verdnatura-chat/ios/Pods/FirebaseCrashlytics/Crashlytics/Shared/FIRCLSMachO/FIRCLSMachO.m

510 lines
14 KiB
Objective-C

// Copyright 2019 Google
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "FIRCLSMachO.h"
#include <Foundation/Foundation.h>
#include <mach-o/dyld.h>
#include <mach-o/fat.h>
#include <mach-o/getsect.h>
#include <mach-o/ldsyms.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <dlfcn.h>
#include <fcntl.h>
#include <stdio.h>
#include <unistd.h>
// This is defined in newer versions of iOS/macOS in usr/include/mach/machine.h
#define CLS_CPU_SUBTYPE_ARM64E ((cpu_subtype_t)2)
static void FIRCLSMachOHeaderValues(FIRCLSMachOSliceRef slice,
const struct load_command** cmds,
uint32_t* cmdCount);
static bool FIRCLSMachOSliceIsValid(FIRCLSMachOSliceRef slice);
bool FIRCLSMachOFileInitWithPath(FIRCLSMachOFileRef file, const char* path) {
struct stat statBuffer;
if (!file || !path) {
return false;
}
file->fd = 0;
file->mappedFile = NULL;
file->mappedSize = 0;
file->fd = open(path, O_RDONLY);
if (file->fd < 0) {
// unable to open mach-o file
return false;
}
if (fstat(file->fd, &statBuffer) == -1) {
close(file->fd);
return false;
}
// We need some minimum size for this to even be a possible mach-o file. I believe
// its probably quite a bit bigger than this, but this at least covers something.
// We also need it to be a regular file.
file->mappedSize = (size_t)statBuffer.st_size;
if (statBuffer.st_size < 16 || !(statBuffer.st_mode & S_IFREG)) {
close(file->fd);
return false;
}
// Map the file to memory. MAP_SHARED can potentially reduce the amount of actual private
// memory needed to do this mapping. Also, be sure to check for the correct failure result.
file->mappedFile = mmap(0, file->mappedSize, PROT_READ, MAP_FILE | MAP_SHARED, file->fd, 0);
if (!file->mappedFile || (file->mappedFile == MAP_FAILED)) {
close(file->fd);
return false;
}
return true;
}
bool FIRCLSMachOFileInitWithCurrent(FIRCLSMachOFileRef file) {
struct FIRCLSMachOSlice slice = FIRCLSMachOSliceGetCurrent();
const char* imagePath = FIRCLSMachOSliceGetExecutablePath(&slice);
return FIRCLSMachOFileInitWithPath(file, imagePath);
}
void FIRCLSMachOFileDestroy(FIRCLSMachOFileRef file) {
if (!file) {
return;
}
if (file->mappedFile && file->mappedSize > 0) {
munmap(file->mappedFile, file->mappedSize);
}
close(file->fd);
}
void FIRCLSMachOFileEnumerateSlices(FIRCLSMachOFileRef file, FIRCLSMachOSliceIterator block) {
FIRCLSMachOEnumerateSlicesAtAddress(file->mappedFile, block);
}
void FIRCLSMachOEnumerateSlicesAtAddress(void* executableData, FIRCLSMachOSliceIterator block) {
// check the magic value, to determine if we have a fat header or not
uint32_t magicValue;
uint32_t archCount;
const struct fat_arch* fatArch;
struct FIRCLSMachOSlice slice;
memset(&slice, 0, sizeof(struct FIRCLSMachOSlice));
magicValue = ((struct fat_header*)executableData)->magic;
if ((magicValue != FAT_MAGIC) && (magicValue != FAT_CIGAM)) {
slice.startAddress = executableData;
// use this to fill in the values
FIRCLSMachOHeaderValues(&slice, NULL, NULL);
block(&slice);
return;
}
archCount = OSSwapBigToHostInt32(((struct fat_header*)executableData)->nfat_arch);
fatArch = executableData + sizeof(struct fat_header);
for (uint32_t i = 0; i < archCount; ++i) {
slice.cputype = OSSwapBigToHostInt32(fatArch->cputype);
slice.cpusubtype = OSSwapBigToHostInt32(fatArch->cpusubtype);
slice.startAddress = executableData + OSSwapBigToHostInt32(fatArch->offset);
block(&slice);
// advance to the next fat_arch structure
fatArch = (struct fat_arch*)((uintptr_t)fatArch + sizeof(struct fat_arch));
}
}
struct FIRCLSMachOSlice FIRCLSMachOFileSliceWithArchitectureName(FIRCLSMachOFileRef file,
const char* name) {
__block struct FIRCLSMachOSlice value;
memset(&value, 0, sizeof(struct FIRCLSMachOSlice));
FIRCLSMachOFileEnumerateSlices(file, ^(FIRCLSMachOSliceRef slice) {
if (strcmp(FIRCLSMachOSliceGetArchitectureName(slice), name) == 0) {
value = *slice;
}
});
return value;
}
static void FIRCLSMachOHeaderValues(FIRCLSMachOSliceRef slice,
const struct load_command** cmds,
uint32_t* cmdCount) {
const struct mach_header* header32 = (const struct mach_header*)slice->startAddress;
const struct mach_header_64* header64 = (const struct mach_header_64*)slice->startAddress;
uint32_t commandCount;
const void* commandsAddress;
if (cmds) {
*cmds = NULL;
}
if (cmdCount) {
*cmdCount = 0;
}
if (!slice->startAddress) {
return;
}
// the 32 and 64 bit versions have an identical structures, so this will work
switch (header32->magic) {
case MH_MAGIC: // 32-bit
case MH_CIGAM:
slice->cputype = header32->cputype;
slice->cpusubtype = header32->cpusubtype;
commandCount = header32->ncmds;
commandsAddress = slice->startAddress + sizeof(struct mach_header);
break;
case MH_MAGIC_64: // 64-bit
case MH_CIGAM_64:
slice->cputype = header64->cputype;
slice->cpusubtype = header64->cpusubtype;
commandCount = header64->ncmds;
commandsAddress = slice->startAddress + sizeof(struct mach_header_64);
break;
default:
// not a valid header
return;
}
// assign everything back by reference
if (cmds) {
*cmds = commandsAddress;
}
if (cmdCount) {
*cmdCount = commandCount;
}
}
static bool FIRCLSMachOSliceIsValid(FIRCLSMachOSliceRef slice) {
if (!slice) {
return false;
}
if (!slice->startAddress) {
return false;
}
return true;
}
void FIRCLSMachOSliceEnumerateLoadCommands(FIRCLSMachOSliceRef slice,
FIRCLSMachOLoadCommandIterator block) {
const struct load_command* cmd;
uint32_t cmdCount;
if (!block) {
return;
}
if (!FIRCLSMachOSliceIsValid(slice)) {
return;
}
FIRCLSMachOHeaderValues(slice, &cmd, &cmdCount);
for (uint32_t i = 0; cmd != NULL && i < cmdCount; ++i) {
block(cmd->cmd, cmd->cmdsize, cmd);
cmd = (struct load_command*)((uintptr_t)cmd + cmd->cmdsize);
}
}
struct FIRCLSMachOSlice FIRCLSMachOSliceGetCurrent(void) {
const NXArchInfo* archInfo;
struct FIRCLSMachOSlice slice;
void* executableSymbol;
Dl_info dlinfo;
archInfo = NXGetLocalArchInfo();
if (archInfo) {
slice.cputype = archInfo->cputype;
slice.cpusubtype = archInfo->cpusubtype;
}
slice.startAddress = NULL;
// This call can fail when Exported Symbols File in Build Settings is missing the symbol value
// defined as _MH_EXECUTE_SYM (if you look in the header the underscored MH_EXECUTE_SYM define is
// there)
executableSymbol = dlsym(RTLD_MAIN_ONLY, MH_EXECUTE_SYM);
// get the address of the main function
if (dladdr(executableSymbol, &dlinfo) != 0) {
slice.startAddress = dlinfo.dli_fbase;
}
return slice;
}
struct FIRCLSMachOSlice FIRCLSMachOSliceWithHeader(void* machHeader) {
struct FIRCLSMachOSlice slice;
slice.startAddress = machHeader;
return slice;
}
const char* FIRCLSMachOSliceGetExecutablePath(FIRCLSMachOSliceRef slice) {
Dl_info info;
if (!FIRCLSMachOSliceIsValid(slice)) {
return NULL;
}
// use dladdr here to look up the information we need for a binary image
if (dladdr(slice->startAddress, &info) == 0) {
return NULL;
}
return info.dli_fname;
}
const char* FIRCLSMachOSliceGetArchitectureName(FIRCLSMachOSliceRef slice) {
const NXArchInfo* archInfo;
// there are some special cases here for types not handled by earlier OSes
if (slice->cputype == CPU_TYPE_ARM && slice->cpusubtype == CPU_SUBTYPE_ARM_V7S) {
return "armv7s";
}
if (slice->cputype == (CPU_TYPE_ARM | CPU_ARCH_ABI64)) {
if (slice->cpusubtype == CLS_CPU_SUBTYPE_ARM64E) {
return "arm64e";
} else if (slice->cpusubtype == CPU_SUBTYPE_ARM64_ALL) {
return "arm64";
}
}
if (slice->cputype == (CPU_TYPE_ARM) && slice->cpusubtype == CPU_SUBTYPE_ARM_V7K) {
return "armv7k";
}
archInfo = NXGetArchInfoFromCpuType(slice->cputype, slice->cpusubtype);
if (!archInfo) {
return "unknown";
}
return archInfo->name;
}
bool FIRCLSMachOSliceIs64Bit(FIRCLSMachOSliceRef slice) {
// I'm pretty sure this is sufficient...
return (slice->cputype & CPU_ARCH_ABI64) == CPU_ARCH_ABI64;
}
bool FIRCLSMachOSliceGetSectionByName(FIRCLSMachOSliceRef slice,
const char* segName,
const char* sectionName,
const void** ptr) {
if (!ptr) {
return false;
}
*ptr = NULL; // make sure this is set before returning
FIRCLSMachOSection section;
if (!FIRCLSMachOSliceInitSectionByName(slice, segName, sectionName, &section)) {
return false;
}
// WARNING: this calculation isn't correct, but is here to maintain backwards
// compatibility for now with callers of FIRCLSMachOSliceGetSectionByName. All new
// users should be calling FIRCLSMachOSliceInitSectionByName
*ptr = (const void*)((uintptr_t)slice->startAddress + section.offset);
return true;
}
bool FIRCLSMachOSliceInitSectionByName(FIRCLSMachOSliceRef slice,
const char* segName,
const char* sectionName,
FIRCLSMachOSection* section) {
if (!FIRCLSMachOSliceIsValid(slice)) {
return false;
}
if (!section) {
return false;
}
memset(section, 0, sizeof(FIRCLSMachOSection));
if (FIRCLSMachOSliceIs64Bit(slice)) {
const struct section_64* sect =
getsectbynamefromheader_64(slice->startAddress, segName, sectionName);
if (!sect) {
return false;
}
section->addr = sect->addr;
section->size = sect->size;
section->offset = sect->offset;
} else {
const struct section* sect = getsectbynamefromheader(slice->startAddress, segName, sectionName);
if (!sect) {
return false;
}
section->addr = sect->addr;
section->size = sect->size;
section->offset = sect->offset;
}
return true;
}
// TODO: this is left in-place just to ensure that old crashltyics + new fabric are still compatible
// with each other. As a happy bonus, if that situation does come up, this will also fix the bug
// that was preventing compact unwind on arm64 + iOS 9 from working correctly.
void FIRCLSMachOSliceGetUnwindInformation(FIRCLSMachOSliceRef slice,
const void** ehFrame,
const void** unwindInfo) {
if (!unwindInfo && !ehFrame) {
return;
}
bool found = false;
intptr_t slide = 0;
// This is inefficient, but we have no other safe way to do this correctly. Modifying the
// FIRCLSMachOSlice structure is tempting, but could introduce weird binary-compatibility issues
// with version mis-matches.
for (uint32_t i = 0; i < _dyld_image_count(); ++i) {
const struct mach_header* header = _dyld_get_image_header(i);
if (header == slice->startAddress) {
found = true;
slide = _dyld_get_image_vmaddr_slide(i);
break;
}
}
// make sure we were able to find a matching value
if (!found) {
return;
}
FIRCLSMachOSection section;
if (unwindInfo) {
if (FIRCLSMachOSliceInitSectionByName(slice, SEG_TEXT, "__unwind_info", &section)) {
*unwindInfo = (void*)(section.addr + slide);
}
}
if (ehFrame) {
if (FIRCLSMachOSliceInitSectionByName(slice, SEG_TEXT, "__eh_frame", &section)) {
*ehFrame = (void*)(section.addr + slide);
}
}
}
uint8_t const* FIRCLSMachOGetUUID(const struct load_command* cmd) {
return ((const struct uuid_command*)cmd)->uuid;
}
const char* FIRCLSMachOGetDylibPath(const struct load_command* cmd) {
const struct dylib_command* dylibcmd = (const struct dylib_command*)cmd;
return (const char*)((uintptr_t)cmd + dylibcmd->dylib.name.offset);
}
bool FIRCLSMachOGetEncrypted(const struct load_command* cmd) {
return ((struct encryption_info_command*)cmd)->cryptid > 0;
}
static FIRCLSMachOVersion FIRCLSMachOVersionFromEncoded(uint32_t encoded) {
FIRCLSMachOVersion version;
version.major = (encoded & 0xffff0000) >> 16;
version.minor = (encoded & 0x0000ff00) >> 8;
version.bugfix = encoded & 0x000000ff;
return version;
}
FIRCLSMachOVersion FIRCLSMachOGetMinimumOSVersion(const struct load_command* cmd) {
return FIRCLSMachOVersionFromEncoded(((const struct version_min_command*)cmd)->version);
}
FIRCLSMachOVersion FIRCLSMachOGetLinkedSDKVersion(const struct load_command* cmd) {
return FIRCLSMachOVersionFromEncoded(((const struct version_min_command*)cmd)->sdk);
}
FIRCLSMachOSegmentCommand FIRCLSMachOGetSegmentCommand(const struct load_command* cmd) {
FIRCLSMachOSegmentCommand segmentCommand;
memset(&segmentCommand, 0, sizeof(FIRCLSMachOSegmentCommand));
if (!cmd) {
return segmentCommand;
}
if (cmd->cmd == LC_SEGMENT) {
struct segment_command* segCmd = (struct segment_command*)cmd;
memcpy(segmentCommand.segname, segCmd->segname, 16);
segmentCommand.vmaddr = segCmd->vmaddr;
segmentCommand.vmsize = segCmd->vmsize;
} else if (cmd->cmd == LC_SEGMENT_64) {
struct segment_command_64* segCmd = (struct segment_command_64*)cmd;
memcpy(segmentCommand.segname, segCmd->segname, 16);
segmentCommand.vmaddr = segCmd->vmaddr;
segmentCommand.vmsize = segCmd->vmsize;
}
return segmentCommand;
}
NSString* FIRCLSMachONormalizeUUID(CFUUIDBytes* uuidBytes) {
CFUUIDRef uuid = CFUUIDCreateFromUUIDBytes(kCFAllocatorDefault, *uuidBytes);
NSString* string = CFBridgingRelease(CFUUIDCreateString(kCFAllocatorDefault, uuid));
CFRelease(uuid);
return [[string stringByReplacingOccurrencesOfString:@"-" withString:@""] lowercaseString];
}
NSString* FIRCLSMachOFormatVersion(FIRCLSMachOVersion* version) {
if (!version) {
return nil;
}
return [NSString stringWithFormat:@"%d.%d.%d", version->major, version->minor, version->bugfix];
}