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
 303                  */
 304                 pg = pgfind(-uap->pid);
 305                 if (pg == NULL) {
 306                         error = ESRCH;
 307                         goto done;
 308                 }
 309                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 310                         if (descend)
 311                                 ret |= ktrsetchildren(td, p, ops, facs, tracenode);
 312                         else
 313                                 ret |= ktrops(td, p, ops, facs, tracenode);
 314                 }
 315         } else {
 316                 /*
 317                  * by pid
 318                  */
 319                 p = pfind(uap->pid);
 320                 if (p == NULL) {
 321                         error = ESRCH;
 322                         goto done;
 323                 }
 324                 if (descend)
 325                         ret |= ktrsetchildren(td, p, ops, facs, tracenode);
 326                 else
 327                         ret |= ktrops(td, p, ops, facs, tracenode);
 328         }
 329         if (!ret)
 330                 error = EPERM;
 331 done:
 332         if (tracenode)
 333                 ktrdestroy(&tracenode);
 334         curp->p_traceflag &= ~KTRFAC_ACTIVE;
 335         rel_mplock();
 336         return (error);
 337 #else
 338         return ENOSYS;
 339 #endif
 340 }
 341
 342 #ifdef KTRACE
 343
 344 /*
 345  * NOTE: NOT MPSAFE (yet)
 346  */
 347 static int
 348 ktrace_clear_callback(struct proc *p, void *data)
 349 {
 350         struct ktrace_clear_info *info = data;
 351
 352         if (p->p_tracenode) {
 353                 if (info->rootclear) {
 354                         if (p->p_tracenode == info->tracenode) {
 355                                 ktrdestroy(&p->p_tracenode);
 356                                 p->p_traceflag = 0;
 357                         }
 358                 } else {
 359                         if (p->p_tracenode->kn_vp == info->tracenode->kn_vp) {
 360                                 if (ktrcanset(curthread, p)) {
 361                                         ktrdestroy(&p->p_tracenode);
 362                                         p->p_traceflag = 0;
 363                                 } else {
 364                                         info->error = EPERM;
 365                                 }
 366                         }
 367                 }
 368         }
 369         return(0);
 370 }
 371
 372 #endif
 373
 374 /*
 375  * utrace system call
 376  *
 377  * MPALMOSTSAFE
 378  */
 379 int
 380 sys_utrace(struct utrace_args *uap)
 381 {
 382 #ifdef KTRACE
 383         struct ktr_header *kth;
 384         struct thread *td = curthread;  /* XXX */
 385         caddr_t cp;
 386
 387         if (!KTRPOINT(td, KTR_USER))
 388                 return (0);
 389         if (uap->len > KTR_USER_MAXLEN)
 390                 return (EINVAL);
 391         td->td_lwp->lwp_traceflag |= KTRFAC_ACTIVE;
 392         kth = ktrgetheader(KTR_USER);
 393         MALLOC(cp, caddr_t, uap->len, M_KTRACE, M_WAITOK);
 394         if (!copyin(uap->addr, cp, uap->len)) {
 395                 kth->ktr_buf = cp;
 396                 kth->ktr_len = uap->len;
 397                 get_mplock();
 398                 ktrwrite(td->td_lwp, kth, NULL);
 399                 rel_mplock();
 400         }
 401         FREE(kth, M_KTRACE);
 402         FREE(cp, M_KTRACE);
 403         td->td_lwp->lwp_traceflag &= ~KTRFAC_ACTIVE;
 404
 405         return (0);
 406 #else
 407         return (ENOSYS);
 408 #endif
 409 }
 410
 411 void
 412 ktrdestroy(struct ktrace_node **tracenodep)
 413 {
 414         ktrace_node_t tracenode;
 415
 416         if ((tracenode = *tracenodep) != NULL) {
 417                 *tracenodep = NULL;
 418                 KKASSERT(tracenode->kn_refs > 0);
 419                 /* XXX not MP safe yet */
 420                 --tracenode->kn_refs;
 421                 if (tracenode->kn_refs == 0) {
 422                         vn_close(tracenode->kn_vp, FREAD|FWRITE);
 423                         tracenode->kn_vp = NULL;
 424                         FREE(tracenode, M_KTRACE);
 425                 }
 426         }
 427 }
 428
 429 /*
 430  * This allows a process to inherit a ref on a tracenode and is also used
 431  * as a temporary ref to prevent a tracenode from being destroyed out from
 432  * under an active operation.
 433  */
 434 ktrace_node_t
 435 ktrinherit(ktrace_node_t tracenode)
 436 {
 437         if (tracenode) {
 438                 KKASSERT(tracenode->kn_refs > 0);
 439                 ++tracenode->kn_refs;
 440         }
 441         return(tracenode);
 442 }
 443
 444 #ifdef KTRACE
 445 static int
 446 ktrops(struct thread *td, struct proc *p, int ops, int facs,
 447        ktrace_node_t tracenode)
 448 {
 449         ktrace_node_t oldnode;
 450
 451         if (!ktrcanset(td, p))
 452                 return (0);
 453         if (ops == KTROP_SET) {
 454                 if ((oldnode = p->p_tracenode) != tracenode) {
 455                         p->p_tracenode = ktrinherit(tracenode);
 456                         ktrdestroy(&oldnode);
 457                 }
 458                 p->p_traceflag |= facs;
 459                 if (td->td_ucred->cr_uid == 0)
 460                         p->p_traceflag |= KTRFAC_ROOT;
 461         } else {
 462                 /* KTROP_CLEAR */
 463                 if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
 464                         /* no more tracing */
 465                         p->p_traceflag = 0;
 466                         ktrdestroy(&p->p_tracenode);
 467                 }
 468         }
 469
 470         return (1);
 471 }
 472
 473 static int
 474 ktrsetchildren(struct thread *td, struct proc *top, int ops, int facs,
 475                ktrace_node_t tracenode)
 476 {
 477         struct proc *p;
 478         int ret = 0;
 479
 480         p = top;
 481         for (;;) {
 482                 ret |= ktrops(td, p, ops, facs, tracenode);
 483                 /*
 484                  * If this process has children, descend to them next,
 485                  * otherwise do any siblings, and if done with this level,
 486                  * follow back up the tree (but not past top).
 487                  */
 488                 if (!LIST_EMPTY(&p->p_children))
 489                         p = LIST_FIRST(&p->p_children);
 490                 else for (;;) {
 491                         if (p == top)
 492                                 return (ret);
 493                         if (LIST_NEXT(p, p_sibling)) {
 494                                 p = LIST_NEXT(p, p_sibling);
 495                                 break;
 496                         }
 497                         p = p->p_pptr;
 498                 }
 499         }
 500         /*NOTREACHED*/
 501 }
 502
 503 static void
 504 ktrwrite(struct lwp *lp, struct ktr_header *kth, struct uio *uio)
 505 {
 506         struct ktrace_clear_info info;
 507         struct uio auio;
 508         struct iovec aiov[2];
 509         int error;
 510         ktrace_node_t tracenode;
 511
 512         /*
 513          * We have to ref our tracenode to prevent it from being ripped out
 514          * from under us while we are trying to use it.   p_tracenode can
 515          * go away at any time if another process gets a write error.
 516          *
 517          * XXX not MP safe
 518          */
 519         if (lp->lwp_proc->p_tracenode == NULL)
 520                 return;
 521         tracenode = ktrinherit(lp->lwp_proc->p_tracenode);
 522         auio.uio_iov = &aiov[0];
 523         auio.uio_offset = 0;
 524         auio.uio_segflg = UIO_SYSSPACE;
 525         auio.uio_rw = UIO_WRITE;
 526         aiov[0].iov_base = (caddr_t)kth;
 527         aiov[0].iov_len = sizeof(struct ktr_header);
 528         auio.uio_resid = sizeof(struct ktr_header);
 529         auio.uio_iovcnt = 1;
 530         auio.uio_td = curthread;
 531         if (kth->ktr_len > 0) {
 532                 auio.uio_iovcnt++;
 533                 aiov[1].iov_base = kth->ktr_buf;
 534                 aiov[1].iov_len = kth->ktr_len;
 535                 auio.uio_resid += kth->ktr_len;
 536                 if (uio != NULL)
 537                         kth->ktr_len += uio->uio_resid;
 538         }
 539         vn_lock(tracenode->kn_vp, LK_EXCLUSIVE | LK_RETRY);
 540         error = VOP_WRITE(tracenode->kn_vp, &auio,
 541                           IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred);
 542         if (error == 0 && uio != NULL) {
 543                 error = VOP_WRITE(tracenode->kn_vp, uio,
 544                               IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred);
 545         }
 546         vn_unlock(tracenode->kn_vp);
 547         if (error) {
 548                 /*
 549                  * If an error occured, give up tracing on all processes
 550                  * using this tracenode.  This is not MP safe but is
 551                  * blocking-safe.
 552                  */
 553                 log(LOG_NOTICE,
 554                     "ktrace write failed, errno %d, tracing stopped\n", error);
 555                 info.tracenode = tracenode;
 556                 info.error = 0;
 557                 info.rootclear = 1;
 558                 allproc_scan(ktrace_clear_callback, &info);
 559         }
 560         ktrdestroy(&tracenode);
 561 }
 562
 563 /*
 564  * Return true if caller has permission to set the ktracing state
 565  * of target.  Essentially, the target can't possess any
 566  * more permissions than the caller.  KTRFAC_ROOT signifies that
 567  * root previously set the tracing status on the target process, and
 568  * so, only root may further change it.
 569  *
 570  * TODO: check groups.  use caller effective gid.
 571  */
 572 static int
 573 ktrcanset(struct thread *calltd, struct proc *targetp)
 574 {
 575         struct ucred *caller = calltd->td_ucred;
 576         struct ucred *target = targetp->p_ucred;
 577
 578         if (!PRISON_CHECK(caller, target))
 579                 return (0);
 580         if ((caller->cr_uid == target->cr_ruid &&
 581              target->cr_ruid == target->cr_svuid &&
 582              caller->cr_rgid == target->cr_rgid &&      /* XXX */
 583              target->cr_rgid == target->cr_svgid &&
 584              (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
 585              (targetp->p_flag & P_SUGID) == 0) ||
 586              caller->cr_uid == 0)
 587                 return (1);
 588
 589         return (0);
 590 }
 591
 592 #endif /* KTRACE */