]>
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 | 31 | #include "drbd_req.h" |
a2972846 | 32 | #include "drbd_state_change.h" |
b8907339 PR |
33 | |
34 | struct after_state_chg_work { | |
35 | struct drbd_work w; | |
84b8c06b | 36 | struct drbd_device *device; |
b8907339 PR |
37 | union drbd_state os; |
38 | union drbd_state ns; | |
39 | enum chg_state_flags flags; | |
40 | struct completion *done; | |
a2972846 | 41 | struct drbd_state_change *state_change; |
b8907339 PR |
42 | }; |
43 | ||
d942ae44 PR |
44 | enum sanitize_state_warnings { |
45 | NO_WARNING, | |
46 | ABORTED_ONLINE_VERIFY, | |
47 | ABORTED_RESYNC, | |
48 | CONNECTION_LOST_NEGOTIATING, | |
49 | IMPLICITLY_UPGRADED_DISK, | |
50 | IMPLICITLY_UPGRADED_PDSK, | |
51 | }; | |
52 | ||
a2972846 AG |
53 | static void count_objects(struct drbd_resource *resource, |
54 | unsigned int *n_devices, | |
55 | unsigned int *n_connections) | |
56 | { | |
57 | struct drbd_device *device; | |
58 | struct drbd_connection *connection; | |
59 | int vnr; | |
60 | ||
61 | *n_devices = 0; | |
62 | *n_connections = 0; | |
63 | ||
64 | idr_for_each_entry(&resource->devices, device, vnr) | |
65 | (*n_devices)++; | |
2b479766 | 66 | for_each_connection(connection, resource) |
a2972846 | 67 | (*n_connections)++; |
a2972846 AG |
68 | } |
69 | ||
70 | static struct drbd_state_change *alloc_state_change(unsigned int n_devices, unsigned int n_connections, gfp_t gfp) | |
71 | { | |
72 | struct drbd_state_change *state_change; | |
73 | unsigned int size, n; | |
74 | ||
75 | size = sizeof(struct drbd_state_change) + | |
76 | n_devices * sizeof(struct drbd_device_state_change) + | |
77 | n_connections * sizeof(struct drbd_connection_state_change) + | |
78 | n_devices * n_connections * sizeof(struct drbd_peer_device_state_change); | |
79 | state_change = kmalloc(size, gfp); | |
80 | if (!state_change) | |
81 | return NULL; | |
82 | state_change->n_devices = n_devices; | |
83 | state_change->n_connections = n_connections; | |
84 | state_change->devices = (void *)(state_change + 1); | |
85 | state_change->connections = (void *)&state_change->devices[n_devices]; | |
86 | state_change->peer_devices = (void *)&state_change->connections[n_connections]; | |
87 | state_change->resource->resource = NULL; | |
88 | for (n = 0; n < n_devices; n++) | |
89 | state_change->devices[n].device = NULL; | |
90 | for (n = 0; n < n_connections; n++) | |
91 | state_change->connections[n].connection = NULL; | |
92 | return state_change; | |
93 | } | |
94 | ||
95 | struct drbd_state_change *remember_old_state(struct drbd_resource *resource, gfp_t gfp) | |
96 | { | |
97 | struct drbd_state_change *state_change; | |
98 | struct drbd_device *device; | |
99 | unsigned int n_devices; | |
100 | struct drbd_connection *connection; | |
101 | unsigned int n_connections; | |
102 | int vnr; | |
103 | ||
104 | struct drbd_device_state_change *device_state_change; | |
105 | struct drbd_peer_device_state_change *peer_device_state_change; | |
106 | struct drbd_connection_state_change *connection_state_change; | |
107 | ||
2b479766 LE |
108 | /* Caller holds req_lock spinlock. |
109 | * No state, no device IDR, no connections lists can change. */ | |
a2972846 | 110 | count_objects(resource, &n_devices, &n_connections); |
a2972846 AG |
111 | state_change = alloc_state_change(n_devices, n_connections, gfp); |
112 | if (!state_change) | |
113 | return NULL; | |
114 | ||
a2972846 AG |
115 | kref_get(&resource->kref); |
116 | state_change->resource->resource = resource; | |
117 | state_change->resource->role[OLD] = | |
118 | conn_highest_role(first_connection(resource)); | |
119 | state_change->resource->susp[OLD] = resource->susp; | |
120 | state_change->resource->susp_nod[OLD] = resource->susp_nod; | |
121 | state_change->resource->susp_fen[OLD] = resource->susp_fen; | |
122 | ||
2b479766 LE |
123 | connection_state_change = state_change->connections; |
124 | for_each_connection(connection, resource) { | |
125 | kref_get(&connection->kref); | |
126 | connection_state_change->connection = connection; | |
127 | connection_state_change->cstate[OLD] = | |
128 | connection->cstate; | |
129 | connection_state_change->peer_role[OLD] = | |
130 | conn_highest_peer(connection); | |
131 | connection_state_change++; | |
132 | } | |
133 | ||
a2972846 AG |
134 | device_state_change = state_change->devices; |
135 | peer_device_state_change = state_change->peer_devices; | |
136 | idr_for_each_entry(&resource->devices, device, vnr) { | |
137 | kref_get(&device->kref); | |
138 | device_state_change->device = device; | |
139 | device_state_change->disk_state[OLD] = device->state.disk; | |
140 | ||
141 | /* The peer_devices for each device have to be enumerated in | |
142 | the order of the connections. We may not use for_each_peer_device() here. */ | |
143 | for_each_connection(connection, resource) { | |
144 | struct drbd_peer_device *peer_device; | |
145 | ||
a2972846 AG |
146 | peer_device = conn_peer_device(connection, device->vnr); |
147 | peer_device_state_change->peer_device = peer_device; | |
148 | peer_device_state_change->disk_state[OLD] = | |
149 | device->state.pdsk; | |
150 | peer_device_state_change->repl_state[OLD] = | |
151 | max_t(enum drbd_conns, | |
152 | C_WF_REPORT_PARAMS, device->state.conn); | |
153 | peer_device_state_change->resync_susp_user[OLD] = | |
154 | device->state.user_isp; | |
155 | peer_device_state_change->resync_susp_peer[OLD] = | |
156 | device->state.peer_isp; | |
157 | peer_device_state_change->resync_susp_dependency[OLD] = | |
158 | device->state.aftr_isp; | |
159 | peer_device_state_change++; | |
160 | } | |
161 | device_state_change++; | |
162 | } | |
163 | ||
a2972846 AG |
164 | return state_change; |
165 | } | |
166 | ||
167 | static void remember_new_state(struct drbd_state_change *state_change) | |
168 | { | |
169 | struct drbd_resource_state_change *resource_state_change; | |
170 | struct drbd_resource *resource; | |
171 | unsigned int n; | |
172 | ||
173 | if (!state_change) | |
174 | return; | |
175 | ||
176 | resource_state_change = &state_change->resource[0]; | |
177 | resource = resource_state_change->resource; | |
178 | ||
179 | resource_state_change->role[NEW] = | |
180 | conn_highest_role(first_connection(resource)); | |
181 | resource_state_change->susp[NEW] = resource->susp; | |
182 | resource_state_change->susp_nod[NEW] = resource->susp_nod; | |
183 | resource_state_change->susp_fen[NEW] = resource->susp_fen; | |
184 | ||
185 | for (n = 0; n < state_change->n_devices; n++) { | |
186 | struct drbd_device_state_change *device_state_change = | |
187 | &state_change->devices[n]; | |
188 | struct drbd_device *device = device_state_change->device; | |
189 | ||
190 | device_state_change->disk_state[NEW] = device->state.disk; | |
191 | } | |
192 | ||
193 | for (n = 0; n < state_change->n_connections; n++) { | |
194 | struct drbd_connection_state_change *connection_state_change = | |
195 | &state_change->connections[n]; | |
196 | struct drbd_connection *connection = | |
197 | connection_state_change->connection; | |
198 | ||
199 | connection_state_change->cstate[NEW] = connection->cstate; | |
200 | connection_state_change->peer_role[NEW] = | |
201 | conn_highest_peer(connection); | |
202 | } | |
203 | ||
204 | for (n = 0; n < state_change->n_devices * state_change->n_connections; n++) { | |
205 | struct drbd_peer_device_state_change *peer_device_state_change = | |
206 | &state_change->peer_devices[n]; | |
207 | struct drbd_device *device = | |
208 | peer_device_state_change->peer_device->device; | |
209 | union drbd_dev_state state = device->state; | |
210 | ||
211 | peer_device_state_change->disk_state[NEW] = state.pdsk; | |
212 | peer_device_state_change->repl_state[NEW] = | |
213 | max_t(enum drbd_conns, C_WF_REPORT_PARAMS, state.conn); | |
214 | peer_device_state_change->resync_susp_user[NEW] = | |
215 | state.user_isp; | |
216 | peer_device_state_change->resync_susp_peer[NEW] = | |
217 | state.peer_isp; | |
218 | peer_device_state_change->resync_susp_dependency[NEW] = | |
219 | state.aftr_isp; | |
220 | } | |
221 | } | |
222 | ||
223 | void copy_old_to_new_state_change(struct drbd_state_change *state_change) | |
224 | { | |
225 | struct drbd_resource_state_change *resource_state_change = &state_change->resource[0]; | |
226 | unsigned int n_device, n_connection, n_peer_device, n_peer_devices; | |
227 | ||
228 | #define OLD_TO_NEW(x) \ | |
229 | (x[NEW] = x[OLD]) | |
230 | ||
231 | OLD_TO_NEW(resource_state_change->role); | |
232 | OLD_TO_NEW(resource_state_change->susp); | |
233 | OLD_TO_NEW(resource_state_change->susp_nod); | |
234 | OLD_TO_NEW(resource_state_change->susp_fen); | |
235 | ||
236 | for (n_connection = 0; n_connection < state_change->n_connections; n_connection++) { | |
237 | struct drbd_connection_state_change *connection_state_change = | |
238 | &state_change->connections[n_connection]; | |
239 | ||
240 | OLD_TO_NEW(connection_state_change->peer_role); | |
241 | OLD_TO_NEW(connection_state_change->cstate); | |
242 | } | |
243 | ||
244 | for (n_device = 0; n_device < state_change->n_devices; n_device++) { | |
245 | struct drbd_device_state_change *device_state_change = | |
246 | &state_change->devices[n_device]; | |
247 | ||
248 | OLD_TO_NEW(device_state_change->disk_state); | |
249 | } | |
250 | ||
251 | n_peer_devices = state_change->n_devices * state_change->n_connections; | |
252 | for (n_peer_device = 0; n_peer_device < n_peer_devices; n_peer_device++) { | |
253 | struct drbd_peer_device_state_change *p = | |
254 | &state_change->peer_devices[n_peer_device]; | |
255 | ||
256 | OLD_TO_NEW(p->disk_state); | |
257 | OLD_TO_NEW(p->repl_state); | |
258 | OLD_TO_NEW(p->resync_susp_user); | |
259 | OLD_TO_NEW(p->resync_susp_peer); | |
260 | OLD_TO_NEW(p->resync_susp_dependency); | |
261 | } | |
262 | ||
263 | #undef OLD_TO_NEW | |
264 | } | |
265 | ||
266 | void forget_state_change(struct drbd_state_change *state_change) | |
267 | { | |
268 | unsigned int n; | |
269 | ||
270 | if (!state_change) | |
271 | return; | |
272 | ||
273 | if (state_change->resource->resource) | |
274 | kref_put(&state_change->resource->resource->kref, drbd_destroy_resource); | |
275 | for (n = 0; n < state_change->n_devices; n++) { | |
276 | struct drbd_device *device = state_change->devices[n].device; | |
277 | ||
278 | if (device) | |
279 | kref_put(&device->kref, drbd_destroy_device); | |
280 | } | |
281 | for (n = 0; n < state_change->n_connections; n++) { | |
282 | struct drbd_connection *connection = | |
283 | state_change->connections[n].connection; | |
284 | ||
285 | if (connection) | |
286 | kref_put(&connection->kref, drbd_destroy_connection); | |
287 | } | |
288 | kfree(state_change); | |
289 | } | |
290 | ||
99920dc5 | 291 | static int w_after_state_ch(struct drbd_work *w, int unused); |
b30ab791 | 292 | static void after_state_ch(struct drbd_device *device, union drbd_state os, |
a2972846 AG |
293 | union drbd_state ns, enum chg_state_flags flags, |
294 | struct drbd_state_change *); | |
54761697 | 295 | static enum drbd_state_rv is_valid_state(struct drbd_device *, union drbd_state); |
bde89a9e | 296 | static enum drbd_state_rv is_valid_soft_transition(union drbd_state, union drbd_state, struct drbd_connection *); |
3509502d | 297 | static enum drbd_state_rv is_valid_transition(union drbd_state os, union drbd_state ns); |
d7fe69c6 PR |
298 | static union drbd_state sanitize_state(struct drbd_device *device, union drbd_state os, |
299 | union drbd_state ns, enum sanitize_state_warnings *warn); | |
b8907339 | 300 | |
2aebfabb PR |
301 | static inline bool is_susp(union drbd_state s) |
302 | { | |
303 | return s.susp || s.susp_nod || s.susp_fen; | |
304 | } | |
305 | ||
bde89a9e | 306 | bool conn_all_vols_unconf(struct drbd_connection *connection) |
0e29d163 | 307 | { |
c06ece6b | 308 | struct drbd_peer_device *peer_device; |
695d08fa | 309 | bool rv = true; |
e90285e0 | 310 | int vnr; |
0e29d163 | 311 | |
695d08fa | 312 | rcu_read_lock(); |
c06ece6b AG |
313 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
314 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
315 | if (device->state.disk != D_DISKLESS || |
316 | device->state.conn != C_STANDALONE || | |
317 | device->state.role != R_SECONDARY) { | |
695d08fa PR |
318 | rv = false; |
319 | break; | |
320 | } | |
0e29d163 | 321 | } |
695d08fa PR |
322 | rcu_read_unlock(); |
323 | ||
324 | return rv; | |
0e29d163 PR |
325 | } |
326 | ||
cb703454 PR |
327 | /* Unfortunately the states where not correctly ordered, when |
328 | they where defined. therefore can not use max_t() here. */ | |
329 | static enum drbd_role max_role(enum drbd_role role1, enum drbd_role role2) | |
330 | { | |
331 | if (role1 == R_PRIMARY || role2 == R_PRIMARY) | |
332 | return R_PRIMARY; | |
333 | if (role1 == R_SECONDARY || role2 == R_SECONDARY) | |
334 | return R_SECONDARY; | |
335 | return R_UNKNOWN; | |
336 | } | |
a2972846 | 337 | |
cb703454 PR |
338 | static enum drbd_role min_role(enum drbd_role role1, enum drbd_role role2) |
339 | { | |
340 | if (role1 == R_UNKNOWN || role2 == R_UNKNOWN) | |
341 | return R_UNKNOWN; | |
342 | if (role1 == R_SECONDARY || role2 == R_SECONDARY) | |
343 | return R_SECONDARY; | |
344 | return R_PRIMARY; | |
345 | } | |
346 | ||
bde89a9e | 347 | enum drbd_role conn_highest_role(struct drbd_connection *connection) |
cb703454 PR |
348 | { |
349 | enum drbd_role role = R_UNKNOWN; | |
c06ece6b | 350 | struct drbd_peer_device *peer_device; |
cb703454 PR |
351 | int vnr; |
352 | ||
695d08fa | 353 | rcu_read_lock(); |
c06ece6b AG |
354 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
355 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 356 | role = max_role(role, device->state.role); |
c06ece6b | 357 | } |
695d08fa | 358 | rcu_read_unlock(); |
cb703454 PR |
359 | |
360 | return role; | |
361 | } | |
362 | ||
bde89a9e | 363 | enum drbd_role conn_highest_peer(struct drbd_connection *connection) |
cb703454 PR |
364 | { |
365 | enum drbd_role peer = R_UNKNOWN; | |
c06ece6b | 366 | struct drbd_peer_device *peer_device; |
cb703454 PR |
367 | int vnr; |
368 | ||
695d08fa | 369 | rcu_read_lock(); |
c06ece6b AG |
370 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
371 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 372 | peer = max_role(peer, device->state.peer); |
c06ece6b | 373 | } |
695d08fa | 374 | rcu_read_unlock(); |
cb703454 PR |
375 | |
376 | return peer; | |
377 | } | |
378 | ||
bde89a9e | 379 | enum drbd_disk_state conn_highest_disk(struct drbd_connection *connection) |
cb703454 | 380 | { |
11f8b2b6 | 381 | enum drbd_disk_state disk_state = D_DISKLESS; |
c06ece6b | 382 | struct drbd_peer_device *peer_device; |
cb703454 PR |
383 | int vnr; |
384 | ||
695d08fa | 385 | rcu_read_lock(); |
c06ece6b AG |
386 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
387 | struct drbd_device *device = peer_device->device; | |
11f8b2b6 | 388 | disk_state = max_t(enum drbd_disk_state, disk_state, device->state.disk); |
c06ece6b | 389 | } |
695d08fa | 390 | rcu_read_unlock(); |
cb703454 | 391 | |
11f8b2b6 | 392 | return disk_state; |
cb703454 PR |
393 | } |
394 | ||
bde89a9e | 395 | enum drbd_disk_state conn_lowest_disk(struct drbd_connection *connection) |
4669265a | 396 | { |
11f8b2b6 | 397 | enum drbd_disk_state disk_state = D_MASK; |
c06ece6b | 398 | struct drbd_peer_device *peer_device; |
4669265a PR |
399 | int vnr; |
400 | ||
695d08fa | 401 | rcu_read_lock(); |
c06ece6b AG |
402 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
403 | struct drbd_device *device = peer_device->device; | |
11f8b2b6 | 404 | disk_state = min_t(enum drbd_disk_state, disk_state, device->state.disk); |
c06ece6b | 405 | } |
695d08fa | 406 | rcu_read_unlock(); |
4669265a | 407 | |
11f8b2b6 | 408 | return disk_state; |
4669265a PR |
409 | } |
410 | ||
bde89a9e | 411 | enum drbd_disk_state conn_highest_pdsk(struct drbd_connection *connection) |
cb703454 | 412 | { |
11f8b2b6 | 413 | enum drbd_disk_state disk_state = D_DISKLESS; |
c06ece6b | 414 | struct drbd_peer_device *peer_device; |
cb703454 PR |
415 | int vnr; |
416 | ||
695d08fa | 417 | rcu_read_lock(); |
c06ece6b AG |
418 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
419 | struct drbd_device *device = peer_device->device; | |
11f8b2b6 | 420 | disk_state = max_t(enum drbd_disk_state, disk_state, device->state.pdsk); |
c06ece6b | 421 | } |
695d08fa | 422 | rcu_read_unlock(); |
cb703454 | 423 | |
11f8b2b6 | 424 | return disk_state; |
cb703454 PR |
425 | } |
426 | ||
bde89a9e | 427 | enum drbd_conns conn_lowest_conn(struct drbd_connection *connection) |
19f83c76 PR |
428 | { |
429 | enum drbd_conns conn = C_MASK; | |
c06ece6b | 430 | struct drbd_peer_device *peer_device; |
19f83c76 PR |
431 | int vnr; |
432 | ||
695d08fa | 433 | rcu_read_lock(); |
c06ece6b AG |
434 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
435 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 436 | conn = min_t(enum drbd_conns, conn, device->state.conn); |
c06ece6b | 437 | } |
695d08fa | 438 | rcu_read_unlock(); |
19f83c76 PR |
439 | |
440 | return conn; | |
441 | } | |
442 | ||
bde89a9e | 443 | static bool no_peer_wf_report_params(struct drbd_connection *connection) |
79702011 | 444 | { |
c06ece6b | 445 | struct drbd_peer_device *peer_device; |
79702011 PR |
446 | int vnr; |
447 | bool rv = true; | |
448 | ||
449 | rcu_read_lock(); | |
c06ece6b AG |
450 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
451 | if (peer_device->device->state.conn == C_WF_REPORT_PARAMS) { | |
79702011 PR |
452 | rv = false; |
453 | break; | |
454 | } | |
455 | rcu_read_unlock(); | |
456 | ||
457 | return rv; | |
458 | } | |
459 | ||
a8821531 PR |
460 | static void wake_up_all_devices(struct drbd_connection *connection) |
461 | { | |
462 | struct drbd_peer_device *peer_device; | |
463 | int vnr; | |
464 | ||
465 | rcu_read_lock(); | |
466 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
467 | wake_up(&peer_device->device->state_wait); | |
468 | rcu_read_unlock(); | |
469 | ||
470 | } | |
471 | ||
79702011 | 472 | |
b8907339 PR |
473 | /** |
474 | * cl_wide_st_chg() - true if the state change is a cluster wide one | |
b30ab791 | 475 | * @device: DRBD device. |
b8907339 PR |
476 | * @os: old (current) state. |
477 | * @ns: new (wanted) state. | |
478 | */ | |
b30ab791 | 479 | static int cl_wide_st_chg(struct drbd_device *device, |
b8907339 PR |
480 | union drbd_state os, union drbd_state ns) |
481 | { | |
482 | return (os.conn >= C_CONNECTED && ns.conn >= C_CONNECTED && | |
483 | ((os.role != R_PRIMARY && ns.role == R_PRIMARY) || | |
484 | (os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
485 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) || | |
e8744f5a | 486 | (os.disk != D_FAILED && ns.disk == D_FAILED))) || |
b8907339 | 487 | (os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) || |
369bea63 PR |
488 | (os.conn == C_CONNECTED && ns.conn == C_VERIFY_S) || |
489 | (os.conn == C_CONNECTED && ns.conn == C_WF_REPORT_PARAMS); | |
b8907339 PR |
490 | } |
491 | ||
56707f9e PR |
492 | static union drbd_state |
493 | apply_mask_val(union drbd_state os, union drbd_state mask, union drbd_state val) | |
494 | { | |
495 | union drbd_state ns; | |
496 | ns.i = (os.i & ~mask.i) | val.i; | |
497 | return ns; | |
498 | } | |
499 | ||
b8907339 | 500 | enum drbd_state_rv |
b30ab791 | 501 | drbd_change_state(struct drbd_device *device, enum chg_state_flags f, |
b8907339 PR |
502 | union drbd_state mask, union drbd_state val) |
503 | { | |
504 | unsigned long flags; | |
56707f9e | 505 | union drbd_state ns; |
b8907339 PR |
506 | enum drbd_state_rv rv; |
507 | ||
0500813f | 508 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
509 | ns = apply_mask_val(drbd_read_state(device), mask, val); |
510 | rv = _drbd_set_state(device, ns, f, NULL); | |
0500813f | 511 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
512 | |
513 | return rv; | |
514 | } | |
515 | ||
516 | /** | |
517 | * drbd_force_state() - Impose a change which happens outside our control on our state | |
b30ab791 | 518 | * @device: DRBD device. |
b8907339 PR |
519 | * @mask: mask of state bits to change. |
520 | * @val: value of new state bits. | |
521 | */ | |
b30ab791 | 522 | void drbd_force_state(struct drbd_device *device, |
b8907339 PR |
523 | union drbd_state mask, union drbd_state val) |
524 | { | |
b30ab791 | 525 | drbd_change_state(device, CS_HARD, mask, val); |
b8907339 PR |
526 | } |
527 | ||
528 | static enum drbd_state_rv | |
b30ab791 | 529 | _req_st_cond(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
530 | union drbd_state val) |
531 | { | |
532 | union drbd_state os, ns; | |
533 | unsigned long flags; | |
534 | enum drbd_state_rv rv; | |
535 | ||
b30ab791 | 536 | if (test_and_clear_bit(CL_ST_CHG_SUCCESS, &device->flags)) |
b8907339 PR |
537 | return SS_CW_SUCCESS; |
538 | ||
b30ab791 | 539 | if (test_and_clear_bit(CL_ST_CHG_FAIL, &device->flags)) |
b8907339 PR |
540 | return SS_CW_FAILED_BY_PEER; |
541 | ||
0500813f | 542 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 543 | os = drbd_read_state(device); |
d7fe69c6 | 544 | ns = sanitize_state(device, os, apply_mask_val(os, mask, val), NULL); |
3509502d | 545 | rv = is_valid_transition(os, ns); |
a3025a27 | 546 | if (rv >= SS_SUCCESS) |
3509502d | 547 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ |
b8907339 | 548 | |
b30ab791 | 549 | if (!cl_wide_st_chg(device, os, ns)) |
b8907339 | 550 | rv = SS_CW_NO_NEED; |
3509502d | 551 | if (rv == SS_UNKNOWN_ERROR) { |
b30ab791 | 552 | rv = is_valid_state(device, ns); |
a3025a27 | 553 | if (rv >= SS_SUCCESS) { |
a6b32bc3 | 554 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
a3025a27 | 555 | if (rv >= SS_SUCCESS) |
b8907339 PR |
556 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ |
557 | } | |
558 | } | |
0500813f | 559 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
560 | |
561 | return rv; | |
562 | } | |
563 | ||
564 | /** | |
565 | * drbd_req_state() - Perform an eventually cluster wide state change | |
b30ab791 | 566 | * @device: DRBD device. |
b8907339 PR |
567 | * @mask: mask of state bits to change. |
568 | * @val: value of new state bits. | |
569 | * @f: flags | |
570 | * | |
571 | * Should not be called directly, use drbd_request_state() or | |
572 | * _drbd_request_state(). | |
573 | */ | |
574 | static enum drbd_state_rv | |
b30ab791 | 575 | drbd_req_state(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
576 | union drbd_state val, enum chg_state_flags f) |
577 | { | |
578 | struct completion done; | |
579 | unsigned long flags; | |
580 | union drbd_state os, ns; | |
581 | enum drbd_state_rv rv; | |
582 | ||
583 | init_completion(&done); | |
584 | ||
585 | if (f & CS_SERIALIZE) | |
b30ab791 | 586 | mutex_lock(device->state_mutex); |
b8907339 | 587 | |
0500813f | 588 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 589 | os = drbd_read_state(device); |
d7fe69c6 | 590 | ns = sanitize_state(device, os, apply_mask_val(os, mask, val), NULL); |
3509502d | 591 | rv = is_valid_transition(os, ns); |
3c5e5f6a | 592 | if (rv < SS_SUCCESS) { |
0500813f | 593 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
3509502d | 594 | goto abort; |
3c5e5f6a | 595 | } |
b8907339 | 596 | |
b30ab791 AG |
597 | if (cl_wide_st_chg(device, os, ns)) { |
598 | rv = is_valid_state(device, ns); | |
b8907339 | 599 | if (rv == SS_SUCCESS) |
a6b32bc3 | 600 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
0500813f | 601 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
602 | |
603 | if (rv < SS_SUCCESS) { | |
604 | if (f & CS_VERBOSE) | |
b30ab791 | 605 | print_st_err(device, os, ns, rv); |
b8907339 PR |
606 | goto abort; |
607 | } | |
608 | ||
69a22773 | 609 | if (drbd_send_state_req(first_peer_device(device), mask, val)) { |
b8907339 PR |
610 | rv = SS_CW_FAILED_BY_PEER; |
611 | if (f & CS_VERBOSE) | |
b30ab791 | 612 | print_st_err(device, os, ns, rv); |
b8907339 PR |
613 | goto abort; |
614 | } | |
615 | ||
b30ab791 AG |
616 | wait_event(device->state_wait, |
617 | (rv = _req_st_cond(device, mask, val))); | |
b8907339 PR |
618 | |
619 | if (rv < SS_SUCCESS) { | |
b8907339 | 620 | if (f & CS_VERBOSE) |
b30ab791 | 621 | print_st_err(device, os, ns, rv); |
b8907339 PR |
622 | goto abort; |
623 | } | |
0500813f | 624 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
625 | ns = apply_mask_val(drbd_read_state(device), mask, val); |
626 | rv = _drbd_set_state(device, ns, f, &done); | |
b8907339 | 627 | } else { |
b30ab791 | 628 | rv = _drbd_set_state(device, ns, f, &done); |
b8907339 PR |
629 | } |
630 | ||
0500813f | 631 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
632 | |
633 | if (f & CS_WAIT_COMPLETE && rv == SS_SUCCESS) { | |
0b0ba1ef | 634 | D_ASSERT(device, current != first_peer_device(device)->connection->worker.task); |
b8907339 PR |
635 | wait_for_completion(&done); |
636 | } | |
637 | ||
638 | abort: | |
639 | if (f & CS_SERIALIZE) | |
b30ab791 | 640 | mutex_unlock(device->state_mutex); |
b8907339 PR |
641 | |
642 | return rv; | |
643 | } | |
644 | ||
645 | /** | |
646 | * _drbd_request_state() - Request a state change (with flags) | |
b30ab791 | 647 | * @device: DRBD device. |
b8907339 PR |
648 | * @mask: mask of state bits to change. |
649 | * @val: value of new state bits. | |
650 | * @f: flags | |
651 | * | |
652 | * Cousin of drbd_request_state(), useful with the CS_WAIT_COMPLETE | |
653 | * flag, or when logging of failed state change requests is not desired. | |
654 | */ | |
655 | enum drbd_state_rv | |
b30ab791 | 656 | _drbd_request_state(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
657 | union drbd_state val, enum chg_state_flags f) |
658 | { | |
659 | enum drbd_state_rv rv; | |
660 | ||
b30ab791 AG |
661 | wait_event(device->state_wait, |
662 | (rv = drbd_req_state(device, mask, val, f)) != SS_IN_TRANSIENT_STATE); | |
b8907339 PR |
663 | |
664 | return rv; | |
665 | } | |
666 | ||
a8821531 PR |
667 | enum drbd_state_rv |
668 | _drbd_request_state_holding_state_mutex(struct drbd_device *device, union drbd_state mask, | |
669 | union drbd_state val, enum chg_state_flags f) | |
670 | { | |
671 | enum drbd_state_rv rv; | |
672 | ||
673 | BUG_ON(f & CS_SERIALIZE); | |
674 | ||
675 | wait_event_cmd(device->state_wait, | |
676 | (rv = drbd_req_state(device, mask, val, f)) != SS_IN_TRANSIENT_STATE, | |
677 | mutex_unlock(device->state_mutex), | |
678 | mutex_lock(device->state_mutex)); | |
679 | ||
680 | return rv; | |
681 | } | |
682 | ||
8ce953aa | 683 | static void print_st(struct drbd_device *device, const char *name, union drbd_state ns) |
b8907339 | 684 | { |
d0180171 | 685 | drbd_err(device, " %s = { cs:%s ro:%s/%s ds:%s/%s %c%c%c%c%c%c }\n", |
b8907339 PR |
686 | name, |
687 | drbd_conn_str(ns.conn), | |
688 | drbd_role_str(ns.role), | |
689 | drbd_role_str(ns.peer), | |
690 | drbd_disk_str(ns.disk), | |
691 | drbd_disk_str(ns.pdsk), | |
692 | is_susp(ns) ? 's' : 'r', | |
693 | ns.aftr_isp ? 'a' : '-', | |
694 | ns.peer_isp ? 'p' : '-', | |
695 | ns.user_isp ? 'u' : '-', | |
696 | ns.susp_fen ? 'F' : '-', | |
697 | ns.susp_nod ? 'N' : '-' | |
698 | ); | |
699 | } | |
700 | ||
b30ab791 | 701 | void print_st_err(struct drbd_device *device, union drbd_state os, |
b8907339 PR |
702 | union drbd_state ns, enum drbd_state_rv err) |
703 | { | |
704 | if (err == SS_IN_TRANSIENT_STATE) | |
705 | return; | |
d0180171 | 706 | drbd_err(device, "State change failed: %s\n", drbd_set_st_err_str(err)); |
b30ab791 AG |
707 | print_st(device, " state", os); |
708 | print_st(device, "wanted", ns); | |
b8907339 PR |
709 | } |
710 | ||
435693e8 | 711 | static long print_state_change(char *pb, union drbd_state os, union drbd_state ns, |
bbeb641c PR |
712 | enum chg_state_flags flags) |
713 | { | |
435693e8 | 714 | char *pbp; |
bbeb641c PR |
715 | pbp = pb; |
716 | *pbp = 0; | |
706cb24c | 717 | |
435693e8 | 718 | if (ns.role != os.role && flags & CS_DC_ROLE) |
bbeb641c PR |
719 | pbp += sprintf(pbp, "role( %s -> %s ) ", |
720 | drbd_role_str(os.role), | |
721 | drbd_role_str(ns.role)); | |
435693e8 | 722 | if (ns.peer != os.peer && flags & CS_DC_PEER) |
bbeb641c PR |
723 | pbp += sprintf(pbp, "peer( %s -> %s ) ", |
724 | drbd_role_str(os.peer), | |
725 | drbd_role_str(ns.peer)); | |
435693e8 | 726 | if (ns.conn != os.conn && flags & CS_DC_CONN) |
bbeb641c PR |
727 | pbp += sprintf(pbp, "conn( %s -> %s ) ", |
728 | drbd_conn_str(os.conn), | |
729 | drbd_conn_str(ns.conn)); | |
435693e8 | 730 | if (ns.disk != os.disk && flags & CS_DC_DISK) |
bbeb641c PR |
731 | pbp += sprintf(pbp, "disk( %s -> %s ) ", |
732 | drbd_disk_str(os.disk), | |
733 | drbd_disk_str(ns.disk)); | |
435693e8 | 734 | if (ns.pdsk != os.pdsk && flags & CS_DC_PDSK) |
bbeb641c PR |
735 | pbp += sprintf(pbp, "pdsk( %s -> %s ) ", |
736 | drbd_disk_str(os.pdsk), | |
737 | drbd_disk_str(ns.pdsk)); | |
706cb24c PR |
738 | |
739 | return pbp - pb; | |
740 | } | |
741 | ||
b30ab791 | 742 | static void drbd_pr_state_change(struct drbd_device *device, union drbd_state os, union drbd_state ns, |
706cb24c PR |
743 | enum chg_state_flags flags) |
744 | { | |
745 | char pb[300]; | |
746 | char *pbp = pb; | |
747 | ||
748 | pbp += print_state_change(pbp, os, ns, flags ^ CS_DC_MASK); | |
749 | ||
bbeb641c PR |
750 | if (ns.aftr_isp != os.aftr_isp) |
751 | pbp += sprintf(pbp, "aftr_isp( %d -> %d ) ", | |
752 | os.aftr_isp, | |
753 | ns.aftr_isp); | |
754 | if (ns.peer_isp != os.peer_isp) | |
755 | pbp += sprintf(pbp, "peer_isp( %d -> %d ) ", | |
756 | os.peer_isp, | |
757 | ns.peer_isp); | |
758 | if (ns.user_isp != os.user_isp) | |
759 | pbp += sprintf(pbp, "user_isp( %d -> %d ) ", | |
760 | os.user_isp, | |
761 | ns.user_isp); | |
435693e8 | 762 | |
706cb24c | 763 | if (pbp != pb) |
d0180171 | 764 | drbd_info(device, "%s\n", pb); |
bbeb641c | 765 | } |
b8907339 | 766 | |
bde89a9e | 767 | static void conn_pr_state_change(struct drbd_connection *connection, union drbd_state os, union drbd_state ns, |
435693e8 PR |
768 | enum chg_state_flags flags) |
769 | { | |
770 | char pb[300]; | |
706cb24c PR |
771 | char *pbp = pb; |
772 | ||
773 | pbp += print_state_change(pbp, os, ns, flags); | |
435693e8 | 774 | |
706cb24c PR |
775 | if (is_susp(ns) != is_susp(os) && flags & CS_DC_SUSP) |
776 | pbp += sprintf(pbp, "susp( %d -> %d ) ", | |
777 | is_susp(os), | |
778 | is_susp(ns)); | |
779 | ||
780 | if (pbp != pb) | |
1ec861eb | 781 | drbd_info(connection, "%s\n", pb); |
435693e8 PR |
782 | } |
783 | ||
784 | ||
b8907339 PR |
785 | /** |
786 | * is_valid_state() - Returns an SS_ error code if ns is not valid | |
b30ab791 | 787 | * @device: DRBD device. |
b8907339 PR |
788 | * @ns: State to consider. |
789 | */ | |
790 | static enum drbd_state_rv | |
b30ab791 | 791 | is_valid_state(struct drbd_device *device, union drbd_state ns) |
b8907339 PR |
792 | { |
793 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
794 | ||
795 | enum drbd_fencing_p fp; | |
796 | enum drbd_state_rv rv = SS_SUCCESS; | |
44ed167d | 797 | struct net_conf *nc; |
b8907339 | 798 | |
daeda1cc | 799 | rcu_read_lock(); |
b8907339 | 800 | fp = FP_DONT_CARE; |
b30ab791 AG |
801 | if (get_ldev(device)) { |
802 | fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
803 | put_ldev(device); | |
b8907339 PR |
804 | } |
805 | ||
a6b32bc3 | 806 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
807 | if (nc) { |
808 | if (!nc->two_primaries && ns.role == R_PRIMARY) { | |
047e95e2 PR |
809 | if (ns.peer == R_PRIMARY) |
810 | rv = SS_TWO_PRIMARIES; | |
a6b32bc3 | 811 | else if (conn_highest_peer(first_peer_device(device)->connection) == R_PRIMARY) |
047e95e2 | 812 | rv = SS_O_VOL_PEER_PRI; |
44ed167d | 813 | } |
b8907339 PR |
814 | } |
815 | ||
816 | if (rv <= 0) | |
4e526a00 | 817 | goto out; /* already found a reason to abort */ |
b30ab791 | 818 | else if (ns.role == R_SECONDARY && device->open_cnt) |
b8907339 PR |
819 | rv = SS_DEVICE_IN_USE; |
820 | ||
821 | else if (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.disk < D_UP_TO_DATE) | |
822 | rv = SS_NO_UP_TO_DATE_DISK; | |
823 | ||
824 | else if (fp >= FP_RESOURCE && | |
825 | ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk >= D_UNKNOWN) | |
826 | rv = SS_PRIMARY_NOP; | |
827 | ||
828 | else if (ns.role == R_PRIMARY && ns.disk <= D_INCONSISTENT && ns.pdsk <= D_INCONSISTENT) | |
829 | rv = SS_NO_UP_TO_DATE_DISK; | |
830 | ||
831 | else if (ns.conn > C_CONNECTED && ns.disk < D_INCONSISTENT) | |
832 | rv = SS_NO_LOCAL_DISK; | |
833 | ||
834 | else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT) | |
835 | rv = SS_NO_REMOTE_DISK; | |
836 | ||
837 | else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) | |
838 | rv = SS_NO_UP_TO_DATE_DISK; | |
839 | ||
840 | else if ((ns.conn == C_CONNECTED || | |
841 | ns.conn == C_WF_BITMAP_S || | |
842 | ns.conn == C_SYNC_SOURCE || | |
843 | ns.conn == C_PAUSED_SYNC_S) && | |
844 | ns.disk == D_OUTDATED) | |
845 | rv = SS_CONNECTED_OUTDATES; | |
846 | ||
847 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
44ed167d | 848 | (nc->verify_alg[0] == 0)) |
b8907339 PR |
849 | rv = SS_NO_VERIFY_ALG; |
850 | ||
851 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
a6b32bc3 | 852 | first_peer_device(device)->connection->agreed_pro_version < 88) |
b8907339 PR |
853 | rv = SS_NOT_SUPPORTED; |
854 | ||
5c4f13d9 PR |
855 | else if (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) |
856 | rv = SS_NO_UP_TO_DATE_DISK; | |
857 | ||
858 | else if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
859 | ns.pdsk == D_UNKNOWN) | |
860 | rv = SS_NEED_CONNECTION; | |
861 | ||
b8907339 PR |
862 | else if (ns.conn >= C_CONNECTED && ns.pdsk == D_UNKNOWN) |
863 | rv = SS_CONNECTED_OUTDATES; | |
864 | ||
4e526a00 | 865 | out: |
44ed167d PR |
866 | rcu_read_unlock(); |
867 | ||
b8907339 PR |
868 | return rv; |
869 | } | |
870 | ||
871 | /** | |
a75f34ad | 872 | * is_valid_soft_transition() - Returns an SS_ error code if the state transition is not possible |
3509502d PR |
873 | * This function limits state transitions that may be declined by DRBD. I.e. |
874 | * user requests (aka soft transitions). | |
b30ab791 | 875 | * @device: DRBD device. |
b8907339 PR |
876 | * @ns: new state. |
877 | * @os: old state. | |
878 | */ | |
879 | static enum drbd_state_rv | |
bde89a9e | 880 | is_valid_soft_transition(union drbd_state os, union drbd_state ns, struct drbd_connection *connection) |
b8907339 PR |
881 | { |
882 | enum drbd_state_rv rv = SS_SUCCESS; | |
883 | ||
884 | if ((ns.conn == C_STARTING_SYNC_T || ns.conn == C_STARTING_SYNC_S) && | |
885 | os.conn > C_CONNECTED) | |
886 | rv = SS_RESYNC_RUNNING; | |
887 | ||
888 | if (ns.conn == C_DISCONNECTING && os.conn == C_STANDALONE) | |
889 | rv = SS_ALREADY_STANDALONE; | |
890 | ||
891 | if (ns.disk > D_ATTACHING && os.disk == D_DISKLESS) | |
892 | rv = SS_IS_DISKLESS; | |
893 | ||
894 | if (ns.conn == C_WF_CONNECTION && os.conn < C_UNCONNECTED) | |
895 | rv = SS_NO_NET_CONFIG; | |
896 | ||
897 | if (ns.disk == D_OUTDATED && os.disk < D_OUTDATED && os.disk != D_ATTACHING) | |
898 | rv = SS_LOWER_THAN_OUTDATED; | |
899 | ||
900 | if (ns.conn == C_DISCONNECTING && os.conn == C_UNCONNECTED) | |
901 | rv = SS_IN_TRANSIENT_STATE; | |
902 | ||
a1096a6e | 903 | /* While establishing a connection only allow cstate to change. |
a8821531 | 904 | Delay/refuse role changes, detach attach etc... (they do not touch cstate) */ |
bde89a9e | 905 | if (test_bit(STATE_SENT, &connection->flags) && |
a8821531 PR |
906 | !((ns.conn == C_WF_REPORT_PARAMS && os.conn == C_WF_CONNECTION) || |
907 | (ns.conn >= C_CONNECTED && os.conn == C_WF_REPORT_PARAMS))) | |
a1096a6e PR |
908 | rv = SS_IN_TRANSIENT_STATE; |
909 | ||
31d64604 LE |
910 | /* Do not promote during resync handshake triggered by "force primary". |
911 | * This is a hack. It should really be rejected by the peer during the | |
912 | * cluster wide state change request. */ | |
913 | if (os.role != R_PRIMARY && ns.role == R_PRIMARY | |
914 | && ns.pdsk == D_UP_TO_DATE | |
915 | && ns.disk != D_UP_TO_DATE && ns.disk != D_DISKLESS | |
916 | && (ns.conn <= C_WF_SYNC_UUID || ns.conn != os.conn)) | |
917 | rv = SS_IN_TRANSIENT_STATE; | |
918 | ||
b8907339 PR |
919 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED) |
920 | rv = SS_NEED_CONNECTION; | |
921 | ||
922 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
923 | ns.conn != os.conn && os.conn > C_CONNECTED) | |
924 | rv = SS_RESYNC_RUNNING; | |
925 | ||
926 | if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
927 | os.conn < C_CONNECTED) | |
928 | rv = SS_NEED_CONNECTION; | |
929 | ||
930 | if ((ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE) | |
931 | && os.conn < C_WF_REPORT_PARAMS) | |
932 | rv = SS_NEED_CONNECTION; /* No NetworkFailure -> SyncTarget etc... */ | |
933 | ||
2bd5ed5d PR |
934 | if (ns.conn == C_DISCONNECTING && ns.pdsk == D_OUTDATED && |
935 | os.conn < C_CONNECTED && os.pdsk > D_OUTDATED) | |
936 | rv = SS_OUTDATE_WO_CONN; | |
937 | ||
b8907339 PR |
938 | return rv; |
939 | } | |
940 | ||
3509502d | 941 | static enum drbd_state_rv |
fda74117 | 942 | is_valid_conn_transition(enum drbd_conns oc, enum drbd_conns nc) |
3509502d | 943 | { |
d9cc6e23 LE |
944 | /* no change -> nothing to do, at least for the connection part */ |
945 | if (oc == nc) | |
946 | return SS_NOTHING_TO_DO; | |
3509502d | 947 | |
d9cc6e23 LE |
948 | /* disconnect of an unconfigured connection does not make sense */ |
949 | if (oc == C_STANDALONE && nc == C_DISCONNECTING) | |
950 | return SS_ALREADY_STANDALONE; | |
951 | ||
952 | /* from C_STANDALONE, we start with C_UNCONNECTED */ | |
953 | if (oc == C_STANDALONE && nc != C_UNCONNECTED) | |
954 | return SS_NEED_CONNECTION; | |
3509502d | 955 | |
25b0d6c8 PR |
956 | /* When establishing a connection we need to go through WF_REPORT_PARAMS! |
957 | Necessary to do the right thing upon invalidate-remote on a disconnected resource */ | |
958 | if (oc < C_WF_REPORT_PARAMS && nc >= C_CONNECTED) | |
959 | return SS_NEED_CONNECTION; | |
960 | ||
3509502d | 961 | /* After a network error only C_UNCONNECTED or C_DISCONNECTING may follow. */ |
fda74117 | 962 | if (oc >= C_TIMEOUT && oc <= C_TEAR_DOWN && nc != C_UNCONNECTED && nc != C_DISCONNECTING) |
d9cc6e23 | 963 | return SS_IN_TRANSIENT_STATE; |
3509502d PR |
964 | |
965 | /* After C_DISCONNECTING only C_STANDALONE may follow */ | |
fda74117 | 966 | if (oc == C_DISCONNECTING && nc != C_STANDALONE) |
d9cc6e23 | 967 | return SS_IN_TRANSIENT_STATE; |
3509502d | 968 | |
d9cc6e23 | 969 | return SS_SUCCESS; |
fda74117 PR |
970 | } |
971 | ||
972 | ||
973 | /** | |
974 | * is_valid_transition() - Returns an SS_ error code if the state transition is not possible | |
975 | * This limits hard state transitions. Hard state transitions are facts there are | |
976 | * imposed on DRBD by the environment. E.g. disk broke or network broke down. | |
977 | * But those hard state transitions are still not allowed to do everything. | |
978 | * @ns: new state. | |
979 | * @os: old state. | |
980 | */ | |
981 | static enum drbd_state_rv | |
982 | is_valid_transition(union drbd_state os, union drbd_state ns) | |
983 | { | |
984 | enum drbd_state_rv rv; | |
985 | ||
986 | rv = is_valid_conn_transition(os.conn, ns.conn); | |
987 | ||
3509502d PR |
988 | /* we cannot fail (again) if we already detached */ |
989 | if (ns.disk == D_FAILED && os.disk == D_DISKLESS) | |
990 | rv = SS_IS_DISKLESS; | |
991 | ||
992 | return rv; | |
993 | } | |
994 | ||
b30ab791 | 995 | static void print_sanitize_warnings(struct drbd_device *device, enum sanitize_state_warnings warn) |
d942ae44 PR |
996 | { |
997 | static const char *msg_table[] = { | |
998 | [NO_WARNING] = "", | |
999 | [ABORTED_ONLINE_VERIFY] = "Online-verify aborted.", | |
1000 | [ABORTED_RESYNC] = "Resync aborted.", | |
1001 | [CONNECTION_LOST_NEGOTIATING] = "Connection lost while negotiating, no data!", | |
1002 | [IMPLICITLY_UPGRADED_DISK] = "Implicitly upgraded disk", | |
1003 | [IMPLICITLY_UPGRADED_PDSK] = "Implicitly upgraded pdsk", | |
1004 | }; | |
1005 | ||
1006 | if (warn != NO_WARNING) | |
d0180171 | 1007 | drbd_warn(device, "%s\n", msg_table[warn]); |
d942ae44 PR |
1008 | } |
1009 | ||
b8907339 PR |
1010 | /** |
1011 | * sanitize_state() - Resolves implicitly necessary additional changes to a state transition | |
b30ab791 | 1012 | * @device: DRBD device. |
b8907339 PR |
1013 | * @os: old state. |
1014 | * @ns: new state. | |
1015 | * @warn_sync_abort: | |
1016 | * | |
1017 | * When we loose connection, we have to set the state of the peers disk (pdsk) | |
1018 | * to D_UNKNOWN. This rule and many more along those lines are in this function. | |
1019 | */ | |
d7fe69c6 PR |
1020 | static union drbd_state sanitize_state(struct drbd_device *device, union drbd_state os, |
1021 | union drbd_state ns, enum sanitize_state_warnings *warn) | |
b8907339 PR |
1022 | { |
1023 | enum drbd_fencing_p fp; | |
1024 | enum drbd_disk_state disk_min, disk_max, pdsk_min, pdsk_max; | |
1025 | ||
d942ae44 PR |
1026 | if (warn) |
1027 | *warn = NO_WARNING; | |
1028 | ||
b8907339 | 1029 | fp = FP_DONT_CARE; |
b30ab791 | 1030 | if (get_ldev(device)) { |
daeda1cc | 1031 | rcu_read_lock(); |
b30ab791 | 1032 | fp = rcu_dereference(device->ldev->disk_conf)->fencing; |
daeda1cc | 1033 | rcu_read_unlock(); |
b30ab791 | 1034 | put_ldev(device); |
b8907339 PR |
1035 | } |
1036 | ||
3509502d | 1037 | /* Implications from connection to peer and peer_isp */ |
b8907339 PR |
1038 | if (ns.conn < C_CONNECTED) { |
1039 | ns.peer_isp = 0; | |
1040 | ns.peer = R_UNKNOWN; | |
1041 | if (ns.pdsk > D_UNKNOWN || ns.pdsk < D_INCONSISTENT) | |
1042 | ns.pdsk = D_UNKNOWN; | |
1043 | } | |
1044 | ||
1045 | /* Clear the aftr_isp when becoming unconfigured */ | |
1046 | if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY) | |
1047 | ns.aftr_isp = 0; | |
1048 | ||
4308a0a3 | 1049 | /* An implication of the disk states onto the connection state */ |
b8907339 | 1050 | /* Abort resync if a disk fails/detaches */ |
4308a0a3 | 1051 | if (ns.conn > C_CONNECTED && (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) { |
d942ae44 PR |
1052 | if (warn) |
1053 | *warn = ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T ? | |
1054 | ABORTED_ONLINE_VERIFY : ABORTED_RESYNC; | |
b8907339 PR |
1055 | ns.conn = C_CONNECTED; |
1056 | } | |
1057 | ||
1058 | /* Connection breaks down before we finished "Negotiating" */ | |
1059 | if (ns.conn < C_CONNECTED && ns.disk == D_NEGOTIATING && | |
b30ab791 AG |
1060 | get_ldev_if_state(device, D_NEGOTIATING)) { |
1061 | if (device->ed_uuid == device->ldev->md.uuid[UI_CURRENT]) { | |
1062 | ns.disk = device->new_state_tmp.disk; | |
1063 | ns.pdsk = device->new_state_tmp.pdsk; | |
b8907339 | 1064 | } else { |
d942ae44 PR |
1065 | if (warn) |
1066 | *warn = CONNECTION_LOST_NEGOTIATING; | |
b8907339 PR |
1067 | ns.disk = D_DISKLESS; |
1068 | ns.pdsk = D_UNKNOWN; | |
1069 | } | |
b30ab791 | 1070 | put_ldev(device); |
b8907339 PR |
1071 | } |
1072 | ||
1073 | /* D_CONSISTENT and D_OUTDATED vanish when we get connected */ | |
1074 | if (ns.conn >= C_CONNECTED && ns.conn < C_AHEAD) { | |
1075 | if (ns.disk == D_CONSISTENT || ns.disk == D_OUTDATED) | |
1076 | ns.disk = D_UP_TO_DATE; | |
1077 | if (ns.pdsk == D_CONSISTENT || ns.pdsk == D_OUTDATED) | |
1078 | ns.pdsk = D_UP_TO_DATE; | |
1079 | } | |
1080 | ||
1081 | /* Implications of the connection stat on the disk states */ | |
1082 | disk_min = D_DISKLESS; | |
1083 | disk_max = D_UP_TO_DATE; | |
1084 | pdsk_min = D_INCONSISTENT; | |
1085 | pdsk_max = D_UNKNOWN; | |
1086 | switch ((enum drbd_conns)ns.conn) { | |
1087 | case C_WF_BITMAP_T: | |
1088 | case C_PAUSED_SYNC_T: | |
1089 | case C_STARTING_SYNC_T: | |
1090 | case C_WF_SYNC_UUID: | |
1091 | case C_BEHIND: | |
1092 | disk_min = D_INCONSISTENT; | |
1093 | disk_max = D_OUTDATED; | |
1094 | pdsk_min = D_UP_TO_DATE; | |
1095 | pdsk_max = D_UP_TO_DATE; | |
1096 | break; | |
1097 | case C_VERIFY_S: | |
1098 | case C_VERIFY_T: | |
1099 | disk_min = D_UP_TO_DATE; | |
1100 | disk_max = D_UP_TO_DATE; | |
1101 | pdsk_min = D_UP_TO_DATE; | |
1102 | pdsk_max = D_UP_TO_DATE; | |
1103 | break; | |
1104 | case C_CONNECTED: | |
1105 | disk_min = D_DISKLESS; | |
1106 | disk_max = D_UP_TO_DATE; | |
1107 | pdsk_min = D_DISKLESS; | |
1108 | pdsk_max = D_UP_TO_DATE; | |
1109 | break; | |
1110 | case C_WF_BITMAP_S: | |
1111 | case C_PAUSED_SYNC_S: | |
1112 | case C_STARTING_SYNC_S: | |
1113 | case C_AHEAD: | |
1114 | disk_min = D_UP_TO_DATE; | |
1115 | disk_max = D_UP_TO_DATE; | |
1116 | pdsk_min = D_INCONSISTENT; | |
1117 | pdsk_max = D_CONSISTENT; /* D_OUTDATED would be nice. But explicit outdate necessary*/ | |
1118 | break; | |
1119 | case C_SYNC_TARGET: | |
1120 | disk_min = D_INCONSISTENT; | |
1121 | disk_max = D_INCONSISTENT; | |
1122 | pdsk_min = D_UP_TO_DATE; | |
1123 | pdsk_max = D_UP_TO_DATE; | |
1124 | break; | |
1125 | case C_SYNC_SOURCE: | |
1126 | disk_min = D_UP_TO_DATE; | |
1127 | disk_max = D_UP_TO_DATE; | |
1128 | pdsk_min = D_INCONSISTENT; | |
1129 | pdsk_max = D_INCONSISTENT; | |
1130 | break; | |
1131 | case C_STANDALONE: | |
1132 | case C_DISCONNECTING: | |
1133 | case C_UNCONNECTED: | |
1134 | case C_TIMEOUT: | |
1135 | case C_BROKEN_PIPE: | |
1136 | case C_NETWORK_FAILURE: | |
1137 | case C_PROTOCOL_ERROR: | |
1138 | case C_TEAR_DOWN: | |
1139 | case C_WF_CONNECTION: | |
1140 | case C_WF_REPORT_PARAMS: | |
1141 | case C_MASK: | |
1142 | break; | |
1143 | } | |
1144 | if (ns.disk > disk_max) | |
1145 | ns.disk = disk_max; | |
1146 | ||
1147 | if (ns.disk < disk_min) { | |
d942ae44 PR |
1148 | if (warn) |
1149 | *warn = IMPLICITLY_UPGRADED_DISK; | |
b8907339 PR |
1150 | ns.disk = disk_min; |
1151 | } | |
1152 | if (ns.pdsk > pdsk_max) | |
1153 | ns.pdsk = pdsk_max; | |
1154 | ||
1155 | if (ns.pdsk < pdsk_min) { | |
d942ae44 PR |
1156 | if (warn) |
1157 | *warn = IMPLICITLY_UPGRADED_PDSK; | |
b8907339 PR |
1158 | ns.pdsk = pdsk_min; |
1159 | } | |
1160 | ||
1161 | if (fp == FP_STONITH && | |
d7fe69c6 PR |
1162 | (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED) && |
1163 | !(os.role == R_PRIMARY && os.conn < C_CONNECTED && os.pdsk > D_OUTDATED)) | |
b8907339 PR |
1164 | ns.susp_fen = 1; /* Suspend IO while fence-peer handler runs (peer lost) */ |
1165 | ||
eb6bea67 | 1166 | if (device->resource->res_opts.on_no_data == OND_SUSPEND_IO && |
d7fe69c6 PR |
1167 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) && |
1168 | !(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE)) | |
b8907339 PR |
1169 | ns.susp_nod = 1; /* Suspend IO while no data available (no accessible data available) */ |
1170 | ||
1171 | if (ns.aftr_isp || ns.peer_isp || ns.user_isp) { | |
1172 | if (ns.conn == C_SYNC_SOURCE) | |
1173 | ns.conn = C_PAUSED_SYNC_S; | |
1174 | if (ns.conn == C_SYNC_TARGET) | |
1175 | ns.conn = C_PAUSED_SYNC_T; | |
1176 | } else { | |
1177 | if (ns.conn == C_PAUSED_SYNC_S) | |
1178 | ns.conn = C_SYNC_SOURCE; | |
1179 | if (ns.conn == C_PAUSED_SYNC_T) | |
1180 | ns.conn = C_SYNC_TARGET; | |
1181 | } | |
1182 | ||
1183 | return ns; | |
1184 | } | |
1185 | ||
b30ab791 | 1186 | void drbd_resume_al(struct drbd_device *device) |
b8907339 | 1187 | { |
b30ab791 | 1188 | if (test_and_clear_bit(AL_SUSPENDED, &device->flags)) |
d0180171 | 1189 | drbd_info(device, "Resumed AL updates\n"); |
b8907339 PR |
1190 | } |
1191 | ||
28bc3b8c | 1192 | /* helper for _drbd_set_state */ |
b30ab791 | 1193 | static void set_ov_position(struct drbd_device *device, enum drbd_conns cs) |
b8907339 | 1194 | { |
a6b32bc3 | 1195 | if (first_peer_device(device)->connection->agreed_pro_version < 90) |
b30ab791 AG |
1196 | device->ov_start_sector = 0; |
1197 | device->rs_total = drbd_bm_bits(device); | |
1198 | device->ov_position = 0; | |
b8907339 PR |
1199 | if (cs == C_VERIFY_T) { |
1200 | /* starting online verify from an arbitrary position | |
1201 | * does not fit well into the existing protocol. | |
1202 | * on C_VERIFY_T, we initialize ov_left and friends | |
1203 | * implicitly in receive_DataRequest once the | |
1204 | * first P_OV_REQUEST is received */ | |
b30ab791 | 1205 | device->ov_start_sector = ~(sector_t)0; |
b8907339 | 1206 | } else { |
b30ab791 AG |
1207 | unsigned long bit = BM_SECT_TO_BIT(device->ov_start_sector); |
1208 | if (bit >= device->rs_total) { | |
1209 | device->ov_start_sector = | |
1210 | BM_BIT_TO_SECT(device->rs_total - 1); | |
1211 | device->rs_total = 1; | |
b8907339 | 1212 | } else |
b30ab791 AG |
1213 | device->rs_total -= bit; |
1214 | device->ov_position = device->ov_start_sector; | |
b8907339 | 1215 | } |
b30ab791 | 1216 | device->ov_left = device->rs_total; |
b8907339 PR |
1217 | } |
1218 | ||
1219 | /** | |
28bc3b8c | 1220 | * _drbd_set_state() - Set a new DRBD state |
b30ab791 | 1221 | * @device: DRBD device. |
b8907339 PR |
1222 | * @ns: new state. |
1223 | * @flags: Flags | |
1224 | * @done: Optional completion, that will get completed after the after_state_ch() finished | |
1225 | * | |
28bc3b8c | 1226 | * Caller needs to hold req_lock. Do not call directly. |
b8907339 PR |
1227 | */ |
1228 | enum drbd_state_rv | |
28bc3b8c AG |
1229 | _drbd_set_state(struct drbd_device *device, union drbd_state ns, |
1230 | enum chg_state_flags flags, struct completion *done) | |
b8907339 | 1231 | { |
44a4d551 LE |
1232 | struct drbd_peer_device *peer_device = first_peer_device(device); |
1233 | struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; | |
b8907339 PR |
1234 | union drbd_state os; |
1235 | enum drbd_state_rv rv = SS_SUCCESS; | |
d942ae44 | 1236 | enum sanitize_state_warnings ssw; |
b8907339 | 1237 | struct after_state_chg_work *ascw; |
a2972846 | 1238 | struct drbd_state_change *state_change; |
b8907339 | 1239 | |
b30ab791 | 1240 | os = drbd_read_state(device); |
b8907339 | 1241 | |
d7fe69c6 | 1242 | ns = sanitize_state(device, os, ns, &ssw); |
b8907339 PR |
1243 | if (ns.i == os.i) |
1244 | return SS_NOTHING_TO_DO; | |
1245 | ||
3509502d PR |
1246 | rv = is_valid_transition(os, ns); |
1247 | if (rv < SS_SUCCESS) | |
1248 | return rv; | |
1249 | ||
b8907339 PR |
1250 | if (!(flags & CS_HARD)) { |
1251 | /* pre-state-change checks ; only look at ns */ | |
1252 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
1253 | ||
b30ab791 | 1254 | rv = is_valid_state(device, ns); |
b8907339 PR |
1255 | if (rv < SS_SUCCESS) { |
1256 | /* If the old state was illegal as well, then let | |
1257 | this happen...*/ | |
1258 | ||
b30ab791 | 1259 | if (is_valid_state(device, os) == rv) |
44a4d551 | 1260 | rv = is_valid_soft_transition(os, ns, connection); |
b8907339 | 1261 | } else |
44a4d551 | 1262 | rv = is_valid_soft_transition(os, ns, connection); |
b8907339 PR |
1263 | } |
1264 | ||
1265 | if (rv < SS_SUCCESS) { | |
1266 | if (flags & CS_VERBOSE) | |
b30ab791 | 1267 | print_st_err(device, os, ns, rv); |
b8907339 PR |
1268 | return rv; |
1269 | } | |
1270 | ||
b30ab791 | 1271 | print_sanitize_warnings(device, ssw); |
b8907339 | 1272 | |
b30ab791 | 1273 | drbd_pr_state_change(device, os, ns, flags); |
b8907339 | 1274 | |
706cb24c PR |
1275 | /* Display changes to the susp* flags that where caused by the call to |
1276 | sanitize_state(). Only display it here if we where not called from | |
1277 | _conn_request_state() */ | |
1278 | if (!(flags & CS_DC_SUSP)) | |
44a4d551 | 1279 | conn_pr_state_change(connection, os, ns, |
a6b32bc3 | 1280 | (flags & ~CS_DC_MASK) | CS_DC_SUSP); |
706cb24c | 1281 | |
b8907339 PR |
1282 | /* if we are going -> D_FAILED or D_DISKLESS, grab one extra reference |
1283 | * on the ldev here, to be sure the transition -> D_DISKLESS resp. | |
1284 | * drbd_ldev_destroy() won't happen before our corresponding | |
1285 | * after_state_ch works run, where we put_ldev again. */ | |
1286 | if ((os.disk != D_FAILED && ns.disk == D_FAILED) || | |
1287 | (os.disk != D_DISKLESS && ns.disk == D_DISKLESS)) | |
b30ab791 | 1288 | atomic_inc(&device->local_cnt); |
b8907339 | 1289 | |
5ab7d2c0 LE |
1290 | if (!is_sync_state(os.conn) && is_sync_state(ns.conn)) |
1291 | clear_bit(RS_DONE, &device->flags); | |
1292 | ||
a2972846 AG |
1293 | /* FIXME: Have any flags been set earlier in this function already? */ |
1294 | state_change = remember_old_state(device->resource, GFP_ATOMIC); | |
1295 | ||
ba3c6fb8 LE |
1296 | /* changes to local_cnt and device flags should be visible before |
1297 | * changes to state, which again should be visible before anything else | |
1298 | * depending on that change happens. */ | |
1299 | smp_wmb(); | |
b30ab791 | 1300 | device->state.i = ns.i; |
6bbf53ca AG |
1301 | device->resource->susp = ns.susp; |
1302 | device->resource->susp_nod = ns.susp_nod; | |
1303 | device->resource->susp_fen = ns.susp_fen; | |
ba3c6fb8 | 1304 | smp_wmb(); |
b8907339 | 1305 | |
a2972846 AG |
1306 | remember_new_state(state_change); |
1307 | ||
2681f7f6 | 1308 | /* put replicated vs not-replicated requests in seperate epochs */ |
ba3c6fb8 LE |
1309 | if (drbd_should_do_remote((union drbd_dev_state)os.i) != |
1310 | drbd_should_do_remote((union drbd_dev_state)ns.i)) | |
44a4d551 | 1311 | start_new_tl_epoch(connection); |
2681f7f6 | 1312 | |
b8907339 | 1313 | if (os.disk == D_ATTACHING && ns.disk >= D_NEGOTIATING) |
b30ab791 | 1314 | drbd_print_uuids(device, "attached to UUIDs"); |
b8907339 | 1315 | |
79702011 PR |
1316 | /* Wake up role changes, that were delayed because of connection establishing */ |
1317 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn != C_WF_REPORT_PARAMS && | |
a8821531 | 1318 | no_peer_wf_report_params(connection)) { |
44a4d551 | 1319 | clear_bit(STATE_SENT, &connection->flags); |
a8821531 PR |
1320 | wake_up_all_devices(connection); |
1321 | } | |
79702011 | 1322 | |
b30ab791 AG |
1323 | wake_up(&device->misc_wait); |
1324 | wake_up(&device->state_wait); | |
44a4d551 | 1325 | wake_up(&connection->ping_wait); |
b8907339 | 1326 | |
58ffa580 LE |
1327 | /* Aborted verify run, or we reached the stop sector. |
1328 | * Log the last position, unless end-of-device. */ | |
b8907339 | 1329 | if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) && |
58ffa580 | 1330 | ns.conn <= C_CONNECTED) { |
b30ab791 AG |
1331 | device->ov_start_sector = |
1332 | BM_BIT_TO_SECT(drbd_bm_bits(device) - device->ov_left); | |
1333 | if (device->ov_left) | |
d0180171 | 1334 | drbd_info(device, "Online Verify reached sector %llu\n", |
b30ab791 | 1335 | (unsigned long long)device->ov_start_sector); |
b8907339 PR |
1336 | } |
1337 | ||
1338 | if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) && | |
1339 | (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) { | |
d0180171 | 1340 | drbd_info(device, "Syncer continues.\n"); |
b30ab791 AG |
1341 | device->rs_paused += (long)jiffies |
1342 | -(long)device->rs_mark_time[device->rs_last_mark]; | |
b8907339 | 1343 | if (ns.conn == C_SYNC_TARGET) |
b30ab791 | 1344 | mod_timer(&device->resync_timer, jiffies); |
b8907339 PR |
1345 | } |
1346 | ||
1347 | if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) && | |
1348 | (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) { | |
d0180171 | 1349 | drbd_info(device, "Resync suspended\n"); |
b30ab791 | 1350 | device->rs_mark_time[device->rs_last_mark] = jiffies; |
b8907339 PR |
1351 | } |
1352 | ||
1353 | if (os.conn == C_CONNECTED && | |
1354 | (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) { | |
1355 | unsigned long now = jiffies; | |
1356 | int i; | |
1357 | ||
b30ab791 AG |
1358 | set_ov_position(device, ns.conn); |
1359 | device->rs_start = now; | |
b30ab791 AG |
1360 | device->rs_last_sect_ev = 0; |
1361 | device->ov_last_oos_size = 0; | |
1362 | device->ov_last_oos_start = 0; | |
b8907339 PR |
1363 | |
1364 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { | |
b30ab791 AG |
1365 | device->rs_mark_left[i] = device->ov_left; |
1366 | device->rs_mark_time[i] = now; | |
b8907339 PR |
1367 | } |
1368 | ||
b30ab791 | 1369 | drbd_rs_controller_reset(device); |
b8907339 PR |
1370 | |
1371 | if (ns.conn == C_VERIFY_S) { | |
d0180171 | 1372 | drbd_info(device, "Starting Online Verify from sector %llu\n", |
b30ab791 AG |
1373 | (unsigned long long)device->ov_position); |
1374 | mod_timer(&device->resync_timer, jiffies); | |
b8907339 PR |
1375 | } |
1376 | } | |
1377 | ||
b30ab791 AG |
1378 | if (get_ldev(device)) { |
1379 | u32 mdf = device->ldev->md.flags & ~(MDF_CONSISTENT|MDF_PRIMARY_IND| | |
b8907339 PR |
1380 | MDF_CONNECTED_IND|MDF_WAS_UP_TO_DATE| |
1381 | MDF_PEER_OUT_DATED|MDF_CRASHED_PRIMARY); | |
1382 | ||
d5d7ebd4 | 1383 | mdf &= ~MDF_AL_CLEAN; |
b30ab791 | 1384 | if (test_bit(CRASHED_PRIMARY, &device->flags)) |
b8907339 | 1385 | mdf |= MDF_CRASHED_PRIMARY; |
b30ab791 AG |
1386 | if (device->state.role == R_PRIMARY || |
1387 | (device->state.pdsk < D_INCONSISTENT && device->state.peer == R_PRIMARY)) | |
b8907339 | 1388 | mdf |= MDF_PRIMARY_IND; |
b30ab791 | 1389 | if (device->state.conn > C_WF_REPORT_PARAMS) |
b8907339 | 1390 | mdf |= MDF_CONNECTED_IND; |
b30ab791 | 1391 | if (device->state.disk > D_INCONSISTENT) |
b8907339 | 1392 | mdf |= MDF_CONSISTENT; |
b30ab791 | 1393 | if (device->state.disk > D_OUTDATED) |
b8907339 | 1394 | mdf |= MDF_WAS_UP_TO_DATE; |
b30ab791 | 1395 | if (device->state.pdsk <= D_OUTDATED && device->state.pdsk >= D_INCONSISTENT) |
b8907339 | 1396 | mdf |= MDF_PEER_OUT_DATED; |
b30ab791 AG |
1397 | if (mdf != device->ldev->md.flags) { |
1398 | device->ldev->md.flags = mdf; | |
1399 | drbd_md_mark_dirty(device); | |
b8907339 PR |
1400 | } |
1401 | if (os.disk < D_CONSISTENT && ns.disk >= D_CONSISTENT) | |
b30ab791 AG |
1402 | drbd_set_ed_uuid(device, device->ldev->md.uuid[UI_CURRENT]); |
1403 | put_ldev(device); | |
b8907339 PR |
1404 | } |
1405 | ||
1406 | /* Peer was forced D_UP_TO_DATE & R_PRIMARY, consider to resync */ | |
1407 | if (os.disk == D_INCONSISTENT && os.pdsk == D_INCONSISTENT && | |
1408 | os.peer == R_SECONDARY && ns.peer == R_PRIMARY) | |
b30ab791 | 1409 | set_bit(CONSIDER_RESYNC, &device->flags); |
b8907339 PR |
1410 | |
1411 | /* Receiver should clean up itself */ | |
1412 | if (os.conn != C_DISCONNECTING && ns.conn == C_DISCONNECTING) | |
44a4d551 | 1413 | drbd_thread_stop_nowait(&connection->receiver); |
b8907339 PR |
1414 | |
1415 | /* Now the receiver finished cleaning up itself, it should die */ | |
1416 | if (os.conn != C_STANDALONE && ns.conn == C_STANDALONE) | |
44a4d551 | 1417 | drbd_thread_stop_nowait(&connection->receiver); |
b8907339 PR |
1418 | |
1419 | /* Upon network failure, we need to restart the receiver. */ | |
823bd832 | 1420 | if (os.conn > C_WF_CONNECTION && |
b8907339 | 1421 | ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT) |
44a4d551 | 1422 | drbd_thread_restart_nowait(&connection->receiver); |
b8907339 PR |
1423 | |
1424 | /* Resume AL writing if we get a connection */ | |
28e448bb | 1425 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { |
b30ab791 | 1426 | drbd_resume_al(device); |
44a4d551 | 1427 | connection->connect_cnt++; |
28e448bb | 1428 | } |
b8907339 | 1429 | |
07be15b1 LE |
1430 | /* remember last attach time so request_timer_fn() won't |
1431 | * kill newly established sessions while we are still trying to thaw | |
1432 | * previously frozen IO */ | |
1433 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && | |
1434 | ns.disk > D_NEGOTIATING) | |
b30ab791 | 1435 | device->last_reattach_jif = jiffies; |
07be15b1 | 1436 | |
b8907339 PR |
1437 | ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC); |
1438 | if (ascw) { | |
1439 | ascw->os = os; | |
1440 | ascw->ns = ns; | |
1441 | ascw->flags = flags; | |
1442 | ascw->w.cb = w_after_state_ch; | |
84b8c06b | 1443 | ascw->device = device; |
b8907339 | 1444 | ascw->done = done; |
a2972846 | 1445 | ascw->state_change = state_change; |
44a4d551 | 1446 | drbd_queue_work(&connection->sender_work, |
84b8c06b | 1447 | &ascw->w); |
b8907339 | 1448 | } else { |
d0180171 | 1449 | drbd_err(device, "Could not kmalloc an ascw\n"); |
b8907339 PR |
1450 | } |
1451 | ||
1452 | return rv; | |
1453 | } | |
1454 | ||
99920dc5 | 1455 | static int w_after_state_ch(struct drbd_work *w, int unused) |
b8907339 PR |
1456 | { |
1457 | struct after_state_chg_work *ascw = | |
1458 | container_of(w, struct after_state_chg_work, w); | |
84b8c06b | 1459 | struct drbd_device *device = ascw->device; |
b8907339 | 1460 | |
a2972846 AG |
1461 | after_state_ch(device, ascw->os, ascw->ns, ascw->flags, ascw->state_change); |
1462 | forget_state_change(ascw->state_change); | |
137975c1 | 1463 | if (ascw->flags & CS_WAIT_COMPLETE) |
b8907339 | 1464 | complete(ascw->done); |
b8907339 PR |
1465 | kfree(ascw); |
1466 | ||
99920dc5 | 1467 | return 0; |
b8907339 PR |
1468 | } |
1469 | ||
b30ab791 | 1470 | static void abw_start_sync(struct drbd_device *device, int rv) |
b8907339 PR |
1471 | { |
1472 | if (rv) { | |
d0180171 | 1473 | drbd_err(device, "Writing the bitmap failed not starting resync.\n"); |
b30ab791 | 1474 | _drbd_request_state(device, NS(conn, C_CONNECTED), CS_VERBOSE); |
b8907339 PR |
1475 | return; |
1476 | } | |
1477 | ||
b30ab791 | 1478 | switch (device->state.conn) { |
b8907339 | 1479 | case C_STARTING_SYNC_T: |
b30ab791 | 1480 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); |
b8907339 PR |
1481 | break; |
1482 | case C_STARTING_SYNC_S: | |
b30ab791 | 1483 | drbd_start_resync(device, C_SYNC_SOURCE); |
b8907339 PR |
1484 | break; |
1485 | } | |
1486 | } | |
1487 | ||
b30ab791 | 1488 | int drbd_bitmap_io_from_worker(struct drbd_device *device, |
54761697 | 1489 | int (*io_fn)(struct drbd_device *), |
b8907339 PR |
1490 | char *why, enum bm_flag flags) |
1491 | { | |
1492 | int rv; | |
1493 | ||
0b0ba1ef | 1494 | D_ASSERT(device, current == first_peer_device(device)->connection->worker.task); |
b8907339 | 1495 | |
b30ab791 | 1496 | /* open coded non-blocking drbd_suspend_io(device); */ |
7dbb4386 | 1497 | atomic_inc(&device->suspend_cnt); |
b8907339 | 1498 | |
b30ab791 AG |
1499 | drbd_bm_lock(device, why, flags); |
1500 | rv = io_fn(device); | |
1501 | drbd_bm_unlock(device); | |
b8907339 | 1502 | |
b30ab791 | 1503 | drbd_resume_io(device); |
b8907339 PR |
1504 | |
1505 | return rv; | |
1506 | } | |
1507 | ||
a2972846 AG |
1508 | void notify_resource_state_change(struct sk_buff *skb, |
1509 | unsigned int seq, | |
1510 | struct drbd_resource_state_change *resource_state_change, | |
1511 | enum drbd_notification_type type) | |
1512 | { | |
1513 | struct drbd_resource *resource = resource_state_change->resource; | |
1514 | struct resource_info resource_info = { | |
1515 | .res_role = resource_state_change->role[NEW], | |
1516 | .res_susp = resource_state_change->susp[NEW], | |
1517 | .res_susp_nod = resource_state_change->susp_nod[NEW], | |
1518 | .res_susp_fen = resource_state_change->susp_fen[NEW], | |
1519 | }; | |
1520 | ||
1521 | notify_resource_state(skb, seq, resource, &resource_info, type); | |
1522 | } | |
1523 | ||
1524 | void notify_connection_state_change(struct sk_buff *skb, | |
1525 | unsigned int seq, | |
1526 | struct drbd_connection_state_change *connection_state_change, | |
1527 | enum drbd_notification_type type) | |
1528 | { | |
1529 | struct drbd_connection *connection = connection_state_change->connection; | |
1530 | struct connection_info connection_info = { | |
1531 | .conn_connection_state = connection_state_change->cstate[NEW], | |
1532 | .conn_role = connection_state_change->peer_role[NEW], | |
1533 | }; | |
1534 | ||
1535 | notify_connection_state(skb, seq, connection, &connection_info, type); | |
1536 | } | |
1537 | ||
1538 | void notify_device_state_change(struct sk_buff *skb, | |
1539 | unsigned int seq, | |
1540 | struct drbd_device_state_change *device_state_change, | |
1541 | enum drbd_notification_type type) | |
1542 | { | |
1543 | struct drbd_device *device = device_state_change->device; | |
1544 | struct device_info device_info = { | |
1545 | .dev_disk_state = device_state_change->disk_state[NEW], | |
1546 | }; | |
1547 | ||
1548 | notify_device_state(skb, seq, device, &device_info, type); | |
1549 | } | |
1550 | ||
1551 | void notify_peer_device_state_change(struct sk_buff *skb, | |
1552 | unsigned int seq, | |
1553 | struct drbd_peer_device_state_change *p, | |
1554 | enum drbd_notification_type type) | |
1555 | { | |
1556 | struct drbd_peer_device *peer_device = p->peer_device; | |
1557 | struct peer_device_info peer_device_info = { | |
1558 | .peer_repl_state = p->repl_state[NEW], | |
1559 | .peer_disk_state = p->disk_state[NEW], | |
1560 | .peer_resync_susp_user = p->resync_susp_user[NEW], | |
1561 | .peer_resync_susp_peer = p->resync_susp_peer[NEW], | |
1562 | .peer_resync_susp_dependency = p->resync_susp_dependency[NEW], | |
1563 | }; | |
1564 | ||
1565 | notify_peer_device_state(skb, seq, peer_device, &peer_device_info, type); | |
1566 | } | |
1567 | ||
1568 | static void broadcast_state_change(struct drbd_state_change *state_change) | |
1569 | { | |
1570 | struct drbd_resource_state_change *resource_state_change = &state_change->resource[0]; | |
1571 | bool resource_state_has_changed; | |
1572 | unsigned int n_device, n_connection, n_peer_device, n_peer_devices; | |
1573 | void (*last_func)(struct sk_buff *, unsigned int, void *, | |
1574 | enum drbd_notification_type) = NULL; | |
1575 | void *uninitialized_var(last_arg); | |
1576 | ||
1577 | #define HAS_CHANGED(state) ((state)[OLD] != (state)[NEW]) | |
1578 | #define FINAL_STATE_CHANGE(type) \ | |
1579 | ({ if (last_func) \ | |
1580 | last_func(NULL, 0, last_arg, type); \ | |
1581 | }) | |
1582 | #define REMEMBER_STATE_CHANGE(func, arg, type) \ | |
1583 | ({ FINAL_STATE_CHANGE(type | NOTIFY_CONTINUES); \ | |
1584 | last_func = (typeof(last_func))func; \ | |
1585 | last_arg = arg; \ | |
1586 | }) | |
1587 | ||
1588 | mutex_lock(¬ification_mutex); | |
1589 | ||
1590 | resource_state_has_changed = | |
1591 | HAS_CHANGED(resource_state_change->role) || | |
1592 | HAS_CHANGED(resource_state_change->susp) || | |
1593 | HAS_CHANGED(resource_state_change->susp_nod) || | |
1594 | HAS_CHANGED(resource_state_change->susp_fen); | |
1595 | ||
1596 | if (resource_state_has_changed) | |
1597 | REMEMBER_STATE_CHANGE(notify_resource_state_change, | |
1598 | resource_state_change, NOTIFY_CHANGE); | |
1599 | ||
1600 | for (n_connection = 0; n_connection < state_change->n_connections; n_connection++) { | |
1601 | struct drbd_connection_state_change *connection_state_change = | |
1602 | &state_change->connections[n_connection]; | |
1603 | ||
1604 | if (HAS_CHANGED(connection_state_change->peer_role) || | |
1605 | HAS_CHANGED(connection_state_change->cstate)) | |
1606 | REMEMBER_STATE_CHANGE(notify_connection_state_change, | |
1607 | connection_state_change, NOTIFY_CHANGE); | |
1608 | } | |
1609 | ||
1610 | for (n_device = 0; n_device < state_change->n_devices; n_device++) { | |
1611 | struct drbd_device_state_change *device_state_change = | |
1612 | &state_change->devices[n_device]; | |
1613 | ||
1614 | if (HAS_CHANGED(device_state_change->disk_state)) | |
1615 | REMEMBER_STATE_CHANGE(notify_device_state_change, | |
1616 | device_state_change, NOTIFY_CHANGE); | |
1617 | } | |
1618 | ||
1619 | n_peer_devices = state_change->n_devices * state_change->n_connections; | |
1620 | for (n_peer_device = 0; n_peer_device < n_peer_devices; n_peer_device++) { | |
1621 | struct drbd_peer_device_state_change *p = | |
1622 | &state_change->peer_devices[n_peer_device]; | |
1623 | ||
1624 | if (HAS_CHANGED(p->disk_state) || | |
1625 | HAS_CHANGED(p->repl_state) || | |
1626 | HAS_CHANGED(p->resync_susp_user) || | |
1627 | HAS_CHANGED(p->resync_susp_peer) || | |
1628 | HAS_CHANGED(p->resync_susp_dependency)) | |
1629 | REMEMBER_STATE_CHANGE(notify_peer_device_state_change, | |
1630 | p, NOTIFY_CHANGE); | |
1631 | } | |
1632 | ||
1633 | FINAL_STATE_CHANGE(NOTIFY_CHANGE); | |
1634 | mutex_unlock(¬ification_mutex); | |
1635 | ||
1636 | #undef HAS_CHANGED | |
1637 | #undef FINAL_STATE_CHANGE | |
1638 | #undef REMEMBER_STATE_CHANGE | |
1639 | } | |
1640 | ||
20004e24 LE |
1641 | /* takes old and new peer disk state */ |
1642 | static bool lost_contact_to_peer_data(enum drbd_disk_state os, enum drbd_disk_state ns) | |
1643 | { | |
1644 | if ((os >= D_INCONSISTENT && os != D_UNKNOWN && os != D_OUTDATED) | |
1645 | && (ns < D_INCONSISTENT || ns == D_UNKNOWN || ns == D_OUTDATED)) | |
1646 | return true; | |
1647 | ||
1648 | /* Scenario, starting with normal operation | |
1649 | * Connected Primary/Secondary UpToDate/UpToDate | |
1650 | * NetworkFailure Primary/Unknown UpToDate/DUnknown (frozen) | |
1651 | * ... | |
1652 | * Connected Primary/Secondary UpToDate/Diskless (resumed; needs to bump uuid!) | |
1653 | */ | |
1654 | if (os == D_UNKNOWN | |
1655 | && (ns == D_DISKLESS || ns == D_FAILED || ns == D_OUTDATED)) | |
1656 | return true; | |
1657 | ||
1658 | return false; | |
1659 | } | |
1660 | ||
b8907339 PR |
1661 | /** |
1662 | * after_state_ch() - Perform after state change actions that may sleep | |
b30ab791 | 1663 | * @device: DRBD device. |
b8907339 PR |
1664 | * @os: old state. |
1665 | * @ns: new state. | |
1666 | * @flags: Flags | |
1667 | */ | |
b30ab791 | 1668 | static void after_state_ch(struct drbd_device *device, union drbd_state os, |
a2972846 AG |
1669 | union drbd_state ns, enum chg_state_flags flags, |
1670 | struct drbd_state_change *state_change) | |
b8907339 | 1671 | { |
6bbf53ca | 1672 | struct drbd_resource *resource = device->resource; |
44a4d551 LE |
1673 | struct drbd_peer_device *peer_device = first_peer_device(device); |
1674 | struct drbd_connection *connection = peer_device ? peer_device->connection : NULL; | |
3b98c0c2 LE |
1675 | struct sib_info sib; |
1676 | ||
a2972846 AG |
1677 | broadcast_state_change(state_change); |
1678 | ||
3b98c0c2 LE |
1679 | sib.sib_reason = SIB_STATE_CHANGE; |
1680 | sib.os = os; | |
1681 | sib.ns = ns; | |
b8907339 | 1682 | |
2f2abeae LE |
1683 | if ((os.disk != D_UP_TO_DATE || os.pdsk != D_UP_TO_DATE) |
1684 | && (ns.disk == D_UP_TO_DATE && ns.pdsk == D_UP_TO_DATE)) { | |
b30ab791 AG |
1685 | clear_bit(CRASHED_PRIMARY, &device->flags); |
1686 | if (device->p_uuid) | |
1687 | device->p_uuid[UI_FLAGS] &= ~((u64)2); | |
b8907339 PR |
1688 | } |
1689 | ||
b8907339 | 1690 | /* Inform userspace about the change... */ |
b30ab791 | 1691 | drbd_bcast_event(device, &sib); |
b8907339 PR |
1692 | |
1693 | if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) && | |
1694 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) | |
b30ab791 | 1695 | drbd_khelper(device, "pri-on-incon-degr"); |
b8907339 PR |
1696 | |
1697 | /* Here we have the actions that are performed after a | |
1698 | state change. This function might sleep */ | |
1699 | ||
b8907339 | 1700 | if (ns.susp_nod) { |
a6d00c8e PR |
1701 | enum drbd_req_event what = NOTHING; |
1702 | ||
0500813f | 1703 | spin_lock_irq(&device->resource->req_lock); |
bde89a9e | 1704 | if (os.conn < C_CONNECTED && conn_lowest_conn(connection) >= C_CONNECTED) |
b8907339 PR |
1705 | what = RESEND; |
1706 | ||
3fb4746d | 1707 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && |
af61494a | 1708 | conn_lowest_disk(connection) == D_UP_TO_DATE) |
b8907339 PR |
1709 | what = RESTART_FROZEN_DISK_IO; |
1710 | ||
6bbf53ca | 1711 | if (resource->susp_nod && what != NOTHING) { |
bde89a9e AG |
1712 | _tl_restart(connection, what); |
1713 | _conn_request_state(connection, | |
892fdd1a PR |
1714 | (union drbd_state) { { .susp_nod = 1 } }, |
1715 | (union drbd_state) { { .susp_nod = 0 } }, | |
1716 | CS_VERBOSE); | |
b8907339 | 1717 | } |
0500813f | 1718 | spin_unlock_irq(&device->resource->req_lock); |
b8907339 PR |
1719 | } |
1720 | ||
88f79ec4 | 1721 | if (ns.susp_fen) { |
0500813f | 1722 | spin_lock_irq(&device->resource->req_lock); |
6bbf53ca | 1723 | if (resource->susp_fen && conn_lowest_conn(connection) >= C_CONNECTED) { |
88f79ec4 | 1724 | /* case2: The connection was established again: */ |
c06ece6b | 1725 | struct drbd_peer_device *peer_device; |
88f79ec4 PR |
1726 | int vnr; |
1727 | ||
1728 | rcu_read_lock(); | |
c06ece6b AG |
1729 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
1730 | clear_bit(NEW_CUR_UUID, &peer_device->device->flags); | |
88f79ec4 | 1731 | rcu_read_unlock(); |
20004e24 LE |
1732 | |
1733 | /* We should actively create a new uuid, _before_ | |
1734 | * we resume/resent, if the peer is diskless | |
1735 | * (recovery from a multiple error scenario). | |
1736 | * Currently, this happens with a slight delay | |
1737 | * below when checking lost_contact_to_peer_data() ... | |
1738 | */ | |
bde89a9e AG |
1739 | _tl_restart(connection, RESEND); |
1740 | _conn_request_state(connection, | |
88f79ec4 PR |
1741 | (union drbd_state) { { .susp_fen = 1 } }, |
1742 | (union drbd_state) { { .susp_fen = 0 } }, | |
1743 | CS_VERBOSE); | |
1744 | } | |
0500813f | 1745 | spin_unlock_irq(&device->resource->req_lock); |
88f79ec4 PR |
1746 | } |
1747 | ||
b8907339 PR |
1748 | /* Became sync source. With protocol >= 96, we still need to send out |
1749 | * the sync uuid now. Need to do that before any drbd_send_state, or | |
1750 | * the other side may go "paused sync" before receiving the sync uuids, | |
1751 | * which is unexpected. */ | |
1752 | if ((os.conn != C_SYNC_SOURCE && os.conn != C_PAUSED_SYNC_S) && | |
1753 | (ns.conn == C_SYNC_SOURCE || ns.conn == C_PAUSED_SYNC_S) && | |
44a4d551 LE |
1754 | connection->agreed_pro_version >= 96 && get_ldev(device)) { |
1755 | drbd_gen_and_send_sync_uuid(peer_device); | |
b30ab791 | 1756 | put_ldev(device); |
b8907339 PR |
1757 | } |
1758 | ||
1759 | /* Do not change the order of the if above and the two below... */ | |
369bea63 PR |
1760 | if (os.pdsk == D_DISKLESS && |
1761 | ns.pdsk > D_DISKLESS && ns.pdsk != D_UNKNOWN) { /* attach on the peer */ | |
a324896b LE |
1762 | /* we probably will start a resync soon. |
1763 | * make sure those things are properly reset. */ | |
b30ab791 AG |
1764 | device->rs_total = 0; |
1765 | device->rs_failed = 0; | |
1766 | atomic_set(&device->rs_pending_cnt, 0); | |
1767 | drbd_rs_cancel_all(device); | |
a324896b | 1768 | |
44a4d551 LE |
1769 | drbd_send_uuids(peer_device); |
1770 | drbd_send_state(peer_device, ns); | |
b8907339 PR |
1771 | } |
1772 | /* No point in queuing send_bitmap if we don't have a connection | |
1773 | * anymore, so check also the _current_ state, not only the new state | |
1774 | * at the time this work was queued. */ | |
1775 | if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S && | |
b30ab791 AG |
1776 | device->state.conn == C_WF_BITMAP_S) |
1777 | drbd_queue_bitmap_io(device, &drbd_send_bitmap, NULL, | |
b8907339 PR |
1778 | "send_bitmap (WFBitMapS)", |
1779 | BM_LOCKED_TEST_ALLOWED); | |
1780 | ||
1781 | /* Lost contact to peer's copy of the data */ | |
20004e24 | 1782 | if (lost_contact_to_peer_data(os.pdsk, ns.pdsk)) { |
b30ab791 | 1783 | if (get_ldev(device)) { |
b8907339 | 1784 | if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && |
b30ab791 AG |
1785 | device->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
1786 | if (drbd_suspended(device)) { | |
1787 | set_bit(NEW_CUR_UUID, &device->flags); | |
b8907339 | 1788 | } else { |
b30ab791 | 1789 | drbd_uuid_new_current(device); |
44a4d551 | 1790 | drbd_send_uuids(peer_device); |
b8907339 PR |
1791 | } |
1792 | } | |
b30ab791 | 1793 | put_ldev(device); |
b8907339 PR |
1794 | } |
1795 | } | |
1796 | ||
b30ab791 | 1797 | if (ns.pdsk < D_INCONSISTENT && get_ldev(device)) { |
9bd2eb2c | 1798 | if (os.peer != R_PRIMARY && ns.peer == R_PRIMARY && |
b30ab791 AG |
1799 | device->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
1800 | drbd_uuid_new_current(device); | |
44a4d551 | 1801 | drbd_send_uuids(peer_device); |
0cfac5dd | 1802 | } |
b8907339 PR |
1803 | /* D_DISKLESS Peer becomes secondary */ |
1804 | if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) | |
1805 | /* We may still be Primary ourselves. | |
1806 | * No harm done if the bitmap still changes, | |
1807 | * redirtied pages will follow later. */ | |
b30ab791 | 1808 | drbd_bitmap_io_from_worker(device, &drbd_bm_write, |
b8907339 | 1809 | "demote diskless peer", BM_LOCKED_SET_ALLOWED); |
b30ab791 | 1810 | put_ldev(device); |
b8907339 PR |
1811 | } |
1812 | ||
1813 | /* Write out all changed bits on demote. | |
1814 | * Though, no need to da that just yet | |
1815 | * if there is a resync going on still */ | |
1816 | if (os.role == R_PRIMARY && ns.role == R_SECONDARY && | |
b30ab791 | 1817 | device->state.conn <= C_CONNECTED && get_ldev(device)) { |
b8907339 PR |
1818 | /* No changes to the bitmap expected this time, so assert that, |
1819 | * even though no harm was done if it did change. */ | |
b30ab791 | 1820 | drbd_bitmap_io_from_worker(device, &drbd_bm_write, |
b8907339 | 1821 | "demote", BM_LOCKED_TEST_ALLOWED); |
b30ab791 | 1822 | put_ldev(device); |
b8907339 PR |
1823 | } |
1824 | ||
1825 | /* Last part of the attaching process ... */ | |
1826 | if (ns.conn >= C_CONNECTED && | |
1827 | os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) { | |
44a4d551 LE |
1828 | drbd_send_sizes(peer_device, 0, 0); /* to start sync... */ |
1829 | drbd_send_uuids(peer_device); | |
1830 | drbd_send_state(peer_device, ns); | |
b8907339 PR |
1831 | } |
1832 | ||
1833 | /* We want to pause/continue resync, tell peer. */ | |
1834 | if (ns.conn >= C_CONNECTED && | |
1835 | ((os.aftr_isp != ns.aftr_isp) || | |
1836 | (os.user_isp != ns.user_isp))) | |
44a4d551 | 1837 | drbd_send_state(peer_device, ns); |
b8907339 PR |
1838 | |
1839 | /* In case one of the isp bits got set, suspend other devices. */ | |
1840 | if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) && | |
1841 | (ns.aftr_isp || ns.peer_isp || ns.user_isp)) | |
b30ab791 | 1842 | suspend_other_sg(device); |
b8907339 PR |
1843 | |
1844 | /* Make sure the peer gets informed about eventual state | |
1845 | changes (ISP bits) while we were in WFReportParams. */ | |
1846 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED) | |
44a4d551 | 1847 | drbd_send_state(peer_device, ns); |
b8907339 PR |
1848 | |
1849 | if (os.conn != C_AHEAD && ns.conn == C_AHEAD) | |
44a4d551 | 1850 | drbd_send_state(peer_device, ns); |
b8907339 PR |
1851 | |
1852 | /* We are in the progress to start a full sync... */ | |
1853 | if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
1854 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S)) | |
1855 | /* no other bitmap changes expected during this phase */ | |
b30ab791 | 1856 | drbd_queue_bitmap_io(device, |
b8907339 PR |
1857 | &drbd_bmio_set_n_write, &abw_start_sync, |
1858 | "set_n_write from StartingSync", BM_LOCKED_TEST_ALLOWED); | |
1859 | ||
b8907339 PR |
1860 | /* first half of local IO error, failure to attach, |
1861 | * or administrative detach */ | |
1862 | if (os.disk != D_FAILED && ns.disk == D_FAILED) { | |
32db80f6 PR |
1863 | enum drbd_io_error_p eh = EP_PASS_ON; |
1864 | int was_io_error = 0; | |
28bc3b8c | 1865 | /* corresponding get_ldev was in _drbd_set_state, to serialize |
32db80f6 PR |
1866 | * our cleanup here with the transition to D_DISKLESS. |
1867 | * But is is still not save to dreference ldev here, since | |
1868 | * we might come from an failed Attach before ldev was set. */ | |
b30ab791 | 1869 | if (device->ldev) { |
32db80f6 | 1870 | rcu_read_lock(); |
b30ab791 | 1871 | eh = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
32db80f6 PR |
1872 | rcu_read_unlock(); |
1873 | ||
b30ab791 | 1874 | was_io_error = test_and_clear_bit(WAS_IO_ERROR, &device->flags); |
32db80f6 | 1875 | |
dc99562a LE |
1876 | /* Intentionally call this handler first, before drbd_send_state(). |
1877 | * See: 2932204 drbd: call local-io-error handler early | |
1878 | * People may chose to hard-reset the box from this handler. | |
1879 | * It is useful if this looks like a "regular node crash". */ | |
6f1a6563 | 1880 | if (was_io_error && eh == EP_CALL_HELPER) |
b30ab791 | 1881 | drbd_khelper(device, "local-io-error"); |
6f1a6563 | 1882 | |
0c849666 LE |
1883 | /* Immediately allow completion of all application IO, |
1884 | * that waits for completion from the local disk, | |
1885 | * if this was a force-detach due to disk_timeout | |
1886 | * or administrator request (drbdsetup detach --force). | |
1887 | * Do NOT abort otherwise. | |
1888 | * Aborting local requests may cause serious problems, | |
1889 | * if requests are completed to upper layers already, | |
1890 | * and then later the already submitted local bio completes. | |
1891 | * This can cause DMA into former bio pages that meanwhile | |
1892 | * have been re-used for other things. | |
1893 | * So aborting local requests may cause crashes, | |
1894 | * or even worse, silent data corruption. | |
1895 | */ | |
b30ab791 AG |
1896 | if (test_and_clear_bit(FORCE_DETACH, &device->flags)) |
1897 | tl_abort_disk_io(device); | |
32db80f6 PR |
1898 | |
1899 | /* current state still has to be D_FAILED, | |
1900 | * there is only one way out: to D_DISKLESS, | |
1901 | * and that may only happen after our put_ldev below. */ | |
b30ab791 | 1902 | if (device->state.disk != D_FAILED) |
d0180171 | 1903 | drbd_err(device, |
32db80f6 | 1904 | "ASSERT FAILED: disk is %s during detach\n", |
b30ab791 | 1905 | drbd_disk_str(device->state.disk)); |
32db80f6 PR |
1906 | |
1907 | if (ns.conn >= C_CONNECTED) | |
44a4d551 | 1908 | drbd_send_state(peer_device, ns); |
32db80f6 | 1909 | |
b30ab791 | 1910 | drbd_rs_cancel_all(device); |
32db80f6 PR |
1911 | |
1912 | /* In case we want to get something to stable storage still, | |
1913 | * this may be the last chance. | |
1914 | * Following put_ldev may transition to D_DISKLESS. */ | |
b30ab791 | 1915 | drbd_md_sync(device); |
32db80f6 | 1916 | } |
b30ab791 | 1917 | put_ldev(device); |
b8907339 PR |
1918 | } |
1919 | ||
b30ab791 AG |
1920 | /* second half of local IO error, failure to attach, |
1921 | * or administrative detach, | |
1922 | * after local_cnt references have reached zero again */ | |
1923 | if (os.disk != D_DISKLESS && ns.disk == D_DISKLESS) { | |
1924 | /* We must still be diskless, | |
1925 | * re-attach has to be serialized with this! */ | |
1926 | if (device->state.disk != D_DISKLESS) | |
d0180171 AG |
1927 | drbd_err(device, |
1928 | "ASSERT FAILED: disk is %s while going diskless\n", | |
1929 | drbd_disk_str(device->state.disk)); | |
b8907339 | 1930 | |
6ab9b1b6 | 1931 | if (ns.conn >= C_CONNECTED) |
44a4d551 | 1932 | drbd_send_state(peer_device, ns); |
b8907339 PR |
1933 | /* corresponding get_ldev in __drbd_set_state |
1934 | * this may finally trigger drbd_ldev_destroy. */ | |
b30ab791 | 1935 | put_ldev(device); |
b8907339 PR |
1936 | } |
1937 | ||
1938 | /* Notify peer that I had a local IO error, and did not detached.. */ | |
6ab9b1b6 | 1939 | if (os.disk == D_UP_TO_DATE && ns.disk == D_INCONSISTENT && ns.conn >= C_CONNECTED) |
44a4d551 | 1940 | drbd_send_state(peer_device, ns); |
b8907339 PR |
1941 | |
1942 | /* Disks got bigger while they were detached */ | |
1943 | if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING && | |
b30ab791 | 1944 | test_and_clear_bit(RESYNC_AFTER_NEG, &device->flags)) { |
b8907339 | 1945 | if (ns.conn == C_CONNECTED) |
b30ab791 | 1946 | resync_after_online_grow(device); |
b8907339 PR |
1947 | } |
1948 | ||
1949 | /* A resync finished or aborted, wake paused devices... */ | |
1950 | if ((os.conn > C_CONNECTED && ns.conn <= C_CONNECTED) || | |
1951 | (os.peer_isp && !ns.peer_isp) || | |
1952 | (os.user_isp && !ns.user_isp)) | |
b30ab791 | 1953 | resume_next_sg(device); |
b8907339 PR |
1954 | |
1955 | /* sync target done with resync. Explicitly notify peer, even though | |
1956 | * it should (at least for non-empty resyncs) already know itself. */ | |
1957 | if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED) | |
44a4d551 | 1958 | drbd_send_state(peer_device, ns); |
b8907339 | 1959 | |
58ffa580 LE |
1960 | /* Verify finished, or reached stop sector. Peer did not know about |
1961 | * the stop sector, and we may even have changed the stop sector during | |
1962 | * verify to interrupt/stop early. Send the new state. */ | |
1963 | if (os.conn == C_VERIFY_S && ns.conn == C_CONNECTED | |
b30ab791 | 1964 | && verify_can_do_stop_sector(device)) |
44a4d551 | 1965 | drbd_send_state(peer_device, ns); |
58ffa580 | 1966 | |
b8907339 PR |
1967 | /* This triggers bitmap writeout of potentially still unwritten pages |
1968 | * if the resync finished cleanly, or aborted because of peer disk | |
be115b69 LE |
1969 | * failure, or on transition from resync back to AHEAD/BEHIND. |
1970 | * | |
1971 | * Connection loss is handled in drbd_disconnected() by the receiver. | |
1972 | * | |
b8907339 PR |
1973 | * For resync aborted because of local disk failure, we cannot do |
1974 | * any bitmap writeout anymore. | |
be115b69 | 1975 | * |
b8907339 PR |
1976 | * No harm done if some bits change during this phase. |
1977 | */ | |
be115b69 LE |
1978 | if ((os.conn > C_CONNECTED && os.conn < C_AHEAD) && |
1979 | (ns.conn == C_CONNECTED || ns.conn >= C_AHEAD) && get_ldev(device)) { | |
b30ab791 | 1980 | drbd_queue_bitmap_io(device, &drbd_bm_write_copy_pages, NULL, |
a220d291 | 1981 | "write from resync_finished", BM_LOCKED_CHANGE_ALLOWED); |
b30ab791 | 1982 | put_ldev(device); |
b8907339 PR |
1983 | } |
1984 | ||
1985 | if (ns.disk == D_DISKLESS && | |
1986 | ns.conn == C_STANDALONE && | |
1987 | ns.role == R_SECONDARY) { | |
1988 | if (os.aftr_isp != ns.aftr_isp) | |
b30ab791 | 1989 | resume_next_sg(device); |
b8907339 PR |
1990 | } |
1991 | ||
b30ab791 | 1992 | drbd_md_sync(device); |
b8907339 PR |
1993 | } |
1994 | ||
bbeb641c | 1995 | struct after_conn_state_chg_work { |
4c007603 | 1996 | struct drbd_work w; |
bbeb641c | 1997 | enum drbd_conns oc; |
8c7e16c3 | 1998 | union drbd_state ns_min; |
b30ab791 | 1999 | union drbd_state ns_max; /* new, max state, over all devices */ |
bbeb641c | 2000 | enum chg_state_flags flags; |
84b8c06b | 2001 | struct drbd_connection *connection; |
a2972846 | 2002 | struct drbd_state_change *state_change; |
bbeb641c PR |
2003 | }; |
2004 | ||
99920dc5 | 2005 | static int w_after_conn_state_ch(struct drbd_work *w, int unused) |
bbeb641c PR |
2006 | { |
2007 | struct after_conn_state_chg_work *acscw = | |
4c007603 | 2008 | container_of(w, struct after_conn_state_chg_work, w); |
84b8c06b | 2009 | struct drbd_connection *connection = acscw->connection; |
bbeb641c | 2010 | enum drbd_conns oc = acscw->oc; |
5f082f98 | 2011 | union drbd_state ns_max = acscw->ns_max; |
c06ece6b | 2012 | struct drbd_peer_device *peer_device; |
a6d00c8e | 2013 | int vnr; |
bbeb641c | 2014 | |
a2972846 AG |
2015 | broadcast_state_change(acscw->state_change); |
2016 | forget_state_change(acscw->state_change); | |
bbeb641c PR |
2017 | kfree(acscw); |
2018 | ||
b8907339 | 2019 | /* Upon network configuration, we need to start the receiver */ |
5f082f98 | 2020 | if (oc == C_STANDALONE && ns_max.conn == C_UNCONNECTED) |
bde89a9e | 2021 | drbd_thread_start(&connection->receiver); |
b8907339 | 2022 | |
f3dfa40a LE |
2023 | if (oc == C_DISCONNECTING && ns_max.conn == C_STANDALONE) { |
2024 | struct net_conf *old_conf; | |
2025 | ||
a2972846 AG |
2026 | mutex_lock(¬ification_mutex); |
2027 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2028 | notify_peer_device_state(NULL, 0, peer_device, NULL, | |
2029 | NOTIFY_DESTROY | NOTIFY_CONTINUES); | |
2030 | notify_connection_state(NULL, 0, connection, NULL, NOTIFY_DESTROY); | |
2031 | mutex_unlock(¬ification_mutex); | |
2032 | ||
0500813f | 2033 | mutex_lock(&connection->resource->conf_update); |
bde89a9e AG |
2034 | old_conf = connection->net_conf; |
2035 | connection->my_addr_len = 0; | |
2036 | connection->peer_addr_len = 0; | |
ccdd6a93 | 2037 | RCU_INIT_POINTER(connection->net_conf, NULL); |
bde89a9e | 2038 | conn_free_crypto(connection); |
0500813f | 2039 | mutex_unlock(&connection->resource->conf_update); |
f3dfa40a LE |
2040 | |
2041 | synchronize_rcu(); | |
2042 | kfree(old_conf); | |
2043 | } | |
2044 | ||
a6d00c8e PR |
2045 | if (ns_max.susp_fen) { |
2046 | /* case1: The outdate peer handler is successful: */ | |
2047 | if (ns_max.pdsk <= D_OUTDATED) { | |
695d08fa | 2048 | rcu_read_lock(); |
c06ece6b AG |
2049 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2050 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2051 | if (test_bit(NEW_CUR_UUID, &device->flags)) { |
2052 | drbd_uuid_new_current(device); | |
2053 | clear_bit(NEW_CUR_UUID, &device->flags); | |
a6d00c8e PR |
2054 | } |
2055 | } | |
695d08fa | 2056 | rcu_read_unlock(); |
0500813f | 2057 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
2058 | _tl_restart(connection, CONNECTION_LOST_WHILE_PENDING); |
2059 | _conn_request_state(connection, | |
8a0bab2a PR |
2060 | (union drbd_state) { { .susp_fen = 1 } }, |
2061 | (union drbd_state) { { .susp_fen = 0 } }, | |
2062 | CS_VERBOSE); | |
0500813f | 2063 | spin_unlock_irq(&connection->resource->req_lock); |
a6d00c8e | 2064 | } |
a6d00c8e | 2065 | } |
05a10ec7 | 2066 | kref_put(&connection->kref, drbd_destroy_connection); |
19fffd7b | 2067 | |
bde89a9e | 2068 | conn_md_sync(connection); |
19fffd7b | 2069 | |
99920dc5 | 2070 | return 0; |
bbeb641c PR |
2071 | } |
2072 | ||
8ce953aa | 2073 | static void conn_old_common_state(struct drbd_connection *connection, union drbd_state *pcs, enum chg_state_flags *pf) |
88ef594e | 2074 | { |
435693e8 | 2075 | enum chg_state_flags flags = ~0; |
c06ece6b | 2076 | struct drbd_peer_device *peer_device; |
435693e8 | 2077 | int vnr, first_vol = 1; |
e0e16653 PR |
2078 | union drbd_dev_state os, cs = { |
2079 | { .role = R_SECONDARY, | |
2080 | .peer = R_UNKNOWN, | |
bde89a9e | 2081 | .conn = connection->cstate, |
e0e16653 PR |
2082 | .disk = D_DISKLESS, |
2083 | .pdsk = D_UNKNOWN, | |
2084 | } }; | |
88ef594e | 2085 | |
695d08fa | 2086 | rcu_read_lock(); |
c06ece6b AG |
2087 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2088 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 2089 | os = device->state; |
88ef594e | 2090 | |
435693e8 PR |
2091 | if (first_vol) { |
2092 | cs = os; | |
2093 | first_vol = 0; | |
2094 | continue; | |
2095 | } | |
2096 | ||
2097 | if (cs.role != os.role) | |
2098 | flags &= ~CS_DC_ROLE; | |
2099 | ||
2100 | if (cs.peer != os.peer) | |
2101 | flags &= ~CS_DC_PEER; | |
2102 | ||
2103 | if (cs.conn != os.conn) | |
2104 | flags &= ~CS_DC_CONN; | |
88ef594e | 2105 | |
435693e8 PR |
2106 | if (cs.disk != os.disk) |
2107 | flags &= ~CS_DC_DISK; | |
88ef594e | 2108 | |
435693e8 PR |
2109 | if (cs.pdsk != os.pdsk) |
2110 | flags &= ~CS_DC_PDSK; | |
2111 | } | |
695d08fa | 2112 | rcu_read_unlock(); |
435693e8 PR |
2113 | |
2114 | *pf |= CS_DC_MASK; | |
2115 | *pf &= flags; | |
da9fbc27 | 2116 | (*pcs).i = cs.i; |
88ef594e | 2117 | } |
bbeb641c | 2118 | |
bd0c824a | 2119 | static enum drbd_state_rv |
bde89a9e | 2120 | conn_is_valid_transition(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
88ef594e | 2121 | enum chg_state_flags flags) |
bbeb641c | 2122 | { |
bd0c824a | 2123 | enum drbd_state_rv rv = SS_SUCCESS; |
bbeb641c | 2124 | union drbd_state ns, os; |
c06ece6b | 2125 | struct drbd_peer_device *peer_device; |
bd0c824a | 2126 | int vnr; |
bbeb641c | 2127 | |
695d08fa | 2128 | rcu_read_lock(); |
c06ece6b AG |
2129 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2130 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 2131 | os = drbd_read_state(device); |
d7fe69c6 | 2132 | ns = sanitize_state(device, os, apply_mask_val(os, mask, val), NULL); |
bd0c824a | 2133 | |
778bcf2e PR |
2134 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
2135 | ns.disk = os.disk; | |
2136 | ||
bd0c824a PR |
2137 | if (ns.i == os.i) |
2138 | continue; | |
bbeb641c | 2139 | |
bd0c824a | 2140 | rv = is_valid_transition(os, ns); |
bd0c824a | 2141 | |
c06ece6b | 2142 | if (rv >= SS_SUCCESS && !(flags & CS_HARD)) { |
b30ab791 | 2143 | rv = is_valid_state(device, ns); |
bd0c824a | 2144 | if (rv < SS_SUCCESS) { |
b30ab791 | 2145 | if (is_valid_state(device, os) == rv) |
bde89a9e | 2146 | rv = is_valid_soft_transition(os, ns, connection); |
bd0c824a | 2147 | } else |
bde89a9e | 2148 | rv = is_valid_soft_transition(os, ns, connection); |
bd0c824a | 2149 | } |
c06ece6b AG |
2150 | |
2151 | if (rv < SS_SUCCESS) { | |
2152 | if (flags & CS_VERBOSE) | |
2153 | print_st_err(device, os, ns, rv); | |
bd0c824a | 2154 | break; |
c06ece6b | 2155 | } |
bbeb641c | 2156 | } |
695d08fa | 2157 | rcu_read_unlock(); |
bbeb641c | 2158 | |
bd0c824a | 2159 | return rv; |
bbeb641c PR |
2160 | } |
2161 | ||
8ce953aa | 2162 | static void |
bde89a9e | 2163 | conn_set_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
8c7e16c3 | 2164 | union drbd_state *pns_min, union drbd_state *pns_max, enum chg_state_flags flags) |
bbeb641c | 2165 | { |
f132f554 | 2166 | union drbd_state ns, os, ns_max = { }; |
8c7e16c3 PR |
2167 | union drbd_state ns_min = { |
2168 | { .role = R_MASK, | |
2169 | .peer = R_MASK, | |
e0e16653 | 2170 | .conn = val.conn, |
8c7e16c3 PR |
2171 | .disk = D_MASK, |
2172 | .pdsk = D_MASK | |
2173 | } }; | |
c06ece6b | 2174 | struct drbd_peer_device *peer_device; |
bbeb641c | 2175 | enum drbd_state_rv rv; |
f132f554 | 2176 | int vnr, number_of_volumes = 0; |
bbeb641c | 2177 | |
07be15b1 LE |
2178 | if (mask.conn == C_MASK) { |
2179 | /* remember last connect time so request_timer_fn() won't | |
2180 | * kill newly established sessions while we are still trying to thaw | |
2181 | * previously frozen IO */ | |
bde89a9e AG |
2182 | if (connection->cstate != C_WF_REPORT_PARAMS && val.conn == C_WF_REPORT_PARAMS) |
2183 | connection->last_reconnect_jif = jiffies; | |
07be15b1 | 2184 | |
bde89a9e | 2185 | connection->cstate = val.conn; |
07be15b1 | 2186 | } |
bd0c824a | 2187 | |
695d08fa | 2188 | rcu_read_lock(); |
c06ece6b AG |
2189 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2190 | struct drbd_device *device = peer_device->device; | |
f132f554 | 2191 | number_of_volumes++; |
b30ab791 | 2192 | os = drbd_read_state(device); |
bd0c824a | 2193 | ns = apply_mask_val(os, mask, val); |
d7fe69c6 | 2194 | ns = sanitize_state(device, os, ns, NULL); |
bbeb641c | 2195 | |
778bcf2e PR |
2196 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
2197 | ns.disk = os.disk; | |
2198 | ||
28bc3b8c | 2199 | rv = _drbd_set_state(device, ns, flags, NULL); |
7e5fec31 | 2200 | BUG_ON(rv < SS_SUCCESS); |
b30ab791 | 2201 | ns.i = device->state.i; |
8c7e16c3 PR |
2202 | ns_max.role = max_role(ns.role, ns_max.role); |
2203 | ns_max.peer = max_role(ns.peer, ns_max.peer); | |
2204 | ns_max.conn = max_t(enum drbd_conns, ns.conn, ns_max.conn); | |
2205 | ns_max.disk = max_t(enum drbd_disk_state, ns.disk, ns_max.disk); | |
2206 | ns_max.pdsk = max_t(enum drbd_disk_state, ns.pdsk, ns_max.pdsk); | |
2207 | ||
2208 | ns_min.role = min_role(ns.role, ns_min.role); | |
2209 | ns_min.peer = min_role(ns.peer, ns_min.peer); | |
2210 | ns_min.conn = min_t(enum drbd_conns, ns.conn, ns_min.conn); | |
2211 | ns_min.disk = min_t(enum drbd_disk_state, ns.disk, ns_min.disk); | |
2212 | ns_min.pdsk = min_t(enum drbd_disk_state, ns.pdsk, ns_min.pdsk); | |
bd0c824a | 2213 | } |
695d08fa | 2214 | rcu_read_unlock(); |
bbeb641c | 2215 | |
f132f554 PR |
2216 | if (number_of_volumes == 0) { |
2217 | ns_min = ns_max = (union drbd_state) { { | |
2218 | .role = R_SECONDARY, | |
2219 | .peer = R_UNKNOWN, | |
2220 | .conn = val.conn, | |
2221 | .disk = D_DISKLESS, | |
2222 | .pdsk = D_UNKNOWN | |
2223 | } }; | |
2224 | } | |
2225 | ||
6bbf53ca AG |
2226 | ns_min.susp = ns_max.susp = connection->resource->susp; |
2227 | ns_min.susp_nod = ns_max.susp_nod = connection->resource->susp_nod; | |
2228 | ns_min.susp_fen = ns_max.susp_fen = connection->resource->susp_fen; | |
8c7e16c3 PR |
2229 | |
2230 | *pns_min = ns_min; | |
2231 | *pns_max = ns_max; | |
bbeb641c PR |
2232 | } |
2233 | ||
df24aa45 | 2234 | static enum drbd_state_rv |
bde89a9e | 2235 | _conn_rq_cond(struct drbd_connection *connection, union drbd_state mask, union drbd_state val) |
df24aa45 | 2236 | { |
88ea685d | 2237 | enum drbd_state_rv err, rv = SS_UNKNOWN_ERROR; /* continue waiting */; |
df24aa45 | 2238 | |
bde89a9e | 2239 | if (test_and_clear_bit(CONN_WD_ST_CHG_OKAY, &connection->flags)) |
88ea685d | 2240 | rv = SS_CW_SUCCESS; |
df24aa45 | 2241 | |
bde89a9e | 2242 | if (test_and_clear_bit(CONN_WD_ST_CHG_FAIL, &connection->flags)) |
88ea685d | 2243 | rv = SS_CW_FAILED_BY_PEER; |
df24aa45 | 2244 | |
88ea685d PR |
2245 | err = conn_is_valid_transition(connection, mask, val, 0); |
2246 | if (err == SS_SUCCESS && connection->cstate == C_WF_REPORT_PARAMS) | |
2247 | return rv; | |
df24aa45 | 2248 | |
88ea685d | 2249 | return err; |
df24aa45 PR |
2250 | } |
2251 | ||
bbeb641c | 2252 | enum drbd_state_rv |
bde89a9e | 2253 | _conn_request_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
bbeb641c PR |
2254 | enum chg_state_flags flags) |
2255 | { | |
2256 | enum drbd_state_rv rv = SS_SUCCESS; | |
bbeb641c | 2257 | struct after_conn_state_chg_work *acscw; |
bde89a9e | 2258 | enum drbd_conns oc = connection->cstate; |
8c7e16c3 | 2259 | union drbd_state ns_max, ns_min, os; |
c02abda2 | 2260 | bool have_mutex = false; |
a2972846 | 2261 | struct drbd_state_change *state_change; |
bbeb641c | 2262 | |
07fc9619 PR |
2263 | if (mask.conn) { |
2264 | rv = is_valid_conn_transition(oc, val.conn); | |
2265 | if (rv < SS_SUCCESS) | |
2266 | goto abort; | |
2267 | } | |
bbeb641c | 2268 | |
bde89a9e | 2269 | rv = conn_is_valid_transition(connection, mask, val, flags); |
bbeb641c PR |
2270 | if (rv < SS_SUCCESS) |
2271 | goto abort; | |
2272 | ||
df24aa45 PR |
2273 | if (oc == C_WF_REPORT_PARAMS && val.conn == C_DISCONNECTING && |
2274 | !(flags & (CS_LOCAL_ONLY | CS_HARD))) { | |
c02abda2 LE |
2275 | |
2276 | /* This will be a cluster-wide state change. | |
2277 | * Need to give up the spinlock, grab the mutex, | |
2278 | * then send the state change request, ... */ | |
0500813f | 2279 | spin_unlock_irq(&connection->resource->req_lock); |
bde89a9e | 2280 | mutex_lock(&connection->cstate_mutex); |
c02abda2 LE |
2281 | have_mutex = true; |
2282 | ||
bde89a9e AG |
2283 | set_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
2284 | if (conn_send_state_req(connection, mask, val)) { | |
c02abda2 | 2285 | /* sending failed. */ |
bde89a9e | 2286 | clear_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
c02abda2 LE |
2287 | rv = SS_CW_FAILED_BY_PEER; |
2288 | /* need to re-aquire the spin lock, though */ | |
2289 | goto abort_unlocked; | |
2290 | } | |
2291 | ||
2292 | if (val.conn == C_DISCONNECTING) | |
bde89a9e | 2293 | set_bit(DISCONNECT_SENT, &connection->flags); |
c02abda2 LE |
2294 | |
2295 | /* ... and re-aquire the spinlock. | |
2296 | * If _conn_rq_cond() returned >= SS_SUCCESS, we must call | |
2297 | * conn_set_state() within the same spinlock. */ | |
0500813f | 2298 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
2299 | wait_event_lock_irq(connection->ping_wait, |
2300 | (rv = _conn_rq_cond(connection, mask, val)), | |
0500813f | 2301 | connection->resource->req_lock); |
bde89a9e | 2302 | clear_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
df24aa45 PR |
2303 | if (rv < SS_SUCCESS) |
2304 | goto abort; | |
2305 | } | |
2306 | ||
a2972846 | 2307 | state_change = remember_old_state(connection->resource, GFP_ATOMIC); |
bde89a9e | 2308 | conn_old_common_state(connection, &os, &flags); |
706cb24c | 2309 | flags |= CS_DC_SUSP; |
bde89a9e AG |
2310 | conn_set_state(connection, mask, val, &ns_min, &ns_max, flags); |
2311 | conn_pr_state_change(connection, os, ns_max, flags); | |
a2972846 | 2312 | remember_new_state(state_change); |
bbeb641c PR |
2313 | |
2314 | acscw = kmalloc(sizeof(*acscw), GFP_ATOMIC); | |
2315 | if (acscw) { | |
435693e8 | 2316 | acscw->oc = os.conn; |
8c7e16c3 | 2317 | acscw->ns_min = ns_min; |
5f082f98 | 2318 | acscw->ns_max = ns_max; |
bbeb641c | 2319 | acscw->flags = flags; |
4c007603 | 2320 | acscw->w.cb = w_after_conn_state_ch; |
bde89a9e | 2321 | kref_get(&connection->kref); |
84b8c06b | 2322 | acscw->connection = connection; |
a2972846 | 2323 | acscw->state_change = state_change; |
4c007603 | 2324 | drbd_queue_work(&connection->sender_work, &acscw->w); |
bbeb641c | 2325 | } else { |
1ec861eb | 2326 | drbd_err(connection, "Could not kmalloc an acscw\n"); |
b8907339 | 2327 | } |
bbeb641c | 2328 | |
a01842eb | 2329 | abort: |
c02abda2 LE |
2330 | if (have_mutex) { |
2331 | /* mutex_unlock() "... must not be used in interrupt context.", | |
2332 | * so give up the spinlock, then re-aquire it */ | |
0500813f | 2333 | spin_unlock_irq(&connection->resource->req_lock); |
c02abda2 | 2334 | abort_unlocked: |
bde89a9e | 2335 | mutex_unlock(&connection->cstate_mutex); |
0500813f | 2336 | spin_lock_irq(&connection->resource->req_lock); |
c02abda2 LE |
2337 | } |
2338 | if (rv < SS_SUCCESS && flags & CS_VERBOSE) { | |
1ec861eb AG |
2339 | drbd_err(connection, "State change failed: %s\n", drbd_set_st_err_str(rv)); |
2340 | drbd_err(connection, " mask = 0x%x val = 0x%x\n", mask.i, val.i); | |
2341 | drbd_err(connection, " old_conn:%s wanted_conn:%s\n", drbd_conn_str(oc), drbd_conn_str(val.conn)); | |
a01842eb | 2342 | } |
bbeb641c PR |
2343 | return rv; |
2344 | } | |
2345 | ||
2346 | enum drbd_state_rv | |
bde89a9e | 2347 | conn_request_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
bbeb641c PR |
2348 | enum chg_state_flags flags) |
2349 | { | |
2350 | enum drbd_state_rv rv; | |
2351 | ||
0500813f | 2352 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 2353 | rv = _conn_request_state(connection, mask, val, flags); |
0500813f | 2354 | spin_unlock_irq(&connection->resource->req_lock); |
bbeb641c PR |
2355 | |
2356 | return rv; | |
b8907339 | 2357 | } |