*
* 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, 2016, 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)
{
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;
}
/* 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
{
#if SOCKNAL_SINGLE_FRAG_TX
- struct iovec scratch;
- struct iovec *scratchiov = &scratch;
- unsigned int niov = 1;
+ struct kvec scratch;
+ struct kvec *scratchiov = &scratch;
+ unsigned int niov = 1;
#else
- struct iovec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
- unsigned int niov = tx->tx_niov;
+ struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
+ unsigned int niov = tx->tx_niov;
#endif
struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
int i;
nob < tx->tx_resid)
msg.msg_flags |= MSG_MORE;
- rc = kernel_sendmsg(sock, &msg, (struct kvec *)scratchiov,
- niov, nob);
+ rc = kernel_sendmsg(sock, &msg, scratchiov, niov, nob);
}
return rc;
}
}
} else {
#if SOCKNAL_SINGLE_FRAG_TX || !SOCKNAL_RISK_KMAP_DEADLOCK
- struct iovec scratch;
- struct iovec *scratchiov = &scratch;
+ struct kvec scratch;
+ struct kvec *scratchiov = &scratch;
unsigned int niov = 1;
#else
#ifdef CONFIG_HIGHMEM
#warning "XXX risk of kmap deadlock on multiple frags..."
#endif
- struct iovec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
+ struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
unsigned int niov = tx->tx_nkiov;
#endif
struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
nob < tx->tx_resid)
msg.msg_flags |= MSG_MORE;
- rc = kernel_sendmsg(sock, &msg, (struct kvec *)scratchiov, niov, nob);
+ rc = kernel_sendmsg(sock, &msg, scratchiov, niov, nob);
for (i = 0; i < niov; i++)
kunmap(kiov[i].kiov_page);
ksocknal_lib_eager_ack (ksock_conn_t *conn)
{
int opt = 1;
- mm_segment_t oldmm = get_fs();
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. */
- set_fs(KERNEL_DS);
- sock->ops->setsockopt (sock, SOL_TCP, TCP_QUICKACK,
- (char *)&opt, sizeof (opt));
- set_fs(oldmm);
+ kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
+ (char *)&opt, sizeof(opt));
}
int
ksocknal_lib_recv_iov (ksock_conn_t *conn)
{
#if SOCKNAL_SINGLE_FRAG_RX
- struct iovec scratch;
- struct iovec *scratchiov = &scratch;
+ struct kvec scratch;
+ struct kvec *scratchiov = &scratch;
unsigned int niov = 1;
#else
- struct iovec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
+ struct kvec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
unsigned int niov = conn->ksnc_rx_niov;
#endif
- struct iovec *iov = conn->ksnc_rx_iov;
+ struct kvec *iov = conn->ksnc_rx_iov;
struct msghdr msg = {
.msg_flags = 0
};
}
LASSERT (nob <= conn->ksnc_rx_nob_wanted);
- rc = kernel_recvmsg(conn->ksnc_sock, &msg,
- (struct kvec *)scratchiov, niov, nob, MSG_DONTWAIT);
+ rc = kernel_recvmsg(conn->ksnc_sock, &msg, scratchiov, niov, nob,
+ MSG_DONTWAIT);
saved_csum = 0;
if (conn->ksnc_proto == &ksocknal_protocol_v2x) {
static void *
ksocknal_lib_kiov_vmap(lnet_kiov_t *kiov, int niov,
- struct iovec *iov, struct page **pages)
+ struct kvec *iov, struct page **pages)
{
void *addr;
int nob;
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;
ksocknal_lib_recv_kiov (ksock_conn_t *conn)
{
#if SOCKNAL_SINGLE_FRAG_RX || !SOCKNAL_RISK_KMAP_DEADLOCK
- struct iovec scratch;
- struct iovec *scratchiov = &scratch;
+ struct kvec scratch;
+ struct kvec *scratchiov = &scratch;
struct page **pages = NULL;
unsigned int niov = 1;
#else
#ifdef CONFIG_HIGHMEM
#warning "XXX risk of kmap deadlock on multiple frags..."
#endif
- struct iovec *scratchiov = conn->ksnc_scheduler->kss_scratch_iov;
+ 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
LASSERT (nob <= conn->ksnc_rx_nob_wanted);
- rc = kernel_recvmsg(conn->ksnc_sock, &msg,
- (struct kvec *)scratchiov, n, nob, MSG_DONTWAIT);
+ rc = kernel_recvmsg(conn->ksnc_sock, &msg, scratchiov, n, nob,
+ MSG_DONTWAIT);
if (conn->ksnc_msg.ksm_csum != 0) {
for (i = 0, sum = rc; sum > 0; i++, sum -= fragnob) {
int
ksocknal_lib_get_conn_tunables (ksock_conn_t *conn, int *txmem, int *rxmem, int *nagle)
{
- mm_segment_t oldmm = get_fs ();
struct socket *sock = conn->ksnc_sock;
int len;
int rc;
rc = lnet_sock_getbuf(sock, txmem, rxmem);
if (rc == 0) {
len = sizeof(*nagle);
- set_fs(KERNEL_DS);
- rc = sock->ops->getsockopt(sock, SOL_TCP, TCP_NODELAY,
- (char *)nagle, &len);
- set_fs(oldmm);
+ rc = kernel_getsockopt(sock, SOL_TCP, TCP_NODELAY,
+ (char *)nagle, &len);
}
ksocknal_connsock_decref(conn);
int
ksocknal_lib_setup_sock (struct socket *sock)
{
- mm_segment_t oldmm = get_fs ();
int rc;
int option;
int keep_idle;
linger.l_onoff = 0;
linger.l_linger = 0;
- set_fs (KERNEL_DS);
- rc = sock_setsockopt (sock, SOL_SOCKET, SO_LINGER,
- (char *)&linger, sizeof (linger));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
+ (char *)&linger, sizeof(linger));
if (rc != 0) {
CERROR ("Can't set SO_LINGER: %d\n", rc);
return (rc);
}
option = -1;
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_LINGER2,
- (char *)&option, sizeof (option));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_TCP, TCP_LINGER2,
+ (char *)&option, sizeof(option));
if (rc != 0) {
CERROR ("Can't set SO_LINGER2: %d\n", rc);
return (rc);
if (!*ksocknal_tunables.ksnd_nagle) {
option = 1;
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_NODELAY,
- (char *)&option, sizeof (option));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
+ (char *)&option, sizeof(option));
if (rc != 0) {
CERROR ("Can't disable nagle: %d\n", rc);
return (rc);
option *= 1000;
#endif
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_BACKOFF_INIT,
- (char *)&option, sizeof (option));
- set_fs (oldmm);
+ 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);
option *= 1000;
#endif
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_BACKOFF_MAX,
- (char *)&option, sizeof (option));
- set_fs (oldmm);
+ 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);
do_keepalive = (keep_idle > 0 && keep_count > 0 && keep_intvl > 0);
option = (do_keepalive ? 1 : 0);
- set_fs (KERNEL_DS);
- rc = sock_setsockopt (sock, SOL_SOCKET, SO_KEEPALIVE,
- (char *)&option, sizeof (option));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
+ (char *)&option, sizeof(option));
if (rc != 0) {
CERROR ("Can't set SO_KEEPALIVE: %d\n", rc);
return (rc);
if (!do_keepalive)
return (0);
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_KEEPIDLE,
- (char *)&keep_idle, sizeof (keep_idle));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
+ (char *)&keep_idle, sizeof(keep_idle));
if (rc != 0) {
CERROR ("Can't set TCP_KEEPIDLE: %d\n", rc);
return (rc);
}
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_KEEPINTVL,
- (char *)&keep_intvl, sizeof (keep_intvl));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
+ (char *)&keep_intvl, sizeof(keep_intvl));
if (rc != 0) {
CERROR ("Can't set TCP_KEEPINTVL: %d\n", rc);
return (rc);
}
- set_fs (KERNEL_DS);
- rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_KEEPCNT,
- (char *)&keep_count, sizeof (keep_count));
- set_fs (oldmm);
+ rc = kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
+ (char *)&keep_count, sizeof(keep_count));
if (rc != 0) {
CERROR ("Can't set TCP_KEEPCNT: %d\n", rc);
return (rc);
int nonagle;
int val = 1;
int rc;
- mm_segment_t oldmm;
rc = ksocknal_connsock_addref(conn);
if (rc != 0) /* being shut down */
tp->nonagle = 1;
release_sock (sk);
- oldmm = get_fs ();
- set_fs (KERNEL_DS);
-
- rc = sk->sk_prot->setsockopt (sk, SOL_TCP, TCP_NODELAY,
- (char *)&val, sizeof (val));
+ rc = kernel_setsockopt(conn->ksnc_sock, SOL_TCP, TCP_NODELAY,
+ (char *)&val, sizeof(val));
LASSERT (rc == 0);
- set_fs (oldmm);
-
lock_sock (sk);
tp->nonagle = nonagle;
release_sock (sk);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff