12:30kale: short qeastion, can nouveau now set the clock frequency for pascal? or is it still awaiting nvidia?
13:08karolherbst: kale: still waiting on nvidia
18:58codehotter: what am I doing wrong if I have no /dev/dri or /sys/class/drm
19:01imirkin: you're not loading a drm driver
19:02imirkin: or your gpu is not recognized by the driver you're loading
19:07codehotter: fixed with dnf install kernel-modules
20:27letterrip: hi all, are there any detailed guides for troubleshooting?
20:28letterrip: for someone without a great deal of familiarity with troubleshooting video graphics on linux?
20:29imirkin: no real guides for anyone, no matter the level of familiarity
20:29letterrip: for kernel modesetting, currently i can only boot with nomodeset -
20:29imirkin: which gpu do you have, and which kernel?
20:32letterrip: ancient - geforce go 7200, kernel is 4.9.0-8-amd64
20:33letterrip: i suspect it might be a incorrect edid
20:34imirkin: and what happens when you don't disable modesetting?
20:34letterrip: if i don't disable mode setting the kernel quits booting after the crng is initialized
20:34imirkin: define 'quits'
20:34imirkin: kernel doesn't boot - it invokes init, which then does stuff.
20:35letterrip: the last message i see printed is about initializing the crng, then all disk activity ceases
20:35imirkin: fun - in my boot, nouveau loads right after crng
20:35imirkin: i suspect that has something to do with crng being the last thing
20:35imirkin: and then nouveau is the first module
20:36karolherbst: letterrip: do you have a second machine? maybe you can ssh into it?
20:36karolherbst: or blacklist nouveau and modprobe it later
20:36karolherbst: and see what dmesg throws at you
20:36imirkin: yeah, the "modprobe later" thing is a decent way of getting proper logs
20:36letterrip: karolherbst: ok will give that a try
20:46letterrip: karolherbst: ok rebooted with nouveau blacklisted
20:46letterrip: what info do you need next?
20:46karolherbst: letterrip: best would be to ssh into the machine and modprobe nouveau
20:46karolherbst: so you don't rely on the displays to still work
20:48letterrip: karolherbst: ok, can give that a try as well
21:26letterrip: karolherbst: is wireless and the ssh server started before i'll get to the graphics? or will i need to do so over a cat5 cable?
21:28karolherbst: letterrip: depends on your system
21:28karolherbst: sometimes the wifi password is encrypted so the users wallet has to be unlocked first
21:31imirkin: letterrip: i like to ssh in, and run "dmesg -w"
21:31imirkin: that way even if it hangs, i still get most of the messages
21:31letterrip: anything else i should no before i try?
21:32imirkin: make sure X isn't running
21:53LetterRip: ok was able to do it
21:53LetterRip: I found a similar issue for amdgpu - relates to the GPU sharing memory with the CPU
21:54LetterRip: so i included output of lspci -vvv for my laptop as well
21:57imirkin: LetterRip: is there something non-standard about your setup?
21:57LetterRip: actually that isn't the full output from the lspci -vvv i just included the GPU part, can add the rest if desired
21:58imirkin: i.e. is this a plain ol' laptop/
21:58LetterRip: imirkin - not that i know of
21:58LetterRip: Sony VAIO AR320-e
21:58LetterRip: so fairly old
21:58imirkin: no funny virtualization
21:58imirkin: could you include the whole boot log?
21:59imirkin: kernel bug in mm/pat.c
21:59imirkin: that means that PAT isn't enabled but somehow the kernel thinks it is?
21:59imirkin: try booting with "nopat"
22:00LetterRip: imirkin - https://pastebin.ubuntu.com/p/PZY83n67P2/
22:00imirkin: [ 0.472088] pci 0000:01:00.0: can't claim BAR 1 [mem 0xb0000000-0xbfffffff 64bit pref]: no compatible bridge window
22:00imirkin: that's bad.
22:01imirkin: that's the gpu pci device
22:02imirkin: unfortunately this is above my pay grade. upgrading your kernel could randomly fix this
22:02LetterRip: imirkin here is also the complete lspci -vvv if that helps - https://pastebin.ubuntu.com/p/9PG3Sttxst/
22:04LetterRip: imirkin - the GPU is shared memory - so 256 MB dedicated, and then apparently it can 'steal' from the GPU, (and that bug report seemed to suggest that the CPU was 'stealing' all of the ram of the GPU- I think?)
22:04LetterRip: er steal from the 'CPU'
22:04LetterRip: something that was fairly common for older laptops
22:04imirkin: that GPU has dedicated vram
22:04imirkin: the stealing is ... a misnomer
22:05imirkin: it was just the terminology of the day
22:05imirkin: when they wanted to say that the gpu can access 256mb ram, even though it only has 128 on-board
22:05imirkin: or whatever
22:05imirkin: the remainder is regular system ram
22:05LetterRip: couldn't think of the right word as i was typing :)
22:05imirkin: it's not "stolen", just allocated.
22:05LetterRip: anyway a shared memory model
22:06imirkin: unlike some IGP's which actually carve out a region of physical memory for themselves
22:06imirkin: anyways, the issue is that some BAR isn't being allocated
22:06imirkin: and it happens to be an important BAR
22:06imirkin: i can't tell you anything about that allocatino process
22:07LetterRip: BAR means?
22:07imirkin: base address register
22:07imirkin: but really it's the way that the cpu can do mmio
22:07imirkin: LetterRip: try booting with "pci=nobar"
22:07imirkin: nobar [X86] Do not assign address space to the
22:07imirkin: BARs that weren't assigned by the BIOS.
22:08imirkin: and separately, try pci=nocrs
22:09LetterRip: will do
22:09imirkin: [ 0.472088] pci 0000:01:00.0: can't claim BAR 1 [mem 0xb0000000-0xbfffffff 64bit pref]: no compatible bridge window
22:10imirkin: basically that's the message you need to "fix"
22:10imirkin: or e.g. [ 0.512286] pci 0000:01:00.0: BAR 1: no space for [mem size 0x10000000 64bit pref]
22:13LetterRip: no luck with pci=nobar, will try the next one
22:14LetterRip: nocrs is looking promising, font changed
22:16LetterRip: imirkin, it completed booting, lightdm loaded, and am in the desktop, the mouse now looks like a square with two rows of chinese characters
22:18LetterRip: yep, greatly appreciated :)
22:19imirkin: pastebin new dmesg?
22:20LetterRip: will do
22:22imirkin: that dmesg doesn't have pci=nocrs btw
22:23imirkin: or nouveau loaded
22:23imirkin: and it's a different machine
22:23imirkin: you forgot to ssh :)
22:24LetterRip: i did a 'reset and clear' - must have killed the ssh connection
22:26LetterRip: imirkin, there you go, sorry about that :)
22:26imirkin: that all seems perfectly happy
22:27LetterRip: issue is similar to this
22:27LetterRip: but my square looks differnt
22:28imirkin: yeah, that makes sense... the cursor somehow got messed up
22:28LetterRip: also the mouseclicks aren't being recieved/acted on
22:28imirkin: what happens when you move the mouse around to areas of the screen where the cursor would change?
22:29imirkin: the hotspot is the top-left area of the square
22:29imirkin: i.e. the literal top-left pixel
22:30LetterRip: the square looks the same regardless of location
22:30LetterRip: but it is transparentish
22:30imirkin: yeah, probably just random data
22:30imirkin: question is ... how did that get messed up
22:31LetterRip: perhaps not 'random' since it looks like two vertical stripes of chinese characters (though probably not actually characters)
22:32imirkin: did nouveau ever work on this hardware?
22:32LetterRip: this is the first i've tried with this laptop
22:32LetterRip: so for me - defintely no - and haven't found online anyone who has provided a report either way
22:32LetterRip: anywho appreciate your help,
22:33imirkin: i'm just reading over some stuff to see what's up
22:33LetterRip: will do some work on my own after dinner
22:33LetterRip: and report back if i find anything interesting
22:34imirkin: one thing to try is to edit drivers/gpu/drm/nouveau/dispnv04/hw.h:nv_fix_nv40_hw_cursor and make it do nothing at all
22:34imirkin: that said, i've never had problems with it. but i've also never tried G7x hardware (only the slightly earlier series), and certainly not a mobile part
22:35LetterRip: will give that a try
22:36LetterRip: thanks again for your help, and have a good day/night
23:05LetterRip: imirkin, it looks like actually the entire display isn't updating
23:06imirkin: you just get a fixed image?
23:06LetterRip: i just realized that the date and time hadn't updated since it booted and some of the icons hadn't moved to their correct spot
23:06LetterRip: yes it is just a fixed image
23:06LetterRip: partway through the desktop initialing
23:06imirkin: so that means the whole modeset failed...
23:07imirkin: what happens if you flip to a vt and back?
23:07imirkin: e.g. ctrl+alt+f1
23:07imirkin: and then ctrl+alt+f7
23:07LetterRip: i can't do that
23:07imirkin: why not/
23:07LetterRip: also realized that that wasn't working
23:07imirkin: can you ssh in
23:07imirkin: and run "chvt 1" as root
23:08imirkin: is this a wayland-based setup by any chance?
23:09LetterRip: running it, but no response at the terminal (block curson in the terminal is waiting)
23:09imirkin: anything new in dmesg?
23:10LetterRip: i don't think so
23:10imirkin: well, first things first --
23:10imirkin: try a fresh kernel
23:11imirkin: i can't think of anything in the past couple years that would have affected this
23:11imirkin: but i also don't want to debug already-fixed bugs
23:12LetterRip: is the current dmesg
23:15LetterRip: imirkin, ok will give that a try