[linux.git] / net / smc / smc_cdc.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Shared Memory Communications over RDMA (SMC-R) and RoCE
 *
 * Connection Data Control (CDC)
 * handles flow control
 *
 * Copyright IBM Corp. 2016
 *
 * Author(s):  Ursula Braun <[email protected]>
 */

#include <linux/spinlock.h>

#include "smc.h"
#include "smc_wr.h"
#include "smc_cdc.h"
#include "smc_tx.h"
#include "smc_rx.h"
#include "smc_close.h"

/********************************** send *************************************/

/* handler for send/transmission completion of a CDC msg */
static void smc_cdc_tx_handler(struct smc_wr_tx_pend_priv *pnd_snd,
			       struct smc_link *link,
			       enum ib_wc_status wc_status)
{
	struct smc_cdc_tx_pend *cdcpend = (struct smc_cdc_tx_pend *)pnd_snd;
	struct smc_connection *conn = cdcpend->conn;
	struct smc_sock *smc;
	int diff;

	smc = container_of(conn, struct smc_sock, conn);
	bh_lock_sock(&smc->sk);
	if (!wc_status) {
		diff = smc_curs_diff(cdcpend->conn->sndbuf_desc->len,
				     &cdcpend->conn->tx_curs_fin,
				     &cdcpend->cursor);
		/* sndbuf_space is decreased in smc_sendmsg */
		smp_mb__before_atomic();
		atomic_add(diff, &cdcpend->conn->sndbuf_space);
		/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
		smp_mb__after_atomic();
		smc_curs_copy(&conn->tx_curs_fin, &cdcpend->cursor, conn);
		smc_curs_copy(&conn->local_tx_ctrl_fin, &cdcpend->p_cursor,
			      conn);
		conn->tx_cdc_seq_fin = cdcpend->ctrl_seq;
	}

	if (atomic_dec_and_test(&conn->cdc_pend_tx_wr) &&
	    unlikely(wq_has_sleeper(&conn->cdc_pend_tx_wq)))
		wake_up(&conn->cdc_pend_tx_wq);
	WARN_ON(atomic_read(&conn->cdc_pend_tx_wr) < 0);

	smc_tx_sndbuf_nonfull(smc);
	bh_unlock_sock(&smc->sk);
}

int smc_cdc_get_free_slot(struct smc_connection *conn,
			  struct smc_link *link,
			  struct smc_wr_buf **wr_buf,
			  struct smc_rdma_wr **wr_rdma_buf,
			  struct smc_cdc_tx_pend **pend)
{
	int rc;

	rc = smc_wr_tx_get_free_slot(link, smc_cdc_tx_handler, wr_buf,
				     wr_rdma_buf,
				     (struct smc_wr_tx_pend_priv **)pend);
	if (conn->killed) {
		/* abnormal termination */
		if (!rc)
			smc_wr_tx_put_slot(link,
					   (struct smc_wr_tx_pend_priv *)pend);
		rc = -EPIPE;
	}
	return rc;
}

static inline void smc_cdc_add_pending_send(struct smc_connection *conn,
					    struct smc_cdc_tx_pend *pend)
{
	BUILD_BUG_ON_MSG(
		sizeof(struct smc_cdc_msg) > SMC_WR_BUF_SIZE,
		"must increase SMC_WR_BUF_SIZE to at least sizeof(struct smc_cdc_msg)");
	BUILD_BUG_ON_MSG(
		offsetofend(struct smc_cdc_msg, reserved) > SMC_WR_TX_SIZE,
		"must adapt SMC_WR_TX_SIZE to sizeof(struct smc_cdc_msg); if not all smc_wr upper layer protocols use the same message size any more, must start to set link->wr_tx_sges[i].length on each individual smc_wr_tx_send()");
	BUILD_BUG_ON_MSG(
		sizeof(struct smc_cdc_tx_pend) > SMC_WR_TX_PEND_PRIV_SIZE,
		"must increase SMC_WR_TX_PEND_PRIV_SIZE to at least sizeof(struct smc_cdc_tx_pend)");
	pend->conn = conn;
	pend->cursor = conn->tx_curs_sent;
	pend->p_cursor = conn->local_tx_ctrl.prod;
	pend->ctrl_seq = conn->tx_cdc_seq;
}

int smc_cdc_msg_send(struct smc_connection *conn,
		     struct smc_wr_buf *wr_buf,
		     struct smc_cdc_tx_pend *pend)
{
	struct smc_link *link = conn->lnk;
	union smc_host_cursor cfed;
	int rc;

	smc_cdc_add_pending_send(conn, pend);

	conn->tx_cdc_seq++;
	conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
	smc_host_msg_to_cdc((struct smc_cdc_msg *)wr_buf, conn, &cfed);

	atomic_inc(&conn->cdc_pend_tx_wr);
	smp_mb__after_atomic(); /* Make sure cdc_pend_tx_wr added before post */

	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
	if (!rc) {
		smc_curs_copy(&conn->rx_curs_confirmed, &cfed, conn);
		conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
	} else {
		conn->tx_cdc_seq--;
		conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
		atomic_dec(&conn->cdc_pend_tx_wr);
	}

	return rc;
}

/* send a validation msg indicating the move of a conn to an other QP link */
int smcr_cdc_msg_send_validation(struct smc_connection *conn,
				 struct smc_cdc_tx_pend *pend,
				 struct smc_wr_buf *wr_buf)
{
	struct smc_host_cdc_msg *local = &conn->local_tx_ctrl;
	struct smc_link *link = conn->lnk;
	struct smc_cdc_msg *peer;
	int rc;

	peer = (struct smc_cdc_msg *)wr_buf;
	peer->common.type = local->common.type;
	peer->len = local->len;
	peer->seqno = htons(conn->tx_cdc_seq_fin); /* seqno last compl. tx */
	peer->token = htonl(local->token);
	peer->prod_flags.failover_validation = 1;

	/* We need to set pend->conn here to make sure smc_cdc_tx_handler()
	 * can handle properly
	 */
	smc_cdc_add_pending_send(conn, pend);

	atomic_inc(&conn->cdc_pend_tx_wr);
	smp_mb__after_atomic(); /* Make sure cdc_pend_tx_wr added before post */

	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
	if (unlikely(rc))
		atomic_dec(&conn->cdc_pend_tx_wr);

	return rc;
}

static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
{
	struct smc_cdc_tx_pend *pend;
	struct smc_wr_buf *wr_buf;
	struct smc_link *link;
	bool again = false;
	int rc;

again:
	link = conn->lnk;
	if (!smc_wr_tx_link_hold(link))
		return -ENOLINK;
	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, NULL, &pend);
	if (rc)
		goto put_out;

	spin_lock_bh(&conn->send_lock);
	if (link != conn->lnk) {
		/* link of connection changed, try again one time*/
		spin_unlock_bh(&conn->send_lock);
		smc_wr_tx_put_slot(link,
				   (struct smc_wr_tx_pend_priv *)pend);
		smc_wr_tx_link_put(link);
		if (again)
			return -ENOLINK;
		again = true;
		goto again;
	}
	rc = smc_cdc_msg_send(conn, wr_buf, pend);
	spin_unlock_bh(&conn->send_lock);
put_out:
	smc_wr_tx_link_put(link);
	return rc;
}

int smc_cdc_get_slot_and_msg_send(struct smc_connection *conn)
{
	int rc;

	if (!conn->lgr || (conn->lgr->is_smcd && conn->lgr->peer_shutdown))
		return -EPIPE;

	if (conn->lgr->is_smcd) {
		spin_lock_bh(&conn->send_lock);
		rc = smcd_cdc_msg_send(conn);
		spin_unlock_bh(&conn->send_lock);
	} else {
		rc = smcr_cdc_get_slot_and_msg_send(conn);
	}

	return rc;
}

void smc_cdc_wait_pend_tx_wr(struct smc_connection *conn)
{
	wait_event(conn->cdc_pend_tx_wq, !atomic_read(&conn->cdc_pend_tx_wr));
}

/* Send a SMC-D CDC header.
 * This increments the free space available in our send buffer.
 * Also update the confirmed receive buffer with what was sent to the peer.
 */
int smcd_cdc_msg_send(struct smc_connection *conn)
{
	struct smc_sock *smc = container_of(conn, struct smc_sock, conn);
	union smc_host_cursor curs;
	struct smcd_cdc_msg cdc;
	int rc, diff;

	memset(&cdc, 0, sizeof(cdc));
	cdc.common.type = SMC_CDC_MSG_TYPE;
	curs.acurs.counter = atomic64_read(&conn->local_tx_ctrl.prod.acurs);
	cdc.prod.wrap = curs.wrap;
	cdc.prod.count = curs.count;
	curs.acurs.counter = atomic64_read(&conn->local_tx_ctrl.cons.acurs);
	cdc.cons.wrap = curs.wrap;
	cdc.cons.count = curs.count;
	cdc.cons.prod_flags = conn->local_tx_ctrl.prod_flags;
	cdc.cons.conn_state_flags = conn->local_tx_ctrl.conn_state_flags;
	rc = smcd_tx_ism_write(conn, &cdc, sizeof(cdc), 0, 1);
	if (rc)
		return rc;
	smc_curs_copy(&conn->rx_curs_confirmed, &curs, conn);
	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
	/* Calculate transmitted data and increment free send buffer space */
	diff = smc_curs_diff(conn->sndbuf_desc->len, &conn->tx_curs_fin,
			     &conn->tx_curs_sent);
	/* increased by confirmed number of bytes */
	smp_mb__before_atomic();
	atomic_add(diff, &conn->sndbuf_space);
	/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
	smp_mb__after_atomic();
	smc_curs_copy(&conn->tx_curs_fin, &conn->tx_curs_sent, conn);

	smc_tx_sndbuf_nonfull(smc);
	return rc;
}

/********************************* receive ***********************************/

static inline bool smc_cdc_before(u16 seq1, u16 seq2)
{
	return (s16)(seq1 - seq2) < 0;
}

static void smc_cdc_handle_urg_data_arrival(struct smc_sock *smc,
					    int *diff_prod)
{
	struct smc_connection *conn = &smc->conn;
	char *base;

	/* new data included urgent business */
	smc_curs_copy(&conn->urg_curs, &conn->local_rx_ctrl.prod, conn);
	conn->urg_state = SMC_URG_VALID;
	if (!sock_flag(&smc->sk, SOCK_URGINLINE))
		/* we'll skip the urgent byte, so don't account for it */
		(*diff_prod)--;
	base = (char *)conn->rmb_desc->cpu_addr + conn->rx_off;
	if (conn->urg_curs.count)
		conn->urg_rx_byte = *(base + conn->urg_curs.count - 1);
	else
		conn->urg_rx_byte = *(base + conn->rmb_desc->len - 1);
	sk_send_sigurg(&smc->sk);
}

static void smc_cdc_msg_validate(struct smc_sock *smc, struct smc_cdc_msg *cdc,
				 struct smc_link *link)
{
	struct smc_connection *conn = &smc->conn;
	u16 recv_seq = ntohs(cdc->seqno);
	s16 diff;

	/* check that seqnum was seen before */
	diff = conn->local_rx_ctrl.seqno - recv_seq;
	if (diff < 0) { /* diff larger than 0x7fff */
		/* drop connection */
		conn->out_of_sync = 1;	/* prevent any further receives */
		spin_lock_bh(&conn->send_lock);
		conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
		conn->lnk = link;
		spin_unlock_bh(&conn->send_lock);
		sock_hold(&smc->sk); /* sock_put in abort_work */
		if (!queue_work(smc_close_wq, &conn->abort_work))
			sock_put(&smc->sk);
	}
}

static void smc_cdc_msg_recv_action(struct smc_sock *smc,
				    struct smc_cdc_msg *cdc)
{
	union smc_host_cursor cons_old, prod_old;
	struct smc_connection *conn = &smc->conn;
	int diff_cons, diff_prod;

	smc_curs_copy(&prod_old, &conn->local_rx_ctrl.prod, conn);
	smc_curs_copy(&cons_old, &conn->local_rx_ctrl.cons, conn);
	smc_cdc_msg_to_host(&conn->local_rx_ctrl, cdc, conn);

	diff_cons = smc_curs_diff(conn->peer_rmbe_size, &cons_old,
				  &conn->local_rx_ctrl.cons);
	if (diff_cons) {
		/* peer_rmbe_space is decreased during data transfer with RDMA
		 * write
		 */
		smp_mb__before_atomic();
		atomic_add(diff_cons, &conn->peer_rmbe_space);
		/* guarantee 0 <= peer_rmbe_space <= peer_rmbe_size */
		smp_mb__after_atomic();
	}

	diff_prod = smc_curs_diff(conn->rmb_desc->len, &prod_old,
				  &conn->local_rx_ctrl.prod);
	if (diff_prod) {
		if (conn->local_rx_ctrl.prod_flags.urg_data_present)
			smc_cdc_handle_urg_data_arrival(smc, &diff_prod);
		/* bytes_to_rcv is decreased in smc_recvmsg */
		smp_mb__before_atomic();
		atomic_add(diff_prod, &conn->bytes_to_rcv);
		/* guarantee 0 <= bytes_to_rcv <= rmb_desc->len */
		smp_mb__after_atomic();
		smc->sk.sk_data_ready(&smc->sk);
	} else {
		if (conn->local_rx_ctrl.prod_flags.write_blocked)
			smc->sk.sk_data_ready(&smc->sk);
		if (conn->local_rx_ctrl.prod_flags.urg_data_pending)
			conn->urg_state = SMC_URG_NOTYET;
	}

	/* trigger sndbuf consumer: RDMA write into peer RMBE and CDC */
	if ((diff_cons && smc_tx_prepared_sends(conn)) ||
	    conn->local_rx_ctrl.prod_flags.cons_curs_upd_req ||
	    conn->local_rx_ctrl.prod_flags.urg_data_pending)
		smc_tx_sndbuf_nonempty(conn);

	if (diff_cons && conn->urg_tx_pend &&
	    atomic_read(&conn->peer_rmbe_space) == conn->peer_rmbe_size) {
		/* urg data confirmed by peer, indicate we're ready for more */
		conn->urg_tx_pend = false;
		smc->sk.sk_write_space(&smc->sk);
	}

	if (conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) {
		smc->sk.sk_err = ECONNRESET;
		conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
	}
	if (smc_cdc_rxed_any_close_or_senddone(conn)) {
		smc->sk.sk_shutdown |= RCV_SHUTDOWN;
		if (smc->clcsock && smc->clcsock->sk)
			smc->clcsock->sk->sk_shutdown |= RCV_SHUTDOWN;
		sock_set_flag(&smc->sk, SOCK_DONE);
		sock_hold(&smc->sk); /* sock_put in close_work */
		if (!queue_work(smc_close_wq, &conn->close_work))
			sock_put(&smc->sk);
	}
}

/* called under tasklet context */
static void smc_cdc_msg_recv(struct smc_sock *smc, struct smc_cdc_msg *cdc)
{
	sock_hold(&smc->sk);
	bh_lock_sock(&smc->sk);
	smc_cdc_msg_recv_action(smc, cdc);
	bh_unlock_sock(&smc->sk);
	sock_put(&smc->sk); /* no free sk in softirq-context */
}

/* Schedule a tasklet for this connection. Triggered from the ISM device IRQ
 * handler to indicate update in the DMBE.
 *
 * Context:
 * - tasklet context
 */
static void smcd_cdc_rx_tsklet(struct tasklet_struct *t)
{
	struct smc_connection *conn = from_tasklet(conn, t, rx_tsklet);
	struct smcd_cdc_msg *data_cdc;
	struct smcd_cdc_msg cdc;
	struct smc_sock *smc;

	if (!conn || conn->killed)
		return;

	data_cdc = (struct smcd_cdc_msg *)conn->rmb_desc->cpu_addr;
	smcd_curs_copy(&cdc.prod, &data_cdc->prod, conn);
	smcd_curs_copy(&cdc.cons, &data_cdc->cons, conn);
	smc = container_of(conn, struct smc_sock, conn);
	smc_cdc_msg_recv(smc, (struct smc_cdc_msg *)&cdc);
}

/* Initialize receive tasklet. Called from ISM device IRQ handler to start
 * receiver side.
 */
void smcd_cdc_rx_init(struct smc_connection *conn)
{
	tasklet_setup(&conn->rx_tsklet, smcd_cdc_rx_tsklet);
}

/***************************** init, exit, misc ******************************/

static void smc_cdc_rx_handler(struct ib_wc *wc, void *buf)
{
	struct smc_link *link = (struct smc_link *)wc->qp->qp_context;
	struct smc_cdc_msg *cdc = buf;
	struct smc_connection *conn;
	struct smc_link_group *lgr;
	struct smc_sock *smc;

	if (wc->byte_len < offsetof(struct smc_cdc_msg, reserved))
		return; /* short message */
	if (cdc->len != SMC_WR_TX_SIZE)
		return; /* invalid message */

	/* lookup connection */
	lgr = smc_get_lgr(link);
	read_lock_bh(&lgr->conns_lock);
	conn = smc_lgr_find_conn(ntohl(cdc->token), lgr);
	read_unlock_bh(&lgr->conns_lock);
	if (!conn || conn->out_of_sync)
		return;
	smc = container_of(conn, struct smc_sock, conn);

	if (cdc->prod_flags.failover_validation) {
		smc_cdc_msg_validate(smc, cdc, link);
		return;
	}
	if (smc_cdc_before(ntohs(cdc->seqno),
			   conn->local_rx_ctrl.seqno))
		/* received seqno is old */
		return;

	smc_cdc_msg_recv(smc, cdc);
}

static struct smc_wr_rx_handler smc_cdc_rx_handlers[] = {
	{
		.handler	= smc_cdc_rx_handler,
		.type		= SMC_CDC_MSG_TYPE
	},
	{
		.handler	= NULL,
	}
};

int __init smc_cdc_init(void)
{
	struct smc_wr_rx_handler *handler;
	int rc = 0;

	for (handler = smc_cdc_rx_handlers; handler->handler; handler++) {
		INIT_HLIST_NODE(&handler->list);
		rc = smc_wr_rx_register_handler(handler);
		if (rc)
			break;
	}
	return rc;
}
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
5f08318f UB	2	/*
	3	* Shared Memory Communications over RDMA (SMC-R) and RoCE
	4	*
	5	* Connection Data Control (CDC)
	6	* handles flow control
	7	*
	8	* Copyright IBM Corp. 2016
	9	*
	10	* Author(s): Ursula Braun <[email protected]>
	11	*/
	12
	13	#include <linux/spinlock.h>
	14
	15	#include "smc.h"
	16	#include "smc_wr.h"
	17	#include "smc_cdc.h"
e6727f39	18	#include "smc_tx.h"
952310cc	19	#include "smc_rx.h"
b38d7324	20	#include "smc_close.h"
5f08318f UB	21
	22	/******************************** send ***********************************/
	23
5f08318f UB	24	/* handler for send/transmission completion of a CDC msg */
	25	static void smc_cdc_tx_handler(struct smc_wr_tx_pend_priv *pnd_snd,
	26	struct smc_link *link,
	27	enum ib_wc_status wc_status)
	28	{
	29	struct smc_cdc_tx_pend cdcpend = (struct smc_cdc_tx_pend )pnd_snd;
bac6de7b	30	struct smc_connection *conn = cdcpend->conn;
5f08318f UB	31	struct smc_sock *smc;
	32	int diff;
	33
bac6de7b	34	smc = container_of(conn, struct smc_sock, conn);
5f08318f UB	35	bh_lock_sock(&smc->sk);
5f08318f UB	36	if (!wc_status) {
69cb7dc0	37	diff = smc_curs_diff(cdcpend->conn->sndbuf_desc->len,
5f08318f UB	38	&cdcpend->conn->tx_curs_fin,
	39	&cdcpend->cursor);
	40	/* sndbuf_space is decreased in smc_sendmsg */
	41	smp_mb__before_atomic();
	42	atomic_add(diff, &cdcpend->conn->sndbuf_space);
69cb7dc0	43	/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
5f08318f	44	smp_mb__after_atomic();
bac6de7b	45	smc_curs_copy(&conn->tx_curs_fin, &cdcpend->cursor, conn);
f0ec4f1d KG	46	smc_curs_copy(&conn->local_tx_ctrl_fin, &cdcpend->p_cursor,
	47	conn);
	48	conn->tx_cdc_seq_fin = cdcpend->ctrl_seq;
5f08318f	49	}
349d4312 DL	50
	51	if (atomic_dec_and_test(&conn->cdc_pend_tx_wr) &&
	52	unlikely(wq_has_sleeper(&conn->cdc_pend_tx_wq)))
	53	wake_up(&conn->cdc_pend_tx_wq);
	54	WARN_ON(atomic_read(&conn->cdc_pend_tx_wr) < 0);
	55
e6727f39	56	smc_tx_sndbuf_nonfull(smc);
5f08318f UB	57	bh_unlock_sock(&smc->sk);
	58	}
	59
51957bc5	60	int smc_cdc_get_free_slot(struct smc_connection *conn,
c6f02ebe	61	struct smc_link *link,
5f08318f	62	struct smc_wr_buf **wr_buf,
ad6f317f	63	struct smc_rdma_wr **wr_rdma_buf,
5f08318f UB	64	struct smc_cdc_tx_pend **pend)
5f08318f UB	65	{
1a0a04c7	66	int rc;
51957bc5	67
1a0a04c7	68	rc = smc_wr_tx_get_free_slot(link, smc_cdc_tx_handler, wr_buf,
ad6f317f	69	wr_rdma_buf,
1a0a04c7	70	(struct smc_wr_tx_pend_priv **)pend);
2bced6ae	71	if (conn->killed) {
1a0a04c7	72	/* abnormal termination */
2bced6ae KG	73	if (!rc)
	74	smc_wr_tx_put_slot(link,
	75	(struct smc_wr_tx_pend_priv *)pend);
1a0a04c7	76	rc = -EPIPE;
2bced6ae	77	}
1a0a04c7	78	return rc;
5f08318f UB	79	}
	80
	81	static inline void smc_cdc_add_pending_send(struct smc_connection *conn,
	82	struct smc_cdc_tx_pend *pend)
	83	{
	84	BUILD_BUG_ON_MSG(
	85	sizeof(struct smc_cdc_msg) > SMC_WR_BUF_SIZE,
	86	"must increase SMC_WR_BUF_SIZE to at least sizeof(struct smc_cdc_msg)");
	87	BUILD_BUG_ON_MSG(
b9a22dd9	88	offsetofend(struct smc_cdc_msg, reserved) > SMC_WR_TX_SIZE,
5f08318f UB	89	"must adapt SMC_WR_TX_SIZE to sizeof(struct smc_cdc_msg); if not all smc_wr upper layer protocols use the same message size any more, must start to set link->wr_tx_sges[i].length on each individual smc_wr_tx_send()");
	90	BUILD_BUG_ON_MSG(
	91	sizeof(struct smc_cdc_tx_pend) > SMC_WR_TX_PEND_PRIV_SIZE,
	92	"must increase SMC_WR_TX_PEND_PRIV_SIZE to at least sizeof(struct smc_cdc_tx_pend)");
	93	pend->conn = conn;
	94	pend->cursor = conn->tx_curs_sent;
	95	pend->p_cursor = conn->local_tx_ctrl.prod;
	96	pend->ctrl_seq = conn->tx_cdc_seq;
	97	}
	98
	99	int smc_cdc_msg_send(struct smc_connection *conn,
	100	struct smc_wr_buf *wr_buf,
	101	struct smc_cdc_tx_pend *pend)
	102	{
387707fd	103	struct smc_link *link = conn->lnk;
b8649efa	104	union smc_host_cursor cfed;
5f08318f UB	105	int rc;
5f08318f UB	106
5f08318f UB	107	smc_cdc_add_pending_send(conn, pend);
	108
	109	conn->tx_cdc_seq++;
	110	conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
ccc8ca9b	111	smc_host_msg_to_cdc((struct smc_cdc_msg *)wr_buf, conn, &cfed);
349d4312 DL	112
	113	atomic_inc(&conn->cdc_pend_tx_wr);
	114	smp_mb__after_atomic(); /* Make sure cdc_pend_tx_wr added before post */
	115
5f08318f	116	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
4dff63c2	117	if (!rc) {
b8649efa	118	smc_curs_copy(&conn->rx_curs_confirmed, &cfed, conn);
4dff63c2	119	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
f0ec4f1d KG	120	} else {
	121	conn->tx_cdc_seq--;
	122	conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
349d4312	123	atomic_dec(&conn->cdc_pend_tx_wr);
4dff63c2	124	}
5f08318f UB	125
	126	return rc;
	127	}
	128
29bd73db	129	/* send a validation msg indicating the move of a conn to an other QP link */
b8ded9de KG	130	int smcr_cdc_msg_send_validation(struct smc_connection *conn,
	131	struct smc_cdc_tx_pend *pend,
	132	struct smc_wr_buf *wr_buf)
29bd73db KG	133	{
	134	struct smc_host_cdc_msg *local = &conn->local_tx_ctrl;
	135	struct smc_link *link = conn->lnk;
29bd73db KG	136	struct smc_cdc_msg *peer;
	137	int rc;
	138
29bd73db KG	139	peer = (struct smc_cdc_msg *)wr_buf;
	140	peer->common.type = local->common.type;
	141	peer->len = local->len;
	142	peer->seqno = htons(conn->tx_cdc_seq_fin); /* seqno last compl. tx */
	143	peer->token = htonl(local->token);
	144	peer->prod_flags.failover_validation = 1;
	145
349d4312 DL	146	/* We need to set pend->conn here to make sure smc_cdc_tx_handler()
	147	* can handle properly
	148	*/
	149	smc_cdc_add_pending_send(conn, pend);
	150
	151	atomic_inc(&conn->cdc_pend_tx_wr);
	152	smp_mb__after_atomic(); /* Make sure cdc_pend_tx_wr added before post */
	153
29bd73db	154	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
349d4312 DL	155	if (unlikely(rc))
	156	atomic_dec(&conn->cdc_pend_tx_wr);
	157
29bd73db KG	158	return rc;
	159	}
	160
be244f28	161	static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
5f08318f UB	162	{
	163	struct smc_cdc_tx_pend *pend;
	164	struct smc_wr_buf *wr_buf;
c6f02ebe KG	165	struct smc_link *link;
c6f02ebe KG	166	bool again = false;
5f08318f UB	167	int rc;
5f08318f UB	168
c6f02ebe KG	169	again:
c6f02ebe KG	170	link = conn->lnk;
95f7f3e7 KG	171	if (!smc_wr_tx_link_hold(link))
95f7f3e7 KG	172	return -ENOLINK;
c6f02ebe	173	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, NULL, &pend);
5f08318f	174	if (rc)
95f7f3e7	175	goto put_out;
5f08318f	176
2dee25af	177	spin_lock_bh(&conn->send_lock);
c6f02ebe KG	178	if (link != conn->lnk) {
	179	/* link of connection changed, try again one time*/
	180	spin_unlock_bh(&conn->send_lock);
	181	smc_wr_tx_put_slot(link,
	182	(struct smc_wr_tx_pend_priv *)pend);
95f7f3e7	183	smc_wr_tx_link_put(link);
c6f02ebe KG	184	if (again)
	185	return -ENOLINK;
	186	again = true;
	187	goto again;
	188	}
2dee25af KG	189	rc = smc_cdc_msg_send(conn, wr_buf, pend);
2dee25af KG	190	spin_unlock_bh(&conn->send_lock);
95f7f3e7 KG	191	put_out:
95f7f3e7 KG	192	smc_wr_tx_link_put(link);
2dee25af	193	return rc;
5f08318f UB	194	}
5f08318f UB	195
be244f28 HW	196	int smc_cdc_get_slot_and_msg_send(struct smc_connection *conn)
	197	{
	198	int rc;
	199
50c6b20e UB	200	if (!conn->lgr \|\| (conn->lgr->is_smcd && conn->lgr->peer_shutdown))
	201	return -EPIPE;
	202
be244f28 HW	203	if (conn->lgr->is_smcd) {
	204	spin_lock_bh(&conn->send_lock);
	205	rc = smcd_cdc_msg_send(conn);
	206	spin_unlock_bh(&conn->send_lock);
	207	} else {
	208	rc = smcr_cdc_get_slot_and_msg_send(conn);
	209	}
	210
	211	return rc;
	212	}
	213
349d4312	214	void smc_cdc_wait_pend_tx_wr(struct smc_connection *conn)
5f08318f	215	{
349d4312	216	wait_event(conn->cdc_pend_tx_wq, !atomic_read(&conn->cdc_pend_tx_wr));
5f08318f UB	217	}
5f08318f UB	218
be244f28 HW	219	/* Send a SMC-D CDC header.
	220	* This increments the free space available in our send buffer.
	221	* Also update the confirmed receive buffer with what was sent to the peer.
	222	*/
	223	int smcd_cdc_msg_send(struct smc_connection *conn)
	224	{
	225	struct smc_sock *smc = container_of(conn, struct smc_sock, conn);
b9a22dd9	226	union smc_host_cursor curs;
be244f28 HW	227	struct smcd_cdc_msg cdc;
	228	int rc, diff;
	229
	230	memset(&cdc, 0, sizeof(cdc));
	231	cdc.common.type = SMC_CDC_MSG_TYPE;
b9a22dd9 UB	232	curs.acurs.counter = atomic64_read(&conn->local_tx_ctrl.prod.acurs);
	233	cdc.prod.wrap = curs.wrap;
	234	cdc.prod.count = curs.count;
	235	curs.acurs.counter = atomic64_read(&conn->local_tx_ctrl.cons.acurs);
	236	cdc.cons.wrap = curs.wrap;
	237	cdc.cons.count = curs.count;
	238	cdc.cons.prod_flags = conn->local_tx_ctrl.prod_flags;
	239	cdc.cons.conn_state_flags = conn->local_tx_ctrl.conn_state_flags;
be244f28 HW	240	rc = smcd_tx_ism_write(conn, &cdc, sizeof(cdc), 0, 1);
	241	if (rc)
	242	return rc;
b9a22dd9	243	smc_curs_copy(&conn->rx_curs_confirmed, &curs, conn);
4dff63c2	244	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
be244f28 HW	245	/* Calculate transmitted data and increment free send buffer space */
	246	diff = smc_curs_diff(conn->sndbuf_desc->len, &conn->tx_curs_fin,
	247	&conn->tx_curs_sent);
	248	/* increased by confirmed number of bytes */
	249	smp_mb__before_atomic();
	250	atomic_add(diff, &conn->sndbuf_space);
	251	/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
	252	smp_mb__after_atomic();
bac6de7b	253	smc_curs_copy(&conn->tx_curs_fin, &conn->tx_curs_sent, conn);
be244f28 HW	254
	255	smc_tx_sndbuf_nonfull(smc);
	256	return rc;
	257	}
	258
5f08318f UB	259	/******************************* receive *********************************/
	260
	261	static inline bool smc_cdc_before(u16 seq1, u16 seq2)
	262	{
	263	return (s16)(seq1 - seq2) < 0;
	264	}
	265
de8474eb SR	266	static void smc_cdc_handle_urg_data_arrival(struct smc_sock *smc,
	267	int *diff_prod)
	268	{
	269	struct smc_connection *conn = &smc->conn;
	270	char *base;
	271
	272	/* new data included urgent business */
bac6de7b	273	smc_curs_copy(&conn->urg_curs, &conn->local_rx_ctrl.prod, conn);
de8474eb SR	274	conn->urg_state = SMC_URG_VALID;
	275	if (!sock_flag(&smc->sk, SOCK_URGINLINE))
	276	/* we'll skip the urgent byte, so don't account for it */
	277	(*diff_prod)--;
be244f28	278	base = (char *)conn->rmb_desc->cpu_addr + conn->rx_off;
de8474eb SR	279	if (conn->urg_curs.count)
	280	conn->urg_rx_byte = *(base + conn->urg_curs.count - 1);
	281	else
	282	conn->urg_rx_byte = *(base + conn->rmb_desc->len - 1);
	283	sk_send_sigurg(&smc->sk);
	284	}
	285
b286a065 KG	286	static void smc_cdc_msg_validate(struct smc_sock smc, struct smc_cdc_msg cdc,
	287	struct smc_link *link)
	288	{
	289	struct smc_connection *conn = &smc->conn;
	290	u16 recv_seq = ntohs(cdc->seqno);
	291	s16 diff;
	292
	293	/* check that seqnum was seen before */
	294	diff = conn->local_rx_ctrl.seqno - recv_seq;
	295	if (diff < 0) { /* diff larger than 0x7fff */
	296	/* drop connection */
	297	conn->out_of_sync = 1; /* prevent any further receives */
	298	spin_lock_bh(&conn->send_lock);
	299	conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
	300	conn->lnk = link;
	301	spin_unlock_bh(&conn->send_lock);
	302	sock_hold(&smc->sk); /* sock_put in abort_work */
22ef473d	303	if (!queue_work(smc_close_wq, &conn->abort_work))
b286a065 KG	304	sock_put(&smc->sk);
	305	}
	306	}
	307
5f08318f	308	static void smc_cdc_msg_recv_action(struct smc_sock *smc,
5f08318f UB	309	struct smc_cdc_msg *cdc)
	310	{
	311	union smc_host_cursor cons_old, prod_old;
	312	struct smc_connection *conn = &smc->conn;
	313	int diff_cons, diff_prod;
	314
bac6de7b SR	315	smc_curs_copy(&prod_old, &conn->local_rx_ctrl.prod, conn);
bac6de7b SR	316	smc_curs_copy(&cons_old, &conn->local_rx_ctrl.cons, conn);
5f08318f UB	317	smc_cdc_msg_to_host(&conn->local_rx_ctrl, cdc, conn);
	318
	319	diff_cons = smc_curs_diff(conn->peer_rmbe_size, &cons_old,
	320	&conn->local_rx_ctrl.cons);
	321	if (diff_cons) {
	322	/* peer_rmbe_space is decreased during data transfer with RDMA
	323	* write
	324	*/
	325	smp_mb__before_atomic();
	326	atomic_add(diff_cons, &conn->peer_rmbe_space);
	327	/* guarantee 0 <= peer_rmbe_space <= peer_rmbe_size */
	328	smp_mb__after_atomic();
	329	}
	330
69cb7dc0	331	diff_prod = smc_curs_diff(conn->rmb_desc->len, &prod_old,
5f08318f UB	332	&conn->local_rx_ctrl.prod);
5f08318f UB	333	if (diff_prod) {
de8474eb SR	334	if (conn->local_rx_ctrl.prod_flags.urg_data_present)
de8474eb SR	335	smc_cdc_handle_urg_data_arrival(smc, &diff_prod);
5f08318f UB	336	/* bytes_to_rcv is decreased in smc_recvmsg */
	337	smp_mb__before_atomic();
	338	atomic_add(diff_prod, &conn->bytes_to_rcv);
69cb7dc0	339	/* guarantee 0 <= bytes_to_rcv <= rmb_desc->len */
5f08318f	340	smp_mb__after_atomic();
952310cc	341	smc->sk.sk_data_ready(&smc->sk);
de8474eb	342	} else {
cf0cfe53 KG	343	if (conn->local_rx_ctrl.prod_flags.write_blocked)
	344	smc->sk.sk_data_ready(&smc->sk);
	345	if (conn->local_rx_ctrl.prod_flags.urg_data_pending)
	346	conn->urg_state = SMC_URG_NOTYET;
5f08318f UB	347	}
5f08318f UB	348
51f1de79	349	/* trigger sndbuf consumer: RDMA write into peer RMBE and CDC */
cf0cfe53 KG	350	if ((diff_cons && smc_tx_prepared_sends(conn)) \|\|
	351	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req \|\|
	352	conn->local_rx_ctrl.prod_flags.urg_data_pending)
51f1de79	353	smc_tx_sndbuf_nonempty(conn);
cf0cfe53	354
de8474eb SR	355	if (diff_cons && conn->urg_tx_pend &&
	356	atomic_read(&conn->peer_rmbe_space) == conn->peer_rmbe_size) {
	357	/* urg data confirmed by peer, indicate we're ready for more */
	358	conn->urg_tx_pend = false;
	359	smc->sk.sk_write_space(&smc->sk);
	360	}
51f1de79	361
b38d7324	362	if (conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) {
5f08318f	363	smc->sk.sk_err = ECONNRESET;
b38d7324 UB	364	conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
b38d7324 UB	365	}
46c28dbd UB	366	if (smc_cdc_rxed_any_close_or_senddone(conn)) {
	367	smc->sk.sk_shutdown \|= RCV_SHUTDOWN;
	368	if (smc->clcsock && smc->clcsock->sk)
	369	smc->clcsock->sk->sk_shutdown \|= RCV_SHUTDOWN;
	370	sock_set_flag(&smc->sk, SOCK_DONE);
51f1de79	371	sock_hold(&smc->sk); /* sock_put in close_work */
22ef473d	372	if (!queue_work(smc_close_wq, &conn->close_work))
51f1de79	373	sock_put(&smc->sk);
b38d7324	374	}
5f08318f UB	375	}
	376
	377	/* called under tasklet context */
d7b0e37c	378	static void smc_cdc_msg_recv(struct smc_sock smc, struct smc_cdc_msg cdc)
5f08318f	379	{
5f08318f	380	sock_hold(&smc->sk);
5f08318f	381	bh_lock_sock(&smc->sk);
d7b0e37c	382	smc_cdc_msg_recv_action(smc, cdc);
5f08318f UB	383	bh_unlock_sock(&smc->sk);
	384	sock_put(&smc->sk); /* no free sk in softirq-context */
	385	}
	386
be244f28 HW	387	/* Schedule a tasklet for this connection. Triggered from the ISM device IRQ
	388	* handler to indicate update in the DMBE.
	389	*
	390	* Context:
	391	* - tasklet context
	392	*/
fcb8e3a3	393	static void smcd_cdc_rx_tsklet(struct tasklet_struct *t)
be244f28	394	{
fcb8e3a3	395	struct smc_connection *conn = from_tasklet(conn, t, rx_tsklet);
b9a22dd9	396	struct smcd_cdc_msg *data_cdc;
be244f28 HW	397	struct smcd_cdc_msg cdc;
	398	struct smc_sock *smc;
	399
b2900980	400	if (!conn \|\| conn->killed)
be244f28 HW	401	return;
be244f28 HW	402
b9a22dd9 UB	403	data_cdc = (struct smcd_cdc_msg *)conn->rmb_desc->cpu_addr;
	404	smcd_curs_copy(&cdc.prod, &data_cdc->prod, conn);
	405	smcd_curs_copy(&cdc.cons, &data_cdc->cons, conn);
be244f28 HW	406	smc = container_of(conn, struct smc_sock, conn);
	407	smc_cdc_msg_recv(smc, (struct smc_cdc_msg *)&cdc);
	408	}
	409
	410	/* Initialize receive tasklet. Called from ISM device IRQ handler to start
	411	* receiver side.
	412	*/
	413	void smcd_cdc_rx_init(struct smc_connection *conn)
	414	{
fcb8e3a3	415	tasklet_setup(&conn->rx_tsklet, smcd_cdc_rx_tsklet);
be244f28 HW	416	}
be244f28 HW	417
5f08318f UB	418	/*************************** init, exit, misc ****************************/
	419
	420	static void smc_cdc_rx_handler(struct ib_wc wc, void buf)
	421	{
	422	struct smc_link link = (struct smc_link )wc->qp->qp_context;
	423	struct smc_cdc_msg *cdc = buf;
d7b0e37c HW	424	struct smc_connection *conn;
	425	struct smc_link_group *lgr;
	426	struct smc_sock *smc;
5f08318f UB	427
	428	if (wc->byte_len < offsetof(struct smc_cdc_msg, reserved))
	429	return; /* short message */
cbba07a7	430	if (cdc->len != SMC_WR_TX_SIZE)
5f08318f	431	return; /* invalid message */
d7b0e37c HW	432
d7b0e37c HW	433	/* lookup connection */
00e5fb26	434	lgr = smc_get_lgr(link);
d7b0e37c HW	435	read_lock_bh(&lgr->conns_lock);
	436	conn = smc_lgr_find_conn(ntohl(cdc->token), lgr);
	437	read_unlock_bh(&lgr->conns_lock);
b286a065	438	if (!conn \|\| conn->out_of_sync)
d7b0e37c HW	439	return;
	440	smc = container_of(conn, struct smc_sock, conn);
	441
b286a065 KG	442	if (cdc->prod_flags.failover_validation) {
	443	smc_cdc_msg_validate(smc, cdc, link);
	444	return;
d7b0e37c	445	}
b286a065 KG	446	if (smc_cdc_before(ntohs(cdc->seqno),
	447	conn->local_rx_ctrl.seqno))
	448	/* received seqno is old */
	449	return;
	450
d7b0e37c	451	smc_cdc_msg_recv(smc, cdc);
5f08318f UB	452	}
	453
	454	static struct smc_wr_rx_handler smc_cdc_rx_handlers[] = {
	455	{
	456	.handler = smc_cdc_rx_handler,
	457	.type = SMC_CDC_MSG_TYPE
	458	},
	459	{
	460	.handler = NULL,
	461	}
	462	};
	463
	464	int __init smc_cdc_init(void)
	465	{
	466	struct smc_wr_rx_handler *handler;
	467	int rc = 0;
	468
	469	for (handler = smc_cdc_rx_handlers; handler->handler; handler++) {
	470	INIT_HLIST_NODE(&handler->list);
	471	rc = smc_wr_rx_register_handler(handler);
	472	if (rc)
	473	break;
	474	}
	475	return rc;
	476	}