diff --git a/webrtc/common_video/h264/h264_bitstream_parser.cc b/webrtc/common_video/h264/h264_bitstream_parser.cc index 28b92e1241..5f89c50af1 100644 --- a/webrtc/common_video/h264/h264_bitstream_parser.cc +++ b/webrtc/common_video/h264/h264_bitstream_parser.cc @@ -21,23 +21,30 @@ namespace webrtc { -#define RETURN_FALSE_ON_FAIL(x) \ +#define RETURN_ON_FAIL(x, res) \ if (!(x)) { \ LOG_F(LS_ERROR) << "FAILED: " #x; \ - return false; \ + return res; \ } +#define RETURN_INV_ON_FAIL(x) RETURN_ON_FAIL(x, kInvalidStream) + H264BitstreamParser::H264BitstreamParser() {} H264BitstreamParser::~H264BitstreamParser() {} -bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, - size_t source_length, - uint8_t nalu_type) { - RTC_CHECK(sps_); - RTC_CHECK(pps_); +H264BitstreamParser::Result H264BitstreamParser::ParseNonParameterSetNalu( + const uint8_t* source, + size_t source_length, + uint8_t nalu_type) { + if (!sps_ || !pps_) + return kInvalidStream; + last_slice_qp_delta_ = rtc::Optional(); std::unique_ptr slice_rbsp( H264::ParseRbsp(source, source_length)); + if (slice_rbsp->size() < H264::kNaluTypeSize) + return kInvalidStream; + rtc::BitBuffer slice_reader(slice_rbsp->data() + H264::kNaluTypeSize, slice_rbsp->size() - H264::kNaluTypeSize); // Check to see if this is an IDR slice, which has an extra field to parse @@ -48,64 +55,64 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, uint32_t bits_tmp; // first_mb_in_slice: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); // slice_type: ue(v) uint32_t slice_type; - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&slice_type)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&slice_type)); // slice_type's 5..9 range is used to indicate that all slices of a picture // have the same value of slice_type % 5, we don't care about that, so we map // to the corresponding 0..4 range. slice_type %= 5; // pic_parameter_set_id: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); if (sps_->separate_colour_plane_flag == 1) { // colour_plane_id - RETURN_FALSE_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 2)); + RETURN_INV_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 2)); } // frame_num: u(v) // Represented by log2_max_frame_num_minus4 + 4 bits. - RETURN_FALSE_ON_FAIL( + RETURN_INV_ON_FAIL( slice_reader.ReadBits(&bits_tmp, sps_->log2_max_frame_num_minus4 + 4)); uint32_t field_pic_flag = 0; if (sps_->frame_mbs_only_flag == 0) { // field_pic_flag: u(1) - RETURN_FALSE_ON_FAIL(slice_reader.ReadBits(&field_pic_flag, 1)); + RETURN_INV_ON_FAIL(slice_reader.ReadBits(&field_pic_flag, 1)); if (field_pic_flag != 0) { // bottom_field_flag: u(1) - RETURN_FALSE_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 1)); + RETURN_INV_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 1)); } } if (is_idr) { // idr_pic_id: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } // pic_order_cnt_lsb: u(v) // Represented by sps_.log2_max_pic_order_cnt_lsb_minus4 + 4 bits. if (sps_->pic_order_cnt_type == 0) { - RETURN_FALSE_ON_FAIL(slice_reader.ReadBits( + RETURN_INV_ON_FAIL(slice_reader.ReadBits( &bits_tmp, sps_->log2_max_pic_order_cnt_lsb_minus4 + 4)); if (pps_->bottom_field_pic_order_in_frame_present_flag && field_pic_flag == 0) { // delta_pic_order_cnt_bottom: se(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } } if (sps_->pic_order_cnt_type == 1 && !sps_->delta_pic_order_always_zero_flag) { // delta_pic_order_cnt[0]: se(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); if (pps_->bottom_field_pic_order_in_frame_present_flag && !field_pic_flag) { // delta_pic_order_cnt[1]: se(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } } if (pps_->redundant_pic_cnt_present_flag) { // redundant_pic_cnt: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } if (slice_type == H264::SliceType::kB) { // direct_spatial_mv_pred_flag: u(1) - RETURN_FALSE_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 1)); + RETURN_INV_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 1)); } switch (slice_type) { case H264::SliceType::kP: @@ -113,14 +120,14 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, case H264::SliceType::kSp: uint32_t num_ref_idx_active_override_flag; // num_ref_idx_active_override_flag: u(1) - RETURN_FALSE_ON_FAIL( + RETURN_INV_ON_FAIL( slice_reader.ReadBits(&num_ref_idx_active_override_flag, 1)); if (num_ref_idx_active_override_flag != 0) { // num_ref_idx_l0_active_minus1: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); if (slice_type == H264::SliceType::kB) { // num_ref_idx_l1_active_minus1: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } } break; @@ -128,8 +135,10 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, break; } // assume nal_unit_type != 20 && nal_unit_type != 21: - RTC_CHECK_NE(nalu_type, 20); - RTC_CHECK_NE(nalu_type, 21); + if (nalu_type == 20 || nalu_type == 21) { + LOG(LS_ERROR) << "Unsupported nal unit type."; + return kUnsupportedStream; + } // if (nal_unit_type == 20 || nal_unit_type == 21) // ref_pic_list_mvc_modification() // else @@ -141,23 +150,21 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, if (slice_type % 5 != 2 && slice_type % 5 != 4) { // ref_pic_list_modification_flag_l0: u(1) uint32_t ref_pic_list_modification_flag_l0; - RETURN_FALSE_ON_FAIL( + RETURN_INV_ON_FAIL( slice_reader.ReadBits(&ref_pic_list_modification_flag_l0, 1)); if (ref_pic_list_modification_flag_l0) { uint32_t modification_of_pic_nums_idc; do { // modification_of_pic_nums_idc: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb( + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb( &modification_of_pic_nums_idc)); if (modification_of_pic_nums_idc == 0 || modification_of_pic_nums_idc == 1) { // abs_diff_pic_num_minus1: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } else if (modification_of_pic_nums_idc == 2) { // long_term_pic_num: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } } while (modification_of_pic_nums_idc != 3); } @@ -165,34 +172,33 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, if (slice_type % 5 == 1) { // ref_pic_list_modification_flag_l1: u(1) uint32_t ref_pic_list_modification_flag_l1; - RETURN_FALSE_ON_FAIL( + RETURN_INV_ON_FAIL( slice_reader.ReadBits(&ref_pic_list_modification_flag_l1, 1)); if (ref_pic_list_modification_flag_l1) { uint32_t modification_of_pic_nums_idc; do { // modification_of_pic_nums_idc: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb( + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb( &modification_of_pic_nums_idc)); if (modification_of_pic_nums_idc == 0 || modification_of_pic_nums_idc == 1) { // abs_diff_pic_num_minus1: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } else if (modification_of_pic_nums_idc == 2) { // long_term_pic_num: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } } while (modification_of_pic_nums_idc != 3); } } } // TODO(pbos): Do we need support for pred_weight_table()? - RTC_CHECK( - !((pps_->weighted_pred_flag && (slice_type == H264::SliceType::kP || - slice_type == H264::SliceType::kSp)) || - (pps_->weighted_bipred_idc != 0 && slice_type == H264::SliceType::kB))) - << "Missing support for pred_weight_table()."; + if ((pps_->weighted_pred_flag && (slice_type == H264::SliceType::kP || + slice_type == H264::SliceType::kSp)) || + (pps_->weighted_bipred_idc == 1 && slice_type == H264::SliceType::kB)) { + LOG(LS_ERROR) << "Streams with pred_weight_table unsupported."; + return kUnsupportedStream; + } // if ((weighted_pred_flag && (slice_type == P || slice_type == SP)) || // (weighted_bipred_idc == 1 && slice_type == B)) { // pred_weight_table() @@ -202,39 +208,35 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, if (is_idr) { // no_output_of_prior_pics_flag: u(1) // long_term_reference_flag: u(1) - RETURN_FALSE_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 2)); + RETURN_INV_ON_FAIL(slice_reader.ReadBits(&bits_tmp, 2)); } else { // adaptive_ref_pic_marking_mode_flag: u(1) uint32_t adaptive_ref_pic_marking_mode_flag; - RETURN_FALSE_ON_FAIL( + RETURN_INV_ON_FAIL( slice_reader.ReadBits(&adaptive_ref_pic_marking_mode_flag, 1)); if (adaptive_ref_pic_marking_mode_flag) { uint32_t memory_management_control_operation; do { // memory_management_control_operation: ue(v) - RETURN_FALSE_ON_FAIL(slice_reader.ReadExponentialGolomb( + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb( &memory_management_control_operation)); if (memory_management_control_operation == 1 || memory_management_control_operation == 3) { // difference_of_pic_nums_minus1: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } if (memory_management_control_operation == 2) { // long_term_pic_num: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } if (memory_management_control_operation == 3 || memory_management_control_operation == 6) { // long_term_frame_idx: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } if (memory_management_control_operation == 4) { // max_long_term_frame_idx_plus1: ue(v) - RETURN_FALSE_ON_FAIL( - slice_reader.ReadExponentialGolomb(&golomb_tmp)); + RETURN_INV_ON_FAIL(slice_reader.ReadExponentialGolomb(&golomb_tmp)); } } while (memory_management_control_operation != 0); } @@ -244,10 +246,10 @@ bool H264BitstreamParser::ParseNonParameterSetNalu(const uint8_t* source, // if (entropy_coding_mode_flag && slice_type != I && slice_type != SI) // cabac_init_idc int32_t last_slice_qp_delta; - RETURN_FALSE_ON_FAIL( + RETURN_INV_ON_FAIL( slice_reader.ReadSignedExponentialGolomb(&last_slice_qp_delta)); last_slice_qp_delta_ = rtc::Optional(last_slice_qp_delta); - return true; + return kOk; } void H264BitstreamParser::ParseSlice(const uint8_t* slice, size_t length) { @@ -257,19 +259,20 @@ void H264BitstreamParser::ParseSlice(const uint8_t* slice, size_t length) { sps_ = SpsParser::ParseSps(slice + H264::kNaluTypeSize, length - H264::kNaluTypeSize); if (!sps_) - FATAL() << "Unable to parse SPS from H264 bitstream."; + LOG(LS_WARNING) << "Unable to parse SPS from H264 bitstream."; break; } case H264::NaluType::kPps: { pps_ = PpsParser::ParsePps(slice + H264::kNaluTypeSize, length - H264::kNaluTypeSize); if (!pps_) - FATAL() << "Unable to parse PPS from H264 bitstream."; + LOG(LS_WARNING) << "Unable to parse PPS from H264 bitstream."; break; } default: - RTC_CHECK(ParseNonParameterSetNalu(slice, length, nalu_type)) - << "Failed to parse picture slice."; + Result res = ParseNonParameterSetNalu(slice, length, nalu_type); + if (res != kOk) + LOG(LS_INFO) << "Failed to parse bitstream. Error: " << res; break; } } @@ -278,7 +281,6 @@ void H264BitstreamParser::ParseBitstream(const uint8_t* bitstream, size_t length) { std::vector nalu_indices = H264::FindNaluIndices(bitstream, length); - RTC_CHECK(!nalu_indices.empty()); for (const H264::NaluIndex& index : nalu_indices) ParseSlice(&bitstream[index.payload_start_offset], index.payload_size); } diff --git a/webrtc/common_video/h264/h264_bitstream_parser.h b/webrtc/common_video/h264/h264_bitstream_parser.h index 0b84e40234..aa0155239a 100644 --- a/webrtc/common_video/h264/h264_bitstream_parser.h +++ b/webrtc/common_video/h264/h264_bitstream_parser.h @@ -31,6 +31,12 @@ namespace webrtc { // bitstreams. class H264BitstreamParser { public: + enum Result { + kOk, + kInvalidStream, + kUnsupportedStream, + }; + H264BitstreamParser(); virtual ~H264BitstreamParser(); @@ -42,9 +48,9 @@ class H264BitstreamParser { protected: void ParseSlice(const uint8_t* slice, size_t length); - bool ParseNonParameterSetNalu(const uint8_t* source, - size_t source_length, - uint8_t nalu_type); + Result ParseNonParameterSetNalu(const uint8_t* source, + size_t source_length, + uint8_t nalu_type); // SPS/PPS state, updated when parsing new SPS/PPS, used to parse slices. rtc::Optional sps_; diff --git a/webrtc/common_video/h264/h264_common.cc b/webrtc/common_video/h264/h264_common.cc index fe55b02d2c..a9cc6a25e6 100644 --- a/webrtc/common_video/h264/h264_common.cc +++ b/webrtc/common_video/h264/h264_common.cc @@ -24,6 +24,7 @@ std::vector FindNaluIndices(const uint8_t* buffer, std::vector sequences; if (buffer_size < kNaluShortStartSequenceSize) return sequences; + const size_t end = buffer_size - kNaluShortStartSequenceSize; for (size_t i = 0; i < end;) { if (buffer[i + 2] > 1) {