* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
+#include <libcfs/linux/linux-mem.h>
#include "socklnd.h"
struct ksock_tx *
if (tx == NULL)
return NULL;
- atomic_set(&tx->tx_refcount, 1);
+ refcount_set(&tx->tx_refcount, 1);
tx->tx_zc_aborted = 0;
tx->tx_zc_capable = 0;
tx->tx_zc_checked = 0;
{
struct ksock_tx *tx;
- tx = ksocknal_alloc_tx(KSOCK_MSG_NOOP, KSOCK_NOOP_TX_SIZE);
- if (tx == NULL) {
- CERROR("Can't allocate noop tx desc\n");
- return NULL;
- }
+ tx = ksocknal_alloc_tx(KSOCK_MSG_NOOP, KSOCK_NOOP_TX_SIZE);
+ if (tx == NULL) {
+ CERROR("Can't allocate noop tx desc\n");
+ return NULL;
+ }
- tx->tx_conn = NULL;
- tx->tx_lnetmsg = NULL;
- tx->tx_kiov = NULL;
- tx->tx_nkiov = 0;
- tx->tx_iov = tx->tx_frags.virt.iov;
- tx->tx_niov = 1;
- tx->tx_nonblk = nonblk;
+ tx->tx_conn = NULL;
+ tx->tx_lnetmsg = NULL;
+ tx->tx_kiov = NULL;
+ tx->tx_nkiov = 0;
+ tx->tx_niov = 1;
+ tx->tx_nonblk = nonblk;
tx->tx_msg.ksm_csum = 0;
tx->tx_msg.ksm_type = KSOCK_MSG_NOOP;
tx->tx_msg.ksm_zc_cookies[0] = 0;
- tx->tx_msg.ksm_zc_cookies[1] = cookie;
+ tx->tx_msg.ksm_zc_cookies[1] = cookie;
- return tx;
+ return tx;
}
}
static int
-ksocknal_send_iov(struct ksock_conn *conn, struct ksock_tx *tx,
+ksocknal_send_hdr(struct ksock_conn *conn, struct ksock_tx *tx,
struct kvec *scratch_iov)
{
- struct kvec *iov = tx->tx_iov;
+ struct kvec *iov = &tx->tx_hdr;
int nob;
int rc;
LASSERT(tx->tx_niov > 0);
- /* Never touch tx->tx_iov inside ksocknal_lib_send_iov() */
- rc = ksocknal_lib_send_iov(conn, tx, scratch_iov);
+ /* Never touch tx->tx_hdr inside ksocknal_lib_send_hdr() */
+ rc = ksocknal_lib_send_hdr(conn, tx, scratch_iov);
if (rc <= 0) /* sent nothing? */
return rc;
tx->tx_resid -= nob;
/* "consume" iov */
- do {
- LASSERT(tx->tx_niov > 0);
+ LASSERT(tx->tx_niov == 1);
- if (nob < (int) iov->iov_len) {
- iov->iov_base += nob;
- iov->iov_len -= nob;
- return rc;
- }
+ if (nob < (int) iov->iov_len) {
+ iov->iov_base += nob;
+ iov->iov_len -= nob;
+ return rc;
+ }
- nob -= iov->iov_len;
- tx->tx_iov = ++iov;
- tx->tx_niov--;
- } while (nob != 0);
+ LASSERT(nob == iov->iov_len);
+ tx->tx_niov--;
return rc;
}
ksocknal_data.ksnd_enomem_tx--;
rc = -EAGAIN;
} else if (tx->tx_niov != 0) {
- rc = ksocknal_send_iov(conn, tx, scratch_iov);
+ rc = ksocknal_send_hdr(conn, tx, scratch_iov);
} else {
rc = ksocknal_send_kiov(conn, tx, scratch_iov);
}
/* allocated send buffer bytes < computed; infer
* something got ACKed */
conn->ksnc_tx_deadline = ktime_get_seconds() +
- lnet_get_lnd_timeout();
+ ksocknal_timeout();
conn->ksnc_peer->ksnp_last_alive = ktime_get_seconds();
conn->ksnc_tx_bufnob = bufnob;
smp_mb();
conn->ksnc_peer->ksnp_last_alive = ktime_get_seconds();
conn->ksnc_rx_deadline = ktime_get_seconds() +
- lnet_get_lnd_timeout();
+ ksocknal_timeout();
smp_mb(); /* order with setting rx_started */
conn->ksnc_rx_started = 1;
conn->ksnc_peer->ksnp_last_alive = ktime_get_seconds();
conn->ksnc_rx_deadline = ktime_get_seconds() +
- lnet_get_lnd_timeout();
+ ksocknal_timeout();
smp_mb(); /* order with setting rx_started */
conn->ksnc_rx_started = 1;
tx->tx_hstatus = LNET_MSG_STATUS_LOCAL_ERROR;
}
- LASSERT(atomic_read(&tx->tx_refcount) == 1);
+ LASSERT(refcount_read(&tx->tx_refcount) == 1);
ksocknal_tx_done(ni, tx, error);
}
}
/* ZC_REQ is going to be pinned to the peer_ni */
tx->tx_deadline = ktime_get_seconds() +
- lnet_get_lnd_timeout();
+ ksocknal_timeout();
LASSERT (tx->tx_msg.ksm_zc_cookies[0] == 0);
counter++; /* exponential backoff warnings */
if ((counter & (-counter)) == counter)
- CWARN("%u ENOMEM tx %p (%u allocated)\n",
- counter, conn, atomic_read(&libcfs_kmemory));
+ CWARN("%u ENOMEM tx %p (%lld allocated)\n",
+ counter, conn, libcfs_kmem_read());
/* Queue on ksnd_enomem_conns for retry after a timeout */
spin_lock_bh(&ksocknal_data.ksnd_reaper_lock);
ksocknal_tx_prep(conn, tx);
- /* Ensure the frags we've been given EXACTLY match the number of
- * bytes we want to send. Many TCP/IP stacks disregard any total
+ /* Ensure the frags we've been given EXACTLY match the number of
+ * bytes we want to send. Many TCP/IP stacks disregard any total
* size parameters passed to them and just look at the frags.
- *
- * We always expect at least 1 mapped fragment containing the
- * complete ksocknal message header. */
- LASSERT (lnet_iov_nob (tx->tx_niov, tx->tx_iov) +
- lnet_kiov_nob(tx->tx_nkiov, tx->tx_kiov) ==
- (unsigned int)tx->tx_nob);
- LASSERT (tx->tx_niov >= 1);
- LASSERT (tx->tx_resid == tx->tx_nob);
+ *
+ * We always expect at least 1 mapped fragment containing the
+ * complete ksocknal message header.
+ */
+ LASSERT(lnet_iov_nob(tx->tx_niov, &tx->tx_hdr) +
+ lnet_kiov_nob(tx->tx_nkiov, tx->tx_kiov) ==
+ (unsigned int)tx->tx_nob);
+ LASSERT(tx->tx_niov >= 1);
+ LASSERT(tx->tx_resid == tx->tx_nob);
CDEBUG (D_NET, "Packet %p type %d, nob %d niov %d nkiov %d\n",
tx, (tx->tx_lnetmsg != NULL) ? tx->tx_lnetmsg->msg_hdr.type:
if (list_empty(&conn->ksnc_tx_queue) && bufnob == 0) {
/* First packet starts the timeout */
conn->ksnc_tx_deadline = ktime_get_seconds() +
- lnet_get_lnd_timeout();
+ ksocknal_timeout();
if (conn->ksnc_tx_bufnob > 0) /* something got ACKed */
conn->ksnc_peer->ksnp_last_alive = ktime_get_seconds();
conn->ksnc_tx_bufnob = 0;
ksocknal_find_connecting_route_locked (peer_ni) != NULL) {
/* the message is going to be pinned to the peer_ni */
tx->tx_deadline = ktime_get_seconds() +
- lnet_get_lnd_timeout();
+ ksocknal_timeout();
/* Queue the message until a connection is established */
list_add_tail(&tx->tx_list, &peer_ni->ksnp_tx_queue);
int
ksocknal_send(struct lnet_ni *ni, void *private, struct lnet_msg *lntmsg)
{
- int mpflag = 1;
+ /* '1' for consistency with code that checks !mpflag to restore */
+ unsigned int mpflag = 1;
int type = lntmsg->msg_type;
struct lnet_process_id target = lntmsg->msg_target;
- unsigned int payload_niov = lntmsg->msg_niov;
- struct kvec *payload_iov = lntmsg->msg_iov;
- struct bio_vec *payload_kiov = lntmsg->msg_kiov;
- unsigned int payload_offset = lntmsg->msg_offset;
- unsigned int payload_nob = lntmsg->msg_len;
- struct ksock_tx *tx;
- int desc_size;
- int rc;
+ unsigned int payload_niov = lntmsg->msg_niov;
+ struct bio_vec *payload_kiov = lntmsg->msg_kiov;
+ unsigned int payload_offset = lntmsg->msg_offset;
+ unsigned int payload_nob = lntmsg->msg_len;
+ struct ksock_tx *tx;
+ int desc_size;
+ int rc;
/* NB 'private' is different depending on what we're sending.
* Just ignore it... */
LASSERT (payload_nob == 0 || payload_niov > 0);
LASSERT (payload_niov <= LNET_MAX_IOV);
- /* payload is either all vaddrs or all pages */
- LASSERT (!(payload_kiov != NULL && payload_iov != NULL));
LASSERT (!in_interrupt ());
- if (payload_iov != NULL)
- desc_size = offsetof(struct ksock_tx,
- tx_frags.virt.iov[1 + payload_niov]);
- else
- desc_size = offsetof(struct ksock_tx,
- tx_frags.paged.kiov[payload_niov]);
+ desc_size = offsetof(struct ksock_tx,
+ tx_payload[payload_niov]);
if (lntmsg->msg_vmflush)
- mpflag = cfs_memory_pressure_get_and_set();
- tx = ksocknal_alloc_tx(KSOCK_MSG_LNET, desc_size);
- if (tx == NULL) {
- CERROR("Can't allocate tx desc type %d size %d\n",
- type, desc_size);
- if (lntmsg->msg_vmflush)
- cfs_memory_pressure_restore(mpflag);
- return (-ENOMEM);
- }
+ mpflag = memalloc_noreclaim_save();
- tx->tx_conn = NULL; /* set when assigned a conn */
- tx->tx_lnetmsg = lntmsg;
-
- if (payload_iov != NULL) {
- tx->tx_kiov = NULL;
- tx->tx_nkiov = 0;
- tx->tx_iov = tx->tx_frags.virt.iov;
- tx->tx_niov = 1 +
- lnet_extract_iov(payload_niov, &tx->tx_iov[1],
- payload_niov, payload_iov,
- payload_offset, payload_nob);
- } else {
- tx->tx_niov = 1;
- tx->tx_iov = &tx->tx_frags.paged.iov;
- tx->tx_kiov = tx->tx_frags.paged.kiov;
- tx->tx_nkiov = lnet_extract_kiov(payload_niov, tx->tx_kiov,
- payload_niov, payload_kiov,
- payload_offset, payload_nob);
-
- if (payload_nob >= *ksocknal_tunables.ksnd_zc_min_payload)
- tx->tx_zc_capable = 1;
- }
+ tx = ksocknal_alloc_tx(KSOCK_MSG_LNET, desc_size);
+ if (tx == NULL) {
+ CERROR("Can't allocate tx desc type %d size %d\n",
+ type, desc_size);
+ if (lntmsg->msg_vmflush)
+ memalloc_noreclaim_restore(mpflag);
+ return -ENOMEM;
+ }
+
+ tx->tx_conn = NULL; /* set when assigned a conn */
+ tx->tx_lnetmsg = lntmsg;
+
+ tx->tx_niov = 1;
+ tx->tx_kiov = tx->tx_payload;
+ tx->tx_nkiov = lnet_extract_kiov(payload_niov, tx->tx_kiov,
+ payload_niov, payload_kiov,
+ payload_offset, payload_nob);
+
+ if (payload_nob >= *ksocknal_tunables.ksnd_zc_min_payload)
+ tx->tx_zc_capable = 1;
tx->tx_msg.ksm_csum = 0;
tx->tx_msg.ksm_type = KSOCK_MSG_LNET;
tx->tx_msg.ksm_zc_cookies[0] = 0;
tx->tx_msg.ksm_zc_cookies[1] = 0;
- /* The first fragment will be set later in pro_pack */
- rc = ksocknal_launch_packet(ni, tx, target);
- if (!mpflag)
- cfs_memory_pressure_restore(mpflag);
+ /* The first fragment will be set later in pro_pack */
+ rc = ksocknal_launch_packet(ni, tx, target);
+ /*
+ * We can't test lntsmg->msg_vmflush again as lntmsg may
+ * have been freed.
+ */
+ if (!mpflag)
+ memalloc_noreclaim_restore(mpflag);
if (rc == 0)
return (0);
int
ksocknal_thread_start(int (*fn)(void *arg), void *arg, char *name)
{
- struct task_struct *task = kthread_run(fn, arg, name);
+ struct task_struct *task = kthread_run(fn, arg, "%s", name);
if (IS_ERR(task))
return PTR_ERR(task);
- write_lock_bh(&ksocknal_data.ksnd_global_lock);
- ksocknal_data.ksnd_nthreads++;
- write_unlock_bh(&ksocknal_data.ksnd_global_lock);
+ atomic_inc(&ksocknal_data.ksnd_nthreads);
return 0;
}
void
ksocknal_thread_fini (void)
{
- write_lock_bh(&ksocknal_data.ksnd_global_lock);
- if (--ksocknal_data.ksnd_nthreads == 0)
+ if (atomic_dec_and_test(&ksocknal_data.ksnd_nthreads))
wake_up_var(&ksocknal_data.ksnd_nthreads);
- write_unlock_bh(&ksocknal_data.ksnd_global_lock);
}
int
struct lnet_process_id *id;
int rc;
- LASSERT (atomic_read(&conn->ksnc_conn_refcount) > 0);
+ LASSERT(refcount_read(&conn->ksnc_conn_refcount) > 0);
/* NB: sched lock NOT held */
/* SOCKNAL_RX_LNET_HEADER is here for backward compatibility */
int
ksocknal_recv(struct lnet_ni *ni, void *private, struct lnet_msg *msg,
- int delayed, unsigned int niov, struct kvec *iov,
+ int delayed, unsigned int niov,
struct bio_vec *kiov, unsigned int offset, unsigned int mlen,
unsigned int rlen)
{
conn->ksnc_rx_nob_wanted = mlen;
conn->ksnc_rx_nob_left = rlen;
- if (mlen == 0 || iov != NULL) {
- conn->ksnc_rx_nkiov = 0;
- conn->ksnc_rx_kiov = NULL;
- conn->ksnc_rx_iov = conn->ksnc_rx_iov_space.iov;
- conn->ksnc_rx_niov =
- lnet_extract_iov(LNET_MAX_IOV, conn->ksnc_rx_iov,
- niov, iov, offset, mlen);
- } else {
- conn->ksnc_rx_niov = 0;
- conn->ksnc_rx_iov = NULL;
- conn->ksnc_rx_kiov = conn->ksnc_rx_iov_space.kiov;
- conn->ksnc_rx_nkiov =
- lnet_extract_kiov(LNET_MAX_IOV, conn->ksnc_rx_kiov,
- niov, kiov, offset, mlen);
- }
+ if (mlen == 0) {
+ conn->ksnc_rx_nkiov = 0;
+ conn->ksnc_rx_kiov = NULL;
+ conn->ksnc_rx_iov = conn->ksnc_rx_iov_space.iov;
+ conn->ksnc_rx_niov = 0;
+ } else {
+ conn->ksnc_rx_niov = 0;
+ conn->ksnc_rx_iov = NULL;
+ conn->ksnc_rx_kiov = conn->ksnc_rx_iov_space.kiov;
+ conn->ksnc_rx_nkiov =
+ lnet_extract_kiov(LNET_MAX_IOV, conn->ksnc_rx_kiov,
+ niov, kiov, offset, mlen);
+ }
LASSERT (mlen ==
lnet_iov_nob (conn->ksnc_rx_niov, conn->ksnc_rx_iov) +
struct ksock_conn *conn;
struct ksock_tx *tx;
int rc;
- int nloops = 0;
long id = (long)arg;
struct page **rx_scratch_pgs;
struct kvec *scratch_iov;
did_something = 1;
}
- if (!did_something || /* nothing to do */
- ++nloops == SOCKNAL_RESCHED) { /* hogging CPU? */
+ if (!did_something || /* nothing to do */
+ need_resched()) { /* hogging CPU? */
spin_unlock_bh(&sched->kss_lock);
- nloops = 0;
-
if (!did_something) { /* wait for something to do */
rc = wait_event_interruptible_exclusive(
sched->kss_waitq,
/* socket type set on active connections - not set on passive */
LASSERT(!active == !(conn->ksnc_type != SOCKLND_CONN_NONE));
- timeout = active ? lnet_get_lnd_timeout() :
+ timeout = active ? ksocknal_timeout() :
lnet_acceptor_timeout();
rc = lnet_sock_read(sock, &hello->kshm_magic,
int retry_later = 0;
int rc = 0;
- deadline = ktime_get_seconds() + lnet_get_lnd_timeout();
+ deadline = ktime_get_seconds() + ksocknal_timeout();
write_lock_bh(&ksocknal_data.ksnd_global_lock);
route->ksnr_connecting = 1;
- for (;;) {
- wanted = ksocknal_route_mask() & ~route->ksnr_connected;
+ for (;;) {
+ wanted = ksocknal_route_mask() & ~route->ksnr_connected;
- /* stop connecting if peer_ni/route got closed under me, or
- * route got connected while queued */
- if (peer_ni->ksnp_closing || route->ksnr_deleted ||
- wanted == 0) {
- retry_later = 0;
- break;
- }
+ /* stop connecting if peer_ni/route got closed under me, or
+ * route got connected while queued */
+ if (peer_ni->ksnp_closing || route->ksnr_deleted ||
+ wanted == 0) {
+ retry_later = 0;
+ break;
+ }
- /* reschedule if peer_ni is connecting to me */
- if (peer_ni->ksnp_accepting > 0) {
- CDEBUG(D_NET,
- "peer_ni %s(%d) already connecting to me, retry later.\n",
- libcfs_nid2str(peer_ni->ksnp_id.nid), peer_ni->ksnp_accepting);
- retry_later = 1;
- }
+ /* reschedule if peer_ni is connecting to me */
+ if (peer_ni->ksnp_accepting > 0) {
+ CDEBUG(D_NET,
+ "peer_ni %s(%d) already connecting to me, retry later.\n",
+ libcfs_nid2str(peer_ni->ksnp_id.nid), peer_ni->ksnp_accepting);
+ retry_later = 1;
+ }
- if (retry_later) /* needs reschedule */
- break;
+ if (retry_later) /* needs reschedule */
+ break;
- if ((wanted & (1 << SOCKLND_CONN_ANY)) != 0) {
- type = SOCKLND_CONN_ANY;
- } else if ((wanted & (1 << SOCKLND_CONN_CONTROL)) != 0) {
- type = SOCKLND_CONN_CONTROL;
- } else if ((wanted & (1 << SOCKLND_CONN_BULK_IN)) != 0) {
- type = SOCKLND_CONN_BULK_IN;
- } else {
- LASSERT ((wanted & (1 << SOCKLND_CONN_BULK_OUT)) != 0);
- type = SOCKLND_CONN_BULK_OUT;
- }
+ if ((wanted & BIT(SOCKLND_CONN_ANY)) != 0) {
+ type = SOCKLND_CONN_ANY;
+ } else if ((wanted & BIT(SOCKLND_CONN_CONTROL)) != 0) {
+ type = SOCKLND_CONN_CONTROL;
+ } else if ((wanted & BIT(SOCKLND_CONN_BULK_IN)) != 0) {
+ type = SOCKLND_CONN_BULK_IN;
+ } else {
+ LASSERT ((wanted & BIT(SOCKLND_CONN_BULK_OUT)) != 0);
+ type = SOCKLND_CONN_BULK_OUT;
+ }
write_unlock_bh(&ksocknal_data.ksnd_global_lock);
if (ktime_get_seconds() >= deadline) {
- rc = -ETIMEDOUT;
- lnet_connect_console_error(rc, peer_ni->ksnp_id.nid,
- route->ksnr_ipaddr,
- route->ksnr_port);
- goto failed;
- }
+ rc = -ETIMEDOUT;
+ lnet_connect_console_error(rc, peer_ni->ksnp_id.nid,
+ route->ksnr_ipaddr,
+ route->ksnr_port);
+ goto failed;
+ }
sock = lnet_connect(peer_ni->ksnp_id.nid,
route->ksnr_myiface,
goto failed;
}
- rc = ksocknal_create_conn(peer_ni->ksnp_ni, route, sock, type);
- if (rc < 0) {
- lnet_connect_console_error(rc, peer_ni->ksnp_id.nid,
- route->ksnr_ipaddr,
- route->ksnr_port);
- goto failed;
- }
+ rc = ksocknal_create_conn(peer_ni->ksnp_ni, route, sock, type);
+ if (rc < 0) {
+ lnet_connect_console_error(rc, peer_ni->ksnp_id.nid,
+ route->ksnr_ipaddr,
+ route->ksnr_port);
+ goto failed;
+ }
- /* A +ve RC means I have to retry because I lost the connection
- * race or I have to renegotiate protocol version */
- retry_later = (rc != 0);
- if (retry_later)
- CDEBUG(D_NET, "peer_ni %s: conn race, retry later.\n",
- libcfs_nid2str(peer_ni->ksnp_id.nid));
+ /* A +ve RC means I have to retry because I lost the connection
+ * race or I have to renegotiate protocol version */
+ retry_later = (rc != 0);
+ if (retry_later)
+ CDEBUG(D_NET, "peer_ni %s: conn race, retry later.\n",
+ libcfs_nid2str(peer_ni->ksnp_id.nid));
write_lock_bh(&ksocknal_data.ksnd_global_lock);
- }
+ }
route->ksnr_scheduled = 0;
route->ksnr_connecting = 0;
spinlock_t *connd_lock = &ksocknal_data.ksnd_connd_lock;
struct ksock_connreq *cr;
wait_queue_entry_t wait;
- int nloops = 0;
int cons_retry = 0;
init_waitqueue_entry(&wait, current);
}
if (dropped_lock) {
- if (++nloops < SOCKNAL_RESCHED)
+ if (!need_resched())
continue;
spin_unlock_bh(connd_lock);
- nloops = 0;
cond_resched();
spin_lock_bh(connd_lock);
continue;
add_wait_queue_exclusive(&ksocknal_data.ksnd_connd_waitq, &wait);
spin_unlock_bh(connd_lock);
- nloops = 0;
schedule_timeout(timeout);
remove_wait_queue(&ksocknal_data.ksnd_connd_waitq, &wait);
* times the timeout interval.
*/
- lnd_timeout = lnet_get_lnd_timeout();
+ lnd_timeout = ksocknal_timeout();
if (lnd_timeout > n * p)
chunk = (chunk * n * p) / lnd_timeout;
if (chunk == 0)
git://git.whamcloud.com / fs / lustre-release.git / blobdiff