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>
57 #include <sys/mplock2.h>
59 static MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
60 MALLOC_DEFINE(M_SESSION, "session", "session header");
61 MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
62 MALLOC_DEFINE(M_LWP, "lwp", "lwp structures");
63 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
65 int ps_showallprocs = 1;
66 static int ps_showallthreads = 1;
67 SYSCTL_INT(_security, OID_AUTO, ps_showallprocs, CTLFLAG_RW,
69 "Unprivileged processes can see proccesses with different UID/GID");
70 SYSCTL_INT(_security, OID_AUTO, ps_showallthreads, CTLFLAG_RW,
71 &ps_showallthreads, 0,
72 "Unprivileged processes can see kernel threads");
74 static void pgdelete(struct pgrp *);
75 static void orphanpg(struct pgrp *pg);
76 static pid_t proc_getnewpid_locked(int random_offset);
81 struct pidhashhead *pidhashtbl;
83 struct pgrphashhead *pgrphashtbl;
85 struct proclist allproc;
86 struct proclist zombproc;
87 struct spinlock allproc_spin;
90 * Random component to nextpid generation. We mix in a random factor to make
91 * it a little harder to predict. We sanity check the modulus value to avoid
92 * doing it in critical paths. Don't let it be too small or we pointlessly
93 * waste randomness entropy, and don't let it be impossibly large. Using a
94 * modulus that is too big causes a LOT more process table scans and slows
95 * down fork processing as the pidchecked caching is defeated.
97 static int randompid = 0;
100 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
105 error = sysctl_handle_int(oidp, &pid, 0, req);
106 if (error || !req->newptr)
108 if (pid < 0 || pid > PID_MAX - 100) /* out of range */
110 else if (pid < 2) /* NOP */
112 else if (pid < 100) /* Make it reasonable */
118 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
119 0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
122 * Initialize global process hashing structures.
128 LIST_INIT(&zombproc);
129 spin_init(&allproc_spin);
131 pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
132 pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
137 * Is p an inferior of the current process?
140 inferior(struct proc *p)
142 for (; p != curproc; p = p->p_pptr)
149 * Locate a process by number
156 LIST_FOREACH(p, PIDHASH(pid), p_hash) {
164 * Locate a process group by number
171 LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
172 if (pgrp->pg_id == pgid)
179 * Move p to a new or existing process group (and session)
182 enterpgrp(struct proc *p, pid_t pgid, int mksess)
184 struct pgrp *pgrp = pgfind(pgid);
186 KASSERT(pgrp == NULL || !mksess,
187 ("enterpgrp: setsid into non-empty pgrp"));
188 KASSERT(!SESS_LEADER(p),
189 ("enterpgrp: session leader attempted setpgrp"));
192 pid_t savepid = p->p_pid;
197 KASSERT(p->p_pid == pgid,
198 ("enterpgrp: new pgrp and pid != pgid"));
199 if ((np = pfind(savepid)) == NULL || np != p)
201 MALLOC(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP,
204 struct session *sess;
209 MALLOC(sess, struct session *, sizeof(struct session),
210 M_SESSION, M_WAITOK);
212 sess->s_sid = p->p_pid;
214 sess->s_ttyvp = NULL;
216 bcopy(p->p_session->s_login, sess->s_login,
217 sizeof(sess->s_login));
218 p->p_flag &= ~P_CONTROLT;
219 pgrp->pg_session = sess;
220 KASSERT(p == curproc,
221 ("enterpgrp: mksession and p != curproc"));
223 pgrp->pg_session = p->p_session;
224 sess_hold(pgrp->pg_session);
227 LIST_INIT(&pgrp->pg_members);
228 LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
230 SLIST_INIT(&pgrp->pg_sigiolst);
231 lockinit(&pgrp->pg_lock, "pgwt", 0, 0);
232 } else if (pgrp == p->p_pgrp)
236 * Adjust eligibility of affected pgrps to participate in job control.
237 * Increment eligibility counts before decrementing, otherwise we
238 * could reach 0 spuriously during the first call.
241 fixjobc(p, p->p_pgrp, 0);
243 LIST_REMOVE(p, p_pglist);
244 if (LIST_EMPTY(&p->p_pgrp->pg_members))
247 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
252 * remove process from process group
255 leavepgrp(struct proc *p)
258 LIST_REMOVE(p, p_pglist);
259 if (LIST_EMPTY(&p->p_pgrp->pg_members))
266 * delete a process group
269 pgdelete(struct pgrp *pgrp)
273 * Reset any sigio structures pointing to us as a result of
274 * F_SETOWN with our pgid.
276 funsetownlst(&pgrp->pg_sigiolst);
278 if (pgrp->pg_session->s_ttyp != NULL &&
279 pgrp->pg_session->s_ttyp->t_pgrp == pgrp)
280 pgrp->pg_session->s_ttyp->t_pgrp = NULL;
281 LIST_REMOVE(pgrp, pg_hash);
282 sess_rele(pgrp->pg_session);
287 * Adjust the ref count on a session structure. When the ref count falls to
288 * zero the tty is disassociated from the session and the session structure
289 * is freed. Note that tty assocation is not itself ref-counted.
292 sess_hold(struct session *sp)
298 sess_rele(struct session *sp)
300 KKASSERT(sp->s_count > 0);
301 if (--sp->s_count == 0) {
302 if (sp->s_ttyp && sp->s_ttyp->t_session) {
303 #ifdef TTY_DO_FULL_CLOSE
304 /* FULL CLOSE, see ttyclearsession() */
305 KKASSERT(sp->s_ttyp->t_session == sp);
306 sp->s_ttyp->t_session = NULL;
308 /* HALF CLOSE, see ttyclearsession() */
309 if (sp->s_ttyp->t_session == sp)
310 sp->s_ttyp->t_session = NULL;
313 kfree(sp, M_SESSION);
318 * Adjust pgrp jobc counters when specified process changes process group.
319 * We count the number of processes in each process group that "qualify"
320 * the group for terminal job control (those with a parent in a different
321 * process group of the same session). If that count reaches zero, the
322 * process group becomes orphaned. Check both the specified process'
323 * process group and that of its children.
324 * entering == 0 => p is leaving specified group.
325 * entering == 1 => p is entering specified group.
328 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
330 struct pgrp *hispgrp;
331 struct session *mysession = pgrp->pg_session;
334 * Check p's parent to see whether p qualifies its own process
335 * group; if so, adjust count for p's process group.
337 if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
338 hispgrp->pg_session == mysession) {
341 else if (--pgrp->pg_jobc == 0)
346 * Check this process' children to see whether they qualify
347 * their process groups; if so, adjust counts for children's
350 LIST_FOREACH(p, &p->p_children, p_sibling)
351 if ((hispgrp = p->p_pgrp) != pgrp &&
352 hispgrp->pg_session == mysession &&
353 p->p_stat != SZOMB) {
356 else if (--hispgrp->pg_jobc == 0)
362 * A process group has become orphaned;
363 * if there are any stopped processes in the group,
364 * hang-up all process in that group.
367 orphanpg(struct pgrp *pg)
371 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
372 if (p->p_stat == SSTOP) {
373 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
383 * Add a new process to the allproc list and the PID hash. This
384 * also assigns a pid to the new process.
386 * MPALMOSTSAFE - acquires mplock for karc4random() call
389 proc_add_allproc(struct proc *p)
393 if ((random_offset = randompid) != 0) {
395 random_offset = karc4random() % random_offset;
399 spin_lock_wr(&allproc_spin);
400 p->p_pid = proc_getnewpid_locked(random_offset);
401 LIST_INSERT_HEAD(&allproc, p, p_list);
402 LIST_INSERT_HEAD(PIDHASH(p->p_pid), p, p_hash);
403 spin_unlock_wr(&allproc_spin);
407 * Calculate a new process pid. This function is integrated into
408 * proc_add_allproc() to guarentee that the new pid is not reused before
409 * the new process can be added to the allproc list.
411 * MPSAFE - must be called with allproc_spin held.
415 proc_getnewpid_locked(int random_offset)
417 static pid_t nextpid;
418 static pid_t pidchecked;
422 * Find an unused process ID. We remember a range of unused IDs
423 * ready to use (from nextpid+1 through pidchecked-1).
425 nextpid = nextpid + 1 + random_offset;
428 * If the process ID prototype has wrapped around,
429 * restart somewhat above 0, as the low-numbered procs
430 * tend to include daemons that don't exit.
432 if (nextpid >= PID_MAX) {
433 nextpid = nextpid % PID_MAX;
438 if (nextpid >= pidchecked) {
441 pidchecked = PID_MAX;
443 * Scan the active and zombie procs to check whether this pid
444 * is in use. Remember the lowest pid that's greater
445 * than nextpid, so we can avoid checking for a while.
447 p = LIST_FIRST(&allproc);
449 for (; p != 0; p = LIST_NEXT(p, p_list)) {
450 while (p->p_pid == nextpid ||
451 p->p_pgrp->pg_id == nextpid ||
452 p->p_session->s_sid == nextpid) {
454 if (nextpid >= pidchecked)
457 if (p->p_pid > nextpid && pidchecked > p->p_pid)
458 pidchecked = p->p_pid;
459 if (p->p_pgrp->pg_id > nextpid &&
460 pidchecked > p->p_pgrp->pg_id)
461 pidchecked = p->p_pgrp->pg_id;
462 if (p->p_session->s_sid > nextpid &&
463 pidchecked > p->p_session->s_sid)
464 pidchecked = p->p_session->s_sid;
468 p = LIST_FIRST(&zombproc);
476 * Called from exit1 to remove a process from the allproc
477 * list and move it to the zombie list.
482 proc_move_allproc_zombie(struct proc *p)
484 spin_lock_wr(&allproc_spin);
486 spin_unlock_wr(&allproc_spin);
487 tsleep(p, 0, "reap1", hz / 10);
488 spin_lock_wr(&allproc_spin);
490 LIST_REMOVE(p, p_list);
491 LIST_INSERT_HEAD(&zombproc, p, p_list);
492 LIST_REMOVE(p, p_hash);
494 spin_unlock_wr(&allproc_spin);
498 * This routine is called from kern_wait() and will remove the process
499 * from the zombie list and the sibling list. This routine will block
500 * if someone has a lock on the proces (p_lock).
505 proc_remove_zombie(struct proc *p)
507 spin_lock_wr(&allproc_spin);
509 spin_unlock_wr(&allproc_spin);
510 tsleep(p, 0, "reap1", hz / 10);
511 spin_lock_wr(&allproc_spin);
513 LIST_REMOVE(p, p_list); /* off zombproc */
514 LIST_REMOVE(p, p_sibling);
515 spin_unlock_wr(&allproc_spin);
519 * Scan all processes on the allproc list. The process is automatically
520 * held for the callback. A return value of -1 terminates the loop.
525 allproc_scan(int (*callback)(struct proc *, void *), void *data)
530 spin_lock_rd(&allproc_spin);
531 LIST_FOREACH(p, &allproc, p_list) {
533 spin_unlock_rd(&allproc_spin);
534 r = callback(p, data);
535 spin_lock_rd(&allproc_spin);
540 spin_unlock_rd(&allproc_spin);
544 * Scan all lwps of processes on the allproc list. The lwp is automatically
545 * held for the callback. A return value of -1 terminates the loop.
547 * possibly not MPSAFE, needs to access foreingn proc structures
550 alllwp_scan(int (*callback)(struct lwp *, void *), void *data)
556 spin_lock_rd(&allproc_spin);
557 LIST_FOREACH(p, &allproc, p_list) {
559 spin_unlock_rd(&allproc_spin);
560 FOREACH_LWP_IN_PROC(lp, p) {
562 r = callback(lp, data);
565 spin_lock_rd(&allproc_spin);
570 spin_unlock_rd(&allproc_spin);
574 * Scan all processes on the zombproc list. The process is automatically
575 * held for the callback. A return value of -1 terminates the loop.
580 zombproc_scan(int (*callback)(struct proc *, void *), void *data)
585 spin_lock_rd(&allproc_spin);
586 LIST_FOREACH(p, &zombproc, p_list) {
588 spin_unlock_rd(&allproc_spin);
589 r = callback(p, data);
590 spin_lock_rd(&allproc_spin);
595 spin_unlock_rd(&allproc_spin);
602 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
608 for (i = 0; i <= pgrphash; i++) {
609 if (!LIST_EMPTY(&pgrphashtbl[i])) {
610 kprintf("\tindx %d\n", i);
611 LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
613 "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
614 (void *)pgrp, (long)pgrp->pg_id,
615 (void *)pgrp->pg_session,
616 pgrp->pg_session->s_count,
617 (void *)LIST_FIRST(&pgrp->pg_members));
618 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
619 kprintf("\t\tpid %ld addr %p pgrp %p\n",
620 (long)p->p_pid, (void *)p,
630 * Locate a process on the zombie list. Return a held process or NULL.
637 LIST_FOREACH(p, &zombproc, p_list)
644 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
646 struct kinfo_proc ki;
648 int skp = 0, had_output = 0;
651 bzero(&ki, sizeof(ki));
652 fill_kinfo_proc(p, &ki);
653 if ((flags & KERN_PROC_FLAG_LWP) == 0)
656 FOREACH_LWP_IN_PROC(lp, p) {
658 fill_kinfo_lwp(lp, &ki.kp_lwp);
660 error = SYSCTL_OUT(req, &ki, sizeof(ki));
667 /* We need to output at least the proc, even if there is no lwp. */
668 if (had_output == 0) {
669 error = SYSCTL_OUT(req, &ki, sizeof(ki));
675 sysctl_out_proc_kthread(struct thread *td, struct sysctl_req *req, int flags)
677 struct kinfo_proc ki;
680 fill_kinfo_proc_kthread(td, &ki);
681 error = SYSCTL_OUT(req, &ki, sizeof(ki));
688 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
690 int *name = (int*) arg1;
691 int oid = oidp->oid_number;
692 u_int namelen = arg2;
694 struct proclist *plist;
696 int doingzomb, flags = 0;
700 struct ucred *cr1 = curproc->p_ucred;
702 flags = oid & KERN_PROC_FLAGMASK;
703 oid &= ~KERN_PROC_FLAGMASK;
705 if ((oid == KERN_PROC_ALL && namelen != 0) ||
706 (oid != KERN_PROC_ALL && namelen != 1))
709 if (oid == KERN_PROC_PID) {
710 p = pfind((pid_t)name[0]);
713 if (!PRISON_CHECK(cr1, p->p_ucred))
716 error = sysctl_out_proc(p, req, flags);
722 /* overestimate by 5 procs */
723 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
727 for (doingzomb = 0; doingzomb <= 1; doingzomb++) {
732 LIST_FOREACH(p, plist, p_list) {
734 * Show a user only their processes.
736 if ((!ps_showallprocs) && p_trespass(cr1, p->p_ucred))
739 * Skip embryonic processes.
741 if (p->p_stat == SIDL)
744 * TODO - make more efficient (see notes below).
749 /* could do this by traversing pgrp */
750 if (p->p_pgrp == NULL ||
751 p->p_pgrp->pg_id != (pid_t)name[0])
756 if ((p->p_flag & P_CONTROLT) == 0 ||
757 p->p_session == NULL ||
758 p->p_session->s_ttyp == NULL ||
759 dev2udev(p->p_session->s_ttyp->t_dev) !=
765 if (p->p_ucred == NULL ||
766 p->p_ucred->cr_uid != (uid_t)name[0])
771 if (p->p_ucred == NULL ||
772 p->p_ucred->cr_ruid != (uid_t)name[0])
777 if (!PRISON_CHECK(cr1, p->p_ucred))
780 error = sysctl_out_proc(p, req, flags);
788 * Iterate over all active cpus and scan their thread list. Start
789 * with the next logical cpu and end with our original cpu. We
790 * migrate our own thread to each target cpu in order to safely scan
791 * its thread list. In the last loop we migrate back to our original
794 origcpu = mycpu->gd_cpuid;
795 if (!ps_showallthreads || jailed(cr1))
797 for (n = 1; n <= ncpus; ++n) {
801 nid = (origcpu + n) % ncpus;
802 if ((smp_active_mask & (1 << nid)) == 0)
804 rgd = globaldata_find(nid);
805 lwkt_setcpu_self(rgd);
807 TAILQ_FOREACH(td, &mycpu->gd_tdallq, td_allq) {
820 error = sysctl_out_proc_kthread(td, req, doingzomb);
831 * This sysctl allows a process to retrieve the argument list or process
832 * title for another process without groping around in the address space
833 * of the other process. It also allow a process to set its own "process
834 * title to a string of its own choice.
837 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
839 int *name = (int*) arg1;
840 u_int namelen = arg2;
844 struct ucred *cr1 = curproc->p_ucred;
849 p = pfind((pid_t)name[0]);
853 if ((!ps_argsopen) && p_trespass(cr1, p->p_ucred))
856 if (req->newptr && curproc != p)
859 if (req->oldptr && p->p_args != NULL)
860 error = SYSCTL_OUT(req, p->p_args->ar_args, p->p_args->ar_length);
861 if (req->newptr == NULL)
864 if (p->p_args && --p->p_args->ar_ref == 0)
865 FREE(p->p_args, M_PARGS);
868 if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
871 MALLOC(pa, struct pargs *, sizeof(struct pargs) + req->newlen,
874 pa->ar_length = req->newlen;
875 error = SYSCTL_IN(req, pa->ar_args, req->newlen);
883 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD, 0, "Process table");
885 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT,
886 0, 0, sysctl_kern_proc, "S,proc", "Return entire process table");
888 SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD,
889 sysctl_kern_proc, "Process table");
891 SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD,
892 sysctl_kern_proc, "Process table");
894 SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD,
895 sysctl_kern_proc, "Process table");
897 SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD,
898 sysctl_kern_proc, "Process table");
900 SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD,
901 sysctl_kern_proc, "Process table");
903 SYSCTL_NODE(_kern_proc, (KERN_PROC_ALL | KERN_PROC_FLAG_LWP), all_lwp, CTLFLAG_RD,
904 sysctl_kern_proc, "Process table");
906 SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_FLAG_LWP), pgrp_lwp, CTLFLAG_RD,
907 sysctl_kern_proc, "Process table");
909 SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_FLAG_LWP), tty_lwp, CTLFLAG_RD,
910 sysctl_kern_proc, "Process table");
912 SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_FLAG_LWP), uid_lwp, CTLFLAG_RD,
913 sysctl_kern_proc, "Process table");
915 SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_FLAG_LWP), ruid_lwp, CTLFLAG_RD,
916 sysctl_kern_proc, "Process table");
918 SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_FLAG_LWP), pid_lwp, CTLFLAG_RD,
919 sysctl_kern_proc, "Process table");
921 SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args, CTLFLAG_RW | CTLFLAG_ANYBODY,
922 sysctl_kern_proc_args, "Process argument list");