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 * $DragonFly: src/sys/kern/kern_ktrace.c,v 1.30 2008/04/14 12:01:50 dillon Exp $ 36 */ 37 38 #include "opt_ktrace.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/sysproto.h> 43 #include <sys/kernel.h> 44 #include <sys/proc.h> 45 #include <sys/fcntl.h> 46 #include <sys/lock.h> 47 #include <sys/nlookup.h> 48 #include <sys/vnode.h> 49 #include <sys/ktrace.h> 50 #include <sys/malloc.h> 51 #include <sys/syslog.h> 52 #include <sys/sysent.h> 53 54 #include <vm/vm_zone.h> 55 static MALLOC_DEFINE(M_KTRACE, "KTRACE", "KTRACE"); 56 57 #ifdef KTRACE 58 static struct ktr_header *ktrgetheader (int type); 59 static void ktrwrite (struct lwp *, struct ktr_header *, struct uio *); 60 static int ktrcanset (struct thread *,struct proc *); 61 static int ktrsetchildren (struct thread *,struct proc *,int,int, ktrace_node_t); 62 static int ktrops (struct thread *,struct proc *,int,int, ktrace_node_t); 63 64 /* 65 * MPSAFE 66 */ 67 static struct ktr_header * 68 ktrgetheader(int type) 69 { 70 struct ktr_header *kth; 71 struct proc *p = curproc; /* XXX */ 72 struct lwp *lp = curthread->td_lwp; 73 74 MALLOC(kth, struct ktr_header *, sizeof (struct ktr_header), 75 M_KTRACE, M_WAITOK); 76 kth->ktr_type = type; 77 /* XXX threaded flag is a hack at the moment */ 78 kth->ktr_flags = (p->p_nthreads > 1) ? KTRH_THREADED : 0; 79 microtime(&kth->ktr_time); 80 kth->ktr_pid = p->p_pid; 81 kth->ktr_tid = lp->lwp_tid; 82 bcopy(p->p_comm, kth->ktr_comm, MAXCOMLEN + 1); 83 return (kth); 84 } 85 86 void 87 ktrsyscall(struct lwp *lp, int code, int narg, register_t args[]) 88 { 89 struct ktr_header *kth; 90 struct ktr_syscall *ktp; 91 int len; 92 register_t *argp; 93 int i; 94 95 len = offsetof(struct ktr_syscall, ktr_args) + 96 (narg * sizeof(register_t)); 97 98 /* 99 * Setting the active bit prevents a ktrace recursion from the 100 * ktracing op itself. 101 */ 102 lp->lwp_traceflag |= KTRFAC_ACTIVE; 103 kth = ktrgetheader(KTR_SYSCALL); 104 MALLOC(ktp, struct ktr_syscall *, len, M_KTRACE, M_WAITOK); 105 ktp->ktr_code = code; 106 ktp->ktr_narg = narg; 107 argp = &ktp->ktr_args[0]; 108 for (i = 0; i < narg; i++) 109 *argp++ = args[i]; 110 kth->ktr_buf = (caddr_t)ktp; 111 kth->ktr_len = len; 112 ktrwrite(lp, kth, NULL); 113 FREE(ktp, M_KTRACE); 114 FREE(kth, M_KTRACE); 115 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 116 } 117 118 void 119 ktrsysret(struct lwp *lp, int code, int error, register_t retval) 120 { 121 struct ktr_header *kth; 122 struct ktr_sysret ktp; 123 124 lp->lwp_traceflag |= KTRFAC_ACTIVE; 125 kth = ktrgetheader(KTR_SYSRET); 126 ktp.ktr_code = code; 127 ktp.ktr_error = error; 128 ktp.ktr_retval = retval; /* what about val2 ? */ 129 130 kth->ktr_buf = (caddr_t)&ktp; 131 kth->ktr_len = sizeof(struct ktr_sysret); 132 133 ktrwrite(lp, kth, NULL); 134 FREE(kth, M_KTRACE); 135 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 136 } 137 138 void 139 ktrnamei(struct lwp *lp, char *path) 140 { 141 struct ktr_header *kth; 142 143 lp->lwp_traceflag |= KTRFAC_ACTIVE; 144 kth = ktrgetheader(KTR_NAMEI); 145 kth->ktr_len = strlen(path); 146 kth->ktr_buf = path; 147 148 ktrwrite(lp, kth, NULL); 149 FREE(kth, M_KTRACE); 150 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 151 } 152 153 void 154 ktrgenio(struct lwp *lp, int fd, enum uio_rw rw, struct uio *uio, int error) 155 { 156 struct ktr_header *kth; 157 struct ktr_genio ktg; 158 159 if (error) 160 return; 161 lp->lwp_traceflag |= KTRFAC_ACTIVE; 162 kth = ktrgetheader(KTR_GENIO); 163 ktg.ktr_fd = fd; 164 ktg.ktr_rw = rw; 165 kth->ktr_buf = (caddr_t)&ktg; 166 kth->ktr_len = sizeof(struct ktr_genio); 167 uio->uio_offset = 0; 168 uio->uio_rw = UIO_WRITE; 169 170 ktrwrite(lp, kth, uio); 171 FREE(kth, M_KTRACE); 172 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 173 } 174 175 void 176 ktrpsig(struct lwp *lp, int sig, sig_t action, sigset_t *mask, int code) 177 { 178 struct ktr_header *kth; 179 struct ktr_psig kp; 180 181 lp->lwp_traceflag |= KTRFAC_ACTIVE; 182 kth = ktrgetheader(KTR_PSIG); 183 kp.signo = (char)sig; 184 kp.action = action; 185 kp.mask = *mask; 186 kp.code = code; 187 kth->ktr_buf = (caddr_t)&kp; 188 kth->ktr_len = sizeof (struct ktr_psig); 189 190 ktrwrite(lp, kth, NULL); 191 FREE(kth, M_KTRACE); 192 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 193 } 194 195 void 196 ktrcsw(struct lwp *lp, int out, int user) 197 { 198 struct ktr_header *kth; 199 struct ktr_csw kc; 200 201 lp->lwp_traceflag |= KTRFAC_ACTIVE; 202 kth = ktrgetheader(KTR_CSW); 203 kc.out = out; 204 kc.user = user; 205 kth->ktr_buf = (caddr_t)&kc; 206 kth->ktr_len = sizeof (struct ktr_csw); 207 208 ktrwrite(lp, kth, NULL); 209 FREE(kth, M_KTRACE); 210 lp->lwp_traceflag &= ~KTRFAC_ACTIVE; 211 } 212 #endif 213 214 /* Interface and common routines */ 215 216 #ifdef KTRACE 217 /* 218 * ktrace system call 219 */ 220 struct ktrace_clear_info { 221 ktrace_node_t tracenode; 222 int rootclear; 223 int error; 224 }; 225 226 static int ktrace_clear_callback(struct proc *p, void *data); 227 228 #endif 229 230 /* 231 * MPALMOSTSAFE 232 */ 233 int 234 sys_ktrace(struct ktrace_args *uap) 235 { 236 #ifdef KTRACE 237 struct ktrace_clear_info info; 238 struct thread *td = curthread; 239 struct proc *curp = td->td_proc; 240 struct proc *p; 241 struct pgrp *pg; 242 int facs = uap->facs & ~KTRFAC_ROOT; 243 int ops = KTROP(uap->ops); 244 int descend = uap->ops & KTRFLAG_DESCEND; 245 int ret = 0; 246 int error = 0; 247 struct nlookupdata nd; 248 ktrace_node_t tracenode = NULL; 249 250 get_mplock(); 251 curp->p_traceflag |= KTRFAC_ACTIVE; 252 if (ops != KTROP_CLEAR) { 253 /* 254 * an operation which requires a file argument. 255 */ 256 error = nlookup_init(&nd, uap->fname, 257 UIO_USERSPACE, NLC_LOCKVP); 258 if (error == 0) 259 error = vn_open(&nd, NULL, FREAD|FWRITE|O_NOFOLLOW, 0); 260 if (error == 0 && nd.nl_open_vp->v_type != VREG) 261 error = EACCES; 262 if (error) { 263 curp->p_traceflag &= ~KTRFAC_ACTIVE; 264 nlookup_done(&nd); 265 goto done; 266 } 267 MALLOC(tracenode, ktrace_node_t, sizeof (struct ktrace_node), 268 M_KTRACE, M_WAITOK | M_ZERO); 269 tracenode->kn_vp = nd.nl_open_vp; 270 tracenode->kn_refs = 1; 271 nd.nl_open_vp = NULL; 272 nlookup_done(&nd); 273 vn_unlock(tracenode->kn_vp); 274 } 275 /* 276 * Clear all uses of the tracefile. Not the most efficient operation 277 * in the world. 278 */ 279 if (ops == KTROP_CLEARFILE) { 280 info.tracenode = tracenode; 281 info.error = 0; 282 info.rootclear = 0; 283 allproc_scan(ktrace_clear_callback, &info); 284 error = info.error; 285 goto done; 286 } 287 /* 288 * need something to (un)trace (XXX - why is this here?) 289 */ 290 if (!facs) { 291 error = EINVAL; 292 goto done; 293 } 294 /* 295 * do it 296 */ 297 if (uap->pid < 0) { 298 /* 299 * by process group 300 */ 301 pg = pgfind(-uap->pid); 302 if (pg == NULL) { 303 error = ESRCH; 304 goto done; 305 } 306 LIST_FOREACH(p, &pg->pg_members, p_pglist) { 307 if (descend) 308 ret |= ktrsetchildren(td, p, ops, facs, tracenode); 309 else 310 ret |= ktrops(td, p, ops, facs, tracenode); 311 } 312 } else { 313 /* 314 * by pid 315 */ 316 p = pfind(uap->pid); 317 if (p == NULL) { 318 error = ESRCH; 319 goto done; 320 } 321 if (descend) 322 ret |= ktrsetchildren(td, p, ops, facs, tracenode); 323 else 324 ret |= ktrops(td, p, ops, facs, tracenode); 325 } 326 if (!ret) 327 error = EPERM; 328 done: 329 if (tracenode) 330 ktrdestroy(&tracenode); 331 curp->p_traceflag &= ~KTRFAC_ACTIVE; 332 rel_mplock(); 333 return (error); 334 #else 335 return ENOSYS; 336 #endif 337 } 338 339 #ifdef KTRACE 340 341 /* 342 * NOTE: NOT MPSAFE (yet) 343 */ 344 static int 345 ktrace_clear_callback(struct proc *p, void *data) 346 { 347 struct ktrace_clear_info *info = data; 348 349 if (p->p_tracenode) { 350 if (info->rootclear) { 351 if (p->p_tracenode == info->tracenode) { 352 ktrdestroy(&p->p_tracenode); 353 p->p_traceflag = 0; 354 } 355 } else { 356 if (p->p_tracenode->kn_vp == info->tracenode->kn_vp) { 357 if (ktrcanset(curthread, p)) { 358 ktrdestroy(&p->p_tracenode); 359 p->p_traceflag = 0; 360 } else { 361 info->error = EPERM; 362 } 363 } 364 } 365 } 366 return(0); 367 } 368 369 #endif 370 371 /* 372 * utrace system call 373 * 374 * MPALMOSTSAFE 375 */ 376 int 377 sys_utrace(struct utrace_args *uap) 378 { 379 #ifdef KTRACE 380 struct ktr_header *kth; 381 struct thread *td = curthread; /* XXX */ 382 caddr_t cp; 383 384 if (!KTRPOINT(td, KTR_USER)) 385 return (0); 386 if (uap->len > KTR_USER_MAXLEN) 387 return (EINVAL); 388 td->td_lwp->lwp_traceflag |= KTRFAC_ACTIVE; 389 kth = ktrgetheader(KTR_USER); 390 MALLOC(cp, caddr_t, uap->len, M_KTRACE, M_WAITOK); 391 if (!copyin(uap->addr, cp, uap->len)) { 392 kth->ktr_buf = cp; 393 kth->ktr_len = uap->len; 394 get_mplock(); 395 ktrwrite(td->td_lwp, kth, NULL); 396 rel_mplock(); 397 } 398 FREE(kth, M_KTRACE); 399 FREE(cp, M_KTRACE); 400 td->td_lwp->lwp_traceflag &= ~KTRFAC_ACTIVE; 401 402 return (0); 403 #else 404 return (ENOSYS); 405 #endif 406 } 407 408 void 409 ktrdestroy(struct ktrace_node **tracenodep) 410 { 411 ktrace_node_t tracenode; 412 413 if ((tracenode = *tracenodep) != NULL) { 414 *tracenodep = NULL; 415 KKASSERT(tracenode->kn_refs > 0); 416 /* XXX not MP safe yet */ 417 --tracenode->kn_refs; 418 if (tracenode->kn_refs == 0) { 419 vn_close(tracenode->kn_vp, FREAD|FWRITE); 420 tracenode->kn_vp = NULL; 421 FREE(tracenode, M_KTRACE); 422 } 423 } 424 } 425 426 /* 427 * This allows a process to inherit a ref on a tracenode and is also used 428 * as a temporary ref to prevent a tracenode from being destroyed out from 429 * under an active operation. 430 */ 431 ktrace_node_t 432 ktrinherit(ktrace_node_t tracenode) 433 { 434 if (tracenode) { 435 KKASSERT(tracenode->kn_refs > 0); 436 ++tracenode->kn_refs; 437 } 438 return(tracenode); 439 } 440 441 #ifdef KTRACE 442 static int 443 ktrops(struct thread *td, struct proc *p, int ops, int facs, 444 ktrace_node_t tracenode) 445 { 446 ktrace_node_t oldnode; 447 448 if (!ktrcanset(td, p)) 449 return (0); 450 if (ops == KTROP_SET) { 451 if ((oldnode = p->p_tracenode) != tracenode) { 452 p->p_tracenode = ktrinherit(tracenode); 453 ktrdestroy(&oldnode); 454 } 455 p->p_traceflag |= facs; 456 if (td->td_ucred->cr_uid == 0) 457 p->p_traceflag |= KTRFAC_ROOT; 458 } else { 459 /* KTROP_CLEAR */ 460 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) { 461 /* no more tracing */ 462 p->p_traceflag = 0; 463 ktrdestroy(&p->p_tracenode); 464 } 465 } 466 467 return (1); 468 } 469 470 static int 471 ktrsetchildren(struct thread *td, struct proc *top, int ops, int facs, 472 ktrace_node_t tracenode) 473 { 474 struct proc *p; 475 int ret = 0; 476 477 p = top; 478 for (;;) { 479 ret |= ktrops(td, p, ops, facs, tracenode); 480 /* 481 * If this process has children, descend to them next, 482 * otherwise do any siblings, and if done with this level, 483 * follow back up the tree (but not past top). 484 */ 485 if (!LIST_EMPTY(&p->p_children)) 486 p = LIST_FIRST(&p->p_children); 487 else for (;;) { 488 if (p == top) 489 return (ret); 490 if (LIST_NEXT(p, p_sibling)) { 491 p = LIST_NEXT(p, p_sibling); 492 break; 493 } 494 p = p->p_pptr; 495 } 496 } 497 /*NOTREACHED*/ 498 } 499 500 static void 501 ktrwrite(struct lwp *lp, struct ktr_header *kth, struct uio *uio) 502 { 503 struct ktrace_clear_info info; 504 struct uio auio; 505 struct iovec aiov[2]; 506 int error; 507 ktrace_node_t tracenode; 508 509 /* 510 * We have to ref our tracenode to prevent it from being ripped out 511 * from under us while we are trying to use it. p_tracenode can 512 * go away at any time if another process gets a write error. 513 * 514 * XXX not MP safe 515 */ 516 if (lp->lwp_proc->p_tracenode == NULL) 517 return; 518 tracenode = ktrinherit(lp->lwp_proc->p_tracenode); 519 auio.uio_iov = &aiov[0]; 520 auio.uio_offset = 0; 521 auio.uio_segflg = UIO_SYSSPACE; 522 auio.uio_rw = UIO_WRITE; 523 aiov[0].iov_base = (caddr_t)kth; 524 aiov[0].iov_len = sizeof(struct ktr_header); 525 auio.uio_resid = sizeof(struct ktr_header); 526 auio.uio_iovcnt = 1; 527 auio.uio_td = curthread; 528 if (kth->ktr_len > 0) { 529 auio.uio_iovcnt++; 530 aiov[1].iov_base = kth->ktr_buf; 531 aiov[1].iov_len = kth->ktr_len; 532 auio.uio_resid += kth->ktr_len; 533 if (uio != NULL) 534 kth->ktr_len += uio->uio_resid; 535 } 536 vn_lock(tracenode->kn_vp, LK_EXCLUSIVE | LK_RETRY); 537 error = VOP_WRITE(tracenode->kn_vp, &auio, 538 IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred); 539 if (error == 0 && uio != NULL) { 540 error = VOP_WRITE(tracenode->kn_vp, uio, 541 IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred); 542 } 543 vn_unlock(tracenode->kn_vp); 544 if (error) { 545 /* 546 * If an error occured, give up tracing on all processes 547 * using this tracenode. This is not MP safe but is 548 * blocking-safe. 549 */ 550 log(LOG_NOTICE, 551 "ktrace write failed, errno %d, tracing stopped\n", error); 552 info.tracenode = tracenode; 553 info.error = 0; 554 info.rootclear = 1; 555 allproc_scan(ktrace_clear_callback, &info); 556 } 557 ktrdestroy(&tracenode); 558 } 559 560 /* 561 * Return true if caller has permission to set the ktracing state 562 * of target. Essentially, the target can't possess any 563 * more permissions than the caller. KTRFAC_ROOT signifies that 564 * root previously set the tracing status on the target process, and 565 * so, only root may further change it. 566 * 567 * TODO: check groups. use caller effective gid. 568 */ 569 static int 570 ktrcanset(struct thread *calltd, struct proc *targetp) 571 { 572 struct ucred *caller = calltd->td_ucred; 573 struct ucred *target = targetp->p_ucred; 574 575 if (!PRISON_CHECK(caller, target)) 576 return (0); 577 if ((caller->cr_uid == target->cr_ruid && 578 target->cr_ruid == target->cr_svuid && 579 caller->cr_rgid == target->cr_rgid && /* XXX */ 580 target->cr_rgid == target->cr_svgid && 581 (targetp->p_traceflag & KTRFAC_ROOT) == 0 && 582 (targetp->p_flag & P_SUGID) == 0) || 583 caller->cr_uid == 0) 584 return (1); 585 586 return (0); 587 } 588 589 #endif /* KTRACE */ 590