[linux-block.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 <ubraun@linux.vnet.ibm.com>
 */

#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"
#include "smc_ism.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_buf_desc *sndbuf_desc;
	struct smc_sock *smc;
	int diff;

	sndbuf_desc = conn->sndbuf_desc;
	smc = container_of(conn, struct smc_sock, conn);
	bh_lock_sock(&smc->sk);
	if (!wc_status && sndbuf_desc) {
		diff = smc_curs_diff(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)) {
		/* If user owns the sock_lock, mark the connection need sending.
		 * User context will later try to send when it release sock_lock
		 * in smc_release_cb()
		 */
		if (sock_owned_by_user(&smc->sk))
			conn->tx_in_release_sock = true;
		else
			smc_tx_pending(conn);

		if (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 (!smc_conn_lgr_valid(conn) ||
	    (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;

	if (smc_ism_support_dmb_nocopy(conn->lgr->smcd))
		/* if local sndbuf shares the same memory region with
		 * peer DMB, then don't update the tx_curs_fin
		 * and sndbuf_space until peer has consumed the data.
		 */
		return 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 0;
}

/********************************* 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, diff_tx;

	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();

		/* if local sndbuf shares the same memory region with
		 * peer RMB, then update tx_curs_fin and sndbuf_space
		 * here since peer has already consumed the data.
		 */
		if (conn->lgr->is_smcd &&
		    smc_ism_support_dmb_nocopy(conn->lgr->smcd)) {
			/* Calculate consumed data and
			 * increment free send buffer space.
			 */
			diff_tx = smc_curs_diff(conn->sndbuf_desc->len,
						&conn->tx_curs_fin,
						&conn->local_rx_ctrl.cons);
			/* increase local sndbuf space and fin_curs */
			smp_mb__before_atomic();
			atomic_add(diff_tx, &conn->sndbuf_space);
			/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
			smp_mb__after_atomic();
			smc_curs_copy(&conn->tx_curs_fin,
				      &conn->local_rx_ctrl.cons, conn);

			smc_tx_sndbuf_nonfull(smc);
		}
	}

	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) {
		if (!sock_owned_by_user(&smc->sk))
			smc_tx_pending(conn);
		else
			conn->tx_in_release_sock = true;
	}

	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;
		smc_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
	1	// SPDX-License-Identifier: GPL-2.0
	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 <ubraun@linux.vnet.ibm.com>
	11	*/
	12
	13	#include <linux/spinlock.h>
	14
	15	#include "smc.h"
	16	#include "smc_wr.h"
	17	#include "smc_cdc.h"
	18	#include "smc_tx.h"
	19	#include "smc_rx.h"
	20	#include "smc_close.h"
	21	#include "smc_ism.h"
	22
	23	/******************************** send ***********************************/
	24
	25	/* handler for send/transmission completion of a CDC msg */
	26	static void smc_cdc_tx_handler(struct smc_wr_tx_pend_priv *pnd_snd,
	27	struct smc_link *link,
	28	enum ib_wc_status wc_status)
	29	{
	30	struct smc_cdc_tx_pend cdcpend = (struct smc_cdc_tx_pend )pnd_snd;
	31	struct smc_connection *conn = cdcpend->conn;
	32	struct smc_buf_desc *sndbuf_desc;
	33	struct smc_sock *smc;
	34	int diff;
	35
	36	sndbuf_desc = conn->sndbuf_desc;
	37	smc = container_of(conn, struct smc_sock, conn);
	38	bh_lock_sock(&smc->sk);
	39	if (!wc_status && sndbuf_desc) {
	40	diff = smc_curs_diff(sndbuf_desc->len,
	41	&cdcpend->conn->tx_curs_fin,
	42	&cdcpend->cursor);
	43	/* sndbuf_space is decreased in smc_sendmsg */
	44	smp_mb__before_atomic();
	45	atomic_add(diff, &cdcpend->conn->sndbuf_space);
	46	/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
	47	smp_mb__after_atomic();
	48	smc_curs_copy(&conn->tx_curs_fin, &cdcpend->cursor, conn);
	49	smc_curs_copy(&conn->local_tx_ctrl_fin, &cdcpend->p_cursor,
	50	conn);
	51	conn->tx_cdc_seq_fin = cdcpend->ctrl_seq;
	52	}
	53
	54	if (atomic_dec_and_test(&conn->cdc_pend_tx_wr)) {
	55	/* If user owns the sock_lock, mark the connection need sending.
	56	* User context will later try to send when it release sock_lock
	57	* in smc_release_cb()
	58	*/
	59	if (sock_owned_by_user(&smc->sk))
	60	conn->tx_in_release_sock = true;
	61	else
	62	smc_tx_pending(conn);
	63
	64	if (unlikely(wq_has_sleeper(&conn->cdc_pend_tx_wq)))
	65	wake_up(&conn->cdc_pend_tx_wq);
	66	}
	67	WARN_ON(atomic_read(&conn->cdc_pend_tx_wr) < 0);
	68
	69	smc_tx_sndbuf_nonfull(smc);
	70	bh_unlock_sock(&smc->sk);
	71	}
	72
	73	int smc_cdc_get_free_slot(struct smc_connection *conn,
	74	struct smc_link *link,
	75	struct smc_wr_buf **wr_buf,
	76	struct smc_rdma_wr **wr_rdma_buf,
	77	struct smc_cdc_tx_pend **pend)
	78	{
	79	int rc;
	80
	81	rc = smc_wr_tx_get_free_slot(link, smc_cdc_tx_handler, wr_buf,
	82	wr_rdma_buf,
	83	(struct smc_wr_tx_pend_priv **)pend);
	84	if (conn->killed) {
	85	/* abnormal termination */
	86	if (!rc)
	87	smc_wr_tx_put_slot(link,
	88	(struct smc_wr_tx_pend_priv )(pend));
	89	rc = -EPIPE;
	90	}
	91	return rc;
	92	}
	93
	94	static inline void smc_cdc_add_pending_send(struct smc_connection *conn,
	95	struct smc_cdc_tx_pend *pend)
	96	{
	97	BUILD_BUG_ON_MSG(
	98	sizeof(struct smc_cdc_msg) > SMC_WR_BUF_SIZE,
	99	"must increase SMC_WR_BUF_SIZE to at least sizeof(struct smc_cdc_msg)");
	100	BUILD_BUG_ON_MSG(
	101	offsetofend(struct smc_cdc_msg, reserved) > SMC_WR_TX_SIZE,
	102	"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()");
	103	BUILD_BUG_ON_MSG(
	104	sizeof(struct smc_cdc_tx_pend) > SMC_WR_TX_PEND_PRIV_SIZE,
	105	"must increase SMC_WR_TX_PEND_PRIV_SIZE to at least sizeof(struct smc_cdc_tx_pend)");
	106	pend->conn = conn;
	107	pend->cursor = conn->tx_curs_sent;
	108	pend->p_cursor = conn->local_tx_ctrl.prod;
	109	pend->ctrl_seq = conn->tx_cdc_seq;
	110	}
	111
	112	int smc_cdc_msg_send(struct smc_connection *conn,
	113	struct smc_wr_buf *wr_buf,
	114	struct smc_cdc_tx_pend *pend)
	115	{
	116	struct smc_link *link = conn->lnk;
	117	union smc_host_cursor cfed;
	118	int rc;
	119
	120	smc_cdc_add_pending_send(conn, pend);
	121
	122	conn->tx_cdc_seq++;
	123	conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
	124	smc_host_msg_to_cdc((struct smc_cdc_msg *)wr_buf, conn, &cfed);
	125
	126	atomic_inc(&conn->cdc_pend_tx_wr);
	127	smp_mb__after_atomic(); /* Make sure cdc_pend_tx_wr added before post */
	128
	129	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
	130	if (!rc) {
	131	smc_curs_copy(&conn->rx_curs_confirmed, &cfed, conn);
	132	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
	133	} else {
	134	conn->tx_cdc_seq--;
	135	conn->local_tx_ctrl.seqno = conn->tx_cdc_seq;
	136	atomic_dec(&conn->cdc_pend_tx_wr);
	137	}
	138
	139	return rc;
	140	}
	141
	142	/* send a validation msg indicating the move of a conn to an other QP link */
	143	int smcr_cdc_msg_send_validation(struct smc_connection *conn,
	144	struct smc_cdc_tx_pend *pend,
	145	struct smc_wr_buf *wr_buf)
	146	{
	147	struct smc_host_cdc_msg *local = &conn->local_tx_ctrl;
	148	struct smc_link *link = conn->lnk;
	149	struct smc_cdc_msg *peer;
	150	int rc;
	151
	152	peer = (struct smc_cdc_msg *)wr_buf;
	153	peer->common.type = local->common.type;
	154	peer->len = local->len;
	155	peer->seqno = htons(conn->tx_cdc_seq_fin); /* seqno last compl. tx */
	156	peer->token = htonl(local->token);
	157	peer->prod_flags.failover_validation = 1;
	158
	159	/* We need to set pend->conn here to make sure smc_cdc_tx_handler()
	160	* can handle properly
	161	*/
	162	smc_cdc_add_pending_send(conn, pend);
	163
	164	atomic_inc(&conn->cdc_pend_tx_wr);
	165	smp_mb__after_atomic(); /* Make sure cdc_pend_tx_wr added before post */
	166
	167	rc = smc_wr_tx_send(link, (struct smc_wr_tx_pend_priv *)pend);
	168	if (unlikely(rc))
	169	atomic_dec(&conn->cdc_pend_tx_wr);
	170
	171	return rc;
	172	}
	173
	174	static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
	175	{
	176	struct smc_cdc_tx_pend *pend;
	177	struct smc_wr_buf *wr_buf;
	178	struct smc_link *link;
	179	bool again = false;
	180	int rc;
	181
	182	again:
	183	link = conn->lnk;
	184	if (!smc_wr_tx_link_hold(link))
	185	return -ENOLINK;
	186	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, NULL, &pend);
	187	if (rc)
	188	goto put_out;
	189
	190	spin_lock_bh(&conn->send_lock);
	191	if (link != conn->lnk) {
	192	/* link of connection changed, try again one time*/
	193	spin_unlock_bh(&conn->send_lock);
	194	smc_wr_tx_put_slot(link,
	195	(struct smc_wr_tx_pend_priv *)pend);
	196	smc_wr_tx_link_put(link);
	197	if (again)
	198	return -ENOLINK;
	199	again = true;
	200	goto again;
	201	}
	202	rc = smc_cdc_msg_send(conn, wr_buf, pend);
	203	spin_unlock_bh(&conn->send_lock);
	204	put_out:
	205	smc_wr_tx_link_put(link);
	206	return rc;
	207	}
	208
	209	int smc_cdc_get_slot_and_msg_send(struct smc_connection *conn)
	210	{
	211	int rc;
	212
	213	if (!smc_conn_lgr_valid(conn) \|\|
	214	(conn->lgr->is_smcd && conn->lgr->peer_shutdown))
	215	return -EPIPE;
	216
	217	if (conn->lgr->is_smcd) {
	218	spin_lock_bh(&conn->send_lock);
	219	rc = smcd_cdc_msg_send(conn);
	220	spin_unlock_bh(&conn->send_lock);
	221	} else {
	222	rc = smcr_cdc_get_slot_and_msg_send(conn);
	223	}
	224
	225	return rc;
	226	}
	227
	228	void smc_cdc_wait_pend_tx_wr(struct smc_connection *conn)
	229	{
	230	wait_event(conn->cdc_pend_tx_wq, !atomic_read(&conn->cdc_pend_tx_wr));
	231	}
	232
	233	/* Send a SMC-D CDC header.
	234	* This increments the free space available in our send buffer.
	235	* Also update the confirmed receive buffer with what was sent to the peer.
	236	*/
	237	int smcd_cdc_msg_send(struct smc_connection *conn)
	238	{
	239	struct smc_sock *smc = container_of(conn, struct smc_sock, conn);
	240	union smc_host_cursor curs;
	241	struct smcd_cdc_msg cdc;
	242	int rc, diff;
	243
	244	memset(&cdc, 0, sizeof(cdc));
	245	cdc.common.type = SMC_CDC_MSG_TYPE;
	246	curs.acurs.counter = atomic64_read(&conn->local_tx_ctrl.prod.acurs);
	247	cdc.prod.wrap = curs.wrap;
	248	cdc.prod.count = curs.count;
	249	curs.acurs.counter = atomic64_read(&conn->local_tx_ctrl.cons.acurs);
	250	cdc.cons.wrap = curs.wrap;
	251	cdc.cons.count = curs.count;
	252	cdc.cons.prod_flags = conn->local_tx_ctrl.prod_flags;
	253	cdc.cons.conn_state_flags = conn->local_tx_ctrl.conn_state_flags;
	254	rc = smcd_tx_ism_write(conn, &cdc, sizeof(cdc), 0, 1);
	255	if (rc)
	256	return rc;
	257	smc_curs_copy(&conn->rx_curs_confirmed, &curs, conn);
	258	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req = 0;
	259
	260	if (smc_ism_support_dmb_nocopy(conn->lgr->smcd))
	261	/* if local sndbuf shares the same memory region with
	262	* peer DMB, then don't update the tx_curs_fin
	263	* and sndbuf_space until peer has consumed the data.
	264	*/
	265	return 0;
	266
	267	/* Calculate transmitted data and increment free send buffer space */
	268	diff = smc_curs_diff(conn->sndbuf_desc->len, &conn->tx_curs_fin,
	269	&conn->tx_curs_sent);
	270	/* increased by confirmed number of bytes */
	271	smp_mb__before_atomic();
	272	atomic_add(diff, &conn->sndbuf_space);
	273	/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
	274	smp_mb__after_atomic();
	275	smc_curs_copy(&conn->tx_curs_fin, &conn->tx_curs_sent, conn);
	276
	277	smc_tx_sndbuf_nonfull(smc);
	278	return 0;
	279	}
	280
	281	/******************************* receive *********************************/
	282
	283	static inline bool smc_cdc_before(u16 seq1, u16 seq2)
	284	{
	285	return (s16)(seq1 - seq2) < 0;
	286	}
	287
	288	static void smc_cdc_handle_urg_data_arrival(struct smc_sock *smc,
	289	int *diff_prod)
	290	{
	291	struct smc_connection *conn = &smc->conn;
	292	char *base;
	293
	294	/* new data included urgent business */
	295	smc_curs_copy(&conn->urg_curs, &conn->local_rx_ctrl.prod, conn);
	296	conn->urg_state = SMC_URG_VALID;
	297	if (!sock_flag(&smc->sk, SOCK_URGINLINE))
	298	/* we'll skip the urgent byte, so don't account for it */
	299	(*diff_prod)--;
	300	base = (char *)conn->rmb_desc->cpu_addr + conn->rx_off;
	301	if (conn->urg_curs.count)
	302	conn->urg_rx_byte = *(base + conn->urg_curs.count - 1);
	303	else
	304	conn->urg_rx_byte = *(base + conn->rmb_desc->len - 1);
	305	sk_send_sigurg(&smc->sk);
	306	}
	307
	308	static void smc_cdc_msg_validate(struct smc_sock smc, struct smc_cdc_msg cdc,
	309	struct smc_link *link)
	310	{
	311	struct smc_connection *conn = &smc->conn;
	312	u16 recv_seq = ntohs(cdc->seqno);
	313	s16 diff;
	314
	315	/* check that seqnum was seen before */
	316	diff = conn->local_rx_ctrl.seqno - recv_seq;
	317	if (diff < 0) { /* diff larger than 0x7fff */
	318	/* drop connection */
	319	conn->out_of_sync = 1; /* prevent any further receives */
	320	spin_lock_bh(&conn->send_lock);
	321	conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
	322	conn->lnk = link;
	323	spin_unlock_bh(&conn->send_lock);
	324	sock_hold(&smc->sk); /* sock_put in abort_work */
	325	if (!queue_work(smc_close_wq, &conn->abort_work))
	326	sock_put(&smc->sk);
	327	}
	328	}
	329
	330	static void smc_cdc_msg_recv_action(struct smc_sock *smc,
	331	struct smc_cdc_msg *cdc)
	332	{
	333	union smc_host_cursor cons_old, prod_old;
	334	struct smc_connection *conn = &smc->conn;
	335	int diff_cons, diff_prod, diff_tx;
	336
	337	smc_curs_copy(&prod_old, &conn->local_rx_ctrl.prod, conn);
	338	smc_curs_copy(&cons_old, &conn->local_rx_ctrl.cons, conn);
	339	smc_cdc_msg_to_host(&conn->local_rx_ctrl, cdc, conn);
	340
	341	diff_cons = smc_curs_diff(conn->peer_rmbe_size, &cons_old,
	342	&conn->local_rx_ctrl.cons);
	343	if (diff_cons) {
	344	/* peer_rmbe_space is decreased during data transfer with RDMA
	345	* write
	346	*/
	347	smp_mb__before_atomic();
	348	atomic_add(diff_cons, &conn->peer_rmbe_space);
	349	/* guarantee 0 <= peer_rmbe_space <= peer_rmbe_size */
	350	smp_mb__after_atomic();
	351
	352	/* if local sndbuf shares the same memory region with
	353	* peer RMB, then update tx_curs_fin and sndbuf_space
	354	* here since peer has already consumed the data.
	355	*/
	356	if (conn->lgr->is_smcd &&
	357	smc_ism_support_dmb_nocopy(conn->lgr->smcd)) {
	358	/* Calculate consumed data and
	359	* increment free send buffer space.
	360	*/
	361	diff_tx = smc_curs_diff(conn->sndbuf_desc->len,
	362	&conn->tx_curs_fin,
	363	&conn->local_rx_ctrl.cons);
	364	/* increase local sndbuf space and fin_curs */
	365	smp_mb__before_atomic();
	366	atomic_add(diff_tx, &conn->sndbuf_space);
	367	/* guarantee 0 <= sndbuf_space <= sndbuf_desc->len */
	368	smp_mb__after_atomic();
	369	smc_curs_copy(&conn->tx_curs_fin,
	370	&conn->local_rx_ctrl.cons, conn);
	371
	372	smc_tx_sndbuf_nonfull(smc);
	373	}
	374	}
	375
	376	diff_prod = smc_curs_diff(conn->rmb_desc->len, &prod_old,
	377	&conn->local_rx_ctrl.prod);
	378	if (diff_prod) {
	379	if (conn->local_rx_ctrl.prod_flags.urg_data_present)
	380	smc_cdc_handle_urg_data_arrival(smc, &diff_prod);
	381	/* bytes_to_rcv is decreased in smc_recvmsg */
	382	smp_mb__before_atomic();
	383	atomic_add(diff_prod, &conn->bytes_to_rcv);
	384	/* guarantee 0 <= bytes_to_rcv <= rmb_desc->len */
	385	smp_mb__after_atomic();
	386	smc->sk.sk_data_ready(&smc->sk);
	387	} else {
	388	if (conn->local_rx_ctrl.prod_flags.write_blocked)
	389	smc->sk.sk_data_ready(&smc->sk);
	390	if (conn->local_rx_ctrl.prod_flags.urg_data_pending)
	391	conn->urg_state = SMC_URG_NOTYET;
	392	}
	393
	394	/* trigger sndbuf consumer: RDMA write into peer RMBE and CDC */
	395	if ((diff_cons && smc_tx_prepared_sends(conn)) \|\|
	396	conn->local_rx_ctrl.prod_flags.cons_curs_upd_req \|\|
	397	conn->local_rx_ctrl.prod_flags.urg_data_pending) {
	398	if (!sock_owned_by_user(&smc->sk))
	399	smc_tx_pending(conn);
	400	else
	401	conn->tx_in_release_sock = true;
	402	}
	403
	404	if (diff_cons && conn->urg_tx_pend &&
	405	atomic_read(&conn->peer_rmbe_space) == conn->peer_rmbe_size) {
	406	/* urg data confirmed by peer, indicate we're ready for more */
	407	conn->urg_tx_pend = false;
	408	smc->sk.sk_write_space(&smc->sk);
	409	}
	410
	411	if (conn->local_rx_ctrl.conn_state_flags.peer_conn_abort) {
	412	smc->sk.sk_err = ECONNRESET;
	413	conn->local_tx_ctrl.conn_state_flags.peer_conn_abort = 1;
	414	}
	415	if (smc_cdc_rxed_any_close_or_senddone(conn)) {
	416	smc->sk.sk_shutdown \|= RCV_SHUTDOWN;
	417	if (smc->clcsock && smc->clcsock->sk)
	418	smc->clcsock->sk->sk_shutdown \|= RCV_SHUTDOWN;
	419	smc_sock_set_flag(&smc->sk, SOCK_DONE);
	420	sock_hold(&smc->sk); /* sock_put in close_work */
	421	if (!queue_work(smc_close_wq, &conn->close_work))
	422	sock_put(&smc->sk);
	423	}
	424	}
	425
	426	/* called under tasklet context */
	427	static void smc_cdc_msg_recv(struct smc_sock smc, struct smc_cdc_msg cdc)
	428	{
	429	sock_hold(&smc->sk);
	430	bh_lock_sock(&smc->sk);
	431	smc_cdc_msg_recv_action(smc, cdc);
	432	bh_unlock_sock(&smc->sk);
	433	sock_put(&smc->sk); /* no free sk in softirq-context */
	434	}
	435
	436	/* Schedule a tasklet for this connection. Triggered from the ISM device IRQ
	437	* handler to indicate update in the DMBE.
	438	*
	439	* Context:
	440	* - tasklet context
	441	*/
	442	static void smcd_cdc_rx_tsklet(struct tasklet_struct *t)
	443	{
	444	struct smc_connection *conn = from_tasklet(conn, t, rx_tsklet);
	445	struct smcd_cdc_msg *data_cdc;
	446	struct smcd_cdc_msg cdc;
	447	struct smc_sock *smc;
	448
	449	if (!conn \|\| conn->killed)
	450	return;
	451
	452	data_cdc = (struct smcd_cdc_msg *)conn->rmb_desc->cpu_addr;
	453	smcd_curs_copy(&cdc.prod, &data_cdc->prod, conn);
	454	smcd_curs_copy(&cdc.cons, &data_cdc->cons, conn);
	455	smc = container_of(conn, struct smc_sock, conn);
	456	smc_cdc_msg_recv(smc, (struct smc_cdc_msg *)&cdc);
	457	}
	458
	459	/* Initialize receive tasklet. Called from ISM device IRQ handler to start
	460	* receiver side.
	461	*/
	462	void smcd_cdc_rx_init(struct smc_connection *conn)
	463	{
	464	tasklet_setup(&conn->rx_tsklet, smcd_cdc_rx_tsklet);
	465	}
	466
	467	/*************************** init, exit, misc ****************************/
	468
	469	static void smc_cdc_rx_handler(struct ib_wc wc, void buf)
	470	{
	471	struct smc_link link = (struct smc_link )wc->qp->qp_context;
	472	struct smc_cdc_msg *cdc = buf;
	473	struct smc_connection *conn;
	474	struct smc_link_group *lgr;
	475	struct smc_sock *smc;
	476
	477	if (wc->byte_len < offsetof(struct smc_cdc_msg, reserved))
	478	return; /* short message */
	479	if (cdc->len != SMC_WR_TX_SIZE)
	480	return; /* invalid message */
	481
	482	/* lookup connection */
	483	lgr = smc_get_lgr(link);
	484	read_lock_bh(&lgr->conns_lock);
	485	conn = smc_lgr_find_conn(ntohl(cdc->token), lgr);
	486	read_unlock_bh(&lgr->conns_lock);
	487	if (!conn \|\| conn->out_of_sync)
	488	return;
	489	smc = container_of(conn, struct smc_sock, conn);
	490
	491	if (cdc->prod_flags.failover_validation) {
	492	smc_cdc_msg_validate(smc, cdc, link);
	493	return;
	494	}
	495	if (smc_cdc_before(ntohs(cdc->seqno),
	496	conn->local_rx_ctrl.seqno))
	497	/* received seqno is old */
	498	return;
	499
	500	smc_cdc_msg_recv(smc, cdc);
	501	}
	502
	503	static struct smc_wr_rx_handler smc_cdc_rx_handlers[] = {
	504	{
	505	.handler = smc_cdc_rx_handler,
	506	.type = SMC_CDC_MSG_TYPE
	507	},
	508	{
	509	.handler = NULL,
	510	}
	511	};
	512
	513	int __init smc_cdc_init(void)
	514	{
	515	struct smc_wr_rx_handler *handler;
	516	int rc = 0;
	517
	518	for (handler = smc_cdc_rx_handlers; handler->handler; handler++) {
	519	INIT_HLIST_NODE(&handler->list);
	520	rc = smc_wr_rx_register_handler(handler);
	521	if (rc)
	522	break;
	523	}
	524	return rc;
	525	}