PREV_LEC

NEXT_LEC

📷 Suggested Image

A terminal screenshot showing the output of ls -lh /boot on a real Linux 6.x system — highlighting vmlinuz, initramfs, System.map and config files. Alt text: Linux /boot directory contents after kernel installation showing vmlinuz initramfs System.map files

💡 How to Check Your Kernel’s Compression Type

# Look in your .config for the compression setting:
grep "CONFIG_KERNEL_" /boot/config-$(uname -r) | grep "=y"

# You will see one of:
# CONFIG_KERNEL_GZIP=y    (classic, widest support)
# CONFIG_KERNEL_LZ4=y     (fast decompression — common on Ubuntu 22+)
# CONFIG_KERNEL_ZSTD=y    (best ratio + speed — Fedora 37+, Debian 12+)

📷 Suggested Image

A diagram or annotated screenshot showing vmlinux in the kernel source tree root vs vmlinuz in /boot — with file sizes visible. Alt text: vmlinux vs vmlinuz comparison showing location size and purpose in Linux kernel development

📷 Suggested Image

A photo of a terminal running systemd-analyze showing the boot timeline, with the initramfs phase highlighted. Alt text: Linux kernel boot timeline showing initramfs phase duration using systemd-analyze

Q: Can I boot Linux without initramfs?

Yes — if all your storage drivers and your root filesystem’s driver are compiled directly into the kernel (=y in .config, not as loadable modules), initramfs is not needed. This is common on minimal embedded systems where the hardware is fixed and known. On general-purpose distros it is always used because they ship a single kernel for thousands of hardware combinations.

Q: Why is the file named initrd.img on Ubuntu but initramfs on Fedora?

This is purely a naming convention difference between distributions. The file format inside is identical cpio-based initramfs in both cases. Ubuntu/Debian inherited the initrd name from history; Fedora/RHEL chose the more accurate initramfs name. The kernel handles both correctly — it detects the format by inspecting the file header, not the filename.

Q: What happens if the initramfs image is missing or corrupt?

The system will fail to boot and drop into a kernel panic with an error like “VFS: Unable to mount root fs”. To fix this, boot from a live USB, mount your root partition, chroot into it, and regenerate the initramfs with update-initramfs -u or dracut --force. This is a common recovery task for Linux admins.

Q: How do I reduce initramfs boot time on an embedded target?

Three practical approaches: (1) Switch compression from gzip to LZ4 (CONFIG_INITRAMFS_COMPRESSION_LZ4=y) — decompression is 3–5x faster. (2) Use dracut --hostonly or prune unnecessary modules from the initramfs to reduce its size. (3) For targets with fixed hardware, skip initramfs entirely and compile all drivers in-kernel. Buildroot and Yocto both have options to generate minimal, stripped-down initramfs images optimized for fast boot.

Q: What is the difference between mkinitramfs and dracut?

mkinitramfs is a shell-script based tool from the Debian ecosystem — simple and reliable but less configurable. dracut is a more modern, modular, distro-agnostic framework used by Fedora, RHEL, Arch, and increasingly Ubuntu. Both produce a functionally equivalent cpio initramfs image. dracut is better for complex setups like LUKS-on-LVM, iSCSI root, multipath, or when you need fine-grained control over what goes into the image.

Q: How is vmlinuz different from bzImage?

bzImage is the name of the compressed kernel image as produced in the kernel build system — specifically at arch/x86/boot/bzImage. When make install runs, that file is copied to /boot/ and renamed to vmlinuz-<version>. So vmlinuz and bzImage are the same binary — just different names at different stages of the workflow. The “bz” in bzImage stands for “big zImage” — it is not related to bzip2 compression.

Q1. What is the difference between vmlinux and vmlinuz?

vmlinux is the uncompressed ELF kernel binary sitting in the root of the kernel source tree after compilation. It is large (often over 100 MB), retains full debug symbols, and is used by tools like gdb, perf, and crash for post-mortem debugging. It is not directly bootable.

vmlinuz is the compressed, self-extracting kernel image placed in /boot. GRUB loads this into RAM at boot time. The “z” hints at compression — historically gzip, though modern kernel 6.x builds commonly use LZ4 or Zstd. On x86 this file is architecturally a bzImage.

Q2. Why does Linux need initramfs? What problem does it solve?

The fundamental boot paradox: the kernel needs to mount the root filesystem to load its files — but mounting the root filesystem may require a kernel module (e.g. the NVMe driver, RAID driver, or filesystem driver) that is stored on the root filesystem. A circular dependency.

initramfs solves this by giving the kernel a tiny, self-contained filesystem in RAM before any disk driver is loaded. Its /init script loads the necessary kernel modules, handles disk decryption if needed, and mounts the real root filesystem. Once that is done, initramfs is discarded and the real system takes over.

Q3. What is System.map and when do you use it as a kernel developer?

System.map is a plain text file generated at compile time that maps every kernel symbol name to its virtual memory address. As a kernel developer you use it primarily when investigating kernel Oops or panic messages — the crash log contains hex addresses, and System.map lets you translate those addresses into function names to identify exactly where in the kernel the crash occurred.

Q4. What is the difference between initrd and initramfs?

initrd (initial RAM disk) was the older approach — it emulated a block device in RAM and required the kernel to have a block device driver and a filesystem driver just to access it. initramfs replaced this in kernel 2.6. It is a cpio archive that the kernel unpacks directly into a tmpfs — no filesystem driver needed to access itself. initramfs is simpler, faster, uses less memory, and is the standard on all modern Linux systems.

Q5. How is initramfs used differently in embedded Linux vs desktop Linux?

On a desktop, initramfs is a temporary bridge — it loads drivers, mounts the real root disk partition, then gets discarded. On embedded targets, initramfs often becomes the permanent root filesystem. The entire userspace (BusyBox, daemons, config files) is packed into the initramfs and embedded directly inside the kernel binary using CONFIG_INITRAMFS_SOURCE. This produces a single deployable file, fast boot, and a clean read-only RAM filesystem — ideal for embedded products.

Q6. Which kernel config enables initramfs support and what is the default?

CONFIG_BLK_DEV_INITRD=y. It is enabled by default in all mainstream kernel configurations. Disabling it means the kernel cannot use an initramfs at all. This is only done on minimal embedded targets where the hardware is fixed and all required drivers can be compiled directly into the kernel image.

Q7. Walk me through what happens from GRUB to the login prompt on a modern Linux 6.x system.

GRUB reads its config, loads vmlinuz and the initramfs image into RAM. The self-extracting vmlinuz decompresses into memory and begins running. The kernel initialises memory, the scheduler, and basic hardware. It unpacks the initramfs into tmpfs and runs /init inside it — loading storage drivers, handling encryption, mounting the real root filesystem. initramfs is then freed, and control is handed to systemd on the real root. systemd starts all services in parallel and eventually the login prompt or desktop appears.

PREV_LEC

NEXT_LEC