Code Wrangling at its Best
In Chapter 30, Toyix gained its first user-mode shell: But the shell could not launch programs yet. The kernel monitor could run programs: but the user shell could not. This chapter adds the first user-facing process-control syscalls: At first, SYS_EXEC still launches programs from the embedded program registry. We are not loading from a filesystem
Continue Reading “Chapter 31 — SYS_EXEC, SYS_WAITPID, and Shell-Launched Programs”
In Chapter 29, we added the first shared userland runtime. This chapter extends that runtime with formatted output and shell-oriented helpers, then uses it to build the first real long-running user-mode program: This first shell will run entirely in ring 3. It will read lines from stdin, parse commands, and execute a few built-in commands.
In Chapter 28, we cleaned up the user-program build system: now automatically builds: But our user programs still duplicate small helper functions: Both demo.c and counter.c contain nearly the same code. This chapter adds the first tiny Toyix userland support library: This is not a real libc yet, but it is the beginning of one.
In Chapter 27, we added a second user program: That proved the program registry can handle more than one embedded ELF. But the Makefile now has duplicated build rules: That will not scale. This chapter cleans up the user-program build system so adding a new user program is mostly: and then adding: The build system
Continue Reading “Chapter 28 — Pattern-Based User Program Build System”
In Chapter 26, we added: That gave us a real background process path. But our only user program, demo, waits for terminal input: That makes it awkward for runbg, because a background process can compete with the monitor for keyboard input. This chapter adds a second embedded user program: counter does not read from stdin.
Continue Reading “Chapter 27 — A Second User Program for Safe Background Execution”
In Chapter 25, the monitor gained an embedded program registry and a foreground run command: That gave us an exec-like path, but only in foreground mode: This chapter adds the next process-management layer: After this chapter, the monitor will support: The kernel still does not have job control, signals, or a filesystem, but it now
Continue Reading “Chapter 26 — Process Table, ps, runbg, and wait PID”
In Chapter 23, we crossed another major boundary: But the user program still started like this: That is not how we want user programs to look long-term. This chapter adds a real initial user stack with: Then crt0.S will pass those arguments to: This moves us closer to a normal process startup model. ELF process
Continue Reading “Chapter 24 — User argc / argv and a Real Initial Stack”
In Chapter 19, we gave each process its own address space: That was a major architectural milestone. But our user program is still not really “loaded.” We are still building a TOYEXE image inside process.c and handing it to the loader. This chapter adds a small executable format before jumping to ELF. We will call
Continue Reading “Chapter 20 — A Tiny Executable Format and User Program Loader”
In Chapter 18, user programs gained a much more realistic syscall interface: But all user processes still shared the same page directory. That means every process saw the same user mappings: This chapter gives each process its own address space. The new model becomes: When the scheduler switches between processes, it will also switch CR3.
Continue Reading “Chapter 19 — Per-Process Address Spaces and CR3 Switching”
In Chapter 17, we added the first minimal process object: That was a big step, but the syscall interface was still a little too artificial. This chapter makes user programs feel more like real programs by adding a tiny file-descriptor-style interface: We will update SYS_WRITE to use a file descriptor, add SYS_READ, and run a
Continue Reading “Chapter 18 — File-Descriptor Syscalls and a Tiny User-Mode Console Program”
Recent Comments