On Wed, Sep 10, 2025 at 03:37:35PM +0000, Maarten Brock wrote: > > From: Tom Rini > > > > On Tue, Sep 09, 2025 at 04:23:56PM +0200, Kory Maincent wrote: > > > On Tue, 9 Sep 2025 12:45:10 +0000 > > > Maarten Brock <maarten.br...@sttls.nl> wrote: > > > > > > > > From: Kory Maincent <kory.mainc...@bootlin.com> > > > > > > > > > > On Mon, 8 Sep 2025 16:05:04 +0000 > > > > > Maarten Brock <maarten.br...@sttls.nl> wrote: > > > > > > > > > > > When PMIC drivers are disabled their functions should not be called. > > > > > > > > > > > > Signed-off-by: Maarten Brock <maarten.br...@sttls.nl> > > > > > > > > > When a board is using a certain PMIC it should be possible to deselect > > > > all other PMICs that are not present. There is no point in keeping > > > > unused > > > > code in the binary. Isn't that the whole idea of the configuration? > > > > The current situation gives a build failure when the unused PMIC is > > > > deselected. > > > > > > > > This file board/ti/am335x/board.c gives the impression to be useable for > > > > all AM335X based boards, including those yet unknown. What if a board > > > > does not need any PMIC configuration, because the default strapped > > > > config > > > > is enough, does one still need to drag those drivers along? > > > > > > Indeed, I hadn't understood that you would face a build error as the PMIC > > > functions are not defined if we disable PMIC configs. > > > > > > I think the first issue is the design of this board.c file, but, well it > > > is > > > U-boot typical design. > > > And scale_vcores_generic() is not a generic function at all ... > > > > > > Ok, your change are legit after all. > > > > A thing to keep in mind is that am335x_evm is intended for the TI EVM > > family of boards and look-alikes (so all of the beagles which are > > kinda-sorta TI EVMs depending on when you asked someone). > > TI and U-boot give the impression that this file could be used as a template > for any AM335x based board. That it produces build failures as soon as you > modify some configuration contradicts that IMHO. > > The same goes for my recent patch about NO_NET and not providing > ft_board_setup(). That also results in an unexpected build failure.
Yes, there is more "delete this code too" required here than on modern TI chips where the goal of having easier to customize evm support was (closer?) to being realized. -- Tom
signature.asc
Description: PGP signature
