10:31 <marex> lynxeye: so it seems that whacking that SRC reset indeed always corrects the gpu2d behavior

10:32 <marex> lynxeye: I tried doing that in TFA, but it seems to be necessary after every PGC cycle, otherwise the GPU2D is starting in undefined state

11:00 <lynxeye> marex: fun... so we need to couple 2d and 3d gpu domains, so they can assert their reset together. :(

11:00 <marex> lynxeye: unless you have a better idea ?

11:03 <lynxeye> marex: The irony is that the GPU section of the reference manual is super proud of the GPU being able to be gated individually ;)

11:04 <lynxeye> marex: No, no better ideas at this point. Maybe we could do the soft reset in etnaviv earlier to get the GPU into a defined state, but I'm not sure if this is sufficient...

11:17 <marex> lynxeye: I dont think thats the problem

11:17 <marex> lynxeye: I think it's the AXI bus that flakes out, hence the SRC reset is necessary

11:18 <marex> lynxeye: or AHB bus as well (?)

12:30 <lynxeye> marex: Yea, most likely the AXI master interface is doing funky stuff, which could not be fixed with the soft reset