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