]> git.saurik.com Git - apple/ld64.git/blob - src/other/machochecker.cpp
projects / apple / ld64.git / blob
? search:


97138e85d6e5c9258bcc7c5739210d6a628f5872
[apple/ld64.git] / src / other / machochecker.cpp
1 /* -*- mode: C++; c-basic-offset: 4; tab-width: 4 -*-
2 *
3 * Copyright (c) 2006-2010 Apple Inc. All rights reserved.
4 *
5 * @APPLE_LICENSE_HEADER_START@
6 *
7 * This file contains Original Code and/or Modifications of Original Code
8 * as defined in and that are subject to the Apple Public Source License
9 * Version 2.0 (the 'License'). You may not use this file except in
10 * compliance with the License. Please obtain a copy of the License at
11 * http://www.opensource.apple.com/apsl/ and read it before using this
12 * file.
13 *
14 * The Original Code and all software distributed under the License are
15 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
16 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
17 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
19 * Please see the License for the specific language governing rights and
20 * limitations under the License.
21 *
22 * @APPLE_LICENSE_HEADER_END@
23 */
24
25 #include <sys/types.h>
26 #include <sys/stat.h>
27 #include <sys/mman.h>
28 #include <stdarg.h>
29 #include <stdio.h>
30 #include <stdlib.h>
31 #include <fcntl.h>
32 #include <unistd.h>
33 #include <errno.h>
34
35 #include <vector>
36 #include <set>
37 #include <unordered_set>
38
39 #include "configure.h"
40
41 #include "MachOFileAbstraction.hpp"
42 #include "Architectures.hpp"
43
44
45 __attribute__((noreturn))
46 void throwf(const char* format, ...)
47 {
48 va_list list;
49 char* p;
50 va_start(list, format);
51 vasprintf(&p, format, list);
52 va_end(list);
53
54 const char* t = p;
55 throw t;
56 }
57
58 static uint64_t read_uleb128(const uint8_t*& p, const uint8_t* end)
59 {
60 uint64_t result = 0;
61 int bit = 0;
62 do {
63 if (p == end)
64 throwf("malformed uleb128");
65
66 uint64_t slice = *p & 0x7f;
67
68 if (bit >= 64 || slice << bit >> bit != slice)
69 throwf("uleb128 too big");
70 else {
71 result |= (slice << bit);
72 bit += 7;
73 }
74 }
75 while (*p++ & 0x80);
76 return result;
77 }
78
79
80 static int64_t read_sleb128(const uint8_t*& p, const uint8_t* end)
81 {
82 int64_t result = 0;
83 int bit = 0;
84 uint8_t byte;
85 do {
86 if (p == end)
87 throwf("malformed sleb128");
88 byte = *p++;
89 result |= ((byte & 0x7f) << bit);
90 bit += 7;
91 } while (byte & 0x80);
92 // sign extend negative numbers
93 if ( (byte & 0x40) != 0 )
94 result |= (-1LL) << bit;
95 return result;
96 }
97
98
99 template <typename A>
100 class MachOChecker
101 {
102 public:
103 static bool validFile(const uint8_t* fileContent);
104 static MachOChecker<A>* make(const uint8_t* fileContent, uint32_t fileLength, const char* path, const char* verifierDstRoot)
105 { return new MachOChecker<A>(fileContent, fileLength, path, verifierDstRoot); }
106 virtual ~MachOChecker() {}
107
108
109 private:
110 typedef typename A::P P;
111 typedef typename A::P::E E;
112 typedef typename A::P::uint_t pint_t;
113
114 // utility classes for using std::unordered_map with c-strings
115 struct CStringHash {
116 size_t operator()(const char* __s) const {
117 size_t __h = 0;
118 for ( ; *__s; ++__s)
119 __h = 5 * __h + *__s;
120 return __h;
121 };
122 };
123 struct CStringEquals
124 {
125 bool operator()(const char* left, const char* right) const { return (strcmp(left, right) == 0); }
126 };
127
128 typedef std::unordered_set<const char*, CStringHash, CStringEquals> StringSet;
129
130 MachOChecker(const uint8_t* fileContent, uint32_t fileLength, const char* path, const char* verifierDstRoot);
131 void checkMachHeader();
132 void checkLoadCommands();
133 void checkSection(const macho_segment_command<P>* segCmd, const macho_section<P>* sect);
134 uint8_t loadCommandSizeMask();
135 void checkSymbolTable();
136 void checkInitTerms();
137 void checkIndirectSymbolTable();
138 void checkRelocations();
139 void checkExternalReloation(const macho_relocation_info<P>* reloc);
140 void checkLocalReloation(const macho_relocation_info<P>* reloc);
141 void verify();
142 void verifyInstallName();
143 void verifyNoRpaths();
144 void verifyNoFlatLookups();
145
146 pint_t relocBase();
147 bool addressInWritableSegment(pint_t address);
148 bool hasTextRelocInRange(pint_t start, pint_t end);
149 pint_t segStartAddress(uint8_t segIndex);
150 bool addressIsRebaseSite(pint_t addr);
151 bool addressIsBindingSite(pint_t addr);
152 pint_t getInitialStackPointer(const macho_thread_command<P>*);
153 pint_t getEntryPoint(const macho_thread_command<P>*);
154 const char* archName();
155
156
157 const char* fPath;
158 const char* fDstRoot;
159 const macho_header<P>* fHeader;
160 uint32_t fLength;
161 const char* fInstallName;
162 const char* fStrings;
163 const char* fStringsEnd;
164 const macho_nlist<P>* fSymbols;
165 uint32_t fSymbolCount;
166 const macho_dysymtab_command<P>* fDynamicSymbolTable;
167 const uint32_t* fIndirectTable;
168 uint32_t fIndirectTableCount;
169 const macho_relocation_info<P>* fLocalRelocations;
170 uint32_t fLocalRelocationsCount;
171 const macho_relocation_info<P>* fExternalRelocations;
172 uint32_t fExternalRelocationsCount;
173 bool fWriteableSegmentWithAddrOver4G;
174 bool fSlidableImage;
175 bool fHasLC_RPATH;
176 const macho_segment_command<P>* fFirstSegment;
177 const macho_segment_command<P>* fFirstWritableSegment;
178 const macho_segment_command<P>* fTEXTSegment;
179 const macho_dyld_info_command<P>* fDyldInfo;
180 uint32_t fSectionCount;
181 std::vector<const macho_segment_command<P>*>fSegments;
182 };
183
184
185 template <>
186 bool MachOChecker<x86>::validFile(const uint8_t* fileContent)
187 {
188 const macho_header<P>* header = (const macho_header<P>*)fileContent;
189 if ( header->magic() != MH_MAGIC )
190 return false;
191 if ( header->cputype() != CPU_TYPE_I386 )
192 return false;
193 switch (header->filetype()) {
194 case MH_EXECUTE:
195 case MH_DYLIB:
196 case MH_BUNDLE:
197 case MH_DYLINKER:
198 return true;
199 }
200 return false;
201 }
202
203 template <>
204 bool MachOChecker<x86_64>::validFile(const uint8_t* fileContent)
205 {
206 const macho_header<P>* header = (const macho_header<P>*)fileContent;
207 if ( header->magic() != MH_MAGIC_64 )
208 return false;
209 if ( header->cputype() != CPU_TYPE_X86_64 )
210 return false;
211 switch (header->filetype()) {
212 case MH_EXECUTE:
213 case MH_DYLIB:
214 case MH_BUNDLE:
215 case MH_DYLINKER:
216 return true;
217 }
218 return false;
219 }
220
221 #if SUPPORT_ARCH_arm_any
222 template <>
223 bool MachOChecker<arm>::validFile(const uint8_t* fileContent)
224 {
225 const macho_header<P>* header = (const macho_header<P>*)fileContent;
226 if ( header->magic() != MH_MAGIC )
227 return false;
228 if ( header->cputype() != CPU_TYPE_ARM )
229 return false;
230 switch (header->filetype()) {
231 case MH_EXECUTE:
232 case MH_DYLIB:
233 case MH_BUNDLE:
234 case MH_DYLINKER:
235 return true;
236 }
237 return false;
238 }
239 #endif
240
241 #if SUPPORT_ARCH_arm64
242 template <>
243 bool MachOChecker<arm64>::validFile(const uint8_t* fileContent)
244 {
245 const macho_header<P>* header = (const macho_header<P>*)fileContent;
246 if ( header->magic() != MH_MAGIC_64 )
247 return false;
248 if ( header->cputype() != CPU_TYPE_ARM64 )
249 return false;
250 switch (header->filetype()) {
251 case MH_EXECUTE:
252 case MH_DYLIB:
253 case MH_BUNDLE:
254 case MH_DYLINKER:
255 return true;
256 }
257 return false;
258 }
259 #endif
260
261
262 template <> uint8_t MachOChecker<ppc>::loadCommandSizeMask() { return 0x03; }
263 template <> uint8_t MachOChecker<ppc64>::loadCommandSizeMask() { return 0x07; }
264 template <> uint8_t MachOChecker<x86>::loadCommandSizeMask() { return 0x03; }
265 template <> uint8_t MachOChecker<x86_64>::loadCommandSizeMask() { return 0x07; }
266 template <> uint8_t MachOChecker<arm>::loadCommandSizeMask() { return 0x03; }
267 template <> uint8_t MachOChecker<arm64>::loadCommandSizeMask() { return 0x07; }
268
269
270 template <>
271 x86::P::uint_t MachOChecker<x86>::getInitialStackPointer(const macho_thread_command<x86::P>* threadInfo)
272 {
273 return threadInfo->thread_register(7);
274 }
275
276 template <>
277 x86_64::P::uint_t MachOChecker<x86_64>::getInitialStackPointer(const macho_thread_command<x86_64::P>* threadInfo)
278 {
279 return threadInfo->thread_register(7);
280 }
281
282 template <>
283 arm::P::uint_t MachOChecker<arm>::getInitialStackPointer(const macho_thread_command<arm::P>* threadInfo)
284 {
285 return threadInfo->thread_register(13);
286 }
287
288 template <>
289 arm64::P::uint_t MachOChecker<arm64>::getInitialStackPointer(const macho_thread_command<arm64::P>* threadInfo)
290 {
291 throw "LC_UNIXTHREAD not supported for arm64";
292 }
293
294
295 template <>
296 ppc::P::uint_t MachOChecker<ppc>::getEntryPoint(const macho_thread_command<ppc::P>* threadInfo)
297 {
298 return threadInfo->thread_register(0);
299 }
300
301
302 template <>
303 x86::P::uint_t MachOChecker<x86>::getEntryPoint(const macho_thread_command<x86::P>* threadInfo)
304 {
305 return threadInfo->thread_register(10);
306 }
307
308 template <>
309 x86_64::P::uint_t MachOChecker<x86_64>::getEntryPoint(const macho_thread_command<x86_64::P>* threadInfo)
310 {
311 return threadInfo->thread_register(16);
312 }
313
314 template <>
315 arm::P::uint_t MachOChecker<arm>::getEntryPoint(const macho_thread_command<arm::P>* threadInfo)
316 {
317 return threadInfo->thread_register(15);
318 }
319
320 template <>
321 arm64::P::uint_t MachOChecker<arm64>::getEntryPoint(const macho_thread_command<arm64::P>* threadInfo)
322 {
323 throw "LC_UNIXTHREAD not supported for arm64";
324 }
325
326
327 template <typename A>
328 const char* MachOChecker<A>::archName()
329 {
330 switch ( fHeader->cputype() ) {
331 case CPU_TYPE_I386:
332 return "i386";
333 case CPU_TYPE_X86_64:
334 if ( fHeader->cpusubtype() == CPU_SUBTYPE_X86_64_H )
335 return "x86_64h";
336 else
337 return "x86_64";
338 case CPU_TYPE_ARM:
339 switch ( fHeader->cpusubtype() ) {
340 case CPU_SUBTYPE_ARM_V7:
341 return "armv7";
342 case CPU_SUBTYPE_ARM_V7S:
343 return "armv7s";
344 case CPU_SUBTYPE_ARM_V7K:
345 return "armv7k";
346 }
347 return "arm";
348 case CPU_TYPE_ARM64:
349 return "arm64";
350 }
351 return "unknown";
352 }
353
354
355 template <typename A>
356 MachOChecker<A>::MachOChecker(const uint8_t* fileContent, uint32_t fileLength, const char* path, const char* verifierDstRoot)
357 : fHeader(NULL), fLength(fileLength), fInstallName(NULL), fStrings(NULL), fSymbols(NULL), fSymbolCount(0), fDynamicSymbolTable(NULL), fIndirectTableCount(0),
358 fLocalRelocations(NULL), fLocalRelocationsCount(0), fExternalRelocations(NULL), fExternalRelocationsCount(0),
359 fWriteableSegmentWithAddrOver4G(false), fSlidableImage(false), fHasLC_RPATH(false), fFirstSegment(NULL), fFirstWritableSegment(NULL),
360 fTEXTSegment(NULL), fDyldInfo(NULL), fSectionCount(0)
361 {
362 // sanity check
363 if ( ! validFile(fileContent) )
364 throw "not a mach-o file that can be checked";
365
366 fPath = strdup(path);
367 fDstRoot = verifierDstRoot ? strdup(verifierDstRoot) : NULL;
368 fHeader = (const macho_header<P>*)fileContent;
369
370 // sanity check header
371 checkMachHeader();
372
373 // check load commands
374 checkLoadCommands();
375
376 checkIndirectSymbolTable();
377
378 checkRelocations();
379
380 checkSymbolTable();
381
382 checkInitTerms();
383
384 if ( verifierDstRoot != NULL )
385 verify();
386 }
387
388
389 template <typename A>
390 void MachOChecker<A>::checkMachHeader()
391 {
392 if ( (fHeader->sizeofcmds() + sizeof(macho_header<P>)) > fLength )
393 throw "sizeofcmds in mach_header is larger than file";
394
395 uint32_t flags = fHeader->flags();
396 const uint32_t invalidBits = MH_INCRLINK | MH_LAZY_INIT | 0xFC000000;
397 if ( flags & invalidBits )
398 throw "invalid bits in mach_header flags";
399 if ( (flags & MH_NO_REEXPORTED_DYLIBS) && (fHeader->filetype() != MH_DYLIB) )
400 throw "MH_NO_REEXPORTED_DYLIBS bit of mach_header flags only valid for dylibs";
401
402 switch ( fHeader->filetype() ) {
403 case MH_EXECUTE:
404 fSlidableImage = ( flags & MH_PIE );
405 break;
406 case MH_DYLIB:
407 case MH_BUNDLE:
408 fSlidableImage = true;
409 break;
410 default:
411 throw "not a mach-o file type supported by this tool";
412 }
413 }
414
415 template <typename A>
416 void MachOChecker<A>::checkLoadCommands()
417 {
418 // check that all load commands fit within the load command space file
419 const macho_encryption_info_command<P>* encryption_info = NULL;
420 const macho_thread_command<P>* threadInfo = NULL;
421 const macho_entry_point_command<P>* entryPoint = NULL;
422 const uint8_t* const endOfFile = (uint8_t*)fHeader + fLength;
423 const uint8_t* const endOfLoadCommands = (uint8_t*)fHeader + sizeof(macho_header<P>) + fHeader->sizeofcmds();
424 const uint32_t cmd_count = fHeader->ncmds();
425 const macho_load_command<P>* const cmds = (macho_load_command<P>*)((uint8_t*)fHeader + sizeof(macho_header<P>));
426 const macho_load_command<P>* cmd = cmds;
427 const macho_dylib_command<P>* dylibID;
428 for (uint32_t i = 0; i < cmd_count; ++i) {
429 uint32_t size = cmd->cmdsize();
430 if ( (size & this->loadCommandSizeMask()) != 0 )
431 throwf("load command #%d has a unaligned size", i);
432 const uint8_t* endOfCmd = ((uint8_t*)cmd)+cmd->cmdsize();
433 if ( endOfCmd > endOfLoadCommands )
434 throwf("load command #%d extends beyond the end of the load commands", i);
435 if ( endOfCmd > endOfFile )
436 throwf("load command #%d extends beyond the end of the file", i);
437 switch ( cmd->cmd() ) {
438 case macho_segment_command<P>::CMD:
439 case LC_SYMTAB:
440 case LC_DYSYMTAB:
441 case LC_LOAD_DYLIB:
442 case LC_ID_DYLINKER:
443 case LC_LOAD_DYLINKER:
444 case macho_routines_command<P>::CMD:
445 case LC_SUB_FRAMEWORK:
446 case LC_SUB_CLIENT:
447 case LC_TWOLEVEL_HINTS:
448 case LC_PREBIND_CKSUM:
449 case LC_LOAD_WEAK_DYLIB:
450 case LC_LAZY_LOAD_DYLIB:
451 case LC_UUID:
452 case LC_REEXPORT_DYLIB:
453 case LC_SEGMENT_SPLIT_INFO:
454 case LC_CODE_SIGNATURE:
455 case LC_LOAD_UPWARD_DYLIB:
456 case LC_VERSION_MIN_MACOSX:
457 case LC_VERSION_MIN_IPHONEOS:
458 case LC_FUNCTION_STARTS:
459 case LC_DYLD_ENVIRONMENT:
460 case LC_DATA_IN_CODE:
461 case LC_DYLIB_CODE_SIGN_DRS:
462 case LC_SOURCE_VERSION:
463 break;
464 case LC_RPATH:
465 fHasLC_RPATH = true;
466 break;
467 case LC_ID_DYLIB:
468 dylibID = (macho_dylib_command<P>*)cmd;
469 if ( dylibID->name_offset() > size )
470 throwf("malformed mach-o: LC_ID_DYLIB load command has offset (%u) outside its size (%u)", dylibID->name_offset(), size);
471 if ( (dylibID->name_offset() + strlen(dylibID->name()) + 1) > size )
472 throwf("malformed mach-o: LC_ID_DYLIB load command string extends beyond end of load command");
473 fInstallName = dylibID->name();
474 break;
475 case LC_DYLD_INFO:
476 case LC_DYLD_INFO_ONLY:
477 fDyldInfo = (macho_dyld_info_command<P>*)cmd;
478 break;
479 case LC_ENCRYPTION_INFO:
480 case LC_ENCRYPTION_INFO_64:
481 encryption_info = (macho_encryption_info_command<P>*)cmd;
482 break;
483 case LC_SUB_UMBRELLA:
484 case LC_SUB_LIBRARY:
485 if ( fHeader->flags() & MH_NO_REEXPORTED_DYLIBS )
486 throw "MH_NO_REEXPORTED_DYLIBS bit of mach_header flags should not be set in an image with LC_SUB_LIBRARY or LC_SUB_UMBRELLA";
487 break;
488 case LC_MAIN:
489 if ( fHeader->filetype() != MH_EXECUTE )
490 throw "LC_MAIN can only be used in MH_EXECUTE file types";
491 entryPoint = (macho_entry_point_command<P>*)cmd;
492 break;
493 case LC_UNIXTHREAD:
494 if ( fHeader->filetype() != MH_EXECUTE )
495 throw "LC_UNIXTHREAD can only be used in MH_EXECUTE file types";
496 threadInfo = (macho_thread_command<P>*)cmd;
497 break;
498 default:
499 throwf("load command #%d is an unknown kind 0x%X", i, cmd->cmd());
500 }
501 cmd = (const macho_load_command<P>*)endOfCmd;
502 }
503
504 // check segments
505 cmd = cmds;
506 std::vector<std::pair<pint_t, pint_t> > segmentAddressRanges;
507 std::vector<std::pair<pint_t, pint_t> > segmentFileOffsetRanges;
508 const macho_segment_command<P>* linkEditSegment = NULL;
509 const macho_segment_command<P>* stackSegment = NULL;
510 for (uint32_t i = 0; i < cmd_count; ++i) {
511 if ( cmd->cmd() == macho_segment_command<P>::CMD ) {
512 const macho_segment_command<P>* segCmd = (const macho_segment_command<P>*)cmd;
513 fSegments.push_back(segCmd);
514 if ( segCmd->cmdsize() != (sizeof(macho_segment_command<P>) + segCmd->nsects() * sizeof(macho_section_content<P>)) )
515 throw "invalid segment load command size";
516
517 // see if this overlaps another segment address range
518 uint64_t startAddr = segCmd->vmaddr();
519 uint64_t endAddr = startAddr + segCmd->vmsize();
520 for (typename std::vector<std::pair<pint_t, pint_t> >::iterator it = segmentAddressRanges.begin(); it != segmentAddressRanges.end(); ++it) {
521 if ( it->first < startAddr ) {
522 if ( it->second > startAddr )
523 throw "overlapping segment vm addresses";
524 }
525 else if ( it->first > startAddr ) {
526 if ( it->first < endAddr )
527 throw "overlapping segment vm addresses";
528 }
529 else {
530 throw "overlapping segment vm addresses";
531 }
532 segmentAddressRanges.push_back(std::make_pair(startAddr, endAddr));
533 }
534 // see if this overlaps another segment file offset range
535 uint64_t startOffset = segCmd->fileoff();
536 uint64_t endOffset = startOffset + segCmd->filesize();
537 for (typename std::vector<std::pair<pint_t, pint_t> >::iterator it = segmentFileOffsetRanges.begin(); it != segmentFileOffsetRanges.end(); ++it) {
538 if ( it->first < startOffset ) {
539 if ( it->second > startOffset )
540 throw "overlapping segment file data";
541 }
542 else if ( it->first > startOffset ) {
543 if ( it->first < endOffset )
544 throw "overlapping segment file data";
545 }
546 else {
547 throw "overlapping segment file data";
548 }
549 segmentFileOffsetRanges.push_back(std::make_pair(startOffset, endOffset));
550 // check is within file bounds
551 if ( (startOffset > fLength) || (endOffset > fLength) )
552 throw "segment file data is past end of file";
553 }
554 // verify it fits in file
555 if ( startOffset > fLength )
556 throw "segment fileoff does not fit in file";
557 if ( endOffset > fLength )
558 throw "segment fileoff+filesize does not fit in file";
559
560 // record special segments
561 if ( strcmp(segCmd->segname(), "__LINKEDIT") == 0 )
562 linkEditSegment = segCmd;
563 else if ( strcmp(segCmd->segname(), "__UNIXSTACK") == 0 )
564 stackSegment = segCmd;
565
566 // cache interesting segments
567 if ( fFirstSegment == NULL )
568 fFirstSegment = segCmd;
569 if ( (fTEXTSegment == NULL) && (strcmp(segCmd->segname(), "__TEXT") == 0) )
570 fTEXTSegment = segCmd;
571 if ( (segCmd->initprot() & VM_PROT_WRITE) != 0 ) {
572 if ( fFirstWritableSegment == NULL )
573 fFirstWritableSegment = segCmd;
574 if ( segCmd->vmaddr() > 0x100000000ULL )
575 fWriteableSegmentWithAddrOver4G = true;
576 }
577
578 // check section ranges
579 const macho_section<P>* const sectionsStart = (macho_section<P>*)((char*)segCmd + sizeof(macho_segment_command<P>));
580 const macho_section<P>* const sectionsEnd = &sectionsStart[segCmd->nsects()];
581 for(const macho_section<P>* sect = sectionsStart; sect < sectionsEnd; ++sect) {
582 // check all non-zero sized sections are within segment
583 if ( sect->addr() < startAddr )
584 throwf("section %s vm address not within segment", sect->sectname());
585 if ( (sect->addr()+sect->size()) > endAddr )
586 throwf("section %s vm address not within segment", sect->sectname());
587 if ( ((sect->flags() & SECTION_TYPE) != S_ZEROFILL)
588 && ((sect->flags() & SECTION_TYPE) != S_THREAD_LOCAL_ZEROFILL)
589 && (segCmd->filesize() != 0)
590 && (sect->size() != 0) ) {
591 if ( sect->offset() < startOffset )
592 throwf("section %s file offset not within segment", sect->sectname());
593 if ( (sect->offset()+sect->size()) > endOffset )
594 throwf("section %s file offset not within segment", sect->sectname());
595 }
596 checkSection(segCmd, sect);
597 ++fSectionCount;
598 }
599 }
600 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
601 }
602
603 // verify there was a LINKEDIT segment
604 if ( linkEditSegment == NULL )
605 throw "no __LINKEDIT segment";
606
607 // verify there was an executable __TEXT segment and load commands are in it
608 if ( fTEXTSegment == NULL )
609 throw "no __TEXT segment";
610 if ( fTEXTSegment->initprot() != (VM_PROT_READ|VM_PROT_EXECUTE) )
611 throw "__TEXT segment does not have r-x init permissions";
612 //if ( fTEXTSegment->maxprot() != (VM_PROT_READ|VM_PROT_EXECUTE|VM_PROT_WRITE) )
613 // throw "__TEXT segment does not have rwx max permissions";
614 if ( fTEXTSegment->fileoff() != 0 )
615 throw "__TEXT segment does not start at mach_header";
616 if ( fTEXTSegment->filesize() < (sizeof(macho_header<P>)+fHeader->sizeofcmds()) )
617 throw "__TEXT segment smaller than load commands";
618
619 // verify if custom stack used, that stack is in __UNIXSTACK segment
620 if ( threadInfo != NULL ) {
621 pint_t initialSP = getInitialStackPointer(threadInfo);
622 if ( initialSP != 0 ) {
623 if ( stackSegment == NULL )
624 throw "LC_UNIXTHREAD specifics custom initial stack pointer, but no __UNIXSTACK segment";
625 if ( (initialSP < stackSegment->vmaddr()) || (initialSP > (stackSegment->vmaddr()+stackSegment->vmsize())) )
626 throw "LC_UNIXTHREAD specifics custom initial stack pointer which does not point into __UNIXSTACK segment";
627 }
628 }
629
630 // verify __UNIXSTACK is zero fill
631 if ( stackSegment != NULL ) {
632 if ( (stackSegment->filesize() != 0) || (stackSegment->fileoff() != 0) )
633 throw "__UNIXSTACK is not a zero-fill segment";
634 if ( stackSegment->vmsize() < 4096 )
635 throw "__UNIXSTACK segment is too small";
636 }
637
638 // verify entry point is in __TEXT segment
639 if ( threadInfo != NULL ) {
640 pint_t initialPC = getEntryPoint(threadInfo);
641 if ( (initialPC < fTEXTSegment->vmaddr()) || (initialPC >= (fTEXTSegment->vmaddr()+fTEXTSegment->vmsize())) )
642 throwf("entry point 0x%0llX is outside __TEXT segment", (long long)initialPC);
643 }
644 else if ( entryPoint != NULL ) {
645 pint_t initialOffset = entryPoint->entryoff();
646 if ( (initialOffset < fTEXTSegment->fileoff()) || (initialOffset >= (fTEXTSegment->fileoff()+fTEXTSegment->filesize())) )
647 throwf("entry point 0x%0llX is outside __TEXT segment", (long long)initialOffset);
648 }
649
650 // checks for executables
651 bool isStaticExecutable = false;
652 if ( fHeader->filetype() == MH_EXECUTE ) {
653 isStaticExecutable = true;
654 cmd = cmds;
655 for (uint32_t i = 0; i < cmd_count; ++i) {
656 switch ( cmd->cmd() ) {
657 case LC_LOAD_DYLINKER:
658 // the existence of a dyld load command makes a executable dynamic
659 isStaticExecutable = false;
660 break;
661 }
662 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
663 }
664 if ( isStaticExecutable ) {
665 if ( (fHeader->flags() != MH_NOUNDEFS) && (fHeader->flags() != (MH_NOUNDEFS|MH_PIE)) )
666 throw "invalid bits in mach_header flags for static executable";
667 }
668 }
669
670 // verify encryption info
671 if ( encryption_info != NULL ) {
672 switch ( fHeader->filetype() ) {
673 case MH_EXECUTE: case MH_DYLIB: case MH_BUNDLE:
674 break; // okay
675 default:
676 throw "LC_ENCRYPTION_INFO load command is not allowed in this file type";
677 }
678 if ( encryption_info->cryptoff() < (sizeof(macho_header<P>) + fHeader->sizeofcmds()) )
679 throw "LC_ENCRYPTION_INFO load command has cryptoff covers some load commands";
680 if ( (encryption_info->cryptoff() % 4096) != 0 )
681 throw "LC_ENCRYPTION_INFO load command has cryptoff which is not page aligned";
682 if ( (encryption_info->cryptsize() % 4096) != 0 )
683 throw "LC_ENCRYPTION_INFO load command has cryptsize which is not page sized";
684 for (typename std::vector<std::pair<pint_t, pint_t> >::iterator it = segmentFileOffsetRanges.begin();
685 it != segmentFileOffsetRanges.end(); ++it) {
686 if ( (it->first <= encryption_info->cryptoff()) && (encryption_info->cryptoff() < it->second) ) {
687 if ( (encryption_info->cryptoff() + encryption_info->cryptsize()) > it->second )
688 throw "LC_ENCRYPTION_INFO load command is not contained within one segment";
689 }
690 }
691 }
692
693 // verify dylib has LC_ID_DYLIB
694 if ( fHeader->filetype() == MH_DYLIB ) {
695 if ( fInstallName == NULL )
696 throw "MH_DYLIB missing LC_ID_DYLIB";
697 }
698 else {
699 if ( fInstallName != NULL )
700 throw "LC_ID_DYLIB found but file type is not MH_DYLIB";
701 }
702
703 // check LC_SYMTAB, LC_DYSYMTAB, and LC_SEGMENT_SPLIT_INFO
704 cmd = cmds;
705 bool foundDynamicSymTab = false;
706 for (uint32_t i = 0; i < cmd_count; ++i) {
707 switch ( cmd->cmd() ) {
708 case LC_SYMTAB:
709 {
710 const macho_symtab_command<P>* symtab = (macho_symtab_command<P>*)cmd;
711 fSymbolCount = symtab->nsyms();
712 fSymbols = (const macho_nlist<P>*)((char*)fHeader + symtab->symoff());
713 if ( symtab->symoff() < linkEditSegment->fileoff() )
714 throw "symbol table not in __LINKEDIT";
715 if ( (symtab->symoff() + fSymbolCount*sizeof(macho_nlist<P>*)) > symtab->stroff() )
716 throw "symbol table overlaps string pool";
717 if ( (symtab->symoff() % sizeof(pint_t)) != 0 )
718 throw "symbol table start not pointer aligned";
719 fStrings = (char*)fHeader + symtab->stroff();
720 fStringsEnd = fStrings + symtab->strsize();
721 if ( symtab->stroff() < linkEditSegment->fileoff() )
722 throw "string pool not in __LINKEDIT";
723 if ( (symtab->stroff()+symtab->strsize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
724 throw "string pool extends beyond __LINKEDIT";
725 if ( (symtab->stroff() % 4) != 0 ) // work around until rdar://problem/4737991 is fixed
726 throw "string pool start not pointer aligned";
727 }
728 break;
729 case LC_DYSYMTAB:
730 {
731 if ( isStaticExecutable &&! fSlidableImage )
732 throw "LC_DYSYMTAB should not be used in static executable";
733 foundDynamicSymTab = true;
734 fDynamicSymbolTable = (macho_dysymtab_command<P>*)cmd;
735 fIndirectTable = (uint32_t*)((char*)fHeader + fDynamicSymbolTable->indirectsymoff());
736 fIndirectTableCount = fDynamicSymbolTable->nindirectsyms();
737 if ( fIndirectTableCount != 0 ) {
738 if ( fDynamicSymbolTable->indirectsymoff() < linkEditSegment->fileoff() )
739 throw "indirect symbol table not in __LINKEDIT";
740 if ( (fDynamicSymbolTable->indirectsymoff()+fIndirectTableCount*8) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
741 throw "indirect symbol table not in __LINKEDIT";
742 if ( (fDynamicSymbolTable->indirectsymoff() % sizeof(pint_t)) != 0 )
743 throw "indirect symbol table not pointer aligned";
744 }
745 fLocalRelocationsCount = fDynamicSymbolTable->nlocrel();
746 if ( fLocalRelocationsCount != 0 ) {
747 fLocalRelocations = (const macho_relocation_info<P>*)((char*)fHeader + fDynamicSymbolTable->locreloff());
748 if ( fDynamicSymbolTable->locreloff() < linkEditSegment->fileoff() )
749 throw "local relocations not in __LINKEDIT";
750 if ( (fDynamicSymbolTable->locreloff()+fLocalRelocationsCount*sizeof(macho_relocation_info<P>)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
751 throw "local relocations not in __LINKEDIT";
752 if ( (fDynamicSymbolTable->locreloff() % sizeof(pint_t)) != 0 )
753 throw "local relocations table not pointer aligned";
754 }
755 fExternalRelocationsCount = fDynamicSymbolTable->nextrel();
756 if ( fExternalRelocationsCount != 0 ) {
757 fExternalRelocations = (const macho_relocation_info<P>*)((char*)fHeader + fDynamicSymbolTable->extreloff());
758 if ( fDynamicSymbolTable->extreloff() < linkEditSegment->fileoff() )
759 throw "external relocations not in __LINKEDIT";
760 if ( (fDynamicSymbolTable->extreloff()+fExternalRelocationsCount*sizeof(macho_relocation_info<P>)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
761 throw "external relocations not in __LINKEDIT";
762 if ( (fDynamicSymbolTable->extreloff() % sizeof(pint_t)) != 0 )
763 throw "external relocations table not pointer aligned";
764 }
765 }
766 break;
767 case LC_SEGMENT_SPLIT_INFO:
768 {
769 if ( isStaticExecutable )
770 throw "LC_SEGMENT_SPLIT_INFO should not be used in static executable";
771 const macho_linkedit_data_command<P>* info = (macho_linkedit_data_command<P>*)cmd;
772 if ( info->dataoff() < linkEditSegment->fileoff() )
773 throw "split seg info not in __LINKEDIT";
774 if ( (info->dataoff()+info->datasize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
775 throw "split seg info not in __LINKEDIT";
776 if ( (info->dataoff() % sizeof(pint_t)) != 0 )
777 throw "split seg info table not pointer aligned";
778 if ( (info->datasize() % sizeof(pint_t)) != 0 )
779 throw "split seg info size not a multiple of pointer size";
780 }
781 break;
782 case LC_FUNCTION_STARTS:
783 {
784 const macho_linkedit_data_command<P>* info = (macho_linkedit_data_command<P>*)cmd;
785 if ( info->dataoff() < linkEditSegment->fileoff() )
786 throw "function starts data not in __LINKEDIT";
787 if ( (info->dataoff()+info->datasize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
788 throw "function starts data not in __LINKEDIT";
789 if ( (info->dataoff() % sizeof(pint_t)) != 0 )
790 throw "function starts data table not pointer aligned";
791 if ( (info->datasize() % sizeof(pint_t)) != 0 )
792 throw "function starts data size not a multiple of pointer size";
793 }
794 break;
795 case LC_DATA_IN_CODE:
796 {
797 const macho_linkedit_data_command<P>* info = (macho_linkedit_data_command<P>*)cmd;
798 if ( info->dataoff() < linkEditSegment->fileoff() )
799 throw "data-in-code data not in __LINKEDIT";
800 if ( (info->dataoff()+info->datasize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
801 throw "data-in-code data not in __LINKEDIT";
802 if ( (info->dataoff() % sizeof(pint_t)) != 0 )
803 throw "data-in-code data table not pointer aligned";
804 if ( (info->datasize() % sizeof(pint_t)) != 0 )
805 throw "data-in-code data size not a multiple of pointer size";
806 }
807 break;
808 case LC_DYLIB_CODE_SIGN_DRS:
809 {
810 const macho_linkedit_data_command<P>* info = (macho_linkedit_data_command<P>*)cmd;
811 if ( info->dataoff() < linkEditSegment->fileoff() )
812 throw "dependent dylib DR data not in __LINKEDIT";
813 if ( (info->dataoff()+info->datasize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
814 throw "dependent dylib DR data not in __LINKEDIT";
815 if ( (info->dataoff() % sizeof(pint_t)) != 0 )
816 throw "dependent dylib DR data table not pointer aligned";
817 if ( (info->datasize() % sizeof(pint_t)) != 0 )
818 throw "dependent dylib DR data size not a multiple of pointer size";
819 }
820 break;
821 }
822 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
823 }
824 if ( !isStaticExecutable && !foundDynamicSymTab )
825 throw "missing dynamic symbol table";
826 if ( fStrings == NULL )
827 throw "missing symbol table";
828
829 }
830
831 template <typename A>
832 void MachOChecker<A>::checkSection(const macho_segment_command<P>* segCmd, const macho_section<P>* sect)
833 {
834 uint8_t sectionType = (sect->flags() & SECTION_TYPE);
835 if ( sectionType == S_ZEROFILL ) {
836 if ( sect->offset() != 0 )
837 throwf("section offset should be zero for zero-fill section %s", sect->sectname());
838 }
839
840 // check section's segment name matches segment
841 // if ( strncmp(sect->segname(), segCmd->segname(), 16) != 0 )
842 // throwf("section %s in segment %s has wrong segment name", sect->sectname(), segCmd->segname());
843
844 // more section tests here
845 }
846
847
848 template <typename A>
849 void MachOChecker<A>::verify()
850 {
851 bool sharedCacheCandidate = false;
852 if ( fInstallName != NULL ) {
853 if ( (strncmp(fInstallName, "/usr/lib/", 9) == 0) || (strncmp(fInstallName, "/System/Library/", 16) == 0) ) {
854 sharedCacheCandidate = true;
855 verifyInstallName();
856 verifyNoRpaths();
857 }
858 }
859 verifyNoFlatLookups();
860 }
861
862
863 template <typename A>
864 void MachOChecker<A>::verifyInstallName()
865 {
866 // Don't allow @rpath to be used as -install_name for OS dylibs
867 if ( strncmp(fInstallName, "@rpath/", 7) == 0 ) {
868 printf("os_dylib_rpath_install_name\tfatal\t-install_name uses @rpath in arch %s\n", archName());
869 }
870 else {
871 // Verify -install_name match actual path of dylib
872 const char* installPathWithinDstRoot = &fPath[strlen(fDstRoot)];
873 if ( strcmp(installPathWithinDstRoot, fInstallName) != 0 ) {
874 // see if install name is a symlink to actual file
875 bool symlinkToDylib = false;
876 char absDstPath[PATH_MAX];
877 if ( realpath(fDstRoot, absDstPath) != NULL ) {
878 char fullInstallNamePath[PATH_MAX];
879 strlcpy(fullInstallNamePath, absDstPath, PATH_MAX);
880 strlcat(fullInstallNamePath, fInstallName, PATH_MAX);
881 char absInstallNamePath[PATH_MAX];
882 if ( realpath(fullInstallNamePath, absInstallNamePath) != NULL ) {
883 char absFPath[PATH_MAX];
884 if ( realpath(fPath, absFPath) != NULL ) {
885 if ( strcmp(absInstallNamePath, absFPath) == 0 )
886 symlinkToDylib = true;
887 }
888 }
889 }
890 if ( !symlinkToDylib )
891 printf("os_dylib_bad_install_name\twarn\t-install_name does not match install location in arch %s\n", archName());
892 }
893 }
894
895 }
896
897 template <typename A>
898 void MachOChecker<A>::verifyNoRpaths()
899 {
900 // Don't allow OS dylibs to add rpaths
901 if ( fHasLC_RPATH ) {
902 printf("os_dylib_rpath\twarn\tcontains LC_RPATH load command in arch %s\n", archName());
903 }
904 }
905
906
907 template <typename A>
908 void MachOChecker<A>::verifyNoFlatLookups()
909 {
910 if ( (fHeader->flags() & MH_TWOLEVEL) == 0 ) {
911 printf("os_dylib_flat_namespace\twarn\tbuilt with -flat_namespace in arch %s\n", archName());
912 return;
913 }
914
915 if ( fDynamicSymbolTable != NULL ) {
916 const macho_nlist<P>* const undefinesStart = &fSymbols[fDynamicSymbolTable->iundefsym()];
917 const macho_nlist<P>* const undefinesEnd = &undefinesStart[fDynamicSymbolTable->nundefsym()];
918 for(const macho_nlist<P>* sym = undefinesStart; sym < undefinesEnd; ++sym) {
919 //printf("0x%04X %s\n", sym->n_desc(), &fStrings[sym->n_strx()]);
920 if ( GET_LIBRARY_ORDINAL(sym->n_desc()) == DYNAMIC_LOOKUP_ORDINAL ) {
921 const char* symName = &fStrings[sym->n_strx()];
922 printf("os_dylib_undefined_dynamic_lookup\twarn\tbuilt with -undefined dynamic_lookup for symbol %s in arch %s\n", symName, archName());
923 }
924 }
925 }
926 }
927
928 template <typename A>
929 void MachOChecker<A>::checkIndirectSymbolTable()
930 {
931 // static executables don't have indirect symbol table
932 if ( fDynamicSymbolTable == NULL )
933 return;
934 const macho_load_command<P>* const cmds = (macho_load_command<P>*)((uint8_t*)fHeader + sizeof(macho_header<P>));
935 const uint32_t cmd_count = fHeader->ncmds();
936 const macho_load_command<P>* cmd = cmds;
937 for (uint32_t i = 0; i < cmd_count; ++i) {
938 if ( cmd->cmd() == macho_segment_command<P>::CMD ) {
939 const macho_segment_command<P>* segCmd = (const macho_segment_command<P>*)cmd;
940 const macho_section<P>* const sectionsStart = (macho_section<P>*)((char*)segCmd + sizeof(macho_segment_command<P>));
941 const macho_section<P>* const sectionsEnd = &sectionsStart[segCmd->nsects()];
942 for(const macho_section<P>* sect = sectionsStart; sect < sectionsEnd; ++sect) {
943 // make sure all magic sections that use indirect symbol table fit within it
944 uint32_t start = 0;
945 uint32_t elementSize = 0;
946 switch ( sect->flags() & SECTION_TYPE ) {
947 case S_SYMBOL_STUBS:
948 elementSize = sect->reserved2();
949 start = sect->reserved1();
950 break;
951 case S_LAZY_SYMBOL_POINTERS:
952 case S_NON_LAZY_SYMBOL_POINTERS:
953 elementSize = sizeof(pint_t);
954 start = sect->reserved1();
955 break;
956 }
957 if ( elementSize != 0 ) {
958 uint32_t count = sect->size() / elementSize;
959 if ( (count*elementSize) != sect->size() )
960 throwf("%s section size is not an even multiple of element size", sect->sectname());
961 if ( (start+count) > fIndirectTableCount )
962 throwf("%s section references beyond end of indirect symbol table (%d > %d)", sect->sectname(), start+count, fIndirectTableCount );
963 }
964 }
965 }
966 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
967 }
968
969
970 if ( fDynamicSymbolTable->ilocalsym() != 0 )
971 throwf("start of local symbols (%d) not at start of symbol table", fDynamicSymbolTable->ilocalsym());
972
973 if ( fDynamicSymbolTable->ilocalsym() > fSymbolCount )
974 throwf("start of local symbols out of range (%d > %d) in indirect symbol table", fDynamicSymbolTable->ilocalsym(), fSymbolCount);
975 if ( fDynamicSymbolTable->ilocalsym() + fDynamicSymbolTable->nlocalsym() > fSymbolCount ) {
976 throwf("local symbols out of range (%d+%d > %d) in indirect symbol table",
977 fDynamicSymbolTable->ilocalsym(), fDynamicSymbolTable->nlocalsym(), fSymbolCount);
978 }
979
980 if ( fDynamicSymbolTable->iextdefsym() > fSymbolCount )
981 throwf("start of extern symbols out of range (%d > %d) in indirect symbol table", fDynamicSymbolTable->iextdefsym(), fSymbolCount);
982 if ( fDynamicSymbolTable->iextdefsym() != fDynamicSymbolTable->ilocalsym() + fDynamicSymbolTable->nlocalsym() ) {
983 throwf("start of extern symbols (%d) not contiguous to local symbols (%d+%d) in indirect symbol table",
984 fDynamicSymbolTable->iextdefsym(), fDynamicSymbolTable->ilocalsym(), fDynamicSymbolTable->nlocalsym() );
985 }
986 if ( fDynamicSymbolTable->iextdefsym() + fDynamicSymbolTable->nextdefsym() > fSymbolCount ) {
987 throwf("extern symbols out of range (%d+%d > %d) in indirect symbol table",
988 fDynamicSymbolTable->iextdefsym(), fDynamicSymbolTable->nextdefsym(), fSymbolCount);
989 }
990
991 if ( fDynamicSymbolTable->iundefsym() > fSymbolCount )
992 throwf("start of undefined symbols out of range (%d > %d) in indirect symbol table", fDynamicSymbolTable->iundefsym(), fSymbolCount);
993 if ( fDynamicSymbolTable->iundefsym() != fDynamicSymbolTable->iextdefsym() + fDynamicSymbolTable->nextdefsym() ) {
994 throwf("start of undefined symbols (%d) not contiguous to extern symbols (%d+%d) in indirect symbol table",
995 fDynamicSymbolTable->iundefsym(), fDynamicSymbolTable->iextdefsym(), fDynamicSymbolTable->nextdefsym());
996 }
997 if ( fDynamicSymbolTable->iundefsym() + fDynamicSymbolTable->nundefsym() > fSymbolCount ) {
998 throwf("undefined symbols out of range (%d+%d > %d) in indirect symbol table",
999 fDynamicSymbolTable->iundefsym(), fDynamicSymbolTable->nundefsym(), fSymbolCount);
1000 }
1001
1002 if ( fDynamicSymbolTable->iundefsym() + fDynamicSymbolTable->nundefsym() != fSymbolCount ) {
1003 throwf("end undefined symbols (%d+%d) not at end of all symbols (%d) in indirect symbol table",
1004 fDynamicSymbolTable->iundefsym(), fDynamicSymbolTable->nundefsym(), fSymbolCount );
1005 }
1006 }
1007
1008
1009
1010
1011 template <typename A>
1012 void MachOChecker<A>::checkSymbolTable()
1013 {
1014 // verify no duplicate external symbol names
1015 if ( fDynamicSymbolTable != NULL ) {
1016 StringSet externalNames;
1017 const macho_nlist<P>* const exportedStart = &fSymbols[fDynamicSymbolTable->iextdefsym()];
1018 const macho_nlist<P>* const exportedEnd = &exportedStart[fDynamicSymbolTable->nextdefsym()];
1019 int i = fDynamicSymbolTable->iextdefsym();
1020 for(const macho_nlist<P>* p = exportedStart; p < exportedEnd; ++p, ++i) {
1021 const char* symName = &fStrings[p->n_strx()];
1022 if ( symName > fStringsEnd )
1023 throw "string index out of range";
1024 //fprintf(stderr, "sym[%d] = %s\n", i, symName);
1025 if ( externalNames.find(symName) != externalNames.end() )
1026 throwf("duplicate external symbol: %s", symName);
1027 if ( (p->n_type() & N_EXT) == 0 )
1028 throwf("non-external symbol in external symbol range: %s", symName);
1029 // don't add N_INDR to externalNames because there is likely an undefine with same name
1030 if ( (p->n_type() & N_INDR) == 0 )
1031 externalNames.insert(symName);
1032 }
1033 // verify no undefines with same name as an external symbol
1034 const macho_nlist<P>* const undefinesStart = &fSymbols[fDynamicSymbolTable->iundefsym()];
1035 const macho_nlist<P>* const undefinesEnd = &undefinesStart[fDynamicSymbolTable->nundefsym()];
1036 for(const macho_nlist<P>* p = undefinesStart; p < undefinesEnd; ++p) {
1037 const char* symName = &fStrings[p->n_strx()];
1038 if ( symName > fStringsEnd )
1039 throw "string index out of range";
1040 if ( externalNames.find(symName) != externalNames.end() )
1041 throwf("undefine with same name as external symbol: %s", symName);
1042 }
1043 // verify all N_SECT values are valid
1044 for(const macho_nlist<P>* p = fSymbols; p < &fSymbols[fSymbolCount]; ++p) {
1045 uint8_t type = p->n_type();
1046 if ( ((type & N_STAB) == 0) && ((type & N_TYPE) == N_SECT) ) {
1047 if ( p->n_sect() > fSectionCount ) {
1048 throwf("symbol '%s' has n_sect=%d which is too large", &fStrings[p->n_strx()], p->n_sect());
1049 }
1050 }
1051 }
1052 }
1053 }
1054
1055
1056 template <typename A>
1057 void MachOChecker<A>::checkInitTerms()
1058 {
1059 const macho_load_command<P>* const cmds = (macho_load_command<P>*)((uint8_t*)fHeader + sizeof(macho_header<P>));
1060 const uint32_t cmd_count = fHeader->ncmds();
1061 const macho_load_command<P>* cmd = cmds;
1062 for (uint32_t i = 0; i < cmd_count; ++i) {
1063 if ( cmd->cmd() == macho_segment_command<P>::CMD ) {
1064 const macho_segment_command<P>* segCmd = (const macho_segment_command<P>*)cmd;
1065 const macho_section<P>* const sectionsStart = (macho_section<P>*)((char*)segCmd + sizeof(macho_segment_command<P>));
1066 const macho_section<P>* const sectionsEnd = &sectionsStart[segCmd->nsects()];
1067 for(const macho_section<P>* sect = sectionsStart; sect < sectionsEnd; ++sect) {
1068 // make sure all magic sections that use indirect symbol table fit within it
1069 uint32_t count;
1070 pint_t* arrayStart;
1071 pint_t* arrayEnd;
1072 const char* kind = "initializer";
1073 switch ( sect->flags() & SECTION_TYPE ) {
1074 case S_MOD_TERM_FUNC_POINTERS:
1075 kind = "terminator";
1076 // fall through
1077 case S_MOD_INIT_FUNC_POINTERS:
1078 count = sect->size() / sizeof(pint_t);
1079 if ( (count*sizeof(pint_t)) != sect->size() )
1080 throwf("%s section size is not an even multiple of element size", sect->sectname());
1081 if ( (sect->addr() % sizeof(pint_t)) != 0 )
1082 throwf("%s section size is not pointer size aligned", sect->sectname());
1083 // check each pointer in array points within TEXT
1084 arrayStart = (pint_t*)((char*)fHeader + sect->offset());
1085 arrayEnd = (pint_t*)((char*)fHeader + sect->offset() + sect->size());
1086 for (pint_t* p=arrayStart; p < arrayEnd; ++p) {
1087 pint_t pointer = P::getP(*p);
1088 if ( (pointer < fTEXTSegment->vmaddr()) || (pointer >= (fTEXTSegment->vmaddr()+fTEXTSegment->vmsize())) )
1089 throwf("%s 0x%08llX points outside __TEXT segment", kind, (long long)pointer);
1090 }
1091 // check each pointer in array will be rebased and not bound
1092 if ( fSlidableImage ) {
1093 pint_t sectionBeginAddr = sect->addr();
1094 pint_t sectionEndddr = sect->addr() + sect->size();
1095 for(pint_t addr = sectionBeginAddr; addr < sectionEndddr; addr += sizeof(pint_t)) {
1096 if ( addressIsBindingSite(addr) )
1097 throwf("%s at 0x%0llX has binding to external symbol", kind, (long long)addr);
1098 if ( ! addressIsRebaseSite(addr) )
1099 throwf("%s at 0x%0llX is not rebased", kind, (long long)addr);
1100 }
1101 }
1102 break;
1103 }
1104 }
1105 }
1106 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
1107 }
1108
1109 }
1110
1111
1112
1113 template <>
1114 x86::P::uint_t MachOChecker<x86>::relocBase()
1115 {
1116 if ( fHeader->flags() & MH_SPLIT_SEGS )
1117 return fFirstWritableSegment->vmaddr();
1118 else
1119 return fFirstSegment->vmaddr();
1120 }
1121
1122 template <>
1123 x86_64::P::uint_t MachOChecker<x86_64>::relocBase()
1124 {
1125 // check for split-seg
1126 return fFirstWritableSegment->vmaddr();
1127 }
1128
1129 template <>
1130 arm::P::uint_t MachOChecker<arm>::relocBase()
1131 {
1132 if ( fHeader->flags() & MH_SPLIT_SEGS )
1133 return fFirstWritableSegment->vmaddr();
1134 else
1135 return fFirstSegment->vmaddr();
1136 }
1137
1138 template <>
1139 arm64::P::uint_t MachOChecker<arm64>::relocBase()
1140 {
1141 return fFirstWritableSegment->vmaddr();
1142 }
1143
1144
1145 template <typename A>
1146 bool MachOChecker<A>::addressInWritableSegment(pint_t address)
1147 {
1148 const macho_load_command<P>* const cmds = (macho_load_command<P>*)((uint8_t*)fHeader + sizeof(macho_header<P>));
1149 const uint32_t cmd_count = fHeader->ncmds();
1150 const macho_load_command<P>* cmd = cmds;
1151 for (uint32_t i = 0; i < cmd_count; ++i) {
1152 if ( cmd->cmd() == macho_segment_command<P>::CMD ) {
1153 const macho_segment_command<P>* segCmd = (const macho_segment_command<P>*)cmd;
1154 if ( (address >= segCmd->vmaddr()) && (address < segCmd->vmaddr()+segCmd->vmsize()) ) {
1155 // if segment is writable, we are fine
1156 if ( (segCmd->initprot() & VM_PROT_WRITE) != 0 )
1157 return true;
1158 // could be a text reloc, make sure section bit is set
1159 const macho_section<P>* const sectionsStart = (macho_section<P>*)((char*)segCmd + sizeof(macho_segment_command<P>));
1160 const macho_section<P>* const sectionsEnd = &sectionsStart[segCmd->nsects()];
1161 for(const macho_section<P>* sect = sectionsStart; sect < sectionsEnd; ++sect) {
1162 if ( (sect->addr() <= address) && (address < (sect->addr()+sect->size())) ) {
1163 // found section for this address, if has relocs we are fine
1164 return ( (sect->flags() & (S_ATTR_EXT_RELOC|S_ATTR_LOC_RELOC)) != 0 );
1165 }
1166 }
1167 }
1168 }
1169 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
1170 }
1171 return false;
1172 }
1173
1174
1175
1176 template <>
1177 void MachOChecker<x86>::checkExternalReloation(const macho_relocation_info<P>* reloc)
1178 {
1179 if ( reloc->r_length() != 2 )
1180 throw "bad external relocation length";
1181 if ( reloc->r_type() != GENERIC_RELOC_VANILLA )
1182 throw "unknown external relocation type";
1183 if ( reloc->r_pcrel() != 0 )
1184 throw "bad external relocation pc_rel";
1185 if ( reloc->r_extern() == 0 )
1186 throw "local relocation found with external relocations";
1187 if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
1188 throw "external relocation address not in writable segment";
1189 // FIX: check r_symbol
1190 }
1191
1192
1193 template <>
1194 void MachOChecker<x86_64>::checkExternalReloation(const macho_relocation_info<P>* reloc)
1195 {
1196 if ( reloc->r_length() != 3 )
1197 throw "bad external relocation length";
1198 if ( reloc->r_type() != X86_64_RELOC_UNSIGNED )
1199 throw "unknown external relocation type";
1200 if ( reloc->r_pcrel() != 0 )
1201 throw "bad external relocation pc_rel";
1202 if ( reloc->r_extern() == 0 )
1203 throw "local relocation found with external relocations";
1204 if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
1205 throw "exernal relocation address not in writable segment";
1206 // FIX: check r_symbol
1207 }
1208
1209 #if SUPPORT_ARCH_arm_any
1210 template <>
1211 void MachOChecker<arm>::checkExternalReloation(const macho_relocation_info<P>* reloc)
1212 {
1213 if ( reloc->r_length() != 2 )
1214 throw "bad external relocation length";
1215 if ( reloc->r_type() != ARM_RELOC_VANILLA )
1216 throw "unknown external relocation type";
1217 if ( reloc->r_pcrel() != 0 )
1218 throw "bad external relocation pc_rel";
1219 if ( reloc->r_extern() == 0 )
1220 throw "local relocation found with external relocations";
1221 if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
1222 throw "external relocation address not in writable segment";
1223 // FIX: check r_symbol
1224 }
1225 #endif
1226
1227 #if SUPPORT_ARCH_arm64
1228 template <>
1229 void MachOChecker<arm64>::checkExternalReloation(const macho_relocation_info<P>* reloc)
1230 {
1231 throw "external relocations not used for arm64";
1232 }
1233 #endif
1234
1235
1236 template <>
1237 void MachOChecker<x86>::checkLocalReloation(const macho_relocation_info<P>* reloc)
1238 {
1239 // FIX
1240 }
1241
1242 template <>
1243 void MachOChecker<x86_64>::checkLocalReloation(const macho_relocation_info<P>* reloc)
1244 {
1245 if ( reloc->r_length() != 3 )
1246 throw "bad local relocation length";
1247 if ( reloc->r_type() != X86_64_RELOC_UNSIGNED )
1248 throw "unknown local relocation type";
1249 if ( reloc->r_pcrel() != 0 )
1250 throw "bad local relocation pc_rel";
1251 if ( reloc->r_extern() != 0 )
1252 throw "external relocation found with local relocations";
1253 if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
1254 throw "local relocation address not in writable segment";
1255 }
1256
1257 #if SUPPORT_ARCH_arm_any
1258 template <>
1259 void MachOChecker<arm>::checkLocalReloation(const macho_relocation_info<P>* reloc)
1260 {
1261 if ( reloc->r_address() & R_SCATTERED ) {
1262 // scattered
1263 const macho_scattered_relocation_info<P>* sreloc = (const macho_scattered_relocation_info<P>*)reloc;
1264 if ( sreloc->r_length() != 2 )
1265 throw "bad local scattered relocation length";
1266 if ( sreloc->r_type() != ARM_RELOC_PB_LA_PTR )
1267 throw "bad local scattered relocation type";
1268 }
1269 else {
1270 if ( reloc->r_length() != 2 )
1271 throw "bad local relocation length";
1272 if ( reloc->r_extern() != 0 )
1273 throw "external relocation found with local relocations";
1274 if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
1275 throw "local relocation address not in writable segment";
1276 }
1277 }
1278 #endif
1279
1280 #if SUPPORT_ARCH_arm64
1281 template <>
1282 void MachOChecker<arm64>::checkLocalReloation(const macho_relocation_info<P>* reloc)
1283 {
1284 throw "local relocations not used for arm64";
1285 }
1286 #endif
1287
1288
1289 template <typename A>
1290 void MachOChecker<A>::checkRelocations()
1291 {
1292 // external relocations should be sorted to minimize dyld symbol lookups
1293 // therefore every reloc with the same r_symbolnum value should be contiguous
1294 std::set<uint32_t> previouslySeenSymbolIndexes;
1295 uint32_t lastSymbolIndex = 0xFFFFFFFF;
1296 const macho_relocation_info<P>* const externRelocsEnd = &fExternalRelocations[fExternalRelocationsCount];
1297 for (const macho_relocation_info<P>* reloc = fExternalRelocations; reloc < externRelocsEnd; ++reloc) {
1298 this->checkExternalReloation(reloc);
1299 if ( reloc->r_symbolnum() != lastSymbolIndex ) {
1300 if ( previouslySeenSymbolIndexes.count(reloc->r_symbolnum()) != 0 )
1301 throw "external relocations not sorted";
1302 previouslySeenSymbolIndexes.insert(lastSymbolIndex);
1303 lastSymbolIndex = reloc->r_symbolnum();
1304 }
1305 }
1306
1307 const macho_relocation_info<P>* const localRelocsEnd = &fLocalRelocations[fLocalRelocationsCount];
1308 for (const macho_relocation_info<P>* reloc = fLocalRelocations; reloc < localRelocsEnd; ++reloc) {
1309 this->checkLocalReloation(reloc);
1310 }
1311
1312 // verify any section with S_ATTR_LOC_RELOC bits set actually has text relocs
1313 const macho_load_command<P>* const cmds = (macho_load_command<P>*)((uint8_t*)fHeader + sizeof(macho_header<P>));
1314 const uint32_t cmd_count = fHeader->ncmds();
1315 const macho_load_command<P>* cmd = cmds;
1316 for (uint32_t i = 0; i < cmd_count; ++i) {
1317 if ( cmd->cmd() == macho_segment_command<P>::CMD ) {
1318 const macho_segment_command<P>* segCmd = (const macho_segment_command<P>*)cmd;
1319 // if segment is writable, we are fine
1320 if ( (segCmd->initprot() & VM_PROT_WRITE) != 0 )
1321 continue;
1322 // look at sections that have text reloc bit set
1323 const macho_section<P>* const sectionsStart = (macho_section<P>*)((char*)segCmd + sizeof(macho_segment_command<P>));
1324 const macho_section<P>* const sectionsEnd = &sectionsStart[segCmd->nsects()];
1325 for(const macho_section<P>* sect = sectionsStart; sect < sectionsEnd; ++sect) {
1326 if ( (sect->flags() & S_ATTR_LOC_RELOC) != 0 ) {
1327 if ( ! hasTextRelocInRange(sect->addr(), sect->addr()+sect->size()) ) {
1328 throwf("section %s has attribute set that it has relocs, but it has none", sect->sectname());
1329 }
1330 }
1331 }
1332 }
1333 cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
1334 }
1335 }
1336
1337 template <typename A>
1338 typename A::P::uint_t MachOChecker<A>::segStartAddress(uint8_t segIndex)
1339 {
1340 if ( segIndex > fSegments.size() )
1341 throw "segment index out of range";
1342 return fSegments[segIndex]->vmaddr();
1343 }
1344
1345 template <typename A>
1346 bool MachOChecker<A>::hasTextRelocInRange(pint_t rangeStart, pint_t rangeEnd)
1347 {
1348 // look at local relocs
1349 const macho_relocation_info<P>* const localRelocsEnd = &fLocalRelocations[fLocalRelocationsCount];
1350 for (const macho_relocation_info<P>* reloc = fLocalRelocations; reloc < localRelocsEnd; ++reloc) {
1351 pint_t relocAddress = reloc->r_address() + this->relocBase();
1352 if ( (rangeStart <= relocAddress) && (relocAddress < rangeEnd) )
1353 return true;
1354 }
1355 // look rebase info
1356 if ( fDyldInfo != NULL ) {
1357 const uint8_t* p = (uint8_t*)fHeader + fDyldInfo->rebase_off();
1358 const uint8_t* end = &p[fDyldInfo->rebase_size()];
1359
1360 uint8_t type = 0;
1361 uint64_t segOffset = 0;
1362 uint32_t count;
1363 uint32_t skip;
1364 int segIndex;
1365 pint_t segStartAddr = 0;
1366 pint_t addr;
1367 bool done = false;
1368 while ( !done && (p < end) ) {
1369 uint8_t immediate = *p & REBASE_IMMEDIATE_MASK;
1370 uint8_t opcode = *p & REBASE_OPCODE_MASK;
1371 ++p;
1372 switch (opcode) {
1373 case REBASE_OPCODE_DONE:
1374 done = true;
1375 break;
1376 case REBASE_OPCODE_SET_TYPE_IMM:
1377 type = immediate;
1378 break;
1379 case REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB:
1380 segIndex = immediate;
1381 segStartAddr = segStartAddress(segIndex);
1382 segOffset = read_uleb128(p, end);
1383 break;
1384 case REBASE_OPCODE_ADD_ADDR_ULEB:
1385 segOffset += read_uleb128(p, end);
1386 break;
1387 case REBASE_OPCODE_ADD_ADDR_IMM_SCALED:
1388 segOffset += immediate*sizeof(pint_t);
1389 break;
1390 case REBASE_OPCODE_DO_REBASE_IMM_TIMES:
1391 for (int i=0; i < immediate; ++i) {
1392 addr = segStartAddr+segOffset;
1393 if ( (rangeStart <= addr) && (addr < rangeEnd) )
1394 return true;
1395 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1396 segOffset += sizeof(pint_t);
1397 }
1398 break;
1399 case REBASE_OPCODE_DO_REBASE_ULEB_TIMES:
1400 count = read_uleb128(p, end);
1401 for (uint32_t i=0; i < count; ++i) {
1402 addr = segStartAddr+segOffset;
1403 if ( (rangeStart <= addr) && (addr < rangeEnd) )
1404 return true;
1405 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1406 segOffset += sizeof(pint_t);
1407 }
1408 break;
1409 case REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB:
1410 addr = segStartAddr+segOffset;
1411 if ( (rangeStart <= addr) && (addr < rangeEnd) )
1412 return true;
1413 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1414 segOffset += read_uleb128(p, end) + sizeof(pint_t);
1415 break;
1416 case REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB:
1417 count = read_uleb128(p, end);
1418 skip = read_uleb128(p, end);
1419 for (uint32_t i=0; i < count; ++i) {
1420 addr = segStartAddr+segOffset;
1421 if ( (rangeStart <= addr) && (addr < rangeEnd) )
1422 return true;
1423 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1424 segOffset += skip + sizeof(pint_t);
1425 }
1426 break;
1427 default:
1428 throwf("bad rebase opcode %d", *p);
1429 }
1430 }
1431 }
1432 return false;
1433 }
1434
1435 template <typename A>
1436 bool MachOChecker<A>::addressIsRebaseSite(pint_t targetAddr)
1437 {
1438 // look at local relocs
1439 const macho_relocation_info<P>* const localRelocsEnd = &fLocalRelocations[fLocalRelocationsCount];
1440 for (const macho_relocation_info<P>* reloc = fLocalRelocations; reloc < localRelocsEnd; ++reloc) {
1441 pint_t relocAddress = reloc->r_address() + this->relocBase();
1442 if ( relocAddress == targetAddr )
1443 return true;
1444 }
1445 // look rebase info
1446 if ( fDyldInfo != NULL ) {
1447 const uint8_t* p = (uint8_t*)fHeader + fDyldInfo->rebase_off();
1448 const uint8_t* end = &p[fDyldInfo->rebase_size()];
1449
1450 uint8_t type = 0;
1451 uint64_t segOffset = 0;
1452 uint32_t count;
1453 uint32_t skip;
1454 int segIndex;
1455 pint_t segStartAddr = 0;
1456 pint_t addr;
1457 bool done = false;
1458 while ( !done && (p < end) ) {
1459 uint8_t immediate = *p & REBASE_IMMEDIATE_MASK;
1460 uint8_t opcode = *p & REBASE_OPCODE_MASK;
1461 ++p;
1462 switch (opcode) {
1463 case REBASE_OPCODE_DONE:
1464 done = true;
1465 break;
1466 case REBASE_OPCODE_SET_TYPE_IMM:
1467 type = immediate;
1468 break;
1469 case REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB:
1470 segIndex = immediate;
1471 segStartAddr = segStartAddress(segIndex);
1472 segOffset = read_uleb128(p, end);
1473 break;
1474 case REBASE_OPCODE_ADD_ADDR_ULEB:
1475 segOffset += read_uleb128(p, end);
1476 break;
1477 case REBASE_OPCODE_ADD_ADDR_IMM_SCALED:
1478 segOffset += immediate*sizeof(pint_t);
1479 break;
1480 case REBASE_OPCODE_DO_REBASE_IMM_TIMES:
1481 for (int i=0; i < immediate; ++i) {
1482 addr = segStartAddr+segOffset;
1483 if ( addr == targetAddr )
1484 return true;
1485 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1486 segOffset += sizeof(pint_t);
1487 }
1488 break;
1489 case REBASE_OPCODE_DO_REBASE_ULEB_TIMES:
1490 count = read_uleb128(p, end);
1491 for (uint32_t i=0; i < count; ++i) {
1492 addr = segStartAddr+segOffset;
1493 if ( addr == targetAddr )
1494 return true;
1495 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1496 segOffset += sizeof(pint_t);
1497 }
1498 break;
1499 case REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB:
1500 addr = segStartAddr+segOffset;
1501 if ( addr == targetAddr )
1502 return true;
1503 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1504 segOffset += read_uleb128(p, end) + sizeof(pint_t);
1505 break;
1506 case REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB:
1507 count = read_uleb128(p, end);
1508 skip = read_uleb128(p, end);
1509 for (uint32_t i=0; i < count; ++i) {
1510 addr = segStartAddr+segOffset;
1511 if ( addr == targetAddr )
1512 return true;
1513 //printf("%-7s %-16s 0x%08llX %s\n", segName, sectionName(segIndex, segStartAddr+segOffset), segStartAddr+segOffset, typeName);
1514 segOffset += skip + sizeof(pint_t);
1515 }
1516 break;
1517 default:
1518 throwf("bad rebase opcode %d", *p);
1519 }
1520 }
1521 }
1522 return false;
1523 }
1524
1525
1526 template <typename A>
1527 bool MachOChecker<A>::addressIsBindingSite(pint_t targetAddr)
1528 {
1529 // look at external relocs
1530 const macho_relocation_info<P>* const externRelocsEnd = &fExternalRelocations[fExternalRelocationsCount];
1531 for (const macho_relocation_info<P>* reloc = fExternalRelocations; reloc < externRelocsEnd; ++reloc) {
1532 pint_t relocAddress = reloc->r_address() + this->relocBase();
1533 if ( relocAddress == targetAddr )
1534 return true;
1535 }
1536 // look bind info
1537 if ( fDyldInfo != NULL ) {
1538 const uint8_t* p = (uint8_t*)fHeader + fDyldInfo->bind_off();
1539 const uint8_t* end = &p[fDyldInfo->bind_size()];
1540
1541 uint8_t type = 0;
1542 uint64_t segOffset = 0;
1543 uint32_t count;
1544 uint32_t skip;
1545 uint8_t flags;
1546 const char* symbolName = NULL;
1547 int libraryOrdinal = 0;
1548 int segIndex;
1549 int64_t addend = 0;
1550 pint_t segStartAddr = 0;
1551 pint_t addr;
1552 bool done = false;
1553 while ( !done && (p < end) ) {
1554 uint8_t immediate = *p & BIND_IMMEDIATE_MASK;
1555 uint8_t opcode = *p & BIND_OPCODE_MASK;
1556 ++p;
1557 switch (opcode) {
1558 case BIND_OPCODE_DONE:
1559 done = true;
1560 break;
1561 case BIND_OPCODE_SET_DYLIB_ORDINAL_IMM:
1562 libraryOrdinal = immediate;
1563 break;
1564 case BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB:
1565 libraryOrdinal = read_uleb128(p, end);
1566 break;
1567 case BIND_OPCODE_SET_DYLIB_SPECIAL_IMM:
1568 // the special ordinals are negative numbers
1569 if ( immediate == 0 )
1570 libraryOrdinal = 0;
1571 else {
1572 int8_t signExtended = BIND_OPCODE_MASK | immediate;
1573 libraryOrdinal = signExtended;
1574 }
1575 break;
1576 case BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM:
1577 symbolName = (char*)p;
1578 while (*p != '\0')
1579 ++p;
1580 ++p;
1581 break;
1582 case BIND_OPCODE_SET_TYPE_IMM:
1583 type = immediate;
1584 break;
1585 case BIND_OPCODE_SET_ADDEND_SLEB:
1586 addend = read_sleb128(p, end);
1587 break;
1588 case BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB:
1589 segIndex = immediate;
1590 segStartAddr = segStartAddress(segIndex);
1591 segOffset = read_uleb128(p, end);
1592 break;
1593 case BIND_OPCODE_ADD_ADDR_ULEB:
1594 segOffset += read_uleb128(p, end);
1595 break;
1596 case BIND_OPCODE_DO_BIND:
1597 if ( (segStartAddr+segOffset) == targetAddr )
1598 return true;
1599 segOffset += sizeof(pint_t);
1600 break;
1601 case BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB:
1602 if ( (segStartAddr+segOffset) == targetAddr )
1603 return true;
1604 segOffset += read_uleb128(p, end) + sizeof(pint_t);
1605 break;
1606 case BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED:
1607 if ( (segStartAddr+segOffset) == targetAddr )
1608 return true;
1609 segOffset += immediate*sizeof(pint_t) + sizeof(pint_t);
1610 break;
1611 case BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB:
1612 count = read_uleb128(p, end);
1613 skip = read_uleb128(p, end);
1614 for (uint32_t i=0; i < count; ++i) {
1615 if ( (segStartAddr+segOffset) == targetAddr )
1616 return true;
1617 segOffset += skip + sizeof(pint_t);
1618 }
1619 break;
1620 default:
1621 throwf("bad bind opcode %d", *p);
1622 }
1623 }
1624 }
1625 return false;
1626 }
1627
1628
1629 static void check(const char* path, const char* verifierDstRoot)
1630 {
1631 struct stat stat_buf;
1632
1633 try {
1634 int fd = ::open(path, O_RDONLY, 0);
1635 if ( fd == -1 )
1636 throw "cannot open file";
1637 if ( ::fstat(fd, &stat_buf) != 0 )
1638 throwf("fstat(%s) failed, errno=%d\n", path, errno);
1639 uint32_t length = stat_buf.st_size;
1640 uint8_t* p = (uint8_t*)::mmap(NULL, stat_buf.st_size, PROT_READ, MAP_FILE | MAP_PRIVATE, fd, 0);
1641 if ( p == ((uint8_t*)(-1)) )
1642 throw "cannot map file";
1643 ::close(fd);
1644 const mach_header* mh = (mach_header*)p;
1645 if ( mh->magic == OSSwapBigToHostInt32(FAT_MAGIC) ) {
1646 const struct fat_header* fh = (struct fat_header*)p;
1647 const struct fat_arch* archs = (struct fat_arch*)(p + sizeof(struct fat_header));
1648 for (unsigned long i=0; i < OSSwapBigToHostInt32(fh->nfat_arch); ++i) {
1649 size_t offset = OSSwapBigToHostInt32(archs[i].offset);
1650 size_t size = OSSwapBigToHostInt32(archs[i].size);
1651 unsigned int cputype = OSSwapBigToHostInt32(archs[i].cputype);
1652
1653 switch(cputype) {
1654 case CPU_TYPE_I386:
1655 if ( MachOChecker<x86>::validFile(p + offset) )
1656 MachOChecker<x86>::make(p + offset, size, path, verifierDstRoot);
1657 else
1658 throw "in universal file, i386 slice does not contain i386 mach-o";
1659 break;
1660 case CPU_TYPE_X86_64:
1661 if ( MachOChecker<x86_64>::validFile(p + offset) )
1662 MachOChecker<x86_64>::make(p + offset, size, path, verifierDstRoot);
1663 else
1664 throw "in universal file, x86_64 slice does not contain x86_64 mach-o";
1665 break;
1666 #if SUPPORT_ARCH_arm_any
1667 case CPU_TYPE_ARM:
1668 if ( MachOChecker<arm>::validFile(p + offset) )
1669 MachOChecker<arm>::make(p + offset, size, path, verifierDstRoot);
1670 else
1671 throw "in universal file, arm slice does not contain arm mach-o";
1672 break;
1673 #endif
1674 #if SUPPORT_ARCH_arm64
1675 case CPU_TYPE_ARM64:
1676 if ( MachOChecker<arm64>::validFile(p + offset) )
1677 MachOChecker<arm64>::make(p + offset, size, path, verifierDstRoot);
1678 else
1679 throw "in universal file, arm64 slice does not contain arm mach-o";
1680 break;
1681 #endif
1682 default:
1683 throwf("in universal file, unknown architecture slice 0x%x\n", cputype);
1684 }
1685 }
1686 }
1687 else if ( MachOChecker<x86>::validFile(p) ) {
1688 MachOChecker<x86>::make(p, length, path, verifierDstRoot);
1689 }
1690 else if ( MachOChecker<x86_64>::validFile(p) ) {
1691 MachOChecker<x86_64>::make(p, length, path, verifierDstRoot);
1692 }
1693 #if SUPPORT_ARCH_arm_any
1694 else if ( MachOChecker<arm>::validFile(p) ) {
1695 MachOChecker<arm>::make(p, length, path, verifierDstRoot);
1696 }
1697 #endif
1698 #if SUPPORT_ARCH_arm64
1699 else if ( MachOChecker<arm64>::validFile(p) ) {
1700 MachOChecker<arm64>::make(p, length, path, verifierDstRoot);
1701 }
1702 #endif
1703 else {
1704 throw "not a known file type";
1705 }
1706 }
1707 catch (const char* msg) {
1708 throwf("%s in %s", msg, path);
1709 }
1710 }
1711
1712
1713 int main(int argc, const char* argv[])
1714 {
1715 bool progress = false;
1716 const char* verifierDstRoot = NULL;
1717 int result = 0;
1718 for(int i=1; i < argc; ++i) {
1719 const char* arg = argv[i];
1720 if ( arg[0] == '-' ) {
1721 if ( strcmp(arg, "-progress") == 0 ) {
1722 progress = true;
1723 }
1724 else if ( strcmp(arg, "-verifier_dstroot") == 0 ) {
1725 verifierDstRoot = argv[++i];
1726 }
1727 else if ( strcmp(arg, "-verifier_error_list") == 0 ) {
1728 printf("os_dylib_rpath_install_name\tOS dylibs (those in /usr/lib/ or /System/Library/) must be built with -install_name that is an absolute path - not an @rpath\n");
1729 printf("os_dylib_bad_install_name\tOS dylibs (those in /usr/lib/ or /System/Library/) must be built with -install_name matching their file system location\n");
1730 printf("os_dylib_rpath\tOS dylibs should not contain LC_RPATH load commands (from -rpath linker option)\n");
1731 printf("os_dylib_flat_namespace\tOS dylibs should not be built with -flat_namespace\n");
1732 printf("os_dylib_undefined_dynamic_lookup\tOS dylibs should not be built with -undefined dynamic_lookup\n");
1733 printf("os_dylib_malformed\the mach-o is malformed\n");
1734 return 0;
1735 }
1736 else {
1737 throwf("unknown option: %s\n", arg);
1738 }
1739 }
1740 else {
1741 bool success = true;
1742 try {
1743 check(arg, verifierDstRoot);
1744 }
1745 catch (const char* msg) {
1746 if ( verifierDstRoot ) {
1747 printf("os_dylib_malformed\twarn\t%s\n", msg);
1748 }
1749 else {
1750 fprintf(stderr, "machocheck failed: %s\n", msg);
1751 result = 1;
1752 success = false;
1753 }
1754 }
1755 if ( success && progress )
1756 printf("ok: %s\n", arg);
1757 }
1758 }
1759
1760 return result;
1761 }
1762
mirror of ld64