1 /* $NetBSD: kern_proc.c,v 1.217 2018/09/04 16:03:56 maxv Exp $ */ 2 3 /*- 4 * Copyright (c) 1999, 2006, 2007, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center, and by Andrew Doran. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1989, 1991, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)kern_proc.c 8.7 (Berkeley) 2/14/95 62 */ 63 64 #include <sys/cdefs.h> 65 __KERNEL_RCSID(0, "$NetBSD: kern_proc.c,v 1.217 2018/09/04 16:03:56 maxv 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 #endif 73 74 #if defined(__HAVE_COMPAT_NETBSD32) && !defined(COMPAT_NETBSD32) \ 75 && !defined(_RUMPKERNEL) 76 #define COMPAT_NETBSD32 77 #endif 78 79 #include <sys/param.h> 80 #include <sys/systm.h> 81 #include <sys/kernel.h> 82 #include <sys/proc.h> 83 #include <sys/resourcevar.h> 84 #include <sys/buf.h> 85 #include <sys/acct.h> 86 #include <sys/wait.h> 87 #include <sys/file.h> 88 #include <ufs/ufs/quota.h> 89 #include <sys/uio.h> 90 #include <sys/pool.h> 91 #include <sys/pset.h> 92 #include <sys/mbuf.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> 108 109 #include <uvm/uvm_extern.h> 110 #include <uvm/uvm.h> 111 112 #ifdef COMPAT_NETBSD32 113 #include <compat/netbsd32/netbsd32.h> 114 #endif 115 116 /* 117 * Process lists. 118 */ 119 120 struct proclist allproc __cacheline_aligned; 121 struct proclist zombproc __cacheline_aligned; 122 123 kmutex_t * proc_lock __cacheline_aligned; 124 125 /* 126 * pid to proc lookup is done by indexing the pid_table array. 127 * Since pid numbers are only allocated when an empty slot 128 * has been found, there is no need to search any lists ever. 129 * (an orphaned pgrp will lock the slot, a session will lock 130 * the pgrp with the same number.) 131 * If the table is too small it is reallocated with twice the 132 * previous size and the entries 'unzipped' into the two halves. 133 * A linked list of free entries is passed through the pt_proc 134 * field of 'free' items - set odd to be an invalid ptr. 135 */ 136 137 struct pid_table { 138 struct proc *pt_proc; 139 struct pgrp *pt_pgrp; 140 pid_t pt_pid; 141 }; 142 #if 1 /* strongly typed cast - should be a noop */ 143 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; } 144 #else 145 #define p2u(p) ((uint)p) 146 #endif 147 #define P_VALID(p) (!(p2u(p) & 1)) 148 #define P_NEXT(p) (p2u(p) >> 1) 149 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1)) 150 151 /* 152 * Table of process IDs (PIDs). 153 */ 154 static struct pid_table *pid_table __read_mostly; 155 156 #define INITIAL_PID_TABLE_SIZE (1 << 5) 157 158 /* Table mask, threshold for growing and number of allocated PIDs. */ 159 static u_int pid_tbl_mask __read_mostly; 160 static u_int pid_alloc_lim __read_mostly; 161 static u_int pid_alloc_cnt __cacheline_aligned; 162 163 /* Next free, last free and maximum PIDs. */ 164 static u_int next_free_pt __cacheline_aligned; 165 static u_int last_free_pt __cacheline_aligned; 166 static pid_t pid_max __read_mostly; 167 168 /* Components of the first process -- never freed. */ 169 170 extern struct emul emul_netbsd; /* defined in kern_exec.c */ 171 172 struct session session0 = { 173 .s_count = 1, 174 .s_sid = 0, 175 }; 176 struct pgrp pgrp0 = { 177 .pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members), 178 .pg_session = &session0, 179 }; 180 filedesc_t filedesc0; 181 struct cwdinfo cwdi0 = { 182 .cwdi_cmask = CMASK, 183 .cwdi_refcnt = 1, 184 }; 185 struct plimit limit0; 186 struct pstats pstat0; 187 struct vmspace vmspace0; 188 struct sigacts sigacts0; 189 struct proc proc0 = { 190 .p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps), 191 .p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters), 192 .p_nlwps = 1, 193 .p_nrlwps = 1, 194 .p_nlwpid = 1, /* must match lwp0.l_lid */ 195 .p_pgrp = &pgrp0, 196 .p_comm = "system", 197 /* 198 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8) 199 * when they exit. init(8) can easily wait them out for us. 200 */ 201 .p_flag = PK_SYSTEM | PK_NOCLDWAIT, 202 .p_stat = SACTIVE, 203 .p_nice = NZERO, 204 .p_emul = &emul_netbsd, 205 .p_cwdi = &cwdi0, 206 .p_limit = &limit0, 207 .p_fd = &filedesc0, 208 .p_vmspace = &vmspace0, 209 .p_stats = &pstat0, 210 .p_sigacts = &sigacts0, 211 #ifdef PROC0_MD_INITIALIZERS 212 PROC0_MD_INITIALIZERS 213 #endif 214 }; 215 kauth_cred_t cred0; 216 217 static const int nofile = NOFILE; 218 static const int maxuprc = MAXUPRC; 219 220 static int sysctl_doeproc(SYSCTLFN_PROTO); 221 static int sysctl_kern_proc_args(SYSCTLFN_PROTO); 222 223 /* 224 * The process list descriptors, used during pid allocation and 225 * by sysctl. No locking on this data structure is needed since 226 * it is completely static. 227 */ 228 const struct proclist_desc proclists[] = { 229 { &allproc }, 230 { &zombproc }, 231 { NULL }, 232 }; 233 234 static struct pgrp * pg_remove(pid_t); 235 static void pg_delete(pid_t); 236 static void orphanpg(struct pgrp *); 237 238 static specificdata_domain_t proc_specificdata_domain; 239 240 static pool_cache_t proc_cache; 241 242 static kauth_listener_t proc_listener; 243 244 static int fill_pathname(struct lwp *, pid_t, void *, size_t *); 245 246 static int 247 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 248 void *arg0, void *arg1, void *arg2, void *arg3) 249 { 250 struct proc *p; 251 int result; 252 253 result = KAUTH_RESULT_DEFER; 254 p = arg0; 255 256 switch (action) { 257 case KAUTH_PROCESS_CANSEE: { 258 enum kauth_process_req req; 259 260 req = (enum kauth_process_req)arg1; 261 262 switch (req) { 263 case KAUTH_REQ_PROCESS_CANSEE_ARGS: 264 case KAUTH_REQ_PROCESS_CANSEE_ENTRY: 265 case KAUTH_REQ_PROCESS_CANSEE_OPENFILES: 266 case KAUTH_REQ_PROCESS_CANSEE_EPROC: 267 result = KAUTH_RESULT_ALLOW; 268 break; 269 270 case KAUTH_REQ_PROCESS_CANSEE_ENV: 271 if (kauth_cred_getuid(cred) != 272 kauth_cred_getuid(p->p_cred) || 273 kauth_cred_getuid(cred) != 274 kauth_cred_getsvuid(p->p_cred)) 275 break; 276 277 result = KAUTH_RESULT_ALLOW; 278 279 break; 280 281 case KAUTH_REQ_PROCESS_CANSEE_KPTR: 282 default: 283 break; 284 } 285 286 break; 287 } 288 289 case KAUTH_PROCESS_FORK: { 290 int lnprocs = (int)(unsigned long)arg2; 291 292 /* 293 * Don't allow a nonprivileged user to use the last few 294 * processes. The variable lnprocs is the current number of 295 * processes, maxproc is the limit. 296 */ 297 if (__predict_false((lnprocs >= maxproc - 5))) 298 break; 299 300 result = KAUTH_RESULT_ALLOW; 301 302 break; 303 } 304 305 case KAUTH_PROCESS_CORENAME: 306 case KAUTH_PROCESS_STOPFLAG: 307 if (proc_uidmatch(cred, p->p_cred) == 0) 308 result = KAUTH_RESULT_ALLOW; 309 310 break; 311 312 default: 313 break; 314 } 315 316 return result; 317 } 318 319 /* 320 * Initialize global process hashing structures. 321 */ 322 void 323 procinit(void) 324 { 325 const struct proclist_desc *pd; 326 u_int i; 327 #define LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1)) 328 329 for (pd = proclists; pd->pd_list != NULL; pd++) 330 LIST_INIT(pd->pd_list); 331 332 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 333 pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE 334 * sizeof(struct pid_table), KM_SLEEP); 335 pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1; 336 pid_max = PID_MAX; 337 338 /* Set free list running through table... 339 Preset 'use count' above PID_MAX so we allocate pid 1 next. */ 340 for (i = 0; i <= pid_tbl_mask; i++) { 341 pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1); 342 pid_table[i].pt_pgrp = 0; 343 pid_table[i].pt_pid = 0; 344 } 345 /* slot 0 is just grabbed */ 346 next_free_pt = 1; 347 /* Need to fix last entry. */ 348 last_free_pt = pid_tbl_mask; 349 pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY); 350 /* point at which we grow table - to avoid reusing pids too often */ 351 pid_alloc_lim = pid_tbl_mask - 1; 352 #undef LINK_EMPTY 353 354 proc_specificdata_domain = specificdata_domain_create(); 355 KASSERT(proc_specificdata_domain != NULL); 356 357 proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0, 358 "procpl", NULL, IPL_NONE, NULL, NULL, NULL); 359 360 proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 361 proc_listener_cb, NULL); 362 } 363 364 void 365 procinit_sysctl(void) 366 { 367 static struct sysctllog *clog; 368 369 sysctl_createv(&clog, 0, NULL, NULL, 370 CTLFLAG_PERMANENT, 371 CTLTYPE_NODE, "proc", 372 SYSCTL_DESCR("System-wide process information"), 373 sysctl_doeproc, 0, NULL, 0, 374 CTL_KERN, KERN_PROC, CTL_EOL); 375 sysctl_createv(&clog, 0, NULL, NULL, 376 CTLFLAG_PERMANENT, 377 CTLTYPE_NODE, "proc2", 378 SYSCTL_DESCR("Machine-independent process information"), 379 sysctl_doeproc, 0, NULL, 0, 380 CTL_KERN, KERN_PROC2, CTL_EOL); 381 sysctl_createv(&clog, 0, NULL, NULL, 382 CTLFLAG_PERMANENT, 383 CTLTYPE_NODE, "proc_args", 384 SYSCTL_DESCR("Process argument information"), 385 sysctl_kern_proc_args, 0, NULL, 0, 386 CTL_KERN, KERN_PROC_ARGS, CTL_EOL); 387 388 /* 389 "nodes" under these: 390 391 KERN_PROC_ALL 392 KERN_PROC_PID pid 393 KERN_PROC_PGRP pgrp 394 KERN_PROC_SESSION sess 395 KERN_PROC_TTY tty 396 KERN_PROC_UID uid 397 KERN_PROC_RUID uid 398 KERN_PROC_GID gid 399 KERN_PROC_RGID gid 400 401 all in all, probably not worth the effort... 402 */ 403 } 404 405 /* 406 * Initialize process 0. 407 */ 408 void 409 proc0_init(void) 410 { 411 struct proc *p; 412 struct pgrp *pg; 413 struct rlimit *rlim; 414 rlim_t lim; 415 int i; 416 417 p = &proc0; 418 pg = &pgrp0; 419 420 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH); 421 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE); 422 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 423 424 rw_init(&p->p_reflock); 425 cv_init(&p->p_waitcv, "wait"); 426 cv_init(&p->p_lwpcv, "lwpwait"); 427 428 LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling); 429 430 pid_table[0].pt_proc = p; 431 LIST_INSERT_HEAD(&allproc, p, p_list); 432 433 pid_table[0].pt_pgrp = pg; 434 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist); 435 436 #ifdef __HAVE_SYSCALL_INTERN 437 (*p->p_emul->e_syscall_intern)(p); 438 #endif 439 440 /* Create credentials. */ 441 cred0 = kauth_cred_alloc(); 442 p->p_cred = cred0; 443 444 /* Create the CWD info. */ 445 rw_init(&cwdi0.cwdi_lock); 446 447 /* Create the limits structures. */ 448 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE); 449 450 rlim = limit0.pl_rlimit; 451 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) { 452 rlim[i].rlim_cur = RLIM_INFINITY; 453 rlim[i].rlim_max = RLIM_INFINITY; 454 } 455 456 rlim[RLIMIT_NOFILE].rlim_max = maxfiles; 457 rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile; 458 459 rlim[RLIMIT_NPROC].rlim_max = maxproc; 460 rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc; 461 462 lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free)); 463 rlim[RLIMIT_RSS].rlim_max = lim; 464 rlim[RLIMIT_MEMLOCK].rlim_max = lim; 465 rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3; 466 467 rlim[RLIMIT_NTHR].rlim_max = maxlwp; 468 rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc; 469 470 /* Note that default core name has zero length. */ 471 limit0.pl_corename = defcorename; 472 limit0.pl_cnlen = 0; 473 limit0.pl_refcnt = 1; 474 limit0.pl_writeable = false; 475 limit0.pl_sv_limit = NULL; 476 477 /* Configure virtual memory system, set vm rlimits. */ 478 uvm_init_limits(p); 479 480 /* Initialize file descriptor table for proc0. */ 481 fd_init(&filedesc0); 482 483 /* 484 * Initialize proc0's vmspace, which uses the kernel pmap. 485 * All kernel processes (which never have user space mappings) 486 * share proc0's vmspace, and thus, the kernel pmap. 487 */ 488 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS), 489 trunc_page(VM_MAXUSER_ADDRESS), 490 #ifdef __USE_TOPDOWN_VM 491 true 492 #else 493 false 494 #endif 495 ); 496 497 /* Initialize signal state for proc0. XXX IPL_SCHED */ 498 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED); 499 siginit(p); 500 501 proc_initspecific(p); 502 kdtrace_proc_ctor(NULL, p); 503 } 504 505 /* 506 * Session reference counting. 507 */ 508 509 void 510 proc_sesshold(struct session *ss) 511 { 512 513 KASSERT(mutex_owned(proc_lock)); 514 ss->s_count++; 515 } 516 517 void 518 proc_sessrele(struct session *ss) 519 { 520 521 KASSERT(mutex_owned(proc_lock)); 522 /* 523 * We keep the pgrp with the same id as the session in order to 524 * stop a process being given the same pid. Since the pgrp holds 525 * a reference to the session, it must be a 'zombie' pgrp by now. 526 */ 527 if (--ss->s_count == 0) { 528 struct pgrp *pg; 529 530 pg = pg_remove(ss->s_sid); 531 mutex_exit(proc_lock); 532 533 kmem_free(pg, sizeof(struct pgrp)); 534 kmem_free(ss, sizeof(struct session)); 535 } else { 536 mutex_exit(proc_lock); 537 } 538 } 539 540 /* 541 * Check that the specified process group is in the session of the 542 * specified process. 543 * Treats -ve ids as process ids. 544 * Used to validate TIOCSPGRP requests. 545 */ 546 int 547 pgid_in_session(struct proc *p, pid_t pg_id) 548 { 549 struct pgrp *pgrp; 550 struct session *session; 551 int error; 552 553 mutex_enter(proc_lock); 554 if (pg_id < 0) { 555 struct proc *p1 = proc_find(-pg_id); 556 if (p1 == NULL) { 557 error = EINVAL; 558 goto fail; 559 } 560 pgrp = p1->p_pgrp; 561 } else { 562 pgrp = pgrp_find(pg_id); 563 if (pgrp == NULL) { 564 error = EINVAL; 565 goto fail; 566 } 567 } 568 session = pgrp->pg_session; 569 error = (session != p->p_pgrp->pg_session) ? EPERM : 0; 570 fail: 571 mutex_exit(proc_lock); 572 return error; 573 } 574 575 /* 576 * p_inferior: is p an inferior of q? 577 */ 578 static inline bool 579 p_inferior(struct proc *p, struct proc *q) 580 { 581 582 KASSERT(mutex_owned(proc_lock)); 583 584 for (; p != q; p = p->p_pptr) 585 if (p->p_pid == 0) 586 return false; 587 return true; 588 } 589 590 /* 591 * proc_find: locate a process by the ID. 592 * 593 * => Must be called with proc_lock held. 594 */ 595 proc_t * 596 proc_find_raw(pid_t pid) 597 { 598 struct pid_table *pt; 599 proc_t *p; 600 601 KASSERT(mutex_owned(proc_lock)); 602 pt = &pid_table[pid & pid_tbl_mask]; 603 p = pt->pt_proc; 604 if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) { 605 return NULL; 606 } 607 return p; 608 } 609 610 proc_t * 611 proc_find(pid_t pid) 612 { 613 proc_t *p; 614 615 p = proc_find_raw(pid); 616 if (__predict_false(p == NULL)) { 617 return NULL; 618 } 619 620 /* 621 * Only allow live processes to be found by PID. 622 * XXX: p_stat might change, since unlocked. 623 */ 624 if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) { 625 return p; 626 } 627 return NULL; 628 } 629 630 /* 631 * pgrp_find: locate a process group by the ID. 632 * 633 * => Must be called with proc_lock held. 634 */ 635 struct pgrp * 636 pgrp_find(pid_t pgid) 637 { 638 struct pgrp *pg; 639 640 KASSERT(mutex_owned(proc_lock)); 641 642 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp; 643 644 /* 645 * Cannot look up a process group that only exists because the 646 * session has not died yet (traditional). 647 */ 648 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) { 649 return NULL; 650 } 651 return pg; 652 } 653 654 static void 655 expand_pid_table(void) 656 { 657 size_t pt_size, tsz; 658 struct pid_table *n_pt, *new_pt; 659 struct proc *proc; 660 struct pgrp *pgrp; 661 pid_t pid, rpid; 662 u_int i; 663 uint new_pt_mask; 664 665 pt_size = pid_tbl_mask + 1; 666 tsz = pt_size * 2 * sizeof(struct pid_table); 667 new_pt = kmem_alloc(tsz, KM_SLEEP); 668 new_pt_mask = pt_size * 2 - 1; 669 670 mutex_enter(proc_lock); 671 if (pt_size != pid_tbl_mask + 1) { 672 /* Another process beat us to it... */ 673 mutex_exit(proc_lock); 674 kmem_free(new_pt, tsz); 675 return; 676 } 677 678 /* 679 * Copy entries from old table into new one. 680 * If 'pid' is 'odd' we need to place in the upper half, 681 * even pid's to the lower half. 682 * Free items stay in the low half so we don't have to 683 * fixup the reference to them. 684 * We stuff free items on the front of the freelist 685 * because we can't write to unmodified entries. 686 * Processing the table backwards maintains a semblance 687 * of issuing pid numbers that increase with time. 688 */ 689 i = pt_size - 1; 690 n_pt = new_pt + i; 691 for (; ; i--, n_pt--) { 692 proc = pid_table[i].pt_proc; 693 pgrp = pid_table[i].pt_pgrp; 694 if (!P_VALID(proc)) { 695 /* Up 'use count' so that link is valid */ 696 pid = (P_NEXT(proc) + pt_size) & ~pt_size; 697 rpid = 0; 698 proc = P_FREE(pid); 699 if (pgrp) 700 pid = pgrp->pg_id; 701 } else { 702 pid = pid_table[i].pt_pid; 703 rpid = pid; 704 } 705 706 /* Save entry in appropriate half of table */ 707 n_pt[pid & pt_size].pt_proc = proc; 708 n_pt[pid & pt_size].pt_pgrp = pgrp; 709 n_pt[pid & pt_size].pt_pid = rpid; 710 711 /* Put other piece on start of free list */ 712 pid = (pid ^ pt_size) & ~pid_tbl_mask; 713 n_pt[pid & pt_size].pt_proc = 714 P_FREE((pid & ~pt_size) | next_free_pt); 715 n_pt[pid & pt_size].pt_pgrp = 0; 716 n_pt[pid & pt_size].pt_pid = 0; 717 718 next_free_pt = i | (pid & pt_size); 719 if (i == 0) 720 break; 721 } 722 723 /* Save old table size and switch tables */ 724 tsz = pt_size * sizeof(struct pid_table); 725 n_pt = pid_table; 726 pid_table = new_pt; 727 pid_tbl_mask = new_pt_mask; 728 729 /* 730 * pid_max starts as PID_MAX (= 30000), once we have 16384 731 * allocated pids we need it to be larger! 732 */ 733 if (pid_tbl_mask > PID_MAX) { 734 pid_max = pid_tbl_mask * 2 + 1; 735 pid_alloc_lim |= pid_alloc_lim << 1; 736 } else 737 pid_alloc_lim <<= 1; /* doubles number of free slots... */ 738 739 mutex_exit(proc_lock); 740 kmem_free(n_pt, tsz); 741 } 742 743 struct proc * 744 proc_alloc(void) 745 { 746 struct proc *p; 747 748 p = pool_cache_get(proc_cache, PR_WAITOK); 749 p->p_stat = SIDL; /* protect against others */ 750 proc_initspecific(p); 751 kdtrace_proc_ctor(NULL, p); 752 p->p_pid = -1; 753 proc_alloc_pid(p); 754 return p; 755 } 756 757 /* 758 * proc_alloc_pid: allocate PID and record the given proc 'p' so that 759 * proc_find_raw() can find it by the PID. 760 */ 761 762 pid_t 763 proc_alloc_pid(struct proc *p) 764 { 765 struct pid_table *pt; 766 pid_t pid; 767 int nxt; 768 769 for (;;expand_pid_table()) { 770 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) 771 /* ensure pids cycle through 2000+ values */ 772 continue; 773 mutex_enter(proc_lock); 774 pt = &pid_table[next_free_pt]; 775 #ifdef DIAGNOSTIC 776 if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp)) 777 panic("proc_alloc: slot busy"); 778 #endif 779 nxt = P_NEXT(pt->pt_proc); 780 if (nxt & pid_tbl_mask) 781 break; 782 /* Table full - expand (NB last entry not used....) */ 783 mutex_exit(proc_lock); 784 } 785 786 /* pid is 'saved use count' + 'size' + entry */ 787 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt; 788 if ((uint)pid > (uint)pid_max) 789 pid &= pid_tbl_mask; 790 next_free_pt = nxt & pid_tbl_mask; 791 792 /* Grab table slot */ 793 pt->pt_proc = p; 794 795 KASSERT(pt->pt_pid == 0); 796 pt->pt_pid = pid; 797 if (p->p_pid == -1) { 798 p->p_pid = pid; 799 } 800 pid_alloc_cnt++; 801 mutex_exit(proc_lock); 802 803 return pid; 804 } 805 806 /* 807 * Free a process id - called from proc_free (in kern_exit.c) 808 * 809 * Called with the proc_lock held. 810 */ 811 void 812 proc_free_pid(pid_t pid) 813 { 814 struct pid_table *pt; 815 816 KASSERT(mutex_owned(proc_lock)); 817 818 pt = &pid_table[pid & pid_tbl_mask]; 819 820 /* save pid use count in slot */ 821 pt->pt_proc = P_FREE(pid & ~pid_tbl_mask); 822 KASSERT(pt->pt_pid == pid); 823 pt->pt_pid = 0; 824 825 if (pt->pt_pgrp == NULL) { 826 /* link last freed entry onto ours */ 827 pid &= pid_tbl_mask; 828 pt = &pid_table[last_free_pt]; 829 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid); 830 pt->pt_pid = 0; 831 last_free_pt = pid; 832 pid_alloc_cnt--; 833 } 834 835 atomic_dec_uint(&nprocs); 836 } 837 838 void 839 proc_free_mem(struct proc *p) 840 { 841 842 kdtrace_proc_dtor(NULL, p); 843 pool_cache_put(proc_cache, p); 844 } 845 846 /* 847 * proc_enterpgrp: move p to a new or existing process group (and session). 848 * 849 * If we are creating a new pgrp, the pgid should equal 850 * the calling process' pid. 851 * If is only valid to enter a process group that is in the session 852 * of the process. 853 * Also mksess should only be set if we are creating a process group 854 * 855 * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return. 856 */ 857 int 858 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess) 859 { 860 struct pgrp *new_pgrp, *pgrp; 861 struct session *sess; 862 struct proc *p; 863 int rval; 864 pid_t pg_id = NO_PGID; 865 866 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL; 867 868 /* Allocate data areas we might need before doing any validity checks */ 869 mutex_enter(proc_lock); /* Because pid_table might change */ 870 if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) { 871 mutex_exit(proc_lock); 872 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP); 873 mutex_enter(proc_lock); 874 } else 875 new_pgrp = NULL; 876 rval = EPERM; /* most common error (to save typing) */ 877 878 /* Check pgrp exists or can be created */ 879 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp; 880 if (pgrp != NULL && pgrp->pg_id != pgid) 881 goto done; 882 883 /* Can only set another process under restricted circumstances. */ 884 if (pid != curp->p_pid) { 885 /* Must exist and be one of our children... */ 886 p = proc_find(pid); 887 if (p == NULL || !p_inferior(p, curp)) { 888 rval = ESRCH; 889 goto done; 890 } 891 /* ... in the same session... */ 892 if (sess != NULL || p->p_session != curp->p_session) 893 goto done; 894 /* ... existing pgid must be in same session ... */ 895 if (pgrp != NULL && pgrp->pg_session != p->p_session) 896 goto done; 897 /* ... and not done an exec. */ 898 if (p->p_flag & PK_EXEC) { 899 rval = EACCES; 900 goto done; 901 } 902 } else { 903 /* ... setsid() cannot re-enter a pgrp */ 904 if (mksess && (curp->p_pgid == curp->p_pid || 905 pgrp_find(curp->p_pid))) 906 goto done; 907 p = curp; 908 } 909 910 /* Changing the process group/session of a session 911 leader is definitely off limits. */ 912 if (SESS_LEADER(p)) { 913 if (sess == NULL && p->p_pgrp == pgrp) 914 /* unless it's a definite noop */ 915 rval = 0; 916 goto done; 917 } 918 919 /* Can only create a process group with id of process */ 920 if (pgrp == NULL && pgid != pid) 921 goto done; 922 923 /* Can only create a session if creating pgrp */ 924 if (sess != NULL && pgrp != NULL) 925 goto done; 926 927 /* Check we allocated memory for a pgrp... */ 928 if (pgrp == NULL && new_pgrp == NULL) 929 goto done; 930 931 /* Don't attach to 'zombie' pgrp */ 932 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members)) 933 goto done; 934 935 /* Expect to succeed now */ 936 rval = 0; 937 938 if (pgrp == p->p_pgrp) 939 /* nothing to do */ 940 goto done; 941 942 /* Ok all setup, link up required structures */ 943 944 if (pgrp == NULL) { 945 pgrp = new_pgrp; 946 new_pgrp = NULL; 947 if (sess != NULL) { 948 sess->s_sid = p->p_pid; 949 sess->s_leader = p; 950 sess->s_count = 1; 951 sess->s_ttyvp = NULL; 952 sess->s_ttyp = NULL; 953 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET; 954 memcpy(sess->s_login, p->p_session->s_login, 955 sizeof(sess->s_login)); 956 p->p_lflag &= ~PL_CONTROLT; 957 } else { 958 sess = p->p_pgrp->pg_session; 959 proc_sesshold(sess); 960 } 961 pgrp->pg_session = sess; 962 sess = NULL; 963 964 pgrp->pg_id = pgid; 965 LIST_INIT(&pgrp->pg_members); 966 #ifdef DIAGNOSTIC 967 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp)) 968 panic("enterpgrp: pgrp table slot in use"); 969 if (__predict_false(mksess && p != curp)) 970 panic("enterpgrp: mksession and p != curproc"); 971 #endif 972 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp; 973 pgrp->pg_jobc = 0; 974 } 975 976 /* 977 * Adjust eligibility of affected pgrps to participate in job control. 978 * Increment eligibility counts before decrementing, otherwise we 979 * could reach 0 spuriously during the first call. 980 */ 981 fixjobc(p, pgrp, 1); 982 fixjobc(p, p->p_pgrp, 0); 983 984 /* Interlock with ttread(). */ 985 mutex_spin_enter(&tty_lock); 986 987 /* Move process to requested group. */ 988 LIST_REMOVE(p, p_pglist); 989 if (LIST_EMPTY(&p->p_pgrp->pg_members)) 990 /* defer delete until we've dumped the lock */ 991 pg_id = p->p_pgrp->pg_id; 992 p->p_pgrp = pgrp; 993 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist); 994 995 /* Done with the swap; we can release the tty mutex. */ 996 mutex_spin_exit(&tty_lock); 997 998 done: 999 if (pg_id != NO_PGID) { 1000 /* Releases proc_lock. */ 1001 pg_delete(pg_id); 1002 } else { 1003 mutex_exit(proc_lock); 1004 } 1005 if (sess != NULL) 1006 kmem_free(sess, sizeof(*sess)); 1007 if (new_pgrp != NULL) 1008 kmem_free(new_pgrp, sizeof(*new_pgrp)); 1009 #ifdef DEBUG_PGRP 1010 if (__predict_false(rval)) 1011 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n", 1012 pid, pgid, mksess, curp->p_pid, rval); 1013 #endif 1014 return rval; 1015 } 1016 1017 /* 1018 * proc_leavepgrp: remove a process from its process group. 1019 * => must be called with the proc_lock held, which will be released; 1020 */ 1021 void 1022 proc_leavepgrp(struct proc *p) 1023 { 1024 struct pgrp *pgrp; 1025 1026 KASSERT(mutex_owned(proc_lock)); 1027 1028 /* Interlock with ttread() */ 1029 mutex_spin_enter(&tty_lock); 1030 pgrp = p->p_pgrp; 1031 LIST_REMOVE(p, p_pglist); 1032 p->p_pgrp = NULL; 1033 mutex_spin_exit(&tty_lock); 1034 1035 if (LIST_EMPTY(&pgrp->pg_members)) { 1036 /* Releases proc_lock. */ 1037 pg_delete(pgrp->pg_id); 1038 } else { 1039 mutex_exit(proc_lock); 1040 } 1041 } 1042 1043 /* 1044 * pg_remove: remove a process group from the table. 1045 * => must be called with the proc_lock held; 1046 * => returns process group to free; 1047 */ 1048 static struct pgrp * 1049 pg_remove(pid_t pg_id) 1050 { 1051 struct pgrp *pgrp; 1052 struct pid_table *pt; 1053 1054 KASSERT(mutex_owned(proc_lock)); 1055 1056 pt = &pid_table[pg_id & pid_tbl_mask]; 1057 pgrp = pt->pt_pgrp; 1058 1059 KASSERT(pgrp != NULL); 1060 KASSERT(pgrp->pg_id == pg_id); 1061 KASSERT(LIST_EMPTY(&pgrp->pg_members)); 1062 1063 pt->pt_pgrp = NULL; 1064 1065 if (!P_VALID(pt->pt_proc)) { 1066 /* Orphaned pgrp, put slot onto free list. */ 1067 KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0); 1068 pg_id &= pid_tbl_mask; 1069 pt = &pid_table[last_free_pt]; 1070 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id); 1071 KASSERT(pt->pt_pid == 0); 1072 last_free_pt = pg_id; 1073 pid_alloc_cnt--; 1074 } 1075 return pgrp; 1076 } 1077 1078 /* 1079 * pg_delete: delete and free a process group. 1080 * => must be called with the proc_lock held, which will be released. 1081 */ 1082 static void 1083 pg_delete(pid_t pg_id) 1084 { 1085 struct pgrp *pg; 1086 struct tty *ttyp; 1087 struct session *ss; 1088 1089 KASSERT(mutex_owned(proc_lock)); 1090 1091 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp; 1092 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) { 1093 mutex_exit(proc_lock); 1094 return; 1095 } 1096 1097 ss = pg->pg_session; 1098 1099 /* Remove reference (if any) from tty to this process group */ 1100 mutex_spin_enter(&tty_lock); 1101 ttyp = ss->s_ttyp; 1102 if (ttyp != NULL && ttyp->t_pgrp == pg) { 1103 ttyp->t_pgrp = NULL; 1104 KASSERT(ttyp->t_session == ss); 1105 } 1106 mutex_spin_exit(&tty_lock); 1107 1108 /* 1109 * The leading process group in a session is freed by proc_sessrele(), 1110 * if last reference. Note: proc_sessrele() releases proc_lock. 1111 */ 1112 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL; 1113 proc_sessrele(ss); 1114 1115 if (pg != NULL) { 1116 /* Free it, if was not done by proc_sessrele(). */ 1117 kmem_free(pg, sizeof(struct pgrp)); 1118 } 1119 } 1120 1121 /* 1122 * Adjust pgrp jobc counters when specified process changes process group. 1123 * We count the number of processes in each process group that "qualify" 1124 * the group for terminal job control (those with a parent in a different 1125 * process group of the same session). If that count reaches zero, the 1126 * process group becomes orphaned. Check both the specified process' 1127 * process group and that of its children. 1128 * entering == 0 => p is leaving specified group. 1129 * entering == 1 => p is entering specified group. 1130 * 1131 * Call with proc_lock held. 1132 */ 1133 void 1134 fixjobc(struct proc *p, struct pgrp *pgrp, int entering) 1135 { 1136 struct pgrp *hispgrp; 1137 struct session *mysession = pgrp->pg_session; 1138 struct proc *child; 1139 1140 KASSERT(mutex_owned(proc_lock)); 1141 1142 /* 1143 * Check p's parent to see whether p qualifies its own process 1144 * group; if so, adjust count for p's process group. 1145 */ 1146 hispgrp = p->p_pptr->p_pgrp; 1147 if (hispgrp != pgrp && hispgrp->pg_session == mysession) { 1148 if (entering) { 1149 pgrp->pg_jobc++; 1150 p->p_lflag &= ~PL_ORPHANPG; 1151 } else if (--pgrp->pg_jobc == 0) 1152 orphanpg(pgrp); 1153 } 1154 1155 /* 1156 * Check this process' children to see whether they qualify 1157 * their process groups; if so, adjust counts for children's 1158 * process groups. 1159 */ 1160 LIST_FOREACH(child, &p->p_children, p_sibling) { 1161 hispgrp = child->p_pgrp; 1162 if (hispgrp != pgrp && hispgrp->pg_session == mysession && 1163 !P_ZOMBIE(child)) { 1164 if (entering) { 1165 child->p_lflag &= ~PL_ORPHANPG; 1166 hispgrp->pg_jobc++; 1167 } else if (--hispgrp->pg_jobc == 0) 1168 orphanpg(hispgrp); 1169 } 1170 } 1171 } 1172 1173 /* 1174 * A process group has become orphaned; 1175 * if there are any stopped processes in the group, 1176 * hang-up all process in that group. 1177 * 1178 * Call with proc_lock held. 1179 */ 1180 static void 1181 orphanpg(struct pgrp *pg) 1182 { 1183 struct proc *p; 1184 1185 KASSERT(mutex_owned(proc_lock)); 1186 1187 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1188 if (p->p_stat == SSTOP) { 1189 p->p_lflag |= PL_ORPHANPG; 1190 psignal(p, SIGHUP); 1191 psignal(p, SIGCONT); 1192 } 1193 } 1194 } 1195 1196 #ifdef DDB 1197 #include <ddb/db_output.h> 1198 void pidtbl_dump(void); 1199 void 1200 pidtbl_dump(void) 1201 { 1202 struct pid_table *pt; 1203 struct proc *p; 1204 struct pgrp *pgrp; 1205 int id; 1206 1207 db_printf("pid table %p size %x, next %x, last %x\n", 1208 pid_table, pid_tbl_mask+1, 1209 next_free_pt, last_free_pt); 1210 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) { 1211 p = pt->pt_proc; 1212 if (!P_VALID(p) && !pt->pt_pgrp) 1213 continue; 1214 db_printf(" id %x: ", id); 1215 if (P_VALID(p)) 1216 db_printf("slotpid %d proc %p id %d (0x%x) %s\n", 1217 pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm); 1218 else 1219 db_printf("next %x use %x\n", 1220 P_NEXT(p) & pid_tbl_mask, 1221 P_NEXT(p) & ~pid_tbl_mask); 1222 if ((pgrp = pt->pt_pgrp)) { 1223 db_printf("\tsession %p, sid %d, count %d, login %s\n", 1224 pgrp->pg_session, pgrp->pg_session->s_sid, 1225 pgrp->pg_session->s_count, 1226 pgrp->pg_session->s_login); 1227 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n", 1228 pgrp, pgrp->pg_id, pgrp->pg_jobc, 1229 LIST_FIRST(&pgrp->pg_members)); 1230 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) { 1231 db_printf("\t\tpid %d addr %p pgrp %p %s\n", 1232 p->p_pid, p, p->p_pgrp, p->p_comm); 1233 } 1234 } 1235 } 1236 } 1237 #endif /* DDB */ 1238 1239 #ifdef KSTACK_CHECK_MAGIC 1240 1241 #define KSTACK_MAGIC 0xdeadbeaf 1242 1243 /* XXX should be per process basis? */ 1244 static int kstackleftmin = KSTACK_SIZE; 1245 static int kstackleftthres = KSTACK_SIZE / 8; 1246 1247 void 1248 kstack_setup_magic(const struct lwp *l) 1249 { 1250 uint32_t *ip; 1251 uint32_t const *end; 1252 1253 KASSERT(l != NULL); 1254 KASSERT(l != &lwp0); 1255 1256 /* 1257 * fill all the stack with magic number 1258 * so that later modification on it can be detected. 1259 */ 1260 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1261 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1262 for (; ip < end; ip++) { 1263 *ip = KSTACK_MAGIC; 1264 } 1265 } 1266 1267 void 1268 kstack_check_magic(const struct lwp *l) 1269 { 1270 uint32_t const *ip, *end; 1271 int stackleft; 1272 1273 KASSERT(l != NULL); 1274 1275 /* don't check proc0 */ /*XXX*/ 1276 if (l == &lwp0) 1277 return; 1278 1279 #ifdef __MACHINE_STACK_GROWS_UP 1280 /* stack grows upwards (eg. hppa) */ 1281 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1282 end = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1283 for (ip--; ip >= end; ip--) 1284 if (*ip != KSTACK_MAGIC) 1285 break; 1286 1287 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip; 1288 #else /* __MACHINE_STACK_GROWS_UP */ 1289 /* stack grows downwards (eg. i386) */ 1290 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l); 1291 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE); 1292 for (; ip < end; ip++) 1293 if (*ip != KSTACK_MAGIC) 1294 break; 1295 1296 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l); 1297 #endif /* __MACHINE_STACK_GROWS_UP */ 1298 1299 if (kstackleftmin > stackleft) { 1300 kstackleftmin = stackleft; 1301 if (stackleft < kstackleftthres) 1302 printf("warning: kernel stack left %d bytes" 1303 "(pid %u:lid %u)\n", stackleft, 1304 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1305 } 1306 1307 if (stackleft <= 0) { 1308 panic("magic on the top of kernel stack changed for " 1309 "pid %u, lid %u: maybe kernel stack overflow", 1310 (u_int)l->l_proc->p_pid, (u_int)l->l_lid); 1311 } 1312 } 1313 #endif /* KSTACK_CHECK_MAGIC */ 1314 1315 int 1316 proclist_foreach_call(struct proclist *list, 1317 int (*callback)(struct proc *, void *arg), void *arg) 1318 { 1319 struct proc marker; 1320 struct proc *p; 1321 int ret = 0; 1322 1323 marker.p_flag = PK_MARKER; 1324 mutex_enter(proc_lock); 1325 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) { 1326 if (p->p_flag & PK_MARKER) { 1327 p = LIST_NEXT(p, p_list); 1328 continue; 1329 } 1330 LIST_INSERT_AFTER(p, &marker, p_list); 1331 ret = (*callback)(p, arg); 1332 KASSERT(mutex_owned(proc_lock)); 1333 p = LIST_NEXT(&marker, p_list); 1334 LIST_REMOVE(&marker, p_list); 1335 } 1336 mutex_exit(proc_lock); 1337 1338 return ret; 1339 } 1340 1341 int 1342 proc_vmspace_getref(struct proc *p, struct vmspace **vm) 1343 { 1344 1345 /* XXXCDC: how should locking work here? */ 1346 1347 /* curproc exception is for coredump. */ 1348 1349 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) || 1350 (p->p_vmspace->vm_refcnt < 1)) { /* XXX */ 1351 return EFAULT; 1352 } 1353 1354 uvmspace_addref(p->p_vmspace); 1355 *vm = p->p_vmspace; 1356 1357 return 0; 1358 } 1359 1360 /* 1361 * Acquire a write lock on the process credential. 1362 */ 1363 void 1364 proc_crmod_enter(void) 1365 { 1366 struct lwp *l = curlwp; 1367 struct proc *p = l->l_proc; 1368 kauth_cred_t oc; 1369 1370 /* Reset what needs to be reset in plimit. */ 1371 if (p->p_limit->pl_corename != defcorename) { 1372 lim_setcorename(p, defcorename, 0); 1373 } 1374 1375 mutex_enter(p->p_lock); 1376 1377 /* Ensure the LWP cached credentials are up to date. */ 1378 if ((oc = l->l_cred) != p->p_cred) { 1379 kauth_cred_hold(p->p_cred); 1380 l->l_cred = p->p_cred; 1381 kauth_cred_free(oc); 1382 } 1383 } 1384 1385 /* 1386 * Set in a new process credential, and drop the write lock. The credential 1387 * must have a reference already. Optionally, free a no-longer required 1388 * credential. The scheduler also needs to inspect p_cred, so we also 1389 * briefly acquire the sched state mutex. 1390 */ 1391 void 1392 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid) 1393 { 1394 struct lwp *l = curlwp, *l2; 1395 struct proc *p = l->l_proc; 1396 kauth_cred_t oc; 1397 1398 KASSERT(mutex_owned(p->p_lock)); 1399 1400 /* Is there a new credential to set in? */ 1401 if (scred != NULL) { 1402 p->p_cred = scred; 1403 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 1404 if (l2 != l) 1405 l2->l_prflag |= LPR_CRMOD; 1406 } 1407 1408 /* Ensure the LWP cached credentials are up to date. */ 1409 if ((oc = l->l_cred) != scred) { 1410 kauth_cred_hold(scred); 1411 l->l_cred = scred; 1412 } 1413 } else 1414 oc = NULL; /* XXXgcc */ 1415 1416 if (sugid) { 1417 /* 1418 * Mark process as having changed credentials, stops 1419 * tracing etc. 1420 */ 1421 p->p_flag |= PK_SUGID; 1422 } 1423 1424 mutex_exit(p->p_lock); 1425 1426 /* If there is a credential to be released, free it now. */ 1427 if (fcred != NULL) { 1428 KASSERT(scred != NULL); 1429 kauth_cred_free(fcred); 1430 if (oc != scred) 1431 kauth_cred_free(oc); 1432 } 1433 } 1434 1435 /* 1436 * proc_specific_key_create -- 1437 * Create a key for subsystem proc-specific data. 1438 */ 1439 int 1440 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor) 1441 { 1442 1443 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor)); 1444 } 1445 1446 /* 1447 * proc_specific_key_delete -- 1448 * Delete a key for subsystem proc-specific data. 1449 */ 1450 void 1451 proc_specific_key_delete(specificdata_key_t key) 1452 { 1453 1454 specificdata_key_delete(proc_specificdata_domain, key); 1455 } 1456 1457 /* 1458 * proc_initspecific -- 1459 * Initialize a proc's specificdata container. 1460 */ 1461 void 1462 proc_initspecific(struct proc *p) 1463 { 1464 int error __diagused; 1465 1466 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref); 1467 KASSERT(error == 0); 1468 } 1469 1470 /* 1471 * proc_finispecific -- 1472 * Finalize a proc's specificdata container. 1473 */ 1474 void 1475 proc_finispecific(struct proc *p) 1476 { 1477 1478 specificdata_fini(proc_specificdata_domain, &p->p_specdataref); 1479 } 1480 1481 /* 1482 * proc_getspecific -- 1483 * Return proc-specific data corresponding to the specified key. 1484 */ 1485 void * 1486 proc_getspecific(struct proc *p, specificdata_key_t key) 1487 { 1488 1489 return (specificdata_getspecific(proc_specificdata_domain, 1490 &p->p_specdataref, key)); 1491 } 1492 1493 /* 1494 * proc_setspecific -- 1495 * Set proc-specific data corresponding to the specified key. 1496 */ 1497 void 1498 proc_setspecific(struct proc *p, specificdata_key_t key, void *data) 1499 { 1500 1501 specificdata_setspecific(proc_specificdata_domain, 1502 &p->p_specdataref, key, data); 1503 } 1504 1505 int 1506 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target) 1507 { 1508 int r = 0; 1509 1510 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) || 1511 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) { 1512 /* 1513 * suid proc of ours or proc not ours 1514 */ 1515 r = EPERM; 1516 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) { 1517 /* 1518 * sgid proc has sgid back to us temporarily 1519 */ 1520 r = EPERM; 1521 } else { 1522 /* 1523 * our rgid must be in target's group list (ie, 1524 * sub-processes started by a sgid process) 1525 */ 1526 int ismember = 0; 1527 1528 if (kauth_cred_ismember_gid(cred, 1529 kauth_cred_getgid(target), &ismember) != 0 || 1530 !ismember) 1531 r = EPERM; 1532 } 1533 1534 return (r); 1535 } 1536 1537 /* 1538 * sysctl stuff 1539 */ 1540 1541 #define KERN_PROCSLOP (5 * sizeof(struct kinfo_proc)) 1542 1543 static const u_int sysctl_flagmap[] = { 1544 PK_ADVLOCK, P_ADVLOCK, 1545 PK_EXEC, P_EXEC, 1546 PK_NOCLDWAIT, P_NOCLDWAIT, 1547 PK_32, P_32, 1548 PK_CLDSIGIGN, P_CLDSIGIGN, 1549 PK_SUGID, P_SUGID, 1550 0 1551 }; 1552 1553 static const u_int sysctl_sflagmap[] = { 1554 PS_NOCLDSTOP, P_NOCLDSTOP, 1555 PS_WEXIT, P_WEXIT, 1556 PS_STOPFORK, P_STOPFORK, 1557 PS_STOPEXEC, P_STOPEXEC, 1558 PS_STOPEXIT, P_STOPEXIT, 1559 0 1560 }; 1561 1562 static const u_int sysctl_slflagmap[] = { 1563 PSL_TRACED, P_TRACED, 1564 PSL_CHTRACED, P_CHTRACED, 1565 PSL_SYSCALL, P_SYSCALL, 1566 0 1567 }; 1568 1569 static const u_int sysctl_lflagmap[] = { 1570 PL_CONTROLT, P_CONTROLT, 1571 PL_PPWAIT, P_PPWAIT, 1572 0 1573 }; 1574 1575 static const u_int sysctl_stflagmap[] = { 1576 PST_PROFIL, P_PROFIL, 1577 0 1578 1579 }; 1580 1581 /* used by kern_lwp also */ 1582 const u_int sysctl_lwpflagmap[] = { 1583 LW_SINTR, L_SINTR, 1584 LW_SYSTEM, L_SYSTEM, 1585 0 1586 }; 1587 1588 /* 1589 * Find the most ``active'' lwp of a process and return it for ps display 1590 * purposes 1591 */ 1592 static struct lwp * 1593 proc_active_lwp(struct proc *p) 1594 { 1595 static const int ostat[] = { 1596 0, 1597 2, /* LSIDL */ 1598 6, /* LSRUN */ 1599 5, /* LSSLEEP */ 1600 4, /* LSSTOP */ 1601 0, /* LSZOMB */ 1602 1, /* LSDEAD */ 1603 7, /* LSONPROC */ 1604 3 /* LSSUSPENDED */ 1605 }; 1606 1607 struct lwp *l, *lp = NULL; 1608 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 1609 KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat)); 1610 if (lp == NULL || 1611 ostat[l->l_stat] > ostat[lp->l_stat] || 1612 (ostat[l->l_stat] == ostat[lp->l_stat] && 1613 l->l_cpticks > lp->l_cpticks)) { 1614 lp = l; 1615 continue; 1616 } 1617 } 1618 return lp; 1619 } 1620 1621 static int 1622 sysctl_doeproc(SYSCTLFN_ARGS) 1623 { 1624 union { 1625 struct kinfo_proc kproc; 1626 struct kinfo_proc2 kproc2; 1627 } *kbuf; 1628 struct proc *p, *next, *marker; 1629 char *where, *dp; 1630 int type, op, arg, error; 1631 u_int elem_size, kelem_size, elem_count; 1632 size_t buflen, needed; 1633 bool match, zombie, mmmbrains; 1634 1635 if (namelen == 1 && name[0] == CTL_QUERY) 1636 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1637 1638 dp = where = oldp; 1639 buflen = where != NULL ? *oldlenp : 0; 1640 error = 0; 1641 needed = 0; 1642 type = rnode->sysctl_num; 1643 1644 if (type == KERN_PROC) { 1645 if (namelen == 0) 1646 return EINVAL; 1647 switch (op = name[0]) { 1648 case KERN_PROC_ALL: 1649 if (namelen != 1) 1650 return EINVAL; 1651 arg = 0; 1652 break; 1653 default: 1654 if (namelen != 2) 1655 return EINVAL; 1656 arg = name[1]; 1657 break; 1658 } 1659 elem_count = 0; /* Hush little compiler, don't you cry */ 1660 kelem_size = elem_size = sizeof(kbuf->kproc); 1661 } else { 1662 if (namelen != 4) 1663 return EINVAL; 1664 op = name[0]; 1665 arg = name[1]; 1666 elem_size = name[2]; 1667 elem_count = name[3]; 1668 kelem_size = sizeof(kbuf->kproc2); 1669 } 1670 1671 sysctl_unlock(); 1672 1673 kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP); 1674 marker = kmem_alloc(sizeof(*marker), KM_SLEEP); 1675 marker->p_flag = PK_MARKER; 1676 1677 mutex_enter(proc_lock); 1678 /* 1679 * Start with zombies to prevent reporting processes twice, in case they 1680 * are dying and being moved from the list of alive processes to zombies. 1681 */ 1682 mmmbrains = true; 1683 for (p = LIST_FIRST(&zombproc);; p = next) { 1684 if (p == NULL) { 1685 if (mmmbrains) { 1686 p = LIST_FIRST(&allproc); 1687 mmmbrains = false; 1688 } 1689 if (p == NULL) 1690 break; 1691 } 1692 next = LIST_NEXT(p, p_list); 1693 if ((p->p_flag & PK_MARKER) != 0) 1694 continue; 1695 1696 /* 1697 * Skip embryonic processes. 1698 */ 1699 if (p->p_stat == SIDL) 1700 continue; 1701 1702 mutex_enter(p->p_lock); 1703 error = kauth_authorize_process(l->l_cred, 1704 KAUTH_PROCESS_CANSEE, p, 1705 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL); 1706 if (error != 0) { 1707 mutex_exit(p->p_lock); 1708 continue; 1709 } 1710 1711 /* 1712 * Hande all the operations in one switch on the cost of 1713 * algorithm complexity is on purpose. The win splitting this 1714 * function into several similar copies makes maintenance burden 1715 * burden, code grow and boost is neglible in practical systems. 1716 */ 1717 switch (op) { 1718 case KERN_PROC_PID: 1719 match = (p->p_pid == (pid_t)arg); 1720 break; 1721 1722 case KERN_PROC_PGRP: 1723 match = (p->p_pgrp->pg_id == (pid_t)arg); 1724 break; 1725 1726 case KERN_PROC_SESSION: 1727 match = (p->p_session->s_sid == (pid_t)arg); 1728 break; 1729 1730 case KERN_PROC_TTY: 1731 match = true; 1732 if (arg == (int) KERN_PROC_TTY_REVOKE) { 1733 if ((p->p_lflag & PL_CONTROLT) == 0 || 1734 p->p_session->s_ttyp == NULL || 1735 p->p_session->s_ttyvp != NULL) { 1736 match = false; 1737 } 1738 } else if ((p->p_lflag & PL_CONTROLT) == 0 || 1739 p->p_session->s_ttyp == NULL) { 1740 if ((dev_t)arg != KERN_PROC_TTY_NODEV) { 1741 match = false; 1742 } 1743 } else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) { 1744 match = false; 1745 } 1746 break; 1747 1748 case KERN_PROC_UID: 1749 match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg); 1750 break; 1751 1752 case KERN_PROC_RUID: 1753 match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg); 1754 break; 1755 1756 case KERN_PROC_GID: 1757 match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg); 1758 break; 1759 1760 case KERN_PROC_RGID: 1761 match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg); 1762 break; 1763 1764 case KERN_PROC_ALL: 1765 match = true; 1766 /* allow everything */ 1767 break; 1768 1769 default: 1770 error = EINVAL; 1771 mutex_exit(p->p_lock); 1772 goto cleanup; 1773 } 1774 if (!match) { 1775 mutex_exit(p->p_lock); 1776 continue; 1777 } 1778 1779 /* 1780 * Grab a hold on the process. 1781 */ 1782 if (mmmbrains) { 1783 zombie = true; 1784 } else { 1785 zombie = !rw_tryenter(&p->p_reflock, RW_READER); 1786 } 1787 if (zombie) { 1788 LIST_INSERT_AFTER(p, marker, p_list); 1789 } 1790 1791 if (buflen >= elem_size && 1792 (type == KERN_PROC || elem_count > 0)) { 1793 if (type == KERN_PROC) { 1794 kbuf->kproc.kp_proc = *p; 1795 fill_eproc(p, &kbuf->kproc.kp_eproc, zombie); 1796 } else { 1797 fill_kproc2(p, &kbuf->kproc2, zombie); 1798 elem_count--; 1799 } 1800 mutex_exit(p->p_lock); 1801 mutex_exit(proc_lock); 1802 /* 1803 * Copy out elem_size, but not larger than kelem_size 1804 */ 1805 error = sysctl_copyout(l, kbuf, dp, 1806 uimin(kelem_size, elem_size)); 1807 mutex_enter(proc_lock); 1808 if (error) { 1809 goto bah; 1810 } 1811 dp += elem_size; 1812 buflen -= elem_size; 1813 } else { 1814 mutex_exit(p->p_lock); 1815 } 1816 needed += elem_size; 1817 1818 /* 1819 * Release reference to process. 1820 */ 1821 if (zombie) { 1822 next = LIST_NEXT(marker, p_list); 1823 LIST_REMOVE(marker, p_list); 1824 } else { 1825 rw_exit(&p->p_reflock); 1826 next = LIST_NEXT(p, p_list); 1827 } 1828 1829 /* 1830 * Short-circuit break quickly! 1831 */ 1832 if (op == KERN_PROC_PID) 1833 break; 1834 } 1835 mutex_exit(proc_lock); 1836 1837 if (where != NULL) { 1838 *oldlenp = dp - where; 1839 if (needed > *oldlenp) { 1840 error = ENOMEM; 1841 goto out; 1842 } 1843 } else { 1844 needed += KERN_PROCSLOP; 1845 *oldlenp = needed; 1846 } 1847 kmem_free(kbuf, sizeof(*kbuf)); 1848 kmem_free(marker, sizeof(*marker)); 1849 sysctl_relock(); 1850 return 0; 1851 bah: 1852 if (zombie) 1853 LIST_REMOVE(marker, p_list); 1854 else 1855 rw_exit(&p->p_reflock); 1856 cleanup: 1857 mutex_exit(proc_lock); 1858 out: 1859 kmem_free(kbuf, sizeof(*kbuf)); 1860 kmem_free(marker, sizeof(*marker)); 1861 sysctl_relock(); 1862 return error; 1863 } 1864 1865 int 1866 copyin_psstrings(struct proc *p, struct ps_strings *arginfo) 1867 { 1868 1869 #ifdef COMPAT_NETBSD32 1870 if (p->p_flag & PK_32) { 1871 struct ps_strings32 arginfo32; 1872 1873 int error = copyin_proc(p, (void *)p->p_psstrp, &arginfo32, 1874 sizeof(arginfo32)); 1875 if (error) 1876 return error; 1877 arginfo->ps_argvstr = (void *)(uintptr_t)arginfo32.ps_argvstr; 1878 arginfo->ps_nargvstr = arginfo32.ps_nargvstr; 1879 arginfo->ps_envstr = (void *)(uintptr_t)arginfo32.ps_envstr; 1880 arginfo->ps_nenvstr = arginfo32.ps_nenvstr; 1881 return 0; 1882 } 1883 #endif 1884 return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo)); 1885 } 1886 1887 static int 1888 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len) 1889 { 1890 void **cookie = cookie_; 1891 struct lwp *l = cookie[0]; 1892 char *dst = cookie[1]; 1893 1894 return sysctl_copyout(l, src, dst + off, len); 1895 } 1896 1897 /* 1898 * sysctl helper routine for kern.proc_args pseudo-subtree. 1899 */ 1900 static int 1901 sysctl_kern_proc_args(SYSCTLFN_ARGS) 1902 { 1903 struct ps_strings pss; 1904 struct proc *p; 1905 pid_t pid; 1906 int type, error; 1907 void *cookie[2]; 1908 1909 if (namelen == 1 && name[0] == CTL_QUERY) 1910 return (sysctl_query(SYSCTLFN_CALL(rnode))); 1911 1912 if (newp != NULL || namelen != 2) 1913 return (EINVAL); 1914 pid = name[0]; 1915 type = name[1]; 1916 1917 switch (type) { 1918 case KERN_PROC_PATHNAME: 1919 sysctl_unlock(); 1920 error = fill_pathname(l, pid, oldp, oldlenp); 1921 sysctl_relock(); 1922 return error; 1923 1924 case KERN_PROC_ARGV: 1925 case KERN_PROC_NARGV: 1926 case KERN_PROC_ENV: 1927 case KERN_PROC_NENV: 1928 /* ok */ 1929 break; 1930 default: 1931 return (EINVAL); 1932 } 1933 1934 sysctl_unlock(); 1935 1936 /* check pid */ 1937 mutex_enter(proc_lock); 1938 if ((p = proc_find(pid)) == NULL) { 1939 error = EINVAL; 1940 goto out_locked; 1941 } 1942 mutex_enter(p->p_lock); 1943 1944 /* Check permission. */ 1945 if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV) 1946 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 1947 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL); 1948 else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV) 1949 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE, 1950 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL); 1951 else 1952 error = EINVAL; /* XXXGCC */ 1953 if (error) { 1954 mutex_exit(p->p_lock); 1955 goto out_locked; 1956 } 1957 1958 if (oldp == NULL) { 1959 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) 1960 *oldlenp = sizeof (int); 1961 else 1962 *oldlenp = ARG_MAX; /* XXX XXX XXX */ 1963 error = 0; 1964 mutex_exit(p->p_lock); 1965 goto out_locked; 1966 } 1967 1968 /* 1969 * Zombies don't have a stack, so we can't read their psstrings. 1970 * System processes also don't have a user stack. 1971 */ 1972 if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) { 1973 error = EINVAL; 1974 mutex_exit(p->p_lock); 1975 goto out_locked; 1976 } 1977 1978 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY; 1979 mutex_exit(p->p_lock); 1980 if (error) { 1981 goto out_locked; 1982 } 1983 mutex_exit(proc_lock); 1984 1985 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) { 1986 int value; 1987 if ((error = copyin_psstrings(p, &pss)) == 0) { 1988 if (type == KERN_PROC_NARGV) 1989 value = pss.ps_nargvstr; 1990 else 1991 value = pss.ps_nenvstr; 1992 error = sysctl_copyout(l, &value, oldp, sizeof(value)); 1993 *oldlenp = sizeof(value); 1994 } 1995 } else { 1996 cookie[0] = l; 1997 cookie[1] = oldp; 1998 error = copy_procargs(p, type, oldlenp, 1999 copy_procargs_sysctl_cb, cookie); 2000 } 2001 rw_exit(&p->p_reflock); 2002 sysctl_relock(); 2003 return error; 2004 2005 out_locked: 2006 mutex_exit(proc_lock); 2007 sysctl_relock(); 2008 return error; 2009 } 2010 2011 int 2012 copy_procargs(struct proc *p, int oid, size_t *limit, 2013 int (*cb)(void *, const void *, size_t, size_t), void *cookie) 2014 { 2015 struct ps_strings pss; 2016 size_t len, i, loaded, entry_len; 2017 struct uio auio; 2018 struct iovec aiov; 2019 int error, argvlen; 2020 char *arg; 2021 char **argv; 2022 vaddr_t user_argv; 2023 struct vmspace *vmspace; 2024 2025 /* 2026 * Allocate a temporary buffer to hold the argument vector and 2027 * the arguments themselve. 2028 */ 2029 arg = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2030 argv = kmem_alloc(PAGE_SIZE, KM_SLEEP); 2031 2032 /* 2033 * Lock the process down in memory. 2034 */ 2035 vmspace = p->p_vmspace; 2036 uvmspace_addref(vmspace); 2037 2038 /* 2039 * Read in the ps_strings structure. 2040 */ 2041 if ((error = copyin_psstrings(p, &pss)) != 0) 2042 goto done; 2043 2044 /* 2045 * Now read the address of the argument vector. 2046 */ 2047 switch (oid) { 2048 case KERN_PROC_ARGV: 2049 user_argv = (uintptr_t)pss.ps_argvstr; 2050 argvlen = pss.ps_nargvstr; 2051 break; 2052 case KERN_PROC_ENV: 2053 user_argv = (uintptr_t)pss.ps_envstr; 2054 argvlen = pss.ps_nenvstr; 2055 break; 2056 default: 2057 error = EINVAL; 2058 goto done; 2059 } 2060 2061 if (argvlen < 0) { 2062 error = EIO; 2063 goto done; 2064 } 2065 2066 2067 /* 2068 * Now copy each string. 2069 */ 2070 len = 0; /* bytes written to user buffer */ 2071 loaded = 0; /* bytes from argv already processed */ 2072 i = 0; /* To make compiler happy */ 2073 entry_len = PROC_PTRSZ(p); 2074 2075 for (; argvlen; --argvlen) { 2076 int finished = 0; 2077 vaddr_t base; 2078 size_t xlen; 2079 int j; 2080 2081 if (loaded == 0) { 2082 size_t rem = entry_len * argvlen; 2083 loaded = MIN(rem, PAGE_SIZE); 2084 error = copyin_vmspace(vmspace, 2085 (const void *)user_argv, argv, loaded); 2086 if (error) 2087 break; 2088 user_argv += loaded; 2089 i = 0; 2090 } 2091 2092 #ifdef COMPAT_NETBSD32 2093 if (p->p_flag & PK_32) { 2094 netbsd32_charp *argv32; 2095 2096 argv32 = (netbsd32_charp *)argv; 2097 base = (vaddr_t)NETBSD32PTR64(argv32[i++]); 2098 } else 2099 #endif 2100 base = (vaddr_t)argv[i++]; 2101 loaded -= entry_len; 2102 2103 /* 2104 * The program has messed around with its arguments, 2105 * possibly deleting some, and replacing them with 2106 * NULL's. Treat this as the last argument and not 2107 * a failure. 2108 */ 2109 if (base == 0) 2110 break; 2111 2112 while (!finished) { 2113 xlen = PAGE_SIZE - (base & PAGE_MASK); 2114 2115 aiov.iov_base = arg; 2116 aiov.iov_len = PAGE_SIZE; 2117 auio.uio_iov = &aiov; 2118 auio.uio_iovcnt = 1; 2119 auio.uio_offset = base; 2120 auio.uio_resid = xlen; 2121 auio.uio_rw = UIO_READ; 2122 UIO_SETUP_SYSSPACE(&auio); 2123 error = uvm_io(&vmspace->vm_map, &auio, 0); 2124 if (error) 2125 goto done; 2126 2127 /* Look for the end of the string */ 2128 for (j = 0; j < xlen; j++) { 2129 if (arg[j] == '\0') { 2130 xlen = j + 1; 2131 finished = 1; 2132 break; 2133 } 2134 } 2135 2136 /* Check for user buffer overflow */ 2137 if (len + xlen > *limit) { 2138 finished = 1; 2139 if (len > *limit) 2140 xlen = 0; 2141 else 2142 xlen = *limit - len; 2143 } 2144 2145 /* Copyout the page */ 2146 error = (*cb)(cookie, arg, len, xlen); 2147 if (error) 2148 goto done; 2149 2150 len += xlen; 2151 base += xlen; 2152 } 2153 } 2154 *limit = len; 2155 2156 done: 2157 kmem_free(argv, PAGE_SIZE); 2158 kmem_free(arg, PAGE_SIZE); 2159 uvmspace_free(vmspace); 2160 return error; 2161 } 2162 2163 #define SET_KERN_ADDR(dst, src, allow) \ 2164 do { \ 2165 if (allow) \ 2166 dst = src; \ 2167 } while (0); 2168 2169 /* 2170 * Fill in an eproc structure for the specified process. 2171 */ 2172 void 2173 fill_eproc(struct proc *p, struct eproc *ep, bool zombie) 2174 { 2175 bool allowaddr; 2176 struct tty *tp; 2177 struct lwp *l; 2178 int error; 2179 2180 KASSERT(mutex_owned(proc_lock)); 2181 KASSERT(mutex_owned(p->p_lock)); 2182 2183 memset(ep, 0, sizeof(*ep)); 2184 2185 /* If not privileged, don't expose kernel addresses. */ 2186 error = kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE, 2187 curproc, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL); 2188 allowaddr = (error == 0); 2189 2190 SET_KERN_ADDR(ep->e_paddr, p, allowaddr); 2191 SET_KERN_ADDR(ep->e_sess, p->p_session, allowaddr); 2192 if (p->p_cred) { 2193 kauth_cred_topcred(p->p_cred, &ep->e_pcred); 2194 kauth_cred_toucred(p->p_cred, &ep->e_ucred); 2195 } 2196 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2197 struct vmspace *vm = p->p_vmspace; 2198 2199 ep->e_vm.vm_rssize = vm_resident_count(vm); 2200 ep->e_vm.vm_tsize = vm->vm_tsize; 2201 ep->e_vm.vm_dsize = vm->vm_dsize; 2202 ep->e_vm.vm_ssize = vm->vm_ssize; 2203 ep->e_vm.vm_map.size = vm->vm_map.size; 2204 2205 /* Pick the primary (first) LWP */ 2206 l = proc_active_lwp(p); 2207 KASSERT(l != NULL); 2208 lwp_lock(l); 2209 if (l->l_wchan) 2210 strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN); 2211 lwp_unlock(l); 2212 } 2213 ep->e_ppid = p->p_ppid; 2214 if (p->p_pgrp && p->p_session) { 2215 ep->e_pgid = p->p_pgrp->pg_id; 2216 ep->e_jobc = p->p_pgrp->pg_jobc; 2217 ep->e_sid = p->p_session->s_sid; 2218 if ((p->p_lflag & PL_CONTROLT) && 2219 (tp = p->p_session->s_ttyp)) { 2220 ep->e_tdev = tp->t_dev; 2221 ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2222 SET_KERN_ADDR(ep->e_tsess, tp->t_session, allowaddr); 2223 } else 2224 ep->e_tdev = (uint32_t)NODEV; 2225 ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0; 2226 if (SESS_LEADER(p)) 2227 ep->e_flag |= EPROC_SLEADER; 2228 strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME); 2229 } 2230 ep->e_xsize = ep->e_xrssize = 0; 2231 ep->e_xccount = ep->e_xswrss = 0; 2232 } 2233 2234 /* 2235 * Fill in a kinfo_proc2 structure for the specified process. 2236 */ 2237 void 2238 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie) 2239 { 2240 struct tty *tp; 2241 struct lwp *l, *l2; 2242 struct timeval ut, st, rt; 2243 sigset_t ss1, ss2; 2244 struct rusage ru; 2245 struct vmspace *vm; 2246 bool allowaddr; 2247 int error; 2248 2249 KASSERT(mutex_owned(proc_lock)); 2250 KASSERT(mutex_owned(p->p_lock)); 2251 2252 /* If not privileged, don't expose kernel addresses. */ 2253 error = kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE, 2254 curproc, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL); 2255 allowaddr = (error == 0); 2256 2257 sigemptyset(&ss1); 2258 sigemptyset(&ss2); 2259 memset(ki, 0, sizeof(*ki)); 2260 2261 SET_KERN_ADDR(ki->p_paddr, PTRTOUINT64(p), allowaddr); 2262 SET_KERN_ADDR(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr); 2263 SET_KERN_ADDR(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr); 2264 SET_KERN_ADDR(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr); 2265 SET_KERN_ADDR(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr); 2266 SET_KERN_ADDR(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr); 2267 SET_KERN_ADDR(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr); 2268 SET_KERN_ADDR(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr); 2269 ki->p_tsess = 0; /* may be changed if controlling tty below */ 2270 SET_KERN_ADDR(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr); 2271 ki->p_eflag = 0; 2272 ki->p_exitsig = p->p_exitsig; 2273 ki->p_flag = L_INMEM; /* Process never swapped out */ 2274 ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag); 2275 ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag); 2276 ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag); 2277 ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag); 2278 ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag); 2279 ki->p_pid = p->p_pid; 2280 ki->p_ppid = p->p_ppid; 2281 ki->p_uid = kauth_cred_geteuid(p->p_cred); 2282 ki->p_ruid = kauth_cred_getuid(p->p_cred); 2283 ki->p_gid = kauth_cred_getegid(p->p_cred); 2284 ki->p_rgid = kauth_cred_getgid(p->p_cred); 2285 ki->p_svuid = kauth_cred_getsvuid(p->p_cred); 2286 ki->p_svgid = kauth_cred_getsvgid(p->p_cred); 2287 ki->p_ngroups = kauth_cred_ngroups(p->p_cred); 2288 kauth_cred_getgroups(p->p_cred, ki->p_groups, 2289 uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])), 2290 UIO_SYSSPACE); 2291 2292 ki->p_uticks = p->p_uticks; 2293 ki->p_sticks = p->p_sticks; 2294 ki->p_iticks = p->p_iticks; 2295 ki->p_tpgid = NO_PGID; /* may be changed if controlling tty below */ 2296 SET_KERN_ADDR(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr); 2297 ki->p_traceflag = p->p_traceflag; 2298 2299 memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t)); 2300 memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t)); 2301 2302 ki->p_cpticks = 0; 2303 ki->p_pctcpu = p->p_pctcpu; 2304 ki->p_estcpu = 0; 2305 ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */ 2306 ki->p_realstat = p->p_stat; 2307 ki->p_nice = p->p_nice; 2308 ki->p_xstat = P_WAITSTATUS(p); 2309 ki->p_acflag = p->p_acflag; 2310 2311 strncpy(ki->p_comm, p->p_comm, 2312 uimin(sizeof(ki->p_comm), sizeof(p->p_comm))); 2313 strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename)); 2314 2315 ki->p_nlwps = p->p_nlwps; 2316 ki->p_realflag = ki->p_flag; 2317 2318 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) { 2319 vm = p->p_vmspace; 2320 ki->p_vm_rssize = vm_resident_count(vm); 2321 ki->p_vm_tsize = vm->vm_tsize; 2322 ki->p_vm_dsize = vm->vm_dsize; 2323 ki->p_vm_ssize = vm->vm_ssize; 2324 ki->p_vm_vsize = atop(vm->vm_map.size); 2325 /* 2326 * Since the stack is initially mapped mostly with 2327 * PROT_NONE and grown as needed, adjust the "mapped size" 2328 * to skip the unused stack portion. 2329 */ 2330 ki->p_vm_msize = 2331 atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize; 2332 2333 /* Pick the primary (first) LWP */ 2334 l = proc_active_lwp(p); 2335 KASSERT(l != NULL); 2336 lwp_lock(l); 2337 ki->p_nrlwps = p->p_nrlwps; 2338 ki->p_forw = 0; 2339 ki->p_back = 0; 2340 SET_KERN_ADDR(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr); 2341 ki->p_stat = l->l_stat; 2342 ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag); 2343 ki->p_swtime = l->l_swtime; 2344 ki->p_slptime = l->l_slptime; 2345 if (l->l_stat == LSONPROC) 2346 ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags; 2347 else 2348 ki->p_schedflags = 0; 2349 ki->p_priority = lwp_eprio(l); 2350 ki->p_usrpri = l->l_priority; 2351 if (l->l_wchan) 2352 strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg)); 2353 SET_KERN_ADDR(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr); 2354 ki->p_cpuid = cpu_index(l->l_cpu); 2355 lwp_unlock(l); 2356 LIST_FOREACH(l, &p->p_lwps, l_sibling) { 2357 /* This is hardly correct, but... */ 2358 sigplusset(&l->l_sigpend.sp_set, &ss1); 2359 sigplusset(&l->l_sigmask, &ss2); 2360 ki->p_cpticks += l->l_cpticks; 2361 ki->p_pctcpu += l->l_pctcpu; 2362 ki->p_estcpu += l->l_estcpu; 2363 } 2364 } 2365 sigplusset(&p->p_sigpend.sp_set, &ss2); 2366 memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t)); 2367 memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t)); 2368 2369 if (p->p_session != NULL) { 2370 ki->p_sid = p->p_session->s_sid; 2371 ki->p__pgid = p->p_pgrp->pg_id; 2372 if (p->p_session->s_ttyvp) 2373 ki->p_eflag |= EPROC_CTTY; 2374 if (SESS_LEADER(p)) 2375 ki->p_eflag |= EPROC_SLEADER; 2376 strncpy(ki->p_login, p->p_session->s_login, 2377 uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login)); 2378 ki->p_jobc = p->p_pgrp->pg_jobc; 2379 if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) { 2380 ki->p_tdev = tp->t_dev; 2381 ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID; 2382 SET_KERN_ADDR(ki->p_tsess, PTRTOUINT64(tp->t_session), 2383 allowaddr); 2384 } else { 2385 ki->p_tdev = (int32_t)NODEV; 2386 } 2387 } 2388 2389 if (!P_ZOMBIE(p) && !zombie) { 2390 ki->p_uvalid = 1; 2391 ki->p_ustart_sec = p->p_stats->p_start.tv_sec; 2392 ki->p_ustart_usec = p->p_stats->p_start.tv_usec; 2393 2394 calcru(p, &ut, &st, NULL, &rt); 2395 ki->p_rtime_sec = rt.tv_sec; 2396 ki->p_rtime_usec = rt.tv_usec; 2397 ki->p_uutime_sec = ut.tv_sec; 2398 ki->p_uutime_usec = ut.tv_usec; 2399 ki->p_ustime_sec = st.tv_sec; 2400 ki->p_ustime_usec = st.tv_usec; 2401 2402 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru)); 2403 ki->p_uru_nvcsw = 0; 2404 ki->p_uru_nivcsw = 0; 2405 LIST_FOREACH(l2, &p->p_lwps, l_sibling) { 2406 ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw); 2407 ki->p_uru_nivcsw += l2->l_nivcsw; 2408 ruadd(&ru, &l2->l_ru); 2409 } 2410 ki->p_uru_maxrss = ru.ru_maxrss; 2411 ki->p_uru_ixrss = ru.ru_ixrss; 2412 ki->p_uru_idrss = ru.ru_idrss; 2413 ki->p_uru_isrss = ru.ru_isrss; 2414 ki->p_uru_minflt = ru.ru_minflt; 2415 ki->p_uru_majflt = ru.ru_majflt; 2416 ki->p_uru_nswap = ru.ru_nswap; 2417 ki->p_uru_inblock = ru.ru_inblock; 2418 ki->p_uru_oublock = ru.ru_oublock; 2419 ki->p_uru_msgsnd = ru.ru_msgsnd; 2420 ki->p_uru_msgrcv = ru.ru_msgrcv; 2421 ki->p_uru_nsignals = ru.ru_nsignals; 2422 2423 timeradd(&p->p_stats->p_cru.ru_utime, 2424 &p->p_stats->p_cru.ru_stime, &ut); 2425 ki->p_uctime_sec = ut.tv_sec; 2426 ki->p_uctime_usec = ut.tv_usec; 2427 } 2428 } 2429 2430 2431 int 2432 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid) 2433 { 2434 int error; 2435 2436 mutex_enter(proc_lock); 2437 if (pid == -1) 2438 *p = l->l_proc; 2439 else 2440 *p = proc_find(pid); 2441 2442 if (*p == NULL) { 2443 if (pid != -1) 2444 mutex_exit(proc_lock); 2445 return ESRCH; 2446 } 2447 if (pid != -1) 2448 mutex_enter((*p)->p_lock); 2449 mutex_exit(proc_lock); 2450 2451 error = kauth_authorize_process(l->l_cred, 2452 KAUTH_PROCESS_CANSEE, *p, 2453 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL); 2454 if (error) { 2455 if (pid != -1) 2456 mutex_exit((*p)->p_lock); 2457 } 2458 return error; 2459 } 2460 2461 static int 2462 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp) 2463 { 2464 int error; 2465 struct proc *p; 2466 2467 if ((error = proc_find_locked(l, &p, pid)) != 0) 2468 return error; 2469 2470 if (p->p_path == NULL) { 2471 if (pid != -1) 2472 mutex_exit(p->p_lock); 2473 return ENOENT; 2474 } 2475 2476 size_t len = strlen(p->p_path) + 1; 2477 if (oldp != NULL) { 2478 error = sysctl_copyout(l, p->p_path, oldp, *oldlenp); 2479 if (error == 0 && *oldlenp < len) 2480 error = ENOSPC; 2481 } 2482 *oldlenp = len; 2483 if (pid != -1) 2484 mutex_exit(p->p_lock); 2485 return error; 2486 } 2487 2488 int 2489 proc_getauxv(struct proc *p, void **buf, size_t *len) 2490 { 2491 struct ps_strings pss; 2492 int error; 2493 void *uauxv, *kauxv; 2494 size_t size; 2495 2496 if ((error = copyin_psstrings(p, &pss)) != 0) 2497 return error; 2498 if (pss.ps_envstr == NULL) 2499 return EIO; 2500 2501 size = p->p_execsw->es_arglen; 2502 if (size == 0) 2503 return EIO; 2504 2505 size_t ptrsz = PROC_PTRSZ(p); 2506 uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz); 2507 2508 kauxv = kmem_alloc(size, KM_SLEEP); 2509 2510 error = copyin_proc(p, uauxv, kauxv, size); 2511 if (error) { 2512 kmem_free(kauxv, size); 2513 return error; 2514 } 2515 2516 *buf = kauxv; 2517 *len = size; 2518 2519 return 0; 2520 } 2521