Hi Tom,
On 01.10.21 14:59, Tom Rini wrote:
On Fri, Oct 01, 2021 at 01:01:37PM +0200, Stefan Roese wrote:
Added to Tom to Cc
On 01.10.21 12:43, Stefan Roese wrote:
On 01.10.21 12:39, Pali Rohár wrote:
Hello!
On Friday 01 October 2021 12:28:34 Stefan Roese wrote:
First the test on "next" without the 39 patches:
[stefan@ryzen u-boot (next)]$ ./tools/kwboot -p -b u-boot-spl.kwb -t
...
DDR3 Training Sequence - Ended Successfully
Trying to boot from BOOTROM
Returning to BootROM (return address 0xffff0aa0)...
+xmodem: Protocol error
Not working!
And now test on "master" without the 39 patches:
[stefan@ryzen u-boot (master)]$ ./tools/kwboot -p -b u-boot-spl.kwb -t
...
Seems to okay.
Thanks,
Stefan
So something is broken in next branch. I have looked at kwboot and
kwbimage code and there is no change in next branch which is not in
master branch.
So it looks like that something not related to kwboot broke it in next
branch. And seems that git bisect on AXP hw would be required. But I do
not own AXP board, so I cannot do it.
I can try to do this here later, if I find the time.
git bisect leads to this problematic commit:
[stefan@ryzen u-boot ((ab92b38a0161...)|BISECTING)]$ git bisect good
a2ac2b964bfbb20d6791ee94b9034a50cfadb5b0 is the first bad commit
commit a2ac2b964bfbb20d6791ee94b9034a50cfadb5b0
Author: Tom Rini <tr...@konsulko.com>
Date: Fri Aug 27 21:18:30 2021 -0400
Convert CONFIG_SKIP_LOWLEVEL_INIT et al to Kconfig
This converts the following to Kconfig:
CONFIG_SKIP_LOWLEVEL_INIT
CONFIG_SKIP_LOWLEVEL_INIT_ONLY
In order to do this, we need to introduce SPL and TPL variants of these
options so that we can clearly disable these options only in SPL in some
cases, and both instances in other cases.
Signed-off-by: Tom Rini <tr...@konsulko.com>
Frankly, I did not analyse this change fully. But it seems to
responsible for this breakage on Armada XP in "next".
OK, I guess the conversion from
arch/arm/mach-mvebu/include/mach/config.h didn't work right. These
boards should have set CONFIG_SPL_SKIP_LOWLEVEL_INIT but don't.
Yes, this seems to be the case here.
Why it was missed, well, I can't spot the magic to include
<mach/config.h> if it exists, but I just missed that, so I'll need to
make sure to take more care on the tricky conversions that also use
mach/config.h. Let me know if you need me to re-convert platforms here,
or now that you know what's missing you can fix it, thanks!
I'll send a fix for "next" to the list shortly.
Thanks,
Stefan
