It's Now Easier To Setup A Linux Payload For Coreboot
Phoronix: It's Now Easier To Setup A Linux Payload For Coreboot
One of the unique benefits of the Coreboot open-source BIOS/firmware project is that it can support loading the Linux kernel directly after initializing the motherboard -- instead of using GRUB2/SeaBIOS, the Linux kernel can be included on the ROM chip. This isn't a new feature to Coreboot, but with a Git commit made today it's now easier to configure...
Putting the Linux kernel on a rom BIOS chip is an exceptionally bad idea.
Seems like the article might have a mistake, misunderstanding in it. They happen.
Really cool but as long as we don't have a mobo/laptop/HW that we can use this its useless.
Originally Posted by plonoma
Originally Posted by http://www.coreboot.org/Payloads#Linux
With this you can boot a minimal Linux system (think no X) in like 1 second only.
I have been amazed this is not the common way to boot your system. With an optimized firmware like coreboot and a slim Linux setup modern hardware can boot in just a couple of seconds. Fedora can load in under 3 seconds, so the most time of boot is spent in BIOS/UEFI/BOOTLOADER on modern systems.
Last edited by varikonniemi; 09-01-2013 at 03:59 AM.
Yes, but every kernel update (which for modern hardware specially, might bring interesting things) implies flashing the BIOS again. I don't know if there's any limit on how many times you can do that without side effects, but I'd expect that to exist. It's not usual to flash your BIOS once every two or three months, I don't think they are designed for such often updates.
Originally Posted by varikonniemi
Flash is typically designed for a couple thousand rewrite cycles, so that should not be a problem. However, updating flash is a comparably risky situation, so I would advise against trying to stuff your distro kernel (which seem to be updated every 4 days on average) into flash every time.
Originally Posted by mrugiero
Adding Linux as a payload (instead of loading Linux via SeaBIOS, GRUB2, FILO, or even Tianocore) is a rather special feature, used for uncommon situations like embedded systems. I guess the main value comes from using the Linux in flash as a loader for an on-disk version - Linux's kexec feature is just made for that. This allows developing the boot process (which may include updating the OS by various means, recovery, signature checking, ...) in a known, common, and time tested environment (read: the Linux userland) instead of whatever the bootloaders give you.
"Regular coreboot users" (to the degree that such people actually exist ;-) ) will probably be more happy with a payload whose modus operandi is supported by their main operating system (that is, SeaBIOS or Tianocore as a payload for common Linux, Windows, BSD, and so on) - after all the OS is what they _really_ want to use.
Disclosure: I'm the author of the new Linux-as-payload mechanism (and as a sidenote, the change is more than just a bunch of Kconfig variables; see the grandparent commit of the Kconfig change).
not a bad idea - in fact having an ELF kernel (preferably linux) in the boot rom/flash together with minimal init code, was the original goal and intended setting of the LinuxBIOS project - motivated by:
Originally Posted by plonoma
the option and associated mechanism to boot an arbitrary operating system being unnecessary in an embedded / clustered environment (the context the project was born in) where the os (or at least the kernel) is fixed
besides being possibly a point of failure, with the bios not exactly being the best when it comes to manageability (think of a headless cluster node stuck in a POST "select boot device" menu or in a "keyboard fail - press F1 to continue" error);
the bios' device probe and initialization feature becoming redundant with linux having, or in the process of getting, code to perform it natively instead of relying on external legacy (16 bit) code;
They made the pragmatic move to being a bootloader for normal, boot-media-installed, Os's - and changed name to CoreBoot - when it became evident that the market (boards and chip manufacturers) would not move to higher capacity bios chips (although for a while the decrease in flash memory cost suggested that chips sufficient for a small operating system, would become available)
Last edited by silix; 09-01-2013 at 09:31 AM.