Re: Yet another [cross] installer
Hello,
2010/3/3 Oliver Kiddle <okiddle@yahoo.co.uk>:
> If you consider what makes the non-x86 hardware different it is the lack
> of a standard way to boot an alternate OS. With a PC BIOS you can instruct
> a PC to boot from a floppy/CD/whatever and we only need a single CD image
> to boot most PCs. Even for headless x86 servers, it's never hard to get
> whatever OS on.
>
> If ARM really think they can challenge Intel Atoms in the netbook market,
> then I think they need to address this because a significant number of
> people will care about being able to upgrade the OS or to choose between
> Android, WinCE, Ubuntu Netbook Remix or MeeGo.
At least some people is aware of such issues and are trying to
convince (arm) kernel people to introduce device trees[0], whilst it
has been widely used by powerpc people (also xilinx is adopting it[1],
I *guess* they'll *probably* use it for new AMBA bus design for their
ARM+FPGA).
> Most of these devices use u-boot (or RedBoot) and as these are
> open source, perhaps it would be wise to make them (in their default
> configuration) do things to make life easy for us. As far as I can tell
> from a quick scan of the documentation, u-boot has support for putting a
> logo on an LCD during boot but not for using that screen and a keyboard as
> a console. If a certain key press during boot switched u-boot to providing
> a console then many manufacturers might leave it in (either intentionally
> or due to ignorance). For me the hardest part of implementing a text
> console or even BIOS style menus in an open source bootloader would be
> getting a setup where I'm not going to brick my shiny new netbook.
The non-x86 installer shall have an interface to customize any kind of
loader or pre-loader, at least have an option to pass a configuration
file for it. I also believe environment variables should be kept in a
separated partition so those can be easily accesible from kernel land.
> Other similar features would be supporting consoles with USB-serial
> converters and network connections at carefully timed points in the boot
> process.
Network/SD/MMC/USB boot is surely a must and those should be first
devices to tackle. Then we also have MTD to think about. Having a
serial debug port is desirable.
> For any device where the manufacturers try to close it off to hackers,
> we're always going to have to jump through hoops specific to that device
> to get Debian installed. They can easily disable bootloader
> features but on netbooks, maybe they won't: PC makers are used to having
> a BIOS with screens saying "Press F1 to enter Setup".
It is nice if we can hack manufacturers toys, but if those are
non-free to hackers, why should we "officially" support them?
Best regards,
[0] http://marc.info/?l=linux-kernel&m=126660762616227&w=2
[1] http://git.xilinx.com/cgi-bin/gitweb.cgi?p=device-tree.git;a=summary
--
Héctor Orón
"Our Sun unleashes tremendous flares expelling hot gas into the Solar
System, which one day will disconnect us."
Reply to: