A data processing method includes the following operations: separately performing convolutional interleaving on n lane data streams to obtain n first data streams, where n>1, and first FEC encoding is performed on all the n lane data streams. Every a codewords obtained through the first FEC encoding are distributed in b lane data streams, where a≤b≤n, and a≥1. z consecutive symbols in each of the first data streams are from z different codewords, and z>1. Then, every K first data streams of the n first data streams are multiplexed to obtain one second data stream, so as to obtain a total of m second data streams. The n first data streams include G first data stream subsets, symbols in different first data stream subsets are from different codewords, m=n/K, K>1, and G>1. y consecutive symbols in each second data stream are from y different codewords, where y>z.