1 /*- 2 * Copyright (c) 2000 David O'Brien 3 * Copyright (c) 1995-1996 S�ren Schmidt 4 * Copyright (c) 1996 Peter Wemm 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer 12 * in this position and unchanged. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. The name of the author may not be used to endorse or promote products 17 * derived from this software withough specific prior written permission 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29 * 30 * $FreeBSD$ 31 */ 32 33 #include <sys/param.h> 34 #include <sys/exec.h> 35 #include <sys/fcntl.h> 36 #include <sys/imgact.h> 37 #include <sys/imgact_elf.h> 38 #include <sys/kernel.h> 39 #include <sys/lock.h> 40 #include <sys/malloc.h> 41 #include <sys/mutex.h> 42 #include <sys/mman.h> 43 #include <sys/namei.h> 44 #include <sys/pioctl.h> 45 #include <sys/proc.h> 46 #include <sys/procfs.h> 47 #include <sys/resourcevar.h> 48 #include <sys/systm.h> 49 #include <sys/signalvar.h> 50 #include <sys/stat.h> 51 #include <sys/sx.h> 52 #include <sys/syscall.h> 53 #include <sys/sysctl.h> 54 #include <sys/sysent.h> 55 #include <sys/vnode.h> 56 57 #include <vm/vm.h> 58 #include <vm/vm_kern.h> 59 #include <vm/vm_param.h> 60 #include <vm/pmap.h> 61 #include <vm/vm_map.h> 62 #include <vm/vm_object.h> 63 #include <vm/vm_extern.h> 64 65 #include <machine/elf.h> 66 #include <machine/md_var.h> 67 68 #define OLD_EI_BRAND 8 69 70 __ElfType(Brandinfo); 71 __ElfType(Auxargs); 72 73 static int elf_check_header(const Elf_Ehdr *hdr); 74 static int elf_freebsd_fixup(register_t **stack_base, 75 struct image_params *imgp); 76 static int elf_load_file(struct proc *p, const char *file, u_long *addr, 77 u_long *entry); 78 static int elf_load_section(struct proc *p, 79 struct vmspace *vmspace, struct vnode *vp, 80 vm_offset_t offset, caddr_t vmaddr, size_t memsz, size_t filsz, 81 vm_prot_t prot); 82 static int exec_elf_imgact(struct image_params *imgp); 83 84 static int elf_trace = 0; 85 SYSCTL_INT(_debug, OID_AUTO, elf_trace, CTLFLAG_RW, &elf_trace, 0, ""); 86 87 struct sysentvec elf_freebsd_sysvec = { 88 SYS_MAXSYSCALL, 89 sysent, 90 0, 91 0, 92 0, 93 0, 94 0, 95 0, 96 elf_freebsd_fixup, 97 sendsig, 98 sigcode, 99 &szsigcode, 100 0, 101 "FreeBSD ELF", 102 elf_coredump, 103 NULL, 104 MINSIGSTKSZ 105 }; 106 107 static Elf_Brandinfo freebsd_brand_info = { 108 ELFOSABI_FREEBSD, 109 "FreeBSD", 110 "", 111 "/usr/libexec/ld-elf.so.1", 112 &elf_freebsd_sysvec 113 }; 114 static Elf_Brandinfo *elf_brand_list[MAX_BRANDS] = { 115 &freebsd_brand_info, 116 NULL, NULL, NULL, 117 NULL, NULL, NULL, NULL 118 }; 119 120 int 121 elf_insert_brand_entry(Elf_Brandinfo *entry) 122 { 123 int i; 124 125 for (i=1; i<MAX_BRANDS; i++) { 126 if (elf_brand_list[i] == NULL) { 127 elf_brand_list[i] = entry; 128 break; 129 } 130 } 131 if (i == MAX_BRANDS) 132 return -1; 133 return 0; 134 } 135 136 int 137 elf_remove_brand_entry(Elf_Brandinfo *entry) 138 { 139 int i; 140 141 for (i=1; i<MAX_BRANDS; i++) { 142 if (elf_brand_list[i] == entry) { 143 elf_brand_list[i] = NULL; 144 break; 145 } 146 } 147 if (i == MAX_BRANDS) 148 return -1; 149 return 0; 150 } 151 152 int 153 elf_brand_inuse(Elf_Brandinfo *entry) 154 { 155 struct proc *p; 156 int rval = FALSE; 157 158 sx_slock(&allproc_lock); 159 LIST_FOREACH(p, &allproc, p_list) { 160 if (p->p_sysent == entry->sysvec) { 161 rval = TRUE; 162 break; 163 } 164 } 165 sx_sunlock(&allproc_lock); 166 167 return (rval); 168 } 169 170 static int 171 elf_check_header(const Elf_Ehdr *hdr) 172 { 173 if (!IS_ELF(*hdr) || 174 hdr->e_ident[EI_CLASS] != ELF_TARG_CLASS || 175 hdr->e_ident[EI_DATA] != ELF_TARG_DATA || 176 hdr->e_ident[EI_VERSION] != EV_CURRENT) 177 return ENOEXEC; 178 179 if (!ELF_MACHINE_OK(hdr->e_machine)) 180 return ENOEXEC; 181 182 if (hdr->e_version != ELF_TARG_VER) 183 return ENOEXEC; 184 185 return 0; 186 } 187 188 static int 189 elf_load_section(struct proc *p, struct vmspace *vmspace, struct vnode *vp, vm_offset_t offset, caddr_t vmaddr, size_t memsz, size_t filsz, vm_prot_t prot) 190 { 191 size_t map_len; 192 vm_offset_t map_addr; 193 int error, rv; 194 size_t copy_len; 195 vm_object_t object; 196 vm_offset_t file_addr; 197 vm_offset_t data_buf = 0; 198 199 GIANT_REQUIRED; 200 201 VOP_GETVOBJECT(vp, &object); 202 error = 0; 203 204 /* 205 * It's necessary to fail if the filsz + offset taken from the 206 * header is greater than the actual file pager object's size. 207 * If we were to allow this, then the vm_map_find() below would 208 * walk right off the end of the file object and into the ether. 209 * 210 * While I'm here, might as well check for something else that 211 * is invalid: filsz cannot be greater than memsz. 212 */ 213 if ((off_t)filsz + offset > object->un_pager.vnp.vnp_size || 214 filsz > memsz) { 215 uprintf("elf_load_section: truncated ELF file\n"); 216 return (ENOEXEC); 217 } 218 219 map_addr = trunc_page((vm_offset_t)vmaddr); 220 file_addr = trunc_page(offset); 221 222 /* 223 * We have two choices. We can either clear the data in the last page 224 * of an oversized mapping, or we can start the anon mapping a page 225 * early and copy the initialized data into that first page. We 226 * choose the second.. 227 */ 228 if (memsz > filsz) 229 map_len = trunc_page(offset+filsz) - file_addr; 230 else 231 map_len = round_page(offset+filsz) - file_addr; 232 233 if (map_len != 0) { 234 vm_object_reference(object); 235 vm_map_lock(&vmspace->vm_map); 236 rv = vm_map_insert(&vmspace->vm_map, 237 object, 238 file_addr, /* file offset */ 239 map_addr, /* virtual start */ 240 map_addr + map_len,/* virtual end */ 241 prot, 242 VM_PROT_ALL, 243 MAP_COPY_ON_WRITE | MAP_PREFAULT); 244 vm_map_unlock(&vmspace->vm_map); 245 if (rv != KERN_SUCCESS) { 246 vm_object_deallocate(object); 247 return EINVAL; 248 } 249 250 /* we can stop now if we've covered it all */ 251 if (memsz == filsz) { 252 return 0; 253 } 254 } 255 256 257 /* 258 * We have to get the remaining bit of the file into the first part 259 * of the oversized map segment. This is normally because the .data 260 * segment in the file is extended to provide bss. It's a neat idea 261 * to try and save a page, but it's a pain in the behind to implement. 262 */ 263 copy_len = (offset + filsz) - trunc_page(offset + filsz); 264 map_addr = trunc_page((vm_offset_t)vmaddr + filsz); 265 map_len = round_page((vm_offset_t)vmaddr + memsz) - map_addr; 266 267 /* This had damn well better be true! */ 268 if (map_len != 0) { 269 vm_map_lock(&vmspace->vm_map); 270 rv = vm_map_insert(&vmspace->vm_map, NULL, 0, 271 map_addr, map_addr + map_len, 272 VM_PROT_ALL, VM_PROT_ALL, 0); 273 vm_map_unlock(&vmspace->vm_map); 274 if (rv != KERN_SUCCESS) { 275 return EINVAL; 276 } 277 } 278 279 if (copy_len != 0) { 280 vm_object_reference(object); 281 rv = vm_map_find(exec_map, 282 object, 283 trunc_page(offset + filsz), 284 &data_buf, 285 PAGE_SIZE, 286 TRUE, 287 VM_PROT_READ, 288 VM_PROT_ALL, 289 MAP_COPY_ON_WRITE | MAP_PREFAULT_PARTIAL); 290 if (rv != KERN_SUCCESS) { 291 vm_object_deallocate(object); 292 return EINVAL; 293 } 294 295 /* send the page fragment to user space */ 296 error = copyout((caddr_t)data_buf, (caddr_t)map_addr, copy_len); 297 vm_map_remove(exec_map, data_buf, data_buf + PAGE_SIZE); 298 if (error) { 299 return (error); 300 } 301 } 302 303 /* 304 * set it to the specified protection 305 */ 306 vm_map_protect(&vmspace->vm_map, map_addr, map_addr + map_len, prot, 307 FALSE); 308 309 return error; 310 } 311 312 /* 313 * Load the file "file" into memory. It may be either a shared object 314 * or an executable. 315 * 316 * The "addr" reference parameter is in/out. On entry, it specifies 317 * the address where a shared object should be loaded. If the file is 318 * an executable, this value is ignored. On exit, "addr" specifies 319 * where the file was actually loaded. 320 * 321 * The "entry" reference parameter is out only. On exit, it specifies 322 * the entry point for the loaded file. 323 */ 324 static int 325 elf_load_file(struct proc *p, const char *file, u_long *addr, u_long *entry) 326 { 327 struct { 328 struct nameidata nd; 329 struct vattr attr; 330 struct image_params image_params; 331 } *tempdata; 332 const Elf_Ehdr *hdr = NULL; 333 const Elf_Phdr *phdr = NULL; 334 struct nameidata *nd; 335 struct vmspace *vmspace = p->p_vmspace; 336 struct vattr *attr; 337 struct image_params *imgp; 338 vm_prot_t prot; 339 u_long rbase; 340 u_long base_addr = 0; 341 int error, i, numsegs; 342 343 if (curthread->td_proc != p) 344 panic("elf_load_file - thread"); /* XXXKSE DIAGNOSTIC */ 345 346 tempdata = malloc(sizeof(*tempdata), M_TEMP, M_WAITOK); 347 nd = &tempdata->nd; 348 attr = &tempdata->attr; 349 imgp = &tempdata->image_params; 350 351 /* 352 * Initialize part of the common data 353 */ 354 imgp->proc = p; 355 imgp->uap = NULL; 356 imgp->attr = attr; 357 imgp->firstpage = NULL; 358 imgp->image_header = (char *)kmem_alloc_wait(exec_map, PAGE_SIZE); 359 360 if (imgp->image_header == NULL) { 361 nd->ni_vp = NULL; 362 error = ENOMEM; 363 goto fail; 364 } 365 366 /* XXXKSE */ 367 NDINIT(nd, LOOKUP, LOCKLEAF|FOLLOW, UIO_SYSSPACE, file, curthread); 368 369 if ((error = namei(nd)) != 0) { 370 nd->ni_vp = NULL; 371 goto fail; 372 } 373 NDFREE(nd, NDF_ONLY_PNBUF); 374 imgp->vp = nd->ni_vp; 375 376 /* 377 * Check permissions, modes, uid, etc on the file, and "open" it. 378 */ 379 error = exec_check_permissions(imgp); 380 if (error) { 381 VOP_UNLOCK(nd->ni_vp, 0, curthread); /* XXXKSE */ 382 goto fail; 383 } 384 385 error = exec_map_first_page(imgp); 386 /* 387 * Also make certain that the interpreter stays the same, so set 388 * its VTEXT flag, too. 389 */ 390 if (error == 0) 391 nd->ni_vp->v_flag |= VTEXT; 392 VOP_UNLOCK(nd->ni_vp, 0, curthread); /* XXXKSE */ 393 if (error) 394 goto fail; 395 396 hdr = (const Elf_Ehdr *)imgp->image_header; 397 if ((error = elf_check_header(hdr)) != 0) 398 goto fail; 399 if (hdr->e_type == ET_DYN) 400 rbase = *addr; 401 else if (hdr->e_type == ET_EXEC) 402 rbase = 0; 403 else { 404 error = ENOEXEC; 405 goto fail; 406 } 407 408 /* Only support headers that fit within first page for now */ 409 if ((hdr->e_phoff > PAGE_SIZE) || 410 (hdr->e_phoff + hdr->e_phentsize * hdr->e_phnum) > PAGE_SIZE) { 411 error = ENOEXEC; 412 goto fail; 413 } 414 415 phdr = (const Elf_Phdr *)(imgp->image_header + hdr->e_phoff); 416 417 for (i = 0, numsegs = 0; i < hdr->e_phnum; i++) { 418 if (phdr[i].p_type == PT_LOAD) { /* Loadable segment */ 419 prot = 0; 420 if (phdr[i].p_flags & PF_X) 421 prot |= VM_PROT_EXECUTE; 422 if (phdr[i].p_flags & PF_W) 423 prot |= VM_PROT_WRITE; 424 if (phdr[i].p_flags & PF_R) 425 prot |= VM_PROT_READ; 426 427 if ((error = elf_load_section(p, vmspace, nd->ni_vp, 428 phdr[i].p_offset, 429 (caddr_t)phdr[i].p_vaddr + 430 rbase, 431 phdr[i].p_memsz, 432 phdr[i].p_filesz, prot)) != 0) 433 goto fail; 434 /* 435 * Establish the base address if this is the 436 * first segment. 437 */ 438 if (numsegs == 0) 439 base_addr = trunc_page(phdr[i].p_vaddr + rbase); 440 numsegs++; 441 } 442 } 443 *addr = base_addr; 444 *entry=(unsigned long)hdr->e_entry + rbase; 445 446 fail: 447 if (imgp->firstpage) 448 exec_unmap_first_page(imgp); 449 if (imgp->image_header) 450 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->image_header, 451 PAGE_SIZE); 452 if (nd->ni_vp) 453 vrele(nd->ni_vp); 454 455 free(tempdata, M_TEMP); 456 457 return error; 458 } 459 460 /* 461 * non static, as it can be overridden by start_init() 462 */ 463 #ifdef __ia64__ 464 int fallback_elf_brand = ELFOSABI_FREEBSD; 465 #else 466 int fallback_elf_brand = -1; 467 #endif 468 SYSCTL_INT(_kern, OID_AUTO, fallback_elf_brand, CTLFLAG_RW, 469 &fallback_elf_brand, -1, 470 "ELF brand of last resort"); 471 472 static int 473 exec_elf_imgact(struct image_params *imgp) 474 { 475 const Elf_Ehdr *hdr = (const Elf_Ehdr *) imgp->image_header; 476 const Elf_Phdr *phdr; 477 Elf_Auxargs *elf_auxargs = NULL; 478 struct vmspace *vmspace; 479 vm_prot_t prot; 480 u_long text_size = 0, data_size = 0; 481 u_long text_addr = 0, data_addr = 0; 482 u_long addr, entry = 0, proghdr = 0; 483 int error, i; 484 const char *interp = NULL; 485 Elf_Brandinfo *brand_info; 486 char *path; 487 488 GIANT_REQUIRED; 489 490 /* 491 * Do we have a valid ELF header ? 492 */ 493 if (elf_check_header(hdr) != 0 || hdr->e_type != ET_EXEC) 494 return -1; 495 496 /* 497 * From here on down, we return an errno, not -1, as we've 498 * detected an ELF file. 499 */ 500 501 if ((hdr->e_phoff > PAGE_SIZE) || 502 (hdr->e_phoff + hdr->e_phentsize * hdr->e_phnum) > PAGE_SIZE) { 503 /* Only support headers in first page for now */ 504 return ENOEXEC; 505 } 506 phdr = (const Elf_Phdr*)(imgp->image_header + hdr->e_phoff); 507 508 /* 509 * From this point on, we may have resources that need to be freed. 510 */ 511 512 /* 513 * Yeah, I'm paranoid. There is every reason in the world to get 514 * VTEXT now since from here on out, there are places we can have 515 * a context switch. Better safe than sorry; I really don't want 516 * the file to change while it's being loaded. 517 */ 518 mtx_lock(&imgp->vp->v_interlock); 519 imgp->vp->v_flag |= VTEXT; 520 mtx_unlock(&imgp->vp->v_interlock); 521 522 if ((error = exec_extract_strings(imgp)) != 0) 523 goto fail; 524 525 exec_new_vmspace(imgp); 526 527 vmspace = imgp->proc->p_vmspace; 528 529 for (i = 0; i < hdr->e_phnum; i++) { 530 switch(phdr[i].p_type) { 531 532 case PT_LOAD: /* Loadable segment */ 533 prot = 0; 534 if (phdr[i].p_flags & PF_X) 535 prot |= VM_PROT_EXECUTE; 536 if (phdr[i].p_flags & PF_W) 537 prot |= VM_PROT_WRITE; 538 if (phdr[i].p_flags & PF_R) 539 prot |= VM_PROT_READ; 540 541 if ((error = elf_load_section(imgp->proc, 542 vmspace, imgp->vp, 543 phdr[i].p_offset, 544 (caddr_t)phdr[i].p_vaddr, 545 phdr[i].p_memsz, 546 phdr[i].p_filesz, prot)) != 0) 547 goto fail; 548 549 /* 550 * Is this .text or .data ?? 551 * 552 * We only handle one each of those yet XXX 553 */ 554 if (hdr->e_entry >= phdr[i].p_vaddr && 555 hdr->e_entry <(phdr[i].p_vaddr+phdr[i].p_memsz)) { 556 text_addr = trunc_page(phdr[i].p_vaddr); 557 text_size = round_page(phdr[i].p_memsz + 558 phdr[i].p_vaddr - 559 text_addr); 560 entry = (u_long)hdr->e_entry; 561 } else { 562 data_addr = trunc_page(phdr[i].p_vaddr); 563 data_size = round_page(phdr[i].p_memsz + 564 phdr[i].p_vaddr - 565 data_addr); 566 } 567 break; 568 case PT_INTERP: /* Path to interpreter */ 569 if (phdr[i].p_filesz > MAXPATHLEN || 570 phdr[i].p_offset + phdr[i].p_filesz > PAGE_SIZE) { 571 error = ENOEXEC; 572 goto fail; 573 } 574 interp = imgp->image_header + phdr[i].p_offset; 575 break; 576 case PT_PHDR: /* Program header table info */ 577 proghdr = phdr[i].p_vaddr; 578 break; 579 default: 580 break; 581 } 582 } 583 584 vmspace->vm_tsize = text_size >> PAGE_SHIFT; 585 vmspace->vm_taddr = (caddr_t)(uintptr_t)text_addr; 586 vmspace->vm_dsize = data_size >> PAGE_SHIFT; 587 vmspace->vm_daddr = (caddr_t)(uintptr_t)data_addr; 588 589 addr = ELF_RTLD_ADDR(vmspace); 590 591 imgp->entry_addr = entry; 592 593 brand_info = NULL; 594 595 /* We support three types of branding -- (1) the ELF EI_OSABI field 596 * that SCO added to the ELF spec, (2) FreeBSD 3.x's traditional string 597 * branding w/in the ELF header, and (3) path of the `interp_path' 598 * field. We should also look for an ".note.ABI-tag" ELF section now 599 * in all Linux ELF binaries, FreeBSD 4.1+, and some NetBSD ones. 600 */ 601 602 /* If the executable has a brand, search for it in the brand list. */ 603 if (brand_info == NULL) { 604 for (i = 0; i < MAX_BRANDS; i++) { 605 Elf_Brandinfo *bi = elf_brand_list[i]; 606 607 if (bi != NULL && 608 (hdr->e_ident[EI_OSABI] == bi->brand 609 || 0 == 610 strncmp((const char *)&hdr->e_ident[OLD_EI_BRAND], 611 bi->compat_3_brand, strlen(bi->compat_3_brand)))) { 612 brand_info = bi; 613 break; 614 } 615 } 616 } 617 618 /* Lacking a known brand, search for a recognized interpreter. */ 619 if (brand_info == NULL && interp != NULL) { 620 for (i = 0; i < MAX_BRANDS; i++) { 621 Elf_Brandinfo *bi = elf_brand_list[i]; 622 623 if (bi != NULL && 624 strcmp(interp, bi->interp_path) == 0) { 625 brand_info = bi; 626 break; 627 } 628 } 629 } 630 631 /* Lacking a recognized interpreter, try the default brand */ 632 if (brand_info == NULL) { 633 for (i = 0; i < MAX_BRANDS; i++) { 634 Elf_Brandinfo *bi = elf_brand_list[i]; 635 636 if (bi != NULL && fallback_elf_brand == bi->brand) { 637 brand_info = bi; 638 break; 639 } 640 } 641 } 642 643 if (brand_info == NULL) { 644 uprintf("ELF binary type \"%u\" not known.\n", 645 hdr->e_ident[EI_OSABI]); 646 error = ENOEXEC; 647 goto fail; 648 } 649 650 imgp->proc->p_sysent = brand_info->sysvec; 651 if (interp != NULL) { 652 path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 653 snprintf(path, MAXPATHLEN, "%s%s", 654 brand_info->emul_path, interp); 655 if ((error = elf_load_file(imgp->proc, path, &addr, 656 &imgp->entry_addr)) != 0) { 657 if ((error = elf_load_file(imgp->proc, interp, &addr, 658 &imgp->entry_addr)) != 0) { 659 uprintf("ELF interpreter %s not found\n", path); 660 free(path, M_TEMP); 661 goto fail; 662 } 663 } 664 free(path, M_TEMP); 665 } 666 667 /* 668 * Construct auxargs table (used by the fixup routine) 669 */ 670 elf_auxargs = malloc(sizeof(Elf_Auxargs), M_TEMP, M_WAITOK); 671 elf_auxargs->execfd = -1; 672 elf_auxargs->phdr = proghdr; 673 elf_auxargs->phent = hdr->e_phentsize; 674 elf_auxargs->phnum = hdr->e_phnum; 675 elf_auxargs->pagesz = PAGE_SIZE; 676 elf_auxargs->base = addr; 677 elf_auxargs->flags = 0; 678 elf_auxargs->entry = entry; 679 elf_auxargs->trace = elf_trace; 680 681 imgp->auxargs = elf_auxargs; 682 imgp->interpreted = 0; 683 684 fail: 685 return error; 686 } 687 688 static int 689 elf_freebsd_fixup(register_t **stack_base, struct image_params *imgp) 690 { 691 Elf_Auxargs *args = (Elf_Auxargs *)imgp->auxargs; 692 register_t *pos; 693 694 pos = *stack_base + (imgp->argc + imgp->envc + 2); 695 696 if (args->trace) { 697 AUXARGS_ENTRY(pos, AT_DEBUG, 1); 698 } 699 if (args->execfd != -1) { 700 AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd); 701 } 702 AUXARGS_ENTRY(pos, AT_PHDR, args->phdr); 703 AUXARGS_ENTRY(pos, AT_PHENT, args->phent); 704 AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum); 705 AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz); 706 AUXARGS_ENTRY(pos, AT_FLAGS, args->flags); 707 AUXARGS_ENTRY(pos, AT_ENTRY, args->entry); 708 AUXARGS_ENTRY(pos, AT_BASE, args->base); 709 AUXARGS_ENTRY(pos, AT_NULL, 0); 710 711 free(imgp->auxargs, M_TEMP); 712 imgp->auxargs = NULL; 713 714 (*stack_base)--; 715 suword(*stack_base, (long) imgp->argc); 716 return 0; 717 } 718 719 /* 720 * Code for generating ELF core dumps. 721 */ 722 723 typedef void (*segment_callback)(vm_map_entry_t, void *); 724 725 /* Closure for cb_put_phdr(). */ 726 struct phdr_closure { 727 Elf_Phdr *phdr; /* Program header to fill in */ 728 Elf_Off offset; /* Offset of segment in core file */ 729 }; 730 731 /* Closure for cb_size_segment(). */ 732 struct sseg_closure { 733 int count; /* Count of writable segments. */ 734 size_t size; /* Total size of all writable segments. */ 735 }; 736 737 static void cb_put_phdr(vm_map_entry_t, void *); 738 static void cb_size_segment(vm_map_entry_t, void *); 739 static void each_writable_segment(struct proc *, segment_callback, void *); 740 static int elf_corehdr(struct thread *, struct vnode *, struct ucred *, 741 int, void *, size_t); 742 static void elf_puthdr(struct proc *, void *, size_t *, 743 const prstatus_t *, const prfpregset_t *, const prpsinfo_t *, int); 744 static void elf_putnote(void *, size_t *, const char *, int, 745 const void *, size_t); 746 747 extern int osreldate; 748 749 int 750 elf_coredump(td, vp, limit) 751 struct thread *td; 752 register struct vnode *vp; 753 off_t limit; 754 { 755 register struct proc *p = td->td_proc; 756 register struct ucred *cred = td->td_ucred; 757 int error = 0; 758 struct sseg_closure seginfo; 759 void *hdr; 760 size_t hdrsize; 761 762 /* Size the program segments. */ 763 seginfo.count = 0; 764 seginfo.size = 0; 765 each_writable_segment(p, cb_size_segment, &seginfo); 766 767 /* 768 * Calculate the size of the core file header area by making 769 * a dry run of generating it. Nothing is written, but the 770 * size is calculated. 771 */ 772 hdrsize = 0; 773 elf_puthdr((struct proc *)NULL, (void *)NULL, &hdrsize, 774 (const prstatus_t *)NULL, (const prfpregset_t *)NULL, 775 (const prpsinfo_t *)NULL, seginfo.count); 776 777 if (hdrsize + seginfo.size >= limit) 778 return (EFAULT); 779 780 /* 781 * Allocate memory for building the header, fill it up, 782 * and write it out. 783 */ 784 hdr = malloc(hdrsize, M_TEMP, M_WAITOK); 785 if (hdr == NULL) { 786 return EINVAL; 787 } 788 error = elf_corehdr(td, vp, cred, seginfo.count, hdr, hdrsize); 789 790 /* Write the contents of all of the writable segments. */ 791 if (error == 0) { 792 Elf_Phdr *php; 793 off_t offset; 794 int i; 795 796 php = (Elf_Phdr *)((char *)hdr + sizeof(Elf_Ehdr)) + 1; 797 offset = hdrsize; 798 for (i = 0; i < seginfo.count; i++) { 799 error = vn_rdwr_inchunks(UIO_WRITE, vp, 800 (caddr_t)php->p_vaddr, 801 php->p_filesz, offset, UIO_USERSPACE, 802 IO_UNIT | IO_DIRECT, cred, (int *)NULL, curthread); /* XXXKSE */ 803 if (error != 0) 804 break; 805 offset += php->p_filesz; 806 php++; 807 } 808 } 809 free(hdr, M_TEMP); 810 811 return error; 812 } 813 814 /* 815 * A callback for each_writable_segment() to write out the segment's 816 * program header entry. 817 */ 818 static void 819 cb_put_phdr(entry, closure) 820 vm_map_entry_t entry; 821 void *closure; 822 { 823 struct phdr_closure *phc = (struct phdr_closure *)closure; 824 Elf_Phdr *phdr = phc->phdr; 825 826 phc->offset = round_page(phc->offset); 827 828 phdr->p_type = PT_LOAD; 829 phdr->p_offset = phc->offset; 830 phdr->p_vaddr = entry->start; 831 phdr->p_paddr = 0; 832 phdr->p_filesz = phdr->p_memsz = entry->end - entry->start; 833 phdr->p_align = PAGE_SIZE; 834 phdr->p_flags = 0; 835 if (entry->protection & VM_PROT_READ) 836 phdr->p_flags |= PF_R; 837 if (entry->protection & VM_PROT_WRITE) 838 phdr->p_flags |= PF_W; 839 if (entry->protection & VM_PROT_EXECUTE) 840 phdr->p_flags |= PF_X; 841 842 phc->offset += phdr->p_filesz; 843 phc->phdr++; 844 } 845 846 /* 847 * A callback for each_writable_segment() to gather information about 848 * the number of segments and their total size. 849 */ 850 static void 851 cb_size_segment(entry, closure) 852 vm_map_entry_t entry; 853 void *closure; 854 { 855 struct sseg_closure *ssc = (struct sseg_closure *)closure; 856 857 ssc->count++; 858 ssc->size += entry->end - entry->start; 859 } 860 861 /* 862 * For each writable segment in the process's memory map, call the given 863 * function with a pointer to the map entry and some arbitrary 864 * caller-supplied data. 865 */ 866 static void 867 each_writable_segment(p, func, closure) 868 struct proc *p; 869 segment_callback func; 870 void *closure; 871 { 872 vm_map_t map = &p->p_vmspace->vm_map; 873 vm_map_entry_t entry; 874 875 for (entry = map->header.next; entry != &map->header; 876 entry = entry->next) { 877 vm_object_t obj; 878 879 if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) || 880 (entry->protection & (VM_PROT_READ|VM_PROT_WRITE)) != 881 (VM_PROT_READ|VM_PROT_WRITE)) 882 continue; 883 884 /* 885 ** Dont include memory segment in the coredump if 886 ** MAP_NOCORE is set in mmap(2) or MADV_NOCORE in 887 ** madvise(2). 888 */ 889 if (entry->eflags & MAP_ENTRY_NOCOREDUMP) 890 continue; 891 892 if ((obj = entry->object.vm_object) == NULL) 893 continue; 894 895 /* Find the deepest backing object. */ 896 while (obj->backing_object != NULL) 897 obj = obj->backing_object; 898 899 /* Ignore memory-mapped devices and such things. */ 900 if (obj->type != OBJT_DEFAULT && 901 obj->type != OBJT_SWAP && 902 obj->type != OBJT_VNODE) 903 continue; 904 905 (*func)(entry, closure); 906 } 907 } 908 909 /* 910 * Write the core file header to the file, including padding up to 911 * the page boundary. 912 */ 913 static int 914 elf_corehdr(td, vp, cred, numsegs, hdr, hdrsize) 915 struct thread *td; 916 struct vnode *vp; 917 struct ucred *cred; 918 int numsegs; 919 size_t hdrsize; 920 void *hdr; 921 { 922 struct { 923 prstatus_t status; 924 prfpregset_t fpregset; 925 prpsinfo_t psinfo; 926 } *tempdata; 927 struct proc *p = td->td_proc; 928 size_t off; 929 prstatus_t *status; 930 prfpregset_t *fpregset; 931 prpsinfo_t *psinfo; 932 933 tempdata = malloc(sizeof(*tempdata), M_TEMP, M_ZERO | M_WAITOK); 934 status = &tempdata->status; 935 fpregset = &tempdata->fpregset; 936 psinfo = &tempdata->psinfo; 937 938 /* Gather the information for the header. */ 939 status->pr_version = PRSTATUS_VERSION; 940 status->pr_statussz = sizeof(prstatus_t); 941 status->pr_gregsetsz = sizeof(gregset_t); 942 status->pr_fpregsetsz = sizeof(fpregset_t); 943 status->pr_osreldate = osreldate; 944 status->pr_cursig = p->p_sig; 945 status->pr_pid = p->p_pid; 946 fill_regs(td, &status->pr_reg); 947 948 fill_fpregs(td, fpregset); 949 950 psinfo->pr_version = PRPSINFO_VERSION; 951 psinfo->pr_psinfosz = sizeof(prpsinfo_t); 952 strncpy(psinfo->pr_fname, p->p_comm, sizeof(psinfo->pr_fname) - 1); 953 954 /* XXX - We don't fill in the command line arguments properly yet. */ 955 strncpy(psinfo->pr_psargs, p->p_comm, PRARGSZ); 956 957 /* Fill in the header. */ 958 bzero(hdr, hdrsize); 959 off = 0; 960 elf_puthdr(p, hdr, &off, status, fpregset, psinfo, numsegs); 961 962 free(tempdata, M_TEMP); 963 964 /* Write it to the core file. */ 965 return vn_rdwr_inchunks(UIO_WRITE, vp, hdr, hdrsize, (off_t)0, 966 UIO_SYSSPACE, IO_UNIT | IO_DIRECT, cred, NULL, td); /* XXXKSE */ 967 } 968 969 static void 970 elf_puthdr(struct proc *p, void *dst, size_t *off, const prstatus_t *status, 971 const prfpregset_t *fpregset, const prpsinfo_t *psinfo, int numsegs) 972 { 973 size_t ehoff; 974 size_t phoff; 975 size_t noteoff; 976 size_t notesz; 977 978 ehoff = *off; 979 *off += sizeof(Elf_Ehdr); 980 981 phoff = *off; 982 *off += (numsegs + 1) * sizeof(Elf_Phdr); 983 984 noteoff = *off; 985 elf_putnote(dst, off, "FreeBSD", NT_PRSTATUS, status, 986 sizeof *status); 987 elf_putnote(dst, off, "FreeBSD", NT_FPREGSET, fpregset, 988 sizeof *fpregset); 989 elf_putnote(dst, off, "FreeBSD", NT_PRPSINFO, psinfo, 990 sizeof *psinfo); 991 notesz = *off - noteoff; 992 993 /* Align up to a page boundary for the program segments. */ 994 *off = round_page(*off); 995 996 if (dst != NULL) { 997 Elf_Ehdr *ehdr; 998 Elf_Phdr *phdr; 999 struct phdr_closure phc; 1000 1001 /* 1002 * Fill in the ELF header. 1003 */ 1004 ehdr = (Elf_Ehdr *)((char *)dst + ehoff); 1005 ehdr->e_ident[EI_MAG0] = ELFMAG0; 1006 ehdr->e_ident[EI_MAG1] = ELFMAG1; 1007 ehdr->e_ident[EI_MAG2] = ELFMAG2; 1008 ehdr->e_ident[EI_MAG3] = ELFMAG3; 1009 ehdr->e_ident[EI_CLASS] = ELF_CLASS; 1010 ehdr->e_ident[EI_DATA] = ELF_DATA; 1011 ehdr->e_ident[EI_VERSION] = EV_CURRENT; 1012 ehdr->e_ident[EI_OSABI] = ELFOSABI_FREEBSD; 1013 ehdr->e_ident[EI_ABIVERSION] = 0; 1014 ehdr->e_ident[EI_PAD] = 0; 1015 ehdr->e_type = ET_CORE; 1016 ehdr->e_machine = ELF_ARCH; 1017 ehdr->e_version = EV_CURRENT; 1018 ehdr->e_entry = 0; 1019 ehdr->e_phoff = phoff; 1020 ehdr->e_flags = 0; 1021 ehdr->e_ehsize = sizeof(Elf_Ehdr); 1022 ehdr->e_phentsize = sizeof(Elf_Phdr); 1023 ehdr->e_phnum = numsegs + 1; 1024 ehdr->e_shentsize = sizeof(Elf_Shdr); 1025 ehdr->e_shnum = 0; 1026 ehdr->e_shstrndx = SHN_UNDEF; 1027 1028 /* 1029 * Fill in the program header entries. 1030 */ 1031 phdr = (Elf_Phdr *)((char *)dst + phoff); 1032 1033 /* The note segement. */ 1034 phdr->p_type = PT_NOTE; 1035 phdr->p_offset = noteoff; 1036 phdr->p_vaddr = 0; 1037 phdr->p_paddr = 0; 1038 phdr->p_filesz = notesz; 1039 phdr->p_memsz = 0; 1040 phdr->p_flags = 0; 1041 phdr->p_align = 0; 1042 phdr++; 1043 1044 /* All the writable segments from the program. */ 1045 phc.phdr = phdr; 1046 phc.offset = *off; 1047 each_writable_segment(p, cb_put_phdr, &phc); 1048 } 1049 } 1050 1051 static void 1052 elf_putnote(void *dst, size_t *off, const char *name, int type, 1053 const void *desc, size_t descsz) 1054 { 1055 Elf_Note note; 1056 1057 note.n_namesz = strlen(name) + 1; 1058 note.n_descsz = descsz; 1059 note.n_type = type; 1060 if (dst != NULL) 1061 bcopy(¬e, (char *)dst + *off, sizeof note); 1062 *off += sizeof note; 1063 if (dst != NULL) 1064 bcopy(name, (char *)dst + *off, note.n_namesz); 1065 *off += roundup2(note.n_namesz, sizeof(Elf_Size)); 1066 if (dst != NULL) 1067 bcopy(desc, (char *)dst + *off, note.n_descsz); 1068 *off += roundup2(note.n_descsz, sizeof(Elf_Size)); 1069 } 1070 1071 /* 1072 * Tell kern_execve.c about it, with a little help from the linker. 1073 */ 1074 static struct execsw elf_execsw = {exec_elf_imgact, "ELF"}; 1075 EXEC_SET(elf, elf_execsw); 1076