Re: [RFC PATCH] linux-user: un-parent OBJECT(cpu) when closing thread

[Laurent Vivier <laurent@vivier.eu>]

On 21/06/2022 20:17, Laurent Vivier wrote:
> Le 10/06/2022 à 16:38, Alex Bennée a écrit :
>> While forcing the CPU to unrealize by hand does trigger the clean-up
>> code we never fully free resources because refcount never reaches
>> zero. This is because QOM automatically added objects without an
>> explicit parent to /unattached/, incrementing the refcount.
>>
>> Instead of manually triggering unrealization just unparent the object
>> and let the device machinery deal with that for us.
>>
>> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/866
>> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>> ---
>>   linux-user/syscall.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/linux-user/syscall.c b/linux-user/syscall.c
>> index f55cdebee5..c413d32311 100644
>> --- a/linux-user/syscall.c
>> +++ b/linux-user/syscall.c
>> @@ -8566,7 +8566,7 @@ static abi_long do_syscall1(CPUArchState *cpu_env, int num, 
>> abi_long arg1,
>>           if (CPU_NEXT(first_cpu)) {
>>               TaskState *ts = cpu->opaque;
>> -            object_property_set_bool(OBJECT(cpu), "realized", false, NULL);
>> +            object_unparent(OBJECT(cpu));
>>               object_unref(OBJECT(cpu));
>>               /*
>>                * At this point the CPU should be unrealized and removed
> 
> Applied to my linux-user-for-7.1 branch.
> 

I remove it from the branch.

It breaks something with arm64 (only) target (any distros: stretch, sid, trusty, bionic), 
the following LTP test cases fail:

bind04, bind05, fcntl34, fcntl34_64, fcntl36, fcntl36_64, inotify09, mmap18, 
sched_setattr01, sched_getattr01, tgkill01, tgkill02, tgkill03, futex_wake02

For instance:

tst_test.c:1250: TINFO: Timeout per run is 0h 05m 00s
bind04.c:117: TINFO: Testing AF_UNIX pathname stream
bind04.c:150: TPASS: Communication successful
tst_test.c:1300: TBROK: Test killed by SIGSEGV!

Thanks,
Laurent