AKA Frequently Questioned Answers
An article was written for Libreboot, to be maintained over time, that accurately describes the current status Libreboot in terms of software freedom, describing any caveats that exist for specific hardware platforms.
Please read the article, thus:
Software and hardware freedom status for each mainboard supported by Libreboot
You may also find this other section of the FAQ useful: What level of software freedom does Libreboot give me?
Refer to the lbmk build instructions.
Refer to the lbmk maintenance manual.
This SPI flasher will damage your chip, and the mainboard that it is connected to.
Read the notes about CH341A on docs/install/spi.html to learn more.
You do not need to be a skilled developer in order to help the project substantially. If you have a board supported by Coreboot, consider porting it to Libreboot.
If you have a board supported in Libreboot then please consider becoming a tester. Testing involves minimal effort and really helps out the project. See the board maintainers documentation if you are interested in testing roms before they are released.
flashrom -p internal for software based flashing, and you get an error related to /dev/mem access, you should reboot with
iomem=relaxed kernel parameter before running flashrom, or use a kernel that has
CONFIG_IO_STRICT_DEVMEM not enabled.
Example flashrom output with both
flashrom v0.9.9-r1955 on Linux 4.11.9-1-ARCH (x86_64) flashrom is free software, get the source code at https://flashrom.org Calibrating delay loop... OK. Error accessing high tables, 0x100000 bytes at 0x000000007fb5d000 /dev/mem mmap failed: Operation not permitted Failed getting access to coreboot high tables. Error accessing DMI Table, 0x1000 bytes at 0x000000007fb27000 /dev/mem mmap failed: Operation not permitted
We don’t know how to detect the correct PWM value to use in coreboot, so we just use the default one in coreboot which has this issue on some CCFL panels, but not LED panels.
You can work around this in your distribution, by following the notes at docs: backlight control.
This was observed on some systems using network-manager. This happens both on the original BIOS and in libreboot. It’s a quirk in the hardware. On debian systems, a workaround is to restart the networking service when you connect the ethernet cable:
sudo service network-manager restart
On systemd-based distros, you might try:
sudo systemctl restart network-manager
(the service name might be different for you, depending on your configuration)
KGPE-D16, KCMA-D8 and KFSN4-DRE ASUS mainboards were removed on 19 November 2022. You can still use older revisions of Libreboot, and older release versions.
Loading the option ROM from the PIKE2008 module on either ASUS KCMA-D8 or KGPE-D16 causes the system to hang at boot. It’s possible to use this in the payload (if you use a linux kernel payload, like linuxboot), or to boot (with SeaGRUB and/or SeaBIOS) from regular SATA and then use it in Linux. The Linux kernel is capable of using the PIKE2008 module without loading the option ROM.
The easiest method of doing so is by using the kernel’s netconsole and reproducing the panic. Netconsole requires two machines, the one that is panicky (source) and the one that will receive crash logs (target). The source has to be connected with an ethernet cable and the target has to be reachable at the time of the panic. To set this system up, execute the following commands as root on the source (
source#) and normal user on the target (
Start a listener server on the target machine (netcat works well):
target$ nc -u -l -p 6666
Mount configfs (only once per boot, you can check if it is already mounted with
mount | grep /sys/kernel/config. This will return no output if it is not).
source# modprobe configfs
source# mkdir -p /sys/kernel/config
source# mount none -t configfs /sys/kernel/config
find source’s ethernet interface name, it should be of the form
ip address or
source# iface="enp0s29f8u1" change this
Fill the target machine’s IPv4 address here:
source# tgtip="192.168.1.2" change this
Create netconsole logging target on the source machine:
source# modprobe netconsole
source# cd /sys/kernel/config/netconsole
source# mkdir target1; cd target1
source# srcip=$(ip -4 addr show dev "$iface" | grep -Eo '[0-9]+\.[0-9]+\.[0-9]+\.[0-9]+')
source# echo "$srcip" > local_ip
source# echo "$tgtip" > remote_ip
source# echo "$iface" > dev_name
source# arping -I "$iface" "$tgtip" -f | grep -o '..:..:..:..:..:..' > remote_mac
source# echo 1 > enabled
Change console loglevel to debugging:
source# dmesg -n debug
Test if the logging works by e.g. inserting or removing an USB device on the source. There should be a few lines appearing in the terminal, in which you started netcat (nc), on the target host.
Try to reproduce the kernel panic.
Any system can easily be added, so compatibility merely refers to whatever boards are integrated in the
lbmk build system, which libreboot uses.
Please read the hardware compatibility list.
Coreboot is nominally Free Software, but requires binary blobs on most x86 targets that it supports, on both Intel and AMD.
NOTE: The information below is slightly out of date. Nowadays, Intel ME does not run on an ARC coprocessor, but instead runs on a modified Intel 486 based architecture, with the ME firmware written for x86 based on the Minix operating system. However, the overall design philosophy and operation is mostly the same.
On most current Intel platforms that have Intel ME, it is now possible to disable Intel ME after BringUp. See:
On all GM45+ICH9M laptops that have an Intel ME in it (additionally, this means X4X+ICH10 desktops), the ME firmware is not needed in the boot flash. Either a modified descriptor is used, which disables the ME and removes the region for it in the boot flash, or a descriptorless setup is used. However, all modern Intel platforms otherwise require an Intel ME image to be present in the main boot flash.
Now onto the main topic:
Introduced in June 2006 in Intel’s 965 Express Chipset Family of (Graphics and) Memory Controller Hubs, or (G)MCHs, and the ICH8 I/O Controller Family, the Intel Management Engine (ME) is a separate computing environment physically located in the (G)MCH chip. In Q3 2009, the first generation of Intel Core i3/i5/i7 (Nehalem) CPUs and the 5 Series Chipset family of Platform Controller Hubs, or PCHs, brought a more tightly integrated ME (now at version 6.0) inside the PCH chip, which itself replaced the ICH. Thus, the ME is present on all Intel desktop, mobile (laptop), and server systems since mid 2006.
The ME consists of an ARC processor core (replaced with other processor cores in later generations of the ME), code and data caches, a timer, and a secure internal bus to which additional devices are connected, including a cryptography engine, internal ROM and RAM, memory controllers, and a direct memory access (DMA) engine to access the host operating system’s memory as well as to reserve a region of protected external memory to supplement the ME’s limited internal RAM. The ME also has network access with its own MAC address through an Intel Gigabit Ethernet Controller. Its boot program, stored on the internal ROM, loads a firmware “manifest” from the PC’s SPI flash chip. This manifest is signed with a strong cryptographic key, which differs between versions of the ME firmware. If the manifest isn’t signed by a specific Intel key, the boot ROM won’t load and execute the firmware and the ME processor core will be halted.
The ME firmware is compressed and consists of modules that are listed in the manifest along with secure cryptographic hashes of their contents. One module is the operating system kernel, which is based on a proprietary real-time operating system (RTOS) kernel called “ThreadX”. The developer, Express Logic, sells licenses and source code for ThreadX. Customers such as Intel are forbidden from disclosing or sublicensing the ThreadX source code. Another module is the Dynamic Application Loader (DAL), which consists of a Java virtual machine and set of preinstalled Java classes for cryptography, secure storage, etc. The DAL module can load and execute additional ME modules from the PC’s HDD or SSD. The ME firmware also includes a number of native application modules within its flash memory space, including Intel Active Management Technology (AMT), an implementation of a Trusted Platform Module (TPM), Intel Boot Guard, and audio and video DRM systems.
The Active Management Technology (AMT) application, part of the Intel “vPro” brand, is a Web server and application code that enables remote users to power on, power off, view information about, and otherwise manage the PC. It can be used remotely even while the PC is powered off (via Wake-on-Lan). Traffic is encrypted using SSL/TLS libraries, but recall that all of the major SSL/TLS implementations have had highly publicized vulnerabilities. The AMT application itself has known vulnerabilities, which have been exploited to develop rootkits and keyloggers and covertly gain encrypted access to the management features of a PC. Remember that the ME has full access to the PC’s RAM. This means that an attacker exploiting any of these vulnerabilities may gain access to everything on the PC as it runs: all open files, all running applications, all keys pressed, and more.
Intel Boot Guard is an ME application introduced in Q2 2013 with ME firmware version 9.0 on 4th Generation Intel Core i3/i5/i7 (Haswell) CPUs. It allows a PC OEM to generate an asymmetric cryptographic keypair, install the public key in the CPU, and prevent the CPU from executing boot firmware that isn’t signed with their private key. This means that coreboot and libreboot are impossible to port to such PCs, without the OEM’s private signing key. Note that systems assembled from separately purchased mainboard and CPU parts are unaffected, since the vendor of the mainboard (on which the boot firmware is stored) can’t possibly affect the public key stored on the CPU.
ME firmware versions 4.0 and later (Intel 4 Series and later chipsets) include an ME application for audio and video DRM called “Protected Audio Video Path” (PAVP). The ME receives from the host operating system an encrypted media stream and encrypted key, decrypts the key, and sends the encrypted media decrypted key to the GPU, which then decrypts the media. PAVP is also used by another ME application to draw an authentication PIN pad directly onto the screen. In this usage, the PAVP application directly controls the graphics that appear on the PC’s screen in a way that the host OS cannot detect. ME firmware version 7.0 on PCHs with 2nd Generation Intel Core i3/i5/i7 (Sandy Bridge) CPUs replaces PAVP with a similar DRM application called “Intel Insider”. Like the AMT application, these DRM applications, which in themselves are defective by design, demonstrate the omnipotent capabilities of the ME: this hardware and its proprietary firmware can access and control everything that is in RAM and even everything that is shown on the screen.
The Intel Management Engine with its proprietary firmware has complete access to and control over the PC: it can power on or shut down the PC, read all open files, examine all running applications, track all keys pressed and mouse movements, and even capture or display images on the screen. And it has a network interface that is demonstrably insecure, which can allow an attacker on the network to inject rootkits that completely compromise the PC and can report to the attacker all activities performed on the PC. It is a threat to freedom, security, and privacy that can’t be ignored.
Before version 6.0 (that is, on systems from 2008/2009 and earlier), the ME can be disabled by setting a couple of values in the SPI flash memory. The ME firmware can then be removed entirely from the flash memory space. The libreboot project does this on the Intel 4 Series systems that it supports, such as the ThinkPad X200 and ThinkPad T400. ME firmware versions 6.0 and later, which are found on all systems with an Intel Core i3/i5/i7 CPU and a PCH, include “ME Ignition” firmware that performs some hardware initialization and power management. If the ME’s boot ROM does not find in the SPI flash memory an ME firmware manifest with a valid Intel signature, the whole PC will shut down after 30 minutes.
Due to the signature verification, developing free replacement firmware for the ME is basically impossible. The only entity capable of replacing the ME firmware is Intel. As previously stated, the ME firmware includes proprietary code licensed from third parties, so Intel couldn’t release the source code even if they wanted to. And even if they developed completely new ME firmware without third-party proprietary code and released its source code, the ME’s boot ROM would reject any modified firmware that isn’t signed by Intel. Thus, the ME firmware is both hopelessly proprietary and “tivoized”.
In summary, the Intel Management Engine and its applications are a backdoor with total access to and control over the rest of the PC. The ME is a threat to freedom, security, and privacy, and the libreboot project strongly recommends avoiding it entirely. Since recent versions of it can’t be removed, this means avoiding all recent generations of Intel hardware.
The above paragraph is only talking about setups where the full Intel ME firmware is used, containing networking code and especially Active Management Technology (AMT).
Use of the
me_cleaner utility is believed to minimize any security risk when using these Intel platforms, and coreboot does contain fully free code for sandybridge/ivybridge platforms. Freedom-wise, these are similar to libreboot compatible ThinkPads, and they are quite nice machines.
More information about the Management Engine can be found on various Web sites, including me.bios.io, unhuffme, coreboot wiki, and Wikipedia. The book Platform Embedded Security Technology Revealed describes in great detail the ME’s hardware architecture and firmware application modules.
If you’re stuck with the ME (non-libreboot system), you might find this interesting: http://hardenedlinux.org/firmware/2016/11/17/neutralize_ME_firmware_on_sandybridge_and_ivybridge.html
On all recent Intel systems, coreboot support has revolved around integrating a blob (for each system) called the FSP (firmware support package), which handles all of the hardware initialization, including memory and CPU initialization. Reverse engineering and replacing this blob is almost impossible, due to how complex it is. Even for the most skilled developer, it would take years to replace. Intel distributes this blob to firmware developers, without source.
Since the FSP is responsible for the early hardware initialization, that means it also handles SMM (System Management Mode). This is a special mode that operates below the operating system level. It’s possible that rootkits could be implemented there, which could perform a number of attacks on the user (the list is endless). Any Intel system that has the proprietary FSP blob cannot be trusted at all. In fact, several SMM rootkits have been demonstrated in the wild (use a search engine to find them).
The microcode configures logic gates in your CPU, to implement an instruction set architecture. Your CPU will already contain them, but it also supplies a way to update the microcode at boot time, fixing bugs and greatly enhancing the general reliability of your system.
Microcode is already discussed in great detail, on the binary blobs policy.
This interesting video talks about how a group of people reverse engineered the microcode on AMD processors:
The git repository for that project is here:
Both the video and the repository give some further insight about CPU microcode. The way it works on AMD will be very similar to Intel.
For years, coreboot has been struggling against Intel. Intel has been shown to be extremely uncooperative in general. Many coreboot developers, and companies, have tried to get Intel to cooperate; namely, releasing source code for the firmware components. Even Google, which sells millions of chromebooks (coreboot preinstalled) have been unable to persuade them.
Even when Intel does cooperate, they still don’t provide source code. They might provide limited information (datasheets) under strict corporate NDA (non-disclosure agreement), but even that is not guaranteed. Even ODMs and IBVs can’t get source code from Intel, in most cases (they will just integrate the blobs that Intel provides).
Newer Intel graphics chipsets also require firmware blobs.
libreboot project does provide some support for newer Intel platforms, but you should be aware of these issues if you choose to run those machines.
AMD has more or less the same problem as Intel, when it comes to software freedom.
This is basically AMD’s own version of the Intel Management Engine. It has all of the same basic security and freedom issues, although the implementation is wildly different.
The Platform Security Processor (PSP) is built in on the AMD CPUs whose architecture is Late Family 16h (Puma), Zen 17h or later (and also on the AMD GPUs which are GCN 5th gen (Vega) or later). On the CPUs, a PSP controls the main x86 core startup. PSP firmware is cryptographically signed with a strong key similar to the Intel ME. If the PSP firmware is not present, or if the AMD signing key is not present, the x86 cores will not be released from reset, rendering the system inoperable.
The PSP is an ARM core with TrustZone technology, built onto the main CPU die. As such, it has the ability to hide its own program code, scratch RAM, and any data it may have taken and stored from the lesser-privileged x86 system RAM (kernel encryption keys, login data, browsing history, keystrokes, who knows!). To make matters worse, the PSP theoretically has access to the entire system memory space (AMD either will not or cannot deny this, and it would seem to be required to allow the DRM “features” to work as intended), which means that it has at minimum MMIO-based access to the network controllers and any other PCI/PCIe peripherals installed on the system.
In theory any malicious entity with access to the AMD signing key would be able to install persistent malware that could not be eradicated without an external flasher and a known good PSP image. Furthermore, multiple security vulnerabilities have been demonstrated in AMD firmware in the past, and there is every reason to assume one or more zero day vulnerabilities are lurking in the PSP firmware. Given the extreme privilege level (ring -2 or ring -3) of the PSP, said vulnerabilities would have the ability to remotely monitor and control any PSP enabled machine completely outside of the user’s knowledge.
Much like with the Intel Boot Guard (an application of the Intel Management Engine), AMD’s PSP can also act as a tyrant by checking signatures on any boot firmware that you flash, making replacement boot firmware (e.g. libreboot, coreboot) impossible on some boards. Early anecdotal reports indicate that AMD’s boot guard counterpart will be used on most OEM hardware, disabled only on so-called “enthusiast” CPUs.
NOTE: This section is oudated, and it is in need of cleanup.
NOTE: This section may be outdated, and it is in need of cleanup.
Handles some power management for PCIe devices (without this, your laptop will not work properly) and several other power management related features.
The firmware is signed, although on older AMD hardware it is a symmetric key, which means that with access to the key (if leaked) you could sign your own modified version and run it. Rudolf Marek (coreboot hacker) found out how to extract this key in this video demonstration, and based on this work, Damien Zammit (another coreboot hacker) partially replaced it with free firmware, but on the relevant system (ASUS F2A85-M) there were still other blobs present (Video BIOS, and others).
NOTE: More needs to be written about this, to reflect the current reality. The situation with AMD has evolved in recent years. The information on this FAQ page is a few years out of date.
This is responsible for virtually all core hardware initialization on modern AMD systems. In 2011, AMD started cooperating with the coreboot project, releasing this as source code under a free license. In 2014, they stopped releasing source code and started releasing AGESA as binary blobs instead. This makes AGESA now equivalent to Intel FSP.
Read the Intel section practically the same, though it was found with much later hardware in AMD that you could run without microcode updates. It’s unknown whether the updates are needed on all AMD boards (depends on CPU).
The libreboot project does not consider microcode updates a problem, and it enables them by default on all supported hardware.
AMD seemed like it was on the right track in 2011 when it started cooperating with and releasing source code for several critical components to the coreboot project. It was not to be. For so-called economic reasons, they decided that it was not worth the time to invest in the coreboot project anymore. Unfortunately they haven’t even shared the source code of AGESA library for a Family 15h Steamroller/Excavator architectures (which, like the earlier fam15h Bulldozer/Piledriver, do not have a PSP) and released it to a coreboot project only as a binary.
For a company to go from being so good, to so bad, in just 3 years, shows that something is seriously wrong with AMD. Like Intel, they do not deserve your money.
Given the current state of Intel hardware with the Management Engine, it is our opinion that all performant x86 hardware newer than the AMD Family 15h CPUs (on AMD’s side) or anything post-2009 on Intel’s side is defective by design and cannot safely be used to store, transmit, or process sensitive data. Sensitive data is any data in which a data breach would cause significant economic harm to the entity which created or was responsible for storing said data, so this would include banks, credit card companies, or retailers (customer account records), in addition to the “usual” engineering and software development firms. This also affects whistleblowers, or anyone who needs actual privacy and security.
If it’s supported by coreboot, you can add it immediately. Read the porting guide for how to port for a new board. If you are able to generate a working rom for your system, please read lbmk maintenance manual for how to add it to libreboot.
If coreboot lacks support for your hardware, you must add support for it. Please consult the coreboot project for guidance.
Externally rewrite 25xx NOR flash via SPI protocol
It’s possible to use a 16-pin SOIC test clip on an 8-pin SOIC chip, if you align the pins properly. The connection is generally more sturdy.
By default, there is no write-protection on a libreboot system. This is for usability reasons, because most people do not have easy access to an external programmer for re-flashing their firmware, or they find it inconvenient to use an external programmer.
On some systems, it is possible to write-protect the firmware, such that it is rendered read-only at the OS level (external flashing is still possible, using dedicated hardware). For example, on current GM45 laptops (e.g. ThinkPad X200, T400), you can write-protect (see ICH9 gen utility).
It’s possible to write-protect on all libreboot systems, but the instructions need to be written. The documentation is in the main git repository, so you are welcome to submit patches adding these instructions.
TODO: Document PRx based flash protection on Intel platforms, and investigate other methods on AMD systems.
Most libreboot setups actually use the GRUB payload. More information about payloads can be found at coreboot.org/Payloads. SeaBIOS is also available. The CMOS config is hardcoded in libreboot.
The libreboot project inherits the modular payload concept from coreboot, which means that pre-OS bare-metal BIOS setup programs are not very practical. Coreboot (and libreboot) does include a utility called nvramtool, which can be used to change some settings. You can find nvramtool under coreboot/util/nvramtool/, in the libreboot source archives.
The -a option in nvramtool will list the available options, and -w can be used to change them. Consult the nvramtool documentation on the coreboot wiki for more information.
In practise, you don’t need to change any of those settings, in most cases.
Default libreboot setups lock the CMOS table, to ensure consistent functionality for all users. You can use:
nvramtool -C yourrom.rom -w somesetting=somevalue
To get a full list of available options, do this:
nvramtool -C yourrom.rom -a
This will change the default inside that ROM image, and then you can re-flash it.
It is advisable to simply use a larger ROM image. This section was written mostly for ASUS KCMA-D8 and KGPE-D16 mainboards, where previously we only provided 2MiB ROM images in libreboot, but we now provide 16MiB ROM images. Other sizes are not provided because in practise, someone upgrading one of these chips will just use a 16MiB one. Larger sizes are available, but 16MiB is the maximum that you can use on all currently supported libreboot systems that use SPI flash.
Required for ROMs where the ROM image is smaller than the flash chip (e.g. writing a 2MiB ROM to a 16MiB flash chip).
Create an empty (00 bytes) file with a size the difference between the ROM and flash chip. The case above, for example:
truncate -s +14MiB pad.bin
For x86 descriptorless images you need to pad from the beginning of the ROM:
cat pad.bin yourrom.rom > yourrom.rom.new
For ARM and x86 with intel flash descriptor, you need to pad after the image:
cat yourrom.rom pad.bin > yourrom.rom.new
Flash the resulting file. Note that cbfstool will not be able to operate on images padded this way so make sure to make all changes to the image, including runtime config, before padding.
To remove padding, for example after reading it off the flash chip, simply use dd(1) to extract only the non-padded portion. Continuing with the examples above, in order to extract a 2MiB x86 descriptorless ROM from a padded 16MiB image do the following:
dd if=flashromread.rom of=yourrom.rom ibs=14MiB skip=1
With padding removed cbfstool will be able to operate on the image as usual.
Most libreboot setups integrate the GRUB bootloader already, as a payload. This means that the GRUB bootloader is actually flashed, as part of the boot firmware (libreboot). This means that you do not have to install a boot loader on the HDD or SSD, when installing a new distribution. You’ll be able to boot just fine, using the bootloader (GRUB) that is in the flash chip.
This also means that even if you remove the HDD or SSD, you’ll still have a functioning bootloader installed which could be used to boot a live distribution installer from a USB flash drive. See How to install Linux on a libreboot system
Nowadays, other payloads are also provided. If you’re using the SeaBIOS payload, then the normal MBR bootsector is used on your HDD or SSD, like you would expect. So the above paragraphs only apply to the GRUB payload.
Not anymore. Recent versions of libreboot (using the GRUB payload) will automatically switch to a GRUB configuration on the HDD or SSD, if it exists. You can also load a different GRUB configuration, from any kind of device that is supported in GRUB (such as a USB flash drive). For more information, see Modifying the GRUB Configuration in libreboot Systems
If you’re using the SeaBIOS payload, it’s even easier. It works just like you would expect. SeaBIOS implements a normal x86 BIOS interface.
You can find photos of various chip types on the following page:
External 25xx NOR flashing guide
This was reported by a user, running Debian 11 with kernel
thinkpad_acpi module was not loading, with the following message:
modprobe: ERROR: could not insert 'thinkpad_acpi': "No such device"
Battery info in
/sys was absent, because of this. The user reported that the following workaround was effective (in Debian).
Add this line to
options thinkpad_acpi force_load=1
You can install the
tlp package and start that service. For example, on Debian:
apt-get install tlp tlp-rdw systemctl enable tlp systemctl start tlp
Now read the manual:
As root, you can do:
This will provide information about the battery.
You can also read information about these in the libreboot binary blob reduction policy, where it goes into more detail about some of them.
The Video BIOS is present on most video cards. For integrated graphics, the VBIOS (special kind of OptionROM) is usually embedded in the main boot firmware. For external graphics, the VBIOS is usually on the graphics card itself. This is usually proprietary; the only difference is that SeaBIOS can execute it (alternatively, you embed it in a coreboot ROM image and have coreboot executes it, if you use a different payload, such as GRUB).
The coreboot project provides free initialization code, on many boards, and libreboot will use this code when it is available, depending on the configuration.
In configurations where SeaBIOS and native GPU init are used together, a special shim VBIOS is added that uses coreboot linear framebuffer.
Most (all?) laptops have this. The EC (embedded controller) is a small, separate processor that basically processes inputs/outputs that are specific to laptops. For example:
EC is present on nearly all laptops. Other devices use, depending on complexity, either EC or variant with firmware in Mask ROM - SuperIO.
HDDs and SSDs have firmware in them, intended to handle the internal workings of the device while exposing a simple, standard interface (such as AHCI/SATA) that the OS software can use, generically. This firmware is transparent to the user of the drive.
HDDs and SSDs are quite complex, and these days contain quite complex hardware which is even capable of running an entire operating system (by this, we mean that the drive itself is capable of running its own embedded OS), even Linux.
SSDs and HDDs are a special case, since they are persistent storage devices as well as computers.
Example attack that malicious firmware could do: substitute your SSH keys, allowing unauthorized remote access by an unknown adversary. Or maybe substitute your GPG keys. SATA drives can also have DMA (through the controller), which means that they could read from system memory; the drive can have its own hidden storage, theoretically, where it could read your LUKS keys and store them unencrypted for future retrieval by an adversary.
With proper IOMMU and use of USB instead of SATA, it might be possible to mitigate any DMA-related issues that could arise.
Some proof of concepts have been demonstrated. For HDDs: https://spritesmods.com/?art=hddhack&page=1 For SSDs: http://www.bunniestudios.com/blog/?p=3554
Viable free replacement firmware is currently unknown to exist. For SSDs, the OpenSSD project may be interesting.
Apparently, SATA drives themselves don’t have DMA but can make use of it through the controller. This http://web.archive.org/web/20170319043915/http://www.lttconn.com/res/lttconn/pdres/201005/20100521170123066.pdf (pages 388-414, 420-421, 427, 446-465, 492-522, 631-638) and this http://www.intel.co.uk/content/dam/www/public/us/en/documents/technical-specifications/serial-ata-ahci-spec-rev1_3.pdf (pages 59, 67, 94, 99).
The following is based on discussion with Peter Stuge (CareBear\) in the coreboot IRC channel on Friday, 18 September 2015, when investigating whether the SATA drive itself can make use of DMA. The following is based on the datasheets linked above:
According to those linked documents, FIS type 39h is “DMA Activate FIS - Device to Host”. It mentions “transfer of data from the host to the device, and goes on to say: Upon receiving a DMA Activate, if the host adapter’s DMA controller has been programmed and armed, the host adapter shall initiate the transmission of a Data FIS and shall transmit in this FIS the data corresponding to the host memory regions indicated by the DMA controller’s context.” FIS is a protocol unit (Frame Information Structure). Based on this, it seems that a drive can tell the host controller that it would like for DMA to happen, but unless the host software has already or will in the future set up this DMA transfer then nothing happens. A drive can also send DMA Setup. If a DMA Setup FIS is sent first, with the Auto-Activate bit set, then it is already set up, and the drive can initiate DMA. The document goes on to say “Upon receiving a DMA Setup, the receiver of the FIS shall validate the received DMA Setup request.” - in other words, the host is supposed to validate; but maybe there’s a bug there. The document goes on to say “The specific implementation of the buffer identifier and buffer/address validation is not specified” - so noone will actually bother. “the receiver of the FIS” - in the case we’re considering, that’s the host controller hardware in the chipset and/or the kernel driver (most likely the kernel driver). All SATA devices have flash-upgradeable firmware, which can usually be updated by running software in your operating system; malicious software running as root could update this firmware, or the firmware could already be malicious. Your HDD or SSD is the perfect place for a malicious adversary to install malware, because it’s a persistent storage device as well as a computer.
Based on this, it’s safe to say that use of USB instead of SATA is advisable if security is a concern. USB 2.0 has plenty of bandwidth for many HDDs (a few high-end ones can use more bandwidth than USB 2.0 is capable of), but for SSDs it might be problematic. USB 3.0 will provide more reasonable performance, though note that depending on the system, you may have to deal with binary blob XHCI firmware in your kernel (if that bothers you).
Use of USB is also not an absolute guarantee of safety, so do beware. The attack surface becomes much smaller, but a malicious drive could still attempt a “fuzzing” attack (e.g. sending malformed USB descriptors, which is how the tyrant DRM on the Playstation 3 was broken, so that users could run their own operating system and run unsigned code). (you’re probably safe, unless there’s a security flaw in the USB library/driver that your OS uses. USB is generally considered one of the safest protocols, precisely because USB devices have no DMA)
It is recommended that you use full disk encryption, on HDDs connected via USB. There are several adapters available online, that allow you to connect SATA HDDs via USB, and Libreboot is capable of booting from them the normal way. Consult the documentation for your Linux/BSD operating system, so that you can know how to install it with full disk encryption.
The current theory (unproven) is that this will at least prevent malicious drives from wrongly manipulating data being read from or written to the drive, since it can’t access your LUKS key if it’s only ever in RAM, provided that the HDD doesn’t have DMA (USB devices don’t have DMA). The worst that it could do in this case is destroy your data. Of course, you should make sure never to put any keyfiles in the LUKS header. Take what this paragraph says with a pinch of salt. This is still under discussion, and none of this is proven.
Ethernet NICs will typically run firmware inside, which is responsible for initializing the device internally. Theoretically, it could be configured to drop packets, or even modify them.
With proper IOMMU, it might be possible to mitigate the DMA-related issues. A USB NIC can also be used, which does not have DMA.
Microcode configures logic gate arrays in a microprocessor, to implement the instruction set architecture. Special decoders in the microprocessor will configure the circuitry, based on that microcode.
The libreboot blob reduction policy goes into great detail about microcode.
Sound hardware (integrated or discrete) typically has firmware on it (DSP) for processing input/output. Again, a USB DAC is a good workaround.
Webcams have firmware integrated into them that process the image input into the camera; adjusting focus, white balancing and so on. Can use USB webcam hardware, to work around potential DMA issues; integrated webcams (on laptops, for instance) are discouraged by the libreboot project, for security reasons.
USB host controllers require firmware. Sometimes, this has to be supplied by coreboot itself.
Some laptops might have a simcard reader in them, with a card for handling WWAN, connecting to a 3g/4g (e.g. GSM) network. This is the same technology used in mobile phones, for remote network access (e.g. internet).
NOTE: not to be confused with wifi. Wifi is a different technology, and entirely unrelated.
The baseband processor inside the WWAN chip will have its own embedded operating system, most likely proprietary. Use of this technology also implies the same privacy issues as with mobile phones (remote tracking by the GSM network, by triangulating the signal).
On some laptops, these cards use USB (internally), so won’t have DMA, but it’s still a massive freedom and privacy issue. If you have an internal WWAN chip/card, the libreboot project recommends that you disable and (ideally, if possible) physically remove the hardware. If you absolutely must use this technology, an external USB dongle is much better because it can be easily removed when you don’t need it, thereby disabling any external entities from tracking your location.
Use of ethernet or wifi is recommended, as opposed to mobile networks, as these are generally much safer.
Absolutely! It is well-tested in libreboot, and highly recommended. See installing Linux and booting Linux.
Any recent distribution should work, as long as it uses KMS (kernel mode setting) for the graphics.
On Fedora, by default the grub.cfg tries to boot linux in 16-bit mode. You just have to modify Fedora’s GRUB configuration. Refer to the Linux page.
Absolutely! The libreboot firmware has good support for FreeBSD, NetBSD and OpenBSD. Other systems are untested, but should work just fine.
Unknown. Perhaps so, but it’s impossible to say without further testing.
Please read the libreboot binary blob minimalisation policy.
Please also read: Software and hardware freedom status for each mainboard supported by Libreboot
The libreboot firmware provides host hardware initialisation inside ROM files, that can be written to NOR flash, but on many systems there exist a lot more small computers on the mainboard running blob firmware. Some of them are not practicable to replace due to being located on Mask ROM. Most laptops have EC (Embedded Controller) firmware, for example.
Besides software itself (embedded in ROM or not), most hardware (from ICs to circuit boards) are not released under open source licenses. We do not have a single device that can be considered be “100% free”, and such absolutes are nearly impossible to reach.
Notable proprietary blobs (not a complete list):
One day, we will live in a world where anyone can get their own chips made, including CPUs but also every other type of IC. Efforts to make homemade chip fabrication a reality are now in their infancy, but such efforts do exist, for example, the work done by Sam Zeloof and the Libre Silicon project:
(Sam literally makes CPUs in his garage)
yt-dlp (an enhanced fork of
youtube-dl) is recommended for links to
youtube.com. See: https://github.com/yt-dlp/yt-dlp
Lastly the most important message to everybody gaining this wonderful new hobby - Secret to Learning Electronics
Markdown file for this page: https://libreboot.org/faq.md
This HTML page was generated by the untitled static site generator.