#include "o2iblnd.h"
+static void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx);
+
static void
kiblnd_tx_done(lnet_ni_t *ni, kib_tx_t *tx)
{
lnet_msg_t *lntmsg[2];
- kib_net_t *net = ni->ni_data;
- int rc;
- int i;
+ kib_net_t *net = ni->ni_data;
+ int rc;
+ int i;
LASSERT(net != NULL);
LASSERT(!in_interrupt());
static kib_tx_t *
kiblnd_get_idle_tx(lnet_ni_t *ni, lnet_nid_t target)
{
- kib_net_t *net = (kib_net_t *)ni->ni_data;
- struct list_head *node;
- kib_tx_t *tx;
- kib_tx_poolset_t *tps;
+ kib_net_t *net = (kib_net_t *)ni->ni_data;
+ struct list_head *node;
+ kib_tx_t *tx;
+ kib_tx_poolset_t *tps;
tps = net->ibn_tx_ps[lnet_cpt_of_nid(target)];
node = kiblnd_pool_alloc_node(&tps->tps_poolset);
LASSERT(tx->tx_conn == NULL);
LASSERT(tx->tx_lntmsg[0] == NULL);
LASSERT(tx->tx_lntmsg[1] == NULL);
- LASSERT(tx->tx_u.pmr == NULL);
LASSERT(tx->tx_nfrags == 0);
return tx;
static void
kiblnd_drop_rx(kib_rx_t *rx)
{
- kib_conn_t *conn = rx->rx_conn;
- struct kib_sched_info *sched = conn->ibc_sched;
- unsigned long flags;
+ kib_conn_t *conn = rx->rx_conn;
+ struct kib_sched_info *sched = conn->ibc_sched;
+ unsigned long flags;
spin_lock_irqsave(&sched->ibs_lock, flags);
LASSERT(conn->ibc_nrx > 0);
int
kiblnd_post_rx(kib_rx_t *rx, int credit)
{
- kib_conn_t *conn = rx->rx_conn;
- kib_net_t *net = conn->ibc_peer->ibp_ni->ni_data;
- struct ib_recv_wr *bad_wrq = NULL;
- struct ib_mr *mr;
- int rc;
+ kib_conn_t *conn = rx->rx_conn;
+ kib_net_t *net = conn->ibc_peer->ibp_ni->ni_data;
+ struct ib_recv_wr *bad_wrq = NULL;
+ struct ib_mr *mr;
+ int rc;
LASSERT(net != NULL);
LASSERT(!in_interrupt());
rx->rx_sge.addr = rx->rx_msgaddr;
rx->rx_sge.length = IBLND_MSG_SIZE;
- rx->rx_wrq.next = NULL;
+ rx->rx_wrq.next = NULL;
rx->rx_wrq.sg_list = &rx->rx_sge;
rx->rx_wrq.num_sge = 1;
- rx->rx_wrq.wr_id = kiblnd_ptr2wreqid(rx, IBLND_WID_RX);
+ rx->rx_wrq.wr_id = kiblnd_ptr2wreqid(rx, IBLND_WID_RX);
LASSERT(conn->ibc_state >= IBLND_CONN_INIT);
LASSERT(rx->rx_nob >= 0); /* not posted */
rx->rx_nob = -1; /* flag posted */
+ /* NB: need an extra reference after ib_post_recv because we don't
+ * own this rx (and rx::rx_conn) anymore, LU-5678.
+ */
+ kiblnd_conn_addref(conn);
rc = ib_post_recv(conn->ibc_cmid->qp, &rx->rx_wrq, &bad_wrq);
- if (rc != 0) {
+ if (unlikely(rc != 0)) {
CERROR("Can't post rx for %s: %d, bad_wrq: %p\n",
libcfs_nid2str(conn->ibc_peer->ibp_nid), rc, bad_wrq);
rx->rx_nob = 0;
}
if (conn->ibc_state < IBLND_CONN_ESTABLISHED) /* Initial post */
- return rc;
+ goto out;
- if (rc != 0) {
+ if (unlikely(rc != 0)) {
kiblnd_close_conn(conn, rc);
kiblnd_drop_rx(rx); /* No more posts for this rx */
- return rc;
+ goto out;
}
if (credit == IBLND_POSTRX_NO_CREDIT)
- return 0;
+ goto out;
spin_lock(&conn->ibc_lock);
if (credit == IBLND_POSTRX_PEER_CREDIT)
spin_unlock(&conn->ibc_lock);
kiblnd_check_sends(conn);
- return 0;
+out:
+ kiblnd_conn_decref(conn);
+ return rc;
}
static kib_tx_t *
kiblnd_find_waiting_tx_locked(kib_conn_t *conn, int txtype, __u64 cookie)
{
- struct list_head *tmp;
+ struct list_head *tmp;
list_for_each(tmp, &conn->ibc_active_txs) {
kib_tx_t *tx = list_entry(tmp, kib_tx_t, tx_list);
static void
kiblnd_handle_completion(kib_conn_t *conn, int txtype, int status, __u64 cookie)
{
- kib_tx_t *tx;
- lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
- int idle;
+ kib_tx_t *tx;
+ lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
+ int idle;
spin_lock(&conn->ibc_lock);
}
if (tx->tx_status == 0) { /* success so far */
- if (status < 0) { /* failed? */
+ if (status < 0) /* failed? */
tx->tx_status = status;
- } else if (txtype == IBLND_MSG_GET_REQ) {
+ else if (txtype == IBLND_MSG_GET_REQ)
lnet_set_reply_msg_len(ni, tx->tx_lntmsg[1], status);
- }
}
tx->tx_waiting = 0;
static void
kiblnd_send_completion(kib_conn_t *conn, int type, int status, __u64 cookie)
{
- lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
- kib_tx_t *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
+ lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
+ kib_tx_t *tx = kiblnd_get_idle_tx(ni, conn->ibc_peer->ibp_nid);
if (tx == NULL) {
CERROR("Can't get tx for completion %x for %s\n",
static void
kiblnd_handle_rx(kib_rx_t *rx)
{
- kib_msg_t *msg = rx->rx_msg;
- kib_conn_t *conn = rx->rx_conn;
- lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
- int credits = msg->ibm_credits;
- kib_tx_t *tx;
- int rc = 0;
- int rc2;
- int post_credit;
+ kib_msg_t *msg = rx->rx_msg;
+ kib_conn_t *conn = rx->rx_conn;
+ lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
+ int credits = msg->ibm_credits;
+ kib_tx_t *tx;
+ int rc = 0;
+ int rc2;
+ int post_credit;
LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED);
static void
kiblnd_rx_complete(kib_rx_t *rx, int status, int nob)
{
- kib_msg_t *msg = rx->rx_msg;
- kib_conn_t *conn = rx->rx_conn;
- lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
- kib_net_t *net = ni->ni_data;
- int rc;
- int err = -EIO;
+ kib_msg_t *msg = rx->rx_msg;
+ kib_conn_t *conn = rx->rx_conn;
+ lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
+ kib_net_t *net = ni->ni_data;
+ int rc;
+ int err = -EIO;
LASSERT(net != NULL);
LASSERT(rx->rx_nob < 0); /* was posted */
/* racing with connection establishment/teardown! */
if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
- rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
- unsigned long flags;
+ rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
+ unsigned long flags;
write_lock_irqsave(g_lock, flags);
/* must check holding global lock to eliminate race */
static int
kiblnd_fmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
{
- kib_hca_dev_t *hdev;
- __u64 *pages = tx->tx_pages;
- kib_fmr_poolset_t *fps;
- int npages;
- int size;
- int cpt;
- int rc;
- int i;
+ kib_hca_dev_t *hdev;
+ __u64 *pages = tx->tx_pages;
+ kib_fmr_poolset_t *fps;
+ int npages;
+ int size;
+ int cpt;
+ int rc;
+ int i;
LASSERT(tx->tx_pool != NULL);
LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
- hdev = tx->tx_pool->tpo_hdev;
+ hdev = tx->tx_pool->tpo_hdev;
for (i = 0, npages = 0; i < rd->rd_nfrags; i++) {
for (size = 0; size < rd->rd_frags[i].rf_nob;
cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
fps = net->ibn_fmr_ps[cpt];
- rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->tx_u.fmr);
+ rc = kiblnd_fmr_pool_map(fps, pages, npages, 0, &tx->fmr);
if (rc != 0) {
CERROR("Can't map %d pages: %d\n", npages, rc);
return rc;
/* If rd is not tx_rd, it's going to get sent to a peer, who will need
* the rkey */
- rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.fmr.fmr_pfmr->fmr->rkey :
- tx->tx_u.fmr.fmr_pfmr->fmr->lkey;
+ rd->rd_key = (rd != tx->tx_rd) ? tx->fmr.fmr_pfmr->fmr->rkey :
+ tx->fmr.fmr_pfmr->fmr->lkey;
rd->rd_frags[0].rf_addr &= ~hdev->ibh_page_mask;
- rd->rd_frags[0].rf_nob = nob;
+ rd->rd_frags[0].rf_nob = nob;
rd->rd_nfrags = 1;
return 0;
}
-static int
-kiblnd_pmr_map_tx(kib_net_t *net, kib_tx_t *tx, kib_rdma_desc_t *rd, int nob)
+static void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx)
{
- kib_hca_dev_t *hdev;
- kib_pmr_poolset_t *pps;
- __u64 iova;
- int cpt;
- int rc;
-
- LASSERT(tx->tx_pool != NULL);
- LASSERT(tx->tx_pool->tpo_pool.po_owner != NULL);
-
- hdev = tx->tx_pool->tpo_hdev;
-
- iova = rd->rd_frags[0].rf_addr & ~hdev->ibh_page_mask;
-
- cpt = tx->tx_pool->tpo_pool.po_owner->ps_cpt;
-
- pps = net->ibn_pmr_ps[cpt];
- rc = kiblnd_pmr_pool_map(pps, hdev, rd, &iova, &tx->tx_u.pmr);
- if (rc != 0) {
- CERROR("Failed to create MR by phybuf: %d\n", rc);
- return rc;
- }
-
- /* If rd is not tx_rd, it's going to get sent to a peer, who will need
- * the rkey */
- rd->rd_key = (rd != tx->tx_rd) ? tx->tx_u.pmr->pmr_mr->rkey :
- tx->tx_u.pmr->pmr_mr->lkey;
- rd->rd_nfrags = 1;
- rd->rd_frags[0].rf_addr = iova;
- rd->rd_frags[0].rf_nob = nob;
-
- return 0;
-}
-
-void
-kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx)
-{
- kib_net_t *net = ni->ni_data;
+ kib_net_t *net = ni->ni_data;
LASSERT(net != NULL);
- if (net->ibn_fmr_ps != NULL && tx->tx_u.fmr.fmr_pfmr != NULL) {
- kiblnd_fmr_pool_unmap(&tx->tx_u.fmr, tx->tx_status);
- tx->tx_u.fmr.fmr_pfmr = NULL;
-
- } else if (net->ibn_pmr_ps != NULL && tx->tx_u.pmr != NULL) {
- kiblnd_pmr_pool_unmap(tx->tx_u.pmr);
- tx->tx_u.pmr = NULL;
+ if (net->ibn_fmr_ps && tx->fmr.fmr_pfmr) {
+ kiblnd_fmr_pool_unmap(&tx->fmr, tx->tx_status);
+ tx->fmr.fmr_pfmr = NULL;
}
if (tx->tx_nfrags != 0) {
}
}
-int
-kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx,
- kib_rdma_desc_t *rd, int nfrags)
+static int kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
+ int nfrags)
{
- kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
- kib_net_t *net = ni->ni_data;
- struct ib_mr *mr = NULL;
- __u32 nob;
- int i;
+ kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
+ kib_net_t *net = ni->ni_data;
+ struct ib_mr *mr = NULL;
+ __u32 nob;
+ int i;
/* If rd is not tx_rd, it's going to get sent to a peer and I'm the
* RDMA sink */
tx->tx_dmadir = (rd != tx->tx_rd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
tx->tx_nfrags = nfrags;
- rd->rd_nfrags =
- kiblnd_dma_map_sg(hdev->ibh_ibdev,
- tx->tx_frags, tx->tx_nfrags, tx->tx_dmadir);
+ rd->rd_nfrags = kiblnd_dma_map_sg(hdev->ibh_ibdev, tx->tx_frags,
+ tx->tx_nfrags, tx->tx_dmadir);
for (i = 0, nob = 0; i < rd->rd_nfrags; i++) {
rd->rd_frags[i].rf_nob = kiblnd_sg_dma_len(
if (net->ibn_fmr_ps != NULL)
return kiblnd_fmr_map_tx(net, tx, rd, nob);
- else if (net->ibn_pmr_ps != NULL)
- return kiblnd_pmr_map_tx(net, tx, rd, nob);
return -EINVAL;
}
-
static int
kiblnd_setup_rd_iov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
unsigned int niov, struct kvec *iov, int offset, int nob)
{
- kib_net_t *net = ni->ni_data;
- struct page *page;
+ kib_net_t *net = ni->ni_data;
+ struct page *page;
struct scatterlist *sg;
- unsigned long vaddr;
- int fragnob;
- int page_offset;
+ unsigned long vaddr;
+ int fragnob;
+ int page_offset;
LASSERT(nob > 0);
LASSERT(niov > 0);
kiblnd_setup_rd_kiov(lnet_ni_t *ni, kib_tx_t *tx, kib_rdma_desc_t *rd,
int nkiov, lnet_kiov_t *kiov, int offset, int nob)
{
- kib_net_t *net = ni->ni_data;
+ kib_net_t *net = ni->ni_data;
struct scatterlist *sg;
- int fragnob;
+ int fragnob;
CDEBUG(D_NET, "niov %d offset %d nob %d\n", nkiov, offset, nob);
__releases(conn->ibc_lock)
__acquires(conn->ibc_lock)
{
- kib_msg_t *msg = tx->tx_msg;
- kib_peer_t *peer = conn->ibc_peer;
- int ver = conn->ibc_version;
- int rc;
- int done;
+ kib_msg_t *msg = tx->tx_msg;
+ kib_peer_t *peer = conn->ibc_peer;
+ int ver = conn->ibc_version;
+ int rc;
+ int done;
struct ib_send_wr *bad_wrq;
LASSERT(tx->tx_queued);
/* close_conn will launch failover */
rc = -ENETDOWN;
} else {
- rc = ib_post_send(conn->ibc_cmid->qp,
- tx->tx_wrq, &bad_wrq);
+ rc = ib_post_send(conn->ibc_cmid->qp, &tx->tx_wrq->wr, &bad_wrq);
}
conn->ibc_last_send = jiffies;
void
kiblnd_check_sends(kib_conn_t *conn)
{
- int ver = conn->ibc_version;
+ int ver = conn->ibc_version;
lnet_ni_t *ni = conn->ibc_peer->ibp_ni;
- kib_tx_t *tx;
+ kib_tx_t *tx;
/* Don't send anything until after the connection is established */
if (conn->ibc_state < IBLND_CONN_ESTABLISHED) {
static void
kiblnd_tx_complete(kib_tx_t *tx, int status)
{
- int failed = (status != IB_WC_SUCCESS);
- kib_conn_t *conn = tx->tx_conn;
- int idle;
+ int failed = (status != IB_WC_SUCCESS);
+ kib_conn_t *conn = tx->tx_conn;
+ int idle;
LASSERT(tx->tx_sending > 0);
void
kiblnd_init_tx_msg(lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob)
{
- kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
- struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq];
- struct ib_send_wr *wrq = &tx->tx_wrq[tx->tx_nwrq];
- int nob = offsetof(kib_msg_t, ibm_u) + body_nob;
- struct ib_mr *mr;
+ kib_hca_dev_t *hdev = tx->tx_pool->tpo_hdev;
+ struct ib_sge *sge = &tx->tx_sge[tx->tx_nwrq];
+ struct ib_rdma_wr *wrq = &tx->tx_wrq[tx->tx_nwrq];
+ int nob = offsetof(kib_msg_t, ibm_u) + body_nob;
+ struct ib_mr *mr;
LASSERT(tx->tx_nwrq >= 0);
LASSERT(tx->tx_nwrq < IBLND_MAX_RDMA_FRAGS + 1);
memset(wrq, 0, sizeof(*wrq));
- wrq->next = NULL;
- wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_TX);
- wrq->sg_list = sge;
- wrq->num_sge = 1;
- wrq->opcode = IB_WR_SEND;
- wrq->send_flags = IB_SEND_SIGNALED;
+ wrq->wr.next = NULL;
+ wrq->wr.wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_TX);
+ wrq->wr.sg_list = sge;
+ wrq->wr.num_sge = 1;
+ wrq->wr.opcode = IB_WR_SEND;
+ wrq->wr.send_flags = IB_SEND_SIGNALED;
tx->tx_nwrq++;
}
kiblnd_init_rdma(kib_conn_t *conn, kib_tx_t *tx, int type,
int resid, kib_rdma_desc_t *dstrd, __u64 dstcookie)
{
- kib_msg_t *ibmsg = tx->tx_msg;
- kib_rdma_desc_t *srcrd = tx->tx_rd;
- struct ib_sge *sge = &tx->tx_sge[0];
- struct ib_send_wr *wrq = &tx->tx_wrq[0];
- int rc = resid;
- int srcidx;
- int dstidx;
- int wrknob;
+ kib_msg_t *ibmsg = tx->tx_msg;
+ kib_rdma_desc_t *srcrd = tx->tx_rd;
+ struct ib_sge *sge = &tx->tx_sge[0];
+ struct ib_rdma_wr *wrq = &tx->tx_wrq[0], *next;
+ int rc = resid;
+ int srcidx;
+ int dstidx;
+ int wrknob;
LASSERT(!in_interrupt());
LASSERT(tx->tx_nwrq == 0);
sge->length = wrknob;
wrq = &tx->tx_wrq[tx->tx_nwrq];
+ next = wrq + 1;
- wrq->next = wrq + 1;
- wrq->wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA);
- wrq->sg_list = sge;
- wrq->num_sge = 1;
- wrq->opcode = IB_WR_RDMA_WRITE;
- wrq->send_flags = 0;
+ wrq->wr.next = &next->wr;
+ wrq->wr.wr_id = kiblnd_ptr2wreqid(tx, IBLND_WID_RDMA);
+ wrq->wr.sg_list = sge;
+ wrq->wr.num_sge = 1;
+ wrq->wr.opcode = IB_WR_RDMA_WRITE;
+ wrq->wr.send_flags = 0;
- wrq->wr.rdma.remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx);
- wrq->wr.rdma.rkey = kiblnd_rd_frag_key(dstrd, dstidx);
+ wrq->remote_addr = kiblnd_rd_frag_addr(dstrd, dstidx);
+ wrq->rkey = kiblnd_rd_frag_key(dstrd, dstidx);
srcidx = kiblnd_rd_consume_frag(srcrd, srcidx, wrknob);
dstidx = kiblnd_rd_consume_frag(dstrd, dstidx, wrknob);
void
kiblnd_queue_tx_locked(kib_tx_t *tx, kib_conn_t *conn)
{
- struct list_head *q;
+ struct list_head *q;
LASSERT(tx->tx_nwrq > 0); /* work items set up */
LASSERT(!tx->tx_queued); /* not queued for sending already */
kiblnd_connect_peer(kib_peer_t *peer)
{
struct rdma_cm_id *cmid;
- kib_dev_t *dev;
- kib_net_t *net = peer->ibp_ni->ni_data;
+ kib_dev_t *dev;
+ kib_net_t *net = peer->ibp_ni->ni_data;
struct sockaddr_in srcaddr;
struct sockaddr_in dstaddr;
- int rc;
+ int rc;
LASSERT(net != NULL);
LASSERT(peer->ibp_connecting > 0);
void
kiblnd_launch_tx(lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid)
{
- kib_peer_t *peer;
- kib_peer_t *peer2;
- kib_conn_t *conn;
- rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
- unsigned long flags;
- int rc;
+ kib_peer_t *peer;
+ kib_peer_t *peer2;
+ kib_conn_t *conn;
+ rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
+ unsigned long flags;
+ int rc;
/* If I get here, I've committed to send, so I complete the tx with
* failure on any problems */
int
kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg)
{
- lnet_hdr_t *hdr = &lntmsg->msg_hdr;
- int type = lntmsg->msg_type;
+ lnet_hdr_t *hdr = &lntmsg->msg_hdr;
+ int type = lntmsg->msg_type;
lnet_process_id_t target = lntmsg->msg_target;
- int target_is_router = lntmsg->msg_target_is_router;
- int routing = lntmsg->msg_routing;
- unsigned int payload_niov = lntmsg->msg_niov;
- struct kvec *payload_iov = lntmsg->msg_iov;
- lnet_kiov_t *payload_kiov = lntmsg->msg_kiov;
- unsigned int payload_offset = lntmsg->msg_offset;
- unsigned int payload_nob = lntmsg->msg_len;
- kib_msg_t *ibmsg;
- kib_tx_t *tx;
- int nob;
- int rc;
+ int target_is_router = lntmsg->msg_target_is_router;
+ int routing = lntmsg->msg_routing;
+ unsigned int payload_niov = lntmsg->msg_niov;
+ struct kvec *payload_iov = lntmsg->msg_iov;
+ lnet_kiov_t *payload_kiov = lntmsg->msg_kiov;
+ unsigned int payload_offset = lntmsg->msg_offset;
+ unsigned int payload_nob = lntmsg->msg_len;
+ kib_msg_t *ibmsg;
+ kib_rdma_desc_t *rd;
+ kib_tx_t *tx;
+ int nob;
+ int rc;
/* NB 'private' is different depending on what we're sending.... */
}
ibmsg = tx->tx_msg;
-
+ rd = &ibmsg->ibm_u.get.ibgm_rd;
if ((lntmsg->msg_md->md_options & LNET_MD_KIOV) == 0)
- rc = kiblnd_setup_rd_iov(ni, tx,
- &ibmsg->ibm_u.get.ibgm_rd,
+ rc = kiblnd_setup_rd_iov(ni, tx, rd,
lntmsg->msg_md->md_niov,
lntmsg->msg_md->md_iov.iov,
0, lntmsg->msg_md->md_length);
else
- rc = kiblnd_setup_rd_kiov(ni, tx,
- &ibmsg->ibm_u.get.ibgm_rd,
+ rc = kiblnd_setup_rd_kiov(ni, tx, rd,
lntmsg->msg_md->md_niov,
lntmsg->msg_md->md_iov.kiov,
0, lntmsg->msg_md->md_length);
return -EIO;
}
- nob = offsetof(kib_get_msg_t, ibgm_rd.rd_frags[tx->tx_nfrags]);
+ nob = offsetof(kib_get_msg_t, ibgm_rd.rd_frags[rd->rd_nfrags]);
ibmsg->ibm_u.get.ibgm_cookie = tx->tx_cookie;
ibmsg->ibm_u.get.ibgm_hdr = *hdr;
kiblnd_reply(lnet_ni_t *ni, kib_rx_t *rx, lnet_msg_t *lntmsg)
{
lnet_process_id_t target = lntmsg->msg_target;
- unsigned int niov = lntmsg->msg_niov;
- struct kvec *iov = lntmsg->msg_iov;
- lnet_kiov_t *kiov = lntmsg->msg_kiov;
- unsigned int offset = lntmsg->msg_offset;
- unsigned int nob = lntmsg->msg_len;
- kib_tx_t *tx;
- int rc;
+ unsigned int niov = lntmsg->msg_niov;
+ struct kvec *iov = lntmsg->msg_iov;
+ lnet_kiov_t *kiov = lntmsg->msg_kiov;
+ unsigned int offset = lntmsg->msg_offset;
+ unsigned int nob = lntmsg->msg_len;
+ kib_tx_t *tx;
+ int rc;
tx = kiblnd_get_idle_tx(ni, rx->rx_conn->ibc_peer->ibp_nid);
if (tx == NULL) {
unsigned int niov, struct kvec *iov, lnet_kiov_t *kiov,
unsigned int offset, unsigned int mlen, unsigned int rlen)
{
- kib_rx_t *rx = private;
- kib_msg_t *rxmsg = rx->rx_msg;
- kib_conn_t *conn = rx->rx_conn;
- kib_tx_t *tx;
- kib_msg_t *txmsg;
- int nob;
- int post_credit = IBLND_POSTRX_PEER_CREDIT;
- int rc = 0;
+ kib_rx_t *rx = private;
+ kib_msg_t *rxmsg = rx->rx_msg;
+ kib_conn_t *conn = rx->rx_conn;
+ kib_tx_t *tx;
+ int nob;
+ int post_credit = IBLND_POSTRX_PEER_CREDIT;
+ int rc = 0;
LASSERT(mlen <= rlen);
LASSERT(!in_interrupt());
lnet_finalize(ni, lntmsg, 0);
break;
- case IBLND_MSG_PUT_REQ:
+ case IBLND_MSG_PUT_REQ: {
+ kib_msg_t *txmsg;
+ kib_rdma_desc_t *rd;
+
if (mlen == 0) {
lnet_finalize(ni, lntmsg, 0);
kiblnd_send_completion(rx->rx_conn, IBLND_MSG_PUT_NAK, 0,
}
txmsg = tx->tx_msg;
+ rd = &txmsg->ibm_u.putack.ibpam_rd;
if (kiov == NULL)
- rc = kiblnd_setup_rd_iov(ni, tx,
- &txmsg->ibm_u.putack.ibpam_rd,
+ rc = kiblnd_setup_rd_iov(ni, tx, rd,
niov, iov, offset, mlen);
else
- rc = kiblnd_setup_rd_kiov(ni, tx,
- &txmsg->ibm_u.putack.ibpam_rd,
+ rc = kiblnd_setup_rd_kiov(ni, tx, rd,
niov, kiov, offset, mlen);
if (rc != 0) {
CERROR("Can't setup PUT sink for %s: %d\n",
break;
}
- nob = offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[tx->tx_nfrags]);
+ nob = offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[rd->rd_nfrags]);
txmsg->ibm_u.putack.ibpam_src_cookie = rxmsg->ibm_u.putreq.ibprm_cookie;
txmsg->ibm_u.putack.ibpam_dst_cookie = tx->tx_cookie;
/* reposted buffer reserved for PUT_DONE */
post_credit = IBLND_POSTRX_NO_CREDIT;
break;
+ }
case IBLND_MSG_GET_REQ:
if (lntmsg != NULL) {
static void
kiblnd_peer_notify(kib_peer_t *peer)
{
- int error = 0;
- unsigned long last_alive = 0;
+ int error = 0;
+ unsigned long last_alive = 0;
unsigned long flags;
read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
* connection to be finished off by the connd. Otherwise the connd is
* already dealing with it (either to set it up or tear it down).
* Caller holds kib_global_lock exclusively in irq context */
- kib_peer_t *peer = conn->ibc_peer;
- kib_dev_t *dev;
- unsigned long flags;
+ kib_peer_t *peer = conn->ibc_peer;
+ kib_dev_t *dev;
+ unsigned long flags;
LASSERT(error != 0 || conn->ibc_state >= IBLND_CONN_ESTABLISHED);
static void
kiblnd_handle_early_rxs(kib_conn_t *conn)
{
- unsigned long flags;
- kib_rx_t *rx;
+ unsigned long flags;
+ kib_rx_t *rx;
kib_rx_t *tmp;
LASSERT(!in_interrupt());
kiblnd_abort_txs(kib_conn_t *conn, struct list_head *txs)
{
LIST_HEAD(zombies);
- struct list_head *tmp;
- struct list_head *nxt;
- kib_tx_t *tx;
+ struct list_head *tmp;
+ struct list_head *nxt;
+ kib_tx_t *tx;
spin_lock(&conn->ibc_lock);
kiblnd_peer_connect_failed(kib_peer_t *peer, int active, int error)
{
LIST_HEAD(zombies);
- unsigned long flags;
+ unsigned long flags;
LASSERT(error != 0);
LASSERT(!in_interrupt());
void
kiblnd_connreq_done(kib_conn_t *conn, int status)
{
- kib_peer_t *peer = conn->ibc_peer;
- kib_tx_t *tx;
+ kib_peer_t *peer = conn->ibc_peer;
+ kib_tx_t *tx;
kib_tx_t *tmp;
- struct list_head txs;
- unsigned long flags;
- int active;
+ struct list_head txs;
+ unsigned long flags;
+ int active;
active = (conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
static void
kiblnd_reject(struct rdma_cm_id *cmid, kib_rej_t *rej)
{
- int rc;
+ int rc;
rc = rdma_reject(cmid, rej, sizeof(*rej));
static int
kiblnd_passive_connect(struct rdma_cm_id *cmid, void *priv, int priv_nob)
{
- rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
- kib_msg_t *reqmsg = priv;
- kib_msg_t *ackmsg;
- kib_dev_t *ibdev;
- kib_peer_t *peer;
- kib_peer_t *peer2;
- kib_conn_t *conn;
- lnet_ni_t *ni = NULL;
- kib_net_t *net = NULL;
- lnet_nid_t nid;
+ rwlock_t *g_lock = &kiblnd_data.kib_global_lock;
+ kib_msg_t *reqmsg = priv;
+ kib_msg_t *ackmsg;
+ kib_dev_t *ibdev;
+ kib_peer_t *peer;
+ kib_peer_t *peer2;
+ kib_conn_t *conn;
+ lnet_ni_t *ni = NULL;
+ kib_net_t *net = NULL;
+ lnet_nid_t nid;
struct rdma_conn_param cp;
- kib_rej_t rej;
- int version = IBLND_MSG_VERSION;
- unsigned long flags;
- int rc;
- struct sockaddr_in *peer_addr;
+ kib_rej_t rej;
+ int version = IBLND_MSG_VERSION;
+ unsigned long flags;
+ int rc;
+ struct sockaddr_in *peer_addr;
+
LASSERT(!in_interrupt());
/* cmid inherits 'context' from the corresponding listener id */
LASSERT(ibdev != NULL);
memset(&rej, 0, sizeof(rej));
- rej.ibr_magic = IBLND_MSG_MAGIC;
- rej.ibr_why = IBLND_REJECT_FATAL;
+ rej.ibr_magic = IBLND_MSG_MAGIC;
+ rej.ibr_why = IBLND_REJECT_FATAL;
rej.ibr_cp.ibcp_max_msg_size = IBLND_MSG_SIZE;
peer_addr = (struct sockaddr_in *)&(cmid->route.addr.dst_addr);
if (*kiblnd_tunables.kib_require_priv_port &&
ntohs(peer_addr->sin_port) >= PROT_SOCK) {
__u32 ip = ntohl(peer_addr->sin_addr.s_addr);
+
CERROR("Peer's port (%pI4h:%hu) is not privileged\n",
&ip, ntohs(peer_addr->sin_port));
goto failed;
}
nid = reqmsg->ibm_srcnid;
- ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid));
+ ni = lnet_net2ni(LNET_NIDNET(reqmsg->ibm_dstnid));
if (ni != NULL) {
net = (kib_net_t *)ni->ni_data;
* CM callback doesn't destroy cmid. */
conn->ibc_incarnation = reqmsg->ibm_srcstamp;
- conn->ibc_credits = IBLND_MSG_QUEUE_SIZE(version);
+ conn->ibc_credits = IBLND_MSG_QUEUE_SIZE(version);
conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(version);
LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(version)
<= IBLND_RX_MSGS(version));
memset(&cp, 0, sizeof(cp));
cp.private_data = ackmsg;
- cp.private_data_len = ackmsg->ibm_nob;
+ cp.private_data_len = ackmsg->ibm_nob;
cp.responder_resources = 0; /* No atomic ops or RDMA reads */
- cp.initiator_depth = 0;
+ cp.initiator_depth = 0;
cp.flow_control = 1;
- cp.retry_count = *kiblnd_tunables.kib_retry_count;
- cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
+ cp.retry_count = *kiblnd_tunables.kib_retry_count;
+ cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
CDEBUG(D_NET, "Accept %s\n", libcfs_nid2str(nid));
if (ni != NULL)
lnet_ni_decref(ni);
- rej.ibr_version = version;
+ rej.ibr_version = version;
rej.ibr_cp.ibcp_queue_depth = IBLND_MSG_QUEUE_SIZE(version);
rej.ibr_cp.ibcp_max_frags = IBLND_RDMA_FRAGS(version);
kiblnd_reject(cmid, &rej);
kiblnd_reconnect(kib_conn_t *conn, int version,
__u64 incarnation, int why, kib_connparams_t *cp)
{
- kib_peer_t *peer = conn->ibc_peer;
- char *reason;
- int retry = 0;
- unsigned long flags;
+ kib_peer_t *peer = conn->ibc_peer;
+ char *reason;
+ int retry = 0;
+ unsigned long flags;
LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
LASSERT(peer->ibp_connecting > 0); /* 'conn' at least */
static void
kiblnd_rejected(kib_conn_t *conn, int reason, void *priv, int priv_nob)
{
- kib_peer_t *peer = conn->ibc_peer;
+ kib_peer_t *peer = conn->ibc_peer;
LASSERT(!in_interrupt());
LASSERT(conn->ibc_state == IBLND_CONN_ACTIVE_CONNECT);
case IB_CM_REJ_CONSUMER_DEFINED:
if (priv_nob >= offsetof(kib_rej_t, ibr_padding)) {
- kib_rej_t *rej = priv;
- kib_connparams_t *cp = NULL;
- int flip = 0;
- __u64 incarnation = -1;
+ kib_rej_t *rej = priv;
+ kib_connparams_t *cp = NULL;
+ int flip = 0;
+ __u64 incarnation = -1;
/* NB. default incarnation is -1 because:
* a) V1 will ignore dst incarnation in connreq.
static void
kiblnd_check_connreply(kib_conn_t *conn, void *priv, int priv_nob)
{
- kib_peer_t *peer = conn->ibc_peer;
- lnet_ni_t *ni = peer->ibp_ni;
- kib_net_t *net = ni->ni_data;
- kib_msg_t *msg = priv;
- int ver = conn->ibc_version;
- int rc = kiblnd_unpack_msg(msg, priv_nob);
- unsigned long flags;
+ kib_peer_t *peer = conn->ibc_peer;
+ lnet_ni_t *ni = peer->ibp_ni;
+ kib_net_t *net = ni->ni_data;
+ kib_msg_t *msg = priv;
+ int ver = conn->ibc_version;
+ int rc = kiblnd_unpack_msg(msg, priv_nob);
+ unsigned long flags;
LASSERT(net != NULL);
goto failed;
}
- conn->ibc_incarnation = msg->ibm_srcstamp;
- conn->ibc_credits =
+ conn->ibc_incarnation = msg->ibm_srcstamp;
+ conn->ibc_credits =
conn->ibc_reserved_credits = IBLND_MSG_QUEUE_SIZE(ver);
LASSERT(conn->ibc_credits + conn->ibc_reserved_credits + IBLND_OOB_MSGS(ver)
<= IBLND_RX_MSGS(ver));
static int
kiblnd_active_connect(struct rdma_cm_id *cmid)
{
- kib_peer_t *peer = (kib_peer_t *)cmid->context;
- kib_conn_t *conn;
- kib_msg_t *msg;
- struct rdma_conn_param cp;
- int version;
- __u64 incarnation;
- unsigned long flags;
- int rc;
+ kib_peer_t *peer = (kib_peer_t *)cmid->context;
+ kib_conn_t *conn;
+ kib_msg_t *msg;
+ struct rdma_conn_param cp;
+ int version;
+ __u64 incarnation;
+ unsigned long flags;
+ int rc;
read_lock_irqsave(&kiblnd_data.kib_global_lock, flags);
incarnation = peer->ibp_incarnation;
- version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION :
- peer->ibp_version;
+ version = (peer->ibp_version == 0) ? IBLND_MSG_VERSION :
+ peer->ibp_version;
read_unlock_irqrestore(&kiblnd_data.kib_global_lock, flags);
cp.private_data_len = msg->ibm_nob;
cp.responder_resources = 0; /* No atomic ops or RDMA reads */
cp.initiator_depth = 0;
- cp.flow_control = 1;
- cp.retry_count = *kiblnd_tunables.kib_retry_count;
+ cp.flow_control = 1;
+ cp.retry_count = *kiblnd_tunables.kib_retry_count;
cp.rnr_retry_count = *kiblnd_tunables.kib_rnr_retry_count;
LASSERT(cmid->context == (void *)conn);
int
kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event)
{
- kib_peer_t *peer;
- kib_conn_t *conn;
- int rc;
+ kib_peer_t *peer;
+ kib_conn_t *conn;
+ int rc;
switch (event->event) {
default:
static int
kiblnd_check_txs_locked(kib_conn_t *conn, struct list_head *txs)
{
- kib_tx_t *tx;
- struct list_head *ttmp;
+ kib_tx_t *tx;
+ struct list_head *ttmp;
list_for_each(ttmp, txs) {
tx = list_entry(ttmp, kib_tx_t, tx_list);
{
LIST_HEAD(closes);
LIST_HEAD(checksends);
- struct list_head *peers = &kiblnd_data.kib_peers[idx];
- struct list_head *ptmp;
- kib_peer_t *peer;
- kib_conn_t *conn;
+ struct list_head *peers = &kiblnd_data.kib_peers[idx];
+ struct list_head *ptmp;
+ kib_peer_t *peer;
+ kib_conn_t *conn;
kib_conn_t *tmp;
- struct list_head *ctmp;
- unsigned long flags;
+ struct list_head *ctmp;
+ unsigned long flags;
/* NB. We expect to have a look at all the peers and not find any
* RDMAs to time out, so we just use a shared lock while we
int
kiblnd_connd(void *arg)
{
- wait_queue_t wait;
- unsigned long flags;
- kib_conn_t *conn;
- int timeout;
- int i;
- int dropped_lock;
- int peer_index = 0;
- unsigned long deadline = jiffies;
+ wait_queue_t wait;
+ unsigned long flags;
+ kib_conn_t *conn;
+ int timeout;
+ int i;
+ int dropped_lock;
+ int peer_index = 0;
+ unsigned long deadline = jiffies;
cfs_block_allsigs();
dropped_lock = 0;
if (!list_empty(&kiblnd_data.kib_connd_zombies)) {
- conn = list_entry(kiblnd_data. \
- kib_connd_zombies.next,
+ conn = list_entry(kiblnd_data.kib_connd_zombies.next,
kib_conn_t, ibc_list);
list_del(&conn->ibc_list);
if (timeout <= 0) {
const int n = 4;
const int p = 1;
- int chunk = kiblnd_data.kib_peer_hash_size;
+ int chunk = kiblnd_data.kib_peer_hash_size;
spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);
dropped_lock = 1;
* consuming my CQ I could be called after all completions have
* occurred. But in this case, ibc_nrx == 0 && ibc_nsends_posted == 0
* and this CQ is about to be destroyed so I NOOP. */
- kib_conn_t *conn = (kib_conn_t *)arg;
- struct kib_sched_info *sched = conn->ibc_sched;
- unsigned long flags;
+ kib_conn_t *conn = arg;
+ struct kib_sched_info *sched = conn->ibc_sched;
+ unsigned long flags;
LASSERT(cq == conn->ibc_cq);
int
kiblnd_scheduler(void *arg)
{
- long id = (long)arg;
- struct kib_sched_info *sched;
- kib_conn_t *conn;
- wait_queue_t wait;
- unsigned long flags;
- struct ib_wc wc;
- int did_something;
- int busy_loops = 0;
- int rc;
+ long id = (long)arg;
+ struct kib_sched_info *sched;
+ kib_conn_t *conn;
+ wait_queue_t wait;
+ unsigned long flags;
+ struct ib_wc wc;
+ int did_something;
+ int busy_loops = 0;
+ int rc;
cfs_block_allsigs();
int
kiblnd_failover_thread(void *arg)
{
- rwlock_t *glock = &kiblnd_data.kib_global_lock;
- kib_dev_t *dev;
- wait_queue_t wait;
- unsigned long flags;
- int rc;
+ rwlock_t *glock = &kiblnd_data.kib_global_lock;
+ kib_dev_t *dev;
+ wait_queue_t wait;
+ unsigned long flags;
+ int rc;
LASSERT(*kiblnd_tunables.kib_dev_failover != 0);
write_lock_irqsave(glock, flags);
while (!kiblnd_data.kib_shutdown) {
- int do_failover = 0;
- int long_sleep;
+ int do_failover = 0;
+ int long_sleep;
list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
ibd_fail_list) {
Code Review / kvmfornfv.git / blobdiff