00:00 <colemickens> I feel like I've seen some notes about black display but I'm not sure where that ended up

00:07 <colemickens> dang, I was so sure this was just going to work too

00:07 <vk6[m]> what happens during this "screen flash?"

00:08 <colemickens> nothing, the messages about loading the device tree and then ... exiting boot services

00:08 <colemickens> that's in the small uefi console fb in the middle of the screen

00:08 <colemickens> adn then there's a whole screen flash, presumably as the initrd/some driver loads

00:09 <colemickens> I actually wonder if i'm too over zealous with my initrd, as I've included all of the various drivers so they load fairly early.

00:10 <colemickens> hm, do I need a custom console= ... it is arm after all

00:11 <vk6[m]> i believe i'm experiencing something similar with my attempts

00:11 <vk6[m]> for me at least, the kernel start loading fine and i can see the fbcon and boot logo for a bit, then the screen goes black

00:12 <vk6[m]> my hypothesis so far is that i was missing the firmware for the gpu

00:12 <colemickens> ah hm, I dont even see boot logo

00:12 <colemickens> you mean the efi/vendor one?

00:12 <colemickens> I guess I don't think nixos is configured to really show that. It just does a verbose normal boot

00:12 <vk6[m]> i mean the line of penguins at the top left of the screen

00:29 <robclark> colemickens: I believe you still need at least some of `clk_ignore_unused pd_ignore_unused arm64.nopauth efi=novamap`

00:29 <robclark> fwiw display should come up without gpu fw

00:30 <robclark> drm/msm just tries to load gpu fw each time the device is opened until it succeeds (to avoid needing fw in initrd)

00:38 <colemickens> I have fw too, from .sera

00:38 <colemickens> I'll get 'er back out and try tagging those on system-boot on the fly

00:42 <colemickens> oh hey, look at that. I have kernel boot log. It looks like I messed up the nixos image in some other potentially silly way, but there it goes

00:42 <colemickens> I'll have to remake image anyway for kernel params and can fixup the partition label mistake I've made or rwhatever it is

00:56 <colemickens> hm, I kind of suspect I'm missing a qcom/usb initrd module

00:56 <colemickens> since I'm usb booting

00:56 <vk6[m]> i managed to get to the sddm login screen

00:57 <vk6[m]> but the keyboard and trackpad don't work

00:59 <colemickens> oh I hope it wasn't just waiting on repart :/

01:00 <FarchordSteveCossette[m]> colemickens: On the vivobook, I know you gotta use USB-C. USB-A is just a no dice.

01:01 <colemickens> this device only has em, and I'm not using an adapter, so hopefully good there

01:01 <FarchordSteveCossette[m]> yeah if I want to use a USB-A drive, I need a USB-C hub lol

01:02 <colemickens> ugh, got a message about a missing fw file that I know is in there

01:03 <colemickens> does qcom still have issues with loading compressed fw?

01:04 <colemickens> yeah, the boot log is telling me it's waiting for a /dev/disk/by-label/nixos to appear and sure enough on my computer its there :(

01:19 <robclark> loading compressed fw is a kconfig thing.. nothing arch specific

01:20 <robclark> I'm using compressed modules and fw here, fwiw

01:21 <adamcstephens> i don't think pdmapper supports zstd compression, which nixos uses, but otherwise compressed firmware should be fine

01:50 <robclark> ok, so userspace pd-mapper needs the json things to not be compressed (but I assume kernel pd-mapper doesn't have the same limitation)

01:51 <robclark> I've not been considering userspace pd-mapper since it is getting replaced with kernel pd-mapper

01:53 <steev> i don't know because i don't do compressed firmware (yet)

01:53 <steev> maybe once debian starts doing it, i will

01:57 <vk6[m]> so i copied the extra firmware files (from https://github.com/Seraphin-/linux-firmware-x1e80100-vivobook) in an attempt to get wifi and the keyboard working, but now it's crashing during the systemd startup because somehow the usb drive is getting disconnected

01:58 <vk6[m]> * (from https://github.com/Seraphin-/linux-firmware-x1e80100-vivobook) to /lib/firmware in an

01:59 <vk6[m]> i'm not really sure what's going on here

02:00 <robclark> so.. I've got rootfs on nvme which might be side stepping some issue with power getting cut to usb at some point in boot (although I've not had problems with usb boot to recover from botched kernels).. clk_ignore_unused pd_ignore_unused might be related

02:01 <robclark> re kb/touchpad input I've had to manually modprobe evdev... no idea why

02:01 <robclark> I have this /etc/modules-load.d/robs-hacks.conf :

02:01 <robclark> https://www.irccloud.com/pastebin/SCTmVPdL/

02:02 <steev> wasn't there a similar issue on x13s when some firmware or other was in the initramfs?

02:02 <vk6[m]> i guess i'll also try booting from nvme to get around that usb power issue

02:03 <robclark> pretty sure you want 'clk_ignore_unused pd_ignore_unused' at this point either way

02:03 <steev> you do

02:03 <vk6[m]> i do have those kernel args already

02:04 <robclark> steev: I moved some things that need fw (like adsp/cdsp) to modules to avoid having to get fw into initrd.. but then I need the modules-load.d trick to get them to load

02:05 <robclark> probably some missing distro things

02:05 <robclark> but I'm just a gpu driver dev so no idea how those things are _supposed_ to work

03:09 <steev> ema: have you tried out the 6.10 kernel at all? my udev rule for the wifi mac address isn't working anymore?

03:09 <steev> ohhh because ip was moved from usr/sbin to usr/bin

09:56 <pstef> hmm, I switched from Fedora Rawhide to Debian Trixie and it's generally nice, but I can't "Enable Bluetooth" under KDE/Plasma (even with "experimental" kernel 6.10mumble)

11:19 <mahmoudajawad[m]> Do we have people running Snapdragon X here? What devices?

11:36 <FarchordSteveCossette[m]> <mahmoudajawad[m]> "Do we have people running..." <- Mostly the Lenovo yoga and the thinkpad I believe. There's also a couple Asus Vivobooks, like myself

11:37 <FarchordSteveCossette[m]> https://gitlab.com/cki-project/kernel-ark/-/merge_requests/3289 Seems like I might start maybe perhaps get a bit of an easier fun at this....

12:38 <mahmoudajawad[m]> <FarchordSteveCossette[m]> "Mostly the Lenovo yoga and the..." <- Awesome! Thanks for replying.

12:39 <mahmoudajawad[m]> Anyone here compiling list of reports on distros, what is working what not, installation steps on these devices?

12:41 <FarchordSteveCossette[m]> mahmoudajawad[m]: Yeah that's been asked before. Info is spread a bit everywhere right now. To be fair, I don't think anything is 100% working, all laptops are in various states

12:42 <FarchordSteveCossette[m]> There's quite a bit of file ripping being needed from Windows right now too

12:42 <FarchordSteveCossette[m]> Once I get my Fedora install running (Sometimes in the next century lol) I'll make a guide

16:10 <steev> pstef: did you do the bluetooth mac addr fix?

16:12 <pstef> steev: I know I tried to, but maybe it was the wrong fix

16:13 <steev> https://github.com/jhovold/linux/wiki/X13s#bluetooth

16:14 <pstef> that helped, so I tried something weird

16:14 <pstef> steev: thanks

16:30 <steev> pstef: fwiw, i followed ema's excellent instructions in https://wiki.debian.org/InstallingDebianOn/Thinkpad/X13s#Wi-Fi_and_Bluetooth

16:31 <steev> the only thing in trixie/sid is that the wifi part of that rule should be /usr/bin/ip now

16:32 <steev> i only noticed the change because i started getting a notification of a new wifi device every time i rebooted it

16:32 <pstef> I tried to do exactly that but failed at reading

16:33 <steev> :D

16:34 <pstef> from a quick glance I was under the impression that the udev part is only for wifi and I never had any problems with it, so I skipped to the systemd unit part... which turns out to depend on the udev part

16:41 <steev> ahhh :D

16:46 <steev> hmm, so debian starting with kernel 6.10 is gonna offer a 16K pagesize kernel it looks like

16:46 <steev> for me, most things work, but zed, for some reason just immediately dies saying it's out of memory

17:35 <dgilmore> fedora has 16k and 64k configs but is not currently building images. I believe that the Ashai remix is building a 16k kernel for use on mac's. RHEL offers 4k and 64k builds

17:56 <FarchordSteveCossette[m]> <dgilmore> "fedora has 16k and 64k configs..." <- I have a really annoying problem with the fedora kernel

17:57 <FarchordSteveCossette[m]> Imagine I download a fresh .raw.xz image of workstation, and I put it on the drive. The image will boot. But imagine I uninstall that kernel (kernel-install remove <name> and reinstall it (depmod -a, kernel-install, fix the boot entry) with that exact same kernel, it no longer boots

17:57 <FarchordSteveCossette[m]> I have no idea why

18:01 <dgilmore> There is many reasons that may happen. But the usecase is not something people typically do

18:29 <FarchordSteveCossette[m]> fair enough XD

19:12 <HdkR> Asahi's kernel also requires 16k pages. So there's no getting away from it there :P

19:36 <steev> you can change it to 4K, just iirc, marcan didn't want to as he wanted to force people to fix their stuff, since it should work (rightly so imo)

19:56 <robclark> something other than 4k is going to be problematic for fex if you want windows games (otherwise I tend to think apple was right with going for larger page size.. it is kinda one of the secret perf advantages that apple has)

20:10 <steev> robclark: as far as i can tell, on the linux side of things, aside from this weird crash saying it's out of host memory, most things work fine (i'm not much of a windows gamer, or gamer in general)

20:13 <JensGlathe[m]> 🤯 video calls are working on element, I'm not safe anymore

20:13 <FarchordSteveCossette[m]> lol

20:46 <robclark> steev: I don't expect windows games or gpu to actually work completely with 4k pages, I'm just interested to compare cpu benchmarks with mX 4k pages vs mX 16k pages vs x1e 4k vs x1e 16k

20:48 <robclark> I know mX gpu mmu and iommu's don't really work with 4k pg size, but that shouldn't matter for cpu benchmarks

21:10 <colemickens> [@_oftc_robclark:matrix.org](https://matrix.to/#/@_oftc_robclark:matrix.org) So I confirmed my image was correct, added the kernel params, enabled emergency console and sure enough, I don't see my USB storage device showing up. Nothing under dev block except the nvme

21:10 <colemickens> If you have any more ideas or someone else I can bug, please let me know

21:52 <HdkR> robclark: Yea, the dart being limited to 16k page size makes anything beyond CPU loads spicy on Apple Silicon. Hacking around the limitations is a lot worse than just using a 4k page size vm :D

21:52 <HdkR> Just need to fix all the issues with the VM now, which is a lot easier

21:54 <robclark> Yeah, like I said I was only interested in CPU benchmarks for that

21:54 <robclark> I want to know what the pagesize uplift is

21:55 <HdkR> Pretty sure they did some benchmarking initially

21:56 <HdkR> There was some mention of a 20% perf hit going from 16k to 4k, but it must have been a pretty spicy bench to hit that number

22:18 <jannau> robclark: iirc I saw a 15% perf uplift for kernel compilation with 16k pages

22:39 <steev> https://browser.geekbench.com/v6/cpu/compare/7086459?baseline=7086686 this is what i see on the x13s - 6459 is the 4K pagesize, 6686 is 16K

22:49 <gabertron> anybody managed to get audio working on the x1 laptops so far? On my t14s and got widevine working and can see the Olympics but can’t hear anything :) was hoping I could look at somebody’s tplg settings for the yoga or something for reference

22:52 <HdkR> steev: 3.5% uplift is more in line with expectations

22:52 <steev> gabertron: the .conf files are the source for the tplg file

22:54 <steev> https://git.codelinaro.org/srinivas.kandagatla/audioreach-topology

22:54 <steev> i suppose the .m4 is the actual source?

23:08 <gabertron> Cool will take a look