ReplicatedPG,rados: add CEPH_OSD_[COPY_FROM]_MAP_SNAP_TO_CLONE

When promoting a clone, we want to use the provided snapid to specify
specify the clone id directly.

Signed-off-by: Samuel Just <sam.just@inktank.com>

diff --git a/src/include/rados.h b/src/include/rados.h
index 49391d947ed7d123df9fab3d1126e46cc1246a1c..0d02b241c1d7d9f4c96d3535d56ca332c3bfbe52 100644 (file)
--- a/src/include/rados.h
+++ b/src/include/rados.h
@@ -372,6 +372,8 @@ enum {
        CEPH_OSD_FLAG_SKIPRWLOCKS =   0x10000,  /* skip rw locks */
        CEPH_OSD_FLAG_IGNORE_OVERLAY =0x20000,  /* ignore pool overlay */
        CEPH_OSD_FLAG_FLUSH =         0x40000,  /* this is part of flush */
+       CEPH_OSD_FLAG_MAP_SNAP_CLONE =0x80000,  /* map snap direct to clone id
+                                                */
 };
 
 enum {
@@ -401,6 +403,8 @@ enum {
        CEPH_OSD_COPY_FROM_FLAG_FLUSH = 1,     /* part of a flush operation */
        CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY = 2,  /* ignore pool overlay */
        CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE = 4, /* ignore osd cache logic */
+       CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE = 8, /* map snap direct to
+                                                    * cloneid */
 };
 
 enum {

diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc
index 06034667bfb9e10672ca41fbf0c89d784ca1e378..1a9260a0028945c502c67d43e067e5c7a638f0cc 100644 (file)
--- a/src/osd/ReplicatedPG.cc
+++ b/src/osd/ReplicatedPG.cc
@@ -1262,7 +1262,10 @@ void ReplicatedPG::do_op(OpRequestRef op)
                m->get_pg().ps(),
                m->get_object_locator().get_pool(),
                m->get_object_locator().nspace);
-  int r = find_object_context(oid, &obc, can_create, &missing_oid);
+  int r = find_object_context(
+    oid, &obc, can_create,
+    m->get_flags() & CEPH_OSD_FLAG_MAP_SNAP_CLONE,
+    &missing_oid);
 
   if (r == -EAGAIN) {
     // If we're not the primary of this OSD, and we have
@@ -1358,7 +1361,8 @@ void ReplicatedPG::do_op(OpRequestRef op)
        if (src_oid.is_head() && is_missing_object(src_oid)) {
          wait_for_unreadable_object(src_oid, op);
        } else if ((r = find_object_context(
-                     src_oid, &sobc, false, &wait_oid)) == -EAGAIN) {
+                     src_oid, &sobc, false, false,
+                     &wait_oid)) == -EAGAIN) {
          // missing the specific snap we need; requeue and wait.
          wait_for_unreadable_object(wait_oid, op);
        } else if (r) {
@@ -1628,7 +1632,8 @@ void ReplicatedPG::promote_object(OpRequestRef op, ObjectContextRef obc,
   oloc.pool = pool.info.tier_of;
   start_copy(cb, obc, obc->obs.oi.soid, oloc, 0,
             CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY |
-            CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE,
+            CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE |
+            CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE,
             obc->obs.oi.soid.snap == CEPH_NOSNAP);
 
   assert(obc->is_blocked());
@@ -4563,7 +4568,7 @@ int ReplicatedPG::_rollback_to(OpContext *ctx, ceph_osd_op& op)
   int ret = find_object_context(
     hobject_t(soid.oid, soid.get_key(), snapid, soid.hash, info.pgid.pool(),
              soid.get_namespace()),
-    &rollback_to, false, &missing_oid);
+    &rollback_to, false, false, &missing_oid);
   if (ret == -EAGAIN) {
     /* clone must be missing */
     assert(is_missing_object(missing_oid));
@@ -5453,6 +5458,8 @@ void ReplicatedPG::_copy_some(ObjectContextRef obc, CopyOpRef cop)
     flags |= CEPH_OSD_FLAG_IGNORE_CACHE;
   if (cop->flags & CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY)
     flags |= CEPH_OSD_FLAG_IGNORE_OVERLAY;
+  if (cop->flags & CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE)
+    flags |= CEPH_OSD_FLAG_MAP_SNAP_CLONE;
 
   C_GatherBuilder gather(g_ceph_context);
 
@@ -6048,7 +6055,8 @@ int ReplicatedPG::start_flush(OpContext *ctx, bool blocking, hobject_t *pmissing
     o.copy_from(soid.oid.name, soid.snap, oloc, oi.user_version,
                CEPH_OSD_COPY_FROM_FLAG_FLUSH |
                CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY |
-               CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE);
+               CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE |
+               CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE);
   }
   C_Flush *fin = new C_Flush(this, soid, get_last_peering_reset());
   object_locator_t base_oloc(soid);
@@ -6971,6 +6979,7 @@ void ReplicatedPG::context_registry_on_change()
 int ReplicatedPG::find_object_context(const hobject_t& oid,
                                      ObjectContextRef *pobc,
                                      bool can_create,
+                                     bool map_snapid_to_clone,
                                      hobject_t *pmissing)
 {
   hobject_t head(oid.oid, oid.get_key(), CEPH_NOSNAP, oid.hash,
@@ -7024,7 +7033,7 @@ int ReplicatedPG::find_object_context(const hobject_t& oid,
   }
 
   // we want a snap
-  if (pool.info.is_removed_snap(oid.snap)) {
+  if (!map_snapid_to_clone && pool.info.is_removed_snap(oid.snap)) {
     dout(10) << __func__ << " snap " << oid.snap << " is removed" << dendl;
     return -ENOENT;
   }
@@ -7037,6 +7046,66 @@ int ReplicatedPG::find_object_context(const hobject_t& oid,
     return -ENOENT;
   }
 
+  if (map_snapid_to_clone) {
+    dout(10) << "find_object_context " << oid << " @" << oid.snap
+            << " snapset " << ssc->snapset
+            << " map_snapid_to_clone=true" << dendl;
+    if (oid.snap > ssc->snapset.seq) {
+      // already must be readable
+      ObjectContextRef obc = get_object_context(head, false);
+      dout(10) << "find_object_context " << oid << " @" << oid.snap
+              << " snapset " << ssc->snapset
+              << " maps to head" << dendl;
+      *pobc = obc;
+      put_snapset_context(ssc);
+      return (obc && obc->obs.exists) ? 0 : -ENOENT;
+    } else {
+      vector<snapid_t>::const_iterator citer = std::find(
+       ssc->snapset.clones.begin(),
+       ssc->snapset.clones.end(),
+       oid.snap);
+      if (citer == ssc->snapset.clones.end()) {
+       dout(10) << "find_object_context " << oid << " @" << oid.snap
+                << " snapset " << ssc->snapset
+                << " maps to nothing" << dendl;
+       put_snapset_context(ssc);
+       return -ENOENT;
+      }
+
+      dout(10) << "find_object_context " << oid << " @" << oid.snap
+              << " snapset " << ssc->snapset
+              << " maps to " << oid << dendl;
+
+      if (pg_log.get_missing().is_missing(oid)) {
+       dout(10) << "find_object_context " << oid << " @" << oid.snap
+                << " snapset " << ssc->snapset
+                << " " << oid << " is missing" << dendl;
+       if (pmissing)
+         *pmissing = oid;
+       put_snapset_context(ssc);
+       return -EAGAIN;
+      }
+
+      ObjectContextRef obc = get_object_context(oid, false);
+      if (!obc || !obc->obs.exists) {
+       dout(10) << "find_object_context " << oid << " @" << oid.snap
+                << " snapset " << ssc->snapset
+                << " " << oid << " is not present" << dendl;
+       if (pmissing)
+         *pmissing = oid;
+       put_snapset_context(ssc);
+       return -ENOENT;
+      }
+      dout(10) << "find_object_context " << oid << " @" << oid.snap
+              << " snapset " << ssc->snapset
+              << " " << oid << " HIT" << dendl;
+      *pobc = obc;
+      put_snapset_context(ssc);
+      return 0;
+    }
+    assert(0); //unreachable
+  }
+
   dout(10) << "find_object_context " << oid << " @" << oid.snap
           << " snapset " << ssc->snapset << dendl;
  

diff --git a/src/osd/ReplicatedPG.h b/src/osd/ReplicatedPG.h
index 239d118c9f0ce42819007e8211b05af5b32f3c64..76d2d1a8f13edb8392e36c1c4a68965c169b5a18 100644 (file)
--- a/src/osd/ReplicatedPG.h
+++ b/src/osd/ReplicatedPG.h
@@ -878,6 +878,7 @@ protected:
   int find_object_context(const hobject_t& oid,
                          ObjectContextRef *pobc,
                          bool can_create,
+                         bool map_snapid_to_clone=false,
                          hobject_t *missing_oid=NULL);
 
   void add_object_context_to_pg_stat(ObjectContextRef obc, pg_stat_t *stat);

diff --git a/src/osd/osd_types.cc b/src/osd/osd_types.cc
index dca7bbf605f3b9128d237cfdd2d7de5ff5c8e358..e0fd85f32d3f58f6c09747ea3bad4df402c568ae 100644 (file)
--- a/src/osd/osd_types.cc
+++ b/src/osd/osd_types.cc
@@ -46,6 +46,7 @@ const char *ceph_osd_flag_name(unsigned flag)
   case CEPH_OSD_FLAG_SKIPRWLOCKS: return "skiprwlocks";
   case CEPH_OSD_FLAG_IGNORE_OVERLAY: return "ignore_overlay";
   case CEPH_OSD_FLAG_FLUSH: return "flush";
+  case CEPH_OSD_FLAG_MAP_SNAP_CLONE: return "map_snap_clone";
   default: return "???";
   }
 }