<robclark>
if this is all in the context of android, maybe it is just better to extend what jstultz started w/ compile option to disable UBWC.. he did that more specifically for nv12, but I guess we could do a similar thing to just disable UBWC across the board
<robclark>
minigbm kinda has the same problem as gralloc, in that it isn't a good support-any-hw-sw-combo solution
<Marijn[m]>
I'm fairly certain plane properties can be queried without kms master
<Marijn[m]>
In this case all I really need is for it to check if the presentation hardware supports the QCOM_COMPRESSED modifier... Seems like a simple and reasonable and generic thing to do... Assuming this buffer was allocated as SCANOUT though, have to see what usage hints are used within gralloc
<Marijn[m]>
As for mesa just refusing to use ubwc-modifier buffers when support for it is not implemented should be enough for now
<robclark>
file a gitlab issue if mesa is just silently ignoring the QCOM_COMPRESSED on a5xx, so I remember to look at that
<robclark>
IME we did have some issues going from sdm845 to sc7180 which were due to incorrect UBWC register settings on kernel side for both dpu and gpu side of things, so it is likely to be the sort of thing that needs to be validated on each SoC, even once GPU and mdp5 support it