From b66623e33efbbf55717df7bfc49882371118b866 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Wed, 8 Aug 2012 21:19:09 +0200 Subject: [PATCH] drbd: Avoid NetworkFailure state during disconnect Disconnecting is a cluster wide state change. In case the peer node agrees to the state transition, it sends back the fact on the meta-data connection and closes both sockets. In case the node node that initiated the state transfer sees the closing action on the data-socket, before the P_STATE_CHG_REPLY packet, it was going into one of the network failure states. At least with the fencing option set to something else thatn "dont-care", the unclean shutdown of the connection causes a short IO freeze or a fence operation. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_int.h | 1 + drivers/block/drbd/drbd_receiver.c | 30 +++++++++++++++++++++++++++++- drivers/block/drbd/drbd_state.c | 3 +++ 3 files changed, 33 insertions(+), 1 deletion(-) diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index b83398d64a9c..37ae87e468ae 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -816,6 +816,7 @@ enum { * so shrink_page_list() would not recurse into, * and potentially deadlock on, this drbd worker. */ + DISCONNECT_SENT, }; struct drbd_tconn { /* is a resource from the config file */ diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c index 0eefbeb65664..1a8f698021a2 100644 --- a/drivers/block/drbd/drbd_receiver.c +++ b/drivers/block/drbd/drbd_receiver.c @@ -522,7 +522,6 @@ static int drbd_recv(struct drbd_tconn *tconn, void *buf, size_t size) conn_err(tconn, "sock_recvmsg returned %d\n", rv); break; } else if (rv == 0) { - conn_info(tconn, "sock was shut down by peer\n"); break; } else { /* signal came in, or peer/link went down, @@ -535,9 +534,25 @@ static int drbd_recv(struct drbd_tconn *tconn, void *buf, size_t size) set_fs(oldfs); + if (rv == 0) { + if (test_bit(DISCONNECT_SENT, &tconn->flags)) { + long t; + rcu_read_lock(); + t = rcu_dereference(tconn->net_conf)->ping_timeo * HZ/10; + rcu_read_unlock(); + + t = wait_event_timeout(tconn->ping_wait, tconn->cstate < C_WF_REPORT_PARAMS, t); + + if (t) + goto out; + } + conn_info(tconn, "sock was shut down by peer\n"); + } + if (rv != size) conn_request_state(tconn, NS(conn, C_BROKEN_PIPE), CS_HARD); +out: return rv; } @@ -894,6 +909,7 @@ static int conn_connect(struct drbd_tconn *tconn) .door_bell = COMPLETION_INITIALIZER_ONSTACK(ad.door_bell), }; + clear_bit(DISCONNECT_SENT, &tconn->flags); if (conn_request_state(tconn, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS) return -2; @@ -5316,6 +5332,18 @@ int drbd_asender(struct drbd_thread *thi) received += rv; buf += rv; } else if (rv == 0) { + if (test_bit(DISCONNECT_SENT, &tconn->flags)) { + long t; + rcu_read_lock(); + t = rcu_dereference(tconn->net_conf)->ping_timeo * HZ/10; + rcu_read_unlock(); + + t = wait_event_timeout(tconn->ping_wait, + tconn->cstate < C_WF_REPORT_PARAMS, + t); + if (t) + break; + } conn_err(tconn, "meta connection shut down by peer.\n"); goto reconnect; } else if (rv == -EAGAIN) { diff --git a/drivers/block/drbd/drbd_state.c b/drivers/block/drbd/drbd_state.c index c16349aec23c..4fda4e2024ec 100644 --- a/drivers/block/drbd/drbd_state.c +++ b/drivers/block/drbd/drbd_state.c @@ -1742,6 +1742,9 @@ conn_cl_wide(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state v goto abort; } + if (val.conn == C_DISCONNECTING) + set_bit(DISCONNECT_SENT, &tconn->flags); + wait_event(tconn->ping_wait, (rv = _conn_rq_cond(tconn, mask, val))); clear_bit(CONN_WD_ST_CHG_REQ, &tconn->flags); -- 2.34.1