mbox series

[0/3] add Ebang EBAZ4205 support

Message ID 20210120194033.26970-1-michael@walle.cc (mailing list archive)
Headers show
Series add Ebang EBAZ4205 support | expand

Message

Michael Walle Jan. 20, 2021, 7:40 p.m. UTC
Add support for the Ebang EBAZ4205 board. This board was once used as a
control board for a bitcoin mining device. Nowawdays it is sold as a cheap
Zynq-7000 eval board.

Michael Walle (3):
  dt-bindings: add ebang vendor prefix
  dt-bindings: arm: add Ebang EBAZ4205 board
  ARM: dts: add Ebang EBAZ4205 device tree

 .../devicetree/bindings/arm/xilinx.yaml       |   1 +
 .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
 arch/arm/boot/dts/Makefile                    |   1 +
 arch/arm/boot/dts/zynq-ebaz4205.dts           | 109 ++++++++++++++++++
 4 files changed, 113 insertions(+)
 create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts

Comments

Michal Simek Jan. 21, 2021, 9:25 a.m. UTC | #1
Hi

On 1/20/21 8:40 PM, Michael Walle wrote:
> Add support for the Ebang EBAZ4205 board. This board was once used as a
> control board for a bitcoin mining device. Nowawdays it is sold as a cheap
> Zynq-7000 eval board.
> 
> Michael Walle (3):
>   dt-bindings: add ebang vendor prefix
>   dt-bindings: arm: add Ebang EBAZ4205 board
>   ARM: dts: add Ebang EBAZ4205 device tree
> 
>  .../devicetree/bindings/arm/xilinx.yaml       |   1 +
>  .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
>  arch/arm/boot/dts/Makefile                    |   1 +
>  arch/arm/boot/dts/zynq-ebaz4205.dts           | 109 ++++++++++++++++++
>  4 files changed, 113 insertions(+)
>  create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts
> 

any link with schematics?

I will let dt guys to comment 1/3 but series look good to me.
The board doesn't look interesting from description point of view that's
why all the time thinking if makes sense to add it to kernel.

Thanks,
Michal
Michael Walle Jan. 21, 2021, 9:35 a.m. UTC | #2
Hi Michal,

Am 2021-01-21 10:25, schrieb Michal Simek:
> On 1/20/21 8:40 PM, Michael Walle wrote:
>> Add support for the Ebang EBAZ4205 board. This board was once used as 
>> a
>> control board for a bitcoin mining device. Nowawdays it is sold as a 
>> cheap
>> Zynq-7000 eval board.
>> 
>> Michael Walle (3):
>>   dt-bindings: add ebang vendor prefix
>>   dt-bindings: arm: add Ebang EBAZ4205 board
>>   ARM: dts: add Ebang EBAZ4205 device tree
>> 
>>  .../devicetree/bindings/arm/xilinx.yaml       |   1 +
>>  .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
>>  arch/arm/boot/dts/Makefile                    |   1 +
>>  arch/arm/boot/dts/zynq-ebaz4205.dts           | 109 
>> ++++++++++++++++++
>>  4 files changed, 113 insertions(+)
>>  create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts
>> 
> 
> any link with schematics?

https://github.com/xjtuecho/EBAZ4205, looks like these are
reverse engineered (from a layout file?) though.

> I will let dt guys to comment 1/3 but series look good to me.
> The board doesn't look interesting from description point of view 
> that's
> why all the time thinking if makes sense to add it to kernel.

What do you want to tell me? That for the time being, it didn't
appear to you to add the board yourself - or do you thing it
doesn't make sense at all. If its the latter, what would be
actual reason to have a board in mainline?

-michael
Michal Simek Jan. 21, 2021, 9:57 a.m. UTC | #3
Hi,

On 1/21/21 10:35 AM, Michael Walle wrote:
> Hi Michal,
> 
> Am 2021-01-21 10:25, schrieb Michal Simek:
>> On 1/20/21 8:40 PM, Michael Walle wrote:
>>> Add support for the Ebang EBAZ4205 board. This board was once used as a
>>> control board for a bitcoin mining device. Nowawdays it is sold as a
>>> cheap
>>> Zynq-7000 eval board.
>>>
>>> Michael Walle (3):
>>>   dt-bindings: add ebang vendor prefix
>>>   dt-bindings: arm: add Ebang EBAZ4205 board
>>>   ARM: dts: add Ebang EBAZ4205 device tree
>>>
>>>  .../devicetree/bindings/arm/xilinx.yaml       |   1 +
>>>  .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
>>>  arch/arm/boot/dts/Makefile                    |   1 +
>>>  arch/arm/boot/dts/zynq-ebaz4205.dts           | 109 ++++++++++++++++++
>>>  4 files changed, 113 insertions(+)
>>>  create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts
>>>
>>
>> any link with schematics?
> 
> https://github.com/xjtuecho/EBAZ4205, looks like these are
> reverse engineered (from a layout file?) though.

Interesting but at least something.

> 
>> I will let dt guys to comment 1/3 but series look good to me.
>> The board doesn't look interesting from description point of view that's
>> why all the time thinking if makes sense to add it to kernel.
> 
> What do you want to tell me? That for the time being, it didn't
> appear to you to add the board yourself - or do you thing it
> doesn't make sense at all. If its the latter, what would be
> actual reason to have a board in mainline?

I have bad experience with for example Avnet boards which people add and
none is really updating them and they are in the same state for years.

Long time ago we agreed that doesn't make sense to describe PL in
upstream projects and we only describe PS part. It means you likely miss
several things which are useful and the reason for using these SoCs is PL.

As you likely know Xilinx has Versal device and I didn't push any device
tree to any upstream project and thinking not to add any description for
boards and stay in sort of space that "virtual" description for SoC
should be enough. Maybe just versal.dtsi and one kitchen sink DT should
be added but not description for all boards.

The same is if make sense to push all DTs for all standard xilinx zynqmp
evaluation boards. If there is something interesting/new I thought it
makes sense to add it as pattern to follow. But for boards which looks
very similar from PS point of view I don't think there is real value to
add and invest time for maintaining.

Back to your case. Board is cheap which is not all the time case for any
xilinx board but you have only uart, sd and partially described ethernet
which doesn't work without PL. Is it worth to have this described?
Especially when it is visible that you need to describe custom PL and DT
overlays are not solid yet.

Thanks,
Michal
Michael Walle Jan. 21, 2021, 10:13 a.m. UTC | #4
Hi,

Am 2021-01-21 10:57, schrieb Michal Simek:
> Hi,
> 
> On 1/21/21 10:35 AM, Michael Walle wrote:
>> Hi Michal,
>> 
>> Am 2021-01-21 10:25, schrieb Michal Simek:
>>> On 1/20/21 8:40 PM, Michael Walle wrote:
>>>> Add support for the Ebang EBAZ4205 board. This board was once used 
>>>> as a
>>>> control board for a bitcoin mining device. Nowawdays it is sold as a
>>>> cheap
>>>> Zynq-7000 eval board.
>>>> 
>>>> Michael Walle (3):
>>>>   dt-bindings: add ebang vendor prefix
>>>>   dt-bindings: arm: add Ebang EBAZ4205 board
>>>>   ARM: dts: add Ebang EBAZ4205 device tree
>>>> 
>>>>  .../devicetree/bindings/arm/xilinx.yaml       |   1 +
>>>>  .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
>>>>  arch/arm/boot/dts/Makefile                    |   1 +
>>>>  arch/arm/boot/dts/zynq-ebaz4205.dts           | 109 
>>>> ++++++++++++++++++
>>>>  4 files changed, 113 insertions(+)
>>>>  create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts
>>>> 
>>> 
>>> any link with schematics?
>> 
>> https://github.com/xjtuecho/EBAZ4205, looks like these are
>> reverse engineered (from a layout file?) though.
> 
> Interesting but at least something.
> 
>> 
>>> I will let dt guys to comment 1/3 but series look good to me.
>>> The board doesn't look interesting from description point of view 
>>> that's
>>> why all the time thinking if makes sense to add it to kernel.
>> 
>> What do you want to tell me? That for the time being, it didn't
>> appear to you to add the board yourself - or do you thing it
>> doesn't make sense at all. If its the latter, what would be
>> actual reason to have a board in mainline?
> 
> I have bad experience with for example Avnet boards which people add 
> and
> none is really updating them and they are in the same state for years.

Wouldn't it be better then to pull the plug at some time and remove 
these
boards.

TBH I was a bit disappointed by your statement. It sounded like "nah
this board isn't worth it". Esp. because it is just one (small) file.
But more below.

> Long time ago we agreed that doesn't make sense to describe PL in
> upstream projects and we only describe PS part. It means you likely 
> miss
> several things which are useful and the reason for using these SoCs is 
> PL.
> 
> As you likely know Xilinx has Versal device and I didn't push any 
> device
> tree to any upstream project and thinking not to add any description 
> for
> boards and stay in sort of space that "virtual" description for SoC
> should be enough. Maybe just versal.dtsi and one kitchen sink DT should
> be added but not description for all boards.
> 
> The same is if make sense to push all DTs for all standard xilinx 
> zynqmp
> evaluation boards. If there is something interesting/new I thought it
> makes sense to add it as pattern to follow. But for boards which looks
> very similar from PS point of view I don't think there is real value to
> add and invest time for maintaining.
> 
> Back to your case. Board is cheap which is not all the time case for 
> any
> xilinx board but you have only uart, sd and partially described 
> ethernet
> which doesn't work without PL. Is it worth to have this described?

I got your point. But it is at least a jump start for the users if that
board boots out of the box. And yes, its unfortunate, that ethernet
just works if the PL is configured. This is already done by the
bootloader, because there I do have the same problem.

> Especially when it is visible that you need to describe custom PL and 
> DT
> overlays are not solid yet.
> 
> Thanks,
> Michal
Michal Simek Jan. 21, 2021, 10:23 a.m. UTC | #5
Hi,

On 1/21/21 11:13 AM, Michael Walle wrote:
> Hi,
> 
> Am 2021-01-21 10:57, schrieb Michal Simek:
>> Hi,
>>
>> On 1/21/21 10:35 AM, Michael Walle wrote:
>>> Hi Michal,
>>>
>>> Am 2021-01-21 10:25, schrieb Michal Simek:
>>>> On 1/20/21 8:40 PM, Michael Walle wrote:
>>>>> Add support for the Ebang EBAZ4205 board. This board was once used
>>>>> as a
>>>>> control board for a bitcoin mining device. Nowawdays it is sold as a
>>>>> cheap
>>>>> Zynq-7000 eval board.
>>>>>
>>>>> Michael Walle (3):
>>>>>   dt-bindings: add ebang vendor prefix
>>>>>   dt-bindings: arm: add Ebang EBAZ4205 board
>>>>>   ARM: dts: add Ebang EBAZ4205 device tree
>>>>>
>>>>>  .../devicetree/bindings/arm/xilinx.yaml       |   1 +
>>>>>  .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
>>>>>  arch/arm/boot/dts/Makefile                    |   1 +
>>>>>  arch/arm/boot/dts/zynq-ebaz4205.dts           | 109
>>>>> ++++++++++++++++++
>>>>>  4 files changed, 113 insertions(+)
>>>>>  create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts
>>>>>
>>>>
>>>> any link with schematics?
>>>
>>> https://github.com/xjtuecho/EBAZ4205, looks like these are
>>> reverse engineered (from a layout file?) though.
>>
>> Interesting but at least something.
>>
>>>
>>>> I will let dt guys to comment 1/3 but series look good to me.
>>>> The board doesn't look interesting from description point of view
>>>> that's
>>>> why all the time thinking if makes sense to add it to kernel.
>>>
>>> What do you want to tell me? That for the time being, it didn't
>>> appear to you to add the board yourself - or do you thing it
>>> doesn't make sense at all. If its the latter, what would be
>>> actual reason to have a board in mainline?
>>
>> I have bad experience with for example Avnet boards which people add and
>> none is really updating them and they are in the same state for years.
> 
> Wouldn't it be better then to pull the plug at some time and remove these
> boards.
> 
> TBH I was a bit disappointed by your statement. It sounded like "nah
> this board isn't worth it". Esp. because it is just one (small) file.
> But more below.
> 
>> Long time ago we agreed that doesn't make sense to describe PL in
>> upstream projects and we only describe PS part. It means you likely miss
>> several things which are useful and the reason for using these SoCs is
>> PL.
>>
>> As you likely know Xilinx has Versal device and I didn't push any device
>> tree to any upstream project and thinking not to add any description for
>> boards and stay in sort of space that "virtual" description for SoC
>> should be enough. Maybe just versal.dtsi and one kitchen sink DT should
>> be added but not description for all boards.
>>
>> The same is if make sense to push all DTs for all standard xilinx zynqmp
>> evaluation boards. If there is something interesting/new I thought it
>> makes sense to add it as pattern to follow. But for boards which looks
>> very similar from PS point of view I don't think there is real value to
>> add and invest time for maintaining.
>>
>> Back to your case. Board is cheap which is not all the time case for any
>> xilinx board but you have only uart, sd and partially described ethernet
>> which doesn't work without PL. Is it worth to have this described?
> 
> I got your point. But it is at least a jump start for the users if that
> board boots out of the box. And yes, its unfortunate, that ethernet
> just works if the PL is configured. This is already done by the
> bootloader, because there I do have the same problem.

Zynq/ZynqMP boards can use U-Boot SPL. "Advantage" of this solution
especially for Zynq is that in u-boot there is open a way for adding
ps7_init file which is determined by device tree name.
I think it would make sense to add these DTs and also ps7_init to U-Boot
project and wire it up with zynq_virt platform and then you can boot
Linux with using U-Boot's DT pointed by $fdtcontroladdr.
Then you will get support from scratch to be able to boot.

Thanks,
Michal
Michael Walle Jan. 21, 2021, 10:41 a.m. UTC | #6
Hi,

Am 2021-01-21 11:23, schrieb Michal Simek:
>>> Back to your case. Board is cheap which is not all the time case for 
>>> any
>>> xilinx board but you have only uart, sd and partially described 
>>> ethernet
>>> which doesn't work without PL. Is it worth to have this described?
>> 
>> I got your point. But it is at least a jump start for the users if 
>> that
>> board boots out of the box. And yes, its unfortunate, that ethernet
>> just works if the PL is configured. This is already done by the
>> bootloader, because there I do have the same problem.
> 
> Zynq/ZynqMP boards can use U-Boot SPL. "Advantage" of this solution
> especially for Zynq is that in u-boot there is open a way for adding
> ps7_init file which is determined by device tree name.
> I think it would make sense to add these DTs and also ps7_init to 
> U-Boot
> project and wire it up with zynq_virt platform and then you can boot
> Linux with using U-Boot's DT pointed by $fdtcontroladdr.
> Then you will get support from scratch to be able to boot.

I already have patches for u-boot (using SPL). But my impression was
that linux is the master for the device trees. Esp. if there are some
problems with the board its often useful to have an in-tree device
tree.

What is the difference between this board and the other zynq boards
in the kernel?

In any case, please make this decision now: will you accept this
device tree or not?

-michael
Michal Simek Jan. 21, 2021, 1:16 p.m. UTC | #7
On 1/21/21 11:41 AM, Michael Walle wrote:
> Hi,
> 
> Am 2021-01-21 11:23, schrieb Michal Simek:
>>>> Back to your case. Board is cheap which is not all the time case for
>>>> any
>>>> xilinx board but you have only uart, sd and partially described
>>>> ethernet
>>>> which doesn't work without PL. Is it worth to have this described?
>>>
>>> I got your point. But it is at least a jump start for the users if that
>>> board boots out of the box. And yes, its unfortunate, that ethernet
>>> just works if the PL is configured. This is already done by the
>>> bootloader, because there I do have the same problem.
>>
>> Zynq/ZynqMP boards can use U-Boot SPL. "Advantage" of this solution
>> especially for Zynq is that in u-boot there is open a way for adding
>> ps7_init file which is determined by device tree name.
>> I think it would make sense to add these DTs and also ps7_init to U-Boot
>> project and wire it up with zynq_virt platform and then you can boot
>> Linux with using U-Boot's DT pointed by $fdtcontroladdr.
>> Then you will get support from scratch to be able to boot.
> 
> I already have patches for u-boot (using SPL). But my impression was
> that linux is the master for the device trees. Esp. if there are some
> problems with the board its often useful to have an in-tree device
> tree.
> 
> What is the difference between this board and the other zynq boards
> in the kernel?
> 
> In any case, please make this decision now: will you accept this
> device tree or not?

If you promise to regularly test it I am fine with it.

Thanks,
Michal
Michael Walle Jan. 21, 2021, 3:14 p.m. UTC | #8
Am 2021-01-21 14:16, schrieb Michal Simek:
> On 1/21/21 11:41 AM, Michael Walle wrote:
>> Am 2021-01-21 11:23, schrieb Michal Simek:
>>>>> Back to your case. Board is cheap which is not all the time case 
>>>>> for
>>>>> any
>>>>> xilinx board but you have only uart, sd and partially described
>>>>> ethernet
>>>>> which doesn't work without PL. Is it worth to have this described?
>>>> 
>>>> I got your point. But it is at least a jump start for the users if 
>>>> that
>>>> board boots out of the box. And yes, its unfortunate, that ethernet
>>>> just works if the PL is configured. This is already done by the
>>>> bootloader, because there I do have the same problem.
>>> 
>>> Zynq/ZynqMP boards can use U-Boot SPL. "Advantage" of this solution
>>> especially for Zynq is that in u-boot there is open a way for adding
>>> ps7_init file which is determined by device tree name.
>>> I think it would make sense to add these DTs and also ps7_init to 
>>> U-Boot
>>> project and wire it up with zynq_virt platform and then you can boot
>>> Linux with using U-Boot's DT pointed by $fdtcontroladdr.
>>> Then you will get support from scratch to be able to boot.
>> 
>> I already have patches for u-boot (using SPL). But my impression was
>> that linux is the master for the device trees. Esp. if there are some
>> problems with the board its often useful to have an in-tree device
>> tree.
>> 
>> What is the difference between this board and the other zynq boards
>> in the kernel?
>> 
>> In any case, please make this decision now: will you accept this
>> device tree or not?
> 
> If you promise to regularly test it I am fine with it.

I might even integrate it into our lava lab.

-michael
Michal Simek Jan. 26, 2021, 1:31 p.m. UTC | #9
On 1/20/21 8:40 PM, Michael Walle wrote:
> Add support for the Ebang EBAZ4205 board. This board was once used as a
> control board for a bitcoin mining device. Nowawdays it is sold as a cheap
> Zynq-7000 eval board.
> 
> Michael Walle (3):
>   dt-bindings: add ebang vendor prefix
>   dt-bindings: arm: add Ebang EBAZ4205 board
>   ARM: dts: add Ebang EBAZ4205 device tree
> 
>  .../devicetree/bindings/arm/xilinx.yaml       |   1 +
>  .../devicetree/bindings/vendor-prefixes.yaml  |   2 +
>  arch/arm/boot/dts/Makefile                    |   1 +
>  arch/arm/boot/dts/zynq-ebaz4205.dts           | 109 ++++++++++++++++++
>  4 files changed, 113 insertions(+)
>  create mode 100644 arch/arm/boot/dts/zynq-ebaz4205.dts
> 

Applied all.

Thanks,
Michal