06:35 <mwalle> marex: yes, i've enabled devcoredump on ls1028a

06:37 <mwalle> marex: but the fault I'm seeing is definitely from the smmu

09:04 <mwalle> austriancoder: have fun :(

09:04 <mwalle> :)

11:03 <mwalle> marex: btw I still get "[ 45.424363] DMA-API: cacheline tracking EEXIST, overlapping mappings aren't supported". haven't digged into that, do you have CONFIG_DMA_API_DEBUG enabled?

11:10 <mwalle> with the smmu handling fixed, I don't get any MMU faults anymore (with a simple test). I'll now let glmark run for some time though

12:34 <marex> mwalle: I run a bit more involved test, running stress-ng --cpu in the background and iperf3 in the background, and then glmark

12:34 <marex> it triggers rather seldom

12:35 <marex> the iommuv2 etnaviv code looks real weird thought

12:35 <marex> *ehough

12:35 <marex> *though

13:05 <mwalle> marex: what are your arguments for iperf3?

13:06 <mwalle> *parameters

14:38 <sravn> austriancoder: you are welcome for a cup of coffe or a pint of beer now you are almost visiting me. Lives in Holstebro in Denmark :-)

16:51 <mwalle> marex: interesting, just noticed this: https://pastebin.com/raw/X3q3Fada

16:52 <mwalle> that SMMU fault with "--off-screen" still seems to be a bug, but during the next run I'm getting a MMU fault..

16:53 <mwalle> mhh actually it doesn't go away anymore

16:58 <mwalle> and it happens when i start glmark2, stop it and start it again (or sometimes I have to do it three or four times)