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)
<qzed> we got lift-of on uefivars: at least on the surface pro x: https://paste.debian.net/1247761/
<qzed> probably wont work on the flex though :/
<qzed> that needs some rpmb or storage callbacks
<qzed> but on the pro x storage seems to be managed fully by tz
<qzed> steev: if you want to give it a try: https://github.com/qzed/linux-qcom-uefivars
<qzed> if you do that, you might want to make this dev_info or something higher: https://github.com/qzed/linux-qcom-uefivars/blob/47d4f490670afabb37dbbe41066d29e1212b496a/qcom_uefivars.c#L122
<szclsya[m]> hmm seems the pixelated sway is due to fractional scaling with software renderer. use a scale factor 2 solves the problem
<szclsya[m]> hopefully we'll have functional gpu soon
<steev> qzed: i'll try to squeeze some time in to try :D
<steev> oh, that's nice
<steev> qzed: it'll need changes when next rolls around
<steev> aha
<steev> shuffling the kernel across to the flex so i can test now
hexdump0815 has joined #aarch64-laptops
hexdump01 has quit [Ping timeout: 480 seconds]
<steev> '[ 163.420185] qcom_uefivars qcom_uefivars: _qcuefi_get_and_print_next: name=COMPAL, guid=B697DE83-1AB6-42C4-9DEE-A806C637818B' it our battery lol
<szclsya[m]> steev: huh bluetooth works?
<steev> szclsya[m]: no :(
<steev> by the looks of it, the bluetooth driver doesn't know about our chipset yet
<steev> qzed: i'll try it on the thinkpad as well, just take a bit more setup there
<szclsya[m]> steev: oh well. I heard that bluetooth interface are pretty universal (from the M2 blog post), does it mean we can just add the hardware id to kernel and expect it to work?
<steev> not quite, we need to also add the bits to the dt (the gpios i asked bjorn about earlier)
<steev> but yeah, once those two things are in place... it should work
<szclsya[m]> ah, makes sense
<steev> i checked in bluetooth-next hoping maybe someone had already added it and it hadn't yet made its way down the line, but doesn't appear to be the case
<steev> it's hard to tell though - i copied the entirety of FileRepository over to my linux install, and the folder i have is qcbtfmuart_hsp8280.inf_arm64_bbb33bd69dc3395b
<steev> but it has both .tlv and .bin files, as well as a bunch of bXX files
<steev> and the .ver files say they are
<steev> BTFM.HSP.1.0.0-00709-QCAHSPROMZ-1
<steev> BTFW.HSP.2.0.0-00607-WOS_PATCHZ-1
<steev> qzed: and the thinkpad - https://paste.debian.net/1247771 (i made the line a dev_info on the thinkpad)
<steev> [ 89.529032] qcom_uefivars qcom_uefivars: _qcuefi_get_and_print_next: name=LenovoHiddenSetting, guid=1827CFC7-4E61-4273-B796-D35F4B0C88FC
<steev> hmmmmm
<szclsya[m]> Lenovo Secret Sauce (TM)
<steev> there are a few in there that make me laugh
<steev> dianders: https://lists.freedesktop.org/archives/dri-devel/2022-July/365201.html blargh, i say R3 in the mail and R2 in the actual change *sigh*
<steev> ugh, i don't know why it's double signed of by :(
<jhovold> steev: thanks for testing. Regarding the warnings for the missing i2c devices, I did no longer see them here, but I may have been running with some upcoming i2c fixes from bamse. Will double check.
<jhovold> szclsya[m]: thanks for confirming!
iivanov has joined #aarch64-laptops
matthias_bgg has joined #aarch64-laptops
<qzed> steev: nice, I didn't expect it to work on the flex but that's a pleasant surprise
<qzed> I guess it make sense since the UEFI partition thing was all zeros
<qzed> next steps are getting GetVariable and SetVariable working and hooking everything up
<qzed> after that you can hopefully dig into all the secret menus :P
SallyAhaj has quit [Remote host closed the connection]
iivanov__ has joined #aarch64-laptops
iivanov has quit [Ping timeout: 480 seconds]
iivanov__ has quit [Quit: Leaving...]
moa5505 has joined #aarch64-laptops
<moa5505> Hello ! Any advice on which ARM laptop to buy (and where) ?
<moa5505> From what I have seen, the x13s looks like the fancier one, the flex 5g looks like a good quality build and the C630 looks like the best bang for buck but it has an older processor.
<moa5505> Compatibility on the C630 also seems to be better than the rest.
<moa5505> (Also most models only seem to be available in the US, and I'm in Europe ...)
matthias_bgg has quit [Ping timeout: 480 seconds]
<rfs613> moa5505: you're probably aware the c630 is discontinued by lenovo, so you're probably looking at used ones
<rfs613> and yes its processor is older than the others, whether this matters depends on your use case
<qzed> steev: reading and writing variables works now too... now I only need to figure out how to hook this up with the kernel and with efivarfs...
<HdkR> Blammo, x13s
matthias_bgg has joined #aarch64-laptops
<moa5505> @rfs613 Yeah, I only saw used or refurbished ones
<szclsya[m]> the soc (8cx gen3) in X13s is significantly more powerful than previous generations, so I'd recommend X13s
<moa5505> Do you think the snapdragon 850 is enough for web browsing, text editing and light developpment ?
<szclsya[m]> (sadly they haven't put it in other laptops or tablets
<moa5505> It's also 1000$ more expensive ^^''
<szclsya[m]> right... then I guess 850 should handle emacs just fine. I doubt it will perform well with language with lsp though
<clover[m]> if you can stand the keyboard/hardware, M1 / M2 macbook air is probably the most sensible choice as it has Asahi Linux support, and that project is thriving.
<clover[m]> i'm really hoping this X13s will be my endgame machine, once more linux support comes
<HdkR> price to perf the bottom tier ARM macbooks are quite good as well
<clover[m]> yeah that's what i am saying too
<HdkR> $1000 for a highend ARM device? That's hard to find :D
<clover[m]> s/keyboard/keyboard layout/
<gwolf> moa5505: I am very happy with my c630; is not the latest and as others said it's discontinued (already was a year ago when I got it), but works quite well
<gwolf> If your needs requires a fast processor for CPU intensive tasks, however... Don't use ARM.
<gwolf> I do use mine for the precise use case you present.
<moa5505> Yeah, Apple really changed the whole game with their M1/M2, even in $/perf they are dominating atm... Which was far from being the case 5 years ago
<moa5505> About the X13s, it really looks amazing I agree
<HdkR> Flashing an image to a USB drive right now, hopefully the x13s works without...too much pain
<moa5505> Thanks gwolf, I think given my tight budget that the C630 is the best choice here ...
<szclsya[m]> moa5505: from what I observe apple's $/perf advantage only holds if you want the bare minimum configuration
<moa5505> Can you dual boot easily with the C630 ?
<szclsya[m]> if you want more ram and drive space the price will quickly climbs up (and their ssd aren't replaceable, so there isn't a choice
<moa5505> @szclsya That's maybe true, but even out of the 8G and swapping the thing is really responsive ...
<clover[m]> HdkR: you using my ISO?
<clover[m]> i used balenaetcher to flash it, btw
<szclsya[m]> moa5505: yeah, I do believe if that is enough for what you do, then it's a fantastic value
<HdkR> clover[m]: You have an ISO?
<clover[m]> yes :>
<HdkR> Then no, didn't even know about it
<HdkR> What does this image provide?
<clover[m]> its a basic archiso with linux-steev
<clover[m]> here's the project dir: https://github.com/ironrobin/archiso-x13s
<clover[m]> no wifi support yet, you'll have to figure that out, i don't even know how
<clover[m]> i think Leo got it working
<HdkR> I can give it a try and then rip the Arch out of it
<clover[m]> lulz ok
<clover[m]> Leo Shen: you got wifi working right? did you mention it was slow?
<clover[m]> i think we have a WiFi 6 card so it should be good
<szclsya[m]> it's... weird. it cannot connect to some hotspots and are extremely slow sometimes for no reason
<szclsya[m]> makes sense since we are hacking firmware for another card onto it
matthias_bgg has quit [Ping timeout: 480 seconds]
<steev> clover[m]: you might need to create a symlink from 2.0 to 2.1 like https://github.com/kvalo/ath11k-firmware/tree/master/WCN6855 does
<szclsya[m]> steev: alarm's linux-firmware package did that by default for some reason, so mine just (well, sorta) works
<steev> hm, i wonder why it's not working for clover :(
<clover[m]> do i have linux-firmware scratches head
<szclsya[m]> I just slam iwd onto it and it just picks up, fyi
<szclsya[m]> But it does freak out from time to time, but simply reload ath11k_pci usually fixes it
<clover[m]> iwd too
<steev> can you post your dmesg output?
<gwolf> moa5505: Yes, the C630 (usually?) comes with Windows installed, and you can install Linux with a (barely modified) debian-installer
<gwolf> you end up booting to GRUB, which presents you with a (working!) chooser menu with Windows and Linu
<clover[m]> steev: am at coffee shop doing paid work rn, but yes, later :)
<clover[m]> hate how WSL2 just crashes randomly, can't wait to move to baremetal linux on this machine
<steev> i haven't experienced that here, but admittedly, as soon as i got linux mostly working, i stay in here. occasionally i go into windows to check for updates, but that's pretty much every tuesday because i can never remember when "patch tuesday" is
<clover[m]> does x13s support 16k page kernels, steev?
<clover[m]> like how the Apple Silicon macs do it?
<steev> i honestly have no idea, but my understanding was that there's a lot of shit (still) broken with that in the linux userland
<clover[m]> yeah, you're not wrong about that
<clover[m]> but the perf boost would be cool
<qzed> steev: I have somewhat feature-complete uefivars now
<steev> qzed: nice!
<steev> HdkR: wtf is that? FEX not likely to be fixed? smh
<qzed> insmod the module, run "mount -t efivarfs none /sys/firmware/efi/efivars", and off you go
<steev> qzed: oh that's awesome
<steev> qzed: it'll still need changes in -next (so 5.20? 5.21?) because qcom scm changes coming
<qzed> ah okay, I'll try to have a look at that
<qzed> next up: moving that into the kernel with DT bindings and then we should no longer need efishell to add a boot entry...
<steev> qzed: https://paste.debian.net/1247871 (my build against jhovold's next tree for the x13s)
<steev> qzed: bindings? or just documentation?
<qzed> ah, did you pull in that one patch?
<qzed> and I need to figure out how to split that module up too... at the moment it does more than it should
<steev> oh! right, that patch doesn't apply
<qzed> got it, I'll try to figure out what changed
<steev> i haven't looked into why just yet
<qzed> we'll probably need some very basic DT bindings for that to decide on which platforms to load it
<steev> but i had a free moment because sucuri blocked me from the web service i need to use from work for trying to exploit my server (allegedly)
<qzed> hmm, I can't spot anything that'd break it... guess I'll have to give it a try
<steev> the hunks at 525 and 555 fail
<qzed> noticed that just now, I'll push a branch once I've fixed that
<qzed> steev: https://github.com/linux-surface/kernel/commit/30b460981243f08196c838a9377a6a4f9d9ea8de (not tested/compile-tested in any way yet though)
<steev> compiles
<clover[m]> nice
<HdkR> steev: What's this about?
<HdkR> Oh, page size stuff
<HdkR> I could introduce some hacks for some partial 16k page support but running x86 applications with a 16k page size will just break things spuriously
<steev> "D
<HdkR> More ideally Linux sprouts support for mixed page size userspace, and IOMMU on Apple devices still ensures 16k IO alignment in the kernel space
<HdkR> Isn't too unheard of since Nvidia devices also require 64k alignments in x86 land
<clover[m]> <steev> "i haven't experienced that here,..." <- i get multiple crashes per day. workaround is to run `wsl --shutdown` and wait for it to kill everything then re-launch WSL2
<steev> qzed: insmods, mount command works, i do see them in there :D
<qzed> nice, thanks for testing!
<steev> whenever you do shove them at the mailing lists, cc me and i can send my T-b
<qzed> will do, but i'll have to do a bit of clean up and thinking about how i structure the secure-os vs uefisecapp stuff (should really be two modules instead of one...)
<steev> worst case, throw it up, and let the powers that be decide if it should be split
<qzed> I mean splitting it up isn't the hard part... it's just properly naming things ...
<steev> yeah, i still let the maintainers decide what to call things :P
<clover[m]> what's in a name? a rose by any other name would smell as sweet...
<HdkR> This X13s really has trouble booting off of USB doesn't it
<HdkR> Sometimes the USB boot device just...doesn't show up
<clover[m]> reboot from windows and give it a sec before pressing f12
<clover[m]> can X13s run 32 bit binaries or 64 bit only
<HdkR> 4xX1 + 4xA78, Supports both 32-bit and 64-bit
<clover[m]> tanks
<HdkR> 8G1 has mixed support, next-gen is when to be concerned.
<steev> i've never had an issue with my usb devices on the x13s - but i use the ones i linked a couple days back, exclusively - and 32g in size because of the broken c630 bios stuff
<clover[m]> if you are using usb-c try flipping it a couple times, i think that really can make a difference
<clover[m]> so apparently it might be the last generation of fast ARM chips that will run 32 bit, HdkR
<HdkR> Pretty much. I'm fine with that
<szclsya[m]> I see no problem with this. It's not like desktop linux on arm has a large amount of legacy apps to carry, same as windows on arm
<steev> i'm just glad they *finally* released native toolchains on WoA
<steev> well, it's still beta, but still
<HdkR> A bunch of Android applications going away is likely the biggest casualty
moa5505 has quit [Remote host closed the connection]
<clover[m]> <HdkR> "4xX1 + 4xA78, Supports both 32-..." <- So X1 are big and A78 are little cores?
<HdkR> It's more of biggest + big arrangement, but sure
<clover[m]> Hehe
<steev> it IS nice to compile alacritty in 2 1/2 minutes
<steev> though the c630 does it in ~6
<steev> still not a massive wait
<clover[m]> Why are you compiling alacrity lol
<clover[m]> You running gentoo?
<HdkR> This X13s throws quite a few kernel backtraces doesn't it
janrinze has joined #aarch64-laptops
<HdkR> panel-edp.c:740, unknown panel IVO 0x854b
<janrinze> clover[m]: let me know when you had time to check the Geekbench 5
<clover[m]> Ok
<HdkR> Bunch of alg_test backtraces, failing to allocate for multiple things
<janrinze> Anyone here running RISC OS on Linux on their aarch64 laptop?
<clover[m]> Welcome janrinze: lots of x13s linux users in here. At least 5 of us 😃
<janrinze> :-D
Lucy[m] has joined #aarch64-laptops
<janrinze> clover[m]: you have no idea how many ARM laptops and SBCs i have and did linux porting for..
<janrinze> It's been a weird but interesting journey
<clover[m]> Are you going to get an x13s, too? :p
<janrinze> Depends.. they're rather pricey a.t.m.
<janrinze> haven't seen any in my regular webshops yet
<robclark> HdkR: unknown panel thing is safe to ignore for now.. panel-edp will fall back to conservative/save (ie. longer) power-on delay but should otherwise work
<janrinze> clover[m]: what's the full typename of the lenovo laptop?
<clover[m]> Dtb is named sc8280xp-thinkpad.dtb
<janrinze> Lenovo has it's own type name system.. i think this was a XXX- H3 --YY something
<janrinze> cna't find it anymore.. The X13 is also an AMD laptop so it's not easy to find one online
<szclsya[m]> 21BX
<szclsya[m]> latter part depends on your configuration and region
<clover[m]> Don't get distracted by the z13 I know it looks nice 😆
<steev> HdkR: yeah, the algo stuff is something - i sent a patch for the splat you see last night - https://lists.freedesktop.org/archives/dri-devel/2022-July/365202.html
<HdkR> Fancy
<steev> apparently i screwed up pretty badly though because it doesn't show up in the dri-devel patchwork?
<robclark> if you aren't subscribed to dri-devel it might mean you have to wait until someone gets around to dealing with the spam queue
<steev> ah, definitely not subscribed
<steev> i know i screwed up anyway, somehow, because my Sob is on it twice?
<robclark> would be nice if mailman had some rule to let through things that look like patches
<clover[m]> Leo Shen: if you have time can you run Geekbench-5.4.4-LinuxARMPreview
<steev> hm, i'm not entirely sure about the algo stuff actually, but crypto is waaaaaaaaaaaaaaaaaaaaaaaay outside my wheelhouse
<steev> i have issues computing 1+1=2 so, you really don't want me doing crypto stuff
<szclsya[m]> not a huge fan of geekbench, but ok
<clover[m]> Ty sir
<janrinze> szclsya[m]: which benchmark tool do you prefer?
<szclsya[m]> I just compile my rust project and check the build time. suit my workload
<janrinze> szclsya[m]: how does that help to compare with other systems?
<janrinze> specifically systems you are considering to buy?
<janrinze> for example I was considering the NVidia AGX Orin but decided to get a Mac Studio instead.. If there were any NVidia Orin laptops I would definitely consider that..
<janrinze> my last ARM laptop was a NVidia K1 laptop.. I don't think that supports 64 bit..
<szclsya[m]> I find building itself is similar to my previous 4750U laptop, but LTO phase is slower
<szclsya[m]> not sure if it's a big.LITTLE thing
<janrinze> the 4750U is a Ryzen?
<szclsya[m]> yup, I'm pretty happy with its performance, but the battery life is just not enough
<janrinze> in my experience compiling large projects are often disk bandwidth limited.
<janrinze> szclsya[m]: are you using ccache?
<szclsya[m]> I'm using cargo, which should do compile cache automatically
<szclsya[m]> I usually do cargo clean before running the benchmark build though, so it would be fair
<szclsya[m]> fyi, https://github.com/szclsya/mpdris2-rs this project took 1m50s on 4750u and 2m23s on x13s (with -O3 and lto
<szclsya[m]> There's some error in information detected in geekbench
<steev> HdkR: oh, you're going to want to grab the .mbns from the windows partition and copy them into /lib/firmware/qcom/sc8280xp/ otherwise your battery won't charge
<HdkR> oops
<steev> you wiped windows didn't you
<HdkR> You know me quite well
<szclsya[m]> hopefully you didn't wipe the recovery partition
<steev> he wiped the entire drive :D
<szclsya[m]> oops
<HdkR> Of course, destroy the internal drive
<janrinze> szclsya[m]: did you cross compile on he 4750u?
<szclsya[m]> no, compile to x64 on 4750u, aarch64 on x13s
<szclsya[m]> but you're right, cross compile to a same arch would be more fair
<janrinze> aarch64 compiling takes a lot more time..
<janrinze> specifically lto
<szclsya[m]> yeah, I guess that's why it takes that much more time on x13s
<steev> szclsya[m]: the thing about HdkR, he goes all in on linux :D
<HdkR> I don't have time to deal with a single Windows
<janrinze> what kind of memory bandwidth does that Lenovo have?
<szclsya[m]> steev: I would happily do the same if more things are functional, to be fair
<HdkR> 86.3GB/s
<steev> it takes time, but it'll get there
<janrinze> HdkR: after testing win10 on ARM, win11 on ARM i noticed that i only installed it just to see if it works.. never actually used it.
<janrinze> HdkR: aha.. so faster than the mac mini. that's good.
<HdkR> Until M2 Mac Mini comes along anyway.
<janrinze> The NVidia AGX is still twice the bandwidth of the mac mini but has pretty outdated cpu by now..
<janrinze> HdkR: Mac Studio .. that's the memory bandwidth king a.t.m.
<HdkR> Of course, Ultra is nutty
<janrinze> yup.. running cpu-only 3d raymarching tests is awesome..
<szclsya[m]> weird why the single core score is much lower compared to the multicore
<szclsya[m]> maybe it's getting allocated to the smaller core?
<janrinze> szclsya[m]: which machine do you refer to?
<szclsya[m]> the x13s, it has 4 large cores and 4 lager cores
<steev> i would like a lager core
<szclsya[m]> (well, large core relative to the efficient cores used by smartphone socs
<janrinze> try using taskset ..
<szclsya[m]> good idea
<steev> robclark: oh, the thing is, it shows up on the mailing list, just not on patchwork
<HdkR> On my 888 board I had to crank the DT capacity on the X1 core just so things got scheduled correctly on that core. But shouldn't have the same problem with the 8cxg3 arrangement :D
<janrinze> will reboot now to test newly built kernel.. brb.
janrinze has quit [Quit: Leaving.]
<steev> and jan was never heard from again :(
<szclsya[m]> :'(
janrinze has joined #aarch64-laptops
<steev> qzed: huh, interesting
<steev> "Adding boot menu entry for EFI firmware configuration" unexpected but welcome change, i guess?
<janrinze> I changed the kernel timer from 250Hz to 1000Hz to see if it changes anything.
<janrinze> can't measure a difference. feels a bit snappier though.
<HdkR> oh me oh my, the storage on the X13s is removable
<steev> o really
<robclark> steev: odd.. sometimes patchwork just missing things, perhaps?
<steev> robclark: indeed :) the msm one does from time to time as well
<HdkR> Even has a second M.2 standoff
<steev> hmmmmmmmm
<HdkR> Oh, that second one isn't M.2
<janrinze> HdkR: it looks mirrored
<HdkR> It's actually a smidge smaller
<janrinze> SATA?
<HdkR> Doubt, probably something custom for a cellular board
<janrinze> I got an enclosure that I thought needed nvme but it was a SATA.. that has actually two notches
<HdkR> I'm surprised this drive is a PCIe 4.0 4x drive
<steev> that big ass heat sink block by the usb ports definitely explains why that side gets quite warm
<janrinze> looks like they did a better job with the heatsink than Apple did with their M2 macbook air :-D
<HdkR> Quite a wide thin heat pipe
<janrinze> good pictures.
<janrinze> the laptop costs around 1400 euro's?
<steev> depending on their sale code each month, yeah
<janrinze> at lenovo.com they show the laptop with 4 different prices.. might be a good idea to add the differences in configuration for each of these..
<HdkR> They don't even have all configurations available yet. There is supposed to be a 400nits display one, and one with 1TB SSD
<janrinze> the 32GB mem model only has 512GB SSD.. weird.
<szclsya[m]> yeah
<szclsya[m]> even worse is that you can't even buy a larger pcie4 2242 ssd right now
<szclsya[m]> the only name brand I can find is Sabrent, but it's only pcie3
<HdkR> Maybe even Lenovo is having problems sourcing the 1TB 2242 drive
<szclsya[m]> wonder if we can make an adapter board to 2280 that utilizes the space if the wwan slot is not occupied
* janrinze is off to bed..
<HdkR> Hm, grub booting from internal drive but only to its bash helper
<HdkR> Oh right, make sure to adjust uuid in EFI
<HdkR> Need to adjust kernel image now and...give an initrd?
<HdkR> I forget, how do I generate an initrd, is it necessary?
<szclsya[m]> it will work if you bundle everything you need into the kernel
<szclsya[m]> *it will work without initrd