Lustre Repository Browser

Viewing: socklnd_lib.c

// SPDX-License-Identifier: GPL-2.0

/* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 2011, 2017, Intel Corporation.
 */

/* This file is part of Lustre, http://www.lustre.org/ */

#include "socklnd.h"

int
ksocknal_lib_get_conn_addrs(struct ksock_conn *conn)
{
	int rc = lnet_sock_getaddr(conn->ksnc_sock, true,
				   &conn->ksnc_peeraddr);

	/* Didn't need the {get,put}connsock dance to deref ksnc_sock... */
	LASSERT(!conn->ksnc_closing);

	if (rc != 0) {
		CERROR("Error getting sock peer_ni IP: rc = %d\n", rc);
		return rc;
	}

	rc = lnet_sock_getaddr(conn->ksnc_sock, false,
			       &conn->ksnc_myaddr);
	if (rc != 0) {
		CERROR("Error getting sock local IP: rc = %d\n", rc);
		return rc;
	}

	return 0;
}

int
ksocknal_lib_zc_capable(struct ksock_conn *conn)
{
	int  caps = conn->ksnc_sock->sk->sk_route_caps;

	if (conn->ksnc_proto == &ksocknal_protocol_v1x)
		return 0;

	/* ZC if the socket supports scatter/gather and doesn't need software
	 * checksums
	 */
	return ((caps & NETIF_F_SG) != 0 && (caps & NETIF_F_CSUM_MASK) != 0);
}

int
ksocknal_lib_send_hdr(struct ksock_conn *conn, struct ksock_tx *tx)
{
	struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
	struct socket *sock = conn->ksnc_sock;
	int nob = 0;

	if (*ksocknal_tunables.ksnd_enable_csum	&&	  /* checksum enabled */
	    conn->ksnc_proto == &ksocknal_protocol_v2x && /* V2.x connection  */
	    tx->tx_nob == tx->tx_resid		       && /* frist sending    */
	    tx->tx_msg.ksm_csum == 0)			  /* not checksummed  */
		ksocknal_lib_csum_tx(tx);

	if (tx->tx_niov)
		nob += tx->tx_hdr.iov_len;

	if (!list_empty(&conn->ksnc_tx_queue) ||
	    nob < tx->tx_resid)
		msg.msg_flags |= MSG_MORE;

	iov_iter_kvec(&msg.msg_iter, WRITE,
		      &tx->tx_hdr, tx->tx_niov, nob);

	return sock_sendmsg(sock, &msg);
}

static int
ksocknal_lib_sendpage(struct socket *sock, struct bio_vec *kiov,
		      int nkiov, int msgflg)
{
#ifdef MSG_SPLICE_PAGES /* added to kernel 6.4.0-rc2 */
	struct msghdr msg = {.msg_flags = msgflg | MSG_SPLICE_PAGES};

	iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, kiov, 1, kiov->bv_len);

	return sock_sendmsg(sock, &msg);
#else
	struct sock *sk = sock->sk;
	struct page *page = kiov->bv_page;
	int offset = kiov->bv_offset;
	int fragsize = kiov->bv_len;

	return sk->sk_prot->sendpage(sk, page, offset, fragsize, msgflg);
#endif
}

int
ksocknal_lib_send_kiov(struct ksock_conn *conn, struct ksock_tx *tx)
{
	struct socket *sock = conn->ksnc_sock;
	struct bio_vec *kiov = tx->tx_kiov;
	int rc;
	int nob;

	/* Not NOOP message */
	LASSERT(tx->tx_lnetmsg);

	/* can't trust socket ops to consume our iovs or leave them alone */
	if (tx->tx_msg.ksm_zc_cookies[0] != 0) {
		/* Zero copy is enabled */
		int msgflg = MSG_DONTWAIT;

		CDEBUG(D_NET, "page %p + offset %x for %d\n",
		       kiov->bv_page, kiov->bv_offset, kiov->bv_len);

		if (!list_empty(&conn->ksnc_tx_queue) ||
		    kiov->bv_len < tx->tx_resid)
			msgflg |= MSG_MORE;

		rc = ksocknal_lib_sendpage(sock, kiov, tx->tx_nkiov, msgflg);
	} else {
		struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
		int i;

		for (nob = i = 0; i < tx->tx_nkiov; i++)
			nob += kiov[i].bv_len;

		if (!list_empty(&conn->ksnc_tx_queue) ||
		    nob < tx->tx_resid)
			msg.msg_flags |= MSG_MORE;

		iov_iter_bvec(&msg.msg_iter, WRITE,
			      kiov, tx->tx_nkiov, nob);
		rc = sock_sendmsg(sock, &msg);
	}
	return rc;
}

void
ksocknal_lib_eager_ack(struct ksock_conn *conn)
{
	struct socket *sock = conn->ksnc_sock;

	/* Remind the socket to ACK eagerly.  If I don't, the socket might
	 * think I'm about to send something it could piggy-back the ACK on,
	 * introducing delay in completing zero-copy sends in my peer_ni.
	 */
	tcp_sock_set_quickack(sock->sk, 1);
}

static int lustre_csum(struct kvec *v, void *context)
{
	struct ksock_conn *conn = context;

	conn->ksnc_rx_csum = crc32_le(conn->ksnc_rx_csum,
				      v->iov_base, v->iov_len);
	return 0;
}

static int ksocknal_iter_for_each_range(struct iov_iter *i, size_t bytes,
					int (*f)(struct kvec *vec,
						 void *context),
					void *context)
{
	size_t skip = i->iov_offset;
	int err = -EINVAL;
	struct kvec w;

	if (!bytes)
		return 0;

	if (iov_iter_type(i) == ITER_BVEC) {
		struct bvec_iter start = {
			.bi_size	= bytes,
			.bi_bvec_done	= skip,
			.bi_idx = 0,
		};
		struct bvec_iter bi;
		struct bio_vec v;

		for_each_bvec(v, i->bvec, bi, start) {
			if (!v.bv_len)
				continue;

			w.iov_base = kmap(v.bv_page) + v.bv_offset;
			w.iov_len = v.bv_len;
			err = f(&w, context);
			kunmap(v.bv_page);
		}
	} else if (iov_iter_type(i) == ITER_KVEC) {
		const struct kvec *kvec = i->kvec;
		size_t wanted = bytes;
		struct kvec v;

		v.iov_len = min(bytes, kvec->iov_len - skip);
		if (likely(v.iov_len)) {
			v.iov_base = kvec->iov_base + skip;
			err = f(&v, context);
			skip += v.iov_len;
			bytes -= v.iov_len;
		}
		while (unlikely(bytes)) {
			kvec++;
			v.iov_len = min(bytes, kvec->iov_len);
			if (unlikely(!v.iov_len))
				continue;
			v.iov_base = kvec->iov_base;
			err = f(&v, context);
			skip = v.iov_len;
			bytes -= v.iov_len;
		}
		bytes = wanted;
	}
	return err;
}

int
ksocknal_lib_recv(struct ksock_conn *conn)
{
	struct msghdr msg = { .msg_iter = conn->ksnc_rx_to };
	u32 saved_csum;
	int rc;

	rc = sock_recvmsg(conn->ksnc_sock, &msg, MSG_DONTWAIT);
	if (rc <= 0)
		return rc;

	saved_csum = conn->ksnc_msg.ksm_csum;
	if (!saved_csum)
		return rc;

	/* header is included only in V2 - V3 checksums only the bulk data */
	if (iov_iter_type(&conn->ksnc_rx_to) != ITER_BVEC &&
	    conn->ksnc_proto != &ksocknal_protocol_v2x)
		return rc;

	/* accumulate checksum */
	conn->ksnc_msg.ksm_csum = 0;
	ksocknal_iter_for_each_range(&conn->ksnc_rx_to, rc, lustre_csum, conn);
	conn->ksnc_msg.ksm_csum = saved_csum;

	return rc;
}

void
ksocknal_lib_csum_tx(struct ksock_tx *tx)
{
	int i;
	__u32 csum;
	void *base;

	LASSERT(tx->tx_hdr.iov_base == (void *)&tx->tx_msg);
	LASSERT(tx->tx_conn != NULL);
	LASSERT(tx->tx_conn->ksnc_proto == &ksocknal_protocol_v2x);

	tx->tx_msg.ksm_csum = 0;

	csum = ksocknal_csum(~0, (void *)tx->tx_hdr.iov_base,
			     tx->tx_hdr.iov_len);

	for (i = 0; i < tx->tx_nkiov; i++) {
		void *kaddr = kmap_local_page(tx->tx_kiov[i].bv_page);

		base = kaddr + tx->tx_kiov[i].bv_offset;
		csum = ksocknal_csum(csum, base, tx->tx_kiov[i].bv_len);
		kunmap_local(kaddr);
	}

	if (*ksocknal_tunables.ksnd_inject_csum_error) {
		csum++;
		*ksocknal_tunables.ksnd_inject_csum_error = 0;
	}

	tx->tx_msg.ksm_csum = csum;
}

int
ksocknal_lib_get_conn_tunables(struct ksock_conn *conn, int *txmem,
			       int *rxmem, int *nagle)
{
	struct socket *sock = conn->ksnc_sock;
	struct tcp_sock *tp = tcp_sk(sock->sk);

	if (ksocknal_connsock_addref(conn) < 0) {
		LASSERT(conn->ksnc_closing);
		*txmem = 0;
		*rxmem = 0;
		*nagle = 0;
		return -ESHUTDOWN;
	}

	lnet_sock_getbuf(sock, txmem, rxmem);

	*nagle = !(tp->nonagle & TCP_NAGLE_OFF);

	ksocknal_connsock_decref(conn);

	return 0;
}

int
ksocknal_lib_setup_sock(struct socket *sock, struct lnet_ni *ni)
{
	int rc;
	int keep_idle;
	int keep_intvl;
	int keep_count;
	int do_keepalive;
	struct tcp_sock *tp = tcp_sk(sock->sk);
	struct lnet_ioctl_config_socklnd_tunables *lndtun;

	sock->sk->sk_allocation = GFP_NOFS;

	/* Ensure this socket aborts active sends immediately when closed. */
	sock_reset_flag(sock->sk, SOCK_LINGER);

	tp->linger2 = -1;

	if (!*ksocknal_tunables.ksnd_nagle)
		tcp_sock_set_nodelay(sock->sk);

	lnet_sock_setbuf(sock,
			 *ksocknal_tunables.ksnd_tx_buffer_size,
			 *ksocknal_tunables.ksnd_rx_buffer_size);

/* TCP_BACKOFF_* sockopt tunables unsupported in stock kernels */
#ifdef SOCKNAL_BACKOFF
	if (*ksocknal_tunables.ksnd_backoff_init > 0) {
		int option = *ksocknal_tunables.ksnd_backoff_init;
#ifdef SOCKNAL_BACKOFF_MS
		option *= 1000;
#endif

		rc = kernel_setsockopt(sock, SOL_TCP, TCP_BACKOFF_INIT,
				       (char *)&option, sizeof(option));
		if (rc != 0) {
			CERROR("Can't set initial tcp backoff %d: %d\n",
			       option, rc);
			return rc;
		}
	}

	if (*ksocknal_tunables.ksnd_backoff_max > 0) {
		int option = *ksocknal_tunables.ksnd_backoff_max;
#ifdef SOCKNAL_BACKOFF_MS
		option *= 1000;
#endif

		rc = kernel_setsockopt(sock, SOL_TCP, TCP_BACKOFF_MAX,
				       (char *)&option, sizeof(option));
		if (rc != 0) {
			CERROR("Can't set maximum tcp backoff %d: %d\n",
			       option, rc);
			return rc;
		}
	}
#endif

	/* snapshot tunables */
	keep_idle  = *ksocknal_tunables.ksnd_keepalive_idle;
	keep_count = *ksocknal_tunables.ksnd_keepalive_count;
	keep_intvl = *ksocknal_tunables.ksnd_keepalive_intvl;

	do_keepalive = (keep_idle > 0 && keep_count > 0 && keep_intvl > 0);

#ifdef HAVE_KERNEL_SETSOCKOPT
	/* open-coded version doesn't work in all kernels, and
	 * there is no helper function, so call kernel_setsockopt()
	 * directly.
	 */
	{
		int option = (do_keepalive ? 1 : 0);

		kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
				  (char *)&option, sizeof(option));
	}
#else
	if (sock->sk->sk_prot->keepalive)
		sock->sk->sk_prot->keepalive(sock->sk, do_keepalive);
	if (do_keepalive)
		sock_set_flag(sock->sk, SOCK_KEEPOPEN);
	else
		sock_reset_flag(sock->sk, SOCK_KEEPOPEN);
#endif /* HAVE_KERNEL_SETSOCKOPT */

	if (!do_keepalive)
		return (0);

	rc = tcp_sock_set_keepidle(sock->sk, keep_idle);
	if (rc != 0) {
		CERROR("Can't set TCP_KEEPIDLE: %d\n", rc);
		return rc;
	}

	rc = tcp_sock_set_keepintvl(sock->sk, keep_intvl);
	if (rc != 0) {
		CERROR("Can't set TCP_KEEPINTVL: %d\n", rc);
		return rc;
	}

	rc = tcp_sock_set_keepcnt(sock->sk, keep_count);
	if (rc != 0) {
		CERROR("Can't set TCP_KEEPCNT: %d\n", rc);
		return rc;
	}

	lndtun = &ni->ni_lnd_tunables.lnd_tun_u.lnd_sock;
	if (lndtun->lnd_tos >= 0)
		ip_sock_set_tos(sock->sk, lndtun->lnd_tos);

	return 0;
}

void
ksocknal_lib_push_conn(struct ksock_conn *conn)
{
	struct sock *sk;
	struct tcp_sock *tp;
	int nonagle;
	int rc;

	rc = ksocknal_connsock_addref(conn);
	if (rc != 0)                            /* being shut down */
		return;

	sk = conn->ksnc_sock->sk;
	tp = tcp_sk(sk);

	lock_sock(sk);
	nonagle = tp->nonagle;
	tp->nonagle = TCP_NAGLE_OFF;
	release_sock(sk);

	tcp_sock_set_nodelay(conn->ksnc_sock->sk);

	lock_sock(sk);
	tp->nonagle = nonagle;
	release_sock(sk);

	ksocknal_connsock_decref(conn);
}

void ksocknal_read_callback(struct ksock_conn *conn);
void ksocknal_write_callback(struct ksock_conn *conn);
/*
 * socket call back in Linux
 */
static void
ksocknal_data_ready(struct sock *sk)
{
	struct ksock_conn  *conn;

	/* interleave correctly with closing sockets... */
	LASSERT(!in_hardirq());
	read_lock_bh(&ksocknal_data.ksnd_global_lock);

	conn = sk->sk_user_data;
	if (conn == NULL) {	/* raced with ksocknal_terminate_conn */
		LASSERT(sk->sk_data_ready != &ksocknal_data_ready);
		sk->sk_data_ready(sk);
	} else
		ksocknal_read_callback(conn);

	read_unlock_bh(&ksocknal_data.ksnd_global_lock);
}

static void
ksocknal_write_space(struct sock *sk)
{
	struct ksock_conn *conn;
	int wspace;
	int min_wpace;

	/* interleave correctly with closing sockets... */
	LASSERT(!in_hardirq());
	read_lock(&ksocknal_data.ksnd_global_lock);

	conn = sk->sk_user_data;
	wspace = sk_stream_wspace(sk);
	min_wpace = sk_stream_min_wspace(sk);

	CDEBUG(D_NET, "sk %p wspace %d low water %d conn %p%s%s%s\n",
	       sk, wspace, min_wpace, conn,
	       (conn == NULL) ? "" : (conn->ksnc_tx_ready ?
				      " ready" : " blocked"),
	       (conn == NULL) ? "" : (conn->ksnc_tx_scheduled ?
				      " scheduled" : " idle"),
	       (conn == NULL) ? "" : (list_empty(&conn->ksnc_tx_queue) ?
				      " empty" : " queued"));

	if (conn == NULL) {             /* raced with ksocknal_terminate_conn */
		LASSERT(sk->sk_write_space != &ksocknal_write_space);
		sk->sk_write_space(sk);

		read_unlock(&ksocknal_data.ksnd_global_lock);
		return;
	}

	if (wspace >= min_wpace) { /* got enough space */
		ksocknal_write_callback(conn);

		/* Clear SOCK_NOSPACE _after_ ksocknal_write_callback so the
		 * ENOMEM check in ksocknal_transmit is race-free (think!).
		 */
		clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
	}

	read_unlock(&ksocknal_data.ksnd_global_lock);
}

void
ksocknal_lib_save_callback(struct socket *sock, struct ksock_conn *conn)
{
	conn->ksnc_saved_data_ready = sock->sk->sk_data_ready;
	conn->ksnc_saved_write_space = sock->sk->sk_write_space;
}

void
ksocknal_lib_set_callback(struct socket *sock,  struct ksock_conn *conn)
{
	sock->sk->sk_user_data = conn;
	sock->sk->sk_data_ready = ksocknal_data_ready;
	sock->sk->sk_write_space = ksocknal_write_space;
}

void
ksocknal_lib_reset_callback(struct socket *sock, struct ksock_conn *conn)
{
	/* Remove conn's network callbacks.
	 * NB I _have_ to restore the callback, rather than storing a noop,
	 * since the socket could survive past this module being unloaded!!
	 */
	sock->sk->sk_data_ready = conn->ksnc_saved_data_ready;
	sock->sk->sk_write_space = conn->ksnc_saved_write_space;

	/* A callback could be in progress already; they hold a read lock
	 * on ksnd_global_lock (to serialise with me) and NOOP if
	 * sk_user_data is NULL.
	 */
	sock->sk->sk_user_data = NULL;
}

int
ksocknal_lib_memory_pressure(struct ksock_conn *conn)
{
	struct ksock_sched *sched;
	int rc = 0;

	sched = conn->ksnc_scheduler;
	spin_lock_bh(&sched->kss_lock);

	if (!test_bit(SOCK_NOSPACE, &conn->ksnc_sock->flags) &&
	    !conn->ksnc_tx_ready) {
		/* SOCK_NOSPACE is set when the socket fills and cleared in the
		 * write_space callback (which also sets ksnc_tx_ready). If
		 * SOCK_NOSPACE and ksnc_tx_ready are BOTH zero, I didn't fill
		 * the socket and write_space won't reschedule me, so I return
		 * -ENOMEM to get my caller to retry after a timeout
		 */
		rc = -ENOMEM;
	}

	spin_unlock_bh(&sched->kss_lock);

	return rc;
}