00:59 <javierm> did anyone else have issues with the msm/drm driver failing to probe if there are a lot of stuff build as a module unless the default deferred_probe_timeout is increased?

01:00 <javierm> I had to add deferred_probe_timeout=30 to my cmdline

02:59 <steev> javierm: on what kernel

03:49 <travmurav[m]> hexdump0815: aceridus re: memory regions - I got an impression that efi shim injects the uefi reserved regions into the system memory map which kinda makes sure os doesn't access blocked regions

03:49 <travmurav[m]> but in my case also blocks like 400mb of otherwise useable ram as it seems windows can reclaim the efi reserved range for the pil loader :/

03:54 <travmurav[m]> aceridus: did you get the modem up?

03:59 <travmurav[m]> so in my case iirc I needed &remoteproc_mpss (Modem, your firmware is probably non "_nm", iirc book go always has lte), then &wifi; memory regions for them.

04:01 <travmurav[m]> after that need to have rmtfs in userspace to kick the modem on, then pd-mapper and tqftpserv so it can download the firmware from the disk.

05:53 <steev> bamse: i take it you haven't had a chance to circle back to the pmic_glink stuff?

05:53 <steev> i don't see it in the kernel and i don't see it on the list anymore

14:39 <leezu> javierm: Regarding the clock issue, Abhinav pointed out https://gitlab.freedesktop.org/drm/msm/-/issues/13 posted by robclark half a year ago. However, it's different in that prior to 6.1 the issue only happens with PANEL_EDP=m but now it happens even with PANEL_EDP=y. On your system, did you observe the issue even prior to 6.1 and with PANEL_EDP=y?

14:43 <Leandro[m]> Got the money back

16:54 <leezu> robclark: Besides the clock issue in 6.1, I notice there are additional regressions in 6.0 (compared to 5.19). At least in 6.0.6, 6.0.7, 6.0.8 (haven't tried others yet) the adreno gpu frequently hangs and fails to initialize correctly. Have you seen these issues before?

16:54 <leezu> [ 13.793781] adreno 5000000.gpu: Adding to iommu group 6

16:54 <leezu> [ 13.820328] adreno 5000000.gpu: [drm:a6xx_gpu_init] *ERROR* failed to read speed-bin (-517). Some OPPs may not be supported by hardware

16:54 <leezu> [ 13.834068] adreno 5000000.gpu: supply vdd not found, using dummy regulator

16:54 <leezu> [ 13.841394] adreno 5000000.gpu: supply vddcx not found, using dummy regulator

16:56 <robclark> dummy regulator msgs can be ignored, that code should probably be removed (it dates back to before gdsc was modeled as genpd.. reading speed-bin probably depends on qfprom? Or whatever reads the efuses

16:58 <robclark> oh, but -517 is -EPROBE_DEFER

17:00 <robclark> probably we want something like:

17:01 <robclark> https://www.irccloud.com/pastebin/VdmxSU3X/

17:01 <robclark> not sure why you see that, maybe drm=y but whatever is needed to read speedbin is =m?

17:02 <robclark> just a guess

17:16 <hexdump0815> Leandro[m]: congratulations - time to get a real galaxy book go now :)

18:16 <leezu> robclark: I see. Thank you. So I should review 'git diff v5.19 v6.0.8 -- arch/arm64/configs/defconfig ' for newly introduced options? Given what you say, for example +CONFIG_INTERCONNECT_QCOM_SC7180=m and +CONFIG_VIDEO_QCOM_VENUS=m look suspicious.

18:19 <leezu> I already had CONFIG_INTERCONNECT_QCOM_SC7180=y https://gitlab.freedesktop.org/drm/msm/uploads/5583d2c7ebaf193240643ac7eb10dd53/config

18:20 <robclark> I guess I'd be more concerned about kconfig changes that didn't have matching defconfig update, and or new reasons why something nvmem related would probe-defer.. tbh I normally use `make olddefconfig` rather than starting with fresh defconfig when moving btwn kernel versions

18:20 <robclark> not sure if that is why I didn't see that problem on 6.0-rc or 6.1-rc

18:22 <robclark> given that you are somehow hitting that path, could you try the diff I pasted and confirm that *eventually* the gpu initializes? I think that is the right thing (we would rather not silently miss out on the higher OPP on 7c gen2 things), but would need to figure out how to trigger that path

18:23 <leezu> I've also started with the working 5.19 config and added CONFIG_NVMEM_QCOM_QFPROM and CONFIG_USB_ONBOARD_HUB per javierm's advice

18:24 <leezu> Yes, the GPU eventually initializes but is subject to frequent freezes. Let me share the full dmesg

18:24 <leezu> https://termbin.com/0r8w dmesg for 6.0.8

18:24 <leezu> [ 59.746137] msm_dpu ae01000.display-controller: [drm:hangcheck_handler] *ERROR* A618: hangcheck detected gpu lockup rb 0!

18:24 <leezu> [ 59.757419] msm_dpu ae01000.display-controller: [drm:hangcheck_handler] *ERROR* A618: completed fence: 4294967048

18:24 <leezu> [ 59.768321] msm_dpu ae01000.display-controller: [drm:hangcheck_handler] *ERROR* A618: submitted fence: 4294967049

18:24 <leezu> [ 59.779243] msm_dpu ae01000.display-controller: [drm:recover_worker] *ERROR* A618: hangcheck recover!

18:24 <leezu> [ 59.788747] msm_dpu ae01000.display-controller: [drm:recover_worker] *ERROR* A618: offending task: gnome-shel:sq0 (/usr/bin/gnome-shell)

18:27 <leezu> robclark: Given you don't hit the issue, would you mind sharing your config?

18:29 <robclark> https://paste.centos.org/view/50b94fbe is my non-debug config for 6.1-rc

18:30 <robclark> [ 14.952127] [drm:adreno_idle] *ERROR* A618: timeout waiting to drain ringbuffer 0 rptr/wptr = 0/C

18:30 <robclark> gpu isn't powering up correctly

20:36 <steev> hm, i can't even get next to boot here on 8280, can't tell what is going on, it's going poof as soon as leaving efi services

20:45 <steev> ardb: your efi changes are in -next right? so that efi=novamap isn't required?

20:50 <swgws> steev: do you have a working defconfig for 6.0 on 8280?

20:51 <steev> The laptop_defcon fit in my repo should?

20:51 <steev> Thanks phone

20:51 <steev> laptop_defconfig

20:52 <swgws> thanks

20:53 <steev> The 6.0.7 tag should work, I still need to push out 6.0.8, but I need to double check the patch set

20:54 <steev> Lenovo 6.0.7 tag* to be specific, because I was stupid and did a --all push to it on accident and now have just about every tag that ever existed

20:56 <Leandro[m]> Lol

21:00 <steev> 5,823 tags :(

21:01 <steev> lenovo-x13s-6.0.7 is the tag i mean

21:24 <leezu> robclark: thank you. Your config indeed avoids the msm and display issues on 6.1-rc4. Wifi still does not work. Does it work for you?

22:29 <steev> javierm: actually, with 6.1, i do see that here too, needing to increase the timeout

22:33 <steev> javierm: 10 works for me, don't need to bump to 30

22:35 <steev> although, the mhi device is being finnicky

23:14 <robclark> leezu: never tried wifi, no haven't checked what fw and userspace bits I'd need to make wifi work.. but wifi isn't so much needed for gpu dev

23:14 <robclark> (well, kinda is for some android games that think if you aren't on wifi you are on LTE, but for those I have to use CrOS image anyways)