ivampiresp/Spdk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

example/nvmf: Replace locked linked list by lockless ring for run queue

Browse Source 
Replace mutex and linked list by lockless ring for nvmf_reactor->threads.

If we create event call, we will be able to keep linked list and
remove only mutex. However, simple replacement of locked linked list
by lockless ring will be enough for now.

Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I590a385ff018ad20c7d6752b0a3a5562e3b68510
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2579
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This commit is contained in:
Shuhei Matsumoto 2020-05-25 06:32:14 +09:00 committed by Tomasz Zawadzki
parent 10190302d2
commit ab6dbff72a
1 changed files with 16 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
examples/nvmf/nvmf/nvmf.c
View File

@ -69,9 +69,8 @@ struct nvmf_lw_thread {
struct nvmf_reactor {
uint32_t core;
pthread_mutex_t mutex;
TAILQ_HEAD(, nvmf_lw_thread) threads;
struct spdk_ring *threads;
TAILQ_ENTRY(nvmf_reactor) link;
};
@ -185,17 +184,12 @@ static int
nvmf_reactor_run(void *arg)
{
struct nvmf_reactor *nvmf_reactor = arg;
struct nvmf_lw_thread *lw_thread, *tmp;
struct nvmf_lw_thread *lw_thread;
struct spdk_thread *thread;
/* run all the lightweight threads in this nvmf_reactor by FIFO. */
do {
pthread_mutex_lock(&nvmf_reactor->mutex);
lw_thread = TAILQ_FIRST(&nvmf_reactor->threads);
if (lw_thread != NULL) {
TAILQ_REMOVE(&nvmf_reactor->threads, lw_thread, link);
pthread_mutex_unlock(&nvmf_reactor->mutex);
if (spdk_ring_dequeue(nvmf_reactor->threads, (void **)&lw_thread, 1)) {
thread = spdk_thread_get_from_ctx(lw_thread);
spdk_thread_poll(thread, 0, 0);
@ -203,21 +197,15 @@ nvmf_reactor_run(void *arg)
if (spdk_unlikely(spdk_thread_is_exited(thread) &&
spdk_thread_is_idle(thread))) {
spdk_thread_destroy(thread);
pthread_mutex_lock(&nvmf_reactor->mutex);
} else {
pthread_mutex_lock(&nvmf_reactor->mutex);
TAILQ_INSERT_TAIL(&nvmf_reactor->threads, lw_thread, link);
spdk_ring_enqueue(nvmf_reactor->threads, (void **)&lw_thread, 1, NULL);
}
}
pthread_mutex_unlock(&nvmf_reactor->mutex);
} while (!g_reactors_exit);
/* free all the lightweight threads */
pthread_mutex_lock(&nvmf_reactor->mutex);
TAILQ_FOREACH_SAFE(lw_thread, &nvmf_reactor->threads, link, tmp) {
while (spdk_ring_dequeue(nvmf_reactor->threads, (void **)&lw_thread, 1)) {
thread = spdk_thread_get_from_ctx(lw_thread);
TAILQ_REMOVE(&nvmf_reactor->threads, lw_thread, link);
spdk_set_thread(thread);
spdk_thread_exit(thread);
while (!spdk_thread_is_exited(thread)) {
@ -225,7 +213,6 @@ nvmf_reactor_run(void *arg)
}
spdk_thread_destroy(thread);
}
pthread_mutex_unlock(&nvmf_reactor->mutex);
return 0;
}
@ -263,9 +250,7 @@ nvmf_schedule_spdk_thread(struct spdk_thread *thread)
/* each spdk_thread has the core affinity */
if (spdk_cpuset_get_cpu(cpumask, nvmf_reactor->core)) {
pthread_mutex_lock(&nvmf_reactor->mutex);
TAILQ_INSERT_TAIL(&nvmf_reactor->threads, lw_thread, link);
pthread_mutex_unlock(&nvmf_reactor->mutex);
spdk_ring_enqueue(nvmf_reactor->threads, (void **)&lw_thread, 1, NULL);
break;
}
}
@ -338,8 +323,15 @@ nvmf_init_threads(void)
}
nvmf_reactor->core = i;
pthread_mutex_init(&nvmf_reactor->mutex, NULL);
TAILQ_INIT(&nvmf_reactor->threads);
nvmf_reactor->threads = spdk_ring_create(SPDK_RING_TYPE_MP_SC, 1024, SPDK_ENV_SOCKET_ID_ANY);
if (!nvmf_reactor->threads) {
fprintf(stderr, "failed to alloc ring\n");
free(nvmf_reactor);
rc = -ENOMEM;
goto err_exit;
}
TAILQ_INSERT_TAIL(&g_reactors, nvmf_reactor, link);
if (i == master_core) {
@ -379,7 +371,7 @@ nvmf_destroy_threads(void)
struct nvmf_reactor *nvmf_reactor, *tmp;
TAILQ_FOREACH_SAFE(nvmf_reactor, &g_reactors, link, tmp) {
pthread_mutex_destroy(&nvmf_reactor->mutex);
spdk_ring_free(nvmf_reactor->threads);
free(nvmf_reactor);
}