DOOM target + launcher code

Reviewed-on: #20

Makefile

@@ -1,3 +1,5 @@
+USER_PROGRAMS := pedicel.raw apex.raw doom.raw
+USER_FILES := wow.txt doom1.wad
 
 BUILDDIR := build
 ELFFILE := pepperk
@@ -16,12 +18,11 @@ endif
 OBJFILES := $(patsubst $(SRC)/%.c, $(BUILDDIR)/%.o, $(SOURCES))
 
 CC := x86_64-elf-gcc
-CC_FLAGS=-Wall -Wextra -std=gnu99 -nostdlib -ffreestanding -fstack-protector -fno-omit-frame-pointer -fno-stack-check -fno-PIC -ffunction-sections -fdata-sections -mcmodel=kernel
+CC_FLAGS=-Wall -Wextra -std=gnu99 -nostdlib -ffreestanding -fstack-protector -fno-omit-frame-pointer -fno-stack-check -fno-PIC -ffunction-sections -fdata-sections -mcmodel=kernel -mno-red-zone
 
 LD := x86_64-elf-ld
 
 $(ELFFILE): $(BUILDDIR) $(OBJFILES)
-	nasm -f bin user/hello.S -o $(BUILDDIR)/hello
 	nasm -f elf64 src/arch/x86/idt.S -o $(BUILDDIR)/idt_stub.o
 	$(LD) -o $(ELFFILE) -T linker.ld $(OBJFILES) $(BUILDDIR)/idt_stub.o
 	# Get the symbols for debugging
@@ -42,13 +43,18 @@ limine/limine:
 	git clone https://github.com/limine-bootloader/limine.git --branch=v9.x-binary --depth=1
 	$(MAKE) -C limine
 
+.PHONY: user
+user:
+	$(MAKE) -C user
+	tar cvf $(BUILDDIR)/initfs.tar -C $(BUILDDIR) $(USER_PROGRAMS) -C ../user $(USER_FILES)
+
 build-iso: limine/limine $(ELFFILE)
 	rm -rf iso_root
 	mkdir -p iso_root/boot
 	cp -v $(ELFFILE) iso_root/boot
 	mkdir -p iso_root/boot/limine
 	cp -v limine.conf iso_root/boot/limine
-	cp $(BUILDDIR)/hello iso_root/boot/
+	cp $(BUILDDIR)/initfs.tar iso_root/boot/
 	mkdir -p iso_root/EFI/BOOT
 	cp -v limine/limine-bios.sys limine/limine-bios-cd.bin limine/limine-uefi-cd.bin iso_root/boot/limine/
 	cp -v limine/BOOTX64.EFI iso_root/EFI/BOOT/

README.md

@@ -1,5 +1,5 @@
 # <img width="40" height="40" alt="red-pepper" src="https://i.ibb.co/mrHH6d1m/pixil-frame-0-4.png" /> pepperOS: "will never be done"
-
+<a href="https://ibb.co/KxbfxmFB"><img src="https://i.ibb.co/GQn8QFcG/pepper.png" alt="pepper" border="0"></a>
 ## Description
 
 PepperOS is a 64-bit freely-licensed monolithic kernel for x86 processors, with round-robin preemptive scheduling and 4-level paging. See the [manual](docs/MANUAL.md) for more.
@@ -23,8 +23,9 @@ CC := gcc
 LD := ld
 ```
 
-Then, to compile the kernel and make an ISO image file, run: `make build-iso`
+Then, to compile the kernel and make an ISO image file, run: `make`.  
-To run it with QEMU, do: `make run`
+To build the user programs and the initial filesystem, do `make user`.  
+To run it with QEMU, do: `make run`.  
 
 ## Trying the kernel on real hardware
 
@@ -48,31 +49,9 @@ These features can be activated by setting them to "true" at the end of the make
 make UBSAN=true
 ```
 
-## TODO
+## Writing software for PepperOS
 
-The basics that I'm targeting are:
+If you want to write software for PepperOS, take a look at the [Software Developer's guide](docs/SOFTWARE.md).
-
-### Basic utility of what we call a "kernel"
-
-- Implement tasks, and task switching + context switching and spinlock acquire/release
-- Load an executable
-- Filesystem (TAR for read-only initfs, then maybe read-write using FAT12/16/32 or easier fs) w/ VFS layer
-- Getting to userspace (ring 3 switching, syscall interface)
-- Porting musl libc or equivalent
-
-### Scalability/maintenance/expansion features
-
-- Documentation
-- SOME error handling in functions
-- Unit tests
-- Good error codes (like Linux kernel: ENOMEM, ENOENT, ...)
-
-### Optional features
-
-In the future, maybe?
-- SMP support (Limine provides functionality to make this easier)
-- Parsing the ACPI tables and using them for something
-- Replacing the PIT timer with APIC
 
 ## Thanks
 

docs/MANUAL.md

@@ -25,6 +25,9 @@ The recommended hardware to run PepperOS is the following:
 
 ## b. Features
 
+- Round robin preemptive scheduling
+- Coexistence of ring 0 and ring 3 processes
+
 ## II. Kernel architecture
 
 ### a. Boot process
@@ -37,4 +40,5 @@ The recommended hardware to run PepperOS is the following:
 
 ## III. Syscall table
 
-Not yet implemented.
+The syscall interface in the Pepper kernel uses the System V ABI convention for argument order.
+It vaguely mimics Unix-like systems. You will find it in [SYSCALLS.md](SYSCALLS.md).

docs/SOFTWARE.md

@@ -0,0 +1,72 @@
+# Writing software for PepperOS
+
+## Why would you want to do that?
+
+Honestly I have no idea. Maybe you have too much free time.  
+Keep in mind that the Pepper kernel is a personal project and it's full of bugs, inconsistencies, weird ways of doing things (and I don't care because it's my toy).  
+Now if you still want to write something for this OS, thank you. Follow along.
+
+## Headers available in userspace
+
+Of course, all of the freestanding headers are available:
+- `<float.h>`: macros for floating-point types
+- `<limits.h>`: macros for integer types
+- `<iso646.h>`: macros for bitwise and logical operators
+- `<stdarg.h>`: variadic function support
+- `<stddef.h>`: definitions for `size_t`, `ptrdiff_t`, and others
+- `<stdbool.h>`: definitions for boolean types
+- `<stdint.h>`: definitions for `int_t` and `uint_t` types
+
+Also available is the `<syscall.h>` header that gives access to low-level system call interface, notably the `syscallX` function family, X being the amount of arguments to use.
+
+(TODO: put the other headers here once libc is more complete)
+
+## 1. Write the source code
+
+PepperOS is able to run programs written in x86 assembly, and C programs.
+
+### x86 Assembly
+
+Start your assembly file with the `bits 64` instruction, to emit 64-bit code.  
+You can add sections `.text`, `.data`, `.bss` as you need.  
+The three things to take in consideration here are:
+- PepperOS does not use the `syscall` instruction, instead it uses the old-fashioned `int 0x80` to trigger a system call.  
+- The entry point should be labelled as `_start`.
+- At the end of the file, there should be an exit system call followed by a loop, like so:
+
+```nasm
+.end:
+    mov rax, 0x3C
+    mov rdi, 0x0
+    int 0x80
+
+.loop:
+    jmp .loop
+```
+
+For an example, look at the file [pedicel.S](../user/pedicel.S).
+
+### C program
+
+You will find relevant headers in the `libc` directory. They contain system call wrappers, utility functions, and more. See what's implemented there and what's not.  
+To invoke a system call you can use the functions defined in `libc/syscall.h`.
+
+## 2. Add the Makefile rule and variable
+
+Now that your code is complete, add a Makefile rule to `user/Makefile` with your program name. You can just copy-paste the rule that applies to you (either from an Assembly source or C source) and change the name of the files (.raw, .elf, etc...) in the rule.  
+For clarity, raw binaries have the `.raw` extension, and ELF ones have `.elf`.  
+You also now have to add the name of the executable to the `USER_PROGRAMS` variable at the top of the global Makefile.  
+Finally, do `make user` to compile your program.
+
+## 3. Run your program
+
+You can now boot up PepperOS, in a VM or on real hardware, and use the kernel's shell to `list` files in the filesystem (to see if your executable was properly added), and then, run it with the `load` command. Congratulations, you made a program for a random hobby OS!
+
+## 4. (Optional) debugging
+
+Use GDB with the `make debug` rule!  
+For your information, user programs are loaded at `0x400000`. Can be good to know to set breakpoints.
+
+## 5. (Optional) contribute!
+
+If you like what you've done and you think it could be nice to add it to PepperOS, send it to me by e-mail: `xamidev (at) riseup (dot) net`. It may or may not be added in a future release... who knows?

docs/STYLE.md

@@ -2,6 +2,14 @@
 
 This document describes the coding style for the Pepper kernel. It is used as a guideline across all source files.
 
+## Setting up a language server (optional)
+
+Before you do anything you might want to setup a language server with your editor. This will save you lots of time correcting errors and stuff. I use `clangd`, and generate my `compile_commands.json` like so:
+
+```
+bear -- make
+```
+
 ## Indentation
 
 Indentations should be 4 characters long.

docs/SYSCALLS.md

@@ -0,0 +1,15 @@
+# Pepper kernel system call table
+
+The following table contains all of the system calls supported by PepperOS, as well as their arguments. The explanation for what every system call does is available as a comment above each function in corresponding files in the `src/syscall` folder.
+
+Name | Number (%rax) | arg0 (%rdi) | arg1 (%rsi) | arg2 (%rdx) | arg3 (%r10) |
+|---|---|---|---|---|---|
+| sys_read | 0 | unsigned int fd | char* buf | size_t count | |
+| sys_write | 1 | unsigned int fd | const char* buf | size_t count | |
+| sys_open | 2 | const char* filename | int flags | | |
+| sys_close | 3 | unsigned int fd | | | |
+| sys_lseek | 8 | unsigned int fd | int offset | int whence | |
+| sys_tell | 9 | unsigned int fd | | | |
+| sys_eof | 10 | unsigned int fd | | | |
+| sys_draw | 11 | const uint8_t* src | int width | int height | int channels |
+| sys_exit | 60 | int error_code | | | |

include/config.h

@@ -9,8 +9,8 @@
 
 /* version */
 #define PEPPEROS_VERSION_MAJOR "0"
-#define PEPPEROS_VERSION_MINOR "0"
+#define PEPPEROS_VERSION_MINOR "1"
-#define PEPPEROS_VERSION_PATCH "109"
+#define PEPPEROS_VERSION_PATCH "121"
 #define PEPPEROS_SPLASH \
 "\x1b[38;5;196m                                      \x1b[38;5;231m____  _____\r\n\x1b[0m"\
 "\x1b[38;5;196m    ____  ___  ____  ____  ___  _____\x1b[38;5;231m/ __ \\/ ___/\r\n\x1b[0m"\
@@ -44,6 +44,8 @@
 #define USER_STACK_TOP 0x80000000
 #define USER_STACK_PAGES 16 // 16*4096 = 64kb
 #define USER_CODE_START 0x400000 // like linux
+#define USER_RAW_EXTRA_PAGES 8192 // Extra writable pages after raw image for .bss/heap
+                                 // TODO: throw this away and make an ELF loader instead bruh
 
 /* paging */
 #define PAGING_MAX_PHYS 0x200000000
@@ -63,4 +65,8 @@
 /* ssp */
 #define STACK_CHK_GUARD 0x7ABA5C007ABA5C00
 
+/* fs */
+#define FDT_MAX 8 // Maximum amount of file descriptors per process
+
+
 #endif

include/fs/initfs.h

@@ -0,0 +1,24 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   PS/2 Keyboard driver
+ * @license GPL-3.0-only
+ */
+
+#ifndef INITFS_H
+#define INITFS_H
+
+#include <limine.h>
+
+int initfs_init(struct limine_file* tar_file);
+int tar_lookup(unsigned char* archive, char* filename, char** out);
+int tar_exists(const char* filename);
+int tar_read(char* filename, char* out, int count, int offset);
+void tar_list();
+
+enum Seek {
+    SEEK_SET,
+    SEEK_CUR,
+    SEEK_END
+};
+
+#endif

include/kernel.h

@@ -8,9 +8,15 @@
 #define KERNEL_H
 
 #include "limine.h"
+
+// Not in POSIX order.
 enum ErrorCodes {
-	ENOMEM,
+	ENOMEM, // No memory
-	EIO
+	EIO,    // Input/output error
+	ENOENT, // No entry
+	EBADFD,  // Bad file descriptor
+	EMFILE,  // Too many open files
+	EINVAL  // Invalid argument
 };
 
 #define CLEAR_INTERRUPTS __asm__ volatile("cli")
@@ -46,6 +52,8 @@ void debug_stack_trace(unsigned int max_frames);
 const char* debug_find_symbol(uintptr_t rip, uintptr_t* offset);
 void boot_mem_display(void);
 
+int loader_load_raw();
+
 #define assert(check) do { if(!(check)) hcf(); } while(0)
 
 struct boot_context {

include/sched/process.h

@@ -11,6 +11,7 @@
 #include <config.h>
 #include <stdint.h>
 #include <limine.h>
+#include <stdbool.h>
 
 typedef enum {
     READY,
@@ -18,6 +19,13 @@ typedef enum {
     DEAD
 } status_t;
 
+struct fd {
+    int fd;
+    char filename[PROCESS_NAME_MAX];  // File opened
+    uint64_t cursor; // Cursor position in file
+    bool open;
+};
+
 struct process {
     size_t pid;
     char name[PROCESS_NAME_MAX];
@@ -26,6 +34,10 @@ struct process {
     struct cpu_status* context;
 	void* root_page_table; // Process PML4 (should contain kernel PML4 in higher half [256-511]
     void* kernel_stack; // Used for interrupts (syscall: int 0x80), defines the TSS RSP0
+    
+    struct fd fdt[FDT_MAX]; // File Descriptor Table
+    size_t next_free_fd;
+    
     struct process* next;
 };
 
@@ -38,6 +50,7 @@ void process_exit(void);
 
 void process_display_list(struct process* processes_list);
 
-void process_create_user(struct limine_file* file);
+
+void process_create_user_raw(char* file, int size, char* name);
 
 #endif

include/string/string.h

@@ -13,5 +13,7 @@ char *strcpy(char *dest, const char *src);
 char *strcat(char *dest, const char *src);
 void strncpy(char* dst, const char* src, size_t n);
 int strncmp(const char* s1, const char* s2, size_t n);
+size_t strlen(const char* str);
+int atoi(const char* str);
 
 #endif

limine.conf

@@ -6,4 +6,4 @@ interface_branding: Welcome to the PepperOS disk!
 	
 	comment: Default configuration (warning: spicy)
 	path: boot():/boot/pepperk
-	module_path: boot():/boot/hello
+	module_path: boot():/boot/initfs.tar

src/arch/x86/idt.c

@@ -264,10 +264,9 @@ struct cpu_status* interrupt_dispatch(struct cpu_status* context)
             // Send an EOI so that we can continue having interrupts
             outb(0x20, 0x20);
 
-            // Scheduler is temporarily disabled to test user trampoline
+            if (ticks % SCHEDULER_QUANTUM == 0) {
-            /* if (ticks % SCHEDULER_QUANTUM == 0) {
                 return scheduler_schedule(context);
-            } */
+            }
 
             break;
 

src/arch/x86/syscall.c

@@ -1,62 +0,0 @@
-/*
- * @author  xamidev <xamidev@riseup.net>
- * @brief   System call handling
- * @license GPL-3.0-only
- */
-
-#include "sched/scheduler.h"
-#include <arch/x86.h>
-#include <kernel.h>
-#include <stddef.h>
-#include <io/term/term.h>
-
-void sys_write(unsigned int fd, const char* buf, size_t count)
-{
-    switch (fd) {
-        case 1: //stdout
-            for (size_t i=0; i<count; i++) {
-                internal_putc(buf[i], NULL);
-            }
-            break;
-
-        case 2: //stderr
-            break;
-    }
-}
-
-/*
- * syscall_handler - System call dispatcher
- * @regs: CPU state
- *
- * This function is called from the interrupt dispatcher,
- * when an interrupt 0x80 is emitted from userland.
- *
- * It switches control to the syscall number provided
- * in %rax. 
- *
- * We try to follow the System V convention here:
- * - syscall number in %rax
- * - args in %rdi, %rsi, %rdx, %r10, %r8, %r9
- * - return value (if any) in %rax
- *
- * Return:
- * <regs> - CPU state after system call
- */
-struct cpu_status* syscall_handler(struct cpu_status* regs)
-{
-    DEBUG("Syscall %lx with (arg0=%lx arg1=%lx)", regs->rax, regs->rdi, regs->rsi);
-
-    switch (regs->rax)
-    {
-        case 0: //sys_read
-            break;
-        case 1: //sys_write
-            sys_write(regs->rdi, (char*)regs->rsi, regs->rdx);
-            break;
-        default:
-            regs->rax = 0xbad515ca11;
-            break;
-    }
-
-    return regs;
-}

src/fs/initfs.c

@@ -0,0 +1,147 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   Initial TAR filesystem (read-only)
+ * @license GPL-3.0-only
+ */
+
+#include <sched/process.h>
+#include <limine.h>
+#include <fs/initfs.h>
+#include <kernel.h>
+#include <mem/utils.h>
+#include <string/string.h>
+
+void* archive_start_addr;
+uint64_t archive_size;
+
+/*
+ * tar_oct2bin - convert octal size string to an integer 
+ * @str:  octal size string
+ * @size: size of string
+ *
+ * Return:
+ * $n - file size as an integer
+ */
+int tar_oct2bin(unsigned char* str, int size)
+{
+    int n = 0;
+    unsigned char* c = str;
+    while (size-- > 0) {
+        n *= 8;
+        n += *c - '0';
+        c++;
+    }
+    return n;
+}
+
+/*
+ * tar_lookup - lookup a file in the TAR file
+ * @archive:  pointer to beginning of the archive
+ * @filename: file to lookup (absolute path)
+ * @out:      where to store file data if found
+ *
+ * Return:
+ * $filesize - size of the file, if found 
+ * $-ENOENT  - file not found
+ */
+int tar_lookup(unsigned char* archive, char* filename, char** out)
+{
+    unsigned char *ptr = archive;
+
+    while (!memcmp(ptr + 257, "ustar", 5)) {
+        int filesize = tar_oct2bin(ptr + 0x7c, 11);
+        if (!memcmp(ptr, filename, strlen(filename) + 1)) {
+            *out = (char*)(ptr + 512);
+            return filesize;
+        }
+        ptr += (((filesize + 511) / 512) + 1) * 512;
+    }
+    return -ENOENT;
+}
+
+/*
+ * tar_list - list all files present in archive
+ */
+void tar_list()
+{
+    printf("++ Contents of initial filesystem ++\r\n\r\n");
+    unsigned char *ptr = archive_start_addr;
+
+    while (!memcmp(ptr + 257, "ustar", 5)) {
+        int filesize = tar_oct2bin(ptr + 0x7c, 11);
+        char* filename = (char*)ptr;
+        printf("file: %s\r\n", filename);
+        ptr += (((filesize + 511) / 512) + 1) * 512;
+    }
+}
+
+/*
+ * tar_read - read a file in the TAR archive
+ * @filename: file to read (absolute path)
+ * @out:      where to store file data if found
+ * @count:    amount of bytes to read
+ * @offset:   read from byte offset (0 for none)
+ *
+ * Return:
+ * $filesize - size of the file, if found 
+ * $-ENOENT  - file not found
+ */
+int tar_read(char* filename, char* out, int count, int offset)
+{
+    char* file_data;
+    int filesize = tar_lookup(archive_start_addr, filename, &file_data);
+
+    if (filesize <= 0) {
+        return filesize;
+    }
+
+    if (offset >= filesize) {
+        return -EINVAL;
+    }
+
+    int remaining = filesize - offset;
+    int to_read = remaining < count ? remaining : count;
+
+    memcpy(out, file_data + offset, to_read);
+
+    return to_read;
+}
+
+/*
+ * tar_exists - check if a file exists in the TAR archive
+ * @filename: file to check (absolute path)
+ *
+ * Return:
+ * $filesize - size of the file, if found 
+ * $-ENOENT  - file not found
+ */
+int tar_exists(const char* filename)
+{
+    unsigned char* ptr = archive_start_addr;
+
+    while (!memcmp(ptr + 257, "ustar", 5)) {
+        int filesize = tar_oct2bin(ptr + 0x7c, 11);
+        if (!memcmp(ptr, filename, strlen(filename) + 1)) {
+            return filesize;
+        }
+        ptr += (((filesize + 511) / 512) + 1) * 512;
+    }
+    return -ENOENT;
+}
+
+
+/*
+ * initfs_init - initialize the TAR initial filesystem
+ * @tar_file: pointer to the Limine-loaded archive
+ *
+ * Return:
+ * $0       - on success
+ */
+int initfs_init(struct limine_file* tar_file)
+{
+    archive_start_addr = tar_file->address;
+    archive_size = tar_file->size;
+
+    DEBUG("Loaded TAR initial filesystem (initfs.tar)");
+    return 0;
+}

src/io/term/term.c

@@ -50,9 +50,17 @@ void internal_putc(int c, void *_)
     if (init.terminal) {
         if (panic_count == 0) {
             spinlock_acquire(&term_lock);
+            if (ch == '\n') {
+                char cr = '\r';
+                flanterm_write(ft_ctx, &cr, 1);
+            }
             flanterm_write(ft_ctx, &ch, 1);
             spinlock_release(&term_lock);
         } else {
+            if (ch == '\n') {
+                char cr = '\r';
+                flanterm_write(ft_ctx, &cr, 1);
+            }
             flanterm_write(ft_ctx, &ch, 1);
         }
     }
@@ -82,9 +90,17 @@ void debug_putc(int c, void *_)
     if (init.terminal && (!init.all || panic_count > 0)) {
         if (panic_count == 0) {
             spinlock_acquire(&term_lock);
+            if (ch == '\n') {
+                char cr = '\r';
+                flanterm_write(ft_ctx, &cr, 1);
+            }
             flanterm_write(ft_ctx, &ch, 1);
             spinlock_release(&term_lock);
         } else {
+            if (ch == '\n') {
+                char cr = '\r';
+                flanterm_write(ft_ctx, &cr, 1);
+            }
             flanterm_write(ft_ctx, &ch, 1);
         }
     }

src/kapps/kshell.c

@@ -4,6 +4,7 @@
  * @license GPL-3.0-only
  */
 
+#include "fs/initfs.h"
 #include <io/term/term.h>
 #include <config.h>
 #include <io/kbd/ps2.h>
@@ -12,6 +13,7 @@
 #include <kernel.h>
 #include <time/date.h>
 #include <mem/kheap.h>
+#include <sched/process.h>
 
 __attribute__((noinline))
 void smash_it()
@@ -22,6 +24,52 @@ void smash_it()
     }
 }
 
+extern struct process* processes_list;
+
+void ps()
+{
+    printf("pid\tname\tstatus\n");
+    struct process* curr = processes_list;
+    while (curr != NULL) {
+        char* status;
+        switch (curr->status) {
+            case READY: status = "READY"; break;
+            case RUNNING: status = "RUNNING"; break;
+            case DEAD: status = "DEAD"; break;
+            default: status = "N/A"; break;
+        }
+        printf("%u\t%s\t%s\n", curr->pid, curr->name, status);
+        if (curr->next != NULL) {
+            curr = curr->next;
+        } else {
+            break;
+        }
+    }
+}
+
+void kill()
+{
+    char input_buf[11] = {0};
+    printf("pid> ");
+    keyboard_getline(input_buf, 10);
+    int pid = atoi(input_buf);
+
+    struct process* curr = processes_list;
+    while (curr != NULL) {
+        if (curr->pid == (size_t)pid) {
+            curr->status = DEAD; // equivalent of SIGKILL (no clean termination like SIGTERM)
+            printf("killed %d\n", pid);
+            break;
+        }
+        if (curr->next != NULL) {
+            curr = curr->next;
+        } else {
+            printf("couldn't find process with PID %d\n", pid);
+            break;
+        }
+    }
+}
+
 /*
  * pedicel_main - Kernel shell main function
  * @arg: argument (optional)
@@ -34,6 +82,7 @@ void smash_it()
  */
 void pedicel_main(void* arg)
 {
+    (void)arg;
     printf("Welcome to the kernel shell!\r\nType 'help' for a list of commands.\r\n");
 
     for (;;) {
@@ -42,16 +91,18 @@ void pedicel_main(void* arg)
         keyboard_getline(input_buf, PEDICEL_INPUT_SIZE);
         
         if (strncmp(input_buf, "help", 4) == 0) {
-            printf("\r\nYou are currently running the test kernel shell. This is not\r\n"
+            printf("++ shell builtins ++\r\n\r\n"
-                        "a fully-fledged shell like you'd find in a complete operating system,\r\n"
+                            "\tclear   - clear the screen\n"
-                        "but rather a toy to play around in the meantime.\r\n\r\n"
+                            "\tpanic   - trigger a test panic\n"
-                            "clear   - clear the screen\r\n"
+                            "\tsyscall - trigger int 0x80\n"
-                            "panic   - trigger a test panic\r\n"
+                            "\tpf      - trigger a page fault\n"
-                            "syscall - trigger int 0x80\r\n"
+                            "\tnow     - get current date\n"
-                            "pf      - trigger a page fault\r\n"
+                            "\tsmash   - smash the stack\n"
-                            "now     - get current date\r\n"
+                            "\tmem     - get used heap info\n"
-                            "smash   - smash the stack\r\n"
+                            "\tload    - load an user executable\n"
-                            "mem     - get used heap info\r\n");
+                            "\tlist    - list initfs.tar contents\n"
+                            "\tps      - list running processes\n"
+                            "\tkill    - kill a running process by PID\n");
             continue;
         }
 
@@ -96,6 +147,26 @@ void pedicel_main(void* arg)
             continue;
         }
 
+        if (strncmp(input_buf, "load", 4) == 0) {
+            loader_load_raw();
+            continue;
+        }
+
+        if (strncmp(input_buf, "list", 4) == 0) {
+            tar_list();
+            continue;
+        }
+
+        if (strncmp(input_buf, "ps", 2) == 0) {
+            ps();
+            continue;
+        }
+
+        if (strncmp(input_buf, "kill", 4) == 0) {
+            kill();
+            continue;
+        }
+
         printf("%s: command not found\r\n", input_buf);
     }
 }

src/kapps/loader.c

@@ -0,0 +1,32 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   Executable loader
+ * @license GPL-3.0-only
+ */         
+
+#include <stddef.h>
+#include <fs/initfs.h>
+#include <kernel.h>
+#include <sched/process.h>
+#include <io/kbd/ps2.h>
+#include <string/string.h>
+
+extern void* archive_start_addr;
+
+int loader_load_raw()
+{
+    char input_buf[PEDICEL_INPUT_SIZE] = {0};
+    do {
+        printf("file> ");
+        keyboard_getline(input_buf, PEDICEL_INPUT_SIZE);   
+    } while (strncmp(input_buf, "", 1) == 0);
+
+    char* data = NULL;
+    int sz = tar_lookup(archive_start_addr, input_buf,&data);
+    if (sz > 0) {
+        process_create_user_raw(data, sz, input_buf);
+        return 0; // TODO: should return something else on error
+    }
+    printf("Couldn't load file '%s'\r\n", input_buf);
+    return 1;
+}

src/kmain.c

@@ -25,6 +25,7 @@
 #include <io/term/flanterm_backends/fb.h>
 #include <arch/x86.h>
 #include <boot/boot.h>
+#include <fs/initfs.h>
 
 // Limine version used
 __attribute__((used, section(".limine_requests")))
@@ -72,19 +73,12 @@ struct process* idle_proc;
 
 void idle_main(void* arg)
 {
+	(void)arg;
 	for (;;) {
 		asm("hlt");
 	}
 }
 
-void thing_main(void* arg)
-{
-	printf("What's your name, pal? ");
-	char name[10];
-	keyboard_getline(name, 10);
-	printf("\r\n{%s} is such a nice name!\r\n", name);
-}
-
 extern uintptr_t kheap_start;
 
 /*
@@ -118,22 +112,19 @@ void kmain()
 	
 	process_init();
 	idle_proc = process_create("idle", (void*)idle_main, 0);
-	process_create("pedicel", (void*)pedicel_main, 0);
-
-	scheduler_init();
 
 	if (!boot_ctx.module) {
-		panic(NULL, "could not load 'hello' executable :(");
+		panic(NULL, "could not load initfs.tar :(");
 	}
+	if (boot_ctx.module->module_count == 1) {
+		initfs_init(boot_ctx.module->modules[0]);
+	}
+	
+	process_create("kshell", (void*)pedicel_main, 0);
+	scheduler_init();
 
 	printf(PEPPEROS_SPLASH);
 	init.all = true;
 
-	if (boot_ctx.module->module_count == 1) {
-		file = boot_ctx.module->modules[0];
-		DEBUG("file: addr=%p size=%u", file->address, file->size);
-		process_create_user(file);
-	}
-
 	idle();
 }

src/mem/vmm.c

@@ -18,6 +18,7 @@ compared to the PMM which allocs/frees 4kb frames ("physical pages").
 #include <mem/paging.h>
 #include <stddef.h>
 #include <mem/pmm.h>
+#include <mem/utils.h>
 #include <kernel.h>
 
 extern uint64_t *kernel_pml4;
@@ -116,6 +117,8 @@ void* vmm_map(uint64_t* pml4, uint64_t virt, uint64_t flags)
         panic(NULL, "VMM/PMM out of memory!");
     }
 
+    memset(PHYS_TO_VIRT(phys), 0, PAGE_SIZE);
+
     paging_map_page(pml4, virt, phys, flags | PTE_PRESENT);
     return (void*)virt;
 }
@@ -246,7 +249,6 @@ uintptr_t vmm_alloc_user_stack(uint64_t* pml4)
     for (size_t i=stack_top; i>stack_top-stack_size; i-=PAGE_SIZE) {
         vmm_map(pml4, i, PTE_PRESENT | PTE_WRITABLE | PTE_USER);
     }
-
     return stack_top;
 }
 
@@ -254,7 +256,11 @@ uintptr_t vmm_alloc_user_code(uint64_t* pml4, void* code_addr, uint64_t code_siz
 {
     uintptr_t code_start = USER_CODE_START;
     
-    for (size_t i=code_start; i<code_start+code_size; i+=PAGE_SIZE) {
+    // Round code_size up to next page boundary
+    uint64_t code_size_aligned = (code_size + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1);
+    uint64_t mapped_size = code_size_aligned + ((uint64_t)USER_RAW_EXTRA_PAGES * PAGE_SIZE);
+    
+    for (uint64_t i=code_start; i<code_start+mapped_size; i+=PAGE_SIZE) {
         vmm_map(pml4, i, PTE_PRESENT | PTE_WRITABLE | PTE_USER);
     } 
     

src/sched/process.c

@@ -98,6 +98,8 @@ struct process* process_create(char* name, void(*function)(void*), void* arg)
     // Kernel PML4 as it already maps code/stack (when switching to userland we'll have to change that)
     proc->root_page_table = kernel_pml4;
 
+    proc->kernel_stack = kalloc_stack();
+
     proc->next = 0;
 
     process_add(&processes_list, proc);
@@ -220,38 +222,77 @@ void process_jump_to_user(uintptr_t stack_top, uintptr_t user_code)
         " :: "r"(stack_top), "r"(user_code));
 }
 
-// Kernel stack used for interrupts from userland process.
-// Should be set in TSS.RSP0 when switching to userland process.
-uint8_t interrupt_stack[0x8000];
-
 extern struct tss tss;
 
 /*
- * process_create_user - Create a new user process
+ * process_create_user_raw - Create a new user process from raw binary
- * @file: pointer to Limine file structure
+ * @file: pointer to beginning of binary
+ * @size: size of the binary
+ * @name: name for the new process
  *
- * This function takes a loaded Limine executable
+ * This function takes an executable loaded in memory
- * module, and maps its code, a user stack, sets the
+ * and maps its code, a user stack, sets the TSS RSP0
- * TSS RSP0 for interrupts, and finally jumps to the
+ * for interrupts, and finally jumps to the user code.
- * user code.
  */
-void process_create_user(struct limine_file* file)
+void process_create_user_raw(char* file, int size, char* name)
 {
-    void* exec_addr = file->address;
+    CLEAR_INTERRUPTS;
-    uint64_t exec_size = file->size;
+    struct process* proc = (struct process*)kmalloc(sizeof(struct process));
+    struct cpu_status* ctx = (struct cpu_status*)kmalloc(sizeof(struct cpu_status));
+
+    if (!proc || !ctx) panic(NULL, "out of memory while creating user process");
+
+    memset(proc, 0, sizeof(struct process));
+    memset(ctx, 0, sizeof(struct cpu_status));
+
+    strncpy(proc->name, name, PROCESS_NAME_MAX);
+    proc->pid = next_free_pid++;
+    proc->status = READY;
+    proc->next = 0;
+    proc->context = ctx;
+    proc->context->iret_ss = USER_DATA_SEGMENT | 3;
+    proc->context->iret_cs = USER_CODE_SEGMENT | 3;
+    proc->context->iret_flags = 0x202; // Interrupt Flag set
+
+        /* Set basic entries for the process's File Descriptor Table */
+    proc->fdt[0].fd = 0;
+    proc->fdt[0].open = true;
+    proc->fdt[0].cursor = 0;
+    strncpy(proc->fdt[0].filename, "stdin", PROCESS_NAME_MAX - 1);
+
+    proc->fdt[1].fd = 1;
+    proc->fdt[1].open = true;
+    proc->fdt[1].cursor = 0;
+    strncpy(proc->fdt[1].filename, "stdout", PROCESS_NAME_MAX - 1);
+
+    proc->fdt[2].fd = 2;
+    proc->fdt[2].open = true;
+    proc->fdt[2].cursor = 0;
+    strncpy(proc->fdt[2].filename, "stderr", PROCESS_NAME_MAX - 1);
+
+    proc->next_free_fd = 3; // file descriptors are also bump-allocated
+
+    void* exec_addr = (void*)file;
+    uint64_t exec_size = size;
 
     uint64_t* user_pml4 = vmm_create_address_space();
+    if (!user_pml4) panic(NULL, "failed to create user address space");
+    proc->root_page_table = user_pml4;
+
     uintptr_t stack_top = vmm_alloc_user_stack(user_pml4);
     uint64_t code = vmm_alloc_user_code(user_pml4, exec_addr, exec_size);
     
-    // Could be kalloc_stack()ed PER PROCESS when we grow that
+    proc->context->iret_rsp = stack_top;
-    tss.rsp0 = (uint64_t)(interrupt_stack + sizeof(interrupt_stack));
+    proc->context->iret_rip = code;
+    proc->kernel_stack = kalloc_stack();
+    if (!proc->kernel_stack) panic(NULL, "failed to allocate kernel stack");
 
-    // Load user_pml4 into cr3 along here??
+    // Copy code into user pages; for that we need to temporarily switch to the user pml4
     load_cr3(VIRT_TO_PHYS((uint64_t)user_pml4));
-
-    // Copy code into user pages
     memcpy((uint64_t*)code, exec_addr, exec_size);
+    load_cr3(VIRT_TO_PHYS((uint64_t)kernel_pml4));
 
-    process_jump_to_user(stack_top, code);
+    process_add(&processes_list, proc);
+    DEBUG("user process '%s' (pid=%u) enqueued for scheduling", name, proc->pid);
+    SET_INTERRUPTS;
 }

src/sched/scheduler.c

@@ -9,17 +9,21 @@
 #include <mem/paging.h>
 #include <stdint.h>
 #include <io/serial/serial.h>
+#include <arch/gdt.h>
 
 extern struct process* processes_list;
 extern struct process* current_process;
 extern struct process* idle_proc;
 
+extern struct tss tss;
+
 /*
  * scheduler_init - Choose the first process
  */
 void scheduler_init()
 {
     current_process = processes_list;
+    DEBUG("scheduler starting with: pid=%u, name='%s', context=%p", current_process->pid, current_process->name, current_process->context);
 }
 
 /*
@@ -39,42 +43,56 @@ struct cpu_status* scheduler_schedule(struct cpu_status* context)
     }
 
     if (current_process == NULL) {
-        // If no more processes, then set IDLE as the current process, that's it.
+        panic(NULL, "current_process is NULL");
-        current_process = idle_proc;
     }
     
-    if (current_process == idle_proc && current_process->next == NULL)
+    if (current_process->context == NULL) {
-    {
+        panic(NULL, "current_process->context is NULL");
-        return idle_proc->context;
     }
 
     current_process->context = context;
 
-    for (;;) {
+    if (current_process->status == DEAD) {
-        struct process* prev_process = current_process;
+        struct process* dead_process = current_process;
-        if (current_process->next != NULL) {
+        struct process* next_process = (dead_process->next != NULL) ? dead_process->next : processes_list;
+
+        process_delete(&processes_list, dead_process);
+
+        if (processes_list == NULL || next_process == dead_process) {
+            current_process = idle_proc;
+            return idle_proc->context;
+        }
+
+        current_process = next_process;
+    } else if (current_process->next != NULL) {
         current_process = current_process->next;
     } else {
         current_process = processes_list;
     }
 
-        if (current_process != NULL && current_process->status == DEAD) {
+    for (;;) {
-            process_delete(&prev_process, current_process);
+        if (current_process->status == DEAD) {
-            current_process = NULL;
+            struct process* dead_process = current_process;
+            struct process* next_process = (current_process->next != NULL) ? current_process->next : processes_list;
+
+            process_delete(&processes_list, dead_process);
+
+            if (processes_list == NULL || next_process == dead_process) {
+                current_process = idle_proc;
                 return idle_proc->context;
-        } else {
+            }
+
+            current_process = next_process;
+            continue;
+        }
+
         current_process->status = RUNNING;
-/*             if (prev_process != current_process) {
-                DEBUG("Changed from {pid=%u, name=%s} to {pid=%u, name=%s}", prev_process->pid, prev_process->name, current_process->pid, current_process->name);
-            } */
         break;
     }
-    }
-
-    //DEBUG("current_process={pid=%u, name='%s', root_page_table[virt]=%p}", current_process->pid, current_process->name, current_process->root_page_table);
 
+    // Here, we chose next running process so we load its kernel stack & page tables
+    tss.rsp0 = (uint64_t)current_process->kernel_stack;
     load_cr3(VIRT_TO_PHYS((uint64_t)current_process->root_page_table));
-    //DEBUG("Loaded process PML4 into CR3");
 
     return current_process->context;
 }

src/string/string.c

@@ -71,7 +71,6 @@ void strncpy(char* dst, const char* src, size_t n)
    while(i++ != n && (*dst++ = *src++));
 }
 
-
 /*
  * strncmp - compare two strings up to n characters
  * @s1: first string
@@ -99,3 +98,37 @@ int strncmp(const char* s1, const char* s2, size_t n)
         return ( *(unsigned char *)s1 - *(unsigned char *)s2 );
     }
 }
+
+/*
+ * strlen - get length of a string (BSD implementation)
+ * @str: NULL-terminated string
+ * 
+ * Return:
+ * %len - length of string
+ */
+size_t strlen(const char* str)
+{
+        const char* s;
+        for (s = str; *s; ++s);
+        return (s - str);
+}
+
+/*
+ * atoi - ascii to integer (PureDOOM implementation)
+ * @str: ASCII string to convert
+ *
+ * Return:
+ * %i - integer represented in @str
+ */
+int atoi(const char* str)
+{
+    int i = 0;
+    int c;
+
+    while ((c = *str++) != 0) {
+        i *= 10;
+        i += c - '0';
+    }
+
+    return i;
+}

src/syscall/files.c

@@ -0,0 +1,252 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   File-related system calls
+ * @license GPL-3.0-only
+ */
+
+#include <stddef.h>
+#include <kernel.h>
+#include <config.h>
+#include <sched/process.h>
+#include <string/string.h>
+#include <fs/initfs.h>
+#include <io/kbd/ps2.h>
+
+extern struct process* current_process;
+
+/*
+ * normalize_path - remove leading slashes from a path
+ * @path: path to normalize
+ *
+ * Return:
+ * %path - normalized path
+ * %NULL - if bad argument
+ */
+static const char* normalize_path(const char* path)
+{
+    if (!path) return NULL;
+
+    while (*path == '/') {
+        path++;
+    }
+
+    return path;
+}
+
+/*
+ * sys_open - Open a file
+ * @filename: Absolute path to file
+ * @flags:    (Not implemented)
+ *
+ * This system call opens a file in read-only mode,
+ * because of the TAR filesystem.
+ *
+ * Return:
+ * %fd - file descriptor refering to file
+ * On error, a negative number representing the error code is returned.
+ */
+int sys_open(const char* filename, int flags)
+{
+    // TODO: support flags (right now everything is read only, O_RDONLY)
+    (void)flags; // gcc shut up
+    const char* path = normalize_path(filename);
+
+    if (tar_exists(path) < 0) {
+        return -ENOENT; // file doesn't exist..
+    }
+    // file exists here!
+    if (current_process->next_free_fd >= FDT_MAX) {
+        return -EMFILE;
+    }
+    int fd = current_process->next_free_fd++;
+    current_process->fdt[fd].fd = fd;
+    current_process->fdt[fd].open = true;
+    current_process->fdt[fd].cursor = 0;
+    strncpy(current_process->fdt[fd].filename, path, PROCESS_NAME_MAX - 1);
+    return fd;
+}
+
+/*
+ * sys_close - Close a file
+ * @fd: file descriptor
+ *
+ * This system call closes the file referred to by
+ * the file descriptor @fd.
+ *
+ * Return:
+ * %0       - file closed
+ * %-EBADFD - bad file descriptor
+ */
+int sys_close(unsigned int fd)
+{
+    if (fd >= FDT_MAX) {
+        return -EBADFD;
+    }
+
+    if (!current_process->fdt[fd].open) {
+        return -EBADFD; // FD not opened in the first place
+    }
+
+    current_process->fdt[fd].open = false;
+    current_process->fdt[fd].filename[0] = '\0';
+    current_process->fdt[fd].cursor = 0;
+    return 0;
+}
+
+/*
+ * sys_lseek - reposition file cursor
+ * @fd:     file descriptor referring to file
+ * @offset: byte amount to be used according to @whence
+ * @whence: resposition directive
+ *
+ * This system call repositions the cursor for @fd by @offset,
+ * from somewhere in the file, depending on the value of @whence:
+ * SEEK_SET (0) -> from the beginning
+ * SEEK_CUR (1) -> from the actual value of the cursor
+ * SEEK_END (2) -> from the end of the file
+ *
+ * Return:
+ * %new_cursor - the new cursor value
+ * On error, a negative value corresponding to an error code is returned.
+ */
+int sys_lseek(unsigned int fd, int offset, int whence)
+{
+    if (fd >= FDT_MAX) {
+        return -EBADFD;
+    }
+    if (!current_process->fdt[fd].open) {
+        return -EBADFD;
+    }
+
+    int filesize = tar_exists(current_process->fdt[fd].filename);
+    if (filesize < 0) {
+        return -ENOENT;
+    }
+
+    int cursor = current_process->fdt[fd].cursor;
+    int base;
+    switch (whence) {
+        case SEEK_SET: base = 0; break;
+        case SEEK_CUR: base = cursor; break;
+        case SEEK_END: base = filesize; break;
+        default: return -EINVAL;
+    }
+
+    int new_cursor = base + offset;
+    if (new_cursor < 0) new_cursor = 0;
+    if (new_cursor > filesize) new_cursor = filesize;
+
+    current_process->fdt[fd].cursor = new_cursor;
+    return new_cursor;
+}
+
+/*
+ * sys_tell - Get cursor position for a file
+ * @fd: file descriptor to use
+ *
+ * Return:
+ * %cursor - cursor position of the file descriptor
+ * On error, a negative value with the corresponding error code is set.
+ */
+int sys_tell(unsigned int fd)
+{
+    if (fd >= FDT_MAX) {
+        return -EBADFD;
+    }
+    if (!current_process->fdt[fd].open) {
+        return -EBADFD;
+    }
+    return current_process->fdt[fd].cursor;
+}
+
+/*
+ * sys_eof - Are we at the end of the file yet?
+ * @fd: file descriptor to use
+ *
+ * This function determines if the cursor for the
+ * file descriptor @fd is at or past the end of
+ * the file it refers to.
+ *
+ * Return:
+ * %>0 - we are at or past EOF
+ * %0  - not yet (still have bytes to read)
+ */
+int sys_eof(unsigned int fd)
+{
+    if (fd >= FDT_MAX) {
+        return -EBADFD;
+    }
+    if (!current_process->fdt[fd].open) {
+        return -EBADFD;
+    }
+
+    int filesize = tar_exists(current_process->fdt[fd].filename);
+    if (filesize < 0) {
+        return -ENOENT;
+    }
+
+    return current_process->fdt[fd].cursor >= (uint64_t)filesize;
+}
+
+/*
+ * sys_read - read from an open file
+ * @fd:    file descriptor to use
+ * @buf:   out buffer for read data
+ * @count: amount of bytes to read
+ *
+ * Return:
+ * %sz - amount of bytes read (on success)
+ * On error, a negative value with an error code is returned.
+ */
+int sys_read(unsigned int fd, char* buf, size_t count)
+{
+    size_t i;
+    switch (fd) {
+        case 0: //read from stdin (keyboard)
+            for (i=0; i<count; i++) {
+                buf[i] = keyboard_getchar();
+            }
+            return i;
+        case 1: // from stdout
+        case 2: // from stderr
+            return -EBADFD;
+        default: // from an open file?
+            if (current_process->fdt[fd].open == false) {
+                return -EBADFD; // File descriptor wasn't open
+            }
+            // Here fd refers to a valid opened file..
+            int sz = tar_read(current_process->fdt[fd].filename, buf, count,
+                            current_process->fdt[fd].cursor);
+            if (sz == 0) {
+                return -ENOENT;
+            } else {
+                current_process->fdt[fd].cursor += sz;
+                return sz;
+            }
+    }
+    return -EBADFD;
+}
+
+/*
+ * sys_write - write to a file
+ * @fd:    file descriptor to write to
+ * @buf:   buffer of bytes to write
+ * @count: number of bytes to write
+ *
+ * Return:
+ * %count - number of bytes written
+ * On error, a negative value with corresponding error code is returned.
+ */
+int sys_write(unsigned int fd, const char* buf, size_t count)
+{
+    switch (fd) {
+        case 1: //stdout
+        case 2: //stderr
+            for (size_t i=0; i<count; i++) {
+                internal_putc(buf[i], NULL);
+            }
+            return count;
+        default:
+            return -EBADFD;
+    }
+}

src/syscall/proc.c

@@ -0,0 +1,28 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   Process-wide system calls
+ * @license GPL-3.0-only
+ */
+
+#include <kernel.h>
+#include <sched/process.h>
+
+extern struct process* current_process;
+
+/*
+ * sys_exit - Terminate the current process
+ * @error_code: error code to return
+ *
+ * This system call exits the running process. Be aware that
+ * it does nothing except that. All resources are supposed to
+ * be cleaned up before calling this.
+ *
+ * Return:
+ * %error_code - the error code to return
+ */
+int sys_exit(int error_code)
+{
+    current_process->status = DEAD;
+    DEBUG("(pid=%u, name=%s)", current_process->pid, current_process->name);
+    return error_code;
+}

src/syscall/screen.c

@@ -0,0 +1,105 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   Screen/framebuffer system calls
+ * @license GPL-3.0-only
+ */
+
+// This is not part of POSIX or anything btw
+
+#include <stdint.h>
+#include <kernel.h>
+
+extern struct boot_context boot_ctx;
+
+/*
+ * pack_rbga_to_fb - Convert rgb values to rgba pixel for framebuffer
+ * @r: red value
+ * @g: green value
+ * @b: blue value
+ *
+ * This function uses the mask sizes and shifts of the Limine
+ * framebuffer to convert raw rgb values into usable pixels.
+ *
+ * Return:
+ * <pixel> - usable pixel
+ */
+static uint32_t pack_rgba_to_fb(uint8_t r, uint8_t g, uint8_t b)
+{
+    uint32_t pixel = 0;
+
+    if (boot_ctx.fb->red_mask_size) {
+        pixel |= ((uint32_t)(r >> (8 - boot_ctx.fb->red_mask_size)) << boot_ctx.fb->red_mask_shift);
+    }
+    if (boot_ctx.fb->green_mask_size) {
+        pixel |= ((uint32_t)(g >> (8 - boot_ctx.fb->green_mask_size)) << boot_ctx.fb->green_mask_shift);
+    }
+    if (boot_ctx.fb->blue_mask_size) {
+        pixel |= ((uint32_t)(b >> (8 - boot_ctx.fb->blue_mask_size)) << boot_ctx.fb->blue_mask_shift);
+    }
+
+    return pixel;
+}
+
+/*
+ * sys_draw - Draw a framebuffer subset
+ * @src:      Source frame
+ * @width:    width of the frame
+ * @height:   height of the frame
+ * @channels: how many channels (RGB, RGBA)
+ *
+ * This system call draws the frame @src, having some
+ * @width and @height, in the top right of the Limine
+ * framebuffer. (Used for DOOM)
+ *
+ * Return:
+ * %0 - on success
+ * On error, a negative value with an error code is returned.
+ */
+int sys_draw(const uint8_t* src, int width, int height, int channels)
+{
+    if (!boot_ctx.fb || !src) {
+        return -EINVAL;
+    }
+    if (channels != 4 || width <= 0 || height <= 0) {
+        return -EINVAL;
+    }
+    if (boot_ctx.fb->bpp < 24) {
+        return -EIO;
+    }
+
+    uint32_t* dst = (uint32_t*)boot_ctx.fb->address;
+    uint64_t dst_w = boot_ctx.fb->width;
+    uint64_t dst_h = boot_ctx.fb->height;
+    uint64_t dst_pitch_px = boot_ctx.fb->pitch / 4;
+    uint64_t scale = 2;
+    uint64_t scaled_w = (uint64_t)width * scale;
+    uint64_t scaled_h = (uint64_t)height * scale;
+
+    if (scaled_w > dst_w || scaled_h > dst_h) {
+        return -EINVAL;
+    }
+
+    uint64_t dst_x = dst_w - scaled_w;
+    uint64_t dst_y = 0;
+
+    for (uint64_t y = 0; y < (uint64_t)height; ++y) {
+        const uint8_t* src_row = src + y * (uint64_t)width * 4ULL;
+        uint32_t* dst_row0 = dst + (dst_y + y * scale) * dst_pitch_px + dst_x;
+        uint32_t* dst_row1 = dst_row0 + dst_pitch_px;
+
+        for (uint64_t x = 0; x < (uint64_t)width; ++x) {
+            const uint8_t* p = src_row + x * 4ULL;
+            uint32_t pixel = pack_rgba_to_fb(p[0], p[1], p[2]);
+            uint64_t dx = x * scale;
+
+            // Write 2x2 block. This is because source DOOM frame (320x200)
+            // is quite small.
+            dst_row0[dx + 0] = pixel;
+            dst_row0[dx + 1] = pixel;
+            dst_row1[dx + 0] = pixel;
+            dst_row1[dx + 1] = pixel;
+        }
+    }
+
+    return 0;
+}

src/syscall/syscall.c

@@ -0,0 +1,98 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   System call handling
+ * @license GPL-3.0-only
+ */
+
+#include "config.h"
+#include "sched/scheduler.h"
+#include <arch/x86.h>
+#include <kernel.h>
+#include <stddef.h>
+#include <io/term/term.h>
+#include <sched/process.h>
+#include <io/kbd/ps2.h>
+#include <fs/initfs.h>
+#include <string/string.h>
+#include <mem/utils.h>
+
+extern struct process* current_process;
+
+int sys_read(unsigned int fd, char* buf, size_t count);
+int sys_write(unsigned int fd, const char* buf, size_t count);
+int sys_open(const char* filename, int flags);
+int sys_close(unsigned int fd);
+int sys_lseek(unsigned int fd, int offset, int whence);
+int sys_tell(unsigned int fd); // needed by doom, therefore TOP PRIORITY
+int sys_eof(unsigned int fd); // same
+int sys_draw(const uint8_t* src, int width, int height, int channels);
+int sys_exit(int error_code);
+
+/*
+ * syscall_handler - System call dispatcher
+ * @regs: CPU state
+ *
+ * This function is called from the interrupt dispatcher,
+ * when an interrupt 0x80 is emitted from userland.
+ *
+ * It switches control to the syscall number provided
+ * in %rax. 
+ *
+ * We try to follow the System V convention here:
+ * - syscall number in %rax
+ * - args in %rdi, %rsi, %rdx, %r10, %r8, %r9
+ * - return value (if any) in %rax
+ *
+ * Return:
+ * <regs> - CPU state after system call
+ */
+struct cpu_status* syscall_handler(struct cpu_status* regs)
+{
+    switch (regs->rax)
+    {
+        case 0:
+            DEBUG("sys_read(fd=%u, buf=%p, count=%u)", regs->rdi, regs->rsi, regs->rdx);
+            regs->rax = sys_read(regs->rdi, (char*)regs->rsi, regs->rdx);
+            break;
+        case 1:
+            DEBUG("sys_write(fd=%u, buf=%p, count=%u)", regs->rdi, regs->rsi, regs->rdx);
+            regs->rax = sys_write(regs->rdi, (char*)regs->rsi, regs->rdx);
+            break;
+        case 2:
+            DEBUG("sys_open(filename=%s, flags=%u)", regs->rdi, regs->rsi);
+            regs->rax = sys_open((const char*)regs->rdi, regs->rsi);
+            break;
+        case 3:
+            DEBUG("sys_close(fd=%u)", regs->rdi);
+            regs->rax = sys_close(regs->rdi);
+            break;
+        case 8:
+            DEBUG("sys_lseek(fd=%u, off=%d, whence=%u)", regs->rdi, regs->rsi, regs->rdx);
+            regs->rax = sys_lseek(regs->rdi, regs->rsi, regs->rdx);
+            break;
+        case 9:
+            DEBUG("sys_tell(fd=%u)", regs->rdi);
+            regs->rax = sys_tell(regs->rdi);
+            break;
+        case 10:
+            DEBUG("sys_eof(fd=%u)", regs->rdi);
+            regs->rax = sys_eof(regs->rdi);
+            break;
+        case 11: // No DEBUG() here because it makes significant overhead
+            regs->rax = sys_draw((const uint8_t*)regs->rdi, regs->rsi, regs->rdx, regs->r10);
+            break;
+        case 60:
+            DEBUG("sys_exit(error_code=%d)", regs->rdi);
+            regs->rax = sys_exit(regs->rdi);
+            break;
+        default:
+            DEBUG("Bad syscall! (rax=%p, rdi=%p, rsi=%p, rdx=%p)",
+                    regs->rax, regs->rdi, regs->rsi, regs->rdx);
+            regs->rax = 0xbad515ca11;
+            break;
+    }
+
+    //DEBUG("returned rax=%p (%u)", regs->rax, regs->rax);
+
+    return regs;
+}

user/Makefile

@@ -0,0 +1,28 @@
+CC := x86_64-elf-gcc
+CC_FLAGS := -ffreestanding -nostdlib -fno-pic -mno-red-zone -Ilibc
+DOOM_CC_FLAGS := $(CC_FLAGS) -std=gnu11
+
+LD := x86_64-elf-ld
+
+BUILDDIR := ../build
+LIBDIR := libc
+
+all: pedicel apex doom
+
+.PHONY: pedicel
+pedicel:
+	nasm -f bin pedicel.S -o $(BUILDDIR)/pedicel.raw
+
+.PHONY: apex
+apex:
+	$(CC) $(CC_FLAGS) -c apex.c -o $(BUILDDIR)/apex.o
+	nasm -f elf64 $(LIBDIR)/crt0.S -o $(BUILDDIR)/crt0.o
+	$(LD) -T $(LIBDIR)/linker.ld $(BUILDDIR)/crt0.o $(BUILDDIR)/apex.o -o $(BUILDDIR)/apex.elf
+	objcopy -O binary $(BUILDDIR)/apex.elf $(BUILDDIR)/apex.raw
+
+.PHONY: doom
+doom:
+	$(CC) $(DOOM_CC_FLAGS) -c doom.c -o $(BUILDDIR)/doom.o
+	nasm -f elf64 $(LIBDIR)/crt0.S -o $(BUILDDIR)/crt0.o
+	$(LD) -T $(LIBDIR)/linker.ld $(BUILDDIR)/crt0.o $(BUILDDIR)/doom.o -o $(BUILDDIR)/doom.elf
+	objcopy -O binary $(BUILDDIR)/doom.elf $(BUILDDIR)/doom.raw

user/apex.c

@@ -0,0 +1,7 @@
+#include <syscall.h>
+
+int main() {
+    const char* msg = "hi from C userland\r\n";
+    write(1, msg, 21);
+    return 42;
+}

user/doom.c

@@ -0,0 +1,147 @@
+#define DOOM_IMPLEMENTATION 
+#include "PureDOOM.h"
+#include <syscall.h>
+#include <stdint.h>
+
+//We use a separate heap because malloc is not yet available in userspace.
+#define DOOM_HEAP_START ((unsigned char*)0x00500000)
+#define DOOM_HEAP_SIZE  (24 * 1024 * 1024) //24mb
+
+static unsigned char* doom_heap_curr = DOOM_HEAP_START;
+static unsigned char* doom_heap_end = DOOM_HEAP_START + DOOM_HEAP_SIZE;
+
+// The following functions are wrappers for system calls made for
+// compatibility with puredoom's function signatures
+static int doom_cstr_len(const char* str)
+{
+    int len = 0;
+    while (str && str[len]) len++;
+    return len;
+}
+
+static void doom_print_cb(const char* str)
+{
+    int len = doom_cstr_len(str);
+    if (len > 0) {
+        write(1, str, len);
+    }
+}
+
+static void* doom_open_cb(const char* filename, const char* mode)
+{
+    (void)mode; // open doesn't support flags/mode (yet)
+    int fd = open(filename, 0);
+    if (fd < 0) return 0;
+    return (void*)(intptr_t)(fd);
+}
+
+static void doom_close_cb(void* handle)
+{
+    int fd = (int)(intptr_t)handle;
+    if (fd >= 0) {
+        close(fd);
+    }
+}
+
+static int doom_read_cb(void* handle, void* buf, int count)
+{
+    int fd = (int)(intptr_t)handle;
+    if (fd < 0) return -1;
+    return read(fd, (char*)buf, count);
+}
+
+static int doom_write_cb(void* handle, const void* buf, int count)
+{
+    int fd = (int)(intptr_t)handle;
+    if (fd < 0) return -1;
+    return (int)write(fd, (const char*)buf, count);
+}
+
+static int doom_seek_cb(void* handle, int offset, doom_seek_t origin)
+{
+    int fd = (int)(intptr_t)handle;
+    if (fd < 0) return -1;
+    return seek(fd, offset, (int)origin);
+}
+
+static int doom_tell_cb(void* handle)
+{
+    int fd = (int)(intptr_t)handle;
+    if (fd < 0) return -1;
+    return tell(fd);
+}
+
+static int doom_eof_cb(void* handle)
+{
+    int fd = (int)(intptr_t)handle;
+    if (fd < 0) return 1;
+    return eof(fd);
+}
+
+// Bump allocator (bump ptr until we're out of heap memory)
+static void* doom_malloc_cb(int size)
+{
+    if (size <= 0) return 0;
+
+    uintptr_t curr = (uintptr_t)doom_heap_curr;
+
+    // 16-byte align allocated blocks
+    curr = (curr + 15ULL) & ~15ULL;
+
+    if (curr + (uintptr_t)size > (uintptr_t)doom_heap_end) {
+        return 0;
+    }
+
+    doom_heap_curr = (unsigned char*)(curr + (uintptr_t)size);
+    return (void*)curr;
+}
+
+// No free
+static void doom_free_cb(void* ptr)
+{
+    (void)ptr;
+}
+
+static void doom_exit_cb(int code)
+{
+    exit(code);
+}
+
+// To get path for the WAD file
+static char* doom_getenv_cb(const char* var)
+{
+    static char home[] = "/";
+    static char waddir[] = "/";
+
+    if (!var) return 0;
+    if (doom_strcmp(var, "HOME") == 0) return home;
+    if (doom_strcmp(var, "DOOMWADDIR") == 0) return waddir;
+    return 0;
+}
+
+int main()
+{
+    char* argv[2] = {"doom", 0};
+
+    // Override default implementations
+    doom_set_print(doom_print_cb);
+    doom_set_malloc(doom_malloc_cb, doom_free_cb);
+    doom_set_file_io(doom_open_cb,
+                     doom_close_cb,
+                     doom_read_cb,
+                     doom_write_cb,
+                     doom_seek_cb,
+                     doom_tell_cb,
+                     doom_eof_cb);
+    doom_set_exit(doom_exit_cb);
+    doom_set_getenv(doom_getenv_cb);
+
+    doom_init(1, argv, 0);
+
+    while (true)
+    {
+        doom_force_update();
+        const uint8_t* framebuffer = doom_get_framebuffer(4);
+        draw(framebuffer, 320, 200, 4);
+    }
+}

user/hello.S

@@ -1,16 +0,0 @@
-bits 64
-
-section .data
-    hi db "hi from userland :) we did it man", 0
-
-section .text
-
-hello:
-    mov rax, 0x1        ;sys_write
-    mov rdi, 0x1        ;stdout
-    lea rsi, [rel hi]   ;char* buf
-    mov rdx, 33         ;count
-    int 0x80
-
-.loop:
-    jmp .loop

user/libc/crt0.S

@@ -0,0 +1,18 @@
+bits 64
+global _start
+extern main
+
+section .text
+
+; Begin the program with main() function
+_start:
+    call main
+
+; Exit the program by exit() syscall
+.exit:
+    mov rdi, rax    ; put the value of "return X;" (rax) as arg1 (error_code)
+    mov rax, 60     ; sys_exit
+    int 0x80
+
+.loop:
+    jmp .loop

user/libc/linker.ld

@@ -0,0 +1,22 @@
+ENTRY(_start)
+
+SECTIONS
+{
+    . = 0x400000;
+
+    .text : {
+        *(.text*)
+    }
+
+    .rodata : {
+        *(.rodata*)
+    }
+
+    .data : {
+        *(.data*)
+    }
+
+    .bss : {
+        *(.bss*)
+    }
+}

user/libc/syscall.h

@@ -0,0 +1,88 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   System call wrappers for userspace
+ * @license GPL-3.0-only
+ */
+
+#pragma once
+// TODO: replace all ifndef/define/endif by pragma once..
+
+#include <stddef.h>
+#include <stdint.h>
+
+// 3-args syscall
+static inline long syscall3(long n, long a, long b, long c) {
+    long ret;
+
+    __asm__ volatile (
+        "int $0x80"
+        : "=a"(ret)
+        : "a"(n), "D"(a), "S"(b), "d"(c) // a = rax, D = rdi, S = rsi, d = rdx
+        : "memory"
+    );
+
+    return ret;
+}
+
+// 4-args syscall
+static inline long syscall4(long n, long a, long b, long c, long d) {
+    long ret;
+    register long r10 __asm__("r10") = d;
+
+    __asm__ volatile (
+        "int $0x80"
+        : "=a"(ret)
+        : "a"(n), "D"(a), "S"(b), "d"(c), "r"(r10)
+        : "memory"
+    );
+
+    return ret;
+}
+
+// Single-arg syscall
+static inline long syscall1(long n, long a) {
+    return syscall3(n, a, 0, 0);
+}
+
+static inline int write(int fd, const char* buf, long len) {
+    return (int)syscall3(1, fd, (long)buf, len);
+}
+
+static inline int read(int fd, char* buf, long len) {
+    return (int)syscall3(0, fd, (long)buf, len);
+}
+
+static inline int open(const char* path, int flags) {
+    return (int)syscall3(2, (long)path, flags, 0);
+}
+
+static inline int close(unsigned int fd) {
+    return (int)syscall3(3, fd, 0, 0);
+}
+
+static inline int seek(int fd, int offset, int whence) {
+    return (int)syscall3(8, fd, offset, whence);
+}
+
+static inline int tell(unsigned int fd) {
+    return (int)syscall1(9, fd);
+}
+
+static inline int eof(unsigned int fd) {
+    return (int)syscall1(10, fd);
+}
+
+static inline int draw(const unsigned char* buf, int width, int height, int channels) {
+    return (int)syscall4(11, (long)buf, width, height, channels);
+}
+
+static inline void exit(int code) {
+    __asm__ volatile (
+        "int $0x80"
+        :
+        : "a"(60), "D"(code)
+        : "memory"
+    );
+
+    for (;;);
+}

user/pedicel.S

@@ -0,0 +1,51 @@
+bits 64
+
+section .data
+hello db 0x0A, 0x0D, "TEST PROGRAM...", 0x0A, 0x0D, 0
+filename db "wow.txt", 0
+
+section .text
+
+_start:
+    mov rax, 0x1    ;sys_write
+    mov rdi, 0x1    ;stdout
+    lea rsi, [rel hello]
+    mov rdx, 19     ;count
+    int 0x80
+
+    ; Open a file
+    mov rax, 0x2            ;sys_open
+    lea rdi, [rel filename] ;filename
+    mov rsi, 0x0            ;flags
+    int 0x80
+
+    mov rdi, rax            ;fd
+    mov rax, 0x0            ;sys_read
+    lea rsi, [rel buf]      ;buf
+    mov rdx, 33             ;count
+    int 0x80
+
+    mov rax, 0x1        ;sys_write
+    mov rdi, 0x1        ;stdout
+    lea rsi, [rel buf]  ;buf
+    mov rdx, 33         ;count
+    int 0x80
+
+    mov rax, 0x3        ;sys_close
+    mov rdi, 0x3        ;fd
+    int 0x80
+
+; when we are ready to have an os specific toolchain,
+; this bit (exit & loop) should be appended at the end of every
+; C program we compile.
+
+.end:
+    mov rax, 0x3C
+    mov rdi, 0x0
+    int 0x80
+
+.loop:
+    jmp .loop
+
+section .bss
+buf resb 10

user/wow.txt

@@ -0,0 +1 @@
+hi from a file opened in usermode