Go Compiler Bug Story

When I read this article, another bug tracing story, I found I really like to follow these kinds of stories. Because I enjoy seeing the progress, and moreover I can understand what they are doing, and I also would do the same things if I were them. Summaries are:

• Very rare kernel panic happened, but initially pretty hard to find any pattern
• They cross checked Go issue tracking and found same issues
• More and more cases accumulated, and finally they see the pattern
• Netlink.Receive was the routine causing the panic
• But code review brought nothing, so they go deeper, in to disassembling
• And found suspicious Splitted Operation for Stack Pointer
• During modifying the SP, if  that routine preempted, the SP is not correct
• Async preemption, as it is preempted because it is entering the I/O
• This SP corruption is happening during unwinding the stack, but garbage collector requires the correct SP even during unwinding