State of ARM port

State of ARM port

[Andrei Borzenkov]

It seems that distributions tend to focus on native U-Boot support for
extlinux-like configuration and direct loading of Linux kernel/FDT.
Recently it came up for openSUSE, reasons were named

1. upstream U-Boot prefers extlinux for loading, CONFIG_API is
considered edge case, deprecated. They do not consider GRUB valid reason
to maintain it :)

2. GRUB requires patching for each board to set valid link address

3. some general issues on specific boards

1 requires active commitment from U-Boot community, apparently it is
lacking. To properly fix 2 we need relocation support in U-Boot; which
again returns us to "GRUB not being interesting to U-Boot community" :)

3 depends on motivation to debug and fix issues; as long as GRUB is not
considered there is none.

So what should we do with this port?

Re: State of ARM port

[Leif Lindholm]

Hi Andrei,

Sorry, have been off on holiday.

On Sun, Nov 29, 2015 at 10:43:54AM +0300, Andrei Borzenkov wrote:
> It seems that distributions tend to focus on native U-Boot support for
> extlinux-like configuration and direct loading of Linux kernel/FDT.
> Recently it came up for openSUSE, reasons were named

Debian don't, and I'm not sure I've heard of any others moving that
direction, so I don't know if it can be called a trend.

> 1. upstream U-Boot prefers extlinux for loading, CONFIG_API is
> considered edge case, deprecated. They do not consider GRUB valid reason
> to maintain it :)

Do you have any references to this?
(Not questioning it, sounds likely, but would like to read through
any existing discussions.)

> 2. GRUB requires patching for each board to set valid link address

Yeah :(

> 3. some general issues on specific boards

Hmm?

> 1 requires active commitment from U-Boot community, apparently it is
> lacking. To properly fix 2 we need relocation support in U-Boot; which
> again returns us to "GRUB not being interesting to U-Boot community" :)

No, 2 would be totally fixable in GRUB. It would just take someone
actually making the grub kernel position independent. And I haven't
managed to find the time in the past two years...

> 3 depends on motivation to debug and fix issues; as long as GRUB is not
> considered there is none.
> 
> So what should we do with this port?

Unsure.

I have had people start asking me for the arm64 variant, but if the
U-Boot community is actively disintirested in providing an API to hook
into, this may not make sense.

Regards,

Leif

Re: State of ARM port

[Andrei Borzenkov]

07.12.2015 20:02, Leif Lindholm пишет:
> Hi Andrei,
> 
> Sorry, have been off on holiday.
> 
> On Sun, Nov 29, 2015 at 10:43:54AM +0300, Andrei Borzenkov wrote:
>> It seems that distributions tend to focus on native U-Boot support for
>> extlinux-like configuration and direct loading of Linux kernel/FDT.
>> Recently it came up for openSUSE, reasons were named
> 
> Debian don't, and I'm not sure I've heard of any others moving that
> direction, so I don't know if it can be called a trend.
> 
>> 1. upstream U-Boot prefers extlinux for loading, CONFIG_API is
>> considered edge case, deprecated. They do not consider GRUB valid reason
>> to maintain it :)
> 
> Do you have any references to this?
> (Not questioning it, sounds likely, but would like to read through
> any existing discussions.)
> 

https://www.marc.info/?l=u-boot&m=140881607132099&w=2
http://lists.denx.de/pipermail/u-boot/2015-October/229915.html (full thread)

https://github.com/openSUSE/perl-bootloader/pull/81#issuecomment-158543744

>> 2. GRUB requires patching for each board to set valid link address
> 
> Yeah :(
> 
>> 3. some general issues on specific boards
> 
> Hmm?
> 
>> 1 requires active commitment from U-Boot community, apparently it is
>> lacking. To properly fix 2 we need relocation support in U-Boot; which
>> again returns us to "GRUB not being interesting to U-Boot community" :)
> 
> No, 2 would be totally fixable in GRUB. It would just take someone
> actually making the grub kernel position independent. And I haven't
> managed to find the time in the past two years...
> 
>> 3 depends on motivation to debug and fix issues; as long as GRUB is not
>> considered there is none.
>>
>> So what should we do with this port?
> 
> Unsure.
> 
> I have had people start asking me for the arm64 variant, but if the

arm64 EFI grub port is used as standard in openSUSE. Only U-Boot
platforms are moving away.

Or do you mean arm64 U-Boot?

> U-Boot community is actively disintirested in providing an API to hook
> into, this may not make sense.
> 
> Regards,
> 
> Leif
> 

Re: State of ARM port

[Leif Lindholm]

On Mon, Dec 07, 2015 at 08:22:59PM +0300, Andrei Borzenkov wrote:
> >> 1. upstream U-Boot prefers extlinux for loading, CONFIG_API is
> >> considered edge case, deprecated. They do not consider GRUB valid reason
> >> to maintain it :)
> > 
> > Do you have any references to this?
> > (Not questioning it, sounds likely, but would like to read through
> > any existing discussions.)
> 
> https://www.marc.info/?l=u-boot&m=140881607132099&w=2
> http://lists.denx.de/pipermail/u-boot/2015-October/229915.html (full thread)

It stops at Rafal pointing out it's needed for FreeBSD though.

But certainly there seems to be an ideological clash here.

> https://github.com/openSUSE/perl-bootloader/pull/81#issuecomment-158543744

Hmm. They haven't reported any issues here, or to the Linaro cross-distro
mailing list.

> >> 3 depends on motivation to debug and fix issues; as long as GRUB is not
> >> considered there is none.
> >>
> >> So what should we do with this port?
> > 
> > Unsure.
> > 
> > I have had people start asking me for the arm64 variant, but if the
> 
> arm64 EFI grub port is used as standard in openSUSE. Only U-Boot
> platforms are moving away.
> 
> Or do you mean arm64 U-Boot?

Yeah, sorry, I meant arm64 U-Boot.

/
    Leif