Re: [Qemu-devel] colo-compare: segfault and assert on colo_compare_finalize

[Hailiang Zhang <zhang.zhanghailiang@huawei.com>]

Hi,

Did you test this branch https://github.com/coloft/qemu/tree/colo-for-qemu-2.10-2017-4-22 ?

This seems to be an already known problem, I'm not quite sure, it may be fixed by this patch

b19456dd0ea4eb418ad093f092adbb882be13054
char: Fix removing wrong GSource that be found by fd_in_tag
We use fd_in_tag to find a GSource, fd_in_tag is return value of g_source_attach(GSource *source, GMainContext *context), the return value is unique only in the same context, so we may get the same values with different 'context' parameters. It is no problem to find the right fd_in_tag by using g_main_context_find_source_by_id(GMainContext *context, guint source_id) while there is only one default main context. But colo-compare tries to create/use its own context, and if we pass wrong 'context' parameter with right fd_in_tag, we will find a wrong GSource to handle. We tried to fix the related codes in commit b43decb015a6efeb9e3cdbdb80f6547ad7248a4c, but it didn't fix the bug completely, because we still have some codes didn't pass *right* context parameter for remove_fd_in_watch(). Let's fix it by record the GSource directly instead of fd_in_tag. Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com> Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com> 
Message-Id: <1492564532-91680-1-git-send-email-zhang.zhanghailiang@huawei.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Actually, we have already re-writed this part, and please follow the later series. Thanks, Hailiang

On 2017/8/8 0:39, Eduardo Otubo wrote:
> (please ignore my last email, looks like mutt wants play games lately)
>
> Hi all,
>
> I have found a problem on colo-compare that leads to segmentation fault
> when calling qemu like this:
>
>    $ qemu-system-x86_64 -S -machine pc -object colo-compare,id=test-object
>
> First I got an assert failed:
>
>    (qemu-system-x86_64:7887): GLib-CRITICAL **: g_main_loop_quit:
> assertion 'loop != NULL' failed
>
>   From this looks like s->compare_loop is NULL on the function
> colo_compare_finalize(), then I just added a check there and the assert
> went away. But then there's the segfault:
>
>    Thread 1 "qemu-system-x86" received signal SIGSEGV, Segmentation fault.
>    0x00007ffff333f79e in pthread_join () from /lib64/libpthread.so.0
>    (gdb) bt
>    #0  0x00007ffff333f79e in pthread_join () at /lib64/libpthread.so.0
>    #1  0x0000555555c379d2 in qemu_thread_join (thread=0x7ffff7ff5160) at
> util/qemu-thread-posix.c:547
>    #2  0x0000555555adfc1a in colo_compare_finalize (obj=0x7ffff7fd3010)
> at net/colo-compare.c:867
>    #3  0x0000555555b2cd87 in object_deinit (obj=0x7ffff7fd3010,
> type=0x5555567432e0) at qom/object.c:453
>    #4  0x0000555555b2cdf9 in object_finalize (data=0x7ffff7fd3010) at
> qom/object.c:467
>    #5  0x0000555555b2dd80 in object_unref (obj=0x7ffff7fd3010) at
> qom/object.c:902
>    #6  0x0000555555b319a5 in user_creatable_add_type (type=0x5555567499a0
> "colo-compare", id=0x555556749960 "test-object", qdict=0x555556835750,
> v=0x55555681a3f0, errp=0x7fffffffde58) at qom/object_interfaces.c:105
>    #7  0x0000555555b31b02 in user_creatable_add_opts
> (opts=0x555556749910, errp=0x7fffffffde58) at qom/object_interfaces.c:135
>    #8  0x0000555555b31bfd in user_creatable_add_opts_foreach
> (opaque=0x5555558e9c39 <object_create_delayed>, opts=0x555556749910,
> errp=0x0) at qom/object_interfaces.c:159
>    #9  0x0000555555c4aecf in qemu_opts_foreach (list=0x555556157ac0
> <qemu_object_opts>, func=0x555555b31b6f
> <user_creatable_add_opts_foreach>, opaque=0x5555558e9c39
> <object_create_delayed>, errp=0x0) at util/qemu-option.c:1104
>    #10 0x00005555558edb75 in main (argc=6, argv=0x7fffffffe2d8,
> envp=0x7fffffffe310) at vl.c:4520
>
> At this point '&s->thread' is '0'. Is this segfault and the above
> mentioned assert trigged because I'm creating a colo-compare object
> without any other parameter? In a positive case, a simple workaround and
> error check should do it. Otherwise I'll debug a little more.
>
> Best regards,