Tue, 06 Jan 2015 21:39:09 +0100
Conditionally force memory storage according to privacy.thirdparty.isolate;
This solves Tor bug #9701, complying with disk avoidance documented in
https://www.torproject.org/projects/torbrowser/design/#disk-avoidance.
michael@0 | 1 | /* -*- Mode: C; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ |
michael@0 | 2 | /* This Source Code Form is subject to the terms of the Mozilla Public |
michael@0 | 3 | * License, v. 2.0. If a copy of the MPL was not distributed with this |
michael@0 | 4 | * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
michael@0 | 5 | |
michael@0 | 6 | #include "nsConverterInputStream.h" |
michael@0 | 7 | #include "nsIInputStream.h" |
michael@0 | 8 | #include "nsICharsetConverterManager.h" |
michael@0 | 9 | #include "nsReadLine.h" |
michael@0 | 10 | #include "nsStreamUtils.h" |
michael@0 | 11 | #include "nsServiceManagerUtils.h" |
michael@0 | 12 | #include <algorithm> |
michael@0 | 13 | |
michael@0 | 14 | #define CONVERTER_BUFFER_SIZE 8192 |
michael@0 | 15 | |
michael@0 | 16 | NS_IMPL_ISUPPORTS(nsConverterInputStream, nsIConverterInputStream, |
michael@0 | 17 | nsIUnicharInputStream, nsIUnicharLineInputStream) |
michael@0 | 18 | |
michael@0 | 19 | |
michael@0 | 20 | NS_IMETHODIMP |
michael@0 | 21 | nsConverterInputStream::Init(nsIInputStream* aStream, |
michael@0 | 22 | const char *aCharset, |
michael@0 | 23 | int32_t aBufferSize, |
michael@0 | 24 | char16_t aReplacementChar) |
michael@0 | 25 | { |
michael@0 | 26 | static NS_DEFINE_CID(kCharsetConverterManagerCID, NS_ICHARSETCONVERTERMANAGER_CID); |
michael@0 | 27 | |
michael@0 | 28 | if (!aCharset) |
michael@0 | 29 | aCharset = "UTF-8"; |
michael@0 | 30 | |
michael@0 | 31 | nsresult rv; |
michael@0 | 32 | |
michael@0 | 33 | if (aBufferSize <=0) aBufferSize=CONVERTER_BUFFER_SIZE; |
michael@0 | 34 | |
michael@0 | 35 | // get the decoder |
michael@0 | 36 | nsCOMPtr<nsICharsetConverterManager> ccm = |
michael@0 | 37 | do_GetService(kCharsetConverterManagerCID, &rv); |
michael@0 | 38 | if (NS_FAILED(rv)) return rv; |
michael@0 | 39 | |
michael@0 | 40 | rv = ccm->GetUnicodeDecoder(aCharset ? aCharset : "ISO-8859-1", getter_AddRefs(mConverter)); |
michael@0 | 41 | if (NS_FAILED(rv)) return rv; |
michael@0 | 42 | |
michael@0 | 43 | // set up our buffers |
michael@0 | 44 | if (!mByteData.SetCapacity(aBufferSize) || |
michael@0 | 45 | !mUnicharData.SetCapacity(aBufferSize)) { |
michael@0 | 46 | return NS_ERROR_OUT_OF_MEMORY; |
michael@0 | 47 | } |
michael@0 | 48 | |
michael@0 | 49 | mInput = aStream; |
michael@0 | 50 | mReplacementChar = aReplacementChar; |
michael@0 | 51 | if (!aReplacementChar || |
michael@0 | 52 | aReplacementChar != mConverter->GetCharacterForUnMapped()) { |
michael@0 | 53 | mConverter->SetInputErrorBehavior(nsIUnicodeDecoder::kOnError_Signal); |
michael@0 | 54 | } |
michael@0 | 55 | |
michael@0 | 56 | return NS_OK; |
michael@0 | 57 | } |
michael@0 | 58 | |
michael@0 | 59 | NS_IMETHODIMP |
michael@0 | 60 | nsConverterInputStream::Close() |
michael@0 | 61 | { |
michael@0 | 62 | nsresult rv = mInput ? mInput->Close() : NS_OK; |
michael@0 | 63 | mLineBuffer = nullptr; |
michael@0 | 64 | mInput = nullptr; |
michael@0 | 65 | mConverter = nullptr; |
michael@0 | 66 | mByteData.Clear(); |
michael@0 | 67 | mUnicharData.Clear(); |
michael@0 | 68 | return rv; |
michael@0 | 69 | } |
michael@0 | 70 | |
michael@0 | 71 | NS_IMETHODIMP |
michael@0 | 72 | nsConverterInputStream::Read(char16_t* aBuf, |
michael@0 | 73 | uint32_t aCount, |
michael@0 | 74 | uint32_t *aReadCount) |
michael@0 | 75 | { |
michael@0 | 76 | NS_ASSERTION(mUnicharDataLength >= mUnicharDataOffset, "unsigned madness"); |
michael@0 | 77 | uint32_t readCount = mUnicharDataLength - mUnicharDataOffset; |
michael@0 | 78 | if (0 == readCount) { |
michael@0 | 79 | // Fill the unichar buffer |
michael@0 | 80 | readCount = Fill(&mLastErrorCode); |
michael@0 | 81 | if (readCount == 0) { |
michael@0 | 82 | *aReadCount = 0; |
michael@0 | 83 | return mLastErrorCode; |
michael@0 | 84 | } |
michael@0 | 85 | } |
michael@0 | 86 | if (readCount > aCount) { |
michael@0 | 87 | readCount = aCount; |
michael@0 | 88 | } |
michael@0 | 89 | memcpy(aBuf, mUnicharData.Elements() + mUnicharDataOffset, |
michael@0 | 90 | readCount * sizeof(char16_t)); |
michael@0 | 91 | mUnicharDataOffset += readCount; |
michael@0 | 92 | *aReadCount = readCount; |
michael@0 | 93 | return NS_OK; |
michael@0 | 94 | } |
michael@0 | 95 | |
michael@0 | 96 | NS_IMETHODIMP |
michael@0 | 97 | nsConverterInputStream::ReadSegments(nsWriteUnicharSegmentFun aWriter, |
michael@0 | 98 | void* aClosure, |
michael@0 | 99 | uint32_t aCount, uint32_t *aReadCount) |
michael@0 | 100 | { |
michael@0 | 101 | NS_ASSERTION(mUnicharDataLength >= mUnicharDataOffset, "unsigned madness"); |
michael@0 | 102 | uint32_t bytesToWrite = mUnicharDataLength - mUnicharDataOffset; |
michael@0 | 103 | nsresult rv; |
michael@0 | 104 | if (0 == bytesToWrite) { |
michael@0 | 105 | // Fill the unichar buffer |
michael@0 | 106 | bytesToWrite = Fill(&rv); |
michael@0 | 107 | if (bytesToWrite <= 0) { |
michael@0 | 108 | *aReadCount = 0; |
michael@0 | 109 | return rv; |
michael@0 | 110 | } |
michael@0 | 111 | } |
michael@0 | 112 | |
michael@0 | 113 | if (bytesToWrite > aCount) |
michael@0 | 114 | bytesToWrite = aCount; |
michael@0 | 115 | |
michael@0 | 116 | uint32_t bytesWritten; |
michael@0 | 117 | uint32_t totalBytesWritten = 0; |
michael@0 | 118 | |
michael@0 | 119 | while (bytesToWrite) { |
michael@0 | 120 | rv = aWriter(this, aClosure, |
michael@0 | 121 | mUnicharData.Elements() + mUnicharDataOffset, |
michael@0 | 122 | totalBytesWritten, bytesToWrite, &bytesWritten); |
michael@0 | 123 | if (NS_FAILED(rv)) { |
michael@0 | 124 | // don't propagate errors to the caller |
michael@0 | 125 | break; |
michael@0 | 126 | } |
michael@0 | 127 | |
michael@0 | 128 | bytesToWrite -= bytesWritten; |
michael@0 | 129 | totalBytesWritten += bytesWritten; |
michael@0 | 130 | mUnicharDataOffset += bytesWritten; |
michael@0 | 131 | |
michael@0 | 132 | } |
michael@0 | 133 | |
michael@0 | 134 | *aReadCount = totalBytesWritten; |
michael@0 | 135 | |
michael@0 | 136 | return NS_OK; |
michael@0 | 137 | } |
michael@0 | 138 | |
michael@0 | 139 | NS_IMETHODIMP |
michael@0 | 140 | nsConverterInputStream::ReadString(uint32_t aCount, nsAString& aString, |
michael@0 | 141 | uint32_t* aReadCount) |
michael@0 | 142 | { |
michael@0 | 143 | NS_ASSERTION(mUnicharDataLength >= mUnicharDataOffset, "unsigned madness"); |
michael@0 | 144 | uint32_t readCount = mUnicharDataLength - mUnicharDataOffset; |
michael@0 | 145 | if (0 == readCount) { |
michael@0 | 146 | // Fill the unichar buffer |
michael@0 | 147 | readCount = Fill(&mLastErrorCode); |
michael@0 | 148 | if (readCount == 0) { |
michael@0 | 149 | *aReadCount = 0; |
michael@0 | 150 | return mLastErrorCode; |
michael@0 | 151 | } |
michael@0 | 152 | } |
michael@0 | 153 | if (readCount > aCount) { |
michael@0 | 154 | readCount = aCount; |
michael@0 | 155 | } |
michael@0 | 156 | const char16_t* buf = mUnicharData.Elements() + mUnicharDataOffset; |
michael@0 | 157 | aString.Assign(buf, readCount); |
michael@0 | 158 | mUnicharDataOffset += readCount; |
michael@0 | 159 | *aReadCount = readCount; |
michael@0 | 160 | return NS_OK; |
michael@0 | 161 | } |
michael@0 | 162 | |
michael@0 | 163 | uint32_t |
michael@0 | 164 | nsConverterInputStream::Fill(nsresult * aErrorCode) |
michael@0 | 165 | { |
michael@0 | 166 | if (nullptr == mInput) { |
michael@0 | 167 | // We already closed the stream! |
michael@0 | 168 | *aErrorCode = NS_BASE_STREAM_CLOSED; |
michael@0 | 169 | return 0; |
michael@0 | 170 | } |
michael@0 | 171 | |
michael@0 | 172 | if (NS_FAILED(mLastErrorCode)) { |
michael@0 | 173 | // We failed to completely convert last time, and error-recovery |
michael@0 | 174 | // is disabled. We will fare no better this time, so... |
michael@0 | 175 | *aErrorCode = mLastErrorCode; |
michael@0 | 176 | return 0; |
michael@0 | 177 | } |
michael@0 | 178 | |
michael@0 | 179 | // We assume a many to one conversion and are using equal sizes for |
michael@0 | 180 | // the two buffers. However if an error happens at the very start |
michael@0 | 181 | // of a byte buffer we may end up in a situation where n bytes lead |
michael@0 | 182 | // to n+1 unicode chars. Thus we need to keep track of the leftover |
michael@0 | 183 | // bytes as we convert. |
michael@0 | 184 | |
michael@0 | 185 | uint32_t nb; |
michael@0 | 186 | *aErrorCode = NS_FillArray(mByteData, mInput, mLeftOverBytes, &nb); |
michael@0 | 187 | if (nb == 0 && mLeftOverBytes == 0) { |
michael@0 | 188 | // No more data |
michael@0 | 189 | *aErrorCode = NS_OK; |
michael@0 | 190 | return 0; |
michael@0 | 191 | } |
michael@0 | 192 | |
michael@0 | 193 | NS_ASSERTION(uint32_t(nb) + mLeftOverBytes == mByteData.Length(), |
michael@0 | 194 | "mByteData is lying to us somewhere"); |
michael@0 | 195 | |
michael@0 | 196 | // Now convert as much of the byte buffer to unicode as possible |
michael@0 | 197 | mUnicharDataOffset = 0; |
michael@0 | 198 | mUnicharDataLength = 0; |
michael@0 | 199 | uint32_t srcConsumed = 0; |
michael@0 | 200 | do { |
michael@0 | 201 | int32_t srcLen = mByteData.Length() - srcConsumed; |
michael@0 | 202 | int32_t dstLen = mUnicharData.Capacity() - mUnicharDataLength; |
michael@0 | 203 | *aErrorCode = mConverter->Convert(mByteData.Elements()+srcConsumed, |
michael@0 | 204 | &srcLen, |
michael@0 | 205 | mUnicharData.Elements()+mUnicharDataLength, |
michael@0 | 206 | &dstLen); |
michael@0 | 207 | mUnicharDataLength += dstLen; |
michael@0 | 208 | // XXX if srcLen is negative, we want to drop the _first_ byte in |
michael@0 | 209 | // the erroneous byte sequence and try again. This is not quite |
michael@0 | 210 | // possible right now -- see bug 160784 |
michael@0 | 211 | srcConsumed += srcLen; |
michael@0 | 212 | if (NS_FAILED(*aErrorCode) && mReplacementChar) { |
michael@0 | 213 | NS_ASSERTION(0 < mUnicharData.Capacity() - mUnicharDataLength, |
michael@0 | 214 | "Decoder returned an error but filled the output buffer! " |
michael@0 | 215 | "Should not happen."); |
michael@0 | 216 | mUnicharData.Elements()[mUnicharDataLength++] = mReplacementChar; |
michael@0 | 217 | ++srcConsumed; |
michael@0 | 218 | // XXX this is needed to make sure we don't underrun our buffer; |
michael@0 | 219 | // bug 160784 again |
michael@0 | 220 | srcConsumed = std::max<uint32_t>(srcConsumed, 0); |
michael@0 | 221 | mConverter->Reset(); |
michael@0 | 222 | } |
michael@0 | 223 | NS_ASSERTION(srcConsumed <= mByteData.Length(), |
michael@0 | 224 | "Whoa. The converter should have returned NS_OK_UDEC_MOREINPUT before this point!"); |
michael@0 | 225 | } while (mReplacementChar && |
michael@0 | 226 | NS_FAILED(*aErrorCode) && |
michael@0 | 227 | mUnicharData.Capacity() > mUnicharDataLength); |
michael@0 | 228 | |
michael@0 | 229 | mLeftOverBytes = mByteData.Length() - srcConsumed; |
michael@0 | 230 | |
michael@0 | 231 | return mUnicharDataLength; |
michael@0 | 232 | } |
michael@0 | 233 | |
michael@0 | 234 | NS_IMETHODIMP |
michael@0 | 235 | nsConverterInputStream::ReadLine(nsAString& aLine, bool* aResult) |
michael@0 | 236 | { |
michael@0 | 237 | if (!mLineBuffer) { |
michael@0 | 238 | mLineBuffer = new nsLineBuffer<char16_t>; |
michael@0 | 239 | } |
michael@0 | 240 | return NS_ReadLine(this, mLineBuffer.get(), aLine, aResult); |
michael@0 | 241 | } |