rsyslog/runtime/atomic.h
Rainer Gerhards 18b12d9f6e runtime: harden atomic helper fallbacks
Why:
The atomic helper header had minor cleanup issues and two no-atomics
fallbacks did not match the contracts of the GCC builtins they replace.
Keeping these semantics aligned matters for portability builds.

Impact:
No behavior change when compiler atomic builtins are available. Builds
without atomic builtins now get matching OR and CAS return semantics.

Before/After:
Before, the fallback OR helper assigned the value, and CAS_VAL returned
the value after the operation. After, they match the builtin behavior.

Technical Overview:
Fix comment and URL typos left in runtime/atomic.h and remove a stray
semicolon from the ATOMIC_CAS_VAL builtin macro. Parenthesize data in the
atomic CAS builtin wrappers for macro consistency. In the mutex fallback,
make ATOMIC_OR_INT_TO_INT apply bitwise OR and make ATOMIC_CAS_VAL save
and return the previous value before any successful replacement.

With the help of AI-Agents: Codex
2026-05-11 20:26:02 +02:00

327 lines
13 KiB
C

/* This header supplies atomic operations. So far, we rely on GCC's
* atomic builtins. During configure, we check if atomic operations are
* available. If they are not, I am making the necessary provisioning to live without them if
* they are not available. Please note that you should only use the macros
* here if you think you can actually live WITHOUT an explicit atomic operation,
* because in the non-presence of them, we simply do it without atomicity.
* Which, for word-aligned data types, usually (but only usually!) should work.
*
* We are using the functions described in
* http://gcc.gnu.org/onlinedocs/gcc/Atomic-Builtins.html
*
* THESE MACROS MUST ONLY BE USED WITH WORD-SIZED DATA TYPES!
*
* Usage guidance (important for portability and static analysis tools):
* - Operations are provided as macros or inline functions. Prefer simple lvalues
* for the "data" argument to avoid double-evaluation surprises.
* - The "phlpmut" argument is a helper mutex used when atomics are unavailable.
* If atomics are supported, these helper mutex macros become no-ops.
* Callers must still pass a valid pointer where required to keep signatures
* uniform across platforms.
* - Some macros are "best effort" (PREFER_*). They may degrade to non-atomic
* operations when no atomic support is available. Use them only for
* statistics or counters where occasional lost updates are acceptable.
* - The remaining ATOMIC_* helpers provide true atomicity when available and
* fall back to mutex-based critical sections when not.
* - Mixing atomic loads/stores with plain reads/writes to the same variable is
* undefined on modern compilers. If you use ATOMIC_* on a variable, access it
* exclusively via these helpers (or other synchronized paths).
*
* Copyright 2008-2012 Rainer Gerhards and Adiscon GmbH.
*
* This file is part of the rsyslog runtime library.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
* -or-
* see COPYING.ASL20 in the source distribution
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef INCLUDED_ATOMIC_H
#define INCLUDED_ATOMIC_H
#include <time.h>
#include "typedefs.h"
/* for this release, we disable atomic calls because there seem to be some
* portability problems and we can not fix that without destabilizing the build.
* They simply came in too late. -- rgerhards, 2008-04-02
*/
#ifdef HAVE_ATOMIC_BUILTINS
#define ATOMIC_SUB(data, val, phlpmut) __sync_fetch_and_sub(data, val)
#define ATOMIC_SUB_unsigned(data, val, phlpmut) __sync_fetch_and_sub(data, val)
#define ATOMIC_ADD(data, val) __sync_fetch_and_add(&(data), val)
#define ATOMIC_INC(data, phlpmut) ((void)__sync_fetch_and_add(data, 1))
#define ATOMIC_INC_AND_FETCH_int(data, phlpmut) __sync_fetch_and_add(data, 1)
#define ATOMIC_INC_AND_FETCH_unsigned(data, phlpmut) __sync_fetch_and_add(data, 1)
#define ATOMIC_DEC(data, phlpmut) ((void)__sync_sub_and_fetch(data, 1))
#define ATOMIC_DEC_AND_FETCH(data, phlpmut) __sync_sub_and_fetch(data, 1)
#define ATOMIC_FETCH_32BIT(data, phlpmut) ((int)__sync_fetch_and_and(data, 0xffffffff))
#define ATOMIC_FETCH_32BIT_unsigned(data, phlpmut) ((unsigned)__sync_fetch_and_and(data, 0xffffffff))
#define ATOMIC_STORE_1_TO_32BIT(data) __sync_lock_test_and_set(&(data), 1)
#define ATOMIC_STORE_0_TO_INT(data, phlpmut) __sync_fetch_and_and(data, 0)
#define ATOMIC_STORE_1_TO_INT(data, phlpmut) __sync_fetch_and_or(data, 1)
#define ATOMIC_STORE_INT(data, phlpmut, val) ((void)__sync_lock_test_and_set((data), (val)))
#define ATOMIC_OR_INT_TO_INT(data, phlpmut, val) __sync_fetch_and_or((data), (val))
#define ATOMIC_CAS(data, oldVal, newVal, phlpmut) __sync_bool_compare_and_swap((data), (oldVal), (newVal))
#define ATOMIC_CAS_time_t(data, oldVal, newVal, phlpmut) __sync_bool_compare_and_swap((data), (oldVal), (newVal))
#define ATOMIC_CAS_VAL(data, oldVal, newVal, phlpmut) __sync_val_compare_and_swap((data), (oldVal), (newVal))
/* Atomic operations for pointers (word-sized on all supported platforms) */
#define ATOMIC_FETCH_PTR(data, phlpmut) ((void *)__sync_fetch_and_add(data, 0))
#define ATOMIC_STORE_PTR(data, phlpmut, val) ((void)__sync_lock_test_and_set((data), (val)))
#define ATOMIC_CAS_PTR(data, oldVal, newVal, phlpmut) __sync_bool_compare_and_swap(data, (oldVal), (newVal))
/* functions below are not needed if we have atomics */
#define DEF_ATOMIC_HELPER_MUT(x)
#define INIT_ATOMIC_HELPER_MUT(x)
#define DESTROY_ATOMIC_HELPER_MUT(x)
/* the following operations should preferably be done atomically, but it is
* not fatal if not -- that means we can live with some missed updates. So be
* sure to use these macros only if that really does not matter!
*/
#define PREFER_ATOMIC_INC(data) ((void)__sync_fetch_and_add(&(data), 1))
#define PREFER_FETCH_32BIT(data) ((unsigned)__sync_fetch_and_and(&(data), 0xffffffff))
#define PREFER_STORE_0_TO_INT(data) __sync_fetch_and_and(data, 0)
#define PREFER_STORE_1_TO_INT(data) __sync_fetch_and_or(data, 1)
#else
/* note that we gained practical proof that theoretical problems DO occur
* if we do not properly address them. See this blog post for details:
* http://blog.gerhards.net/2009/01/rsyslog-data-race-analysis.html
* The bottom line is that if there are no atomics available, we should NOT
* simply go ahead and do without them - use mutexes or other things. The
* code needs to be checked against all those cases. -- rgerhards, 2009-01-30
*/
#include <pthread.h>
#define ATOMIC_INC(data, phlpmut) \
{ \
pthread_mutex_lock(phlpmut); \
++(*(data)); \
pthread_mutex_unlock(phlpmut); \
}
#define ATOMIC_STORE_0_TO_INT(data, hlpmut) \
{ \
pthread_mutex_lock(hlpmut); \
*(data) = 0; \
pthread_mutex_unlock(hlpmut); \
}
#define ATOMIC_STORE_1_TO_INT(data, hlpmut) \
{ \
pthread_mutex_lock(hlpmut); \
*(data) = 1; \
pthread_mutex_unlock(hlpmut); \
}
#define ATOMIC_STORE_INT(data, hlpmut, val) \
do { \
pthread_mutex_lock(hlpmut); \
*(data) = (val); \
pthread_mutex_unlock(hlpmut); \
} while (0)
#define ATOMIC_OR_INT_TO_INT(data, hlpmut, val) \
{ \
pthread_mutex_lock(hlpmut); \
*(data) |= (val); \
pthread_mutex_unlock(hlpmut); \
}
static inline int ATOMIC_CAS(int *data, int oldVal, int newVal, pthread_mutex_t *phlpmut) {
int bSuccess;
pthread_mutex_lock(phlpmut);
if (*data == oldVal) {
*data = newVal;
bSuccess = 1;
} else {
bSuccess = 0;
}
pthread_mutex_unlock(phlpmut);
return (bSuccess);
}
static inline int ATOMIC_CAS_time_t(time_t *data, time_t oldVal, time_t newVal, pthread_mutex_t *phlpmut) {
int bSuccess;
pthread_mutex_lock(phlpmut);
if (*data == oldVal) {
*data = newVal;
bSuccess = 1;
} else {
bSuccess = 0;
}
pthread_mutex_unlock(phlpmut);
return (bSuccess);
}
static inline int ATOMIC_CAS_VAL(int *data, int oldVal, int newVal, pthread_mutex_t *phlpmut) {
int val;
pthread_mutex_lock(phlpmut);
val = *data;
if (*data == oldVal) {
*data = newVal;
}
pthread_mutex_unlock(phlpmut);
return (val);
}
#define ATOMIC_DEC(data, phlpmut) \
{ \
pthread_mutex_lock(phlpmut); \
--(*(data)); \
pthread_mutex_unlock(phlpmut); \
}
static inline int ATOMIC_INC_AND_FETCH_int(int *data, pthread_mutex_t *phlpmut) {
int val;
pthread_mutex_lock(phlpmut);
val = ++(*data);
pthread_mutex_unlock(phlpmut);
return (val);
}
static inline unsigned ATOMIC_INC_AND_FETCH_unsigned(unsigned *data, pthread_mutex_t *phlpmut) {
unsigned val;
pthread_mutex_lock(phlpmut);
val = ++(*data);
pthread_mutex_unlock(phlpmut);
return (val);
}
static inline int ATOMIC_DEC_AND_FETCH(int *data, pthread_mutex_t *phlpmut) {
int val;
pthread_mutex_lock(phlpmut);
val = --(*data);
pthread_mutex_unlock(phlpmut);
return (val);
}
static inline int ATOMIC_FETCH_32BIT(int *data, pthread_mutex_t *phlpmut) {
int val;
pthread_mutex_lock(phlpmut);
val = (*data);
pthread_mutex_unlock(phlpmut);
return (val);
}
static inline int ATOMIC_FETCH_32BIT_unsigned(unsigned *data, pthread_mutex_t *phlpmut) {
int val;
pthread_mutex_lock(phlpmut);
val = (*data);
pthread_mutex_unlock(phlpmut);
return (val);
}
static inline void ATOMIC_SUB(int *data, int val, pthread_mutex_t *phlpmut) {
pthread_mutex_lock(phlpmut);
(*data) -= val;
pthread_mutex_unlock(phlpmut);
}
static inline void ATOMIC_SUB_unsigned(unsigned *data, int val, pthread_mutex_t *phlpmut) {
pthread_mutex_lock(phlpmut);
(*data) -= val;
pthread_mutex_unlock(phlpmut);
}
/* Atomic operations for pointers - fallback to mutex-protected operations */
static inline void *ATOMIC_FETCH_PTR(void **data, pthread_mutex_t *phlpmut) {
void *val;
pthread_mutex_lock(phlpmut);
val = *data;
pthread_mutex_unlock(phlpmut);
return val;
}
static inline void ATOMIC_STORE_PTR(void **data, pthread_mutex_t *phlpmut, void *val) {
pthread_mutex_lock(phlpmut);
*data = val;
pthread_mutex_unlock(phlpmut);
}
static inline int ATOMIC_CAS_PTR(void **data, void *oldVal, void *newVal, pthread_mutex_t *phlpmut) {
int bSuccess;
pthread_mutex_lock(phlpmut);
if (*data == oldVal) {
*data = newVal;
bSuccess = 1;
} else {
bSuccess = 0;
}
pthread_mutex_unlock(phlpmut);
return bSuccess;
}
#define DEF_ATOMIC_HELPER_MUT(x) pthread_mutex_t x
#define INIT_ATOMIC_HELPER_MUT(x) pthread_mutex_init(&(x), NULL)
#define DESTROY_ATOMIC_HELPER_MUT(x) pthread_mutex_destroy(&(x))
#define PREFER_ATOMIC_INC(data) ((void)++data)
#define PREFER_FETCH_32BIT(data) ((unsigned)(data))
#define PREFER_STORE_0_TO_INT(data) (*(data) = 0)
#define PREFER_STORE_1_TO_INT(data) (*(data) = 1)
#endif
/* we need to handle 64bit atomics seperately as some platforms have
* 32 bit atomics, but not 64 bit ones... -- rgerhards, 2010-12-01
*/
#ifdef HAVE_ATOMIC_BUILTINS64
#define ATOMIC_INC_uint64(data, phlpmut) ((void)__sync_fetch_and_add(data, 1))
#define ATOMIC_ADD_uint64(data, phlpmut, value) ((void)__sync_fetch_and_add(data, value))
#define ATOMIC_DEC_uint64(data, phlpmut) ((void)__sync_sub_and_fetch(data, 1))
#define ATOMIC_INC_AND_FETCH_uint64(data, phlpmut) __sync_fetch_and_add(data, 1)
#define ATOMIC_STORE_uint64(data, phlpmut, value) ((void)__sync_lock_test_and_set((data), (value)))
#define DEF_ATOMIC_HELPER_MUT64(x)
#define INIT_ATOMIC_HELPER_MUT64(x)
#define DESTROY_ATOMIC_HELPER_MUT64(x)
#else
#define ATOMIC_INC_uint64(data, phlpmut) \
{ \
pthread_mutex_lock(phlpmut); \
++(*(data)); \
pthread_mutex_unlock(phlpmut); \
}
#define ATOMIC_ADD_uint64(data, phlpmut, value) \
{ \
pthread_mutex_lock(phlpmut); \
*data += value; \
pthread_mutex_unlock(phlpmut); \
}
#define ATOMIC_DEC_uint64(data, phlpmut) \
{ \
pthread_mutex_lock(phlpmut); \
--(*(data)); \
pthread_mutex_unlock(phlpmut); \
}
#define ATOMIC_STORE_uint64(data, phlpmut, value) \
{ \
pthread_mutex_lock(phlpmut); \
*(data) = (value); \
pthread_mutex_unlock(phlpmut); \
}
static inline unsigned ATOMIC_INC_AND_FETCH_uint64(uint64 *data, pthread_mutex_t *phlpmut) {
uint64 val;
pthread_mutex_lock(phlpmut);
val = ++(*data);
pthread_mutex_unlock(phlpmut);
return (val);
}
#define DEF_ATOMIC_HELPER_MUT64(x) pthread_mutex_t x
#define INIT_ATOMIC_HELPER_MUT64(x) pthread_mutex_init(&(x), NULL)
#define DESTROY_ATOMIC_HELPER_MUT64(x) pthread_mutex_destroy(&(x))
#endif /* #ifdef HAVE_ATOMIC_BUILTINS64 */
#endif /* #ifndef INCLUDED_ATOMIC_H */