1 /* $NetBSD: kern_proc.c,v 1.260 2020/09/05 16:30:12 riastradh 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 1.260 2020/09/05 16:30:12 riastradh 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/futex.h> 110 #include <sys/pserialize.h> 111 112 #include <uvm/uvm_extern.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 static pserialize_t proc_psz; 123 124 /* 125 * pid to lwp/proc lookup is done by indexing the pid_table array. 126 * Since pid numbers are only allocated when an empty slot 127 * has been found, there is no need to search any lists ever. 128 * (an orphaned pgrp will lock the slot, a session will lock 129 * the pgrp with the same number.) 130 * If the table is too small it is reallocated with twice the 131 * previous size and the entries 'unzipped' into the two halves. 132 * A linked list of free entries is passed through the pt_lwp 133 * field of 'free' items - set odd to be an invalid ptr. Two 134 * additional bits are also used to indicate if the slot is 135 * currently occupied by a proc or lwp, and if the PID is 136 * hidden from certain kinds of lookups. We thus require a 137 * minimum alignment for proc and lwp structures (LWPs are 138 * at least 32-byte aligned). 139 */ 140 141 struct pid_table { 142 uintptr_t pt_slot; 143 struct pgrp *pt_pgrp; 144 pid_t pt_pid; 145 }; 146 147 #define PT_F_FREE ((uintptr_t)__BIT(0)) 148 #define PT_F_LWP 0 /* pseudo-flag */ 149 #define PT_F_PROC ((uintptr_t)__BIT(1)) 150 151 #define PT_F_TYPEBITS (PT_F_FREE|PT_F_PROC) 152 #define PT_F_ALLBITS (PT_F_FREE|PT_F_PROC) 153 154 #define PT_VALID(s) (((s) & PT_F_FREE) == 0) 155 #define PT_RESERVED(s) ((s) == 0) 156 #define PT_NEXT(s) ((u_int)(s) >> 1) 157 #define PT_SET_FREE(pid) (((pid) << 1) | PT_F_FREE) 158 #define PT_SET_LWP(l) ((uintptr_t)(l)) 159 #define PT_SET_PROC(p) (((uintptr_t)(p)) | PT_F_PROC) 160 #define PT_SET_RESERVED 0 161 #define PT_GET_LWP(s) ((struct lwp *)((s) & ~PT_F_ALLBITS)) 162 #define PT_GET_PROC(s) ((struct proc *)((s) & ~PT_F_ALLBITS)) 163 #define PT_GET_TYPE(s) ((s) & PT_F_TYPEBITS) 164 #define PT_IS_LWP(s) (PT_GET_TYPE(s) == PT_F_LWP && (s) != 0) 165 #define PT_IS_PROC(s) (PT_GET_TYPE(s) == PT_F_PROC) 166 167 #define MIN_PROC_ALIGNMENT (PT_F_ALLBITS + 1) 168 169 /* 170 * Table of process IDs (PIDs). 171 */ 172 static struct pid_table *pid_table __read_mostly; 173 174 #define INITIAL_PID_TABLE_SIZE (1 << 5) 175 176 /* Table mask, threshold for growing and number of allocated PIDs. */ 177 static u_int pid_tbl_mask __read_mostly; 178 static u_int pid_alloc_lim __read_mostly; 179 static u_int pid_alloc_cnt __cacheline_aligned; 180 181 /* Next free, last free and maximum PIDs. */ 182 static u_int next_free_pt __cacheline_aligned; 183 static u_int last_free_pt __cacheline_aligned; 184 static pid_t pid_max __read_mostly; 185 186 /* Components of the first process -- never freed. */ 187 188 extern struct emul emul_netbsd; /* defined in kern_exec.c */ 189 190 struct session session0 = { 191 .s_count = 1, 192 .s_sid = 0, 193 }; 194 struct pgrp pgrp0 = { 195 .pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members), 196 .pg_session = &session0, 197 }; 198 filedesc_t filedesc0; 199 struct cwdinfo cwdi0 = { 200 .cwdi_cmask = CMASK, 201 .cwdi_refcnt = 1, 202 }; 203 struct plimit limit0; 204 struct pstats pstat0; 205 struct vmspace vmspace0; 206 struct sigacts sigacts0; 207 struct proc proc0 = { 208 .p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps), 209 .p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters), 210 .p_nlwps = 1, 211 .p_nrlwps = 1, 212 .p_pgrp = &pgrp0, 213 .p_comm = "system", 214 /* 215 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8) 216 * when they exit. init(8) can easily wait them out for us. 217 */ 218 .p_flag = PK_SYSTEM | PK_NOCLDWAIT, 219 .p_stat = SACTIVE, 220 .p_nice = NZERO, 221 .p_emul = &emul_netbsd, 222 .p_cwdi = &cwdi0, 223 .p_limit = &limit0, 224 .p_fd = &filedesc0, 225 .p_vmspace = &vmspace0, 226 .p_stats = &pstat0, 227 .p_sigacts = &sigacts0, 228 #ifdef PROC0_MD_INITIALIZERS 229 PROC0_MD_INITIALIZERS 230 #endif 231 }; 232 kauth_cred_t cred0; 233 234 static const int nofile = NOFILE; 235 static const int maxuprc = MAXUPRC; 236 237 static int sysctl_doeproc(SYSCTLFN_PROTO); 238 static int sysctl_kern_proc_args(SYSCTLFN_PROTO); 239 static int sysctl_security_expose_address(SYSCTLFN_PROTO); 240 241 #ifdef KASLR 242 static int kern_expose_address = 0; 243 #else 244 static int kern_expose_address = 1; 245 #endif 246 /* 247 * The process list descriptors, used during pid allocation and 248 * by sysctl. No locking on this data structure is needed since 249 * it is completely static. 250 */ 251 const struct proclist_desc proclists[] = { 252 { &allproc }, 253 { &zombproc }, 254 { NULL }, 255 }; 256 257 static struct pgrp * pg_remove(pid_t); 258 static void pg_delete(pid_t); 259 static void orphanpg(struct pgrp *); 260 261 static specificdata_domain_t proc_specificdata_domain; 262 263 static pool_cache_t proc_cache; 264 265 static kauth_listener_t proc_listener; 266 267 static void fill_proc(const struct proc *, struct proc *, bool); 268 static int fill_pathname(struct lwp *, pid_t, void *, size_t *); 269 static int fill_cwd(struct lwp *, pid_t, void *, size_t *); 270 271 static int 272 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 273 void *arg0, void *arg1, void *arg2, void *arg3) 274 { 275 struct proc *p; 276 int result; 277 278 result = KAUTH_RESULT_DEFER; 279 p = arg0; 280 281 switch (action) { 282 case KAUTH_PROCESS_CANSEE: { 283 enum kauth_process_req req; 284 285 req = (enum kauth_process_req)(uintptr_t)arg1; 286 287 switch (req) { 288 case KAUTH_REQ_PROCESS_CANSEE_ARGS: 289 case KAUTH_REQ_PROCESS_CANSEE_ENTRY: 290 case KAUTH_REQ_PROCESS_CANSEE_OPENFILES: 291 case KAUTH_REQ_PROCESS_CANSEE_EPROC: 292 result = KAUTH_RESULT_ALLOW; 293 break; 294 295 case KAUTH_REQ_PROCESS_CANSEE_ENV: 296 if (kauth_cred_getuid(cred) != 297 kauth_cred_getuid(p->p_cred) || 298 kauth_cred_getuid(cred) != 299 kauth_cred_getsvuid(p->p_cred)) 300 break; 301 302 result = KAUTH_RESULT_ALLOW; 303 304 break; 305 306 case KAUTH_REQ_PROCESS_CANSEE_KPTR: 307 if (!kern_expose_address) 308 break; 309 310 if (kern_expose_address == 1 && !(p->p_flag & PK_KMEM)) 311 break; 312 313 result = KAUTH_RESULT_ALLOW; 314 315 break; 316 317 default: 318 break; 319 } 320 321 break; 322 } 323 324 case KAUTH_PROCESS_FORK: { 325 int lnprocs = (int)(unsigned long)arg2; 326 327 /* 328 * Don't allow a nonprivileged user to use the last few 329 * processes. The variable lnprocs is the current number of 330 * processes, maxproc is the limit. 331 */ 332 if (__predict_false((lnprocs >= maxproc - 5))) 333 break; 334 335 result = KAUTH_RESULT_ALLOW; 336 337 break; 338 } 339 340 case KAUTH_PROCESS_CORENAME: 341 case KAUTH_PROCESS_STOPFLAG: 342 if (proc_uidmatch(cred, p->p_cred) == 0) 343 result = KAUTH_RESULT_ALLOW; 344 345 break; 346 347 default: 348 break; 349 } 350 351 return result; 352 } 353 354 static int 355 proc_ctor(void *arg __unused, void *obj, int flags __unused) 356 { 357 memset(obj, 0, sizeof(struct proc)); 358 return 0; 359 } 360 361 static pid_t proc_alloc_pid_slot(struct proc *, uintptr_t); 362 363 /* 364 * Initialize global process hashing structures. 365 */ 366 void 367 procinit(void) 368 { 369 const struct proclist_desc *pd; 370 u_int i; 371 #define LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1)) 372 373 for (pd = proclists; pd->pd_list != NULL; pd++) 374 LIST_INIT(pd->pd_list); 375 376 mutex_init(&proc_lock, MUTEX_DEFAULT, IPL_NONE); 377 378 proc_psz = pserialize_create(); 379 380 pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE 381 * sizeof(struct pid_table), KM_SLEEP); 382 pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1; 383 pid_max = PID_MAX; 384 385 /* Set free list running through table... 386 Preset 'use count' above PID_MAX so we allocate pid 1 next. */ 387 for (i = 0; i <= pid_tbl_mask; i++) { 388 pid_table[i].pt_slot = PT_SET_FREE(LINK_EMPTY + i + 1); 389 pid_table[i].pt_pgrp = 0; 390 pid_table[i].pt_pid = 0; 391 } 392 /* slot 0 is just grabbed */ 393 next_free_pt = 1; 394 /* Need to fix last entry. */ 395 last_free_pt = pid_tbl_mask; 396 pid_table[last_free_pt].pt_slot = PT_SET_FREE(LINK_EMPTY); 397 /* point at which we grow table - to avoid reusing pids too often */ 398 pid_alloc_lim = pid_tbl_mask - 1; 399 #undef LINK_EMPTY 400 401 /* Reserve PID 1 for init(8). */ /* XXX slightly gross */ 402 mutex_enter(&proc_lock); 403 if (proc_alloc_pid_slot(&proc0, PT_SET_RESERVED) != 1) 404 panic("failed to reserve PID 1 for init(8)"); 405 mutex_exit(&proc_lock); 406 407 proc_specificdata_domain = specificdata_domain_create(); 408 KASSERT(proc_specificdata_domain != NULL); 409 410 size_t proc_alignment = coherency_unit; 411 if (proc_alignment < MIN_PROC_ALIGNMENT) 412 proc_alignment = MIN_PROC_ALIGNMENT; 413 414 proc_cache = pool_cache_init(sizeof(struct proc), proc_alignment, 0, 0, 415 "procpl", NULL, IPL_NONE, proc_ctor, NULL, NULL); 416 417 proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 418 proc_listener_cb, NULL); 419 } 420 421 void 422 procinit_sysctl(void) 423 { 424 static struct sysctllog *clog; 425 426 sysctl_createv(&clog, 0, NULL, NULL, 427 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 428 CTLTYPE_INT, "expose_address", 429 SYSCTL_DESCR("Enable exposing kernel addresses"), 430 sysctl_security_expose_address, 0, 431 &kern_expose_address, 0, CTL_KERN, CTL_CREATE, CTL_EOL); 432 sysctl_createv(&clog, 0, NULL, NULL, 433 CTLFLAG_PERMANENT, 434 CTLTYPE_NODE, "proc", 435 SYSCTL_DESCR("System-wide process information"), 436 sysctl_doeproc, 0, NULL, 0, 437 CTL_KERN, KERN_PROC, CTL_EOL); 438 sysctl_createv(&clog, 0, NULL, NULL, 439 CTLFLAG_PERMANENT, 440 CTLTYPE_NODE, "proc2", 441 SYSCTL_DESCR("Machine-independent process information"), 442 sysctl_doeproc, 0, NULL, 0, 443 CTL_KERN, KERN_PROC2, CTL_EOL); 444 sysctl_createv(&clog, 0, NULL, NULL, 445 CTLFLAG_PERMANENT, 446 CTLTYPE_NODE, "proc_args", 447 SYSCTL_DESCR("Process argument information"), 448 sysctl_kern_proc_args, 0, NULL, 0, 449 CTL_KERN, KERN_PROC_ARGS, CTL_EOL); 450 451 /* 452 "nodes" under these: 453 454 KERN_PROC_ALL 455 KERN_PROC_PID pid 456 KERN_PROC_PGRP pgrp 457 KERN_PROC_SESSION sess 458 KERN_PROC_TTY tty 459 KERN_PROC_UID uid 460 KERN_PROC_RUID uid 461 KERN_PROC_GID gid 462 KERN_PROC_RGID gid 463 464 all in all, probably not worth the effort... 465 */ 466 } 467 468 /* 469 * Initialize process 0. 470 */ 471 void 472 proc0_init(void) 473 { 474 struct proc *p; 475 struct pgrp *pg; 476 struct rlimit *rlim; 477 rlim_t lim; 478 int i; 479 480 p = &proc0; 481 pg = &pgrp0; 482 483 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 484 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 485 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 486 487 rw_init(&p->p_reflock); 488 cv_init(&p->p_waitcv, "wait"); 489 cv_init(&p->p_lwpcv, "lwpwait"); 490 491 LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling); 492 493 KASSERT(lwp0.l_lid == 0); 494 pid_table[lwp0.l_lid].pt_slot = PT_SET_LWP(&lwp0); 495 LIST_INSERT_HEAD(&allproc, p, p_list); 496 497 pid_table[lwp0.l_lid].pt_pgrp = pg; 498 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist); 499 500 #ifdef __HAVE_SYSCALL_INTERN 501 (*p->p_emul->e_syscall_intern)(p); 502 #endif 503 504 /* Create credentials. */ 505 cred0 = kauth_cred_alloc(); 506 p->p_cred = cred0; 507 508 /* Create the CWD info. */ 509 rw_init(&cwdi0.cwdi_lock); 510 511 /* Create the limits structures. */ 512 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE); 513 514 rlim = limit0.pl_rlimit; 515 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) { 516 rlim[i].rlim_cur = RLIM_INFINITY; 517 rlim[i].rlim_max = RLIM_INFINITY; 518 } 519 520 rlim[RLIMIT_NOFILE].rlim_max = maxfiles; 521 rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile; 522 523 rlim[RLIMIT_NPROC].rlim_max = maxproc; 524 rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc; 525 526 lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvm_availmem(false))); 527 rlim[RLIMIT_RSS].rlim_max = lim; 528 rlim[RLIMIT_MEMLOCK].rlim_max = lim; 529 rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 530 531 rlim[RLIMIT_NTHR].rlim_max = maxlwp; 532 rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc; 533 534 /* Note that default core name has zero length. */ 535 limit0.pl_corename = defcorename; 536 limit0.pl_cnlen = 0; 537 limit0.pl_refcnt = 1; 538 limit0.pl_writeable = false; 539 limit0.pl_sv_limit = NULL; 540 541 /* Configure virtual memory system, set vm rlimits. */ 542 uvm_init_limits(p); 543 544 /* Initialize file descriptor table for proc0. */ 545 fd_init(&filedesc0); 546 547 /* 548 * Initialize proc0's vmspace, which uses the kernel pmap. 549 * All kernel processes (which never have user space mappings) 550 * share proc0's vmspace, and thus, the kernel pmap. 551 */ 552 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 553 trunc_page(VM_MAXUSER_ADDRESS), 554 #ifdef __USE_TOPDOWN_VM 555 true 556 #else 557 false 558 #endif 559 ); 560 561 /* Initialize signal state for proc0. XXX IPL_SCHED */ 562 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED); 563 siginit(p); 564 565 proc_initspecific(p); 566 kdtrace_proc_ctor(NULL, p); 567 } 568 569 /* 570 * Session reference counting. 571 */ 572 573 void 574 proc_sesshold(struct session *ss) 575 { 576 577 KASSERT(mutex_owned(&proc_lock)); 578 ss->s_count++; 579 } 580 581 void 582 proc_sessrele(struct session *ss) 583 { 584 struct pgrp *pg; 585 586 KASSERT(mutex_owned(&proc_lock)); 587 KASSERT(ss->s_count > 0); 588 589 /* 590 * We keep the pgrp with the same id as the session in order to 591 * stop a process being given the same pid. Since the pgrp holds 592 * a reference to the session, it must be a 'zombie' pgrp by now. 593 */ 594 if (--ss->s_count == 0) { 595 pg = pg_remove(ss->s_sid); 596 } else { 597 pg = NULL; 598 ss = NULL; 599 } 600 601 mutex_exit(&proc_lock); 602 603 if (pg) 604 kmem_free(pg, sizeof(struct pgrp)); 605 if (ss) 606 kmem_free(ss, sizeof(struct session)); 607 } 608 609 /* 610 * Check that the specified process group is in the session of the 611 * specified process. 612 * Treats -ve ids as process ids. 613 * Used to validate TIOCSPGRP requests. 614 */ 615 int 616 pgid_in_session(struct proc *p, pid_t pg_id) 617 { 618 struct pgrp *pgrp; 619 struct session *session; 620 int error; 621 622 mutex_enter(&proc_lock); 623 if (pg_id < 0) { 624 struct proc *p1 = proc_find(-pg_id); 625 if (p1 == NULL) { 626 error = EINVAL; 627 goto fail; 628 } 629 pgrp = p1->p_pgrp; 630 } else { 631 pgrp = pgrp_find(pg_id); 632 if (pgrp == NULL) { 633 error = EINVAL; 634 goto fail; 635 } 636 } 637 session = pgrp->pg_session; 638 error = (session != p->p_pgrp->pg_session) ? EPERM : 0; 639 fail: 640 mutex_exit(&proc_lock); 641 return error; 642 } 643 644 /* 645 * p_inferior: is p an inferior of q? 646 */ 647 static inline bool 648 p_inferior(struct proc *p, struct proc *q) 649 { 650 651 KASSERT(mutex_owned(&proc_lock)); 652 653 for (; p != q; p = p->p_pptr) 654 if (p->p_pid == 0) 655 return false; 656 return true; 657 } 658 659 /* 660 * proc_find_lwp: locate an lwp in said proc by the ID. 661 * 662 * => Must be called with p::p_lock held. 663 * => LSIDL lwps are not returned because they are only partially 664 * constructed while occupying the slot. 665 * => Callers need to be careful about lwp::l_stat of the returned 666 * lwp. 667 */ 668 struct lwp * 669 proc_find_lwp(proc_t *p, pid_t pid) 670 { 671 struct pid_table *pt; 672 struct lwp *l = NULL; 673 uintptr_t slot; 674 int s; 675 676 KASSERT(mutex_owned(p->p_lock)); 677 678 /* 679 * Look in the pid_table. This is done unlocked inside a pserialize 680 * read section covering pid_table's memory allocation only, so take 681 * care to read the slot atomically and only once. This issues a 682 * memory barrier for dependent loads on alpha. 683 */ 684 s = pserialize_read_enter(); 685 pt = &pid_table[pid & pid_tbl_mask]; 686 slot = atomic_load_consume(&pt->pt_slot); 687 if (__predict_false(!PT_IS_LWP(slot))) { 688 pserialize_read_exit(s); 689 return NULL; 690 } 691 692 /* 693 * Check to see if the LWP is from the correct process. We won't 694 * see entries in pid_table from a prior process that also used "p", 695 * by virtue of the fact that allocating "p" means all prior updates 696 * to dependant data structures are visible to this thread. 697 */ 698 l = PT_GET_LWP(slot); 699 if (__predict_false(atomic_load_relaxed(&l->l_proc) != p)) { 700 pserialize_read_exit(s); 701 return NULL; 702 } 703 704 /* 705 * We now know that p->p_lock holds this LWP stable. 706 * 707 * If the status is not LSIDL, it means the LWP is intended to be 708 * findable by LID and l_lid cannot change behind us. 709 * 710 * No need to acquire the LWP's lock to check for LSIDL, as 711 * p->p_lock must be held to transition in and out of LSIDL. 712 * Any other observed state of is no particular interest. 713 */ 714 pserialize_read_exit(s); 715 return l->l_stat != LSIDL && l->l_lid == pid ? l : NULL; 716 } 717 718 /* 719 * proc_find_lwp_unlocked: locate an lwp in said proc by the ID. 720 * 721 * => Called in a pserialize read section with no locks held. 722 * => LSIDL lwps are not returned because they are only partially 723 * constructed while occupying the slot. 724 * => Callers need to be careful about lwp::l_stat of the returned 725 * lwp. 726 * => If an LWP is found, it's returned locked. 727 */ 728 struct lwp * 729 proc_find_lwp_unlocked(proc_t *p, pid_t pid) 730 { 731 struct pid_table *pt; 732 struct lwp *l = NULL; 733 uintptr_t slot; 734 735 KASSERT(pserialize_in_read_section()); 736 737 /* 738 * Look in the pid_table. This is done unlocked inside a pserialize 739 * read section covering pid_table's memory allocation only, so take 740 * care to read the slot atomically and only once. This issues a 741 * memory barrier for dependent loads on alpha. 742 */ 743 pt = &pid_table[pid & pid_tbl_mask]; 744 slot = atomic_load_consume(&pt->pt_slot); 745 if (__predict_false(!PT_IS_LWP(slot))) { 746 return NULL; 747 } 748 749 /* 750 * Lock the LWP we found to get it stable. If it's embryonic or 751 * reaped (LSIDL) then none of the other fields can safely be 752 * checked. 753 */ 754 l = PT_GET_LWP(slot); 755 lwp_lock(l); 756 if (__predict_false(l->l_stat == LSIDL)) { 757 lwp_unlock(l); 758 return NULL; 759 } 760 761 /* 762 * l_proc and l_lid are now known stable because the LWP is not 763 * LSIDL, so check those fields too to make sure we found the 764 * right thing. 765 */ 766 if (__predict_false(l->l_proc != p || l->l_lid != pid)) { 767 lwp_unlock(l); 768 return NULL; 769 } 770 771 /* Everything checks out, return it locked. */ 772 return l; 773 } 774 775 /* 776 * proc_find_lwp_acquire_proc: locate an lwp and acquire a lock 777 * on its containing proc. 778 * 779 * => Similar to proc_find_lwp(), but does not require you to have 780 * the proc a priori. 781 * => Also returns proc * to caller, with p::p_lock held. 782 * => Same caveats apply. 783 */ 784 struct lwp * 785 proc_find_lwp_acquire_proc(pid_t pid, struct proc **pp) 786 { 787 struct pid_table *pt; 788 struct proc *p = NULL; 789 struct lwp *l = NULL; 790 uintptr_t slot; 791 792 KASSERT(pp != NULL); 793 mutex_enter(&proc_lock); 794 pt = &pid_table[pid & pid_tbl_mask]; 795 796 slot = pt->pt_slot; 797 if (__predict_true(PT_IS_LWP(slot) && pt->pt_pid == pid)) { 798 l = PT_GET_LWP(slot); 799 p = l->l_proc; 800 mutex_enter(p->p_lock); 801 if (__predict_false(l->l_stat == LSIDL)) { 802 mutex_exit(p->p_lock); 803 l = NULL; 804 p = NULL; 805 } 806 } 807 mutex_exit(&proc_lock); 808 809 KASSERT(p == NULL || mutex_owned(p->p_lock)); 810 *pp = p; 811 return l; 812 } 813 814 /* 815 * proc_find_raw_pid_table_locked: locate a process by the ID. 816 * 817 * => Must be called with proc_lock held. 818 */ 819 static proc_t * 820 proc_find_raw_pid_table_locked(pid_t pid, bool any_lwpid) 821 { 822 struct pid_table *pt; 823 proc_t *p = NULL; 824 uintptr_t slot; 825 826 /* No - used by DDB. KASSERT(mutex_owned(&proc_lock)); */ 827 pt = &pid_table[pid & pid_tbl_mask]; 828 829 slot = pt->pt_slot; 830 if (__predict_true(PT_IS_LWP(slot) && pt->pt_pid == pid)) { 831 /* 832 * When looking up processes, require a direct match 833 * on the PID assigned to the proc, not just one of 834 * its LWPs. 835 * 836 * N.B. We require lwp::l_proc of LSIDL LWPs to be 837 * valid here. 838 */ 839 p = PT_GET_LWP(slot)->l_proc; 840 if (__predict_false(p->p_pid != pid && !any_lwpid)) 841 p = NULL; 842 } else if (PT_IS_PROC(slot) && pt->pt_pid == pid) { 843 p = PT_GET_PROC(slot); 844 } 845 return p; 846 } 847 848 proc_t * 849 proc_find_raw(pid_t pid) 850 { 851 852 return proc_find_raw_pid_table_locked(pid, false); 853 } 854 855 static proc_t * 856 proc_find_internal(pid_t pid, bool any_lwpid) 857 { 858 proc_t *p; 859 860 KASSERT(mutex_owned(&proc_lock)); 861 862 p = proc_find_raw_pid_table_locked(pid, any_lwpid); 863 if (__predict_false(p == NULL)) { 864 return NULL; 865 } 866 867 /* 868 * Only allow live processes to be found by PID. 869 * XXX: p_stat might change, since proc unlocked. 870 */ 871 if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) { 872 return p; 873 } 874 return NULL; 875 } 876 877 proc_t * 878 proc_find(pid_t pid) 879 { 880 return proc_find_internal(pid, false); 881 } 882 883 proc_t * 884 proc_find_lwpid(pid_t pid) 885 { 886 return proc_find_internal(pid, true); 887 } 888 889 /* 890 * pgrp_find: locate a process group by the ID. 891 * 892 * => Must be called with proc_lock held. 893 */ 894 struct pgrp * 895 pgrp_find(pid_t pgid) 896 { 897 struct pgrp *pg; 898 899 KASSERT(mutex_owned(&proc_lock)); 900 901 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp; 902 903 /* 904 * Cannot look up a process group that only exists because the 905 * session has not died yet (traditional). 906 */ 907 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) { 908 return NULL; 909 } 910 return pg; 911 } 912 913 static void 914 expand_pid_table(void) 915 { 916 size_t pt_size, tsz; 917 struct pid_table *n_pt, *new_pt; 918 uintptr_t slot; 919 struct pgrp *pgrp; 920 pid_t pid, rpid; 921 u_int i; 922 uint new_pt_mask; 923 924 KASSERT(mutex_owned(&proc_lock)); 925 926 /* Unlock the pid_table briefly to allocate memory. */ 927 pt_size = pid_tbl_mask + 1; 928 mutex_exit(&proc_lock); 929 930 tsz = pt_size * 2 * sizeof(struct pid_table); 931 new_pt = kmem_alloc(tsz, KM_SLEEP); 932 new_pt_mask = pt_size * 2 - 1; 933 934 /* XXX For now. The pratical limit is much lower anyway. */ 935 KASSERT(new_pt_mask <= FUTEX_TID_MASK); 936 937 mutex_enter(&proc_lock); 938 if (pt_size != pid_tbl_mask + 1) { 939 /* Another process beat us to it... */ 940 mutex_exit(&proc_lock); 941 kmem_free(new_pt, tsz); 942 goto out; 943 } 944 945 /* 946 * Copy entries from old table into new one. 947 * If 'pid' is 'odd' we need to place in the upper half, 948 * even pid's to the lower half. 949 * Free items stay in the low half so we don't have to 950 * fixup the reference to them. 951 * We stuff free items on the front of the freelist 952 * because we can't write to unmodified entries. 953 * Processing the table backwards maintains a semblance 954 * of issuing pid numbers that increase with time. 955 */ 956 i = pt_size - 1; 957 n_pt = new_pt + i; 958 for (; ; i--, n_pt--) { 959 slot = pid_table[i].pt_slot; 960 pgrp = pid_table[i].pt_pgrp; 961 if (!PT_VALID(slot)) { 962 /* Up 'use count' so that link is valid */ 963 pid = (PT_NEXT(slot) + pt_size) & ~pt_size; 964 rpid = 0; 965 slot = PT_SET_FREE(pid); 966 if (pgrp) 967 pid = pgrp->pg_id; 968 } else { 969 pid = pid_table[i].pt_pid; 970 rpid = pid; 971 } 972 973 /* Save entry in appropriate half of table */ 974 n_pt[pid & pt_size].pt_slot = slot; 975 n_pt[pid & pt_size].pt_pgrp = pgrp; 976 n_pt[pid & pt_size].pt_pid = rpid; 977 978 /* Put other piece on start of free list */ 979 pid = (pid ^ pt_size) & ~pid_tbl_mask; 980 n_pt[pid & pt_size].pt_slot = 981 PT_SET_FREE((pid & ~pt_size) | next_free_pt); 982 n_pt[pid & pt_size].pt_pgrp = 0; 983 n_pt[pid & pt_size].pt_pid = 0; 984 985 next_free_pt = i | (pid & pt_size); 986 if (i == 0) 987 break; 988 } 989 990 /* Save old table size and switch tables */ 991 tsz = pt_size * sizeof(struct pid_table); 992 n_pt = pid_table; 993 pid_table = new_pt; 994 pid_tbl_mask = new_pt_mask; 995 996 /* 997 * pid_max starts as PID_MAX (= 30000), once we have 16384 998 * allocated pids we need it to be larger! 999 */ 1000 if (pid_tbl_mask > PID_MAX) { 1001 pid_max = pid_tbl_mask * 2 + 1; 1002 pid_alloc_lim |= pid_alloc_lim << 1; 1003 } else 1004 pid_alloc_lim <<= 1; /* doubles number of free slots... */ 1005 1006 mutex_exit(&proc_lock); 1007 1008 /* 1009 * Make sure that unlocked access to the old pid_table is complete 1010 * and then free it. 1011 */ 1012 pserialize_perform(proc_psz); 1013 kmem_free(n_pt, tsz); 1014 1015 out: /* Return with proc_lock held again. */ 1016 mutex_enter(&proc_lock); 1017 } 1018 1019 struct proc * 1020 proc_alloc(void) 1021 { 1022 struct proc *p; 1023 1024 p = pool_cache_get(proc_cache, PR_WAITOK); 1025 p->p_stat = SIDL; /* protect against others */ 1026 proc_initspecific(p); 1027 kdtrace_proc_ctor(NULL, p); 1028 1029 /* 1030 * Allocate a placeholder in the pid_table. When we create the 1031 * first LWP for this process, it will take ownership of the 1032 * slot. 1033 */ 1034 if (__predict_false(proc_alloc_pid(p) == -1)) { 1035 /* Allocating the PID failed; unwind. */ 1036 proc_finispecific(p); 1037 proc_free_mem(p); 1038 p = NULL; 1039 } 1040 return p; 1041 } 1042 1043 /* 1044 * proc_alloc_pid_slot: allocate PID and record the occcupant so that 1045 * proc_find_raw() can find it by the PID. 1046 */ 1047 static pid_t __noinline 1048 proc_alloc_pid_slot(struct proc *p, uintptr_t slot) 1049 { 1050 struct pid_table *pt; 1051 pid_t pid; 1052 int nxt; 1053 1054 KASSERT(mutex_owned(&proc_lock)); 1055 1056 for (;;expand_pid_table()) { 1057 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) { 1058 /* ensure pids cycle through 2000+ values */ 1059 continue; 1060 } 1061 /* 1062 * The first user process *must* be given PID 1. 1063 * it has already been reserved for us. This 1064 * will be coming in from the proc_alloc() call 1065 * above, and the entry will be usurped later when 1066 * the first user LWP is created. 1067 * XXX this is slightly gross. 1068 */ 1069 if (__predict_false(PT_RESERVED(pid_table[1].pt_slot) && 1070 p != &proc0)) { 1071 KASSERT(PT_IS_PROC(slot)); 1072 pt = &pid_table[1]; 1073 pt->pt_slot = slot; 1074 return 1; 1075 } 1076 pt = &pid_table[next_free_pt]; 1077 #ifdef DIAGNOSTIC 1078 if (__predict_false(PT_VALID(pt->pt_slot) || pt->pt_pgrp)) 1079 panic("proc_alloc: slot busy"); 1080 #endif 1081 nxt = PT_NEXT(pt->pt_slot); 1082 if (nxt & pid_tbl_mask) 1083 break; 1084 /* Table full - expand (NB last entry not used....) */ 1085 } 1086 1087 /* pid is 'saved use count' + 'size' + entry */ 1088 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt; 1089 if ((uint)pid > (uint)pid_max) 1090 pid &= pid_tbl_mask; 1091 next_free_pt = nxt & pid_tbl_mask; 1092 1093 /* XXX For now. The pratical limit is much lower anyway. */ 1094 KASSERT(pid <= FUTEX_TID_MASK); 1095 1096 /* Grab table slot */ 1097 pt->pt_slot = slot; 1098 1099 KASSERT(pt->pt_pid == 0); 1100 pt->pt_pid = pid; 1101 pid_alloc_cnt++; 1102 1103 return pid; 1104 } 1105 1106 pid_t 1107 proc_alloc_pid(struct proc *p) 1108 { 1109 pid_t pid; 1110 1111 KASSERT((((uintptr_t)p) & PT_F_ALLBITS) == 0); 1112 KASSERT(p->p_stat == SIDL); 1113 1114 mutex_enter(&proc_lock); 1115 pid = proc_alloc_pid_slot(p, PT_SET_PROC(p)); 1116 if (pid != -1) 1117 p->p_pid = pid; 1118 mutex_exit(&proc_lock); 1119 1120 return pid; 1121 } 1122 1123 pid_t 1124 proc_alloc_lwpid(struct proc *p, struct lwp *l) 1125 { 1126 struct pid_table *pt; 1127 pid_t pid; 1128 1129 KASSERT((((uintptr_t)l) & PT_F_ALLBITS) == 0); 1130 KASSERT(l->l_proc == p); 1131 KASSERT(l->l_stat == LSIDL); 1132 1133 /* 1134 * For unlocked lookup in proc_find_lwp(), make sure l->l_proc 1135 * is globally visible before the LWP becomes visible via the 1136 * pid_table. 1137 */ 1138 #ifndef __HAVE_ATOMIC_AS_MEMBAR 1139 membar_producer(); 1140 #endif 1141 1142 /* 1143 * If the slot for p->p_pid currently points to the proc, 1144 * then we should usurp this ID for the LWP. This happens 1145 * at least once per process (for the first LWP), and can 1146 * happen again if the first LWP for a process exits and 1147 * before the process creates another. 1148 */ 1149 mutex_enter(&proc_lock); 1150 pid = p->p_pid; 1151 pt = &pid_table[pid & pid_tbl_mask]; 1152 KASSERT(pt->pt_pid == pid); 1153 if (PT_IS_PROC(pt->pt_slot)) { 1154 KASSERT(PT_GET_PROC(pt->pt_slot) == p); 1155 l->l_lid = pid; 1156 pt->pt_slot = PT_SET_LWP(l); 1157 } else { 1158 /* Need to allocate a new slot. */ 1159 pid = proc_alloc_pid_slot(p, PT_SET_LWP(l)); 1160 if (pid != -1) 1161 l->l_lid = pid; 1162 } 1163 mutex_exit(&proc_lock); 1164 1165 return pid; 1166 } 1167 1168 static void __noinline 1169 proc_free_pid_internal(pid_t pid, uintptr_t type __diagused) 1170 { 1171 struct pid_table *pt; 1172 1173 pt = &pid_table[pid & pid_tbl_mask]; 1174 1175 KASSERT(PT_GET_TYPE(pt->pt_slot) == type); 1176 KASSERT(pt->pt_pid == pid); 1177 1178 /* save pid use count in slot */ 1179 pt->pt_slot = PT_SET_FREE(pid & ~pid_tbl_mask); 1180 pt->pt_pid = 0; 1181 1182 if (pt->pt_pgrp == NULL) { 1183 /* link last freed entry onto ours */ 1184 pid &= pid_tbl_mask; 1185 pt = &pid_table[last_free_pt]; 1186 pt->pt_slot = PT_SET_FREE(PT_NEXT(pt->pt_slot) | pid); 1187 pt->pt_pid = 0; 1188 last_free_pt = pid; 1189 pid_alloc_cnt--; 1190 } 1191 } 1192 1193 /* 1194 * Free a process id - called from proc_free (in kern_exit.c) 1195 * 1196 * Called with the proc_lock held. 1197 */ 1198 void 1199 proc_free_pid(pid_t pid) 1200 { 1201 1202 KASSERT(mutex_owned(&proc_lock)); 1203 proc_free_pid_internal(pid, PT_F_PROC); 1204 } 1205 1206 /* 1207 * Free a process id used by an LWP. If this was the process's 1208 * first LWP, we convert the slot to point to the process; the 1209 * entry will get cleaned up later when the process finishes exiting. 1210 * 1211 * If not, then it's the same as proc_free_pid(). 1212 */ 1213 void 1214 proc_free_lwpid(struct proc *p, pid_t pid) 1215 { 1216 1217 KASSERT(mutex_owned(&proc_lock)); 1218 1219 if (__predict_true(p->p_pid == pid)) { 1220 struct pid_table *pt; 1221 1222 pt = &pid_table[pid & pid_tbl_mask]; 1223 1224 KASSERT(pt->pt_pid == pid); 1225 KASSERT(PT_IS_LWP(pt->pt_slot)); 1226 KASSERT(PT_GET_LWP(pt->pt_slot)->l_proc == p); 1227 1228 pt->pt_slot = PT_SET_PROC(p); 1229 return; 1230 } 1231 proc_free_pid_internal(pid, PT_F_LWP); 1232 } 1233 1234 void 1235 proc_free_mem(struct proc *p) 1236 { 1237 1238 kdtrace_proc_dtor(NULL, p); 1239 pool_cache_put(proc_cache, p); 1240 } 1241 1242 /* 1243 * proc_enterpgrp: move p to a new or existing process group (and session). 1244 * 1245 * If we are creating a new pgrp, the pgid should equal 1246 * the calling process' pid. 1247 * If is only valid to enter a process group that is in the session 1248 * of the process. 1249 * Also mksess should only be set if we are creating a process group 1250 * 1251 * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return. 1252 */ 1253 int 1254 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess) 1255 { 1256 struct pgrp *new_pgrp, *pgrp; 1257 struct session *sess; 1258 struct proc *p; 1259 int rval; 1260 pid_t pg_id = NO_PGID; 1261 1262 /* Allocate data areas we might need before doing any validity checks */ 1263 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL; 1264 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP); 1265 1266 mutex_enter(&proc_lock); 1267 rval = EPERM; /* most common error (to save typing) */ 1268 1269 /* Check pgrp exists or can be created */ 1270 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp; 1271 if (pgrp != NULL && pgrp->pg_id != pgid) 1272 goto done; 1273 1274 /* Can only set another process under restricted circumstances. */ 1275 if (pid != curp->p_pid) { 1276 /* Must exist and be one of our children... */ 1277 p = proc_find_internal(pid, false); 1278 if (p == NULL || !p_inferior(p, curp)) { 1279 rval = ESRCH; 1280 goto done; 1281 } 1282 /* ... in the same session... */ 1283 if (sess != NULL || p->p_session != curp->p_session) 1284 goto done; 1285 /* ... existing pgid must be in same session ... */ 1286 if (pgrp != NULL && pgrp->pg_session != p->p_session) 1287 goto done; 1288 /* ... and not done an exec. */ 1289 if (p->p_flag & PK_EXEC) { 1290 rval = EACCES; 1291 goto done; 1292 } 1293 } else { 1294 /* ... setsid() cannot re-enter a pgrp */ 1295 if (mksess && (curp->p_pgid == curp->p_pid || 1296 pgrp_find(curp->p_pid))) 1297 goto done; 1298 p = curp; 1299 } 1300 1301 /* Changing the process group/session of a session 1302 leader is definitely off limits. */ 1303 if (SESS_LEADER(p)) { 1304 if (sess == NULL && p->p_pgrp == pgrp) 1305 /* unless it's a definite noop */ 1306 rval = 0; 1307 goto done; 1308 } 1309 1310 /* Can only create a process group with id of process */ 1311 if (pgrp == NULL && pgid != pid) 1312 goto done; 1313 1314 /* Can only create a session if creating pgrp */ 1315 if (sess != NULL && pgrp != NULL) 1316 goto done; 1317 1318 /* Check we allocated memory for a pgrp... */ 1319 if (pgrp == NULL && new_pgrp == NULL) 1320 goto done; 1321 1322 /* Don't attach to 'zombie' pgrp */ 1323 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members)) 1324 goto done; 1325 1326 /* Expect to succeed now */ 1327 rval = 0; 1328 1329 if (pgrp == p->p_pgrp) 1330 /* nothing to do */ 1331 goto done; 1332 1333 /* Ok all setup, link up required structures */ 1334 1335 if (pgrp == NULL) { 1336 pgrp = new_pgrp; 1337 new_pgrp = NULL; 1338 if (sess != NULL) { 1339 sess->s_sid = p->p_pid; 1340 sess->s_leader = p; 1341 sess->s_count = 1; 1342 sess->s_ttyvp = NULL; 1343 sess->s_ttyp = NULL; 1344 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET; 1345 memcpy(sess->s_login, p->p_session->s_login, 1346 sizeof(sess->s_login)); 1347 p->p_lflag &= ~PL_CONTROLT; 1348 } else { 1349 sess = p->p_pgrp->pg_session; 1350 proc_sesshold(sess); 1351 } 1352 pgrp->pg_session = sess; 1353 sess = NULL; 1354 1355 pgrp->pg_id = pgid; 1356 LIST_INIT(&pgrp->pg_members); 1357 #ifdef DIAGNOSTIC 1358 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp)) 1359 panic("enterpgrp: pgrp table slot in use"); 1360 if (__predict_false(mksess && p != curp)) 1361 panic("enterpgrp: mksession and p != curproc"); 1362 #endif 1363 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp; 1364 pgrp->pg_jobc = 0; 1365 } 1366 1367 /* 1368 * Adjust eligibility of affected pgrps to participate in job control. 1369 * Increment eligibility counts before decrementing, otherwise we 1370 * could reach 0 spuriously during the first call. 1371 */ 1372 fixjobc(p, pgrp, 1); 1373 fixjobc(p, p->p_pgrp, 0); 1374 1375 /* Interlock with ttread(). */ 1376 mutex_spin_enter(&tty_lock); 1377 1378 /* Move process to requested group. */ 1379 LIST_REMOVE(p, p_pglist); 1380 if (LIST_EMPTY(&p->p_pgrp->pg_members)) 1381 /* defer delete until we've dumped the lock */ 1382 pg_id = p->p_pgrp->pg_id; 1383 p->p_pgrp = pgrp; 1384 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 1385 1386 /* Done with the swap; we can release the tty mutex. */ 1387 mutex_spin_exit(&tty_lock); 1388 1389 done: 1390 if (pg_id != NO_PGID) { 1391 /* Releases proc_lock. */ 1392 pg_delete(pg_id); 1393 } else { 1394 mutex_exit(&proc_lock); 1395 } 1396 if (sess != NULL) 1397 kmem_free(sess, sizeof(*sess)); 1398 if (new_pgrp != NULL) 1399 kmem_free(new_pgrp, sizeof(*new_pgrp)); 1400 #ifdef DEBUG_PGRP 1401 if (__predict_false(rval)) 1402 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n", 1403 pid, pgid, mksess, curp->p_pid, rval); 1404 #endif 1405 return rval; 1406 } 1407 1408 /* 1409 * proc_leavepgrp: remove a process from its process group. 1410 * => must be called with the proc_lock held, which will be released; 1411 */ 1412 void 1413 proc_leavepgrp(struct proc *p) 1414 { 1415 struct pgrp *pgrp; 1416 1417 KASSERT(mutex_owned(&proc_lock)); 1418 1419 /* Interlock with ttread() */ 1420 mutex_spin_enter(&tty_lock); 1421 pgrp = p->p_pgrp; 1422 LIST_REMOVE(p, p_pglist); 1423 p->p_pgrp = NULL; 1424 mutex_spin_exit(&tty_lock); 1425 1426 if (LIST_EMPTY(&pgrp->pg_members)) { 1427 /* Releases proc_lock. */ 1428 pg_delete(pgrp->pg_id); 1429 } else { 1430 mutex_exit(&proc_lock); 1431 } 1432 } 1433 1434 /* 1435 * pg_remove: remove a process group from the table. 1436 * => must be called with the proc_lock held; 1437 * => returns process group to free; 1438 */ 1439 static struct pgrp * 1440 pg_remove(pid_t pg_id) 1441 { 1442 struct pgrp *pgrp; 1443 struct pid_table *pt; 1444 1445 KASSERT(mutex_owned(&proc_lock)); 1446 1447 pt = &pid_table[pg_id & pid_tbl_mask]; 1448 pgrp = pt->pt_pgrp; 1449 1450 KASSERT(pgrp != NULL); 1451 KASSERT(pgrp->pg_id == pg_id); 1452 KASSERT(LIST_EMPTY(&pgrp->pg_members)); 1453 1454 pt->pt_pgrp = NULL; 1455 1456 if (!PT_VALID(pt->pt_slot)) { 1457 /* Orphaned pgrp, put slot onto free list. */ 1458 KASSERT((PT_NEXT(pt->pt_slot) & pid_tbl_mask) == 0); 1459 pg_id &= pid_tbl_mask; 1460 pt = &pid_table[last_free_pt]; 1461 pt->pt_slot = PT_SET_FREE(PT_NEXT(pt->pt_slot) | pg_id); 1462 KASSERT(pt->pt_pid == 0); 1463 last_free_pt = pg_id; 1464 pid_alloc_cnt--; 1465 } 1466 return pgrp; 1467 } 1468 1469 /* 1470 * pg_delete: delete and free a process group. 1471 * => must be called with the proc_lock held, which will be released. 1472 */ 1473 static void 1474 pg_delete(pid_t pg_id) 1475 { 1476 struct pgrp *pg; 1477 struct tty *ttyp; 1478 struct session *ss; 1479 1480 KASSERT(mutex_owned(&proc_lock)); 1481 1482 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp; 1483 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) { 1484 mutex_exit(&proc_lock); 1485 return; 1486 } 1487 1488 ss = pg->pg_session; 1489 1490 /* Remove reference (if any) from tty to this process group */ 1491 mutex_spin_enter(&tty_lock); 1492 ttyp = ss->s_ttyp; 1493 if (ttyp != NULL && ttyp->t_pgrp == pg) { 1494 ttyp->t_pgrp = NULL; 1495 KASSERT(ttyp->t_session == ss); 1496 } 1497 mutex_spin_exit(&tty_lock); 1498 1499 /* 1500 * The leading process group in a session is freed by proc_sessrele(), 1501 * if last reference. It will also release the locks. 1502 */ 1503 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL; 1504 proc_sessrele(ss); 1505 1506 if (pg != NULL) { 1507 /* Free it, if was not done above. */ 1508 kmem_free(pg, sizeof(struct pgrp)); 1509 } 1510 } 1511 1512 /* 1513 * Adjust pgrp jobc counters when specified process changes process group. 1514 * We count the number of processes in each process group that "qualify" 1515 * the group for terminal job control (those with a parent in a different 1516 * process group of the same session). If that count reaches zero, the 1517 * process group becomes orphaned. Check both the specified process' 1518 * process group and that of its children. 1519 * entering == 0 => p is leaving specified group. 1520 * entering == 1 => p is entering specified group. 1521 * 1522 * Call with proc_lock held. 1523 */ 1524 void 1525 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 1526 { 1527 struct pgrp *hispgrp; 1528 struct session *mysession = pgrp->pg_session; 1529 struct proc *child; 1530 1531 KASSERT(mutex_owned(&proc_lock)); 1532 1533 /* 1534 * Check p's parent to see whether p qualifies its own process 1535 * group; if so, adjust count for p's process group. 1536 */ 1537 hispgrp = p->p_pptr->p_pgrp; 1538 if (hispgrp != pgrp && hispgrp->pg_session == mysession) { 1539 if (entering) { 1540 pgrp->pg_jobc++; 1541 p->p_lflag &= ~PL_ORPHANPG; 1542 } else { 1543 KASSERT(pgrp->pg_jobc > 0); 1544 if (--pgrp->pg_jobc == 0) 1545 orphanpg(pgrp); 1546 } 1547 } 1548 1549 /* 1550 * Check this process' children to see whether they qualify 1551 * their process groups; if so, adjust counts for children's 1552 * process groups. 1553 */ 1554 LIST_FOREACH(child, &p->p_children, p_sibling) { 1555 hispgrp = child->p_pgrp; 1556 if (hispgrp != pgrp && hispgrp->pg_session == mysession && 1557 !P_ZOMBIE(child)) { 1558 if (entering) { 1559 child->p_lflag &= ~PL_ORPHANPG; 1560 hispgrp->pg_jobc++; 1561 } else { 1562 KASSERT(hispgrp->pg_jobc > 0); 1563 if (--hispgrp->pg_jobc == 0) 1564 orphanpg(hispgrp); 1565 } 1566 } 1567 } 1568 } 1569 1570 /* 1571 * A process group has become orphaned; 1572 * if there are any stopped processes in the group, 1573 * hang-up all process in that group. 1574 * 1575 * Call with proc_lock held. 1576 */ 1577 static void 1578 orphanpg(struct pgrp *pg) 1579 { 1580 struct proc *p; 1581 1582 KASSERT(mutex_owned(&proc_lock)); 1583 1584 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1585 if (p->p_stat == SSTOP) { 1586 p->p_lflag |= PL_ORPHANPG; 1587 psignal(p, SIGHUP); 1588 psignal(p, SIGCONT); 1589 } 1590 } 1591 } 1592 1593 #ifdef DDB 1594 #include <ddb/db_output.h> 1595 void pidtbl_dump(void); 1596 void 1597 pidtbl_dump(void) 1598 { 1599 struct pid_table *pt; 1600 struct proc *p; 1601 struct pgrp *pgrp; 1602 uintptr_t slot; 1603 int id; 1604 1605 db_printf("pid table %p size %x, next %x, last %x\n", 1606 pid_table, pid_tbl_mask+1, 1607 next_free_pt, last_free_pt); 1608 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) { 1609 slot = pt->pt_slot; 1610 if (!PT_VALID(slot) && !pt->pt_pgrp) 1611 continue; 1612 if (PT_IS_LWP(slot)) { 1613 p = PT_GET_LWP(slot)->l_proc; 1614 } else if (PT_IS_PROC(slot)) { 1615 p = PT_GET_PROC(slot); 1616 } else { 1617 p = NULL; 1618 } 1619 db_printf(" id %x: ", id); 1620 if (p != NULL) 1621 db_printf("slotpid %d proc %p id %d (0x%x) %s\n", 1622 pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm); 1623 else 1624 db_printf("next %x use %x\n", 1625 PT_NEXT(slot) & pid_tbl_mask, 1626 PT_NEXT(slot) & ~pid_tbl_mask); 1627 if ((pgrp = pt->pt_pgrp)) { 1628 db_printf("\tsession %p, sid %d, count %d, login %s\n", 1629 pgrp->pg_session, pgrp->pg_session->s_sid, 1630 pgrp->pg_session->s_count, 1631 pgrp->pg_session->s_login); 1632 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n", 1633 pgrp, pgrp->pg_id, pgrp->pg_jobc, 1634 LIST_FIRST(&pgrp->pg_members)); 1635 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 1636 db_printf("\t\tpid %d addr %p pgrp %p %s\n", 1637 p->p_pid, p, p->p_pgrp, p->p_comm); 1638 } 1639 } 1640 } 1641 } 1642 #endif /* DDB */ 1643 1644 #ifdef KSTACK_CHECK_MAGIC 1645 1646 #define KSTACK_MAGIC 0xdeadbeaf 1647 1648 /* XXX should be per process basis? */ 1649 static int kstackleftmin = KSTACK_SIZE; 1650 static int kstackleftthres = KSTACK_SIZE / 8; 1651 1652 void 1653 kstack_setup_magic(const struct lwp *l) 1654 { 1655 uint32_t *ip; 1656 uint32_t const *end; 1657 1658 KASSERT(l != NULL); 1659 KASSERT(l != &lwp0); 1660 1661 /* 1662 * fill all the stack with magic number 1663 * so that later modification on it can be detected. 1664 */ 1665 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1666 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1667 for (; ip < end; ip++) { 1668 *ip = KSTACK_MAGIC; 1669 } 1670 } 1671 1672 void 1673 kstack_check_magic(const struct lwp *l) 1674 { 1675 uint32_t const *ip, *end; 1676 int stackleft; 1677 1678 KASSERT(l != NULL); 1679 1680 /* don't check proc0 */ /*XXX*/ 1681 if (l == &lwp0) 1682 return; 1683 1684 #ifdef __MACHINE_STACK_GROWS_UP 1685 /* stack grows upwards (eg. hppa) */ 1686 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1687 end = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1688 for (ip--; ip >= end; ip--) 1689 if (*ip != KSTACK_MAGIC) 1690 break; 1691 1692 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip; 1693 #else /* __MACHINE_STACK_GROWS_UP */ 1694 /* stack grows downwards (eg. i386) */ 1695 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1696 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1697 for (; ip < end; ip++) 1698 if (*ip != KSTACK_MAGIC) 1699 break; 1700 1701 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l); 1702 #endif /* __MACHINE_STACK_GROWS_UP */ 1703 1704 if (kstackleftmin > stackleft) { 1705 kstackleftmin = stackleft; 1706 if (stackleft < kstackleftthres) 1707 printf("warning: kernel stack left %d bytes" 1708 "(pid %u:lid %u)\n", stackleft, 1709 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1710 } 1711 1712 if (stackleft <= 0) { 1713 panic("magic on the top of kernel stack changed for " 1714 "pid %u, lid %u: maybe kernel stack overflow", 1715 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1716 } 1717 } 1718 #endif /* KSTACK_CHECK_MAGIC */ 1719 1720 int 1721 proclist_foreach_call(struct proclist *list, 1722 int (*callback)(struct proc *, void *arg), void *arg) 1723 { 1724 struct proc marker; 1725 struct proc *p; 1726 int ret = 0; 1727 1728 marker.p_flag = PK_MARKER; 1729 mutex_enter(&proc_lock); 1730 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) { 1731 if (p->p_flag & PK_MARKER) { 1732 p = LIST_NEXT(p, p_list); 1733 continue; 1734 } 1735 LIST_INSERT_AFTER(p, &marker, p_list); 1736 ret = (*callback)(p, arg); 1737 KASSERT(mutex_owned(&proc_lock)); 1738 p = LIST_NEXT(&marker, p_list); 1739 LIST_REMOVE(&marker, p_list); 1740 } 1741 mutex_exit(&proc_lock); 1742 1743 return ret; 1744 } 1745 1746 int 1747 proc_vmspace_getref(struct proc *p, struct vmspace **vm) 1748 { 1749 1750 /* XXXCDC: how should locking work here? */ 1751 1752 /* curproc exception is for coredump. */ 1753 1754 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) || 1755 (p->p_vmspace->vm_refcnt < 1)) { 1756 return EFAULT; 1757 } 1758 1759 uvmspace_addref(p->p_vmspace); 1760 *vm = p->p_vmspace; 1761 1762 return 0; 1763 } 1764 1765 /* 1766 * Acquire a write lock on the process credential. 1767 */ 1768 void 1769 proc_crmod_enter(void) 1770 { 1771 struct lwp *l = curlwp; 1772 struct proc *p = l->l_proc; 1773 kauth_cred_t oc; 1774 1775 /* Reset what needs to be reset in plimit. */ 1776 if (p->p_limit->pl_corename != defcorename) { 1777 lim_setcorename(p, defcorename, 0); 1778 } 1779 1780 mutex_enter(p->p_lock); 1781 1782 /* Ensure the LWP cached credentials are up to date. */ 1783 if ((oc = l->l_cred) != p->p_cred) { 1784 kauth_cred_hold(p->p_cred); 1785 l->l_cred = p->p_cred; 1786 kauth_cred_free(oc); 1787 } 1788 } 1789 1790 /* 1791 * Set in a new process credential, and drop the write lock. The credential 1792 * must have a reference already. Optionally, free a no-longer required 1793 * credential. The scheduler also needs to inspect p_cred, so we also 1794 * briefly acquire the sched state mutex. 1795 */ 1796 void 1797 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid) 1798 { 1799 struct lwp *l = curlwp, *l2; 1800 struct proc *p = l->l_proc; 1801 kauth_cred_t oc; 1802 1803 KASSERT(mutex_owned(p->p_lock)); 1804 1805 /* Is there a new credential to set in? */ 1806 if (scred != NULL) { 1807 p->p_cred = scred; 1808 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 1809 if (l2 != l) 1810 l2->l_prflag |= LPR_CRMOD; 1811 } 1812 1813 /* Ensure the LWP cached credentials are up to date. */ 1814 if ((oc = l->l_cred) != scred) { 1815 kauth_cred_hold(scred); 1816 l->l_cred = scred; 1817 } 1818 } else 1819 oc = NULL; /* XXXgcc */ 1820 1821 if (sugid) { 1822 /* 1823 * Mark process as having changed credentials, stops 1824 * tracing etc. 1825 */ 1826 p->p_flag |= PK_SUGID; 1827 } 1828 1829 mutex_exit(p->p_lock); 1830 1831 /* If there is a credential to be released, free it now. */ 1832 if (fcred != NULL) { 1833 KASSERT(scred != NULL); 1834 kauth_cred_free(fcred); 1835 if (oc != scred) 1836 kauth_cred_free(oc); 1837 } 1838 } 1839 1840 /* 1841 * proc_specific_key_create -- 1842 * Create a key for subsystem proc-specific data. 1843 */ 1844 int 1845 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 1846 { 1847 1848 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor)); 1849 } 1850 1851 /* 1852 * proc_specific_key_delete -- 1853 * Delete a key for subsystem proc-specific data. 1854 */ 1855 void 1856 proc_specific_key_delete(specificdata_key_t key) 1857 { 1858 1859 specificdata_key_delete(proc_specificdata_domain, key); 1860 } 1861 1862 /* 1863 * proc_initspecific -- 1864 * Initialize a proc's specificdata container. 1865 */ 1866 void 1867 proc_initspecific(struct proc *p) 1868 { 1869 int error __diagused; 1870 1871 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref); 1872 KASSERT(error == 0); 1873 } 1874 1875 /* 1876 * proc_finispecific -- 1877 * Finalize a proc's specificdata container. 1878 */ 1879 void 1880 proc_finispecific(struct proc *p) 1881 { 1882 1883 specificdata_fini(proc_specificdata_domain, &p->p_specdataref); 1884 } 1885 1886 /* 1887 * proc_getspecific -- 1888 * Return proc-specific data corresponding to the specified key. 1889 */ 1890 void * 1891 proc_getspecific(struct proc *p, specificdata_key_t key) 1892 { 1893 1894 return (specificdata_getspecific(proc_specificdata_domain, 1895 &p->p_specdataref, key)); 1896 } 1897 1898 /* 1899 * proc_setspecific -- 1900 * Set proc-specific data corresponding to the specified key. 1901 */ 1902 void 1903 proc_setspecific(struct proc *p, specificdata_key_t key, void *data) 1904 { 1905 1906 specificdata_setspecific(proc_specificdata_domain, 1907 &p->p_specdataref, key, data); 1908 } 1909 1910 int 1911 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target) 1912 { 1913 int r = 0; 1914 1915 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) || 1916 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) { 1917 /* 1918 * suid proc of ours or proc not ours 1919 */ 1920 r = EPERM; 1921 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) { 1922 /* 1923 * sgid proc has sgid back to us temporarily 1924 */ 1925 r = EPERM; 1926 } else { 1927 /* 1928 * our rgid must be in target's group list (ie, 1929 * sub-processes started by a sgid process) 1930 */ 1931 int ismember = 0; 1932 1933 if (kauth_cred_ismember_gid(cred, 1934 kauth_cred_getgid(target), &ismember) != 0 || 1935 !ismember) 1936 r = EPERM; 1937 } 1938 1939 return (r); 1940 } 1941 1942 /* 1943 * sysctl stuff 1944 */ 1945 1946 #define KERN_PROCSLOP (5 * sizeof(struct kinfo_proc)) 1947 1948 static const u_int sysctl_flagmap[] = { 1949 PK_ADVLOCK, P_ADVLOCK, 1950 PK_EXEC, P_EXEC, 1951 PK_NOCLDWAIT, P_NOCLDWAIT, 1952 PK_32, P_32, 1953 PK_CLDSIGIGN, P_CLDSIGIGN, 1954 PK_SUGID, P_SUGID, 1955 0 1956 }; 1957 1958 static const u_int sysctl_sflagmap[] = { 1959 PS_NOCLDSTOP, P_NOCLDSTOP, 1960 PS_WEXIT, P_WEXIT, 1961 PS_STOPFORK, P_STOPFORK, 1962 PS_STOPEXEC, P_STOPEXEC, 1963 PS_STOPEXIT, P_STOPEXIT, 1964 0 1965 }; 1966 1967 static const u_int sysctl_slflagmap[] = { 1968 PSL_TRACED, P_TRACED, 1969 PSL_CHTRACED, P_CHTRACED, 1970 PSL_SYSCALL, P_SYSCALL, 1971 0 1972 }; 1973 1974 static const u_int sysctl_lflagmap[] = { 1975 PL_CONTROLT, P_CONTROLT, 1976 PL_PPWAIT, P_PPWAIT, 1977 0 1978 }; 1979 1980 static const u_int sysctl_stflagmap[] = { 1981 PST_PROFIL, P_PROFIL, 1982 0 1983 1984 }; 1985 1986 /* used by kern_lwp also */ 1987 const u_int sysctl_lwpflagmap[] = { 1988 LW_SINTR, L_SINTR, 1989 LW_SYSTEM, L_SYSTEM, 1990 0 1991 }; 1992 1993 /* 1994 * Find the most ``active'' lwp of a process and return it for ps display 1995 * purposes 1996 */ 1997 static struct lwp * 1998 proc_active_lwp(struct proc *p) 1999 { 2000 static const int ostat[] = { 2001 0, 2002 2, /* LSIDL */ 2003 6, /* LSRUN */ 2004 5, /* LSSLEEP */ 2005 4, /* LSSTOP */ 2006 0, /* LSZOMB */ 2007 1, /* LSDEAD */ 2008 7, /* LSONPROC */ 2009 3 /* LSSUSPENDED */ 2010 }; 2011 2012 struct lwp *l, *lp = NULL; 2013 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 2014 KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat)); 2015 if (lp == NULL || 2016 ostat[l->l_stat] > ostat[lp->l_stat] || 2017 (ostat[l->l_stat] == ostat[lp->l_stat] && 2018 l->l_cpticks > lp->l_cpticks)) { 2019 lp = l; 2020 continue; 2021 } 2022 } 2023 return lp; 2024 } 2025 2026 static int 2027 sysctl_doeproc(SYSCTLFN_ARGS) 2028 { 2029 union { 2030 struct kinfo_proc kproc; 2031 struct kinfo_proc2 kproc2; 2032 } *kbuf; 2033 struct proc *p, *next, *marker; 2034 char *where, *dp; 2035 int type, op, arg, error; 2036 u_int elem_size, kelem_size, elem_count; 2037 size_t buflen, needed; 2038 bool match, zombie, mmmbrains; 2039 const bool allowaddr = get_expose_address(curproc); 2040 2041 if (namelen == 1 && name[0] == CTL_QUERY) 2042 return (sysctl_query(SYSCTLFN_CALL(rnode))); 2043 2044 dp = where = oldp; 2045 buflen = where != NULL ? *oldlenp : 0; 2046 error = 0; 2047 needed = 0; 2048 type = rnode->sysctl_num; 2049 2050 if (type == KERN_PROC) { 2051 if (namelen == 0) 2052 return EINVAL; 2053 switch (op = name[0]) { 2054 case KERN_PROC_ALL: 2055 if (namelen != 1) 2056 return EINVAL; 2057 arg = 0; 2058 break; 2059 default: 2060 if (namelen != 2) 2061 return EINVAL; 2062 arg = name[1]; 2063 break; 2064 } 2065 elem_count = 0; /* Hush little compiler, don't you cry */ 2066 kelem_size = elem_size = sizeof(kbuf->kproc); 2067 } else { 2068 if (namelen != 4) 2069 return EINVAL; 2070 op = name[0]; 2071 arg = name[1]; 2072 elem_size = name[2]; 2073 elem_count = name[3]; 2074 kelem_size = sizeof(kbuf->kproc2); 2075 } 2076 2077 sysctl_unlock(); 2078 2079 kbuf = kmem_zalloc(sizeof(*kbuf), KM_SLEEP); 2080 marker = kmem_alloc(sizeof(*marker), KM_SLEEP); 2081 marker->p_flag = PK_MARKER; 2082 2083 mutex_enter(&proc_lock); 2084 /* 2085 * Start with zombies to prevent reporting processes twice, in case they 2086 * are dying and being moved from the list of alive processes to zombies. 2087 */ 2088 mmmbrains = true; 2089 for (p = LIST_FIRST(&zombproc);; p = next) { 2090 if (p == NULL) { 2091 if (mmmbrains) { 2092 p = LIST_FIRST(&allproc); 2093 mmmbrains = false; 2094 } 2095 if (p == NULL) 2096 break; 2097 } 2098 next = LIST_NEXT(p, p_list); 2099 if ((p->p_flag & PK_MARKER) != 0) 2100 continue; 2101 2102 /* 2103 * Skip embryonic processes. 2104 */ 2105 if (p->p_stat == SIDL) 2106 continue; 2107 2108 mutex_enter(p->p_lock); 2109 error = kauth_authorize_process(l->l_cred, 2110 KAUTH_PROCESS_CANSEE, p, 2111 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL); 2112 if (error != 0) { 2113 mutex_exit(p->p_lock); 2114 continue; 2115 } 2116 2117 /* 2118 * Hande all the operations in one switch on the cost of 2119 * algorithm complexity is on purpose. The win splitting this 2120 * function into several similar copies makes maintenance burden 2121 * burden, code grow and boost is neglible in practical systems. 2122 */ 2123 switch (op) { 2124 case KERN_PROC_PID: 2125 match = (p->p_pid == (pid_t)arg); 2126 break; 2127 2128 case KERN_PROC_PGRP: 2129 match = (p->p_pgrp->pg_id == (pid_t)arg); 2130 break; 2131 2132 case KERN_PROC_SESSION: 2133 match = (p->p_session->s_sid == (pid_t)arg); 2134 break; 2135 2136 case KERN_PROC_TTY: 2137 match = true; 2138 if (arg == (int) KERN_PROC_TTY_REVOKE) { 2139 if ((p->p_lflag & PL_CONTROLT) == 0 || 2140 p->p_session->s_ttyp == NULL || 2141 p->p_session->s_ttyvp != NULL) { 2142 match = false; 2143 } 2144 } else if ((p->p_lflag & PL_CONTROLT) == 0 || 2145 p->p_session->s_ttyp == NULL) { 2146 if ((dev_t)arg != KERN_PROC_TTY_NODEV) { 2147 match = false; 2148 } 2149 } else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) { 2150 match = false; 2151 } 2152 break; 2153 2154 case KERN_PROC_UID: 2155 match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg); 2156 break; 2157 2158 case KERN_PROC_RUID: 2159 match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg); 2160 break; 2161 2162 case KERN_PROC_GID: 2163 match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg); 2164 break; 2165 2166 case KERN_PROC_RGID: 2167 match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg); 2168 break; 2169 2170 case KERN_PROC_ALL: 2171 match = true; 2172 /* allow everything */ 2173 break; 2174 2175 default: 2176 error = EINVAL; 2177 mutex_exit(p->p_lock); 2178 goto cleanup; 2179 } 2180 if (!match) { 2181 mutex_exit(p->p_lock); 2182 continue; 2183 } 2184 2185 /* 2186 * Grab a hold on the process. 2187 */ 2188 if (mmmbrains) { 2189 zombie = true; 2190 } else { 2191 zombie = !rw_tryenter(&p->p_reflock, RW_READER); 2192 } 2193 if (zombie) { 2194 LIST_INSERT_AFTER(p, marker, p_list); 2195 } 2196 2197 if (buflen >= elem_size && 2198 (type == KERN_PROC || elem_count > 0)) { 2199 ruspace(p); /* Update process vm resource use */ 2200 2201 if (type == KERN_PROC) { 2202 fill_proc(p, &kbuf->kproc.kp_proc, allowaddr); 2203 fill_eproc(p, &kbuf->kproc.kp_eproc, zombie, 2204 allowaddr); 2205 } else { 2206 fill_kproc2(p, &kbuf->kproc2, zombie, 2207 allowaddr); 2208 elem_count--; 2209 } 2210 mutex_exit(p->p_lock); 2211 mutex_exit(&proc_lock); 2212 /* 2213 * Copy out elem_size, but not larger than kelem_size 2214 */ 2215 error = sysctl_copyout(l, kbuf, dp, 2216 uimin(kelem_size, elem_size)); 2217 mutex_enter(&proc_lock); 2218 if (error) { 2219 goto bah; 2220 } 2221 dp += elem_size; 2222 buflen -= elem_size; 2223 } else { 2224 mutex_exit(p->p_lock); 2225 } 2226 needed += elem_size; 2227 2228 /* 2229 * Release reference to process. 2230 */ 2231 if (zombie) { 2232 next = LIST_NEXT(marker, p_list); 2233 LIST_REMOVE(marker, p_list); 2234 } else { 2235 rw_exit(&p->p_reflock); 2236 next = LIST_NEXT(p, p_list); 2237 } 2238 2239 /* 2240 * Short-circuit break quickly! 2241 */ 2242 if (op == KERN_PROC_PID) 2243 break; 2244 } 2245 mutex_exit(&proc_lock); 2246 2247 if (where != NULL) { 2248 *oldlenp = dp - where; 2249 if (needed > *oldlenp) { 2250 error = ENOMEM; 2251 goto out; 2252 } 2253 } else { 2254 needed += KERN_PROCSLOP; 2255 *oldlenp = needed; 2256 } 2257 kmem_free(kbuf, sizeof(*kbuf)); 2258 kmem_free(marker, sizeof(*marker)); 2259 sysctl_relock(); 2260 return 0; 2261 bah: 2262 if (zombie) 2263 LIST_REMOVE(marker, p_list); 2264 else 2265 rw_exit(&p->p_reflock); 2266 cleanup: 2267 mutex_exit(&proc_lock); 2268 out: 2269 kmem_free(kbuf, sizeof(*kbuf)); 2270 kmem_free(marker, sizeof(*marker)); 2271 sysctl_relock(); 2272 return error; 2273 } 2274 2275 int 2276 copyin_psstrings(struct proc *p, struct ps_strings *arginfo) 2277 { 2278 #if !defined(_RUMPKERNEL) 2279 int retval; 2280 2281 if (p->p_flag & PK_32) { 2282 MODULE_HOOK_CALL(kern_proc32_copyin_hook, (p, arginfo), 2283 enosys(), retval); 2284 return retval; 2285 } 2286 #endif /* !defined(_RUMPKERNEL) */ 2287 2288 return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo)); 2289 } 2290 2291 static int 2292 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len) 2293 { 2294 void **cookie = cookie_; 2295 struct lwp *l = cookie[0]; 2296 char *dst = cookie[1]; 2297 2298 return sysctl_copyout(l, src, dst + off, len); 2299 } 2300 2301 /* 2302 * sysctl helper routine for kern.proc_args pseudo-subtree. 2303 */ 2304 static int 2305 sysctl_kern_proc_args(SYSCTLFN_ARGS) 2306 { 2307 struct ps_strings pss; 2308 struct proc *p; 2309 pid_t pid; 2310 int type, error; 2311 void *cookie[2]; 2312 2313 if (namelen == 1 && name[0] == CTL_QUERY) 2314 return (sysctl_query(SYSCTLFN_CALL(rnode))); 2315 2316 if (newp != NULL || namelen != 2) 2317 return (EINVAL); 2318 pid = name[0]; 2319 type = name[1]; 2320 2321 switch (type) { 2322 case KERN_PROC_PATHNAME: 2323 sysctl_unlock(); 2324 error = fill_pathname(l, pid, oldp, oldlenp); 2325 sysctl_relock(); 2326 return error; 2327 2328 case KERN_PROC_CWD: 2329 sysctl_unlock(); 2330 error = fill_cwd(l, pid, oldp, oldlenp); 2331 sysctl_relock(); 2332 return error; 2333 2334 case KERN_PROC_ARGV: 2335 case KERN_PROC_NARGV: 2336 case KERN_PROC_ENV: 2337 case KERN_PROC_NENV: 2338 /* ok */ 2339 break; 2340 default: 2341 return (EINVAL); 2342 } 2343 2344 sysctl_unlock(); 2345 2346 /* check pid */ 2347 mutex_enter(&proc_lock); 2348 if ((p = proc_find(pid)) == NULL) { 2349 error = EINVAL; 2350 goto out_locked; 2351 } 2352 mutex_enter(p->p_lock); 2353 2354 /* Check permission. */ 2355 if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV) 2356 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 2357 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL); 2358 else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV) 2359 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 2360 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL); 2361 else 2362 error = EINVAL; /* XXXGCC */ 2363 if (error) { 2364 mutex_exit(p->p_lock); 2365 goto out_locked; 2366 } 2367 2368 if (oldp == NULL) { 2369 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) 2370 *oldlenp = sizeof (int); 2371 else 2372 *oldlenp = ARG_MAX; /* XXX XXX XXX */ 2373 error = 0; 2374 mutex_exit(p->p_lock); 2375 goto out_locked; 2376 } 2377 2378 /* 2379 * Zombies don't have a stack, so we can't read their psstrings. 2380 * System processes also don't have a user stack. 2381 */ 2382 if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) { 2383 error = EINVAL; 2384 mutex_exit(p->p_lock); 2385 goto out_locked; 2386 } 2387 2388 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 2389 mutex_exit(p->p_lock); 2390 if (error) { 2391 goto out_locked; 2392 } 2393 mutex_exit(&proc_lock); 2394 2395 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) { 2396 int value; 2397 if ((error = copyin_psstrings(p, &pss)) == 0) { 2398 if (type == KERN_PROC_NARGV) 2399 value = pss.ps_nargvstr; 2400 else 2401 value = pss.ps_nenvstr; 2402 error = sysctl_copyout(l, &value, oldp, sizeof(value)); 2403 *oldlenp = sizeof(value); 2404 } 2405 } else { 2406 cookie[0] = l; 2407 cookie[1] = oldp; 2408 error = copy_procargs(p, type, oldlenp, 2409 copy_procargs_sysctl_cb, cookie); 2410 } 2411 rw_exit(&p->p_reflock); 2412 sysctl_relock(); 2413 return error; 2414 2415 out_locked: 2416 mutex_exit(&proc_lock); 2417 sysctl_relock(); 2418 return error; 2419 } 2420 2421 int 2422 copy_procargs(struct proc *p, int oid, size_t *limit, 2423 int (*cb)(void *, const void *, size_t, size_t), void *cookie) 2424 { 2425 struct ps_strings pss; 2426 size_t len, i, loaded, entry_len; 2427 struct uio auio; 2428 struct iovec aiov; 2429 int error, argvlen; 2430 char *arg; 2431 char **argv; 2432 vaddr_t user_argv; 2433 struct vmspace *vmspace; 2434 2435 /* 2436 * Allocate a temporary buffer to hold the argument vector and 2437 * the arguments themselve. 2438 */ 2439 arg = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2440 argv = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2441 2442 /* 2443 * Lock the process down in memory. 2444 */ 2445 vmspace = p->p_vmspace; 2446 uvmspace_addref(vmspace); 2447 2448 /* 2449 * Read in the ps_strings structure. 2450 */ 2451 if ((error = copyin_psstrings(p, &pss)) != 0) 2452 goto done; 2453 2454 /* 2455 * Now read the address of the argument vector. 2456 */ 2457 switch (oid) { 2458 case KERN_PROC_ARGV: 2459 user_argv = (uintptr_t)pss.ps_argvstr; 2460 argvlen = pss.ps_nargvstr; 2461 break; 2462 case KERN_PROC_ENV: 2463 user_argv = (uintptr_t)pss.ps_envstr; 2464 argvlen = pss.ps_nenvstr; 2465 break; 2466 default: 2467 error = EINVAL; 2468 goto done; 2469 } 2470 2471 if (argvlen < 0) { 2472 error = EIO; 2473 goto done; 2474 } 2475 2476 2477 /* 2478 * Now copy each string. 2479 */ 2480 len = 0; /* bytes written to user buffer */ 2481 loaded = 0; /* bytes from argv already processed */ 2482 i = 0; /* To make compiler happy */ 2483 entry_len = PROC_PTRSZ(p); 2484 2485 for (; argvlen; --argvlen) { 2486 int finished = 0; 2487 vaddr_t base; 2488 size_t xlen; 2489 int j; 2490 2491 if (loaded == 0) { 2492 size_t rem = entry_len * argvlen; 2493 loaded = MIN(rem, PAGE_SIZE); 2494 error = copyin_vmspace(vmspace, 2495 (const void *)user_argv, argv, loaded); 2496 if (error) 2497 break; 2498 user_argv += loaded; 2499 i = 0; 2500 } 2501 2502 #if !defined(_RUMPKERNEL) 2503 if (p->p_flag & PK_32) 2504 MODULE_HOOK_CALL(kern_proc32_base_hook, 2505 (argv, i++), 0, base); 2506 else 2507 #endif /* !defined(_RUMPKERNEL) */ 2508 base = (vaddr_t)argv[i++]; 2509 loaded -= entry_len; 2510 2511 /* 2512 * The program has messed around with its arguments, 2513 * possibly deleting some, and replacing them with 2514 * NULL's. Treat this as the last argument and not 2515 * a failure. 2516 */ 2517 if (base == 0) 2518 break; 2519 2520 while (!finished) { 2521 xlen = PAGE_SIZE - (base & PAGE_MASK); 2522 2523 aiov.iov_base = arg; 2524 aiov.iov_len = PAGE_SIZE; 2525 auio.uio_iov = &aiov; 2526 auio.uio_iovcnt = 1; 2527 auio.uio_offset = base; 2528 auio.uio_resid = xlen; 2529 auio.uio_rw = UIO_READ; 2530 UIO_SETUP_SYSSPACE(&auio); 2531 error = uvm_io(&vmspace->vm_map, &auio, 0); 2532 if (error) 2533 goto done; 2534 2535 /* Look for the end of the string */ 2536 for (j = 0; j < xlen; j++) { 2537 if (arg[j] == '\0') { 2538 xlen = j + 1; 2539 finished = 1; 2540 break; 2541 } 2542 } 2543 2544 /* Check for user buffer overflow */ 2545 if (len + xlen > *limit) { 2546 finished = 1; 2547 if (len > *limit) 2548 xlen = 0; 2549 else 2550 xlen = *limit - len; 2551 } 2552 2553 /* Copyout the page */ 2554 error = (*cb)(cookie, arg, len, xlen); 2555 if (error) 2556 goto done; 2557 2558 len += xlen; 2559 base += xlen; 2560 } 2561 } 2562 *limit = len; 2563 2564 done: 2565 kmem_free(argv, PAGE_SIZE); 2566 kmem_free(arg, PAGE_SIZE); 2567 uvmspace_free(vmspace); 2568 return error; 2569 } 2570 2571 /* 2572 * Fill in a proc structure for the specified process. 2573 */ 2574 static void 2575 fill_proc(const struct proc *psrc, struct proc *p, bool allowaddr) 2576 { 2577 COND_SET_STRUCT(p->p_list, psrc->p_list, allowaddr); 2578 memset(&p->p_auxlock, 0, sizeof(p->p_auxlock)); 2579 COND_SET_STRUCT(p->p_lock, psrc->p_lock, allowaddr); 2580 memset(&p->p_stmutex, 0, sizeof(p->p_stmutex)); 2581 memset(&p->p_reflock, 0, sizeof(p->p_reflock)); 2582 COND_SET_STRUCT(p->p_waitcv, psrc->p_waitcv, allowaddr); 2583 COND_SET_STRUCT(p->p_lwpcv, psrc->p_lwpcv, allowaddr); 2584 COND_SET_PTR(p->p_cred, psrc->p_cred, allowaddr); 2585 COND_SET_PTR(p->p_fd, psrc->p_fd, allowaddr); 2586 COND_SET_PTR(p->p_cwdi, psrc->p_cwdi, allowaddr); 2587 COND_SET_PTR(p->p_stats, psrc->p_stats, allowaddr); 2588 COND_SET_PTR(p->p_limit, psrc->p_limit, allowaddr); 2589 COND_SET_PTR(p->p_vmspace, psrc->p_vmspace, allowaddr); 2590 COND_SET_PTR(p->p_sigacts, psrc->p_sigacts, allowaddr); 2591 COND_SET_PTR(p->p_aio, psrc->p_aio, allowaddr); 2592 p->p_mqueue_cnt = psrc->p_mqueue_cnt; 2593 memset(&p->p_specdataref, 0, sizeof(p->p_specdataref)); 2594 p->p_exitsig = psrc->p_exitsig; 2595 p->p_flag = psrc->p_flag; 2596 p->p_sflag = psrc->p_sflag; 2597 p->p_slflag = psrc->p_slflag; 2598 p->p_lflag = psrc->p_lflag; 2599 p->p_stflag = psrc->p_stflag; 2600 p->p_stat = psrc->p_stat; 2601 p->p_trace_enabled = psrc->p_trace_enabled; 2602 p->p_pid = psrc->p_pid; 2603 COND_SET_STRUCT(p->p_pglist, psrc->p_pglist, allowaddr); 2604 COND_SET_PTR(p->p_pptr, psrc->p_pptr, allowaddr); 2605 COND_SET_STRUCT(p->p_sibling, psrc->p_sibling, allowaddr); 2606 COND_SET_STRUCT(p->p_children, psrc->p_children, allowaddr); 2607 COND_SET_STRUCT(p->p_lwps, psrc->p_lwps, allowaddr); 2608 COND_SET_PTR(p->p_raslist, psrc->p_raslist, allowaddr); 2609 p->p_nlwps = psrc->p_nlwps; 2610 p->p_nzlwps = psrc->p_nzlwps; 2611 p->p_nrlwps = psrc->p_nrlwps; 2612 p->p_nlwpwait = psrc->p_nlwpwait; 2613 p->p_ndlwps = psrc->p_ndlwps; 2614 p->p_nstopchild = psrc->p_nstopchild; 2615 p->p_waited = psrc->p_waited; 2616 COND_SET_PTR(p->p_zomblwp, psrc->p_zomblwp, allowaddr); 2617 COND_SET_PTR(p->p_vforklwp, psrc->p_vforklwp, allowaddr); 2618 COND_SET_PTR(p->p_sched_info, psrc->p_sched_info, allowaddr); 2619 p->p_estcpu = psrc->p_estcpu; 2620 p->p_estcpu_inherited = psrc->p_estcpu_inherited; 2621 p->p_forktime = psrc->p_forktime; 2622 p->p_pctcpu = psrc->p_pctcpu; 2623 COND_SET_PTR(p->p_opptr, psrc->p_opptr, allowaddr); 2624 COND_SET_PTR(p->p_timers, psrc->p_timers, allowaddr); 2625 p->p_rtime = psrc->p_rtime; 2626 p->p_uticks = psrc->p_uticks; 2627 p->p_sticks = psrc->p_sticks; 2628 p->p_iticks = psrc->p_iticks; 2629 p->p_xutime = psrc->p_xutime; 2630 p->p_xstime = psrc->p_xstime; 2631 p->p_traceflag = psrc->p_traceflag; 2632 COND_SET_PTR(p->p_tracep, psrc->p_tracep, allowaddr); 2633 COND_SET_PTR(p->p_textvp, psrc->p_textvp, allowaddr); 2634 COND_SET_PTR(p->p_emul, psrc->p_emul, allowaddr); 2635 COND_SET_PTR(p->p_emuldata, psrc->p_emuldata, allowaddr); 2636 COND_SET_CPTR(p->p_execsw, psrc->p_execsw, allowaddr); 2637 COND_SET_STRUCT(p->p_klist, psrc->p_klist, allowaddr); 2638 COND_SET_STRUCT(p->p_sigwaiters, psrc->p_sigwaiters, allowaddr); 2639 COND_SET_STRUCT(p->p_sigpend.sp_info, psrc->p_sigpend.sp_info, 2640 allowaddr); 2641 p->p_sigpend.sp_set = psrc->p_sigpend.sp_set; 2642 COND_SET_PTR(p->p_lwpctl, psrc->p_lwpctl, allowaddr); 2643 p->p_ppid = psrc->p_ppid; 2644 p->p_oppid = psrc->p_oppid; 2645 COND_SET_PTR(p->p_path, psrc->p_path, allowaddr); 2646 p->p_sigctx = psrc->p_sigctx; 2647 p->p_nice = psrc->p_nice; 2648 memcpy(p->p_comm, psrc->p_comm, sizeof(p->p_comm)); 2649 COND_SET_PTR(p->p_pgrp, psrc->p_pgrp, allowaddr); 2650 COND_SET_VALUE(p->p_psstrp, psrc->p_psstrp, allowaddr); 2651 p->p_pax = psrc->p_pax; 2652 p->p_xexit = psrc->p_xexit; 2653 p->p_xsig = psrc->p_xsig; 2654 p->p_acflag = psrc->p_acflag; 2655 COND_SET_STRUCT(p->p_md, psrc->p_md, allowaddr); 2656 p->p_stackbase = psrc->p_stackbase; 2657 COND_SET_PTR(p->p_dtrace, psrc->p_dtrace, allowaddr); 2658 } 2659 2660 /* 2661 * Fill in an eproc structure for the specified process. 2662 */ 2663 void 2664 fill_eproc(struct proc *p, struct eproc *ep, bool zombie, bool allowaddr) 2665 { 2666 struct tty *tp; 2667 struct lwp *l; 2668 2669 KASSERT(mutex_owned(&proc_lock)); 2670 KASSERT(mutex_owned(p->p_lock)); 2671 2672 COND_SET_PTR(ep->e_paddr, p, allowaddr); 2673 COND_SET_PTR(ep->e_sess, p->p_session, allowaddr); 2674 if (p->p_cred) { 2675 kauth_cred_topcred(p->p_cred, &ep->e_pcred); 2676 kauth_cred_toucred(p->p_cred, &ep->e_ucred); 2677 } 2678 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2679 struct vmspace *vm = p->p_vmspace; 2680 2681 ep->e_vm.vm_rssize = vm_resident_count(vm); 2682 ep->e_vm.vm_tsize = vm->vm_tsize; 2683 ep->e_vm.vm_dsize = vm->vm_dsize; 2684 ep->e_vm.vm_ssize = vm->vm_ssize; 2685 ep->e_vm.vm_map.size = vm->vm_map.size; 2686 2687 /* Pick the primary (first) LWP */ 2688 l = proc_active_lwp(p); 2689 KASSERT(l != NULL); 2690 lwp_lock(l); 2691 if (l->l_wchan) 2692 strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN); 2693 lwp_unlock(l); 2694 } 2695 ep->e_ppid = p->p_ppid; 2696 if (p->p_pgrp && p->p_session) { 2697 ep->e_pgid = p->p_pgrp->pg_id; 2698 ep->e_jobc = p->p_pgrp->pg_jobc; 2699 ep->e_sid = p->p_session->s_sid; 2700 if ((p->p_lflag & PL_CONTROLT) && 2701 (tp = p->p_session->s_ttyp)) { 2702 ep->e_tdev = tp->t_dev; 2703 ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2704 COND_SET_PTR(ep->e_tsess, tp->t_session, allowaddr); 2705 } else 2706 ep->e_tdev = (uint32_t)NODEV; 2707 ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0; 2708 if (SESS_LEADER(p)) 2709 ep->e_flag |= EPROC_SLEADER; 2710 strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME); 2711 } 2712 ep->e_xsize = ep->e_xrssize = 0; 2713 ep->e_xccount = ep->e_xswrss = 0; 2714 } 2715 2716 /* 2717 * Fill in a kinfo_proc2 structure for the specified process. 2718 */ 2719 void 2720 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie, bool allowaddr) 2721 { 2722 struct tty *tp; 2723 struct lwp *l, *l2; 2724 struct timeval ut, st, rt; 2725 sigset_t ss1, ss2; 2726 struct rusage ru; 2727 struct vmspace *vm; 2728 2729 KASSERT(mutex_owned(&proc_lock)); 2730 KASSERT(mutex_owned(p->p_lock)); 2731 2732 sigemptyset(&ss1); 2733 sigemptyset(&ss2); 2734 2735 COND_SET_VALUE(ki->p_paddr, PTRTOUINT64(p), allowaddr); 2736 COND_SET_VALUE(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr); 2737 COND_SET_VALUE(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr); 2738 COND_SET_VALUE(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr); 2739 COND_SET_VALUE(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr); 2740 COND_SET_VALUE(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr); 2741 COND_SET_VALUE(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr); 2742 COND_SET_VALUE(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr); 2743 ki->p_tsess = 0; /* may be changed if controlling tty below */ 2744 COND_SET_VALUE(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr); 2745 ki->p_eflag = 0; 2746 ki->p_exitsig = p->p_exitsig; 2747 ki->p_flag = L_INMEM; /* Process never swapped out */ 2748 ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag); 2749 ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag); 2750 ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag); 2751 ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag); 2752 ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag); 2753 ki->p_pid = p->p_pid; 2754 ki->p_ppid = p->p_ppid; 2755 ki->p_uid = kauth_cred_geteuid(p->p_cred); 2756 ki->p_ruid = kauth_cred_getuid(p->p_cred); 2757 ki->p_gid = kauth_cred_getegid(p->p_cred); 2758 ki->p_rgid = kauth_cred_getgid(p->p_cred); 2759 ki->p_svuid = kauth_cred_getsvuid(p->p_cred); 2760 ki->p_svgid = kauth_cred_getsvgid(p->p_cred); 2761 ki->p_ngroups = kauth_cred_ngroups(p->p_cred); 2762 kauth_cred_getgroups(p->p_cred, ki->p_groups, 2763 uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])), 2764 UIO_SYSSPACE); 2765 2766 ki->p_uticks = p->p_uticks; 2767 ki->p_sticks = p->p_sticks; 2768 ki->p_iticks = p->p_iticks; 2769 ki->p_tpgid = NO_PGID; /* may be changed if controlling tty below */ 2770 COND_SET_VALUE(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr); 2771 ki->p_traceflag = p->p_traceflag; 2772 2773 memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t)); 2774 memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t)); 2775 2776 ki->p_cpticks = 0; 2777 ki->p_pctcpu = p->p_pctcpu; 2778 ki->p_estcpu = 0; 2779 ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */ 2780 ki->p_realstat = p->p_stat; 2781 ki->p_nice = p->p_nice; 2782 ki->p_xstat = P_WAITSTATUS(p); 2783 ki->p_acflag = p->p_acflag; 2784 2785 strncpy(ki->p_comm, p->p_comm, 2786 uimin(sizeof(ki->p_comm), sizeof(p->p_comm))); 2787 strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename)); 2788 2789 ki->p_nlwps = p->p_nlwps; 2790 ki->p_realflag = ki->p_flag; 2791 2792 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2793 vm = p->p_vmspace; 2794 ki->p_vm_rssize = vm_resident_count(vm); 2795 ki->p_vm_tsize = vm->vm_tsize; 2796 ki->p_vm_dsize = vm->vm_dsize; 2797 ki->p_vm_ssize = vm->vm_ssize; 2798 ki->p_vm_vsize = atop(vm->vm_map.size); 2799 /* 2800 * Since the stack is initially mapped mostly with 2801 * PROT_NONE and grown as needed, adjust the "mapped size" 2802 * to skip the unused stack portion. 2803 */ 2804 ki->p_vm_msize = 2805 atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize; 2806 2807 /* Pick the primary (first) LWP */ 2808 l = proc_active_lwp(p); 2809 KASSERT(l != NULL); 2810 lwp_lock(l); 2811 ki->p_nrlwps = p->p_nrlwps; 2812 ki->p_forw = 0; 2813 ki->p_back = 0; 2814 COND_SET_VALUE(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr); 2815 ki->p_stat = l->l_stat; 2816 ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag); 2817 ki->p_swtime = l->l_swtime; 2818 ki->p_slptime = l->l_slptime; 2819 if (l->l_stat == LSONPROC) 2820 ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags; 2821 else 2822 ki->p_schedflags = 0; 2823 ki->p_priority = lwp_eprio(l); 2824 ki->p_usrpri = l->l_priority; 2825 if (l->l_wchan) 2826 strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg)); 2827 COND_SET_VALUE(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr); 2828 ki->p_cpuid = cpu_index(l->l_cpu); 2829 lwp_unlock(l); 2830 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 2831 /* This is hardly correct, but... */ 2832 sigplusset(&l->l_sigpend.sp_set, &ss1); 2833 sigplusset(&l->l_sigmask, &ss2); 2834 ki->p_cpticks += l->l_cpticks; 2835 ki->p_pctcpu += l->l_pctcpu; 2836 ki->p_estcpu += l->l_estcpu; 2837 } 2838 } 2839 sigplusset(&p->p_sigpend.sp_set, &ss1); 2840 memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t)); 2841 memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t)); 2842 2843 if (p->p_session != NULL) { 2844 ki->p_sid = p->p_session->s_sid; 2845 ki->p__pgid = p->p_pgrp->pg_id; 2846 if (p->p_session->s_ttyvp) 2847 ki->p_eflag |= EPROC_CTTY; 2848 if (SESS_LEADER(p)) 2849 ki->p_eflag |= EPROC_SLEADER; 2850 strncpy(ki->p_login, p->p_session->s_login, 2851 uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login)); 2852 ki->p_jobc = p->p_pgrp->pg_jobc; 2853 if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) { 2854 ki->p_tdev = tp->t_dev; 2855 ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2856 COND_SET_VALUE(ki->p_tsess, PTRTOUINT64(tp->t_session), 2857 allowaddr); 2858 } else { 2859 ki->p_tdev = (int32_t)NODEV; 2860 } 2861 } 2862 2863 if (!P_ZOMBIE(p) && !zombie) { 2864 ki->p_uvalid = 1; 2865 ki->p_ustart_sec = p->p_stats->p_start.tv_sec; 2866 ki->p_ustart_usec = p->p_stats->p_start.tv_usec; 2867 2868 calcru(p, &ut, &st, NULL, &rt); 2869 ki->p_rtime_sec = rt.tv_sec; 2870 ki->p_rtime_usec = rt.tv_usec; 2871 ki->p_uutime_sec = ut.tv_sec; 2872 ki->p_uutime_usec = ut.tv_usec; 2873 ki->p_ustime_sec = st.tv_sec; 2874 ki->p_ustime_usec = st.tv_usec; 2875 2876 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 2877 ki->p_uru_nvcsw = 0; 2878 ki->p_uru_nivcsw = 0; 2879 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 2880 ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw); 2881 ki->p_uru_nivcsw += l2->l_nivcsw; 2882 ruadd(&ru, &l2->l_ru); 2883 } 2884 ki->p_uru_maxrss = ru.ru_maxrss; 2885 ki->p_uru_ixrss = ru.ru_ixrss; 2886 ki->p_uru_idrss = ru.ru_idrss; 2887 ki->p_uru_isrss = ru.ru_isrss; 2888 ki->p_uru_minflt = ru.ru_minflt; 2889 ki->p_uru_majflt = ru.ru_majflt; 2890 ki->p_uru_nswap = ru.ru_nswap; 2891 ki->p_uru_inblock = ru.ru_inblock; 2892 ki->p_uru_oublock = ru.ru_oublock; 2893 ki->p_uru_msgsnd = ru.ru_msgsnd; 2894 ki->p_uru_msgrcv = ru.ru_msgrcv; 2895 ki->p_uru_nsignals = ru.ru_nsignals; 2896 2897 timeradd(&p->p_stats->p_cru.ru_utime, 2898 &p->p_stats->p_cru.ru_stime, &ut); 2899 ki->p_uctime_sec = ut.tv_sec; 2900 ki->p_uctime_usec = ut.tv_usec; 2901 } 2902 } 2903 2904 2905 int 2906 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid) 2907 { 2908 int error; 2909 2910 mutex_enter(&proc_lock); 2911 if (pid == -1) 2912 *p = l->l_proc; 2913 else 2914 *p = proc_find(pid); 2915 2916 if (*p == NULL) { 2917 if (pid != -1) 2918 mutex_exit(&proc_lock); 2919 return ESRCH; 2920 } 2921 if (pid != -1) 2922 mutex_enter((*p)->p_lock); 2923 mutex_exit(&proc_lock); 2924 2925 error = kauth_authorize_process(l->l_cred, 2926 KAUTH_PROCESS_CANSEE, *p, 2927 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 2928 if (error) { 2929 if (pid != -1) 2930 mutex_exit((*p)->p_lock); 2931 } 2932 return error; 2933 } 2934 2935 static int 2936 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2937 { 2938 int error; 2939 struct proc *p; 2940 2941 if ((error = proc_find_locked(l, &p, pid)) != 0) 2942 return error; 2943 2944 if (p->p_path == NULL) { 2945 if (pid != -1) 2946 mutex_exit(p->p_lock); 2947 return ENOENT; 2948 } 2949 2950 size_t len = strlen(p->p_path) + 1; 2951 if (oldp != NULL) { 2952 size_t copylen = uimin(len, *oldlenp); 2953 error = sysctl_copyout(l, p->p_path, oldp, copylen); 2954 if (error == 0 && *oldlenp < len) 2955 error = ENOSPC; 2956 } 2957 *oldlenp = len; 2958 if (pid != -1) 2959 mutex_exit(p->p_lock); 2960 return error; 2961 } 2962 2963 static int 2964 fill_cwd(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2965 { 2966 int error; 2967 struct proc *p; 2968 char *path; 2969 char *bp, *bend; 2970 struct cwdinfo *cwdi; 2971 struct vnode *vp; 2972 size_t len, lenused; 2973 2974 if ((error = proc_find_locked(l, &p, pid)) != 0) 2975 return error; 2976 2977 len = MAXPATHLEN * 4; 2978 2979 path = kmem_alloc(len, KM_SLEEP); 2980 2981 bp = &path[len]; 2982 bend = bp; 2983 *(--bp) = '\0'; 2984 2985 cwdi = p->p_cwdi; 2986 rw_enter(&cwdi->cwdi_lock, RW_READER); 2987 vp = cwdi->cwdi_cdir; 2988 error = getcwd_common(vp, NULL, &bp, path, len/2, 0, l); 2989 rw_exit(&cwdi->cwdi_lock); 2990 2991 if (error) 2992 goto out; 2993 2994 lenused = bend - bp; 2995 2996 if (oldp != NULL) { 2997 size_t copylen = uimin(lenused, *oldlenp); 2998 error = sysctl_copyout(l, bp, oldp, copylen); 2999 if (error == 0 && *oldlenp < lenused) 3000 error = ENOSPC; 3001 } 3002 *oldlenp = lenused; 3003 out: 3004 if (pid != -1) 3005 mutex_exit(p->p_lock); 3006 kmem_free(path, len); 3007 return error; 3008 } 3009 3010 int 3011 proc_getauxv(struct proc *p, void **buf, size_t *len) 3012 { 3013 struct ps_strings pss; 3014 int error; 3015 void *uauxv, *kauxv; 3016 size_t size; 3017 3018 if ((error = copyin_psstrings(p, &pss)) != 0) 3019 return error; 3020 if (pss.ps_envstr == NULL) 3021 return EIO; 3022 3023 size = p->p_execsw->es_arglen; 3024 if (size == 0) 3025 return EIO; 3026 3027 size_t ptrsz = PROC_PTRSZ(p); 3028 uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz); 3029 3030 kauxv = kmem_alloc(size, KM_SLEEP); 3031 3032 error = copyin_proc(p, uauxv, kauxv, size); 3033 if (error) { 3034 kmem_free(kauxv, size); 3035 return error; 3036 } 3037 3038 *buf = kauxv; 3039 *len = size; 3040 3041 return 0; 3042 } 3043 3044 3045 static int 3046 sysctl_security_expose_address(SYSCTLFN_ARGS) 3047 { 3048 int expose_address, error; 3049 struct sysctlnode node; 3050 3051 node = *rnode; 3052 node.sysctl_data = &expose_address; 3053 expose_address = *(int *)rnode->sysctl_data; 3054 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 3055 if (error || newp == NULL) 3056 return error; 3057 3058 if (kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_KERNADDR, 3059 0, NULL, NULL, NULL)) 3060 return EPERM; 3061 3062 switch (expose_address) { 3063 case 0: 3064 case 1: 3065 case 2: 3066 break; 3067 default: 3068 return EINVAL; 3069 } 3070 3071 *(int *)rnode->sysctl_data = expose_address; 3072 3073 return 0; 3074 } 3075 3076 bool 3077 get_expose_address(struct proc *p) 3078 { 3079 /* allow only if sysctl variable is set or privileged */ 3080 return kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE, 3081 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL) == 0; 3082 } 3083