13:52 tagr: pabs: tegra114-roth is the old SHIELD controller, not the actual SHIELD Tablet
13:52 tagr: I don't think we ever upstreamed support for that, though the DTS files might be available somewhere still
13:53 tagr: I think acourbot had worked on it at some point and I very vaguely recall looking at getting the mini-HDMI to work (and fail at it)
13:55 tagr: my recollection is that you had to have some fairly elaborate hardware trickery to get some sort of debugging UART out of it and acourbot might have been one of the very few people ever booting mainline Linux on it (I don't think I ever had the necessary UART hardware and tried to get it to work mostly blindly because I don't think the display panel was working)
13:55 tagr: unfortunately my ST8 broke a couple of years back, so I don't have access to that hardware anymore
14:02 tagr: pabs: looks like acourbot (actually, I think he goes/went by gnurou on IRC) no longer has that ST8 branch that he used to keep
14:02 tagr: there's a reference to it here: https://github.com/NVIDIA/tegra-nouveau-rootfs/issues/22 but it 404's
14:10 tagr: pabs: hah! I found an old copy of that branch on a local repo (I knew that keeping archives of all this would come in handy at some point): https://github.com/thierryreding/linux/tree/st8
14:12 tagr: looking at the ST8 DTSI file in there it might actually still apply correctly and looks minimal enough to still work on a recent linux-next
14:12 tagr: and you might even be able to get very basic display support via simplefb
15:31 karolherbst: tagr: any idea what 2D_LAYER_EXCEEDS_DEPTH means?
15:31 karolherbst: I think that term was taken from nvgpu
15:32 karolherbst: ehh
15:32 karolherbst: open gpu doc actually
15:33 karolherbst: https://github.com/NVIDIA/open-gpu-doc/blob/master/manuals/ampere/ga102/dev_ce.ref.txt#L206
15:34 karolherbst: doing a blocklinear copy...
15:34 karolherbst: ehhhhh
15:34 karolherbst: no, that should be fine
15:37 karolherbst: it works with pitch...
15:37 karolherbst: annoying
16:14 tagr: can't say that I've heard of that one
16:20 tagr: karolherbst: but yeah, sounds like an error you'd get when you try to do an pitch-linear only operation on a blocklinear buffer
16:41 karolherbst: tagr: ohh, that could explain it
16:41 karolherbst: but in gallium we always use blocklinear
16:41 karolherbst: maybe there is state I never set
16:42 karolherbst: and the hw just uses fields never set and throws that error
16:42 karolherbst: I might trying to set more stuff
18:31 Wally: tagr: So there is need to support older kernels
19:28 Wally: Are GEM memory allocations blocklinear? Are they compressed?