os/bluestore: list obj which equals to pend

otherwise we could have failures like

scrub : stat mismatch, got 3/4 objects, 1/2 clones, 3/4 dirty, 3/4 omap, 0/0 pinned, 0/0 hit_set_archive, 0/0 whiteouts, 49/56 bytes, 0/0 manifest objects, 0/0 hit_set_archive bytes."

where the numbers of scrubbed object, clones, dirty and omap are always
less than the total number of corresponding numbers, if the PG contains
object(s) whose hash happens to be 0xffffffff.

in this change, if the calculated hash of the upper bound is greater
than the maximum possible number represented by uint32_t, in addition to
setting the hash of the upper bound hobj to 0xffffffff, we also set the
nspace of hobj of the upper bound to "\xff", so that the upper bound
is greater than an hobj whose hash happens to be 0xfffffff. please note,
the nspace of "\xff" is not an ascii string, so it's not likely to be
less than a real-world nspace of an hobj.

with this new *greater* upper bound, we are able to include the previous
missing hobj when listing the objects in a PG. so the scrub won't be
annoyed when the number of objects does not match.

Fixes: https://tracker.ceph.com/issues/52705
Signed-off-by: Mykola Golub <mykola.golub@clyso.com>
Signed-off-by: Kefu Chai <tchaikov@gmail.com>
(cherry picked from commit ffab13bcd9006c1f961a24b8016df9d1fe06ba1d)

 Conflicts:
	src/os/bluestore/BlueStore.cc
 - get_coll_range function signature alignment

diff --git a/src/os/bluestore/BlueStore.cc b/src/os/bluestore/BlueStore.cc
index b3a4625920c82b912b72086d9386708d385689d9..acb79228ae2bd8f4d9b50fede64a347d910ff241 100644 (file)
--- a/src/os/bluestore/BlueStore.cc
+++ b/src/os/bluestore/BlueStore.cc
@@ -296,10 +296,14 @@ static const char *_key_decode_shard(const char *key, shard_id_t *pshard)
 }
 
 static void get_coll_range(const coll_t& cid, int bits,
-                           ghobject_t *temp_start, ghobject_t *temp_end,
-                           ghobject_t *start, ghobject_t *end)
+  ghobject_t *temp_start, ghobject_t *temp_end,
+  ghobject_t *start, ghobject_t *end, bool legacy)
 {
   spg_t pgid;
+  constexpr uint32_t MAX_HASH = std::numeric_limits<uint32_t>::max();
+  // use different nspaces due to we use different schemes when encoding
+  // keys for listing objects
+  const std::string_view MAX_NSPACE = legacy ? "\x7f" : "\xff";
   if (cid.is_pg(&pgid)) {
     start->shard_id = pgid.shard;
     *temp_start = *start;
@@ -315,19 +319,23 @@ static void get_coll_range(const coll_t& cid, int bits,
     temp_start->hobj.set_bitwise_key_u32(reverse_hash);
 
     uint64_t end_hash = reverse_hash  + (1ull << (32 - bits));
-    if (end_hash > 0xffffffffull)
-      end_hash = 0xffffffffull;
-
-    end->hobj.set_bitwise_key_u32(end_hash);
-    temp_end->hobj.set_bitwise_key_u32(end_hash);
+    if (end_hash > MAX_HASH) {
+      // make sure end hobj is even greater than the maximum possible hobj
+      end->hobj.set_bitwise_key_u32(MAX_HASH);
+      temp_end->hobj.set_bitwise_key_u32(MAX_HASH);
+      end->hobj.nspace = MAX_NSPACE;
+    } else {
+      end->hobj.set_bitwise_key_u32(end_hash);
+      temp_end->hobj.set_bitwise_key_u32(end_hash);
+    }
   } else {
     start->shard_id = shard_id_t::NO_SHARD;
     start->hobj.pool = -1ull;
 
     *end = *start;
     start->hobj.set_bitwise_key_u32(0);
-    end->hobj.set_bitwise_key_u32(0xffffffff);
-
+    end->hobj.set_bitwise_key_u32(MAX_HASH);
+    end->hobj.nspace = MAX_NSPACE;
     // no separate temp section
     *temp_start = *end;
     *temp_end = *end;
@@ -10982,7 +10990,7 @@ int BlueStore::_collection_list(
     return 0;
   }
   get_coll_range(c->cid, c->cnode.bits, &coll_range_temp_start,
-                 &coll_range_temp_end, &coll_range_start, &coll_range_end);
+                 &coll_range_temp_end, &coll_range_start, &coll_range_end, legacy);
   dout(20) << __func__
     << " range " << coll_range_temp_start
     << " to " << coll_range_temp_end

diff --git a/src/test/objectstore/store_test.cc b/src/test/objectstore/store_test.cc
index 846185a077e76d6f6c508d5e9476f603daf961ad..a2cdfcc1da1eff4bc37427f90ba62e93bd20041f 100644 (file)
--- a/src/test/objectstore/store_test.cc
+++ b/src/test/objectstore/store_test.cc
@@ -2928,6 +2928,61 @@ TEST_P(StoreTest, ListEndTest) {
   }
 }
 
+TEST_P(StoreTest, List_0xfffffff_Hash_Test_in_meta) {
+  int r = 0;
+  coll_t cid;
+  auto ch = store->create_new_collection(cid);
+  {
+    ObjectStore::Transaction t;
+    t.create_collection(cid, 0);
+    r = queue_transaction(store, ch, std::move(t));
+    ASSERT_EQ(r, 0);
+  }
+  {
+    ObjectStore::Transaction t;
+    ghobject_t hoid(hobject_t(sobject_t("obj", CEPH_NOSNAP),
+                             "", UINT32_C(0xffffffff), -1, "nspace"));
+    t.touch(cid, hoid);
+    r = queue_transaction(store, ch, std::move(t));
+    ASSERT_EQ(r, 0);
+  }
+  {
+    vector<ghobject_t> objects;
+    r = collection_list(store, ch, ghobject_t(), ghobject_t::get_max(), INT_MAX,
+                       &objects, nullptr, true);
+    ASSERT_EQ(r, 0);
+    ASSERT_EQ(objects.size(), 1);
+  }
+}
+
+TEST_P(StoreTest, List_0xfffffff_Hash_Test_in_PG) {
+  int r = 0;
+  const int64_t poolid = 1;
+  coll_t cid(spg_t(pg_t(0, poolid), shard_id_t::NO_SHARD));
+  auto ch = store->create_new_collection(cid);
+  {
+    ObjectStore::Transaction t;
+    t.create_collection(cid, 0);
+    r = queue_transaction(store, ch, std::move(t));
+    ASSERT_EQ(r, 0);
+  }
+  {
+    ObjectStore::Transaction t;
+    ghobject_t hoid(hobject_t(sobject_t("obj", CEPH_NOSNAP),
+                             "", UINT32_C(0xffffffff), poolid, "nspace"));
+    t.touch(cid, hoid);
+    r = queue_transaction(store, ch, std::move(t));
+    ASSERT_EQ(r, 0);
+  }
+  {
+    vector<ghobject_t> objects;
+    r = collection_list(store, ch, ghobject_t(), ghobject_t::get_max(), INT_MAX,
+                       &objects, nullptr, true);
+    ASSERT_EQ(r, 0);
+    ASSERT_EQ(objects.size(), 1);
+  }
+}
+
 TEST_P(StoreTest, Sort) {
   {
     hobject_t a(sobject_t("a", CEPH_NOSNAP));