Disclosed are a method, a device, and a computer readable storage medium for detecting and processing a system fault. The method includes: an interrupt service routine sending a first stage kicking dog signal, and receiving a second stage kicking dog signal for a system detection task (S101); and when task dead loop or task abnormity is detected, performing system abnormity processing according to a preset processing policy, wherein when the interrupt service routine fails to receive the second stage kicking dog signal within a set period of time, the interrupt service routine stops sending the first stage kicking dog signal, and the system reboots (S102).