i#513 drx_buf Part 3: Fixes incorrect locking

Commit log for first patchset:
---------------
i#513 drx_buf Part 3: Fixes incorrect locking

Fixes a minor locking bug in fault_event_helper(), and modifies drx_buf
to use rwlocks as opposed to the general built-in drvector lock.
---------------

[toshi, 2017-01-10 07:04:22]

On 2017/01/10 07:00:43, toshi wrote:

Note that the rwlocking stuff isn't strictly necessary. I wrote it and
benchmarked it a while ago, but when I tested it recently it didn't seem to help
much. I believe I was testing on a VM before, so could be the reason I stuck
with the patch then.

Nevertheless, I sent it all in for review. If need be I can remove the rwlocking
code, and just leave the bugfix.

[bruening, 2017-01-10 17:28:40]

LGTM

https://codereview.appspot.com/315360043/diff/1/ext/drx/drx_buf.c
File ext/drx/drx_buf.c (left):

https://codereview.appspot.com/315360043/diff/1/ext/drx/drx_buf.c#oldcode211
ext/drx/drx_buf.c:211: if (!any_bufs_created)
Generally it's faster to have this kind of read check before a write: I'm
curious why you're removing it?

[toshi, 2017-01-10 21:01:11]

https://codereview.appspot.com/315360043/diff/1/ext/drx/drx_buf.c
File ext/drx/drx_buf.c (left):

https://codereview.appspot.com/315360043/diff/1/ext/drx/drx_buf.c#oldcode211
ext/drx/drx_buf.c:211: if (!any_bufs_created)
On 2017/01/10 17:28:40, bruening wrote:
> Generally it's faster to have this kind of read check before a write: I'm
> curious why you're removing it?

Whoops, my bad. Will put it back.

[toshi, 2017-01-11 02:06:08]

Committed as
https://github.com/DynamoRIO/dynamorio/commit/410b2322590a71b547545c8d6a07d33...

Final commit log: 
---------------
i#513 drx_buf Part 3: Fixes incorrect locking

Fixes a minor locking bug in fault_event_helper(), and modifies drx_buf
to use rwlocks as opposed to the general built-in drvector lock.

Review-URL: https://codereview.appspot.com/315360043
---------------