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
Mathew has quit [Quit: Leaving]
mcbridematt has joined #aarch64-laptops
<steev>
6.11-rc1 is tagged
<steev>
clover[m]: i pushed 6.10.2
hexdump0815 has joined #aarch64-laptops
hexdump01 has quit [Ping timeout: 480 seconds]
<clover[m]>
Cool. 16k?
<steev>
nope
<steev>
that's a debian config thing; i might bring back the laptop_defconfig, if you wanna try to give that a whirl for if it breaks your stuff again
<steev>
but that wonn't be til later this week
<steev>
there's still some software out there that doesn't like 16K pages - i think asahi keeps a list somewhere in their wiki, and, it seems like some things are unhappy with vulkan on freedreno when you use them too
<HdkR>
Huh, I wonder why Vulkan things would break
<steev>
that part i'm not sure at all, and i need to ask the software peeps how to get more debug info into the log they have
<steev>
but the basic gist is, when it does memory check on the vulkan device, with 16K pages, it just reports that it's out of host memory
<HdkR>
Weird
iivanov has joined #aarch64-laptops
<lollaritits[m]>
<steev> "clover: i pushed 6.10.2" <- x13s?
iivanov has quit [Quit: Leaving...]
iivanov has joined #aarch64-laptops
<steev>
lollaritits[m]: yeah
<steev>
i don't have an x1e :)
<lollaritits[m]>
ah good
<steev>
though i'm guessing johan's stuff is gonna start having the x13s bits that still haven't made it in, as well as x1e stuff for 6.11
<lollaritits[m]>
finally no more outdate 6.10 kernel
<lollaritits[m]>
like a real archlinux user 🫣
<lollaritits[m]>
making kernels on the InterCityDirect to Amsterdam 🫣
<steev>
heck yeah
<steev>
sadly, i'm pretty much at my place, or my sister's
jhovold has joined #aarch64-laptops
<lollaritits[m]>
im currently on my way to Antwerpen Centraal to get the ICD
<gwolf>
steev: and... do you have any tree for linux >= 6.7 for the C630? :-)
<steev>
gwolf: kind of :)
<steev>
debian's 6.10 should *mostly* work, but you might need to blacklist ipa, which would suck if you use the modem
<steev>
no audio though
<gwolf>
OK! That is great news. I will try!
<steev>
lumag has done a couple of merge requests recently that get most of the qualcomm stuff up to snuff, but for some reason there's an rcu stall when ipa comes up, something to do with coresight i think he said, but i'm not sure
<gwolf>
For blacklisting, I gues mentioning it in modprobe.d should work?
<gwolf>
I guess I still need to do the install-dtbs whenever I update or switch a kernel, right?
<steev>
not sure, honestly. i just do modprobe.blacklist=ipa on the command line, if modprobe.d works, use it :)
<steev>
yeah - actually let me double check
<gwolf>
OK! I'll play with it :-)
<steev>
ah, no
<steev>
the EC stuff is still outstanding
<steev>
maybe it's in 6.11?
<steev>
yah, 6.11
<steev>
i really need to go to bed (2am, meeting at 9am) but tomorrow i'll push my debian kernel stuff to salsa
<steev>
it's 6.10 but with the thinkpad and c630 stuff applied (but still no audio and the ipa thing)
<gwolf>
right, Korean time is not compatible with you :-)
<gwolf>
OK, I'll wait for .11
<gwolf>
thanks as always!
<steev>
i would like to know why it keeps spamming Starting disk, but yes, bed first
<steev>
oh wonderful :) it happens on the c630, which has ufs :)
<ema>
steev: hey! Nope, I haven't tried 6.10 on the X13s yet. Is it just the wifi mac address not working as usual or is there anything else you've noticed?
<ema>
"not working as usual" -> heh, language is ambiguous :-) "not working the way it usually does" is what I meant
<JosDehaes[m]>
could you run the upstream 6.11-rc1 tree on yoga 🤔
<\[m]>
so I called today to lenovo belgium, the last the sales person heard was beginning of August the t14s will be available regionally ánd configurable
<HdkR>
Oh, that's nice
<JosDehaes[m]>
will they have another keyboard option than azerty?
<\[m]>
which I guess might be indicative of a global availability ? since you can't order the t14s here yet
<\[m]>
JosDehaes[m]: they usually have an option, EU English
<Jasper[m]>
\[m]: Very cool
<JosDehaes[m]>
well none of the X1E laptops for sale now have any other option than belgian azerty
<JosDehaes[m]>
(except for dell)
<Jasper[m]>
The Dutch store may sell ansi with eur on 5
<JosDehaes[m]>
but then they don't ship to Belgium
<JosDehaes[m]>
In 2018, I wanted to buy a lenovo, and in Belgium I could NOT order with qwerty (I called them even), and in Netherlands I could not get it to ship to Belgium. I was ready to give them money and they weren't able to accept it
<JosDehaes[m]>
(and I want US qwerty, not intl, with the backslash/pipe above the enter)
<konradybcio>
you can use shipping forwarding services
<konradybcio>
i.e. you ship to a NL address and they ship it onwards to you
<JosDehaes[m]>
I bought my yoga slim 7x on eBay from a US seller 😁
<lollaritits[m]>
<\[m]> "so I called today to lenovo..." <- fancy
<JosDehaes[m]>
ok, so v6.11-rc1 boots, but doesn't seem to work (I had no kb/mouse at GDM screen, but it seemed the machine froze)
<FarchordSteveCossette[m]>
<JosDehaes[m]> "ok, so v6.11-rc1 boots, but..." <- XD I have a similar situation here. But I can use an external keyboard and mouse
<FarchordSteveCossette[m]>
.... plugged in through a USB-C hub XD
<\[m]>
<JosDehaes[m]> "(and I want US qwerty, not intl,..." <- dude is that an option !?!!
<\[m]>
I thought this was just random
<\[m]>
maybe diff keyboard comes with configurability
<\[m]>
very familiar with this, I drove to Netherlands to get my first macbook air m1 with qwerty keyboard, impossible to get it shipped from coolbue even - so close yet so far
<\[m]>
thought I actually don't care, it was for resellability
<\[m]>
but that enter / backslash layout I prefer US layout then I think
<JosDehaes[m]>
I got lucky with the only second hand yoga slim 7x available
<JosDehaes[m]>
Even with shipping and customs it was only about 1000 euro
<\[m]>
haha with your story, I've been checking ebay too - but I guess I'm glad in the end, I just want 14 inch thinkpad really
<\[m]>
but for that keyboard layout, could you reship from USA? might be an option though it'll be premium price
<FarchordSteveCossette[m]>
<\[m]> "but for that keyboard layout..." <- There may be companies that do that. Or you would have to trust an individual to buy it for you and then ship it to you
<FarchordSteveCossette[m]>
But there'S some negatives to it: For example, you'd need to pay the local US/CA Sales tax, and then the shipping to the Netherlands
<FarchordSteveCossette[m]>
And if you need warranty, in most cases you'd be SOL
dump_stack has joined #aarch64-laptops
<FarchordSteveCossette[m]>
Oh and ofc customs fees
<JosDehaes[m]>
eBay global shipping arranges the VAT for you
<JosDehaes[m]>
no customs
<FarchordSteveCossette[m]>
oh yeah we're talking outside of ebay
<FarchordSteveCossette[m]>
or at least I am
<dump_stack>
\[m]: bhphotovideo ships to the EU, I got T14s from them a few weeks ago.
dump_stack has quit [Quit: Page closed]
<steev>
ema: neither, sorry i forgot to ping the fix back - ip no longer carries the legacy /usr/sbin/ip symlink, so the udev rule needed to be fixed to be /usr/bin/ip :)
<ema>
steev: good to hear!
<ema>
wiki page updated, thanks steev
<steev>
:)
zdykstra has quit [Ping timeout: 480 seconds]
<jhovold>
Here's an updated wip branch for the X13s:
<jhovold>
- fix battery thermal zone warnings (regression in 6.10 final)
<jhovold>
- fix broken efivars support (6.11-rc1 regression)
<jhovold>
- fix PCIe4 PHY supply
<jhovold>
- fix PCIe domain numbers
<jhovold>
- fix PCIe perst pull-down leakage
<jhovold>
- fix LLCC settings
<jhovold>
- add support for PCIe5 and CRD modem
<jhovold>
- enable 4-lane PCIe for NVMe
<jhovold>
Notes:
<jhovold>
- tcsrcc_x1e80100 must now be included in initramfs (or built in)
<jhovold>
- in-kernel pd-mapper is not enabled as it causes regressions
<steev>
konradybcio: it's still only 27 patches on top so far :D
<konradybcio>
jhovold could you double check that the adsp fw is loaded after kernel pd mapper? maybe we're catching it in a bootloader-initialized state or something
<konradybcio>
s/after/before
echanude has joined #aarch64-laptops
<abelvesa>
robclark: (and everyone else who was using my kernel tree for x1e): please use jhovold's tree from now on
<jhovold>
konradybcio: pd-mapper appears to be started before the adsp fw is up
<jhovold>
so could be some kind of race there
<steev>
konradybcio: oh thanks for the 16K page workaround on sdm845, i actually had booted it last night and saw the same thing and figured i'd look at it today after work
<konradybcio>
jhovold what if you call qcom_scm_pas_shutdown(1) before pd-mapper is registered :P
<konradybcio>
steev tips fedora
<jhovold>
I'm leaving that for someone else to debug while I go on holiday
<lumag>
it is fine to start pd-mapper before the *dsp
<lumag>
I see. I don't have x1e8, so somebody would need to debug that.
<jhovold>
don't you then risk quering the fw based on what the boot fw was running as konradybcio suggested?
<lumag>
Querying for what?
<jhovold>
lumag: sgerhold may be able to help with debugging
<lumag>
pd-mapper just provides the maps. It is the firmware who queries pd-mapper for the data
<lumag>
and for the expected services
<jhovold>
ok, but changes in timing means that drivers may not start talking to the fw sooner than before
<jhovold>
i'll leave you guys to sort it out, but do note that I saw a new error once also on the X13s which may be related to this regression
<jhovold>
this shoud not be x1e80100 specific
<lumag>
jhovold, I'm running pd-mapper on my x13s, no issues observed
<jhovold>
how often do you scan the log for new errors on boot? ;)
<jhovold>
but that X13s could be unrelated, and random changes in timing could impact this (e.g. kernel config)
<jhovold>
that X13s error*
<jhovold>
something is broken in any case
<JensGlathe[m]>
hmm try to boot from a type-c stick (also in swapped config) and you get some different timings
<Jasper[m]>
<jhovold> " - fix battery thermal zone..." <- Will this one be a bugfix for 6.10 aswell upstream?
<Jasper[m]>
Not keeping watch on the mailing list enough, how do you even subscribe to these?
<FarchordSteveCossette[m]>
You know, i thought this channel was just full of enthusiasts trying stuff to make those laptops work… i just noticed a bunch of you here are linaro devs/employees lol
<Jasper[m]>
And which ones are interesting in this case
<Jasper[m]>
FarchordSteveCossette[m]: These things are mutually exclusive :p
<lumag>
jhovold, I usually see them, thanks for you making sure that boot log is quiet
<Jasper[m]>
FarchordSteveCossette[m]: Many combo's possible! Though I don't think there are people here that work at Linaro and hate their aarch64-laptops related job
<Jasper[m]>
Unless you do, and I'm sorry for any lurkers that fall into that category
echanude has joined #aarch64-laptops
<FarchordSteveCossette[m]>
Well, I know it’s your jobs, but nonetheless, thank you for your work! 😊
<sgerhold>
lumag: If the firmware misbehaves based on slight pd-mapper timing differences on X1E then we probably have little choice but to workaround it somehow :/
<sgerhold>
I'll have to try reproducing it myself and see if I can make it go away
<lumag>
sgerhold, thanks!
<steev>
Jasper[m]: you can use lei to watch stuff you're interested in
<steev>
without having to subscribe to mailing lists
<steev>
you can also check the different patchworks and keep in mind the qcom maintainers for each
<steev>
FarchordSteveCossette[m]: for the record i'm a linux dev, but distro, not kernel (well i have some kernel commits, but i'm not a professional kernel dev, definitely an enthusiast who bumbles his way around things)
<FarchordSteveCossette[m]>
If you dont mind me asking, which distro?
<steev>
kali
<steev>
i'm very transparent about it :)
<abby>
yeah it's a good mix of linaro, distro, and enthusiasts here
<Nios34[m]>
steev: What is lei 👀
<steev>
i don't think there's an actual webpage for it, but https://josefbacik.github.io/kernel/2021/10/18/lei-and-b4.html explains it pretty well (while it's centered on paying attention to things for development, end users can definitely have a look over things!)
<steev>
basic gist is, you say "i care about these things" and it pulls down all the related patches to "these things"
<robclark>
jhovold: hmm, with wip/x1e80100-6.11-rc1 I'm not getting battery status.. is that related to pd-mapper?
<steev>
it is
echanude has quit [Quit: WeeChat 4.3.5]
<robclark>
hmm, I do still have CONFIG_QCOM_PD_MAPPER=y in my local config
<steev>
check if you're getting the messages that johan mentioned? here, at least, i don't see them, with PD_MAPPER=m
<robclark>
not seeing those msgs..
<robclark>
HdkR: side note.. gallium a7xx landed yesterday, you probably want to re-cherry-pick the patches since I fixed a few things before they were merged
<sgerhold>
robclark: the pd-mapper changes for X1E weren't applied upstream yet, so they're probably not in Johan's branch yet if he had regressions
<robclark>
hmm, I have the config option.. maybe I don't have the patch adding x1e
* robclark
checks
<craftyguy>
jhovold: does bad stuff happen if I start the userspace pd mapper thing when using 6.11-rc1?
<sgerhold>
yeah the config option is in 6.11-rc1, but the x1e addition is not
<steev>
craftyguy: in my experience they can co-exist
<sgerhold>
I don't have any of those and it works fine for me :o
<robclark>
probably not all of it needed.. I did have non-working kb/touchpad without evdev
<robclark>
and ath12k I needed
<robclark>
tbh I'm not sure what is supposed to clue the kernel in about what modules to load.. so could be somethign wrong on my end
<sgerhold>
robclark: the kernel usually sends uevents with module aliases (modalias, e.g. the DT compatible) and then udev loads the modules
<sgerhold>
it should just work automagically
<robclark>
could be something broken with fedora?
iivanov has quit [Remote host closed the connection]
<robclark>
or maybe just some linux-next breakage at the time.. although evidence from this morning suggests I at least still need the last line
<FarchordSteveCossette[m]>
robclark: You remember the issue I had with the system no longer booting? I'm trying to pinpoint it right now, but how do you generate your initramfs when you install a custom kernel?
<FarchordSteveCossette[m]>
I usually put the kernel in /lib/modules/, depmod -a, kernel-install add <path to kernel> <path to vmlinuz> but doing this, the new image will refuse to boot
<FarchordSteveCossette[m]>
Now just gotta figure out how to "install" the kernel in a custom location (As I don't want it to overwrite my system'S kernel lol)
zdykstra has joined #aarch64-laptops
<steev>
tag it with a different version string
<steev>
LOCALVERSION="-farchord " or some such
<FarchordSteveCossette[m]>
steev: Oh so that will put it in my folders but I can simply move it out..... right?
<steev>
that will set the version string to X.Y.Z-farchord and it would default to installing to /lib/modules/X.Y.Z-farchord
<robclark>
FarchordSteveCossette[m]: I guess the custom kernel isn't the same version as distro... but otherwise use LOCALVERSION as steev suggests
<steev>
not sure what you mean by move it out
<FarchordSteveCossette[m]>
@steev: I'm building on my Ryzen machine, then I'll move it over to my USB-C SSD
<FarchordSteveCossette[m]>
I assume make install will move it to /lib/modules/
<steev>
if you're installing it on your local machine, yeah, but you could do like rob where you set O= so that it sets an output directory, instead of just installing it into /lib/modules
<FarchordSteveCossette[m]>
I thought O= is the arch?
<steev>
it is not
<FarchordSteveCossette[m]>
ok, how do I make sure I'm building the aarch64 version? XD
<steev>
make ARCH=arm64 CROSS_COMPILE=whateverfedoradoes LOCALVERSION=-farchord O=aarch64 ...
<FarchordSteveCossette[m]>
Cool thanks! Just gotta figure out "whateverfedoradoes" :) I'll google it!
<steev>
that's the cross compiler's prefix
<steev>
on debian it's aarch64-linux-gnu-
<steev>
it's probably the same, but i don't know fedora at all, sorry :)
<steev>
you can also point it to a specific path that has a cross compiler extracted, assuming you aren't using the distro one
laine__ has joined #aarch64-laptops
<robclark>
FarchordSteveCossette[m]: alias amake='make INSTALL_PATH=`pwd`/aarch64/target O=aarch64 INSTALL_MOD_PATH=`pwd`/aarch64/target CROSS_COMPILE=aarch64-redhat-linux-'
<FarchordSteveCossette[m]>
robclark: thanks lol
<robclark>
the INSTALL_PATH/INSTALL_MOD_PATH will get it to install kernel somewhere other than where your host kernel is
<FarchordSteveCossette[m]>
robclark: mmm says the redhat compiler is not found lmc
<steev>
make sure you have the cross compiler installed
<FarchordSteveCossette[m]>
ahh not same thing, one sec
<robclark>
note that I'm assuming your host system is fedora also
<steev>
but if you're building a kernel on your ryzen, and not using a cross compiler, and not specifying arch, it's probably building an amd64 kernel
<FarchordSteveCossette[m]>
robclark: Yep it is
<FarchordSteveCossette[m]>
The compiler tag was wrong
<steev>
can usually figure it out by just running aarch64-<tab> (assuming your shell has tab completion)
<FarchordSteveCossette[m]>
You are building kernel with non-retpoline compiler.
<FarchordSteveCossette[m]>
GEN arch/x86/include/generated/asm/orc_hash.h
<FarchordSteveCossette[m]>
Please update your compiler.
<FarchordSteveCossette[m]>
lol
<robclark>
O=arm64
<robclark>
or rather... ARCH=arm64
<robclark>
sorry I left that out, I simplified the alias once I no longer needed to cross compile
<robclark>
FarchordSteveCossette[m]: ^^^
<FarchordSteveCossette[m]>
Alright.... building XD thanks robclark
<FarchordSteveCossette[m]>
I added the alias in the zshrc I figure I'll need it again later lol
* FarchordSteveCossette[m]
wonders how long compiling will take on a r9 7900x.
<robclark>
compile time really depends on # of drivers enabled
<robclark>
I was curious why defconfig kernel compile was much slower on x1e than zen5... then I realized arm65 defconfig enables ~6.5x as many drivers
<steev>
that's why johan makes his defconfig only do the stuff needed for the device(s)
<FarchordSteveCossette[m]>
Yep and I used that
<FarchordSteveCossette[m]>
compile got done in 15mins XD
<FarchordSteveCossette[m]>
robclark: so, this is what I ended up getting, and sorry in advance irc people:
<FarchordSteveCossette[m]>
Am I right to assume I can simply bring that over to the target system, and from the build folder, make my initrd/...?
<FarchordSteveCossette[m]>
amake install and module_install fail because it can't find the modules in /lib/modules/<version>
<robclark>
hmm, it has been a while since I had to cross-compile, I think when I was I was extracting initrd, copying over new modules, and then re-cpio'ing it
<JosDehaes[m]>
you can do this: INSTALL_MOD_PATH=~/tmp/initrd-clean/ make modules_install
<robclark>
right, I'm assuming he is doing that part.. but then how to get modules into initrd?
<robclark>
(I vaguely remember at some point I had to strip debug syms from modules to prevent initrd from being too big.. OTOH then I was dealing with a board that had a lot less RAM)
iivanov has quit [Remote host closed the connection]
<FarchordSteveCossette[m]>
<JosDehaes[m]> "you can do this: INSTALL_MOD_PA..." <- yeah install_mod_path is already in the alias robclark gave me. Essentially, it copies all the drivers, but fails when it tries to run dracut
<FarchordSteveCossette[m]>
because the kernel on this is something else XD
<JosDehaes[m]>
no idea about dracut, I stopped using redhat based distros in 1999 :-D
<FarchordSteveCossette[m]>
No problem, thanks anyway!
<steev>
maybe on the box just dracut --regenerate-all ?
<FarchordSteveCossette[m]>
Thanks! Yeah I checked my current kernels and I think right now I got a decent footing. I'll try it later!
<craftyguy>
Where is sysrq on the x13s keyboard?
<craftyguy>
I found some reference to Fn+S but either I don't have magic sysrq enabled or it's not the right key :/
<steev>
make sure it's enabled in the kernel
<craftyguy>
yeah, hard to do when I need it 😅
<craftyguy>
will check after rebooting the harsh way
<craftyguy>
ah it's not configured in the kernel :(
<steev>
you could ask johan to enable it in his defconfig :)
<craftyguy>
it's fine, I'm maintaining my own config in pmOS for it
kalebris_ has joined #aarch64-laptops
iivanov has joined #aarch64-laptops
kalebris has quit [Ping timeout: 480 seconds]
kalebris_ is now known as kalebris
iivanov has quit [Read error: No route to host]
iivanov has joined #aarch64-laptops
iivanov has quit [Ping timeout: 480 seconds]
kettenis has quit [Remote host closed the connection]
iivanov has joined #aarch64-laptops
iivanov has quit [Ping timeout: 480 seconds]
iivanov has joined #aarch64-laptops
iivanov has quit [Ping timeout: 480 seconds]
nirik[m] is now known as nirik
<nirik>
my slim 7x arrived. :) poked around seeing if I could get it booting with a fedora rawhide usb... well, sort of, it boots and I get to a point where it can't find the rootfs. I thnk it's resetting usb or something... I have a nvme showing up tomorrow, will swap that in for the windows one and see if I can get further.