Merge pull request 'should be right?' (#20) from sys_fix into main

Reviewed-on: #20

Makefile

@@ -1,3 +1,5 @@
+USER_PROGRAMS := pedicel.raw apex.raw
+USER_FILES := wow.txt
 
 BUILDDIR := build
 ELFFILE := pepperk
@@ -16,7 +18,7 @@ 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
 
@@ -41,12 +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)/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 +40,13 @@ 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.
+
+Name | Number (%rax) | arg0 (%rdi) | arg1 (%rsi) | arg2 (%rdx) |
+|---|---|---|---|---|
+| 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_exit | 60 | int error_code | | |

docs/SOFTWARE.md

@@ -0,0 +1,57 @@
+# 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.
+
+## 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.

include/arch/gdt.h

@@ -13,19 +13,32 @@
 // we'll only use this as a requirement for paging, not more.
 // This means base 0 and no limit (whole address space)
 
-#define NUM_GDT_ENTRIES 5
+#define NUM_GDT_ENTRIES 7
 
 #define NULL_SELECTOR 0x00
 #define KERNEL_CODE_SEGMENT 0x08
 #define KERNEL_DATA_SEGMENT 0x10
 #define USER_CODE_SEGMENT 0x18
 #define USER_DATA_SEGMENT 0x20
+#define TSS_SEGMENT 0x28
 
 struct GDTR {
     uint16_t limit;
     uint64_t address;
 } __attribute__((packed));
 
+struct tss {
+    uint32_t reserved0;
+    uint64_t rsp0;
+    uint64_t rsp1;
+    uint64_t rsp2;
+    uint64_t reserved1;
+    uint64_t ist[7];
+    uint64_t reserved2;
+    uint16_t reserved3;
+    uint16_t iopb;
+} __attribute__((packed));
+
 void gdt_init(void);
 
 #endif

include/arch/x86.h

@@ -10,7 +10,6 @@ void wrmsr(uint32_t msr, uint64_t value);
 bool x86_has_msr();
 void x86_arch_init();
 
-
 void x86_cpu_identification();
 int cpuid_get_vendor_string(char* str);
 
@@ -36,7 +35,7 @@ struct idtr {
 // All general-purpose registers (except rsp) as stored on the stack,
 // plus the values we pushed (vector number, error code) and the iret frame
 // In reverse order because the stack grows downwards.
-struct cpu_status_t {
+struct cpu_status {
     uint64_t r15;
     uint64_t r14;
     uint64_t r13;
@@ -63,6 +62,6 @@ struct cpu_status_t {
     uint64_t iret_ss;
 };
 
-struct cpu_status_t* syscall_handler(struct cpu_status_t* regs);
+struct cpu_status* syscall_handler(struct cpu_status* regs);
 
 #endif

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"\
@@ -40,6 +40,11 @@
 #define KERNEL_STACK_SIZE 65536
 #define KERNEL_IDT_ENTRIES 33
 
+/* user */
+#define USER_STACK_TOP 0x80000000
+#define USER_STACK_PAGES 16 // 16*4096 = 64kb
+#define USER_CODE_START 0x400000 // like linux
+
 /* paging */
 #define PAGING_MAX_PHYS 0x200000000
 
@@ -58,4 +63,8 @@
 /* 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 +1,18 @@
+/*
+ * @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();
+
+#endif

include/kernel.h

@@ -8,9 +8,14 @@
 #define KERNEL_H
 
 #include "limine.h"
+
 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")
@@ -35,7 +40,7 @@ extern volatile uint64_t ticks;
 
 // printf("debug: [%s]: " log "\n", __FILE__, ##__VA_ARGS__);
 
-void panic(struct cpu_status_t* ctx, const char* str);
+void panic(struct cpu_status* ctx, const char* str);
 void hcf(void);
 void idle(void);
 
@@ -46,6 +51,8 @@ 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 {
@@ -54,6 +61,7 @@ struct boot_context {
 	struct limine_hhdm_response* hhdm;
 	struct limine_kernel_address_response* kaddr;
 	struct limine_boot_time_response* bootdate;
+	struct limine_module_response* module;
 };
 
 // Are these modules initialized yet?

include/mem/kheap.h

@@ -16,11 +16,11 @@
 #include <stddef.h>
 #include <stdint.h>
 
-struct heap_block_t {
+struct heap_block {
     size_t size;
     bool free; // 1byte
     uint8_t reserved[7]; // (7+1 = 8 bytes)
-    struct heap_block_t* next;
+    struct heap_block* next;
 } __attribute__((aligned(16)));
 
 void kheap_init(void);
@@ -28,5 +28,6 @@ void* kmalloc(size_t size);
 void kfree(void* ptr);
 void* kalloc_stack(void);
 void kheap_map_page(void);
+void kheap_info();
 
 #endif

include/mem/paging.h

@@ -34,7 +34,7 @@ extern uint64_t hhdm_off;
 #define PAGE_ALIGN_DOWN(x) ((x) & PTE_ADDR_MASK)
 
 #define ALIGN(size) ALIGN_UP(size, 16)
-#define BLOCK_MIN_SIZE (sizeof(struct heap_block_t) + 16)
+#define BLOCK_MIN_SIZE (sizeof(struct heap_block) + 16)
 
 #define PML4_INDEX(x) (((x) >> 39) & 0x1FF)
 #define PDPT_INDEX(x) (((x) >> 30) & 0x1FF)

include/mem/vmm.h

@@ -22,6 +22,8 @@ void vmm_unmap(uint64_t* pml4, uint64_t virt);
 void* vmm_map(uint64_t* pml4, uint64_t virt, uint64_t flags);
 uint64_t* vmm_create_address_space();
 uint64_t vmm_virt_to_phys(uint64_t* pml4, uint64_t virt);
+uintptr_t vmm_alloc_user_stack(uint64_t* pml4);
+uintptr_t vmm_alloc_user_code(uint64_t* pml4, void* code_addr, uint64_t code_size);
 
 #define VMM_USER_SPACE_START 0x0000000000001000 
 #define VMM_USER_SPACE_END   0x00007FFFFFFFF000

include/sched/process.h

@@ -10,6 +10,8 @@
 #include <stddef.h>
 #include <config.h>
 #include <stdint.h>
+#include <limine.h>
+#include <stdbool.h>
 
 typedef enum {
     READY,
@@ -17,23 +19,38 @@ typedef enum {
     DEAD
 } status_t;
 
-struct process_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];
 
     status_t status;
-    struct cpu_status_t* context;
+    struct cpu_status* context;
 	void* root_page_table; // Process PML4 (should contain kernel PML4 in higher half [256-511]
-    struct process_t* next;
+    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;
 };
 
 void process_init(void);
-struct process_t* process_create(char* name, void(*function)(void*), void* arg);
+struct process* process_create(char* name, void(*function)(void*), void* arg);
-void process_add(struct process_t** processes_list, struct process_t* process);
+void process_add(struct process** processes_list, struct process* process);
-void process_delete(struct process_t** processes_list, struct process_t* process);
+void process_delete(struct process** processes_list, struct process* process);
-struct process_t* process_get_next(struct process_t* process);
+struct process* process_get_next(struct process* process);
 void process_exit(void);
 
-void process_display_list(struct process_t* processes_list);
+void process_display_list(struct process* processes_list);
+
+
+void process_create_user_raw(char* file, int size, char* name);
 
 #endif

include/sched/scheduler.h

@@ -7,7 +7,7 @@
 #ifndef SCHEDULER_H
 #define SCHEDULER_H
 
-struct cpu_status_t* scheduler_schedule(struct cpu_status_t* context);
+struct cpu_status* scheduler_schedule(struct cpu_status* context);
 void scheduler_init(void);
 
 #endif

include/sched/spinlock.h

@@ -1,4 +1,4 @@
-/*
+        /*
  * @author  xamidev <xamidev@riseup.net>
  * @brief   Spinlock implementation
  * @license GPL-3.0-only
@@ -10,13 +10,13 @@
 #include <stdbool.h>
 #include <stdint.h>
 
-struct spinlock_t
+struct spinlock
 {
     bool locked;
     uint64_t rflags;
 };
 
-void spinlock_acquire(struct spinlock_t* lock);
+void spinlock_acquire(struct spinlock* lock);
-void spinlock_release(struct spinlock_t* lock);
+void spinlock_release(struct spinlock* lock);
 
 #endif

include/security/ubsan.h

@@ -62,5 +62,4 @@ struct ubsan_overflow_data
 	struct ubsan_type_descriptor* type;
 };
 
-
 #endif

include/string/string.h

@@ -13,5 +13,6 @@ 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);
 
 #endif

limine.conf

@@ -6,3 +6,4 @@ interface_branding: Welcome to the PepperOS disk!
 	
 	comment: Default configuration (warning: spicy)
 	path: boot():/boot/pepperk
+	module_path: boot():/boot/initfs.tar

src/arch/x86/gdt.c

@@ -0,0 +1,138 @@
+/*
+ * @author  xamidev <xamidev@riseup.net>
+ * @brief   Global Descriptor Table and Task State Segment setup
+ * @license GPL-3.0-only
+ */
+
+#include <arch/gdt.h>
+#include <stdint.h>
+#include <io/serial/serial.h>
+#include <kernel.h>
+
+// Descriptors are 8-byte wide (64bits)
+// So the selectors will be (in bytes): 0x0, 0x8, 0x10, 0x18, etc..
+uint64_t gdt_entries[NUM_GDT_ENTRIES];
+struct GDTR gdtr;
+
+struct tss tss = {0};
+
+/*
+ * gdt_load - Loads Global Descriptor Table
+ */
+static void gdt_load()
+{
+    asm("lgdt %0" : : "m"(gdtr));
+}
+
+/*
+ * gdt_flush - Flushes the Global Descriptor Table
+ *
+ * This function loads new Segment Selectors to make
+ * the GDT changes take effect
+ */
+static void gdt_flush()
+{
+    // Here, 0x8 is the kernel code selector
+    // and 0x10 is the kernel data selector
+    asm volatile (
+        "mov $0x10, %%ax \n" // Reload segments with kernel data selector
+        "mov %%ax, %%ds \n"
+        "mov %%ax, %%es \n"
+        "mov %%ax, %%fs \n"
+        "mov %%ax, %%gs \n"
+        "mov %%ax, %%ss \n"
+
+        "pushq $0x8      \n"   // CS reload
+        "lea 1f(%%rip), %%rax \n"
+        "push %%rax     \n"
+        "lretq          \n"
+        "1:             \n" // Execution continues here after CS reload
+        :
+        :
+        : "rax", "memory"
+    );
+}
+
+/*
+ * get_set_entry - Sets a GDT entry
+ * @num:    Number of the entry (index in GDT)
+ * @flags:  Flags (Granularity, Size, Long mode)
+ * @access: Access byte (contains Descriptor Privilege Level)
+ *
+ * This function fills a GDT entry with the specified @flags
+ * and @access byte. The base and limit fields are left to zero
+ * because we don't use segmentation for memory management.
+ */
+
+static void gdt_set_entry(int num, uint8_t flags, uint8_t access)
+{
+    uint64_t gdt_entry = 0;
+
+    gdt_entry |= (access << 8);
+    gdt_entry |= (flags << 20);
+    
+    // Rest (base, limit) is always zero
+    gdt_entries[num] = gdt_entry << 32;
+}
+
+/*
+ * gdt_set_tss - Setup the TSS entry in the GDT
+ * @num: Number of the entry (index in GDT)
+ *
+ * This function sets up a Task State Segment entry
+ * in the Global Descriptor Table.
+ *
+ * The entry is 128-bit long, so it actually takes
+ * two 64-bit GDT entries.
+ */
+static void gdt_set_tss(int num)
+{
+    uint64_t tss_base = (uint64_t)&tss;
+    uint64_t tss_limit = sizeof(struct tss) - 1;
+
+    tss.iopb = sizeof(struct tss);
+
+    uint64_t tss_low = 0;
+    tss_low |= (tss_limit & 0xFFFFULL);
+    tss_low |= (tss_base & 0xFFFFFFULL) << 16;
+    tss_low |= 0x89ULL << 40;
+    tss_low |= ((tss_limit >> 16) & 0xFULL) << 48;
+    tss_low |= ((tss_base >> 24) & 0xFFULL) << 56;
+
+    uint64_t tss_high = (tss_base >> 32) & 0xFFFFFFFFULL;
+
+    gdt_entries[num] = tss_low;
+    gdt_entries[num + 1] = tss_high;
+}
+
+/*
+ * gdt_init - Global Descriptor Table initialization 
+ *
+ * This function loads a new GDT in the CPU.
+ * It contains a null descriptor, kernel code and data
+ * segments, and user code and data segments.
+ * However, we do not use segmentation to manage memory on
+ * 64-bit x86, as it's deprecated. Instead, we use paging. 
+ */
+void gdt_init()
+{
+    gdt_set_entry(0, 0, 0); // Null descriptor  (0x0)
+    gdt_set_entry(1, 0xA, 0x9B); // Kernel code (0x8)
+    gdt_set_entry(2, 0xC, 0x93); // Kernel data (0x10)
+    gdt_set_entry(3, 0xA, 0xFB); // User code   (0x18)
+    gdt_set_entry(4, 0xC, 0xF3); // User data   (0x20)
+    gdt_set_tss(5);                            // TSS         (0x28)
+
+    // The -1 subtraction is some wizardry explained in the OSDev wiki -> GDT
+    gdtr.limit = NUM_GDT_ENTRIES * sizeof(uint64_t) - 1;
+    gdtr.address = (uint64_t)gdt_entries;
+
+    // Load the GDT we created, flush the old one
+    gdt_load();
+    gdt_flush();
+
+    // Load task register with new TSS
+    asm volatile("ltr %%ax" : : "a"(TSS_SEGMENT) : "memory");
+
+    DEBUG("GDT initialized");
+}

src/arch/x86/idt.c

@@ -77,7 +77,7 @@ void idt_init()
         idt_set_entry(i, vector_0_handler + (i*16), 0);
     }
 
-    idt_set_entry(0x80, vector_128_handler, 0);
+    idt_set_entry(0x80, vector_128_handler, 3);
 
     idt_load(&idt);
     DEBUG("IDT initialized");
@@ -108,7 +108,7 @@ static inline uint64_t read_cr2(void)
  * Also displays an interpretation of the thrown error code.
  * Then halts the system. We could implement demand paging later.
  */
-static void page_fault_handler(struct cpu_status_t* ctx)
+static void page_fault_handler(struct cpu_status* ctx)
 {
     // It could be used to remap pages etc. to fix the fault, but right now what I'm more
     // interested in is getting more info out of those numbers cause i'm lost each time i have 
@@ -149,7 +149,7 @@ static void page_fault_handler(struct cpu_status_t* ctx)
  * Shows detail about a General Protection Fault,
  * and what may have caused it. Halts the system.
  */
-static void gp_fault_handler(struct cpu_status_t* ctx)
+static void gp_fault_handler(struct cpu_status* ctx)
 {
     DEBUG("\x1b[38;5;231mGeneral Protection Fault at rip=0x%p, err=%u (%s)\x1b[0m",
     ctx->iret_rip,
@@ -185,7 +185,7 @@ static void gp_fault_handler(struct cpu_status_t* ctx)
  * Return:
  * <context> - CPU context after interrupt
  */
-struct cpu_status_t* interrupt_dispatch(struct cpu_status_t* context)
+struct cpu_status* interrupt_dispatch(struct cpu_status* context)
 {
     if (context == NULL) {
         panic(NULL, "Interrupt dispatch recieved NULL context!");

src/arch/x86/init.c

@@ -4,7 +4,7 @@
  * @license GPL-3.0-only
  */
 
-#include <mem/gdt.h>
+#include <arch/gdt.h>
 #include <stdint.h>
 #include <arch/x86.h>
 #include <kernel.h>
@@ -29,6 +29,26 @@ static void x86_overwrite_pat()
     wrmsr(0x277, pat);
 }
 
+/*
+ * x86_enable_fpu - Enable Floating Point Unit
+ *
+ * This function enables the Floating Point Unit,
+ * which allows the CPU to do floating point
+ * operations.
+ *
+ * Here we do not check for FPU support but we
+ * should. However most processors support it.
+ */
+static void x86_enable_fpu()
+{
+    size_t cr4;
+	__asm__ volatile("mov %%cr4, %0" : "=r"(cr4));
+	cr4 |= 0x200;
+	__asm__ volatile("mov %0, %%cr4" :: "r"(cr4));
+    uint16_t cw = 0x37F; // control word
+    asm volatile("fldcw %0" :: "m"(cw));
+}
+
 /*
  * x86_arch_init - Initialize x86 CPU structures
  *
@@ -42,6 +62,7 @@ static void x86_overwrite_pat()
 void x86_arch_init()
 {
     x86_overwrite_pat();
+    x86_enable_fpu();
     x86_cpu_identification();
     idt_init();
     gdt_init();

src/arch/x86/syscall.c

@@ -4,23 +4,157 @@
  * @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>
 
-struct cpu_status_t* syscall_handler(struct cpu_status_t* regs)
+extern struct process* current_process;
+
+// Return fd on success, -errno on error
+int sys_open(const char* filename, int flags)
 {
-    DEBUG("Syscall %lx with argument %lx", regs->rdi, regs->rsi);
+    if (tar_exists(filename) < 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, filename, PROCESS_NAME_MAX - 1);
+    return fd;
+}
 
-    switch (regs->rdi)
+// Return 0 on success, -EBADFD if invalid FD
+int sys_close(int fd)
+{
+    if (fd < 0 || 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;
+}
+
+// Should return the number of bytes read
+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;
+}
+
+// TODO: Should have a return value: number of bytes written on success, -1 on error (errno set) 
+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;
+    }
+}
+
+int sys_exit(int error_code)
+{
+    current_process->status = DEAD;
+    DEBUG("(pid=%u, name=%s)", current_process->pid, current_process->name);
+    return 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 60:
+            DEBUG("sys_exit(error_code=%d)", regs->rdi);
+            regs->rax = sys_exit(regs->rdi);
             break;
         default:
-            regs->rsi = 0xdeadbeef;
+            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;
 }

src/boot/boot.c

@@ -13,6 +13,9 @@
 #include <limine.h>
 #include <stddef.h>
 
+__attribute__((used, section(".limine_requests_start")))
+volatile LIMINE_REQUESTS_START_MARKER;
+
 __attribute__((used, section(".limine_requests")))
 volatile struct limine_framebuffer_request framebuffer_request = {
 	.id = LIMINE_FRAMEBUFFER_REQUEST,
@@ -43,8 +46,11 @@ volatile struct limine_boot_time_request date_request = {
 	.revision = 0
 };
 
-__attribute__((used, section(".limine_requests_start")))
+__attribute__((used, section(".limine_requests")))
-volatile LIMINE_REQUESTS_START_MARKER;
+volatile struct limine_module_request module_request = {
+	.id = LIMINE_MODULE_REQUEST,
+	.revision = 0
+};
 
 __attribute__((used, section(".limine_requests_end")))
 volatile LIMINE_REQUESTS_END_MARKER;
@@ -58,4 +64,5 @@ void populate_boot_context(struct boot_context* ctx)
 	ctx->hhdm = hhdm_request.response ? hhdm_request.response : NULL;
 	ctx->kaddr = kerneladdr_request.response ? kerneladdr_request.response : NULL;
 	ctx->bootdate = date_request.response ? date_request.response : NULL;
+	ctx->module = module_request.response ? module_request.response : NULL;
 }

src/debug/panic.c

@@ -18,7 +18,7 @@ extern int panic_count;
  */
 void read_rflags(uint64_t rflags)
 {
-	DEBUG("\x1b[38;5;226m%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\x1b[38;5;231m",
+	printf("\x1b[38;5;226m%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\x1b[38;5;231m",
 		CHECK_BIT(rflags, 0) ? "CF " : "", /*carry flag*/
 		CHECK_BIT(rflags, 2) ? "PF " : "", /*parity flag*/
 		CHECK_BIT(rflags, 4) ? "AF " : "", /*auxiliary carry flag*/
@@ -47,7 +47,7 @@ void read_rflags(uint64_t rflags)
  * Will display to terminal if it is initialized, otherwise serial only.
  * Can be called with or without a CPU context.
  */
-void panic(struct cpu_status_t* ctx, const char* str)
+void panic(struct cpu_status* ctx, const char* str)
 {
 	CLEAR_INTERRUPTS;
 	panic_count += 1;

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/kbd/ps2.c

@@ -289,15 +289,21 @@ int keyboard_getline(char* output, size_t size)
     
     // Read until Enter is pressed
     while ((c = keyboard_getchar()) != 0x0A) {
-        if (index == size-1) {
+        if (c == '\b') {
-            output[index] = c;
+            if (index > 0) {
-            output[index+1] = '\0';
+                index--;
-            return index;
+                output[index] = '\0';
+                printf(" \b");
             }
-        output[index] = c;
+            continue;
-        index++;
         }
-    output[index+1] = '\0';
+
+        if (index >= size-1) {
+            continue;
+        }
+        output[index++] = c;
+    }
+    output[index] = '\0';
     return index;
 }
 

src/io/serial/serial.c

@@ -11,7 +11,7 @@
 extern struct init_status init;
 
 extern int panic_count;
-struct spinlock_t serial_lock = {0};
+struct spinlock serial_lock = {0};
 
 /*
  * outb - Writes a byte to a CPU port

src/io/term/term.c

@@ -29,8 +29,8 @@ because this shitty implementation will be replaced one day by Flanterm
 extern struct flanterm_context* ft_ctx;
 extern struct init_status init;
 
-struct spinlock_t term_lock = {0};
+struct spinlock term_lock = {0};
-struct spinlock_t printf_lock = {0};
+struct spinlock printf_lock = {0};
 
 extern int panic_count;
 

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>
@@ -11,6 +12,7 @@
 #include <stdint.h>
 #include <kernel.h>
 #include <time/date.h>
+#include <mem/kheap.h>
 
 __attribute__((noinline))
 void smash_it()
@@ -41,15 +43,16 @@ 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\r\n"
-                        "but rather a toy to play around in the meantime.\r\n\r\n"
+                            "\tpanic   - trigger a test panic\r\n"
-                            "clear   - clear the screen\r\n"
+                            "\tsyscall - trigger int 0x80\r\n"
-                            "panic   - trigger a test panic\r\n"
+                            "\tpf      - trigger a page fault\r\n"
-                            "syscall - trigger int 0x80\r\n"
+                            "\tnow     - get current date\r\n"
-                            "pf      - trigger a page fault\r\n"
+                            "\tsmash   - smash the stack\r\n"
-                            "now     - get current date\r\n"
+                            "\tmem     - get used heap info\r\n"
-                            "smash   - smash the stack\r\n");
+                            "\tload    - load an user executable\r\n"
+                            "\tlist    - list initfs.tar contents\r\n");
             continue;
         }
 
@@ -89,6 +92,21 @@ void pedicel_main(void* arg)
             continue;
         }
 
+        if (strncmp(input_buf, "mem", 3) == 0) {
+            kheap_info();
+            continue;
+        }
+
+        if (strncmp(input_buf, "load", 4) == 0) {
+            loader_load_raw();
+            continue;
+        }
+
+        if (strncmp(input_buf, "list", 4) == 0) {
+            tar_list();
+            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

@@ -9,7 +9,7 @@
 #include <limine.h>
 #include <io/term/term.h>
 #include <io/serial/serial.h>
-#include <mem/gdt.h>
+#include <arch/gdt.h>
 #include <mem/utils.h>
 #include <kernel.h>
 #include <time/timer.h>
@@ -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")))
@@ -62,10 +63,13 @@ extern volatile struct limine_memmap_request memmap_request;
 extern volatile struct limine_hhdm_request hhdm_request;
 extern volatile struct limine_kernel_address_request kerneladdr_request;
 extern volatile struct limine_boot_time_request date_request;
+extern volatile struct limine_module_request module_request;
 
-extern struct process_t* processes_list;
+struct limine_file* file;
-extern struct process_t* current_process;
+
-struct process_t* idle_proc;
+extern struct process* processes_list;
+extern struct process* current_process;
+struct process* idle_proc;
 
 void idle_main(void* arg)
 {
@@ -74,14 +78,6 @@ void idle_main(void* arg)
 	}
 }
 
-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;
 
 /*
@@ -115,8 +111,15 @@ void kmain()
 	
 	process_init();
 	idle_proc = process_create("idle", (void*)idle_main, 0);
-	process_create("pedicel", (void*)pedicel_main, 0);
 
+	if (!boot_ctx.module) {
+		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);

src/mem/gdt.c

@@ -1,107 +0,0 @@
-/*
- * @author  xamidev <xamidev@riseup.net>
- * @brief   Global Descriptor Table (for legacy reasons)
- * @license GPL-3.0-only
- */
-
-#include <mem/gdt.h>
-#include <stdint.h>
-#include <io/serial/serial.h>
-#include <kernel.h>
-
-// Descriptors are 8-byte wide (64bits)
-// So the selectors will be (in bytes): 0x0, 0x8, 0x10, 0x18, etc..
-uint64_t gdt_entries[NUM_GDT_ENTRIES];
-struct GDTR gdtr;
-
-/*
- * gdt_load - Loads Global Descriptor Table
- */
-static void gdt_load()
-{
-    asm("lgdt %0" : : "m"(gdtr));
-}
-
-/*
- * gdt_flush - Flushes the Global Descriptor Table
- *
- * This function loads new Segment Selectors to make
- * the GDT changes take effect
- */
-static void gdt_flush()
-{
-    // Here, 0x8 is the kernel code selector
-    // and 0x10 is the kernel data selector
-    asm volatile (
-        "mov $0x10, %%ax \n" // Reload segments with kernel data selector
-        "mov %%ax, %%ds \n"
-        "mov %%ax, %%es \n"
-        "mov %%ax, %%fs \n"
-        "mov %%ax, %%gs \n"
-        "mov %%ax, %%ss \n"
-
-        "pushq $0x8      \n"   // CS reload
-        "lea 1f(%%rip), %%rax \n"
-        "push %%rax     \n"
-        "lretq          \n"
-        "1:             \n" // Execution continues here after CS reload
-        :
-        :
-        : "rax", "memory"
-    );
-}
-
-/*
- * gdt_init - Global Descriptor Table initialization 
- *
- * This function loads a new GDT in the CPU.
- * It contains a null descriptor, kernel code and data
- * segments, and user code and data segments.
- * However, we do not use segmentation to manage memory on
- * 64-bit x86, as it's deprecated. Instead, we use paging. 
- */
-void gdt_init()
-{
-    // Null descriptor (required)
-    gdt_entries[0] = 0;
-
-    // Kernel code segment
-    uint64_t kernel_code = 0;
-    kernel_code |= 0b1101 << 8; // Selector type: accessed, read-enable, no conforming
-    kernel_code |= 1 << 12; // not a system descriptor
-    kernel_code |= 0 << 13; // DPL field = 0
-    kernel_code |= 1 << 15; // Present
-    kernel_code |= 1 << 21; // Long mode
-
-    // Left shift 32 bits so we place our stuff in the upper 32 bits of the descriptor.
-    // The lower 32 bits contain limit and part of base and therefore are ignored in Long Mode
-    // (because we'll use paging; segmentation is used only for legacy)
-    gdt_entries[1] = kernel_code << 32;
-
-    uint64_t kernel_data = 0;
-    kernel_data |= 0b0011 << 8;
-    kernel_data |= 1 << 12;
-    kernel_data |= 0 << 13;
-    kernel_data |= 1 << 15;
-    kernel_data |= 1 << 21;
-
-    gdt_entries[2] = kernel_data << 32;
-
-    // We re-use the kernel descriptors here, and just update their DPL fields
-    // (Descriptor privilege level) from ring 0 -> to ring 3 (userspace)
-    uint64_t user_code = kernel_code | (3 << 13);
-    gdt_entries[3] = user_code;
-
-    uint64_t user_data = kernel_data | (3 << 13);
-    gdt_entries[4] = user_data;
-
-    // The -1 subtraction is some wizardry explained in the OSDev wiki -> GDT
-    gdtr.limit = NUM_GDT_ENTRIES * sizeof(uint64_t) - 1;
-    gdtr.address = (uint64_t)gdt_entries;
-
-    // Load the GDT we created, flush the old one
-    gdt_load();
-    gdt_flush();
-
-    DEBUG("GDT initialized");
-}

src/mem/kheap.c

@@ -17,7 +17,7 @@ extern uint64_t kernel_virt_base;
 
 uintptr_t kheap_start;
 
-static struct heap_block_t* head = NULL;
+static struct heap_block* head = NULL;
 static uintptr_t end;
 
 // Kernel root table (level 4)
@@ -55,8 +55,8 @@ void kheap_init()
     end = current_addr;
 
     // Give linked list head its properties
-    head = (struct heap_block_t*)kheap_start;
+    head = (struct heap_block*)kheap_start;
-    head->size = (end-kheap_start) - sizeof(struct heap_block_t);
+    head->size = (end-kheap_start) - sizeof(struct heap_block);
     head->free = true;
     head->next = NULL;
     DEBUG("Kernel heap initialized, head=0x%p, size=%u bytes", head, head->size);
@@ -80,16 +80,16 @@ void* kmalloc(size_t size)
     if (!size) return NULL;
     size = ALIGN(size);
 
-    struct heap_block_t* curr = head;
+    struct heap_block* curr = head;
 
     while (curr) {
         // Is block free and big enough for us?
         if (curr->free && curr->size >= size) {
             // We split the block if it is big enough
-            if (curr->size >= size + sizeof(struct heap_block_t) + 16) {
+            if (curr->size >= size + sizeof(struct heap_block) + 16) {
-                struct heap_block_t* split = (struct heap_block_t*)((uintptr_t)curr + sizeof(struct heap_block_t) + size);
+                struct heap_block* split = (struct heap_block*)((uintptr_t)curr + sizeof(struct heap_block) + size);
                 
-                split->size = curr->size - size - sizeof(struct heap_block_t);
+                split->size = curr->size - size - sizeof(struct heap_block);
                 split->free = true;
                 split->next = curr->next;
 
@@ -99,7 +99,7 @@ void* kmalloc(size_t size)
 
             // Found a good block, we return it
             curr->free = false;
-            return (void*)((uintptr_t)curr + sizeof(struct heap_block_t));
+            return (void*)((uintptr_t)curr + sizeof(struct heap_block));
         }
         // Continue browsing the list if nothing good was found yet
         curr = curr->next;
@@ -127,11 +127,11 @@ void kfree(void* ptr)
     if (!ptr) return;
 
     // Set it free!
-    struct heap_block_t* block = (struct heap_block_t*)((uintptr_t)ptr - sizeof(struct heap_block_t));
+    struct heap_block* block = (struct heap_block*)((uintptr_t)ptr - sizeof(struct heap_block));
     block->free = true;
 
     // merge adjacent free blocks (coalescing)
-    struct heap_block_t* curr = head;
+    struct heap_block* curr = head;
     while (curr && curr->next) {
         if (curr->free && curr->next->free) {
             curr->size += sizeof(*curr) + curr->next->size;
@@ -158,3 +158,32 @@ void* kalloc_stack()
     uint8_t* ptr = kmalloc(PROCESS_STACK_SIZE); // As it's out of kmalloc, stack is already mapped into kernel space
     return ptr ? ptr+PROCESS_STACK_SIZE : NULL;
 }
+
+/*
+ * kheap_info - Display heap info
+ *
+ * This function writes the size of the heap (total),
+ * the number of allocated bytes, and the number of
+ * free bytes to the standard output.
+ */
+void kheap_info()
+{
+    uint64_t free_bytes = 0;
+    struct heap_block* curr = (struct heap_block*)kheap_start;
+
+    while (curr) {
+        if (curr->free == true) {
+            free_bytes += curr->size;
+        }
+        curr = curr->next;
+    }
+
+    uint64_t total = end-kheap_start;
+
+    printf("total=% 8u bytes (%u kB)\r\n"
+                "alloc=% 8u bytes (%u kB)\r\n"
+                " free=% 8u bytes (%u kB)\r\n",
+                total, (total)/1000,
+                total-free_bytes, (total-free_bytes)/1000,
+                free_bytes, free_bytes/1000);
+}

src/mem/paging.c

@@ -99,30 +99,46 @@ void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_
 
     uint64_t *pdpt, *pd, *pt;
 
+    // Any parent entry on a userspace mapping must also carry PTE_USER,
+    // otherwise CPL3 accesses fault even if the final PTE is user.
+    uint64_t parent_flags = PTE_PRESENT | PTE_WRITABLE;
+    if (flags & PTE_USER) {
+        parent_flags |= PTE_USER;
+    }
+
     // PML4
     // If the entry at index is not present, allocate enough space for it
     // then populate the entry with correct addr + flags
     if (!(root_table[pml4_i] & PTE_PRESENT)) {
         pdpt = alloc_page_table();
-        root_table[pml4_i] = VIRT_TO_PHYS(pdpt) | PTE_PRESENT | PTE_WRITABLE;
+        root_table[pml4_i] = VIRT_TO_PHYS(pdpt) | parent_flags;
     } else {
         pdpt = (uint64_t *)PHYS_TO_VIRT(root_table[pml4_i] & PTE_ADDR_MASK);
+        if (flags & PTE_USER) {
+            root_table[pml4_i] |= PTE_USER;
+        }
     }
 
     // PDPT: same here
     if (!(pdpt[pdpt_i] & PTE_PRESENT)) {
         pd = alloc_page_table();
-        pdpt[pdpt_i] = VIRT_TO_PHYS(pd) | PTE_PRESENT | PTE_WRITABLE;
+        pdpt[pdpt_i] = VIRT_TO_PHYS(pd) | parent_flags;
     } else {
         pd = (uint64_t *)PHYS_TO_VIRT(pdpt[pdpt_i] & PTE_ADDR_MASK);
+        if (flags & PTE_USER) {
+            pdpt[pdpt_i] |= PTE_USER;
+        }
     }
 
     // PD: and here
     if (!(pd[pd_i] & PTE_PRESENT)) {
         pt = alloc_page_table();
-        pd[pd_i] = VIRT_TO_PHYS(pt) | PTE_PRESENT | PTE_WRITABLE;
+        pd[pd_i] = VIRT_TO_PHYS(pt) | parent_flags;
     } else {
         pt = (uint64_t *)PHYS_TO_VIRT(pd[pd_i] & PTE_ADDR_MASK);
+        if (flags & PTE_USER) {
+            pd[pd_i] |= PTE_USER;
+        }
     }
 
     // PT: finally, populate the page table entry

src/mem/vmm.c

@@ -13,6 +13,7 @@ in a specified virtual space
 compared to the PMM which allocs/frees 4kb frames ("physical pages").
 */
 
+#include "config.h"
 #include <mem/vmm.h>
 #include <mem/paging.h>
 #include <stddef.h>
@@ -225,6 +226,43 @@ void* vmm_alloc_region(uint64_t* pml4, size_t pages, uint64_t flags)
     return NULL;
 }
 
+/*
+ * vmm_map_user_stack - Map a user stack
+ * @pml4: the user process's PML4
+ *
+ * This function maps and allocates a userspace
+ * stack in the user @pml4 provided, according
+ * to constants USER_STACK_TOP and USER_STACK_PAGES.
+ *
+ * Return:
+ * <addr> - User stack top address
+ */
+
+uintptr_t vmm_alloc_user_stack(uint64_t* pml4)
+{
+    uintptr_t stack_top = USER_STACK_TOP;
+    size_t stack_size = USER_STACK_PAGES*PAGE_SIZE;
+
+    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;
+}
+
+uintptr_t vmm_alloc_user_code(uint64_t* pml4, void* code_addr, uint64_t code_size)
+{
+    uintptr_t code_start = USER_CODE_START;
+    
+    // Round code_size up to next page boundary
+    uint64_t code_size_aligned = (code_size + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1);
+    
+    for (uint64_t i=code_start; i<code_start+code_size_aligned; i+=PAGE_SIZE) {
+        vmm_map(pml4, i, PTE_PRESENT | PTE_WRITABLE | PTE_USER);
+    } 
+    
+    return code_start;
+}
+
 void vmm_init()
 {
     // NO U

src/sched/process.c

@@ -4,20 +4,23 @@
  * @license GPL-3.0-only
  */
 
+#include "mem/paging.h"
+#include "mem/vmm.h"
 #include <stddef.h>
 #include <sched/process.h>
 #include <mem/kheap.h>
 #include <kernel.h>
 #include <string/string.h>
-#include <mem/gdt.h>
+#include <arch/gdt.h>
 #include <config.h>
 #include <io/serial/serial.h>
-
 #include <io/term/flanterm.h>
+#include <mem/utils.h>
+
 extern struct flanterm_context* ft_ctx;
 
-struct process_t* processes_list;
+struct process* processes_list;
-struct process_t* current_process;
+struct process* current_process;
 
 extern uint64_t *kernel_pml4;
 
@@ -39,10 +42,10 @@ void process_init()
  * This function prints the linked list of processes
  * to the DEBUG output.
  */
-void process_display_list(struct process_t* processes_list)
+void process_display_list(struct process* processes_list)
 {
     int process_view_id = 0;
-    struct process_t* tmp = processes_list;
+    struct process* tmp = processes_list;
     while (tmp != NULL) {
         DEBUG("{%d: %p} -> ", process_view_id, tmp);
         tmp = tmp->next;
@@ -64,11 +67,11 @@ void process_display_list(struct process_t* processes_list)
  * Return:
  * <proc> - pointer to created process
  */
-struct process_t* process_create(char* name, void(*function)(void*), void* arg)
+struct process* process_create(char* name, void(*function)(void*), void* arg)
 {
     CLEAR_INTERRUPTS;
-    struct process_t* proc = (struct process_t*)kmalloc(sizeof(struct process_t));
+    struct process* proc = (struct process*)kmalloc(sizeof(struct process));
-    struct cpu_status_t* ctx = (struct cpu_status_t*)kmalloc(sizeof(struct cpu_status_t));
+    struct cpu_status* ctx = (struct cpu_status*)kmalloc(sizeof(struct cpu_status));
 
     // No more memory?
     if (!proc) return NULL;
@@ -95,6 +98,8 @@ struct process_t* 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);
@@ -108,7 +113,7 @@ struct process_t* process_create(char* name, void(*function)(void*), void* arg)
  * @processes_list: pointer to the head of the linked list
  * @process:        process to add at the end of the linked list
  */
-void process_add(struct process_t** processes_list, struct process_t* process)
+void process_add(struct process** processes_list, struct process* process)
 {
     if (!process) return;
     process->next = NULL;
@@ -119,7 +124,7 @@ void process_add(struct process_t** processes_list, struct process_t* process)
         return;
     }
 
-    struct process_t* tmp = *processes_list;
+    struct process* tmp = *processes_list;
     while (tmp->next != NULL) {
         tmp = tmp->next;
     }
@@ -132,7 +137,7 @@ void process_add(struct process_t** processes_list, struct process_t* process)
  * @processes_list: pointer to head of linked list
  * @process:        the process to delete from the list
  */
-void process_delete(struct process_t** processes_list, struct process_t* process)
+void process_delete(struct process** processes_list, struct process* process)
 {
     if (!processes_list || !*processes_list || !process) return;
 
@@ -144,7 +149,7 @@ void process_delete(struct process_t** processes_list, struct process_t* process
         return;
     }
 
-    struct process_t* tmp = *processes_list;
+    struct process* tmp = *processes_list;
     while (tmp->next && tmp->next != process) {
         tmp = tmp->next;
     }
@@ -167,7 +172,7 @@ void process_delete(struct process_t** processes_list, struct process_t* process
  * Return:
  * <process->next> - process right after the one specified
  */
-struct process_t* process_get_next(struct process_t* process)
+struct process* process_get_next(struct process* process)
 {
     if (!process) return NULL;
     return process->next;
@@ -197,3 +202,97 @@ void process_exit()
         asm("hlt");
     }
 }
+
+/*
+ * process_jump_to_user - Jump to userland
+ * @stack_top: Address of the top of the user stack
+ * @user_code: Address of the first instruction of user code
+ */
+void process_jump_to_user(uintptr_t stack_top, uintptr_t user_code)
+{
+    // 0x20 | 3 = 0x23 (user data segment | 3)
+    // 0x18 | 3 = 0x1B (user code segment | 3)
+    asm volatile(" \
+        push $0x23 \n\
+        push %0 \n\
+        push $0x202 \n\
+        push $0x1B \n\
+        push %1 \n\
+        iretq \n\
+        " :: "r"(stack_top), "r"(user_code));
+}
+
+extern struct tss tss;
+
+/*
+ * process_create_user_raw - Create a new user process from raw binary
+ * @file: pointer to beginning of binary
+ * @size: size of the binary
+ * @name: name for the new process
+ *
+ * This function takes an executable loaded in memory
+ * and maps its code, a user stack, sets the TSS RSP0
+ * for interrupts, and finally jumps to the user code.
+ */
+void process_create_user_raw(char* file, int size, char* name)
+{
+    CLEAR_INTERRUPTS;
+    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);
+    
+    proc->context->iret_rsp = stack_top;
+    proc->context->iret_rip = code;
+    proc->kernel_stack = kalloc_stack();
+    if (!proc->kernel_stack) panic(NULL, "failed to allocate kernel stack");
+
+    // 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));
+    memcpy((uint64_t*)code, exec_addr, exec_size);
+    load_cr3(VIRT_TO_PHYS((uint64_t)kernel_pml4));
+
+    process_add(&processes_list, proc);
+    DEBUG("user process '%s' (pid=%u) enqueued for scheduling", name, proc->pid);
+    SET_INTERRUPTS;
+}

src/sched/scheduler.c

@@ -9,10 +9,13 @@
 #include <mem/paging.h>
 #include <stdint.h>
 #include <io/serial/serial.h>
+#include <arch/gdt.h>
 
-extern struct process_t* processes_list;
+extern struct process* processes_list;
-extern struct process_t* current_process;
+extern struct process* current_process;
-extern struct process_t* idle_proc;
+extern struct process* idle_proc;
+
+extern struct tss tss;
 
 /*
  * scheduler_init - Choose the first process
@@ -20,6 +23,7 @@ extern struct process_t* idle_proc;
 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);
 }
 
 /*
@@ -32,49 +36,63 @@ void scheduler_init()
  * Return:
  * <context> - CPU context for next process
  */
-struct cpu_status_t* scheduler_schedule(struct cpu_status_t* context)
+struct cpu_status* scheduler_schedule(struct cpu_status* context)
 {
     if (context == NULL) {
         panic(NULL, "Scheduler called with NULL 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_t* 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/sched/spinlock.c

@@ -16,7 +16,7 @@
  * Saves the RFLAGS register, then acquires a lock.
  * Pause instruction is used to ease the CPU.
  */
-void spinlock_acquire(struct spinlock_t* lock)
+void spinlock_acquire(struct spinlock* lock)
 {
     uint64_t rflags;
     asm volatile("pushfq ; pop %0 ; cli" : "=rm"(rflags) : : "memory");
@@ -36,7 +36,7 @@ void spinlock_acquire(struct spinlock_t* lock)
  * unlocks it (clears locked state).
  * RFLAGS is then restored.
  */
-void spinlock_release(struct spinlock_t* lock)
+void spinlock_release(struct spinlock* lock)
 {
     uint64_t rflags = lock->rflags;
     __atomic_clear(&lock->locked, __ATOMIC_RELEASE);

src/string/string.c

@@ -99,3 +99,11 @@ int strncmp(const char* s1, const char* s2, size_t n)
         return ( *(unsigned char *)s1 - *(unsigned char *)s2 );
     }
 }
+
+// BSD implementation
+size_t strlen(const char* str)
+{
+        const char* s;
+        for (s = str; *s; ++s);
+        return (s - str);
+}

user/Makefile

@@ -0,0 +1,18 @@
+CC := x86_64-elf-gcc
+CC_FLAGS := -ffreestanding -nostdlib -fno-pic -mno-red-zone -Ilibc
+
+LD := x86_64-elf-ld
+
+BUILDDIR := ../build
+LIBDIR := libc
+
+all: pedicel apex
+
+pedicel:
+	nasm -f bin pedicel.S -o $(BUILDDIR)/pedicel.raw
+
+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

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/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 +1,30 @@
+#pragma once
+
+// 3 because 3 arguments to the call, get it??
+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)
+        : "memory"
+    );
+
+    return ret;
+}
+
+static inline void write(int fd, const char* buf, long len) {
+    syscall3(1, fd, (long)buf, len);
+}
+
+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