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)
aceridus has quit [Remote host closed the connection]
iivanov has quit [Ping timeout: 480 seconds]
iivanov has joined #aarch64-laptops
iivanov has quit [Quit: Leaving...]
iivanov has joined #aarch64-laptops
iivanov has quit [Remote host closed the connection]
iivanov has joined #aarch64-laptops
systwi has quit [Read error: Connection reset by peer]
systwi has joined #aarch64-laptops
pbox has joined #aarch64-laptops
<pbox>
hello, I'm looking for a laptop that can churn a lot of data (i.e., RAM) and works well with linux. however, I have little to no idea about what is in the market, what I can expect to find, etc.
<HdkR>
How much ram we talking here?
<pbox>
would anyone like to share suggestions for specs and models?
<pbox>
I dont know, I think at least 32
<pbox>
but i'd personally go for 64
<HdkR>
Then your options have been severely cut down
<HdkR>
64GB you only have Apple laptops
<pbox>
oh
<HdkR>
32GB you have Lenovo X13s and...One of the Surface Pro devices?
<pbox>
I see! I will check both
<HdkR>
Oh, looks like Surface Pro ARM devices capped out at 16GB
<pbox>
I never had an ARM machine other than a phone - how well do they perform?
<HdkR>
Apple devices compete with x86 devices in performance
<HdkR>
Snapdragon devices not quite as much.
<pbox>
the lenovo has a snapdragon cpu
<HdkR>
Indeed, it definitely won't have an Apple CPU
<HdkR>
Snapdragon CPU performance is usually 2-3 generations behind Apple hardware. Apple hardware is only 1-2 generations behind x86 device performance.
<HdkR>
perf/w is a different story if that matters, but if peak perf isn't great then perf/w is less interesting :P
pbox has quit [Remote host closed the connection]
<Sobek[m]>
For what it’s worth perf/W is an important metric in quite a few cases. And unless you are in the market for saw, a top of the line i9 or Xeon, there’s probably an Apple CPU which can match its performance. GPU is a bit harder on Apple especially on the game side of things.
<HdkR>
Need that Asahi GPU to come up quickly :D
<HdkR>
But you also wouldn't get a GPU on X13s today either so eh
<robclark>
HdkR: seems like we might be getting closer.. at least it is to the point of crashing the gpu ;-)
<HdkR>
nice
<HdkR>
Better than crashing the device without and debug information
<Caterpillar>
pbox: I don't understand what X1 gen is the URL about
<robclark>
pbox: the topic here is arm laptops, but IME if you want an x86 laptop lenovo (and maybe some other vendors) do have some that ship with linux.. I've got an xps13 that I've had for a few years with no problems running linux (and I know others with newer and later gens of xps13)
<pbox>
gen10
<pbox>
oh wow - sorry to be off topic
<robclark>
Caterpillar: nope but I work on drm/msm and mesa and was trying to do what I could to remote debug gpu bring-up for a690 ;-)
<Caterpillar>
robclark: I have an x13s, as soon I manage to boot Linux I am happy to help testing
<clover[m]>
i think his issue is he is trying to boot with ACPI and needs to use a dtb instead
aceridus has joined #aarch64-laptops
<aceridus>
clover[m]: I just finished reading and agree. I want to message him and suggest a few kernel debug flags so he can at least see where it is choking
<aceridus>
I am extremely tempted to grab one of the dev kits myself, just to tinker
<aceridus>
The price for the specs is actually quite nice!
pbox has quit [Remote host closed the connection]
<clover[m]>
i don't think he will ever be able to use KVM though. i think something about the qualcomm chips blocks us from doing that
<aceridus>
clover[m]: Do the qualcomm chips have virtualization capabilities? If so, I am guessing that is what Windows uses, at least in part, for running x86 apps via their emulation layers?
<jenneron[m]>
it should be something windows uses for wsl2
<clover[m]>
>what steev says... afaik there's some signed blob "kernel extension" thing that essentially acts as a mini-hypervisor in el2 and needs to be loaded by the OS, or something like that
<clover[m]>
unfortunately that's not documented anywhere
<clover[m]>
-qzed
<aceridus>
Ugh, that is annoying.
<robclark>
qcom chips do support hypervisor.. but unfortunately not in a way that KVM can use with android or windows firmware (because linux kernel starts in EL1, not EL2).. chromebooks w/ qcom SoC have different firmware where kernel starts in EL2 (because we use KVM extensively)
<aceridus>
robclark: Interesting, so EL2 is a lower or higher level than EL1? Also, what is EL?
<javierm>
aceridus: Exception Level
<broonie>
Higher, it stands for Exception Level. ELn is the supervisor for ELn-1 in aarch64.
<broonie>
So EL0 is userspace, EL1,the kernel, EL2 the hypervisor and EL3 the firmware.
<aceridus>
Awesome info, thanks!
<aceridus>
I presume in order to get the Linux kernel running at EL2 we would need qualcomm's blessing and signed blobs from them to allow such a thing?
<javierm>
robclark: do you know of any other snapdragon 8 device where VHE is enabled for Linux/KVM to run in EL2?
<javierm>
IMO this fact makes for example the HP X2 Chromebook to be much more appealing than let's say the thinkpad X13s
<javierm>
s/8// actually since the HP X2 has a snapdragon 7 IIRC
<aceridus>
javierm: Thanks! That looks very useful for understanding more about this topic.
<robclark>
javierm: it is basically CrOS firmware vs "everything else" unfortunately.. I'm actually a bit curious how virt runs on windows, but I guess it is a sort of sibling vm situation, where windows is a sibling vm to WSL/linux
<javierm>
robclark: that's what I thought, thanks for the confirmation
<Caterpillar>
very discouraging the things you are saying about KVM. If I knew that before I probably had bought an AMD Ryzen X1 carbon instead of the X13s ARM
<aceridus>
So, is KVM on Linux not a possibility on the Windows devices unless Microsoft and/or Qualcomm were to help?
<robclark>
I guess we would need some hints from MS about how it works (ie. confirm the sibling-vm theory and give some hints about how primary "host" OS VM can setup ringbuffers/etc with the sibling VMs that it spins up
<Libre___>
Hey, I have the Windows Dev kit 2023 as well
<Libre___>
was looking to get linux booting on it, any tips welcome
<Libre___>
(not the person from the blog)
<Libre___>
it does support virtualization, I tried booting fedora36 aarch64 inside hyper-v and it worked fine
<Libre___>
but the same iso did not want to boot on bare metal
<Libre___>
just get a Synchronous Exception
<Guest85>
where did it get stuck?
<Libre___>
and an address
Guest85 is now known as qzed
<Libre___>
after grub
<Libre___>
reading up on linux DTBs later :)
<robclark>
Libre___: yeah, windows obvious supports virt on those things.. just not in a way that linux as the "host" OS knows how to use.. with hyper-v windows is running as a guest VM as well
<qzed>
not sure about that one... but if it gets stuck on "Exiting boot services..." you'll need to set efi=novamap as kernel argument
<robclark>
maybe 8cx gen4 will support the nested-vm aarch64 extension, which I guess would make this all easier
<robclark>
PSA for anyone running upstream kernel.. when you move to v6.1-rc 28fc7c986f01fdcfd28af648be2597624cac0e27 means olddefconfig will miss CONFIG_NVMEM_QCOM_QFPROM so you'll need to add it manually
<steev>
that's probably what i missed when i was doing my testings
jhovold has quit [Ping timeout: 480 seconds]
<Libre___>
robclark: so what you're saying is that kvm/xen virtualization on most of the snapdragon devices on the market today is unlikely to be possible at all?
<Libre___>
If you at all get linux to boot on the bare metal
<javierm>
Libre___: yes, that's what he is saying
<Libre___>
aight, that's a shame
<Libre___>
but getting linux to run on the bm could be useful in itself
<javierm>
robclark: I also saw that android is exploring with a pKVM that's different, running the host in EL1 but part of KVM in EL2
<javierm>
because the threat model is different, and the guest don't trust the host so don't want to allow mapping the guest memory
<javierm>
steev: you can also check the options I needed to enable in the fedora kernel for 6.1, among those was CONFIG_NVMEM_QCOM_QFPROM that robclark mentioned
<robclark>
javierm: I believe pKVM is also based on linux kernel running crosvm in the hypervisor, with devices mapped passthru to VM, so similar to manatee (except maybe not also involving additional sibling VMs?)
<javierm>
AFAIU pKVM is the old nVHE from ARMv8.0 but with more stuff in EL2 than just the world-switch logic
<robclark>
that is possible, I haven't looked too closely at pKVM and when I did it was long ago
<javierm>
robclark: it seems to be a WIP though, so VHE / a Linux host in EL2 is what I was testing last week
<broonie>
Yeah, very much WIP
<javierm>
broonie: got it. Thanks
<steev>
javierm: your mr seems to not enable it though?
<steev>
or i'm misreading because of the way fedora does the configs
<steev>
oh, i see, it's moved out of generic
<javierm>
steev: yup
iivanov has quit [Quit: Leaving...]
ardb_ has joined #aarch64-laptops
robclark has quit [Ping timeout: 480 seconds]
robher has quit [Ping timeout: 480 seconds]
ardb has quit [Ping timeout: 480 seconds]
jhugo___ has quit [Ping timeout: 480 seconds]
ardb_ is now known as ardb
jbowen has quit [Ping timeout: 480 seconds]
jhugo___ has joined #aarch64-laptops
jbowen has joined #aarch64-laptops
robher has joined #aarch64-laptops
robclark has joined #aarch64-laptops
<hexdump0815>
aceridus: some galaxy book go updates
<hexdump0815>
it looks like your kernel config is the arm64 defconfig plus CONFIG_INTERCONNECT_QCOM_SC7180=y which is very nice, as not much more seems to be required beyond defconfig
<hexdump0815>
i noticed that the usb-c port (tested with the right one) works for me if i plugin an usb-c to multiple usb-a hub into it and the usb drive to boot from into the usb-a port
<hexdump0815>
i think i have read something similar about the galaxy book s - maybe jenneron[m] can say more about it?
<hexdump0815>
maybe we should look at the other samsung woa device bringups for hints on how to get things working - there might be some similarities maybe (which makes sense as all is from samsung)
<hexdump0815>
with usb-a to boot from and a hub on the usb-c port the system is quite useable running xfce and usb mouse and eth
<hexdump0815>
the cpu seems to be clocked at max speed at boot already - at least it looks like this accoring to 7z b results - so compiling on it should work quite well :)
<hexdump0815>
when the keyboard errors at boot appear and i hit escape i see some blank gray image with three dots on the screen so maybe the i2c traffic triggers something at firmware level? that screen disappears when hitting esc again and it continues booting
<steev>
the blank with 3 dots is likely plymouth
<Leandro[m]>
<hexdump0815> "maybe we should look at the..." <- I could probably take some A52/A72 work
<hexdump0815>
steev: this is debian bookworm, so i guess plymouth would be debian themed - looked more like some low level system screen maybe?
<steev>
unlikely, i recall that screen here too
<steev>
the x13s boots too fast for plymouth to start, and the sc8180x and sdm850 both run kali, so it uses kali's plymouth theme :(
<Caterpillar>
steev: GPU support is still on developemnt isn't it? And concerning my personal experience I am having troubles in applying that Linaro guide, as I wrote in the Google Doc comments
<steev>
yes, there are things still in development, that doesn't mean it's *unusable*
<Caterpillar>
steev: which distro/kernel are you using?
<Caterpillar>
that guide is the only one I found and it's not working. Other people running Linux are not sharing much to help other to run Linux
<steev>
debian stable, kernel is currently 6.1 rc3 with various patches from the mailing lists on top
<steev>
i'm not sure what comments are yours, and really, aside from having to find the correct fs: in efi, it has worked just fine here?
<Leandro[m]>
Hmm
<steev>
this is also the first time i saw the guide, so i'm not even sure what it says; i'll look at it more in depth later.
<Leandro[m]>
I think I'm going to get T2 to work on Galaxy book go some time
<Caterpillar>
steev: I am stuck at page 4 c. FS5:\> ls EFI\debian There is no way on my system to find DtbLoader.efi and grubaa64.efi files under such folder
<Caterpillar>
the name I use for the comments is Germano Massullo
<steev>
note that it says you may have to try other fsX:
<Caterpillar>
steev: indeed I had to use another one which is fs15
<Caterpillar>
but I can only find DtbLoader.efi in the root folder
<Leandro[m]>
Wait how do I disable secure boot on the galaxy, any exploits I have to execute?
<steev>
go into the bios and disable it?
<Leandro[m]>
...that thing has a bios menu?
<steev>
probably, mash f2/f8/f12
<Leandro[m]>
I'm... Amazed
<steev>
assuming it's a WoA device, it should
<steev>
Caterpillar: that sounds like you're not on the actual device, but on the iso
<Leandro[m]>
Tell that Microsoft from 2012 xD
<steev>
it has been 10 years since 2012 :) (despite the past couple years feeling like 20 itself)
<qzed>
luckily for us, the ballmer days are over :)
<Caterpillar>
steev: maybe you're right since I had to use chainloader(cd0,msdos2))/efi/shell/Shell.efi before because there was no way to use chainloader(hdX,msdosX))/efi/shell/Shell.efi
<steev>
did you already a) disable bitlocker b) disable secureboot ?
<Caterpillar>
steev: a)I have never enabled bitlocker b) yes
<robclark>
PSA part two, CONFIG_USB_ONBOARD_HUB seems like a thing you need to enable now too (for sc7180 chromebooks at least) <-- javierm I guess at least this matters for you
<steev>
a is enabled by default on windows pro, afaik
<Caterpillar>
steev: mine is home edition
<steev>
weird, mine came with pro
<Caterpillar>
steev: damn, the bitlocker is enabled. I now will disable it and restart everything from the scratch. Thank you for your help
<steev>
note
<steev>
when you boot back into windows, it will require you to go to some website and enter the bitlocker key
<Caterpillar>
:O
<hexdump0815>
Leandro[m]: the go has a bios with f2 and a boot chooser with f10 - disabling secure boot is easyly done in the bios - nothing special required
aceridus has quit [Remote host closed the connection]
aceridus has joined #aarch64-laptops
<Leandro[m]>
😳
<Leandro[m]>
That's awesome
<aceridus>
hexdump0815: Nice progress! I just checked my right hand usb-c and it works with a wireless keyboard through an a-to-c adapter. Not sure why I thought it didn't before, but I'll take it!
<Leandro[m]>
Dam
<Leandro[m]>
Y'all are quick
<Leandro[m]>
Can I get the current status?
<Leandro[m]>
By the time I get one you guys are almost done I think
<javierm>
robclark: yes, CONFIG_USB_ONBOARD_HUB was already enabled in the fedora kernel
<javierm>
but I noticed that without it some regulators needed for USB were disabled due being unused
<aceridus>
Leandro[m]: USB type-a and right hand type-c are working. The keyboard appears to be fully functional, but has some odd issues (incomplete descriptors on boot but a keypress clears it, periodically gets bus arbitration failures but still works). The touchpad still needs help due to an external regulator that I can't figure out how to enable. I haven't checked WiFi or GPU but will be surprised if they work straight away
<Leandro[m]>
GPU can probably be done like on A52
<Leandro[m]>
About wifi, which chip is used?
<hexdump0815>
i think gpu should be possible - travmurav[m] has it working on his 7c woa acer aspire 1, so should be doable maybe
<Leandro[m]>
Hmm
<hexdump0815>
the good thing about 7c is that there is already a lot of mainline code for the 7c chromebooks
<robclark>
javierm: I guess the needed regulator was marked always-on before 6.1
<javierm>
robclark: yeah I guess so, or not defined and left enabled by the firmware ?
<javierm>
I didn't check why started to happen in v6.1 though
<Caterpillar>
steev: your solution worked! Now I am running Linux"
<Caterpillar>
!
<alpernebbi>
hexdump0815: blank with three dots is some plymouth fallback when it has serial consoles, try having only console=tty0 in your kernel cmdline
<steev>
Caterpillar: :D
<steev>
congrats on your useful system
<hexdump0815>
alpernebbi: steev: ok, so then it seems to really be plymouth then - thanks for the info
<steev>
alpernebbi: aha! that's exactly what it is
<Caterpillar>
Windows update just updated Eliantech firmware, I don't know what it is, but it took 10 minutes in UEFI reboot to install it
<steev>
that's why i can never remmeber when i see it, because i don't often have serial console and graphics at the same time
<steev>
that's the touchpad firmware
SSJ_GZ has quit [Read error: Connection reset by peer]
SSJ_GZ has joined #aarch64-laptops
SSJ_GZ has quit [Ping timeout: 480 seconds]
<Caterpillar>
tomorrow I will run some phoronix test suite benchs