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 /*
  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->lwp_proc, 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->lwp_proc, 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->lwp_proc, 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->lwp_proc, 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->lwp_proc, 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->lwp_proc, 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(curp, p, ops, facs, tracenode);
 309                         else
 310                                 ret |= ktrops(curp, 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(curp, p, ops, facs, tracenode);
 323                 else
 324                         ret |= ktrops(curp, 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(curproc, 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         struct proc *p = td->td_proc;
 383         caddr_t cp;
 384
 385         if (!KTRPOINT(td, KTR_USER))
 386                 return (0);
 387         if (uap->len > KTR_USER_MAXLEN)
 388                 return (EINVAL);
 389         td->td_lwp->lwp_traceflag |= KTRFAC_ACTIVE;
 390         kth = ktrgetheader(KTR_USER);
 391         MALLOC(cp, caddr_t, uap->len, M_KTRACE, M_WAITOK);
 392         if (!copyin(uap->addr, cp, uap->len)) {
 393                 kth->ktr_buf = cp;
 394                 kth->ktr_len = uap->len;
 395                 get_mplock();
 396                 ktrwrite(p, kth, NULL);
 397                 rel_mplock();
 398         }
 399         FREE(kth, M_KTRACE);
 400         FREE(cp, M_KTRACE);
 401         td->td_lwp->lwp_traceflag &= ~KTRFAC_ACTIVE;
 402
 403         return (0);
 404 #else
 405         return (ENOSYS);
 406 #endif
 407 }
 408
 409 void
 410 ktrdestroy(struct ktrace_node **tracenodep)
 411 {
 412         ktrace_node_t tracenode;
 413
 414         if ((tracenode = *tracenodep) != NULL) {
 415                 *tracenodep = NULL;
 416                 KKASSERT(tracenode->kn_refs > 0);
 417                 /* XXX not MP safe yet */
 418                 --tracenode->kn_refs;
 419                 if (tracenode->kn_refs == 0) {
 420                         vn_close(tracenode->kn_vp, FREAD|FWRITE);
 421                         tracenode->kn_vp = NULL;
 422                         FREE(tracenode, M_KTRACE);
 423                 }
 424         }
 425 }
 426
 427 /*
 428  * This allows a process to inherit a ref on a tracenode and is also used
 429  * as a temporary ref to prevent a tracenode from being destroyed out from
 430  * under an active operation.
 431  */
 432 ktrace_node_t
 433 ktrinherit(ktrace_node_t tracenode)
 434 {
 435         if (tracenode) {
 436                 KKASSERT(tracenode->kn_refs > 0);
 437                 ++tracenode->kn_refs;
 438         }
 439         return(tracenode);
 440 }
 441
 442 #ifdef KTRACE
 443 static int
 444 ktrops(struct proc *curp, struct proc *p, int ops, int facs,
 445        ktrace_node_t tracenode)
 446 {
 447         ktrace_node_t oldnode;
 448
 449         if (!ktrcanset(curp, p))
 450                 return (0);
 451         if (ops == KTROP_SET) {
 452                 if ((oldnode = p->p_tracenode) != tracenode) {
 453                         p->p_tracenode = ktrinherit(tracenode);
 454                         ktrdestroy(&oldnode);
 455                 }
 456                 p->p_traceflag |= facs;
 457                 if (curp->p_ucred->cr_uid == 0)
 458                         p->p_traceflag |= KTRFAC_ROOT;
 459         } else {
 460                 /* KTROP_CLEAR */
 461                 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
 462                         /* no more tracing */
 463                         p->p_traceflag = 0;
 464                         ktrdestroy(&p->p_tracenode);
 465                 }
 466         }
 467
 468         return (1);
 469 }
 470
 471 static int
 472 ktrsetchildren(struct proc *curp, struct proc *top, int ops, int facs,
 473                ktrace_node_t tracenode)
 474 {
 475         struct proc *p;
 476         int ret = 0;
 477
 478         p = top;
 479         for (;;) {
 480                 ret |= ktrops(curp, p, ops, facs, tracenode);
 481                 /*
 482                  * If this process has children, descend to them next,
 483                  * otherwise do any siblings, and if done with this level,
 484                  * follow back up the tree (but not past top).
 485                  */
 486                 if (!LIST_EMPTY(&p->p_children))
 487                         p = LIST_FIRST(&p->p_children);
 488                 else for (;;) {
 489                         if (p == top)
 490                                 return (ret);
 491                         if (LIST_NEXT(p, p_sibling)) {
 492                                 p = LIST_NEXT(p, p_sibling);
 493                                 break;
 494                         }
 495                         p = p->p_pptr;
 496                 }
 497         }
 498         /*NOTREACHED*/
 499 }
 500
 501 static void
 502 ktrwrite(struct proc *p, struct ktr_header *kth, struct uio *uio)
 503 {
 504         struct ktrace_clear_info info;
 505         struct uio auio;
 506         struct iovec aiov[2];
 507         int error;
 508         ktrace_node_t tracenode;
 509
 510         /*
 511          * We have to ref our tracenode to prevent it from being ripped out
 512          * from under us while we are trying to use it.   p_tracenode can
 513          * go away at any time if another process gets a write error.
 514          *
 515          * XXX not MP safe
 516          */
 517         if (p->p_tracenode == NULL)
 518                 return;
 519         tracenode = ktrinherit(p->p_tracenode);
 520         auio.uio_iov = &aiov[0];
 521         auio.uio_offset = 0;
 522         auio.uio_segflg = UIO_SYSSPACE;
 523         auio.uio_rw = UIO_WRITE;
 524         aiov[0].iov_base = (caddr_t)kth;
 525         aiov[0].iov_len = sizeof(struct ktr_header);
 526         auio.uio_resid = sizeof(struct ktr_header);
 527         auio.uio_iovcnt = 1;
 528         auio.uio_td = curthread;
 529         if (kth->ktr_len > 0) {
 530                 auio.uio_iovcnt++;
 531                 aiov[1].iov_base = kth->ktr_buf;
 532                 aiov[1].iov_len = kth->ktr_len;
 533                 auio.uio_resid += kth->ktr_len;
 534                 if (uio != NULL)
 535                         kth->ktr_len += uio->uio_resid;
 536         }
 537         vn_lock(tracenode->kn_vp, LK_EXCLUSIVE | LK_RETRY);
 538         error = VOP_WRITE(tracenode->kn_vp, &auio,
 539                           IO_UNIT | IO_APPEND, p->p_ucred);
 540         if (error == 0 && uio != NULL) {
 541                 error = VOP_WRITE(tracenode->kn_vp, uio,
 542                                   IO_UNIT | IO_APPEND, p->p_ucred);
 543         }
 544         vn_unlock(tracenode->kn_vp);
 545         if (error) {
 546                 /*
 547                  * If an error occured, give up tracing on all processes
 548                  * using this tracenode.  This is not MP safe but is
 549                  * blocking-safe.
 550                  */
 551                 log(LOG_NOTICE,
 552                     "ktrace write failed, errno %d, tracing stopped\n", error);
 553                 info.tracenode = tracenode;
 554                 info.error = 0;
 555                 info.rootclear = 1;
 556                 allproc_scan(ktrace_clear_callback, &info);
 557         }
 558         ktrdestroy(&tracenode);
 559 }
 560
 561 /*
 562  * Return true if caller has permission to set the ktracing state
 563  * of target.  Essentially, the target can't possess any
 564  * more permissions than the caller.  KTRFAC_ROOT signifies that
 565  * root previously set the tracing status on the target process, and
 566  * so, only root may further change it.
 567  *
 568  * TODO: check groups.  use caller effective gid.
 569  */
 570 static int
 571 ktrcanset(struct proc *callp, struct proc *targetp)
 572 {
 573         struct ucred *caller = callp->p_ucred;
 574         struct ucred *target = targetp->p_ucred;
 575
 576         if (!PRISON_CHECK(caller, target))
 577                 return (0);
 578         if ((caller->cr_uid == target->cr_ruid &&
 579              target->cr_ruid == target->cr_svuid &&
 580              caller->cr_rgid == target->cr_rgid &&      /* XXX */
 581              target->cr_rgid == target->cr_svgid &&
 582              (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
 583              (targetp->p_flag & P_SUGID) == 0) ||
 584              caller->cr_uid == 0)
 585                 return (1);
 586
 587         return (0);
 588 }
 589
 590 #endif /* KTRACE */