[fedora-arm] u-boot plans for Fedora 22 ?

[Hans de Goede]

Hi,

On 03/21/2015 02:49 PM, Dennis Gilmore wrote:
> On Saturday, March 21, 2015 03:40:20 PM Hans de Goede wrote:
>> Hi,
>>
>> On 21-03-15 15:21, Hans de Goede wrote:
>>> Hi,
>>>
>>> On 06-03-15 23:50, Peter Robinson wrote:
>>>>> I assume that we will be rebasing u-boot to the just released v2015.01
>>>>> for F-22? But I was wondering if there is any chance we can jump to
>>>>> v2015.04 ? The reason I'm asking is that things are progressing
>>>>> quite rapidly on the u-boot side, at least with Allwinner SoC support,
>>>>> I've just send a pull-req for v2015.04 with the following highlights:
>>>>>
>>>>> 1) Improved sun6i (A31) support, including support for the A31s variant
>>>>> and
>>>>>
>>>>>      automatic assignment of a SoC serial based MAC address for ethernet
>>>>>
>>>>> 2) Full sun8i (A23) support including DRAM controller init and SPL, so
>>>>> now
>>>>>
>>>>>      people can boot these boards using a full FOSS solution
>>>>>
>>>>> 3) Many improvement to the graphical console support, automatic
>>>>> selection
>>>>>
>>>>>      of the native mode for HDMI/DVI monitors via DDC + EDID, LCD panel
>>>>>
>>>>> support,
>>>>>
>>>>>      VGA output support
>>>>>
>>>>> 4) Preparation work for OTG controller support, together with 3) this
>>>>> allows using
>>>>>
>>>>>      u-boot on tablets effortlessly. The rest of the OTG support is going
>>>>>
>>>>> upstream
>>>>>
>>>>>      through the usb tree
>>>>>
>>>>> And if possible I would like to see this end up in Fedora 22 :)
>>>> An initial build of 2015.04rc3 will be in rawhide tomorrow, so please
>>>> test, I've enabled a few extra new devices and I'll be reviewing the
>>>> rest of the new devices over the next week or so.
>>>>
>>>> Please test, once it's settled down and we're fairly certain all the
>>>> usual suspects haven't regressed we'll move it into F-22 before beta
>>>> starts to get locked down.
>>> I noticed that -rc4 has also been build, so I've given that a test-run
>>> on sunxi instead:
>>> http://koji.fedoraproject.org/koji/buildinfo?buildID=622061
>>>
>>>   From a sunxi pov this build looks good.
>> I have to take this back, it seems that this build is appending:
>>
>> " console=ttyS0,115200"
>>
>> To the kernel cmdline, I've just double-checked and this is not upstream
>> behavior, is this being done by Fedora specific patches ?
>>
>> This breaks kernel output and systemd status messages output on
>> tty0 / the hdmi output, resulting in a blackscreen until gdm
>> starts.
>>
>> As discussed a while back, the proper way to do is set
>> stdout-path in the devicetree to the serial console, then the
>> kernel will automatically make it a second console, next to tty0
>> and output messages on both, and systemd will output messages
>> and spawn gettys on both too.
>>
>> Upstream u-boot is already setting stdout-path for all sunxi
>> devices, so from a sunxi pov the appending of " console=ttyS0,115200"
>> is a regression. If this is done for some other boards, it would
>> be better to either patch those boards dts files to set stdout-path,
>> or u-boot to set stdout-path rather then appending " console=ttyS0,115200"
>> and breaking video output. If you can tell me which specific boards
>> need work here I can whip up a patch (for others to test).
 >
> I suspect that no other boards are doing so, I was not aware that you had
> gotten that upstream yet.

For sunxi I've added a patch to u-boot which patches it into the dts as
u-boot loads the dts, u-boot already has infrastructure / code for
this, so it is just setting a single #define in include/configs/<board.h>,
see:

http://git.denx.de/?p=u-boot.git;a=commitdiff;h=f3133962f469a8b6b9ba237ba670f0ca7c88a02e

And there are also plans for sunxi to just outright add it to the dts files
as shipped with the kernel, so that it will also work with older u-boot
versions.

 > we will need to drop the patch that automatically
> adds the console from u-boot. every other system will need to be looked at.

Or wrap it in in #ifndef CONFIG_ARCH_SUNXI as an interim workaround,
TBH I was a bit surprised that we're carrying such a patch as what
to exactly put on the cmdline differs per board (the ttyS0 name is
not the same everywhere), or are you using the 'console' env
variable for this ?

Anyways at a minimum please wrap the patch automatically adding the
console= argument in #ifndef CONFIG_ARCH_SUNXI

Note that using stdout-path on devices which have hdmi (or similar video)
out is better because it gives a kernel console on both the hdmi and
the serial port, either way let me know if you need any help with this.

Regards,

Hans