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 (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
32 * Copyright (c) 2011, Whamcloud, Inc.
35 * This file is part of Lustre, http://www.lustre.org/
36 * Lustre is a trademark of Sun Microsystems, Inc.
38 /** \defgroup obd_import PtlRPC import definitions
39 * Imports are client-side representation of remote obd target.
47 /** \defgroup export export
52 #include <lustre_handles.h>
53 #include <lustre/lustre_idl.h>
57 * Adaptive Timeout stuff
61 #define D_ADAPTTO D_OTHER
62 #define AT_BINS 4 /* "bin" means "N seconds of history" */
63 #define AT_FLG_NOHIST 0x1 /* use last reported value only */
65 struct adaptive_timeout {
66 time_t at_binstart; /* bin start time */
67 unsigned int at_hist[AT_BINS]; /* timeout history bins */
68 unsigned int at_flags;
69 unsigned int at_current; /* current timeout value */
70 unsigned int at_worst_ever; /* worst-ever timeout value */
71 time_t at_worst_time; /* worst-ever timeout timestamp */
72 cfs_spinlock_t at_lock;
75 struct ptlrpc_at_array {
76 cfs_list_t *paa_reqs_array; /** array to hold requests */
77 __u32 paa_size; /** the size of array */
78 __u32 paa_count; /** the total count of reqs */
79 time_t paa_deadline; /** the earliest deadline of reqs */
80 __u32 *paa_reqs_count; /** the count of reqs in each entry */
83 #define IMP_AT_MAX_PORTALS 8
85 int iat_portal[IMP_AT_MAX_PORTALS];
86 struct adaptive_timeout iat_net_latency;
87 struct adaptive_timeout iat_service_estimate[IMP_AT_MAX_PORTALS];
93 /** Possible import states */
94 enum lustre_imp_state {
95 LUSTRE_IMP_CLOSED = 1,
97 LUSTRE_IMP_DISCON = 3,
98 LUSTRE_IMP_CONNECTING = 4,
99 LUSTRE_IMP_REPLAY = 5,
100 LUSTRE_IMP_REPLAY_LOCKS = 6,
101 LUSTRE_IMP_REPLAY_WAIT = 7,
102 LUSTRE_IMP_RECOVER = 8,
104 LUSTRE_IMP_EVICTED = 10,
107 /** Returns test string representation of numeric import state \a state */
108 static inline char * ptlrpc_import_state_name(enum lustre_imp_state state)
110 static char* import_state_names[] = {
111 "<UNKNOWN>", "CLOSED", "NEW", "DISCONN",
112 "CONNECTING", "REPLAY", "REPLAY_LOCKS", "REPLAY_WAIT",
113 "RECOVER", "FULL", "EVICTED",
116 LASSERT (state <= LUSTRE_IMP_EVICTED);
117 return import_state_names[state];
121 * List of import event types
123 enum obd_import_event {
124 IMP_EVENT_DISCON = 0x808001,
125 IMP_EVENT_INACTIVE = 0x808002,
126 IMP_EVENT_INVALIDATE = 0x808003,
127 IMP_EVENT_ACTIVE = 0x808004,
128 IMP_EVENT_OCD = 0x808005,
129 IMP_EVENT_DEACTIVATE = 0x808006,
130 IMP_EVENT_ACTIVATE = 0x808007,
134 * Definition of import connection structure
136 struct obd_import_conn {
137 /** Item for linking connections together */
139 /** Pointer to actual PortalRPC connection */
140 struct ptlrpc_connection *oic_conn;
141 /** uuid of remote side */
142 struct obd_uuid oic_uuid;
144 * Time (64 bit jiffies) of last connection attempt on this connection
146 __u64 oic_last_attempt;
150 #define IMP_STATE_HIST_LEN 16
151 struct import_state_hist {
152 enum lustre_imp_state ish_state;
157 * Defintion of PortalRPC import structure.
158 * Imports are representing client-side view to remote target.
161 /** Local handle (== id) for this import. */
162 struct portals_handle imp_handle;
163 /** Reference counter */
164 cfs_atomic_t imp_refcount;
165 struct lustre_handle imp_dlm_handle; /* client's ldlm export */
166 /** Currently active connection */
167 struct ptlrpc_connection *imp_connection;
168 /** PortalRPC client structure for this import */
169 struct ptlrpc_client *imp_client;
170 /** List element for linking into pinger chain */
171 cfs_list_t imp_pinger_chain;
172 /** List element for linking into chain for destruction */
173 cfs_list_t imp_zombie_chain;
176 * Lists of requests that are retained for replay, waiting for a reply,
177 * or waiting for recovery to complete, respectively.
180 cfs_list_t imp_replay_list;
181 cfs_list_t imp_sending_list;
182 cfs_list_t imp_delayed_list;
185 /** obd device for this import */
186 struct obd_device *imp_obd;
189 * some seciruty-related fields
192 struct ptlrpc_sec *imp_sec;
193 cfs_mutex_t imp_sec_mutex;
194 cfs_time_t imp_sec_expire;
197 /** Wait queue for those who need to wait for recovery completion */
198 cfs_waitq_t imp_recovery_waitq;
200 /** Number of requests currently in-flight */
201 cfs_atomic_t imp_inflight;
202 /** Number of requests currently unregistering */
203 cfs_atomic_t imp_unregistering;
204 /** Number of replay requests inflight */
205 cfs_atomic_t imp_replay_inflight;
206 /** Number of currently happening import invalidations */
207 cfs_atomic_t imp_inval_count;
208 /** Numbner of request timeouts */
209 cfs_atomic_t imp_timeouts;
210 /** Current import state */
211 enum lustre_imp_state imp_state;
212 /** History of import states */
213 struct import_state_hist imp_state_hist[IMP_STATE_HIST_LEN];
214 int imp_state_hist_idx;
215 /** Current import generation. Incremented on every reconnect */
217 /** Incremented every time we send reconnection request */
220 * \see ptlrpc_free_committed remembers imp_generation value here
221 * after a check to save on unnecessary replay list iterations
223 int imp_last_generation_checked;
224 /** Last tranno we replayed */
225 __u64 imp_last_replay_transno;
226 /** Last transno committed on remote side */
227 __u64 imp_peer_committed_transno;
229 * \see ptlrpc_free_committed remembers last_transno since its last
230 * check here and if last_transno did not change since last run of
231 * ptlrpc_free_committed and import generation is the same, we can
232 * skip looking for requests to remove from replay list as optimisation
234 __u64 imp_last_transno_checked;
236 * Remote export handle. This is how remote side knows what export
237 * we are talking to. Filled from response to connect request
239 struct lustre_handle imp_remote_handle;
240 /** When to perform next ping. time in jiffies. */
241 cfs_time_t imp_next_ping;
242 /** When we last succesfully connected. time in 64bit jiffies */
243 __u64 imp_last_success_conn;
245 /** List of all possible connection for import. */
246 cfs_list_t imp_conn_list;
248 * Current connection. \a imp_connection is imp_conn_current->oic_conn
250 struct obd_import_conn *imp_conn_current;
252 /** Protects flags, level, generation, conn_cnt, *_list */
253 cfs_spinlock_t imp_lock;
256 unsigned long imp_no_timeout:1, /* timeouts are disabled */
257 imp_invalid:1, /* evicted */
258 imp_deactive:1, /* administratively disabled */
259 imp_replayable:1, /* try to recover the import */
260 imp_dlm_fake:1, /* don't run recovery (timeout instead) */
261 imp_server_timeout:1, /* use 1/2 timeout on MDS' OSCs */
262 imp_delayed_recovery:1, /* VBR: imp in delayed recovery */
263 imp_no_lock_replay:1, /* VBR: if gap was found then no lock replays */
264 imp_vbr_failed:1, /* recovery by versions was failed */
265 imp_force_verify:1, /* force an immidiate ping */
266 imp_pingable:1, /* pingable */
267 imp_resend_replay:1, /* resend for replay */
268 imp_no_pinger_recover:1,/* disable normal recovery, for test only. */
269 imp_force_reconnect:1; /* import must be reconnected instead of chouse new connection */
270 __u32 imp_connect_op;
271 struct obd_connect_data imp_connect_data;
272 __u64 imp_connect_flags_orig;
273 int imp_connect_error;
276 __u32 imp_msghdr_flags; /* adjusted based on server capability */
278 struct ptlrpc_request_pool *imp_rq_pool; /* emergency request pool */
280 struct imp_at imp_at; /* adaptive timeout data */
281 time_t imp_last_reply_time; /* for health check */
284 typedef void (*obd_import_callback)(struct obd_import *imp, void *closure,
285 int event, void *event_arg, void *cb_data);
288 * Structure for import observer.
289 * It is possible to register "observer" on an import and every time
290 * something happens to an import (like connect/evict/disconnect)
291 * obderver will get its callback called with event type
293 struct obd_import_observer {
294 cfs_list_t oio_chain;
295 obd_import_callback oio_cb;
299 void class_observe_import(struct obd_import *imp, obd_import_callback cb,
301 void class_unobserve_import(struct obd_import *imp, obd_import_callback cb,
303 void class_notify_import_observers(struct obd_import *imp, int event,
307 static inline unsigned int at_est2timeout(unsigned int val)
309 /* add an arbitrary minimum: 125% +5 sec */
310 return (val + (val >> 2) + 5);
313 static inline unsigned int at_timeout2est(unsigned int val)
315 /* restore estimate value from timeout: e=4/5(t-5) */
317 return (max((val << 2) / 5, 5U) - 4);
320 static inline void at_init(struct adaptive_timeout *at, int val, int flags) {
321 memset(at, 0, sizeof(*at));
322 at->at_current = val;
323 at->at_worst_ever = val;
324 at->at_worst_time = cfs_time_current_sec();
325 at->at_flags = flags;
326 cfs_spin_lock_init(&at->at_lock);
328 extern unsigned int at_min;
329 static inline int at_get(struct adaptive_timeout *at) {
330 return (at->at_current > at_min) ? at->at_current : at_min;
332 int at_measured(struct adaptive_timeout *at, unsigned int val);
333 int import_at_get_index(struct obd_import *imp, int portal);
334 extern unsigned int at_max;
335 #define AT_OFF (at_max == 0)
339 extern struct obd_import *class_exp2cliimp(struct obd_export *);
340 extern struct obd_import *class_conn2cliimp(struct lustre_handle *);
344 #endif /* __IMPORT_H */