19:12 <bl4ckb0ne> do I need to do something in particular to get `viv_interpose.so`? following libvivhook's README gives `viv_interpose.la` only

19:36 <austriancoder> bl4ckb0ne: it should work as written in the readme

19:36 <bl4ckb0ne> might be my cross compile setup hm

19:36 <austriancoder> bl4ckb0ne: btw do you also see lot of "swiotlb buffer is full" messages around mmu faults?

19:37 <bl4ckb0ne> nope

19:37 <austriancoder> bl4ckb0ne: mind sharing your dts?

19:37 <bl4ckb0ne> i see a log message about gpu idle with 6.1 around the hang

19:38 <austriancoder> bl4ckb0ne: for me kmscube seems to work fine (with vkms)

19:38 <bl4ckb0ne> havent tried kmscube yet, will give a go

19:38 <bl4ckb0ne> i dont think I can share the dts sorry

19:39 <austriancoder> bl4ckb0ne: anything special in the dts about reserved memory .. maybe GPU related?

19:39 <bl4ckb0ne> i can take a look

19:43 <ManMower> gpu_reserved: gpu_reserved@0x8800000000 { no-map; reg = <0x8 0x80000000 0 0x20000000>;

19:43 <ManMower> };

19:43 <ManMower> (answering on behalf of bl4ckb0ne with a really sketchy cut and paste...)

19:44 <ManMower> there's a linux,cma section as well, is that relevant?

19:56 <austriancoder> DDR on the imx8qm starts at 0x880000000 and there are some mmu limitations: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpu/drm/etnaviv/etnaviv_drv.c?h=v6.9-rc7#n623

19:57 <austriancoder> I am hitting this https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpu/drm/etnaviv/etnaviv_gpu.c?h=v6.9-rc7#n835 but my dma32 zone is empty

19:58 <austriancoder> looks like there is an other ddr addres in the reference manual: 0x80000000

20:01 <ManMower> is our dts wrong/bad?

20:17 <austriancoder> That gpu_reserved label is used in galore driver (in downstream kernel). etnaviv does nothing with it

20:19 <austriancoder> I know that the userspace part runs without problems on that SoC. So I would focus on this gpu_reserved label.

21:41 <bl4ckb0ne> austriancoder: lynxeye told us that the userspace might not set an address register properly and that could cause the mmu fault

21:47 <austriancoder> bl4ckb0ne: that's the generic answer why there is an mmu fault with address 0x0. Look at the generated cmd stream and search for a 0x0.

21:49 <austriancoder> bl4ckb0ne: kmscube works for you?

21:50 <bl4ckb0ne> havent tried, cooking dinner

21:50 <bl4ckb0ne> will do later

21:50 <austriancoder> bl4ckb0ne: k..

22:03 <austriancoder> bl4ckb0ne: from a hwdb view I see no reason why the cmd stream should be that different to the other gc7000.

22:03 <austriancoder> Bed time here..

22:05 <bl4ckb0ne> have a good one, ill update the chan tonight