kern_proc.c revision 1.239.2.1
1 /* $NetBSD: kern_proc.c,v 1.239.2.1 2020/01/25 15:54:03 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2006, 2007, 2008, 2020 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 1.239.2.1 2020/01/25 15:54:03 ad 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 #include <sys/vnode.h> 110 111 #include <uvm/uvm_extern.h> 112 #include <uvm/uvm.h> 113 114 /* 115 * Process lists. 116 */ 117 118 struct proclist allproc __cacheline_aligned; 119 struct proclist zombproc __cacheline_aligned; 120 121 kmutex_t * proc_lock __cacheline_aligned; 122 123 /* 124 * pid to proc lookup is done by indexing the pid_table array. 125 * Since pid numbers are only allocated when an empty slot 126 * has been found, there is no need to search any lists ever. 127 * (an orphaned pgrp will lock the slot, a session will lock 128 * the pgrp with the same number.) 129 * If the table is too small it is reallocated with twice the 130 * previous size and the entries 'unzipped' into the two halves. 131 * A linked list of free entries is passed through the pt_proc 132 * field of 'free' items - set odd to be an invalid ptr. 133 */ 134 135 struct pid_table { 136 struct proc *pt_proc; 137 struct pgrp *pt_pgrp; 138 pid_t pt_pid; 139 }; 140 #if 1 /* strongly typed cast - should be a noop */ 141 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; } 142 #else 143 #define p2u(p) ((uint)p) 144 #endif 145 #define P_VALID(p) (!(p2u(p) & 1)) 146 #define P_NEXT(p) (p2u(p) >> 1) 147 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1)) 148 149 /* 150 * Table of process IDs (PIDs). 151 */ 152 static struct pid_table *pid_table __read_mostly; 153 154 #define INITIAL_PID_TABLE_SIZE (1 << 5) 155 156 /* Table mask, threshold for growing and number of allocated PIDs. */ 157 static u_int pid_tbl_mask __read_mostly; 158 static u_int pid_alloc_lim __read_mostly; 159 static u_int pid_alloc_cnt __cacheline_aligned; 160 161 /* Next free, last free and maximum PIDs. */ 162 static u_int next_free_pt __cacheline_aligned; 163 static u_int last_free_pt __cacheline_aligned; 164 static pid_t pid_max __read_mostly; 165 166 /* Components of the first process -- never freed. */ 167 168 extern struct emul emul_netbsd; /* defined in kern_exec.c */ 169 170 struct session session0 = { 171 .s_count = 1, 172 .s_sid = 0, 173 }; 174 struct pgrp pgrp0 = { 175 .pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members), 176 .pg_session = &session0, 177 }; 178 filedesc_t filedesc0; 179 struct cwdinfo cwdi0 = { 180 .cwdi_cmask = CMASK, 181 .cwdi_refcnt = 1, 182 }; 183 struct plimit limit0; 184 struct pstats pstat0; 185 struct vmspace vmspace0; 186 struct sigacts sigacts0; 187 struct proc proc0 = { 188 .p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps), 189 .p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters), 190 .p_nlwps = 1, 191 .p_nrlwps = 1, 192 .p_nlwpid = 1, /* must match lwp0.l_lid */ 193 .p_pgrp = &pgrp0, 194 .p_comm = "system", 195 /* 196 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8) 197 * when they exit. init(8) can easily wait them out for us. 198 */ 199 .p_flag = PK_SYSTEM | PK_NOCLDWAIT, 200 .p_stat = SACTIVE, 201 .p_nice = NZERO, 202 .p_emul = &emul_netbsd, 203 .p_cwdi = &cwdi0, 204 .p_limit = &limit0, 205 .p_fd = &filedesc0, 206 .p_vmspace = &vmspace0, 207 .p_stats = &pstat0, 208 .p_sigacts = &sigacts0, 209 #ifdef PROC0_MD_INITIALIZERS 210 PROC0_MD_INITIALIZERS 211 #endif 212 }; 213 kauth_cred_t cred0; 214 215 static const int nofile = NOFILE; 216 static const int maxuprc = MAXUPRC; 217 218 static int sysctl_doeproc(SYSCTLFN_PROTO); 219 static int sysctl_kern_proc_args(SYSCTLFN_PROTO); 220 static int sysctl_security_expose_address(SYSCTLFN_PROTO); 221 222 #ifdef KASLR 223 static int kern_expose_address = 0; 224 #else 225 static int kern_expose_address = 1; 226 #endif 227 /* 228 * The process list descriptors, used during pid allocation and 229 * by sysctl. No locking on this data structure is needed since 230 * it is completely static. 231 */ 232 const struct proclist_desc proclists[] = { 233 { &allproc }, 234 { &zombproc }, 235 { NULL }, 236 }; 237 238 static struct pgrp * pg_remove(pid_t); 239 static void pg_delete(pid_t); 240 static void orphanpg(struct pgrp *); 241 242 static specificdata_domain_t proc_specificdata_domain; 243 244 static pool_cache_t proc_cache; 245 246 static kauth_listener_t proc_listener; 247 248 static void fill_proc(const struct proc *, struct proc *, bool); 249 static int fill_pathname(struct lwp *, pid_t, void *, size_t *); 250 static int fill_cwd(struct lwp *, pid_t, void *, size_t *); 251 252 static int 253 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 254 void *arg0, void *arg1, void *arg2, void *arg3) 255 { 256 struct proc *p; 257 int result; 258 259 result = KAUTH_RESULT_DEFER; 260 p = arg0; 261 262 switch (action) { 263 case KAUTH_PROCESS_CANSEE: { 264 enum kauth_process_req req; 265 266 req = (enum kauth_process_req)arg1; 267 268 switch (req) { 269 case KAUTH_REQ_PROCESS_CANSEE_ARGS: 270 case KAUTH_REQ_PROCESS_CANSEE_ENTRY: 271 case KAUTH_REQ_PROCESS_CANSEE_OPENFILES: 272 case KAUTH_REQ_PROCESS_CANSEE_EPROC: 273 result = KAUTH_RESULT_ALLOW; 274 break; 275 276 case KAUTH_REQ_PROCESS_CANSEE_ENV: 277 if (kauth_cred_getuid(cred) != 278 kauth_cred_getuid(p->p_cred) || 279 kauth_cred_getuid(cred) != 280 kauth_cred_getsvuid(p->p_cred)) 281 break; 282 283 result = KAUTH_RESULT_ALLOW; 284 285 break; 286 287 case KAUTH_REQ_PROCESS_CANSEE_KPTR: 288 if (!kern_expose_address) 289 break; 290 291 if (kern_expose_address == 1 && !(p->p_flag & PK_KMEM)) 292 break; 293 294 result = KAUTH_RESULT_ALLOW; 295 296 break; 297 298 default: 299 break; 300 } 301 302 break; 303 } 304 305 case KAUTH_PROCESS_FORK: { 306 int lnprocs = (int)(unsigned long)arg2; 307 308 /* 309 * Don't allow a nonprivileged user to use the last few 310 * processes. The variable lnprocs is the current number of 311 * processes, maxproc is the limit. 312 */ 313 if (__predict_false((lnprocs >= maxproc - 5))) 314 break; 315 316 result = KAUTH_RESULT_ALLOW; 317 318 break; 319 } 320 321 case KAUTH_PROCESS_CORENAME: 322 case KAUTH_PROCESS_STOPFLAG: 323 if (proc_uidmatch(cred, p->p_cred) == 0) 324 result = KAUTH_RESULT_ALLOW; 325 326 break; 327 328 default: 329 break; 330 } 331 332 return result; 333 } 334 335 static int 336 proc_ctor(void *arg __unused, void *obj, int flags __unused) 337 { 338 memset(obj, 0, sizeof(struct proc)); 339 return 0; 340 } 341 342 /* 343 * Initialize global process hashing structures. 344 */ 345 void 346 procinit(void) 347 { 348 const struct proclist_desc *pd; 349 u_int i; 350 #define LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1)) 351 352 for (pd = proclists; pd->pd_list != NULL; pd++) 353 LIST_INIT(pd->pd_list); 354 355 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 356 pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE 357 * sizeof(struct pid_table), KM_SLEEP); 358 pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1; 359 pid_max = PID_MAX; 360 361 /* Set free list running through table... 362 Preset 'use count' above PID_MAX so we allocate pid 1 next. */ 363 for (i = 0; i <= pid_tbl_mask; i++) { 364 pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1); 365 pid_table[i].pt_pgrp = 0; 366 pid_table[i].pt_pid = 0; 367 } 368 /* slot 0 is just grabbed */ 369 next_free_pt = 1; 370 /* Need to fix last entry. */ 371 last_free_pt = pid_tbl_mask; 372 pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY); 373 /* point at which we grow table - to avoid reusing pids too often */ 374 pid_alloc_lim = pid_tbl_mask - 1; 375 #undef LINK_EMPTY 376 377 proc_specificdata_domain = specificdata_domain_create(); 378 KASSERT(proc_specificdata_domain != NULL); 379 380 proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0, 381 "procpl", NULL, IPL_NONE, proc_ctor, NULL, NULL); 382 383 proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 384 proc_listener_cb, NULL); 385 } 386 387 void 388 procinit_sysctl(void) 389 { 390 static struct sysctllog *clog; 391 392 sysctl_createv(&clog, 0, NULL, NULL, 393 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 394 CTLTYPE_INT, "expose_address", 395 SYSCTL_DESCR("Enable exposing kernel addresses"), 396 sysctl_security_expose_address, 0, 397 &kern_expose_address, 0, CTL_KERN, CTL_CREATE, CTL_EOL); 398 sysctl_createv(&clog, 0, NULL, NULL, 399 CTLFLAG_PERMANENT, 400 CTLTYPE_NODE, "proc", 401 SYSCTL_DESCR("System-wide process information"), 402 sysctl_doeproc, 0, NULL, 0, 403 CTL_KERN, KERN_PROC, CTL_EOL); 404 sysctl_createv(&clog, 0, NULL, NULL, 405 CTLFLAG_PERMANENT, 406 CTLTYPE_NODE, "proc2", 407 SYSCTL_DESCR("Machine-independent process information"), 408 sysctl_doeproc, 0, NULL, 0, 409 CTL_KERN, KERN_PROC2, CTL_EOL); 410 sysctl_createv(&clog, 0, NULL, NULL, 411 CTLFLAG_PERMANENT, 412 CTLTYPE_NODE, "proc_args", 413 SYSCTL_DESCR("Process argument information"), 414 sysctl_kern_proc_args, 0, NULL, 0, 415 CTL_KERN, KERN_PROC_ARGS, CTL_EOL); 416 417 /* 418 "nodes" under these: 419 420 KERN_PROC_ALL 421 KERN_PROC_PID pid 422 KERN_PROC_PGRP pgrp 423 KERN_PROC_SESSION sess 424 KERN_PROC_TTY tty 425 KERN_PROC_UID uid 426 KERN_PROC_RUID uid 427 KERN_PROC_GID gid 428 KERN_PROC_RGID gid 429 430 all in all, probably not worth the effort... 431 */ 432 } 433 434 /* 435 * Initialize process 0. 436 */ 437 void 438 proc0_init(void) 439 { 440 struct proc *p; 441 struct pgrp *pg; 442 struct rlimit *rlim; 443 rlim_t lim; 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 cv_init(&p->p_waitcv, "wait"); 455 cv_init(&p->p_lwpcv, "lwpwait"); 456 457 LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling); 458 459 pid_table[0].pt_proc = p; 460 LIST_INSERT_HEAD(&allproc, p, p_list); 461 462 pid_table[0].pt_pgrp = pg; 463 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist); 464 465 #ifdef __HAVE_SYSCALL_INTERN 466 (*p->p_emul->e_syscall_intern)(p); 467 #endif 468 469 /* Create credentials. */ 470 cred0 = kauth_cred_alloc(); 471 p->p_cred = cred0; 472 473 /* Create the CWD info. */ 474 mutex_init(&cwdi0.cwdi_lock, MUTEX_DEFAULT, IPL_NONE); 475 476 /* Create the limits structures. */ 477 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE); 478 479 rlim = limit0.pl_rlimit; 480 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) { 481 rlim[i].rlim_cur = RLIM_INFINITY; 482 rlim[i].rlim_max = RLIM_INFINITY; 483 } 484 485 rlim[RLIMIT_NOFILE].rlim_max = maxfiles; 486 rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile; 487 488 rlim[RLIMIT_NPROC].rlim_max = maxproc; 489 rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc; 490 491 lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvm_availmem())); 492 rlim[RLIMIT_RSS].rlim_max = lim; 493 rlim[RLIMIT_MEMLOCK].rlim_max = lim; 494 rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 495 496 rlim[RLIMIT_NTHR].rlim_max = maxlwp; 497 rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc; 498 499 /* Note that default core name has zero length. */ 500 limit0.pl_corename = defcorename; 501 limit0.pl_cnlen = 0; 502 limit0.pl_refcnt = 1; 503 limit0.pl_writeable = false; 504 limit0.pl_sv_limit = NULL; 505 506 /* Configure virtual memory system, set vm rlimits. */ 507 uvm_init_limits(p); 508 509 /* Initialize file descriptor table for proc0. */ 510 fd_init(&filedesc0); 511 512 /* 513 * Initialize proc0's vmspace, which uses the kernel pmap. 514 * All kernel processes (which never have user space mappings) 515 * share proc0's vmspace, and thus, the kernel pmap. 516 */ 517 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 518 trunc_page(VM_MAXUSER_ADDRESS), 519 #ifdef __USE_TOPDOWN_VM 520 true 521 #else 522 false 523 #endif 524 ); 525 526 /* Initialize signal state for proc0. XXX IPL_SCHED */ 527 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED); 528 siginit(p); 529 530 proc_initspecific(p); 531 kdtrace_proc_ctor(NULL, p); 532 } 533 534 /* 535 * Session reference counting. 536 */ 537 538 void 539 proc_sesshold(struct session *ss) 540 { 541 542 KASSERT(mutex_owned(proc_lock)); 543 ss->s_count++; 544 } 545 546 void 547 proc_sessrele(struct session *ss) 548 { 549 550 KASSERT(mutex_owned(proc_lock)); 551 /* 552 * We keep the pgrp with the same id as the session in order to 553 * stop a process being given the same pid. Since the pgrp holds 554 * a reference to the session, it must be a 'zombie' pgrp by now. 555 */ 556 if (--ss->s_count == 0) { 557 struct pgrp *pg; 558 559 pg = pg_remove(ss->s_sid); 560 mutex_exit(proc_lock); 561 562 kmem_free(pg, sizeof(struct pgrp)); 563 kmem_free(ss, sizeof(struct session)); 564 } else { 565 mutex_exit(proc_lock); 566 } 567 } 568 569 /* 570 * Check that the specified process group is in the session of the 571 * specified process. 572 * Treats -ve ids as process ids. 573 * Used to validate TIOCSPGRP requests. 574 */ 575 int 576 pgid_in_session(struct proc *p, pid_t pg_id) 577 { 578 struct pgrp *pgrp; 579 struct session *session; 580 int error; 581 582 mutex_enter(proc_lock); 583 if (pg_id < 0) { 584 struct proc *p1 = proc_find(-pg_id); 585 if (p1 == NULL) { 586 error = EINVAL; 587 goto fail; 588 } 589 pgrp = p1->p_pgrp; 590 } else { 591 pgrp = pgrp_find(pg_id); 592 if (pgrp == NULL) { 593 error = EINVAL; 594 goto fail; 595 } 596 } 597 session = pgrp->pg_session; 598 error = (session != p->p_pgrp->pg_session) ? EPERM : 0; 599 fail: 600 mutex_exit(proc_lock); 601 return error; 602 } 603 604 /* 605 * p_inferior: is p an inferior of q? 606 */ 607 static inline bool 608 p_inferior(struct proc *p, struct proc *q) 609 { 610 611 KASSERT(mutex_owned(proc_lock)); 612 613 for (; p != q; p = p->p_pptr) 614 if (p->p_pid == 0) 615 return false; 616 return true; 617 } 618 619 /* 620 * proc_find: locate a process by the ID. 621 * 622 * => Must be called with proc_lock held. 623 */ 624 proc_t * 625 proc_find_raw(pid_t pid) 626 { 627 struct pid_table *pt; 628 proc_t *p; 629 630 KASSERT(mutex_owned(proc_lock)); 631 pt = &pid_table[pid & pid_tbl_mask]; 632 p = pt->pt_proc; 633 if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) { 634 return NULL; 635 } 636 return p; 637 } 638 639 proc_t * 640 proc_find(pid_t pid) 641 { 642 proc_t *p; 643 644 p = proc_find_raw(pid); 645 if (__predict_false(p == NULL)) { 646 return NULL; 647 } 648 649 /* 650 * Only allow live processes to be found by PID. 651 * XXX: p_stat might change, since unlocked. 652 */ 653 if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) { 654 return p; 655 } 656 return NULL; 657 } 658 659 /* 660 * pgrp_find: locate a process group by the ID. 661 * 662 * => Must be called with proc_lock held. 663 */ 664 struct pgrp * 665 pgrp_find(pid_t pgid) 666 { 667 struct pgrp *pg; 668 669 KASSERT(mutex_owned(proc_lock)); 670 671 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp; 672 673 /* 674 * Cannot look up a process group that only exists because the 675 * session has not died yet (traditional). 676 */ 677 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) { 678 return NULL; 679 } 680 return pg; 681 } 682 683 static void 684 expand_pid_table(void) 685 { 686 size_t pt_size, tsz; 687 struct pid_table *n_pt, *new_pt; 688 struct proc *proc; 689 struct pgrp *pgrp; 690 pid_t pid, rpid; 691 u_int i; 692 uint new_pt_mask; 693 694 pt_size = pid_tbl_mask + 1; 695 tsz = pt_size * 2 * sizeof(struct pid_table); 696 new_pt = kmem_alloc(tsz, KM_SLEEP); 697 new_pt_mask = pt_size * 2 - 1; 698 699 mutex_enter(proc_lock); 700 if (pt_size != pid_tbl_mask + 1) { 701 /* Another process beat us to it... */ 702 mutex_exit(proc_lock); 703 kmem_free(new_pt, tsz); 704 return; 705 } 706 707 /* 708 * Copy entries from old table into new one. 709 * If 'pid' is 'odd' we need to place in the upper half, 710 * even pid's to the lower half. 711 * Free items stay in the low half so we don't have to 712 * fixup the reference to them. 713 * We stuff free items on the front of the freelist 714 * because we can't write to unmodified entries. 715 * Processing the table backwards maintains a semblance 716 * of issuing pid numbers that increase with time. 717 */ 718 i = pt_size - 1; 719 n_pt = new_pt + i; 720 for (; ; i--, n_pt--) { 721 proc = pid_table[i].pt_proc; 722 pgrp = pid_table[i].pt_pgrp; 723 if (!P_VALID(proc)) { 724 /* Up 'use count' so that link is valid */ 725 pid = (P_NEXT(proc) + pt_size) & ~pt_size; 726 rpid = 0; 727 proc = P_FREE(pid); 728 if (pgrp) 729 pid = pgrp->pg_id; 730 } else { 731 pid = pid_table[i].pt_pid; 732 rpid = pid; 733 } 734 735 /* Save entry in appropriate half of table */ 736 n_pt[pid & pt_size].pt_proc = proc; 737 n_pt[pid & pt_size].pt_pgrp = pgrp; 738 n_pt[pid & pt_size].pt_pid = rpid; 739 740 /* Put other piece on start of free list */ 741 pid = (pid ^ pt_size) & ~pid_tbl_mask; 742 n_pt[pid & pt_size].pt_proc = 743 P_FREE((pid & ~pt_size) | next_free_pt); 744 n_pt[pid & pt_size].pt_pgrp = 0; 745 n_pt[pid & pt_size].pt_pid = 0; 746 747 next_free_pt = i | (pid & pt_size); 748 if (i == 0) 749 break; 750 } 751 752 /* Save old table size and switch tables */ 753 tsz = pt_size * sizeof(struct pid_table); 754 n_pt = pid_table; 755 pid_table = new_pt; 756 pid_tbl_mask = new_pt_mask; 757 758 /* 759 * pid_max starts as PID_MAX (= 30000), once we have 16384 760 * allocated pids we need it to be larger! 761 */ 762 if (pid_tbl_mask > PID_MAX) { 763 pid_max = pid_tbl_mask * 2 + 1; 764 pid_alloc_lim |= pid_alloc_lim << 1; 765 } else 766 pid_alloc_lim <<= 1; /* doubles number of free slots... */ 767 768 mutex_exit(proc_lock); 769 kmem_free(n_pt, tsz); 770 } 771 772 struct proc * 773 proc_alloc(void) 774 { 775 struct proc *p; 776 777 p = pool_cache_get(proc_cache, PR_WAITOK); 778 p->p_stat = SIDL; /* protect against others */ 779 proc_initspecific(p); 780 kdtrace_proc_ctor(NULL, p); 781 p->p_pid = -1; 782 proc_alloc_pid(p); 783 return p; 784 } 785 786 /* 787 * proc_alloc_pid: allocate PID and record the given proc 'p' so that 788 * proc_find_raw() can find it by the PID. 789 */ 790 791 pid_t 792 proc_alloc_pid(struct proc *p) 793 { 794 struct pid_table *pt; 795 pid_t pid; 796 int nxt; 797 798 for (;;expand_pid_table()) { 799 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) 800 /* ensure pids cycle through 2000+ values */ 801 continue; 802 mutex_enter(proc_lock); 803 pt = &pid_table[next_free_pt]; 804 #ifdef DIAGNOSTIC 805 if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp)) 806 panic("proc_alloc: slot busy"); 807 #endif 808 nxt = P_NEXT(pt->pt_proc); 809 if (nxt & pid_tbl_mask) 810 break; 811 /* Table full - expand (NB last entry not used....) */ 812 mutex_exit(proc_lock); 813 } 814 815 /* pid is 'saved use count' + 'size' + entry */ 816 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt; 817 if ((uint)pid > (uint)pid_max) 818 pid &= pid_tbl_mask; 819 next_free_pt = nxt & pid_tbl_mask; 820 821 /* Grab table slot */ 822 pt->pt_proc = p; 823 824 KASSERT(pt->pt_pid == 0); 825 pt->pt_pid = pid; 826 if (p->p_pid == -1) { 827 p->p_pid = pid; 828 } 829 pid_alloc_cnt++; 830 mutex_exit(proc_lock); 831 832 return pid; 833 } 834 835 /* 836 * Free a process id - called from proc_free (in kern_exit.c) 837 * 838 * Called with the proc_lock held. 839 */ 840 void 841 proc_free_pid(pid_t pid) 842 { 843 struct pid_table *pt; 844 845 KASSERT(mutex_owned(proc_lock)); 846 847 pt = &pid_table[pid & pid_tbl_mask]; 848 849 /* save pid use count in slot */ 850 pt->pt_proc = P_FREE(pid & ~pid_tbl_mask); 851 KASSERT(pt->pt_pid == pid); 852 pt->pt_pid = 0; 853 854 if (pt->pt_pgrp == NULL) { 855 /* link last freed entry onto ours */ 856 pid &= pid_tbl_mask; 857 pt = &pid_table[last_free_pt]; 858 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid); 859 pt->pt_pid = 0; 860 last_free_pt = pid; 861 pid_alloc_cnt--; 862 } 863 864 atomic_dec_uint(&nprocs); 865 } 866 867 void 868 proc_free_mem(struct proc *p) 869 { 870 871 kdtrace_proc_dtor(NULL, p); 872 pool_cache_put(proc_cache, p); 873 } 874 875 /* 876 * proc_enterpgrp: move p to a new or existing process group (and session). 877 * 878 * If we are creating a new pgrp, the pgid should equal 879 * the calling process' pid. 880 * If is only valid to enter a process group that is in the session 881 * of the process. 882 * Also mksess should only be set if we are creating a process group 883 * 884 * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return. 885 */ 886 int 887 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess) 888 { 889 struct pgrp *new_pgrp, *pgrp; 890 struct session *sess; 891 struct proc *p; 892 int rval; 893 pid_t pg_id = NO_PGID; 894 895 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL; 896 897 /* Allocate data areas we might need before doing any validity checks */ 898 mutex_enter(proc_lock); /* Because pid_table might change */ 899 if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) { 900 mutex_exit(proc_lock); 901 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP); 902 mutex_enter(proc_lock); 903 } else 904 new_pgrp = NULL; 905 rval = EPERM; /* most common error (to save typing) */ 906 907 /* Check pgrp exists or can be created */ 908 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp; 909 if (pgrp != NULL && pgrp->pg_id != pgid) 910 goto done; 911 912 /* Can only set another process under restricted circumstances. */ 913 if (pid != curp->p_pid) { 914 /* Must exist and be one of our children... */ 915 p = proc_find(pid); 916 if (p == NULL || !p_inferior(p, curp)) { 917 rval = ESRCH; 918 goto done; 919 } 920 /* ... in the same session... */ 921 if (sess != NULL || p->p_session != curp->p_session) 922 goto done; 923 /* ... existing pgid must be in same session ... */ 924 if (pgrp != NULL && pgrp->pg_session != p->p_session) 925 goto done; 926 /* ... and not done an exec. */ 927 if (p->p_flag & PK_EXEC) { 928 rval = EACCES; 929 goto done; 930 } 931 } else { 932 /* ... setsid() cannot re-enter a pgrp */ 933 if (mksess && (curp->p_pgid == curp->p_pid || 934 pgrp_find(curp->p_pid))) 935 goto done; 936 p = curp; 937 } 938 939 /* Changing the process group/session of a session 940 leader is definitely off limits. */ 941 if (SESS_LEADER(p)) { 942 if (sess == NULL && p->p_pgrp == pgrp) 943 /* unless it's a definite noop */ 944 rval = 0; 945 goto done; 946 } 947 948 /* Can only create a process group with id of process */ 949 if (pgrp == NULL && pgid != pid) 950 goto done; 951 952 /* Can only create a session if creating pgrp */ 953 if (sess != NULL && pgrp != NULL) 954 goto done; 955 956 /* Check we allocated memory for a pgrp... */ 957 if (pgrp == NULL && new_pgrp == NULL) 958 goto done; 959 960 /* Don't attach to 'zombie' pgrp */ 961 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members)) 962 goto done; 963 964 /* Expect to succeed now */ 965 rval = 0; 966 967 if (pgrp == p->p_pgrp) 968 /* nothing to do */ 969 goto done; 970 971 /* Ok all setup, link up required structures */ 972 973 if (pgrp == NULL) { 974 pgrp = new_pgrp; 975 new_pgrp = NULL; 976 if (sess != NULL) { 977 sess->s_sid = p->p_pid; 978 sess->s_leader = p; 979 sess->s_count = 1; 980 sess->s_ttyvp = NULL; 981 sess->s_ttyp = NULL; 982 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET; 983 memcpy(sess->s_login, p->p_session->s_login, 984 sizeof(sess->s_login)); 985 p->p_lflag &= ~PL_CONTROLT; 986 } else { 987 sess = p->p_pgrp->pg_session; 988 proc_sesshold(sess); 989 } 990 pgrp->pg_session = sess; 991 sess = NULL; 992 993 pgrp->pg_id = pgid; 994 LIST_INIT(&pgrp->pg_members); 995 #ifdef DIAGNOSTIC 996 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp)) 997 panic("enterpgrp: pgrp table slot in use"); 998 if (__predict_false(mksess && p != curp)) 999 panic("enterpgrp: mksession and p != curproc"); 1000 #endif 1001 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp; 1002 pgrp->pg_jobc = 0; 1003 } 1004 1005 /* 1006 * Adjust eligibility of affected pgrps to participate in job control. 1007 * Increment eligibility counts before decrementing, otherwise we 1008 * could reach 0 spuriously during the first call. 1009 */ 1010 fixjobc(p, pgrp, 1); 1011 fixjobc(p, p->p_pgrp, 0); 1012 1013 /* Interlock with ttread(). */ 1014 mutex_spin_enter(&tty_lock); 1015 1016 /* Move process to requested group. */ 1017 LIST_REMOVE(p, p_pglist); 1018 if (LIST_EMPTY(&p->p_pgrp->pg_members)) 1019 /* defer delete until we've dumped the lock */ 1020 pg_id = p->p_pgrp->pg_id; 1021 p->p_pgrp = pgrp; 1022 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 1023 1024 /* Done with the swap; we can release the tty mutex. */ 1025 mutex_spin_exit(&tty_lock); 1026 1027 done: 1028 if (pg_id != NO_PGID) { 1029 /* Releases proc_lock. */ 1030 pg_delete(pg_id); 1031 } else { 1032 mutex_exit(proc_lock); 1033 } 1034 if (sess != NULL) 1035 kmem_free(sess, sizeof(*sess)); 1036 if (new_pgrp != NULL) 1037 kmem_free(new_pgrp, sizeof(*new_pgrp)); 1038 #ifdef DEBUG_PGRP 1039 if (__predict_false(rval)) 1040 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n", 1041 pid, pgid, mksess, curp->p_pid, rval); 1042 #endif 1043 return rval; 1044 } 1045 1046 /* 1047 * proc_leavepgrp: remove a process from its process group. 1048 * => must be called with the proc_lock held, which will be released; 1049 */ 1050 void 1051 proc_leavepgrp(struct proc *p) 1052 { 1053 struct pgrp *pgrp; 1054 1055 KASSERT(mutex_owned(proc_lock)); 1056 1057 /* Interlock with ttread() */ 1058 mutex_spin_enter(&tty_lock); 1059 pgrp = p->p_pgrp; 1060 LIST_REMOVE(p, p_pglist); 1061 p->p_pgrp = NULL; 1062 mutex_spin_exit(&tty_lock); 1063 1064 if (LIST_EMPTY(&pgrp->pg_members)) { 1065 /* Releases proc_lock. */ 1066 pg_delete(pgrp->pg_id); 1067 } else { 1068 mutex_exit(proc_lock); 1069 } 1070 } 1071 1072 /* 1073 * pg_remove: remove a process group from the table. 1074 * => must be called with the proc_lock held; 1075 * => returns process group to free; 1076 */ 1077 static struct pgrp * 1078 pg_remove(pid_t pg_id) 1079 { 1080 struct pgrp *pgrp; 1081 struct pid_table *pt; 1082 1083 KASSERT(mutex_owned(proc_lock)); 1084 1085 pt = &pid_table[pg_id & pid_tbl_mask]; 1086 pgrp = pt->pt_pgrp; 1087 1088 KASSERT(pgrp != NULL); 1089 KASSERT(pgrp->pg_id == pg_id); 1090 KASSERT(LIST_EMPTY(&pgrp->pg_members)); 1091 1092 pt->pt_pgrp = NULL; 1093 1094 if (!P_VALID(pt->pt_proc)) { 1095 /* Orphaned pgrp, put slot onto free list. */ 1096 KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0); 1097 pg_id &= pid_tbl_mask; 1098 pt = &pid_table[last_free_pt]; 1099 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id); 1100 KASSERT(pt->pt_pid == 0); 1101 last_free_pt = pg_id; 1102 pid_alloc_cnt--; 1103 } 1104 return pgrp; 1105 } 1106 1107 /* 1108 * pg_delete: delete and free a process group. 1109 * => must be called with the proc_lock held, which will be released. 1110 */ 1111 static void 1112 pg_delete(pid_t pg_id) 1113 { 1114 struct pgrp *pg; 1115 struct tty *ttyp; 1116 struct session *ss; 1117 1118 KASSERT(mutex_owned(proc_lock)); 1119 1120 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp; 1121 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) { 1122 mutex_exit(proc_lock); 1123 return; 1124 } 1125 1126 ss = pg->pg_session; 1127 1128 /* Remove reference (if any) from tty to this process group */ 1129 mutex_spin_enter(&tty_lock); 1130 ttyp = ss->s_ttyp; 1131 if (ttyp != NULL && ttyp->t_pgrp == pg) { 1132 ttyp->t_pgrp = NULL; 1133 KASSERT(ttyp->t_session == ss); 1134 } 1135 mutex_spin_exit(&tty_lock); 1136 1137 /* 1138 * The leading process group in a session is freed by proc_sessrele(), 1139 * if last reference. Note: proc_sessrele() releases proc_lock. 1140 */ 1141 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL; 1142 proc_sessrele(ss); 1143 1144 if (pg != NULL) { 1145 /* Free it, if was not done by proc_sessrele(). */ 1146 kmem_free(pg, sizeof(struct pgrp)); 1147 } 1148 } 1149 1150 /* 1151 * Adjust pgrp jobc counters when specified process changes process group. 1152 * We count the number of processes in each process group that "qualify" 1153 * the group for terminal job control (those with a parent in a different 1154 * process group of the same session). If that count reaches zero, the 1155 * process group becomes orphaned. Check both the specified process' 1156 * process group and that of its children. 1157 * entering == 0 => p is leaving specified group. 1158 * entering == 1 => p is entering specified group. 1159 * 1160 * Call with proc_lock held. 1161 */ 1162 void 1163 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 1164 { 1165 struct pgrp *hispgrp; 1166 struct session *mysession = pgrp->pg_session; 1167 struct proc *child; 1168 1169 KASSERT(mutex_owned(proc_lock)); 1170 1171 /* 1172 * Check p's parent to see whether p qualifies its own process 1173 * group; if so, adjust count for p's process group. 1174 */ 1175 hispgrp = p->p_pptr->p_pgrp; 1176 if (hispgrp != pgrp && hispgrp->pg_session == mysession) { 1177 if (entering) { 1178 pgrp->pg_jobc++; 1179 p->p_lflag &= ~PL_ORPHANPG; 1180 } else if (--pgrp->pg_jobc == 0) 1181 orphanpg(pgrp); 1182 } 1183 1184 /* 1185 * Check this process' children to see whether they qualify 1186 * their process groups; if so, adjust counts for children's 1187 * process groups. 1188 */ 1189 LIST_FOREACH(child, &p->p_children, p_sibling) { 1190 hispgrp = child->p_pgrp; 1191 if (hispgrp != pgrp && hispgrp->pg_session == mysession && 1192 !P_ZOMBIE(child)) { 1193 if (entering) { 1194 child->p_lflag &= ~PL_ORPHANPG; 1195 hispgrp->pg_jobc++; 1196 } else if (--hispgrp->pg_jobc == 0) 1197 orphanpg(hispgrp); 1198 } 1199 } 1200 } 1201 1202 /* 1203 * A process group has become orphaned; 1204 * if there are any stopped processes in the group, 1205 * hang-up all process in that group. 1206 * 1207 * Call with proc_lock held. 1208 */ 1209 static void 1210 orphanpg(struct pgrp *pg) 1211 { 1212 struct proc *p; 1213 1214 KASSERT(mutex_owned(proc_lock)); 1215 1216 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1217 if (p->p_stat == SSTOP) { 1218 p->p_lflag |= PL_ORPHANPG; 1219 psignal(p, SIGHUP); 1220 psignal(p, SIGCONT); 1221 } 1222 } 1223 } 1224 1225 #ifdef DDB 1226 #include <ddb/db_output.h> 1227 void pidtbl_dump(void); 1228 void 1229 pidtbl_dump(void) 1230 { 1231 struct pid_table *pt; 1232 struct proc *p; 1233 struct pgrp *pgrp; 1234 int id; 1235 1236 db_printf("pid table %p size %x, next %x, last %x\n", 1237 pid_table, pid_tbl_mask+1, 1238 next_free_pt, last_free_pt); 1239 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) { 1240 p = pt->pt_proc; 1241 if (!P_VALID(p) && !pt->pt_pgrp) 1242 continue; 1243 db_printf(" id %x: ", id); 1244 if (P_VALID(p)) 1245 db_printf("slotpid %d proc %p id %d (0x%x) %s\n", 1246 pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm); 1247 else 1248 db_printf("next %x use %x\n", 1249 P_NEXT(p) & pid_tbl_mask, 1250 P_NEXT(p) & ~pid_tbl_mask); 1251 if ((pgrp = pt->pt_pgrp)) { 1252 db_printf("\tsession %p, sid %d, count %d, login %s\n", 1253 pgrp->pg_session, pgrp->pg_session->s_sid, 1254 pgrp->pg_session->s_count, 1255 pgrp->pg_session->s_login); 1256 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n", 1257 pgrp, pgrp->pg_id, pgrp->pg_jobc, 1258 LIST_FIRST(&pgrp->pg_members)); 1259 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 1260 db_printf("\t\tpid %d addr %p pgrp %p %s\n", 1261 p->p_pid, p, p->p_pgrp, p->p_comm); 1262 } 1263 } 1264 } 1265 } 1266 #endif /* DDB */ 1267 1268 #ifdef KSTACK_CHECK_MAGIC 1269 1270 #define KSTACK_MAGIC 0xdeadbeaf 1271 1272 /* XXX should be per process basis? */ 1273 static int kstackleftmin = KSTACK_SIZE; 1274 static int kstackleftthres = KSTACK_SIZE / 8; 1275 1276 void 1277 kstack_setup_magic(const struct lwp *l) 1278 { 1279 uint32_t *ip; 1280 uint32_t const *end; 1281 1282 KASSERT(l != NULL); 1283 KASSERT(l != &lwp0); 1284 1285 /* 1286 * fill all the stack with magic number 1287 * so that later modification on it can be detected. 1288 */ 1289 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1290 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1291 for (; ip < end; ip++) { 1292 *ip = KSTACK_MAGIC; 1293 } 1294 } 1295 1296 void 1297 kstack_check_magic(const struct lwp *l) 1298 { 1299 uint32_t const *ip, *end; 1300 int stackleft; 1301 1302 KASSERT(l != NULL); 1303 1304 /* don't check proc0 */ /*XXX*/ 1305 if (l == &lwp0) 1306 return; 1307 1308 #ifdef __MACHINE_STACK_GROWS_UP 1309 /* stack grows upwards (eg. hppa) */ 1310 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1311 end = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1312 for (ip--; ip >= end; ip--) 1313 if (*ip != KSTACK_MAGIC) 1314 break; 1315 1316 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip; 1317 #else /* __MACHINE_STACK_GROWS_UP */ 1318 /* stack grows downwards (eg. i386) */ 1319 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1320 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1321 for (; ip < end; ip++) 1322 if (*ip != KSTACK_MAGIC) 1323 break; 1324 1325 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l); 1326 #endif /* __MACHINE_STACK_GROWS_UP */ 1327 1328 if (kstackleftmin > stackleft) { 1329 kstackleftmin = stackleft; 1330 if (stackleft < kstackleftthres) 1331 printf("warning: kernel stack left %d bytes" 1332 "(pid %u:lid %u)\n", stackleft, 1333 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1334 } 1335 1336 if (stackleft <= 0) { 1337 panic("magic on the top of kernel stack changed for " 1338 "pid %u, lid %u: maybe kernel stack overflow", 1339 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1340 } 1341 } 1342 #endif /* KSTACK_CHECK_MAGIC */ 1343 1344 int 1345 proclist_foreach_call(struct proclist *list, 1346 int (*callback)(struct proc *, void *arg), void *arg) 1347 { 1348 struct proc marker; 1349 struct proc *p; 1350 int ret = 0; 1351 1352 marker.p_flag = PK_MARKER; 1353 mutex_enter(proc_lock); 1354 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) { 1355 if (p->p_flag & PK_MARKER) { 1356 p = LIST_NEXT(p, p_list); 1357 continue; 1358 } 1359 LIST_INSERT_AFTER(p, &marker, p_list); 1360 ret = (*callback)(p, arg); 1361 KASSERT(mutex_owned(proc_lock)); 1362 p = LIST_NEXT(&marker, p_list); 1363 LIST_REMOVE(&marker, p_list); 1364 } 1365 mutex_exit(proc_lock); 1366 1367 return ret; 1368 } 1369 1370 int 1371 proc_vmspace_getref(struct proc *p, struct vmspace **vm) 1372 { 1373 1374 /* XXXCDC: how should locking work here? */ 1375 1376 /* curproc exception is for coredump. */ 1377 1378 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) || 1379 (p->p_vmspace->vm_refcnt < 1)) { /* XXX */ 1380 return EFAULT; 1381 } 1382 1383 uvmspace_addref(p->p_vmspace); 1384 *vm = p->p_vmspace; 1385 1386 return 0; 1387 } 1388 1389 /* 1390 * Acquire a write lock on the process credential. 1391 */ 1392 void 1393 proc_crmod_enter(void) 1394 { 1395 struct lwp *l = curlwp; 1396 struct proc *p = l->l_proc; 1397 kauth_cred_t oc; 1398 1399 /* Reset what needs to be reset in plimit. */ 1400 if (p->p_limit->pl_corename != defcorename) { 1401 lim_setcorename(p, defcorename, 0); 1402 } 1403 1404 mutex_enter(p->p_lock); 1405 1406 /* Ensure the LWP cached credentials are up to date. */ 1407 if ((oc = l->l_cred) != p->p_cred) { 1408 kauth_cred_hold(p->p_cred); 1409 l->l_cred = p->p_cred; 1410 kauth_cred_free(oc); 1411 } 1412 } 1413 1414 /* 1415 * Set in a new process credential, and drop the write lock. The credential 1416 * must have a reference already. Optionally, free a no-longer required 1417 * credential. The scheduler also needs to inspect p_cred, so we also 1418 * briefly acquire the sched state mutex. 1419 */ 1420 void 1421 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid) 1422 { 1423 struct lwp *l = curlwp, *l2; 1424 struct proc *p = l->l_proc; 1425 kauth_cred_t oc; 1426 1427 KASSERT(mutex_owned(p->p_lock)); 1428 1429 /* Is there a new credential to set in? */ 1430 if (scred != NULL) { 1431 p->p_cred = scred; 1432 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 1433 if (l2 != l) 1434 l2->l_prflag |= LPR_CRMOD; 1435 } 1436 1437 /* Ensure the LWP cached credentials are up to date. */ 1438 if ((oc = l->l_cred) != scred) { 1439 kauth_cred_hold(scred); 1440 l->l_cred = scred; 1441 } 1442 } else 1443 oc = NULL; /* XXXgcc */ 1444 1445 if (sugid) { 1446 /* 1447 * Mark process as having changed credentials, stops 1448 * tracing etc. 1449 */ 1450 p->p_flag |= PK_SUGID; 1451 } 1452 1453 mutex_exit(p->p_lock); 1454 1455 /* If there is a credential to be released, free it now. */ 1456 if (fcred != NULL) { 1457 KASSERT(scred != NULL); 1458 kauth_cred_free(fcred); 1459 if (oc != scred) 1460 kauth_cred_free(oc); 1461 } 1462 } 1463 1464 /* 1465 * proc_specific_key_create -- 1466 * Create a key for subsystem proc-specific data. 1467 */ 1468 int 1469 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 1470 { 1471 1472 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor)); 1473 } 1474 1475 /* 1476 * proc_specific_key_delete -- 1477 * Delete a key for subsystem proc-specific data. 1478 */ 1479 void 1480 proc_specific_key_delete(specificdata_key_t key) 1481 { 1482 1483 specificdata_key_delete(proc_specificdata_domain, key); 1484 } 1485 1486 /* 1487 * proc_initspecific -- 1488 * Initialize a proc's specificdata container. 1489 */ 1490 void 1491 proc_initspecific(struct proc *p) 1492 { 1493 int error __diagused; 1494 1495 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref); 1496 KASSERT(error == 0); 1497 } 1498 1499 /* 1500 * proc_finispecific -- 1501 * Finalize a proc's specificdata container. 1502 */ 1503 void 1504 proc_finispecific(struct proc *p) 1505 { 1506 1507 specificdata_fini(proc_specificdata_domain, &p->p_specdataref); 1508 } 1509 1510 /* 1511 * proc_getspecific -- 1512 * Return proc-specific data corresponding to the specified key. 1513 */ 1514 void * 1515 proc_getspecific(struct proc *p, specificdata_key_t key) 1516 { 1517 1518 return (specificdata_getspecific(proc_specificdata_domain, 1519 &p->p_specdataref, key)); 1520 } 1521 1522 /* 1523 * proc_setspecific -- 1524 * Set proc-specific data corresponding to the specified key. 1525 */ 1526 void 1527 proc_setspecific(struct proc *p, specificdata_key_t key, void *data) 1528 { 1529 1530 specificdata_setspecific(proc_specificdata_domain, 1531 &p->p_specdataref, key, data); 1532 } 1533 1534 int 1535 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target) 1536 { 1537 int r = 0; 1538 1539 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) || 1540 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) { 1541 /* 1542 * suid proc of ours or proc not ours 1543 */ 1544 r = EPERM; 1545 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) { 1546 /* 1547 * sgid proc has sgid back to us temporarily 1548 */ 1549 r = EPERM; 1550 } else { 1551 /* 1552 * our rgid must be in target's group list (ie, 1553 * sub-processes started by a sgid process) 1554 */ 1555 int ismember = 0; 1556 1557 if (kauth_cred_ismember_gid(cred, 1558 kauth_cred_getgid(target), &ismember) != 0 || 1559 !ismember) 1560 r = EPERM; 1561 } 1562 1563 return (r); 1564 } 1565 1566 /* 1567 * sysctl stuff 1568 */ 1569 1570 #define KERN_PROCSLOP (5 * sizeof(struct kinfo_proc)) 1571 1572 static const u_int sysctl_flagmap[] = { 1573 PK_ADVLOCK, P_ADVLOCK, 1574 PK_EXEC, P_EXEC, 1575 PK_NOCLDWAIT, P_NOCLDWAIT, 1576 PK_32, P_32, 1577 PK_CLDSIGIGN, P_CLDSIGIGN, 1578 PK_SUGID, P_SUGID, 1579 0 1580 }; 1581 1582 static const u_int sysctl_sflagmap[] = { 1583 PS_NOCLDSTOP, P_NOCLDSTOP, 1584 PS_WEXIT, P_WEXIT, 1585 PS_STOPFORK, P_STOPFORK, 1586 PS_STOPEXEC, P_STOPEXEC, 1587 PS_STOPEXIT, P_STOPEXIT, 1588 0 1589 }; 1590 1591 static const u_int sysctl_slflagmap[] = { 1592 PSL_TRACED, P_TRACED, 1593 PSL_CHTRACED, P_CHTRACED, 1594 PSL_SYSCALL, P_SYSCALL, 1595 0 1596 }; 1597 1598 static const u_int sysctl_lflagmap[] = { 1599 PL_CONTROLT, P_CONTROLT, 1600 PL_PPWAIT, P_PPWAIT, 1601 0 1602 }; 1603 1604 static const u_int sysctl_stflagmap[] = { 1605 PST_PROFIL, P_PROFIL, 1606 0 1607 1608 }; 1609 1610 /* used by kern_lwp also */ 1611 const u_int sysctl_lwpflagmap[] = { 1612 LW_SINTR, L_SINTR, 1613 LW_SYSTEM, L_SYSTEM, 1614 0 1615 }; 1616 1617 /* 1618 * Find the most ``active'' lwp of a process and return it for ps display 1619 * purposes 1620 */ 1621 static struct lwp * 1622 proc_active_lwp(struct proc *p) 1623 { 1624 static const int ostat[] = { 1625 0, 1626 2, /* LSIDL */ 1627 6, /* LSRUN */ 1628 5, /* LSSLEEP */ 1629 4, /* LSSTOP */ 1630 0, /* LSZOMB */ 1631 1, /* LSDEAD */ 1632 7, /* LSONPROC */ 1633 3 /* LSSUSPENDED */ 1634 }; 1635 1636 struct lwp *l, *lp = NULL; 1637 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 1638 KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat)); 1639 if (lp == NULL || 1640 ostat[l->l_stat] > ostat[lp->l_stat] || 1641 (ostat[l->l_stat] == ostat[lp->l_stat] && 1642 l->l_cpticks > lp->l_cpticks)) { 1643 lp = l; 1644 continue; 1645 } 1646 } 1647 return lp; 1648 } 1649 1650 static int 1651 sysctl_doeproc(SYSCTLFN_ARGS) 1652 { 1653 union { 1654 struct kinfo_proc kproc; 1655 struct kinfo_proc2 kproc2; 1656 } *kbuf; 1657 struct proc *p, *next, *marker; 1658 char *where, *dp; 1659 int type, op, arg, error; 1660 u_int elem_size, kelem_size, elem_count; 1661 size_t buflen, needed; 1662 bool match, zombie, mmmbrains; 1663 const bool allowaddr = get_expose_address(curproc); 1664 1665 if (namelen == 1 && name[0] == CTL_QUERY) 1666 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1667 1668 dp = where = oldp; 1669 buflen = where != NULL ? *oldlenp : 0; 1670 error = 0; 1671 needed = 0; 1672 type = rnode->sysctl_num; 1673 1674 if (type == KERN_PROC) { 1675 if (namelen == 0) 1676 return EINVAL; 1677 switch (op = name[0]) { 1678 case KERN_PROC_ALL: 1679 if (namelen != 1) 1680 return EINVAL; 1681 arg = 0; 1682 break; 1683 default: 1684 if (namelen != 2) 1685 return EINVAL; 1686 arg = name[1]; 1687 break; 1688 } 1689 elem_count = 0; /* Hush little compiler, don't you cry */ 1690 kelem_size = elem_size = sizeof(kbuf->kproc); 1691 } else { 1692 if (namelen != 4) 1693 return EINVAL; 1694 op = name[0]; 1695 arg = name[1]; 1696 elem_size = name[2]; 1697 elem_count = name[3]; 1698 kelem_size = sizeof(kbuf->kproc2); 1699 } 1700 1701 sysctl_unlock(); 1702 1703 kbuf = kmem_zalloc(sizeof(*kbuf), KM_SLEEP); 1704 marker = kmem_alloc(sizeof(*marker), KM_SLEEP); 1705 marker->p_flag = PK_MARKER; 1706 1707 mutex_enter(proc_lock); 1708 /* 1709 * Start with zombies to prevent reporting processes twice, in case they 1710 * are dying and being moved from the list of alive processes to zombies. 1711 */ 1712 mmmbrains = true; 1713 for (p = LIST_FIRST(&zombproc);; p = next) { 1714 if (p == NULL) { 1715 if (mmmbrains) { 1716 p = LIST_FIRST(&allproc); 1717 mmmbrains = false; 1718 } 1719 if (p == NULL) 1720 break; 1721 } 1722 next = LIST_NEXT(p, p_list); 1723 if ((p->p_flag & PK_MARKER) != 0) 1724 continue; 1725 1726 /* 1727 * Skip embryonic processes. 1728 */ 1729 if (p->p_stat == SIDL) 1730 continue; 1731 1732 mutex_enter(p->p_lock); 1733 error = kauth_authorize_process(l->l_cred, 1734 KAUTH_PROCESS_CANSEE, p, 1735 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL); 1736 if (error != 0) { 1737 mutex_exit(p->p_lock); 1738 continue; 1739 } 1740 1741 /* 1742 * Hande all the operations in one switch on the cost of 1743 * algorithm complexity is on purpose. The win splitting this 1744 * function into several similar copies makes maintenance burden 1745 * burden, code grow and boost is neglible in practical systems. 1746 */ 1747 switch (op) { 1748 case KERN_PROC_PID: 1749 match = (p->p_pid == (pid_t)arg); 1750 break; 1751 1752 case KERN_PROC_PGRP: 1753 match = (p->p_pgrp->pg_id == (pid_t)arg); 1754 break; 1755 1756 case KERN_PROC_SESSION: 1757 match = (p->p_session->s_sid == (pid_t)arg); 1758 break; 1759 1760 case KERN_PROC_TTY: 1761 match = true; 1762 if (arg == (int) KERN_PROC_TTY_REVOKE) { 1763 if ((p->p_lflag & PL_CONTROLT) == 0 || 1764 p->p_session->s_ttyp == NULL || 1765 p->p_session->s_ttyvp != NULL) { 1766 match = false; 1767 } 1768 } else if ((p->p_lflag & PL_CONTROLT) == 0 || 1769 p->p_session->s_ttyp == NULL) { 1770 if ((dev_t)arg != KERN_PROC_TTY_NODEV) { 1771 match = false; 1772 } 1773 } else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) { 1774 match = false; 1775 } 1776 break; 1777 1778 case KERN_PROC_UID: 1779 match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg); 1780 break; 1781 1782 case KERN_PROC_RUID: 1783 match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg); 1784 break; 1785 1786 case KERN_PROC_GID: 1787 match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg); 1788 break; 1789 1790 case KERN_PROC_RGID: 1791 match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg); 1792 break; 1793 1794 case KERN_PROC_ALL: 1795 match = true; 1796 /* allow everything */ 1797 break; 1798 1799 default: 1800 error = EINVAL; 1801 mutex_exit(p->p_lock); 1802 goto cleanup; 1803 } 1804 if (!match) { 1805 mutex_exit(p->p_lock); 1806 continue; 1807 } 1808 1809 /* 1810 * Grab a hold on the process. 1811 */ 1812 if (mmmbrains) { 1813 zombie = true; 1814 } else { 1815 zombie = !rw_tryenter(&p->p_reflock, RW_READER); 1816 } 1817 if (zombie) { 1818 LIST_INSERT_AFTER(p, marker, p_list); 1819 } 1820 1821 if (buflen >= elem_size && 1822 (type == KERN_PROC || elem_count > 0)) { 1823 ruspace(p); /* Update process vm resource use */ 1824 1825 if (type == KERN_PROC) { 1826 fill_proc(p, &kbuf->kproc.kp_proc, allowaddr); 1827 fill_eproc(p, &kbuf->kproc.kp_eproc, zombie, 1828 allowaddr); 1829 } else { 1830 fill_kproc2(p, &kbuf->kproc2, zombie, 1831 allowaddr); 1832 elem_count--; 1833 } 1834 mutex_exit(p->p_lock); 1835 mutex_exit(proc_lock); 1836 /* 1837 * Copy out elem_size, but not larger than kelem_size 1838 */ 1839 error = sysctl_copyout(l, kbuf, dp, 1840 uimin(kelem_size, elem_size)); 1841 mutex_enter(proc_lock); 1842 if (error) { 1843 goto bah; 1844 } 1845 dp += elem_size; 1846 buflen -= elem_size; 1847 } else { 1848 mutex_exit(p->p_lock); 1849 } 1850 needed += elem_size; 1851 1852 /* 1853 * Release reference to process. 1854 */ 1855 if (zombie) { 1856 next = LIST_NEXT(marker, p_list); 1857 LIST_REMOVE(marker, p_list); 1858 } else { 1859 rw_exit(&p->p_reflock); 1860 next = LIST_NEXT(p, p_list); 1861 } 1862 1863 /* 1864 * Short-circuit break quickly! 1865 */ 1866 if (op == KERN_PROC_PID) 1867 break; 1868 } 1869 mutex_exit(proc_lock); 1870 1871 if (where != NULL) { 1872 *oldlenp = dp - where; 1873 if (needed > *oldlenp) { 1874 error = ENOMEM; 1875 goto out; 1876 } 1877 } else { 1878 needed += KERN_PROCSLOP; 1879 *oldlenp = needed; 1880 } 1881 kmem_free(kbuf, sizeof(*kbuf)); 1882 kmem_free(marker, sizeof(*marker)); 1883 sysctl_relock(); 1884 return 0; 1885 bah: 1886 if (zombie) 1887 LIST_REMOVE(marker, p_list); 1888 else 1889 rw_exit(&p->p_reflock); 1890 cleanup: 1891 mutex_exit(proc_lock); 1892 out: 1893 kmem_free(kbuf, sizeof(*kbuf)); 1894 kmem_free(marker, sizeof(*marker)); 1895 sysctl_relock(); 1896 return error; 1897 } 1898 1899 int 1900 copyin_psstrings(struct proc *p, struct ps_strings *arginfo) 1901 { 1902 #if !defined(_RUMPKERNEL) 1903 int retval; 1904 1905 if (p->p_flag & PK_32) { 1906 MODULE_HOOK_CALL(kern_proc32_copyin_hook, (p, arginfo), 1907 enosys(), retval); 1908 return retval; 1909 } 1910 #endif /* !defined(_RUMPKERNEL) */ 1911 1912 return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo)); 1913 } 1914 1915 static int 1916 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len) 1917 { 1918 void **cookie = cookie_; 1919 struct lwp *l = cookie[0]; 1920 char *dst = cookie[1]; 1921 1922 return sysctl_copyout(l, src, dst + off, len); 1923 } 1924 1925 /* 1926 * sysctl helper routine for kern.proc_args pseudo-subtree. 1927 */ 1928 static int 1929 sysctl_kern_proc_args(SYSCTLFN_ARGS) 1930 { 1931 struct ps_strings pss; 1932 struct proc *p; 1933 pid_t pid; 1934 int type, error; 1935 void *cookie[2]; 1936 1937 if (namelen == 1 && name[0] == CTL_QUERY) 1938 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1939 1940 if (newp != NULL || namelen != 2) 1941 return (EINVAL); 1942 pid = name[0]; 1943 type = name[1]; 1944 1945 switch (type) { 1946 case KERN_PROC_PATHNAME: 1947 sysctl_unlock(); 1948 error = fill_pathname(l, pid, oldp, oldlenp); 1949 sysctl_relock(); 1950 return error; 1951 1952 case KERN_PROC_CWD: 1953 sysctl_unlock(); 1954 error = fill_cwd(l, pid, oldp, oldlenp); 1955 sysctl_relock(); 1956 return error; 1957 1958 case KERN_PROC_ARGV: 1959 case KERN_PROC_NARGV: 1960 case KERN_PROC_ENV: 1961 case KERN_PROC_NENV: 1962 /* ok */ 1963 break; 1964 default: 1965 return (EINVAL); 1966 } 1967 1968 sysctl_unlock(); 1969 1970 /* check pid */ 1971 mutex_enter(proc_lock); 1972 if ((p = proc_find(pid)) == NULL) { 1973 error = EINVAL; 1974 goto out_locked; 1975 } 1976 mutex_enter(p->p_lock); 1977 1978 /* Check permission. */ 1979 if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV) 1980 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 1981 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL); 1982 else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV) 1983 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 1984 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL); 1985 else 1986 error = EINVAL; /* XXXGCC */ 1987 if (error) { 1988 mutex_exit(p->p_lock); 1989 goto out_locked; 1990 } 1991 1992 if (oldp == NULL) { 1993 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) 1994 *oldlenp = sizeof (int); 1995 else 1996 *oldlenp = ARG_MAX; /* XXX XXX XXX */ 1997 error = 0; 1998 mutex_exit(p->p_lock); 1999 goto out_locked; 2000 } 2001 2002 /* 2003 * Zombies don't have a stack, so we can't read their psstrings. 2004 * System processes also don't have a user stack. 2005 */ 2006 if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) { 2007 error = EINVAL; 2008 mutex_exit(p->p_lock); 2009 goto out_locked; 2010 } 2011 2012 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 2013 mutex_exit(p->p_lock); 2014 if (error) { 2015 goto out_locked; 2016 } 2017 mutex_exit(proc_lock); 2018 2019 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) { 2020 int value; 2021 if ((error = copyin_psstrings(p, &pss)) == 0) { 2022 if (type == KERN_PROC_NARGV) 2023 value = pss.ps_nargvstr; 2024 else 2025 value = pss.ps_nenvstr; 2026 error = sysctl_copyout(l, &value, oldp, sizeof(value)); 2027 *oldlenp = sizeof(value); 2028 } 2029 } else { 2030 cookie[0] = l; 2031 cookie[1] = oldp; 2032 error = copy_procargs(p, type, oldlenp, 2033 copy_procargs_sysctl_cb, cookie); 2034 } 2035 rw_exit(&p->p_reflock); 2036 sysctl_relock(); 2037 return error; 2038 2039 out_locked: 2040 mutex_exit(proc_lock); 2041 sysctl_relock(); 2042 return error; 2043 } 2044 2045 int 2046 copy_procargs(struct proc *p, int oid, size_t *limit, 2047 int (*cb)(void *, const void *, size_t, size_t), void *cookie) 2048 { 2049 struct ps_strings pss; 2050 size_t len, i, loaded, entry_len; 2051 struct uio auio; 2052 struct iovec aiov; 2053 int error, argvlen; 2054 char *arg; 2055 char **argv; 2056 vaddr_t user_argv; 2057 struct vmspace *vmspace; 2058 2059 /* 2060 * Allocate a temporary buffer to hold the argument vector and 2061 * the arguments themselve. 2062 */ 2063 arg = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2064 argv = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2065 2066 /* 2067 * Lock the process down in memory. 2068 */ 2069 vmspace = p->p_vmspace; 2070 uvmspace_addref(vmspace); 2071 2072 /* 2073 * Read in the ps_strings structure. 2074 */ 2075 if ((error = copyin_psstrings(p, &pss)) != 0) 2076 goto done; 2077 2078 /* 2079 * Now read the address of the argument vector. 2080 */ 2081 switch (oid) { 2082 case KERN_PROC_ARGV: 2083 user_argv = (uintptr_t)pss.ps_argvstr; 2084 argvlen = pss.ps_nargvstr; 2085 break; 2086 case KERN_PROC_ENV: 2087 user_argv = (uintptr_t)pss.ps_envstr; 2088 argvlen = pss.ps_nenvstr; 2089 break; 2090 default: 2091 error = EINVAL; 2092 goto done; 2093 } 2094 2095 if (argvlen < 0) { 2096 error = EIO; 2097 goto done; 2098 } 2099 2100 2101 /* 2102 * Now copy each string. 2103 */ 2104 len = 0; /* bytes written to user buffer */ 2105 loaded = 0; /* bytes from argv already processed */ 2106 i = 0; /* To make compiler happy */ 2107 entry_len = PROC_PTRSZ(p); 2108 2109 for (; argvlen; --argvlen) { 2110 int finished = 0; 2111 vaddr_t base; 2112 size_t xlen; 2113 int j; 2114 2115 if (loaded == 0) { 2116 size_t rem = entry_len * argvlen; 2117 loaded = MIN(rem, PAGE_SIZE); 2118 error = copyin_vmspace(vmspace, 2119 (const void *)user_argv, argv, loaded); 2120 if (error) 2121 break; 2122 user_argv += loaded; 2123 i = 0; 2124 } 2125 2126 #if !defined(_RUMPKERNEL) 2127 if (p->p_flag & PK_32) 2128 MODULE_HOOK_CALL(kern_proc32_base_hook, 2129 (argv, i++), 0, base); 2130 else 2131 #endif /* !defined(_RUMPKERNEL) */ 2132 base = (vaddr_t)argv[i++]; 2133 loaded -= entry_len; 2134 2135 /* 2136 * The program has messed around with its arguments, 2137 * possibly deleting some, and replacing them with 2138 * NULL's. Treat this as the last argument and not 2139 * a failure. 2140 */ 2141 if (base == 0) 2142 break; 2143 2144 while (!finished) { 2145 xlen = PAGE_SIZE - (base & PAGE_MASK); 2146 2147 aiov.iov_base = arg; 2148 aiov.iov_len = PAGE_SIZE; 2149 auio.uio_iov = &aiov; 2150 auio.uio_iovcnt = 1; 2151 auio.uio_offset = base; 2152 auio.uio_resid = xlen; 2153 auio.uio_rw = UIO_READ; 2154 UIO_SETUP_SYSSPACE(&auio); 2155 error = uvm_io(&vmspace->vm_map, &auio, 0); 2156 if (error) 2157 goto done; 2158 2159 /* Look for the end of the string */ 2160 for (j = 0; j < xlen; j++) { 2161 if (arg[j] == '\0') { 2162 xlen = j + 1; 2163 finished = 1; 2164 break; 2165 } 2166 } 2167 2168 /* Check for user buffer overflow */ 2169 if (len + xlen > *limit) { 2170 finished = 1; 2171 if (len > *limit) 2172 xlen = 0; 2173 else 2174 xlen = *limit - len; 2175 } 2176 2177 /* Copyout the page */ 2178 error = (*cb)(cookie, arg, len, xlen); 2179 if (error) 2180 goto done; 2181 2182 len += xlen; 2183 base += xlen; 2184 } 2185 } 2186 *limit = len; 2187 2188 done: 2189 kmem_free(argv, PAGE_SIZE); 2190 kmem_free(arg, PAGE_SIZE); 2191 uvmspace_free(vmspace); 2192 return error; 2193 } 2194 2195 /* 2196 * Fill in a proc structure for the specified process. 2197 */ 2198 static void 2199 fill_proc(const struct proc *psrc, struct proc *p, bool allowaddr) 2200 { 2201 COND_SET_VALUE(p->p_list, psrc->p_list, allowaddr); 2202 COND_SET_VALUE(p->p_auxlock, psrc->p_auxlock, allowaddr); 2203 COND_SET_VALUE(p->p_lock, psrc->p_lock, allowaddr); 2204 COND_SET_VALUE(p->p_stmutex, psrc->p_stmutex, allowaddr); 2205 COND_SET_VALUE(p->p_reflock, psrc->p_reflock, allowaddr); 2206 COND_SET_VALUE(p->p_waitcv, psrc->p_waitcv, allowaddr); 2207 COND_SET_VALUE(p->p_lwpcv, psrc->p_lwpcv, allowaddr); 2208 COND_SET_VALUE(p->p_cred, psrc->p_cred, allowaddr); 2209 COND_SET_VALUE(p->p_fd, psrc->p_fd, allowaddr); 2210 COND_SET_VALUE(p->p_cwdi, psrc->p_cwdi, allowaddr); 2211 COND_SET_VALUE(p->p_stats, psrc->p_stats, allowaddr); 2212 COND_SET_VALUE(p->p_limit, psrc->p_limit, allowaddr); 2213 COND_SET_VALUE(p->p_vmspace, psrc->p_vmspace, allowaddr); 2214 COND_SET_VALUE(p->p_sigacts, psrc->p_sigacts, allowaddr); 2215 COND_SET_VALUE(p->p_aio, psrc->p_aio, allowaddr); 2216 p->p_mqueue_cnt = psrc->p_mqueue_cnt; 2217 COND_SET_VALUE(p->p_specdataref, psrc->p_specdataref, allowaddr); 2218 p->p_exitsig = psrc->p_exitsig; 2219 p->p_flag = psrc->p_flag; 2220 p->p_sflag = psrc->p_sflag; 2221 p->p_slflag = psrc->p_slflag; 2222 p->p_lflag = psrc->p_lflag; 2223 p->p_stflag = psrc->p_stflag; 2224 p->p_stat = psrc->p_stat; 2225 p->p_trace_enabled = psrc->p_trace_enabled; 2226 p->p_pid = psrc->p_pid; 2227 COND_SET_VALUE(p->p_pglist, psrc->p_pglist, allowaddr); 2228 COND_SET_VALUE(p->p_pptr, psrc->p_pptr, allowaddr); 2229 COND_SET_VALUE(p->p_sibling, psrc->p_sibling, allowaddr); 2230 COND_SET_VALUE(p->p_children, psrc->p_children, allowaddr); 2231 COND_SET_VALUE(p->p_lwps, psrc->p_lwps, allowaddr); 2232 COND_SET_VALUE(p->p_raslist, psrc->p_raslist, allowaddr); 2233 p->p_nlwps = psrc->p_nlwps; 2234 p->p_nzlwps = psrc->p_nzlwps; 2235 p->p_nrlwps = psrc->p_nrlwps; 2236 p->p_nlwpwait = psrc->p_nlwpwait; 2237 p->p_ndlwps = psrc->p_ndlwps; 2238 p->p_nlwpid = psrc->p_nlwpid; 2239 p->p_nstopchild = psrc->p_nstopchild; 2240 p->p_waited = psrc->p_waited; 2241 COND_SET_VALUE(p->p_zomblwp, psrc->p_zomblwp, allowaddr); 2242 COND_SET_VALUE(p->p_vforklwp, psrc->p_vforklwp, allowaddr); 2243 COND_SET_VALUE(p->p_sched_info, psrc->p_sched_info, allowaddr); 2244 p->p_estcpu = psrc->p_estcpu; 2245 p->p_estcpu_inherited = psrc->p_estcpu_inherited; 2246 p->p_forktime = psrc->p_forktime; 2247 p->p_pctcpu = psrc->p_pctcpu; 2248 COND_SET_VALUE(p->p_opptr, psrc->p_opptr, allowaddr); 2249 COND_SET_VALUE(p->p_timers, psrc->p_timers, allowaddr); 2250 p->p_rtime = psrc->p_rtime; 2251 p->p_uticks = psrc->p_uticks; 2252 p->p_sticks = psrc->p_sticks; 2253 p->p_iticks = psrc->p_iticks; 2254 p->p_xutime = psrc->p_xutime; 2255 p->p_xstime = psrc->p_xstime; 2256 p->p_traceflag = psrc->p_traceflag; 2257 COND_SET_VALUE(p->p_tracep, psrc->p_tracep, allowaddr); 2258 COND_SET_VALUE(p->p_textvp, psrc->p_textvp, allowaddr); 2259 COND_SET_VALUE(p->p_emul, psrc->p_emul, allowaddr); 2260 COND_SET_VALUE(p->p_emuldata, psrc->p_emuldata, allowaddr); 2261 COND_SET_VALUE(p->p_execsw, psrc->p_execsw, allowaddr); 2262 COND_SET_VALUE(p->p_klist, psrc->p_klist, allowaddr); 2263 COND_SET_VALUE(p->p_sigwaiters, psrc->p_sigwaiters, allowaddr); 2264 COND_SET_VALUE(p->p_sigpend, psrc->p_sigpend, allowaddr); 2265 COND_SET_VALUE(p->p_lwpctl, psrc->p_lwpctl, allowaddr); 2266 p->p_ppid = psrc->p_ppid; 2267 COND_SET_VALUE(p->p_path, psrc->p_path, allowaddr); 2268 COND_SET_VALUE(p->p_sigctx, psrc->p_sigctx, allowaddr); 2269 p->p_nice = psrc->p_nice; 2270 memcpy(p->p_comm, psrc->p_comm, sizeof(p->p_comm)); 2271 COND_SET_VALUE(p->p_pgrp, psrc->p_pgrp, allowaddr); 2272 COND_SET_VALUE(p->p_psstrp, psrc->p_psstrp, allowaddr); 2273 p->p_pax = psrc->p_pax; 2274 p->p_xexit = psrc->p_xexit; 2275 p->p_xsig = psrc->p_xsig; 2276 p->p_acflag = psrc->p_acflag; 2277 COND_SET_VALUE(p->p_md, psrc->p_md, allowaddr); 2278 p->p_stackbase = psrc->p_stackbase; 2279 COND_SET_VALUE(p->p_dtrace, psrc->p_dtrace, allowaddr); 2280 } 2281 2282 /* 2283 * Fill in an eproc structure for the specified process. 2284 */ 2285 void 2286 fill_eproc(struct proc *p, struct eproc *ep, bool zombie, bool allowaddr) 2287 { 2288 struct tty *tp; 2289 struct lwp *l; 2290 2291 KASSERT(mutex_owned(proc_lock)); 2292 KASSERT(mutex_owned(p->p_lock)); 2293 2294 COND_SET_VALUE(ep->e_paddr, p, allowaddr); 2295 COND_SET_VALUE(ep->e_sess, p->p_session, allowaddr); 2296 if (p->p_cred) { 2297 kauth_cred_topcred(p->p_cred, &ep->e_pcred); 2298 kauth_cred_toucred(p->p_cred, &ep->e_ucred); 2299 } 2300 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2301 struct vmspace *vm = p->p_vmspace; 2302 2303 ep->e_vm.vm_rssize = vm_resident_count(vm); 2304 ep->e_vm.vm_tsize = vm->vm_tsize; 2305 ep->e_vm.vm_dsize = vm->vm_dsize; 2306 ep->e_vm.vm_ssize = vm->vm_ssize; 2307 ep->e_vm.vm_map.size = vm->vm_map.size; 2308 2309 /* Pick the primary (first) LWP */ 2310 l = proc_active_lwp(p); 2311 KASSERT(l != NULL); 2312 lwp_lock(l); 2313 if (l->l_wchan) 2314 strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN); 2315 lwp_unlock(l); 2316 } 2317 ep->e_ppid = p->p_ppid; 2318 if (p->p_pgrp && p->p_session) { 2319 ep->e_pgid = p->p_pgrp->pg_id; 2320 ep->e_jobc = p->p_pgrp->pg_jobc; 2321 ep->e_sid = p->p_session->s_sid; 2322 if ((p->p_lflag & PL_CONTROLT) && 2323 (tp = p->p_session->s_ttyp)) { 2324 ep->e_tdev = tp->t_dev; 2325 ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2326 COND_SET_VALUE(ep->e_tsess, tp->t_session, allowaddr); 2327 } else 2328 ep->e_tdev = (uint32_t)NODEV; 2329 ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0; 2330 if (SESS_LEADER(p)) 2331 ep->e_flag |= EPROC_SLEADER; 2332 strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME); 2333 } 2334 ep->e_xsize = ep->e_xrssize = 0; 2335 ep->e_xccount = ep->e_xswrss = 0; 2336 } 2337 2338 /* 2339 * Fill in a kinfo_proc2 structure for the specified process. 2340 */ 2341 void 2342 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie, bool allowaddr) 2343 { 2344 struct tty *tp; 2345 struct lwp *l, *l2; 2346 struct timeval ut, st, rt; 2347 sigset_t ss1, ss2; 2348 struct rusage ru; 2349 struct vmspace *vm; 2350 2351 KASSERT(mutex_owned(proc_lock)); 2352 KASSERT(mutex_owned(p->p_lock)); 2353 2354 sigemptyset(&ss1); 2355 sigemptyset(&ss2); 2356 2357 COND_SET_VALUE(ki->p_paddr, PTRTOUINT64(p), allowaddr); 2358 COND_SET_VALUE(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr); 2359 COND_SET_VALUE(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr); 2360 COND_SET_VALUE(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr); 2361 COND_SET_VALUE(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr); 2362 COND_SET_VALUE(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr); 2363 COND_SET_VALUE(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr); 2364 COND_SET_VALUE(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr); 2365 ki->p_tsess = 0; /* may be changed if controlling tty below */ 2366 COND_SET_VALUE(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr); 2367 ki->p_eflag = 0; 2368 ki->p_exitsig = p->p_exitsig; 2369 ki->p_flag = L_INMEM; /* Process never swapped out */ 2370 ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag); 2371 ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag); 2372 ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag); 2373 ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag); 2374 ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag); 2375 ki->p_pid = p->p_pid; 2376 ki->p_ppid = p->p_ppid; 2377 ki->p_uid = kauth_cred_geteuid(p->p_cred); 2378 ki->p_ruid = kauth_cred_getuid(p->p_cred); 2379 ki->p_gid = kauth_cred_getegid(p->p_cred); 2380 ki->p_rgid = kauth_cred_getgid(p->p_cred); 2381 ki->p_svuid = kauth_cred_getsvuid(p->p_cred); 2382 ki->p_svgid = kauth_cred_getsvgid(p->p_cred); 2383 ki->p_ngroups = kauth_cred_ngroups(p->p_cred); 2384 kauth_cred_getgroups(p->p_cred, ki->p_groups, 2385 uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])), 2386 UIO_SYSSPACE); 2387 2388 ki->p_uticks = p->p_uticks; 2389 ki->p_sticks = p->p_sticks; 2390 ki->p_iticks = p->p_iticks; 2391 ki->p_tpgid = NO_PGID; /* may be changed if controlling tty below */ 2392 COND_SET_VALUE(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr); 2393 ki->p_traceflag = p->p_traceflag; 2394 2395 memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t)); 2396 memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t)); 2397 2398 ki->p_cpticks = 0; 2399 ki->p_pctcpu = p->p_pctcpu; 2400 ki->p_estcpu = 0; 2401 ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */ 2402 ki->p_realstat = p->p_stat; 2403 ki->p_nice = p->p_nice; 2404 ki->p_xstat = P_WAITSTATUS(p); 2405 ki->p_acflag = p->p_acflag; 2406 2407 strncpy(ki->p_comm, p->p_comm, 2408 uimin(sizeof(ki->p_comm), sizeof(p->p_comm))); 2409 strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename)); 2410 2411 ki->p_nlwps = p->p_nlwps; 2412 ki->p_realflag = ki->p_flag; 2413 2414 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2415 vm = p->p_vmspace; 2416 ki->p_vm_rssize = vm_resident_count(vm); 2417 ki->p_vm_tsize = vm->vm_tsize; 2418 ki->p_vm_dsize = vm->vm_dsize; 2419 ki->p_vm_ssize = vm->vm_ssize; 2420 ki->p_vm_vsize = atop(vm->vm_map.size); 2421 /* 2422 * Since the stack is initially mapped mostly with 2423 * PROT_NONE and grown as needed, adjust the "mapped size" 2424 * to skip the unused stack portion. 2425 */ 2426 ki->p_vm_msize = 2427 atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize; 2428 2429 /* Pick the primary (first) LWP */ 2430 l = proc_active_lwp(p); 2431 KASSERT(l != NULL); 2432 lwp_lock(l); 2433 ki->p_nrlwps = p->p_nrlwps; 2434 ki->p_forw = 0; 2435 ki->p_back = 0; 2436 COND_SET_VALUE(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr); 2437 ki->p_stat = l->l_stat; 2438 ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag); 2439 ki->p_swtime = l->l_swtime; 2440 ki->p_slptime = l->l_slptime; 2441 if (l->l_stat == LSONPROC) 2442 ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags; 2443 else 2444 ki->p_schedflags = 0; 2445 ki->p_priority = lwp_eprio(l); 2446 ki->p_usrpri = l->l_priority; 2447 if (l->l_wchan) 2448 strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg)); 2449 COND_SET_VALUE(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr); 2450 ki->p_cpuid = cpu_index(l->l_cpu); 2451 lwp_unlock(l); 2452 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 2453 /* This is hardly correct, but... */ 2454 sigplusset(&l->l_sigpend.sp_set, &ss1); 2455 sigplusset(&l->l_sigmask, &ss2); 2456 ki->p_cpticks += l->l_cpticks; 2457 ki->p_pctcpu += l->l_pctcpu; 2458 ki->p_estcpu += l->l_estcpu; 2459 } 2460 } 2461 sigplusset(&p->p_sigpend.sp_set, &ss1); 2462 memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t)); 2463 memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t)); 2464 2465 if (p->p_session != NULL) { 2466 ki->p_sid = p->p_session->s_sid; 2467 ki->p__pgid = p->p_pgrp->pg_id; 2468 if (p->p_session->s_ttyvp) 2469 ki->p_eflag |= EPROC_CTTY; 2470 if (SESS_LEADER(p)) 2471 ki->p_eflag |= EPROC_SLEADER; 2472 strncpy(ki->p_login, p->p_session->s_login, 2473 uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login)); 2474 ki->p_jobc = p->p_pgrp->pg_jobc; 2475 if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) { 2476 ki->p_tdev = tp->t_dev; 2477 ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2478 COND_SET_VALUE(ki->p_tsess, PTRTOUINT64(tp->t_session), 2479 allowaddr); 2480 } else { 2481 ki->p_tdev = (int32_t)NODEV; 2482 } 2483 } 2484 2485 if (!P_ZOMBIE(p) && !zombie) { 2486 ki->p_uvalid = 1; 2487 ki->p_ustart_sec = p->p_stats->p_start.tv_sec; 2488 ki->p_ustart_usec = p->p_stats->p_start.tv_usec; 2489 2490 calcru(p, &ut, &st, NULL, &rt); 2491 ki->p_rtime_sec = rt.tv_sec; 2492 ki->p_rtime_usec = rt.tv_usec; 2493 ki->p_uutime_sec = ut.tv_sec; 2494 ki->p_uutime_usec = ut.tv_usec; 2495 ki->p_ustime_sec = st.tv_sec; 2496 ki->p_ustime_usec = st.tv_usec; 2497 2498 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 2499 ki->p_uru_nvcsw = 0; 2500 ki->p_uru_nivcsw = 0; 2501 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 2502 ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw); 2503 ki->p_uru_nivcsw += l2->l_nivcsw; 2504 ruadd(&ru, &l2->l_ru); 2505 } 2506 ki->p_uru_maxrss = ru.ru_maxrss; 2507 ki->p_uru_ixrss = ru.ru_ixrss; 2508 ki->p_uru_idrss = ru.ru_idrss; 2509 ki->p_uru_isrss = ru.ru_isrss; 2510 ki->p_uru_minflt = ru.ru_minflt; 2511 ki->p_uru_majflt = ru.ru_majflt; 2512 ki->p_uru_nswap = ru.ru_nswap; 2513 ki->p_uru_inblock = ru.ru_inblock; 2514 ki->p_uru_oublock = ru.ru_oublock; 2515 ki->p_uru_msgsnd = ru.ru_msgsnd; 2516 ki->p_uru_msgrcv = ru.ru_msgrcv; 2517 ki->p_uru_nsignals = ru.ru_nsignals; 2518 2519 timeradd(&p->p_stats->p_cru.ru_utime, 2520 &p->p_stats->p_cru.ru_stime, &ut); 2521 ki->p_uctime_sec = ut.tv_sec; 2522 ki->p_uctime_usec = ut.tv_usec; 2523 } 2524 } 2525 2526 2527 int 2528 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid) 2529 { 2530 int error; 2531 2532 mutex_enter(proc_lock); 2533 if (pid == -1) 2534 *p = l->l_proc; 2535 else 2536 *p = proc_find(pid); 2537 2538 if (*p == NULL) { 2539 if (pid != -1) 2540 mutex_exit(proc_lock); 2541 return ESRCH; 2542 } 2543 if (pid != -1) 2544 mutex_enter((*p)->p_lock); 2545 mutex_exit(proc_lock); 2546 2547 error = kauth_authorize_process(l->l_cred, 2548 KAUTH_PROCESS_CANSEE, *p, 2549 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 2550 if (error) { 2551 if (pid != -1) 2552 mutex_exit((*p)->p_lock); 2553 } 2554 return error; 2555 } 2556 2557 static int 2558 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2559 { 2560 int error; 2561 struct proc *p; 2562 2563 if ((error = proc_find_locked(l, &p, pid)) != 0) 2564 return error; 2565 2566 if (p->p_path == NULL) { 2567 if (pid != -1) 2568 mutex_exit(p->p_lock); 2569 return ENOENT; 2570 } 2571 2572 size_t len = strlen(p->p_path) + 1; 2573 if (oldp != NULL) { 2574 size_t copylen = uimin(len, *oldlenp); 2575 error = sysctl_copyout(l, p->p_path, oldp, copylen); 2576 if (error == 0 && *oldlenp < len) 2577 error = ENOSPC; 2578 } 2579 *oldlenp = len; 2580 if (pid != -1) 2581 mutex_exit(p->p_lock); 2582 return error; 2583 } 2584 2585 static int 2586 fill_cwd(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2587 { 2588 int error; 2589 struct proc *p; 2590 char *path; 2591 char *bp, *bend; 2592 const struct cwdinfo *cwdi; 2593 struct vnode *vp; 2594 size_t len, lenused; 2595 2596 if ((error = proc_find_locked(l, &p, pid)) != 0) 2597 return error; 2598 2599 len = MAXPATHLEN * 4; 2600 2601 path = kmem_alloc(len, KM_SLEEP); 2602 2603 bp = &path[len]; 2604 bend = bp; 2605 *(--bp) = '\0'; 2606 2607 cwdi = cwdlock(p); 2608 vp = cwdi->cwdi_cdir; 2609 vref(vp); 2610 cwdunlock(p); 2611 error = getcwd_common(vp, NULL, &bp, path, len/2, 0, l); 2612 vrele(vp); 2613 2614 if (error) 2615 goto out; 2616 2617 lenused = bend - bp; 2618 2619 if (oldp != NULL) { 2620 size_t copylen = uimin(lenused, *oldlenp); 2621 error = sysctl_copyout(l, bp, oldp, copylen); 2622 if (error == 0 && *oldlenp < lenused) 2623 error = ENOSPC; 2624 } 2625 *oldlenp = lenused; 2626 out: 2627 if (pid != -1) 2628 mutex_exit(p->p_lock); 2629 kmem_free(path, len); 2630 return error; 2631 } 2632 2633 int 2634 proc_getauxv(struct proc *p, void **buf, size_t *len) 2635 { 2636 struct ps_strings pss; 2637 int error; 2638 void *uauxv, *kauxv; 2639 size_t size; 2640 2641 if ((error = copyin_psstrings(p, &pss)) != 0) 2642 return error; 2643 if (pss.ps_envstr == NULL) 2644 return EIO; 2645 2646 size = p->p_execsw->es_arglen; 2647 if (size == 0) 2648 return EIO; 2649 2650 size_t ptrsz = PROC_PTRSZ(p); 2651 uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz); 2652 2653 kauxv = kmem_alloc(size, KM_SLEEP); 2654 2655 error = copyin_proc(p, uauxv, kauxv, size); 2656 if (error) { 2657 kmem_free(kauxv, size); 2658 return error; 2659 } 2660 2661 *buf = kauxv; 2662 *len = size; 2663 2664 return 0; 2665 } 2666 2667 2668 static int 2669 sysctl_security_expose_address(SYSCTLFN_ARGS) 2670 { 2671 int expose_address, error; 2672 struct sysctlnode node; 2673 2674 node = *rnode; 2675 node.sysctl_data = &expose_address; 2676 expose_address = *(int *)rnode->sysctl_data; 2677 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2678 if (error || newp == NULL) 2679 return error; 2680 2681 if (kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_KERNADDR, 2682 0, NULL, NULL, NULL)) 2683 return EPERM; 2684 2685 switch (expose_address) { 2686 case 0: 2687 case 1: 2688 case 2: 2689 break; 2690 default: 2691 return EINVAL; 2692 } 2693 2694 *(int *)rnode->sysctl_data = expose_address; 2695 2696 return 0; 2697 } 2698 2699 bool 2700 get_expose_address(struct proc *p) 2701 { 2702 /* allow only if sysctl variable is set or privileged */ 2703 return kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE, 2704 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL) == 0; 2705 } 2706
Indexes created Tue Sep 23 14:10:03 GMT 2025