Summary: | Celeron N3150 : random display off instants and occasional full-system freezes | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Product: | DRI | Reporter: | René J.V. Bertin <rjvbertin> | ||||||||
Component: | DRM/Intel | Assignee: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||||
Status: | CLOSED FIXED | QA Contact: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||||
Severity: | major | ||||||||||
Priority: | medium | CC: | intel-gfx-bugs, mattias.slabanja, rjvbertin | ||||||||
Version: | unspecified | ||||||||||
Hardware: | x86-64 (AMD64) | ||||||||||
OS: | Linux (All) | ||||||||||
Whiteboard: | |||||||||||
i915 platform: | BSW/CHT | i915 features: | |||||||||
Attachments: |
|
Description
René J.V. Bertin
2016-06-15 18:21:18 UTC
(In reply to René J.V. Bertin from comment #0) > - mainline kernel 4.5.6, home-built Please try 4.6.2 or newer. Building right now, will take >3h on my hardware. In the meantime, is there somewhere where one can see exactly what classifies as "preliminary hardware" w.r.t. a kernel release? (In reply to René J.V. Bertin from comment #2) > Building right now, will take >3h on my hardware. Feels like you must be doing something wrong. For trying out stuff, it's usually not necessary to use the full distro config, for example. > In the meantime, is there somewhere where one can see exactly what > classifies as "preliminary hardware" w.r.t. a kernel release? It's more like "preliminary support" for certain hardware in certain kernel versions. There's no table or anything you could look at, though one could certainly be created. See drivers/gpu/drm/i915/i915_drv.c and look for .is_preliminary = 1 to see for which platforms the support is still considered preliminary (in the kernel sources you have). 'git log --grep=preliminary -- drivers/gpu/drm/i915/i915_drv.c' will give you a fairly accurate list of when the support was considered good enough not to be preliminary for each platform. 'git tag --contains <commit-id> | grep "^v" | sort -V | head -1' on a commit-id to see which kernel release that happened. Oh, I know I could build a minimal kernel, but I was planning to upgrade to 4.6x anyway once it hit 4.6.2 or above, also for the wife's machine which has a 6th gen i3 and thus stands to gain even more from a newer kernel. And which doesn't mind churning away on the build while I use my system for other things. I'll have a look at your "preliminary suggestions" and see if that tells me things that I know how to link to specific hardware that actually means something to me. I search the 4.6.2 changelog for i915 already, and that only confirmed that quite a few issues were addressed ^^ One other thing: I've been presuming that my Celeron won't choke on code compiled with -march=core2 - not mistakingly I hope? I just noticed this while creating the 4.6.2 initrd:
W: Possible missing firmware /lib/firmware/i915/skl_guc_ver4.bin for module i915
Supposing that's a relevant missing file, where could I find it? I have these:
> ls /lib/firmware/i915/
bxt_dmc_ver1_04.bin skl_dmc_ver1_19.bin skl_dmc_ver1_21.bin skl_guc_ver1_1059.bin
bxt_dmc_ver1.bin skl_dmc_ver1_20.bin skl_dmc_ver1.bin skl_guc_ver1.bin
(In reply to René J.V. Bertin from comment #6) > I just noticed this while creating the 4.6.2 initrd: > > W: Possible missing firmware /lib/firmware/i915/skl_guc_ver4.bin for module > i915 > > Supposing that's a relevant missing file, where could I find it? I have > these: > > > ls /lib/firmware/i915/ > bxt_dmc_ver1_04.bin skl_dmc_ver1_19.bin skl_dmc_ver1_21.bin > skl_guc_ver1_1059.bin > bxt_dmc_ver1.bin skl_dmc_ver1_20.bin skl_dmc_ver1.bin > skl_guc_ver1.bin Only relevant for Skylake. It should be in linux-firmware or https://01.org/linuxgraphics/downloads Well, that was a painful experience. I installed the 4.6.2 kernel plus the 1.57 firmware (cf. https://bbs.archlinux.org/viewtopic.php?pid=1533863#p1533863), rebooted. Apart from the illegible text during the initial initrd loading which 4.6.2 didn't change all looked fine until I was about to come report here. Switching tabs a couple of times my display froze except for the mouse cursor. I tried the sleep/wake trick, and saw something about the GPU being banned for some reason on the virtual terminal, and then when back in X11 everything started to flicker until, apparently, KWin crashed. I coud quit most KDE apps, but Chrome no longer responded to keystrokes. I booted back into 4.5.6, and had similar things happen; compositing effect no longer worked, switching to a virtual terminal and back to X11 gave me a black screen, and those are just the dysfunctions I remember. It took multiple reboots and quite a bit of trial and error to get back to where I came from. I tried one last time with 4.6.2, and again everything was fine until I started Chrome. This time I just let it start, and at some point KWin crashed and Chrome again couldn't be quit. I've put back the firmware files I had before, and kicked 4.6.2 off the system (even if I know it shouldn't be required). My next attempt will be to test the UXA acceleration again. Clearly 4.6.2 isn't ripe yet for my hardware (it seems to run fine on the wife's i3). Normally I wouldn't mind filing bug reports and testing patches, but not if it's so hard to get back to a working state. I've tried once more, this time from the external boot/install I use with the i3 laptop (which also excludes potential artefacts from running off a ZFS pool). Same thing: as soon as I started Chrome and surfed to youtube.com I started getting problems. I'll be attaching kern.log and dmesg output; it looks like 4.6.2 causes GPU crashes as a result of certain compositing (DRI) operations. The crashing PID mentioned in kern.log is the X server. Created attachment 124605 [details]
kern.log with 4.6.2
Created attachment 124606 [details]
dmesg, kernel 4.6.2
Reporter, is this still issue with latest kernel? I can't tell, I've been too busy with other things to follow kernel development on a secondary install. My main installation runs off a ZFS root which for now keeps me at the 4.5 kernel series, currently 4.5.7 which runs fine. I'm hoping the pending 4.7+ support will land soon in an official ZoL release. I do think though that the full-system freezes were unrelated to graphics, but the result of thermald trying to throttle down the CPU. I have now been able to update my ZFS drivers, build, install and test the 4.8.15 kernel for a while. No display deactivations, no freezes. As I said in my previous comment, the freezes were most likely due to the thermald. I stopped getting them when I uninstalled thermald, and since this is a notebook where thermal and energy control is handled fully by the firmware I'm not seeing overheating problems either. I haven't had display deactivations for a while either, probably since I upgraded to Ubuntu's lts-xenial Xorg and Mesa. I'm closing this as fixed (would be nice to know what was fixed of course). I hope I won't have to reopen it when I try 4.9.2 ;) (In reply to René J.V. Bertin from comment #14) > I have now been able to update my ZFS drivers, build, install and test the > 4.8.15 kernel for a while. > > No display deactivations, no freezes. > > As I said in my previous comment, the freezes were most likely due to the > thermald. I stopped getting them when I uninstalled thermald, and since this > is a notebook where thermal and energy control is handled fully by the > firmware I'm not seeing overheating problems either. > > I haven't had display deactivations for a while either, probably since I > upgraded to Ubuntu's lts-xenial Xorg and Mesa. > > I'm closing this as fixed (would be nice to know what was fixed of course). > I hope I won't have to reopen it when I try 4.9.2 ;) thanks René for your feedback and confirmation. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.