03:24 <mothenjoyer69> <Dylanger> "Keep in mind at least to my..." <- oh this is disappointing to hear; i wonder if it also extends to the Microsoft Dev Kit?

04:27 <HdkR> mothenjoyer69: Yes

04:27 <HdkR> You'll get virtualization inside Windows but not outside of it.

05:02 <mothenjoyer69> fair enough :(

05:08 <HdkR> Well I couldn't call it fair, but it's the ruthless facts

05:09 <HdkR> At least Jetson gives you EL2 I guess

06:02 <swgws> Hi! I was directed this way as a resource for linux on arm64 - I'm trying to boot linux on the Windows Dev Kit 2023 - which has what appears to be a modified qcom 8cx gen 3 inside. I've tried booting both ACPI and with the devicetree from another 8cx gen 3 device, the lenovo thinkpad x13s. I keep getting to the point where arm-smmu is doing some stuff, then the screen blanks for a few seconds and the device resets. Here's a picture of the final

06:02 <swgws> lines I can get out on efifb: https://user-images.githubusercontent.com/2041026/200714923-db4f830c-b612-4f6e-8596-ac822a544eef.png

06:03 <swgws> Anyone have pointers to something I can look at or ideas on what to try? I've been stripping out things like backlight and HID-over-I2c input nodes from the DTS, but I can't get past the line above in boot

06:06 <steev> you gotta add an override for the smmu

06:06 <steev> is gonna be the big thing

06:11 <swgws> is there somewhere I can read up on that? Or an example where it's been done for a different board?

07:57 <steev> not trying to be evasive, i'm heading to bed, but i believe whitequark logs this channel, you could definitely check there, it's been mentioned within the past month because of someone working on getting the galaxy book go working

07:59 <swgws> I'll take a look, thanks!

11:47 <hexdump0815> swgws: maybe have a look at https://oftc.irclog.whitequark.org/aarch64-laptops/2022-10-18#31534740 and around, especially the following days have some bringup info in them - good luck

13:16 <leezu> laine: as it's a windows laptop, you'll need to follow a similar journey to the ongoing work on the samsung galaxy go https://github.com/aarch64-laptops/debian-cdimage/issues/21#issuecomment-1294284859

14:43 <leezu> javierm, hexdump0815: thank you. I've rebuilt with CONFIG_NVMEM_QCOM_QFPROM and CONFIG_USB_ONBOARD_HUB but the backtraces and also the 'pp3300_hub: disabling" remain. The full config is https://termbin.com/jcz1o and the updated dmesg is https://termbin.com/nzg3. I tried passing clk_ignore_unused, but something is going wrong with the kernel command line parameters.

14:43 <leezu> clk_ignore_unused is missing but a cros_secure paramatere is present that I haven't configured (cros_secure is also present on my currently working 5.19 kernel

14:49 <leezu> robclark: do you have any idea about the drivers/firmware/google/coreboot_table.c:104 memcpy: detected field-spanning write (size 208) of single field "&device->entry" issue on lazor with 6.1-rc4? I see there's no change to the coreboot firmware driver

14:54 <robclark> leezu: some compiler issue or something like that, makes it warn about something it doesn't understand AFAIU? We have a (probably internal) bug about it, not qc specific but AFAIU it is harmless and can be ignored

14:58 <leezu> robclark: Ok, that's good to know. Thank you. For "disp_cc_mdss_mdp_clk status stuck at 'off'" at drivers/clk/qcom/clk-branch.c:91 I suspect it's a regression in the qcom clk driver affecting sc7180? What's your take?

15:05 <ardb> robclark: that sizeof() in the coreboot code looks wrong

15:06 <ardb> hmm or maybe not

15:07 <ardb> in any case, that's just the string functions getting better runtime diagnostics thanks to kees

15:10 <javierm> ardb: wonder if gets confused because ptr_entry is void * instead of struct coreboot_table_entry * ?

15:10 <javierm> maybe just needs a cast when memcpy() ?

15:11 <javierm> or just declare ptr_entry as coreboot_table_entry *

15:11 <ardb> entry is part of a union, and the union is apparently larger

15:11 <ardb> so the target of the memcpy() should be the union itself not one of its members

15:12 <javierm> ardb: ah, that makes sense

15:12 <javierm> or make the union just pointers to the different types and devm managed allocations for these?

15:13 <ardb> that seems like a lot of work for something that is just being misdiagnosed and perfectly safe

15:13 <ardb> (assuming the union's size >= 208)

15:14 <javierm> ardb: right. I meant just to prevent these kind of issues. Since in that case the union will always be sizeof(void *)

15:14 <javierm> but agree that it may be overkill

15:16 <ardb> generally, using memcpy() to assign multiple struct fields at a time is a bit dodgy

15:17 <robclark> leezu: I guess this is w/ drm stuff built as a module? There is a known issue with that, which AFAIU the current thinking is that a fw fix is needed, but I haven't been following it too closely

15:17 <javierm> robclark: yeah, but clk_ignore_unused should be enough to workaround that

15:17 <javierm> and AFAIU leezu is already using that

15:20 <robclark> hmm, I didn't see any new issues like that w/ 6.1 but have been using =y.. maybe abhinav (who doesn't seem to be in this channel) would know

15:22 <javierm> robclark: we have everything built as a module in fedora and with clk_ignore_unused it works for me. Tried to fix the underlaying issue but couldn't due missing docs

15:22 <javierm> the least bad option was to mark a clock as critical

15:23 <javierm> but that's a hack, abhinav said that they are working on a proper solution

16:45 <hexdump0815> leezu: regarding cros_secure in your kernel cmdline - i guess this gets automatically parsed in by the bootloader, but can safely be ignored on a mainliner kernel i guess

16:46 <hexdump0815> leezu: i think it is also possible to add kernel cmdline parameters in the kernel config - this might be a way to get clk_ignore_unused in in case the bootloader is really parsing it away if you add it

16:51 <leezu> javierm: I wasn't able to pass clk_ignore_unused correctly. I'll try the workaround via kernel config hexdump0815 suggested

16:52 <leezu> robclark: All options starting with CONFIG_DRM are =y or unset. Full config at https://termbin.com/jcz1o