robclark 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
<gabertron>
steev: not sure what WiFi issue you are referring to but are you seeing flakey WiFi? Maybe try: iw dev wlan0 set power_save off
<steev>
gabertron: the bug linked, it's something we've seen for a long time, it's not related to power saving
<gabertron>
Oh ok
<steev>
it started (became far more prevalent) around 6.9?
<steev>
fwiw it's already off :)
<gabertron>
Ok great :)
<krei-se>
oh hi it's the same guys as on libreboot
alfredo has joined #aarch64-laptops
jhovold has joined #aarch64-laptops
alfredo has quit [Read error: Connection reset by peer]
<jhovold>
steev: I haven't heard anything more about the wifi issue (despite a reminder on list)
<steev>
i've noticed it happening a whole lot more in the 6.10 rc era
<JensGlathe[m]>
~ once a week maybe, not very often
cyrinux has joined #aarch64-laptops
<jhovold>
I still haven't hit it on my X13s during normal use (only saw it twice when I did some wifi benchmarking with GIC ITS last autumn)
<albsen[m]>
<jhovold> "I still haven't hit it on my..." <- different models of x13s have different behaviors. the 16gb 5g version never had any issues. the 32gb (non 5g) has it constantly
Guest912 is now known as qzed
<steev>
huh, i'm actually seeing the messages right now, but the system has been idle since i logged it in an hour an 10 minutes ago
<steev>
usually when i see them though, the wifi disconnects
<JensGlathe[m]>
there is a between phase where performance degrades
<steev>
kernel wise there's just nothing though... 01:11 i get the rfkill input handler disabled message, and then 01:47 it starts spitting out the insufficient length
<steev>
i wouldn't know because it was just sitting there open, logged in to gnome but display was off (nothing but the power adapter plugged in)
<steev>
is the performance degredation from it spamming the hell out of dmesg? because it has been spamming said message since 01:47
<JensGlathe[m]>
yes it usually does
smpl has joined #aarch64-laptops
<jhovold>
albsen[m]: my x13s has also got 32 GB and no modem, even if it's a preproduction machine
<steev>
jhovold: are you letting it sleep, or do you keep it powered on and not sleep (i mostly do the latter... if it's plugged in it doesn't sleep, if it's not, it does)
<jhovold>
I suspend it when I move it around or during lunch etc, but I tend to power it off at night
<steev>
i should do that more
srinik has joined #aarch64-laptops
mcbridematt has quit [Quit: Leaving]
<albsen[m]>
<jhovold> "albsen: my x13s has also got..." <- interesting, no clue why its different between models (and even the same model). I noticed it on the 32gb one due to me migrating the ssd over from the 16gb to the 32gb model. having said that, since 6.10.rc3 I only had to forcefully restart and reload the wifi chip once. so, maybe its actually fixed and this was just one odd one out
<jhovold>
albsen[m]: I don't think it's fixed, there's a race somewhere that triggers this and random changes in timing may hide or make the problem worse
<jhovold>
is what I think
mcbridematt has joined #aarch64-laptops
jhovold has quit [Quit: WeeChat 4.2.1]
ektor5 has quit [Quit: WeeChat 4.3.4]
ektor5 has joined #aarch64-laptops
hightower3 has quit []
<spawacz>
My x13s just arrived!!!
<spawacz>
Do you have a guide on setting Linux on that?
<spawacz>
I see Debian page on x13s. I usually debootstrap the system from another system/usb. I wonder if I could plug the disk to my other machine, bootstrap there and just plug it back to the thinkpad? I don't have a USB-A stick
<JensGlathe[m]>
I assume you mean type-c stick. you could unpack the image onto the ssd (if you remove it from the laptop anyway), it would boot. But this a little out of scope for this image... there's also an official Ubuntu x13s image (23.10 works) that gives you a real installer
<JensGlathe[m]>
But for the official ISO you need a type-c stick, a USB hub or an external ssd
<JensGlathe[m]>
I would recommend any of this for the image, too
craftyguy has quit [Remote host closed the connection]
craftyguy has joined #aarch64-laptops
craftyguy has quit [Remote host closed the connection]
craftyguy has joined #aarch64-laptops
jhovold has joined #aarch64-laptops
srinik has quit [Remote host closed the connection]
srinik has joined #aarch64-laptops
<kettenis>
so it seems that for the vivobook s15 the battmgr info notification message gained another 32-bits at the end
<kettenis>
I'm also seeing messages from an "unknown
<kettenis>
" glink owner 0x800e (32782)
<kettenis>
anyone around here who would be able to figure out what those mean?
<spawacz>
should I keep windows on the x13s for firmware updates etc, or Debian will suffice?
<robclark>
spawacz: I _think_ all the /lib/firmware stuff is upstream.. and there is a way to update uefi fw without windows but I'm failing at remembering where the github tree was for that
<jhovold>
the venus video acceleration firmware is not yet in linux-firmware
<jhovold>
and needs to be copied from windows for now, if you want to enable that (support is only in my wip branches due to the missing fw)
<jhovold>
spawacz: ^
<travmurav[m]>
btw why can't it be submitted "in advance" with the firmware-name of a yet-non-existent blob? then when it finally lands it will load fine and until then the driver will just safely fail to probe like on all other devices when the fw is missing?
<robclark>
imho, if you wait for the fw you might be waiting forever
<robclark>
I guess lenovo has moved on from the x13s at this point
<jhovold>
most people will fail to notice that the interconnect sync state will not be reached and lose hours of battery time
<jhovold>
when the fw is missing
<jhovold>
can be addressed on the command line and in the kernel config, but someone is bound to not do so
srinik has quit [Ping timeout: 480 seconds]
<Jasper[m]>
Welp I hope it lands at some point
<Jasper[m]>
I'm not too jazzed about never having upstream support for something like that :/
<jhovold>
robclark: we're not waiting on lenovo, it's qualcomm that's blocking the release, supposedly to be released "soon"
<robclark>
oh, ok.. well that is good
<robclark>
that said, maybe someone should figure out a fix for sync state when fw is missing.. because there is always bound to be someone somewhere that is missing fw
<jhovold>
agree, someone should, there's just too many things to do and not always someone that's willing to pay for everything
<jhovold>
at least not at once
<konradybcio>
robclark this is not always feasible
<konradybcio>
for example we can't really pull the plug on pcie buses because xbl enabled these and the devices are poking the host
<konradybcio>
(or expect the link to be up or equivalent.. the symptom is that the platform dies if we dont shut things down cleanly)
<robclark>
well, I mean we should be able to at least make it not worse than it would be with driver disabled in kconfig
<jhovold>
it's the same issue there
<jhovold>
if the driver fails to probe for whatever reason you will not hit the sync state (depending on your config)
<jhovold>
currently we err on the safe side by default at the cost of power consumption when something is missing
<robclark>
I mean, what are the odds of convincing _every_ OEM to push their signed fw to l-f?
<robclark>
it would be nice if future things allowed unsigned fw with protected-content features disabled... that would make this less of a mess
srinik has joined #aarch64-laptops
alfredo has joined #aarch64-laptops
<travmurav[m]>
jhovold: ah right so the forever deferred on firmware driver will hang the votes, this is kinda sad but makes sense then, right...
<jhovold>
it's related to all the hand over and unused resources discussion, I talked a bit about this at kernel receipes last year
<jhovold>
"someone" should address all of this, just need to find the time and resources
<robclark>
is venus really on when kernel starts? Display I could understand, but I'd expect most other things to be off
<jhovold>
it's off, but the interrconnect never reach the sync state until all consumers in DT have been successfully probed
<jhovold>
so we'll be running at whatever power state the boot fw left us with even when we could go lower
<robclark>
maybe userspace could tell interconnect when to give up on deferred devices after rootfs pivot?
<jhovold>
right, that's one approach (a listed a few in my talk)
<robclark>
ahh
<jhovold>
but we probably need a timeout as well, can't depend on user space generally
<jhovold>
and we should probably use the same mechanism for unused clocks and power domains as well
<jhovold>
so that we can get rid of clk_ignore_unused etc
<jhovold>
I hope I can get a chance to look at it in a not too distant future
<robclark>
yeah, I suppose clk/pd thing with _ignore_unused would follow the existing precedent
<jhovold>
that bit is still unresolved, so we should be able to settle on a common mechanism and policy
<jhovold>
regulators also has it's own (the 30 s timeout, simple and get's the job done most of the time)
* travmurav[m]
had a hack patch for 30s timeout for clk to workaround that old issue with clk stuck in mdp somewhere
alfredo has quit [Quit: alfredo]
<JensGlathe[m]>
<spawacz> "should I keep windows on the..." <- For X13s its sorta safe to not keep the Windows install. I keep it, though, also for updating firmwares when they appear.
* colemickens
was dreaming about protesting laptops until they support LVFS, but I probably won't live that long.
<spawacz>
do I need that firmware from my own thinkpad or I can download it from somewhere?
<craftyguy>
ah ok, I saw it mentioned above and wasn't sure if it was still required if the venus fw was available, but I just got confused :P
<craftyguy>
spawacz: I'm grabbing it from lenovo, and extracting it
<jhovold>
I think it can be extracted from files from the lenovo webpage, someone may have uploaded it somewhere (even though they probably should not)
<steev>
yeah, it can be downloaded from lenovo's support page for the x13s and innoxtract used on it
<craftyguy>
grab the .exe from lenovo, use "innoextract -I drivers/qcvss8280.mbn /path/to/qcom/graphics/driver.exe"
<spawacz>
what is missing without it?
<spawacz>
acceleration in mpv for example?
<craftyguy>
more runtime power management is disabled, it sounds like
<steev>
if using mainline kernel, nothing. if using jhovold or mine, the pm as johan mentioned
<travmurav[m]>
btw did anyone happen to figure out the firefox v4l2 support? as I understood it, upstream ffmpeg misses something that FF uses since it's v4l2 m2m support was made for rpi which has a fork for those (iirc it was dmabuf support)
<travmurav[m]>
I quickly tried to place the rpi forked ffmpeg into my system a while back but it didn't seem like ff has picked it up
<travmurav[m]>
though not sure if alpine ff vendors those in so aggressively or something
neobrain[m] has quit []
<konradybcio>
bod said firefox 128 nightly has the required fix and joined the last couple meetings from the x13s
<konradybcio>
travmurav: ^
<konradybcio>
I had issues with alpine packages before so doublecheck that too
<travmurav[m]>
<konradybcio> "bod said firefox 128 nightly has..." <- I mean Venus h264 accel not camera
<travmurav[m]>
(it's absolutely amazing to see the swisp work go so far though)
psydroid[m] has quit []
<spawacz>
I'm booting the installer, change quiet to arm64.nopauth but the laptop reboots after 5 lines are printed
<JensGlathe[m]>
looks like its not finding the dtb. Is it specified?
<spawacz>
no
<spawacz>
I'm followin debian's page
<steev>
did you copy it to the dtb? and did you enable the linux option in bios?
<steev>
er, copy the dtb to the usb*
<spawacz>
no, i copied it to efi parition from windows
<steev>
the efi partition on the usb
<steev>
it doesn't read from the nvme
<spawacz>
no, i downloaded the daily debian testing netinst iso
<steev>
er, no, i'm wrong, you're right
jjardon[m] has quit [Quit: Client limit exceeded: 20000]
<spawacz>
I see in the log before reboot "Generating empty DTB"
<spawacz>
it disappears so fast
<steev>
that means it's not getting the dtb from bios like it should be
<JensGlathe[m]>
nvme also has an ESP, might be taken first?
<steev>
no no, he did the right thing
<steev>
i misremembered and pulled up the wiki page :)
<JensGlathe[m]>
I booted from USB with the dtb in the nvme ESP, and Linux option selected in boot menu
<steev>
it's copied to nvme's esp, bios (with linux option enabled) will pass that in. if linux option isn't set, it won't, afaik
<steev>
it should be the root of the esp, not in a subdir
<spawacz>
I am confident i copied it to D:/
<spawacz>
Let me recheck
<steev>
in b4 it's sittinng on c:\
<steev>
i'd also double check that the linux option is enabled in the bios
<spawacz>
"You have been denied permission to access this folder", give me a break
<steev>
elevated cmd/psh
<spawacz>
omg, stupid edge downlaoded it as dtb.txt
<steev>
oh lol
<JensGlathe[m]>
😈
<steev>
amazing
<spawacz>
Yayyyyyy dmesg rolls across the screen
srinik has quit [Ping timeout: 480 seconds]
<steev>
nice
<steev>
welcome to a decent debian arm64 box
<spawacz>
I think initramfs is stuck, writing that "script something waiting block"
arisu has quit []
<spawacz>
Begin: Running /scripts/local-block ... done - around 20 times, then screen goes blacvk
<steev>
could be an issue with the daily, you could try a previous day... or month, or try a sid iso
<spawacz>
I will boot from usb, chroot inside and take a look around
<spawacz>
I added some kernel command line options and im dropped into initramfs shell, unfortunately keyboard does not work
ajhalaney[m] has quit []
<spawacz>
I can mount vfat partition but not ext4 from the intaller shell... why?
matrix638[m] has quit []
juergh1 has quit [Quit: Client limit exceeded: 20000]
kalebris has joined #aarch64-laptops
nekogirl[m] has quit [Quit: Client limit exceeded: 20000]
cmeerw[m] has quit []
<steev>
because debian installer doesn't add the utilities until it reaches the step of partitioning
<spawacz>
So now I've wiped the whole disk, the dtb is missing. Can I copy it to the USB's EFI partition?
<steev>
i *think* so
<spawacz>
I can't copy it there because tehre's no space left on the EFI parition of the debian installer
<spawacz>
/dev/sdb2 3768 3762 6 100% /mnt
<JensGlathe[m]>
sounds like starting again
<steev>
we should probably ask them to increase that ever so slightly, instead of compressing it as small as possible
<steev>
you could create an ESP partition on the NVME and put it in there
<spawacz>
you mean, unscrew the laptop and extract the disk?
<spawacz>
or can I copy a file from USB to the partition via the bios maybe?
<steev>
i was thinking doing it from a different usb. i don't recall if there is a uefi shell on these or not
<JensGlathe[m]>
you get a shell right? you could partition it as desired
<steev>
he can't partition though, debian installer doesn't load the partition tools until it gets to its partitioning stage
<steev>
i don't recall how to install the udebs manually
<spawacz>
no, I cant boot the installed because there is no dtb on the efi partition atm
<spawacz>
installer*
<steev>
ah, right
<steev>
you could write jens ubuntu image to a usb drive, and then install debian from that
<spawacz>
So now I will check the older netinst, if they leave enogugh space to squeeze the dtb there, if not, I willresize the usb's efi parition, copy it there and check
<spawacz>
hmmmm that is an excellent idea steev
<steev>
assuming his image doesn't require the dtb to be in the esp
<JensGlathe[m]>
no it doesn't, its in the image, with grub
<JensGlathe[m]>
and you get the nice Ubuntu GUI
<steev>
SOME DAY GRUB DEVS WILL ACCEPT THAT PATCH
<steev>
a guy can dream, anyway
<JensGlathe[m]>
are there any grub devs?
<abby>
which patch?
<steev>
the one that auto populates the devicetree line
<JensGlathe[m]>
devicetree command
<steev>
you can manually do it
<steev>
it was nacked upstream but i don't recall why
<frozen_cheese[m]>
i just built grub from git the other day, it has the devicetree command in it
<abby>
i stole the patch from fedora for void
<JensGlathe[m]>
frozen_cheese[m]: but its not a release yet?
<abby>
devicetree command is in 2.12
<steev>
the command itself has been in for a long time
<frozen_cheese[m]>
that's fair, it might not be a "release" version
<steev>
we aren't talking about the command, we're talking about ubuntu's patch. void and ubuntu do whatever, but debian don't want that maintenance burden, so insist on getting upstream to accept the patch before they will
<JensGlathe[m]>
principles, so important
<steev>
agreed
<abby>
debian not wanting to maintain a patch? what twilight zone have i wandered into
<steev>
in kali we do whatever (we're a bunch of hackers anyway), but at the same time, we fork as few packages as possible, because while we are hackers... we only have so much time between the 8 of us
underpantsgnome[m] has quit []
LucasTreffenstdt[m] has joined #aarch64-laptops
<spawacz>
Booted!!!!! :D
<spawacz>
i did not add the moduels to initramfs
<spawacz>
and i've recovered from the missing dtb file by flashing edk to a pendrive and copying it back this way :)
<steev>
nice!
hightower2 has joined #aarch64-laptops
<spawacz>
Now I don't know if it is charging. The leds near F1, F2 and the power button are blinking ocassionally
<spawacz>
does that mean it is, or I should try a different cable?
<spawacz>
the red dot near between the usb ports is on
<spawacz>
it does not wantto power on now
<spawacz>
different cable seemed to work
jhovold has quit [Ping timeout: 480 seconds]
<steev>
if you aren't using my branch (or anything before 6.11), you need the power-domain-mapper package or whatever from the wiki
<steev>
spawacz: ah, important to note, if pd-mapper service isn’t running, charging won’t happen either
<steev>
It’ll be a kernel driver once 6.11 comes out though, so no more userland daemon needed then
smpl has quit [Ping timeout: 480 seconds]
<craftyguy>
that will be really nice :D
<steev>
i've got it in my tree already
<steev>
i'm debating asking debian to cherry pick it, since i've tested it on 6.9 (and 6.10), and since it's in -next there should be less pushback
<abby>
I pushed it on 6.8+ on void because it would be a pain at a later date to have everyone switch over, especially when void has almost every supported kernel available
<steev>
yeah it works on 6.8+ but most people aren't gonna go back to that :D