From: Arnout Vandecappelle <arnout@mind.be>
To: buildroot@busybox.net
Subject: [Buildroot] Compiling grub on x86_64
Date: Thu, 28 Jun 2012 10:09:57 +0200 [thread overview]
Message-ID: <4FEC1155.4020008@mind.be> (raw)
In-Reply-To: <20120625100458.4a3e62c8@skate>
On 06/25/12 10:04, Thomas Petazzoni wrote:
> Hello,
>
> Le Mon, 25 Jun 2012 09:51:52 +0200,
> Arnout Vandecappelle<arnout@mind.be> a ?crit :
>
> > I have an x86 project using and internal toolchain and grub as the boot loader. We recently switched to x86_64 (for
> > syscall performance reasons). Unfortunately, grub fails to even configure: the compiler can't generate executables.
> > It's missing -lgcc_s. I guess this is because the toolchain is built without multilib support and grub uses the -m32
> > switch.
> >
> > I'm unusually stumped at this point. Building a toolchain with multilib support seems the easy way out, but that's a
> > bit heavy just to be able to compile grub. A 64-bit build of grub will probably not work because BIOSes don't
> > understand 64-bit. So the best thing I can come up with is to have a separate buildroot config just to build grub.
> > Does anyone have better ideas?
>
> No, it's a difficult problem. And the issue came up several times on
> the list already. How much effort would it be to also generate a
> multilib 32bits/64bits toolchain when x86_64 is used?
It's certainly not trivial... Here's a list of issues I can come up with:
* It requires a different implementation for internal, crosstool-NG and
external toolchains.
* For external toolchains, there should be a check if it is multilib or not.
* For external toolchains, the wrapper has to change a lot because
the -m32 may imply a different -march and -mtune.
* crosstool-NG itself has no support for multilib AFAICS. Support would
have to be added.
* For internal toolchains, uClibc has to be built twice.
* Final gcc stage should be configured with the multilib paths.
* Are the multilib libraries installed to the target? For grub this isn't
really needed, but if something else compiles with -m32...
* Are other libraries compiled with multilib variants?
>
> > With this in mind, it's probably best to remove the BR2_x86_64 option from the grub config, or at least limit it to
> > external toolchains that may work (i.e. Sourcery).
>
> At least for now, yes, making sure that Grub can't be selected with
> internal toolchains on x86_64 seems like a good idea.
Turns out that it currently also doesn't work with multilib external toolchains.
The wrapper sets --sysroot to a path which doesn't contain the 32-bit libs.
For the time being, I guess grub should be disabled for all 64-bit configurations.
Regards,
Arnout
--
Arnout Vandecappelle arnout at mind be
Senior Embedded Software Architect +32-16-286540
Essensium/Mind http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint: 7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F
next prev parent reply other threads:[~2012-06-28 8:09 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-06-25 7:51 [Buildroot] Compiling grub on x86_64 Arnout Vandecappelle
2012-06-25 8:04 ` Thomas Petazzoni
2012-06-28 8:09 ` Arnout Vandecappelle [this message]
2012-06-28 9:46 ` Thomas Petazzoni
2012-06-28 9:56 ` Thomas Petazzoni
2012-06-28 12:44 ` Arnout Vandecappelle
2012-06-28 13:06 ` Thomas Petazzoni
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4FEC1155.4020008@mind.be \
--to=arnout@mind.be \
--cc=buildroot@busybox.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox