1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 /** \defgroup obd_import PtlRPC import definitions
37 * Imports are client-side representation of remote obd target.
45 /** \defgroup export export
50 #include <lustre_handles.h>
51 #include <lustre/lustre_idl.h>
55 * Adaptive Timeout stuff
59 #define D_ADAPTTO D_OTHER
60 #define AT_BINS 4 /* "bin" means "N seconds of history" */
61 #define AT_FLG_NOHIST 0x1 /* use last reported value only */
63 struct adaptive_timeout {
64 time_t at_binstart; /* bin start time */
65 unsigned int at_hist[AT_BINS]; /* timeout history bins */
66 unsigned int at_flags;
67 unsigned int at_current; /* current timeout value */
68 unsigned int at_worst_ever; /* worst-ever timeout value */
69 time_t at_worst_time; /* worst-ever timeout timestamp */
70 cfs_spinlock_t at_lock;
73 struct ptlrpc_at_array {
74 cfs_list_t *paa_reqs_array; /** array to hold requests */
75 __u32 paa_size; /** the size of array */
76 __u32 paa_count; /** the total count of reqs */
77 time_t paa_deadline; /** the earliest deadline of reqs */
78 __u32 *paa_reqs_count; /** the count of reqs in each entry */
81 #define IMP_AT_MAX_PORTALS 8
83 int iat_portal[IMP_AT_MAX_PORTALS];
84 struct adaptive_timeout iat_net_latency;
85 struct adaptive_timeout iat_service_estimate[IMP_AT_MAX_PORTALS];
91 /** Possible import states */
92 enum lustre_imp_state {
93 LUSTRE_IMP_CLOSED = 1,
95 LUSTRE_IMP_DISCON = 3,
96 LUSTRE_IMP_CONNECTING = 4,
97 LUSTRE_IMP_REPLAY = 5,
98 LUSTRE_IMP_REPLAY_LOCKS = 6,
99 LUSTRE_IMP_REPLAY_WAIT = 7,
100 LUSTRE_IMP_RECOVER = 8,
102 LUSTRE_IMP_EVICTED = 10,
105 /** Returns test string representation of numeric import state \a state */
106 static inline char * ptlrpc_import_state_name(enum lustre_imp_state state)
108 static char* import_state_names[] = {
109 "<UNKNOWN>", "CLOSED", "NEW", "DISCONN",
110 "CONNECTING", "REPLAY", "REPLAY_LOCKS", "REPLAY_WAIT",
111 "RECOVER", "FULL", "EVICTED",
114 LASSERT (state <= LUSTRE_IMP_EVICTED);
115 return import_state_names[state];
119 * List of import event types
121 enum obd_import_event {
122 IMP_EVENT_DISCON = 0x808001,
123 IMP_EVENT_INACTIVE = 0x808002,
124 IMP_EVENT_INVALIDATE = 0x808003,
125 IMP_EVENT_ACTIVE = 0x808004,
126 IMP_EVENT_OCD = 0x808005,
130 * Definition of import connection structure
132 struct obd_import_conn {
133 /** Item for linking connections together */
135 /** Pointer to actual PortalRPC connection */
136 struct ptlrpc_connection *oic_conn;
137 /** uuid of remote side */
138 struct obd_uuid oic_uuid;
140 * Time (64 bit jiffies) of last connection attempt on this connection
142 __u64 oic_last_attempt;
146 #define IMP_STATE_HIST_LEN 16
147 struct import_state_hist {
148 enum lustre_imp_state ish_state;
153 * Defintion of PortalRPC import structure.
154 * Imports are representing client-side view to remote target.
157 /** Local handle (== id) for this import. */
158 struct portals_handle imp_handle;
159 /** Reference counter */
160 cfs_atomic_t imp_refcount;
161 struct lustre_handle imp_dlm_handle; /* client's ldlm export */
162 /** Currently active connection */
163 struct ptlrpc_connection *imp_connection;
164 /** PortalRPC client structure for this import */
165 struct ptlrpc_client *imp_client;
166 /** List element for linking into pinger chain */
167 cfs_list_t imp_pinger_chain;
168 /** List element for linking into chain for destruction */
169 cfs_list_t imp_zombie_chain;
172 * Lists of requests that are retained for replay, waiting for a reply,
173 * or waiting for recovery to complete, respectively.
176 cfs_list_t imp_replay_list;
177 cfs_list_t imp_sending_list;
178 cfs_list_t imp_delayed_list;
181 /** obd device for this import */
182 struct obd_device *imp_obd;
185 * some seciruty-related fields
188 struct ptlrpc_sec *imp_sec;
189 cfs_semaphore_t imp_sec_mutex;
190 cfs_time_t imp_sec_expire;
193 /** Wait queue for those who need to wait for recovery completion */
194 cfs_waitq_t imp_recovery_waitq;
196 /** Number of requests currently in-flight */
197 cfs_atomic_t imp_inflight;
198 /** Number of requests currently unregistering */
199 cfs_atomic_t imp_unregistering;
200 /** Number of replay requests inflight */
201 cfs_atomic_t imp_replay_inflight;
202 /** Number of currently happening import invalidations */
203 cfs_atomic_t imp_inval_count;
204 /** Numbner of request timeouts */
205 cfs_atomic_t imp_timeouts;
206 /** Current import state */
207 enum lustre_imp_state imp_state;
208 /** History of import states */
209 struct import_state_hist imp_state_hist[IMP_STATE_HIST_LEN];
210 int imp_state_hist_idx;
211 /** Current import generation. Incremented on every reconnect */
213 /** Incremented every time we send reconnection request */
216 * \see ptlrpc_free_committed remembers imp_generation value here
217 * after a check to save on unnecessary replay list iterations
219 int imp_last_generation_checked;
220 /** Last tranno we replayed */
221 __u64 imp_last_replay_transno;
222 /** Last transno committed on remote side */
223 __u64 imp_peer_committed_transno;
225 * \see ptlrpc_free_committed remembers last_transno since its last
226 * check here and if last_transno did not change since last run of
227 * ptlrpc_free_committed and import generation is the same, we can
228 * skip looking for requests to remove from replay list as optimisation
230 __u64 imp_last_transno_checked;
232 * Remote export handle. This is how remote side knows what export
233 * we are talking to. Filled from response to connect request
235 struct lustre_handle imp_remote_handle;
236 /** When to perform next ping. time in jiffies. */
237 cfs_time_t imp_next_ping;
238 /** When we last succesfully connected. time in 64bit jiffies */
239 __u64 imp_last_success_conn;
241 /** List of all possible connection for import. */
242 cfs_list_t imp_conn_list;
244 * Current connection. \a imp_connection is imp_conn_current->oic_conn
246 struct obd_import_conn *imp_conn_current;
248 /** Protects flags, level, generation, conn_cnt, *_list */
249 cfs_spinlock_t imp_lock;
252 unsigned long imp_no_timeout:1, /* timeouts are disabled */
253 imp_invalid:1, /* evicted */
254 imp_deactive:1, /* administratively disabled */
255 imp_replayable:1, /* try to recover the import */
256 imp_dlm_fake:1, /* don't run recovery (timeout instead) */
257 imp_server_timeout:1, /* use 1/2 timeout on MDS' OSCs */
258 imp_delayed_recovery:1, /* VBR: imp in delayed recovery */
259 imp_no_lock_replay:1, /* VBR: if gap was found then no lock replays */
260 imp_vbr_failed:1, /* recovery by versions was failed */
261 imp_force_verify:1, /* force an immidiate ping */
262 imp_pingable:1, /* pingable */
263 imp_resend_replay:1, /* resend for replay */
264 imp_force_reconnect:1; /* import must be reconnected instead of chouse new connection */
265 __u32 imp_connect_op;
266 struct obd_connect_data imp_connect_data;
267 __u64 imp_connect_flags_orig;
268 int imp_connect_error;
271 __u32 imp_msghdr_flags; /* adjusted based on server capability */
273 struct ptlrpc_request_pool *imp_rq_pool; /* emergency request pool */
275 struct imp_at imp_at; /* adaptive timeout data */
276 time_t imp_last_reply_time; /* for health check */
279 typedef void (*obd_import_callback)(struct obd_import *imp, void *closure,
280 int event, void *event_arg, void *cb_data);
283 * Structure for import observer.
284 * It is possible to register "observer" on an import and every time
285 * something happens to an import (like connect/evict/disconnect)
286 * obderver will get its callback called with event type
288 struct obd_import_observer {
289 cfs_list_t oio_chain;
290 obd_import_callback oio_cb;
294 void class_observe_import(struct obd_import *imp, obd_import_callback cb,
296 void class_unobserve_import(struct obd_import *imp, obd_import_callback cb,
298 void class_notify_import_observers(struct obd_import *imp, int event,
302 static inline unsigned int at_est2timeout(unsigned int val)
304 /* add an arbitrary minimum: 125% +5 sec */
305 return (val + (val >> 2) + 5);
308 static inline unsigned int at_timeout2est(unsigned int val)
310 /* restore estimate value from timeout: e=4/5(t-5) */
312 return (max((val << 2) / 5, 5U) - 4);
315 static inline void at_init(struct adaptive_timeout *at, int val, int flags) {
316 memset(at, 0, sizeof(*at));
317 at->at_current = val;
318 at->at_worst_ever = val;
319 at->at_worst_time = cfs_time_current_sec();
320 at->at_flags = flags;
321 cfs_spin_lock_init(&at->at_lock);
323 static inline int at_get(struct adaptive_timeout *at) {
324 return at->at_current;
326 int at_measured(struct adaptive_timeout *at, unsigned int val);
327 int import_at_get_index(struct obd_import *imp, int portal);
328 extern unsigned int at_max;
329 #define AT_OFF (at_max == 0)
333 extern struct obd_import *class_exp2cliimp(struct obd_export *);
334 extern struct obd_import *class_conn2cliimp(struct lustre_handle *);
338 #endif /* __IMPORT_H */