10:23 <danielt> Can anyone offer some clues on the getting the GPU running on Thinkpad X13s. I've seen the success photos but can't replicate them. Initially I got the black screen discussed yesterday... but I switched to a less modular kernel (based on johan_defconfig). That does at least boot but I get messages saying it couldn't power up the GPU. https://gist.github.com/daniel-thompson/4bb12ac567152f63095086df9eccbd71

10:24 <danielt> If anyone can compare my logs to known working (or just share some known working info about kernel config or firmware) that would be awesome!

11:59 <juergh[m]> danielt: maybe this?

11:59 <juergh[m]> [ 16.956974] panel-simple-dp-aux aux-aea0000.displayport-controller: Couldn't identify panel via EDID

11:59 <juergh[m]> [ 17.219386] panel-simple-dp-aux aux-aea0000.displayport-controller: Unknown panel BOE 0x0a84, using conservative timings

12:01 <juergh[m]> jhovold: still puzzled. I can't seem to replicate the black screen anymore. different question: how can I prevent the early console switch so that the initramfs console is actually usable?

12:16 <jhovold> juergh[m]: you need to include the display deps in the initramfs, see the commit adding johan_defconfig for details

12:16 <jhovold> this is specifically needed to prevent regulator core from disabling seemingly unused regulators after 30 s

12:18 <jhovold> juergh[m]: that "Unknown panel" warning should be benign, but you can add the panel timings to the driver to suppress the warning

12:21 <jhovold> danielt: bamse found another bug related to the msm drm driver late component bind failure ("probe deferral") which you may need when enabling the GPU unless you make sure to load (and bind) the panel driver before the drm driver

12:21 <jhovold> https://lore.kernel.org/linux-arm-msm/20230203181245.3523937-1-quic_bjorande@quicinc.com/

12:21 <danielt> juergh: Could be (although I think that's probably just the USB-C dock failing to come up on the boot I captured... and it also fails for me without the dock plugged in). Overall I don't know what "good" boot with GPU looks like so I'm not sure what to compare against. If you have one you are willing to share that would be useful.

12:23 <danielt> jhovold: Thanks. I'll take a look at that.

12:38 <juergh[m]> jhovold: I'm aware of the initrd modules requirement. but my question is is it not possible to prevent the console switch and stay on the fb console? not sure if that make sense, don't know much about graphics.

12:39 <jhovold> juergh[m]: you'd currently need to mark the corresponding regulators as always-on in the DT if you want them to stay on when no driver has claimed them

12:40 <jhovold> there's been some talk about addressing this similar to how we handle interconnects, that is, to keep the regulators on until all drivers are bound

12:41 <jhovold> but there's no such support in mainline currently

12:43 <juergh[m]> ok. thanks for the elaboration.

16:19 <HdkR> danielt: I used https://github.com/steev/linux/commits/sc8280xp-next-20230207-gpu this branch for the GPU working pictures

16:19 <HdkR> And Mesa branch https://gitlab.freedesktop.org/robclark/mesa/-/tree/fd/a690

16:22 <danielt> HdkR: Thanks. I'll take a look over the weekend. After help from jhovold the GPU looks OK in the logs but hasn't come to life to that should help me figure out what I overlooked!

16:22 <danielt> s/life to/life so/

16:30 <steev> if you don't have the mesa stuff, that's likely it

17:09 <bamse> ndec: fwiw, changing bt address dynamically works fine...

17:17 <steev> i suppose we could set one in the dt, or maybe we could do something like the rtc and store it in nvmem?

17:18 <bamse> steev: the default one works fine...until you have more than one device with the default one...

17:18 <steev> i'd rather have one that starts with winstrom's oui at least

17:18 <bamse> steev: it's stored somewhere...

17:25 <steev> i was thinking lenovofprdata but that's probably the fingerprint data... maybe in one of the lenovo* variables, like maybe lenovowolinfo

17:28 <clover[m]> steev: is wayland/mutter still borked even with gpu for external display?

17:30 <steev> yes

17:30 <clover[m]> rip

17:30 <steev> bamse: possibly lBoot0014

17:33 <clover[m]> have you noticed better battery life with gpu?

17:40 <steev> i haven't used it closely enough yet to see

17:44 <steev> i wanna say yes because i went to sleep with it suspended, and woke up with it not dead

17:44 <steev> and it wasn't at 100% when i suspended it

20:42 <steev> huh

20:42 <steev> didn't get the frame reassembly timeout on 20230210 next

20:57 <rfs613> steev: maybe some elves came in overnight and tightened all the screws to manufacturer spec? :P

21:18 <steev> it's probably a fluke of the boot :)