[yocto] [PATCH] EFI support for IA

[Darren Hart]

On 11/29/2011 01:06 AM, Josef Ahmad wrote:
> Darren,
> 
> My bad. I added the two configuration statements into local.conf.
> Now I moved the two statements into my machine configuration files, and
> I confirm it works. The system image is deployed including bootia32.efi
> and configuration file, and I successfully booted the kernel.
> 

Excellent!

> Good work!
> Regarding Grub vs Syslinux as the adopted bootloader, I agree with
> choosing the second one as the default, as soon as an EFI release of
> Syslinux shows up.
> At the same time though, I reckon it might be worthy to keep Grub as a
> second choice for EFI bootloading, rather than drop its support. Also,
> because of the way you refactored the boot image generation procedure,
> adding further bootloaders doesn't require any longer a big effort.

Great, we're of the same mind here.

--
Darren

> 
> Josef
> 
> On 28 November 2011 14:58, Darren Hart <dvhart at linux.intel.com
> <mailto:dvhart at linux.intel.com>> wrote:
> 
> 
> 
>     On 11/28/2011 01:06 AM, Josef Ahmad wrote:
>     > Hi Darren,
>     >
>     > I checked out your dvhart/efi branch and attempted to build my BSP.
>     > I inlcuded "efi" to my MACHINE_FEATURES, and "live" to my
>     IMAGE_FSTYPES,
>     > so as to get an .hddimg file.
>     >
>     > I haven't found GRUB in my deployed hddimg though. Maybe I'm missing
>     > something?
> 
> 
>     Hi Josef,
> 
>     Do you see bootia32.efi and grub.cfg in the root directory of the
>     hddimg? That is all it installs.
> 
>     --
>     Darren
> 
>     >
>     > Thanks
>     >
>     > Josef
>     >
>     > On 24 November 2011 16:01, Darren Hart <dvhart at linux.intel.com
>     <mailto:dvhart at linux.intel.com>
>     > <mailto:dvhart at linux.intel.com <mailto:dvhart at linux.intel.com>>>
>     wrote:
>     >
>     >     On 11/24/2011 12:48 AM, Josef Ahmad wrote:
>     >     > Hi Darren,
>     >     >
>     >     > I'll track your branch to sync our efforts.
>     >
>     >     Hi Josef,
>     >
>     >     Great. I just sent an RFC patch series to this list last
>     night. Please
>     >     try it out and see how it goes.
>     >
>     >     >
>     >     > I meant to only generate the Grub EFI binary for the target,
>     without
>     >     > including any tool/library into the root filesystem: I assumed
>     >     that it's
>     >     > sufficient to deploy the bootloader onto the top-level image.
>     >     That's why
>     >
>     >     Agreed.
>     >
>     >     > I build natively the mkimage tool and run that on the host
>     to get the
>     >     > EFI executable. Now, as you justly point out, there's a target
>     >     mismatch:
>     >     > mkimage incorrectly calls out the host architecture under the -O
>     >     switch,
>     >     > whereas it should have been something like ${TARGET_ARCH}-efi.
>     >
>     >     Right. That took some doing, but it is now resolved in my branch.
>     >
>     >     >
>     >     > That said, surely a solution that embodies the GRUB tools
>     into the
>     >     root
>     >     > filesystem is desirable.
>     >
>     >     That can be accomplished using the grub_1.99 recipe. Perhaps
>     we should
>     >     enable building with EFI if MACHINE_FEATURES includes "efi". I
>     consider
>     >     that to be separate from this effort.
>     >
>     >
>     >     > Feel free to merge my grub config generator into your branch, as
>     >     well as
>     >     > any of my contrib you may like. Speaking of grub.bbclass, I
>     named
>     >     it so
>     >     > as it's not specifically tied to the EFI mode. It basically
>     builds up
>     >     > the grub configuration menu, based off machine-specific
>     parameters. I
>     >     > implemented it among the lines of syslinux.bbclass.
>     >
>     >     While the menu creation isn't necessarily EFI specific, the
>     rest of the
>     >     bbclass is. If we wanted to, we could create another
>     grub.bbclass that
>     >     can do a live image with GRUB legacy booting. However, I'm
>     also working
>     >     to use the syslinux family of loaders wherever possible. I'm using
>     >     grub-efi as a stop-gap until an EFI syslinux becomes
>     available. The idea
>     >     here is to keep things consistent from the live image to the
>     installed
>     >     image, as well as reduce the number of bootloaders and
>     configuration
>     >     files we need deal with for a given BSP.
>     >
>     >     Please review the patch series I sent out and let me know if
>     it works
>     >     for your purposes. General code review is needed as well.
>     >
>     >     Thanks Josef!
>     >
>     >     --
>     >     Darren Hart
>     >     Intel Open Source Technology Center
>     >     Yocto Project - Linux Kernel
>     >     _______________________________________________
>     >     yocto mailing list
>     >     yocto at yoctoproject.org <mailto:yocto at yoctoproject.org>
>     <mailto:yocto at yoctoproject.org <mailto:yocto at yoctoproject.org>>
>     >     https://lists.yoctoproject.org/listinfo/yocto
>     >
>     >
> 
>     --
>     Darren Hart
>     Intel Open Source Technology Center
>     Yocto Project - Linux Kernel
> 
> 

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel