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