i#1982 bbbuf.c: Fixes incorrect wrapping

Commit log for first patchset:
---------------
i#1982 bbbuf.c: Fixes incorrect wrapping

Fixes the incorrect wrapping on the buffer by bbbuf.c. Previously it
used a uqadd16, which saturates instead of wraps, resulting in the
buffer never wrapping. We now use drx_buf which optimizes for buffers of
16KB to produce the same instrumentation.

Also optimizes drx_buf to fast-fail any signal handlers if only the
optimized circular buffer has been created.

Fixes #1982
---------------

[bruening, 2017-01-11 17:18:25]

LGTM w/ suggestions for more comments and some unused code removal

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c
File api/samples/bbbuf.c (right):

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode54
api/samples/bbbuf.c:54: #define MINSERT instrlist_meta_preinsert
This can be removed?

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode56
api/samples/bbbuf.c:56: static drx_buf_t *buf;
suggest: add a comment explaining that we use drx_buf's circular buffer which
makes our work easy (this is a sample, good to have extra comments)

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode58
api/samples/bbbuf.c:58: #define BUF_64K_BYTE (1 << 16)
This can be removed, right?

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode66
api/samples/bbbuf.c:66: /* We need a 2nd scratch reg for several operations */
Update the comment to say that we only need this for ARM

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode80
api/samples/bbbuf.c:80: if (drreg_reserve_register(drcontext, bb, inst, NULL,
&reg2) != DRREG_SUCCESS) {
suggest: maybe also have a comment here saying that the drx_buf routines need a
scratch reg for ARM

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode95
api/samples/bbbuf.c:95: * 16 bits of the pointer
suggest: prefer capitalized and punctuated for all but single-line comments: so
"Internally ... pointer."

[toshi, 2017-01-11 18:46:12]

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c
File api/samples/bbbuf.c (right):

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode54
api/samples/bbbuf.c:54: #define MINSERT instrlist_meta_preinsert
On 2017/01/11 17:18:25, bruening wrote:
> This can be removed?

Yup, done.

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode56
api/samples/bbbuf.c:56: static drx_buf_t *buf;
On 2017/01/11 17:18:25, bruening wrote:
> suggest: add a comment explaining that we use drx_buf's circular buffer which
> makes our work easy (this is a sample, good to have extra comments)

Done.

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode58
api/samples/bbbuf.c:58: #define BUF_64K_BYTE (1 << 16)
On 2017/01/11 17:18:25, bruening wrote:
> This can be removed, right?

Done.

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode66
api/samples/bbbuf.c:66: /* We need a 2nd scratch reg for several operations */
On 2017/01/11 17:18:25, bruening wrote:
> Update the comment to say that we only need this for ARM

Done.

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode80
api/samples/bbbuf.c:80: if (drreg_reserve_register(drcontext, bb, inst, NULL,
&reg2) != DRREG_SUCCESS) {
On 2017/01/11 17:18:25, bruening wrote:
> suggest: maybe also have a comment here saying that the drx_buf routines need
a
> scratch reg for ARM

Done.

https://codereview.appspot.com/317080043/diff/1/api/samples/bbbuf.c#newcode95
api/samples/bbbuf.c:95: * 16 bits of the pointer
On 2017/01/11 17:18:25, bruening wrote:
> suggest: prefer capitalized and punctuated for all but single-line comments:
so
> "Internally ... pointer."

Done.

[toshi, 2017-01-11 19:05:11]

Committed as
https://github.com/DynamoRIO/dynamorio/commit/2380ed9e72b56acd07b51956d3b201c...

Final commit log: 
---------------
i#1982 bbbuf.c: Fixes incorrect wrapping

Fixes the incorrect wrapping on the buffer by bbbuf.c. Previously it
used a uqadd16, which saturates instead of wraps, resulting in the
buffer never wrapping. We now use drx_buf which optimizes for buffers of
16KB to produce the same instrumentation.

Also optimizes drx_buf to fast-fail any signal handlers if only the
optimized circular buffer has been created.

Fixes #1982

Review-URL: https://codereview.appspot.com/317080043
---------------