LinuxBoot: a new Linux Foundation project for boot firmware
Firmware has always had a simple purpose: to boot the OS. Achieving that has become much more difficult due to increasing complexity of both hardware and deployment. Firmware often must set up many components in the system, interface with more varieties of boot media, including high-speed storage and networking interfaces, and support advanced protocols and security features. LinuxBoot addresses the often slow, often error-prone, obscured code that executes these steps with a Linux kernel. The result is a system that boots in a fraction of the time of a typical system, and with greater reliability."
Posted Jan 27, 2018 3:34 UTC (Sat)
by unixbhaskar (guest, #44758)
[Link] (2 responses)
Posted Jan 27, 2018 11:15 UTC (Sat)
by unixbhaskar (guest, #44758)
[Link] (1 responses)
Posted Jan 28, 2018 22:08 UTC (Sun)
by jaymzh (subscriber, #57150)
[Link]
"
Initially, all three of these projects were called NERF. Because this would have only become more confusing, and because we do not want to be prescriptive of the initramfs, we named the Linux kernel in firmware LinuxBoot."
Posted Jan 27, 2018 6:45 UTC (Sat)
by flussence (guest, #85566)
[Link] (2 responses)
There's a similar “value-add” in my slow Atom netbook, where an upstreamed kernel driver lets me squeeze a very needed few hundred MHz out of it at runtime by twiddling a sysfs file. But that's also an interface to something proprietary in the BIOS.
I'm all for anything that gets rid of UEFI and replaces it with software that doesn't feel frankensteined together in a weekend — but they need to understand they're competing against incumbents using this 25 year old marketing trick to maintain control.
Posted Jan 27, 2018 12:34 UTC (Sat)
by merge (subscriber, #65339)
[Link]
You'd only use the parts up to the DXE interface (implemented by the vendor; you have the binary only). LinuxBoot uses that and takes over from there. I can imagine binary-exaction or -modification tools to be possible. ... or am I totally mistaken? :)
Posted Jan 27, 2018 21:36 UTC (Sat)
by xtifr (guest, #143)
[Link]
Yes, for something like this to have any chance, it would probably have to be backed by a huge consortium whose members include many of the biggest players in tech. Companies like Google, Intel, IBM, Oracle, Dell, Cisco, Samsung, AMD, Comcast, HP, Lenovo, Sony, Netflix, Facebook, and maybe even Microsoft.
Something like...say...the Linux Foundation! :D
Of course, being a LF project doesn't guarantee success--many of the companies that make up the Foundation may not be interested in or supportive of this particular project. But the backing of the Foundation is about the best possible support such a project could have. I have a lot of doubts about and mistrust of LF, but their backing here seems like a pretty good thing overall.
Posted Jan 27, 2018 7:56 UTC (Sat)
by russell (guest, #10458)
[Link] (15 responses)
Wouldn't this be just another piece of firmware that does not get updated?
Why not eliminate all firmware and have your choosen operating system install it's loader in eeprom, as the very first instruction executed?
Machine configuration could be given to the operating system as data tables and ONLY data tables?
This would be great for security. Firmware is never updated, but operating systems do update.
Posted Jan 27, 2018 8:46 UTC (Sat)
by alison (subscriber, #63752)
[Link] (3 responses)
Hopefully when we can purchase ARM64 laptops (for I do believe they're coming), it will be possible to boot them with device-trees rather than ACPI tables.
Posted Jan 27, 2018 11:55 UTC (Sat)
by amacater (subscriber, #790)
[Link] (2 responses)
I just hope that this doesn't become another LF "only commercial consortia need really apply" project like some of the container/networking projects.
Posted Jan 27, 2018 18:09 UTC (Sat)
by khim (subscriber, #9252)
[Link] (1 responses)
Posted Jan 27, 2018 21:37 UTC (Sat)
by alison (subscriber, #63752)
[Link]
Posted Jan 27, 2018 11:14 UTC (Sat)
by unixbhaskar (guest, #44758)
[Link] (2 responses)
please take a peek...
Posted Jan 27, 2018 16:31 UTC (Sat)
by Wol (subscriber, #4433)
[Link] (1 responses)
So we'll have linux-based coreboot, firing up a linux-based UEFI, firing up a linux-based grub, firing up linux itself ...
way to go!
Cheers,
Posted Jan 27, 2018 19:54 UTC (Sat)
by darwish (guest, #102479)
[Link]
Posted Jan 27, 2018 12:49 UTC (Sat)
by dowdle (subscriber, #659)
[Link] (1 responses)
With the Intel ME issues from a month or two ago... and the more recent Meltdown / Spectre stuff, Dell has pretty much produced two BIOS updates for all of their makes and models released in the last 5 years. I think they were working on going even further back when Intel announced that they were withdrawing the microcode for Spectre variant 2... so Dell pulled the updates that included that and are working on the next round with whatever Intel comes up with. One particular model I've had to update is the OptiPlex 9010 which has a BIOS version of A27. While they might have skipped a version or two, I'd guess there have been 20+ BIOS updates in the 5 years that model has been out.
While it is true some manufacturers don't update their firmware, you should be avoiding those vendors rather than rewarding them.
Posted Jan 27, 2018 21:38 UTC (Sat)
by Felix (guest, #36445)
[Link]
That means you can usually upgrade the UEFI pretty easily from Linux (unfortunately while they support the Latitude E5X60 and E7X70 my E5X70 is not on the list). It would be great if they could also add their servers to LVFS but I guess there is no enterprise distro with fwupd support right now so it might be a bit pointless...
Posted Jan 27, 2018 16:04 UTC (Sat)
by khim (subscriber, #9252)
[Link]
Posted Jan 27, 2018 19:39 UTC (Sat)
by flussence (guest, #85566)
[Link] (1 responses)
Posted Jan 27, 2018 21:49 UTC (Sat)
by alison (subscriber, #63752)
[Link]
Listening to Minnich's ELCE presentation, my takeaway was the code on the PCH is not being replaced, but only that which runs on the main processor. AFAICT, me_cleaner removes some FW from the PCH, but NERF does not replace it.
Posted Jan 29, 2018 9:29 UTC (Mon)
by mfuzzey (subscriber, #57966)
[Link] (2 responses)
People often want to run multiple operating systems on the same hardware.
Also the boot firmware is relatively independent of the OS being be used.
So, it makes more sense to have OS independent bootfirware and bootloaders that can be used for multiple operating systems (at one time or not depending on the user's wishes).
This also makes it easier to "safely" experiment with new operating systems on existing hardware if you can boot them without updating the firmware.
Finally, updating the boot firmware is a risky operation. If something goes wrong in the update procedure, or if the new firmware is simply sufficiently buggy you can brick your hardware and be stuck unless you have the right tools (JTAG probe, or even a soldering iron) to fix it.
Posted Jan 29, 2018 15:12 UTC (Mon)
by tao (subscriber, #17563)
[Link] (1 responses)
ITYM:
"A small subset of the already small subset of people running something else than Windows on their hardware"
Posted Jan 31, 2018 14:40 UTC (Wed)
by nim-nim (subscriber, #34454)
[Link]
Posted Jan 27, 2018 10:51 UTC (Sat)
by lamby (subscriber, #42621)
[Link] (2 responses)
Posted Jan 27, 2018 20:19 UTC (Sat)
by pturmel (guest, #95781)
[Link]
Posted Feb 14, 2018 21:57 UTC (Wed)
by thomasg (guest, #114185)
[Link]
Its very hard to imagine google developing GPL code when not bound to do so.
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
* LinuxBoot is the project that replaces specific firmware functionality with a Linux kernel. LinuxBoot is agnostic to what initramfs is used with the kernel.
* NERF is LinuxBoot with u-root as the initramfs. u-root contains boot policy tools in Go (e.g. PXE booting, booting via GRUB config) among standard busybox-like utilities rewritten in Go.
* HEADS is a secure runtime that can be used as the initramfs for LinuxBoot. Take a look at the LinuxBoot branch of HEADS. See osresearch.net for more documentation on HEADS.
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
Wol
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
Microsoft (rightly) got a lot of flack some time back for silently overwriting the MBR when installing Windows.
Overwriting boot firmware would be even worse.
Most of the work is setting up the hardware and loading stuff from media.
Only the last stage, actually passing control, with any data that may be required, may depend on the OS (and standards like UEFI can make even that OS independent).
So, it's probably not a good idea to force firmware updates as part of a OS update.
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
2. All the repair and maintenance tools also need booting (and it's idiotic to boot them from the broken system they're supposed to repair)
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
LinuxBoot: a new Linux Foundation project for boot firmware
