CVE-2026-76850

LMDeploy deserializes disaggregated-serving peer messages with pickle. The handlezmqrecv coroutine in lmdeploy/pytorch/disagg/conn/engineconn.py reads peer-to-peer cache-free requests with recvpyobj(), which deserializes the received bytes with pickle.loads(), and the isinstance check against DistServeCacheFreeRequest runs only after deserialization has already completed. The peer that supplies those bytes is caller-controlled: p2pconnect passes remoteengineendpointinfo.zmqaddress from the request body to connect() on the ZMQ PULL socket, and the POST /distserve/p2pinitialize and /distserve/p2pconnect endpoints in lmdeploy/serve/openai/apiserver.py apply no authentication unless the server is started with api_keys, which defaults to None. A remote attacker can direct an engine to pull from a ZMQ endpoint under their control and execute arbitrary code in the engine process. Deployments that do not enable disaggregated serving are not affected, because the receive loop is only started once the migration backend accepts the connection.