ChanServ changed the topic of #aarch64-laptops to: Linux support for AArch64 Laptops (Asus NovaGo TP370QL - HP Envy x2 - Lenovo Mixx 630 - Lenovo Yoga C630)
<steev>
jenneron[m]: fwiw... i can't reproduce agl7's issue, it built fine here
Melody91 has quit [Read error: No route to host]
<cenunix[m]>
I'm trying to get a nixos configuration for the x13s up and running, and so far I haven't even seen the bootloader, could this possibly be related to missing firmware steev? I'm just trying to get this thing to boot for now, i've tried grub and systemd. Whenever i select a boot option in the f12 menu it returns me to the menu in half a second
<cenunix[m]>
i'll upload a gist in a minute and link it so you can see the entire configuration^ including the kernel parameters and the modules
<cenunix[m]>
and the kernel version*
<clover[m]>
The latest archiso I've made have done the same behavior. I didn't really look into it though
<clover[m]>
Even with old kernel
<cenunix[m]>
huh, what changes were made with the latest archiso?
<clover[m]>
Nothing that would cause issues before boot I don't think?
<cenunix[m]>
eh, yeah I have no idea. Usually I can at least get to a bootloader screen before things blow up. I even updated the bios and everything so I could enable the linux boot option
<cenunix[m]>
I tried with an old bios too and specifying the dtb^ so im almost out of ideas
<clover[m]>
You have to specify dtb if booting USB linux option or not
<clover[m]>
I know that much
<cenunix[m]>
ah, ill try specifying it again
<cenunix[m]>
wouldn't I still at least make it to the bootloader though? seems odd
<cenunix[m]>
secure boot is disabled of course
<cenunix[m]>
makes me wonder if i messed something up when partitioning
baozich has joined #aarch64-laptops
hexdump0815 has joined #aarch64-laptops
hexdump01 has quit [Ping timeout: 480 seconds]
<steev>
if it's immediately returning, i don't see that being a firmware issue, where it isn't even throwing in the bootloader; how are you adding the efi vars?
<steev>
what i mean is, the bootloader is before the kernel even comes into play
<travmurav[m]>
agl7-x13s: I wonder if your problem is due to something stumbling on a bracket "\_(\_GALAXY)" in the file path
<travmurav[m]>
hm I guess it shouldn't matter for a "normal" build with no special wrapper tooling
<vkoul>
steev: jenneron[m] bamse has picked it up, so it should be in -next now
<cenunix[m]>
steev: not sure exactly what you mean by setting efi vars, however here are my nixos configs
<cenunix[m]>
these dont boot ofc ^ but they might be useful for reference. I'm assuming giving my lack of awareness around efi vars and me not reaching the bootloader that is most likely the issu
kitsune-tsuki has quit [Remote host closed the connection]
agl7 has joined #aarch64-laptops
<konradybcio>
steev can you provide more details on your battery life? "not near what it was" when?
iivanov has quit [Remote host closed the connection]
abcdw has quit [Remote host closed the connection]
abcdw has joined #aarch64-laptops
iivanov has joined #aarch64-laptops
baozich has quit [Ping timeout: 480 seconds]
<ajhalaney[m]>
robclark: fedora kernel definitely won't work. we have a "day of learning" and I'm trying to "learn" what's not happy but I've been struggling to debug early boot stuff without a serial the last two nights, so I wouldn't hold my breath
baozich has joined #aarch64-laptops
baozich has quit [Remote host closed the connection]
baozich has joined #aarch64-laptops
<konradybcio>
ajhalaney: checking \*QCOM\*, \*MSM\* and \*8280\* would be good for starters
<konradybcio>
in .config
<ajhalaney[m]>
konradybcio: it's almost all on but some stuff are modules. I tried the fedora config with a sha from steev that boots fine with laptop_config and it died on me. I'm gonna flip all the obvious modules over to built-in to match the working config and then walk it back. Most of the obvious stuff _seems_ to be in the initramfs but /me shrugs
<konradybcio>
i'd start with all RPMH stuff to =y, also perhaps INTERCONNECT and GCC and PINCTRL
<ajhalaney[m]>
thanks ill give that a whirl!
<javierm>
ajhalaney[m]: what's your cmdline to get debug output ?
<javierm>
ajhalaney[m]: maybe you can check the MR I proposed some time ago to make the QC Chromebook (SC7180) so I believe the driver set should be quite similar
<ajhalaney[m]>
javierm: the only real addition to pd_ignore_unused, clk_ignore_unused, arm64.nopauth, is earlycon=efifb. I'm a bit confused how console handoff stuff works tbh. I'll peek your MR too, thanks
<javierm>
ajhalaney[m]: this is what I use fwiw: earlycon efi=debug debug initcall_debug log_buf_len=16M ignore_loglevel keep_bootcon
<javierm>
ajhalaney[m]: earlycon=efifb isn't correct for fedora since we got rid of all the fbdev drivers :)
<javierm>
we are just using simpledrm + the DRM fbdev emulation layer
<ajhalaney[m]>
javierm: lmao that's good to know I've been living on a false dream :P
<javierm>
ajhalaney[m]: you can also add boot_delay=100 in case you want to delay the printouts
<javierm>
for the Chromebook I've the SuzyQ cable that gives me serial output, that's pretty handy. I wish all the consumer laptops had something like that
<travmurav[m]>
I thought the firmware/efi/earlycon.c is not related to fbdev
<javierm>
travmurav[m]: oh, you are correct indeed
<javierm>
got confused because drivers/video/fbdev/efifb.c is named the same
<travmurav[m]>
so it should work at least until the "violent takeover" of msm driver where the display stuff gets reset
<javierm>
ajhalaney[m]: scratch that comment then :) but still earlycon without efifb I believe will work and use the default console for earlycon
<javierm>
travmurav[m]: exactly
<ajhalaney[m]>
I don't think stdout-path is set for the thinkpad (since it doesn't have a serial like you luckily do), but I'll check! thanks :)
<ajhalaney[m]>
"violent takeover" lol
<javierm>
travmurav[m], ajhalaney[m]: you can also deny list msm and just use simpledrm if the firmware/bootloader provided a generic system framebuffer
<javierm>
if you are booting with EFI that means having an EFI-GOP set up
<travmurav[m]>
might also want to detach/disable the backlight in case Linux decides to kill it (not sure how it's set up on x13s)
<javierm>
travmurav[m]: agreed
<ajhalaney[m]>
Yeah, I had to do that travmurav (just killed the callback in source)
<javierm>
another trick could be to prevent the IOMMU drivers to get probed since sometimes that can also kill the generic framebuffer set by the firmware
<javierm>
initcall_blacklist=iommu_subsys_init
leezu has quit [Quit: WeeChat 3.0]
leezu has joined #aarch64-laptops
<steev>
konradybcio: the c630 used to get 54 hours of uptime with the system suspended; currently we get < 24
<stirl>
clover[m] -- So I went through the installer last night and when I rebooted there was no option for linux, it just went straight in to windows. Hoping I could get some things clarified as I am not much of an arch wizard.
<stirl>
When doing the pacstrap step, I just installed base and then chrooted to install the linux-x13s package. should I have installed linux-firmware in the pacstrap step? I was assuming that I didn't want to install linux but I wasn't sure about firmware
<stirl>
2nd thing was the bootloader step. The wiki was pretty vague. It just said pick a bootloader so I installed grub.
<stirl>
didn't get grub on reboot though
<stirl>
Other minor things: didn't seem to have any access to any of the RTC files. none of the commands about setting the clock were working. (not sure how important that is) Also I was getting a warning that I didn't have any console fonts installed
<clover[m]>
Grabbing a coffee then I'll reply :)
Xyaon has quit [Remote host closed the connection]
agl7_ has joined #aarch64-laptops
agl7 has quit [Remote host closed the connection]
<steev>
cenunix[m]: i have no idea how nix works exactly; i guess i meant, i don't understand why my kernel stuff comes into play when it doesn't seem to be booting at all? if it's from f12-> whatever; that's pre kernel stuff
<clover[m]>
stirl: so you are able to boot the live environment? to update the clock you need to connect to internet and run timedatectl set-ntp true
<clover[m]>
then if you run date command you can verify your clock is synced
agl7_ has quit [Ping timeout: 480 seconds]
<clover[m]>
after this step you should go ahead and add my repo (steps in the readme) and run a pacman -Syu to sync and update everything
<stirl>
clover[m] -- and all of that happens when chrooted into the installation. Live environment booted up with no issues.
<stirl>
okay I did all of the time and partitioning without chrooting.
<stirl>
do you add your repo before chrooting?
<stirl>
that way pacstrap can install everything?
<clover[m]>
yes because during pacstrap (step 2.2) you should add linux-x13s and linux-x13s-firmware which come from my repo
<stirl>
okay. I just installed base with pacstrap, then chrooted and installed everything from your repo
<konradybcio>
steev: would you be willing and able to bisect that regression?
<konradybcio>
And are you sure it doesn't have to do with your bat getting older? :P
<stirl>
should I still install base? or just the x13s stuff
<clover[m]>
yes try to stick to the wiki as closely as possible :)
<steev>
konradybcio: i'm scared it's the latter :( the c630 is 5 years old? if not more at this point; but at the same time... it'll take forever and a day t bisect since it needs to basically sit there until the battery dies
<stirl>
got it. I was just a bit unsure on where to do the x13s specific steps
<stirl>
I will attempt again tonight
<clover[m]>
err, linux-x13s and x13s-firmware *
<clover[m]>
cool good luck
baozich has quit [Ping timeout: 480 seconds]
<clover[m]>
also idk about grub, i used systemd-boot personally, and i've never installed grub myself before
<stirl>
I will go with systemd then. I didn't have any reasoning behind selecting grub. Just picked one
<stirl>
You have all been very helpful! I have always been kinda scared of the irc chats but this one has been generally great :)
<clover[m]>
mount your root partition then mount your EFI partition inside of your mounted root partition at /boot, then chroot in and bootctl install
<stirl>
so mount root, then chroot /mnt, then mount the efi partition to /boot
<clover[m]>
e.g., mount root partition to /mnt, mount EFI partition to /mnt/boot then chroot to /mnt, run bootctl install (if systemd-package is installed i cant remember, might need to pacman -S systemd-boot first)
<clover[m]>
should be part of base IIRC
hightower4 has joined #aarch64-laptops
baozich has joined #aarch64-laptops
<stirl>
Got it
<stirl>
Got it
<clover[m]>
oh when you chroot make sure to do arch-chroot instead, adds some extra goodness
<clover[m]>
and here are the contents of my /boot/loader/entries/arch.conf
<clover[m]>
your root partition uuid will be different of course. you can find that with blkid iirc
hightower3 has quit [Ping timeout: 480 seconds]
<stirl>
and thanks
<clover[m]>
note the missing dtb= kernel parameter. thats cause i have my linux option switched on in the bios. if you have it off, you will need to add dtb=sc8280xp-lenovo-thinkpad-x13s.dtb
<clover[m]>
thats literally the only thing linux option does lmao
<clover[m]>
if i ever get around to restructuring my disk free space, i might do a tutorial write up. idk if i want to keep supporting arch for x13s with all of the boot weirdness like what travmurav ran into yesterday
<clover[m]>
also package updates seem to be ... broken? right now in alarm with no communication when or if things will be fixed
<konradybcio>
steev: does the c630 not have means of displaying battery current consumpion through some pmic?
* travmurav[m]
is not sure if the ping for the boot weirdness was wrong or the timezones are this funny
<clover[m]>
whoops, that was cenunix. sorry for the confusing ping travmurav. D:
baozich has quit [Ping timeout: 480 seconds]
<travmurav[m]>
np :)
<stirl>
clover[m] -- is the goal to move to fedora or debian?
<clover[m]>
debian is a bit boring, i was hoping someone in here had some success with fedora but nothing reproducible has popped up. also nixOS might be an option but seems kinda esoteric
<stirl>
fedora I think would be a good place for this if we could get it working. The first thing I tried was the default fedora image but didn't boot
<stirl>
agree on the debian = boring take
hightower4 has quit [Ping timeout: 480 seconds]
<clover[m]>
the goal right now is just pray arch linux arm fixes their build infrastructure to account for the recent git migration upstream arch did. if that happens i don't need to distro hop
<stirl>
is there any expectation that support for these soc's will make it into the mainline kernel at some point
<clover[m]>
yes
<clover[m]>
we have multiple talented devs mainlining the x13s right now :)
<clover[m]>
sometimes they pop in and say what they are accomplishing, its pretty cool
<robclark>
bamse, konradybcio: hmm, looks like a690_gmu.bin is missing from linux-firmware?
<robclark>
ok, useful to know.. but I thought fedora's /lib/firmware was missing something or out of date.. until I realized it was also missing in linux-firmware.git
<qzed>
nope, that's the one I extracted from the windows driver ages ago (with the a680 one as well)... not sure if there's a way to get that upstream
<qzed>
IIRC it was possible to fall back to some other fw
<qzed>
robclark: ^
<robclark>
clover[m]: jfwiw kernel will look for a660_sqe.fw and a690_gmu.bin, the former is already in l-f .. no need for an a690_sqe.fw
<robclark>
qzed: ahh, ok.. I guess hopefully linaro will sort out a "proper" a690_gmu.bin
<clover[m]>
make a PR
<clover[m]>
:P jk i will test that out
<qzed>
I'm just not sure about the legalities of all that, I think that should probably go through the proper channels xD
<qzed>
and lenovo hopefully has some weight to throw around to get that sorted
<steev>
konradybcio: oh, it might
<qzed>
steev, jenneron: I have some new/updated branches with the sc8180x patches from vkoul, usb-mp from krishna, and qseecom/uefisecapp v4
<steev>
oh nice
<qzed>
spx/v6.3 or tag surface/spx/v6.3.5 and spx/next-20230602
<qzed>
things are slowly getting there
<robclark>
afaict lenovo themselves pushed all of the signed fw's.. I guess gmu is one that got as-is from qcom (it is not signed).. but linaro has handled sqe/gmu fw before so hopefully they can do it again
<steev>
afaik, there's work being done there, but nothing is out
<qzed>
right, I guess that makes sense, lenovo might not even have the file themselves since it's hidden in the driver
<konradybcio>
robclark: i wonder if we need to ship more than one gmu fw anyway
<konradybcio>
A730 boots fine with a740's
<konradybcio>
I'll perform some experiments..
<robclark>
hmm.. I'm not sure if bamse decided we needed a690_gmu.bin just based on the file extracted being different or if there is some legit reason for it being different
agl7 has joined #aarch64-laptops
<agl7>
jenneron[m]: Are you here?
<konradybcio>
From what I understand, GMU firmware is just a software construct and we have to have a high enough version to support a given gpu
<jenneron[m]>
agl7: yes?
<agl7>
jenneron[m]: I'am in the Clubhaus, and at home I can not access the x13s because it is off. I want to try to build once more your linux kernel for sc8180x (Galaxy-Book-S) on a NanoPi R6s Computer (I have access at home over the Internet wit rustdesk). Can you give me once more the URLs where I can download your kernel, because this are on the sSD of the x13s where I have no access.
<robclark>
konradybcio: yeah, and a single gmu fw does seem to have support for multiple different gpu's, at least within the same sub-generation/family.. but idk if qc realized they needed some fixes for a690 gmu which aren't in the original a660_gmu.bin or something along those lines?
<robclark>
ajhalaney[m]: so what do I need to do after `make install` of new kernel to make systemd-boot happy.. seems like `make install` step generates /boot/loader fragment but sd-boot is looking instead in the ESP (not sure if it is just using EFI services for file access, in which case it probably does not have ext4 support to load things from /boot)
<robclark>
and are using the uki thing or separate initrd/dtb/img
<agl7>
jenneron[m]: I have DietPi OS (a Debian derivat) on the NanoPi R6S!
<ajhalaney[m]>
I'm using separate stuff not the UKI robclark . I manually move the vmlinuz-*, initramfs-*, and loader/entries/* stuff over to the ESP
<ajhalaney[m]>
It is using EFI services and unfortunately no ext4 there
<robclark>
ajhalaney[m]: ok, yeah it works with manual copy and artisan crafted loader/entries .conf file.. was just kinda hoping there was a better way.. but at least I can boot now w/out jerry rigged arch boot partition on usb-c pointing at rootfs on nvme
<ajhalaney[m]>
definitely a better way, i just have this thing duct taped together right now lol
<robclark>
clover[m]: that could be because fw not in initrd (which is fine, it will try again later).. or because you are missing fw.. or because /lib/firmware is using compressed files but option for that not enabled in kernel
<jenneron[m]>
i have, but it is too long and i'm tired, i have just missed the question
<jenneron[m]>
you can use whitequark for history of the room
<agl7>
ok
<robclark>
clover[m]: fwiw, looks like laptop_defconfig does not enable module xy compression.. so if $distro you are using does that could be the issue
<agl7>
jenneron[m]: Thank you :-)
<robclark>
err, no, not module compression, there is a separate fw compression option
<robclark>
CONFIG_FW_LOADER_COMPRESS
<konradybcio>
robclark: gmu fw is versioned, so I'd assume higher fw contains all the fixes for lower fw ver
<konradybcio>
a740 was like v4.1 or 4.2
<konradybcio>
a6xx was at v2.x
<robclark>
I try to not make too many assumptions about how they manage their fw versioning since in the past it has been a mess ;-)
<robclark>
(or whether they actually test newer fw versions on older devices, etc)
<clover[m]>
not a huge deal either way, my desktop environment is accelerated and that's enough for me heh
<robclark>
if thing are accelerated it means it eventually managed to load fw, so all is good
<robclark>
konradybcio: well, a690_gmu.bin as a hard link to a660_gmu.bin at least works well enough to get to gnome-shell desktop
<agl7>
jenneron[m]: On the "NanoPi R6S" with "DietPi OS (Debian)" I can compile your kernel. On the ThinkPad X13s shortly after beginn compiling there came yesterday a Shell-Error (by /bin/sh), now all works.
<agl7>
jenneron[m]: I don't know why there came an error because steev's kernel are compliled without any error.
<stirl>
clover[m] -- will the uuid I need in the config file be the uuid for the root partition or the boot partition
Evaia63101 has quit [Quit: Hack the Gibson]
Evaia63101 has joined #aarch64-laptops
<clover[m]>
Root
<robclark>
bamse: btw, scp'ing things to the x13s (via usb-c eth) and ksoftirqd is using like 100% of a core
<robclark>
which seems a bit... excessive
<clover[m]>
says so right in the parameter name ^_^
<stirl>
Thanks! I saw it was in the parameter name but just wanted to double check
<HdkR>
robclark: Welcome to my world when trying to run games off the network
<HdkR>
robclark: iommu.strict=0 in the boot options should make it less apparent but it won't really improve the perf
<robclark>
huh, 80.88% in _raw_spin_unlock_irqrestore()
<robclark>
at any rate, if it is iommu related, I remember we had to do some tuning around that on sc7180.. something about wait-for-safe logic or so
stirl has quit [Read error: Connection reset by peer]
iivanov has quit [Ping timeout: 480 seconds]
<robclark>
so, for the all important and rigorous "how fast does it compile mesa" benchmark (not exactly distro config but most drivers enabled).. 2m41.356s .. vs my xps13 (kbl i7-8550U) at 6m31.153s .. color me impressed
<HdkR>
Those X1C and A78C cores are pretty beefy
<HdkR>
But also that KBL supports AVX2, so not really feature parity :)
<ardb>
robclark: _raw_spin_unlock_irqrestore() is where the perf sampling interrupt gets taken
<robclark>
hmm ok
<robclark>
HdkR: heh.. compared stk fps.. the kbl laptop is a bit lower res (only 1920x1080).. but ~180fps vs ~40fps !! I knew intel gpu hw wasn't great but that is a big difference
leezu has quit [Quit: WeeChat 3.0]
<robclark>
that was with the xps13 wheezing too (fan running) :-P
<HdkR>
robclark: I've had a few cases where the X13s beats out the SteamDeck as well. It's just a really beefy GPU
<HdkR>
Until it starts thermally throttling anyway
<robclark>
heh, they just need to put one of these chips in a steam deck with some active cooling ;-)
<HdkR>
I can't guarantee that level of back compat yet. Users would be really angry if they can't play their latest game running anti-emulation code
<clover[m]>
robclark: fedora arm success? What did you have to do?
<robclark>
what _didn't_ I have to do.. but yeah, got it installed
<robclark>
had to use arch iso img to boot strap it, and it involved at one point manually smashing kernel modules from arch initrd into a fedora one
<robclark>
and some lolz because for systemd-boot the arch ESP partition is _way_ too small
<clover[m]>
Did you use my archiso or an official one?
<clover[m]>
Cause I might be able to make the ESP partition larger