + 1;
2479 kve->kve_structsize = sizeof(*kve);
2480 kve->kve_structsize = roundup(kve->kve_structsize,
2483 /* Halt filling and truncate rather than exceeding maxlen */
2484 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2486 vm_map_lock_read(map);
2488 } else if (maxlen != -1)
2489 maxlen -= kve->kve_structsize;
2491 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2493 vm_map_lock_read(map);
2496 if (last_timestamp != map->timestamp) {
2497 vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2501 vm_map_unlock_read(map);
2509 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2513 int error, error2, *name;
2516 sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2517 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2518 error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2523 error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2524 error2 = sbuf_finish(&sb);
2526 return (error != 0 ? error : error2);
2529 #if defined(STACK) || defined(DDB)
2531 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2533 struct kinfo_kstack *kkstp;
2534 int error, i, *name, numthreads;
2535 lwpid_t *lwpidarray;
2542 error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2546 kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2547 st = stack_create();
2552 if (lwpidarray != NULL) {
2553 free(lwpidarray, M_TEMP);
2556 numthreads = p->p_numthreads;
2558 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2561 } while (numthreads < p->p_numthreads);
2564 * XXXRW: During the below loop, execve(2) and countless other sorts
2565 * of changes could have taken place. Should we check to see if the
2566 * vmspace has been replaced, or the like, in order to prevent
2567 * giving a snapshot that spans, say, execve(2), with some threads
2568 * before and some after? Among other things, the credentials could
2569 * have changed, in which case the right to extract debug info might
2570 * no longer be assured.
2573 FOREACH_THREAD_IN_PROC(p, td) {
2574 KASSERT(i < numthreads,
2575 ("sysctl_kern_proc_kstack: numthreads"));
2576 lwpidarray[i] = td->td_tid;
2580 for (i = 0; i < numthreads; i++) {
2581 td = thread_find(p, lwpidarray[i]);
2585 bzero(kkstp, sizeof(*kkstp));
2586 (void)sbuf_new(&sb, kkstp->kkst_trace,
2587 sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2589 kkstp->kkst_tid = td->td_tid;
2590 if (TD_IS_SWAPPED(td)) {
2591 kkstp->kkst_state = KKST_STATE_SWAPPED;
2592 } else if (TD_IS_RUNNING(td)) {
2593 if (stack_save_td_running(st, td) == 0)
2594 kkstp->kkst_state = KKST_STATE_STACKOK;
2596 kkstp->kkst_state = KKST_STATE_RUNNING;
2598 kkstp->kkst_state = KKST_STATE_STACKOK;
2599 stack_save_td(st, td);
2603 stack_sbuf_print(&sb, st);
2606 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2613 if (lwpidarray != NULL)
2614 free(lwpidarray, M_TEMP);
2616 free(kkstp, M_TEMP);
2622 * This sysctl allows a process to retrieve the full list of groups from
2623 * itself or another process.
2626 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2628 pid_t *pidp = (pid_t *)arg1;
2629 unsigned int arglen = arg2;
2636 if (*pidp == -1) { /* -1 means this process */
2637 p = req->td->td_proc;
2640 error = pget(*pidp, PGET_CANSEE, &p);
2645 cred = crhold(p->p_ucred);
2648 error = SYSCTL_OUT(req, cred->cr_groups,
2649 cred->cr_ngroups * sizeof(gid_t));
2655 * This sysctl allows a process to retrieve or/and set the resource limit for
2659 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2661 int *name = (int *)arg1;
2662 u_int namelen = arg2;
2671 which = (u_int)name[1];
2672 if (which >= RLIM_NLIMITS)
2675 if (req->newptr != NULL && req->newlen != sizeof(rlim))
2678 flags = PGET_HOLD | PGET_NOTWEXIT;
2679 if (req->newptr != NULL)
2680 flags |= PGET_CANDEBUG;
2682 flags |= PGET_CANSEE;
2683 error = pget((pid_t)name[0], flags, &p);
2690 if (req->oldptr != NULL) {
2692 lim_rlimit_proc(p, which, &rlim);
2695 error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2702 if (req->newptr != NULL) {
2703 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2705 error = kern_proc_setrlimit(curthread, p, which, &rlim);
2714 * This sysctl allows a process to retrieve ps_strings structure location of
2718 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2720 int *name = (int *)arg1;
2721 u_int namelen = arg2;
2723 vm_offset_t ps_strings;
2725 #ifdef COMPAT_FREEBSD32
2726 uint32_t ps_strings32;
2732 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2735 #ifdef COMPAT_FREEBSD32
2736 if ((req->flags & SCTL_MASK32) != 0) {
2738 * We return 0 if the 32 bit emulation request is for a 64 bit
2741 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2742 PTROUT(p->p_sysent->sv_psstrings) : 0;
2744 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2748 ps_strings = p->p_sysent->sv_psstrings;
2750 error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2755 * This sysctl allows a process to retrieve umask of another process.
2758 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2760 int *name = (int *)arg1;
2761 u_int namelen = arg2;
2769 error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2773 FILEDESC_SLOCK(p->p_fd);
2774 fd_cmask = p->p_fd->fd_cmask;
2775 FILEDESC_SUNLOCK(p->p_fd);
2777 error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2782 * This sysctl allows a process to set and retrieve binary osreldate of
2786 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2788 int *name = (int *)arg1;
2789 u_int namelen = arg2;
2791 int flags, error, osrel;
2796 if (req->newptr != NULL && req->newlen != sizeof(osrel))
2799 flags = PGET_HOLD | PGET_NOTWEXIT;
2800 if (req->newptr != NULL)
2801 flags |= PGET_CANDEBUG;
2803 flags |= PGET_CANSEE;
2804 error = pget((pid_t)name[0], flags, &p);
2808 error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2812 if (req->newptr != NULL) {
2813 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2828 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2830 int *name = (int *)arg1;
2831 u_int namelen = arg2;
2833 struct kinfo_sigtramp kst;
2834 const struct sysentvec *sv;
2836 #ifdef COMPAT_FREEBSD32
2837 struct kinfo_sigtramp32 kst32;
2843 error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2847 #ifdef COMPAT_FREEBSD32
2848 if ((req->flags & SCTL_MASK32) != 0) {
2849 bzero(&kst32, sizeof(kst32));
2850 if (SV_PROC_FLAG(p, SV_ILP32)) {
2851 if (sv->sv_sigcode_base != 0) {
2852 kst32.ksigtramp_start = sv->sv_sigcode_base;
2853 kst32.ksigtramp_end = sv->sv_sigcode_base +
2856 kst32.ksigtramp_start = sv->sv_psstrings -
2858 kst32.ksigtramp_end = sv->sv_psstrings;
2862 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2866 bzero(&kst, sizeof(kst));
2867 if (sv->sv_sigcode_base != 0) {
2868 kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2869 kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2872 kst.ksigtramp_start = (char *)sv->sv_psstrings -
2874 kst.ksigtramp_end = (char *)sv->sv_psstrings;
2877 error = SYSCTL_OUT(req, &kst, sizeof(kst));
2881 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table");
2883 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2884 CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2885 "Return entire process table");
2887 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2888 sysctl_kern_proc, "Process table");
2890 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2891 sysctl_kern_proc, "Process table");
2893 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2894 sysctl_kern_proc, "Process table");
2896 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2897 CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2899 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2900 sysctl_kern_proc, "Process table");
2902 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2903 sysctl_kern_proc, "Process table");
2905 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2906 sysctl_kern_proc, "Process table");
2908 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2909 sysctl_kern_proc, "Process table");
2911 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2912 sysctl_kern_proc, "Return process table, no threads");
2914 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2915 CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2916 sysctl_kern_proc_args, "Process argument list");
2918 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2919 sysctl_kern_proc_env, "Process environment");
2921 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2922 CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2924 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2925 CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2927 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2928 CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2929 "Process syscall vector name (ABI type)");
2931 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2932 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2934 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2935 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2937 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2938 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2940 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2941 sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2943 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2944 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2946 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2947 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2949 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2950 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2952 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2953 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2955 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2956 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2957 "Return process table, no threads");
2959 #ifdef COMPAT_FREEBSD7
2960 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2961 CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2964 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2965 CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2967 #if defined(STACK) || defined(DDB)
2968 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2969 CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2972 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2973 CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2975 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2976 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2977 "Process resource limits");
2979 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2980 CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2981 "Process ps_strings location");
2983 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2984 CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2986 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2987 CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2988 "Process binary osreldate");
2990 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2991 CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2992 "Process signal trampoline location");
2997 * stop_all_proc() purpose is to stop all process which have usermode,
2998 * except current process for obvious reasons. This makes it somewhat
2999 * unreliable when invoked from multithreaded process. The service
3000 * must not be user-callable anyway.
3005 struct proc *cp, *p;
3007 bool restart, seen_stopped, seen_exiting, stopped_some;
3011 sx_xlock(&allproc_lock);
3013 seen_exiting = seen_stopped = stopped_some = restart = false;
3014 LIST_REMOVE(cp, p_list);
3015 LIST_INSERT_HEAD(&allproc, cp, p_list);
3017 p = LIST_NEXT(cp, p_list);
3020 LIST_REMOVE(cp, p_list);
3021 LIST_INSERT_AFTER(p, cp, p_list);
3023 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3027 if ((p->p_flag & P_WEXIT) != 0) {
3028 seen_exiting = true;
3032 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3034 * Stopped processes are tolerated when there
3035 * are no other processes which might continue
3036 * them. P_STOPPED_SINGLE but not
3037 * P_TOTAL_STOP process still has at least one
3040 seen_stopped = true;
3045 sx_xunlock(&allproc_lock);
3046 r = thread_single(p, SINGLE_ALLPROC);
3050 stopped_some = true;
3053 sx_xlock(&allproc_lock);
3055 /* Catch forked children we did not see in iteration. */
3056 if (gen != allproc_gen)
3058 sx_xunlock(&allproc_lock);
3059 if (restart || stopped_some || seen_exiting || seen_stopped) {
3060 kern_yield(PRI_USER);
3066 resume_all_proc(void)
3068 struct proc *cp, *p;
3071 sx_xlock(&allproc_lock);
3072 LIST_REMOVE(cp, p_list);
3073 LIST_INSERT_HEAD(&allproc, cp, p_list);
3075 p = LIST_NEXT(cp, p_list);
3078 LIST_REMOVE(cp, p_list);
3079 LIST_INSERT_AFTER(p, cp, p_list);
3081 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3082 sx_xunlock(&allproc_lock);
3084 thread_single_end(p, SINGLE_ALLPROC);
3087 sx_xlock(&allproc_lock);
3092 sx_xunlock(&allproc_lock);
3095 /* #define TOTAL_STOP_DEBUG 1 */
3096 #ifdef TOTAL_STOP_DEBUG
3097 volatile static int ap_resume;
3098 #include <sys/mount.h>
3101 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3107 error = sysctl_handle_int(oidp, &val, 0, req);
3108 if (error != 0 || req->newptr == NULL)
3113 while (ap_resume == 0)
3121 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3122 CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3123 sysctl_debug_stop_all_proc, "I",