static void rds_ib_tasklet_fn_recv(unsigned long data)
{
struct rds_ib_connection *ic = (struct rds_ib_connection *)data;
struct rds_connection *conn = ic->conn;
struct rds_ib_device *rds_ibdev = ic->rds_ibdev;
struct rds_ib_ack_state state;
if (!rds_ibdev)
rds_conn_drop(conn);
rds_ib_stats_inc(s_ib_tasklet_call);
memset(&state, 0, sizeof(state));
poll_rcq(ic, ic->i_recv_cq, ic->i_recv_wc, &state);
ib_req_notify_cq(ic->i_recv_cq, IB_CQ_SOLICITED);
poll_rcq(ic, ic->i_recv_cq, ic->i_recv_wc, &state);
if (state.ack_next_valid)
rds_ib_set_ack(ic, state.ack_next, state.ack_required);
if (state.ack_recv_valid && state.ack_recv > ic->i_ack_recv) {
rds_send_drop_acked(conn, state.ack_recv, NULL);
ic->i_ack_recv = state.ack_recv;
}
if (rds_conn_up(conn))
rds_ib_attempt_ack(ic);
}
void rds_ib_mr_cqe_handler(struct rds_ib_connection *ic, struct ib_wc *wc)
{
struct rds_ib_mr *ibmr = (void *)(unsigned long)wc->wr_id;
struct rds_ib_frmr *frmr = &ibmr->u.frmr;
if (wc->status != IB_WC_SUCCESS) {
frmr->fr_state = FRMR_IS_STALE;
if (rds_conn_up(ic->conn))
rds_ib_conn_error(ic->conn,
"frmr completion <%pI4,%pI4> status %u(%s), vendor_err 0x%x, disconnecting and reconnecting\n",
&ic->conn->c_laddr,
&ic->conn->c_faddr,
wc->status,
ib_wc_status_msg(wc->status),
wc->vendor_err);
}
if (frmr->fr_inv) {
frmr->fr_state = FRMR_IS_FREE;
frmr->fr_inv = false;
atomic_inc(&ic->i_fastreg_wrs);
} else {
atomic_inc(&ic->i_fastunreg_wrs);
}
}
int rds_tcp_conn_connect(struct rds_connection *conn)
{
struct socket *sock = NULL;
struct sockaddr_in src, dest;
int ret;
struct rds_tcp_connection *tc = conn->c_transport_data;
mutex_lock(&tc->t_conn_lock);
if (rds_conn_up(conn)) {
mutex_unlock(&tc->t_conn_lock);
return 0;
}
ret = sock_create_kern(rds_conn_net(conn), PF_INET,
SOCK_STREAM, IPPROTO_TCP, &sock);
if (ret < 0)
goto out;
rds_tcp_tune(sock);
src.sin_family = AF_INET;
src.sin_addr.s_addr = (__force u32)conn->c_laddr;
src.sin_port = (__force u16)htons(0);
ret = sock->ops->bind(sock, (struct sockaddr *)&src, sizeof(src));
if (ret) {
rdsdebug("bind failed with %d at address %pI4\n",
ret, &conn->c_laddr);
goto out;
}
dest.sin_family = AF_INET;
dest.sin_addr.s_addr = (__force u32)conn->c_faddr;
dest.sin_port = (__force u16)htons(RDS_TCP_PORT);
/*
* once we call connect() we can start getting callbacks and they
* own the socket
*/
rds_tcp_set_callbacks(sock, conn);
ret = sock->ops->connect(sock, (struct sockaddr *)&dest, sizeof(dest),
O_NONBLOCK);
rdsdebug("connect to address %pI4 returned %d\n", &conn->c_faddr, ret);
if (ret == -EINPROGRESS)
ret = 0;
if (ret == 0) {
rds_tcp_keepalive(sock);
sock = NULL;
} else {
rds_tcp_restore_callbacks(sock, conn->c_transport_data);
}
out:
mutex_unlock(&tc->t_conn_lock);
if (sock)
sock_release(sock);
return ret;
}
static void rds_ib_tasklet_fn_send(unsigned long data)
{
struct rds_ib_connection *ic = (struct rds_ib_connection *)data;
struct rds_connection *conn = ic->conn;
rds_ib_stats_inc(s_ib_tasklet_call);
poll_scq(ic, ic->i_send_cq, ic->i_send_wc);
ib_req_notify_cq(ic->i_send_cq, IB_CQ_NEXT_COMP);
poll_scq(ic, ic->i_send_cq, ic->i_send_wc);
if (rds_conn_up(conn) &&
(!test_bit(RDS_LL_SEND_FULL, &conn->c_flags) ||
test_bit(0, &conn->c_map_queued)))
rds_send_xmit(ic->conn);
}
static void rds_ib_tasklet_fn_send(unsigned long data)
{
struct rds_ib_connection *ic = (struct rds_ib_connection *)data;
struct rds_connection *conn = ic->conn;
rds_ib_stats_inc(s_ib_tasklet_call);
/* if cq has been already reaped, ignore incoming cq event */
if (atomic_read(&ic->i_cq_quiesce))
return;
poll_scq(ic, ic->i_send_cq, ic->i_send_wc);
ib_req_notify_cq(ic->i_send_cq, IB_CQ_NEXT_COMP);
poll_scq(ic, ic->i_send_cq, ic->i_send_wc);
if (rds_conn_up(conn) &&
(!test_bit(RDS_LL_SEND_FULL, &conn->c_flags) ||
test_bit(0, &conn->c_map_queued)))
rds_send_xmit(&ic->conn->c_path[0]);
}
/*
* The _oldest/_free ring operations here race cleanly with the alloc/unalloc
* operations performed in the send path. As the sender allocs and potentially
* unallocs the next free entry in the ring it doesn't alter which is
* the next to be freed, which is what this is concerned with.
*/
void rds_ib_send_cq_comp_handler(struct ib_cq *cq, void *context)
{
struct rds_connection *conn = context;
struct rds_ib_connection *ic = conn->c_transport_data;
struct ib_wc wc;
struct rds_ib_send_work *send;
u32 completed;
u32 oldest;
u32 i = 0;
int ret;
rdsdebug("cq %p conn %p\n", cq, conn);
rds_ib_stats_inc(s_ib_tx_cq_call);
ret = ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
if (ret)
rdsdebug("ib_req_notify_cq send failed: %d\n", ret);
while (ib_poll_cq(cq, 1, &wc) > 0) {
rdsdebug("wc wr_id 0x%llx status %u byte_len %u imm_data %u\n",
(unsigned long long)wc.wr_id, wc.status, wc.byte_len,
be32_to_cpu(wc.ex.imm_data));
rds_ib_stats_inc(s_ib_tx_cq_event);
if (wc.wr_id == RDS_IB_ACK_WR_ID) {
if (ic->i_ack_queued + HZ/2 < jiffies)
rds_ib_stats_inc(s_ib_tx_stalled);
rds_ib_ack_send_complete(ic);
continue;
}
oldest = rds_ib_ring_oldest(&ic->i_send_ring);
completed = rds_ib_ring_completed(&ic->i_send_ring, wc.wr_id, oldest);
for (i = 0; i < completed; i++) {
send = &ic->i_sends[oldest];
/* In the error case, wc.opcode sometimes contains garbage */
switch (send->s_wr.opcode) {
case IB_WR_SEND:
if (send->s_rm)
rds_ib_send_unmap_rm(ic, send, wc.status);
break;
case IB_WR_RDMA_WRITE:
case IB_WR_RDMA_READ:
/* Nothing to be done - the SG list will be unmapped
* when the SEND completes. */
break;
default:
if (printk_ratelimit())
printk(KERN_NOTICE
"RDS/IB: %s: unexpected opcode 0x%x in WR!\n",
__func__, send->s_wr.opcode);
break;
}
send->s_wr.opcode = 0xdead;
send->s_wr.num_sge = 1;
if (send->s_queued + HZ/2 < jiffies)
rds_ib_stats_inc(s_ib_tx_stalled);
/* If a RDMA operation produced an error, signal this right
* away. If we don't, the subsequent SEND that goes with this
* RDMA will be canceled with ERR_WFLUSH, and the application
* never learn that the RDMA failed. */
if (unlikely(wc.status == IB_WC_REM_ACCESS_ERR && send->s_op)) {
struct rds_message *rm;
rm = rds_send_get_message(conn, send->s_op);
if (rm)
rds_ib_send_rdma_complete(rm, wc.status);
}
oldest = (oldest + 1) % ic->i_send_ring.w_nr;
}
rds_ib_ring_free(&ic->i_send_ring, completed);
if (test_and_clear_bit(RDS_LL_SEND_FULL, &conn->c_flags)
|| test_bit(0, &conn->c_map_queued))
queue_delayed_work(rds_wq, &conn->c_send_w, 0);
/* We expect errors as the qp is drained during shutdown */
if (wc.status != IB_WC_SUCCESS && rds_conn_up(conn)) {
rds_ib_conn_error(conn,
"send completion on %pI4 "
"had status %u, disconnecting and reconnecting\n",
&conn->c_faddr, wc.status);
}
}
}
