GMP scheduling model of golang

I. Briefly describe the GMP scheduling model of go language

G: A G represents a goroutine. The essence of a coroutine is a user-mode thread, and the user has control over it. It occupies less memory and has low switching costs.
M: Kernel state thread, an M represents a kernel thread, which is equivalent to a system thread. All Gs must be placed on M to run.
P: A logical processor, which can carry several Gs, and make these Gs dock with M in a timely manner. A P represents the context required by M.The number of P depends on the set GOMAXPROCS. The new version of go uses the maximum number of cores by default. For example, if you have an 8-core processor, then the number of P is 8

The number of M does not necessarily match the P. You can set a lot of M, and the M and P can be bound to run, and the excess M is in the dormant state.
The system will find G from the global queue and assign it to an idle M through the scheduler. P will select an M to run. The number of M and G varies. P will have a local queue to represent the unprocessed G and M itself.There is a G being processed, and each time M finishes processing a G, it takes a G from the local queue and updates the schedtick field of P. If there is no G in the local queue, the number of G/P is taken from the global queue at one time.G, if there is nothing in the global queue, take half of it from other P's local queues.

Second, the goroutine of golang: if the Goroutine blocks, will the corresponding M also block?

In the following cases:

Goroutine is blocked due to atomic, mutex or channel operation calls
G blocking in this case will not block kernel thread M, so it will not cause context switching of M to involve G switching

Goroutine is blocked due to network request file IO, time.sleep, and ticker timer operations
The G blocking in this case will not block the kernel thread M, so it will not cause the context switch of M to involve the switch of G

Goroutine blocking due to calling the blocked system
This situation will cause M to block, the kernel will switch M, and the P associated with M will not wait for the blocking of M, which means that the current under PAll G cannot be executed, so at this time, P will be disassociated from the current M, and will be associated with another kernel thread M2. M2 may be a newly created kernel thread, or a previously idle kernel thread may be awakened to execute P'sG.
When M's system call blocking ends, the G will try to obtain an idle P for execution and put it into the P's local queue.If P cannot be obtained, then the thread M becomes dormant, joins the idle thread, and then the G is put into the global queue.

Third, how to block a Goroutine

Method 1: Receive data from a channel that does not send data
Method 2: Send data to a channel that does not receive data
Method 3: Receive data from an empty channel
Method 4: Send data toSend data in an empty channel
Method 5: Use select