On 25/11/2022 15:57, Sam Protsenko wrote: > On Fri, 25 Nov 2022 at 08:47, Krzysztof Kozlowski > <krzysztof.kozlowski@xxxxxxxxxx> wrote: >> >> On 25/11/2022 15:22, Sam Protsenko wrote: >>> On Fri, 25 Nov 2022 at 05:22, Krzysztof Kozlowski >>> <krzysztof.kozlowski@xxxxxxxxxx> wrote: >>>> >>>> Exynos5433 has several different SYSREGs, so use dedicated compatibles >>>> for them. >>>> >>>> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> >>>> >>>> --- >>>> >>>> Cc: Sriranjani P <sriranjani.p@xxxxxxxxxxx> >>>> Cc: Chanho Park <chanho61.park@xxxxxxxxxxx> >>>> Cc: Sam Protsenko <semen.protsenko@xxxxxxxxxx> >>>> --- >>> >>> Hi Krzysztof, >>> >>> Just curious: what is the rationale for adding those more specific >>> sysregs? AFAIR, e.g. in Exynos850, different SysReg instances have >>> pretty much the same register layout. >>> >> >> On Exynos5433 all these blocks have different registers. Are you saying >> that Exynos850 has four (or more) sysregs which are exactly the same? >> Same registers? Why would they duplicate it? >> > > Ah, no, you are right. Just checked it, they are different. Just first > couple of registers are similar between blocks, that's why I memorized > it wrong. > > So as I understand, adding those new compatibles follows "describe HW, > not a driver" rule? Because AFAIU, right now it'll fallback to > "syscon" compatible anyway. Yes, they describe hardware. Of course all of these sysregs are similar as they are just bunch of SFR/MMIO-region, but they have different roles/features. For example some other devices (users) of syscon/sysreg should reference specific device, not any sysreg. On several other architectures we use specific compatibles, so I think for Samsung we should do the same. Different case was for Exynos 3/4/5 where there was only one SYSREG. Best regards, Krzysztof
