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" | |
30 | #include "drbd_req.h" | |
31 | ||
cdfda633 | 32 | /* in drbd_main.c */ |
71fc7eed | 33 | extern void tl_abort_disk_io(struct drbd_conf *mdev); |
cdfda633 | 34 | |
b8907339 PR |
35 | struct after_state_chg_work { |
36 | struct drbd_work w; | |
37 | union drbd_state os; | |
38 | union drbd_state ns; | |
39 | enum chg_state_flags flags; | |
40 | struct completion *done; | |
41 | }; | |
42 | ||
d942ae44 PR |
43 | enum sanitize_state_warnings { |
44 | NO_WARNING, | |
45 | ABORTED_ONLINE_VERIFY, | |
46 | ABORTED_RESYNC, | |
47 | CONNECTION_LOST_NEGOTIATING, | |
48 | IMPLICITLY_UPGRADED_DISK, | |
49 | IMPLICITLY_UPGRADED_PDSK, | |
50 | }; | |
51 | ||
99920dc5 | 52 | static int w_after_state_ch(struct drbd_work *w, int unused); |
b8907339 PR |
53 | static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, |
54 | union drbd_state ns, enum chg_state_flags flags); | |
a75f34ad PR |
55 | static enum drbd_state_rv is_valid_state(struct drbd_conf *, union drbd_state); |
56 | static enum drbd_state_rv is_valid_soft_transition(union drbd_state, union drbd_state); | |
3509502d | 57 | static enum drbd_state_rv is_valid_transition(union drbd_state os, union drbd_state ns); |
4308a0a3 | 58 | static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state ns, |
d942ae44 | 59 | enum sanitize_state_warnings *warn); |
b8907339 | 60 | |
2aebfabb PR |
61 | static inline bool is_susp(union drbd_state s) |
62 | { | |
63 | return s.susp || s.susp_nod || s.susp_fen; | |
64 | } | |
65 | ||
d0456c72 | 66 | bool conn_all_vols_unconf(struct drbd_tconn *tconn) |
0e29d163 PR |
67 | { |
68 | struct drbd_conf *mdev; | |
695d08fa | 69 | bool rv = true; |
e90285e0 | 70 | int vnr; |
0e29d163 | 71 | |
695d08fa | 72 | rcu_read_lock(); |
e90285e0 | 73 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
d0456c72 LE |
74 | if (mdev->state.disk != D_DISKLESS || |
75 | mdev->state.conn != C_STANDALONE || | |
695d08fa PR |
76 | mdev->state.role != R_SECONDARY) { |
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 | ||
105 | enum drbd_role conn_highest_role(struct drbd_tconn *tconn) | |
106 | { | |
107 | enum drbd_role role = R_UNKNOWN; | |
108 | struct drbd_conf *mdev; | |
109 | int vnr; | |
110 | ||
695d08fa | 111 | rcu_read_lock(); |
cb703454 PR |
112 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
113 | role = max_role(role, mdev->state.role); | |
695d08fa | 114 | rcu_read_unlock(); |
cb703454 PR |
115 | |
116 | return role; | |
117 | } | |
118 | ||
119 | enum drbd_role conn_highest_peer(struct drbd_tconn *tconn) | |
120 | { | |
121 | enum drbd_role peer = R_UNKNOWN; | |
122 | struct drbd_conf *mdev; | |
123 | int vnr; | |
124 | ||
695d08fa | 125 | rcu_read_lock(); |
cb703454 PR |
126 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
127 | peer = max_role(peer, mdev->state.peer); | |
695d08fa | 128 | rcu_read_unlock(); |
cb703454 PR |
129 | |
130 | return peer; | |
131 | } | |
132 | ||
133 | enum drbd_disk_state conn_highest_disk(struct drbd_tconn *tconn) | |
134 | { | |
135 | enum drbd_disk_state ds = D_DISKLESS; | |
136 | struct drbd_conf *mdev; | |
137 | int vnr; | |
138 | ||
695d08fa | 139 | rcu_read_lock(); |
cb703454 PR |
140 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
141 | ds = max_t(enum drbd_disk_state, ds, mdev->state.disk); | |
695d08fa | 142 | rcu_read_unlock(); |
cb703454 PR |
143 | |
144 | return ds; | |
145 | } | |
146 | ||
4669265a PR |
147 | enum drbd_disk_state conn_lowest_disk(struct drbd_tconn *tconn) |
148 | { | |
149 | enum drbd_disk_state ds = D_MASK; | |
150 | struct drbd_conf *mdev; | |
151 | int vnr; | |
152 | ||
695d08fa | 153 | rcu_read_lock(); |
4669265a PR |
154 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
155 | ds = min_t(enum drbd_disk_state, ds, mdev->state.disk); | |
695d08fa | 156 | rcu_read_unlock(); |
4669265a PR |
157 | |
158 | return ds; | |
159 | } | |
160 | ||
cb703454 PR |
161 | enum drbd_disk_state conn_highest_pdsk(struct drbd_tconn *tconn) |
162 | { | |
163 | enum drbd_disk_state ds = D_DISKLESS; | |
164 | struct drbd_conf *mdev; | |
165 | int vnr; | |
166 | ||
695d08fa | 167 | rcu_read_lock(); |
cb703454 PR |
168 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
169 | ds = max_t(enum drbd_disk_state, ds, mdev->state.pdsk); | |
695d08fa | 170 | rcu_read_unlock(); |
cb703454 PR |
171 | |
172 | return ds; | |
173 | } | |
174 | ||
19f83c76 PR |
175 | enum drbd_conns conn_lowest_conn(struct drbd_tconn *tconn) |
176 | { | |
177 | enum drbd_conns conn = C_MASK; | |
178 | struct drbd_conf *mdev; | |
179 | int vnr; | |
180 | ||
695d08fa | 181 | rcu_read_lock(); |
19f83c76 PR |
182 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
183 | conn = min_t(enum drbd_conns, conn, mdev->state.conn); | |
695d08fa | 184 | rcu_read_unlock(); |
19f83c76 PR |
185 | |
186 | return conn; | |
187 | } | |
188 | ||
b8907339 PR |
189 | /** |
190 | * cl_wide_st_chg() - true if the state change is a cluster wide one | |
191 | * @mdev: DRBD device. | |
192 | * @os: old (current) state. | |
193 | * @ns: new (wanted) state. | |
194 | */ | |
195 | static int cl_wide_st_chg(struct drbd_conf *mdev, | |
196 | union drbd_state os, union drbd_state ns) | |
197 | { | |
198 | return (os.conn >= C_CONNECTED && ns.conn >= C_CONNECTED && | |
199 | ((os.role != R_PRIMARY && ns.role == R_PRIMARY) || | |
200 | (os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
201 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) || | |
202 | (os.disk != D_DISKLESS && ns.disk == D_DISKLESS))) || | |
203 | (os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) || | |
369bea63 PR |
204 | (os.conn == C_CONNECTED && ns.conn == C_VERIFY_S) || |
205 | (os.conn == C_CONNECTED && ns.conn == C_WF_REPORT_PARAMS); | |
b8907339 PR |
206 | } |
207 | ||
56707f9e PR |
208 | static union drbd_state |
209 | apply_mask_val(union drbd_state os, union drbd_state mask, union drbd_state val) | |
210 | { | |
211 | union drbd_state ns; | |
212 | ns.i = (os.i & ~mask.i) | val.i; | |
213 | return ns; | |
214 | } | |
215 | ||
b8907339 PR |
216 | enum drbd_state_rv |
217 | drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f, | |
218 | union drbd_state mask, union drbd_state val) | |
219 | { | |
220 | unsigned long flags; | |
56707f9e | 221 | union drbd_state ns; |
b8907339 PR |
222 | enum drbd_state_rv rv; |
223 | ||
224 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); | |
78bae59b | 225 | ns = apply_mask_val(drbd_read_state(mdev), mask, val); |
b8907339 | 226 | rv = _drbd_set_state(mdev, ns, f, NULL); |
b8907339 PR |
227 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); |
228 | ||
229 | return rv; | |
230 | } | |
231 | ||
232 | /** | |
233 | * drbd_force_state() - Impose a change which happens outside our control on our state | |
234 | * @mdev: DRBD device. | |
235 | * @mask: mask of state bits to change. | |
236 | * @val: value of new state bits. | |
237 | */ | |
238 | void drbd_force_state(struct drbd_conf *mdev, | |
239 | union drbd_state mask, union drbd_state val) | |
240 | { | |
241 | drbd_change_state(mdev, CS_HARD, mask, val); | |
242 | } | |
243 | ||
244 | static enum drbd_state_rv | |
245 | _req_st_cond(struct drbd_conf *mdev, union drbd_state mask, | |
246 | union drbd_state val) | |
247 | { | |
248 | union drbd_state os, ns; | |
249 | unsigned long flags; | |
250 | enum drbd_state_rv rv; | |
251 | ||
252 | if (test_and_clear_bit(CL_ST_CHG_SUCCESS, &mdev->flags)) | |
253 | return SS_CW_SUCCESS; | |
254 | ||
255 | if (test_and_clear_bit(CL_ST_CHG_FAIL, &mdev->flags)) | |
256 | return SS_CW_FAILED_BY_PEER; | |
257 | ||
b8907339 | 258 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); |
78bae59b | 259 | os = drbd_read_state(mdev); |
56707f9e | 260 | ns = sanitize_state(mdev, apply_mask_val(os, mask, val), NULL); |
3509502d PR |
261 | rv = is_valid_transition(os, ns); |
262 | if (rv == SS_SUCCESS) | |
263 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ | |
b8907339 PR |
264 | |
265 | if (!cl_wide_st_chg(mdev, os, ns)) | |
266 | rv = SS_CW_NO_NEED; | |
3509502d | 267 | if (rv == SS_UNKNOWN_ERROR) { |
b8907339 PR |
268 | rv = is_valid_state(mdev, ns); |
269 | if (rv == SS_SUCCESS) { | |
a75f34ad | 270 | rv = is_valid_soft_transition(os, ns); |
b8907339 PR |
271 | if (rv == SS_SUCCESS) |
272 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ | |
273 | } | |
274 | } | |
275 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
276 | ||
277 | return rv; | |
278 | } | |
279 | ||
280 | /** | |
281 | * drbd_req_state() - Perform an eventually cluster wide state change | |
282 | * @mdev: DRBD device. | |
283 | * @mask: mask of state bits to change. | |
284 | * @val: value of new state bits. | |
285 | * @f: flags | |
286 | * | |
287 | * Should not be called directly, use drbd_request_state() or | |
288 | * _drbd_request_state(). | |
289 | */ | |
290 | static enum drbd_state_rv | |
291 | drbd_req_state(struct drbd_conf *mdev, union drbd_state mask, | |
292 | union drbd_state val, enum chg_state_flags f) | |
293 | { | |
294 | struct completion done; | |
295 | unsigned long flags; | |
296 | union drbd_state os, ns; | |
297 | enum drbd_state_rv rv; | |
298 | ||
299 | init_completion(&done); | |
300 | ||
301 | if (f & CS_SERIALIZE) | |
8410da8f | 302 | mutex_lock(mdev->state_mutex); |
b8907339 PR |
303 | |
304 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); | |
78bae59b | 305 | os = drbd_read_state(mdev); |
56707f9e | 306 | ns = sanitize_state(mdev, apply_mask_val(os, mask, val), NULL); |
3509502d | 307 | rv = is_valid_transition(os, ns); |
3c5e5f6a LE |
308 | if (rv < SS_SUCCESS) { |
309 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
3509502d | 310 | goto abort; |
3c5e5f6a | 311 | } |
b8907339 PR |
312 | |
313 | if (cl_wide_st_chg(mdev, os, ns)) { | |
314 | rv = is_valid_state(mdev, ns); | |
315 | if (rv == SS_SUCCESS) | |
a75f34ad | 316 | rv = is_valid_soft_transition(os, ns); |
b8907339 PR |
317 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); |
318 | ||
319 | if (rv < SS_SUCCESS) { | |
320 | if (f & CS_VERBOSE) | |
321 | print_st_err(mdev, os, ns, rv); | |
322 | goto abort; | |
323 | } | |
324 | ||
d24ae219 | 325 | if (drbd_send_state_req(mdev, mask, val)) { |
b8907339 PR |
326 | rv = SS_CW_FAILED_BY_PEER; |
327 | if (f & CS_VERBOSE) | |
328 | print_st_err(mdev, os, ns, rv); | |
329 | goto abort; | |
330 | } | |
331 | ||
332 | wait_event(mdev->state_wait, | |
333 | (rv = _req_st_cond(mdev, mask, val))); | |
334 | ||
335 | if (rv < SS_SUCCESS) { | |
b8907339 PR |
336 | if (f & CS_VERBOSE) |
337 | print_st_err(mdev, os, ns, rv); | |
338 | goto abort; | |
339 | } | |
340 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); | |
78bae59b | 341 | ns = apply_mask_val(drbd_read_state(mdev), mask, val); |
b8907339 | 342 | rv = _drbd_set_state(mdev, ns, f, &done); |
b8907339 PR |
343 | } else { |
344 | rv = _drbd_set_state(mdev, ns, f, &done); | |
345 | } | |
346 | ||
347 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
348 | ||
349 | if (f & CS_WAIT_COMPLETE && rv == SS_SUCCESS) { | |
350 | D_ASSERT(current != mdev->tconn->worker.task); | |
351 | wait_for_completion(&done); | |
352 | } | |
353 | ||
354 | abort: | |
355 | if (f & CS_SERIALIZE) | |
8410da8f | 356 | mutex_unlock(mdev->state_mutex); |
b8907339 PR |
357 | |
358 | return rv; | |
359 | } | |
360 | ||
361 | /** | |
362 | * _drbd_request_state() - Request a state change (with flags) | |
363 | * @mdev: DRBD device. | |
364 | * @mask: mask of state bits to change. | |
365 | * @val: value of new state bits. | |
366 | * @f: flags | |
367 | * | |
368 | * Cousin of drbd_request_state(), useful with the CS_WAIT_COMPLETE | |
369 | * flag, or when logging of failed state change requests is not desired. | |
370 | */ | |
371 | enum drbd_state_rv | |
372 | _drbd_request_state(struct drbd_conf *mdev, union drbd_state mask, | |
373 | union drbd_state val, enum chg_state_flags f) | |
374 | { | |
375 | enum drbd_state_rv rv; | |
376 | ||
377 | wait_event(mdev->state_wait, | |
378 | (rv = drbd_req_state(mdev, mask, val, f)) != SS_IN_TRANSIENT_STATE); | |
379 | ||
380 | return rv; | |
381 | } | |
382 | ||
383 | static void print_st(struct drbd_conf *mdev, char *name, union drbd_state ns) | |
384 | { | |
385 | dev_err(DEV, " %s = { cs:%s ro:%s/%s ds:%s/%s %c%c%c%c%c%c }\n", | |
386 | name, | |
387 | drbd_conn_str(ns.conn), | |
388 | drbd_role_str(ns.role), | |
389 | drbd_role_str(ns.peer), | |
390 | drbd_disk_str(ns.disk), | |
391 | drbd_disk_str(ns.pdsk), | |
392 | is_susp(ns) ? 's' : 'r', | |
393 | ns.aftr_isp ? 'a' : '-', | |
394 | ns.peer_isp ? 'p' : '-', | |
395 | ns.user_isp ? 'u' : '-', | |
396 | ns.susp_fen ? 'F' : '-', | |
397 | ns.susp_nod ? 'N' : '-' | |
398 | ); | |
399 | } | |
400 | ||
401 | void print_st_err(struct drbd_conf *mdev, union drbd_state os, | |
402 | union drbd_state ns, enum drbd_state_rv err) | |
403 | { | |
404 | if (err == SS_IN_TRANSIENT_STATE) | |
405 | return; | |
406 | dev_err(DEV, "State change failed: %s\n", drbd_set_st_err_str(err)); | |
407 | print_st(mdev, " state", os); | |
408 | print_st(mdev, "wanted", ns); | |
409 | } | |
410 | ||
435693e8 | 411 | static long print_state_change(char *pb, union drbd_state os, union drbd_state ns, |
bbeb641c PR |
412 | enum chg_state_flags flags) |
413 | { | |
435693e8 | 414 | char *pbp; |
bbeb641c PR |
415 | pbp = pb; |
416 | *pbp = 0; | |
706cb24c | 417 | |
435693e8 | 418 | if (ns.role != os.role && flags & CS_DC_ROLE) |
bbeb641c PR |
419 | pbp += sprintf(pbp, "role( %s -> %s ) ", |
420 | drbd_role_str(os.role), | |
421 | drbd_role_str(ns.role)); | |
435693e8 | 422 | if (ns.peer != os.peer && flags & CS_DC_PEER) |
bbeb641c PR |
423 | pbp += sprintf(pbp, "peer( %s -> %s ) ", |
424 | drbd_role_str(os.peer), | |
425 | drbd_role_str(ns.peer)); | |
435693e8 | 426 | if (ns.conn != os.conn && flags & CS_DC_CONN) |
bbeb641c PR |
427 | pbp += sprintf(pbp, "conn( %s -> %s ) ", |
428 | drbd_conn_str(os.conn), | |
429 | drbd_conn_str(ns.conn)); | |
435693e8 | 430 | if (ns.disk != os.disk && flags & CS_DC_DISK) |
bbeb641c PR |
431 | pbp += sprintf(pbp, "disk( %s -> %s ) ", |
432 | drbd_disk_str(os.disk), | |
433 | drbd_disk_str(ns.disk)); | |
435693e8 | 434 | if (ns.pdsk != os.pdsk && flags & CS_DC_PDSK) |
bbeb641c PR |
435 | pbp += sprintf(pbp, "pdsk( %s -> %s ) ", |
436 | drbd_disk_str(os.pdsk), | |
437 | drbd_disk_str(ns.pdsk)); | |
706cb24c PR |
438 | |
439 | return pbp - pb; | |
440 | } | |
441 | ||
442 | static void drbd_pr_state_change(struct drbd_conf *mdev, union drbd_state os, union drbd_state ns, | |
443 | enum chg_state_flags flags) | |
444 | { | |
445 | char pb[300]; | |
446 | char *pbp = pb; | |
447 | ||
448 | pbp += print_state_change(pbp, os, ns, flags ^ CS_DC_MASK); | |
449 | ||
bbeb641c PR |
450 | if (ns.aftr_isp != os.aftr_isp) |
451 | pbp += sprintf(pbp, "aftr_isp( %d -> %d ) ", | |
452 | os.aftr_isp, | |
453 | ns.aftr_isp); | |
454 | if (ns.peer_isp != os.peer_isp) | |
455 | pbp += sprintf(pbp, "peer_isp( %d -> %d ) ", | |
456 | os.peer_isp, | |
457 | ns.peer_isp); | |
458 | if (ns.user_isp != os.user_isp) | |
459 | pbp += sprintf(pbp, "user_isp( %d -> %d ) ", | |
460 | os.user_isp, | |
461 | ns.user_isp); | |
435693e8 | 462 | |
706cb24c | 463 | if (pbp != pb) |
bbeb641c PR |
464 | dev_info(DEV, "%s\n", pb); |
465 | } | |
b8907339 | 466 | |
435693e8 PR |
467 | static void conn_pr_state_change(struct drbd_tconn *tconn, union drbd_state os, union drbd_state ns, |
468 | enum chg_state_flags flags) | |
469 | { | |
470 | char pb[300]; | |
706cb24c PR |
471 | char *pbp = pb; |
472 | ||
473 | pbp += print_state_change(pbp, os, ns, flags); | |
435693e8 | 474 | |
706cb24c PR |
475 | if (is_susp(ns) != is_susp(os) && flags & CS_DC_SUSP) |
476 | pbp += sprintf(pbp, "susp( %d -> %d ) ", | |
477 | is_susp(os), | |
478 | is_susp(ns)); | |
479 | ||
480 | if (pbp != pb) | |
435693e8 PR |
481 | conn_info(tconn, "%s\n", pb); |
482 | } | |
483 | ||
484 | ||
b8907339 PR |
485 | /** |
486 | * is_valid_state() - Returns an SS_ error code if ns is not valid | |
487 | * @mdev: DRBD device. | |
488 | * @ns: State to consider. | |
489 | */ | |
490 | static enum drbd_state_rv | |
491 | is_valid_state(struct drbd_conf *mdev, union drbd_state ns) | |
492 | { | |
493 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
494 | ||
495 | enum drbd_fencing_p fp; | |
496 | enum drbd_state_rv rv = SS_SUCCESS; | |
44ed167d | 497 | struct net_conf *nc; |
b8907339 | 498 | |
daeda1cc | 499 | rcu_read_lock(); |
b8907339 PR |
500 | fp = FP_DONT_CARE; |
501 | if (get_ldev(mdev)) { | |
daeda1cc | 502 | fp = rcu_dereference(mdev->ldev->disk_conf)->fencing; |
b8907339 PR |
503 | put_ldev(mdev); |
504 | } | |
505 | ||
44ed167d PR |
506 | nc = rcu_dereference(mdev->tconn->net_conf); |
507 | if (nc) { | |
508 | if (!nc->two_primaries && ns.role == R_PRIMARY) { | |
047e95e2 PR |
509 | if (ns.peer == R_PRIMARY) |
510 | rv = SS_TWO_PRIMARIES; | |
cb703454 | 511 | else if (conn_highest_peer(mdev->tconn) == R_PRIMARY) |
047e95e2 | 512 | rv = SS_O_VOL_PEER_PRI; |
44ed167d | 513 | } |
b8907339 PR |
514 | } |
515 | ||
516 | if (rv <= 0) | |
517 | /* already found a reason to abort */; | |
518 | else if (ns.role == R_SECONDARY && mdev->open_cnt) | |
519 | rv = SS_DEVICE_IN_USE; | |
520 | ||
521 | else if (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.disk < D_UP_TO_DATE) | |
522 | rv = SS_NO_UP_TO_DATE_DISK; | |
523 | ||
524 | else if (fp >= FP_RESOURCE && | |
525 | ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk >= D_UNKNOWN) | |
526 | rv = SS_PRIMARY_NOP; | |
527 | ||
528 | else if (ns.role == R_PRIMARY && ns.disk <= D_INCONSISTENT && ns.pdsk <= D_INCONSISTENT) | |
529 | rv = SS_NO_UP_TO_DATE_DISK; | |
530 | ||
531 | else if (ns.conn > C_CONNECTED && ns.disk < D_INCONSISTENT) | |
532 | rv = SS_NO_LOCAL_DISK; | |
533 | ||
534 | else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT) | |
535 | rv = SS_NO_REMOTE_DISK; | |
536 | ||
537 | else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) | |
538 | rv = SS_NO_UP_TO_DATE_DISK; | |
539 | ||
540 | else if ((ns.conn == C_CONNECTED || | |
541 | ns.conn == C_WF_BITMAP_S || | |
542 | ns.conn == C_SYNC_SOURCE || | |
543 | ns.conn == C_PAUSED_SYNC_S) && | |
544 | ns.disk == D_OUTDATED) | |
545 | rv = SS_CONNECTED_OUTDATES; | |
546 | ||
547 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
44ed167d | 548 | (nc->verify_alg[0] == 0)) |
b8907339 PR |
549 | rv = SS_NO_VERIFY_ALG; |
550 | ||
551 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
552 | mdev->tconn->agreed_pro_version < 88) | |
553 | rv = SS_NOT_SUPPORTED; | |
554 | ||
555 | else if (ns.conn >= C_CONNECTED && ns.pdsk == D_UNKNOWN) | |
556 | rv = SS_CONNECTED_OUTDATES; | |
557 | ||
44ed167d PR |
558 | rcu_read_unlock(); |
559 | ||
b8907339 PR |
560 | return rv; |
561 | } | |
562 | ||
563 | /** | |
a75f34ad | 564 | * is_valid_soft_transition() - Returns an SS_ error code if the state transition is not possible |
3509502d PR |
565 | * This function limits state transitions that may be declined by DRBD. I.e. |
566 | * user requests (aka soft transitions). | |
b8907339 PR |
567 | * @mdev: DRBD device. |
568 | * @ns: new state. | |
569 | * @os: old state. | |
570 | */ | |
571 | static enum drbd_state_rv | |
a75f34ad | 572 | is_valid_soft_transition(union drbd_state os, union drbd_state ns) |
b8907339 PR |
573 | { |
574 | enum drbd_state_rv rv = SS_SUCCESS; | |
575 | ||
576 | if ((ns.conn == C_STARTING_SYNC_T || ns.conn == C_STARTING_SYNC_S) && | |
577 | os.conn > C_CONNECTED) | |
578 | rv = SS_RESYNC_RUNNING; | |
579 | ||
580 | if (ns.conn == C_DISCONNECTING && os.conn == C_STANDALONE) | |
581 | rv = SS_ALREADY_STANDALONE; | |
582 | ||
583 | if (ns.disk > D_ATTACHING && os.disk == D_DISKLESS) | |
584 | rv = SS_IS_DISKLESS; | |
585 | ||
586 | if (ns.conn == C_WF_CONNECTION && os.conn < C_UNCONNECTED) | |
587 | rv = SS_NO_NET_CONFIG; | |
588 | ||
589 | if (ns.disk == D_OUTDATED && os.disk < D_OUTDATED && os.disk != D_ATTACHING) | |
590 | rv = SS_LOWER_THAN_OUTDATED; | |
591 | ||
592 | if (ns.conn == C_DISCONNECTING && os.conn == C_UNCONNECTED) | |
593 | rv = SS_IN_TRANSIENT_STATE; | |
594 | ||
2325eb66 PR |
595 | /* if (ns.conn == os.conn && ns.conn == C_WF_REPORT_PARAMS) |
596 | rv = SS_IN_TRANSIENT_STATE; */ | |
b8907339 PR |
597 | |
598 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED) | |
599 | rv = SS_NEED_CONNECTION; | |
600 | ||
601 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
602 | ns.conn != os.conn && os.conn > C_CONNECTED) | |
603 | rv = SS_RESYNC_RUNNING; | |
604 | ||
605 | if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
606 | os.conn < C_CONNECTED) | |
607 | rv = SS_NEED_CONNECTION; | |
608 | ||
609 | if ((ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE) | |
610 | && os.conn < C_WF_REPORT_PARAMS) | |
611 | rv = SS_NEED_CONNECTION; /* No NetworkFailure -> SyncTarget etc... */ | |
612 | ||
613 | return rv; | |
614 | } | |
615 | ||
3509502d | 616 | static enum drbd_state_rv |
fda74117 | 617 | is_valid_conn_transition(enum drbd_conns oc, enum drbd_conns nc) |
3509502d | 618 | { |
d9cc6e23 LE |
619 | /* no change -> nothing to do, at least for the connection part */ |
620 | if (oc == nc) | |
621 | return SS_NOTHING_TO_DO; | |
3509502d | 622 | |
d9cc6e23 LE |
623 | /* disconnect of an unconfigured connection does not make sense */ |
624 | if (oc == C_STANDALONE && nc == C_DISCONNECTING) | |
625 | return SS_ALREADY_STANDALONE; | |
626 | ||
627 | /* from C_STANDALONE, we start with C_UNCONNECTED */ | |
628 | if (oc == C_STANDALONE && nc != C_UNCONNECTED) | |
629 | return SS_NEED_CONNECTION; | |
3509502d | 630 | |
25b0d6c8 PR |
631 | /* When establishing a connection we need to go through WF_REPORT_PARAMS! |
632 | Necessary to do the right thing upon invalidate-remote on a disconnected resource */ | |
633 | if (oc < C_WF_REPORT_PARAMS && nc >= C_CONNECTED) | |
634 | return SS_NEED_CONNECTION; | |
635 | ||
3509502d | 636 | /* After a network error only C_UNCONNECTED or C_DISCONNECTING may follow. */ |
fda74117 | 637 | if (oc >= C_TIMEOUT && oc <= C_TEAR_DOWN && nc != C_UNCONNECTED && nc != C_DISCONNECTING) |
d9cc6e23 | 638 | return SS_IN_TRANSIENT_STATE; |
3509502d PR |
639 | |
640 | /* After C_DISCONNECTING only C_STANDALONE may follow */ | |
fda74117 | 641 | if (oc == C_DISCONNECTING && nc != C_STANDALONE) |
d9cc6e23 | 642 | return SS_IN_TRANSIENT_STATE; |
3509502d | 643 | |
d9cc6e23 | 644 | return SS_SUCCESS; |
fda74117 PR |
645 | } |
646 | ||
647 | ||
648 | /** | |
649 | * is_valid_transition() - Returns an SS_ error code if the state transition is not possible | |
650 | * This limits hard state transitions. Hard state transitions are facts there are | |
651 | * imposed on DRBD by the environment. E.g. disk broke or network broke down. | |
652 | * But those hard state transitions are still not allowed to do everything. | |
653 | * @ns: new state. | |
654 | * @os: old state. | |
655 | */ | |
656 | static enum drbd_state_rv | |
657 | is_valid_transition(union drbd_state os, union drbd_state ns) | |
658 | { | |
659 | enum drbd_state_rv rv; | |
660 | ||
661 | rv = is_valid_conn_transition(os.conn, ns.conn); | |
662 | ||
3509502d PR |
663 | /* we cannot fail (again) if we already detached */ |
664 | if (ns.disk == D_FAILED && os.disk == D_DISKLESS) | |
665 | rv = SS_IS_DISKLESS; | |
666 | ||
667 | return rv; | |
668 | } | |
669 | ||
d942ae44 PR |
670 | static void print_sanitize_warnings(struct drbd_conf *mdev, enum sanitize_state_warnings warn) |
671 | { | |
672 | static const char *msg_table[] = { | |
673 | [NO_WARNING] = "", | |
674 | [ABORTED_ONLINE_VERIFY] = "Online-verify aborted.", | |
675 | [ABORTED_RESYNC] = "Resync aborted.", | |
676 | [CONNECTION_LOST_NEGOTIATING] = "Connection lost while negotiating, no data!", | |
677 | [IMPLICITLY_UPGRADED_DISK] = "Implicitly upgraded disk", | |
678 | [IMPLICITLY_UPGRADED_PDSK] = "Implicitly upgraded pdsk", | |
679 | }; | |
680 | ||
681 | if (warn != NO_WARNING) | |
682 | dev_warn(DEV, "%s\n", msg_table[warn]); | |
683 | } | |
684 | ||
b8907339 PR |
685 | /** |
686 | * sanitize_state() - Resolves implicitly necessary additional changes to a state transition | |
687 | * @mdev: DRBD device. | |
688 | * @os: old state. | |
689 | * @ns: new state. | |
690 | * @warn_sync_abort: | |
691 | * | |
692 | * When we loose connection, we have to set the state of the peers disk (pdsk) | |
693 | * to D_UNKNOWN. This rule and many more along those lines are in this function. | |
694 | */ | |
4308a0a3 | 695 | static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state ns, |
d942ae44 | 696 | enum sanitize_state_warnings *warn) |
b8907339 PR |
697 | { |
698 | enum drbd_fencing_p fp; | |
699 | enum drbd_disk_state disk_min, disk_max, pdsk_min, pdsk_max; | |
700 | ||
d942ae44 PR |
701 | if (warn) |
702 | *warn = NO_WARNING; | |
703 | ||
b8907339 PR |
704 | fp = FP_DONT_CARE; |
705 | if (get_ldev(mdev)) { | |
daeda1cc PR |
706 | rcu_read_lock(); |
707 | fp = rcu_dereference(mdev->ldev->disk_conf)->fencing; | |
708 | rcu_read_unlock(); | |
b8907339 PR |
709 | put_ldev(mdev); |
710 | } | |
711 | ||
3509502d | 712 | /* Implications from connection to peer and peer_isp */ |
b8907339 PR |
713 | if (ns.conn < C_CONNECTED) { |
714 | ns.peer_isp = 0; | |
715 | ns.peer = R_UNKNOWN; | |
716 | if (ns.pdsk > D_UNKNOWN || ns.pdsk < D_INCONSISTENT) | |
717 | ns.pdsk = D_UNKNOWN; | |
718 | } | |
719 | ||
720 | /* Clear the aftr_isp when becoming unconfigured */ | |
721 | if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY) | |
722 | ns.aftr_isp = 0; | |
723 | ||
4308a0a3 | 724 | /* An implication of the disk states onto the connection state */ |
b8907339 | 725 | /* Abort resync if a disk fails/detaches */ |
4308a0a3 | 726 | if (ns.conn > C_CONNECTED && (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) { |
d942ae44 PR |
727 | if (warn) |
728 | *warn = ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T ? | |
729 | ABORTED_ONLINE_VERIFY : ABORTED_RESYNC; | |
b8907339 PR |
730 | ns.conn = C_CONNECTED; |
731 | } | |
732 | ||
733 | /* Connection breaks down before we finished "Negotiating" */ | |
734 | if (ns.conn < C_CONNECTED && ns.disk == D_NEGOTIATING && | |
735 | get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
736 | if (mdev->ed_uuid == mdev->ldev->md.uuid[UI_CURRENT]) { | |
737 | ns.disk = mdev->new_state_tmp.disk; | |
738 | ns.pdsk = mdev->new_state_tmp.pdsk; | |
739 | } else { | |
d942ae44 PR |
740 | if (warn) |
741 | *warn = CONNECTION_LOST_NEGOTIATING; | |
b8907339 PR |
742 | ns.disk = D_DISKLESS; |
743 | ns.pdsk = D_UNKNOWN; | |
744 | } | |
745 | put_ldev(mdev); | |
746 | } | |
747 | ||
748 | /* D_CONSISTENT and D_OUTDATED vanish when we get connected */ | |
749 | if (ns.conn >= C_CONNECTED && ns.conn < C_AHEAD) { | |
750 | if (ns.disk == D_CONSISTENT || ns.disk == D_OUTDATED) | |
751 | ns.disk = D_UP_TO_DATE; | |
752 | if (ns.pdsk == D_CONSISTENT || ns.pdsk == D_OUTDATED) | |
753 | ns.pdsk = D_UP_TO_DATE; | |
754 | } | |
755 | ||
756 | /* Implications of the connection stat on the disk states */ | |
757 | disk_min = D_DISKLESS; | |
758 | disk_max = D_UP_TO_DATE; | |
759 | pdsk_min = D_INCONSISTENT; | |
760 | pdsk_max = D_UNKNOWN; | |
761 | switch ((enum drbd_conns)ns.conn) { | |
762 | case C_WF_BITMAP_T: | |
763 | case C_PAUSED_SYNC_T: | |
764 | case C_STARTING_SYNC_T: | |
765 | case C_WF_SYNC_UUID: | |
766 | case C_BEHIND: | |
767 | disk_min = D_INCONSISTENT; | |
768 | disk_max = D_OUTDATED; | |
769 | pdsk_min = D_UP_TO_DATE; | |
770 | pdsk_max = D_UP_TO_DATE; | |
771 | break; | |
772 | case C_VERIFY_S: | |
773 | case C_VERIFY_T: | |
774 | disk_min = D_UP_TO_DATE; | |
775 | disk_max = D_UP_TO_DATE; | |
776 | pdsk_min = D_UP_TO_DATE; | |
777 | pdsk_max = D_UP_TO_DATE; | |
778 | break; | |
779 | case C_CONNECTED: | |
780 | disk_min = D_DISKLESS; | |
781 | disk_max = D_UP_TO_DATE; | |
782 | pdsk_min = D_DISKLESS; | |
783 | pdsk_max = D_UP_TO_DATE; | |
784 | break; | |
785 | case C_WF_BITMAP_S: | |
786 | case C_PAUSED_SYNC_S: | |
787 | case C_STARTING_SYNC_S: | |
788 | case C_AHEAD: | |
789 | disk_min = D_UP_TO_DATE; | |
790 | disk_max = D_UP_TO_DATE; | |
791 | pdsk_min = D_INCONSISTENT; | |
792 | pdsk_max = D_CONSISTENT; /* D_OUTDATED would be nice. But explicit outdate necessary*/ | |
793 | break; | |
794 | case C_SYNC_TARGET: | |
795 | disk_min = D_INCONSISTENT; | |
796 | disk_max = D_INCONSISTENT; | |
797 | pdsk_min = D_UP_TO_DATE; | |
798 | pdsk_max = D_UP_TO_DATE; | |
799 | break; | |
800 | case C_SYNC_SOURCE: | |
801 | disk_min = D_UP_TO_DATE; | |
802 | disk_max = D_UP_TO_DATE; | |
803 | pdsk_min = D_INCONSISTENT; | |
804 | pdsk_max = D_INCONSISTENT; | |
805 | break; | |
806 | case C_STANDALONE: | |
807 | case C_DISCONNECTING: | |
808 | case C_UNCONNECTED: | |
809 | case C_TIMEOUT: | |
810 | case C_BROKEN_PIPE: | |
811 | case C_NETWORK_FAILURE: | |
812 | case C_PROTOCOL_ERROR: | |
813 | case C_TEAR_DOWN: | |
814 | case C_WF_CONNECTION: | |
815 | case C_WF_REPORT_PARAMS: | |
816 | case C_MASK: | |
817 | break; | |
818 | } | |
819 | if (ns.disk > disk_max) | |
820 | ns.disk = disk_max; | |
821 | ||
822 | if (ns.disk < disk_min) { | |
d942ae44 PR |
823 | if (warn) |
824 | *warn = IMPLICITLY_UPGRADED_DISK; | |
b8907339 PR |
825 | ns.disk = disk_min; |
826 | } | |
827 | if (ns.pdsk > pdsk_max) | |
828 | ns.pdsk = pdsk_max; | |
829 | ||
830 | if (ns.pdsk < pdsk_min) { | |
d942ae44 PR |
831 | if (warn) |
832 | *warn = IMPLICITLY_UPGRADED_PDSK; | |
b8907339 PR |
833 | ns.pdsk = pdsk_min; |
834 | } | |
835 | ||
836 | if (fp == FP_STONITH && | |
4308a0a3 | 837 | (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED)) |
b8907339 PR |
838 | ns.susp_fen = 1; /* Suspend IO while fence-peer handler runs (peer lost) */ |
839 | ||
f399002e | 840 | if (mdev->tconn->res_opts.on_no_data == OND_SUSPEND_IO && |
4308a0a3 | 841 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) |
b8907339 PR |
842 | ns.susp_nod = 1; /* Suspend IO while no data available (no accessible data available) */ |
843 | ||
844 | if (ns.aftr_isp || ns.peer_isp || ns.user_isp) { | |
845 | if (ns.conn == C_SYNC_SOURCE) | |
846 | ns.conn = C_PAUSED_SYNC_S; | |
847 | if (ns.conn == C_SYNC_TARGET) | |
848 | ns.conn = C_PAUSED_SYNC_T; | |
849 | } else { | |
850 | if (ns.conn == C_PAUSED_SYNC_S) | |
851 | ns.conn = C_SYNC_SOURCE; | |
852 | if (ns.conn == C_PAUSED_SYNC_T) | |
853 | ns.conn = C_SYNC_TARGET; | |
854 | } | |
855 | ||
856 | return ns; | |
857 | } | |
858 | ||
859 | void drbd_resume_al(struct drbd_conf *mdev) | |
860 | { | |
861 | if (test_and_clear_bit(AL_SUSPENDED, &mdev->flags)) | |
862 | dev_info(DEV, "Resumed AL updates\n"); | |
863 | } | |
864 | ||
865 | /* helper for __drbd_set_state */ | |
866 | static void set_ov_position(struct drbd_conf *mdev, enum drbd_conns cs) | |
867 | { | |
868 | if (mdev->tconn->agreed_pro_version < 90) | |
869 | mdev->ov_start_sector = 0; | |
870 | mdev->rs_total = drbd_bm_bits(mdev); | |
871 | mdev->ov_position = 0; | |
872 | if (cs == C_VERIFY_T) { | |
873 | /* starting online verify from an arbitrary position | |
874 | * does not fit well into the existing protocol. | |
875 | * on C_VERIFY_T, we initialize ov_left and friends | |
876 | * implicitly in receive_DataRequest once the | |
877 | * first P_OV_REQUEST is received */ | |
878 | mdev->ov_start_sector = ~(sector_t)0; | |
879 | } else { | |
880 | unsigned long bit = BM_SECT_TO_BIT(mdev->ov_start_sector); | |
881 | if (bit >= mdev->rs_total) { | |
882 | mdev->ov_start_sector = | |
883 | BM_BIT_TO_SECT(mdev->rs_total - 1); | |
884 | mdev->rs_total = 1; | |
885 | } else | |
886 | mdev->rs_total -= bit; | |
887 | mdev->ov_position = mdev->ov_start_sector; | |
888 | } | |
889 | mdev->ov_left = mdev->rs_total; | |
890 | } | |
891 | ||
892 | /** | |
893 | * __drbd_set_state() - Set a new DRBD state | |
894 | * @mdev: DRBD device. | |
895 | * @ns: new state. | |
896 | * @flags: Flags | |
897 | * @done: Optional completion, that will get completed after the after_state_ch() finished | |
898 | * | |
899 | * Caller needs to hold req_lock, and global_state_lock. Do not call directly. | |
900 | */ | |
901 | enum drbd_state_rv | |
902 | __drbd_set_state(struct drbd_conf *mdev, union drbd_state ns, | |
903 | enum chg_state_flags flags, struct completion *done) | |
904 | { | |
905 | union drbd_state os; | |
906 | enum drbd_state_rv rv = SS_SUCCESS; | |
d942ae44 | 907 | enum sanitize_state_warnings ssw; |
b8907339 PR |
908 | struct after_state_chg_work *ascw; |
909 | ||
78bae59b | 910 | os = drbd_read_state(mdev); |
b8907339 | 911 | |
d942ae44 | 912 | ns = sanitize_state(mdev, ns, &ssw); |
b8907339 PR |
913 | if (ns.i == os.i) |
914 | return SS_NOTHING_TO_DO; | |
915 | ||
3509502d PR |
916 | rv = is_valid_transition(os, ns); |
917 | if (rv < SS_SUCCESS) | |
918 | return rv; | |
919 | ||
b8907339 PR |
920 | if (!(flags & CS_HARD)) { |
921 | /* pre-state-change checks ; only look at ns */ | |
922 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
923 | ||
924 | rv = is_valid_state(mdev, ns); | |
925 | if (rv < SS_SUCCESS) { | |
926 | /* If the old state was illegal as well, then let | |
927 | this happen...*/ | |
928 | ||
929 | if (is_valid_state(mdev, os) == rv) | |
a75f34ad | 930 | rv = is_valid_soft_transition(os, ns); |
b8907339 | 931 | } else |
a75f34ad | 932 | rv = is_valid_soft_transition(os, ns); |
b8907339 PR |
933 | } |
934 | ||
935 | if (rv < SS_SUCCESS) { | |
936 | if (flags & CS_VERBOSE) | |
937 | print_st_err(mdev, os, ns, rv); | |
938 | return rv; | |
939 | } | |
940 | ||
d942ae44 | 941 | print_sanitize_warnings(mdev, ssw); |
b8907339 | 942 | |
435693e8 | 943 | drbd_pr_state_change(mdev, os, ns, flags); |
b8907339 | 944 | |
706cb24c PR |
945 | /* Display changes to the susp* flags that where caused by the call to |
946 | sanitize_state(). Only display it here if we where not called from | |
947 | _conn_request_state() */ | |
948 | if (!(flags & CS_DC_SUSP)) | |
949 | conn_pr_state_change(mdev->tconn, os, ns, (flags & ~CS_DC_MASK) | CS_DC_SUSP); | |
950 | ||
b8907339 PR |
951 | /* if we are going -> D_FAILED or D_DISKLESS, grab one extra reference |
952 | * on the ldev here, to be sure the transition -> D_DISKLESS resp. | |
953 | * drbd_ldev_destroy() won't happen before our corresponding | |
954 | * after_state_ch works run, where we put_ldev again. */ | |
955 | if ((os.disk != D_FAILED && ns.disk == D_FAILED) || | |
956 | (os.disk != D_DISKLESS && ns.disk == D_DISKLESS)) | |
957 | atomic_inc(&mdev->local_cnt); | |
958 | ||
da9fbc27 | 959 | mdev->state.i = ns.i; |
8e0af25f PR |
960 | mdev->tconn->susp = ns.susp; |
961 | mdev->tconn->susp_nod = ns.susp_nod; | |
962 | mdev->tconn->susp_fen = ns.susp_fen; | |
b8907339 PR |
963 | |
964 | if (os.disk == D_ATTACHING && ns.disk >= D_NEGOTIATING) | |
965 | drbd_print_uuids(mdev, "attached to UUIDs"); | |
966 | ||
967 | wake_up(&mdev->misc_wait); | |
968 | wake_up(&mdev->state_wait); | |
2a67d8b9 | 969 | wake_up(&mdev->tconn->ping_wait); |
b8907339 PR |
970 | |
971 | /* aborted verify run. log the last position */ | |
972 | if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) && | |
973 | ns.conn < C_CONNECTED) { | |
974 | mdev->ov_start_sector = | |
975 | BM_BIT_TO_SECT(drbd_bm_bits(mdev) - mdev->ov_left); | |
976 | dev_info(DEV, "Online Verify reached sector %llu\n", | |
977 | (unsigned long long)mdev->ov_start_sector); | |
978 | } | |
979 | ||
980 | if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) && | |
981 | (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) { | |
982 | dev_info(DEV, "Syncer continues.\n"); | |
983 | mdev->rs_paused += (long)jiffies | |
984 | -(long)mdev->rs_mark_time[mdev->rs_last_mark]; | |
985 | if (ns.conn == C_SYNC_TARGET) | |
986 | mod_timer(&mdev->resync_timer, jiffies); | |
987 | } | |
988 | ||
989 | if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) && | |
990 | (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) { | |
991 | dev_info(DEV, "Resync suspended\n"); | |
992 | mdev->rs_mark_time[mdev->rs_last_mark] = jiffies; | |
993 | } | |
994 | ||
995 | if (os.conn == C_CONNECTED && | |
996 | (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) { | |
997 | unsigned long now = jiffies; | |
998 | int i; | |
999 | ||
1000 | set_ov_position(mdev, ns.conn); | |
1001 | mdev->rs_start = now; | |
1002 | mdev->rs_last_events = 0; | |
1003 | mdev->rs_last_sect_ev = 0; | |
1004 | mdev->ov_last_oos_size = 0; | |
1005 | mdev->ov_last_oos_start = 0; | |
1006 | ||
1007 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { | |
1008 | mdev->rs_mark_left[i] = mdev->ov_left; | |
1009 | mdev->rs_mark_time[i] = now; | |
1010 | } | |
1011 | ||
1012 | drbd_rs_controller_reset(mdev); | |
1013 | ||
1014 | if (ns.conn == C_VERIFY_S) { | |
1015 | dev_info(DEV, "Starting Online Verify from sector %llu\n", | |
1016 | (unsigned long long)mdev->ov_position); | |
1017 | mod_timer(&mdev->resync_timer, jiffies); | |
1018 | } | |
1019 | } | |
1020 | ||
1021 | if (get_ldev(mdev)) { | |
1022 | u32 mdf = mdev->ldev->md.flags & ~(MDF_CONSISTENT|MDF_PRIMARY_IND| | |
1023 | MDF_CONNECTED_IND|MDF_WAS_UP_TO_DATE| | |
1024 | MDF_PEER_OUT_DATED|MDF_CRASHED_PRIMARY); | |
1025 | ||
d5d7ebd4 | 1026 | mdf &= ~MDF_AL_CLEAN; |
b8907339 PR |
1027 | if (test_bit(CRASHED_PRIMARY, &mdev->flags)) |
1028 | mdf |= MDF_CRASHED_PRIMARY; | |
1029 | if (mdev->state.role == R_PRIMARY || | |
1030 | (mdev->state.pdsk < D_INCONSISTENT && mdev->state.peer == R_PRIMARY)) | |
1031 | mdf |= MDF_PRIMARY_IND; | |
1032 | if (mdev->state.conn > C_WF_REPORT_PARAMS) | |
1033 | mdf |= MDF_CONNECTED_IND; | |
1034 | if (mdev->state.disk > D_INCONSISTENT) | |
1035 | mdf |= MDF_CONSISTENT; | |
1036 | if (mdev->state.disk > D_OUTDATED) | |
1037 | mdf |= MDF_WAS_UP_TO_DATE; | |
1038 | if (mdev->state.pdsk <= D_OUTDATED && mdev->state.pdsk >= D_INCONSISTENT) | |
1039 | mdf |= MDF_PEER_OUT_DATED; | |
1040 | if (mdf != mdev->ldev->md.flags) { | |
1041 | mdev->ldev->md.flags = mdf; | |
1042 | drbd_md_mark_dirty(mdev); | |
1043 | } | |
1044 | if (os.disk < D_CONSISTENT && ns.disk >= D_CONSISTENT) | |
1045 | drbd_set_ed_uuid(mdev, mdev->ldev->md.uuid[UI_CURRENT]); | |
1046 | put_ldev(mdev); | |
1047 | } | |
1048 | ||
1049 | /* Peer was forced D_UP_TO_DATE & R_PRIMARY, consider to resync */ | |
1050 | if (os.disk == D_INCONSISTENT && os.pdsk == D_INCONSISTENT && | |
1051 | os.peer == R_SECONDARY && ns.peer == R_PRIMARY) | |
1052 | set_bit(CONSIDER_RESYNC, &mdev->flags); | |
1053 | ||
1054 | /* Receiver should clean up itself */ | |
1055 | if (os.conn != C_DISCONNECTING && ns.conn == C_DISCONNECTING) | |
1056 | drbd_thread_stop_nowait(&mdev->tconn->receiver); | |
1057 | ||
1058 | /* Now the receiver finished cleaning up itself, it should die */ | |
1059 | if (os.conn != C_STANDALONE && ns.conn == C_STANDALONE) | |
1060 | drbd_thread_stop_nowait(&mdev->tconn->receiver); | |
1061 | ||
1062 | /* Upon network failure, we need to restart the receiver. */ | |
823bd832 | 1063 | if (os.conn > C_WF_CONNECTION && |
b8907339 PR |
1064 | ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT) |
1065 | drbd_thread_restart_nowait(&mdev->tconn->receiver); | |
1066 | ||
1067 | /* Resume AL writing if we get a connection */ | |
1068 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) | |
1069 | drbd_resume_al(mdev); | |
1070 | ||
1071 | ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC); | |
1072 | if (ascw) { | |
1073 | ascw->os = os; | |
1074 | ascw->ns = ns; | |
1075 | ascw->flags = flags; | |
1076 | ascw->w.cb = w_after_state_ch; | |
a21e9298 | 1077 | ascw->w.mdev = mdev; |
b8907339 PR |
1078 | ascw->done = done; |
1079 | drbd_queue_work(&mdev->tconn->data.work, &ascw->w); | |
1080 | } else { | |
bbeb641c | 1081 | dev_err(DEV, "Could not kmalloc an ascw\n"); |
b8907339 PR |
1082 | } |
1083 | ||
1084 | return rv; | |
1085 | } | |
1086 | ||
99920dc5 | 1087 | static int w_after_state_ch(struct drbd_work *w, int unused) |
b8907339 PR |
1088 | { |
1089 | struct after_state_chg_work *ascw = | |
1090 | container_of(w, struct after_state_chg_work, w); | |
00d56944 | 1091 | struct drbd_conf *mdev = w->mdev; |
b8907339 PR |
1092 | |
1093 | after_state_ch(mdev, ascw->os, ascw->ns, ascw->flags); | |
1094 | if (ascw->flags & CS_WAIT_COMPLETE) { | |
1095 | D_ASSERT(ascw->done != NULL); | |
1096 | complete(ascw->done); | |
1097 | } | |
1098 | kfree(ascw); | |
1099 | ||
99920dc5 | 1100 | return 0; |
b8907339 PR |
1101 | } |
1102 | ||
1103 | static void abw_start_sync(struct drbd_conf *mdev, int rv) | |
1104 | { | |
1105 | if (rv) { | |
1106 | dev_err(DEV, "Writing the bitmap failed not starting resync.\n"); | |
1107 | _drbd_request_state(mdev, NS(conn, C_CONNECTED), CS_VERBOSE); | |
1108 | return; | |
1109 | } | |
1110 | ||
1111 | switch (mdev->state.conn) { | |
1112 | case C_STARTING_SYNC_T: | |
1113 | _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); | |
1114 | break; | |
1115 | case C_STARTING_SYNC_S: | |
1116 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
1117 | break; | |
1118 | } | |
1119 | } | |
1120 | ||
1121 | int drbd_bitmap_io_from_worker(struct drbd_conf *mdev, | |
1122 | int (*io_fn)(struct drbd_conf *), | |
1123 | char *why, enum bm_flag flags) | |
1124 | { | |
1125 | int rv; | |
1126 | ||
1127 | D_ASSERT(current == mdev->tconn->worker.task); | |
1128 | ||
1129 | /* open coded non-blocking drbd_suspend_io(mdev); */ | |
1130 | set_bit(SUSPEND_IO, &mdev->flags); | |
1131 | ||
1132 | drbd_bm_lock(mdev, why, flags); | |
1133 | rv = io_fn(mdev); | |
1134 | drbd_bm_unlock(mdev); | |
1135 | ||
1136 | drbd_resume_io(mdev); | |
1137 | ||
1138 | return rv; | |
1139 | } | |
1140 | ||
1141 | /** | |
1142 | * after_state_ch() - Perform after state change actions that may sleep | |
1143 | * @mdev: DRBD device. | |
1144 | * @os: old state. | |
1145 | * @ns: new state. | |
1146 | * @flags: Flags | |
1147 | */ | |
1148 | static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, | |
1149 | union drbd_state ns, enum chg_state_flags flags) | |
1150 | { | |
3b98c0c2 LE |
1151 | struct sib_info sib; |
1152 | ||
1153 | sib.sib_reason = SIB_STATE_CHANGE; | |
1154 | sib.os = os; | |
1155 | sib.ns = ns; | |
b8907339 PR |
1156 | |
1157 | if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) { | |
1158 | clear_bit(CRASHED_PRIMARY, &mdev->flags); | |
1159 | if (mdev->p_uuid) | |
1160 | mdev->p_uuid[UI_FLAGS] &= ~((u64)2); | |
1161 | } | |
1162 | ||
b8907339 | 1163 | /* Inform userspace about the change... */ |
3b98c0c2 | 1164 | drbd_bcast_event(mdev, &sib); |
b8907339 PR |
1165 | |
1166 | if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) && | |
1167 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) | |
1168 | drbd_khelper(mdev, "pri-on-incon-degr"); | |
1169 | ||
1170 | /* Here we have the actions that are performed after a | |
1171 | state change. This function might sleep */ | |
1172 | ||
b8907339 | 1173 | if (ns.susp_nod) { |
a6d00c8e PR |
1174 | enum drbd_req_event what = NOTHING; |
1175 | ||
1176 | if (os.conn < C_CONNECTED && conn_lowest_conn(mdev->tconn) >= C_CONNECTED) | |
b8907339 PR |
1177 | what = RESEND; |
1178 | ||
3fb4746d PR |
1179 | if ((os.disk == D_ATTACHING || os.disk == D_NEGOTIATING) && |
1180 | conn_lowest_disk(mdev->tconn) > D_NEGOTIATING) | |
b8907339 PR |
1181 | what = RESTART_FROZEN_DISK_IO; |
1182 | ||
a6d00c8e | 1183 | if (what != NOTHING) { |
b8907339 | 1184 | spin_lock_irq(&mdev->tconn->req_lock); |
a6d00c8e PR |
1185 | _tl_restart(mdev->tconn, what); |
1186 | _drbd_set_state(_NS(mdev, susp_nod, 0), CS_VERBOSE, NULL); | |
b8907339 PR |
1187 | spin_unlock_irq(&mdev->tconn->req_lock); |
1188 | } | |
b8907339 PR |
1189 | } |
1190 | ||
1191 | /* Became sync source. With protocol >= 96, we still need to send out | |
1192 | * the sync uuid now. Need to do that before any drbd_send_state, or | |
1193 | * the other side may go "paused sync" before receiving the sync uuids, | |
1194 | * which is unexpected. */ | |
1195 | if ((os.conn != C_SYNC_SOURCE && os.conn != C_PAUSED_SYNC_S) && | |
1196 | (ns.conn == C_SYNC_SOURCE || ns.conn == C_PAUSED_SYNC_S) && | |
1197 | mdev->tconn->agreed_pro_version >= 96 && get_ldev(mdev)) { | |
1198 | drbd_gen_and_send_sync_uuid(mdev); | |
1199 | put_ldev(mdev); | |
1200 | } | |
1201 | ||
1202 | /* Do not change the order of the if above and the two below... */ | |
369bea63 PR |
1203 | if (os.pdsk == D_DISKLESS && |
1204 | ns.pdsk > D_DISKLESS && ns.pdsk != D_UNKNOWN) { /* attach on the peer */ | |
b8907339 | 1205 | drbd_send_uuids(mdev); |
43de7c85 | 1206 | drbd_send_state(mdev, ns); |
b8907339 PR |
1207 | } |
1208 | /* No point in queuing send_bitmap if we don't have a connection | |
1209 | * anymore, so check also the _current_ state, not only the new state | |
1210 | * at the time this work was queued. */ | |
1211 | if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S && | |
1212 | mdev->state.conn == C_WF_BITMAP_S) | |
1213 | drbd_queue_bitmap_io(mdev, &drbd_send_bitmap, NULL, | |
1214 | "send_bitmap (WFBitMapS)", | |
1215 | BM_LOCKED_TEST_ALLOWED); | |
1216 | ||
1217 | /* Lost contact to peer's copy of the data */ | |
1218 | if ((os.pdsk >= D_INCONSISTENT && | |
1219 | os.pdsk != D_UNKNOWN && | |
1220 | os.pdsk != D_OUTDATED) | |
1221 | && (ns.pdsk < D_INCONSISTENT || | |
1222 | ns.pdsk == D_UNKNOWN || | |
1223 | ns.pdsk == D_OUTDATED)) { | |
1224 | if (get_ldev(mdev)) { | |
1225 | if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && | |
1226 | mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { | |
2aebfabb | 1227 | if (drbd_suspended(mdev)) { |
b8907339 PR |
1228 | set_bit(NEW_CUR_UUID, &mdev->flags); |
1229 | } else { | |
1230 | drbd_uuid_new_current(mdev); | |
1231 | drbd_send_uuids(mdev); | |
1232 | } | |
1233 | } | |
1234 | put_ldev(mdev); | |
1235 | } | |
1236 | } | |
1237 | ||
1238 | if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) { | |
0cfac5dd PR |
1239 | if (os.peer == R_SECONDARY && ns.peer == R_PRIMARY && |
1240 | mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { | |
1241 | drbd_uuid_new_current(mdev); | |
1242 | drbd_send_uuids(mdev); | |
1243 | } | |
b8907339 PR |
1244 | /* D_DISKLESS Peer becomes secondary */ |
1245 | if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) | |
1246 | /* We may still be Primary ourselves. | |
1247 | * No harm done if the bitmap still changes, | |
1248 | * redirtied pages will follow later. */ | |
1249 | drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, | |
1250 | "demote diskless peer", BM_LOCKED_SET_ALLOWED); | |
1251 | put_ldev(mdev); | |
1252 | } | |
1253 | ||
1254 | /* Write out all changed bits on demote. | |
1255 | * Though, no need to da that just yet | |
1256 | * if there is a resync going on still */ | |
1257 | if (os.role == R_PRIMARY && ns.role == R_SECONDARY && | |
1258 | mdev->state.conn <= C_CONNECTED && get_ldev(mdev)) { | |
1259 | /* No changes to the bitmap expected this time, so assert that, | |
1260 | * even though no harm was done if it did change. */ | |
1261 | drbd_bitmap_io_from_worker(mdev, &drbd_bm_write, | |
1262 | "demote", BM_LOCKED_TEST_ALLOWED); | |
1263 | put_ldev(mdev); | |
1264 | } | |
1265 | ||
1266 | /* Last part of the attaching process ... */ | |
1267 | if (ns.conn >= C_CONNECTED && | |
1268 | os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) { | |
1269 | drbd_send_sizes(mdev, 0, 0); /* to start sync... */ | |
1270 | drbd_send_uuids(mdev); | |
43de7c85 | 1271 | drbd_send_state(mdev, ns); |
b8907339 PR |
1272 | } |
1273 | ||
1274 | /* We want to pause/continue resync, tell peer. */ | |
1275 | if (ns.conn >= C_CONNECTED && | |
1276 | ((os.aftr_isp != ns.aftr_isp) || | |
1277 | (os.user_isp != ns.user_isp))) | |
43de7c85 | 1278 | drbd_send_state(mdev, ns); |
b8907339 PR |
1279 | |
1280 | /* In case one of the isp bits got set, suspend other devices. */ | |
1281 | if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) && | |
1282 | (ns.aftr_isp || ns.peer_isp || ns.user_isp)) | |
1283 | suspend_other_sg(mdev); | |
1284 | ||
1285 | /* Make sure the peer gets informed about eventual state | |
1286 | changes (ISP bits) while we were in WFReportParams. */ | |
1287 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED) | |
43de7c85 | 1288 | drbd_send_state(mdev, ns); |
b8907339 PR |
1289 | |
1290 | if (os.conn != C_AHEAD && ns.conn == C_AHEAD) | |
43de7c85 | 1291 | drbd_send_state(mdev, ns); |
b8907339 PR |
1292 | |
1293 | /* We are in the progress to start a full sync... */ | |
1294 | if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
1295 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S)) | |
1296 | /* no other bitmap changes expected during this phase */ | |
1297 | drbd_queue_bitmap_io(mdev, | |
1298 | &drbd_bmio_set_n_write, &abw_start_sync, | |
1299 | "set_n_write from StartingSync", BM_LOCKED_TEST_ALLOWED); | |
1300 | ||
1301 | /* We are invalidating our self... */ | |
1302 | if (os.conn < C_CONNECTED && ns.conn < C_CONNECTED && | |
1303 | os.disk > D_INCONSISTENT && ns.disk == D_INCONSISTENT) | |
1304 | /* other bitmap operation expected during this phase */ | |
1305 | drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, | |
1306 | "set_n_write from invalidate", BM_LOCKED_MASK); | |
1307 | ||
1308 | /* first half of local IO error, failure to attach, | |
1309 | * or administrative detach */ | |
1310 | if (os.disk != D_FAILED && ns.disk == D_FAILED) { | |
32db80f6 PR |
1311 | enum drbd_io_error_p eh = EP_PASS_ON; |
1312 | int was_io_error = 0; | |
b8907339 | 1313 | /* corresponding get_ldev was in __drbd_set_state, to serialize |
32db80f6 PR |
1314 | * our cleanup here with the transition to D_DISKLESS. |
1315 | * But is is still not save to dreference ldev here, since | |
1316 | * we might come from an failed Attach before ldev was set. */ | |
1317 | if (mdev->ldev) { | |
1318 | rcu_read_lock(); | |
1319 | eh = rcu_dereference(mdev->ldev->disk_conf)->on_io_error; | |
1320 | rcu_read_unlock(); | |
1321 | ||
1322 | was_io_error = test_and_clear_bit(WAS_IO_ERROR, &mdev->flags); | |
1323 | ||
1324 | /* Immediately allow completion of all application IO, that waits | |
1325 | for completion from the local disk. */ | |
1326 | tl_abort_disk_io(mdev); | |
1327 | ||
1328 | /* current state still has to be D_FAILED, | |
1329 | * there is only one way out: to D_DISKLESS, | |
1330 | * and that may only happen after our put_ldev below. */ | |
1331 | if (mdev->state.disk != D_FAILED) | |
1332 | dev_err(DEV, | |
1333 | "ASSERT FAILED: disk is %s during detach\n", | |
1334 | drbd_disk_str(mdev->state.disk)); | |
1335 | ||
1336 | if (ns.conn >= C_CONNECTED) | |
1337 | drbd_send_state(mdev, ns); | |
1338 | ||
1339 | drbd_rs_cancel_all(mdev); | |
1340 | ||
1341 | /* In case we want to get something to stable storage still, | |
1342 | * this may be the last chance. | |
1343 | * Following put_ldev may transition to D_DISKLESS. */ | |
1344 | drbd_md_sync(mdev); | |
1345 | } | |
b8907339 PR |
1346 | put_ldev(mdev); |
1347 | ||
1348 | if (was_io_error && eh == EP_CALL_HELPER) | |
1349 | drbd_khelper(mdev, "local-io-error"); | |
1350 | } | |
1351 | ||
1352 | /* second half of local IO error, failure to attach, | |
1353 | * or administrative detach, | |
1354 | * after local_cnt references have reached zero again */ | |
1355 | if (os.disk != D_DISKLESS && ns.disk == D_DISKLESS) { | |
1356 | /* We must still be diskless, | |
1357 | * re-attach has to be serialized with this! */ | |
1358 | if (mdev->state.disk != D_DISKLESS) | |
1359 | dev_err(DEV, | |
1360 | "ASSERT FAILED: disk is %s while going diskless\n", | |
1361 | drbd_disk_str(mdev->state.disk)); | |
1362 | ||
1363 | mdev->rs_total = 0; | |
1364 | mdev->rs_failed = 0; | |
1365 | atomic_set(&mdev->rs_pending_cnt, 0); | |
1366 | ||
6ab9b1b6 PR |
1367 | if (ns.conn >= C_CONNECTED) |
1368 | drbd_send_state(mdev, ns); | |
b8907339 PR |
1369 | /* corresponding get_ldev in __drbd_set_state |
1370 | * this may finally trigger drbd_ldev_destroy. */ | |
1371 | put_ldev(mdev); | |
1372 | } | |
1373 | ||
1374 | /* Notify peer that I had a local IO error, and did not detached.. */ | |
6ab9b1b6 | 1375 | if (os.disk == D_UP_TO_DATE && ns.disk == D_INCONSISTENT && ns.conn >= C_CONNECTED) |
43de7c85 | 1376 | drbd_send_state(mdev, ns); |
b8907339 PR |
1377 | |
1378 | /* Disks got bigger while they were detached */ | |
1379 | if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING && | |
1380 | test_and_clear_bit(RESYNC_AFTER_NEG, &mdev->flags)) { | |
1381 | if (ns.conn == C_CONNECTED) | |
1382 | resync_after_online_grow(mdev); | |
1383 | } | |
1384 | ||
1385 | /* A resync finished or aborted, wake paused devices... */ | |
1386 | if ((os.conn > C_CONNECTED && ns.conn <= C_CONNECTED) || | |
1387 | (os.peer_isp && !ns.peer_isp) || | |
1388 | (os.user_isp && !ns.user_isp)) | |
1389 | resume_next_sg(mdev); | |
1390 | ||
1391 | /* sync target done with resync. Explicitly notify peer, even though | |
1392 | * it should (at least for non-empty resyncs) already know itself. */ | |
1393 | if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED) | |
43de7c85 | 1394 | drbd_send_state(mdev, ns); |
b8907339 PR |
1395 | |
1396 | /* This triggers bitmap writeout of potentially still unwritten pages | |
1397 | * if the resync finished cleanly, or aborted because of peer disk | |
1398 | * failure, or because of connection loss. | |
1399 | * For resync aborted because of local disk failure, we cannot do | |
1400 | * any bitmap writeout anymore. | |
1401 | * No harm done if some bits change during this phase. | |
1402 | */ | |
1403 | if (os.conn > C_CONNECTED && ns.conn <= C_CONNECTED && get_ldev(mdev)) { | |
1404 | drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, | |
1405 | "write from resync_finished", BM_LOCKED_SET_ALLOWED); | |
1406 | put_ldev(mdev); | |
1407 | } | |
1408 | ||
1409 | if (ns.disk == D_DISKLESS && | |
1410 | ns.conn == C_STANDALONE && | |
1411 | ns.role == R_SECONDARY) { | |
1412 | if (os.aftr_isp != ns.aftr_isp) | |
1413 | resume_next_sg(mdev); | |
1414 | } | |
1415 | ||
b8907339 PR |
1416 | drbd_md_sync(mdev); |
1417 | } | |
1418 | ||
bbeb641c PR |
1419 | struct after_conn_state_chg_work { |
1420 | struct drbd_work w; | |
1421 | enum drbd_conns oc; | |
8c7e16c3 | 1422 | union drbd_state ns_min; |
5f082f98 | 1423 | union drbd_state ns_max; /* new, max state, over all mdevs */ |
bbeb641c PR |
1424 | enum chg_state_flags flags; |
1425 | }; | |
1426 | ||
99920dc5 | 1427 | static int w_after_conn_state_ch(struct drbd_work *w, int unused) |
bbeb641c PR |
1428 | { |
1429 | struct after_conn_state_chg_work *acscw = | |
1430 | container_of(w, struct after_conn_state_chg_work, w); | |
1431 | struct drbd_tconn *tconn = w->tconn; | |
1432 | enum drbd_conns oc = acscw->oc; | |
5f082f98 | 1433 | union drbd_state ns_max = acscw->ns_max; |
a6d00c8e PR |
1434 | union drbd_state ns_min = acscw->ns_min; |
1435 | struct drbd_conf *mdev; | |
1436 | int vnr; | |
bbeb641c PR |
1437 | |
1438 | kfree(acscw); | |
1439 | ||
b8907339 | 1440 | /* Upon network configuration, we need to start the receiver */ |
5f082f98 | 1441 | if (oc == C_STANDALONE && ns_max.conn == C_UNCONNECTED) |
b8907339 PR |
1442 | drbd_thread_start(&tconn->receiver); |
1443 | ||
f3dfa40a LE |
1444 | if (oc == C_DISCONNECTING && ns_max.conn == C_STANDALONE) { |
1445 | struct net_conf *old_conf; | |
1446 | ||
a0095508 | 1447 | mutex_lock(&tconn->conf_update); |
f3dfa40a | 1448 | old_conf = tconn->net_conf; |
089c075d AG |
1449 | tconn->my_addr_len = 0; |
1450 | tconn->peer_addr_len = 0; | |
f3dfa40a LE |
1451 | rcu_assign_pointer(tconn->net_conf, NULL); |
1452 | conn_free_crypto(tconn); | |
a0095508 | 1453 | mutex_unlock(&tconn->conf_update); |
f3dfa40a LE |
1454 | |
1455 | synchronize_rcu(); | |
1456 | kfree(old_conf); | |
1457 | } | |
1458 | ||
a6d00c8e PR |
1459 | if (ns_max.susp_fen) { |
1460 | /* case1: The outdate peer handler is successful: */ | |
1461 | if (ns_max.pdsk <= D_OUTDATED) { | |
1462 | tl_clear(tconn); | |
695d08fa | 1463 | rcu_read_lock(); |
a6d00c8e PR |
1464 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1465 | if (test_bit(NEW_CUR_UUID, &mdev->flags)) { | |
1466 | drbd_uuid_new_current(mdev); | |
1467 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
1468 | } | |
1469 | } | |
695d08fa | 1470 | rcu_read_unlock(); |
a6d00c8e PR |
1471 | conn_request_state(tconn, |
1472 | (union drbd_state) { { .susp_fen = 1 } }, | |
1473 | (union drbd_state) { { .susp_fen = 0 } }, | |
1474 | CS_VERBOSE); | |
1475 | } | |
1476 | /* case2: The connection was established again: */ | |
1477 | if (ns_min.conn >= C_CONNECTED) { | |
695d08fa | 1478 | rcu_read_lock(); |
a6d00c8e PR |
1479 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
1480 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
695d08fa | 1481 | rcu_read_unlock(); |
a6d00c8e PR |
1482 | spin_lock_irq(&tconn->req_lock); |
1483 | _tl_restart(tconn, RESEND); | |
1484 | _conn_request_state(tconn, | |
1485 | (union drbd_state) { { .susp_fen = 1 } }, | |
1486 | (union drbd_state) { { .susp_fen = 0 } }, | |
1487 | CS_VERBOSE); | |
1488 | spin_unlock_irq(&tconn->req_lock); | |
1489 | } | |
1490 | } | |
9dc9fbb3 | 1491 | kref_put(&tconn->kref, &conn_destroy); |
99920dc5 | 1492 | return 0; |
bbeb641c PR |
1493 | } |
1494 | ||
435693e8 | 1495 | void conn_old_common_state(struct drbd_tconn *tconn, union drbd_state *pcs, enum chg_state_flags *pf) |
88ef594e | 1496 | { |
435693e8 | 1497 | enum chg_state_flags flags = ~0; |
88ef594e | 1498 | struct drbd_conf *mdev; |
435693e8 | 1499 | int vnr, first_vol = 1; |
e0e16653 PR |
1500 | union drbd_dev_state os, cs = { |
1501 | { .role = R_SECONDARY, | |
1502 | .peer = R_UNKNOWN, | |
1503 | .conn = tconn->cstate, | |
1504 | .disk = D_DISKLESS, | |
1505 | .pdsk = D_UNKNOWN, | |
1506 | } }; | |
88ef594e | 1507 | |
695d08fa | 1508 | rcu_read_lock(); |
88ef594e PR |
1509 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1510 | os = mdev->state; | |
1511 | ||
435693e8 PR |
1512 | if (first_vol) { |
1513 | cs = os; | |
1514 | first_vol = 0; | |
1515 | continue; | |
1516 | } | |
1517 | ||
1518 | if (cs.role != os.role) | |
1519 | flags &= ~CS_DC_ROLE; | |
1520 | ||
1521 | if (cs.peer != os.peer) | |
1522 | flags &= ~CS_DC_PEER; | |
1523 | ||
1524 | if (cs.conn != os.conn) | |
1525 | flags &= ~CS_DC_CONN; | |
88ef594e | 1526 | |
435693e8 PR |
1527 | if (cs.disk != os.disk) |
1528 | flags &= ~CS_DC_DISK; | |
88ef594e | 1529 | |
435693e8 PR |
1530 | if (cs.pdsk != os.pdsk) |
1531 | flags &= ~CS_DC_PDSK; | |
1532 | } | |
695d08fa | 1533 | rcu_read_unlock(); |
435693e8 PR |
1534 | |
1535 | *pf |= CS_DC_MASK; | |
1536 | *pf &= flags; | |
da9fbc27 | 1537 | (*pcs).i = cs.i; |
88ef594e | 1538 | } |
bbeb641c | 1539 | |
bd0c824a PR |
1540 | static enum drbd_state_rv |
1541 | conn_is_valid_transition(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val, | |
88ef594e | 1542 | enum chg_state_flags flags) |
bbeb641c | 1543 | { |
bd0c824a | 1544 | enum drbd_state_rv rv = SS_SUCCESS; |
bbeb641c | 1545 | union drbd_state ns, os; |
bd0c824a PR |
1546 | struct drbd_conf *mdev; |
1547 | int vnr; | |
bbeb641c | 1548 | |
695d08fa | 1549 | rcu_read_lock(); |
bd0c824a | 1550 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
78bae59b | 1551 | os = drbd_read_state(mdev); |
bd0c824a PR |
1552 | ns = sanitize_state(mdev, apply_mask_val(os, mask, val), NULL); |
1553 | ||
778bcf2e PR |
1554 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
1555 | ns.disk = os.disk; | |
1556 | ||
bd0c824a PR |
1557 | if (ns.i == os.i) |
1558 | continue; | |
bbeb641c | 1559 | |
bd0c824a PR |
1560 | rv = is_valid_transition(os, ns); |
1561 | if (rv < SS_SUCCESS) | |
1562 | break; | |
1563 | ||
1564 | if (!(flags & CS_HARD)) { | |
1565 | rv = is_valid_state(mdev, ns); | |
1566 | if (rv < SS_SUCCESS) { | |
1567 | if (is_valid_state(mdev, os) == rv) | |
1568 | rv = is_valid_soft_transition(os, ns); | |
1569 | } else | |
1570 | rv = is_valid_soft_transition(os, ns); | |
1571 | } | |
1572 | if (rv < SS_SUCCESS) | |
1573 | break; | |
bbeb641c | 1574 | } |
695d08fa | 1575 | rcu_read_unlock(); |
bbeb641c | 1576 | |
bd0c824a PR |
1577 | if (rv < SS_SUCCESS && flags & CS_VERBOSE) |
1578 | print_st_err(mdev, os, ns, rv); | |
1579 | ||
1580 | return rv; | |
bbeb641c PR |
1581 | } |
1582 | ||
8c7e16c3 | 1583 | void |
bd0c824a | 1584 | conn_set_state(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val, |
8c7e16c3 | 1585 | union drbd_state *pns_min, union drbd_state *pns_max, enum chg_state_flags flags) |
bbeb641c | 1586 | { |
f132f554 | 1587 | union drbd_state ns, os, ns_max = { }; |
8c7e16c3 PR |
1588 | union drbd_state ns_min = { |
1589 | { .role = R_MASK, | |
1590 | .peer = R_MASK, | |
e0e16653 | 1591 | .conn = val.conn, |
8c7e16c3 PR |
1592 | .disk = D_MASK, |
1593 | .pdsk = D_MASK | |
1594 | } }; | |
bd0c824a | 1595 | struct drbd_conf *mdev; |
bbeb641c | 1596 | enum drbd_state_rv rv; |
f132f554 | 1597 | int vnr, number_of_volumes = 0; |
bbeb641c | 1598 | |
bd0c824a PR |
1599 | if (mask.conn == C_MASK) |
1600 | tconn->cstate = val.conn; | |
1601 | ||
695d08fa | 1602 | rcu_read_lock(); |
bd0c824a | 1603 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
f132f554 | 1604 | number_of_volumes++; |
78bae59b | 1605 | os = drbd_read_state(mdev); |
bd0c824a PR |
1606 | ns = apply_mask_val(os, mask, val); |
1607 | ns = sanitize_state(mdev, ns, NULL); | |
bbeb641c | 1608 | |
778bcf2e PR |
1609 | if (flags & CS_IGN_OUTD_FAIL && ns.disk == D_OUTDATED && os.disk < D_OUTDATED) |
1610 | ns.disk = os.disk; | |
1611 | ||
bd0c824a PR |
1612 | rv = __drbd_set_state(mdev, ns, flags, NULL); |
1613 | if (rv < SS_SUCCESS) | |
1614 | BUG(); | |
bbeb641c | 1615 | |
8c7e16c3 PR |
1616 | ns.i = mdev->state.i; |
1617 | ns_max.role = max_role(ns.role, ns_max.role); | |
1618 | ns_max.peer = max_role(ns.peer, ns_max.peer); | |
1619 | ns_max.conn = max_t(enum drbd_conns, ns.conn, ns_max.conn); | |
1620 | ns_max.disk = max_t(enum drbd_disk_state, ns.disk, ns_max.disk); | |
1621 | ns_max.pdsk = max_t(enum drbd_disk_state, ns.pdsk, ns_max.pdsk); | |
1622 | ||
1623 | ns_min.role = min_role(ns.role, ns_min.role); | |
1624 | ns_min.peer = min_role(ns.peer, ns_min.peer); | |
1625 | ns_min.conn = min_t(enum drbd_conns, ns.conn, ns_min.conn); | |
1626 | ns_min.disk = min_t(enum drbd_disk_state, ns.disk, ns_min.disk); | |
1627 | ns_min.pdsk = min_t(enum drbd_disk_state, ns.pdsk, ns_min.pdsk); | |
bd0c824a | 1628 | } |
695d08fa | 1629 | rcu_read_unlock(); |
bbeb641c | 1630 | |
f132f554 PR |
1631 | if (number_of_volumes == 0) { |
1632 | ns_min = ns_max = (union drbd_state) { { | |
1633 | .role = R_SECONDARY, | |
1634 | .peer = R_UNKNOWN, | |
1635 | .conn = val.conn, | |
1636 | .disk = D_DISKLESS, | |
1637 | .pdsk = D_UNKNOWN | |
1638 | } }; | |
1639 | } | |
1640 | ||
8c7e16c3 PR |
1641 | ns_min.susp = ns_max.susp = tconn->susp; |
1642 | ns_min.susp_nod = ns_max.susp_nod = tconn->susp_nod; | |
1643 | ns_min.susp_fen = ns_max.susp_fen = tconn->susp_fen; | |
1644 | ||
1645 | *pns_min = ns_min; | |
1646 | *pns_max = ns_max; | |
bbeb641c PR |
1647 | } |
1648 | ||
df24aa45 PR |
1649 | static enum drbd_state_rv |
1650 | _conn_rq_cond(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val) | |
1651 | { | |
df24aa45 PR |
1652 | enum drbd_state_rv rv; |
1653 | ||
1654 | if (test_and_clear_bit(CONN_WD_ST_CHG_OKAY, &tconn->flags)) | |
1655 | return SS_CW_SUCCESS; | |
1656 | ||
1657 | if (test_and_clear_bit(CONN_WD_ST_CHG_FAIL, &tconn->flags)) | |
1658 | return SS_CW_FAILED_BY_PEER; | |
1659 | ||
df24aa45 PR |
1660 | spin_lock_irq(&tconn->req_lock); |
1661 | rv = tconn->cstate != C_WF_REPORT_PARAMS ? SS_CW_NO_NEED : SS_UNKNOWN_ERROR; | |
1662 | ||
1663 | if (rv == SS_UNKNOWN_ERROR) | |
435693e8 | 1664 | rv = conn_is_valid_transition(tconn, mask, val, 0); |
df24aa45 | 1665 | |
bd0c824a PR |
1666 | if (rv == SS_SUCCESS) |
1667 | rv = SS_UNKNOWN_ERROR; /* cont waiting, otherwise fail. */ | |
df24aa45 PR |
1668 | |
1669 | spin_unlock_irq(&tconn->req_lock); | |
1670 | ||
1671 | return rv; | |
1672 | } | |
1673 | ||
1674 | static enum drbd_state_rv | |
1675 | conn_cl_wide(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val, | |
1676 | enum chg_state_flags f) | |
1677 | { | |
1678 | enum drbd_state_rv rv; | |
1679 | ||
1680 | spin_unlock_irq(&tconn->req_lock); | |
1681 | mutex_lock(&tconn->cstate_mutex); | |
1682 | ||
4d0fc3fd | 1683 | set_bit(CONN_WD_ST_CHG_REQ, &tconn->flags); |
caee1c3a | 1684 | if (conn_send_state_req(tconn, mask, val)) { |
4d0fc3fd | 1685 | clear_bit(CONN_WD_ST_CHG_REQ, &tconn->flags); |
df24aa45 PR |
1686 | rv = SS_CW_FAILED_BY_PEER; |
1687 | /* if (f & CS_VERBOSE) | |
1688 | print_st_err(mdev, os, ns, rv); */ | |
1689 | goto abort; | |
1690 | } | |
1691 | ||
1692 | wait_event(tconn->ping_wait, (rv = _conn_rq_cond(tconn, mask, val))); | |
4d0fc3fd | 1693 | clear_bit(CONN_WD_ST_CHG_REQ, &tconn->flags); |
df24aa45 PR |
1694 | |
1695 | abort: | |
1696 | mutex_unlock(&tconn->cstate_mutex); | |
1697 | spin_lock_irq(&tconn->req_lock); | |
1698 | ||
1699 | return rv; | |
1700 | } | |
1701 | ||
bbeb641c PR |
1702 | enum drbd_state_rv |
1703 | _conn_request_state(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val, | |
1704 | enum chg_state_flags flags) | |
1705 | { | |
1706 | enum drbd_state_rv rv = SS_SUCCESS; | |
bbeb641c PR |
1707 | struct after_conn_state_chg_work *acscw; |
1708 | enum drbd_conns oc = tconn->cstate; | |
8c7e16c3 | 1709 | union drbd_state ns_max, ns_min, os; |
bbeb641c | 1710 | |
bbeb641c PR |
1711 | rv = is_valid_conn_transition(oc, val.conn); |
1712 | if (rv < SS_SUCCESS) | |
1713 | goto abort; | |
1714 | ||
88ef594e | 1715 | rv = conn_is_valid_transition(tconn, mask, val, flags); |
bbeb641c PR |
1716 | if (rv < SS_SUCCESS) |
1717 | goto abort; | |
1718 | ||
df24aa45 PR |
1719 | if (oc == C_WF_REPORT_PARAMS && val.conn == C_DISCONNECTING && |
1720 | !(flags & (CS_LOCAL_ONLY | CS_HARD))) { | |
1721 | rv = conn_cl_wide(tconn, mask, val, flags); | |
1722 | if (rv < SS_SUCCESS) | |
1723 | goto abort; | |
1724 | } | |
1725 | ||
435693e8 | 1726 | conn_old_common_state(tconn, &os, &flags); |
706cb24c | 1727 | flags |= CS_DC_SUSP; |
8c7e16c3 | 1728 | conn_set_state(tconn, mask, val, &ns_min, &ns_max, flags); |
5f082f98 | 1729 | conn_pr_state_change(tconn, os, ns_max, flags); |
bbeb641c PR |
1730 | |
1731 | acscw = kmalloc(sizeof(*acscw), GFP_ATOMIC); | |
1732 | if (acscw) { | |
435693e8 | 1733 | acscw->oc = os.conn; |
8c7e16c3 | 1734 | acscw->ns_min = ns_min; |
5f082f98 | 1735 | acscw->ns_max = ns_max; |
bbeb641c PR |
1736 | acscw->flags = flags; |
1737 | acscw->w.cb = w_after_conn_state_ch; | |
9dc9fbb3 | 1738 | kref_get(&tconn->kref); |
bbeb641c PR |
1739 | acscw->w.tconn = tconn; |
1740 | drbd_queue_work(&tconn->data.work, &acscw->w); | |
1741 | } else { | |
1742 | conn_err(tconn, "Could not kmalloc an acscw\n"); | |
b8907339 | 1743 | } |
bbeb641c | 1744 | |
a01842eb PR |
1745 | return rv; |
1746 | abort: | |
1747 | if (flags & CS_VERBOSE) { | |
1748 | conn_err(tconn, "State change failed: %s\n", drbd_set_st_err_str(rv)); | |
1749 | conn_err(tconn, " state = { cs:%s }\n", drbd_conn_str(oc)); | |
1750 | conn_err(tconn, "wanted = { cs:%s }\n", drbd_conn_str(val.conn)); | |
1751 | } | |
bbeb641c PR |
1752 | return rv; |
1753 | } | |
1754 | ||
1755 | enum drbd_state_rv | |
1756 | conn_request_state(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val, | |
1757 | enum chg_state_flags flags) | |
1758 | { | |
1759 | enum drbd_state_rv rv; | |
1760 | ||
1761 | spin_lock_irq(&tconn->req_lock); | |
1762 | rv = _conn_request_state(tconn, mask, val, flags); | |
1763 | spin_unlock_irq(&tconn->req_lock); | |
1764 | ||
1765 | return rv; | |
b8907339 | 1766 | } |