<SolidHal>
if they don't come back, we know how it went :D
<alyssa>
SolidHal: That's not very nice :p
<icecream95>
alyssa: It booted first time, with no obvious problems yet. Now to try and undervolt the GPU a bit
<alyssa>
booting a kernel on a chromebook first try? legend :o
<icecream95>
perf works again :D
<alyssa>
🎉
<alyssa>
undervolting a t760? you sure I couldn't interest you in a shiny new bifrost? :p
<icecream95>
alyssa: Sure, I'll expect it in the mail in a couple of weeks
<HdkR>
icecream95: You need a Bifrost? I'll fund it, which one do you want? :)
<HdkR>
Apparently both the GoA and ODROID-N2 boards are in stock again
<HdkR>
Or do you want a Lenovo Duet?
<icecream95>
HdkR: Well, my veyron-speedy is falling apart a bit...
<alyssa>
icecream95: OOI does your trackpad still work?
<icecream95>
alyssa: Trackpad is fine, charger and screen hinge not so much
<HdkR>
So Chromebook Duet? Looks like Best Buy has them in stock
<macc24>
,_,
<macc24>
how high do i have to be in nerd hierarchy to get hardware to develop on?
<alyssa>
icecream95: Ah. Trackpad went first for me, charger second. Don't think I ever had issues with the hinge.
<HdkR>
macc24: Just need some contributions and someone to yeet them at you :P
<macc24>
HdkR: is making the only known to me distro with mainline linux and u-boot for odroid go advance enough?
<HdkR>
Sure, you want a duet?
<HdkR>
Shipping outside of the US is difficult :P
<SolidHal>
lol *free hardware for everyone*
<macc24>
hm
<macc24>
perhaps
<HdkR>
Or some other piece of hardware in mind?
<macc24>
wait, this is serious?
<HdkR>
Yes, I throw hardware at people all the time
<alyssa>
^^ can confirm
* icecream95
wonders if it's worth waiting for Valhall chromebooks
<HdkR>
It's what little contribution I can do since time is harder
<alyssa>
But only if they're located within 10 meters of his residence
<alyssa>
Otherwise he won't manage to throw it that far.
<HdkR>
UPS Instructions: "Please yeet directly at customer"
<macc24>
i need to think it through at better time than 2am
<HdkR>
hah sure. I guess if it is a hardkernel board, shipping other than US is easy :P
* icecream95
sees that the Chromebook Duet is available locally, at the usual inflated pricing
<HdkR>
Not uncommon
<kinkinkijkin>
i remember running kde plasma 5 on my odroid xu4 with libmali
<kinkinkijkin>
it was pain don't follow my example
<kinkinkijkin>
i had to compile a very specific patchset for libgbm and then every time i launched it, it lasted 10 minutes before hitting a complexity limit and locking the board with lag
<alyssa>
t_t
<kinkinkijkin>
the patchset wasn't one you could just find either, i scoured google's repos for early chromeos patches of libgbm
<kinkinkijkin>
to find a set of patches that worked
<kinkinkijkin>
it honestly probably taken less time and effort to write the patches by hand tbh
<kinkinkijkin>
in the end it wasn't even worth it because i couldn't get it to throw any errors, useful or not
<kinkinkijkin>
by the way, are there any new arm laptops lately? other than apple's (announcement? leak?)
<alyssa>
kinkinkijkin: chromebook? :p
stikonas has quit [Remote host closed the connection]
<kinkinkijkin>
another arm chromebook is out?
<kinkinkijkin>
finally
<HdkR>
The Lenovo Chromebook Duet is pretty nice
<alyssa>
you didn't see the glob of them on the CN tower? :p
<kinkinkijkin>
i haven't gazed towards toronto in a few weeks :p
<kinkinkijkin>
oh crap i can't click that i'm in weechat on my odroid-xu4
<HdkR>
icecream95: How can i help you get that local duet? :)
<kinkinkijkin>
is the latest pinebook still going out btw? is there another, even?
<kinkinkijkin>
last i checked was the second majour version, third overall version including that weird split between the types of the first pinebook
popolon has quit [Quit: WeeChat 2.9]
<macc24>
kinkinkijkin: you can ctrl+l to switch to "plain" mode and then double click to select link and paste with middle mouse button
<kinkinkijkin>
that's not working
<kinkinkijkin>
i'm using terminator on sway with an alternate keyboard layout
<kinkinkijkin>
i've found nothing supports dvorak-fr by default unless it just reads key names, which is considered "bad" practice for whatever reason
<kinkinkijkin>
apparently the duet is only like $400, that seems cheap for a laptop
<HdkR>
Chromebooks are on average towards the cheaper side of things
<macc24>
i've seen cheaper
<kinkinkijkin>
i haven't bought a new computer since $1000 got you 1080p and an amd athlon 2 x2 220 at 2.2ghz
<kinkinkijkin>
wait nevermind i bought my odroid xu4 new
<kinkinkijkin>
for like... $150 cause i got a bad source
<alyssa>
does anyone else pronouncing xu4 as "zoo four"?
<kinkinkijkin>
i will now
<kinkinkijkin>
dang
<alyssa>
😻
<alyssa>
i must vocalize everything
kaspter has joined #panfrost
<kinkinkijkin>
horrible time to acquire a fancy for some new hardware
<alyssa>
uh, have names become harder to pronounce?
<kinkinkijkin>
moving in two months, far north even
<macc24>
what's the name of that helio p60t chip in duet? mt6771?
<alyssa>
mt8183
<alyssa>
"mitt eight one eight three"
<macc24>
ohhhh mt8183
kaspter has quit [Remote host closed the connection]
kaspter has joined #panfrost
kaspter has quit [Excess Flood]
kaspter has joined #panfrost
<kinkinkijkin>
time to test firefox like i meant to some large number of hours ago
<kinkinkijkin>
firefox works
unoccupied has joined #panfrost
unoccupied is now known as leah
leah is now known as _4of7
davidlt has joined #panfrost
_4of7 has quit [Remote host closed the connection]
vstehle has quit [Ping timeout: 272 seconds]
<icecream95>
With Linux 5.9, scrolling in Firefox (with OpenGL compositing) leaks shared memory which is never freed, not even after rmmod'ing panfrost
<chewitt>
kinkinkijkin are you running panfrost on xu4? .. I wasn't aware that was working
<chewitt>
would be fab if it is now .. I'd be able to retire all the mail/blob stuff in my distro :)
davidlt has quit [Ping timeout: 240 seconds]
<macc24>
chewitt: for my board with g31mp2 panfrost is running fine
<chewitt>
xu4 has a midgard t628 chip
<chewitt>
it's an early chip with a larger collection of errata to work around
<macc24>
huh
* macc24
noted
<chewitt>
last time I tried panfrost with it, it crashed and kodi failed to run
<chewitt>
but that was some time ago
<macc24>
tip: LIBGL_ALWAYS_SOFTWARE=1 makes everything work :D
chewitt has quit [Quit: Zzz..]
icecream95 has quit [Ping timeout: 240 seconds]
kaspter has quit [Ping timeout: 260 seconds]
kaspter has joined #panfrost
vstehle has joined #panfrost
icecream95 has joined #panfrost
<dstzd>
icecream95: i've noticed something similar in evince (pdf reader) but wasn't sure what was causing it.
<tomeu>
sounds pretty scary :/
<anarsoul>
icecream95: ouch :(
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 272 seconds]
camus1 is now known as kaspter
<archetech>
macc24 tip: LIBGL_ALWAYS_SOFTWARE=1 makes everything work :D cuz it turns off real gpu and uses cpu
<archetech>
which is crap
alpernebbi has joined #panfrost
<dstzd>
does anyone know if pages is supposed to be freed somewhere from the "pages = kvmalloc_array" in drivers/gpu/drm/panfrost/panfrost_mmu.c ?
davidlt has joined #panfrost
<macc24>
does mediatek do weird stuff like amlogic?
<cyrozap>
HdkR: In addition to the Lenovo Chromebook Duet, Acer has a new MT8183 Chromebook as well (an actual Chromebook with built-in keyboard, not a tablet like the Duet), but it doesn't seem to be available just yet: https://www.acer.com/ac/en/US/content/model/NX.HUVAA.003
<cyrozap>
Also, the 9th gen/2019 Kindle Fire HD 10 also has an MT8183, and as an added bonus it has JTAG and UART0 exposed internally, and due to a quirk of the MT8183 boot ROM, JTAG can _never_ be disabled.
<macc24>
cyrozap: duet has uart externally afaik
<macc24>
i'm curious if that port on bottom is just usb or something more
<cyrozap>
And because of that, so long as you can connect to JTAG within 15 ms of booting the SoC, you can bypass signature checks (set breakpoint on signature function, jump over it, set result to "valid", continue execution).
<cyrozap>
alyssa, macc24: Technically, the MT8183 and the MT6771 are the same chip. As in, same chip ID, same boot ROM, same memory map. My guess is the MT8183 is just an MT6771 with either the cellular parts not verified/failed validation, or just sold without the RF support chips to support cellular functions.
<macc24>
noted
<cyrozap>
macc24: Yeah, supposedly on these Chomebooks they have that Google H1 chip, so you can do closed-case debugging through USB-C and some fancy cable.
<cyrozap>
HdkR: Ah, yeah, it does have a nicer screen, but the lack of a built-in keyboard is kind of a dealbreaker for me. At that point, I might as well just get a cheaper tablet-only device like the Fire HD 10.
<HdkR>
Sadly there are a lot of cheap devices that are doing the whole tablet + keyboard cover thing
<HdkR>
2-in-1 is so much nicer instead of that
<cyrozap>
Also, if anyone wants to hear my stream-of-consciousness while I explore the inner details of MediaTek chips, I post that stuff over in #postmarketos-lowlevel. Currently I'm working on MT7668 WiFi chips and hacking support for them into mainline, but I'll get back to the SoC stuff eventually.
<macc24>
tablet + keyboard cover has one advantage
<macc24>
no code for disabling keyboard is needed in tablet mode ;)_
<HdkR>
hah
raster has joined #panfrost
<HdkR>
Alternative opinion, you have to RE a full microcontroller like the ProX to even get the keyboard
<macc24>
huh?
<cyrozap>
Heh, I'm currently using a 2-in-1 that (I think) disables its keyboard in hardware when switched into tablet mode, but the touchpad (including buttons) stays enabled, so tablet mode ends up being not that useful.
<HdkR>
Surface devices talk to their keyboard cover over some i2c interface rather than standard USB HID...They call it SAM
<HdkR>
Surface Accessory Module or something
<macc24>
i2c hid?
<HdkR>
I have the x86 version partially ported to ARM, but it can't find the device because I failed to DTS properly :P
Venemo_ is now known as Venemo
<archetech>
cyrozap: That's a cool offer I'd like that but for panfrost dev
<archetech>
like to watch how a dev fixes this color issue for example
<archetech>
or things that cause freezing
<cyrozap>
archetech: Well, I guess it's more like I use that channel to document stuff as I learn it, since it can sometimes take a while for me to organize the information into a more useful form.
<archetech>
I did that with Linux from Scratch
<archetech>
which I do have working on the N2
chewitt has joined #panfrost
<chewitt>
cyrozap I'm also interested in MT7668 SDIO support as it started to appear in recent Amlogic STB (Android) boxes
<chewitt>
do you have code in a repo anywhere?
stikonas has joined #panfrost
<cyrozap>
chewitt: Oh, lol, I think I may have responded to you on the LKML, though it was more of a "I'm also interested in this" with some links to downstream code and where to buy (MT7668 USB) hardware.
<cyrozap>
I don't have any code in a repo yet, on account of "the support for MT7668 clobbers the MT7663 code" and the fact that at most I've been able to get it to try to load some firmware.
<cyrozap>
Like, literally I just rewrote some of the MT7663 code to work with the MT7668 XD
<chewitt>
I have no requirement for MT7663 support so that wouldn't be an issue :)
<cyrozap>
But I do have a correct register map, and I can dump the EFUSEs, so that's pretty neat. I'm basically just working my way through the code, fixing things where they fail, and adding register/bit definitions as needed.
<cyrozap>
And last night I figured out how to get Qemu to boot my modified kernel and talk to the device so I don't have to reboot my computer every time I want to test a new change.
<chewitt>
I don't have test hardware myself, but other people in the LibreELEC ecosystem do, so it's easy to get things tested by people who are (mostly) good at testing
<chewitt>
it's used in some of the Amlogic S905X3 devices .. so I will probably get one
<chewitt>
the boxes use the SDIO interface not USB tho
<cyrozap>
For now I'm focusing on USB support, since that's all the hardware I have supports. In theory it shouldn't be too difficult to get the SDIO support working once USB is done, but of course I don't have the hardware to test that.
<chewitt>
what's the hardware you're trying to support?
<cyrozap>
It's a module made by LG for use in several of their TV sets. It uses the USB variant of the MT7668, but only USB 2.0 is exposed.
<chewitt>
this is just the dev device tho, right?
<chewitt>
or you're trying to run mainline on TVs?
<cyrozap>
Oh, yeah, it's just the device.
<cyrozap>
This is mostly practice before I try to get SoC internal WiFi working, since the software interface is basically the same, just on an internal not-quite-SDIO bus instead of USB or real SDIO.
<cyrozap>
Also, DBDC 802.11ac connectable via USB is kind of nice, even if you can't get full 3.0 speeds.
<cyrozap>
Also also, I wanted to play around with the firmware loading and try to get my own code running on the chip, since it has an ARM Cortex-R4 and an Andes NDS32 and the whole chip seems like it'd be fun to play around with :)
<cyrozap>
chewitt: We should probably discuss this more in #postmarketos-lowlevel, since this is somewhat off-topic for #panfrost.
<chewitt>
I'll read up on what postmarketos is :)
<chewitt>
I joined anyway..
<cyrozap>
The tl;dr of postmarketos is "GNU/Linux on smartphones and other devices". The lowlevel channel is for "lower than the kernel" stuff, like bootloaders, firmware, reverse engineering, etc.
icecream95 has quit [Quit: leaving]
<alyssa>
cyrozap: I don't mind :)
<tomeu>
dstzd: I think in drm_gem_shmem_put_pages
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 272 seconds]
camus1 is now known as kaspter
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
gcl_ has joined #panfrost
gcl has quit [Ping timeout: 258 seconds]
rando25892 has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 260 seconds]
camus1 is now known as kaspter
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 256 seconds]
gcl_ has joined #panfrost
gcl has quit [Ping timeout: 260 seconds]
kaspter has quit [Ping timeout: 256 seconds]
kaspter has joined #panfrost
rando25892 has quit [Ping timeout: 240 seconds]
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 260 seconds]
gcl_ has joined #panfrost
gcl has quit [Ping timeout: 258 seconds]
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 260 seconds]
rando25892 has joined #panfrost
archetech has quit [Quit: Konversation terminated!]
<kinkinkijkin>
chewitt yes panfrost is working on xu4, with a lot of visual bugs though
gcl_ has joined #panfrost
gcl has quit [Ping timeout: 240 seconds]
gcl_ has quit [Ping timeout: 246 seconds]
gcl has joined #panfrost
gcl has quit [Ping timeout: 256 seconds]
<kinkinkijkin>
okay, so a full report on current visual bugs on t682 on 32-bit, in multiple lines as i remember them:
<kinkinkijkin>
z-fighting and buffer texture buffer rewind, personified as flickering, in 2d rendering such as sway
gcl has joined #panfrost
<bbrezillon>
stepri01: hm, so it looks like we can't really have drm_sched_start() outside of the section protected by the reset lock
<kinkinkijkin>
texture buffer scrambling shortly after creation, but only shortly after creation. curiously enough, resizing fixes this
<kinkinkijkin>
hanging under low load on latest stable kernel (deadlock possibly?)
<kinkinkijkin>
occasional render skipping of seemingly random textures, part of the flickering
gcl has quit [Ping timeout: 260 seconds]
<kinkinkijkin>
somewhat strangely, 3d rendering seems to pass fine though i haven't tested it extensively yet
gcl has joined #panfrost
<stepri01>
bbrezillon: yes - if we can do the drm_sched_start() inside the lock that would solve the problem. I'm unsure whether that causes further problems though ;)
<kinkinkijkin>
and seemingly exclusive to pcsx, opening egl contexts fails every open of the emulator after the first one
<kinkinkijkin>
es2gears is running into some minor vertex rounding issues it seems, causing black lines to appear on polygonal borders occasionally
<kinkinkijkin>
exact same result with glxgears
<bbrezillon>
stepri01: that's how it was done before :P
<stepri01>
bbrezillon: yes - where I was suggesting it causes other problems ;) Sorry going to be away from my computer for a while, I'll take another look on Wednesday
archetech has joined #panfrost
stepri01 has quit [Quit: leaving]
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 256 seconds]
gcl_ has joined #panfrost
gcl has quit [Ping timeout: 260 seconds]
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 260 seconds]
raster has joined #panfrost
gcl_ has joined #panfrost
<alyssa>
kinkinkijkin: Oh boy
gcl has quit [Ping timeout: 260 seconds]
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 265 seconds]
felipealmeida has quit [Ping timeout: 260 seconds]
gcl has quit [Ping timeout: 260 seconds]
felipealmeida has joined #panfrost
Elpaulo has quit [Remote host closed the connection]
Elpaulo has joined #panfrost
gcl has joined #panfrost
gcl_ has joined #panfrost
gcl has quit [Ping timeout: 264 seconds]
<kinkinkijkin>
wait im an idiot who doesn't update their source tree correctly
<kinkinkijkin>
just a second gonna retest everything with the properly updated source tree
<kinkinkijkin>
i will return shortly
kinkinkijkin has quit [Quit: WeeChat 2.3]
kinkinkijkin has joined #panfrost
kinkinkijkin has quit [Quit: WeeChat 2.3]
cphealy has quit [Ping timeout: 260 seconds]
kaspter has quit [Read error: Connection reset by peer]
kaspter has joined #panfrost
kaspter has quit [Ping timeout: 272 seconds]
kaspter has joined #panfrost
purringChaos has joined #panfrost
davidlt has quit [Ping timeout: 240 seconds]
kinkinkijkin has joined #panfrost
<purringChaos>
alyssa, you said on the gist you can reproduce it with more generality, can you elaborate?
<alyssa>
mysterious buffer leaks that don't necesarily clear when the process dies is a known but unknown cause issue
<alyssa>
it's not just webkit
<purringChaos>
i think vscod(ium) also exibits the same behaviour and firefox after a week too
<purringChaos>
some apps more than others i guess
<alyssa>
bbrezillon: wonder if we can make time to explore memory stability at some point.
<alyssa>
overallocation in mesa contributes to the problem but as is becoming clear the kernel isn't handling things properly either
<purringChaos>
the kernel should be reclaiming stuff, even if every process but init is killed, it still wont reclaim it
<kinkinkijkin>
it's faces disappearing and backing up
<kinkinkijkin>
no z-fighting beyond the normal that i can find
<kinkinkijkin>
i just "played" a video game for about an amount of time im not quite sure of that must've been short, the highlight of that session was managing to get through the menus which didn't have proper masking and so were jusst a bunch of colourful squares
<kinkinkijkin>
i got to "playing" the video game for a good solid "huh"
<kinkinkijkin>
i can send a video of the issue for better examination if sent equipment to record this such as an hdmi capture method and a bottle of brandy the size of the amount of recording id have to do
<kinkinkijkin>
attempting to examine that in realtime was pain incarnate and im pretty sure it lasted like 2 minutes but it felt like 20 years
<kinkinkijkin>
i did not choose a good game to test with, for my eyes
ric96 has quit [*.net *.split]
ric96 has joined #panfrost
<kinkinkijkin>
why did i go play more
<alyssa>
🙉
<kinkinkijkin>
ill be back in a few minutes on my x86 desktop all this flickering is giving me a migraine
kinkinkijkin has quit [Quit: WeeChat 2.3]
purringChaos has quit [Remote host closed the connection]
kinkinkijkin has joined #panfrost
kinkinkijkin has quit [Ping timeout: 260 seconds]
kinkinkijkin has joined #panfrost
<kinkinkijkin>
accidentally unplugged my computer trying to turn on jack