1 /* $NetBSD: kern_ktrace.c,v 1.156 2011/04/27 00:00:46 joerg 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 1.156 2011/04/27 00:00:46 joerg 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 static void 677 ktr_kmem(lwp_t *l, int type, const void *bf, size_t len) 678 { 679 struct ktrace_entry *kte; 680 void *buf; 681 682 if (ktealloc(&kte, &buf, l, type, len)) 683 return; 684 memcpy(buf, bf, len); 685 ktraddentry(l, kte, KTA_WAITOK); 686 } 687 688 static void 689 ktr_io(lwp_t *l, int fd, enum uio_rw rw, struct iovec *iov, size_t len) 690 { 691 struct ktrace_entry *kte; 692 struct ktr_genio *ktp; 693 size_t resid = len, cnt, buflen; 694 char *cp; 695 696 next: 697 buflen = min(PAGE_SIZE, resid + sizeof(struct ktr_genio)); 698 699 if (ktealloc(&kte, (void *)&ktp, l, KTR_GENIO, buflen)) 700 return; 701 702 ktp->ktr_fd = fd; 703 ktp->ktr_rw = rw; 704 705 cp = (void *)(ktp + 1); 706 buflen -= sizeof(struct ktr_genio); 707 kte->kte_kth.ktr_len = sizeof(struct ktr_genio); 708 709 while (buflen > 0) { 710 cnt = min(iov->iov_len, buflen); 711 if (copyin(iov->iov_base, cp, cnt) != 0) 712 goto out; 713 kte->kte_kth.ktr_len += cnt; 714 cp += cnt; 715 buflen -= cnt; 716 resid -= cnt; 717 iov->iov_len -= cnt; 718 if (iov->iov_len == 0) 719 iov++; 720 else 721 iov->iov_base = (char *)iov->iov_base + cnt; 722 } 723 724 /* 725 * Don't push so many entry at once. It will cause kmem map 726 * shortage. 727 */ 728 ktraddentry(l, kte, KTA_WAITOK | KTA_LARGE); 729 if (resid > 0) { 730 if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD) { 731 (void)ktrenter(l); 732 preempt(); 733 ktrexit(l); 734 } 735 736 goto next; 737 } 738 739 return; 740 741 out: 742 ktefree(kte); 743 ktrexit(l); 744 } 745 746 void 747 ktr_genio(int fd, enum uio_rw rw, const void *addr, size_t len, int error) 748 { 749 lwp_t *l = curlwp; 750 struct iovec iov; 751 752 if (!KTRPOINT(l->l_proc, KTR_GENIO) || error != 0) 753 return; 754 iov.iov_base = __UNCONST(addr); 755 iov.iov_len = len; 756 ktr_io(l, fd, rw, &iov, len); 757 } 758 759 void 760 ktr_geniov(int fd, enum uio_rw rw, struct iovec *iov, size_t len, int error) 761 { 762 lwp_t *l = curlwp; 763 764 if (!KTRPOINT(l->l_proc, KTR_GENIO) || error != 0) 765 return; 766 ktr_io(l, fd, rw, iov, len); 767 } 768 769 void 770 ktr_mibio(int fd, enum uio_rw rw, const void *addr, size_t len, int error) 771 { 772 lwp_t *l = curlwp; 773 struct iovec iov; 774 775 if (!KTRPOINT(l->l_proc, KTR_MIB) || error != 0) 776 return; 777 iov.iov_base = __UNCONST(addr); 778 iov.iov_len = len; 779 ktr_io(l, fd, rw, &iov, len); 780 } 781 782 void 783 ktr_psig(int sig, sig_t action, const sigset_t *mask, 784 const ksiginfo_t *ksi) 785 { 786 struct ktrace_entry *kte; 787 lwp_t *l = curlwp; 788 struct { 789 struct ktr_psig kp; 790 siginfo_t si; 791 } *kbuf; 792 793 if (!KTRPOINT(l->l_proc, KTR_PSIG)) 794 return; 795 796 if (ktealloc(&kte, (void *)&kbuf, l, KTR_PSIG, sizeof(*kbuf))) 797 return; 798 799 kbuf->kp.signo = (char)sig; 800 kbuf->kp.action = action; 801 kbuf->kp.mask = *mask; 802 803 if (ksi) { 804 kbuf->kp.code = KSI_TRAPCODE(ksi); 805 (void)memset(&kbuf->si, 0, sizeof(kbuf->si)); 806 kbuf->si._info = ksi->ksi_info; 807 kte->kte_kth.ktr_len = sizeof(*kbuf); 808 } else { 809 kbuf->kp.code = 0; 810 kte->kte_kth.ktr_len = sizeof(struct ktr_psig); 811 } 812 813 ktraddentry(l, kte, KTA_WAITOK); 814 } 815 816 void 817 ktr_csw(int out, int user) 818 { 819 lwp_t *l = curlwp; 820 struct proc *p = l->l_proc; 821 struct ktrace_entry *kte; 822 struct ktr_csw *kc; 823 824 if (!KTRPOINT(p, KTR_CSW)) 825 return; 826 827 /* 828 * Don't record context switches resulting from blocking on 829 * locks; it's too easy to get duff results. 830 */ 831 if (l->l_syncobj == &mutex_syncobj || l->l_syncobj == &rw_syncobj) 832 return; 833 834 /* 835 * We can't sleep if we're already going to sleep (if original 836 * condition is met during sleep, we hang up). 837 * 838 * XXX This is not ideal: it would be better to maintain a pool 839 * of ktes and actually push this to the kthread when context 840 * switch happens, however given the points where we are called 841 * from that is difficult to do. 842 */ 843 if (out) { 844 if (ktrenter(l)) 845 return; 846 847 nanotime(&l->l_ktrcsw); 848 l->l_pflag |= LP_KTRCSW; 849 if (user) 850 l->l_pflag |= LP_KTRCSWUSER; 851 else 852 l->l_pflag &= ~LP_KTRCSWUSER; 853 854 ktrexit(l); 855 return; 856 } 857 858 /* 859 * On the way back in, we need to record twice: once for entry, and 860 * once for exit. 861 */ 862 if ((l->l_pflag & LP_KTRCSW) != 0) { 863 struct timespec *ts; 864 l->l_pflag &= ~LP_KTRCSW; 865 866 if (ktealloc(&kte, (void *)&kc, l, KTR_CSW, sizeof(*kc))) 867 return; 868 869 kc->out = 1; 870 kc->user = ((l->l_pflag & LP_KTRCSWUSER) != 0); 871 872 ts = &l->l_ktrcsw; 873 switch (KTRFAC_VERSION(p->p_traceflag)) { 874 case 0: 875 kte->kte_kth.ktr_otv.tv_sec = ts->tv_sec; 876 kte->kte_kth.ktr_otv.tv_usec = ts->tv_nsec / 1000; 877 break; 878 case 1: 879 kte->kte_kth.ktr_ots.tv_sec = ts->tv_sec; 880 kte->kte_kth.ktr_ots.tv_nsec = ts->tv_nsec; 881 break; 882 case 2: 883 kte->kte_kth.ktr_ts.tv_sec = ts->tv_sec; 884 kte->kte_kth.ktr_ts.tv_nsec = ts->tv_nsec; 885 break; 886 default: 887 break; 888 } 889 890 ktraddentry(l, kte, KTA_WAITOK); 891 } 892 893 if (ktealloc(&kte, (void *)&kc, l, KTR_CSW, sizeof(*kc))) 894 return; 895 896 kc->out = 0; 897 kc->user = user; 898 899 ktraddentry(l, kte, KTA_WAITOK); 900 } 901 902 bool 903 ktr_point(int fac_bit) 904 { 905 return curlwp->l_proc->p_traceflag & fac_bit; 906 } 907 908 int 909 ktruser(const char *id, void *addr, size_t len, int ustr) 910 { 911 struct ktrace_entry *kte; 912 struct ktr_user *ktp; 913 lwp_t *l = curlwp; 914 void *user_dta; 915 int error; 916 917 if (!KTRPOINT(l->l_proc, KTR_USER)) 918 return 0; 919 920 if (len > KTR_USER_MAXLEN) 921 return ENOSPC; 922 923 error = ktealloc(&kte, (void *)&ktp, l, KTR_USER, sizeof(*ktp) + len); 924 if (error != 0) 925 return error; 926 927 if (ustr) { 928 if (copyinstr(id, ktp->ktr_id, KTR_USER_MAXIDLEN, NULL) != 0) 929 ktp->ktr_id[0] = '\0'; 930 } else 931 strncpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN); 932 ktp->ktr_id[KTR_USER_MAXIDLEN-1] = '\0'; 933 934 user_dta = (void *)(ktp + 1); 935 if ((error = copyin(addr, (void *)user_dta, len)) != 0) 936 len = 0; 937 938 ktraddentry(l, kte, KTA_WAITOK); 939 return error; 940 } 941 942 void 943 ktr_kuser(const char *id, void *addr, size_t len) 944 { 945 struct ktrace_entry *kte; 946 struct ktr_user *ktp; 947 lwp_t *l = curlwp; 948 int error; 949 950 if (!KTRPOINT(l->l_proc, KTR_USER)) 951 return; 952 953 if (len > KTR_USER_MAXLEN) 954 return; 955 956 error = ktealloc(&kte, (void *)&ktp, l, KTR_USER, sizeof(*ktp) + len); 957 if (error != 0) 958 return; 959 960 strlcpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN); 961 962 memcpy(ktp + 1, addr, len); 963 964 ktraddentry(l, kte, KTA_WAITOK); 965 } 966 967 void 968 ktr_saupcall(struct lwp *l, int type, int nevent, int nint, void *sas, 969 void *ap, void *ksas) 970 { 971 struct ktrace_entry *kte; 972 struct ktr_saupcall *ktp; 973 size_t len, sz; 974 struct sa_t **sapp; 975 int i; 976 977 if (!KTRPOINT(l->l_proc, KTR_SAUPCALL)) 978 return; 979 980 len = sizeof(struct ktr_saupcall); 981 sz = len + sizeof(struct sa_t) * (nevent + nint + 1); 982 983 if (ktealloc(&kte, (void *)&ktp, l, KTR_SAUPCALL, sz)) 984 return; 985 986 ktp->ktr_type = type; 987 ktp->ktr_nevent = nevent; 988 ktp->ktr_nint = nint; 989 ktp->ktr_sas = sas; 990 ktp->ktr_ap = ap; 991 992 /* Copy the sa_t's */ 993 sapp = (struct sa_t **) ksas; 994 995 for (i = nevent + nint; i >= 0; i--) { 996 memcpy((char *)ktp + len, *sapp, sizeof(struct sa_t)); 997 len += sizeof(struct sa_t); 998 sapp++; 999 } 1000 1001 kte->kte_kth.ktr_len = len; 1002 ktraddentry(l, kte, KTA_WAITOK); 1003 } 1004 1005 void 1006 ktr_mib(const int *name, u_int namelen) 1007 { 1008 struct ktrace_entry *kte; 1009 int *namep; 1010 size_t size; 1011 lwp_t *l = curlwp; 1012 1013 if (!KTRPOINT(l->l_proc, KTR_MIB)) 1014 return; 1015 1016 size = namelen * sizeof(*name); 1017 1018 if (ktealloc(&kte, (void *)&namep, l, KTR_MIB, size)) 1019 return; 1020 1021 (void)memcpy(namep, name, namelen * sizeof(*name)); 1022 1023 ktraddentry(l, kte, KTA_WAITOK); 1024 } 1025 1026 /* Interface and common routines */ 1027 1028 int 1029 ktrace_common(lwp_t *curl, int ops, int facs, int pid, file_t *fp) 1030 { 1031 struct proc *curp; 1032 struct proc *p; 1033 struct pgrp *pg; 1034 struct ktr_desc *ktd = NULL; 1035 int ret = 0; 1036 int error = 0; 1037 int descend; 1038 1039 curp = curl->l_proc; 1040 descend = ops & KTRFLAG_DESCEND; 1041 facs = facs & ~((unsigned) KTRFAC_PERSISTENT); 1042 1043 (void)ktrenter(curl); 1044 1045 switch (KTROP(ops)) { 1046 1047 case KTROP_CLEARFILE: 1048 /* 1049 * Clear all uses of the tracefile 1050 */ 1051 mutex_enter(&ktrace_lock); 1052 ktd = ktd_lookup(fp); 1053 mutex_exit(&ktrace_lock); 1054 if (ktd == NULL) 1055 goto done; 1056 error = ktrderefall(ktd, 1); 1057 goto done; 1058 1059 case KTROP_SET: 1060 mutex_enter(&ktrace_lock); 1061 ktd = ktd_lookup(fp); 1062 mutex_exit(&ktrace_lock); 1063 if (ktd == NULL) { 1064 ktd = kmem_alloc(sizeof(*ktd), KM_SLEEP); 1065 TAILQ_INIT(&ktd->ktd_queue); 1066 callout_init(&ktd->ktd_wakch, CALLOUT_MPSAFE); 1067 cv_init(&ktd->ktd_cv, "ktrwait"); 1068 cv_init(&ktd->ktd_sync_cv, "ktrsync"); 1069 ktd->ktd_flags = 0; 1070 ktd->ktd_qcount = 0; 1071 ktd->ktd_error = 0; 1072 ktd->ktd_errcnt = 0; 1073 ktd->ktd_delayqcnt = ktd_delayqcnt; 1074 ktd->ktd_wakedelay = mstohz(ktd_wakedelay); 1075 ktd->ktd_intrwakdl = mstohz(ktd_intrwakdl); 1076 ktd->ktd_ref = 0; 1077 ktd->ktd_fp = fp; 1078 mutex_enter(&ktrace_lock); 1079 ktdref(ktd); 1080 mutex_exit(&ktrace_lock); 1081 1082 /* 1083 * XXX: not correct. needs an way to detect 1084 * whether ktruss or ktrace. 1085 */ 1086 if (fp->f_type == DTYPE_PIPE) 1087 ktd->ktd_flags |= KTDF_INTERACTIVE; 1088 1089 mutex_enter(&fp->f_lock); 1090 fp->f_count++; 1091 mutex_exit(&fp->f_lock); 1092 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 1093 ktrace_thread, ktd, &ktd->ktd_lwp, "ktrace"); 1094 if (error != 0) { 1095 kmem_free(ktd, sizeof(*ktd)); 1096 mutex_enter(&fp->f_lock); 1097 fp->f_count--; 1098 mutex_exit(&fp->f_lock); 1099 goto done; 1100 } 1101 1102 mutex_enter(&ktrace_lock); 1103 if (ktd_lookup(fp) != NULL) { 1104 ktdrel(ktd); 1105 ktd = NULL; 1106 } else 1107 TAILQ_INSERT_TAIL(&ktdq, ktd, ktd_list); 1108 if (ktd == NULL) 1109 cv_wait(&lbolt, &ktrace_lock); 1110 mutex_exit(&ktrace_lock); 1111 if (ktd == NULL) 1112 goto done; 1113 } 1114 break; 1115 1116 case KTROP_CLEAR: 1117 break; 1118 } 1119 1120 /* 1121 * need something to (un)trace (XXX - why is this here?) 1122 */ 1123 if (!facs) { 1124 error = EINVAL; 1125 goto done; 1126 } 1127 1128 /* 1129 * do it 1130 */ 1131 mutex_enter(proc_lock); 1132 if (pid < 0) { 1133 /* 1134 * by process group 1135 */ 1136 pg = pgrp_find(-pid); 1137 if (pg == NULL) 1138 error = ESRCH; 1139 else { 1140 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1141 if (descend) 1142 ret |= ktrsetchildren(curl, p, ops, 1143 facs, ktd); 1144 else 1145 ret |= ktrops(curl, p, ops, facs, 1146 ktd); 1147 } 1148 } 1149 1150 } else { 1151 /* 1152 * by pid 1153 */ 1154 p = proc_find(pid); 1155 if (p == NULL) 1156 error = ESRCH; 1157 else if (descend) 1158 ret |= ktrsetchildren(curl, p, ops, facs, ktd); 1159 else 1160 ret |= ktrops(curl, p, ops, facs, ktd); 1161 } 1162 mutex_exit(proc_lock); 1163 if (error == 0 && !ret) 1164 error = EPERM; 1165 done: 1166 if (ktd != NULL) { 1167 mutex_enter(&ktrace_lock); 1168 if (error != 0) { 1169 /* 1170 * Wakeup the thread so that it can be die if we 1171 * can't trace any process. 1172 */ 1173 ktd_wakeup(ktd); 1174 } 1175 if (KTROP(ops) == KTROP_SET || KTROP(ops) == KTROP_CLEARFILE) 1176 ktdrel(ktd); 1177 mutex_exit(&ktrace_lock); 1178 } 1179 ktrexit(curl); 1180 return (error); 1181 } 1182 1183 /* 1184 * fktrace system call 1185 */ 1186 /* ARGSUSED */ 1187 int 1188 sys_fktrace(struct lwp *l, const struct sys_fktrace_args *uap, register_t *retval) 1189 { 1190 /* { 1191 syscallarg(int) fd; 1192 syscallarg(int) ops; 1193 syscallarg(int) facs; 1194 syscallarg(int) pid; 1195 } */ 1196 file_t *fp; 1197 int error, fd; 1198 1199 fd = SCARG(uap, fd); 1200 if ((fp = fd_getfile(fd)) == NULL) 1201 return (EBADF); 1202 if ((fp->f_flag & FWRITE) == 0) 1203 error = EBADF; 1204 else 1205 error = ktrace_common(l, SCARG(uap, ops), 1206 SCARG(uap, facs), SCARG(uap, pid), fp); 1207 fd_putfile(fd); 1208 return error; 1209 } 1210 1211 /* 1212 * ktrace system call 1213 */ 1214 /* ARGSUSED */ 1215 int 1216 sys_ktrace(struct lwp *l, const struct sys_ktrace_args *uap, register_t *retval) 1217 { 1218 /* { 1219 syscallarg(const char *) fname; 1220 syscallarg(int) ops; 1221 syscallarg(int) facs; 1222 syscallarg(int) pid; 1223 } */ 1224 struct vnode *vp = NULL; 1225 file_t *fp = NULL; 1226 struct pathbuf *pb; 1227 struct nameidata nd; 1228 int error = 0; 1229 int fd; 1230 1231 if (ktrenter(l)) 1232 return EAGAIN; 1233 1234 if (KTROP(SCARG(uap, ops)) != KTROP_CLEAR) { 1235 /* 1236 * an operation which requires a file argument. 1237 */ 1238 error = pathbuf_copyin(SCARG(uap, fname), &pb); 1239 if (error) { 1240 ktrexit(l); 1241 return (error); 1242 } 1243 NDINIT(&nd, LOOKUP, FOLLOW, pb); 1244 if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) { 1245 pathbuf_destroy(pb); 1246 ktrexit(l); 1247 return (error); 1248 } 1249 vp = nd.ni_vp; 1250 pathbuf_destroy(pb); 1251 VOP_UNLOCK(vp); 1252 if (vp->v_type != VREG) { 1253 vn_close(vp, FREAD|FWRITE, l->l_cred); 1254 ktrexit(l); 1255 return (EACCES); 1256 } 1257 /* 1258 * This uses up a file descriptor slot in the 1259 * tracing process for the duration of this syscall. 1260 * This is not expected to be a problem. 1261 */ 1262 if ((error = fd_allocfile(&fp, &fd)) != 0) { 1263 vn_close(vp, FWRITE, l->l_cred); 1264 ktrexit(l); 1265 return error; 1266 } 1267 fp->f_flag = FWRITE; 1268 fp->f_type = DTYPE_VNODE; 1269 fp->f_ops = &vnops; 1270 fp->f_data = (void *)vp; 1271 vp = NULL; 1272 } 1273 error = ktrace_common(l, SCARG(uap, ops), SCARG(uap, facs), 1274 SCARG(uap, pid), fp); 1275 if (fp != NULL) { 1276 if (error != 0) { 1277 /* File unused. */ 1278 fd_abort(curproc, fp, fd); 1279 } else { 1280 /* File was used. */ 1281 fd_abort(curproc, NULL, fd); 1282 } 1283 } 1284 return (error); 1285 } 1286 1287 int 1288 ktrops(lwp_t *curl, struct proc *p, int ops, int facs, 1289 struct ktr_desc *ktd) 1290 { 1291 int vers = ops & KTRFAC_VER_MASK; 1292 int error = 0; 1293 1294 mutex_enter(p->p_lock); 1295 mutex_enter(&ktrace_lock); 1296 1297 if (!ktrcanset(curl, p)) 1298 goto out; 1299 1300 switch (vers) { 1301 case KTRFACv0: 1302 case KTRFACv1: 1303 case KTRFACv2: 1304 break; 1305 default: 1306 error = EINVAL; 1307 goto out; 1308 } 1309 1310 if (KTROP(ops) == KTROP_SET) { 1311 if (p->p_tracep != ktd) { 1312 /* 1313 * if trace file already in use, relinquish 1314 */ 1315 ktrderef(p); 1316 p->p_tracep = ktd; 1317 ktradref(p); 1318 } 1319 p->p_traceflag |= facs; 1320 if (kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_KTRACE, 1321 p, KAUTH_ARG(KAUTH_REQ_PROCESS_KTRACE_PERSISTENT), NULL, 1322 NULL) == 0) 1323 p->p_traceflag |= KTRFAC_PERSISTENT; 1324 } else { 1325 /* KTROP_CLEAR */ 1326 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) { 1327 /* no more tracing */ 1328 ktrderef(p); 1329 } 1330 } 1331 1332 if (p->p_traceflag) 1333 p->p_traceflag |= vers; 1334 /* 1335 * Emit an emulation record, every time there is a ktrace 1336 * change/attach request. 1337 */ 1338 if (KTRPOINT(p, KTR_EMUL)) 1339 p->p_traceflag |= KTRFAC_TRC_EMUL; 1340 1341 p->p_trace_enabled = trace_is_enabled(p); 1342 #ifdef __HAVE_SYSCALL_INTERN 1343 (*p->p_emul->e_syscall_intern)(p); 1344 #endif 1345 1346 out: 1347 mutex_exit(&ktrace_lock); 1348 mutex_exit(p->p_lock); 1349 1350 return (1); 1351 } 1352 1353 int 1354 ktrsetchildren(lwp_t *curl, struct proc *top, int ops, int facs, 1355 struct ktr_desc *ktd) 1356 { 1357 struct proc *p; 1358 int ret = 0; 1359 1360 KASSERT(mutex_owned(proc_lock)); 1361 1362 p = top; 1363 for (;;) { 1364 ret |= ktrops(curl, p, ops, facs, ktd); 1365 /* 1366 * If this process has children, descend to them next, 1367 * otherwise do any siblings, and if done with this level, 1368 * follow back up the tree (but not past top). 1369 */ 1370 if (LIST_FIRST(&p->p_children) != NULL) { 1371 p = LIST_FIRST(&p->p_children); 1372 continue; 1373 } 1374 for (;;) { 1375 if (p == top) 1376 return (ret); 1377 if (LIST_NEXT(p, p_sibling) != NULL) { 1378 p = LIST_NEXT(p, p_sibling); 1379 break; 1380 } 1381 p = p->p_pptr; 1382 } 1383 } 1384 /*NOTREACHED*/ 1385 } 1386 1387 void 1388 ktrwrite(struct ktr_desc *ktd, struct ktrace_entry *kte) 1389 { 1390 size_t hlen; 1391 struct uio auio; 1392 struct iovec aiov[64], *iov; 1393 struct ktrace_entry *top = kte; 1394 struct ktr_header *kth; 1395 file_t *fp = ktd->ktd_fp; 1396 int error; 1397 next: 1398 auio.uio_iov = iov = &aiov[0]; 1399 auio.uio_offset = 0; 1400 auio.uio_rw = UIO_WRITE; 1401 auio.uio_resid = 0; 1402 auio.uio_iovcnt = 0; 1403 UIO_SETUP_SYSSPACE(&auio); 1404 do { 1405 struct timespec ts; 1406 lwpid_t lid; 1407 kth = &kte->kte_kth; 1408 1409 hlen = sizeof(struct ktr_header); 1410 switch (kth->ktr_version) { 1411 case 0: 1412 ts = kth->ktr_time; 1413 1414 kth->ktr_otv.tv_sec = ts.tv_sec; 1415 kth->ktr_otv.tv_usec = ts.tv_nsec / 1000; 1416 kth->ktr_unused = NULL; 1417 hlen -= sizeof(kth->_v) - 1418 MAX(sizeof(kth->_v._v0), sizeof(kth->_v._v1)); 1419 break; 1420 case 1: 1421 ts = kth->ktr_time; 1422 lid = kth->ktr_lid; 1423 1424 kth->ktr_ots.tv_sec = ts.tv_sec; 1425 kth->ktr_ots.tv_nsec = ts.tv_nsec; 1426 kth->ktr_olid = lid; 1427 hlen -= sizeof(kth->_v) - 1428 MAX(sizeof(kth->_v._v0), sizeof(kth->_v._v1)); 1429 break; 1430 } 1431 iov->iov_base = (void *)kth; 1432 iov++->iov_len = hlen; 1433 auio.uio_resid += hlen; 1434 auio.uio_iovcnt++; 1435 if (kth->ktr_len > 0) { 1436 iov->iov_base = kte->kte_buf; 1437 iov++->iov_len = kth->ktr_len; 1438 auio.uio_resid += kth->ktr_len; 1439 auio.uio_iovcnt++; 1440 } 1441 } while ((kte = TAILQ_NEXT(kte, kte_list)) != NULL && 1442 auio.uio_iovcnt < sizeof(aiov) / sizeof(aiov[0]) - 1); 1443 1444 again: 1445 error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio, 1446 fp->f_cred, FOF_UPDATE_OFFSET); 1447 switch (error) { 1448 1449 case 0: 1450 if (auio.uio_resid > 0) 1451 goto again; 1452 if (kte != NULL) 1453 goto next; 1454 break; 1455 1456 case EWOULDBLOCK: 1457 kpause("ktrzzz", false, 1, NULL); 1458 goto again; 1459 1460 default: 1461 /* 1462 * If error encountered, give up tracing on this 1463 * vnode. Don't report EPIPE as this can easily 1464 * happen with fktrace()/ktruss. 1465 */ 1466 #ifndef DEBUG 1467 if (error != EPIPE) 1468 #endif 1469 log(LOG_NOTICE, 1470 "ktrace write failed, errno %d, tracing stopped\n", 1471 error); 1472 (void)ktrderefall(ktd, 0); 1473 } 1474 1475 while ((kte = top) != NULL) { 1476 top = TAILQ_NEXT(top, kte_list); 1477 ktefree(kte); 1478 } 1479 } 1480 1481 void 1482 ktrace_thread(void *arg) 1483 { 1484 struct ktr_desc *ktd = arg; 1485 file_t *fp = ktd->ktd_fp; 1486 struct ktrace_entry *kte; 1487 int ktrerr, errcnt; 1488 1489 mutex_enter(&ktrace_lock); 1490 for (;;) { 1491 kte = TAILQ_FIRST(&ktd->ktd_queue); 1492 if (kte == NULL) { 1493 if (ktd->ktd_flags & KTDF_WAIT) { 1494 ktd->ktd_flags &= ~(KTDF_WAIT | KTDF_BLOCKING); 1495 cv_broadcast(&ktd->ktd_sync_cv); 1496 } 1497 if (ktd->ktd_ref == 0) 1498 break; 1499 cv_wait(&ktd->ktd_cv, &ktrace_lock); 1500 continue; 1501 } 1502 TAILQ_INIT(&ktd->ktd_queue); 1503 ktd->ktd_qcount = 0; 1504 ktrerr = ktd->ktd_error; 1505 errcnt = ktd->ktd_errcnt; 1506 ktd->ktd_error = ktd->ktd_errcnt = 0; 1507 mutex_exit(&ktrace_lock); 1508 1509 if (ktrerr) { 1510 log(LOG_NOTICE, 1511 "ktrace failed, fp %p, error 0x%x, total %d\n", 1512 fp, ktrerr, errcnt); 1513 } 1514 ktrwrite(ktd, kte); 1515 mutex_enter(&ktrace_lock); 1516 } 1517 1518 TAILQ_REMOVE(&ktdq, ktd, ktd_list); 1519 mutex_exit(&ktrace_lock); 1520 1521 /* 1522 * ktrace file descriptor can't be watched (are not visible to 1523 * userspace), so no kqueue stuff here 1524 * XXX: The above comment is wrong, because the fktrace file 1525 * descriptor is available in userland. 1526 */ 1527 closef(fp); 1528 1529 cv_destroy(&ktd->ktd_sync_cv); 1530 cv_destroy(&ktd->ktd_cv); 1531 1532 callout_stop(&ktd->ktd_wakch); 1533 callout_destroy(&ktd->ktd_wakch); 1534 kmem_free(ktd, sizeof(*ktd)); 1535 1536 kthread_exit(0); 1537 } 1538 1539 /* 1540 * Return true if caller has permission to set the ktracing state 1541 * of target. Essentially, the target can't possess any 1542 * more permissions than the caller. KTRFAC_PERSISTENT signifies that 1543 * the tracing will persist on sugid processes during exec; it is only 1544 * settable by a process with appropriate credentials. 1545 * 1546 * TODO: check groups. use caller effective gid. 1547 */ 1548 int 1549 ktrcanset(lwp_t *calll, struct proc *targetp) 1550 { 1551 KASSERT(mutex_owned(targetp->p_lock)); 1552 KASSERT(mutex_owned(&ktrace_lock)); 1553 1554 if (kauth_authorize_process(calll->l_cred, KAUTH_PROCESS_KTRACE, 1555 targetp, NULL, NULL, NULL) == 0) 1556 return (1); 1557 1558 return (0); 1559 } 1560 1561 /* 1562 * Put user defined entry to ktrace records. 1563 */ 1564 int 1565 sys_utrace(struct lwp *l, const struct sys_utrace_args *uap, register_t *retval) 1566 { 1567 /* { 1568 syscallarg(const char *) label; 1569 syscallarg(void *) addr; 1570 syscallarg(size_t) len; 1571 } */ 1572 1573 return ktruser(SCARG(uap, label), SCARG(uap, addr), 1574 SCARG(uap, len), 1); 1575 } 1576