kern_proc.c revision 1.157
1 /* $NetBSD: kern_proc.c,v 1.157 2009/11/04 15:35:09 pooka 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 1.157 2009/11/04 15:35:09 pooka 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 lwp *l; 363 rlim_t lim; 364 int i; 365 366 p = &proc0; 367 pg = &pgrp0; 368 l = &lwp0; 369 370 KASSERT(l->l_lid == p->p_nlwpid); 371 372 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 373 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 374 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 375 376 rw_init(&p->p_reflock); 377 cv_init(&p->p_waitcv, "wait"); 378 cv_init(&p->p_lwpcv, "lwpwait"); 379 380 LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling); 381 382 pid_table[0].pt_proc = p; 383 LIST_INSERT_HEAD(&allproc, p, p_list); 384 LIST_INSERT_HEAD(&alllwp, l, l_list); 385 386 pid_table[0].pt_pgrp = pg; 387 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist); 388 389 #ifdef __HAVE_SYSCALL_INTERN 390 (*p->p_emul->e_syscall_intern)(p); 391 #endif 392 393 callout_init(&l->l_timeout_ch, CALLOUT_MPSAFE); 394 callout_setfunc(&l->l_timeout_ch, sleepq_timeout, l); 395 cv_init(&l->l_sigcv, "sigwait"); 396 397 /* Create credentials. */ 398 cred0 = kauth_cred_alloc(); 399 p->p_cred = cred0; 400 kauth_cred_hold(cred0); 401 l->l_cred = cred0; 402 403 /* Create the CWD info. */ 404 rw_init(&cwdi0.cwdi_lock); 405 406 /* Create the limits structures. */ 407 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE); 408 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) 409 limit0.pl_rlimit[i].rlim_cur = 410 limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; 411 412 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles; 413 limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = 414 maxfiles < nofile ? maxfiles : nofile; 415 416 limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc; 417 limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = 418 maxproc < maxuprc ? maxproc : maxuprc; 419 420 lim = ptoa(uvmexp.free); 421 limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim; 422 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim; 423 limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 424 limit0.pl_corename = defcorename; 425 limit0.pl_refcnt = 1; 426 limit0.pl_sv_limit = NULL; 427 428 /* Configure virtual memory system, set vm rlimits. */ 429 uvm_init_limits(p); 430 431 /* Initialize file descriptor table for proc0. */ 432 fd_init(&filedesc0); 433 434 /* 435 * Initialize proc0's vmspace, which uses the kernel pmap. 436 * All kernel processes (which never have user space mappings) 437 * share proc0's vmspace, and thus, the kernel pmap. 438 */ 439 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 440 trunc_page(VM_MAX_ADDRESS)); 441 442 l->l_addr = proc0paddr; /* XXX */ 443 444 /* Initialize signal state for proc0. XXX IPL_SCHED */ 445 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED); 446 siginit(p); 447 448 proc_initspecific(p); 449 lwp_initspecific(l); 450 451 SYSCALL_TIME_LWP_INIT(l); 452 } 453 454 /* 455 * Session reference counting. 456 */ 457 458 void 459 proc_sesshold(struct session *ss) 460 { 461 462 KASSERT(mutex_owned(proc_lock)); 463 ss->s_count++; 464 } 465 466 void 467 proc_sessrele(struct session *ss) 468 { 469 470 KASSERT(mutex_owned(proc_lock)); 471 /* 472 * We keep the pgrp with the same id as the session in order to 473 * stop a process being given the same pid. Since the pgrp holds 474 * a reference to the session, it must be a 'zombie' pgrp by now. 475 */ 476 if (--ss->s_count == 0) { 477 struct pgrp *pg; 478 479 pg = pg_remove(ss->s_sid); 480 mutex_exit(proc_lock); 481 482 kmem_free(pg, sizeof(struct pgrp)); 483 kmem_free(ss, sizeof(struct session)); 484 } else { 485 mutex_exit(proc_lock); 486 } 487 } 488 489 /* 490 * Check that the specified process group is in the session of the 491 * specified process. 492 * Treats -ve ids as process ids. 493 * Used to validate TIOCSPGRP requests. 494 */ 495 int 496 pgid_in_session(struct proc *p, pid_t pg_id) 497 { 498 struct pgrp *pgrp; 499 struct session *session; 500 int error; 501 502 mutex_enter(proc_lock); 503 if (pg_id < 0) { 504 struct proc *p1 = p_find(-pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL); 505 if (p1 == NULL) 506 return EINVAL; 507 pgrp = p1->p_pgrp; 508 } else { 509 pgrp = pg_find(pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL); 510 if (pgrp == NULL) 511 return EINVAL; 512 } 513 session = pgrp->pg_session; 514 if (session != p->p_pgrp->pg_session) 515 error = EPERM; 516 else 517 error = 0; 518 mutex_exit(proc_lock); 519 520 return error; 521 } 522 523 /* 524 * p_inferior: is p an inferior of q? 525 */ 526 static inline bool 527 p_inferior(struct proc *p, struct proc *q) 528 { 529 530 KASSERT(mutex_owned(proc_lock)); 531 532 for (; p != q; p = p->p_pptr) 533 if (p->p_pid == 0) 534 return false; 535 return true; 536 } 537 538 /* 539 * Locate a process by number 540 */ 541 struct proc * 542 p_find(pid_t pid, uint flags) 543 { 544 struct proc *p; 545 char stat; 546 547 if (!(flags & PFIND_LOCKED)) 548 mutex_enter(proc_lock); 549 550 p = pid_table[pid & pid_tbl_mask].pt_proc; 551 552 /* Only allow live processes to be found by pid. */ 553 /* XXXSMP p_stat */ 554 if (P_VALID(p) && p->p_pid == pid && ((stat = p->p_stat) == SACTIVE || 555 stat == SSTOP || ((flags & PFIND_ZOMBIE) && 556 (stat == SZOMB || stat == SDEAD || stat == SDYING)))) { 557 if (flags & PFIND_UNLOCK_OK) 558 mutex_exit(proc_lock); 559 return p; 560 } 561 if (flags & PFIND_UNLOCK_FAIL) 562 mutex_exit(proc_lock); 563 return NULL; 564 } 565 566 567 /* 568 * Locate a process group by number 569 */ 570 struct pgrp * 571 pg_find(pid_t pgid, uint flags) 572 { 573 struct pgrp *pg; 574 575 if (!(flags & PFIND_LOCKED)) 576 mutex_enter(proc_lock); 577 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp; 578 /* 579 * Can't look up a pgrp that only exists because the session 580 * hasn't died yet (traditional) 581 */ 582 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) { 583 if (flags & PFIND_UNLOCK_FAIL) 584 mutex_exit(proc_lock); 585 return NULL; 586 } 587 588 if (flags & PFIND_UNLOCK_OK) 589 mutex_exit(proc_lock); 590 return pg; 591 } 592 593 static void 594 expand_pid_table(void) 595 { 596 size_t pt_size, tsz; 597 struct pid_table *n_pt, *new_pt; 598 struct proc *proc; 599 struct pgrp *pgrp; 600 pid_t pid; 601 u_int i; 602 603 pt_size = pid_tbl_mask + 1; 604 tsz = pt_size * 2 * sizeof(struct pid_table); 605 new_pt = kmem_alloc(tsz, KM_SLEEP); 606 607 mutex_enter(proc_lock); 608 if (pt_size != pid_tbl_mask + 1) { 609 /* Another process beat us to it... */ 610 mutex_exit(proc_lock); 611 kmem_free(new_pt, tsz); 612 return; 613 } 614 615 /* 616 * Copy entries from old table into new one. 617 * If 'pid' is 'odd' we need to place in the upper half, 618 * even pid's to the lower half. 619 * Free items stay in the low half so we don't have to 620 * fixup the reference to them. 621 * We stuff free items on the front of the freelist 622 * because we can't write to unmodified entries. 623 * Processing the table backwards maintains a semblance 624 * of issueing pid numbers that increase with time. 625 */ 626 i = pt_size - 1; 627 n_pt = new_pt + i; 628 for (; ; i--, n_pt--) { 629 proc = pid_table[i].pt_proc; 630 pgrp = pid_table[i].pt_pgrp; 631 if (!P_VALID(proc)) { 632 /* Up 'use count' so that link is valid */ 633 pid = (P_NEXT(proc) + pt_size) & ~pt_size; 634 proc = P_FREE(pid); 635 if (pgrp) 636 pid = pgrp->pg_id; 637 } else 638 pid = proc->p_pid; 639 640 /* Save entry in appropriate half of table */ 641 n_pt[pid & pt_size].pt_proc = proc; 642 n_pt[pid & pt_size].pt_pgrp = pgrp; 643 644 /* Put other piece on start of free list */ 645 pid = (pid ^ pt_size) & ~pid_tbl_mask; 646 n_pt[pid & pt_size].pt_proc = 647 P_FREE((pid & ~pt_size) | next_free_pt); 648 n_pt[pid & pt_size].pt_pgrp = 0; 649 next_free_pt = i | (pid & pt_size); 650 if (i == 0) 651 break; 652 } 653 654 /* Save old table size and switch tables */ 655 tsz = pt_size * sizeof(struct pid_table); 656 n_pt = pid_table; 657 pid_table = new_pt; 658 pid_tbl_mask = pt_size * 2 - 1; 659 660 /* 661 * pid_max starts as PID_MAX (= 30000), once we have 16384 662 * allocated pids we need it to be larger! 663 */ 664 if (pid_tbl_mask > PID_MAX) { 665 pid_max = pid_tbl_mask * 2 + 1; 666 pid_alloc_lim |= pid_alloc_lim << 1; 667 } else 668 pid_alloc_lim <<= 1; /* doubles number of free slots... */ 669 670 mutex_exit(proc_lock); 671 kmem_free(n_pt, tsz); 672 } 673 674 struct proc * 675 proc_alloc(void) 676 { 677 struct proc *p; 678 int nxt; 679 pid_t pid; 680 struct pid_table *pt; 681 682 p = pool_cache_get(proc_cache, PR_WAITOK); 683 p->p_stat = SIDL; /* protect against others */ 684 685 proc_initspecific(p); 686 /* allocate next free pid */ 687 688 for (;;expand_pid_table()) { 689 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) 690 /* ensure pids cycle through 2000+ values */ 691 continue; 692 mutex_enter(proc_lock); 693 pt = &pid_table[next_free_pt]; 694 #ifdef DIAGNOSTIC 695 if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp)) 696 panic("proc_alloc: slot busy"); 697 #endif 698 nxt = P_NEXT(pt->pt_proc); 699 if (nxt & pid_tbl_mask) 700 break; 701 /* Table full - expand (NB last entry not used....) */ 702 mutex_exit(proc_lock); 703 } 704 705 /* pid is 'saved use count' + 'size' + entry */ 706 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt; 707 if ((uint)pid > (uint)pid_max) 708 pid &= pid_tbl_mask; 709 p->p_pid = pid; 710 next_free_pt = nxt & pid_tbl_mask; 711 712 /* Grab table slot */ 713 pt->pt_proc = p; 714 pid_alloc_cnt++; 715 716 mutex_exit(proc_lock); 717 718 return p; 719 } 720 721 /* 722 * Free a process id - called from proc_free (in kern_exit.c) 723 * 724 * Called with the proc_lock held. 725 */ 726 void 727 proc_free_pid(struct proc *p) 728 { 729 pid_t pid = p->p_pid; 730 struct pid_table *pt; 731 732 KASSERT(mutex_owned(proc_lock)); 733 734 pt = &pid_table[pid & pid_tbl_mask]; 735 #ifdef DIAGNOSTIC 736 if (__predict_false(pt->pt_proc != p)) 737 panic("proc_free: pid_table mismatch, pid %x, proc %p", 738 pid, p); 739 #endif 740 /* save pid use count in slot */ 741 pt->pt_proc = P_FREE(pid & ~pid_tbl_mask); 742 743 if (pt->pt_pgrp == NULL) { 744 /* link last freed entry onto ours */ 745 pid &= pid_tbl_mask; 746 pt = &pid_table[last_free_pt]; 747 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid); 748 last_free_pt = pid; 749 pid_alloc_cnt--; 750 } 751 752 atomic_dec_uint(&nprocs); 753 } 754 755 void 756 proc_free_mem(struct proc *p) 757 { 758 759 pool_cache_put(proc_cache, p); 760 } 761 762 /* 763 * proc_enterpgrp: move p to a new or existing process group (and session). 764 * 765 * If we are creating a new pgrp, the pgid should equal 766 * the calling process' pid. 767 * If is only valid to enter a process group that is in the session 768 * of the process. 769 * Also mksess should only be set if we are creating a process group 770 * 771 * Only called from sys_setsid and sys_setpgid. 772 */ 773 int 774 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess) 775 { 776 struct pgrp *new_pgrp, *pgrp; 777 struct session *sess; 778 struct proc *p; 779 int rval; 780 pid_t pg_id = NO_PGID; 781 782 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL; 783 784 /* Allocate data areas we might need before doing any validity checks */ 785 mutex_enter(proc_lock); /* Because pid_table might change */ 786 if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) { 787 mutex_exit(proc_lock); 788 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP); 789 mutex_enter(proc_lock); 790 } else 791 new_pgrp = NULL; 792 rval = EPERM; /* most common error (to save typing) */ 793 794 /* Check pgrp exists or can be created */ 795 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp; 796 if (pgrp != NULL && pgrp->pg_id != pgid) 797 goto done; 798 799 /* Can only set another process under restricted circumstances. */ 800 if (pid != curp->p_pid) { 801 /* must exist and be one of our children... */ 802 if ((p = p_find(pid, PFIND_LOCKED)) == NULL || 803 !p_inferior(p, curp)) { 804 rval = ESRCH; 805 goto done; 806 } 807 /* ... in the same session... */ 808 if (sess != NULL || p->p_session != curp->p_session) 809 goto done; 810 /* ... existing pgid must be in same session ... */ 811 if (pgrp != NULL && pgrp->pg_session != p->p_session) 812 goto done; 813 /* ... and not done an exec. */ 814 if (p->p_flag & PK_EXEC) { 815 rval = EACCES; 816 goto done; 817 } 818 } else { 819 /* ... setsid() cannot re-enter a pgrp */ 820 if (mksess && (curp->p_pgid == curp->p_pid || 821 pg_find(curp->p_pid, PFIND_LOCKED))) 822 goto done; 823 p = curp; 824 } 825 826 /* Changing the process group/session of a session 827 leader is definitely off limits. */ 828 if (SESS_LEADER(p)) { 829 if (sess == NULL && p->p_pgrp == pgrp) 830 /* unless it's a definite noop */ 831 rval = 0; 832 goto done; 833 } 834 835 /* Can only create a process group with id of process */ 836 if (pgrp == NULL && pgid != pid) 837 goto done; 838 839 /* Can only create a session if creating pgrp */ 840 if (sess != NULL && pgrp != NULL) 841 goto done; 842 843 /* Check we allocated memory for a pgrp... */ 844 if (pgrp == NULL && new_pgrp == NULL) 845 goto done; 846 847 /* Don't attach to 'zombie' pgrp */ 848 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members)) 849 goto done; 850 851 /* Expect to succeed now */ 852 rval = 0; 853 854 if (pgrp == p->p_pgrp) 855 /* nothing to do */ 856 goto done; 857 858 /* Ok all setup, link up required structures */ 859 860 if (pgrp == NULL) { 861 pgrp = new_pgrp; 862 new_pgrp = NULL; 863 if (sess != NULL) { 864 sess->s_sid = p->p_pid; 865 sess->s_leader = p; 866 sess->s_count = 1; 867 sess->s_ttyvp = NULL; 868 sess->s_ttyp = NULL; 869 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET; 870 memcpy(sess->s_login, p->p_session->s_login, 871 sizeof(sess->s_login)); 872 p->p_lflag &= ~PL_CONTROLT; 873 } else { 874 sess = p->p_pgrp->pg_session; 875 proc_sesshold(sess); 876 } 877 pgrp->pg_session = sess; 878 sess = NULL; 879 880 pgrp->pg_id = pgid; 881 LIST_INIT(&pgrp->pg_members); 882 #ifdef DIAGNOSTIC 883 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp)) 884 panic("enterpgrp: pgrp table slot in use"); 885 if (__predict_false(mksess && p != curp)) 886 panic("enterpgrp: mksession and p != curproc"); 887 #endif 888 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp; 889 pgrp->pg_jobc = 0; 890 } 891 892 /* 893 * Adjust eligibility of affected pgrps to participate in job control. 894 * Increment eligibility counts before decrementing, otherwise we 895 * could reach 0 spuriously during the first call. 896 */ 897 fixjobc(p, pgrp, 1); 898 fixjobc(p, p->p_pgrp, 0); 899 900 /* Interlock with ttread(). */ 901 mutex_spin_enter(&tty_lock); 902 903 /* Move process to requested group. */ 904 LIST_REMOVE(p, p_pglist); 905 if (LIST_EMPTY(&p->p_pgrp->pg_members)) 906 /* defer delete until we've dumped the lock */ 907 pg_id = p->p_pgrp->pg_id; 908 p->p_pgrp = pgrp; 909 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 910 911 /* Done with the swap; we can release the tty mutex. */ 912 mutex_spin_exit(&tty_lock); 913 914 done: 915 if (pg_id != NO_PGID) { 916 /* Releases proc_lock. */ 917 pg_delete(pg_id); 918 } else { 919 mutex_exit(proc_lock); 920 } 921 if (sess != NULL) 922 kmem_free(sess, sizeof(*sess)); 923 if (new_pgrp != NULL) 924 kmem_free(new_pgrp, sizeof(*new_pgrp)); 925 #ifdef DEBUG_PGRP 926 if (__predict_false(rval)) 927 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n", 928 pid, pgid, mksess, curp->p_pid, rval); 929 #endif 930 return rval; 931 } 932 933 /* 934 * proc_leavepgrp: remove a process from its process group. 935 * => must be called with the proc_lock held, which will be released; 936 */ 937 void 938 proc_leavepgrp(struct proc *p) 939 { 940 struct pgrp *pgrp; 941 942 KASSERT(mutex_owned(proc_lock)); 943 944 /* Interlock with ttread() */ 945 mutex_spin_enter(&tty_lock); 946 pgrp = p->p_pgrp; 947 LIST_REMOVE(p, p_pglist); 948 p->p_pgrp = NULL; 949 mutex_spin_exit(&tty_lock); 950 951 if (LIST_EMPTY(&pgrp->pg_members)) { 952 /* Releases proc_lock. */ 953 pg_delete(pgrp->pg_id); 954 } else { 955 mutex_exit(proc_lock); 956 } 957 } 958 959 /* 960 * pg_remove: remove a process group from the table. 961 * => must be called with the proc_lock held; 962 * => returns process group to free; 963 */ 964 static struct pgrp * 965 pg_remove(pid_t pg_id) 966 { 967 struct pgrp *pgrp; 968 struct pid_table *pt; 969 970 KASSERT(mutex_owned(proc_lock)); 971 972 pt = &pid_table[pg_id & pid_tbl_mask]; 973 pgrp = pt->pt_pgrp; 974 975 KASSERT(pgrp != NULL); 976 KASSERT(pgrp->pg_id == pg_id); 977 KASSERT(LIST_EMPTY(&pgrp->pg_members)); 978 979 pt->pt_pgrp = NULL; 980 981 if (!P_VALID(pt->pt_proc)) { 982 /* Orphaned pgrp, put slot onto free list. */ 983 KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0); 984 pg_id &= pid_tbl_mask; 985 pt = &pid_table[last_free_pt]; 986 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id); 987 last_free_pt = pg_id; 988 pid_alloc_cnt--; 989 } 990 return pgrp; 991 } 992 993 /* 994 * pg_delete: delete and free a process group. 995 * => must be called with the proc_lock held, which will be released. 996 */ 997 static void 998 pg_delete(pid_t pg_id) 999 { 1000 struct pgrp *pg; 1001 struct tty *ttyp; 1002 struct session *ss; 1003 1004 KASSERT(mutex_owned(proc_lock)); 1005 1006 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp; 1007 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) { 1008 mutex_exit(proc_lock); 1009 return; 1010 } 1011 1012 ss = pg->pg_session; 1013 1014 /* Remove reference (if any) from tty to this process group */ 1015 mutex_spin_enter(&tty_lock); 1016 ttyp = ss->s_ttyp; 1017 if (ttyp != NULL && ttyp->t_pgrp == pg) { 1018 ttyp->t_pgrp = NULL; 1019 KASSERT(ttyp->t_session == ss); 1020 } 1021 mutex_spin_exit(&tty_lock); 1022 1023 /* 1024 * The leading process group in a session is freed by proc_sessrele(), 1025 * if last reference. Note: proc_sessrele() releases proc_lock. 1026 */ 1027 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL; 1028 proc_sessrele(ss); 1029 1030 if (pg != NULL) { 1031 /* Free it, if was not done by proc_sessrele(). */ 1032 kmem_free(pg, sizeof(struct pgrp)); 1033 } 1034 } 1035 1036 /* 1037 * Adjust pgrp jobc counters when specified process changes process group. 1038 * We count the number of processes in each process group that "qualify" 1039 * the group for terminal job control (those with a parent in a different 1040 * process group of the same session). If that count reaches zero, the 1041 * process group becomes orphaned. Check both the specified process' 1042 * process group and that of its children. 1043 * entering == 0 => p is leaving specified group. 1044 * entering == 1 => p is entering specified group. 1045 * 1046 * Call with proc_lock held. 1047 */ 1048 void 1049 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 1050 { 1051 struct pgrp *hispgrp; 1052 struct session *mysession = pgrp->pg_session; 1053 struct proc *child; 1054 1055 KASSERT(mutex_owned(proc_lock)); 1056 1057 /* 1058 * Check p's parent to see whether p qualifies its own process 1059 * group; if so, adjust count for p's process group. 1060 */ 1061 hispgrp = p->p_pptr->p_pgrp; 1062 if (hispgrp != pgrp && hispgrp->pg_session == mysession) { 1063 if (entering) { 1064 pgrp->pg_jobc++; 1065 p->p_lflag &= ~PL_ORPHANPG; 1066 } else if (--pgrp->pg_jobc == 0) 1067 orphanpg(pgrp); 1068 } 1069 1070 /* 1071 * Check this process' children to see whether they qualify 1072 * their process groups; if so, adjust counts for children's 1073 * process groups. 1074 */ 1075 LIST_FOREACH(child, &p->p_children, p_sibling) { 1076 hispgrp = child->p_pgrp; 1077 if (hispgrp != pgrp && hispgrp->pg_session == mysession && 1078 !P_ZOMBIE(child)) { 1079 if (entering) { 1080 child->p_lflag &= ~PL_ORPHANPG; 1081 hispgrp->pg_jobc++; 1082 } else if (--hispgrp->pg_jobc == 0) 1083 orphanpg(hispgrp); 1084 } 1085 } 1086 } 1087 1088 /* 1089 * A process group has become orphaned; 1090 * if there are any stopped processes in the group, 1091 * hang-up all process in that group. 1092 * 1093 * Call with proc_lock held. 1094 */ 1095 static void 1096 orphanpg(struct pgrp *pg) 1097 { 1098 struct proc *p; 1099 1100 KASSERT(mutex_owned(proc_lock)); 1101 1102 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1103 if (p->p_stat == SSTOP) { 1104 p->p_lflag |= PL_ORPHANPG; 1105 psignal(p, SIGHUP); 1106 psignal(p, SIGCONT); 1107 } 1108 } 1109 } 1110 1111 #ifdef DDB 1112 #include <ddb/db_output.h> 1113 void pidtbl_dump(void); 1114 void 1115 pidtbl_dump(void) 1116 { 1117 struct pid_table *pt; 1118 struct proc *p; 1119 struct pgrp *pgrp; 1120 int id; 1121 1122 db_printf("pid table %p size %x, next %x, last %x\n", 1123 pid_table, pid_tbl_mask+1, 1124 next_free_pt, last_free_pt); 1125 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) { 1126 p = pt->pt_proc; 1127 if (!P_VALID(p) && !pt->pt_pgrp) 1128 continue; 1129 db_printf(" id %x: ", id); 1130 if (P_VALID(p)) 1131 db_printf("proc %p id %d (0x%x) %s\n", 1132 p, p->p_pid, p->p_pid, p->p_comm); 1133 else 1134 db_printf("next %x use %x\n", 1135 P_NEXT(p) & pid_tbl_mask, 1136 P_NEXT(p) & ~pid_tbl_mask); 1137 if ((pgrp = pt->pt_pgrp)) { 1138 db_printf("\tsession %p, sid %d, count %d, login %s\n", 1139 pgrp->pg_session, pgrp->pg_session->s_sid, 1140 pgrp->pg_session->s_count, 1141 pgrp->pg_session->s_login); 1142 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n", 1143 pgrp, pgrp->pg_id, pgrp->pg_jobc, 1144 LIST_FIRST(&pgrp->pg_members)); 1145 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 1146 db_printf("\t\tpid %d addr %p pgrp %p %s\n", 1147 p->p_pid, p, p->p_pgrp, p->p_comm); 1148 } 1149 } 1150 } 1151 } 1152 #endif /* DDB */ 1153 1154 #ifdef KSTACK_CHECK_MAGIC 1155 #include <sys/user.h> 1156 1157 #define KSTACK_MAGIC 0xdeadbeaf 1158 1159 /* XXX should be per process basis? */ 1160 static int kstackleftmin = KSTACK_SIZE; 1161 static int kstackleftthres = KSTACK_SIZE / 8; 1162 1163 void 1164 kstack_setup_magic(const struct lwp *l) 1165 { 1166 uint32_t *ip; 1167 uint32_t const *end; 1168 1169 KASSERT(l != NULL); 1170 KASSERT(l != &lwp0); 1171 1172 /* 1173 * fill all the stack with magic number 1174 * so that later modification on it can be detected. 1175 */ 1176 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1177 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1178 for (; ip < end; ip++) { 1179 *ip = KSTACK_MAGIC; 1180 } 1181 } 1182 1183 void 1184 kstack_check_magic(const struct lwp *l) 1185 { 1186 uint32_t const *ip, *end; 1187 int stackleft; 1188 1189 KASSERT(l != NULL); 1190 1191 /* don't check proc0 */ /*XXX*/ 1192 if (l == &lwp0) 1193 return; 1194 1195 #ifdef __MACHINE_STACK_GROWS_UP 1196 /* stack grows upwards (eg. hppa) */ 1197 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1198 end = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1199 for (ip--; ip >= end; ip--) 1200 if (*ip != KSTACK_MAGIC) 1201 break; 1202 1203 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip; 1204 #else /* __MACHINE_STACK_GROWS_UP */ 1205 /* stack grows downwards (eg. i386) */ 1206 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1207 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1208 for (; ip < end; ip++) 1209 if (*ip != KSTACK_MAGIC) 1210 break; 1211 1212 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l); 1213 #endif /* __MACHINE_STACK_GROWS_UP */ 1214 1215 if (kstackleftmin > stackleft) { 1216 kstackleftmin = stackleft; 1217 if (stackleft < kstackleftthres) 1218 printf("warning: kernel stack left %d bytes" 1219 "(pid %u:lid %u)\n", stackleft, 1220 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1221 } 1222 1223 if (stackleft <= 0) { 1224 panic("magic on the top of kernel stack changed for " 1225 "pid %u, lid %u: maybe kernel stack overflow", 1226 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1227 } 1228 } 1229 #endif /* KSTACK_CHECK_MAGIC */ 1230 1231 int 1232 proclist_foreach_call(struct proclist *list, 1233 int (*callback)(struct proc *, void *arg), void *arg) 1234 { 1235 struct proc marker; 1236 struct proc *p; 1237 int ret = 0; 1238 1239 marker.p_flag = PK_MARKER; 1240 mutex_enter(proc_lock); 1241 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) { 1242 if (p->p_flag & PK_MARKER) { 1243 p = LIST_NEXT(p, p_list); 1244 continue; 1245 } 1246 LIST_INSERT_AFTER(p, &marker, p_list); 1247 ret = (*callback)(p, arg); 1248 KASSERT(mutex_owned(proc_lock)); 1249 p = LIST_NEXT(&marker, p_list); 1250 LIST_REMOVE(&marker, p_list); 1251 } 1252 mutex_exit(proc_lock); 1253 1254 return ret; 1255 } 1256 1257 int 1258 proc_vmspace_getref(struct proc *p, struct vmspace **vm) 1259 { 1260 1261 /* XXXCDC: how should locking work here? */ 1262 1263 /* curproc exception is for coredump. */ 1264 1265 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) || 1266 (p->p_vmspace->vm_refcnt < 1)) { /* XXX */ 1267 return EFAULT; 1268 } 1269 1270 uvmspace_addref(p->p_vmspace); 1271 *vm = p->p_vmspace; 1272 1273 return 0; 1274 } 1275 1276 /* 1277 * Acquire a write lock on the process credential. 1278 */ 1279 void 1280 proc_crmod_enter(void) 1281 { 1282 struct lwp *l = curlwp; 1283 struct proc *p = l->l_proc; 1284 struct plimit *lim; 1285 kauth_cred_t oc; 1286 char *cn; 1287 1288 /* Reset what needs to be reset in plimit. */ 1289 if (p->p_limit->pl_corename != defcorename) { 1290 lim_privatise(p, false); 1291 lim = p->p_limit; 1292 mutex_enter(&lim->pl_lock); 1293 cn = lim->pl_corename; 1294 lim->pl_corename = defcorename; 1295 mutex_exit(&lim->pl_lock); 1296 if (cn != defcorename) 1297 free(cn, M_TEMP); 1298 } 1299 1300 mutex_enter(p->p_lock); 1301 1302 /* Ensure the LWP cached credentials are up to date. */ 1303 if ((oc = l->l_cred) != p->p_cred) { 1304 kauth_cred_hold(p->p_cred); 1305 l->l_cred = p->p_cred; 1306 kauth_cred_free(oc); 1307 } 1308 1309 } 1310 1311 /* 1312 * Set in a new process credential, and drop the write lock. The credential 1313 * must have a reference already. Optionally, free a no-longer required 1314 * credential. The scheduler also needs to inspect p_cred, so we also 1315 * briefly acquire the sched state mutex. 1316 */ 1317 void 1318 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid) 1319 { 1320 struct lwp *l = curlwp, *l2; 1321 struct proc *p = l->l_proc; 1322 kauth_cred_t oc; 1323 1324 KASSERT(mutex_owned(p->p_lock)); 1325 1326 /* Is there a new credential to set in? */ 1327 if (scred != NULL) { 1328 p->p_cred = scred; 1329 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 1330 if (l2 != l) 1331 l2->l_prflag |= LPR_CRMOD; 1332 } 1333 1334 /* Ensure the LWP cached credentials are up to date. */ 1335 if ((oc = l->l_cred) != scred) { 1336 kauth_cred_hold(scred); 1337 l->l_cred = scred; 1338 } 1339 } else 1340 oc = NULL; /* XXXgcc */ 1341 1342 if (sugid) { 1343 /* 1344 * Mark process as having changed credentials, stops 1345 * tracing etc. 1346 */ 1347 p->p_flag |= PK_SUGID; 1348 } 1349 1350 mutex_exit(p->p_lock); 1351 1352 /* If there is a credential to be released, free it now. */ 1353 if (fcred != NULL) { 1354 KASSERT(scred != NULL); 1355 kauth_cred_free(fcred); 1356 if (oc != scred) 1357 kauth_cred_free(oc); 1358 } 1359 } 1360 1361 /* 1362 * proc_specific_key_create -- 1363 * Create a key for subsystem proc-specific data. 1364 */ 1365 int 1366 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 1367 { 1368 1369 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor)); 1370 } 1371 1372 /* 1373 * proc_specific_key_delete -- 1374 * Delete a key for subsystem proc-specific data. 1375 */ 1376 void 1377 proc_specific_key_delete(specificdata_key_t key) 1378 { 1379 1380 specificdata_key_delete(proc_specificdata_domain, key); 1381 } 1382 1383 /* 1384 * proc_initspecific -- 1385 * Initialize a proc's specificdata container. 1386 */ 1387 void 1388 proc_initspecific(struct proc *p) 1389 { 1390 int error; 1391 1392 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref); 1393 KASSERT(error == 0); 1394 } 1395 1396 /* 1397 * proc_finispecific -- 1398 * Finalize a proc's specificdata container. 1399 */ 1400 void 1401 proc_finispecific(struct proc *p) 1402 { 1403 1404 specificdata_fini(proc_specificdata_domain, &p->p_specdataref); 1405 } 1406 1407 /* 1408 * proc_getspecific -- 1409 * Return proc-specific data corresponding to the specified key. 1410 */ 1411 void * 1412 proc_getspecific(struct proc *p, specificdata_key_t key) 1413 { 1414 1415 return (specificdata_getspecific(proc_specificdata_domain, 1416 &p->p_specdataref, key)); 1417 } 1418 1419 /* 1420 * proc_setspecific -- 1421 * Set proc-specific data corresponding to the specified key. 1422 */ 1423 void 1424 proc_setspecific(struct proc *p, specificdata_key_t key, void *data) 1425 { 1426 1427 specificdata_setspecific(proc_specificdata_domain, 1428 &p->p_specdataref, key, data); 1429 } 1430 1431 int 1432 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target) 1433 { 1434 int r = 0; 1435 1436 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) || 1437 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) { 1438 /* 1439 * suid proc of ours or proc not ours 1440 */ 1441 r = EPERM; 1442 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) { 1443 /* 1444 * sgid proc has sgid back to us temporarily 1445 */ 1446 r = EPERM; 1447 } else { 1448 /* 1449 * our rgid must be in target's group list (ie, 1450 * sub-processes started by a sgid process) 1451 */ 1452 int ismember = 0; 1453 1454 if (kauth_cred_ismember_gid(cred, 1455 kauth_cred_getgid(target), &ismember) != 0 || 1456 !ismember) 1457 r = EPERM; 1458 } 1459 1460 return (r); 1461 } 1462 1463
Indexes created Sun Sep 21 16:09:51 GMT 2025