kern_proc.c revision 1.241
1 /* $NetBSD: kern_proc.c,v 1.241 2020/02/21 00:26:22 joerg 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 1.241 2020/02/21 00:26:22 joerg Exp $"); 66 67 #ifdef _KERNEL_OPT 68 #include "opt_kstack.h" 69 #include "opt_maxuprc.h" 70 #include "opt_dtrace.h" 71 #include "opt_compat_netbsd32.h" 72 #include "opt_kaslr.h" 73 #endif 74 75 #if defined(__HAVE_COMPAT_NETBSD32) && !defined(COMPAT_NETBSD32) \ 76 && !defined(_RUMPKERNEL) 77 #define COMPAT_NETBSD32 78 #endif 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/kernel.h> 83 #include <sys/proc.h> 84 #include <sys/resourcevar.h> 85 #include <sys/buf.h> 86 #include <sys/acct.h> 87 #include <sys/wait.h> 88 #include <sys/file.h> 89 #include <ufs/ufs/quota.h> 90 #include <sys/uio.h> 91 #include <sys/pool.h> 92 #include <sys/pset.h> 93 #include <sys/ioctl.h> 94 #include <sys/tty.h> 95 #include <sys/signalvar.h> 96 #include <sys/ras.h> 97 #include <sys/filedesc.h> 98 #include <sys/syscall_stats.h> 99 #include <sys/kauth.h> 100 #include <sys/sleepq.h> 101 #include <sys/atomic.h> 102 #include <sys/kmem.h> 103 #include <sys/namei.h> 104 #include <sys/dtrace_bsd.h> 105 #include <sys/sysctl.h> 106 #include <sys/exec.h> 107 #include <sys/cpu.h> 108 #include <sys/compat_stub.h> 109 110 #include <uvm/uvm_extern.h> 111 #include <uvm/uvm.h> 112 113 /* 114 * Process lists. 115 */ 116 117 struct proclist allproc __cacheline_aligned; 118 struct proclist zombproc __cacheline_aligned; 119 120 kmutex_t * proc_lock __cacheline_aligned; 121 122 /* 123 * pid to proc lookup is done by indexing the pid_table array. 124 * Since pid numbers are only allocated when an empty slot 125 * has been found, there is no need to search any lists ever. 126 * (an orphaned pgrp will lock the slot, a session will lock 127 * the pgrp with the same number.) 128 * If the table is too small it is reallocated with twice the 129 * previous size and the entries 'unzipped' into the two halves. 130 * A linked list of free entries is passed through the pt_proc 131 * field of 'free' items - set odd to be an invalid ptr. 132 */ 133 134 struct pid_table { 135 struct proc *pt_proc; 136 struct pgrp *pt_pgrp; 137 pid_t pt_pid; 138 }; 139 #if 1 /* strongly typed cast - should be a noop */ 140 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; } 141 #else 142 #define p2u(p) ((uint)p) 143 #endif 144 #define P_VALID(p) (!(p2u(p) & 1)) 145 #define P_NEXT(p) (p2u(p) >> 1) 146 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1)) 147 148 /* 149 * Table of process IDs (PIDs). 150 */ 151 static struct pid_table *pid_table __read_mostly; 152 153 #define INITIAL_PID_TABLE_SIZE (1 << 5) 154 155 /* Table mask, threshold for growing and number of allocated PIDs. */ 156 static u_int pid_tbl_mask __read_mostly; 157 static u_int pid_alloc_lim __read_mostly; 158 static u_int pid_alloc_cnt __cacheline_aligned; 159 160 /* Next free, last free and maximum PIDs. */ 161 static u_int next_free_pt __cacheline_aligned; 162 static u_int last_free_pt __cacheline_aligned; 163 static pid_t pid_max __read_mostly; 164 165 /* Components of the first process -- never freed. */ 166 167 extern struct emul emul_netbsd; /* defined in kern_exec.c */ 168 169 struct session session0 = { 170 .s_count = 1, 171 .s_sid = 0, 172 }; 173 struct pgrp pgrp0 = { 174 .pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members), 175 .pg_session = &session0, 176 }; 177 filedesc_t filedesc0; 178 struct cwdinfo cwdi0 = { 179 .cwdi_cmask = CMASK, 180 .cwdi_refcnt = 1, 181 }; 182 struct plimit limit0; 183 struct pstats pstat0; 184 struct vmspace vmspace0; 185 struct sigacts sigacts0; 186 struct proc proc0 = { 187 .p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps), 188 .p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters), 189 .p_nlwps = 1, 190 .p_nrlwps = 1, 191 .p_nlwpid = 1, /* must match lwp0.l_lid */ 192 .p_pgrp = &pgrp0, 193 .p_comm = "system", 194 /* 195 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8) 196 * when they exit. init(8) can easily wait them out for us. 197 */ 198 .p_flag = PK_SYSTEM | PK_NOCLDWAIT, 199 .p_stat = SACTIVE, 200 .p_nice = NZERO, 201 .p_emul = &emul_netbsd, 202 .p_cwdi = &cwdi0, 203 .p_limit = &limit0, 204 .p_fd = &filedesc0, 205 .p_vmspace = &vmspace0, 206 .p_stats = &pstat0, 207 .p_sigacts = &sigacts0, 208 #ifdef PROC0_MD_INITIALIZERS 209 PROC0_MD_INITIALIZERS 210 #endif 211 }; 212 kauth_cred_t cred0; 213 214 static const int nofile = NOFILE; 215 static const int maxuprc = MAXUPRC; 216 217 static int sysctl_doeproc(SYSCTLFN_PROTO); 218 static int sysctl_kern_proc_args(SYSCTLFN_PROTO); 219 static int sysctl_security_expose_address(SYSCTLFN_PROTO); 220 221 #ifdef KASLR 222 static int kern_expose_address = 0; 223 #else 224 static int kern_expose_address = 1; 225 #endif 226 /* 227 * The process list descriptors, used during pid allocation and 228 * by sysctl. No locking on this data structure is needed since 229 * it is completely static. 230 */ 231 const struct proclist_desc proclists[] = { 232 { &allproc }, 233 { &zombproc }, 234 { NULL }, 235 }; 236 237 static struct pgrp * pg_remove(pid_t); 238 static void pg_delete(pid_t); 239 static void orphanpg(struct pgrp *); 240 241 static specificdata_domain_t proc_specificdata_domain; 242 243 static pool_cache_t proc_cache; 244 245 static kauth_listener_t proc_listener; 246 247 static void fill_proc(const struct proc *, struct proc *, bool); 248 static int fill_pathname(struct lwp *, pid_t, void *, size_t *); 249 static int fill_cwd(struct lwp *, pid_t, void *, size_t *); 250 251 static int 252 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 253 void *arg0, void *arg1, void *arg2, void *arg3) 254 { 255 struct proc *p; 256 int result; 257 258 result = KAUTH_RESULT_DEFER; 259 p = arg0; 260 261 switch (action) { 262 case KAUTH_PROCESS_CANSEE: { 263 enum kauth_process_req req; 264 265 req = (enum kauth_process_req)(uintptr_t)arg1; 266 267 switch (req) { 268 case KAUTH_REQ_PROCESS_CANSEE_ARGS: 269 case KAUTH_REQ_PROCESS_CANSEE_ENTRY: 270 case KAUTH_REQ_PROCESS_CANSEE_OPENFILES: 271 case KAUTH_REQ_PROCESS_CANSEE_EPROC: 272 result = KAUTH_RESULT_ALLOW; 273 break; 274 275 case KAUTH_REQ_PROCESS_CANSEE_ENV: 276 if (kauth_cred_getuid(cred) != 277 kauth_cred_getuid(p->p_cred) || 278 kauth_cred_getuid(cred) != 279 kauth_cred_getsvuid(p->p_cred)) 280 break; 281 282 result = KAUTH_RESULT_ALLOW; 283 284 break; 285 286 case KAUTH_REQ_PROCESS_CANSEE_KPTR: 287 if (!kern_expose_address) 288 break; 289 290 if (kern_expose_address == 1 && !(p->p_flag & PK_KMEM)) 291 break; 292 293 result = KAUTH_RESULT_ALLOW; 294 295 break; 296 297 default: 298 break; 299 } 300 301 break; 302 } 303 304 case KAUTH_PROCESS_FORK: { 305 int lnprocs = (int)(unsigned long)arg2; 306 307 /* 308 * Don't allow a nonprivileged user to use the last few 309 * processes. The variable lnprocs is the current number of 310 * processes, maxproc is the limit. 311 */ 312 if (__predict_false((lnprocs >= maxproc - 5))) 313 break; 314 315 result = KAUTH_RESULT_ALLOW; 316 317 break; 318 } 319 320 case KAUTH_PROCESS_CORENAME: 321 case KAUTH_PROCESS_STOPFLAG: 322 if (proc_uidmatch(cred, p->p_cred) == 0) 323 result = KAUTH_RESULT_ALLOW; 324 325 break; 326 327 default: 328 break; 329 } 330 331 return result; 332 } 333 334 static int 335 proc_ctor(void *arg __unused, void *obj, int flags __unused) 336 { 337 memset(obj, 0, sizeof(struct proc)); 338 return 0; 339 } 340 341 /* 342 * Initialize global process hashing structures. 343 */ 344 void 345 procinit(void) 346 { 347 const struct proclist_desc *pd; 348 u_int i; 349 #define LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1)) 350 351 for (pd = proclists; pd->pd_list != NULL; pd++) 352 LIST_INIT(pd->pd_list); 353 354 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 355 pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE 356 * sizeof(struct pid_table), KM_SLEEP); 357 pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1; 358 pid_max = PID_MAX; 359 360 /* Set free list running through table... 361 Preset 'use count' above PID_MAX so we allocate pid 1 next. */ 362 for (i = 0; i <= pid_tbl_mask; i++) { 363 pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1); 364 pid_table[i].pt_pgrp = 0; 365 pid_table[i].pt_pid = 0; 366 } 367 /* slot 0 is just grabbed */ 368 next_free_pt = 1; 369 /* Need to fix last entry. */ 370 last_free_pt = pid_tbl_mask; 371 pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY); 372 /* point at which we grow table - to avoid reusing pids too often */ 373 pid_alloc_lim = pid_tbl_mask - 1; 374 #undef LINK_EMPTY 375 376 proc_specificdata_domain = specificdata_domain_create(); 377 KASSERT(proc_specificdata_domain != NULL); 378 379 proc_cache = pool_cache_init(sizeof(struct proc), coherency_unit, 0, 0, 380 "procpl", NULL, IPL_NONE, proc_ctor, NULL, NULL); 381 382 proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 383 proc_listener_cb, NULL); 384 } 385 386 void 387 procinit_sysctl(void) 388 { 389 static struct sysctllog *clog; 390 391 sysctl_createv(&clog, 0, NULL, NULL, 392 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 393 CTLTYPE_INT, "expose_address", 394 SYSCTL_DESCR("Enable exposing kernel addresses"), 395 sysctl_security_expose_address, 0, 396 &kern_expose_address, 0, CTL_KERN, CTL_CREATE, CTL_EOL); 397 sysctl_createv(&clog, 0, NULL, NULL, 398 CTLFLAG_PERMANENT, 399 CTLTYPE_NODE, "proc", 400 SYSCTL_DESCR("System-wide process information"), 401 sysctl_doeproc, 0, NULL, 0, 402 CTL_KERN, KERN_PROC, CTL_EOL); 403 sysctl_createv(&clog, 0, NULL, NULL, 404 CTLFLAG_PERMANENT, 405 CTLTYPE_NODE, "proc2", 406 SYSCTL_DESCR("Machine-independent process information"), 407 sysctl_doeproc, 0, NULL, 0, 408 CTL_KERN, KERN_PROC2, CTL_EOL); 409 sysctl_createv(&clog, 0, NULL, NULL, 410 CTLFLAG_PERMANENT, 411 CTLTYPE_NODE, "proc_args", 412 SYSCTL_DESCR("Process argument information"), 413 sysctl_kern_proc_args, 0, NULL, 0, 414 CTL_KERN, KERN_PROC_ARGS, CTL_EOL); 415 416 /* 417 "nodes" under these: 418 419 KERN_PROC_ALL 420 KERN_PROC_PID pid 421 KERN_PROC_PGRP pgrp 422 KERN_PROC_SESSION sess 423 KERN_PROC_TTY tty 424 KERN_PROC_UID uid 425 KERN_PROC_RUID uid 426 KERN_PROC_GID gid 427 KERN_PROC_RGID gid 428 429 all in all, probably not worth the effort... 430 */ 431 } 432 433 /* 434 * Initialize process 0. 435 */ 436 void 437 proc0_init(void) 438 { 439 struct proc *p; 440 struct pgrp *pg; 441 struct rlimit *rlim; 442 rlim_t lim; 443 int error __diagused; 444 int i; 445 446 p = &proc0; 447 pg = &pgrp0; 448 449 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 450 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 451 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 452 453 rw_init(&p->p_reflock); 454 rw_init(&p->p_treelock); 455 cv_init(&p->p_waitcv, "wait"); 456 cv_init(&p->p_lwpcv, "lwpwait"); 457 458 LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling); 459 radix_tree_init_tree(&p->p_lwptree); 460 error = radix_tree_insert_node(&p->p_lwptree, 461 (uint64_t)(lwp0.l_lid - 1), &lwp0); 462 KASSERT(error == 0); 463 464 pid_table[0].pt_proc = p; 465 LIST_INSERT_HEAD(&allproc, p, p_list); 466 467 pid_table[0].pt_pgrp = pg; 468 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist); 469 470 #ifdef __HAVE_SYSCALL_INTERN 471 (*p->p_emul->e_syscall_intern)(p); 472 #endif 473 474 /* Create credentials. */ 475 cred0 = kauth_cred_alloc(); 476 p->p_cred = cred0; 477 478 /* Create the CWD info. */ 479 rw_init(&cwdi0.cwdi_lock); 480 481 /* Create the limits structures. */ 482 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE); 483 484 rlim = limit0.pl_rlimit; 485 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) { 486 rlim[i].rlim_cur = RLIM_INFINITY; 487 rlim[i].rlim_max = RLIM_INFINITY; 488 } 489 490 rlim[RLIMIT_NOFILE].rlim_max = maxfiles; 491 rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile; 492 493 rlim[RLIMIT_NPROC].rlim_max = maxproc; 494 rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc; 495 496 lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvm_availmem())); 497 rlim[RLIMIT_RSS].rlim_max = lim; 498 rlim[RLIMIT_MEMLOCK].rlim_max = lim; 499 rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 500 501 rlim[RLIMIT_NTHR].rlim_max = maxlwp; 502 rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc; 503 504 /* Note that default core name has zero length. */ 505 limit0.pl_corename = defcorename; 506 limit0.pl_cnlen = 0; 507 limit0.pl_refcnt = 1; 508 limit0.pl_writeable = false; 509 limit0.pl_sv_limit = NULL; 510 511 /* Configure virtual memory system, set vm rlimits. */ 512 uvm_init_limits(p); 513 514 /* Initialize file descriptor table for proc0. */ 515 fd_init(&filedesc0); 516 517 /* 518 * Initialize proc0's vmspace, which uses the kernel pmap. 519 * All kernel processes (which never have user space mappings) 520 * share proc0's vmspace, and thus, the kernel pmap. 521 */ 522 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 523 trunc_page(VM_MAXUSER_ADDRESS), 524 #ifdef __USE_TOPDOWN_VM 525 true 526 #else 527 false 528 #endif 529 ); 530 531 /* Initialize signal state for proc0. XXX IPL_SCHED */ 532 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED); 533 siginit(p); 534 535 proc_initspecific(p); 536 kdtrace_proc_ctor(NULL, p); 537 } 538 539 /* 540 * Session reference counting. 541 */ 542 543 void 544 proc_sesshold(struct session *ss) 545 { 546 547 KASSERT(mutex_owned(proc_lock)); 548 ss->s_count++; 549 } 550 551 void 552 proc_sessrele(struct session *ss) 553 { 554 555 KASSERT(mutex_owned(proc_lock)); 556 /* 557 * We keep the pgrp with the same id as the session in order to 558 * stop a process being given the same pid. Since the pgrp holds 559 * a reference to the session, it must be a 'zombie' pgrp by now. 560 */ 561 if (--ss->s_count == 0) { 562 struct pgrp *pg; 563 564 pg = pg_remove(ss->s_sid); 565 mutex_exit(proc_lock); 566 567 kmem_free(pg, sizeof(struct pgrp)); 568 kmem_free(ss, sizeof(struct session)); 569 } else { 570 mutex_exit(proc_lock); 571 } 572 } 573 574 /* 575 * Check that the specified process group is in the session of the 576 * specified process. 577 * Treats -ve ids as process ids. 578 * Used to validate TIOCSPGRP requests. 579 */ 580 int 581 pgid_in_session(struct proc *p, pid_t pg_id) 582 { 583 struct pgrp *pgrp; 584 struct session *session; 585 int error; 586 587 mutex_enter(proc_lock); 588 if (pg_id < 0) { 589 struct proc *p1 = proc_find(-pg_id); 590 if (p1 == NULL) { 591 error = EINVAL; 592 goto fail; 593 } 594 pgrp = p1->p_pgrp; 595 } else { 596 pgrp = pgrp_find(pg_id); 597 if (pgrp == NULL) { 598 error = EINVAL; 599 goto fail; 600 } 601 } 602 session = pgrp->pg_session; 603 error = (session != p->p_pgrp->pg_session) ? EPERM : 0; 604 fail: 605 mutex_exit(proc_lock); 606 return error; 607 } 608 609 /* 610 * p_inferior: is p an inferior of q? 611 */ 612 static inline bool 613 p_inferior(struct proc *p, struct proc *q) 614 { 615 616 KASSERT(mutex_owned(proc_lock)); 617 618 for (; p != q; p = p->p_pptr) 619 if (p->p_pid == 0) 620 return false; 621 return true; 622 } 623 624 /* 625 * proc_find: locate a process by the ID. 626 * 627 * => Must be called with proc_lock held. 628 */ 629 proc_t * 630 proc_find_raw(pid_t pid) 631 { 632 struct pid_table *pt; 633 proc_t *p; 634 635 KASSERT(mutex_owned(proc_lock)); 636 pt = &pid_table[pid & pid_tbl_mask]; 637 p = pt->pt_proc; 638 if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) { 639 return NULL; 640 } 641 return p; 642 } 643 644 proc_t * 645 proc_find(pid_t pid) 646 { 647 proc_t *p; 648 649 p = proc_find_raw(pid); 650 if (__predict_false(p == NULL)) { 651 return NULL; 652 } 653 654 /* 655 * Only allow live processes to be found by PID. 656 * XXX: p_stat might change, since unlocked. 657 */ 658 if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) { 659 return p; 660 } 661 return NULL; 662 } 663 664 /* 665 * pgrp_find: locate a process group by the ID. 666 * 667 * => Must be called with proc_lock held. 668 */ 669 struct pgrp * 670 pgrp_find(pid_t pgid) 671 { 672 struct pgrp *pg; 673 674 KASSERT(mutex_owned(proc_lock)); 675 676 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp; 677 678 /* 679 * Cannot look up a process group that only exists because the 680 * session has not died yet (traditional). 681 */ 682 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) { 683 return NULL; 684 } 685 return pg; 686 } 687 688 static void 689 expand_pid_table(void) 690 { 691 size_t pt_size, tsz; 692 struct pid_table *n_pt, *new_pt; 693 struct proc *proc; 694 struct pgrp *pgrp; 695 pid_t pid, rpid; 696 u_int i; 697 uint new_pt_mask; 698 699 pt_size = pid_tbl_mask + 1; 700 tsz = pt_size * 2 * sizeof(struct pid_table); 701 new_pt = kmem_alloc(tsz, KM_SLEEP); 702 new_pt_mask = pt_size * 2 - 1; 703 704 mutex_enter(proc_lock); 705 if (pt_size != pid_tbl_mask + 1) { 706 /* Another process beat us to it... */ 707 mutex_exit(proc_lock); 708 kmem_free(new_pt, tsz); 709 return; 710 } 711 712 /* 713 * Copy entries from old table into new one. 714 * If 'pid' is 'odd' we need to place in the upper half, 715 * even pid's to the lower half. 716 * Free items stay in the low half so we don't have to 717 * fixup the reference to them. 718 * We stuff free items on the front of the freelist 719 * because we can't write to unmodified entries. 720 * Processing the table backwards maintains a semblance 721 * of issuing pid numbers that increase with time. 722 */ 723 i = pt_size - 1; 724 n_pt = new_pt + i; 725 for (; ; i--, n_pt--) { 726 proc = pid_table[i].pt_proc; 727 pgrp = pid_table[i].pt_pgrp; 728 if (!P_VALID(proc)) { 729 /* Up 'use count' so that link is valid */ 730 pid = (P_NEXT(proc) + pt_size) & ~pt_size; 731 rpid = 0; 732 proc = P_FREE(pid); 733 if (pgrp) 734 pid = pgrp->pg_id; 735 } else { 736 pid = pid_table[i].pt_pid; 737 rpid = pid; 738 } 739 740 /* Save entry in appropriate half of table */ 741 n_pt[pid & pt_size].pt_proc = proc; 742 n_pt[pid & pt_size].pt_pgrp = pgrp; 743 n_pt[pid & pt_size].pt_pid = rpid; 744 745 /* Put other piece on start of free list */ 746 pid = (pid ^ pt_size) & ~pid_tbl_mask; 747 n_pt[pid & pt_size].pt_proc = 748 P_FREE((pid & ~pt_size) | next_free_pt); 749 n_pt[pid & pt_size].pt_pgrp = 0; 750 n_pt[pid & pt_size].pt_pid = 0; 751 752 next_free_pt = i | (pid & pt_size); 753 if (i == 0) 754 break; 755 } 756 757 /* Save old table size and switch tables */ 758 tsz = pt_size * sizeof(struct pid_table); 759 n_pt = pid_table; 760 pid_table = new_pt; 761 pid_tbl_mask = new_pt_mask; 762 763 /* 764 * pid_max starts as PID_MAX (= 30000), once we have 16384 765 * allocated pids we need it to be larger! 766 */ 767 if (pid_tbl_mask > PID_MAX) { 768 pid_max = pid_tbl_mask * 2 + 1; 769 pid_alloc_lim |= pid_alloc_lim << 1; 770 } else 771 pid_alloc_lim <<= 1; /* doubles number of free slots... */ 772 773 mutex_exit(proc_lock); 774 kmem_free(n_pt, tsz); 775 } 776 777 struct proc * 778 proc_alloc(void) 779 { 780 struct proc *p; 781 782 p = pool_cache_get(proc_cache, PR_WAITOK); 783 p->p_stat = SIDL; /* protect against others */ 784 proc_initspecific(p); 785 kdtrace_proc_ctor(NULL, p); 786 p->p_pid = -1; 787 proc_alloc_pid(p); 788 return p; 789 } 790 791 /* 792 * proc_alloc_pid: allocate PID and record the given proc 'p' so that 793 * proc_find_raw() can find it by the PID. 794 */ 795 796 pid_t 797 proc_alloc_pid(struct proc *p) 798 { 799 struct pid_table *pt; 800 pid_t pid; 801 int nxt; 802 803 for (;;expand_pid_table()) { 804 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) 805 /* ensure pids cycle through 2000+ values */ 806 continue; 807 mutex_enter(proc_lock); 808 pt = &pid_table[next_free_pt]; 809 #ifdef DIAGNOSTIC 810 if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp)) 811 panic("proc_alloc: slot busy"); 812 #endif 813 nxt = P_NEXT(pt->pt_proc); 814 if (nxt & pid_tbl_mask) 815 break; 816 /* Table full - expand (NB last entry not used....) */ 817 mutex_exit(proc_lock); 818 } 819 820 /* pid is 'saved use count' + 'size' + entry */ 821 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt; 822 if ((uint)pid > (uint)pid_max) 823 pid &= pid_tbl_mask; 824 next_free_pt = nxt & pid_tbl_mask; 825 826 /* Grab table slot */ 827 pt->pt_proc = p; 828 829 KASSERT(pt->pt_pid == 0); 830 pt->pt_pid = pid; 831 if (p->p_pid == -1) { 832 p->p_pid = pid; 833 } 834 pid_alloc_cnt++; 835 mutex_exit(proc_lock); 836 837 return pid; 838 } 839 840 /* 841 * Free a process id - called from proc_free (in kern_exit.c) 842 * 843 * Called with the proc_lock held. 844 */ 845 void 846 proc_free_pid(pid_t pid) 847 { 848 struct pid_table *pt; 849 850 KASSERT(mutex_owned(proc_lock)); 851 852 pt = &pid_table[pid & pid_tbl_mask]; 853 854 /* save pid use count in slot */ 855 pt->pt_proc = P_FREE(pid & ~pid_tbl_mask); 856 KASSERT(pt->pt_pid == pid); 857 pt->pt_pid = 0; 858 859 if (pt->pt_pgrp == NULL) { 860 /* link last freed entry onto ours */ 861 pid &= pid_tbl_mask; 862 pt = &pid_table[last_free_pt]; 863 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid); 864 pt->pt_pid = 0; 865 last_free_pt = pid; 866 pid_alloc_cnt--; 867 } 868 869 atomic_dec_uint(&nprocs); 870 } 871 872 void 873 proc_free_mem(struct proc *p) 874 { 875 876 kdtrace_proc_dtor(NULL, p); 877 pool_cache_put(proc_cache, p); 878 } 879 880 /* 881 * proc_enterpgrp: move p to a new or existing process group (and session). 882 * 883 * If we are creating a new pgrp, the pgid should equal 884 * the calling process' pid. 885 * If is only valid to enter a process group that is in the session 886 * of the process. 887 * Also mksess should only be set if we are creating a process group 888 * 889 * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return. 890 */ 891 int 892 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess) 893 { 894 struct pgrp *new_pgrp, *pgrp; 895 struct session *sess; 896 struct proc *p; 897 int rval; 898 pid_t pg_id = NO_PGID; 899 900 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL; 901 902 /* Allocate data areas we might need before doing any validity checks */ 903 mutex_enter(proc_lock); /* Because pid_table might change */ 904 if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) { 905 mutex_exit(proc_lock); 906 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP); 907 mutex_enter(proc_lock); 908 } else 909 new_pgrp = NULL; 910 rval = EPERM; /* most common error (to save typing) */ 911 912 /* Check pgrp exists or can be created */ 913 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp; 914 if (pgrp != NULL && pgrp->pg_id != pgid) 915 goto done; 916 917 /* Can only set another process under restricted circumstances. */ 918 if (pid != curp->p_pid) { 919 /* Must exist and be one of our children... */ 920 p = proc_find(pid); 921 if (p == NULL || !p_inferior(p, curp)) { 922 rval = ESRCH; 923 goto done; 924 } 925 /* ... in the same session... */ 926 if (sess != NULL || p->p_session != curp->p_session) 927 goto done; 928 /* ... existing pgid must be in same session ... */ 929 if (pgrp != NULL && pgrp->pg_session != p->p_session) 930 goto done; 931 /* ... and not done an exec. */ 932 if (p->p_flag & PK_EXEC) { 933 rval = EACCES; 934 goto done; 935 } 936 } else { 937 /* ... setsid() cannot re-enter a pgrp */ 938 if (mksess && (curp->p_pgid == curp->p_pid || 939 pgrp_find(curp->p_pid))) 940 goto done; 941 p = curp; 942 } 943 944 /* Changing the process group/session of a session 945 leader is definitely off limits. */ 946 if (SESS_LEADER(p)) { 947 if (sess == NULL && p->p_pgrp == pgrp) 948 /* unless it's a definite noop */ 949 rval = 0; 950 goto done; 951 } 952 953 /* Can only create a process group with id of process */ 954 if (pgrp == NULL && pgid != pid) 955 goto done; 956 957 /* Can only create a session if creating pgrp */ 958 if (sess != NULL && pgrp != NULL) 959 goto done; 960 961 /* Check we allocated memory for a pgrp... */ 962 if (pgrp == NULL && new_pgrp == NULL) 963 goto done; 964 965 /* Don't attach to 'zombie' pgrp */ 966 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members)) 967 goto done; 968 969 /* Expect to succeed now */ 970 rval = 0; 971 972 if (pgrp == p->p_pgrp) 973 /* nothing to do */ 974 goto done; 975 976 /* Ok all setup, link up required structures */ 977 978 if (pgrp == NULL) { 979 pgrp = new_pgrp; 980 new_pgrp = NULL; 981 if (sess != NULL) { 982 sess->s_sid = p->p_pid; 983 sess->s_leader = p; 984 sess->s_count = 1; 985 sess->s_ttyvp = NULL; 986 sess->s_ttyp = NULL; 987 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET; 988 memcpy(sess->s_login, p->p_session->s_login, 989 sizeof(sess->s_login)); 990 p->p_lflag &= ~PL_CONTROLT; 991 } else { 992 sess = p->p_pgrp->pg_session; 993 proc_sesshold(sess); 994 } 995 pgrp->pg_session = sess; 996 sess = NULL; 997 998 pgrp->pg_id = pgid; 999 LIST_INIT(&pgrp->pg_members); 1000 #ifdef DIAGNOSTIC 1001 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp)) 1002 panic("enterpgrp: pgrp table slot in use"); 1003 if (__predict_false(mksess && p != curp)) 1004 panic("enterpgrp: mksession and p != curproc"); 1005 #endif 1006 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp; 1007 pgrp->pg_jobc = 0; 1008 } 1009 1010 /* 1011 * Adjust eligibility of affected pgrps to participate in job control. 1012 * Increment eligibility counts before decrementing, otherwise we 1013 * could reach 0 spuriously during the first call. 1014 */ 1015 fixjobc(p, pgrp, 1); 1016 fixjobc(p, p->p_pgrp, 0); 1017 1018 /* Interlock with ttread(). */ 1019 mutex_spin_enter(&tty_lock); 1020 1021 /* Move process to requested group. */ 1022 LIST_REMOVE(p, p_pglist); 1023 if (LIST_EMPTY(&p->p_pgrp->pg_members)) 1024 /* defer delete until we've dumped the lock */ 1025 pg_id = p->p_pgrp->pg_id; 1026 p->p_pgrp = pgrp; 1027 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 1028 1029 /* Done with the swap; we can release the tty mutex. */ 1030 mutex_spin_exit(&tty_lock); 1031 1032 done: 1033 if (pg_id != NO_PGID) { 1034 /* Releases proc_lock. */ 1035 pg_delete(pg_id); 1036 } else { 1037 mutex_exit(proc_lock); 1038 } 1039 if (sess != NULL) 1040 kmem_free(sess, sizeof(*sess)); 1041 if (new_pgrp != NULL) 1042 kmem_free(new_pgrp, sizeof(*new_pgrp)); 1043 #ifdef DEBUG_PGRP 1044 if (__predict_false(rval)) 1045 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n", 1046 pid, pgid, mksess, curp->p_pid, rval); 1047 #endif 1048 return rval; 1049 } 1050 1051 /* 1052 * proc_leavepgrp: remove a process from its process group. 1053 * => must be called with the proc_lock held, which will be released; 1054 */ 1055 void 1056 proc_leavepgrp(struct proc *p) 1057 { 1058 struct pgrp *pgrp; 1059 1060 KASSERT(mutex_owned(proc_lock)); 1061 1062 /* Interlock with ttread() */ 1063 mutex_spin_enter(&tty_lock); 1064 pgrp = p->p_pgrp; 1065 LIST_REMOVE(p, p_pglist); 1066 p->p_pgrp = NULL; 1067 mutex_spin_exit(&tty_lock); 1068 1069 if (LIST_EMPTY(&pgrp->pg_members)) { 1070 /* Releases proc_lock. */ 1071 pg_delete(pgrp->pg_id); 1072 } else { 1073 mutex_exit(proc_lock); 1074 } 1075 } 1076 1077 /* 1078 * pg_remove: remove a process group from the table. 1079 * => must be called with the proc_lock held; 1080 * => returns process group to free; 1081 */ 1082 static struct pgrp * 1083 pg_remove(pid_t pg_id) 1084 { 1085 struct pgrp *pgrp; 1086 struct pid_table *pt; 1087 1088 KASSERT(mutex_owned(proc_lock)); 1089 1090 pt = &pid_table[pg_id & pid_tbl_mask]; 1091 pgrp = pt->pt_pgrp; 1092 1093 KASSERT(pgrp != NULL); 1094 KASSERT(pgrp->pg_id == pg_id); 1095 KASSERT(LIST_EMPTY(&pgrp->pg_members)); 1096 1097 pt->pt_pgrp = NULL; 1098 1099 if (!P_VALID(pt->pt_proc)) { 1100 /* Orphaned pgrp, put slot onto free list. */ 1101 KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0); 1102 pg_id &= pid_tbl_mask; 1103 pt = &pid_table[last_free_pt]; 1104 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id); 1105 KASSERT(pt->pt_pid == 0); 1106 last_free_pt = pg_id; 1107 pid_alloc_cnt--; 1108 } 1109 return pgrp; 1110 } 1111 1112 /* 1113 * pg_delete: delete and free a process group. 1114 * => must be called with the proc_lock held, which will be released. 1115 */ 1116 static void 1117 pg_delete(pid_t pg_id) 1118 { 1119 struct pgrp *pg; 1120 struct tty *ttyp; 1121 struct session *ss; 1122 1123 KASSERT(mutex_owned(proc_lock)); 1124 1125 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp; 1126 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) { 1127 mutex_exit(proc_lock); 1128 return; 1129 } 1130 1131 ss = pg->pg_session; 1132 1133 /* Remove reference (if any) from tty to this process group */ 1134 mutex_spin_enter(&tty_lock); 1135 ttyp = ss->s_ttyp; 1136 if (ttyp != NULL && ttyp->t_pgrp == pg) { 1137 ttyp->t_pgrp = NULL; 1138 KASSERT(ttyp->t_session == ss); 1139 } 1140 mutex_spin_exit(&tty_lock); 1141 1142 /* 1143 * The leading process group in a session is freed by proc_sessrele(), 1144 * if last reference. Note: proc_sessrele() releases proc_lock. 1145 */ 1146 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL; 1147 proc_sessrele(ss); 1148 1149 if (pg != NULL) { 1150 /* Free it, if was not done by proc_sessrele(). */ 1151 kmem_free(pg, sizeof(struct pgrp)); 1152 } 1153 } 1154 1155 /* 1156 * Adjust pgrp jobc counters when specified process changes process group. 1157 * We count the number of processes in each process group that "qualify" 1158 * the group for terminal job control (those with a parent in a different 1159 * process group of the same session). If that count reaches zero, the 1160 * process group becomes orphaned. Check both the specified process' 1161 * process group and that of its children. 1162 * entering == 0 => p is leaving specified group. 1163 * entering == 1 => p is entering specified group. 1164 * 1165 * Call with proc_lock held. 1166 */ 1167 void 1168 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 1169 { 1170 struct pgrp *hispgrp; 1171 struct session *mysession = pgrp->pg_session; 1172 struct proc *child; 1173 1174 KASSERT(mutex_owned(proc_lock)); 1175 1176 /* 1177 * Check p's parent to see whether p qualifies its own process 1178 * group; if so, adjust count for p's process group. 1179 */ 1180 hispgrp = p->p_pptr->p_pgrp; 1181 if (hispgrp != pgrp && hispgrp->pg_session == mysession) { 1182 if (entering) { 1183 pgrp->pg_jobc++; 1184 p->p_lflag &= ~PL_ORPHANPG; 1185 } else if (--pgrp->pg_jobc == 0) 1186 orphanpg(pgrp); 1187 } 1188 1189 /* 1190 * Check this process' children to see whether they qualify 1191 * their process groups; if so, adjust counts for children's 1192 * process groups. 1193 */ 1194 LIST_FOREACH(child, &p->p_children, p_sibling) { 1195 hispgrp = child->p_pgrp; 1196 if (hispgrp != pgrp && hispgrp->pg_session == mysession && 1197 !P_ZOMBIE(child)) { 1198 if (entering) { 1199 child->p_lflag &= ~PL_ORPHANPG; 1200 hispgrp->pg_jobc++; 1201 } else if (--hispgrp->pg_jobc == 0) 1202 orphanpg(hispgrp); 1203 } 1204 } 1205 } 1206 1207 /* 1208 * A process group has become orphaned; 1209 * if there are any stopped processes in the group, 1210 * hang-up all process in that group. 1211 * 1212 * Call with proc_lock held. 1213 */ 1214 static void 1215 orphanpg(struct pgrp *pg) 1216 { 1217 struct proc *p; 1218 1219 KASSERT(mutex_owned(proc_lock)); 1220 1221 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1222 if (p->p_stat == SSTOP) { 1223 p->p_lflag |= PL_ORPHANPG; 1224 psignal(p, SIGHUP); 1225 psignal(p, SIGCONT); 1226 } 1227 } 1228 } 1229 1230 #ifdef DDB 1231 #include <ddb/db_output.h> 1232 void pidtbl_dump(void); 1233 void 1234 pidtbl_dump(void) 1235 { 1236 struct pid_table *pt; 1237 struct proc *p; 1238 struct pgrp *pgrp; 1239 int id; 1240 1241 db_printf("pid table %p size %x, next %x, last %x\n", 1242 pid_table, pid_tbl_mask+1, 1243 next_free_pt, last_free_pt); 1244 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) { 1245 p = pt->pt_proc; 1246 if (!P_VALID(p) && !pt->pt_pgrp) 1247 continue; 1248 db_printf(" id %x: ", id); 1249 if (P_VALID(p)) 1250 db_printf("slotpid %d proc %p id %d (0x%x) %s\n", 1251 pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm); 1252 else 1253 db_printf("next %x use %x\n", 1254 P_NEXT(p) & pid_tbl_mask, 1255 P_NEXT(p) & ~pid_tbl_mask); 1256 if ((pgrp = pt->pt_pgrp)) { 1257 db_printf("\tsession %p, sid %d, count %d, login %s\n", 1258 pgrp->pg_session, pgrp->pg_session->s_sid, 1259 pgrp->pg_session->s_count, 1260 pgrp->pg_session->s_login); 1261 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n", 1262 pgrp, pgrp->pg_id, pgrp->pg_jobc, 1263 LIST_FIRST(&pgrp->pg_members)); 1264 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 1265 db_printf("\t\tpid %d addr %p pgrp %p %s\n", 1266 p->p_pid, p, p->p_pgrp, p->p_comm); 1267 } 1268 } 1269 } 1270 } 1271 #endif /* DDB */ 1272 1273 #ifdef KSTACK_CHECK_MAGIC 1274 1275 #define KSTACK_MAGIC 0xdeadbeaf 1276 1277 /* XXX should be per process basis? */ 1278 static int kstackleftmin = KSTACK_SIZE; 1279 static int kstackleftthres = KSTACK_SIZE / 8; 1280 1281 void 1282 kstack_setup_magic(const struct lwp *l) 1283 { 1284 uint32_t *ip; 1285 uint32_t const *end; 1286 1287 KASSERT(l != NULL); 1288 KASSERT(l != &lwp0); 1289 1290 /* 1291 * fill all the stack with magic number 1292 * so that later modification on it can be detected. 1293 */ 1294 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1295 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1296 for (; ip < end; ip++) { 1297 *ip = KSTACK_MAGIC; 1298 } 1299 } 1300 1301 void 1302 kstack_check_magic(const struct lwp *l) 1303 { 1304 uint32_t const *ip, *end; 1305 int stackleft; 1306 1307 KASSERT(l != NULL); 1308 1309 /* don't check proc0 */ /*XXX*/ 1310 if (l == &lwp0) 1311 return; 1312 1313 #ifdef __MACHINE_STACK_GROWS_UP 1314 /* stack grows upwards (eg. hppa) */ 1315 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1316 end = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1317 for (ip--; ip >= end; ip--) 1318 if (*ip != KSTACK_MAGIC) 1319 break; 1320 1321 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip; 1322 #else /* __MACHINE_STACK_GROWS_UP */ 1323 /* stack grows downwards (eg. i386) */ 1324 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1325 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1326 for (; ip < end; ip++) 1327 if (*ip != KSTACK_MAGIC) 1328 break; 1329 1330 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l); 1331 #endif /* __MACHINE_STACK_GROWS_UP */ 1332 1333 if (kstackleftmin > stackleft) { 1334 kstackleftmin = stackleft; 1335 if (stackleft < kstackleftthres) 1336 printf("warning: kernel stack left %d bytes" 1337 "(pid %u:lid %u)\n", stackleft, 1338 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1339 } 1340 1341 if (stackleft <= 0) { 1342 panic("magic on the top of kernel stack changed for " 1343 "pid %u, lid %u: maybe kernel stack overflow", 1344 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1345 } 1346 } 1347 #endif /* KSTACK_CHECK_MAGIC */ 1348 1349 int 1350 proclist_foreach_call(struct proclist *list, 1351 int (*callback)(struct proc *, void *arg), void *arg) 1352 { 1353 struct proc marker; 1354 struct proc *p; 1355 int ret = 0; 1356 1357 marker.p_flag = PK_MARKER; 1358 mutex_enter(proc_lock); 1359 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) { 1360 if (p->p_flag & PK_MARKER) { 1361 p = LIST_NEXT(p, p_list); 1362 continue; 1363 } 1364 LIST_INSERT_AFTER(p, &marker, p_list); 1365 ret = (*callback)(p, arg); 1366 KASSERT(mutex_owned(proc_lock)); 1367 p = LIST_NEXT(&marker, p_list); 1368 LIST_REMOVE(&marker, p_list); 1369 } 1370 mutex_exit(proc_lock); 1371 1372 return ret; 1373 } 1374 1375 int 1376 proc_vmspace_getref(struct proc *p, struct vmspace **vm) 1377 { 1378 1379 /* XXXCDC: how should locking work here? */ 1380 1381 /* curproc exception is for coredump. */ 1382 1383 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) || 1384 (p->p_vmspace->vm_refcnt < 1)) { /* XXX */ 1385 return EFAULT; 1386 } 1387 1388 uvmspace_addref(p->p_vmspace); 1389 *vm = p->p_vmspace; 1390 1391 return 0; 1392 } 1393 1394 /* 1395 * Acquire a write lock on the process credential. 1396 */ 1397 void 1398 proc_crmod_enter(void) 1399 { 1400 struct lwp *l = curlwp; 1401 struct proc *p = l->l_proc; 1402 kauth_cred_t oc; 1403 1404 /* Reset what needs to be reset in plimit. */ 1405 if (p->p_limit->pl_corename != defcorename) { 1406 lim_setcorename(p, defcorename, 0); 1407 } 1408 1409 mutex_enter(p->p_lock); 1410 1411 /* Ensure the LWP cached credentials are up to date. */ 1412 if ((oc = l->l_cred) != p->p_cred) { 1413 kauth_cred_hold(p->p_cred); 1414 l->l_cred = p->p_cred; 1415 kauth_cred_free(oc); 1416 } 1417 } 1418 1419 /* 1420 * Set in a new process credential, and drop the write lock. The credential 1421 * must have a reference already. Optionally, free a no-longer required 1422 * credential. The scheduler also needs to inspect p_cred, so we also 1423 * briefly acquire the sched state mutex. 1424 */ 1425 void 1426 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid) 1427 { 1428 struct lwp *l = curlwp, *l2; 1429 struct proc *p = l->l_proc; 1430 kauth_cred_t oc; 1431 1432 KASSERT(mutex_owned(p->p_lock)); 1433 1434 /* Is there a new credential to set in? */ 1435 if (scred != NULL) { 1436 p->p_cred = scred; 1437 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 1438 if (l2 != l) 1439 l2->l_prflag |= LPR_CRMOD; 1440 } 1441 1442 /* Ensure the LWP cached credentials are up to date. */ 1443 if ((oc = l->l_cred) != scred) { 1444 kauth_cred_hold(scred); 1445 l->l_cred = scred; 1446 } 1447 } else 1448 oc = NULL; /* XXXgcc */ 1449 1450 if (sugid) { 1451 /* 1452 * Mark process as having changed credentials, stops 1453 * tracing etc. 1454 */ 1455 p->p_flag |= PK_SUGID; 1456 } 1457 1458 mutex_exit(p->p_lock); 1459 1460 /* If there is a credential to be released, free it now. */ 1461 if (fcred != NULL) { 1462 KASSERT(scred != NULL); 1463 kauth_cred_free(fcred); 1464 if (oc != scred) 1465 kauth_cred_free(oc); 1466 } 1467 } 1468 1469 /* 1470 * proc_specific_key_create -- 1471 * Create a key for subsystem proc-specific data. 1472 */ 1473 int 1474 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 1475 { 1476 1477 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor)); 1478 } 1479 1480 /* 1481 * proc_specific_key_delete -- 1482 * Delete a key for subsystem proc-specific data. 1483 */ 1484 void 1485 proc_specific_key_delete(specificdata_key_t key) 1486 { 1487 1488 specificdata_key_delete(proc_specificdata_domain, key); 1489 } 1490 1491 /* 1492 * proc_initspecific -- 1493 * Initialize a proc's specificdata container. 1494 */ 1495 void 1496 proc_initspecific(struct proc *p) 1497 { 1498 int error __diagused; 1499 1500 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref); 1501 KASSERT(error == 0); 1502 } 1503 1504 /* 1505 * proc_finispecific -- 1506 * Finalize a proc's specificdata container. 1507 */ 1508 void 1509 proc_finispecific(struct proc *p) 1510 { 1511 1512 specificdata_fini(proc_specificdata_domain, &p->p_specdataref); 1513 } 1514 1515 /* 1516 * proc_getspecific -- 1517 * Return proc-specific data corresponding to the specified key. 1518 */ 1519 void * 1520 proc_getspecific(struct proc *p, specificdata_key_t key) 1521 { 1522 1523 return (specificdata_getspecific(proc_specificdata_domain, 1524 &p->p_specdataref, key)); 1525 } 1526 1527 /* 1528 * proc_setspecific -- 1529 * Set proc-specific data corresponding to the specified key. 1530 */ 1531 void 1532 proc_setspecific(struct proc *p, specificdata_key_t key, void *data) 1533 { 1534 1535 specificdata_setspecific(proc_specificdata_domain, 1536 &p->p_specdataref, key, data); 1537 } 1538 1539 int 1540 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target) 1541 { 1542 int r = 0; 1543 1544 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) || 1545 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) { 1546 /* 1547 * suid proc of ours or proc not ours 1548 */ 1549 r = EPERM; 1550 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) { 1551 /* 1552 * sgid proc has sgid back to us temporarily 1553 */ 1554 r = EPERM; 1555 } else { 1556 /* 1557 * our rgid must be in target's group list (ie, 1558 * sub-processes started by a sgid process) 1559 */ 1560 int ismember = 0; 1561 1562 if (kauth_cred_ismember_gid(cred, 1563 kauth_cred_getgid(target), &ismember) != 0 || 1564 !ismember) 1565 r = EPERM; 1566 } 1567 1568 return (r); 1569 } 1570 1571 /* 1572 * sysctl stuff 1573 */ 1574 1575 #define KERN_PROCSLOP (5 * sizeof(struct kinfo_proc)) 1576 1577 static const u_int sysctl_flagmap[] = { 1578 PK_ADVLOCK, P_ADVLOCK, 1579 PK_EXEC, P_EXEC, 1580 PK_NOCLDWAIT, P_NOCLDWAIT, 1581 PK_32, P_32, 1582 PK_CLDSIGIGN, P_CLDSIGIGN, 1583 PK_SUGID, P_SUGID, 1584 0 1585 }; 1586 1587 static const u_int sysctl_sflagmap[] = { 1588 PS_NOCLDSTOP, P_NOCLDSTOP, 1589 PS_WEXIT, P_WEXIT, 1590 PS_STOPFORK, P_STOPFORK, 1591 PS_STOPEXEC, P_STOPEXEC, 1592 PS_STOPEXIT, P_STOPEXIT, 1593 0 1594 }; 1595 1596 static const u_int sysctl_slflagmap[] = { 1597 PSL_TRACED, P_TRACED, 1598 PSL_CHTRACED, P_CHTRACED, 1599 PSL_SYSCALL, P_SYSCALL, 1600 0 1601 }; 1602 1603 static const u_int sysctl_lflagmap[] = { 1604 PL_CONTROLT, P_CONTROLT, 1605 PL_PPWAIT, P_PPWAIT, 1606 0 1607 }; 1608 1609 static const u_int sysctl_stflagmap[] = { 1610 PST_PROFIL, P_PROFIL, 1611 0 1612 1613 }; 1614 1615 /* used by kern_lwp also */ 1616 const u_int sysctl_lwpflagmap[] = { 1617 LW_SINTR, L_SINTR, 1618 LW_SYSTEM, L_SYSTEM, 1619 0 1620 }; 1621 1622 /* 1623 * Find the most ``active'' lwp of a process and return it for ps display 1624 * purposes 1625 */ 1626 static struct lwp * 1627 proc_active_lwp(struct proc *p) 1628 { 1629 static const int ostat[] = { 1630 0, 1631 2, /* LSIDL */ 1632 6, /* LSRUN */ 1633 5, /* LSSLEEP */ 1634 4, /* LSSTOP */ 1635 0, /* LSZOMB */ 1636 1, /* LSDEAD */ 1637 7, /* LSONPROC */ 1638 3 /* LSSUSPENDED */ 1639 }; 1640 1641 struct lwp *l, *lp = NULL; 1642 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 1643 KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat)); 1644 if (lp == NULL || 1645 ostat[l->l_stat] > ostat[lp->l_stat] || 1646 (ostat[l->l_stat] == ostat[lp->l_stat] && 1647 l->l_cpticks > lp->l_cpticks)) { 1648 lp = l; 1649 continue; 1650 } 1651 } 1652 return lp; 1653 } 1654 1655 static int 1656 sysctl_doeproc(SYSCTLFN_ARGS) 1657 { 1658 union { 1659 struct kinfo_proc kproc; 1660 struct kinfo_proc2 kproc2; 1661 } *kbuf; 1662 struct proc *p, *next, *marker; 1663 char *where, *dp; 1664 int type, op, arg, error; 1665 u_int elem_size, kelem_size, elem_count; 1666 size_t buflen, needed; 1667 bool match, zombie, mmmbrains; 1668 const bool allowaddr = get_expose_address(curproc); 1669 1670 if (namelen == 1 && name[0] == CTL_QUERY) 1671 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1672 1673 dp = where = oldp; 1674 buflen = where != NULL ? *oldlenp : 0; 1675 error = 0; 1676 needed = 0; 1677 type = rnode->sysctl_num; 1678 1679 if (type == KERN_PROC) { 1680 if (namelen == 0) 1681 return EINVAL; 1682 switch (op = name[0]) { 1683 case KERN_PROC_ALL: 1684 if (namelen != 1) 1685 return EINVAL; 1686 arg = 0; 1687 break; 1688 default: 1689 if (namelen != 2) 1690 return EINVAL; 1691 arg = name[1]; 1692 break; 1693 } 1694 elem_count = 0; /* Hush little compiler, don't you cry */ 1695 kelem_size = elem_size = sizeof(kbuf->kproc); 1696 } else { 1697 if (namelen != 4) 1698 return EINVAL; 1699 op = name[0]; 1700 arg = name[1]; 1701 elem_size = name[2]; 1702 elem_count = name[3]; 1703 kelem_size = sizeof(kbuf->kproc2); 1704 } 1705 1706 sysctl_unlock(); 1707 1708 kbuf = kmem_zalloc(sizeof(*kbuf), KM_SLEEP); 1709 marker = kmem_alloc(sizeof(*marker), KM_SLEEP); 1710 marker->p_flag = PK_MARKER; 1711 1712 mutex_enter(proc_lock); 1713 /* 1714 * Start with zombies to prevent reporting processes twice, in case they 1715 * are dying and being moved from the list of alive processes to zombies. 1716 */ 1717 mmmbrains = true; 1718 for (p = LIST_FIRST(&zombproc);; p = next) { 1719 if (p == NULL) { 1720 if (mmmbrains) { 1721 p = LIST_FIRST(&allproc); 1722 mmmbrains = false; 1723 } 1724 if (p == NULL) 1725 break; 1726 } 1727 next = LIST_NEXT(p, p_list); 1728 if ((p->p_flag & PK_MARKER) != 0) 1729 continue; 1730 1731 /* 1732 * Skip embryonic processes. 1733 */ 1734 if (p->p_stat == SIDL) 1735 continue; 1736 1737 mutex_enter(p->p_lock); 1738 error = kauth_authorize_process(l->l_cred, 1739 KAUTH_PROCESS_CANSEE, p, 1740 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL); 1741 if (error != 0) { 1742 mutex_exit(p->p_lock); 1743 continue; 1744 } 1745 1746 /* 1747 * Hande all the operations in one switch on the cost of 1748 * algorithm complexity is on purpose. The win splitting this 1749 * function into several similar copies makes maintenance burden 1750 * burden, code grow and boost is neglible in practical systems. 1751 */ 1752 switch (op) { 1753 case KERN_PROC_PID: 1754 match = (p->p_pid == (pid_t)arg); 1755 break; 1756 1757 case KERN_PROC_PGRP: 1758 match = (p->p_pgrp->pg_id == (pid_t)arg); 1759 break; 1760 1761 case KERN_PROC_SESSION: 1762 match = (p->p_session->s_sid == (pid_t)arg); 1763 break; 1764 1765 case KERN_PROC_TTY: 1766 match = true; 1767 if (arg == (int) KERN_PROC_TTY_REVOKE) { 1768 if ((p->p_lflag & PL_CONTROLT) == 0 || 1769 p->p_session->s_ttyp == NULL || 1770 p->p_session->s_ttyvp != NULL) { 1771 match = false; 1772 } 1773 } else if ((p->p_lflag & PL_CONTROLT) == 0 || 1774 p->p_session->s_ttyp == NULL) { 1775 if ((dev_t)arg != KERN_PROC_TTY_NODEV) { 1776 match = false; 1777 } 1778 } else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) { 1779 match = false; 1780 } 1781 break; 1782 1783 case KERN_PROC_UID: 1784 match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg); 1785 break; 1786 1787 case KERN_PROC_RUID: 1788 match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg); 1789 break; 1790 1791 case KERN_PROC_GID: 1792 match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg); 1793 break; 1794 1795 case KERN_PROC_RGID: 1796 match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg); 1797 break; 1798 1799 case KERN_PROC_ALL: 1800 match = true; 1801 /* allow everything */ 1802 break; 1803 1804 default: 1805 error = EINVAL; 1806 mutex_exit(p->p_lock); 1807 goto cleanup; 1808 } 1809 if (!match) { 1810 mutex_exit(p->p_lock); 1811 continue; 1812 } 1813 1814 /* 1815 * Grab a hold on the process. 1816 */ 1817 if (mmmbrains) { 1818 zombie = true; 1819 } else { 1820 zombie = !rw_tryenter(&p->p_reflock, RW_READER); 1821 } 1822 if (zombie) { 1823 LIST_INSERT_AFTER(p, marker, p_list); 1824 } 1825 1826 if (buflen >= elem_size && 1827 (type == KERN_PROC || elem_count > 0)) { 1828 ruspace(p); /* Update process vm resource use */ 1829 1830 if (type == KERN_PROC) { 1831 fill_proc(p, &kbuf->kproc.kp_proc, allowaddr); 1832 fill_eproc(p, &kbuf->kproc.kp_eproc, zombie, 1833 allowaddr); 1834 } else { 1835 fill_kproc2(p, &kbuf->kproc2, zombie, 1836 allowaddr); 1837 elem_count--; 1838 } 1839 mutex_exit(p->p_lock); 1840 mutex_exit(proc_lock); 1841 /* 1842 * Copy out elem_size, but not larger than kelem_size 1843 */ 1844 error = sysctl_copyout(l, kbuf, dp, 1845 uimin(kelem_size, elem_size)); 1846 mutex_enter(proc_lock); 1847 if (error) { 1848 goto bah; 1849 } 1850 dp += elem_size; 1851 buflen -= elem_size; 1852 } else { 1853 mutex_exit(p->p_lock); 1854 } 1855 needed += elem_size; 1856 1857 /* 1858 * Release reference to process. 1859 */ 1860 if (zombie) { 1861 next = LIST_NEXT(marker, p_list); 1862 LIST_REMOVE(marker, p_list); 1863 } else { 1864 rw_exit(&p->p_reflock); 1865 next = LIST_NEXT(p, p_list); 1866 } 1867 1868 /* 1869 * Short-circuit break quickly! 1870 */ 1871 if (op == KERN_PROC_PID) 1872 break; 1873 } 1874 mutex_exit(proc_lock); 1875 1876 if (where != NULL) { 1877 *oldlenp = dp - where; 1878 if (needed > *oldlenp) { 1879 error = ENOMEM; 1880 goto out; 1881 } 1882 } else { 1883 needed += KERN_PROCSLOP; 1884 *oldlenp = needed; 1885 } 1886 kmem_free(kbuf, sizeof(*kbuf)); 1887 kmem_free(marker, sizeof(*marker)); 1888 sysctl_relock(); 1889 return 0; 1890 bah: 1891 if (zombie) 1892 LIST_REMOVE(marker, p_list); 1893 else 1894 rw_exit(&p->p_reflock); 1895 cleanup: 1896 mutex_exit(proc_lock); 1897 out: 1898 kmem_free(kbuf, sizeof(*kbuf)); 1899 kmem_free(marker, sizeof(*marker)); 1900 sysctl_relock(); 1901 return error; 1902 } 1903 1904 int 1905 copyin_psstrings(struct proc *p, struct ps_strings *arginfo) 1906 { 1907 #if !defined(_RUMPKERNEL) 1908 int retval; 1909 1910 if (p->p_flag & PK_32) { 1911 MODULE_HOOK_CALL(kern_proc32_copyin_hook, (p, arginfo), 1912 enosys(), retval); 1913 return retval; 1914 } 1915 #endif /* !defined(_RUMPKERNEL) */ 1916 1917 return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo)); 1918 } 1919 1920 static int 1921 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len) 1922 { 1923 void **cookie = cookie_; 1924 struct lwp *l = cookie[0]; 1925 char *dst = cookie[1]; 1926 1927 return sysctl_copyout(l, src, dst + off, len); 1928 } 1929 1930 /* 1931 * sysctl helper routine for kern.proc_args pseudo-subtree. 1932 */ 1933 static int 1934 sysctl_kern_proc_args(SYSCTLFN_ARGS) 1935 { 1936 struct ps_strings pss; 1937 struct proc *p; 1938 pid_t pid; 1939 int type, error; 1940 void *cookie[2]; 1941 1942 if (namelen == 1 && name[0] == CTL_QUERY) 1943 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1944 1945 if (newp != NULL || namelen != 2) 1946 return (EINVAL); 1947 pid = name[0]; 1948 type = name[1]; 1949 1950 switch (type) { 1951 case KERN_PROC_PATHNAME: 1952 sysctl_unlock(); 1953 error = fill_pathname(l, pid, oldp, oldlenp); 1954 sysctl_relock(); 1955 return error; 1956 1957 case KERN_PROC_CWD: 1958 sysctl_unlock(); 1959 error = fill_cwd(l, pid, oldp, oldlenp); 1960 sysctl_relock(); 1961 return error; 1962 1963 case KERN_PROC_ARGV: 1964 case KERN_PROC_NARGV: 1965 case KERN_PROC_ENV: 1966 case KERN_PROC_NENV: 1967 /* ok */ 1968 break; 1969 default: 1970 return (EINVAL); 1971 } 1972 1973 sysctl_unlock(); 1974 1975 /* check pid */ 1976 mutex_enter(proc_lock); 1977 if ((p = proc_find(pid)) == NULL) { 1978 error = EINVAL; 1979 goto out_locked; 1980 } 1981 mutex_enter(p->p_lock); 1982 1983 /* Check permission. */ 1984 if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV) 1985 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 1986 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL); 1987 else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV) 1988 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 1989 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL); 1990 else 1991 error = EINVAL; /* XXXGCC */ 1992 if (error) { 1993 mutex_exit(p->p_lock); 1994 goto out_locked; 1995 } 1996 1997 if (oldp == NULL) { 1998 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) 1999 *oldlenp = sizeof (int); 2000 else 2001 *oldlenp = ARG_MAX; /* XXX XXX XXX */ 2002 error = 0; 2003 mutex_exit(p->p_lock); 2004 goto out_locked; 2005 } 2006 2007 /* 2008 * Zombies don't have a stack, so we can't read their psstrings. 2009 * System processes also don't have a user stack. 2010 */ 2011 if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) { 2012 error = EINVAL; 2013 mutex_exit(p->p_lock); 2014 goto out_locked; 2015 } 2016 2017 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 2018 mutex_exit(p->p_lock); 2019 if (error) { 2020 goto out_locked; 2021 } 2022 mutex_exit(proc_lock); 2023 2024 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) { 2025 int value; 2026 if ((error = copyin_psstrings(p, &pss)) == 0) { 2027 if (type == KERN_PROC_NARGV) 2028 value = pss.ps_nargvstr; 2029 else 2030 value = pss.ps_nenvstr; 2031 error = sysctl_copyout(l, &value, oldp, sizeof(value)); 2032 *oldlenp = sizeof(value); 2033 } 2034 } else { 2035 cookie[0] = l; 2036 cookie[1] = oldp; 2037 error = copy_procargs(p, type, oldlenp, 2038 copy_procargs_sysctl_cb, cookie); 2039 } 2040 rw_exit(&p->p_reflock); 2041 sysctl_relock(); 2042 return error; 2043 2044 out_locked: 2045 mutex_exit(proc_lock); 2046 sysctl_relock(); 2047 return error; 2048 } 2049 2050 int 2051 copy_procargs(struct proc *p, int oid, size_t *limit, 2052 int (*cb)(void *, const void *, size_t, size_t), void *cookie) 2053 { 2054 struct ps_strings pss; 2055 size_t len, i, loaded, entry_len; 2056 struct uio auio; 2057 struct iovec aiov; 2058 int error, argvlen; 2059 char *arg; 2060 char **argv; 2061 vaddr_t user_argv; 2062 struct vmspace *vmspace; 2063 2064 /* 2065 * Allocate a temporary buffer to hold the argument vector and 2066 * the arguments themselve. 2067 */ 2068 arg = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2069 argv = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2070 2071 /* 2072 * Lock the process down in memory. 2073 */ 2074 vmspace = p->p_vmspace; 2075 uvmspace_addref(vmspace); 2076 2077 /* 2078 * Read in the ps_strings structure. 2079 */ 2080 if ((error = copyin_psstrings(p, &pss)) != 0) 2081 goto done; 2082 2083 /* 2084 * Now read the address of the argument vector. 2085 */ 2086 switch (oid) { 2087 case KERN_PROC_ARGV: 2088 user_argv = (uintptr_t)pss.ps_argvstr; 2089 argvlen = pss.ps_nargvstr; 2090 break; 2091 case KERN_PROC_ENV: 2092 user_argv = (uintptr_t)pss.ps_envstr; 2093 argvlen = pss.ps_nenvstr; 2094 break; 2095 default: 2096 error = EINVAL; 2097 goto done; 2098 } 2099 2100 if (argvlen < 0) { 2101 error = EIO; 2102 goto done; 2103 } 2104 2105 2106 /* 2107 * Now copy each string. 2108 */ 2109 len = 0; /* bytes written to user buffer */ 2110 loaded = 0; /* bytes from argv already processed */ 2111 i = 0; /* To make compiler happy */ 2112 entry_len = PROC_PTRSZ(p); 2113 2114 for (; argvlen; --argvlen) { 2115 int finished = 0; 2116 vaddr_t base; 2117 size_t xlen; 2118 int j; 2119 2120 if (loaded == 0) { 2121 size_t rem = entry_len * argvlen; 2122 loaded = MIN(rem, PAGE_SIZE); 2123 error = copyin_vmspace(vmspace, 2124 (const void *)user_argv, argv, loaded); 2125 if (error) 2126 break; 2127 user_argv += loaded; 2128 i = 0; 2129 } 2130 2131 #if !defined(_RUMPKERNEL) 2132 if (p->p_flag & PK_32) 2133 MODULE_HOOK_CALL(kern_proc32_base_hook, 2134 (argv, i++), 0, base); 2135 else 2136 #endif /* !defined(_RUMPKERNEL) */ 2137 base = (vaddr_t)argv[i++]; 2138 loaded -= entry_len; 2139 2140 /* 2141 * The program has messed around with its arguments, 2142 * possibly deleting some, and replacing them with 2143 * NULL's. Treat this as the last argument and not 2144 * a failure. 2145 */ 2146 if (base == 0) 2147 break; 2148 2149 while (!finished) { 2150 xlen = PAGE_SIZE - (base & PAGE_MASK); 2151 2152 aiov.iov_base = arg; 2153 aiov.iov_len = PAGE_SIZE; 2154 auio.uio_iov = &aiov; 2155 auio.uio_iovcnt = 1; 2156 auio.uio_offset = base; 2157 auio.uio_resid = xlen; 2158 auio.uio_rw = UIO_READ; 2159 UIO_SETUP_SYSSPACE(&auio); 2160 error = uvm_io(&vmspace->vm_map, &auio, 0); 2161 if (error) 2162 goto done; 2163 2164 /* Look for the end of the string */ 2165 for (j = 0; j < xlen; j++) { 2166 if (arg[j] == '\0') { 2167 xlen = j + 1; 2168 finished = 1; 2169 break; 2170 } 2171 } 2172 2173 /* Check for user buffer overflow */ 2174 if (len + xlen > *limit) { 2175 finished = 1; 2176 if (len > *limit) 2177 xlen = 0; 2178 else 2179 xlen = *limit - len; 2180 } 2181 2182 /* Copyout the page */ 2183 error = (*cb)(cookie, arg, len, xlen); 2184 if (error) 2185 goto done; 2186 2187 len += xlen; 2188 base += xlen; 2189 } 2190 } 2191 *limit = len; 2192 2193 done: 2194 kmem_free(argv, PAGE_SIZE); 2195 kmem_free(arg, PAGE_SIZE); 2196 uvmspace_free(vmspace); 2197 return error; 2198 } 2199 2200 /* 2201 * Fill in a proc structure for the specified process. 2202 */ 2203 static void 2204 fill_proc(const struct proc *psrc, struct proc *p, bool allowaddr) 2205 { 2206 COND_SET_VALUE(p->p_list, psrc->p_list, allowaddr); 2207 COND_SET_VALUE(p->p_auxlock, psrc->p_auxlock, allowaddr); 2208 COND_SET_VALUE(p->p_lock, psrc->p_lock, allowaddr); 2209 COND_SET_VALUE(p->p_stmutex, psrc->p_stmutex, allowaddr); 2210 COND_SET_VALUE(p->p_reflock, psrc->p_reflock, allowaddr); 2211 COND_SET_VALUE(p->p_waitcv, psrc->p_waitcv, allowaddr); 2212 COND_SET_VALUE(p->p_lwpcv, psrc->p_lwpcv, allowaddr); 2213 COND_SET_VALUE(p->p_cred, psrc->p_cred, allowaddr); 2214 COND_SET_VALUE(p->p_fd, psrc->p_fd, allowaddr); 2215 COND_SET_VALUE(p->p_cwdi, psrc->p_cwdi, allowaddr); 2216 COND_SET_VALUE(p->p_stats, psrc->p_stats, allowaddr); 2217 COND_SET_VALUE(p->p_limit, psrc->p_limit, allowaddr); 2218 COND_SET_VALUE(p->p_vmspace, psrc->p_vmspace, allowaddr); 2219 COND_SET_VALUE(p->p_sigacts, psrc->p_sigacts, allowaddr); 2220 COND_SET_VALUE(p->p_aio, psrc->p_aio, allowaddr); 2221 p->p_mqueue_cnt = psrc->p_mqueue_cnt; 2222 COND_SET_VALUE(p->p_specdataref, psrc->p_specdataref, allowaddr); 2223 p->p_exitsig = psrc->p_exitsig; 2224 p->p_flag = psrc->p_flag; 2225 p->p_sflag = psrc->p_sflag; 2226 p->p_slflag = psrc->p_slflag; 2227 p->p_lflag = psrc->p_lflag; 2228 p->p_stflag = psrc->p_stflag; 2229 p->p_stat = psrc->p_stat; 2230 p->p_trace_enabled = psrc->p_trace_enabled; 2231 p->p_pid = psrc->p_pid; 2232 COND_SET_VALUE(p->p_pglist, psrc->p_pglist, allowaddr); 2233 COND_SET_VALUE(p->p_pptr, psrc->p_pptr, allowaddr); 2234 COND_SET_VALUE(p->p_sibling, psrc->p_sibling, allowaddr); 2235 COND_SET_VALUE(p->p_children, psrc->p_children, allowaddr); 2236 COND_SET_VALUE(p->p_lwps, psrc->p_lwps, allowaddr); 2237 COND_SET_VALUE(p->p_raslist, psrc->p_raslist, allowaddr); 2238 p->p_nlwps = psrc->p_nlwps; 2239 p->p_nzlwps = psrc->p_nzlwps; 2240 p->p_nrlwps = psrc->p_nrlwps; 2241 p->p_nlwpwait = psrc->p_nlwpwait; 2242 p->p_ndlwps = psrc->p_ndlwps; 2243 p->p_nlwpid = psrc->p_nlwpid; 2244 p->p_nstopchild = psrc->p_nstopchild; 2245 p->p_waited = psrc->p_waited; 2246 COND_SET_VALUE(p->p_zomblwp, psrc->p_zomblwp, allowaddr); 2247 COND_SET_VALUE(p->p_vforklwp, psrc->p_vforklwp, allowaddr); 2248 COND_SET_VALUE(p->p_sched_info, psrc->p_sched_info, allowaddr); 2249 p->p_estcpu = psrc->p_estcpu; 2250 p->p_estcpu_inherited = psrc->p_estcpu_inherited; 2251 p->p_forktime = psrc->p_forktime; 2252 p->p_pctcpu = psrc->p_pctcpu; 2253 COND_SET_VALUE(p->p_opptr, psrc->p_opptr, allowaddr); 2254 COND_SET_VALUE(p->p_timers, psrc->p_timers, allowaddr); 2255 p->p_rtime = psrc->p_rtime; 2256 p->p_uticks = psrc->p_uticks; 2257 p->p_sticks = psrc->p_sticks; 2258 p->p_iticks = psrc->p_iticks; 2259 p->p_xutime = psrc->p_xutime; 2260 p->p_xstime = psrc->p_xstime; 2261 p->p_traceflag = psrc->p_traceflag; 2262 COND_SET_VALUE(p->p_tracep, psrc->p_tracep, allowaddr); 2263 COND_SET_VALUE(p->p_textvp, psrc->p_textvp, allowaddr); 2264 COND_SET_VALUE(p->p_emul, psrc->p_emul, allowaddr); 2265 COND_SET_VALUE(p->p_emuldata, psrc->p_emuldata, allowaddr); 2266 COND_SET_VALUE(p->p_execsw, psrc->p_execsw, allowaddr); 2267 COND_SET_VALUE(p->p_klist, psrc->p_klist, allowaddr); 2268 COND_SET_VALUE(p->p_sigwaiters, psrc->p_sigwaiters, allowaddr); 2269 COND_SET_VALUE(p->p_sigpend, psrc->p_sigpend, allowaddr); 2270 COND_SET_VALUE(p->p_lwpctl, psrc->p_lwpctl, allowaddr); 2271 p->p_ppid = psrc->p_ppid; 2272 COND_SET_VALUE(p->p_path, psrc->p_path, allowaddr); 2273 COND_SET_VALUE(p->p_sigctx, psrc->p_sigctx, allowaddr); 2274 p->p_nice = psrc->p_nice; 2275 memcpy(p->p_comm, psrc->p_comm, sizeof(p->p_comm)); 2276 COND_SET_VALUE(p->p_pgrp, psrc->p_pgrp, allowaddr); 2277 COND_SET_VALUE(p->p_psstrp, psrc->p_psstrp, allowaddr); 2278 p->p_pax = psrc->p_pax; 2279 p->p_xexit = psrc->p_xexit; 2280 p->p_xsig = psrc->p_xsig; 2281 p->p_acflag = psrc->p_acflag; 2282 COND_SET_VALUE(p->p_md, psrc->p_md, allowaddr); 2283 p->p_stackbase = psrc->p_stackbase; 2284 COND_SET_VALUE(p->p_dtrace, psrc->p_dtrace, allowaddr); 2285 } 2286 2287 /* 2288 * Fill in an eproc structure for the specified process. 2289 */ 2290 void 2291 fill_eproc(struct proc *p, struct eproc *ep, bool zombie, bool allowaddr) 2292 { 2293 struct tty *tp; 2294 struct lwp *l; 2295 2296 KASSERT(mutex_owned(proc_lock)); 2297 KASSERT(mutex_owned(p->p_lock)); 2298 2299 COND_SET_VALUE(ep->e_paddr, p, allowaddr); 2300 COND_SET_VALUE(ep->e_sess, p->p_session, allowaddr); 2301 if (p->p_cred) { 2302 kauth_cred_topcred(p->p_cred, &ep->e_pcred); 2303 kauth_cred_toucred(p->p_cred, &ep->e_ucred); 2304 } 2305 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2306 struct vmspace *vm = p->p_vmspace; 2307 2308 ep->e_vm.vm_rssize = vm_resident_count(vm); 2309 ep->e_vm.vm_tsize = vm->vm_tsize; 2310 ep->e_vm.vm_dsize = vm->vm_dsize; 2311 ep->e_vm.vm_ssize = vm->vm_ssize; 2312 ep->e_vm.vm_map.size = vm->vm_map.size; 2313 2314 /* Pick the primary (first) LWP */ 2315 l = proc_active_lwp(p); 2316 KASSERT(l != NULL); 2317 lwp_lock(l); 2318 if (l->l_wchan) 2319 strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN); 2320 lwp_unlock(l); 2321 } 2322 ep->e_ppid = p->p_ppid; 2323 if (p->p_pgrp && p->p_session) { 2324 ep->e_pgid = p->p_pgrp->pg_id; 2325 ep->e_jobc = p->p_pgrp->pg_jobc; 2326 ep->e_sid = p->p_session->s_sid; 2327 if ((p->p_lflag & PL_CONTROLT) && 2328 (tp = p->p_session->s_ttyp)) { 2329 ep->e_tdev = tp->t_dev; 2330 ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2331 COND_SET_VALUE(ep->e_tsess, tp->t_session, allowaddr); 2332 } else 2333 ep->e_tdev = (uint32_t)NODEV; 2334 ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0; 2335 if (SESS_LEADER(p)) 2336 ep->e_flag |= EPROC_SLEADER; 2337 strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME); 2338 } 2339 ep->e_xsize = ep->e_xrssize = 0; 2340 ep->e_xccount = ep->e_xswrss = 0; 2341 } 2342 2343 /* 2344 * Fill in a kinfo_proc2 structure for the specified process. 2345 */ 2346 void 2347 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie, bool allowaddr) 2348 { 2349 struct tty *tp; 2350 struct lwp *l, *l2; 2351 struct timeval ut, st, rt; 2352 sigset_t ss1, ss2; 2353 struct rusage ru; 2354 struct vmspace *vm; 2355 2356 KASSERT(mutex_owned(proc_lock)); 2357 KASSERT(mutex_owned(p->p_lock)); 2358 2359 sigemptyset(&ss1); 2360 sigemptyset(&ss2); 2361 2362 COND_SET_VALUE(ki->p_paddr, PTRTOUINT64(p), allowaddr); 2363 COND_SET_VALUE(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr); 2364 COND_SET_VALUE(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr); 2365 COND_SET_VALUE(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr); 2366 COND_SET_VALUE(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr); 2367 COND_SET_VALUE(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr); 2368 COND_SET_VALUE(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr); 2369 COND_SET_VALUE(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr); 2370 ki->p_tsess = 0; /* may be changed if controlling tty below */ 2371 COND_SET_VALUE(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr); 2372 ki->p_eflag = 0; 2373 ki->p_exitsig = p->p_exitsig; 2374 ki->p_flag = L_INMEM; /* Process never swapped out */ 2375 ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag); 2376 ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag); 2377 ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag); 2378 ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag); 2379 ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag); 2380 ki->p_pid = p->p_pid; 2381 ki->p_ppid = p->p_ppid; 2382 ki->p_uid = kauth_cred_geteuid(p->p_cred); 2383 ki->p_ruid = kauth_cred_getuid(p->p_cred); 2384 ki->p_gid = kauth_cred_getegid(p->p_cred); 2385 ki->p_rgid = kauth_cred_getgid(p->p_cred); 2386 ki->p_svuid = kauth_cred_getsvuid(p->p_cred); 2387 ki->p_svgid = kauth_cred_getsvgid(p->p_cred); 2388 ki->p_ngroups = kauth_cred_ngroups(p->p_cred); 2389 kauth_cred_getgroups(p->p_cred, ki->p_groups, 2390 uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])), 2391 UIO_SYSSPACE); 2392 2393 ki->p_uticks = p->p_uticks; 2394 ki->p_sticks = p->p_sticks; 2395 ki->p_iticks = p->p_iticks; 2396 ki->p_tpgid = NO_PGID; /* may be changed if controlling tty below */ 2397 COND_SET_VALUE(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr); 2398 ki->p_traceflag = p->p_traceflag; 2399 2400 memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t)); 2401 memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t)); 2402 2403 ki->p_cpticks = 0; 2404 ki->p_pctcpu = p->p_pctcpu; 2405 ki->p_estcpu = 0; 2406 ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */ 2407 ki->p_realstat = p->p_stat; 2408 ki->p_nice = p->p_nice; 2409 ki->p_xstat = P_WAITSTATUS(p); 2410 ki->p_acflag = p->p_acflag; 2411 2412 strncpy(ki->p_comm, p->p_comm, 2413 uimin(sizeof(ki->p_comm), sizeof(p->p_comm))); 2414 strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename)); 2415 2416 ki->p_nlwps = p->p_nlwps; 2417 ki->p_realflag = ki->p_flag; 2418 2419 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2420 vm = p->p_vmspace; 2421 ki->p_vm_rssize = vm_resident_count(vm); 2422 ki->p_vm_tsize = vm->vm_tsize; 2423 ki->p_vm_dsize = vm->vm_dsize; 2424 ki->p_vm_ssize = vm->vm_ssize; 2425 ki->p_vm_vsize = atop(vm->vm_map.size); 2426 /* 2427 * Since the stack is initially mapped mostly with 2428 * PROT_NONE and grown as needed, adjust the "mapped size" 2429 * to skip the unused stack portion. 2430 */ 2431 ki->p_vm_msize = 2432 atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize; 2433 2434 /* Pick the primary (first) LWP */ 2435 l = proc_active_lwp(p); 2436 KASSERT(l != NULL); 2437 lwp_lock(l); 2438 ki->p_nrlwps = p->p_nrlwps; 2439 ki->p_forw = 0; 2440 ki->p_back = 0; 2441 COND_SET_VALUE(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr); 2442 ki->p_stat = l->l_stat; 2443 ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag); 2444 ki->p_swtime = l->l_swtime; 2445 ki->p_slptime = l->l_slptime; 2446 if (l->l_stat == LSONPROC) 2447 ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags; 2448 else 2449 ki->p_schedflags = 0; 2450 ki->p_priority = lwp_eprio(l); 2451 ki->p_usrpri = l->l_priority; 2452 if (l->l_wchan) 2453 strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg)); 2454 COND_SET_VALUE(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr); 2455 ki->p_cpuid = cpu_index(l->l_cpu); 2456 lwp_unlock(l); 2457 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 2458 /* This is hardly correct, but... */ 2459 sigplusset(&l->l_sigpend.sp_set, &ss1); 2460 sigplusset(&l->l_sigmask, &ss2); 2461 ki->p_cpticks += l->l_cpticks; 2462 ki->p_pctcpu += l->l_pctcpu; 2463 ki->p_estcpu += l->l_estcpu; 2464 } 2465 } 2466 sigplusset(&p->p_sigpend.sp_set, &ss1); 2467 memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t)); 2468 memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t)); 2469 2470 if (p->p_session != NULL) { 2471 ki->p_sid = p->p_session->s_sid; 2472 ki->p__pgid = p->p_pgrp->pg_id; 2473 if (p->p_session->s_ttyvp) 2474 ki->p_eflag |= EPROC_CTTY; 2475 if (SESS_LEADER(p)) 2476 ki->p_eflag |= EPROC_SLEADER; 2477 strncpy(ki->p_login, p->p_session->s_login, 2478 uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login)); 2479 ki->p_jobc = p->p_pgrp->pg_jobc; 2480 if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) { 2481 ki->p_tdev = tp->t_dev; 2482 ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2483 COND_SET_VALUE(ki->p_tsess, PTRTOUINT64(tp->t_session), 2484 allowaddr); 2485 } else { 2486 ki->p_tdev = (int32_t)NODEV; 2487 } 2488 } 2489 2490 if (!P_ZOMBIE(p) && !zombie) { 2491 ki->p_uvalid = 1; 2492 ki->p_ustart_sec = p->p_stats->p_start.tv_sec; 2493 ki->p_ustart_usec = p->p_stats->p_start.tv_usec; 2494 2495 calcru(p, &ut, &st, NULL, &rt); 2496 ki->p_rtime_sec = rt.tv_sec; 2497 ki->p_rtime_usec = rt.tv_usec; 2498 ki->p_uutime_sec = ut.tv_sec; 2499 ki->p_uutime_usec = ut.tv_usec; 2500 ki->p_ustime_sec = st.tv_sec; 2501 ki->p_ustime_usec = st.tv_usec; 2502 2503 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 2504 ki->p_uru_nvcsw = 0; 2505 ki->p_uru_nivcsw = 0; 2506 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 2507 ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw); 2508 ki->p_uru_nivcsw += l2->l_nivcsw; 2509 ruadd(&ru, &l2->l_ru); 2510 } 2511 ki->p_uru_maxrss = ru.ru_maxrss; 2512 ki->p_uru_ixrss = ru.ru_ixrss; 2513 ki->p_uru_idrss = ru.ru_idrss; 2514 ki->p_uru_isrss = ru.ru_isrss; 2515 ki->p_uru_minflt = ru.ru_minflt; 2516 ki->p_uru_majflt = ru.ru_majflt; 2517 ki->p_uru_nswap = ru.ru_nswap; 2518 ki->p_uru_inblock = ru.ru_inblock; 2519 ki->p_uru_oublock = ru.ru_oublock; 2520 ki->p_uru_msgsnd = ru.ru_msgsnd; 2521 ki->p_uru_msgrcv = ru.ru_msgrcv; 2522 ki->p_uru_nsignals = ru.ru_nsignals; 2523 2524 timeradd(&p->p_stats->p_cru.ru_utime, 2525 &p->p_stats->p_cru.ru_stime, &ut); 2526 ki->p_uctime_sec = ut.tv_sec; 2527 ki->p_uctime_usec = ut.tv_usec; 2528 } 2529 } 2530 2531 2532 int 2533 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid) 2534 { 2535 int error; 2536 2537 mutex_enter(proc_lock); 2538 if (pid == -1) 2539 *p = l->l_proc; 2540 else 2541 *p = proc_find(pid); 2542 2543 if (*p == NULL) { 2544 if (pid != -1) 2545 mutex_exit(proc_lock); 2546 return ESRCH; 2547 } 2548 if (pid != -1) 2549 mutex_enter((*p)->p_lock); 2550 mutex_exit(proc_lock); 2551 2552 error = kauth_authorize_process(l->l_cred, 2553 KAUTH_PROCESS_CANSEE, *p, 2554 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 2555 if (error) { 2556 if (pid != -1) 2557 mutex_exit((*p)->p_lock); 2558 } 2559 return error; 2560 } 2561 2562 static int 2563 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2564 { 2565 int error; 2566 struct proc *p; 2567 2568 if ((error = proc_find_locked(l, &p, pid)) != 0) 2569 return error; 2570 2571 if (p->p_path == NULL) { 2572 if (pid != -1) 2573 mutex_exit(p->p_lock); 2574 return ENOENT; 2575 } 2576 2577 size_t len = strlen(p->p_path) + 1; 2578 if (oldp != NULL) { 2579 size_t copylen = uimin(len, *oldlenp); 2580 error = sysctl_copyout(l, p->p_path, oldp, copylen); 2581 if (error == 0 && *oldlenp < len) 2582 error = ENOSPC; 2583 } 2584 *oldlenp = len; 2585 if (pid != -1) 2586 mutex_exit(p->p_lock); 2587 return error; 2588 } 2589 2590 static int 2591 fill_cwd(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2592 { 2593 int error; 2594 struct proc *p; 2595 char *path; 2596 char *bp, *bend; 2597 struct cwdinfo *cwdi; 2598 struct vnode *vp; 2599 size_t len, lenused; 2600 2601 if ((error = proc_find_locked(l, &p, pid)) != 0) 2602 return error; 2603 2604 len = MAXPATHLEN * 4; 2605 2606 path = kmem_alloc(len, KM_SLEEP); 2607 2608 bp = &path[len]; 2609 bend = bp; 2610 *(--bp) = '\0'; 2611 2612 cwdi = p->p_cwdi; 2613 rw_enter(&cwdi->cwdi_lock, RW_READER); 2614 vp = cwdi->cwdi_cdir; 2615 error = getcwd_common(vp, NULL, &bp, path, len/2, 0, l); 2616 rw_exit(&cwdi->cwdi_lock); 2617 2618 if (error) 2619 goto out; 2620 2621 lenused = bend - bp; 2622 2623 if (oldp != NULL) { 2624 size_t copylen = uimin(lenused, *oldlenp); 2625 error = sysctl_copyout(l, bp, oldp, copylen); 2626 if (error == 0 && *oldlenp < lenused) 2627 error = ENOSPC; 2628 } 2629 *oldlenp = lenused; 2630 out: 2631 if (pid != -1) 2632 mutex_exit(p->p_lock); 2633 kmem_free(path, len); 2634 return error; 2635 } 2636 2637 int 2638 proc_getauxv(struct proc *p, void **buf, size_t *len) 2639 { 2640 struct ps_strings pss; 2641 int error; 2642 void *uauxv, *kauxv; 2643 size_t size; 2644 2645 if ((error = copyin_psstrings(p, &pss)) != 0) 2646 return error; 2647 if (pss.ps_envstr == NULL) 2648 return EIO; 2649 2650 size = p->p_execsw->es_arglen; 2651 if (size == 0) 2652 return EIO; 2653 2654 size_t ptrsz = PROC_PTRSZ(p); 2655 uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz); 2656 2657 kauxv = kmem_alloc(size, KM_SLEEP); 2658 2659 error = copyin_proc(p, uauxv, kauxv, size); 2660 if (error) { 2661 kmem_free(kauxv, size); 2662 return error; 2663 } 2664 2665 *buf = kauxv; 2666 *len = size; 2667 2668 return 0; 2669 } 2670 2671 2672 static int 2673 sysctl_security_expose_address(SYSCTLFN_ARGS) 2674 { 2675 int expose_address, error; 2676 struct sysctlnode node; 2677 2678 node = *rnode; 2679 node.sysctl_data = &expose_address; 2680 expose_address = *(int *)rnode->sysctl_data; 2681 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2682 if (error || newp == NULL) 2683 return error; 2684 2685 if (kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_KERNADDR, 2686 0, NULL, NULL, NULL)) 2687 return EPERM; 2688 2689 switch (expose_address) { 2690 case 0: 2691 case 1: 2692 case 2: 2693 break; 2694 default: 2695 return EINVAL; 2696 } 2697 2698 *(int *)rnode->sysctl_data = expose_address; 2699 2700 return 0; 2701 } 2702 2703 bool 2704 get_expose_address(struct proc *p) 2705 { 2706 /* allow only if sysctl variable is set or privileged */ 2707 return kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE, 2708 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL) == 0; 2709 } 2710
Indexes created Sun Sep 21 18:09:52 GMT 2025