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 1.30 2007/02/19 01:14:23 corecode 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 lwp *lp; 277 struct iovec iov; 278 struct uio uio; 279 struct ptrace_io_desc *piod; 280 int error = 0; 281 int write, tmp; 282 283 write = 0; 284 if (req == PT_TRACE_ME) { 285 p = curp; 286 } else { 287 if ((p = pfind(pid)) == NULL) 288 return ESRCH; 289 } 290 if (!PRISON_CHECK(curp->p_ucred, p->p_ucred)) 291 return (ESRCH); 292 293 /* Can't trace a process that's currently exec'ing. */ 294 if ((p->p_flag & P_INEXEC) != 0) 295 return EAGAIN; 296 297 /* 298 * Permissions check 299 */ 300 switch (req) { 301 case PT_TRACE_ME: 302 /* Always legal. */ 303 break; 304 305 case PT_ATTACH: 306 /* Self */ 307 if (p->p_pid == curp->p_pid) 308 return EINVAL; 309 310 /* Already traced */ 311 if (p->p_flag & P_TRACED) 312 return EBUSY; 313 314 if (curp->p_flag & P_TRACED) 315 for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr) 316 if (pp == p) 317 return (EINVAL); 318 319 /* not owned by you, has done setuid (unless you're root) */ 320 if ((p->p_ucred->cr_ruid != curp->p_ucred->cr_ruid) || 321 (p->p_flag & P_SUGID)) { 322 if ((error = suser_cred(curp->p_ucred, 0)) != 0) 323 return error; 324 } 325 326 /* can't trace init when securelevel > 0 */ 327 if (securelevel > 0 && p->p_pid == 1) 328 return EPERM; 329 330 /* OK */ 331 break; 332 333 case PT_READ_I: 334 case PT_READ_D: 335 case PT_WRITE_I: 336 case PT_WRITE_D: 337 case PT_IO: 338 case PT_CONTINUE: 339 case PT_KILL: 340 case PT_STEP: 341 case PT_DETACH: 342 #ifdef PT_GETREGS 343 case PT_GETREGS: 344 #endif 345 #ifdef PT_SETREGS 346 case PT_SETREGS: 347 #endif 348 #ifdef PT_GETFPREGS 349 case PT_GETFPREGS: 350 #endif 351 #ifdef PT_SETFPREGS 352 case PT_SETFPREGS: 353 #endif 354 #ifdef PT_GETDBREGS 355 case PT_GETDBREGS: 356 #endif 357 #ifdef PT_SETDBREGS 358 case PT_SETDBREGS: 359 #endif 360 /* not being traced... */ 361 if ((p->p_flag & P_TRACED) == 0) 362 return EPERM; 363 364 /* not being traced by YOU */ 365 if (p->p_pptr != curp) 366 return EBUSY; 367 368 /* not currently stopped */ 369 if (p->p_stat != SSTOP || 370 (p->p_flag & P_WAITED) == 0) { 371 return EBUSY; 372 } 373 374 /* OK */ 375 break; 376 377 default: 378 return EINVAL; 379 } 380 381 /* XXX lwp */ 382 lp = FIRST_LWP_IN_PROC(p); 383 #ifdef FIX_SSTEP 384 /* 385 * Single step fixup ala procfs 386 */ 387 FIX_SSTEP(lp); 388 #endif 389 390 /* 391 * Actually do the requests 392 */ 393 394 *res = 0; 395 396 switch (req) { 397 case PT_TRACE_ME: 398 /* set my trace flag and "owner" so it can read/write me */ 399 p->p_flag |= P_TRACED; 400 p->p_oppid = p->p_pptr->p_pid; 401 return 0; 402 403 case PT_ATTACH: 404 /* security check done above */ 405 p->p_flag |= P_TRACED; 406 p->p_oppid = p->p_pptr->p_pid; 407 if (p->p_pptr != curp) 408 proc_reparent(p, curp); 409 data = SIGSTOP; 410 goto sendsig; /* in PT_CONTINUE below */ 411 412 case PT_STEP: 413 case PT_CONTINUE: 414 case PT_DETACH: 415 /* Zero means do not send any signal */ 416 if (data < 0 || data > _SIG_MAXSIG) 417 return EINVAL; 418 419 LWPHOLD(lp); 420 421 if (req == PT_STEP) { 422 if ((error = ptrace_single_step (lp))) { 423 LWPRELE(lp); 424 return error; 425 } 426 } 427 428 if (addr != (void *)1) { 429 if ((error = ptrace_set_pc (lp, 430 (u_long)(uintfptr_t)addr))) { 431 LWPRELE(lp); 432 return error; 433 } 434 } 435 LWPRELE(lp); 436 437 if (req == PT_DETACH) { 438 /* reset process parent */ 439 if (p->p_oppid != p->p_pptr->p_pid) { 440 struct proc *pp; 441 442 pp = pfind(p->p_oppid); 443 proc_reparent(p, pp ? pp : initproc); 444 } 445 446 p->p_flag &= ~(P_TRACED | P_WAITED); 447 p->p_oppid = 0; 448 449 /* should we send SIGCHLD? */ 450 } 451 452 sendsig: 453 /* 454 * Deliver or queue signal. If the process is stopped 455 * force it to be SACTIVE again. 456 */ 457 crit_enter(); 458 if (p->p_stat == SSTOP) { 459 p->p_xstat = data; 460 lp->lwp_flag |= LWP_BREAKTSLEEP; 461 proc_unstop(p); 462 } else if (data) { 463 ksignal(p, data); 464 } 465 crit_exit(); 466 return 0; 467 468 case PT_WRITE_I: 469 case PT_WRITE_D: 470 write = 1; 471 /* fallthrough */ 472 case PT_READ_I: 473 case PT_READ_D: 474 /* 475 * NOTE! uio_offset represents the offset in the target 476 * process. The iov is in the current process (the guy 477 * making the ptrace call) so uio_td must be the current 478 * process (though for a SYSSPACE transfer it doesn't 479 * really matter). 480 */ 481 tmp = 0; 482 /* write = 0 set above */ 483 iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp; 484 iov.iov_len = sizeof(int); 485 uio.uio_iov = &iov; 486 uio.uio_iovcnt = 1; 487 uio.uio_offset = (off_t)(uintptr_t)addr; 488 uio.uio_resid = sizeof(int); 489 uio.uio_segflg = UIO_SYSSPACE; 490 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 491 uio.uio_td = curthread; 492 error = procfs_domem(curp, lp, NULL, &uio); 493 if (uio.uio_resid != 0) { 494 /* 495 * XXX procfs_domem() doesn't currently return ENOSPC, 496 * so I think write() can bogusly return 0. 497 * XXX what happens for short writes? We don't want 498 * to write partial data. 499 * XXX procfs_domem() returns EPERM for other invalid 500 * addresses. Convert this to EINVAL. Does this 501 * clobber returns of EPERM for other reasons? 502 */ 503 if (error == 0 || error == ENOSPC || error == EPERM) 504 error = EINVAL; /* EOF */ 505 } 506 if (!write) 507 *res = tmp; 508 return (error); 509 510 case PT_IO: 511 /* 512 * NOTE! uio_offset represents the offset in the target 513 * process. The iov is in the current process (the guy 514 * making the ptrace call) so uio_td must be the current 515 * process. 516 */ 517 piod = addr; 518 iov.iov_base = piod->piod_addr; 519 iov.iov_len = piod->piod_len; 520 uio.uio_iov = &iov; 521 uio.uio_iovcnt = 1; 522 uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs; 523 uio.uio_resid = piod->piod_len; 524 uio.uio_segflg = UIO_USERSPACE; 525 uio.uio_td = curthread; 526 switch (piod->piod_op) { 527 case PIOD_READ_D: 528 case PIOD_READ_I: 529 uio.uio_rw = UIO_READ; 530 break; 531 case PIOD_WRITE_D: 532 case PIOD_WRITE_I: 533 uio.uio_rw = UIO_WRITE; 534 break; 535 default: 536 return (EINVAL); 537 } 538 error = procfs_domem(curp, lp, NULL, &uio); 539 piod->piod_len -= uio.uio_resid; 540 return (error); 541 542 case PT_KILL: 543 data = SIGKILL; 544 goto sendsig; /* in PT_CONTINUE above */ 545 546 #ifdef PT_SETREGS 547 case PT_SETREGS: 548 write = 1; 549 /* fallthrough */ 550 #endif /* PT_SETREGS */ 551 #ifdef PT_GETREGS 552 case PT_GETREGS: 553 /* write = 0 above */ 554 #endif /* PT_SETREGS */ 555 #if defined(PT_SETREGS) || defined(PT_GETREGS) 556 if (!procfs_validregs(lp)) /* no P_SYSTEM procs please */ 557 return EINVAL; 558 else { 559 iov.iov_base = addr; 560 iov.iov_len = sizeof(struct reg); 561 uio.uio_iov = &iov; 562 uio.uio_iovcnt = 1; 563 uio.uio_offset = 0; 564 uio.uio_resid = sizeof(struct reg); 565 uio.uio_segflg = UIO_SYSSPACE; 566 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 567 uio.uio_td = curthread; 568 return (procfs_doregs(curp, lp, NULL, &uio)); 569 } 570 #endif /* defined(PT_SETREGS) || defined(PT_GETREGS) */ 571 572 #ifdef PT_SETFPREGS 573 case PT_SETFPREGS: 574 write = 1; 575 /* fallthrough */ 576 #endif /* PT_SETFPREGS */ 577 #ifdef PT_GETFPREGS 578 case PT_GETFPREGS: 579 /* write = 0 above */ 580 #endif /* PT_SETFPREGS */ 581 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS) 582 if (!procfs_validfpregs(lp)) /* no P_SYSTEM procs please */ 583 return EINVAL; 584 else { 585 iov.iov_base = addr; 586 iov.iov_len = sizeof(struct fpreg); 587 uio.uio_iov = &iov; 588 uio.uio_iovcnt = 1; 589 uio.uio_offset = 0; 590 uio.uio_resid = sizeof(struct fpreg); 591 uio.uio_segflg = UIO_SYSSPACE; 592 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 593 uio.uio_td = curthread; 594 return (procfs_dofpregs(curp, lp, NULL, &uio)); 595 } 596 #endif /* defined(PT_SETFPREGS) || defined(PT_GETFPREGS) */ 597 598 #ifdef PT_SETDBREGS 599 case PT_SETDBREGS: 600 write = 1; 601 /* fallthrough */ 602 #endif /* PT_SETDBREGS */ 603 #ifdef PT_GETDBREGS 604 case PT_GETDBREGS: 605 /* write = 0 above */ 606 #endif /* PT_SETDBREGS */ 607 #if defined(PT_SETDBREGS) || defined(PT_GETDBREGS) 608 if (!procfs_validdbregs(lp)) /* no P_SYSTEM procs please */ 609 return EINVAL; 610 else { 611 iov.iov_base = addr; 612 iov.iov_len = sizeof(struct dbreg); 613 uio.uio_iov = &iov; 614 uio.uio_iovcnt = 1; 615 uio.uio_offset = 0; 616 uio.uio_resid = sizeof(struct dbreg); 617 uio.uio_segflg = UIO_SYSSPACE; 618 uio.uio_rw = write ? UIO_WRITE : UIO_READ; 619 uio.uio_td = curthread; 620 return (procfs_dodbregs(curp, lp, NULL, &uio)); 621 } 622 #endif /* defined(PT_SETDBREGS) || defined(PT_GETDBREGS) */ 623 624 default: 625 break; 626 } 627 628 return 0; 629 } 630 631 int 632 trace_req(struct proc *p) 633 { 634 return 1; 635 } 636 637 /* 638 * stopevent() 639 * 640 * Stop a process because of a procfs event. Stay stopped until p->p_step 641 * is cleared (cleared by PIOCCONT in procfs). 642 * 643 * MPSAFE 644 */ 645 void 646 stopevent(struct proc *p, unsigned int event, unsigned int val) 647 { 648 p->p_step = 1; 649 650 do { 651 crit_enter(); 652 wakeup(&p->p_stype); /* Wake up any PIOCWAIT'ing procs */ 653 p->p_xstat = val; 654 p->p_stype = event; /* Which event caused the stop? */ 655 tsleep(&p->p_step, 0, "stopevent", 0); 656 crit_exit(); 657 } while (p->p_step); 658 } 659 660