sys/kern/kern_proc.c

   1 /*-
   2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   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.
  16  *
  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
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
  30  */
  31
  32 #include <sys/cdefs.h>
  33 __FBSDID("$FreeBSD$");
  34
  35 #include "opt_compat.h"
  36 #include "opt_ddb.h"
  37 #include "opt_ktrace.h"
  38 #include "opt_kstack_pages.h"
  39 #include "opt_stack.h"
  40
  41 #include <sys/param.h>
  42 #include <sys/systm.h>
  43 #include <sys/elf.h>
  44 #include <sys/eventhandler.h>
  45 #include <sys/exec.h>
  46 #include <sys/jail.h>
  47 #include <sys/kernel.h>
  48 #include <sys/limits.h>
  49 #include <sys/lock.h>
  50 #include <sys/loginclass.h>
  51 #include <sys/malloc.h>
  52 #include <sys/mman.h>
  53 #include <sys/mount.h>
  54 #include <sys/mutex.h>
  55 #include <sys/proc.h>
  56 #include <sys/ptrace.h>
  57 #include <sys/refcount.h>
  58 #include <sys/resourcevar.h>
  59 #include <sys/rwlock.h>
  60 #include <sys/sbuf.h>
  61 #include <sys/sysent.h>
  62 #include <sys/sched.h>
  63 #include <sys/smp.h>
  64 #include <sys/stack.h>
  65 #include <sys/stat.h>
  66 #include <sys/sysctl.h>
  67 #include <sys/filedesc.h>
  68 #include <sys/tty.h>
  69 #include <sys/signalvar.h>
  70 #include <sys/sdt.h>
  71 #include <sys/sx.h>
  72 #include <sys/user.h>
  73 #include <sys/vnode.h>
  74 #include <sys/wait.h>
  75
  76 #ifdef DDB
  77 #include <ddb/ddb.h>
  78 #endif
  79
  80 #include <vm/vm.h>
  81 #include <vm/vm_param.h>
  82 #include <vm/vm_extern.h>
  83 #include <vm/pmap.h>
  84 #include <vm/vm_map.h>
  85 #include <vm/vm_object.h>
  86 #include <vm/vm_page.h>
  87 #include <vm/uma.h>
  88
  89 #ifdef COMPAT_FREEBSD32
  90 #include <compat/freebsd32/freebsd32.h>
  91 #include <compat/freebsd32/freebsd32_util.h>
  92 #endif
  93
  94 SDT_PROVIDER_DEFINE(proc);
  95 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
  96     "int");
  97 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
  98     "int");
  99 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
 100     "struct thread *");
 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");
 104
 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");
 109
 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,
 115     int preferthread);
 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);
 124
 125 /*
 126  * Other process lists
 127  */
 128 struct pidhashhead *pidhashtbl;
 129 u_long pidhash;
 130 struct pgrphashhead *pgrphashtbl;
 131 u_long pgrphash;
 132 struct proclist allproc;
 133 struct proclist zombproc;
 134 struct sx __exclusive_cache_line allproc_lock;
 135 struct sx __exclusive_cache_line proctree_lock;
 136 struct mtx __exclusive_cache_line ppeers_lock;
 137 uma_zone_t proc_zone;
 138
 139 /*
 140  * The offset of various fields in struct proc and struct thread.
 141  * These are used by kernel debuggers to enumerate kernel threads and
 142  * processes.
 143  */
 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);
 153
 154 EVENTHANDLER_LIST_DEFINE(process_ctor);
 155 EVENTHANDLER_LIST_DEFINE(process_dtor);
 156 EVENTHANDLER_LIST_DEFINE(process_init);
 157 EVENTHANDLER_LIST_DEFINE(process_fini);
 158 EVENTHANDLER_LIST_DEFINE(process_exit);
 159 EVENTHANDLER_LIST_DEFINE(process_fork);
 160 EVENTHANDLER_LIST_DEFINE(process_exec);
 161
 162 EVENTHANDLER_LIST_DECLARE(thread_ctor);
 163 EVENTHANDLER_LIST_DECLARE(thread_dtor);
 164
 165 int kstack_pages = KSTACK_PAGES;
 166 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
 167     "Kernel stack size in pages");
 168 static int vmmap_skip_res_cnt = 0;
 169 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
 170     &vmmap_skip_res_cnt, 0,
 171     "Skip calculation of the pages resident count in kern.proc.vmmap");
 172
 173 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
 174 #ifdef COMPAT_FREEBSD32
 175 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
 176 #endif
 177
 178 /*
 179  * Initialize global process hashing structures.
 180  */
 181 void
 182 procinit(void)
 183 {
 184
 185         sx_init(&allproc_lock, "allproc");
 186         sx_init(&proctree_lock, "proctree");
 187         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
 188         LIST_INIT(&allproc);
 189         LIST_INIT(&zombproc);
 190         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
 191         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
 192         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
 193             proc_ctor, proc_dtor, proc_init, proc_fini,
 194             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
 195         uihashinit();
 196 }
 197
 198 /*
 199  * Prepare a proc for use.
 200  */
 201 static int
 202 proc_ctor(void *mem, int size, void *arg, int flags)
 203 {
 204         struct proc *p;
 205         struct thread *td;
 206
 207         p = (struct proc *)mem;
 208         SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
 209         EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
 210         SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
 211         td = FIRST_THREAD_IN_PROC(p);
 212         if (td != NULL) {
 213                 /* Make sure all thread constructors are executed */
 214                 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
 215         }
 216         return (0);
 217 }
 218
 219 /*
 220  * Reclaim a proc after use.
 221  */
 222 static void
 223 proc_dtor(void *mem, int size, void *arg)
 224 {
 225         struct proc *p;
 226         struct thread *td;
 227
 228         /* INVARIANTS checks go here */
 229         p = (struct proc *)mem;
 230         td = FIRST_THREAD_IN_PROC(p);
 231         SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
 232         if (td != NULL) {
 233 #ifdef INVARIANTS
 234                 KASSERT((p->p_numthreads == 1),
 235                     ("bad number of threads in exiting process"));
 236                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
 237 #endif
 238                 /* Free all OSD associated to this thread. */
 239                 osd_thread_exit(td);
 240                 td_softdep_cleanup(td);
 241                 MPASS(td->td_su == NULL);
 242
 243                 /* Make sure all thread destructors are executed */
 244                 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
 245         }
 246         EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
 247         if (p->p_ksi != NULL)
 248                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
 249         SDT_PROBE3(proc, , dtor, return, p, size, arg);
 250 }
 251
 252 /*
 253  * Initialize type-stable parts of a proc (when newly created).
 254  */
 255 static int
 256 proc_init(void *mem, int size, int flags)
 257 {
 258         struct proc *p;
 259
 260         p = (struct proc *)mem;
 261         SDT_PROBE3(proc, , init, entry, p, size, flags);
 262         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
 263         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
 264         mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
 265         mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
 266         mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
 267         cv_init(&p->p_pwait, "ppwait");
 268         cv_init(&p->p_dbgwait, "dbgwait");
 269         TAILQ_INIT(&p->p_threads);           /* all threads in proc */
 270         EVENTHANDLER_DIRECT_INVOKE(process_init, p);
 271         p->p_stats = pstats_alloc();
 272         p->p_pgrp = NULL;
 273         SDT_PROBE3(proc, , init, return, p, size, flags);
 274         return (0);
 275 }
 276
 277 /*
 278  * UMA should ensure that this function is never called.
 279  * Freeing a proc structure would violate type stability.
 280  */
 281 static void
 282 proc_fini(void *mem, int size)
 283 {
 284 #ifdef notnow
 285         struct proc *p;
 286
 287         p = (struct proc *)mem;
 288         EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
 289         pstats_free(p->p_stats);
 290         thread_free(FIRST_THREAD_IN_PROC(p));
 291         mtx_destroy(&p->p_mtx);
 292         if (p->p_ksi != NULL)
 293                 ksiginfo_free(p->p_ksi);
 294 #else
 295         panic("proc reclaimed");
 296 #endif
 297 }
 298
 299 /*
 300  * Is p an inferior of the current process?
 301  */
 302 int
 303 inferior(struct proc *p)
 304 {
 305
 306         sx_assert(&proctree_lock, SX_LOCKED);
 307         PROC_LOCK_ASSERT(p, MA_OWNED);
 308         for (; p != curproc; p = proc_realparent(p)) {
 309                 if (p->p_pid == 0)
 310                         return (0);
 311         }
 312         return (1);
 313 }
 314
 315 struct proc *
 316 pfind_locked(pid_t pid)
 317 {
 318         struct proc *p;
 319
 320         sx_assert(&allproc_lock, SX_LOCKED);
 321         LIST_FOREACH(p, PIDHASH(pid), p_hash) {
 322                 if (p->p_pid == pid) {
 323                         PROC_LOCK(p);
 324                         if (p->p_state == PRS_NEW) {
 325                                 PROC_UNLOCK(p);
 326                                 p = NULL;
 327                         }
 328                         break;
 329                 }
 330         }
 331         return (p);
 332 }
 333
 334 /*
 335  * Locate a process by number; return only "live" processes -- i.e., neither
 336  * zombies nor newly born but incompletely initialized processes.  By not
 337  * returning processes in the PRS_NEW state, we allow callers to avoid
 338  * testing for that condition to avoid dereferencing p_ucred, et al.
 339  */
 340 struct proc *
 341 pfind(pid_t pid)
 342 {
 343         struct proc *p;
 344
 345         sx_slock(&allproc_lock);
 346         p = pfind_locked(pid);
 347         sx_sunlock(&allproc_lock);
 348         return (p);
 349 }
 350
 351 static struct proc *
 352 pfind_tid_locked(pid_t tid)
 353 {
 354         struct proc *p;
 355         struct thread *td;
 356
 357         sx_assert(&allproc_lock, SX_LOCKED);
 358         FOREACH_PROC_IN_SYSTEM(p) {
 359                 PROC_LOCK(p);
 360                 if (p->p_state == PRS_NEW) {
 361                         PROC_UNLOCK(p);
 362                         continue;
 363                 }
 364                 FOREACH_THREAD_IN_PROC(p, td) {
 365                         if (td->td_tid == tid)
 366                                 goto found;
 367                 }
 368                 PROC_UNLOCK(p);
 369         }
 370 found:
 371         return (p);
 372 }
 373
 374 /*
 375  * Locate a process group by number.
 376  * The caller must hold proctree_lock.
 377  */
 378 struct pgrp *
 379 pgfind(pid_t pgid)
 380 {
 381         struct pgrp *pgrp;
 382
 383         sx_assert(&proctree_lock, SX_LOCKED);
 384
 385         LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
 386                 if (pgrp->pg_id == pgid) {
 387                         PGRP_LOCK(pgrp);
 388                         return (pgrp);
 389                 }
 390         }
 391         return (NULL);
 392 }
 393
 394 /*
 395  * Locate process and do additional manipulations, depending on flags.
 396  */
 397 int
 398 pget(pid_t pid, int flags, struct proc **pp)
 399 {
 400         struct proc *p;
 401         int error;
 402
 403         p = curproc;
 404         if (p->p_pid == pid) {
 405                 PROC_LOCK(p);
 406         } else {
 407                 sx_slock(&allproc_lock);
 408                 if (pid <= PID_MAX) {
 409                         p = pfind_locked(pid);
 410                         if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
 411                                 p = zpfind_locked(pid);
 412                 } else if ((flags & PGET_NOTID) == 0) {
 413                         p = pfind_tid_locked(pid);
 414                 } else {
 415                         p = NULL;
 416                 }
 417                 sx_sunlock(&allproc_lock);
 418                 if (p == NULL)
 419                         return (ESRCH);
 420                 if ((flags & PGET_CANSEE) != 0) {
 421                         error = p_cansee(curthread, p);
 422                         if (error != 0)
 423                                 goto errout;
 424                 }
 425         }
 426         if ((flags & PGET_CANDEBUG) != 0) {
 427                 error = p_candebug(curthread, p);
 428                 if (error != 0)
 429                         goto errout;
 430         }
 431         if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
 432                 error = EPERM;
 433                 goto errout;
 434         }
 435         if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
 436                 error = ESRCH;
 437                 goto errout;
 438         }
 439         if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
 440                 /*
 441                  * XXXRW: Not clear ESRCH is the right error during proc
 442                  * execve().
 443                  */
 444                 error = ESRCH;
 445                 goto errout;
 446         }
 447         if ((flags & PGET_HOLD) != 0) {
 448                 _PHOLD(p);
 449                 PROC_UNLOCK(p);
 450         }
 451         *pp = p;
 452         return (0);
 453 errout:
 454         PROC_UNLOCK(p);
 455         return (error);
 456 }
 457
 458 /*
 459  * Create a new process group.
 460  * pgid must be equal to the pid of p.
 461  * Begin a new session if required.
 462  */
 463 int
 464 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
 465 {
 466
 467         sx_assert(&proctree_lock, SX_XLOCKED);
 468
 469         KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
 470         KASSERT(p->p_pid == pgid,
 471             ("enterpgrp: new pgrp and pid != pgid"));
 472         KASSERT(pgfind(pgid) == NULL,
 473             ("enterpgrp: pgrp with pgid exists"));
 474         KASSERT(!SESS_LEADER(p),
 475             ("enterpgrp: session leader attempted setpgrp"));
 476
 477         mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
 478
 479         if (sess != NULL) {
 480                 /*
 481                  * new session
 482                  */
 483                 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
 484                 PROC_LOCK(p);
 485                 p->p_flag &= ~P_CONTROLT;
 486                 PROC_UNLOCK(p);
 487                 PGRP_LOCK(pgrp);
 488                 sess->s_leader = p;
 489                 sess->s_sid = p->p_pid;
 490                 refcount_init(&sess->s_count, 1);
 491                 sess->s_ttyvp = NULL;
 492                 sess->s_ttydp = NULL;
 493                 sess->s_ttyp = NULL;
 494                 bcopy(p->p_session->s_login, sess->s_login,
 495                             sizeof(sess->s_login));
 496                 pgrp->pg_session = sess;
 497                 KASSERT(p == curproc,
 498                     ("enterpgrp: mksession and p != curproc"));
 499         } else {
 500                 pgrp->pg_session = p->p_session;
 501                 sess_hold(pgrp->pg_session);
 502                 PGRP_LOCK(pgrp);
 503         }
 504         pgrp->pg_id = pgid;
 505         LIST_INIT(&pgrp->pg_members);
 506
 507         /*
 508          * As we have an exclusive lock of proctree_lock,
 509          * this should not deadlock.
 510          */
 511         LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
 512         pgrp->pg_jobc = 0;
 513         SLIST_INIT(&pgrp->pg_sigiolst);
 514         PGRP_UNLOCK(pgrp);
 515
 516         doenterpgrp(p, pgrp);
 517
 518         return (0);
 519 }
 520
 521 /*
 522  * Move p to an existing process group
 523  */
 524 int
 525 enterthispgrp(struct proc *p, struct pgrp *pgrp)
 526 {
 527
 528         sx_assert(&proctree_lock, SX_XLOCKED);
 529         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 530         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 531         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 532         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 533         KASSERT(pgrp->pg_session == p->p_session,
 534                 ("%s: pgrp's session %p, p->p_session %p.\n",
 535                 __func__,
 536                 pgrp->pg_session,
 537                 p->p_session));
 538         KASSERT(pgrp != p->p_pgrp,
 539                 ("%s: p belongs to pgrp.", __func__));
 540
 541         doenterpgrp(p, pgrp);
 542
 543         return (0);
 544 }
 545
 546 /*
 547  * Move p to a process group
 548  */
 549 static void
 550 doenterpgrp(struct proc *p, struct pgrp *pgrp)
 551 {
 552         struct pgrp *savepgrp;
 553
 554         sx_assert(&proctree_lock, SX_XLOCKED);
 555         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 556         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 557         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 558         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 559
 560         savepgrp = p->p_pgrp;
 561
 562         /*
 563          * Adjust eligibility of affected pgrps to participate in job control.
 564          * Increment eligibility counts before decrementing, otherwise we
 565          * could reach 0 spuriously during the first call.
 566          */
 567         fixjobc(p, pgrp, 1);
 568         fixjobc(p, p->p_pgrp, 0);
 569
 570         PGRP_LOCK(pgrp);
 571         PGRP_LOCK(savepgrp);
 572         PROC_LOCK(p);
 573         LIST_REMOVE(p, p_pglist);
 574         p->p_pgrp = pgrp;
 575         PROC_UNLOCK(p);
 576         LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
 577         PGRP_UNLOCK(savepgrp);
 578         PGRP_UNLOCK(pgrp);
 579         if (LIST_EMPTY(&savepgrp->pg_members))
 580                 pgdelete(savepgrp);
 581 }
 582
 583 /*
 584  * remove process from process group
 585  */
 586 int
 587 leavepgrp(struct proc *p)
 588 {
 589         struct pgrp *savepgrp;
 590
 591         sx_assert(&proctree_lock, SX_XLOCKED);
 592         savepgrp = p->p_pgrp;
 593         PGRP_LOCK(savepgrp);
 594         PROC_LOCK(p);
 595         LIST_REMOVE(p, p_pglist);
 596         p->p_pgrp = NULL;
 597         PROC_UNLOCK(p);
 598         PGRP_UNLOCK(savepgrp);
 599         if (LIST_EMPTY(&savepgrp->pg_members))
 600                 pgdelete(savepgrp);
 601         return (0);
 602 }
 603
 604 /*
 605  * delete a process group
 606  */
 607 static void
 608 pgdelete(struct pgrp *pgrp)
 609 {
 610         struct session *savesess;
 611         struct tty *tp;
 612
 613         sx_assert(&proctree_lock, SX_XLOCKED);
 614         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 615         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 616
 617         /*
 618          * Reset any sigio structures pointing to us as a result of
 619          * F_SETOWN with our pgid.
 620          */
 621         funsetownlst(&pgrp->pg_sigiolst);
 622
 623         PGRP_LOCK(pgrp);
 624         tp = pgrp->pg_session->s_ttyp;
 625         LIST_REMOVE(pgrp, pg_hash);
 626         savesess = pgrp->pg_session;
 627         PGRP_UNLOCK(pgrp);
 628
 629         /* Remove the reference to the pgrp before deallocating it. */
 630         if (tp != NULL) {
 631                 tty_lock(tp);
 632                 tty_rel_pgrp(tp, pgrp);
 633         }
 634
 635         mtx_destroy(&pgrp->pg_mtx);
 636         free(pgrp, M_PGRP);
 637         sess_release(savesess);
 638 }
 639
 640 static void
 641 pgadjustjobc(struct pgrp *pgrp, int entering)
 642 {
 643
 644         PGRP_LOCK(pgrp);
 645         if (entering)
 646                 pgrp->pg_jobc++;
 647         else {
 648                 --pgrp->pg_jobc;
 649                 if (pgrp->pg_jobc == 0)
 650                         orphanpg(pgrp);
 651         }
 652         PGRP_UNLOCK(pgrp);
 653 }
 654
 655 /*
 656  * Adjust pgrp jobc counters when specified process changes process group.
 657  * We count the number of processes in each process group that "qualify"
 658  * the group for terminal job control (those with a parent in a different
 659  * process group of the same session).  If that count reaches zero, the
 660  * process group becomes orphaned.  Check both the specified process'
 661  * process group and that of its children.
 662  * entering == 0 => p is leaving specified group.
 663  * entering == 1 => p is entering specified group.
 664  */
 665 void
 666 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
 667 {
 668         struct pgrp *hispgrp;
 669         struct session *mysession;
 670         struct proc *q;
 671
 672         sx_assert(&proctree_lock, SX_LOCKED);
 673         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 674         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 675         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 676
 677         /*
 678          * Check p's parent to see whether p qualifies its own process
 679          * group; if so, adjust count for p's process group.
 680          */
 681         mysession = pgrp->pg_session;
 682         if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
 683             hispgrp->pg_session == mysession)
 684                 pgadjustjobc(pgrp, entering);
 685
 686         /*
 687          * Check this process' children to see whether they qualify
 688          * their process groups; if so, adjust counts for children's
 689          * process groups.
 690          */
 691         LIST_FOREACH(q, &p->p_children, p_sibling) {
 692                 hispgrp = q->p_pgrp;
 693                 if (hispgrp == pgrp ||
 694                     hispgrp->pg_session != mysession)
 695                         continue;
 696                 if (q->p_state == PRS_ZOMBIE)
 697                         continue;
 698                 pgadjustjobc(hispgrp, entering);
 699         }
 700 }
 701
 702 void
 703 killjobc(void)
 704 {
 705         struct session *sp;
 706         struct tty *tp;
 707         struct proc *p;
 708         struct vnode *ttyvp;
 709
 710         p = curproc;
 711         MPASS(p->p_flag & P_WEXIT);
 712         /*
 713          * Do a quick check to see if there is anything to do with the
 714          * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
 715          */
 716         PROC_LOCK(p);
 717         if (!SESS_LEADER(p) &&
 718             (p->p_pgrp == p->p_pptr->p_pgrp) &&
 719             LIST_EMPTY(&p->p_children)) {
 720                 PROC_UNLOCK(p);
 721                 return;
 722         }
 723         PROC_UNLOCK(p);
 724
 725         sx_xlock(&proctree_lock);
 726         if (SESS_LEADER(p)) {
 727                 sp = p->p_session;
 728
 729                 /*
 730                  * s_ttyp is not zero'd; we use this to indicate that
 731                  * the session once had a controlling terminal. (for
 732                  * logging and informational purposes)
 733                  */
 734                 SESS_LOCK(sp);
 735                 ttyvp = sp->s_ttyvp;
 736                 tp = sp->s_ttyp;
 737                 sp->s_ttyvp = NULL;
 738                 sp->s_ttydp = NULL;
 739                 sp->s_leader = NULL;
 740                 SESS_UNLOCK(sp);
 741
 742                 /*
 743                  * Signal foreground pgrp and revoke access to
 744                  * controlling terminal if it has not been revoked
 745                  * already.
 746                  *
 747                  * Because the TTY may have been revoked in the mean
 748                  * time and could already have a new session associated
 749                  * with it, make sure we don't send a SIGHUP to a
 750                  * foreground process group that does not belong to this
 751                  * session.
 752                  */
 753
 754                 if (tp != NULL) {
 755                         tty_lock(tp);
 756                         if (tp->t_session == sp)
 757                                 tty_signal_pgrp(tp, SIGHUP);
 758                         tty_unlock(tp);
 759                 }
 760
 761                 if (ttyvp != NULL) {
 762                         sx_xunlock(&proctree_lock);
 763                         if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
 764                                 VOP_REVOKE(ttyvp, REVOKEALL);
 765                                 VOP_UNLOCK(ttyvp, 0);
 766                         }
 767                         vrele(ttyvp);
 768                         sx_xlock(&proctree_lock);
 769                 }
 770         }
 771         fixjobc(p, p->p_pgrp, 0);
 772         sx_xunlock(&proctree_lock);
 773 }
 774
 775 /*
 776  * A process group has become orphaned;
 777  * if there are any stopped processes in the group,
 778  * hang-up all process in that group.
 779  */
 780 static void
 781 orphanpg(struct pgrp *pg)
 782 {
 783         struct proc *p;
 784
 785         PGRP_LOCK_ASSERT(pg, MA_OWNED);
 786
 787         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 788                 PROC_LOCK(p);
 789                 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
 790                         PROC_UNLOCK(p);
 791                         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 792                                 PROC_LOCK(p);
 793                                 kern_psignal(p, SIGHUP);
 794                                 kern_psignal(p, SIGCONT);
 795                                 PROC_UNLOCK(p);
 796                         }
 797                         return;
 798                 }
 799                 PROC_UNLOCK(p);
 800         }
 801 }
 802
 803 void
 804 sess_hold(struct session *s)
 805 {
 806
 807         refcount_acquire(&s->s_count);
 808 }
 809
 810 void
 811 sess_release(struct session *s)
 812 {
 813
 814         if (refcount_release(&s->s_count)) {
 815                 if (s->s_ttyp != NULL) {
 816                         tty_lock(s->s_ttyp);
 817                         tty_rel_sess(s->s_ttyp, s);
 818                 }
 819                 mtx_destroy(&s->s_mtx);
 820                 free(s, M_SESSION);
 821         }
 822 }
 823
 824 #ifdef DDB
 825
 826 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
 827 {
 828         struct pgrp *pgrp;
 829         struct proc *p;
 830         int i;
 831
 832         for (i = 0; i <= pgrphash; i++) {
 833                 if (!LIST_EMPTY(&pgrphashtbl[i])) {
 834                         printf("\tindx %d\n", i);
 835                         LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
 836                                 printf(
 837                         "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
 838                                     (void *)pgrp, (long)pgrp->pg_id,
 839                                     (void *)pgrp->pg_session,
 840                                     pgrp->pg_session->s_count,
 841                                     (void *)LIST_FIRST(&pgrp->pg_members));
 842                                 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
 843                                         printf("\t\tpid %ld addr %p pgrp %p\n",
 844                                             (long)p->p_pid, (void *)p,
 845                                             (void *)p->p_pgrp);
 846                                 }
 847                         }
 848                 }
 849         }
 850 }
 851 #endif /* DDB */
 852
 853 /*
 854  * Calculate the kinfo_proc members which contain process-wide
 855  * informations.
 856  * Must be called with the target process locked.
 857  */
 858 static void
 859 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
 860 {
 861         struct thread *td;
 862
 863         PROC_LOCK_ASSERT(p, MA_OWNED);
 864
 865         kp->ki_estcpu = 0;
 866         kp->ki_pctcpu = 0;
 867         FOREACH_THREAD_IN_PROC(p, td) {
 868                 thread_lock(td);
 869                 kp->ki_pctcpu += sched_pctcpu(td);
 870                 kp->ki_estcpu += sched_estcpu(td);
 871                 thread_unlock(td);
 872         }
 873 }
 874
 875 /*
 876  * Clear kinfo_proc and fill in any information that is common
 877  * to all threads in the process.
 878  * Must be called with the target process locked.
 879  */
 880 static void
 881 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
 882 {
 883         struct thread *td0;
 884         struct tty *tp;
 885         struct session *sp;
 886         struct ucred *cred;
 887         struct sigacts *ps;
 888         struct timeval boottime;
 889
 890         /* For proc_realparent. */
 891         sx_assert(&proctree_lock, SX_LOCKED);
 892         PROC_LOCK_ASSERT(p, MA_OWNED);
 893         bzero(kp, sizeof(*kp));
 894
 895         kp->ki_structsize = sizeof(*kp);
 896         kp->ki_paddr = p;
 897         kp->ki_addr =/* p->p_addr; */0; /* XXX */
 898         kp->ki_args = p->p_args;
 899         kp->ki_textvp = p->p_textvp;
 900 #ifdef KTRACE
 901         kp->ki_tracep = p->p_tracevp;
 902         kp->ki_traceflag = p->p_traceflag;
 903 #endif
 904         kp->ki_fd = p->p_fd;
 905         kp->ki_vmspace = p->p_vmspace;
 906         kp->ki_flag = p->p_flag;
 907         kp->ki_flag2 = p->p_flag2;
 908         cred = p->p_ucred;
 909         if (cred) {
 910                 kp->ki_uid = cred->cr_uid;
 911                 kp->ki_ruid = cred->cr_ruid;
 912                 kp->ki_svuid = cred->cr_svuid;
 913                 kp->ki_cr_flags = 0;
 914                 if (cred->cr_flags & CRED_FLAG_CAPMODE)
 915                         kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
 916                 /* XXX bde doesn't like KI_NGROUPS */
 917                 if (cred->cr_ngroups > KI_NGROUPS) {
 918                         kp->ki_ngroups = KI_NGROUPS;
 919                         kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
 920                 } else
 921                         kp->ki_ngroups = cred->cr_ngroups;
 922                 bcopy(cred->cr_groups, kp->ki_groups,
 923                     kp->ki_ngroups * sizeof(gid_t));
 924                 kp->ki_rgid = cred->cr_rgid;
 925                 kp->ki_svgid = cred->cr_svgid;
 926                 /* If jailed(cred), emulate the old P_JAILED flag. */
 927                 if (jailed(cred)) {
 928                         kp->ki_flag |= P_JAILED;
 929                         /* If inside the jail, use 0 as a jail ID. */
 930                         if (cred->cr_prison != curthread->td_ucred->cr_prison)
 931                                 kp->ki_jid = cred->cr_prison->pr_id;
 932                 }
 933                 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
 934                     sizeof(kp->ki_loginclass));
 935         }
 936         ps = p->p_sigacts;
 937         if (ps) {
 938                 mtx_lock(&ps->ps_mtx);
 939                 kp->ki_sigignore = ps->ps_sigignore;
 940                 kp->ki_sigcatch = ps->ps_sigcatch;
 941                 mtx_unlock(&ps->ps_mtx);
 942         }
 943         if (p->p_state != PRS_NEW &&
 944             p->p_state != PRS_ZOMBIE &&
 945             p->p_vmspace != NULL) {
 946                 struct vmspace *vm = p->p_vmspace;
 947
 948                 kp->ki_size = vm->vm_map.size;
 949                 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
 950                 FOREACH_THREAD_IN_PROC(p, td0) {
 951                         if (!TD_IS_SWAPPED(td0))
 952                                 kp->ki_rssize += td0->td_kstack_pages;
 953                 }
 954                 kp->ki_swrss = vm->vm_swrss;
 955                 kp->ki_tsize = vm->vm_tsize;
 956                 kp->ki_dsize = vm->vm_dsize;
 957                 kp->ki_ssize = vm->vm_ssize;
 958         } else if (p->p_state == PRS_ZOMBIE)
 959                 kp->ki_stat = SZOMB;
 960         if (kp->ki_flag & P_INMEM)
 961                 kp->ki_sflag = PS_INMEM;
 962         else
 963                 kp->ki_sflag = 0;
 964         /* Calculate legacy swtime as seconds since 'swtick'. */
 965         kp->ki_swtime = (ticks - p->p_swtick) / hz;
 966         kp->ki_pid = p->p_pid;
 967         kp->ki_nice = p->p_nice;
 968         kp->ki_fibnum = p->p_fibnum;
 969         kp->ki_start = p->p_stats->p_start;
 970         getboottime(&boottime);
 971         timevaladd(&kp->ki_start, &boottime);
 972         PROC_STATLOCK(p);
 973         rufetch(p, &kp->ki_rusage);
 974         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
 975         calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
 976         PROC_STATUNLOCK(p);
 977         calccru(p, &kp->ki_childutime, &kp->ki_childstime);
 978         /* Some callers want child times in a single value. */
 979         kp->ki_childtime = kp->ki_childstime;
 980         timevaladd(&kp->ki_childtime, &kp->ki_childutime);
 981
 982         FOREACH_THREAD_IN_PROC(p, td0)
 983                 kp->ki_cow += td0->td_cow;
 984
 985         tp = NULL;
 986         if (p->p_pgrp) {
 987                 kp->ki_pgid = p->p_pgrp->pg_id;
 988                 kp->ki_jobc = p->p_pgrp->pg_jobc;
 989                 sp = p->p_pgrp->pg_session;
 990
 991                 if (sp != NULL) {
 992                         kp->ki_sid = sp->s_sid;
 993                         SESS_LOCK(sp);
 994                         strlcpy(kp->ki_login, sp->s_login,
 995                             sizeof(kp->ki_login));
 996                         if (sp->s_ttyvp)
 997                                 kp->ki_kiflag |= KI_CTTY;
 998                         if (SESS_LEADER(p))
 999                                 kp->ki_kiflag |= KI_SLEADER;
1000                         /* XXX proctree_lock */
1001                         tp = sp->s_ttyp;
1002                         SESS_UNLOCK(sp);
1003                 }
1004         }
1005         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1006                 kp->ki_tdev = tty_udev(tp);
1007                 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1008                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1009                 if (tp->t_session)
1010                         kp->ki_tsid = tp->t_session->s_sid;
1011         } else {
1012                 kp->ki_tdev = NODEV;
1013                 kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1014         }
1015         if (p->p_comm[0] != '\0')
1016                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1017         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1018             p->p_sysent->sv_name[0] != '\0')
1019                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1020         kp->ki_siglist = p->p_siglist;
1021         kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1022         kp->ki_acflag = p->p_acflag;
1023         kp->ki_lock = p->p_lock;
1024         if (p->p_pptr) {
1025                 kp->ki_ppid = proc_realparent(p)->p_pid;
1026                 if (p->p_flag & P_TRACED)
1027                         kp->ki_tracer = p->p_pptr->p_pid;
1028         }
1029 }
1030
1031 /*
1032  * Fill in information that is thread specific.  Must be called with
1033  * target process locked.  If 'preferthread' is set, overwrite certain
1034  * process-related fields that are maintained for both threads and
1035  * processes.
1036  */
1037 static void
1038 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1039 {
1040         struct proc *p;
1041
1042         p = td->td_proc;
1043         kp->ki_tdaddr = td;
1044         PROC_LOCK_ASSERT(p, MA_OWNED);
1045
1046         if (preferthread)
1047                 PROC_STATLOCK(p);
1048         thread_lock(td);
1049         if (td->td_wmesg != NULL)
1050                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1051         else
1052                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1053         if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1054             sizeof(kp->ki_tdname)) {
1055                 strlcpy(kp->ki_moretdname,
1056                     td->td_name + sizeof(kp->ki_tdname) - 1,
1057                     sizeof(kp->ki_moretdname));
1058         } else {
1059                 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1060         }
1061         if (TD_ON_LOCK(td)) {
1062                 kp->ki_kiflag |= KI_LOCKBLOCK;
1063                 strlcpy(kp->ki_lockname, td->td_lockname,
1064                     sizeof(kp->ki_lockname));
1065         } else {
1066                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
1067                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1068         }
1069
1070         if (p->p_state == PRS_NORMAL) { /* approximate. */
1071                 if (TD_ON_RUNQ(td) ||
1072                     TD_CAN_RUN(td) ||
1073                     TD_IS_RUNNING(td)) {
1074                         kp->ki_stat = SRUN;
1075                 } else if (P_SHOULDSTOP(p)) {
1076                         kp->ki_stat = SSTOP;
1077                 } else if (TD_IS_SLEEPING(td)) {
1078                         kp->ki_stat = SSLEEP;
1079                 } else if (TD_ON_LOCK(td)) {
1080                         kp->ki_stat = SLOCK;
1081                 } else {
1082                         kp->ki_stat = SWAIT;
1083                 }
1084         } else if (p->p_state == PRS_ZOMBIE) {
1085                 kp->ki_stat = SZOMB;
1086         } else {
1087                 kp->ki_stat = SIDL;
1088         }
1089
1090         /* Things in the thread */
1091         kp->ki_wchan = td->td_wchan;
1092         kp->ki_pri.pri_level = td->td_priority;
1093         kp->ki_pri.pri_native = td->td_base_pri;
1094
1095         /*
1096          * Note: legacy fields; clamp at the old NOCPU value and/or
1097          * the maximum u_char CPU value.
1098          */
1099         if (td->td_lastcpu == NOCPU)
1100                 kp->ki_lastcpu_old = NOCPU_OLD;
1101         else if (td->td_lastcpu > MAXCPU_OLD)
1102                 kp->ki_lastcpu_old = MAXCPU_OLD;
1103         else
1104                 kp->ki_lastcpu_old = td->td_lastcpu;
1105
1106         if (td->td_oncpu == NOCPU)
1107                 kp->ki_oncpu_old = NOCPU_OLD;
1108         else if (td->td_oncpu > MAXCPU_OLD)
1109                 kp->ki_oncpu_old = MAXCPU_OLD;
1110         else
1111                 kp->ki_oncpu_old = td->td_oncpu;
1112
1113         kp->ki_lastcpu = td->td_lastcpu;
1114         kp->ki_oncpu = td->td_oncpu;
1115         kp->ki_tdflags = td->td_flags;
1116         kp->ki_tid = td->td_tid;
1117         kp->ki_numthreads = p->p_numthreads;
1118         kp->ki_pcb = td->td_pcb;
1119         kp->ki_kstack = (void *)td->td_kstack;
1120         kp->ki_slptime = (ticks - td->td_slptick) / hz;
1121         kp->ki_pri.pri_class = td->td_pri_class;
1122         kp->ki_pri.pri_user = td->td_user_pri;
1123
1124         if (preferthread) {
1125                 rufetchtd(td, &kp->ki_rusage);
1126                 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1127                 kp->ki_pctcpu = sched_pctcpu(td);
1128                 kp->ki_estcpu = sched_estcpu(td);
1129                 kp->ki_cow = td->td_cow;
1130         }
1131
1132         /* We can't get this anymore but ps etc never used it anyway. */
1133         kp->ki_rqindex = 0;
1134
1135         if (preferthread)
1136                 kp->ki_siglist = td->td_siglist;
1137         kp->ki_sigmask = td->td_sigmask;
1138         thread_unlock(td);
1139         if (preferthread)
1140                 PROC_STATUNLOCK(p);
1141 }
1142
1143 /*
1144  * Fill in a kinfo_proc structure for the specified process.
1145  * Must be called with the target process locked.
1146  */
1147 void
1148 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1149 {
1150
1151         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1152
1153         fill_kinfo_proc_only(p, kp);
1154         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1155         fill_kinfo_aggregate(p, kp);
1156 }
1157
1158 struct pstats *
1159 pstats_alloc(void)
1160 {
1161
1162         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1163 }
1164
1165 /*
1166  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1167  */
1168 void
1169 pstats_fork(struct pstats *src, struct pstats *dst)
1170 {
1171
1172         bzero(&dst->pstat_startzero,
1173             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1174         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1175             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1176 }
1177
1178 void
1179 pstats_free(struct pstats *ps)
1180 {
1181
1182         free(ps, M_SUBPROC);
1183 }
1184
1185 static struct proc *
1186 zpfind_locked(pid_t pid)
1187 {
1188         struct proc *p;
1189
1190         sx_assert(&allproc_lock, SX_LOCKED);
1191         LIST_FOREACH(p, &zombproc, p_list) {
1192                 if (p->p_pid == pid) {
1193                         PROC_LOCK(p);
1194                         break;
1195                 }
1196         }
1197         return (p);
1198 }
1199
1200 /*
1201  * Locate a zombie process by number
1202  */
1203 struct proc *
1204 zpfind(pid_t pid)
1205 {
1206         struct proc *p;
1207
1208         sx_slock(&allproc_lock);
1209         p = zpfind_locked(pid);
1210         sx_sunlock(&allproc_lock);
1211         return (p);
1212 }
1213
1214 #ifdef COMPAT_FREEBSD32
1215
1216 /*
1217  * This function is typically used to copy out the kernel address, so
1218  * it can be replaced by assignment of zero.
1219  */
1220 static inline uint32_t
1221 ptr32_trim(void *ptr)
1222 {
1223         uintptr_t uptr;
1224
1225         uptr = (uintptr_t)ptr;
1226         return ((uptr > UINT_MAX) ? 0 : uptr);
1227 }
1228
1229 #define PTRTRIM_CP(src,dst,fld) \
1230         do { (dst).fld = ptr32_trim((src).fld); } while (0)
1231
1232 static void
1233 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1234 {
1235         int i;
1236
1237         bzero(ki32, sizeof(struct kinfo_proc32));
1238         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1239         CP(*ki, *ki32, ki_layout);
1240         PTRTRIM_CP(*ki, *ki32, ki_args);
1241         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1242         PTRTRIM_CP(*ki, *ki32, ki_addr);
1243         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1244         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1245         PTRTRIM_CP(*ki, *ki32, ki_fd);
1246         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1247         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1248         CP(*ki, *ki32, ki_pid);
1249         CP(*ki, *ki32, ki_ppid);
1250         CP(*ki, *ki32, ki_pgid);
1251         CP(*ki, *ki32, ki_tpgid);
1252         CP(*ki, *ki32, ki_sid);
1253         CP(*ki, *ki32, ki_tsid);
1254         CP(*ki, *ki32, ki_jobc);
1255         CP(*ki, *ki32, ki_tdev);
1256         CP(*ki, *ki32, ki_tdev_freebsd11);
1257         CP(*ki, *ki32, ki_siglist);
1258         CP(*ki, *ki32, ki_sigmask);
1259         CP(*ki, *ki32, ki_sigignore);
1260         CP(*ki, *ki32, ki_sigcatch);
1261         CP(*ki, *ki32, ki_uid);
1262         CP(*ki, *ki32, ki_ruid);
1263         CP(*ki, *ki32, ki_svuid);
1264         CP(*ki, *ki32, ki_rgid);
1265         CP(*ki, *ki32, ki_svgid);
1266         CP(*ki, *ki32, ki_ngroups);
1267         for (i = 0; i < KI_NGROUPS; i++)
1268                 CP(*ki, *ki32, ki_groups[i]);
1269         CP(*ki, *ki32, ki_size);
1270         CP(*ki, *ki32, ki_rssize);
1271         CP(*ki, *ki32, ki_swrss);
1272         CP(*ki, *ki32, ki_tsize);
1273         CP(*ki, *ki32, ki_dsize);
1274         CP(*ki, *ki32, ki_ssize);
1275         CP(*ki, *ki32, ki_xstat);
1276         CP(*ki, *ki32, ki_acflag);
1277         CP(*ki, *ki32, ki_pctcpu);
1278         CP(*ki, *ki32, ki_estcpu);
1279         CP(*ki, *ki32, ki_slptime);
1280         CP(*ki, *ki32, ki_swtime);
1281         CP(*ki, *ki32, ki_cow);
1282         CP(*ki, *ki32, ki_runtime);
1283         TV_CP(*ki, *ki32, ki_start);
1284         TV_CP(*ki, *ki32, ki_childtime);
1285         CP(*ki, *ki32, ki_flag);
1286         CP(*ki, *ki32, ki_kiflag);
1287         CP(*ki, *ki32, ki_traceflag);
1288         CP(*ki, *ki32, ki_stat);
1289         CP(*ki, *ki32, ki_nice);
1290         CP(*ki, *ki32, ki_lock);
1291         CP(*ki, *ki32, ki_rqindex);
1292         CP(*ki, *ki32, ki_oncpu);
1293         CP(*ki, *ki32, ki_lastcpu);
1294
1295         /* XXX TODO: wrap cpu value as appropriate */
1296         CP(*ki, *ki32, ki_oncpu_old);
1297         CP(*ki, *ki32, ki_lastcpu_old);
1298
1299         bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1300         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1301         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1302         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1303         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1304         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1305         bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1306         bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1307         CP(*ki, *ki32, ki_tracer);
1308         CP(*ki, *ki32, ki_flag2);
1309         CP(*ki, *ki32, ki_fibnum);
1310         CP(*ki, *ki32, ki_cr_flags);
1311         CP(*ki, *ki32, ki_jid);
1312         CP(*ki, *ki32, ki_numthreads);
1313         CP(*ki, *ki32, ki_tid);
1314         CP(*ki, *ki32, ki_pri);
1315         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1316         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1317         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1318         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1319         PTRTRIM_CP(*ki, *ki32, ki_udata);
1320         PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1321         CP(*ki, *ki32, ki_sflag);
1322         CP(*ki, *ki32, ki_tdflags);
1323 }
1324 #endif
1325
1326 int
1327 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1328 {
1329         struct thread *td;
1330         struct kinfo_proc ki;
1331 #ifdef COMPAT_FREEBSD32
1332         struct kinfo_proc32 ki32;
1333 #endif
1334         int error;
1335
1336         PROC_LOCK_ASSERT(p, MA_OWNED);
1337         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1338
1339         error = 0;
1340         fill_kinfo_proc(p, &ki);
1341         if ((flags & KERN_PROC_NOTHREADS) != 0) {
1342 #ifdef COMPAT_FREEBSD32
1343                 if ((flags & KERN_PROC_MASK32) != 0) {
1344                         freebsd32_kinfo_proc_out(&ki, &ki32);
1345                         if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1346                                 error = ENOMEM;
1347                 } else
1348 #endif
1349                         if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1350                                 error = ENOMEM;
1351         } else {
1352                 FOREACH_THREAD_IN_PROC(p, td) {
1353                         fill_kinfo_thread(td, &ki, 1);
1354 #ifdef COMPAT_FREEBSD32
1355                         if ((flags & KERN_PROC_MASK32) != 0) {
1356                                 freebsd32_kinfo_proc_out(&ki, &ki32);
1357                                 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1358                                         error = ENOMEM;
1359                         } else
1360 #endif
1361                                 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1362                                         error = ENOMEM;
1363                         if (error != 0)
1364                                 break;
1365                 }
1366         }
1367         PROC_UNLOCK(p);
1368         return (error);
1369 }
1370
1371 static int
1372 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1373     int doingzomb)
1374 {
1375         struct sbuf sb;
1376         struct kinfo_proc ki;
1377         struct proc *np;
1378         int error, error2;
1379         pid_t pid;
1380
1381         pid = p->p_pid;
1382         sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1383         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1384         error = kern_proc_out(p, &sb, flags);
1385         error2 = sbuf_finish(&sb);
1386         sbuf_delete(&sb);
1387         if (error != 0)
1388                 return (error);
1389         else if (error2 != 0)
1390                 return (error2);
1391         if (doingzomb)
1392                 np = zpfind(pid);
1393         else {
1394                 if (pid == 0)
1395                         return (0);
1396                 np = pfind(pid);
1397         }
1398         if (np == NULL)
1399                 return (ESRCH);
1400         if (np != p) {
1401                 PROC_UNLOCK(np);
1402                 return (ESRCH);
1403         }
1404         PROC_UNLOCK(np);
1405         return (0);
1406 }
1407
1408 static int
1409 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1410 {
1411         int *name = (int *)arg1;
1412         u_int namelen = arg2;
1413         struct proc *p;
1414         int flags, doingzomb, oid_number;
1415         int error = 0;
1416
1417         oid_number = oidp->oid_number;
1418         if (oid_number != KERN_PROC_ALL &&
1419             (oid_number & KERN_PROC_INC_THREAD) == 0)
1420                 flags = KERN_PROC_NOTHREADS;
1421         else {
1422                 flags = 0;
1423                 oid_number &= ~KERN_PROC_INC_THREAD;
1424         }
1425 #ifdef COMPAT_FREEBSD32
1426         if (req->flags & SCTL_MASK32)
1427                 flags |= KERN_PROC_MASK32;
1428 #endif
1429         if (oid_number == KERN_PROC_PID) {
1430                 if (namelen != 1)
1431                         return (EINVAL);
1432                 error = sysctl_wire_old_buffer(req, 0);
1433                 if (error)
1434                         return (error);
1435                 sx_slock(&proctree_lock);
1436                 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1437                 if (error == 0)
1438                         error = sysctl_out_proc(p, req, flags, 0);
1439                 sx_sunlock(&proctree_lock);
1440                 return (error);
1441         }
1442
1443         switch (oid_number) {
1444         case KERN_PROC_ALL:
1445                 if (namelen != 0)
1446                         return (EINVAL);
1447                 break;
1448         case KERN_PROC_PROC:
1449                 if (namelen != 0 && namelen != 1)
1450                         return (EINVAL);
1451                 break;
1452         default:
1453                 if (namelen != 1)
1454                         return (EINVAL);
1455                 break;
1456         }
1457
1458         if (!req->oldptr) {
1459                 /* overestimate by 5 procs */
1460                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1461                 if (error)
1462                         return (error);
1463         }
1464         error = sysctl_wire_old_buffer(req, 0);
1465         if (error != 0)
1466                 return (error);
1467         sx_slock(&proctree_lock);
1468         sx_slock(&allproc_lock);
1469         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1470                 if (!doingzomb)
1471                         p = LIST_FIRST(&allproc);
1472                 else
1473                         p = LIST_FIRST(&zombproc);
1474                 for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1475                         /*
1476                          * Skip embryonic processes.
1477                          */
1478                         PROC_LOCK(p);
1479                         if (p->p_state == PRS_NEW) {
1480                                 PROC_UNLOCK(p);
1481                                 continue;
1482                         }
1483                         KASSERT(p->p_ucred != NULL,
1484                             ("process credential is NULL for non-NEW proc"));
1485                         /*
1486                          * Show a user only appropriate processes.
1487                          */
1488                         if (p_cansee(curthread, p)) {
1489                                 PROC_UNLOCK(p);
1490                                 continue;
1491                         }
1492                         /*
1493                          * TODO - make more efficient (see notes below).
1494                          * do by session.
1495                          */
1496                         switch (oid_number) {
1497
1498                         case KERN_PROC_GID:
1499                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1500                                         PROC_UNLOCK(p);
1501                                         continue;
1502                                 }
1503                                 break;
1504
1505                         case KERN_PROC_PGRP:
1506                                 /* could do this by traversing pgrp */
1507                                 if (p->p_pgrp == NULL ||
1508                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1509                                         PROC_UNLOCK(p);
1510                                         continue;
1511                                 }
1512                                 break;
1513
1514                         case KERN_PROC_RGID:
1515                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1516                                         PROC_UNLOCK(p);
1517                                         continue;
1518                                 }
1519                                 break;
1520
1521                         case KERN_PROC_SESSION:
1522                                 if (p->p_session == NULL ||
1523                                     p->p_session->s_sid != (pid_t)name[0]) {
1524                                         PROC_UNLOCK(p);
1525                                         continue;
1526                                 }
1527                                 break;
1528
1529                         case KERN_PROC_TTY:
1530                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1531                                     p->p_session == NULL) {
1532                                         PROC_UNLOCK(p);
1533                                         continue;
1534                                 }
1535                                 /* XXX proctree_lock */
1536                                 SESS_LOCK(p->p_session);
1537                                 if (p->p_session->s_ttyp == NULL ||
1538                                     tty_udev(p->p_session->s_ttyp) !=
1539                                     (dev_t)name[0]) {
1540                                         SESS_UNLOCK(p->p_session);
1541                                         PROC_UNLOCK(p);
1542                                         continue;
1543                                 }
1544                                 SESS_UNLOCK(p->p_session);
1545                                 break;
1546
1547                         case KERN_PROC_UID:
1548                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1549                                         PROC_UNLOCK(p);
1550                                         continue;
1551                                 }
1552                                 break;
1553
1554                         case KERN_PROC_RUID:
1555                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1556                                         PROC_UNLOCK(p);
1557                                         continue;
1558                                 }
1559                                 break;
1560
1561                         case KERN_PROC_PROC:
1562                                 break;
1563
1564                         default:
1565                                 break;
1566
1567                         }
1568
1569                         error = sysctl_out_proc(p, req, flags, doingzomb);
1570                         if (error) {
1571                                 sx_sunlock(&allproc_lock);
1572                                 sx_sunlock(&proctree_lock);
1573                                 return (error);
1574                         }
1575                 }
1576         }
1577         sx_sunlock(&allproc_lock);
1578         sx_sunlock(&proctree_lock);
1579         return (0);
1580 }
1581
1582 struct pargs *
1583 pargs_alloc(int len)
1584 {
1585         struct pargs *pa;
1586
1587         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1588                 M_WAITOK);
1589         refcount_init(&pa->ar_ref, 1);
1590         pa->ar_length = len;
1591         return (pa);
1592 }
1593
1594 static void
1595 pargs_free(struct pargs *pa)
1596 {
1597
1598         free(pa, M_PARGS);
1599 }
1600
1601 void
1602 pargs_hold(struct pargs *pa)
1603 {
1604
1605         if (pa == NULL)
1606                 return;
1607         refcount_acquire(&pa->ar_ref);
1608 }
1609
1610 void
1611 pargs_drop(struct pargs *pa)
1612 {
1613
1614         if (pa == NULL)
1615                 return;
1616         if (refcount_release(&pa->ar_ref))
1617                 pargs_free(pa);
1618 }
1619
1620 static int
1621 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1622     size_t len)
1623 {
1624         ssize_t n;
1625
1626         /*
1627          * This may return a short read if the string is shorter than the chunk
1628          * and is aligned at the end of the page, and the following page is not
1629          * mapped.
1630          */
1631         n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1632         if (n <= 0)
1633                 return (ENOMEM);
1634         return (0);
1635 }
1636
1637 #define PROC_AUXV_MAX   256     /* Safety limit on auxv size. */
1638
1639 enum proc_vector_type {
1640         PROC_ARG,
1641         PROC_ENV,
1642         PROC_AUX,
1643 };
1644
1645 #ifdef COMPAT_FREEBSD32
1646 static int
1647 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1648     size_t *vsizep, enum proc_vector_type type)
1649 {
1650         struct freebsd32_ps_strings pss;
1651         Elf32_Auxinfo aux;
1652         vm_offset_t vptr, ptr;
1653         uint32_t *proc_vector32;
1654         char **proc_vector;
1655         size_t vsize, size;
1656         int i, error;
1657
1658         error = 0;
1659         if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1660             sizeof(pss)) != sizeof(pss))
1661                 return (ENOMEM);
1662         switch (type) {
1663         case PROC_ARG:
1664                 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1665                 vsize = pss.ps_nargvstr;
1666                 if (vsize > ARG_MAX)
1667                         return (ENOEXEC);
1668                 size = vsize * sizeof(int32_t);
1669                 break;
1670         case PROC_ENV:
1671                 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1672                 vsize = pss.ps_nenvstr;
1673                 if (vsize > ARG_MAX)
1674                         return (ENOEXEC);
1675                 size = vsize * sizeof(int32_t);
1676                 break;
1677         case PROC_AUX:
1678                 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1679                     (pss.ps_nenvstr + 1) * sizeof(int32_t);
1680                 if (vptr % 4 != 0)
1681                         return (ENOEXEC);
1682                 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1683                         if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1684                             sizeof(aux))
1685                                 return (ENOMEM);
1686                         if (aux.a_type == AT_NULL)
1687                                 break;
1688                         ptr += sizeof(aux);
1689                 }
1690                 if (aux.a_type != AT_NULL)
1691                         return (ENOEXEC);
1692                 vsize = i + 1;
1693                 size = vsize * sizeof(aux);
1694                 break;
1695         default:
1696                 KASSERT(0, ("Wrong proc vector type: %d", type));
1697                 return (EINVAL);
1698         }
1699         proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1700         if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1701                 error = ENOMEM;
1702                 goto done;
1703         }
1704         if (type == PROC_AUX) {
1705                 *proc_vectorp = (char **)proc_vector32;
1706                 *vsizep = vsize;
1707                 return (0);
1708         }
1709         proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1710         for (i = 0; i < (int)vsize; i++)
1711                 proc_vector[i] = PTRIN(proc_vector32[i]);
1712         *proc_vectorp = proc_vector;
1713         *vsizep = vsize;
1714 done:
1715         free(proc_vector32, M_TEMP);
1716         return (error);
1717 }
1718 #endif
1719
1720 static int
1721 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1722     size_t *vsizep, enum proc_vector_type type)
1723 {
1724         struct ps_strings pss;
1725         Elf_Auxinfo aux;
1726         vm_offset_t vptr, ptr;
1727         char **proc_vector;
1728         size_t vsize, size;
1729         int i;
1730
1731 #ifdef COMPAT_FREEBSD32
1732         if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1733                 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1734 #endif
1735         if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1736             sizeof(pss)) != sizeof(pss))
1737                 return (ENOMEM);
1738         switch (type) {
1739         case PROC_ARG:
1740                 vptr = (vm_offset_t)pss.ps_argvstr;
1741                 vsize = pss.ps_nargvstr;
1742                 if (vsize > ARG_MAX)
1743                         return (ENOEXEC);
1744                 size = vsize * sizeof(char *);
1745                 break;
1746         case PROC_ENV:
1747                 vptr = (vm_offset_t)pss.ps_envstr;
1748                 vsize = pss.ps_nenvstr;
1749                 if (vsize > ARG_MAX)
1750                         return (ENOEXEC);
1751                 size = vsize * sizeof(char *);
1752                 break;
1753         case PROC_AUX:
1754                 /*
1755                  * The aux array is just above env array on the stack. Check
1756                  * that the address is naturally aligned.
1757                  */
1758                 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1759                     * sizeof(char *);
1760 #if __ELF_WORD_SIZE == 64
1761                 if (vptr % sizeof(uint64_t) != 0)
1762 #else
1763                 if (vptr % sizeof(uint32_t) != 0)
1764 #endif
1765                         return (ENOEXEC);
1766                 /*
1767                  * We count the array size reading the aux vectors from the
1768                  * stack until AT_NULL vector is returned.  So (to keep the code
1769                  * simple) we read the process stack twice: the first time here
1770                  * to find the size and the second time when copying the vectors
1771                  * to the allocated proc_vector.
1772                  */
1773                 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1774                         if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1775                             sizeof(aux))
1776                                 return (ENOMEM);
1777                         if (aux.a_type == AT_NULL)
1778                                 break;
1779                         ptr += sizeof(aux);
1780                 }
1781                 /*
1782                  * If the PROC_AUXV_MAX entries are iterated over, and we have
1783                  * not reached AT_NULL, it is most likely we are reading wrong
1784                  * data: either the process doesn't have auxv array or data has
1785                  * been modified. Return the error in this case.
1786                  */
1787                 if (aux.a_type != AT_NULL)
1788                         return (ENOEXEC);
1789                 vsize = i + 1;
1790                 size = vsize * sizeof(aux);
1791                 break;
1792         default:
1793                 KASSERT(0, ("Wrong proc vector type: %d", type));
1794                 return (EINVAL); /* In case we are built without INVARIANTS. */
1795         }
1796         proc_vector = malloc(size, M_TEMP, M_WAITOK);
1797         if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1798                 free(proc_vector, M_TEMP);
1799                 return (ENOMEM);
1800         }
1801         *proc_vectorp = proc_vector;
1802         *vsizep = vsize;
1803
1804         return (0);
1805 }
1806
1807 #define GET_PS_STRINGS_CHUNK_SZ 256     /* Chunk size (bytes) for ps_strings operations. */
1808
1809 static int
1810 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1811     enum proc_vector_type type)
1812 {
1813         size_t done, len, nchr, vsize;
1814         int error, i;
1815         char **proc_vector, *sptr;
1816         char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1817
1818         PROC_ASSERT_HELD(p);
1819
1820         /*
1821          * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1822          */
1823         nchr = 2 * (PATH_MAX + ARG_MAX);
1824
1825         error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1826         if (error != 0)
1827                 return (error);
1828         for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1829                 /*
1830                  * The program may have scribbled into its argv array, e.g. to
1831                  * remove some arguments.  If that has happened, break out
1832                  * before trying to read from NULL.
1833                  */
1834                 if (proc_vector[i] == NULL)
1835                         break;
1836                 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1837                         error = proc_read_string(td, p, sptr, pss_string,
1838                             sizeof(pss_string));
1839                         if (error != 0)
1840                                 goto done;
1841                         len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1842                         if (done + len >= nchr)
1843                                 len = nchr - done - 1;
1844                         sbuf_bcat(sb, pss_string, len);
1845                         if (len != GET_PS_STRINGS_CHUNK_SZ)
1846                                 break;
1847                         done += GET_PS_STRINGS_CHUNK_SZ;
1848                 }
1849                 sbuf_bcat(sb, "", 1);
1850                 done += len + 1;
1851         }
1852 done:
1853         free(proc_vector, M_TEMP);
1854         return (error);
1855 }
1856
1857 int
1858 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1859 {
1860
1861         return (get_ps_strings(curthread, p, sb, PROC_ARG));
1862 }
1863
1864 int
1865 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1866 {
1867
1868         return (get_ps_strings(curthread, p, sb, PROC_ENV));
1869 }
1870
1871 int
1872 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1873 {
1874         size_t vsize, size;
1875         char **auxv;
1876         int error;
1877
1878         error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1879         if (error == 0) {
1880 #ifdef COMPAT_FREEBSD32
1881                 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1882                         size = vsize * sizeof(Elf32_Auxinfo);
1883                 else
1884 #endif
1885                         size = vsize * sizeof(Elf_Auxinfo);
1886                 if (sbuf_bcat(sb, auxv, size) != 0)
1887                         error = ENOMEM;
1888                 free(auxv, M_TEMP);
1889         }
1890         return (error);
1891 }
1892
1893 /*
1894  * This sysctl allows a process to retrieve the argument list or process
1895  * title for another process without groping around in the address space
1896  * of the other process.  It also allow a process to set its own "process
1897  * title to a string of its own choice.
1898  */
1899 static int
1900 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1901 {
1902         int *name = (int *)arg1;
1903         u_int namelen = arg2;
1904         struct pargs *newpa, *pa;
1905         struct proc *p;
1906         struct sbuf sb;
1907         int flags, error = 0, error2;
1908
1909         if (namelen != 1)
1910                 return (EINVAL);
1911
1912         flags = PGET_CANSEE;
1913         if (req->newptr != NULL)
1914                 flags |= PGET_ISCURRENT;
1915         error = pget((pid_t)name[0], flags, &p);
1916         if (error)
1917                 return (error);
1918
1919         pa = p->p_args;
1920         if (pa != NULL) {
1921                 pargs_hold(pa);
1922                 PROC_UNLOCK(p);
1923                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1924                 pargs_drop(pa);
1925         } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1926                 _PHOLD(p);
1927                 PROC_UNLOCK(p);
1928                 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1929                 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1930                 error = proc_getargv(curthread, p, &sb);
1931                 error2 = sbuf_finish(&sb);
1932                 PRELE(p);
1933                 sbuf_delete(&sb);
1934                 if (error == 0 && error2 != 0)
1935                         error = error2;
1936         } else {
1937                 PROC_UNLOCK(p);
1938         }
1939         if (error != 0 || req->newptr == NULL)
1940                 return (error);
1941
1942         if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1943                 return (ENOMEM);
1944         newpa = pargs_alloc(req->newlen);
1945         error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1946         if (error != 0) {
1947                 pargs_free(newpa);
1948                 return (error);
1949         }
1950         PROC_LOCK(p);
1951         pa = p->p_args;
1952         p->p_args = newpa;
1953         PROC_UNLOCK(p);
1954         pargs_drop(pa);
1955         return (0);
1956 }
1957
1958 /*
1959  * This sysctl allows a process to retrieve environment of another process.
1960  */
1961 static int
1962 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1963 {
1964         int *name = (int *)arg1;
1965         u_int namelen = arg2;
1966         struct proc *p;
1967         struct sbuf sb;
1968         int error, error2;
1969
1970         if (namelen != 1)
1971                 return (EINVAL);
1972
1973         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1974         if (error != 0)
1975                 return (error);
1976         if ((p->p_flag & P_SYSTEM) != 0) {
1977                 PRELE(p);
1978                 return (0);
1979         }
1980
1981         sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1982         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1983         error = proc_getenvv(curthread, p, &sb);
1984         error2 = sbuf_finish(&sb);
1985         PRELE(p);
1986         sbuf_delete(&sb);
1987         return (error != 0 ? error : error2);
1988 }
1989
1990 /*
1991  * This sysctl allows a process to retrieve ELF auxiliary vector of
1992  * another process.
1993  */
1994 static int
1995 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1996 {
1997         int *name = (int *)arg1;
1998         u_int namelen = arg2;
1999         struct proc *p;
2000         struct sbuf sb;
2001         int error, error2;
2002
2003         if (namelen != 1)
2004                 return (EINVAL);
2005
2006         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2007         if (error != 0)
2008                 return (error);
2009         if ((p->p_flag & P_SYSTEM) != 0) {
2010                 PRELE(p);
2011                 return (0);
2012         }
2013         sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2014         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2015         error = proc_getauxv(curthread, p, &sb);
2016         error2 = sbuf_finish(&sb);
2017         PRELE(p);
2018         sbuf_delete(&sb);
2019         return (error != 0 ? error : error2);
2020 }
2021
2022 /*
2023  * This sysctl allows a process to retrieve the path of the executable for
2024  * itself or another process.
2025  */
2026 static int
2027 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2028 {
2029         pid_t *pidp = (pid_t *)arg1;
2030         unsigned int arglen = arg2;
2031         struct proc *p;
2032         struct vnode *vp;
2033         char *retbuf, *freebuf;
2034         int error;
2035
2036         if (arglen != 1)
2037                 return (EINVAL);
2038         if (*pidp == -1) {      /* -1 means this process */
2039                 p = req->td->td_proc;
2040         } else {
2041                 error = pget(*pidp, PGET_CANSEE, &p);
2042                 if (error != 0)
2043                         return (error);
2044         }
2045
2046         vp = p->p_textvp;
2047         if (vp == NULL) {
2048                 if (*pidp != -1)
2049                         PROC_UNLOCK(p);
2050                 return (0);
2051         }
2052         vref(vp);
2053         if (*pidp != -1)
2054                 PROC_UNLOCK(p);
2055         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2056         vrele(vp);
2057         if (error)
2058                 return (error);
2059         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2060         free(freebuf, M_TEMP);
2061         return (error);
2062 }
2063
2064 static int
2065 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2066 {
2067         struct proc *p;
2068         char *sv_name;
2069         int *name;
2070         int namelen;
2071         int error;
2072
2073         namelen = arg2;
2074         if (namelen != 1)
2075                 return (EINVAL);
2076
2077         name = (int *)arg1;
2078         error = pget((pid_t)name[0], PGET_CANSEE, &p);
2079         if (error != 0)
2080                 return (error);
2081         sv_name = p->p_sysent->sv_name;
2082         PROC_UNLOCK(p);
2083         return (sysctl_handle_string(oidp, sv_name, 0, req));
2084 }
2085
2086 #ifdef KINFO_OVMENTRY_SIZE
2087 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2088 #endif
2089
2090 #ifdef COMPAT_FREEBSD7
2091 static int
2092 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2093 {
2094         vm_map_entry_t entry, tmp_entry;
2095         unsigned int last_timestamp;
2096         char *fullpath, *freepath;
2097         struct kinfo_ovmentry *kve;
2098         struct vattr va;
2099         struct ucred *cred;
2100         int error, *name;
2101         struct vnode *vp;
2102         struct proc *p;
2103         vm_map_t map;
2104         struct vmspace *vm;
2105
2106         name = (int *)arg1;
2107         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2108         if (error != 0)
2109                 return (error);
2110         vm = vmspace_acquire_ref(p);
2111         if (vm == NULL) {
2112                 PRELE(p);
2113                 return (ESRCH);
2114         }
2115         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2116
2117         map = &vm->vm_map;
2118         vm_map_lock_read(map);
2119         for (entry = map->header.next; entry != &map->header;
2120             entry = entry->next) {
2121                 vm_object_t obj, tobj, lobj;
2122                 vm_offset_t addr;
2123
2124                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2125                         continue;
2126
2127                 bzero(kve, sizeof(*kve));
2128                 kve->kve_structsize = sizeof(*kve);
2129
2130                 kve->kve_private_resident = 0;
2131                 obj = entry->object.vm_object;
2132                 if (obj != NULL) {
2133                         VM_OBJECT_RLOCK(obj);
2134                         if (obj->shadow_count == 1)
2135                                 kve->kve_private_resident =
2136                                     obj->resident_page_count;
2137                 }
2138                 kve->kve_resident = 0;
2139                 addr = entry->start;
2140                 while (addr < entry->end) {
2141                         if (pmap_extract(map->pmap, addr))
2142                                 kve->kve_resident++;
2143                         addr += PAGE_SIZE;
2144                 }
2145
2146                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2147                         if (tobj != obj)
2148                                 VM_OBJECT_RLOCK(tobj);
2149                         if (lobj != obj)
2150                                 VM_OBJECT_RUNLOCK(lobj);
2151                         lobj = tobj;
2152                 }
2153
2154                 kve->kve_start = (void*)entry->start;
2155                 kve->kve_end = (void*)entry->end;
2156                 kve->kve_offset = (off_t)entry->offset;
2157
2158                 if (entry->protection & VM_PROT_READ)
2159                         kve->kve_protection |= KVME_PROT_READ;
2160                 if (entry->protection & VM_PROT_WRITE)
2161                         kve->kve_protection |= KVME_PROT_WRITE;
2162                 if (entry->protection & VM_PROT_EXECUTE)
2163                         kve->kve_protection |= KVME_PROT_EXEC;
2164
2165                 if (entry->eflags & MAP_ENTRY_COW)
2166                         kve->kve_flags |= KVME_FLAG_COW;
2167                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2168                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2169                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2170                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2171
2172                 last_timestamp = map->timestamp;
2173                 vm_map_unlock_read(map);
2174
2175                 kve->kve_fileid = 0;
2176                 kve->kve_fsid = 0;
2177                 freepath = NULL;
2178                 fullpath = "";
2179                 if (lobj) {
2180                         vp = NULL;
2181                         switch (lobj->type) {
2182                         case OBJT_DEFAULT:
2183                                 kve->kve_type = KVME_TYPE_DEFAULT;
2184                                 break;
2185                         case OBJT_VNODE:
2186                                 kve->kve_type = KVME_TYPE_VNODE;
2187                                 vp = lobj->handle;
2188                                 vref(vp);
2189                                 break;
2190                         case OBJT_SWAP:
2191                                 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2192                                         kve->kve_type = KVME_TYPE_VNODE;
2193                                         if ((lobj->flags & OBJ_TMPFS) != 0) {
2194                                                 vp = lobj->un_pager.swp.swp_tmpfs;
2195                                                 vref(vp);
2196                                         }
2197                                 } else {
2198                                         kve->kve_type = KVME_TYPE_SWAP;
2199                                 }
2200                                 break;
2201                         case OBJT_DEVICE:
2202                                 kve->kve_type = KVME_TYPE_DEVICE;
2203                                 break;
2204                         case OBJT_PHYS:
2205                                 kve->kve_type = KVME_TYPE_PHYS;
2206                                 break;
2207                         case OBJT_DEAD:
2208                                 kve->kve_type = KVME_TYPE_DEAD;
2209                                 break;
2210                         case OBJT_SG:
2211                                 kve->kve_type = KVME_TYPE_SG;
2212                                 break;
2213                         default:
2214                                 kve->kve_type = KVME_TYPE_UNKNOWN;
2215                                 break;
2216                         }
2217                         if (lobj != obj)
2218                                 VM_OBJECT_RUNLOCK(lobj);
2219
2220                         kve->kve_ref_count = obj->ref_count;
2221                         kve->kve_shadow_count = obj->shadow_count;
2222                         VM_OBJECT_RUNLOCK(obj);
2223                         if (vp != NULL) {
2224                                 vn_fullpath(curthread, vp, &fullpath,
2225                                     &freepath);
2226                                 cred = curthread->td_ucred;
2227                                 vn_lock(vp, LK_SHARED | LK_RETRY);
2228                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
2229                                         kve->kve_fileid = va.va_fileid;
2230                                         /* truncate */
2231                                         kve->kve_fsid = va.va_fsid;
2232                                 }
2233                                 vput(vp);
2234                         }
2235                 } else {
2236                         kve->kve_type = KVME_TYPE_NONE;
2237                         kve->kve_ref_count = 0;
2238                         kve->kve_shadow_count = 0;
2239                 }
2240
2241                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2242                 if (freepath != NULL)
2243                         free(freepath, M_TEMP);
2244
2245                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2246                 vm_map_lock_read(map);
2247                 if (error)
2248                         break;
2249                 if (last_timestamp != map->timestamp) {
2250                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2251                         entry = tmp_entry;
2252                 }
2253         }
2254         vm_map_unlock_read(map);
2255         vmspace_free(vm);
2256         PRELE(p);
2257         free(kve, M_TEMP);
2258         return (error);
2259 }
2260 #endif  /* COMPAT_FREEBSD7 */
2261
2262 #ifdef KINFO_VMENTRY_SIZE
2263 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2264 #endif
2265
2266 static void
2267 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2268     struct kinfo_vmentry *kve)
2269 {
2270         vm_object_t obj, tobj;
2271         vm_page_t m, m_adv;
2272         vm_offset_t addr;
2273         vm_paddr_t locked_pa;
2274         vm_pindex_t pi, pi_adv, pindex;
2275
2276         locked_pa = 0;
2277         obj = entry->object.vm_object;
2278         addr = entry->start;
2279         m_adv = NULL;
2280         pi = OFF_TO_IDX(entry->offset);
2281         for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2282                 if (m_adv != NULL) {
2283                         m = m_adv;
2284                 } else {
2285                         pi_adv = atop(entry->end - addr);
2286                         pindex = pi;
2287                         for (tobj = obj;; tobj = tobj->backing_object) {
2288                                 m = vm_page_find_least(tobj, pindex);
2289                                 if (m != NULL) {
2290                                         if (m->pindex == pindex)
2291                                                 break;
2292                                         if (pi_adv > m->pindex - pindex) {
2293                                                 pi_adv = m->pindex - pindex;
2294                                                 m_adv = m;
2295                                         }
2296                                 }
2297                                 if (tobj->backing_object == NULL)
2298                                         goto next;
2299                                 pindex += OFF_TO_IDX(tobj->
2300                                     backing_object_offset);
2301                         }
2302                 }
2303                 m_adv = NULL;
2304                 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2305                     (addr & (pagesizes[1] - 1)) == 0 &&
2306                     (pmap_mincore(map->pmap, addr, &locked_pa) &
2307                     MINCORE_SUPER) != 0) {
2308                         kve->kve_flags |= KVME_FLAG_SUPER;
2309                         pi_adv = atop(pagesizes[1]);
2310                 } else {
2311                         /*
2312                          * We do not test the found page on validity.
2313                          * Either the page is busy and being paged in,
2314                          * or it was invalidated.  The first case
2315                          * should be counted as resident, the second
2316                          * is not so clear; we do account both.
2317                          */
2318                         pi_adv = 1;
2319                 }
2320                 kve->kve_resident += pi_adv;
2321 next:;
2322         }
2323         PA_UNLOCK_COND(locked_pa);
2324 }
2325
2326 /*
2327  * Must be called with the process locked and will return unlocked.
2328  */
2329 int
2330 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2331 {
2332         vm_map_entry_t entry, tmp_entry;
2333         struct vattr va;
2334         vm_map_t map;
2335         vm_object_t obj, tobj, lobj;
2336         char *fullpath, *freepath;
2337         struct kinfo_vmentry *kve;
2338         struct ucred *cred;
2339         struct vnode *vp;
2340         struct vmspace *vm;
2341         vm_offset_t addr;
2342         unsigned int last_timestamp;
2343         int error;
2344
2345         PROC_LOCK_ASSERT(p, MA_OWNED);
2346
2347         _PHOLD(p);
2348         PROC_UNLOCK(p);
2349         vm = vmspace_acquire_ref(p);
2350         if (vm == NULL) {
2351                 PRELE(p);
2352                 return (ESRCH);
2353         }
2354         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2355
2356         error = 0;
2357         map = &vm->vm_map;
2358         vm_map_lock_read(map);
2359         for (entry = map->header.next; entry != &map->header;
2360             entry = entry->next) {
2361                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2362                         continue;
2363
2364                 addr = entry->end;
2365                 bzero(kve, sizeof(*kve));
2366                 obj = entry->object.vm_object;
2367                 if (obj != NULL) {
2368                         for (tobj = obj; tobj != NULL;
2369                             tobj = tobj->backing_object) {
2370                                 VM_OBJECT_RLOCK(tobj);
2371                                 lobj = tobj;
2372                         }
2373                         if (obj->backing_object == NULL)
2374                                 kve->kve_private_resident =
2375                                     obj->resident_page_count;
2376                         if (!vmmap_skip_res_cnt)
2377                                 kern_proc_vmmap_resident(map, entry, kve);
2378                         for (tobj = obj; tobj != NULL;
2379                             tobj = tobj->backing_object) {
2380                                 if (tobj != obj && tobj != lobj)
2381                                         VM_OBJECT_RUNLOCK(tobj);
2382                         }
2383                 } else {
2384                         lobj = NULL;
2385                 }
2386
2387                 kve->kve_start = entry->start;
2388                 kve->kve_end = entry->end;
2389                 kve->kve_offset = entry->offset;
2390
2391                 if (entry->protection & VM_PROT_READ)
2392                         kve->kve_protection |= KVME_PROT_READ;
2393                 if (entry->protection & VM_PROT_WRITE)
2394                         kve->kve_protection |= KVME_PROT_WRITE;
2395                 if (entry->protection & VM_PROT_EXECUTE)
2396                         kve->kve_protection |= KVME_PROT_EXEC;
2397
2398                 if (entry->eflags & MAP_ENTRY_COW)
2399                         kve->kve_flags |= KVME_FLAG_COW;
2400                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2401                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2402                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2403                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2404                 if (entry->eflags & MAP_ENTRY_GROWS_UP)
2405                         kve->kve_flags |= KVME_FLAG_GROWS_UP;
2406                 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2407                         kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2408
2409                 last_timestamp = map->timestamp;
2410                 vm_map_unlock_read(map);
2411
2412                 freepath = NULL;
2413                 fullpath = "";
2414                 if (lobj != NULL) {
2415                         vp = NULL;
2416                         switch (lobj->type) {
2417                         case OBJT_DEFAULT:
2418                                 kve->kve_type = KVME_TYPE_DEFAULT;
2419                                 break;
2420                         case OBJT_VNODE:
2421                                 kve->kve_type = KVME_TYPE_VNODE;
2422                                 vp = lobj->handle;
2423                                 vref(vp);
2424                                 break;
2425                         case OBJT_SWAP:
2426                                 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2427                                         kve->kve_type = KVME_TYPE_VNODE;
2428                                         if ((lobj->flags & OBJ_TMPFS) != 0) {
2429                                                 vp = lobj->un_pager.swp.swp_tmpfs;
2430                                                 vref(vp);
2431                                         }
2432                                 } else {
2433                                         kve->kve_type = KVME_TYPE_SWAP;
2434                                 }
2435                                 break;
2436                         case OBJT_DEVICE:
2437                                 kve->kve_type = KVME_TYPE_DEVICE;
2438                                 break;
2439                         case OBJT_PHYS:
2440                                 kve->kve_type = KVME_TYPE_PHYS;
2441                                 break;
2442                         case OBJT_DEAD:
2443                                 kve->kve_type = KVME_TYPE_DEAD;
2444                                 break;
2445                         case OBJT_SG:
2446                                 kve->kve_type = KVME_TYPE_SG;
2447                                 break;
2448                         case OBJT_MGTDEVICE:
2449                                 kve->kve_type = KVME_TYPE_MGTDEVICE;
2450                                 break;
2451                         default:
2452                                 kve->kve_type = KVME_TYPE_UNKNOWN;
2453                                 break;
2454                         }
2455                         if (lobj != obj)
2456                                 VM_OBJECT_RUNLOCK(lobj);
2457
2458                         kve->kve_ref_count = obj->ref_count;
2459                         kve->kve_shadow_count = obj->shadow_count;
2460                         VM_OBJECT_RUNLOCK(obj);
2461                         if (vp != NULL) {
2462                                 vn_fullpath(curthread, vp, &fullpath,
2463                                     &freepath);
2464                                 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2465                                 cred = curthread->td_ucred;
2466                                 vn_lock(vp, LK_SHARED | LK_RETRY);
2467                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
2468                                         kve->kve_vn_fileid = va.va_fileid;
2469                                         kve->kve_vn_fsid = va.va_fsid;
2470                                         kve->kve_vn_fsid_freebsd11 =
2471                                             kve->kve_vn_fsid; /* truncate */
2472                                         kve->kve_vn_mode =
2473                                             MAKEIMODE(va.va_type, va.va_mode);
2474                                         kve->kve_vn_size = va.va_size;
2475                                         kve->kve_vn_rdev = va.va_rdev;
2476                                         kve->kve_vn_rdev_freebsd11 =
2477                                             kve->kve_vn_rdev; /* truncate */
2478                                         kve->kve_status = KF_ATTR_VALID;
2479                                 }
2480                                 vput(vp);
2481                         }
2482                 } else {
2483                         kve->kve_type = KVME_TYPE_NONE;
2484                         kve->kve_ref_count = 0;
2485                         kve->kve_shadow_count = 0;
2486                 }
2487
2488                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2489                 if (freepath != NULL)
2490                         free(freepath, M_TEMP);
2491
2492                 /* Pack record size down */
2493                 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2494                         kve->kve_structsize =
2495                             offsetof(struct kinfo_vmentry, kve_path) +
2496                             strlen(kve->kve_path) + 1;
2497                 else
2498                         kve->kve_structsize = sizeof(*kve);
2499                 kve->kve_structsize = roundup(kve->kve_structsize,
2500                     sizeof(uint64_t));
2501
2502                 /* Halt filling and truncate rather than exceeding maxlen */
2503                 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2504                         error = 0;
2505                         vm_map_lock_read(map);
2506                         break;
2507                 } else if (maxlen != -1)
2508                         maxlen -= kve->kve_structsize;
2509
2510                 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2511                         error = ENOMEM;
2512                 vm_map_lock_read(map);
2513                 if (error != 0)
2514                         break;
2515                 if (last_timestamp != map->timestamp) {
2516                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2517                         entry = tmp_entry;
2518                 }
2519         }
2520         vm_map_unlock_read(map);
2521         vmspace_free(vm);
2522         PRELE(p);
2523         free(kve, M_TEMP);
2524         return (error);
2525 }
2526
2527 static int
2528 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2529 {
2530         struct proc *p;
2531         struct sbuf sb;
2532         int error, error2, *name;
2533
2534         name = (int *)arg1;
2535         sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2536         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2537         error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2538         if (error != 0) {
2539                 sbuf_delete(&sb);
2540                 return (error);
2541         }
2542         error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2543         error2 = sbuf_finish(&sb);
2544         sbuf_delete(&sb);
2545         return (error != 0 ? error : error2);
2546 }
2547
2548 #if defined(STACK) || defined(DDB)
2549 static int
2550 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2551 {
2552         struct kinfo_kstack *kkstp;
2553         int error, i, *name, numthreads;
2554         lwpid_t *lwpidarray;
2555         struct thread *td;
2556         struct stack *st;
2557         struct sbuf sb;
2558         struct proc *p;
2559
2560         name = (int *)arg1;
2561         error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2562         if (error != 0)
2563                 return (error);
2564
2565         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2566         st = stack_create(M_WAITOK);
2567
2568         lwpidarray = NULL;
2569         PROC_LOCK(p);
2570         do {
2571                 if (lwpidarray != NULL) {
2572                         free(lwpidarray, M_TEMP);
2573                         lwpidarray = NULL;
2574                 }
2575                 numthreads = p->p_numthreads;
2576                 PROC_UNLOCK(p);
2577                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2578                     M_WAITOK | M_ZERO);
2579                 PROC_LOCK(p);
2580         } while (numthreads < p->p_numthreads);
2581
2582         /*
2583          * XXXRW: During the below loop, execve(2) and countless other sorts
2584          * of changes could have taken place.  Should we check to see if the
2585          * vmspace has been replaced, or the like, in order to prevent
2586          * giving a snapshot that spans, say, execve(2), with some threads
2587          * before and some after?  Among other things, the credentials could
2588          * have changed, in which case the right to extract debug info might
2589          * no longer be assured.
2590          */
2591         i = 0;
2592         FOREACH_THREAD_IN_PROC(p, td) {
2593                 KASSERT(i < numthreads,
2594                     ("sysctl_kern_proc_kstack: numthreads"));
2595                 lwpidarray[i] = td->td_tid;
2596                 i++;
2597         }
2598         numthreads = i;
2599         for (i = 0; i < numthreads; i++) {
2600                 td = thread_find(p, lwpidarray[i]);
2601                 if (td == NULL) {
2602                         continue;
2603                 }
2604                 bzero(kkstp, sizeof(*kkstp));
2605                 (void)sbuf_new(&sb, kkstp->kkst_trace,
2606                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2607                 thread_lock(td);
2608                 kkstp->kkst_tid = td->td_tid;
2609                 if (TD_IS_SWAPPED(td)) {
2610                         kkstp->kkst_state = KKST_STATE_SWAPPED;
2611                 } else if (TD_IS_RUNNING(td)) {
2612                         if (stack_save_td_running(st, td) == 0)
2613                                 kkstp->kkst_state = KKST_STATE_STACKOK;
2614                         else
2615                                 kkstp->kkst_state = KKST_STATE_RUNNING;
2616                 } else {
2617                         kkstp->kkst_state = KKST_STATE_STACKOK;
2618                         stack_save_td(st, td);
2619                 }
2620                 thread_unlock(td);
2621                 PROC_UNLOCK(p);
2622                 stack_sbuf_print(&sb, st);
2623                 sbuf_finish(&sb);
2624                 sbuf_delete(&sb);
2625                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2626                 PROC_LOCK(p);
2627                 if (error)
2628                         break;
2629         }
2630         _PRELE(p);
2631         PROC_UNLOCK(p);
2632         if (lwpidarray != NULL)
2633                 free(lwpidarray, M_TEMP);
2634         stack_destroy(st);
2635         free(kkstp, M_TEMP);
2636         return (error);
2637 }
2638 #endif
2639
2640 /*
2641  * This sysctl allows a process to retrieve the full list of groups from
2642  * itself or another process.
2643  */
2644 static int
2645 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2646 {
2647         pid_t *pidp = (pid_t *)arg1;
2648         unsigned int arglen = arg2;
2649         struct proc *p;
2650         struct ucred *cred;
2651         int error;
2652
2653         if (arglen != 1)
2654                 return (EINVAL);
2655         if (*pidp == -1) {      /* -1 means this process */
2656                 p = req->td->td_proc;
2657                 PROC_LOCK(p);
2658         } else {
2659                 error = pget(*pidp, PGET_CANSEE, &p);
2660                 if (error != 0)
2661                         return (error);
2662         }
2663
2664         cred = crhold(p->p_ucred);
2665         PROC_UNLOCK(p);
2666
2667         error = SYSCTL_OUT(req, cred->cr_groups,
2668             cred->cr_ngroups * sizeof(gid_t));
2669         crfree(cred);
2670         return (error);
2671 }
2672
2673 /*
2674  * This sysctl allows a process to retrieve or/and set the resource limit for
2675  * another process.
2676  */
2677 static int
2678 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2679 {
2680         int *name = (int *)arg1;
2681         u_int namelen = arg2;
2682         struct rlimit rlim;
2683         struct proc *p;
2684         u_int which;
2685         int flags, error;
2686
2687         if (namelen != 2)
2688                 return (EINVAL);
2689
2690         which = (u_int)name[1];
2691         if (which >= RLIM_NLIMITS)
2692                 return (EINVAL);
2693
2694         if (req->newptr != NULL && req->newlen != sizeof(rlim))
2695                 return (EINVAL);
2696
2697         flags = PGET_HOLD | PGET_NOTWEXIT;
2698         if (req->newptr != NULL)
2699                 flags |= PGET_CANDEBUG;
2700         else
2701                 flags |= PGET_CANSEE;
2702         error = pget((pid_t)name[0], flags, &p);
2703         if (error != 0)
2704                 return (error);
2705
2706         /*
2707          * Retrieve limit.
2708          */
2709         if (req->oldptr != NULL) {
2710                 PROC_LOCK(p);
2711                 lim_rlimit_proc(p, which, &rlim);
2712                 PROC_UNLOCK(p);
2713         }
2714         error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2715         if (error != 0)
2716                 goto errout;
2717
2718         /*
2719          * Set limit.
2720          */
2721         if (req->newptr != NULL) {
2722                 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2723                 if (error == 0)
2724                         error = kern_proc_setrlimit(curthread, p, which, &rlim);
2725         }
2726
2727 errout:
2728         PRELE(p);
2729         return (error);
2730 }
2731
2732 /*
2733  * This sysctl allows a process to retrieve ps_strings structure location of
2734  * another process.
2735  */
2736 static int
2737 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2738 {
2739         int *name = (int *)arg1;
2740         u_int namelen = arg2;
2741         struct proc *p;
2742         vm_offset_t ps_strings;
2743         int error;
2744 #ifdef COMPAT_FREEBSD32
2745         uint32_t ps_strings32;
2746 #endif
2747
2748         if (namelen != 1)
2749                 return (EINVAL);
2750
2751         error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2752         if (error != 0)
2753                 return (error);
2754 #ifdef COMPAT_FREEBSD32
2755         if ((req->flags & SCTL_MASK32) != 0) {
2756                 /*
2757                  * We return 0 if the 32 bit emulation request is for a 64 bit
2758                  * process.
2759                  */
2760                 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2761                     PTROUT(p->p_sysent->sv_psstrings) : 0;
2762                 PROC_UNLOCK(p);
2763                 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2764                 return (error);
2765         }
2766 #endif
2767         ps_strings = p->p_sysent->sv_psstrings;
2768         PROC_UNLOCK(p);
2769         error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2770         return (error);
2771 }
2772
2773 /*
2774  * This sysctl allows a process to retrieve umask of another process.
2775  */
2776 static int
2777 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2778 {
2779         int *name = (int *)arg1;
2780         u_int namelen = arg2;
2781         struct proc *p;
2782         int error;
2783         u_short fd_cmask;
2784         pid_t pid;
2785
2786         if (namelen != 1)
2787                 return (EINVAL);
2788
2789         pid = (pid_t)name[0];
2790         p = curproc;
2791         if (pid == p->p_pid || pid == 0) {
2792                 fd_cmask = p->p_fd->fd_cmask;
2793                 goto out;
2794         }
2795
2796         error = pget(pid, PGET_WANTREAD, &p);
2797         if (error != 0)
2798                 return (error);
2799
2800         fd_cmask = p->p_fd->fd_cmask;
2801         PRELE(p);
2802 out:
2803         error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2804         return (error);
2805 }
2806
2807 /*
2808  * This sysctl allows a process to set and retrieve binary osreldate of
2809  * another process.
2810  */
2811 static int
2812 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2813 {
2814         int *name = (int *)arg1;
2815         u_int namelen = arg2;
2816         struct proc *p;
2817         int flags, error, osrel;
2818
2819         if (namelen != 1)
2820                 return (EINVAL);
2821
2822         if (req->newptr != NULL && req->newlen != sizeof(osrel))
2823                 return (EINVAL);
2824
2825         flags = PGET_HOLD | PGET_NOTWEXIT;
2826         if (req->newptr != NULL)
2827                 flags |= PGET_CANDEBUG;
2828         else
2829                 flags |= PGET_CANSEE;
2830         error = pget((pid_t)name[0], flags, &p);
2831         if (error != 0)
2832                 return (error);
2833
2834         error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2835         if (error != 0)
2836                 goto errout;
2837
2838         if (req->newptr != NULL) {
2839                 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2840                 if (error != 0)
2841                         goto errout;
2842                 if (osrel < 0) {
2843                         error = EINVAL;
2844                         goto errout;
2845                 }
2846                 p->p_osrel = osrel;
2847         }
2848 errout:
2849         PRELE(p);
2850         return (error);
2851 }
2852
2853 static int
2854 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2855 {
2856         int *name = (int *)arg1;
2857         u_int namelen = arg2;
2858         struct proc *p;
2859         struct kinfo_sigtramp kst;
2860         const struct sysentvec *sv;
2861         int error;
2862 #ifdef COMPAT_FREEBSD32
2863         struct kinfo_sigtramp32 kst32;
2864 #endif
2865
2866         if (namelen != 1)
2867                 return (EINVAL);
2868
2869         error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2870         if (error != 0)
2871                 return (error);
2872         sv = p->p_sysent;
2873 #ifdef COMPAT_FREEBSD32
2874         if ((req->flags & SCTL_MASK32) != 0) {
2875                 bzero(&kst32, sizeof(kst32));
2876                 if (SV_PROC_FLAG(p, SV_ILP32)) {
2877                         if (sv->sv_sigcode_base != 0) {
2878                                 kst32.ksigtramp_start = sv->sv_sigcode_base;
2879                                 kst32.ksigtramp_end = sv->sv_sigcode_base +
2880                                     *sv->sv_szsigcode;
2881                         } else {
2882                                 kst32.ksigtramp_start = sv->sv_psstrings -
2883                                     *sv->sv_szsigcode;
2884                                 kst32.ksigtramp_end = sv->sv_psstrings;
2885                         }
2886                 }
2887                 PROC_UNLOCK(p);
2888                 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2889                 return (error);
2890         }
2891 #endif
2892         bzero(&kst, sizeof(kst));
2893         if (sv->sv_sigcode_base != 0) {
2894                 kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2895                 kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2896                     *sv->sv_szsigcode;
2897         } else {
2898                 kst.ksigtramp_start = (char *)sv->sv_psstrings -
2899                     *sv->sv_szsigcode;
2900                 kst.ksigtramp_end = (char *)sv->sv_psstrings;
2901         }
2902         PROC_UNLOCK(p);
2903         error = SYSCTL_OUT(req, &kst, sizeof(kst));
2904         return (error);
2905 }
2906
2907 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2908
2909 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2910         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2911         "Return entire process table");
2912
2913 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2914         sysctl_kern_proc, "Process table");
2915
2916 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2917         sysctl_kern_proc, "Process table");
2918
2919 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2920         sysctl_kern_proc, "Process table");
2921
2922 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2923         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2924
2925 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2926         sysctl_kern_proc, "Process table");
2927
2928 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2929         sysctl_kern_proc, "Process table");
2930
2931 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2932         sysctl_kern_proc, "Process table");
2933
2934 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2935         sysctl_kern_proc, "Process table");
2936
2937 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2938         sysctl_kern_proc, "Return process table, no threads");
2939
2940 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2941         CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2942         sysctl_kern_proc_args, "Process argument list");
2943
2944 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2945         sysctl_kern_proc_env, "Process environment");
2946
2947 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2948         CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2949
2950 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2951         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2952
2953 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2954         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2955         "Process syscall vector name (ABI type)");
2956
2957 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2958         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2959
2960 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2961         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2962
2963 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2964         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2965
2966 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2967         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2968
2969 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2970         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2971
2972 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2973         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2974
2975 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2976         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2977
2978 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2979         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2980
2981 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2982         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2983         "Return process table, no threads");
2984
2985 #ifdef COMPAT_FREEBSD7
2986 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2987         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2988 #endif
2989
2990 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2991         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2992
2993 #if defined(STACK) || defined(DDB)
2994 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2995         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2996 #endif
2997
2998 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2999         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3000
3001 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3002         CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3003         "Process resource limits");
3004
3005 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3006         CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3007         "Process ps_strings location");
3008
3009 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3010         CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3011
3012 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3013         CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3014         "Process binary osreldate");
3015
3016 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3017         CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3018         "Process signal trampoline location");
3019
3020 int allproc_gen;
3021
3022 /*
3023  * stop_all_proc() purpose is to stop all process which have usermode,
3024  * except current process for obvious reasons.  This makes it somewhat
3025  * unreliable when invoked from multithreaded process.  The service
3026  * must not be user-callable anyway.
3027  */
3028 void
3029 stop_all_proc(void)
3030 {
3031         struct proc *cp, *p;
3032         int r, gen;
3033         bool restart, seen_stopped, seen_exiting, stopped_some;
3034
3035         cp = curproc;
3036 allproc_loop:
3037         sx_xlock(&allproc_lock);
3038         gen = allproc_gen;
3039         seen_exiting = seen_stopped = stopped_some = restart = false;
3040         LIST_REMOVE(cp, p_list);
3041         LIST_INSERT_HEAD(&allproc, cp, p_list);
3042         for (;;) {
3043                 p = LIST_NEXT(cp, p_list);
3044                 if (p == NULL)
3045                         break;
3046                 LIST_REMOVE(cp, p_list);
3047                 LIST_INSERT_AFTER(p, cp, p_list);
3048                 PROC_LOCK(p);
3049                 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3050                         PROC_UNLOCK(p);
3051                         continue;
3052                 }
3053                 if ((p->p_flag & P_WEXIT) != 0) {
3054                         seen_exiting = true;
3055                         PROC_UNLOCK(p);
3056                         continue;
3057                 }
3058                 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3059                         /*
3060                          * Stopped processes are tolerated when there
3061                          * are no other processes which might continue
3062                          * them.  P_STOPPED_SINGLE but not
3063                          * P_TOTAL_STOP process still has at least one
3064                          * thread running.
3065                          */
3066                         seen_stopped = true;
3067                         PROC_UNLOCK(p);
3068                         continue;
3069                 }
3070                 _PHOLD(p);
3071                 sx_xunlock(&allproc_lock);
3072                 r = thread_single(p, SINGLE_ALLPROC);
3073                 if (r != 0)
3074                         restart = true;
3075                 else
3076                         stopped_some = true;
3077                 _PRELE(p);
3078                 PROC_UNLOCK(p);
3079                 sx_xlock(&allproc_lock);
3080         }
3081         /* Catch forked children we did not see in iteration. */
3082         if (gen != allproc_gen)
3083                 restart = true;
3084         sx_xunlock(&allproc_lock);
3085         if (restart || stopped_some || seen_exiting || seen_stopped) {
3086                 kern_yield(PRI_USER);
3087                 goto allproc_loop;
3088         }
3089 }
3090
3091 void
3092 resume_all_proc(void)
3093 {
3094         struct proc *cp, *p;
3095
3096         cp = curproc;
3097         sx_xlock(&allproc_lock);
3098         LIST_REMOVE(cp, p_list);
3099         LIST_INSERT_HEAD(&allproc, cp, p_list);
3100         for (;;) {
3101                 p = LIST_NEXT(cp, p_list);
3102                 if (p == NULL)
3103                         break;
3104                 LIST_REMOVE(cp, p_list);
3105                 LIST_INSERT_AFTER(p, cp, p_list);
3106                 PROC_LOCK(p);
3107                 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3108                         sx_xunlock(&allproc_lock);
3109                         _PHOLD(p);
3110                         thread_single_end(p, SINGLE_ALLPROC);
3111                         _PRELE(p);
3112                         PROC_UNLOCK(p);
3113                         sx_xlock(&allproc_lock);
3114                 } else {
3115                         PROC_UNLOCK(p);
3116                 }
3117         }
3118         sx_xunlock(&allproc_lock);
3119 }
3120
3121 /* #define      TOTAL_STOP_DEBUG        1 */
3122 #ifdef TOTAL_STOP_DEBUG
3123 volatile static int ap_resume;
3124 #include <sys/mount.h>
3125
3126 static int
3127 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3128 {
3129         int error, val;
3130
3131         val = 0;
3132         ap_resume = 0;
3133         error = sysctl_handle_int(oidp, &val, 0, req);
3134         if (error != 0 || req->newptr == NULL)
3135                 return (error);
3136         if (val != 0) {
3137                 stop_all_proc();
3138                 syncer_suspend();
3139                 while (ap_resume == 0)
3140                         ;
3141                 syncer_resume();
3142                 resume_all_proc();
3143         }
3144         return (0);
3145 }
3146
3147 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3148     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3149     sysctl_debug_stop_all_proc, "I",
3150     "");
3151 #endif