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)
<shawnguo>
steev: yeah, I will leave laptops-5.13 there as it is, and am picking up lmh v4 for laptops-5.14 that I'm building right now.
<steev>
v5
<steev>
i hope
<shawnguo>
I meant v5, yes :)
<steev>
:)
<steev>
cool, i haven't attempted it because of the rcg clk thing that someone is supposed to be looking into, i pinged abhinav, but they haven't responded about it, and i don't wanna be pushy
<bamse>
steev: seems my wifi fix only works on saturdays :/
<steev>
interesting way of adding parental controls into the kernel :P
<bamse>
steev: but managed to crash the primus and grab a ramdump, so i have an error message now
<bamse>
hehe
<steev>
well that's good, i've also been watching the stuff in freedreno or dridevel, one of the two
<steev>
and running git fetch --all hourly
<steev>
ohhh, 5.14 came out today, that explains why shawn's working on it
<bamse>
i can share my branch if you're interested, it's just that it still doesn't have stable wifi
<steev>
i don't mind the crash every few hours or so, it actually doesn't happen super often when i tell it not to suspend
<bamse>
pushing a new branch for you to play with
<bamse>
don't think there's anything new since last time...just newer linux-next and some of the patches have been updated based on feedback from the list
<steev>
shouldn't it just... not support it instead of throw that error? (also why does it hang if bam_dma and dev_qce aren't built in)
<shawnguo>
well, it's warn not error, and that's a result of xts(aes) driver being registered but some particular case is not supported
<shawnguo>
hang with bam_dma and dev_qce being module is a separate issue that I will take another look
<steev>
what i mean is, that is a "scary" output for end users, and it's an expected outcome, so it shouldn't produce "scary" output
<shawnguo>
that's what crypto testmgr decided to do, I'm not sure it's reasonable either
<steev>
bamse: hm, the defconfig seems to be missing "something", i ended up in a busybox shell, i'm guessing ufs stuff isn't in my initrd or something, because it's looking for localblock... i think?
<steev>
this weird double screen thing makes it kinda hard to read
<steev>
took the old config, lets see how this does, odd that the ccache seems to be invalidated though cuz it's super slow
<bamse>
double screen?
<steev>
at first boot, it's like the display is split down the center and mirrored - sometimes 3 times, the same thing happens on the c630
<bamse>
can you snap a picture?
<steev>
sure give me a sec
<steev>
i rebooted to grab the rc2 config that works
<bamse>
i don't think i have any functional difference, so the -rc2 config should work
<bamse>
hmm, that's not the same bug that i had earlier...
<bamse>
well the bug i had wouldn't show up on c630
<steev>
that only happens during the initial boot
<steev>
once X inits or the plymouth splash screen does, it *usually* goes away
<steev>
it might be a grub bug, because it does seem to happen when booting into windows too
<bamse>
hmm, yeah i think the penguins goes away when we probe the msm drm driver
<bamse>
yeah, that sounds more likely...still worrying though
<steev>
i can't recall when it started, but i'm gonna guess latest grub update past bullseye release
<steev>
i would *expect* the same on a sid system, but sometimes stuff jumps into testing skipping unstable, so not sure
<bamse>
well it does resemble one of the many artifacts i saw before getting the display sorted out...but that would not explain why you see it on the c630
<steev>
okay, definitely had the wifis crash quickly there
<bamse>
you're welcome ;)
<steev>
but something that i couldn't figure out... on the previous kernel
<steev>
the first time i logged in, i was *always* shoved onto X11
<bamse>
i pushed one more patch now
<steev>
i log out and back in, and it was on wayland
<steev>
it now logs in to wayland
<bamse>
yeah you said that, really odd
<bamse>
the only thing you changed was the kernel?
<steev>
yep
<bamse>
hmm, odd
<steev>
i'm on 21.2.1 i think, of mesa, with the patch for adding in the adreno
<bamse>
but fetch again and do your testing with the "disable cluster collapse" change included
<steev>
will do
<bamse>
looks much better so far
<bamse>
well, on saturday the primus didn't crash for a full hour...so who knows
FizzBuzz has quit []
<steev>
just rebooted
<steev>
i was simply looking through dmesg, and attempting to use bluetooth to scan (after installing rfkill since it was softblocked)
<steev>
more of a "does bluetooth work" than any other reason
<bamse>
hmm, odd
<bamse>
although i just got a NULL pointer dereference in __sys_shutdown when trying to start chromium
<steev>
oh, i definitely got that too
<bamse>
so i guess next-20210819 might not be the best kernel we have produced
<steev>
well, i assume that's the same as what i got since i was remote
<steev>
firefox
<steev>
nevermind
<steev>
firefox at least opened its window?
<steev>
but then same null pointer
<bamse>
i think i found the fix on the linux-next mailing list, checking now
<bamse>
yeah, now chromium starts again
<bamse>
so let me push that
<bamse>
done
<bamse>
and the damn thing froze again
<bamse>
hmm, i did revert one hack in the interconnect provider between testing on the primus and flex
<bamse>
and the primus is at 2400 seconds pinging strong...so perhaps...
<bamse>
time to get some sleep, maybe that solves the issue
rfs613 has quit [Quit: restart]
rfs613 has joined #aarch64-laptops
<HdkR>
Do any of the 7c chromebooks ship with 16GB of ram?
<HdkR>
Or any other Snapdragon Chromebook I guess?
alpernebbi has joined #aarch64-laptops
<HdkR>
Or a snapdragon laptop that can run upstream linux with a little bit of faff? The ProX isn't that :D
<HdkR>
and 16GB of ram
<arnd>
HdkR: it's hard to even find any with 8GB at the moment
<arnd>
They are often twice the price of the 4GB models
<HdkR>
Hm
<HdkR>
Speaking of which, phire seems to have noticed that their c630 is missing ~1GB of visible RAM. DTS configuration issue or something else consuming it?
<HdkR>
only 6.9GB is a bit rough
<HdkR>
Probably running some hecking old kernel from when I first set it up :)
<bamse>
HdkR: running "free -m" on my c630 reports that i have 7548 rams...
<bamse>
which is still not much...
<HdkR>
hm
<robclark>
HdkR: max is 8GB..
<HdkR>
I'm disappointed this day
<robclark>
well, 8GB + zram swap
<HdkR>
that's not real ram :P
<robclark>
I don't think there are many x86 chromebooks w/ 16GB
<HdkR>
Need a Pixelbook replacement with a good ARM SoC
<robclark>
you can get the hp x2 11 with 8GB.. and a stylus!
<HdkR>
How much ram does the stylus have in it?
<HdkR>
Better be 8GB
<robclark>
heh, I guess it is a bit less than that
<HdkR>
Oh, I completely forgot that the Pixelbook Go was already a replacement for the Pixelbook. Just need a replacement for that with a 3rd gen 8cx :D
<HdkR>
9cx? For ARMv9 compute?
<bamse>
isn't the pro x the only "laptop" you can get with 16gb ram?
<HdkR>
That I'm aware of
<HdkR>
I've not been tracking where the SPX currently is in support
<HdkR>
Last I knew it still had SAM issues
<bamse>
i was hoping to "finish" up the flex work and then write up a prox dt to try it out...
<HdkR>
If ProX bringing up is in a better place than I expected then that could work
<HdkR>
I can at least export that one to places
<bamse>
i don't know anything about the SAM stuff, but other than that it seems pretty similar to the flex...
<bamse>
got nvme instead of ufs...but that i have validated that on the primus, which has both
<steev>
HdkR: there are some HP snapdragons with 16GB
<steev>
since no one has gotten any of us any of them :P
<HdkR>
Ugh, Windows.
<HdkR>
Windows doesn't work at all for what I want :P
<HdkR>
ChromeOS I can at least just tell someone to hack a shell on to it
<bamse>
$1949.00...
<robclark>
and people complain about not having a $300 chromebook w/ 8cx gen3 :-P
<HdkR>
Might as well as wait for the Macbooks to have Linux running on it :)
<HdkR>
I'm willing to pay whatever it takes, there just aren't options D:
<steev>
it's gonna be a while still
<steev>
there are options if the hardware is gotten into the correct hands :P
<bamse>
HdkR: clicking their web thingie indicates that it would save you $200 (on getting something with 16gb ram)
<bamse>
HdkR: but then you have to live with the fact that it would be much faster as well...
<steev>
but yeah, even 200 off is still outta my price range
<HdkR>
oh no, web thingies
<steev>
i just bought my tickets for my trip this month so all my extra spending monies is gone
<HdkR>
steev: Are you volunteering for devce bringup?
<steev>
it shouldn't be *that* difficult to base on what bamse has for the flex/primus
<HdkR>
I mostly just want to ship laptops to my coworkers that aren't old or a massive PITA to setup
<steev>
that makes sense
<HdkR>
c630 meets the old criteria by not being sold anymore
<robclark>
gen3 is actual changes, compared to gen1 vs gen2.. on gpu side, I expect a690 is probably similar to a660 but with more SP cores.. not sure what the delta is on the display side.. perhaps some things can be reused from sc7280 bring up work, but you probably want either an codearora kernel or docs
<bamse>
HdkR: the c630 is still nice and was priced well...but as you say, it no longer possible to get hold of
<steev>
bamse: shawnguo: so, i backed out alllll of my changes to my 5.13, and have put in just the minimum of laptops and while i've seen one single error [drm:dpu_core_irq_callback_handler [msm]] *ERROR* no registered cb, idx:27 enable_count:0 - i haven't had it do that horrible thing i complained about before
<steev>
the horrible thing where it starts spewing some message and the system becomes unusable
<bamse>
robclark: so they actually bumped 640->660 based adreno on the 8cx respin?
<robclark>
that is my expectation
<robclark>
assuming there is some relation to sc7280
<bamse>
they just throw ip blocks in a bingo ball and out comes a new platform...
<HdkR>
Also the 7c chromebook isn't interesting mostly because the CPU improvements aren't really there compared to SDM850 in the C630
<steev>
wow
<steev>
just going right for the juggular
<robclark>
bamse: pretty much
<HdkR>
It's midrange versus flagship chip, it happens like that :P
<robclark>
HdkR: I think I'd expect the current 7c things to be a bit faster in singled (or dual) threaded..
<robclark>
than sdm850
<HdkR>
Since it would heavily be used for code compiling though, it wouldn't be very nice
<robclark>
yeah, you might want more performance cores for that
<HdkR>
2+6 arrangement isn't great
<bamse>
they should do 5+1 instead
<robclark>
7280 has more fun arrangement, if you pay attention to the dts
<steev>
#nvidia
<bamse>
sorry, 7+1...
<bamse>
robclark: another 4+3+1 ?
<robclark>
that's what it looks like to me
<steev>
is that the one that stephen just sent a patch to base some numbers in reality?
<bamse>
robclark: they have that on most socs these days...think 8cx is one of the few that has 4+4
<robclark>
I assume that is 4 prime + 4 fast?
<bamse>
steev: nah, that was sc7180 still...if you mean the patch from doug
<robclark>
or, you mean 8c[x] gen1/2?
<bamse>
robclark: i don't know what they are classed as, but i hope so
<bamse>
robclark: i mean sc8180x :)
<robclark>
ahh, well if the smaller cores are OoO they are not littles
<bamse>
but i think the real benefit that sc8180x should have is that there's more arrows going to the memory
<robclark>
not sure what all the kryo core names map to in cortex lingo
<robclark>
yeah
<bamse>
i've just given up on that
<bamse>
i'm trying to get it to work, then we benchmark it and see that we can squeeze the juice out
<bamse>
and right now, geekbench tells me that the c630 is faster, so there's more juice to be sqeezed
<bamse>
in particular there's more joules to be released
<robclark>
interconnect or cpufreq?
<bamse>
i added interconnect references from cpufreq and that almost got the memory bandwidth benchmarks up to what i see on sm8150...so there's something missing, but it's closer
<bamse>
but when i'm idling the system it's at 35C and when i'm running geekbench it launches all the way to 37C
<bamse>
so something is broken
<robclark>
heh, yeah, it's not letting those joules out
<bamse>
good for the environment, i guess
<steev>
bamse: ah yeah it was doug
<steev>
i actually still had the patch open in a tab :D
<steev>
i noticed that the flex doesn't get very hot either
<steev>
not that i can actually see the thermals
<HdkR>
robclark: So what you're saying is that I need an SC7280 chromebook today? :thonk:
<bamse>
steev: you can't see them?
<steev>
yes
<steev>
bamse: correct
<bamse>
steev: /sys/class/thermal is empty?
<steev>
let me boot it back up, pretty sure it is though
<steev>
this is on the rc2 where it lasts a bit longer before wifi breaks
<bamse>
steev: i "solved" that with an usb ethernet dongle...
<steev>
hm, no thermal isn't empty
<bamse>
no idea why...do you have tsens in the dts? do you have CONFIG_QCOM_TSENSE enabled?
<steev>
tsens is yes, but bottom says no sensors found
* bamse
closes that tab and pretend like it never happened
<robclark>
triple-plus-good
<steev>
oh weird
<steev>
i'm seeing pdmapper refusing to start
<steev>
not refusing, failing, i connected a usb ethernet, lets see if i can get into the machine
<steev>
oh, it's rmtfs that's failing to start, failed to get rprocfd
<bamse>
sounds like you don't have your modem remoteproc probed
<steev>
rmnet and ipa are both definitely modules
<bamse>
and i think we have a problem with the hid device on &i2c7 address 5...i.e the thing that shows up as keyboard, wireless radio control and ""
<bamse>
i'm idling the flex and i get hundreds of interrupts per second from it...
<steev>
wow
<bamse>
perhaps it's not level triggered, even though i would have gotten that from the acpi tables
<steev>
oh weird
<steev>
bamse: so when it does ^^ the display is getting set to 1280x720 for some reason
<steev>
Screen 0: minimum 1280 x 720, current 1280 x 720, maximum 1280 x 720
<steev>
it's like, it's not updating from efifb
<steev>
[ 0.193435] efifb: mode is 1280x720x32, linelength=5120, pages=1
<steev>
bleh, it was a kali thing
<steev>
in grub at least
<steev>
so in kali, we force 1280x720 so that people with 4k displays aren't like "omg i can't see anything" to choose in grub
<steev>
so for whatever reason, when it's going from grub's 720p... it never updates to 1080p
<steev>
bamse: also, for giggles, i'm pretty sure the stuff is older, and not based on the latest, but maybe it is? i'm trying out the laptops-5.14 on the flex5g, and i'm currently syncing my kernel repos across machines, and not seeing the... "lag" of the cursor while the network traffic is happening
<steev>
hm, and only seeing the 3 cooling devices, so i guess i need to look what's different
<steev>
ah, yeah older :)
<steev>
bamse: ah, think i may have found the (my) issue
<steev>
gpu-thermal-bottom
<steev>
steev@meatwad:/sys/class/thermal/thermal_zone25$ cat type
<bamse>
does heim give you any useful complaints, or computer just says no?
<bamse>
looks like my keyboard has reached 7466351 interrupts now...i should fix that
<bamse>
given that the lid has been closed pretty much the entire time
<steev>
as with all rust back traces, it's fucking useless
<steev>
https://bpa.st/AWWA which just points to "let sensor = sensor.unwrap();"
<steev>
hm
<steev>
actually, i *think* for temperatures it's looking under hwmon and not thermal ?
<steev>
and for hwmon we only have hwmon0 which is the ath10k
<steev>
steev@limitless:~$ ls /sys/class/hwmon
<steev>
hwmon0 hwmon1 hwmon2
<steev>
hm, no that can't be right
<steev>
okay, it only seems to use thermal if hwmon returns nothing, maybe, i can't really read this code tbh
<bamse>
steev: there's a code path in there that looks for class/thermal/thermal_zone*
<steev>
yeah
<bamse>
steev: but i have no idea if it's triggered
<steev>
well the c630 has hwmon stuff, so
<steev>
it should showw the same issue
<steev>
i'm just gonna open an issue with upstream and blame them
<bamse>
and forgo this opportunity to learn the language of the future?
<steev>
yes
<steev>
though rust in the kernel now!
<bamse>
madness
<steev>
yeah, there is definitely some performance improvements to be made on the flex, it takes 107 minutes to build a pinebook kali image on it, and it takes roughly half that on the c630
<steev>
stupid question... would lmh apply on the flex? or is it specific to 850 and below