mirror/openssl
2
0
Fork 0
mirror of https://github.com/openssl/openssl.git synced 2024-12-27 06:21:43 +08:00
Code Issues Packages Projects Releases Wiki Activity
dab96a4f60
openssl/doc/man3/CRYPTO_THREAD_run_once.pod
Neil Horman cc4ea5e000 Introduce new internal hashtable implementation 
Create a new hashtable that is more efficient than the existing LHASH_OF
implementation.  the new ossl_ht api offers several new features that
improve performance opportunistically

* A more generalized hash function.  Currently using fnv1a, provides a
  more general hash function, but can still be overridden where needed

* Improved locking and reference counting.  This hash table is
  internally locked with an RCU lock, and optionally reference counts
  elements, allowing for users to not have to create and manage their
  own read/write locks

* Lockless operation.  The hash table can be configured to operate
  locklessly on the read side, improving performance, at the sacrifice
  of the ability to grow the hash table or delete elements from it

* A filter function allowing for the retrieval of several elements at a
  time matching a given criteria without having to hold a lock
  permanently

* a doall_until iterator variant, that allows callers which need to
  iterate over the entire hash table until a given condition is met (as
  defined by the return value of the iterator callback).  This allows
  for callers attempting to do expensive cache searches for a small
  number of elements to terminate the iteration early, saving cpu cycles

* Dynamic type safety.  The hash table provides operations to set and
  get data of a specific type without having to define a type at the
  instatiation point

* Multiple data type storage.  The hash table can store multiple data
  types allowing for more flexible usage

* Ubsan safety.  Because the API deals with concrete single types
  (HT_KEY and HT_VALUE), leaving specific type casting to the call
  recipient with dynamic type validation, this implementation is safe
  from the ubsan undefined behavior warnings that require additional
  thunking on callbacks.

Testing of this new hashtable with an equivalent hash function, I can
observe approximately a 6% performance improvement in the lhash_test

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/23671)
2024-04-24 12:03:30 +10:00

255 lines
8.5 KiB
Plaintext
Raw Blame History

=pod
=head1 NAME
CRYPTO_THREAD_run_once,
CRYPTO_THREAD_lock_new, CRYPTO_THREAD_read_lock, CRYPTO_THREAD_write_lock,
CRYPTO_THREAD_unlock, CRYPTO_THREAD_lock_free,
CRYPTO_atomic_add, CRYPTO_atomic_or, CRYPTO_atomic_load, CRYPTO_atomic_store,
CRYPTO_atomic_load_int,
OSSL_set_max_threads, OSSL_get_max_threads,
OSSL_get_thread_support_flags, OSSL_THREAD_SUPPORT_FLAG_THREAD_POOL,
OSSL_THREAD_SUPPORT_FLAG_DEFAULT_SPAWN - OpenSSL thread support
=head1 SYNOPSIS
#include <openssl/crypto.h>
CRYPTO_ONCE CRYPTO_ONCE_STATIC_INIT;
int CRYPTO_THREAD_run_once(CRYPTO_ONCE *once, void (*init)(void));
CRYPTO_RWLOCK *CRYPTO_THREAD_lock_new(void);
int CRYPTO_THREAD_read_lock(CRYPTO_RWLOCK *lock);
int CRYPTO_THREAD_write_lock(CRYPTO_RWLOCK *lock);
int CRYPTO_THREAD_unlock(CRYPTO_RWLOCK *lock);
void CRYPTO_THREAD_lock_free(CRYPTO_RWLOCK *lock);
int CRYPTO_atomic_add(int *val, int amount, int *ret, CRYPTO_RWLOCK *lock);
int CRYPTO_atomic_or(uint64_t *val, uint64_t op, uint64_t *ret,
CRYPTO_RWLOCK *lock);
int CRYPTO_atomic_load(uint64_t *val, uint64_t *ret, CRYPTO_RWLOCK *lock);
int CRYPTO_atomic_store(uint64_t *dst, uint64_t val, CRYPTO_RWLOCK *lock);
int CRYPTO_atomic_load_int(int *val, int *ret, CRYPTO_RWLOCK *lock);
int OSSL_set_max_threads(OSSL_LIB_CTX *ctx, uint64_t max_threads);
uint64_t OSSL_get_max_threads(OSSL_LIB_CTX *ctx);
uint32_t OSSL_get_thread_support_flags(void);
#define OSSL_THREAD_SUPPORT_FLAG_THREAD_POOL
#define OSSL_THREAD_SUPPORT_FLAG_DEFAULT_SPAWN
=head1 DESCRIPTION
OpenSSL can be safely used in multi-threaded applications provided that
support for the underlying OS threading API is built-in. Currently, OpenSSL
supports the pthread and Windows APIs. OpenSSL can also be built without
any multi-threading support, for example on platforms that don't provide
any threading support or that provide a threading API that is not yet
supported by OpenSSL.
The following multi-threading function are provided:
=over 2
=item *
CRYPTO_THREAD_run_once() can be used to perform one-time initialization.
The I<once> argument must be a pointer to a static object of type
B<CRYPTO_ONCE> that was statically initialized to the value
B<CRYPTO_ONCE_STATIC_INIT>.
The I<init> argument is a pointer to a function that performs the desired
exactly once initialization.
In particular, this can be used to allocate locks in a thread-safe manner,
which can then be used with the locking functions below.
=item *
CRYPTO_THREAD_lock_new() allocates, initializes and returns a new read/write
lock.
=item *
CRYPTO_THREAD_read_lock() locks the provided I<lock> for reading.
=item *
CRYPTO_THREAD_write_lock() locks the provided I<lock> for writing.
=item *
CRYPTO_THREAD_unlock() unlocks the previously locked I<lock>.
=item *
CRYPTO_THREAD_lock_free() frees the provided I<lock>.
=item *
CRYPTO_atomic_add() atomically adds I<amount> to I<*val> and returns the
result of the operation in I<*ret>. I<lock> will be locked, unless atomic
operations are supported on the specific platform. Because of this, if a
variable is modified by CRYPTO_atomic_add() then CRYPTO_atomic_add() must
be the only way that the variable is modified. If atomic operations are not
supported and I<lock> is NULL, then the function will fail.
=item *
CRYPTO_atomic_or() performs an atomic bitwise or of I<op> and I<*val> and stores
the result back in I<*val>. It also returns the result of the operation in
I<*ret>. I<lock> will be locked, unless atomic operations are supported on the
specific platform. Because of this, if a variable is modified by
CRYPTO_atomic_or() or read by CRYPTO_atomic_load() then CRYPTO_atomic_or() must
be the only way that the variable is modified. If atomic operations are not
supported and I<lock> is NULL, then the function will fail.
=item *
CRYPTO_atomic_load() atomically loads the contents of I<*val> into I<*ret>.
I<lock> will be locked, unless atomic operations are supported on the specific
platform. Because of this, if a variable is modified by CRYPTO_atomic_or() or
read by CRYPTO_atomic_load() then CRYPTO_atomic_load() must be the only way that
the variable is read. If atomic operations are not supported and I<lock> is
NULL, then the function will fail.
=item *
CRYPTO_atomic_store() atomically stores the contents of I<val> into I<*dst>.
I<lock> will be locked, unless atomic operations are supported on the specific
plaform.
=item *
CRYPTO_atomic_load_int() works identically to CRYPTO_atomic_load() but operates
on an I<int> value instead of a I<uint64_t> value.
=item *
OSSL_set_max_threads() sets the maximum number of threads to be used by the
thread pool. If the argument is 0, thread pooling is disabled. OpenSSL will
not create any threads and existing threads in the thread pool will be torn
down. The maximum thread count is a limit, not a target. Threads will not be
spawned unless (and until) there is demand. Thread polling is disabled by
default. To enable threading you must call OSSL_set_max_threads() explicitly.
Under no circumstances is this done for you.
=item *
OSSL_get_thread_support_flags() determines what thread pool functionality
OpenSSL is compiled with and is able to support in the current run time
environment. B<OSSL_THREAD_SUPPORT_FLAG_THREAD_POOL> indicates that the base
thread pool functionality is available, and
B<OSSL_THREAD_SUPPORT_FLAG_DEFAULT_SPAWN> indicates that the default thread pool
model is available. The default thread pool model is currently the only model
available, therefore both of these flags must be set for thread pool
functionality to be used.
=back
=head1 RETURN VALUES
CRYPTO_THREAD_run_once() returns 1 on success, or 0 on error.
CRYPTO_THREAD_lock_new() returns the allocated lock, or NULL on error.
CRYPTO_THREAD_lock_free() returns no value.
OSSL_set_max_threads() returns 1 on success and 0 on failure. Returns failure
if OpenSSL-managed thread pooling is not supported (for example, if it is not
supported on the current platform, or because OpenSSL is not built with the
necessary support).
OSSL_get_max_threads() returns the maximum number of threads currently allowed
to be used by the thread pool. If thread pooling is disabled or not available,
returns 0.
OSSL_get_thread_support_flags() returns zero or more B<OSSL_THREAD_SUPPORT_FLAG>
values.
The other functions return 1 on success, or 0 on error.
=head1 NOTES
On Windows platforms the CRYPTO_THREAD_* types and functions in the
F<< <openssl/crypto.h> >> header are dependent on some of the types
customarily made available by including F<< <windows.h> >>. The application
developer is likely to require control over when the latter is included,
commonly as one of the first included headers. Therefore, it is defined as an
application developer's responsibility to include F<< <windows.h> >> prior to
F<< <openssl/crypto.h> >> where use of CRYPTO_THREAD_* types and functions is
required.
=head1 EXAMPLES
You can find out if OpenSSL was configured with thread support:
#include <openssl/opensslconf.h>
#if defined(OPENSSL_THREADS)
/* thread support enabled */
#else
/* no thread support */
#endif
This example safely initializes and uses a lock.
#ifdef _WIN32
# include <windows.h>
#endif
#include <openssl/crypto.h>
static CRYPTO_ONCE once = CRYPTO_ONCE_STATIC_INIT;
static CRYPTO_RWLOCK *lock;
static void myinit(void)
{
lock = CRYPTO_THREAD_lock_new();
}
static int mylock(void)
{
if (!CRYPTO_THREAD_run_once(&once, void init) || lock == NULL)
return 0;
return CRYPTO_THREAD_write_lock(lock);
}
static int myunlock(void)
{
return CRYPTO_THREAD_unlock(lock);
}
int serialized(void)
{
int ret = 0;
if (mylock()) {
/* Your code here, do not return without releasing the lock! */
ret = ... ;
}
myunlock();
return ret;
}
Finalization of locks is an advanced topic, not covered in this example.
This can only be done at process exit or when a dynamically loaded library is
no longer in use and is unloaded.
The simplest solution is to just "leak" the lock in applications and not
repeatedly load/unload shared libraries that allocate locks.
=head1 SEE ALSO
L<crypto(7)>, L<openssl-threads(7)>.
=head1 HISTORY
CRYPTO_atomic_store() was added in OpenSSL 3.4.0
=head1 COPYRIGHT
Copyright 2000-2023 The OpenSSL Project Authors. All Rights Reserved.
Licensed under the Apache License 2.0 (the "License"). You may not use
this file except in compliance with the License. You can obtain a copy
in the file LICENSE in the source distribution or at
L<https://www.openssl.org/source/license.html>.
=cut
Reference in New Issue View Git Blame Copy Permalink