sys/kern/kern_ktrace.c

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