[Qemu-devel] [Bug 1811244] [NEW] qemu 3.1/i386 crashes when MTTCG is enabled

[Jakub Jermar]

Public bug reported:

When MTTCG is enabled, QEMU 3.1.0 sometimes crashes when running the
following command line:

qemu-system-i386 -kernel
/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/bootstrap
-append bootstrap -initrd
"/home/jermar/work/software/l4/fiasco/.build-i386/fiasco
-serial_esc,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/sigma0
,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/moe
rom/ahci.cfg,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/ned
,test_env.lua ,/home/jermar/Kernkonzept/software/l4/pkg/ahci-
driver/examples/md5sum/ahci.cfg
,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/l4re
,/home/jermar/Kernkonzept/software/l4/pkg/ahci-
driver/examples/md5sum/ahci.io
,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/io
,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/ahci-
drv ,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f
/ahci-md5-sync" -smp 4 -accel tcg,thread=multi -device ahci,id=ahci0
-drive if=none,file=/home/jermar/Kernkonzept/software/l4/.build-i386/pkg
/ahci-driver/test/examples/test_ahci.img,format=raw,id=drive-sata0-0-0
-device ide-drive,bus=ahci0.0,drive=drive-sata0-0-0,id=sata0-0-0 -serial
stdio -nographic -monitor none

The host is x86_64.

The stack at the time of the crash (core dump and debug binary linked
below[1]):

Core was generated by `qemu-system-i386 -kernel 
/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  io_writex (address@hidden, address@hidden, mmu_idx=2, address@hidden, 
address@hidden, address@hidden, recheck=false, size=4)
    at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cputlb.c:791
791         if (mr->global_locking && !qemu_mutex_iothread_locked()) {
[Current thread is 1 (Thread 0x7fc5af7fe700 (LWP 3625719))]
Missing separate debuginfos, use: dnf debuginfo-install 
SDL2-2.0.9-1.fc29.x86_64 at-spi2-atk-2.30.0-1.fc29.x86_64 
at-spi2-core-2.30.0-2.fc29.x86_64 atk-2.30.0-1.fc29.x86_64 
bzip2-libs-1.0.6-28.fc29.x86_64 cairo4
(gdb) bt
#0  0x0000565354f5f365 in io_writex
    (address@hidden, address@hidden, mmu_idx=2, address@hidden, address@hidden, 
address@hidden, recheck=false, size=4)
    at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cputlb.c:791
#1  0x0000565354f621b2 in io_writel (recheck=<optimized out>, 
retaddr=140487132809203, addr=3938451632, val=0, index=0, mmu_idx=2, 
env=0x565355ca0140)
    at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/softmmu_template.h:310
#2  0x0000565354f621b2 in helper_le_stl_mmu (env=0x565355ca0140, 
addr=<optimized out>, val=0, oi=34, retaddr=140487132809203)
    at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/softmmu_template.h:310
#3  0x00007fc5b5a587f3 in code_gen_buffer ()
#4  0x0000565354f75fd0 in cpu_tb_exec (itb=<optimized out>, cpu=0x7fc5b5a5aa40 
<code_gen_buffer+12266006>) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cpu-exec.c:171
#5  0x0000565354f75fd0 in cpu_loop_exec_tb (tb_exit=<synthetic pointer>, 
last_tb=<synthetic pointer>, tb=<optimized out>, cpu=0x7fc5b5a5aa40 
<code_gen_buffer+12266006>)
    at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cpu-exec.c:615
#6  0x0000565354f75fd0 in cpu_exec (address@hidden) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cpu-exec.c:725
#7  0x0000565354f33b1f in tcg_cpu_exec (cpu=0x565355c97e90) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/cpus.c:1429
#8  0x0000565354f35e83 in qemu_tcg_cpu_thread_fn (arg=0x565355c97e90) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/cpus.c:1733
#9  0x0000565354f35e83 in qemu_tcg_cpu_thread_fn (address@hidden) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/cpus.c:1707
#10 0x00005653552ec5da in qemu_thread_start (args=<optimized out>) at 
util/qemu-thread-posix.c:498
#11 0x00007fc5b858a58e in start_thread () at /lib64/libpthread.so.0
#12 0x00007fc5b84b96a3 in clone () at /lib64/libc.so.6

Another symptom that occurs more often than this crash is that the guest
hangs while waiting for another CPU to complete a cross-CPU call.
Disabling MTTCG makes both symptoms go away.

[1] Core file + debug binary: http://jermar.eu/ref/qemu-mttcg-
core.tar.xz

** Affects: qemu
     Importance: Undecided
         Status: New


** Tags: core i386 mttcg

-- 
You received this bug notification because you are a member of qemu-
devel-ml, which is subscribed to QEMU.
https://bugs.launchpad.net/bugs/1811244

Title:
  qemu 3.1/i386 crashes when MTTCG is enabled

Status in QEMU:
  New

Bug description:
  When MTTCG is enabled, QEMU 3.1.0 sometimes crashes when running the
  following command line:

  qemu-system-i386 -kernel
  /home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/bootstrap
  -append bootstrap -initrd
  "/home/jermar/work/software/l4/fiasco/.build-i386/fiasco
  
-serial_esc,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/sigma0
  ,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/moe
  
rom/ahci.cfg,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/ned
  ,test_env.lua ,/home/jermar/Kernkonzept/software/l4/pkg/ahci-
  driver/examples/md5sum/ahci.cfg
  ,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/l4re
  ,/home/jermar/Kernkonzept/software/l4/pkg/ahci-
  driver/examples/md5sum/ahci.io
  ,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f/io
  ,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f
  /ahci-drv
  ,/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x86_gen/l4f
  /ahci-md5-sync" -smp 4 -accel tcg,thread=multi -device ahci,id=ahci0
  -drive
  if=none,file=/home/jermar/Kernkonzept/software/l4/.build-i386/pkg
  /ahci-driver/test/examples/test_ahci.img,format=raw,id=drive-sata0-0-0
  -device ide-drive,bus=ahci0.0,drive=drive-sata0-0-0,id=sata0-0-0
  -serial stdio -nographic -monitor none

  The host is x86_64.

  The stack at the time of the crash (core dump and debug binary linked
  below[1]):

  Core was generated by `qemu-system-i386 -kernel 
/home/jermar/Kernkonzept/software/l4/.build-i386/bin/x'.
  Program terminated with signal SIGSEGV, Segmentation fault.
  #0  io_writex (address@hidden, address@hidden, mmu_idx=2, address@hidden, 
address@hidden, address@hidden, recheck=false, size=4)
      at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cputlb.c:791
  791       if (mr->global_locking && !qemu_mutex_iothread_locked()) {
  [Current thread is 1 (Thread 0x7fc5af7fe700 (LWP 3625719))]
  Missing separate debuginfos, use: dnf debuginfo-install 
SDL2-2.0.9-1.fc29.x86_64 at-spi2-atk-2.30.0-1.fc29.x86_64 
at-spi2-core-2.30.0-2.fc29.x86_64 atk-2.30.0-1.fc29.x86_64 
bzip2-libs-1.0.6-28.fc29.x86_64 cairo4
  (gdb) bt
  #0  0x0000565354f5f365 in io_writex
      (address@hidden, address@hidden, mmu_idx=2, address@hidden, 
address@hidden, address@hidden, recheck=false, size=4)
      at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cputlb.c:791
  #1  0x0000565354f621b2 in io_writel (recheck=<optimized out>, 
retaddr=140487132809203, addr=3938451632, val=0, index=0, mmu_idx=2, 
env=0x565355ca0140)
      at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/softmmu_template.h:310
  #2  0x0000565354f621b2 in helper_le_stl_mmu (env=0x565355ca0140, 
addr=<optimized out>, val=0, oi=34, retaddr=140487132809203)
      at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/softmmu_template.h:310
  #3  0x00007fc5b5a587f3 in code_gen_buffer ()
  #4  0x0000565354f75fd0 in cpu_tb_exec (itb=<optimized out>, 
cpu=0x7fc5b5a5aa40 <code_gen_buffer+12266006>) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cpu-exec.c:171
  #5  0x0000565354f75fd0 in cpu_loop_exec_tb (tb_exit=<synthetic pointer>, 
last_tb=<synthetic pointer>, tb=<optimized out>, cpu=0x7fc5b5a5aa40 
<code_gen_buffer+12266006>)
      at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cpu-exec.c:615
  #6  0x0000565354f75fd0 in cpu_exec (address@hidden) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/accel/tcg/cpu-exec.c:725
  #7  0x0000565354f33b1f in tcg_cpu_exec (cpu=0x565355c97e90) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/cpus.c:1429
  #8  0x0000565354f35e83 in qemu_tcg_cpu_thread_fn (arg=0x565355c97e90) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/cpus.c:1733
  #9  0x0000565354f35e83 in qemu_tcg_cpu_thread_fn (address@hidden) at 
/home/jermar/software/HelenOS/helenos.git/contrib/qemu/qemu-3.1.0/cpus.c:1707
  #10 0x00005653552ec5da in qemu_thread_start (args=<optimized out>) at 
util/qemu-thread-posix.c:498
  #11 0x00007fc5b858a58e in start_thread () at /lib64/libpthread.so.0
  #12 0x00007fc5b84b96a3 in clone () at /lib64/libc.so.6

  Another symptom that occurs more often than this crash is that the
  guest hangs while waiting for another CPU to complete a cross-CPU
  call. Disabling MTTCG makes both symptoms go away.

  [1] Core file + debug binary: http://jermar.eu/ref/qemu-mttcg-
  core.tar.xz

To manage notifications about this bug go to:
https://bugs.launchpad.net/qemu/+bug/1811244/+subscriptions