*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#include "socklnd.h"
-# if defined(CONFIG_SYSCTL) && !CFS_SYSFS_MODULE_PARM
-
-static struct ctl_table ksocknal_ctl_table[] = {
- {
- INIT_CTL_NAME
- .procname = "timeout",
- .data = &ksocknal_tunables.ksnd_timeout,
- .maxlen = sizeof (int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "credits",
- .data = &ksocknal_tunables.ksnd_credits,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "peer_credits",
- .data = &ksocknal_tunables.ksnd_peertxcredits,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "peer_buffer_credits",
- .data = &ksocknal_tunables.ksnd_peerrtrcredits,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "peer_timeout",
- .data = &ksocknal_tunables.ksnd_peertimeout,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "nconnds",
- .data = &ksocknal_tunables.ksnd_nconnds,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "min_reconnectms",
- .data = &ksocknal_tunables.ksnd_min_reconnectms,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "max_reconnectms",
- .data = &ksocknal_tunables.ksnd_max_reconnectms,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "eager_ack",
- .data = &ksocknal_tunables.ksnd_eager_ack,
- .maxlen = sizeof (int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "zero_copy",
- .data = &ksocknal_tunables.ksnd_zc_min_payload,
- .maxlen = sizeof (int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "zero_copy_recv",
- .data = &ksocknal_tunables.ksnd_zc_recv,
- .maxlen = sizeof (int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "zero_copy_recv",
- .data = &ksocknal_tunables.ksnd_zc_recv_min_nfrags,
- .maxlen = sizeof (int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "typed",
- .data = &ksocknal_tunables.ksnd_typed_conns,
- .maxlen = sizeof (int),
- .mode = 0444,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "min_bulk",
- .data = &ksocknal_tunables.ksnd_min_bulk,
- .maxlen = sizeof (int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "rx_buffer_size",
- .data = &ksocknal_tunables.ksnd_rx_buffer_size,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "tx_buffer_size",
- .data = &ksocknal_tunables.ksnd_tx_buffer_size,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "nagle",
- .data = &ksocknal_tunables.ksnd_nagle,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
-#ifdef CPU_AFFINITY
- {
- INIT_CTL_NAME
- .procname = "irq_affinity",
- .data = &ksocknal_tunables.ksnd_irq_affinity,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
-#endif
- {
- INIT_CTL_NAME
- .procname = "round_robin",
- .data = &ksocknal_tunables.ksnd_round_robin,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "keepalive",
- .data = &ksocknal_tunables.ksnd_keepalive,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "keepalive_idle",
- .data = &ksocknal_tunables.ksnd_keepalive_idle,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "keepalive_count",
- .data = &ksocknal_tunables.ksnd_keepalive_count,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "keepalive_intvl",
- .data = &ksocknal_tunables.ksnd_keepalive_intvl,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
-#ifdef SOCKNAL_BACKOFF
- {
- INIT_CTL_NAME
- .procname = "backoff_init",
- .data = &ksocknal_tunables.ksnd_backoff_init,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
- {
- INIT_CTL_NAME
- .procname = "backoff_max",
- .data = &ksocknal_tunables.ksnd_backoff_max,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
-#endif
-#if SOCKNAL_VERSION_DEBUG
- {
- INIT_CTL_NAME
- .procname = "protocol",
- .data = &ksocknal_tunables.ksnd_protocol,
- .maxlen = sizeof(int),
- .mode = 0644,
- .proc_handler = &proc_dointvec,
- INIT_STRATEGY
- },
-#endif
- { 0 }
-};
-
-struct ctl_table ksocknal_top_ctl_table[] = {
- {
- INIT_CTL_NAME
- .procname = "socknal",
- .data = NULL,
- .maxlen = 0,
- .mode = 0555,
- .child = ksocknal_ctl_table
- },
- { 0 }
-};
-
int
-ksocknal_lib_tunables_init ()
-{
- if (!*ksocknal_tunables.ksnd_typed_conns) {
- int rc = -EINVAL;
-#if SOCKNAL_VERSION_DEBUG
- if (*ksocknal_tunables.ksnd_protocol < 3)
- rc = 0;
-#endif
- if (rc != 0) {
- CERROR("Protocol V3.x MUST have typed connections\n");
- return rc;
- }
- }
-
- if (*ksocknal_tunables.ksnd_zc_recv_min_nfrags < 2)
- *ksocknal_tunables.ksnd_zc_recv_min_nfrags = 2;
- if (*ksocknal_tunables.ksnd_zc_recv_min_nfrags > LNET_MAX_IOV)
- *ksocknal_tunables.ksnd_zc_recv_min_nfrags = LNET_MAX_IOV;
-
- ksocknal_tunables.ksnd_sysctl =
- register_sysctl_table(ksocknal_top_ctl_table);
-
- if (ksocknal_tunables.ksnd_sysctl == NULL)
- CWARN("Can't setup /proc tunables\n");
-
- return 0;
-}
-
-void
-ksocknal_lib_tunables_fini ()
-{
- if (ksocknal_tunables.ksnd_sysctl != NULL)
- unregister_sysctl_table(ksocknal_tunables.ksnd_sysctl);
-}
-#else
-int
-ksocknal_lib_tunables_init ()
-{
- return 0;
-}
-
-void
-ksocknal_lib_tunables_fini ()
-{
-}
-#endif /* # if CONFIG_SYSCTL && !CFS_SYSFS_MODULE_PARM */
-
-int
-ksocknal_lib_get_conn_addrs (ksock_conn_t *conn)
+ksocknal_lib_get_conn_addrs(struct ksock_conn *conn)
{
int rc = lnet_sock_getaddr(conn->ksnc_sock, true,
- &conn->ksnc_ipaddr,
- &conn->ksnc_port);
+ &conn->ksnc_ipaddr,
+ &conn->ksnc_port);
/* Didn't need the {get,put}connsock dance to deref ksnc_sock... */
LASSERT (!conn->ksnc_closing);
if (rc != 0) {
- CERROR ("Error %d getting sock peer IP\n", rc);
+ CERROR ("Error %d getting sock peer_ni IP\n", rc);
return rc;
}
}
int
-ksocknal_lib_zc_capable(ksock_conn_t *conn)
+ksocknal_lib_zc_capable(struct ksock_conn *conn)
{
int caps = conn->ksnc_sock->sk->sk_route_caps;
/* ZC if the socket supports scatter/gather and doesn't need software
* checksums */
- return ((caps & NETIF_F_SG) != 0 && (caps & NETIF_F_ALL_CSUM) != 0);
+ return ((caps & NETIF_F_SG) != 0 && (caps & NETIF_F_CSUM_MASK) != 0);
}
int
-ksocknal_lib_send_iov(ksock_conn_t *conn, ksock_tx_t *tx)
+ksocknal_lib_send_iov(struct ksock_conn *conn, struct ksock_tx *tx,
+ struct kvec *scratchiov)
{
struct socket *sock = conn->ksnc_sock;
int nob;
struct kvec *scratchiov = &scratch;
unsigned int niov = 1;
#else
- struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
unsigned int niov = tx->tx_niov;
#endif
struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
}
int
-ksocknal_lib_send_kiov(ksock_conn_t *conn, ksock_tx_t *tx)
+ksocknal_lib_send_kiov(struct ksock_conn *conn, struct ksock_tx *tx,
+ struct kvec *scratchiov)
{
- struct socket *sock = conn->ksnc_sock;
- lnet_kiov_t *kiov = tx->tx_kiov;
- int rc;
- int nob;
+ struct socket *sock = conn->ksnc_sock;
+ lnet_kiov_t *kiov = tx->tx_kiov;
+ int rc;
+ int nob;
- /* Not NOOP message */
- LASSERT (tx->tx_lnetmsg != NULL);
+ /* Not NOOP message */
+ LASSERT(tx->tx_lnetmsg != NULL);
- /* NB we can't trust socket ops to either consume our iovs
- * or leave them alone. */
- if (tx->tx_msg.ksm_zc_cookies[0] != 0) {
- /* Zero copy is enabled */
- struct sock *sk = sock->sk;
- struct page *page = kiov->kiov_page;
- int offset = kiov->kiov_offset;
- int fragsize = kiov->kiov_len;
- int msgflg = MSG_DONTWAIT;
+ /* NB we can't trust socket ops to either consume our iovs
+ * or leave them alone. */
+ if (tx->tx_msg.ksm_zc_cookies[0] != 0) {
+ /* Zero copy is enabled */
+ struct sock *sk = sock->sk;
+ struct page *page = kiov->kiov_page;
+ int offset = kiov->kiov_offset;
+ int fragsize = kiov->kiov_len;
+ int msgflg = MSG_DONTWAIT;
- CDEBUG(D_NET, "page %p + offset %x for %d\n",
- page, offset, kiov->kiov_len);
+ CDEBUG(D_NET, "page %p + offset %x for %d\n",
+ page, offset, kiov->kiov_len);
if (!list_empty(&conn->ksnc_tx_queue) ||
- fragsize < tx->tx_resid)
- msgflg |= MSG_MORE;
-
- if (sk->sk_prot->sendpage != NULL) {
- rc = sk->sk_prot->sendpage(sk, page,
- offset, fragsize, msgflg);
- } else {
- rc = cfs_tcp_sendpage(sk, page, offset, fragsize,
- msgflg);
- }
- } else {
+ fragsize < tx->tx_resid)
+ msgflg |= MSG_MORE;
+
+ if (sk->sk_prot->sendpage != NULL) {
+ rc = sk->sk_prot->sendpage(sk, page,
+ offset, fragsize, msgflg);
+ } else {
+ rc = tcp_sendpage(sk, page, offset, fragsize, msgflg);
+ }
+ } else {
#if SOCKNAL_SINGLE_FRAG_TX || !SOCKNAL_RISK_KMAP_DEADLOCK
struct kvec scratch;
struct kvec *scratchiov = &scratch;
#ifdef CONFIG_HIGHMEM
#warning "XXX risk of kmap deadlock on multiple frags..."
#endif
- struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
unsigned int niov = tx->tx_nkiov;
#endif
struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
}
void
-ksocknal_lib_eager_ack (ksock_conn_t *conn)
+ksocknal_lib_eager_ack(struct ksock_conn *conn)
{
- int opt = 1;
- struct socket *sock = conn->ksnc_sock;
+ int opt = 1;
+ 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. */
+ * peer_ni. */
kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
(char *)&opt, sizeof(opt));
}
int
-ksocknal_lib_recv_iov (ksock_conn_t *conn)
+ksocknal_lib_recv_iov(struct ksock_conn *conn, struct kvec *scratchiov)
{
#if SOCKNAL_SINGLE_FRAG_RX
struct kvec scratch;
struct kvec *scratchiov = &scratch;
unsigned int niov = 1;
#else
- struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
unsigned int niov = conn->ksnc_rx_niov;
#endif
struct kvec *iov = conn->ksnc_rx_iov;
for (nob = i = 0; i < niov; i++) {
if ((kiov[i].kiov_offset != 0 && i > 0) ||
(kiov[i].kiov_offset + kiov[i].kiov_len !=
- PAGE_CACHE_SIZE && i < niov - 1))
+ PAGE_SIZE && i < niov - 1))
return NULL;
pages[i] = kiov[i].kiov_page;
}
int
-ksocknal_lib_recv_kiov (ksock_conn_t *conn)
+ksocknal_lib_recv_kiov(struct ksock_conn *conn, struct page **pages,
+ struct kvec *scratchiov)
{
#if SOCKNAL_SINGLE_FRAG_RX || !SOCKNAL_RISK_KMAP_DEADLOCK
struct kvec scratch;
#ifdef CONFIG_HIGHMEM
#warning "XXX risk of kmap deadlock on multiple frags..."
#endif
- struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
- struct page **pages = conn->ksnc_scheduler->kss_rx_scratch_pgs;
unsigned int niov = conn->ksnc_rx_nkiov;
#endif
lnet_kiov_t *kiov = conn->ksnc_rx_kiov;
}
void
-ksocknal_lib_csum_tx(ksock_tx_t *tx)
+ksocknal_lib_csum_tx(struct ksock_tx *tx)
{
int i;
__u32 csum;
}
int
-ksocknal_lib_get_conn_tunables (ksock_conn_t *conn, int *txmem, int *rxmem, int *nagle)
+ksocknal_lib_get_conn_tunables(struct ksock_conn *conn, int *txmem, int *rxmem, int *nagle)
{
struct socket *sock = conn->ksnc_sock;
int len;
}
void
-ksocknal_lib_push_conn (ksock_conn_t *conn)
+ksocknal_lib_push_conn(struct ksock_conn *conn)
{
- struct sock *sk;
- struct tcp_sock *tp;
- int nonagle;
- int val = 1;
- int rc;
+ struct sock *sk;
+ struct tcp_sock *tp;
+ int nonagle;
+ int val = 1;
+ int rc;
rc = ksocknal_connsock_addref(conn);
if (rc != 0) /* being shut down */
ksocknal_connsock_decref(conn);
}
-extern void ksocknal_read_callback (ksock_conn_t *conn);
-extern void ksocknal_write_callback (ksock_conn_t *conn);
+void ksocknal_read_callback(struct ksock_conn *conn);
+void ksocknal_write_callback(struct ksock_conn *conn);
/*
* socket call back in Linux
*/
ksocknal_data_ready(struct sock *sk, int n)
#endif
{
- ksock_conn_t *conn;
+ struct ksock_conn *conn;
ENTRY;
/* interleave correctly with closing sockets... */
static void
ksocknal_write_space (struct sock *sk)
{
- ksock_conn_t *conn;
+ struct ksock_conn *conn;
int wspace;
int min_wpace;
}
void
-ksocknal_lib_save_callback(struct socket *sock, ksock_conn_t *conn)
+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, ksock_conn_t *conn)
+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;
}
void
-ksocknal_lib_reset_callback(struct socket *sock, ksock_conn_t *conn)
+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,
}
int
-ksocknal_lib_memory_pressure(ksock_conn_t *conn)
+ksocknal_lib_memory_pressure(struct ksock_conn *conn)
{
int rc = 0;
- ksock_sched_t *sched;
+ struct ksock_sched *sched;
sched = conn->ksnc_scheduler;
spin_lock_bh(&sched->kss_lock);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff