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: src/sys/kern/kern_ktrace.c,v 1.35.2.6 2002/07/05 22:36:38 darrenr Exp $ 35 */ 36 37 #include "opt_ktrace.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/sysproto.h> 42 #include <sys/kernel.h> 43 #include <sys/proc.h> 44 #include <sys/fcntl.h> 45 #include <sys/lock.h> 46 #include <sys/nlookup.h> 47 #include <sys/vnode.h> 48 #include <sys/ktrace.h> 49 #include <sys/malloc.h> 50 #include <sys/syslog.h> 51 #include <sys/sysent.h> 52 53 #include <vm/vm_zone.h> 54 55 #include <sys/mplock2.h> 56 57 static MALLOC_DEFINE(M_KTRACE, "KTRACE", "KTRACE"); 58 59 #ifdef KTRACE 60 static struct ktr_header *ktrgetheader (int type); 61 static void ktrwrite (struct lwp *, struct ktr_header *, struct uio *); 62 static int ktrcanset (struct thread *,struct proc *); 63 static int ktrsetchildren (struct thread *,struct proc *,int,int, ktrace_node_t); 64 static int ktrops (struct thread *,struct proc *,int,int, ktrace_node_t); 65 66 /* 67 * MPSAFE 68 */ 69 static struct ktr_header * 70 ktrgetheader(int type) 71 { 72 struct ktr_header *kth; 73 struct proc *p = curproc; /* XXX */ 74 struct lwp *lp = curthread->td_lwp; 75 76 kth = kmalloc(sizeof(struct ktr_header), M_KTRACE, M_WAITOK); 77 kth->ktr_type = type; 78 /* XXX threaded flag is a hack at the moment */ 79 kth->ktr_flags = (p->p_nthreads > 1) ? KTRH_THREADED : 0; 80 microtime(&kth->ktr_time); 81 kth->ktr_pid = p->p_pid; 82 kth->ktr_tid = lp->lwp_tid; 83 bcopy(p->p_comm, kth->ktr_comm, MAXCOMLEN + 1); 84 return (kth); 85 } 86 87 void 88 ktrsyscall(struct lwp *lp, int code, int narg, register_t args[]) 89 { 90 struct ktr_header *kth; 91 struct ktr_syscall *ktp; 92 int len; 93 register_t *argp; 94 int i; 95 96 len = offsetof(struct ktr_syscall, ktr_args) + 97 (narg * sizeof(register_t)); 98 99 /* 100 * Setting the active bit prevents a ktrace recursion from the 101 * ktracing op itself. 102 */ 103 lp->lwp_traceflag |= KTRFAC_ACTIVE; 104 kth = ktrgetheader(KTR_SYSCALL); 105 ktp = kmalloc(len, M_KTRACE, M_WAITOK); 106 ktp->ktr_code = code; 107 ktp->ktr_narg = narg; 108 argp = &ktp->ktr_args[0]; 109 for (i = 0; i < narg; i++) 110 *argp++ = args[i]; 111 kth->ktr_buf = (caddr_t)ktp; 112 kth->ktr_len = len; 113 ktrwrite(lp, kth, NULL); 114 kfree(ktp, M_KTRACE); 115 kfree(kth, M_KTRACE); 116 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 117 } 118 119 void 120 ktrsysret(struct lwp *lp, int code, int error, register_t retval) 121 { 122 struct ktr_header *kth; 123 struct ktr_sysret ktp; 124 125 lp->lwp_traceflag |= KTRFAC_ACTIVE; 126 kth = ktrgetheader(KTR_SYSRET); 127 ktp.ktr_code = code; 128 ktp.ktr_error = error; 129 if (error == 0) 130 ktp.ktr_retval = retval; /* what about val2 ? */ 131 else 132 ktp.ktr_retval = 0; 133 134 kth->ktr_buf = (caddr_t)&ktp; 135 kth->ktr_len = sizeof(struct ktr_sysret); 136 137 ktrwrite(lp, kth, NULL); 138 kfree(kth, M_KTRACE); 139 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 140 } 141 142 void 143 ktrnamei(struct lwp *lp, char *path) 144 { 145 struct ktr_header *kth; 146 147 lp->lwp_traceflag |= KTRFAC_ACTIVE; 148 kth = ktrgetheader(KTR_NAMEI); 149 kth->ktr_len = strlen(path); 150 kth->ktr_buf = path; 151 152 ktrwrite(lp, kth, NULL); 153 kfree(kth, M_KTRACE); 154 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 155 } 156 157 void 158 ktrgenio(struct lwp *lp, int fd, enum uio_rw rw, struct uio *uio, int error) 159 { 160 struct ktr_header *kth; 161 struct ktr_genio ktg; 162 163 if (error) 164 return; 165 lp->lwp_traceflag |= KTRFAC_ACTIVE; 166 kth = ktrgetheader(KTR_GENIO); 167 ktg.ktr_fd = fd; 168 ktg.ktr_rw = rw; 169 kth->ktr_buf = (caddr_t)&ktg; 170 kth->ktr_len = sizeof(struct ktr_genio); 171 uio->uio_offset = 0; 172 uio->uio_rw = UIO_WRITE; 173 174 ktrwrite(lp, kth, uio); 175 kfree(kth, M_KTRACE); 176 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 177 } 178 179 void 180 ktrpsig(struct lwp *lp, int sig, sig_t action, sigset_t *mask, int code) 181 { 182 struct ktr_header *kth; 183 struct ktr_psig kp; 184 185 lp->lwp_traceflag |= KTRFAC_ACTIVE; 186 kth = ktrgetheader(KTR_PSIG); 187 kp.signo = (char)sig; 188 kp.action = action; 189 kp.mask = *mask; 190 kp.code = code; 191 kth->ktr_buf = (caddr_t)&kp; 192 kth->ktr_len = sizeof (struct ktr_psig); 193 194 ktrwrite(lp, kth, NULL); 195 kfree(kth, M_KTRACE); 196 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 197 } 198 199 void 200 ktrcsw(struct lwp *lp, int out, int user) 201 { 202 struct ktr_header *kth; 203 struct ktr_csw kc; 204 205 lp->lwp_traceflag |= KTRFAC_ACTIVE; 206 kth = ktrgetheader(KTR_CSW); 207 kc.out = out; 208 kc.user = user; 209 kth->ktr_buf = (caddr_t)&kc; 210 kth->ktr_len = sizeof (struct ktr_csw); 211 212 ktrwrite(lp, kth, NULL); 213 kfree(kth, M_KTRACE); 214 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 215 } 216 #endif 217 218 /* Interface and common routines */ 219 220 #ifdef KTRACE 221 /* 222 * ktrace system call 223 */ 224 struct ktrace_clear_info { 225 ktrace_node_t tracenode; 226 int rootclear; 227 int error; 228 }; 229 230 static int ktrace_clear_callback(struct proc *p, void *data); 231 232 #endif 233 234 /* 235 * MPALMOSTSAFE 236 */ 237 int 238 sys_ktrace(struct ktrace_args *uap) 239 { 240 #ifdef KTRACE 241 struct ktrace_clear_info info; 242 struct thread *td = curthread; 243 struct proc *curp = td->td_proc; 244 struct proc *p; 245 struct pgrp *pg; 246 int facs = uap->facs & ~KTRFAC_ROOT; 247 int ops = KTROP(uap->ops); 248 int descend = uap->ops & KTRFLAG_DESCEND; 249 int ret = 0; 250 int error = 0; 251 struct nlookupdata nd; 252 ktrace_node_t tracenode = NULL; 253 254 get_mplock(); 255 curp->p_traceflag |= KTRFAC_ACTIVE; 256 if (ops != KTROP_CLEAR) { 257 /* 258 * an operation which requires a file argument. 259 */ 260 error = nlookup_init(&nd, uap->fname, 261 UIO_USERSPACE, NLC_LOCKVP); 262 if (error == 0) 263 error = vn_open(&nd, NULL, FREAD|FWRITE|O_NOFOLLOW, 0); 264 if (error == 0 && nd.nl_open_vp->v_type != VREG) 265 error = EACCES; 266 if (error) { 267 curp->p_traceflag &= ~KTRFAC_ACTIVE; 268 nlookup_done(&nd); 269 goto done; 270 } 271 tracenode = kmalloc(sizeof(struct ktrace_node), M_KTRACE, 272 M_WAITOK | M_ZERO); 273 tracenode->kn_vp = nd.nl_open_vp; 274 tracenode->kn_refs = 1; 275 nd.nl_open_vp = NULL; 276 nlookup_done(&nd); 277 vn_unlock(tracenode->kn_vp); 278 } 279 /* 280 * Clear all uses of the tracefile. Not the most efficient operation 281 * in the world. 282 */ 283 if (ops == KTROP_CLEARFILE) { 284 info.tracenode = tracenode; 285 info.error = 0; 286 info.rootclear = 0; 287 allproc_scan(ktrace_clear_callback, &info); 288 error = info.error; 289 goto done; 290 } 291 /* 292 * need something to (un)trace (XXX - why is this here?) 293 */ 294 if (!facs) { 295 error = EINVAL; 296 goto done; 297 } 298 /* 299 * do it 300 */ 301 if (uap->pid < 0) { 302 /* 303 * By process group. Process group is referenced, preventing 304 * disposal. 305 */ 306 pg = pgfind(-uap->pid); 307 if (pg == NULL) { 308 error = ESRCH; 309 goto done; 310 } 311 lwkt_gettoken(&pg->pg_token); 312 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 313 PHOLD(p); 314 if (descend) 315 ret |= ktrsetchildren(td, p, ops, facs, tracenode); 316 else 317 ret |= ktrops(td, p, ops, facs, tracenode); 318 PRELE(p); 319 } 320 lwkt_reltoken(&pg->pg_token); 321 pgrel(pg); 322 } else { 323 /* 324 * by pid 325 */ 326 p = pfind(uap->pid); 327 if (p == NULL) { 328 error = ESRCH; 329 goto done; 330 } 331 if (descend) 332 ret |= ktrsetchildren(td, p, ops, facs, tracenode); 333 else 334 ret |= ktrops(td, p, ops, facs, tracenode); 335 PRELE(p); 336 } 337 if (!ret) 338 error = EPERM; 339 done: 340 if (tracenode) 341 ktrdestroy(&tracenode); 342 curp->p_traceflag &= ~KTRFAC_ACTIVE; 343 rel_mplock(); 344 return (error); 345 #else 346 return ENOSYS; 347 #endif 348 } 349 350 #ifdef KTRACE 351 352 /* 353 * NOTE: NOT MPSAFE (yet) 354 */ 355 static int 356 ktrace_clear_callback(struct proc *p, void *data) 357 { 358 struct ktrace_clear_info *info = data; 359 360 if (p->p_tracenode) { 361 if (info->rootclear) { 362 if (p->p_tracenode == info->tracenode) { 363 ktrdestroy(&p->p_tracenode); 364 p->p_traceflag = 0; 365 } 366 } else { 367 if (p->p_tracenode->kn_vp == info->tracenode->kn_vp) { 368 if (ktrcanset(curthread, p)) { 369 ktrdestroy(&p->p_tracenode); 370 p->p_traceflag = 0; 371 } else { 372 info->error = EPERM; 373 } 374 } 375 } 376 } 377 return(0); 378 } 379 380 #endif 381 382 /* 383 * utrace system call 384 * 385 * MPALMOSTSAFE 386 */ 387 int 388 sys_utrace(struct utrace_args *uap) 389 { 390 #ifdef KTRACE 391 struct ktr_header *kth; 392 struct thread *td = curthread; /* XXX */ 393 caddr_t cp; 394 395 if (!KTRPOINT(td, KTR_USER)) 396 return (0); 397 if (uap->len > KTR_USER_MAXLEN) 398 return (EINVAL); 399 td->td_lwp->lwp_traceflag |= KTRFAC_ACTIVE; 400 kth = ktrgetheader(KTR_USER); 401 cp = kmalloc(uap->len, M_KTRACE, M_WAITOK); 402 if (!copyin(uap->addr, cp, uap->len)) { 403 kth->ktr_buf = cp; 404 kth->ktr_len = uap->len; 405 ktrwrite(td->td_lwp, kth, NULL); 406 } 407 kfree(kth, M_KTRACE); 408 kfree(cp, M_KTRACE); 409 td->td_lwp->lwp_traceflag &= ~KTRFAC_ACTIVE; 410 411 return (0); 412 #else 413 return (ENOSYS); 414 #endif 415 } 416 417 void 418 ktrdestroy(struct ktrace_node **tracenodep) 419 { 420 ktrace_node_t tracenode; 421 422 if ((tracenode = *tracenodep) != NULL) { 423 *tracenodep = NULL; 424 KKASSERT(tracenode->kn_refs > 0); 425 if (atomic_fetchadd_int(&tracenode->kn_refs, -1) == 1) { 426 vn_close(tracenode->kn_vp, FREAD|FWRITE); 427 tracenode->kn_vp = NULL; 428 kfree(tracenode, M_KTRACE); 429 } 430 } 431 } 432 433 /* 434 * This allows a process to inherit a ref on a tracenode and is also used 435 * as a temporary ref to prevent a tracenode from being destroyed out from 436 * under an active operation. 437 */ 438 ktrace_node_t 439 ktrinherit(ktrace_node_t tracenode) 440 { 441 if (tracenode) { 442 KKASSERT(tracenode->kn_refs > 0); 443 atomic_add_int(&tracenode->kn_refs, 1); 444 } 445 return(tracenode); 446 } 447 448 #ifdef KTRACE 449 static int 450 ktrops(struct thread *td, struct proc *p, int ops, int facs, 451 ktrace_node_t tracenode) 452 { 453 ktrace_node_t oldnode; 454 455 if (!ktrcanset(td, p)) 456 return (0); 457 if (ops == KTROP_SET) { 458 if ((oldnode = p->p_tracenode) != tracenode) { 459 p->p_tracenode = ktrinherit(tracenode); 460 ktrdestroy(&oldnode); 461 } 462 p->p_traceflag |= facs; 463 if (td->td_ucred->cr_uid == 0) 464 p->p_traceflag |= KTRFAC_ROOT; 465 } else { 466 /* KTROP_CLEAR */ 467 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) { 468 /* no more tracing */ 469 p->p_traceflag = 0; 470 ktrdestroy(&p->p_tracenode); 471 } 472 } 473 474 return (1); 475 } 476 477 static int 478 ktrsetchildren(struct thread *td, struct proc *top, int ops, int facs, 479 ktrace_node_t tracenode) 480 { 481 struct proc *p; 482 struct proc *np; 483 int ret = 0; 484 485 p = top; 486 PHOLD(p); 487 lwkt_gettoken(&p->p_token); 488 489 for (;;) { 490 ret |= ktrops(td, p, ops, facs, tracenode); 491 492 /* 493 * If this process has children, descend to them next, 494 * otherwise do any siblings, and if done with this level, 495 * follow back up the tree (but not past top). 496 */ 497 if ((np = LIST_FIRST(&p->p_children)) != NULL) { 498 PHOLD(np); 499 } 500 while (np == NULL) { 501 if (p == top) 502 break; 503 if ((np = LIST_NEXT(p, p_sibling)) != NULL) { 504 PHOLD(np); 505 break; 506 } 507 508 /* 509 * recurse up to parent, set p in our inner 510 * loop when doing this. np can be NULL if 511 * we race a reparenting to init (thus 'top' 512 * is skipped past and never encountered). 513 */ 514 np = p->p_pptr; 515 if (np == NULL) 516 break; 517 PHOLD(np); 518 lwkt_reltoken(&p->p_token); 519 PRELE(p); 520 p = np; 521 lwkt_gettoken(&p->p_token); 522 np = NULL; 523 } 524 lwkt_reltoken(&p->p_token); 525 PRELE(p); 526 p = np; 527 if (p == NULL) 528 break; 529 /* Already held, but we need the token too */ 530 lwkt_gettoken(&p->p_token); 531 } 532 return (ret); 533 } 534 535 static void 536 ktrwrite(struct lwp *lp, struct ktr_header *kth, struct uio *uio) 537 { 538 struct ktrace_clear_info info; 539 struct uio auio; 540 struct iovec aiov[2]; 541 int error; 542 ktrace_node_t tracenode; 543 544 /* 545 * We have to ref our tracenode to prevent it from being ripped out 546 * from under us while we are trying to use it. p_tracenode can 547 * go away at any time if another process gets a write error. 548 * 549 * XXX not MP safe 550 */ 551 if (lp->lwp_proc->p_tracenode == NULL) 552 return; 553 tracenode = ktrinherit(lp->lwp_proc->p_tracenode); 554 auio.uio_iov = &aiov[0]; 555 auio.uio_offset = 0; 556 auio.uio_segflg = UIO_SYSSPACE; 557 auio.uio_rw = UIO_WRITE; 558 aiov[0].iov_base = (caddr_t)kth; 559 aiov[0].iov_len = sizeof(struct ktr_header); 560 auio.uio_resid = sizeof(struct ktr_header); 561 auio.uio_iovcnt = 1; 562 auio.uio_td = curthread; 563 if (kth->ktr_len > 0) { 564 auio.uio_iovcnt++; 565 aiov[1].iov_base = kth->ktr_buf; 566 aiov[1].iov_len = kth->ktr_len; 567 auio.uio_resid += kth->ktr_len; 568 if (uio != NULL) 569 kth->ktr_len += uio->uio_resid; 570 } 571 vn_lock(tracenode->kn_vp, LK_EXCLUSIVE | LK_RETRY); 572 error = VOP_WRITE(tracenode->kn_vp, &auio, 573 IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred); 574 if (error == 0 && uio != NULL) { 575 error = VOP_WRITE(tracenode->kn_vp, uio, 576 IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred); 577 } 578 vn_unlock(tracenode->kn_vp); 579 if (error) { 580 /* 581 * If an error occured, give up tracing on all processes 582 * using this tracenode. This is not MP safe but is 583 * blocking-safe. 584 */ 585 log(LOG_NOTICE, 586 "ktrace write failed, errno %d, tracing stopped\n", error); 587 info.tracenode = tracenode; 588 info.error = 0; 589 info.rootclear = 1; 590 allproc_scan(ktrace_clear_callback, &info); 591 } 592 ktrdestroy(&tracenode); 593 } 594 595 /* 596 * Return true if caller has permission to set the ktracing state 597 * of target. Essentially, the target can't possess any 598 * more permissions than the caller. KTRFAC_ROOT signifies that 599 * root previously set the tracing status on the target process, and 600 * so, only root may further change it. 601 * 602 * TODO: check groups. use caller effective gid. 603 */ 604 static int 605 ktrcanset(struct thread *calltd, struct proc *targetp) 606 { 607 struct ucred *caller = calltd->td_ucred; 608 struct ucred *target = targetp->p_ucred; 609 610 if (!PRISON_CHECK(caller, target)) 611 return (0); 612 if ((caller->cr_uid == target->cr_ruid && 613 target->cr_ruid == target->cr_svuid && 614 caller->cr_rgid == target->cr_rgid && /* XXX */ 615 target->cr_rgid == target->cr_svgid && 616 (targetp->p_traceflag & KTRFAC_ROOT) == 0 && 617 (targetp->p_flags & P_SUGID) == 0) || 618 caller->cr_uid == 0) 619 return (1); 620 621 return (0); 622 } 623 624 #endif /* KTRACE */ 625