/* -*- mode: C++; c-basic-offset: 4; tab-width: 4 -*-
*
- * Copyright (c) 2006 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2006-2007 Apple Inc. All rights reserved.
*
* @APPLE_LICENSE_HEADER_START@
*
#include <mach-o/x86_64/reloc.h>
#include <vector>
+#include <set>
+#include <ext/hash_set>
#include "MachOFileAbstraction.hpp"
#include "Architectures.hpp"
typedef typename A::P::E E;
typedef typename A::P::uint_t pint_t;
+ class CStringEquals
+ {
+ public:
+ bool operator()(const char* left, const char* right) const { return (strcmp(left, right) == 0); }
+ };
+
+ typedef __gnu_cxx::hash_set<const char*, __gnu_cxx::hash<const char*>, CStringEquals> StringSet;
+
MachOChecker(const uint8_t* fileContent, uint32_t fileLength, const char* path);
void checkMachHeader();
void checkLoadCommands();
void checkSection(const macho_segment_command<P>* segCmd, const macho_section<P>* sect);
uint8_t loadCommandSizeMask();
+ void checkSymbolTable();
void checkIndirectSymbolTable();
void checkRelocations();
void checkExternalReloation(const macho_relocation_info<P>* reloc);
const char* fStringsEnd;
const macho_nlist<P>* fSymbols;
uint32_t fSymbolCount;
+ const macho_dysymtab_command<P>* fDynamicSymbolTable;
const uint32_t* fIndirectTable;
uint32_t fIndirectTableCount;
const macho_relocation_info<P>* fLocalRelocations;
uint32_t fLocalRelocationsCount;
const macho_relocation_info<P>* fExternalRelocations;
uint32_t fExternalRelocationsCount;
- pint_t fRelocBase;
bool fWriteableSegmentWithAddrOver4G;
const macho_segment_command<P>* fFirstSegment;
const macho_segment_command<P>* fFirstWritableSegment;
template <typename A>
MachOChecker<A>::MachOChecker(const uint8_t* fileContent, uint32_t fileLength, const char* path)
- : fHeader(NULL), fLength(fileLength), fStrings(NULL), fSymbols(NULL), fSymbolCount(0), fIndirectTableCount(0),
+ : fHeader(NULL), fLength(fileLength), fStrings(NULL), fSymbols(NULL), fSymbolCount(0), fDynamicSymbolTable(NULL), fIndirectTableCount(0),
fLocalRelocations(NULL), fLocalRelocationsCount(0), fExternalRelocations(NULL), fExternalRelocationsCount(0),
- fRelocBase(0), fWriteableSegmentWithAddrOver4G(false), fFirstSegment(NULL), fFirstWritableSegment(NULL)
+ fWriteableSegmentWithAddrOver4G(false), fFirstSegment(NULL), fFirstWritableSegment(NULL)
{
// sanity check
if ( ! validFile(fileContent) )
checkIndirectSymbolTable();
checkRelocations();
+
+ checkSymbolTable();
}
throw "sizeofcmds in mach_header is larger than file";
uint32_t flags = fHeader->flags();
- uint32_t invalidBits = MH_INCRLINK | MH_LAZY_INIT | 0xFFFC0000;
+ const uint32_t invalidBits = MH_INCRLINK | MH_LAZY_INIT | 0xFFE00000;
if ( flags & invalidBits )
throw "invalid bits in mach_header flags";
-
+ if ( (flags & MH_NO_REEXPORTED_DYLIBS) && (fHeader->filetype() != MH_DYLIB) )
+ throw "MH_NO_REEXPORTED_DYLIBS bit of mach_header flags only valid for dylibs";
}
template <typename A>
case LC_ID_DYLINKER:
case macho_routines_command<P>::CMD:
case LC_SUB_FRAMEWORK:
- case LC_SUB_UMBRELLA:
case LC_SUB_CLIENT:
case LC_TWOLEVEL_HINTS:
case LC_PREBIND_CKSUM:
case LC_LOAD_WEAK_DYLIB:
case LC_UUID:
+ case LC_REEXPORT_DYLIB:
+ case LC_SEGMENT_SPLIT_INFO:
+ break;
+ case LC_SUB_UMBRELLA:
+ case LC_SUB_LIBRARY:
+ if ( fHeader->flags() & MH_NO_REEXPORTED_DYLIBS )
+ 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";
break;
default:
throwf("load command #%d is an unknown kind 0x%X", i, cmd->cmd());
// cache interesting segments
if ( fFirstSegment == NULL )
fFirstSegment = segCmd;
- if ( (fFirstWritableSegment == NULL) && ((segCmd->initprot() & VM_PROT_WRITE) != 0) )
- fFirstWritableSegment = segCmd;
-
+ if ( (segCmd->initprot() & VM_PROT_WRITE) != 0 ) {
+ if ( fFirstWritableSegment == NULL )
+ fFirstWritableSegment = segCmd;
+ if ( segCmd->vmaddr() > 0x100000000ULL )
+ fWriteableSegmentWithAddrOver4G = true;
+ }
+
// check section ranges
const macho_section<P>* const sectionsStart = (macho_section<P>*)((char*)segCmd + sizeof(macho_segment_command<P>));
const macho_section<P>* const sectionsEnd = §ionsStart[segCmd->nsects()];
throwf("section %s vm address not within segment", sect->sectname());
if ( (sect->addr()+sect->size()) > endAddr )
throwf("section %s vm address not within segment", sect->sectname());
- if ( (sect->flags() &SECTION_TYPE) != S_ZEROFILL ) {
+ if ( ((sect->flags() &SECTION_TYPE) != S_ZEROFILL) && (segCmd->filesize() != 0) ) {
if ( sect->offset() < startOffset )
throwf("section %s file offset not within segment", sect->sectname());
if ( (sect->offset()+sect->size()) > endOffset )
}
}
- // check LC_SYMTAB and LC_DYSYMTAB
+ // check LC_SYMTAB, LC_DYSYMTAB, and LC_SEGMENT_SPLIT_INFO
cmd = cmds;
bool foundDynamicSymTab = false;
for (uint32_t i = 0; i < cmd_count; ++i) {
throw "symbol table not in __LINKEDIT";
if ( (symtab->symoff() + fSymbolCount*sizeof(macho_nlist<P>*)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
throw "symbol table end not in __LINKEDIT";
+ if ( (symtab->symoff() % sizeof(pint_t)) != 0 )
+ throw "symbol table start not pointer aligned";
fStrings = (char*)fHeader + symtab->stroff();
fStringsEnd = fStrings + symtab->strsize();
if ( symtab->stroff() < linkEditSegment->fileoff() )
throw "string pool not in __LINKEDIT";
if ( (symtab->stroff()+symtab->strsize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
throw "string pool extends beyond __LINKEDIT";
+ if ( (symtab->stroff() % 4) != 0 ) // work around until rdar://problem/4737991 is fixed
+ throw "string pool start not pointer aligned";
+ if ( (symtab->strsize() % sizeof(pint_t)) != 0 )
+ throw "string pool size not a multiple of pointer size";
}
break;
case LC_DYSYMTAB:
if ( isStaticExecutable )
throw "LC_DYSYMTAB should not be used in static executable";
foundDynamicSymTab = true;
- const macho_dysymtab_command<P>* dsymtab = (struct macho_dysymtab_command<P>*)cmd;
- fIndirectTable = (uint32_t*)((char*)fHeader + dsymtab->indirectsymoff());
- fIndirectTableCount = dsymtab->nindirectsyms();
- if ( dsymtab->indirectsymoff() < linkEditSegment->fileoff() )
- throw "indirect symbol table not in __LINKEDIT";
- if ( (dsymtab->indirectsymoff()+fIndirectTableCount*8) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
- throw "indirect symbol table not in __LINKEDIT";
- fLocalRelocationsCount = dsymtab->nlocrel();
+ fDynamicSymbolTable = (struct macho_dysymtab_command<P>*)cmd;
+ fIndirectTable = (uint32_t*)((char*)fHeader + fDynamicSymbolTable->indirectsymoff());
+ fIndirectTableCount = fDynamicSymbolTable->nindirectsyms();
+ if ( fIndirectTableCount != 0 ) {
+ if ( fDynamicSymbolTable->indirectsymoff() < linkEditSegment->fileoff() )
+ throw "indirect symbol table not in __LINKEDIT";
+ if ( (fDynamicSymbolTable->indirectsymoff()+fIndirectTableCount*8) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
+ throw "indirect symbol table not in __LINKEDIT";
+ if ( (fDynamicSymbolTable->indirectsymoff() % sizeof(pint_t)) != 0 )
+ throw "indirect symbol table not pointer aligned";
+ }
+ fLocalRelocationsCount = fDynamicSymbolTable->nlocrel();
if ( fLocalRelocationsCount != 0 ) {
- fLocalRelocations = (const macho_relocation_info<P>*)((char*)fHeader + dsymtab->locreloff());
- if ( dsymtab->locreloff() < linkEditSegment->fileoff() )
+ fLocalRelocations = (const macho_relocation_info<P>*)((char*)fHeader + fDynamicSymbolTable->locreloff());
+ if ( fDynamicSymbolTable->locreloff() < linkEditSegment->fileoff() )
throw "local relocations not in __LINKEDIT";
- if ( (
dsymtab->locreloff()+fLocalRelocationsCount*sizeof(macho_relocation_info<P>)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
+ if ( (
fDynamicSymbolTable->locreloff()+fLocalRelocationsCount*sizeof(macho_relocation_info<P>)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
throw "local relocations not in __LINKEDIT";
+ if ( (fDynamicSymbolTable->locreloff() % sizeof(pint_t)) != 0 )
+ throw "local relocations table not pointer aligned";
}
- fExternalRelocationsCount = dsymtab->nextrel();
+ fExternalRelocationsCount = fDynamicSymbolTable->nextrel();
if ( fExternalRelocationsCount != 0 ) {
- fExternalRelocations = (const macho_relocation_info<P>*)((char*)fHeader + dsymtab->extreloff());
- if ( dsymtab->extreloff() < linkEditSegment->fileoff() )
- throw "local relocations not in __LINKEDIT";
- if ( (dsymtab->extreloff()+fExternalRelocationsCount*sizeof(macho_relocation_info<P>)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
- throw "local relocations not in __LINKEDIT";
+ fExternalRelocations = (const macho_relocation_info<P>*)((char*)fHeader + fDynamicSymbolTable->extreloff());
+ if ( fDynamicSymbolTable->extreloff() < linkEditSegment->fileoff() )
+ throw "external relocations not in __LINKEDIT";
+ if ( (fDynamicSymbolTable->extreloff()+fExternalRelocationsCount*sizeof(macho_relocation_info<P>)) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
+ throw "external relocations not in __LINKEDIT";
+ if ( (fDynamicSymbolTable->extreloff() % sizeof(pint_t)) != 0 )
+ throw "external relocations table not pointer aligned";
}
}
break;
+ case LC_SEGMENT_SPLIT_INFO:
+ {
+ if ( isStaticExecutable )
+ throw "LC_SEGMENT_SPLIT_INFO should not be used in static executable";
+ const macho_linkedit_data_command<P>* info = (struct macho_linkedit_data_command<P>*)cmd;
+ if ( info->dataoff() < linkEditSegment->fileoff() )
+ throw "split seg info not in __LINKEDIT";
+ if ( (info->dataoff()+info->datasize()) > (linkEditSegment->fileoff()+linkEditSegment->filesize()) )
+ throw "split seg info not in __LINKEDIT";
+ if ( (info->dataoff() % sizeof(pint_t)) != 0 )
+ throw "split seg info table not pointer aligned";
+ if ( (info->datasize() % sizeof(pint_t)) != 0 )
+ throw "split seg info size not a multiple of pointer size";
+ }
+ break;
}
cmd = (const macho_load_command<P>*)(((uint8_t*)cmd)+cmd->cmdsize());
}
throw "missing dynamic symbol table";
if ( fStrings == NULL )
throw "missing symbol table";
-
- fRelocBase = this->relocBase();
-
+
}
template <typename A>
}
+template <typename A>
+void MachOChecker<A>::checkSymbolTable()
+{
+ // verify no duplicate external symbol names
+ if ( fDynamicSymbolTable != NULL ) {
+ StringSet externalNames;
+ const macho_nlist<P>* const exportedStart = &fSymbols[fDynamicSymbolTable->iextdefsym()];
+ const macho_nlist<P>* const exportedEnd = &exportedStart[fDynamicSymbolTable->nextdefsym()];
+ for(const macho_nlist<P>* p = exportedStart; p < exportedEnd; ++p) {
+ const char* symName = &fStrings[p->n_strx()];
+ if ( externalNames.find(symName) != externalNames.end() )
+ throwf("duplicate external symbol: %s", symName);
+ externalNames.insert(symName);
+ }
+ }
+}
+
template <>
ppc::P::uint_t MachOChecker<ppc>::relocBase()
template <>
void MachOChecker<ppc>::checkExternalReloation(const macho_relocation_info<P>* reloc)
{
- // FIX
+ if ( reloc->r_length() != 2 )
+ throw "bad external relocation length";
+ if ( reloc->r_type() != GENERIC_RELOC_VANILLA )
+ throw "unknown external relocation type";
+ if ( reloc->r_pcrel() != 0 )
+ throw "bad external relocation pc_rel";
+ if ( reloc->r_extern() == 0 )
+ throw "local relocation found with external relocations";
+ if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
+ throw "external relocation address not in writable segment";
+ // FIX: check r_symbol
}
template <>
if ( reloc->r_extern() == 0 )
throw "local relocation found with external relocations";
if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
- throw "local relocation address not in writable segment";
+ throw "external relocation address not in writable segment";
// FIX: check r_symbol
}
template <>
void MachOChecker<x86>::checkExternalReloation(const macho_relocation_info<P>* reloc)
{
- // FIX
+ if ( reloc->r_length() != 2 )
+ throw "bad external relocation length";
+ if ( reloc->r_type() != GENERIC_RELOC_VANILLA )
+ throw "unknown external relocation type";
+ if ( reloc->r_pcrel() != 0 )
+ throw "bad external relocation pc_rel";
+ if ( reloc->r_extern() == 0 )
+ throw "local relocation found with external relocations";
+ if ( ! this->addressInWritableSegment(reloc->r_address() + this->relocBase()) )
+ throw "external relocation address not in writable segment";
+ // FIX: check r_symbol
}
template <typename A>
void MachOChecker<A>::checkRelocations()
{
+ // external relocations should be sorted to minimize dyld symbol lookups
+ // therefore every reloc with the same r_symbolnum value should be contiguous
+ std::set<uint32_t> previouslySeenSymbolIndexes;
+ uint32_t lastSymbolIndex = 0xFFFFFFFF;
const macho_relocation_info<P>* const externRelocsEnd = &fExternalRelocations[fExternalRelocationsCount];
for (const macho_relocation_info<P>* reloc = fExternalRelocations; reloc < externRelocsEnd; ++reloc) {
this->checkExternalReloation(reloc);
+ if ( reloc->r_symbolnum() != lastSymbolIndex ) {
+ if ( previouslySeenSymbolIndexes.count(reloc->r_symbolnum()) != 0 )
+ throw "external relocations not sorted";
+ previouslySeenSymbolIndexes.insert(lastSymbolIndex);
+ lastSymbolIndex = reloc->r_symbolnum();
+ }
}
const macho_relocation_info<P>* const localRelocsEnd = &fLocalRelocations[fLocalRelocationsCount];
if ( mh->magic == OSSwapBigToHostInt32(FAT_MAGIC) ) {
const struct fat_header* fh = (struct fat_header*)p;
const struct fat_arch* archs = (struct fat_arch*)(p + sizeof(struct fat_header));
- for (unsigned long i=0; i < fh->nfat_arch; ++i) {
- if ( archs[i].cputype == CPU_TYPE_POWERPC ) {
- if ( MachOChecker<ppc>::validFile(p + archs[i].offset) )
- MachOChecker<ppc>::make(p + archs[i].offset, archs[i].size, path);
+ for (unsigned long i=0; i < OSSwapBigToHostInt32(fh->nfat_arch); ++i) {
+ size_t offset = OSSwapBigToHostInt32(archs[i].offset);
+ size_t size = OSSwapBigToHostInt32(archs[i].size);
+ unsigned int cputype = OSSwapBigToHostInt32(archs[i].cputype);
+
+ switch(cputype) {
+ case CPU_TYPE_POWERPC:
+ if ( MachOChecker<ppc>::validFile(p + offset) )
+ MachOChecker<ppc>::make(p + offset, size, path);
else
throw "in universal file, ppc slice does not contain ppc mach-o";
- }
- else if ( archs[i].cputype == CPU_TYPE_I386 ) {
- if ( MachOChecker<x86>::validFile(p + archs[i].offset) )
- MachOChecker<x86>::make(p + archs[i].offset, archs[i].size, path);
+ break;
+ case CPU_TYPE_I386:
+ if ( MachOChecker<x86>::validFile(p + offset) )
+ MachOChecker<x86>::make(p + offset, size, path);
else
throw "in universal file, i386 slice does not contain i386 mach-o";
- }
- else if ( archs[i].cputype == CPU_TYPE_POWERPC64 ) {
- if ( MachOChecker<ppc64>::validFile(p + archs[i].offset) )
- MachOChecker<ppc64>::make(p + archs[i].offset, archs[i].size, path);
+ break;
+ case CPU_TYPE_POWERPC64:
+ if ( MachOChecker<ppc64>::validFile(p + offset) )
+ MachOChecker<ppc64>::make(p + offset, size, path);
else
throw "in universal file, ppc64 slice does not contain ppc64 mach-o";
- }
- else if ( archs[i].cputype == CPU_TYPE_X86_64 ) {
- if ( MachOChecker<x86_64>::validFile(p + archs[i].offset) )
- MachOChecker<x86_64>::make(p + archs[i].offset, archs[i].size, path);
+ break;
+ case CPU_TYPE_X86_64:
+ if ( MachOChecker<x86_64>::validFile(p + offset) )
+ MachOChecker<x86_64>::make(p + offset, size, path);
else
throw "in universal file, x86_64 slice does not contain x86_64 mach-o";
- }
- else {
- throw "in universal file, unknown architecture slice";
+ break;
+ default:
+ throwf("in universal file, unknown architecture slice 0x%x\n", cputype);
}
}
}
projects / apple / ld64.git / blobdiff