chromium中HTTP网络资源的加载过程

来源:互联网 发布:matlab矩阵逻辑运算 编辑:程序博客网 时间:2024/04/28 11:10

chromium中HTTP网络资源的加载主要分两部分,一部分是缓存的网络资源,一部分是线上的网络资源。

我们访问http网页的时候,首先访问httpcache,看缓存中的数据是否有效,如果有效,那么我们加载这个数据,如果无效,那么我们访问网络去重新加载资源,当然chromium中HTTP网络资源的加载并没有说起来这么简单,实际上架构设计还是比较复杂的。首先我们先看看HttpCache::Transaction的相关设计及实现,研究一下缓存的处理。

HttpCache::Transaction

在源码分析之前,我们先看一个小示例,从使用者的角度看看cache的使用过程,那么就需要一个切入点。

情景分析

首先我们在windbg中启动chromium,然后设置多进程调试。

0:000> .childdbg 1Processes created by the current process will be debugged0:000> g

运行后待程序稳定后,中断到调试器,之后在主进程上下断

0:011> bp chrome_7feed810000!net::HttpCache::Transaction::Start

之后运行起来,然后在浏览器的地址窗口中输入一个字符“c”,这是我假定的设置的,因为浏览器的地址窗口默认设置了百度的搜索引擎,当用户输入一个字符的时候会触发一次搜索。而实际上确实是这样。当输入字符后,我的程序中断了。

0:028> kc # Call Site00 chrome_7feed810000!net::HttpCache::Transaction::Start01 chrome_7feed810000!net::URLRequestHttpJob::StartTransactionInternal02 chrome_7feed810000!net::URLRequestHttpJob::MaybeStartTransactionInternal03 chrome_7feed810000!net::URLRequestHttpJob::StartTransaction04 chrome_7feed810000!net::URLRequestHttpJob::SetCookieHeaderAndStart05 chrome_7feed810000!base::internal::RunnableAdapter<void (__cdecl LocalDataContainer::*)(std::list<content::CacheStorageUsageInfo,std::allocator<content::CacheStorageUsageInfo> > const &)>::Run06 chrome_7feed810000!base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl LocalDataContainer::*)(std::list<content::CacheStorageUsageInfo,std::allocator<content::CacheStorageUsageInfo> > const &)> >::MakeItSo07 chrome_7feed810000!base::internal::Invoker<base::IndexSequence<0>,base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl LocalDataContainer::*)(std::list<content::CacheStorageUsageInfo,std::allocator<content::CacheStorageUsageInfo> > const & __ptr64) __ptr64>,void __cdecl(LocalDataContainer * __ptr64,std::list<content::CacheStorageUsageInfo,std::allocator<content::CacheStorageUsageInfo> > const & __ptr64),base::WeakPtr<LocalDataContainer> >,base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl LocalDataContainer::*)(std::list<content::CacheStorageUsageInfo,std::allocator<content::CacheStorageUsageInfo> > const & __ptr64) __ptr64> >,void __cdecl(std::list<content::CacheStorageUsageInfo,std::allocator<content::CacheStorageUsageInfo> > const & __ptr64)>::Run08 chrome_7feed810000!base::Callback<void __cdecl(void)>::Run09 chrome_7feed810000!net::CookieMonster::CookieMonsterTask::InvokeCallback0a chrome_7feed810000!net::CookieMonster::GetCookieListWithOptionsTask::Run0b chrome_7feed810000!net::CookieMonster::DoCookieTaskForURL0c chrome_7feed810000!net::CookieMonster::GetCookieListWithOptionsAsync0d chrome_7feed810000!net::URLRequestHttpJob::AddCookieHeaderAndStart0e chrome_7feed810000!net::URLRequestHttpJob::Start0f chrome_7feed810000!net::URLRequest::StartJob10 chrome_7feed810000!net::URLRequest::BeforeRequestComplete11 chrome_7feed810000!net::URLRequest::Start12 chrome_7feed810000!net::URLFetcherCore::StartURLRequest13 chrome_7feed810000!net::URLFetcherCore::StartURLRequestWhenAppropriate14 chrome_7feed810000!net::URLFetcherCore::DidInitializeWriter15 chrome_7feed810000!net::URLFetcherCore::StartOnIOThread16 chrome_7feed810000!base::Callback<void __cdecl(void)>::Run17 chrome_7feed810000!base::debug::TaskAnnotator::RunTask18 chrome_7feed810000!base::MessageLoop::RunTask19 chrome_7feed810000!base::MessageLoop::DeferOrRunPendingTask1a chrome_7feed810000!base::MessageLoop::DoWork1b chrome_7feed810000!base::MessagePumpForIO::DoRunLoop1c chrome_7feed810000!base::MessagePumpWin::Run1d chrome_7feed810000!base::MessageLoop::RunHandler1e chrome_7feed810000!base::RunLoop::Run1f chrome_7feed810000!base::MessageLoop::Run20 chrome_7feed810000!base::Thread::Run21 chrome_7feed810000!content::BrowserThreadImpl::IOThreadRun22 chrome_7feed810000!content::BrowserThreadImpl::Run23 chrome_7feed810000!base::Thread::ThreadMain24 chrome_7feed810000!base::`anonymous namespace'::ThreadFunc25 kernel32!BaseThreadInitThunk26 ntdll!RtlUserThreadStart

研究一下堆栈信息,其实就已经知道了大抵上的调用层次,这个请求的方式和我之前介绍过的chromium中FTP网络资源的加载 文章中很相似。

在一个IO线程中启动了URL的加载请求,这是一个内部请求,使用的是URLFetcherCore,然后启动URLRequest,之后开启一个特定的job,在这里这个job是URLRequestHttpJob,之后启动一个传输,而这个传输就是我们要说的HttpCache::Transaction,内部数据的请求基于HttpCache。

0:028> ~~[1244]s;.frame 0n0;dv /t /vchrome_7feed810000!net::HttpCache::Transaction::Start:000007fe`ee46c6e4 48895c2410      mov     qword ptr [rsp+10h],rbx ss:00000000`0961c408=0000000012695a6000 00000000`0961c3f8 000007fe`ee4bf027 chrome_7feed810000!net::HttpCache::Transaction::Start [c:\b\build\slave\win64\build\src\net\http\http_cache_transaction.cc @ 254]@rcx              class net::HttpCache::Transaction * this = 0x00000000`1269c170@rdx              struct net::HttpRequestInfo * request = 0x00000000`12695c28@r8               class base::Callback<void __cdecl(int)> * callback = 0x00000000`12695d40@r9               class net::BoundNetLog * net_log = 0x00000000`12691ec8<unavailable>     int rv = <value unavailable>0:028> dx -id 0,0 -r1 (*((chrome_7feed810000!net::HttpRequestInfo *)0x12695c28))(*((chrome_7feed810000!net::HttpRequestInfo *)0x12695c28))                 [Type: net::HttpRequestInfo]    [+0x000] url              [Type: GURL]    [+0x078] method           : "GET" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x098] extra_headers    [Type: net::HttpRequestHeaders]    [+0x0b0] upload_data_stream : 0x0 [Type: net::UploadDataStream *]    [+0x0b8] load_flags       : 64    [+0x0bc] motivation       : NORMAL_MOTIVATION (2) [Type: net::HttpRequestInfo::RequestMotivation]    [+0x0c0] privacy_mode     : PRIVACY_MODE_DISABLED (0) [Type: net::PrivacyMode]0:028> dx -id 0,0 -r1 (*((chrome_7feed810000!GURL *)0x12695c28))(*((chrome_7feed810000!GURL *)0x12695c28))                 [Type: GURL]    [+0x000] spec_            : "http://suggestion.baidu.com/su?wd=c&action=opensearch&ie=UTF-8" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x020] is_valid_        : true    [+0x028] parsed_          [Type: url::Parsed]    [+0x070] inner_url_       [Type: scoped_ptr<GURL,std::default_delete<GURL> >]

接着,我们来看一下,当我们在浏览器中输入一个字符的时候,浏览器的一个URL的请求过程,上面的网址就是要请求的URL,“http://suggestion.baidu.com/su?wd=c&action=opensearch&ie=UTF-8”,这是一个http GET命令。Opensearch 字符”c”.
当然,浏览器地址栏的结果数据的提供过程并不是我们讨论的主题,我们是通过这个切入点去详细了解其中一点,http网络资源的加载过程。

源码分析

HttpCache::Transaction::Start

上面我们看到了最后一个函数栈,是一个缓存传输启动的过程,我们从这里开始,已知的一点是我们向启动函数中传递了请求信息,包含了URL和method。

int HttpCache::Transaction::Start(const HttpRequestInfo* request,                                  const CompletionCallback& callback,                                  const BoundNetLog& net_log) {......  if (!cache_.get())    return ERR_UNEXPECTED;  SetRequest(net_log, request);  // We have to wait until the backend is initialized so we start the SM.  next_state_ = STATE_GET_BACKEND;  int rv = DoLoop(OK);  // Setting this here allows us to check for the existence of a callback_ to  // determine if we are still inside Start.  if (rv == ERR_IO_PENDING)    callback_ = callback;  return rv;}

我们来看一下源码,这里我们设置了下一跳状态,然后启动状态机循环,这里多少和FTP资源的加载有些相似,FTP的加载要较为简单些。

HttpCache::Transaction::State

  enum State {    // Normally, states are traversed in approximately this order.    STATE_NONE,    STATE_GET_BACKEND,    STATE_GET_BACKEND_COMPLETE,    STATE_INIT_ENTRY,    STATE_OPEN_ENTRY,    STATE_OPEN_ENTRY_COMPLETE,    STATE_DOOM_ENTRY,    STATE_DOOM_ENTRY_COMPLETE,    STATE_CREATE_ENTRY,    STATE_CREATE_ENTRY_COMPLETE,    STATE_ADD_TO_ENTRY,    STATE_ADD_TO_ENTRY_COMPLETE,    .....    }

HttpCache的传输过程也是状态改变的过程,状态比较多,情况就比较多。
chromium源码作者还是比较贴心的,在源码状态机循环函数前以注释的方式列出了各种情况。

// 1. Not-cached entry://   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> CreateEntry* -> AddToEntry* ->//   SendRequest* -> SuccessfulSendRequest -> OverwriteCachedResponse ->//   CacheWriteResponse* -> TruncateCachedData* -> TruncateCachedMetadata* ->//   PartialHeadersReceived////   Read()://   NetworkRead* -> CacheWriteData*//// 2. Cached entry, no validation://   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> AddToEntry* -> CacheReadResponse*//   -> CacheDispatchValidation -> BeginPartialCacheValidation() ->//   BeginCacheValidation() -> SetupEntryForRead()////   Read()://   CacheReadData*//// 3. Cached entry, validation (304)://   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> AddToEntry* -> CacheReadResponse*//   -> CacheDispatchValidation -> BeginPartialCacheValidation() ->//   BeginCacheValidation() -> SendRequest* -> SuccessfulSendRequest ->//   UpdateCachedResponse -> CacheWriteUpdatedResponse* ->//   UpdateCachedResponseComplete -> OverwriteCachedResponse ->//   PartialHeadersReceived////   Read()://   CacheReadData*//// 4. Cached entry, validation and replace (200)://   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> AddToEntry* -> CacheReadResponse*//   -> CacheDispatchValidation -> BeginPartialCacheValidation() ->//   BeginCacheValidation() -> SendRequest* -> SuccessfulSendRequest ->//   OverwriteCachedResponse -> CacheWriteResponse* -> DoTruncateCachedData* ->//   TruncateCachedMetadata* -> PartialHeadersReceived////   Read()://   NetworkRead* -> CacheWriteData*//// 5. Sparse entry, partially cached, byte range request://   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> AddToEntry* -> CacheReadResponse*//   -> CacheDispatchValidation -> BeginPartialCacheValidation() ->//   CacheQueryData* -> ValidateEntryHeadersAndContinue() ->//   StartPartialCacheValidation -> CompletePartialCacheValidation ->//   BeginCacheValidation() -> SendRequest* -> SuccessfulSendRequest ->//   UpdateCachedResponse -> CacheWriteUpdatedResponse* ->//   UpdateCachedResponseComplete -> OverwriteCachedResponse ->//   PartialHeadersReceived////   Read() 1://   NetworkRead* -> CacheWriteData*////   Read() 2://   NetworkRead* -> CacheWriteData* -> StartPartialCacheValidation ->//   CompletePartialCacheValidation -> CacheReadData* ->////   Read() 3://   CacheReadData* -> StartPartialCacheValidation ->//   CompletePartialCacheValidation -> BeginCacheValidation() -> SendRequest* ->//   SuccessfulSendRequest -> UpdateCachedResponse* -> OverwriteCachedResponse//   -> PartialHeadersReceived -> NetworkRead* -> CacheWriteData*//// 6. HEAD. Not-cached entry://   Pass through. Don't save a HEAD by itself.//   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> SendRequest*//// 7. HEAD. Cached entry, no validation://   Start()://   The same flow as for a GET request (example #2)////   Read()://   CacheReadData (returns 0)//// 8. HEAD. Cached entry, validation (304)://   The request updates the stored headers.//   Start(): Same as for a GET request (example #3)////   Read()://   CacheReadData (returns 0)//// 9. HEAD. Cached entry, validation and replace (200)://   Pass through. The request dooms the old entry, as a HEAD won't be stored by//   itself.//   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> AddToEntry* -> CacheReadResponse*//   -> CacheDispatchValidation -> BeginPartialCacheValidation() ->//   BeginCacheValidation() -> SendRequest* -> SuccessfulSendRequest ->//   OverwriteCachedResponse//// 10. HEAD. Sparse entry, partially cached://   Serve the request from the cache, as long as it doesn't require//   revalidation. Ignore missing ranges when deciding to revalidate. If the//   entry requires revalidation, ignore the whole request and go to full pass//   through (the result of the HEAD request will NOT update the entry).////   Start(): Basically the same as example 7, as we never create a partial_//   object for this request.//// 11. Prefetch, not-cached entry://   The same as example 1. The "unused_since_prefetch" bit is stored as true in//   UpdateCachedResponse.//// 12. Prefetch, cached entry://   Like examples 2-4, only CacheToggleUnusedSincePrefetch* is inserted between//   CacheReadResponse* and CacheDispatchValidation if the unused_since_prefetch//   bit is unset.//// 13. Cached entry less than 5 minutes old, unused_since_prefetch is true://   Skip validation, similar to example 2.//   GetBackend* -> InitEntry -> OpenEntry* -> AddToEntry* -> CacheReadResponse*//   -> CacheToggleUnusedSincePrefetch* -> CacheDispatchValidation ->//   BeginPartialCacheValidation() -> BeginCacheValidation() ->//   SetupEntryForRead()////   Read()://   CacheReadData*//// 14. Cached entry more than 5 minutes old, unused_since_prefetch is true://   Like examples 2-4, only CacheToggleUnusedSincePrefetch* is inserted between//   CacheReadResponse* and CacheDispatchValidation.

简要看看,一般的缓存传输基本上都要经历的过程是获取Backend,然后初始化Entry,之后打开,然后从中加载进来,之后进行验证,后面的过程就有些不一样了。自然,从地址栏中输入单个字符引发搜索,URL资源请求也就符合其中一种情况。

操作Backend准备Cache数据

base::WeakPtr<HttpCache> cache_;
int HttpCache::Transaction::DoGetBackend() {  cache_pending_ = true;  next_state_ = STATE_GET_BACKEND_COMPLETE;  net_log_.BeginEvent(NetLog::TYPE_HTTP_CACHE_GET_BACKEND);  return cache_->GetBackendForTransaction(this);}int HttpCache::Transaction::DoInitEntry() {  DCHECK(!new_entry_);  if (!cache_.get())    return ERR_UNEXPECTED;  if (mode_ == WRITE) {    next_state_ = STATE_DOOM_ENTRY;    return OK;  }  next_state_ = STATE_OPEN_ENTRY;  return OK;}int HttpCache::Transaction::DoOpenEntry() {  DCHECK(!new_entry_);  next_state_ = STATE_OPEN_ENTRY_COMPLETE;  cache_pending_ = true;  net_log_.BeginEvent(NetLog::TYPE_HTTP_CACHE_OPEN_ENTRY);  first_cache_access_since_ = TimeTicks::Now();  return cache_->OpenEntry(cache_key_, &new_entry_, this);}

Httpcache的传输实际上操作httpCache来完成具体的功能,做前期的准备工作,最后通过cache_key_,来查询某一项。

[+0x3a0] cache_key_       : "http://suggestion.baidu.com/su?wd=c&action=opensearch&ie=UTF-8" 

这里面的cache_key_其实就是URL,在HttpCache中URL唯一检索资源,就是通过这个key值来找到对应的Entry。

int HttpCache::Transaction::DoAddToEntry() {  DCHECK(new_entry_);  cache_pending_ = true;  next_state_ = STATE_ADD_TO_ENTRY_COMPLETE;  net_log_.BeginEvent(NetLog::TYPE_HTTP_CACHE_ADD_TO_ENTRY);  DCHECK(entry_lock_waiting_since_.is_null());  entry_lock_waiting_since_ = TimeTicks::Now();  int rv = cache_->AddTransactionToEntry(new_entry_, this);  if (rv == ERR_IO_PENDING) {    if (bypass_lock_for_test_) {      OnAddToEntryTimeout(entry_lock_waiting_since_);    } else {      int timeout_milliseconds = 20 * 1000;      if (partial_ && new_entry_->writer &&          new_entry_->writer->range_requested_) {        timeout_milliseconds = 25;      }      base::ThreadTaskRunnerHandle::Get()->PostDelayedTask(          FROM_HERE,          base::Bind(&HttpCache::Transaction::OnAddToEntryTimeout,                     weak_factory_.GetWeakPtr(), entry_lock_waiting_since_),          TimeDelta::FromMilliseconds(timeout_milliseconds));    }  }  return rv;}

当我们找到对应的项后,我们具体的加载这个项的数据,如果返回IO_PENDING,那么设置超时回调函数。设置下一状态为STATE_ADD_TO_ENTRY_COMPLETE,在完成函数中,设置下一跳状态,记录处理时间。

HttpCache::Transaction Read Entry and validation

int HttpCache::Transaction::DoCacheReadResponse() {  DCHECK(entry_);  next_state_ = STATE_CACHE_READ_RESPONSE_COMPLETE;  io_buf_len_ = entry_->disk_entry->GetDataSize(kResponseInfoIndex);  read_buf_ = new IOBuffer(io_buf_len_);  net_log_.BeginEvent(NetLog::TYPE_HTTP_CACHE_READ_INFO);  return entry_->disk_entry->ReadData(kResponseInfoIndex, 0, read_buf_.get(),                                      io_buf_len_, io_callback_);}

当一切都准备完成后,开始从缓存中读取数据,函数中设置下一跳,然后从entry中获取到缓存的大小,申请空间,之后操作entry来读取数据。

int HttpCache::Transaction::DoCacheDispatchValidation() {  int result = ERR_FAILED;  switch (mode_) {    case READ:      UpdateTransactionPattern(PATTERN_ENTRY_USED);      result = BeginCacheRead();      break;    case READ_WRITE:      result = BeginPartialCacheValidation();      break;    case UPDATE:      result = BeginExternallyConditionalizedRequest();      break;    case WRITE:    default:      NOTREACHED();  }  return result;}

当我们读取完数据后,就开始了验证数据的过程。这里是个读写的情况。在派发验证的过程中,跳过了验证,符合上面的情况2,Cached entry, no validation。

HttpCache::Transaction 数据读取

当我们准备cache,操作并查找到对应的项,然后读取项,之后判断是否验证项,如果这都通过了之后,下层会通知上层来提取数据。

0:026> kc # Call Site00 chrome_7feed900000!net::HttpCache::Transaction::DoCacheReadData01 chrome_7feed900000!net::HttpCache::Transaction::DoLoop02 chrome_7feed900000!net::HttpCache::Transaction::Read03 chrome_7feed900000!net::URLRequestHttpJob::ReadRawData04 chrome_7feed900000!net::URLRequestJob::ReadRawDataHelper05 chrome_7feed900000!net::URLRequestJob::Read06 chrome_7feed900000!net::URLRequest::Read07 chrome_7feed900000!net::URLFetcherCore::ReadResponse08 chrome_7feed900000!net::URLFetcherCore::OnResponseStarted09 chrome_7feed900000!net::URLRequest::NotifyResponseStarted0a chrome_7feed900000!net::URLRequestJob::NotifyHeadersComplete0b chrome_7feed900000!net::URLRequestHttpJob::NotifyHeadersComplete0c chrome_7feed900000!net::URLRequestHttpJob::SaveCookiesAndNotifyHeadersComplete0d chrome_7feed900000!net::URLRequestHttpJob::OnStartCompleted0e chrome_7feed900000!base::Callback<void __cdecl(int)>::Run0f chrome_7feed900000!net::HttpCache::Transaction::DoLoop10 chrome_7feed900000!base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)>::Run11 chrome_7feed900000!base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)> >::MakeItSo12 chrome_7feed900000!base::internal::Invoker<base::IndexSequence<0>,base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64>,void __cdecl(extensions::NativeMessageProcessHost * __ptr64,int),base::WeakPtr<extensions::NativeMessageProcessHost> >,base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64> >,void __cdecl(int const & __ptr64)>::Run13 chrome_7feed900000!base::Callback<void __cdecl(int)>::Run14 chrome_7feed900000!disk_cache::InFlightBackendIO::OnOperationComplete15 chrome_7feed900000!disk_cache::BackgroundIO::OnIOSignalled16 chrome_7feed900000!base::Callback<void __cdecl(void)>::Run17 chrome_7feed900000!base::debug::TaskAnnotator::RunTask18 chrome_7feed900000!base::MessageLoop::RunTask19 chrome_7feed900000!base::MessageLoop::DeferOrRunPendingTask1a chrome_7feed900000!base::MessageLoop::DoWork1b chrome_7feed900000!base::MessagePumpForIO::DoRunLoop1c chrome_7feed900000!base::MessagePumpWin::Run1d chrome_7feed900000!base::MessageLoop::RunHandler1e chrome_7feed900000!base::RunLoop::Run1f chrome_7feed900000!base::MessageLoop::Run20 chrome_7feed900000!base::Thread::Run21 chrome_7feed900000!content::BrowserThreadImpl::IOThreadRun22 chrome_7feed900000!content::BrowserThreadImpl::Run23 chrome_7feed900000!base::Thread::ThreadMain24 chrome_7feed900000!base::`anonymous namespace'::ThreadFunc25 kernel32!BaseThreadInitThunk26 ntdll!RtlUserThreadStart

disk_cache 是HttpCache实际存储在硬盘上的数据组织结构,当完成后通知URLRequestHttpJob启动完成,然后再通过HttpCache进行读取数据。

int HttpCache::Transaction::DoCacheReadData() {  if (request_->method == "HEAD")    return 0;  DCHECK(entry_);  next_state_ = STATE_CACHE_READ_DATA_COMPLETE;  if (net_log_.IsCapturing())    net_log_.BeginEvent(NetLog::TYPE_HTTP_CACHE_READ_DATA);  if (partial_) {    return partial_->CacheRead(entry_->disk_entry, read_buf_.get(), io_buf_len_,                               io_callback_);  }  return entry_->disk_entry->ReadData(kResponseContentIndex, read_offset_,                                      read_buf_.get(), io_buf_len_,                                      io_callback_);}

而数据的读取操作的实际处理也是通过disk_cache进行实际的数据处理,这里HttpCache做中间的一个传输的过程。

HttpCache::Transaction 小结

这里简要介绍了HttpCache的传输过程,仅介绍了其中的一种情况(Cached entry, no validation ),当我们在地址栏中输入字符的时候,会触发一个URL的请求,而这个请求事先也是要通过HttpCache的传输的,请求去查询cache中的项,然后读取出来,在验证的时候跳过了验证机制,之后开始读取实际的数据,完成数据的传输。

HttpNetworkTransaction

上面我们分析了HttpCache的传输方式,如果HttpCache中没有我们要查阅的数据呢,因为我们要访问的是一个我们从没有访问过的URL,又或者我们访问的cache数据已经失效了,那么怎么办呢,此时就要通过网络来加载数据,需要通过http网络传输来完成具体的数据加载。

情景分析

现假设我们访问chromium中FTP网络资源的加载 这篇文章,我的HttpCache中并没有这个URL相关的信息,我们可以在浏览器的地址栏中输入chrome://cache/ 来看HttpCache的存储情况。
我们没有这个缓存,目前我们符合上面的情况一,当我们检索cache后发现没有数据,那么我们要启动网络传输的请求。

0:026> kc # Call Site00 chrome_7feed900000!net::HttpNetworkTransaction::Start01 chrome_7feed900000!DevToolsNetworkTransaction::Start02 chrome_7feed900000!net::HttpCache::Transaction::DoSendRequest03 chrome_7feed900000!net::HttpCache::Transaction::DoLoop04 chrome_7feed900000!base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)>::Run05 chrome_7feed900000!base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)> >::MakeItSo06 chrome_7feed900000!base::internal::Invoker<base::IndexSequence<0>,base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64>,void __cdecl(extensions::NativeMessageProcessHost * __ptr64,int),base::WeakPtr<extensions::NativeMessageProcessHost> >,base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64> >,void __cdecl(int const & __ptr64)>::Run07 chrome_7feed900000!base::Callback<void __cdecl(int)>::Run08 chrome_7feed900000!net::HttpCache::WorkItem::NotifyTransaction09 chrome_7feed900000!net::HttpCache::OnIOComplete0a chrome_7feed900000!net::HttpCache::OnPendingOpComplete0b chrome_7feed900000!base::Callback<void __cdecl(int)>::Run0c chrome_7feed900000!disk_cache::InFlightBackendIO::OnOperationComplete0d chrome_7feed900000!disk_cache::BackgroundIO::OnIOSignalled0e chrome_7feed900000!base::Callback<void __cdecl(void)>::Run0f chrome_7feed900000!base::debug::TaskAnnotator::RunTask10 chrome_7feed900000!base::MessageLoop::RunTask11 chrome_7feed900000!base::MessageLoop::DeferOrRunPendingTask12 chrome_7feed900000!base::MessageLoop::DoWork13 chrome_7feed900000!base::MessagePumpForIO::DoRunLoop14 chrome_7feed900000!base::MessagePumpWin::Run15 chrome_7feed900000!base::MessageLoop::RunHandler16 chrome_7feed900000!base::RunLoop::Run17 chrome_7feed900000!base::MessageLoop::Run18 chrome_7feed900000!base::Thread::Run19 chrome_7feed900000!content::BrowserThreadImpl::IOThreadRun1a chrome_7feed900000!content::BrowserThreadImpl::Run1b chrome_7feed900000!base::Thread::ThreadMain1c chrome_7feed900000!base::`anonymous namespace'::ThreadFunc1d kernel32!BaseThreadInitThunk1e ntdll!RtlUserThreadStart

我们看一下局部变量信息,看看运行时的状态

0:026> ~~[10c4]s;.frame 0n0;dv /t /vchrome_7feed900000!net::HttpNetworkTransaction::Start:000007fe`ee58d66c 48895c2410      mov     qword ptr [rsp+10h],rbx ss:00000000`0914e158=000000000aec661000 00000000`0914e148 000007fe`ee28f812 chrome_7feed900000!net::HttpNetworkTransaction::Start [c:\b\build\slave\win64\build\src\net\http\http_network_transaction.cc @ 191]@rcx              class net::HttpNetworkTransaction * this = 0x00000000`11a21640@rdx              struct net::HttpRequestInfo * request_info = 0x00000000`0aac8ff8@r8               class base::Callback<void __cdecl(int)> * callback = 0x00000000`11d18b80@r9               class net::BoundNetLog * net_log = 0x00000000`11d187a0<unavailable>     int rv = <value unavailable>0:026> dx -id 0,0 -r1 (*((chrome_7feed900000!net::HttpRequestInfo *)0xaac8ff8))(*((chrome_7feed900000!net::HttpRequestInfo *)0xaac8ff8))                 [Type: net::HttpRequestInfo]    [+0x000] url              [Type: GURL]    [+0x078] method           : "GET" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x098] extra_headers    [Type: net::HttpRequestHeaders]    [+0x0b0] upload_data_stream : 0x0 [Type: net::UploadDataStream *]    [+0x0b8] load_flags       : 4352    [+0x0bc] motivation       : NORMAL_MOTIVATION (2) [Type: net::HttpRequestInfo::RequestMotivation]    [+0x0c0] privacy_mode     : PRIVACY_MODE_DISABLED (0) [Type: net::PrivacyMode]0:026> dx -id 0,0 -r1 (*((chrome_7feed900000!GURL *)0xaac8ff8))(*((chrome_7feed900000!GURL *)0xaac8ff8))                 [Type: GURL]    [+0x000] spec_            : "http://blog.csdn.net/feiniao251314/article/details/52230012" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x020] is_valid_        : true    [+0x028] parsed_          [Type: url::Parsed]    [+0x070] inner_url_       [Type: scoped_ptr<GURL,std::default_delete<GURL> >]

在cache中没有找到URL相关的资源,那么我们将请求信息传递到HttpNetworkTransaction中,试图通过网络来请求数据。

源码分析

在源码分析之前,我们看一下HttpNetworkTransaction的状态,根据状态,我们大体上就能了解网络HttpNetworkTransaction要处理的网络情况

HttpNetworkTransaction::State

  enum State {    STATE_NOTIFY_BEFORE_CREATE_STREAM,    STATE_CREATE_STREAM,    STATE_CREATE_STREAM_COMPLETE,    STATE_INIT_STREAM,    STATE_INIT_STREAM_COMPLETE,    STATE_GENERATE_PROXY_AUTH_TOKEN,    STATE_GENERATE_PROXY_AUTH_TOKEN_COMPLETE,    STATE_GENERATE_SERVER_AUTH_TOKEN,    STATE_GENERATE_SERVER_AUTH_TOKEN_COMPLETE,    STATE_GET_TOKEN_BINDING_KEY,    STATE_GET_TOKEN_BINDING_KEY_COMPLETE,    STATE_INIT_REQUEST_BODY,    STATE_INIT_REQUEST_BODY_COMPLETE,    STATE_BUILD_REQUEST,    STATE_BUILD_REQUEST_COMPLETE,    STATE_SEND_REQUEST,    STATE_SEND_REQUEST_COMPLETE,    STATE_READ_HEADERS,    STATE_READ_HEADERS_COMPLETE,    STATE_READ_BODY,    STATE_READ_BODY_COMPLETE,    STATE_DRAIN_BODY_FOR_AUTH_RESTART,    STATE_DRAIN_BODY_FOR_AUTH_RESTART_COMPLETE,    STATE_NONE  };

状态要比HttpCache要少许多,和ftp网络传输状态相当。分创建,初始化,生成key,初始化请求,建立请求,发送请求,读取请求等等。我们从中简要分析其中的过程。

启动与初始化

int HttpNetworkTransaction::Start(const HttpRequestInfo* request_info,                                  const CompletionCallback& callback,                                  const BoundNetLog& net_log) {  net_log_ = net_log;  request_ = request_info;.......  next_state_ = STATE_NOTIFY_BEFORE_CREATE_STREAM;  int rv = DoLoop(OK);  if (rv == ERR_IO_PENDING)    callback_ = callback;  return rv;}int HttpNetworkTransaction::DoNotifyBeforeCreateStream() {  next_state_ = STATE_CREATE_STREAM;  bool defer = false;  if (!before_network_start_callback_.is_null())    before_network_start_callback_.Run(&defer);  if (!defer)    return OK;  return ERR_IO_PENDING;}int HttpNetworkTransaction::DoCreateStream() {  // TODO(mmenke): Remove ScopedTracker below once crbug.com/424359 is fixed.  tracked_objects::ScopedTracker tracking_profile(      FROM_HERE_WITH_EXPLICIT_FUNCTION(          "424359 HttpNetworkTransaction::DoCreateStream"));  response_.network_accessed = true;  next_state_ = STATE_CREATE_STREAM_COMPLETE;  if (ForWebSocketHandshake()) {    stream_request_.reset(        session_->http_stream_factory_for_websocket()            ->RequestWebSocketHandshakeStream(                  *request_,                  priority_,                  server_ssl_config_,                  proxy_ssl_config_,                  this,                  websocket_handshake_stream_base_create_helper_,                  net_log_));  } else {    stream_request_.reset(        session_->http_stream_factory()->RequestStream(            *request_,            priority_,            server_ssl_config_,            proxy_ssl_config_,            this,            net_log_));  }  DCHECK(stream_request_.get());  return ERR_IO_PENDING;}

创建比较简单,使用会话session_来创建一个请求Stream,内部创建了socket,我们暂且不向下深挖,暂时先分析http这一层次的传输过程,下面给出创建socket的堆栈示意:

0:026> kc # Call Site00 chrome_7feed900000!net::CreatePlatformSocket01 chrome_7feed900000!net::UDPSocketWin::Open02 chrome_7feed900000!net::UDPClientSocket::Connect03 chrome_7feed900000!net::`anonymous namespace'::IsGloballyReachable04 chrome_7feed900000!net::HostResolverImpl::IsIPv6Reachable05 chrome_7feed900000!net::HostResolverImpl::GetEffectiveKeyForRequest06 chrome_7feed900000!net::HostResolverImpl::Resolve07 chrome_7feed900000!net::SingleRequestHostResolver::Resolve08 chrome_7feed900000!net::TransportConnectJobHelper::DoResolveHost09 chrome_7feed900000!net::TransportConnectJob::DoResolveHost0a chrome_7feed900000!net::TransportConnectJobHelper::DoLoop<net::TransportConnectJob>0b chrome_7feed900000!net::ConnectJob::Connect0c chrome_7feed900000!net::internal::ClientSocketPoolBaseHelper::RequestSocketInternal0d chrome_7feed900000!net::internal::ClientSocketPoolBaseHelper::RequestSocket0e chrome_7feed900000!net::ClientSocketPoolBase<net::TransportSocketParams>::RequestSocket0f chrome_7feed900000!net::ClientSocketHandle::Init<net::SSLClientSocketPool>10 chrome_7feed900000!net::`anonymous namespace'::InitSocketPoolHelper11 chrome_7feed900000!net::InitSocketHandleForHttpRequest12 chrome_7feed900000!net::HttpStreamFactoryImpl::Job::DoInitConnection13 chrome_7feed900000!net::HttpStreamFactoryImpl::Job::DoLoop14 chrome_7feed900000!net::HttpStreamFactoryImpl::Job::RunLoop15 chrome_7feed900000!net::HttpStreamFactoryImpl::RequestStreamInternal16 chrome_7feed900000!net::HttpStreamFactoryImpl::RequestStream17 chrome_7feed900000!net::HttpNetworkTransaction::DoCreateStream18 chrome_7feed900000!net::HttpNetworkTransaction::DoLoop19 chrome_7feed900000!net::HttpNetworkTransaction::Start1a chrome_7feed900000!DevToolsNetworkTransaction::Start1b chrome_7feed900000!net::HttpCache::Transaction::DoSendRequest1c chrome_7feed900000!net::HttpCache::Transaction::DoLoop1d chrome_7feed900000!base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)>::Run1e chrome_7feed900000!base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)> >::MakeItSo1f chrome_7feed900000!base::internal::Invoker<base::IndexSequence<0>,base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64>,void __cdecl(extensions::NativeMessageProcessHost * __ptr64,int),base::WeakPtr<extensions::NativeMessageProcessHost> >,base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64> >,void __cdecl(int const & __ptr64)>::Run20 chrome_7feed900000!base::Callback<void __cdecl(int)>::Run21 chrome_7feed900000!net::HttpCache::WorkItem::NotifyTransaction22 chrome_7feed900000!net::HttpCache::OnIOComplete23 chrome_7feed900000!net::HttpCache::OnPendingOpComplete24 chrome_7feed900000!base::Callback<void __cdecl(int)>::Run25 chrome_7feed900000!disk_cache::InFlightBackendIO::OnOperationComplete26 chrome_7feed900000!disk_cache::BackgroundIO::OnIOSignalled27 chrome_7feed900000!base::Callback<void __cdecl(void)>::Run28 chrome_7feed900000!base::debug::TaskAnnotator::RunTask29 chrome_7feed900000!base::MessageLoop::RunTask2a chrome_7feed900000!base::MessageLoop::DeferOrRunPendingTask2b chrome_7feed900000!base::MessageLoop::DoWork2c chrome_7feed900000!base::MessagePumpForIO::DoRunLoop2d chrome_7feed900000!base::MessagePumpWin::Run2e chrome_7feed900000!base::MessageLoop::RunHandler2f chrome_7feed900000!base::RunLoop::Run30 chrome_7feed900000!base::MessageLoop::Run31 chrome_7feed900000!base::Thread::Run32 chrome_7feed900000!content::BrowserThreadImpl::IOThreadRun33 chrome_7feed900000!content::BrowserThreadImpl::Run34 chrome_7feed900000!base::Thread::ThreadMain35 chrome_7feed900000!base::`anonymous namespace'::ThreadFunc36 kernel32!BaseThreadInitThunk37 ntdll!RtlUserThreadStart
int HttpNetworkTransaction::DoInitStream() {  DCHECK(stream_.get());  next_state_ = STATE_INIT_STREAM_COMPLETE;  stream_->GetRemoteEndpoint(&remote_endpoint_);  return stream_->InitializeStream(request_, priority_, net_log_, io_callback_);}

创建请求完成后,初始化Stream,然后生成服务器和客户端的Token.

发送请求

int HttpNetworkTransaction::DoSendRequest() {  // TODO(mmenke): Remove ScopedTracker below once crbug.com/424359 is fixed.  tracked_objects::ScopedTracker tracking_profile(      FROM_HERE_WITH_EXPLICIT_FUNCTION(          "424359 HttpNetworkTransaction::DoSendRequest"));  send_start_time_ = base::TimeTicks::Now();  next_state_ = STATE_SEND_REQUEST_COMPLETE;  return stream_->SendRequest(request_headers_, &response_, io_callback_);}

这里我们可以查看一下request_headers_的相关信息,目前是connection请求。

0:025> dx -id 0,0 -r1 (*((chrome_7fee4550000!net::HttpRequestHeaders::HeaderKeyValuePair *)0x114f49e0))(*((chrome_7fee4550000!net::HttpRequestHeaders::HeaderKeyValuePair *)0x114f49e0))                 [Type: net::HttpRequestHeaders::HeaderKeyValuePair]    [+0x000] key              : "Host" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x020] value            : "blog.csdn.net" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]0:025> dx -id 0,0 -r1 (*((chrome_7fee4550000!net::HttpRequestHeaders::HeaderKeyValuePair *)0x114f4a20))(*((chrome_7fee4550000!net::HttpRequestHeaders::HeaderKeyValuePair *)0x114f4a20))                 [Type: net::HttpRequestHeaders::HeaderKeyValuePair]    [+0x000] key              : "Connection" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x020] value            : "keep-alive" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]0:025> dx -id 0,0 -r1 (*((chrome_7fee4550000!net::HttpRequestHeaders::HeaderKeyValuePair *)0x114f4a60))(*((chrome_7fee4550000!net::HttpRequestHeaders::HeaderKeyValuePair *)0x114f4a60))                 [Type: net::HttpRequestHeaders::HeaderKeyValuePair]    [+0x000] key              : "Accept" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]    [+0x020] value            : "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8" [Type: std::basic_string<char,std::char_traits<char>,std::allocator<char> >]

接收数据

int HttpNetworkTransaction::DoSendRequestComplete(int result) {  send_end_time_ = base::TimeTicks::Now();  if (result < 0)    return HandleIOError(result);  next_state_ = STATE_READ_HEADERS;  return OK;}int HttpNetworkTransaction::DoReadHeaders() {  next_state_ = STATE_READ_HEADERS_COMPLETE;  return stream_->ReadResponseHeaders(io_callback_);}

当发送请求完成后,会设置下一跳状态为读取头部状态,然后返回状态循环,开始读取头部信息。

HttpNetworkTransaction 小结

当在HttpCache中无法获取到有效信息的时候,那么就会生成一个HttpNetworkTransaction请求,通过网络来请求最新数据,HttpNetworkTransaction负责建立连接,初始化终端,发送和接收数据,将数据组织好。

网络请求结果的转储

上面我们说了HttpCache::transaction和HttpNetworkTransaction的情况,一个是本地缓存传输,一个是网络数据传输,当本地数据无效的时候,会请求网络传输,而当网络传输完毕的时候会通知本地缓存处理。

情景分析

我们可以如下下断

0:025> bp chrome_7fee4550000!net::HttpCache::Transaction::DoSuccessfulSendRequest

这样,当HttpNetworkTransaction传输完成的时候,会中断到调试器,这样跟踪一下,看看实际的运行过程。
下面是堆栈示意。

0:025> kc # Call Site00 chrome_7fee4550000!net::HttpCache::Transaction::DoSuccessfulSendRequest01 chrome_7fee4550000!net::HttpCache::Transaction::DoLoop02 chrome_7fee4550000!base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)>::Run03 chrome_7fee4550000!base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int)> >::MakeItSo04 chrome_7fee4550000!base::internal::Invoker<base::IndexSequence<0>,base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64>,void __cdecl(extensions::NativeMessageProcessHost * __ptr64,int),base::WeakPtr<extensions::NativeMessageProcessHost> >,base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl extensions::NativeMessageProcessHost::*)(int) __ptr64> >,void __cdecl(int const & __ptr64)>::Run05 chrome_7fee4550000!base::Callback<void __cdecl(int)>::Run06 chrome_7fee4550000!net::HttpNetworkTransaction::DoCallback07 chrome_7fee4550000!net::HttpNetworkTransaction::OnIOComplete08 chrome_7fee4550000!net::HttpNetworkTransaction::OnStreamReady09 chrome_7fee4550000!net::HttpStreamFactoryImpl::Job::OnStreamReadyCallback0a chrome_7fee4550000!base::internal::RunnableAdapter<void (__cdecl browser_sync::BookmarkModelAssociator::*)(void)>::Run0b chrome_7fee4550000!base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl browser_sync::BookmarkModelAssociator::*)(void)> >::MakeItSo0c chrome_7fee4550000!base::internal::Invoker<base::IndexSequence<0>,base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl browser_sync::BookmarkModelAssociator::*)(void) __ptr64>,void __cdecl(browser_sync::BookmarkModelAssociator * __ptr64),base::WeakPtr<browser_sync::BookmarkModelAssociator> >,base::internal::InvokeHelper<1,void,base::internal::RunnableAdapter<void (__cdecl browser_sync::BookmarkModelAssociator::*)(void) __ptr64> >,void __cdecl(void)>::Run0d chrome_7fee4550000!base::Callback<void __cdecl(void)>::Run0e chrome_7fee4550000!base::debug::TaskAnnotator::RunTask0f chrome_7fee4550000!base::MessageLoop::RunTask10 chrome_7fee4550000!base::MessageLoop::DeferOrRunPendingTask11 chrome_7fee4550000!base::MessageLoop::DoWork12 chrome_7fee4550000!base::MessagePumpForIO::DoRunLoop13 chrome_7fee4550000!base::MessagePumpWin::Run14 chrome_7fee4550000!base::MessageLoop::RunHandler15 chrome_7fee4550000!base::RunLoop::Run16 chrome_7fee4550000!base::MessageLoop::Run17 chrome_7fee4550000!base::Thread::Run18 chrome_7fee4550000!content::BrowserThreadImpl::IOThreadRun19 chrome_7fee4550000!content::BrowserThreadImpl::Run1a chrome_7fee4550000!base::Thread::ThreadMain1b chrome_7fee4550000!base::`anonymous namespace'::ThreadFunc1c kernel32!BaseThreadInitThunk1d ntdll!RtlUserThreadStart

在HttpNetworkTransaction完成的时候会调用callback,这个是在启动HttpNetworkTransaction请求的时候设置的,完成时回调到HttpCache::transaction,进行数据的转储处理。

源码分析

调用过程

// 1. Not-cached entry://   Start()://   GetBackend* -> InitEntry -> OpenEntry* -> CreateEntry* -> AddToEntry* ->//   SendRequest* -> SuccessfulSendRequest -> OverwriteCachedResponse ->//   CacheWriteResponse* -> TruncateCachedData* -> TruncateCachedMetadata* ->//   PartialHeadersReceived////   Read()://   NetworkRead* -> CacheWriteData*

我们回顾一下情况一,如上,我们收到了HttpNetworkTransaction请求完成后,会重写缓存,但是我们的缓存中并没有这一项,就没有实际的重写操作,而是进行下一状态,STATE_CACHE_WRITE_RESPONSE.

int HttpCache::Transaction::DoCacheWriteResponse() {  next_state_ = STATE_CACHE_WRITE_RESPONSE_COMPLETE;  return WriteResponseInfoToEntry(truncated_);}

这一状态的处理就是实际的写响应信息到HttpCache的Entry中。

int HttpCache::Transaction::WriteResponseInfoToEntry(bool truncated) {  if (!entry_)    return OK;......  bool skip_transient_headers = true;  scoped_refptr<PickledIOBuffer> data(new PickledIOBuffer());  response_.Persist(data->pickle(), skip_transient_headers, truncated);  data->Done();  io_buf_len_ = data->pickle()->size();  return entry_->disk_entry->WriteData(kResponseInfoIndex, 0, data.get(),                                       io_buf_len_, io_callback_, true);}

在这个函数中实际的写数据操作通过Entry来实现,将数据写到disk_cache中。

转储小结

就是这样,当HttpCache中没有有效数据的时候,就通过HttpNetworkTransaction请求网络数据,当网络数据请求完毕时,回调到HttpCache::transaction中将数据存储到HttpCache中,以便下次读取。

样本的版本信息

chrome 50.0.2661.87

符号路径:
https://chromium-browser-symsrv.commondatastorage.googleapis.com

源码网站:
www.chromium.org

感谢chromium项目相关作者及人员。

1 0
原创粉丝点击