[PATCH v2 01/22] apalis-imx8: fix booting caused by missing dm_event

[Francesco Dolcini]

Hello Tom, 

On Sat, Jan 14, 2023 at 01:23:11PM -0500, Tom Rini wrote:
> On Sat, Jan 14, 2023 at 07:00:08PM +0100, Heinrich Schuchardt wrote:
> > On 1/14/23 18:22, Tom Rini wrote:
> > > On Sat, Jan 14, 2023 at 11:57:20AM -0500, Tom Rini wrote:
> > > > On Sat, Jan 14, 2023 at 08:27:25AM -0500, Tom Rini wrote:
> > > > > On Sat, Jan 14, 2023 at 08:25:19AM -0300, Fabio Estevam wrote:
> > > > > > Hi Francesco,
> > > > > > 
> > > > > > On Fri, Jan 13, 2023 at 2:18 PM Francesco Dolcini <francesco at dolcini.it> wrote:
> > > > > > > 
> > > > > > > From: Marcel Ziswiler <marcel.ziswiler at toradex.com>
> > > > > > > 
> > > > > > > Without the DM_EVENT absolutely no output whatsoever and the system does
> > > > > > > not boot at all.
> > > > > > 
> > > > > > Why do we need to select CONFIG_EVENT=y in the first place?
> > > > > > 
> > > > > > Oliver on CC reported the same issue for his board, and we have other
> > > > > > broken boards too
> > > > > > that do not select CONFIG_EVENT=y.
> > > > > > 
> > > > > > Can we have a generic fix for this?
> > > > > > 
> > > > > > Tom, Heinrich?
> > > > > 
> > > > > Yes, if someone can explain what the problem is, we can correct the
> > > > > dependency.  I gather most platforms have whatever this problem is
> > > > > ignored due to enabling EFI_LOADER.
> > > > 
> > > > The more I look at this, the more confused I get. With DEBUG_UART
> > > > enabled, where do thing fail without CONFIG_EVENT (and CONFIG_DM_EVENT,
> > > > which is then enabled due to "imply DM_EVENT" in arch/arm/Kconfig, under
> > > > ARCH_IMX8) ? How about if you turn DM_EVENT back off?
> > > 
> > > The only other data point I can add right now is that it's not entirely
> > > generic, I think. I can disable EFI_LOADER and EVENT and DM_EVENT on
> > > am65x_evm (both r5 and a53 portions) and still boot up fine. I don't
> > > have something imx8 in my lab, however.
> > > 
> > 
> > I cannot find event_register() used outside of
> > lib/efi_driver/efi_block_device.c and test/common/event.c)
> > 
> > Why should any board with CONFIG_EFI_LOADER=n depend on events?
> 
> There's other events, using EVENT_SPY(...).  I'm working on a patch to
> fix things up now.

Given that this is now solved by your patch Tom, we can just drop this,
together with patch 10. The question here is if I should send another
series with those 2 removed or you just do not apply those.

Francesco