| Summary: | kwin crash while disabling lvds1 | ||
|---|---|---|---|
| Product: | [Plasma] kwin | Reporter: | George <georgecrum> |
| Component: | general | Assignee: | KWin default assignee <kwin-bugs-null> |
| Status: | RESOLVED UPSTREAM | ||
| Severity: | crash | CC: | ahepas1999, anjos.nuno, bastianholst, bernie, bitzzylvn, cebacam, christofk, css, esteve.farres, fermatslittletheorem, francisco.ospina, giluxxx, headcrabextra, helio.loureiro, hpfrantzy5, jc, jonathon.l.walker, jvnetsl, kai, kde, kde, kdebugs, kdebugs, ken_speakman, kollix, martin.holec, michael.zippe, northsoft, null, orchegion, oskargibson, peter, ryanskingsbury+kde, sethbass83, stakanov.s, stefan.walser, stephan.diestelhorst, talldalry, timsmeets, xgarve, zwaen |
| Priority: | NOR | Flags: | thomas.luebking:
Intel+
|
| Version: | 4.8.2 | ||
| Target Milestone: | --- | ||
| Platform: | Ubuntu | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
| Attachments: |
4.10 doesn't solve opengl problems
New crash information added by DrKonqi |
||
|
Description
George
2012-05-03 21:57:37 UTC
crashes in the driver. Which GPU and driver are you using? kinfocenter reports: 3D Accelerator Vendor: Intel Device: Core Processor Intergrated Graphic Controller Driver Vendor: Tungsten Graphics, Inc Renderer: Mesa DRI Intel Ironlake Mobile OpenGL/ES Version: 2.1 Mesa 8.0.2 Kernel module: i915 From lspci: 00:02.0 VGA compatible controller: Intel Corporation Core Processor Integrated Graphics Controller (rev 02) From lsmod: video 19596 1 i915 I think it has ultimately even the same source as bug #252817 (there're many results for "gg:brw_update_renderbuffer_surface segfault" and esp. ubuntu 12) With a little luck, 4.9 implicitly fixes this, because it doesn't restart the compositor on screensize changes. @George Before you detach the screen, simply suspend compositing (Shift+Alt+F12) and resume it afterwards (same combo) - you could also try with "kwin_gles --replace" "Resolved upstream" does not mean it's fixed anywhere but that it's a bug in a dependency we cannot fix ourselves. Hi Thomas, Thanks for the tips. However Shift+Alt+F12 before and after screen detachment doesn't seem to help. The command kwin_gles --replace works until I Ctrl+C then some windows do not have trim and can't be moved (even when using Alt+mouse left click hold). Kickoff application launcher menu is visible but does not work. Chrome browser still has window trim and can be moved around. Understood about upstream dependencies. I'll avoid detaching screens in the meantime. Thanks all for you attention to this. -George (In reply to comment #4) > Hi Thomas, > However Shift+Alt+F12 before and after screen detachment doesn't seem to help. You *cannot* get the same crash w/o a running (GL) compositor, what precisely happens when removing the screen while the compositor is suspended? > The command kwin_gles --replace works until I Ctrl+C errr - yes. "kwin_gles" is the window manager compiled for GLES (OpenGL subset) - if you quit it, you've no window manager anymore ;-) append and ampersand to run stuff from konsole forked to background "kwin_gles --replace &" - but you can remove the screen w/o getting a crash while running kwin_gles? *** Bug 300886 has been marked as a duplicate of this bug. *** *** Bug 301131 has been marked as a duplicate of this bug. *** *** Bug 301403 has been marked as a duplicate of this bug. *** *** Bug 301529 has been marked as a duplicate of this bug. *** *** Bug 302169 has been marked as a duplicate of this bug. *** *** Bug 303041 has been marked as a duplicate of this bug. *** (In reply to comment #1) > crashes right inside the driver. Not much we can do about it. Make sure you > have "unredirection of fullscreen windows" disabled. > > *** This bug has been marked as a duplicate of bug 299333 *** how I disable or enable "unredirection of fullscreen windows" ? *** Bug 304241 has been marked as a duplicate of this bug. *** btw I have one Intel video card xorg-x11-drv-intel-2.19.0-5.fc17.x86_64 Everyone has, for this is an intel bug about deactivating unredirection: run "kcmshell4 kwincompositing" and uncheck it (advanced tab) (In reply to comment #15) > about deactivating unredirection: > run "kcmshell4 kwincompositing" and uncheck it (advanced tab) on "Configure desktop effects" on advance tab, I just see . Composite type : Xrender or OpenGL and two options , keep thumbs and scale method one checkbox which says "suspend desktop effects on full windows" and two opengl options are disable . none "unredirection" option (In reply to comment #16) > one checkbox which says "suspend desktop effects on full windows" that's the one to disable. The technical feature is called "unredirection" of full-screen windows. *** Bug 306867 has been marked as a duplicate of this bug. *** *** Bug 306927 has been marked as a duplicate of this bug. *** *** Bug 307027 has been marked as a duplicate of this bug. *** *** Bug 309569 has been marked as a duplicate of this bug. *** *** Bug 310417 has been marked as a duplicate of this bug. *** *** Bug 311671 has been marked as a duplicate of this bug. *** *** Bug 311682 has been marked as a duplicate of this bug. *** https://bugs.freedesktop.org/show_bug.cgi?id=58780 <- corresponding bug report for the intel driver *** Bug 312335 has been marked as a duplicate of this bug. *** *** Bug 312336 has been marked as a duplicate of this bug. *** *** Bug 312468 has been marked as a duplicate of this bug. *** Git commit 276dfc615030465f01358a11ac98b11075e5ddec by Thomas Lübking. Committed on 29/12/2012 at 15:36. Pushed by luebking into branch 'KDE/4.10'. skip reset states if client untabbed for release FIXED-IN: 4.10 REVIEW: 108003 M +3 -1 kwin/client.cpp M +1 -1 kwin/client.h M +1 -1 kwin/workspace.cpp http://commits.kde.org/kde-workspace/276dfc615030465f01358a11ac98b11075e5ddec Wrong BUG in commit comment. *** Bug 312686 has been marked as a duplicate of this bug. *** *** Bug 312797 has been marked as a duplicate of this bug. *** *** Bug 313059 has been marked as a duplicate of this bug. *** *** Bug 313416 has been marked as a duplicate of this bug. *** *** Bug 314018 has been marked as a duplicate of this bug. *** *** Bug 314196 has been marked as a duplicate of this bug. *** *** Bug 314673 has been marked as a duplicate of this bug. *** *** Bug 314730 has been marked as a duplicate of this bug. *** *** Bug 316327 has been marked as a duplicate of this bug. *** Looking back at OP: > 3. Disabling LVDS1 and applying causes [...] both monitor going into power saving mode. > 4. Need to use FN+F7 combination to get screens back. And the other bugs around xrandr i fear the bug isn't even in the intel driver but we get some memory corruption from calling XRRUpdateConfiguration() See esp. bug #293209 *** Bug 316866 has been marked as a duplicate of this bug. *** *** Bug 317130 has been marked as a duplicate of this bug. *** *** Bug 317160 has been marked as a duplicate of this bug. *** Created attachment 78284 [details]
4.10 doesn't solve opengl problems
I have updated to KDE 4.10
opengl still doesn't work
Activating opengl nativ seems to be accepted. But after activating "Animation Arbeitsflächen-Würfel" it shows that this cannot be activated because of missing opengl.
Firefox and Thunderbird still look very ugly.
Sorry, the commit that set those vars didn't belong here. This bug is in the driver - updating KDE won't fix it ever, sorry. Created attachment 78338 [details]
New crash information added by DrKonqi
kwin (4.9.5) on KDE Platform 4.9.5 using Qt 4.8.4
- What I was doing when the application crashed:
This crash happens almost everytime a leave a fullscreen opengl game. Maybe related: The option to disable compositing during fullscreen mode is enabled.
-- Backtrace (Reduced):
#6 brw_update_renderbuffer_surface (brw=0x156df40, rb=0x155e280, unit=0) at brw_wm_surface_state.c:1109
#7 0x00007f0d8269a3c0 in brw_update_renderbuffer_surfaces (brw=0x156df40) at brw_wm_surface_state.c:1205
#8 0x00007f0d826835c2 in brw_upload_state (brw=0x156df40) at brw_state_upload.c:498
#9 0x00007f0d826705c7 in brw_try_draw_prims (max_index=<optimized out>, min_index=<optimized out>, ib=0x7fff802da758, nr_prims=22524616, prim=0x7fff802da740, arrays=<optimized out>, ctx=0x156df40) at brw_draw.c:493
#10 brw_draw_prims (ctx=0x156df40, prim=0x7fff802da740, nr_prims=22524616, ib=0x7fff802da758, index_bounds_valid=<optimized out>, min_index=0, max_index=35, tfb_vertcount=0x0) at brw_draw.c:589
do the "fullscreen" games run a different resolution or do you have "suspend compositing for fullscreen windows" checked in "kcmshell4 kwincompositing", 3rd tab? I also get a lot of crashes when I come out of fullscreen from watching videos in smplayer. There is no resolution change there, but it doesn't happen all the time. @Thomas They run same resolution but I have indeed the suspend compositing setting set... I think we're talking about the same thing. I have "Suspend desktop effects for fullscreen windows" set in System Settings > Desktop Effects > Advanced. disable that, your facing bug #252817 @Thomas. Thanks. I seem to recall having the same problem even with this disabled, but I'll let you know if I hit it again. *** Bug 317373 has been marked as a duplicate of this bug. *** After disabling "unredirection" (what it really is, instead of what the words state in the UI), the crash now longer occurs. After reading Martin's blog post http://blog.martin-graesslin.com/blog/2011/04/turning-compositing-off-in-the-right-way/ I understand that unredirection of windows is only properly supported by drivers supporting two OpenGL contexts at the same time - what most drivers don't. I've instead set a kwin rule to block compositing while Steam is running (actually that is what I run most games with) and it works fine since then (and performance is better, too). I've tried Steam on a Windows PC lately, using a two-monitor setup. And what did I discover?! Even Windows disables compositing (Aero) as soon as a game goes fullscreen. And a nice bonus is that the annoying flickering of the screen is gone now when watching fullscreen videos or playing games in fullscreen mode while some notification pops up (most annoying is Skype, I wished it had some switch to not do notifications during fullscreen sessions). BTW: Why doesn't kwin detect fullscreen applications and blocks compositing? I've read the blog post but didn't actually get the point. If there's a WM hint since months/years why isn't it used by current games, or why isn't it detected if usedm why don't native KDE applications use it (e.g. the video player while watching fullscreen)? Not sure how to differentiate it. I wonder if there is some "magic" way to build a kwin script which detects fullscreen games and blocks compositing. I would really enjoy and favor it. (In reply to comment #54) > supported by drivers supporting two OpenGL contexts at the same time - what > most drivers don't. Actually they (should) do but are buggy on it, afair it's however more a MESA than a particular driver issue. > BTW: Why doesn't kwin detect fullscreen applications and blocks compositing? Matter of focus - it would be wrong for a fullscreened browser etc. (ie. when there will be popups and dialogs etc.) > If there's a WM hint since months/years why isn't it used by current games, or why isn't it > detected if usedm why don't native KDE applications use it You'll have to ask app devs about that - most likely because it did not become NETWM canonical (iirc Sam was interested in it for compiz, but he's dropped compiz development - not sure what Canonical wants to do now) It's however (for KDE apps) as simple as calling KWindowSystem::setBlockingCompositing(winId(), true); so just tell everybody when you're interested. > wonder if there is some "magic" way to build a kwin script which detects > fullscreen games and blocks compositing. I would really enjoy and favor it. Automatic "game" detection isn't possible, but script is here: http://kde-look.org/content/show.php/GameMode?content=156659 *** Bug 317761 has been marked as a duplicate of this bug. *** *** Bug 317796 has been marked as a duplicate of this bug. *** FTR, bug #317796 has a slight deeper backtrace: #6 gen7_update_renderbuffer_surface (brw=0x8c2a5d8, rb=0x8bd9050, unit=0) at gen7_wm_surface_state.c:471 #7 0xaeb8ac03 in brw_update_renderbuffer_surfaces (brw=0x8c2a5d8) at brw_wm_surface_state.c:1205 #8 0xaeb72795 in brw_upload_state (brw=brw@entry=0x8c2a5d8) at brw_state_upload.c:498 #9 0xaeb5eb47 in brw_try_draw_prims (max_index=11, min_index=0, ib=0x0, nr_prims=1, prim=0xbfc81e20, arrays=0x8c2a5d8, ctx=0x8c2a5d8) at brw_draw.c:493 #10 brw_draw_prims (ctx=0x8c2a5d8, prim=0xbfc81e20, nr_prims=1, ib=0x0, index_bounds_valid=1 '\001', min_index=0, max_index=11, tfb_vertcount=0x0) at brw_draw.c:589 #11 0xae5067b2 in vbo_draw_arrays (ctx=ctx@entry=0x8c2a5d8, mode=mode@entry=4, start=start@entry=0, count=count@entry=12, numInstances=numInstances@entry=1, baseInstance=baseInstance@entry=0) at ../../../src/mesa/vbo/vbo_exec_array.c:645 #12 0xae5068a4 in vbo_exec_DrawArrays (mode=4, start=0, count=12) at ../../../src/mesa/vbo/vbo_exec_array.c:675 #13 0xb64cede6 in KWin::GLVertexBufferPrivate::corePainting (this=0x8e14088, region=..., primitiveMode=4, hardwareClipping=false) at /usr/src/debug/kde-workspace-4.10.0/kwin/libkwineffects/kwinglutils.cpp:1254 #14 0xb64cf0b5 in KWin::GLVertexBuffer::render (this=0x8e13528, region=..., primitiveMode=4, hardwareClipping=false) at /usr/src/debug/kde-workspace-4.10.0/kwin/libkwineffects/kwinglutils.cpp:1363 *** Bug 319007 has been marked as a duplicate of this bug. *** *** Bug 319758 has been marked as a duplicate of this bug. *** <dev note>I recently started to see problems like this crash on my notebook. They are triggered by a hung GPU (only solution to the problem: reboot). The reason for the hung GPU is an overflowing X event queue and dropped events. The reasons for the too many events is the synaptics (!) driver. So in future we might check in the duplicates whether the user has a touchpad and whether /var/log/Xorg.0.log contains information pointing to the problem.</dev note> *** Bug 321352 has been marked as a duplicate of this bug. *** *** Bug 322168 has been marked as a duplicate of this bug. *** *** Bug 322487 has been marked as a duplicate of this bug. *** *** Bug 322959 has been marked as a duplicate of this bug. *** My duplicate report asked for qdbus supportInformation. Can be found here: http://paste.kde.org/p6a2655d8/ *** Bug 323133 has been marked as a duplicate of this bug. *** *** Bug 323303 has been marked as a duplicate of this bug. *** *** Bug 323424 has been marked as a duplicate of this bug. *** *** Bug 325776 has been marked as a duplicate of this bug. *** *** Bug 331031 has been marked as a duplicate of this bug. *** *** Bug 331189 has been marked as a duplicate of this bug. *** *** Bug 331837 has been marked as a duplicate of this bug. *** *** Bug 331881 has been marked as a duplicate of this bug. *** We've now driver crashes through primusrun in this as well. Can anybody entountering this please attach the output of "xprop" and "xwininfo" on such window (maybe there's a way to detect them automatically) *** Bug 341556 has been marked as a duplicate of this bug. *** *** Bug 348871 has been marked as a duplicate of this bug. *** |