[U-Boot] [PATCH v3 7/7] cmd: Add bind/unbind commands to bind a device to a driver from the command line

[Michal Simek]

On 11.7.2018 14:46, Tom Rini wrote:
> On Wed, Jul 11, 2018 at 07:57:13AM +0200, Michal Simek wrote:
>> On 10.7.2018 18:40, Tom Rini wrote:
>>> On Mon, Jul 09, 2018 at 11:59:57AM -0500, Joe Hershberger wrote:
>>>> On Mon, Jul 9, 2018 at 9:43 AM, Tom Rini <trini at konsulko.com> wrote:
>>>>> On Mon, Jul 09, 2018 at 08:19:44AM +0200, Michal Simek wrote:
>>>>>> On 30.6.2018 06:19, Simon Glass wrote:
>>>>>>> On 27 June 2018 at 07:13, Michal Simek <michal.simek at xilinx.com> wrote:
>>>>>>>> On 22.6.2018 14:25, Jean-Jacques Hiblot wrote:
>>>>>>>>> In some cases it can be useful to be able to bind a device to a driver from
>>>>>>>>> the command line.
>>>>>>>>> The obvious example is for versatile devices such as USB gadget.
>>>>>>>>> Another use case is when the devices are not yet ready at startup and
>>>>>>>>> require some setup before the drivers are bound (ex: FPGA which bitsream is
>>>>>>>>> fetched from a mass storage or ethernet)
>>>>>>>>>
>>>>>>>>> usage example:
>>>>>>>>>
>>>>>>>>> bind usb_dev_generic 0 usb_ether
>>>>>>>>> unbind usb_dev_generic 0 usb_ether
>>>>>>>>> or
>>>>>>>>> unbind eth 1
>>>>>>>>>
>>>>>>>>> bind /ocp/omap_dwc3 at 48380000/usb at 48390000 usb_ether
>>>>>>>>> unbind /ocp/omap_dwc3 at 48380000/usb at 48390000
>>>>>>>>>
>>>>>>>>> Signed-off-by: Jean-Jacques Hiblot <jjhiblot at ti.com>
>>>>>>>>>
>>>>>>>>> ---
>>>>>>>>>
>>>>>>>>> Changes in v3:
>>>>>>>>> - factorize code based on comments from ML
>>>>>>>>> - remove the devices before unbinding them
>>>>>>>>> - use device_find_global_by_ofnode() to get a device by its node.
>>>>>>>>> - Added tests
>>>>>>>>>
>>>>>>>>> Changes in v2:
>>>>>>>>> - Make the bind/unbind command generic, not specific to usb device.
>>>>>>>>> - Update the API to be able to bind/unbind based on DTS node path
>>>>>>>>> - Add a Kconfig option to select the bind/unbind commands
>>>>>>>>>
>>>>>>>>>  arch/sandbox/dts/test.dts  |  11 ++
>>>>>>>>>  cmd/Kconfig                |   9 ++
>>>>>>>>>  cmd/Makefile               |   1 +
>>>>>>>>>  cmd/bind.c                 | 255 +++++++++++++++++++++++++++++++++++++++++++++
>>>>>>>>>  configs/sandbox_defconfig  |   1 +
>>>>>>>>>  test/py/tests/test_bind.py | 178 +++++++++++++++++++++++++++++++
>>>>>>>>>  6 files changed, 455 insertions(+)
>>>>>>>>>  create mode 100644 cmd/bind.c
>>>>>>>>>  create mode 100644 test/py/tests/test_bind.py
>>>>>>>
>>>>>>> Reviewed-by: Simon Glass <sjg at chromium.org>
>>>>>>>
>>>>>>> Nice work
>>>>>>>
>>>>>>> [...]
>>>>>>>
>>>>>>>>
>>>>>>>> I have tested bind/unbind with dwc3 on zynqmp for ethernet gadget and it
>>>>>>>> is working fine for me.
>>>>>>>> I have also tried to use bind/unbind for gpio zynqmp driver and it is
>>>>>>>> also working fine.
>>>>>>>>
>>>>>>>> It means
>>>>>>>> Tested-by: Michal Simek <michal.simek at xilinx.com>
>>>>>>>>
>>>>>>>> I would prefer if these commands are called as dm bind and dm unbind
>>>>>>>> instead of simply bind/unbind which should also fit to dm command
>>>>>>>> description "dm - Driver model low level access".
>>>>>>>
>>>>>>> Yes i can see the point. But after thinking about it, maybe it is best
>>>>>>> as it is? After all driver model is fundamental to U-Boot and it's not
>>>>>>> clear what else we might bind/unbind.
>>>>>>>
>>>>>>> I'd like to get other opinions here, too.
>>>>>>
>>>>>> Tom/Marek: Any opinion?
>>>>>
>>>>> I think dm bind/unbind makes sense, yes.  "bind" and "unbind" are pretty
>>>>> generic terms and making it clear it's part of working "inside" of DM to
>>>>> hook/unhook things by making it a sub-command of dm sounds good.
>>>>> Thanks!
>>>>
>>>> I agree with Simon here. I think bind and unbind won't have any
>>>> plausible other meaning in U-Boot and DM is core to U-Boot
>>>> functionality in the new world. I think it would be OK to have "dm
>>>> bind" alias to "bind" if that's a point of confusion, but having it
>>>> top-level seems good to me.
>>>
>>> They're still very generic words for something that's part of working
>>> under the dm framework.  That said, looking at test/dm/cmd_dm.c and that
>>> it's supposed to be only for test/debug type work, yes, OK, I'll change
>>> my mind.
>>
>> I would expect that almost everybody will enable CMD_DM where symbol is
>> in cmd/Kconfig but implementation in test/dm/cmd_dm.c (it is a question
>> if this is the right place for this file).
> 
> It might well really belong as cmd/dm.c, but content wise it's debug
> information that is also useful in the bind/unbind case (so you know
> where U-Boot sees things as being at).

Then we should really not enable it by default for all boards with DM on.

 640 config CMD_DM
 641         bool "dm - Access to driver model information"
 642         depends on DM
 643         default y

Thanks,
Michal