[Qemu-devel] crash using qemu-aarch64-softmmu

[Qemu-devel] crash using qemu-aarch64-softmmu

[Pranith Kumar]

[-- Attachment #1: Type: text/plain, Size: 3249 bytes --]

Hi,

I occasionally get the following crash while running an AArch64 softmmu on
an x86-64 system. I am using version 2.2 and cannot update to the latest
version. Did anyone else see this happening? If this is fixed, I would love
to get the patch backported.

Thanks!

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffc3b94700 (LWP 4409)]
0x00007ffff54b75b9 in tb_phys_invalidate (tb=0x7fffc4679010,
page_addr=18446744073709551615) at
/mnt/devops/code/qsim/qemu/translate-all.c:965
965             tb2 = tb1->jmp_next[n1];
(gdb) bt
#0  0x00007ffff54b75b9 in tb_phys_invalidate (tb=0x7fffc4679010,
page_addr=18446744073709551615) at
/mnt/devops/code/qsim/qemu/translate-all.c:965
#1  0x00007ffff54b7c0e in tb_invalidate_phys_page_range (start=946623904,
end=946623905, is_cpu_write_access=1) at
/mnt/devops/code/qsim/qemu/translate-all.c:1178
#2  0x00007ffff54b7d82 in tb_invalidate_phys_page_fast (start=946623904,
len=1) at /mnt/devops/code/qsim/qemu/translate-all.c:1236
#3  0x00007ffff54b301e in notdirty_mem_write (opaque=0x0,
ram_addr=946623904, val=0, size=1) at /mnt/devops/code/qsim/qemu/exec.c:1647
#4  0x00007ffff5500872 in memory_region_write_accessor (mr=0x7ffff5e93180
<io_mem_notdirty>, addr=946623904, value=0x7fffc3b93688, size=1, shift=0,
mask=255) at /mnt/devops/code/qsim/qemu/memory.c:443
#5  0x00007ffff55009ba in access_with_adjusted_size (addr=946623904,
value=0x7fffc3b93688, size=1, access_size_min=1, access_size_max=4,
access=0x7ffff55007da <memory_region_write_accessor>,
    mr=0x7ffff5e93180 <io_mem_notdirty>) at
/mnt/devops/code/qsim/qemu/memory.c:480
#6  0x00007ffff55039ec in memory_region_dispatch_write (mr=0x7ffff5e93180
<io_mem_notdirty>, addr=946623904, data=0, size=1) at
/mnt/devops/code/qsim/qemu/memory.c:1117
#7  0x00007ffff55073fc in io_mem_write (mr=0x7ffff5e93180
<io_mem_notdirty>, addr=946623904, val=0, size=1) at
/mnt/devops/code/qsim/qemu/memory.c:1973
#8  0x00007ffff550dfd5 in io_writeb (env=0x707048, physaddr=946623904,
val=0 '\000', addr=18446743799778268576, retaddr=140736744817534) at
/mnt/devops/code/qsim/qemu/softmmu_template.h:381
#9  0x00007ffff550e319 in helper_ret_stb_mmu (env=0x707048,
addr=18446743799778268576, val=0 '\000', mmu_idx=1,
retaddr=140736744817534) at
/mnt/devops/code/qsim/qemu/softmmu_template.h:420
#10 0x00007ffff5614550 in helper_dc_zva (env=0x707048,
vaddr_in=18446743799778268544) at
/mnt/devops/code/qsim/qemu/target-arm/helper.c:5086
#11 0x00007fffd3ae8380 in code_gen_buffer ()
#12 0x00007ffff54b9c5c in cpu_tb_exec (cpu=0x6fee00, tb_ptr=0x7fffd3ae81f0
<code_gen_buffer+4649456> "A\213n\374\205\355\017\205*\001") at
/mnt/devops/code/qsim/qemu/cpu-exec.c:171
#13 0x00007ffff54ba6bc in cpu_arm_exec (env=0x707048) at
/mnt/devops/code/qsim/qemu/cpu-exec.c:482
#14 0x00007ffff54e9e66 in tcg_cpu_exec (env=0x707048) at
/mnt/devops/code/qsim/qemu/cpus.c:1354
#15 0x00007ffff54e9f7d in tcg_exec_all () at
/mnt/devops/code/qsim/qemu/cpus.c:1387
#16 0x00007ffff54e926f in qemu_tcg_cpu_thread_fn (arg=0x6fee00) at
/mnt/devops/code/qsim/qemu/cpus.c:1033
#17 0x00007ffff69ee182 in start_thread (arg=0x7fffc3b94700) at
pthread_create.c:312
#18 0x00007ffff671b47d in clone () at
../sysdeps/unix/sysv/linux/x86_64/clone.S:111


-- 
Pranith

[-- Attachment #2: Type: text/html, Size: 3511 bytes --]

Re: [Qemu-devel] crash using qemu-aarch64-softmmu

[Peter Maydell]

On 15 July 2015 at 20:17, Pranith Kumar <bobby.prani@gmail.com> wrote:
> Hi,
>
> I occasionally get the following crash while running an AArch64 softmmu on
> an x86-64 system. I am using version 2.2 and cannot update to the latest
> version. Did anyone else see this happening? If this is fixed, I would love
> to get the patch backported.
>
> Thanks!
>
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7fffc3b94700 (LWP 4409)]
> 0x00007ffff54b75b9 in tb_phys_invalidate (tb=0x7fffc4679010,
> page_addr=18446744073709551615) at
> /mnt/devops/code/qsim/qemu/translate-all.c:965
> 965             tb2 = tb1->jmp_next[n1];

Googling suggests "qsim" is "a project which aims, as part of the
Manifold simulation effort at Georgia Tech, to create a thread safe
multicore emulation library based on the QEMU emulator".

My immediate guess is that this is buggy and you're causing
QEMU to corrupt some of its data structures by updating
and/or using them from multiple threads at once.

Can you reproduce the problem with a stock upstream QEMU?

thanks
-- PMM

Re: [Qemu-devel] crash using qemu-aarch64-softmmu

[Pranith Kumar]

[-- Attachment #1: Type: text/plain, Size: 693 bytes --]

On Wed, Jul 15, 2015 at 4:28 PM, Peter Maydell <peter.maydell@linaro.org>
wrote:

> Googling suggests "qsim" is "a project which aims, as part of the
> Manifold simulation effort at Georgia Tech, to create a thread safe
> multicore emulation library based on the QEMU emulator".
>
> My immediate guess is that this is buggy and you're causing
> QEMU to corrupt some of its data structures by updating
> and/or using them from multiple threads at once.
>
> Can you reproduce the problem with a stock upstream QEMU?
>
>
The qsim changes AFAIU are not causing this. But I understand the
apprehension that might be. I will try to reproduce this with an upstream
virgin QEMU.

Thanks!

-- 
Pranith

[-- Attachment #2: Type: text/html, Size: 1127 bytes --]