Fix deadlock with coalesce and vdi activation

If the VDI is being coalesced, the VDI activation would want to
interrupt it.
But the LVMSR coalesce needs the lvchange-p lock of the SR.
Moved the check for chain coalesce before taking the lock.

Signed-off-by: Damien Thenot <damien.thenot@vates.tech>

Author: Nambrok

drivers/blktap2.py

@@ -1772,6 +1772,8 @@ def _activate_locked(self, sr_uuid, vdi_uuid, options):
 
             vdi_type = self.target.get_vdi_type()
 
+            if not self._check_journal_coalesce_chain(sr_uuid, vdi_uuid):
+                return False
 
             # Take lvchange-p Lock before running
             # tap-ctl open
@@ -1784,11 +1786,6 @@ def _activate_locked(self, sr_uuid, vdi_uuid, options):
                 lock = Lock("lvchange-p", NS_PREFIX_LVM + sr_uuid)
                 lock.acquire()
 
-            if not self._check_journal_coalesce_chain(sr_uuid, vdi_uuid):
-                return False
-            # we could return false from here if we need to retry after relink
-            # #TODO: handling error here
-
             # When we attach a static VDI for HA, we cannot communicate with
             # xapi, because has not started yet. These VDIs are raw.
             if VdiType.isCowImage(vdi_type):