diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 09:22:09 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-07 09:22:09 +0000 |
commit | 43a97878ce14b72f0981164f87f2e35e14151312 (patch) | |
tree | 620249daf56c0258faa40cbdcf9cfba06de2a846 /netwerk/streamconv/converters/nsUnknownDecoder.cpp | |
parent | Initial commit. (diff) | |
download | firefox-43a97878ce14b72f0981164f87f2e35e14151312.tar.xz firefox-43a97878ce14b72f0981164f87f2e35e14151312.zip |
Adding upstream version 110.0.1.upstream/110.0.1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rw-r--r-- | netwerk/streamconv/converters/nsUnknownDecoder.cpp | 852 |
1 files changed, 852 insertions, 0 deletions
diff --git a/netwerk/streamconv/converters/nsUnknownDecoder.cpp b/netwerk/streamconv/converters/nsUnknownDecoder.cpp new file mode 100644 index 0000000000..a9d919456a --- /dev/null +++ b/netwerk/streamconv/converters/nsUnknownDecoder.cpp @@ -0,0 +1,852 @@ +/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#include "nsUnknownDecoder.h" +#include "nsIPipe.h" +#include "nsIInputStream.h" +#include "nsIOutputStream.h" +#include "nsMimeTypes.h" + +#include "nsCRT.h" + +#include "nsIMIMEService.h" + +#include "nsIViewSourceChannel.h" +#include "nsIHttpChannel.h" +#include "nsIForcePendingChannel.h" +#include "nsIEncodedChannel.h" +#include "nsIURI.h" +#include "nsStringStream.h" +#include "nsNetCID.h" +#include "nsNetUtil.h" +#include "nsQueryObject.h" +#include "nsComponentManagerUtils.h" +#include "nsServiceManagerUtils.h" + +#include <algorithm> + +#define MAX_BUFFER_SIZE 512u + +using namespace mozilla; + +NS_IMPL_ISUPPORTS(nsUnknownDecoder::ConvertedStreamListener, nsIStreamListener, + nsIRequestObserver) + +nsUnknownDecoder::ConvertedStreamListener::ConvertedStreamListener( + nsUnknownDecoder* aDecoder) { + mDecoder = aDecoder; +} + +nsresult nsUnknownDecoder::ConvertedStreamListener::AppendDataToString( + nsIInputStream* inputStream, void* closure, const char* rawSegment, + uint32_t toOffset, uint32_t count, uint32_t* writeCount) { + nsCString* decodedData = static_cast<nsCString*>(closure); + decodedData->Append(rawSegment, count); + *writeCount = count; + return NS_OK; +} + +NS_IMETHODIMP +nsUnknownDecoder::ConvertedStreamListener::OnStartRequest(nsIRequest* request) { + return NS_OK; +} + +NS_IMETHODIMP +nsUnknownDecoder::ConvertedStreamListener::OnDataAvailable( + nsIRequest* request, nsIInputStream* stream, uint64_t offset, + uint32_t count) { + uint32_t read; + nsAutoCString decodedData; + { + MutexAutoLock lock(mDecoder->mMutex); + decodedData = mDecoder->mDecodedData; + } + nsresult rv = + stream->ReadSegments(AppendDataToString, &decodedData, count, &read); + if (NS_FAILED(rv)) { + return rv; + } + MutexAutoLock lock(mDecoder->mMutex); + mDecoder->mDecodedData = decodedData; + return NS_OK; +} + +NS_IMETHODIMP +nsUnknownDecoder::ConvertedStreamListener::OnStopRequest(nsIRequest* request, + nsresult status) { + return NS_OK; +} + +nsUnknownDecoder::nsUnknownDecoder(nsIStreamListener* aListener) + : mNextListener(aListener), + mBuffer(nullptr), + mBufferLen(0), + mMutex("nsUnknownDecoder"), + mDecodedData("") {} + +nsUnknownDecoder::~nsUnknownDecoder() { + if (mBuffer) { + delete[] mBuffer; + mBuffer = nullptr; + } +} + +// ---- +// +// nsISupports implementation... +// +// ---- + +NS_IMPL_ADDREF(nsUnknownDecoder) +NS_IMPL_RELEASE(nsUnknownDecoder) + +NS_INTERFACE_MAP_BEGIN(nsUnknownDecoder) + NS_INTERFACE_MAP_ENTRY(nsIStreamConverter) + NS_INTERFACE_MAP_ENTRY(nsIStreamListener) + NS_INTERFACE_MAP_ENTRY(nsIRequestObserver) + NS_INTERFACE_MAP_ENTRY(nsIContentSniffer) + NS_INTERFACE_MAP_ENTRY(nsIThreadRetargetableStreamListener) + NS_INTERFACE_MAP_ENTRY_AMBIGUOUS(nsISupports, nsIStreamListener) +NS_INTERFACE_MAP_END + +// ---- +// +// nsIStreamConverter methods... +// +// ---- + +NS_IMETHODIMP +nsUnknownDecoder::Convert(nsIInputStream* aFromStream, const char* aFromType, + const char* aToType, nsISupports* aCtxt, + nsIInputStream** aResultStream) { + return NS_ERROR_NOT_IMPLEMENTED; +} + +NS_IMETHODIMP +nsUnknownDecoder::AsyncConvertData(const char* aFromType, const char* aToType, + nsIStreamListener* aListener, + nsISupports* aCtxt) { + NS_ASSERTION(aListener && aFromType && aToType, + "null pointer passed into multi mixed converter"); + // hook up our final listener. this guy gets the various On*() calls we want + // to throw at him. + // + + MutexAutoLock lock(mMutex); + mNextListener = aListener; + return (aListener) ? NS_OK : NS_ERROR_FAILURE; +} + +NS_IMETHODIMP +nsUnknownDecoder::GetConvertedType(const nsACString& aFromType, + nsIChannel* aChannel, nsACString& aToType) { + return NS_ERROR_NOT_IMPLEMENTED; +} + +// ---- +// +// nsIStreamListener methods... +// +// ---- + +NS_IMETHODIMP +nsUnknownDecoder::OnDataAvailable(nsIRequest* request, nsIInputStream* aStream, + uint64_t aSourceOffset, uint32_t aCount) { + nsresult rv = NS_OK; + + bool contentTypeEmpty; + { + MutexAutoLock lock(mMutex); + if (!mNextListener) return NS_ERROR_FAILURE; + + contentTypeEmpty = mContentType.IsEmpty(); + } + + if (contentTypeEmpty) { + uint32_t count, len; + + // If the buffer has not been allocated by now, just fail... + if (!mBuffer) return NS_ERROR_OUT_OF_MEMORY; + + // + // Determine how much of the stream should be read to fill up the + // sniffer buffer... + // + if (mBufferLen + aCount >= MAX_BUFFER_SIZE) { + count = MAX_BUFFER_SIZE - mBufferLen; + } else { + count = aCount; + } + + // Read the data into the buffer... + rv = aStream->Read((mBuffer + mBufferLen), count, &len); + if (NS_FAILED(rv)) return rv; + + mBufferLen += len; + aCount -= len; + + if (aCount) { + // + // Adjust the source offset... The call to FireListenerNotifications(...) + // will make the first OnDataAvailable(...) call with an offset of 0. + // So, this offset needs to be adjusted to reflect that... + // + aSourceOffset += mBufferLen; + + DetermineContentType(request); + + rv = FireListenerNotifications(request, nullptr); + } + } + + // Must not fire ODA again if it failed once + if (aCount && NS_SUCCEEDED(rv)) { +#ifdef DEBUG + { + MutexAutoLock lock(mMutex); + NS_ASSERTION(!mContentType.IsEmpty(), + "Content type should be known by now."); + } +#endif + + nsCOMPtr<nsIStreamListener> listener; + { + MutexAutoLock lock(mMutex); + listener = mNextListener; + } + rv = listener->OnDataAvailable(request, aStream, aSourceOffset, aCount); + } + + return rv; +} + +// ---- +// +// nsIRequestObserver methods... +// +// ---- + +NS_IMETHODIMP +nsUnknownDecoder::OnStartRequest(nsIRequest* request) { + nsresult rv = NS_OK; + + { + MutexAutoLock lock(mMutex); + if (!mNextListener) return NS_ERROR_FAILURE; + } + + // Allocate the sniffer buffer... + if (NS_SUCCEEDED(rv) && !mBuffer) { + mBuffer = new char[MAX_BUFFER_SIZE]; + + if (!mBuffer) { + rv = NS_ERROR_OUT_OF_MEMORY; + } + } + + // Do not pass the OnStartRequest on to the next listener (yet)... + return rv; +} + +NS_IMETHODIMP +nsUnknownDecoder::OnStopRequest(nsIRequest* request, nsresult aStatus) { + nsresult rv = NS_OK; + + bool contentTypeEmpty; + { + MutexAutoLock lock(mMutex); + if (!mNextListener) return NS_ERROR_FAILURE; + + contentTypeEmpty = mContentType.IsEmpty(); + } + + // + // The total amount of data is less than the size of the sniffer buffer. + // Analyze the buffer now... + // + if (contentTypeEmpty) { + DetermineContentType(request); + + // Make sure channel listeners see channel as pending while we call + // OnStartRequest/OnDataAvailable, even though the underlying channel + // has already hit OnStopRequest. + nsCOMPtr<nsIForcePendingChannel> forcePendingChannel = + do_QueryInterface(request); + if (forcePendingChannel) { + forcePendingChannel->ForcePending(true); + } + + rv = FireListenerNotifications(request, nullptr); + + if (NS_FAILED(rv)) { + aStatus = rv; + } + + // now we need to set pending state to false before calling OnStopRequest + if (forcePendingChannel) { + forcePendingChannel->ForcePending(false); + } + } + + nsCOMPtr<nsIStreamListener> listener; + { + MutexAutoLock lock(mMutex); + listener = mNextListener; + mNextListener = nullptr; + } + rv = listener->OnStopRequest(request, aStatus); + + return rv; +} + +// ---- +// +// nsIContentSniffer methods... +// +// ---- +NS_IMETHODIMP +nsUnknownDecoder::GetMIMETypeFromContent(nsIRequest* aRequest, + const uint8_t* aData, uint32_t aLength, + nsACString& type) { + // This is only used by sniffer, therefore we do not need to lock anything + // here. + nsCOMPtr<nsIChannel> channel(do_QueryInterface(aRequest)); + if (channel) { + nsCOMPtr<nsILoadInfo> loadInfo = channel->LoadInfo(); + if (loadInfo->GetSkipContentSniffing()) { + return NS_ERROR_NOT_AVAILABLE; + } + } + + mBuffer = const_cast<char*>(reinterpret_cast<const char*>(aData)); + mBufferLen = aLength; + DetermineContentType(aRequest); + mBuffer = nullptr; + mBufferLen = 0; + type.Assign(mContentType); + mContentType.Truncate(); + return type.IsEmpty() ? NS_ERROR_NOT_AVAILABLE : NS_OK; +} + +// Actual sniffing code + +/** + * This is the array of sniffer entries that depend on "magic numbers" + * in the file. Each entry has either a type associated with it (set + * these with the SNIFFER_ENTRY macro) or a function to be executed + * (set these with the SNIFFER_ENTRY_WITH_FUNC macro). The function + * should take a single nsIRequest* and returns bool -- true if + * it sets mContentType, false otherwise + */ +nsUnknownDecoder::nsSnifferEntry nsUnknownDecoder::sSnifferEntries[] = { + SNIFFER_ENTRY("%PDF-", APPLICATION_PDF), + + SNIFFER_ENTRY("%!PS-Adobe-", APPLICATION_POSTSCRIPT), + + // Files that start with mailbox delimiters let's provisionally call + // text/plain + SNIFFER_ENTRY("From", TEXT_PLAIN), SNIFFER_ENTRY(">From", TEXT_PLAIN), + + // If the buffer begins with "#!" or "%!" then it is a script of + // some sort... "Scripts" can include arbitrary data to be passed + // to an interpreter, so we need to decide whether we can call this + // text or whether it's data. + SNIFFER_ENTRY_WITH_FUNC("#!", &nsUnknownDecoder::LastDitchSniff), + + // XXXbz should (and can) we also include the various ways that <?xml can + // appear as UTF-16 and such? See http://www.w3.org/TR/REC-xml#sec-guessing + SNIFFER_ENTRY_WITH_FUNC("<?xml", &nsUnknownDecoder::SniffForXML)}; + +uint32_t nsUnknownDecoder::sSnifferEntryNum = + sizeof(nsUnknownDecoder::sSnifferEntries) / + sizeof(nsUnknownDecoder::nsSnifferEntry); + +void nsUnknownDecoder::DetermineContentType(nsIRequest* aRequest) { + { + MutexAutoLock lock(mMutex); + NS_ASSERTION(mContentType.IsEmpty(), "Content type is already known."); + if (!mContentType.IsEmpty()) return; + } + + nsCOMPtr<nsIChannel> channel(do_QueryInterface(aRequest)); + if (channel) { + nsCOMPtr<nsILoadInfo> loadInfo = channel->LoadInfo(); + if (loadInfo->GetSkipContentSniffing()) { + /* + * If we did not get a useful Content-Type from the server + * but also have sniffing disabled, just determine whether + * to use text/plain or octetstream and log an error to the Console + */ + LastDitchSniff(aRequest); + + nsCOMPtr<nsIHttpChannel> httpChannel(do_QueryInterface(aRequest)); + if (httpChannel) { + nsAutoCString type; + httpChannel->GetContentType(type); + nsCOMPtr<nsIURI> requestUri; + httpChannel->GetURI(getter_AddRefs(requestUri)); + nsAutoCString spec; + requestUri->GetSpec(spec); + if (spec.Length() > 50) { + spec.Truncate(50); + spec.AppendLiteral("..."); + } + httpChannel->LogMimeTypeMismatch( + "XTCOWithMIMEValueMissing"_ns, false, NS_ConvertUTF8toUTF16(spec), + // Type is not used in the Error Message but required + NS_ConvertUTF8toUTF16(type)); + } + return; + } + } + + const char* testData = mBuffer; + uint32_t testDataLen = mBufferLen; + // Check if data are compressed. + nsAutoCString decodedData; + + if (channel) { + // ConvertEncodedData is always called only on a single thread for each + // instance of an object. + nsresult rv = ConvertEncodedData(aRequest, mBuffer, mBufferLen); + if (NS_SUCCEEDED(rv)) { + MutexAutoLock lock(mMutex); + decodedData = mDecodedData; + } + if (!decodedData.IsEmpty()) { + testData = decodedData.get(); + testDataLen = std::min<uint32_t>(decodedData.Length(), MAX_BUFFER_SIZE); + } + } + + // First, run through all the types we can detect reliably based on + // magic numbers + uint32_t i; + for (i = 0; i < sSnifferEntryNum; ++i) { + if (testDataLen >= sSnifferEntries[i].mByteLen && // enough data + memcmp(testData, sSnifferEntries[i].mBytes, + sSnifferEntries[i].mByteLen) == 0) { // and type matches + NS_ASSERTION( + sSnifferEntries[i].mMimeType || + sSnifferEntries[i].mContentTypeSniffer, + "Must have either a type string or a function to set the type"); + NS_ASSERTION(!sSnifferEntries[i].mMimeType || + !sSnifferEntries[i].mContentTypeSniffer, + "Both a type string and a type sniffing function set;" + " using type string"); + if (sSnifferEntries[i].mMimeType) { + MutexAutoLock lock(mMutex); + mContentType = sSnifferEntries[i].mMimeType; + NS_ASSERTION(!mContentType.IsEmpty(), + "Content type should be known by now."); + return; + } + if ((this->*(sSnifferEntries[i].mContentTypeSniffer))(aRequest)) { +#ifdef DEBUG + MutexAutoLock lock(mMutex); + NS_ASSERTION(!mContentType.IsEmpty(), + "Content type should be known by now."); +#endif + return; + } + } + } + + nsAutoCString sniffedType; + NS_SniffContent(NS_DATA_SNIFFER_CATEGORY, aRequest, (const uint8_t*)testData, + testDataLen, sniffedType); + { + MutexAutoLock lock(mMutex); + mContentType = sniffedType; + if (!mContentType.IsEmpty()) { + return; + } + } + + if (SniffForHTML(aRequest)) { +#ifdef DEBUG + MutexAutoLock lock(mMutex); + NS_ASSERTION(!mContentType.IsEmpty(), + "Content type should be known by now."); +#endif + return; + } + + // We don't know what this is yet. Before we just give up, try + // the URI from the request. + if (SniffURI(aRequest)) { +#ifdef DEBUG + MutexAutoLock lock(mMutex); + NS_ASSERTION(!mContentType.IsEmpty(), + "Content type should be known by now."); +#endif + return; + } + + LastDitchSniff(aRequest); +#ifdef DEBUG + MutexAutoLock lock(mMutex); + NS_ASSERTION(!mContentType.IsEmpty(), "Content type should be known by now."); +#endif +} + +bool nsUnknownDecoder::SniffForHTML(nsIRequest* aRequest) { + MutexAutoLock lock(mMutex); + + // Now look for HTML. + const char* str; + const char* end; + if (mDecodedData.IsEmpty()) { + str = mBuffer; + end = mBuffer + mBufferLen; + } else { + str = mDecodedData.get(); + end = mDecodedData.get() + + std::min<uint32_t>(mDecodedData.Length(), MAX_BUFFER_SIZE); + } + + // skip leading whitespace + while (str != end && nsCRT::IsAsciiSpace(*str)) { + ++str; + } + + // did we find something like a start tag? + if (str == end || *str != '<' || ++str == end) { + return false; + } + + // If we seem to be SGML or XML and we got down here, just pretend we're HTML + if (*str == '!' || *str == '?') { + mContentType = TEXT_HTML; + return true; + } + + uint32_t bufSize = end - str; + // We use sizeof(_tagstr) below because that's the length of _tagstr + // with the one char " " or ">" appended. +#define MATCHES_TAG(_tagstr) \ + (bufSize >= sizeof(_tagstr) && \ + (nsCRT::strncasecmp(str, _tagstr " ", sizeof(_tagstr)) == 0 || \ + nsCRT::strncasecmp(str, _tagstr ">", sizeof(_tagstr)) == 0)) + + if (MATCHES_TAG("html") || MATCHES_TAG("frameset") || MATCHES_TAG("body") || + MATCHES_TAG("head") || MATCHES_TAG("script") || MATCHES_TAG("iframe") || + MATCHES_TAG("a") || MATCHES_TAG("img") || MATCHES_TAG("table") || + MATCHES_TAG("title") || MATCHES_TAG("link") || MATCHES_TAG("base") || + MATCHES_TAG("style") || MATCHES_TAG("div") || MATCHES_TAG("p") || + MATCHES_TAG("font") || MATCHES_TAG("applet") || MATCHES_TAG("meta") || + MATCHES_TAG("center") || MATCHES_TAG("form") || MATCHES_TAG("isindex") || + MATCHES_TAG("h1") || MATCHES_TAG("h2") || MATCHES_TAG("h3") || + MATCHES_TAG("h4") || MATCHES_TAG("h5") || MATCHES_TAG("h6") || + MATCHES_TAG("b") || MATCHES_TAG("pre")) { + mContentType = TEXT_HTML; + return true; + } + +#undef MATCHES_TAG + + return false; +} + +bool nsUnknownDecoder::SniffForXML(nsIRequest* aRequest) { + // First see whether we can glean anything from the uri... + if (!SniffURI(aRequest)) { + // Oh well; just generic XML will have to do + MutexAutoLock lock(mMutex); + mContentType = TEXT_XML; + } + + return true; +} + +bool nsUnknownDecoder::SniffURI(nsIRequest* aRequest) { + nsCOMPtr<nsIChannel> channel(do_QueryInterface(aRequest)); + nsCOMPtr<nsILoadInfo> loadInfo = channel->LoadInfo(); + if (loadInfo->GetSkipContentSniffing()) { + return false; + } + nsCOMPtr<nsIMIMEService> mimeService(do_GetService("@mozilla.org/mime;1")); + if (mimeService) { + nsCOMPtr<nsIChannel> channel = do_QueryInterface(aRequest); + if (channel) { + nsCOMPtr<nsIURI> uri; + nsresult result = channel->GetURI(getter_AddRefs(uri)); + if (NS_SUCCEEDED(result) && uri) { + nsAutoCString type; + result = mimeService->GetTypeFromURI(uri, type); + if (NS_SUCCEEDED(result)) { + MutexAutoLock lock(mMutex); + mContentType = type; + return true; + } + } + } + } + + return false; +} + +// This macro is based on RFC 2046 Section 4.1.2. Treat any char 0-31 +// except the 9-13 range (\t, \n, \v, \f, \r) and char 27 (used by +// encodings like Shift_JIS) as non-text +#define IS_TEXT_CHAR(ch) \ + (((unsigned char)(ch)) > 31 || (9 <= (ch) && (ch) <= 13) || (ch) == 27) + +bool nsUnknownDecoder::LastDitchSniff(nsIRequest* aRequest) { + // All we can do now is try to guess whether this is text/plain or + // application/octet-stream + + MutexAutoLock lock(mMutex); + + const char* testData; + uint32_t testDataLen; + if (mDecodedData.IsEmpty()) { + testData = mBuffer; + // Since some legacy text files end with 0x1A, reading the entire buffer + // will lead misdetection. + testDataLen = std::min<uint32_t>(mBufferLen, MAX_BUFFER_SIZE); + } else { + testData = mDecodedData.get(); + testDataLen = std::min<uint32_t>(mDecodedData.Length(), MAX_BUFFER_SIZE); + } + + // First, check for a BOM. If we see one, assume this is text/plain + // in whatever encoding. If there is a BOM _and_ text we will + // always have at least 4 bytes in the buffer (since the 2-byte BOMs + // are for 2-byte encodings and the UTF-8 BOM is 3 bytes). + if (testDataLen >= 4) { + const unsigned char* buf = (const unsigned char*)testData; + if ((buf[0] == 0xFE && buf[1] == 0xFF) || // UTF-16, Big Endian + (buf[0] == 0xFF && buf[1] == 0xFE) || // UTF-16 or UCS-4, Little Endian + (buf[0] == 0xEF && buf[1] == 0xBB && buf[2] == 0xBF) || // UTF-8 + (buf[0] == 0 && buf[1] == 0 && buf[2] == 0xFE && + buf[3] == 0xFF)) { // UCS-4, Big Endian + + mContentType = TEXT_PLAIN; + return true; + } + } + + // Now see whether the buffer has any non-text chars. If not, then let's + // just call it text/plain... + // + uint32_t i; + for (i = 0; i < testDataLen && IS_TEXT_CHAR(testData[i]); i++) { + } + + if (i == testDataLen) { + mContentType = TEXT_PLAIN; + } else { + mContentType = APPLICATION_OCTET_STREAM; + } + + return true; +} + +nsresult nsUnknownDecoder::FireListenerNotifications(nsIRequest* request, + nsISupports* aCtxt) { + nsresult rv = NS_OK; + + nsCOMPtr<nsIStreamListener> listener; + nsAutoCString contentType; + { + MutexAutoLock lock(mMutex); + if (!mNextListener) return NS_ERROR_FAILURE; + + listener = mNextListener; + contentType = mContentType; + } + + if (!contentType.IsEmpty()) { + nsCOMPtr<nsIViewSourceChannel> viewSourceChannel = + do_QueryInterface(request); + if (viewSourceChannel) { + rv = viewSourceChannel->SetOriginalContentType(contentType); + } else { + nsCOMPtr<nsIChannel> channel = do_QueryInterface(request, &rv); + if (NS_SUCCEEDED(rv)) { + // Set the new content type on the channel... + rv = channel->SetContentType(contentType); + } + } + + NS_ASSERTION(NS_SUCCEEDED(rv), "Unable to set content type on channel!"); + + if (NS_FAILED(rv)) { + // Cancel the request to make sure it has the correct status if + // mNextListener looks at it. + request->Cancel(rv); + listener->OnStartRequest(request); + return rv; + } + } + + // Fire the OnStartRequest(...) + rv = listener->OnStartRequest(request); + + if (NS_SUCCEEDED(rv)) { + // install stream converter if required + nsCOMPtr<nsIEncodedChannel> encodedChannel = do_QueryInterface(request); + if (encodedChannel) { + nsCOMPtr<nsIStreamListener> listenerNew; + rv = encodedChannel->DoApplyContentConversions( + listener, getter_AddRefs(listenerNew), aCtxt); + if (NS_SUCCEEDED(rv) && listenerNew) { + MutexAutoLock lock(mMutex); + mNextListener = listenerNew; + listener = listenerNew; + } + } + } + + if (!mBuffer) return NS_ERROR_OUT_OF_MEMORY; + + // If the request was canceled, then we need to treat that equivalently + // to an error returned by OnStartRequest. + if (NS_SUCCEEDED(rv)) request->GetStatus(&rv); + + // Fire the first OnDataAvailable for the data that was read from the + // stream into the sniffer buffer... + if (NS_SUCCEEDED(rv) && (mBufferLen > 0)) { + uint32_t len = 0; + nsCOMPtr<nsIInputStream> in; + nsCOMPtr<nsIOutputStream> out; + + // Create a pipe and fill it with the data from the sniffer buffer. + NS_NewPipe(getter_AddRefs(in), getter_AddRefs(out), MAX_BUFFER_SIZE, + MAX_BUFFER_SIZE); + + rv = out->Write(mBuffer, mBufferLen, &len); + if (NS_SUCCEEDED(rv)) { + if (len == mBufferLen) { + rv = listener->OnDataAvailable(request, in, 0, len); + } else { + NS_ERROR("Unable to write all the data into the pipe."); + rv = NS_ERROR_FAILURE; + } + } + } + + delete[] mBuffer; + mBuffer = nullptr; + mBufferLen = 0; + + return rv; +} + +nsresult nsUnknownDecoder::ConvertEncodedData(nsIRequest* request, + const char* data, + uint32_t length) { + nsresult rv = NS_OK; + + { + MutexAutoLock lock(mMutex); + mDecodedData = ""; + } + nsCOMPtr<nsIEncodedChannel> encodedChannel(do_QueryInterface(request)); + if (encodedChannel) { + RefPtr<ConvertedStreamListener> strListener = + new ConvertedStreamListener(this); + + nsCOMPtr<nsIStreamListener> listener; + rv = encodedChannel->DoApplyContentConversions( + strListener, getter_AddRefs(listener), nullptr); + + if (NS_FAILED(rv)) { + return rv; + } + + if (listener) { + listener->OnStartRequest(request); + + if (length) { + nsCOMPtr<nsIStringInputStream> rawStream = + do_CreateInstance(NS_STRINGINPUTSTREAM_CONTRACTID); + if (!rawStream) return NS_ERROR_FAILURE; + + rv = rawStream->SetData((const char*)data, length); + NS_ENSURE_SUCCESS(rv, rv); + + rv = listener->OnDataAvailable(request, rawStream, 0, length); + NS_ENSURE_SUCCESS(rv, rv); + } + + listener->OnStopRequest(request, NS_OK); + } + } + return rv; +} + +// +// nsIThreadRetargetableStreamListener methods +// +NS_IMETHODIMP +nsUnknownDecoder::CheckListenerChain() { + nsCOMPtr<nsIThreadRetargetableStreamListener> listener; + { + MutexAutoLock lock(mMutex); + listener = do_QueryInterface(mNextListener); + } + if (!listener) { + return NS_ERROR_NO_INTERFACE; + } + + return listener->CheckListenerChain(); +} + +void nsBinaryDetector::DetermineContentType(nsIRequest* aRequest) { + nsCOMPtr<nsIHttpChannel> httpChannel = do_QueryInterface(aRequest); + if (!httpChannel) { + return; + } + + nsCOMPtr<nsILoadInfo> loadInfo = httpChannel->LoadInfo(); + if (loadInfo->GetSkipContentSniffing()) { + LastDitchSniff(aRequest); + return; + } + // It's an HTTP channel. Check for the text/plain mess + nsAutoCString contentTypeHdr; + Unused << httpChannel->GetResponseHeader("Content-Type"_ns, contentTypeHdr); + nsAutoCString contentType; + httpChannel->GetContentType(contentType); + + // Make sure to do a case-sensitive exact match comparison here. Apache + // 1.x just sends text/plain for "unknown", while Apache 2.x sends + // text/plain with a ISO-8859-1 charset. Debian's Apache version, just to + // be different, sends text/plain with iso-8859-1 charset. For extra fun, + // FC7, RHEL4, and Ubuntu Feisty send charset=UTF-8. Don't do general + // case-insensitive comparison, since we really want to apply this crap as + // rarely as we can. + if (!contentType.EqualsLiteral("text/plain") || + (!contentTypeHdr.EqualsLiteral("text/plain") && + !contentTypeHdr.EqualsLiteral("text/plain; charset=ISO-8859-1") && + !contentTypeHdr.EqualsLiteral("text/plain; charset=iso-8859-1") && + !contentTypeHdr.EqualsLiteral("text/plain; charset=UTF-8"))) { + return; + } + + // Check whether we have content-encoding. If we do, don't try to + // detect the type. + // XXXbz we could improve this by doing a local decompress if we + // wanted, I'm sure. + nsAutoCString contentEncoding; + Unused << httpChannel->GetResponseHeader("Content-Encoding"_ns, + contentEncoding); + if (!contentEncoding.IsEmpty()) { + return; + } + + LastDitchSniff(aRequest); + MutexAutoLock lock(mMutex); + if (mContentType.EqualsLiteral(APPLICATION_OCTET_STREAM)) { + // We want to guess at it instead + mContentType = APPLICATION_GUESS_FROM_EXT; + } else { + // Let the text/plain type we already have be, so that other content + // sniffers can also get a shot at this data. + mContentType.Truncate(); + } +} |