ChanServ changed the topic of #aarch64-laptops to: Linux support for AArch64 Laptops (Chrome OS Trogdor Devices - Asus NovaGo TP370QL - HP Envy x2 - Lenovo Mixx 630 - Lenovo Yoga C630 - Lenovo ThinkPad X13s - and various other snapdragon laptops) - https://oftc.irclog.whitequark.org/aarch64-laptops
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
I'm at EL1 but can address the full 64ish gigs of ram
<programmerin-wonderland>
I thought that was only addressable on EL2
<programmerin-wonderland>
Also rfkill doesn't even report my bluetooth adapter at all so I think there might be a deeper issue than the mac thing, I'm going to check my devicetree real quick
programmerin-wonderland has quit [Quit: Konversation terminated!]
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
actually I just realized I have no idea what I'm doing. I tried swapping the device tree from tj's surface laptop 7 kernel (https://github.com/tj90241/linux/releases/tag/v6.12.43rc1) and that took out wifi from my kernel entirely (and ofc didn't add bluetooth)
<programmerin-wonderland>
Also btmgmt info just says Index list with 0 items which makes me think its a different issue from the one in the GitHub issue thread
<programmerin-wonderland>
Additionally, btmgmt -i 0 public-addr a4:0d:0c:69:e4:86 returns Set Public Address for hci0 failed with status 0x11 (Invalid Index)
<programmerin-wonderland>
But it does seem like I am lacking the bluetooth seciton in my devicetree, comparing to tj's devicetree
<\[m]>
> Yes I know some Snapdragon laptops have the ability to install Linux kind-of easily... but it's not universal, and I'd still call that setup a 'hack' as it's not directly supported by Qualcomm. It's annoying, because Linaro's done a ton of work on Linux support for Snapdragon! ↩︎
<JensGlathe[m]>
He's not wrong
<SpieringsAE>
it would be awesome if qualcomm could somehow integrate the newest devicetrees into their firmware, at least thats how I think that it mostly works. Qualcomm provides a lot and then the vendors just dress it up
<JensGlathe[m]>
… aaand they don‘t, except for maybe two top-tier devices
<Treibholz[m]>
this looks interesting. But if 8-12GB RAM are sufficient, then a VM on a MacMini M4 still has the best performance/price ratio, to run linux on aarch64...
<\[m]>
I thought mac os on arm64 didn't support to use gpu over tb/usb4
<\[m]>
but yeah for hardware raid can be cool but also not
<\[m]>
I sold my external pcie because of that and I followed it up halfway through the M releases, didn't seem to be on their roadmap
<Jasper[m]>
\[m]: I think the bigger problem is the lack of drivers
<Jasper[m]>
Dunno about barspace
<\[m]>
yes the drivers indeed not os
<\[m]>
and / or hardware
<Treibholz[m]>
But I'm not sure it would work. Unfortunatelly my Lenovo Universal TB4 Dock does not work as expected with the MacMini, so I need to recable my monitors every time.
<\[m]>
I can only imagine that it's not in their business interests to make it modularly usable, want gpu, you pay apple $$$
<Jasper[m]>
Treibholz[m]: There are other issues with apple and DP MST support 😄
<Treibholz[m]>
If you want GPU with RAM, a Mac with 64GB is still cheaper than just a GPU with 64GB RAM...
<Treibholz[m]>
<Jasper[m]> "There are other issues with..." <- With my work-device (MacBook Pro 2020 - M1 Pro) the Lenovo Dock works fine.
pbrobinson has quit [Ping timeout: 480 seconds]
pbrobinson has joined #aarch64-laptops
pbrobinson has quit []
pbrobinson has joined #aarch64-laptops
jglathe_t14s has joined #aarch64-laptops
pbrobinson_ has joined #aarch64-laptops
pbrobinson has quit [Read error: No route to host]
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
Jasper: Oh I usually am not "online" but I am looking at the irc log. I'm a college student so I dont have a desktop always online but I'm typically always checking the web irc log from my phone LOL
<programmerin-wonderland>
konradybcio: It doesn't seem to be a firmware thing but a kernel thing for me
<programmerin-wonderland>
If I use tj's kernel I cant boot unless I cutmem my top 32GB
<programmerin-wonderland>
but if I use the ubuntu concept kernel all is well, even since the original (well, first I tried personally) 6.14 ubuntu concept kernel release
<programmerin-wonderland>
thats actually why I'm using ubuntu concept in the first place
<programmerin-wonderland>
But I'll check to make sure I have all my aliases in check
<programmerin-wonderland>
SpieringsAE: "it would be awesome if qualcomm could somehow integrate..." It would be even cooler if we had sane ACPI tables in the first place and just booted like x86 laptops </3
<tobhe[m]>
I don't think I have any extra magic in the kernel that would make it work, maybe a different bug?
<tobhe[m]>
The easiest way to trigger the 64GB issue is by dd-ing stuff to or from a usb drive
<tobhe[m]>
it usually results in a blue screen after a while
<programmerin-wonderland>
I haven't run into it yet but I also haven't done more than simple web browsing on KDE + Chromium. But I might try just doing work stuff on my linux partition and see how far that gets me in that case
<programmerin-wonderland>
Do we know what the EL1 bug is? (and also, how do I boot in EL2? I kind of do want KVM)
<tobhe[m]>
the bug seems to be in gunyah or rather the emulated smmu, something wrong with the address calculation but we have no details
<programmerin-wonderland>
I was going to say "it's surprising we haven't seen android or ChromeOS devices with this issue" but I can't say I've seen android or ARM ChromeOS devices with 64GB ram either
<tobhe[m]>
right. and windows gets around it by switching to el2 early
<tobhe[m]>
every EFI app should be affected though
<programmerin-wonderland>
is it just if the EFI app alloc's enough memory or is it effected regardless? Because I've not seen grub crash or anything
<programmerin-wonderland>
slight follow up question: Do qualcomm employee's know about the bug anyways?
<tobhe[m]>
grub does crash unless you are using a patched version, not enough memory more like the wrong memory
<tobhe[m]>
booting from USB is a good way to trigger the issue
<programmerin-wonderland>
oh that also might be why I can never start anything but ubuntu concept in the first place then, I cant even get grub to boot from the debian arm livecd even though it seems EFI-ish
<programmerin-wonderland>
I thought it was just broken for everyone
<tobhe[m]>
that will break other hardware though, I am working on finding a better solution
<programmerin-wonderland>
👍 is 4GB significant in any way or is it a " it's fine up to 32GB but Grub doesn't need any more than that anyways"
<tobhe[m]>
you can actually add one f to make it 32G, doesn't have to be 4G. I think the one we ship in the PPA actually goes up to 32 nowadays
<programmerin-wonderland>
Not sure if I had the chance to say this yet but I really do appreciate all the work you and the rest of the Ubuntu concept team puts in to make these devices usable on linux!
<tobhe[m]>
Thanks 🙂 much appreciated
<steveej[m]>
hey, is there any info about jhovold's github repo? it looks like it hasn't received updates in a while and i wonder whether i shall switch to a different repo for x13s sources
<steveej[m]>
i guess jhovold_defconfig will not be on mainline. so defconfig should be fine there for x13s?
<JensGlathe[m]>
you could use jhovold_defconfig since its nicely small, but ned to update things
<JensGlathe[m]>
... eventually
<tobhe[m]>
The linaro tree just patches the regular arm64 defconfig I think
programmerin-wonderland has quit [Read error: Connection reset by peer]
SpieringsAE has quit [Quit: SpieringsAE]
programmerin-wonderland has joined #aarch64-laptops
programmerin-wonderland has quit [Read error: Connection reset by peer]
programmerin-wonderland has joined #aarch64-laptops
programmerin-wonderland has quit [Read error: Connection reset by peer]
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
JensGlathe: checked my devicetree and yeah I'm missing my uart14 label thanks! I'll try patching it when I get back from classes today
ungeskriptet_ has joined #aarch64-laptops
ungeskriptet has quit [Ping timeout: 480 seconds]
programmerin-wonderland has quit [Read error: Connection reset by peer]
abelvesa has quit [Ping timeout: 480 seconds]
abelvesa has joined #aarch64-laptops
programmerin-wonderland has joined #aarch64-laptops
programmerin-wonderland has quit [Read error: Connection reset by peer]
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
alright so I added bluetooth to my devicetree properly now and now theres no error on boot after adding serial1=&uart14; to my devicetree's aliases. Still nothing in rfkill but I think that was an important prereq regardless
<programmerin-wonderland>
ooh good call let me check
<valpackett>
<Jasper[m]> "Dunno about barspace" <- the macs have the cacheable/device mapping thing pretty thoroughly screwed up. i actually wonder how it is on x1e?
<mmediouni[m]>
meh, it's better than on ampere altra in practice tbh, better to fault early than to munge data
<programmerin-wonderland>
yes many firmwares there but I dont see anything about a wcn7850 there
<programmerin-wonderland>
in /lib/firmwares/qca/ atleast
<programmerin-wonderland>
I see a wcn7850 mention in ath12k though but that makes sense and I do have wifi already
<programmerin-wonderland>
I'm a bit uneducated on the topic my apologies but what is a splat?
<abby>
error traceback
<programmerin-wonderland>
ohh
<programmerin-wonderland>
yeah I have one in my dmesg from boot too, something about a WARNING: CPU: 1 PID: 124 at drivers/gpu/drm/panel/panel-edp.c:814 generic_edp_panel_probe+0x140/0x2a0
<programmerin-wonderland>
but my gpu is working fine regardless
<JensGlathe[m]>
That would be the panel unknown
<JensGlathe[m]>
does backlight brightness work? also after suspend/resume? That's the one I'm fighting wit on the Ideapad 5
<programmerin-wonderland>
Oh I'm one of those people that just never suspends so I've never tested. Been too scared to suspend incase it breaks my session
<programmerin-wonderland>
but I can give it a shot :P
programmerin-wonderland has quit [Quit: Konversation terminated!]
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
yes! it still works after a suspend
<valpackett>
<mmediouni[m]> "not upstream (yet?) but works..." <- i thought the macs had a way worse bug than that altra erratum
<programmerin-wonderland>
yes! it still works after a suspend
<programmerin-wonderland>
oops I think that sent twice
<JensGlathe[m]>
nice
<mmediouni[m]>
valpackett: The workaround for that erratum is to go to the Mac situation w/ using device memory for PCIe BARs (to be able to trap the unaligned accesses)
<programmerin-wonderland>
it's weird because it actually works better than on windows. On windows my external monitors brightness cant be adjusted but it can be here for whatever reason
<programmerin-wonderland>
(over HDMI)
<JensGlathe[m]>
whut
<JensGlathe[m]>
nice
<JensGlathe[m]>
Ever tried Twinkle Tray on Windows?
* travmurav[m]
is still upset ddc/ci brightness control was never pushed to mainstream
<programmerin-wonderland>
No pretty much all I have installed is powertoys and I do everything else in wsl
<JensGlathe[m]>
but you can can dkms-build the driver, I did on the TB16 when I tried to get a clue on brightness there (pointless, I know now)
MelodyOwO has joined #aarch64-laptops
<steev>
programmerin-wonderland: can you paste the full dmesg?
<programmerin-wonderland>
Huh I clicked that link, firefox showed some anime fox girl and crashed
<steev>
lol
<steev>
thats the anti-ai check
<steev>
in your uart14 definition, you should probably have (above status = "okay") something along the lines of pinctrl-0 = <&uart14_default>;\npinctrl-names = "default";
<steev>
that does assume that whatever dtsi has a uart14_default defined
<programmerin-wonderland>
I see, I'll try adding that to my devicetree brb
<programmerin-wonderland>
hm I dont have a uart14_default anywhere in my devicetree
<JensGlathe[m]>
then its missing
<steev>
it should be in the dtsi
<JensGlathe[m]>
its not in mine
<steev>
uart14: ?
<clover[m]>
now i want to watch alice in wonderland
<programmerin-wonderland>
Can I just copy the uart2-default-state from the lenovo-thinkpad-x13s devicetree?
<steev>
you need the correct gpios
<steev>
you could look at other x1e devices maybe and see
<programmerin-wonderland>
:+1:
<mmediouni[m]>
On the X13s I'm getting about half the expected memory bandwidth from a GPU microbenchmark (on Linux). Is some memory bandwidth scaling stuff known to be missing?
<steev>
hm, actually, i don't see the definitions in any other x1e here
<steev>
at least, not in torvalds tree
<steev>
i should clone the ubuntu concept kernel sauce
<programmerin-wonderland>
theres a web viewer too but it doesn't have search I think
<programmerin-wonderland>
every other 80-100 device I'm seeing here seem to be using gpio116 (sample size 2 so not very reliable )
<programmerin-wonderland>
or gpio 117
<steev>
couldn't hurt to try :)
<programmerin-wonderland>
new to arm stuff, this wont damage my hardware right 😅
<steev>
i didn't break my x13s when i had the wrong info... as far as i know
jglathe_t14s has quit [Ping timeout: 480 seconds]
<steev>
(i wrote the initial bluetooth support for the x13s)
<steev>
which was all guess work :D
<programmerin-wonderland>
Well, I guess you only live once! brb
programmerin-wonderland has quit [Remote host closed the connection]
programmerin-wonderland has joined #aarch64-laptops
<programmerin-wonderland>
ultimately did nothing for rfkill
<programmerin-wonderland>
alice@alice-Microsoft-Surface-Laptop-7th-Edition:~$ sudo dmesg | grep -i bluetooth
<programmerin-wonderland>
wow
<robclark>
mmediouni[m]: looks like interconnect scaling is wired up in sc8280xp.dtsi .. not sure what you are running but there ofc plenty of ways to get suboptimal performance from a gpu
<programmerin-wonderland>
thats actually a bit of a downgrade
<programmerin-wonderland>
nothing at all in logs for bluetooth now
<mmediouni[m]>
robclark: just regular clpeak (with the latest mesa)
<programmerin-wonderland>
I added this line to uart14 pinctrl-0 = <&wcn_wlan_bt_en>; followed by pinctrl-names = "default"; wcn_wlan_bt_en was already defined in the file
<steev>
as you are finding out, that is not correct :)
<programmerin-wonderland>
sad but true lol
<alexVinarskis[m]>
You should be able to get these from dsdt dumps. Search anything uart related, my device had Bluetooth node that noted which uart it is and a gpio iirc.
<programmerin-wonderland>
actually is there any info in windows device manager that might be relevant here
<programmerin-wonderland>
dsdt is acpi table stuff right?
<steev>
yes
<robclark>
mmediouni[m]: which clpeak benchmark?
<alexVinarskis[m]>
Yes, in windows dev manager you can find uart node in dev properties, though so far i think its always uart14?
<mmediouni[m]>
Am observing ~28.5GB/sec of memory bandwidth when I'd have expected double that
<alexVinarskis[m]>
programmerin-wonderland: then maybe compare that dsdt to one from your device, anything uart/Bluetooth related. If its 1:1 same, then perhaps issue is something else
<programmerin-wonderland>
yeah I'm only missing one section so I'm going to try that and if it doesn't work I think I'll have to do some more digging
<programmerin-wonderland>
it's the &pcie4_port0 section
programmerin-wonderland has quit [Remote host closed the connection]
programmerin-wonderland has joined #aarch64-laptops
<robclark>
mmediouni[m]: --global-bandwidth?
<programmerin-wonderland>
yup still no dice, going to have to look a bit deeper then
<mmediouni[m]>
robclark: Yes
<steev>
programmerin-wonderland: might be silly but maybe check the firmware that windows loads is actually there in /lib/firmware/qca ? I know someone asked, but e.g. for a long time the bluetooth firmware for the x13s had to be pulled from windows
<Jasper[m]>
(otherwise the qualcomm-reference-drivers github repo and cabextract is the way to go)
<programmerin-wonderland>
I don't see any files that contain `wcn7850` in qca I thought the firmware for that would be the same as the ath12k
jglathe_t14s has joined #aarch64-laptops
<robclark>
mmediouni[m]: hmm, I'm seeing ~53GB/s
<mmediouni[m]>
robclark: I'm testing on mesa 25.2 with the Ubuntu 25.10 kernel (6.16). Is that too old?
<mmediouni[m]>
And it's a 32GB RAM model if that matters
<programmerin-wonderland>
\[m]: "Hey… quick question, why are anime catgirls blocking my access to the Linux kernel" I ask myself that every day
<steev>
programmerin-wonderland: i have no idea what the proper firmware would be :) but you can look for similar files in the windows system32 directory somewhere
<robclark>
same as mine
<programmerin-wonderland>
alright
<programmerin-wonderland>
going to look in driverstore and hope to find a clue
<robclark>
mmediouni[m]: install nvtop, and use that to check if gpu utilization hits 100%?
<programmerin-wonderland>
thanks for all the help everyone!
<programmerin-wonderland>
"./qcbluetooth8380/BSRC_BT.bin" Well that's a pretty suspicious file name there microsoft
<robclark>
mmediouni[m]: hmm, ok.. I'm actually on a bit older kernel + mesa on my x13s (use my x1e laptop most of the time these days).. I'd have to go back and check if I had any extra patches in my kernel build (and test a newer mesa), but that will have to be later today
<steev>
robclark: i also see ~25 here with debian testing (25.0.7)
<robclark>
idk if it matters but fedora has clpeak v1.1.4... there have been some rusticl fixes to avoid stalls which might matter.. and in general a bunch of fixes since 25.0.x
<robclark>
(25.1 is already eol ;-))
<clover[m]>
steev now that we are both laid off lmk if you want to start a company together :D
<steev>
clover[m]: oh shit
<steev>
robclark: yeah, 25.2.2 is in unstable but trixie release and whatnot
<f_>
Steev & Clover, Mainlining consultancy :>
<f_>
clover[m]: laid off from kali too?
<steev>
they're an arch btw kinda guy, not kali
<f_>
ah right :p
<f_>
as you can see I don't watch this channel very often
<f_>
xD
<steev>
:D
<steev>
to be fair, the kali team was only 6 of us so
<Jasper[m]>
@clover all the strength to you too
<steev>
now they're 5
<f_>
clover[m]: yes, best of luck in finding a new job
<f_>
steev: excited for when they get the count to 3
<f_>
and then 0
<f_>
(seriously it was quite rude)
<steev>
no animosity for the kali team themselves, the call was from offsec
<f_>
yeah I know
<clover[m]>
oh im fine, excited to put more time towards open source stuff
<f_>
Whatever you end up doing I wish you glhf
<f_>
being laid off is not fun
<gwolf>
Oh, I was unaware Kali was _not_ a community project...
<gwolf>
(or rather, why should have I assumed so? Well, selection bias perhaps? ;-) )
<steev>
it... is a weird spot, for sure
<steev>
offsec was created to keep kali always available for free, and not having to rely on donations where people may try to exert pressure to remove a competing tool or some such
<Jasper[m]>
Instead they live off my neverending OSCP attempts :(
<steev>
where are you falling short? AD set? methodology? enumeration?
<Jasper[m]>
I 60/100'd both attempts, but I figured out what I read over in the AD set that gets me access to something
<Jasper[m]>
Let's just say the facepalm I did still echoes across the world
<steev>
:D
<Jasper[m]>
I read over a very blatant credential combo in my second winPEAS output
<Jasper[m]>
The general set goes well every time basically
<robclark>
mmediouni[m], steev: while running `clpeak --global-bandwidth` can you `cat /sys/kernel/debug/interconnect/interconnect_summary | grep gpu` ?
jglathe_t14s has quit [Remote host closed the connection]