BIND 10 #2836: update in-memory zone loading so it can work with shmem segment

[BIND 10 Development]

#2836: update in-memory zone loading so it can work with shmem segment
-------------------------------------+-------------------------------------
            Reporter:  jinmei        |                        Owner:
                Type:  task          |  vorner
            Priority:  medium        |                       Status:
           Component:  data source   |  reviewing
            Keywords:                |                    Milestone:
           Sensitive:  0             |  Sprint-20130514
         Sub-Project:  DNS           |                   Resolution:
Estimated Difficulty:  4             |                 CVSS Scoring:
         Total Hours:  0             |              Defect Severity:  N/A
                                     |  Feature Depending on Ticket:
                                     |  shared memory data source
                                     |          Add Hours to Ticket:  0
                                     |                    Internal?:  0
-------------------------------------+-------------------------------------

Comment (by vorner):

 Hello

 Replying to [comment:15 jinmei]:
 > As you mentioned a crash on the daily call, I've looked into the
 > latest trac2836 branch (at 2f220f9).
 >
 > I'm not sure what exactly happened to you, but it caused a crash
 > in my environment, too.  In my case, address reallocation (mapped on a
 > different address) never happened until growing the segment finally
 > fails, at which point the `MemorySegmentMapped` object breaks its
 > integrity (already reset the underlying boost segment).  So the
 > subsequent cleanup attempt triggered NULL pointer dereference.

 Thank you, I'll look into this. Actually, my crash is in a code I did not
 commit yet, but I'll investigate and see if it might be related. But I
 think not, my crash is in this code:

 {{{#!c++
 // Load bunch of small zones, hoping some of the relocation will happen
 // during the memory creation, not only Rdata creation.
 TEST(ZoneDataLoaterTest, relocate) {
     MemorySegmentMapped segment(mapped_file,
 isc::util::MemorySegmentMapped::CREATE_ONLY,
                                 4096);
     const size_t zone_count = 10000;
     typedef SegmentObjectHolder<ZoneData, RRClass> Holder;
     typedef boost::shared_ptr<Holder> HolderPtr;
     std::vector<HolderPtr> zones;
     for (size_t i = 0; i < zone_count; ++i) {
         // Load some zone
         ZoneData* data = loadZoneData(segment, RRClass::IN(),
                                       Name("example.org"),
                                       TEST_DATA_DIR
                                       "/example.org-nsec3-signed.zone");
         // Store it, so it is cleaned up later
         zones.push_back(HolderPtr(new Holder(segment, data,
                                              RRClass::IN())));

     }
     // Deallocate all the zones now.
     zones.clear();
     EXPECT_TRUE(segment.allMemoryDeallocated());
     EXPECT_EQ(0, unlink(mapped_file));
 }
 }}}

 I tried to put a breakpoint onto the place that throws the SegmentGrown
 exception and it actually never happened. What is even more strange, it
 does not crash with the local segment. I get a segfault somewhere in the
 rbtree, but my valgrind is broken and refuses to start.

-- 
Ticket URL: <http://bind10.isc.org/ticket/2836#comment:16>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development