See “Version” in the documentation
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-libre, 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 Parabola, you can try:
$ sudo systemctl restart network-manager
(the service name might be different for you, depending on your configuration)
Libreboot 20160818, 20160902 and 20160907 all have a bug: in SeaBIOS, PCI options ROMs are loaded when available, by default. This is not technically a problem, because an option ROM can be free or non-free. In practise, though, they are usually non-free.
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, or petitboot), or to boot (with SeaGRUB and/or SeaBIOS) from regular SATA and then use it in GNU+Linux. The Linux kernel is capable of using the PIKE2008 module without loading the option ROM.
Libreboot-unstable (or git) now disables loading PCI option ROMs, but previous releases with SeaGRUB (20160818-20160907) do not. You can work around this by running the following command:
$ ./cbfstool yourrom.rom add-int -i 0 -n etc/pci-optionrom-exec
You can find cbfstool in the _util archive with the libreboot release that you are using.
See the hardware compatibility list.
Short answer: no.
There are severe privacy, security and freedom issues with these laptops, due to the Intel chipsets that they use. See:
Most notably, these laptops also use the Intel FSP binary blob, for the entire hardware initialization. Coreboot does support a particular revision of one of their laptops, but most are either unsupported or rely on binary blobs for most of the hardware initialization.
In particular, the Intel Management Engine is a severe threat to privacy and security, not to mention freedom, since it is a remote backdoor that provides Intel remote access to a computer where it is present.
Intel themselves even admitted it, publicly.
The Libreboot project recommends avoiding all hardware sold by Purism.
It is unlikely that any post-2008 Intel hardware will ever be supported in libreboot, due to severe security and freedom issues; so severe, that the libreboot project recommends avoiding all modern Intel hardware. If you have an Intel based system affected by the problems described below, then you should get rid of it as soon as possible. The main issues are as follows:
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. libreboot does this on the Intel 4 Series systems that it supports, such as the Libreboot X200 and Libreboot 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.
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
Also see (effort to disable the ME): https://www.coreboot.org/pipermail/coreboot/2016-November/082331.html - look at the whole thread
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).
All modern x86 CPUs (from Intel and AMD) use what is called microcode. CPUs are extremely complex, and difficult to get right, so the circuitry is designed in a very generic way, where only basic instructions are handled in hardware. Most of the instruction set is implemented using microcode, which is low-level software running inside the CPU that can specify how the circuitry is to be used, for each instruction. The built-in microcode is part of the hardware, and read-only. Both the circuitry and the microcode can have bugs, which could cause reliability issues.
Microcode updates are proprietary blobs, uploaded to the CPU at boot time, which patches the built-in microcode and disables buggy parts of the CPU to improve reliability. In the past, these updates were handled by the operating system kernel, but on all recent systems it is the boot firmware that must perform this task. Coreboot does distribute microcode updates for Intel and AMD CPUs, but libreboot cannot, because the whole point of libreboot is to be 100% free software.
On some older Intel CPUs, it is possible to exclude the microcode updates and not have any reliability issues in practise. All current libreboot systems work without microcode updates (otherwise, they wouldn’t be supported in libreboot). However, all modern Intel CPUs require the microcode updates, otherwise the system will not boot at all, or it will be extremely unstable (memory corruption, for example).
Intel CPU microcode updates are signed, which means that you could not even run a modified version, even if you had the source code. If you try to upload your own modified updates, the CPU will reject them.
The microcode updates alter the way instructions behave on the CPU. That means they affect the way the CPU works, in a very fundamental way. That makes it software. The updates are proprietary, and are software, so we exclude them from libreboot. The microcode built into the CPU already is not so much of an issue, since we can’t change it anyway (it’s read-only).
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).
Recent Intel graphics chipsets also require firmware blobs.
Intel is only going to get worse when it comes to user freedom. Libreboot has no support recent Intel platforms, precisely because of the problems described above. The only way to solve this is to get Intel to change their policies and to be more friendly to the free software community. Reverse engineering won’t solve anything long-term, unfortunately, but we need to keep doing it anyway. Moving forward, Intel hardware is a non-option unless a radical change happens within Intel.
Basically, all Intel hardware from year 2010 and beyond will never be supported by libreboot. The libreboot project is actively ignoring all modern Intel hardware at this point, and focusing on alternative platforms.
It is extremely unlikely that any post-2013 AMD hardware will ever be supported in libreboot, due to severe security and freedom issues; so severe, that the libreboot project recommends avoiding all modern AMD hardware. If you have an AMD based system affected by the problems described below, then you should get rid of it as soon as possible. The main issues are as follows:
We call on AMD to release source code and specs for the new AMD Ryzen platforms! We call on the community to put pressure on AMD. Click here to read more
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 all Family 16h + systems (basically anything post-2013), and 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.
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) preventing the hardware from being supported in libreboot.
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).
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.
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.
Libreboot has support for fam15h AMD hardware (~2012 gen) and some older Intel platforms like Napa, Montevina, Eagle Lake, Lakeport (2004-2006). We also have support for some ARM chipsets (rk3288). On the Intel side, we’re also interested in some of the chipsets that use Atom CPUs (rebranded from older chipsets, mostly using ich7-based southbridges).
Short answer: yes. These laptops also have an Intel GPU inside, which libreboot uses. The ATI GPU is ignored by libreboot.
These laptops use what is called switchable graphics, where it will have both an Intel and ATI GPU. Coreboot will allow you to set (using nvramtool) a parameter, specifying whether you would like to use Intel or ATI. The ATI GPU lacks free native graphics initialization in coreboot, unlike the Intel GPU.
Libreboot modifies coreboot, in such a way where this nvramtool setting is ignored. Libreboot will just assume that you want to use the Intel GPU. Therefore, the ATI GPU is completely disabled on these laptops. Intel is used instead, with the free native graphics initialization (VBIOS replacement) that exists in coreboot.
Libreboot now supports desktop hardware: (see list) (with full native video initialization).
A common issue with desktop hardware is the Video BIOS, when no onboard video is present, since every video card has a different Video BIOS. Onboard GPUs also require one, so those still have to be replaced with free software (non-trivial task). Libreboot has to initialize the graphics chipset, but most graphics cards lack a free Video BIOS for this purpose. Some desktop motherboards supported in coreboot do have onboard graphics chipsets, but these also require a proprietary Video BIOS, in most cases.
Most likely not. First, you must consult coreboot’s own hardware compatibility list at http://www.coreboot.org/Supported_Motherboards and, if it is supported, check whether it can run without any proprietary blobs in the ROM image. If it can: wonderful! Libreboot can support it, and you can add support for it. If not, then you will need to figure out how to reverse engineer and replace (or remove) those blobs that do still exist, in such a way where the system is still usable in some defined way.
For those systems where no coreboot support exists, you must first port it to coreboot and, if it can then run without any blobs in the ROM image, it can be added to libreboot. See: Motherboard Porting Guide (this is just the tip of the iceberg!)
Please note that board development should be done upstream (in coreboot) and merged downstream (into libreboot). This is the correct way to do it, and it is how the libreboot project is coordinated so as to avoid too much forking of the coreboot source code.
Libreboot has support for some ARM based laptops, using the Rockchip RK3288 SoC. Check the libreboot hardware compatibility list, for more information.
See installation guide
SPI flash chips can be programmed with the BeagleBone Black or the Raspberry Pi.
If you are using the GRUB payload, you can add a username and password (salted, hashed) to your GRUB configuration that resides inside the flash chip. The following guides (which also cover full disk encryption, including the /boot/ directory) show how to set a boot password in GRUB: (Installing Debian or Devuan with FDE) and (Installing Parabola or Arch GNU+Linux-Libre, with FDE)
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.
Libreboot actually uses the GRUB payload. More information about payloads can be found at coreboot.org/Payloads.
Libreboot 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.
Libreboot locks the CMOS table, to ensure consistent functionality for all users. You can use:
$ nvramtool -C yourrom.rom -w somesetting=somevalue
This will change the default inside that ROM image, and then you can re-flash it.
Libreboot integrates 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 GNU+Linux on a libreboot system
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
SOIC-8 SPI flash chip:
SOIC-16 SPI flash chip:
See the license information.
The Libreboot logo is available as a bitmap, a vector, or a greyscale vector.
Libreboot Inside stickers are available as a PDF or a vector
The main freedom issue on any system, is the boot firmware (usually referred to as a BIOS or UEFI). Libreboot replaces the boot firmware with fully free code, but even with libreboot, there may still be other hardware components in the system (e.g. laptop) that run their own dedicated firmware, sometimes proprietary. These are on secondary processors, where the firmware is usually read-only, written for very specific tasks. While these are unrelated to libreboot, technically speaking, it makes sense to document some of the issues here.
Note that these issues are not unique to libreboot systems. They apply universally, to most systems. The issues described below are the most common (or otherwise critical).
Dealing with these problems will most likely be handled by a separate project.
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).
On current libreboot systems, instead of VBIOS, coreboot native GPU init is used, which is currently only implemented for Intel GPUs. Other cards with proper KMS drivers can be initialized once Linux boots, but copy of VBIOS may be still needed to fetch proper VRAM frequency and other similar parameters (without executing VBIOS code).
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:
Alexander Couzens from coreboot (lynxis on coreboot IRC) is working on a free EC firmware replacement for the ThinkPads that are supported in libreboot. See: https://github.com/lynxis/h8s-ec (not ready yet).
Most (all?) chromebooks have free EC firmware. Libreboot is currently looking into supporting a few ARM-based chromebooks.
EC is only present on laptops. On desktop/server boards it is absent (not required).
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 GNU+Linux or BusyBox/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://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 (unless you’re using USB 3.0, which is not yet usable in freedom. See
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. Libreboot documents how to install several distributions with full disk encryption. You can adapt these for use with USB drives:
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.
Implements an instruction set. See description. Here we mean microcode built in to the CPU. We are not talking about the updates supplied by the boot firmware (libreboot does not include microcode updates, and only supports systems that will work without it) Microcode can be very powerful. No proof that it’s malicious, but it could theoretically
There isn’t really a way to solve this, unless you use a CPU which does not have microcode. (ARM CPUs don’t, but most ARM systems require blobs for the graphics hardware at present, and typically have other things like soldered wifi which might require blobs)
CPUs often on modern systems have a processor inside it for things like power management. ARM for example, has lots of these.
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.
Doesn’t really apply to current libreboot systems (none of them have USB 3.0 at the moment), but USB 3.0 host controllers typically rely on firmware to implement the XHCI specification. Some newer coreboot ports also require this blob, if you want to use USB 3.0.
This doesn’t affect libreboot at the moment, because all current systems that are supported only have older versions of USB available. USB devices also don’t have DMA (but the USB host controller itself does).
With proper IOMMU, it might be possible to mitigate the DMA-related issues (with the host controller).
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.
On all current libreboot laptops, it is possible to remove the WWAN card and sim card if it exists. The WWAN card is next to the wifi card, and the sim card (if installed) will be in a slot underneath the battery, or next to the RAM.
Absolutely! It is well-tested in libreboot, and highly recommended. See installing GNU+Linux and booting GNU+Linux.
Any recent distribution should work, as long as it uses KMS (kernel mode setting) for the graphics.
Absolutely! Libreboot has native support for NetBSD, OpenBSD and LibertyBSD. Other distros are untested.
Unknown. Probably not.
Edit this page – License – Template – Authors – Conduct Guidelines – Governance – Peers Community