01:44Hydr0p0nX: I'm running RHEL 7 (Dev subscription) with an AMD and NVidia graphics card (both physical), I want to use my AMD as passthrough to Win7 VM hosted in KVM. When I blacklist the radeon driver and add it to pci-stub.ids, I get an X crash; with the radeon driver enabled, I get an X display on both the nvidia and radeon displays with no problem, xorg log is at http://paste.ubuntu.com/23047954/ ... i've remved the nomodeset settings and a
02:55Tom^: Hydr0p0nX: for some reason its still trying to load up on the radeon card.
02:57Tom^: so i would assume the kernel still thinks it exists and you havent isolated / blacklisted it properly yet
03:15Hydr0p0nX: this is after uninstalling the xorg ati driver and blacklisting in grub and /etc/modprobe.d - http://paste.ubuntu.com/23048055/
03:44Tom^: Hydr0p0nX: what nvidia gpu is this?
11:09RSpliet: Hydr0p0nX: might sound silly, but file your bug with Red Hat in this case. It just so happens to be that the person responsible for the kernel side of nouveau is a Red Hat employee ;-)
11:10RSpliet: also, he lives in Australia, so bug tracker is the best means of getting his attention
11:12RSpliet: having said that, in this case your graphics stack might simply be too much out of date to properly support your particular NVIDIA GPU
11:16RSpliet: in particular, your kernel might not properly recognise your GPU, check your dmesg logs
12:05karolherbst_work: RSpliet: odd, his gpu should be a gk107 according to the pci id, but we can't really rely on that 100%
12:31RSpliet: karolherbst_work: I know, GTX650 popped up when I googled it
12:32karolherbst_work: 3.10 kernel :O
12:33karolherbst_work: how did I miss that
12:34karolherbst_work: mhh, 3.10 has support for nvf0
12:34RSpliet: yeah, but nobody truly knows what that really means. Red Hat stable kernels tended to be patched up so strongly it might even be FreeBSD at this point ;-)
12:35karolherbst_work: well I checked the kernel longterm branch
12:35RSpliet: Ben probably knows
12:35karolherbst_work: I guess they don't remove such features?
12:35RSpliet: I don't know what nouveau bits have been backported
12:35karolherbst_work: right, then I check plain 3.10
12:37karolherbst_work: yep, gk107 should be supported
12:37karolherbst_work: but maybe the ddx is too ol
12:37RSpliet: ddx is 1.0.11, should be new enough
12:38RSpliet: but better to wait for feedback and let Ben handle this one
17:41NanoSector: hey, it's the guy with reclocking issues from yesterday... anyone in for some sleep issues? :D
17:43NanoSector: I know it's related because I can sleep with nvidia just fine but nouveau just gets stuck on resume
17:48Calinou: usually it's the other way around
17:48Calinou: ie. sleep works fine with open source drivers, almost never with proprietary drivers
17:57imirkin: sounds like sleep works just fine...
17:57imirkin: waking up is the issue
17:57imirkin: i get that a lot myself
17:57imirkin: stupid alarms
18:02NanoSector: imirkin: yeah sorry
18:07imirkin: NanoSector: i forget, are you on a laptop? if so, does this happen with runpm=0?
18:07NanoSector: i'm on a laptop yeah, i'll try with runpm=0
18:10NanoSector: imirkin: no, still freezes on resume
18:11imirkin: NanoSector: any chance you can get logs? also, iirc, there's a way to to a suspend/resume without *actually* doing it
18:11NanoSector: i can journalctl previous boots
18:11imirkin: alllso, is this s2ram or s2disk?
18:11imirkin: well, can you e.g. ssh into it?
18:12NanoSector: note to self: .149
18:12NanoSector: i'll try
18:12imirkin: hope dhcp is stable :)
18:13NanoSector: oh, enabling sshd helps
18:13imirkin: just stick a /bin/sh -i into inetd.conf :)
18:19imirkin: [that was a joke, btw]
18:19imirkin: [before you go and do that]
18:20NanoSector: yeah it's dead on resume, no sshing into it
18:20imirkin: ok =/
18:21imirkin: one thing you could do to get logs is set up netconsole
18:21imirkin: hopefully it resumes enough to send logs over network
18:21imirkin: definitely wired would be preferable in that case ;)
18:25NanoSector: could be TLP maybe screwing things up, I'll disable that
18:25NanoSector: i'll also give reclocking another shot with it disabled
18:26NanoSector: nope not tlp
18:26NanoSector: i'll set up netconsole
18:27NanoSector: the kernel does come up because sometimes i get a picture of just my cursor on the screen
18:31imirkin: i wonder if it's an intel issue, since the main screen is intel..
18:32imirkin: [classy move #1: blame the other guy]
18:34NanoSector: sorry, got distracted, i'll set up netconsole now
18:34NanoSector: youtube's such a wonderful palce
18:34NanoSector: *place, even
18:38NanoSector: note to self again: enp5s0f1
18:38NanoSector: who invented these device names :|
18:38imirkin: annoying people
18:38imirkin: you can fix it
18:38imirkin: add empty files in /etc/udev/rules.d/ called 80-net-name-slot.rules and 80-net-setup-link.rules
18:38imirkin: and everything will be fixed
18:39karolherbst: NanoSector: I do hope I didn't mess up suspending again, but it should work actually... or maybe if you use my branch, it is outdated?
18:40NanoSector: i'm using your reclocking branch, yes
18:40NanoSector: also sorry if you find me annoying for throwing all these things at you; I'm just hoping to help you fix some issues for other people :)
18:41karolherbst: nah, people writing stuff can't annoy me, usually. Because you have always the choice to not read something on the internet :p
18:42NanoSector: haha true
18:42imirkin: sounds like a challenge...
18:42NanoSector: so I booted with firstname.lastname@example.org/enp5s0f1,email@example.com/
18:43NanoSector: where .149 is my laptop, and .221 is my desktop
18:43NanoSector: and I ran 'nc -u -l -p 6666' on my desktop
18:43NanoSector: but nothing happens?
18:43imirkin: i usually do netconsole=@/firstname.lastname@example.org/ or something along those lines
18:43imirkin: i wonder if it doesn't work coz it wants eth0 or something
18:43NanoSector: oh my link is down
18:44karolherbst: oh my
18:45NanoSector: work dammit :S
18:45karolherbst: it only gets new stuff though
18:45karolherbst: push stuff through /dev/kmsg
18:45NanoSector: just 'echo hello world > /dev/kmsg?
18:45karolherbst: anything should do
18:45karolherbst: it should at least appear in dmesgt
18:46NanoSector: it did
18:46NanoSector: not on my desktop tho
18:46NanoSector: i'll try with the more minimal command from imirkin
18:47imirkin: i don't remember the exact thing unfortunately
18:47imirkin: but basically the majority of those parameters are optional
18:47imirkin: but the separators are required
18:47karolherbst: NanoSector: you can reconfigure netconsole at runtiome though
18:48karolherbst: NanoSector: do you have netconsole loaded at least?
18:48NanoSector: that's the structure according to arch wiki
18:48imirkin: right, so you do...
18:48imirkin: and that will send udp to port 6666
18:49NanoSector: nothing on desktop, will check if netconsole is loaded
18:49NanoSector: i usually make those rookie mistakes :x
18:49NanoSector: no not loaded
18:49imirkin: ah, oops
18:49imirkin: also do like 'dmesg -n 8'
18:51karolherbst: just found the best message in dmesg: "[311102.829462] done." ...
18:54NanoSector: huh, systemd-boot might be cutting off my command line
18:54NanoSector: bad D:
18:55karolherbst: it shouldn't
18:55karolherbst: you just need to load netconsole
18:55NanoSector: yeah but if i include a trailing slash in the parameter
18:55NanoSector: and check via dmesg
18:55Tom^: lol just found a cute systemd bug
18:55NanoSector: it's not there
18:55Tom^: Startup finished in 3min 29.679s (firmware) + 3.104s (loader) + 3.732s (kernel) + 642ms (userspace) = 3min 37.158s
18:55Tom^: maths? plz.
18:55NanoSector: yeah that's the same here
18:56NanoSector: well, it behaves now
18:56NanoSector: Startup finished in 10.420s (firmware) + 3.598s (loader) + 4.453s (kernel) + 4.233s (userspace) = 22.706s
18:56karolherbst: Tom^: are you serious? :D
18:56Tom^: it takes about 30 seconds but yea it seem to think it takes 3 mins
18:57karolherbst: guess what the first number means
18:57NanoSector: i think the uefi firmware reports the amount it took to the bootloader tho, so it's not really systemd's fault
18:57karolherbst: as if systemd could messure time before even the firmware was started
18:57karolherbst: if that would be true, I would have bigger worries
18:58NanoSector: by the time we get apps in UEFI firmware i start to get worried
18:58karolherbst: Tom^: uefi firmwares are one of the worst pieces of code... ever
18:58karolherbst: NanoSector: we already have like forever?
18:58NanoSector: like installable apps
18:58karolherbst: NanoSector: like grub? or the kernel?
18:58karolherbst: grub is installable
18:58Tom^: karolherbst: well yea, everytime i reboot a new entry is added for my bootloader
18:59NanoSector: no i don't mean bootloader :P something which nests itself inside uefi
18:59Tom^: so i sort of have to clean it out a bit with efibootmgr from time to time, because im worried its gonna buffer overflow my uefi
18:59NanoSector: or whatever
18:59karolherbst: NanoSector: linux kernel?
18:59NanoSector: Tom^: systemd-boot does that on my board too, grub does not
18:59Tom^: NanoSector: im not alone then it seems
18:59NanoSector: karolherbst: i guess
18:59NanoSector: Tom^: no
19:00NanoSector: Tom^: Gigabyte board, I assume?
19:00Tom^: uhm, i think so yea
19:00karolherbst: "/boot/EFI/LINUX/LINUX.efi: Linux kernel x86 boot executable bzImage, version 4.7.0-gentoo (root@) #1 SMP PREEMPT Sun Jul 31 11:22:09 CEST 20, RO-rootFS, swap_dev 0x7, Normal VGA" :p
19:00NanoSector: karolherbst: efistub?
19:01NanoSector: always was too lazy to set that up
19:01NanoSector: do i also set loglevel 8 or soemthing?
19:01karolherbst: modules signed with one time keys :3
19:01NanoSector: because no matter what i try, nothing happens on desktop pc
19:02karolherbst: does netconsole show in lsmod?
19:02NanoSector: it even shows up in dmesg
19:03karolherbst: well there is always wireshark or so
19:06NanoSector: it doesn't even work if i use localhost
19:06karolherbst: how could it
19:07NanoSector: and netconsole shows up in lsmod
19:07NanoSector: I mean on my laptop
19:07karolherbst: how could it
19:07karolherbst: the packages are sent to 192.168.1.221
19:07karolherbst: not 127.0.0.1 :p
19:08NanoSector: no i set localhost/127.0.0.1 in the kernel parameters as target ip
19:08karolherbst: mhh k
19:08karolherbst: no idea if netconsole sends to loopback though
19:10NanoSector: oh setting it up manually works
19:10NanoSector: so there's that
19:11NanoSector: nothing on resume
19:11NanoSector: like, no output to desktop
19:11karolherbst: does that hapen with nouveau not loaded?
19:12karolherbst: wait, we could do something else
19:12NanoSector: *reboot, take 50 :D*
19:12karolherbst: NanoSector: do you have pstore mounted?
19:12NanoSector: doubti t
19:12NanoSector: *doubt it
19:12karolherbst: systemd will mount it automatically
19:12NanoSector: where should it be mounted?
19:13NanoSector: yes, it's empty
19:13karolherbst: is it mounted?
19:13NanoSector: 'df' shows nothing mounted to that directory
19:13NanoSector: it exists though
19:13karolherbst: df won't show it
19:14karolherbst: check mount
19:14NanoSector: yes, it's mounted
19:15NanoSector: pstore on /sys/fs/pstore type pstore (rw,nosuid,nodev,noexec,relatime)
19:15karolherbst: what does this show:
19:15karolherbst: cat /proc/config.gz | gzip --decompress | grep PSTORE
19:18Tom^: hm just realized that my rpi would be a quite neat little netconsole device and have it hookedup on the second monitor input
19:20NanoSector: wb karolherbst
19:20karolherbst: seriously, how hard can it be to provide stable internet...
19:20karolherbst: I get like every day a disconnect of a few minutes
19:20karolherbst: from the ISP
19:21karolherbst: NanoSector: .... those guys
19:21karolherbst: why is there no CONFIG_EFI_VARS_PSTORE=y ...
19:21Tom^: it isnt your router then?
19:21karolherbst: Tom^: nope
19:21karolherbst: Tom^: From 192.0.0.2 icmp_seq=163 Destination Net Unreachable
19:22NanoSector: karolherbst: dunno, this is the stock Arch kernel
19:22karolherbst: it is stupid
19:23karolherbst: NanoSector: mind recompiling the kernel with that option set?
19:24NanoSector: karolherbst: I would if I understood the process and had time :P
19:24karolherbst: those arch guys, but I think it is wrongly configured like everywhere
19:24NanoSector: compiling a kernel will probably take years on this thing
19:25karolherbst: it shouldn't
19:25karolherbst: 1 hour is a long time
19:25NanoSector: although might be a good time to finally get a kernel based on modprobed-db
19:28NanoSector: ...if i had that installed
19:28NanoSector: well i'll grab linux-mainline from the aur and follow the wiki then I guess
19:28NanoSector: oh that's 4.8 now
19:29NanoSector: karolherbst: don't suppose your kepler reclocking branch works on 4.8 or is it obsolete by then?
19:30karolherbst: 4.8 isn't even released yezt
19:30NanoSector: rc1 that is
19:31NanoSector: ok it's building with that option
19:31NanoSector: i'll also rebuild regular linux with it
19:31NanoSector: with abs
19:32Tom^: modprobed-db is quite neat actually
19:32NanoSector: it is
19:32Tom^: have had it running for a while and got most if not all my modules in it by now
19:33NanoSector: ran it on my homeserver, reduced module directory size from 180 MB to just 17MB
19:33NanoSector: i don't plug in devices to it except a usb keyboard at times
19:33karolherbst: yeah, and then you don't get netconsole, cause you never used it before :p
19:33karolherbst: 17MB, how much
19:33Tom^: then you add it to modprobe-db and recompile. ;)
19:34karolherbst: I have 19MB, and nvidia is 17MB of that :p
19:34Tom^: it takes like 5 minutes when you have things set up
19:34karolherbst: without nvidia I have 7.8MB kernel + 2MB modules
19:34NanoSector: oh the actual modules are 3.6 MB
19:34NanoSector: 12.4mb kernel
19:35karolherbst: mine isn't even optimized for size and uses the crappiest of all compression algos
19:36NanoSector: so laptop is now compiling 2 kernels at once
19:38NanoSector: heh biggest driver on server is i915
20:17urmet: size optimization sounds kind of good on paper. smaller code fits in cpu cache = fast.
20:17urmet: and compression is all about decompression speed
20:37karolherbst: urmet: well optimized for size usually creates rather bad code
20:38urmet: it was a bigger benefit before. when memory was slower
20:39urmet: and caches were smaller
20:39karolherbst: memory is slow today
20:39NanoSector: it's still compiling >.>
20:39NanoSector: told you it'll take a year if not mroe
20:39urmet: slow but faster than before
20:39karolherbst: urmet: well, flash drives are also faster than before
20:40karolherbst: but compared to CPUs, memory speed increased slower
20:40urmet: most of kernel text is not pageable?
20:40urmet: so disk speed doesn't matter
20:40karolherbst: that wasn't the point
20:41urmet: benchmarking real-world situations is the only true guide
20:44urmet: i like going all out gentoo on my kernels - -march=native supposedly it should result in faster and/or smaller code :)
20:45karolherbst: smaller code isn't always better, especially if that leads to more branching
20:46karolherbst: loop unrolling is an optimization which usually leads to faster code, but bigger code size
20:48urmet: and on desktop use you won't notice either way :P
20:51urmet: or just ignore me
20:51urmet:has mush for brains - 5 hours drive from vacation to help with office moving for 12 hours
20:52imirkin: you sure know how to vacation
20:52urmet: i know more about networking than any other guy here. so i'm kind of needed
20:53urmet: today was hardware, tomorrow - probably 12 hours of network configuration
20:54Tom^: NanoSector: did you forget to change makeflags?
20:55Tom^: NanoSector: MAKEFLAGS in /etc/makepkg.conf
20:55Tom^: ok good, :p
20:55Tom^: NanoSector: get modprobed-db =D
20:56NanoSector: still running
20:56NanoSector: and yes
20:56NanoSector: i just did that
20:56NanoSector: fuck this
20:56urmet: modprobed-db ?
20:56Tom^: urmet: keeps a database of modules you have loaded, so localmodconfig can then use those
20:56urmet: ooh, shiny
20:56Tom^: or rather it modprobes it all before localmodconfig :p
20:56urmet: too bad I don't use arch
20:57Tom^: it wouldnt be to hard to code something like this yourself tho
20:57NanoSector: really thinking of building myself a mainline kernel with modprobed-db on my desktop
20:57NanoSector: iirc 4.8 rc1 has experimental amdgpu for gcn 1.0
20:58urmet: Tom^: seems like effort. I like the ease of use that gentoo provides me
20:58urmet:should go to sleep
21:08NanoSector: it's done in less than an eternity
21:12NanoSector: karolherbst: enabled that kernel option and there's still nothing in /sys/fs/pstore
21:12NanoSector: and it's still mounted
21:12NanoSector: and yes i have rebooted
21:18imirkin: urmet: ask for a raise :)
21:20urmet: 1) configure new network amazingly
21:20urmet: 2) ...
21:20urmet: 3) profit
21:55karolherbst: NanoSector: well, you have to crash the kernel
21:55karolherbst: it won't magically create entries
22:21NanoSector: karolherbst: still empty after hanging on resume and rebooting
22:25NanoSector: thinking i might just give up on that... lol