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_resourceProvider(config->resourceProvider.release())
+    , m_receivingDataOnlyFromResourceProvider(false)
 {
+    // Tell the resource provider that we're ready to start receiving
+    // chunks of data directly from the parser thread.
+    if (m_resourceProvider)
+        m_resourceProvider->setBackgroundClient(this);
 }
 
 BackgroundHTMLParser::~BackgroundHTMLParser()
 {
+    if (m_resourceProvider) {
+        // Tell the resource provider that calling into this instance
+        // is no longer safe and any further data received should be
+        // ignored.
+        m_resourceProvider->setBackgroundClient(0);
+        // The resource provider needs to be destructed on the main thread,
+        // as the filter it installs on the I/O thread can only be removed

[abarth-chromium, 2014/02/03 21:17:36]

Blink doesn't know anything about I/O threads.

[oystein (OOO til 10th of July), 2014/02/05 00:05:00]

Removed that part of the comment; it was just a clarification, not relevant to
Blink.

+        // 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::destroyResourceProvider, m_resourceProvider.release()));
+    }
+}
+
+void BackgroundHTMLParser::didReceivedData(const char* data, size_t length)
+{
+    if (m_receivingDataOnlyFromResourceProvider && m_decoder) {
+        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. This is also necessary if we've lost our
+    // decoder, until we get a new one passed from the main thread.
+
+    OwnPtr<Vector<char> > buffer = adoptPtr(new Vector<char>(length));
+    memcpy(buffer->data(), data, length);

[abarth-chromium, 2014/02/03 21:17:36]

buffer.append(data, length)

[oystein (OOO til 10th of July), 2014/02/05 00:05:00]

Done.

+    m_queuedData.append(buffer.release());
 }
 
 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_receivingDataOnlyFromResourceProvider);
     updateDocument(m_decoder->decode(buffer->data(), buffer->size()));
 }
 
 void BackgroundHTMLParser::setDecoder(PassOwnPtr<TextResourceDecoder> decoder)
 {
     m_decoder = decoder;
+
+    // If our decoder got reset in mid-stream and got recreated, we may have
+    // some pending data here.
+    if (m_decoder)
+        flushQueuedData();
 }
 
 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::flushQueuedData()
+{
+    // We only process queued chunks if we have a decoder and if we know
+    // we won't be receiving any more chunks for the main thread. Otherwise
+    // we hang on to the chunks for a bit longer.
+    if (!m_decoder || !m_receivingDataOnlyFromResourceProvider)
+        return;
+
+    for (Vector<OwnPtr<Vector<char> > >::iterator entry = m_queuedData.begin(); entry != m_queuedData.end(); ++entry)
+        updateDocument(m_decoder->decode((*entry)->data(), (*entry)->size()));

[abarth-chromium, 2014/02/03 21:17:36]

SharedBuffer has a better way of doing this sort of thing.

[oystein (OOO til 10th of July), 2014/02/05 00:05:00]

Done.

+
+    m_queuedData.clear();
+}
+
+void BackgroundHTMLParser::resourceFilterAdded()
+{
+    m_receivingDataOnlyFromResourceProvider = true;
+
+    // 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.
+    flushQueuedData();
+}
+
 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);
 }
 
 }