Reviewed-by: Richard Levitte <levitte@openssl.org>

Though the callers check the function return value and ignore the
size_t output argument on failure, it is still often not ideal to
store (-1) in size_t on error.  That might signal an unduly large
buffer.  Instead set the size_t to 0, to indicate no space.

Reviewed-by: Richard Levitte <levitte@openssl.org>

Implements Thread Local Storage in the windows async port. This also has
some knock on effects to the posix and null implementations.

Reviewed-by: Rich Salz <rsalz@openssl.org>

In theory the pthreads approach for Thread Local Storage should be more
portable.

This also changes some APIs in order to accommodate this change. In
particular ASYNC_init_pool is renamed ASYNC_init_thread and
ASYNC_free_pool is renamed ASYNC_cleanup_thread. Also introduced ASYNC_init
and ASYNC_cleanup.

Reviewed-by: Rich Salz <rsalz@openssl.org>

A lot of the pool handling code was in the arch specific files, but was
actually boiler plate and the same across the implementations. This commit
moves as much code as possible out of the arch specific files.

Reviewed-by: Rich Salz <rsalz@openssl.org>

We were using _pipe to create a pipe on windows. This uses the "int" type
for its file descriptor for compatibility. However most windows functions
expect to use a "HANDLE". Probably we could get away with just casting but
it seems more robust to use the proper type and main stream windows
functions.

Reviewed-by: Rich Salz <rsalz@openssl.org>

There were a number of places where the async code did not conform to the
OpenSSL coding style.

Reviewed-by: Rich Salz <rsalz@openssl.org>

The rand code can aquire locks and then attempt crypto operations. This
can end up in a deadlock if we are using an async engine, because control
returns back to the user code whilst still holding the lock. We need to
force synchronous operation for these sections of code.

Reviewed-by: Rich Salz <rsalz@openssl.org>

There are potential deadlock situations that can occur if code executing
within the context of a job aquires a lock, and then pauses the job. This
adds an ability to temporarily block pauses from occuring whilst performing
work and holding a lock.

Reviewed-by: Rich Salz <rsalz@openssl.org>

During development some functions got added and then later taken away.
Since these will never appear in a production version there is no reason
for them to appear in libeay.num flagged as "NOEXIST".

Reviewed-by: Rich Salz <rsalz@openssl.org>

Even with _XOPEN_SOURCE defined OS-X still displays warnings that
makecontext and friends are deprecated. This isn't a problem until you
try and build with --strict-warnings, and the build fails. This change
suppresses the warnings. We know they are deprecated but there is no
alternative!

Reviewed-by: Rich Salz <rsalz@openssl.org>

async_fibre_makecontext was initialise the fibre first and then calling
getcontext(). It should be the other way around because the getcontext
call may overwrite some of the things we just initialised. This didn't
cause an issue on Linux and so the problem went unnoticed. On OS-X it
causes a crash.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Make it clear that this function is ssl specific.

Reviewed-by: Rich Salz <rsalz@openssl.org>

For some reason the dasync sha1 functions did not start with the
dasync prefix like all of the other functions do. Changed for
consistency.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Clarify that you must only call this after all async jobs have
completed - otherwise you could get memory leaks.

Reviewed-by: Rich Salz <rsalz@openssl.org>

OS-X complains if we don't have _XOPEN_SOURCE defined.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Make clang build without errors in the async code

Reviewed-by: Rich Salz <rsalz@openssl.org>

_longjmp/_setjmp do not manipulate the signal mask whilst
longjmp/setjmp may do. Online sources suggest this could result
in a significant speed up in the context switching.

Reviewed-by: Rich Salz <rsalz@openssl.org>

If config'd without -d (--debug), asynctest was crashing with:
*** longjmp causes uninitialized stack frame ***

This is because gcc will add certain checks for some functions
(including longjmp). The checks assume you can only longjmp down the
stack not up. However, if we are actually jumping to a different
fibre then it can appear as if we are going up the stack when we are
not really. This change disables the check.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add some clarifications to the async documentation. Also changed
ASYNC_pause_job() so that it returns success if you are not within the
context of a job. This is so that engines can be used either asynchronously
or synchronously and can treat an error from ASYNC_pause_job() as a real
error.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Tidy up the libssl async calls and make sure all IO functions are covered.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add ASYNCerr support to give some meaningful error message in the event of
a failure.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add a CHANGES entry for the new async code.

Reviewed-by: Rich Salz <rsalz@openssl.org>

s_server already had the ability to wait on an async file descriptor. This
adds it to s_client too.

Reviewed-by: Rich Salz <rsalz@openssl.org>

0 is a valid file descriptor so SSL_get_async_wait_fd should instead return
-1 on error.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Document the libssl and command line application aspects of async.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Tidied up the naming of functions and structures to be consistent

Reviewed-by: Rich Salz <rsalz@openssl.org>

Fix some warnings in the async code when compiling on windows.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Don't recreate a new ASYNC_CTX every time we call ASYNC_start_job() - the
same one can be used for the life of the thread. Instead we only free it
up when we call ASYNC_free_pool().

Reviewed-by: Rich Salz <rsalz@openssl.org>

The ASYNC null implementation has not kept pace with the rest of the async
development and so was failing to compile.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Add a suite of tests for the ASYNC_* functions

Reviewed-by: Rich Salz <rsalz@openssl.org>

The ASYNC_in_job() function is redundant. The same effect can be achieved by
using ASYNC_get_current_job().

Reviewed-by: Rich Salz <rsalz@openssl.org>

Reviewed-by: Rich Salz <rsalz@openssl.org>

Reviewed-by: Rich Salz <rsalz@openssl.org>

Port the async pool and notify code to windows.

Reviewed-by: Rich Salz <rsalz@openssl.org>

s_client was not freeing up the async pool if async mode was enabled.

Reviewed-by: Rich Salz <rsalz@openssl.org>

If an async event occurs during a renegotiation in SSL_read then s_server
was looping around, detecting we were in init and calling
init_ssl_connection instead of re-calling SSL_read.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Implement the ASYNC_JOB as a local thread pool. Remove the API support
for global pools.

Reviewed-by: Rich Salz <rsalz@openssl.org>

Initial API implemented for notifying applications that an ASYNC_JOB
has completed. Currently only s_server is using this. The Dummy Async
engine "cheats" in that it notifies that it has completed *before* it
pauses the job. A normal async engine would not do that.

Only the posix version of this has been implemented so far, so it will
probably fail to compile on Windows at the moment.

Reviewed-by: Rich Salz <rsalz@openssl.org>