17:43 <aka_[m]> just another day hoping to catch sboyd and be able to ask question

19:12 <lumag> konradybcio, Marijn[m]: I noticed that the adreno-smmu is not used for the sdm630/sdm660 mmss_smmu. Is it just for the same reason as we don't use ttbr1 for msm8996, the lack of separate pagetables?

19:12 <lumag> or is it some hardware issue?

19:12 <lumag> (or firmware)

19:14 <konradybcio> it's been too long to remember

19:14 <konradybcio> does it work differently if you add adreno-smmu?

19:14 <konradybcio> (with all of our downstream patches ofc..)

19:15 <lumag> konradybcio, no hardware at hand :-(

19:15 <aka_[m]> with adreno smmu it should kick with these tables and not sure if they work on a5xx stuff, not on a506 for sure

19:16 <konradybcio> I think Angelo did some black magic to make 8998/a540 go faster with some iommu optimizations

19:16 <konradybcio> but you'd have to ping him directly

19:19 <lumag> konradybcio, ok, I

19:19 <lumag> I'll just document it as is and let somebody else check the issue

19:50 <sboyd> aka_[m]: yo! I have not worked at qcom for many years. I'd ask bamse

20:22 <lumag> kholk[m], konradybcio: another iommu question. you've added sdm630-v2/msm8998-v2 to the qcom_smmu_impl table. However that means that these devices are now getting all qcom-500 specifics, including the arm-mmu500 reset. Is this expected or not?

20:55 <konradybcio> I think it got there by mistake, as the quirks did get there in the end

20:55 <konradybcio> though it of course should be there, provided the aforementioned quirks also get impl