From c2cb3345069d0ba6c0029ee5f4f9ebc89e42dcad Mon Sep 17 00:00:00 2001
From: Sears Russell <sears@cs.berkeley.edu>
Date: Wed, 25 Aug 2010 16:37:05 +0000
Subject: [PATCH] add comment to explain mysterious policy in concurrentWrapper

---
 src/stasis/replacementPolicy/concurrentWrapper.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/stasis/replacementPolicy/concurrentWrapper.c b/src/stasis/replacementPolicy/concurrentWrapper.c
index 5a1a436..b4aabfd 100644
--- a/src/stasis/replacementPolicy/concurrentWrapper.c
+++ b/src/stasis/replacementPolicy/concurrentWrapper.c
@@ -72,6 +72,10 @@ static void* cwGetStaleHelper(struct replacementPolicy* impl, void*(*func)(struc
     }
   }
   if(bucket != oldbucket) {
+    // note that, even on success, we increment the bucket.  Otherwise, we could
+    // (would) eventually get unlucky, and some caller would do a getStaleAndRemove,
+    // fail to get a latch, insert it back, and the next getStaleAndRemove would
+    // deterministically return the same page again, leading to an infinite loop.
     pthread_setspecific(rp->next_bucket, (void*) bucket);
   }
   return ret;