X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lnet%2Fklnds%2Fo2iblnd%2Fo2iblnd.h;h=9eb45788a211c4df060ca5c448145577067a2d77;hp=ce80d761a58ee04be7dd1f3ff205226dd1f2f873;hb=c6e4c21c4f8b04abc53c1010a697eb3ada4fb315;hpb=11c2c0ec77125041e9c8143a80e7e51aede653ea diff --git a/lnet/klnds/o2iblnd/o2iblnd.h b/lnet/klnds/o2iblnd/o2iblnd.h index ce80d76..9eb4578 100644 --- a/lnet/klnds/o2iblnd/o2iblnd.h +++ b/lnet/klnds/o2iblnd/o2iblnd.h @@ -15,11 +15,7 @@ * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see - * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ @@ -27,7 +23,7 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, Whamcloud, Inc. + * Copyright (c) 2011, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -38,8 +34,18 @@ * Author: Eric Barton */ +#ifdef HAVE_COMPAT_RDMA +#include + +#ifdef LINUX_3_17_COMPAT_H +#undef NEED_KTIME_GET_REAL_NS +#endif + +#endif + #include #include +#include #include #include #include @@ -47,7 +53,6 @@ #include #include -#include #include #include @@ -63,129 +68,94 @@ #include #include -#define DEBUG_SUBSYSTEM S_LND - -#include -#include -#include -#include - #include #include #include #include -#define IBLND_PEER_HASH_SIZE 101 /* # peer lists */ +#define DEBUG_SUBSYSTEM S_LND + +#include +#include + +#define IBLND_PEER_HASH_SIZE 101 /* # peer_ni lists */ /* # scheduler loops before reschedule */ #define IBLND_RESCHED 100 #define IBLND_N_SCHED 2 #define IBLND_N_SCHED_HIGH 4 -typedef struct -{ - int *kib_dev_failover; /* HCA failover */ - unsigned int *kib_service; /* IB service number */ - int *kib_min_reconnect_interval; /* first failed connection retry... */ - int *kib_max_reconnect_interval; /* ...exponentially increasing to this */ - int *kib_cksum; /* checksum kib_msg_t? */ - int *kib_timeout; /* comms timeout (seconds) */ - int *kib_keepalive; /* keepalive timeout (seconds) */ - int *kib_ntx; /* # tx descs */ - int *kib_credits; /* # concurrent sends */ - int *kib_peertxcredits; /* # concurrent sends to 1 peer */ - int *kib_peerrtrcredits; /* # per-peer router buffer credits */ - int *kib_peercredits_hiw; /* # when eagerly to return credits */ - int *kib_peertimeout; /* seconds to consider peer dead */ - char **kib_default_ipif; /* default IPoIB interface */ - int *kib_retry_count; - int *kib_rnr_retry_count; - int *kib_concurrent_sends; /* send work queue sizing */ - int *kib_ib_mtu; /* IB MTU */ - int *kib_map_on_demand; /* map-on-demand if RD has more fragments - * than this value, 0 disable map-on-demand */ - int *kib_pmr_pool_size; /* # physical MR in pool */ - int *kib_fmr_pool_size; /* # FMRs in pool */ - int *kib_fmr_flush_trigger; /* When to trigger FMR flush */ - int *kib_fmr_cache; /* enable FMR pool cache? */ -#if defined(CONFIG_SYSCTL) && !CFS_SYSFS_MODULE_PARM - cfs_sysctl_table_header_t *kib_sysctl; /* sysctl interface */ -#endif - int *kib_require_priv_port;/* accept only privileged ports */ - int *kib_use_priv_port; /* use privileged port for active connect */ +struct kib_tunables { + int *kib_dev_failover; /* HCA failover */ + unsigned int *kib_service; /* IB service number */ + int *kib_cksum; /* checksum struct kib_msg? */ + int *kib_timeout; /* comms timeout (seconds) */ + int *kib_keepalive; /* keepalive timeout (seconds) */ + char **kib_default_ipif; /* default IPoIB interface */ + int *kib_retry_count; + int *kib_rnr_retry_count; + int *kib_ib_mtu; /* IB MTU */ + int *kib_require_priv_port;/* accept only privileged ports */ + int *kib_use_priv_port; /* use privileged port for active connect */ /* # threads on each CPT */ int *kib_nscheds; -} kib_tunables_t; + int *kib_wrq_sge; /* # sg elements per wrq */ + int *kib_use_fastreg_gaps; /* enable discontiguous fastreg fragment support */ +}; -extern kib_tunables_t kiblnd_tunables; +extern struct kib_tunables kiblnd_tunables; #define IBLND_MSG_QUEUE_SIZE_V1 8 /* V1 only : # messages/RDMAs in-flight */ #define IBLND_CREDIT_HIGHWATER_V1 7 /* V1 only : when eagerly to return credits */ -#define IBLND_CREDITS_DEFAULT 8 /* default # of peer credits */ -#define IBLND_CREDITS_MAX ((typeof(((kib_msg_t*) 0)->ibm_credits)) - 1) /* Max # of peer credits */ +#define IBLND_CREDITS_DEFAULT 8 /* default # of peer_ni credits */ +#define IBLND_CREDITS_MAX ((typeof(((struct kib_msg *) 0)->ibm_credits)) - 1) /* Max # of peer_ni credits */ -#define IBLND_MSG_QUEUE_SIZE(v) ((v) == IBLND_MSG_VERSION_1 ? \ - IBLND_MSG_QUEUE_SIZE_V1 : \ - *kiblnd_tunables.kib_peertxcredits) /* # messages/RDMAs in-flight */ -#define IBLND_CREDITS_HIGHWATER(v) ((v) == IBLND_MSG_VERSION_1 ? \ - IBLND_CREDIT_HIGHWATER_V1 : \ - *kiblnd_tunables.kib_peercredits_hiw) /* when eagerly to return credits */ +/* when eagerly to return credits */ +#define IBLND_CREDITS_HIGHWATER(t, conn) ((conn->ibc_version) == IBLND_MSG_VERSION_1 ? \ + IBLND_CREDIT_HIGHWATER_V1 : \ + min(t->lnd_peercredits_hiw, (__u32)conn->ibc_queue_depth - 1)) -#ifdef HAVE_RDMA_CREATE_ID_4ARG -#define kiblnd_rdma_create_id(cb, dev, ps, qpt) rdma_create_id(cb, dev, ps, qpt) +#ifdef HAVE_RDMA_CREATE_ID_5ARG +# define kiblnd_rdma_create_id(ns, cb, dev, ps, qpt) \ + rdma_create_id((ns) ? (ns) : &init_net, cb, dev, ps, qpt) #else -#define kiblnd_rdma_create_id(cb, dev, ps, qpt) rdma_create_id(cb, dev, ps) +# ifdef HAVE_RDMA_CREATE_ID_4ARG +# define kiblnd_rdma_create_id(ns, cb, dev, ps, qpt) \ + rdma_create_id(cb, dev, ps, qpt) +# else +# define kiblnd_rdma_create_id(ns, cb, dev, ps, qpt) \ + rdma_create_id(cb, dev, ps) +# endif #endif -static inline int -kiblnd_concurrent_sends_v1(void) -{ - if (*kiblnd_tunables.kib_concurrent_sends > IBLND_MSG_QUEUE_SIZE_V1 * 2) - return IBLND_MSG_QUEUE_SIZE_V1 * 2; - - if (*kiblnd_tunables.kib_concurrent_sends < IBLND_MSG_QUEUE_SIZE_V1 / 2) - return IBLND_MSG_QUEUE_SIZE_V1 / 2; - - return *kiblnd_tunables.kib_concurrent_sends; -} - -#define IBLND_CONCURRENT_SENDS(v) ((v) == IBLND_MSG_VERSION_1 ? \ - kiblnd_concurrent_sends_v1() : \ - *kiblnd_tunables.kib_concurrent_sends) /* 2 OOB shall suffice for 1 keepalive and 1 returning credits */ #define IBLND_OOB_CAPABLE(v) ((v) != IBLND_MSG_VERSION_1) #define IBLND_OOB_MSGS(v) (IBLND_OOB_CAPABLE(v) ? 2 : 0) #define IBLND_MSG_SIZE (4<<10) /* max size of queued messages (inc hdr) */ #define IBLND_MAX_RDMA_FRAGS LNET_MAX_IOV /* max # of fragments supported */ -#define IBLND_CFG_RDMA_FRAGS (*kiblnd_tunables.kib_map_on_demand != 0 ? \ - *kiblnd_tunables.kib_map_on_demand : \ - IBLND_MAX_RDMA_FRAGS) /* max # of fragments configured by user */ -#define IBLND_RDMA_FRAGS(v) ((v) == IBLND_MSG_VERSION_1 ? \ - IBLND_MAX_RDMA_FRAGS : IBLND_CFG_RDMA_FRAGS) /************************/ /* derived constants... */ /* Pools (shared by connections on each CPT) */ /* These pools can grow at runtime, so don't need give a very large value */ #define IBLND_TX_POOL 256 -#define IBLND_PMR_POOL 256 #define IBLND_FMR_POOL 256 #define IBLND_FMR_POOL_FLUSH 192 -/* TX messages (shared by all connections) */ -#define IBLND_TX_MSGS() (*kiblnd_tunables.kib_ntx) - /* RX messages (per connection) */ -#define IBLND_RX_MSGS(v) (IBLND_MSG_QUEUE_SIZE(v) * 2 + IBLND_OOB_MSGS(v)) -#define IBLND_RX_MSG_BYTES(v) (IBLND_RX_MSGS(v) * IBLND_MSG_SIZE) -#define IBLND_RX_MSG_PAGES(v) ((IBLND_RX_MSG_BYTES(v) + PAGE_SIZE - 1) / PAGE_SIZE) +#define IBLND_RX_MSGS(c) \ + ((c->ibc_queue_depth) * 2 + IBLND_OOB_MSGS(c->ibc_version)) +#define IBLND_RX_MSG_BYTES(c) (IBLND_RX_MSGS(c) * IBLND_MSG_SIZE) +#define IBLND_RX_MSG_PAGES(c) \ + ((IBLND_RX_MSG_BYTES(c) + PAGE_SIZE - 1) / PAGE_SIZE) /* WRs and CQEs (per connection) */ -#define IBLND_RECV_WRS(v) IBLND_RX_MSGS(v) -#define IBLND_SEND_WRS(v) ((IBLND_RDMA_FRAGS(v) + 1) * IBLND_CONCURRENT_SENDS(v)) -#define IBLND_CQ_ENTRIES(v) (IBLND_RECV_WRS(v) + IBLND_SEND_WRS(v)) +#define IBLND_RECV_WRS(c) IBLND_RX_MSGS(c) + +/* 2 = LNet msg + Transfer chain */ +#define IBLND_CQ_ENTRIES(c) (IBLND_RECV_WRS(c) + kiblnd_send_wrs(c)) struct kib_hca_dev; @@ -196,60 +166,64 @@ struct kib_hca_dev; #define KIB_IFNAME_SIZE 256 #endif -typedef struct -{ - cfs_list_t ibd_list; /* chain on kib_devs */ - cfs_list_t ibd_fail_list; /* chain on kib_failed_devs */ - __u32 ibd_ifip; /* IPoIB interface IP */ - /** IPoIB interface name */ - char ibd_ifname[KIB_IFNAME_SIZE]; - int ibd_nnets; /* # nets extant */ - - cfs_time_t ibd_next_failover; - int ibd_failed_failover; /* # failover failures */ - unsigned int ibd_failover; /* failover in progress */ - unsigned int ibd_can_failover; /* IPoIB interface is a bonding master */ - cfs_list_t ibd_nets; - struct kib_hca_dev *ibd_hdev; -} kib_dev_t; - -typedef struct kib_hca_dev -{ - struct rdma_cm_id *ibh_cmid; /* listener cmid */ - struct ib_device *ibh_ibdev; /* IB device */ - int ibh_page_shift; /* page shift of current HCA */ - int ibh_page_size; /* page size of current HCA */ - __u64 ibh_page_mask; /* page mask of current HCA */ - int ibh_mr_shift; /* bits shift of max MR size */ - __u64 ibh_mr_size; /* size of MR */ - int ibh_nmrs; /* # of global MRs */ - struct ib_mr **ibh_mrs; /* global MR */ - struct ib_pd *ibh_pd; /* PD */ - kib_dev_t *ibh_dev; /* owner */ - cfs_atomic_t ibh_ref; /* refcount */ -} kib_hca_dev_t; +enum kib_dev_caps { + IBLND_DEV_CAPS_FASTREG_ENABLED = BIT(0), + IBLND_DEV_CAPS_FASTREG_GAPS_SUPPORT = BIT(1), + IBLND_DEV_CAPS_FMR_ENABLED = BIT(2), +}; + +struct kib_dev { + struct list_head ibd_list; /* chain on kib_devs */ + struct list_head ibd_fail_list; /* chain on kib_failed_devs */ + __u32 ibd_ifip; /* IPoIB interface IP */ + /** IPoIB interface name */ + char ibd_ifname[KIB_IFNAME_SIZE]; + int ibd_nnets; /* # nets extant */ + + time64_t ibd_next_failover; + /* # failover failures */ + int ibd_failed_failover; + /* failover in progress */ + unsigned int ibd_failover; + /* IPoIB interface is a bonding master */ + unsigned int ibd_can_failover; + struct list_head ibd_nets; + struct kib_hca_dev *ibd_hdev; + enum kib_dev_caps ibd_dev_caps; +}; + +struct kib_hca_dev { + struct rdma_cm_id *ibh_cmid; /* listener cmid */ + struct ib_device *ibh_ibdev; /* IB device */ + int ibh_page_shift; /* page shift of current HCA */ + int ibh_page_size; /* page size of current HCA */ + __u64 ibh_page_mask; /* page mask of current HCA */ + __u64 ibh_mr_size; /* size of MR */ + int ibh_max_qp_wr; /* maximum work requests size */ +#ifdef HAVE_IB_GET_DMA_MR + struct ib_mr *ibh_mrs; /* global MR */ +#endif + struct ib_pd *ibh_pd; /* PD */ + u8 ibh_port; /* port number */ + struct ib_event_handler + ibh_event_handler; /* IB event handler */ + int ibh_state; /* device status */ +#define IBLND_DEV_PORT_DOWN 0 +#define IBLND_DEV_PORT_ACTIVE 1 +#define IBLND_DEV_FATAL 2 + struct kib_dev *ibh_dev; /* owner */ + atomic_t ibh_ref; /* refcount */ +}; /** # of seconds to keep pool alive */ #define IBLND_POOL_DEADLINE 300 /** # of seconds to retry if allocation failed */ #define IBLND_POOL_RETRY 1 -typedef struct -{ +struct kib_pages { int ibp_npages; /* # pages */ struct page *ibp_pages[0]; /* page array */ -} kib_pages_t; - -struct kib_pmr_pool; - -typedef struct { - cfs_list_t pmr_list; /* chain node */ - struct ib_phys_buf *pmr_ipb; /* physical buffer */ - struct ib_mr *pmr_mr; /* IB MR */ - struct kib_pmr_pool *pmr_pool; /* owner of this MR */ - __u64 pmr_iova; /* Virtual I/O address */ - int pmr_refcount; /* reference count */ -} kib_phys_mr_t; +}; struct kib_pool; struct kib_poolset; @@ -257,111 +231,157 @@ struct kib_poolset; typedef int (*kib_ps_pool_create_t)(struct kib_poolset *ps, int inc, struct kib_pool **pp_po); typedef void (*kib_ps_pool_destroy_t)(struct kib_pool *po); -typedef void (*kib_ps_node_init_t)(struct kib_pool *po, cfs_list_t *node); -typedef void (*kib_ps_node_fini_t)(struct kib_pool *po, cfs_list_t *node); +typedef void (*kib_ps_node_init_t)(struct kib_pool *po, struct list_head *node); +typedef void (*kib_ps_node_fini_t)(struct kib_pool *po, struct list_head *node); struct kib_net; #define IBLND_POOL_NAME_LEN 32 -typedef struct kib_poolset -{ - cfs_spinlock_t ps_lock; /* serialize */ - struct kib_net *ps_net; /* network it belongs to */ - char ps_name[IBLND_POOL_NAME_LEN]; /* pool set name */ - cfs_list_t ps_pool_list; /* list of pools */ - cfs_list_t ps_failed_pool_list; /* failed pool list */ - cfs_time_t ps_next_retry; /* time stamp for retry if failed to allocate */ - int ps_increasing; /* is allocating new pool */ - int ps_pool_size; /* new pool size */ - int ps_cpt; /* CPT id */ - - kib_ps_pool_create_t ps_pool_create; /* create a new pool */ - kib_ps_pool_destroy_t ps_pool_destroy; /* destroy a pool */ - kib_ps_node_init_t ps_node_init; /* initialize new allocated node */ - kib_ps_node_fini_t ps_node_fini; /* finalize node */ -} kib_poolset_t; - -typedef struct kib_pool -{ - cfs_list_t po_list; /* chain on pool list */ - cfs_list_t po_free_list; /* pre-allocated node */ - kib_poolset_t *po_owner; /* pool_set of this pool */ - cfs_time_t po_deadline; /* deadline of this pool */ - int po_allocated; /* # of elements in use */ - int po_failed; /* pool is created on failed HCA */ - int po_size; /* # of pre-allocated elements */ -} kib_pool_t; - -typedef struct { - kib_poolset_t tps_poolset; /* pool-set */ +struct kib_poolset { + /* serialize */ + spinlock_t ps_lock; + /* network it belongs to */ + struct kib_net *ps_net; + /* pool set name */ + char ps_name[IBLND_POOL_NAME_LEN]; + /* list of pools */ + struct list_head ps_pool_list; + /* failed pool list */ + struct list_head ps_failed_pool_list; + /* time stamp for retry if failed to allocate */ + time64_t ps_next_retry; + /* is allocating new pool */ + int ps_increasing; + /* new pool size */ + int ps_pool_size; + /* CPT id */ + int ps_cpt; + + /* create a new pool */ + kib_ps_pool_create_t ps_pool_create; + /* destroy a pool */ + kib_ps_pool_destroy_t ps_pool_destroy; + /* initialize new allocated node */ + kib_ps_node_init_t ps_node_init; + /* finalize node */ + kib_ps_node_fini_t ps_node_fini; +}; + +struct kib_pool { + /* chain on pool list */ + struct list_head po_list; + /* pre-allocated node */ + struct list_head po_free_list; + /* pool_set of this pool */ + struct kib_poolset *po_owner; + /* deadline of this pool */ + time64_t po_deadline; + /* # of elements in use */ + int po_allocated; + /* pool is created on failed HCA */ + int po_failed; + /* # of pre-allocated elements */ + int po_size; +}; + +struct kib_tx_poolset { + struct kib_poolset tps_poolset; /* pool-set */ __u64 tps_next_tx_cookie; /* cookie of TX */ -} kib_tx_poolset_t; +}; -typedef struct { - kib_pool_t tpo_pool; /* pool */ +struct kib_tx_pool { + struct kib_pool tpo_pool; /* pool */ struct kib_hca_dev *tpo_hdev; /* device for this pool */ struct kib_tx *tpo_tx_descs; /* all the tx descriptors */ - kib_pages_t *tpo_tx_pages; /* premapped tx msg pages */ -} kib_tx_pool_t; - -typedef struct { - kib_poolset_t pps_poolset; /* pool-set */ -} kib_pmr_poolset_t; - -typedef struct kib_pmr_pool { - struct kib_hca_dev *ppo_hdev; /* device for this pool */ - kib_pool_t ppo_pool; /* pool */ -} kib_pmr_pool_t; + struct kib_pages *tpo_tx_pages; /* premapped tx msg pages */ +}; -typedef struct -{ - cfs_spinlock_t fps_lock; /* serialize */ - struct kib_net *fps_net; /* IB network */ - cfs_list_t fps_pool_list; /* FMR pool list */ - cfs_list_t fps_failed_pool_list; /* FMR pool list */ - __u64 fps_version; /* validity stamp */ +struct kib_fmr_poolset { + spinlock_t fps_lock; /* serialize */ + struct kib_net *fps_net; /* IB network */ + struct list_head fps_pool_list; /* FMR pool list */ + struct list_head fps_failed_pool_list; /* FMR pool list */ + __u64 fps_version; /* validity stamp */ int fps_cpt; /* CPT id */ int fps_pool_size; int fps_flush_trigger; + int fps_cache; /* is allocating new pool */ int fps_increasing; /* time stamp for retry if failed to allocate */ - cfs_time_t fps_next_retry; -} kib_fmr_poolset_t; + time64_t fps_next_retry; +}; -typedef struct -{ - cfs_list_t fpo_list; /* chain on pool list */ - struct kib_hca_dev *fpo_hdev; /* device for this pool */ - kib_fmr_poolset_t *fpo_owner; /* owner of this pool */ - struct ib_fmr_pool *fpo_fmr_pool; /* IB FMR pool */ - cfs_time_t fpo_deadline; /* deadline of this pool */ - int fpo_failed; /* fmr pool is failed */ - int fpo_map_count; /* # of mapped FMR */ -} kib_fmr_pool_t; - -typedef struct { - struct ib_pool_fmr *fmr_pfmr; /* IB pool fmr */ - kib_fmr_pool_t *fmr_pool; /* pool of FMR */ -} kib_fmr_t; - -typedef struct kib_net -{ - cfs_list_t ibn_list; /* chain on kib_dev_t::ibd_nets */ - __u64 ibn_incarnation; /* my epoch */ - int ibn_init; /* initialisation state */ - int ibn_shutdown; /* shutting down? */ +#ifndef HAVE_IB_RDMA_WR +struct ib_rdma_wr { + struct ib_send_wr wr; +}; +#endif - cfs_atomic_t ibn_npeers; /* # peers extant */ - cfs_atomic_t ibn_nconns; /* # connections extant */ +struct kib_fast_reg_descriptor { /* For fast registration */ + struct list_head frd_list; + struct ib_rdma_wr frd_inv_wr; +#ifdef HAVE_IB_MAP_MR_SG + struct ib_reg_wr frd_fastreg_wr; +#else + struct ib_rdma_wr frd_fastreg_wr; + struct ib_fast_reg_page_list *frd_frpl; +#endif + struct ib_mr *frd_mr; + bool frd_valid; +}; - kib_tx_poolset_t **ibn_tx_ps; /* tx pool-set */ - kib_fmr_poolset_t **ibn_fmr_ps; /* fmr pool-set */ - kib_pmr_poolset_t **ibn_pmr_ps; /* pmr pool-set */ +struct kib_fmr_pool { + struct list_head fpo_list; /* chain on pool list */ + struct kib_hca_dev *fpo_hdev; /* device for this pool */ + struct kib_fmr_poolset *fpo_owner; /* owner of this pool */ + union { + struct { + struct ib_fmr_pool *fpo_fmr_pool; /* IB FMR pool */ + } fmr; + struct { /* For fast registration */ + struct list_head fpo_pool_list; + int fpo_pool_size; + } fast_reg; + }; + time64_t fpo_deadline; /* deadline of this pool */ + int fpo_failed; /* fmr pool is failed */ + int fpo_map_count; /* # of mapped FMR */ + bool fpo_is_fmr; /* True if FMR pools allocated */ +}; + +struct kib_fmr { + struct kib_fmr_pool *fmr_pool; /* pool of FMR */ + struct ib_pool_fmr *fmr_pfmr; /* IB pool fmr */ + struct kib_fast_reg_descriptor *fmr_frd; + u32 fmr_key; +}; + +#ifdef HAVE_ORACLE_OFED_EXTENSIONS +#define kib_fmr_pool_map(pool, pgs, n, iov) \ + ib_fmr_pool_map_phys((pool), (pgs), (n), (iov), NULL) +#else +#define kib_fmr_pool_map(pool, pgs, n, iov) \ + ib_fmr_pool_map_phys((pool), (pgs), (n), (iov)) +#endif + +struct kib_net { + /* chain on struct kib_dev::ibd_nets */ + struct list_head ibn_list; + __u64 ibn_incarnation;/* my epoch */ + int ibn_init; /* initialisation state */ + int ibn_shutdown; /* shutting down? */ + + atomic_t ibn_npeers; /* # peers extant */ + atomic_t ibn_nconns; /* # connections extant */ - kib_dev_t *ibn_dev; /* underlying IB device */ -} kib_net_t; + struct kib_tx_poolset **ibn_tx_ps; /* tx pool-set */ + struct kib_fmr_poolset **ibn_fmr_ps; /* fmr pool-set */ + + struct kib_dev *ibn_dev; /* underlying IB device */ + struct lnet_ni *ibn_ni; /* LNet interface */ +}; #define KIB_THREAD_SHIFT 16 #define KIB_THREAD_ID(cpt, tid) ((cpt) << KIB_THREAD_SHIFT | (tid)) @@ -370,11 +390,11 @@ typedef struct kib_net struct kib_sched_info { /* serialise */ - cfs_spinlock_t ibs_lock; + spinlock_t ibs_lock; /* schedulers sleep here */ - cfs_waitq_t ibs_waitq; + wait_queue_head_t ibs_waitq; /* conns to check for rx completions */ - cfs_list_t ibs_conns; + struct list_head ibs_conns; /* number of scheduler threads */ int ibs_nthreads; /* max allowed scheduler threads */ @@ -382,35 +402,43 @@ struct kib_sched_info { int ibs_cpt; /* CPT id */ }; -typedef struct -{ +struct kib_data { int kib_init; /* initialisation state */ int kib_shutdown; /* shut down? */ - cfs_list_t kib_devs; /* IB devices extant */ + struct list_head kib_devs; /* IB devices extant */ /* list head of failed devices */ - cfs_list_t kib_failed_devs; + struct list_head kib_failed_devs; /* schedulers sleep here */ - cfs_waitq_t kib_failover_waitq; - cfs_atomic_t kib_nthreads; /* # live threads */ - /* stabilize net/dev/peer/conn ops */ - cfs_rwlock_t kib_global_lock; + wait_queue_head_t kib_failover_waitq; + atomic_t kib_nthreads; /* # live threads */ + /* stabilize net/dev/peer_ni/conn ops */ + rwlock_t kib_global_lock; /* hash table of all my known peers */ - cfs_list_t *kib_peers; + struct list_head *kib_peers; /* size of kib_peers */ int kib_peer_hash_size; /* the connd task (serialisation assertions) */ void *kib_connd; /* connections to setup/teardown */ - cfs_list_t kib_connd_conns; + struct list_head kib_connd_conns; /* connections with zero refcount */ - cfs_list_t kib_connd_zombies; + struct list_head kib_connd_zombies; + /* connections to reconnect */ + struct list_head kib_reconn_list; + /* peers wait for reconnection */ + struct list_head kib_reconn_wait; + /* + * The second that peers are pulled out from \a kib_reconn_wait + * for reconnection. + */ + time64_t kib_reconn_sec; /* connection daemon sleeps here */ - cfs_waitq_t kib_connd_waitq; - cfs_spinlock_t kib_connd_lock; /* serialise */ + wait_queue_head_t kib_connd_waitq; + spinlock_t kib_connd_lock; /* serialise */ struct ib_qp_attr kib_error_qpa; /* QP->ERROR */ /* percpt data for schedulers */ struct kib_sched_info **kib_scheds; -} kib_data_t; +}; #define IBLND_INIT_NOTHING 0 #define IBLND_INIT_DATA 1 @@ -421,60 +449,51 @@ typedef struct * These are sent in sender's byte order (i.e. receiver flips). */ -typedef struct kib_connparams -{ +struct kib_connparams { __u16 ibcp_queue_depth; __u16 ibcp_max_frags; __u32 ibcp_max_msg_size; -} WIRE_ATTR kib_connparams_t; +} WIRE_ATTR; -typedef struct -{ - lnet_hdr_t ibim_hdr; /* portals header */ - char ibim_payload[0]; /* piggy-backed payload */ -} WIRE_ATTR kib_immediate_msg_t; +struct kib_immediate_msg { + struct lnet_hdr ibim_hdr; /* portals header */ + char ibim_payload[0];/* piggy-backed payload */ +} WIRE_ATTR; -typedef struct -{ +struct kib_rdma_frag { __u32 rf_nob; /* # bytes this frag */ __u64 rf_addr; /* CAVEAT EMPTOR: misaligned!! */ -} WIRE_ATTR kib_rdma_frag_t; +} WIRE_ATTR; -typedef struct -{ - __u32 rd_key; /* local/remote key */ - __u32 rd_nfrags; /* # fragments */ - kib_rdma_frag_t rd_frags[0]; /* buffer frags */ -} WIRE_ATTR kib_rdma_desc_t; +struct kib_rdma_desc { + __u32 rd_key; /* local/remote key */ + __u32 rd_nfrags; /* # fragments */ + struct kib_rdma_frag rd_frags[0]; /* buffer frags */ +} WIRE_ATTR; -typedef struct -{ - lnet_hdr_t ibprm_hdr; /* portals header */ - __u64 ibprm_cookie; /* opaque completion cookie */ -} WIRE_ATTR kib_putreq_msg_t; +struct kib_putreq_msg { + struct lnet_hdr ibprm_hdr; /* portals header */ + __u64 ibprm_cookie; /* opaque completion cookie */ +} WIRE_ATTR; -typedef struct -{ +struct kib_putack_msg { __u64 ibpam_src_cookie; /* reflected completion cookie */ __u64 ibpam_dst_cookie; /* opaque completion cookie */ - kib_rdma_desc_t ibpam_rd; /* sender's sink buffer */ -} WIRE_ATTR kib_putack_msg_t; + struct kib_rdma_desc ibpam_rd; /* sender's sink buffer */ +} WIRE_ATTR; -typedef struct -{ - lnet_hdr_t ibgm_hdr; /* portals header */ - __u64 ibgm_cookie; /* opaque completion cookie */ - kib_rdma_desc_t ibgm_rd; /* rdma descriptor */ -} WIRE_ATTR kib_get_msg_t; +struct kib_get_msg { + struct lnet_hdr ibgm_hdr; /* portals header */ + __u64 ibgm_cookie; /* opaque completion cookie */ + struct kib_rdma_desc ibgm_rd; /* rdma descriptor */ +} WIRE_ATTR; -typedef struct -{ +struct kib_completion_msg { __u64 ibcm_cookie; /* opaque completion cookie */ __s32 ibcm_status; /* < 0 failure: >= 0 length */ -} WIRE_ATTR kib_completion_msg_t; +} WIRE_ATTR; -typedef struct -{ +struct kib_msg { /* First 2 fields fixed FOR ALL TIME */ __u32 ibm_magic; /* I'm an ibnal message */ __u16 ibm_version; /* this is my version number */ @@ -489,14 +508,14 @@ typedef struct __u64 ibm_dststamp; /* destination's incarnation */ union { - kib_connparams_t connparams; - kib_immediate_msg_t immediate; - kib_putreq_msg_t putreq; - kib_putack_msg_t putack; - kib_get_msg_t get; - kib_completion_msg_t completion; + struct kib_connparams connparams; + struct kib_immediate_msg immediate; + struct kib_putreq_msg putreq; + struct kib_putack_msg putack; + struct kib_get_msg get; + struct kib_completion_msg completion; } WIRE_ATTR ibm_u; -} WIRE_ATTR kib_msg_t; +} WIRE_ATTR; #define IBLND_MSG_MAGIC LNET_PROTO_IB_MAGIC /* unique magic */ @@ -515,122 +534,191 @@ typedef struct #define IBLND_MSG_GET_REQ 0xd6 /* getreq (sink->src) */ #define IBLND_MSG_GET_DONE 0xd7 /* completion (src->sink: all OK) */ -typedef struct { +struct kib_rej { __u32 ibr_magic; /* sender's magic */ __u16 ibr_version; /* sender's version */ __u8 ibr_why; /* reject reason */ __u8 ibr_padding; /* padding */ - __u64 ibr_incarnation; /* incarnation of peer */ - kib_connparams_t ibr_cp; /* connection parameters */ -} WIRE_ATTR kib_rej_t; + __u64 ibr_incarnation; /* incarnation of peer_ni */ + struct kib_connparams ibr_cp; /* connection parameters */ +} WIRE_ATTR; /* connection rejection reasons */ #define IBLND_REJECT_CONN_RACE 1 /* You lost connection race */ #define IBLND_REJECT_NO_RESOURCES 2 /* Out of memory/conns etc */ #define IBLND_REJECT_FATAL 3 /* Anything else */ -#define IBLND_REJECT_CONN_UNCOMPAT 4 /* incompatible version peer */ -#define IBLND_REJECT_CONN_STALE 5 /* stale peer */ +#define IBLND_REJECT_CONN_UNCOMPAT 4 /* incompatible version peer_ni */ +#define IBLND_REJECT_CONN_STALE 5 /* stale peer_ni */ -#define IBLND_REJECT_RDMA_FRAGS 6 /* Fatal: peer's rdma frags can't match mine */ -#define IBLND_REJECT_MSG_QUEUE_SIZE 7 /* Fatal: peer's msg queue size can't match mine */ +/* peer_ni's rdma frags doesn't match mine */ +#define IBLND_REJECT_RDMA_FRAGS 6 +/* peer_ni's msg queue size doesn't match mine */ +#define IBLND_REJECT_MSG_QUEUE_SIZE 7 +#define IBLND_REJECT_INVALID_SRV_ID 8 /***********************************************************************/ -typedef struct kib_rx /* receive message */ -{ - cfs_list_t rx_list; /* queue for attention */ - struct kib_conn *rx_conn; /* owning conn */ - int rx_nob; /* # bytes received (-1 while posted) */ - enum ib_wc_status rx_status; /* completion status */ - kib_msg_t *rx_msg; /* message buffer (host vaddr) */ - __u64 rx_msgaddr; /* message buffer (I/O addr) */ - DECLARE_PCI_UNMAP_ADDR (rx_msgunmap); /* for dma_unmap_single() */ - struct ib_recv_wr rx_wrq; /* receive work item... */ - struct ib_sge rx_sge; /* ...and its memory */ -} kib_rx_t; +struct kib_rx { /* receive message */ + /* queue for attention */ + struct list_head rx_list; + /* owning conn */ + struct kib_conn *rx_conn; + /* # bytes received (-1 while posted) */ + int rx_nob; + /* message buffer (host vaddr) */ + struct kib_msg *rx_msg; + /* message buffer (I/O addr) */ + __u64 rx_msgaddr; + /* for dma_unmap_single() */ + DEFINE_DMA_UNMAP_ADDR(rx_msgunmap); + /* receive work item... */ + struct ib_recv_wr rx_wrq; + /* ...and its memory */ + struct ib_sge rx_sge; +}; #define IBLND_POSTRX_DONT_POST 0 /* don't post */ #define IBLND_POSTRX_NO_CREDIT 1 /* post: no credits */ -#define IBLND_POSTRX_PEER_CREDIT 2 /* post: give peer back 1 credit */ +#define IBLND_POSTRX_PEER_CREDIT 2 /* post: give peer_ni back 1 credit */ #define IBLND_POSTRX_RSRVD_CREDIT 3 /* post: give myself back 1 reserved credit */ -typedef struct kib_tx /* transmit message */ -{ - cfs_list_t tx_list; /* queue on idle_txs ibc_tx_queue etc. */ - kib_tx_pool_t *tx_pool; /* pool I'm from */ - struct kib_conn *tx_conn; /* owning conn */ - short tx_sending; /* # tx callbacks outstanding */ - short tx_queued; /* queued for sending */ - short tx_waiting; /* waiting for peer */ - int tx_status; /* LNET completion status */ - unsigned long tx_deadline; /* completion deadline */ - __u64 tx_cookie; /* completion cookie */ - lnet_msg_t *tx_lntmsg[2]; /* lnet msgs to finalize on completion */ - kib_msg_t *tx_msg; /* message buffer (host vaddr) */ - __u64 tx_msgaddr; /* message buffer (I/O addr) */ - DECLARE_PCI_UNMAP_ADDR (tx_msgunmap); /* for dma_unmap_single() */ - int tx_nwrq; /* # send work items */ - struct ib_send_wr *tx_wrq; /* send work items... */ - struct ib_sge *tx_sge; /* ...and their memory */ - kib_rdma_desc_t *tx_rd; /* rdma descriptor */ - int tx_nfrags; /* # entries in... */ - struct scatterlist *tx_frags; /* dma_map_sg descriptor */ - __u64 *tx_pages; /* rdma phys page addrs */ - union { - kib_phys_mr_t *pmr; /* MR for physical buffer */ - kib_fmr_t fmr; /* FMR */ - } tx_u; - int tx_dmadir; /* dma direction */ -} kib_tx_t; +struct kib_tx { /* transmit message */ + /* queue on idle_txs ibc_tx_queue etc. */ + struct list_head tx_list; + /* pool I'm from */ + struct kib_tx_pool *tx_pool; + /* owning conn */ + struct kib_conn *tx_conn; + /* # tx callbacks outstanding */ + short tx_sending; + /* queued for sending */ + short tx_queued; + /* waiting for peer_ni */ + short tx_waiting; + /* LNET completion status */ + int tx_status; + /* health status of the transmit */ + enum lnet_msg_hstatus tx_hstatus; + /* completion deadline */ + ktime_t tx_deadline; + /* completion cookie */ + __u64 tx_cookie; + /* lnet msgs to finalize on completion */ + struct lnet_msg *tx_lntmsg[2]; + /* message buffer (host vaddr) */ + struct kib_msg *tx_msg; + /* message buffer (I/O addr) */ + __u64 tx_msgaddr; + /* for dma_unmap_single() */ + DEFINE_DMA_UNMAP_ADDR(tx_msgunmap); + /** sge for tx_msgaddr */ + struct ib_sge tx_msgsge; + /* # send work items */ + int tx_nwrq; + /* # used scatter/gather elements */ + int tx_nsge; + /* send work items... */ + struct ib_rdma_wr *tx_wrq; + /* ...and their memory */ + struct ib_sge *tx_sge; + /* rdma descriptor */ + struct kib_rdma_desc *tx_rd; + /* # entries in... */ + int tx_nfrags; + /* dma_map_sg descriptor */ + struct scatterlist *tx_frags; + /* rdma phys page addrs */ + __u64 *tx_pages; + /* gaps in fragments */ + bool tx_gaps; + /* FMR */ + struct kib_fmr tx_fmr; + /* dma direction */ + int tx_dmadir; +}; -typedef struct kib_connvars -{ +struct kib_connvars { /* connection-in-progress variables */ - kib_msg_t cv_msg; -} kib_connvars_t; + struct kib_msg cv_msg; +}; -typedef struct kib_conn -{ - struct kib_sched_info *ibc_sched; /* scheduler information */ - struct kib_peer *ibc_peer; /* owning peer */ - kib_hca_dev_t *ibc_hdev; /* HCA bound on */ - cfs_list_t ibc_list; /* stash on peer's conn list */ - cfs_list_t ibc_sched_list; /* schedule for attention */ - __u16 ibc_version; /* version of connection */ - __u64 ibc_incarnation; /* which instance of the peer */ - cfs_atomic_t ibc_refcount; /* # users */ - int ibc_state; /* what's happening */ - int ibc_nsends_posted; /* # uncompleted sends */ - int ibc_noops_posted; /* # uncompleted NOOPs */ - int ibc_credits; /* # credits I have */ - int ibc_outstanding_credits; /* # credits to return */ - int ibc_reserved_credits;/* # ACK/DONE msg credits */ - int ibc_comms_error; /* set on comms error */ - int ibc_nrx:16; /* receive buffers owned */ - int ibc_scheduled:1; /* scheduled for attention */ - int ibc_ready:1; /* CQ callback fired */ - /* time of last send */ - unsigned long ibc_last_send; - /** link chain for kiblnd_check_conns only */ - cfs_list_t ibc_connd_list; - /** rxs completed before ESTABLISHED */ - cfs_list_t ibc_early_rxs; - /** IBLND_MSG_NOOPs for IBLND_MSG_VERSION_1 */ - cfs_list_t ibc_tx_noops; - cfs_list_t ibc_tx_queue; /* sends that need a credit */ - cfs_list_t ibc_tx_queue_nocred;/* sends that don't need a credit */ - cfs_list_t ibc_tx_queue_rsrvd; /* sends that need to reserve an ACK/DONE msg */ - cfs_list_t ibc_active_txs; /* active tx awaiting completion */ - cfs_spinlock_t ibc_lock; /* serialise */ - kib_rx_t *ibc_rxs; /* the rx descs */ - kib_pages_t *ibc_rx_pages; /* premapped rx msg pages */ - - struct rdma_cm_id *ibc_cmid; /* CM id */ - struct ib_cq *ibc_cq; /* completion queue */ - - kib_connvars_t *ibc_connvars; /* in-progress connection state */ -} kib_conn_t; +struct kib_conn { + /* scheduler information */ + struct kib_sched_info *ibc_sched; + /* owning peer_ni */ + struct kib_peer_ni *ibc_peer; + /* HCA bound on */ + struct kib_hca_dev *ibc_hdev; + /* stash on peer_ni's conn list */ + struct list_head ibc_list; + /* schedule for attention */ + struct list_head ibc_sched_list; + /* version of connection */ + __u16 ibc_version; + /* reconnect later */ + __u16 ibc_reconnect:1; + /* which instance of the peer */ + __u64 ibc_incarnation; + /* # users */ + atomic_t ibc_refcount; + /* what's happening */ + int ibc_state; + /* # uncompleted sends */ + int ibc_nsends_posted; + /* # uncompleted NOOPs */ + int ibc_noops_posted; + /* # credits I have */ + int ibc_credits; + /* # credits to return */ + int ibc_outstanding_credits; + /* # ACK/DONE msg credits */ + int ibc_reserved_credits; + /* set on comms error */ + int ibc_comms_error; + /* connections queue depth */ + __u16 ibc_queue_depth; + /* connections max frags */ + __u16 ibc_max_frags; + /* receive buffers owned */ + unsigned int ibc_nrx:16; + /* scheduled for attention */ + unsigned int ibc_scheduled:1; + /* CQ callback fired */ + unsigned int ibc_ready:1; + /* time of last send */ + ktime_t ibc_last_send; + /** link chain for kiblnd_check_conns only */ + struct list_head ibc_connd_list; + /** rxs completed before ESTABLISHED */ + struct list_head ibc_early_rxs; + /** IBLND_MSG_NOOPs for IBLND_MSG_VERSION_1 */ + struct list_head ibc_tx_noops; + /* sends that need a credit */ + struct list_head ibc_tx_queue; + /* sends that don't need a credit */ + struct list_head ibc_tx_queue_nocred; + /* sends that need to reserve an ACK/DONE msg */ + struct list_head ibc_tx_queue_rsrvd; + /* active tx awaiting completion */ + struct list_head ibc_active_txs; + /* zombie tx awaiting done */ + struct list_head ibc_zombie_txs; + /* serialise */ + spinlock_t ibc_lock; + /* the rx descs */ + struct kib_rx *ibc_rxs; + /* premapped rx msg pages */ + struct kib_pages *ibc_rx_pages; + + /* CM id */ + struct rdma_cm_id *ibc_cmid; + /* completion queue */ + struct ib_cq *ibc_cq; + + /* in-progress connection state */ + struct kib_connvars *ibc_connvars; +}; #define IBLND_CONN_INIT 0 /* being initialised */ #define IBLND_CONN_ACTIVE_CONNECT 1 /* active sending req */ @@ -639,45 +727,103 @@ typedef struct kib_conn #define IBLND_CONN_CLOSING 4 /* being closed */ #define IBLND_CONN_DISCONNECTED 5 /* disconnected */ -typedef struct kib_peer +struct kib_peer_ni { + /* stash on global peer_ni list */ + struct list_head ibp_list; + /* who's on the other end(s) */ + lnet_nid_t ibp_nid; + /* LNet interface */ + struct lnet_ni *ibp_ni; + /* all active connections */ + struct list_head ibp_conns; + /* next connection to send on for round robin */ + struct kib_conn *ibp_next_conn; + /* msgs waiting for a conn */ + struct list_head ibp_tx_queue; + /* incarnation of peer_ni */ + __u64 ibp_incarnation; + /* when (in seconds) I was last alive */ + time64_t ibp_last_alive; + /* # users */ + atomic_t ibp_refcount; + /* version of peer_ni */ + __u16 ibp_version; + /* current passive connection attempts */ + unsigned short ibp_accepting; + /* current active connection attempts */ + unsigned short ibp_connecting; + /* reconnect this peer_ni later */ + unsigned char ibp_reconnecting; + /* counter of how many times we triggered a conn race */ + unsigned char ibp_races; + /* # consecutive reconnection attempts to this peer */ + unsigned int ibp_reconnected; + /* errno on closing this peer_ni */ + int ibp_error; + /* max map_on_demand */ + __u16 ibp_max_frags; + /* max_peer_credits */ + __u16 ibp_queue_depth; +}; + +#ifndef HAVE_IB_INC_RKEY +/** + * ib_inc_rkey - increments the key portion of the given rkey. Can be used + * for calculating a new rkey for type 2 memory windows. + * @rkey - the rkey to increment. + */ +static inline u32 ib_inc_rkey(u32 rkey) +{ + const u32 mask = 0x000000ff; + return ((rkey + 1) & mask) | (rkey & ~mask); +} +#endif + +extern struct kib_data kiblnd_data; + +extern void kiblnd_hdev_destroy(struct kib_hca_dev *hdev); + +int kiblnd_msg_queue_size(int version, struct lnet_ni *ni); + +static inline int +kiblnd_concurrent_sends(int version, struct lnet_ni *ni) { - cfs_list_t ibp_list; /* stash on global peer list */ - lnet_nid_t ibp_nid; /* who's on the other end(s) */ - lnet_ni_t *ibp_ni; /* LNet interface */ - cfs_atomic_t ibp_refcount; /* # users */ - cfs_list_t ibp_conns; /* all active connections */ - cfs_list_t ibp_tx_queue; /* msgs waiting for a conn */ - __u16 ibp_version; /* version of peer */ - __u64 ibp_incarnation; /* incarnation of peer */ - int ibp_connecting; /* current active connection attempts */ - int ibp_accepting; /* current passive connection attempts */ - int ibp_error; /* errno on closing this peer */ - cfs_time_t ibp_last_alive; /* when (in jiffies) I was last alive */ -} kib_peer_t; - -extern kib_data_t kiblnd_data; - -extern void kiblnd_hdev_destroy(kib_hca_dev_t *hdev); + struct lnet_ioctl_config_o2iblnd_tunables *tunables; + int concurrent_sends; + + tunables = &ni->ni_lnd_tunables.lnd_tun_u.lnd_o2ib; + concurrent_sends = tunables->lnd_concurrent_sends; + + if (version == IBLND_MSG_VERSION_1) { + if (concurrent_sends > IBLND_MSG_QUEUE_SIZE_V1 * 2) + return IBLND_MSG_QUEUE_SIZE_V1 * 2; + + if (concurrent_sends < IBLND_MSG_QUEUE_SIZE_V1 / 2) + return IBLND_MSG_QUEUE_SIZE_V1 / 2; + } + + return concurrent_sends; +} static inline void -kiblnd_hdev_addref_locked(kib_hca_dev_t *hdev) +kiblnd_hdev_addref_locked(struct kib_hca_dev *hdev) { - LASSERT (cfs_atomic_read(&hdev->ibh_ref) > 0); - cfs_atomic_inc(&hdev->ibh_ref); + LASSERT(atomic_read(&hdev->ibh_ref) > 0); + atomic_inc(&hdev->ibh_ref); } static inline void -kiblnd_hdev_decref(kib_hca_dev_t *hdev) +kiblnd_hdev_decref(struct kib_hca_dev *hdev) { - LASSERT (cfs_atomic_read(&hdev->ibh_ref) > 0); - if (cfs_atomic_dec_and_test(&hdev->ibh_ref)) - kiblnd_hdev_destroy(hdev); + LASSERT(atomic_read(&hdev->ibh_ref) > 0); + if (atomic_dec_and_test(&hdev->ibh_ref)) + kiblnd_hdev_destroy(hdev); } static inline int -kiblnd_dev_can_failover(kib_dev_t *dev) +kiblnd_dev_can_failover(struct kib_dev *dev) { - if (!cfs_list_empty(&dev->ibd_fail_list)) /* already scheduled */ + if (!list_empty(&dev->ibd_fail_list)) /* already scheduled */ return 0; if (*kiblnd_tunables.kib_dev_failover == 0) /* disabled */ @@ -692,98 +838,127 @@ kiblnd_dev_can_failover(kib_dev_t *dev) #define kiblnd_conn_addref(conn) \ do { \ CDEBUG(D_NET, "conn[%p] (%d)++\n", \ - (conn), cfs_atomic_read(&(conn)->ibc_refcount)); \ - cfs_atomic_inc(&(conn)->ibc_refcount); \ + (conn), atomic_read(&(conn)->ibc_refcount)); \ + atomic_inc(&(conn)->ibc_refcount); \ } while (0) -#define kiblnd_conn_decref(conn) \ -do { \ - unsigned long flags; \ - \ - CDEBUG(D_NET, "conn[%p] (%d)--\n", \ - (conn), cfs_atomic_read(&(conn)->ibc_refcount)); \ - LASSERT_ATOMIC_POS(&(conn)->ibc_refcount); \ - if (cfs_atomic_dec_and_test(&(conn)->ibc_refcount)) { \ - cfs_spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags); \ - cfs_list_add_tail(&(conn)->ibc_list, \ - &kiblnd_data.kib_connd_zombies); \ - cfs_waitq_signal(&kiblnd_data.kib_connd_waitq); \ - cfs_spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);\ - } \ +#define kiblnd_conn_decref(conn) \ +do { \ + unsigned long flags; \ + \ + CDEBUG(D_NET, "conn[%p] (%d)--\n", \ + (conn), atomic_read(&(conn)->ibc_refcount)); \ + LASSERT_ATOMIC_POS(&(conn)->ibc_refcount); \ + if (atomic_dec_and_test(&(conn)->ibc_refcount)) { \ + spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags); \ + list_add_tail(&(conn)->ibc_list, \ + &kiblnd_data.kib_connd_zombies); \ + wake_up(&kiblnd_data.kib_connd_waitq); \ + spin_unlock_irqrestore(&kiblnd_data.kib_connd_lock, flags);\ + } \ } while (0) -#define kiblnd_peer_addref(peer) \ +#define kiblnd_peer_addref(peer_ni) \ do { \ - CDEBUG(D_NET, "peer[%p] -> %s (%d)++\n", \ - (peer), libcfs_nid2str((peer)->ibp_nid), \ - cfs_atomic_read (&(peer)->ibp_refcount)); \ - cfs_atomic_inc(&(peer)->ibp_refcount); \ + CDEBUG(D_NET, "peer_ni[%p] -> %s (%d)++\n", \ + (peer_ni), libcfs_nid2str((peer_ni)->ibp_nid), \ + atomic_read (&(peer_ni)->ibp_refcount)); \ + atomic_inc(&(peer_ni)->ibp_refcount); \ } while (0) -#define kiblnd_peer_decref(peer) \ +#define kiblnd_peer_decref(peer_ni) \ do { \ - CDEBUG(D_NET, "peer[%p] -> %s (%d)--\n", \ - (peer), libcfs_nid2str((peer)->ibp_nid), \ - cfs_atomic_read (&(peer)->ibp_refcount)); \ - LASSERT_ATOMIC_POS(&(peer)->ibp_refcount); \ - if (cfs_atomic_dec_and_test(&(peer)->ibp_refcount)) \ - kiblnd_destroy_peer(peer); \ + CDEBUG(D_NET, "peer_ni[%p] -> %s (%d)--\n", \ + (peer_ni), libcfs_nid2str((peer_ni)->ibp_nid), \ + atomic_read (&(peer_ni)->ibp_refcount)); \ + LASSERT_ATOMIC_POS(&(peer_ni)->ibp_refcount); \ + if (atomic_dec_and_test(&(peer_ni)->ibp_refcount)) \ + kiblnd_destroy_peer(peer_ni); \ } while (0) -static inline cfs_list_t * +static inline bool +kiblnd_peer_connecting(struct kib_peer_ni *peer_ni) +{ + return peer_ni->ibp_connecting != 0 || + peer_ni->ibp_reconnecting != 0 || + peer_ni->ibp_accepting != 0; +} + +static inline bool +kiblnd_peer_idle(struct kib_peer_ni *peer_ni) +{ + return !kiblnd_peer_connecting(peer_ni) && list_empty(&peer_ni->ibp_conns); +} + +static inline struct list_head * kiblnd_nid2peerlist (lnet_nid_t nid) { - unsigned int hash = - ((unsigned int)nid) % kiblnd_data.kib_peer_hash_size; + unsigned int hash = + ((unsigned int)nid) % kiblnd_data.kib_peer_hash_size; - return (&kiblnd_data.kib_peers [hash]); + return &kiblnd_data.kib_peers[hash]; } static inline int -kiblnd_peer_active (kib_peer_t *peer) +kiblnd_peer_active(struct kib_peer_ni *peer_ni) { - /* Am I in the peer hash table? */ - return (!cfs_list_empty(&peer->ibp_list)); + /* Am I in the peer_ni hash table? */ + return !list_empty(&peer_ni->ibp_list); } -static inline kib_conn_t * -kiblnd_get_conn_locked (kib_peer_t *peer) +static inline struct kib_conn * +kiblnd_get_conn_locked(struct kib_peer_ni *peer_ni) { - LASSERT (!cfs_list_empty(&peer->ibp_conns)); + struct list_head *next; + + LASSERT(!list_empty(&peer_ni->ibp_conns)); - /* just return the first connection */ - return cfs_list_entry(peer->ibp_conns.next, kib_conn_t, ibc_list); + /* Advance to next connection, be sure to skip the head node */ + if (!peer_ni->ibp_next_conn || + peer_ni->ibp_next_conn->ibc_list.next == &peer_ni->ibp_conns) + next = peer_ni->ibp_conns.next; + else + next = peer_ni->ibp_next_conn->ibc_list.next; + peer_ni->ibp_next_conn = list_entry(next, struct kib_conn, ibc_list); + + return peer_ni->ibp_next_conn; } static inline int -kiblnd_send_keepalive(kib_conn_t *conn) +kiblnd_send_keepalive(struct kib_conn *conn) { - return (*kiblnd_tunables.kib_keepalive > 0) && - cfs_time_after(jiffies, conn->ibc_last_send + - *kiblnd_tunables.kib_keepalive*CFS_HZ); + s64 keepalive_ns = *kiblnd_tunables.kib_keepalive * NSEC_PER_SEC; + + return (*kiblnd_tunables.kib_keepalive > 0) && + ktime_after(ktime_get(), + ktime_add_ns(conn->ibc_last_send, keepalive_ns)); } static inline int -kiblnd_need_noop(kib_conn_t *conn) +kiblnd_need_noop(struct kib_conn *conn) { - LASSERT (conn->ibc_state >= IBLND_CONN_ESTABLISHED); + struct lnet_ni *ni = conn->ibc_peer->ibp_ni; + struct lnet_ioctl_config_o2iblnd_tunables *tunables; + + LASSERT(conn->ibc_state >= IBLND_CONN_ESTABLISHED); + tunables = &ni->ni_lnd_tunables.lnd_tun_u.lnd_o2ib; if (conn->ibc_outstanding_credits < - IBLND_CREDITS_HIGHWATER(conn->ibc_version) && + IBLND_CREDITS_HIGHWATER(tunables, conn) && !kiblnd_send_keepalive(conn)) return 0; /* No need to send NOOP */ if (IBLND_OOB_CAPABLE(conn->ibc_version)) { - if (!cfs_list_empty(&conn->ibc_tx_queue_nocred)) + if (!list_empty(&conn->ibc_tx_queue_nocred)) return 0; /* NOOP can be piggybacked */ /* No tx to piggyback NOOP onto or no credit to send a tx */ - return (cfs_list_empty(&conn->ibc_tx_queue) || + return (list_empty(&conn->ibc_tx_queue) || conn->ibc_credits == 0); } - if (!cfs_list_empty(&conn->ibc_tx_noops) || /* NOOP already queued */ - !cfs_list_empty(&conn->ibc_tx_queue_nocred) || /* piggyback NOOP */ + if (!list_empty(&conn->ibc_tx_noops) || /* NOOP already queued */ + !list_empty(&conn->ibc_tx_queue_nocred) || /* piggyback NOOP */ conn->ibc_credits == 0) /* no credit */ return 0; @@ -792,42 +967,44 @@ kiblnd_need_noop(kib_conn_t *conn) return 0; /* No tx to piggyback NOOP onto or no credit to send a tx */ - return (cfs_list_empty(&conn->ibc_tx_queue) || conn->ibc_credits == 1); + return (list_empty(&conn->ibc_tx_queue) || conn->ibc_credits == 1); } static inline void -kiblnd_abort_receives(kib_conn_t *conn) +kiblnd_abort_receives(struct kib_conn *conn) { ib_modify_qp(conn->ibc_cmid->qp, &kiblnd_data.kib_error_qpa, IB_QP_STATE); } static inline const char * -kiblnd_queue2str (kib_conn_t *conn, cfs_list_t *q) +kiblnd_queue2str(struct kib_conn *conn, struct list_head *q) { - if (q == &conn->ibc_tx_queue) - return "tx_queue"; + if (q == &conn->ibc_tx_queue) + return "tx_queue"; - if (q == &conn->ibc_tx_queue_rsrvd) - return "tx_queue_rsrvd"; + if (q == &conn->ibc_tx_queue_rsrvd) + return "tx_queue_rsrvd"; - if (q == &conn->ibc_tx_queue_nocred) - return "tx_queue_nocred"; + if (q == &conn->ibc_tx_queue_nocred) + return "tx_queue_nocred"; - if (q == &conn->ibc_active_txs) - return "active_txs"; + if (q == &conn->ibc_active_txs) + return "active_txs"; - LBUG(); - return NULL; + LBUG(); + return NULL; } /* CAVEAT EMPTOR: We rely on descriptor alignment to allow us to use the * lowest bits of the work request id to stash the work item type. */ -#define IBLND_WID_TX 0 -#define IBLND_WID_RDMA 1 -#define IBLND_WID_RX 2 -#define IBLND_WID_MASK 3UL +#define IBLND_WID_INVAL 0 +#define IBLND_WID_TX 1 +#define IBLND_WID_RX 2 +#define IBLND_WID_RDMA 3 +#define IBLND_WID_MR 4 +#define IBLND_WID_MASK 7UL static inline __u64 kiblnd_ptr2wreqid (void *ptr, int type) @@ -852,21 +1029,21 @@ kiblnd_wreqid2type (__u64 wreqid) } static inline void -kiblnd_set_conn_state (kib_conn_t *conn, int state) +kiblnd_set_conn_state(struct kib_conn *conn, int state) { - conn->ibc_state = state; - cfs_mb(); + conn->ibc_state = state; + smp_mb(); } static inline void -kiblnd_init_msg (kib_msg_t *msg, int type, int body_nob) +kiblnd_init_msg(struct kib_msg *msg, int type, int body_nob) { msg->ibm_type = type; - msg->ibm_nob = offsetof(kib_msg_t, ibm_u) + body_nob; + msg->ibm_nob = offsetof(struct kib_msg, ibm_u) + body_nob; } static inline int -kiblnd_rd_size (kib_rdma_desc_t *rd) +kiblnd_rd_size(struct kib_rdma_desc *rd) { int i; int size; @@ -878,25 +1055,25 @@ kiblnd_rd_size (kib_rdma_desc_t *rd) } static inline __u64 -kiblnd_rd_frag_addr(kib_rdma_desc_t *rd, int index) +kiblnd_rd_frag_addr(struct kib_rdma_desc *rd, int index) { return rd->rd_frags[index].rf_addr; } static inline __u32 -kiblnd_rd_frag_size(kib_rdma_desc_t *rd, int index) +kiblnd_rd_frag_size(struct kib_rdma_desc *rd, int index) { return rd->rd_frags[index].rf_nob; } static inline __u32 -kiblnd_rd_frag_key(kib_rdma_desc_t *rd, int index) +kiblnd_rd_frag_key(struct kib_rdma_desc *rd, int index) { return rd->rd_key; } static inline int -kiblnd_rd_consume_frag(kib_rdma_desc_t *rd, int index, __u32 nob) +kiblnd_rd_consume_frag(struct kib_rdma_desc *rd, int index, __u32 nob) { if (nob < rd->rd_frags[index].rf_nob) { rd->rd_frags[index].rf_addr += nob; @@ -909,18 +1086,16 @@ kiblnd_rd_consume_frag(kib_rdma_desc_t *rd, int index, __u32 nob) } static inline int -kiblnd_rd_msg_size(kib_rdma_desc_t *rd, int msgtype, int n) +kiblnd_rd_msg_size(struct kib_rdma_desc *rd, int msgtype, int n) { LASSERT (msgtype == IBLND_MSG_GET_REQ || msgtype == IBLND_MSG_PUT_ACK); return msgtype == IBLND_MSG_GET_REQ ? - offsetof(kib_get_msg_t, ibgm_rd.rd_frags[n]) : - offsetof(kib_putack_msg_t, ibpam_rd.rd_frags[n]); + offsetof(struct kib_get_msg, ibgm_rd.rd_frags[n]) : + offsetof(struct kib_putack_msg, ibpam_rd.rd_frags[n]); } -#ifdef HAVE_OFED_IB_DMA_MAP - static inline __u64 kiblnd_dma_mapping_error(struct ib_device *dev, u64 dma_addr) { @@ -958,6 +1133,12 @@ static inline void kiblnd_dma_unmap_sg(struct ib_device *dev, ib_dma_unmap_sg(dev, sg, nents, direction); } +#ifndef HAVE_IB_SG_DMA_ADDRESS +#include +#define ib_sg_dma_address(dev, sg) sg_dma_address(sg) +#define ib_sg_dma_len(dev, sg) sg_dma_len(sg) +#endif + static inline __u64 kiblnd_sg_dma_address(struct ib_device *dev, struct scatterlist *sg) { @@ -977,145 +1158,66 @@ static inline unsigned int kiblnd_sg_dma_len(struct ib_device *dev, #define KIBLND_CONN_PARAM(e) ((e)->param.conn.private_data) #define KIBLND_CONN_PARAM_LEN(e) ((e)->param.conn.private_data_len) -#else - -static inline __u64 -kiblnd_dma_mapping_error(struct ib_device *dev, dma_addr_t dma_addr) -{ - return dma_mapping_error(dma_addr); -} - -static inline dma_addr_t kiblnd_dma_map_single(struct ib_device *dev, - void *msg, size_t size, - enum dma_data_direction direction) -{ - return dma_map_single(dev->dma_device, msg, size, direction); -} - -static inline void kiblnd_dma_unmap_single(struct ib_device *dev, - dma_addr_t addr, size_t size, - enum dma_data_direction direction) -{ - dma_unmap_single(dev->dma_device, addr, size, direction); -} - -#define KIBLND_UNMAP_ADDR_SET(p, m, a) pci_unmap_addr_set(p, m, a) -#define KIBLND_UNMAP_ADDR(p, m, a) pci_unmap_addr(p, m) - -static inline int kiblnd_dma_map_sg(struct ib_device *dev, - struct scatterlist *sg, int nents, - enum dma_data_direction direction) -{ - return dma_map_sg(dev->dma_device, sg, nents, direction); -} - -static inline void kiblnd_dma_unmap_sg(struct ib_device *dev, - struct scatterlist *sg, int nents, - enum dma_data_direction direction) -{ - return dma_unmap_sg(dev->dma_device, sg, nents, direction); -} - - -static inline dma_addr_t kiblnd_sg_dma_address(struct ib_device *dev, - struct scatterlist *sg) -{ - return sg_dma_address(sg); -} - - -static inline unsigned int kiblnd_sg_dma_len(struct ib_device *dev, - struct scatterlist *sg) -{ - return sg_dma_len(sg); -} - -#define KIBLND_CONN_PARAM(e) ((e)->private_data) -#define KIBLND_CONN_PARAM_LEN(e) ((e)->private_data_len) - -#endif +void kiblnd_abort_txs(struct kib_conn *conn, struct list_head *txs); +void kiblnd_map_rx_descs(struct kib_conn *conn); +void kiblnd_unmap_rx_descs(struct kib_conn *conn); +void kiblnd_pool_free_node(struct kib_pool *pool, struct list_head *node); +struct list_head *kiblnd_pool_alloc_node(struct kib_poolset *ps); -struct ib_mr *kiblnd_find_rd_dma_mr(kib_hca_dev_t *hdev, - kib_rdma_desc_t *rd); -struct ib_mr *kiblnd_find_dma_mr(kib_hca_dev_t *hdev, - __u64 addr, __u64 size); -void kiblnd_map_rx_descs(kib_conn_t *conn); -void kiblnd_unmap_rx_descs(kib_conn_t *conn); -int kiblnd_map_tx(lnet_ni_t *ni, kib_tx_t *tx, - kib_rdma_desc_t *rd, int nfrags); -void kiblnd_unmap_tx(lnet_ni_t *ni, kib_tx_t *tx); -void kiblnd_pool_free_node(kib_pool_t *pool, cfs_list_t *node); -cfs_list_t *kiblnd_pool_alloc_node(kib_poolset_t *ps); - -int kiblnd_fmr_pool_map(kib_fmr_poolset_t *fps, __u64 *pages, - int npages, __u64 iov, kib_fmr_t *fmr); -void kiblnd_fmr_pool_unmap(kib_fmr_t *fmr, int status); - -int kiblnd_pmr_pool_map(kib_pmr_poolset_t *pps, kib_hca_dev_t *hdev, - kib_rdma_desc_t *rd, __u64 *iova, kib_phys_mr_t **pp_pmr); -void kiblnd_pmr_pool_unmap(kib_phys_mr_t *pmr); - -int kiblnd_startup (lnet_ni_t *ni); -void kiblnd_shutdown (lnet_ni_t *ni); -int kiblnd_ctl (lnet_ni_t *ni, unsigned int cmd, void *arg); -void kiblnd_query (struct lnet_ni *ni, lnet_nid_t nid, cfs_time_t *when); +int kiblnd_fmr_pool_map(struct kib_fmr_poolset *fps, struct kib_tx *tx, + struct kib_rdma_desc *rd, u32 nob, u64 iov, + struct kib_fmr *fmr); +void kiblnd_fmr_pool_unmap(struct kib_fmr *fmr, int status); +int kiblnd_tunables_setup(struct lnet_ni *ni); int kiblnd_tunables_init(void); -void kiblnd_tunables_fini(void); int kiblnd_connd (void *arg); int kiblnd_scheduler(void *arg); -int kiblnd_thread_start (int (*fn)(void *arg), void *arg); +int kiblnd_thread_start(int (*fn)(void *arg), void *arg, char *name); int kiblnd_failover_thread (void *arg); -int kiblnd_alloc_pages(kib_pages_t **pp, int cpt, int npages); -void kiblnd_free_pages (kib_pages_t *p); +int kiblnd_alloc_pages(struct kib_pages **pp, int cpt, int npages); int kiblnd_cm_callback(struct rdma_cm_id *cmid, struct rdma_cm_event *event); int kiblnd_translate_mtu(int value); -int kiblnd_dev_failover(kib_dev_t *dev); -int kiblnd_create_peer (lnet_ni_t *ni, kib_peer_t **peerp, lnet_nid_t nid); -void kiblnd_destroy_peer (kib_peer_t *peer); -void kiblnd_destroy_dev (kib_dev_t *dev); -void kiblnd_unlink_peer_locked (kib_peer_t *peer); -void kiblnd_peer_alive (kib_peer_t *peer); -kib_peer_t *kiblnd_find_peer_locked (lnet_nid_t nid); -void kiblnd_peer_connect_failed (kib_peer_t *peer, int active, int error); -int kiblnd_close_stale_conns_locked (kib_peer_t *peer, - int version, __u64 incarnation); -int kiblnd_close_peer_conns_locked (kib_peer_t *peer, int why); - -void kiblnd_connreq_done(kib_conn_t *conn, int status); -kib_conn_t *kiblnd_create_conn (kib_peer_t *peer, struct rdma_cm_id *cmid, - int state, int version); -void kiblnd_destroy_conn (kib_conn_t *conn); -void kiblnd_close_conn (kib_conn_t *conn, int error); -void kiblnd_close_conn_locked (kib_conn_t *conn, int error); - -int kiblnd_init_rdma (kib_conn_t *conn, kib_tx_t *tx, int type, - int nob, kib_rdma_desc_t *dstrd, __u64 dstcookie); - -void kiblnd_launch_tx (lnet_ni_t *ni, kib_tx_t *tx, lnet_nid_t nid); -void kiblnd_queue_tx_locked (kib_tx_t *tx, kib_conn_t *conn); -void kiblnd_queue_tx (kib_tx_t *tx, kib_conn_t *conn); -void kiblnd_init_tx_msg (lnet_ni_t *ni, kib_tx_t *tx, int type, int body_nob); -void kiblnd_txlist_done (lnet_ni_t *ni, cfs_list_t *txlist, - int status); -void kiblnd_check_sends (kib_conn_t *conn); +int kiblnd_dev_failover(struct kib_dev *dev, struct net *ns); +int kiblnd_create_peer(struct lnet_ni *ni, struct kib_peer_ni **peerp, + lnet_nid_t nid); +void kiblnd_destroy_peer(struct kib_peer_ni *peer); +bool kiblnd_reconnect_peer(struct kib_peer_ni *peer); +void kiblnd_destroy_dev(struct kib_dev *dev); +void kiblnd_unlink_peer_locked(struct kib_peer_ni *peer_ni); +struct kib_peer_ni *kiblnd_find_peer_locked(struct lnet_ni *ni, lnet_nid_t nid); +int kiblnd_close_stale_conns_locked(struct kib_peer_ni *peer_ni, + int version, u64 incarnation); +int kiblnd_close_peer_conns_locked(struct kib_peer_ni *peer_ni, int why); + +struct kib_conn *kiblnd_create_conn(struct kib_peer_ni *peer_ni, + struct rdma_cm_id *cmid, + int state, int version); +void kiblnd_destroy_conn(struct kib_conn *conn); +void kiblnd_close_conn(struct kib_conn *conn, int error); +void kiblnd_close_conn_locked(struct kib_conn *conn, int error); + +void kiblnd_launch_tx(struct lnet_ni *ni, struct kib_tx *tx, lnet_nid_t nid); +void kiblnd_txlist_done(struct list_head *txlist, int status, + enum lnet_msg_hstatus hstatus); void kiblnd_qp_event(struct ib_event *event, void *arg); void kiblnd_cq_event(struct ib_event *event, void *arg); void kiblnd_cq_completion(struct ib_cq *cq, void *arg); -void kiblnd_pack_msg (lnet_ni_t *ni, kib_msg_t *msg, int version, - int credits, lnet_nid_t dstnid, __u64 dststamp); -int kiblnd_unpack_msg(kib_msg_t *msg, int nob); -int kiblnd_post_rx (kib_rx_t *rx, int credit); +void kiblnd_pack_msg(struct lnet_ni *ni, struct kib_msg *msg, int version, + int credits, lnet_nid_t dstnid, __u64 dststamp); +int kiblnd_unpack_msg(struct kib_msg *msg, int nob); +int kiblnd_post_rx(struct kib_rx *rx, int credit); -int kiblnd_send(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg); -int kiblnd_recv(lnet_ni_t *ni, void *private, lnet_msg_t *lntmsg, int delayed, - unsigned int niov, struct iovec *iov, lnet_kiov_t *kiov, - unsigned int offset, unsigned int mlen, unsigned int rlen); +int kiblnd_send(struct lnet_ni *ni, void *private, struct lnet_msg *lntmsg); +int kiblnd_recv(struct lnet_ni *ni, void *private, struct lnet_msg *lntmsg, + int delayed, unsigned int niov, struct kvec *iov, + lnet_kiov_t *kiov, unsigned int offset, unsigned int mlen, + unsigned int rlen);