Hi Stefan,

On 14:42 Sat 08 Feb     , Stefan Wahren wrote:

Hi Andrea,

Am 07.02.25 um 22:31 schrieb Andrea della Porta:

quoted

The RP1 driver uses the infrastructure enabled by OF_OVERLAY config
option. Enable that option in defconfig in order to produce a kernel
usable on RaspberryPi5 avoiding to enable it separately.

Signed-off-by: Andrea della Porta <andrea.porta@suse.com>
---
This patch is *OPTIONAL* since I'm not sure if OF_OVERLAY is a desirable
feature to have enabled by default. It would be advisable to have it included
so that 'make defconfig' can produce a kernel config that will work out
of the box on Rpi5, otherwise OF_OVERLAY has to be enabled separately.

I think this isn't a good approach to convince the arm64 maintainer.
This change is not really optional for the Raspberry Pi 5 and possible
users/testers/CI rely on a working default configuration.

So my first suggestion would be to provide a scripts/bloat-o-meter
output (before/after). Based on this the maintainer can better decided.

I agree with you. IIUC what you are proposing, the following metrics should
help to decide. The defconfig kernel with CONFIG_OF_OVERLAY=y added (wrt to
the defconfig one without that set) has:

- same uncompressed kernel image size (due to ELF section alignment I guess)

- ~7Kb bigger Image.gz

- 3 new modules (all related to RP1, i.e.: clk-rp1, pinctrl-rp1 and rp1-pci)

- 27 added symbols, of which 5 exported
 
hoping this is enough to gather a rough idea of the impact.

In case this change is still rejected, we still have the option of
something like this [1]

I'm ready to go down this way should the previous numbers not be convincing
enough.

Many thanks,
Andrea

Best regards

[1] -
https://patchwork.kernel.org/project/linux-kbuild/patch/20200203184820.4433-3-nsaenzjulienne@suse.de/