-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
* GPL HEADER START
*
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
/*
* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2013, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* Author: Eric Barton <eric@bartonsoftware.com>
*/
-#ifndef EXPORT_SYMTAB
-# define EXPORT_SYMTAB
-#endif
-#ifndef AUTOCONF_INCLUDED
-#include <linux/config.h>
-#endif
+#include <linux/version.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/stat.h>
#include <linux/errno.h>
-#include <linux/smp_lock.h>
#include <linux/unistd.h>
#include <linux/uio.h>
-#include <asm/system.h>
#include <asm/uaccess.h>
#include <asm/io.h>
#include <linux/kmod.h>
#include <linux/sysctl.h>
#include <linux/pci.h>
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,32)
+#include <linux/pci-dma.h>
+#endif
#include <net/sock.h>
#include <linux/in.h>
#include <lnet/lib-lnet.h>
#include <lnet/lnet-sysctl.h>
-#if !HAVE_GFP_T
-typedef int gfp_t;
+#ifdef HAVE_COMPAT_RDMA
+#include <linux/compat-2.6.h>
#endif
-
#include <rdma/rdma_cm.h>
#include <rdma/ib_cm.h>
#include <rdma/ib_verbs.h>
#include <rdma/ib_fmr_pool.h>
-/* tunables fixed at compile time */
-#ifdef CONFIG_SMP
-# define IBLND_N_SCHED cfs_num_online_cpus() /* # schedulers */
-#else
-# define IBLND_N_SCHED 1 /* # schedulers */
-#endif
+#define IBLND_PEER_HASH_SIZE 101 /* # peer lists */
+/* # scheduler loops before reschedule */
+#define IBLND_RESCHED 100
-#define IBLND_PEER_HASH_SIZE 101 /* # peer lists */
-#define IBLND_RESCHED 100 /* # scheduler loops before reschedule */
+#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? */
+ 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 */
+ struct ctl_table_header *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 */
+ 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;
extern kib_tunables_t kiblnd_tunables;
/************************/
/* 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)
struct kib_hca_dev;
+/* o2iblnd can run over aliased interface */
+#ifdef IFALIASZ
+#define KIB_IFNAME_SIZE IFALIASZ
+#else
+#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 */
- char ibd_ifname[32]; /* IPoIB interface name */
+ /** IPoIB interface name */
+ char ibd_ifname[KIB_IFNAME_SIZE];
int ibd_nnets; /* # nets extant */
cfs_time_t ibd_next_failover;
struct kib_pool;
struct kib_poolset;
-typedef int (*kib_ps_pool_create_t)(struct kib_poolset *ps, int inc, struct kib_pool **pp_po);
+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, cfs_list_t *node);
+typedef void (*kib_ps_node_fini_t)(struct kib_pool *po, cfs_list_t *node);
struct kib_net;
typedef struct kib_poolset
{
- cfs_spinlock_t ps_lock; /* serialize */
+ 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_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 */
typedef struct
{
- cfs_spinlock_t fps_lock; /* serialize */
+ 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 */
- int fps_increasing; /* is allocating new pool */
- cfs_time_t fps_next_retry; /* time stamp for retry if failed to allocate */
+ int fps_cpt; /* CPT id */
+ int fps_pool_size;
+ int fps_flush_trigger;
+ /* 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;
typedef struct
__u64 ibn_incarnation; /* my epoch */
int ibn_init; /* initialisation state */
int ibn_shutdown; /* shutting down? */
- unsigned int ibn_with_fmr:1; /* FMR? */
- unsigned int ibn_with_pmr:1; /* PMR? */
- cfs_atomic_t ibn_npeers; /* # peers extant */
- cfs_atomic_t ibn_nconns; /* # connections extant */
+ cfs_atomic_t ibn_npeers; /* # peers extant */
+ cfs_atomic_t ibn_nconns; /* # connections extant */
- 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 */
+ 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 */
- kib_dev_t *ibn_dev; /* underlying IB device */
+ kib_dev_t *ibn_dev; /* underlying IB device */
} kib_net_t;
+#define KIB_THREAD_SHIFT 16
+#define KIB_THREAD_ID(cpt, tid) ((cpt) << KIB_THREAD_SHIFT | (tid))
+#define KIB_THREAD_CPT(id) ((id) >> KIB_THREAD_SHIFT)
+#define KIB_THREAD_TID(id) ((id) & ((1UL << KIB_THREAD_SHIFT) - 1))
+
+struct kib_sched_info {
+ /* serialise */
+ spinlock_t ibs_lock;
+ /* schedulers sleep here */
+ wait_queue_head_t ibs_waitq;
+ /* conns to check for rx completions */
+ cfs_list_t ibs_conns;
+ /* number of scheduler threads */
+ int ibs_nthreads;
+ /* max allowed scheduler threads */
+ int ibs_nthreads_max;
+ int ibs_cpt; /* CPT id */
+};
+
typedef struct
{
- int kib_init; /* initialisation state */
- int kib_shutdown; /* shut down? */
- cfs_list_t kib_devs; /* IB devices extant */
- cfs_list_t kib_failed_devs; /* list head of failed devices */
- cfs_atomic_t kib_nthreads; /* # live threads */
- cfs_rwlock_t kib_global_lock; /* stabilize net/dev/peer/conn ops */
-
- cfs_list_t *kib_peers; /* hash table of all my known peers */
- int kib_peer_hash_size;/* size of kib_peers */
-
- void *kib_connd; /* the connd task (serialisation assertions) */
- cfs_list_t kib_connd_conns; /* connections to setup/teardown */
- cfs_list_t kib_connd_zombies;/* connections with zero refcount */
- cfs_waitq_t kib_connd_waitq; /* connection daemon sleeps here */
- cfs_spinlock_t kib_connd_lock; /* serialise */
-
- cfs_waitq_t kib_sched_waitq; /* schedulers sleep here */
- cfs_list_t kib_sched_conns; /* conns to check for rx completions */
- cfs_spinlock_t kib_sched_lock; /* serialise */
- cfs_waitq_t kib_failover_waitq; /* schedulers sleep here */
-
- struct ib_qp_attr kib_error_qpa; /* QP->ERROR */
+ int kib_init; /* initialisation state */
+ int kib_shutdown; /* shut down? */
+ cfs_list_t kib_devs; /* IB devices extant */
+ /* list head of failed devices */
+ cfs_list_t kib_failed_devs;
+ /* schedulers sleep here */
+ wait_queue_head_t kib_failover_waitq;
+ cfs_atomic_t kib_nthreads; /* # live threads */
+ /* stabilize net/dev/peer/conn ops */
+ rwlock_t kib_global_lock;
+ /* hash table of all my known peers */
+ cfs_list_t *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;
+ /* connections with zero refcount */
+ cfs_list_t kib_connd_zombies;
+ /* connection daemon sleeps here */
+ 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
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 */
+ 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 */
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 */
- unsigned long ibc_last_send; /* time of last send */
- cfs_list_t ibc_early_rxs; /* rxs completed before ESTABLISHED */
- cfs_list_t ibc_tx_noops; /* IBLND_MSG_NOOPs for IBLND_MSG_VERSION_1 */
+ unsigned int ibc_nrx:16; /* receive buffers owned */
+ unsigned int ibc_scheduled:1; /* scheduled for attention */
+ unsigned 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 */
+ spinlock_t ibc_lock; /* serialise */
kib_rx_t *ibc_rxs; /* the rx descs */
kib_pages_t *ibc_rx_pages; /* premapped rx msg pages */
kib_connvars_t *ibc_connvars; /* in-progress connection state */
} kib_conn_t;
-#define IBLND_CONN_INIT 0 /* being intialised */
+#define IBLND_CONN_INIT 0 /* being initialised */
#define IBLND_CONN_ACTIVE_CONNECT 1 /* active sending req */
#define IBLND_CONN_PASSIVE_WAIT 2 /* passive waiting for rtu */
#define IBLND_CONN_ESTABLISHED 3 /* connection established */
cfs_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), cfs_atomic_read(&(conn)->ibc_refcount)); \
+ LASSERT_ATOMIC_POS(&(conn)->ibc_refcount); \
+ if (cfs_atomic_dec_and_test(&(conn)->ibc_refcount)) { \
+ spin_lock_irqsave(&kiblnd_data.kib_connd_lock, flags); \
+ cfs_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) \
static inline int
kiblnd_send_keepalive(kib_conn_t *conn)
{
- return (*kiblnd_tunables.kib_keepalive > 0) &&
- cfs_time_after(jiffies, conn->ibc_last_send +
- *kiblnd_tunables.kib_keepalive*CFS_HZ);
+ return (*kiblnd_tunables.kib_keepalive > 0) &&
+ cfs_time_after(jiffies, conn->ibc_last_send +
+ *kiblnd_tunables.kib_keepalive*HZ);
}
static inline int
-kiblnd_send_noop(kib_conn_t *conn)
+kiblnd_need_noop(kib_conn_t *conn)
{
LASSERT (conn->ibc_state >= IBLND_CONN_ESTABLISHED);
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) || conn->ibc_credits == 0);
+ return (cfs_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) || /* can be piggybacked */
+ if (!cfs_list_empty(&conn->ibc_tx_noops) || /* NOOP already queued */
+ !cfs_list_empty(&conn->ibc_tx_queue_nocred) || /* piggyback NOOP */
conn->ibc_credits == 0) /* no credit */
return 0;
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 npages);
+int kiblnd_alloc_pages(kib_pages_t **pp, int cpt, int npages);
void kiblnd_free_pages (kib_pages_t *p);
int kiblnd_cm_callback(struct rdma_cm_id *cmid,
git://git.whamcloud.com / fs / lustre-release.git / blobdiff