Kshell: load executable command

Reviewed-on: #18

Makefile

@@ -16,12 +16,15 @@ 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 bin user/pedicel.S -o $(BUILDDIR)/pedicel
+	tar cvf $(BUILDDIR)/initfs.tar -C $(BUILDDIR) hello pedicel
+
 	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 +45,15 @@ limine/limine:
 	git clone https://github.com/limine-bootloader/limine.git --branch=v9.x-binary --depth=1
 	$(MAKE) -C limine
 
+initfs:
+
 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/

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 +40,9 @@ 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.
+
+Name | Number (%rax) | arg0 (%rdi) | arg1 (%rsi) | arg2 (%rdx) |
+|---|---|---|---|---|
+| sys_write | 1 | unsigned int fd | const char* buf | size_t count | | 
+| 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_PATCH "109"
+#define PEPPEROS_VERSION_MINOR "1"
+#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"\

include/fs/initfs.h

@@ -0,0 +1,15 @@
+/*
+ * @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);
+
+#endif

include/kernel.h

@@ -10,7 +10,8 @@
 #include "limine.h"
 enum ErrorCodes {
 	ENOMEM,
-	EIO
+	EIO,
+	ENOENT
 };
 
 #define CLEAR_INTERRUPTS __asm__ volatile("cli")
@@ -46,6 +47,8 @@ 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

@@ -38,6 +38,7 @@ void process_exit(void);
 
 void process_display_list(struct process* processes_list);
 
-void process_create_user(struct limine_file* file);
+void process_create_user(struct limine_file* file, char* name);
+void process_create_user_raw(char* file, int size, char* name);
 
 #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

@@ -6,4 +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

@@ -9,6 +9,9 @@
 #include <kernel.h>
 #include <stddef.h>
 #include <io/term/term.h>
+#include <sched/process.h>
+
+extern struct process* current_process;
 
 void sys_write(unsigned int fd, const char* buf, size_t count)
 {
@@ -20,10 +23,19 @@ void sys_write(unsigned int fd, const char* buf, size_t count)
             break;
 
         case 2: //stderr
+            for (size_t i=0; i<count; i++) {
+                internal_putc(buf[i], NULL);
+            }
             break;
     }
 }
 
+void sys_exit(int error_code)
+{
+    current_process->status = DEAD;
+    DEBUG("exiting process PID=%u name=%s", current_process->pid, current_process->name);
+}
+
 /*
  * syscall_handler - System call dispatcher
  * @regs: CPU state
@@ -53,6 +65,9 @@ struct cpu_status* syscall_handler(struct cpu_status* regs)
         case 1: //sys_write
             sys_write(regs->rdi, (char*)regs->rsi, regs->rdx);
             break;
+        case 60: //sys_exit
+            sys_exit(regs->rdi);
+            break;
         default:
             regs->rax = 0xbad515ca11;
             break;

src/fs/initfs.c

@@ -0,0 +1,76 @@
+/*
+ * @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 
+ * $0        - 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 0;
+}
+
+/*
+ * 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/kapps/kshell.c

@@ -51,7 +51,8 @@ void pedicel_main(void* arg)
                             "pf      - trigger a page fault\r\n"
                             "now     - get current date\r\n"
                             "smash   - smash the stack\r\n"
-                            "mem     - get used heap info\r\n");
+                            "mem     - get used heap info\r\n"
+                            "load    - load an user executable\r\n");
             continue;
         }
 
@@ -96,6 +97,11 @@ void pedicel_main(void* arg)
             continue;
         }
 
+        if (strncmp(input_buf, "load", 4) == 0) {
+            loader_load_raw();
+            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")))
@@ -118,22 +119,19 @@ 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

@@ -246,17 +246,19 @@ 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;
 }
 
 uintptr_t vmm_alloc_user_code(uint64_t* pml4, void* code_addr, uint64_t code_size)
 {
     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);
+    
+    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;
 }

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 +222,101 @@ 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
  * @file: pointer to Limine file structure
+ * @name: name for the new process
  *
  * This function takes a loaded Limine executable
  * module, and maps its code, a user stack, sets the
  * TSS RSP0 for interrupts, and finally jumps to the
  * user code.
  */
-void process_create_user(struct limine_file* file)
+void process_create_user(struct limine_file* file, 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");
+
+    strncpy(proc->name, name, PROCESS_NAME_MAX);
+    memset(ctx, 0, sizeof(struct cpu_status)); // set GP registers to zero
+    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
+
     void* exec_addr = file->address;
     uint64_t exec_size = file->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
-    tss.rsp0 = (uint64_t)(interrupt_stack + sizeof(interrupt_stack));
+    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");
 
-    // 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;
+}
+
+// Same as above but for a raw data pointer (pointing to raw binary, no ELF)
+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");
+
+    strncpy(proc->name, name, PROCESS_NAME_MAX);
+    memset(ctx, 0, sizeof(struct cpu_status)); // set GP registers to zero
+    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
+
+    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

@@ -9,17 +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.
-        current_process = idle_proc;
+        panic(NULL, "current_process is NULL");
     }
-
-    if (current_process == idle_proc && current_process->next == NULL)
-    {
-        return idle_proc->context;
+    
+    if (current_process->context == NULL) {
+        panic(NULL, "current_process->context is NULL");
     }
 
     current_process->context = context;
 
-    for (;;) {
-        struct process* prev_process = current_process;
-        if (current_process->next != NULL) {
-            current_process = current_process->next;
-        } else {
-            current_process = processes_list;
-        }
-        
-        if (current_process != NULL && current_process->status == DEAD) {
-            process_delete(&prev_process, current_process);
-            current_process = NULL;
+    if (current_process->status == DEAD) {
+        struct process* dead_process = current_process;
+        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;
-        } else {
-            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;
         }
+
+        current_process = next_process;
+    } else if (current_process->next != NULL) {
+        current_process = current_process->next;
+    } else {
+        current_process = processes_list;
     }
 
-    //DEBUG("current_process={pid=%u, name='%s', root_page_table[virt]=%p}", current_process->pid, current_process->name, current_process->root_page_table);
+    for (;;) {
+        if (current_process->status == DEAD) {
+            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;
+            }
+
+            current_process = next_process;
+            continue;
+        }
+
+        current_process->status = RUNNING;
+        break;
+    }
+
+    // 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

@@ -98,4 +98,12 @@ int strncmp(const char* s1, const char* s2, size_t n)
     else {
         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/hello.S

@@ -1,7 +1,7 @@
 bits 64
 
 section .data
-    hi db "hi from userland :) we did it man", 0
+    hi db "hi from userland :) we did it man", 0x0A, 0x0d, 0
 
 section .text
 
@@ -9,7 +9,12 @@ hello:
     mov rax, 0x1        ;sys_write
     mov rdi, 0x1        ;stdout
     lea rsi, [rel hi]   ;char* buf
-    mov rdx, 33         ;count
+    mov rdx, 35         ;count
+    int 0x80
+
+.end:
+    mov rax, 0x3C       ;sys_exit
+    mov rdi, 0x0        ;error_code
     int 0x80
 
 .loop:

user/pedicel.S

@@ -0,0 +1,25 @@
+bits 64
+
+section .data
+hello db 0x0A, 0x0D, "User program 2 speaking", 0x0A, 0x0D, 0
+
+section .text
+
+_start:
+    mov rax, 0x1    ;sys_write
+    mov rdi, 0x1    ;stdout
+    lea rsi, [rel hello]
+    mov rdx, 27     ;count
+    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