Re: [PATCH v2] arm64: defconfig: Enable TISCI Interrupt Router and Aggregator

[Vaishnav Achath <vaishnav.a@ti.com>]

Hi Geert, Nishanth,

On 05/02/25 09:21, Nishanth Menon wrote:
> On 08:42-20250204, Geert Uytterhoeven wrote:
>> Hi Vaishnav,
>>
>> Thanks for your patch!
>>
>> On Mon, 3 Feb 2025 at 16:08, Vaishnav Achath <vaishnav.a@ti.com> wrote:
>>> Enable TISCI Interrupt Router and Interrupt Aggregator drivers.
>>> These IPs are found in all TI K3 SoCs like J721E, AM62X and is required
>>> for core functionality like DMA, GPIO Interrupts which is necessary
>>> during boot, thus make them built-in.
>>>
>>> bloat-o-meter summary on vmlinux:
>>>          add/remove: 460/1 grow/shrink: 4/0 up/down: 162483/-8 (162475)
>>>          ...
>>>          Total: Before=31615984, After=31778459, chg +0.51%
>>
>> My first thought was: that's a hefty size increase for two rather
>> small drivers!  So I enabled these two options in one of my own kernel
>> configs, and saw a size increase of only 7872 bytes...
>>
>>> Fixes: 2d95ffaecbc2 ("irqchip/ti-sci-intr: Add module build support")
>>> Fixes: b8b26ae398c4 ("irqchip/ti-sci-inta : Add module build support")
>>>
>>> Signed-off-by: Vaishnav Achath <vaishnav.a@ti.com>
>>> ---
>>>
>>> V1 : https://lore.kernel.org/all/20250119182121.3956546-1-vaishnav.a@ti.com/
>>>
>>> V1->V2:
>>>          * Add Fixes tag.
>>>          * Add bloat-o-meter summary to commit message.
>>>
>>> Previously both of these configs were selected for ARCH_K3 in respective
>>> Kconfigs, but commit b8b26ae398c4 ("irqchip/ti-sci-inta : Add module build support")
>>> and commit 2d95ffaecbc2 ("irqchip/ti-sci-intr: Add module build support")
>>> in dropped this recently without adding those back in arm64
>>> defconfig, this causes boot failure in all TI K3 platforms.
>>>
>>> Passing bootlog with configs enabled:
>>> https://gist.github.com/vaishnavachath/03d8d3ce0f22219643c2f4d1dd243fd1
>>>
>>>   arch/arm64/configs/defconfig | 2 ++
>>>   1 file changed, 2 insertions(+)
>>>
>>> diff --git a/arch/arm64/configs/defconfig b/arch/arm64/configs/defconfig
>>> index cb7da4415599..1f25423de383 100644
>>> --- a/arch/arm64/configs/defconfig
>>> +++ b/arch/arm64/configs/defconfig
>>> @@ -1551,6 +1551,8 @@ CONFIG_PWM_VISCONTI=m
>>>   CONFIG_SL28CPLD_INTC=y
>>>   CONFIG_QCOM_PDC=y
>>>   CONFIG_QCOM_MPM=y
>>> +CONFIG_TI_SCI_INTR_IRQCHIP=y
>>> +CONFIG_TI_SCI_INTA_IRQCHIP=y
>>>   CONFIG_RESET_GPIO=m
>>>   CONFIG_RESET_IMX7=y
>>>   CONFIG_RESET_QCOM_AOSS=y
>>
>> However, on arm64 defconfig, this has indeed a much bigger impact,
>> as it also (re)enables several other drivers that depend on the
>> irqchip symbols.
>>
>> You probably want to mention that in the commit message...
> 
> Thanks Geert and I agree. This however just restores the status quo as
> it was in 6.13 tag kernel (and ofcourse the default boot status).
> 
> Will be good to mention that as well - in effect, we are just restoring
> the behavior, not introducing new features.
> 

Thank you for the feedback, I have sent a v3 with updated commit message 
mentioning the reason for size increase and that we are reinstating 
previous behavior, all the dependencies getting re-enabled were already 
part of the arm64 defconfig.

V3 - https://lore.kernel.org/all/20250205062229.3869081-1-vaishnav.a@ti.com/

Thanks and Regards,
Vaishnav

> Side note: We are still working towards fixing up other driver modules
> and some time in near future, hope to transition to initramfs boot as
> default in a single patch.. but till then let us maintain the status
> quo.
>