Mailing-List: contact reviews-help@impala.incubator.apache.org; run by ezmlm
Precedence: bulk
Message-Id: <201708050024.v750OOAH018021@ip-10-146-233-104.ec2.internal>
Date: Sat, 5 Aug 2017 00:24:24 +0000
From: "Henry Robinson (Code Review)" <gerrit@cloudera.org>
To: impala-cr@cloudera.com, reviews@impala.incubator.apache.org
CC: Tim Armstrong <tarmstrong@cloudera.com>
Reply-To: henry@cloudera.com
Subject: =?UTF-8?Q?=5BImpala-ASF-CR=5D_IMPALA-5666=3A_ASAN_poisoning_for_MemPool_and_BufferPool=0A?=
In-Reply-To: <gerrit.1501882968000.I8d5a28dfee2b7c631981aac75524bde9acc0b36a@gerrit.cloudera.org>
References: <gerrit.1501882968000.I8d5a28dfee2b7c631981aac75524bde9acc0b36a@gerrit.cloudera.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Content-Disposition: inline
User-Agent: Gerrit/2.12.7
archived-at: Sat, 05 Aug 2017 00:24:29 -0000

Henry Robinson has uploaded a new patch set (#2).

Change subject: IMPALA-5666: ASAN poisoning for MemPool and BufferPool
......................................................................

IMPALA-5666: ASAN poisoning for MemPool and BufferPool

"If you poison us, do we not die?"

* Use ASAN_[UN]POISON_MEMORY_REGION to indicate to ASAN runtime that
  memory is not 'valid' from Impala's perspective (but still valid from
  kernel's perspective).

* Add this to MemPool and BufferPoolAllocator. For both, memory goes
  through the following lifecycle: when it is allocated from the OS and
  returned to the user, it must be unpoisoned. When the user returns it
  to a cache, it becomes poisoned. When the cache is freed to the OS, it
  must be unpoisoned again (otherwise future allocations elsewhere in
  the system might return poisoned memory).

* Also add checks to FreePool which uses a MemPool underneath, but has
  its own freelist cache. Fix a couple of bugs in free-pool-test that
  these checks uncovered.

Testing:

* Tests that only run if ASAN is enabled to confirm that poisoning
  catches simple use-after-return errors. These are 'death' tests which
  catch expected process exits.

* Fix one bug found in StringFunctions::FindInSet which would read one
  byte past the end of a stringvalue.

Change-Id: I8d5a28dfee2b7c631981aac75524bde9acc0b36a
---
M be/src/exprs/string-functions-ir.cc
M be/src/runtime/bufferpool/buffer-allocator-test.cc
M be/src/runtime/bufferpool/buffer-allocator.cc
M be/src/runtime/bufferpool/buffer-pool.h
M be/src/runtime/bufferpool/free-list.h
M be/src/runtime/free-pool-test.cc
M be/src/runtime/free-pool.h
M be/src/runtime/mem-pool-test.cc
M be/src/runtime/mem-pool.cc
M be/src/runtime/mem-pool.h
10 files changed, 146 insertions(+), 18 deletions(-)


  git pull ssh://gerrit.cloudera.org:29418/Impala-ASF refs/changes/91/7591/2
-- 
To view, visit http://gerrit.cloudera.org:8080/7591
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: newpatchset
Gerrit-Change-Id: I8d5a28dfee2b7c631981aac75524bde9acc0b36a
Gerrit-PatchSet: 2
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Henry Robinson <henry@cloudera.com>
Gerrit-Reviewer: Tim Armstrong <tarmstrong@cloudera.com>