alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
vstehle has quit [Ping timeout: 268 seconds]
<alyssa> urjaman: rainbow can be any of the 8 parts of speech
<alyssa> Noun: The rainbow is pretty.
<alyssa> Adjective: The rainbow shirt was adorable.
<alyssa> Verb: I rainbowed and now I have a girlfriend.
<alyssa> Adverb: The rainbow red shirt was neat.
<alyssa> Pronoun: I was talking to my friend, and rainbow said rainbow was rainbow.
<alyssa> Preposition: The colours rainbow me are incredible.
<alyssa> Conjunction: I think, rainbow I am.
<alyssa> Interjection: RAINBOW!!!!!!!!!!
<alyssa> Speaking of rainbows, plymouth on a chromebook is a sight alright.
* alyssa is getting nostalgic from using Fedora years ago
<urjaman> i have 5.4-rc2 running on my veyron
<urjaman> didnt test panfrost yet but neat they didnt break anything super-horribly (for veyron) in the meantime :P
<alyssa> urjaman: Hah!
<alyssa> urjaman: But Do You Have Plymouth? :^)
<alyssa> (Super thanks to Alper, who should totally come on IRC :p)
<urjaman> ooooh the icons are no longer glitching :)
<urjaman> and yeah panfrost runs (so far) :P
<alyssa> :P
<alyssa> Grr random kernel errors disrupting my perfect boot
<alyssa> [ okay it's far from perfect, I admit ]
<urjaman> okay whoever does the kernel, that "Purging %lu bytes\n" message in panfrost_gem_shrinker.c maaybe could use a "panfrost:" prefix or something
<alyssa> ALso whoever does the kernel
<alyssa> having panfrost.ko in my debain initramfs causes the machine to hang booting
<alyssa> If I keep it out of the initramfs, machine boots fine (and then it loads off /lib/modules and I still get spinny gears, plymouth doesn't need gl or anything)
<urjaman> (also not sure if it needs to be output at info level...)
<urjaman> do you have the rockchip DRM bits as a module?
<alyssa> I do
<alyssa> I broke things
<alyssa> f'naaaaa
<urjaman> yeah i was just thinking maybe something weird could happen if you load/initialize panfrost before the rockchip ... or something similar *shrug*
<alyssa> ....actually though that wasn't supposed to break anything
<urjaman> yeah it's not supposed to but
<alyssa> No I meant enabling gdm
<urjaman> and it shouldnt matter that panfrost gets loaded during initramfs phase
<urjaman> ah okay
<alyssa> which I mean
<alyssa> I swore gdm3 at least worked at one point
<alyssa> lightdm is ok
<urjaman> but yeah so far (*knocking on wood*) I havent noticed anything that makes me need to go disabling glamor, so this is neat
<alyssa> gdm3 works fine on Kevin II
<alyssa> why not Kevin i?
<alyssa> "Works on my machine" but only one
<urjaman> :D
<urjaman> Works for me but also doesnt :P
<alyssa> GNOME itself doesn't want to start
NeuroScr has quit [Quit: NeuroScr]
<alyssa> I will say the break from gfx to just mess with linux is fun :p
<alyssa> Ah ha, assertion 0 failed!!
NeuroScr has joined #panfrost
<urjaman> why is a kbuild test robot building random branches i push to my github
<alyssa> Aside ... I think GNOME is faster on Kevin I than Kevin II? Am I imagining things?
<urjaman> like I know the build failure because i used that branch to transfer the thing for makepkg to test and yeah that failed :P
<alyssa> dfjhksuigrspgyrioutrog
<alyssa> I can do arithmetic this is fine
* alyssa knew she said she wasn't going to do graphics tonight but .... segfault presented itself ...
<alyssa> Okay I think it's fixed so what's this unrelated(?) segfault
<alyssa> GNOME shell crashing with touch screen
<alyssa> ??
<alyssa> Only affects Kevin I
<alyssa> I can't even tell if this is a panfrost issue or a gnome issue
* alyssa is past the point of things making sense
<alyssa> Uhm
<alyssa> Kevin II, to Kevin I: ...works for me
NeuroScr has quit [Quit: NeuroScr]
<alyssa> Uhm, with X it works, wayland no??
<alyssa> Okay I have it working with X but it's super slow..?
* alyssa isn't sure why gdm be slow :/
NeuroScr has joined #panfrost
megi has quit [Ping timeout: 240 seconds]
NeuroScr has quit [Quit: NeuroScr]
vstehle has joined #panfrost
davidlt has joined #panfrost
NeuroScr has joined #panfrost
NeuroScr has quit [Quit: NeuroScr]
NeuroScr has joined #panfrost
NeuroScr has joined #panfrost
NeuroScr has quit [Quit: NeuroScr]
yann has joined #panfrost
jailbox has joined #panfrost
adjtm_ has quit [Ping timeout: 240 seconds]
<EmilKarlson> yes, it seems there are slight differences
adjtm_ has joined #panfrost
stikonas has joined #panfrost
yann has quit [Ping timeout: 240 seconds]
raster has joined #panfrost
adjtm_ has quit [Quit: Leaving]
adjtm has joined #panfrost
megi has joined #panfrost
warpme_ has joined #panfrost
<urjaman> weird
<urjaman> i rebooted with a new kernel (changed nothing graphical, only USB things) and X has no accel "because" '(EE) modeset(0): eglInitialize() failed'
<urjaman> the panfrost kernel module loaded just fine and the dmesg has nothing from it
<urjaman> that is after loading / when Xorg initializes ...
<urjaman> except for '[ 21.617420] broken atomic modeset userspace detected, disabling atomic' but i think 5.4 said that before too
<alyssa> EmilKarlson: Differences between what and what?
<EmilKarlson> Kevins
<urjaman> odd ... powered off and tried again and the same thing, logged out and logged in (=restarted X) and the same thing ... did a "sudo ninja install" from the same mesa build dir as before, logged out and logged in and now it works ?????
<EmilKarlson> at least cros_ec protocol version, which is probably not your issue though
<alyssa> EmilKarlson: Oh, mm
<alyssa> I feel like differences in software stack (exact kernel version, etc) are more likely..?
<EmilKarlson> sure
<robmur01> alyssa: re modules: if your config depends on /dev/dri/card0 being display and dev/dri/card1 being GPU, module load order can mess things up
<alyssa> robmur01: That seems like a likely culprit, but display *does* come up, just login doesn't (it hangs on logs, or a blackscreen if you installed plymouth)
<robmur01> certainly kmscube doesn't like panfrost loading first
<alyssa> Maybe the Debian initramfs makes more assumptions
* alyssa can't tell if this is a Debian bug or a Panfrost bug
<robmur01> AFAICS the only visible difference should be the DRI numbering, so I'd be inclined to point the finger at GDM
<alyssa> robmur01: This occurs even with multi-user.target (just booting to a console, no GDM)
yann has joined #panfrost
adjtm has quit [Ping timeout: 240 seconds]
<robmur01> OK, that is indeed odd - I can't see how DRM drivers should have any influence there, unless the target's been given some weird dependencies.
<alyssa> Yeah, I dunno
<alyssa> I don't know Debian or the kernel well
<alyssa> ...which does make nervous about trying to switch to a Debian kernel with disk encryption :X
<robmur01> not being able to boot is good for security :P
<alyssa> :d
<alyssa> Okay, reproducibly,
<alyssa> Touching the touchscreen crashes gnome-shell on Kevin I, but works fine on Kevin II
<robmur01> different revisions with different touchscreen HW?
<alyssa> I mean, maybe?
<alyssa> Why would G-S care?
<robmur01> different drivers (or wrong driver in one case) == different behaviour
<robmur01> usually
<alyssa> It works fine in not G-S
<alyssa> (I have used this touchscreen extensively in MATE)
<robmur01> my fallback reasoning is "computers are the worst"
<alyssa> There's that
<paulk-leonov> maybe try with a fresh gnome session?
<paulk-leonov> just to troubleshoot
<paulk-leonov> gnome tends to be quite stateful apparently
adjtm has joined #panfrost
<bbrezillon> tomeu: I had a look at the "gitlab-ci job filtering" we talked about yesterday. Looks like we can have fine grained selection through variables through the "only: variables: " job section, but that implies tweaking each job to define the matching pattern
<bbrezillon> I tried doing it at the .ci-run-policy level http://code.bulix.org/45j45o-902472, but it doesn't seem to work
<tomeu> bbrezillon: any ideas why?
<tomeu> bbrezillon, alyssa: btw, did you see this? https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2253
<tomeu> seems to be some crash
<bbrezillon> tomeu: for the CI stuff, I'm not sure the right value of a gitlab regex is allowed to be a variable
<tomeu> yeah, could be
<bbrezillon> I tried various things, and everytime I tried to put a $SOMETHING as a right operand of the =~ it evalutes to false
<daniels> i wonder if you want $$
<alyssa> paulk-leonov: fresh how?
<bbrezillon> daniels: I can try that
<bbrezillon> daniels: BTW, do you know where I can find the code parsing/evaluating those conditions?
<narmstrong> bbrezillon: looking at the docs, seems the reges are only for the only: variables case
<narmstrong> *regex
<bbrezillon> narmstrong: yep
<bbrezillon> that's what I use it for
<bbrezillon> daniels: doesn't seem to help
<bbrezillon> also, it seems that even with the manual regex on job1, both job1 and job2 are executed
<bbrezillon> but it might be me struggling with regex
<narmstrong> bbrezillon: why on the same line ?
<paulk-leonov> alyssa: like, new user fresh or "mv .config .config-old" type of fresh :)
<narmstrong> only:
<narmstrong> - $MESA_CI_SCOPE == null
<narmstrong> variables:
<narmstrong> - $MESA_CI_SCOPE =~ /job1/
<paulk-leonov> alyssa: maybe .local too
<bbrezillon> hm, I really want the ||
<narmstrong> `Conjunction / Disjunction (introduced in GitLab 12.0)`
<narmstrong> is fd.o running gitlab 12.0 ?
<daniels> 12.3.5
<alyssa> paulk-leonov: Alright ..
<alyssa> Just as I was getting comfy :p
* alyssa is trying to switch to GNOME for her personal machine
<alyssa> we'll see how this goes
<daniels> bbrezillon: gitlab-runner != gitlab
<bbrezillon> daniels: oops
<paulk-leonov> alyssa: :)
<daniels> bbrezillon: the version you see there is the version of the local CI job execution engine on whatever is executing your jobs, not of the core gitlab installation
<bbrezillon> anyway, the || and && operators seem to work fine based on the tests I've done
<bbrezillon> daniels: thanks for the pointer
<bbrezillon> I think I have 2 problems here
<bbrezillon> - $PLOP =~ $PLIP does not work (at least not as I expect)
<bbrezillon> - $CI_JOB_NAME is not updated to job2 when it comes to "only:variables: " evaluation (the echo command prints the right value though)
alyssa_ has joined #panfrost
<alyssa_> Hello from Panfrost adventures...
<bbrezillon> having the "only: variables:" copied in each job would solves that, but it's likely to be omitted when new jobs are created
alyssa_ has left #panfrost [#panfrost]
<tomeu> alyssa: do you find it usable?
<paulk-leonov> alyssa: did moving the config/local folders do the trick to have it stop crashing?
raster has quit [Remote host closed the connection]
raster has joined #panfrost
Kwiboo has quit [Ping timeout: 240 seconds]
<robmur01> urjaman: I needed a moment of distraction, so here ya go - https://patchwork.freedesktop.org/patch/334873/
adjtm has quit [Ping timeout: 264 seconds]
<urjaman> robmur01: thanks (after half a day of running panfrost those messages are actually most of my dmesg :P)
<urjaman> but there was something else in there ... at some point (i havent been running anything more than firefox and xfce-terminal) there happened a handful of 'panfrost ffa30000.gpu: matching BO is not heap type (GPU VA = 14a00000)'s (with a warning stack trace) plus a TRANSLATION_FAULT_LEVEL2 after each one
<robmur01> yeah, I think there's a race on stopping jobs where we unmap stuff before the GPU has definitely stopped accessing it.
<robmur01> with a super-slow FPGA GPU this seems more or less 100% reproducible
adjtm has joined #panfrost
davidlt has quit [Ping timeout: 240 seconds]
Kwiboo has joined #panfrost
yann has quit [Ping timeout: 265 seconds]
<urjaman> okay umm trying to scroll (with the touchpad) a text file in mousepad causes Xorg to crash with "Segmentation fault at address 0x0"
* urjaman is somewhat amused that it's the text editor that manages to crash the graphics driver (or well something related) :P
yann has joined #panfrost
raster has quit [Remote host closed the connection]
davidlt has joined #panfrost
davidlt has quit [Read error: Connection reset by peer]
yann has quit [Ping timeout: 276 seconds]
yann|work has joined #panfrost
NeuroScr has joined #panfrost
NeuroScr has quit [Quit: NeuroScr]
adjtm_ has joined #panfrost
adjtm has quit [Quit: Leaving]
jailbox has quit [Ping timeout: 240 seconds]
AreaScout_ has quit [Ping timeout: 268 seconds]
jailbox has joined #panfrost
Depau has quit [Ping timeout: 240 seconds]
Depau has joined #panfrost
<urjaman> https://urja.dev/panfrost_oops_wq.txt managed to disable the kernel quite well (whilst still being alive enough to give me a dmesg over ssh)
<urjaman> panfrost_job_timedout -> drm_sched_increase_karma -> arch_spin_lock -> oops and that hung the workqueue is how i read that
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
mrfixit2001 has joined #panfrost