Linux Kernel Architecture

Chapter 2: Process Management and Scheduling

InterpretingBinary Formats

Each binary format is represented in the Linux kernel by an instance of the following (simplified) data

structure:

<binfmts.h>

struct linux_binfmt {

struct linux_binfmt * next;

struct module *module;

int (*load_binary)(struct linux_binprm *, struct pt_regs * regs);

int (*load_shlib)(struct file *);

int (*core_dump)(long signr, struct pt_regs * regs, struct file * file);

unsigned long min_coredump; /* minimal dump size */

};

Each binary format must provide three functions:

1. load_binaryto load normal programs.


2. load_shlibto load ashared library, that is, a dynamic library.


3. core_dumpto write a core dump if there is a program error. This dump can subsequently be
   analyzed using a debugger (e.g.,gdb) for troubleshooting purposes.min_coredumpis a lower
   bound on the core file size from which a coredump will be generated (usually, this is the size
   of a single memory page).

Each binary format must first be registered in the kernel usingregister_binfmt. The purpose of this

function is to add a new binary format to a linked list whose list head is represented by theformats

global variable fromfs/exec.c.Thelinux_binfmtinstances are linked with each other by means of

theirnextelement.

2.4.4 Exiting Processes

Processes must terminate with theexitsystem call. This gives the kernel the opportunity to free the

resources used by the processes to the system.^19 The entry point for this call is thesys_exitfunction

that requires an error code as its parameter in orderto exit the process. Its definition is architecture-

independent and is held inkernel/exit.c. Its implementation is not particularly interesting because it

immediately delegates its work todo_exit.

Suffice it to say that the implementation of this function consists essentially of decrementing reference

counters and returning memory areas to memory management once the reference counter has reverted

to 0 and the corresponding structure is no longer being used by any process in the system.

2.5 Implementation of the Scheduler

A unique description of each process is held in memory and is linked with other processes by means of

several structures. This is the situation facing the scheduler, whose task is to share CPU time between

the programs to create the illusion of concurrent execution. As discussed above, this task is split into two

different parts — one relating to the scheduling policy and the other to context switching.

(^19) exitcan be called explicitly by the programmer. However, the compiler automatically adds a corresponding call to the end of
themainfunction (or to the main function used by the particular language).