}
}
-static int kiblnd_unpack_rd(struct kib_msg *msg, int flip)
+static int kiblnd_unpack_rd(struct kib_msg *msg, bool flip)
{
struct kib_rdma_desc *rd;
- int nob;
- int n;
- int i;
+ int nob;
+ int n;
+ int i;
- LASSERT (msg->ibm_type == IBLND_MSG_GET_REQ ||
- msg->ibm_type == IBLND_MSG_PUT_ACK);
+ LASSERT(msg->ibm_type == IBLND_MSG_GET_REQ ||
+ msg->ibm_type == IBLND_MSG_PUT_ACK);
- rd = msg->ibm_type == IBLND_MSG_GET_REQ ?
- &msg->ibm_u.get.ibgm_rd :
- &msg->ibm_u.putack.ibpam_rd;
+ rd = msg->ibm_type == IBLND_MSG_GET_REQ ?
+ &msg->ibm_u.get.ibgm_rd :
+ &msg->ibm_u.putack.ibpam_rd;
- if (flip) {
- __swab32s(&rd->rd_key);
- __swab32s(&rd->rd_nfrags);
- }
+ if (flip) {
+ __swab32s(&rd->rd_key);
+ __swab32s(&rd->rd_nfrags);
+ }
- n = rd->rd_nfrags;
+ n = rd->rd_nfrags;
- if (n <= 0 || n > IBLND_MAX_RDMA_FRAGS) {
- CERROR("Bad nfrags: %d, should be 0 < n <= %d\n",
- n, IBLND_MAX_RDMA_FRAGS);
- return 1;
- }
+ if (n <= 0 || n > IBLND_MAX_RDMA_FRAGS) {
+ CERROR("Bad nfrags: %d, should be 0 < n <= %d\n",
+ n, IBLND_MAX_RDMA_FRAGS);
+ return 1;
+ }
nob = offsetof(struct kib_msg, ibm_u) +
- kiblnd_rd_msg_size(rd, msg->ibm_type, n);
+ kiblnd_rd_msg_size(rd, msg->ibm_type, n);
- if (msg->ibm_nob < nob) {
- CERROR("Short %s: %d(%d)\n",
- kiblnd_msgtype2str(msg->ibm_type), msg->ibm_nob, nob);
- return 1;
- }
+ if (msg->ibm_nob < nob) {
+ CERROR("Short %s: %d(%d)\n",
+ kiblnd_msgtype2str(msg->ibm_type), msg->ibm_nob, nob);
+ return 1;
+ }
- if (!flip)
- return 0;
+ if (!flip)
+ return 0;
- for (i = 0; i < n; i++) {
- __swab32s(&rd->rd_frags[i].rf_nob);
- __swab64s(&rd->rd_frags[i].rf_addr);
- }
+ for (i = 0; i < n; i++) {
+ __swab32s(&rd->rd_frags[i].rf_nob);
+ __swab64s(&rd->rd_frags[i].rf_addr);
+ }
- return 0;
+ return 0;
}
void kiblnd_pack_msg(struct lnet_ni *ni, struct kib_msg *msg, int version,
int kiblnd_unpack_msg(struct kib_msg *msg, int nob)
{
const int hdr_size = offsetof(struct kib_msg, ibm_u);
- __u32 msg_cksum;
- __u16 version;
- int msg_nob;
- int flip;
-
- /* 6 bytes are enough to have received magic + version */
- if (nob < 6) {
- CERROR("Short message: %d\n", nob);
- return -EPROTO;
- }
+ __u32 msg_cksum;
+ __u16 version;
+ int msg_nob;
+ bool flip;
- if (msg->ibm_magic == IBLND_MSG_MAGIC) {
- flip = 0;
- } else if (msg->ibm_magic == __swab32(IBLND_MSG_MAGIC)) {
- flip = 1;
- } else {
- CERROR("Bad magic: %08x\n", msg->ibm_magic);
- return -EPROTO;
- }
+ /* 6 bytes are enough to have received magic + version */
+ if (nob < 6) {
+ CERROR("Short message: %d\n", nob);
+ return -EPROTO;
+ }
- version = flip ? __swab16(msg->ibm_version) : msg->ibm_version;
- if (version != IBLND_MSG_VERSION &&
- version != IBLND_MSG_VERSION_1) {
- CERROR("Bad version: %x\n", version);
- return -EPROTO;
- }
+ if (msg->ibm_magic == IBLND_MSG_MAGIC) {
+ flip = false;
+ } else if (msg->ibm_magic == __swab32(IBLND_MSG_MAGIC)) {
+ flip = true;
+ } else {
+ CERROR("Bad magic: %08x\n", msg->ibm_magic);
+ return -EPROTO;
+ }
- if (nob < hdr_size) {
- CERROR("Short message: %d\n", nob);
- return -EPROTO;
- }
+ version = flip ? __swab16(msg->ibm_version) : msg->ibm_version;
+ if (version != IBLND_MSG_VERSION &&
+ version != IBLND_MSG_VERSION_1) {
+ CERROR("Bad version: %x\n", version);
+ return -EPROTO;
+ }
- msg_nob = flip ? __swab32(msg->ibm_nob) : msg->ibm_nob;
- if (msg_nob > nob) {
- CERROR("Short message: got %d, wanted %d\n", nob, msg_nob);
- return -EPROTO;
- }
+ if (nob < hdr_size) {
+ CERROR("Short message: %d\n", nob);
+ return -EPROTO;
+ }
- /* checksum must be computed with ibm_cksum zero and BEFORE anything
- * gets flipped */
- msg_cksum = flip ? __swab32(msg->ibm_cksum) : msg->ibm_cksum;
- msg->ibm_cksum = 0;
- if (msg_cksum != 0 &&
- msg_cksum != kiblnd_cksum(msg, msg_nob)) {
- CERROR("Bad checksum\n");
- return -EPROTO;
- }
+ msg_nob = flip ? __swab32(msg->ibm_nob) : msg->ibm_nob;
+ if (msg_nob > nob) {
+ CERROR("Short message: got %d, wanted %d\n", nob, msg_nob);
+ return -EPROTO;
+ }
- msg->ibm_cksum = msg_cksum;
+ /* checksum must be computed with ibm_cksum zero and BEFORE anything
+ * gets flipped
+ */
+ msg_cksum = flip ? __swab32(msg->ibm_cksum) : msg->ibm_cksum;
+ msg->ibm_cksum = 0;
+ if (msg_cksum != 0 &&
+ msg_cksum != kiblnd_cksum(msg, msg_nob)) {
+ CERROR("Bad checksum\n");
+ return -EPROTO;
+ }
+
+ msg->ibm_cksum = msg_cksum;
- if (flip) {
- /* leave magic unflipped as a clue to peer_ni endianness */
- msg->ibm_version = version;
+ if (flip) {
+ /* leave magic unflipped as a clue to peer_ni endianness */
+ msg->ibm_version = version;
BUILD_BUG_ON(sizeof(msg->ibm_type) != 1);
BUILD_BUG_ON(sizeof(msg->ibm_credits) != 1);
- msg->ibm_nob = msg_nob;
- __swab64s(&msg->ibm_srcnid);
- __swab64s(&msg->ibm_srcstamp);
- __swab64s(&msg->ibm_dstnid);
- __swab64s(&msg->ibm_dststamp);
- }
+ msg->ibm_nob = msg_nob;
+ __swab64s(&msg->ibm_srcnid);
+ __swab64s(&msg->ibm_srcstamp);
+ __swab64s(&msg->ibm_dstnid);
+ __swab64s(&msg->ibm_dststamp);
+ }
- if (msg->ibm_srcnid == LNET_NID_ANY) {
- CERROR("Bad src nid: %s\n", libcfs_nid2str(msg->ibm_srcnid));
- return -EPROTO;
- }
+ if (msg->ibm_srcnid == LNET_NID_ANY) {
+ CERROR("Bad src nid: %s\n", libcfs_nid2str(msg->ibm_srcnid));
+ return -EPROTO;
+ }
- if (msg_nob < kiblnd_msgtype2size(msg->ibm_type)) {
- CERROR("Short %s: %d(%d)\n", kiblnd_msgtype2str(msg->ibm_type),
- msg_nob, kiblnd_msgtype2size(msg->ibm_type));
- return -EPROTO;
- }
+ if (msg_nob < kiblnd_msgtype2size(msg->ibm_type)) {
+ CERROR("Short %s: %d(%d)\n", kiblnd_msgtype2str(msg->ibm_type),
+ msg_nob, kiblnd_msgtype2size(msg->ibm_type));
+ return -EPROTO;
+ }
- switch (msg->ibm_type) {
- default:
- CERROR("Unknown message type %x\n", msg->ibm_type);
- return -EPROTO;
+ switch (msg->ibm_type) {
+ default:
+ CERROR("Unknown message type %x\n", msg->ibm_type);
+ return -EPROTO;
- case IBLND_MSG_NOOP:
- case IBLND_MSG_IMMEDIATE:
- case IBLND_MSG_PUT_REQ:
- break;
+ case IBLND_MSG_NOOP:
+ case IBLND_MSG_IMMEDIATE:
+ case IBLND_MSG_PUT_REQ:
+ break;
- case IBLND_MSG_PUT_ACK:
- case IBLND_MSG_GET_REQ:
- if (kiblnd_unpack_rd(msg, flip))
- return -EPROTO;
- break;
+ case IBLND_MSG_PUT_ACK:
+ case IBLND_MSG_GET_REQ:
+ if (kiblnd_unpack_rd(msg, flip))
+ return -EPROTO;
+ break;
- case IBLND_MSG_PUT_NAK:
- case IBLND_MSG_PUT_DONE:
- case IBLND_MSG_GET_DONE:
- if (flip)
- __swab32s(&msg->ibm_u.completion.ibcm_status);
- break;
+ case IBLND_MSG_PUT_NAK:
+ case IBLND_MSG_PUT_DONE:
+ case IBLND_MSG_GET_DONE:
+ if (flip)
+ __swab32s(&msg->ibm_u.completion.ibcm_status);
+ break;
- case IBLND_MSG_CONNREQ:
- case IBLND_MSG_CONNACK:
- if (flip) {
- __swab16s(&msg->ibm_u.connparams.ibcp_queue_depth);
- __swab16s(&msg->ibm_u.connparams.ibcp_max_frags);
- __swab32s(&msg->ibm_u.connparams.ibcp_max_msg_size);
- }
- break;
- }
- return 0;
+ case IBLND_MSG_CONNREQ:
+ case IBLND_MSG_CONNACK:
+ if (flip) {
+ __swab16s(&msg->ibm_u.connparams.ibcp_queue_depth);
+ __swab16s(&msg->ibm_u.connparams.ibcp_max_frags);
+ __swab32s(&msg->ibm_u.connparams.ibcp_max_msg_size);
+ }
+ break;
+ }
+ return 0;
}
int
bool is_rx = (rd != tx->tx_rd);
#ifdef HAVE_FMR_POOL_API
__u64 *pages = tx->tx_pages;
- bool tx_pages_mapped = 0;
+ bool tx_pages_mapped = false;
int npages = 0;
#endif
int rc;
if (!tx_pages_mapped) {
npages = kiblnd_map_tx_pages(tx, rd);
- tx_pages_mapped = 1;
+ tx_pages_mapped = true;
}
pfmr = kib_fmr_pool_map(fpo->fmr.fpo_fmr_pool,
pages, npages, iov);
if (likely(!IS_ERR(pfmr))) {
fmr->fmr_key = is_rx ? pfmr->fmr->rkey
- : pfmr->fmr->lkey;
+ : pfmr->fmr->lkey;
fmr->fmr_frd = NULL;
fmr->fmr_pfmr = pfmr;
fmr->fmr_pool = fpo;
#endif
struct ib_mr *mr;
- frd = list_first_entry(&fpo->fast_reg.fpo_pool_list,
- struct kib_fast_reg_descriptor,
- frd_list);
+ frd = list_first_entry(
+ &fpo->fast_reg.fpo_pool_list,
+ struct kib_fast_reg_descriptor,
+ frd_list);
list_del(&frd->frd_list);
spin_unlock(&fps->fps_lock);
rd->rd_nfrags, PAGE_SIZE);
#endif /* HAVE_IB_MAP_MR_SG_5ARGS */
if (unlikely(n != rd->rd_nfrags)) {
- CERROR("Failed to map mr %d/%d "
- "elements\n", n, rd->rd_nfrags);
+ CERROR("Failed to map mr %d/%d elements\n",
+ n, rd->rd_nfrags);
return n < 0 ? n : -EINVAL;
}
#else /* HAVE_IB_MAP_MR_SG */
if (!tx_pages_mapped) {
npages = kiblnd_map_tx_pages(tx, rd);
- tx_pages_mapped = 1;
+ tx_pages_mapped = true;
}
LASSERT(npages <= frpl->max_page_list_len);
memcpy(frpl->page_list, pages,
- sizeof(*pages) * npages);
+ sizeof(*pages) * npages);
/* Prepare FastReg WR */
wr = &frd->frd_fastreg_wr;
wr->wr.wr.fast_reg.page_shift = PAGE_SHIFT;
wr->wr.wr.fast_reg.length = nob;
wr->wr.wr.fast_reg.rkey =
- is_rx ? mr->rkey : mr->lkey;
+ is_rx ? mr->rkey : mr->lkey;
wr->wr.wr.fast_reg.access_flags =
- (IB_ACCESS_LOCAL_WRITE |
- IB_ACCESS_REMOTE_WRITE);
+ (IB_ACCESS_LOCAL_WRITE |
+ IB_ACCESS_REMOTE_WRITE);
#endif /* HAVE_IB_MAP_MR_SG */
fmr->fmr_key = is_rx ? mr->rkey : mr->lkey;
kiblnd_txlist_done(&zombies, -ECONNABORTED, LNET_MSG_STATUS_OK);
}
-static int
+static bool
kiblnd_tx_may_discard(struct kib_conn *conn)
{
- int rc = 0;
+ bool rc = false;
struct kib_tx *nxt;
struct kib_tx *tx;
if (tx->tx_sending == 0) {
kiblnd_conn_decref(tx->tx_conn);
tx->tx_conn = NULL;
- rc = 1;
+ rc = true;
}
}
}
*kiblnd_tunables.kib_service);
break;
- case IB_CM_REJ_CONSUMER_DEFINED:
+ case IB_CM_REJ_CONSUMER_DEFINED:
if (priv_nob >= offsetof(struct kib_rej, ibr_padding)) {
struct kib_rej *rej = priv;
struct kib_connparams *cp = NULL;
- int flip = 0;
- __u64 incarnation = -1;
-
- /* NB. default incarnation is -1 because:
- * a) V1 will ignore dst incarnation in connreq.
- * b) V2 will provide incarnation while rejecting me,
- * -1 will be overwrote.
- *
- * if I try to connect to a V1 peer_ni with V2 protocol,
- * it rejected me then upgrade to V2, I have no idea
- * about the upgrading and try to reconnect with V1,
- * in this case upgraded V2 can find out I'm trying to
- * talk to the old guy and reject me(incarnation is -1).
- */
-
- if (rej->ibr_magic == __swab32(IBLND_MSG_MAGIC) ||
- rej->ibr_magic == __swab32(LNET_PROTO_MAGIC)) {
- __swab32s(&rej->ibr_magic);
- __swab16s(&rej->ibr_version);
- flip = 1;
- }
+ bool flip = false;
+ __u64 incarnation = -1;
+
+ /* NB. default incarnation is -1 because:
+ * a) V1 will ignore dst incarnation in connreq.
+ * b) V2 will provide incarnation while rejecting me,
+ * -1 will be overwrote.
+ *
+ * if I try to connect to a V1 peer_ni with V2 protocol,
+ * it rejected me then upgrade to V2, I have no idea
+ * about the upgrading and try to reconnect with V1,
+ * in this case upgraded V2 can find out I'm trying to
+ * talk to the old guy and reject me(incarnation is -1).
+ */
+
+ if (rej->ibr_magic == __swab32(IBLND_MSG_MAGIC) ||
+ rej->ibr_magic == __swab32(LNET_PROTO_MAGIC)) {
+ __swab32s(&rej->ibr_magic);
+ __swab16s(&rej->ibr_version);
+ flip = true;
+ }
if (priv_nob >= sizeof(struct kib_rej) &&
- rej->ibr_version > IBLND_MSG_VERSION_1) {
- /* priv_nob is always 148 in current version
- * of OFED, so we still need to check version.
- * (define of IB_CM_REJ_PRIVATE_DATA_SIZE) */
- cp = &rej->ibr_cp;
-
- if (flip) {
- __swab64s(&rej->ibr_incarnation);
- __swab16s(&cp->ibcp_queue_depth);
- __swab16s(&cp->ibcp_max_frags);
- __swab32s(&cp->ibcp_max_msg_size);
- }
-
- incarnation = rej->ibr_incarnation;
- }
-
- if (rej->ibr_magic != IBLND_MSG_MAGIC &&
- rej->ibr_magic != LNET_PROTO_MAGIC) {
- CERROR("%s rejected: consumer defined fatal error\n",
- libcfs_nid2str(peer_ni->ibp_nid));
- break;
- }
-
- if (rej->ibr_version != IBLND_MSG_VERSION &&
- rej->ibr_version != IBLND_MSG_VERSION_1) {
- CERROR("%s rejected: o2iblnd version %x error\n",
- libcfs_nid2str(peer_ni->ibp_nid),
- rej->ibr_version);
- break;
- }
-
- if (rej->ibr_why == IBLND_REJECT_FATAL &&
- rej->ibr_version == IBLND_MSG_VERSION_1) {
- CDEBUG(D_NET, "rejected by old version peer_ni %s: %x\n",
- libcfs_nid2str(peer_ni->ibp_nid), rej->ibr_version);
-
- if (conn->ibc_version != IBLND_MSG_VERSION_1)
- rej->ibr_why = IBLND_REJECT_CONN_UNCOMPAT;
- }
-
- switch (rej->ibr_why) {
- case IBLND_REJECT_CONN_RACE:
- case IBLND_REJECT_CONN_STALE:
- case IBLND_REJECT_CONN_UNCOMPAT:
+ rej->ibr_version > IBLND_MSG_VERSION_1) {
+ /* priv_nob is always 148 in current version
+ * of OFED, so we still need to check version.
+ * (define of IB_CM_REJ_PRIVATE_DATA_SIZE)
+ */
+ cp = &rej->ibr_cp;
+
+ if (flip) {
+ __swab64s(&rej->ibr_incarnation);
+ __swab16s(&cp->ibcp_queue_depth);
+ __swab16s(&cp->ibcp_max_frags);
+ __swab32s(&cp->ibcp_max_msg_size);
+ }
+
+ incarnation = rej->ibr_incarnation;
+ }
+
+ if (rej->ibr_magic != IBLND_MSG_MAGIC &&
+ rej->ibr_magic != LNET_PROTO_MAGIC) {
+ CERROR("%s rejected: consumer defined fatal error\n",
+ libcfs_nid2str(peer_ni->ibp_nid));
+ break;
+ }
+
+ if (rej->ibr_version != IBLND_MSG_VERSION &&
+ rej->ibr_version != IBLND_MSG_VERSION_1) {
+ CERROR("%s rejected: o2iblnd version %x error\n",
+ libcfs_nid2str(peer_ni->ibp_nid),
+ rej->ibr_version);
+ break;
+ }
+
+ if (rej->ibr_why == IBLND_REJECT_FATAL &&
+ rej->ibr_version == IBLND_MSG_VERSION_1) {
+ CDEBUG(D_NET, "rejected by old version peer_ni %s: %x\n",
+ libcfs_nid2str(peer_ni->ibp_nid),
+ rej->ibr_version);
+
+ if (conn->ibc_version != IBLND_MSG_VERSION_1)
+ rej->ibr_why = IBLND_REJECT_CONN_UNCOMPAT;
+ }
+
+ switch (rej->ibr_why) {
+ case IBLND_REJECT_CONN_RACE:
+ case IBLND_REJECT_CONN_STALE:
+ case IBLND_REJECT_CONN_UNCOMPAT:
case IBLND_REJECT_MSG_QUEUE_SIZE:
case IBLND_REJECT_RDMA_FRAGS:
kiblnd_check_reconnect(conn, rej->ibr_version,
- incarnation, rej->ibr_why, cp);
- break;
-
- case IBLND_REJECT_NO_RESOURCES:
- CERROR("%s rejected: o2iblnd no resources\n",
- libcfs_nid2str(peer_ni->ibp_nid));
- break;
-
- case IBLND_REJECT_FATAL:
- CERROR("%s rejected: o2iblnd fatal error\n",
- libcfs_nid2str(peer_ni->ibp_nid));
- break;
-
- default:
- CERROR("%s rejected: o2iblnd reason %d\n",
- libcfs_nid2str(peer_ni->ibp_nid),
- rej->ibr_why);
- break;
- }
- break;
- }
- /* fall through */
- default:
- CNETERR("%s rejected: reason %d, size %d\n",
- libcfs_nid2str(peer_ni->ibp_nid), reason, priv_nob);
- break;
- }
+ incarnation,
+ rej->ibr_why, cp);
+ break;
+
+ case IBLND_REJECT_NO_RESOURCES:
+ CERROR("%s rejected: o2iblnd no resources\n",
+ libcfs_nid2str(peer_ni->ibp_nid));
+ break;
+
+ case IBLND_REJECT_FATAL:
+ CERROR("%s rejected: o2iblnd fatal error\n",
+ libcfs_nid2str(peer_ni->ibp_nid));
+ break;
+
+ default:
+ CERROR("%s rejected: o2iblnd reason %d\n",
+ libcfs_nid2str(peer_ni->ibp_nid),
+ rej->ibr_why);
+ break;
+ }
+ break;
+ }
+ /* fall through */
+ default:
+ CNETERR("%s rejected: reason %d, size %d\n",
+ libcfs_nid2str(peer_ni->ibp_nid), reason, priv_nob);
+ break;
+ }
- kiblnd_connreq_done(conn, -ECONNREFUSED);
+ kiblnd_connreq_done(conn, -ECONNREFUSED);
}
static void
struct kib_conn *conn;
int timeout;
int i;
- int dropped_lock;
+ bool dropped_lock;
int peer_index = 0;
unsigned long deadline = jiffies;
while (!kiblnd_data.kib_shutdown) {
int reconn = 0;
- dropped_lock = 0;
+ dropped_lock = false;
if (!list_empty(&kiblnd_data.kib_connd_zombies)) {
struct kib_peer_ni *peer_ni = NULL;
}
spin_unlock_irqrestore(lock, flags);
- dropped_lock = 1;
+ dropped_lock = true;
kiblnd_destroy_conn(conn);
list_del(&conn->ibc_list);
spin_unlock_irqrestore(lock, flags);
- dropped_lock = 1;
+ dropped_lock = true;
kiblnd_disconnect_conn(conn);
wait = conn->ibc_waits;
list_del(&conn->ibc_list);
spin_unlock_irqrestore(lock, flags);
- dropped_lock = 1;
+ dropped_lock = true;
reconn += kiblnd_reconnect_peer(conn->ibc_peer);
kiblnd_peer_decref(conn->ibc_peer);
kiblnd_conn_decref(conn);
spin_lock_irqsave(lock, flags);
- if (dropped_lock == 0)
+ if (!dropped_lock)
list_add_tail(&conn->ibc_list,
&kiblnd_data.kib_connd_waits);
}
unsigned int lnd_timeout;
spin_unlock_irqrestore(lock, flags);
- dropped_lock = 1;
+ dropped_lock = true;
/* Time to check for RDMA timeouts on a few more
* peers: I do checks every 'p' seconds on a
int
kiblnd_scheduler(void *arg)
{
- long id = (long)arg;
- struct kib_sched_info *sched;
+ long id = (long)arg;
+ struct kib_sched_info *sched;
struct kib_conn *conn;
- wait_queue_entry_t wait;
- unsigned long flags;
- struct ib_wc wc;
- int did_something;
- int rc;
+ wait_queue_entry_t wait;
+ unsigned long flags;
+ struct ib_wc wc;
+ bool did_something;
+ int rc;
init_wait(&wait);
rc = cfs_cpt_bind(lnet_cpt_table(), sched->ibs_cpt);
if (rc != 0) {
- CWARN("Unable to bind on CPU partition %d, please verify "
- "whether all CPUs are healthy and reload modules if "
- "necessary, otherwise your system might under risk of "
- "low performance\n", sched->ibs_cpt);
+ CWARN("Unable to bind on CPU partition %d, please verify whether all CPUs are healthy and reload modules if necessary, otherwise your system might under risk of low performance\n", sched->ibs_cpt);
}
spin_lock_irqsave(&sched->ibs_lock, flags);
spin_lock_irqsave(&sched->ibs_lock, flags);
}
- did_something = 0;
+ did_something = false;
if (!list_empty(&sched->ibs_conns)) {
conn = list_entry(sched->ibs_conns.next,
wc.wr_id = IBLND_WID_INVAL;
- rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
- if (rc == 0) {
- rc = ib_req_notify_cq(conn->ibc_cq,
- IB_CQ_NEXT_COMP);
- if (rc < 0) {
- CWARN("%s: ib_req_notify_cq failed: %d, "
- "closing connection\n",
- libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
- kiblnd_close_conn(conn, -EIO);
- kiblnd_conn_decref(conn);
+ rc = ib_poll_cq(conn->ibc_cq, 1, &wc);
+ if (rc == 0) {
+ rc = ib_req_notify_cq(conn->ibc_cq,
+ IB_CQ_NEXT_COMP);
+ if (rc < 0) {
+ CWARN("%s: ib_req_notify_cq failed: %d, closing connection\n",
+ libcfs_nid2str(conn->ibc_peer->ibp_nid), rc);
+ kiblnd_close_conn(conn, -EIO);
+ kiblnd_conn_decref(conn);
spin_lock_irqsave(&sched->ibs_lock,
- flags);
+ flags);
continue;
}
}
if (rc < 0) {
- CWARN("%s: ib_poll_cq failed: %d, "
- "closing connection\n",
+ CWARN("%s: ib_poll_cq failed: %d, closing connection\n",
libcfs_nid2str(conn->ibc_peer->ibp_nid),
rc);
kiblnd_close_conn(conn, -EIO);
/* +1 ref for sched_conns */
kiblnd_conn_addref(conn);
list_add_tail(&conn->ibc_sched_list,
- &sched->ibs_conns);
+ &sched->ibs_conns);
if (waitqueue_active(&sched->ibs_waitq))
wake_up(&sched->ibs_waitq);
} else {
kiblnd_complete(&wc);
spin_lock_irqsave(&sched->ibs_lock, flags);
- }
+ }
- kiblnd_conn_decref(conn); /* ...drop my ref from above */
- did_something = 1;
- }
+ kiblnd_conn_decref(conn); /* ..drop my ref from above */
+ did_something = true;
+ }
- if (did_something)
- continue;
+ if (did_something)
+ continue;
set_current_state(TASK_INTERRUPTIBLE);
add_wait_queue_exclusive(&sched->ibs_waitq, &wait);
int
kiblnd_failover_thread(void *arg)
{
- rwlock_t *glock = &kiblnd_data.kib_global_lock;
+ rwlock_t *glock = &kiblnd_data.kib_global_lock;
struct kib_dev *dev;
struct net *ns = arg;
wait_queue_entry_t wait;
- unsigned long flags;
- int rc;
+ unsigned long flags;
+ int rc;
LASSERT(*kiblnd_tunables.kib_dev_failover != 0);
init_wait(&wait);
write_lock_irqsave(glock, flags);
- while (!kiblnd_data.kib_shutdown) {
- int do_failover = 0;
- int long_sleep;
+ while (!kiblnd_data.kib_shutdown) {
+ bool do_failover = false;
+ int long_sleep;
list_for_each_entry(dev, &kiblnd_data.kib_failed_devs,
- ibd_fail_list) {
+ ibd_fail_list) {
if (ktime_get_seconds() < dev->ibd_next_failover)
- continue;
- do_failover = 1;
- break;
- }
+ continue;
+ do_failover = true;
+ break;
+ }
- if (do_failover) {
+ if (do_failover) {
list_del_init(&dev->ibd_fail_list);
- dev->ibd_failover = 1;
+ dev->ibd_failover = 1;
write_unlock_irqrestore(glock, flags);
rc = kiblnd_dev_failover(dev, ns);
write_lock_irqsave(glock, flags);
- LASSERT (dev->ibd_failover);
- dev->ibd_failover = 0;
- if (rc >= 0) { /* Device is OK or failover succeed */
+ LASSERT(dev->ibd_failover);
+ dev->ibd_failover = 0;
+ if (rc >= 0) { /* Device is OK or failover succeed */
dev->ibd_next_failover = ktime_get_seconds() + 3;
- continue;
- }
+ continue;
+ }
- /* failed to failover, retry later */
+ /* failed to failover, retry later */
dev->ibd_next_failover = ktime_get_seconds() +
- min(dev->ibd_failed_failover, 10);
- if (kiblnd_dev_can_failover(dev)) {
+ min(dev->ibd_failed_failover, 10);
+ if (kiblnd_dev_can_failover(dev)) {
list_add_tail(&dev->ibd_fail_list,
- &kiblnd_data.kib_failed_devs);
- }
+ &kiblnd_data.kib_failed_devs);
+ }
- continue;
- }
+ continue;
+ }
- /* long sleep if no more pending failover */
+ /* long sleep if no more pending failover */
long_sleep = list_empty(&kiblnd_data.kib_failed_devs);
set_current_state(TASK_INTERRUPTIBLE);
write_unlock_irqrestore(glock, flags);
rc = schedule_timeout(long_sleep ? cfs_time_seconds(10) :
- cfs_time_seconds(1));
+ cfs_time_seconds(1));
set_current_state(TASK_RUNNING);
remove_wait_queue(&kiblnd_data.kib_failover_waitq, &wait);
write_lock_irqsave(glock, flags);
- if (!long_sleep || rc != 0)
- continue;
+ if (!long_sleep || rc != 0)
+ continue;
- /* have a long sleep, routine check all active devices,
- * we need checking like this because if there is not active
- * connection on the dev and no SEND from local, we may listen
- * on wrong HCA for ever while there is a bonding failover */
+ /* have a long sleep, routine check all active devices,
+ * we need checking like this because if there is not active
+ * connection on the dev and no SEND from local, we may listen
+ * on wrong HCA for ever while there is a bonding failover
+ */
list_for_each_entry(dev, &kiblnd_data.kib_devs, ibd_list) {
- if (kiblnd_dev_can_failover(dev)) {
+ if (kiblnd_dev_can_failover(dev)) {
list_add_tail(&dev->ibd_fail_list,
- &kiblnd_data.kib_failed_devs);
- }
- }
- }
+ &kiblnd_data.kib_failed_devs);
+ }
+ }
+ }
write_unlock_irqrestore(glock, flags);
- kiblnd_thread_fini();
- return 0;
+ kiblnd_thread_fini();
+ return 0;
}
git://git.whamcloud.com / fs / lustre-release.git / commitdiff