Redirect HTML resource bytes directly to parser thread (Blink side CL)

Source/core/html/parser/BackgroundHTMLParser.cpp

 /*
  * Copyright (C) 2013 Google, Inc. All Rights Reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
  * 1. Redistributions of source code must retain the above copyright
  *    notice, this list of conditions and the following disclaimer.
  * 2. Redistributions in binary form must reproduce the above copyright
  *    notice, this list of conditions and the following disclaimer in the
  *    documentation and/or other materials provided with the distribution.
  *
  * THIS SOFTWARE IS PROVIDED BY GOOGLE INC. ``AS IS'' AND ANY
  * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL GOOGLE INC. OR
  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
  * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
 #include "config.h"
 #include "core/html/parser/BackgroundHTMLParser.h"
 
 #include "core/html/parser/HTMLDocumentParser.h"
-#include "core/html/parser/HTMLParserThread.h"
 #include "core/html/parser/TextResourceDecoder.h"
 #include "core/html/parser/XSSAuditor.h"
 #include "wtf/MainThread.h"
 #include "wtf/text/TextPosition.h"
 
 namespace WebCore {
 
 // On a network with high latency and high bandwidth, using a device
 // with a fast CPU, we could end up speculatively tokenizing
 // the whole document, well ahead of when the main-thread actually needs it.
@@ skipping 46 matching lines @@
     : m_weakFactory(reference, this)
     , m_token(adoptPtr(new HTMLToken))
     , m_tokenizer(HTMLTokenizer::create(config->options))
     , m_treeBuilderSimulator(config->options)
     , m_options(config->options)
     , m_parser(config->parser)
     , m_pendingTokens(adoptPtr(new CompactHTMLTokenStream))
     , m_xssAuditor(config->xssAuditor.release())
     , m_preloadScanner(config->preloadScanner.release())
     , m_decoder(config->decoder.release())
+    , m_resourceBridge(config->resourceBridge.release())
+    , m_parserThreadIsStandalone(false)
 {
+    // Tell the resource bridge that we're ready to start receiving
+    // chunks of data.
+    if (m_resourceBridge)
+        m_resourceBridge->setPeer(this);
 }
 
 BackgroundHTMLParser::~BackgroundHTMLParser()
 {
+    if (m_resourceBridge) {
+        // Tell the resource bridge that calling into this instance
+        // is no longer safe and any further data received should be
+        // ignored.
+        m_resourceBridge->setPeer(0);
+        // The resource bridge needs to be destructed on the main thread,
+        // as the filter it installs on the I/O thread can only be removed
+        // from the main thread...
+        // Note we can only call static functions on the HTMLDocumentParser at this point
+        // as the weakptr we have for it is likely invalidated now.
+        callOnMainThread(bind(&HTMLDocumentParser::destructResourceBridge, m_resourceBridge.release()));
+    }
+}
+
+void BackgroundHTMLParser::OnReceivedData(const char* data, size_t length)
+{
+    if (m_parserThreadIsStandalone) {
+        updateDocument(m_decoder->decode(data, length));
+        return;
+    }
+
+    // If the parser thread is not standalone yet, it means we may get
+    // further data packets from the main thread and will need to queue
+    // up any data coming directly from the parser thread until we're
+    // sure we're complete.
+
+    // Storing the raw pointer here is safe as we keep the shared
+    // memory alive for as long as this object is alive (via its
+    // attached WebParserResourceBridge)

[abarth-chromium, 2013/12/18 18:28:49]

That sounds very fragile.  How about we just make a copy of the data that this
object can own instead of having this complex lifetime coupling.

[oystein (OOO til 10th of July), 2014/01/13 23:19:50]

The lifetime coupling is still needed to be able to safely make the calls
between the Chrome and Blink parts of this, so making a copy here wouldn't
simplify any code elsewhere as far as I can tell.

+    m_queuedData.push_back(std::make_pair(data, length));
 }
 
 void BackgroundHTMLParser::append(const String& input)
 {
     ASSERT(!m_input.current().isClosed());
     m_input.append(input);
     pumpTokenizer();
 }
 
 void BackgroundHTMLParser::appendBytes(PassOwnPtr<Vector<char> > buffer)
 {
+    ASSERT(!m_parserThreadIsStandalone);
     updateDocument(m_decoder->decode(buffer->data(), buffer->size()));
 }
 
 void BackgroundHTMLParser::setDecoder(PassOwnPtr<TextResourceDecoder> decoder)
 {
     m_decoder = decoder;
 }
 
 void BackgroundHTMLParser::flush()
 {
     updateDocument(m_decoder->flush());
 }
 
 void BackgroundHTMLParser::updateDocument(const String& decodedData)
 {
     DocumentEncodingData encodingData(*m_decoder.get());
 
     if (encodingData != m_lastSeenEncodingData) {
         m_lastSeenEncodingData = encodingData;
-
         m_xssAuditor->setEncoding(encodingData.encoding());
         callOnMainThread(bind(&HTMLDocumentParser::didReceiveEncodingDataFromBackgroundParser, m_parser, encodingData));
     }
 
     if (decodedData.isEmpty())
         return;
 
     append(decodedData);
 }
 
@@ skipping 20 matching lines @@
 {
     markEndOfFile();
     pumpTokenizer();
 }
 
 void BackgroundHTMLParser::stop()
 {
     delete this;
 }
 
+void BackgroundHTMLParser::resourceFilterAdded()
+{
+    m_parserThreadIsStandalone = true;

[abarth-chromium, 2013/12/18 18:28:49]

I'm not sure what this state variable means.  It looks like a switch between
receiving data from the main thread and from receiving data directly from the
network stack.

[oystein (OOO til 10th of July), 2014/01/13 23:19:50]

Renamed to m_receivingDataOnlyFromResourceProvider. A bit verbose though, but
best I could come up with.

+
+    // At this point we know for sure that no further data will
+    // be coming from the main thread, so we can process anything
+    // we've received directly on the parser thread in the meantime.
+    for (DataQueue::iterator entry = m_queuedData.begin(); entry != m_queuedData.end(); ++entry)
+        updateDocument(m_decoder->decode(entry->first, entry->second));

[abarth-chromium, 2013/12/18 18:28:49]

Do we need to worry about one of these updateDocument calls stopping parsing?

[oystein (OOO til 10th of July), 2014/01/13 23:19:50]

Shouldn't matter. As soon as a parsed chunk is processed on the main thread and
signals for the parsing to stop, it'll invalidate the weakptrs that the
BackgroundHTMLParser holds, so any further updates will be ignored.

+
+    // After this point we'll never need m_queuedData again, so let's release its
+    // memory.
+    DataQueue emptyQueue;
+    std::swap(emptyQueue, m_queuedData);

[abarth-chromium, 2013/12/18 18:28:49]

These idioms will be different once you switch to using WTF collections.

+}
+
 void BackgroundHTMLParser::forcePlaintextForTextDocument()
 {
     // This is only used by the TextDocumentParser (a subclass of HTMLDocumentParser)
     // to force us into the PLAINTEXT state w/o using a <plaintext> tag.
     // The TextDocumentParser uses a <pre> tag for historical/compatibility reasons.
     m_tokenizer->setState(HTMLTokenizer::PLAINTEXTState);
 }
 
 void BackgroundHTMLParser::markEndOfFile()
 {
@@ skipping 61 matching lines @@
     chunk->treeBuilderState = m_treeBuilderSimulator.state();
     chunk->inputCheckpoint = m_input.createCheckpoint(m_pendingTokens->size());
     chunk->preloadScannerCheckpoint = m_preloadScanner->createCheckpoint();
     chunk->tokens = m_pendingTokens.release();
     callOnMainThread(bind(&HTMLDocumentParser::didReceiveParsedChunkFromBackgroundParser, m_parser, chunk.release()));
 
     m_pendingTokens = adoptPtr(new CompactHTMLTokenStream);
 }
 
 }