glusterfs: Use gcc builtin ATOMIC operator to increase/decreate refcount.

Problem: In glusterfs code base we call mutex_lock/unlock to take
         reference/dereference for a object.Sometime it could be
         reason for lock contention also.

Solution: There is no need to use mutex to increase/decrease ref
          counter, instead of using mutex use gcc builtin ATOMIC
          operation.

Test:   I have not observed yet how much performance gain after apply
        this patch specific to glusterfs but i have tested same
        with below small program(mutex and atomic both) and
        get good difference.

static int numOuterLoops;
static void *
threadFunc(void *arg)
{
    int j;

    for (j = 0; j < numOuterLoops; j++) {
            __atomic_add_fetch (&glob, 1,__ATOMIC_ACQ_REL);
    }
    return NULL;
}

int
main(int argc, char *argv[])
{
    int opt, s, j;
    int numThreads;
    pthread_t *thread;
    int verbose;
    int64_t n = 0;

    if (argc < 2 ) {
     printf(" Please provide 2 args Num of threads && Outer Loop\n");
     exit (-1);
    }
    numThreads = atoi(argv[1]);
    numOuterLoops = atoi (argv[2]);

    if (1) {
        printf("\tthreads: %d; outer loops: %d;\n",
                numThreads, numOuterLoops);
    }

    thread = calloc(numThreads, sizeof(pthread_t));
    if (thread == NULL) {
        printf ("calloc error so exit\n");
        exit (-1);
    }

    __atomic_store (&glob, &n, __ATOMIC_RELEASE);
    for (j = 0; j < numThreads; j++) {
        s = pthread_create(&thread[j], NULL, threadFunc, NULL);
        if (s != 0) {
            printf ("pthread_create failed so exit\n");
            exit (-1);
        }
    }

    for (j = 0; j < numThreads; j++) {
        s = pthread_join(thread[j], NULL);
        if (s != 0) {
            printf ("pthread_join failed so exit\n");
            exit (-1);
        }
    }
    printf("glob value is %ld\n",__atomic_load_n (&glob,__ATOMIC_RELAXED));

    exit(0);
}

   time ./thr_count 800 800000
   threads: 800; outer loops: 800000;
   glob value is 640000000

real	1m10.288s
user	0m57.269s
sys	3m31.565s

time ./thr_count_atomic 800 800000
     threads: 800; outer loops: 800000;
glob value is 640000000

real	0m20.313s
user	1m20.558s
sys	0m0.028

Change-Id: Ie5030a52ea264875e002e108dd4b207b15ab7cc7
Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>

9 files changed, 42 insertions, 84 deletions

diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 3d6fbb1fd79..cc7d41b3bfa 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -87,8 +87,8 @@ struct dht_layout {
          */
         int                gen;
         int                type;
-        int                ref; /* use with dht_conf_t->layout_lock */
-        uint32_t           search_unhashed;
+        gf_atomic_t        ref; /* use with dht_conf_t->layout_lock */
+        gf_boolean_t       search_unhashed;
         struct {
                 int        err;   /* 0 = normal
                                      -1 = dir exists and no xattr
diff --git a/xlators/cluster/dht/src/dht-layout.c b/xlators/cluster/dht/src/dht-layout.c
index 419ce329a1d..97b98e01451 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -48,12 +48,12 @@ dht_layout_new (xlator_t *this, int cnt)
                 layout->gen = conf->gen;
         }
 
-        layout->ref = 1;
+        GF_ATOMIC_INIT (layout->ref, 1);
 
         ENSURE(NULL != layout);
         ENSURE(layout->type == DHT_HASH_TYPE_DM);
         ENSURE(layout->cnt == cnt);
-        ENSURE(layout->ref == 1);
+        ENSURE(GF_ATOMIC_GET (layout->ref) == 1);
 out:
         return layout;
 }
@@ -62,24 +62,13 @@ out:
 dht_layout_t *
 dht_layout_get (xlator_t *this, inode_t *inode)
 {
-        dht_conf_t   *conf = NULL;
         dht_layout_t *layout = NULL;
         int           ret = 0;
 
-        conf = this->private;
-        if (!conf)
-                goto out;
-
-        LOCK (&conf->layout_lock);
-        {
-                ret = dht_inode_ctx_layout_get (inode, this, &layout);
-                if ((!ret) && layout) {
-                        layout->ref++;
-                }
+        ret = dht_inode_ctx_layout_get (inode, this, &layout);
+        if ((!ret) && layout) {
+                GF_ATOMIC_INC (layout->ref);
         }
-        UNLOCK (&conf->layout_lock);
-
-out:
         return layout;
 }
 
@@ -100,7 +89,7 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
         {
                 oldret = dht_inode_ctx_layout_get (inode, this, &old_layout);
                 if (layout)
-                        layout->ref++;
+                        GF_ATOMIC_INC (layout->ref);
                 ret = dht_inode_ctx_layout_set (inode, this, layout);
         }
         UNLOCK (&conf->layout_lock);
@@ -108,6 +97,8 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
         if (!oldret) {
                 dht_layout_unref (this, old_layout);
         }
+        if (ret)
+                GF_ATOMIC_DEC (layout->ref);
 
 out:
         return ret;
@@ -117,19 +108,12 @@ out:
 void
 dht_layout_unref (xlator_t *this, dht_layout_t *layout)
 {
-        dht_conf_t  *conf = NULL;
         int          ref = 0;
 
         if (!layout || layout->preset || !this->private)
                 return;
 
-        conf = this->private;
-
-        LOCK (&conf->layout_lock);
-        {
-                ref = --layout->ref;
-        }
-        UNLOCK (&conf->layout_lock);
+        ref = GF_ATOMIC_DEC (layout->ref);
 
         if (!ref)
                 GF_FREE (layout);
@@ -139,17 +123,10 @@ dht_layout_unref (xlator_t *this, dht_layout_t *layout)
 dht_layout_t *
 dht_layout_ref (xlator_t *this, dht_layout_t *layout)
 {
-        dht_conf_t  *conf = NULL;
-
         if (layout->preset || !this->private)
                 return layout;
 
-        conf = this->private;
-        LOCK (&conf->layout_lock);
-        {
-                layout->ref++;
-        }
-        UNLOCK (&conf->layout_lock);
+        GF_ATOMIC_INC (layout->ref);
 
         return layout;
 }
diff --git a/xlators/cluster/dht/src/unittest/dht_layout_unittest.c b/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
index 84a89160e38..fb6445ce1f3 100644
--- a/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
+++ b/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
@@ -90,7 +90,7 @@ test_dht_layout_new(void **state)
     assert_non_null(layout);
     assert_int_equal(layout->type, DHT_HASH_TYPE_DM);
     assert_int_equal(layout->cnt, cnt);
-    assert_int_equal(layout->ref, 1);
+    assert_int_equal(GF_ATOMIC_GET (layout->ref), 1);
     assert_int_equal(layout->gen, 0);
     assert_int_equal(layout->spread_cnt, 0);
     free(layout);
@@ -107,7 +107,7 @@ test_dht_layout_new(void **state)
     assert_non_null(layout);
     assert_int_equal(layout->type, DHT_HASH_TYPE_DM);
     assert_int_equal(layout->cnt, cnt);
-    assert_int_equal(layout->ref, 1);
+    assert_int_equal(GF_ATOMIC_GET (layout->ref), 1);
     assert_int_equal(layout->gen, conf->gen);
     assert_int_equal(layout->spread_cnt, conf->dir_spread_cnt);
     free(layout);
diff --git a/xlators/debug/io-stats/src/io-stats.c b/xlators/debug/io-stats/src/io-stats.c
index c0cd0e2477b..4864c67a101 100644
--- a/xlators/debug/io-stats/src/io-stats.c
+++ b/xlators/debug/io-stats/src/io-stats.c
@@ -74,7 +74,7 @@ struct ios_stat {
         char           *filename;
         gf_atomic_t     counters[IOS_STATS_TYPE_MAX];
         struct ios_stat_lat thru_counters [IOS_STATS_THRU_MAX];
-        int             refcnt;
+        gf_atomic_t     refcnt;
 };
 
 struct ios_stat_list {
@@ -333,13 +333,10 @@ ios_fd_ctx_set (fd_t *fd, xlator_t *this, struct ios_fd *iosfd)
 static int
 ios_stat_ref (struct ios_stat *iosstat)
 {
-        LOCK (&iosstat->lock);
-        {
-                iosstat->refcnt++;
-        }
-        UNLOCK (&iosstat->lock);
+        uint64_t refcnt = 0;
+        refcnt = GF_ATOMIC_INC (iosstat->refcnt);
 
-        return iosstat->refcnt;
+        return refcnt;
 }
 
 
@@ -347,18 +344,16 @@ static int
 ios_stat_unref (struct ios_stat *iosstat)
 {
         int cleanup = 0;
-        LOCK (&iosstat->lock);
-        {
-                iosstat->refcnt--;
-                if (iosstat->refcnt == 0) {
-                        if (iosstat->filename) {
-                                GF_FREE (iosstat->filename);
-                                iosstat->filename = NULL;
-                        }
-                        cleanup = 1;
+        uint64_t refcnt = 0;
+
+        refcnt = GF_ATOMIC_DEC (iosstat->refcnt);
+        if (refcnt == 0) {
+                if (iosstat->filename) {
+                        GF_FREE (iosstat->filename);
+                        iosstat->filename = NULL;
                 }
+                cleanup = 1;
         }
-        UNLOCK (&iosstat->lock);
 
         if (cleanup) {
                 LOCK_DESTROY (&iosstat->lock);
diff --git a/xlators/features/changelog/src/changelog-ev-handle.h b/xlators/features/changelog/src/changelog-ev-handle.h
index eef0492a9ee..e89af8793a8 100644
--- a/xlators/features/changelog/src/changelog-ev-handle.h
+++ b/xlators/features/changelog/src/changelog-ev-handle.h
@@ -24,7 +24,7 @@ typedef struct changelog_rpc_clnt {
 
         gf_lock_t lock;
 
-        unsigned long ref;
+        gf_atomic_t   ref;
         gf_boolean_t  disconnected;
 
         unsigned int filter;
@@ -43,11 +43,7 @@ typedef struct changelog_rpc_clnt {
 static inline void
 changelog_rpc_clnt_ref (changelog_rpc_clnt_t *crpc)
 {
-        LOCK (&crpc->lock);
-        {
-                ++crpc->ref;
-        }
-        UNLOCK (&crpc->lock);
+        GF_ATOMIC_INC (crpc->ref);
 }
 
 static inline void
@@ -66,16 +62,14 @@ static inline void
 changelog_rpc_clnt_unref (changelog_rpc_clnt_t *crpc)
 {
         gf_boolean_t gone = _gf_false;
+        uint64_t     ref  = 0;
+
+        ref = GF_ATOMIC_DEC (crpc->ref);
 
-        LOCK (&crpc->lock);
-        {
-                if (!(--crpc->ref)
-                    && changelog_rpc_clnt_is_disconnected (crpc)) {
-                        list_del (&crpc->list);
-                        gone = _gf_true;
-                }
+        if (!ref && changelog_rpc_clnt_is_disconnected (crpc)) {
+                list_del (&crpc->list);
+                gone = _gf_true;
         }
-        UNLOCK (&crpc->lock);
 
         if (gone)
                 crpc->cleanup (crpc);
diff --git a/xlators/features/changelog/src/changelog-rpc.c b/xlators/features/changelog/src/changelog-rpc.c
index 5524e433cbb..72b5cdbc7cb 100644
--- a/xlators/features/changelog/src/changelog-rpc.c
+++ b/xlators/features/changelog/src/changelog-rpc.c
@@ -211,7 +211,7 @@ changelog_rpc_clnt_init (xlator_t *this,
         /* Take a ref, the last unref will be on RPC_CLNT_DESTROY
          * which comes as a result of last rpc_clnt_unref.
          */
-        crpc->ref = 1;
+        GF_ATOMIC_INIT (crpc->ref, 1);
         changelog_set_disconnect_flag (crpc, _gf_false);
 
         crpc->filter = rpc_req->filter;
diff --git a/xlators/nfs/server/src/nfs3.c b/xlators/nfs/server/src/nfs3.c
index b053eb3250e..6428a77442e 100644
--- a/xlators/nfs/server/src/nfs3.c
+++ b/xlators/nfs/server/src/nfs3.c
@@ -539,8 +539,8 @@ static void
 __nfs3_call_state_wipe (nfs3_call_state_t *cs)
 {
         if (cs->fd) {
-                gf_msg_trace (GF_NFS3, 0, "fd 0x%lx ref: %d",
-                        (long)cs->fd, cs->fd->refcount);
+                gf_msg_trace (GF_NFS3, 0, "fd 0x%lx ref: %"PRId64,
+                        (long)cs->fd, GF_ATOMIC_GET (cs->fd->refcount));
                 fd_unref (cs->fd);
         }
 
diff --git a/xlators/protocol/client/src/client-handshake.c b/xlators/protocol/client/src/client-handshake.c
index ca3d140d08e..6429dc2e8ae 100644
--- a/xlators/protocol/client/src/client-handshake.c
+++ b/xlators/protocol/client/src/client-handshake.c
@@ -32,7 +32,7 @@ int client_set_lk_version_cbk (struct rpc_req *req, struct iovec *iov,
 int client_set_lk_version (xlator_t *this);
 
 typedef struct client_fd_lk_local {
-        int             ref;
+        gf_atomic_t     ref;
         gf_boolean_t    error;
         gf_lock_t       lock;
         clnt_fd_ctx_t *fdctx;
@@ -285,11 +285,7 @@ clnt_fd_lk_local_ref (xlator_t *this, clnt_fd_lk_local_t *local)
 {
         GF_VALIDATE_OR_GOTO (this->name, local, out);
 
-        LOCK (&local->lock);
-        {
-                local->ref++;
-        }
-        UNLOCK (&local->lock);
+        GF_ATOMIC_INC (local->ref);
 out:
         return local;
 }
@@ -301,11 +297,7 @@ clnt_fd_lk_local_unref (xlator_t *this, clnt_fd_lk_local_t *local)
 
         GF_VALIDATE_OR_GOTO (this->name, local, out);
 
-        LOCK (&local->lock);
-        {
-                ref = --local->ref;
-        }
-        UNLOCK (&local->lock);
+        ref = GF_ATOMIC_DEC (local->ref);
 
         if (ref == 0) {
                 LOCK_DESTROY (&local->lock);
@@ -325,7 +317,7 @@ clnt_fd_lk_local_create (clnt_fd_ctx_t *fdctx)
         if (!local)
                 goto out;
 
-        local->ref    = 1;
+        GF_ATOMIC_INIT (local->ref, 1);
         local->error  = _gf_false;
         local->fdctx = fdctx;
 
diff --git a/xlators/protocol/client/src/client.c b/xlators/protocol/client/src/client.c
index 8df697ed883..8ec9eaeb8a9 100644
--- a/xlators/protocol/client/src/client.c
+++ b/xlators/protocol/client/src/client.c
@@ -2844,7 +2844,7 @@ client_fd_lk_ctx_dump (xlator_t *this, fd_lk_ctx_t *lk_ctx, int nth_fd)
         fd_lk_ctx_node_t        *plock                   = NULL;
         char                    key[GF_DUMP_MAX_BUF_LEN] = {0,};
 
-        lk_ctx_ref = fd_lk_ctx_try_ref (lk_ctx);
+        lk_ctx_ref = fd_lk_ctx_ref (lk_ctx);
         if (!lk_ctx_ref)
                 return;