00:03 <Jasper[m]> <jglathe_volterra> "https://drive.google.com/file/d/..."; <- huh, looks almost correct?

05:44 <steev> konradybcio: is the point of the lmh on sc8280xp irq just to... eventually have that number change if you can figure out how to really push it?

06:45 <jglathe_sdbox2> Jasper[m], yes, almost. Next test window will be fun

15:28 <jenneron[m]> wcd934x-codec.5.auto: MBHC Over Current on HPHR detected

15:28 <jenneron[m]> what does that even mean

15:28 <jenneron[m]> sounds really bad, but i don't understand how it can happen

15:32 <jenneron[m]> my heaphones are not just a wire between HPHR/L and GND :P

15:33 <jenneron[m]> but the impedance is 16 Ohm, might be too low

19:19 <jglathe__> EL2 news: Volterra is running on 6.8-rc4 on EL2 *with* arm-smmu-v3 active and PCIe functional as it seems. But, no display. VM is running, appears to be stable enough.

19:19 <jglathe__> the journalctl -b: https://drive.google.com/file/d/1hikUhgP2qW8qH0NPCSF72l20Ydvpgd64/view?usp=sharing

19:21 <jglathe__> Great debug work with travmurav[m]

19:26 <Jasper[m]> NICE

19:26 <Jasper[m]> congrats

19:34 <jglathe__> feels a little bleeding edge, without smmuv3 (and pcie) it is stable and usable for VM work.

19:39 <steev> okay, c630 restored... fedora didn't work for me though, was really weird, it only recognized the usb the first 2 times i tried to boot from it, after that it just acted like it wasn't plugged in. i ended up just going with the debian installer from the debiancd repo with the ancient kernel, and just waited for partitioning, then mounted the target, copied the new efi binary in, and used that

19:39 <steev> somehow i was so tired when i finished... i sent that to the debian arm channel instead :/

19:57 <jglathe_volterra> https://github.com/jglathe/linux_ms_dev_kit/tree/jg/el2-iommu-v6.8

20:14 <Jasper[m]> <jglathe_volterra> "https://github.com/jglathe/linux..."; <- @travmurav does this have dtbhack potential?

20:28 <jglathe_volterra> no need afaik. I created a separate dts (in this tree) without the zap shader node. This is still quite unclean, but will be condensed into the device dts as far as possible.

20:29 <jglathe_volterra> I don't use dtbhack to boot, I just load slbounce.efi, exit to grub and on it goes with 6.8.0-rc4-EL2-IOMMU+

20:29 <Jasper[m]> jglathe_volterra: Yeah, but how would you get this upstream?

20:30 <Jasper[m]> Ideally that'd be the end goal, but you can hardly create a ton of dt's that are used depending on the exception level

20:30 <jglathe_volterra> When this works (we're not that far away some work needs to be spent on booting up directly with slbounce, but its still hackery

20:30 <Jasper[m]> Currently those kinds of changes are in dtbhack

20:35 <jglathe_volterra> upstream. my brother in atheist... The Volterra patch is in limbo 'cause Merck Hung doesn't answer konradybcio I guess

20:40 <jglathe_volterra> I have no issue creating a dt for tested machines as long as this is necessary. You could add logic in the kernel for qcom hyp-dependent nodes, and do things for EL2 only (like <linux, EL2=>; IMO would be a good way to make it transparent, but that's a topic for when we've managed loading co-procesor fw and things like these

20:45 <jglathe_volterra> Who would guide these decisions?

20:53 <Segfault[m]> would it not be better for the relevant drivers to automatically detect whether the hypervisor is active and change their behaviour accordingly? i.e. skip trying to load the zap shader if there's no hypervisor?

20:55 <jglathe__> exactly. There are others, though, like arm-smmu-v3, which are reserved nrmally, but have to be loaded and set up in EL2

21:00 <jglathe_x13s> X13s is on 6.8-rc4 now

21:00 <jglathe_x13s> where was the test focus?

23:17 <robclark> jglathe_x13s, Jasper[m]: it would maybe be interesting to be able to have el2 vs el1 conditionals in dtb.. I briefly looked into adding a check into driver to ignore zap node if we are in el2 but needed fxn isn't exported to modules (and we probably want to skip some reserved memory regions, etc.. so that is probably better handled in core rather than driver).. but with sc7180 WoA vs CrOS laptops we have some precedent

23:17 <robclark> for this sort of differentiation, I think

23:23 <dgilmore> steev: I have found that sometimes on the c630 I have to boot windows and delete the efi entries for fedora, usually it happens if I remove the usb drive, when I plug it back in it gets skipped