Mitigate libsrtp wraparound with loss decryption failure (#1438)

versatica · Aug 29, 2024 · d3a7ae8 · d3a7ae8
1 parent 3604b1e
commit d3a7ae8
Show file tree

Hide file tree

Showing 12 changed files with 783 additions and 23 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -3,6 +3,7 @@
 ### NEXT
 
 - `Worker`: Fix `io_uring` support detection ([PR #1445](https://github.com/versatica/mediasoup/pull/1445)).
+- Mitigate libsrtp wraparound with loss decryption failure ([PR #1438](https://github.com/versatica/mediasoup/pull/1438)).
 
 ### 3.14.11
 

diff --git a/worker/include/RTC/PipeConsumer.hpp b/worker/include/RTC/PipeConsumer.hpp
@@ -73,7 +73,8 @@ namespace RTC
 		absl::flat_hash_map<uint32_t, RTC::RtpStreamSend*> mapSsrcRtpStream;
 		bool keyFrameSupported{ false };
 		absl::flat_hash_map<RTC::RtpStreamSend*, bool> mapRtpStreamSyncRequired;
-		absl::flat_hash_map<RTC::RtpStreamSend*, RTC::SeqManager<uint16_t>> mapRtpStreamRtpSeqManager;
+		absl::flat_hash_map<RTC::RtpStreamSend*, std::unique_ptr<RTC::SeqManager<uint16_t>>>
+		  mapRtpStreamRtpSeqManager;
 	};
 } // namespace RTC
 

diff --git a/worker/include/RTC/SeqManager.hpp b/worker/include/RTC/SeqManager.hpp
@@ -36,6 +36,7 @@ namespace RTC
 
 	public:
 		SeqManager() = default;
+		SeqManager(T initialOutput);
 
 	public:
 		void Sync(T input);
@@ -50,6 +51,7 @@ namespace RTC
 	private:
 		// Whether at least a sequence number has been inserted.
 		bool started{ false };
+		T initialOutput{ 0 };
 		T base{ 0 };
 		T maxOutput{ 0 };
 		T maxInput{ 0 };

diff --git a/worker/include/RTC/SimpleConsumer.hpp b/worker/include/RTC/SimpleConsumer.hpp
@@ -87,7 +87,7 @@ namespace RTC
 		RTC::RtpStreamRecv* producerRtpStream{ nullptr };
 		bool keyFrameSupported{ false };
 		bool syncRequired{ false };
-		RTC::SeqManager<uint16_t> rtpSeqManager;
+		std::unique_ptr<RTC::SeqManager<uint16_t>> rtpSeqManager;
 		bool managingBitrate{ false };
 		std::unique_ptr<RTC::Codecs::EncodingContext> encodingContext;
 	};

diff --git a/worker/include/RTC/SimulcastConsumer.hpp b/worker/include/RTC/SimulcastConsumer.hpp
@@ -115,7 +115,7 @@ namespace RTC
 		bool syncRequired{ false };
 		int16_t spatialLayerToSync{ -1 };
 		bool lastSentPacketHasMarker{ false };
-		RTC::SeqManager<uint16_t> rtpSeqManager;
+		std::unique_ptr<RTC::SeqManager<uint16_t>> rtpSeqManager;
 		int16_t preferredSpatialLayer{ -1 };
 		int16_t preferredTemporalLayer{ -1 };
 		int16_t provisionalTargetSpatialLayer{ -1 };

diff --git a/worker/include/RTC/SvcConsumer.hpp b/worker/include/RTC/SvcConsumer.hpp
@@ -100,7 +100,7 @@ namespace RTC
 		std::vector<RTC::RtpStreamSend*> rtpStreams;
 		RTC::RtpStreamRecv* producerRtpStream{ nullptr };
 		bool syncRequired{ false };
-		RTC::SeqManager<uint16_t> rtpSeqManager;
+		std::unique_ptr<RTC::SeqManager<uint16_t>> rtpSeqManager;
 		int16_t preferredSpatialLayer{ -1 };
 		int16_t preferredTemporalLayer{ -1 };
 		int16_t provisionalTargetSpatialLayer{ -1 };

diff --git a/worker/src/RTC/PipeConsumer.cpp b/worker/src/RTC/PipeConsumer.cpp
@@ -4,6 +4,7 @@
 #include "RTC/PipeConsumer.hpp"
 #include "Logger.hpp"
 #include "MediaSoupErrors.hpp"
+#include "Utils.hpp"
 #include "RTC/Codecs/Tools.hpp"
 
 namespace RTC
@@ -264,7 +265,7 @@ namespace RTC
 		// Packets with only padding are not forwarded.
 		if (packet->GetPayloadLength() == 0)
 		{
-			rtpSeqManager.Drop(packet->GetSequenceNumber());
+			rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 			packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::EMPTY_PAYLOAD);
@@ -284,15 +285,15 @@ namespace RTC
 				MS_DEBUG_TAG(rtp, "sync key frame received");
 			}
 
-			rtpSeqManager.Sync(packet->GetSequenceNumber() - 1);
+			rtpSeqManager->Sync(packet->GetSequenceNumber() - 1);
 
 			syncRequired = false;
 		}
 
 		// Update RTP seq number and timestamp.
 		uint16_t seq;
 
-		rtpSeqManager.Input(packet->GetSequenceNumber(), seq);
+		rtpSeqManager->Input(packet->GetSequenceNumber(), seq);
 
 		// Save original packet fields.
 		auto origSsrc = packet->GetSsrc();
@@ -680,7 +681,15 @@ namespace RTC
 			this->mapMappedSsrcSsrc[consumableEncoding.ssrc] = encoding.ssrc;
 			this->mapSsrcRtpStream[encoding.ssrc]            = rtpStream;
 			this->mapRtpStreamSyncRequired[rtpStream]        = false;
-			this->mapRtpStreamRtpSeqManager[rtpStream];
+
+			// Let's choose an initial output seq number between 1000 and 32768 to avoid
+			// libsrtp bug:
+			// https://github.com/versatica/mediasoup/issues/1437
+			const uint16_t initialOutputSeq =
+			  Utils::Crypto::GetRandomUInt(1000u, std::numeric_limits<uint16_t>::max() / 2);
+
+			this->mapRtpStreamRtpSeqManager[rtpStream].reset(
+			  new RTC::SeqManager<uint16_t>(initialOutputSeq));
 		}
 	}
 

diff --git a/worker/src/RTC/SeqManager.cpp b/worker/src/RTC/SeqManager.cpp
@@ -39,9 +39,17 @@ namespace RTC
 		return isSeqHigherThan(lhs, rhs);
 	}
 
+	template<typename T, uint8_t N>
+	SeqManager<T, N>::SeqManager(T initialOutput) : initialOutput(initialOutput)
+	{
+		MS_TRACE();
+	}
+
 	template<typename T, uint8_t N>
 	void SeqManager<T, N>::Sync(T input)
 	{
+		MS_TRACE();
+
 		// Update base.
 		this->base = (this->maxOutput - input) & MaxValue;
 
@@ -55,6 +63,8 @@ namespace RTC
 	template<typename T, uint8_t N>
 	void SeqManager<T, N>::Drop(T input)
 	{
+		MS_TRACE();
+
 		// Mark as dropped if 'input' is higher than anyone already processed.
 		if (SeqManager<T, N>::IsSeqHigherThan(input, this->maxInput))
 		{
@@ -71,6 +81,8 @@ namespace RTC
 	template<typename T, uint8_t N>
 	bool SeqManager<T, N>::Input(T input, T& output)
 	{
+		MS_TRACE();
+
 		auto base = this->base;
 
 		// No dropped inputs to consider.
@@ -140,6 +152,8 @@ namespace RTC
 			}
 		}
 
+		output = (output + this->initialOutput) & MaxValue;
+
 		return true;
 	}
 
@@ -162,6 +176,8 @@ namespace RTC
 	template<typename T, uint8_t N>
 	void SeqManager<T, N>::ClearDropped()
 	{
+		MS_TRACE();
+
 		// Cleanup dropped values.
 		if (this->dropped.empty())
 		{

diff --git a/worker/src/RTC/SimpleConsumer.cpp b/worker/src/RTC/SimpleConsumer.cpp
@@ -5,6 +5,7 @@
 #include "DepLibUV.hpp"
 #include "Logger.hpp"
 #include "MediaSoupErrors.hpp"
+#include "Utils.hpp"
 #include "RTC/Codecs/Tools.hpp"
 #include "RTC/SimpleConsumer.hpp"
 
@@ -36,6 +37,14 @@ namespace RTC
 		// Create RtpStreamSend instance for sending a single stream to the remote.
 		CreateRtpStream();
 
+		// Let's chosee an initial output seq number between 1000 and 32768 to avoid
+		// libsrtp bug:
+		// https://github.com/versatica/mediasoup/issues/1437
+		const uint16_t initialOutputSeq =
+		  Utils::Crypto::GetRandomUInt(1000u, std::numeric_limits<uint16_t>::max() / 2);
+
+		this->rtpSeqManager.reset(new RTC::SeqManager<uint16_t>(initialOutputSeq));
+
 		// Create the encoding context for Opus.
 		if (
 		  mediaCodec->mimeType.type == RTC::RtpCodecMimeType::Type::AUDIO &&
@@ -340,7 +349,7 @@ namespace RTC
 			  packet->GetSequenceNumber(),
 			  packet->GetTimestamp());
 
-			this->rtpSeqManager.Drop(packet->GetSequenceNumber());
+			this->rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 			packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::DROPPED_BY_CODEC);
@@ -363,7 +372,7 @@ namespace RTC
 		// Packets with only padding are not forwarded.
 		if (packet->GetPayloadLength() == 0)
 		{
-			this->rtpSeqManager.Drop(packet->GetSequenceNumber());
+			this->rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 			packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::EMPTY_PAYLOAD);
@@ -383,15 +392,15 @@ namespace RTC
 				MS_DEBUG_TAG(rtp, "sync key frame received");
 			}
 
-			this->rtpSeqManager.Sync(packet->GetSequenceNumber() - 1);
+			this->rtpSeqManager->Sync(packet->GetSequenceNumber() - 1);
 
 			this->syncRequired = false;
 		}
 
 		// Update RTP seq number and timestamp.
 		uint16_t seq;
 
-		this->rtpSeqManager.Input(packet->GetSequenceNumber(), seq);
+		this->rtpSeqManager->Input(packet->GetSequenceNumber(), seq);
 
 		// Save original packet fields.
 		auto origSsrc = packet->GetSsrc();

diff --git a/worker/src/RTC/SimulcastConsumer.cpp b/worker/src/RTC/SimulcastConsumer.cpp
@@ -5,6 +5,7 @@
 #include "DepLibUV.hpp"
 #include "Logger.hpp"
 #include "MediaSoupErrors.hpp"
+#include "Utils.hpp"
 #include "RTC/Codecs/Tools.hpp"
 
 namespace RTC
@@ -98,6 +99,14 @@ namespace RTC
 			  "%s codec not supported for simulcast", mediaCodec->mimeType.ToString().c_str());
 		}
 
+		// Let's chosee an initial output seq number between 1000 and 32768 to avoid
+		// libsrtp bug:
+		// https://github.com/versatica/mediasoup/issues/1437
+		const uint16_t initialOutputSeq =
+		  Utils::Crypto::GetRandomUInt(1000u, std::numeric_limits<uint16_t>::max() / 2);
+
+		this->rtpSeqManager.reset(new RTC::SeqManager<uint16_t>(initialOutputSeq));
+
 		RTC::Codecs::EncodingContext::Params params;
 
 		params.spatialLayers  = encoding.spatialLayers;
@@ -802,7 +811,7 @@ namespace RTC
 		// not have payload other than padding, then drop it.
 		if (spatialLayer == this->currentSpatialLayer && packet->GetPayloadLength() == 0)
 		{
-			this->rtpSeqManager.Drop(packet->GetSequenceNumber());
+			this->rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 			packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::EMPTY_PAYLOAD);
@@ -951,7 +960,7 @@ namespace RTC
 			// 'packet->GetSequenceNumber() -2' may increase SeqManager::base and
 			// increase the output sequence number.
 			// https://github.com/versatica/mediasoup/issues/408
-			this->rtpSeqManager.Sync(packet->GetSequenceNumber() - (this->lastSentPacketHasMarker ? 1 : 2));
+			this->rtpSeqManager->Sync(packet->GetSequenceNumber() - (this->lastSentPacketHasMarker ? 1 : 2));
 
 			this->encodingContext->SyncRequired();
 
@@ -1013,7 +1022,7 @@ namespace RTC
 			// Rewrite payload if needed. Drop packet if necessary.
 			if (!packet->ProcessPayload(this->encodingContext.get(), marker))
 			{
-				this->rtpSeqManager.Drop(packet->GetSequenceNumber());
+				this->rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 				packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::DROPPED_BY_CODEC);
@@ -1032,7 +1041,7 @@ namespace RTC
 		uint16_t seq;
 		const uint32_t timestamp = packet->GetTimestamp() - this->tsOffset;
 
-		this->rtpSeqManager.Input(packet->GetSequenceNumber(), seq);
+		this->rtpSeqManager->Input(packet->GetSequenceNumber(), seq);
 
 		// Save original packet fields.
 		auto origSsrc      = packet->GetSsrc();
@@ -1066,7 +1075,7 @@ namespace RTC
 		// Process the packet.
 		if (this->rtpStream->ReceivePacket(packet, sharedPacket))
 		{
-			if (this->rtpSeqManager.GetMaxOutput() == packet->GetSequenceNumber())
+			if (this->rtpSeqManager->GetMaxOutput() == packet->GetSequenceNumber())
 			{
 				this->lastSentPacketHasMarker = packet->HasMarker();
 			}

diff --git a/worker/src/RTC/SvcConsumer.cpp b/worker/src/RTC/SvcConsumer.cpp
@@ -5,6 +5,7 @@
 #include "DepLibUV.hpp"
 #include "Logger.hpp"
 #include "MediaSoupErrors.hpp"
+#include "Utils.hpp"
 #include "RTC/Codecs/Tools.hpp"
 
 namespace RTC
@@ -79,6 +80,14 @@ namespace RTC
 			MS_THROW_TYPE_ERROR("%s codec not supported for svc", mediaCodec->mimeType.ToString().c_str());
 		}
 
+		// Let's chosee an initial output seq number between 1000 and 32768 to avoid
+		// libsrtp bug:
+		// https://github.com/versatica/mediasoup/issues/1437
+		const uint16_t initialOutputSeq =
+		  Utils::Crypto::GetRandomUInt(1000u, std::numeric_limits<uint16_t>::max() / 2);
+
+		this->rtpSeqManager.reset(new RTC::SeqManager<uint16_t>(initialOutputSeq));
+
 		RTC::Codecs::EncodingContext::Params params;
 
 		params.spatialLayers  = encoding.spatialLayers;
@@ -681,7 +690,7 @@ namespace RTC
 		// Packets with only padding are not forwarded.
 		if (packet->GetPayloadLength() == 0)
 		{
-			this->rtpSeqManager.Drop(packet->GetSequenceNumber());
+			this->rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 			packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::EMPTY_PAYLOAD);
@@ -701,7 +710,7 @@ namespace RTC
 				MS_DEBUG_TAG(rtp, "sync key frame received");
 			}
 
-			this->rtpSeqManager.Sync(packet->GetSequenceNumber() - 1);
+			this->rtpSeqManager->Sync(packet->GetSequenceNumber() - 1);
 			this->encodingContext->SyncRequired();
 
 			this->syncRequired = false;
@@ -715,7 +724,7 @@ namespace RTC
 
 		if (!packet->ProcessPayload(this->encodingContext.get(), marker))
 		{
-			this->rtpSeqManager.Drop(packet->GetSequenceNumber());
+			this->rtpSeqManager->Drop(packet->GetSequenceNumber());
 
 #ifdef MS_RTC_LOGGER_RTP
 			packet->logger.Dropped(RtcLogger::RtpPacket::DropReason::DROPPED_BY_CODEC);
@@ -738,7 +747,7 @@ namespace RTC
 		// Update RTP seq number and timestamp based on NTP offset.
 		uint16_t seq;
 
-		this->rtpSeqManager.Input(packet->GetSequenceNumber(), seq);
+		this->rtpSeqManager->Input(packet->GetSequenceNumber(), seq);
 
 		// Save original packet fields.
 		auto origSsrc = packet->GetSsrc();