summaryrefslogtreecommitdiff
path: root/net/rds
diff options
context:
space:
mode:
Diffstat (limited to 'net/rds')
-rw-r--r--net/rds/connection.c3
-rw-r--r--net/rds/iw_rdma.c40
-rw-r--r--net/rds/rds.h8
-rw-r--r--net/rds/recv.c4
-rw-r--r--net/rds/send.c36
5 files changed, 62 insertions, 29 deletions
diff --git a/net/rds/connection.c b/net/rds/connection.c
index 378c3a6acf84..14f041398ca1 100644
--- a/net/rds/connection.c
+++ b/net/rds/connection.c
@@ -130,7 +130,7 @@ static struct rds_connection *__rds_conn_create(__be32 laddr, __be32 faddr,
rcu_read_lock();
conn = rds_conn_lookup(head, laddr, faddr, trans);
if (conn && conn->c_loopback && conn->c_trans != &rds_loop_transport &&
- !is_outgoing) {
+ laddr == faddr && !is_outgoing) {
/* This is a looped back IB connection, and we're
* called by the code handling the incoming connect.
* We need a second connection object into which we
@@ -193,6 +193,7 @@ static struct rds_connection *__rds_conn_create(__be32 laddr, __be32 faddr,
}
atomic_set(&conn->c_state, RDS_CONN_DOWN);
+ conn->c_send_gen = 0;
conn->c_reconnect_jiffies = 0;
INIT_DELAYED_WORK(&conn->c_send_w, rds_send_worker);
INIT_DELAYED_WORK(&conn->c_recv_w, rds_recv_worker);
diff --git a/net/rds/iw_rdma.c b/net/rds/iw_rdma.c
index a817705ce2d0..dba8d0864f18 100644
--- a/net/rds/iw_rdma.c
+++ b/net/rds/iw_rdma.c
@@ -88,7 +88,9 @@ static unsigned int rds_iw_unmap_fastreg_list(struct rds_iw_mr_pool *pool,
int *unpinned);
static void rds_iw_destroy_fastreg(struct rds_iw_mr_pool *pool, struct rds_iw_mr *ibmr);
-static int rds_iw_get_device(struct rds_sock *rs, struct rds_iw_device **rds_iwdev, struct rdma_cm_id **cm_id)
+static int rds_iw_get_device(struct sockaddr_in *src, struct sockaddr_in *dst,
+ struct rds_iw_device **rds_iwdev,
+ struct rdma_cm_id **cm_id)
{
struct rds_iw_device *iwdev;
struct rds_iw_cm_id *i_cm_id;
@@ -112,15 +114,15 @@ static int rds_iw_get_device(struct rds_sock *rs, struct rds_iw_device **rds_iwd
src_addr->sin_port,
dst_addr->sin_addr.s_addr,
dst_addr->sin_port,
- rs->rs_bound_addr,
- rs->rs_bound_port,
- rs->rs_conn_addr,
- rs->rs_conn_port);
+ src->sin_addr.s_addr,
+ src->sin_port,
+ dst->sin_addr.s_addr,
+ dst->sin_port);
#ifdef WORKING_TUPLE_DETECTION
- if (src_addr->sin_addr.s_addr == rs->rs_bound_addr &&
- src_addr->sin_port == rs->rs_bound_port &&
- dst_addr->sin_addr.s_addr == rs->rs_conn_addr &&
- dst_addr->sin_port == rs->rs_conn_port) {
+ if (src_addr->sin_addr.s_addr == src->sin_addr.s_addr &&
+ src_addr->sin_port == src->sin_port &&
+ dst_addr->sin_addr.s_addr == dst->sin_addr.s_addr &&
+ dst_addr->sin_port == dst->sin_port) {
#else
/* FIXME - needs to compare the local and remote
* ipaddr/port tuple, but the ipaddr is the only
@@ -128,7 +130,7 @@ static int rds_iw_get_device(struct rds_sock *rs, struct rds_iw_device **rds_iwd
* zero'ed. It doesn't appear to be properly populated
* during connection setup...
*/
- if (src_addr->sin_addr.s_addr == rs->rs_bound_addr) {
+ if (src_addr->sin_addr.s_addr == src->sin_addr.s_addr) {
#endif
spin_unlock_irq(&iwdev->spinlock);
*rds_iwdev = iwdev;
@@ -180,19 +182,13 @@ int rds_iw_update_cm_id(struct rds_iw_device *rds_iwdev, struct rdma_cm_id *cm_i
{
struct sockaddr_in *src_addr, *dst_addr;
struct rds_iw_device *rds_iwdev_old;
- struct rds_sock rs;
struct rdma_cm_id *pcm_id;
int rc;
src_addr = (struct sockaddr_in *)&cm_id->route.addr.src_addr;
dst_addr = (struct sockaddr_in *)&cm_id->route.addr.dst_addr;
- rs.rs_bound_addr = src_addr->sin_addr.s_addr;
- rs.rs_bound_port = src_addr->sin_port;
- rs.rs_conn_addr = dst_addr->sin_addr.s_addr;
- rs.rs_conn_port = dst_addr->sin_port;
-
- rc = rds_iw_get_device(&rs, &rds_iwdev_old, &pcm_id);
+ rc = rds_iw_get_device(src_addr, dst_addr, &rds_iwdev_old, &pcm_id);
if (rc)
rds_iw_remove_cm_id(rds_iwdev, cm_id);
@@ -598,9 +594,17 @@ void *rds_iw_get_mr(struct scatterlist *sg, unsigned long nents,
struct rds_iw_device *rds_iwdev;
struct rds_iw_mr *ibmr = NULL;
struct rdma_cm_id *cm_id;
+ struct sockaddr_in src = {
+ .sin_addr.s_addr = rs->rs_bound_addr,
+ .sin_port = rs->rs_bound_port,
+ };
+ struct sockaddr_in dst = {
+ .sin_addr.s_addr = rs->rs_conn_addr,
+ .sin_port = rs->rs_conn_port,
+ };
int ret;
- ret = rds_iw_get_device(rs, &rds_iwdev, &cm_id);
+ ret = rds_iw_get_device(&src, &dst, &rds_iwdev, &cm_id);
if (ret || !cm_id) {
ret = -ENODEV;
goto out;
diff --git a/net/rds/rds.h b/net/rds/rds.h
index c2a5eef41343..0d41155a2258 100644
--- a/net/rds/rds.h
+++ b/net/rds/rds.h
@@ -110,6 +110,7 @@ struct rds_connection {
void *c_transport_data;
atomic_t c_state;
+ unsigned long c_send_gen;
unsigned long c_flags;
unsigned long c_reconnect_jiffies;
struct delayed_work c_send_w;
@@ -702,8 +703,8 @@ void rds_inc_init(struct rds_incoming *inc, struct rds_connection *conn,
void rds_inc_put(struct rds_incoming *inc);
void rds_recv_incoming(struct rds_connection *conn, __be32 saddr, __be32 daddr,
struct rds_incoming *inc, gfp_t gfp);
-int rds_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
- size_t size, int msg_flags);
+int rds_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
+ int msg_flags);
void rds_clear_recv_queue(struct rds_sock *rs);
int rds_notify_queue_get(struct rds_sock *rs, struct msghdr *msg);
void rds_inc_info_copy(struct rds_incoming *inc,
@@ -711,8 +712,7 @@ void rds_inc_info_copy(struct rds_incoming *inc,
__be32 saddr, __be32 daddr, int flip);
/* send.c */
-int rds_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
- size_t payload_len);
+int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len);
void rds_send_reset(struct rds_connection *conn);
int rds_send_xmit(struct rds_connection *conn);
struct sockaddr_in;
diff --git a/net/rds/recv.c b/net/rds/recv.c
index f9ec1acd801c..a00462b0d01d 100644
--- a/net/rds/recv.c
+++ b/net/rds/recv.c
@@ -395,8 +395,8 @@ static int rds_cmsg_recv(struct rds_incoming *inc, struct msghdr *msg)
return 0;
}
-int rds_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
- size_t size, int msg_flags)
+int rds_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
+ int msg_flags)
{
struct sock *sk = sock->sk;
struct rds_sock *rs = rds_sk_to_rs(sk);
diff --git a/net/rds/send.c b/net/rds/send.c
index 42f65d4305c8..e9430f537f9c 100644
--- a/net/rds/send.c
+++ b/net/rds/send.c
@@ -140,8 +140,11 @@ int rds_send_xmit(struct rds_connection *conn)
struct scatterlist *sg;
int ret = 0;
LIST_HEAD(to_be_dropped);
+ int batch_count;
+ unsigned long send_gen = 0;
restart:
+ batch_count = 0;
/*
* sendmsg calls here after having queued its message on the send
@@ -157,6 +160,17 @@ restart:
}
/*
+ * we record the send generation after doing the xmit acquire.
+ * if someone else manages to jump in and do some work, we'll use
+ * this to avoid a goto restart farther down.
+ *
+ * The acquire_in_xmit() check above ensures that only one
+ * caller can increment c_send_gen at any time.
+ */
+ conn->c_send_gen++;
+ send_gen = conn->c_send_gen;
+
+ /*
* rds_conn_shutdown() sets the conn state and then tests RDS_IN_XMIT,
* we do the opposite to avoid races.
*/
@@ -202,6 +216,16 @@ restart:
if (!rm) {
unsigned int len;
+ batch_count++;
+
+ /* we want to process as big a batch as we can, but
+ * we also want to avoid softlockups. If we've been
+ * through a lot of messages, lets back off and see
+ * if anyone else jumps in
+ */
+ if (batch_count >= 1024)
+ goto over_batch;
+
spin_lock_irqsave(&conn->c_lock, flags);
if (!list_empty(&conn->c_send_queue)) {
@@ -357,9 +381,9 @@ restart:
}
}
+over_batch:
if (conn->c_trans->xmit_complete)
conn->c_trans->xmit_complete(conn);
-
release_in_xmit(conn);
/* Nuke any messages we decided not to retransmit. */
@@ -380,10 +404,15 @@ restart:
* If the transport cannot continue (i.e ret != 0), then it must
* call us when more room is available, such as from the tx
* completion handler.
+ *
+ * We have an extra generation check here so that if someone manages
+ * to jump in after our release_in_xmit, we'll see that they have done
+ * some work and we will skip our goto
*/
if (ret == 0) {
smp_mb();
- if (!list_empty(&conn->c_send_queue)) {
+ if (!list_empty(&conn->c_send_queue) &&
+ send_gen == conn->c_send_gen) {
rds_stats_inc(s_send_lock_queue_raced);
goto restart;
}
@@ -920,8 +949,7 @@ static int rds_cmsg_send(struct rds_sock *rs, struct rds_message *rm,
return ret;
}
-int rds_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg,
- size_t payload_len)
+int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
{
struct sock *sk = sock->sk;
struct rds_sock *rs = rds_sk_to_rs(sk);
generated by cgit v1.2.3 (git 2.39.0) at 2025-02-14 12:48:22 +0300