[PATCH 7/7] verdin-imx8mp_defconfig: Enable PCIe/NVMe support

[Marek Vasut]

On 2/21/24 13:12, Sumit Garg wrote:
> On Wed, 21 Feb 2024 at 16:11, Marek Vasut <marex at denx.de> wrote:
>>
>> On 2/21/24 10:18, Marcel Ziswiler wrote:
>>> Hi Sumit
>>>
>>> On Wed, 2024-02-21 at 08:55 +0100, Francesco Dolcini wrote:
>>>> Hello Sumit,
>>>>
>>>> On Tue, Feb 20, 2024 at 06:40:56PM +0530, Sumit Garg wrote:
>>>>> Also, enable reset driver which is a prerequisite for PCIe support.
>>>>>
>>>>> Signed-off-by: Sumit Garg <sumit.garg at linaro.org>
>>>>> ---
>>>>>    configs/verdin-imx8mp_defconfig | 9 +++++++++
>>>>>    1 file changed, 9 insertions(+)
>>>>>
>>>>> diff --git a/configs/verdin-imx8mp_defconfig b/configs/verdin-imx8mp_defconfig
>>>>> index 22b8a334dfa..d8bd644322b 100644
>>>>> --- a/configs/verdin-imx8mp_defconfig
>>>>> +++ b/configs/verdin-imx8mp_defconfig
>>>>> @@ -185,3 +185,12 @@ CONFIG_USB_GADGET_VENDOR_NUM=0x1b67
>>>>>    CONFIG_USB_GADGET_PRODUCT_NUM=0x4000
>>>>>    CONFIG_IMX_WATCHDOG=y
>>>>>    CONFIG_HEXDUMP=y
>>>>> +CONFIG_DM_RESET=y
>>>>> +CONFIG_RESET_IMX=y
>>>>> +CONFIG_PCI=y
>>>>> +CONFIG_PCIE_DW_IMX8=y
>>>>> +CONFIG_PHY_IMX8M_PCIE=y
>>>>> +CONFIG_CMD_PCI=y
>>>>> +CONFIG_NVME=y
>>>>> +CONFIG_NVME_PCI=y
>>>>> +CONFIG_CMD_NVME=y
>>>>
>>>> This will increase the u-boot proper size
>>>
>>> Yes, I checked and it is actually slightly more than 32 K.
>>>
>>>> and marginally increase the
>>>> boot time (because of a bigger binary to be read from the eMMC).
>>>
>>> That was also my concern.
>>>
>>>> Apart of that do you expect any other impact on those changes? SPL
>>>> binary size should not be affected, correct?
>>>>
>>>> Asking this out loudly to confirm that nothing unexpected is going to
>>>> happen because of these changes.
>>>
>>> Other than that I actually gave it a quick try and PCIe/NVMe does indeed work and the regular boot is not
>>> affected (other than the slight size and boot time increase, of course).
>>>
>>>> For my curiosity, care to share what's the use case? Do you plan to have
>>>> the OS stored into an NVME device?
>>>
>>> For us the question is basically whether that use case does mandate enforcing such changes for each and every
>>> customer. Plus the regular expected maintenance effort any such change brings with it, of course.
>>
>> You can always enable this support on MX8MP EVK, it has M2 slot and this
>> would add build coverage of this code too, without impacting Verdin.
> 
> I would have chosen that as the base platform to enable but
> unfortunately I don't have that at my desk. However, if someone is
> willing to test this patch-set on MX8MP EVK then I am happy to extend
> corresponding defconfig too.

I think we can surely find a platform on which this can be enabled by 
default and tested by people.