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 1.28 2006/08/12 00:26:20 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
  70         MALLOC(kth, struct ktr_header *, sizeof (struct ktr_header),
  71                 M_KTRACE, M_WAITOK);
  72         kth->ktr_type = type;
  73         microtime(&kth->ktr_time);
  74         kth->ktr_pid = p->p_pid;
  75         bcopy(p->p_comm, kth->ktr_comm, MAXCOMLEN + 1);
  76         return (kth);
  77 }
  78
  79 void
  80 ktrsyscall(struct proc *p, int code, int narg, register_t args[])
  81 {
  82         struct  ktr_header *kth;
  83         struct  ktr_syscall *ktp;
  84         int len;
  85         register_t *argp;
  86         int i;
  87
  88         len = offsetof(struct ktr_syscall, ktr_args) +
  89               (narg * sizeof(register_t));
  90
  91         /*
  92          * Setting the active bit prevents a ktrace recursion from the
  93          * ktracing op itself.
  94          */
  95         p->p_traceflag |= KTRFAC_ACTIVE;
  96         kth = ktrgetheader(KTR_SYSCALL);
  97         MALLOC(ktp, struct ktr_syscall *, len, M_KTRACE, M_WAITOK);
  98         ktp->ktr_code = code;
  99         ktp->ktr_narg = narg;
 100         argp = &ktp->ktr_args[0];
 101         for (i = 0; i < narg; i++)
 102                 *argp++ = args[i];
 103         kth->ktr_buf = (caddr_t)ktp;
 104         kth->ktr_len = len;
 105         ktrwrite(p, kth, NULL);
 106         FREE(ktp, M_KTRACE);
 107         FREE(kth, M_KTRACE);
 108         p->p_traceflag &= ~KTRFAC_ACTIVE;
 109 }
 110
 111 void
 112 ktrsysret(struct proc *p, int code, int error, register_t retval)
 113 {
 114         struct ktr_header *kth;
 115         struct ktr_sysret ktp;
 116
 117         p->p_traceflag |= KTRFAC_ACTIVE;
 118         kth = ktrgetheader(KTR_SYSRET);
 119         ktp.ktr_code = code;
 120         ktp.ktr_error = error;
 121         ktp.ktr_retval = retval;                /* what about val2 ? */
 122
 123         kth->ktr_buf = (caddr_t)&ktp;
 124         kth->ktr_len = sizeof(struct ktr_sysret);
 125
 126         ktrwrite(p, kth, NULL);
 127         FREE(kth, M_KTRACE);
 128         p->p_traceflag &= ~KTRFAC_ACTIVE;
 129 }
 130
 131 void
 132 ktrnamei(struct proc *p, char *path)
 133 {
 134         struct ktr_header *kth;
 135
 136         p->p_traceflag |= KTRFAC_ACTIVE;
 137         kth = ktrgetheader(KTR_NAMEI);
 138         kth->ktr_len = strlen(path);
 139         kth->ktr_buf = path;
 140
 141         ktrwrite(p, kth, NULL);
 142         FREE(kth, M_KTRACE);
 143         p->p_traceflag &= ~KTRFAC_ACTIVE;
 144 }
 145
 146 void
 147 ktrgenio(struct proc *p, int fd, enum uio_rw rw, struct uio *uio, int error)
 148 {
 149         struct ktr_header *kth;
 150         struct ktr_genio ktg;
 151
 152         if (error)
 153                 return;
 154         p->p_traceflag |= KTRFAC_ACTIVE;
 155         kth = ktrgetheader(KTR_GENIO);
 156         ktg.ktr_fd = fd;
 157         ktg.ktr_rw = rw;
 158         kth->ktr_buf = (caddr_t)&ktg;
 159         kth->ktr_len = sizeof(struct ktr_genio);
 160         uio->uio_offset = 0;
 161         uio->uio_rw = UIO_WRITE;
 162
 163         ktrwrite(p, kth, uio);
 164         FREE(kth, M_KTRACE);
 165         p->p_traceflag &= ~KTRFAC_ACTIVE;
 166 }
 167
 168 void
 169 ktrpsig(struct proc *p, int sig, sig_t action, sigset_t *mask, int code)
 170 {
 171         struct ktr_header *kth;
 172         struct ktr_psig kp;
 173
 174         p->p_traceflag |= KTRFAC_ACTIVE;
 175         kth = ktrgetheader(KTR_PSIG);
 176         kp.signo = (char)sig;
 177         kp.action = action;
 178         kp.mask = *mask;
 179         kp.code = code;
 180         kth->ktr_buf = (caddr_t)&kp;
 181         kth->ktr_len = sizeof (struct ktr_psig);
 182
 183         ktrwrite(p, kth, NULL);
 184         FREE(kth, M_KTRACE);
 185         p->p_traceflag &= ~KTRFAC_ACTIVE;
 186 }
 187
 188 void
 189 ktrcsw(struct proc *p, int out, int user)
 190 {
 191         struct ktr_header *kth;
 192         struct  ktr_csw kc;
 193
 194         p->p_traceflag |= KTRFAC_ACTIVE;
 195         kth = ktrgetheader(KTR_CSW);
 196         kc.out = out;
 197         kc.user = user;
 198         kth->ktr_buf = (caddr_t)&kc;
 199         kth->ktr_len = sizeof (struct ktr_csw);
 200
 201         ktrwrite(p, kth, NULL);
 202         FREE(kth, M_KTRACE);
 203         p->p_traceflag &= ~KTRFAC_ACTIVE;
 204 }
 205 #endif
 206
 207 /* Interface and common routines */
 208
 209 #ifdef KTRACE
 210 /*
 211  * ktrace system call
 212  */
 213 struct ktrace_clear_info {
 214         ktrace_node_t tracenode;
 215         int rootclear;
 216         int error;
 217 };
 218
 219 static int ktrace_clear_callback(struct proc *p, void *data);
 220
 221 #endif
 222
 223 int
 224 sys_ktrace(struct ktrace_args *uap)
 225 {
 226 #ifdef KTRACE
 227         struct ktrace_clear_info info;
 228         struct thread *td = curthread;
 229         struct proc *curp = td->td_proc;
 230         struct proc *p;
 231         struct pgrp *pg;
 232         int facs = uap->facs & ~KTRFAC_ROOT;
 233         int ops = KTROP(uap->ops);
 234         int descend = uap->ops & KTRFLAG_DESCEND;
 235         int ret = 0;
 236         int error = 0;
 237         struct nlookupdata nd;
 238         ktrace_node_t tracenode = NULL;
 239
 240         curp->p_traceflag |= KTRFAC_ACTIVE;
 241         if (ops != KTROP_CLEAR) {
 242                 /*
 243                  * an operation which requires a file argument.
 244                  */
 245                 error = nlookup_init(&nd, uap->fname,
 246                                         UIO_USERSPACE, NLC_LOCKVP);
 247                 if (error == 0)
 248                         error = vn_open(&nd, NULL, FREAD|FWRITE|O_NOFOLLOW, 0);
 249                 if (error == 0 && nd.nl_open_vp->v_type != VREG)
 250                         error = EACCES;
 251                 if (error) {
 252                         curp->p_traceflag &= ~KTRFAC_ACTIVE;
 253                         nlookup_done(&nd);
 254                         return (error);
 255                 }
 256                 MALLOC(tracenode, ktrace_node_t, sizeof (struct ktrace_node),
 257                        M_KTRACE, M_WAITOK | M_ZERO);
 258                 tracenode->kn_vp = nd.nl_open_vp;
 259                 tracenode->kn_refs = 1;
 260                 nd.nl_open_vp = NULL;
 261                 nlookup_done(&nd);
 262                 vn_unlock(tracenode->kn_vp);
 263         }
 264         /*
 265          * Clear all uses of the tracefile.  Not the most efficient operation
 266          * in the world.
 267          */
 268         if (ops == KTROP_CLEARFILE) {
 269                 info.tracenode = tracenode;
 270                 info.error = 0;
 271                 info.rootclear = 0;
 272                 allproc_scan(ktrace_clear_callback, &info);
 273                 error = info.error;
 274                 goto done;
 275         }
 276         /*
 277          * need something to (un)trace (XXX - why is this here?)
 278          */
 279         if (!facs) {
 280                 error = EINVAL;
 281                 goto done;
 282         }
 283         /*
 284          * do it
 285          */
 286         if (uap->pid < 0) {
 287                 /*
 288                  * by process group
 289                  */
 290                 pg = pgfind(-uap->pid);
 291                 if (pg == NULL) {
 292                         error = ESRCH;
 293                         goto done;
 294                 }
 295                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 296                         if (descend)
 297                                 ret |= ktrsetchildren(curp, p, ops, facs, tracenode);
 298                         else
 299                                 ret |= ktrops(curp, p, ops, facs, tracenode);
 300                 }
 301         } else {
 302                 /*
 303                  * by pid
 304                  */
 305                 p = pfind(uap->pid);
 306                 if (p == NULL) {
 307                         error = ESRCH;
 308                         goto done;
 309                 }
 310                 if (descend)
 311                         ret |= ktrsetchildren(curp, p, ops, facs, tracenode);
 312                 else
 313                         ret |= ktrops(curp, p, ops, facs, tracenode);
 314         }
 315         if (!ret)
 316                 error = EPERM;
 317 done:
 318         if (tracenode)
 319                 ktrdestroy(&tracenode);
 320         curp->p_traceflag &= ~KTRFAC_ACTIVE;
 321         return (error);
 322 #else
 323         return ENOSYS;
 324 #endif
 325 }
 326
 327 #ifdef KTRACE
 328
 329 /*
 330  * NOTE: NOT MPSAFE (yet)
 331  */
 332 static int
 333 ktrace_clear_callback(struct proc *p, void *data)
 334 {
 335         struct ktrace_clear_info *info = data;
 336
 337         if (info->rootclear) {
 338                 if (p->p_tracenode == info->tracenode) {
 339                         ktrdestroy(&p->p_tracenode);
 340                         p->p_traceflag = 0;
 341                 }
 342         } else {
 343                 if (p->p_tracenode->kn_vp == info->tracenode->kn_vp) {
 344                         if (ktrcanset(curproc, p)) {
 345                                 ktrdestroy(&p->p_tracenode);
 346                                 p->p_traceflag = 0;
 347                         } else {
 348                                 info->error = EPERM;
 349                         }
 350                 }
 351         }
 352         return(0);
 353 }
 354
 355 #endif
 356
 357 /*
 358  * utrace system call
 359  */
 360 /* ARGSUSED */
 361 int
 362 sys_utrace(struct utrace_args *uap)
 363 {
 364 #ifdef KTRACE
 365         struct ktr_header *kth;
 366         struct thread *td = curthread;  /* XXX */
 367         struct proc *p = td->td_proc;
 368         caddr_t cp;
 369
 370         if (!KTRPOINT(td, KTR_USER))
 371                 return (0);
 372         if (uap->len > KTR_USER_MAXLEN)
 373                 return (EINVAL);
 374         p->p_traceflag |= KTRFAC_ACTIVE;
 375         kth = ktrgetheader(KTR_USER);
 376         MALLOC(cp, caddr_t, uap->len, M_KTRACE, M_WAITOK);
 377         if (!copyin(uap->addr, cp, uap->len)) {
 378                 kth->ktr_buf = cp;
 379                 kth->ktr_len = uap->len;
 380                 ktrwrite(p, kth, NULL);
 381         }
 382         FREE(kth, M_KTRACE);
 383         FREE(cp, M_KTRACE);
 384         p->p_traceflag &= ~KTRFAC_ACTIVE;
 385
 386         return (0);
 387 #else
 388         return (ENOSYS);
 389 #endif
 390 }
 391
 392 void
 393 ktrdestroy(struct ktrace_node **tracenodep)
 394 {
 395         ktrace_node_t tracenode;
 396
 397         if ((tracenode = *tracenodep) != NULL) {
 398                 *tracenodep = NULL;
 399                 KKASSERT(tracenode->kn_refs > 0);
 400                 /* XXX not MP safe yet */
 401                 --tracenode->kn_refs;
 402                 if (tracenode->kn_refs == 0) {
 403                         vn_close(tracenode->kn_vp, FREAD|FWRITE);
 404                         tracenode->kn_vp = NULL;
 405                         FREE(tracenode, M_KTRACE);
 406                 }
 407         }
 408 }
 409
 410 /*
 411  * This allows a process to inherit a ref on a tracenode and is also used
 412  * as a temporary ref to prevent a tracenode from being destroyed out from
 413  * under an active operation.
 414  */
 415 ktrace_node_t
 416 ktrinherit(ktrace_node_t tracenode)
 417 {
 418         if (tracenode) {
 419                 KKASSERT(tracenode->kn_refs > 0);
 420                 ++tracenode->kn_refs;
 421         }
 422         return(tracenode);
 423 }
 424
 425 #ifdef KTRACE
 426 static int
 427 ktrops(struct proc *curp, struct proc *p, int ops, int facs,
 428        ktrace_node_t tracenode)
 429 {
 430         ktrace_node_t oldnode;
 431
 432         if (!ktrcanset(curp, p))
 433                 return (0);
 434         if (ops == KTROP_SET) {
 435                 if ((oldnode = p->p_tracenode) != tracenode) {
 436                         p->p_tracenode = ktrinherit(tracenode);
 437                         ktrdestroy(&oldnode);
 438                 }
 439                 p->p_traceflag |= facs;
 440                 if (curp->p_ucred->cr_uid == 0)
 441                         p->p_traceflag |= KTRFAC_ROOT;
 442         } else {
 443                 /* KTROP_CLEAR */
 444                 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
 445                         /* no more tracing */
 446                         p->p_traceflag = 0;
 447                         ktrdestroy(&p->p_tracenode);
 448                 }
 449         }
 450
 451         return (1);
 452 }
 453
 454 static int
 455 ktrsetchildren(struct proc *curp, struct proc *top, int ops, int facs,
 456                ktrace_node_t tracenode)
 457 {
 458         struct proc *p;
 459         int ret = 0;
 460
 461         p = top;
 462         for (;;) {
 463                 ret |= ktrops(curp, p, ops, facs, tracenode);
 464                 /*
 465                  * If this process has children, descend to them next,
 466                  * otherwise do any siblings, and if done with this level,
 467                  * follow back up the tree (but not past top).
 468                  */
 469                 if (!LIST_EMPTY(&p->p_children))
 470                         p = LIST_FIRST(&p->p_children);
 471                 else for (;;) {
 472                         if (p == top)
 473                                 return (ret);
 474                         if (LIST_NEXT(p, p_sibling)) {
 475                                 p = LIST_NEXT(p, p_sibling);
 476                                 break;
 477                         }
 478                         p = p->p_pptr;
 479                 }
 480         }
 481         /*NOTREACHED*/
 482 }
 483
 484 static void
 485 ktrwrite(struct proc *p, struct ktr_header *kth, struct uio *uio)
 486 {
 487         struct ktrace_clear_info info;
 488         struct uio auio;
 489         struct iovec aiov[2];
 490         int error;
 491         ktrace_node_t tracenode;
 492
 493         /*
 494          * We have to ref our tracenode to prevent it from being ripped out
 495          * from under us while we are trying to use it.   p_tracenode can
 496          * go away at any time if another process gets a write error.
 497          *
 498          * XXX not MP safe
 499          */
 500         if (p->p_tracenode == NULL)
 501                 return;
 502         tracenode = ktrinherit(p->p_tracenode);
 503         auio.uio_iov = &aiov[0];
 504         auio.uio_offset = 0;
 505         auio.uio_segflg = UIO_SYSSPACE;
 506         auio.uio_rw = UIO_WRITE;
 507         aiov[0].iov_base = (caddr_t)kth;
 508         aiov[0].iov_len = sizeof(struct ktr_header);
 509         auio.uio_resid = sizeof(struct ktr_header);
 510         auio.uio_iovcnt = 1;
 511         auio.uio_td = curthread;
 512         if (kth->ktr_len > 0) {
 513                 auio.uio_iovcnt++;
 514                 aiov[1].iov_base = kth->ktr_buf;
 515                 aiov[1].iov_len = kth->ktr_len;
 516                 auio.uio_resid += kth->ktr_len;
 517                 if (uio != NULL)
 518                         kth->ktr_len += uio->uio_resid;
 519         }
 520         vn_lock(tracenode->kn_vp, LK_EXCLUSIVE | LK_RETRY);
 521         error = VOP_WRITE(tracenode->kn_vp, &auio,
 522                           IO_UNIT | IO_APPEND, p->p_ucred);
 523         if (error == 0 && uio != NULL) {
 524                 error = VOP_WRITE(tracenode->kn_vp, uio,
 525                                   IO_UNIT | IO_APPEND, p->p_ucred);
 526         }
 527         vn_unlock(tracenode->kn_vp);
 528         if (error) {
 529                 /*
 530                  * If an error occured, give up tracing on all processes
 531                  * using this tracenode.  This is not MP safe but is
 532                  * blocking-safe.
 533                  */
 534                 log(LOG_NOTICE,
 535                     "ktrace write failed, errno %d, tracing stopped\n", error);
 536                 info.tracenode = tracenode;
 537                 info.error = 0;
 538                 info.rootclear = 1;
 539                 allproc_scan(ktrace_clear_callback, &info);
 540         }
 541         ktrdestroy(&tracenode);
 542 }
 543
 544 /*
 545  * Return true if caller has permission to set the ktracing state
 546  * of target.  Essentially, the target can't possess any
 547  * more permissions than the caller.  KTRFAC_ROOT signifies that
 548  * root previously set the tracing status on the target process, and
 549  * so, only root may further change it.
 550  *
 551  * TODO: check groups.  use caller effective gid.
 552  */
 553 static int
 554 ktrcanset(struct proc *callp, struct proc *targetp)
 555 {
 556         struct ucred *caller = callp->p_ucred;
 557         struct ucred *target = targetp->p_ucred;
 558
 559         if (!PRISON_CHECK(caller, target))
 560                 return (0);
 561         if ((caller->cr_uid == target->cr_ruid &&
 562              target->cr_ruid == target->cr_svuid &&
 563              caller->cr_rgid == target->cr_rgid &&      /* XXX */
 564              target->cr_rgid == target->cr_svgid &&
 565              (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
 566              (targetp->p_flag & P_SUGID) == 0) ||
 567              caller->cr_uid == 0)
 568                 return (1);
 569
 570         return (0);
 571 }
 572
 573 #endif /* KTRACE */