Stivale Bare Bones
| Difficulty level |
|---|
 Beginner |
| Kernel Designs |
|---|
| Models |
| Other Concepts |
What is stivale?
stivale means "boot" in Italian. It is a boot protocol designed to overcome shortcomings of common boot protocols used by hobbyist OS developers, such as Multiboot.
There are actually 2 revisions of the stivale boot protocol, namely: stivale, and stivale2. The earlier stivale revision (simply "stivale", but for clarity we are going to call it "stivale1" from now on), is a very simple "KISS" boot protocol only supporting what was deemed necessary at the time, but versioning and expandability issues made creating stivale2 a necessity. stivale2 makes use of tags for bootloader writers' and kernel writers' convenience, and to make future expandability and revisioning easier.
stivale is firmware and architecture agnostic, though the only bootloader fully supporting the stivale protocols as of writing this article (2021/01/09) is Limine, which is an x86/x86_64 BIOS and UEFI bootloader, with TomatBoot (now archived) and Sabaton offering limited stivale compliance for x86_64 UEFI exclusively, and aarch64, respectively.
This article will demonstrate how to write 2 small 64-bit higher half stivale kernels in (GNU) C, using the stivale1 and stivale2 protocols, respectively, and boot them using the Limine bootloader.
Should I use stivale1 or stivale2?
Whilst stivale1 is easier to implement (for both bootloader programmers and kernel programmers), it is limited in design, and it lacks some features which are exclusive to stivale2 such as SMP support (multicore), support for mapping the framebuffer as write combining using the MTRRs, and likely more features in the future as they will be added exclusively to stivale2 and not stivale1.
stivale1 is demonstrated here for completeness's sake, but stivale2 should be the go to choice in the future.
stivale1 example
Overview
For this example, we will create these 2 files and place them in the same directory:
- kernel.c
- linker.ld
As one may notice, there is no "entry point" assembly stub, as one is not necessary with either stivale1 or 2, when using a language which can make use of a standard SysV x86 calling convention.
Furthermore, we will download the header file stivale.h which defines structures that we will use to interact with the bootloader from here, and place it in the same directory as the other files.
Obviously, this is just a bare bones example, and one should always refer to the stivale1 specification for more details and information.
kernel.c
This is the kernel "main".
#include <stdint.h>
#include <stddef.h>
#include <stivale.h>
// We need to tell the stivale bootloader where we want our stack to be.
// We are going to allocate our stack as an uninitialised array in .bss.
static uint8_t stack[4096];
// The stivale specification says we need to define a "header structure".
// This structure needs to reside in the .stivalehdr ELF section in order
// for the bootloader to find it. We use this __attribute__ directive to
// tell the compiler to put the following structure in said section.
__attribute__((section(".stivalehdr"), used))
static struct stivale_header stivale_hdr = {
// Let's tell the bootloader where our stack is.
// We need to add the sizeof(stack) since in x86(_64) the stack grows
// downwards.
.stack = (uintptr_t)stack + sizeof(stack),
// These flags can be used to tell the bootloader to enable or disable
// certain features. We turn on bit 0 as we are going to ask for a
// graphical framebuffer. Not setting this bit will make the
// bootloader default to a CGA-compatible text mode instead.
.flags = (1 << 0),
// We set all the framebuffer specifics to 0 as we want the bootloader
// to pick the best it can.
.framebuffer_width = 0,
.framebuffer_height = 0,
.framebuffer_bpp = 0,
// The entry_point member is used to specify an alternative entry
// point that the bootloader should jump to instead of the executable's
// ELF entry point. We do not care about that so we leave it zeroed.
.entry_point = 0
};
// The following will be our kernel's entry point.
void _start(struct stivale_struct *stivale_struct) {
// Let's get the address of the framebuffer.
uint8_t *fb_addr = (uint8_t *)stivale_struct->framebuffer_addr;
// Let's try to paint a few pixels white in the top left, so we know
// that we booted correctly.
for (size_t i = 0; i < 128; i++) {
fb_addr[i] = 0xff;
}
// We're done, just hang...
for (;;) {
asm ("hlt");
}
}
linker.ld
This is going to be our linker script describing where our sections will end up in memory.
/* Tell the linker that we want the symbol _start to be our entry point */
ENTRY(_start)
SECTIONS
{
/* We wanna be placed in the higher half, 2MiB above 0 in physical memory. */
. = 0xffffffff80200000;
/* We place the .stivalehdr section containing the header in its own section, */
/* and we use the KEEP directive on it to make sure it doesn't get discarded. */
.stivalehdr : {
KEEP(*(.stivalehdr))
}
/* Then let's place all the other traditional executable sections afterwards. */
.text : {
*(.text*)
}
.rodata : {
*(.rodata*)
}
.data : {
*(.data*)
}
.bss : {
*(COMMON)
*(.bss*)
}
}
stivale2 example
Overview
For this example, we will create these 2 files and place them in the same directory:
- kernel.c
- linker.ld
As one may notice, there is no "entry point" assembly stub, as one is not necessary with either stivale1 or 2, when using a language which can make use of a standard SysV x86 calling convention.
Furthermore, we will download the header file stivale2.h which defines structures that we will use to interact with the bootloader from here, and place it in the same directory as the other files.
Obviously, this is just a bare bones example, and one should always refer to the stivale2 specification for more details and information.
kernel.c
This is the kernel "main".
#include <stdint.h>
#include <stddef.h>
#include <stivale2.h>
// We need to tell the stivale bootloader where we want our stack to be.
// We are going to allocate our stack as an uninitialised array in .bss.
static uint8_t stack[4096];
// stivale2 uses a linked list of tags for both communicating TO the
// bootloader, or receiving info FROM it. More information about these tags
// is found in the stivale2 specification.
// stivale2 offers a runtime terminal service which can be ditched at any
// time, but it provides an easy way to print out to graphical terminal,
// especially during early boot.
static struct stivale2_header_tag_terminal terminal_hdr_tag = {
// All tags need to begin with an identifier and a pointer to the next tag.
.tag = {
// Identification constant defined in stivale2.h and the specification.
.identifier = STIVALE2_HEADER_TAG_TERMINAL_ID,
// If next is 0, it marks the end of the linked list of header tags.
.next = 0
},
// The terminal header tag possesses a flags field, leave it as 0 for now
// as it is unused.
.flags = 0
};
// We are now going to define a framebuffer header tag, which is mandatory when
// using the stivale2 terminal.
// This tag tells the bootloader that we want a graphical framebuffer instead
// of a CGA-compatible text mode. Omitting this tag will make the bootloader
// default to text mode, if available.
static struct stivale2_header_tag_framebuffer framebuffer_hdr_tag = {
// Same as above.
.tag = {
.identifier = STIVALE2_HEADER_TAG_FRAMEBUFFER_ID,
// Instead of 0, we now point to the previous header tag. The order in
// which header tags are linked does not matter.
.next = (uint64_t)&terminal_hdr_tag
},
// We set all the framebuffer specifics to 0 as we want the bootloader
// to pick the best it can.
.framebuffer_width = 0,
.framebuffer_height = 0,
.framebuffer_bpp = 0
};
// The stivale2 specification says we need to define a "header structure".
// This structure needs to reside in the .stivale2hdr ELF section in order
// for the bootloader to find it. We use this __attribute__ directive to
// tell the compiler to put the following structure in said section.
__attribute__((section(".stivale2hdr"), used))
static struct stivale2_header stivale_hdr = {
// The entry_point member is used to specify an alternative entry
// point that the bootloader should jump to instead of the executable's
// ELF entry point. We do not care about that so we leave it zeroed.
.entry_point = 0,
// Let's tell the bootloader where our stack is.
// We need to add the sizeof(stack) since in x86(_64) the stack grows
// downwards.
.stack = (uintptr_t)stack + sizeof(stack),
// No flags are currently defined as per spec and should be left to 0.
.flags = 0,
// This header structure is the root of the linked list of header tags and
// points to the first one in the linked list.
.tags = (uintptr_t)&framebuffer_hdr_tag
};
// We will now write a helper function which will allow us to scan for tags
// that we want FROM the bootloader (structure tags).
void *stivale2_get_tag(struct stivale2_struct *stivale2_struct, uint64_t id) {
struct stivale2_tag *current_tag = (void *)stivale2_struct->tags;
for (;;) {
// If the tag pointer is NULL (end of linked list), we did not find
// the tag. Return NULL to signal this.
if (current_tag == NULL) {
return NULL;
}
// Check whether the identifier matches. If it does, return a pointer
// to the matching tag.
if (current_tag->identifier == id) {
return current_tag;
}
// Get a pointer to the next tag in the linked list and repeat.
current_tag = (void *)current_tag->next;
}
}
// The following will be our kernel's entry point.
void _start(struct stivale2_struct *stivale2_struct) {
// Let's get the terminal structure tag from the bootloader.
struct stivale2_struct_tag_terminal *term_str_tag;
term_str_tag = stivale2_get_tag(stivale2_struct, STIVALE2_STRUCT_TAG_TERMINAL_ID);
// Check if the tag was actually found.
if (term_str_tag == NULL) {
// It wasn't found, just hang...
for (;;) {
asm ("hlt");
}
}
// Let's get the address of the terminal write function.
void *term_write_ptr = (void *)term_str_tag->term_write;
// Now, let's assign this pointer to a function pointer which
// matches the prototype described in the stivale2 specification for
// the stivale2_term_write function.
void (*term_write)(const char *string, size_t length) = term_write_ptr;
// We should now be able to call the above function pointer to print out
// a simple "Hello World" to screen.
term_write("Hello World", 11);
// We're done, just hang...
for (;;) {
asm ("hlt");
}
}
linker.ld
This is going to be our linker script describing where our sections will end up in memory.
/* Tell the linker that we want the symbol _start to be our entry point */
ENTRY(_start)
SECTIONS
{
/* We wanna be placed in the higher half, 2MiB above 0 in physical memory. */
/* Since we are going to use PIE, this is just the base load address, but the */
/* bootloader will be able to relocate us as it sees fit. */
. = 0xffffffff80200000;
/* We place the .stivale2hdr section containing the header in its own section, */
/* and we use the KEEP directive on it to make sure it doesn't get discarded. */
.stivale2hdr : {
KEEP(*(.stivale2hdr))
}
/* Then let's place all the other traditional executable sections afterwards. */
.text : {
*(.text*)
}
.rodata : {
*(.rodata*)
}
.data : {
*(.data*)
}
.bss : {
*(COMMON)
*(.bss*)
}
}
Building the kernel and creating an image
Makefile
In order to build our kernel, we are going to use a Makefile. This Makefile can be used for both stivale1 and stivale2 bare bones.
# This is the name that our final kernel executable will have.
# Change as needed.
KERNEL := myos.elf
# It is highly recommended to use a custom built cross toolchain to build a kernel.
# We are only using "cc" as a placeholder here. It may work by using
# the host system's toolchain, but this is not guaranteed.
CC = cc
# User controllable CFLAGS.
CFLAGS = -Wall -Wextra -O2 -pipe
# Internal link flags that should not be changed by the user.
INTERNALLDFLAGS := \
-fno-pic -fpie \
-Wl,-static,-pie,--no-dynamic-linker,-ztext \
-static-pie \
-nostdlib \
-Tlinker.ld \
-z max-page-size=0x1000
# Internal C flags that should not be changed by the user.
INTERNALCFLAGS := \
-I. \
-std=gnu11 \
-ffreestanding \
-fno-stack-protector \
-fno-pic -fpie \
-mgeneral-regs-only \
-mno-red-zone
# Use find to glob all *.c files in the directory and extract the object names.
CFILES := $(shell find ./ -type f -name '*.c')
OBJ := $(CFILES:.c=.o)
# Targets that do not actually build a file of the same name.
.PHONY: all clean
# Default target.
all: $(KERNEL)
# Link rules for the final kernel executable.
$(KERNEL): $(OBJ)
$(CC) $(INTERNALLDFLAGS) $(OBJ) -o $@
# Compilation rules for *.c files.
%.o: %.c
$(CC) $(CFLAGS) $(INTERNALCFLAGS) -c $< -o $@
# Remove object files and the final executable.
clean:
rm -rf $(KERNEL) $(OBJ)
limine.cfg
This file is parsed by Limine and it describes boot entries and other bootloader configuration variables. Further information here.
# Timeout in seconds that Limine will use before automatically booting.
TIMEOUT=5
# The entry name that will be displayed in the boot menu
:myOS
# Change the protocol line depending on the used protocol.
PROTOCOL=stivale2
# Path to the kernel to boot. boot:/// represents the partition on which limine.cfg is located.
KERNEL_PATH=boot:///myos.elf
# Remove the following line to enable kernel address layout randomisation.
KASLR=no
Compiling the kernel
We can now build our example kernel by running make. This command, if successful, should generate a file called myos.elf (or the chosen kernel name). This is our stivale(2)-compliant kernel executable.
Creating the image
We can now create either an ISO or a hard disk/USB drive image with our kernel on it. Limine can boot on both BIOS and UEFI if the image is set up to do so, which is what we are going to do.
Creating an ISO
In this example we are going to create a CD-ROM ISO capable of booting on both UEFI and legacy BIOS systems.
For this to work, we will need the xorriso utility.
These are shell commands. They can also be compiled into a script or Makefile.
# Download the latest Limine binary release.
git clone https://github.com/limine-bootloader/limine.git --branch=v2.0-branch-binary --depth=1
# Build limine-install.
make -C limine
# Create a directory which will be our ISO root.
mkdir -p iso_root
# Copy the relevant files over.
cp -v myos.elf limine.cfg limine/limine.sys \
limine/limine-cd.bin limine/limine-eltorito-efi.bin iso_root/
# Create the bootable ISO.
xorriso -as mkisofs -b limine-cd.bin \
-no-emul-boot -boot-load-size 4 -boot-info-table \
-eltorito-alt-boot -e limine-eltorito-efi.bin \
-no-emul-boot iso_root -o image.iso
Creating a hard disk/USB drive image
In this example we'll create a GPT partition table using parted, containing a single FAT partition, also known as the ESP in EFI terminology, which will store our kernel, configs, and bootloader.
This example is more involved and is made up of more steps than creating an ISO image.
These are shell commands. They can also be compiled into a script or Makefile.
# Create an empty zeroed out 64MiB image file.
dd if=/dev/zero bs=1M count=0 seek=64 of=image.hdd
# Create a GPT partition table.
parted -s image.hdd mklabel gpt
# Create an ESP partition that spans the whole disk.
parted -s image.hdd mkpart ESP fat32 2048s 100%
parted -s image.hdd set 1 esp on
# Download the latest Limine binary release.
git clone https://github.com/limine-bootloader/limine.git --branch=v2.0-branch-binary --depth=1
# Build limine-install.
make -C limine
# Install the Limine BIOS stages onto the image.
./limine/limine-install image.hdd
# Mount the loopback device.
USED_LOOPBACK=$(sudo losetup -Pf --show image.hdd)
# Format the ESP partition as FAT32.
sudo mkfs.fat -F 32 ${USED_LOOPBACK}p1
# Mount the partition itself.
mkdir -p img_mount
sudo mount ${USED_LOOPBACK}p1 img_mount
# Copy the relevant files over.
sudo mkdir -p img_mount/EFI/BOOT
sudo cp -v myos.elf limine.cfg limine/limine.sys img_mount/
sudo cp -v limine/BOOTX64.EFI img_mount/EFI/BOOT/
# Sync system cache and unmount partition and loopback device.
sync
sudo umount img_mount
sudo losetup -d ${USED_LOOPBACK}
Conclusions
If everything above has been completed successfully, you should now have a bootable ISO or hard drive/USB image containing your 64-bit higher half stivale(2) kernel and Limine to boot it. Once the kernel is successfully booted, you should see a small row of white pixels in the top left of the screen when it comes to stivale1, while for stivale2 you should be seeing "Hello World".