X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lnet%2Finclude%2Flnet%2Flib-types.h;h=5b8b8d84e829bab6845c221d023b1f56845621f5;hp=588ea0029381fec460b5c9fc167923c4f35acdd3;hb=0d816af574b7063c0ce339b67d2066b229d20f59;hpb=76adbed805e71995d521d1a26e1e3d93f3dfd7b7 diff --git a/lnet/include/lnet/lib-types.h b/lnet/include/lnet/lib-types.h index 588ea00..5b8b8d8 100644 --- a/lnet/include/lnet/lib-types.h +++ b/lnet/include/lnet/lib-types.h @@ -23,7 +23,7 @@ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2012, 2016, Intel Corporation. + * Copyright (c) 2012, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -44,36 +44,88 @@ #include #include +#include #include -#include +#include +#include +#include /* Max payload size */ -#ifndef CONFIG_LNET_MAX_PAYLOAD -# error "CONFIG_LNET_MAX_PAYLOAD must be defined in config.h" -#endif +#define LNET_MAX_PAYLOAD LNET_MTU -#define LNET_MAX_PAYLOAD CONFIG_LNET_MAX_PAYLOAD -#if (LNET_MAX_PAYLOAD < LNET_MTU) -# error "LNET_MAX_PAYLOAD too small - error in configure --with-max-payload-mb" -#elif (LNET_MAX_PAYLOAD > (PAGE_SIZE * LNET_MAX_IOV)) -# error "LNET_MAX_PAYLOAD too large - error in configure --with-max-payload-mb" -#endif +#define LNET_MAX_IOV (LNET_MAX_PAYLOAD >> PAGE_SHIFT) + +/* + * This is the maximum health value. + * All local and peer NIs created have their health default to this value. + */ +#define LNET_MAX_HEALTH_VALUE 1000 /* forward refs */ struct lnet_libmd; -typedef struct lnet_msg { +enum lnet_msg_hstatus { + LNET_MSG_STATUS_OK = 0, + LNET_MSG_STATUS_LOCAL_INTERRUPT, + LNET_MSG_STATUS_LOCAL_DROPPED, + LNET_MSG_STATUS_LOCAL_ABORTED, + LNET_MSG_STATUS_LOCAL_NO_ROUTE, + LNET_MSG_STATUS_LOCAL_ERROR, + LNET_MSG_STATUS_LOCAL_TIMEOUT, + LNET_MSG_STATUS_REMOTE_ERROR, + LNET_MSG_STATUS_REMOTE_DROPPED, + LNET_MSG_STATUS_REMOTE_TIMEOUT, + LNET_MSG_STATUS_NETWORK_TIMEOUT, + LNET_MSG_STATUS_END, +}; + +struct lnet_rsp_tracker { + /* chain on the waiting list */ + struct list_head rspt_on_list; + /* cpt to lock */ + int rspt_cpt; + /* nid of next hop */ + lnet_nid_t rspt_next_hop_nid; + /* deadline of the REPLY/ACK */ + ktime_t rspt_deadline; + /* parent MD */ + struct lnet_handle_md rspt_mdh; +}; + +struct lnet_msg { struct list_head msg_activelist; struct list_head msg_list; /* Q for credits/MD */ - lnet_process_id_t msg_target; + struct lnet_process_id msg_target; /* Primary NID of the source. */ lnet_nid_t msg_initiator; /* where is it from, it's only for building event */ lnet_nid_t msg_from; __u32 msg_type; + /* + * hold parameters in case message is with held due + * to discovery + */ + lnet_nid_t msg_src_nid_param; + lnet_nid_t msg_rtr_nid_param; + + /* + * Deadline for the message after which it will be finalized if it + * has not completed. + */ + ktime_t msg_deadline; + + /* The message health status. */ + enum lnet_msg_hstatus msg_health_status; + /* This is a recovery message */ + bool msg_recovery; + /* the number of times a transmission has been retried */ + int msg_retry_count; + /* flag to indicate that we do not want to resend this message */ + bool msg_no_resend; + /* committed for sending */ unsigned int msg_tx_committed:1; /* CPT # this message committed for sending */ @@ -118,73 +170,74 @@ typedef struct lnet_msg { struct kvec *msg_iov; lnet_kiov_t *msg_kiov; - lnet_event_t msg_ev; - lnet_hdr_t msg_hdr; -} lnet_msg_t; - + struct lnet_event msg_ev; + struct lnet_hdr msg_hdr; +}; -typedef struct lnet_libhandle { +struct lnet_libhandle { struct list_head lh_hash_chain; __u64 lh_cookie; -} lnet_libhandle_t; +}; #define lh_entry(ptr, type, member) \ ((type *)((char *)(ptr)-(char *)(&((type *)0)->member))) -typedef struct lnet_eq { +struct lnet_eq { struct list_head eq_list; - lnet_libhandle_t eq_lh; - lnet_seq_t eq_enq_seq; - lnet_seq_t eq_deq_seq; + struct lnet_libhandle eq_lh; + unsigned long eq_enq_seq; + unsigned long eq_deq_seq; unsigned int eq_size; lnet_eq_handler_t eq_callback; - lnet_event_t *eq_events; + struct lnet_event *eq_events; int **eq_refs; /* percpt refcount for EQ */ -} lnet_eq_t; +}; -typedef struct lnet_me { +struct lnet_me { struct list_head me_list; - lnet_libhandle_t me_lh; - lnet_process_id_t me_match_id; + int me_cpt; + struct lnet_process_id me_match_id; unsigned int me_portal; unsigned int me_pos; /* hash offset in mt_hash */ __u64 me_match_bits; __u64 me_ignore_bits; - lnet_unlink_t me_unlink; + enum lnet_unlink me_unlink; struct lnet_libmd *me_md; -} lnet_me_t; - -typedef struct lnet_libmd { - struct list_head md_list; - lnet_libhandle_t md_lh; - lnet_me_t *md_me; - char *md_start; - unsigned int md_offset; - unsigned int md_length; - unsigned int md_max_size; - int md_threshold; - int md_refcount; - unsigned int md_options; - unsigned int md_flags; - unsigned int md_niov; /* # frags at end of struct */ - void *md_user_ptr; - lnet_eq_t *md_eq; +}; + +struct lnet_libmd { + struct list_head md_list; + struct lnet_libhandle md_lh; + struct lnet_me *md_me; + char *md_start; + unsigned int md_offset; + unsigned int md_length; + unsigned int md_max_size; + int md_threshold; + int md_refcount; + unsigned int md_options; + unsigned int md_flags; + unsigned int md_niov; /* # frags at end of struct */ + void *md_user_ptr; + struct lnet_rsp_tracker *md_rspt_ptr; + struct lnet_eq *md_eq; + struct lnet_handle_md md_bulk_handle; union { - struct kvec iov[LNET_MAX_IOV]; - lnet_kiov_t kiov[LNET_MAX_IOV]; + struct kvec iov[LNET_MAX_IOV]; + lnet_kiov_t kiov[LNET_MAX_IOV]; } md_iov; -} lnet_libmd_t; +}; #define LNET_MD_FLAG_ZOMBIE (1 << 0) #define LNET_MD_FLAG_AUTO_UNLINK (1 << 1) #define LNET_MD_FLAG_ABORTED (1 << 2) -typedef struct { +struct lnet_test_peer { /* info about peers we are trying to fail */ struct list_head tp_list; /* ln_test_peers */ lnet_nid_t tp_nid; /* matching nid */ unsigned int tp_threshold; /* # failures to simulate */ -} lnet_test_peer_t; +}; #define LNET_COOKIE_TYPE_MD 1 #define LNET_COOKIE_TYPE_ME 2 @@ -195,12 +248,7 @@ typedef struct { struct lnet_ni; /* forward ref */ struct socket; -typedef struct lnet_lnd -{ - /* fields managed by portals */ - struct list_head lnd_list; /* stash in the LND table */ - int lnd_refcount; /* # active instances */ - +struct lnet_lnd { /* fields initialized by the LND */ __u32 lnd_type; @@ -224,14 +272,15 @@ typedef struct lnet_lnd * and 'private' is the 'private' passed to lnet_parse(). Return * non-zero for immediate failure, otherwise complete later with * lnet_finalize() */ - int (*lnd_send)(struct lnet_ni *ni, void *private, lnet_msg_t *msg); + int (*lnd_send)(struct lnet_ni *ni, void *private, + struct lnet_msg *msg); /* Start receiving 'mlen' bytes of payload data, skipping the following * 'rlen' - 'mlen' bytes. 'private' is the 'private' passed to * lnet_parse(). Return non-zero for immedaite failure, otherwise * complete later with lnet_finalize(). This also gives back a receive * credit if the LND does flow control. */ - int (*lnd_recv)(struct lnet_ni *ni, void *private, lnet_msg_t *msg, + int (*lnd_recv)(struct lnet_ni *ni, void *private, struct lnet_msg *msg, int delayed, unsigned int niov, struct kvec *iov, lnet_kiov_t *kiov, unsigned int offset, unsigned int mlen, unsigned int rlen); @@ -242,24 +291,15 @@ typedef struct lnet_lnd * for success and do NOT give back a receive credit; that has to wait * until lnd_recv() gets called. On failure return < 0 and * release resources; lnd_recv() will not be called. */ - int (*lnd_eager_recv)(struct lnet_ni *ni, void *private, lnet_msg_t *msg, - void **new_privatep); + int (*lnd_eager_recv)(struct lnet_ni *ni, void *private, + struct lnet_msg *msg, void **new_privatep); - /* notification of peer health */ - void (*lnd_notify)(struct lnet_ni *ni, lnet_nid_t peer, int alive); - - /* query of peer aliveness */ - void (*lnd_query)(struct lnet_ni *ni, lnet_nid_t peer, cfs_time_t *when); + /* notification of peer down */ + void (*lnd_notify_peer_down)(lnet_nid_t peer); /* accept a new connection */ int (*lnd_accept)(struct lnet_ni *ni, struct socket *sock); -} lnd_t; - -typedef struct lnet_ni_status { - lnet_nid_t ns_nid; - __u32 ns_status; - __u32 ns_unused; -} WIRE_ATTR lnet_ni_status_t; +}; struct lnet_tx_queue { int tq_credits; /* # tx credits free */ @@ -280,22 +320,51 @@ enum lnet_net_state { }; enum lnet_ni_state { - /* set when NI block is allocated */ + /* initial state when NI is created */ LNET_NI_STATE_INIT = 0, - /* set when NI is started successfully */ + /* set when NI is brought up */ LNET_NI_STATE_ACTIVE, - /* set when LND notifies NI failed */ - LNET_NI_STATE_FAILED, - /* set when LND notifies NI degraded */ - LNET_NI_STATE_DEGRADED, - /* set when shuttding down NI */ - LNET_NI_STATE_DELETING + /* set when NI is being shutdown */ + LNET_NI_STATE_DELETING, +}; + +#define LNET_NI_RECOVERY_PENDING BIT(0) +#define LNET_NI_RECOVERY_FAILED BIT(1) + +enum lnet_stats_type { + LNET_STATS_TYPE_SEND = 0, + LNET_STATS_TYPE_RECV, + LNET_STATS_TYPE_DROP +}; + +struct lnet_comm_count { + atomic_t co_get_count; + atomic_t co_put_count; + atomic_t co_reply_count; + atomic_t co_ack_count; + atomic_t co_hello_count; }; struct lnet_element_stats { - atomic_t send_count; - atomic_t recv_count; - atomic_t drop_count; + struct lnet_comm_count el_send_stats; + struct lnet_comm_count el_recv_stats; + struct lnet_comm_count el_drop_stats; +}; + +struct lnet_health_local_stats { + atomic_t hlt_local_interrupt; + atomic_t hlt_local_dropped; + atomic_t hlt_local_aborted; + atomic_t hlt_local_no_route; + atomic_t hlt_local_timeout; + atomic_t hlt_local_error; +}; + +struct lnet_health_remote_stats { + atomic_t hlt_remote_dropped; + atomic_t hlt_remote_timeout; + atomic_t hlt_remote_error; + atomic_t hlt_network_timeout; }; struct lnet_net { @@ -327,7 +396,7 @@ struct lnet_net { bool net_tunables_set; /* procedural interface */ - lnd_t *net_lnd; + const struct lnet_lnd *net_lnd; /* list of NIs on this net */ struct list_head net_ni_list; @@ -338,16 +407,22 @@ struct lnet_net { /* dying LND instances */ struct list_head net_ni_zombie; - /* network state */ - enum lnet_net_state net_state; + /* when I was last alive */ + time64_t net_last_alive; + + /* protects access to net_last_alive */ + spinlock_t net_lock; }; -typedef struct lnet_ni { +struct lnet_ni { /* chain on the lnet_net structure */ struct list_head ni_netlist; - /* chain on net_ni_cpt */ - struct list_head ni_cptlist; + /* chain on the recovery queue */ + struct list_head ni_recovery; + + /* MD handle for recovery ping */ + struct lnet_handle_md ni_ping_mdh; spinlock_t ni_lock; @@ -372,18 +447,18 @@ typedef struct lnet_ni { /* percpt reference count */ int **ni_refs; - /* when I was last alive */ - long ni_last_alive; - /* pointer to parent network */ struct lnet_net *ni_net; /* my health status */ - lnet_ni_status_t *ni_status; + struct lnet_ni_status *ni_status; - /* NI FSM */ + /* NI FSM. Protected by lnet_ni_lock() */ enum lnet_ni_state ni_state; + /* Recovery state. Protected by lnet_ni_lock() */ + __u32 ni_recovery_state; + /* per NI LND tunables */ struct lnet_lnd_tunables ni_lnd_tunables; @@ -392,6 +467,7 @@ typedef struct lnet_ni { /* NI statistics */ struct lnet_element_stats ni_stats; + struct lnet_health_local_stats ni_hstats; /* physical device CPT */ int ni_dev_cpt; @@ -400,140 +476,280 @@ typedef struct lnet_ni { __u32 ni_seq; /* + * health value + * initialized to LNET_MAX_HEALTH_VALUE + * Value is decremented every time we fail to send a message over + * this NI because of a NI specific failure. + * Value is incremented if we successfully send a message. + */ + atomic_t ni_healthv; + + /* + * Set to 1 by the LND when it receives an event telling it the device + * has gone into a fatal state. Set to 0 when the LND receives an + * even telling it the device is back online. + */ + atomic_t ni_fatal_error_on; + + /* * equivalent interfaces to use * This is an array because socklnd bonding can still be configured */ - char *ni_interfaces[LNET_MAX_INTERFACES]; + char *ni_interfaces[LNET_INTERFACES_NUM]; struct net *ni_net_ns; /* original net namespace */ -} lnet_ni_t; +}; #define LNET_PROTO_PING_MATCHBITS 0x8000000000000000LL -/* NB: value of these features equal to LNET_PROTO_PING_VERSION_x - * of old LNet, so there shouldn't be any compatibility issue */ -#define LNET_PING_FEAT_INVAL (0) /* no feature */ -#define LNET_PING_FEAT_BASE (1 << 0) /* just a ping */ -#define LNET_PING_FEAT_NI_STATUS (1 << 1) /* return NI status */ -#define LNET_PING_FEAT_RTE_DISABLED (1 << 2) /* Routing enabled */ - -#define LNET_PING_FEAT_MASK (LNET_PING_FEAT_BASE | \ - LNET_PING_FEAT_NI_STATUS) - -typedef struct lnet_ping_info { - __u32 pi_magic; - __u32 pi_features; - lnet_pid_t pi_pid; - __u32 pi_nnis; - struct lnet_ni_status pi_ni[0]; -} WIRE_ATTR lnet_ping_info_t; - -/* router checker data, per router */ -#define LNET_MAX_RTR_NIS 16 -#define LNET_PINGINFO_SIZE offsetof(struct lnet_ping_info, pi_ni[LNET_MAX_RTR_NIS]) -typedef struct { - /* chain on the_lnet.ln_zombie_rcd or ln_deathrow_rcd */ - struct list_head rcd_list; - lnet_handle_md_t rcd_mdh; /* ping buffer MD */ - struct lnet_peer_ni *rcd_gateway; /* reference to gateway */ - struct lnet_ping_info *rcd_pinginfo; /* ping buffer */ -} lnet_rc_data_t; +/* + * Descriptor of a ping info buffer: keep a separate indicator of the + * size and a reference count. The type is used both as a source and + * sink of data, so we need to keep some information outside of the + * area that may be overwritten by network data. + */ +struct lnet_ping_buffer { + int pb_nnis; + atomic_t pb_refcnt; + struct lnet_ping_info pb_info; +}; + +#define LNET_PING_BUFFER_SIZE(NNIDS) \ + offsetof(struct lnet_ping_buffer, pb_info.pi_ni[NNIDS]) +#define LNET_PING_BUFFER_LONI(PBUF) ((PBUF)->pb_info.pi_ni[0].ns_nid) +#define LNET_PING_BUFFER_SEQNO(PBUF) ((PBUF)->pb_info.pi_ni[0].ns_status) + +#define LNET_PING_INFO_TO_BUFFER(PINFO) \ + container_of((PINFO), struct lnet_ping_buffer, pb_info) struct lnet_peer_ni { - /* chain on peer_net */ - struct list_head lpni_on_peer_net_list; + /* chain on lpn_peer_nis */ + struct list_head lpni_peer_nis; /* chain on remote peer list */ struct list_head lpni_on_remote_peer_ni_list; + /* chain on recovery queue */ + struct list_head lpni_recovery; /* chain on peer hash */ struct list_head lpni_hashlist; /* messages blocking for tx credits */ struct list_head lpni_txq; - /* messages blocking for router credits */ - struct list_head lpni_rtrq; - /* chain on router list */ - struct list_head lpni_rtr_list; /* pointer to peer net I'm part of */ struct lnet_peer_net *lpni_peer_net; /* statistics kept on each peer NI */ struct lnet_element_stats lpni_stats; - /* spin lock protecting credits and lpni_txq / lpni_rtrq */ + struct lnet_health_remote_stats lpni_hstats; + /* spin lock protecting credits and lpni_txq */ spinlock_t lpni_lock; /* # tx credits available */ int lpni_txcredits; /* low water mark */ int lpni_mintxcredits; + /* + * Each peer_ni in a gateway maintains its own credits. This + * allows more traffic to gateways that have multiple interfaces. + */ /* # router credits */ int lpni_rtrcredits; /* low water mark */ int lpni_minrtrcredits; /* bytes queued for sending */ long lpni_txqnob; - /* alive/dead? */ - bool lpni_alive; - /* notification outstanding? */ - bool lpni_notify; - /* outstanding notification for LND? */ - bool lpni_notifylnd; - /* some thread is handling notification */ - bool lpni_notifying; - /* SEND event outstanding from ping */ - bool lpni_ping_notsent; - /* # times router went dead<->alive. Protected with lpni_lock */ - int lpni_alive_count; - /* time of last aliveness news */ - cfs_time_t lpni_timestamp; - /* time of last ping attempt */ - cfs_time_t lpni_ping_timestamp; - /* != 0 if ping reply expected */ - cfs_time_t lpni_ping_deadline; - /* when I was last alive */ - cfs_time_t lpni_last_alive; - /* when lpni_ni was queried last time */ - cfs_time_t lpni_last_query; /* network peer is on */ struct lnet_net *lpni_net; /* peer's NID */ lnet_nid_t lpni_nid; /* # refs */ atomic_t lpni_refcount; + /* health value for the peer */ + atomic_t lpni_healthv; + /* recovery ping mdh */ + struct lnet_handle_md lpni_recovery_ping_mdh; /* CPT this peer attached on */ int lpni_cpt; - /* # refs from lnet_route_t::lr_gateway */ - int lpni_rtr_refcount; + /* state flags -- protected by lpni_lock */ + unsigned lpni_state; + /* status of the peer NI as reported by the peer */ + __u32 lpni_ns_status; /* sequence number used to round robin over peer nis within a net */ __u32 lpni_seq; /* sequence number used to round robin over gateways */ __u32 lpni_gw_seq; - /* health flag */ - bool lpni_healthy; /* returned RC ping features. Protected with lpni_lock */ unsigned int lpni_ping_feats; - /* routes on this peer */ - struct list_head lpni_routes; - /* array of preferred local nids */ - lnet_nid_t *lpni_pref_nids; + /* time last message was received from the peer */ + time64_t lpni_last_alive; + /* preferred local nids: if only one, use lpni_pref.nid */ + union lpni_pref { + lnet_nid_t nid; + lnet_nid_t *nids; + } lpni_pref; /* number of preferred NIDs in lnpi_pref_nids */ __u32 lpni_pref_nnids; - /* router checker state */ - lnet_rc_data_t *lpni_rcd; }; +/* Preferred path added due to traffic on non-MR peer_ni */ +#define LNET_PEER_NI_NON_MR_PREF (1 << 0) +/* peer is being recovered. */ +#define LNET_PEER_NI_RECOVERY_PENDING (1 << 1) +/* recovery ping failed */ +#define LNET_PEER_NI_RECOVERY_FAILED (1 << 2) +/* peer is being deleted */ +#define LNET_PEER_NI_DELETING (1 << 3) + struct lnet_peer { - /* chain on global peer list */ - struct list_head lp_on_lnet_peer_list; + /* chain on pt_peer_list */ + struct list_head lp_peer_list; /* list of peer nets */ struct list_head lp_peer_nets; + /* list of messages pending discovery*/ + struct list_head lp_dc_pendq; + + /* chain on router list */ + struct list_head lp_rtr_list; + /* primary NID of the peer */ lnet_nid_t lp_primary_nid; - /* peer is Multi-Rail enabled peer */ - bool lp_multi_rail; + /* net to perform discovery on */ + __u32 lp_disc_net_id; + + /* CPT of peer_table */ + int lp_cpt; + + /* number of NIDs on this peer */ + int lp_nnis; + + /* # refs from lnet_route::lr_gateway */ + int lp_rtr_refcount; + + /* + * peer specific health sensitivity value to decrement peer nis in + * this peer with if set to something other than 0 + */ + __u32 lp_health_sensitivity; + + /* messages blocking for router credits */ + struct list_head lp_rtrq; + + /* routes on this peer */ + struct list_head lp_routes; + + /* reference count */ + atomic_t lp_refcount; + + /* lock protecting peer state flags and lpni_rtrq */ + spinlock_t lp_lock; + + /* peer state flags */ + unsigned lp_state; + + /* buffer for data pushed by peer */ + struct lnet_ping_buffer *lp_data; + + /* MD handle for ping in progress */ + struct lnet_handle_md lp_ping_mdh; + + /* MD handle for push in progress */ + struct lnet_handle_md lp_push_mdh; + + /* number of NIDs for sizing push data */ + int lp_data_nnis; + + /* NI config sequence number of peer */ + __u32 lp_peer_seqno; + + /* Local NI config sequence number acked by peer */ + __u32 lp_node_seqno; + + /* Local NI config sequence number sent to peer */ + __u32 lp_node_seqno_sent; + + /* Ping error encountered during discovery. */ + int lp_ping_error; + + /* Push error encountered during discovery. */ + int lp_push_error; + + /* Error encountered during discovery. */ + int lp_dc_error; + + /* time it was put on the ln_dc_working queue */ + time64_t lp_last_queued; + + /* link on discovery-related lists */ + struct list_head lp_dc_list; + + /* tasks waiting on discovery of this peer */ + wait_queue_head_t lp_dc_waitq; + + /* cached peer aliveness */ + bool lp_alive; }; +/* + * The status flags in lp_state. Their semantics have chosen so that + * lp_state can be zero-initialized. + * + * A peer is marked MULTI_RAIL in two cases: it was configured using DLC + * as multi-rail aware, or the LNET_PING_FEAT_MULTI_RAIL bit was set. + * + * A peer is marked NO_DISCOVERY if the LNET_PING_FEAT_DISCOVERY bit was + * NOT set when the peer was pinged by discovery. + * + * A peer is marked ROUTER if it indicates so in the feature bit. + */ +#define LNET_PEER_MULTI_RAIL (1 << 0) /* Multi-rail aware */ +#define LNET_PEER_NO_DISCOVERY (1 << 1) /* Peer disabled discovery */ +#define LNET_PEER_ROUTER_ENABLED (1 << 2) /* router feature enabled */ + +/* + * A peer is marked CONFIGURED if it was configured by DLC. + * + * In addition, a peer is marked DISCOVERED if it has fully passed + * through Peer Discovery. + * + * When Peer Discovery is disabled, the discovery thread will mark + * peers REDISCOVER to indicate that they should be re-examined if + * discovery is (re)enabled on the node. + * + * A peer that was created as the result of inbound traffic will not + * be marked at all. + */ +#define LNET_PEER_CONFIGURED (1 << 3) /* Configured via DLC */ +#define LNET_PEER_DISCOVERED (1 << 4) /* Peer was discovered */ +#define LNET_PEER_REDISCOVER (1 << 5) /* Discovery was disabled */ +/* + * A peer is marked DISCOVERING when discovery is in progress. + * The other flags below correspond to stages of discovery. + */ +#define LNET_PEER_DISCOVERING (1 << 6) /* Discovering */ +#define LNET_PEER_DATA_PRESENT (1 << 7) /* Remote peer data present */ +#define LNET_PEER_NIDS_UPTODATE (1 << 8) /* Remote peer info uptodate */ +#define LNET_PEER_PING_SENT (1 << 9) /* Waiting for REPLY to Ping */ +#define LNET_PEER_PUSH_SENT (1 << 10) /* Waiting for ACK of Push */ +#define LNET_PEER_PING_FAILED (1 << 11) /* Ping send failure */ +#define LNET_PEER_PUSH_FAILED (1 << 12) /* Push send failure */ +/* + * A ping can be forced as a way to fix up state, or as a manual + * intervention by an admin. + * A push can be forced in circumstances that would normally not + * allow for one to happen. + */ +#define LNET_PEER_FORCE_PING (1 << 13) /* Forced Ping */ +#define LNET_PEER_FORCE_PUSH (1 << 14) /* Forced Push */ + +/* force delete even if router */ +#define LNET_PEER_RTR_NI_FORCE_DEL (1 << 15) + +/* gw undergoing alive discovery */ +#define LNET_PEER_RTR_DISCOVERY (1 << 16) +/* gw has undergone discovery (does not indicate success or failure) */ +#define LNET_PEER_RTR_DISCOVERED (1 << 17) + struct lnet_peer_net { - /* chain on peer block */ - struct list_head lpn_on_peer_list; + /* chain on lp_peer_nets */ + struct list_head lpn_peer_nets; /* list of peer_nis on this network */ struct list_head lpn_peer_nis; @@ -543,57 +759,86 @@ struct lnet_peer_net { /* Net ID */ __u32 lpn_net_id; + + /* time of last router net check attempt */ + time64_t lpn_rtrcheck_timestamp; + + /* selection sequence number */ + __u32 lpn_seq; + + /* reference count */ + atomic_t lpn_refcount; }; /* peer hash size */ #define LNET_PEER_HASH_BITS 9 #define LNET_PEER_HASH_SIZE (1 << LNET_PEER_HASH_BITS) -/* peer hash table */ +/* + * peer hash table - one per CPT + * + * protected by lnet_net_lock/EX for update + * pt_version + * pt_hash[...] + * pt_peer_list + * pt_peers + * protected by pt_zombie_lock: + * pt_zombie_list + * pt_zombies + * + * pt_zombie lock nests inside lnet_net_lock + */ struct lnet_peer_table { int pt_version; /* /proc validity stamp */ - atomic_t pt_number; /* # peers extant */ struct list_head *pt_hash; /* NID->peer hash */ - struct list_head pt_zombie_list; /* zombie peers */ - int pt_zombies; /* # zombie peers */ + struct list_head pt_peer_list; /* peers */ + int pt_peers; /* # peers */ + struct list_head pt_zombie_list; /* zombie peer_ni */ + int pt_zombies; /* # zombie peers_ni */ spinlock_t pt_zombie_lock; /* protect list and count */ }; /* peer aliveness is enabled only on routers for peers in a network where the - * lnet_ni_t::ni_peertimeout has been set to a positive value */ + * struct lnet_ni::ni_peertimeout has been set to a positive value + */ #define lnet_peer_aliveness_enabled(lp) (the_lnet.ln_routing != 0 && \ - (lp)->lpni_net->net_tunables.lct_peer_timeout > 0) + ((lp)->lpni_net) && \ + (lp)->lpni_net->net_tunables.lct_peer_timeout > 0) -typedef struct { +struct lnet_route { struct list_head lr_list; /* chain on net */ struct list_head lr_gwlist; /* chain on gateway */ - struct lnet_peer_ni *lr_gateway; /* router node */ + struct lnet_peer *lr_gateway; /* router node */ + lnet_nid_t lr_nid; /* NID used to add route */ __u32 lr_net; /* remote network number */ + __u32 lr_lnet; /* local network number */ int lr_seq; /* sequence for round-robin */ - unsigned int lr_downis; /* number of down NIs */ __u32 lr_hops; /* how far I am */ unsigned int lr_priority; /* route priority */ -} lnet_route_t; + bool lr_alive; /* cached route aliveness */ +}; #define LNET_REMOTE_NETS_HASH_DEFAULT (1U << 7) #define LNET_REMOTE_NETS_HASH_MAX (1U << 16) #define LNET_REMOTE_NETS_HASH_SIZE (1 << the_lnet.ln_remote_nets_hbits) -typedef struct { +struct lnet_remotenet { /* chain on ln_remote_nets_hash */ struct list_head lrn_list; /* routes to me */ struct list_head lrn_routes; /* my net number */ __u32 lrn_net; -} lnet_remotenet_t; +}; /** lnet message has credit and can be submitted to lnd for send/receive */ #define LNET_CREDIT_OK 0 /** lnet message is waiting for credit */ #define LNET_CREDIT_WAIT 1 +/** lnet message is waiting for discovery */ +#define LNET_DC_WAIT 2 -typedef struct { +struct lnet_rtrbufpool { /* my free buffer pool */ struct list_head rbp_bufs; /* messages blocking for a buffer */ @@ -608,17 +853,17 @@ typedef struct { int rbp_credits; /* low water mark */ int rbp_mincredits; -} lnet_rtrbufpool_t; +}; -typedef struct { +struct lnet_rtrbuf { struct list_head rb_list; /* chain on rbp_bufs */ - lnet_rtrbufpool_t *rb_pool; /* owning pool */ + struct lnet_rtrbufpool *rb_pool; /* owning pool */ lnet_kiov_t rb_kiov[0]; /* the buffer space */ -} lnet_rtrbuf_t; +}; #define LNET_PEER_HASHSIZE 503 /* prime! */ -enum { +enum lnet_match_flags { /* Didn't match anything */ LNET_MATCHMD_NONE = (1 << 0), /* Matched OK */ @@ -631,7 +876,7 @@ enum { LNET_MATCHMD_FINISH = (LNET_MATCHMD_OK | LNET_MATCHMD_DROP), }; -/* Options for lnet_portal_t::ptl_options */ +/* Options for struct lnet_portal::ptl_options */ #define LNET_PTL_LAZY (1 << 0) #define LNET_PTL_MATCH_UNIQUE (1 << 1) /* unique match, for RDMA */ #define LNET_PTL_MATCH_WILDCARD (1 << 2) /* wildcard match, request portal */ @@ -639,7 +884,7 @@ enum { /* parameter for matching operations (GET, PUT) */ struct lnet_match_info { __u64 mi_mbits; - lnet_process_id_t mi_id; + struct lnet_process_id mi_id; unsigned int mi_cpt; unsigned int mi_opc; unsigned int mi_portal; @@ -684,7 +929,7 @@ struct lnet_match_table { /* dispatch routed PUT message by hashing source NID for wildcard portals */ #define LNET_PTL_ROTOR_HASH_RT 3 -typedef struct lnet_portal { +struct lnet_portal { spinlock_t ptl_lock; unsigned int ptl_index; /* portal ID, reserved */ /* flags on this portal: lazy, unique... */ @@ -701,7 +946,7 @@ typedef struct lnet_portal { int ptl_mt_nmaps; /* array of active entries' cpu-partition-id */ int ptl_mt_maps[0]; -} lnet_portal_t; +}; #define LNET_LH_HASH_BITS 12 #define LNET_LH_HASH_SIZE (1ULL << LNET_LH_HASH_BITS) @@ -722,18 +967,31 @@ struct lnet_msg_container { int msc_nfinalizers; /* msgs waiting to complete finalizing */ struct list_head msc_finalizing; + /* msgs waiting to be resent */ + struct list_head msc_resending; struct list_head msc_active; /* active message list */ /* threads doing finalization */ void **msc_finalizers; + /* threads doing resends */ + void **msc_resenders; }; +/* Peer Discovery states */ +#define LNET_DC_STATE_SHUTDOWN 0 /* not started */ +#define LNET_DC_STATE_RUNNING 1 /* started up OK */ +#define LNET_DC_STATE_STOPPING 2 /* telling thread to stop */ + /* Router Checker states */ -#define LNET_RC_STATE_SHUTDOWN 0 /* not started */ -#define LNET_RC_STATE_RUNNING 1 /* started up OK */ -#define LNET_RC_STATE_STOPPING 2 /* telling thread to stop */ +#define LNET_MT_STATE_SHUTDOWN 0 /* not started */ +#define LNET_MT_STATE_RUNNING 1 /* started up OK */ +#define LNET_MT_STATE_STOPPING 2 /* telling thread to stop */ -typedef struct -{ +/* LNet states */ +#define LNET_STATE_SHUTDOWN 0 /* not started */ +#define LNET_STATE_RUNNING 1 /* started up OK */ +#define LNET_STATE_STOPPING 2 /* telling thread to stop */ + +struct lnet { /* CPU partition table of LNet */ struct cfs_cpt_table *ln_cpt_table; /* number of CPTs in ln_cpt_table */ @@ -745,9 +1003,7 @@ typedef struct /* # portals */ int ln_nportals; /* the vector of portals */ - lnet_portal_t **ln_portals; - /* percpt ME containers */ - struct lnet_res_container **ln_me_containers; + struct lnet_portal **ln_portals; /* percpt MD container */ struct lnet_res_container **ln_md_containers; @@ -762,10 +1018,8 @@ typedef struct struct cfs_percpt_lock *ln_net_lock; /* percpt message containers for active/finalizing/freed message */ struct lnet_msg_container **ln_msg_containers; - lnet_counters_t **ln_counters; + struct lnet_counters **ln_counters; struct lnet_peer_table **ln_peer_tables; - /* list of configured or discovered peers */ - struct list_head ln_peers; /* list of peer nis not on a local network */ struct list_head ln_remote_peer_ni_list; /* failure simulation */ @@ -778,6 +1032,10 @@ typedef struct struct lnet_ni *ln_loni; /* network zombie list */ struct list_head ln_net_zombie; + /* resend messages list */ + struct list_head ln_msg_resend; + /* spin lock to protect the msg resend list */ + spinlock_t ln_msg_resend_lock; /* remote networks with routes to them */ struct list_head *ln_remote_nets_hash; @@ -788,22 +1046,50 @@ typedef struct /* validity stamp */ __u64 ln_routers_version; /* percpt router buffer pools */ - lnet_rtrbufpool_t **ln_rtrpools; - - lnet_handle_md_t ln_ping_target_md; - lnet_handle_eq_t ln_ping_target_eq; - struct lnet_ping_info *ln_ping_info; - - /* router checker startup/shutdown state */ - int ln_rc_state; - /* router checker's event queue */ - lnet_handle_eq_t ln_rc_eqh; - /* rcd still pending on net */ - struct list_head ln_rcd_deathrow; - /* rcd ready for free */ - struct list_head ln_rcd_zombie; + struct lnet_rtrbufpool **ln_rtrpools; + + /* + * Ping target / Push source + * + * The ping target and push source share a single buffer. The + * ln_ping_target is protected against concurrent updates by + * ln_api_mutex. + */ + struct lnet_handle_md ln_ping_target_md; + struct lnet_handle_eq ln_ping_target_eq; + struct lnet_ping_buffer *ln_ping_target; + atomic_t ln_ping_target_seqno; + + /* + * Push Target + * + * ln_push_nnis contains the desired size of the push target. + * The lnet_net_lock is used to handle update races. The old + * buffer may linger a while after it has been unlinked, in + * which case the event handler cleans up. + */ + struct lnet_handle_eq ln_push_target_eq; + struct lnet_handle_md ln_push_target_md; + struct lnet_ping_buffer *ln_push_target; + int ln_push_target_nnis; + + /* discovery event queue handle */ + struct lnet_handle_eq ln_dc_eqh; + /* discovery requests */ + struct list_head ln_dc_request; + /* discovery working list */ + struct list_head ln_dc_working; + /* discovery expired list */ + struct list_head ln_dc_expired; + /* discovery thread wait queue */ + wait_queue_head_t ln_dc_waitq; + /* discovery startup/shutdown state */ + int ln_dc_state; + + /* monitor thread startup/shutdown state */ + int ln_mt_state; /* serialise startup/shutdown */ - struct semaphore ln_rc_signal; + struct semaphore ln_mt_signal; struct mutex ln_api_mutex; struct mutex ln_lnd_mutex; @@ -811,18 +1097,18 @@ typedef struct int ln_niinit_self; /* LNetNIInit/LNetNIFini counter */ int ln_refcount; - /* shutdown in progress */ - int ln_shutdown; + /* SHUTDOWN/RUNNING/STOPPING */ + int ln_state; int ln_routing; /* am I a router? */ lnet_pid_t ln_pid; /* requested pid */ /* uniquely identifies this ni in this epoch */ __u64 ln_interface_cookie; /* registered LNDs */ - struct list_head ln_lnds; + const struct lnet_lnd *ln_lnds[NUM_LNDS]; /* test protocol compatibility flags */ - int ln_testprotocompat; + unsigned long ln_testprotocompat; /* 0 - load the NIs from the mod params * 1 - do not load the NIs from the mod params @@ -831,10 +1117,41 @@ typedef struct */ bool ln_nis_from_mod_params; - /* waitq for router checker. As long as there are no routes in - * the list, the router checker will sleep on this queue. when - * routes are added the thread will wake up */ - wait_queue_head_t ln_rc_waitq; -} lnet_t; + /* + * completion for the monitor thread. The monitor thread takes care of + * checking routes, timedout messages and resending messages. + */ + struct completion ln_mt_wait_complete; + + /* per-cpt resend queues */ + struct list_head **ln_mt_resendqs; + /* local NIs to recover */ + struct list_head ln_mt_localNIRecovq; + /* local NIs to recover */ + struct list_head ln_mt_peerNIRecovq; + /* + * An array of queues for GET/PUT waiting for REPLY/ACK respectively. + * There are CPT number of queues. Since response trackers will be + * added on the fast path we can't afford to grab the exclusive + * net lock to protect these queues. The CPT will be calculated + * based on the mdh cookie. + */ + struct list_head **ln_mt_rstq; + /* + * A response tracker becomes a zombie when the associated MD is queued + * for unlink before the response tracker is detached from the MD. An + * entry on a zombie list can be freed when either the remaining + * operations on the MD complete or when LNet has shut down. + */ + struct list_head **ln_mt_zombie_rstqs; + /* recovery eq handler */ + struct lnet_handle_eq ln_mt_eqh; + + /* + * Completed when the discovery and monitor threads can enter their + * work loops + */ + struct completion ln_started; +}; #endif