1 /* $NetBSD: kern_ktrace.c,v 1.142 2008/04/24 18:39:24 ad 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 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39 /* 40 * Copyright (c) 1989, 1993 41 * The Regents of the University of California. All rights reserved. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. Neither the name of the University nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 65 * SUCH DAMAGE. 66 * 67 * @(#)kern_ktrace.c 8.5 (Berkeley) 5/14/95 68 */ 69 70 #include <sys/cdefs.h> 71 __KERNEL_RCSID(0, "$NetBSD: kern_ktrace.c,v 1.142 2008/04/24 18:39:24 ad Exp $"); 72 73 #include <sys/param.h> 74 #include <sys/systm.h> 75 #include <sys/proc.h> 76 #include <sys/file.h> 77 #include <sys/namei.h> 78 #include <sys/vnode.h> 79 #include <sys/kernel.h> 80 #include <sys/kthread.h> 81 #include <sys/ktrace.h> 82 #include <sys/kmem.h> 83 #include <sys/syslog.h> 84 #include <sys/filedesc.h> 85 #include <sys/ioctl.h> 86 #include <sys/callout.h> 87 #include <sys/kauth.h> 88 89 #include <sys/mount.h> 90 #include <sys/syscallargs.h> 91 92 /* 93 * TODO: 94 * - need better error reporting? 95 * - userland utility to sort ktrace.out by timestamp. 96 * - keep minimum information in ktrace_entry when rest of alloc failed. 97 * - per trace control of configurable parameters. 98 */ 99 100 struct ktrace_entry { 101 TAILQ_ENTRY(ktrace_entry) kte_list; 102 struct ktr_header kte_kth; 103 void *kte_buf; 104 size_t kte_bufsz; 105 #define KTE_SPACE 32 106 uint8_t kte_space[KTE_SPACE]; 107 }; 108 109 struct ktr_desc { 110 TAILQ_ENTRY(ktr_desc) ktd_list; 111 int ktd_flags; 112 #define KTDF_WAIT 0x0001 113 #define KTDF_DONE 0x0002 114 #define KTDF_BLOCKING 0x0004 115 #define KTDF_INTERACTIVE 0x0008 116 int ktd_error; 117 #define KTDE_ENOMEM 0x0001 118 #define KTDE_ENOSPC 0x0002 119 int ktd_errcnt; 120 int ktd_ref; /* # of reference */ 121 int ktd_qcount; /* # of entry in the queue */ 122 123 /* 124 * Params to control behaviour. 125 */ 126 int ktd_delayqcnt; /* # of entry allowed to delay */ 127 int ktd_wakedelay; /* delay of wakeup in *tick* */ 128 int ktd_intrwakdl; /* ditto, but when interactive */ 129 130 file_t *ktd_fp; /* trace output file */ 131 lwp_t *ktd_lwp; /* our kernel thread */ 132 TAILQ_HEAD(, ktrace_entry) ktd_queue; 133 callout_t ktd_wakch; /* delayed wakeup */ 134 kcondvar_t ktd_sync_cv; 135 kcondvar_t ktd_cv; 136 }; 137 138 static int ktealloc(struct ktrace_entry **, void **, lwp_t *, int, 139 size_t); 140 static void ktrwrite(struct ktr_desc *, struct ktrace_entry *); 141 static int ktrace_common(lwp_t *, int, int, int, file_t *); 142 static int ktrops(lwp_t *, struct proc *, int, int, 143 struct ktr_desc *); 144 static int ktrsetchildren(lwp_t *, struct proc *, int, int, 145 struct ktr_desc *); 146 static int ktrcanset(lwp_t *, struct proc *); 147 static int ktrsamefile(file_t *, file_t *); 148 static void ktr_kmem(lwp_t *, int, const void *, size_t); 149 static void ktr_io(lwp_t *, int, enum uio_rw, struct iovec *, size_t); 150 151 static struct ktr_desc * 152 ktd_lookup(file_t *); 153 static void ktdrel(struct ktr_desc *); 154 static void ktdref(struct ktr_desc *); 155 static void ktraddentry(lwp_t *, struct ktrace_entry *, int); 156 /* Flags for ktraddentry (3rd arg) */ 157 #define KTA_NOWAIT 0x0000 158 #define KTA_WAITOK 0x0001 159 #define KTA_LARGE 0x0002 160 static void ktefree(struct ktrace_entry *); 161 static void ktd_logerrl(struct ktr_desc *, int); 162 static void ktrace_thread(void *); 163 static int ktrderefall(struct ktr_desc *, int); 164 165 /* 166 * Default vaules. 167 */ 168 #define KTD_MAXENTRY 1000 /* XXX: tune */ 169 #define KTD_TIMEOUT 5 /* XXX: tune */ 170 #define KTD_DELAYQCNT 100 /* XXX: tune */ 171 #define KTD_WAKEDELAY 5000 /* XXX: tune */ 172 #define KTD_INTRWAKDL 100 /* XXX: tune */ 173 174 /* 175 * Patchable variables. 176 */ 177 int ktd_maxentry = KTD_MAXENTRY; /* max # of entry in the queue */ 178 int ktd_timeout = KTD_TIMEOUT; /* timeout in seconds */ 179 int ktd_delayqcnt = KTD_DELAYQCNT; /* # of entry allowed to delay */ 180 int ktd_wakedelay = KTD_WAKEDELAY; /* delay of wakeup in *ms* */ 181 int ktd_intrwakdl = KTD_INTRWAKDL; /* ditto, but when interactive */ 182 183 kmutex_t ktrace_lock; 184 int ktrace_on; 185 static TAILQ_HEAD(, ktr_desc) ktdq = TAILQ_HEAD_INITIALIZER(ktdq); 186 187 MALLOC_DEFINE(M_KTRACE, "ktrace", "ktrace data buffer"); 188 POOL_INIT(kte_pool, sizeof(struct ktrace_entry), 0, 0, 0, 189 "ktepl", &pool_allocator_nointr, IPL_NONE); 190 191 static void 192 ktd_wakeup(struct ktr_desc *ktd) 193 { 194 195 callout_stop(&ktd->ktd_wakch); 196 cv_signal(&ktd->ktd_cv); 197 } 198 199 static void 200 ktd_callout(void *arg) 201 { 202 203 mutex_enter(&ktrace_lock); 204 ktd_wakeup(arg); 205 mutex_exit(&ktrace_lock); 206 } 207 208 static void 209 ktd_logerrl(struct ktr_desc *ktd, int error) 210 { 211 212 ktd->ktd_error |= error; 213 ktd->ktd_errcnt++; 214 } 215 216 #if 0 217 static void 218 ktd_logerr(struct proc *p, int error) 219 { 220 struct ktr_desc *ktd; 221 222 KASSERT(mutex_owned(&ktrace_lock)); 223 224 ktd = p->p_tracep; 225 if (ktd == NULL) 226 return; 227 228 ktd_logerrl(ktd, error); 229 } 230 #endif 231 232 static inline int 233 ktrenter(lwp_t *l) 234 { 235 236 if ((l->l_pflag & LP_KTRACTIVE) != 0) 237 return 1; 238 l->l_pflag |= LP_KTRACTIVE; 239 return 0; 240 } 241 242 static inline void 243 ktrexit(lwp_t *l) 244 { 245 246 l->l_pflag &= ~LP_KTRACTIVE; 247 } 248 249 /* 250 * Initialise the ktrace system. 251 */ 252 void 253 ktrinit(void) 254 { 255 256 mutex_init(&ktrace_lock, MUTEX_DEFAULT, IPL_NONE); 257 } 258 259 /* 260 * Release a reference. Called with ktrace_lock held. 261 */ 262 void 263 ktdrel(struct ktr_desc *ktd) 264 { 265 266 KASSERT(mutex_owned(&ktrace_lock)); 267 268 KDASSERT(ktd->ktd_ref != 0); 269 KASSERT(ktd->ktd_ref > 0); 270 KASSERT(ktrace_on > 0); 271 ktrace_on--; 272 if (--ktd->ktd_ref <= 0) { 273 ktd->ktd_flags |= KTDF_DONE; 274 cv_signal(&ktd->ktd_cv); 275 } 276 } 277 278 void 279 ktdref(struct ktr_desc *ktd) 280 { 281 282 KASSERT(mutex_owned(&ktrace_lock)); 283 284 ktd->ktd_ref++; 285 ktrace_on++; 286 } 287 288 struct ktr_desc * 289 ktd_lookup(file_t *fp) 290 { 291 struct ktr_desc *ktd; 292 293 KASSERT(mutex_owned(&ktrace_lock)); 294 295 for (ktd = TAILQ_FIRST(&ktdq); ktd != NULL; 296 ktd = TAILQ_NEXT(ktd, ktd_list)) { 297 if (ktrsamefile(ktd->ktd_fp, fp)) { 298 ktdref(ktd); 299 break; 300 } 301 } 302 303 return (ktd); 304 } 305 306 void 307 ktraddentry(lwp_t *l, struct ktrace_entry *kte, int flags) 308 { 309 struct proc *p = l->l_proc; 310 struct ktr_desc *ktd; 311 #ifdef DEBUG 312 struct timeval t1, t2; 313 #endif 314 315 mutex_enter(&ktrace_lock); 316 317 if (p->p_traceflag & KTRFAC_TRC_EMUL) { 318 /* Add emulation trace before first entry for this process */ 319 p->p_traceflag &= ~KTRFAC_TRC_EMUL; 320 mutex_exit(&ktrace_lock); 321 ktrexit(l); 322 ktremul(); 323 (void)ktrenter(l); 324 mutex_enter(&ktrace_lock); 325 } 326 327 /* Tracing may have been cancelled. */ 328 ktd = p->p_tracep; 329 if (ktd == NULL) 330 goto freekte; 331 332 /* 333 * Bump reference count so that the object will remain while 334 * we are here. Note that the trace is controlled by other 335 * process. 336 */ 337 ktdref(ktd); 338 339 if (ktd->ktd_flags & KTDF_DONE) 340 goto relktd; 341 342 if (ktd->ktd_qcount > ktd_maxentry) { 343 ktd_logerrl(ktd, KTDE_ENOSPC); 344 goto relktd; 345 } 346 TAILQ_INSERT_TAIL(&ktd->ktd_queue, kte, kte_list); 347 ktd->ktd_qcount++; 348 if (ktd->ktd_flags & KTDF_BLOCKING) 349 goto skip_sync; 350 351 if (flags & KTA_WAITOK && 352 (/* flags & KTA_LARGE */0 || ktd->ktd_flags & KTDF_WAIT || 353 ktd->ktd_qcount > ktd_maxentry >> 1)) 354 /* 355 * Sync with writer thread since we're requesting rather 356 * big one or many requests are pending. 357 */ 358 do { 359 ktd->ktd_flags |= KTDF_WAIT; 360 ktd_wakeup(ktd); 361 #ifdef DEBUG 362 getmicrouptime(&t1); 363 #endif 364 if (cv_timedwait(&ktd->ktd_sync_cv, &ktrace_lock, 365 ktd_timeout * hz) != 0) { 366 ktd->ktd_flags |= KTDF_BLOCKING; 367 /* 368 * Maybe the writer thread is blocking 369 * completely for some reason, but 370 * don't stop target process forever. 371 */ 372 log(LOG_NOTICE, "ktrace timeout\n"); 373 break; 374 } 375 #ifdef DEBUG 376 getmicrouptime(&t2); 377 timersub(&t2, &t1, &t2); 378 if (t2.tv_sec > 0) 379 log(LOG_NOTICE, 380 "ktrace long wait: %ld.%06ld\n", 381 t2.tv_sec, t2.tv_usec); 382 #endif 383 } while (p->p_tracep == ktd && 384 (ktd->ktd_flags & (KTDF_WAIT | KTDF_DONE)) == KTDF_WAIT); 385 else { 386 /* Schedule delayed wakeup */ 387 if (ktd->ktd_qcount > ktd->ktd_delayqcnt) 388 ktd_wakeup(ktd); /* Wakeup now */ 389 else if (!callout_pending(&ktd->ktd_wakch)) 390 callout_reset(&ktd->ktd_wakch, 391 ktd->ktd_flags & KTDF_INTERACTIVE ? 392 ktd->ktd_intrwakdl : ktd->ktd_wakedelay, 393 ktd_callout, ktd); 394 } 395 396 skip_sync: 397 ktdrel(ktd); 398 mutex_exit(&ktrace_lock); 399 ktrexit(l); 400 return; 401 402 relktd: 403 ktdrel(ktd); 404 405 freekte: 406 mutex_exit(&ktrace_lock); 407 ktefree(kte); 408 ktrexit(l); 409 } 410 411 void 412 ktefree(struct ktrace_entry *kte) 413 { 414 415 if (kte->kte_buf != kte->kte_space) 416 kmem_free(kte->kte_buf, kte->kte_bufsz); 417 pool_put(&kte_pool, kte); 418 } 419 420 /* 421 * "deep" compare of two files for the purposes of clearing a trace. 422 * Returns true if they're the same open file, or if they point at the 423 * same underlying vnode/socket. 424 */ 425 426 int 427 ktrsamefile(file_t *f1, file_t *f2) 428 { 429 430 return ((f1 == f2) || 431 ((f1 != NULL) && (f2 != NULL) && 432 (f1->f_type == f2->f_type) && 433 (f1->f_data == f2->f_data))); 434 } 435 436 void 437 ktrderef(struct proc *p) 438 { 439 struct ktr_desc *ktd = p->p_tracep; 440 441 KASSERT(mutex_owned(&ktrace_lock)); 442 443 p->p_traceflag = 0; 444 if (ktd == NULL) 445 return; 446 p->p_tracep = NULL; 447 448 cv_broadcast(&ktd->ktd_sync_cv); 449 ktdrel(ktd); 450 } 451 452 void 453 ktradref(struct proc *p) 454 { 455 struct ktr_desc *ktd = p->p_tracep; 456 457 KASSERT(mutex_owned(&ktrace_lock)); 458 459 ktdref(ktd); 460 } 461 462 int 463 ktrderefall(struct ktr_desc *ktd, int auth) 464 { 465 lwp_t *curl = curlwp; 466 struct proc *p; 467 int error = 0; 468 469 mutex_enter(proc_lock); 470 PROCLIST_FOREACH(p, &allproc) { 471 if (p->p_tracep != ktd) 472 continue; 473 mutex_enter(p->p_lock); 474 mutex_enter(&ktrace_lock); 475 if (p->p_tracep == ktd) { 476 if (!auth || ktrcanset(curl, p)) 477 ktrderef(p); 478 else 479 error = EPERM; 480 } 481 mutex_exit(&ktrace_lock); 482 mutex_exit(p->p_lock); 483 } 484 mutex_exit(proc_lock); 485 486 return error; 487 } 488 489 int 490 ktealloc(struct ktrace_entry **ktep, void **bufp, lwp_t *l, int type, 491 size_t sz) 492 { 493 struct proc *p = l->l_proc; 494 struct ktrace_entry *kte; 495 struct ktr_header *kth; 496 void *buf; 497 498 if (ktrenter(l)) 499 return EAGAIN; 500 501 kte = pool_get(&kte_pool, PR_WAITOK); 502 if (sz > sizeof(kte->kte_space)) { 503 if ((buf = kmem_alloc(sz, KM_SLEEP)) == NULL) { 504 pool_put(&kte_pool, kte); 505 ktrexit(l); 506 return ENOMEM; 507 } 508 } else 509 buf = kte->kte_space; 510 511 kte->kte_bufsz = sz; 512 kte->kte_buf = buf; 513 514 kth = &kte->kte_kth; 515 (void)memset(kth, 0, sizeof(*kth)); 516 kth->ktr_len = sz; 517 kth->ktr_type = type; 518 kth->ktr_pid = p->p_pid; 519 memcpy(kth->ktr_comm, p->p_comm, MAXCOMLEN); 520 kth->ktr_version = KTRFAC_VERSION(p->p_traceflag); 521 522 switch (KTRFAC_VERSION(p->p_traceflag)) { 523 case 0: 524 /* This is the original format */ 525 microtime(&kth->ktr_tv); 526 break; 527 case 1: 528 kth->ktr_lid = l->l_lid; 529 nanotime(&kth->ktr_time); 530 break; 531 default: 532 break; 533 } 534 535 *ktep = kte; 536 *bufp = buf; 537 538 return 0; 539 } 540 541 void 542 ktr_syscall(register_t code, const register_t args[], int narg) 543 { 544 lwp_t *l = curlwp; 545 struct proc *p = l->l_proc; 546 struct ktrace_entry *kte; 547 struct ktr_syscall *ktp; 548 register_t *argp; 549 size_t len; 550 u_int i; 551 552 if (!KTRPOINT(p, KTR_SYSCALL)) 553 return; 554 555 len = sizeof(struct ktr_syscall) + narg * sizeof argp[0]; 556 557 if (ktealloc(&kte, (void *)&ktp, l, KTR_SYSCALL, len)) 558 return; 559 560 ktp->ktr_code = code; 561 ktp->ktr_argsize = narg * sizeof argp[0]; 562 argp = (register_t *)(ktp + 1); 563 for (i = 0; i < narg; i++) 564 *argp++ = args[i]; 565 566 ktraddentry(l, kte, KTA_WAITOK); 567 } 568 569 void 570 ktr_sysret(register_t code, int error, register_t *retval) 571 { 572 lwp_t *l = curlwp; 573 struct ktrace_entry *kte; 574 struct ktr_sysret *ktp; 575 576 if (!KTRPOINT(l->l_proc, KTR_SYSRET)) 577 return; 578 579 if (ktealloc(&kte, (void *)&ktp, l, KTR_SYSRET, 580 sizeof(struct ktr_sysret))) 581 return; 582 583 ktp->ktr_code = code; 584 ktp->ktr_eosys = 0; /* XXX unused */ 585 ktp->ktr_error = error; 586 ktp->ktr_retval = retval ? retval[0] : 0; 587 ktp->ktr_retval_1 = retval ? retval[1] : 0; 588 589 ktraddentry(l, kte, KTA_WAITOK); 590 } 591 592 void 593 ktr_namei(const char *path, size_t pathlen) 594 { 595 lwp_t *l = curlwp; 596 597 if (!KTRPOINT(l->l_proc, KTR_NAMEI)) 598 return; 599 600 ktr_kmem(l, KTR_NAMEI, path, pathlen); 601 } 602 603 void 604 ktr_namei2(const char *eroot, size_t erootlen, 605 const char *path, size_t pathlen) 606 { 607 lwp_t *l = curlwp; 608 struct ktrace_entry *kte; 609 void *buf; 610 611 if (!KTRPOINT(l->l_proc, KTR_NAMEI)) 612 return; 613 614 if (ktealloc(&kte, &buf, l, KTR_NAMEI, erootlen + pathlen)) 615 return; 616 memcpy(buf, eroot, erootlen); 617 buf = (char *)buf + erootlen; 618 memcpy(buf, path, pathlen); 619 ktraddentry(l, kte, KTA_WAITOK); 620 } 621 622 void 623 ktr_emul(void) 624 { 625 lwp_t *l = curlwp; 626 const char *emul = l->l_proc->p_emul->e_name; 627 628 if (!KTRPOINT(l->l_proc, KTR_EMUL)) 629 return; 630 631 ktr_kmem(l, KTR_EMUL, emul, strlen(emul)); 632 } 633 634 void 635 ktr_execarg(const void *bf, size_t len) 636 { 637 lwp_t *l = curlwp; 638 639 if (!KTRPOINT(l->l_proc, KTR_EXEC_ARG)) 640 return; 641 642 ktr_kmem(l, KTR_EXEC_ARG, bf, len); 643 } 644 645 void 646 ktr_execenv(const void *bf, size_t len) 647 { 648 lwp_t *l = curlwp; 649 650 if (!KTRPOINT(l->l_proc, KTR_EXEC_ENV)) 651 return; 652 653 ktr_kmem(l, KTR_EXEC_ENV, bf, len); 654 } 655 656 static void 657 ktr_kmem(lwp_t *l, int type, const void *bf, size_t len) 658 { 659 struct ktrace_entry *kte; 660 void *buf; 661 662 if (ktealloc(&kte, &buf, l, type, len)) 663 return; 664 memcpy(buf, bf, len); 665 ktraddentry(l, kte, KTA_WAITOK); 666 } 667 668 static void 669 ktr_io(lwp_t *l, int fd, enum uio_rw rw, struct iovec *iov, size_t len) 670 { 671 struct ktrace_entry *kte; 672 struct ktr_genio *ktp; 673 size_t resid = len, cnt, buflen; 674 void *cp; 675 676 next: 677 buflen = min(PAGE_SIZE, resid + sizeof(struct ktr_genio)); 678 679 if (ktealloc(&kte, (void *)&ktp, l, KTR_GENIO, buflen)) 680 return; 681 682 ktp->ktr_fd = fd; 683 ktp->ktr_rw = rw; 684 685 cp = (void *)(ktp + 1); 686 buflen -= sizeof(struct ktr_genio); 687 kte->kte_kth.ktr_len = sizeof(struct ktr_genio); 688 689 while (buflen > 0) { 690 cnt = min(iov->iov_len, buflen); 691 if (copyin(iov->iov_base, cp, cnt) != 0) 692 goto out; 693 kte->kte_kth.ktr_len += cnt; 694 buflen -= cnt; 695 resid -= cnt; 696 iov->iov_len -= cnt; 697 if (iov->iov_len == 0) 698 iov++; 699 else 700 iov->iov_base = (char *)iov->iov_base + cnt; 701 } 702 703 /* 704 * Don't push so many entry at once. It will cause kmem map 705 * shortage. 706 */ 707 ktraddentry(l, kte, KTA_WAITOK | KTA_LARGE); 708 if (resid > 0) { 709 if (curcpu()->ci_schedstate.spc_flags & SPCF_SHOULDYIELD) { 710 (void)ktrenter(l); 711 preempt(); 712 ktrexit(l); 713 } 714 715 goto next; 716 } 717 718 return; 719 720 out: 721 ktefree(kte); 722 ktrexit(l); 723 } 724 725 void 726 ktr_genio(int fd, enum uio_rw rw, const void *addr, size_t len, int error) 727 { 728 lwp_t *l = curlwp; 729 struct iovec iov; 730 731 if (!KTRPOINT(l->l_proc, KTR_GENIO) || error != 0) 732 return; 733 iov.iov_base = __UNCONST(addr); 734 iov.iov_len = len; 735 ktr_io(l, fd, rw, &iov, len); 736 } 737 738 void 739 ktr_geniov(int fd, enum uio_rw rw, struct iovec *iov, size_t len, int error) 740 { 741 lwp_t *l = curlwp; 742 743 if (!KTRPOINT(l->l_proc, KTR_GENIO) || error != 0) 744 return; 745 ktr_io(l, fd, rw, iov, len); 746 } 747 748 void 749 ktr_mibio(int fd, enum uio_rw rw, const void *addr, size_t len, int error) 750 { 751 lwp_t *l = curlwp; 752 struct iovec iov; 753 754 if (!KTRPOINT(l->l_proc, KTR_MIB) || error != 0) 755 return; 756 iov.iov_base = __UNCONST(addr); 757 iov.iov_len = len; 758 ktr_io(l, fd, rw, &iov, len); 759 } 760 761 void 762 ktr_psig(int sig, sig_t action, const sigset_t *mask, 763 const ksiginfo_t *ksi) 764 { 765 struct ktrace_entry *kte; 766 lwp_t *l = curlwp; 767 struct { 768 struct ktr_psig kp; 769 siginfo_t si; 770 } *kbuf; 771 772 if (!KTRPOINT(l->l_proc, KTR_PSIG)) 773 return; 774 775 if (ktealloc(&kte, (void *)&kbuf, l, KTR_PSIG, sizeof(*kbuf))) 776 return; 777 778 kbuf->kp.signo = (char)sig; 779 kbuf->kp.action = action; 780 kbuf->kp.mask = *mask; 781 782 if (ksi) { 783 kbuf->kp.code = KSI_TRAPCODE(ksi); 784 (void)memset(&kbuf->si, 0, sizeof(kbuf->si)); 785 kbuf->si._info = ksi->ksi_info; 786 kte->kte_kth.ktr_len = sizeof(*kbuf); 787 } else { 788 kbuf->kp.code = 0; 789 kte->kte_kth.ktr_len = sizeof(struct ktr_psig); 790 } 791 792 ktraddentry(l, kte, KTA_WAITOK); 793 } 794 795 void 796 ktr_csw(int out, int user) 797 { 798 lwp_t *l = curlwp; 799 struct proc *p = l->l_proc; 800 struct ktrace_entry *kte; 801 struct ktr_csw *kc; 802 803 if (!KTRPOINT(p, KTR_CSW)) 804 return; 805 806 /* 807 * Don't record context switches resulting from blocking on 808 * locks; it's too easy to get duff results. 809 */ 810 if (l->l_syncobj == &mutex_syncobj || l->l_syncobj == &rw_syncobj) 811 return; 812 813 /* 814 * We can't sleep if we're already going to sleep (if original 815 * condition is met during sleep, we hang up). 816 * 817 * XXX This is not ideal: it would be better to maintain a pool 818 * of ktes and actually push this to the kthread when context 819 * switch happens, however given the points where we are called 820 * from that is difficult to do. 821 */ 822 if (out) { 823 if (ktrenter(l)) 824 return; 825 826 switch (KTRFAC_VERSION(p->p_traceflag)) { 827 case 0: 828 /* This is the original format */ 829 microtime(&l->l_ktrcsw.tv); 830 l->l_pflag |= LP_KTRCSW; 831 break; 832 case 1: 833 nanotime(&l->l_ktrcsw.ts); 834 l->l_pflag |= LP_KTRCSW; 835 break; 836 default: 837 break; 838 } 839 840 if (user) 841 l->l_pflag |= LP_KTRCSWUSER; 842 else 843 l->l_pflag &= ~LP_KTRCSWUSER; 844 845 ktrexit(l); 846 return; 847 } 848 849 /* 850 * On the way back in, we need to record twice: once for entry, and 851 * once for exit. 852 */ 853 if ((l->l_pflag & LP_KTRCSW) != 0) { 854 l->l_pflag &= ~LP_KTRCSW; 855 856 if (ktealloc(&kte, (void *)&kc, l, KTR_CSW, sizeof(*kc))) 857 return; 858 859 kc->out = 1; 860 kc->user = ((l->l_pflag & LP_KTRCSWUSER) != 0); 861 862 switch (KTRFAC_VERSION(p->p_traceflag)) { 863 case 0: 864 /* This is the original format */ 865 memcpy(&kte->kte_kth.ktr_tv, &l->l_ktrcsw.tv, 866 sizeof(kte->kte_kth.ktr_tv)); 867 break; 868 case 1: 869 memcpy(&kte->kte_kth.ktr_time, &l->l_ktrcsw.ts, 870 sizeof(kte->kte_kth.ktr_time)); 871 break; 872 default: 873 break; 874 } 875 876 ktraddentry(l, kte, KTA_WAITOK); 877 } 878 879 if (ktealloc(&kte, (void *)&kc, l, KTR_CSW, sizeof(*kc))) 880 return; 881 882 kc->out = 0; 883 kc->user = user; 884 885 ktraddentry(l, kte, KTA_WAITOK); 886 } 887 888 bool 889 ktr_point(int fac_bit) 890 { 891 return curlwp->l_proc->p_traceflag & fac_bit; 892 } 893 894 int 895 ktruser(const char *id, void *addr, size_t len, int ustr) 896 { 897 struct ktrace_entry *kte; 898 struct ktr_user *ktp; 899 lwp_t *l = curlwp; 900 void *user_dta; 901 int error; 902 903 if (!KTRPOINT(l->l_proc, KTR_USER)) 904 return 0; 905 906 if (len > KTR_USER_MAXLEN) 907 return ENOSPC; 908 909 error = ktealloc(&kte, (void *)&ktp, l, KTR_USER, sizeof(*ktp) + len); 910 if (error != 0) 911 return error; 912 913 if (ustr) { 914 if (copyinstr(id, ktp->ktr_id, KTR_USER_MAXIDLEN, NULL) != 0) 915 ktp->ktr_id[0] = '\0'; 916 } else 917 strncpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN); 918 ktp->ktr_id[KTR_USER_MAXIDLEN-1] = '\0'; 919 920 user_dta = (void *)(ktp + 1); 921 if ((error = copyin(addr, (void *)user_dta, len)) != 0) 922 len = 0; 923 924 ktraddentry(l, kte, KTA_WAITOK); 925 return error; 926 } 927 928 void 929 ktr_kuser(const char *id, void *addr, size_t len) 930 { 931 struct ktrace_entry *kte; 932 struct ktr_user *ktp; 933 lwp_t *l = curlwp; 934 int error; 935 936 if (!KTRPOINT(l->l_proc, KTR_USER)) 937 return; 938 939 if (len > KTR_USER_MAXLEN) 940 return; 941 942 error = ktealloc(&kte, (void *)&ktp, l, KTR_USER, sizeof(*ktp) + len); 943 if (error != 0) 944 return; 945 946 strlcpy(ktp->ktr_id, id, KTR_USER_MAXIDLEN); 947 948 memcpy(ktp + 1, addr, len); 949 950 ktraddentry(l, kte, KTA_WAITOK); 951 } 952 953 void 954 ktr_mmsg(const void *msgh, size_t size) 955 { 956 lwp_t *l = curlwp; 957 958 if (!KTRPOINT(l->l_proc, KTR_MMSG)) 959 return; 960 961 ktr_kmem(l, KTR_MMSG, msgh, size); 962 } 963 964 void 965 ktr_mool(const void *kaddr, size_t size, const void *uaddr) 966 { 967 struct ktrace_entry *kte; 968 struct ktr_mool *kp; 969 struct ktr_mool *bf; 970 lwp_t *l = curlwp; 971 972 if (!KTRPOINT(l->l_proc, KTR_MOOL)) 973 return; 974 975 if (ktealloc(&kte, (void *)&kp, l, KTR_MOOL, size + sizeof(*kp))) 976 return; 977 978 kp->uaddr = uaddr; 979 kp->size = size; 980 bf = kp + 1; /* Skip uaddr and size */ 981 (void)memcpy(bf, kaddr, size); 982 983 ktraddentry(l, kte, KTA_WAITOK); 984 } 985 986 void 987 ktr_mib(const int *name, u_int namelen) 988 { 989 struct ktrace_entry *kte; 990 int *namep; 991 size_t size; 992 lwp_t *l = curlwp; 993 994 if (!KTRPOINT(l->l_proc, KTR_MIB)) 995 return; 996 997 size = namelen * sizeof(*name); 998 999 if (ktealloc(&kte, (void *)&namep, l, KTR_MIB, size)) 1000 return; 1001 1002 (void)memcpy(namep, name, namelen * sizeof(*name)); 1003 1004 ktraddentry(l, kte, KTA_WAITOK); 1005 } 1006 1007 /* Interface and common routines */ 1008 1009 int 1010 ktrace_common(lwp_t *curl, int ops, int facs, int pid, file_t *fp) 1011 { 1012 struct proc *curp; 1013 struct proc *p; 1014 struct pgrp *pg; 1015 struct ktr_desc *ktd = NULL; 1016 int ret = 0; 1017 int error = 0; 1018 int descend; 1019 1020 curp = curl->l_proc; 1021 descend = ops & KTRFLAG_DESCEND; 1022 facs = facs & ~((unsigned) KTRFAC_PERSISTENT); 1023 1024 (void)ktrenter(curl); 1025 1026 switch (KTROP(ops)) { 1027 1028 case KTROP_CLEARFILE: 1029 /* 1030 * Clear all uses of the tracefile 1031 */ 1032 mutex_enter(&ktrace_lock); 1033 ktd = ktd_lookup(fp); 1034 mutex_exit(&ktrace_lock); 1035 if (ktd == NULL) 1036 goto done; 1037 error = ktrderefall(ktd, 1); 1038 goto done; 1039 1040 case KTROP_SET: 1041 mutex_enter(&ktrace_lock); 1042 ktd = ktd_lookup(fp); 1043 mutex_exit(&ktrace_lock); 1044 if (ktd == NULL) { 1045 ktd = kmem_alloc(sizeof(*ktd), KM_SLEEP); 1046 TAILQ_INIT(&ktd->ktd_queue); 1047 callout_init(&ktd->ktd_wakch, CALLOUT_MPSAFE); 1048 cv_init(&ktd->ktd_cv, "ktrwait"); 1049 cv_init(&ktd->ktd_sync_cv, "ktrsync"); 1050 ktd->ktd_flags = 0; 1051 ktd->ktd_qcount = 0; 1052 ktd->ktd_error = 0; 1053 ktd->ktd_errcnt = 0; 1054 ktd->ktd_delayqcnt = ktd_delayqcnt; 1055 ktd->ktd_wakedelay = mstohz(ktd_wakedelay); 1056 ktd->ktd_intrwakdl = mstohz(ktd_intrwakdl); 1057 ktd->ktd_ref = 0; 1058 ktd->ktd_fp = fp; 1059 mutex_enter(&ktrace_lock); 1060 ktdref(ktd); 1061 mutex_exit(&ktrace_lock); 1062 1063 /* 1064 * XXX: not correct. needs an way to detect 1065 * whether ktruss or ktrace. 1066 */ 1067 if (fp->f_type == DTYPE_PIPE) 1068 ktd->ktd_flags |= KTDF_INTERACTIVE; 1069 1070 mutex_enter(&fp->f_lock); 1071 fp->f_count++; 1072 mutex_exit(&fp->f_lock); 1073 error = kthread_create(PRI_NONE, KTHREAD_MPSAFE, NULL, 1074 ktrace_thread, ktd, &ktd->ktd_lwp, "ktrace"); 1075 if (error != 0) { 1076 kmem_free(ktd, sizeof(*ktd)); 1077 mutex_enter(&fp->f_lock); 1078 fp->f_count--; 1079 mutex_exit(&fp->f_lock); 1080 goto done; 1081 } 1082 1083 mutex_enter(&ktrace_lock); 1084 if (ktd_lookup(fp) != NULL) { 1085 ktdrel(ktd); 1086 ktd = NULL; 1087 } else 1088 TAILQ_INSERT_TAIL(&ktdq, ktd, ktd_list); 1089 if (ktd == NULL) 1090 cv_wait(&lbolt, &ktrace_lock); 1091 mutex_exit(&ktrace_lock); 1092 if (ktd == NULL) 1093 goto done; 1094 } 1095 break; 1096 1097 case KTROP_CLEAR: 1098 break; 1099 } 1100 1101 /* 1102 * need something to (un)trace (XXX - why is this here?) 1103 */ 1104 if (!facs) { 1105 error = EINVAL; 1106 goto done; 1107 } 1108 1109 /* 1110 * do it 1111 */ 1112 mutex_enter(proc_lock); 1113 if (pid < 0) { 1114 /* 1115 * by process group 1116 */ 1117 pg = pg_find(-pid, PFIND_LOCKED); 1118 if (pg == NULL) 1119 error = ESRCH; 1120 else { 1121 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 1122 if (descend) 1123 ret |= ktrsetchildren(curl, p, ops, 1124 facs, ktd); 1125 else 1126 ret |= ktrops(curl, p, ops, facs, 1127 ktd); 1128 } 1129 } 1130 1131 } else { 1132 /* 1133 * by pid 1134 */ 1135 p = p_find(pid, PFIND_LOCKED); 1136 if (p == NULL) 1137 error = ESRCH; 1138 else if (descend) 1139 ret |= ktrsetchildren(curl, p, ops, facs, ktd); 1140 else 1141 ret |= ktrops(curl, p, ops, facs, ktd); 1142 } 1143 mutex_exit(proc_lock); 1144 if (error == 0 && !ret) 1145 error = EPERM; 1146 done: 1147 if (ktd != NULL) { 1148 mutex_enter(&ktrace_lock); 1149 if (error != 0) { 1150 /* 1151 * Wakeup the thread so that it can be die if we 1152 * can't trace any process. 1153 */ 1154 ktd_wakeup(ktd); 1155 } 1156 if (KTROP(ops) == KTROP_SET || KTROP(ops) == KTROP_CLEARFILE) 1157 ktdrel(ktd); 1158 mutex_exit(&ktrace_lock); 1159 } 1160 ktrexit(curl); 1161 return (error); 1162 } 1163 1164 /* 1165 * fktrace system call 1166 */ 1167 /* ARGSUSED */ 1168 int 1169 sys_fktrace(struct lwp *l, const struct sys_fktrace_args *uap, register_t *retval) 1170 { 1171 /* { 1172 syscallarg(int) fd; 1173 syscallarg(int) ops; 1174 syscallarg(int) facs; 1175 syscallarg(int) pid; 1176 } */ 1177 file_t *fp; 1178 int error, fd; 1179 1180 fd = SCARG(uap, fd); 1181 if ((fp = fd_getfile(fd)) == NULL) 1182 return (EBADF); 1183 if ((fp->f_flag & FWRITE) == 0) 1184 error = EBADF; 1185 else 1186 error = ktrace_common(l, SCARG(uap, ops), 1187 SCARG(uap, facs), SCARG(uap, pid), fp); 1188 fd_putfile(fd); 1189 return error; 1190 } 1191 1192 /* 1193 * ktrace system call 1194 */ 1195 /* ARGSUSED */ 1196 int 1197 sys_ktrace(struct lwp *l, const struct sys_ktrace_args *uap, register_t *retval) 1198 { 1199 /* { 1200 syscallarg(const char *) fname; 1201 syscallarg(int) ops; 1202 syscallarg(int) facs; 1203 syscallarg(int) pid; 1204 } */ 1205 struct vnode *vp = NULL; 1206 file_t *fp = NULL; 1207 struct nameidata nd; 1208 int error = 0; 1209 int fd; 1210 1211 if (ktrenter(l)) 1212 return EAGAIN; 1213 1214 if (KTROP(SCARG(uap, ops)) != KTROP_CLEAR) { 1215 /* 1216 * an operation which requires a file argument. 1217 */ 1218 NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, fname)); 1219 if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) { 1220 ktrexit(l); 1221 return (error); 1222 } 1223 vp = nd.ni_vp; 1224 VOP_UNLOCK(vp, 0); 1225 if (vp->v_type != VREG) { 1226 vn_close(vp, FREAD|FWRITE, l->l_cred); 1227 ktrexit(l); 1228 return (EACCES); 1229 } 1230 /* 1231 * This uses up a file descriptor slot in the 1232 * tracing process for the duration of this syscall. 1233 * This is not expected to be a problem. 1234 */ 1235 if ((error = fd_allocfile(&fp, &fd)) != 0) { 1236 vn_close(vp, FWRITE, l->l_cred); 1237 ktrexit(l); 1238 return error; 1239 } 1240 fp->f_flag = FWRITE; 1241 fp->f_type = DTYPE_VNODE; 1242 fp->f_ops = &vnops; 1243 fp->f_data = (void *)vp; 1244 vp = NULL; 1245 } 1246 error = ktrace_common(l, SCARG(uap, ops), SCARG(uap, facs), 1247 SCARG(uap, pid), fp); 1248 if (fp != NULL) { 1249 if (error != 0) { 1250 /* File unused. */ 1251 fd_abort(curproc, fp, fd); 1252 } else { 1253 /* File was used. */ 1254 fd_abort(curproc, NULL, fd); 1255 } 1256 } 1257 return (error); 1258 } 1259 1260 int 1261 ktrops(lwp_t *curl, struct proc *p, int ops, int facs, 1262 struct ktr_desc *ktd) 1263 { 1264 int vers = ops & KTRFAC_VER_MASK; 1265 int error = 0; 1266 1267 mutex_enter(p->p_lock); 1268 mutex_enter(&ktrace_lock); 1269 1270 if (!ktrcanset(curl, p)) 1271 goto out; 1272 1273 switch (vers) { 1274 case KTRFACv0: 1275 case KTRFACv1: 1276 break; 1277 default: 1278 error = EINVAL; 1279 goto out; 1280 } 1281 1282 if (KTROP(ops) == KTROP_SET) { 1283 if (p->p_tracep != ktd) { 1284 /* 1285 * if trace file already in use, relinquish 1286 */ 1287 ktrderef(p); 1288 p->p_tracep = ktd; 1289 ktradref(p); 1290 } 1291 p->p_traceflag |= facs; 1292 if (kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_KTRACE, 1293 p, KAUTH_ARG(KAUTH_REQ_PROCESS_KTRACE_PERSISTENT), NULL, 1294 NULL) == 0) 1295 p->p_traceflag |= KTRFAC_PERSISTENT; 1296 } else { 1297 /* KTROP_CLEAR */ 1298 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) { 1299 /* no more tracing */ 1300 ktrderef(p); 1301 } 1302 } 1303 1304 if (p->p_traceflag) 1305 p->p_traceflag |= vers; 1306 /* 1307 * Emit an emulation record, every time there is a ktrace 1308 * change/attach request. 1309 */ 1310 if (KTRPOINT(p, KTR_EMUL)) 1311 p->p_traceflag |= KTRFAC_TRC_EMUL; 1312 1313 p->p_trace_enabled = trace_is_enabled(p); 1314 #ifdef __HAVE_SYSCALL_INTERN 1315 (*p->p_emul->e_syscall_intern)(p); 1316 #endif 1317 1318 out: 1319 mutex_exit(&ktrace_lock); 1320 mutex_exit(p->p_lock); 1321 1322 return (1); 1323 } 1324 1325 int 1326 ktrsetchildren(lwp_t *curl, struct proc *top, int ops, int facs, 1327 struct ktr_desc *ktd) 1328 { 1329 struct proc *p; 1330 int ret = 0; 1331 1332 KASSERT(mutex_owned(proc_lock)); 1333 1334 p = top; 1335 for (;;) { 1336 ret |= ktrops(curl, p, ops, facs, ktd); 1337 /* 1338 * If this process has children, descend to them next, 1339 * otherwise do any siblings, and if done with this level, 1340 * follow back up the tree (but not past top). 1341 */ 1342 if (LIST_FIRST(&p->p_children) != NULL) { 1343 p = LIST_FIRST(&p->p_children); 1344 continue; 1345 } 1346 for (;;) { 1347 if (p == top) 1348 return (ret); 1349 if (LIST_NEXT(p, p_sibling) != NULL) { 1350 p = LIST_NEXT(p, p_sibling); 1351 break; 1352 } 1353 p = p->p_pptr; 1354 } 1355 } 1356 /*NOTREACHED*/ 1357 } 1358 1359 void 1360 ktrwrite(struct ktr_desc *ktd, struct ktrace_entry *kte) 1361 { 1362 struct uio auio; 1363 struct iovec aiov[64], *iov; 1364 struct ktrace_entry *top = kte; 1365 struct ktr_header *kth; 1366 file_t *fp = ktd->ktd_fp; 1367 int error; 1368 next: 1369 auio.uio_iov = iov = &aiov[0]; 1370 auio.uio_offset = 0; 1371 auio.uio_rw = UIO_WRITE; 1372 auio.uio_resid = 0; 1373 auio.uio_iovcnt = 0; 1374 UIO_SETUP_SYSSPACE(&auio); 1375 do { 1376 kth = &kte->kte_kth; 1377 1378 if (kth->ktr_version == 0) { 1379 /* 1380 * Convert back to the old format fields 1381 */ 1382 TIMESPEC_TO_TIMEVAL(&kth->ktr_tv, &kth->ktr_time); 1383 kth->ktr_unused = NULL; 1384 } 1385 iov->iov_base = (void *)kth; 1386 iov++->iov_len = sizeof(struct ktr_header); 1387 auio.uio_resid += sizeof(struct ktr_header); 1388 auio.uio_iovcnt++; 1389 if (kth->ktr_len > 0) { 1390 iov->iov_base = kte->kte_buf; 1391 iov++->iov_len = kth->ktr_len; 1392 auio.uio_resid += kth->ktr_len; 1393 auio.uio_iovcnt++; 1394 } 1395 } while ((kte = TAILQ_NEXT(kte, kte_list)) != NULL && 1396 auio.uio_iovcnt < sizeof(aiov) / sizeof(aiov[0]) - 1); 1397 1398 again: 1399 error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio, 1400 fp->f_cred, FOF_UPDATE_OFFSET); 1401 switch (error) { 1402 1403 case 0: 1404 if (auio.uio_resid > 0) 1405 goto again; 1406 if (kte != NULL) 1407 goto next; 1408 break; 1409 1410 case EWOULDBLOCK: 1411 kpause("ktrzzz", false, 1, NULL); 1412 goto again; 1413 1414 default: 1415 /* 1416 * If error encountered, give up tracing on this 1417 * vnode. Don't report EPIPE as this can easily 1418 * happen with fktrace()/ktruss. 1419 */ 1420 #ifndef DEBUG 1421 if (error != EPIPE) 1422 #endif 1423 log(LOG_NOTICE, 1424 "ktrace write failed, errno %d, tracing stopped\n", 1425 error); 1426 (void)ktrderefall(ktd, 0); 1427 } 1428 1429 while ((kte = top) != NULL) { 1430 top = TAILQ_NEXT(top, kte_list); 1431 ktefree(kte); 1432 } 1433 } 1434 1435 void 1436 ktrace_thread(void *arg) 1437 { 1438 struct ktr_desc *ktd = arg; 1439 file_t *fp = ktd->ktd_fp; 1440 struct ktrace_entry *kte; 1441 int ktrerr, errcnt; 1442 1443 mutex_enter(&ktrace_lock); 1444 for (;;) { 1445 kte = TAILQ_FIRST(&ktd->ktd_queue); 1446 if (kte == NULL) { 1447 if (ktd->ktd_flags & KTDF_WAIT) { 1448 ktd->ktd_flags &= ~(KTDF_WAIT | KTDF_BLOCKING); 1449 cv_broadcast(&ktd->ktd_sync_cv); 1450 } 1451 if (ktd->ktd_ref == 0) 1452 break; 1453 cv_wait(&ktd->ktd_cv, &ktrace_lock); 1454 continue; 1455 } 1456 TAILQ_INIT(&ktd->ktd_queue); 1457 ktd->ktd_qcount = 0; 1458 ktrerr = ktd->ktd_error; 1459 errcnt = ktd->ktd_errcnt; 1460 ktd->ktd_error = ktd->ktd_errcnt = 0; 1461 mutex_exit(&ktrace_lock); 1462 1463 if (ktrerr) { 1464 log(LOG_NOTICE, 1465 "ktrace failed, fp %p, error 0x%x, total %d\n", 1466 fp, ktrerr, errcnt); 1467 } 1468 ktrwrite(ktd, kte); 1469 mutex_enter(&ktrace_lock); 1470 } 1471 1472 TAILQ_REMOVE(&ktdq, ktd, ktd_list); 1473 mutex_exit(&ktrace_lock); 1474 1475 /* 1476 * ktrace file descriptor can't be watched (are not visible to 1477 * userspace), so no kqueue stuff here 1478 * XXX: The above comment is wrong, because the fktrace file 1479 * descriptor is available in userland. 1480 */ 1481 closef(fp); 1482 1483 callout_stop(&ktd->ktd_wakch); 1484 callout_destroy(&ktd->ktd_wakch); 1485 kmem_free(ktd, sizeof(*ktd)); 1486 1487 kthread_exit(0); 1488 } 1489 1490 /* 1491 * Return true if caller has permission to set the ktracing state 1492 * of target. Essentially, the target can't possess any 1493 * more permissions than the caller. KTRFAC_PERSISTENT signifies that 1494 * the tracing will persist on sugid processes during exec; it is only 1495 * settable by a process with appropriate credentials. 1496 * 1497 * TODO: check groups. use caller effective gid. 1498 */ 1499 int 1500 ktrcanset(lwp_t *calll, struct proc *targetp) 1501 { 1502 KASSERT(mutex_owned(targetp->p_lock)); 1503 KASSERT(mutex_owned(&ktrace_lock)); 1504 1505 if (kauth_authorize_process(calll->l_cred, KAUTH_PROCESS_KTRACE, 1506 targetp, NULL, NULL, NULL) == 0) 1507 return (1); 1508 1509 return (0); 1510 } 1511 1512 /* 1513 * Put user defined entry to ktrace records. 1514 */ 1515 int 1516 sys_utrace(struct lwp *l, const struct sys_utrace_args *uap, register_t *retval) 1517 { 1518 /* { 1519 syscallarg(const char *) label; 1520 syscallarg(void *) addr; 1521 syscallarg(size_t) len; 1522 } */ 1523 1524 return ktruser(SCARG(uap, label), SCARG(uap, addr), 1525 SCARG(uap, len), 1); 1526 } 1527