1 /* $OpenBSD: kern_ktrace.c,v 1.104 2020/09/13 09:48:39 claudio Exp $ */ 2 /* $NetBSD: kern_ktrace.c,v 1.23 1996/02/09 18:59:36 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1989, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)kern_ktrace.c 8.2 (Berkeley) 9/23/93 33 */ 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/proc.h> 38 #include <sys/sched.h> 39 #include <sys/fcntl.h> 40 #include <sys/namei.h> 41 #include <sys/vnode.h> 42 #include <sys/lock.h> 43 #include <sys/ktrace.h> 44 #include <sys/malloc.h> 45 #include <sys/syslog.h> 46 #include <sys/sysctl.h> 47 #include <sys/pledge.h> 48 49 #include <sys/mount.h> 50 #include <sys/syscall.h> 51 #include <sys/syscallargs.h> 52 53 #include <uvm/uvm_extern.h> 54 55 void ktrinitheaderraw(struct ktr_header *, uint, pid_t, pid_t); 56 void ktrinitheader(struct ktr_header *, struct proc *, int); 57 int ktrstart(struct proc *, struct vnode *, struct ucred *); 58 int ktrops(struct proc *, struct process *, int, int, struct vnode *, 59 struct ucred *); 60 int ktrsetchildren(struct proc *, struct process *, int, int, 61 struct vnode *, struct ucred *); 62 int ktrwrite(struct proc *, struct ktr_header *, const void *, size_t); 63 int ktrwrite2(struct proc *, struct ktr_header *, const void *, size_t, 64 const void *, size_t); 65 int ktrwriteraw(struct proc *, struct vnode *, struct ucred *, 66 struct ktr_header *, struct iovec *); 67 int ktrcanset(struct proc *, struct process *); 68 69 /* 70 * Clear the trace settings in a correct way (to avoid races). 71 */ 72 void 73 ktrcleartrace(struct process *pr) 74 { 75 struct vnode *vp; 76 struct ucred *cred; 77 78 if (pr->ps_tracevp != NULL) { 79 vp = pr->ps_tracevp; 80 cred = pr->ps_tracecred; 81 82 pr->ps_traceflag = 0; 83 pr->ps_tracevp = NULL; 84 pr->ps_tracecred = NULL; 85 86 vp->v_writecount--; 87 vrele(vp); 88 crfree(cred); 89 } 90 } 91 92 /* 93 * Change the trace setting in a correct way (to avoid races). 94 */ 95 void 96 ktrsettrace(struct process *pr, int facs, struct vnode *newvp, 97 struct ucred *newcred) 98 { 99 struct vnode *oldvp; 100 struct ucred *oldcred; 101 102 KASSERT(newvp != NULL); 103 KASSERT(newcred != NULL); 104 105 pr->ps_traceflag |= facs; 106 107 /* nothing to change about where the trace goes? */ 108 if (pr->ps_tracevp == newvp && pr->ps_tracecred == newcred) 109 return; 110 111 vref(newvp); 112 crhold(newcred); 113 newvp->v_writecount++; 114 115 oldvp = pr->ps_tracevp; 116 oldcred = pr->ps_tracecred; 117 118 pr->ps_tracevp = newvp; 119 pr->ps_tracecred = newcred; 120 121 if (oldvp != NULL) { 122 oldvp->v_writecount--; 123 vrele(oldvp); 124 crfree(oldcred); 125 } 126 } 127 128 void 129 ktrinitheaderraw(struct ktr_header *kth, uint type, pid_t pid, pid_t tid) 130 { 131 memset(kth, 0, sizeof(struct ktr_header)); 132 kth->ktr_type = type; 133 nanotime(&kth->ktr_time); 134 kth->ktr_pid = pid; 135 kth->ktr_tid = tid; 136 } 137 138 void 139 ktrinitheader(struct ktr_header *kth, struct proc *p, int type) 140 { 141 struct process *pr = p->p_p; 142 143 ktrinitheaderraw(kth, type, pr->ps_pid, p->p_tid + THREAD_PID_OFFSET); 144 memcpy(kth->ktr_comm, pr->ps_comm, MAXCOMLEN); 145 } 146 147 int 148 ktrstart(struct proc *p, struct vnode *vp, struct ucred *cred) 149 { 150 struct ktr_header kth; 151 152 ktrinitheaderraw(&kth, htobe32(KTR_START), -1, -1); 153 return (ktrwriteraw(p, vp, cred, &kth, NULL)); 154 } 155 156 void 157 ktrsyscall(struct proc *p, register_t code, size_t argsize, register_t args[]) 158 { 159 struct ktr_header kth; 160 struct ktr_syscall *ktp; 161 size_t len = sizeof(struct ktr_syscall) + argsize; 162 register_t *argp; 163 u_int nargs = 0; 164 int i; 165 166 if (code == SYS_sysctl) { 167 /* 168 * The sysctl encoding stores the mib[] 169 * array because it is interesting. 170 */ 171 if (args[1] > 0) 172 nargs = lmin(args[1], CTL_MAXNAME); 173 len += nargs * sizeof(int); 174 } 175 atomic_setbits_int(&p->p_flag, P_INKTR); 176 ktrinitheader(&kth, p, KTR_SYSCALL); 177 ktp = malloc(len, M_TEMP, M_WAITOK); 178 ktp->ktr_code = code; 179 ktp->ktr_argsize = argsize; 180 argp = (register_t *)((char *)ktp + sizeof(struct ktr_syscall)); 181 for (i = 0; i < (argsize / sizeof *argp); i++) 182 *argp++ = args[i]; 183 if (nargs && copyin((void *)args[0], argp, nargs * sizeof(int))) 184 memset(argp, 0, nargs * sizeof(int)); 185 ktrwrite(p, &kth, ktp, len); 186 free(ktp, M_TEMP, len); 187 atomic_clearbits_int(&p->p_flag, P_INKTR); 188 } 189 190 void 191 ktrsysret(struct proc *p, register_t code, int error, 192 const register_t retval[2]) 193 { 194 struct ktr_header kth; 195 struct ktr_sysret ktp; 196 int len; 197 198 atomic_setbits_int(&p->p_flag, P_INKTR); 199 ktrinitheader(&kth, p, KTR_SYSRET); 200 ktp.ktr_code = code; 201 ktp.ktr_error = error; 202 if (error) 203 len = 0; 204 else if (code == SYS_lseek) 205 /* the one exception: lseek on ILP32 needs more */ 206 len = sizeof(long long); 207 else 208 len = sizeof(register_t); 209 ktrwrite2(p, &kth, &ktp, sizeof(ktp), retval, len); 210 atomic_clearbits_int(&p->p_flag, P_INKTR); 211 } 212 213 void 214 ktrnamei(struct proc *p, char *path) 215 { 216 struct ktr_header kth; 217 218 atomic_setbits_int(&p->p_flag, P_INKTR); 219 ktrinitheader(&kth, p, KTR_NAMEI); 220 ktrwrite(p, &kth, path, strlen(path)); 221 atomic_clearbits_int(&p->p_flag, P_INKTR); 222 } 223 224 void 225 ktrgenio(struct proc *p, int fd, enum uio_rw rw, struct iovec *iov, 226 ssize_t len) 227 { 228 struct ktr_header kth; 229 struct ktr_genio ktp; 230 caddr_t cp; 231 int count, error; 232 int buflen; 233 234 atomic_setbits_int(&p->p_flag, P_INKTR); 235 236 /* beware overflow */ 237 if (len > PAGE_SIZE) 238 buflen = PAGE_SIZE; 239 else 240 buflen = len + sizeof(struct ktr_genio); 241 242 ktrinitheader(&kth, p, KTR_GENIO); 243 ktp.ktr_fd = fd; 244 ktp.ktr_rw = rw; 245 246 cp = malloc(buflen, M_TEMP, M_WAITOK); 247 while (len > 0) { 248 /* 249 * Don't allow this process to hog the cpu when doing 250 * huge I/O. 251 */ 252 sched_pause(preempt); 253 254 count = lmin(iov->iov_len, buflen); 255 if (count > len) 256 count = len; 257 if (copyin(iov->iov_base, cp, count)) 258 break; 259 260 KERNEL_LOCK(); 261 error = ktrwrite2(p, &kth, &ktp, sizeof(ktp), cp, count); 262 KERNEL_UNLOCK(); 263 if (error != 0) 264 break; 265 266 iov->iov_len -= count; 267 iov->iov_base = (caddr_t)iov->iov_base + count; 268 269 if (iov->iov_len == 0) 270 iov++; 271 272 len -= count; 273 } 274 275 free(cp, M_TEMP, buflen); 276 atomic_clearbits_int(&p->p_flag, P_INKTR); 277 } 278 279 void 280 ktrpsig(struct proc *p, int sig, sig_t action, int mask, int code, 281 siginfo_t *si) 282 { 283 struct ktr_header kth; 284 struct ktr_psig kp; 285 286 atomic_setbits_int(&p->p_flag, P_INKTR); 287 ktrinitheader(&kth, p, KTR_PSIG); 288 kp.signo = (char)sig; 289 kp.action = action; 290 kp.mask = mask; 291 kp.code = code; 292 kp.si = *si; 293 294 KERNEL_LOCK(); 295 ktrwrite(p, &kth, &kp, sizeof(kp)); 296 KERNEL_UNLOCK(); 297 atomic_clearbits_int(&p->p_flag, P_INKTR); 298 } 299 300 void 301 ktrstruct(struct proc *p, const char *name, const void *data, size_t datalen) 302 { 303 struct ktr_header kth; 304 305 atomic_setbits_int(&p->p_flag, P_INKTR); 306 ktrinitheader(&kth, p, KTR_STRUCT); 307 308 if (data == NULL) 309 datalen = 0; 310 KERNEL_LOCK(); 311 ktrwrite2(p, &kth, name, strlen(name) + 1, data, datalen); 312 KERNEL_UNLOCK(); 313 atomic_clearbits_int(&p->p_flag, P_INKTR); 314 } 315 316 int 317 ktruser(struct proc *p, const char *id, const void *addr, size_t len) 318 { 319 struct ktr_header kth; 320 struct ktr_user ktp; 321 int error; 322 void *memp; 323 #define STK_PARAMS 128 324 long long stkbuf[STK_PARAMS / sizeof(long long)]; 325 326 if (!KTRPOINT(p, KTR_USER)) 327 return (0); 328 if (len > KTR_USER_MAXLEN) 329 return (EINVAL); 330 331 atomic_setbits_int(&p->p_flag, P_INKTR); 332 ktrinitheader(&kth, p, KTR_USER); 333 memset(ktp.ktr_id, 0, KTR_USER_MAXIDLEN); 334 error = copyinstr(id, ktp.ktr_id, KTR_USER_MAXIDLEN, NULL); 335 if (error == 0) { 336 if (len > sizeof(stkbuf)) 337 memp = malloc(len, M_TEMP, M_WAITOK); 338 else 339 memp = stkbuf; 340 error = copyin(addr, memp, len); 341 if (error == 0) 342 ktrwrite2(p, &kth, &ktp, sizeof(ktp), memp, len); 343 if (memp != stkbuf) 344 free(memp, M_TEMP, len); 345 } 346 atomic_clearbits_int(&p->p_flag, P_INKTR); 347 return (error); 348 } 349 350 void 351 ktrexec(struct proc *p, int type, const char *data, ssize_t len) 352 { 353 struct ktr_header kth; 354 int count; 355 int buflen; 356 357 assert(type == KTR_EXECARGS || type == KTR_EXECENV); 358 atomic_setbits_int(&p->p_flag, P_INKTR); 359 360 /* beware overflow */ 361 if (len > PAGE_SIZE) 362 buflen = PAGE_SIZE; 363 else 364 buflen = len; 365 366 ktrinitheader(&kth, p, type); 367 368 while (len > 0) { 369 /* 370 * Don't allow this process to hog the cpu when doing 371 * huge I/O. 372 */ 373 sched_pause(preempt); 374 375 count = lmin(len, buflen); 376 if (ktrwrite(p, &kth, data, count) != 0) 377 break; 378 379 len -= count; 380 data += count; 381 } 382 383 atomic_clearbits_int(&p->p_flag, P_INKTR); 384 } 385 386 void 387 ktrpledge(struct proc *p, int error, uint64_t code, int syscall) 388 { 389 struct ktr_header kth; 390 struct ktr_pledge kp; 391 392 atomic_setbits_int(&p->p_flag, P_INKTR); 393 ktrinitheader(&kth, p, KTR_PLEDGE); 394 kp.error = error; 395 kp.code = code; 396 kp.syscall = syscall; 397 398 KERNEL_LOCK(); 399 ktrwrite(p, &kth, &kp, sizeof(kp)); 400 KERNEL_UNLOCK(); 401 atomic_clearbits_int(&p->p_flag, P_INKTR); 402 } 403 404 /* Interface and common routines */ 405 406 int 407 doktrace(struct vnode *vp, int ops, int facs, pid_t pid, struct proc *p) 408 { 409 struct process *pr = NULL; 410 struct ucred *cred = NULL; 411 struct pgrp *pg; 412 int descend = ops & KTRFLAG_DESCEND; 413 int ret = 0; 414 int error = 0; 415 416 facs = facs & ~((unsigned)KTRFAC_ROOT); 417 ops = KTROP(ops); 418 419 if (ops != KTROP_CLEAR) { 420 /* 421 * an operation which requires a file argument. 422 */ 423 cred = p->p_ucred; 424 if (!vp) { 425 error = EINVAL; 426 goto done; 427 } 428 if (vp->v_type != VREG) { 429 error = EACCES; 430 goto done; 431 } 432 } 433 /* 434 * Clear all uses of the tracefile 435 */ 436 if (ops == KTROP_CLEARFILE) { 437 LIST_FOREACH(pr, &allprocess, ps_list) { 438 if (pr->ps_tracevp == vp) { 439 if (ktrcanset(p, pr)) 440 ktrcleartrace(pr); 441 else 442 error = EPERM; 443 } 444 } 445 goto done; 446 } 447 /* 448 * need something to (un)trace (XXX - why is this here?) 449 */ 450 if (!facs) { 451 error = EINVAL; 452 goto done; 453 } 454 if (ops == KTROP_SET) { 455 if (suser(p) == 0) 456 facs |= KTRFAC_ROOT; 457 error = ktrstart(p, vp, cred); 458 if (error != 0) 459 goto done; 460 } 461 /* 462 * do it 463 */ 464 if (pid < 0) { 465 /* 466 * by process group 467 */ 468 pg = pgfind(-pid); 469 if (pg == NULL) { 470 error = ESRCH; 471 goto done; 472 } 473 LIST_FOREACH(pr, &pg->pg_members, ps_pglist) { 474 if (descend) 475 ret |= ktrsetchildren(p, pr, ops, facs, vp, 476 cred); 477 else 478 ret |= ktrops(p, pr, ops, facs, vp, cred); 479 } 480 } else { 481 /* 482 * by pid 483 */ 484 pr = prfind(pid); 485 if (pr == NULL) { 486 error = ESRCH; 487 goto done; 488 } 489 if (descend) 490 ret |= ktrsetchildren(p, pr, ops, facs, vp, cred); 491 else 492 ret |= ktrops(p, pr, ops, facs, vp, cred); 493 } 494 if (!ret) 495 error = EPERM; 496 done: 497 return (error); 498 } 499 500 /* 501 * ktrace system call 502 */ 503 int 504 sys_ktrace(struct proc *p, void *v, register_t *retval) 505 { 506 struct sys_ktrace_args /* { 507 syscallarg(const char *) fname; 508 syscallarg(int) ops; 509 syscallarg(int) facs; 510 syscallarg(pid_t) pid; 511 } */ *uap = v; 512 struct vnode *vp = NULL; 513 const char *fname = SCARG(uap, fname); 514 struct ucred *cred = NULL; 515 int error; 516 517 if (fname) { 518 struct nameidata nd; 519 520 cred = p->p_ucred; 521 NDINIT(&nd, 0, 0, UIO_USERSPACE, fname, p); 522 nd.ni_pledge = PLEDGE_CPATH | PLEDGE_WPATH; 523 nd.ni_unveil = UNVEIL_CREATE | UNVEIL_WRITE; 524 if ((error = vn_open(&nd, FWRITE|O_NOFOLLOW, 0)) != 0) 525 return error; 526 vp = nd.ni_vp; 527 528 VOP_UNLOCK(vp); 529 } 530 531 error = doktrace(vp, SCARG(uap, ops), SCARG(uap, facs), 532 SCARG(uap, pid), p); 533 if (vp != NULL) 534 (void)vn_close(vp, FWRITE, cred, p); 535 536 return error; 537 } 538 539 int 540 ktrops(struct proc *curp, struct process *pr, int ops, int facs, 541 struct vnode *vp, struct ucred *cred) 542 { 543 if (!ktrcanset(curp, pr)) 544 return (0); 545 if (ops == KTROP_SET) 546 ktrsettrace(pr, facs, vp, cred); 547 else { 548 /* KTROP_CLEAR */ 549 pr->ps_traceflag &= ~facs; 550 if ((pr->ps_traceflag & KTRFAC_MASK) == 0) { 551 /* cleared all the facility bits, so stop completely */ 552 ktrcleartrace(pr); 553 } 554 } 555 556 return (1); 557 } 558 559 int 560 ktrsetchildren(struct proc *curp, struct process *top, int ops, int facs, 561 struct vnode *vp, struct ucred *cred) 562 { 563 struct process *pr; 564 int ret = 0; 565 566 pr = top; 567 for (;;) { 568 ret |= ktrops(curp, pr, ops, facs, vp, cred); 569 /* 570 * If this process has children, descend to them next, 571 * otherwise do any siblings, and if done with this level, 572 * follow back up the tree (but not past top). 573 */ 574 if (!LIST_EMPTY(&pr->ps_children)) 575 pr = LIST_FIRST(&pr->ps_children); 576 else for (;;) { 577 if (pr == top) 578 return (ret); 579 if (LIST_NEXT(pr, ps_sibling) != NULL) { 580 pr = LIST_NEXT(pr, ps_sibling); 581 break; 582 } 583 pr = pr->ps_pptr; 584 } 585 } 586 /*NOTREACHED*/ 587 } 588 589 int 590 ktrwrite(struct proc *p, struct ktr_header *kth, const void *aux, size_t len) 591 { 592 struct vnode *vp = p->p_p->ps_tracevp; 593 struct ucred *cred = p->p_p->ps_tracecred; 594 struct iovec data[2]; 595 int error; 596 597 if (vp == NULL) 598 return 0; 599 crhold(cred); 600 data[0].iov_base = (void *)aux; 601 data[0].iov_len = len; 602 data[1].iov_len = 0; 603 kth->ktr_len = len; 604 error = ktrwriteraw(p, vp, cred, kth, data); 605 crfree(cred); 606 return (error); 607 } 608 609 int 610 ktrwrite2(struct proc *p, struct ktr_header *kth, const void *aux1, 611 size_t len1, const void *aux2, size_t len2) 612 { 613 struct vnode *vp = p->p_p->ps_tracevp; 614 struct ucred *cred = p->p_p->ps_tracecred; 615 struct iovec data[2]; 616 int error; 617 618 if (vp == NULL) 619 return 0; 620 crhold(cred); 621 data[0].iov_base = (void *)aux1; 622 data[0].iov_len = len1; 623 data[1].iov_base = (void *)aux2; 624 data[1].iov_len = len2; 625 kth->ktr_len = len1 + len2; 626 error = ktrwriteraw(p, vp, cred, kth, data); 627 crfree(cred); 628 return (error); 629 } 630 631 int 632 ktrwriteraw(struct proc *curp, struct vnode *vp, struct ucred *cred, 633 struct ktr_header *kth, struct iovec *data) 634 { 635 struct uio auio; 636 struct iovec aiov[3]; 637 struct process *pr; 638 int error; 639 640 KERNEL_ASSERT_LOCKED(); 641 642 auio.uio_iov = &aiov[0]; 643 auio.uio_offset = 0; 644 auio.uio_segflg = UIO_SYSSPACE; 645 auio.uio_rw = UIO_WRITE; 646 aiov[0].iov_base = (caddr_t)kth; 647 aiov[0].iov_len = sizeof(struct ktr_header); 648 auio.uio_resid = sizeof(struct ktr_header); 649 auio.uio_iovcnt = 1; 650 auio.uio_procp = curp; 651 if (kth->ktr_len > 0) { 652 aiov[1] = data[0]; 653 aiov[2] = data[1]; 654 auio.uio_iovcnt++; 655 if (aiov[2].iov_len > 0) 656 auio.uio_iovcnt++; 657 auio.uio_resid += kth->ktr_len; 658 } 659 error = vget(vp, LK_EXCLUSIVE | LK_RETRY); 660 if (error) 661 goto bad; 662 error = VOP_WRITE(vp, &auio, IO_UNIT|IO_APPEND, cred); 663 vput(vp); 664 if (error) 665 goto bad; 666 667 return (0); 668 669 bad: 670 /* 671 * If error encountered, give up tracing on this vnode. 672 */ 673 log(LOG_NOTICE, "ktrace write failed, errno %d, tracing stopped\n", 674 error); 675 LIST_FOREACH(pr, &allprocess, ps_list) { 676 if (pr == curp->p_p) 677 continue; 678 if (pr->ps_tracevp == vp && pr->ps_tracecred == cred) 679 ktrcleartrace(pr); 680 } 681 ktrcleartrace(curp->p_p); 682 return (error); 683 } 684 685 /* 686 * Return true if caller has permission to set the ktracing state 687 * of target. Essentially, the target can't possess any 688 * more permissions than the caller. KTRFAC_ROOT signifies that 689 * root previously set the tracing status on the target process, and 690 * so, only root may further change it. 691 * 692 * TODO: check groups. use caller effective gid. 693 */ 694 int 695 ktrcanset(struct proc *callp, struct process *targetpr) 696 { 697 struct ucred *caller = callp->p_ucred; 698 struct ucred *target = targetpr->ps_ucred; 699 700 if ((caller->cr_uid == target->cr_ruid && 701 target->cr_ruid == target->cr_svuid && 702 caller->cr_rgid == target->cr_rgid && /* XXX */ 703 target->cr_rgid == target->cr_svgid && 704 (targetpr->ps_traceflag & KTRFAC_ROOT) == 0 && 705 !ISSET(targetpr->ps_flags, PS_SUGID)) || 706 caller->cr_uid == 0) 707 return (1); 708 709 return (0); 710 } 711