Bug 1633536

Description of problem:
After hot plugging a backend image with 'blockdev-add', migrate the guest from src to dst with shared storage(I have tried nfs/rbd), it's found qemu core dump in src side when the migration nearly finishes:

# Core dump info in src side:
(qemu) qemu-kvm: block.c:855: bdrv_child_cb_inactivate: Assertion `bs->open_flags & 0x0800' failed.
./vm-mig1.sh: line 28: 30951 Aborted                 (core dumped) /usr/libexec/qemu-kvm

# Error info in dst side:
(qemu) qemu-kvm: Failed to load virtio_pci/modern_queue_state:avail
qemu-kvm: Failed to load virtio_pci/modern_state:vqs
qemu-kvm: Failed to load virtio/extra_state:extra_state
qemu-kvm: Failed to load virtio-net:virtio
qemu-kvm: load of migration failed: Input/output error


Version-Release number of selected component (if applicable):
qemu-kvm-rhev-2.12.0-18.el7.x86_64

How reproducible:
100%

Steps to Reproduce:
1. Boot up both src and dst guests, the images are on shared storage:
/usr/libexec/qemu-kvm \
    -name 'avocado-vt-vm1'  \
    -sandbox off  \
    -machine pc  \
    -nodefaults  \
    -vga std  \
    -chardev socket,id=qmp_id_qmpmonitor1,path=/var/tmp/avocado1,server,nowait \
    -mon chardev=qmp_id_qmpmonitor1,mode=control  \
    -device nec-usb-xhci,id=usb1,bus=pci.0,addr=0x4\
    -device virtio-serial-pci,id=virtio-serial0,bus=pci.0,addr=0x5 \
    -chardev socket,path=/tmp/qga.sock,server,nowait,id=qga0 \
    -device virtserialport,bus=virtio-serial0.0,chardev=qga0,id=qemu-ga0,name=org.qemu.guest_agent.0 \
    -device virtio-scsi-pci,id=virtio_scsi_pci0,bus=pci.0,addr=0x6 \
    -blockdev node-name=disk0,file.driver=file,driver=qcow2,file.filename=/home/kvm_autotest_root/images/rhel75-64-virtio-scsi.qcow2 \
    -device scsi-hd,drive=disk0,id=image1 \
    -blockdev node-name=disk1,file.driver=file,driver=qcow2,file.filename=/home/kvm_autotest_root/images/hd1 \
    -device scsi-hd,drive=disk1,id=image11 \
    -device virtio-net-pci,mac=9a:78:79:7a:7b:7d,id=id8e5D72,vectors=4,netdev=idrYUYaH,bus=pci.0,addr=0x3 \
    -netdev tap,id=idrYUYaH,vhost=on \
    -m 4096  \
    -smp 4,maxcpus=4,cores=2,threads=1,sockets=2  \
    -device usb-tablet,id=usb-tablet1,bus=usb1.0,port=1  \
    -cpu Penryn \
    -vnc :20  \
    -rtc base=localtime,clock=host,driftfix=slew  \
    -boot menu=off,strict=off,order=cdn,once=c \
    -enable-kvm \
    -monitor stdio

    **-incoming tcp:0:5200** #Only for dst guest

2. In src side, create an image with qemu-img:
# qemu-img create -f qcow2 /home/kvm_autotest_root/images/fullbackup.qcow2 2G

3. In both src and dst sides, hot plug the image with blockdev-add in qmp:
{"execute":"blockdev-add","arguments":{"driver":"file","node-name":"fullbackup","filename":"/home/kvm_autotest_root/images/fullbackup.qcow2"}}
{"return": {}}
{ 'execute':'blockdev-add','arguments':{'driver':'qcow2','node-name':'fbk','file':'fullbackup'}}
{"return": {}}

4. In src side, do migration and check its status in qmp:
{"execute":"migrate","arguments":{"uri":"tcp:10.66.8.198:5200"}}
{"execute":"query-migrate"}


Actual results:
In step4, qemu core dump when the migration is nearly to finish, the detailed debug info is as follows:
(gdb) bt full
#0  0x00007f00f62fc207 in raise () at /lib64/libc.so.6
#1  0x00007f00f62fd8f8 in abort () at /lib64/libc.so.6
#2  0x00007f00f62f5026 in __assert_fail_base () at /lib64/libc.so.6
#3  0x00007f00f62f50d2 in  () at /lib64/libc.so.6
#4  0x00005598f83b10fe in bdrv_child_cb_inactivate (child=<optimized out>) at block.c:855
        child = <optimized out>
        bs = <optimized out>
#5  0x00005598f83b2e82 in bdrv_inactivate_recurse (bs=0x5598fa902800, setting_flag=setting_flag@entry=true) at block.c:4460
        perm = <optimized out>
        shared_perm = <optimized out>
        child = <optimized out>
        parent = 0x5598face2500
        ret = <optimized out>
#6  0x00005598f83b51ed in bdrv_inactivate_all () at block.c:4512
        bs = <optimized out>
        it = {phase = BDRV_NEXT_MONITOR_OWNED, blk = 0x0, bs = 0x5598fa902800}
        ret = <optimized out>
        pass = 1
        aio_ctxs = <optimized out>
        ctx = <optimized out>
#7  0x00005598f83564c2 in qemu_savevm_state_complete_precopy (f=0x5598fb512000, iterable_only=<optimized out>, inactivate_disks=<optimized out>) at migration/savevm.c:1198
        vmdesc = 0x5598fc98e9b0
        vmdesc_len = <optimized out>
        se = 0x0
        ret = <optimized out>
        in_postcopy = false
        __func__ = "qemu_savevm_state_complete_precopy"
#8  0x00005598f835235e in migration_thread (opaque=0x5598fa5ec500) at migration/migration.c:2144
        s = 0x5598fa5ec500
        setup_start = <optimized out>
#9  0x00007f00f669add5 in start_thread () at /lib64/libpthread.so.0
#10 0x00007f00f63c3ead in clone () at /lib64/libc.so.6
(gdb) 


Expected results:
No core dump when do migration after hot plug the backend image.

Additional info:
1. Migrating guest after hot plugging both backend image and the frontend one(with device_add) is without the bug problem.
2. Migrating guest after hot plugging an backend image with '__com.redhat_drive_add'(while without the frontend) is without the bug problem.