RE: [PATCH 06/16] arm64: dts: exynos: Rename the term elbi to appl

From: Shradha Todi
Date: Fri Mar 03 2023 - 01:25:42 EST




> -----Original Message-----
> From: Krzysztof Kozlowski [mailto:krzysztof.kozlowski@xxxxxxxxxx]
> Sent: 16 February 2023 16:34
> To: Shradha Todi <shradha.t@xxxxxxxxxxx>; lpieralisi@xxxxxxxxxx;
> kw@xxxxxxxxx; robh@xxxxxxxxxx; bhelgaas@xxxxxxxxxx;
> krzysztof.kozlowski+dt@xxxxxxxxxx; alim.akhtar@xxxxxxxxxxx;
> jingoohan1@xxxxxxxxx; Sergey.Semin@xxxxxxxxxxxxxxxxxxxx;
> lukas.bulwahn@xxxxxxxxx; hongxing.zhu@xxxxxxx; tglx@xxxxxxxxxxxxx;
> m.szyprowski@xxxxxxxxxxx; jh80.chung@xxxxxxxxxx;
> pankaj.dubey@xxxxxxxxxxx
> Cc: linux-pci@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-arm-
> kernel@xxxxxxxxxxxxxxxxxxx; linux-samsung-soc@xxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 06/16] arm64: dts: exynos: Rename the term elbi to appl
>
> On 14/02/2023 13:13, Shradha Todi wrote:
> > DT uses the name elbi in reg-names for application logic registers
> > which is a wrong nomenclature. This patch fixes the same.
> >
> > This commit shouldn't be applied without changes
> > "dt-bindings: PCI: Rename the term elbi to appl" and
> > "PCI: samsung: Rename the term elbi to appl"
>
> Dependencies and patch ordering goes after '---', because there is no point
> to store it in git history.
>

Understood will take care in next set of patches.

> Anyway, that's an ABI break and Exynos5433 is quite stable, so without clear
> indication of fixed bug, we should not do this.
>

We have strong technical reason to do so.

As per DWC PCIe UM, ELBI delivers an inbound register RD/WR received by the controller to external application registers when the controller
is expected to generate the PCIe completion of this register RD/WR.
In this driver register space which is currently marked as ELBI, is not used for this purpose (Not sure why original author has named this set of registers as ELBI)
So to keep this technically correct, it should be marked as application specific wrapper register space.
We used name as "appl" taking reference from intel-gw-pcie.yaml's similar register space named as "app", whereas in nvidia,tegra194-pcie.yaml it's named "appl".

So our argument is if a future Samsung manufactured SoC having DWC PCIe controller comes with support of real ELBI interface, we need to use the name elbi.
We know such SoC exists but they are not yet upstreamed.

Ready to adopt the best possible suggested method to make this happen but I really think the name ELBI is misleading.

>
> Best regards,
> Krzysztof