[05:35] KitsuWhooa: netbooting recent live cds over nfs seems broken, but it works with aoe, provided one enters 2 commands manually as initramfs-tools doesn't have code for that... [05:55] The command line I'm using is: APPEND ro initrd=pxelinux.cfg/casper/initrd boot=casper break=mount root=/dev/ether/e0.0p1 [05:55] And the commands needed at the initramfs breakpoint are: ipconfig enp0s17; modprobe aoe; exit [05:58] Unfortunately mate stopped producing an i386 iso so I can't test cosmic there, I'll try with xubuntu [06:06] Corrupted output in this one too: 01:00.0 VGA compatible controller [0300]: NVIDIA Corporation NV18 [GeForce4 MX 4000] [10de:0185] (rev c1) [06:10] Hmm, the corruption is related to the resolution. [06:10] Initial=1366x768=corrupted. Then I set 1024x768, works. Then I set again 1366x768, works again! [06:10] Same card, another monitor with 1280x1024=works without doing anything [06:46] KitsuWhooa: netbooting recent live cds over nfs seems broken <-- since I set up ltsp, I don't think it matters anymore [06:47] and I got tired yesterday and didn't look into it any further [06:47] computer was crashing too, so I need to run memtest86+ for a bit [06:47] I literally built it yesterday to test the graphics cards and picked random memory sticks that I'm not sure if they were in a working condition or not [06:52] (09:46:45 πμ) KitsuWhooa: since I set up ltsp, I don't think it matters anymore => eh, it's easier to download/test a bunch of ubuntu live cds than build a lot of different chroots (if one wants to test 16.04/18.04/17.10/18.10 etc) [06:52] that is true [07:18] alkisg: I just tried another nvidia card and got the same result [07:18] except when I tried to switch to a tty I saw "gpu lockup" and the machine froze :p [07:18] so I can't tell what card it is yet [07:23] ltsp also doesn't seem to let me log in over a tty [07:27] KitsuWhooa: ltsp-config lts.conf, sudo nano /var/lib/tftpboot/ltsp/i386/lts.conf [07:28] In lts.conf you can specify either SCREEN_02=shell to get a tty (you have an example there) [07:28] Ah, didn't know that [07:28] or even INIT_COMMAND_ROOT_HASH="sed 's!^root:[^:]*:!root:\$6\$p2LdWE6j\$PDd1TUzGvvIkj9SE8wbw1gA/MD66tHHlStqi1.qyv860oK47UnKcafSKqGp7cbgZUPlgyPv6giCVyCSCdJt1b0:!' -i /etc/shadow" [07:28] thanks [07:28] ...to allow you to login with root/qwer1234 [07:29] in the meantime, I blindly opened a terminal in the X session and piped the output to netcat [07:29] 01:00.0 VGA compatible controller: NVIDIA Corporation NV34 [GeForce FX 5200] (rev a1) [07:29] Yeah I think I saw the corrupted output with this in a school [07:29] KitsuWhooa, what's the default resolution? [07:29] Like, 1366x768? [07:29] 1680x1050 [07:29] Hrm, there goes my theory about "width not divisible by 8" :D [07:31] So, all three of my nvidia agp graphics cards exhibit the same issue [07:31] and one is a completely different architecture [07:31] I guess I'll try compiling older versions of software and trying it [07:32] also, not sure what you mean by "to get a tty" [07:32] KitsuWhooa: do you have another monitor around, to see if it works there? [07:32] I already have a tty1 on the client [07:32] Do install epoptes, it's rather handy [07:32] You right click > open terminal remotely [07:32] I might as well at this point [07:32] what kind of monitor are you thinking? [07:32] And you can run xorg commands on the client while sitting in your server [07:33] Anything else; in the school lab I'm trying, one model gets corrupted output, 3 other models don't [07:33] I have a few 1920x1080 ones, and a CRT that's 1024x768/1280x1024 [07:33] Great, I think you'll find at least one that works fine [07:33] let me try those and see what happens [07:34] I'll be heading to the school now, to test things locally [07:34] alright [07:34] I'll try to join irc from there [07:38] 1920x1080 also has the same problem [07:44] alkisg: sure enough, it renders fine at 1024x768@60 [07:53] MX200 has the same issue at 1680x1050@60 [08:40] KitsuWhooa: xrandr -s other-resolution; xrandr -s problematic-resolution fixes the issue for me, [08:41] but I've also seen that some monitors that dont have the issue, when you move a window in them, it shows termporary corruption, which goes away if I do the 2 xrandrs combo again there [08:41] So it surely sounds like something isn't initialized properly, and it is when I issue xrandr [08:46] Hm [08:52] KitsuWhooa: can you try this in lts.conf? So far it solves all problems for me: X_OPTION_01="\"PageFlip\" \"off\"" [08:56] Yeah that fixes it [08:56] I'm going to make this default for all nvidia cards in all schools here :D [08:56] How much of a performance loss does it cause? [08:57] I'm not sure [08:57] lol, dragging glxgears around makes it drop from 60 to 25FPS [08:59] I get the feeling this is an X server issue [09:16] I saw no performance loss with pageflip (youtube/glxgears), so I'll automatically set that from an ltsp initscript when i see nvidia cards [09:16] I'm going to try to bisect it if I have the time [09:17] currently dealing with inted crashing every 5 minutes [09:17] *inetd [09:17] and telling whoopsie to not report anything also failed [09:17] Meh, who needs that? :D [09:17] I don't know, it's what got installed in the VM [09:17] :p [09:20] alkisg: how do I get an ltsp client to boot an older kernel? [09:20] Apparently it doesn't use the one the host is running [09:23] unless it's in the squashfs image and I need to update it [09:33] Ah, looks like it's ltsp-update-kernels [09:48] (back) yup, that's it [09:49] It's still booting the new one after running it [09:49] unless I need to pass specific args [09:49] I didn't see anything regarding the kernel in lts.conf either [09:50] lts.conf is loaded after the kernel [09:50] I guess I can manually edit pxelinux.cfg/* [09:50] Yes, or you can use symlinks [09:50] I tried symlinking the initrd and vmlinuz files, but it loads the 4.15 ones directly [09:50] Just make sure the squashfs image does have the new kernel, and it's not just in tftp [09:51] so the symlinks were ignored [09:51] You'd need to write a "default" file with vmlinuz there, instead of vmlinuz-version [09:51] So both symlinks + manual pxelinux.cfg/default... [09:51] Ah I see [09:51] (if pxelinux.cfg/default isn't a symlink, ltsp doesn't touch it anymore, it's preserved) [09:53] Just make sure the squashfs image does have the new kernel, and it's not just in tftp <-- I updated the image while booted to the kernel I want to run, so I hope that will do it [09:54] Sure then it should contain all kernels, not just the one you booted [09:55] excellent [10:05] okay it's definitely not a drm issue [10:06] bug happens on 4.4.155-0404155-generic [10:08] (11:59:19 πμ) KitsuWhooa: I get the feeling this is an X server issue => what does this mean, why does it happen only with nvidia cards then? [10:08] I was thinking that the bug would be in the package xserver-xorg-video-nouveau, isn't that the case? [10:09] My guess is something changed in X that broke xf86-video-nouveau [10:10] Ah [10:11] Looking at the commits after the release of xserver-xorg-video-nouveau that's in 16.04, I don't see anything that would directly affect DRI2 [10:11] but I am going to try compiling the version that comes with 16.04 to make sure [10:12] Can't we just dpkg -i the 16.04 packages over the running ltsp client, and killall xorg? [10:12] That probably won't work [10:13] Ignoring deb version dependencies, there's probably an abi incompatibility [10:16] abi with what, if we dpkg all of xorg including the drivers? [10:16] i.e. xserver-xorg-*... [10:16] Ah [10:16] that might work, assuming any of these packages don't depend on anything else specifically [10:16] but what is the point of that? [10:17] Maybe verify that kernel/drm don't matter? [10:17] I don't know the subsystems well enough to know :D [10:17] Well, considering I just downgraded to 4.4 and it's still broken, I doubt it's the kernel [10:17] Btw, I wonder if we can force triple buffering and see if it helps [10:17] you cahn [10:17] *can [10:18] PageFlip -> on [10:18] SwapLimit -> 2 [10:18] ty, will try it tomorrow when I go back to the office or to a school [10:18] I'll try it now :p [10:18] Great! :) [10:18] I assume X_OPTION_02 will work too? [10:18] Yes [10:19] To make sure, cat /run/ltsp-xorg.conf after it boots [10:22] looks like X isn't starting :p [10:22] oh I'm dumb [10:22] I set PageFlip to off [10:25] okay yeah, it's not doing anything with SwapLimit set to 2 [10:27] Meaning output corrupted, or no output? [10:27] just stays in the tty, login screen never pops up [10:28] can you cat the xorg.conf file, to make sure there's no error due to e.g. bad escaping of quotes? [10:28] cat /run/ltsp-xorg.conf [10:28] Or, getltscfg -a | grep OPTION [10:29] no such file or directory for the conf file [10:29] looks like something broke [10:29] ...wait [10:29] I commented both X_OPTION lines and it still doesn't start [10:30] What's the output of `getltscfg -a | nc termbin.com 9999` [10:30] give me a bit, I rebooted again [10:32] Okay nevermind, I forgot to uncomment the ldm line when I enabled the tty [10:32] let's try the X options again [10:33] I have to say though, I wasn't expecting netboot to be this fast. USB1.1 is very painful [10:34] It's about as fast as booting from brand new rotational hard disk [10:34] Except if you're using gigabit, when it's faster [10:35] Right, I checked the ltsp-xorg.conf [10:35] it looks okay [10:35] but it's still not booting with PageFlip 2 [10:35] er [10:35] SwapLimit 21 [10:35] *2 [10:35] I can't type [10:35] check /var/log/Xorg.7.log.old [10:35] Xorg log says parse error, 2 is not a valid keyword [10:36] yeah was just looking at that [10:36] maybe I need to put 2 in quotes too [10:36] even though the manpage says integer [10:36] Possibly... [10:37] nope, same error [10:37] cat /run/ltsp-xorg.conf | nc termbin.com 9999 [10:37] I'll just send it to my desktop instead :p [10:38] https://tasossah.com/ltsp-xorg.conf [10:38] it looks okay to me [10:39] only other thing I can think of is that SwapLimit was removed [10:40] Apparently 2 is the default for Xorg server > 1.12 [10:43] I just rebooted again and it worked this time. What?! [10:43] Worked as in X started. Screen is still garbled [12:17] I compiled xf86-video-nouveau .13 and it segfaults after showing a bit of corruption, so my guess is it was either broken between .12 and .13, or it's on X