Loading...
--- dyld/dyld-960/other-tools/dyld_shared_cache_util.cpp
+++ dyld/dyld-1330/other-tools/dyld_shared_cache_util.cpp
@@ -48,23 +48,34 @@
#include <vector>
#include <iostream>
#include <optional>
-
-//#include "ClosureBuilder.h"
+#include <ranges>
+
+#include "ClosureFileSystemPhysical.h"
#include "DyldSharedCache.h"
#include "JSONWriter.h"
#include "Trie.hpp"
#include "dsc_extractor.h"
#include "dyld_introspection.h"
+#include "OptimizerObjC.h"
#include "OptimizerSwift.h"
#include "PrebuiltLoader.h"
#include "DyldProcessConfig.h"
#include "DyldRuntimeState.h"
+#include "Utilities.h"
#include "objc-shared-cache.h"
#include "OptimizerObjC.h"
+#include "ObjCVisitor.h"
+#include "SymbolicatedImage.h"
+
using namespace dyld4;
+
+using other_tools::SymbolicatedImage;
+using mach_o::Header;
+using mach_o::Platform;
+using mach_o::Version32;
#if TARGET_OS_OSX
#define DSC_BUNDLE_REL_PATH "../../lib/dsc_extractor.bundle"
@@ -72,30 +83,6 @@
#define DSC_BUNDLE_REL_PATH "../lib/dsc_extractor.bundle"
#endif
-// In newer shared caches, relative method list selectors are offsets from the magic selector in libobjc
-static uint64_t getSharedCacheRelativeSelectorBaseVMAddress(const DyldSharedCache* dyldCache, bool rebased)
-{
- __block uint64_t sharedCacheRelativeSelectorBaseVMAddress = 0;
- if ( dyldCache->header.mappingOffset >= __offsetof(dyld_cache_header, symbolFileUUID) ) {
- constexpr std::string_view magicSelector = "\xf0\x9f\xa4\xaf";
- dyld3::MachOAnalyzer::VMAddrConverter vmAddrConverter = dyldCache->makeVMAddrConverter(rebased);
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
- if ( !strcmp(installName, "/usr/lib/libobjc.A.dylib") ) {
- const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
- uintptr_t slide = ma->getSlide();
- Diagnostics diag;
- ma->forEachObjCSelectorReference(diag, vmAddrConverter,
- ^(uint64_t selRefVMAddr, uint64_t selRefTargetVMAddr, bool &stop) {
- const char* selString = (const char*)selRefTargetVMAddr + slide;
- if ( selString == magicSelector ) {
- sharedCacheRelativeSelectorBaseVMAddress = selRefTargetVMAddr;
- }
- });
- }
- });
- }
- return sharedCacheRelativeSelectorBaseVMAddress;
-}
enum Mode {
modeNone,
@@ -104,24 +91,43 @@
modeDependencies,
modeSlideInfo,
modeVerboseSlideInfo,
+ modeFixupsInDylib,
modeTextInfo,
modeLinkEdit,
modeLocalSymbols,
modeJSONMap,
+ modeVerboseJSONMap,
modeJSONDependents,
modeSectionSizes,
modeStrings,
modeInfo,
+ modeTPROInfo,
+ modeStats,
modeSize,
modeObjCInfo,
modeObjCProtocols,
modeObjCImpCaches,
modeObjCClasses,
+ modeObjCClassLayout,
+ modeObjCClassMethodLists,
+ modeObjCClassHashTable,
modeObjCSelectors,
modeSwiftProtocolConformances,
+ modeSwiftPtrTables,
+ modeLookupVA,
modeExtract,
modePatchTable,
- modeListDylibsWithSection
+ modeDumpPatchTable,
+ modeRootsCost,
+ modeListDylibsWithSection,
+ modeDuplicates,
+ modeDuplicatesSummary,
+ modeMachHeaders,
+ modeCacheHeader,
+ modeDylibSymbols,
+ modeFunctionStarts,
+ modeFunctionVariants,
+ modePrewarmingData,
};
struct Options {
@@ -131,6 +137,9 @@
const char* segmentName;
const char* sectionName;
const char* rootPath = nullptr;
+ const char* fixupsInDylib;
+ const char* rootsCostOfDylib = nullptr;
+ const char* lookupVA = nullptr;
bool printUUIDs;
bool printVMAddrs;
bool printDylibVersions;
@@ -139,7 +148,50 @@
static void usage() {
- fprintf(stderr, "Usage: dyld_shared_cache_util -list [ -uuid ] [-vmaddr] | -dependents <dylib-path> [ -versions ] | -linkedit | -map | -slide_info | -verbose_slide_info | -info | -extract <dylib-dir> [ shared-cache-file ] \n");
+ fprintf(stderr, "Usage: dyld_shared_cache_util <command> [-fs-root] [-inode] [-versions] [-vmaddr] [shared-cache-file]\n"
+ " Commands:\n"
+ " -list [-uuid] [-vmaddr] list images\n"
+ " -dependents <dylib-path> list dependents of dylib\n"
+ " -linkedit print linkedit contents\n"
+ " -info print shared cache info\n"
+ " -stats print size stats\n"
+ " -slide_info print slide info\n"
+ " -verbose_slide_info print verbose slide info\n"
+ " -fixups_in_dylib <dylib-path> print fixups in dylib\n"
+ " -text_info print locations of TEXT segments\n"
+ " -local_symbols print local symbols and locations\n"
+ " -strings print C strings in images\n"
+ " -sections print summary of section sizes\n"
+ " -exports list exported symbols in images\n"
+ " -duplicate_exports list symbols exported by multiple images\n"
+ " -duplicate_exports_summary print number of duplicated symbols per image\n"
+ " -map print map of segment locations\n"
+ " -json-map print map of segment locations in JSON format\n"
+ " -verbose-json-map print map of segment and section locations in JSON format\n"
+ " -json-dependents print dependents in JSON format\n"
+ " -size print the size of each image\n"
+ " -objc-info print summary of ObjC content\n"
+ " -objc-protocols list ObjC protocols\n"
+ " -objc-imp-caches print contents of ObjC method caches\n"
+ " -objc-classes print ObjC class names and methods in JSON format\n"
+ " -objc-class-layout print size, start offset, and ivars of ObjC classes\n"
+ " -objc-class-method-lists print methods and properties of ObjC classes\n"
+ " -objc-class-hash-table print the contents of the ObjC class table\n"
+ " -objc-selectors print all ObjC selector names and locations in JSON format\n"
+ " -swift-proto print Swift protocol conformance table\n"
+ " -swift-ptrtables print Swift pointer tables\n"
+ " -lookup-va lookup range and symbols at the given virtual address\n"
+ " -extract <directory> extract images into the given directory\n"
+ " -patch_table print symbol patch table\n"
+ " -list_dylibs_with_section <seg> <sect> list images that contain the given section\n"
+ " -mach_headers summarize mach header of each image\n"
+ " -load_commands summarize load commands of each image\n"
+ " -cache_header print header of each shared cache file\n"
+ " -dylib_symbols print all symbol names and locations\n"
+ " -function_starts print address of beginning of each function\n"
+ " -function_variants print function variants fixups\n"
+ " -roots_cost <dylib-path> estimate the dirty memory cost of installing a dylib root\n"
+ " -prewarming_data print VM prewarming ranges\n");
}
static void checkMode(Mode mode) {
@@ -150,28 +202,159 @@
}
}
+struct SymbolicatedCache
+{
+ struct Range
+ {
+ uint64_t startAddr;
+ uint64_t endAddr;
+ std::optional<size_t> imageIndex;
+ std::string_view segmentName;
+ std::string_view sectName;
+
+ bool operator<(const Range& other) const
+ {
+ return startAddr < other.startAddr;
+ }
+ };
+
+ SymbolicatedCache(const DyldSharedCache* cache, bool isCacheOnDisk);
+
+ std::optional<size_t> findClosestRange(uint64_t addr) const;
+ SymbolicatedImage::SymbolLoc findClosestSymbol(uint64_t addr, const SymbolicatedImage*& image) const;
+
+ std::string symbolNameAt(uint64_t addr) const;
+
+ std::vector<Range> ranges;
+ std::vector<Image> machoImages;
+ std::vector<SymbolicatedImage> images;
+ uint64_t cacheBaseAddr;
+};
+
+SymbolicatedCache::SymbolicatedCache(const DyldSharedCache* cache, bool isCacheOnDisk)
+{
+ cacheBaseAddr = cache->unslidLoadAddress();
+
+ machoImages.reserve(cache->imagesCount());
+ images.reserve(cache->imagesCount());
+ cache->forEachImage(^(const Header* hdr, const char* installName) {
+ machoImages.emplace_back((void*)hdr, (size_t)-1, isCacheOnDisk ? Image::MappingKind::dyldLoadedPreFixups : Image::MappingKind::dyldLoadedPostFixups);
+ });
+
+ for ( const Image& image : machoImages )
+ images.emplace_back(image);
+
+ for ( size_t i = 0; i < images.size(); ++i ) {
+ const SymbolicatedImage& im = images[i];
+ im.image().header()->forEachSection(^(const Header::SectionInfo& sectInfo, bool &stop) {
+ if ( sectInfo.size == 0 )
+ return;
+ ranges.push_back({ .imageIndex = i, .startAddr = sectInfo.address, .endAddr = sectInfo.address + sectInfo.size, .segmentName = sectInfo.segmentName, .sectName = sectInfo.sectionName });
+ });
+ }
+
+ std::sort(ranges.begin(), ranges.end());
+ for ( size_t i = 1; i < ranges.size(); ++i ) {
+ if ( ranges[i-1].endAddr > ranges[i].startAddr ) {
+ assert(false && "overlapping image ranges");
+ }
+ }
+}
+
+std::string SymbolicatedCache::symbolNameAt(uint64_t addr) const
+{
+ const SymbolicatedImage* image = nullptr;
+ SymbolicatedImage::SymbolLoc loc = findClosestSymbol(addr, image);
+ if ( loc.name.empty() ) {
+ if ( image ) {
+ return std::string(image->image().header()->installName()) + "+" + json::hex(loc.inSymbolOffset);
+ }
+ return json::hex( addr );
+ }
+
+ std::string nameWithImage = std::string(image->image().header()->installName()) + "`" + std::string(loc.name);
+ if ( loc.inSymbolOffset != 0 )
+ return nameWithImage + "+" + json::hex(loc.inSymbolOffset);
+ return nameWithImage;
+}
+
+std::optional<size_t> SymbolicatedCache::findClosestRange(uint64_t addr) const
+{
+ auto it = std::lower_bound(ranges.begin(), ranges.end(), addr, [](const Range& range, uint64_t cmpAddr) -> bool {
+ return range.startAddr <= cmpAddr;
+ });
+ // lower_bound returns the range after the one we need
+ if ( (it != ranges.end()) && (it != ranges.begin()) ) {
+ --it;
+ } else {
+ it = ranges.begin();
+ }
+
+ if ( addr < it->startAddr || addr >= it->endAddr )
+ return std::nullopt;
+
+ return std::distance(ranges.begin(), it);
+}
+
+SymbolicatedImage::SymbolLoc SymbolicatedCache::findClosestSymbol(uint64_t addr, const SymbolicatedImage*& image) const
+{
+ SymbolicatedImage::SymbolLoc loc;
+ image = nullptr;
+ if ( ranges.empty() )
+ return loc;
+
+ std::optional<size_t> rangeIndex = findClosestRange(addr);
+ if ( !rangeIndex )
+ return loc;
+
+ const Range& range = ranges[*rangeIndex];
+ if ( range.imageIndex == std::nullopt )
+ return loc;
+
+ size_t imageIndex = *range.imageIndex;
+ assert(imageIndex < images.size());
+ //fprintf(stderr, "debug symbol lookup at offset: %llu, abs: 0x%llX, image: %s\n", runtimeOffset, addr, images[imageIndex].image().header()->installName());
+ image = &images[imageIndex];
+ loc = images[imageIndex].findClosestSymbol(addr);
+
+ if ( loc.name.empty() ) {
+ loc.inSymbolOffset = (uint32_t)(addr - image->prefLoadAddress());
+ }
+ return loc;
+}
+
+
struct SegmentInfo
{
- uint64_t vmAddr;
- uint64_t vmSize;
- const char* installName;
- const char* segName;
+ uint64_t vmAddr;
+ uint64_t vmSize;
+ const char* installName;
+ std::string_view segName;
};
-static void buildSegmentInfo(const DyldSharedCache* dyldCache, std::vector<SegmentInfo>& segInfos)
+static void sortSegmentInfo(std::vector<SegmentInfo>& segInfos)
{
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
- dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)mh;
- ma->forEachSegment(^(const dyld3::MachOAnalyzer::SegmentInfo& info, bool& stop) {
- // Note, we subtract 1 from the vmSize so that lower_bound doesn't include the end of the segment
- // as being a match for a given address.
- segInfos.push_back({info.vmAddr, info.vmSize - 1, installName, info.segName});
- });
- });
-
std::sort(segInfos.begin(), segInfos.end(), [](const SegmentInfo& l, const SegmentInfo& r) -> bool {
return l.vmAddr < r.vmAddr;
});
+}
+
+static void buildSegmentInfo(const Header* hdr, std::vector<SegmentInfo>& segInfos)
+{
+ const char* installName = hdr->installName();
+ hdr->forEachSegment(^(const Header::SegmentInfo& info, bool& stop) {
+ // Note, we subtract 1 from the vmSize so that lower_bound doesn't include the end of the segment
+ // as being a match for a given address.
+ segInfos.push_back({info.vmaddr, info.vmsize - 1, installName, info.segmentName});
+ });
+}
+
+static void buildSegmentInfo(const DyldSharedCache* dyldCache, std::vector<SegmentInfo>& segInfos)
+{
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ buildSegmentInfo(hdr, segInfos);
+ });
+ sortSegmentInfo(segInfos);
}
static void printSlideInfoForDataRegion(const DyldSharedCache* dyldCache, uint64_t dataStartAddress, uint64_t dataSize,
@@ -189,6 +372,20 @@
for(int j=0; j < slideInfoHeader->entries_size; ++j)
printf("%02X", entry->bits[j]);
printf("\n");
+ if ( verboseSlideInfo ) {
+ uint8_t* pageContent = (uint8_t*)(long)(dataPagesStart + (4096 * i));
+ for(int j=0; j < slideInfoHeader->entries_size; ++j) {
+ uint8_t bitmask = entry->bits[j];
+ for (unsigned k = 0; k != 8; ++k) {
+ if ( bitmask & (1 << k) ) {
+ uint32_t pageOffset = ((j * 8) + k) * 4;
+ uint8_t* loc = pageContent + pageOffset;
+ uint32_t rawValue = *((uint32_t*)loc);
+ printf(" [% 5d + 0x%04llX]: 0x%016llX\n", i, (uint64_t)(pageOffset), (uint64_t)rawValue);
+ }
+ }
+ }
+ }
}
}
else if ( slideInfoHeader->version == 2 ) {
@@ -282,9 +479,6 @@
if ( loc->auth.authenticated ) {
uint64_t target = authValueAdd + loc->auth.offsetFromSharedCacheBase;
uint64_t targetValue = target;
-#if __has_feature(ptrauth_calls)
- targetValue = ptr.arm64e.signPointer((void*)loc, target);
-#endif
printf(" [% 5d + 0x%04llX]: 0x%016llX (JOP: diversity %d, address %s, %s)\n",
i, (uint64_t)((const uint8_t*)loc - pageStart), targetValue,
ptr.arm64e.authBind.diversity, ptr.arm64e.authBind.addrDiv ? "true" : "false",
@@ -367,20 +561,735 @@
}
}
}
+ else if ( slideInfoHeader->version == 5 ) {
+ const dyld_cache_slide_info5* slideInfo = (dyld_cache_slide_info5*)(slideInfoHeader);
+ printf("page_size=%d\n", slideInfo->page_size);
+ printf("page_starts_count=%d\n", slideInfo->page_starts_count);
+ printf("auth_value_add=0x%016llX\n", slideInfo->value_add);
+ const uintptr_t valueAdd = (uintptr_t)(slideInfo->value_add);
+ for (int i=0; i < slideInfo->page_starts_count; ++i) {
+ uint16_t delta = slideInfo->page_starts[i];
+ if ( delta == DYLD_CACHE_SLIDE_V5_PAGE_ATTR_NO_REBASE ) {
+ printf("page[% 5d]: no rebasing\n", i);
+ continue;
+ }
+
+ printf("page[% 5d]: start=0x%04X\n", i, delta);
+ if ( !verboseSlideInfo )
+ continue;
+
+ delta = delta/sizeof(uint64_t); // initial offset is byte based
+ const uint8_t* pageStart = dataPagesStart + (i * slideInfo->page_size);
+ const dyld_cache_slide_pointer5* loc = (dyld_cache_slide_pointer5*)pageStart;
+
+ do {
+ loc += delta;
+ delta = loc->regular.next;
+
+ dyld3::MachOLoaded::ChainedFixupPointerOnDisk ptr;
+ ptr.raw64 = *((uint64_t*)loc);
+ PointerMetaData pmd(&ptr, DYLD_CHAINED_PTR_ARM64E_SHARED_CACHE);
+
+ uint64_t targetValue = valueAdd + loc->regular.runtimeOffset;
+ if ( pmd.authenticated ) {
+ printf(" [% 5d + 0x%04llX]: 0x%016llX (JOP: diversity %d, address %s, %s)\n",
+ i, (uint64_t)((const uint8_t*)loc - pageStart), targetValue,
+ ptr.cache64e.auth.diversity, ptr.cache64e.auth.addrDiv ? "true" : "false",
+ ptr.cache64e.keyName());
+ } else {
+ targetValue = targetValue | ptr.cache64e.high8();
+ printf(" [% 5d + 0x%04llX]: 0x%016llX\n", i, (uint64_t)((const uint8_t*)loc - pageStart), targetValue);
+ }
+ } while (delta != 0);
+ }
+ }
}
-
-static void findImageAndSegment(const DyldSharedCache* dyldCache, const std::vector<SegmentInfo>& segInfos, uint64_t cacheOffset, SegmentInfo* found)
+static void forEachSlidValue(const DyldSharedCache* dyldCache, uint64_t dataStartAddress, uint64_t dataSize,
+ const uint8_t* dataPagesStart,
+ const dyld_cache_slide_info* slideInfoHeader,
+ void (^callback)(uint64_t fixupVMAddr, uint64_t targetVMAddr,
+ PointerMetaData PMD))
+{
+ if ( slideInfoHeader->version == 1 ) {
+ const dyld_cache_slide_info_entry* entries = (dyld_cache_slide_info_entry*)((char*)slideInfoHeader + slideInfoHeader->entries_offset);
+ const uint16_t* tocs = (uint16_t*)((char*)slideInfoHeader + slideInfoHeader->toc_offset);
+ for(int i=0; i < slideInfoHeader->toc_count; ++i) {
+ const dyld_cache_slide_info_entry* entry = &entries[tocs[i]];
+ uint8_t* pageContent = (uint8_t*)(long)(dataPagesStart + (4096 * i));
+ for(int j=0; j < slideInfoHeader->entries_size; ++j) {
+ uint8_t bitmask = entry->bits[j];
+ for (unsigned k = 0; k != 8; ++k) {
+ if ( bitmask & (1 << k) ) {
+ uint32_t pageOffset = ((j * 8) + k) * 4;
+ uint8_t* loc = pageContent + pageOffset;
+ uint32_t rawValue = *((uint32_t*)loc);
+
+ uint64_t offsetInDataRegion = loc - dataPagesStart;
+ uint64_t fixupVMAddr = dataStartAddress + offsetInDataRegion;
+ uint64_t targetVMAddr = rawValue;
+ callback(fixupVMAddr, targetVMAddr, PointerMetaData());
+ }
+ }
+ }
+ }
+ }
+ else if ( slideInfoHeader->version == 2 ) {
+ const dyld_cache_slide_info2* slideInfo = (dyld_cache_slide_info2*)(slideInfoHeader);
+ const uint16_t* starts = (uint16_t* )((char*)slideInfo + slideInfo->page_starts_offset);
+ const uint16_t* extras = (uint16_t* )((char*)slideInfo + slideInfo->page_extras_offset);
+ for (int i=0; i < slideInfo->page_starts_count; ++i) {
+ const uint16_t start = starts[i];
+ auto rebaseChain = [&](uint8_t* pageContent, uint16_t startOffset)
+ {
+ uintptr_t slideAmount = 0;
+ const uintptr_t deltaMask = (uintptr_t)(slideInfo->delta_mask);
+ const uintptr_t valueMask = ~deltaMask;
+ const uintptr_t valueAdd = (uintptr_t)(slideInfo->value_add);
+ const unsigned deltaShift = __builtin_ctzll(deltaMask) - 2;
+
+ uint32_t pageOffset = startOffset;
+ uint32_t delta = 1;
+ while ( delta != 0 ) {
+ uint8_t* loc = pageContent + pageOffset;
+ uintptr_t rawValue = *((uintptr_t*)loc);
+ delta = (uint32_t)((rawValue & deltaMask) >> deltaShift);
+ uintptr_t value = (rawValue & valueMask);
+ if ( value != 0 ) {
+ value += valueAdd;
+ value += slideAmount;
+ }
+ pageOffset += delta;
+
+ uint64_t offsetInDataRegion = loc - dataPagesStart;
+ uint64_t fixupVMAddr = dataStartAddress + offsetInDataRegion;
+ uint64_t targetVMAddr = value;
+ callback(fixupVMAddr, targetVMAddr, PointerMetaData());
+ }
+ };
+ if ( start == DYLD_CACHE_SLIDE_PAGE_ATTR_NO_REBASE ) {
+ // Nothing to do here
+ }
+ else if ( start & DYLD_CACHE_SLIDE_PAGE_ATTR_EXTRA ) {
+ int j=(start & 0x3FFF);
+ bool done = false;
+ do {
+ uint16_t aStart = extras[j];
+ uint8_t* page = (uint8_t*)(long)(dataPagesStart + (slideInfo->page_size*i));
+ uint16_t pageStartOffset = (aStart & 0x3FFF)*4;
+ rebaseChain(page, pageStartOffset);
+ done = (extras[j] & DYLD_CACHE_SLIDE_PAGE_ATTR_END);
+ ++j;
+ } while ( !done );
+ }
+ else {
+ uint8_t* page = (uint8_t*)(long)(dataPagesStart + (slideInfo->page_size*i));
+ uint16_t pageStartOffset = start*4;
+ rebaseChain(page, pageStartOffset);
+ }
+ }
+ }
+ else if ( slideInfoHeader->version == 3 ) {
+ const dyld_cache_slide_info3* slideInfo = (dyld_cache_slide_info3*)(slideInfoHeader);
+ const uintptr_t authValueAdd = (uintptr_t)(slideInfo->auth_value_add);
+ for (int i=0; i < slideInfo->page_starts_count; ++i) {
+ uint16_t delta = slideInfo->page_starts[i];
+ if ( delta == DYLD_CACHE_SLIDE_V3_PAGE_ATTR_NO_REBASE ) {
+ // Nothing to do here
+ continue;
+ }
+
+ delta = delta/sizeof(uint64_t); // initial offset is byte based
+ const uint8_t* pageStart = dataPagesStart + (i * slideInfo->page_size);
+ const dyld_cache_slide_pointer3* loc = (dyld_cache_slide_pointer3*)pageStart;
+ do {
+ loc += delta;
+ delta = loc->plain.offsetToNextPointer;
+ dyld3::MachOLoaded::ChainedFixupPointerOnDisk ptr;
+ ptr.raw64 = *((uint64_t*)loc);
+ if ( loc->auth.authenticated ) {
+ uint64_t targetVMAddr = authValueAdd + loc->auth.offsetFromSharedCacheBase;
+
+ PointerMetaData pmd(&ptr, DYLD_CHAINED_PTR_ARM64E);
+ uint64_t offsetInDataRegion = (const uint8_t*)loc - dataPagesStart;
+ uint64_t fixupVMAddr = dataStartAddress + offsetInDataRegion;
+ callback(fixupVMAddr, targetVMAddr, pmd);
+ }
+ else {
+ uint64_t targetVMAddr = ptr.arm64e.unpackTarget();
+
+ uint64_t offsetInDataRegion = (const uint8_t*)loc - dataPagesStart;
+ uint64_t fixupVMAddr = dataStartAddress + offsetInDataRegion;
+ callback(fixupVMAddr, targetVMAddr, PointerMetaData());
+ }
+ } while (delta != 0);
+ }
+ }
+ else if ( slideInfoHeader->version == 4 ) {
+ const dyld_cache_slide_info4* slideInfo = (dyld_cache_slide_info4*)(slideInfoHeader);
+ const uint16_t* starts = (uint16_t* )((char*)slideInfo + slideInfo->page_starts_offset);
+ const uint16_t* extras = (uint16_t* )((char*)slideInfo + slideInfo->page_extras_offset);
+ for (int i=0; i < slideInfo->page_starts_count; ++i) {
+ const uint16_t start = starts[i];
+ auto rebaseChainV4 = [&](uint8_t* pageContent, uint16_t startOffset)
+ {
+ uintptr_t slideAmount = 0;
+ const uintptr_t deltaMask = (uintptr_t)(slideInfo->delta_mask);
+ const uintptr_t valueMask = ~deltaMask;
+ const uintptr_t valueAdd = (uintptr_t)(slideInfo->value_add);
+ const unsigned deltaShift = __builtin_ctzll(deltaMask) - 2;
+
+ uint32_t pageOffset = startOffset;
+ uint32_t delta = 1;
+ while ( delta != 0 ) {
+ uint8_t* loc = pageContent + pageOffset;
+ uint32_t rawValue = *((uint32_t*)loc);
+ delta = (uint32_t)((rawValue & deltaMask) >> deltaShift);
+ uintptr_t value = (rawValue & valueMask);
+ if ( (value & 0xFFFF8000) == 0 ) {
+ // small positive non-pointer, use as-is
+ }
+ else if ( (value & 0x3FFF8000) == 0x3FFF8000 ) {
+ // small negative non-pointer
+ value |= 0xC0000000;
+ }
+ else {
+ value += valueAdd;
+ value += slideAmount;
+
+ uint64_t offsetInDataRegion = (const uint8_t*)loc - dataPagesStart;
+ uint64_t fixupVMAddr = dataStartAddress + offsetInDataRegion;
+ uint64_t targetVMAddr = value;
+ callback(fixupVMAddr, targetVMAddr, PointerMetaData());
+ }
+ pageOffset += delta;
+ }
+ };
+ if ( start == DYLD_CACHE_SLIDE4_PAGE_NO_REBASE ) {
+ // Nothing to do here
+ }
+ else if ( start & DYLD_CACHE_SLIDE4_PAGE_USE_EXTRA ) {
+ int j=(start & DYLD_CACHE_SLIDE4_PAGE_INDEX);
+ bool done = false;
+ do {
+ uint16_t aStart = extras[j];
+ uint8_t* page = (uint8_t*)(long)(dataPagesStart + (slideInfo->page_size*i));
+ uint16_t pageStartOffset = (aStart & DYLD_CACHE_SLIDE4_PAGE_INDEX)*4;
+ rebaseChainV4(page, pageStartOffset);
+ done = (extras[j] & DYLD_CACHE_SLIDE4_PAGE_EXTRA_END);
+ ++j;
+ } while ( !done );
+ }
+ else {
+ uint8_t* page = (uint8_t*)(long)(dataPagesStart + (slideInfo->page_size*i));
+ uint16_t pageStartOffset = start*4;
+ rebaseChainV4(page, pageStartOffset);
+ }
+ }
+ }
+ else if ( slideInfoHeader->version == 5 ) {
+ const dyld_cache_slide_info5* slideInfo = (dyld_cache_slide_info5*)(slideInfoHeader);
+ const uintptr_t valueAdd = (uintptr_t)(slideInfo->value_add);
+ for (int i=0; i < slideInfo->page_starts_count; ++i) {
+ uint16_t delta = slideInfo->page_starts[i];
+ if ( delta == DYLD_CACHE_SLIDE_V5_PAGE_ATTR_NO_REBASE ) {
+ // Nothing to do here
+ continue;
+ }
+
+ delta = delta/sizeof(uint64_t); // initial offset is byte based
+ const uint8_t* pageStart = dataPagesStart + (i * slideInfo->page_size);
+ const dyld_cache_slide_pointer5* loc = (dyld_cache_slide_pointer5*)pageStart;
+ do {
+ loc += delta;
+ delta = loc->regular.next;
+
+ dyld3::MachOLoaded::ChainedFixupPointerOnDisk ptr;
+ ptr.raw64 = *((uint64_t*)loc);
+ PointerMetaData pmd(&ptr, DYLD_CHAINED_PTR_ARM64E_SHARED_CACHE);
+
+ uint64_t offsetInDataRegion = (const uint8_t*)loc - dataPagesStart;
+ uint64_t fixupVMAddr = dataStartAddress + offsetInDataRegion;
+ uint64_t targetVMAddr = valueAdd + loc->auth.runtimeOffset + ((uint64_t)pmd.high8 << 56);
+ callback(fixupVMAddr, targetVMAddr, pmd);
+ } while (delta != 0);
+ }
+ }
+}
+
+
+static bool findImageAndSegment(const DyldSharedCache* dyldCache, const std::vector<SegmentInfo>& segInfos, uint64_t cacheOffset, SegmentInfo* found)
{
const uint64_t locVmAddr = dyldCache->unslidLoadAddress() + cacheOffset;
- const SegmentInfo target = { locVmAddr, 0, NULL, NULL };
+ const SegmentInfo target = { locVmAddr, 0, NULL, { } };
const auto lowIt = std::lower_bound(segInfos.begin(), segInfos.end(), target,
[](const SegmentInfo& l, const SegmentInfo& r) -> bool {
return l.vmAddr+l.vmSize < r.vmAddr+r.vmSize;
});
+
+ if ( lowIt == segInfos.end() )
+ return false;
+
+ if ( locVmAddr < lowIt->vmAddr )
+ return false;
+ if ( locVmAddr >= (lowIt->vmAddr + lowIt->vmSize) )
+ return false;
+
*found = *lowIt;
+ return true;
}
+static void dumpObjCClassLayout(const DyldSharedCache* dyldCache)
+{
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
+ Diagnostics diag;
+
+ uint64_t sharedCacheRelativeSelectorBaseVMAddress = dyldCache->sharedCacheRelativeSelectorBaseVMAddress();
+ __block objc_visitor::Visitor visitor(dyldCache, ma, VMAddress(sharedCacheRelativeSelectorBaseVMAddress));
+ visitor.forEachClassAndMetaClass(^(const objc_visitor::Class& objcClass, bool& stopClass) {
+ const char* className = objcClass.getName(visitor);
+ bool isMetaClass = objcClass.isMetaClass;
+ uint32_t instanceStart = objcClass.getInstanceStart(visitor);
+ objc_visitor::IVarList ivars = objcClass.getIVars(visitor);
+
+ printf("%s (%s): start 0x%x\n", className, isMetaClass ? "metaclass" : "class", instanceStart);
+ std::optional<metadata_visitor::ResolvedValue> superClassValue = objcClass.getSuperclass(visitor);
+ if ( superClassValue.has_value() ) {
+ bool unusedIsPatchable = false;
+ objc_visitor::Class superClass(superClassValue.value(), isMetaClass, unusedIsPatchable);
+ const char* superClassName = superClass.getName(visitor);
+ uint32_t superStart = superClass.getInstanceStart(visitor);
+ uint32_t superSize = superClass.getInstanceSize(visitor);
+ printf(" super %s (%s): start 0x%x, size 0x%x\n", superClassName,
+ isMetaClass ? "metaclass" : "class", superStart, superSize);
+ }
+
+ uint32_t numIVars = ivars.numIVars();
+ for ( uint32_t i = 0; i != numIVars; ++i ) {
+ objc_visitor::IVar ivar = ivars.getIVar(visitor, i);
+ std::optional<uint32_t> ivarStart = ivar.getOffset(visitor);
+ const char* name = ivar.getName(visitor);
+ printf(" ivar %s: 0x%x (start + 0x%d), alignment %d\n",
+ name,
+ ivarStart.has_value() ? ivarStart.value() : -1,
+ ivarStart.has_value() ? (ivarStart.value() - instanceStart) : -1,
+ ivar.getAlignment(visitor));
+ }
+ });
+ });
+}
+
+template<typename ListTy>
+static ListTy skipListsOfLists(ListTy&& list, const objc_visitor::Visitor& visitor)
+{
+ // we only want the class list. Ignore all ther other lists of lists
+ if ( list.isListOfLists() ) {
+ const ListOfListsEntry* listHeader = (ListOfListsEntry*)((uint8_t*) ((uint64_t)list.getLocation() & ~1));
+ VMAddress methodListVMAddr = list.getVMAddress().value() - VMOffset(1ULL);
+
+ if ( listHeader->count != 0 ) {
+ uint32_t classListIndex = listHeader->count - 1;
+
+ const ListOfListsEntry& listEntry = (listHeader + 1)[classListIndex];
+
+ // The list entry is a relative offset to the target
+ // Work out the VMAddress of that target
+ VMOffset listEntryVMOffset{(uint64_t)&listEntry - (uint64_t)listHeader};
+ VMAddress listEntryVMAddr = methodListVMAddr + listEntryVMOffset;
+ VMAddress targetVMAddr = listEntryVMAddr + VMOffset((uint64_t)listEntry.offset);
+
+ metadata_visitor::ResolvedValue classMethodListValue = visitor.getValueFor(targetVMAddr);
+ ListTy classMethodList(classMethodListValue);
+
+ return classMethodList;
+ } else {
+ return { std::nullopt };
+ }
+ }
+
+ return list;
+}
+
+static void dumpObjCClassMethodLists(const DyldSharedCache* dyldCache)
+{
+ // Map from vmAddr to the category name for that address
+
+ __block std::unordered_map<VMAddress, std::string, VMAddressHash, VMAddressEqual> categoryMap;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
+ Diagnostics diag;
+
+ const char* leafName = strrchr(installName, '/');
+ if ( leafName == NULL )
+ leafName = installName;
+ else
+ leafName++;
+
+ uint64_t sharedCacheRelativeSelectorBaseVMAddress = dyldCache->sharedCacheRelativeSelectorBaseVMAddress();
+ __block objc_visitor::Visitor visitor(dyldCache, ma, VMAddress(sharedCacheRelativeSelectorBaseVMAddress));
+ visitor.forEachCategory(^(const objc_visitor::Category& objcCategory, bool& stopCategory) {
+ const char* categoryName = objcCategory.getName(visitor);
+ {
+ objc_visitor::MethodList methodList = objcCategory.getClassMethods(visitor);
+ std::optional<VMAddress> vmAddr = methodList.getVMAddress();
+ if ( vmAddr.has_value() ) {
+ categoryMap[vmAddr.value()] = std::string(categoryName) + " - " + leafName;
+ }
+ }
+ {
+ objc_visitor::MethodList methodList = objcCategory.getInstanceMethods(visitor);
+ std::optional<VMAddress> vmAddr = methodList.getVMAddress();
+ if ( vmAddr.has_value() ) {
+ categoryMap[vmAddr.value()] = std::string(categoryName) + " - " + leafName;
+ }
+ }
+ {
+ objc_visitor::ProtocolList protocolList = objcCategory.getProtocols(visitor);
+ std::optional<VMAddress> vmAddr = protocolList.getVMAddress();
+ if ( vmAddr.has_value() ) {
+ categoryMap[vmAddr.value()] = std::string(categoryName) + " - " + leafName;
+ }
+ }
+ {
+ objc_visitor::PropertyList propertyList = objcCategory.getClassProperties(visitor);
+ std::optional<VMAddress> vmAddr = propertyList.getVMAddress();
+ if ( vmAddr.has_value() ) {
+ categoryMap[vmAddr.value()] = std::string(categoryName) + " - " + leafName;
+ }
+ }
+ {
+ objc_visitor::PropertyList propertyList = objcCategory.getInstanceProperties(visitor);
+ std::optional<VMAddress> vmAddr = propertyList.getVMAddress();
+ if ( vmAddr.has_value() ) {
+ categoryMap[vmAddr.value()] = std::string(categoryName) + " - " + leafName;
+ }
+ }
+ });
+ });
+
+ __block std::map<uint64_t, const char*> dylibVMAddrMap;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ if ( hdr->hasObjC() )
+ dylibVMAddrMap[hdr->preferredLoadAddress()] = installName;
+ });
+
+#if 0
+ // Get a map of all dylibs in the cache from their "objc index" to install name
+ __block std::map<uint16_t, const char*> dylibMap;
+
+ const objc::HeaderInfoRO* headerInfoRO = dyldCache->objcHeaderInfoRO();
+ const bool is64 = (strstr(dyldCache->archName(), "64") != nullptr) && (strstr(dyldCache->archName(), "64_32") == nullptr);
+ if ( is64 ) {
+ const auto* headerInfo64 = (objc::objc_headeropt_ro_t<uint64_t>*)headerInfoRO;
+ uint64_t headerInfoVMAddr = dyldCache->unslidLoadAddress();
+ headerInfoVMAddr += (uint64_t)headerInfo64 - (uint64_t)dyldCache;
+ for ( std::pair<uint64_t, const char*> vmAddrAndName : dylibVMAddrMap ) {
+ const objc::objc_header_info_ro_t<uint64_t>* element = headerInfo64->get(headerInfoVMAddr, vmAddrAndName.first);
+ if ( element != nullptr ) {
+ dylibMap[headerInfo64->index(element)] = vmAddrAndName.second;
+ }
+ }
+ } else {
+ const auto* headerInfo32 = (objc::objc_headeropt_ro_t<uint32_t>*)headerInfoRO;
+ uint64_t headerInfoVMAddr = dyldCache->unslidLoadAddress();
+ headerInfoVMAddr += (uint64_t)headerInfo32 - (uint64_t)dyldCache;
+ for ( std::pair<uint64_t, const char*> vmAddrAndName : dylibVMAddrMap ) {
+ const objc::objc_header_info_ro_t<uint32_t>* element = headerInfo32->get(headerInfoVMAddr, vmAddrAndName.first);
+ if ( element != nullptr ) {
+ dylibMap[headerInfo32->index(element)] = vmAddrAndName.second;
+ }
+ }
+ }
+ }
+#endif
+
+ // Print all method lists in the shared cache
+
+ struct ListOfListsEntry {
+ union {
+ struct {
+ uint64_t imageIndex: 16;
+ int64_t offset: 48;
+ };
+ struct {
+ uint32_t entsize;
+ uint32_t count;
+ };
+ };
+ };
+
+ __block std::unordered_set<VMAddress, VMAddressHash, VMAddressEqual> seenCategories;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
+ Diagnostics diag;
+
+ printf("--- %s ---\n", installName);
+
+ uint64_t sharedCacheRelativeSelectorBaseVMAddress = dyldCache->sharedCacheRelativeSelectorBaseVMAddress();
+ __block objc_visitor::Visitor visitor(dyldCache, ma, VMAddress(sharedCacheRelativeSelectorBaseVMAddress));
+ visitor.forEachClassAndMetaClass(^(const objc_visitor::Class& objcClass, bool& stopClass) {
+ const char* className = objcClass.getName(visitor);
+ bool isMetaClass = objcClass.isMetaClass;
+
+ printf("%s (%s):\n", className, isMetaClass ? "metaclass" : "class");
+ // method lists
+ {
+ objc_visitor::MethodList methodList = objcClass.getBaseMethods(visitor);
+ if ( methodList.isListOfLists() ) {
+ const ListOfListsEntry* listHeader = (ListOfListsEntry*)((uint8_t*) ((uint64_t)methodList.getLocation() & ~1));
+ VMAddress methodListVMAddr = methodList.getVMAddress().value() - VMOffset(1ULL);
+
+ printf("(list of %d lists) {\n", listHeader->count);
+ for ( uint32_t i = 0; i != listHeader->count; ++i ) {
+ const ListOfListsEntry& listEntry = (listHeader + 1)[i];
+
+ // The list entry is a relative offset to the target
+ // Work out the VMAddress of that target
+ VMOffset listEntryVMOffset{(uint64_t)&listEntry - (uint64_t)listHeader};
+ VMAddress listEntryVMAddr = methodListVMAddr + listEntryVMOffset;
+ VMAddress targetVMAddr = listEntryVMAddr + VMOffset((uint64_t)listEntry.offset);
+
+ auto categoryIt = categoryMap.find(targetVMAddr);
+ if ( categoryIt != categoryMap.end() ) {
+ seenCategories.insert(targetVMAddr);
+ printf(" (category methods: image (%d) %s) {\n", listEntry.imageIndex, categoryIt->second.c_str());
+
+ metadata_visitor::ResolvedValue catMethodListValue = visitor.getValueFor(targetVMAddr);
+ objc_visitor::MethodList catMethodList(catMethodListValue);
+ uint32_t numMethods = catMethodList.numMethods();
+ for ( uint32_t methodIndex = 0; methodIndex != numMethods; ++methodIndex ) {
+ objc_visitor::Method method = catMethodList.getMethod(visitor, methodIndex);
+ const char* name = method.getName(visitor);
+ printf(" %s\n", name);
+ }
+
+ printf(" }\n");
+ } else {
+ // If we didn't find a category then we must be processing the class
+ // methods. These have to be last
+ if ( (i + 1) != listHeader->count ) {
+ fprintf(stderr, "Invalid method list on %s in %s\n", className, installName);
+ exit(1);
+ }
+ printf(" (class methods: image (%d)) {\n", listEntry.imageIndex);
+
+ metadata_visitor::ResolvedValue classMethodListValue = visitor.getValueFor(targetVMAddr);
+ objc_visitor::MethodList classMethodList(classMethodListValue);
+ uint32_t numMethods = classMethodList.numMethods();
+ for ( uint32_t methodIndex = 0; methodIndex != numMethods; ++methodIndex ) {
+ objc_visitor::Method method = classMethodList.getMethod(visitor, methodIndex);
+ const char* name = method.getName(visitor);
+ printf(" %s\n", name);
+ }
+
+ printf(" }\n");
+ }
+ }
+ printf("}\n");
+ } else {
+ printf("(class methods) {\n");
+ uint32_t numMethods = methodList.numMethods();
+ for ( uint32_t methodIndex = 0; methodIndex != numMethods; ++methodIndex ) {
+ objc_visitor::Method method = methodList.getMethod(visitor, methodIndex);
+ const char* name = method.getName(visitor);
+ printf(" %s\n", name);
+ }
+ printf("}\n");
+ }
+ }
+
+ // protocol lists
+ if ( !isMetaClass) {
+ objc_visitor::ProtocolList protocolList = objcClass.getBaseProtocols(visitor);
+ if ( protocolList.isListOfLists() ) {
+
+ const ListOfListsEntry* listHeader = (ListOfListsEntry*)((uint8_t*) ((uint64_t)protocolList.getLocation() & ~1));
+ VMAddress protocolListVMAddr = protocolList.getVMAddress().value() - VMOffset(1ULL);
+
+ printf("(list of %d lists) {\n", listHeader->count);
+ for ( uint32_t i = 0; i != listHeader->count; ++i ) {
+ const ListOfListsEntry& listEntry = (listHeader + 1)[i];
+
+ // The list entry is a relative offset to the target
+ // Work out the VMAddress of that target
+ VMOffset listEntryVMOffset{(uint64_t)&listEntry - (uint64_t)listHeader};
+ VMAddress listEntryVMAddr = protocolListVMAddr + listEntryVMOffset;
+ VMAddress targetVMAddr = listEntryVMAddr + VMOffset((uint64_t)listEntry.offset);
+
+ auto categoryIt = categoryMap.find(targetVMAddr);
+ if ( categoryIt != categoryMap.end() ) {
+ seenCategories.insert(targetVMAddr);
+ printf(" (category protocols: image (%d) %s) {\n", listEntry.imageIndex, categoryIt->second.c_str());
+
+ metadata_visitor::ResolvedValue catProtocolListValue = visitor.getValueFor(targetVMAddr);
+ objc_visitor::ProtocolList catProtocolList(catProtocolListValue);
+ uint64_t numProtocols = catProtocolList.numProtocols(visitor);
+ for ( uint64_t protocolIndex = 0; protocolIndex != numProtocols; ++protocolIndex ) {
+ objc_visitor::Protocol protocol = catProtocolList.getProtocol(visitor, protocolIndex);
+ const char* name = protocol.getName(visitor);
+ printf(" %s\n", name);
+ }
+
+ printf(" }\n");
+ } else {
+ // If we didn't find a category then we must be processing the class
+ // protocols. These have to be last
+ if ( (i + 1) != listHeader->count ) {
+ fprintf(stderr, "Invalid protocol list on %s in %s\n", className, installName);
+ exit(1);
+ }
+ printf(" (class protocols: image (%d)) {\n", listEntry.imageIndex);
+
+ metadata_visitor::ResolvedValue classProtocolListValue = visitor.getValueFor(targetVMAddr);
+ objc_visitor::ProtocolList classProtocolList(classProtocolListValue);
+ uint64_t numProtocols = classProtocolList.numProtocols(visitor);
+ for ( uint64_t protocolIndex = 0; protocolIndex != numProtocols; ++protocolIndex ) {
+ objc_visitor::Protocol protocol = classProtocolList.getProtocol(visitor, protocolIndex);
+ const char* name = protocol.getName(visitor);
+ printf(" %s\n", name);
+ }
+
+ printf(" }\n");
+ }
+ }
+ printf("}\n");
+ } else {
+ printf("(class protocols) {\n");
+ uint64_t numProtocols = protocolList.numProtocols(visitor);
+ for ( uint64_t protocolIndex = 0; protocolIndex != numProtocols; ++protocolIndex ) {
+ objc_visitor::Protocol protocol = protocolList.getProtocol(visitor, protocolIndex);
+ const char* name = protocol.getName(visitor);
+ printf(" %s\n", name);
+ }
+ printf("}\n");
+ }
+ }
+ // property lists
+ {
+ objc_visitor::PropertyList propertyList = objcClass.getBaseProperties(visitor);
+ if ( propertyList.isListOfLists() ) {
+ const ListOfListsEntry* listHeader = (ListOfListsEntry*)((uint8_t*) ((uint64_t)propertyList.getLocation() & ~1));
+ VMAddress propertyListVMAddr = propertyList.getVMAddress().value() - VMOffset(1ULL);
+
+ printf("(list of %d lists) {\n", listHeader->count);
+ for ( uint32_t i = 0; i != listHeader->count; ++i ) {
+ const ListOfListsEntry& listEntry = (listHeader + 1)[i];
+
+ // The list entry is a relative offset to the target
+ // Work out the VMAddress of that target
+ VMOffset listEntryVMOffset{(uint64_t)&listEntry - (uint64_t)listHeader};
+ VMAddress listEntryVMAddr = propertyListVMAddr + listEntryVMOffset;
+ VMAddress targetVMAddr = listEntryVMAddr + VMOffset((uint64_t)listEntry.offset);
+
+ auto categoryIt = categoryMap.find(targetVMAddr);
+ if ( categoryIt != categoryMap.end() ) {
+ seenCategories.insert(targetVMAddr);
+ printf(" (category properties: image (%d) %s) {\n", listEntry.imageIndex, categoryIt->second.c_str());
+
+ metadata_visitor::ResolvedValue catPropertyListValue = visitor.getValueFor(targetVMAddr);
+ objc_visitor::PropertyList catPropertyList(catPropertyListValue);
+ uint32_t numProperties = catPropertyList.numProperties();
+ for ( uint32_t propertyIndex = 0; propertyIndex != numProperties; ++propertyIndex ) {
+ objc_visitor::Property property = catPropertyList.getProperty(visitor, propertyIndex);
+ const char* name = property.getName(visitor);
+ printf(" %s\n", name);
+ }
+
+ printf(" }\n");
+ } else {
+ // If we didn't find a category then we must be processing the class
+ // properties. These have to be last
+ if ( (i + 1) != listHeader->count ) {
+ fprintf(stderr, "Invalid property list on %s in %s\n", className, installName);
+ exit(1);
+ }
+ printf(" (class properties: image (%d)) {\n", listEntry.imageIndex);
+
+ metadata_visitor::ResolvedValue classPropertyListValue = visitor.getValueFor(targetVMAddr);
+ objc_visitor::PropertyList classPropertyList(classPropertyListValue);
+ uint32_t numProperties = classPropertyList.numProperties();
+ for ( uint32_t propertyIndex = 0; propertyIndex != numProperties; ++propertyIndex ) {
+ objc_visitor::Property property = classPropertyList.getProperty(visitor, propertyIndex);
+ const char* name = property.getName(visitor);
+ printf(" %s\n", name);
+ }
+
+ printf(" }\n");
+ }
+ }
+ printf("}\n");
+ } else {
+ printf("(class properties) {\n");
+ uint32_t numProperties = propertyList.numProperties();
+ for ( uint32_t propertyIndex = 0; propertyIndex != numProperties; ++propertyIndex ) {
+ objc_visitor::Property property = propertyList.getProperty(visitor, propertyIndex);
+ const char* name = property.getName(visitor);
+ printf(" %s\n", name);
+ }
+ printf("}\n");
+ }
+ }
+ });
+ });
+
+ // Check if any categories weren't attached
+ bool badCategory = false;
+ for ( auto& [vmAddr, name] : categoryMap ) {
+ if ( seenCategories.count(vmAddr) )
+ continue;
+
+ badCategory = true;
+ fprintf(stderr, "Failed to find class with category: %s\n", name.c_str());
+ }
+
+ if ( badCategory )
+ exit(1);
+}
+
+static bool patchKindIsRootOptimized(PatchKind patchKind)
+{
+ return (patchKind == PatchKind::cfObj2) || (patchKind == PatchKind::objcClass);
+}
+
+struct PatchPageMetric
+{
+ std::unordered_set<uint64_t> pages;
+ size_t numUses = 0;
+
+ void addUse(uint64_t cacheOffset)
+ {
+ ++numUses;
+ pages.insert(cacheOffset & ~0x3FFF);
+ }
+};
+
+static void logCostOfRootPerSymbol(const std::unordered_map<CString, PatchPageMetric>& pagesPerSymbol)
+{
+ printf("\n%ld symbols page coverage:\n", pagesPerSymbol.size());
+
+ struct Metric
+ {
+ CString symName;
+ size_t numPages = 0;
+ size_t numUses = 0;
+ };
+
+ std::vector<Metric> metrics;
+ metrics.reserve(pagesPerSymbol.size());
+
+ std::ranges::transform(pagesPerSymbol, std::back_inserter(metrics), [](auto& entry) {
+ return Metric{entry.first, (size_t)entry.second.pages.size(), entry.second.numUses};
+ });
+ std::ranges::sort(metrics, [](auto& l, auto& r) {
+ if ( l.numPages == r.numPages )
+ return l.symName < r.symName;
+ return l.numPages < r.numPages;
+ });
+
+ for ( auto& metric : metrics )
+ printf("%s is on %ld pages, %ld uses\n", metric.symName.c_str(), metric.numPages, metric.numUses);
+}
int main (int argc, const char* argv[]) {
@@ -423,6 +1332,14 @@
checkMode(options.mode);
options.mode = modeInfo;
}
+ else if (strcmp(opt, "-tpro") == 0) {
+ checkMode(options.mode);
+ options.mode = modeTPROInfo;
+ }
+ else if (strcmp(opt, "-stats") == 0) {
+ checkMode(options.mode);
+ options.mode = modeStats;
+ }
else if (strcmp(opt, "-slide_info") == 0) {
checkMode(options.mode);
options.mode = modeSlideInfo;
@@ -430,6 +1347,16 @@
else if (strcmp(opt, "-verbose_slide_info") == 0) {
checkMode(options.mode);
options.mode = modeVerboseSlideInfo;
+ }
+ else if (strcmp(opt, "-fixups_in_dylib") == 0) {
+ checkMode(options.mode);
+ options.mode = modeFixupsInDylib;
+ options.fixupsInDylib = argv[++i];
+ if ( i >= argc ) {
+ fprintf(stderr, "Error: option -fixups_in_dylib requires a path argument\n");
+ usage();
+ exit(1);
+ }
}
else if (strcmp(opt, "-text_info") == 0) {
checkMode(options.mode);
@@ -455,6 +1382,12 @@
options.mode = modeStrings;
printExports = true;
}
+ else if (strcmp(opt, "-duplicate_exports") == 0) {
+ options.mode = modeDuplicates;
+ }
+ else if (strcmp(opt, "-duplicate_exports_summary") == 0) {
+ options.mode = modeDuplicatesSummary;
+ }
else if (strcmp(opt, "-map") == 0) {
checkMode(options.mode);
options.mode = modeMap;
@@ -463,6 +1396,10 @@
checkMode(options.mode);
options.mode = modeJSONMap;
}
+ else if (strcmp(opt, "-verbose-json-map") == 0) {
+ checkMode(options.mode);
+ options.mode = modeVerboseJSONMap;
+ }
else if (strcmp(opt, "-json-dependents") == 0) {
checkMode(options.mode);
options.mode = modeJSONDependents;
@@ -487,6 +1424,18 @@
checkMode(options.mode);
options.mode = modeObjCClasses;
}
+ else if (strcmp(opt, "-objc-class-layout") == 0) {
+ checkMode(options.mode);
+ options.mode = modeObjCClassLayout;
+ }
+ else if (strcmp(opt, "-objc-class-method-lists") == 0) {
+ checkMode(options.mode);
+ options.mode = modeObjCClassMethodLists;
+ }
+ else if (strcmp(opt, "-objc-class-hash-table") == 0) {
+ checkMode(options.mode);
+ options.mode = modeObjCClassHashTable;
+ }
else if (strcmp(opt, "-objc-selectors") == 0) {
checkMode(options.mode);
options.mode = modeObjCSelectors;
@@ -497,6 +1446,20 @@
else if (strcmp(opt, "-swift-proto") == 0) {
checkMode(options.mode);
options.mode = modeSwiftProtocolConformances;
+ }
+ else if (strcmp(opt, "-swift-ptrtables") == 0) {
+ checkMode(options.mode);
+ options.mode = modeSwiftPtrTables;
+ }
+ else if (strcmp(opt, "-lookup-va") == 0) {
+ checkMode(options.mode);
+ options.mode = modeLookupVA;
+ options.lookupVA = argv[++i];
+ if ( i >= argc ) {
+ fprintf(stderr, "Error: option -lookup-va requires an address argument\n");
+ usage();
+ exit(1);
+ }
}
else if (strcmp(opt, "-extract") == 0) {
checkMode(options.mode);
@@ -523,6 +1486,22 @@
else if (strcmp(opt, "-patch_table") == 0) {
options.mode = modePatchTable;
}
+ else if (strcmp(opt, "-dump_patch_table") == 0) {
+ options.mode = modeDumpPatchTable;
+ }
+ else if (strcmp(opt, "-function_variants") == 0) {
+ options.mode = modeFunctionVariants;
+ }
+ else if (strcmp(opt, "-roots_cost") == 0) {
+ checkMode(options.mode);
+ options.mode = modeRootsCost;
+ options.rootsCostOfDylib = argv[++i];
+ if ( i >= argc ) {
+ fprintf(stderr, "Error: option -roots_cost requires a path argument\n");
+ usage();
+ exit(1);
+ }
+ }
else if (strcmp(opt, "-list_dylibs_with_section") == 0) {
options.mode = modeListDylibsWithSection;
options.segmentName = argv[++i];
@@ -533,6 +1512,27 @@
exit(1);
}
}
+ else if (strcmp(opt, "-mach_headers") == 0) {
+ checkMode(options.mode);
+ options.mode = modeMachHeaders;
+ }
+ else if (strcmp(opt, "-load_commands") == 0) {
+ fprintf(stderr, "dyld_shared_cache_util -load_commands is deprecated. Use dyld_info -load_commands instead\n");
+ }
+ else if (strcmp(opt, "-cache_header") == 0) {
+ checkMode(options.mode);
+ options.mode = modeCacheHeader;
+ }
+ else if (strcmp(opt, "-dylib_symbols") == 0) {
+ checkMode(options.mode);
+ options.mode = modeDylibSymbols;
+ }
+ else if (strcmp(opt, "-function_starts") == 0) {
+ options.mode = modeFunctionStarts;
+ }
+ else if (strcmp(opt, "-prewarming_data") == 0) {
+ options.mode = modePrewarmingData;
+ }
else {
fprintf(stderr, "Error: unrecognized option %s\n", opt);
usage();
@@ -545,7 +1545,7 @@
}
if ( options.mode == modeNone ) {
- fprintf(stderr, "Error: select one of -list, -dependents, -info, -linkedit, or -map\n");
+ fprintf(stderr, "Error: no command selected\n");
usage();
exit(1);
}
@@ -570,12 +1570,14 @@
__block std::vector<const DyldSharedCache*> dyldCaches;
const DyldSharedCache* dyldCache = nullptr;
+ bool cacheOnDisk = false;
if ( sharedCachePath != nullptr ) {
dyldCaches = DyldSharedCache::mapCacheFiles(sharedCachePath);
// mapCacheFile prints an error if something goes wrong, so just return in that case.
if ( dyldCaches.empty() )
return 1;
dyldCache = dyldCaches.front();
+ cacheOnDisk = true;
}
else {
size_t cacheLength;
@@ -588,25 +1590,116 @@
fprintf(stderr, "Cannot use -objc-classes with a live cache. Please run with a path to an on-disk cache file\n");
return 1;
}
+ if ( options.mode == modeObjCClassLayout ) {
+ fprintf(stderr, "Cannot use -objc-class-layout with a live cache. Please run with a path to an on-disk cache file\n");
+ return 1;
+ }
+ if ( options.mode == modeObjCClassMethodLists ) {
+ fprintf(stderr, "Cannot use -objc-class-method-lists with a live cache. Please run with a path to an on-disk cache file\n");
+ return 1;
+ }
+ if ( options.mode == modeVerboseSlideInfo ) {
+ fprintf(stderr, "Cannot use -verbose_slide_info with a live cache. Please run with a path to an on-disk cache file\n");
+ return 1;
+ }
// The in-use cache might be the first cache file of many. In that case, also add the sub caches
dyldCache->forEachCache(^(const DyldSharedCache *cache, bool& stopCache) {
dyldCaches.push_back(dyldCache);
});
+ cacheOnDisk = false;
}
if ( options.mode == modeSlideInfo || options.mode == modeVerboseSlideInfo ) {
- if ( !dyldCache->hasSlideInfo() ) {
- fprintf(stderr, "Error: dyld shared cache does not contain slide info\n");
+ if ( dyldCache->numSubCaches() == 0 ) {
+ if ( !dyldCache->hasSlideInfo() ) {
+ fprintf(stderr, "Error: dyld shared cache does not contain slide info\n");
+ exit(1);
+ }
+ }
+
+ const bool verboseSlideInfo = (options.mode == modeVerboseSlideInfo);
+ dyldCache->forEachCache(^(const DyldSharedCache *cache, bool& stopCache) {
+ cache->forEachSlideInfo(^(uint64_t mappingStartAddress, uint64_t mappingSize, const uint8_t *mappingPagesStart,
+ uint64_t slideInfoOffset, uint64_t slideInfoSize, const dyld_cache_slide_info *slideInfoHeader) {
+ printSlideInfoForDataRegion(cache, mappingStartAddress, mappingSize, mappingPagesStart,
+ slideInfoHeader, verboseSlideInfo);
+ });
+ });
+ return 0;
+ }
+ else if ( options.mode == modeFixupsInDylib ) {
+ if ( dyldCache->numSubCaches() == 0 ) {
+ if ( !dyldCache->hasSlideInfo() ) {
+ fprintf(stderr, "Error: dyld shared cache does not contain slide info\n");
+ exit(1);
+ }
+ }
+
+ uint32_t imageIndex = ~0U;
+ if ( !dyldCache->hasImagePath(options.fixupsInDylib, imageIndex) ) {
+ fprintf(stderr, "Error: dyld shared cache does not contain image: %s\n",
+ options.fixupsInDylib);
exit(1);
}
- const bool verboseSlideInfo = (options.mode == modeVerboseSlideInfo);
- dyldCache->forEachSlideInfo(^(uint64_t mappingStartAddress, uint64_t mappingSize, const uint8_t *mappingPagesStart,
- uint64_t slideInfoOffset, uint64_t slideInfoSize, const dyld_cache_slide_info *slideInfoHeader) {
- printSlideInfoForDataRegion(dyldCache, mappingStartAddress, mappingSize, mappingPagesStart,
- slideInfoHeader, verboseSlideInfo);
+ const Header* hdr = (const Header*)dyldCache->getIndexedImageEntry(imageIndex);
+
+ __block std::vector<SegmentInfo> dylibSegInfo;
+ buildSegmentInfo(hdr, dylibSegInfo);
+ sortSegmentInfo(dylibSegInfo);
+
+ __block std::vector<SegmentInfo> cacheSegInfo;
+ buildSegmentInfo(dyldCache, cacheSegInfo);
+
+ uint64_t cacheBaseAddress = dyldCache->unslidLoadAddress();
+ auto handler = ^(uint64_t fixupVMAddr, uint64_t targetVMAddr,
+ PointerMetaData pmd)
+ {
+ SegmentInfo fixupAt;
+ if ( !findImageAndSegment(dyldCache, dylibSegInfo, fixupVMAddr - cacheBaseAddress, &fixupAt) ) {
+ // Fixup is not in the given dylib
+ return;
+ }
+
+ // Remove high8 if we have it
+ uint64_t high8 = targetVMAddr >> 56;
+ targetVMAddr = targetVMAddr & 0x00FFFFFFFFFFFFFF;
+
+ SegmentInfo targetAt;
+ if ( !findImageAndSegment(dyldCache, cacheSegInfo, targetVMAddr - cacheBaseAddress, &targetAt) ) {
+ return;
+ }
+
+ if ( pmd.authenticated ) {
+ static const char* keyNames[] = {
+ "IA", "IB", "DA", "DB"
+ };
+ printf("%.*s(0x%04llX) -> %.*s(0x%04llX):%s; (PAC: div=%d, addr=%s, key=%s)\n",
+ (int)fixupAt.segName.size(), fixupAt.segName.data(), fixupVMAddr - fixupAt.vmAddr,
+ (int)targetAt.segName.size(), targetAt.segName.data(), targetVMAddr - targetAt.vmAddr, targetAt.installName,
+ pmd.diversity, pmd.usesAddrDiversity ? "true" : "false", keyNames[pmd.key]);
+ } else {
+ if ( high8 != 0 ) {
+ printf("%.*s(0x%04llX) -> %.*s(0x%04llX):%s; (high8: 0x%02llX)\n",
+ (int)fixupAt.segName.size(), fixupAt.segName.data(), fixupVMAddr - fixupAt.vmAddr,
+ (int)targetAt.segName.size(), targetAt.segName.data(), targetVMAddr - targetAt.vmAddr, targetAt.installName,
+ high8);
+ } else {
+ printf("%.*s(0x%04llX) -> %.*s(0x%04llX):%s\n",
+ (int)fixupAt.segName.size(), fixupAt.segName.data(), fixupVMAddr - fixupAt.vmAddr,
+ (int)targetAt.segName.size(), targetAt.segName.data(), targetVMAddr - targetAt.vmAddr, targetAt.installName);
+ }
+ }
+ };
+
+ dyldCache->forEachCache(^(const DyldSharedCache *cache, bool& stopCache) {
+ cache->forEachSlideInfo(^(uint64_t mappingStartAddress, uint64_t mappingSize, const uint8_t *mappingPagesStart,
+ uint64_t slideInfoOffset, uint64_t slideInfoSize, const dyld_cache_slide_info *slideInfoHeader) {
+ forEachSlidValue(cache, mappingStartAddress, mappingSize, mappingPagesStart,
+ slideInfoHeader, handler);
+ });
});
return 0;
}
@@ -615,12 +1708,14 @@
uuid_string_t uuidString;
uuid_unparse_upper(header->uuid, uuidString);
printf("uuid: %s\n", uuidString);
-
- dyld3::Platform platform = dyldCache->platform();
- printf("platform: %s\n", dyld3::MachOFile::platformName(platform));
+ printf("platform: %s\n", dyldCache->platform().name().c_str());
printf("built by: %s\n", header->locallyBuiltCache ? "local machine" : "B&I");
- printf("cache type: %s\n", header->cacheType ? "production" : "development");
- if ( header->mappingOffset >= __offsetof(dyld_cache_header, imagesCount) ) {
+ printf("cache type: %s\n", DyldSharedCache::getCacheTypeName(header->cacheType));
+ if ( header->dylibsExpectedOnDisk )
+ printf("dylibs expected on disk: true\n");
+ if ( header->cacheType == kDyldSharedCacheTypeUniversal )
+ printf("cache sub-type: %s\n", DyldSharedCache::getCacheTypeName(header->cacheSubType));
+ if ( header->mappingOffset >= offsetof(dyld_cache_header, imagesCount) ) {
printf("image count: %u\n", header->imagesCount);
} else {
printf("image count: %u\n", header->imagesCountOld);
@@ -636,25 +1731,39 @@
entropyBits = __builtin_clz(possibleSlideValues - 1);
printf("ASLR entropy: %u-bits (%lldMB)\n", entropyBits, header->maxSlide >> 20);
}
-
printf("mappings:\n");
dyldCache->forEachRange(^(const char *mappingName, uint64_t unslidVMAddr, uint64_t vmSize,
uint32_t cacheFileIndex, uint64_t fileOffset, uint32_t initProt, uint32_t maxProt, bool& stopRange) {
- printf("%16s %4lluMB, file offset: #%u/0x%08llX -> 0x%08llX, address: 0x%08llX -> 0x%08llX\n",
- mappingName, vmSize / (1024*1024), cacheFileIndex, fileOffset, fileOffset + vmSize, unslidVMAddr, unslidVMAddr + vmSize);
+ std::string initProtString;
+ initProtString += (initProt & VM_PROT_READ) ? "r" : "-";
+ initProtString += (initProt & VM_PROT_WRITE) ? "w" : "-";
+ initProtString += (initProt & VM_PROT_EXECUTE) ? "x" : "-";
+
+ std::string maxProtString;
+ maxProtString += (maxProt & VM_PROT_READ) ? "r" : "-";
+ maxProtString += (maxProt & VM_PROT_WRITE) ? "w" : "-";
+ maxProtString += (maxProt & VM_PROT_EXECUTE) ? "x" : "-";
+
+ printf("%20s %4lluMB, file offset: #%u/0x%08llX -> 0x%08llX, address: 0x%08llX -> 0x%08llX, %s -> %s\n",
+ mappingName, vmSize / (1024*1024), cacheFileIndex, fileOffset, fileOffset + vmSize,
+ unslidVMAddr, unslidVMAddr + vmSize, initProtString.c_str(), maxProtString.c_str());
+ if (header->mappingOffset >= offsetof(dyld_cache_header, dynamicDataOffset)) {
+ if ( (unslidVMAddr + vmSize) == (header->sharedRegionStart + header->dynamicDataOffset) ) {
+ printf(" dynamic config %4lluKB, address: 0x%08llX -> 0x%08llX\n",
+ header->dynamicDataMaxSize/1024, header->sharedRegionStart + header->dynamicDataOffset,
+ header->sharedRegionStart + header->dynamicDataOffset + header->dynamicDataMaxSize);
+ }
+ }
}, ^(const DyldSharedCache* subCache, uint32_t cacheFileIndex) {
const dyld_cache_header* subCacheHeader = &subCache->header;
- if ( subCacheHeader->codeSignatureOffset != 0 ) {
- uint64_t size = subCacheHeader->codeSignatureSize;
- uint64_t csAddr = subCache->getCodeSignAddress();
- if ( size != 0 )
- printf("%16s %4lluMB, file offset: #%u/0x%08llX -> 0x%08llX, address: 0x%08llX -> 0x%08llX\n",
- "code sign", size/(1024*1024), cacheFileIndex,
- subCacheHeader->codeSignatureOffset, subCacheHeader->codeSignatureOffset + size, csAddr, csAddr + size);
- }
-
- if ( subCacheHeader->mappingOffset > __offsetof(dyld_cache_header, rosettaReadOnlySize) ) {
+ if ( subCacheHeader->codeSignatureSize != 0) {
+ printf("%20s %4lluMB, file offset: #%u/0x%08llX -> 0x%08llX\n",
+ "code sign", subCacheHeader->codeSignatureSize/(1024*1024), cacheFileIndex,
+ subCacheHeader->codeSignatureOffset, subCacheHeader->codeSignatureOffset + subCacheHeader->codeSignatureSize);
+ }
+
+ if ( subCacheHeader->mappingOffset > offsetof(dyld_cache_header, rosettaReadOnlySize) ) {
if ( subCacheHeader->rosettaReadOnlySize != 0 ) {
printf("Rosetta RO: %4lluMB, address: 0x%08llX -> 0x%08llX\n",
subCacheHeader->rosettaReadOnlySize/(1024*1024), subCacheHeader->rosettaReadOnlyAddr,
@@ -678,6 +1787,38 @@
subCacheHeader->localSymbolsSize/(1024*1024), cacheFileIndex,
subCacheHeader->localSymbolsOffset, subCacheHeader->localSymbolsOffset + subCacheHeader->localSymbolsSize);
});
+ }
+ else if ( options.mode == modeTPROInfo ) {
+ printf("TPRO mappings:\n");
+ __block bool foundMapping = false;
+ dyldCache->forEachTPRORegion(^(const void *content, uint64_t unslidVMAddr, uint64_t vmSize, bool &stopRegion) {
+ printf(" %4lluKB, address: 0x%08llX -> 0x%08llX\n", vmSize / 1024, unslidVMAddr, unslidVMAddr + vmSize);
+ foundMapping = true;
+ });
+ if ( !foundMapping )
+ printf(" none found\n");
+ }
+ else if ( options.mode == modeStats ) {
+ __block std::map<std::string_view, uint64_t> mappingSizes;
+ __block uint64_t totalFileSize = 0;
+ __block uint64_t minVMAddr = UINT64_MAX;
+ __block uint64_t maxVMAddr = 0;
+
+ dyldCache->forEachRange(^(const char *mappingName, uint64_t unslidVMAddr, uint64_t vmSize,
+ uint32_t cacheFileIndex, uint64_t fileOffset, uint32_t initProt, uint32_t maxProt, bool& stopRange) {
+ mappingSizes[mappingName] += vmSize;
+ totalFileSize += vmSize;
+ minVMAddr = std::min(minVMAddr, unslidVMAddr);
+ maxVMAddr = std::max(maxVMAddr, unslidVMAddr + vmSize);
+ }, nullptr);
+
+ uint64_t totalVMSize = maxVMAddr - minVMAddr;
+
+ printf("-stats:\n");
+ printf(" total file size: %lldMB\n", totalFileSize >> 20);
+ printf(" total VM size: %lldMB\n", totalVMSize >> 20);
+ for ( const auto& mappingNameAndSize : mappingSizes )
+ printf(" total VM size (%s): %lldMB\n", mappingNameAndSize.first.data(), mappingNameAndSize.second >> 20);
}
else if ( options.mode == modeTextInfo ) {
const dyld_cache_header* header = &dyldCache->header;
@@ -775,8 +1916,11 @@
printf("local symbols by dylib (count=%d):\n", entriesCount);
#endif
}
- else if ( options.mode == modeJSONMap ) {
- std::string buffer = dyldCache->generateJSONMap("unknown");
+ else if ( (options.mode == modeJSONMap) || (options.mode == modeVerboseJSONMap) ) {
+ bool verbose = (options.mode == modeVerboseJSONMap);
+ uuid_t uuid;
+ dyldCache->getUUID(uuid);
+ std::string buffer = dyldCache->generateJSONMap("unknown", uuid, verbose);
printf("%s\n", buffer.c_str());
}
else if ( options.mode == modeJSONDependents ) {
@@ -787,26 +1931,20 @@
// The cache has not been slid if we loaded it from disk
bool cacheRebased = (sharedCachePath == nullptr);
dyld3::MachOAnalyzer::VMAddrConverter vmAddrConverter = dyldCache->makeVMAddrConverter(cacheRebased);
- if ( !cacheRebased )
- dyldCache->applyCacheRebases();
-
- uint64_t sharedCacheRelativeSelectorBaseVMAddress = getSharedCacheRelativeSelectorBaseVMAddress(dyldCache, cacheRebased);
-
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+
+ uint64_t sharedCacheRelativeSelectorBaseVMAddress = dyldCache->sharedCacheRelativeSelectorBaseVMAddress();
+
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
__block std::unordered_set<std::string_view> seenStrings;
- const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)mh;
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
int64_t slide = ma->getSlide();
uint32_t pointerSize = ma->pointerSize();
- ma->forEachSection(^(const dyld3::MachOAnalyzer::SectionInfo& info, bool malformedSectionRange, bool& stop) {
- if ( ( (info.sectFlags & SECTION_TYPE) == S_CSTRING_LITERALS ) ) {
- if ( malformedSectionRange ) {
- stop = true;
- return;
- }
- const uint8_t* content = (uint8_t*)(info.sectAddr + slide);
+ ((const Header*)ma)->forEachSection(^(const Header::SectionInfo& info, bool& stop) {
+ if ( ( (info.flags & SECTION_TYPE) == S_CSTRING_LITERALS ) ) {
+ const uint8_t* content = (uint8_t*)(info.address + slide);
const char* s = (char*)content;
- const char* end = s + info.sectSize;
+ const char* end = s + info.size;
while ( s < end ) {
printf("%s: %s\n", installName, s);
seenStrings.insert(s);
@@ -882,12 +2020,12 @@
}
if (printExports) {
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
- const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)mh;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
uint32_t exportTrieRuntimeOffset;
uint32_t exportTrieSize;
if ( ma->hasExportTrie(exportTrieRuntimeOffset, exportTrieSize) ) {
- const uint8_t* start = (uint8_t*)mh + exportTrieRuntimeOffset;
+ const uint8_t* start = (uint8_t*)hdr + exportTrieRuntimeOffset;
const uint8_t* end = start + exportTrieSize;
std::vector<ExportInfoTrie::Entry> exports;
if ( !ExportInfoTrie::parseTrie(start, end, exports) ) {
@@ -895,7 +2033,12 @@
}
for (const ExportInfoTrie::Entry& entry: exports) {
- printf("%s: %s\n", installName, entry.name.c_str());
+ const char* flags = "";
+ if ( entry.info.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER )
+ flags = " (resolver)";
+ if ( entry.info.flags & EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION )
+ flags = " (weak-def)";
+ printf("%s: %s%s\n", installName, entry.name.c_str(), flags);
}
}
});
@@ -903,11 +2046,11 @@
}
else if ( options.mode == modeSectionSizes ) {
__block std::map<std::string, uint64_t> sectionSizes;
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
- const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
- ma->forEachSection(^(const dyld3::MachOAnalyzer::SectionInfo §Info, bool malformedSectionRange, bool &stop) {
- std::string section = std::string(sectInfo.segInfo.segName) + " " + sectInfo.sectName;
- sectionSizes[section] += sectInfo.sectSize;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)hdr;
+ ((const Header*)ma)->forEachSection(^(const Header::SectionInfo §Info, bool &stop) {
+ std::string section = std::string(sectInfo.segmentName) + " " + std::string(sectInfo.sectionName);
+ sectionSizes[section] += sectInfo.size;
});
});
for (const auto& keyAndValue : sectionSizes) {
@@ -915,87 +2058,230 @@
}
}
else if ( options.mode == modeObjCInfo ) {
- const objc_opt::objc_opt_t* objcOpt = dyldCache->objcOpt();
- if ( objcOpt == nullptr ) {
+ if ( !dyldCache->hasOptimizedObjC() ) {
fprintf(stderr, "Error: could not get optimized objc\n");
return 1;
}
- printf("version: %u\n", objcOpt->version);
- printf("flags: 0x%08x\n", objcOpt->flags);
- if ( const objc::SelectorHashTable* selectors = dyldCache->objcOpt()->selectorOpt() ) {
+ printf("version: %u\n", dyldCache->objcOptVersion());
+ printf("flags: 0x%08x\n", dyldCache->objcOptFlags());
+ if ( const objc::SelectorHashTable* selectors = dyldCache->objcSelectorHashTable() ) {
printf("num selectors: %u\n", selectors->occupancy());
- }
- if ( const objc::ClassHashTable* classes = dyldCache->objcOpt()->classOpt() ) {
+ printf("selectors base address: 0x%llx\n",
+ dyldCache->unslidLoadAddress() + ((uint64_t)selectors - (uint64_t)dyldCache));
+ }
+ if ( const objc::ClassHashTable* classes = dyldCache->objcClassHashTable() ) {
printf("num classes: %u\n", classes->occupancy());
- }
- if ( const objc::ProtocolHashTable* protocols = dyldCache->objcOpt()->protocolOpt() ) {
+ printf("classes base address: 0x%llx\n",
+ dyldCache->unslidLoadAddress() + ((uint64_t)classes - (uint64_t)dyldCache));
+ }
+ if ( const objc::ProtocolHashTable* protocols = dyldCache->objcProtocolHashTable() ) {
printf("num protocols: %u\n", protocols->occupancy());
- }
- if ( const void* relativeMethodListSelectorBase = dyldCache->objcOpt()->relativeMethodListsBaseAddress() ) {
+ printf("protocols base address: 0x%llx\n",
+ dyldCache->unslidLoadAddress() + ((uint64_t)protocols - (uint64_t)dyldCache));
+ }
+ if ( const void* relativeMethodListSelectorBase = dyldCache->objcRelativeMethodListsBaseAddress() ) {
printf("method list selector base address: 0x%llx\n", dyldCache->unslidLoadAddress() + ((uint64_t)relativeMethodListSelectorBase - (uint64_t)dyldCache));
printf("method list selector base value: \"%s\"\n", (const char*)relativeMethodListSelectorBase);
}
+
+ // Dump the objc indices
+
+ __block std::map<uint64_t, const char*> dylibVMAddrMap;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ if ( hdr->hasObjC() )
+ dylibVMAddrMap[hdr->preferredLoadAddress()] = installName;
+ });
+
+ std::vector<std::pair<std::string_view, const objc::objc_image_info*>> objcDylibs;
+
+ const objc::HeaderInfoRO* headerInfoRO = dyldCache->objcHeaderInfoRO();
+ const bool is64 = (strstr(dyldCache->archName(), "64") != nullptr) && (strstr(dyldCache->archName(), "64_32") == nullptr);
+ if ( is64 ) {
+ const auto* headerInfo64 = (objc::objc_headeropt_ro_t<uint64_t>*)headerInfoRO;
+ uint64_t headerInfoVMAddr = dyldCache->unslidLoadAddress();
+ headerInfoVMAddr += (uint64_t)headerInfo64 - (uint64_t)dyldCache;
+ for ( std::pair<uint64_t, const char*> vmAddrAndName : dylibVMAddrMap ) {
+ const objc::objc_header_info_ro_t<uint64_t>* element = headerInfo64->get(headerInfoVMAddr, vmAddrAndName.first);
+ if ( element != nullptr ) {
+ objcDylibs.resize(headerInfo64->index(element) + 1);
+ objcDylibs[headerInfo64->index(element)] = { vmAddrAndName.second, (const objc::objc_image_info*)element->imageInfo() };
+ }
+ }
+ } else {
+ const auto* headerInfo32 = (objc::objc_headeropt_ro_t<uint32_t>*)headerInfoRO;
+ uint64_t headerInfoVMAddr = dyldCache->unslidLoadAddress();
+ headerInfoVMAddr += (uint64_t)headerInfo32 - (uint64_t)dyldCache;
+ for ( std::pair<uint64_t, const char*> vmAddrAndName : dylibVMAddrMap ) {
+ const objc::objc_header_info_ro_t<uint32_t>* element = headerInfo32->get(headerInfoVMAddr, vmAddrAndName.first);
+ if ( element != nullptr ) {
+ objcDylibs.resize(headerInfo32->index(element) + 1);
+ objcDylibs[headerInfo32->index(element)] = { vmAddrAndName.second, (const objc::objc_image_info*)element->imageInfo() };
+ }
+ }
+ }
+
+ printf("num objc dylibs: %lu\n", objcDylibs.size());
+ for ( uint32_t i = 0; i != objcDylibs.size(); ++i ) {
+ const std::pair<std::string_view, const objc::objc_image_info*> objcDylib = objcDylibs[i];
+
+ // Try work out which flags we have
+ std::string flagsStr;
+ uint32_t flags = objcDylib.second->flags;
+ std::pair<uint32_t, const char*> flagComponents[] = {
+ { 1 << 0, "dyldCategories" },
+ { 1 << 1, "supportsGC" },
+ { 1 << 2, "requiresGC" },
+ { 1 << 3, "optimizedByDyld" },
+ { 1 << 4, "signedClassRO" },
+ { 1 << 5, "isSimulated" },
+ { 1 << 6, "hasCategoryClassProperties" },
+ { 1 << 7, "optimizedByDyldClosure" },
+ { 0xFF << 8, "swiftUnstableVersion" },
+ { 0xFFFF << 16, "swiftVersion" },
+ };
+ bool needsSeparator = false;
+ for ( auto [mask, name] : flagComponents ) {
+ if ( (flags & mask) != 0 ) {
+ if ( needsSeparator )
+ flagsStr += " | ";
+ needsSeparator = true;
+
+ flagsStr += name;
+ }
+ }
+ printf("dylib[%d]: { 0x%x, 0x%08x } (%s) %s\n",
+ i, objcDylib.second->version, objcDylib.second->flags,
+ flagsStr.c_str(), objcDylib.first.data());
+ }
}
else if ( options.mode == modeObjCProtocols ) {
- if ( dyldCache->objcOpt() == nullptr ) {
+ if ( !dyldCache->hasOptimizedObjC() ) {
fprintf(stderr, "Error: could not get optimized objc\n");
return 1;
}
- const objc::ProtocolHashTable* protocols = dyldCache->objcOpt()->protocolOpt();
+ const objc::ProtocolHashTable* protocols = dyldCache->objcProtocolHashTable();
if ( protocols == nullptr ) {
fprintf(stderr, "Error: could not get optimized objc protocols\n");
return 1;
}
- protocols->forEachProtocol(^(uint32_t bucketIndex, const char* protocolName, const dyld3::Array<uint64_t>& implCacheOffsets) {
-
- if ( implCacheOffsets.empty() ) {
+ __block std::map<uint64_t, const char*> dylibVMAddrMap;
+ dyldCache->forEachImage(^(const Header *hdr, const char *installName) {
+ if ( hdr->hasObjC() )
+ dylibVMAddrMap[hdr->preferredLoadAddress()] = installName;
+ });
+
+ __block std::map<uint16_t, const char*> dylibMap;
+
+ const objc::HeaderInfoRO* headerInfoRO = dyldCache->objcHeaderInfoRO();
+ const bool is64 = (strstr(dyldCache->archName(), "64") != nullptr) && (strstr(dyldCache->archName(), "64_32") == nullptr);
+ if ( is64 ) {
+ const auto* headerInfo64 = (objc::objc_headeropt_ro_t<uint64_t>*)headerInfoRO;
+ uint64_t headerInfoVMAddr = dyldCache->unslidLoadAddress();
+ headerInfoVMAddr += (uint64_t)headerInfo64 - (uint64_t)dyldCache;
+ for ( std::pair<uint64_t, const char*> vmAddrAndName : dylibVMAddrMap ) {
+ const objc::objc_header_info_ro_t<uint64_t>* element = headerInfo64->get(headerInfoVMAddr, vmAddrAndName.first);
+ if ( element != nullptr ) {
+ dylibMap[headerInfo64->index(element)] = vmAddrAndName.second;
+ }
+ }
+ } else {
+ const auto* headerInfo32 = (objc::objc_headeropt_ro_t<uint32_t>*)headerInfoRO;
+ uint64_t headerInfoVMAddr = dyldCache->unslidLoadAddress();
+ headerInfoVMAddr += (uint64_t)headerInfo32 - (uint64_t)dyldCache;
+ for ( std::pair<uint64_t, const char*> vmAddrAndName : dylibVMAddrMap ) {
+ const objc::objc_header_info_ro_t<uint32_t>* element = headerInfo32->get(headerInfoVMAddr, vmAddrAndName.first);
+ if ( element != nullptr ) {
+ dylibMap[headerInfo32->index(element)] = vmAddrAndName.second;
+ }
+ }
+ }
+
+ typedef objc::ProtocolHashTable::ObjectAndDylibIndex ObjectAndDylibIndex;
+ protocols->forEachProtocol(^(uint32_t bucketIndex, const char* protocolName,
+ const dyld3::Array<ObjectAndDylibIndex>& implCacheInfos) {
+
+ if ( implCacheInfos.empty() ) {
// Empty bucket
- fprintf(stderr, "[% 5d]\n", bucketIndex);
+ printf("[% 5d]\n", bucketIndex);
return;
}
- if ( implCacheOffsets.count() == 1 ) {
+ if ( implCacheInfos.count() == 1 ) {
// No duplicates
- fprintf(stderr, "[% 5d] -> (% 8lld) = %s\n", bucketIndex, implCacheOffsets[0], protocolName);
+ printf("[% 5d] -> (% 8lld, %4d) = %s (in %s)\n",
+ bucketIndex, implCacheInfos[0].first, implCacheInfos[0].second, protocolName,
+ dylibMap.at(implCacheInfos[0].second));
return;
}
// class appears in more than one header
- fprintf(stderr, "[% 5d] -> %lu duplicates = %s\n", bucketIndex, implCacheOffsets.count(), protocolName);
- for (uint64_t cacheOffset : implCacheOffsets) {
- fprintf(stderr, " - [% 5d] -> (% 8lld) = %s\n", bucketIndex, cacheOffset, protocolName);
+ fprintf(stderr, "[% 5d] -> %llu duplicates = %s\n", bucketIndex, implCacheInfos.count(), protocolName);
+ for (const ObjectAndDylibIndex& objectInfo : implCacheInfos) {
+ printf(" - [% 5d] -> (% 8lld, %4d) = %s in (%s)\n",
+ bucketIndex, objectInfo.first, objectInfo.second, protocolName,
+ dylibMap.at(objectInfo.second));
}
});
}
+ else if ( options.mode == modeObjCClassHashTable ) {
+ if ( !dyldCache->hasOptimizedObjC() ) {
+ fprintf(stderr, "Error: could not get optimized objc\n");
+ return 1;
+ }
+ const objc::ClassHashTable* classes = dyldCache->objcClassHashTable();
+ if ( classes == nullptr ) {
+ fprintf(stderr, "Error: could not get optimized objc classes\n");
+ return 1;
+ }
+
+ typedef objc::ClassHashTable::ObjectAndDylibIndex ObjectAndDylibIndex;
+ classes->forEachClass(^(uint32_t bucketIndex, const char* className,
+ const dyld3::Array<ObjectAndDylibIndex>& implCacheInfos) {
+ if ( implCacheInfos.empty() ) {
+ // Empty bucket
+ printf("[% 5d]\n", bucketIndex);
+ return;
+ }
+
+ if ( implCacheInfos.count() == 1 ) {
+ // No duplicates
+ printf("[% 5d] -> (% 8lld, %4d) = %s\n",
+ bucketIndex, implCacheInfos[0].first, implCacheInfos[0].second, className);
+ return;
+ }
+
+ // class appears in more than one header
+ printf("[% 5d] -> %llu duplicates = %s\n", bucketIndex, implCacheInfos.count(), className);
+ for (const ObjectAndDylibIndex& objectInfo : implCacheInfos) {
+ printf(" - [% 5d] -> (% 8lld, %4d) = %s\n",
+ bucketIndex, objectInfo.first, objectInfo.second, className);
+ }
+ });
+ }
else if ( options.mode == modeObjCClasses ) {
// If we are running on macOS against a cache for another device, then we need a root path to find on-disk dylibs/executables
- if ( (dyld3::Platform)dyld_get_active_platform() != dyldCache->platform() ) {
+ if ( Platform(dyld_get_active_platform()) != dyldCache->platform() ) {
if ( options.rootPath == nullptr ) {
fprintf(stderr, "Analyzing cache file requires a root path for on-disk binaries. Rerun with -fs-root *path*\n");
return 1;
}
}
- dyldCache->applyCacheRebases();
-
- auto getString = ^const char *(const dyld3::MachOAnalyzer* ma, uint64_t nameVMAddr){
+ auto getString = ^const char *(const dyld3::MachOAnalyzer* ma, VMAddress nameVMAddr){
dyld3::MachOAnalyzer::PrintableStringResult result;
- const char* name = ma->getPrintableString(nameVMAddr, result);
+ const char* name = ma->getPrintableString(nameVMAddr.rawValue(), result);
if (result == dyld3::MachOAnalyzer::PrintableStringResult::CanPrint)
return name;
return nullptr;
};
- // We don't actually slide the cache. It still contains unslid VMAddr's
- const bool rebased = false;
-
- uint64_t sharedCacheRelativeSelectorBaseVMAddress = getSharedCacheRelativeSelectorBaseVMAddress(dyldCache, rebased);
-
- using dyld3::json::Node;
- using dyld3::json::NodeValueType;
+ uint64_t sharedCacheRelativeSelectorBaseVMAddress = dyldCache->sharedCacheRelativeSelectorBaseVMAddress();
+
+ using json::Node;
+ using json::NodeValueType;
std::string instancePrefix("-");
std::string classPrefix("+");
@@ -1004,26 +2290,23 @@
// name of the class they are attaching to
__block std::unordered_map<uint64_t, const char*> classVMAddrToName;
__block std::unordered_map<uint64_t, const char*> metaclassVMAddrToName;
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
- const uint32_t pointerSize = ma->pointerSize();
-
- auto visitClass = ^(uint64_t classVMAddr,
- uint64_t classSuperclassVMAddr, uint64_t classDataVMAddr,
- const dyld3::MachOAnalyzer::ObjCClassInfo& objcClass, bool isMetaClass,
- bool& stop) {
- if (auto className = getString(ma, objcClass.nameVMAddr(pointerSize))) {
- if (isMetaClass)
- metaclassVMAddrToName[classVMAddr] = className;
+
+ __block objc_visitor::Visitor visitor(dyldCache, ma, VMAddress(sharedCacheRelativeSelectorBaseVMAddress));
+
+ auto visitClass = ^(const objc_visitor::Class& objcClass, bool& stopClass) {
+ VMAddress classVMAddr = objcClass.getVMAddress();
+ VMAddress nameVMAddr = objcClass.getNameVMAddr(visitor);
+ if ( auto className = getString(ma, nameVMAddr) ) {
+ if ( objcClass.isMetaClass )
+ metaclassVMAddrToName[classVMAddr.rawValue()] = className;
else
- classVMAddrToName[classVMAddr] = className;
+ classVMAddrToName[classVMAddr.rawValue()] = className;
}
};
- Diagnostics diag;
-
- dyld3::MachOAnalyzer::VMAddrConverter vmAddrConverter = dyldCache->makeVMAddrConverter(rebased);
- ma->forEachObjCClass(diag, vmAddrConverter, visitClass);
+ visitor.forEachClassAndMetaClass(visitClass);
});
// These are used only for the on-disk binaries we analyze
@@ -1031,116 +2314,110 @@
__block std::unordered_map<uint64_t, const char*> onDiskClassVMAddrToName;
__block std::unordered_map<uint64_t, const char*> onDiskMetaclassVMAddrToName;
- auto getProperties = ^(const dyld3::MachOAnalyzer* ma, uint64_t propertiesVMAddr,
- const dyld3::MachOAnalyzer::VMAddrConverter& vmAddrConverter) {
- __block Node propertiesNode;
- auto visitProperty = ^(uint64_t propertyVMAddr, const dyld3::MachOAnalyzer::ObjCProperty& property) {
+ auto getProperties = ^(const dyld3::MachOAnalyzer* ma, const objc_visitor::PropertyList& propertyList,
+ objc_visitor::Visitor& visitor) {
+ Node propertiesNode;
+
+ for ( uint32_t i = 0, numProperties = propertyList.numProperties(); i != numProperties; ++i ) {
+ objc_visitor::Property property = propertyList.getProperty(visitor, i);
+
// Get the name && attributes
- auto propertyName = getString(ma, property.nameVMAddr);
- auto propertyAttributes = getString(ma, property.attributesVMAddr);
-
- if (!propertyName || !propertyAttributes)
- return;
+ const char* propertyName = property.getName(visitor);
+ const char* propertyAttributes = property.getAttributes(visitor);
Node propertyNode;
propertyNode.map["name"] = Node{propertyName};
propertyNode.map["attributes"] = Node{propertyAttributes};
propertiesNode.array.push_back(propertyNode);
- };
- ma->forEachObjCProperty(propertiesVMAddr, vmAddrConverter, visitProperty);
+ }
+
return propertiesNode.array.empty() ? std::optional<Node>() : propertiesNode;
};
- auto getClassProtocols = ^(const dyld3::MachOAnalyzer* ma, uint64_t protocolsVMAddr,
- const dyld3::MachOAnalyzer::VMAddrConverter& vmAddrConverter) {
- __block Node protocolsNode;
-
- auto visitProtocol = ^(uint64_t protocolVMAddr, const dyld3::MachOAnalyzer::ObjCProtocol& protocol) {
- if (const char *name = getString(ma, protocol.nameVMAddr)) {
+ auto getClassProtocols = ^(const dyld3::MachOAnalyzer* ma, const objc_visitor::ProtocolList& protocolList,
+ objc_visitor::Visitor& visitor) {
+ Node protocolsNode;
+
+ for ( uint64_t i = 0, numProtocols = protocolList.numProtocols(visitor); i != numProtocols; ++i ) {
+ objc_visitor::Protocol protocol = protocolList.getProtocol(visitor, i);
+
+ if ( const char *name = getString(ma, protocol.getNameVMAddr(visitor)) ) {
protocolsNode.array.push_back(Node{name});
}
- };
-
- ma->forEachObjCProtocol(protocolsVMAddr, vmAddrConverter, visitProtocol);
+ }
return protocolsNode.array.empty() ? std::optional<Node>() : protocolsNode;
};
auto getProtocols = ^(const dyld3::MachOAnalyzer* ma,
- const dyld3::MachOAnalyzer::VMAddrConverter& vmAddrConverter) {
- __block Node protocols;
-
- auto getMethods = ^(const dyld3::MachOAnalyzer* mh, uint64_t methodListVMAddr, const std::string &prefix, Node &node){
- auto visitMethod = ^(uint64_t methodVMAddr, const dyld3::MachOAnalyzer::ObjCMethod& method,
- bool& stopMethod) {
- if (auto name = getString(mh, method.nameVMAddr)) {
+ objc_visitor::Visitor& visitor) {
+ auto getMethods = ^(const dyld3::MachOAnalyzer* mh, objc_visitor::MethodList methodList,
+ const std::string &prefix, Node &node){
+ for ( uint32_t i = 0, numMethods = methodList.numMethods(); i != numMethods; ++i ) {
+ objc_visitor::Method objcMethod = methodList.getMethod(visitor, i);
+
+ if ( auto name = getString(mh, objcMethod.getNameVMAddr(visitor)) ) {
node.array.push_back(Node{prefix + name});
}
- };
-
- ma->forEachObjCMethod(methodListVMAddr, vmAddrConverter, sharedCacheRelativeSelectorBaseVMAddress, visitMethod);
+ }
};
- auto visitProtocol = ^(uint64_t protoVMAddr,
- const dyld3::MachOAnalyzer::ObjCProtocol& objcProto,
- bool& stopProtocol) {
- const char* protoName = getString(ma, objcProto.nameVMAddr);
- if (!protoName)
+ __block Node protocolsNode;
+ auto visitProtocol = ^(const objc_visitor::Protocol& objcProtocol, bool& stopProtocol) {
+ const char* protoName = getString(ma, objcProtocol.getNameVMAddr(visitor));
+ if ( !protoName )
return;
Node entry;
entry.map["protocolName"] = Node{protoName};
- if ( objcProto.protocolsVMAddr != 0 ) {
- __block Node visitedProtocols;
-
- auto visitProtocolInner = ^(uint64_t protocolRefVMAddr, const dyld3::MachOAnalyzer::ObjCProtocol& protocol) {
- if (auto name = getString(ma, protocol.nameVMAddr)) {
+ objc_visitor::ProtocolList protocolList = objcProtocol.getProtocols(visitor);
+ if ( uint64_t numProtocols = protocolList.numProtocols(visitor); numProtocols != 0 ) {
+ Node visitedProtocols;
+
+ for ( uint32_t i = 0; i != numProtocols; ++i ) {
+ objc_visitor::Protocol innerProtocol = protocolList.getProtocol(visitor, i);
+
+ if ( const char* name = getString(ma, innerProtocol.getNameVMAddr(visitor)) )
visitedProtocols.array.push_back(Node{name});
- }
- };
-
- ma->forEachObjCProtocol(objcProto.protocolsVMAddr, vmAddrConverter, visitProtocolInner);
+ }
+
if (!visitedProtocols.array.empty()) {
entry.map["protocols"] = visitedProtocols;
}
}
Node methods;
- getMethods(ma, objcProto.instanceMethodsVMAddr, instancePrefix, methods);
- getMethods(ma, objcProto.classMethodsVMAddr, classPrefix, methods);
+ getMethods(ma, objcProtocol.getInstanceMethods(visitor), instancePrefix, methods);
+ getMethods(ma, objcProtocol.getClassMethods(visitor), classPrefix, methods);
if (!methods.array.empty()) {
entry.map["methods"] = methods;
}
Node optMethods;
- getMethods(ma, objcProto.optionalInstanceMethodsVMAddr, instancePrefix, optMethods);
- getMethods(ma, objcProto.optionalClassMethodsVMAddr, classPrefix, optMethods);
+ getMethods(ma, objcProtocol.getOptionalInstanceMethods(visitor), instancePrefix, optMethods);
+ getMethods(ma, objcProtocol.getOptionalClassMethods(visitor), classPrefix, optMethods);
if (!optMethods.array.empty()) {
entry.map["optionalMethods"] = optMethods;
}
- protocols.array.push_back(entry);
+ protocolsNode.array.push_back(entry);
};
- Diagnostics diag;
- ma->forEachObjCProtocol(diag, vmAddrConverter, visitProtocol);
-
- return protocols.array.empty() ? std::optional<Node>() : protocols;
+ visitor.forEachProtocol(visitProtocol);
+
+ return protocolsNode.array.empty() ? std::optional<Node>() : protocolsNode;
};
auto getSelRefs = ^(const dyld3::MachOAnalyzer* ma,
- const dyld3::MachOAnalyzer::VMAddrConverter& vmAddrConverter) {
+ objc_visitor::Visitor& visitor) {
__block std::vector<const char *> selNames;
- auto visitSelRef = ^(uint64_t selRefVMAddr, uint64_t selRefTargetVMAddr, bool& stop) {
- if (auto selValue = getString(ma, selRefTargetVMAddr)) {
+ visitor.forEachSelectorReference(^(VMAddress selRefVMAddr, VMAddress selRefTargetVMAddr, const char *selectorString) {
+ if ( auto selValue = getString(ma, selRefTargetVMAddr) ) {
selNames.push_back(selValue);
}
- };
-
- Diagnostics diag;
- ma->forEachObjCSelectorReference(diag, vmAddrConverter, visitSelRef);
+ });
std::sort(selNames.begin(), selNames.end(),
[](const char* a, const char* b) {
@@ -1156,15 +2433,15 @@
};
auto getClasses = ^(const dyld3::MachOAnalyzer* ma,
- const dyld3::MachOAnalyzer::VMAddrConverter& vmAddrConverter) {
- Diagnostics diag;
+ objc_visitor::Visitor& visitor) {
const uint32_t pointerSize = ma->pointerSize();
+ const uint16_t chainedPointerFormat = ma->hasChainedFixups() ? ma->chainedPointerFormat() : 0;
// Get the vmAddrs for all exported symbols as we want to know if classes
// are exported
std::set<uint64_t> exportedSymbolVMAddrs;
{
- uint64_t loadAddress = ma->preferredLoadAddress();
+ uint64_t loadAddress = ((const Header*)ma)->preferredLoadAddress();
uint32_t exportTrieRuntimeOffset;
uint32_t exportTrieSize;
@@ -1182,11 +2459,8 @@
__block Node classesNode;
__block bool skippedPreviousClass = false;
- auto visitClass = ^(uint64_t classVMAddr,
- uint64_t classSuperclassVMAddr, uint64_t classDataVMAddr,
- const dyld3::MachOAnalyzer::ObjCClassInfo& objcClass, bool isMetaClass,
- bool& stopClass) {
- if (isMetaClass) {
+ auto visitClass = ^(const objc_visitor::Class& objcClass, bool& stopClass) {
+ if ( objcClass.isMetaClass ) {
if (skippedPreviousClass) {
// If the class was bad, then skip the meta class too
skippedPreviousClass = false;
@@ -1197,93 +2471,111 @@
}
std::string classType = "-";
- if (isMetaClass)
+ if ( objcClass.isMetaClass )
classType = "+";
+
+ VMAddress classVMAddr = objcClass.getVMAddress();
+ VMAddress nameVMAddr = objcClass.getNameVMAddr(visitor);
+
dyld3::MachOAnalyzer::PrintableStringResult classNameResult;
- const char* className = ma->getPrintableString(objcClass.nameVMAddr(pointerSize), classNameResult);
- if (classNameResult != dyld3::MachOAnalyzer::PrintableStringResult::CanPrint) {
+ const char* className = ma->getPrintableString(nameVMAddr.rawValue(), classNameResult);
+ if ( classNameResult != dyld3::MachOAnalyzer::PrintableStringResult::CanPrint ) {
return;
}
- const char* superClassName = nullptr;
+ __block const char* superClassName = nullptr;
if ( DyldSharedCache::inDyldCache(dyldCache, ma) ) {
- if ( objcClass.superclassVMAddr != 0 ) {
- if (isMetaClass) {
+ std::optional<VMAddress> superclassVMAddr = objcClass.getSuperclassVMAddr(visitor);
+ if ( superclassVMAddr.has_value() ) {
+ if ( objcClass.isMetaClass ) {
// If we are root class, then our superclass should actually point to our own class
- const uint32_t RO_ROOT = (1<<1);
- if ( objcClass.flags(pointerSize) & RO_ROOT ) {
- auto it = classVMAddrToName.find(objcClass.superclassVMAddr);
+ if ( objcClass.isRootClass(visitor) ) {
+ auto it = classVMAddrToName.find(superclassVMAddr.value().rawValue());
assert(it != classVMAddrToName.end());
superClassName = it->second;
} else {
- auto it = metaclassVMAddrToName.find(objcClass.superclassVMAddr);
+ auto it = metaclassVMAddrToName.find(superclassVMAddr.value().rawValue());
assert(it != metaclassVMAddrToName.end());
superClassName = it->second;
}
} else {
- auto it = classVMAddrToName.find(objcClass.superclassVMAddr);
+ auto it = classVMAddrToName.find(superclassVMAddr.value().rawValue());
assert(it != classVMAddrToName.end());
superClassName = it->second;
}
}
} else {
// On-disk binary. Lets crack the chain to work out what we are pointing at
- dyld3::MachOAnalyzer::ChainedFixupPointerOnDisk fixup;
- if ( pointerSize == 8 )
- fixup.raw64 = objcClass.superclassVMAddr;
- else
- fixup.raw32 = (uint32_t)objcClass.superclassVMAddr;
- uint32_t bindOrdinal;
- int64_t embeddedAddend;
- if (fixup.isBind(vmAddrConverter.chainedPointerFormat, bindOrdinal, embeddedAddend)) {
- // Bind to another image. Use the bind table to work out which name to bind to
- const char* symbolName = onDiskChainedFixupBindTargets[(size_t)bindOrdinal];
- if (isMetaClass) {
- if ( strstr(symbolName, "_OBJC_METACLASS_$_") == symbolName ) {
- superClassName = symbolName + strlen("_OBJC_METACLASS_$_");
+ objcClass.withSuperclass(visitor, ^(const dyld3::MachOFile::ChainedFixupPointerOnDisk* fixup, uint16_t) {
+ if ( (pointerSize == 8) && (fixup->raw64 == 0) )
+ return;
+ else if ( (pointerSize == 4) && (fixup->raw32 == 0) )
+ return;
+
+ uint32_t bindOrdinal;
+ int64_t embeddedAddend;
+ if ( fixup->isBind(chainedPointerFormat, bindOrdinal, embeddedAddend) ) {
+ // Bind to another image. Use the bind table to work out which name to bind to
+ const char* symbolName = onDiskChainedFixupBindTargets[(size_t)bindOrdinal];
+ if ( objcClass.isMetaClass ) {
+ if ( strstr(symbolName, "_OBJC_METACLASS_$_") == symbolName ) {
+ superClassName = symbolName + strlen("_OBJC_METACLASS_$_");
+ } else {
+ // Swift classes don't start with these prefixes so just skip them
+ if ( objcClass.isSwiftLegacy(visitor) || objcClass.isSwiftStable(visitor) )
+ return;
+ }
} else {
- // Swift classes don't start with these prefixes so just skip them
- if (objcClass.isSwiftLegacy || objcClass.isSwiftStable)
- return;
+ if ( strstr(symbolName, "_OBJC_CLASS_$_") == symbolName ) {
+ superClassName = symbolName + strlen("_OBJC_CLASS_$_");
+ } else {
+ // Swift classes don't start with these prefixes so just skip them
+ if ( objcClass.isSwiftLegacy(visitor) || objcClass.isSwiftStable(visitor) )
+ return;
+ }
}
} else {
- if ( strstr(symbolName, "_OBJC_CLASS_$_") == symbolName ) {
- superClassName = symbolName + strlen("_OBJC_CLASS_$_");
+ // Rebase within this image.
+ std::optional<VMAddress> superclassVMAddr = objcClass.getSuperclassVMAddr(visitor);
+ if ( objcClass.isMetaClass ) {
+ auto it = onDiskMetaclassVMAddrToName.find(superclassVMAddr.value().rawValue());
+ assert(it != onDiskMetaclassVMAddrToName.end());
+ superClassName = it->second;
} else {
- // Swift classes don't start with these prefixes so just skip them
- if (objcClass.isSwiftLegacy || objcClass.isSwiftStable)
- return;
+ auto it = onDiskClassVMAddrToName.find(superclassVMAddr.value().rawValue());
+ assert(it != onDiskClassVMAddrToName.end());
+ superClassName = it->second;
}
}
- } else {
- // Rebase within this image.
- if (isMetaClass) {
- auto it = onDiskMetaclassVMAddrToName.find(objcClass.superclassVMAddr);
- assert(it != onDiskMetaclassVMAddrToName.end());
- superClassName = it->second;
- } else {
- auto it = onDiskClassVMAddrToName.find(objcClass.superclassVMAddr);
- assert(it != onDiskClassVMAddrToName.end());
- superClassName = it->second;
- }
+ });
+
+ if ( !superClassName ) {
+ // Probably a swift class we want to skip
+ return;
}
}
// Print the methods on this class
- __block Node methodsNode;
- auto visitMethod = ^(uint64_t methodVMAddr, const dyld3::MachOAnalyzer::ObjCMethod& method, bool& stopMethod) {
+ Node methodsNode;
+
+ objc_visitor::MethodList objcMethodList = skipListsOfLists(objcClass.getBaseMethods(visitor), visitor);
+
+ for ( uint32_t i = 0, numMethods = objcMethodList.numMethods(); i != numMethods; ++i ) {
+ objc_visitor::Method objcMethod = objcMethodList.getMethod(visitor, i);
+
dyld3::MachOAnalyzer::PrintableStringResult methodNameResult;
- const char* methodName = ma->getPrintableString(method.nameVMAddr, methodNameResult);
+ const char* methodName = ma->getPrintableString(objcMethod.getNameVMAddr(visitor).rawValue(),
+ methodNameResult);
if (methodNameResult != dyld3::MachOAnalyzer::PrintableStringResult::CanPrint)
- return;
+ continue;
+
methodsNode.array.push_back(Node{classType + methodName});
- };
- ma->forEachObjCMethod(objcClass.baseMethodsVMAddr(pointerSize), vmAddrConverter,
- sharedCacheRelativeSelectorBaseVMAddress, visitMethod);
-
- std::optional<Node> properties = getProperties(ma, objcClass.basePropertiesVMAddr(pointerSize), vmAddrConverter);
-
- if (isMetaClass) {
+ }
+
+ objc_visitor::PropertyList propertyList = skipListsOfLists(objcClass.getBaseProperties(visitor), visitor);
+ std::optional<Node> properties = getProperties(ma, propertyList, visitor);
+
+ if ( objcClass.isMetaClass ) {
assert(!classesNode.array.empty());
Node& currentClassNode = classesNode.array.back();
assert(currentClassNode.map["className"].value == className);
@@ -1302,6 +2594,8 @@
return;
}
+ objc_visitor::ProtocolList protocolList = skipListsOfLists(objcClass.getBaseProtocols(visitor), visitor);
+
Node currentClassNode;
currentClassNode.map["className"] = Node{className};
if ( superClassName != nullptr )
@@ -1310,10 +2604,10 @@
currentClassNode.map["methods"] = methodsNode;
if (properties.has_value())
currentClassNode.map["properties"] = properties.value();
- if (std::optional<Node> protocols = getClassProtocols(ma, objcClass.baseProtocolsVMAddr(pointerSize), vmAddrConverter))
+ if (std::optional<Node> protocols = getClassProtocols(ma, protocolList, visitor))
currentClassNode.map["protocols"] = protocols.value();
- currentClassNode.map["exported"] = Node{exportedSymbolVMAddrs.count(classVMAddr) != 0};
+ currentClassNode.map["exported"] = Node{exportedSymbolVMAddrs.count(classVMAddr.rawValue()) != 0};
// We didn't skip this class so mark it as such
skippedPreviousClass = false;
@@ -1321,32 +2615,37 @@
classesNode.array.push_back(currentClassNode);
};
- ma->forEachObjCClass(diag, vmAddrConverter, visitClass);
+ visitor.forEachClassAndMetaClass(visitClass);
+
return classesNode.array.empty() ? std::optional<Node>() : classesNode;
};
auto getCategories = ^(const dyld3::MachOAnalyzer* ma,
- const dyld3::MachOAnalyzer::VMAddrConverter& vmAddrConverter) {
- Diagnostics diag;
-
+ objc_visitor::Visitor& visitor) {
const uint32_t pointerSize = ma->pointerSize();
+ const uint16_t chainedPointerFormat = ma->hasChainedFixups() ? ma->chainedPointerFormat() : 0;
__block Node categoriesNode;
- auto visitCategory = ^(uint64_t categoryVMAddr,
- const dyld3::MachOAnalyzer::ObjCCategory& objcCategory,
- bool& stopCategory) {
+ auto visitCategory = ^(const objc_visitor::Category& objcCategory, bool& stopCategory) {
+ VMAddress nameVMAddr = objcCategory.getNameVMAddr(visitor);
+
dyld3::MachOAnalyzer::PrintableStringResult categoryNameResult;
- const char* categoryName = ma->getPrintableString(objcCategory.nameVMAddr, categoryNameResult);
+ const char* categoryName = ma->getPrintableString(nameVMAddr.rawValue(), categoryNameResult);
if (categoryNameResult != dyld3::MachOAnalyzer::PrintableStringResult::CanPrint)
return;
- const char* className = nullptr;
- if ( DyldSharedCache::inDyldCache(dyldCache, ma) ) {
- // The class might be missing if the target is not in the shared cache. So just skip these ones
- if ( objcCategory.clsVMAddr == 0 )
- return;
-
- auto it = classVMAddrToName.find(objcCategory.clsVMAddr);
+ __block const char* className = nullptr;
+ if ( DyldSharedCache::inDyldCache(dyldCache, ma) ) {
+ // The class might be missing if the target is not in the shared cache. So just skip these ones
+ std::optional<VMAddress> clsVMAddr = objcCategory.getClassVMAddr(visitor);
+ if ( !clsVMAddr.has_value() )
+ return;
+
+ if ( objcCategory.isForSwiftStubClass() ) {
+ // We don't have a class for stub classes, so just use a marker
+ className = "unknown swift stub class";
+ } else {
+ auto it = classVMAddrToName.find(clsVMAddr.value().rawValue());
if (it == classVMAddrToName.end()) {
// This is an odd binary with perhaps a Swift class. Just skip this entry
// Specifically, categories can be attached to "stub classes" which are not in the
@@ -1355,17 +2654,19 @@
return;
}
className = it->second;
- } else {
- // On-disk binary. Lets crack the chain to work out what we are pointing at
- dyld3::MachOAnalyzer::ChainedFixupPointerOnDisk fixup;
- fixup.raw64 = objcCategory.clsVMAddr;
- if ( pointerSize == 8 )
- fixup.raw64 = objcCategory.clsVMAddr;
- else
- fixup.raw32 = (uint32_t)objcCategory.clsVMAddr;
+ }
+ } else {
+ // On-disk binary. Lets crack the chain to work out what we are pointing at
+ objcCategory.withClass(visitor,
+ ^(const dyld3::MachOFile::ChainedFixupPointerOnDisk* fixup, uint16_t) {
+ if ( (pointerSize == 8) && (fixup->raw64 == 0) )
+ return;
+ else if ( (pointerSize == 4) && (fixup->raw32 == 0) )
+ return;
+
uint32_t bindOrdinal;
int64_t embeddedAddend;
- if (fixup.isBind(vmAddrConverter.chainedPointerFormat, bindOrdinal, embeddedAddend)) {
+ if ( fixup->isBind(chainedPointerFormat, bindOrdinal, embeddedAddend) ) {
// Bind to another image. Use the bind table to work out which name to bind to
const char* symbolName = onDiskChainedFixupBindTargets[(size_t)bindOrdinal];
if ( strstr(symbolName, "_OBJC_CLASS_$_") == symbolName ) {
@@ -1376,65 +2677,78 @@
return;
}
} else {
- auto it = onDiskClassVMAddrToName.find(objcCategory.clsVMAddr);
+ std::optional<VMAddress> clsVMAddr = objcCategory.getClassVMAddr(visitor);
+ auto it = onDiskClassVMAddrToName.find(clsVMAddr.value().rawValue());
if (it == onDiskClassVMAddrToName.end()) {
// This is an odd binary with perhaps a Swift class. Just skip this entry
return;
}
className = it->second;
}
- }
+ });
+
+ if ( !className ) {
+ // Probably a swift class we want to skip
+ return;
+ }
+ }
// Print the instance methods on this category
__block Node methodsNode;
- auto visitInstanceMethod = ^(uint64_t methodVMAddr, const dyld3::MachOAnalyzer::ObjCMethod& method, bool& stopMethod) {
- if (auto methodName = getString(ma, method.nameVMAddr))
- methodsNode.array.push_back(Node{instancePrefix + methodName});
- };
- ma->forEachObjCMethod(objcCategory.instanceMethodsVMAddr, vmAddrConverter,
- sharedCacheRelativeSelectorBaseVMAddress, visitInstanceMethod);
-
- // Print the instance methods on this category
- __block Node classMethodsNode;
- auto visitClassMethod = ^(uint64_t methodVMAddr, const dyld3::MachOAnalyzer::ObjCMethod& method, bool& stopMethod) {
- if (auto methodName = getString(ma, method.nameVMAddr))
- methodsNode.array.push_back(Node{classPrefix + methodName});
- };
- ma->forEachObjCMethod(objcCategory.classMethodsVMAddr, vmAddrConverter,
- sharedCacheRelativeSelectorBaseVMAddress, visitClassMethod);
+ {
+ objc_visitor::MethodList objcMethodList = objcCategory.getInstanceMethods(visitor);
+ for ( uint32_t i = 0, numMethods = objcMethodList.numMethods(); i != numMethods; ++i ) {
+ objc_visitor::Method objcMethod = objcMethodList.getMethod(visitor, i);
+
+ if ( auto methodName = getString(ma, objcMethod.getNameVMAddr(visitor)) )
+ methodsNode.array.push_back(Node{instancePrefix + methodName});
+ }
+ }
+
+ // Print the class methods on this category
+ {
+ objc_visitor::MethodList objcMethodList = objcCategory.getClassMethods(visitor);
+ for ( uint32_t i = 0, numMethods = objcMethodList.numMethods(); i != numMethods; ++i ) {
+ objc_visitor::Method objcMethod = objcMethodList.getMethod(visitor, i);
+
+ if ( auto methodName = getString(ma, objcMethod.getNameVMAddr(visitor)) )
+ methodsNode.array.push_back(Node{classPrefix + methodName});
+ }
+ }
Node currentCategoryNode;
currentCategoryNode.map["categoryName"] = Node{categoryName};
currentCategoryNode.map["className"] = Node{className};
if (!methodsNode.array.empty())
currentCategoryNode.map["methods"] = methodsNode;
- if (std::optional<Node> properties = getProperties(ma, objcCategory.instancePropertiesVMAddr, vmAddrConverter))
+ if (std::optional<Node> properties = getProperties(ma, objcCategory.getInstanceProperties(visitor), visitor))
currentCategoryNode.map["properties"] = properties.value();
- if (std::optional<Node> protocols = getClassProtocols(ma, objcCategory.protocolsVMAddr, vmAddrConverter))
+ if (std::optional<Node> protocols = getClassProtocols(ma, objcCategory.getProtocols(visitor), visitor))
currentCategoryNode.map["protocols"] = protocols.value();
categoriesNode.array.push_back(currentCategoryNode);
};
- ma->forEachObjCCategory(diag, vmAddrConverter, visitCategory);
+ visitor.forEachCategory(visitCategory);
return categoriesNode.array.empty() ? std::optional<Node>() : categoriesNode;
};
__block bool needsComma = false;
- dyld3::json::streamArrayBegin(needsComma);
-
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
- dyld3::MachOAnalyzer::VMAddrConverter vmAddrConverter = dyldCache->makeVMAddrConverter(rebased);
+ json::streamArrayBegin(needsComma);
+
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
+
+ objc_visitor::Visitor visitor(dyldCache, ma, VMAddress(sharedCacheRelativeSelectorBaseVMAddress));
Node imageRecord;
imageRecord.map["imagePath"] = Node{installName};
imageRecord.map["imageType"] = Node{"cache-dylib"};
- std::optional<Node> classes = getClasses(ma, vmAddrConverter);
- std::optional<Node> categories = getCategories(ma, vmAddrConverter);
- std::optional<Node> protocols = getProtocols(ma, vmAddrConverter);
- std::optional<Node> selrefs = getSelRefs(ma, vmAddrConverter);
+ std::optional<Node> classes = getClasses(ma, visitor);
+ std::optional<Node> categories = getCategories(ma, visitor);
+ std::optional<Node> protocols = getProtocols(ma, visitor);
+ std::optional<Node> selrefs = getSelRefs(ma, visitor);
// Skip emitting images with no objc data
if (!classes.has_value() && !categories.has_value() && !protocols.has_value() && !selrefs.has_value())
@@ -1448,7 +2762,7 @@
if (selrefs.has_value())
imageRecord.map["selrefs"] = selrefs.value();
- dyld3::json::streamArrayNode(needsComma, imageRecord);
+ json::streamArrayNode(needsComma, imageRecord);
});
const dyld3::MachOAnalyzer* mainMA = nullptr;
@@ -1469,22 +2783,36 @@
}
KernelArgs kernArgs(mainMA, {"test.exe"}, {}, {});
+ Allocator& alloc = MemoryManager::memoryManager().defaultAllocator();
SyscallDelegate osDelegate;
osDelegate._dyldCache = dyldCache;
osDelegate._rootPath = options.rootPath;
-
- __block ProcessConfig config(&kernArgs, osDelegate);
- RuntimeState stateObject(config);
- RuntimeState& state = stateObject;
+ __block ProcessConfig config(&kernArgs, osDelegate, alloc);
+ RuntimeLocks locks;
+ RuntimeState stateObject(config, locks, alloc);
+ RuntimeState& state = stateObject;
config.dyldCache.addr->forEachLaunchLoaderSet(^(const char* executableRuntimePath, const PrebuiltLoaderSet* pbls) {
__block Diagnostics diag;
bool checkIfOSBinary = state.config.process.archs->checksOSBinary();
- state.config.syscall.withReadOnlyMappedFile(diag, executableRuntimePath, checkIfOSBinary, ^(const void* mapping, size_t mappedSize, bool isOSBinary, const FileID& fileID, const char* canonicalPath) {
- if ( const dyld3::MachOFile* mf = dyld3::MachOFile::compatibleSlice(diag, mapping, mappedSize, executableRuntimePath, state.config.process.platform, isOSBinary, *state.config.process.archs) ) {
- const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mf;
- uint32_t pointerSize = ma->pointerSize();
+ state.config.syscall.withReadOnlyMappedFile(diag, executableRuntimePath, checkIfOSBinary, ^(const void* mapping, size_t mappedSize, bool isOSBinary, const FileID& fileID, const char* canonicalPath, const int fileDescriptor) {
+ uint64_t sliceOffset;
+ uint64_t sliceSize;
+ if ( const dyld3::MachOFile* mf = dyld3::MachOFile::compatibleSlice(diag, sliceOffset, sliceSize, mapping, mappedSize, executableRuntimePath, state.config.process.platform, isOSBinary, *state.config.process.archs) ) {
+ dyld3::closure::FileSystemPhysical fileSystem;
+ dyld3::closure::LoadedFileInfo fileInfo = {
+ .fileContent = (void*)mf,
+ .fileContentLen = sliceSize,
+ .sliceOffset = 0,
+ .sliceLen = sliceSize,
+ .isOSBinary = false,
+ .inode = 0,
+ .mtime = 0,
+ .unload = nullptr,
+ .path = executableRuntimePath
+ };
+ const dyld3::MachOAnalyzer* ma = ((const dyld3::MachOAnalyzer*)mf)->remapIfZeroFill(diag, fileSystem, fileInfo);
// Populate the bind targets for classes from other images
onDiskChainedFixupBindTargets.clear();
@@ -1497,30 +2825,29 @@
// Populate the rebase targets for class names
onDiskMetaclassVMAddrToName.clear();
onDiskClassVMAddrToName.clear();
- auto visitClass = ^(uint64_t classVMAddr,
- uint64_t classSuperclassVMAddr, uint64_t classDataVMAddr,
- const dyld3::MachOAnalyzer::ObjCClassInfo& objcClass, bool isMetaClass,
- bool& stopClass) {
- if (auto className = getString(ma, objcClass.nameVMAddr(pointerSize))) {
- if (isMetaClass)
- onDiskMetaclassVMAddrToName[classVMAddr] = className;
+
+ __block objc_visitor::Visitor visitor(dyldCache, ma, std::nullopt);
+
+ auto visitClass = ^(const objc_visitor::Class& objcClass, bool& stopClass) {
+ VMAddress classVMAddr = objcClass.getVMAddress();
+ VMAddress nameVMAddr = objcClass.getNameVMAddr(visitor);
+ if ( auto className = getString(ma, nameVMAddr) ) {
+ if ( objcClass.isMetaClass )
+ onDiskMetaclassVMAddrToName[classVMAddr.rawValue()] = className;
else
- onDiskClassVMAddrToName[classVMAddr] = className;
+ onDiskClassVMAddrToName[classVMAddr.rawValue()] = className;
}
};
- // Get a vmAddrConverter for this on-disk binary. We can't use the shared cache one
- dyld3::MachOAnalyzer::VMAddrConverter onDiskVMAddrConverter = ma->makeVMAddrConverter(rebased);
-
- ma->forEachObjCClass(diag, onDiskVMAddrConverter, visitClass);
+ visitor.forEachClassAndMetaClass(visitClass);
Node imageRecord;
imageRecord.map["imagePath"] = Node{executableRuntimePath};
imageRecord.map["imageType"] = Node{"executable"};
- std::optional<Node> classes = getClasses(ma, onDiskVMAddrConverter);
- std::optional<Node> categories = getCategories(ma, onDiskVMAddrConverter);
+ std::optional<Node> classes = getClasses(ma, visitor);
+ std::optional<Node> categories = getCategories(ma, visitor);
// TODO: protocols
- std::optional<Node> selrefs = getSelRefs(ma, onDiskVMAddrConverter);
+ std::optional<Node> selrefs = getSelRefs(ma, visitor);
// Skip emitting images with no objc data
if (!classes.has_value() && !categories.has_value() && !selrefs.has_value())
@@ -1532,19 +2859,25 @@
if (selrefs.has_value())
imageRecord.map["selrefs"] = selrefs.value();
- dyld3::json::streamArrayNode(needsComma, imageRecord);
+ json::streamArrayNode(needsComma, imageRecord);
}
});
});
- dyld3::json::streamArrayEnd(needsComma);
+ json::streamArrayEnd(needsComma);
+ }
+ else if ( options.mode == modeObjCClassLayout ) {
+ dumpObjCClassLayout(dyldCache);
+ }
+ else if ( options.mode == modeObjCClassMethodLists ) {
+ dumpObjCClassMethodLists(dyldCache);
}
else if ( options.mode == modeObjCSelectors ) {
- if ( dyldCache->objcOpt() == nullptr ) {
+ if ( !dyldCache->hasOptimizedObjC() ) {
fprintf(stderr, "Error: could not get optimized objc\n");
return 1;
}
- const objc::SelectorHashTable* selectors = dyldCache->objcOpt()->selectorOpt();
+ const objc::SelectorHashTable* selectors = dyldCache->objcSelectorHashTable();
if ( selectors == nullptr ) {
fprintf(stderr, "Error: could not get optimized objc selectors\n");
return 1;
@@ -1561,21 +2894,21 @@
return a < b;
});
- dyld3::json::Node root;
+ json::Node root;
for (const char* selName : selNames) {
- dyld3::json::Node selNode;
- selNode.map["selectorName"] = dyld3::json::Node{selName};
- selNode.map["offset"] = dyld3::json::Node{(int64_t)selName - (int64_t)dyldCache};
+ json::Node selNode;
+ selNode.map["selectorName"] = json::Node{selName};
+ selNode.map["offset"] = json::Node{(int64_t)selName - (int64_t)dyldCache};
root.array.push_back(selNode);
}
- dyld3::json::printJSON(root, 0, std::cout);
+ json::printJSON(root, 0, std::cout);
}
else if ( options.mode == modeSwiftProtocolConformances ) {
#if 0
// This would dump the conformances in each binary, not the table in the shared cache
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
Diagnostics diag;
@@ -1603,7 +2936,7 @@
// Find all the symbols. This maps from VM Addresses to symbol name
__block std::unordered_map<uint64_t, std::string_view> symbols;
__block std::unordered_map<uint64_t, std::string_view> dylibs;
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
Diagnostics diag;
ma->forEachGlobalSymbol(diag, ^(const char *symbolName, uint64_t n_value, uint8_t n_type, uint8_t n_sect, uint16_t n_desc, bool &stop) {
@@ -1667,12 +3000,11 @@
auto getDylibForAddress = ^(uint64_t vmAddress) {
__block std::string_view dylibName;
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
if ( !dylibName.empty() )
return;
- const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
- ma->forEachSegment(^(const dyld3::MachOAnalyzer::SegmentInfo &info, bool &stop) {
- if ( (vmAddress >= info.vmAddr) && (vmAddress < (info.vmAddr+info.vmSize)) ) {
+ ((const Header*)mh)->forEachSegment(^(const Header::SegmentInfo &info, bool &stop) {
+ if ( (vmAddress >= info.vmaddr) && (vmAddress < (info.vmaddr + info.vmsize)) ) {
dylibName = installName;
stop = true;
}
@@ -1689,7 +3021,7 @@
return 0;
}
printf("Swift optimization version: %d\n", swiftOptHeader->version);
- if ( swiftOptHeader->version == 1 ) {
+ if ( swiftOptHeader->version == 1 || swiftOptHeader->version == 2 || swiftOptHeader->version == 3 ) {
printf("Type hash table\n");
const SwiftHashTable* typeHashTable = (const SwiftHashTable*)((uint8_t*)dyldCache + swiftOptHeader->typeConformanceHashTableCacheOffset);
typeHashTable->forEachValue(^(uint32_t bucketIndex, const dyld3::Array<SwiftTypeProtocolConformanceLocation>& impls) {
@@ -1842,8 +3174,84 @@
conformance.data(), protoLoc.protocolConformanceCacheOffset, conformanceDylib.data());
}
});
+
+ if ( swiftOptHeader->version >= 2 )
+ printf("Swift prespecialization data offset: 0x%llx\n", swiftOptHeader->prespecializationDataCacheOffset);
} else {
printf("Unhandled version\n");
+ }
+ }
+ else if ( options.mode == modeSwiftPtrTables ) {
+ uint64_t cacheBaseAddr = dyldCache->unslidLoadAddress();
+ const SwiftOptimizationHeader* swiftOptHeader = dyldCache->swiftOpt();
+ if ( swiftOptHeader == nullptr ) {
+ printf("No Swift optimization information present\n");
+ return 0;
+ }
+ printf("Swift optimization version: %d\n", swiftOptHeader->version);
+ if ( swiftOptHeader->version == 3 ) {
+ SymbolicatedCache symbolicatedCache(dyldCache, cacheOnDisk);
+ for ( size_t i = 0; i < SwiftOptimizationHeader::MAX_PRESPECIALIZED_METADATA_TABLES; ++i ) {
+ uint64_t ptrTableOffset = swiftOptHeader->prespecializedMetadataHashTableCacheOffsets[i];
+ if ( ptrTableOffset == 0 )
+ continue;
+
+ printf("Swift prespecialized metadata hash table #%lu\n", i);
+ const SwiftHashTable* ptrTable = (const SwiftHashTable*)((uint8_t*)dyldCache + ptrTableOffset);
+
+ ptrTable->forEachValue(^(uint32_t bucketIndex, const dyld3::Array<PointerHashTableValue>& values) {
+ for ( const PointerHashTableValue& value : values ) {
+ printf(" - k: [ ");
+ const uint64_t* keys = ptrTable->getCacheOffsets(value);
+ for ( size_t numKey = 0; numKey < value.numOffsets; ++numKey ) {
+ if ( numKey > 0 )
+ printf(", ");
+ printf("%s (0x%llx)", symbolicatedCache.symbolNameAt(cacheBaseAddr + keys[numKey]).c_str(), cacheBaseAddr + keys[numKey]);
+ }
+ printf(" ]\n v: %s (0x%llx)\n", symbolicatedCache.symbolNameAt(cacheBaseAddr + value.cacheOffset).c_str(), cacheBaseAddr + value.cacheOffset);
+ }
+ });
+ }
+ } else {
+ printf("Unhandled version\n");
+ }
+ }
+ else if ( options.mode == modeLookupVA ) {
+ CString vaString = options.lookupVA;
+
+ SymbolicatedCache symbolicatedCache(dyldCache, cacheOnDisk);
+
+ while ( !vaString.empty() ) {
+ char* endptr = nullptr;
+ uint64_t addr = strtoull(vaString.c_str(), &endptr, 16);
+ if ( addr == 0 )
+ break;
+
+ if ( endptr )
+ ++endptr;
+ vaString = endptr;
+
+ printf("0x%llx\n", addr);
+ std::optional<size_t> rangeIndexOpt = symbolicatedCache.findClosestRange(addr);
+ if ( !rangeIndexOpt )
+ return 0;
+
+ size_t rangeIndex = *rangeIndexOpt;
+
+ const SymbolicatedCache::Range& range = symbolicatedCache.ranges[rangeIndex];
+ const SymbolicatedImage* image = nullptr;
+ if ( range.imageIndex ) {
+ image = &symbolicatedCache.images[*range.imageIndex];
+ printf(" %15s %s\n", "in:", image->image().header()->installName());
+ printf(" %15s 0x%llx\n", "image base:", image->image().header()->preferredLoadAddress());
+ }
+ if ( !range.segmentName.empty() ) {
+ printf(" %15s %.*s,%.*s\n", "segment name:",
+ (int)range.segmentName.size(), range.segmentName.data(),
+ (int)range.sectName.size(), range.sectName.data());
+ }
+ printf(" %15s 0x%llx - 0x%llx\n", "range:", range.startAddr, range.endAddr);
+ printf(" %15s %s\n", "symbol:", symbolicatedCache.symbolNameAt(addr).c_str());
}
}
else if ( options.mode == modeExtract ) {
@@ -1862,11 +3270,16 @@
// Get the base pointers from the magic section in objc
__block uint64_t objcCacheOffsetsSize = 0;
__block const void* objcCacheOffsets = nullptr;
+ __block int impCachesVersion = 1;
__block Diagnostics diag;
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
if ( !strcmp(installName, "/usr/lib/libobjc.A.dylib") ) {
- const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)mh;
+ const dyld3::MachOAnalyzer* ma = (const dyld3::MachOAnalyzer*)hdr;
objcCacheOffsets = ma->findSectionContent("__DATA_CONST", "__objc_scoffs", objcCacheOffsetsSize);
+ dyld3::MachOAnalyzer::FoundSymbol foundInfo;
+ if (ma->findExportedSymbol(diag, "_objc_opt_preopt_caches_version", false, foundInfo, nullptr)) {
+ impCachesVersion = *(int*)((uint8_t*)ma + foundInfo.value);
+ }
}
});
@@ -1882,12 +3295,15 @@
dyld3::MachOAnalyzer::VMAddrConverter vmAddrConverter = dyldCache->makeVMAddrConverter(contentRebased);
- uint64_t selectorStringVMAddrStart = vmAddrConverter.convertToVMAddr(((uint64_t*)objcCacheOffsets)[0]);
- uint64_t selectorStringVMAddrEnd = vmAddrConverter.convertToVMAddr(((uint64_t*)objcCacheOffsets)[1]);
+ uint8_t selectorStringIndex = 0;
+ if (impCachesVersion > 1)
+ selectorStringIndex = 1;
+ uint64_t selectorStringVMAddrStart = vmAddrConverter.convertToVMAddr(((uint64_t*)objcCacheOffsets)[selectorStringIndex]);
+ uint64_t selectorStringVMAddrEnd = vmAddrConverter.convertToVMAddr(((uint64_t*)objcCacheOffsets)[selectorStringIndex+1]);
uint64_t sharedCacheRelativeSelectorBaseVMAddress = dyldCache->sharedCacheRelativeSelectorBaseVMAddress();
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
if (diag.hasError())
return;
@@ -1933,7 +3349,8 @@
if (diag.hasError())
return 1;
- dyldCache->forEachImage(^(const mach_header *mh, const char *installName) {
+ __block bool anyBadEntry = false;
+ dyldCache->forEachImage(^(const Header *mh, const char *installName) {
if (diag.hasError())
return;
@@ -1956,54 +3373,90 @@
return;
}
- struct Bucket {
- uint32_t selOffset;
- uint32_t impOffset;
- };
- struct ImpCache {
- int32_t fallback_class_offset;
- uint32_t cache_shift : 5;
- uint32_t cache_mask : 11;
- uint32_t occupied : 14;
- uint32_t has_inlines : 1;
- uint32_t bit_one : 1;
- struct Bucket buckets[];
- };
-
- const ImpCache* impCache = (const ImpCache*)(objcClass.methodCacheVMAddr + slide);
- printf("%s (%s): %d buckets\n", className, type, impCache->cache_mask + 1);
-
- if ((classVMAddr + impCache->fallback_class_offset) != objcClass.superclassVMAddr) {
- printf("Flattening fallback: %s\n", classVMAddrToNameMap[classVMAddr + impCache->fallback_class_offset]);
- }
+ const uint8_t* impCacheBuffer = (const uint8_t*)(objcClass.methodCacheVMAddr + slide);
+ uint32_t cacheMask = 0;
+ uint32_t cacheShift = 0;
+ uint32_t sizeOfHeader = 0;
+ if (impCachesVersion < 3) {
+ const ImpCacheHeader_v1* impCache = (const ImpCacheHeader_v1*)impCacheBuffer;
+ cacheMask = impCache->cache_mask;
+ cacheShift = impCache->cache_shift;
+ sizeOfHeader = sizeof(ImpCacheHeader_v1);
+
+ printf("%s (%s): %d buckets, 0x%X mask, 0x%X shift\n", className, type, impCache->cache_mask + 1, cacheMask, cacheShift);
+ if ((classVMAddr + impCache->fallback_class_offset) != objcClass.superclassVMAddr)
+ printf("Flattening fallback: %s\n", classVMAddrToNameMap[classVMAddr + impCache->fallback_class_offset]);
+ } else {
+ const ImpCacheHeader_v2* impCache = (const ImpCacheHeader_v2*)impCacheBuffer;
+ cacheMask = impCache->cache_mask;
+ cacheShift = impCache->cache_shift;
+ sizeOfHeader = sizeof(ImpCacheHeader_v2);
+
+ printf("%s (%s): %d buckets, 0x%X mask, 0x%X shift\n", className, type, impCache->cache_mask + 1, cacheMask, cacheShift);
+ if ((classVMAddr + impCache->fallback_class_offset) != objcClass.superclassVMAddr)
+ printf("Flattening fallback: %s\n", classVMAddrToNameMap[classVMAddr + impCache->fallback_class_offset]);
+ }
+
+ const uint8_t* buckets = impCacheBuffer + sizeOfHeader;
// Buckets are a 32-bit offset from the impcache itself
- for (uint32_t i = 0; i <= impCache->cache_mask ; ++i) {
- const Bucket& b = impCache->buckets[i];
- uint64_t sel = (uint64_t)b.selOffset + selectorStringVMAddrStart;
- uint64_t imp = classVMAddr - (uint64_t)b.impOffset;
- if (b.selOffset == 0xFFFFFFFF) {
+ for (uint32_t i = 0; i <= cacheMask ; ++i) {
+ uint64_t sel = 0;
+ uint64_t imp = 0;
+ bool empty = false;
+ uint32_t selOff=0;
+ if (impCachesVersion == 1) {
+ const ImpCacheEntry_v1* bucket = (ImpCacheEntry_v1*)buckets + i;
+ selOff = bucket->selOffset;
+ sel = selectorStringVMAddrStart + bucket->selOffset;
+ imp = classVMAddr - bucket->impOffset;
+ if (bucket->selOffset == 0xFFFFFFF && bucket->impOffset == 0)
+ empty = true;
+ } else {
+ const ImpCacheEntry_v2* bucket = (ImpCacheEntry_v2*)buckets + i;
+ selOff = bucket->selOffset;
+ sel = selectorStringVMAddrStart + bucket->selOffset;
+ imp = classVMAddr - (bucket->impOffset << 2);
+ if (bucket->selOffset == 0x3FFFFFF && bucket->impOffset == 0)
+ empty = true;
+ }
+
+ if ( empty ) {
// Empty bucket
printf(" - 0x%016llx: %s\n", 0ULL, "");
} else {
assert(sel < selectorStringVMAddrEnd);
-
auto it = methodToClassMap.find(imp);
if (it == methodToClassMap.end()) {
fprintf(stderr, "Could not find IMP %llx (for %s)\n", imp, (const char*)(sel + slide));
}
assert(it != methodToClassMap.end());
- printf(" - 0x%016llx: %s (from %s)\n", imp, (const char*)(sel + slide), it->second);
+ printf(" - 0x%016llx: %s (from %s) (selOff: 0x%X)\n", imp, (const char*)(sel + slide), it->second, selOff);
+ uint32_t expectedBucket = (selOff >> cacheShift) & cacheMask;
+ if ( expectedBucket != i ) {
+ fprintf(stderr, " [BAD IMP CACHE] - 0x%016llx: %s (from %s) (selOff: 0x%X) (bucket: %d vs expected: %d)\n", imp, (const char*)(sel + slide), it->second, selOff, i, expectedBucket);
+ anyBadEntry = true;
+ }
}
}
};
ma->forEachObjCClass(diag, vmAddrConverter, visitClass);
});
+ if ( anyBadEntry ) {
+ fprintf(stderr, "IMP cache is malformed\n");
+ exit(1);
+ }
} else {
switch ( options.mode ) {
case modeList: {
+ if ( options.printInodes && !dyldCache->header.dylibsExpectedOnDisk ) {
+ fprintf(stderr, "Error: '-inode' option only valid on simulator shared caches\n");
+ return 1;
+ }
// list all dylibs, including their aliases (symlinks to them) with option vmaddr
__block std::vector<std::unordered_set<std::string>> indexToPaths;
__block std::vector<uint64_t> indexToAddr;
+ __block std::unordered_map<uint64_t, uint64_t> indexToINode;
+ __block std::unordered_map<uint64_t, uint64_t> indexToModTime;
__block std::vector<std::string> indexToUUID;
dyldCache->forEachImageTextSegment(^(uint64_t loadAddressUnslid, uint64_t textSegmentSize, const unsigned char* dylibUUID, const char* installName, bool& stop) {
std::unordered_set<std::string> empty;
@@ -2019,6 +3472,12 @@
});
dyldCache->forEachDylibPath(^(const char* dylibPath, uint32_t index) {
indexToPaths[index].insert(dylibPath);
+
+ uint64_t mTime = ~0ULL;
+ uint64_t inode = ~0ULL;
+ dyldCache->getIndexedImageEntry(index, mTime, inode);
+ indexToINode[index] = inode;
+ indexToModTime[index] = mTime;
});
int index = 0;
for (const std::unordered_set<std::string>& paths : indexToPaths) {
@@ -2027,6 +3486,8 @@
printf("0x%08llX ", indexToAddr[index]);
if ( options.printUUIDs )
printf("<%s> ", indexToUUID[index].c_str());
+ if ( options.printInodes )
+ printf("0x%08llX 0x%08llX ", indexToINode[index], indexToModTime[index]);
printf("%s\n", path.c_str());
}
++index;
@@ -2034,10 +3495,9 @@
break;
}
case modeListDylibsWithSection: {
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
- dyld3::MachOFile* mf = (dyld3::MachOFile*)mh;
- mf->forEachSection(^(const dyld3::MachOFile::SectionInfo& sectInfo, bool malformedSectionRange, bool& stop) {
- if ( (strcmp(sectInfo.sectName, options.sectionName) == 0) && (strcmp(sectInfo.segInfo.segName, options.segmentName) == 0) ) {
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ hdr->forEachSection(^(const Header::SectionInfo& sectInfo, bool& stop) {
+ if ( (sectInfo.sectionName == options.sectionName) && (sectInfo.segmentName == options.segmentName) ) {
printf("%s\n", installName);
stop = true;
}
@@ -2048,13 +3508,14 @@
case modeMap: {
__block std::map<uint64_t, const char*> dataSegNames;
__block std::map<uint64_t, uint64_t> dataSegEnds;
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
- dyld3::MachOFile* mf = (dyld3::MachOFile*)mh;
- mf->forEachSegment(^(const dyld3::MachOAnalyzer::SegmentInfo &info, bool &stop) {
- printf("0x%08llX - 0x%08llX %s %s\n", info.vmAddr, info.vmAddr + info.vmSize, info.segName, installName);
- if ( strncmp(info.segName, "__DATA", 6) == 0 ) {
- dataSegNames[info.vmAddr] = installName;
- dataSegEnds[info.vmAddr] = info.vmAddr + info.vmSize;
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ hdr->forEachSegment(^(const Header::SegmentInfo &info, bool &stop) {
+ printf("0x%08llX - 0x%08llX %.*s %s\n", info.vmaddr, info.vmaddr + info.vmsize,
+ (int)info.segmentName.size(), info.segmentName.data(),
+ installName);
+ if ( info.segmentName.starts_with("__DATA") ) {
+ dataSegNames[info.vmaddr] = installName;
+ dataSegEnds[info.vmaddr] = info.vmaddr + info.vmsize;
}
});
});
@@ -2071,7 +3532,7 @@
}
case modeDependencies: {
__block bool dependentTargetFound = false;
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
if ( strcmp(options.dependentsOfPath, installName) != 0 )
return;
dependentTargetFound = true;
@@ -2099,14 +3560,14 @@
}
};
- dyld3::MachOFile* mf = (dyld3::MachOFile*)mh;
+ dyld3::MachOFile* mf = (dyld3::MachOFile*)hdr;
// First print out our dylib and version.
const char* dylibInstallName;
- uint32_t currentVersion;
- uint32_t compatVersion;
- if ( mf->getDylibInstallName(&dylibInstallName, &compatVersion, ¤tVersion) ) {
- printDep(dylibInstallName, compatVersion, currentVersion);
+ Version32 currentVersion;
+ Version32 compatVersion;
+ if ( hdr->getDylibInstallName(&dylibInstallName, &compatVersion, ¤tVersion) ) {
+ printDep(dylibInstallName, compatVersion.value(), currentVersion.value());
}
// Then the dependent dylibs.
@@ -2138,8 +3599,8 @@
}
};
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
- dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)mh;
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
Diagnostics diag;
dyld3::MachOAnalyzer::LinkEditInfo leInfo;
ma->getLinkEditPointers(diag, leInfo);
@@ -2156,28 +3617,28 @@
//printf("export_off=0x%X\n", leInfo.dyldInfo->export_off());
uint32_t exportPageOffsetStart = leInfo.dyldInfo->export_off & (-4096);
uint32_t exportPageOffsetEnd = (leInfo.dyldInfo->export_off + leInfo.dyldInfo->export_size) & (-4096);
- sprintf(message, "exports from %s", shortName);
+ snprintf(message, sizeof(message), "exports from %s", shortName);
add_linkedit(exportPageOffsetStart, exportPageOffsetEnd, message);
}
// add binding info
if ( leInfo.dyldInfo->bind_size != 0 ) {
uint32_t bindPageOffsetStart = leInfo.dyldInfo->bind_off & (-4096);
uint32_t bindPageOffsetEnd = (leInfo.dyldInfo->bind_off + leInfo.dyldInfo->bind_size) & (-4096);
- sprintf(message, "bindings from %s", shortName);
+ snprintf(message, sizeof(message), "bindings from %s", shortName);
add_linkedit(bindPageOffsetStart, bindPageOffsetEnd, message);
}
// add lazy binding info
if ( leInfo.dyldInfo->lazy_bind_size != 0 ) {
uint32_t lazybindPageOffsetStart = leInfo.dyldInfo->lazy_bind_off & (-4096);
uint32_t lazybindPageOffsetEnd = (leInfo.dyldInfo->lazy_bind_off + leInfo.dyldInfo->lazy_bind_size) & (-4096);
- sprintf(message, "lazy bindings from %s", shortName);
+ snprintf(message, sizeof(message), "lazy bindings from %s", shortName);
add_linkedit(lazybindPageOffsetStart, lazybindPageOffsetEnd, message);
}
// add weak binding info
if ( leInfo.dyldInfo->weak_bind_size != 0 ) {
uint32_t weakbindPageOffsetStart = leInfo.dyldInfo->weak_bind_off & (-4096);
uint32_t weakbindPageOffsetEnd = (leInfo.dyldInfo->weak_bind_off + leInfo.dyldInfo->weak_bind_size) & (-4096);
- sprintf(message, "weak bindings from %s", shortName);
+ snprintf(message, sizeof(message), "weak bindings from %s", shortName);
add_linkedit(weakbindPageOffsetStart, weakbindPageOffsetEnd, message);
}
} else {
@@ -2186,7 +3647,7 @@
//printf("export_off=0x%X\n", leInfo.exportsTrie->export_off());
uint32_t exportPageOffsetStart = leInfo.exportsTrie->dataoff & (-4096);
uint32_t exportPageOffsetEnd = (leInfo.exportsTrie->dataoff + leInfo.exportsTrie->datasize) & (-4096);
- sprintf(message, "exports from %s", shortName);
+ snprintf(message, sizeof(message), "exports from %s", shortName);
add_linkedit(exportPageOffsetStart, exportPageOffsetEnd, message);
}
// Chained fixups are stripped from cache binaries, so no need to check for them here
@@ -2204,11 +3665,9 @@
const char* path;
};
__block std::vector<TextInfo> textSegments;
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
-
- dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)mh;
- ma->forEachSegment(^(const dyld3::MachOAnalyzer::SegmentInfo &info, bool &stop) {
- if ( strcmp(info.segName, "__TEXT") != 0 )
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ hdr->forEachSegment(^(const Header::SegmentInfo &info, bool &stop) {
+ if ( info.segmentName != "__TEXT" )
return;
textSegments.push_back({ info.fileSize, installName });
});
@@ -2221,53 +3680,495 @@
}
break;
}
+ case modeFunctionVariants: {
+ printf("Function Variant table size: %lld bytes\n", dyldCache->header.functionVariantInfoSize);
+ uintptr_t cacheSlide = dyldCache->slide();
+ dyldCache->forEachFunctionVariantPatchLocation(^(const void* loc, PointerMetaData pmd, const mach_o::FunctionVariants& fvs, const mach_o::Header* dylibHdr, int variantIndex, bool& stop) {
+ if ( pmd.authenticated ) {
+ printf(" fixup-loc=%p (key=%d, addr=%d, diversity=0x%04X), header-of-dylib-with-variant=%p, variant-index=%d\n",
+ (void*)((uintptr_t)loc - cacheSlide), pmd.key, pmd.usesAddrDiversity, pmd.diversity, (void*)((uintptr_t)dylibHdr - cacheSlide), variantIndex);
+ }
+ else {
+ printf(" fixup-loc=%p, header-of-dylib-with-variant=%p, variant-index=%d\n", (void*)((uintptr_t)loc - cacheSlide), (void*)((uintptr_t)dylibHdr - cacheSlide), variantIndex);
+ }
+ });
+ break;
+ }
case modePatchTable: {
+ printf("Patch table size: %lld bytes\n", dyldCache->header.patchInfoSize);
+
+ uint64_t cacheBaseAddress = dyldCache->unslidLoadAddress();
std::vector<SegmentInfo> segInfos;
buildSegmentInfo(dyldCache, segInfos);
__block uint32_t imageIndex = 0;
- dyldCache->forEachImage(^(const mach_header* mh, const char* installName) {
+ __block std::unordered_map<CString, PatchPageMetric> pagesPerSymbol;
+
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
printf("%s:\n", installName);
- uint64_t cacheBaseAddress = dyldCache->unslidLoadAddress();
- uint64_t dylibBaseAddress = ((dyld3::MachOAnalyzer*)mh)->preferredLoadAddress();
- dyldCache->forEachPatchableExport(imageIndex, ^(uint32_t dylibVMOffsetOfImpl, const char* exportName) {
+ uint64_t dylibBaseAddress = hdr->preferredLoadAddress();
+ dyldCache->forEachPatchableExport(imageIndex, ^(uint32_t dylibVMOffsetOfImpl, const char* exportName,
+ PatchKind patchKind) {
uint64_t cacheOffsetOfImpl = (dylibBaseAddress + dylibVMOffsetOfImpl) - cacheBaseAddress;
- printf(" export: 0x%08llX %s\n", cacheOffsetOfImpl, exportName);
+ printf(" export: 0x%08llX%s %s\n", cacheOffsetOfImpl,
+ PatchTable::patchKindName(patchKind), exportName);
+
+ CString exportNameStr(exportName);
dyldCache->forEachPatchableUseOfExport(imageIndex, dylibVMOffsetOfImpl,
^(uint32_t userImageIndex, uint32_t userVMOffset,
- dyld3::MachOLoaded::PointerMetaData pmd, uint64_t addend) {
+ dyld3::MachOLoaded::PointerMetaData pmd, uint64_t addend,
+ bool isWeakImport) {
// Get the image so that we can convert from dylib offset to cache offset
uint64_t mTime;
uint64_t inode;
- const dyld3::MachOAnalyzer* imageMA = (dyld3::MachOAnalyzer*)(dyldCache->getIndexedImageEntry(userImageIndex, mTime, inode));
- if ( imageMA == nullptr )
+ const Header* imageHdr = (const Header*)(dyldCache->getIndexedImageEntry(userImageIndex, mTime, inode));
+ if ( imageHdr == nullptr )
return;
SegmentInfo usageAt;
- const uint64_t patchLocVmAddr = imageMA->preferredLoadAddress() + userVMOffset;
+ const uint64_t patchLocVmAddr = imageHdr->preferredLoadAddress() + userVMOffset;
const uint64_t patchLocCacheOffset = patchLocVmAddr - cacheBaseAddress;
findImageAndSegment(dyldCache, segInfos, patchLocCacheOffset, &usageAt);
+ if ( !patchKindIsRootOptimized(patchKind) )
+ pagesPerSymbol[exportNameStr].addUse(patchLocVmAddr);
+
// Verify that findImage and the callback image match
- std::string_view userInstallName = imageMA->installName();
- assert(userInstallName == usageAt.installName);
-
- if ( addend == 0 )
- printf(" used by: %s+0x%04llX in %s\n", usageAt.segName, patchLocVmAddr-usageAt.vmAddr, usageAt.installName);
- else
- printf(" used by: %s+0x%04llX (addend=%lld) in %s\n", usageAt.segName, patchLocVmAddr-usageAt.vmAddr, addend, usageAt.installName);
+ std::string_view userInstallName = imageHdr->installName();
+
+ // FIXME: We can't get this from MachoLoaded without having a fixup location to call
+ static const char* keyNames[] = {
+ "IA", "IB", "DA", "DB"
+ };
+
+ uint64_t sectionOffset = patchLocVmAddr-usageAt.vmAddr;
+
+ const char* weakImportString = isWeakImport ? " (weak-import)" : "";
+
+ if ( addend == 0 ) {
+ if ( pmd.authenticated ) {
+ printf(" used by: %.*s(0x%04llX)%s (PAC: div=%d, addr=%s, key=%s) in %s\n",
+ (int)usageAt.segName.size(), usageAt.segName.data(),
+ sectionOffset, weakImportString,
+ pmd.diversity, pmd.usesAddrDiversity ? "true" : "false", keyNames[pmd.key],
+ userInstallName.data());
+ } else {
+ printf(" used by: %.*s(0x%04llX)%s in %s\n",
+ (int)usageAt.segName.size(), usageAt.segName.data(),
+ sectionOffset, weakImportString, userInstallName.data());
+ }
+ } else {
+ if ( pmd.authenticated ) {
+ printf(" used by: %.*s(0x%04llX)%s (addend=%lld) (PAC: div=%d, addr=%s, key=%s) in %s\n",
+ (int)usageAt.segName.size(), usageAt.segName.data(),
+ sectionOffset, weakImportString, addend,
+ pmd.diversity, pmd.usesAddrDiversity ? "true" : "false", keyNames[pmd.key],
+ userInstallName.data());
+ } else {
+ printf(" used by: %.*s(0x%04llX)%s (addend=%lld) in %s\n",
+ (int)usageAt.segName.size(), usageAt.segName.data(),
+ sectionOffset, weakImportString,
+ addend, userInstallName.data());
+ }
+ }
+ });
+
+ // Print GOT uses
+ dyldCache->forEachPatchableGOTUseOfExport(imageIndex, dylibVMOffsetOfImpl,
+ ^(uint64_t cacheVMOffset,
+ dyld3::MachOLoaded::PointerMetaData pmd, uint64_t addend,
+ bool isWeakImport) {
+
+ // FIXME: We can't get this from MachoLoaded without having a fixup location to call
+ static const char* keyNames[] = {
+ "IA", "IB", "DA", "DB"
+ };
+
+ if ( !patchKindIsRootOptimized(patchKind) )
+ pagesPerSymbol["GOT"].addUse(cacheBaseAddress + cacheVMOffset);
+
+ const char* weakImportString = isWeakImport ? " (weak-import)" : "";
+ if ( addend == 0 ) {
+ if ( pmd.authenticated ) {
+ printf(" used by: GOT%s (0x%llx) (PAC: div=%d, addr=%s, key=%s)\n",
+ weakImportString, cacheVMOffset, pmd.diversity, pmd.usesAddrDiversity ? "true" : "false", keyNames[pmd.key]);
+ } else {
+ printf(" used by: GOT%s (0x%llx)\n", weakImportString, cacheVMOffset);
+ }
+ } else {
+ if ( pmd.authenticated ) {
+ printf(" used by: GOT%s (0x%llx) (addend=%lld) (PAC: div=%d, addr=%s, key=%s)\n",
+ weakImportString, cacheVMOffset, addend, pmd.diversity, pmd.usesAddrDiversity ? "true" : "false", keyNames[pmd.key]);
+ } else {
+ printf(" used by: GOT%s (0x%llx) (addend=%lld)\n", weakImportString, cacheVMOffset, addend);
+ }
+ }
});
});
++imageIndex;
});
+
+ logCostOfRootPerSymbol(pagesPerSymbol);
break;
}
+ case modeDumpPatchTable: {
+ PatchTable patchTable(dyldCache->patchTable(), dyldCache->header.patchInfoAddr);
+ patchTable.dump();
+ break;
+ }
+ case modeRootsCost: {
+ std::vector<SegmentInfo> segInfos;
+ buildSegmentInfo(dyldCache, segInfos);
+
+ __block std::optional<uint32_t> rootImageIndex;
+ {
+ __block uint32_t imageIndex = 0;
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ if ( strcmp(installName, options.rootsCostOfDylib) == 0 )
+ rootImageIndex = imageIndex;
+ ++imageIndex;
+ });
+ }
+
+ if ( !rootImageIndex.has_value() ) {
+ fprintf(stderr, "Could not find image '%s' in shared cache\n", options.rootsCostOfDylib);
+ return 1;
+ }
+
+ // For each page of the cache, we record if it was dirtied by patching, and by which dylibs
+ typedef std::pair<const char*, std::string_view> InstallNameAndSegment;
+ __block std::map<uint64_t, std::set<InstallNameAndSegment>> pages;
+ __block std::unordered_map<CString, PatchPageMetric> pagesPerSymbol;
+
+ uint64_t cacheBaseAddress = dyldCache->unslidLoadAddress();
+ dyldCache->forEachPatchableExport(rootImageIndex.value(),
+ ^(uint32_t dylibVMOffsetOfImpl, const char* exportName,
+ PatchKind patchKind) {
+ if ( patchKindIsRootOptimized(patchKind) )
+ return;
+
+ CString exportNameStr(exportName);
+ dyldCache->forEachPatchableUseOfExport(rootImageIndex.value(), dylibVMOffsetOfImpl,
+ ^(uint32_t userImageIndex, uint32_t userVMOffset,
+ dyld3::MachOLoaded::PointerMetaData pmd, uint64_t addend,
+ bool isWeakImport) {
+ // Get the image so that we can convert from dylib offset to cache offset
+ uint64_t mTime;
+ uint64_t inode;
+ const dyld3::MachOAnalyzer* imageMA = (dyld3::MachOAnalyzer*)(dyldCache->getIndexedImageEntry(userImageIndex, mTime, inode));
+ if ( imageMA == nullptr )
+ return;
+
+ SegmentInfo usageAt;
+ const uint64_t patchLocVmAddr = ((const Header*)imageMA)->preferredLoadAddress() + userVMOffset;
+ const uint64_t patchLocCacheOffset = patchLocVmAddr - cacheBaseAddress;
+ findImageAndSegment(dyldCache, segInfos, patchLocCacheOffset, &usageAt);
+
+ // Round to the 16KB page we dirty
+ //clientUsedPages[userImageIndex].insert(usageAt.vmAddr & ~0x3FFF);
+ uint64_t pageAddr = patchLocVmAddr & ~0x3FFF;
+ pages[pageAddr].insert({ usageAt.installName, usageAt.segName });
+ pagesPerSymbol[exportNameStr].addUse(pageAddr);
+ });
+
+ // Print GOT uses
+ dyldCache->forEachPatchableGOTUseOfExport(rootImageIndex.value(), dylibVMOffsetOfImpl,
+ ^(uint64_t cacheVMOffset,
+ dyld3::MachOLoaded::PointerMetaData pmd, uint64_t addend,
+ bool isWeakImport) {
+ // Round to the 16KB page we dirty
+ //gotUsedPages.insert((cacheBaseAddress + cacheVMOffset) & ~0x3FFF);
+ uint64_t pageAddr = (cacheBaseAddress + cacheVMOffset) & ~0x3FFF;
+ pages[pageAddr].insert({ "GOT", "" });
+ pagesPerSymbol["GOT"].addUse(pageAddr);
+ });
+ });
+
+ // Print the results
+ printf("Cost of root of '%s' is %lld pages:\n", options.rootsCostOfDylib, (uint64_t)pages.size());
+
+ for ( const auto& page : pages ) {
+ printf("0x%08llx ", page.first);
+
+ bool needsComma = false;
+ for ( const InstallNameAndSegment& installNameAndSegment : page.second ) {
+ if ( needsComma )
+ printf(", ");
+ needsComma = true;
+
+ const char* leafName = strrchr(installNameAndSegment.first, '/');
+ if ( leafName == NULL )
+ leafName = installNameAndSegment.first;
+ else
+ leafName++;
+
+ if ( !installNameAndSegment.second.empty() )
+ printf("%s(%.*s)", leafName, (int)installNameAndSegment.second.size(), installNameAndSegment.second.data());
+ else
+ printf("%s", leafName);
+ }
+ printf("\n");
+ }
+
+ logCostOfRootPerSymbol(pagesPerSymbol);
+ break;
+ }
+ case modeMachHeaders: {
+ auto printRow = [](const char* magic, const char* arch, const char* filetype,
+ const char* ncmds, const char* sizeofcmds , const char* flags,
+ const char* installname) {
+ printf("%12s %8s %8s %8s %12s %12s %8s\n",
+ magic, arch, filetype, ncmds, sizeofcmds , flags, installname);
+ };
+
+ printRow("magic", "arch", "filetype", "ncmds", "sizeofcmds", "flags", "installname");
+ dyldCache->forEachDylib(^(const Header *mh, const char *installName, uint32_t imageIndex, uint64_t inode, uint64_t mtime, bool &stop) {
+ const dyld3::MachOFile* mf = (const dyld3::MachOFile*)mh;
+ const char* magic = nullptr;
+ if ( mf->magic == MH_MAGIC )
+ magic = "MH_MAGIC";
+ else if ( mf->magic == MH_MAGIC_64 )
+ magic = "MH_MAGIC_64";
+ else if ( mf->magic == MH_CIGAM )
+ magic = "MH_CIGAM";
+ else if ( mf->magic == MH_CIGAM_64 )
+ magic = "MH_CIGAM_64";
+
+ const char* arch = mf->archName();
+ const char* filetype = mf->isDylib() ? "DYLIB" : "UNKNOWN";
+ std::string ncmds = json::decimal(mf->ncmds);
+ std::string sizeofcmds = json::decimal(mf->sizeofcmds);
+ std::string flags = json::hex(mf->flags);
+
+ printRow(magic, arch, filetype, ncmds.c_str(), sizeofcmds.c_str(), flags.c_str(), installName);
+ });
+
+ break;
+ }
+ case modeCacheHeader: {
+ __block uint32_t cacheIndex = 0;
+ dyldCache->forEachCache(^(const DyldSharedCache *cache, bool &stopCache) {
+ printf("Cache #%d\n", cacheIndex);
+
+ uuid_string_t uuidString;
+ uuid_unparse_upper(dyldCache->header.uuid, uuidString);
+
+ uuid_string_t symbolFileUUIDString;
+ uuid_unparse_upper(dyldCache->header.symbolFileUUID, symbolFileUUIDString);
+
+ printf(" - magic: %s\n", dyldCache->header.magic);
+ printf(" - mappingOffset: 0x%llx\n", (uint64_t)dyldCache->header.mappingOffset);
+ printf(" - mappingCount: 0x%llx\n", (uint64_t)dyldCache->header.mappingCount);
+ printf(" - imagesOffsetOld: 0x%llx\n", (uint64_t)dyldCache->header.imagesOffsetOld);
+ printf(" - imagesCountOld: 0x%llx\n", (uint64_t)dyldCache->header.imagesCountOld);
+ printf(" - dyldBaseAddress: 0x%llx\n", (uint64_t)dyldCache->header.dyldBaseAddress);
+ printf(" - codeSignatureOffset: 0x%llx\n", (uint64_t)dyldCache->header.codeSignatureOffset);
+ printf(" - codeSignatureSize: 0x%llx\n", (uint64_t)dyldCache->header.codeSignatureSize);
+ printf(" - slideInfoOffsetUnused: 0x%llx\n", (uint64_t)dyldCache->header.slideInfoOffsetUnused);
+ printf(" - slideInfoSizeUnused: 0x%llx\n", (uint64_t)dyldCache->header.slideInfoSizeUnused);
+ printf(" - localSymbolsOffset: 0x%llx\n", (uint64_t)dyldCache->header.localSymbolsOffset);
+ printf(" - localSymbolsSize: 0x%llx\n", (uint64_t)dyldCache->header.localSymbolsSize);
+ printf(" - uuid: %s\n", uuidString);
+ printf(" - cacheType: 0x%llx\n", (uint64_t)dyldCache->header.cacheType);
+ printf(" - branchPoolsOffset: 0x%llx\n", (uint64_t)dyldCache->header.branchPoolsOffset);
+ printf(" - branchPoolsCount: 0x%llx\n", (uint64_t)dyldCache->header.branchPoolsCount);
+ printf(" - dyldInCacheMH: 0x%llx\n", (uint64_t)dyldCache->header.dyldInCacheMH);
+ printf(" - dyldInCacheEntry: 0x%llx\n", (uint64_t)dyldCache->header.dyldInCacheEntry);
+ printf(" - imagesTextOffset: 0x%llx\n", (uint64_t)dyldCache->header.imagesTextOffset);
+ printf(" - imagesTextCount: 0x%llx\n", (uint64_t)dyldCache->header.imagesTextCount);
+ printf(" - patchInfoAddr: 0x%llx\n", (uint64_t)dyldCache->header.patchInfoAddr);
+ printf(" - patchInfoSize: 0x%llx\n", (uint64_t)dyldCache->header.patchInfoSize);
+ printf(" - otherImageGroupAddrUnused: 0x%llx\n", (uint64_t)dyldCache->header.otherImageGroupAddrUnused);
+ printf(" - otherImageGroupSizeUnused: 0x%llx\n", (uint64_t)dyldCache->header.otherImageGroupSizeUnused);
+ printf(" - progClosuresAddr: 0x%llx\n", (uint64_t)dyldCache->header.progClosuresAddr);
+ printf(" - progClosuresSize: 0x%llx\n", (uint64_t)dyldCache->header.progClosuresSize);
+ printf(" - progClosuresTrieAddr: 0x%llx\n", (uint64_t)dyldCache->header.progClosuresTrieAddr);
+ printf(" - progClosuresTrieSize: 0x%llx\n", (uint64_t)dyldCache->header.progClosuresTrieSize);
+ printf(" - platform: 0x%llx\n", (uint64_t)dyldCache->header.platform);
+ printf(" - formatVersion: 0x%llx\n", (uint64_t)dyldCache->header.formatVersion);
+ printf(" - dylibsExpectedOnDisk: 0x%llx\n", (uint64_t)dyldCache->header.dylibsExpectedOnDisk);
+ printf(" - simulator: 0x%llx\n", (uint64_t)dyldCache->header.simulator);
+ printf(" - locallyBuiltCache: 0x%llx\n", (uint64_t)dyldCache->header.locallyBuiltCache);
+ printf(" - builtFromChainedFixups: 0x%llx\n", (uint64_t)dyldCache->header.builtFromChainedFixups);
+ printf(" - padding: 0x%llx\n", (uint64_t)dyldCache->header.padding);
+ printf(" - sharedRegionStart: 0x%llx\n", (uint64_t)dyldCache->header.sharedRegionStart);
+ printf(" - sharedRegionSize: 0x%llx\n", (uint64_t)dyldCache->header.sharedRegionSize);
+ printf(" - maxSlide: 0x%llx\n", (uint64_t)dyldCache->header.maxSlide);
+ printf(" - dylibsImageArrayAddr: 0x%llx\n", (uint64_t)dyldCache->header.dylibsImageArrayAddr);
+ printf(" - dylibsImageArraySize: 0x%llx\n", (uint64_t)dyldCache->header.dylibsImageArraySize);
+ printf(" - dylibsTrieAddr: 0x%llx\n", (uint64_t)dyldCache->header.dylibsTrieAddr);
+ printf(" - dylibsTrieSize: 0x%llx\n", (uint64_t)dyldCache->header.dylibsTrieSize);
+ printf(" - otherImageArrayAddr: 0x%llx\n", (uint64_t)dyldCache->header.otherImageArrayAddr);
+ printf(" - otherImageArraySize: 0x%llx\n", (uint64_t)dyldCache->header.otherImageArraySize);
+ printf(" - otherTrieAddr: 0x%llx\n", (uint64_t)dyldCache->header.otherTrieAddr);
+ printf(" - otherTrieSize: 0x%llx\n", (uint64_t)dyldCache->header.otherTrieSize);
+ printf(" - mappingWithSlideOffset: 0x%llx\n", (uint64_t)dyldCache->header.mappingWithSlideOffset);
+ printf(" - mappingWithSlideCount: 0x%llx\n", (uint64_t)dyldCache->header.mappingWithSlideCount);
+ printf(" - dylibsPBLStateArrayAddrUnused: 0x%llx\n", (uint64_t)dyldCache->header.dylibsPBLStateArrayAddrUnused);
+ printf(" - dylibsPBLSetAddr: 0x%llx\n", (uint64_t)dyldCache->header.dylibsPBLSetAddr);
+ printf(" - programsPBLSetPoolAddr: 0x%llx\n", (uint64_t)dyldCache->header.programsPBLSetPoolAddr);
+ printf(" - programsPBLSetPoolSize: 0x%llx\n", (uint64_t)dyldCache->header.programsPBLSetPoolSize);
+ printf(" - programTrieAddr: 0x%llx\n", (uint64_t)dyldCache->header.programTrieAddr);
+ printf(" - programTrieSize: 0x%llx\n", (uint64_t)dyldCache->header.programTrieSize);
+ printf(" - osVersion: 0x%llx\n", (uint64_t)dyldCache->header.osVersion);
+ printf(" - altPlatform: 0x%llx\n", (uint64_t)dyldCache->header.altPlatform);
+ printf(" - altOsVersion: 0x%llx\n", (uint64_t)dyldCache->header.altOsVersion);
+ printf(" - swiftOptsOffset: 0x%llx\n", (uint64_t)dyldCache->header.swiftOptsOffset);
+ printf(" - swiftOptsSize: 0x%llx\n", (uint64_t)dyldCache->header.swiftOptsSize);
+ printf(" - subCacheArrayOffset: 0x%llx\n", (uint64_t)dyldCache->header.subCacheArrayOffset);
+ printf(" - subCacheArrayCount: 0x%llx\n", (uint64_t)dyldCache->header.subCacheArrayCount);
+ printf(" - symbolFileUUID: %s\n", symbolFileUUIDString);
+ printf(" - rosettaReadOnlyAddr: 0x%llx\n", (uint64_t)dyldCache->header.rosettaReadOnlyAddr);
+ printf(" - rosettaReadOnlySize: 0x%llx\n", (uint64_t)dyldCache->header.rosettaReadOnlySize);
+ printf(" - rosettaReadWriteAddr: 0x%llx\n", (uint64_t)dyldCache->header.rosettaReadWriteAddr);
+ printf(" - rosettaReadWriteSize: 0x%llx\n", (uint64_t)dyldCache->header.rosettaReadWriteSize);
+ printf(" - imagesOffset: 0x%llx\n", (uint64_t)dyldCache->header.imagesOffset);
+ printf(" - imagesCount: 0x%llx\n", (uint64_t)dyldCache->header.imagesCount);
+ printf(" - cacheSubType: 0x%llx\n", (uint64_t)dyldCache->header.cacheSubType);
+ printf(" - objcOptsOffset: 0x%llx\n", (uint64_t)dyldCache->header.objcOptsOffset);
+ printf(" - cacheAtlasOffset: 0x%llx\n", (uint64_t)dyldCache->header.cacheAtlasOffset);
+ printf(" - cacheAtlasSize: 0x%llx\n", (uint64_t)dyldCache->header.cacheAtlasSize);
+ printf(" - dynamicDataOffset: 0x%llx\n", (uint64_t)dyldCache->header.dynamicDataOffset);
+ printf(" - dynamicDataMaxSize: 0x%llx\n", (uint64_t)dyldCache->header.dynamicDataMaxSize);
+ printf(" - tproMappingsOffset: 0x%llx\n", (uint64_t)dyldCache->header.tproMappingsOffset);
+ printf(" - tproMappingsCount: 0x%llx\n", (uint64_t)dyldCache->header.tproMappingsCount);
+ printf(" - functionVariantInfoAddr: 0x%llx\n", (uint64_t)dyldCache->header.functionVariantInfoAddr);
+ printf(" - functionVariantInfoSize: 0x%llx\n", (uint64_t)dyldCache->header.functionVariantInfoSize);
+ ++cacheIndex;
+ });
+ break;
+ }
+ case modeDylibSymbols:
+ {
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
+ Diagnostics diag;
+
+ printf("%s globals:\n", installName);
+ ma->forEachGlobalSymbol(diag, ^(const char *symbolName, uint64_t n_value, uint8_t n_type, uint8_t n_sect, uint16_t n_desc, bool &stop) {
+ printf(" 0x%08llX: %s\n", n_value, symbolName);
+ });
+ printf("%s locals:\n", installName);
+ ma->forEachLocalSymbol(diag, ^(const char *symbolName, uint64_t n_value, uint8_t n_type, uint8_t n_sect, uint16_t n_desc, bool &stop) {
+ printf(" 0x%08llX: %s\n", n_value, symbolName);
+ });
+ printf("%s undefs:\n", installName);
+ ma->forEachImportedSymbol(diag, ^(const char *symbolName, uint64_t n_value, uint8_t n_type, uint8_t n_sect, uint16_t n_desc, bool &stop) {
+ printf(" undef: %s\n", symbolName);
+ });
+ });
+ break;
+ }
+ case modeFunctionStarts: {
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ printf("%s:\n", installName);
+ uint64_t loadAddress = hdr->preferredLoadAddress();
+ Diagnostics diag;
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
+ ma->forEachFunctionStart(^(uint64_t runtimeOffset) {
+ uint64_t targetVMAddr = loadAddress + runtimeOffset;
+ printf(" 0x%08llX\n", targetVMAddr);
+ });
+ });
+ break;
+ }
+ case modePrewarmingData: {
+ printf("prewarming_data:\n");
+ dyldCache->forEachPrewarmingEntry(^(const void *content, uint64_t unslidVMAddr, uint64_t vmSize) {
+ printf("0x%08llx -> 0x%08llx\n", unslidVMAddr, unslidVMAddr + vmSize);
+ });
+ break;
+ }
+ case modeDuplicates:
+ case modeDuplicatesSummary:
+ {
+ __block std::map<std::string, std::vector<const char*>> symbolsToInstallNames;
+ __block std::set<std::string> weakDefSymbols;
+ dyldCache->forEachImage(^(const Header* hdr, const char* installName) {
+ const dyld3::MachOAnalyzer* ma = (dyld3::MachOAnalyzer*)hdr;
+ uint32_t exportTrieRuntimeOffset;
+ uint32_t exportTrieSize;
+ if ( ma->hasExportTrie(exportTrieRuntimeOffset, exportTrieSize) ) {
+ const uint8_t* start = (uint8_t*)hdr + exportTrieRuntimeOffset;
+ const uint8_t* end = start + exportTrieSize;
+ std::vector<ExportInfoTrie::Entry> exports;
+ if ( ExportInfoTrie::parseTrie(start, end, exports) ) {
+ for (const ExportInfoTrie::Entry& entry: exports) {
+ if ( (entry.info.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) == 0 ) {
+ symbolsToInstallNames[entry.name].push_back(installName);
+ if ( entry.info.flags & EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION ) {
+ weakDefSymbols.insert(entry.name);
+ }
+ }
+ }
+ }
+ }
+ });
+ // filter out unzippered twins
+ std::set<std::string> okTwinSymbols;
+ for (const auto& pos : symbolsToInstallNames) {
+ const std::vector<const char*>& paths = pos.second;
+ if ( paths.size() == 2 ) {
+ // ignore unzippered twins
+ const char* one = paths[0];
+ const char* two = paths[1];
+ if ( (strncmp(one, "/System/iOSSupport/", 19) == 0) || (strncmp(two, "/System/iOSSupport/", 19) == 0) ) {
+ if ( const char* tailOne = Utils::strrstr(one, ".framework/") ) {
+ if ( const char* tailTwo = Utils::strrstr(two, ".framework/") ) {
+ if ( strcmp(tailOne, tailTwo) == 0 )
+ okTwinSymbols.insert(pos.first);
+ }
+ }
+ }
+ }
+ }
+ std::erase_if(symbolsToInstallNames, [&](auto const& pos) { return okTwinSymbols.count(pos.first); });
+
+ if ( options.mode == modeDuplicatesSummary ) {
+ __block std::map<std::string, int> dylibDuplicatesCount;
+ for (const auto& pos : symbolsToInstallNames) {
+ const std::vector<const char*>& paths = pos.second;
+ if ( paths.size() <= 1 )
+ continue;
+ for (const char* path : paths) {
+ dylibDuplicatesCount[path] += 1;
+ }
+ }
+ struct DupCount { const char* path; int count; };
+ std::vector<DupCount> summary;
+ for (const auto& pos : dylibDuplicatesCount) {
+ summary.push_back({pos.first.c_str(), pos.second});
+ }
+ std::sort(summary.begin(), summary.end(), [](const DupCount& l, const DupCount& r) -> bool {
+ return (l.count > r.count);
+ });
+ for ( const DupCount& entry : summary)
+ printf("% 5d %s\n", entry.count, entry.path);
+ }
+ else {
+ for (const auto& pos : symbolsToInstallNames) {
+ const std::vector<const char*>& paths = pos.second;
+ if ( paths.size() > 1 ) {
+ bool isWeakDef = (weakDefSymbols.count(pos.first) != 0);
+ printf("%s%s\n", pos.first.c_str(), (isWeakDef ? " [weak-def]" : ""));
+ for (const char* path : paths)
+ printf(" %s\n", path);
+ }
+ }
+ }
+ }
+ break;
+
case modeNone:
case modeInfo:
+ case modeTPROInfo:
+ case modeStats:
case modeSlideInfo:
case modeVerboseSlideInfo:
+ case modeFixupsInDylib:
case modeTextInfo:
case modeLocalSymbols:
case modeJSONMap:
+ case modeVerboseJSONMap:
case modeJSONDependents:
case modeSectionSizes:
case modeStrings:
@@ -2275,9 +4176,14 @@
case modeObjCProtocols:
case modeObjCImpCaches:
case modeObjCClasses:
+ case modeObjCClassLayout:
+ case modeObjCClassMethodLists:
+ case modeObjCClassHashTable:
case modeObjCSelectors:
case modeSwiftProtocolConformances:
+ case modeSwiftPtrTables:
case modeExtract:
+ case modeLookupVA:
break;
}
}