iscsi-target-ramdisk

Overview

This project adds a preconfigured x86_64 OpenWrt ramdisk image to your boot menu that automatically serves your Linux kernels via PXE and storage via iSCSI, allowing you to PXE boot your regular OS over the network on another computer using Dracut netroot=iscsi:.... Windows and FreeBSD iSCSI SAN boot are also supported.

For example, you can run your laptop OS on your more powerful desktop while still having access to all your laptop's files and programs.

You can also customize the OpenWrt ramdisk with any additional network configuration, files and packages you want.

Quickstart

(Using Fedora as an example):

git clone https://github.com/jwmullally/iscsi-target-ramdisk.git
cd iscsi-target-ramdisk

# Change boot_partition, boot_path and cmdline_default
gedit rootfs/etc/uci-defaults/90-pxe_menu
# Change tgt.1_1.device
gedit rootfs/etc/uci-defaults/85-tgt

sudo dependencies/fedora/build.sh
make images
sudo dependencies/fedora/install.sh
sudo ./install.sh

Usage

• Connect your two computers via Ethernet. This can be either a direct connection or through an Ethernet switch already providing DHCP.

• On the computer where this is installed (target), power on and select iSCSI Target Ramdisk from the boot menu.

• Power on the other computer (initiator), select the BIOS's built-in PXE boot.

  • Both UEFI and Legacy/CSM/BIOS boot are supported, depending on how your target OS is configured.

• The target OS should now be running on the initiator.

If you want to share network/WiFi connections or other services from the target to the initiator:

• Connect to http://192.168.200.1 (or the DHCP assigned IP) to access the OpenWrt Admin UI and configure network routing / WiFi, etc.

If the initiator computer doesn't have network ports, you can also boot via an iSCSI-to-USB bridge like this.

Requirements

• Any modern Linux OS that supports dracut iSCSI.

  • Tested on: Debian 11, Ubuntu 22.04, Fedora 36, Fedora Silverblue 36 and Arch Linux 2022.06.01.

• Also verified to work with: Windows 10, FreeBSD 13.1.

• BIOS/UEFI PXE Boot.

• Disable SecureBoot (see TODO below).

Precautions

Try this out first with the VM images in test to see how it works.

!! This solution makes slight changes to your /boot files. Before continuing, make sure you have a bootable LiveCD in case anything goes wrong.

!! Currently there is NO ENCRYPTION for the iSCSI endpoint. See TODO below. For now, only run this on a trusted network with trusted hosts. Even if your main root partition is encrypted (e.g. with LUKS), MITM attacks are still be possible on the boot files.

While remote booting, treat disconnecting the network cable like unplugging your harddrive while your computer is running. The iscsid initiator with the default settings should be able to recover from disconnects and reboots of the target (you can experiment with this in the test VMs by the disabling/enabling ethernet links). Changing the settings of the network interface carrying the iSCSI traffic can also bring the interface down, so where we can we set it to unmanaged with permanent DHCP leases.

Installation

Download the latest release, or clone this repository and checkout the latest release tag.

The only changes needed for your OS are to add Dracut iSCSI initiator support to your initramfs, and to create a boot entry for the iSCSI Target Ramdisk kernel and initramfs. The install.sh script takes care of this.

Review and adjust the configuration files in this project to match your system.

You'll mainly just want to update:

• rootfs/etc/uci-defaults/90-pxe_menu

  • boot_partition

    • Run sudo blkid and replace the UUID with the one from your partition containing /boot.

  • boot_path

    • Usually / if on its own partition, or /boot/ if its on the main root filesystem.

  • cmdline_default

    • Find your default Kernel command line from /etc/default/grub or /proc/cmdline. Note, this is ignored when it can be read fully from the /boot/loader/entries files.

• rootfs/etc/uci-defaults/85-tgt

  • Update the list of of block devices to share as iSCSI LUNs.

• rootfs/usr/lib/opkg/info/custom-password.postinst

And review:

• src/dracut.conf

• src/bootloaderspec-entry.conf

• src/grub-entry.sh

• Makefile (e.g. to adjust the package selection to support more network cards)

Then proceed with the build and installation steps below.

Debian

sudo dependencies/debian/build.sh
make images
sudo dependencies/debian/install.sh
sudo ./install.sh

NOTE: This replaces initramfs-tools with dracut (README.Debian).

Ubuntu

sudo dependencies/ubuntu/build.sh
make images
sudo dependencies/ubuntu/install.sh
sudo ./install.sh

NOTE: This replaces initramfs-tools with dracut (README.Debian).

Fedora

sudo dependencies/fedora/build.sh
make images
sudo dependencies/fedora/install.sh
sudo ./install.sh

Fedora Silverblue

dependencies/silverblue/build.sh
toolbox run --container iscsi-target-ramdisk-build make images
sudo dependencies/silverblue/install.sh
sudo ./install.sh

Arch

sudo dependencies/archlinux/build.sh
make images
sudo dependencies/archlinux/install.sh
sudo ./install.sh

NOTE: This replaces mkinitcpio with Dracut. (See here for instructions on using mkinitcpio).

After the install, you will need to move the Dracut generated initramfs into place. This can be done automatically for future kernel installs with dracut-hook.

mv /boot/initramfs-linux.img /boot/initramfs-backup-linux.img
mv /boot/initramfs-$(uname -r).img /boot/initramfs-linux.img
grub-mkconfig -o /boot/grub/grub.cfg

Installation on FreeBSD

These instructions are for UEFI BIOS, MBR is TODO.

Ensure you are using /dev/gpt/* labels in /etc/fstab, as device names will be different on different hosts. You can use gpart modify -l to set them. See here for details on how to do this.

Enable iSCSI Boot:

# pkg install net/isboot-kmod
git clone https://github.com/jnielsendotnet/isboot.git
cd isboot/src
make
make install
cat > /boot/loader.conf.d/isboot.conf <<EOF
isboot_load="YES"
EOF

Build iscsi-target-ramdisk.efi:

pkg install sysutils/debootstrap
kldload linux64 fdescfs linprocfs linsysfs tmpfs
service linux onestart
mkdir debian_build
cd debian_build
mkdir dev proc sys tmp
mount -t linprocfs none `pwd`/proc
mount -t devfs none `pwd`/dev
mount -t linsysfs none `pwd`/sys
mount -t tmpfs none `pwd`/tmp
chmod 777 tmp
chmod +t tmp
debootstrap bullseye . http://deb.debian.org/debian
chroot . /bin/bash
apt update
apt install git
cd /root
git clone https://github.com/jwmullally/iscsi-target-ramdisk.git
cd iscsi-target-ramdisk

# Set boot_partition="" to skip Linux kernel detection
vi rootfs/etc/uci-defaults/90-pxe_menu
# Change tgt.1_1.device to discover disks using GPT UUID
vi rootfs/etc/uci-defaults/85-tgt

dependencies/debian/build.sh
make images
make efi
exit

Install iscsi-target-ramdisk.efi:

mkdir /boot/efi/EFI/iscsi-target-ramdisk
cp debian_build/root/iscsi-target-ramdisk/build/images/iscsi-target-ramdisk.efi /boot/efi/EFI/iscsi-target-ramdisk
efibootmgr -a -c -l /boot/efi/EFI/iscsi-target-ramdisk/iscsi-target-ramdisk.efi -L iscsi-target-ramdisk

This will add the iscsi-target-ramdisk to your UEFI boot entries as the default. You should later change the boot order back to your main OS.

Disable DHCP for iSCSI Boot network interface:

install -m 0755 debian_build/root/iscsi-target-ramdisk/src/isbootifname /usr/local/etc/rc.d/isbootifname
sysrc ifconfig_bootnet0="NOAUTO"

Now your system should be ready for remote boot.

Installation on Windows 10

Windows is a little more complicated and requires these steps to be done to prepare it for remote iSCSI boot:

• Building iscsi-target-ramdisk.efi
• Adding a boot entry for the EFI image
• Configure the iSCSI Target
• Install the initiator NIC driver and devices
• Disable Pagefile

Because the IP of the iSCSI target is fixed in iscsicpl, you should run iscsi-target-ramdisk in static IP mode (i.e. 192.168.200.1) or with a static DHCP lease from your local router.

Building iscsi-target-ramdisk.efi

If you have a Linux installation and use GRUB, you should instead follow the Linux installation instructions elsewhere in this guide.

Under Windows, you can build the project in the Windows Subsystem for Linux, and copy the iscsi-target-ramdisk.efi image to the EFI system partition and add a boot entry.

Install the default WSL Ubuntu installation:

wsl --install

Download, configure and build iscsi-target-ramdisk.efi:

git clone https://github.com/jwmullally/iscsi-target-ramdisk.git
cd iscsi-target-ramdisk

# Set boot_partition="" to skip Linux kernel detection
nano rootfs/etc/uci-defaults/90-pxe_menu
# Change tgt.1_1.device
nano rootfs/etc/uci-defaults/85-tgt

sudo dependencies/ubuntu/build.sh
# Workaround spaces in WSL PATH: https://openwrt.org/docs/guide-developer/toolchain/wsl
PATH=/usr/sbin:/usr/bin:/sbin:/bin make images
make efi

WIN_USER="$(cmd.exe /c 'echo %USERNAME%' | tr -d '[:space:]')"
cp build/images/iscsi-target-ramdisk.efi "/mnt/c/Users/$WIN_USER/Downloads/"

Alternatively, you can build the ISO and flash to a USB key and boot from that:

sudo dependencies/ubuntu/build-iso.sh
make iso

WIN_USER="$(cmd.exe /c 'echo %USERNAME%' | tr -d '[:space:]')"
cp build/images/iscsi-target-ramdisk.iso "/mnt/c/Users/$WIN_USER/Downloads/"

Adding a boot entry for the EFI image

Copy the image to the EFI boot partition:

mountvol S: /s
mkdir S:\EFI\iscsi-target-ramdisk
copy %USERPROFILE%\Downloads\iscsi-target-ramdisk.efi S:\EFI\iscsi-target-ramdisk\iscsi-target-ramdisk.efi
mountvol S: /d

Add a new entry to the system EFI Boot Manager (Reference).

bcdedit /copy "{bootmgr}" /d "iscsi-target-ramdisk"
# The entry was successfully copied to {34e8383c-73a7-11e9-9cb0-94de8078a7b5}.

mountvol S: /s
bcdedit /set "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" device partition=S:
mountvol S: /d
bcdedit /set "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" path \EFI\iscsi-target-ramdisk\iscsi-target-ramdisk.efi

bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" locale
bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" inherit
bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" default
bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" resumeobject
bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" displayorder
bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" toolsdisplayorder
bcdedit /deletevalue "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}" timeout

(The above is roughly equivalent to the following under Linux: efibootmgr --create --disk /dev/sda --part 2 --label 'iscsi-target-ramdisk' --loader '\EFI\iscsi-target-ramdisk\iscsi-target-ramdisk.efi')

Reboot the computer and enter the system Boot menu, and you should see the new iscsi-target-ramdisk entry.

Some older systems will not show these extra EFI boot options. If this is the case for you, you should remove this bcdedit entry and install an EFI boot loader like rEFInd or GRUB.

To remove the boot entry, do the following:

bcdedit /delete "{34e8383d-73a7-11e9-9cb0-94de8078a7b5}"

To remove the EFI image, do:

mountvol S: /s
del S:\EFI\iscsi-target-ramdisk\iscsi-target-ramdisk.efi
rmdir S:\EFI\iscsi-target-ramdisk
mountvol S: /d

Configure the iSCSI Target

To prepare an existing regular Windows 10 installation for iSCSI boot, a persistent iSCSI connection to the target must be configured. When Windows is booted normally from local storage, this iSCSI connection will stay disconnected and unused.

• Temporarily boot the image/ISO/USB on another system connected directly via Ethernet.

  • (Or alternatively, boot the ISO on the same system as a Hyper-V VM):

New-VMSwitch -Name InternalSwitch -SwitchType Internal
New-VM -Name iscsi-target-ramdisk-iso -MemoryStartupBytes 2GB -BootDevice CD -NewVHDPath "C:\Users\Public\Documents\Hyper-V\Virtual Hard Disks\iscsi-target-ramdisk.vhdx" -NewVHDSizeBytes 1GB -Generation 1 -Switch InternalSwitch
Set-VMDvdDrive -VMName iscsi-target-ramdisk-iso -Path "$env:USERPROFILE\Downloads\iscsi-target-ramdisk.iso"

Start-VM -Name iscsi-target-ramdisk-iso
Restart-NetAdapter -Name "vEthernet (InternalSwitch)"
# Interface should receive a 192.168.200.0/24 IP address
Get-NetIPAddress | Where-Object {$_.InterfaceAlias -eq "vEthernet (InternalSwitch)"}

• It should start using the default static IP settings 192.168.200.1.

• This will be used to help verify all the iSCSI settings are correct in the iSCSI Initiator Windows UI app.

• Ensure the iSCSI Target Service is running with /etc/init.d/tgt show.

• Visit http://192.168.200.1:81/cgi-bin/get-menu-ipxe to enable iSCSI access through the firewall for your IP, or disable the iSCSI firewall rule with uci set firewall.block_iscsi.enabled=0; /etc/init.d/firewall restart.

• Open the iSCSI Initiator app (iscsicpl.exe) and connect to the target, using the initiator settings from /etc/config/tgt:

  • Warning: This will overwrite your existing Windows iSCSI Initiator Name and Reverse CHAP Secret. To preserve them, set them first in /etc/config/tgt to match your existing Windows settings.

  • Initiator name: target -> <allow_name>.

  • iSCSI Initiator Mutual CHAP Secret: user_out -> <password>.

  • Quick Connect Target: 192.168.200.1.

  • Connect -> Advanced Options -> Enable CHAP log on: Name, Target secret = "user_in" -> <user>, <password>.

  • Perform mutual authentication: Enabled.

  • Add this connection to the list of Favorite Targets: Enabled.

• Or alternatively, copy + paste the commands from http://192.168.200.1:81/cgi-bin/iscsi-windows.ps1 into an Administrator Powershell), or use the following example:

Start-Service -Name MSiSCSI
Set-Service -Name MSiSCSI -StartupType Automatic

Set-InitiatorPort `
    -NodeAddress (Get-InitiatorPort).NodeAddress `
    -NewNodeAddress "iqn.2018-04.org.example:initiator-host"

Set-IscsiChapSecret `
    -ChapSecret "password5678"

New-IscsiTargetPortal `
    -TargetPortalAddress "192.168.200.1"

Connect-IscsiTarget `
    -TargetPortalAddress "192.168.200.1" `
    -NodeAddress "iqn.2018-04.org.example:target-host" `
    -AuthenticationType MUTUALCHAP `
    -ChapUsername "iscsiuser_in" `
    -ChapSecret "password1234" `
    -IsPersistent $true

Update-HostStorageCache
Get-Disk

Install the initiator NIC driver and devices

• Install the driver for the network card used by the initiator. This is complicated as it requires some other way of booting the target OS on the initiator, but only needs to be done once. This can be done in one of the following ways:

  • Attach the target storage directly to the initiator.

  • Boot the iSCSI target as a USB Mass Storage device.

  • Boot iscsi-target-ramdisk on the target and a Linux Live CD on the initiator (e.g. Fedora or Ubuntu). Install libvirtd and virt-manager, attach the iSCSI drive using the script from http://192.168.200.1:81/cgi-bin/iscsistart.sh, create a Win10 VM with the existing /dev/sd* iSCSI drive, add the network card as a "PCI Host Device", then boot Windows in the VM once to install the network driver. This might not work if the hardware is too dissimilar and other system drivers are needed.

• Set your network card driver to start during early boot. (Seems not always necessary, first try PXE booting without this).

  • E.g. Device Manager -> Intel 82574L -> Driver Details: e1i65x64.sys:

  • Regedit: [HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\e1i65x64]: Start = 0

  • Repeat for each ControlSet001, ControlSet002 etc.

Disable the pagefile

Disable the pagefile. Without doing this, you may encounter IRQL_NOT_LESS_OR_EQUAL STOP codes on network boot.

• System -> Advanced System Properties -> Performance -> Pagefile: Disabled.

  • (Or alternatively in PowerShell):

$cs = gwmi Win32_ComputerSystem
if ($cs.AutomaticManagedPagefile) {
    $cs.AutomaticManagedPagefile = $False
    $cs.Put()
}
$pg = gwmi win32_pagefilesetting
if ($pg) {
    $pg.Delete()
}

PXE booting Windows

After preparing the system with the instructions above, you can now use PXE to iSCSI boot the Windows system.

• On the same Windows system, boot the iSCSI Target Ramdisk image from UEFI, GRUB or USB.

• On the initiator host, boot using PXE (BIOS/CSM/Legacy or UEFI depending on your Windows installation) and choose the iBFT SAN boot option.

• Windows should now start running.

Post Installation

Updating

After the initial install, this solution should work indefinitely even as you upgrade your kernels.

If you want to make changes to the OpenWrt configuration, you will only need to update /boot/iscsi-target-ramdisk-kernel.bin and /boot/iscsi-target-ramdisk-initrd.img by doing the following:

make images
sudo ./update.sh

Uninstalling

sudo /usr/local/sbin/uninstall-iscsi-target-ramdisk.sh

Disabling NetworkManager management of BOOTIF interface

It's important that the BOOTIF interface stay running as the root filesystem can't perform read or writes when its down. This can cause problems when NetworkManager takes over the interface, and various events can cause it to be restarted (e.g. even the user refreshing the interface in the UI), which can lock the system. To prevent NetworkManager from managing the interface, this service is installed automatically.

Troubleshooting

On the OpenWrt target:

• Check /srv/pxe/pxe_menu/menu.ipxe and /srv/pxe/pxe_menu contain the expected kernels.

  • Update /etc/config/pxe_menu and rerun /etc/init.d/pxe_menu restart to try find them again.

• Use logread -f to keep an eye on the DHCP and TFTP requests.

• Use tcpdump port 81 to check for incoming PXE related HTTP requests (unfortunately OpenWrt uHTTPd does not support request logging).

• Check the console or dmesg output for Ethernet interface corruption warnings (e.g. some e1000e models have flaky offloading that needs to be disabled with ethtool.)

• To set breakpoints during initrd, change /etc/config/pxe_menu cmdline_iscsi to include rd.shell or rd.break=.... See dracut.cmdline(7) for the list of break points.

On the initiator PXE boot menu:

• Remove quiet from the kernel cmdline to see more debug output during boot.

On the initiator:

• Use iscsiadm -m session -P 1 --print=3 to view the iSCSI connection status and parameters

  • For live updates during testing: watch -n 0.2 iscsiadm -m session -P 1 --print=3

How it works

On the target host (containing the OS to remote boot):

• iSCSI Target Ramdisk boots with its own kernel and stateless initramfs.
• /etc/init.d/pxe_menu is run which discovers the OS kernel images from the /boot partition, copies them to /srv/pxe/pxe_menu and creates entries in /srv/pxe/pxe_menu/menu.ipxe.
  • Configuration: /etc/uci-defaults/90-pxe_menu.
  • If /boot/loader/entries is found, all BootLoaderSpec files are parsed to identify kernel images and cmdline arguments. If not found, entries are created for all kernels matching /boot/vmlinuz-* along with their matching initramfs file and the cmdline_default arguments.
  • The contents of cmdline_iscsi are appended to the cmdline, which include the netroot:iscsi:... paramaters.
• /etc/init.d/tgt starts which exports the disk block devices as iSCSI LUN targets.
  • Configuration: /etc/uci-defaults/85-tgt.
• /etc/init.d/network starts, which sets the first LAN interface to DHCP by default.
  • Configuration: /etc/uci-defaults/80-network
• /etc/init.d/dnsmasq starts which provides PXE DHCP Proxy boot (to work alongside existing DHCP servers) and serves /srv/pxe via TFTP. Regular DHCP allocations are disabled by default.
  • Configuration: /etc/uci-defaults/90-dhcp.
• /etc/init.d/uhttpd starts and serves /srv/pxe via HTTP access.
  • Configuration: /etc/uci-defaults/95-uhttpd.
  • HTTP BASIC authentication is used to protect /srv/pxe/pxe_menu and /srv/pxe/cgi-bin containing the boot images and configuration.
  • The /etc/init.d/pxe_access service can be used to enable/disable access to these files.
• /etc/init.d/dhcpfallback starts, which sets LAN to a static IP if no existing DHCP servers were found during the specified time frame.
  • If activated, /etc/config/dhcp is also changed from PXE DHCP Proxy mode back to regular DHCP server mode.

On the initiator host (the one to run the OS on):

• The BIOS starts PXE boot.
• The PXE ROM requests and receives a DHCP boot response, pointing to the iPXE binary on TFTP.
  • If an existing DHCP server is present, that server will send a regular DHCP response, and the target host will send a separate PXE DHCP Proxy boot response containing just the boot information.
• iPXE is downloaded and executed, which issues another DHCP request and fetches /srv/pxe/ipxe/entry.ipxe over TFTP.
• The user enters a username / password which are used as the authorization for the PXE HTTP requests.
  • Caution: Boot files and iSCSI credentials can still be sniffed as they are transferred over the network. Only use on a physically secure network or direct connection.
• iPXE chainloads /srv/pxe/cgi-bin/get-menu-ipxe over HTTP.
  • iSCSI access for the requesting initiator host's IP address is allowed through the firewall.
  • /srv/pxe/pxe_menu/menu.ipxe is returned and executed by iPXE.
• (Optional, default) The iSCSI target connection details are stored in the iBFT ACPI table.
• The user selects a kernel to boot.
• iPXE fetches the kernel and associated initramfs over HTTP.
• iPXE launches the kernel using the included cmdline arguments, which contain the extra netroot:iscsi:... parameters.
• The kernel starts, unpacks and launches the init process in the initramfs.
• The Dracut modules are executed.
• The dracut-network iSCSI module sees the netroot:iscsi:... arguments and uses them to start an Open iSCSI initiator connection to the iSCSI Target Ramdisk host. If successful, the iSCSI target LUN devices now appear as local block devices.
• Booting continues as normal, mounting the root filesystem using the UUID and other regularly supplied cmdline arguments.
• (NetworkManager) The BOOTIF interface is set to unmanaged with this service to prevent automatic reconfiguration
• The target OS is now fully loaded on the initiator host.

Background

Typical Linux distributions use a simple boot loader (e.g. GRUB) to load the Linux Kernel and an Initial ramdisk mini root file system. The purpose of this root filesystem is to do everything necessary to prepare the storage block devices and mount the real root filesystem. This provides the OS with great flexibility about how the root filesystem is stored, for example on different types of network storage, logical volumes, RAID arrays, encrypted filesystems etc. All the configuration and complexity is handled by software in the Initial Ramdisk; all the kernel needs is a final mounted directory it can chroot, and continue the init boot sequence.

Here we add the existing Dracut iSCSI initiator module to the OS's initramfs, which is designed for booting systems installed to remote iSCSI block devices. It stays deactivated and out of the way unless the netroot:iscsi:... kernel cmdline arguments are supplied. When activated, the disk's block devices show up on the system just as if they were locally attached. Modern Linux distributions use UUID-based partition identification in /etc/fstab, which makes mounting work deterministically regardless of the names of the underlying block devices (e.g. /dev/sda, /dev/sdb ordering can change based on the order drives or USB keys are inserted). In practice, this means you can do upgrades, kernel updates, bootloader changes, etc. as if you were doing them on the original computer. On modern systems with a 1GbE connection, you should get full 1Gb/s transfer speed and relatively low IOP latency. As modern Linux distributions are mostly plug and play, there should be little issue with your OS seeing a completely different set of hardware.

To share the OS drives, we can't use the original OS itself, as only one system can be reading/writing the devices at a time (otherwise disk corruption would result), so instead we use a stateless ramdisk image built using the OpenWrt ImageBuilder. OpenWrt is a very flexible embedded system platform, specializing in network routing with a large number of packages available. This image has been configured to automatically share the drives with iSCSI and the kernel+initramfs files with PXE after booting. The separate OpenWrt system also means you don't have to reconfigure your OS to do all this sharing, and you can easily customize it further by adding more files and packages.

FAQ

• Q: What about support for Fibre Channel over Ethernet (FCoE)?
  • A: iPXE and Dracut support FCoE, but tgtd does not, so LIO would need to be included in OpenWrt to support this. However the benefits (if any) would be marginal and iSCSI is much easier to work with, e.g. "FCoE vs iSCSI?".

Developing

Patches are welcome.

Test with the sample VMs in test before opening a pull request.

Match OpenWrt structure and conventions as much as possible.

TODO

• Debian: Tips for disabling default open-iscsi service by default during normal use to prevent error.

• MACSEC L2 encryption or iSCSI + TLS.

• SecureBoot. (Unlikely?). Provide instructions for self-signed images with mokutil?

• Sort "iSCSI Target Ramdisk" entry under OS entries in bootloader menu.

• OpenWrt tgtd: Support CRC32 Header + Data digests.

• OpenWrt kernel: Add LIO iSCSI target support, use instead of TGT (rough comparison).

• Set custom DHCP user class to always fetch newest iPXE.

• dracut iscsi: Re-add support for rd.iscsi.param when rd.iscsi.firmware is used (for setting node.session.timeo.replacement_timeout=0 for iBFT).

• Make procd scripts non-blocking.

Reference

iPXE - Documentation.

bootup - System bootup process.

dracut.cmdline(7).

dracut.conf(5).

BootLoaderSpec.

systemd kernel-command-line(7).

OpenWrt.

OpenWrt - UCI configuration.

OpenWrt - DHCP.

OpenWrt - iSCSI.

OpenWrt - Image Builder.

iSCSI Security.

Author

Copyright (C) 2022 Joseph Mullally

License: GPLv2

Project: https://github.com/jwmullally/iscsi-target-ramdisk