sys/kern/kern_procctl.c

   1 /*-
   2  * Copyright (c) 2014 John Baldwin
   3  * Copyright (c) 2014 The FreeBSD Foundation
   4  *
   5  * Portions of this software were developed by Konstantin Belousov
   6  * under sponsorship from the FreeBSD Foundation.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 AUTHOR 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
  30 #include <sys/cdefs.h>
  31 __FBSDID("$FreeBSD$");
  32
  33 #include <sys/param.h>
  34 #include <sys/systm.h>
  35 #include <sys/capsicum.h>
  36 #include <sys/lock.h>
  37 #include <sys/mutex.h>
  38 #include <sys/priv.h>
  39 #include <sys/proc.h>
  40 #include <sys/procctl.h>
  41 #include <sys/sx.h>
  42 #include <sys/syscallsubr.h>
  43 #include <sys/sysproto.h>
  44 #include <sys/wait.h>
  45
  46 static int
  47 protect_setchild(struct thread *td, struct proc *p, int flags)
  48 {
  49
  50         PROC_LOCK_ASSERT(p, MA_OWNED);
  51         if (p->p_flag & P_SYSTEM || p_cansched(td, p) != 0)
  52                 return (0);
  53         if (flags & PPROT_SET) {
  54                 p->p_flag |= P_PROTECTED;
  55                 if (flags & PPROT_INHERIT)
  56                         p->p_flag2 |= P2_INHERIT_PROTECTED;
  57         } else {
  58                 p->p_flag &= ~P_PROTECTED;
  59                 p->p_flag2 &= ~P2_INHERIT_PROTECTED;
  60         }
  61         return (1);
  62 }
  63
  64 static int
  65 protect_setchildren(struct thread *td, struct proc *top, int flags)
  66 {
  67         struct proc *p;
  68         int ret;
  69
  70         p = top;
  71         ret = 0;
  72         sx_assert(&proctree_lock, SX_LOCKED);
  73         for (;;) {
  74                 ret |= protect_setchild(td, p, flags);
  75                 PROC_UNLOCK(p);
  76                 /*
  77                  * If this process has children, descend to them next,
  78                  * otherwise do any siblings, and if done with this level,
  79                  * follow back up the tree (but not past top).
  80                  */
  81                 if (!LIST_EMPTY(&p->p_children))
  82                         p = LIST_FIRST(&p->p_children);
  83                 else for (;;) {
  84                         if (p == top) {
  85                                 PROC_LOCK(p);
  86                                 return (ret);
  87                         }
  88                         if (LIST_NEXT(p, p_sibling)) {
  89                                 p = LIST_NEXT(p, p_sibling);
  90                                 break;
  91                         }
  92                         p = p->p_pptr;
  93                 }
  94                 PROC_LOCK(p);
  95         }
  96 }
  97
  98 static int
  99 protect_set(struct thread *td, struct proc *p, int flags)
 100 {
 101         int error, ret;
 102
 103         switch (PPROT_OP(flags)) {
 104         case PPROT_SET:
 105         case PPROT_CLEAR:
 106                 break;
 107         default:
 108                 return (EINVAL);
 109         }
 110
 111         if ((PPROT_FLAGS(flags) & ~(PPROT_DESCEND | PPROT_INHERIT)) != 0)
 112                 return (EINVAL);
 113
 114         error = priv_check(td, PRIV_VM_MADV_PROTECT);
 115         if (error)
 116                 return (error);
 117
 118         if (flags & PPROT_DESCEND)
 119                 ret = protect_setchildren(td, p, flags);
 120         else
 121                 ret = protect_setchild(td, p, flags);
 122         if (ret == 0)
 123                 return (EPERM);
 124         return (0);
 125 }
 126
 127 static int
 128 reap_acquire(struct thread *td, struct proc *p)
 129 {
 130
 131         sx_assert(&proctree_lock, SX_XLOCKED);
 132         if (p != curproc)
 133                 return (EPERM);
 134         if ((p->p_treeflag & P_TREE_REAPER) != 0)
 135                 return (EBUSY);
 136         p->p_treeflag |= P_TREE_REAPER;
 137         /*
 138          * We do not reattach existing children and the whole tree
 139          * under them to us, since p->p_reaper already seen them.
 140          */
 141         return (0);
 142 }
 143
 144 static int
 145 reap_release(struct thread *td, struct proc *p)
 146 {
 147
 148         sx_assert(&proctree_lock, SX_XLOCKED);
 149         if (p != curproc)
 150                 return (EPERM);
 151         if (p == initproc)
 152                 return (EINVAL);
 153         if ((p->p_treeflag & P_TREE_REAPER) == 0)
 154                 return (EINVAL);
 155         reaper_abandon_children(p, false);
 156         return (0);
 157 }
 158
 159 static int
 160 reap_status(struct thread *td, struct proc *p,
 161     struct procctl_reaper_status *rs)
 162 {
 163         struct proc *reap, *p2, *first_p;
 164
 165         sx_assert(&proctree_lock, SX_LOCKED);
 166         bzero(rs, sizeof(*rs));
 167         if ((p->p_treeflag & P_TREE_REAPER) == 0) {
 168                 reap = p->p_reaper;
 169         } else {
 170                 reap = p;
 171                 rs->rs_flags |= REAPER_STATUS_OWNED;
 172         }
 173         if (reap == initproc)
 174                 rs->rs_flags |= REAPER_STATUS_REALINIT;
 175         rs->rs_reaper = reap->p_pid;
 176         rs->rs_descendants = 0;
 177         rs->rs_children = 0;
 178         if (!LIST_EMPTY(&reap->p_reaplist)) {
 179                 first_p = LIST_FIRST(&reap->p_children);
 180                 if (first_p == NULL)
 181                         first_p = LIST_FIRST(&reap->p_reaplist);
 182                 rs->rs_pid = first_p->p_pid;
 183                 LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
 184                         if (proc_realparent(p2) == reap)
 185                                 rs->rs_children++;
 186                         rs->rs_descendants++;
 187                 }
 188         } else {
 189                 rs->rs_pid = -1;
 190         }
 191         return (0);
 192 }
 193
 194 static int
 195 reap_getpids(struct thread *td, struct proc *p, struct procctl_reaper_pids *rp)
 196 {
 197         struct proc *reap, *p2;
 198         struct procctl_reaper_pidinfo *pi, *pip;
 199         u_int i, n;
 200         int error;
 201
 202         sx_assert(&proctree_lock, SX_LOCKED);
 203         PROC_UNLOCK(p);
 204         reap = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
 205         n = i = 0;
 206         error = 0;
 207         LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling)
 208                 n++;
 209         sx_unlock(&proctree_lock);
 210         if (rp->rp_count < n)
 211                 n = rp->rp_count;
 212         pi = malloc(n * sizeof(*pi), M_TEMP, M_WAITOK);
 213         sx_slock(&proctree_lock);
 214         LIST_FOREACH(p2, &reap->p_reaplist, p_reapsibling) {
 215                 if (i == n)
 216                         break;
 217                 pip = &pi[i];
 218                 bzero(pip, sizeof(*pip));
 219                 pip->pi_pid = p2->p_pid;
 220                 pip->pi_subtree = p2->p_reapsubtree;
 221                 pip->pi_flags = REAPER_PIDINFO_VALID;
 222                 if (proc_realparent(p2) == reap)
 223                         pip->pi_flags |= REAPER_PIDINFO_CHILD;
 224                 i++;
 225         }
 226         sx_sunlock(&proctree_lock);
 227         error = copyout(pi, rp->rp_pids, i * sizeof(*pi));
 228         free(pi, M_TEMP);
 229         sx_slock(&proctree_lock);
 230         PROC_LOCK(p);
 231         return (error);
 232 }
 233
 234 static int
 235 reap_kill(struct thread *td, struct proc *p, struct procctl_reaper_kill *rk)
 236 {
 237         struct proc *reap, *p2;
 238         ksiginfo_t ksi;
 239         int error, error1;
 240
 241         sx_assert(&proctree_lock, SX_LOCKED);
 242         if (IN_CAPABILITY_MODE(td))
 243                 return (ECAPMODE);
 244         if (rk->rk_sig <= 0 || rk->rk_sig > _SIG_MAXSIG)
 245                 return (EINVAL);
 246         if ((rk->rk_flags & ~REAPER_KILL_CHILDREN) != 0)
 247                 return (EINVAL);
 248         PROC_UNLOCK(p);
 249         reap = (p->p_treeflag & P_TREE_REAPER) == 0 ? p->p_reaper : p;
 250         ksiginfo_init(&ksi);
 251         ksi.ksi_signo = rk->rk_sig;
 252         ksi.ksi_code = SI_USER;
 253         ksi.ksi_pid = td->td_proc->p_pid;
 254         ksi.ksi_uid = td->td_ucred->cr_ruid;
 255         error = ESRCH;
 256         rk->rk_killed = 0;
 257         rk->rk_fpid = -1;
 258         for (p2 = (rk->rk_flags & REAPER_KILL_CHILDREN) != 0 ?
 259             LIST_FIRST(&reap->p_children) : LIST_FIRST(&reap->p_reaplist);
 260             p2 != NULL;
 261             p2 = (rk->rk_flags & REAPER_KILL_CHILDREN) != 0 ?
 262             LIST_NEXT(p2, p_sibling) : LIST_NEXT(p2, p_reapsibling)) {
 263                 if ((rk->rk_flags & REAPER_KILL_SUBTREE) != 0 &&
 264                     p2->p_reapsubtree != rk->rk_subtree)
 265                         continue;
 266                 PROC_LOCK(p2);
 267                 error1 = p_cansignal(td, p2, rk->rk_sig);
 268                 if (error1 == 0) {
 269                         pksignal(p2, rk->rk_sig, &ksi);
 270                         rk->rk_killed++;
 271                         error = error1;
 272                 } else if (error == ESRCH) {
 273                         error = error1;
 274                         rk->rk_fpid = p2->p_pid;
 275                 }
 276                 PROC_UNLOCK(p2);
 277                 /* Do not end the loop on error, signal everything we can. */
 278         }
 279         PROC_LOCK(p);
 280         return (error);
 281 }
 282
 283 static int
 284 trace_ctl(struct thread *td, struct proc *p, int state)
 285 {
 286
 287         PROC_LOCK_ASSERT(p, MA_OWNED);
 288
 289         /*
 290          * Ktrace changes p_traceflag from or to zero under the
 291          * process lock, so the test does not need to acquire ktrace
 292          * mutex.
 293          */
 294         if ((p->p_flag & P_TRACED) != 0 || p->p_traceflag != 0)
 295                 return (EBUSY);
 296
 297         switch (state) {
 298         case PROC_TRACE_CTL_ENABLE:
 299                 if (td->td_proc != p)
 300                         return (EPERM);
 301                 p->p_flag2 &= ~(P2_NOTRACE | P2_NOTRACE_EXEC);
 302                 break;
 303         case PROC_TRACE_CTL_DISABLE_EXEC:
 304                 p->p_flag2 |= P2_NOTRACE_EXEC | P2_NOTRACE;
 305                 break;
 306         case PROC_TRACE_CTL_DISABLE:
 307                 if ((p->p_flag2 & P2_NOTRACE_EXEC) != 0) {
 308                         KASSERT((p->p_flag2 & P2_NOTRACE) != 0,
 309                             ("dandling P2_NOTRACE_EXEC"));
 310                         if (td->td_proc != p)
 311                                 return (EPERM);
 312                         p->p_flag2 &= ~P2_NOTRACE_EXEC;
 313                 } else {
 314                         p->p_flag2 |= P2_NOTRACE;
 315                 }
 316                 break;
 317         default:
 318                 return (EINVAL);
 319         }
 320         return (0);
 321 }
 322
 323 static int
 324 trace_status(struct thread *td, struct proc *p, int *data)
 325 {
 326
 327         if ((p->p_flag2 & P2_NOTRACE) != 0) {
 328                 KASSERT((p->p_flag & P_TRACED) == 0,
 329                     ("%d traced but tracing disabled", p->p_pid));
 330                 *data = -1;
 331         } else if ((p->p_flag & P_TRACED) != 0) {
 332                 *data = p->p_pptr->p_pid;
 333         } else {
 334                 *data = 0;
 335         }
 336         return (0);
 337 }
 338
 339 #ifndef _SYS_SYSPROTO_H_
 340 struct procctl_args {
 341         idtype_t idtype;
 342         id_t    id;
 343         int     com;
 344         void    *data;
 345 };
 346 #endif
 347 /* ARGSUSED */
 348 int
 349 sys_procctl(struct thread *td, struct procctl_args *uap)
 350 {
 351         void *data;
 352         union {
 353                 struct procctl_reaper_status rs;
 354                 struct procctl_reaper_pids rp;
 355                 struct procctl_reaper_kill rk;
 356         } x;
 357         int error, error1, flags;
 358
 359         switch (uap->com) {
 360         case PROC_SPROTECT:
 361         case PROC_TRACE_CTL:
 362                 error = copyin(uap->data, &flags, sizeof(flags));
 363                 if (error != 0)
 364                         return (error);
 365                 data = &flags;
 366                 break;
 367         case PROC_REAP_ACQUIRE:
 368         case PROC_REAP_RELEASE:
 369                 if (uap->data != NULL)
 370                         return (EINVAL);
 371                 data = NULL;
 372                 break;
 373         case PROC_REAP_STATUS:
 374                 data = &x.rs;
 375                 break;
 376         case PROC_REAP_GETPIDS:
 377                 error = copyin(uap->data, &x.rp, sizeof(x.rp));
 378                 if (error != 0)
 379                         return (error);
 380                 data = &x.rp;
 381                 break;
 382         case PROC_REAP_KILL:
 383                 error = copyin(uap->data, &x.rk, sizeof(x.rk));
 384                 if (error != 0)
 385                         return (error);
 386                 data = &x.rk;
 387                 break;
 388         case PROC_TRACE_STATUS:
 389                 data = &flags;
 390                 break;
 391         default:
 392                 return (EINVAL);
 393         }
 394         error = kern_procctl(td, uap->idtype, uap->id, uap->com, data);
 395         switch (uap->com) {
 396         case PROC_REAP_STATUS:
 397                 if (error == 0)
 398                         error = copyout(&x.rs, uap->data, sizeof(x.rs));
 399                 break;
 400         case PROC_REAP_KILL:
 401                 error1 = copyout(&x.rk, uap->data, sizeof(x.rk));
 402                 if (error == 0)
 403                         error = error1;
 404                 break;
 405         case PROC_TRACE_STATUS:
 406                 if (error == 0)
 407                         error = copyout(&flags, uap->data, sizeof(flags));
 408                 break;
 409         }
 410         return (error);
 411 }
 412
 413 static int
 414 kern_procctl_single(struct thread *td, struct proc *p, int com, void *data)
 415 {
 416
 417         PROC_LOCK_ASSERT(p, MA_OWNED);
 418         switch (com) {
 419         case PROC_SPROTECT:
 420                 return (protect_set(td, p, *(int *)data));
 421         case PROC_REAP_ACQUIRE:
 422                 return (reap_acquire(td, p));
 423         case PROC_REAP_RELEASE:
 424                 return (reap_release(td, p));
 425         case PROC_REAP_STATUS:
 426                 return (reap_status(td, p, data));
 427         case PROC_REAP_GETPIDS:
 428                 return (reap_getpids(td, p, data));
 429         case PROC_REAP_KILL:
 430                 return (reap_kill(td, p, data));
 431         case PROC_TRACE_CTL:
 432                 return (trace_ctl(td, p, *(int *)data));
 433         case PROC_TRACE_STATUS:
 434                 return (trace_status(td, p, data));
 435         default:
 436                 return (EINVAL);
 437         }
 438 }
 439
 440 int
 441 kern_procctl(struct thread *td, idtype_t idtype, id_t id, int com, void *data)
 442 {
 443         struct pgrp *pg;
 444         struct proc *p;
 445         int error, first_error, ok;
 446         bool tree_locked;
 447
 448         switch (com) {
 449         case PROC_REAP_ACQUIRE:
 450         case PROC_REAP_RELEASE:
 451         case PROC_REAP_STATUS:
 452         case PROC_REAP_GETPIDS:
 453         case PROC_REAP_KILL:
 454         case PROC_TRACE_STATUS:
 455                 if (idtype != P_PID)
 456                         return (EINVAL);
 457         }
 458
 459         switch (com) {
 460         case PROC_SPROTECT:
 461         case PROC_REAP_STATUS:
 462         case PROC_REAP_GETPIDS:
 463         case PROC_REAP_KILL:
 464         case PROC_TRACE_CTL:
 465                 sx_slock(&proctree_lock);
 466                 tree_locked = true;
 467                 break;
 468         case PROC_REAP_ACQUIRE:
 469         case PROC_REAP_RELEASE:
 470                 sx_xlock(&proctree_lock);
 471                 tree_locked = true;
 472                 break;
 473         case PROC_TRACE_STATUS:
 474                 tree_locked = false;
 475                 break;
 476         default:
 477                 return (EINVAL);
 478         }
 479
 480         switch (idtype) {
 481         case P_PID:
 482                 p = pfind(id);
 483                 if (p == NULL) {
 484                         error = ESRCH;
 485                         break;
 486                 }
 487                 error = p_cansee(td, p);
 488                 if (error == 0)
 489                         error = kern_procctl_single(td, p, com, data);
 490                 PROC_UNLOCK(p);
 491                 break;
 492         case P_PGID:
 493                 /*
 494                  * Attempt to apply the operation to all members of the
 495                  * group.  Ignore processes in the group that can't be
 496                  * seen.  Ignore errors so long as at least one process is
 497                  * able to complete the request successfully.
 498                  */
 499                 pg = pgfind(id);
 500                 if (pg == NULL) {
 501                         error = ESRCH;
 502                         break;
 503                 }
 504                 PGRP_UNLOCK(pg);
 505                 ok = 0;
 506                 first_error = 0;
 507                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 508                         PROC_LOCK(p);
 509                         if (p->p_state == PRS_NEW || p_cansee(td, p) != 0) {
 510                                 PROC_UNLOCK(p);
 511                                 continue;
 512                         }
 513                         error = kern_procctl_single(td, p, com, data);
 514                         PROC_UNLOCK(p);
 515                         if (error == 0)
 516                                 ok = 1;
 517                         else if (first_error == 0)
 518                                 first_error = error;
 519                 }
 520                 if (ok)
 521                         error = 0;
 522                 else if (first_error != 0)
 523                         error = first_error;
 524                 else
 525                         /*
 526                          * Was not able to see any processes in the
 527                          * process group.
 528                          */
 529                         error = ESRCH;
 530                 break;
 531         default:
 532                 error = EINVAL;
 533                 break;
 534         }
 535         if (tree_locked)
 536                 sx_unlock(&proctree_lock);
 537         return (error);
 538 }