From a4d67a1e0572a3bdb353c03eb3cff9646c6e2a5d Mon Sep 17 00:00:00 2001 From: Ravishankar N Date: Mon, 2 Jul 2018 12:40:41 +0530 Subject: afr: add quorum checks in pre-op Backport of https://review.gluster.org/#/c/19781/ Problem: We seem to be winding the FOP if pre-op did not succeed on quorum bricks and then failing the FOP with EROFS since the fop did not meet quorum. This essentially masks the actual error due to which pre-op failed. (See BZ). Fix: Skip FOP phase if pre-op quorum is not met and go to post-op. Change-Id: Ie58a41e8fa1ad79aa06093706e96db8eef61b6d9 BUG: 1597154 Signed-off-by: Ravishankar N --- xlators/cluster/afr/src/afr-transaction.c | 60 ++++++++++++++++--------------- 1 file changed, 31 insertions(+), 29 deletions(-) diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c index 19740e18b18..98886246df4 100644 --- a/xlators/cluster/afr/src/afr-transaction.c +++ b/xlators/cluster/afr/src/afr-transaction.c @@ -137,6 +137,29 @@ afr_needs_changelog_update (afr_local_t *local) return _gf_false; } +gf_boolean_t +afr_changelog_has_quorum (afr_local_t *local, xlator_t *this) +{ + afr_private_t *priv = NULL; + int i = 0; + unsigned char *success_children = NULL; + + priv = this->private; + success_children = alloca0 (priv->child_count); + + for (i = 0; i < priv->child_count; i++) { + if (!local->transaction.failed_subvols[i]) { + success_children[i] = 1; + } + } + + if (afr_has_quorum (success_children, this)) { + return _gf_true; + } + + return _gf_false; +} + int __afr_txn_write_fop (call_frame_t *frame, xlator_t *this) { @@ -154,7 +177,10 @@ __afr_txn_write_fop (call_frame_t *frame, xlator_t *this) call_count = priv->child_count - AFR_COUNT (failed_subvols, priv->child_count); - if (call_count == 0) { + /* Fail if pre-op did not succeed on quorum no. of bricks. */ + if (!afr_changelog_has_quorum (local, this) || !call_count) { + local->op_ret = -1; + /* local->op_errno is already captured in changelog cbk. */ local->transaction.resume (frame, this); return 0; } @@ -533,33 +559,6 @@ afr_lock_server_count (afr_private_t *priv, afr_transaction_type type) /* {{{ pending */ - -void -afr_handle_post_op_quorum (afr_local_t *local, xlator_t *this) -{ - afr_private_t *priv = NULL; - int i = 0; - unsigned char *post_op_children = NULL; - - priv = this->private; - post_op_children = alloca0 (priv->child_count); - - for (i = 0; i < priv->child_count; i++) { - if (!local->transaction.failed_subvols[i]) { - post_op_children[i] = 1; - } - } - - if (afr_has_quorum (post_op_children, this)) { - return; - } - - local->op_ret = -1; - /*local->op_errno is already captured in post-op callback.*/ - - return; -} - int afr_changelog_post_op_done (call_frame_t *frame, xlator_t *this) { @@ -572,7 +571,10 @@ afr_changelog_post_op_done (call_frame_t *frame, xlator_t *this) int_lock = &local->internal_lock; /* Fail the FOP if post-op did not succeed on quorum no. of bricks. */ - afr_handle_post_op_quorum (local, this); + if (!afr_changelog_has_quorum (local, this)) { + local->op_ret = -1; + /*local->op_errno is already captured in changelog cbk*/ + } if (local->transaction.resume_stub) { call_resume (local->transaction.resume_stub); -- cgit