Re: [Buildroot] [PATCH] configs/roc_pc_rk3399_defconfig: expand rootfs size to 120M

From: Arnout Vandecappelle <arnout@mind.be>
To: Giulio Benetti <giulio.benetti@benettiengineering.com>,
	buildroot@buildroot.org, Peter Korsgaard <peter@korsgaard.com>,
	"Yann E. MORIN" <yann.morin.1998@free.fr>
Cc: Michael Nazzareno Trimarchi <michael@amarulasolutions.com>,
	Danilo Bargen <mail@dbrgn.ch>,
	Suniel Mahesh <sunil@amarulasolutions.com>
Subject: Re: [Buildroot] [PATCH] configs/roc_pc_rk3399_defconfig: expand rootfs size to 120M
Date: Mon, 14 Feb 2022 21:13:33 +0100	[thread overview]
Message-ID: <6edcbc6d-c9bb-e802-48f1-d4866a698aa9@mind.be> (raw)
In-Reply-To: <cb93cb6a-c8ae-0975-6ae0-14ad369de9e2@benettiengineering.com>

On 14/02/2022 01:03, Giulio Benetti wrote:
> Hi Arnout,
> 
> On 13/02/22 12:35, Arnout Vandecappelle wrote:
>>
>>
>> On 02/02/2022 18:10, Giulio Benetti wrote:
>>> Default 60M for rootfs are not enough and generate a build failure, so
>>> let's expand rootfs size to 120M.
>>
>>    Did you find out _why_ 64M wasn't enough for just the kernel and busybox? Is
>> it because of all the stupid modules in the default arm64 defconfig? But most
>> other arm64 defconfigs don't seem to need this...
> 
> It is because of modules(30M) and linux "Image", it's not compressed, so it's 
> 25M. So only for Linux+modules => 55M, plus the few other stuff(Busybox+libc 
> etc.)1M

  Yeah those giant aarch64 default configs are mightily annoying...

  I'm thinking, perhaps we should just bite the bullet and move the default ext2 
size to 120M...

  Peter, Yann, what do you think?

  Regards,
  Arnout

> 
> Also, nothing changed since the beginning, so I don't know how could this happened:
> # git log --oneline configs/roc_pc_rk3399_defconfig
> f51cbc33d9 configs/roc_pc_rk3399: fix U-Boot dependencies
> 80ae1aedcb configs/roc_rk3399_pc: new defconfig
> # git log --oneline roc-rk3399-pc/
> 80ae1aedcb configs/roc_rk3399_pc: new defconfig
> 
> U-Boot dependencies are host dependencies, so it doesn't impact.
> 
> # git log --oneline fs/ext2:
> 5ece6be60b fs/ext2: fir (again) namespace for variables
> db7d786140 fs/ext2: fix namespace for variables
> 5631c38024 fs/ext2: pass a default string to filesystem label
> ...
> here board has been added
> 
> So it doesn't seem a ext2 problem too.
> I can't understand how it could be building in the beginning.
> 
> We have 2 solutions:
> - expand rootfs
> - use a compressed linux Image.gz(25M => 9M), but need target testing
> 
> Honestly I don't think systems with SoCs like RK3399 have problems with 56M 
> more, so I would go with expanding rootfs to 120M and if maintainer in Cc wants 
> to shrink Linux and target testing it is welcome to send a patch.
> 
> What do you think?
> 
> Best regards
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot