I have laptop with i7-3520M, Ubuntu 16.04 and latest oibaf mesa/drivers. From time to time (not always) when laptop comes back from sleep majority of letters doesn't get rendered. All decorations and everything is and letters are lost - so I see empty button frames, for example. In some cases some letters may be visible immediately after resume but when I try to interact with this particular window they disappear. This doesn't happen all the time but is quite often and I also have another laptop with Skylake CPU and similar software setup and I've never seen it happening there.
Nikolay can you attach xorg and kernel logs?
I'll do this next time it happens, but there were no obvious errors.
Created attachment 126003 [details] xorg log
Created attachment 126004 [details] Kernel log, only drm messages In kernel log I've filtered messages that contain drm. Please let me know if you need any other parts of kernel log
Please let me know if you require any more information.
Can you add "drm.debug=0x1e log_buf_len=1M" in your boot parameters to collect more information ? Even filtering your kernel log, I don't see anything after 1.501655 from the boot, do you mean that the issue already occurred? It looks like you have also an external display connected via DP, does the issue is happening on both screen (ie laptop and external one?)
I will add parameters you requested and reboot at once. Could you please clarify where should I collect resulting log from once this happens again? Will that get dumped into dmesg as well. Yes, this happens on both physical screens: laptop's internal and DP. And now that you mention that: I also have 'virtual' screen shared via VNC to another laptop and I cannot immediately say if problem affects that as well - but I'll pay more attention next time! Also, interestingly enough it looks like some apps are not affected: one notable example of that is firefox - I can still see majority (I guess even all) of the letters on web pages, but other apps like terminal, emacs, etc become totally broken. As far as kernel log goes: no, problem occurred much later that than 1.501655 but there are no messages about that in kernel log. Nothing that would even remotely relate to video/screens as far as I can see.
Created attachment 126006 [details] Kernel log Uploading same kernel log with little bit less information filtered out - maybe this would be more useful.
(In reply to Nikolay from comment #7) > I will add parameters you requested and reboot at once. Could you please > clarify where should I collect resulting log from once this happens again? > Will that get dumped into dmesg as well. Yes you are right :) > Also, interestingly enough it looks like some apps are not affected: one > notable example of that is firefox - I can still see majority (I guess even > all) of the letters on web pages, but other apps like terminal, emacs, etc > become totally broken. > > As far as kernel log goes: no, problem occurred much later that than > 1.501655 but there are no messages about that in kernel log. Nothing that > would even remotely relate to video/screens as far as I can see. you are right, there is no error/ warn message (except on bt, for instance) on gfx side after resume happening ~8735 sec from boot. you may try also, separately (ie set, reboot and test if anything is changing): 1- edit /etc/environment and add COGL_ATLAS_DEFAULT_BLIT_MODE=framebuffer (see bug 88584) 2- try to change accel method from sna to uxa by editing your configuration file:/etc/X11/xorg.conf.d/20-intel.conf and set Option "AccelMethod" "uxa"
I've tried running with "drm.debug=0x1e log_buf_len=1M" but it seems to be producing insane amount of logs in /var/log/syslog. Since problem doesn't occur on every resume I'm not sure I'll be able to run in this configuration long enough before my disk is full. Maybe there's some what to reduce amount of logging but still see relevant things?
I've tried 'COGL_ATLAS_DEFAULT_BLIT_MODE=framebuffer` and problem still persists
I've also tried uxa acceleration method. Unfortunately it is not really usable in my setup: once screens when to power saving mode they never woke up after that. No errors in any logs. Also uxa doesn't seem to support 'virtual' screens so I cannot extend my desctop to third 'virtual' screen.
Chris, any hints you can point here?
Hi. I've noticed one thing: sometimes switching to text console and back shows similar problem. I've see the problem after following actions: 1) Cold restart. 2) Login and start applications (notably FF with compositor enabled), wait for many tabs to load. 3) Switch to text console - at this point screens go black - there is no text console prompt, at this point text console is broken. 4) Switch back to X and observe broken fonts. Unfortunately these steps do not always exhibit the problem. Also I've see text console going bad, but fonts in X not breaking. So I'm not sure if that is related. But somehow I thought that following is possible: switching to text console allocates memory somewhere and somehow fails/allocates it in wrong way. And this breaks fonts (sometimes). Now: wake up is only related because maybe it switches for text console for a brief moment during awakening. Is 'text console going black' some sort of a known problem? Can it be related to this one? Should I file a separate ticket? Thanks!
(In reply to Nikolay from comment #0) > I have laptop with i7-3520M, Ubuntu 16.04 and latest oibaf mesa/drivers. I.e. gallium drivers? Please reproduce with Intel drivers.
> I.e. gallium drivers? Please reproduce with Intel drivers. Could you please clarify that this means? Just in case my /etc/X11/xorg.conf.d/20-intel.conf config: Section "Device" Identifier "Intel Graphics" Driver "intel" Option "AccelMethod" "sna" Option "TearFree" "true" Option "DRI" "3" EndSection does it still use gallium?
*shrug* I was looking at [1] which says "i915 gallium driver replaces by default i915 classic driver" [1] https://launchpad.net/~oibaf/+archive/ubuntu/graphics-drivers
Is there any way to check if it is running with gallium driver or not? I've verified xorg logs and glxinfo - none mention gallium
Nikolay - I'm sorry about this long delay until getting back to you. Is problem still valid with the latest components (kernel + userspace) ? If yes then please set status to REOPENED and attach latest logs, otherwise mark bug as RESOLVED.
Unfortunately I do not have that HW anymore, so I'm not able to immediately test this. I'll reopen if I see this problem again. Thanks!
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.