WT-3261 add a checkpoint epoch to avoid draining the eviction queue (#3370)

4 files changed, 30 insertions, 28 deletions

diff --git a/src/include/btree.h b/src/include/btree.h
index 19db27d84a2..8ce77b5ecd3 100644
--- a/src/include/btree.h
+++ b/src/include/btree.h
@@ -151,7 +151,7 @@ struct __wt_btree {
 	volatile uint32_t evict_busy;	/* Count of threads in eviction */
 	int	    evict_start_type;	/* Start position for eviction walk
 					   (see WT_EVICT_WALK_START). */
-	enum {
+	volatile enum {
 		WT_CKPT_OFF, WT_CKPT_PREPARE, WT_CKPT_RUNNING
 	} checkpointing;		/* Checkpoint in progress */
 
@@ -163,19 +163,18 @@ struct __wt_btree {
 	WT_SPINLOCK	flush_lock;	/* Lock to flush the tree's pages */
 
 	/* Flags values up to 0xff are reserved for WT_DHANDLE_* */
-#define	WT_BTREE_ALLOW_SPLITS	0x000100 /* Allow splits, even with no evict */
-#define	WT_BTREE_BULK		0x000200 /* Bulk-load handle */
-#define	WT_BTREE_CLOSED		0x000400 /* Handle closed */
-#define	WT_BTREE_IGNORE_CACHE	0x000800 /* Cache-resident object */
-#define	WT_BTREE_IN_MEMORY	0x001000 /* Cache-resident object */
-#define	WT_BTREE_LOOKASIDE	0x002000 /* Look-aside table */
-#define	WT_BTREE_NO_CHECKPOINT	0x004000 /* Disable checkpoints */
-#define	WT_BTREE_NO_LOGGING	0x008000 /* Disable logging */
-#define	WT_BTREE_REBALANCE	0x020000 /* Handle is for rebalance */
-#define	WT_BTREE_SALVAGE	0x040000 /* Handle is for salvage */
-#define	WT_BTREE_SKIP_CKPT	0x080000 /* Handle skipped checkpoint */
-#define	WT_BTREE_UPGRADE	0x100000 /* Handle is for upgrade */
-#define	WT_BTREE_VERIFY		0x200000 /* Handle is for verify */
+#define	WT_BTREE_BULK		0x000100 /* Bulk-load handle */
+#define	WT_BTREE_CLOSED		0x000200 /* Handle closed */
+#define	WT_BTREE_IGNORE_CACHE	0x000400 /* Cache-resident object */
+#define	WT_BTREE_IN_MEMORY	0x000800 /* Cache-resident object */
+#define	WT_BTREE_LOOKASIDE	0x001000 /* Look-aside table */
+#define	WT_BTREE_NO_CHECKPOINT	0x002000 /* Disable checkpoints */
+#define	WT_BTREE_NO_LOGGING	0x004000 /* Disable logging */
+#define	WT_BTREE_REBALANCE	0x008000 /* Handle is for rebalance */
+#define	WT_BTREE_SALVAGE	0x010000 /* Handle is for salvage */
+#define	WT_BTREE_SKIP_CKPT	0x020000 /* Handle skipped checkpoint */
+#define	WT_BTREE_UPGRADE	0x040000 /* Handle is for upgrade */
+#define	WT_BTREE_VERIFY		0x080000 /* Handle is for verify */
 	uint32_t flags;
 };
 
diff --git a/src/include/btree.i b/src/include/btree.i
index 474b40bf805..d4db65b2033 100644
--- a/src/include/btree.i
+++ b/src/include/btree.i
@@ -1286,6 +1286,16 @@ __wt_page_can_evict(
 		return (true);
 
 	/*
+	 * We can't split or evict multiblock row-store pages where the parent's
+	 * key for the page is an overflow item, because the split into the
+	 * parent frees the backing blocks for any no-longer-used overflow keys,
+	 * which will corrupt the checkpoint's block management.
+	 */
+	if (btree->checkpointing != WT_CKPT_OFF &&
+	    F_ISSET_ATOMIC(ref->home, WT_PAGE_OVERFLOW_KEYS))
+		return (false);
+
+	/*
 	 * Check for in-memory splits before other eviction tests. If the page
 	 * should split in-memory, return success immediately and skip more
 	 * detailed eviction tests. We don't need further tests since the page
@@ -1312,16 +1322,6 @@ __wt_page_can_evict(
 	}
 
 	/*
-	 * We can't evict clean, multiblock row-store pages where the parent's
-	 * key for the page is an overflow item, because the split into the
-	 * parent frees the backing blocks for any no-longer-used overflow keys,
-	 * which will corrupt the checkpoint's block management.
-	 */
-	if (btree->checkpointing != WT_CKPT_OFF &&
-	    F_ISSET_ATOMIC(ref->home, WT_PAGE_OVERFLOW_KEYS))
-		return (false);
-
-	/*
 	 * If a split created new internal pages, those newly created internal
 	 * pages cannot be evicted until all threads are known to have exited
 	 * the original parent page's index, because evicting an internal page
diff --git a/src/include/extern.h b/src/include/extern.h
index ff835bacc56..4f38b7ac433 100644
--- a/src/include/extern.h
+++ b/src/include/extern.h
@@ -649,6 +649,8 @@ extern int __wt_unexpected_object_type( WT_SESSION_IMPL *session, const char *ur
 extern void __wt_gen_init(WT_SESSION_IMPL *session) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
 extern uint64_t __wt_gen(WT_SESSION_IMPL *session, int which) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
 extern uint64_t __wt_gen_next(WT_SESSION_IMPL *session, int which) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
+extern uint64_t __wt_gen_next_drain(WT_SESSION_IMPL *session, int which) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
+extern void __wt_gen_drain(WT_SESSION_IMPL *session, int which, uint64_t generation) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
 extern uint64_t __wt_gen_oldest(WT_SESSION_IMPL *session, int which) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
 extern uint64_t __wt_session_gen(WT_SESSION_IMPL *session, int which) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
 extern void __wt_session_gen_enter(WT_SESSION_IMPL *session, int which) WT_GCC_FUNC_DECL_ATTRIBUTE((visibility("hidden")));
diff --git a/src/include/session.h b/src/include/session.h
index adef5e39068..de2c1463684 100644
--- a/src/include/session.h
+++ b/src/include/session.h
@@ -169,10 +169,11 @@ struct __wt_session_impl {
 
 					/* Generations manager */
 #define	WT_GEN_CHECKPOINT	0	/* Checkpoint generation */
-#define	WT_GEN_HAZARD		1	/* Hazard pointer */
-#define	WT_GEN_SCHEMA		2	/* Schema version */
-#define	WT_GEN_SPLIT		3	/* Page splits */
-#define	WT_GENERATIONS		4	/* Total generation manager entries */
+#define	WT_GEN_EVICT		1	/* Eviction generation */
+#define	WT_GEN_HAZARD		2	/* Hazard pointer */
+#define	WT_GEN_SCHEMA		3	/* Schema version */
+#define	WT_GEN_SPLIT		4	/* Page splits */
+#define	WT_GENERATIONS		5	/* Total generation manager entries */
 	volatile uint64_t generations[WT_GENERATIONS];
 
 	/*