On Fri, Apr 24, 2020 at 02:04:57PM +0200, Anatolij Gustschin wrote:
Hi Dario,
On Thu, 23 Apr 2020 19:27:23 +0200 (CEST) Dario dario86@tutamail.com wrote:
Hello,
On 2017/02/07 commit af2cbfd6b982f7a52414ea50b2ece0290d4b748f broke gigabit ethernet for Olimex A20-OLinuXino-LIME2. This commit likely prevents the activation of the option CONFIG_RTL8211X_PHY_FORCE_MASTER needed for proper gigabit networking on this device. Since the only modification introduced by this commit was moving Kconfig files, I suspect this change may affect many more devices as well for which proper build options may not be enabled.
Thanks for the regression report. We do appreciate these whenever they come in.
Apart from how this went unnoticed for such a long time, what really puzzles me is commit 8728c97eff5bd95f58320f886ae305f17931a374 submitted on 2017/03/13 which removes CONFIG_RTL8211X_PHY_FORCE_MASTER from configs/A20-OLinuXino-Lime2_defconfig altogether.
since the first mentioned commit CONFIG_RTL8211X_PHY_FORCE_MASTER depends on CONFIG_PHY_REALTEK, but this was not enabled in board defconfig. This is probably the reason why it was dropped in 8728c97eff5b.
Right, it got removed because it couldn't be selected anymore. This means that the first commit didn't move everything right.
I think it is not by chance that Arch Linux ARM kept the package uboot-a20-olinuxino-lime2 at version 2017.01-2 which is the last working version before the aforementioned commit broke gigabit ethernet for this board.
My A20-OLinuXino-Lime2 is the only device I have available for testing. My fear is that many more devices may be affected since this same option was removed from A20-Olimex-SOM-EVB_defconfig as well along with several other options from many other devices.
Please test patch [1]. If it helps, we can see how to fix other boards. When reporting, please also Cc board maintainers (see board/sunxi/MAINTAINERS). Thanks!
[1] http://patchwork.ozlabs.org/project/uboot/patch/20200424115411.11102-1-agust...
Please let us know if that works but I'm going to take a look at what else might be broken by the first commit and the patch above implies we need to migrate that symbol in a few more cases.