1813500 - glxtest process can hang which blocks whole Firefox start

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Martin Stránský [:stransky] (ni? me)]

Downstream bug: https://bugzilla.redhat.com/show_bug.cgi?id=2147344
glxtest bactrace https://retrace.fedoraproject.org/faf/reports/608452/

glxtest is blocked at:

0 pthread_cond_signal@@GLIBC_2.3.2
1 cnd_signal	/usr/lib64/dri/radeonsi_dri.so
2 util_queue_add_job	/usr/lib64/dri/radeonsi_dri.so
3 si_schedule_initial_compile	/usr/lib64/dri/radeonsi_dri.so
4 si_create_shader_selector(pipe_context*, pipe_shader_state const*)/usr/lib64/dri/radeonsi_dri.so
5 util_live_shader_cache_get	/usr/lib64/dri/radeonsi_dri.so
6 si_create_shader(pipe_context*, pipe_shader_state const*) /usr/lib64/dri/radeonsi_dri.so
7 ureg_create_shader /usr/lib64/dri/r600_drv_video.so
8 create_vert_shader	/usr/lib64/dri/r600_drv_video.so
9 vl_compositor_init	/usr/lib64/dri/r600_drv_video.so
10 __vaDriverInit_1_16	/usr/lib64/dri/r600_drv_video.so
11 va_openDriver	/lib64/libva.so.2
12 vaInitialize 

Comment 1

[Martin Stránský [:stransky] (ni? me)]

Firefox is blocked here:

#0  __GI___libc_read (nbytes=2047, buf=0x7ffc4e276480, fd=3) at ../sysdeps/unix/sysv/linux/read.c:26
#1  __GI___libc_read (fd=3, buf=buf@entry=0x7ffc4e276480, nbytes=nbytes@entry=2047) at ../sysdeps/unix/sysv/linux/read.c:24
#2  0x00007fcd423fe512 in read (__nbytes=2047, __buf=0x7ffc4e276480, __fd=<optimized out>) at /home/komat/.mozbuild/sysroot-x86_64-linux-gnu/usr/include/x86_64-linux-gnu/bits/unistd.h:44
#3  mozilla::widget::GfxInfo::GetData() (this=0x7fcd4a8abac0) at /raid/src3/widget/gtk/GfxInfo.cpp:106
#4  0x00007fcd42400bb8 in mozilla::widget::GfxInfo::GetFeatureStatusImpl(int, int*, nsTSubstring<char16_t>&, nsTArray<mozilla::widget::GfxDriverInfo> const&, nsTSubstring<char>&, mozilla::widget::OperatingSystem*)
    (this=0x7fcd4a8abac0, aFeature=19, aStatus=0x7ffc4e276f64, aSuggestedDriverVersion=<gNullChar> u"", aDriverInfo=const nsTArray<mozilla::widget::GfxDriverInfo> &, aFailureId=<gNullChar> "", aOS=0x0) at /raid/src3/widget/gtk/GfxInfo.cpp:875
#5  0x00007fcd4235cdb6 in mozilla::widget::GfxInfoBase::GetFeatureStatus(int, nsTSubstring<char>&, int*)
    (this=0x7fcd4a8abac0, aFeature=19, aFailureId=<gNullChar> "", aStatus=0x7ffc4e276f64) at /raid/src3/widget/GfxInfoBase.cpp:841
#6  0x00007fcd40144e4c in IsFeatureSupported(long, bool) (aFeature=aFeature@entry=19, aDefault=aDefault@entry=false) at /raid/src3/objdir-opt-gcc/dist/include/nsCOMPtr.h:851
#7  0x00007fcd40150094 in gfxPlatform::IsDXInterop2Blocked() () at /raid/src3/gfx/thebes/gfxPlatform.cpp:1129
#8  gfxPlatform::Init() () at /raid/src3/gfx/thebes/gfxPlatform.cpp:899
#9  0x00007fcd40150816 in gfxPlatform::GetPlatform() () at /raid/src3/gfx/thebes/gfxPlatform.cpp:462
#10 0x00007fcd423dbb0b in nsWindow::Create(nsIWidget*, void*, mozilla::gfx::IntRectTyped<mozilla::LayoutDevicePixel> const&, mozilla::widget::InitData*)
    (this=0x7fcd4a873800, aParent=<optimized out>, aNativeParent=<optimized out>, aRect=<optimized out>, aInitData=0x7ffc4e2773f0) at /raid/src3/widget/gtk/nsWindow.cpp:5883
#11 0x00007fcd43502d46 in nsIWidget::Create(nsIWidget*, void*, mozilla::gfx::IntRectTyped<mozilla::DesktopPixel> const&, mozilla::widget::InitData*)
    (aInitData=0x7ffc4e2773f0, aRect=..., aNativeParent=0x0, aParent=0x0, this=0x7fcd4a873800) at /raid/src3/objdir-opt-gcc/dist/include/nsIWidget.h:463
#12 mozilla::AppWindow::Initialize(nsIAppWindow*, nsIAppWindow*, int, int, bool, mozilla::widget::InitData&)
    (this=this@entry=0x7fcd2c7f4600, aParent=<optimized out>, aOpener=aOpener@entry=0x0, aInitialWidth=aInitialWidth@entry=1, aInitialHeight=aInitialHeight@entry=1, aIsHiddenWindow=aIsHiddenWindow@entry=false, widgetInitData=...) at /raid/src3/xpfe/appshell/AppWindow.cpp:212

Comment 2

[Robert Mader [:rmader]]

This looks like a Mesa issue to me - is there an issue on their bugtracker already?

Comment 3

[Martin Stránský [:stransky] (ni? me)]

(In reply to Robert Mader [:rmader] from comment #2)

This looks like a Mesa issue to me - is there an issue on their bugtracker already?

I don't think so.

Comment 4

[Martin Stránský [:stransky] (ni? me)]

Attachment: Bug 1813500 [Linux] Don't hang Firefox when glxtest process hangs r?emilio,rmader

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/c37e07978bca
[Linux] Don't hang Firefox when glxtest process hangs r=emilio

Comment 6

[Marian-Vasile Laza]

Backed out changeset c37e07978bca (Bug 1813500) for causing bustages on GfxInfo.cpp.
Backout link
Push with failures <--> Bp-hybrid
Failure Log

Comment 7

[Martin Stránský [:stransky] (ni? me)]

Ohh, missing headers. Should be fixed now, thanks.

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/371e407a353d
[Linux] Don't hang Firefox when glxtest process hangs r=emilio

Comment 9

[Atila Butkovits]

https://hg.mozilla.org/mozilla-central/rev/371e407a353d