+/*
+ * Do a health check on the message:
+ * return -1 if we're not going to handle the error or
+ * if we've reached the maximum number of retries.
+ * success case will return -1 as well
+ * return 0 if it the message is requeued for send
+ */
+static int
+lnet_health_check(struct lnet_msg *msg)
+{
+ enum lnet_msg_hstatus hstatus = msg->msg_health_status;
+ struct lnet_peer_ni *lpni;
+ struct lnet_ni *ni;
+ bool lo = false;
+ bool attempt_local_resend;
+ bool attempt_remote_resend;
+ bool handle_local_health;
+ bool handle_remote_health;
+
+ /* if we're shutting down no point in handling health. */
+ if (the_lnet.ln_mt_state != LNET_MT_STATE_RUNNING)
+ return -1;
+
+ LASSERT(msg->msg_tx_committed || msg->msg_rx_committed);
+
+ /*
+ * if we're sending to the LOLND then the msg_txpeer will not be
+ * set. So no need to sanity check it.
+ */
+ if (msg->msg_tx_committed && msg->msg_txni->ni_nid != LNET_NID_LO_0)
+ LASSERT(msg->msg_txpeer);
+ else if (msg->msg_tx_committed &&
+ msg->msg_txni->ni_nid == LNET_NID_LO_0)
+ lo = true;
+
+ if (hstatus != LNET_MSG_STATUS_OK &&
+ ktime_compare(ktime_get(), msg->msg_deadline) >= 0)
+ return -1;
+
+ /*
+ * always prefer txni/txpeer if they message is committed for both
+ * directions.
+ */
+ if (msg->msg_tx_committed) {
+ ni = msg->msg_txni;
+ lpni = msg->msg_txpeer;
+ attempt_local_resend = attempt_remote_resend = true;
+ } else {
+ ni = msg->msg_rxni;
+ lpni = msg->msg_rxpeer;
+ attempt_local_resend = attempt_remote_resend = false;
+ }
+
+ /* Don't further decrement the health value if a recovery message
+ * failed.
+ */
+ if (msg->msg_recovery)
+ handle_local_health = handle_remote_health = false;
+ else
+ handle_local_health = handle_remote_health = true;
+
+ /* For local failures, health/recovery/resends are not needed if I only
+ * have a single (non-lolnd) interface. NB: pb_nnis includes the lolnd
+ * interface, so a single-rail node would have pb_nnis == 2.
+ */
+ if (the_lnet.ln_ping_target->pb_nnis <= 2) {
+ handle_local_health = false;
+ attempt_local_resend = false;
+ }
+
+ /* For remote failures, health/recovery/resends are not needed if the
+ * peer only has a single interface. Special case for routers where we
+ * rely on health feature to manage route aliveness. NB: unlike pb_nnis
+ * above, lp_nnis does _not_ include the lolnd, so a single-rail node
+ * would have lp_nnis == 1.
+ */
+ if (lpni && lpni->lpni_peer_net->lpn_peer->lp_nnis <= 1) {
+ attempt_remote_resend = false;
+ if (!lnet_isrouter(lpni))
+ handle_remote_health = false;
+ }
+
+ if (!lo)
+ LASSERT(ni && lpni);
+ else
+ LASSERT(ni);
+
+ CDEBUG(D_NET, "health check: %s->%s: %s: %s\n",
+ libcfs_nid2str(ni->ni_nid),
+ (lo) ? "self" : libcfs_nid2str(lpni->lpni_nid),
+ lnet_msgtyp2str(msg->msg_type),
+ lnet_health_error2str(hstatus));
+
+ /*
+ * stats are only incremented for errors so avoid wasting time
+ * incrementing statistics if there is no error.
+ */
+ if (hstatus != LNET_MSG_STATUS_OK) {
+ lnet_net_lock(0);
+ lnet_incr_hstats(ni, lpni, hstatus);
+ lnet_net_unlock(0);
+ }
+
+ switch (hstatus) {
+ case LNET_MSG_STATUS_OK:
+ /*
+ * increment the local ni health weather we successfully
+ * received or sent a message on it.
+ */
+ lnet_inc_healthv(&ni->ni_healthv, lnet_health_sensitivity);
+ /*
+ * It's possible msg_txpeer is NULL in the LOLND
+ * case. Only increment the peer's health if we're
+ * receiving a message from it. It's the only sure way to
+ * know that a remote interface is up.
+ * If this interface is part of a router, then take that
+ * as indication that the router is fully healthy.
+ */
+ if (lpni && msg->msg_rx_committed) {
+ /*
+ * If we're receiving a message from the router or
+ * I'm a router, then set that lpni's health to
+ * maximum so we can commence communication
+ */
+ lnet_net_lock(0);
+ if (lnet_isrouter(lpni) || the_lnet.ln_routing) {
+ lnet_set_lpni_healthv_locked(lpni,
+ LNET_MAX_HEALTH_VALUE);
+ } else {
+ __u32 sensitivity = lpni->lpni_peer_net->
+ lpn_peer->lp_health_sensitivity;
+
+ lnet_inc_lpni_healthv_locked(lpni,
+ (sensitivity) ? sensitivity :
+ lnet_health_sensitivity);
+ /* This peer NI may have previously aged out
+ * of recovery. Now that we've received a
+ * message from it, we can continue recovery
+ * if its health value is still below the
+ * maximum.
+ */
+ lnet_peer_ni_add_to_recoveryq_locked(lpni,
+ &the_lnet.ln_mt_peerNIRecovq,
+ ktime_get_seconds());
+ }
+ lnet_net_unlock(0);
+ }
+
+ /* we can finalize this message */
+ return -1;
+ case LNET_MSG_STATUS_LOCAL_INTERRUPT:
+ case LNET_MSG_STATUS_LOCAL_DROPPED:
+ case LNET_MSG_STATUS_LOCAL_ABORTED:
+ case LNET_MSG_STATUS_LOCAL_NO_ROUTE:
+ case LNET_MSG_STATUS_LOCAL_TIMEOUT:
+ if (handle_local_health)
+ lnet_handle_local_failure(ni);
+ if (attempt_local_resend)
+ return lnet_attempt_msg_resend(msg);
+ break;
+ case LNET_MSG_STATUS_LOCAL_ERROR:
+ if (handle_local_health)
+ lnet_handle_local_failure(ni);
+ return -1;
+ case LNET_MSG_STATUS_REMOTE_DROPPED:
+ if (handle_remote_health)
+ lnet_handle_remote_failure(lpni);
+ if (attempt_remote_resend)
+ return lnet_attempt_msg_resend(msg);
+ break;
+ case LNET_MSG_STATUS_REMOTE_ERROR:
+ case LNET_MSG_STATUS_REMOTE_TIMEOUT:
+ if (handle_remote_health)
+ lnet_handle_remote_failure(lpni);
+ return -1;
+ case LNET_MSG_STATUS_NETWORK_TIMEOUT:
+ if (handle_remote_health)
+ lnet_handle_remote_failure(lpni);
+ if (handle_local_health)
+ lnet_handle_local_failure(ni);
+ return -1;
+ default:
+ LBUG();
+ }
+
+ /* no resend is needed */
+ return -1;
+}
+
+static void
+lnet_msg_detach_md(struct lnet_msg *msg, int status)
+{
+ struct lnet_libmd *md = msg->msg_md;
+ lnet_handler_t handler = NULL;
+ int cpt = lnet_cpt_of_cookie(md->md_lh.lh_cookie);
+ int unlink;
+
+ lnet_res_lock(cpt);
+ while (md->md_flags & LNET_MD_FLAG_HANDLING)
+ /* An event handler is running - wait for it to
+ * complete to avoid races.
+ */
+ lnet_md_wait_handling(md, cpt);
+
+ /* Now it's safe to drop my caller's ref */
+ md->md_refcount--;
+ LASSERT(md->md_refcount >= 0);
+
+ unlink = lnet_md_unlinkable(md);
+ if (md->md_handler) {
+ if ((md->md_flags & LNET_MD_FLAG_ABORTED) && !status) {
+ msg->msg_ev.status = -ETIMEDOUT;
+ CDEBUG(D_NET, "md 0x%p already unlinked\n", md);
+ } else {
+ msg->msg_ev.status = status;
+ }
+ msg->msg_ev.unlinked = unlink;
+ handler = md->md_handler;
+ if (!unlink)
+ md->md_flags |= LNET_MD_FLAG_HANDLING;
+ }
+
+ if (unlink || (md->md_refcount == 0 &&
+ md->md_threshold == LNET_MD_THRESH_INF))
+ lnet_detach_rsp_tracker(md, cpt);
+
+ msg->msg_md = NULL;
+ if (unlink)
+ lnet_md_unlink(md);
+
+ lnet_res_unlock(cpt);
+
+ if (handler) {
+ handler(&msg->msg_ev);
+ if (!unlink) {
+ lnet_res_lock(cpt);
+ md->md_flags &= ~LNET_MD_FLAG_HANDLING;
+ wake_up_var(md);
+ lnet_res_unlock(cpt);
+ }
+ }
+}
+
+static bool
+lnet_is_health_check(struct lnet_msg *msg)
+{
+ bool hc = true;
+ int status = msg->msg_ev.status;
+
+ if ((!msg->msg_tx_committed && !msg->msg_rx_committed) ||
+ !msg->msg_onactivelist) {
+ CDEBUG(D_NET, "msg %p not committed for send or receive\n",
+ msg);
+ return false;
+ }
+
+ if ((msg->msg_tx_committed && !msg->msg_txpeer) ||
+ (msg->msg_rx_committed && !msg->msg_rxpeer)) {
+ /* The optimized GET case does not set msg_rxpeer, but status
+ * could be zero. Only print the error message if we have a
+ * non-zero status.
+ */
+ if (status)
+ CDEBUG(D_NET, "msg %p status %d cannot retry\n", msg,
+ status);
+ return false;
+ }
+
+ /* Check for status inconsistencies */
+ if ((!status && msg->msg_health_status != LNET_MSG_STATUS_OK) ||
+ (status && msg->msg_health_status == LNET_MSG_STATUS_OK)) {
+ CDEBUG(D_NET, "Msg %p is in inconsistent state, don't perform health "
+ "checking (%d, %d)\n", msg, status,
+ msg->msg_health_status);
+ hc = false;
+ }
+
+ CDEBUG(D_NET, "health check = %d, status = %d, hstatus = %d\n",
+ hc, status, msg->msg_health_status);
+
+ return hc;
+}
+
+char *
+lnet_health_error2str(enum lnet_msg_hstatus hstatus)
+{
+ switch (hstatus) {
+ case LNET_MSG_STATUS_LOCAL_INTERRUPT:
+ return "LOCAL_INTERRUPT";
+ case LNET_MSG_STATUS_LOCAL_DROPPED:
+ return "LOCAL_DROPPED";
+ case LNET_MSG_STATUS_LOCAL_ABORTED:
+ return "LOCAL_ABORTED";
+ case LNET_MSG_STATUS_LOCAL_NO_ROUTE:
+ return "LOCAL_NO_ROUTE";
+ case LNET_MSG_STATUS_LOCAL_TIMEOUT:
+ return "LOCAL_TIMEOUT";
+ case LNET_MSG_STATUS_LOCAL_ERROR:
+ return "LOCAL_ERROR";
+ case LNET_MSG_STATUS_REMOTE_DROPPED:
+ return "REMOTE_DROPPED";
+ case LNET_MSG_STATUS_REMOTE_ERROR:
+ return "REMOTE_ERROR";
+ case LNET_MSG_STATUS_REMOTE_TIMEOUT:
+ return "REMOTE_TIMEOUT";
+ case LNET_MSG_STATUS_NETWORK_TIMEOUT:
+ return "NETWORK_TIMEOUT";
+ case LNET_MSG_STATUS_OK:
+ return "OK";
+ default:
+ return "<UNKNOWN>";
+ }
+}
+
+bool
+lnet_send_error_simulation(struct lnet_msg *msg,
+ enum lnet_msg_hstatus *hstatus)
+{
+ if (!msg)
+ return false;
+
+ if (list_empty(&the_lnet.ln_drop_rules))
+ return false;
+
+ /* match only health rules */
+ if (!lnet_drop_rule_match(&msg->msg_hdr, LNET_NID_ANY,
+ hstatus))
+ return false;
+
+ CDEBUG(D_NET, "src %s(%s)->dst %s: %s simulate health error: %s\n",
+ libcfs_nid2str(msg->msg_hdr.src_nid),
+ libcfs_nid2str(msg->msg_txni->ni_nid),
+ libcfs_nid2str(msg->msg_hdr.dest_nid),
+ lnet_msgtyp2str(msg->msg_type),
+ lnet_health_error2str(*hstatus));
+
+ return true;
+}
+EXPORT_SYMBOL(lnet_send_error_simulation);
+