#ifndef _LUSTRE_NET_H
#define _LUSTRE_NET_H
+/** \defgroup net net
+ *
+ * @{
+ */
+
#if defined(__linux__)
#include <linux/lustre_net.h>
#elif defined(__APPLE__)
#include <lu_object.h>
#include <lustre_req_layout.h>
+#include <obd_support.h>
+#include <lustre_ver.h>
+
/* MD flags we _always_ use */
#define PTLRPC_MD_OPTIONS 0
* considered full when less than ?_MAXREQSIZE is left in them.
*/
-#define LDLM_THREADS_AUTO_MIN \
- min((int)(num_online_cpus() * num_online_cpus() * 2), 8)
-#define LDLM_THREADS_AUTO_MAX (LDLM_THREADS_AUTO_MIN * 16)
+#define LDLM_THREADS_AUTO_MIN (2)
+#define LDLM_THREADS_AUTO_MAX min_t(unsigned, cfs_num_online_cpus() * \
+ cfs_num_online_cpus() * 32, 128)
#define LDLM_BL_THREADS LDLM_THREADS_AUTO_MIN
-#define LDLM_NBUFS (64 * num_online_cpus())
+#define LDLM_NBUFS (64 * cfs_num_online_cpus())
#define LDLM_BUFSIZE (8 * 1024)
#define LDLM_MAXREQSIZE (5 * 1024)
#define LDLM_MAXREPSIZE (1024)
#define MDT_MIN_THREADS 2UL
#define MDT_MAX_THREADS 512UL
#define MDT_NUM_THREADS max(min_t(unsigned long, MDT_MAX_THREADS, \
- num_physpages >> (25 - CFS_PAGE_SHIFT)), 2UL)
-#define FLD_NUM_THREADS max(min_t(unsigned long, MDT_MAX_THREADS, \
- num_physpages >> (25 - CFS_PAGE_SHIFT)), 2UL)
-#define SEQ_NUM_THREADS max(min_t(unsigned long, MDT_MAX_THREADS, \
- num_physpages >> (25 - CFS_PAGE_SHIFT)), 2UL)
+ cfs_num_physpages >> (25 - CFS_PAGE_SHIFT)), \
+ 2UL)
/* Absolute limits */
#define MDS_THREADS_MIN 2
#define MDS_THREADS_MAX 512
#define MDS_THREADS_MIN_READPAGE 2
-#define MDS_NBUFS (64 * num_online_cpus())
+#define MDS_NBUFS (64 * cfs_num_online_cpus())
#define MDS_BUFSIZE (8 * 1024)
/* Assume file name length = FNAME_MAX = 256 (true for ext3).
* path name length = PATH_MAX = 4096
* except in the open case where there are a large number of OSTs in a LOV.
*/
#define MDS_MAXREQSIZE (5 * 1024)
-#define MDS_MAXREPSIZE max(9 * 1024, 280 + LOV_MAX_STRIPE_COUNT * 56)
+#define MDS_MAXREPSIZE max(9 * 1024, 362 + LOV_MAX_STRIPE_COUNT * 56)
/* FLD_MAXREQSIZE == lustre_msg + __u32 padding + ptlrpc_body + opc + md_fld */
#define FLD_MAXREQSIZE (160)
/* SEQ_MAXREPSIZE == lustre_msg + ptlrpc_body + lu_range */
#define SEQ_MAXREPSIZE (152)
-#define MGS_THREADS_AUTO_MIN 2
+/* MGS threads must be >= 3, see bug 22458 comment #28 */
+#define MGS_THREADS_AUTO_MIN 3
#define MGS_THREADS_AUTO_MAX 32
-#define MGS_NBUFS (64 * num_online_cpus())
+#define MGS_NBUFS (64 * cfs_num_online_cpus())
#define MGS_BUFSIZE (8 * 1024)
#define MGS_MAXREQSIZE (7 * 1024)
#define MGS_MAXREPSIZE (9 * 1024)
/* Absolute limits */
-#define OSS_THREADS_MIN 2
+#define OSS_THREADS_MIN 3 /* difficult replies, HPQ, others */
#define OSS_THREADS_MAX 512
-#define OST_NBUFS (64 * num_online_cpus())
+#define OST_NBUFS (64 * cfs_num_online_cpus())
#define OST_BUFSIZE (8 * 1024)
/* OST_MAXREQSIZE ~= 4768 bytes =
* lustre_msg + obdo + 16 * obd_ioobj + 256 * niobuf_remote
#define ptlrpc_req_async_args(req) ((void *)&req->rq_async_args)
struct ptlrpc_connection {
- struct hlist_node c_hash;
+ cfs_hlist_node_t c_hash;
lnet_nid_t c_self;
lnet_process_id_t c_peer;
struct obd_uuid c_remote_uuid;
- atomic_t c_refcount;
+ cfs_atomic_t c_refcount;
};
struct ptlrpc_client {
* big enough. For _tons_ of context, OBD_ALLOC a struct and store
* a pointer to it here. The pointer_arg ensures this struct is at
* least big enough for that. */
- void *pointer_arg[9];
- __u64 space[4];
+ void *pointer_arg[11];
+ __u64 space[6];
};
struct ptlrpc_request_set;
typedef int (*set_interpreter_func)(struct ptlrpc_request_set *, void *, int);
struct ptlrpc_request_set {
- int set_remaining; /* # uncompleted requests */
- cfs_waitq_t set_waitq;
- cfs_waitq_t *set_wakeup_ptr;
- struct list_head set_requests;
- struct list_head set_cblist; /* list of completion callbacks */
- set_interpreter_func set_interpret; /* completion callback */
- void *set_arg; /* completion context */
+ cfs_atomic_t set_remaining; /* # uncompleted requests */
+ cfs_waitq_t set_waitq;
+ cfs_waitq_t *set_wakeup_ptr;
+ cfs_list_t set_requests;
+ cfs_list_t set_cblist; /* list of completion callbacks */
+ set_interpreter_func set_interpret; /* completion callback */
+ void *set_arg; /* completion context */
/* locked so that any old caller can communicate requests to
* the set holder who can then fold them into the lock-free set */
- spinlock_t set_new_req_lock;
- struct list_head set_new_requests;
+ cfs_spinlock_t set_new_req_lock;
+ cfs_list_t set_new_requests;
};
struct ptlrpc_set_cbdata {
- struct list_head psc_item;
+ cfs_list_t psc_item;
set_interpreter_func psc_interpret;
void *psc_data;
};
struct ptlrpc_reply_state {
struct ptlrpc_cb_id rs_cb_id;
- struct list_head rs_list;
- struct list_head rs_exp_list;
- struct list_head rs_obd_list;
+ cfs_list_t rs_list;
+ cfs_list_t rs_exp_list;
+ cfs_list_t rs_obd_list;
#if RS_DEBUG
- struct list_head rs_debug_list;
+ cfs_list_t rs_debug_list;
#endif
- /* updates to following flag serialised by srv_request_lock */
+ /* A spinlock to protect the reply state flags */
+ cfs_spinlock_t rs_lock;
+ /* Reply state flags */
unsigned long rs_difficult:1; /* ACK/commit stuff */
+ unsigned long rs_no_ack:1; /* no ACK, even for
+ difficult requests */
unsigned long rs_scheduled:1; /* being handled? */
unsigned long rs_scheduled_ever:1;/* any schedule attempts? */
unsigned long rs_handled:1; /* been handled yet? */
unsigned long rs_on_net:1; /* reply_out_callback pending? */
unsigned long rs_prealloc:1; /* rs from prealloc list */
-
+ unsigned long rs_committed:1;/* the transaction was committed
+ and the rs was dispatched
+ by ptlrpc_commit_replies */
int rs_size;
+ __u32 rs_opc;
__u64 rs_transno;
__u64 rs_xid;
struct obd_export *rs_export;
struct ptlrpc_service *rs_service;
lnet_handle_md_t rs_md_h;
- atomic_t rs_refcount;
+ cfs_atomic_t rs_refcount;
struct ptlrpc_svc_ctx *rs_svc_ctx;
struct lustre_msg *rs_repbuf; /* wrapper */
struct ptlrpc_thread;
enum rq_phase {
- RQ_PHASE_NEW = 0xebc0de00,
- RQ_PHASE_RPC = 0xebc0de01,
- RQ_PHASE_BULK = 0xebc0de02,
- RQ_PHASE_INTERPRET = 0xebc0de03,
- RQ_PHASE_COMPLETE = 0xebc0de04,
+ RQ_PHASE_NEW = 0xebc0de00,
+ RQ_PHASE_RPC = 0xebc0de01,
+ RQ_PHASE_BULK = 0xebc0de02,
+ RQ_PHASE_INTERPRET = 0xebc0de03,
+ RQ_PHASE_COMPLETE = 0xebc0de04,
+ RQ_PHASE_UNREGISTERING = 0xebc0de05,
+ RQ_PHASE_UNDEFINED = 0xebc0de06
};
/** Type of request interpreter call-back */
void *arg, int rc);
struct ptlrpc_request_pool {
- spinlock_t prp_lock;
- struct list_head prp_req_list; /* list of ptlrpc_request structs */
+ cfs_spinlock_t prp_lock;
+ cfs_list_t prp_req_list; /* list of ptlrpc_request structs */
int prp_rq_size;
void (*prp_populate)(struct ptlrpc_request_pool *, int);
};
struct lu_context;
struct lu_env;
+struct ldlm_lock;
+
+struct ptlrpc_hpreq_ops {
+ /**
+ * Check if the lock handle of the given lock is the same as
+ * taken from the request.
+ */
+ int (*hpreq_lock_match)(struct ptlrpc_request *, struct ldlm_lock *);
+ /**
+ * Check if the request is a high priority one.
+ */
+ int (*hpreq_check)(struct ptlrpc_request *);
+};
+
/**
* Represents remote procedure call.
*/
struct ptlrpc_request {
int rq_type; /* one of PTL_RPC_MSG_* */
- struct list_head rq_list;
- struct list_head rq_timed_list; /* server-side early replies */
- struct list_head rq_history_list; /* server-side history */
- __u64 rq_history_seq; /* history sequence # */
+ cfs_list_t rq_list;
+ cfs_list_t rq_timed_list; /* server-side early replies */
+ cfs_list_t rq_history_list; /* server-side history */
+ cfs_list_t rq_exp_list; /* server-side per-export list */
+ struct ptlrpc_hpreq_ops *rq_ops; /* server-side hp handlers */
+ __u64 rq_history_seq; /* history sequence # */
+ /* the index of service's srv_at_array into which request is linked */
+ time_t rq_at_index;
int rq_status;
- spinlock_t rq_lock;
+ cfs_spinlock_t rq_lock;
/* client-side flags are serialized by rq_lock */
unsigned long rq_intr:1, rq_replied:1, rq_err:1,
rq_timedout:1, rq_resend:1, rq_restart:1,
* after server commits corresponding transaction. This is
* used for operations that require sequence of multiple
* requests to be replayed. The only example currently is file
- * open/close/dw/setattr. When last request in such a sequence
- * is committed, ->rq_replay is cleared on all requests in the
+ * open/close. When last request in such a sequence is
+ * committed, ->rq_replay is cleared on all requests in the
* sequence.
*/
rq_replay:1,
- /* this is the last request in the sequence. */
- rq_sequence:1,
rq_no_resend:1, rq_waiting:1, rq_receiving_reply:1,
rq_no_delay:1, rq_net_err:1, rq_wait_ctx:1,
rq_early:1, rq_must_unlink:1,
+ rq_fake:1, /* this fake req */
/* server-side flags */
rq_packed_final:1, /* packed final reply */
- rq_sent_final:1; /* stop sending early replies */
+ rq_hp:1, /* high priority RPC */
+ rq_at_linked:1, /* link into service's srv_at_array */
+ rq_reply_truncate:1,
+ rq_committed:1,
+ /* whether the "rq_set" is a valid one */
+ rq_invalid_rqset:1;
enum rq_phase rq_phase; /* one of RQ_PHASE_* */
- atomic_t rq_refcount; /* client-side refcount for SENT race,
- server-side refcounf for multiple replies */
+ enum rq_phase rq_next_phase; /* one of RQ_PHASE_* to be used next */
+ cfs_atomic_t rq_refcount;/* client-side refcount for SENT race,
+ server-side refcounf for multiple replies */
struct ptlrpc_thread *rq_svc_thread; /* initial thread servicing req */
int rq_request_portal; /* XXX FIXME bug 249 */
int rq_reply_portal; /* XXX FIXME bug 249 */
- int rq_nob_received; /* client-side # reply bytes actually received */
-
+ int rq_nob_received; /* client-side:
+ * !rq_truncate : # reply bytes actually received,
+ * rq_truncate : required repbuf_len for resend */
int rq_reqlen;
struct lustre_msg *rq_reqmsg;
struct lustre_msg *rq_repmsg;
__u64 rq_transno;
__u64 rq_xid;
- struct list_head rq_replay_list;
- struct list_head rq_mod_list;
+ cfs_list_t rq_replay_list;
struct ptlrpc_cli_ctx *rq_cli_ctx; /* client's half ctx */
struct ptlrpc_svc_ctx *rq_svc_ctx; /* server's half ctx */
- struct list_head rq_ctx_chain; /* link to waited ctx */
+ cfs_list_t rq_ctx_chain; /* link to waited ctx */
struct sptlrpc_flavor rq_flvr; /* client & server */
enum lustre_sec_part rq_sp_from;
rq_pack_udesc:1,
rq_pack_bulk:1,
/* doesn't expect reply FIXME */
- rq_no_reply:1;
+ rq_no_reply:1,
+ rq_pill_init:1; /* pill initialized */
uid_t rq_auth_uid; /* authed uid */
uid_t rq_auth_mapped_uid; /* authed uid mapped to */
/* client+server request */
lnet_handle_md_t rq_req_md_h;
struct ptlrpc_cb_id rq_req_cbid;
+ cfs_duration_t rq_delay_limit; /* optional time limit for send attempts */
+ cfs_time_t rq_queued_time; /* time request was first queued */
/* server-side... */
struct timeval rq_arrival_time; /* request arrival time */
volatile time_t rq_deadline; /* when request must finish. volatile
so that servers' early reply updates to the deadline aren't
kept in per-cpu cache */
+ time_t rq_reply_deadline; /* when req reply unlink must finish. */
+ time_t rq_bulk_deadline; /* when req bulk unlink must finish. */
int rq_timeout; /* service time estimate (secs) */
/* Multi-rpc bits */
- struct list_head rq_set_chain;
+ cfs_list_t rq_set_chain;
+ cfs_waitq_t rq_set_waitq;
struct ptlrpc_request_set *rq_set;
/** Async completion handler */
ptlrpc_interpterer_t rq_interpret_reply;
struct ptlrpc_request_pool *rq_pool; /* Pool if request from
preallocated list */
struct lu_context rq_session;
+ struct lu_context rq_recov_session;
/* request format */
struct req_capsule rq_pill;
};
-static inline void ptlrpc_close_replay_seq(struct ptlrpc_request *req)
+static inline int ptlrpc_req_interpret(const struct lu_env *env,
+ struct ptlrpc_request *req, int rc)
{
- spin_lock(&req->rq_lock);
- req->rq_replay = 0;
- req->rq_sequence = 1;
- spin_unlock(&req->rq_lock);
+ if (req->rq_interpret_reply != NULL) {
+ req->rq_status = req->rq_interpret_reply(env, req,
+ &req->rq_async_args,
+ rc);
+ return req->rq_status;
+ }
+ return rc;
}
-static inline void lustre_set_req_swabbed(struct ptlrpc_request *req, int index)
+static inline int lustre_req_swabbed(struct ptlrpc_request *req, int index)
{
LASSERT(index < sizeof(req->rq_req_swab_mask) * 8);
- LASSERT((req->rq_req_swab_mask & (1 << index)) == 0);
- req->rq_req_swab_mask |= 1 << index;
+ return req->rq_req_swab_mask & (1 << index);
}
-static inline void lustre_set_rep_swabbed(struct ptlrpc_request *req, int index)
+static inline int lustre_rep_swabbed(struct ptlrpc_request *req, int index)
{
LASSERT(index < sizeof(req->rq_rep_swab_mask) * 8);
- LASSERT((req->rq_rep_swab_mask & (1 << index)) == 0);
- req->rq_rep_swab_mask |= 1 << index;
+ return req->rq_rep_swab_mask & (1 << index);
}
-static inline int lustre_req_swabbed(struct ptlrpc_request *req, int index)
+static inline int ptlrpc_req_need_swab(struct ptlrpc_request *req)
+{
+ return lustre_req_swabbed(req, MSG_PTLRPC_HEADER_OFF);
+}
+
+static inline int ptlrpc_rep_need_swab(struct ptlrpc_request *req)
+{
+ return lustre_rep_swabbed(req, MSG_PTLRPC_HEADER_OFF);
+}
+
+static inline void lustre_set_req_swabbed(struct ptlrpc_request *req, int index)
{
LASSERT(index < sizeof(req->rq_req_swab_mask) * 8);
- return req->rq_req_swab_mask & (1 << index);
+ LASSERT((req->rq_req_swab_mask & (1 << index)) == 0);
+ req->rq_req_swab_mask |= 1 << index;
}
-static inline int lustre_rep_swabbed(struct ptlrpc_request *req, int index)
+static inline void lustre_set_rep_swabbed(struct ptlrpc_request *req, int index)
{
LASSERT(index < sizeof(req->rq_rep_swab_mask) * 8);
- return req->rq_rep_swab_mask & (1 << index);
+ LASSERT((req->rq_rep_swab_mask & (1 << index)) == 0);
+ req->rq_rep_swab_mask |= 1 << index;
}
static inline const char *
-ptlrpc_rqphase2str(const struct ptlrpc_request *req)
+ptlrpc_phase2str(enum rq_phase phase)
{
- switch (req->rq_phase) {
+ switch (phase) {
case RQ_PHASE_NEW:
return "New";
case RQ_PHASE_RPC:
return "Interpret";
case RQ_PHASE_COMPLETE:
return "Complete";
+ case RQ_PHASE_UNREGISTERING:
+ return "Unregistering";
default:
return "?Phase?";
}
}
+static inline const char *
+ptlrpc_rqphase2str(struct ptlrpc_request *req)
+{
+ return ptlrpc_phase2str(req->rq_phase);
+}
+
/* Spare the preprocessor, spoil the bugs. */
#define FLAG(field, str) (field ? str : "")
FLAG(req->rq_restart, "T"), FLAG(req->rq_replay, "P"), \
FLAG(req->rq_no_resend, "N"), \
FLAG(req->rq_waiting, "W"), \
- FLAG(req->rq_wait_ctx, "C")
+ FLAG(req->rq_wait_ctx, "C"), FLAG(req->rq_hp, "H"), \
+ FLAG(req->rq_committed, "M")
-#define REQ_FLAGS_FMT "%s:%s%s%s%s%s%s%s%s%s%s"
+#define REQ_FLAGS_FMT "%s:%s%s%s%s%s%s%s%s%s%s%s%s"
void _debug_req(struct ptlrpc_request *req, __u32 mask,
struct libcfs_debug_msg_data *data, const char *fmt, ...)
#define debug_req(cdls, level, req, file, func, line, fmt, a...) \
do { \
- CHECK_STACK(); \
+ CFS_CHECK_STACK(); \
\
if (((level) & D_CANTMASK) != 0 || \
((libcfs_debug & (level)) != 0 && \
} while (0)
struct ptlrpc_bulk_page {
- struct list_head bp_link;
+ cfs_list_t bp_link;
int bp_buflen;
int bp_pageoffset; /* offset within a page */
struct page *bp_page;
unsigned long bd_network_rw:1; /* accessible to the network */
unsigned long bd_type:2; /* {put,get}{source,sink} */
unsigned long bd_registered:1; /* client side */
- spinlock_t bd_lock; /* serialise with callback */
+ cfs_spinlock_t bd_lock; /* serialise with callback */
int bd_import_generation;
struct obd_export *bd_export;
struct obd_import *bd_import;
lnet_handle_md_t bd_md_h; /* associated MD */
lnet_nid_t bd_sender; /* stash event::sender */
- cfs_page_t **bd_enc_pages;
#if defined(__KERNEL__)
+ /*
+ * encrypt iov, size is either 0 or bd_iov_count.
+ */
+ lnet_kiov_t *bd_enc_iov;
+
lnet_kiov_t bd_iov[0];
#else
lnet_md_iovec_t bd_iov[0];
};
struct ptlrpc_thread {
-
- struct list_head t_link; /* active threads in svc->srv_threads */
-
- void *t_data; /* thread-private data (preallocated memory) */
+ /**
+ * active threads in svc->srv_threads
+ */
+ cfs_list_t t_link;
+ /**
+ * thread-private data (preallocated memory)
+ */
+ void *t_data;
__u32 t_flags;
-
- unsigned int t_id; /* service thread index, from ptlrpc_start_threads */
+ /**
+ * service thread index, from ptlrpc_start_threads
+ */
+ unsigned int t_id;
+ /**
+ * service thread pid
+ */
+ pid_t t_pid;
+ /**
+ * put watchdog in the structure per thread b=14840
+ */
+ struct lc_watchdog *t_watchdog;
+ /**
+ * the svc this thread belonged to b=18582
+ */
+ struct ptlrpc_service *t_svc;
cfs_waitq_t t_ctl_waitq;
struct lu_env *t_env;
};
struct ptlrpc_request_buffer_desc {
- struct list_head rqbd_list;
- struct list_head rqbd_reqs;
+ cfs_list_t rqbd_list;
+ cfs_list_t rqbd_reqs;
struct ptlrpc_service *rqbd_service;
lnet_handle_md_t rqbd_md_h;
int rqbd_refcount;
typedef int (*svc_handler_t)(struct ptlrpc_request *req);
typedef void (*svcreq_printfn_t)(void *, struct ptlrpc_request *);
+typedef int (*svc_hpreq_handler_t)(struct ptlrpc_request *);
+
+#define PTLRPC_SVC_HP_RATIO 10
struct ptlrpc_service {
- struct list_head srv_list; /* chain thru all services */
+ cfs_list_t srv_list; /* chain thru all services */
int srv_max_req_size; /* biggest request to receive */
int srv_max_reply_size; /* biggest reply to send */
int srv_buf_size; /* size of individual buffers */
int srv_threads_max; /* thread upper limit */
int srv_threads_started; /* index of last started thread */
int srv_threads_running; /* # running threads */
- int srv_n_difficult_replies; /* # 'difficult' replies */
+ cfs_atomic_t srv_n_difficult_replies; /* # 'difficult' replies */
int srv_n_active_reqs; /* # reqs being served */
+ int srv_n_hpreq; /* # HPreqs being served */
cfs_duration_t srv_rqbd_timeout; /* timeout before re-posting reqs, in tick */
- int srv_watchdog_factor; /* soft watchdog timeout mutiplier */
+ int srv_watchdog_factor; /* soft watchdog timeout multiplier */
unsigned srv_cpu_affinity:1; /* bind threads to CPUs */
unsigned srv_at_check:1; /* check early replies */
+ unsigned srv_is_stopping:1; /* under unregister_service */
cfs_time_t srv_at_checktime; /* debug */
__u32 srv_req_portal;
/* AT stuff */
struct adaptive_timeout srv_at_estimate;/* estimated rpc service time */
- spinlock_t srv_at_lock;
- struct list_head srv_at_list; /* reqs waiting for replies */
- cfs_timer_t srv_at_timer; /* early reply timer */
-
- int srv_n_queued_reqs; /* # reqs in either of the queues below */
- struct list_head srv_req_in_queue; /* incoming reqs */
- struct list_head srv_request_queue; /* reqs waiting for service */
-
- struct list_head srv_request_history; /* request history */
- __u64 srv_request_seq; /* next request sequence # */
- __u64 srv_request_max_cull_seq; /* highest seq culled from history */
- svcreq_printfn_t srv_request_history_print_fn; /* service-specific print fn */
-
- struct list_head srv_idle_rqbds; /* request buffers to be reposted */
- struct list_head srv_active_rqbds; /* req buffers receiving */
- struct list_head srv_history_rqbds; /* request buffer history */
- int srv_nrqbd_receiving; /* # posted request buffers */
- int srv_n_history_rqbds; /* # request buffers in history */
- int srv_max_history_rqbds;/* max # request buffers in history */
-
- atomic_t srv_outstanding_replies;
- struct list_head srv_active_replies; /* all the active replies */
- struct list_head srv_reply_queue; /* replies waiting for service */
-
- cfs_waitq_t srv_waitq; /* all threads sleep on this. This
- * wait-queue is signalled when new
- * incoming request arrives and when
- * difficult reply has to be handled. */
-
- struct list_head srv_threads; /* service thread list */
- svc_handler_t srv_handler;
+ cfs_spinlock_t srv_at_lock;
+ struct ptlrpc_at_array srv_at_array; /* reqs waiting for replies */
+ cfs_timer_t srv_at_timer; /* early reply timer */
+
+ int srv_n_queued_reqs; /* # reqs in either of the queues below */
+ int srv_hpreq_count; /* # hp requests handled */
+ int srv_hpreq_ratio; /* # hp per lp reqs to handle */
+ cfs_list_t srv_req_in_queue; /* incoming reqs */
+ cfs_list_t srv_request_queue; /* reqs waiting for service */
+ cfs_list_t srv_request_hpq; /* high priority queue */
+
+ cfs_list_t srv_request_history; /* request history */
+ __u64 srv_request_seq; /* next request sequence # */
+ __u64 srv_request_max_cull_seq; /* highest seq culled from history */
+ svcreq_printfn_t srv_request_history_print_fn; /* service-specific print fn */
+
+ cfs_list_t srv_idle_rqbds; /* request buffers to be reposted */
+ cfs_list_t srv_active_rqbds; /* req buffers receiving */
+ cfs_list_t srv_history_rqbds; /* request buffer history */
+ int srv_nrqbd_receiving; /* # posted request buffers */
+ int srv_n_history_rqbds; /* # request buffers in history */
+ int srv_max_history_rqbds;/* max # request buffers in history */
+
+ cfs_atomic_t srv_outstanding_replies;
+ cfs_list_t srv_active_replies; /* all the active replies */
+#ifndef __KERNEL__
+ cfs_list_t srv_reply_queue; /* replies waiting for service */
+#endif
+ cfs_waitq_t srv_waitq; /* all threads sleep on this. This
+ * wait-queue is signalled when new
+ * incoming request arrives and when
+ * difficult reply has to be handled. */
+
+ cfs_list_t srv_threads; /* service thread list */
+ svc_handler_t srv_handler;
+ svc_hpreq_handler_t srv_hpreq_handler; /* hp request handler */
char *srv_name; /* only statically allocated strings here; we don't clean them */
char *srv_thread_name; /* only statically allocated strings here; we don't clean them */
- spinlock_t srv_lock;
+ cfs_spinlock_t srv_lock;
- cfs_proc_dir_entry_t *srv_procroot;
- struct lprocfs_stats *srv_stats;
+ cfs_proc_dir_entry_t *srv_procroot;
+ struct lprocfs_stats *srv_stats;
/* List of free reply_states */
- struct list_head srv_free_rs_list;
+ cfs_list_t srv_free_rs_list;
/* waitq to run, when adding stuff to srv_free_rs_list */
- cfs_waitq_t srv_free_rs_waitq;
+ cfs_waitq_t srv_free_rs_waitq;
/*
* Tags for lu_context associated with this thread, see struct
* lu_context.
*/
- __u32 srv_ctx_tags;
+ __u32 srv_ctx_tags;
/*
* if non-NULL called during thread creation (ptlrpc_start_thread())
* to initialize service specific per-thread state.
struct ptlrpcd_ctl {
/**
- * Ptlrpc thread control flags (LIOD_START, LIOD_STOP, LIOD_STOP_FORCE)
+ * Ptlrpc thread control flags (LIOD_START, LIOD_STOP, LIOD_FORCE)
*/
unsigned long pc_flags;
/**
* Thread lock protecting structure fields.
*/
- spinlock_t pc_lock;
+ cfs_spinlock_t pc_lock;
/**
* Start completion.
*/
- struct completion pc_starting;
+ cfs_completion_t pc_starting;
/**
* Stop completion.
*/
- struct completion pc_finishing;
+ cfs_completion_t pc_finishing;
/**
* Thread requests set.
*/
*/
LIOD_STOP = 1 << 1,
/**
- * Ptlrpc thread stop force flag. This will cause also
- * aborting any inflight rpcs handled by thread.
+ * Ptlrpc thread force flag (only stop force so far).
+ * This will cause aborting any inflight rpcs handled
+ * by thread if LIOD_STOP is specified.
*/
- LIOD_STOP_FORCE = 1 << 2
+ LIOD_FORCE = 1 << 2,
+ /**
+ * This is a recovery ptlrpc thread.
+ */
+ LIOD_RECOVERY = 1 << 3
};
/* ptlrpc/events.c */
int ptlrpc_start_bulk_transfer(struct ptlrpc_bulk_desc *desc);
void ptlrpc_abort_bulk(struct ptlrpc_bulk_desc *desc);
int ptlrpc_register_bulk(struct ptlrpc_request *req);
-void ptlrpc_unregister_bulk (struct ptlrpc_request *req);
+int ptlrpc_unregister_bulk(struct ptlrpc_request *req, int async);
-static inline int ptlrpc_bulk_active (struct ptlrpc_bulk_desc *desc)
+static inline int ptlrpc_server_bulk_active(struct ptlrpc_bulk_desc *desc)
{
- int rc;
+ int rc;
+
+ LASSERT(desc != NULL);
+
+ cfs_spin_lock(&desc->bd_lock);
+ rc = desc->bd_network_rw;
+ cfs_spin_unlock(&desc->bd_lock);
+ return rc;
+}
+
+static inline int ptlrpc_client_bulk_active(struct ptlrpc_request *req)
+{
+ struct ptlrpc_bulk_desc *desc = req->rq_bulk;
+ int rc;
+
+ LASSERT(req != NULL);
+
+ if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_BULK_UNLINK) &&
+ req->rq_bulk_deadline > cfs_time_current_sec())
+ return 1;
- spin_lock(&desc->bd_lock);
+ if (!desc)
+ return 0;
+
+ cfs_spin_lock(&desc->bd_lock);
rc = desc->bd_network_rw;
- spin_unlock(&desc->bd_lock);
- return (rc);
+ cfs_spin_unlock(&desc->bd_lock);
+ return rc;
}
#define PTLRPC_REPLY_MAYBE_DIFFICULT 0x01
void ptlrpc_cleanup_client(struct obd_import *imp);
struct ptlrpc_connection *ptlrpc_uuid_to_connection(struct obd_uuid *uuid);
-static inline int
-ptlrpc_client_recv_or_unlink (struct ptlrpc_request *req)
-{
- int rc;
-
- spin_lock(&req->rq_lock);
- rc = req->rq_receiving_reply || req->rq_must_unlink;
- spin_unlock(&req->rq_lock);
- return (rc);
-}
-
-static inline void
-ptlrpc_wake_client_req (struct ptlrpc_request *req)
-{
- if (req->rq_set == NULL)
- cfs_waitq_signal(&req->rq_reply_waitq);
- else
- cfs_waitq_signal(&req->rq_set->set_waitq);
-}
-
int ptlrpc_queue_wait(struct ptlrpc_request *req);
int ptlrpc_replay_req(struct ptlrpc_request *req);
-void ptlrpc_unregister_reply(struct ptlrpc_request *req);
+int ptlrpc_unregister_reply(struct ptlrpc_request *req, int async);
void ptlrpc_restart_req(struct ptlrpc_request *req);
void ptlrpc_abort_inflight(struct obd_import *imp);
+void ptlrpc_cleanup_imp(struct obd_import *imp);
void ptlrpc_abort_set(struct ptlrpc_request_set *set);
struct ptlrpc_request_set *ptlrpc_prep_set(void);
int ptlrpc_request_bufs_pack(struct ptlrpc_request *request,
__u32 version, int opcode, char **bufs,
struct ptlrpc_cli_ctx *ctx);
+struct ptlrpc_request *ptlrpc_prep_fakereq(struct obd_import *imp,
+ unsigned int timeout,
+ ptlrpc_interpterer_t interpreter);
+void ptlrpc_fakereq_finished(struct ptlrpc_request *req);
+
struct ptlrpc_request *ptlrpc_prep_req(struct obd_import *imp, __u32 version,
int opcode, int count, __u32 *lengths,
char **bufs);
/* ptlrpc/service.c */
void ptlrpc_save_lock (struct ptlrpc_request *req,
- struct lustre_handle *lock, int mode);
-void ptlrpc_commit_replies (struct obd_device *obd);
+ struct lustre_handle *lock, int mode, int no_ack);
+void ptlrpc_commit_replies(struct obd_export *exp);
+void ptlrpc_dispatch_difficult_reply (struct ptlrpc_reply_state *rs);
void ptlrpc_schedule_difficult_reply (struct ptlrpc_reply_state *rs);
struct ptlrpc_service *ptlrpc_init_svc_conf(struct ptlrpc_service_conf *c,
svc_handler_t h, char *name,
cfs_proc_dir_entry_t *proc_entry,
svcreq_printfn_t,
int min_threads, int max_threads,
- char *threadname, __u32 ctx_tags);
+ char *threadname, __u32 ctx_tags,
+ svc_hpreq_handler_t);
void ptlrpc_stop_all_threads(struct ptlrpc_service *svc);
int ptlrpc_start_threads(struct obd_device *dev, struct ptlrpc_service *svc);
int liblustre_check_services (void *arg);
void ptlrpc_daemonize(char *name);
int ptlrpc_service_health_check(struct ptlrpc_service *);
+void ptlrpc_hpreq_reorder(struct ptlrpc_request *req);
+void ptlrpc_server_active_request_inc(struct ptlrpc_request *req);
+void ptlrpc_server_active_request_dec(struct ptlrpc_request *req);
+void ptlrpc_server_drop_request(struct ptlrpc_request *req);
+#ifdef __KERNEL__
+int ptlrpc_hr_init(void);
+void ptlrpc_hr_fini(void);
+#else
+# define ptlrpc_hr_init() (0)
+# define ptlrpc_hr_fini() do {} while(0)
+#endif
struct ptlrpc_svc_data {
char *name;
/* ptlrpc/pack_generic.c */
int ptlrpc_reconnect_import(struct obd_import *imp);
-int lustre_msg_swabbed(struct lustre_msg *msg);
+
+/** ptlrpc mgs buffer swab interface */
+int ptlrpc_buf_need_swab(struct ptlrpc_request *req, const int inout,
+ int index);
+void ptlrpc_buf_set_swabbed(struct ptlrpc_request *req, const int inout,
+ int index);
+int ptlrpc_unpack_rep_msg(struct ptlrpc_request *req, int len);
+int ptlrpc_unpack_req_msg(struct ptlrpc_request *req, int len);
+
int lustre_msg_check_version(struct lustre_msg *msg, __u32 version);
void lustre_init_msg_v2(struct lustre_msg_v2 *msg, int count, __u32 *lens,
char **bufs);
int lustre_shrink_msg(struct lustre_msg *msg, int segment,
unsigned int newlen, int move_data);
void lustre_free_reply_state(struct ptlrpc_reply_state *rs);
+int __lustre_unpack_msg(struct lustre_msg *m, int len);
int lustre_msg_hdr_size(__u32 magic, int count);
int lustre_msg_size(__u32 magic, int count, __u32 *lengths);
int lustre_msg_size_v2(int count, __u32 *lengths);
int lustre_packed_msg_size(struct lustre_msg *msg);
int lustre_msg_early_size(void);
-int lustre_unpack_msg(struct lustre_msg *m, int len);
void *lustre_msg_buf_v2(struct lustre_msg_v2 *m, int n, int min_size);
void *lustre_msg_buf(struct lustre_msg *m, int n, int minlen);
int lustre_msg_buflen(struct lustre_msg *m, int n);
void lustre_msg_set_buflen(struct lustre_msg *m, int n, int len);
int lustre_msg_bufcount(struct lustre_msg *m);
char *lustre_msg_string (struct lustre_msg *m, int n, int max_len);
-void *lustre_swab_buf(struct lustre_msg *, int n, int minlen, void *swabber);
-void *lustre_swab_reqbuf(struct ptlrpc_request *req, int n, int minlen,
- void *swabber);
-void *lustre_swab_repbuf(struct ptlrpc_request *req, int n, int minlen,
- void *swabber);
__u32 lustre_msghdr_get_flags(struct lustre_msg *msg);
void lustre_msghdr_set_flags(struct lustre_msg *msg, __u32 flags);
__u32 lustre_msg_get_flags(struct lustre_msg *msg);
__u32 lustre_msg_get_opc(struct lustre_msg *msg);
__u64 lustre_msg_get_last_xid(struct lustre_msg *msg);
__u64 lustre_msg_get_last_committed(struct lustre_msg *msg);
+__u64 *lustre_msg_get_versions(struct lustre_msg *msg);
__u64 lustre_msg_get_transno(struct lustre_msg *msg);
__u64 lustre_msg_get_slv(struct lustre_msg *msg);
__u32 lustre_msg_get_limit(struct lustre_msg *msg);
__u32 lustre_msg_get_timeout(struct lustre_msg *msg);
__u32 lustre_msg_get_service_time(struct lustre_msg *msg);
__u32 lustre_msg_get_cksum(struct lustre_msg *msg);
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 0, 0)
+__u32 lustre_msg_calc_cksum(struct lustre_msg *msg, int compat18);
+#else
+# warning "remove checksum compatibility support for b1_8"
__u32 lustre_msg_calc_cksum(struct lustre_msg *msg);
+#endif
void lustre_msg_set_handle(struct lustre_msg *msg,struct lustre_handle *handle);
void lustre_msg_set_type(struct lustre_msg *msg, __u32 type);
void lustre_msg_set_opc(struct lustre_msg *msg, __u32 opc);
void lustre_msg_set_last_xid(struct lustre_msg *msg, __u64 last_xid);
void lustre_msg_set_last_committed(struct lustre_msg *msg,__u64 last_committed);
+void lustre_msg_set_versions(struct lustre_msg *msg, __u64 *versions);
void lustre_msg_set_transno(struct lustre_msg *msg, __u64 transno);
void lustre_msg_set_status(struct lustre_msg *msg, __u32 status);
void lustre_msg_set_conn_cnt(struct lustre_msg *msg, __u32 conn_cnt);
}
static inline void
+ptlrpc_rqphase_move(struct ptlrpc_request *req, enum rq_phase new_phase)
+{
+ if (req->rq_phase == new_phase)
+ return;
+
+ if (new_phase == RQ_PHASE_UNREGISTERING) {
+ req->rq_next_phase = req->rq_phase;
+ if (req->rq_import)
+ cfs_atomic_inc(&req->rq_import->imp_unregistering);
+ }
+
+ if (req->rq_phase == RQ_PHASE_UNREGISTERING) {
+ if (req->rq_import)
+ cfs_atomic_dec(&req->rq_import->imp_unregistering);
+ }
+
+ DEBUG_REQ(D_INFO, req, "move req \"%s\" -> \"%s\"",
+ ptlrpc_rqphase2str(req), ptlrpc_phase2str(new_phase));
+
+ req->rq_phase = new_phase;
+}
+
+static inline int
+ptlrpc_client_early(struct ptlrpc_request *req)
+{
+ if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_REPL_UNLINK) &&
+ req->rq_reply_deadline > cfs_time_current_sec())
+ return 0;
+ return req->rq_early;
+}
+
+static inline int
+ptlrpc_client_replied(struct ptlrpc_request *req)
+{
+ if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_REPL_UNLINK) &&
+ req->rq_reply_deadline > cfs_time_current_sec())
+ return 0;
+ return req->rq_replied;
+}
+
+static inline int
+ptlrpc_client_recv(struct ptlrpc_request *req)
+{
+ if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_REPL_UNLINK) &&
+ req->rq_reply_deadline > cfs_time_current_sec())
+ return 1;
+ return req->rq_receiving_reply;
+}
+
+static inline int
+ptlrpc_client_recv_or_unlink(struct ptlrpc_request *req)
+{
+ int rc;
+
+ cfs_spin_lock(&req->rq_lock);
+ if (OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_LONG_REPL_UNLINK) &&
+ req->rq_reply_deadline > cfs_time_current_sec()) {
+ cfs_spin_unlock(&req->rq_lock);
+ return 1;
+ }
+ rc = req->rq_receiving_reply || req->rq_must_unlink;
+ cfs_spin_unlock(&req->rq_lock);
+ return rc;
+}
+
+static inline void
+ptlrpc_client_wake_req(struct ptlrpc_request *req)
+{
+ if (req->rq_set == NULL)
+ cfs_waitq_signal(&req->rq_reply_waitq);
+ else
+ cfs_waitq_signal(&req->rq_set->set_waitq);
+}
+
+static inline void
ptlrpc_rs_addref(struct ptlrpc_reply_state *rs)
{
- LASSERT(atomic_read(&rs->rs_refcount) > 0);
- atomic_inc(&rs->rs_refcount);
+ LASSERT(cfs_atomic_read(&rs->rs_refcount) > 0);
+ cfs_atomic_inc(&rs->rs_refcount);
}
static inline void
ptlrpc_rs_decref(struct ptlrpc_reply_state *rs)
{
- LASSERT(atomic_read(&rs->rs_refcount) > 0);
- if (atomic_dec_and_test(&rs->rs_refcount))
+ LASSERT(cfs_atomic_read(&rs->rs_refcount) > 0);
+ if (cfs_atomic_dec_and_test(&rs->rs_refcount))
lustre_free_reply_state(rs);
}
}
}
+static inline int ptlrpc_send_limit_expired(struct ptlrpc_request *req)
+{
+ if (req->rq_delay_limit != 0 &&
+ cfs_time_before(cfs_time_add(req->rq_queued_time,
+ cfs_time_seconds(req->rq_delay_limit)),
+ cfs_time_current())) {
+ return 1;
+ }
+ return 0;
+}
+
+static inline int ptlrpc_no_resend(struct ptlrpc_request *req)
+{
+ if (!req->rq_no_resend && ptlrpc_send_limit_expired(req)) {
+ cfs_spin_lock(&req->rq_lock);
+ req->rq_no_resend = 1;
+ cfs_spin_unlock(&req->rq_lock);
+ }
+ return req->rq_no_resend;
+}
+
/* ldlm/ldlm_lib.c */
int client_obd_setup(struct obd_device *obddev, struct lustre_cfg *lcfg);
int client_obd_cleanup(struct obd_device *obddev);
int client_connect_import(const struct lu_env *env,
- struct lustre_handle *conn, struct obd_device *obd,
+ struct obd_export **exp, struct obd_device *obd,
struct obd_uuid *cluuid, struct obd_connect_data *,
void *localdata);
int client_disconnect_export(struct obd_export *exp);
int priority);
int client_import_del_conn(struct obd_import *imp, struct obd_uuid *uuid);
int import_set_conn_priority(struct obd_import *imp, struct obd_uuid *uuid);
+void client_destroy_import(struct obd_import *imp);
+
+int server_disconnect_export(struct obd_export *exp);
/* ptlrpc/pinger.c */
+enum timeout_event {
+ TIMEOUT_GRANT = 1
+};
+struct timeout_item;
+typedef int (*timeout_cb_t)(struct timeout_item *, void *);
int ptlrpc_pinger_add_import(struct obd_import *imp);
int ptlrpc_pinger_del_import(struct obd_import *imp);
+int ptlrpc_add_timeout_client(int time, enum timeout_event event,
+ timeout_cb_t cb, void *data,
+ cfs_list_t *obd_list);
+int ptlrpc_del_timeout_client(cfs_list_t *obd_list,
+ enum timeout_event event);
+struct ptlrpc_request * ptlrpc_prep_ping(struct obd_import *imp);
+int ptlrpc_obd_ping(struct obd_device *obd);
cfs_time_t ptlrpc_suspend_wakeup_time(void);
#ifdef __KERNEL__
void ping_evictor_start(void);
int ptlrpc_check_and_wait_suspend(struct ptlrpc_request *req);
/* ptlrpc/ptlrpcd.c */
-int ptlrpcd_start(char *name, struct ptlrpcd_ctl *pc);
+
+/**
+ * Ptlrpcd scope is a set of two threads: ptlrpcd-foo and ptlrpcd-foo-rcv,
+ * these threads are used to asynchronously send requests queued with
+ * ptlrpcd_add_req(req, PCSOPE_FOO), and to handle completion call-backs for
+ * such requests. Multiple scopes are needed to avoid dead-locks.
+ */
+enum ptlrpcd_scope {
+ /** Scope of bulk read-write rpcs. */
+ PSCOPE_BRW,
+ /** Everything else. */
+ PSCOPE_OTHER,
+ PSCOPE_NR
+};
+
+int ptlrpcd_start(const char *name, struct ptlrpcd_ctl *pc);
void ptlrpcd_stop(struct ptlrpcd_ctl *pc, int force);
void ptlrpcd_wake(struct ptlrpc_request *req);
-void ptlrpcd_add_req(struct ptlrpc_request *req);
+int ptlrpcd_add_req(struct ptlrpc_request *req, enum ptlrpcd_scope scope);
+void ptlrpcd_add_rqset(struct ptlrpc_request_set *set);
int ptlrpcd_addref(void);
void ptlrpcd_decref(void);
/* ptlrpc/llog_client.c */
extern struct llog_operations llog_client_ops;
+/** @} net */
+
#endif