Re:Replacing x86 firmware with Linux and Go
by jabl on Tuesday November 21, 2017 @02:35PM
(#15750819)
> Linux has MILLIONS of lines of code. Please don't. EFI is already bloated as it is.
Well, if you're going to run Linux anyway, running Linux as your firmware + bootloader doesn't increase your attack surface. And, it can be argued that e.g. the Linux networking stack is more battle tested than the UEFI one.
> Support efforts like coreboot instead.
Ron Minnich is the father of coreboot. If it were possible to run coreboot on modern Intel server platforms, I'm certain that's what he would propose. As a sibling commenter mentioned, he views NERF as a backup solution if using coreboot isn't possible.
> And FFS, firmware should not persist once the operating system boots. Persistent firmware is cancer.
In NERF, the Linux kernel burned on the flash rom kexec()'s the final distro kernel. IOW, it replaces itself by the new kernel, it doesn't linger around in the background.
Well, if you're going to run Linux anyway, running Linux as your firmware + bootloader doesn't increase your attack surface. And, it can be argued that e.g. the Linux networking stack is more battle tested than the UEFI one.
> Support efforts like coreboot instead.
Ron Minnich is the father of coreboot. If it were possible to run coreboot on modern Intel server platforms, I'm certain that's what he would propose. As a sibling commenter mentioned, he views NERF as a backup solution if using coreboot isn't possible.
> And FFS, firmware should not persist once the operating system boots. Persistent firmware is cancer.
In NERF, the Linux kernel burned on the flash rom kexec()'s the final distro kernel. IOW, it replaces itself by the new kernel, it doesn't linger around in the background.