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. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
34 * $FreeBSD: src/sys/kern/kern_proc.c,v 1.63.2.9 2003/05/08 07:47:16 kbyanc Exp $
35 * $DragonFly: src/sys/kern/kern_proc.c,v 1.45 2008/06/12 23:25:02 dillon Exp $
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/sysctl.h>
42 #include <sys/malloc.h>
45 #include <sys/filedesc.h>
47 #include <sys/signalvar.h>
48 #include <sys/spinlock.h>
52 #include <vm/vm_map.h>
54 #include <machine/smp.h>
56 #include <sys/spinlock2.h>
58 static MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
59 MALLOC_DEFINE(M_SESSION, "session", "session header");
60 MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
61 MALLOC_DEFINE(M_LWP, "lwp", "lwp structures");
62 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
64 int ps_showallprocs = 1;
65 static int ps_showallthreads = 1;
66 SYSCTL_INT(_security, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
68 "Unprivileged processes can see proccesses with different UID/GID");
69 SYSCTL_INT(_security, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
70 &ps_showallthreads, 0,
71 "Unprivileged processes can see kernel threads");
73 static void pgdelete(struct pgrp *);
74 static void orphanpg(struct pgrp *pg);
75 static pid_t proc_getnewpid_locked(int random_offset);
80 struct pidhashhead *pidhashtbl;
82 struct pgrphashhead *pgrphashtbl;
84 struct proclist allproc;
85 struct proclist zombproc;
86 struct spinlock allproc_spin;
89 * Random component to nextpid generation. We mix in a random factor to make
90 * it a little harder to predict. We sanity check the modulus value to avoid
91 * doing it in critical paths. Don't let it be too small or we pointlessly
92 * waste randomness entropy, and don't let it be impossibly large. Using a
93 * modulus that is too big causes a LOT more process table scans and slows
94 * down fork processing as the pidchecked caching is defeated.
96 static int randompid = 0;
99 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
104 error = sysctl_handle_int(oidp, &pid, 0, req);
105 if (error || !req->newptr)
107 if (pid < 0 || pid > PID_MAX - 100) /* out of range */
109 else if (pid < 2) /* NOP */
111 else if (pid < 100) /* Make it reasonable */
117 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
118 0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
121 * Initialize global process hashing structures.
127 LIST_INIT(&zombproc);
128 spin_init(&allproc_spin);
130 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
131 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
136 * Is p an inferior of the current process?
139 inferior(struct proc *p)
141 for (; p != curproc; p = p->p_pptr)
148 * Locate a process by number
155 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
163 * Locate a process group by number
170 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
171 if (pgrp->pg_id == pgid)
178 * Move p to a new or existing process group (and session)
181 enterpgrp(struct proc *p, pid_t pgid, int mksess)
183 struct pgrp *pgrp = pgfind(pgid);
185 KASSERT(pgrp == NULL || !mksess,
186 ("enterpgrp: setsid into non-empty pgrp"));
187 KASSERT(!SESS_LEADER(p),
188 ("enterpgrp: session leader attempted setpgrp"));
191 pid_t savepid = p->p_pid;
196 KASSERT(p->p_pid == pgid,
197 ("enterpgrp: new pgrp and pid != pgid"));
198 if ((np = pfind(savepid)) == NULL || np != p)
200 MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP,
203 struct session *sess;
208 MALLOC(sess, struct session *, sizeof(struct session),
209 M_SESSION, M_WAITOK);
211 sess->s_sid = p->p_pid;
213 sess->s_ttyvp = NULL;
215 bcopy(p->p_session->s_login, sess->s_login,
216 sizeof(sess->s_login));
217 p->p_flag &= ~P_CONTROLT;
218 pgrp->pg_session = sess;
219 KASSERT(p == curproc,
220 ("enterpgrp: mksession and p != curproc"));
222 pgrp->pg_session = p->p_session;
223 sess_hold(pgrp->pg_session);
226 LIST_INIT(&pgrp->pg_members);
227 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
229 SLIST_INIT(&pgrp->pg_sigiolst);
230 lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
231 } else if (pgrp == p->p_pgrp)
235 * Adjust eligibility of affected pgrps to participate in job control.
236 * Increment eligibility counts before decrementing, otherwise we
237 * could reach 0 spuriously during the first call.
240 fixjobc(p, p->p_pgrp, 0);
242 LIST_REMOVE(p, p_pglist);
243 if (LIST_EMPTY(&p->p_pgrp->pg_members))
246 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
251 * remove process from process group
254 leavepgrp(struct proc *p)
257 LIST_REMOVE(p, p_pglist);
258 if (LIST_EMPTY(&p->p_pgrp->pg_members))
265 * delete a process group
268 pgdelete(struct pgrp *pgrp)
272 * Reset any sigio structures pointing to us as a result of
273 * F_SETOWN with our pgid.
275 funsetownlst(&pgrp->pg_sigiolst);
277 if (pgrp->pg_session->s_ttyp != NULL &&
278 pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
279 pgrp->pg_session->s_ttyp->t_pgrp = NULL;
280 LIST_REMOVE(pgrp, pg_hash);
281 sess_rele(pgrp->pg_session);
286 * Adjust the ref count on a session structure. When the ref count falls to
287 * zero the tty is disassociated from the session and the session structure
288 * is freed. Note that tty assocation is not itself ref-counted.
291 sess_hold(struct session *sp)
297 sess_rele(struct session *sp)
299 KKASSERT(sp->s_count > 0);
300 if (--sp->s_count == 0) {
301 if (sp->s_ttyp && sp->s_ttyp->t_session) {
302 #ifdef TTY_DO_FULL_CLOSE
303 /* FULL CLOSE, see ttyclearsession() */
304 KKASSERT(sp->s_ttyp->t_session == sp);
305 sp->s_ttyp->t_session = NULL;
307 /* HALF CLOSE, see ttyclearsession() */
308 if (sp->s_ttyp->t_session == sp)
309 sp->s_ttyp->t_session = NULL;
312 kfree(sp, M_SESSION);
317 * Adjust pgrp jobc counters when specified process changes process group.
318 * We count the number of processes in each process group that "qualify"
319 * the group for terminal job control (those with a parent in a different
320 * process group of the same session). If that count reaches zero, the
321 * process group becomes orphaned. Check both the specified process'
322 * process group and that of its children.
323 * entering == 0 => p is leaving specified group.
324 * entering == 1 => p is entering specified group.
327 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
329 struct pgrp *hispgrp;
330 struct session *mysession = pgrp->pg_session;
333 * Check p's parent to see whether p qualifies its own process
334 * group; if so, adjust count for p's process group.
336 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
337 hispgrp->pg_session == mysession) {
340 else if (--pgrp->pg_jobc == 0)
345 * Check this process' children to see whether they qualify
346 * their process groups; if so, adjust counts for children's
349 LIST_FOREACH(p, &p->p_children, p_sibling)
350 if ((hispgrp = p->p_pgrp) != pgrp &&
351 hispgrp->pg_session == mysession &&
352 p->p_stat != SZOMB) {
355 else if (--hispgrp->pg_jobc == 0)
361 * A process group has become orphaned;
362 * if there are any stopped processes in the group,
363 * hang-up all process in that group.
366 orphanpg(struct pgrp *pg)
370 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
371 if (p->p_stat == SSTOP) {
372 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
382 * Add a new process to the allproc list and the PID hash. This
383 * also assigns a pid to the new process.
385 * MPALMOSTSAFE - acquires mplock for karc4random() call
388 proc_add_allproc(struct proc *p)
392 if ((random_offset = randompid) != 0) {
394 random_offset = karc4random() % random_offset;
398 spin_lock_wr(&allproc_spin);
399 p->p_pid = proc_getnewpid_locked(random_offset);
400 LIST_INSERT_HEAD(&allproc, p, p_list);
401 LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
402 spin_unlock_wr(&allproc_spin);
406 * Calculate a new process pid. This function is integrated into
407 * proc_add_allproc() to guarentee that the new pid is not reused before
408 * the new process can be added to the allproc list.
410 * MPSAFE - must be called with allproc_spin held.
414 proc_getnewpid_locked(int random_offset)
416 static pid_t nextpid;
417 static pid_t pidchecked;
421 * Find an unused process ID. We remember a range of unused IDs
422 * ready to use (from nextpid+1 through pidchecked-1).
424 nextpid = nextpid + 1 + random_offset;
427 * If the process ID prototype has wrapped around,
428 * restart somewhat above 0, as the low-numbered procs
429 * tend to include daemons that don't exit.
431 if (nextpid >= PID_MAX) {
432 nextpid = nextpid % PID_MAX;
437 if (nextpid >= pidchecked) {
440 pidchecked = PID_MAX;
442 * Scan the active and zombie procs to check whether this pid
443 * is in use. Remember the lowest pid that's greater
444 * than nextpid, so we can avoid checking for a while.
446 p = LIST_FIRST(&allproc);
448 for (; p != 0; p = LIST_NEXT(p, p_list)) {
449 while (p->p_pid == nextpid ||
450 p->p_pgrp->pg_id == nextpid ||
451 p->p_session->s_sid == nextpid) {
453 if (nextpid >= pidchecked)
456 if (p->p_pid > nextpid && pidchecked > p->p_pid)
457 pidchecked = p->p_pid;
458 if (p->p_pgrp->pg_id > nextpid &&
459 pidchecked > p->p_pgrp->pg_id)
460 pidchecked = p->p_pgrp->pg_id;
461 if (p->p_session->s_sid > nextpid &&
462 pidchecked > p->p_session->s_sid)
463 pidchecked = p->p_session->s_sid;
467 p = LIST_FIRST(&zombproc);
475 * Called from exit1 to remove a process from the allproc
476 * list and move it to the zombie list.
481 proc_move_allproc_zombie(struct proc *p)
483 spin_lock_wr(&allproc_spin);
485 spin_unlock_wr(&allproc_spin);
486 tsleep(p, 0, "reap1", hz / 10);
487 spin_lock_wr(&allproc_spin);
489 LIST_REMOVE(p, p_list);
490 LIST_INSERT_HEAD(&zombproc, p, p_list);
491 LIST_REMOVE(p, p_hash);
493 spin_unlock_wr(&allproc_spin);
497 * This routine is called from kern_wait() and will remove the process
498 * from the zombie list and the sibling list. This routine will block
499 * if someone has a lock on the proces (p_lock).
504 proc_remove_zombie(struct proc *p)
506 spin_lock_wr(&allproc_spin);
508 spin_unlock_wr(&allproc_spin);
509 tsleep(p, 0, "reap1", hz / 10);
510 spin_lock_wr(&allproc_spin);
512 LIST_REMOVE(p, p_list); /* off zombproc */
513 LIST_REMOVE(p, p_sibling);
514 spin_unlock_wr(&allproc_spin);
518 * Scan all processes on the allproc list. The process is automatically
519 * held for the callback. A return value of -1 terminates the loop.
524 allproc_scan(int (*callback)(struct proc *, void *), void *data)
529 spin_lock_rd(&allproc_spin);
530 LIST_FOREACH(p, &allproc, p_list) {
532 spin_unlock_rd(&allproc_spin);
533 r = callback(p, data);
534 spin_lock_rd(&allproc_spin);
539 spin_unlock_rd(&allproc_spin);
543 * Scan all lwps of processes on the allproc list. The lwp is automatically
544 * held for the callback. A return value of -1 terminates the loop.
546 * possibly not MPSAFE, needs to access foreingn proc structures
549 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
555 spin_lock_rd(&allproc_spin);
556 LIST_FOREACH(p, &allproc, p_list) {
558 spin_unlock_rd(&allproc_spin);
559 FOREACH_LWP_IN_PROC(lp, p) {
561 r = callback(lp, data);
564 spin_lock_rd(&allproc_spin);
569 spin_unlock_rd(&allproc_spin);
573 * Scan all processes on the zombproc list. The process is automatically
574 * held for the callback. A return value of -1 terminates the loop.
579 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
584 spin_lock_rd(&allproc_spin);
585 LIST_FOREACH(p, &zombproc, p_list) {
587 spin_unlock_rd(&allproc_spin);
588 r = callback(p, data);
589 spin_lock_rd(&allproc_spin);
594 spin_unlock_rd(&allproc_spin);
601 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
607 for (i = 0; i <= pgrphash; i++) {
608 if (!LIST_EMPTY(&pgrphashtbl[i])) {
609 kprintf("\tindx %d\n", i);
610 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
612 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
613 (void *)pgrp, (long)pgrp->pg_id,
614 (void *)pgrp->pg_session,
615 pgrp->pg_session->s_count,
616 (void *)LIST_FIRST(&pgrp->pg_members));
617 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
618 kprintf("\t\tpid %ld addr %p pgrp %p\n",
619 (long)p->p_pid, (void *)p,
629 * Locate a process on the zombie list. Return a held process or NULL.
636 LIST_FOREACH(p, &zombproc, p_list)
643 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
645 struct kinfo_proc ki;
647 int skp = 0, had_output = 0;
650 bzero(&ki, sizeof(ki));
651 fill_kinfo_proc(p, &ki);
652 if ((flags & KERN_PROC_FLAG_LWP) == 0)
655 FOREACH_LWP_IN_PROC(lp, p) {
657 fill_kinfo_lwp(lp, &ki.kp_lwp);
659 error = SYSCTL_OUT(req, &ki, sizeof(ki));
666 /* We need to output at least the proc, even if there is no lwp. */
667 if (had_output == 0) {
668 error = SYSCTL_OUT(req, &ki, sizeof(ki));
674 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
676 struct kinfo_proc ki;
679 fill_kinfo_proc_kthread(td, &ki);
680 error = SYSCTL_OUT(req, &ki, sizeof(ki));
687 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
689 int *name = (int*) arg1;
690 int oid = oidp->oid_number;
691 u_int namelen = arg2;
693 struct proclist *plist;
695 int doingzomb, flags = 0;
699 struct ucred *cr1 = curproc->p_ucred;
701 flags = oid & KERN_PROC_FLAGMASK;
702 oid &= ~KERN_PROC_FLAGMASK;
704 if ((oid == KERN_PROC_ALL && namelen != 0) ||
705 (oid != KERN_PROC_ALL && namelen != 1))
708 if (oid == KERN_PROC_PID) {
709 p = pfind((pid_t)name[0]);
712 if (!PRISON_CHECK(cr1, p->p_ucred))
715 error = sysctl_out_proc(p, req, flags);
721 /* overestimate by 5 procs */
722 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
726 for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
731 LIST_FOREACH(p, plist, p_list) {
733 * Show a user only their processes.
735 if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
738 * Skip embryonic processes.
740 if (p->p_stat == SIDL)
743 * TODO - make more efficient (see notes below).
748 /* could do this by traversing pgrp */
749 if (p->p_pgrp == NULL ||
750 p->p_pgrp->pg_id != (pid_t)name[0])
755 if ((p->p_flag & P_CONTROLT) == 0 ||
756 p->p_session == NULL ||
757 p->p_session->s_ttyp == NULL ||
758 dev2udev(p->p_session->s_ttyp->t_dev) !=
764 if (p->p_ucred == NULL ||
765 p->p_ucred->cr_uid != (uid_t)name[0])
770 if (p->p_ucred == NULL ||
771 p->p_ucred->cr_ruid != (uid_t)name[0])
776 if (!PRISON_CHECK(cr1, p->p_ucred))
779 error = sysctl_out_proc(p, req, flags);
787 * Iterate over all active cpus and scan their thread list. Start
788 * with the next logical cpu and end with our original cpu. We
789 * migrate our own thread to each target cpu in order to safely scan
790 * its thread list. In the last loop we migrate back to our original
793 origcpu = mycpu->gd_cpuid;
794 if (!ps_showallthreads || jailed(cr1))
796 for (n = 1; n <= ncpus; ++n) {
800 nid = (origcpu + n) % ncpus;
801 if ((smp_active_mask & (1 << nid)) == 0)
803 rgd = globaldata_find(nid);
804 lwkt_setcpu_self(rgd);
806 TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
819 error = sysctl_out_proc_kthread(td, req, doingzomb);
830 * This sysctl allows a process to retrieve the argument list or process
831 * title for another process without groping around in the address space
832 * of the other process. It also allow a process to set its own "process
833 * title to a string of its own choice.
836 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
838 int *name = (int*) arg1;
839 u_int namelen = arg2;
843 struct ucred *cr1 = curproc->p_ucred;
848 p = pfind((pid_t)name[0]);
852 if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
855 if (req->newptr && curproc != p)
858 if (req->oldptr && p->p_args != NULL)
859 error = SYSCTL_OUT(req, p->p_args->ar_args, p->p_args->ar_length);
860 if (req->newptr == NULL)
863 if (p->p_args && --p->p_args->ar_ref == 0)
864 FREE(p->p_args, M_PARGS);
867 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
870 MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
873 pa->ar_length = req->newlen;
874 error = SYSCTL_IN(req, pa->ar_args, req->newlen);
882 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table");
884 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
885 0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
887 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
888 sysctl_kern_proc, "Process table");
890 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
891 sysctl_kern_proc, "Process table");
893 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
894 sysctl_kern_proc, "Process table");
896 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
897 sysctl_kern_proc, "Process table");
899 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
900 sysctl_kern_proc, "Process table");
902 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
903 sysctl_kern_proc, "Process table");
905 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
906 sysctl_kern_proc, "Process table");
908 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
909 sysctl_kern_proc, "Process table");
911 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
912 sysctl_kern_proc, "Process table");
914 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
915 sysctl_kern_proc, "Process table");
917 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
918 sysctl_kern_proc, "Process table");
920 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
921 sysctl_kern_proc_args, "Process argument list");