The Console
Executing a program starts with the user typing a command in the console/shell. According to the comet book (Operating Systems – Three easy Pieces) the shell is a normal user mode program.
It will by some means find the executable binary file that implements the command. It calls the kernel function fork() to copy itself. It will then call the kernel function exec() to replace the code segment of the copy with the code segment read from the executable program that implements the command. It will the let the new forked process run and wait on its termination. It will then show the commands return value on the console.
No console
To make things easier during development, interaction with the operating system is not needed and the console can be left out. Identifying a program binary to execute, calling fork() and exec() can just be done in the kernel_main() method. The console can be left out of the picture during early developments.
Under Linux, the first process that is started is the init process. A first process could be started from the kernel_main() method similar to init.
Why fork() and exec()?
The question is could an application not be started by creation of a completely new process without forking an existing process and without using exec()? The comet book says, the approach of fork() and exec() is just the right thing to do. Is that true or not? Could it be done easier?