qemu 2.12.0 qemu-system-ppc illegal instruction on ppc64le, crashes emulator

[Cameron Kaiser]

% uname -a
Linux tim.floodgap.com 4.16.14-300.fc28.ppc64le #1 SMP Tue Jun 5 15:59:48 UTC 2018 ppc64le ppc64le ppc64le GNU/Linux

STR:
Start QEMU and boot Mac OS X 10.4.11.
Download the current version of TenFourFox (I used G3 so that AltiVec was not a confounder).
Try to start TenFourFox in safe mode (hold down Option as you double-click while the icon bounces in the Dock).

Expected:
TenFourFox starts.

Actual:
The entire emulator exits with an illegal instruction error.

Trace of session (including some disassembly so you can see where TCG went wrong):

tim:/home/spectre/src/qemu-2.12.0/ppc-softmmu/% gdb --args ./qemu-system-ppc -M mac99,accel=tcg -m 2048 -prom-env boot-args=-v -boot c -drive file=tigerhd.img,format=raw,cache=none -netdev user,id=mynet0 -device usb-net,netdev=mynet0 -usb -device usb-tablet

GNU gdb (GDB) Fedora 8.1-15.fc28
[...]
Reading symbols from ./qemu-system-ppc...done.
(gdb) run
[...]

Thread 6 "qemu-system-ppc" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff242ea30 (LWP 7017)]
0xfffffffffffffffc in ?? ()
#0 0xfffffffffffffffc in ()
#1 0x00007fffd4edec00 in code_gen_buffer ()
#2 0x00000000100c9e20 in cpu_tb_exec (itb=<optimized out>, cpu=<optimized out>) at /home/spectre/src/qemu-2.12.0/accel/tcg/cpu-exec.c:169
#3 0x00000000100c9e20 in cpu_loop_exec_tb (tb_exit=<synthetic pointer>, last_tb=<synthetic pointer>, tb=<optimized out>, cpu=<optimized out>)
    at /home/spectre/src/qemu-2.12.0/accel/tcg/cpu-exec.c:626
#4 0x00000000100c9e20 in cpu_exec (cpu=<optimized out>)
    at /home/spectre/src/qemu-2.12.0/accel/tcg/cpu-exec.c:734
#5 0x000000001007decc in tcg_cpu_exec (cpu=0x11774e10)
    at /home/spectre/src/qemu-2.12.0/cpus.c:1362
(gdb) disas 0x00007fffd4edebf0, 0x00007fffd4edec10
Dump of assembler code from 0x7fffd4edebf0 to 0x7fffd4edec10:
   0x00007fffd4edebf0 <code_gen_buffer+284027700>: addi r0,r4,3
   0x00007fffd4edebf4 <code_gen_buffer+284027704>: rlwinm r0,r0,0,0,19
   0x00007fffd4edebf8 <code_gen_buffer+284027708>: cmplw cr7,r0,r12
   0x00007fffd4edebfc <code_gen_buffer+284027712>: bnel cr7,0x7fffd4ed8b64 <code_gen_buffer+284002984>
   0x00007fffd4edec00 <code_gen_buffer+284027716>: lwbrx r14,r3,r4
   0x00007fffd4edec04 <code_gen_buffer+284027720>: stw r14,40(r27)
   0x00007fffd4edec08 <code_gen_buffer+284027724>: clrldi r4,r14,32
   0x00007fffd4edec0c <code_gen_buffer+284027728>: rlwinm r3,r4,25,19,26
End of assembler dump.
(gdb) disas 0x7fffd4ed8b60, 0x7fffd4ed8b70
Dump of assembler code from 0x7fffd4ed8b60 to 0x7fffd4ed8b70:
   0x00007fffd4ed8b60 <code_gen_buffer+284002980>: bctrl
   0x00007fffd4ed8b64 <code_gen_buffer+284002984>: mtctr r3
   0x00007fffd4ed8b68 <code_gen_buffer+284002988>: mr r31,r3
   0x00007fffd4ed8b6c <code_gen_buffer+284002992>: li r3,0
End of assembler dump.
(gdb) i reg ctr
ctr 0xffffffffffffffff 18446744073709551615

It appears that the branch at 0x00007fffd4edebfc caused a jump back (a return?) through CTR, but CTR has -1 in it, hence setting PC to 0xfffffffffffffffc. I am not sure how to debug this further.

[Cameron Kaiser (classichasclass)]

Sorry, more complete disassembly of the apparent actual fault:

   0x00007fffd4ed8b64 <code_gen_buffer+284002984>: mtctr r3
   0x00007fffd4ed8b68 <code_gen_buffer+284002988>: mr r31,r3
   0x00007fffd4ed8b6c <code_gen_buffer+284002992>: li r3,0
   0x00007fffd4ed8b70 <code_gen_buffer+284002996>: bctr

[Murilo Opsfelder Araújo (mopsfelder)]

Hi, Cameron.

The step "Start QEMU and boot Mac OS X 10.4.11" is not clear to me. Is there a location where one could download such image and boot?

I wonder how one without access to a Mac image can reproduce this issue.

Cheers
Murilo

[Thomas Huth (th-huth)]

The QEMU project is currently considering to move its bug tracking to another system. For this we need to know which bugs are still valid and which could be closed already. Thus we are setting older bugs to "Incomplete" now.
If you still think this bug report here is valid, then please switch the state back to "New" within the next 60 days, otherwise this report will be marked as "Expired". Or mark it as "Fix Released" if the problem has been solved with a newer version of QEMU already. Thank you and sorry for the inconvenience.

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]