diff options
| author |   Sowmini Varadhan <sowmini.varadhan@oracle.com> | 2018-01-04 06:53:00 -0800 |
|---|---|---|
| committer |   David S. Miller <davem@davemloft.net> | 2018-01-05 13:39:18 -0500 |
| commit | 3db6e0d172c94bd9953a1347c55ffb64b1d2e74f (patch) | |
| tree | 67f61aa4784da3fba56aa43e3626d3f3b6562c78 /net/rds/cong.c | |
| parent | rds: Use atomic flag to track connections being destroyed (diff) | |
| download | linux-dev-3db6e0d172c94bd9953a1347c55ffb64b1d2e74f.tar.xz linux-dev-3db6e0d172c94bd9953a1347c55ffb64b1d2e74f.zip | |
rds: use RCU to synchronize work-enqueue with connection teardown
rds_sendmsg() can enqueue work on cp_send_w from process context, but
it should not enqueue this work if connection teardown has commenced
(else we risk enquing work after rds_conn_path_destroy() has assumed that
all work has been cancelled/flushed).
Similarly some other functions like rds_cong_queue_updates
and rds_tcp_data_ready are called in softirq context, and may end
up enqueuing work on rds_wq after rds_conn_path_destroy() has assumed
that all workqs are quiesced.
Check the RDS_DESTROY_PENDING bit and use rcu synchronization to avoid
all these races.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/rds/cong.c')
| -rw-r--r-- | net/rds/cong.c | 10 |
1 files changed, 7 insertions, 3 deletions
diff --git a/net/rds/cong.c b/net/rds/cong.c index 8398fee7c866..8d19fd25dce3 100644 --- a/net/rds/cong.c +++ b/net/rds/cong.c @@ -219,7 +219,11 @@ void rds_cong_queue_updates(struct rds_cong_map *map) spin_lock_irqsave(&rds_cong_lock, flags); list_for_each_entry(conn, &map->m_conn_list, c_map_item) { - if (!test_and_set_bit(0, &conn->c_map_queued)) { + struct rds_conn_path *cp = &conn->c_path[0]; + + rcu_read_lock(); + if (!test_and_set_bit(0, &conn->c_map_queued) && + !test_bit(RDS_DESTROY_PENDING, &cp->cp_flags)) { rds_stats_inc(s_cong_update_queued); /* We cannot inline the call to rds_send_xmit() here * for two reasons (both pertaining to a TCP transport): @@ -235,9 +239,9 @@ void rds_cong_queue_updates(struct rds_cong_map *map) * therefore trigger warnings. * Defer the xmit to rds_send_worker() instead. */ - queue_delayed_work(rds_wq, - &conn->c_path[0].cp_send_w, 0); + queue_delayed_work(rds_wq, &cp->cp_send_w, 0); } + rcu_read_unlock(); } spin_unlock_irqrestore(&rds_cong_lock, flags); |