RFC: Inclusion of bootloader utilities in task-base

RFC: Inclusion of bootloader utilities in task-base

[Rod Whitby]

Some MACHINE configurations have FEATURES defined for 'uboot' and
'redboot'.  It looks like these are intended to identify the bootloader
used on that machine.

These turn on additions to RDEPENDS in task-base for some bootloader
utilities.

RFC #1: Is putting bootloader utilities in task-base a good idea in the
first place?  Surely it's something that you either need for first boot
(in which case it should be in the more primitive task-boot list), or
it's something you use rarely after first boot (in which case it
shouldn't be  wasting precious flash space in a task-base list but
should just be ipkg installed when needed).

RFC #2: Shouldn't these be RRECOMMENDS unless they are absolutely
critical, so that you can ipkg remove them if you don't want them.

RFC #3: I want to add an 'apex' FEATURE for the Apex bootloader, and
have that enable the 'apex-env' utility.  Any objections?  It would be
done in the way decided by RFC #2, and subject to an RFC #1 decision.

RFC #4: If a machine (like the ixp4xx) can use a number of different
bootloaders (redboot and apex in the case of the NSLU2), should they all
be listed as features in machine.conf?

-- Rod

Re: RFC: Inclusion of bootloader utilities in task-base

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Rod Whitby schreef:
> Some MACHINE configurations have FEATURES defined for 'uboot' and
> 'redboot'.  It looks like these are intended to identify the bootloader
> used on that machine.
> 
> These turn on additions to RDEPENDS in task-base for some bootloader
> utilities.
> 
> RFC #1: Is putting bootloader utilities in task-base a good idea in the
> first place?  Surely it's something that you either need for first boot
> (in which case it should be in the more primitive task-boot list), or
> it's something you use rarely after first boot (in which case it
> shouldn't be  wasting precious flash space in a task-base list but
> should just be ipkg installed when needed).
> 
> RFC #2: Shouldn't these be RRECOMMENDS unless they are absolutely
> critical, so that you can ipkg remove them if you don't want them.

I'm not familiar with the finer points of bootloaders, so I'm not really
qualified to respond to this, but if the utils are _critical_, they
should be RDEPENDS in task-boot, not task base.

regards,

Koen


> RFC #3: I want to add an 'apex' FEATURE for the Apex bootloader, and
> have that enable the 'apex-env' utility.  Any objections?  It would be
> done in the way decided by RFC #2, and subject to an RFC #1 decision.
> 
> RFC #4: If a machine (like the ixp4xx) can use a number of different
> bootloaders (redboot and apex in the case of the NSLU2), should they all
> be listed as features in machine.conf?
> 
> -- Rod
> 
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
> 


- --
koen@dominion.kabel.utwente.nl will go go away in december 2007, please
use k.kooi@student.utwente.nl instead.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFHUmqDMkyGM64RGpERApTWAJ0Tv04Q+IepkAvT3ciF0TZRMKaPQwCbB1r4
b/wlDDn2RfoPM/di65PDBFM=
=H8Wb
-----END PGP SIGNATURE-----

Re: RFC: Inclusion of bootloader utilities in task-base

[Rod Whitby]

Koen Kooi wrote:
> Rod Whitby schreef:
>> Some MACHINE configurations have FEATURES defined for 'uboot' and
>> 'redboot'.  It looks like these are intended to identify the bootloader
>> used on that machine.
> 
>> These turn on additions to RDEPENDS in task-base for some bootloader
>> utilities.
> 
>> RFC #2: Shouldn't these be RRECOMMENDS unless they are absolutely
>> critical, so that you can ipkg remove them if you don't want them.
> 
> I'm not familiar with the finer points of bootloaders, so I'm not really
> qualified to respond to this, but if the utils are _critical_, they
> should be RDEPENDS in task-boot, not task base.

Agreed.  In the case of redboot and apex, they are not that critical.

The redboot util 'fis' is used to reflash fis directory partitions.

The apex util 'apex-env' is used to change the kernel cmdline from userspace.

Dunno about the uboot util package.

-- Rod

Re: RFC: Inclusion of bootloader utilities in task-base

[Paul Sokolovsky]

Hello Rod,

Sunday, December 2, 2007, 12:06:52 PM, you wrote:

> Koen Kooi wrote:
>> Rod Whitby schreef:
>>> Some MACHINE configurations have FEATURES defined for 'uboot' and
>>> 'redboot'.  It looks like these are intended to identify the bootloader
>>> used on that machine.
>> 
>>> These turn on additions to RDEPENDS in task-base for some bootloader
>>> utilities.
>> 
>>> RFC #2: Shouldn't these be RRECOMMENDS unless they are absolutely
>>> critical, so that you can ipkg remove them if you don't want them.
>> 
>> I'm not familiar with the finer points of bootloaders, so I'm not really
>> qualified to respond to this, but if the utils are _critical_, they
>> should be RDEPENDS in task-boot, not task base.

> Agreed.  In the case of redboot and apex, they are not that critical.

> The redboot util 'fis' is used to reflash fis directory partitions.

> The apex util 'apex-env' is used to change the kernel cmdline from userspace.

  This raises the question if user should be allowed (far less
required) to change bootloader's standard cmdline as *the normal
course of using distro* (== task-base). So, maybe the 1st choice ("Is
it good idea at all") is worth more attention.

> Dunno about the uboot util package.

> -- Rod


-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com

Re: RFC: Inclusion of bootloader utilities in task-base

[Rod Whitby]

Paul Sokolovsky wrote:
>   This raises the question if user should be allowed (far less
> required) to change bootloader's standard cmdline as *the normal
> course of using distro* (== task-base). So, maybe the 1st choice ("Is
> it good idea at all") is worth more attention.

Ah, this is a question about which parts of task-base a DISTRO should
include or not (by use of it's DISTRO_FEATURES).  Not about whether
task-base should give distros the choice in the first place.

Is Angstrom (the only DISTRO which currently has this feature enabled)
meant to be a hand-holding-for-naive-users distro, or a power-users distro?

And yes, it's most likely that an end-user distro would put a shell script
wrapper around the bootloader utility to safeguard the user for normal
end-user operations like changing the rootfs load location, or setting up
nfsroot operation.

-- Rod

Re: RFC: Inclusion of bootloader utilities in task-base

[Paul Sokolovsky]

Hello Rod,

Sunday, December 2, 2007, 11:25:12 PM, you wrote:

> Paul Sokolovsky wrote:
>>   This raises the question if user should be allowed (far less
>> required) to change bootloader's standard cmdline as *the normal
>> course of using distro* (== task-base). So, maybe the 1st choice ("Is
>> it good idea at all") is worth more attention.

> Ah, this is a question about which parts of task-base a DISTRO should
> include or not (by use of it's DISTRO_FEATURES).  Not about whether
> task-base should give distros the choice in the first place.

  Ok.

> Is Angstrom (the only DISTRO which currently has this feature enabled)
> meant to be a hand-holding-for-naive-users distro, or a power-users distro?

  Apparently, leaning more towards the first, than to the second. On my
memory, at least some MTD-related utility was cut from Angstrom for
the reasons that users may shoot themselves in foot with it. Other
util (ipkg-link) is not incuded for purely idiosyncratic reasons, bit
with the same motivation - "users may mess up with it".

  But well, I'm not arguing if you should include the utils in
question or not, but rather on more general question what should go
into standard image (or images of different types), and what not.
Then, it would be easier to decide on particulars, and assess, if some
exception is worth here or there. Motivation is that it's not possible
to fit any useful util at all, and space is precious.

> And yes, it's most likely that an end-user distro would put a shell script
> wrapper around the bootloader utility to safeguard the user for normal
> end-user operations like changing the rootfs load location, or setting up
> nfsroot operation.

> -- Rod


-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com

Re: RFC: Inclusion of bootloader utilities in task-base

[Rod Whitby]

Paul Sokolovsky wrote:
>   But well, I'm not arguing if you should include the utils in
> question or not, but rather on more general question what should go
> into standard image (or images of different types), and what not.

That's my goal too.  A couple of sentences at the top of each image file
giving it's intent would go a long way to answering these questions.

For example, I'm writing a nas-server-image.bb file right now, and I've
put at the top:

# This image is intended to provide a configuration
# for a network attached storage server device.
#
# Key features are:
#
# 1) Must be able to mount attached storage devices like
# internal disks, external USB disks, etc.  Should support
# various filesystem choices, but ext2 at a minimum.
#
# 2) Must be able to boot from internal flash, or directly
# from a filesystem stored on the attached storage.
#
# 3) Must be able to share that storage across the network
# using various protocols like Samba, NFS, etc.
#
# 4) Bonus points for being able to support other central
# network services like a central gateway machine might.
#
# Although it is tested with the Angstrom distro, it is
# intended to be distro-agnostic.

-- Rod