Use a multi-word bitset to track which services are mapped to which
lcores, allowing the RTE_SERVICE_NUM_MAX compile-time constant to be >
64.
Replace array-of-bytes service-currently-active flags with a more
compact multi-word bitset-based representation, reducing memory
footprint somewhat.
Signed-off-by: Mattias Rönnblom <mattias.ronnb...@ericsson.com>
---
lib/eal/common/rte_service.c | 70 ++++++++++++++----------------------
1 file changed, 27 insertions(+), 43 deletions(-)
diff --git a/lib/eal/common/rte_service.c b/lib/eal/common/rte_service.c
index 56379930b6..ec0f47e141 100644
--- a/lib/eal/common/rte_service.c
+++ b/lib/eal/common/rte_service.c
@@ -11,6 +11,7 @@
#include <eal_trace_internal.h>
#include <rte_lcore.h>
+#include <rte_bitset.h>
#include <rte_branch_prediction.h>
#include <rte_common.h>
#include <rte_cycles.h>
@@ -63,11 +64,11 @@ struct service_stats {
/* the internal values of a service core */
struct __rte_cache_aligned core_state {
/* map of services IDs are run on this core */
- uint64_t service_mask;
+ RTE_BITSET_DECLARE(mapped_services, RTE_SERVICE_NUM_MAX);
RTE_ATOMIC(uint8_t) runstate; /* running or stopped */
RTE_ATOMIC(uint8_t) thread_active; /* indicates when thread is in
service_run() */
uint8_t is_service_core; /* set if core is currently a service core */
- uint8_t service_active_on_lcore[RTE_SERVICE_NUM_MAX];
+ RTE_BITSET_DECLARE(service_active_on_lcore, RTE_SERVICE_NUM_MAX);
RTE_ATOMIC(uint64_t) loops;
RTE_ATOMIC(uint64_t) cycles;
struct service_stats service_stats[RTE_SERVICE_NUM_MAX];
@@ -81,11 +82,6 @@ static uint32_t rte_service_library_initialized;
int32_t
rte_service_init(void)
{
- /* Hard limit due to the use of an uint64_t-based bitmask (and the
- * clzl intrinsic).
- */
- RTE_BUILD_BUG_ON(RTE_SERVICE_NUM_MAX > 64);
-
if (rte_service_library_initialized) {
EAL_LOG(NOTICE,
"service library init() called, init flag %d",
@@ -296,7 +292,7 @@ rte_service_component_unregister(uint32_t id)
/* clear the run-bit in all cores */
for (i = 0; i < RTE_MAX_LCORE; i++)
- lcore_states[i].service_mask &= ~(UINT64_C(1) << id);
+ rte_bitset_clear(lcore_states[i].mapped_services, id);
memset(&rte_services[id], 0, sizeof(struct rte_service_spec_impl));
@@ -410,7 +406,7 @@ service_runner_do_callback(struct rte_service_spec_impl *s,
/* Expects the service 's' is valid. */
static int32_t
-service_run(uint32_t i, struct core_state *cs, uint64_t service_mask,
+service_run(uint32_t i, struct core_state *cs, const uint64_t *mapped_services,
struct rte_service_spec_impl *s, uint32_t serialize_mt_unsafe)
{
if (!s)
@@ -424,12 +420,12 @@ service_run(uint32_t i, struct core_state *cs, uint64_t
service_mask,
RUNSTATE_RUNNING ||
rte_atomic_load_explicit(&s->app_runstate,
rte_memory_order_acquire) !=
RUNSTATE_RUNNING ||
- !(service_mask & (UINT64_C(1) << i))) {
- cs->service_active_on_lcore[i] = 0;
+ !rte_bitset_test(mapped_services, i)) {
+ rte_bitset_clear(cs->service_active_on_lcore, i);
return -ENOEXEC;
}
- cs->service_active_on_lcore[i] = 1;
+ rte_bitset_set(cs->service_active_on_lcore, i);
if ((service_mt_safe(s) == 0) && (serialize_mt_unsafe == 1)) {
if (!rte_spinlock_trylock(&s->execute_lock))
@@ -454,7 +450,7 @@ rte_service_may_be_active(uint32_t id)
return -EINVAL;
for (i = 0; i < lcore_count; i++) {
- if (lcore_states[ids[i]].service_active_on_lcore[id])
+ if
(rte_bitset_test(lcore_states[ids[i]].service_active_on_lcore, id))
return 1;
}
@@ -474,7 +470,9 @@ rte_service_run_iter_on_app_lcore(uint32_t id, uint32_t
serialize_mt_unsafe)
*/
rte_atomic_fetch_add_explicit(&s->num_mapped_cores, 1,
rte_memory_order_relaxed);
- int ret = service_run(id, cs, UINT64_MAX, s, serialize_mt_unsafe);
+ RTE_BITSET_DECLARE(all_services, RTE_SERVICE_NUM_MAX);
+ rte_bitset_set_all(all_services, RTE_SERVICE_NUM_MAX);
+ int ret = service_run(id, cs, all_services, s, serialize_mt_unsafe);
rte_atomic_fetch_sub_explicit(&s->num_mapped_cores, 1,
rte_memory_order_relaxed);
@@ -485,7 +483,6 @@ static int32_t
service_runner_func(void *arg)
{
RTE_SET_USED(arg);
- uint8_t i;
const int lcore = rte_lcore_id();
struct core_state *cs = &lcore_states[lcore];
@@ -497,20 +494,11 @@ service_runner_func(void *arg)
*/
while (rte_atomic_load_explicit(&cs->runstate,
rte_memory_order_acquire) ==
RUNSTATE_RUNNING) {
+ ssize_t id;
- const uint64_t service_mask = cs->service_mask;
- uint8_t start_id;
- uint8_t end_id;
-
- if (service_mask == 0)
- continue;
-
- start_id = rte_ctz64(service_mask);
- end_id = 64 - rte_clz64(service_mask);
-
- for (i = start_id; i < end_id; i++) {
+ RTE_BITSET_FOREACH_SET(id, cs->mapped_services,
RTE_SERVICE_NUM_MAX) {
/* return value ignored as no change to code flow */
- service_run(i, cs, service_mask, service_get(i), 1);
+ service_run(id, cs, cs->mapped_services,
service_get(id), 1);
}
rte_atomic_store_explicit(&cs->loops, cs->loops + 1,
rte_memory_order_relaxed);
@@ -519,8 +507,7 @@ service_runner_func(void *arg)
/* Switch off this core for all services, to ensure that future
* calls to may_be_active() know this core is switched off.
*/
- for (i = 0; i < RTE_SERVICE_NUM_MAX; i++)
- cs->service_active_on_lcore[i] = 0;
+ rte_bitset_clear_all(cs->service_active_on_lcore, RTE_SERVICE_NUM_MAX);
/* Use SEQ CST memory ordering to avoid any re-ordering around
* this store, ensuring that once this store is visible, the service
@@ -586,7 +573,7 @@ rte_service_lcore_count_services(uint32_t lcore)
if (!cs->is_service_core)
return -ENOTSUP;
- return rte_popcount64(cs->service_mask);
+ return rte_bitset_count_set(cs->mapped_services, RTE_SERVICE_NUM_MAX);
}
int32_t
@@ -639,25 +626,23 @@ service_update(uint32_t sid, uint32_t lcore, uint32_t
*set, uint32_t *enabled)
!lcore_states[lcore].is_service_core)
return -EINVAL;
- uint64_t sid_mask = UINT64_C(1) << sid;
if (set) {
- uint64_t lcore_mapped = lcore_states[lcore].service_mask &
- sid_mask;
+ uint64_t lcore_mapped =
rte_bitset_test(lcore_states[lcore].mapped_services, sid);
if (*set && !lcore_mapped) {
- lcore_states[lcore].service_mask |= sid_mask;
+ rte_bitset_set(lcore_states[lcore].mapped_services,
sid);
rte_atomic_fetch_add_explicit(&rte_services[sid].num_mapped_cores,
1, rte_memory_order_relaxed);
}
if (!*set && lcore_mapped) {
- lcore_states[lcore].service_mask &= ~(sid_mask);
+ rte_bitset_clear(lcore_states[lcore].mapped_services,
sid);
rte_atomic_fetch_sub_explicit(&rte_services[sid].num_mapped_cores,
1, rte_memory_order_relaxed);
}
}
if (enabled)
- *enabled = !!(lcore_states[lcore].service_mask & (sid_mask));
+ *enabled = rte_bitset_test(lcore_states[lcore].mapped_services,
sid);
return 0;
}
@@ -699,11 +684,11 @@ set_lcore_state(uint32_t lcore, int32_t state)
int32_t
rte_service_lcore_reset_all(void)
{
- /* loop over cores, reset all to mask 0 */
+ /* loop over cores, reset all mapped services */
uint32_t i;
for (i = 0; i < RTE_MAX_LCORE; i++) {
if (lcore_states[i].is_service_core) {
- lcore_states[i].service_mask = 0;
+ rte_bitset_clear_all(lcore_states[i].mapped_services,
RTE_SERVICE_NUM_MAX);
set_lcore_state(i, ROLE_RTE);
/* runstate act as guard variable Use
* store-release memory order here to synchronize
@@ -731,7 +716,7 @@ rte_service_lcore_add(uint32_t lcore)
set_lcore_state(lcore, ROLE_SERVICE);
/* ensure that after adding a core the mask and state are defaults */
- lcore_states[lcore].service_mask = 0;
+ rte_bitset_clear_all(lcore_states[lcore].mapped_services,
RTE_SERVICE_NUM_MAX);
/* Use store-release memory order here to synchronize with
* load-acquire in runstate read functions.
*/
@@ -814,12 +799,11 @@ rte_service_lcore_stop(uint32_t lcore)
uint32_t i;
struct core_state *cs = &lcore_states[lcore];
- uint64_t service_mask = cs->service_mask;
for (i = 0; i < RTE_SERVICE_NUM_MAX; i++) {
- int32_t enabled = service_mask & (UINT64_C(1) << i);
- int32_t service_running = rte_service_runstate_get(i);
- int32_t only_core = (1 ==
+ bool enabled = rte_bitset_test(cs->mapped_services, i);
+ bool service_running = rte_service_runstate_get(i);
+ bool only_core = (1 ==
rte_atomic_load_explicit(&rte_services[i].num_mapped_cores,
rte_memory_order_relaxed));
--
2.34.1
