4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 /** \defgroup obd_import PtlRPC import definitions
33 * Imports are client-side representation of remote obd target.
41 /** \defgroup export export
45 #include <linux/workqueue.h>
47 #include <lustre_handles.h>
48 #include <uapi/linux/lustre/lustre_idl.h>
51 * Adaptive Timeout stuff
55 #define D_ADAPTTO D_OTHER
56 #define AT_BINS 4 /* "bin" means "N seconds of history" */
57 #define AT_FLG_NOHIST 0x1 /* use last reported value only */
59 struct adaptive_timeout {
60 time64_t at_binstart; /* bin start time */
61 unsigned int at_hist[AT_BINS]; /* timeout history bins */
62 unsigned int at_flags;
63 unsigned int at_current; /* current timeout value */
64 unsigned int at_worst_ever; /* worst-ever timeout value */
65 time64_t at_worst_time; /* worst-ever timeout timestamp */
69 enum lustre_at_flags {
74 struct ptlrpc_at_array {
75 struct list_head *paa_reqs_array; /** array to hold requests */
76 __u32 paa_size; /** the size of array */
77 __u32 paa_count; /** the total count of reqs */
78 time64_t paa_deadline; /** the earliest deadline of reqs */
79 __u32 *paa_reqs_count; /** the count of reqs in each entry */
82 #define IMP_AT_MAX_PORTALS 8
84 int iat_portal[IMP_AT_MAX_PORTALS];
85 struct adaptive_timeout iat_net_latency;
86 struct adaptive_timeout iat_service_estimate[IMP_AT_MAX_PORTALS];
92 /** Possible import states */
93 enum lustre_imp_state {
94 LUSTRE_IMP_CLOSED = 1,
96 LUSTRE_IMP_DISCON = 3,
97 LUSTRE_IMP_CONNECTING = 4,
98 LUSTRE_IMP_REPLAY = 5,
99 LUSTRE_IMP_REPLAY_LOCKS = 6,
100 LUSTRE_IMP_REPLAY_WAIT = 7,
101 LUSTRE_IMP_RECOVER = 8,
103 LUSTRE_IMP_EVICTED = 10,
104 LUSTRE_IMP_IDLE = 11,
108 /** Returns test string representation of numeric import state \a state */
109 static inline char * ptlrpc_import_state_name(enum lustre_imp_state state)
111 static char *import_state_names[] = {
112 "<UNKNOWN>", "CLOSED", "NEW", "DISCONN",
113 "CONNECTING", "REPLAY", "REPLAY_LOCKS", "REPLAY_WAIT",
114 "RECOVER", "FULL", "EVICTED", "IDLE",
117 LASSERT(state < LUSTRE_IMP_LAST);
118 return import_state_names[state];
122 * List of import event types
124 enum obd_import_event {
125 IMP_EVENT_DISCON = 0x808001,
126 IMP_EVENT_INACTIVE = 0x808002,
127 IMP_EVENT_INVALIDATE = 0x808003,
128 IMP_EVENT_ACTIVE = 0x808004,
129 IMP_EVENT_OCD = 0x808005,
130 IMP_EVENT_DEACTIVATE = 0x808006,
131 IMP_EVENT_ACTIVATE = 0x808007,
135 * Definition of import connection structure
137 struct obd_import_conn {
138 /** Item for linking connections together */
139 struct list_head oic_item;
140 /** Pointer to actual PortalRPC connection */
141 struct ptlrpc_connection *oic_conn;
142 /** uuid of remote side */
143 struct obd_uuid oic_uuid;
145 * Time (64 bit seconds) of last connection attempt on this connection
147 time64_t oic_last_attempt;
151 #define IMP_STATE_HIST_LEN 16
152 struct import_state_hist {
153 enum lustre_imp_state ish_state;
158 * Defintion of PortalRPC import structure.
159 * Imports are representing client-side view to remote target.
162 /** Local handle (== id) for this import. */
163 struct portals_handle imp_handle;
164 /** Reference counter */
165 atomic_t imp_refcount;
166 struct lustre_handle imp_dlm_handle; /* client's ldlm export */
167 /** Currently active connection */
168 struct ptlrpc_connection *imp_connection;
169 /** PortalRPC client structure for this import */
170 struct ptlrpc_client *imp_client;
171 /** List element for linking into pinger chain */
172 struct list_head imp_pinger_chain;
173 /** work struct for destruction of import */
174 struct work_struct imp_zombie_work;
177 * Lists of requests that are retained for replay, waiting for a reply,
178 * or waiting for recovery to complete, respectively.
181 struct list_head imp_replay_list;
182 struct list_head imp_sending_list;
183 struct list_head imp_delayed_list;
187 * List of requests that are retained for committed open replay. Once
188 * open is committed, open replay request will be moved from the
189 * imp_replay_list into the imp_committed_list.
190 * The imp_replay_cursor is for accelerating searching during replay.
193 struct list_head imp_committed_list;
194 struct list_head *imp_replay_cursor;
197 /** List of not replied requests */
198 struct list_head imp_unreplied_list;
199 /** Known maximal replied XID */
200 __u64 imp_known_replied_xid;
202 /** obd device for this import */
203 struct obd_device *imp_obd;
206 * some seciruty-related fields
209 struct ptlrpc_sec *imp_sec;
210 struct mutex imp_sec_mutex;
211 time64_t imp_sec_expire;
212 pid_t imp_sec_refpid;
215 /** Wait queue for those who need to wait for recovery completion */
216 wait_queue_head_t imp_recovery_waitq;
218 /** Number of requests currently in-flight */
219 atomic_t imp_inflight;
220 /** Number of requests currently unregistering */
221 atomic_t imp_unregistering;
222 /** Number of replay requests inflight */
223 atomic_t imp_replay_inflight;
224 /** Number of currently happening import invalidations */
225 atomic_t imp_inval_count;
226 /** Numbner of request timeouts */
227 atomic_t imp_timeouts;
228 /** Current import state */
229 enum lustre_imp_state imp_state;
230 /** Last replay state */
231 enum lustre_imp_state imp_replay_state;
232 /** History of import states */
233 struct import_state_hist imp_state_hist[IMP_STATE_HIST_LEN];
234 int imp_state_hist_idx;
235 /** Current import generation. Incremented on every reconnect */
237 /** Idle connection initiated at this generation */
238 int imp_initiated_at;
239 /** Incremented every time we send reconnection request */
242 * \see ptlrpc_free_committed remembers imp_generation value here
243 * after a check to save on unnecessary replay list iterations
245 int imp_last_generation_checked;
246 /** Last tranno we replayed */
247 __u64 imp_last_replay_transno;
248 /** Last transno committed on remote side */
249 __u64 imp_peer_committed_transno;
251 * \see ptlrpc_free_committed remembers last_transno since its last
252 * check here and if last_transno did not change since last run of
253 * ptlrpc_free_committed and import generation is the same, we can
254 * skip looking for requests to remove from replay list as optimisation
256 __u64 imp_last_transno_checked;
258 * Remote export handle. This is how remote side knows what export
259 * we are talking to. Filled from response to connect request
261 struct lustre_handle imp_remote_handle;
262 /** When to perform next ping. time in jiffies. */
263 time64_t imp_next_ping;
264 /** When we last successfully connected. time in 64bit jiffies */
265 time64_t imp_last_success_conn;
267 /** List of all possible connection for import. */
268 struct list_head imp_conn_list;
270 * Current connection. \a imp_connection is imp_conn_current->oic_conn
272 struct obd_import_conn *imp_conn_current;
274 /** Protects flags, level, generation, conn_cnt, *_list */
278 unsigned long imp_no_timeout:1, /* timeouts are disabled */
279 imp_invalid:1, /* evicted */
280 /* administratively disabled */
282 /* try to recover the import */
284 /* don't run recovery (timeout instead) */
286 /* use 1/2 timeout on MDS' OSCs */
287 imp_server_timeout:1,
288 /* VBR: imp in delayed recovery */
289 imp_delayed_recovery:1,
290 /* recovery by versions was failed */
292 /* force an immidiate ping */
294 /* force a scheduled ping */
295 imp_force_next_verify:1,
298 /* resend for replay */
300 /* disable normal recovery, for test only. */
301 imp_no_pinger_recover:1,
302 /* import must be reconnected instead of
303 * chouse new connection */
304 imp_force_reconnect:1,
305 /* import has tried to connect with server */
307 /* connected but not FULL yet */
309 __u32 imp_connect_op;
310 __u32 imp_idle_timeout;
311 struct obd_connect_data imp_connect_data;
312 __u64 imp_connect_flags_orig;
313 __u64 imp_connect_flags2_orig;
314 int imp_connect_error;
316 enum lustre_msg_magic imp_msg_magic;
317 /* adjusted based on server capability */
318 enum lustre_msghdr imp_msghdr_flags;
320 /* adaptive timeout data */
321 struct imp_at imp_at;
322 time64_t imp_last_reply_time; /* for health check */
326 static inline unsigned int at_est2timeout(unsigned int val)
328 /* add an arbitrary minimum: 125% +5 sec */
329 return (val + (val >> 2) + 5);
332 static inline unsigned int at_timeout2est(unsigned int val)
334 /* restore estimate value from timeout: e=4/5(t-5) */
336 return (max((val << 2) / 5, 5U) - 4);
339 static inline void at_reset_nolock(struct adaptive_timeout *at, int val)
341 at->at_current = val;
342 at->at_worst_ever = val;
343 at->at_worst_time = ktime_get_real_seconds();
346 static inline void at_reset(struct adaptive_timeout *at, int val)
348 spin_lock(&at->at_lock);
349 at_reset_nolock(at, val);
350 spin_unlock(&at->at_lock);
353 static inline void at_init(struct adaptive_timeout *at, int val, int flags) {
354 memset(at, 0, sizeof(*at));
355 spin_lock_init(&at->at_lock);
356 at->at_flags = flags;
360 static inline void at_reinit(struct adaptive_timeout *at, int val, int flags)
362 spin_lock(&at->at_lock);
364 memset(at->at_hist, 0, sizeof(at->at_hist));
365 at->at_flags = flags;
366 at_reset_nolock(at, val);
367 spin_unlock(&at->at_lock);
370 extern unsigned int at_min;
371 static inline int at_get(struct adaptive_timeout *at) {
372 return (at->at_current > at_min) ? at->at_current : at_min;
374 int at_measured(struct adaptive_timeout *at, unsigned int val);
375 int import_at_get_index(struct obd_import *imp, int portal);
376 extern unsigned int at_max;
377 #define AT_OFF (at_max == 0)
381 extern struct obd_import *class_exp2cliimp(struct obd_export *);
382 extern struct obd_import *class_conn2cliimp(struct lustre_handle *);
386 #endif /* __IMPORT_H */