Commit | Line | Data |
---|---|---|
b8907339 PR |
1 | /* |
2 | drbd_state.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | Thanks to Carter Burden, Bart Grantham and Gennadiy Nerubayev | |
11 | from Logicworks, Inc. for making SDP replication support possible. | |
12 | ||
13 | drbd is free software; you can redistribute it and/or modify | |
14 | it under the terms of the GNU General Public License as published by | |
15 | the Free Software Foundation; either version 2, or (at your option) | |
16 | any later version. | |
17 | ||
18 | drbd is distributed in the hope that it will be useful, | |
19 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
20 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
21 | GNU General Public License for more details. | |
22 | ||
23 | You should have received a copy of the GNU General Public License | |
24 | along with drbd; see the file COPYING. If not, write to | |
25 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
26 | */ | |
27 | ||
28 | #include <linux/drbd_limits.h> | |
29 | #include "drbd_int.h" | |
a3603a6e | 30 | #include "drbd_protocol.h" |
b8907339 PR |
31 | #include "drbd_req.h" |
32 | ||
33 | struct after_state_chg_work { | |
34 | struct drbd_work w; | |
35 | union drbd_state os; | |
36 | union drbd_state ns; | |
37 | enum chg_state_flags flags; | |
38 | struct completion *done; | |
39 | }; | |
40 | ||
d942ae44 PR |
41 | enum sanitize_state_warnings { |
42 | NO_WARNING, | |
43 | ABORTED_ONLINE_VERIFY, | |
44 | ABORTED_RESYNC, | |
45 | CONNECTION_LOST_NEGOTIATING, | |
46 | IMPLICITLY_UPGRADED_DISK, | |
47 | IMPLICITLY_UPGRADED_PDSK, | |
48 | }; | |
49 | ||
99920dc5 | 50 | static int w_after_state_ch(struct drbd_work *w, int unused); |
b30ab791 | 51 | static void after_state_ch(struct drbd_device *device, union drbd_state os, |
b8907339 | 52 | union drbd_state ns, enum chg_state_flags flags); |
54761697 | 53 | static enum drbd_state_rv is_valid_state(struct drbd_device *, union drbd_state); |
bde89a9e | 54 | static enum drbd_state_rv is_valid_soft_transition(union drbd_state, union drbd_state, struct drbd_connection *); |
3509502d | 55 | static enum drbd_state_rv is_valid_transition(union drbd_state os, union drbd_state ns); |
b30ab791 | 56 | static union drbd_state sanitize_state(struct drbd_device *device, union drbd_state ns, |
d942ae44 | 57 | enum sanitize_state_warnings *warn); |
b8907339 | 58 | |
2aebfabb PR |
59 | static inline bool is_susp(union drbd_state s) |
60 | { | |
61 | return s.susp || s.susp_nod || s.susp_fen; | |
62 | } | |
63 | ||
bde89a9e | 64 | bool conn_all_vols_unconf(struct drbd_connection *connection) |
0e29d163 | 65 | { |
c06ece6b | 66 | struct drbd_peer_device *peer_device; |
695d08fa | 67 | bool rv = true; |
e90285e0 | 68 | int vnr; |
0e29d163 | 69 | |
695d08fa | 70 | rcu_read_lock(); |
c06ece6b AG |
71 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
72 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
73 | if (device->state.disk != D_DISKLESS || |
74 | device->state.conn != C_STANDALONE || | |
75 | device->state.role != R_SECONDARY) { | |
695d08fa PR |
76 | rv = false; |
77 | break; | |
78 | } | |
0e29d163 | 79 | } |
695d08fa PR |
80 | rcu_read_unlock(); |
81 | ||
82 | return rv; | |
0e29d163 PR |
83 | } |
84 | ||
cb703454 PR |
85 | /* Unfortunately the states where not correctly ordered, when |
86 | they where defined. therefore can not use max_t() here. */ | |
87 | static enum drbd_role max_role(enum drbd_role role1, enum drbd_role role2) | |
88 | { | |
89 | if (role1 == R_PRIMARY || role2 == R_PRIMARY) | |
90 | return R_PRIMARY; | |
91 | if (role1 == R_SECONDARY || role2 == R_SECONDARY) | |
92 | return R_SECONDARY; | |
93 | return R_UNKNOWN; | |
94 | } | |
95 | static enum drbd_role min_role(enum drbd_role role1, enum drbd_role role2) | |
96 | { | |
97 | if (role1 == R_UNKNOWN || role2 == R_UNKNOWN) | |
98 | return R_UNKNOWN; | |
99 | if (role1 == R_SECONDARY || role2 == R_SECONDARY) | |
100 | return R_SECONDARY; | |
101 | return R_PRIMARY; | |
102 | } | |
103 | ||
bde89a9e | 104 | enum drbd_role conn_highest_role(struct drbd_connection *connection) |
cb703454 PR |
105 | { |
106 | enum drbd_role role = R_UNKNOWN; | |
c06ece6b | 107 | struct drbd_peer_device *peer_device; |
cb703454 PR |
108 | int vnr; |
109 | ||
695d08fa | 110 | rcu_read_lock(); |
c06ece6b AG |
111 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
112 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 113 | role = max_role(role, device->state.role); |
c06ece6b | 114 | } |
695d08fa | 115 | rcu_read_unlock(); |
cb703454 PR |
116 | |
117 | return role; | |
118 | } | |
119 | ||
bde89a9e | 120 | enum drbd_role conn_highest_peer(struct drbd_connection *connection) |
cb703454 PR |
121 | { |
122 | enum drbd_role peer = R_UNKNOWN; | |
c06ece6b | 123 | struct drbd_peer_device *peer_device; |
cb703454 PR |
124 | int vnr; |
125 | ||
695d08fa | 126 | rcu_read_lock(); |
c06ece6b AG |
127 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
128 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 129 | peer = max_role(peer, device->state.peer); |
c06ece6b | 130 | } |
695d08fa | 131 | rcu_read_unlock(); |
cb703454 PR |
132 | |
133 | return peer; | |
134 | } | |
135 | ||
bde89a9e | 136 | enum drbd_disk_state conn_highest_disk(struct drbd_connection *connection) |
cb703454 PR |
137 | { |
138 | enum drbd_disk_state ds = D_DISKLESS; | |
c06ece6b | 139 | struct drbd_peer_device *peer_device; |
cb703454 PR |
140 | int vnr; |
141 | ||
695d08fa | 142 | rcu_read_lock(); |
c06ece6b AG |
143 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
144 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 145 | ds = max_t(enum drbd_disk_state, ds, device->state.disk); |
c06ece6b | 146 | } |
695d08fa | 147 | rcu_read_unlock(); |
cb703454 PR |
148 | |
149 | return ds; | |
150 | } | |
151 | ||
bde89a9e | 152 | enum drbd_disk_state conn_lowest_disk(struct drbd_connection *connection) |
4669265a PR |
153 | { |
154 | enum drbd_disk_state ds = D_MASK; | |
c06ece6b | 155 | struct drbd_peer_device *peer_device; |
4669265a PR |
156 | int vnr; |
157 | ||
695d08fa | 158 | rcu_read_lock(); |
c06ece6b AG |
159 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
160 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 161 | ds = min_t(enum drbd_disk_state, ds, device->state.disk); |
c06ece6b | 162 | } |
695d08fa | 163 | rcu_read_unlock(); |
4669265a PR |
164 | |
165 | return ds; | |
166 | } | |
167 | ||
bde89a9e | 168 | enum drbd_disk_state conn_highest_pdsk(struct drbd_connection *connection) |
cb703454 PR |
169 | { |
170 | enum drbd_disk_state ds = D_DISKLESS; | |
c06ece6b | 171 | struct drbd_peer_device *peer_device; |
cb703454 PR |
172 | int vnr; |
173 | ||
695d08fa | 174 | rcu_read_lock(); |
c06ece6b AG |
175 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
176 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 177 | ds = max_t(enum drbd_disk_state, ds, device->state.pdsk); |
c06ece6b | 178 | } |
695d08fa | 179 | rcu_read_unlock(); |
cb703454 PR |
180 | |
181 | return ds; | |
182 | } | |
183 | ||
bde89a9e | 184 | enum drbd_conns conn_lowest_conn(struct drbd_connection *connection) |
19f83c76 PR |
185 | { |
186 | enum drbd_conns conn = C_MASK; | |
c06ece6b | 187 | struct drbd_peer_device *peer_device; |
19f83c76 PR |
188 | int vnr; |
189 | ||
695d08fa | 190 | rcu_read_lock(); |
c06ece6b AG |
191 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
192 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 193 | conn = min_t(enum drbd_conns, conn, device->state.conn); |
c06ece6b | 194 | } |
695d08fa | 195 | rcu_read_unlock(); |
19f83c76 PR |
196 | |
197 | return conn; | |
198 | } | |
199 | ||
bde89a9e | 200 | static bool no_peer_wf_report_params(struct drbd_connection *connection) |
79702011 | 201 | { |
c06ece6b | 202 | struct drbd_peer_device *peer_device; |
79702011 PR |
203 | int vnr; |
204 | bool rv = true; | |
205 | ||
206 | rcu_read_lock(); | |
c06ece6b AG |
207 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
208 | if (peer_device->device->state.conn == C_WF_REPORT_PARAMS) { | |
79702011 PR |
209 | rv = false; |
210 | break; | |
211 | } | |
212 | rcu_read_unlock(); | |
213 | ||
214 | return rv; | |
215 | } | |
216 | ||
217 | ||
b8907339 PR |
218 | /** |
219 | * cl_wide_st_chg() - true if the state change is a cluster wide one | |
b30ab791 | 220 | * @device: DRBD device. |
b8907339 PR |
221 | * @os: old (current) state. |
222 | * @ns: new (wanted) state. | |
223 | */ | |
b30ab791 | 224 | static int cl_wide_st_chg(struct drbd_device *device, |
b8907339 PR |
225 | union drbd_state os, union drbd_state ns) |
226 | { | |
227 | return (os.conn >= C_CONNECTED && ns.conn >= C_CONNECTED && | |
228 | ((os.role != R_PRIMARY && ns.role == R_PRIMARY) || | |
229 | (os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
230 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) || | |
e8744f5a | 231 | (os.disk != D_FAILED && ns.disk == D_FAILED))) || |
b8907339 | 232 | (os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) || |
369bea63 PR |
233 | (os.conn == C_CONNECTED && ns.conn == C_VERIFY_S) || |
234 | (os.conn == C_CONNECTED && ns.conn == C_WF_REPORT_PARAMS); | |
b8907339 PR |
235 | } |
236 | ||
56707f9e PR |
237 | static union drbd_state |
238 | apply_mask_val(union drbd_state os, union drbd_state mask, union drbd_state val) | |
239 | { | |
240 | union drbd_state ns; | |
241 | ns.i = (os.i & ~mask.i) | val.i; | |
242 | return ns; | |
243 | } | |
244 | ||
b8907339 | 245 | enum drbd_state_rv |
b30ab791 | 246 | drbd_change_state(struct drbd_device *device, enum chg_state_flags f, |
b8907339 PR |
247 | union drbd_state mask, union drbd_state val) |
248 | { | |
249 | unsigned long flags; | |
56707f9e | 250 | union drbd_state ns; |
b8907339 PR |
251 | enum drbd_state_rv rv; |
252 | ||
0500813f | 253 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
254 | ns = apply_mask_val(drbd_read_state(device), mask, val); |
255 | rv = _drbd_set_state(device, ns, f, NULL); | |
0500813f | 256 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
257 | |
258 | return rv; | |
259 | } | |
260 | ||
261 | /** | |
262 | * drbd_force_state() - Impose a change which happens outside our control on our state | |
b30ab791 | 263 | * @device: DRBD device. |
b8907339 PR |
264 | * @mask: mask of state bits to change. |
265 | * @val: value of new state bits. | |
266 | */ | |
b30ab791 | 267 | void drbd_force_state(struct drbd_device *device, |
b8907339 PR |
268 | union drbd_state mask, union drbd_state val) |
269 | { | |
b30ab791 | 270 | drbd_change_state(device, CS_HARD, mask, val); |
b8907339 PR |
271 | } |
272 | ||
273 | static enum drbd_state_rv | |
b30ab791 | 274 | _req_st_cond(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
275 | union drbd_state val) |
276 | { | |
277 | union drbd_state os, ns; | |
278 | unsigned long flags; | |
279 | enum drbd_state_rv rv; | |
280 | ||
b30ab791 | 281 | if (test_and_clear_bit(CL_ST_CHG_SUCCESS, &device->flags)) |
b8907339 PR |
282 | return SS_CW_SUCCESS; |
283 | ||
b30ab791 | 284 | if (test_and_clear_bit(CL_ST_CHG_FAIL, &device->flags)) |
b8907339 PR |
285 | return SS_CW_FAILED_BY_PEER; |
286 | ||
0500813f | 287 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
288 | os = drbd_read_state(device); |
289 | ns = sanitize_state(device, apply_mask_val(os, mask, val), NULL); | |
3509502d | 290 | rv = is_valid_transition(os, ns); |
a3025a27 | 291 | if (rv >= SS_SUCCESS) |
3509502d | 292 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ |
b8907339 | 293 | |
b30ab791 | 294 | if (!cl_wide_st_chg(device, os, ns)) |
b8907339 | 295 | rv = SS_CW_NO_NEED; |
3509502d | 296 | if (rv == SS_UNKNOWN_ERROR) { |
b30ab791 | 297 | rv = is_valid_state(device, ns); |
a3025a27 | 298 | if (rv >= SS_SUCCESS) { |
a6b32bc3 | 299 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
a3025a27 | 300 | if (rv >= SS_SUCCESS) |
b8907339 PR |
301 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ |
302 | } | |
303 | } | |
0500813f | 304 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
305 | |
306 | return rv; | |
307 | } | |
308 | ||
309 | /** | |
310 | * drbd_req_state() - Perform an eventually cluster wide state change | |
b30ab791 | 311 | * @device: DRBD device. |
b8907339 PR |
312 | * @mask: mask of state bits to change. |
313 | * @val: value of new state bits. | |
314 | * @f: flags | |
315 | * | |
316 | * Should not be called directly, use drbd_request_state() or | |
317 | * _drbd_request_state(). | |
318 | */ | |
319 | static enum drbd_state_rv | |
b30ab791 | 320 | drbd_req_state(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
321 | union drbd_state val, enum chg_state_flags f) |
322 | { | |
323 | struct completion done; | |
324 | unsigned long flags; | |
325 | union drbd_state os, ns; | |
326 | enum drbd_state_rv rv; | |
327 | ||
328 | init_completion(&done); | |
329 | ||
330 | if (f & CS_SERIALIZE) | |
b30ab791 | 331 | mutex_lock(device->state_mutex); |
b8907339 | 332 | |
0500813f | 333 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
334 | os = drbd_read_state(device); |
335 | ns = sanitize_state(device, apply_mask_val(os, mask, val), NULL); | |
3509502d | 336 | rv = is_valid_transition(os, ns); |
3c5e5f6a | 337 | if (rv < SS_SUCCESS) { |
0500813f | 338 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
3509502d | 339 | goto abort; |
3c5e5f6a | 340 | } |
b8907339 | 341 | |
b30ab791 AG |
342 | if (cl_wide_st_chg(device, os, ns)) { |
343 | rv = is_valid_state(device, ns); | |
b8907339 | 344 | if (rv == SS_SUCCESS) |
a6b32bc3 | 345 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
0500813f | 346 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
347 | |
348 | if (rv < SS_SUCCESS) { | |
349 | if (f & CS_VERBOSE) | |
b30ab791 | 350 | print_st_err(device, os, ns, rv); |
b8907339 PR |
351 | goto abort; |
352 | } | |
353 | ||
69a22773 | 354 | if (drbd_send_state_req(first_peer_device(device), mask, val)) { |
b8907339 PR |
355 | rv = SS_CW_FAILED_BY_PEER; |
356 | if (f & CS_VERBOSE) | |
b30ab791 | 357 | print_st_err(device, os, ns, rv); |
b8907339 PR |
358 | goto abort; |
359 | } | |
360 | ||
b30ab791 AG |
361 | wait_event(device->state_wait, |
362 | (rv = _req_st_cond(device, mask, val))); | |
b8907339 PR |
363 | |
364 | if (rv < SS_SUCCESS) { | |
b8907339 | 365 | if (f & CS_VERBOSE) |
b30ab791 | 366 | print_st_err(device, os, ns, rv); |
b8907339 PR |
367 | goto abort; |
368 | } | |
0500813f | 369 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
370 | ns = apply_mask_val(drbd_read_state(device), mask, val); |
371 | rv = _drbd_set_state(device, ns, f, &done); | |
b8907339 | 372 | } else { |
b30ab791 | 373 | rv = _drbd_set_state(device, ns, f, &done); |
b8907339 PR |
374 | } |
375 | ||
0500813f | 376 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
377 | |
378 | if (f & CS_WAIT_COMPLETE && rv == SS_SUCCESS) { | |
0b0ba1ef | 379 | D_ASSERT(device, current != first_peer_device(device)->connection->worker.task); |
b8907339 PR |
380 | wait_for_completion(&done); |
381 | } | |
382 | ||
383 | abort: | |
384 | if (f & CS_SERIALIZE) | |
b30ab791 | 385 | mutex_unlock(device->state_mutex); |
b8907339 PR |
386 | |
387 | return rv; | |
388 | } | |
389 | ||
390 | /** | |
391 | * _drbd_request_state() - Request a state change (with flags) | |
b30ab791 | 392 | * @device: DRBD device. |
b8907339 PR |
393 | * @mask: mask of state bits to change. |
394 | * @val: value of new state bits. | |
395 | * @f: flags | |
396 | * | |
397 | * Cousin of drbd_request_state(), useful with the CS_WAIT_COMPLETE | |
398 | * flag, or when logging of failed state change requests is not desired. | |
399 | */ | |
400 | enum drbd_state_rv | |
b30ab791 | 401 | _drbd_request_state(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
402 | union drbd_state val, enum chg_state_flags f) |
403 | { | |
404 | enum drbd_state_rv rv; | |
405 | ||
b30ab791 AG |
406 | wait_event(device->state_wait, |
407 | (rv = drbd_req_state(device, mask, val, f)) != SS_IN_TRANSIENT_STATE); | |
b8907339 PR |
408 | |
409 | return rv; | |
410 | } | |
411 | ||
b30ab791 | 412 | static void print_st(struct drbd_device *device, char *name, union drbd_state ns) |
b8907339 | 413 | { |
d0180171 | 414 | drbd_err(device, " %s = { cs:%s ro:%s/%s ds:%s/%s %c%c%c%c%c%c }\n", |
b8907339 PR |
415 | name, |
416 | drbd_conn_str(ns.conn), | |
417 | drbd_role_str(ns.role), | |
418 | drbd_role_str(ns.peer), | |
419 | drbd_disk_str(ns.disk), | |
420 | drbd_disk_str(ns.pdsk), | |
421 | is_susp(ns) ? 's' : 'r', | |
422 | ns.aftr_isp ? 'a' : '-', | |
423 | ns.peer_isp ? 'p' : '-', | |
424 | ns.user_isp ? 'u' : '-', | |
425 | ns.susp_fen ? 'F' : '-', | |
426 | ns.susp_nod ? 'N' : '-' | |
427 | ); | |
428 | } | |
429 | ||
b30ab791 | 430 | void print_st_err(struct drbd_device *device, union drbd_state os, |
b8907339 PR |
431 | union drbd_state ns, enum drbd_state_rv err) |
432 | { | |
433 | if (err == SS_IN_TRANSIENT_STATE) | |
434 | return; | |
d0180171 | 435 | drbd_err(device, "State change failed: %s\n", drbd_set_st_err_str(err)); |
b30ab791 AG |
436 | print_st(device, " state", os); |
437 | print_st(device, "wanted", ns); | |
b8907339 PR |
438 | } |
439 | ||
435693e8 | 440 | static long print_state_change(char *pb, union drbd_state os, union drbd_state ns, |
bbeb641c PR |
441 | enum chg_state_flags flags) |
442 | { | |
435693e8 | 443 | char *pbp; |
bbeb641c PR |
444 | pbp = pb; |
445 | *pbp = 0; | |
706cb24c | 446 | |
435693e8 | 447 | if (ns.role != os.role && flags & CS_DC_ROLE) |
bbeb641c PR |
448 | pbp += sprintf(pbp, "role( %s -> %s ) ", |
449 | drbd_role_str(os.role), | |
450 | drbd_role_str(ns.role)); | |
435693e8 | 451 | if (ns.peer != os.peer && flags & CS_DC_PEER) |
bbeb641c PR |
452 | pbp += sprintf(pbp, "peer( %s -> %s ) ", |
453 | drbd_role_str(os.peer), | |
454 | drbd_role_str(ns.peer)); | |
435693e8 | 455 | if (ns.conn != os.conn && flags & CS_DC_CONN) |
bbeb641c PR |
456 | pbp += sprintf(pbp, "conn( %s -> %s ) ", |
457 | drbd_conn_str(os.conn), | |
458 | drbd_conn_str(ns.conn)); | |
435693e8 | 459 | if (ns.disk != os.disk && flags & CS_DC_DISK) |
bbeb641c PR |
460 | pbp += sprintf(pbp, "disk( %s -> %s ) ", |
461 | drbd_disk_str(os.disk), | |
462 | drbd_disk_str(ns.disk)); | |
435693e8 | 463 | if (ns.pdsk != os.pdsk && flags & CS_DC_PDSK) |
bbeb641c PR |
464 | pbp += sprintf(pbp, "pdsk( %s -> %s ) ", |
465 | drbd_disk_str(os.pdsk), | |
466 | drbd_disk_str(ns.pdsk)); | |
706cb24c PR |
467 | |
468 | return pbp - pb; | |
469 | } | |
470 | ||
b30ab791 | 471 | static void drbd_pr_state_change(struct drbd_device *device, union drbd_state os, union drbd_state ns, |
706cb24c PR |
472 | enum chg_state_flags flags) |
473 | { | |
474 | char pb[300]; | |
475 | char *pbp = pb; | |
476 | ||
477 | pbp += print_state_change(pbp, os, ns, flags ^ CS_DC_MASK); | |
478 | ||
bbeb641c PR |
479 | if (ns.aftr_isp != os.aftr_isp) |
480 | pbp += sprintf(pbp, "aftr_isp( %d -> %d ) ", | |
481 | os.aftr_isp, | |
482 | ns.aftr_isp); | |
483 | if (ns.peer_isp != os.peer_isp) | |
484 | pbp += sprintf(pbp, "peer_isp( %d -> %d ) ", | |
485 | os.peer_isp, | |
486 | ns.peer_isp); | |
487 | if (ns.user_isp != os.user_isp) | |
488 | pbp += sprintf(pbp, "user_isp( %d -> %d ) ", | |
489 | os.user_isp, | |
490 | ns.user_isp); | |
435693e8 | 491 | |
706cb24c | 492 | if (pbp != pb) |
d0180171 | 493 | drbd_info(device, "%s\n", pb); |
bbeb641c | 494 | } |
b8907339 | 495 | |
bde89a9e | 496 | static void conn_pr_state_change(struct drbd_connection *connection, union drbd_state os, union drbd_state ns, |
435693e8 PR |
497 | enum chg_state_flags flags) |
498 | { | |
499 | char pb[300]; | |
706cb24c PR |
500 | char *pbp = pb; |
501 | ||
502 | pbp += print_state_change(pbp, os, ns, flags); | |
435693e8 | 503 | |
706cb24c PR |
504 | if (is_susp(ns) != is_susp(os) && flags & CS_DC_SUSP) |
505 | pbp += sprintf(pbp, "susp( %d -> %d ) ", | |
506 | is_susp(os), | |
507 | is_susp(ns)); | |
508 | ||
509 | if (pbp != pb) | |
1ec861eb | 510 | drbd_info(connection, "%s\n", pb); |
435693e8 PR |
511 | } |
512 | ||
513 | ||
b8907339 PR |
514 | /** |
515 | * is_valid_state() - Returns an SS_ error code if ns is not valid | |
b30ab791 | 516 | * @device: DRBD device. |
b8907339 PR |
517 | * @ns: State to consider. |
518 | */ | |
519 | static enum drbd_state_rv | |
b30ab791 | 520 | is_valid_state(struct drbd_device *device, union drbd_state ns) |
b8907339 PR |
521 | { |
522 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
523 | ||
524 | enum drbd_fencing_p fp; | |
525 | enum drbd_state_rv rv = SS_SUCCESS; | |
44ed167d | 526 | struct net_conf *nc; |
b8907339 | 527 | |
daeda1cc | 528 | rcu_read_lock(); |
b8907339 | 529 | fp = FP_DONT_CARE; |
b30ab791 AG |
530 | if (get_ldev(device)) { |
531 | fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
532 | put_ldev(device); | |
b8907339 PR |
533 | } |
534 | ||
a6b32bc3 | 535 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
536 | if (nc) { |
537 | if (!nc->two_primaries && ns.role == R_PRIMARY) { | |
047e95e2 PR |
538 | if (ns.peer == R_PRIMARY) |
539 | rv = SS_TWO_PRIMARIES; | |
a6b32bc3 | 540 | else if (conn_highest_peer(first_peer_device(device)->connection) == R_PRIMARY) |
047e95e2 | 541 | rv = SS_O_VOL_PEER_PRI; |
44ed167d | 542 | } |
b8907339 PR |
543 | } |
544 | ||
545 | if (rv <= 0) | |
546 | /* already found a reason to abort */; | |
b30ab791 | 547 | else if (ns.role == R_SECONDARY && device->open_cnt) |
b8907339 PR |
548 | rv = SS_DEVICE_IN_USE; |
549 | ||
550 | else if (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.disk < D_UP_TO_DATE) | |
551 | rv = SS_NO_UP_TO_DATE_DISK; | |
552 | ||
553 | else if (fp >= FP_RESOURCE && | |
554 | ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk >= D_UNKNOWN) | |
555 | rv = SS_PRIMARY_NOP; | |
556 | ||
557 | else if (ns.role == R_PRIMARY && ns.disk <= D_INCONSISTENT && ns.pdsk <= D_INCONSISTENT) | |
558 | rv = SS_NO_UP_TO_DATE_DISK; | |
559 | ||
560 | else if (ns.conn > C_CONNECTED && ns.disk < D_INCONSISTENT) | |
561 | rv = SS_NO_LOCAL_DISK; | |
562 | ||
563 | else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT) | |
564 | rv = SS_NO_REMOTE_DISK; | |
565 | ||
566 | else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) | |
567 | rv = SS_NO_UP_TO_DATE_DISK; | |
568 | ||
569 | else if ((ns.conn == C_CONNECTED || | |
570 | ns.conn == C_WF_BITMAP_S || | |
571 | ns.conn == C_SYNC_SOURCE || | |
572 | ns.conn == C_PAUSED_SYNC_S) && | |
573 | ns.disk == D_OUTDATED) | |
574 | rv = SS_CONNECTED_OUTDATES; | |
575 | ||
576 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
44ed167d | 577 | (nc->verify_alg[0] == 0)) |
b8907339 PR |
578 | rv = SS_NO_VERIFY_ALG; |
579 | ||
580 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
a6b32bc3 | 581 | first_peer_device(device)->connection->agreed_pro_version < 88) |
b8907339 PR |
582 | rv = SS_NOT_SUPPORTED; |
583 | ||
5c4f13d9 PR |
584 | else if (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) |
585 | rv = SS_NO_UP_TO_DATE_DISK; | |
586 | ||
587 | else if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
588 | ns.pdsk == D_UNKNOWN) | |
589 | rv = SS_NEED_CONNECTION; | |
590 | ||
b8907339 PR |
591 | else if (ns.conn >= C_CONNECTED && ns.pdsk == D_UNKNOWN) |
592 | rv = SS_CONNECTED_OUTDATES; | |
593 | ||
44ed167d PR |
594 | rcu_read_unlock(); |
595 | ||
b8907339 PR |
596 | return rv; |
597 | } | |
598 | ||
599 | /** | |
a75f34ad | 600 | * is_valid_soft_transition() - Returns an SS_ error code if the state transition is not possible |
3509502d PR |
601 | * This function limits state transitions that may be declined by DRBD. I.e. |
602 | * user requests (aka soft transitions). | |
b30ab791 | 603 | * @device: DRBD device. |
b8907339 PR |
604 | * @ns: new state. |
605 | * @os: old state. | |
606 | */ | |
607 | static enum drbd_state_rv | |
bde89a9e | 608 | is_valid_soft_transition(union drbd_state os, union drbd_state ns, struct drbd_connection *connection) |
b8907339 PR |
609 | { |
610 | enum drbd_state_rv rv = SS_SUCCESS; | |
611 | ||
612 | if ((ns.conn == C_STARTING_SYNC_T || ns.conn == C_STARTING_SYNC_S) && | |
613 | os.conn > C_CONNECTED) | |
614 | rv = SS_RESYNC_RUNNING; | |
615 | ||
616 | if (ns.conn == C_DISCONNECTING && os.conn == C_STANDALONE) | |
617 | rv = SS_ALREADY_STANDALONE; | |
618 | ||
619 | if (ns.disk > D_ATTACHING && os.disk == D_DISKLESS) | |
620 | rv = SS_IS_DISKLESS; | |
621 | ||
622 | if (ns.conn == C_WF_CONNECTION && os.conn < C_UNCONNECTED) | |
623 | rv = SS_NO_NET_CONFIG; | |
624 | ||
625 | if (ns.disk == D_OUTDATED && os.disk < D_OUTDATED && os.disk != D_ATTACHING) | |
626 | rv = SS_LOWER_THAN_OUTDATED; | |
627 | ||
628 | if (ns.conn == C_DISCONNECTING && os.conn == C_UNCONNECTED) | |
629 | rv = SS_IN_TRANSIENT_STATE; | |
630 | ||
2325eb66 PR |
631 | /* if (ns.conn == os.conn && ns.conn == C_WF_REPORT_PARAMS) |
632 | rv = SS_IN_TRANSIENT_STATE; */ | |
b8907339 | 633 | |
a1096a6e PR |
634 | /* While establishing a connection only allow cstate to change. |
635 | Delay/refuse role changes, detach attach etc... */ | |
bde89a9e | 636 | if (test_bit(STATE_SENT, &connection->flags) && |
a1096a6e PR |
637 | !(os.conn == C_WF_REPORT_PARAMS || |
638 | (ns.conn == C_WF_REPORT_PARAMS && os.conn == C_WF_CONNECTION))) | |
639 | rv = SS_IN_TRANSIENT_STATE; | |
640 | ||
b8907339 PR |
641 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED) |
642 | rv = SS_NEED_CONNECTION; | |
643 | ||
644 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
645 | ns.conn != os.conn && os.conn > C_CONNECTED) | |
646 | rv = SS_RESYNC_RUNNING; | |
647 | ||
648 | if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
649 | os.conn < C_CONNECTED) | |
650 | rv = SS_NEED_CONNECTION; | |
651 | ||
652 | if ((ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE) | |
653 | && os.conn < C_WF_REPORT_PARAMS) | |
654 | rv = SS_NEED_CONNECTION; /* No NetworkFailure -> SyncTarget etc... */ | |
655 | ||
2bd5ed5d PR |
656 | if (ns.conn == C_DISCONNECTING && ns.pdsk == D_OUTDATED && |
657 | os.conn < C_CONNECTED && os.pdsk > D_OUTDATED) | |
658 | rv = SS_OUTDATE_WO_CONN; | |
659 | ||
b8907339 PR |
660 | return rv; |
661 | } | |
662 | ||
3509502d | 663 | static enum drbd_state_rv |
fda74117 | 664 | is_valid_conn_transition(enum drbd_conns oc, enum drbd_conns nc) |
3509502d | 665 | { |
d9cc6e23 LE |
666 | /* no change -> nothing to do, at least for the connection part */ |
667 | if (oc == nc) | |
668 | return SS_NOTHING_TO_DO; | |
3509502d | 669 | |
d9cc6e23 LE |
670 | /* disconnect of an unconfigured connection does not make sense */ |
671 | if (oc == C_STANDALONE && nc == C_DISCONNECTING) | |
672 | return SS_ALREADY_STANDALONE; | |
673 | ||
674 | /* from C_STANDALONE, we start with C_UNCONNECTED */ | |
675 | if (oc == C_STANDALONE && nc != C_UNCONNECTED) | |
676 | return SS_NEED_CONNECTION; | |
3509502d | 677 | |
25b0d6c8 PR |
678 | /* When establishing a connection we need to go through WF_REPORT_PARAMS! |
679 | Necessary to do the right thing upon invalidate-remote on a disconnected resource */ | |
680 | if (oc < C_WF_REPORT_PARAMS && nc >= C_CONNECTED) | |
681 | return SS_NEED_CONNECTION; | |
682 | ||
3509502d | 683 | /* After a network error only C_UNCONNECTED or C_DISCONNECTING may follow. */ |
fda74117 | 684 | if (oc >= C_TIMEOUT && oc <= C_TEAR_DOWN && nc != C_UNCONNECTED && nc != C_DISCONNECTING) |
d9cc6e23 | 685 | return SS_IN_TRANSIENT_STATE; |
3509502d PR |
686 | |
687 | /* After C_DISCONNECTING only C_STANDALONE may follow */ | |
fda74117 | 688 | if (oc == C_DISCONNECTING && nc != C_STANDALONE) |
d9cc6e23 | 689 | return SS_IN_TRANSIENT_STATE; |
3509502d | 690 | |
d9cc6e23 | 691 | return SS_SUCCESS; |
fda74117 PR |
692 | } |
693 | ||
694 | ||
695 | /** | |
696 | * is_valid_transition() - Returns an SS_ error code if the state transition is not possible | |
697 | * This limits hard state transitions. Hard state transitions are facts there are | |
698 | * imposed on DRBD by the environment. E.g. disk broke or network broke down. | |
699 | * But those hard state transitions are still not allowed to do everything. | |
700 | * @ns: new state. | |
701 | * @os: old state. | |
702 | */ | |
703 | static enum drbd_state_rv | |
704 | is_valid_transition(union drbd_state os, union drbd_state ns) | |
705 | { | |
706 | enum drbd_state_rv rv; | |
707 | ||
708 | rv = is_valid_conn_transition(os.conn, ns.conn); | |
709 | ||
3509502d PR |
710 | /* we cannot fail (again) if we already detached */ |
711 | if (ns.disk == D_FAILED && os.disk == D_DISKLESS) | |
712 | rv = SS_IS_DISKLESS; | |
713 | ||
714 | return rv; | |
715 | } | |
716 | ||
b30ab791 | 717 | static void print_sanitize_warnings(struct drbd_device *device, enum sanitize_state_warnings warn) |
d942ae44 PR |
718 | { |
719 | static const char *msg_table[] = { | |
720 | [NO_WARNING] = "", | |
721 | [ABORTED_ONLINE_VERIFY] = "Online-verify aborted.", | |
722 | [ABORTED_RESYNC] = "Resync aborted.", | |
723 | [CONNECTION_LOST_NEGOTIATING] = "Connection lost while negotiating, no data!", | |
724 | [IMPLICITLY_UPGRADED_DISK] = "Implicitly upgraded disk", | |
725 | [IMPLICITLY_UPGRADED_PDSK] = "Implicitly upgraded pdsk", | |
726 | }; | |
727 | ||
728 | if (warn != NO_WARNING) | |
d0180171 | 729 | drbd_warn(device, "%s\n", msg_table[warn]); |
d942ae44 PR |
730 | } |
731 | ||
b8907339 PR |
732 | /** |
733 | * sanitize_state() - Resolves implicitly necessary additional changes to a state transition | |
b30ab791 | 734 | * @device: DRBD device. |
b8907339 PR |
735 | * @os: old state. |
736 | * @ns: new state. | |
737 | * @warn_sync_abort: | |
738 | * | |
739 | * When we loose connection, we have to set the state of the peers disk (pdsk) | |
740 | * to D_UNKNOWN. This rule and many more along those lines are in this function. | |
741 | */ | |
b30ab791 | 742 | static union drbd_state sanitize_state(struct drbd_device *device, union drbd_state ns, |
d942ae44 | 743 | enum sanitize_state_warnings *warn) |
b8907339 PR |
744 | { |
745 | enum drbd_fencing_p fp; | |
746 | enum drbd_disk_state disk_min, disk_max, pdsk_min, pdsk_max; | |
747 | ||
d942ae44 PR |
748 | if (warn) |
749 | *warn = NO_WARNING; | |
750 | ||
b8907339 | 751 | fp = FP_DONT_CARE; |
b30ab791 | 752 | if (get_ldev(device)) { |
daeda1cc | 753 | rcu_read_lock(); |
b30ab791 | 754 | fp = rcu_dereference(device->ldev->disk_conf)->fencing; |
daeda1cc | 755 | rcu_read_unlock(); |
b30ab791 | 756 | put_ldev(device); |
b8907339 PR |
757 | } |
758 | ||
3509502d | 759 | /* Implications from connection to peer and peer_isp */ |
b8907339 PR |
760 | if (ns.conn < C_CONNECTED) { |
761 | ns.peer_isp = 0; | |
762 | ns.peer = R_UNKNOWN; | |
763 | if (ns.pdsk > D_UNKNOWN || ns.pdsk < D_INCONSISTENT) | |
764 | ns.pdsk = D_UNKNOWN; | |
765 | } | |
766 | ||
767 | /* Clear the aftr_isp when becoming unconfigured */ | |
768 | if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY) | |
769 | ns.aftr_isp = 0; | |
770 | ||
4308a0a3 | 771 | /* An implication of the disk states onto the connection state */ |
b8907339 | 772 | /* Abort resync if a disk fails/detaches */ |
4308a0a3 | 773 | if (ns.conn > C_CONNECTED && (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) { |
d942ae44 PR |
774 | if (warn) |
775 | *warn = ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T ? | |
776 | ABORTED_ONLINE_VERIFY : ABORTED_RESYNC; | |
b8907339 PR |
777 | ns.conn = C_CONNECTED; |
778 | } | |
779 | ||
780 | /* Connection breaks down before we finished "Negotiating" */ | |
781 | if (ns.conn < C_CONNECTED && ns.disk == D_NEGOTIATING && | |
b30ab791 AG |
782 | get_ldev_if_state(device, D_NEGOTIATING)) { |
783 | if (device->ed_uuid == device->ldev->md.uuid[UI_CURRENT]) { | |
784 | ns.disk = device->new_state_tmp.disk; | |
785 | ns.pdsk = device->new_state_tmp.pdsk; | |
b8907339 | 786 | } else { |
d942ae44 PR |
787 | if (warn) |
788 | *warn = CONNECTION_LOST_NEGOTIATING; | |
b8907339 PR |
789 | ns.disk = D_DISKLESS; |
790 | ns.pdsk = D_UNKNOWN; | |
791 | } | |
b30ab791 | 792 | put_ldev(device); |
b8907339 PR |
793 | } |
794 | ||
795 | /* D_CONSISTENT and D_OUTDATED vanish when we get connected */ | |
796 | if (ns.conn >= C_CONNECTED && ns.conn < C_AHEAD) { | |
797 | if (ns.disk == D_CONSISTENT || ns.disk == D_OUTDATED) | |
798 | ns.disk = D_UP_TO_DATE; | |
799 | if (ns.pdsk == D_CONSISTENT || ns.pdsk == D_OUTDATED) | |
800 | ns.pdsk = D_UP_TO_DATE; | |
801 | } | |
802 | ||
803 | /* Implications of the connection stat on the disk states */ | |
804 | disk_min = D_DISKLESS; | |
805 | disk_max = D_UP_TO_DATE; | |
806 | pdsk_min = D_INCONSISTENT; | |
807 | pdsk_max = D_UNKNOWN; | |
808 | switch ((enum drbd_conns)ns.conn) { | |
809 | case C_WF_BITMAP_T: | |
810 | case C_PAUSED_SYNC_T: | |
811 | case C_STARTING_SYNC_T: | |
812 | case C_WF_SYNC_UUID: | |
813 | case C_BEHIND: | |
814 | disk_min = D_INCONSISTENT; | |
815 | disk_max = D_OUTDATED; | |
816 | pdsk_min = D_UP_TO_DATE; | |
817 | pdsk_max = D_UP_TO_DATE; | |
818 | break; | |
819 | case C_VERIFY_S: | |
820 | case C_VERIFY_T: | |
821 | disk_min = D_UP_TO_DATE; | |
822 | disk_max = D_UP_TO_DATE; | |
823 | pdsk_min = D_UP_TO_DATE; | |
824 | pdsk_max = D_UP_TO_DATE; | |
825 | break; | |
826 | case C_CONNECTED: | |
827 | disk_min = D_DISKLESS; | |
828 | disk_max = D_UP_TO_DATE; | |
829 | pdsk_min = D_DISKLESS; | |
830 | pdsk_max = D_UP_TO_DATE; | |
831 | break; | |
832 | case C_WF_BITMAP_S: | |
833 | case C_PAUSED_SYNC_S: | |
834 | case C_STARTING_SYNC_S: | |
835 | case C_AHEAD: | |
836 | disk_min = D_UP_TO_DATE; | |
837 | disk_max = D_UP_TO_DATE; | |
838 | pdsk_min = D_INCONSISTENT; | |
839 | pdsk_max = D_CONSISTENT; /* D_OUTDATED would be nice. But explicit outdate necessary*/ | |
840 | break; | |
841 | case C_SYNC_TARGET: | |
842 | disk_min = D_INCONSISTENT; | |
843 | disk_max = D_INCONSISTENT; | |
844 | pdsk_min = D_UP_TO_DATE; | |
845 | pdsk_max = D_UP_TO_DATE; | |
846 | break; | |
847 | case C_SYNC_SOURCE: | |
848 | disk_min = D_UP_TO_DATE; | |
849 | disk_max = D_UP_TO_DATE; | |
850 | pdsk_min = D_INCONSISTENT; | |
851 | pdsk_max = D_INCONSISTENT; | |
852 | break; | |
853 | case C_STANDALONE: | |
854 | case C_DISCONNECTING: | |
855 | case C_UNCONNECTED: | |
856 | case C_TIMEOUT: | |
857 | case C_BROKEN_PIPE: | |
858 | case C_NETWORK_FAILURE: | |
859 | case C_PROTOCOL_ERROR: | |
860 | case C_TEAR_DOWN: | |
861 | case C_WF_CONNECTION: | |
862 | case C_WF_REPORT_PARAMS: | |
863 | case C_MASK: | |
864 | break; | |
865 | } | |
866 | if (ns.disk > disk_max) | |
867 | ns.disk = disk_max; | |
868 | ||
869 | if (ns.disk < disk_min) { | |
d942ae44 PR |
870 | if (warn) |
871 | *warn = IMPLICITLY_UPGRADED_DISK; | |
b8907339 PR |
872 | ns.disk = disk_min; |
873 | } | |
874 | if (ns.pdsk > pdsk_max) | |
875 | ns.pdsk = pdsk_max; | |
876 | ||
877 | if (ns.pdsk < pdsk_min) { | |
d942ae44 PR |
878 | if (warn) |
879 | *warn = IMPLICITLY_UPGRADED_PDSK; | |
b8907339 PR |
880 | ns.pdsk = pdsk_min; |
881 | } | |
882 | ||
883 | if (fp == FP_STONITH && | |
4308a0a3 | 884 | (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED)) |
b8907339 PR |
885 | ns.susp_fen = 1; /* Suspend IO while fence-peer handler runs (peer lost) */ |
886 | ||
eb6bea67 | 887 | if (device->resource->res_opts.on_no_data == OND_SUSPEND_IO && |
4308a0a3 | 888 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) |
b8907339 PR |
889 | ns.susp_nod = 1; /* Suspend IO while no data available (no accessible data available) */ |
890 | ||
891 | if (ns.aftr_isp || ns.peer_isp || ns.user_isp) { | |
892 | if (ns.conn == C_SYNC_SOURCE) | |
893 | ns.conn = C_PAUSED_SYNC_S; | |
894 | if (ns.conn == C_SYNC_TARGET) | |
895 | ns.conn = C_PAUSED_SYNC_T; | |
896 | } else { | |
897 | if (ns.conn == C_PAUSED_SYNC_S) | |
898 | ns.conn = C_SYNC_SOURCE; | |
899 | if (ns.conn == C_PAUSED_SYNC_T) | |
900 | ns.conn = C_SYNC_TARGET; | |
901 | } | |
902 | ||
903 | return ns; | |
904 | } | |
905 | ||
b30ab791 | 906 | void drbd_resume_al(struct drbd_device *device) |
b8907339 | 907 | { |
b30ab791 | 908 | if (test_and_clear_bit(AL_SUSPENDED, &device->flags)) |
d0180171 | 909 | drbd_info(device, "Resumed AL updates\n"); |
b8907339 PR |
910 | } |
911 | ||
912 | /* helper for __drbd_set_state */ | |
b30ab791 | 913 | static void set_ov_position(struct drbd_device *device, enum drbd_conns cs) |
b8907339 | 914 | { |
a6b32bc3 | 915 | if (first_peer_device(device)->connection->agreed_pro_version < 90) |
b30ab791 AG |
916 | device->ov_start_sector = 0; |
917 | device->rs_total = drbd_bm_bits(device); | |
918 | device->ov_position = 0; | |
b8907339 PR |
919 | if (cs == C_VERIFY_T) { |
920 | /* starting online verify from an arbitrary position | |
921 | * does not fit well into the existing protocol. | |
922 | * on C_VERIFY_T, we initialize ov_left and friends | |
923 | * implicitly in receive_DataRequest once the | |
924 | * first P_OV_REQUEST is received */ | |
b30ab791 | 925 | device->ov_start_sector = ~(sector_t)0; |
b8907339 | 926 | } else { |
b30ab791 AG |
927 | unsigned long bit = BM_SECT_TO_BIT(device->ov_start_sector); |
928 | if (bit >= device->rs_total) { | |
929 | device->ov_start_sector = | |
930 | BM_BIT_TO_SECT(device->rs_total - 1); | |
931 | device->rs_total = 1; | |
b8907339 | 932 | } else |
b30ab791 AG |
933 | device->rs_total -= bit; |
934 | device->ov_position = device->ov_start_sector; | |
b8907339 | 935 | } |
b30ab791 | 936 | device->ov_left = device->rs_total; |
b8907339 PR |
937 | } |
938 | ||
939 | /** | |
940 | * __drbd_set_state() - Set a new DRBD state | |
b30ab791 | 941 | * @device: DRBD device. |
b8907339 PR |
942 | * @ns: new state. |
943 | * @flags: Flags | |
944 | * @done: Optional completion, that will get completed after the after_state_ch() finished | |
945 | * | |
946 | * Caller needs to hold req_lock, and global_state_lock. Do not call directly. | |
947 | */ | |
948 | enum drbd_state_rv | |
b30ab791 | 949 | __drbd_set_state(struct drbd_device *device, union drbd_state ns, |
b8907339 PR |
950 | enum chg_state_flags flags, struct completion *done) |
951 | { | |
952 | union drbd_state os; | |
953 | enum drbd_state_rv rv = SS_SUCCESS; | |
d942ae44 | 954 | enum sanitize_state_warnings ssw; |
b8907339 | 955 | struct after_state_chg_work *ascw; |
2681f7f6 | 956 | bool did_remote, should_do_remote; |
b8907339 | 957 | |
b30ab791 | 958 | os = drbd_read_state(device); |
b8907339 | 959 | |
b30ab791 | 960 | ns = sanitize_state(device, ns, &ssw); |
b8907339 PR |
961 | if (ns.i == os.i) |
962 | return SS_NOTHING_TO_DO; | |
963 | ||
3509502d PR |
964 | rv = is_valid_transition(os, ns); |
965 | if (rv < SS_SUCCESS) | |
966 | return rv; | |
967 | ||
b8907339 PR |
968 | if (!(flags & CS_HARD)) { |
969 | /* pre-state-change checks ; only look at ns */ | |
970 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
971 | ||
b30ab791 | 972 | rv = is_valid_state(device, ns); |
b8907339 PR |
973 | if (rv < SS_SUCCESS) { |
974 | /* If the old state was illegal as well, then let | |
975 | this happen...*/ | |
976 | ||
b30ab791 | 977 | if (is_valid_state(device, os) == rv) |
a6b32bc3 | 978 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
b8907339 | 979 | } else |
a6b32bc3 | 980 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
b8907339 PR |
981 | } |
982 | ||
983 | if (rv < SS_SUCCESS) { | |
984 | if (flags & CS_VERBOSE) | |
b30ab791 | 985 | print_st_err(device, os, ns, rv); |
b8907339 PR |
986 | return rv; |
987 | } | |
988 | ||
b30ab791 | 989 | print_sanitize_warnings(device, ssw); |
b8907339 | 990 | |
b30ab791 | 991 | drbd_pr_state_change(device, os, ns, flags); |
b8907339 | 992 | |
706cb24c PR |
993 | /* Display changes to the susp* flags that where caused by the call to |
994 | sanitize_state(). Only display it here if we where not called from | |
995 | _conn_request_state() */ | |
996 | if (!(flags & CS_DC_SUSP)) | |
a6b32bc3 AG |
997 | conn_pr_state_change(first_peer_device(device)->connection, os, ns, |
998 | (flags & ~CS_DC_MASK) | CS_DC_SUSP); | |
706cb24c | 999 | |
b8907339 PR |
1000 | /* if we are going -> D_FAILED or D_DISKLESS, grab one extra reference |
1001 | * on the ldev here, to be sure the transition -> D_DISKLESS resp. | |
1002 | * drbd_ldev_destroy() won't happen before our corresponding | |
1003 | * after_state_ch works run, where we put_ldev again. */ | |
1004 | if ((os.disk != D_FAILED && ns.disk == D_FAILED) || | |
1005 | (os.disk != D_DISKLESS && ns.disk == D_DISKLESS)) | |
b30ab791 | 1006 | atomic_inc(&device->local_cnt); |
b8907339 | 1007 | |
b30ab791 AG |
1008 | did_remote = drbd_should_do_remote(device->state); |
1009 | device->state.i = ns.i; | |
1010 | should_do_remote = drbd_should_do_remote(device->state); | |
6bbf53ca AG |
1011 | device->resource->susp = ns.susp; |
1012 | device->resource->susp_nod = ns.susp_nod; | |
1013 | device->resource->susp_fen = ns.susp_fen; | |
b8907339 | 1014 | |
2681f7f6 LE |
1015 | /* put replicated vs not-replicated requests in seperate epochs */ |
1016 | if (did_remote != should_do_remote) | |
a6b32bc3 | 1017 | start_new_tl_epoch(first_peer_device(device)->connection); |
2681f7f6 | 1018 | |
b8907339 | 1019 | if (os.disk == D_ATTACHING && ns.disk >= D_NEGOTIATING) |
b30ab791 | 1020 | drbd_print_uuids(device, "attached to UUIDs"); |
b8907339 | 1021 | |
79702011 PR |
1022 | /* Wake up role changes, that were delayed because of connection establishing */ |
1023 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn != C_WF_REPORT_PARAMS && | |
a6b32bc3 AG |
1024 | no_peer_wf_report_params(first_peer_device(device)->connection)) |
1025 | clear_bit(STATE_SENT, &first_peer_device(device)->connection->flags); | |
79702011 | 1026 | |
b30ab791 AG |
1027 | wake_up(&device->misc_wait); |
1028 | wake_up(&device->state_wait); | |
a6b32bc3 | 1029 | wake_up(&first_peer_device(device)->connection->ping_wait); |
b8907339 | 1030 | |
58ffa580 LE |
1031 | /* Aborted verify run, or we reached the stop sector. |
1032 | * Log the last position, unless end-of-device. */ | |
b8907339 | 1033 | if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) && |
58ffa580 | 1034 | ns.conn <= C_CONNECTED) { |
b30ab791 AG |
1035 | device->ov_start_sector = |
1036 | BM_BIT_TO_SECT(drbd_bm_bits(device) - device->ov_left); | |
1037 | if (device->ov_left) | |
d0180171 | 1038 | drbd_info(device, "Online Verify reached sector %llu\n", |
b30ab791 | 1039 | (unsigned long long)device->ov_start_sector); |
b8907339 PR |
1040 | } |
1041 | ||
1042 | if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) && | |
1043 | (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) { | |
d0180171 | 1044 | drbd_info(device, "Syncer continues.\n"); |
b30ab791 AG |
1045 | device->rs_paused += (long)jiffies |
1046 | -(long)device->rs_mark_time[device->rs_last_mark]; | |
b8907339 | 1047 | if (ns.conn == C_SYNC_TARGET) |
b30ab791 | 1048 | mod_timer(&device->resync_timer, jiffies); |
b8907339 PR |
1049 | } |
1050 | ||
1051 | if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) && | |
1052 | (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) { | |
d0180171 | 1053 | drbd_info(device, "Resync suspended\n"); |
b30ab791 | 1054 | device->rs_mark_time[device->rs_last_mark] = jiffies; |
b8907339 PR |
1055 | } |
1056 | ||
1057 | if (os.conn == C_CONNECTED && | |
1058 | (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) { | |
1059 | unsigned long now = jiffies; | |
1060 | int i; | |
1061 | ||
b30ab791 AG |
1062 | set_ov_position(device, ns.conn); |
1063 | device->rs_start = now; | |
1064 | device->rs_last_events = 0; | |
1065 | device->rs_last_sect_ev = 0; | |
1066 | device->ov_last_oos_size = 0; | |
1067 | device->ov_last_oos_start = 0; | |
b8907339 PR |
1068 | |
1069 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { | |
b30ab791 AG |
1070 | device->rs_mark_left[i] = device->ov_left; |
1071 | device->rs_mark_time[i] = now; | |
b8907339 PR |
1072 | } |
1073 | ||
b30ab791 | 1074 | drbd_rs_controller_reset(device); |
b8907339 PR |
1075 | |
1076 | if (ns.conn == C_VERIFY_S) { | |
d0180171 | 1077 | drbd_info(device, "Starting Online Verify from sector %llu\n", |
b30ab791 AG |
1078 | (unsigned long long)device->ov_position); |
1079 | mod_timer(&device->resync_timer, jiffies); | |
b8907339 PR |
1080 | } |
1081 | } | |
1082 | ||
b30ab791 AG |
1083 | if (get_ldev(device)) { |
1084 | u32 mdf = device->ldev->md.flags & ~(MDF_CONSISTENT|MDF_PRIMARY_IND| | |
b8907339 PR |
1085 | MDF_CONNECTED_IND|MDF_WAS_UP_TO_DATE| |
1086 | MDF_PEER_OUT_DATED|MDF_CRASHED_PRIMARY); | |
1087 | ||
d5d7ebd4 | 1088 | mdf &= ~MDF_AL_CLEAN; |
b30ab791 | 1089 | if (test_bit(CRASHED_PRIMARY, &device->flags)) |
b8907339 | 1090 | mdf |= MDF_CRASHED_PRIMARY; |
b30ab791 AG |
1091 | if (device->state.role == R_PRIMARY || |
1092 | (device->state.pdsk < D_INCONSISTENT && device->state.peer == R_PRIMARY)) | |
b8907339 | 1093 | mdf |= MDF_PRIMARY_IND; |
b30ab791 | 1094 | if (device->state.conn > C_WF_REPORT_PARAMS) |
b8907339 | 1095 | mdf |= MDF_CONNECTED_IND; |
b30ab791 | 1096 | if (device->state.disk > D_INCONSISTENT) |
b8907339 | 1097 | mdf |= MDF_CONSISTENT; |
b30ab791 | 1098 | if (device->state.disk > D_OUTDATED) |
b8907339 | 1099 | mdf |= MDF_WAS_UP_TO_DATE; |
b30ab791 | 1100 | if (device->state.pdsk <= D_OUTDATED && device->state.pdsk >= D_INCONSISTENT) |
b8907339 | 1101 | mdf |= MDF_PEER_OUT_DATED; |
b30ab791 AG |
1102 | if (mdf != device->ldev->md.flags) { |
1103 | device->ldev->md.flags = mdf; | |
1104 | drbd_md_mark_dirty(device); | |
b8907339 PR |
1105 | } |
1106 | if (os.disk < D_CONSISTENT && ns.disk >= D_CONSISTENT) | |
b30ab791 AG |
1107 | drbd_set_ed_uuid(device, device->ldev->md.uuid[UI_CURRENT]); |
1108 | put_ldev(device); | |
b8907339 PR |
1109 | } |
1110 | ||
1111 | /* Peer was forced D_UP_TO_DATE & R_PRIMARY, consider to resync */ | |
1112 | if (os.disk == D_INCONSISTENT && os.pdsk == D_INCONSISTENT && | |
1113 | os.peer == R_SECONDARY && ns.peer == R_PRIMARY) | |
b30ab791 | 1114 | set_bit(CONSIDER_RESYNC, &device->flags); |
b8907339 PR |
1115 | |
1116 | /* Receiver should clean up itself */ | |
1117 | if (os.conn != C_DISCONNECTING && ns.conn == C_DISCONNECTING) | |
a6b32bc3 | 1118 | drbd_thread_stop_nowait(&first_peer_device(device)->connection->receiver); |
b8907339 PR |
1119 | |
1120 | /* Now the receiver finished cleaning up itself, it should die */ | |
1121 | if (os.conn != C_STANDALONE && ns.conn == C_STANDALONE) | |
a6b32bc3 | 1122 | drbd_thread_stop_nowait(&first_peer_device(device)->connection->receiver); |
b8907339 PR |
1123 | |
1124 | /* Upon network failure, we need to restart the receiver. */ | |
823bd832 | 1125 | if (os.conn > C_WF_CONNECTION && |
b8907339 | 1126 | ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT) |
a6b32bc3 | 1127 | drbd_thread_restart_nowait(&first_peer_device(device)->connection->receiver); |
b8907339 PR |
1128 | |
1129 | /* Resume AL writing if we get a connection */ | |
28e448bb | 1130 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { |
b30ab791 | 1131 | drbd_resume_al(device); |
a6b32bc3 | 1132 | first_peer_device(device)->connection->connect_cnt++; |
28e448bb | 1133 | } |
b8907339 | 1134 | |
07be15b1 LE |
1135 | /* remember last attach time so request_timer_fn() won't |
1136 | * kill newly established sessions while we are still trying to thaw | |
1137 | * previously frozen IO */ | |
1138 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && | |
1139 | ns.disk > D_NEGOTIATING) | |
b30ab791 | 1140 | device->last_reattach_jif = jiffies; |
07be15b1 | 1141 | |
b8907339 PR |
1142 | ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC); |
1143 | if (ascw) { | |
1144 | ascw->os = os; | |
1145 | ascw->ns = ns; | |
1146 | ascw->flags = flags; | |
1147 | ascw->w.cb = w_after_state_ch; | |
b30ab791 | 1148 | ascw->w.device = device; |
b8907339 | 1149 | ascw->done = done; |
a6b32bc3 | 1150 | drbd_queue_work(&first_peer_device(device)->connection->sender_work, &ascw->w); |
b8907339 | 1151 | } else { |
d0180171 | 1152 | drbd_err(device, "Could not kmalloc an ascw\n"); |
b8907339 PR |
1153 | } |
1154 | ||
1155 | return rv; | |
1156 | } | |
1157 | ||
99920dc5 | 1158 | static int w_after_state_ch(struct drbd_work *w, int unused) |
b8907339 PR |
1159 | { |
1160 | struct after_state_chg_work *ascw = | |
1161 | container_of(w, struct after_state_chg_work, w); | |
b30ab791 | 1162 | struct drbd_device *device = w->device; |
b8907339 | 1163 | |
b30ab791 | 1164 | after_state_ch(device, ascw->os, ascw->ns, ascw->flags); |
b8907339 | 1165 | if (ascw->flags & CS_WAIT_COMPLETE) { |
0b0ba1ef | 1166 | D_ASSERT(device, ascw->done != NULL); |
b8907339 PR |
1167 | complete(ascw->done); |
1168 | } | |
1169 | kfree(ascw); | |
1170 | ||
99920dc5 | 1171 | return 0; |
b8907339 PR |
1172 | } |
1173 | ||
b30ab791 | 1174 | static void abw_start_sync(struct drbd_device *device, int rv) |
b8907339 PR |
1175 | { |
1176 | if (rv) { | |
d0180171 | 1177 | drbd_err(device, "Writing the bitmap failed not starting resync.\n"); |
b30ab791 | 1178 | _drbd_request_state(device, NS(conn, C_CONNECTED), CS_VERBOSE); |
b8907339 PR |
1179 | return; |
1180 | } | |
1181 | ||
b30ab791 | 1182 | switch (device->state.conn) { |
b8907339 | 1183 | case C_STARTING_SYNC_T: |
b30ab791 | 1184 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); |
b8907339 PR |
1185 | break; |
1186 | case C_STARTING_SYNC_S: | |
b30ab791 | 1187 | drbd_start_resync(device, C_SYNC_SOURCE); |
b8907339 PR |
1188 | break; |
1189 | } | |
1190 | } | |
1191 | ||
b30ab791 | 1192 | int drbd_bitmap_io_from_worker(struct drbd_device *device, |
54761697 | 1193 | int (*io_fn)(struct drbd_device *), |
b8907339 PR |
1194 | char *why, enum bm_flag flags) |
1195 | { | |
1196 | int rv; | |
1197 | ||
0b0ba1ef | 1198 | D_ASSERT(device, current == first_peer_device(device)->connection->worker.task); |
b8907339 | 1199 | |
b30ab791 AG |
1200 | /* open coded non-blocking drbd_suspend_io(device); */ |
1201 | set_bit(SUSPEND_IO, &device->flags); | |
b8907339 | 1202 | |
b30ab791 AG |
1203 | drbd_bm_lock(device, why, flags); |
1204 | rv = io_fn(device); | |
1205 | drbd_bm_unlock(device); | |
b8907339 | 1206 | |
b30ab791 | 1207 | drbd_resume_io(device); |
b8907339 PR |
1208 | |
1209 | return rv; | |
1210 | } | |
1211 | ||
1212 | /** | |
1213 | * after_state_ch() - Perform after state change actions that may sleep | |
b30ab791 | 1214 | * @device: DRBD device. |
b8907339 PR |
1215 | * @os: old state. |
1216 | * @ns: new state. | |
1217 | * @flags: Flags | |
1218 | */ | |
b30ab791 | 1219 | static void after_state_ch(struct drbd_device *device, union drbd_state os, |
b8907339 PR |
1220 | union drbd_state ns, enum chg_state_flags flags) |
1221 | { | |
6bbf53ca | 1222 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
1223 | struct sib_info sib; |
1224 | ||
1225 | sib.sib_reason = SIB_STATE_CHANGE; | |
1226 | sib.os = os; | |
1227 | sib.ns = ns; | |
b8907339 PR |
1228 | |
1229 | if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) { | |
b30ab791 AG |
1230 | clear_bit(CRASHED_PRIMARY, &device->flags); |
1231 | if (device->p_uuid) | |
1232 | device->p_uuid[UI_FLAGS] &= ~((u64)2); | |
b8907339 PR |
1233 | } |
1234 | ||
b8907339 | 1235 | /* Inform userspace about the change... */ |
b30ab791 | 1236 | drbd_bcast_event(device, &sib); |
b8907339 PR |
1237 | |
1238 | if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) && | |
1239 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) | |
b30ab791 | 1240 | drbd_khelper(device, "pri-on-incon-degr"); |
b8907339 PR |
1241 | |
1242 | /* Here we have the actions that are performed after a | |
1243 | state change. This function might sleep */ | |
1244 | ||
b8907339 | 1245 | if (ns.susp_nod) { |
a6b32bc3 | 1246 | struct drbd_connection *connection = first_peer_device(device)->connection; |
a6d00c8e PR |
1247 | enum drbd_req_event what = NOTHING; |
1248 | ||
0500813f | 1249 | spin_lock_irq(&device->resource->req_lock); |
bde89a9e | 1250 | if (os.conn < C_CONNECTED && conn_lowest_conn(connection) >= C_CONNECTED) |
b8907339 PR |
1251 | what = RESEND; |
1252 | ||
3fb4746d | 1253 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && |
bde89a9e | 1254 | conn_lowest_disk(connection) > D_NEGOTIATING) |
b8907339 PR |
1255 | what = RESTART_FROZEN_DISK_IO; |
1256 | ||
6bbf53ca | 1257 | if (resource->susp_nod && what != NOTHING) { |
bde89a9e AG |
1258 | _tl_restart(connection, what); |
1259 | _conn_request_state(connection, | |
892fdd1a PR |
1260 | (union drbd_state) { { .susp_nod = 1 } }, |
1261 | (union drbd_state) { { .susp_nod = 0 } }, | |
1262 | CS_VERBOSE); | |
b8907339 | 1263 | } |
0500813f | 1264 | spin_unlock_irq(&device->resource->req_lock); |
b8907339 PR |
1265 | } |
1266 | ||
88f79ec4 | 1267 | if (ns.susp_fen) { |
a6b32bc3 | 1268 | struct drbd_connection *connection = first_peer_device(device)->connection; |
88f79ec4 | 1269 | |
0500813f | 1270 | spin_lock_irq(&device->resource->req_lock); |
6bbf53ca | 1271 | if (resource->susp_fen && conn_lowest_conn(connection) >= C_CONNECTED) { |
88f79ec4 | 1272 | /* case2: The connection was established again: */ |
c06ece6b | 1273 | struct drbd_peer_device *peer_device; |
88f79ec4 PR |
1274 | int vnr; |
1275 | ||
1276 | rcu_read_lock(); | |
c06ece6b AG |
1277 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
1278 | clear_bit(NEW_CUR_UUID, &peer_device->device->flags); | |
88f79ec4 | 1279 | rcu_read_unlock(); |
bde89a9e AG |
1280 | _tl_restart(connection, RESEND); |
1281 | _conn_request_state(connection, | |
88f79ec4 PR |
1282 | (union drbd_state) { { .susp_fen = 1 } }, |
1283 | (union drbd_state) { { .susp_fen = 0 } }, | |
1284 | CS_VERBOSE); | |
1285 | } | |
0500813f | 1286 | spin_unlock_irq(&device->resource->req_lock); |
88f79ec4 PR |
1287 | } |
1288 | ||
b8907339 PR |
1289 | /* Became sync source. With protocol >= 96, we still need to send out |
1290 | * the sync uuid now. Need to do that before any drbd_send_state, or | |
1291 | * the other side may go "paused sync" before receiving the sync uuids, | |
1292 | * which is unexpected. */ | |
1293 | if ((os.conn != C_SYNC_SOURCE && os.conn != C_PAUSED_SYNC_S) && | |
1294 | (ns.conn == C_SYNC_SOURCE || ns.conn == C_PAUSED_SYNC_S) && | |
a6b32bc3 | 1295 | first_peer_device(device)->connection->agreed_pro_version >= 96 && get_ldev(device)) { |
69a22773 | 1296 | drbd_gen_and_send_sync_uuid(first_peer_device(device)); |
b30ab791 | 1297 | put_ldev(device); |
b8907339 PR |
1298 | } |
1299 | ||
1300 | /* Do not change the order of the if above and the two below... */ | |
369bea63 PR |
1301 | if (os.pdsk == D_DISKLESS && |
1302 | ns.pdsk > D_DISKLESS && ns.pdsk != D_UNKNOWN) { /* attach on the peer */ | |
a324896b LE |
1303 | /* we probably will start a resync soon. |
1304 | * make sure those things are properly reset. */ | |
b30ab791 AG |
1305 | device->rs_total = 0; |
1306 | device->rs_failed = 0; | |
1307 | atomic_set(&device->rs_pending_cnt, 0); | |
1308 | drbd_rs_cancel_all(device); | |
a324896b | 1309 | |
69a22773 AG |
1310 | drbd_send_uuids(first_peer_device(device)); |
1311 | drbd_send_state(first_peer_device(device), ns); | |
b8907339 PR |
1312 | } |
1313 | /* No point in queuing send_bitmap if we don't have a connection | |
1314 | * anymore, so check also the _current_ state, not only the new state | |
1315 | * at the time this work was queued. */ | |
1316 | if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S && | |
b30ab791 AG |
1317 | device->state.conn == C_WF_BITMAP_S) |
1318 | drbd_queue_bitmap_io(device, &drbd_send_bitmap, NULL, | |
b8907339 PR |
1319 | "send_bitmap (WFBitMapS)", |
1320 | BM_LOCKED_TEST_ALLOWED); | |
1321 | ||
1322 | /* Lost contact to peer's copy of the data */ | |
1323 | if ((os.pdsk >= D_INCONSISTENT && | |
1324 | os.pdsk != D_UNKNOWN && | |
1325 | os.pdsk != D_OUTDATED) | |
1326 | && (ns.pdsk < D_INCONSISTENT || | |
1327 | ns.pdsk == D_UNKNOWN || | |
1328 | ns.pdsk == D_OUTDATED)) { | |
b30ab791 | 1329 | if (get_ldev(device)) { |
b8907339 | 1330 | if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && |
b30ab791 AG |
1331 | device->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
1332 | if (drbd_suspended(device)) { | |
1333 | set_bit(NEW_CUR_UUID, &device->flags); | |
b8907339 | 1334 | } else { |
b30ab791 | 1335 | drbd_uuid_new_current(device); |
69a22773 | 1336 | drbd_send_uuids(first_peer_device(device)); |
b8907339 PR |
1337 | } |
1338 | } | |
b30ab791 | 1339 | put_ldev(device); |
b8907339 PR |
1340 | } |
1341 | } | |
1342 | ||
b30ab791 | 1343 | if (ns.pdsk < D_INCONSISTENT && get_ldev(device)) { |
0cfac5dd | 1344 | if (os.peer == R_SECONDARY && ns.peer == R_PRIMARY && |
b30ab791 AG |
1345 | device->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
1346 | drbd_uuid_new_current(device); | |
69a22773 | 1347 | drbd_send_uuids(first_peer_device(device)); |
0cfac5dd | 1348 | } |
b8907339 PR |
1349 | /* D_DISKLESS Peer becomes secondary */ |
1350 | if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) | |
1351 | /* We may still be Primary ourselves. | |
1352 | * No harm done if the bitmap still changes, | |
1353 | * redirtied pages will follow later. */ | |
b30ab791 | 1354 | drbd_bitmap_io_from_worker(device, &drbd_bm_write, |
b8907339 | 1355 | "demote diskless peer", BM_LOCKED_SET_ALLOWED); |
b30ab791 | 1356 | put_ldev(device); |
b8907339 PR |
1357 | } |
1358 | ||
1359 | /* Write out all changed bits on demote. | |
1360 | * Though, no need to da that just yet | |
1361 | * if there is a resync going on still */ | |
1362 | if (os.role == R_PRIMARY && ns.role == R_SECONDARY && | |
b30ab791 | 1363 | device->state.conn <= C_CONNECTED && get_ldev(device)) { |
b8907339 PR |
1364 | /* No changes to the bitmap expected this time, so assert that, |
1365 | * even though no harm was done if it did change. */ | |
b30ab791 | 1366 | drbd_bitmap_io_from_worker(device, &drbd_bm_write, |
b8907339 | 1367 | "demote", BM_LOCKED_TEST_ALLOWED); |
b30ab791 | 1368 | put_ldev(device); |
b8907339 PR |
1369 | } |
1370 | ||
1371 | /* Last part of the attaching process ... */ | |
1372 | if (ns.conn >= C_CONNECTED && | |
1373 | os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) { | |
69a22773 AG |
1374 | drbd_send_sizes(first_peer_device(device), 0, 0); /* to start sync... */ |
1375 | drbd_send_uuids(first_peer_device(device)); | |
1376 | drbd_send_state(first_peer_device(device), ns); | |
b8907339 PR |
1377 | } |
1378 | ||
1379 | /* We want to pause/continue resync, tell peer. */ | |
1380 | if (ns.conn >= C_CONNECTED && | |
1381 | ((os.aftr_isp != ns.aftr_isp) || | |
1382 | (os.user_isp != ns.user_isp))) | |
69a22773 | 1383 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1384 | |
1385 | /* In case one of the isp bits got set, suspend other devices. */ | |
1386 | if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) && | |
1387 | (ns.aftr_isp || ns.peer_isp || ns.user_isp)) | |
b30ab791 | 1388 | suspend_other_sg(device); |
b8907339 PR |
1389 | |
1390 | /* Make sure the peer gets informed about eventual state | |
1391 | changes (ISP bits) while we were in WFReportParams. */ | |
1392 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED) | |
69a22773 | 1393 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1394 | |
1395 | if (os.conn != C_AHEAD && ns.conn == C_AHEAD) | |
69a22773 | 1396 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1397 | |
1398 | /* We are in the progress to start a full sync... */ | |
1399 | if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
1400 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S)) | |
1401 | /* no other bitmap changes expected during this phase */ | |
b30ab791 | 1402 | drbd_queue_bitmap_io(device, |
b8907339 PR |
1403 | &drbd_bmio_set_n_write, &abw_start_sync, |
1404 | "set_n_write from StartingSync", BM_LOCKED_TEST_ALLOWED); | |
1405 | ||
b8907339 PR |
1406 | /* first half of local IO error, failure to attach, |
1407 | * or administrative detach */ | |
1408 | if (os.disk != D_FAILED && ns.disk == D_FAILED) { | |
32db80f6 PR |
1409 | enum drbd_io_error_p eh = EP_PASS_ON; |
1410 | int was_io_error = 0; | |
b8907339 | 1411 | /* corresponding get_ldev was in __drbd_set_state, to serialize |
32db80f6 PR |
1412 | * our cleanup here with the transition to D_DISKLESS. |
1413 | * But is is still not save to dreference ldev here, since | |
1414 | * we might come from an failed Attach before ldev was set. */ | |
b30ab791 | 1415 | if (device->ldev) { |
32db80f6 | 1416 | rcu_read_lock(); |
b30ab791 | 1417 | eh = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
32db80f6 PR |
1418 | rcu_read_unlock(); |
1419 | ||
b30ab791 | 1420 | was_io_error = test_and_clear_bit(WAS_IO_ERROR, &device->flags); |
32db80f6 | 1421 | |
6f1a6563 | 1422 | if (was_io_error && eh == EP_CALL_HELPER) |
b30ab791 | 1423 | drbd_khelper(device, "local-io-error"); |
6f1a6563 | 1424 | |
0c849666 LE |
1425 | /* Immediately allow completion of all application IO, |
1426 | * that waits for completion from the local disk, | |
1427 | * if this was a force-detach due to disk_timeout | |
1428 | * or administrator request (drbdsetup detach --force). | |
1429 | * Do NOT abort otherwise. | |
1430 | * Aborting local requests may cause serious problems, | |
1431 | * if requests are completed to upper layers already, | |
1432 | * and then later the already submitted local bio completes. | |
1433 | * This can cause DMA into former bio pages that meanwhile | |
1434 | * have been re-used for other things. | |
1435 | * So aborting local requests may cause crashes, | |
1436 | * or even worse, silent data corruption. | |
1437 | */ | |
b30ab791 AG |
1438 | if (test_and_clear_bit(FORCE_DETACH, &device->flags)) |
1439 | tl_abort_disk_io(device); | |
32db80f6 PR |
1440 | |
1441 | /* current state still has to be D_FAILED, | |
1442 | * there is only one way out: to D_DISKLESS, | |
1443 | * and that may only happen after our put_ldev below. */ | |
b30ab791 | 1444 | if (device->state.disk != D_FAILED) |
d0180171 | 1445 | drbd_err(device, |
32db80f6 | 1446 | "ASSERT FAILED: disk is %s during detach\n", |
b30ab791 | 1447 | drbd_disk_str(device->state.disk)); |
32db80f6 PR |
1448 | |
1449 | if (ns.conn >= C_CONNECTED) | |
69a22773 | 1450 | drbd_send_state(first_peer_device(device), ns); |
32db80f6 | 1451 | |
b30ab791 | 1452 | drbd_rs_cancel_all(device); |
32db80f6 PR |
1453 | |
1454 | /* In case we want to get something to stable storage still, | |
1455 | * this may be the last chance. | |
1456 | * Following put_ldev may transition to D_DISKLESS. */ | |
b30ab791 | 1457 | drbd_md_sync(device); |
32db80f6 | 1458 | } |
b30ab791 | 1459 | put_ldev(device); |
b8907339 PR |
1460 | } |
1461 | ||
b30ab791 AG |
1462 | /* second half of local IO error, failure to attach, |
1463 | * or administrative detach, | |
1464 | * after local_cnt references have reached zero again */ | |
1465 | if (os.disk != D_DISKLESS && ns.disk == D_DISKLESS) { | |
1466 | /* We must still be diskless, | |
1467 | * re-attach has to be serialized with this! */ | |
1468 | if (device->state.disk != D_DISKLESS) | |
d0180171 AG |
1469 | drbd_err(device, |
1470 | "ASSERT FAILED: disk is %s while going diskless\n", | |
1471 | drbd_disk_str(device->state.disk)); | |
b8907339 | 1472 | |
6ab9b1b6 | 1473 | if (ns.conn >= C_CONNECTED) |
69a22773 | 1474 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1475 | /* corresponding get_ldev in __drbd_set_state |
1476 | * this may finally trigger drbd_ldev_destroy. */ | |
b30ab791 | 1477 | put_ldev(device); |
b8907339 PR |
1478 | } |
1479 | ||
1480 | /* Notify peer that I had a local IO error, and did not detached.. */ | |
6ab9b1b6 | 1481 | if (os.disk == D_UP_TO_DATE && ns.disk == D_INCONSISTENT && ns.conn >= C_CONNECTED) |
69a22773 | 1482 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1483 | |
1484 | /* Disks got bigger while they were detached */ | |
1485 | if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING && | |
b30ab791 | 1486 | test_and_clear_bit(RESYNC_AFTER_NEG, &device->flags)) { |
b8907339 | 1487 | if (ns.conn == C_CONNECTED) |
b30ab791 | 1488 | resync_after_online_grow(device); |
b8907339 PR |
1489 | } |
1490 | ||
1491 | /* A resync finished or aborted, wake paused devices... */ | |
1492 | if ((os.conn > C_CONNECTED && ns.conn <= C_CONNECTED) || | |
1493 | (os.peer_isp && !ns.peer_isp) || | |
1494 | (os.user_isp && !ns.user_isp)) | |
b30ab791 | 1495 | resume_next_sg(device); |
b8907339 PR |
1496 | |
1497 | /* sync target done with resync. Explicitly notify peer, even though | |
1498 | * it should (at least for non-empty resyncs) already know itself. */ | |
1499 | if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED) | |
69a22773 | 1500 | drbd_send_state(first_peer_device(device), ns); |
b8907339 | 1501 | |
58ffa580 LE |
1502 | /* Verify finished, or reached stop sector. Peer did not know about |
1503 | * the stop sector, and we may even have changed the stop sector during | |
1504 | * verify to interrupt/stop early. Send the new state. */ | |
1505 | if (os.conn == C_VERIFY_S && ns.conn == C_CONNECTED | |
b30ab791 | 1506 | && verify_can_do_stop_sector(device)) |
69a22773 | 1507 | drbd_send_state(first_peer_device(device), ns); |
58ffa580 | 1508 | |
b8907339 PR |
1509 | /* This triggers bitmap writeout of potentially still unwritten pages |
1510 | * if the resync finished cleanly, or aborted because of peer disk | |
1511 | * failure, or because of connection loss. | |
1512 | * For resync aborted because of local disk failure, we cannot do | |
1513 | * any bitmap writeout anymore. | |
1514 | * No harm done if some bits change during this phase. | |
1515 | */ | |
b30ab791 AG |
1516 | if (os.conn > C_CONNECTED && ns.conn <= C_CONNECTED && get_ldev(device)) { |
1517 | drbd_queue_bitmap_io(device, &drbd_bm_write_copy_pages, NULL, | |
a220d291 | 1518 | "write from resync_finished", BM_LOCKED_CHANGE_ALLOWED); |
b30ab791 | 1519 | put_ldev(device); |
b8907339 PR |
1520 | } |
1521 | ||
1522 | if (ns.disk == D_DISKLESS && | |
1523 | ns.conn == C_STANDALONE && | |
1524 | ns.role == R_SECONDARY) { | |
1525 | if (os.aftr_isp != ns.aftr_isp) | |
b30ab791 | 1526 | resume_next_sg(device); |
b8907339 PR |
1527 | } |
1528 | ||
b30ab791 | 1529 | drbd_md_sync(device); |
b8907339 PR |
1530 | } |
1531 | ||
bbeb641c PR |
1532 | struct after_conn_state_chg_work { |
1533 | struct drbd_work w; | |
1534 | enum drbd_conns oc; | |
8c7e16c3 | 1535 | union drbd_state ns_min; |
b30ab791 | 1536 | union drbd_state ns_max; /* new, max state, over all devices */ |
bbeb641c PR |
1537 | enum chg_state_flags flags; |
1538 | }; | |
1539 | ||
99920dc5 | 1540 | static int w_after_conn_state_ch(struct drbd_work *w, int unused) |
bbeb641c PR |
1541 | { |
1542 | struct after_conn_state_chg_work *acscw = | |
1543 | container_of(w, struct after_conn_state_chg_work, w); | |
bde89a9e | 1544 | struct drbd_connection *connection = w->connection; |
bbeb641c | 1545 | enum drbd_conns oc = acscw->oc; |
5f082f98 | 1546 | union drbd_state ns_max = acscw->ns_max; |
c06ece6b | 1547 | struct drbd_peer_device *peer_device; |
a6d00c8e | 1548 | int vnr; |
bbeb641c PR |
1549 | |
1550 | kfree(acscw); | |
1551 | ||
b8907339 | 1552 | /* Upon network configuration, we need to start the receiver */ |
5f082f98 | 1553 | if (oc == C_STANDALONE && ns_max.conn == C_UNCONNECTED) |
bde89a9e | 1554 | drbd_thread_start(&connection->receiver); |
b8907339 | 1555 | |
f3dfa40a LE |
1556 | if (oc == C_DISCONNECTING && ns_max.conn == C_STANDALONE) { |
1557 | struct net_conf *old_conf; | |
1558 | ||
0500813f | 1559 | mutex_lock(&connection->resource->conf_update); |
bde89a9e AG |
1560 | old_conf = connection->net_conf; |
1561 | connection->my_addr_len = 0; | |
1562 | connection->peer_addr_len = 0; | |
1563 | rcu_assign_pointer(connection->net_conf, NULL); | |
1564 | conn_free_crypto(connection); | |
0500813f | 1565 | mutex_unlock(&connection->resource->conf_update); |
f3dfa40a LE |
1566 | |
1567 | synchronize_rcu(); | |
1568 | kfree(old_conf); | |
1569 | } | |
1570 | ||
a6d00c8e PR |
1571 | if (ns_max.susp_fen) { |
1572 | /* case1: The outdate peer handler is successful: */ | |
1573 | if (ns_max.pdsk <= D_OUTDATED) { | |
695d08fa | 1574 | rcu_read_lock(); |
c06ece6b AG |
1575 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1576 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1577 | if (test_bit(NEW_CUR_UUID, &device->flags)) { |
1578 | drbd_uuid_new_current(device); | |
1579 | clear_bit(NEW_CUR_UUID, &device->flags); | |
a6d00c8e PR |
1580 | } |
1581 | } | |
695d08fa | 1582 | rcu_read_unlock(); |
0500813f | 1583 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1584 | _tl_restart(connection, CONNECTION_LOST_WHILE_PENDING); |
1585 | _conn_request_state(connection, | |
8a0bab2a PR |
1586 | (union drbd_state) { { .susp_fen = 1 } }, |
1587 | (union drbd_state) { { .susp_fen = 0 } }, | |
1588 | CS_VERBOSE); | |
0500813f | 1589 | spin_unlock_irq(&connection->resource->req_lock); |
a6d00c8e | 1590 | } |
a6d00c8e | 1591 | } |
05a10ec7 | 1592 | kref_put(&connection->kref, drbd_destroy_connection); |
19fffd7b | 1593 | |
bde89a9e | 1594 | conn_md_sync(connection); |
19fffd7b | 1595 | |
99920dc5 | 1596 | return 0; |
bbeb641c PR |
1597 | } |
1598 | ||
bde89a9e | 1599 | void conn_old_common_state(struct drbd_connection *connection, union drbd_state *pcs, enum chg_state_flags *pf) |
88ef594e | 1600 | { |
435693e8 | 1601 | enum chg_state_flags flags = ~0; |
c06ece6b | 1602 | struct drbd_peer_device *peer_device; |
435693e8 | 1603 | int vnr, first_vol = 1; |
e0e16653 PR |
1604 | union drbd_dev_state os, cs = { |
1605 | { .role = R_SECONDARY, | |
1606 | .peer = R_UNKNOWN, | |
bde89a9e | 1607 | .conn = connection->cstate, |
e0e16653 PR |
1608 | .disk = D_DISKLESS, |
1609 | .pdsk = D_UNKNOWN, | |
1610 | } }; | |
88ef594e | 1611 | |
695d08fa | 1612 | rcu_read_lock(); |
c06ece6b AG |
1613 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1614 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 1615 | os = device->state; |
88ef594e | 1616 | |
435693e8 PR |
1617 | if (first_vol) { |
1618 | cs = os; | |
1619 | first_vol = 0; | |
1620 | continue; | |
1621 | } | |
1622 | ||
1623 | if (cs.role != os.role) | |
1624 | flags &= ~CS_DC_ROLE; | |
1625 | ||
1626 | if (cs.peer != os.peer) | |
1627 | flags &= ~CS_DC_PEER; | |
1628 | ||
1629 | if (cs.conn != os.conn) | |
1630 | flags &= ~CS_DC_CONN; | |
88ef594e | 1631 | |
435693e8 PR |
1632 | if (cs.disk != os.disk) |
1633 | flags &= ~CS_DC_DISK; | |
88ef594e | 1634 | |
435693e8 PR |
1635 | if (cs.pdsk != os.pdsk) |
1636 | flags &= ~CS_DC_PDSK; | |
1637 | } | |
695d08fa | 1638 | rcu_read_unlock(); |
435693e8 PR |
1639 | |
1640 | *pf |= CS_DC_MASK; | |
1641 | *pf &= flags; | |
da9fbc27 | 1642 | (*pcs).i = cs.i; |
88ef594e | 1643 | } |
bbeb641c | 1644 | |
bd0c824a | 1645 | static enum drbd_state_rv |
bde89a9e | 1646 | conn_is_valid_transition(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
88ef594e | 1647 | enum chg_state_flags flags) |
bbeb641c | 1648 | { |
bd0c824a | 1649 | enum drbd_state_rv rv = SS_SUCCESS; |
bbeb641c | 1650 | union drbd_state ns, os; |
c06ece6b | 1651 | struct drbd_peer_device *peer_device; |
bd0c824a | 1652 | int vnr; |
bbeb641c | 1653 | |
695d08fa | 1654 | rcu_read_lock(); |
c06ece6b AG |
1655 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1656 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1657 | os = drbd_read_state(device); |
1658 | ns = sanitize_state(device, apply_mask_val(os, mask, val), NULL); | |
bd0c824a | 1659 | |
778bcf2e PR |
1660 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
1661 | ns.disk = os.disk; | |
1662 | ||
bd0c824a PR |
1663 | if (ns.i == os.i) |
1664 | continue; | |
bbeb641c | 1665 | |
bd0c824a | 1666 | rv = is_valid_transition(os, ns); |
bd0c824a | 1667 | |
c06ece6b | 1668 | if (rv >= SS_SUCCESS && !(flags & CS_HARD)) { |
b30ab791 | 1669 | rv = is_valid_state(device, ns); |
bd0c824a | 1670 | if (rv < SS_SUCCESS) { |
b30ab791 | 1671 | if (is_valid_state(device, os) == rv) |
bde89a9e | 1672 | rv = is_valid_soft_transition(os, ns, connection); |
bd0c824a | 1673 | } else |
bde89a9e | 1674 | rv = is_valid_soft_transition(os, ns, connection); |
bd0c824a | 1675 | } |
c06ece6b AG |
1676 | |
1677 | if (rv < SS_SUCCESS) { | |
1678 | if (flags & CS_VERBOSE) | |
1679 | print_st_err(device, os, ns, rv); | |
bd0c824a | 1680 | break; |
c06ece6b | 1681 | } |
bbeb641c | 1682 | } |
695d08fa | 1683 | rcu_read_unlock(); |
bbeb641c | 1684 | |
bd0c824a | 1685 | return rv; |
bbeb641c PR |
1686 | } |
1687 | ||
8c7e16c3 | 1688 | void |
bde89a9e | 1689 | conn_set_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
8c7e16c3 | 1690 | union drbd_state *pns_min, union drbd_state *pns_max, enum chg_state_flags flags) |
bbeb641c | 1691 | { |
f132f554 | 1692 | union drbd_state ns, os, ns_max = { }; |
8c7e16c3 PR |
1693 | union drbd_state ns_min = { |
1694 | { .role = R_MASK, | |
1695 | .peer = R_MASK, | |
e0e16653 | 1696 | .conn = val.conn, |
8c7e16c3 PR |
1697 | .disk = D_MASK, |
1698 | .pdsk = D_MASK | |
1699 | } }; | |
c06ece6b | 1700 | struct drbd_peer_device *peer_device; |
bbeb641c | 1701 | enum drbd_state_rv rv; |
f132f554 | 1702 | int vnr, number_of_volumes = 0; |
bbeb641c | 1703 | |
07be15b1 LE |
1704 | if (mask.conn == C_MASK) { |
1705 | /* remember last connect time so request_timer_fn() won't | |
1706 | * kill newly established sessions while we are still trying to thaw | |
1707 | * previously frozen IO */ | |
bde89a9e AG |
1708 | if (connection->cstate != C_WF_REPORT_PARAMS && val.conn == C_WF_REPORT_PARAMS) |
1709 | connection->last_reconnect_jif = jiffies; | |
07be15b1 | 1710 | |
bde89a9e | 1711 | connection->cstate = val.conn; |
07be15b1 | 1712 | } |
bd0c824a | 1713 | |
695d08fa | 1714 | rcu_read_lock(); |
c06ece6b AG |
1715 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1716 | struct drbd_device *device = peer_device->device; | |
f132f554 | 1717 | number_of_volumes++; |
b30ab791 | 1718 | os = drbd_read_state(device); |
bd0c824a | 1719 | ns = apply_mask_val(os, mask, val); |
b30ab791 | 1720 | ns = sanitize_state(device, ns, NULL); |
bbeb641c | 1721 | |
778bcf2e PR |
1722 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
1723 | ns.disk = os.disk; | |
1724 | ||
b30ab791 | 1725 | rv = __drbd_set_state(device, ns, flags, NULL); |
bd0c824a PR |
1726 | if (rv < SS_SUCCESS) |
1727 | BUG(); | |
bbeb641c | 1728 | |
b30ab791 | 1729 | ns.i = device->state.i; |
8c7e16c3 PR |
1730 | ns_max.role = max_role(ns.role, ns_max.role); |
1731 | ns_max.peer = max_role(ns.peer, ns_max.peer); | |
1732 | ns_max.conn = max_t(enum drbd_conns, ns.conn, ns_max.conn); | |
1733 | ns_max.disk = max_t(enum drbd_disk_state, ns.disk, ns_max.disk); | |
1734 | ns_max.pdsk = max_t(enum drbd_disk_state, ns.pdsk, ns_max.pdsk); | |
1735 | ||
1736 | ns_min.role = min_role(ns.role, ns_min.role); | |
1737 | ns_min.peer = min_role(ns.peer, ns_min.peer); | |
1738 | ns_min.conn = min_t(enum drbd_conns, ns.conn, ns_min.conn); | |
1739 | ns_min.disk = min_t(enum drbd_disk_state, ns.disk, ns_min.disk); | |
1740 | ns_min.pdsk = min_t(enum drbd_disk_state, ns.pdsk, ns_min.pdsk); | |
bd0c824a | 1741 | } |
695d08fa | 1742 | rcu_read_unlock(); |
bbeb641c | 1743 | |
f132f554 PR |
1744 | if (number_of_volumes == 0) { |
1745 | ns_min = ns_max = (union drbd_state) { { | |
1746 | .role = R_SECONDARY, | |
1747 | .peer = R_UNKNOWN, | |
1748 | .conn = val.conn, | |
1749 | .disk = D_DISKLESS, | |
1750 | .pdsk = D_UNKNOWN | |
1751 | } }; | |
1752 | } | |
1753 | ||
6bbf53ca AG |
1754 | ns_min.susp = ns_max.susp = connection->resource->susp; |
1755 | ns_min.susp_nod = ns_max.susp_nod = connection->resource->susp_nod; | |
1756 | ns_min.susp_fen = ns_max.susp_fen = connection->resource->susp_fen; | |
8c7e16c3 PR |
1757 | |
1758 | *pns_min = ns_min; | |
1759 | *pns_max = ns_max; | |
bbeb641c PR |
1760 | } |
1761 | ||
df24aa45 | 1762 | static enum drbd_state_rv |
bde89a9e | 1763 | _conn_rq_cond(struct drbd_connection *connection, union drbd_state mask, union drbd_state val) |
df24aa45 | 1764 | { |
df24aa45 PR |
1765 | enum drbd_state_rv rv; |
1766 | ||
bde89a9e | 1767 | if (test_and_clear_bit(CONN_WD_ST_CHG_OKAY, &connection->flags)) |
df24aa45 PR |
1768 | return SS_CW_SUCCESS; |
1769 | ||
bde89a9e | 1770 | if (test_and_clear_bit(CONN_WD_ST_CHG_FAIL, &connection->flags)) |
df24aa45 PR |
1771 | return SS_CW_FAILED_BY_PEER; |
1772 | ||
bde89a9e AG |
1773 | rv = conn_is_valid_transition(connection, mask, val, 0); |
1774 | if (rv == SS_SUCCESS && connection->cstate == C_WF_REPORT_PARAMS) | |
2bd5ed5d | 1775 | rv = SS_UNKNOWN_ERROR; /* continue waiting */ |
df24aa45 | 1776 | |
df24aa45 PR |
1777 | return rv; |
1778 | } | |
1779 | ||
bbeb641c | 1780 | enum drbd_state_rv |
bde89a9e | 1781 | _conn_request_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
bbeb641c PR |
1782 | enum chg_state_flags flags) |
1783 | { | |
1784 | enum drbd_state_rv rv = SS_SUCCESS; | |
bbeb641c | 1785 | struct after_conn_state_chg_work *acscw; |
bde89a9e | 1786 | enum drbd_conns oc = connection->cstate; |
8c7e16c3 | 1787 | union drbd_state ns_max, ns_min, os; |
c02abda2 | 1788 | bool have_mutex = false; |
bbeb641c | 1789 | |
07fc9619 PR |
1790 | if (mask.conn) { |
1791 | rv = is_valid_conn_transition(oc, val.conn); | |
1792 | if (rv < SS_SUCCESS) | |
1793 | goto abort; | |
1794 | } | |
bbeb641c | 1795 | |
bde89a9e | 1796 | rv = conn_is_valid_transition(connection, mask, val, flags); |
bbeb641c PR |
1797 | if (rv < SS_SUCCESS) |
1798 | goto abort; | |
1799 | ||
df24aa45 PR |
1800 | if (oc == C_WF_REPORT_PARAMS && val.conn == C_DISCONNECTING && |
1801 | !(flags & (CS_LOCAL_ONLY | CS_HARD))) { | |
c02abda2 LE |
1802 | |
1803 | /* This will be a cluster-wide state change. | |
1804 | * Need to give up the spinlock, grab the mutex, | |
1805 | * then send the state change request, ... */ | |
0500813f | 1806 | spin_unlock_irq(&connection->resource->req_lock); |
bde89a9e | 1807 | mutex_lock(&connection->cstate_mutex); |
c02abda2 LE |
1808 | have_mutex = true; |
1809 | ||
bde89a9e AG |
1810 | set_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
1811 | if (conn_send_state_req(connection, mask, val)) { | |
c02abda2 | 1812 | /* sending failed. */ |
bde89a9e | 1813 | clear_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
c02abda2 LE |
1814 | rv = SS_CW_FAILED_BY_PEER; |
1815 | /* need to re-aquire the spin lock, though */ | |
1816 | goto abort_unlocked; | |
1817 | } | |
1818 | ||
1819 | if (val.conn == C_DISCONNECTING) | |
bde89a9e | 1820 | set_bit(DISCONNECT_SENT, &connection->flags); |
c02abda2 LE |
1821 | |
1822 | /* ... and re-aquire the spinlock. | |
1823 | * If _conn_rq_cond() returned >= SS_SUCCESS, we must call | |
1824 | * conn_set_state() within the same spinlock. */ | |
0500813f | 1825 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1826 | wait_event_lock_irq(connection->ping_wait, |
1827 | (rv = _conn_rq_cond(connection, mask, val)), | |
0500813f | 1828 | connection->resource->req_lock); |
bde89a9e | 1829 | clear_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
df24aa45 PR |
1830 | if (rv < SS_SUCCESS) |
1831 | goto abort; | |
1832 | } | |
1833 | ||
bde89a9e | 1834 | conn_old_common_state(connection, &os, &flags); |
706cb24c | 1835 | flags |= CS_DC_SUSP; |
bde89a9e AG |
1836 | conn_set_state(connection, mask, val, &ns_min, &ns_max, flags); |
1837 | conn_pr_state_change(connection, os, ns_max, flags); | |
bbeb641c PR |
1838 | |
1839 | acscw = kmalloc(sizeof(*acscw), GFP_ATOMIC); | |
1840 | if (acscw) { | |
435693e8 | 1841 | acscw->oc = os.conn; |
8c7e16c3 | 1842 | acscw->ns_min = ns_min; |
5f082f98 | 1843 | acscw->ns_max = ns_max; |
bbeb641c PR |
1844 | acscw->flags = flags; |
1845 | acscw->w.cb = w_after_conn_state_ch; | |
bde89a9e AG |
1846 | kref_get(&connection->kref); |
1847 | acscw->w.connection = connection; | |
1848 | drbd_queue_work(&connection->sender_work, &acscw->w); | |
bbeb641c | 1849 | } else { |
1ec861eb | 1850 | drbd_err(connection, "Could not kmalloc an acscw\n"); |
b8907339 | 1851 | } |
bbeb641c | 1852 | |
a01842eb | 1853 | abort: |
c02abda2 LE |
1854 | if (have_mutex) { |
1855 | /* mutex_unlock() "... must not be used in interrupt context.", | |
1856 | * so give up the spinlock, then re-aquire it */ | |
0500813f | 1857 | spin_unlock_irq(&connection->resource->req_lock); |
c02abda2 | 1858 | abort_unlocked: |
bde89a9e | 1859 | mutex_unlock(&connection->cstate_mutex); |
0500813f | 1860 | spin_lock_irq(&connection->resource->req_lock); |
c02abda2 LE |
1861 | } |
1862 | if (rv < SS_SUCCESS && flags & CS_VERBOSE) { | |
1ec861eb AG |
1863 | drbd_err(connection, "State change failed: %s\n", drbd_set_st_err_str(rv)); |
1864 | drbd_err(connection, " mask = 0x%x val = 0x%x\n", mask.i, val.i); | |
1865 | drbd_err(connection, " old_conn:%s wanted_conn:%s\n", drbd_conn_str(oc), drbd_conn_str(val.conn)); | |
a01842eb | 1866 | } |
bbeb641c PR |
1867 | return rv; |
1868 | } | |
1869 | ||
1870 | enum drbd_state_rv | |
bde89a9e | 1871 | conn_request_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
bbeb641c PR |
1872 | enum chg_state_flags flags) |
1873 | { | |
1874 | enum drbd_state_rv rv; | |
1875 | ||
0500813f | 1876 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 1877 | rv = _conn_request_state(connection, mask, val, flags); |
0500813f | 1878 | spin_unlock_irq(&connection->resource->req_lock); |
bbeb641c PR |
1879 | |
1880 | return rv; | |
b8907339 | 1881 | } |