Emulator hangs with GPU emulation on

[Miodrag Dinic]

Hello to everyone,

Some recent changes to GPU emulation engine introduced regression which causes emulator to completely hang.

This issue can be observed with the latest emulator built from emu-master-dev sources and run with -gpu on.

It seems that it is some race condition in question because it can not be reproduced 100% of the time. 

Verified that it is happening for both ARM64 and MIPS/MIPS64 with the Ranchu (qemu2) backend, could not reproduce the hang with the Goldfish (classic) backend.

It is however easier to reproduce with MIPS/MIPS64 emulation using qemu2 backend.

To reproduce the issue with MIPS64 emulation, just run the emulator built from emu-master-dev sources :

cd <AOSP>

source build/envsetup.sh

lunch aosp_mips64-eng

cd <EMU_MASTER_DEV>/external/qemu/

./objs/emulator -show-kernel -gpu on

Almost 100% of the times it will hang during bootanimation.

To increase a chance to reproduce the issue with ARM64, run the emulation through GDB like this:

cd <AOSP>

source build/envsetup.sh

lunch aosp_arm64-eng

cd <EMU_MASTER_DEV>/external/qemu/

gdb -args ./objs/emulator -show-kernel -gpu on

...

Reading symbols from ./objs/emulator...(no debugging symbols found)...done.

(gdb) handle SIGUSR1 nostop

Signal        Stop Print Pass to program Description

SIGUSR1       No Yes Yes User defined signal 1

(gdb) set pagination off

(gdb) run

...

[Thread 0x7fff767fc700 (LWP 2913) exited]

[Thread 0x7fff757fa700 (LWP 2915) exited]

[Thread 0x7fff75ffb700 (LWP 2914) exited]

[Thread 0x7fffc4fff700 (LWP 2917) exited]

This is a point where the emulation freezes:

Program received signal SIGINT, Interrupt.

[Switching to Thread 0x7ffff1a26700 (LWP 2895)]

__lll_lock_wait () at ../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:135

135 ../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S: No such file or directory.

(gdb) bt

#0  __lll_lock_wait () at ../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:135

#1  0x00007ffff7bc6657 in _L_lock_909 () from /lib/x86_64-linux-gnu/libpthread.so.0

#2  0x00007ffff7bc6480 in __GI___pthread_mutex_lock (mutex=0x11b2e80) at ../nptl/pthread_mutex_lock.c:79

#3  0x00000000006c97b9 in ?? ()

#4  0x000000000043dc5b in ?? ()

#5  0x000000000067709b in ?? ()

#6  0x000000000048f117 in ?? ()

#7  0x00000000005b2b4a in ?? ()

#8  0x00007ffff644262f in ?? () from /home/mousewho/MIPS_SUPPORT/sources/Android/workspace-linux/studio-dev/external/qemu/objs/lib64/qt/lib/libQt5Core.so.5

#9  0x00007ffff7bc4182 in start_thread (arg=0x7ffff1a26700) at pthread_create.c:312

#10 0x00007ffff567e47d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

(gdb) info threads 

  Id   Target Id         Frame 

  17   Thread 0x7fff54ffd700 (LWP 2912) "MainLoopThread" pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185

  16   Thread 0x7fff567fe700 (LWP 2911) "MainLoopThread" pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185

  15   Thread 0x7fff56fff700 (LWP 2910) "MainLoopThread" pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185

  14   Thread 0x7fff74ff9700 (LWP 2909) "MainLoopThread" pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185

  6    Thread 0x7fffe89d3700 (LWP 2898) "MainLoopThread" 0x00000000004f7e68 in ?? ()

  4    Thread 0x7ffff1225700 (LWP 2896) "MainLoopThread" pthread_cond_wait@@GLIBC_2.3.2 () at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:185

* 3    Thread 0x7ffff1a26700 (LWP 2895) "MainLoopThread" __lll_lock_wait () at ../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:135

  2    Thread 0x7ffff227d700 (LWP 2893) "QXcbEventReader" 0x00007ffff567112d in poll () at ../sysdeps/unix/syscall-template.S:81

  1    Thread 0x7ffff7ee2980 (LWP 2886) "qemu-system-aar" 0x00007ffff5675da3 in select () at ../sysdeps/unix/syscall-template.S:81

(gdb)

We also identified, that the following changes are most likely to introduced the issue :

https://android-review.googlesource.com/#/q/topic:commonlib+status:merged

If you checkout the sources as follows, issue disappears :

external/qemu:

git checkout 1bd2d4e581f~1

external/qemu-android:

git checkout 288dfebe905~1

Is anybody aware of this issue and is looking into it? Should I submit a bug in  http://code.google.com/p/android/issues/entry?template=Android%20Emulator%20Bug ?

Kind regards,

Miodrag

[David Turner]

Thanks Miodrag,

This is a very recent issue, we were not aware of it but I've created an entry for it: https://code.google.com/p/android/issues/detail?id=209097

--
You received this message because you are subscribed to the Google Groups "Android emulator development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to android-emulator...@googlegroups.com.
To post to this group, send email to android-em...@googlegroups.com.
Visit this group at https://groups.google.com/group/android-emulator-dev.
For more options, visit https://groups.google.com/d/optout.