[PATCH v2] CONFIG_NR_DRAM_BANKS: Remove unreferenced code as its always defined

[Pali Rohár]

Hello!

On Tuesday 11 August 2020 08:37:46 Stefan Roese wrote:
> Hi Tom,
> Hi Pali,
> 
> (added Pali because of the Nokia RX51 issue)

Could you please send me a link to "problematic" patch? As you have not
inlined it in this email.

> On 07.08.20 21:21, Tom Rini wrote:
> > On Thu, Jul 30, 2020 at 12:51:10PM +0200, Stefan Roese wrote:
> > 
> > > Since commit 86cf1c82850f ("configs: Migrate CONFIG_NR_DRAM_BANKS") &
> > > commit 999a772d9f24 ("Kconfig: Migrate CONFIG_NR_DRAM_BANKS"),
> > > CONFIG_NR_DRAM_BANKS is always defined with a value (4 is default).
> > > It makes no sense to still carry code that is guarded with
> > > "#ifndef CONFIG_NR_DRAM_BANKS" (and similar). This patch removes
> > > all these unreferenced code paths.
> > > 
> > > Also complete remove bi_memstart & bi_memsize from the board-info
> > > struct. As now bi_dram[] is always enabled and should be used instead.
> > > This removes the redundant varriables.
> > > 
> > > Signed-off-by: Stefan Roese <sr at denx.de>
> > > Cc: Daniel Schwierzeck <daniel.schwierzeck at gmail.com>
> > > Cc: Tom Rini <trini at konsulko.com>
> > > Cc: Ramon Fried <ramon.fried at gmail.com>
> > > Cc: Simon Glass <sjg at chromium.org>
> > > Cc: Michal Simek <michal.simek at xilinx.com>
> > > Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck at gmail.com>
> > 
> > I don't see quite how, but this is breaking running
> > test/nokia_rx51_test.sh (or, my fixup of this to apply to current master
> > was wrong, and is what's breaking the test).  Please rebase and confirm
> > that test passes as well, thanks!
> 
> I've checked the issue with nokia_rx51_test.sh and could not find any
> issues in the patch. My assumption now is, that the very old Linux
> kernel (2.6.28) that is used in this Nokia test, still uses the bd_info
> struct in Linux to pass the memory information (via bd_memstart &
> bi_memsize), as was also done in the very old PowerPC days. With this
> patch now and the removal of these fields from bd_info, this might
> explain why this kernel does not boot any more (no output on the serial
> console at all).
> 
> Pali, could you please check if my assumption is correct here? And if
> yes, could please switch the test to using a newer kernel version? Or
> remove the Linux kernel booting from the test?

Yes, Nokia N900 uses "old" ATAGs. But we cannot remove it. New kernel
version does not contain Maemo patches which are required for Maemo
system which is still widely used. And yes, people are using it with
U-Boot.

Second reason why we cannot remove support for ATAGs is that Nokia's
signed first stage bootloader pass other setup data via ATAGs for kernel
and U-Boot N900 board code parses it, reuse it and pass to kernel.

And replacing first stage bootloader is not possible because it is
signed and signing keys are secret (now probably lost).

> I've pushed the latest patch version into this branch (based on top of
> the latest TOT):
> 
> https://gitlab.denx.de/u-boot/custodians/u-boot-marvell/-/commits/remove-config-nr-dram-banks-v3-2020-08-11
> 
> The failing Azure CI report can be found here:
> 
> https://dev.azure.com/sr0718/u-boot/_build/results?buildId=20&view=results
> 
> A working Azure report is here for comparison:
> 
> https://dev.azure.com/sr0718/u-boot/_build/results?buildId=18&view=results
> 
> Thanks,
> Stefan