Re: [PATCH 2/3] rockchip: Disable DISTRO_DEFAULTS for rockpro64

[Tom Rini <trini@konsulko.com>]

[-- Attachment #1: Type: text/plain, Size: 3500 bytes --]

On Wed, Feb 22, 2023 at 12:03:58PM -0700, Simon Glass wrote:
> Hi Tom,
> 
> On Wed, 22 Feb 2023 at 12:01, Tom Rini <trini@konsulko.com> wrote:
> >
> > On Wed, Feb 22, 2023 at 11:58:37AM -0700, Simon Glass wrote:
> > > Hi,
> > >
> > > On Wed, 22 Feb 2023 at 11:56, Jonas Karlman <jonas@kwiboo.se> wrote:
> > > >
> > > > On 2023-02-22 19:24, Tom Rini wrote:
> > > > > On Wed, Feb 22, 2023 at 10:19:03AM -0800, Vagrant Cascadian wrote:
> > > > >> On 2023-02-21, Vagrant Cascadian wrote:
> > > > >>> On 2023-02-21, Simon Glass wrote:
> > > > >>>> This board has moved to standard boot but the old 'distro_bootcmd'
> > > > >>>> command is still active. Disable DISTRO_DEFAULTS to fix this.
> > > > >>>
> > > > >>> Works for booting rockpro64-rk3399, thanks!
> > > > >>
> > > > >> I can also confirm that applying a very similar patch for
> > > > >> pinebook-pro-rk3399 works booting with bootstd.
> > > > >>
> > > > >> Seems worth adding if there is a v2 of the patch series.
> > > > >>
> > > > >> Alternately, rather than doing this on a board-by-board basis, is there
> > > > >> some way to disable CONFIG_DISTRO_DEFAULTS when a board is using
> > > > >> BOOTSTD?
> > > > >
> > > > > I think that's possibly a bit dangerous? DISTRO_DEFAULTS and
> > > > > BOOTSTD_DEFAULTS need to be reconciled to a single new symbol, and the
> > > > > DISTRO_DEFAULTS method of select not imply for most things is also I
> > > > > suspect the right path.
> > > > >
> > > >
> > > > Another option would be to set BOOTSTD_BOOTCOMMAND=y, this is the main
> > > > issue with rk3399 at the moment, distro_bootcmd got removed from env,
> > > > yet bootcmd is still "run distro_bootcmd" due to DISTRO_DEFAULTS=y.
> > > >
> > > >
> > > > config BOOTCOMMAND
> > > >         default "bootflow scan -lb" if BOOTSTD_BOOTCOMMAND && CMD_BOOTFLOW_FULL
> > > >         default "bootflow scan" if BOOTSTD_BOOTCOMMAND && !CMD_BOOTFLOW_FULL
> > > >         default "run distro_bootcmd" if !BOOTSTD_BOOTCOMMAND && DISTRO_DEFAULTS
> > > >
> > > > config BOOTSTD_BOOTCOMMAND
> > > >         default y if !DISTRO_DEFAULTS
> > >
> > > That seems reasonable to me, along with a 'depends on BOOTSTD'
> > >
> > > At present DISTRO_DEFAULTS is both enabled in defconfig files (about
> > > 350) and Kconfig (another 300). For rockchip it is the latter.
> > >
> > > It doesn't really make sense to use DISTRO_DEFAULTS when using
> > > BOOTSTD_DEFAULTS. One is 'run distro_bootcmd' and the other is
> > > 'bootflow scan', possibly with -lb flags. But we really don't want
> > > people flipping back and forward.
> >
> > It doesn't make linguistic sense to have "distro defaults" and "bootstd
> > defaults" enabled.  It makes compute sense to have these two symbols
> > combined as they're nearly identical and have a similar conceptual need
> > (make sure we have things like bootz/booti/ext4/etc support).
> 
> Oh, so you are talking about DISTRO_DEFAULTS, not the actual command?
> In that case I can do a patch to make them both use a common symbol to
> bring in most stuff, although bootstd does not need hush and a few
> other things that are needed by distro_bootcmd.
> 
> Let me know if that makes sense.

Yeah, the Kconfig symbols DISTRO_DEFAULTS and BOOTSTD_DEFAULTS need to
be reconciled and combined. This is separate from setting bootcmd to
"run distro_bootcmd", which should be separate, and yes, not be the
chosen symbol when we have bootstd enabled.

-- 
Tom

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 659 bytes --]