Commit | Line | Data |
---|---|---|
b8907339 PR |
1 | /* |
2 | drbd_state.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | Thanks to Carter Burden, Bart Grantham and Gennadiy Nerubayev | |
11 | from Logicworks, Inc. for making SDP replication support possible. | |
12 | ||
13 | drbd is free software; you can redistribute it and/or modify | |
14 | it under the terms of the GNU General Public License as published by | |
15 | the Free Software Foundation; either version 2, or (at your option) | |
16 | any later version. | |
17 | ||
18 | drbd is distributed in the hope that it will be useful, | |
19 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
20 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
21 | GNU General Public License for more details. | |
22 | ||
23 | You should have received a copy of the GNU General Public License | |
24 | along with drbd; see the file COPYING. If not, write to | |
25 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
26 | */ | |
27 | ||
28 | #include <linux/drbd_limits.h> | |
29 | #include "drbd_int.h" | |
a3603a6e | 30 | #include "drbd_protocol.h" |
b8907339 PR |
31 | #include "drbd_req.h" |
32 | ||
33 | struct after_state_chg_work { | |
34 | struct drbd_work w; | |
84b8c06b | 35 | struct drbd_device *device; |
b8907339 PR |
36 | union drbd_state os; |
37 | union drbd_state ns; | |
38 | enum chg_state_flags flags; | |
39 | struct completion *done; | |
40 | }; | |
41 | ||
d942ae44 PR |
42 | enum sanitize_state_warnings { |
43 | NO_WARNING, | |
44 | ABORTED_ONLINE_VERIFY, | |
45 | ABORTED_RESYNC, | |
46 | CONNECTION_LOST_NEGOTIATING, | |
47 | IMPLICITLY_UPGRADED_DISK, | |
48 | IMPLICITLY_UPGRADED_PDSK, | |
49 | }; | |
50 | ||
99920dc5 | 51 | static int w_after_state_ch(struct drbd_work *w, int unused); |
b30ab791 | 52 | static void after_state_ch(struct drbd_device *device, union drbd_state os, |
b8907339 | 53 | union drbd_state ns, enum chg_state_flags flags); |
54761697 | 54 | static enum drbd_state_rv is_valid_state(struct drbd_device *, union drbd_state); |
bde89a9e | 55 | static enum drbd_state_rv is_valid_soft_transition(union drbd_state, union drbd_state, struct drbd_connection *); |
3509502d | 56 | static enum drbd_state_rv is_valid_transition(union drbd_state os, union drbd_state ns); |
d7fe69c6 PR |
57 | static union drbd_state sanitize_state(struct drbd_device *device, union drbd_state os, |
58 | union drbd_state ns, enum sanitize_state_warnings *warn); | |
b8907339 | 59 | |
2aebfabb PR |
60 | static inline bool is_susp(union drbd_state s) |
61 | { | |
62 | return s.susp || s.susp_nod || s.susp_fen; | |
63 | } | |
64 | ||
bde89a9e | 65 | bool conn_all_vols_unconf(struct drbd_connection *connection) |
0e29d163 | 66 | { |
c06ece6b | 67 | struct drbd_peer_device *peer_device; |
695d08fa | 68 | bool rv = true; |
e90285e0 | 69 | int vnr; |
0e29d163 | 70 | |
695d08fa | 71 | rcu_read_lock(); |
c06ece6b AG |
72 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
73 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
74 | if (device->state.disk != D_DISKLESS || |
75 | device->state.conn != C_STANDALONE || | |
76 | device->state.role != R_SECONDARY) { | |
695d08fa PR |
77 | rv = false; |
78 | break; | |
79 | } | |
0e29d163 | 80 | } |
695d08fa PR |
81 | rcu_read_unlock(); |
82 | ||
83 | return rv; | |
0e29d163 PR |
84 | } |
85 | ||
cb703454 PR |
86 | /* Unfortunately the states where not correctly ordered, when |
87 | they where defined. therefore can not use max_t() here. */ | |
88 | static enum drbd_role max_role(enum drbd_role role1, enum drbd_role role2) | |
89 | { | |
90 | if (role1 == R_PRIMARY || role2 == R_PRIMARY) | |
91 | return R_PRIMARY; | |
92 | if (role1 == R_SECONDARY || role2 == R_SECONDARY) | |
93 | return R_SECONDARY; | |
94 | return R_UNKNOWN; | |
95 | } | |
96 | static enum drbd_role min_role(enum drbd_role role1, enum drbd_role role2) | |
97 | { | |
98 | if (role1 == R_UNKNOWN || role2 == R_UNKNOWN) | |
99 | return R_UNKNOWN; | |
100 | if (role1 == R_SECONDARY || role2 == R_SECONDARY) | |
101 | return R_SECONDARY; | |
102 | return R_PRIMARY; | |
103 | } | |
104 | ||
bde89a9e | 105 | enum drbd_role conn_highest_role(struct drbd_connection *connection) |
cb703454 PR |
106 | { |
107 | enum drbd_role role = R_UNKNOWN; | |
c06ece6b | 108 | struct drbd_peer_device *peer_device; |
cb703454 PR |
109 | int vnr; |
110 | ||
695d08fa | 111 | rcu_read_lock(); |
c06ece6b AG |
112 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
113 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 114 | role = max_role(role, device->state.role); |
c06ece6b | 115 | } |
695d08fa | 116 | rcu_read_unlock(); |
cb703454 PR |
117 | |
118 | return role; | |
119 | } | |
120 | ||
bde89a9e | 121 | enum drbd_role conn_highest_peer(struct drbd_connection *connection) |
cb703454 PR |
122 | { |
123 | enum drbd_role peer = R_UNKNOWN; | |
c06ece6b | 124 | struct drbd_peer_device *peer_device; |
cb703454 PR |
125 | int vnr; |
126 | ||
695d08fa | 127 | rcu_read_lock(); |
c06ece6b AG |
128 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
129 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 130 | peer = max_role(peer, device->state.peer); |
c06ece6b | 131 | } |
695d08fa | 132 | rcu_read_unlock(); |
cb703454 PR |
133 | |
134 | return peer; | |
135 | } | |
136 | ||
bde89a9e | 137 | enum drbd_disk_state conn_highest_disk(struct drbd_connection *connection) |
cb703454 PR |
138 | { |
139 | enum drbd_disk_state ds = D_DISKLESS; | |
c06ece6b | 140 | struct drbd_peer_device *peer_device; |
cb703454 PR |
141 | int vnr; |
142 | ||
695d08fa | 143 | rcu_read_lock(); |
c06ece6b AG |
144 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
145 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 146 | ds = max_t(enum drbd_disk_state, ds, device->state.disk); |
c06ece6b | 147 | } |
695d08fa | 148 | rcu_read_unlock(); |
cb703454 PR |
149 | |
150 | return ds; | |
151 | } | |
152 | ||
bde89a9e | 153 | enum drbd_disk_state conn_lowest_disk(struct drbd_connection *connection) |
4669265a PR |
154 | { |
155 | enum drbd_disk_state ds = D_MASK; | |
c06ece6b | 156 | struct drbd_peer_device *peer_device; |
4669265a PR |
157 | int vnr; |
158 | ||
695d08fa | 159 | rcu_read_lock(); |
c06ece6b AG |
160 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
161 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 162 | ds = min_t(enum drbd_disk_state, ds, device->state.disk); |
c06ece6b | 163 | } |
695d08fa | 164 | rcu_read_unlock(); |
4669265a PR |
165 | |
166 | return ds; | |
167 | } | |
168 | ||
bde89a9e | 169 | enum drbd_disk_state conn_highest_pdsk(struct drbd_connection *connection) |
cb703454 PR |
170 | { |
171 | enum drbd_disk_state ds = D_DISKLESS; | |
c06ece6b | 172 | struct drbd_peer_device *peer_device; |
cb703454 PR |
173 | int vnr; |
174 | ||
695d08fa | 175 | rcu_read_lock(); |
c06ece6b AG |
176 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
177 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 178 | ds = max_t(enum drbd_disk_state, ds, device->state.pdsk); |
c06ece6b | 179 | } |
695d08fa | 180 | rcu_read_unlock(); |
cb703454 PR |
181 | |
182 | return ds; | |
183 | } | |
184 | ||
bde89a9e | 185 | enum drbd_conns conn_lowest_conn(struct drbd_connection *connection) |
19f83c76 PR |
186 | { |
187 | enum drbd_conns conn = C_MASK; | |
c06ece6b | 188 | struct drbd_peer_device *peer_device; |
19f83c76 PR |
189 | int vnr; |
190 | ||
695d08fa | 191 | rcu_read_lock(); |
c06ece6b AG |
192 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
193 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 194 | conn = min_t(enum drbd_conns, conn, device->state.conn); |
c06ece6b | 195 | } |
695d08fa | 196 | rcu_read_unlock(); |
19f83c76 PR |
197 | |
198 | return conn; | |
199 | } | |
200 | ||
bde89a9e | 201 | static bool no_peer_wf_report_params(struct drbd_connection *connection) |
79702011 | 202 | { |
c06ece6b | 203 | struct drbd_peer_device *peer_device; |
79702011 PR |
204 | int vnr; |
205 | bool rv = true; | |
206 | ||
207 | rcu_read_lock(); | |
c06ece6b AG |
208 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
209 | if (peer_device->device->state.conn == C_WF_REPORT_PARAMS) { | |
79702011 PR |
210 | rv = false; |
211 | break; | |
212 | } | |
213 | rcu_read_unlock(); | |
214 | ||
215 | return rv; | |
216 | } | |
217 | ||
218 | ||
b8907339 PR |
219 | /** |
220 | * cl_wide_st_chg() - true if the state change is a cluster wide one | |
b30ab791 | 221 | * @device: DRBD device. |
b8907339 PR |
222 | * @os: old (current) state. |
223 | * @ns: new (wanted) state. | |
224 | */ | |
b30ab791 | 225 | static int cl_wide_st_chg(struct drbd_device *device, |
b8907339 PR |
226 | union drbd_state os, union drbd_state ns) |
227 | { | |
228 | return (os.conn >= C_CONNECTED && ns.conn >= C_CONNECTED && | |
229 | ((os.role != R_PRIMARY && ns.role == R_PRIMARY) || | |
230 | (os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
231 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) || | |
e8744f5a | 232 | (os.disk != D_FAILED && ns.disk == D_FAILED))) || |
b8907339 | 233 | (os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) || |
369bea63 PR |
234 | (os.conn == C_CONNECTED && ns.conn == C_VERIFY_S) || |
235 | (os.conn == C_CONNECTED && ns.conn == C_WF_REPORT_PARAMS); | |
b8907339 PR |
236 | } |
237 | ||
56707f9e PR |
238 | static union drbd_state |
239 | apply_mask_val(union drbd_state os, union drbd_state mask, union drbd_state val) | |
240 | { | |
241 | union drbd_state ns; | |
242 | ns.i = (os.i & ~mask.i) | val.i; | |
243 | return ns; | |
244 | } | |
245 | ||
b8907339 | 246 | enum drbd_state_rv |
b30ab791 | 247 | drbd_change_state(struct drbd_device *device, enum chg_state_flags f, |
b8907339 PR |
248 | union drbd_state mask, union drbd_state val) |
249 | { | |
250 | unsigned long flags; | |
56707f9e | 251 | union drbd_state ns; |
b8907339 PR |
252 | enum drbd_state_rv rv; |
253 | ||
0500813f | 254 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
255 | ns = apply_mask_val(drbd_read_state(device), mask, val); |
256 | rv = _drbd_set_state(device, ns, f, NULL); | |
0500813f | 257 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
258 | |
259 | return rv; | |
260 | } | |
261 | ||
262 | /** | |
263 | * drbd_force_state() - Impose a change which happens outside our control on our state | |
b30ab791 | 264 | * @device: DRBD device. |
b8907339 PR |
265 | * @mask: mask of state bits to change. |
266 | * @val: value of new state bits. | |
267 | */ | |
b30ab791 | 268 | void drbd_force_state(struct drbd_device *device, |
b8907339 PR |
269 | union drbd_state mask, union drbd_state val) |
270 | { | |
b30ab791 | 271 | drbd_change_state(device, CS_HARD, mask, val); |
b8907339 PR |
272 | } |
273 | ||
274 | static enum drbd_state_rv | |
b30ab791 | 275 | _req_st_cond(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
276 | union drbd_state val) |
277 | { | |
278 | union drbd_state os, ns; | |
279 | unsigned long flags; | |
280 | enum drbd_state_rv rv; | |
281 | ||
b30ab791 | 282 | if (test_and_clear_bit(CL_ST_CHG_SUCCESS, &device->flags)) |
b8907339 PR |
283 | return SS_CW_SUCCESS; |
284 | ||
b30ab791 | 285 | if (test_and_clear_bit(CL_ST_CHG_FAIL, &device->flags)) |
b8907339 PR |
286 | return SS_CW_FAILED_BY_PEER; |
287 | ||
0500813f | 288 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 289 | os = drbd_read_state(device); |
d7fe69c6 | 290 | ns = sanitize_state(device, os, apply_mask_val(os, mask, val), NULL); |
3509502d | 291 | rv = is_valid_transition(os, ns); |
a3025a27 | 292 | if (rv >= SS_SUCCESS) |
3509502d | 293 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ |
b8907339 | 294 | |
b30ab791 | 295 | if (!cl_wide_st_chg(device, os, ns)) |
b8907339 | 296 | rv = SS_CW_NO_NEED; |
3509502d | 297 | if (rv == SS_UNKNOWN_ERROR) { |
b30ab791 | 298 | rv = is_valid_state(device, ns); |
a3025a27 | 299 | if (rv >= SS_SUCCESS) { |
a6b32bc3 | 300 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
a3025a27 | 301 | if (rv >= SS_SUCCESS) |
b8907339 PR |
302 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ |
303 | } | |
304 | } | |
0500813f | 305 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
306 | |
307 | return rv; | |
308 | } | |
309 | ||
310 | /** | |
311 | * drbd_req_state() - Perform an eventually cluster wide state change | |
b30ab791 | 312 | * @device: DRBD device. |
b8907339 PR |
313 | * @mask: mask of state bits to change. |
314 | * @val: value of new state bits. | |
315 | * @f: flags | |
316 | * | |
317 | * Should not be called directly, use drbd_request_state() or | |
318 | * _drbd_request_state(). | |
319 | */ | |
320 | static enum drbd_state_rv | |
b30ab791 | 321 | drbd_req_state(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
322 | union drbd_state val, enum chg_state_flags f) |
323 | { | |
324 | struct completion done; | |
325 | unsigned long flags; | |
326 | union drbd_state os, ns; | |
327 | enum drbd_state_rv rv; | |
328 | ||
329 | init_completion(&done); | |
330 | ||
331 | if (f & CS_SERIALIZE) | |
b30ab791 | 332 | mutex_lock(device->state_mutex); |
b8907339 | 333 | |
0500813f | 334 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 335 | os = drbd_read_state(device); |
d7fe69c6 | 336 | ns = sanitize_state(device, os, apply_mask_val(os, mask, val), NULL); |
3509502d | 337 | rv = is_valid_transition(os, ns); |
3c5e5f6a | 338 | if (rv < SS_SUCCESS) { |
0500813f | 339 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
3509502d | 340 | goto abort; |
3c5e5f6a | 341 | } |
b8907339 | 342 | |
b30ab791 AG |
343 | if (cl_wide_st_chg(device, os, ns)) { |
344 | rv = is_valid_state(device, ns); | |
b8907339 | 345 | if (rv == SS_SUCCESS) |
a6b32bc3 | 346 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
0500813f | 347 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
348 | |
349 | if (rv < SS_SUCCESS) { | |
350 | if (f & CS_VERBOSE) | |
b30ab791 | 351 | print_st_err(device, os, ns, rv); |
b8907339 PR |
352 | goto abort; |
353 | } | |
354 | ||
69a22773 | 355 | if (drbd_send_state_req(first_peer_device(device), mask, val)) { |
b8907339 PR |
356 | rv = SS_CW_FAILED_BY_PEER; |
357 | if (f & CS_VERBOSE) | |
b30ab791 | 358 | print_st_err(device, os, ns, rv); |
b8907339 PR |
359 | goto abort; |
360 | } | |
361 | ||
b30ab791 AG |
362 | wait_event(device->state_wait, |
363 | (rv = _req_st_cond(device, mask, val))); | |
b8907339 PR |
364 | |
365 | if (rv < SS_SUCCESS) { | |
b8907339 | 366 | if (f & CS_VERBOSE) |
b30ab791 | 367 | print_st_err(device, os, ns, rv); |
b8907339 PR |
368 | goto abort; |
369 | } | |
0500813f | 370 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 AG |
371 | ns = apply_mask_val(drbd_read_state(device), mask, val); |
372 | rv = _drbd_set_state(device, ns, f, &done); | |
b8907339 | 373 | } else { |
b30ab791 | 374 | rv = _drbd_set_state(device, ns, f, &done); |
b8907339 PR |
375 | } |
376 | ||
0500813f | 377 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b8907339 PR |
378 | |
379 | if (f & CS_WAIT_COMPLETE && rv == SS_SUCCESS) { | |
0b0ba1ef | 380 | D_ASSERT(device, current != first_peer_device(device)->connection->worker.task); |
b8907339 PR |
381 | wait_for_completion(&done); |
382 | } | |
383 | ||
384 | abort: | |
385 | if (f & CS_SERIALIZE) | |
b30ab791 | 386 | mutex_unlock(device->state_mutex); |
b8907339 PR |
387 | |
388 | return rv; | |
389 | } | |
390 | ||
391 | /** | |
392 | * _drbd_request_state() - Request a state change (with flags) | |
b30ab791 | 393 | * @device: DRBD device. |
b8907339 PR |
394 | * @mask: mask of state bits to change. |
395 | * @val: value of new state bits. | |
396 | * @f: flags | |
397 | * | |
398 | * Cousin of drbd_request_state(), useful with the CS_WAIT_COMPLETE | |
399 | * flag, or when logging of failed state change requests is not desired. | |
400 | */ | |
401 | enum drbd_state_rv | |
b30ab791 | 402 | _drbd_request_state(struct drbd_device *device, union drbd_state mask, |
b8907339 PR |
403 | union drbd_state val, enum chg_state_flags f) |
404 | { | |
405 | enum drbd_state_rv rv; | |
406 | ||
b30ab791 AG |
407 | wait_event(device->state_wait, |
408 | (rv = drbd_req_state(device, mask, val, f)) != SS_IN_TRANSIENT_STATE); | |
b8907339 PR |
409 | |
410 | return rv; | |
411 | } | |
412 | ||
b30ab791 | 413 | static void print_st(struct drbd_device *device, char *name, union drbd_state ns) |
b8907339 | 414 | { |
d0180171 | 415 | drbd_err(device, " %s = { cs:%s ro:%s/%s ds:%s/%s %c%c%c%c%c%c }\n", |
b8907339 PR |
416 | name, |
417 | drbd_conn_str(ns.conn), | |
418 | drbd_role_str(ns.role), | |
419 | drbd_role_str(ns.peer), | |
420 | drbd_disk_str(ns.disk), | |
421 | drbd_disk_str(ns.pdsk), | |
422 | is_susp(ns) ? 's' : 'r', | |
423 | ns.aftr_isp ? 'a' : '-', | |
424 | ns.peer_isp ? 'p' : '-', | |
425 | ns.user_isp ? 'u' : '-', | |
426 | ns.susp_fen ? 'F' : '-', | |
427 | ns.susp_nod ? 'N' : '-' | |
428 | ); | |
429 | } | |
430 | ||
b30ab791 | 431 | void print_st_err(struct drbd_device *device, union drbd_state os, |
b8907339 PR |
432 | union drbd_state ns, enum drbd_state_rv err) |
433 | { | |
434 | if (err == SS_IN_TRANSIENT_STATE) | |
435 | return; | |
d0180171 | 436 | drbd_err(device, "State change failed: %s\n", drbd_set_st_err_str(err)); |
b30ab791 AG |
437 | print_st(device, " state", os); |
438 | print_st(device, "wanted", ns); | |
b8907339 PR |
439 | } |
440 | ||
435693e8 | 441 | static long print_state_change(char *pb, union drbd_state os, union drbd_state ns, |
bbeb641c PR |
442 | enum chg_state_flags flags) |
443 | { | |
435693e8 | 444 | char *pbp; |
bbeb641c PR |
445 | pbp = pb; |
446 | *pbp = 0; | |
706cb24c | 447 | |
435693e8 | 448 | if (ns.role != os.role && flags & CS_DC_ROLE) |
bbeb641c PR |
449 | pbp += sprintf(pbp, "role( %s -> %s ) ", |
450 | drbd_role_str(os.role), | |
451 | drbd_role_str(ns.role)); | |
435693e8 | 452 | if (ns.peer != os.peer && flags & CS_DC_PEER) |
bbeb641c PR |
453 | pbp += sprintf(pbp, "peer( %s -> %s ) ", |
454 | drbd_role_str(os.peer), | |
455 | drbd_role_str(ns.peer)); | |
435693e8 | 456 | if (ns.conn != os.conn && flags & CS_DC_CONN) |
bbeb641c PR |
457 | pbp += sprintf(pbp, "conn( %s -> %s ) ", |
458 | drbd_conn_str(os.conn), | |
459 | drbd_conn_str(ns.conn)); | |
435693e8 | 460 | if (ns.disk != os.disk && flags & CS_DC_DISK) |
bbeb641c PR |
461 | pbp += sprintf(pbp, "disk( %s -> %s ) ", |
462 | drbd_disk_str(os.disk), | |
463 | drbd_disk_str(ns.disk)); | |
435693e8 | 464 | if (ns.pdsk != os.pdsk && flags & CS_DC_PDSK) |
bbeb641c PR |
465 | pbp += sprintf(pbp, "pdsk( %s -> %s ) ", |
466 | drbd_disk_str(os.pdsk), | |
467 | drbd_disk_str(ns.pdsk)); | |
706cb24c PR |
468 | |
469 | return pbp - pb; | |
470 | } | |
471 | ||
b30ab791 | 472 | static void drbd_pr_state_change(struct drbd_device *device, union drbd_state os, union drbd_state ns, |
706cb24c PR |
473 | enum chg_state_flags flags) |
474 | { | |
475 | char pb[300]; | |
476 | char *pbp = pb; | |
477 | ||
478 | pbp += print_state_change(pbp, os, ns, flags ^ CS_DC_MASK); | |
479 | ||
bbeb641c PR |
480 | if (ns.aftr_isp != os.aftr_isp) |
481 | pbp += sprintf(pbp, "aftr_isp( %d -> %d ) ", | |
482 | os.aftr_isp, | |
483 | ns.aftr_isp); | |
484 | if (ns.peer_isp != os.peer_isp) | |
485 | pbp += sprintf(pbp, "peer_isp( %d -> %d ) ", | |
486 | os.peer_isp, | |
487 | ns.peer_isp); | |
488 | if (ns.user_isp != os.user_isp) | |
489 | pbp += sprintf(pbp, "user_isp( %d -> %d ) ", | |
490 | os.user_isp, | |
491 | ns.user_isp); | |
435693e8 | 492 | |
706cb24c | 493 | if (pbp != pb) |
d0180171 | 494 | drbd_info(device, "%s\n", pb); |
bbeb641c | 495 | } |
b8907339 | 496 | |
bde89a9e | 497 | static void conn_pr_state_change(struct drbd_connection *connection, union drbd_state os, union drbd_state ns, |
435693e8 PR |
498 | enum chg_state_flags flags) |
499 | { | |
500 | char pb[300]; | |
706cb24c PR |
501 | char *pbp = pb; |
502 | ||
503 | pbp += print_state_change(pbp, os, ns, flags); | |
435693e8 | 504 | |
706cb24c PR |
505 | if (is_susp(ns) != is_susp(os) && flags & CS_DC_SUSP) |
506 | pbp += sprintf(pbp, "susp( %d -> %d ) ", | |
507 | is_susp(os), | |
508 | is_susp(ns)); | |
509 | ||
510 | if (pbp != pb) | |
1ec861eb | 511 | drbd_info(connection, "%s\n", pb); |
435693e8 PR |
512 | } |
513 | ||
514 | ||
b8907339 PR |
515 | /** |
516 | * is_valid_state() - Returns an SS_ error code if ns is not valid | |
b30ab791 | 517 | * @device: DRBD device. |
b8907339 PR |
518 | * @ns: State to consider. |
519 | */ | |
520 | static enum drbd_state_rv | |
b30ab791 | 521 | is_valid_state(struct drbd_device *device, union drbd_state ns) |
b8907339 PR |
522 | { |
523 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
524 | ||
525 | enum drbd_fencing_p fp; | |
526 | enum drbd_state_rv rv = SS_SUCCESS; | |
44ed167d | 527 | struct net_conf *nc; |
b8907339 | 528 | |
daeda1cc | 529 | rcu_read_lock(); |
b8907339 | 530 | fp = FP_DONT_CARE; |
b30ab791 AG |
531 | if (get_ldev(device)) { |
532 | fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
533 | put_ldev(device); | |
b8907339 PR |
534 | } |
535 | ||
a6b32bc3 | 536 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
537 | if (nc) { |
538 | if (!nc->two_primaries && ns.role == R_PRIMARY) { | |
047e95e2 PR |
539 | if (ns.peer == R_PRIMARY) |
540 | rv = SS_TWO_PRIMARIES; | |
a6b32bc3 | 541 | else if (conn_highest_peer(first_peer_device(device)->connection) == R_PRIMARY) |
047e95e2 | 542 | rv = SS_O_VOL_PEER_PRI; |
44ed167d | 543 | } |
b8907339 PR |
544 | } |
545 | ||
546 | if (rv <= 0) | |
547 | /* already found a reason to abort */; | |
b30ab791 | 548 | else if (ns.role == R_SECONDARY && device->open_cnt) |
b8907339 PR |
549 | rv = SS_DEVICE_IN_USE; |
550 | ||
551 | else if (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.disk < D_UP_TO_DATE) | |
552 | rv = SS_NO_UP_TO_DATE_DISK; | |
553 | ||
554 | else if (fp >= FP_RESOURCE && | |
555 | ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk >= D_UNKNOWN) | |
556 | rv = SS_PRIMARY_NOP; | |
557 | ||
558 | else if (ns.role == R_PRIMARY && ns.disk <= D_INCONSISTENT && ns.pdsk <= D_INCONSISTENT) | |
559 | rv = SS_NO_UP_TO_DATE_DISK; | |
560 | ||
561 | else if (ns.conn > C_CONNECTED && ns.disk < D_INCONSISTENT) | |
562 | rv = SS_NO_LOCAL_DISK; | |
563 | ||
564 | else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT) | |
565 | rv = SS_NO_REMOTE_DISK; | |
566 | ||
567 | else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) | |
568 | rv = SS_NO_UP_TO_DATE_DISK; | |
569 | ||
570 | else if ((ns.conn == C_CONNECTED || | |
571 | ns.conn == C_WF_BITMAP_S || | |
572 | ns.conn == C_SYNC_SOURCE || | |
573 | ns.conn == C_PAUSED_SYNC_S) && | |
574 | ns.disk == D_OUTDATED) | |
575 | rv = SS_CONNECTED_OUTDATES; | |
576 | ||
577 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
44ed167d | 578 | (nc->verify_alg[0] == 0)) |
b8907339 PR |
579 | rv = SS_NO_VERIFY_ALG; |
580 | ||
581 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
a6b32bc3 | 582 | first_peer_device(device)->connection->agreed_pro_version < 88) |
b8907339 PR |
583 | rv = SS_NOT_SUPPORTED; |
584 | ||
5c4f13d9 PR |
585 | else if (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) |
586 | rv = SS_NO_UP_TO_DATE_DISK; | |
587 | ||
588 | else if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
589 | ns.pdsk == D_UNKNOWN) | |
590 | rv = SS_NEED_CONNECTION; | |
591 | ||
b8907339 PR |
592 | else if (ns.conn >= C_CONNECTED && ns.pdsk == D_UNKNOWN) |
593 | rv = SS_CONNECTED_OUTDATES; | |
594 | ||
44ed167d PR |
595 | rcu_read_unlock(); |
596 | ||
b8907339 PR |
597 | return rv; |
598 | } | |
599 | ||
600 | /** | |
a75f34ad | 601 | * is_valid_soft_transition() - Returns an SS_ error code if the state transition is not possible |
3509502d PR |
602 | * This function limits state transitions that may be declined by DRBD. I.e. |
603 | * user requests (aka soft transitions). | |
b30ab791 | 604 | * @device: DRBD device. |
b8907339 PR |
605 | * @ns: new state. |
606 | * @os: old state. | |
607 | */ | |
608 | static enum drbd_state_rv | |
bde89a9e | 609 | is_valid_soft_transition(union drbd_state os, union drbd_state ns, struct drbd_connection *connection) |
b8907339 PR |
610 | { |
611 | enum drbd_state_rv rv = SS_SUCCESS; | |
612 | ||
613 | if ((ns.conn == C_STARTING_SYNC_T || ns.conn == C_STARTING_SYNC_S) && | |
614 | os.conn > C_CONNECTED) | |
615 | rv = SS_RESYNC_RUNNING; | |
616 | ||
617 | if (ns.conn == C_DISCONNECTING && os.conn == C_STANDALONE) | |
618 | rv = SS_ALREADY_STANDALONE; | |
619 | ||
620 | if (ns.disk > D_ATTACHING && os.disk == D_DISKLESS) | |
621 | rv = SS_IS_DISKLESS; | |
622 | ||
623 | if (ns.conn == C_WF_CONNECTION && os.conn < C_UNCONNECTED) | |
624 | rv = SS_NO_NET_CONFIG; | |
625 | ||
626 | if (ns.disk == D_OUTDATED && os.disk < D_OUTDATED && os.disk != D_ATTACHING) | |
627 | rv = SS_LOWER_THAN_OUTDATED; | |
628 | ||
629 | if (ns.conn == C_DISCONNECTING && os.conn == C_UNCONNECTED) | |
630 | rv = SS_IN_TRANSIENT_STATE; | |
631 | ||
2325eb66 PR |
632 | /* if (ns.conn == os.conn && ns.conn == C_WF_REPORT_PARAMS) |
633 | rv = SS_IN_TRANSIENT_STATE; */ | |
b8907339 | 634 | |
a1096a6e PR |
635 | /* While establishing a connection only allow cstate to change. |
636 | Delay/refuse role changes, detach attach etc... */ | |
bde89a9e | 637 | if (test_bit(STATE_SENT, &connection->flags) && |
a1096a6e PR |
638 | !(os.conn == C_WF_REPORT_PARAMS || |
639 | (ns.conn == C_WF_REPORT_PARAMS && os.conn == C_WF_CONNECTION))) | |
640 | rv = SS_IN_TRANSIENT_STATE; | |
641 | ||
b8907339 PR |
642 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED) |
643 | rv = SS_NEED_CONNECTION; | |
644 | ||
645 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
646 | ns.conn != os.conn && os.conn > C_CONNECTED) | |
647 | rv = SS_RESYNC_RUNNING; | |
648 | ||
649 | if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
650 | os.conn < C_CONNECTED) | |
651 | rv = SS_NEED_CONNECTION; | |
652 | ||
653 | if ((ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE) | |
654 | && os.conn < C_WF_REPORT_PARAMS) | |
655 | rv = SS_NEED_CONNECTION; /* No NetworkFailure -> SyncTarget etc... */ | |
656 | ||
2bd5ed5d PR |
657 | if (ns.conn == C_DISCONNECTING && ns.pdsk == D_OUTDATED && |
658 | os.conn < C_CONNECTED && os.pdsk > D_OUTDATED) | |
659 | rv = SS_OUTDATE_WO_CONN; | |
660 | ||
b8907339 PR |
661 | return rv; |
662 | } | |
663 | ||
3509502d | 664 | static enum drbd_state_rv |
fda74117 | 665 | is_valid_conn_transition(enum drbd_conns oc, enum drbd_conns nc) |
3509502d | 666 | { |
d9cc6e23 LE |
667 | /* no change -> nothing to do, at least for the connection part */ |
668 | if (oc == nc) | |
669 | return SS_NOTHING_TO_DO; | |
3509502d | 670 | |
d9cc6e23 LE |
671 | /* disconnect of an unconfigured connection does not make sense */ |
672 | if (oc == C_STANDALONE && nc == C_DISCONNECTING) | |
673 | return SS_ALREADY_STANDALONE; | |
674 | ||
675 | /* from C_STANDALONE, we start with C_UNCONNECTED */ | |
676 | if (oc == C_STANDALONE && nc != C_UNCONNECTED) | |
677 | return SS_NEED_CONNECTION; | |
3509502d | 678 | |
25b0d6c8 PR |
679 | /* When establishing a connection we need to go through WF_REPORT_PARAMS! |
680 | Necessary to do the right thing upon invalidate-remote on a disconnected resource */ | |
681 | if (oc < C_WF_REPORT_PARAMS && nc >= C_CONNECTED) | |
682 | return SS_NEED_CONNECTION; | |
683 | ||
3509502d | 684 | /* After a network error only C_UNCONNECTED or C_DISCONNECTING may follow. */ |
fda74117 | 685 | if (oc >= C_TIMEOUT && oc <= C_TEAR_DOWN && nc != C_UNCONNECTED && nc != C_DISCONNECTING) |
d9cc6e23 | 686 | return SS_IN_TRANSIENT_STATE; |
3509502d PR |
687 | |
688 | /* After C_DISCONNECTING only C_STANDALONE may follow */ | |
fda74117 | 689 | if (oc == C_DISCONNECTING && nc != C_STANDALONE) |
d9cc6e23 | 690 | return SS_IN_TRANSIENT_STATE; |
3509502d | 691 | |
d9cc6e23 | 692 | return SS_SUCCESS; |
fda74117 PR |
693 | } |
694 | ||
695 | ||
696 | /** | |
697 | * is_valid_transition() - Returns an SS_ error code if the state transition is not possible | |
698 | * This limits hard state transitions. Hard state transitions are facts there are | |
699 | * imposed on DRBD by the environment. E.g. disk broke or network broke down. | |
700 | * But those hard state transitions are still not allowed to do everything. | |
701 | * @ns: new state. | |
702 | * @os: old state. | |
703 | */ | |
704 | static enum drbd_state_rv | |
705 | is_valid_transition(union drbd_state os, union drbd_state ns) | |
706 | { | |
707 | enum drbd_state_rv rv; | |
708 | ||
709 | rv = is_valid_conn_transition(os.conn, ns.conn); | |
710 | ||
3509502d PR |
711 | /* we cannot fail (again) if we already detached */ |
712 | if (ns.disk == D_FAILED && os.disk == D_DISKLESS) | |
713 | rv = SS_IS_DISKLESS; | |
714 | ||
715 | return rv; | |
716 | } | |
717 | ||
b30ab791 | 718 | static void print_sanitize_warnings(struct drbd_device *device, enum sanitize_state_warnings warn) |
d942ae44 PR |
719 | { |
720 | static const char *msg_table[] = { | |
721 | [NO_WARNING] = "", | |
722 | [ABORTED_ONLINE_VERIFY] = "Online-verify aborted.", | |
723 | [ABORTED_RESYNC] = "Resync aborted.", | |
724 | [CONNECTION_LOST_NEGOTIATING] = "Connection lost while negotiating, no data!", | |
725 | [IMPLICITLY_UPGRADED_DISK] = "Implicitly upgraded disk", | |
726 | [IMPLICITLY_UPGRADED_PDSK] = "Implicitly upgraded pdsk", | |
727 | }; | |
728 | ||
729 | if (warn != NO_WARNING) | |
d0180171 | 730 | drbd_warn(device, "%s\n", msg_table[warn]); |
d942ae44 PR |
731 | } |
732 | ||
b8907339 PR |
733 | /** |
734 | * sanitize_state() - Resolves implicitly necessary additional changes to a state transition | |
b30ab791 | 735 | * @device: DRBD device. |
b8907339 PR |
736 | * @os: old state. |
737 | * @ns: new state. | |
738 | * @warn_sync_abort: | |
739 | * | |
740 | * When we loose connection, we have to set the state of the peers disk (pdsk) | |
741 | * to D_UNKNOWN. This rule and many more along those lines are in this function. | |
742 | */ | |
d7fe69c6 PR |
743 | static union drbd_state sanitize_state(struct drbd_device *device, union drbd_state os, |
744 | union drbd_state ns, enum sanitize_state_warnings *warn) | |
b8907339 PR |
745 | { |
746 | enum drbd_fencing_p fp; | |
747 | enum drbd_disk_state disk_min, disk_max, pdsk_min, pdsk_max; | |
748 | ||
d942ae44 PR |
749 | if (warn) |
750 | *warn = NO_WARNING; | |
751 | ||
b8907339 | 752 | fp = FP_DONT_CARE; |
b30ab791 | 753 | if (get_ldev(device)) { |
daeda1cc | 754 | rcu_read_lock(); |
b30ab791 | 755 | fp = rcu_dereference(device->ldev->disk_conf)->fencing; |
daeda1cc | 756 | rcu_read_unlock(); |
b30ab791 | 757 | put_ldev(device); |
b8907339 PR |
758 | } |
759 | ||
3509502d | 760 | /* Implications from connection to peer and peer_isp */ |
b8907339 PR |
761 | if (ns.conn < C_CONNECTED) { |
762 | ns.peer_isp = 0; | |
763 | ns.peer = R_UNKNOWN; | |
764 | if (ns.pdsk > D_UNKNOWN || ns.pdsk < D_INCONSISTENT) | |
765 | ns.pdsk = D_UNKNOWN; | |
766 | } | |
767 | ||
768 | /* Clear the aftr_isp when becoming unconfigured */ | |
769 | if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY) | |
770 | ns.aftr_isp = 0; | |
771 | ||
4308a0a3 | 772 | /* An implication of the disk states onto the connection state */ |
b8907339 | 773 | /* Abort resync if a disk fails/detaches */ |
4308a0a3 | 774 | if (ns.conn > C_CONNECTED && (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) { |
d942ae44 PR |
775 | if (warn) |
776 | *warn = ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T ? | |
777 | ABORTED_ONLINE_VERIFY : ABORTED_RESYNC; | |
b8907339 PR |
778 | ns.conn = C_CONNECTED; |
779 | } | |
780 | ||
781 | /* Connection breaks down before we finished "Negotiating" */ | |
782 | if (ns.conn < C_CONNECTED && ns.disk == D_NEGOTIATING && | |
b30ab791 AG |
783 | get_ldev_if_state(device, D_NEGOTIATING)) { |
784 | if (device->ed_uuid == device->ldev->md.uuid[UI_CURRENT]) { | |
785 | ns.disk = device->new_state_tmp.disk; | |
786 | ns.pdsk = device->new_state_tmp.pdsk; | |
b8907339 | 787 | } else { |
d942ae44 PR |
788 | if (warn) |
789 | *warn = CONNECTION_LOST_NEGOTIATING; | |
b8907339 PR |
790 | ns.disk = D_DISKLESS; |
791 | ns.pdsk = D_UNKNOWN; | |
792 | } | |
b30ab791 | 793 | put_ldev(device); |
b8907339 PR |
794 | } |
795 | ||
796 | /* D_CONSISTENT and D_OUTDATED vanish when we get connected */ | |
797 | if (ns.conn >= C_CONNECTED && ns.conn < C_AHEAD) { | |
798 | if (ns.disk == D_CONSISTENT || ns.disk == D_OUTDATED) | |
799 | ns.disk = D_UP_TO_DATE; | |
800 | if (ns.pdsk == D_CONSISTENT || ns.pdsk == D_OUTDATED) | |
801 | ns.pdsk = D_UP_TO_DATE; | |
802 | } | |
803 | ||
804 | /* Implications of the connection stat on the disk states */ | |
805 | disk_min = D_DISKLESS; | |
806 | disk_max = D_UP_TO_DATE; | |
807 | pdsk_min = D_INCONSISTENT; | |
808 | pdsk_max = D_UNKNOWN; | |
809 | switch ((enum drbd_conns)ns.conn) { | |
810 | case C_WF_BITMAP_T: | |
811 | case C_PAUSED_SYNC_T: | |
812 | case C_STARTING_SYNC_T: | |
813 | case C_WF_SYNC_UUID: | |
814 | case C_BEHIND: | |
815 | disk_min = D_INCONSISTENT; | |
816 | disk_max = D_OUTDATED; | |
817 | pdsk_min = D_UP_TO_DATE; | |
818 | pdsk_max = D_UP_TO_DATE; | |
819 | break; | |
820 | case C_VERIFY_S: | |
821 | case C_VERIFY_T: | |
822 | disk_min = D_UP_TO_DATE; | |
823 | disk_max = D_UP_TO_DATE; | |
824 | pdsk_min = D_UP_TO_DATE; | |
825 | pdsk_max = D_UP_TO_DATE; | |
826 | break; | |
827 | case C_CONNECTED: | |
828 | disk_min = D_DISKLESS; | |
829 | disk_max = D_UP_TO_DATE; | |
830 | pdsk_min = D_DISKLESS; | |
831 | pdsk_max = D_UP_TO_DATE; | |
832 | break; | |
833 | case C_WF_BITMAP_S: | |
834 | case C_PAUSED_SYNC_S: | |
835 | case C_STARTING_SYNC_S: | |
836 | case C_AHEAD: | |
837 | disk_min = D_UP_TO_DATE; | |
838 | disk_max = D_UP_TO_DATE; | |
839 | pdsk_min = D_INCONSISTENT; | |
840 | pdsk_max = D_CONSISTENT; /* D_OUTDATED would be nice. But explicit outdate necessary*/ | |
841 | break; | |
842 | case C_SYNC_TARGET: | |
843 | disk_min = D_INCONSISTENT; | |
844 | disk_max = D_INCONSISTENT; | |
845 | pdsk_min = D_UP_TO_DATE; | |
846 | pdsk_max = D_UP_TO_DATE; | |
847 | break; | |
848 | case C_SYNC_SOURCE: | |
849 | disk_min = D_UP_TO_DATE; | |
850 | disk_max = D_UP_TO_DATE; | |
851 | pdsk_min = D_INCONSISTENT; | |
852 | pdsk_max = D_INCONSISTENT; | |
853 | break; | |
854 | case C_STANDALONE: | |
855 | case C_DISCONNECTING: | |
856 | case C_UNCONNECTED: | |
857 | case C_TIMEOUT: | |
858 | case C_BROKEN_PIPE: | |
859 | case C_NETWORK_FAILURE: | |
860 | case C_PROTOCOL_ERROR: | |
861 | case C_TEAR_DOWN: | |
862 | case C_WF_CONNECTION: | |
863 | case C_WF_REPORT_PARAMS: | |
864 | case C_MASK: | |
865 | break; | |
866 | } | |
867 | if (ns.disk > disk_max) | |
868 | ns.disk = disk_max; | |
869 | ||
870 | if (ns.disk < disk_min) { | |
d942ae44 PR |
871 | if (warn) |
872 | *warn = IMPLICITLY_UPGRADED_DISK; | |
b8907339 PR |
873 | ns.disk = disk_min; |
874 | } | |
875 | if (ns.pdsk > pdsk_max) | |
876 | ns.pdsk = pdsk_max; | |
877 | ||
878 | if (ns.pdsk < pdsk_min) { | |
d942ae44 PR |
879 | if (warn) |
880 | *warn = IMPLICITLY_UPGRADED_PDSK; | |
b8907339 PR |
881 | ns.pdsk = pdsk_min; |
882 | } | |
883 | ||
884 | if (fp == FP_STONITH && | |
d7fe69c6 PR |
885 | (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED) && |
886 | !(os.role == R_PRIMARY && os.conn < C_CONNECTED && os.pdsk > D_OUTDATED)) | |
b8907339 PR |
887 | ns.susp_fen = 1; /* Suspend IO while fence-peer handler runs (peer lost) */ |
888 | ||
eb6bea67 | 889 | if (device->resource->res_opts.on_no_data == OND_SUSPEND_IO && |
d7fe69c6 PR |
890 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) && |
891 | !(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE)) | |
b8907339 PR |
892 | ns.susp_nod = 1; /* Suspend IO while no data available (no accessible data available) */ |
893 | ||
894 | if (ns.aftr_isp || ns.peer_isp || ns.user_isp) { | |
895 | if (ns.conn == C_SYNC_SOURCE) | |
896 | ns.conn = C_PAUSED_SYNC_S; | |
897 | if (ns.conn == C_SYNC_TARGET) | |
898 | ns.conn = C_PAUSED_SYNC_T; | |
899 | } else { | |
900 | if (ns.conn == C_PAUSED_SYNC_S) | |
901 | ns.conn = C_SYNC_SOURCE; | |
902 | if (ns.conn == C_PAUSED_SYNC_T) | |
903 | ns.conn = C_SYNC_TARGET; | |
904 | } | |
905 | ||
906 | return ns; | |
907 | } | |
908 | ||
b30ab791 | 909 | void drbd_resume_al(struct drbd_device *device) |
b8907339 | 910 | { |
b30ab791 | 911 | if (test_and_clear_bit(AL_SUSPENDED, &device->flags)) |
d0180171 | 912 | drbd_info(device, "Resumed AL updates\n"); |
b8907339 PR |
913 | } |
914 | ||
915 | /* helper for __drbd_set_state */ | |
b30ab791 | 916 | static void set_ov_position(struct drbd_device *device, enum drbd_conns cs) |
b8907339 | 917 | { |
a6b32bc3 | 918 | if (first_peer_device(device)->connection->agreed_pro_version < 90) |
b30ab791 AG |
919 | device->ov_start_sector = 0; |
920 | device->rs_total = drbd_bm_bits(device); | |
921 | device->ov_position = 0; | |
b8907339 PR |
922 | if (cs == C_VERIFY_T) { |
923 | /* starting online verify from an arbitrary position | |
924 | * does not fit well into the existing protocol. | |
925 | * on C_VERIFY_T, we initialize ov_left and friends | |
926 | * implicitly in receive_DataRequest once the | |
927 | * first P_OV_REQUEST is received */ | |
b30ab791 | 928 | device->ov_start_sector = ~(sector_t)0; |
b8907339 | 929 | } else { |
b30ab791 AG |
930 | unsigned long bit = BM_SECT_TO_BIT(device->ov_start_sector); |
931 | if (bit >= device->rs_total) { | |
932 | device->ov_start_sector = | |
933 | BM_BIT_TO_SECT(device->rs_total - 1); | |
934 | device->rs_total = 1; | |
b8907339 | 935 | } else |
b30ab791 AG |
936 | device->rs_total -= bit; |
937 | device->ov_position = device->ov_start_sector; | |
b8907339 | 938 | } |
b30ab791 | 939 | device->ov_left = device->rs_total; |
b8907339 PR |
940 | } |
941 | ||
942 | /** | |
943 | * __drbd_set_state() - Set a new DRBD state | |
b30ab791 | 944 | * @device: DRBD device. |
b8907339 PR |
945 | * @ns: new state. |
946 | * @flags: Flags | |
947 | * @done: Optional completion, that will get completed after the after_state_ch() finished | |
948 | * | |
949 | * Caller needs to hold req_lock, and global_state_lock. Do not call directly. | |
950 | */ | |
951 | enum drbd_state_rv | |
b30ab791 | 952 | __drbd_set_state(struct drbd_device *device, union drbd_state ns, |
b8907339 PR |
953 | enum chg_state_flags flags, struct completion *done) |
954 | { | |
955 | union drbd_state os; | |
956 | enum drbd_state_rv rv = SS_SUCCESS; | |
d942ae44 | 957 | enum sanitize_state_warnings ssw; |
b8907339 | 958 | struct after_state_chg_work *ascw; |
2681f7f6 | 959 | bool did_remote, should_do_remote; |
b8907339 | 960 | |
b30ab791 | 961 | os = drbd_read_state(device); |
b8907339 | 962 | |
d7fe69c6 | 963 | ns = sanitize_state(device, os, ns, &ssw); |
b8907339 PR |
964 | if (ns.i == os.i) |
965 | return SS_NOTHING_TO_DO; | |
966 | ||
3509502d PR |
967 | rv = is_valid_transition(os, ns); |
968 | if (rv < SS_SUCCESS) | |
969 | return rv; | |
970 | ||
b8907339 PR |
971 | if (!(flags & CS_HARD)) { |
972 | /* pre-state-change checks ; only look at ns */ | |
973 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
974 | ||
b30ab791 | 975 | rv = is_valid_state(device, ns); |
b8907339 PR |
976 | if (rv < SS_SUCCESS) { |
977 | /* If the old state was illegal as well, then let | |
978 | this happen...*/ | |
979 | ||
b30ab791 | 980 | if (is_valid_state(device, os) == rv) |
a6b32bc3 | 981 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
b8907339 | 982 | } else |
a6b32bc3 | 983 | rv = is_valid_soft_transition(os, ns, first_peer_device(device)->connection); |
b8907339 PR |
984 | } |
985 | ||
986 | if (rv < SS_SUCCESS) { | |
987 | if (flags & CS_VERBOSE) | |
b30ab791 | 988 | print_st_err(device, os, ns, rv); |
b8907339 PR |
989 | return rv; |
990 | } | |
991 | ||
b30ab791 | 992 | print_sanitize_warnings(device, ssw); |
b8907339 | 993 | |
b30ab791 | 994 | drbd_pr_state_change(device, os, ns, flags); |
b8907339 | 995 | |
706cb24c PR |
996 | /* Display changes to the susp* flags that where caused by the call to |
997 | sanitize_state(). Only display it here if we where not called from | |
998 | _conn_request_state() */ | |
999 | if (!(flags & CS_DC_SUSP)) | |
a6b32bc3 AG |
1000 | conn_pr_state_change(first_peer_device(device)->connection, os, ns, |
1001 | (flags & ~CS_DC_MASK) | CS_DC_SUSP); | |
706cb24c | 1002 | |
b8907339 PR |
1003 | /* if we are going -> D_FAILED or D_DISKLESS, grab one extra reference |
1004 | * on the ldev here, to be sure the transition -> D_DISKLESS resp. | |
1005 | * drbd_ldev_destroy() won't happen before our corresponding | |
1006 | * after_state_ch works run, where we put_ldev again. */ | |
1007 | if ((os.disk != D_FAILED && ns.disk == D_FAILED) || | |
1008 | (os.disk != D_DISKLESS && ns.disk == D_DISKLESS)) | |
b30ab791 | 1009 | atomic_inc(&device->local_cnt); |
b8907339 | 1010 | |
b30ab791 AG |
1011 | did_remote = drbd_should_do_remote(device->state); |
1012 | device->state.i = ns.i; | |
1013 | should_do_remote = drbd_should_do_remote(device->state); | |
6bbf53ca AG |
1014 | device->resource->susp = ns.susp; |
1015 | device->resource->susp_nod = ns.susp_nod; | |
1016 | device->resource->susp_fen = ns.susp_fen; | |
b8907339 | 1017 | |
2681f7f6 LE |
1018 | /* put replicated vs not-replicated requests in seperate epochs */ |
1019 | if (did_remote != should_do_remote) | |
a6b32bc3 | 1020 | start_new_tl_epoch(first_peer_device(device)->connection); |
2681f7f6 | 1021 | |
b8907339 | 1022 | if (os.disk == D_ATTACHING && ns.disk >= D_NEGOTIATING) |
b30ab791 | 1023 | drbd_print_uuids(device, "attached to UUIDs"); |
b8907339 | 1024 | |
79702011 PR |
1025 | /* Wake up role changes, that were delayed because of connection establishing */ |
1026 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn != C_WF_REPORT_PARAMS && | |
a6b32bc3 AG |
1027 | no_peer_wf_report_params(first_peer_device(device)->connection)) |
1028 | clear_bit(STATE_SENT, &first_peer_device(device)->connection->flags); | |
79702011 | 1029 | |
b30ab791 AG |
1030 | wake_up(&device->misc_wait); |
1031 | wake_up(&device->state_wait); | |
a6b32bc3 | 1032 | wake_up(&first_peer_device(device)->connection->ping_wait); |
b8907339 | 1033 | |
58ffa580 LE |
1034 | /* Aborted verify run, or we reached the stop sector. |
1035 | * Log the last position, unless end-of-device. */ | |
b8907339 | 1036 | if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) && |
58ffa580 | 1037 | ns.conn <= C_CONNECTED) { |
b30ab791 AG |
1038 | device->ov_start_sector = |
1039 | BM_BIT_TO_SECT(drbd_bm_bits(device) - device->ov_left); | |
1040 | if (device->ov_left) | |
d0180171 | 1041 | drbd_info(device, "Online Verify reached sector %llu\n", |
b30ab791 | 1042 | (unsigned long long)device->ov_start_sector); |
b8907339 PR |
1043 | } |
1044 | ||
1045 | if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) && | |
1046 | (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) { | |
d0180171 | 1047 | drbd_info(device, "Syncer continues.\n"); |
b30ab791 AG |
1048 | device->rs_paused += (long)jiffies |
1049 | -(long)device->rs_mark_time[device->rs_last_mark]; | |
b8907339 | 1050 | if (ns.conn == C_SYNC_TARGET) |
b30ab791 | 1051 | mod_timer(&device->resync_timer, jiffies); |
b8907339 PR |
1052 | } |
1053 | ||
1054 | if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) && | |
1055 | (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) { | |
d0180171 | 1056 | drbd_info(device, "Resync suspended\n"); |
b30ab791 | 1057 | device->rs_mark_time[device->rs_last_mark] = jiffies; |
b8907339 PR |
1058 | } |
1059 | ||
1060 | if (os.conn == C_CONNECTED && | |
1061 | (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) { | |
1062 | unsigned long now = jiffies; | |
1063 | int i; | |
1064 | ||
b30ab791 AG |
1065 | set_ov_position(device, ns.conn); |
1066 | device->rs_start = now; | |
1067 | device->rs_last_events = 0; | |
1068 | device->rs_last_sect_ev = 0; | |
1069 | device->ov_last_oos_size = 0; | |
1070 | device->ov_last_oos_start = 0; | |
b8907339 PR |
1071 | |
1072 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { | |
b30ab791 AG |
1073 | device->rs_mark_left[i] = device->ov_left; |
1074 | device->rs_mark_time[i] = now; | |
b8907339 PR |
1075 | } |
1076 | ||
b30ab791 | 1077 | drbd_rs_controller_reset(device); |
b8907339 PR |
1078 | |
1079 | if (ns.conn == C_VERIFY_S) { | |
d0180171 | 1080 | drbd_info(device, "Starting Online Verify from sector %llu\n", |
b30ab791 AG |
1081 | (unsigned long long)device->ov_position); |
1082 | mod_timer(&device->resync_timer, jiffies); | |
b8907339 PR |
1083 | } |
1084 | } | |
1085 | ||
b30ab791 AG |
1086 | if (get_ldev(device)) { |
1087 | u32 mdf = device->ldev->md.flags & ~(MDF_CONSISTENT|MDF_PRIMARY_IND| | |
b8907339 PR |
1088 | MDF_CONNECTED_IND|MDF_WAS_UP_TO_DATE| |
1089 | MDF_PEER_OUT_DATED|MDF_CRASHED_PRIMARY); | |
1090 | ||
d5d7ebd4 | 1091 | mdf &= ~MDF_AL_CLEAN; |
b30ab791 | 1092 | if (test_bit(CRASHED_PRIMARY, &device->flags)) |
b8907339 | 1093 | mdf |= MDF_CRASHED_PRIMARY; |
b30ab791 AG |
1094 | if (device->state.role == R_PRIMARY || |
1095 | (device->state.pdsk < D_INCONSISTENT && device->state.peer == R_PRIMARY)) | |
b8907339 | 1096 | mdf |= MDF_PRIMARY_IND; |
b30ab791 | 1097 | if (device->state.conn > C_WF_REPORT_PARAMS) |
b8907339 | 1098 | mdf |= MDF_CONNECTED_IND; |
b30ab791 | 1099 | if (device->state.disk > D_INCONSISTENT) |
b8907339 | 1100 | mdf |= MDF_CONSISTENT; |
b30ab791 | 1101 | if (device->state.disk > D_OUTDATED) |
b8907339 | 1102 | mdf |= MDF_WAS_UP_TO_DATE; |
b30ab791 | 1103 | if (device->state.pdsk <= D_OUTDATED && device->state.pdsk >= D_INCONSISTENT) |
b8907339 | 1104 | mdf |= MDF_PEER_OUT_DATED; |
b30ab791 AG |
1105 | if (mdf != device->ldev->md.flags) { |
1106 | device->ldev->md.flags = mdf; | |
1107 | drbd_md_mark_dirty(device); | |
b8907339 PR |
1108 | } |
1109 | if (os.disk < D_CONSISTENT && ns.disk >= D_CONSISTENT) | |
b30ab791 AG |
1110 | drbd_set_ed_uuid(device, device->ldev->md.uuid[UI_CURRENT]); |
1111 | put_ldev(device); | |
b8907339 PR |
1112 | } |
1113 | ||
1114 | /* Peer was forced D_UP_TO_DATE & R_PRIMARY, consider to resync */ | |
1115 | if (os.disk == D_INCONSISTENT && os.pdsk == D_INCONSISTENT && | |
1116 | os.peer == R_SECONDARY && ns.peer == R_PRIMARY) | |
b30ab791 | 1117 | set_bit(CONSIDER_RESYNC, &device->flags); |
b8907339 PR |
1118 | |
1119 | /* Receiver should clean up itself */ | |
1120 | if (os.conn != C_DISCONNECTING && ns.conn == C_DISCONNECTING) | |
a6b32bc3 | 1121 | drbd_thread_stop_nowait(&first_peer_device(device)->connection->receiver); |
b8907339 PR |
1122 | |
1123 | /* Now the receiver finished cleaning up itself, it should die */ | |
1124 | if (os.conn != C_STANDALONE && ns.conn == C_STANDALONE) | |
a6b32bc3 | 1125 | drbd_thread_stop_nowait(&first_peer_device(device)->connection->receiver); |
b8907339 PR |
1126 | |
1127 | /* Upon network failure, we need to restart the receiver. */ | |
823bd832 | 1128 | if (os.conn > C_WF_CONNECTION && |
b8907339 | 1129 | ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT) |
a6b32bc3 | 1130 | drbd_thread_restart_nowait(&first_peer_device(device)->connection->receiver); |
b8907339 PR |
1131 | |
1132 | /* Resume AL writing if we get a connection */ | |
28e448bb | 1133 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { |
b30ab791 | 1134 | drbd_resume_al(device); |
a6b32bc3 | 1135 | first_peer_device(device)->connection->connect_cnt++; |
28e448bb | 1136 | } |
b8907339 | 1137 | |
07be15b1 LE |
1138 | /* remember last attach time so request_timer_fn() won't |
1139 | * kill newly established sessions while we are still trying to thaw | |
1140 | * previously frozen IO */ | |
1141 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && | |
1142 | ns.disk > D_NEGOTIATING) | |
b30ab791 | 1143 | device->last_reattach_jif = jiffies; |
07be15b1 | 1144 | |
b8907339 PR |
1145 | ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC); |
1146 | if (ascw) { | |
1147 | ascw->os = os; | |
1148 | ascw->ns = ns; | |
1149 | ascw->flags = flags; | |
1150 | ascw->w.cb = w_after_state_ch; | |
84b8c06b | 1151 | ascw->device = device; |
b8907339 | 1152 | ascw->done = done; |
84b8c06b AG |
1153 | drbd_queue_work(&first_peer_device(device)->connection->sender_work, |
1154 | &ascw->w); | |
b8907339 | 1155 | } else { |
d0180171 | 1156 | drbd_err(device, "Could not kmalloc an ascw\n"); |
b8907339 PR |
1157 | } |
1158 | ||
1159 | return rv; | |
1160 | } | |
1161 | ||
99920dc5 | 1162 | static int w_after_state_ch(struct drbd_work *w, int unused) |
b8907339 PR |
1163 | { |
1164 | struct after_state_chg_work *ascw = | |
1165 | container_of(w, struct after_state_chg_work, w); | |
84b8c06b | 1166 | struct drbd_device *device = ascw->device; |
b8907339 | 1167 | |
b30ab791 | 1168 | after_state_ch(device, ascw->os, ascw->ns, ascw->flags); |
137975c1 | 1169 | if (ascw->flags & CS_WAIT_COMPLETE) |
b8907339 | 1170 | complete(ascw->done); |
b8907339 PR |
1171 | kfree(ascw); |
1172 | ||
99920dc5 | 1173 | return 0; |
b8907339 PR |
1174 | } |
1175 | ||
b30ab791 | 1176 | static void abw_start_sync(struct drbd_device *device, int rv) |
b8907339 PR |
1177 | { |
1178 | if (rv) { | |
d0180171 | 1179 | drbd_err(device, "Writing the bitmap failed not starting resync.\n"); |
b30ab791 | 1180 | _drbd_request_state(device, NS(conn, C_CONNECTED), CS_VERBOSE); |
b8907339 PR |
1181 | return; |
1182 | } | |
1183 | ||
b30ab791 | 1184 | switch (device->state.conn) { |
b8907339 | 1185 | case C_STARTING_SYNC_T: |
b30ab791 | 1186 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); |
b8907339 PR |
1187 | break; |
1188 | case C_STARTING_SYNC_S: | |
b30ab791 | 1189 | drbd_start_resync(device, C_SYNC_SOURCE); |
b8907339 PR |
1190 | break; |
1191 | } | |
1192 | } | |
1193 | ||
b30ab791 | 1194 | int drbd_bitmap_io_from_worker(struct drbd_device *device, |
54761697 | 1195 | int (*io_fn)(struct drbd_device *), |
b8907339 PR |
1196 | char *why, enum bm_flag flags) |
1197 | { | |
1198 | int rv; | |
1199 | ||
0b0ba1ef | 1200 | D_ASSERT(device, current == first_peer_device(device)->connection->worker.task); |
b8907339 | 1201 | |
b30ab791 AG |
1202 | /* open coded non-blocking drbd_suspend_io(device); */ |
1203 | set_bit(SUSPEND_IO, &device->flags); | |
b8907339 | 1204 | |
b30ab791 AG |
1205 | drbd_bm_lock(device, why, flags); |
1206 | rv = io_fn(device); | |
1207 | drbd_bm_unlock(device); | |
b8907339 | 1208 | |
b30ab791 | 1209 | drbd_resume_io(device); |
b8907339 PR |
1210 | |
1211 | return rv; | |
1212 | } | |
1213 | ||
1214 | /** | |
1215 | * after_state_ch() - Perform after state change actions that may sleep | |
b30ab791 | 1216 | * @device: DRBD device. |
b8907339 PR |
1217 | * @os: old state. |
1218 | * @ns: new state. | |
1219 | * @flags: Flags | |
1220 | */ | |
b30ab791 | 1221 | static void after_state_ch(struct drbd_device *device, union drbd_state os, |
b8907339 PR |
1222 | union drbd_state ns, enum chg_state_flags flags) |
1223 | { | |
6bbf53ca | 1224 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
1225 | struct sib_info sib; |
1226 | ||
1227 | sib.sib_reason = SIB_STATE_CHANGE; | |
1228 | sib.os = os; | |
1229 | sib.ns = ns; | |
b8907339 PR |
1230 | |
1231 | if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) { | |
b30ab791 AG |
1232 | clear_bit(CRASHED_PRIMARY, &device->flags); |
1233 | if (device->p_uuid) | |
1234 | device->p_uuid[UI_FLAGS] &= ~((u64)2); | |
b8907339 PR |
1235 | } |
1236 | ||
b8907339 | 1237 | /* Inform userspace about the change... */ |
b30ab791 | 1238 | drbd_bcast_event(device, &sib); |
b8907339 PR |
1239 | |
1240 | if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) && | |
1241 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) | |
b30ab791 | 1242 | drbd_khelper(device, "pri-on-incon-degr"); |
b8907339 PR |
1243 | |
1244 | /* Here we have the actions that are performed after a | |
1245 | state change. This function might sleep */ | |
1246 | ||
b8907339 | 1247 | if (ns.susp_nod) { |
a6b32bc3 | 1248 | struct drbd_connection *connection = first_peer_device(device)->connection; |
a6d00c8e PR |
1249 | enum drbd_req_event what = NOTHING; |
1250 | ||
0500813f | 1251 | spin_lock_irq(&device->resource->req_lock); |
bde89a9e | 1252 | if (os.conn < C_CONNECTED && conn_lowest_conn(connection) >= C_CONNECTED) |
b8907339 PR |
1253 | what = RESEND; |
1254 | ||
3fb4746d | 1255 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && |
bde89a9e | 1256 | conn_lowest_disk(connection) > D_NEGOTIATING) |
b8907339 PR |
1257 | what = RESTART_FROZEN_DISK_IO; |
1258 | ||
6bbf53ca | 1259 | if (resource->susp_nod && what != NOTHING) { |
bde89a9e AG |
1260 | _tl_restart(connection, what); |
1261 | _conn_request_state(connection, | |
892fdd1a PR |
1262 | (union drbd_state) { { .susp_nod = 1 } }, |
1263 | (union drbd_state) { { .susp_nod = 0 } }, | |
1264 | CS_VERBOSE); | |
b8907339 | 1265 | } |
0500813f | 1266 | spin_unlock_irq(&device->resource->req_lock); |
b8907339 PR |
1267 | } |
1268 | ||
88f79ec4 | 1269 | if (ns.susp_fen) { |
a6b32bc3 | 1270 | struct drbd_connection *connection = first_peer_device(device)->connection; |
88f79ec4 | 1271 | |
0500813f | 1272 | spin_lock_irq(&device->resource->req_lock); |
6bbf53ca | 1273 | if (resource->susp_fen && conn_lowest_conn(connection) >= C_CONNECTED) { |
88f79ec4 | 1274 | /* case2: The connection was established again: */ |
c06ece6b | 1275 | struct drbd_peer_device *peer_device; |
88f79ec4 PR |
1276 | int vnr; |
1277 | ||
1278 | rcu_read_lock(); | |
c06ece6b AG |
1279 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
1280 | clear_bit(NEW_CUR_UUID, &peer_device->device->flags); | |
88f79ec4 | 1281 | rcu_read_unlock(); |
bde89a9e AG |
1282 | _tl_restart(connection, RESEND); |
1283 | _conn_request_state(connection, | |
88f79ec4 PR |
1284 | (union drbd_state) { { .susp_fen = 1 } }, |
1285 | (union drbd_state) { { .susp_fen = 0 } }, | |
1286 | CS_VERBOSE); | |
1287 | } | |
0500813f | 1288 | spin_unlock_irq(&device->resource->req_lock); |
88f79ec4 PR |
1289 | } |
1290 | ||
b8907339 PR |
1291 | /* Became sync source. With protocol >= 96, we still need to send out |
1292 | * the sync uuid now. Need to do that before any drbd_send_state, or | |
1293 | * the other side may go "paused sync" before receiving the sync uuids, | |
1294 | * which is unexpected. */ | |
1295 | if ((os.conn != C_SYNC_SOURCE && os.conn != C_PAUSED_SYNC_S) && | |
1296 | (ns.conn == C_SYNC_SOURCE || ns.conn == C_PAUSED_SYNC_S) && | |
a6b32bc3 | 1297 | first_peer_device(device)->connection->agreed_pro_version >= 96 && get_ldev(device)) { |
69a22773 | 1298 | drbd_gen_and_send_sync_uuid(first_peer_device(device)); |
b30ab791 | 1299 | put_ldev(device); |
b8907339 PR |
1300 | } |
1301 | ||
1302 | /* Do not change the order of the if above and the two below... */ | |
369bea63 PR |
1303 | if (os.pdsk == D_DISKLESS && |
1304 | ns.pdsk > D_DISKLESS && ns.pdsk != D_UNKNOWN) { /* attach on the peer */ | |
a324896b LE |
1305 | /* we probably will start a resync soon. |
1306 | * make sure those things are properly reset. */ | |
b30ab791 AG |
1307 | device->rs_total = 0; |
1308 | device->rs_failed = 0; | |
1309 | atomic_set(&device->rs_pending_cnt, 0); | |
1310 | drbd_rs_cancel_all(device); | |
a324896b | 1311 | |
69a22773 AG |
1312 | drbd_send_uuids(first_peer_device(device)); |
1313 | drbd_send_state(first_peer_device(device), ns); | |
b8907339 PR |
1314 | } |
1315 | /* No point in queuing send_bitmap if we don't have a connection | |
1316 | * anymore, so check also the _current_ state, not only the new state | |
1317 | * at the time this work was queued. */ | |
1318 | if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S && | |
b30ab791 AG |
1319 | device->state.conn == C_WF_BITMAP_S) |
1320 | drbd_queue_bitmap_io(device, &drbd_send_bitmap, NULL, | |
b8907339 PR |
1321 | "send_bitmap (WFBitMapS)", |
1322 | BM_LOCKED_TEST_ALLOWED); | |
1323 | ||
1324 | /* Lost contact to peer's copy of the data */ | |
1325 | if ((os.pdsk >= D_INCONSISTENT && | |
1326 | os.pdsk != D_UNKNOWN && | |
1327 | os.pdsk != D_OUTDATED) | |
1328 | && (ns.pdsk < D_INCONSISTENT || | |
1329 | ns.pdsk == D_UNKNOWN || | |
1330 | ns.pdsk == D_OUTDATED)) { | |
b30ab791 | 1331 | if (get_ldev(device)) { |
b8907339 | 1332 | if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && |
b30ab791 AG |
1333 | device->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
1334 | if (drbd_suspended(device)) { | |
1335 | set_bit(NEW_CUR_UUID, &device->flags); | |
b8907339 | 1336 | } else { |
b30ab791 | 1337 | drbd_uuid_new_current(device); |
69a22773 | 1338 | drbd_send_uuids(first_peer_device(device)); |
b8907339 PR |
1339 | } |
1340 | } | |
b30ab791 | 1341 | put_ldev(device); |
b8907339 PR |
1342 | } |
1343 | } | |
1344 | ||
b30ab791 | 1345 | if (ns.pdsk < D_INCONSISTENT && get_ldev(device)) { |
0cfac5dd | 1346 | if (os.peer == R_SECONDARY && ns.peer == R_PRIMARY && |
b30ab791 AG |
1347 | device->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
1348 | drbd_uuid_new_current(device); | |
69a22773 | 1349 | drbd_send_uuids(first_peer_device(device)); |
0cfac5dd | 1350 | } |
b8907339 PR |
1351 | /* D_DISKLESS Peer becomes secondary */ |
1352 | if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) | |
1353 | /* We may still be Primary ourselves. | |
1354 | * No harm done if the bitmap still changes, | |
1355 | * redirtied pages will follow later. */ | |
b30ab791 | 1356 | drbd_bitmap_io_from_worker(device, &drbd_bm_write, |
b8907339 | 1357 | "demote diskless peer", BM_LOCKED_SET_ALLOWED); |
b30ab791 | 1358 | put_ldev(device); |
b8907339 PR |
1359 | } |
1360 | ||
1361 | /* Write out all changed bits on demote. | |
1362 | * Though, no need to da that just yet | |
1363 | * if there is a resync going on still */ | |
1364 | if (os.role == R_PRIMARY && ns.role == R_SECONDARY && | |
b30ab791 | 1365 | device->state.conn <= C_CONNECTED && get_ldev(device)) { |
b8907339 PR |
1366 | /* No changes to the bitmap expected this time, so assert that, |
1367 | * even though no harm was done if it did change. */ | |
b30ab791 | 1368 | drbd_bitmap_io_from_worker(device, &drbd_bm_write, |
b8907339 | 1369 | "demote", BM_LOCKED_TEST_ALLOWED); |
b30ab791 | 1370 | put_ldev(device); |
b8907339 PR |
1371 | } |
1372 | ||
1373 | /* Last part of the attaching process ... */ | |
1374 | if (ns.conn >= C_CONNECTED && | |
1375 | os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) { | |
69a22773 AG |
1376 | drbd_send_sizes(first_peer_device(device), 0, 0); /* to start sync... */ |
1377 | drbd_send_uuids(first_peer_device(device)); | |
1378 | drbd_send_state(first_peer_device(device), ns); | |
b8907339 PR |
1379 | } |
1380 | ||
1381 | /* We want to pause/continue resync, tell peer. */ | |
1382 | if (ns.conn >= C_CONNECTED && | |
1383 | ((os.aftr_isp != ns.aftr_isp) || | |
1384 | (os.user_isp != ns.user_isp))) | |
69a22773 | 1385 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1386 | |
1387 | /* In case one of the isp bits got set, suspend other devices. */ | |
1388 | if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) && | |
1389 | (ns.aftr_isp || ns.peer_isp || ns.user_isp)) | |
b30ab791 | 1390 | suspend_other_sg(device); |
b8907339 PR |
1391 | |
1392 | /* Make sure the peer gets informed about eventual state | |
1393 | changes (ISP bits) while we were in WFReportParams. */ | |
1394 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED) | |
69a22773 | 1395 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1396 | |
1397 | if (os.conn != C_AHEAD && ns.conn == C_AHEAD) | |
69a22773 | 1398 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1399 | |
1400 | /* We are in the progress to start a full sync... */ | |
1401 | if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
1402 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S)) | |
1403 | /* no other bitmap changes expected during this phase */ | |
b30ab791 | 1404 | drbd_queue_bitmap_io(device, |
b8907339 PR |
1405 | &drbd_bmio_set_n_write, &abw_start_sync, |
1406 | "set_n_write from StartingSync", BM_LOCKED_TEST_ALLOWED); | |
1407 | ||
b8907339 PR |
1408 | /* first half of local IO error, failure to attach, |
1409 | * or administrative detach */ | |
1410 | if (os.disk != D_FAILED && ns.disk == D_FAILED) { | |
32db80f6 PR |
1411 | enum drbd_io_error_p eh = EP_PASS_ON; |
1412 | int was_io_error = 0; | |
b8907339 | 1413 | /* corresponding get_ldev was in __drbd_set_state, to serialize |
32db80f6 PR |
1414 | * our cleanup here with the transition to D_DISKLESS. |
1415 | * But is is still not save to dreference ldev here, since | |
1416 | * we might come from an failed Attach before ldev was set. */ | |
b30ab791 | 1417 | if (device->ldev) { |
32db80f6 | 1418 | rcu_read_lock(); |
b30ab791 | 1419 | eh = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
32db80f6 PR |
1420 | rcu_read_unlock(); |
1421 | ||
b30ab791 | 1422 | was_io_error = test_and_clear_bit(WAS_IO_ERROR, &device->flags); |
32db80f6 | 1423 | |
6f1a6563 | 1424 | if (was_io_error && eh == EP_CALL_HELPER) |
b30ab791 | 1425 | drbd_khelper(device, "local-io-error"); |
6f1a6563 | 1426 | |
0c849666 LE |
1427 | /* Immediately allow completion of all application IO, |
1428 | * that waits for completion from the local disk, | |
1429 | * if this was a force-detach due to disk_timeout | |
1430 | * or administrator request (drbdsetup detach --force). | |
1431 | * Do NOT abort otherwise. | |
1432 | * Aborting local requests may cause serious problems, | |
1433 | * if requests are completed to upper layers already, | |
1434 | * and then later the already submitted local bio completes. | |
1435 | * This can cause DMA into former bio pages that meanwhile | |
1436 | * have been re-used for other things. | |
1437 | * So aborting local requests may cause crashes, | |
1438 | * or even worse, silent data corruption. | |
1439 | */ | |
b30ab791 AG |
1440 | if (test_and_clear_bit(FORCE_DETACH, &device->flags)) |
1441 | tl_abort_disk_io(device); | |
32db80f6 PR |
1442 | |
1443 | /* current state still has to be D_FAILED, | |
1444 | * there is only one way out: to D_DISKLESS, | |
1445 | * and that may only happen after our put_ldev below. */ | |
b30ab791 | 1446 | if (device->state.disk != D_FAILED) |
d0180171 | 1447 | drbd_err(device, |
32db80f6 | 1448 | "ASSERT FAILED: disk is %s during detach\n", |
b30ab791 | 1449 | drbd_disk_str(device->state.disk)); |
32db80f6 PR |
1450 | |
1451 | if (ns.conn >= C_CONNECTED) | |
69a22773 | 1452 | drbd_send_state(first_peer_device(device), ns); |
32db80f6 | 1453 | |
b30ab791 | 1454 | drbd_rs_cancel_all(device); |
32db80f6 PR |
1455 | |
1456 | /* In case we want to get something to stable storage still, | |
1457 | * this may be the last chance. | |
1458 | * Following put_ldev may transition to D_DISKLESS. */ | |
b30ab791 | 1459 | drbd_md_sync(device); |
32db80f6 | 1460 | } |
b30ab791 | 1461 | put_ldev(device); |
b8907339 PR |
1462 | } |
1463 | ||
b30ab791 AG |
1464 | /* second half of local IO error, failure to attach, |
1465 | * or administrative detach, | |
1466 | * after local_cnt references have reached zero again */ | |
1467 | if (os.disk != D_DISKLESS && ns.disk == D_DISKLESS) { | |
1468 | /* We must still be diskless, | |
1469 | * re-attach has to be serialized with this! */ | |
1470 | if (device->state.disk != D_DISKLESS) | |
d0180171 AG |
1471 | drbd_err(device, |
1472 | "ASSERT FAILED: disk is %s while going diskless\n", | |
1473 | drbd_disk_str(device->state.disk)); | |
b8907339 | 1474 | |
6ab9b1b6 | 1475 | if (ns.conn >= C_CONNECTED) |
69a22773 | 1476 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1477 | /* corresponding get_ldev in __drbd_set_state |
1478 | * this may finally trigger drbd_ldev_destroy. */ | |
b30ab791 | 1479 | put_ldev(device); |
b8907339 PR |
1480 | } |
1481 | ||
1482 | /* Notify peer that I had a local IO error, and did not detached.. */ | |
6ab9b1b6 | 1483 | if (os.disk == D_UP_TO_DATE && ns.disk == D_INCONSISTENT && ns.conn >= C_CONNECTED) |
69a22773 | 1484 | drbd_send_state(first_peer_device(device), ns); |
b8907339 PR |
1485 | |
1486 | /* Disks got bigger while they were detached */ | |
1487 | if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING && | |
b30ab791 | 1488 | test_and_clear_bit(RESYNC_AFTER_NEG, &device->flags)) { |
b8907339 | 1489 | if (ns.conn == C_CONNECTED) |
b30ab791 | 1490 | resync_after_online_grow(device); |
b8907339 PR |
1491 | } |
1492 | ||
1493 | /* A resync finished or aborted, wake paused devices... */ | |
1494 | if ((os.conn > C_CONNECTED && ns.conn <= C_CONNECTED) || | |
1495 | (os.peer_isp && !ns.peer_isp) || | |
1496 | (os.user_isp && !ns.user_isp)) | |
b30ab791 | 1497 | resume_next_sg(device); |
b8907339 PR |
1498 | |
1499 | /* sync target done with resync. Explicitly notify peer, even though | |
1500 | * it should (at least for non-empty resyncs) already know itself. */ | |
1501 | if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED) | |
69a22773 | 1502 | drbd_send_state(first_peer_device(device), ns); |
b8907339 | 1503 | |
58ffa580 LE |
1504 | /* Verify finished, or reached stop sector. Peer did not know about |
1505 | * the stop sector, and we may even have changed the stop sector during | |
1506 | * verify to interrupt/stop early. Send the new state. */ | |
1507 | if (os.conn == C_VERIFY_S && ns.conn == C_CONNECTED | |
b30ab791 | 1508 | && verify_can_do_stop_sector(device)) |
69a22773 | 1509 | drbd_send_state(first_peer_device(device), ns); |
58ffa580 | 1510 | |
b8907339 PR |
1511 | /* This triggers bitmap writeout of potentially still unwritten pages |
1512 | * if the resync finished cleanly, or aborted because of peer disk | |
1513 | * failure, or because of connection loss. | |
1514 | * For resync aborted because of local disk failure, we cannot do | |
1515 | * any bitmap writeout anymore. | |
1516 | * No harm done if some bits change during this phase. | |
1517 | */ | |
b30ab791 AG |
1518 | if (os.conn > C_CONNECTED && ns.conn <= C_CONNECTED && get_ldev(device)) { |
1519 | drbd_queue_bitmap_io(device, &drbd_bm_write_copy_pages, NULL, | |
a220d291 | 1520 | "write from resync_finished", BM_LOCKED_CHANGE_ALLOWED); |
b30ab791 | 1521 | put_ldev(device); |
b8907339 PR |
1522 | } |
1523 | ||
1524 | if (ns.disk == D_DISKLESS && | |
1525 | ns.conn == C_STANDALONE && | |
1526 | ns.role == R_SECONDARY) { | |
1527 | if (os.aftr_isp != ns.aftr_isp) | |
b30ab791 | 1528 | resume_next_sg(device); |
b8907339 PR |
1529 | } |
1530 | ||
b30ab791 | 1531 | drbd_md_sync(device); |
b8907339 PR |
1532 | } |
1533 | ||
bbeb641c | 1534 | struct after_conn_state_chg_work { |
4c007603 | 1535 | struct drbd_work w; |
bbeb641c | 1536 | enum drbd_conns oc; |
8c7e16c3 | 1537 | union drbd_state ns_min; |
b30ab791 | 1538 | union drbd_state ns_max; /* new, max state, over all devices */ |
bbeb641c | 1539 | enum chg_state_flags flags; |
84b8c06b | 1540 | struct drbd_connection *connection; |
bbeb641c PR |
1541 | }; |
1542 | ||
99920dc5 | 1543 | static int w_after_conn_state_ch(struct drbd_work *w, int unused) |
bbeb641c PR |
1544 | { |
1545 | struct after_conn_state_chg_work *acscw = | |
4c007603 | 1546 | container_of(w, struct after_conn_state_chg_work, w); |
84b8c06b | 1547 | struct drbd_connection *connection = acscw->connection; |
bbeb641c | 1548 | enum drbd_conns oc = acscw->oc; |
5f082f98 | 1549 | union drbd_state ns_max = acscw->ns_max; |
c06ece6b | 1550 | struct drbd_peer_device *peer_device; |
a6d00c8e | 1551 | int vnr; |
bbeb641c PR |
1552 | |
1553 | kfree(acscw); | |
1554 | ||
b8907339 | 1555 | /* Upon network configuration, we need to start the receiver */ |
5f082f98 | 1556 | if (oc == C_STANDALONE && ns_max.conn == C_UNCONNECTED) |
bde89a9e | 1557 | drbd_thread_start(&connection->receiver); |
b8907339 | 1558 | |
f3dfa40a LE |
1559 | if (oc == C_DISCONNECTING && ns_max.conn == C_STANDALONE) { |
1560 | struct net_conf *old_conf; | |
1561 | ||
0500813f | 1562 | mutex_lock(&connection->resource->conf_update); |
bde89a9e AG |
1563 | old_conf = connection->net_conf; |
1564 | connection->my_addr_len = 0; | |
1565 | connection->peer_addr_len = 0; | |
1566 | rcu_assign_pointer(connection->net_conf, NULL); | |
1567 | conn_free_crypto(connection); | |
0500813f | 1568 | mutex_unlock(&connection->resource->conf_update); |
f3dfa40a LE |
1569 | |
1570 | synchronize_rcu(); | |
1571 | kfree(old_conf); | |
1572 | } | |
1573 | ||
a6d00c8e PR |
1574 | if (ns_max.susp_fen) { |
1575 | /* case1: The outdate peer handler is successful: */ | |
1576 | if (ns_max.pdsk <= D_OUTDATED) { | |
695d08fa | 1577 | rcu_read_lock(); |
c06ece6b AG |
1578 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1579 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1580 | if (test_bit(NEW_CUR_UUID, &device->flags)) { |
1581 | drbd_uuid_new_current(device); | |
1582 | clear_bit(NEW_CUR_UUID, &device->flags); | |
a6d00c8e PR |
1583 | } |
1584 | } | |
695d08fa | 1585 | rcu_read_unlock(); |
0500813f | 1586 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1587 | _tl_restart(connection, CONNECTION_LOST_WHILE_PENDING); |
1588 | _conn_request_state(connection, | |
8a0bab2a PR |
1589 | (union drbd_state) { { .susp_fen = 1 } }, |
1590 | (union drbd_state) { { .susp_fen = 0 } }, | |
1591 | CS_VERBOSE); | |
0500813f | 1592 | spin_unlock_irq(&connection->resource->req_lock); |
a6d00c8e | 1593 | } |
a6d00c8e | 1594 | } |
05a10ec7 | 1595 | kref_put(&connection->kref, drbd_destroy_connection); |
19fffd7b | 1596 | |
bde89a9e | 1597 | conn_md_sync(connection); |
19fffd7b | 1598 | |
99920dc5 | 1599 | return 0; |
bbeb641c PR |
1600 | } |
1601 | ||
bde89a9e | 1602 | void conn_old_common_state(struct drbd_connection *connection, union drbd_state *pcs, enum chg_state_flags *pf) |
88ef594e | 1603 | { |
435693e8 | 1604 | enum chg_state_flags flags = ~0; |
c06ece6b | 1605 | struct drbd_peer_device *peer_device; |
435693e8 | 1606 | int vnr, first_vol = 1; |
e0e16653 PR |
1607 | union drbd_dev_state os, cs = { |
1608 | { .role = R_SECONDARY, | |
1609 | .peer = R_UNKNOWN, | |
bde89a9e | 1610 | .conn = connection->cstate, |
e0e16653 PR |
1611 | .disk = D_DISKLESS, |
1612 | .pdsk = D_UNKNOWN, | |
1613 | } }; | |
88ef594e | 1614 | |
695d08fa | 1615 | rcu_read_lock(); |
c06ece6b AG |
1616 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1617 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 1618 | os = device->state; |
88ef594e | 1619 | |
435693e8 PR |
1620 | if (first_vol) { |
1621 | cs = os; | |
1622 | first_vol = 0; | |
1623 | continue; | |
1624 | } | |
1625 | ||
1626 | if (cs.role != os.role) | |
1627 | flags &= ~CS_DC_ROLE; | |
1628 | ||
1629 | if (cs.peer != os.peer) | |
1630 | flags &= ~CS_DC_PEER; | |
1631 | ||
1632 | if (cs.conn != os.conn) | |
1633 | flags &= ~CS_DC_CONN; | |
88ef594e | 1634 | |
435693e8 PR |
1635 | if (cs.disk != os.disk) |
1636 | flags &= ~CS_DC_DISK; | |
88ef594e | 1637 | |
435693e8 PR |
1638 | if (cs.pdsk != os.pdsk) |
1639 | flags &= ~CS_DC_PDSK; | |
1640 | } | |
695d08fa | 1641 | rcu_read_unlock(); |
435693e8 PR |
1642 | |
1643 | *pf |= CS_DC_MASK; | |
1644 | *pf &= flags; | |
da9fbc27 | 1645 | (*pcs).i = cs.i; |
88ef594e | 1646 | } |
bbeb641c | 1647 | |
bd0c824a | 1648 | static enum drbd_state_rv |
bde89a9e | 1649 | conn_is_valid_transition(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
88ef594e | 1650 | enum chg_state_flags flags) |
bbeb641c | 1651 | { |
bd0c824a | 1652 | enum drbd_state_rv rv = SS_SUCCESS; |
bbeb641c | 1653 | union drbd_state ns, os; |
c06ece6b | 1654 | struct drbd_peer_device *peer_device; |
bd0c824a | 1655 | int vnr; |
bbeb641c | 1656 | |
695d08fa | 1657 | rcu_read_lock(); |
c06ece6b AG |
1658 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1659 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 1660 | os = drbd_read_state(device); |
d7fe69c6 | 1661 | ns = sanitize_state(device, os, apply_mask_val(os, mask, val), NULL); |
bd0c824a | 1662 | |
778bcf2e PR |
1663 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
1664 | ns.disk = os.disk; | |
1665 | ||
bd0c824a PR |
1666 | if (ns.i == os.i) |
1667 | continue; | |
bbeb641c | 1668 | |
bd0c824a | 1669 | rv = is_valid_transition(os, ns); |
bd0c824a | 1670 | |
c06ece6b | 1671 | if (rv >= SS_SUCCESS && !(flags & CS_HARD)) { |
b30ab791 | 1672 | rv = is_valid_state(device, ns); |
bd0c824a | 1673 | if (rv < SS_SUCCESS) { |
b30ab791 | 1674 | if (is_valid_state(device, os) == rv) |
bde89a9e | 1675 | rv = is_valid_soft_transition(os, ns, connection); |
bd0c824a | 1676 | } else |
bde89a9e | 1677 | rv = is_valid_soft_transition(os, ns, connection); |
bd0c824a | 1678 | } |
c06ece6b AG |
1679 | |
1680 | if (rv < SS_SUCCESS) { | |
1681 | if (flags & CS_VERBOSE) | |
1682 | print_st_err(device, os, ns, rv); | |
bd0c824a | 1683 | break; |
c06ece6b | 1684 | } |
bbeb641c | 1685 | } |
695d08fa | 1686 | rcu_read_unlock(); |
bbeb641c | 1687 | |
bd0c824a | 1688 | return rv; |
bbeb641c PR |
1689 | } |
1690 | ||
8c7e16c3 | 1691 | void |
bde89a9e | 1692 | conn_set_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
8c7e16c3 | 1693 | union drbd_state *pns_min, union drbd_state *pns_max, enum chg_state_flags flags) |
bbeb641c | 1694 | { |
f132f554 | 1695 | union drbd_state ns, os, ns_max = { }; |
8c7e16c3 PR |
1696 | union drbd_state ns_min = { |
1697 | { .role = R_MASK, | |
1698 | .peer = R_MASK, | |
e0e16653 | 1699 | .conn = val.conn, |
8c7e16c3 PR |
1700 | .disk = D_MASK, |
1701 | .pdsk = D_MASK | |
1702 | } }; | |
c06ece6b | 1703 | struct drbd_peer_device *peer_device; |
bbeb641c | 1704 | enum drbd_state_rv rv; |
f132f554 | 1705 | int vnr, number_of_volumes = 0; |
bbeb641c | 1706 | |
07be15b1 LE |
1707 | if (mask.conn == C_MASK) { |
1708 | /* remember last connect time so request_timer_fn() won't | |
1709 | * kill newly established sessions while we are still trying to thaw | |
1710 | * previously frozen IO */ | |
bde89a9e AG |
1711 | if (connection->cstate != C_WF_REPORT_PARAMS && val.conn == C_WF_REPORT_PARAMS) |
1712 | connection->last_reconnect_jif = jiffies; | |
07be15b1 | 1713 | |
bde89a9e | 1714 | connection->cstate = val.conn; |
07be15b1 | 1715 | } |
bd0c824a | 1716 | |
695d08fa | 1717 | rcu_read_lock(); |
c06ece6b AG |
1718 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1719 | struct drbd_device *device = peer_device->device; | |
f132f554 | 1720 | number_of_volumes++; |
b30ab791 | 1721 | os = drbd_read_state(device); |
bd0c824a | 1722 | ns = apply_mask_val(os, mask, val); |
d7fe69c6 | 1723 | ns = sanitize_state(device, os, ns, NULL); |
bbeb641c | 1724 | |
778bcf2e PR |
1725 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
1726 | ns.disk = os.disk; | |
1727 | ||
b30ab791 | 1728 | rv = __drbd_set_state(device, ns, flags, NULL); |
bd0c824a PR |
1729 | if (rv < SS_SUCCESS) |
1730 | BUG(); | |
bbeb641c | 1731 | |
b30ab791 | 1732 | ns.i = device->state.i; |
8c7e16c3 PR |
1733 | ns_max.role = max_role(ns.role, ns_max.role); |
1734 | ns_max.peer = max_role(ns.peer, ns_max.peer); | |
1735 | ns_max.conn = max_t(enum drbd_conns, ns.conn, ns_max.conn); | |
1736 | ns_max.disk = max_t(enum drbd_disk_state, ns.disk, ns_max.disk); | |
1737 | ns_max.pdsk = max_t(enum drbd_disk_state, ns.pdsk, ns_max.pdsk); | |
1738 | ||
1739 | ns_min.role = min_role(ns.role, ns_min.role); | |
1740 | ns_min.peer = min_role(ns.peer, ns_min.peer); | |
1741 | ns_min.conn = min_t(enum drbd_conns, ns.conn, ns_min.conn); | |
1742 | ns_min.disk = min_t(enum drbd_disk_state, ns.disk, ns_min.disk); | |
1743 | ns_min.pdsk = min_t(enum drbd_disk_state, ns.pdsk, ns_min.pdsk); | |
bd0c824a | 1744 | } |
695d08fa | 1745 | rcu_read_unlock(); |
bbeb641c | 1746 | |
f132f554 PR |
1747 | if (number_of_volumes == 0) { |
1748 | ns_min = ns_max = (union drbd_state) { { | |
1749 | .role = R_SECONDARY, | |
1750 | .peer = R_UNKNOWN, | |
1751 | .conn = val.conn, | |
1752 | .disk = D_DISKLESS, | |
1753 | .pdsk = D_UNKNOWN | |
1754 | } }; | |
1755 | } | |
1756 | ||
6bbf53ca AG |
1757 | ns_min.susp = ns_max.susp = connection->resource->susp; |
1758 | ns_min.susp_nod = ns_max.susp_nod = connection->resource->susp_nod; | |
1759 | ns_min.susp_fen = ns_max.susp_fen = connection->resource->susp_fen; | |
8c7e16c3 PR |
1760 | |
1761 | *pns_min = ns_min; | |
1762 | *pns_max = ns_max; | |
bbeb641c PR |
1763 | } |
1764 | ||
df24aa45 | 1765 | static enum drbd_state_rv |
bde89a9e | 1766 | _conn_rq_cond(struct drbd_connection *connection, union drbd_state mask, union drbd_state val) |
df24aa45 | 1767 | { |
88ea685d | 1768 | enum drbd_state_rv err, rv = SS_UNKNOWN_ERROR; /* continue waiting */; |
df24aa45 | 1769 | |
bde89a9e | 1770 | if (test_and_clear_bit(CONN_WD_ST_CHG_OKAY, &connection->flags)) |
88ea685d | 1771 | rv = SS_CW_SUCCESS; |
df24aa45 | 1772 | |
bde89a9e | 1773 | if (test_and_clear_bit(CONN_WD_ST_CHG_FAIL, &connection->flags)) |
88ea685d | 1774 | rv = SS_CW_FAILED_BY_PEER; |
df24aa45 | 1775 | |
88ea685d PR |
1776 | err = conn_is_valid_transition(connection, mask, val, 0); |
1777 | if (err == SS_SUCCESS && connection->cstate == C_WF_REPORT_PARAMS) | |
1778 | return rv; | |
df24aa45 | 1779 | |
88ea685d | 1780 | return err; |
df24aa45 PR |
1781 | } |
1782 | ||
bbeb641c | 1783 | enum drbd_state_rv |
bde89a9e | 1784 | _conn_request_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
bbeb641c PR |
1785 | enum chg_state_flags flags) |
1786 | { | |
1787 | enum drbd_state_rv rv = SS_SUCCESS; | |
bbeb641c | 1788 | struct after_conn_state_chg_work *acscw; |
bde89a9e | 1789 | enum drbd_conns oc = connection->cstate; |
8c7e16c3 | 1790 | union drbd_state ns_max, ns_min, os; |
c02abda2 | 1791 | bool have_mutex = false; |
bbeb641c | 1792 | |
07fc9619 PR |
1793 | if (mask.conn) { |
1794 | rv = is_valid_conn_transition(oc, val.conn); | |
1795 | if (rv < SS_SUCCESS) | |
1796 | goto abort; | |
1797 | } | |
bbeb641c | 1798 | |
bde89a9e | 1799 | rv = conn_is_valid_transition(connection, mask, val, flags); |
bbeb641c PR |
1800 | if (rv < SS_SUCCESS) |
1801 | goto abort; | |
1802 | ||
df24aa45 PR |
1803 | if (oc == C_WF_REPORT_PARAMS && val.conn == C_DISCONNECTING && |
1804 | !(flags & (CS_LOCAL_ONLY | CS_HARD))) { | |
c02abda2 LE |
1805 | |
1806 | /* This will be a cluster-wide state change. | |
1807 | * Need to give up the spinlock, grab the mutex, | |
1808 | * then send the state change request, ... */ | |
0500813f | 1809 | spin_unlock_irq(&connection->resource->req_lock); |
bde89a9e | 1810 | mutex_lock(&connection->cstate_mutex); |
c02abda2 LE |
1811 | have_mutex = true; |
1812 | ||
bde89a9e AG |
1813 | set_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
1814 | if (conn_send_state_req(connection, mask, val)) { | |
c02abda2 | 1815 | /* sending failed. */ |
bde89a9e | 1816 | clear_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
c02abda2 LE |
1817 | rv = SS_CW_FAILED_BY_PEER; |
1818 | /* need to re-aquire the spin lock, though */ | |
1819 | goto abort_unlocked; | |
1820 | } | |
1821 | ||
1822 | if (val.conn == C_DISCONNECTING) | |
bde89a9e | 1823 | set_bit(DISCONNECT_SENT, &connection->flags); |
c02abda2 LE |
1824 | |
1825 | /* ... and re-aquire the spinlock. | |
1826 | * If _conn_rq_cond() returned >= SS_SUCCESS, we must call | |
1827 | * conn_set_state() within the same spinlock. */ | |
0500813f | 1828 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1829 | wait_event_lock_irq(connection->ping_wait, |
1830 | (rv = _conn_rq_cond(connection, mask, val)), | |
0500813f | 1831 | connection->resource->req_lock); |
bde89a9e | 1832 | clear_bit(CONN_WD_ST_CHG_REQ, &connection->flags); |
df24aa45 PR |
1833 | if (rv < SS_SUCCESS) |
1834 | goto abort; | |
1835 | } | |
1836 | ||
bde89a9e | 1837 | conn_old_common_state(connection, &os, &flags); |
706cb24c | 1838 | flags |= CS_DC_SUSP; |
bde89a9e AG |
1839 | conn_set_state(connection, mask, val, &ns_min, &ns_max, flags); |
1840 | conn_pr_state_change(connection, os, ns_max, flags); | |
bbeb641c PR |
1841 | |
1842 | acscw = kmalloc(sizeof(*acscw), GFP_ATOMIC); | |
1843 | if (acscw) { | |
435693e8 | 1844 | acscw->oc = os.conn; |
8c7e16c3 | 1845 | acscw->ns_min = ns_min; |
5f082f98 | 1846 | acscw->ns_max = ns_max; |
bbeb641c | 1847 | acscw->flags = flags; |
4c007603 | 1848 | acscw->w.cb = w_after_conn_state_ch; |
bde89a9e | 1849 | kref_get(&connection->kref); |
84b8c06b | 1850 | acscw->connection = connection; |
4c007603 | 1851 | drbd_queue_work(&connection->sender_work, &acscw->w); |
bbeb641c | 1852 | } else { |
1ec861eb | 1853 | drbd_err(connection, "Could not kmalloc an acscw\n"); |
b8907339 | 1854 | } |
bbeb641c | 1855 | |
a01842eb | 1856 | abort: |
c02abda2 LE |
1857 | if (have_mutex) { |
1858 | /* mutex_unlock() "... must not be used in interrupt context.", | |
1859 | * so give up the spinlock, then re-aquire it */ | |
0500813f | 1860 | spin_unlock_irq(&connection->resource->req_lock); |
c02abda2 | 1861 | abort_unlocked: |
bde89a9e | 1862 | mutex_unlock(&connection->cstate_mutex); |
0500813f | 1863 | spin_lock_irq(&connection->resource->req_lock); |
c02abda2 LE |
1864 | } |
1865 | if (rv < SS_SUCCESS && flags & CS_VERBOSE) { | |
1ec861eb AG |
1866 | drbd_err(connection, "State change failed: %s\n", drbd_set_st_err_str(rv)); |
1867 | drbd_err(connection, " mask = 0x%x val = 0x%x\n", mask.i, val.i); | |
1868 | drbd_err(connection, " old_conn:%s wanted_conn:%s\n", drbd_conn_str(oc), drbd_conn_str(val.conn)); | |
a01842eb | 1869 | } |
bbeb641c PR |
1870 | return rv; |
1871 | } | |
1872 | ||
1873 | enum drbd_state_rv | |
bde89a9e | 1874 | conn_request_state(struct drbd_connection *connection, union drbd_state mask, union drbd_state val, |
bbeb641c PR |
1875 | enum chg_state_flags flags) |
1876 | { | |
1877 | enum drbd_state_rv rv; | |
1878 | ||
0500813f | 1879 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 1880 | rv = _conn_request_state(connection, mask, val, flags); |
0500813f | 1881 | spin_unlock_irq(&connection->resource->req_lock); |
bbeb641c PR |
1882 | |
1883 | return rv; | |
b8907339 | 1884 | } |