afr: make flush non-transactional

Flush is historically a transaction to ensure all previous writes
were complete. This is no longer required as write-behind has
learned to make flush a barrier operation (re: conversation w/
Avati).

Flush taking a full file lock causes VMs running on afr volumes
to stall when a migration occurs and self-heal is in progress.
Make afr_flush() a non-transactional operation.

BUG: 874045
Change-Id: If2db83823e280c86b1b29b41361eed7081601632
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-on: http://review.gluster.org/4261
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

3 files changed, 38 insertions, 139 deletions

diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 352010854..46f492bc5 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -2490,137 +2490,45 @@ afr_fd_ctx_set (xlator_t *this, fd_t *fd)
 /* {{{ flush */
 
 int
-afr_flush_unwind (call_frame_t *frame, xlator_t *this)
+afr_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_local_t *   local = NULL;
-        call_frame_t   *main_frame = NULL;
-
-        local = frame->local;
-
-        LOCK (&frame->lock);
-        {
-                if (local->transaction.main_frame)
-                        main_frame = local->transaction.main_frame;
-                local->transaction.main_frame = NULL;
-        }
-        UNLOCK (&frame->lock);
-
-        if (main_frame) {
-                AFR_STACK_UNWIND (flush, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  NULL);
-        }
-
-        return 0;
-}
-
-
-int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
-        afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
         int call_count  = -1;
-        int child_index = (long) cookie;
-        int need_unwind = 0;
 
         local = frame->local;
-        priv  = this->private;
 
         LOCK (&frame->lock);
         {
-                if (afr_fop_failed (op_ret, op_errno))
-                        afr_transaction_fop_failed (frame, this, child_index);
-
                 if (op_ret != -1) {
                         if (local->success_count == 0) {
                                 local->op_ret = op_ret;
                         }
                         local->success_count++;
-
-                        if (local->success_count == priv->wait_count) {
-                                need_unwind = 1;
-                        }
                 }
 
                 local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
-        if (need_unwind)
-                afr_flush_unwind (frame, this);
-
-        call_count = afr_frame_return (frame);
-
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-        }
-
-        return 0;
-}
-
-
-int
-afr_flush_wind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
-        int i = 0;
-        int call_count = -1;
-
-        local = frame->local;
-        priv = this->private;
-
-        call_count = afr_up_children_count (local->child_up, priv->child_count);
-
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
-
-        local->call_count = call_count;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_flush_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->flush,
-                                           local->fd, NULL);
-
-                        if (!--call_count)
-                                break;
-                }
-        }
-
-        return 0;
-}
-
+	call_count = afr_frame_return (frame);
 
-int
-afr_flush_done (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-
-        local = frame->local;
-
-        local->transaction.unwind (frame, this);
-
-        AFR_STACK_DESTROY (frame);
+	if (call_count == 0)
+		AFR_STACK_UNWIND(flush, frame, local->op_ret,
+				 local->op_errno, NULL);
 
         return 0;
 }
 
-
 int
 afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
         afr_local_t   *local = NULL;
-        call_frame_t  *transaction_frame = NULL;
         int            ret        = -1;
         int            op_errno   = 0;
+	int	       call_count = -1;
+	int	       i = 0;
 
         VALIDATE_OR_GOTO (frame, out);
         VALIDATE_OR_GOTO (this, out);
@@ -2628,50 +2536,40 @@ afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 
         priv = this->private;
 
-        transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                op_errno = ENOMEM;
-                goto out;
-        }
+	AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+	local = frame->local;
 
-        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
-        local = transaction_frame->local;
+	ret = afr_local_init(local, priv, &op_errno);
+	if (ret < 0)
+		goto out;
 
-        ret = afr_local_init (local, priv, &op_errno);
-        if (ret < 0)
-                goto out;
-
-        local->op = GF_FOP_FLUSH;
-
-        local->transaction.fop    = afr_flush_wind;
-        local->transaction.done   = afr_flush_done;
-        local->transaction.unwind = afr_flush_unwind;
+	local->fd = fd_ref(fd);
+        call_count = local->call_count;
 
-        local->fd                 = fd_ref (fd);
+	/*
+	 * Ideally we should synchronize flush against completion of writing
+	 * the delayed changelog, but for now we just push it out first...
+	 */
+	afr_delayed_changelog_wake_up(this, fd);
 
-        local->transaction.main_frame = frame;
-        local->transaction.start  = 0;
-        local->transaction.len    = 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_flush_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->flush,
+                                           local->fd, NULL);
 
-        ret = afr_open_fd_fix (transaction_frame, this, _gf_false);
-        if (ret) {
-                op_errno = -ret;
-                goto out;
-        }
-        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
-        if (ret < 0) {
-            op_errno = -ret;
-            goto out;
+                        if (!--call_count)
+                                break;
+                }
         }
 
-        ret = 0;
-out:
-        if (ret < 0) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
+	ret = 0;
 
-                AFR_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
-        }
+out:
+	if (ret < 0)
+		AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
 
         return 0;
 }
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 32fb8d9a4..3ba8d9b56 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -1301,9 +1301,6 @@ void
 afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd);
 
 void
-afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
-
-void
 afr_delayed_changelog_wake_up_cbk (void *data)
 {
 	fd_t           *fd = NULL;
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 102599633..e95bc5b14 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -33,4 +33,8 @@ afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
                       int child, afr_xattrop_type_t op);
 void
 afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this);
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
+
 #endif /* __TRANSACTION_H__ */