kern_ktrace.c revision 112199
1/* 2 * Copyright (c) 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)kern_ktrace.c 8.2 (Berkeley) 9/23/93 34 * $FreeBSD: head/sys/kern/kern_ktrace.c 112199 2003-03-13 18:31:15Z jhb $ 35 */ 36 37#include "opt_ktrace.h" 38#include "opt_mac.h" 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/fcntl.h> 43#include <sys/jail.h> 44#include <sys/kernel.h> 45#include <sys/kthread.h> 46#include <sys/lock.h> 47#include <sys/mutex.h> 48#include <sys/mac.h> 49#include <sys/malloc.h> 50#include <sys/namei.h> 51#include <sys/proc.h> 52#include <sys/unistd.h> 53#include <sys/vnode.h> 54#include <sys/ktrace.h> 55#include <sys/sema.h> 56#include <sys/sx.h> 57#include <sys/sysctl.h> 58#include <sys/syslog.h> 59#include <sys/sysproto.h> 60 61static MALLOC_DEFINE(M_KTRACE, "KTRACE", "KTRACE"); 62 63#ifdef KTRACE 64 65#ifndef KTRACE_REQUEST_POOL 66#define KTRACE_REQUEST_POOL 100 67#endif 68 69struct ktr_request { 70 struct ktr_header ktr_header; 71 struct ucred *ktr_cred; 72 struct vnode *ktr_vp; 73 union { 74 struct ktr_syscall ktr_syscall; 75 struct ktr_sysret ktr_sysret; 76 struct ktr_genio ktr_genio; 77 struct ktr_psig ktr_psig; 78 struct ktr_csw ktr_csw; 79 } ktr_data; 80 STAILQ_ENTRY(ktr_request) ktr_list; 81}; 82 83static int data_lengths[] = { 84 0, /* none */ 85 offsetof(struct ktr_syscall, ktr_args), /* KTR_SYSCALL */ 86 sizeof(struct ktr_sysret), /* KTR_SYSRET */ 87 0, /* KTR_NAMEI */ 88 sizeof(struct ktr_genio), /* KTR_GENIO */ 89 sizeof(struct ktr_psig), /* KTR_PSIG */ 90 sizeof(struct ktr_csw), /* KTR_CSW */ 91 0 /* KTR_USER */ 92}; 93 94static STAILQ_HEAD(, ktr_request) ktr_todo; 95static STAILQ_HEAD(, ktr_request) ktr_free; 96 97SYSCTL_NODE(_kern, OID_AUTO, ktrace, CTLFLAG_RD, 0, "KTRACE options"); 98 99static uint ktr_requestpool = KTRACE_REQUEST_POOL; 100TUNABLE_INT("kern.ktrace.request_pool", &ktr_requestpool); 101 102static uint ktr_geniosize = PAGE_SIZE; 103TUNABLE_INT("kern.ktrace.genio_size", &ktr_geniosize); 104SYSCTL_UINT(_kern_ktrace, OID_AUTO, genio_size, CTLFLAG_RW, &ktr_geniosize, 105 0, "Maximum size of genio event payload"); 106 107static int print_message = 1; 108struct mtx ktrace_mtx; 109static struct sema ktrace_sema; 110 111static void ktrace_init(void *dummy); 112static int sysctl_kern_ktrace_request_pool(SYSCTL_HANDLER_ARGS); 113static uint ktrace_resize_pool(uint newsize); 114static struct ktr_request *ktr_getrequest(int type); 115static void ktr_submitrequest(struct ktr_request *req); 116static void ktr_freerequest(struct ktr_request *req); 117static void ktr_loop(void *dummy); 118static void ktr_writerequest(struct ktr_request *req); 119static int ktrcanset(struct thread *,struct proc *); 120static int ktrsetchildren(struct thread *,struct proc *,int,int,struct vnode *); 121static int ktrops(struct thread *,struct proc *,int,int,struct vnode *); 122 123static void 124ktrace_init(void *dummy) 125{ 126 struct ktr_request *req; 127 int i; 128 129 mtx_init(&ktrace_mtx, "ktrace", NULL, MTX_DEF | MTX_QUIET); 130 sema_init(&ktrace_sema, 0, "ktrace"); 131 STAILQ_INIT(&ktr_todo); 132 STAILQ_INIT(&ktr_free); 133 for (i = 0; i < ktr_requestpool; i++) { 134 req = malloc(sizeof(struct ktr_request), M_KTRACE, M_WAITOK); 135 STAILQ_INSERT_HEAD(&ktr_free, req, ktr_list); 136 } 137 kthread_create(ktr_loop, NULL, NULL, RFHIGHPID, 0, "ktrace"); 138} 139SYSINIT(ktrace_init, SI_SUB_KTRACE, SI_ORDER_ANY, ktrace_init, NULL); 140 141static int 142sysctl_kern_ktrace_request_pool(SYSCTL_HANDLER_ARGS) 143{ 144 struct thread *td; 145 uint newsize, oldsize, wantsize; 146 int error; 147 148 /* Handle easy read-only case first to avoid warnings from GCC. */ 149 if (!req->newptr) { 150 mtx_lock(&ktrace_mtx); 151 oldsize = ktr_requestpool; 152 mtx_unlock(&ktrace_mtx); 153 return (SYSCTL_OUT(req, &oldsize, sizeof(uint))); 154 } 155 156 error = SYSCTL_IN(req, &wantsize, sizeof(uint)); 157 if (error) 158 return (error); 159 td = curthread; 160 td->td_inktrace = 1; 161 mtx_lock(&ktrace_mtx); 162 oldsize = ktr_requestpool; 163 newsize = ktrace_resize_pool(wantsize); 164 mtx_unlock(&ktrace_mtx); 165 td->td_inktrace = 0; 166 error = SYSCTL_OUT(req, &oldsize, sizeof(uint)); 167 if (error) 168 return (error); 169 if (newsize != wantsize) 170 return (ENOSPC); 171 return (0); 172} 173SYSCTL_PROC(_kern_ktrace, OID_AUTO, request_pool, CTLTYPE_UINT|CTLFLAG_RW, 174 &ktr_requestpool, 0, sysctl_kern_ktrace_request_pool, "IU", ""); 175 176static uint 177ktrace_resize_pool(uint newsize) 178{ 179 struct ktr_request *req; 180 181 mtx_assert(&ktrace_mtx, MA_OWNED); 182 print_message = 1; 183 if (newsize == ktr_requestpool) 184 return (newsize); 185 if (newsize < ktr_requestpool) 186 /* Shrink pool down to newsize if possible. */ 187 while (ktr_requestpool > newsize) { 188 req = STAILQ_FIRST(&ktr_free); 189 if (req == NULL) 190 return (ktr_requestpool); 191 STAILQ_REMOVE_HEAD(&ktr_free, ktr_list); 192 ktr_requestpool--; 193 mtx_unlock(&ktrace_mtx); 194 free(req, M_KTRACE); 195 mtx_lock(&ktrace_mtx); 196 } 197 else 198 /* Grow pool up to newsize. */ 199 while (ktr_requestpool < newsize) { 200 mtx_unlock(&ktrace_mtx); 201 req = malloc(sizeof(struct ktr_request), M_KTRACE, 202 M_WAITOK); 203 mtx_lock(&ktrace_mtx); 204 STAILQ_INSERT_HEAD(&ktr_free, req, ktr_list); 205 ktr_requestpool++; 206 } 207 return (ktr_requestpool); 208} 209 210static struct ktr_request * 211ktr_getrequest(int type) 212{ 213 struct ktr_request *req; 214 struct thread *td = curthread; 215 struct proc *p = td->td_proc; 216 int pm; 217 218 td->td_inktrace = 1; 219 mtx_lock(&ktrace_mtx); 220 if (!KTRCHECK(td, type)) { 221 mtx_unlock(&ktrace_mtx); 222 td->td_inktrace = 0; 223 return (NULL); 224 } 225 req = STAILQ_FIRST(&ktr_free); 226 if (req != NULL) { 227 STAILQ_REMOVE_HEAD(&ktr_free, ktr_list); 228 req->ktr_header.ktr_type = type; 229 if (p->p_traceflag & KTRFAC_DROP) { 230 req->ktr_header.ktr_type |= KTR_DROP; 231 p->p_traceflag &= ~KTRFAC_DROP; 232 } 233 KASSERT(p->p_tracevp != NULL, ("ktrace: no trace vnode")); 234 KASSERT(p->p_tracecred != NULL, ("ktrace: no trace cred")); 235 req->ktr_vp = p->p_tracevp; 236 VREF(p->p_tracevp); 237 req->ktr_cred = crhold(p->p_tracecred); 238 mtx_unlock(&ktrace_mtx); 239 microtime(&req->ktr_header.ktr_time); 240 req->ktr_header.ktr_pid = p->p_pid; 241 bcopy(p->p_comm, req->ktr_header.ktr_comm, MAXCOMLEN + 1); 242 req->ktr_header.ktr_buffer = NULL; 243 req->ktr_header.ktr_len = 0; 244 } else { 245 p->p_traceflag |= KTRFAC_DROP; 246 pm = print_message; 247 print_message = 0; 248 mtx_unlock(&ktrace_mtx); 249 if (pm) 250 printf("Out of ktrace request objects.\n"); 251 td->td_inktrace = 0; 252 } 253 return (req); 254} 255 256static void 257ktr_submitrequest(struct ktr_request *req) 258{ 259 260 mtx_lock(&ktrace_mtx); 261 STAILQ_INSERT_TAIL(&ktr_todo, req, ktr_list); 262 sema_post(&ktrace_sema); 263 mtx_unlock(&ktrace_mtx); 264 curthread->td_inktrace = 0; 265} 266 267static void 268ktr_freerequest(struct ktr_request *req) 269{ 270 271 crfree(req->ktr_cred); 272 if (req->ktr_vp != NULL) { 273 mtx_lock(&Giant); 274 vrele(req->ktr_vp); 275 mtx_unlock(&Giant); 276 } 277 if (req->ktr_header.ktr_buffer != NULL) 278 free(req->ktr_header.ktr_buffer, M_KTRACE); 279 mtx_lock(&ktrace_mtx); 280 STAILQ_INSERT_HEAD(&ktr_free, req, ktr_list); 281 mtx_unlock(&ktrace_mtx); 282} 283 284static void 285ktr_loop(void *dummy) 286{ 287 struct ktr_request *req; 288 struct thread *td; 289 struct ucred *cred; 290 291 /* Only cache these values once. */ 292 td = curthread; 293 cred = td->td_ucred; 294 for (;;) { 295 sema_wait(&ktrace_sema); 296 mtx_lock(&ktrace_mtx); 297 req = STAILQ_FIRST(&ktr_todo); 298 STAILQ_REMOVE_HEAD(&ktr_todo, ktr_list); 299 KASSERT(req != NULL, ("got a NULL request")); 300 mtx_unlock(&ktrace_mtx); 301 /* 302 * It is not enough just to pass the cached cred 303 * to the VOP's in ktr_writerequest(). Some VFS 304 * operations use curthread->td_ucred, so we need 305 * to modify our thread's credentials as well. 306 * Evil. 307 */ 308 td->td_ucred = req->ktr_cred; 309 ktr_writerequest(req); 310 td->td_ucred = cred; 311 ktr_freerequest(req); 312 } 313} 314 315/* 316 * MPSAFE 317 */ 318void 319ktrsyscall(code, narg, args) 320 int code, narg; 321 register_t args[]; 322{ 323 struct ktr_request *req; 324 struct ktr_syscall *ktp; 325 size_t buflen; 326 char *buf = NULL; 327 328 buflen = sizeof(register_t) * narg; 329 if (buflen > 0) { 330 buf = malloc(buflen, M_KTRACE, M_WAITOK); 331 bcopy(args, buf, buflen); 332 } 333 req = ktr_getrequest(KTR_SYSCALL); 334 if (req == NULL) { 335 if (buf != NULL) 336 free(buf, M_KTRACE); 337 return; 338 } 339 ktp = &req->ktr_data.ktr_syscall; 340 ktp->ktr_code = code; 341 ktp->ktr_narg = narg; 342 if (buflen > 0) { 343 req->ktr_header.ktr_len = buflen; 344 req->ktr_header.ktr_buffer = buf; 345 } 346 ktr_submitrequest(req); 347} 348 349/* 350 * MPSAFE 351 */ 352void 353ktrsysret(code, error, retval) 354 int code, error; 355 register_t retval; 356{ 357 struct ktr_request *req; 358 struct ktr_sysret *ktp; 359 360 req = ktr_getrequest(KTR_SYSRET); 361 if (req == NULL) 362 return; 363 ktp = &req->ktr_data.ktr_sysret; 364 ktp->ktr_code = code; 365 ktp->ktr_error = error; 366 ktp->ktr_retval = retval; /* what about val2 ? */ 367 ktr_submitrequest(req); 368} 369 370void 371ktrnamei(path) 372 char *path; 373{ 374 struct ktr_request *req; 375 int namelen; 376 char *buf = NULL; 377 378 namelen = strlen(path); 379 if (namelen > 0) { 380 buf = malloc(namelen, M_KTRACE, M_WAITOK); 381 bcopy(path, buf, namelen); 382 } 383 req = ktr_getrequest(KTR_NAMEI); 384 if (req == NULL) { 385 if (buf != NULL) 386 free(buf, M_KTRACE); 387 return; 388 } 389 if (namelen > 0) { 390 req->ktr_header.ktr_len = namelen; 391 req->ktr_header.ktr_buffer = buf; 392 } 393 ktr_submitrequest(req); 394} 395 396/* 397 * Since the uio may not stay valid, we can not hand off this request to 398 * the thread and need to process it synchronously. However, we wish to 399 * keep the relative order of records in a trace file correct, so we 400 * do put this request on the queue (if it isn't empty) and then block. 401 * The ktrace thread waks us back up when it is time for this event to 402 * be posted and blocks until we have completed writing out the event 403 * and woken it back up. 404 */ 405void 406ktrgenio(fd, rw, uio, error) 407 int fd; 408 enum uio_rw rw; 409 struct uio *uio; 410 int error; 411{ 412 struct ktr_request *req; 413 struct ktr_genio *ktg; 414 int datalen; 415 char *buf; 416 417 if (error) 418 return; 419 uio->uio_offset = 0; 420 uio->uio_rw = UIO_WRITE; 421 datalen = imin(uio->uio_resid, ktr_geniosize); 422 buf = malloc(datalen, M_KTRACE, M_WAITOK); 423 if (uiomove(buf, datalen, uio)) { 424 free(buf, M_KTRACE); 425 return; 426 } 427 req = ktr_getrequest(KTR_GENIO); 428 if (req == NULL) { 429 free(buf, M_KTRACE); 430 return; 431 } 432 ktg = &req->ktr_data.ktr_genio; 433 ktg->ktr_fd = fd; 434 ktg->ktr_rw = rw; 435 req->ktr_header.ktr_len = datalen; 436 req->ktr_header.ktr_buffer = buf; 437 ktr_submitrequest(req); 438} 439 440void 441ktrpsig(sig, action, mask, code) 442 int sig; 443 sig_t action; 444 sigset_t *mask; 445 int code; 446{ 447 struct ktr_request *req; 448 struct ktr_psig *kp; 449 450 req = ktr_getrequest(KTR_PSIG); 451 if (req == NULL) 452 return; 453 kp = &req->ktr_data.ktr_psig; 454 kp->signo = (char)sig; 455 kp->action = action; 456 kp->mask = *mask; 457 kp->code = code; 458 ktr_submitrequest(req); 459} 460 461void 462ktrcsw(out, user) 463 int out, user; 464{ 465 struct ktr_request *req; 466 struct ktr_csw *kc; 467 468 req = ktr_getrequest(KTR_CSW); 469 if (req == NULL) 470 return; 471 kc = &req->ktr_data.ktr_csw; 472 kc->out = out; 473 kc->user = user; 474 ktr_submitrequest(req); 475} 476#endif 477 478/* Interface and common routines */ 479 480/* 481 * ktrace system call 482 */ 483#ifndef _SYS_SYSPROTO_H_ 484struct ktrace_args { 485 char *fname; 486 int ops; 487 int facs; 488 int pid; 489}; 490#endif 491/* ARGSUSED */ 492int 493ktrace(td, uap) 494 struct thread *td; 495 register struct ktrace_args *uap; 496{ 497#ifdef KTRACE 498 register struct vnode *vp = NULL; 499 register struct proc *p; 500 struct pgrp *pg; 501 int facs = uap->facs & ~KTRFAC_ROOT; 502 int ops = KTROP(uap->ops); 503 int descend = uap->ops & KTRFLAG_DESCEND; 504 int ret = 0; 505 int flags, error = 0; 506 struct nameidata nd; 507 struct ucred *cred; 508 509 td->td_inktrace = 1; 510 if (ops != KTROP_CLEAR) { 511 /* 512 * an operation which requires a file argument. 513 */ 514 NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->fname, td); 515 flags = FREAD | FWRITE | O_NOFOLLOW; 516 error = vn_open(&nd, &flags, 0); 517 if (error) { 518 td->td_inktrace = 0; 519 return (error); 520 } 521 NDFREE(&nd, NDF_ONLY_PNBUF); 522 vp = nd.ni_vp; 523 VOP_UNLOCK(vp, 0, td); 524 if (vp->v_type != VREG) { 525 (void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td); 526 td->td_inktrace = 0; 527 return (EACCES); 528 } 529 } 530 /* 531 * Clear all uses of the tracefile. 532 */ 533 if (ops == KTROP_CLEARFILE) { 534 sx_slock(&allproc_lock); 535 LIST_FOREACH(p, &allproc, p_list) { 536 PROC_LOCK(p); 537 if (p->p_tracevp == vp) { 538 if (ktrcanset(td, p)) { 539 mtx_lock(&ktrace_mtx); 540 cred = p->p_tracecred; 541 p->p_tracecred = NULL; 542 p->p_tracevp = NULL; 543 p->p_traceflag = 0; 544 mtx_unlock(&ktrace_mtx); 545 PROC_UNLOCK(p); 546 (void) vn_close(vp, FREAD|FWRITE, 547 cred, td); 548 crfree(cred); 549 } else { 550 PROC_UNLOCK(p); 551 error = EPERM; 552 } 553 } else 554 PROC_UNLOCK(p); 555 } 556 sx_sunlock(&allproc_lock); 557 goto done; 558 } 559 /* 560 * need something to (un)trace (XXX - why is this here?) 561 */ 562 if (!facs) { 563 error = EINVAL; 564 goto done; 565 } 566 /* 567 * do it 568 */ 569 if (uap->pid < 0) { 570 /* 571 * by process group 572 */ 573 sx_slock(&proctree_lock); 574 pg = pgfind(-uap->pid); 575 if (pg == NULL) { 576 sx_sunlock(&proctree_lock); 577 error = ESRCH; 578 goto done; 579 } 580 /* 581 * ktrops() may call vrele(). Lock pg_members 582 * by the proctree_lock rather than pg_mtx. 583 */ 584 PGRP_UNLOCK(pg); 585 LIST_FOREACH(p, &pg->pg_members, p_pglist) 586 if (descend) 587 ret |= ktrsetchildren(td, p, ops, facs, vp); 588 else 589 ret |= ktrops(td, p, ops, facs, vp); 590 sx_sunlock(&proctree_lock); 591 } else { 592 /* 593 * by pid 594 */ 595 p = pfind(uap->pid); 596 if (p == NULL) { 597 error = ESRCH; 598 goto done; 599 } 600 PROC_UNLOCK(p); 601 /* XXX: UNLOCK above has a race */ 602 if (descend) 603 ret |= ktrsetchildren(td, p, ops, facs, vp); 604 else 605 ret |= ktrops(td, p, ops, facs, vp); 606 } 607 if (!ret) 608 error = EPERM; 609done: 610 if (vp != NULL) 611 (void) vn_close(vp, FWRITE, td->td_ucred, td); 612 td->td_inktrace = 0; 613 return (error); 614#else 615 return ENOSYS; 616#endif 617} 618 619/* 620 * utrace system call 621 */ 622/* ARGSUSED */ 623int 624utrace(td, uap) 625 struct thread *td; 626 register struct utrace_args *uap; 627{ 628 629#ifdef KTRACE 630 struct ktr_request *req; 631 void *cp; 632 int error; 633 634 if (!KTRPOINT(td, KTR_USER)) 635 return (0); 636 if (uap->len > KTR_USER_MAXLEN) 637 return (EINVAL); 638 cp = malloc(uap->len, M_KTRACE, M_WAITOK); 639 error = copyin(uap->addr, cp, uap->len); 640 if (error) { 641 free(cp, M_KTRACE); 642 return (error); 643 } 644 req = ktr_getrequest(KTR_USER); 645 if (req == NULL) { 646 free(cp, M_KTRACE); 647 return (0); 648 } 649 req->ktr_header.ktr_buffer = cp; 650 req->ktr_header.ktr_len = uap->len; 651 ktr_submitrequest(req); 652 return (0); 653#else 654 return (ENOSYS); 655#endif 656} 657 658#ifdef KTRACE 659static int 660ktrops(td, p, ops, facs, vp) 661 struct thread *td; 662 struct proc *p; 663 int ops, facs; 664 struct vnode *vp; 665{ 666 struct vnode *tracevp = NULL; 667 struct ucred *tracecred = NULL; 668 669 PROC_LOCK(p); 670 if (!ktrcanset(td, p)) { 671 PROC_UNLOCK(p); 672 return (0); 673 } 674 mtx_lock(&ktrace_mtx); 675 if (ops == KTROP_SET) { 676 if (p->p_tracevp != vp) { 677 /* 678 * if trace file already in use, relinquish below 679 */ 680 tracevp = p->p_tracevp; 681 VREF(vp); 682 p->p_tracevp = vp; 683 } 684 if (p->p_tracecred != td->td_ucred) { 685 tracecred = p->p_tracecred; 686 p->p_tracecred = crhold(td->td_ucred); 687 } 688 p->p_traceflag |= facs; 689 if (td->td_ucred->cr_uid == 0) 690 p->p_traceflag |= KTRFAC_ROOT; 691 } else { 692 /* KTROP_CLEAR */ 693 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) { 694 /* no more tracing */ 695 p->p_traceflag = 0; 696 tracevp = p->p_tracevp; 697 p->p_tracevp = NULL; 698 tracecred = p->p_tracecred; 699 p->p_tracecred = NULL; 700 } 701 } 702 mtx_unlock(&ktrace_mtx); 703 PROC_UNLOCK(p); 704 if (tracevp != NULL) 705 vrele(tracevp); 706 if (tracecred != NULL) 707 crfree(tracecred); 708 709 return (1); 710} 711 712static int 713ktrsetchildren(td, top, ops, facs, vp) 714 struct thread *td; 715 struct proc *top; 716 int ops, facs; 717 struct vnode *vp; 718{ 719 register struct proc *p; 720 register int ret = 0; 721 722 p = top; 723 sx_slock(&proctree_lock); 724 for (;;) { 725 ret |= ktrops(td, p, ops, facs, vp); 726 /* 727 * If this process has children, descend to them next, 728 * otherwise do any siblings, and if done with this level, 729 * follow back up the tree (but not past top). 730 */ 731 if (!LIST_EMPTY(&p->p_children)) 732 p = LIST_FIRST(&p->p_children); 733 else for (;;) { 734 if (p == top) { 735 sx_sunlock(&proctree_lock); 736 return (ret); 737 } 738 if (LIST_NEXT(p, p_sibling)) { 739 p = LIST_NEXT(p, p_sibling); 740 break; 741 } 742 p = p->p_pptr; 743 } 744 } 745 /*NOTREACHED*/ 746} 747 748static void 749ktr_writerequest(struct ktr_request *req) 750{ 751 struct ktr_header *kth; 752 struct vnode *vp; 753 struct proc *p; 754 struct thread *td; 755 struct ucred *cred; 756 struct uio auio; 757 struct iovec aiov[3]; 758 struct mount *mp; 759 int datalen, buflen, vrele_count; 760 int error; 761 762 vp = req->ktr_vp; 763 /* 764 * If vp is NULL, the vp has been cleared out from under this 765 * request, so just drop it. 766 */ 767 if (vp == NULL) 768 return; 769 kth = &req->ktr_header; 770 datalen = data_lengths[(ushort)kth->ktr_type & ~KTR_DROP]; 771 buflen = kth->ktr_len; 772 cred = req->ktr_cred; 773 td = curthread; 774 auio.uio_iov = &aiov[0]; 775 auio.uio_offset = 0; 776 auio.uio_segflg = UIO_SYSSPACE; 777 auio.uio_rw = UIO_WRITE; 778 aiov[0].iov_base = (caddr_t)kth; 779 aiov[0].iov_len = sizeof(struct ktr_header); 780 auio.uio_resid = sizeof(struct ktr_header); 781 auio.uio_iovcnt = 1; 782 auio.uio_td = td; 783 if (datalen != 0) { 784 aiov[1].iov_base = (caddr_t)&req->ktr_data; 785 aiov[1].iov_len = datalen; 786 auio.uio_resid += datalen; 787 auio.uio_iovcnt++; 788 kth->ktr_len += datalen; 789 } 790 if (buflen != 0) { 791 KASSERT(kth->ktr_buffer != NULL, ("ktrace: nothing to write")); 792 aiov[auio.uio_iovcnt].iov_base = kth->ktr_buffer; 793 aiov[auio.uio_iovcnt].iov_len = buflen; 794 auio.uio_resid += buflen; 795 auio.uio_iovcnt++; 796 } 797 mtx_lock(&Giant); 798 vn_start_write(vp, &mp, V_WAIT); 799 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 800 (void)VOP_LEASE(vp, td, cred, LEASE_WRITE); 801#ifdef MAC 802 error = mac_check_vnode_write(cred, NOCRED, vp); 803 if (error == 0) 804#endif 805 error = VOP_WRITE(vp, &auio, IO_UNIT | IO_APPEND, cred); 806 VOP_UNLOCK(vp, 0, td); 807 vn_finished_write(mp); 808 mtx_unlock(&Giant); 809 if (!error) 810 return; 811 /* 812 * If error encountered, give up tracing on this vnode. We defer 813 * all the vrele()'s on the vnode until after we are finished walking 814 * the various lists to avoid needlessly holding locks. 815 */ 816 log(LOG_NOTICE, "ktrace write failed, errno %d, tracing stopped\n", 817 error); 818 vrele_count = 0; 819 /* 820 * First, clear this vnode from being used by any processes in the 821 * system. 822 * XXX - If one process gets an EPERM writing to the vnode, should 823 * we really do this? Other processes might have suitable 824 * credentials for the operation. 825 */ 826 cred = NULL; 827 sx_slock(&allproc_lock); 828 LIST_FOREACH(p, &allproc, p_list) { 829 PROC_LOCK(p); 830 if (p->p_tracevp == vp) { 831 mtx_lock(&ktrace_mtx); 832 p->p_tracevp = NULL; 833 p->p_traceflag = 0; 834 cred = p->p_tracecred; 835 p->p_tracecred = NULL; 836 mtx_unlock(&ktrace_mtx); 837 vrele_count++; 838 } 839 PROC_UNLOCK(p); 840 if (cred != NULL) { 841 crfree(cred); 842 cred = NULL; 843 } 844 } 845 sx_sunlock(&allproc_lock); 846 /* 847 * Second, clear this vnode from any pending requests. 848 */ 849 mtx_lock(&ktrace_mtx); 850 STAILQ_FOREACH(req, &ktr_todo, ktr_list) { 851 if (req->ktr_vp == vp) { 852 req->ktr_vp = NULL; 853 vrele_count++; 854 } 855 } 856 mtx_unlock(&ktrace_mtx); 857 mtx_lock(&Giant); 858 while (vrele_count-- > 0) 859 vrele(vp); 860 mtx_unlock(&Giant); 861} 862 863/* 864 * Return true if caller has permission to set the ktracing state 865 * of target. Essentially, the target can't possess any 866 * more permissions than the caller. KTRFAC_ROOT signifies that 867 * root previously set the tracing status on the target process, and 868 * so, only root may further change it. 869 */ 870static int 871ktrcanset(td, targetp) 872 struct thread *td; 873 struct proc *targetp; 874{ 875 876 PROC_LOCK_ASSERT(targetp, MA_OWNED); 877 if (targetp->p_traceflag & KTRFAC_ROOT && 878 suser_cred(td->td_ucred, PRISON_ROOT)) 879 return (0); 880 881 if (p_candebug(td, targetp) != 0) 882 return (0); 883 884 return (1); 885} 886 887#endif /* KTRACE */ 888