All of lore.kernel.org
 help / color / mirror / Atom feed
* Re: RK3572 / OK3572-C support on top of your rk3562 branch
       [not found] <tencent_907535C1D3F4C91824948CA407902292FA0A@qq.com>
@ 2026-10-22 19:16 ` Jonas Karlman
  2026-11-20 16:43   ` Quentin Schulz
  0 siblings, 1 reply; 3+ messages in thread
From: Jonas Karlman @ 2026-10-22 19:16 UTC (permalink / raw)
  To: Cunhao Lu; +Cc: u-boot

Hi Cunhao,

On 10/19/2026 9:49 AM, Cunhao Lu wrote:
> Hi Jonas,
> 
> I've been working on U-Boot support for the Forlinx OK3572-C board, which
> uses the FET3572-C module. After finding your RK3572 support, I rebased my
> work onto your rk3562 branch and now use your SoC, clock, pinctrl and RAM
> drivers.
> 
> I have three follow-up patches:
> 
> - Add optional RBOOT + DDR + SPL three-stage image support to mkimage.
> - Add optional RBOOT firmware support to binman.
> - Add the OK3572-C device tree, defconfig and build documentation.
>   (With Forlinx OK3572-C BSP Alpha1)
> 
> I have verified the complete boot flow from eMMC on the OK3572-C. The
> bootloader, kernel and root filesystem are all stored on eMMC, and the
> board boots into a minimal Debian 13 system.
> 
> Do you have a submission plan for the RK3572 support in your branch?
> Would you prefer me to send these changes as a separate RFC series based
> on your branch, or coordinate them with your SoC support series?

For Rockchip platform we typically require device tree for new SoCs and
boards to be picked from upstream Linux (the devicetree-rebasing tree).

So the first step is usually to upstream dt-bindings, device tree and
core drivers to Linux before patches for the SoC are submitted to U-Boot.

My rk3562 branch contains a long queue of new RK SoC support, some are
waiting on Linux upstream support, others are waiting on open U-Boot RK
refactoring series and/or other mailing list discussions. eMMC/SD-card,
Ethernet, RNG, OTP, SARADC and USB2 should be fully working for all the
pending SoCs in the rk3562 branch. All (or most) patches in rk3562
should eventually hit the mailing list, at least that's the goal.

Please note that for RK3572 e.g. the reset header, pinctrl dtsi, usb2phy
and other DT nodes have been adjusted/corrected for mainline purposes,
likely some other parts to make it more consistent with existing SoCs.

I am currently using [1] with boot_merger to get dboot/rboot merged as
part of my generic-rk3572_defconfig build and test workflow [2][3].

[1] https://github.com/Kwiboo/scripts/blob/main/u-boot/RK3572MINIALL-ramboot.ini
[2] https://github.com/Kwiboo/scripts/blob/main/u-boot.sh#L66-L68
[3] https://github.com/Kwiboo/scripts/blob/main/u-boot.sh#L99-L101

Regards,
Jonas

> 
> I'd be happy to share the patches and help test further changes on this
> board.
> 
> Thanks for your work on this.
> 
> Best regards,
> Cunhao Lu



^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: RK3572 / OK3572-C support on top of your rk3562 branch
  2026-10-22 19:16 ` RK3572 / OK3572-C support on top of your rk3562 branch Jonas Karlman
@ 2026-11-20 16:43   ` Quentin Schulz
  2026-11-20 18:54     ` Jonas Karlman
  0 siblings, 1 reply; 3+ messages in thread
From: Quentin Schulz @ 2026-11-20 16:43 UTC (permalink / raw)
  To: Jonas Karlman, Cunhao Lu; +Cc: u-boot

Hi Jonas, Cunhao,

On 10/22/26 9:16 PM, Jonas Karlman wrote:
> Hi Cunhao,
> 
> On 10/19/2026 9:49 AM, Cunhao Lu wrote:
>> Hi Jonas,
>>
>> I've been working on U-Boot support for the Forlinx OK3572-C board, which
>> uses the FET3572-C module. After finding your RK3572 support, I rebased my
>> work onto your rk3562 branch and now use your SoC, clock, pinctrl and RAM
>> drivers.
>>
>> I have three follow-up patches:
>>
>> - Add optional RBOOT + DDR + SPL three-stage image support to mkimage.
>> - Add optional RBOOT firmware support to binman.
>> - Add the OK3572-C device tree, defconfig and build documentation.
>>    (With Forlinx OK3572-C BSP Alpha1)
>>
>> I have verified the complete boot flow from eMMC on the OK3572-C. The
>> bootloader, kernel and root filesystem are all stored on eMMC, and the
>> board boots into a minimal Debian 13 system.
>>
>> Do you have a submission plan for the RK3572 support in your branch?
>> Would you prefer me to send these changes as a separate RFC series based
>> on your branch, or coordinate them with your SoC support series?
> 
> For Rockchip platform we typically require device tree for new SoCs and
> boards to be picked from upstream Linux (the devicetree-rebasing tree).
> 
> So the first step is usually to upstream dt-bindings, device tree and
> core drivers to Linux before patches for the SoC are submitted to U-Boot.
> 

Just to be clear, the limitation only applies to dt-bindings and the 
various Device Trees, but not to drivers. You're free to submit driver 
support for new SoC whenever you feel like it. Of course, without a 
dt-bindings agreed with upstream, the driver may not be compatible with 
the dt-bindings once finalized, hence why it's a good idea to wait at 
least for the dt-bindings for the driver to be merged in the Linux 
kernel before sending patches for it in U-Boot. I think it's a 
technicality since they (driver and its binding) are usually contributed 
together in a single series, but you could do that, that's fine on my side.

> My rk3562 branch contains a long queue of new RK SoC support, some are
> waiting on Linux upstream support, others are waiting on open U-Boot RK
> refactoring series and/or other mailing list discussions. eMMC/SD-card,
> Ethernet, RNG, OTP, SARADC and USB2 should be fully working for all the
> pending SoCs in the rk3562 branch. All (or most) patches in rk3562
> should eventually hit the mailing list, at least that's the goal.
> 

I think I've caught up with all your patches? Three series awaiting a 
new version (or feedback) as far as I could tell:
- https://lore.kernel.org/u-boot/20260805174509.2968252-1-jonas@kwiboo.se/
- https://lore.kernel.org/u-boot/20260804214442.2592597-1-jonas@kwiboo.se/
- https://lore.kernel.org/u-boot/20260804105721.2402406-1-jonas@kwiboo.se/

Is there something I missed that I should be doing? Let me know if I 
somehow skipped a series or if I need to act on something.

@Cunhao, the typical policy we have is that the first who contributes a 
patch usually gets to work on it until it's satisfactory and can be 
merged. We also typically do not care about unofficial trees, however 
close they are to upstream: if you have patches, and you feel they are 
good enough, you can send them even if another contributor has their own 
patch for the same thing in their tree (and haven't sent patches yet). 
Of course, do not go to a contributor tree and get patches from there 
without having permission to do so. Please do not use LLMs "to be the 
first" (or at all, we don't really like those here).

I know Jonas has many many many patches in his tree and he's been a 
relentless contributor and reviewer for anything Rockchip in U-Boot (and 
I know he's very active in other communities as well), I'm very grateful 
for that. I'm pretty sure Jonas was careful not to send too many patches 
and overload me, so that is also very much appreciated.
I'd like to both avoid duplicated work (Jonas and someone else doing the 
same work) and also putting too much pressure on Jonas by contributors 
"having to wait" for patches to land on the mailing list to start/resume 
bring up of SoCs. "Putting all eggs in the same basket" kind of issues :)

No promise on when I can look at patches, I'm the sole custodian at the 
moment and there's some decent backlog already :) That shouldn't prevent 
you from sending patches to the mailing list and/or reviewing other 
people's patches to help out, it's just that things will take time to 
eventually get merged.

Cheers,
Quentin

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: RK3572 / OK3572-C support on top of your rk3562 branch
  2026-11-20 16:43   ` Quentin Schulz
@ 2026-11-20 18:54     ` Jonas Karlman
  0 siblings, 0 replies; 3+ messages in thread
From: Jonas Karlman @ 2026-11-20 18:54 UTC (permalink / raw)
  To: Quentin Schulz, Cunhao Lu; +Cc: u-boot

Hi Quentin, Cunhao,

On 11/20/2026 5:43 PM, Quentin Schulz wrote:
> Hi Jonas, Cunhao,
> 
> On 10/22/26 9:16 PM, Jonas Karlman wrote:
>> Hi Cunhao,
>>
>> On 10/19/2026 9:49 AM, Cunhao Lu wrote:
>>> Hi Jonas,
>>>
>>> I've been working on U-Boot support for the Forlinx OK3572-C board, which
>>> uses the FET3572-C module. After finding your RK3572 support, I rebased my
>>> work onto your rk3562 branch and now use your SoC, clock, pinctrl and RAM
>>> drivers.
>>>
>>> I have three follow-up patches:
>>>
>>> - Add optional RBOOT + DDR + SPL three-stage image support to mkimage.
>>> - Add optional RBOOT firmware support to binman.
>>> - Add the OK3572-C device tree, defconfig and build documentation.
>>>    (With Forlinx OK3572-C BSP Alpha1)
>>>
>>> I have verified the complete boot flow from eMMC on the OK3572-C. The
>>> bootloader, kernel and root filesystem are all stored on eMMC, and the
>>> board boots into a minimal Debian 13 system.
>>>
>>> Do you have a submission plan for the RK3572 support in your branch?
>>> Would you prefer me to send these changes as a separate RFC series based
>>> on your branch, or coordinate them with your SoC support series?
>>
>> For Rockchip platform we typically require device tree for new SoCs and
>> boards to be picked from upstream Linux (the devicetree-rebasing tree).
>>
>> So the first step is usually to upstream dt-bindings, device tree and
>> core drivers to Linux before patches for the SoC are submitted to U-Boot.
>>
> 
> Just to be clear, the limitation only applies to dt-bindings and the 
> various Device Trees, but not to drivers. You're free to submit driver 
> support for new SoC whenever you feel like it. Of course, without a 
> dt-bindings agreed with upstream, the driver may not be compatible with 
> the dt-bindings once finalized, hence why it's a good idea to wait at 
> least for the dt-bindings for the driver to be merged in the Linux 
> kernel before sending patches for it in U-Boot. I think it's a 
> technicality since they (driver and its binding) are usually contributed 
> together in a single series, but you could do that, that's fine on my side.
> 
>> My rk3562 branch contains a long queue of new RK SoC support, some are
>> waiting on Linux upstream support, others are waiting on open U-Boot RK
>> refactoring series and/or other mailing list discussions. eMMC/SD-card,
>> Ethernet, RNG, OTP, SARADC and USB2 should be fully working for all the
>> pending SoCs in the rk3562 branch. All (or most) patches in rk3562
>> should eventually hit the mailing list, at least that's the goal.
>>
> 
> I think I've caught up with all your patches? Three series awaiting a 
> new version (or feedback) as far as I could tell:
> - https://lore.kernel.org/u-boot/20260805174509.2968252-1-jonas@kwiboo.se/
> - https://lore.kernel.org/u-boot/20260804214442.2592597-1-jonas@kwiboo.se/
> - https://lore.kernel.org/u-boot/20260804105721.2402406-1-jonas@kwiboo.se/
> 
> Is there something I missed that I should be doing? Let me know if I 
> somehow skipped a series or if I need to act on something.

Not that I know of, will continue and refresh series now that clock fixes
have landed, thanks!

Currently have following series in mind for a refresh and posting:
- misc fixes/improvements for rk3506 support
- cleanup: remove unneeded syscon driver for rk35xx
- cleanup: remove uses of rockchip_get_cru() helper
- refactor rk3506 clock driver (baseline for rk35xx clock drivers)

And then continue with following:
- add support for rk3562, rv1126b, rv1106, rv1103b
  - held back pending: cleanups and clock driver baseline
- add support for rk3538 and rk3572
  - held back pending: upstream dt-bindings
- continued clock driver cleanup/refactor for existing socs
- consistent tpl handling

> @Cunhao, the typical policy we have is that the first who contributes a 
> patch usually gets to work on it until it's satisfactory and can be 
> merged. We also typically do not care about unofficial trees, however 
> close they are to upstream: if you have patches, and you feel they are 
> good enough, you can send them even if another contributor has their own 
> patch for the same thing in their tree (and haven't sent patches yet). 
> Of course, do not go to a contributor tree and get patches from there 
> without having permission to do so. Please do not use LLMs "to be the 
> first" (or at all, we don't really like those here).

Personally I typically have no issue with others picking non-WIP marked
patches from my tree to possible help speed things along, at least not
after a quick ping. I have several Linux patches that landed upstream
that where submitted by others :-)

> I know Jonas has many many many patches in his tree and he's been a 
> relentless contributor and reviewer for anything Rockchip in U-Boot (and 
> I know he's very active in other communities as well), I'm very grateful 
> for that. I'm pretty sure Jonas was careful not to send too many patches 
> and overload me, so that is also very much appreciated.

Correct, I have tried not to keep too many series alive at the same time and
switched focus to fixes when issues with existing code was discovered.

> I'd like to both avoid duplicated work (Jonas and someone else doing the 
> same work) and also putting too much pressure on Jonas by contributors 
> "having to wait" for patches to land on the mailing list to start/resume 
> bring up of SoCs. "Putting all eggs in the same basket" kind of issues :)

I mostly just want to have consistent code and features for any new and
existing SoCs (focusing on newer RK35xx SoCs). To avoid spending too
much future time on possible cleanups of inconsistent code, I just
completed similar steps that I had already done to help bring up prior
RK35xx SoCs.

Patches still not being posted is in part due to trying to prioritize
what I though was going to be quick cleanups and fixes ahead of adding
any new SoC support, and also getting consensus on clock driver handling
using patches to refactor an existing SoC (rk3506) instead of tying to
sneak in inconsistent clock handling code using a new SoC.

I will try to reduce my pending queue by posting a few updated and new
series this weekend.

Regards,
Jonas

> No promise on when I can look at patches, I'm the sole custodian at the 
> moment and there's some decent backlog already :) That shouldn't prevent 
> you from sending patches to the mailing list and/or reviewing other 
> people's patches to help out, it's just that things will take time to 
> eventually get merged.
> 
> Cheers,
> Quentin


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-20 18:54 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <tencent_907535C1D3F4C91824948CA407902292FA0A@qq.com>
2026-10-22 19:16 ` RK3572 / OK3572-C support on top of your rk3562 branch Jonas Karlman
2026-11-20 16:43   ` Quentin Schulz
2026-11-20 18:54     ` Jonas Karlman

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.