原文链接
当 App 启动的时候，系统首先会加载 APP 的可执行文件，然后获得 dyld 所在路径，加载 dyld，接着后面的事情就交给 dyld 了。dylb 是什么呢？
Dylb
全称 the dynamic link editor，动态链接器， 源码在这里 。
image 镜像
可以理解为程序中对应实例，可以是可执行文件、Framework、dylib、bundle 文件。一个 App 包含很多镜像，比如：Foundation、 CoreServices 等等。
ImageLoader
用来加载 image 镜像的工具。
Mach-O
全称为 Mach Object，是 MAC 下可执行文件格式，类似 windows exe 文件，主要包括以下三种类型。
-
executable程序可执行文件。 -
dylib动态链接库，类似windows dll、linux so。 -
bundle资源文件，使用dlopen加载。
引用 wwdc2016/406 ：
File Types:
Executable—Main binary for application
Dylib—Dynamic library (aka DSO or DLL)
Bundle—Dylib that cannot be linked, only dlopen(), e.g. plug-ins
Dylb Setup
我们回到 App 启动，在系统内核做好程序准备工作之后，交由 dyld 负责剩下的工作，我们看下 dyld 做了哪些事情。
image
以上摘自 wwdc2016/406 pdf 第59页。
load dylibs
Map all dependent dylibs, recurse，递归找到所有依赖的 dylibs（动态库）
rebase
Rebase all images，调整所有镜像内的指针，添加一个 slide 偏差值，对于 slide 的解释，WWDC 是这样说的：
Slide = actual_address - preferred_address
可能是出于安全考虑，引入了 ASLR，全称 Address Space Layout Randomization。
ASLR
.Address Space Layout Randomization
.Images load at random address
大概意思就是镜像会加载在随机的地址上，和 actual_address 会有一个偏差(slide)，dyld 需要修正这个偏差，来指向正确的地址。
bind
Bind all images，查询符号表，设置指向镜像外部的指针。
objc prepare images
objc prepare images，通知 runtime 准备镜像，这里做的事情比较多，主要是 runtime 的初始化，引用 WWDC 解释：
.Most ObjC set up done via rebasing and binding
.All ObjC class definitions are registered
.Non-fragile ivars offsets updated
.Categories are inserted into method lists
.Selectors are uniqued
大部分 ObjC 的初始化工作已经完成，接下来注册所有的 objc class，更新 ivars 偏移（runtime 2.0 新特性，二进制兼容），把分类的方法插入到方法列表里，再检查 selector 唯一性，具体实现可以看 map_images 。
initializers
到了这一步，dylib 开始调用 C++ 静态构造函数，然后调用 class load（父类优先调用），再调用 category load，接着调用 __attribute__((constructor) 的构造函数。最后调用 main 。
.C++ generates initializer for statically allocated objects
.ObjC +load methods
.Run "bottom up" so each initializer can call dylibs below it
.Lastly, Dyld calls main() in executable
到这里，整个启动流程基本就结束了。
_objc_init 补充
在启动初始化的时候，会调用 _objc_init，这里做一些准备工作，比如说加载环境变量、初始化静态构造函数，注册 镜像映射、镜像加载、镜像卸载回调等等。我们打开 objc-os.mm，找到以下代码：
/***********************************************************************
* _objc_init
* Bootstrap initialization. Registers our image notifier with dyld.
* Called by libSystem BEFORE library initialization time
**********************************************************************/
void _objc_init(void)
{
static bool initialized = false;
if (initialized) return;
initialized = true;
// fixme defer initialization until an objc-using image is found?
environ_init();//环境变量
tls_init();
static_init();//静态构造函数
lock_init();
exception_init();
//注册 images 回调
_dyld_objc_notify_register(&map_images, load_images, unmap_image);
}
environ_init
读取环境配置方法，读取在 Xcode 中配置的环境变量参数。具体可以看 objc-env.h 文件。
OPTION( PrintImages, OBJC_PRINT_IMAGES, "log image and library names as they are loaded")
OPTION( PrintImageTimes, OBJC_PRINT_IMAGE_TIMES, "measure duration of image loading steps")
OPTION( PrintLoading, OBJC_PRINT_LOAD_METHODS, "log calls to class and category +load methods")
OPTION( PrintInitializing, OBJC_PRINT_INITIALIZE_METHODS, "log calls to class +initialize methods")
//省略...
tls_init
初始化线程的析构函数，具体可以看 objc-runtime.mm 文件。
void tls_init(void)
{
#if SUPPORT_DIRECT_THREAD_KEYS
_objc_pthread_key = TLS_DIRECT_KEY;
pthread_key_init_np(TLS_DIRECT_KEY, &_objc_pthread_destroyspecific);
#else
_objc_pthread_key = tls_create(&_objc_pthread_destroyspecific);
#endif
}
static_init
调用 C++的静态构造函数，具体可以看 objc-os.mm 文件。
/***********************************************************************
* static_init
* Run C++ static constructor functions.
* libc calls _objc_init() before dyld would call our static constructors,
* so we have to do it ourselves.
**********************************************************************/
static void static_init()
{
size_t count;
Initializer *inits = getLibobjcInitializers(&_mh_dylib_header, &count);
for (size_t i = 0; i < count; i++) {
inits[i]();
}
}
lock_init
初始化锁，具体可以看 objc-runtime-new.mm 文件。
void lock_init(void)
{
#if SUPPORT_QOS_HACK
BackgroundPriority = _pthread_qos_class_encode(QOS_CLASS_BACKGROUND, 0, 0);
MainPriority = _pthread_qos_class_encode(qos_class_main(), 0, 0);
# if DEBUG
pthread_key_init_np(QOS_KEY, &destroyQOSKey);
# endif
#endif
}
exception_init
初始化 exception handle，具体可以看 objc-exception.mm 文件。
/***********************************************************************
* exception_init
* Initialize libobjc's exception handling system.
* Called by map_images().
**********************************************************************/
void exception_init(void)
{
old_terminate = std::set_terminate(&_objc_terminate);
}
_dyld_objc_notify_register
注册 准备镜像、加载镜像、卸载镜像的回调，每当有新的镜像被加载的时候，都会调用这些回调。
map_images
当 runtime 收到 dylb 的准备镜像通知 的时候，开始初始化 runtime，注册 objc class，更新 ivars offset，把 category 方法合到主类等等，打开 objc-runtime-new.mm，找到以下代码：
/***********************************************************************
* map_images
* Process the given images which are being mapped in by dyld.
* Calls ABI-agnostic code after taking ABI-specific locks.
*
* Locking: write-locks runtimeLock
**********************************************************************/
void
map_images(unsigned count, const char * const paths[],
const struct mach_header * const mhdrs[])
{
rwlock_writer_t lock(runtimeLock);
return map_images_nolock(count, paths, mhdrs);
}
进入方法，先加锁，这里使用了读写锁，然后交给 map_images_nolock 处理。
map_images_nolock
准备镜像具体实现，实现共享内存优化，默认方法注册、自动释放池和散列表初始化及类的加载等等操作。方法比较长，截取了部分：
void
map_images_nolock(unsigned mhCount, const char * const mhPaths[],
const struct mach_header * const mhdrs[])
{
static bool firstTime = YES;
header_info *hList[mhCount];
uint32_t hCount;
size_t selrefCount = 0;
// Perform first-time initialization if necessary.
// This function is called before ordinary library initializers.
// fixme defer initialization until an objc-using image is found?
if (firstTime) {
//预优化初始化
preopt_init();
}
//统计class 数量
// Find all images with Objective-C metadata.
hCount = 0;
// Count classes. Size various table based on the total.
int totalClasses = 0;
// Perform one-time runtime initialization that must be deferred until
// the executable itself is found. This needs to be done before
// further initialization.
// (The executable may not be present in this infoList if the
// executable does not contain Objective-C code but Objective-C
// is dynamically loaded later.
if (firstTime) {
sel_init(selrefCount);
// 自动释放池和散列表初始化
arr_init();
//读取镜像
if (hCount > 0) {
_read_images(hList, hCount, totalClasses, unoptimizedTotalClasses);
}
firstTime = NO;
}
//自动释放池、散列表初始化
void arr_init(void)
{
AutoreleasePoolPage::init();
SideTableInit();
}
最后读取镜像，调用 _read_images。
_read_images
读取镜像，方法内做了很多事情，加载类、注册方法、加载虚函数表、加载协议 Protocol 和非延迟类方法、加载静态实例、加载分类。代码太多，截取了部分，具体可以看 objc-runtime-new.mm。
/***********************************************************************
* _read_images
* Perform initial processing of the headers in the linked
* list beginning with headerList.
*
* Called by: map_images_nolock
*
* Locking: runtimeLock acquired by map_images
**********************************************************************/
void _read_images(header_info **hList, uint32_t hCount, int totalClasses, int unoptimizedTotalClasses)
{
header_info *hi;
uint32_t hIndex;
size_t count;
size_t i;
Class *resolvedFutureClasses = nil;
//加载类
for (EACH_HEADER) {
if (! mustReadClasses(hi)) {
// Image is sufficiently optimized that we need not call readClass()
continue;
}
bool headerIsBundle = hi->isBundle();
bool headerIsPreoptimized = hi->isPreoptimized();
classref_t *classlist = _getObjc2ClassList(hi, &count);
for (i = 0; i < count; i++) {
Class cls = (Class)classlist[i];
Class newCls = readClass(cls, headerIsBundle, headerIsPreoptimized);
if (newCls != cls && newCls) {
// Class was moved but not deleted. Currently this occurs
// only when the new class resolved a future class.
// Non-lazily realize the class below.
resolvedFutureClasses = (Class *)
realloc(resolvedFutureClasses,
(resolvedFutureClassCount+1) * sizeof(Class));
resolvedFutureClasses[resolvedFutureClassCount++] = newCls;
}
}
}
//注册方法
// Fix up @selector references
static size_t UnfixedSelectors;
sel_lock();
for (EACH_HEADER) {
if (hi->isPreoptimized()) continue;
bool isBundle = hi->isBundle();
SEL *sels = _getObjc2SelectorRefs(hi, &count);
UnfixedSelectors += count;
for (i = 0; i < count; i++) {
const char *name = sel_cname(sels[i]);
sels[i] = sel_registerNameNoLock(name, isBundle);
}
}
sel_unlock();
//加载虚函数表
#if SUPPORT_FIXUP
// Fix up old objc_msgSend_fixup call sites
for (EACH_HEADER) {
message_ref_t *refs = _getObjc2MessageRefs(hi, &count);
if (count == 0) continue;
if (PrintVtables) {
_objc_inform("VTABLES: repairing %zu unsupported vtable dispatch "
"call sites in %s", count, hi->fname());
}
for (i = 0; i < count; i++) {
fixupMessageRef(refs+i);
}
}
#endif
//加载协议
// Discover protocols. Fix up protocol refs.
for (EACH_HEADER) {
extern objc_class OBJC_CLASS_$_Protocol;
Class cls = (Class)&OBJC_CLASS_$_Protocol;
assert(cls);
NXMapTable *protocol_map = protocols();
bool isPreoptimized = hi->isPreoptimized();
bool isBundle = hi->isBundle();
protocol_t **protolist = _getObjc2ProtocolList(hi, &count);
for (i = 0; i < count; i++) {
readProtocol(protolist[i], cls, protocol_map,
isPreoptimized, isBundle);
}
}
// Fix up @protocol references
// Preoptimized images may have the right
// answer already but we don't know for sure.
for (EACH_HEADER) {
protocol_t **protolist = _getObjc2ProtocolRefs(hi, &count);
for (i = 0; i < count; i++) {
remapProtocolRef(&protolist[i]);
}
}
//重新布局class
// Realize non-lazy classes (for +load methods and static instances)
for (EACH_HEADER) {
classref_t *classlist =
_getObjc2NonlazyClassList(hi, &count);
for (i = 0; i < count; i++) {
Class cls = remapClass(classlist[i]);
if (!cls) continue;
// hack for class __ARCLite__, which didn't get this above
#if TARGET_OS_SIMULATOR
if (cls->cache._buckets == (void*)&_objc_empty_cache &&
(cls->cache._mask || cls->cache._occupied))
{
cls->cache._mask = 0;
cls->cache._occupied = 0;
}
if (cls->ISA()->cache._buckets == (void*)&_objc_empty_cache &&
(cls->ISA()->cache._mask || cls->ISA()->cache._occupied))
{
cls->ISA()->cache._mask = 0;
cls->ISA()->cache._occupied = 0;
}
#endif
realizeClass(cls);
}
}
//加载分类 category
// Discover categories.
for (EACH_HEADER) {
category_t **catlist =
_getObjc2CategoryList(hi, &count);
bool hasClassProperties = hi->info()->hasCategoryClassProperties();
for (i = 0; i < count; i++) {
category_t *cat = catlist[i];
Class cls = remapClass(cat->cls);
if (!cls) {
// Category's target class is missing (probably weak-linked).
// Disavow any knowledge of this category.
catlist[i] = nil;
if (PrintConnecting) {
_objc_inform("CLASS: IGNORING category \?\?\?(%s) %p with "
"missing weak-linked target class",
cat->name, cat);
}
continue;
}
// Process this category.
// First, register the category with its target class.
// Then, rebuild the class's method lists (etc) if
// the class is realized.
bool classExists = NO;
if (cat->instanceMethods || cat->protocols
|| cat->instanceProperties)
{
addUnattachedCategoryForClass(cat, cls, hi);
if (cls->isRealized()) {
remethodizeClass(cls);
classExists = YES;
}
if (PrintConnecting) {
_objc_inform("CLASS: found category -%s(%s) %s",
cls->nameForLogging(), cat->name,
classExists ? "on existing class" : "");
}
}
if (cat->classMethods || cat->protocols
|| (hasClassProperties && cat->_classProperties))
{
addUnattachedCategoryForClass(cat, cls->ISA(), hi);
if (cls->ISA()->isRealized()) {
remethodizeClass(cls->ISA());
}
if (PrintConnecting) {
_objc_inform("CLASS: found category +%s(%s)",
cls->nameForLogging(), cat->name);
}
}
}
}
}
到此镜像映射完成。
load_images
当 runtime 收到 dylb 的加载镜像通知 的时候，会调用这个方法，作用是加载镜像，打开 objc-runtime-new.mm，找到以下代码：
void
load_images(const char *path __unused, const struct mach_header *mh)
{
// Return without taking locks if there are no +load methods here.
if (!hasLoadMethods((const headerType *)mh)) return;
recursive_mutex_locker_t lock(loadMethodLock);
// Discover load methods
{
rwlock_writer_t lock2(runtimeLock);
prepare_load_methods((const headerType *)mh);
}
// Call +load methods (without runtimeLock - re-entrant)
call_load_methods();
}
不难发现 load_images 主要做了两件事，先调用 prepare_load_methods 进行 load 准备，接着调用 call_load_methods 执行所有的 load 方法。
prepare_load_methods
//load 预处理
void prepare_load_methods(const headerType *mhdr)
{
size_t count, i;
runtimeLock.assertWriting();
classref_t *classlist =
_getObjc2NonlazyClassList(mhdr, &count);
//这里先处理所有class load
for (i = 0; i < count; i++) {
schedule_class_load(remapClass(classlist[i]));
}
//再处理 category load
category_t **categorylist = _getObjc2NonlazyCategoryList(mhdr, &count);
for (i = 0; i < count; i++) {
category_t *cat = categorylist[i];
Class cls = remapClass(cat->cls);
if (!cls) continue; // category for ignored weak-linked class
realizeClass(cls);
assert(cls->ISA()->isRealized());
//添加到 loadable_categories 全局结构体里
add_category_to_loadable_list(cat);
}
}
不难发现，先调用 schedule_class_load 处理 class load，然后再处理 category load。那么处理 class load，父类和子类是什么顺序呢？我们继续看代码：
//处理 class load，superclass 的在前面
static void schedule_class_load(Class cls)
{
if (!cls) return;
assert(cls->isRealized()); // _read_images should realize
if (cls->data()->flags & RW_LOADED) return;
//这里先添加 superClass 的 load
// Ensure superclass-first ordering
schedule_class_load(cls->superclass);
//再添加 class load
add_class_to_loadable_list(cls);
cls->setInfo(RW_LOADED);
}
//处理 category load
void add_category_to_loadable_list(Category cat)
{
IMP method;
loadMethodLock.assertLocked();
method = _category_getLoadMethod(cat);
if (loadable_categories_used == loadable_categories_allocated) {
loadable_categories_allocated = loadable_categories_allocated*2 + 16;
loadable_categories = (struct loadable_category *)
realloc(loadable_categories,
loadable_categories_allocated *
sizeof(struct loadable_category));
}
loadable_categories[loadable_categories_used].cat = cat;
loadable_categories[loadable_categories_used].method = method;
loadable_categories_used++;
}
schedule_class_load 内部写的很清楚，优先存储 superclass load，然后再调用 add_class_to_loadable_list 存储自身的 class load。到这里 load 准备工作做完了。
call_load_methods
接下来开始调用所有的 load，我们先看源码：
//调用所有的 load 方法
void call_load_methods(void)
{
static bool loading = NO;
bool more_categories;
loadMethodLock.assertLocked();
// Re-entrant calls do nothing; the outermost call will finish the job.
if (loading) return;
loading = YES;
void *pool = objc_autoreleasePoolPush();
do {
// 1\. Repeatedly call class +loads until there aren't any more
while (loadable_classes_used > 0) {
//循环调用 class 的 load 方法，直到完成所有调用为止
call_class_loads();
}
// 调用分类的 load 方法
// 2\. Call category +loads ONCE
more_categories = call_category_loads();
// 3\. Run more +loads if there are classes OR more untried categories
} while (loadable_classes_used > 0 || more_categories);
objc_autoreleasePoolPop(pool);
loading = NO;
}
很明显，这里先用一个 while 循环 执行 class load，然后再调用 category load。
注意
-
class load和category load不是通过方法查找调用load方法，而是通过方法列表去到load方法的IMP方法地址直接调用。所以类的和分类的load方法 都会调用。
总结
最后，最后我们总结一下：
- 准备
load方法的时候，先处理superclass load，再处理子类；最后准备category load，单独存储。
- 调用类的
load方法。- 先编译先调用。
- 先调用父类的
load方法，再调用子类的load方法。
- 调用Category的
load方法。- 先编译先调用。
-
class load和category load不是通过方法查找调用load方法，而是通过方法列表去到load方法的IMP方法地址直接调用。所以类的和分类的load方法 都会调用。 - 一个类的只调用
load方法一次。
参考链接
Optimizing App Startup Time
dyld