Demystifying the Linux CPU Scheduler

# Demystifying the Linux CPU Scheduler ## 1. Basic of the Linux Kernel ### 1.1 Operating Systems Page 2 :::danger :radioactive_sign: one could start to wonder ~~what is the actual responsibility of operating systems~~. :+1: one could start to wonder **what the actual responsibility of operating systems is**. ::: :::warning :-1: we first load in the OS kernel at boot time and then bring in the application software afterward. :+1: **We should have the OS kernel loaded into the DRAM at boot time (by boot loader), and make CPU jump to the first instruction of it for execution (arch/arm64/kernel/head.S). Finally, the kernel will bring the application software to the world (EL1 --> EL0).** ::: :::warning :-1: In fact, to ensure reliable interactions between hardware ... :+1: In fact, to ensure **the** reliable interactions between hardware ... ::: :::warning :-1: The concept of computation isolation, which serves to prevent ~~undesired~~ interactions while still supporting necessary interactions, has a long-standing heritage. :+1: **The concept of computation isolation has a long-standing heritage in supporting necessary interactions while preventing *undesirable* ones.** ::: Page 3 :::warning :-1: The kernel has to fulfill the requests for resources in an exact order to provide an abstraction for cooperation from multiple upper-layer applications, exploiting the computing resources. :+1: **The kernel has to fulfill the requests for the resources in an exact order by exploiting the computing resources to provide an abstraction for the cooperation from multiple upper-layer applications.** ::: <br/> ### 1.2 A general overview Page 3 :::warning :-1: it is necessary to grant the kernel the full access privilege to all hardware resources ... :+1: it is necessary to grant **kernel the privilege of the full access** to all hardware resources ... ::: :::warning :-1: the operating system seeks a protection mechanism to fulfill the stability and security needs of the whole system, separating the design into the kernel and user application codes. :+1: the operating system seeks a protection mechanism **by separating *the codes between kernel design and user application* to fulfill the stability and security needs *for* the whole system**. ::: :::warning :-1: While, in practice, ... :+1: **In practice,** depending on the configuration settings at compile time, ... ::: :::warning :-1: Such a memory partitioning mechanism ensures that the kernel data and user data do not interfere with each other. :+1: Such a memory partitioning mechanism ensures that **the data between kernel space and user space** do not interfere with each other. ::: Page 4 :::warning :-1: Also, it is a security measure to protect the entire system from a malfunctioning or malicious user program. :+1: **Moreover**, it is a security measure ... ::: #### 1.2.1 System calls Page 6 :::warning :-1: The process execution is suspended and the control passes to the kernel (kernel mode), which will look up the entry 128 in the interrupt vector table. :+1: The process execution is suspended and the control passes to the kernel (kernel mode) **for looking up the entry 128 in the interrupt vector table**. ::: #### 1.2.2 File systems Page 7 :::warning :-1: Nevertheless, a limited number of writable files exist, which permit modifications to the kernel variables. :+1: **While** a substantial portion of the files within this filesystem exhibit read-only properties, **a limited number of writable files exist for kernel modifications.** ::: :::warning :-1: The directory /proc/sys houses several writable files that corresponds to kernel variables, allowing for direct modification by writing to them or through utilization of the sysctl utility. :+1: The directory /proc/sys houses several writable files **for** kernel variables, **and direct modification is allowed either by writing to them or through the sysctl utility.** ::: #### 1.2.3 A different kind of software Page 8 :::warning :-1: it is shown an example of code where the panic() rotine is called. :+1: **An exmple of code is shown for how the panic() routine is called.** ::: :::danger :radioactive_sign: printk() is also able to reliably handle~~s~~ concurrent writes from different cores,... :+1: printk() is also able to reliably **handle** concurrent writes from different cores,... ::: Page 10 :::warning :-1: the answer is straightforward: first, to use a small amount of kernel memory, and secondly, because of fragmentation. :+1: the answer is straightforward: **one reason is to use a small amount of kernel memory, and the other is to avoid fragmentation.** ::: Page 11 :::danger :radioactive_sign: , and communication between the servers, applications, and the kernel ~~rely~~ on message passing. :+1: , and **the** communication **among** the servers, applications, and the kernel **relies** on message passing. ::: ## 2. The Linux CPU scheduler ## 3. ## 4. ## 5. ## 6. Tracing with ftrace Page 176 :::warning :-1: Also, it is a great educational tool, not just to peek at what happends in the kernel, but also to help approach the source code by observing the function flow. :+1: ::: :::warning :-1: To understand what they do and why they are useful, it is necessary to understand ftrace, which is the tool that uses them. :+1: ::: ### 6.1 How does it work? ### 6.2 Enable the kernel tracing infrastructure ### 6.3 Interfacing with ftace page 180 :::danger :radioactive_sign: Tracing necessitates an extremely lightweight mechanism, as it is ~~common~~ used to monitor events that occur frequently. :+1: Tracing necessitates an extremely lightweight mechanism, as it is **commonly** used to monitor events that occur frequently. ::: :::warning :-1: the ftrace output, which is produced from the kernel, is read from user space. :+1: **the ftace output produced in the kernel is read from user space**. ::: :::warning :-1: tracefs is a specialized filesystem that resides in another filesystem sysfs, which contains files and directories to interact with ftrace. :+1: tracefs is a specialized filesystem that resides in another filesystem sysfs, **and it contains files and directories to interact with ftrace**. ::: :::warning :-1: The purpose of some of these files is not to set options. Rather, it is to list available options. :+1: **Instead of setting options to those files, some of those files are only to list available options.** ::: ### 6.4 Ftrace usage #### 6.4.1 Function tracing