1 /* $NetBSD: kern_ktrace.c,v 1.157 2011/06/01 21:25:00 alnsn Exp $ */ 2 3 /*- 4 * Copyright (c) 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 Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)kern_ktrace.c 8.5 (Berkeley) 5/14/95 61 */ 62 63 #include <sys/cdefs.h> 64 __KERNEL_RCSID(0, "$NetBSD: kern_ktrace.c,v 1.157 2011/06/01 21:25:00 alnsn Exp $"); 65 66 #include <sys/param.h> 67 #include <sys/systm.h> 68 #include <sys/proc.h> 69 #include <sys/file.h> 70 #include <sys/namei.h> 71 #include <sys/vnode.h> 72 #include <sys/kernel.h> 73 #include <sys/kthread.h> 74 #include <sys/ktrace.h> 75 #include <sys/kmem.h> 76 #include <sys/syslog.h> 77 #include <sys/filedesc.h> 78 #include <sys/ioctl.h> 79 #include <sys/callout.h> 80 #include <sys/kauth.h> 81 82 #include <sys/mount.h> 83 #include <sys/sa.h> 84 #include <sys/syscallargs.h> 85 86 /* 87 * TODO: 88 * - need better error reporting? 89 * - userland utility to sort ktrace.out by timestamp. 90 * - keep minimum information in ktrace_entry when rest of alloc failed. 91 * - per trace control of configurable parameters. 92 */ 93 94 struct ktrace_entry { 95 TAILQ_ENTRY(ktrace_entry) kte_list; 96 struct ktr_header kte_kth; 97 void *kte_buf; 98 size_t kte_bufsz; 99 #define KTE_SPACE 32 100 uint8_t kte_space[KTE_SPACE]; 101 }; 102 103 struct ktr_desc { 104 TAILQ_ENTRY(ktr_desc) ktd_list; 105 int ktd_flags; 106 #define KTDF_WAIT 0x0001 107 #define KTDF_DONE 0x0002 108 #define KTDF_BLOCKING 0x0004 109 #define KTDF_INTERACTIVE 0x0008 110 int ktd_error; 111 #define KTDE_ENOMEM 0x0001 112 #define KTDE_ENOSPC 0x0002 113 int ktd_errcnt; 114 int ktd_ref; /* # of reference */ 115 int ktd_qcount; /* # of entry in the queue */ 116 117 /* 118 * Params to control behaviour. 119 */ 120 int ktd_delayqcnt; /* # of entry allowed to delay */ 121 int ktd_wakedelay; /* delay of wakeup in *tick* */ 122 int ktd_intrwakdl; /* ditto, but when interactive */ 123 124 file_t *ktd_fp; /* trace output file */ 125 lwp_t *ktd_lwp; /* our kernel thread */ 126 TAILQ_HEAD(, ktrace_entry) ktd_queue; 127 callout_t ktd_wakch; /* delayed wakeup */ 128 kcondvar_t ktd_sync_cv; 129 kcondvar_t ktd_cv; 130 }; 131 132 static int ktealloc(struct ktrace_entry **, void **, lwp_t *, int, 133 size_t); 134 static void ktrwrite(struct ktr_desc *, struct ktrace_entry *); 135 static int ktrace_common(lwp_t *, int, int, int, file_t *); 136 static int ktrops(lwp_t *, struct proc *, int, int, 137 struct ktr_desc *); 138 static int ktrsetchildren(lwp_t *, struct proc *, int, int, 139 struct ktr_desc *); 140 static int ktrcanset(lwp_t *, struct proc *); 141 static int ktrsamefile(file_t *, file_t *); 142 static void ktr_kmem(lwp_t *, int, const void *, size_t); 143 static void ktr_io(lwp_t *, int, enum uio_rw, struct iovec *, size_t); 144 145 static struct ktr_desc * 146 ktd_lookup(file_t *); 147 static void ktdrel(struct ktr_desc *); 148 static void ktdref(struct ktr_desc *); 149 static void ktraddentry(lwp_t *, struct ktrace_entry *, int); 150 /* Flags for ktraddentry (3rd arg) */ 151 #define KTA_NOWAIT 0x0000 152 #define KTA_WAITOK 0x0001 153 #define KTA_LARGE 0x0002 154 static void ktefree(struct ktrace_entry *); 155 static void ktd_logerrl(struct ktr_desc *, int); 156 static void ktrace_thread(void *); 157 static int ktrderefall(struct ktr_desc *, int); 158 159 /* 160 * Default vaules. 161 */ 162 #define KTD_MAXENTRY 1000 /* XXX: tune */ 163 #define KTD_TIMEOUT 5 /* XXX: tune */ 164 #define KTD_DELAYQCNT 100 /* XXX: tune */ 165 #define KTD_WAKEDELAY 5000 /* XXX: tune */ 166 #define KTD_INTRWAKDL 100 /* XXX: tune */ 167 168 /* 169 * Patchable variables. 170 */ 171 int ktd_maxentry = KTD_MAXENTRY; /* max # of entry in the queue */ 172 int ktd_timeout = KTD_TIMEOUT; /* timeout in seconds */ 173 int ktd_delayqcnt = KTD_DELAYQCNT; /* # of entry allowed to delay */ 174 int ktd_wakedelay = KTD_WAKEDELAY; /* delay of wakeup in *ms* */ 175 int ktd_intrwakdl = KTD_INTRWAKDL; /* ditto, but when interactive */ 176 177 kmutex_t ktrace_lock; 178 int ktrace_on; 179 static TAILQ_HEAD(, ktr_desc) ktdq = TAILQ_HEAD_INITIALIZER(ktdq); 180 static pool_cache_t kte_cache; 181 182 static kauth_listener_t ktrace_listener; 183 184 static void 185 ktd_wakeup(struct ktr_desc *ktd) 186 { 187 188 callout_stop(&ktd->ktd_wakch); 189 cv_signal(&ktd->ktd_cv); 190 } 191 192 static void 193 ktd_callout(void *arg) 194 { 195 196 mutex_enter(&ktrace_lock); 197 ktd_wakeup(arg); 198 mutex_exit(&ktrace_lock); 199 } 200 201 static void 202 ktd_logerrl(struct ktr_desc *ktd, int error) 203 { 204 205 ktd->ktd_error |= error; 206 ktd->ktd_errcnt++; 207 } 208 209 #if 0 210 static void 211 ktd_logerr(struct proc *p, int error) 212 { 213 struct ktr_desc *ktd; 214 215 KASSERT(mutex_owned(&ktrace_lock)); 216 217 ktd = p->p_tracep; 218 if (ktd == NULL) 219 return; 220 221 ktd_logerrl(ktd, error); 222 } 223 #endif 224 225 static inline int 226 ktrenter(lwp_t *l) 227 { 228 229 if ((l->l_pflag & LP_KTRACTIVE) != 0) 230 return 1; 231 l->l_pflag |= LP_KTRACTIVE; 232 return 0; 233 } 234 235 static inline void 236 ktrexit(lwp_t *l) 237 { 238 239 l->l_pflag &= ~LP_KTRACTIVE; 240 } 241 242 static int 243 ktrace_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 244 void *arg0, void *arg1, void *arg2, void *arg3) 245 { 246 struct proc *p; 247 int result; 248 enum kauth_process_req req; 249 250 result = KAUTH_RESULT_DEFER; 251 p = arg0; 252 253 if (action != KAUTH_PROCESS_KTRACE) 254 return result; 255 256 req = (enum kauth_process_req)(unsigned long)arg1; 257 258 /* Privileged; secmodel should handle these. */ 259 if (req == KAUTH_REQ_PROCESS_KTRACE_PERSISTENT) 260 return result; 261 262 if ((p->p_traceflag & KTRFAC_PERSISTENT) || 263 (p->p_flag & PK_SUGID)) 264 return result; 265 266 if (kauth_cred_geteuid(cred) == kauth_cred_getuid(p->p_cred) && 267 kauth_cred_getuid(cred) == kauth_cred_getsvuid(p->p_cred) && 268 kauth_cred_getgid(cred) == kauth_cred_getgid(p->p_cred) && 269 kauth_cred_getgid(cred) == kauth_cred_getsvgid(p->p_cred)) 270 result = KAUTH_RESULT_ALLOW; 271 272 return result; 273 } 274 275 /* 276 * Initialise the ktrace system. 277 */ 278 void 279 ktrinit(void) 280 { 281 282 mutex_init(&ktrace_lock, MUTEX_DEFAULT, IPL_NONE); 283 kte_cache = pool_cache_init(sizeof(struct ktrace_entry), 0, 0, 0, 284 "ktrace", &pool_allocator_nointr, IPL_NONE, NULL, NULL, NULL); 285 286 ktrace_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 287 ktrace_listener_cb, NULL); 288 } 289 290 /* 291 * Release a reference. Called with ktrace_lock held. 292 */ 293 void 294 ktdrel(struct ktr_desc *ktd) 295 { 296 297 KASSERT(mutex_owned(&ktrace_lock)); 298 299 KDASSERT(ktd->ktd_ref != 0); 300 KASSERT(ktd->ktd_ref > 0); 301 KASSERT(ktrace_on > 0); 302 ktrace_on--; 303 if (--ktd->ktd_ref <= 0) { 304 ktd->ktd_flags |= KTDF_DONE; 305 cv_signal(&ktd->ktd_cv); 306 } 307 } 308 309 void 310 ktdref(struct ktr_desc *ktd) 311 { 312 313 KASSERT(mutex_owned(&ktrace_lock)); 314 315 ktd->ktd_ref++; 316 ktrace_on++; 317 } 318 319 struct ktr_desc * 320 ktd_lookup(file_t *fp) 321 { 322 struct ktr_desc *ktd; 323 324 KASSERT(mutex_owned(&ktrace_lock)); 325 326 for (ktd = TAILQ_FIRST(&ktdq); ktd != NULL; 327 ktd = TAILQ_NEXT(ktd, ktd_list)) { 328 if (ktrsamefile(ktd->ktd_fp, fp)) { 329 ktdref(ktd); 330 break; 331 } 332 } 333 334 return (ktd); 335 } 336 337 void 338 ktraddentry(lwp_t *l, struct ktrace_entry *kte, int flags) 339 { 340 struct proc *p = l->l_proc; 341 struct ktr_desc *ktd; 342 #ifdef DEBUG 343 struct timeval t1, t2; 344 #endif 345 346 mutex_enter(&ktrace_lock); 347 348 if (p->p_traceflag & KTRFAC_TRC_EMUL) { 349 /* Add emulation trace before first entry for this process */ 350 p->p_traceflag &= ~KTRFAC_TRC_EMUL; 351 mutex_exit(&ktrace_lock); 352 ktrexit(l); 353 ktremul(); 354 (void)ktrenter(l); 355 mutex_enter(&ktrace_lock); 356 } 357 358 /* Tracing may have been cancelled. */ 359 ktd = p->p_tracep; 360 if (ktd == NULL) 361 goto freekte; 362 363 /* 364 * Bump reference count so that the object will remain while 365 * we are here. Note that the trace is controlled by other 366 * process. 367 */ 368 ktdref(ktd); 369 370 if (ktd->ktd_flags & KTDF_DONE) 371 goto relktd; 372 373 if (ktd->ktd_qcount > ktd_maxentry) { 374 ktd_logerrl(ktd, KTDE_ENOSPC); 375 goto relktd; 376 } 377 TAILQ_INSERT_TAIL(&ktd->ktd_queue, kte, kte_list); 378 ktd->ktd_qcount++; 379 if (ktd->ktd_flags & KTDF_BLOCKING) 380 goto skip_sync; 381 382 if (flags & KTA_WAITOK && 383 (/* flags & KTA_LARGE */0 || ktd->ktd_flags & KTDF_WAIT || 384 ktd->ktd_qcount > ktd_maxentry >> 1)) 385 /* 386 * Sync with writer thread since we're requesting rather 387 * big one or many requests are pending. 388 */ 389 do { 390 ktd->ktd_flags |= KTDF_WAIT; 391 ktd_wakeup(ktd); 392 #ifdef DEBUG 393 getmicrouptime(&t1); 394 #endif 395 if (cv_timedwait(&ktd->ktd_sync_cv, &ktrace_lock, 396 ktd_timeout * hz) != 0) { 397 ktd->ktd_flags |= KTDF_BLOCKING; 398 /* 399 * Maybe the writer thread is blocking 400 * completely for some reason, but 401 * don't stop target process forever. 402 */ 403 log(LOG_NOTICE, "ktrace timeout\n"); 404 break; 405 } 406 #ifdef DEBUG 407 getmicrouptime(&t2); 408 timersub(&t2, &t1, &t2); 409 if (t2.tv_sec > 0) 410 log(LOG_NOTICE, 411 "ktrace long wait: %lld.%06ld\n", 412 (long long)t2.tv_sec, (long)t2.tv_usec); 413 #endif 414 } while (p->p_tracep == ktd && 415 (ktd->ktd_flags & (KTDF_WAIT | KTDF_DONE)) == KTDF_WAIT); 416 else { 417 /* Schedule delayed wakeup */ 418 if (ktd->ktd_qcount > ktd->ktd_delayqcnt) 419 ktd_wakeup(ktd); /* Wakeup now */ 420 else if (!callout_pending(&ktd->ktd_wakch)) 421 callout_reset(&ktd->ktd_wakch, 422 ktd->ktd_flags & KTDF_INTERACTIVE ? 423 ktd->ktd_intrwakdl : ktd->ktd_wakedelay, 424 ktd_callout, ktd); 425 } 426 427 skip_sync: 428 ktdrel(ktd); 429 mutex_exit(&ktrace_lock); 430 ktrexit(l); 431 return; 432 433 relktd: 434 ktdrel(ktd); 435 436 freekte: 437 mutex_exit(&ktrace_lock); 438 ktefree(kte); 439 ktrexit(l); 440 } 441 442 void 443 ktefree(struct ktrace_entry *kte) 444 { 445 446 if (kte->kte_buf != kte->kte_space) 447 kmem_free(kte->kte_buf, kte->kte_bufsz); 448 pool_cache_put(kte_cache, kte); 449 } 450 451 /* 452 * "deep" compare of two files for the purposes of clearing a trace. 453 * Returns true if they're the same open file, or if they point at the 454 * same underlying vnode/socket. 455 */ 456 457 int 458 ktrsamefile(file_t *f1, file_t *f2) 459 { 460 461 return ((f1 == f2) || 462 ((f1 != NULL) && (f2 != NULL) && 463 (f1->f_type == f2->f_type) && 464 (f1->f_data == f2->f_data))); 465 } 466 467 void 468 ktrderef(struct proc *p) 469 { 470 struct ktr_desc *ktd = p->p_tracep; 471 472 KASSERT(mutex_owned(&ktrace_lock)); 473 474 p->p_traceflag = 0; 475 if (ktd == NULL) 476 return; 477 p->p_tracep = NULL; 478 479 cv_broadcast(&ktd->ktd_sync_cv); 480 ktdrel(ktd); 481 } 482 483 void 484 ktradref(struct proc *p) 485 { 486 struct ktr_desc *ktd = p->p_tracep; 487 488 KASSERT(mutex_owned(&ktrace_lock)); 489 490 ktdref(ktd); 491 } 492 493 int 494 ktrderefall(struct ktr_desc *ktd, int auth) 495 { 496 lwp_t *curl = curlwp; 497 struct proc *p; 498 int error = 0; 499 500 mutex_enter(proc_lock); 501 PROCLIST_FOREACH(p, &allproc) { 502 if (p->p_tracep != ktd) 503 continue; 504 mutex_enter(p->p_lock); 505 mutex_enter(&ktrace_lock); 506 if (p->p_tracep == ktd) { 507 if (!auth || ktrcanset(curl, p)) 508 ktrderef(p); 509 else 510 error = EPERM; 511 } 512 mutex_exit(&ktrace_lock); 513 mutex_exit(p->p_lock); 514 } 515 mutex_exit(proc_lock); 516 517 return error; 518 } 519 520 int 521 ktealloc(struct ktrace_entry **ktep, void **bufp, lwp_t *l, int type, 522 size_t sz) 523 { 524 struct proc *p = l->l_proc; 525 struct ktrace_entry *kte; 526 struct ktr_header *kth; 527 void *buf; 528 529 if (ktrenter(l)) 530 return EAGAIN; 531 532 kte = pool_cache_get(kte_cache, PR_WAITOK); 533 if (sz > sizeof(kte->kte_space)) { 534 if ((buf = kmem_alloc(sz, KM_SLEEP)) == NULL) { 535 pool_cache_put(kte_cache, kte); 536 ktrexit(l); 537 return ENOMEM; 538 } 539 } else 540 buf = kte->kte_space; 541 542 kte->kte_bufsz = sz; 543 kte->kte_buf = buf; 544 545 kth = &kte->kte_kth; 546 (void)memset(kth, 0, sizeof(*kth)); 547 kth->ktr_len = sz; 548 kth->ktr_type = type; 549 kth->ktr_pid = p->p_pid; 550 memcpy(kth->ktr_comm, p->p_comm, MAXCOMLEN); 551 kth->ktr_version = KTRFAC_VERSION(p->p_traceflag); 552 kth->ktr_lid = l->l_lid; 553 nanotime(&kth->ktr_ts); 554 555 *ktep = kte; 556 *bufp = buf; 557 558 return 0; 559 } 560 561 void 562 ktr_syscall(register_t code, const register_t args[], int narg) 563 { 564 lwp_t *l = curlwp; 565 struct proc *p = l->l_proc; 566 struct ktrace_entry *kte; 567 struct ktr_syscall *ktp; 568 register_t *argp; 569 size_t len; 570 u_int i; 571 572 if (!KTRPOINT(p, KTR_SYSCALL)) 573 return; 574 575 len = sizeof(struct ktr_syscall) + narg * sizeof argp[0]; 576 577 if (ktealloc(&kte, (void *)&ktp, l, KTR_SYSCALL, len)) 578 return; 579 580 ktp->ktr_code = code; 581 ktp->ktr_argsize = narg * sizeof argp[0]; 582 argp = (register_t *)(ktp + 1); 583 for (i = 0; i < narg; i++) 584 *argp++ = args[i]; 585 586 ktraddentry(l, kte, KTA_WAITOK); 587 } 588 589 void 590 ktr_sysret(register_t code, int error, register_t *retval) 591 { 592 lwp_t *l = curlwp; 593 struct ktrace_entry *kte; 594 struct ktr_sysret *ktp; 595 596 if (!KTRPOINT(l->l_proc, KTR_SYSRET)) 597 return; 598 599 if (ktealloc(&kte, (void *)&ktp, l, KTR_SYSRET, 600 sizeof(struct ktr_sysret))) 601 return; 602 603 ktp->ktr_code = code; 604 ktp->ktr_eosys = 0; /* XXX unused */ 605 ktp->ktr_error = error; 606 ktp->ktr_retval = retval ? retval[0] : 0; 607 ktp->ktr_retval_1 = retval ? retval[1] : 0; 608 609 ktraddentry(l, kte, KTA_WAITOK); 610 } 611 612 void 613 ktr_namei(const char *path, size_t pathlen) 614 { 615 lwp_t *l = curlwp; 616 617 if (!KTRPOINT(l->l_proc, KTR_NAMEI)) 618 return; 619 620 ktr_kmem(l, KTR_NAMEI, path, pathlen); 621 } 622 623 void 624 ktr_namei2(const char *eroot, size_t erootlen, 625 const char *path, size_t pathlen) 626 { 627 lwp_t *l = curlwp; 628 struct ktrace_entry *kte; 629 void *buf; 630 631 if (!KTRPOINT(l->l_proc, KTR_NAMEI)) 632 return; 633 634 if (ktealloc(&kte, &buf, l, KTR_NAMEI, erootlen + pathlen)) 635 return; 636 memcpy(buf, eroot, erootlen); 637 buf = (char *)buf + erootlen; 638 memcpy(buf, path, pathlen); 639 ktraddentry(l, kte, KTA_WAITOK); 640 } 641 642 void 643 ktr_emul(void) 644 { 645 lwp_t *l = curlwp; 646 const char *emul = l->l_proc->p_emul->e_name; 647 648 if (!KTRPOINT(l->l_proc, KTR_EMUL)) 649 return; 650 651 ktr_kmem(l, KTR_EMUL, emul, strlen(emul)); 652 } 653 654 void 655 ktr_execarg(const void *bf, size_t len) 656 { 657 lwp_t *l = curlwp; 658 659 if (!KTRPOINT(l->l_proc, KTR_EXEC_ARG)) 660 return; 661 662 ktr_kmem(l, KTR_EXEC_ARG, bf, len); 663 } 664 665 void 666 ktr_execenv(const void *bf, size_t len) 667 { 668 lwp_t *l = curlwp; 669 670 if (!KTRPOINT(l->l_proc, KTR_EXEC_ENV)) 671 return; 672 673 ktr_kmem(l, KTR_EXEC_ENV, bf, len); 674 } 675 676 void 677 ktr_execfd(int fd, u_int dtype) 678 { 679 struct ktrace_entry *kte; 680 struct ktr_execfd* ktp; 681 682 lwp_t *l = curlwp; 683 684 if (!KTRPOINT(l->l_proc, KTR_EXEC_FD)) 685 return; 686 687 if (ktealloc(&kte, (void *)&ktp, l, KTR_EXEC_FD, sizeof(*ktp))) 688 return; 689 690 ktp->ktr_fd = fd; 691 ktp->ktr_dtype = dtype; 692 ktraddentry(l, kte, KTA_WAITOK); 693 } 694 695 static void 696 ktr_kmem(lwp_t *l, int type, const void *bf, size_t len) 697 { 698 struct ktrace_entry *kte; 699 void *buf; 700 701 if (ktealloc(&kte, &buf, l, type, len)) 702 return; 703 memcpy(buf, bf, len); 704 ktraddentry(l, kte, KTA_WAITOK); 705 } 706 707 static void 708 ktr_io(lwp_t *l, int fd, enum uio_rw rw, struct iovec *iov, size_t len) 709 { 710 struct ktrace_entry *kte; 711 struct ktr_genio *ktp; 712 size_t resid = len, cnt, buflen; 713 char *cp; 714 715 next: 716 buflen = min(PAGE_SIZE, resid + sizeof(struct ktr_genio)); 717 718 if (ktealloc(&kte, (void *)&ktp, l, KTR_GENIO, buflen)) 719 return; 720 721 ktp->ktr_fd = fd; 722 ktp->ktr_rw = rw; 723 724 cp = (void *)(ktp + 1); 725 buflen -= sizeof(struct ktr_genio); 726 kte->kte_kth.ktr_len = sizeof(struct ktr_genio); 727 728 while (buflen > 0) { 729 cnt = min(iov->iov_len, buflen); 730 if (copyin(iov->iov_base, cp, cnt) != 0) 731 goto out; 732 kte->kte_kth.ktr_len += cnt; 733 cp += cnt; 734 buflen -= cnt; 735 resid -= cnt; 736 iov->iov_len -= cnt; 737 if (iov->iov_len == 0) 738 iov++; 739 else 740 iov->iov_base = (char *)iov->iov_base + cnt; 741 } 742 743 /* 744 * Don't push so many entry at once. It will cause kmem map 745 * shortage. 746 */ 747 ktraddentry(l, kte, KTA_WAITOK | KTA_LARGE); 748 if (resid > 0) { 749 if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD) { 750 (void)ktrenter(l); 751 preempt(); 752 ktrexit(l); 753 } 754 755 goto next; 756 } 757 758 return; 759 760 out: 761 ktefree(kte); 762 ktrexit(l); 763 } 764 765 void 766 ktr_genio(int fd, enum uio_rw rw, const void *addr, size_t len, int error) 767 { 768 lwp_t *l = curlwp; 769 struct iovec iov; 770 771 if (!KTRPOINT(l->l_proc, KTR_GENIO) || error != 0) 772 return; 773 iov.iov_base = __UNCONST(addr); 774 iov.iov_len = len; 775 ktr_io(l, fd, rw, &iov, len); 776 } 777 778 void 779 ktr_geniov(int fd, enum uio_rw rw, struct iovec *iov, size_t len, int error) 780 { 781 lwp_t *l = curlwp; 782 783 if (!KTRPOINT(l->l_proc, KTR_GENIO) || error != 0) 784 return; 785 ktr_io(l, fd, rw, iov, len); 786 } 787 788 void 789 ktr_mibio(int fd, enum uio_rw rw, const void *addr, size_t len, int error) 790 { 791 lwp_t *l = curlwp; 792 struct iovec iov; 793 794 if (!KTRPOINT(l->l_proc, KTR_MIB) || error != 0) 795 return; 796 iov.iov_base = __UNCONST(addr); 797 iov.iov_len = len; 798 ktr_io(l, fd, rw, &iov, len); 799 } 800 801 void 802 ktr_psig(int sig, sig_t action, const sigset_t *mask, 803 const ksiginfo_t *ksi) 804 { 805 struct ktrace_entry *kte; 806 lwp_t *l = curlwp; 807 struct { 808 struct ktr_psig kp; 809 siginfo_t si; 810 } *kbuf; 811 812 if (!KTRPOINT(l->l_proc, KTR_PSIG)) 813 return; 814 815 if (ktealloc(&kte, (void *)&kbuf, l, KTR_PSIG, sizeof(*kbuf))) 816 return; 817 818 kbuf->kp.signo = (char)sig; 819 kbuf->kp.action = action; 820 kbuf->kp.mask = *mask; 821 822 if (ksi) { 823 kbuf->kp.code = KSI_TRAPCODE(ksi); 824 (void)memset(&kbuf->si, 0, sizeof(kbuf->si)); 825 kbuf->si._info = ksi->ksi_info; 826 kte->kte_kth.ktr_len = sizeof(*kbuf); 827 } else { 828 kbuf->kp.code = 0; 829 kte->kte_kth.ktr_len = sizeof(struct ktr_psig); 830 } 831 832 ktraddentry(l, kte, KTA_WAITOK); 833 } 834 835 void 836 ktr_csw(int out, int user) 837 { 838 lwp_t *l = curlwp; 839 struct proc *p = l->l_proc; 840 struct ktrace_entry *kte; 841 struct ktr_csw *kc; 842 843 if (!KTRPOINT(p, KTR_CSW)) 844 return; 845 846 /* 847 * Don't record context switches resulting from blocking on 848 * locks; it's too easy to get duff results. 849 */ 850 if (l->l_syncobj == &mutex_syncobj || l->l_syncobj == &rw_syncobj) 851 return; 852 853 /* 854 * We can't sleep if we're already going to sleep (if original 855 * condition is met during sleep, we hang up). 856 * 857 * XXX This is not ideal: it would be better to maintain a pool 858 * of ktes and actually push this to the kthread when context 859 * switch happens, however given the points where we are called 860 * from that is difficult to do. 861 */ 862 if (out) { 863 if (ktrenter(l)) 864 return; 865 866 nanotime(&l->l_ktrcsw); 867 l->l_pflag |= LP_KTRCSW; 868 if (user) 869 l->l_pflag |= LP_KTRCSWUSER; 870 else 871 l->l_pflag &= ~LP_KTRCSWUSER; 872 873 ktrexit(l); 874 return; 875 } 876 877 /* 878 * On the way back in, we need to record twice: once for entry, and 879 * once for exit. 880 */ 881 if ((l->l_pflag & LP_KTRCSW) != 0) { 882 struct timespec *ts; 883 l->l_pflag &= ~LP_KTRCSW; 884 885 if (ktealloc(&kte, (void *)&kc, l, KTR_CSW, sizeof(*kc))) 886 return; 887 888 kc->out = 1; 889 kc->user = ((l->l_pflag & LP_KTRCSWUSER) != 0); 890 891 ts = &l->l_ktrcsw; 892 switch (KTRFAC_VERSION(p->p_traceflag)) { 893 case 0: 894 kte->kte_kth.ktr_otv.tv_sec = ts->tv_sec; 895 kte->kte_kth.ktr_otv.tv_usec = ts->tv_nsec / 1000; 896 break; 897 case 1: 898 kte->kte_kth.ktr_ots.tv_sec = ts->tv_sec; 899 kte->kte_kth.ktr_ots.tv_nsec = ts->tv_nsec; 900 break; 901 case 2: 902 kte->kte_kth.ktr_ts.tv_sec = ts->tv_sec; 903 kte->kte_kth.ktr_ts.tv_nsec = ts->tv_nsec; 904 break; 905 default: 906 break; 907 } 908 909 ktraddentry(l, kte, KTA_WAITOK); 910 } 911 912 if (ktealloc(&kte, (void *)&kc, l, KTR_CSW, sizeof(*kc))) 913 return; 914 915 kc->out = 0; 916 kc->user = user; 917 918 ktraddentry(l, kte, KTA_WAITOK); 919 } 920 921 bool 922 ktr_point(int fac_bit) 923 { 924 return curlwp->l_proc->p_traceflag & fac_bit; 925 } 926 927 int 928 ktruser(const char *id, void *addr, size_t len, int ustr) 929 { 930 struct ktrace_entry *kte; 931 struct ktr_user *ktp; 932 lwp_t *l = curlwp; 933 void *user_dta; 934 int error; 935 936 if (!KTRPOINT(l->l_proc, KTR_USER)) 937 return 0; 938 939 if (len > KTR_USER_MAXLEN) 940 return ENOSPC; 941 942 error = ktealloc(&kte, (void *)&ktp, l, KTR_USER, sizeof(*ktp) + len); 943 if (error != 0) 944 return error; 945 946 if (ustr) { 947 if (copyinstr(id, ktp->ktr_id, KTR_USER_MAXIDLEN, NULL) != 0) 948 ktp->ktr_id[0] = '\0'; 949 } else 950 strncpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN); 951 ktp->ktr_id[KTR_USER_MAXIDLEN-1] = '\0'; 952 953 user_dta = (void *)(ktp + 1); 954 if ((error = copyin(addr, (void *)user_dta, len)) != 0) 955 len = 0; 956 957 ktraddentry(l, kte, KTA_WAITOK); 958 return error; 959 } 960 961 void 962 ktr_kuser(const char *id, void *addr, size_t len) 963 { 964 struct ktrace_entry *kte; 965 struct ktr_user *ktp; 966 lwp_t *l = curlwp; 967 int error; 968 969 if (!KTRPOINT(l->l_proc, KTR_USER)) 970 return; 971 972 if (len > KTR_USER_MAXLEN) 973 return; 974 975 error = ktealloc(&kte, (void *)&ktp, l, KTR_USER, sizeof(*ktp) + len); 976 if (error != 0) 977 return; 978 979 strlcpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN); 980 981 memcpy(ktp + 1, addr, len); 982 983 ktraddentry(l, kte, KTA_WAITOK); 984 } 985 986 void 987 ktr_saupcall(struct lwp *l, int type, int nevent, int nint, void *sas, 988 void *ap, void *ksas) 989 { 990 struct ktrace_entry *kte; 991 struct ktr_saupcall *ktp; 992 size_t len, sz; 993 struct sa_t **sapp; 994 int i; 995 996 if (!KTRPOINT(l->l_proc, KTR_SAUPCALL)) 997 return; 998 999 len = sizeof(struct ktr_saupcall); 1000 sz = len + sizeof(struct sa_t) * (nevent + nint + 1); 1001 1002 if (ktealloc(&kte, (void *)&ktp, l, KTR_SAUPCALL, sz)) 1003 return; 1004 1005 ktp->ktr_type = type; 1006 ktp->ktr_nevent = nevent; 1007 ktp->ktr_nint = nint; 1008 ktp->ktr_sas = sas; 1009 ktp->ktr_ap = ap; 1010 1011 /* Copy the sa_t's */ 1012 sapp = (struct sa_t **) ksas; 1013 1014 for (i = nevent + nint; i >= 0; i--) { 1015 memcpy((char *)ktp + len, *sapp, sizeof(struct sa_t)); 1016 len += sizeof(struct sa_t); 1017 sapp++; 1018 } 1019 1020 kte->kte_kth.ktr_len = len; 1021 ktraddentry(l, kte, KTA_WAITOK); 1022 } 1023 1024 void 1025 ktr_mib(const int *name, u_int namelen) 1026 { 1027 struct ktrace_entry *kte; 1028 int *namep; 1029 size_t size; 1030 lwp_t *l = curlwp; 1031 1032 if (!KTRPOINT(l->l_proc, KTR_MIB)) 1033 return; 1034 1035 size = namelen * sizeof(*name); 1036 1037 if (ktealloc(&kte, (void *)&namep, l, KTR_MIB, size)) 1038 return; 1039 1040 (void)memcpy(namep, name, namelen * sizeof(*name)); 1041 1042 ktraddentry(l, kte, KTA_WAITOK); 1043 } 1044 1045 /* Interface and common routines */ 1046 1047 int 1048 ktrace_common(lwp_t *curl, int ops, int facs, int pid, file_t *fp) 1049 { 1050 struct proc *curp; 1051 struct proc *p; 1052 struct pgrp *pg; 1053 struct ktr_desc *ktd = NULL; 1054 int ret = 0; 1055 int error = 0; 1056 int descend; 1057 1058 curp = curl->l_proc; 1059 descend = ops & KTRFLAG_DESCEND; 1060 facs = facs & ~((unsigned) KTRFAC_PERSISTENT); 1061 1062 (void)ktrenter(curl); 1063 1064 switch (KTROP(ops)) { 1065 1066 case KTROP_CLEARFILE: 1067 /* 1068 * Clear all uses of the tracefile 1069 */ 1070 mutex_enter(&ktrace_lock); 1071 ktd = ktd_lookup(fp); 1072 mutex_exit(&ktrace_lock); 1073 if (ktd == NULL) 1074 goto done; 1075 error = ktrderefall(ktd, 1); 1076 goto done; 1077 1078 case KTROP_SET: 1079 mutex_enter(&ktrace_lock); 1080 ktd = ktd_lookup(fp); 1081 mutex_exit(&ktrace_lock); 1082 if (ktd == NULL) { 1083 ktd = kmem_alloc(sizeof(*ktd), KM_SLEEP); 1084 TAILQ_INIT(&ktd->ktd_queue); 1085 callout_init(&ktd->ktd_wakch, CALLOUT_MPSAFE); 1086 cv_init(&ktd->ktd_cv, "ktrwait"); 1087 cv_init(&ktd->ktd_sync_cv, "ktrsync"); 1088 ktd->ktd_flags = 0; 1089 ktd->ktd_qcount = 0; 1090 ktd->ktd_error = 0; 1091 ktd->ktd_errcnt = 0; 1092 ktd->ktd_delayqcnt = ktd_delayqcnt; 1093 ktd->ktd_wakedelay = mstohz(ktd_wakedelay); 1094 ktd->ktd_intrwakdl = mstohz(ktd_intrwakdl); 1095 ktd->ktd_ref = 0; 1096 ktd->ktd_fp = fp; 1097 mutex_enter(&ktrace_lock); 1098 ktdref(ktd); 1099 mutex_exit(&ktrace_lock); 1100 1101 /* 1102 * XXX: not correct. needs an way to detect 1103 * whether ktruss or ktrace. 1104 */ 1105 if (fp->f_type == DTYPE_PIPE) 1106 ktd->ktd_flags |= KTDF_INTERACTIVE; 1107 1108 mutex_enter(&fp->f_lock); 1109 fp->f_count++; 1110 mutex_exit(&fp->f_lock); 1111 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 1112 ktrace_thread, ktd, &ktd->ktd_lwp, "ktrace"); 1113 if (error != 0) { 1114 kmem_free(ktd, sizeof(*ktd)); 1115 mutex_enter(&fp->f_lock); 1116 fp->f_count--; 1117 mutex_exit(&fp->f_lock); 1118 goto done; 1119 } 1120 1121 mutex_enter(&ktrace_lock); 1122 if (ktd_lookup(fp) != NULL) { 1123 ktdrel(ktd); 1124 ktd = NULL; 1125 } else 1126 TAILQ_INSERT_TAIL(&ktdq, ktd, ktd_list); 1127 if (ktd == NULL) 1128 cv_wait(&lbolt, &ktrace_lock); 1129 mutex_exit(&ktrace_lock); 1130 if (ktd == NULL) 1131 goto done; 1132 } 1133 break; 1134 1135 case KTROP_CLEAR: 1136 break; 1137 } 1138 1139 /* 1140 * need something to (un)trace (XXX - why is this here?) 1141 */ 1142 if (!facs) { 1143 error = EINVAL; 1144 goto done; 1145 } 1146 1147 /* 1148 * do it 1149 */ 1150 mutex_enter(proc_lock); 1151 if (pid < 0) { 1152 /* 1153 * by process group 1154 */ 1155 pg = pgrp_find(-pid); 1156 if (pg == NULL) 1157 error = ESRCH; 1158 else { 1159 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1160 if (descend) 1161 ret |= ktrsetchildren(curl, p, ops, 1162 facs, ktd); 1163 else 1164 ret |= ktrops(curl, p, ops, facs, 1165 ktd); 1166 } 1167 } 1168 1169 } else { 1170 /* 1171 * by pid 1172 */ 1173 p = proc_find(pid); 1174 if (p == NULL) 1175 error = ESRCH; 1176 else if (descend) 1177 ret |= ktrsetchildren(curl, p, ops, facs, ktd); 1178 else 1179 ret |= ktrops(curl, p, ops, facs, ktd); 1180 } 1181 mutex_exit(proc_lock); 1182 if (error == 0 && !ret) 1183 error = EPERM; 1184 done: 1185 if (ktd != NULL) { 1186 mutex_enter(&ktrace_lock); 1187 if (error != 0) { 1188 /* 1189 * Wakeup the thread so that it can be die if we 1190 * can't trace any process. 1191 */ 1192 ktd_wakeup(ktd); 1193 } 1194 if (KTROP(ops) == KTROP_SET || KTROP(ops) == KTROP_CLEARFILE) 1195 ktdrel(ktd); 1196 mutex_exit(&ktrace_lock); 1197 } 1198 ktrexit(curl); 1199 return (error); 1200 } 1201 1202 /* 1203 * fktrace system call 1204 */ 1205 /* ARGSUSED */ 1206 int 1207 sys_fktrace(struct lwp *l, const struct sys_fktrace_args *uap, register_t *retval) 1208 { 1209 /* { 1210 syscallarg(int) fd; 1211 syscallarg(int) ops; 1212 syscallarg(int) facs; 1213 syscallarg(int) pid; 1214 } */ 1215 file_t *fp; 1216 int error, fd; 1217 1218 fd = SCARG(uap, fd); 1219 if ((fp = fd_getfile(fd)) == NULL) 1220 return (EBADF); 1221 if ((fp->f_flag & FWRITE) == 0) 1222 error = EBADF; 1223 else 1224 error = ktrace_common(l, SCARG(uap, ops), 1225 SCARG(uap, facs), SCARG(uap, pid), fp); 1226 fd_putfile(fd); 1227 return error; 1228 } 1229 1230 /* 1231 * ktrace system call 1232 */ 1233 /* ARGSUSED */ 1234 int 1235 sys_ktrace(struct lwp *l, const struct sys_ktrace_args *uap, register_t *retval) 1236 { 1237 /* { 1238 syscallarg(const char *) fname; 1239 syscallarg(int) ops; 1240 syscallarg(int) facs; 1241 syscallarg(int) pid; 1242 } */ 1243 struct vnode *vp = NULL; 1244 file_t *fp = NULL; 1245 struct pathbuf *pb; 1246 struct nameidata nd; 1247 int error = 0; 1248 int fd; 1249 1250 if (ktrenter(l)) 1251 return EAGAIN; 1252 1253 if (KTROP(SCARG(uap, ops)) != KTROP_CLEAR) { 1254 /* 1255 * an operation which requires a file argument. 1256 */ 1257 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1258 if (error) { 1259 ktrexit(l); 1260 return (error); 1261 } 1262 NDINIT(&nd, LOOKUP, FOLLOW, pb); 1263 if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) { 1264 pathbuf_destroy(pb); 1265 ktrexit(l); 1266 return (error); 1267 } 1268 vp = nd.ni_vp; 1269 pathbuf_destroy(pb); 1270 VOP_UNLOCK(vp); 1271 if (vp->v_type != VREG) { 1272 vn_close(vp, FREAD|FWRITE, l->l_cred); 1273 ktrexit(l); 1274 return (EACCES); 1275 } 1276 /* 1277 * This uses up a file descriptor slot in the 1278 * tracing process for the duration of this syscall. 1279 * This is not expected to be a problem. 1280 */ 1281 if ((error = fd_allocfile(&fp, &fd)) != 0) { 1282 vn_close(vp, FWRITE, l->l_cred); 1283 ktrexit(l); 1284 return error; 1285 } 1286 fp->f_flag = FWRITE; 1287 fp->f_type = DTYPE_VNODE; 1288 fp->f_ops = &vnops; 1289 fp->f_data = (void *)vp; 1290 vp = NULL; 1291 } 1292 error = ktrace_common(l, SCARG(uap, ops), SCARG(uap, facs), 1293 SCARG(uap, pid), fp); 1294 if (fp != NULL) { 1295 if (error != 0) { 1296 /* File unused. */ 1297 fd_abort(curproc, fp, fd); 1298 } else { 1299 /* File was used. */ 1300 fd_abort(curproc, NULL, fd); 1301 } 1302 } 1303 return (error); 1304 } 1305 1306 int 1307 ktrops(lwp_t *curl, struct proc *p, int ops, int facs, 1308 struct ktr_desc *ktd) 1309 { 1310 int vers = ops & KTRFAC_VER_MASK; 1311 int error = 0; 1312 1313 mutex_enter(p->p_lock); 1314 mutex_enter(&ktrace_lock); 1315 1316 if (!ktrcanset(curl, p)) 1317 goto out; 1318 1319 switch (vers) { 1320 case KTRFACv0: 1321 case KTRFACv1: 1322 case KTRFACv2: 1323 break; 1324 default: 1325 error = EINVAL; 1326 goto out; 1327 } 1328 1329 if (KTROP(ops) == KTROP_SET) { 1330 if (p->p_tracep != ktd) { 1331 /* 1332 * if trace file already in use, relinquish 1333 */ 1334 ktrderef(p); 1335 p->p_tracep = ktd; 1336 ktradref(p); 1337 } 1338 p->p_traceflag |= facs; 1339 if (kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_KTRACE, 1340 p, KAUTH_ARG(KAUTH_REQ_PROCESS_KTRACE_PERSISTENT), NULL, 1341 NULL) == 0) 1342 p->p_traceflag |= KTRFAC_PERSISTENT; 1343 } else { 1344 /* KTROP_CLEAR */ 1345 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) { 1346 /* no more tracing */ 1347 ktrderef(p); 1348 } 1349 } 1350 1351 if (p->p_traceflag) 1352 p->p_traceflag |= vers; 1353 /* 1354 * Emit an emulation record, every time there is a ktrace 1355 * change/attach request. 1356 */ 1357 if (KTRPOINT(p, KTR_EMUL)) 1358 p->p_traceflag |= KTRFAC_TRC_EMUL; 1359 1360 p->p_trace_enabled = trace_is_enabled(p); 1361 #ifdef __HAVE_SYSCALL_INTERN 1362 (*p->p_emul->e_syscall_intern)(p); 1363 #endif 1364 1365 out: 1366 mutex_exit(&ktrace_lock); 1367 mutex_exit(p->p_lock); 1368 1369 return (1); 1370 } 1371 1372 int 1373 ktrsetchildren(lwp_t *curl, struct proc *top, int ops, int facs, 1374 struct ktr_desc *ktd) 1375 { 1376 struct proc *p; 1377 int ret = 0; 1378 1379 KASSERT(mutex_owned(proc_lock)); 1380 1381 p = top; 1382 for (;;) { 1383 ret |= ktrops(curl, p, ops, facs, ktd); 1384 /* 1385 * If this process has children, descend to them next, 1386 * otherwise do any siblings, and if done with this level, 1387 * follow back up the tree (but not past top). 1388 */ 1389 if (LIST_FIRST(&p->p_children) != NULL) { 1390 p = LIST_FIRST(&p->p_children); 1391 continue; 1392 } 1393 for (;;) { 1394 if (p == top) 1395 return (ret); 1396 if (LIST_NEXT(p, p_sibling) != NULL) { 1397 p = LIST_NEXT(p, p_sibling); 1398 break; 1399 } 1400 p = p->p_pptr; 1401 } 1402 } 1403 /*NOTREACHED*/ 1404 } 1405 1406 void 1407 ktrwrite(struct ktr_desc *ktd, struct ktrace_entry *kte) 1408 { 1409 size_t hlen; 1410 struct uio auio; 1411 struct iovec aiov[64], *iov; 1412 struct ktrace_entry *top = kte; 1413 struct ktr_header *kth; 1414 file_t *fp = ktd->ktd_fp; 1415 int error; 1416 next: 1417 auio.uio_iov = iov = &aiov[0]; 1418 auio.uio_offset = 0; 1419 auio.uio_rw = UIO_WRITE; 1420 auio.uio_resid = 0; 1421 auio.uio_iovcnt = 0; 1422 UIO_SETUP_SYSSPACE(&auio); 1423 do { 1424 struct timespec ts; 1425 lwpid_t lid; 1426 kth = &kte->kte_kth; 1427 1428 hlen = sizeof(struct ktr_header); 1429 switch (kth->ktr_version) { 1430 case 0: 1431 ts = kth->ktr_time; 1432 1433 kth->ktr_otv.tv_sec = ts.tv_sec; 1434 kth->ktr_otv.tv_usec = ts.tv_nsec / 1000; 1435 kth->ktr_unused = NULL; 1436 hlen -= sizeof(kth->_v) - 1437 MAX(sizeof(kth->_v._v0), sizeof(kth->_v._v1)); 1438 break; 1439 case 1: 1440 ts = kth->ktr_time; 1441 lid = kth->ktr_lid; 1442 1443 kth->ktr_ots.tv_sec = ts.tv_sec; 1444 kth->ktr_ots.tv_nsec = ts.tv_nsec; 1445 kth->ktr_olid = lid; 1446 hlen -= sizeof(kth->_v) - 1447 MAX(sizeof(kth->_v._v0), sizeof(kth->_v._v1)); 1448 break; 1449 } 1450 iov->iov_base = (void *)kth; 1451 iov++->iov_len = hlen; 1452 auio.uio_resid += hlen; 1453 auio.uio_iovcnt++; 1454 if (kth->ktr_len > 0) { 1455 iov->iov_base = kte->kte_buf; 1456 iov++->iov_len = kth->ktr_len; 1457 auio.uio_resid += kth->ktr_len; 1458 auio.uio_iovcnt++; 1459 } 1460 } while ((kte = TAILQ_NEXT(kte, kte_list)) != NULL && 1461 auio.uio_iovcnt < sizeof(aiov) / sizeof(aiov[0]) - 1); 1462 1463 again: 1464 error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio, 1465 fp->f_cred, FOF_UPDATE_OFFSET); 1466 switch (error) { 1467 1468 case 0: 1469 if (auio.uio_resid > 0) 1470 goto again; 1471 if (kte != NULL) 1472 goto next; 1473 break; 1474 1475 case EWOULDBLOCK: 1476 kpause("ktrzzz", false, 1, NULL); 1477 goto again; 1478 1479 default: 1480 /* 1481 * If error encountered, give up tracing on this 1482 * vnode. Don't report EPIPE as this can easily 1483 * happen with fktrace()/ktruss. 1484 */ 1485 #ifndef DEBUG 1486 if (error != EPIPE) 1487 #endif 1488 log(LOG_NOTICE, 1489 "ktrace write failed, errno %d, tracing stopped\n", 1490 error); 1491 (void)ktrderefall(ktd, 0); 1492 } 1493 1494 while ((kte = top) != NULL) { 1495 top = TAILQ_NEXT(top, kte_list); 1496 ktefree(kte); 1497 } 1498 } 1499 1500 void 1501 ktrace_thread(void *arg) 1502 { 1503 struct ktr_desc *ktd = arg; 1504 file_t *fp = ktd->ktd_fp; 1505 struct ktrace_entry *kte; 1506 int ktrerr, errcnt; 1507 1508 mutex_enter(&ktrace_lock); 1509 for (;;) { 1510 kte = TAILQ_FIRST(&ktd->ktd_queue); 1511 if (kte == NULL) { 1512 if (ktd->ktd_flags & KTDF_WAIT) { 1513 ktd->ktd_flags &= ~(KTDF_WAIT | KTDF_BLOCKING); 1514 cv_broadcast(&ktd->ktd_sync_cv); 1515 } 1516 if (ktd->ktd_ref == 0) 1517 break; 1518 cv_wait(&ktd->ktd_cv, &ktrace_lock); 1519 continue; 1520 } 1521 TAILQ_INIT(&ktd->ktd_queue); 1522 ktd->ktd_qcount = 0; 1523 ktrerr = ktd->ktd_error; 1524 errcnt = ktd->ktd_errcnt; 1525 ktd->ktd_error = ktd->ktd_errcnt = 0; 1526 mutex_exit(&ktrace_lock); 1527 1528 if (ktrerr) { 1529 log(LOG_NOTICE, 1530 "ktrace failed, fp %p, error 0x%x, total %d\n", 1531 fp, ktrerr, errcnt); 1532 } 1533 ktrwrite(ktd, kte); 1534 mutex_enter(&ktrace_lock); 1535 } 1536 1537 TAILQ_REMOVE(&ktdq, ktd, ktd_list); 1538 mutex_exit(&ktrace_lock); 1539 1540 /* 1541 * ktrace file descriptor can't be watched (are not visible to 1542 * userspace), so no kqueue stuff here 1543 * XXX: The above comment is wrong, because the fktrace file 1544 * descriptor is available in userland. 1545 */ 1546 closef(fp); 1547 1548 cv_destroy(&ktd->ktd_sync_cv); 1549 cv_destroy(&ktd->ktd_cv); 1550 1551 callout_stop(&ktd->ktd_wakch); 1552 callout_destroy(&ktd->ktd_wakch); 1553 kmem_free(ktd, sizeof(*ktd)); 1554 1555 kthread_exit(0); 1556 } 1557 1558 /* 1559 * Return true if caller has permission to set the ktracing state 1560 * of target. Essentially, the target can't possess any 1561 * more permissions than the caller. KTRFAC_PERSISTENT signifies that 1562 * the tracing will persist on sugid processes during exec; it is only 1563 * settable by a process with appropriate credentials. 1564 * 1565 * TODO: check groups. use caller effective gid. 1566 */ 1567 int 1568 ktrcanset(lwp_t *calll, struct proc *targetp) 1569 { 1570 KASSERT(mutex_owned(targetp->p_lock)); 1571 KASSERT(mutex_owned(&ktrace_lock)); 1572 1573 if (kauth_authorize_process(calll->l_cred, KAUTH_PROCESS_KTRACE, 1574 targetp, NULL, NULL, NULL) == 0) 1575 return (1); 1576 1577 return (0); 1578 } 1579 1580 /* 1581 * Put user defined entry to ktrace records. 1582 */ 1583 int 1584 sys_utrace(struct lwp *l, const struct sys_utrace_args *uap, register_t *retval) 1585 { 1586 /* { 1587 syscallarg(const char *) label; 1588 syscallarg(void *) addr; 1589 syscallarg(size_t) len; 1590 } */ 1591 1592 return ktruser(SCARG(uap, label), SCARG(uap, addr), 1593 SCARG(uap, len), 1); 1594 } 1595