1 /* 2 * Copyright (c) 1994, Sean Eric Fagan 3 * 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 Sean Eric Fagan. 16 * 4. The name of the author may not be used to endorse or promote products 17 * derived from this software without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * $FreeBSD: src/sys/kern/sys_process.c,v 1.51.2.6 2003/01/08 03:06:45 kan Exp $ 32 * $DragonFly: src/sys/kern/sys_process.c,v 1.26 2006/12/28 21:24:01 dillon Exp $ 33 */ 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/sysproto.h> 38 #include <sys/proc.h> 39 #include <sys/vnode.h> 40 #include <sys/ptrace.h> 41 #include <sys/reg.h> 42 #include <sys/lock.h> 43 44 #include <vm/vm.h> 45 #include <vm/pmap.h> 46 #include <vm/vm_map.h> 47 #include <vm/vm_page.h> 48 49 #include <sys/user.h> 50 #include <vfs/procfs/procfs.h> 51 #include <sys/thread2.h> 52 53 /* use the equivalent procfs code */ 54 #if 0 55 static int 56 pread (struct proc *procp, unsigned int addr, unsigned int *retval) { 57 int rv; 58 vm_map_t map, tmap; 59 vm_object_t object; 60 vm_offset_t kva = 0; 61 int page_offset; /* offset into page */ 62 vm_offset_t pageno; /* page number */ 63 vm_map_entry_t out_entry; 64 vm_prot_t out_prot; 65 boolean_t wired; 66 vm_pindex_t pindex; 67 68 /* Map page into kernel space */ 69 70 map = &procp->p_vmspace->vm_map; 71 72 page_offset = addr - trunc_page(addr); 73 pageno = trunc_page(addr); 74 75 tmap = map; 76 rv = vm_map_lookup (&tmap, pageno, VM_PROT_READ, &out_entry, 77 &object, &pindex, &out_prot, &wired); 78 79 if (rv != KERN_SUCCESS) 80 return EINVAL; 81 82 vm_map_lookup_done (tmap, out_entry, 0); 83 84 /* Find space in kernel_map for the page we're interested in */ 85 rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex), 86 &kva, PAGE_SIZE, 87 0, 88 VM_MAPTYPE_NORMAL, 89 VM_PROT_ALL, VM_PROT_ALL, 90 0); 91 92 if (!rv) { 93 vm_object_reference (object); 94 95 rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0); 96 if (!rv) { 97 *retval = 0; 98 bcopy ((caddr_t)kva + page_offset, 99 retval, sizeof *retval); 100 } 101 vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE); 102 } 103 104 return rv; 105 } 106 107 static int 108 pwrite (struct proc *procp, unsigned int addr, unsigned int datum) { 109 int rv; 110 vm_map_t map, tmap; 111 vm_object_t object; 112 vm_offset_t kva = 0; 113 int page_offset; /* offset into page */ 114 vm_offset_t pageno; /* page number */ 115 vm_map_entry_t out_entry; 116 vm_prot_t out_prot; 117 boolean_t wired; 118 vm_pindex_t pindex; 119 boolean_t fix_prot = 0; 120 121 /* Map page into kernel space */ 122 123 map = &procp->p_vmspace->vm_map; 124 125 page_offset = addr - trunc_page(addr); 126 pageno = trunc_page(addr); 127 128 /* 129 * Check the permissions for the area we're interested in. 130 */ 131 132 if (vm_map_check_protection (map, pageno, pageno + PAGE_SIZE, 133 VM_PROT_WRITE) == FALSE) { 134 /* 135 * If the page was not writable, we make it so. 136 * XXX It is possible a page may *not* be read/executable, 137 * if a process changes that! 138 */ 139 fix_prot = 1; 140 /* The page isn't writable, so let's try making it so... */ 141 if ((rv = vm_map_protect (map, pageno, pageno + PAGE_SIZE, 142 VM_PROT_ALL, 0)) != KERN_SUCCESS) 143 return EFAULT; /* I guess... */ 144 } 145 146 /* 147 * Now we need to get the page. out_entry, out_prot, wired, and 148 * single_use aren't used. One would think the vm code would be 149 * a *bit* nicer... We use tmap because vm_map_lookup() can 150 * change the map argument. 151 */ 152 153 tmap = map; 154 rv = vm_map_lookup (&tmap, pageno, VM_PROT_WRITE, &out_entry, 155 &object, &pindex, &out_prot, &wired); 156 if (rv != KERN_SUCCESS) { 157 return EINVAL; 158 } 159 160 /* 161 * Okay, we've got the page. Let's release tmap. 162 */ 163 164 vm_map_lookup_done (tmap, out_entry, 0); 165 166 /* 167 * Fault the page in... 168 */ 169 170 rv = vm_fault(map, pageno, VM_PROT_WRITE|VM_PROT_READ, FALSE); 171 if (rv != KERN_SUCCESS) 172 return EFAULT; 173 174 /* Find space in kernel_map for the page we're interested in */ 175 rv = vm_map_find (&kernel_map, object, IDX_TO_OFF(pindex), 176 &kva, PAGE_SIZE, 177 0, 178 VM_MAPTYPE_NORMAL, 179 VM_PROT_ALL, VM_PROT_ALL, 180 0); 181 if (!rv) { 182 vm_object_reference (object); 183 184 rv = vm_map_wire (&kernel_map, kva, kva + PAGE_SIZE, 0); 185 if (!rv) { 186 bcopy (&datum, (caddr_t)kva + page_offset, sizeof datum); 187 } 188 vm_map_remove (&kernel_map, kva, kva + PAGE_SIZE); 189 } 190 191 if (fix_prot) 192 vm_map_protect (map, pageno, pageno + PAGE_SIZE, 193 VM_PROT_READ|VM_PROT_EXECUTE, 0); 194 return rv; 195 } 196 #endif 197 198 /* 199 * Process debugging system call. 200 */ 201 int 202 sys_ptrace(struct ptrace_args *uap) 203 { 204 struct proc *p = curproc; 205 206 /* 207 * XXX this obfuscation is to reduce stack usage, but the register 208 * structs may be too large to put on the stack anyway. 209 */ 210 union { 211 struct ptrace_io_desc piod; 212 struct dbreg dbreg; 213 struct fpreg fpreg; 214 struct reg reg; 215 } r; 216 void *addr; 217 int error = 0; 218 219 addr = &r; 220 switch (uap->req) { 221 case PT_GETREGS: 222 case PT_GETFPREGS: 223 #ifdef PT_GETDBREGS 224 case PT_GETDBREGS: 225 #endif 226 break; 227 case PT_SETREGS: 228 error = copyin(uap->addr, &r.reg, sizeof r.reg); 229 break; 230 case PT_SETFPREGS: 231 error = copyin(uap->addr, &r.fpreg, sizeof r.fpreg); 232 break; 233 #ifdef PT_SETDBREGS 234 case PT_SETDBREGS: 235 error = copyin(uap->addr, &r.dbreg, sizeof r.dbreg); 236 break; 237 #endif 238 case PT_IO: 239 error = copyin(uap->addr, &r.piod, sizeof r.piod); 240 break; 241 default: 242 addr = uap->addr; 243 } 244 if (error) 245 return (error); 246 247 error = kern_ptrace(p, uap->req, uap->pid, addr, uap->data, 248 &uap->sysmsg_result); 249 if (error) 250 return (error); 251 252 switch (uap->req) { 253 case PT_IO: 254 (void)copyout(&r.piod, uap->addr, sizeof r.piod); 255 break; 256 case PT_GETREGS: 257 error = copyout(&r.reg, uap->addr, sizeof r.reg); 258 break; 259 case PT_GETFPREGS: 260 error = copyout(&r.fpreg, uap->addr, sizeof r.fpreg); 261 break; 262 #ifdef PT_GETDBREGS 263 case PT_GETDBREGS: 264 error = copyout(&r.dbreg, uap->addr, sizeof r.dbreg); 265 break; 266 #endif 267 } 268 269 return (error); 270 } 271 272 int 273 kern_ptrace(struct proc *curp, int req, pid_t pid, void *addr, int data, int *res) 274 { 275 struct proc *p, *pp; 276 struct iovec iov; 277 struct uio uio; 278 struct ptrace_io_desc *piod; 279 int error = 0; 280 int write, tmp; 281 282 write = 0; 283 if (req == PT_TRACE_ME) { 284 p = curp; 285 } else { 286 if ((p = pfind(pid)) == NULL) 287 return ESRCH; 288 } 289 if (!PRISON_CHECK(curp->p_ucred, p->p_ucred)) 290 return (ESRCH); 291 292 /* Can't trace a process that's currently exec'ing. */ 293 if ((p->p_flag & P_INEXEC) != 0) 294 return EAGAIN; 295 296 /* 297 * Permissions check 298 */ 299 switch (req) { 300 case PT_TRACE_ME: 301 /* Always legal. */ 302 break; 303 304 case PT_ATTACH: 305 /* Self */ 306 if (p->p_pid == curp->p_pid) 307 return EINVAL; 308 309 /* Already traced */ 310 if (p->p_flag & P_TRACED) 311 return EBUSY; 312 313 if (curp->p_flag & P_TRACED) 314 for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr) 315 if (pp == p) 316 return (EINVAL); 317 318 /* not owned by you, has done setuid (unless you're root) */ 319 if ((p->p_ucred->cr_ruid != curp->p_ucred->cr_ruid) || 320 (p->p_flag & P_SUGID)) { 321 if ((error = suser_cred(curp->p_ucred, 0)) != 0) 322 return error; 323 } 324 325 /* can't trace init when securelevel > 0 */ 326 if (securelevel > 0 && p->p_pid == 1) 327 return EPERM; 328 329 /* OK */ 330 break; 331 332 case PT_READ_I: 333 case PT_READ_D: 334 case PT_WRITE_I: 335 case PT_WRITE_D: 336 case PT_IO: 337 case PT_CONTINUE: 338 case PT_KILL: 339 case PT_STEP: 340 case PT_DETACH: 341 #ifdef PT_GETREGS 342 case PT_GETREGS: 343 #endif 344 #ifdef PT_SETREGS 345 case PT_SETREGS: 346 #endif 347 #ifdef PT_GETFPREGS 348 case PT_GETFPREGS: 349 #endif 350 #ifdef PT_SETFPREGS 351 case PT_SETFPREGS: 352 #endif 353 #ifdef PT_GETDBREGS 354 case PT_GETDBREGS: 355 #endif 356 #ifdef PT_SETDBREGS 357 case PT_SETDBREGS: 358 #endif 359 /* not being traced... */ 360 if ((p->p_flag & P_TRACED) == 0) 361 return EPERM; 362 363 /* not being traced by YOU */ 364 if (p->p_pptr != curp) 365 return EBUSY; 366 367 /* not currently stopped */ 368 if ((p->p_flag & P_STOPPED) == 0 || 369 (p->p_flag & P_WAITED) == 0) { 370 return EBUSY; 371 } 372 373 /* OK */ 374 break; 375 376 default: 377 return EINVAL; 378 } 379 380 #ifdef FIX_SSTEP 381 /* 382 * Single step fixup ala procfs 383 */ 384 FIX_SSTEP(p); 385 #endif 386 387 /* 388 * Actually do the requests 389 */ 390 391 *res = 0; 392 393 switch (req) { 394 case PT_TRACE_ME: 395 /* set my trace flag and "owner" so it can read/write me */ 396 p->p_flag |= P_TRACED; 397 p->p_oppid = p->p_pptr->p_pid; 398 return 0; 399 400 case PT_ATTACH: 401 /* security check done above */ 402 p->p_flag |= P_TRACED; 403 p->p_oppid = p->p_pptr->p_pid; 404 if (p->p_pptr != curp) 405 proc_reparent(p, curp); 406 data = SIGSTOP; 407 goto sendsig; /* in PT_CONTINUE below */ 408 409 case PT_STEP: 410 case PT_CONTINUE: 411 case PT_DETACH: 412 /* Zero means do not send any signal */ 413 if (data < 0 || data > _SIG_MAXSIG) 414 return EINVAL; 415 416 PHOLD(p); 417 418 if (req == PT_STEP) { 419 if ((error = ptrace_single_step (&p->p_lwp))) { 420 PRELE(p); 421 return error; 422 } 423 } 424 425 if (addr != (void *)1) { 426 if ((error = ptrace_set_pc (p, 427 (u_long)(uintfptr_t)addr))) { 428 PRELE(p); 429 return error; 430 } 431 } 432 PRELE(p); 433 434 if (req == PT_DETACH) { 435 /* reset process parent */ 436 if (p->p_oppid != p->p_pptr->p_pid) { 437 struct proc *pp; 438 439 pp = pfind(p->p_oppid); 440 proc_reparent(p, pp ? pp : initproc); 441 } 442 443 p->p_flag &= ~(P_TRACED | P_WAITED); 444 p->p_oppid = 0; 445 446 /* should we send SIGCHLD? */ 447 } 448 449 sendsig: 450 /* 451 * Deliver or queue signal. If the process is stopped 452 * force it to SRUN again. 453 */ 454 crit_enter(); 455 if (p->p_flag & P_STOPPED) { 456 p->p_xstat = data; 457 p->p_flag |= P_BREAKTSLEEP; 458 setrunnable(p); 459 } else if (data) { 460 ksignal(p, data); 461 } 462 crit_exit(); 463 return 0; 464 465 case PT_WRITE_I: 466 case PT_WRITE_D: 467 write = 1; 468 /* fallthrough */ 469 case PT_READ_I: 470 case PT_READ_D: 471 /* 472 * NOTE! uio_offset represents the offset in the target 473 * process. The iov is in the current process (the guy 474 * making the ptrace call) so uio_td must be the current 475 * process (though for a SYSSPACE transfer it doesn't 476 * really matter). 477 */ 478 tmp = 0; 479 /* write = 0 set above */ 480 iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp; 481 iov.iov_len = sizeof(int); 482 uio.uio_iov = &iov; 483 uio.uio_iovcnt = 1; 484 uio.uio_offset = (off_t)(uintptr_t)addr; 485 uio.uio_resid = sizeof(int); 486 uio.uio_segflg = UIO_SYSSPACE; 487 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 488 uio.uio_td = curp->p_thread; 489 error = procfs_domem(curp, p, NULL, &uio); 490 if (uio.uio_resid != 0) { 491 /* 492 * XXX procfs_domem() doesn't currently return ENOSPC, 493 * so I think write() can bogusly return 0. 494 * XXX what happens for short writes? We don't want 495 * to write partial data. 496 * XXX procfs_domem() returns EPERM for other invalid 497 * addresses. Convert this to EINVAL. Does this 498 * clobber returns of EPERM for other reasons? 499 */ 500 if (error == 0 || error == ENOSPC || error == EPERM) 501 error = EINVAL; /* EOF */ 502 } 503 if (!write) 504 *res = tmp; 505 return (error); 506 507 case PT_IO: 508 /* 509 * NOTE! uio_offset represents the offset in the target 510 * process. The iov is in the current process (the guy 511 * making the ptrace call) so uio_td must be the current 512 * process. 513 */ 514 piod = addr; 515 iov.iov_base = piod->piod_addr; 516 iov.iov_len = piod->piod_len; 517 uio.uio_iov = &iov; 518 uio.uio_iovcnt = 1; 519 uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs; 520 uio.uio_resid = piod->piod_len; 521 uio.uio_segflg = UIO_USERSPACE; 522 uio.uio_td = curp->p_thread; 523 switch (piod->piod_op) { 524 case PIOD_READ_D: 525 case PIOD_READ_I: 526 uio.uio_rw = UIO_READ; 527 break; 528 case PIOD_WRITE_D: 529 case PIOD_WRITE_I: 530 uio.uio_rw = UIO_WRITE; 531 break; 532 default: 533 return (EINVAL); 534 } 535 error = procfs_domem(curp, p, NULL, &uio); 536 piod->piod_len -= uio.uio_resid; 537 return (error); 538 539 case PT_KILL: 540 data = SIGKILL; 541 goto sendsig; /* in PT_CONTINUE above */ 542 543 #ifdef PT_SETREGS 544 case PT_SETREGS: 545 write = 1; 546 /* fallthrough */ 547 #endif /* PT_SETREGS */ 548 #ifdef PT_GETREGS 549 case PT_GETREGS: 550 /* write = 0 above */ 551 #endif /* PT_SETREGS */ 552 #if defined(PT_SETREGS) || defined(PT_GETREGS) 553 if (!procfs_validregs(p)) /* no P_SYSTEM procs please */ 554 return EINVAL; 555 else { 556 iov.iov_base = addr; 557 iov.iov_len = sizeof(struct reg); 558 uio.uio_iov = &iov; 559 uio.uio_iovcnt = 1; 560 uio.uio_offset = 0; 561 uio.uio_resid = sizeof(struct reg); 562 uio.uio_segflg = UIO_SYSSPACE; 563 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 564 uio.uio_td = curp->p_thread; 565 return (procfs_doregs(curp, p, NULL, &uio)); 566 } 567 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */ 568 569 #ifdef PT_SETFPREGS 570 case PT_SETFPREGS: 571 write = 1; 572 /* fallthrough */ 573 #endif /* PT_SETFPREGS */ 574 #ifdef PT_GETFPREGS 575 case PT_GETFPREGS: 576 /* write = 0 above */ 577 #endif /* PT_SETFPREGS */ 578 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS) 579 if (!procfs_validfpregs(p)) /* no P_SYSTEM procs please */ 580 return EINVAL; 581 else { 582 iov.iov_base = addr; 583 iov.iov_len = sizeof(struct fpreg); 584 uio.uio_iov = &iov; 585 uio.uio_iovcnt = 1; 586 uio.uio_offset = 0; 587 uio.uio_resid = sizeof(struct fpreg); 588 uio.uio_segflg = UIO_SYSSPACE; 589 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 590 uio.uio_td = curp->p_thread; 591 return (procfs_dofpregs(curp, p, NULL, &uio)); 592 } 593 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */ 594 595 #ifdef PT_SETDBREGS 596 case PT_SETDBREGS: 597 write = 1; 598 /* fallthrough */ 599 #endif /* PT_SETDBREGS */ 600 #ifdef PT_GETDBREGS 601 case PT_GETDBREGS: 602 /* write = 0 above */ 603 #endif /* PT_SETDBREGS */ 604 #if defined(PT_SETDBREGS) || defined(PT_GETDBREGS) 605 if (!procfs_validdbregs(p)) /* no P_SYSTEM procs please */ 606 return EINVAL; 607 else { 608 iov.iov_base = addr; 609 iov.iov_len = sizeof(struct dbreg); 610 uio.uio_iov = &iov; 611 uio.uio_iovcnt = 1; 612 uio.uio_offset = 0; 613 uio.uio_resid = sizeof(struct dbreg); 614 uio.uio_segflg = UIO_SYSSPACE; 615 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 616 uio.uio_td = curp->p_thread; 617 return (procfs_dodbregs(curp, p, NULL, &uio)); 618 } 619 #endif /* defined(PT_SETDBREGS) || defined(PT_GETDBREGS) */ 620 621 default: 622 break; 623 } 624 625 return 0; 626 } 627 628 int 629 trace_req(struct proc *p) 630 { 631 return 1; 632 } 633 634 /* 635 * stopevent() 636 * 637 * Stop a process because of a procfs event. Stay stopped until p->p_step 638 * is cleared (cleared by PIOCCONT in procfs). 639 * 640 * MPSAFE 641 */ 642 void 643 stopevent(struct proc *p, unsigned int event, unsigned int val) 644 { 645 p->p_step = 1; 646 647 do { 648 crit_enter(); 649 wakeup(&p->p_stype); /* Wake up any PIOCWAIT'ing procs */ 650 p->p_xstat = val; 651 p->p_stype = event; /* Which event caused the stop? */ 652 tsleep(&p->p_step, 0, "stopevent", 0); 653 crit_exit(); 654 } while (p->p_step); 655 } 656 657