1 /* $NetBSD: kern_proc.c,v 1.155 2009/10/21 21:12:06 rmind Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2006, 2007, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1989, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: kern_proc.c,v 1.155 2009/10/21 21:12:06 rmind Exp $"); 66 67 #include "opt_kstack.h" 68 #include "opt_maxuprc.h" 69 70 #include <sys/param.h> 71 #include <sys/systm.h> 72 #include <sys/kernel.h> 73 #include <sys/proc.h> 74 #include <sys/resourcevar.h> 75 #include <sys/buf.h> 76 #include <sys/acct.h> 77 #include <sys/wait.h> 78 #include <sys/file.h> 79 #include <ufs/ufs/quota.h> 80 #include <sys/uio.h> 81 #include <sys/pool.h> 82 #include <sys/pset.h> 83 #include <sys/mbuf.h> 84 #include <sys/ioctl.h> 85 #include <sys/tty.h> 86 #include <sys/signalvar.h> 87 #include <sys/ras.h> 88 #include <sys/sa.h> 89 #include <sys/savar.h> 90 #include <sys/filedesc.h> 91 #include "sys/syscall_stats.h" 92 #include <sys/kauth.h> 93 #include <sys/sleepq.h> 94 #include <sys/atomic.h> 95 #include <sys/kmem.h> 96 97 #include <uvm/uvm.h> 98 #include <uvm/uvm_extern.h> 99 100 /* 101 * Other process lists 102 */ 103 104 struct proclist allproc; 105 struct proclist zombproc; /* resources have been freed */ 106 107 kmutex_t *proc_lock; 108 109 /* 110 * pid to proc lookup is done by indexing the pid_table array. 111 * Since pid numbers are only allocated when an empty slot 112 * has been found, there is no need to search any lists ever. 113 * (an orphaned pgrp will lock the slot, a session will lock 114 * the pgrp with the same number.) 115 * If the table is too small it is reallocated with twice the 116 * previous size and the entries 'unzipped' into the two halves. 117 * A linked list of free entries is passed through the pt_proc 118 * field of 'free' items - set odd to be an invalid ptr. 119 */ 120 121 struct pid_table { 122 struct proc *pt_proc; 123 struct pgrp *pt_pgrp; 124 }; 125 #if 1 /* strongly typed cast - should be a noop */ 126 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; } 127 #else 128 #define p2u(p) ((uint)p) 129 #endif 130 #define P_VALID(p) (!(p2u(p) & 1)) 131 #define P_NEXT(p) (p2u(p) >> 1) 132 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1)) 133 134 #define INITIAL_PID_TABLE_SIZE (1 << 5) 135 static struct pid_table *pid_table; 136 static uint pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1; 137 static uint pid_alloc_lim; /* max we allocate before growing table */ 138 static uint pid_alloc_cnt; /* number of allocated pids */ 139 140 /* links through free slots - never empty! */ 141 static uint next_free_pt, last_free_pt; 142 static pid_t pid_max = PID_MAX; /* largest value we allocate */ 143 144 /* Components of the first process -- never freed. */ 145 146 extern struct emul emul_netbsd; /* defined in kern_exec.c */ 147 148 struct session session0 = { 149 .s_count = 1, 150 .s_sid = 0, 151 }; 152 struct pgrp pgrp0 = { 153 .pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members), 154 .pg_session = &session0, 155 }; 156 filedesc_t filedesc0; 157 struct cwdinfo cwdi0 = { 158 .cwdi_cmask = CMASK, /* see cmask below */ 159 .cwdi_refcnt = 1, 160 }; 161 struct plimit limit0; 162 struct pstats pstat0; 163 struct vmspace vmspace0; 164 struct sigacts sigacts0; 165 struct turnstile turnstile0; 166 struct proc proc0 = { 167 .p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps), 168 .p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters), 169 .p_nlwps = 1, 170 .p_nrlwps = 1, 171 .p_nlwpid = 1, /* must match lwp0.l_lid */ 172 .p_pgrp = &pgrp0, 173 .p_comm = "system", 174 /* 175 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8) 176 * when they exit. init(8) can easily wait them out for us. 177 */ 178 .p_flag = PK_SYSTEM | PK_NOCLDWAIT, 179 .p_stat = SACTIVE, 180 .p_nice = NZERO, 181 .p_emul = &emul_netbsd, 182 .p_cwdi = &cwdi0, 183 .p_limit = &limit0, 184 .p_fd = &filedesc0, 185 .p_vmspace = &vmspace0, 186 .p_stats = &pstat0, 187 .p_sigacts = &sigacts0, 188 }; 189 struct lwp lwp0 __aligned(MIN_LWP_ALIGNMENT) = { 190 #ifdef LWP0_CPU_INFO 191 .l_cpu = LWP0_CPU_INFO, 192 #endif 193 .l_proc = &proc0, 194 .l_lid = 1, 195 .l_flag = LW_SYSTEM, 196 .l_stat = LSONPROC, 197 .l_ts = &turnstile0, 198 .l_syncobj = &sched_syncobj, 199 .l_refcnt = 1, 200 .l_priority = PRI_USER + NPRI_USER - 1, 201 .l_inheritedprio = -1, 202 .l_class = SCHED_OTHER, 203 .l_psid = PS_NONE, 204 .l_pi_lenders = SLIST_HEAD_INITIALIZER(&lwp0.l_pi_lenders), 205 .l_name = __UNCONST("swapper"), 206 .l_fd = &filedesc0, 207 }; 208 kauth_cred_t cred0; 209 210 extern struct user *proc0paddr; 211 212 int nofile = NOFILE; 213 int maxuprc = MAXUPRC; 214 int cmask = CMASK; 215 216 MALLOC_DEFINE(M_EMULDATA, "emuldata", "Per-process emulation data"); 217 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures"); 218 219 /* 220 * The process list descriptors, used during pid allocation and 221 * by sysctl. No locking on this data structure is needed since 222 * it is completely static. 223 */ 224 const struct proclist_desc proclists[] = { 225 { &allproc }, 226 { &zombproc }, 227 { NULL }, 228 }; 229 230 static struct pgrp * pg_remove(pid_t); 231 static void pg_delete(pid_t); 232 static void orphanpg(struct pgrp *); 233 234 static specificdata_domain_t proc_specificdata_domain; 235 236 static pool_cache_t proc_cache; 237 238 static kauth_listener_t proc_listener; 239 240 static int 241 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 242 void *arg0, void *arg1, void *arg2, void *arg3) 243 { 244 struct proc *p; 245 int result; 246 247 result = KAUTH_RESULT_DEFER; 248 p = arg0; 249 250 switch (action) { 251 case KAUTH_PROCESS_CANSEE: { 252 enum kauth_process_req req; 253 254 req = (enum kauth_process_req)arg1; 255 256 switch (req) { 257 case KAUTH_REQ_PROCESS_CANSEE_ARGS: 258 case KAUTH_REQ_PROCESS_CANSEE_ENTRY: 259 case KAUTH_REQ_PROCESS_CANSEE_OPENFILES: 260 result = KAUTH_RESULT_ALLOW; 261 262 break; 263 264 case KAUTH_REQ_PROCESS_CANSEE_ENV: 265 if (kauth_cred_getuid(cred) != 266 kauth_cred_getuid(p->p_cred) || 267 kauth_cred_getuid(cred) != 268 kauth_cred_getsvuid(p->p_cred)) 269 break; 270 271 result = KAUTH_RESULT_ALLOW; 272 273 break; 274 275 default: 276 break; 277 } 278 279 break; 280 } 281 282 case KAUTH_PROCESS_FORK: { 283 int lnprocs = (int)(unsigned long)arg2; 284 285 /* 286 * Don't allow a nonprivileged user to use the last few 287 * processes. The variable lnprocs is the current number of 288 * processes, maxproc is the limit. 289 */ 290 if (__predict_false((lnprocs >= maxproc - 5))) 291 break; 292 293 result = KAUTH_RESULT_ALLOW; 294 295 break; 296 } 297 298 case KAUTH_PROCESS_CORENAME: 299 case KAUTH_PROCESS_STOPFLAG: 300 if (proc_uidmatch(cred, p->p_cred) == 0) 301 result = KAUTH_RESULT_ALLOW; 302 303 break; 304 305 default: 306 break; 307 } 308 309 return result; 310 } 311 312 /* 313 * Initialize global process hashing structures. 314 */ 315 void 316 procinit(void) 317 { 318 const struct proclist_desc *pd; 319 u_int i; 320 #define LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1)) 321 322 for (pd = proclists; pd->pd_list != NULL; pd++) 323 LIST_INIT(pd->pd_list); 324 325 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 326 pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE 327 * sizeof(struct pid_table), KM_SLEEP); 328 329 /* Set free list running through table... 330 Preset 'use count' above PID_MAX so we allocate pid 1 next. */ 331 for (i = 0; i <= pid_tbl_mask; i++) { 332 pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1); 333 pid_table[i].pt_pgrp = 0; 334 } 335 /* slot 0 is just grabbed */ 336 next_free_pt = 1; 337 /* Need to fix last entry. */ 338 last_free_pt = pid_tbl_mask; 339 pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY); 340 /* point at which we grow table - to avoid reusing pids too often */ 341 pid_alloc_lim = pid_tbl_mask - 1; 342 #undef LINK_EMPTY 343 344 proc_specificdata_domain = specificdata_domain_create(); 345 KASSERT(proc_specificdata_domain != NULL); 346 347 proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0, 348 "procpl", NULL, IPL_NONE, NULL, NULL, NULL); 349 350 proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 351 proc_listener_cb, NULL); 352 } 353 354 /* 355 * Initialize process 0. 356 */ 357 void 358 proc0_init(void) 359 { 360 struct proc *p; 361 struct pgrp *pg; 362 struct session *sess; 363 struct lwp *l; 364 rlim_t lim; 365 int i; 366 367 p = &proc0; 368 pg = &pgrp0; 369 sess = &session0; 370 l = &lwp0; 371 372 KASSERT(l->l_lid == p->p_nlwpid); 373 374 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 375 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 376 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 377 378 rw_init(&p->p_reflock); 379 cv_init(&p->p_waitcv, "wait"); 380 cv_init(&p->p_lwpcv, "lwpwait"); 381 382 LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling); 383 384 pid_table[0].pt_proc = p; 385 LIST_INSERT_HEAD(&allproc, p, p_list); 386 LIST_INSERT_HEAD(&alllwp, l, l_list); 387 388 pid_table[0].pt_pgrp = pg; 389 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist); 390 391 #ifdef __HAVE_SYSCALL_INTERN 392 (*p->p_emul->e_syscall_intern)(p); 393 #endif 394 395 callout_init(&l->l_timeout_ch, CALLOUT_MPSAFE); 396 callout_setfunc(&l->l_timeout_ch, sleepq_timeout, l); 397 cv_init(&l->l_sigcv, "sigwait"); 398 399 /* Create credentials. */ 400 cred0 = kauth_cred_alloc(); 401 p->p_cred = cred0; 402 kauth_cred_hold(cred0); 403 l->l_cred = cred0; 404 405 /* Create the CWD info. */ 406 rw_init(&cwdi0.cwdi_lock); 407 408 /* Create the limits structures. */ 409 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE); 410 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) 411 limit0.pl_rlimit[i].rlim_cur = 412 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; 413 414 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 415 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = 416 maxfiles < nofile ? maxfiles : nofile; 417 418 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 419 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = 420 maxproc < maxuprc ? maxproc : maxuprc; 421 422 lim = ptoa(uvmexp.free); 423 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim; 424 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim; 425 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 426 limit0.pl_corename = defcorename; 427 limit0.pl_refcnt = 1; 428 limit0.pl_sv_limit = NULL; 429 430 /* Configure virtual memory system, set vm rlimits. */ 431 uvm_init_limits(p); 432 433 /* Initialize file descriptor table for proc0. */ 434 fd_init(&filedesc0); 435 436 /* 437 * Initialize proc0's vmspace, which uses the kernel pmap. 438 * All kernel processes (which never have user space mappings) 439 * share proc0's vmspace, and thus, the kernel pmap. 440 */ 441 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 442 trunc_page(VM_MAX_ADDRESS)); 443 444 l->l_addr = proc0paddr; /* XXX */ 445 446 /* Initialize signal state for proc0. XXX IPL_SCHED */ 447 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED); 448 siginit(p); 449 450 proc_initspecific(p); 451 lwp_initspecific(l); 452 453 SYSCALL_TIME_LWP_INIT(l); 454 } 455 456 /* 457 * Session reference counting. 458 */ 459 460 void 461 proc_sesshold(struct session *ss) 462 { 463 464 KASSERT(mutex_owned(proc_lock)); 465 ss->s_count++; 466 } 467 468 void 469 proc_sessrele(struct session *ss) 470 { 471 472 KASSERT(mutex_owned(proc_lock)); 473 /* 474 * We keep the pgrp with the same id as the session in order to 475 * stop a process being given the same pid. Since the pgrp holds 476 * a reference to the session, it must be a 'zombie' pgrp by now. 477 */ 478 if (--ss->s_count == 0) { 479 struct pgrp *pg; 480 481 pg = pg_remove(ss->s_sid); 482 mutex_exit(proc_lock); 483 484 kmem_free(pg, sizeof(struct pgrp)); 485 kmem_free(ss, sizeof(struct session)); 486 } else { 487 mutex_exit(proc_lock); 488 } 489 } 490 491 /* 492 * Check that the specified process group is in the session of the 493 * specified process. 494 * Treats -ve ids as process ids. 495 * Used to validate TIOCSPGRP requests. 496 */ 497 int 498 pgid_in_session(struct proc *p, pid_t pg_id) 499 { 500 struct pgrp *pgrp; 501 struct session *session; 502 int error; 503 504 mutex_enter(proc_lock); 505 if (pg_id < 0) { 506 struct proc *p1 = p_find(-pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL); 507 if (p1 == NULL) 508 return EINVAL; 509 pgrp = p1->p_pgrp; 510 } else { 511 pgrp = pg_find(pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL); 512 if (pgrp == NULL) 513 return EINVAL; 514 } 515 session = pgrp->pg_session; 516 if (session != p->p_pgrp->pg_session) 517 error = EPERM; 518 else 519 error = 0; 520 mutex_exit(proc_lock); 521 522 return error; 523 } 524 525 /* 526 * p_inferior: is p an inferior of q? 527 */ 528 static inline bool 529 p_inferior(struct proc *p, struct proc *q) 530 { 531 532 KASSERT(mutex_owned(proc_lock)); 533 534 for (; p != q; p = p->p_pptr) 535 if (p->p_pid == 0) 536 return false; 537 return true; 538 } 539 540 /* 541 * Locate a process by number 542 */ 543 struct proc * 544 p_find(pid_t pid, uint flags) 545 { 546 struct proc *p; 547 char stat; 548 549 if (!(flags & PFIND_LOCKED)) 550 mutex_enter(proc_lock); 551 552 p = pid_table[pid & pid_tbl_mask].pt_proc; 553 554 /* Only allow live processes to be found by pid. */ 555 /* XXXSMP p_stat */ 556 if (P_VALID(p) && p->p_pid == pid && ((stat = p->p_stat) == SACTIVE || 557 stat == SSTOP || ((flags & PFIND_ZOMBIE) && 558 (stat == SZOMB || stat == SDEAD || stat == SDYING)))) { 559 if (flags & PFIND_UNLOCK_OK) 560 mutex_exit(proc_lock); 561 return p; 562 } 563 if (flags & PFIND_UNLOCK_FAIL) 564 mutex_exit(proc_lock); 565 return NULL; 566 } 567 568 569 /* 570 * Locate a process group by number 571 */ 572 struct pgrp * 573 pg_find(pid_t pgid, uint flags) 574 { 575 struct pgrp *pg; 576 577 if (!(flags & PFIND_LOCKED)) 578 mutex_enter(proc_lock); 579 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp; 580 /* 581 * Can't look up a pgrp that only exists because the session 582 * hasn't died yet (traditional) 583 */ 584 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) { 585 if (flags & PFIND_UNLOCK_FAIL) 586 mutex_exit(proc_lock); 587 return NULL; 588 } 589 590 if (flags & PFIND_UNLOCK_OK) 591 mutex_exit(proc_lock); 592 return pg; 593 } 594 595 static void 596 expand_pid_table(void) 597 { 598 size_t pt_size, tsz; 599 struct pid_table *n_pt, *new_pt; 600 struct proc *proc; 601 struct pgrp *pgrp; 602 pid_t pid; 603 u_int i; 604 605 pt_size = pid_tbl_mask + 1; 606 tsz = pt_size * 2 * sizeof(struct pid_table); 607 new_pt = kmem_alloc(tsz, KM_SLEEP); 608 609 mutex_enter(proc_lock); 610 if (pt_size != pid_tbl_mask + 1) { 611 /* Another process beat us to it... */ 612 mutex_exit(proc_lock); 613 kmem_free(new_pt, tsz); 614 return; 615 } 616 617 /* 618 * Copy entries from old table into new one. 619 * If 'pid' is 'odd' we need to place in the upper half, 620 * even pid's to the lower half. 621 * Free items stay in the low half so we don't have to 622 * fixup the reference to them. 623 * We stuff free items on the front of the freelist 624 * because we can't write to unmodified entries. 625 * Processing the table backwards maintains a semblance 626 * of issueing pid numbers that increase with time. 627 */ 628 i = pt_size - 1; 629 n_pt = new_pt + i; 630 for (; ; i--, n_pt--) { 631 proc = pid_table[i].pt_proc; 632 pgrp = pid_table[i].pt_pgrp; 633 if (!P_VALID(proc)) { 634 /* Up 'use count' so that link is valid */ 635 pid = (P_NEXT(proc) + pt_size) & ~pt_size; 636 proc = P_FREE(pid); 637 if (pgrp) 638 pid = pgrp->pg_id; 639 } else 640 pid = proc->p_pid; 641 642 /* Save entry in appropriate half of table */ 643 n_pt[pid & pt_size].pt_proc = proc; 644 n_pt[pid & pt_size].pt_pgrp = pgrp; 645 646 /* Put other piece on start of free list */ 647 pid = (pid ^ pt_size) & ~pid_tbl_mask; 648 n_pt[pid & pt_size].pt_proc = 649 P_FREE((pid & ~pt_size) | next_free_pt); 650 n_pt[pid & pt_size].pt_pgrp = 0; 651 next_free_pt = i | (pid & pt_size); 652 if (i == 0) 653 break; 654 } 655 656 /* Save old table size and switch tables */ 657 tsz = pt_size * sizeof(struct pid_table); 658 n_pt = pid_table; 659 pid_table = new_pt; 660 pid_tbl_mask = pt_size * 2 - 1; 661 662 /* 663 * pid_max starts as PID_MAX (= 30000), once we have 16384 664 * allocated pids we need it to be larger! 665 */ 666 if (pid_tbl_mask > PID_MAX) { 667 pid_max = pid_tbl_mask * 2 + 1; 668 pid_alloc_lim |= pid_alloc_lim << 1; 669 } else 670 pid_alloc_lim <<= 1; /* doubles number of free slots... */ 671 672 mutex_exit(proc_lock); 673 kmem_free(n_pt, tsz); 674 } 675 676 struct proc * 677 proc_alloc(void) 678 { 679 struct proc *p; 680 int nxt; 681 pid_t pid; 682 struct pid_table *pt; 683 684 p = pool_cache_get(proc_cache, PR_WAITOK); 685 p->p_stat = SIDL; /* protect against others */ 686 687 proc_initspecific(p); 688 /* allocate next free pid */ 689 690 for (;;expand_pid_table()) { 691 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) 692 /* ensure pids cycle through 2000+ values */ 693 continue; 694 mutex_enter(proc_lock); 695 pt = &pid_table[next_free_pt]; 696 #ifdef DIAGNOSTIC 697 if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp)) 698 panic("proc_alloc: slot busy"); 699 #endif 700 nxt = P_NEXT(pt->pt_proc); 701 if (nxt & pid_tbl_mask) 702 break; 703 /* Table full - expand (NB last entry not used....) */ 704 mutex_exit(proc_lock); 705 } 706 707 /* pid is 'saved use count' + 'size' + entry */ 708 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt; 709 if ((uint)pid > (uint)pid_max) 710 pid &= pid_tbl_mask; 711 p->p_pid = pid; 712 next_free_pt = nxt & pid_tbl_mask; 713 714 /* Grab table slot */ 715 pt->pt_proc = p; 716 pid_alloc_cnt++; 717 718 mutex_exit(proc_lock); 719 720 return p; 721 } 722 723 /* 724 * Free a process id - called from proc_free (in kern_exit.c) 725 * 726 * Called with the proc_lock held. 727 */ 728 void 729 proc_free_pid(struct proc *p) 730 { 731 pid_t pid = p->p_pid; 732 struct pid_table *pt; 733 734 KASSERT(mutex_owned(proc_lock)); 735 736 pt = &pid_table[pid & pid_tbl_mask]; 737 #ifdef DIAGNOSTIC 738 if (__predict_false(pt->pt_proc != p)) 739 panic("proc_free: pid_table mismatch, pid %x, proc %p", 740 pid, p); 741 #endif 742 /* save pid use count in slot */ 743 pt->pt_proc = P_FREE(pid & ~pid_tbl_mask); 744 745 if (pt->pt_pgrp == NULL) { 746 /* link last freed entry onto ours */ 747 pid &= pid_tbl_mask; 748 pt = &pid_table[last_free_pt]; 749 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid); 750 last_free_pt = pid; 751 pid_alloc_cnt--; 752 } 753 754 atomic_dec_uint(&nprocs); 755 } 756 757 void 758 proc_free_mem(struct proc *p) 759 { 760 761 pool_cache_put(proc_cache, p); 762 } 763 764 /* 765 * proc_enterpgrp: move p to a new or existing process group (and session). 766 * 767 * If we are creating a new pgrp, the pgid should equal 768 * the calling process' pid. 769 * If is only valid to enter a process group that is in the session 770 * of the process. 771 * Also mksess should only be set if we are creating a process group 772 * 773 * Only called from sys_setsid and sys_setpgid. 774 */ 775 int 776 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess) 777 { 778 struct pgrp *new_pgrp, *pgrp; 779 struct session *sess; 780 struct proc *p; 781 int rval; 782 pid_t pg_id = NO_PGID; 783 784 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL; 785 786 /* Allocate data areas we might need before doing any validity checks */ 787 mutex_enter(proc_lock); /* Because pid_table might change */ 788 if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) { 789 mutex_exit(proc_lock); 790 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP); 791 mutex_enter(proc_lock); 792 } else 793 new_pgrp = NULL; 794 rval = EPERM; /* most common error (to save typing) */ 795 796 /* Check pgrp exists or can be created */ 797 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp; 798 if (pgrp != NULL && pgrp->pg_id != pgid) 799 goto done; 800 801 /* Can only set another process under restricted circumstances. */ 802 if (pid != curp->p_pid) { 803 /* must exist and be one of our children... */ 804 if ((p = p_find(pid, PFIND_LOCKED)) == NULL || 805 !p_inferior(p, curp)) { 806 rval = ESRCH; 807 goto done; 808 } 809 /* ... in the same session... */ 810 if (sess != NULL || p->p_session != curp->p_session) 811 goto done; 812 /* ... existing pgid must be in same session ... */ 813 if (pgrp != NULL && pgrp->pg_session != p->p_session) 814 goto done; 815 /* ... and not done an exec. */ 816 if (p->p_flag & PK_EXEC) { 817 rval = EACCES; 818 goto done; 819 } 820 } else { 821 /* ... setsid() cannot re-enter a pgrp */ 822 if (mksess && (curp->p_pgid == curp->p_pid || 823 pg_find(curp->p_pid, PFIND_LOCKED))) 824 goto done; 825 p = curp; 826 } 827 828 /* Changing the process group/session of a session 829 leader is definitely off limits. */ 830 if (SESS_LEADER(p)) { 831 if (sess == NULL && p->p_pgrp == pgrp) 832 /* unless it's a definite noop */ 833 rval = 0; 834 goto done; 835 } 836 837 /* Can only create a process group with id of process */ 838 if (pgrp == NULL && pgid != pid) 839 goto done; 840 841 /* Can only create a session if creating pgrp */ 842 if (sess != NULL && pgrp != NULL) 843 goto done; 844 845 /* Check we allocated memory for a pgrp... */ 846 if (pgrp == NULL && new_pgrp == NULL) 847 goto done; 848 849 /* Don't attach to 'zombie' pgrp */ 850 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members)) 851 goto done; 852 853 /* Expect to succeed now */ 854 rval = 0; 855 856 if (pgrp == p->p_pgrp) 857 /* nothing to do */ 858 goto done; 859 860 /* Ok all setup, link up required structures */ 861 862 if (pgrp == NULL) { 863 pgrp = new_pgrp; 864 new_pgrp = NULL; 865 if (sess != NULL) { 866 sess->s_sid = p->p_pid; 867 sess->s_leader = p; 868 sess->s_count = 1; 869 sess->s_ttyvp = NULL; 870 sess->s_ttyp = NULL; 871 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET; 872 memcpy(sess->s_login, p->p_session->s_login, 873 sizeof(sess->s_login)); 874 p->p_lflag &= ~PL_CONTROLT; 875 } else { 876 sess = p->p_pgrp->pg_session; 877 proc_sesshold(sess); 878 } 879 pgrp->pg_session = sess; 880 sess = NULL; 881 882 pgrp->pg_id = pgid; 883 LIST_INIT(&pgrp->pg_members); 884 #ifdef DIAGNOSTIC 885 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp)) 886 panic("enterpgrp: pgrp table slot in use"); 887 if (__predict_false(mksess && p != curp)) 888 panic("enterpgrp: mksession and p != curproc"); 889 #endif 890 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp; 891 pgrp->pg_jobc = 0; 892 } 893 894 /* 895 * Adjust eligibility of affected pgrps to participate in job control. 896 * Increment eligibility counts before decrementing, otherwise we 897 * could reach 0 spuriously during the first call. 898 */ 899 fixjobc(p, pgrp, 1); 900 fixjobc(p, p->p_pgrp, 0); 901 902 /* Interlock with ttread(). */ 903 mutex_spin_enter(&tty_lock); 904 905 /* Move process to requested group. */ 906 LIST_REMOVE(p, p_pglist); 907 if (LIST_EMPTY(&p->p_pgrp->pg_members)) 908 /* defer delete until we've dumped the lock */ 909 pg_id = p->p_pgrp->pg_id; 910 p->p_pgrp = pgrp; 911 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 912 913 /* Done with the swap; we can release the tty mutex. */ 914 mutex_spin_exit(&tty_lock); 915 916 done: 917 if (pg_id != NO_PGID) { 918 /* Releases proc_lock. */ 919 pg_delete(pg_id); 920 } else { 921 mutex_exit(proc_lock); 922 } 923 if (sess != NULL) 924 kmem_free(sess, sizeof(*sess)); 925 if (new_pgrp != NULL) 926 kmem_free(new_pgrp, sizeof(*new_pgrp)); 927 #ifdef DEBUG_PGRP 928 if (__predict_false(rval)) 929 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n", 930 pid, pgid, mksess, curp->p_pid, rval); 931 #endif 932 return rval; 933 } 934 935 /* 936 * proc_leavepgrp: remove a process from its process group. 937 * => must be called with the proc_lock held, which will be released; 938 */ 939 void 940 proc_leavepgrp(struct proc *p) 941 { 942 struct pgrp *pgrp; 943 944 KASSERT(mutex_owned(proc_lock)); 945 946 /* Interlock with ttread() */ 947 mutex_spin_enter(&tty_lock); 948 pgrp = p->p_pgrp; 949 LIST_REMOVE(p, p_pglist); 950 p->p_pgrp = NULL; 951 mutex_spin_exit(&tty_lock); 952 953 if (LIST_EMPTY(&pgrp->pg_members)) { 954 /* Releases proc_lock. */ 955 pg_delete(pgrp->pg_id); 956 } else { 957 mutex_exit(proc_lock); 958 } 959 } 960 961 /* 962 * pg_remove: remove a process group from the table. 963 * => must be called with the proc_lock held; 964 * => returns process group to free; 965 */ 966 static struct pgrp * 967 pg_remove(pid_t pg_id) 968 { 969 struct pgrp *pgrp; 970 struct pid_table *pt; 971 972 KASSERT(mutex_owned(proc_lock)); 973 974 pt = &pid_table[pg_id & pid_tbl_mask]; 975 pgrp = pt->pt_pgrp; 976 977 KASSERT(pgrp != NULL); 978 KASSERT(pgrp->pg_id == pg_id); 979 KASSERT(LIST_EMPTY(&pgrp->pg_members)); 980 981 pt->pt_pgrp = NULL; 982 983 if (!P_VALID(pt->pt_proc)) { 984 /* Orphaned pgrp, put slot onto free list. */ 985 KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0); 986 pg_id &= pid_tbl_mask; 987 pt = &pid_table[last_free_pt]; 988 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id); 989 last_free_pt = pg_id; 990 pid_alloc_cnt--; 991 } 992 return pgrp; 993 } 994 995 /* 996 * pg_delete: delete and free a process group. 997 * => must be called with the proc_lock held, which will be released. 998 */ 999 static void 1000 pg_delete(pid_t pg_id) 1001 { 1002 struct pgrp *pg; 1003 struct tty *ttyp; 1004 struct session *ss; 1005 1006 KASSERT(mutex_owned(proc_lock)); 1007 1008 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp; 1009 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) { 1010 mutex_exit(proc_lock); 1011 return; 1012 } 1013 1014 ss = pg->pg_session; 1015 1016 /* Remove reference (if any) from tty to this process group */ 1017 mutex_spin_enter(&tty_lock); 1018 ttyp = ss->s_ttyp; 1019 if (ttyp != NULL && ttyp->t_pgrp == pg) { 1020 ttyp->t_pgrp = NULL; 1021 KASSERT(ttyp->t_session == ss); 1022 } 1023 mutex_spin_exit(&tty_lock); 1024 1025 /* 1026 * The leading process group in a session is freed by proc_sessrele(), 1027 * if last reference. Note: proc_sessrele() releases proc_lock. 1028 */ 1029 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL; 1030 proc_sessrele(ss); 1031 1032 if (pg != NULL) { 1033 /* Free it, if was not done by proc_sessrele(). */ 1034 kmem_free(pg, sizeof(struct pgrp)); 1035 } 1036 } 1037 1038 /* 1039 * Adjust pgrp jobc counters when specified process changes process group. 1040 * We count the number of processes in each process group that "qualify" 1041 * the group for terminal job control (those with a parent in a different 1042 * process group of the same session). If that count reaches zero, the 1043 * process group becomes orphaned. Check both the specified process' 1044 * process group and that of its children. 1045 * entering == 0 => p is leaving specified group. 1046 * entering == 1 => p is entering specified group. 1047 * 1048 * Call with proc_lock held. 1049 */ 1050 void 1051 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 1052 { 1053 struct pgrp *hispgrp; 1054 struct session *mysession = pgrp->pg_session; 1055 struct proc *child; 1056 1057 KASSERT(mutex_owned(proc_lock)); 1058 1059 /* 1060 * Check p's parent to see whether p qualifies its own process 1061 * group; if so, adjust count for p's process group. 1062 */ 1063 hispgrp = p->p_pptr->p_pgrp; 1064 if (hispgrp != pgrp && hispgrp->pg_session == mysession) { 1065 if (entering) { 1066 pgrp->pg_jobc++; 1067 p->p_lflag &= ~PL_ORPHANPG; 1068 } else if (--pgrp->pg_jobc == 0) 1069 orphanpg(pgrp); 1070 } 1071 1072 /* 1073 * Check this process' children to see whether they qualify 1074 * their process groups; if so, adjust counts for children's 1075 * process groups. 1076 */ 1077 LIST_FOREACH(child, &p->p_children, p_sibling) { 1078 hispgrp = child->p_pgrp; 1079 if (hispgrp != pgrp && hispgrp->pg_session == mysession && 1080 !P_ZOMBIE(child)) { 1081 if (entering) { 1082 child->p_lflag &= ~PL_ORPHANPG; 1083 hispgrp->pg_jobc++; 1084 } else if (--hispgrp->pg_jobc == 0) 1085 orphanpg(hispgrp); 1086 } 1087 } 1088 } 1089 1090 /* 1091 * A process group has become orphaned; 1092 * if there are any stopped processes in the group, 1093 * hang-up all process in that group. 1094 * 1095 * Call with proc_lock held. 1096 */ 1097 static void 1098 orphanpg(struct pgrp *pg) 1099 { 1100 struct proc *p; 1101 int doit; 1102 1103 KASSERT(mutex_owned(proc_lock)); 1104 1105 doit = 0; 1106 1107 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1108 if (p->p_stat == SSTOP) { 1109 p->p_lflag |= PL_ORPHANPG; 1110 psignal(p, SIGHUP); 1111 psignal(p, SIGCONT); 1112 } 1113 } 1114 } 1115 1116 #ifdef DDB 1117 #include <ddb/db_output.h> 1118 void pidtbl_dump(void); 1119 void 1120 pidtbl_dump(void) 1121 { 1122 struct pid_table *pt; 1123 struct proc *p; 1124 struct pgrp *pgrp; 1125 int id; 1126 1127 db_printf("pid table %p size %x, next %x, last %x\n", 1128 pid_table, pid_tbl_mask+1, 1129 next_free_pt, last_free_pt); 1130 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) { 1131 p = pt->pt_proc; 1132 if (!P_VALID(p) && !pt->pt_pgrp) 1133 continue; 1134 db_printf(" id %x: ", id); 1135 if (P_VALID(p)) 1136 db_printf("proc %p id %d (0x%x) %s\n", 1137 p, p->p_pid, p->p_pid, p->p_comm); 1138 else 1139 db_printf("next %x use %x\n", 1140 P_NEXT(p) & pid_tbl_mask, 1141 P_NEXT(p) & ~pid_tbl_mask); 1142 if ((pgrp = pt->pt_pgrp)) { 1143 db_printf("\tsession %p, sid %d, count %d, login %s\n", 1144 pgrp->pg_session, pgrp->pg_session->s_sid, 1145 pgrp->pg_session->s_count, 1146 pgrp->pg_session->s_login); 1147 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n", 1148 pgrp, pgrp->pg_id, pgrp->pg_jobc, 1149 LIST_FIRST(&pgrp->pg_members)); 1150 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 1151 db_printf("\t\tpid %d addr %p pgrp %p %s\n", 1152 p->p_pid, p, p->p_pgrp, p->p_comm); 1153 } 1154 } 1155 } 1156 } 1157 #endif /* DDB */ 1158 1159 #ifdef KSTACK_CHECK_MAGIC 1160 #include <sys/user.h> 1161 1162 #define KSTACK_MAGIC 0xdeadbeaf 1163 1164 /* XXX should be per process basis? */ 1165 static int kstackleftmin = KSTACK_SIZE; 1166 static int kstackleftthres = KSTACK_SIZE / 8; 1167 1168 void 1169 kstack_setup_magic(const struct lwp *l) 1170 { 1171 uint32_t *ip; 1172 uint32_t const *end; 1173 1174 KASSERT(l != NULL); 1175 KASSERT(l != &lwp0); 1176 1177 /* 1178 * fill all the stack with magic number 1179 * so that later modification on it can be detected. 1180 */ 1181 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1182 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1183 for (; ip < end; ip++) { 1184 *ip = KSTACK_MAGIC; 1185 } 1186 } 1187 1188 void 1189 kstack_check_magic(const struct lwp *l) 1190 { 1191 uint32_t const *ip, *end; 1192 int stackleft; 1193 1194 KASSERT(l != NULL); 1195 1196 /* don't check proc0 */ /*XXX*/ 1197 if (l == &lwp0) 1198 return; 1199 1200 #ifdef __MACHINE_STACK_GROWS_UP 1201 /* stack grows upwards (eg. hppa) */ 1202 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1203 end = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1204 for (ip--; ip >= end; ip--) 1205 if (*ip != KSTACK_MAGIC) 1206 break; 1207 1208 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip; 1209 #else /* __MACHINE_STACK_GROWS_UP */ 1210 /* stack grows downwards (eg. i386) */ 1211 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1212 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1213 for (; ip < end; ip++) 1214 if (*ip != KSTACK_MAGIC) 1215 break; 1216 1217 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l); 1218 #endif /* __MACHINE_STACK_GROWS_UP */ 1219 1220 if (kstackleftmin > stackleft) { 1221 kstackleftmin = stackleft; 1222 if (stackleft < kstackleftthres) 1223 printf("warning: kernel stack left %d bytes" 1224 "(pid %u:lid %u)\n", stackleft, 1225 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1226 } 1227 1228 if (stackleft <= 0) { 1229 panic("magic on the top of kernel stack changed for " 1230 "pid %u, lid %u: maybe kernel stack overflow", 1231 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1232 } 1233 } 1234 #endif /* KSTACK_CHECK_MAGIC */ 1235 1236 int 1237 proclist_foreach_call(struct proclist *list, 1238 int (*callback)(struct proc *, void *arg), void *arg) 1239 { 1240 struct proc marker; 1241 struct proc *p; 1242 int ret = 0; 1243 1244 marker.p_flag = PK_MARKER; 1245 mutex_enter(proc_lock); 1246 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) { 1247 if (p->p_flag & PK_MARKER) { 1248 p = LIST_NEXT(p, p_list); 1249 continue; 1250 } 1251 LIST_INSERT_AFTER(p, &marker, p_list); 1252 ret = (*callback)(p, arg); 1253 KASSERT(mutex_owned(proc_lock)); 1254 p = LIST_NEXT(&marker, p_list); 1255 LIST_REMOVE(&marker, p_list); 1256 } 1257 mutex_exit(proc_lock); 1258 1259 return ret; 1260 } 1261 1262 int 1263 proc_vmspace_getref(struct proc *p, struct vmspace **vm) 1264 { 1265 1266 /* XXXCDC: how should locking work here? */ 1267 1268 /* curproc exception is for coredump. */ 1269 1270 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) || 1271 (p->p_vmspace->vm_refcnt < 1)) { /* XXX */ 1272 return EFAULT; 1273 } 1274 1275 uvmspace_addref(p->p_vmspace); 1276 *vm = p->p_vmspace; 1277 1278 return 0; 1279 } 1280 1281 /* 1282 * Acquire a write lock on the process credential. 1283 */ 1284 void 1285 proc_crmod_enter(void) 1286 { 1287 struct lwp *l = curlwp; 1288 struct proc *p = l->l_proc; 1289 struct plimit *lim; 1290 kauth_cred_t oc; 1291 char *cn; 1292 1293 /* Reset what needs to be reset in plimit. */ 1294 if (p->p_limit->pl_corename != defcorename) { 1295 lim_privatise(p, false); 1296 lim = p->p_limit; 1297 mutex_enter(&lim->pl_lock); 1298 cn = lim->pl_corename; 1299 lim->pl_corename = defcorename; 1300 mutex_exit(&lim->pl_lock); 1301 if (cn != defcorename) 1302 free(cn, M_TEMP); 1303 } 1304 1305 mutex_enter(p->p_lock); 1306 1307 /* Ensure the LWP cached credentials are up to date. */ 1308 if ((oc = l->l_cred) != p->p_cred) { 1309 kauth_cred_hold(p->p_cred); 1310 l->l_cred = p->p_cred; 1311 kauth_cred_free(oc); 1312 } 1313 1314 } 1315 1316 /* 1317 * Set in a new process credential, and drop the write lock. The credential 1318 * must have a reference already. Optionally, free a no-longer required 1319 * credential. The scheduler also needs to inspect p_cred, so we also 1320 * briefly acquire the sched state mutex. 1321 */ 1322 void 1323 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid) 1324 { 1325 struct lwp *l = curlwp, *l2; 1326 struct proc *p = l->l_proc; 1327 kauth_cred_t oc; 1328 1329 KASSERT(mutex_owned(p->p_lock)); 1330 1331 /* Is there a new credential to set in? */ 1332 if (scred != NULL) { 1333 p->p_cred = scred; 1334 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 1335 if (l2 != l) 1336 l2->l_prflag |= LPR_CRMOD; 1337 } 1338 1339 /* Ensure the LWP cached credentials are up to date. */ 1340 if ((oc = l->l_cred) != scred) { 1341 kauth_cred_hold(scred); 1342 l->l_cred = scred; 1343 } 1344 } else 1345 oc = NULL; /* XXXgcc */ 1346 1347 if (sugid) { 1348 /* 1349 * Mark process as having changed credentials, stops 1350 * tracing etc. 1351 */ 1352 p->p_flag |= PK_SUGID; 1353 } 1354 1355 mutex_exit(p->p_lock); 1356 1357 /* If there is a credential to be released, free it now. */ 1358 if (fcred != NULL) { 1359 KASSERT(scred != NULL); 1360 kauth_cred_free(fcred); 1361 if (oc != scred) 1362 kauth_cred_free(oc); 1363 } 1364 } 1365 1366 /* 1367 * proc_specific_key_create -- 1368 * Create a key for subsystem proc-specific data. 1369 */ 1370 int 1371 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 1372 { 1373 1374 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor)); 1375 } 1376 1377 /* 1378 * proc_specific_key_delete -- 1379 * Delete a key for subsystem proc-specific data. 1380 */ 1381 void 1382 proc_specific_key_delete(specificdata_key_t key) 1383 { 1384 1385 specificdata_key_delete(proc_specificdata_domain, key); 1386 } 1387 1388 /* 1389 * proc_initspecific -- 1390 * Initialize a proc's specificdata container. 1391 */ 1392 void 1393 proc_initspecific(struct proc *p) 1394 { 1395 int error; 1396 1397 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref); 1398 KASSERT(error == 0); 1399 } 1400 1401 /* 1402 * proc_finispecific -- 1403 * Finalize a proc's specificdata container. 1404 */ 1405 void 1406 proc_finispecific(struct proc *p) 1407 { 1408 1409 specificdata_fini(proc_specificdata_domain, &p->p_specdataref); 1410 } 1411 1412 /* 1413 * proc_getspecific -- 1414 * Return proc-specific data corresponding to the specified key. 1415 */ 1416 void * 1417 proc_getspecific(struct proc *p, specificdata_key_t key) 1418 { 1419 1420 return (specificdata_getspecific(proc_specificdata_domain, 1421 &p->p_specdataref, key)); 1422 } 1423 1424 /* 1425 * proc_setspecific -- 1426 * Set proc-specific data corresponding to the specified key. 1427 */ 1428 void 1429 proc_setspecific(struct proc *p, specificdata_key_t key, void *data) 1430 { 1431 1432 specificdata_setspecific(proc_specificdata_domain, 1433 &p->p_specdataref, key, data); 1434 } 1435 1436 int 1437 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target) 1438 { 1439 int r = 0; 1440 1441 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) || 1442 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) { 1443 /* 1444 * suid proc of ours or proc not ours 1445 */ 1446 r = EPERM; 1447 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) { 1448 /* 1449 * sgid proc has sgid back to us temporarily 1450 */ 1451 r = EPERM; 1452 } else { 1453 /* 1454 * our rgid must be in target's group list (ie, 1455 * sub-processes started by a sgid process) 1456 */ 1457 int ismember = 0; 1458 1459 if (kauth_cred_ismember_gid(cred, 1460 kauth_cred_getgid(target), &ismember) != 0 || 1461 !ismember) 1462 r = EPERM; 1463 } 1464 1465 return (r); 1466 } 1467 1468