osd: prime pg_slots for to-be-split children

Once we know which PGs are about to be created, we instantiate their
pg_slot and mark them waiting_pg, which blocks all incoming events until
the split completes, the PG is installed, and we call wake_pg_waiters().

Signed-off-by: Sage Weil <sage@redhat.com>

diff --git a/src/osd/OSD.cc b/src/osd/OSD.cc
index 427418672c25179d5403dc5e6243462d19c259b1..c26c26af975ad583cb85f04988122078e745178f 100644 (file)
--- a/src/osd/OSD.cc
+++ b/src/osd/OSD.cc
@@ -8149,6 +8149,9 @@ void OSD::split_pgs(
   OSDMapRef nextmap,
   PG::RecoveryCtx *rctx)
 {
+  // make sure to-be-split children are blocked in wq
+  op_shardedwq.prime_splits(childpgids);
+
   unsigned pg_num = nextmap->get_pg_num(
     parent->pg_id.pool());
   parent->update_snap_mapper_bits(
@@ -9477,6 +9480,18 @@ void OSD::ShardedOpWQ::wake_pg_waiters(spg_t pgid)
   }
 }
 
+void OSD::ShardedOpWQ::prime_splits(const set<spg_t>& pgs)
+{
+  dout(20) << __func__ << " " << pgs << dendl;
+  for (auto pgid : pgs) {
+    unsigned shard_index = pgid.hash_to_shard(shard_list.size());
+    ShardData* sdata = shard_list[shard_index];
+    Mutex::Locker l(sdata->sdata_op_ordering_lock);
+    ShardData::pg_slot& slot = sdata->pg_slots[pgid];
+    slot.waiting_for_pg = true;
+  }
+}
+
 void OSD::ShardedOpWQ::prune_pg_waiters(OSDMapRef osdmap, int whoami)
 {
   unsigned pushes_to_free = 0;

diff --git a/src/osd/OSD.h b/src/osd/OSD.h
index 83df4fd503162044b41c54bd8408bee15db413b8..1b16133fc6a950bb8ffbaf6b0495c9fd8b0c3a71 100644 (file)
--- a/src/osd/OSD.h
+++ b/src/osd/OSD.h
@@ -1680,6 +1680,9 @@ private:
     void _wake_pg_slot(spg_t pgid, ShardData *sdata, ShardData::pg_slot& slot,
                       unsigned *pushes_to_free);
 
+    /// prime slots for splitting pgs
+    void prime_splits(const set<spg_t>& pgs);
+
     /// prune ops (and possibly pg_slots) for pgs that shouldn't be here
     void prune_pg_waiters(OSDMapRef osdmap, int whoami);