libvirtd: fix potential deadlock when reloading

libvirtd: fix potential deadlock when reloading

It is possible to deadlock libvirtd when concurrently starting a domain
and restarting the daemon. Threads involved in the deadlock are

Thread 4 (Thread 0x7fc13b53e700 (LWP 64084)):
/lib64/libpthread.so.0
    at util/virthread.c:154
    at qemu/qemu_monitor.c:1083
    cmd=0x7fc110017700, scm_fd=-1, reply=0x7fc13b53d318) at
qemu/qemu_monitor_json.c:305
cmd=0x7fc110017700,
    reply=0x7fc13b53d318) at qemu/qemu_monitor_json.c:335
    at qemu/qemu_monitor_json.c:1298
    at qemu/qemu_monitor.c:1697
    vm=0x7fc110003d00, asyncJob=QEMU_ASYNC_JOB_START) at qemu/qemu_process.c:1763
vm=0x7fc110003d00,
    asyncJob=6, logCtxt=0x7fc1100089c0) at qemu/qemu_process.c:1835
    vm=0x7fc110003d00, asyncJob=6, logCtxt=0x7fc1100089c0) at
qemu/qemu_process.c:2180
driver=0x7fc12004e1e0,
    vm=0x7fc110003d00, asyncJob=QEMU_ASYNC_JOB_START, incoming=0x0, snapshot=0x0,
    vmop=VIR_NETDEV_VPORT_PROFILE_OP_CREATE, flags=17) at qemu/qemu_process.c:6111
driver=0x7fc12004e1e0,
    vm=0x7fc110003d00, updatedCPU=0x0, asyncJob=QEMU_ASYNC_JOB_START,
migrateFrom=0x0,
    migrateFd=-1, migratePath=0x0, snapshot=0x0,
vmop=VIR_NETDEV_VPORT_PROFILE_OP_CREATE,
    flags=17) at qemu/qemu_process.c:6334
    xml=0x7fc110000ed0 "<!--\nWARNING: THIS IS AN AUTO-GENERATED FILE.
CHANGES TO IT ARE LIKELY TO BE\nOVERWRITTEN AND LOST. Changes to this xml
configuration should be made using:\n  virsh edit testvv\nor other
applicati"..., flags=0) at qemu/qemu_driver.c:1776
...

Thread 1 (Thread 0x7fc143c66880 (LWP 64081)):
/lib64/libpthread.so.0
    at util/virthread.c:122
conf/nwfilter_conf.c:159
sig=0x7ffe0a831e30,
    opaque=0x0) at remote/remote_daemon.c:724
    opaque=0x558c5328b230) at rpc/virnetdaemon.c:654
    at util/vireventpoll.c:508
rpc/virnetdaemon.c:858
remote/remote_daemon.c:1496
(gdb) thr 1
[Switching to thread 1 (Thread 0x7fc143c66880 (LWP 64081))]
/lib64/libpthread.so.0
(gdb) f 1
    at util/virthread.c:122
122	    pthread_rwlock_wrlock(&m->lock);
(gdb) p updateLock
$1 = {lock = {__data = {__lock = 0, __nr_readers = 1, __readers_wakeup = 0,
      __writer_wakeup = 0, __nr_readers_queued = 0, __nr_writers_queued = 1,
__writer = 0,
      __shared = 0, __rwelision = 0 '\000', __pad1 = "\000\000\000\000\000\000",
      __pad2 = 0, __flags = 0},
    __size = "\000\000\000\000\001", '\000' <repeats 15 times>, "\001",
'\000' <repeats 34 times>, __align = 4294967296}}

Reloading of the nwfilter driver is stuck waiting for a write lock, which
already has a reader (from qemuDomainCreateXML) in the critical section.
Since the reload occurs in the context of the main event loop thread,
libvirtd becomes deadlocked. The deadlock can be avoided by offloading
the reload work to a thread.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>