From bec66e9e69fe483a66ecfa5ad8189ebabaad1a9a Mon Sep 17 00:00:00 2001 From: H1JK Date: Mon, 16 Jan 2023 11:42:10 +0800 Subject: [PATCH] adjust: Improve WebSocket mask --- common/net/websocket.go | 131 +++++++++++++++++++++++++++++++++++ transport/vmess/websocket.go | 4 +- 2 files changed, 133 insertions(+), 2 deletions(-) create mode 100644 common/net/websocket.go diff --git a/common/net/websocket.go b/common/net/websocket.go new file mode 100644 index 00000000..b002310a --- /dev/null +++ b/common/net/websocket.go @@ -0,0 +1,131 @@ +package net + +import ( + "encoding/binary" + "math/bits" +) + +// kanged from https://github.com/nhooyr/websocket/blob/master/frame.go +// License: MIT + +// MaskWebSocket applies the WebSocket masking algorithm to p +// with the given key. +// See https://tools.ietf.org/html/rfc6455#section-5.3 +// +// The returned value is the correctly rotated key to +// to continue to mask/unmask the message. +// +// It is optimized for LittleEndian and expects the key +// to be in little endian. +// +// See https://github.com/golang/go/issues/31586 +func MaskWebSocket(key uint32, b []byte) uint32 { + if len(b) >= 8 { + key64 := uint64(key)<<32 | uint64(key) + + // At some point in the future we can clean these unrolled loops up. + // See https://github.com/golang/go/issues/31586#issuecomment-487436401 + + // Then we xor until b is less than 128 bytes. + for len(b) >= 128 { + v := binary.LittleEndian.Uint64(b) + binary.LittleEndian.PutUint64(b, v^key64) + v = binary.LittleEndian.Uint64(b[8:16]) + binary.LittleEndian.PutUint64(b[8:16], v^key64) + v = binary.LittleEndian.Uint64(b[16:24]) + binary.LittleEndian.PutUint64(b[16:24], v^key64) + v = binary.LittleEndian.Uint64(b[24:32]) + binary.LittleEndian.PutUint64(b[24:32], v^key64) + v = binary.LittleEndian.Uint64(b[32:40]) + binary.LittleEndian.PutUint64(b[32:40], v^key64) + v = binary.LittleEndian.Uint64(b[40:48]) + binary.LittleEndian.PutUint64(b[40:48], v^key64) + v = binary.LittleEndian.Uint64(b[48:56]) + binary.LittleEndian.PutUint64(b[48:56], v^key64) + v = binary.LittleEndian.Uint64(b[56:64]) + binary.LittleEndian.PutUint64(b[56:64], v^key64) + v = binary.LittleEndian.Uint64(b[64:72]) + binary.LittleEndian.PutUint64(b[64:72], v^key64) + v = binary.LittleEndian.Uint64(b[72:80]) + binary.LittleEndian.PutUint64(b[72:80], v^key64) + v = binary.LittleEndian.Uint64(b[80:88]) + binary.LittleEndian.PutUint64(b[80:88], v^key64) + v = binary.LittleEndian.Uint64(b[88:96]) + binary.LittleEndian.PutUint64(b[88:96], v^key64) + v = binary.LittleEndian.Uint64(b[96:104]) + binary.LittleEndian.PutUint64(b[96:104], v^key64) + v = binary.LittleEndian.Uint64(b[104:112]) + binary.LittleEndian.PutUint64(b[104:112], v^key64) + v = binary.LittleEndian.Uint64(b[112:120]) + binary.LittleEndian.PutUint64(b[112:120], v^key64) + v = binary.LittleEndian.Uint64(b[120:128]) + binary.LittleEndian.PutUint64(b[120:128], v^key64) + b = b[128:] + } + + // Then we xor until b is less than 64 bytes. + for len(b) >= 64 { + v := binary.LittleEndian.Uint64(b) + binary.LittleEndian.PutUint64(b, v^key64) + v = binary.LittleEndian.Uint64(b[8:16]) + binary.LittleEndian.PutUint64(b[8:16], v^key64) + v = binary.LittleEndian.Uint64(b[16:24]) + binary.LittleEndian.PutUint64(b[16:24], v^key64) + v = binary.LittleEndian.Uint64(b[24:32]) + binary.LittleEndian.PutUint64(b[24:32], v^key64) + v = binary.LittleEndian.Uint64(b[32:40]) + binary.LittleEndian.PutUint64(b[32:40], v^key64) + v = binary.LittleEndian.Uint64(b[40:48]) + binary.LittleEndian.PutUint64(b[40:48], v^key64) + v = binary.LittleEndian.Uint64(b[48:56]) + binary.LittleEndian.PutUint64(b[48:56], v^key64) + v = binary.LittleEndian.Uint64(b[56:64]) + binary.LittleEndian.PutUint64(b[56:64], v^key64) + b = b[64:] + } + + // Then we xor until b is less than 32 bytes. + for len(b) >= 32 { + v := binary.LittleEndian.Uint64(b) + binary.LittleEndian.PutUint64(b, v^key64) + v = binary.LittleEndian.Uint64(b[8:16]) + binary.LittleEndian.PutUint64(b[8:16], v^key64) + v = binary.LittleEndian.Uint64(b[16:24]) + binary.LittleEndian.PutUint64(b[16:24], v^key64) + v = binary.LittleEndian.Uint64(b[24:32]) + binary.LittleEndian.PutUint64(b[24:32], v^key64) + b = b[32:] + } + + // Then we xor until b is less than 16 bytes. + for len(b) >= 16 { + v := binary.LittleEndian.Uint64(b) + binary.LittleEndian.PutUint64(b, v^key64) + v = binary.LittleEndian.Uint64(b[8:16]) + binary.LittleEndian.PutUint64(b[8:16], v^key64) + b = b[16:] + } + + // Then we xor until b is less than 8 bytes. + for len(b) >= 8 { + v := binary.LittleEndian.Uint64(b) + binary.LittleEndian.PutUint64(b, v^key64) + b = b[8:] + } + } + + // Then we xor until b is less than 4 bytes. + for len(b) >= 4 { + v := binary.LittleEndian.Uint32(b) + binary.LittleEndian.PutUint32(b, v^key) + b = b[4:] + } + + // xor remaining bytes. + for i := range b { + b[i] ^= byte(key) + key = bits.RotateLeft32(key, -8) + } + + return key +} diff --git a/transport/vmess/websocket.go b/transport/vmess/websocket.go index ce5615d5..90cd7772 100644 --- a/transport/vmess/websocket.go +++ b/transport/vmess/websocket.go @@ -123,8 +123,8 @@ func (wsc *websocketConn) WriteBuffer(buffer *buf.Buffer) error { } maskKey := rand.Uint32() - binary.BigEndian.PutUint32(header[1+payloadBitLength:], maskKey) - maskBytes(*(*[4]byte)(header[1+payloadBitLength:]), 0, data) + binary.LittleEndian.PutUint32(header[1+payloadBitLength:], maskKey) + N.MaskWebSocket(maskKey, data) wsc.wMux.Lock() defer wsc.wMux.Unlock()