Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd.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 | ||
b411b363 | 29 | #include <linux/module.h> |
b411b363 PR |
30 | #include <linux/drbd.h> |
31 | #include <asm/uaccess.h> | |
32 | #include <asm/types.h> | |
33 | #include <net/sock.h> | |
34 | #include <linux/ctype.h> | |
35 | #include <linux/smp_lock.h> | |
36 | #include <linux/fs.h> | |
37 | #include <linux/file.h> | |
38 | #include <linux/proc_fs.h> | |
39 | #include <linux/init.h> | |
40 | #include <linux/mm.h> | |
41 | #include <linux/memcontrol.h> | |
42 | #include <linux/mm_inline.h> | |
43 | #include <linux/slab.h> | |
44 | #include <linux/random.h> | |
45 | #include <linux/reboot.h> | |
46 | #include <linux/notifier.h> | |
47 | #include <linux/kthread.h> | |
48 | ||
49 | #define __KERNEL_SYSCALLS__ | |
50 | #include <linux/unistd.h> | |
51 | #include <linux/vmalloc.h> | |
52 | ||
53 | #include <linux/drbd_limits.h> | |
54 | #include "drbd_int.h" | |
b411b363 PR |
55 | #include "drbd_req.h" /* only for _req_mod in tl_release and tl_clear */ |
56 | ||
57 | #include "drbd_vli.h" | |
58 | ||
59 | struct after_state_chg_work { | |
60 | struct drbd_work w; | |
61 | union drbd_state os; | |
62 | union drbd_state ns; | |
63 | enum chg_state_flags flags; | |
64 | struct completion *done; | |
65 | }; | |
66 | ||
67 | int drbdd_init(struct drbd_thread *); | |
68 | int drbd_worker(struct drbd_thread *); | |
69 | int drbd_asender(struct drbd_thread *); | |
70 | ||
71 | int drbd_init(void); | |
72 | static int drbd_open(struct block_device *bdev, fmode_t mode); | |
73 | static int drbd_release(struct gendisk *gd, fmode_t mode); | |
74 | static int w_after_state_ch(struct drbd_conf *mdev, struct drbd_work *w, int unused); | |
75 | static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, | |
76 | union drbd_state ns, enum chg_state_flags flags); | |
77 | static int w_md_sync(struct drbd_conf *mdev, struct drbd_work *w, int unused); | |
78 | static void md_sync_timer_fn(unsigned long data); | |
79 | static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused); | |
e9e6f3ec | 80 | static int w_go_diskless(struct drbd_conf *mdev, struct drbd_work *w, int unused); |
b411b363 | 81 | |
b411b363 PR |
82 | MODULE_AUTHOR("Philipp Reisner <phil@linbit.com>, " |
83 | "Lars Ellenberg <lars@linbit.com>"); | |
84 | MODULE_DESCRIPTION("drbd - Distributed Replicated Block Device v" REL_VERSION); | |
85 | MODULE_VERSION(REL_VERSION); | |
86 | MODULE_LICENSE("GPL"); | |
87 | MODULE_PARM_DESC(minor_count, "Maximum number of drbd devices (1-255)"); | |
88 | MODULE_ALIAS_BLOCKDEV_MAJOR(DRBD_MAJOR); | |
89 | ||
90 | #include <linux/moduleparam.h> | |
91 | /* allow_open_on_secondary */ | |
92 | MODULE_PARM_DESC(allow_oos, "DONT USE!"); | |
93 | /* thanks to these macros, if compiled into the kernel (not-module), | |
94 | * this becomes the boot parameter drbd.minor_count */ | |
95 | module_param(minor_count, uint, 0444); | |
96 | module_param(disable_sendpage, bool, 0644); | |
97 | module_param(allow_oos, bool, 0); | |
98 | module_param(cn_idx, uint, 0444); | |
99 | module_param(proc_details, int, 0644); | |
100 | ||
101 | #ifdef CONFIG_DRBD_FAULT_INJECTION | |
102 | int enable_faults; | |
103 | int fault_rate; | |
104 | static int fault_count; | |
105 | int fault_devs; | |
106 | /* bitmap of enabled faults */ | |
107 | module_param(enable_faults, int, 0664); | |
108 | /* fault rate % value - applies to all enabled faults */ | |
109 | module_param(fault_rate, int, 0664); | |
110 | /* count of faults inserted */ | |
111 | module_param(fault_count, int, 0664); | |
112 | /* bitmap of devices to insert faults on */ | |
113 | module_param(fault_devs, int, 0644); | |
114 | #endif | |
115 | ||
116 | /* module parameter, defined */ | |
117 | unsigned int minor_count = 32; | |
118 | int disable_sendpage; | |
119 | int allow_oos; | |
120 | unsigned int cn_idx = CN_IDX_DRBD; | |
121 | int proc_details; /* Detail level in proc drbd*/ | |
122 | ||
123 | /* Module parameter for setting the user mode helper program | |
124 | * to run. Default is /sbin/drbdadm */ | |
125 | char usermode_helper[80] = "/sbin/drbdadm"; | |
126 | ||
127 | module_param_string(usermode_helper, usermode_helper, sizeof(usermode_helper), 0644); | |
128 | ||
129 | /* in 2.6.x, our device mapping and config info contains our virtual gendisks | |
130 | * as member "struct gendisk *vdisk;" | |
131 | */ | |
132 | struct drbd_conf **minor_table; | |
133 | ||
134 | struct kmem_cache *drbd_request_cache; | |
135 | struct kmem_cache *drbd_ee_cache; /* epoch entries */ | |
136 | struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ | |
137 | struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ | |
138 | mempool_t *drbd_request_mempool; | |
139 | mempool_t *drbd_ee_mempool; | |
140 | ||
141 | /* I do not use a standard mempool, because: | |
142 | 1) I want to hand out the pre-allocated objects first. | |
143 | 2) I want to be able to interrupt sleeping allocation with a signal. | |
144 | Note: This is a single linked list, the next pointer is the private | |
145 | member of struct page. | |
146 | */ | |
147 | struct page *drbd_pp_pool; | |
148 | spinlock_t drbd_pp_lock; | |
149 | int drbd_pp_vacant; | |
150 | wait_queue_head_t drbd_pp_wait; | |
151 | ||
152 | DEFINE_RATELIMIT_STATE(drbd_ratelimit_state, 5 * HZ, 5); | |
153 | ||
7d4e9d09 | 154 | static const struct block_device_operations drbd_ops = { |
b411b363 PR |
155 | .owner = THIS_MODULE, |
156 | .open = drbd_open, | |
157 | .release = drbd_release, | |
158 | }; | |
159 | ||
160 | #define ARRY_SIZE(A) (sizeof(A)/sizeof(A[0])) | |
161 | ||
162 | #ifdef __CHECKER__ | |
163 | /* When checking with sparse, and this is an inline function, sparse will | |
164 | give tons of false positives. When this is a real functions sparse works. | |
165 | */ | |
166 | int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins) | |
167 | { | |
168 | int io_allowed; | |
169 | ||
170 | atomic_inc(&mdev->local_cnt); | |
171 | io_allowed = (mdev->state.disk >= mins); | |
172 | if (!io_allowed) { | |
173 | if (atomic_dec_and_test(&mdev->local_cnt)) | |
174 | wake_up(&mdev->misc_wait); | |
175 | } | |
176 | return io_allowed; | |
177 | } | |
178 | ||
179 | #endif | |
180 | ||
181 | /** | |
182 | * DOC: The transfer log | |
183 | * | |
184 | * The transfer log is a single linked list of &struct drbd_tl_epoch objects. | |
185 | * mdev->newest_tle points to the head, mdev->oldest_tle points to the tail | |
186 | * of the list. There is always at least one &struct drbd_tl_epoch object. | |
187 | * | |
188 | * Each &struct drbd_tl_epoch has a circular double linked list of requests | |
189 | * attached. | |
190 | */ | |
191 | static int tl_init(struct drbd_conf *mdev) | |
192 | { | |
193 | struct drbd_tl_epoch *b; | |
194 | ||
195 | /* during device minor initialization, we may well use GFP_KERNEL */ | |
196 | b = kmalloc(sizeof(struct drbd_tl_epoch), GFP_KERNEL); | |
197 | if (!b) | |
198 | return 0; | |
199 | INIT_LIST_HEAD(&b->requests); | |
200 | INIT_LIST_HEAD(&b->w.list); | |
201 | b->next = NULL; | |
202 | b->br_number = 4711; | |
7e602c0a | 203 | b->n_writes = 0; |
b411b363 PR |
204 | b->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */ |
205 | ||
206 | mdev->oldest_tle = b; | |
207 | mdev->newest_tle = b; | |
208 | INIT_LIST_HEAD(&mdev->out_of_sequence_requests); | |
209 | ||
210 | mdev->tl_hash = NULL; | |
211 | mdev->tl_hash_s = 0; | |
212 | ||
213 | return 1; | |
214 | } | |
215 | ||
216 | static void tl_cleanup(struct drbd_conf *mdev) | |
217 | { | |
218 | D_ASSERT(mdev->oldest_tle == mdev->newest_tle); | |
219 | D_ASSERT(list_empty(&mdev->out_of_sequence_requests)); | |
220 | kfree(mdev->oldest_tle); | |
221 | mdev->oldest_tle = NULL; | |
222 | kfree(mdev->unused_spare_tle); | |
223 | mdev->unused_spare_tle = NULL; | |
224 | kfree(mdev->tl_hash); | |
225 | mdev->tl_hash = NULL; | |
226 | mdev->tl_hash_s = 0; | |
227 | } | |
228 | ||
229 | /** | |
230 | * _tl_add_barrier() - Adds a barrier to the transfer log | |
231 | * @mdev: DRBD device. | |
232 | * @new: Barrier to be added before the current head of the TL. | |
233 | * | |
234 | * The caller must hold the req_lock. | |
235 | */ | |
236 | void _tl_add_barrier(struct drbd_conf *mdev, struct drbd_tl_epoch *new) | |
237 | { | |
238 | struct drbd_tl_epoch *newest_before; | |
239 | ||
240 | INIT_LIST_HEAD(&new->requests); | |
241 | INIT_LIST_HEAD(&new->w.list); | |
242 | new->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */ | |
243 | new->next = NULL; | |
7e602c0a | 244 | new->n_writes = 0; |
b411b363 PR |
245 | |
246 | newest_before = mdev->newest_tle; | |
247 | /* never send a barrier number == 0, because that is special-cased | |
248 | * when using TCQ for our write ordering code */ | |
249 | new->br_number = (newest_before->br_number+1) ?: 1; | |
250 | if (mdev->newest_tle != new) { | |
251 | mdev->newest_tle->next = new; | |
252 | mdev->newest_tle = new; | |
253 | } | |
254 | } | |
255 | ||
256 | /** | |
257 | * tl_release() - Free or recycle the oldest &struct drbd_tl_epoch object of the TL | |
258 | * @mdev: DRBD device. | |
259 | * @barrier_nr: Expected identifier of the DRBD write barrier packet. | |
260 | * @set_size: Expected number of requests before that barrier. | |
261 | * | |
262 | * In case the passed barrier_nr or set_size does not match the oldest | |
263 | * &struct drbd_tl_epoch objects this function will cause a termination | |
264 | * of the connection. | |
265 | */ | |
266 | void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr, | |
267 | unsigned int set_size) | |
268 | { | |
269 | struct drbd_tl_epoch *b, *nob; /* next old barrier */ | |
270 | struct list_head *le, *tle; | |
271 | struct drbd_request *r; | |
272 | ||
273 | spin_lock_irq(&mdev->req_lock); | |
274 | ||
275 | b = mdev->oldest_tle; | |
276 | ||
277 | /* first some paranoia code */ | |
278 | if (b == NULL) { | |
279 | dev_err(DEV, "BAD! BarrierAck #%u received, but no epoch in tl!?\n", | |
280 | barrier_nr); | |
281 | goto bail; | |
282 | } | |
283 | if (b->br_number != barrier_nr) { | |
284 | dev_err(DEV, "BAD! BarrierAck #%u received, expected #%u!\n", | |
285 | barrier_nr, b->br_number); | |
286 | goto bail; | |
287 | } | |
7e602c0a PR |
288 | if (b->n_writes != set_size) { |
289 | dev_err(DEV, "BAD! BarrierAck #%u received with n_writes=%u, expected n_writes=%u!\n", | |
290 | barrier_nr, set_size, b->n_writes); | |
b411b363 PR |
291 | goto bail; |
292 | } | |
293 | ||
294 | /* Clean up list of requests processed during current epoch */ | |
295 | list_for_each_safe(le, tle, &b->requests) { | |
296 | r = list_entry(le, struct drbd_request, tl_requests); | |
297 | _req_mod(r, barrier_acked); | |
298 | } | |
299 | /* There could be requests on the list waiting for completion | |
300 | of the write to the local disk. To avoid corruptions of | |
301 | slab's data structures we have to remove the lists head. | |
302 | ||
303 | Also there could have been a barrier ack out of sequence, overtaking | |
304 | the write acks - which would be a bug and violating write ordering. | |
305 | To not deadlock in case we lose connection while such requests are | |
306 | still pending, we need some way to find them for the | |
307 | _req_mode(connection_lost_while_pending). | |
308 | ||
309 | These have been list_move'd to the out_of_sequence_requests list in | |
310 | _req_mod(, barrier_acked) above. | |
311 | */ | |
312 | list_del_init(&b->requests); | |
313 | ||
314 | nob = b->next; | |
315 | if (test_and_clear_bit(CREATE_BARRIER, &mdev->flags)) { | |
316 | _tl_add_barrier(mdev, b); | |
317 | if (nob) | |
318 | mdev->oldest_tle = nob; | |
319 | /* if nob == NULL b was the only barrier, and becomes the new | |
320 | barrier. Therefore mdev->oldest_tle points already to b */ | |
321 | } else { | |
322 | D_ASSERT(nob != NULL); | |
323 | mdev->oldest_tle = nob; | |
324 | kfree(b); | |
325 | } | |
326 | ||
327 | spin_unlock_irq(&mdev->req_lock); | |
328 | dec_ap_pending(mdev); | |
329 | ||
330 | return; | |
331 | ||
332 | bail: | |
333 | spin_unlock_irq(&mdev->req_lock); | |
334 | drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR)); | |
335 | } | |
336 | ||
b411b363 | 337 | /** |
11b58e73 | 338 | * _tl_restart() - Walks the transfer log, and applies an action to all requests |
b411b363 | 339 | * @mdev: DRBD device. |
11b58e73 | 340 | * @what: The action/event to perform with all request objects |
b411b363 | 341 | * |
11b58e73 PR |
342 | * @what might be one of connection_lost_while_pending, resend, fail_frozen_disk_io, |
343 | * restart_frozen_disk_io. | |
b411b363 | 344 | */ |
11b58e73 | 345 | static void _tl_restart(struct drbd_conf *mdev, enum drbd_req_event what) |
b411b363 | 346 | { |
11b58e73 | 347 | struct drbd_tl_epoch *b, *tmp, **pn; |
b9b98716 | 348 | struct list_head *le, *tle, carry_reads; |
11b58e73 PR |
349 | struct drbd_request *req; |
350 | int rv, n_writes, n_reads; | |
b411b363 PR |
351 | |
352 | b = mdev->oldest_tle; | |
11b58e73 | 353 | pn = &mdev->oldest_tle; |
b411b363 | 354 | while (b) { |
11b58e73 PR |
355 | n_writes = 0; |
356 | n_reads = 0; | |
b9b98716 | 357 | INIT_LIST_HEAD(&carry_reads); |
b411b363 | 358 | list_for_each_safe(le, tle, &b->requests) { |
11b58e73 PR |
359 | req = list_entry(le, struct drbd_request, tl_requests); |
360 | rv = _req_mod(req, what); | |
361 | ||
362 | n_writes += (rv & MR_WRITE) >> MR_WRITE_SHIFT; | |
363 | n_reads += (rv & MR_READ) >> MR_READ_SHIFT; | |
b411b363 PR |
364 | } |
365 | tmp = b->next; | |
366 | ||
b9b98716 | 367 | if (n_writes) { |
11b58e73 PR |
368 | if (what == resend) { |
369 | b->n_writes = n_writes; | |
370 | if (b->w.cb == NULL) { | |
371 | b->w.cb = w_send_barrier; | |
372 | inc_ap_pending(mdev); | |
373 | set_bit(CREATE_BARRIER, &mdev->flags); | |
374 | } | |
375 | ||
376 | drbd_queue_work(&mdev->data.work, &b->w); | |
377 | } | |
378 | pn = &b->next; | |
379 | } else { | |
b9b98716 PR |
380 | if (n_reads) |
381 | list_add(&carry_reads, &b->requests); | |
11b58e73 PR |
382 | /* there could still be requests on that ring list, |
383 | * in case local io is still pending */ | |
384 | list_del(&b->requests); | |
385 | ||
386 | /* dec_ap_pending corresponding to queue_barrier. | |
387 | * the newest barrier may not have been queued yet, | |
388 | * in which case w.cb is still NULL. */ | |
389 | if (b->w.cb != NULL) | |
390 | dec_ap_pending(mdev); | |
391 | ||
392 | if (b == mdev->newest_tle) { | |
393 | /* recycle, but reinit! */ | |
394 | D_ASSERT(tmp == NULL); | |
395 | INIT_LIST_HEAD(&b->requests); | |
b9b98716 | 396 | list_splice(&carry_reads, &b->requests); |
11b58e73 PR |
397 | INIT_LIST_HEAD(&b->w.list); |
398 | b->w.cb = NULL; | |
399 | b->br_number = net_random(); | |
400 | b->n_writes = 0; | |
401 | ||
402 | *pn = b; | |
403 | break; | |
404 | } | |
405 | *pn = tmp; | |
406 | kfree(b); | |
b411b363 | 407 | } |
b411b363 | 408 | b = tmp; |
b9b98716 | 409 | list_splice(&carry_reads, &b->requests); |
b411b363 | 410 | } |
11b58e73 PR |
411 | } |
412 | ||
413 | ||
414 | /** | |
415 | * tl_clear() - Clears all requests and &struct drbd_tl_epoch objects out of the TL | |
416 | * @mdev: DRBD device. | |
417 | * | |
418 | * This is called after the connection to the peer was lost. The storage covered | |
419 | * by the requests on the transfer gets marked as our of sync. Called from the | |
420 | * receiver thread and the worker thread. | |
421 | */ | |
422 | void tl_clear(struct drbd_conf *mdev) | |
423 | { | |
424 | struct list_head *le, *tle; | |
425 | struct drbd_request *r; | |
426 | ||
427 | spin_lock_irq(&mdev->req_lock); | |
428 | ||
429 | _tl_restart(mdev, connection_lost_while_pending); | |
b411b363 PR |
430 | |
431 | /* we expect this list to be empty. */ | |
432 | D_ASSERT(list_empty(&mdev->out_of_sequence_requests)); | |
433 | ||
434 | /* but just in case, clean it up anyways! */ | |
435 | list_for_each_safe(le, tle, &mdev->out_of_sequence_requests) { | |
436 | r = list_entry(le, struct drbd_request, tl_requests); | |
437 | /* It would be nice to complete outside of spinlock. | |
438 | * But this is easier for now. */ | |
439 | _req_mod(r, connection_lost_while_pending); | |
440 | } | |
441 | ||
442 | /* ensure bit indicating barrier is required is clear */ | |
443 | clear_bit(CREATE_BARRIER, &mdev->flags); | |
444 | ||
288f422e PR |
445 | memset(mdev->app_reads_hash, 0, APP_R_HSIZE*sizeof(void *)); |
446 | ||
b411b363 PR |
447 | spin_unlock_irq(&mdev->req_lock); |
448 | } | |
449 | ||
11b58e73 PR |
450 | void tl_restart(struct drbd_conf *mdev, enum drbd_req_event what) |
451 | { | |
452 | spin_lock_irq(&mdev->req_lock); | |
453 | _tl_restart(mdev, what); | |
454 | spin_unlock_irq(&mdev->req_lock); | |
455 | } | |
456 | ||
b411b363 PR |
457 | /** |
458 | * cl_wide_st_chg() - TRUE if the state change is a cluster wide one | |
459 | * @mdev: DRBD device. | |
460 | * @os: old (current) state. | |
461 | * @ns: new (wanted) state. | |
462 | */ | |
463 | static int cl_wide_st_chg(struct drbd_conf *mdev, | |
464 | union drbd_state os, union drbd_state ns) | |
465 | { | |
466 | return (os.conn >= C_CONNECTED && ns.conn >= C_CONNECTED && | |
467 | ((os.role != R_PRIMARY && ns.role == R_PRIMARY) || | |
468 | (os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
469 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) || | |
470 | (os.disk != D_DISKLESS && ns.disk == D_DISKLESS))) || | |
471 | (os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) || | |
472 | (os.conn == C_CONNECTED && ns.conn == C_VERIFY_S); | |
473 | } | |
474 | ||
475 | int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f, | |
476 | union drbd_state mask, union drbd_state val) | |
477 | { | |
478 | unsigned long flags; | |
479 | union drbd_state os, ns; | |
480 | int rv; | |
481 | ||
482 | spin_lock_irqsave(&mdev->req_lock, flags); | |
483 | os = mdev->state; | |
484 | ns.i = (os.i & ~mask.i) | val.i; | |
485 | rv = _drbd_set_state(mdev, ns, f, NULL); | |
486 | ns = mdev->state; | |
487 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
488 | ||
489 | return rv; | |
490 | } | |
491 | ||
492 | /** | |
493 | * drbd_force_state() - Impose a change which happens outside our control on our state | |
494 | * @mdev: DRBD device. | |
495 | * @mask: mask of state bits to change. | |
496 | * @val: value of new state bits. | |
497 | */ | |
498 | void drbd_force_state(struct drbd_conf *mdev, | |
499 | union drbd_state mask, union drbd_state val) | |
500 | { | |
501 | drbd_change_state(mdev, CS_HARD, mask, val); | |
502 | } | |
503 | ||
504 | static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns); | |
505 | static int is_valid_state_transition(struct drbd_conf *, | |
506 | union drbd_state, union drbd_state); | |
507 | static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os, | |
02bc7174 | 508 | union drbd_state ns, const char **warn_sync_abort); |
b411b363 PR |
509 | int drbd_send_state_req(struct drbd_conf *, |
510 | union drbd_state, union drbd_state); | |
511 | ||
512 | static enum drbd_state_ret_codes _req_st_cond(struct drbd_conf *mdev, | |
513 | union drbd_state mask, union drbd_state val) | |
514 | { | |
515 | union drbd_state os, ns; | |
516 | unsigned long flags; | |
517 | int rv; | |
518 | ||
519 | if (test_and_clear_bit(CL_ST_CHG_SUCCESS, &mdev->flags)) | |
520 | return SS_CW_SUCCESS; | |
521 | ||
522 | if (test_and_clear_bit(CL_ST_CHG_FAIL, &mdev->flags)) | |
523 | return SS_CW_FAILED_BY_PEER; | |
524 | ||
525 | rv = 0; | |
526 | spin_lock_irqsave(&mdev->req_lock, flags); | |
527 | os = mdev->state; | |
528 | ns.i = (os.i & ~mask.i) | val.i; | |
529 | ns = sanitize_state(mdev, os, ns, NULL); | |
530 | ||
531 | if (!cl_wide_st_chg(mdev, os, ns)) | |
532 | rv = SS_CW_NO_NEED; | |
533 | if (!rv) { | |
534 | rv = is_valid_state(mdev, ns); | |
535 | if (rv == SS_SUCCESS) { | |
536 | rv = is_valid_state_transition(mdev, ns, os); | |
537 | if (rv == SS_SUCCESS) | |
538 | rv = 0; /* cont waiting, otherwise fail. */ | |
539 | } | |
540 | } | |
541 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
542 | ||
543 | return rv; | |
544 | } | |
545 | ||
546 | /** | |
547 | * drbd_req_state() - Perform an eventually cluster wide state change | |
548 | * @mdev: DRBD device. | |
549 | * @mask: mask of state bits to change. | |
550 | * @val: value of new state bits. | |
551 | * @f: flags | |
552 | * | |
553 | * Should not be called directly, use drbd_request_state() or | |
554 | * _drbd_request_state(). | |
555 | */ | |
556 | static int drbd_req_state(struct drbd_conf *mdev, | |
557 | union drbd_state mask, union drbd_state val, | |
558 | enum chg_state_flags f) | |
559 | { | |
560 | struct completion done; | |
561 | unsigned long flags; | |
562 | union drbd_state os, ns; | |
563 | int rv; | |
564 | ||
565 | init_completion(&done); | |
566 | ||
567 | if (f & CS_SERIALIZE) | |
568 | mutex_lock(&mdev->state_mutex); | |
569 | ||
570 | spin_lock_irqsave(&mdev->req_lock, flags); | |
571 | os = mdev->state; | |
572 | ns.i = (os.i & ~mask.i) | val.i; | |
573 | ns = sanitize_state(mdev, os, ns, NULL); | |
574 | ||
575 | if (cl_wide_st_chg(mdev, os, ns)) { | |
576 | rv = is_valid_state(mdev, ns); | |
577 | if (rv == SS_SUCCESS) | |
578 | rv = is_valid_state_transition(mdev, ns, os); | |
579 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
580 | ||
581 | if (rv < SS_SUCCESS) { | |
582 | if (f & CS_VERBOSE) | |
583 | print_st_err(mdev, os, ns, rv); | |
584 | goto abort; | |
585 | } | |
586 | ||
587 | drbd_state_lock(mdev); | |
588 | if (!drbd_send_state_req(mdev, mask, val)) { | |
589 | drbd_state_unlock(mdev); | |
590 | rv = SS_CW_FAILED_BY_PEER; | |
591 | if (f & CS_VERBOSE) | |
592 | print_st_err(mdev, os, ns, rv); | |
593 | goto abort; | |
594 | } | |
595 | ||
596 | wait_event(mdev->state_wait, | |
597 | (rv = _req_st_cond(mdev, mask, val))); | |
598 | ||
599 | if (rv < SS_SUCCESS) { | |
600 | drbd_state_unlock(mdev); | |
601 | if (f & CS_VERBOSE) | |
602 | print_st_err(mdev, os, ns, rv); | |
603 | goto abort; | |
604 | } | |
605 | spin_lock_irqsave(&mdev->req_lock, flags); | |
606 | os = mdev->state; | |
607 | ns.i = (os.i & ~mask.i) | val.i; | |
608 | rv = _drbd_set_state(mdev, ns, f, &done); | |
609 | drbd_state_unlock(mdev); | |
610 | } else { | |
611 | rv = _drbd_set_state(mdev, ns, f, &done); | |
612 | } | |
613 | ||
614 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
615 | ||
616 | if (f & CS_WAIT_COMPLETE && rv == SS_SUCCESS) { | |
617 | D_ASSERT(current != mdev->worker.task); | |
618 | wait_for_completion(&done); | |
619 | } | |
620 | ||
621 | abort: | |
622 | if (f & CS_SERIALIZE) | |
623 | mutex_unlock(&mdev->state_mutex); | |
624 | ||
625 | return rv; | |
626 | } | |
627 | ||
628 | /** | |
629 | * _drbd_request_state() - Request a state change (with flags) | |
630 | * @mdev: DRBD device. | |
631 | * @mask: mask of state bits to change. | |
632 | * @val: value of new state bits. | |
633 | * @f: flags | |
634 | * | |
635 | * Cousin of drbd_request_state(), useful with the CS_WAIT_COMPLETE | |
636 | * flag, or when logging of failed state change requests is not desired. | |
637 | */ | |
638 | int _drbd_request_state(struct drbd_conf *mdev, union drbd_state mask, | |
639 | union drbd_state val, enum chg_state_flags f) | |
640 | { | |
641 | int rv; | |
642 | ||
643 | wait_event(mdev->state_wait, | |
644 | (rv = drbd_req_state(mdev, mask, val, f)) != SS_IN_TRANSIENT_STATE); | |
645 | ||
646 | return rv; | |
647 | } | |
648 | ||
649 | static void print_st(struct drbd_conf *mdev, char *name, union drbd_state ns) | |
650 | { | |
651 | dev_err(DEV, " %s = { cs:%s ro:%s/%s ds:%s/%s %c%c%c%c }\n", | |
652 | name, | |
653 | drbd_conn_str(ns.conn), | |
654 | drbd_role_str(ns.role), | |
655 | drbd_role_str(ns.peer), | |
656 | drbd_disk_str(ns.disk), | |
657 | drbd_disk_str(ns.pdsk), | |
fb22c402 | 658 | is_susp(ns) ? 's' : 'r', |
b411b363 PR |
659 | ns.aftr_isp ? 'a' : '-', |
660 | ns.peer_isp ? 'p' : '-', | |
661 | ns.user_isp ? 'u' : '-' | |
662 | ); | |
663 | } | |
664 | ||
665 | void print_st_err(struct drbd_conf *mdev, | |
666 | union drbd_state os, union drbd_state ns, int err) | |
667 | { | |
668 | if (err == SS_IN_TRANSIENT_STATE) | |
669 | return; | |
670 | dev_err(DEV, "State change failed: %s\n", drbd_set_st_err_str(err)); | |
671 | print_st(mdev, " state", os); | |
672 | print_st(mdev, "wanted", ns); | |
673 | } | |
674 | ||
675 | ||
676 | #define drbd_peer_str drbd_role_str | |
677 | #define drbd_pdsk_str drbd_disk_str | |
678 | ||
679 | #define drbd_susp_str(A) ((A) ? "1" : "0") | |
680 | #define drbd_aftr_isp_str(A) ((A) ? "1" : "0") | |
681 | #define drbd_peer_isp_str(A) ((A) ? "1" : "0") | |
682 | #define drbd_user_isp_str(A) ((A) ? "1" : "0") | |
683 | ||
684 | #define PSC(A) \ | |
685 | ({ if (ns.A != os.A) { \ | |
686 | pbp += sprintf(pbp, #A "( %s -> %s ) ", \ | |
687 | drbd_##A##_str(os.A), \ | |
688 | drbd_##A##_str(ns.A)); \ | |
689 | } }) | |
690 | ||
691 | /** | |
692 | * is_valid_state() - Returns an SS_ error code if ns is not valid | |
693 | * @mdev: DRBD device. | |
694 | * @ns: State to consider. | |
695 | */ | |
696 | static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns) | |
697 | { | |
698 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
699 | ||
700 | enum drbd_fencing_p fp; | |
701 | int rv = SS_SUCCESS; | |
702 | ||
703 | fp = FP_DONT_CARE; | |
704 | if (get_ldev(mdev)) { | |
705 | fp = mdev->ldev->dc.fencing; | |
706 | put_ldev(mdev); | |
707 | } | |
708 | ||
709 | if (get_net_conf(mdev)) { | |
710 | if (!mdev->net_conf->two_primaries && | |
711 | ns.role == R_PRIMARY && ns.peer == R_PRIMARY) | |
712 | rv = SS_TWO_PRIMARIES; | |
713 | put_net_conf(mdev); | |
714 | } | |
715 | ||
716 | if (rv <= 0) | |
717 | /* already found a reason to abort */; | |
718 | else if (ns.role == R_SECONDARY && mdev->open_cnt) | |
719 | rv = SS_DEVICE_IN_USE; | |
720 | ||
721 | else if (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.disk < D_UP_TO_DATE) | |
722 | rv = SS_NO_UP_TO_DATE_DISK; | |
723 | ||
724 | else if (fp >= FP_RESOURCE && | |
725 | ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk >= D_UNKNOWN) | |
726 | rv = SS_PRIMARY_NOP; | |
727 | ||
728 | else if (ns.role == R_PRIMARY && ns.disk <= D_INCONSISTENT && ns.pdsk <= D_INCONSISTENT) | |
729 | rv = SS_NO_UP_TO_DATE_DISK; | |
730 | ||
731 | else if (ns.conn > C_CONNECTED && ns.disk < D_INCONSISTENT) | |
732 | rv = SS_NO_LOCAL_DISK; | |
733 | ||
734 | else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT) | |
735 | rv = SS_NO_REMOTE_DISK; | |
736 | ||
8d4ce82b LE |
737 | else if (ns.conn > C_CONNECTED && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) |
738 | rv = SS_NO_UP_TO_DATE_DISK; | |
739 | ||
b411b363 PR |
740 | else if ((ns.conn == C_CONNECTED || |
741 | ns.conn == C_WF_BITMAP_S || | |
742 | ns.conn == C_SYNC_SOURCE || | |
743 | ns.conn == C_PAUSED_SYNC_S) && | |
744 | ns.disk == D_OUTDATED) | |
745 | rv = SS_CONNECTED_OUTDATES; | |
746 | ||
747 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
748 | (mdev->sync_conf.verify_alg[0] == 0)) | |
749 | rv = SS_NO_VERIFY_ALG; | |
750 | ||
751 | else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
752 | mdev->agreed_pro_version < 88) | |
753 | rv = SS_NOT_SUPPORTED; | |
754 | ||
755 | return rv; | |
756 | } | |
757 | ||
758 | /** | |
759 | * is_valid_state_transition() - Returns an SS_ error code if the state transition is not possible | |
760 | * @mdev: DRBD device. | |
761 | * @ns: new state. | |
762 | * @os: old state. | |
763 | */ | |
764 | static int is_valid_state_transition(struct drbd_conf *mdev, | |
765 | union drbd_state ns, union drbd_state os) | |
766 | { | |
767 | int rv = SS_SUCCESS; | |
768 | ||
769 | if ((ns.conn == C_STARTING_SYNC_T || ns.conn == C_STARTING_SYNC_S) && | |
770 | os.conn > C_CONNECTED) | |
771 | rv = SS_RESYNC_RUNNING; | |
772 | ||
773 | if (ns.conn == C_DISCONNECTING && os.conn == C_STANDALONE) | |
774 | rv = SS_ALREADY_STANDALONE; | |
775 | ||
776 | if (ns.disk > D_ATTACHING && os.disk == D_DISKLESS) | |
777 | rv = SS_IS_DISKLESS; | |
778 | ||
779 | if (ns.conn == C_WF_CONNECTION && os.conn < C_UNCONNECTED) | |
780 | rv = SS_NO_NET_CONFIG; | |
781 | ||
782 | if (ns.disk == D_OUTDATED && os.disk < D_OUTDATED && os.disk != D_ATTACHING) | |
783 | rv = SS_LOWER_THAN_OUTDATED; | |
784 | ||
785 | if (ns.conn == C_DISCONNECTING && os.conn == C_UNCONNECTED) | |
786 | rv = SS_IN_TRANSIENT_STATE; | |
787 | ||
788 | if (ns.conn == os.conn && ns.conn == C_WF_REPORT_PARAMS) | |
789 | rv = SS_IN_TRANSIENT_STATE; | |
790 | ||
791 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED) | |
792 | rv = SS_NEED_CONNECTION; | |
793 | ||
794 | if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && | |
795 | ns.conn != os.conn && os.conn > C_CONNECTED) | |
796 | rv = SS_RESYNC_RUNNING; | |
797 | ||
798 | if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) && | |
799 | os.conn < C_CONNECTED) | |
800 | rv = SS_NEED_CONNECTION; | |
801 | ||
802 | return rv; | |
803 | } | |
804 | ||
805 | /** | |
806 | * sanitize_state() - Resolves implicitly necessary additional changes to a state transition | |
807 | * @mdev: DRBD device. | |
808 | * @os: old state. | |
809 | * @ns: new state. | |
810 | * @warn_sync_abort: | |
811 | * | |
812 | * When we loose connection, we have to set the state of the peers disk (pdsk) | |
813 | * to D_UNKNOWN. This rule and many more along those lines are in this function. | |
814 | */ | |
815 | static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os, | |
02bc7174 | 816 | union drbd_state ns, const char **warn_sync_abort) |
b411b363 PR |
817 | { |
818 | enum drbd_fencing_p fp; | |
819 | ||
820 | fp = FP_DONT_CARE; | |
821 | if (get_ldev(mdev)) { | |
822 | fp = mdev->ldev->dc.fencing; | |
823 | put_ldev(mdev); | |
824 | } | |
825 | ||
826 | /* Disallow Network errors to configure a device's network part */ | |
827 | if ((ns.conn >= C_TIMEOUT && ns.conn <= C_TEAR_DOWN) && | |
828 | os.conn <= C_DISCONNECTING) | |
829 | ns.conn = os.conn; | |
830 | ||
f2906e18 LE |
831 | /* After a network error (+C_TEAR_DOWN) only C_UNCONNECTED or C_DISCONNECTING can follow. |
832 | * If you try to go into some Sync* state, that shall fail (elsewhere). */ | |
b411b363 | 833 | if (os.conn >= C_TIMEOUT && os.conn <= C_TEAR_DOWN && |
f2906e18 | 834 | ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING && ns.conn <= C_TEAR_DOWN) |
b411b363 PR |
835 | ns.conn = os.conn; |
836 | ||
837 | /* After C_DISCONNECTING only C_STANDALONE may follow */ | |
838 | if (os.conn == C_DISCONNECTING && ns.conn != C_STANDALONE) | |
839 | ns.conn = os.conn; | |
840 | ||
841 | if (ns.conn < C_CONNECTED) { | |
842 | ns.peer_isp = 0; | |
843 | ns.peer = R_UNKNOWN; | |
844 | if (ns.pdsk > D_UNKNOWN || ns.pdsk < D_INCONSISTENT) | |
845 | ns.pdsk = D_UNKNOWN; | |
846 | } | |
847 | ||
848 | /* Clear the aftr_isp when becoming unconfigured */ | |
849 | if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY) | |
850 | ns.aftr_isp = 0; | |
851 | ||
b411b363 PR |
852 | /* Abort resync if a disk fails/detaches */ |
853 | if (os.conn > C_CONNECTED && ns.conn > C_CONNECTED && | |
854 | (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) { | |
855 | if (warn_sync_abort) | |
02bc7174 LE |
856 | *warn_sync_abort = |
857 | os.conn == C_VERIFY_S || os.conn == C_VERIFY_T ? | |
858 | "Online-verify" : "Resync"; | |
b411b363 PR |
859 | ns.conn = C_CONNECTED; |
860 | } | |
861 | ||
862 | if (ns.conn >= C_CONNECTED && | |
863 | ((ns.disk == D_CONSISTENT || ns.disk == D_OUTDATED) || | |
864 | (ns.disk == D_NEGOTIATING && ns.conn == C_WF_BITMAP_T))) { | |
865 | switch (ns.conn) { | |
866 | case C_WF_BITMAP_T: | |
867 | case C_PAUSED_SYNC_T: | |
868 | ns.disk = D_OUTDATED; | |
869 | break; | |
870 | case C_CONNECTED: | |
871 | case C_WF_BITMAP_S: | |
872 | case C_SYNC_SOURCE: | |
873 | case C_PAUSED_SYNC_S: | |
874 | ns.disk = D_UP_TO_DATE; | |
875 | break; | |
876 | case C_SYNC_TARGET: | |
877 | ns.disk = D_INCONSISTENT; | |
878 | dev_warn(DEV, "Implicitly set disk state Inconsistent!\n"); | |
879 | break; | |
880 | } | |
881 | if (os.disk == D_OUTDATED && ns.disk == D_UP_TO_DATE) | |
882 | dev_warn(DEV, "Implicitly set disk from Outdated to UpToDate\n"); | |
883 | } | |
884 | ||
885 | if (ns.conn >= C_CONNECTED && | |
886 | (ns.pdsk == D_CONSISTENT || ns.pdsk == D_OUTDATED)) { | |
887 | switch (ns.conn) { | |
888 | case C_CONNECTED: | |
889 | case C_WF_BITMAP_T: | |
890 | case C_PAUSED_SYNC_T: | |
891 | case C_SYNC_TARGET: | |
892 | ns.pdsk = D_UP_TO_DATE; | |
893 | break; | |
894 | case C_WF_BITMAP_S: | |
895 | case C_PAUSED_SYNC_S: | |
e0f83012 LE |
896 | /* remap any consistent state to D_OUTDATED, |
897 | * but disallow "upgrade" of not even consistent states. | |
898 | */ | |
899 | ns.pdsk = | |
900 | (D_DISKLESS < os.pdsk && os.pdsk < D_OUTDATED) | |
901 | ? os.pdsk : D_OUTDATED; | |
b411b363 PR |
902 | break; |
903 | case C_SYNC_SOURCE: | |
904 | ns.pdsk = D_INCONSISTENT; | |
905 | dev_warn(DEV, "Implicitly set pdsk Inconsistent!\n"); | |
906 | break; | |
907 | } | |
908 | if (os.pdsk == D_OUTDATED && ns.pdsk == D_UP_TO_DATE) | |
909 | dev_warn(DEV, "Implicitly set pdsk from Outdated to UpToDate\n"); | |
910 | } | |
911 | ||
912 | /* Connection breaks down before we finished "Negotiating" */ | |
913 | if (ns.conn < C_CONNECTED && ns.disk == D_NEGOTIATING && | |
914 | get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
915 | if (mdev->ed_uuid == mdev->ldev->md.uuid[UI_CURRENT]) { | |
916 | ns.disk = mdev->new_state_tmp.disk; | |
917 | ns.pdsk = mdev->new_state_tmp.pdsk; | |
918 | } else { | |
919 | dev_alert(DEV, "Connection lost while negotiating, no data!\n"); | |
920 | ns.disk = D_DISKLESS; | |
921 | ns.pdsk = D_UNKNOWN; | |
922 | } | |
923 | put_ldev(mdev); | |
924 | } | |
925 | ||
926 | if (fp == FP_STONITH && | |
0a492166 PR |
927 | (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED) && |
928 | !(os.role == R_PRIMARY && os.conn < C_CONNECTED && os.pdsk > D_OUTDATED)) | |
fb22c402 | 929 | ns.susp_fen = 1; /* Suspend IO while fence-peer handler runs (peer lost) */ |
265be2d0 PR |
930 | |
931 | if (mdev->sync_conf.on_no_data == OND_SUSPEND_IO && | |
932 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE) && | |
933 | !(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE)) | |
fb22c402 | 934 | ns.susp_nod = 1; /* Suspend IO while no data available (no accessible data available) */ |
b411b363 PR |
935 | |
936 | if (ns.aftr_isp || ns.peer_isp || ns.user_isp) { | |
937 | if (ns.conn == C_SYNC_SOURCE) | |
938 | ns.conn = C_PAUSED_SYNC_S; | |
939 | if (ns.conn == C_SYNC_TARGET) | |
940 | ns.conn = C_PAUSED_SYNC_T; | |
941 | } else { | |
942 | if (ns.conn == C_PAUSED_SYNC_S) | |
943 | ns.conn = C_SYNC_SOURCE; | |
944 | if (ns.conn == C_PAUSED_SYNC_T) | |
945 | ns.conn = C_SYNC_TARGET; | |
946 | } | |
947 | ||
948 | return ns; | |
949 | } | |
950 | ||
951 | /* helper for __drbd_set_state */ | |
952 | static void set_ov_position(struct drbd_conf *mdev, enum drbd_conns cs) | |
953 | { | |
954 | if (cs == C_VERIFY_T) { | |
955 | /* starting online verify from an arbitrary position | |
956 | * does not fit well into the existing protocol. | |
957 | * on C_VERIFY_T, we initialize ov_left and friends | |
958 | * implicitly in receive_DataRequest once the | |
959 | * first P_OV_REQUEST is received */ | |
960 | mdev->ov_start_sector = ~(sector_t)0; | |
961 | } else { | |
962 | unsigned long bit = BM_SECT_TO_BIT(mdev->ov_start_sector); | |
963 | if (bit >= mdev->rs_total) | |
964 | mdev->ov_start_sector = | |
965 | BM_BIT_TO_SECT(mdev->rs_total - 1); | |
966 | mdev->ov_position = mdev->ov_start_sector; | |
967 | } | |
968 | } | |
969 | ||
0778286a PR |
970 | static void drbd_resume_al(struct drbd_conf *mdev) |
971 | { | |
972 | if (test_and_clear_bit(AL_SUSPENDED, &mdev->flags)) | |
973 | dev_info(DEV, "Resumed AL updates\n"); | |
974 | } | |
975 | ||
b411b363 PR |
976 | /** |
977 | * __drbd_set_state() - Set a new DRBD state | |
978 | * @mdev: DRBD device. | |
979 | * @ns: new state. | |
980 | * @flags: Flags | |
981 | * @done: Optional completion, that will get completed after the after_state_ch() finished | |
982 | * | |
983 | * Caller needs to hold req_lock, and global_state_lock. Do not call directly. | |
984 | */ | |
985 | int __drbd_set_state(struct drbd_conf *mdev, | |
986 | union drbd_state ns, enum chg_state_flags flags, | |
987 | struct completion *done) | |
988 | { | |
989 | union drbd_state os; | |
990 | int rv = SS_SUCCESS; | |
02bc7174 | 991 | const char *warn_sync_abort = NULL; |
b411b363 PR |
992 | struct after_state_chg_work *ascw; |
993 | ||
994 | os = mdev->state; | |
995 | ||
996 | ns = sanitize_state(mdev, os, ns, &warn_sync_abort); | |
997 | ||
998 | if (ns.i == os.i) | |
999 | return SS_NOTHING_TO_DO; | |
1000 | ||
1001 | if (!(flags & CS_HARD)) { | |
1002 | /* pre-state-change checks ; only look at ns */ | |
1003 | /* See drbd_state_sw_errors in drbd_strings.c */ | |
1004 | ||
1005 | rv = is_valid_state(mdev, ns); | |
1006 | if (rv < SS_SUCCESS) { | |
1007 | /* If the old state was illegal as well, then let | |
1008 | this happen...*/ | |
1009 | ||
1616a254 | 1010 | if (is_valid_state(mdev, os) == rv) |
b411b363 | 1011 | rv = is_valid_state_transition(mdev, ns, os); |
b411b363 PR |
1012 | } else |
1013 | rv = is_valid_state_transition(mdev, ns, os); | |
1014 | } | |
1015 | ||
1016 | if (rv < SS_SUCCESS) { | |
1017 | if (flags & CS_VERBOSE) | |
1018 | print_st_err(mdev, os, ns, rv); | |
1019 | return rv; | |
1020 | } | |
1021 | ||
1022 | if (warn_sync_abort) | |
02bc7174 | 1023 | dev_warn(DEV, "%s aborted.\n", warn_sync_abort); |
b411b363 PR |
1024 | |
1025 | { | |
1026 | char *pbp, pb[300]; | |
1027 | pbp = pb; | |
1028 | *pbp = 0; | |
1029 | PSC(role); | |
1030 | PSC(peer); | |
1031 | PSC(conn); | |
1032 | PSC(disk); | |
1033 | PSC(pdsk); | |
fb22c402 PR |
1034 | if (is_susp(ns) != is_susp(os)) |
1035 | pbp += sprintf(pbp, "susp( %s -> %s ) ", | |
1036 | drbd_susp_str(is_susp(os)), | |
1037 | drbd_susp_str(is_susp(ns))); | |
b411b363 PR |
1038 | PSC(aftr_isp); |
1039 | PSC(peer_isp); | |
1040 | PSC(user_isp); | |
1041 | dev_info(DEV, "%s\n", pb); | |
1042 | } | |
1043 | ||
1044 | /* solve the race between becoming unconfigured, | |
1045 | * worker doing the cleanup, and | |
1046 | * admin reconfiguring us: | |
1047 | * on (re)configure, first set CONFIG_PENDING, | |
1048 | * then wait for a potentially exiting worker, | |
1049 | * start the worker, and schedule one no_op. | |
1050 | * then proceed with configuration. | |
1051 | */ | |
1052 | if (ns.disk == D_DISKLESS && | |
1053 | ns.conn == C_STANDALONE && | |
1054 | ns.role == R_SECONDARY && | |
1055 | !test_and_set_bit(CONFIG_PENDING, &mdev->flags)) | |
1056 | set_bit(DEVICE_DYING, &mdev->flags); | |
1057 | ||
1058 | mdev->state.i = ns.i; | |
1059 | wake_up(&mdev->misc_wait); | |
1060 | wake_up(&mdev->state_wait); | |
1061 | ||
b411b363 PR |
1062 | /* aborted verify run. log the last position */ |
1063 | if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) && | |
1064 | ns.conn < C_CONNECTED) { | |
1065 | mdev->ov_start_sector = | |
1066 | BM_BIT_TO_SECT(mdev->rs_total - mdev->ov_left); | |
1067 | dev_info(DEV, "Online Verify reached sector %llu\n", | |
1068 | (unsigned long long)mdev->ov_start_sector); | |
1069 | } | |
1070 | ||
1071 | if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) && | |
1072 | (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) { | |
1073 | dev_info(DEV, "Syncer continues.\n"); | |
1d7734a0 LE |
1074 | mdev->rs_paused += (long)jiffies |
1075 | -(long)mdev->rs_mark_time[mdev->rs_last_mark]; | |
63106d3c PR |
1076 | if (ns.conn == C_SYNC_TARGET) |
1077 | mod_timer(&mdev->resync_timer, jiffies); | |
b411b363 PR |
1078 | } |
1079 | ||
1080 | if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) && | |
1081 | (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) { | |
1082 | dev_info(DEV, "Resync suspended\n"); | |
1d7734a0 | 1083 | mdev->rs_mark_time[mdev->rs_last_mark] = jiffies; |
b411b363 PR |
1084 | } |
1085 | ||
1086 | if (os.conn == C_CONNECTED && | |
1087 | (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) { | |
1d7734a0 LE |
1088 | unsigned long now = jiffies; |
1089 | int i; | |
1090 | ||
b411b363 | 1091 | mdev->ov_position = 0; |
1d7734a0 | 1092 | mdev->rs_total = drbd_bm_bits(mdev); |
b411b363 PR |
1093 | if (mdev->agreed_pro_version >= 90) |
1094 | set_ov_position(mdev, ns.conn); | |
1095 | else | |
1096 | mdev->ov_start_sector = 0; | |
1097 | mdev->ov_left = mdev->rs_total | |
1098 | - BM_SECT_TO_BIT(mdev->ov_position); | |
1d7734a0 | 1099 | mdev->rs_start = now; |
0f0601f4 LE |
1100 | mdev->rs_last_events = 0; |
1101 | mdev->rs_last_sect_ev = 0; | |
b411b363 PR |
1102 | mdev->ov_last_oos_size = 0; |
1103 | mdev->ov_last_oos_start = 0; | |
1104 | ||
1d7734a0 LE |
1105 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { |
1106 | mdev->rs_mark_left[i] = mdev->rs_total; | |
1107 | mdev->rs_mark_time[i] = now; | |
1108 | } | |
1109 | ||
b411b363 PR |
1110 | if (ns.conn == C_VERIFY_S) { |
1111 | dev_info(DEV, "Starting Online Verify from sector %llu\n", | |
1112 | (unsigned long long)mdev->ov_position); | |
1113 | mod_timer(&mdev->resync_timer, jiffies); | |
1114 | } | |
1115 | } | |
1116 | ||
1117 | if (get_ldev(mdev)) { | |
1118 | u32 mdf = mdev->ldev->md.flags & ~(MDF_CONSISTENT|MDF_PRIMARY_IND| | |
1119 | MDF_CONNECTED_IND|MDF_WAS_UP_TO_DATE| | |
1120 | MDF_PEER_OUT_DATED|MDF_CRASHED_PRIMARY); | |
1121 | ||
1122 | if (test_bit(CRASHED_PRIMARY, &mdev->flags)) | |
1123 | mdf |= MDF_CRASHED_PRIMARY; | |
1124 | if (mdev->state.role == R_PRIMARY || | |
1125 | (mdev->state.pdsk < D_INCONSISTENT && mdev->state.peer == R_PRIMARY)) | |
1126 | mdf |= MDF_PRIMARY_IND; | |
1127 | if (mdev->state.conn > C_WF_REPORT_PARAMS) | |
1128 | mdf |= MDF_CONNECTED_IND; | |
1129 | if (mdev->state.disk > D_INCONSISTENT) | |
1130 | mdf |= MDF_CONSISTENT; | |
1131 | if (mdev->state.disk > D_OUTDATED) | |
1132 | mdf |= MDF_WAS_UP_TO_DATE; | |
1133 | if (mdev->state.pdsk <= D_OUTDATED && mdev->state.pdsk >= D_INCONSISTENT) | |
1134 | mdf |= MDF_PEER_OUT_DATED; | |
1135 | if (mdf != mdev->ldev->md.flags) { | |
1136 | mdev->ldev->md.flags = mdf; | |
1137 | drbd_md_mark_dirty(mdev); | |
1138 | } | |
1139 | if (os.disk < D_CONSISTENT && ns.disk >= D_CONSISTENT) | |
1140 | drbd_set_ed_uuid(mdev, mdev->ldev->md.uuid[UI_CURRENT]); | |
1141 | put_ldev(mdev); | |
1142 | } | |
1143 | ||
1144 | /* Peer was forced D_UP_TO_DATE & R_PRIMARY, consider to resync */ | |
1145 | if (os.disk == D_INCONSISTENT && os.pdsk == D_INCONSISTENT && | |
1146 | os.peer == R_SECONDARY && ns.peer == R_PRIMARY) | |
1147 | set_bit(CONSIDER_RESYNC, &mdev->flags); | |
1148 | ||
1149 | /* Receiver should clean up itself */ | |
1150 | if (os.conn != C_DISCONNECTING && ns.conn == C_DISCONNECTING) | |
1151 | drbd_thread_stop_nowait(&mdev->receiver); | |
1152 | ||
1153 | /* Now the receiver finished cleaning up itself, it should die */ | |
1154 | if (os.conn != C_STANDALONE && ns.conn == C_STANDALONE) | |
1155 | drbd_thread_stop_nowait(&mdev->receiver); | |
1156 | ||
1157 | /* Upon network failure, we need to restart the receiver. */ | |
1158 | if (os.conn > C_TEAR_DOWN && | |
1159 | ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT) | |
1160 | drbd_thread_restart_nowait(&mdev->receiver); | |
1161 | ||
0778286a PR |
1162 | /* Resume AL writing if we get a connection */ |
1163 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) | |
1164 | drbd_resume_al(mdev); | |
1165 | ||
b411b363 PR |
1166 | ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC); |
1167 | if (ascw) { | |
1168 | ascw->os = os; | |
1169 | ascw->ns = ns; | |
1170 | ascw->flags = flags; | |
1171 | ascw->w.cb = w_after_state_ch; | |
1172 | ascw->done = done; | |
1173 | drbd_queue_work(&mdev->data.work, &ascw->w); | |
1174 | } else { | |
1175 | dev_warn(DEV, "Could not kmalloc an ascw\n"); | |
1176 | } | |
1177 | ||
1178 | return rv; | |
1179 | } | |
1180 | ||
1181 | static int w_after_state_ch(struct drbd_conf *mdev, struct drbd_work *w, int unused) | |
1182 | { | |
1183 | struct after_state_chg_work *ascw = | |
1184 | container_of(w, struct after_state_chg_work, w); | |
1185 | after_state_ch(mdev, ascw->os, ascw->ns, ascw->flags); | |
1186 | if (ascw->flags & CS_WAIT_COMPLETE) { | |
1187 | D_ASSERT(ascw->done != NULL); | |
1188 | complete(ascw->done); | |
1189 | } | |
1190 | kfree(ascw); | |
1191 | ||
1192 | return 1; | |
1193 | } | |
1194 | ||
1195 | static void abw_start_sync(struct drbd_conf *mdev, int rv) | |
1196 | { | |
1197 | if (rv) { | |
1198 | dev_err(DEV, "Writing the bitmap failed not starting resync.\n"); | |
1199 | _drbd_request_state(mdev, NS(conn, C_CONNECTED), CS_VERBOSE); | |
1200 | return; | |
1201 | } | |
1202 | ||
1203 | switch (mdev->state.conn) { | |
1204 | case C_STARTING_SYNC_T: | |
1205 | _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); | |
1206 | break; | |
1207 | case C_STARTING_SYNC_S: | |
1208 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
1209 | break; | |
1210 | } | |
1211 | } | |
1212 | ||
1213 | /** | |
1214 | * after_state_ch() - Perform after state change actions that may sleep | |
1215 | * @mdev: DRBD device. | |
1216 | * @os: old state. | |
1217 | * @ns: new state. | |
1218 | * @flags: Flags | |
1219 | */ | |
1220 | static void after_state_ch(struct drbd_conf *mdev, union drbd_state os, | |
1221 | union drbd_state ns, enum chg_state_flags flags) | |
1222 | { | |
1223 | enum drbd_fencing_p fp; | |
67098930 | 1224 | enum drbd_req_event what = nothing; |
fb22c402 | 1225 | union drbd_state nsm = (union drbd_state){ .i = -1 }; |
b411b363 PR |
1226 | |
1227 | if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) { | |
1228 | clear_bit(CRASHED_PRIMARY, &mdev->flags); | |
1229 | if (mdev->p_uuid) | |
1230 | mdev->p_uuid[UI_FLAGS] &= ~((u64)2); | |
1231 | } | |
1232 | ||
1233 | fp = FP_DONT_CARE; | |
1234 | if (get_ldev(mdev)) { | |
1235 | fp = mdev->ldev->dc.fencing; | |
1236 | put_ldev(mdev); | |
1237 | } | |
1238 | ||
1239 | /* Inform userspace about the change... */ | |
1240 | drbd_bcast_state(mdev, ns); | |
1241 | ||
1242 | if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) && | |
1243 | (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE)) | |
1244 | drbd_khelper(mdev, "pri-on-incon-degr"); | |
1245 | ||
1246 | /* Here we have the actions that are performed after a | |
1247 | state change. This function might sleep */ | |
1248 | ||
fb22c402 PR |
1249 | nsm.i = -1; |
1250 | if (ns.susp_nod) { | |
265be2d0 | 1251 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { |
67098930 | 1252 | if (ns.conn == C_CONNECTED) |
fb22c402 | 1253 | what = resend, nsm.susp_nod = 0; |
67098930 | 1254 | else /* ns.conn > C_CONNECTED */ |
265be2d0 PR |
1255 | dev_err(DEV, "Unexpected Resynd going on!\n"); |
1256 | } | |
1257 | ||
67098930 | 1258 | if (os.disk == D_ATTACHING && ns.disk > D_ATTACHING) |
fb22c402 PR |
1259 | what = restart_frozen_disk_io, nsm.susp_nod = 0; |
1260 | ||
265be2d0 PR |
1261 | } |
1262 | ||
fb22c402 | 1263 | if (ns.susp_fen) { |
43a5182c PR |
1264 | /* case1: The outdate peer handler is successful: */ |
1265 | if (os.pdsk > D_OUTDATED && ns.pdsk <= D_OUTDATED) { | |
b411b363 | 1266 | tl_clear(mdev); |
43a5182c PR |
1267 | if (test_bit(NEW_CUR_UUID, &mdev->flags)) { |
1268 | drbd_uuid_new_current(mdev); | |
1269 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
1270 | drbd_md_sync(mdev); | |
1271 | } | |
b411b363 | 1272 | spin_lock_irq(&mdev->req_lock); |
fb22c402 | 1273 | _drbd_set_state(_NS(mdev, susp_fen, 0), CS_VERBOSE, NULL); |
b411b363 PR |
1274 | spin_unlock_irq(&mdev->req_lock); |
1275 | } | |
43a5182c PR |
1276 | /* case2: The connection was established again: */ |
1277 | if (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED) { | |
1278 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
67098930 | 1279 | what = resend; |
fb22c402 | 1280 | nsm.susp_fen = 0; |
43a5182c | 1281 | } |
b411b363 | 1282 | } |
67098930 PR |
1283 | |
1284 | if (what != nothing) { | |
1285 | spin_lock_irq(&mdev->req_lock); | |
1286 | _tl_restart(mdev, what); | |
fb22c402 PR |
1287 | nsm.i &= mdev->state.i; |
1288 | _drbd_set_state(mdev, nsm, CS_VERBOSE, NULL); | |
67098930 PR |
1289 | spin_unlock_irq(&mdev->req_lock); |
1290 | } | |
1291 | ||
b411b363 PR |
1292 | /* Do not change the order of the if above and the two below... */ |
1293 | if (os.pdsk == D_DISKLESS && ns.pdsk > D_DISKLESS) { /* attach on the peer */ | |
1294 | drbd_send_uuids(mdev); | |
1295 | drbd_send_state(mdev); | |
1296 | } | |
1297 | if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S) | |
1298 | drbd_queue_bitmap_io(mdev, &drbd_send_bitmap, NULL, "send_bitmap (WFBitMapS)"); | |
1299 | ||
1300 | /* Lost contact to peer's copy of the data */ | |
1301 | if ((os.pdsk >= D_INCONSISTENT && | |
1302 | os.pdsk != D_UNKNOWN && | |
1303 | os.pdsk != D_OUTDATED) | |
1304 | && (ns.pdsk < D_INCONSISTENT || | |
1305 | ns.pdsk == D_UNKNOWN || | |
1306 | ns.pdsk == D_OUTDATED)) { | |
b411b363 PR |
1307 | if (get_ldev(mdev)) { |
1308 | if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) && | |
2c8d1967 | 1309 | mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) { |
fb22c402 | 1310 | if (is_susp(mdev->state)) { |
43a5182c PR |
1311 | set_bit(NEW_CUR_UUID, &mdev->flags); |
1312 | } else { | |
1313 | drbd_uuid_new_current(mdev); | |
1314 | drbd_send_uuids(mdev); | |
1315 | } | |
2c8d1967 | 1316 | } |
b411b363 PR |
1317 | put_ldev(mdev); |
1318 | } | |
1319 | } | |
1320 | ||
1321 | if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) { | |
18a50fa2 | 1322 | if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0) { |
2c8d1967 | 1323 | drbd_uuid_new_current(mdev); |
18a50fa2 PR |
1324 | drbd_send_uuids(mdev); |
1325 | } | |
b411b363 PR |
1326 | |
1327 | /* D_DISKLESS Peer becomes secondary */ | |
1328 | if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY) | |
1329 | drbd_al_to_on_disk_bm(mdev); | |
1330 | put_ldev(mdev); | |
1331 | } | |
1332 | ||
1333 | /* Last part of the attaching process ... */ | |
1334 | if (ns.conn >= C_CONNECTED && | |
1335 | os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) { | |
e89b591c | 1336 | drbd_send_sizes(mdev, 0, 0); /* to start sync... */ |
b411b363 PR |
1337 | drbd_send_uuids(mdev); |
1338 | drbd_send_state(mdev); | |
1339 | } | |
1340 | ||
1341 | /* We want to pause/continue resync, tell peer. */ | |
1342 | if (ns.conn >= C_CONNECTED && | |
1343 | ((os.aftr_isp != ns.aftr_isp) || | |
1344 | (os.user_isp != ns.user_isp))) | |
1345 | drbd_send_state(mdev); | |
1346 | ||
1347 | /* In case one of the isp bits got set, suspend other devices. */ | |
1348 | if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) && | |
1349 | (ns.aftr_isp || ns.peer_isp || ns.user_isp)) | |
1350 | suspend_other_sg(mdev); | |
1351 | ||
1352 | /* Make sure the peer gets informed about eventual state | |
1353 | changes (ISP bits) while we were in WFReportParams. */ | |
1354 | if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED) | |
1355 | drbd_send_state(mdev); | |
1356 | ||
1357 | /* We are in the progress to start a full sync... */ | |
1358 | if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) || | |
1359 | (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S)) | |
1360 | drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, &abw_start_sync, "set_n_write from StartingSync"); | |
1361 | ||
1362 | /* We are invalidating our self... */ | |
1363 | if (os.conn < C_CONNECTED && ns.conn < C_CONNECTED && | |
1364 | os.disk > D_INCONSISTENT && ns.disk == D_INCONSISTENT) | |
1365 | drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, "set_n_write from invalidate"); | |
1366 | ||
e9e6f3ec | 1367 | /* first half of local IO error */ |
b411b363 | 1368 | if (os.disk > D_FAILED && ns.disk == D_FAILED) { |
e9e6f3ec LE |
1369 | enum drbd_io_error_p eh = EP_PASS_ON; |
1370 | ||
1371 | if (drbd_send_state(mdev)) | |
1372 | dev_warn(DEV, "Notified peer that my disk is broken.\n"); | |
1373 | else | |
1374 | dev_err(DEV, "Sending state for drbd_io_error() failed\n"); | |
1375 | ||
1376 | drbd_rs_cancel_all(mdev); | |
b411b363 | 1377 | |
b411b363 PR |
1378 | if (get_ldev_if_state(mdev, D_FAILED)) { |
1379 | eh = mdev->ldev->dc.on_io_error; | |
1380 | put_ldev(mdev); | |
1381 | } | |
e9e6f3ec LE |
1382 | if (eh == EP_CALL_HELPER) |
1383 | drbd_khelper(mdev, "local-io-error"); | |
1384 | } | |
b411b363 | 1385 | |
e9e6f3ec LE |
1386 | |
1387 | /* second half of local IO error handling, | |
1388 | * after local_cnt references have reached zero: */ | |
1389 | if (os.disk == D_FAILED && ns.disk == D_DISKLESS) { | |
b411b363 PR |
1390 | mdev->rs_total = 0; |
1391 | mdev->rs_failed = 0; | |
1392 | atomic_set(&mdev->rs_pending_cnt, 0); | |
b411b363 PR |
1393 | } |
1394 | ||
1395 | if (os.disk > D_DISKLESS && ns.disk == D_DISKLESS) { | |
9d282875 LE |
1396 | /* We must still be diskless, |
1397 | * re-attach has to be serialized with this! */ | |
1398 | if (mdev->state.disk != D_DISKLESS) | |
1399 | dev_err(DEV, | |
1400 | "ASSERT FAILED: disk is %s while going diskless\n", | |
1401 | drbd_disk_str(mdev->state.disk)); | |
1402 | ||
1403 | /* we cannot assert local_cnt == 0 here, as get_ldev_if_state | |
1404 | * will inc/dec it frequently. Since we became D_DISKLESS, no | |
1405 | * one has touched the protected members anymore, though, so we | |
1406 | * are safe to free them here. */ | |
e9e6f3ec LE |
1407 | if (drbd_send_state(mdev)) |
1408 | dev_warn(DEV, "Notified peer that I detached my disk.\n"); | |
1409 | else | |
1410 | dev_err(DEV, "Sending state for detach failed\n"); | |
b411b363 PR |
1411 | |
1412 | lc_destroy(mdev->resync); | |
1413 | mdev->resync = NULL; | |
1414 | lc_destroy(mdev->act_log); | |
1415 | mdev->act_log = NULL; | |
1416 | __no_warn(local, | |
1417 | drbd_free_bc(mdev->ldev); | |
1418 | mdev->ldev = NULL;); | |
1419 | ||
f65363cf | 1420 | if (mdev->md_io_tmpp) { |
b411b363 | 1421 | __free_page(mdev->md_io_tmpp); |
f65363cf LE |
1422 | mdev->md_io_tmpp = NULL; |
1423 | } | |
b411b363 PR |
1424 | } |
1425 | ||
1426 | /* Disks got bigger while they were detached */ | |
1427 | if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING && | |
1428 | test_and_clear_bit(RESYNC_AFTER_NEG, &mdev->flags)) { | |
1429 | if (ns.conn == C_CONNECTED) | |
1430 | resync_after_online_grow(mdev); | |
1431 | } | |
1432 | ||
1433 | /* A resync finished or aborted, wake paused devices... */ | |
1434 | if ((os.conn > C_CONNECTED && ns.conn <= C_CONNECTED) || | |
1435 | (os.peer_isp && !ns.peer_isp) || | |
1436 | (os.user_isp && !ns.user_isp)) | |
1437 | resume_next_sg(mdev); | |
1438 | ||
af85e8e8 LE |
1439 | /* sync target done with resync. Explicitly notify peer, even though |
1440 | * it should (at least for non-empty resyncs) already know itself. */ | |
1441 | if (os.disk < D_UP_TO_DATE && os.conn >= C_SYNC_SOURCE && ns.conn == C_CONNECTED) | |
1442 | drbd_send_state(mdev); | |
1443 | ||
f70b3511 | 1444 | /* free tl_hash if we Got thawed and are C_STANDALONE */ |
fb22c402 | 1445 | if (ns.conn == C_STANDALONE && !is_susp(ns) && mdev->tl_hash) |
f70b3511 PR |
1446 | drbd_free_tl_hash(mdev); |
1447 | ||
b411b363 PR |
1448 | /* Upon network connection, we need to start the receiver */ |
1449 | if (os.conn == C_STANDALONE && ns.conn == C_UNCONNECTED) | |
1450 | drbd_thread_start(&mdev->receiver); | |
1451 | ||
1452 | /* Terminate worker thread if we are unconfigured - it will be | |
1453 | restarted as needed... */ | |
1454 | if (ns.disk == D_DISKLESS && | |
1455 | ns.conn == C_STANDALONE && | |
1456 | ns.role == R_SECONDARY) { | |
1457 | if (os.aftr_isp != ns.aftr_isp) | |
1458 | resume_next_sg(mdev); | |
1459 | /* set in __drbd_set_state, unless CONFIG_PENDING was set */ | |
1460 | if (test_bit(DEVICE_DYING, &mdev->flags)) | |
1461 | drbd_thread_stop_nowait(&mdev->worker); | |
1462 | } | |
1463 | ||
1464 | drbd_md_sync(mdev); | |
1465 | } | |
1466 | ||
1467 | ||
1468 | static int drbd_thread_setup(void *arg) | |
1469 | { | |
1470 | struct drbd_thread *thi = (struct drbd_thread *) arg; | |
1471 | struct drbd_conf *mdev = thi->mdev; | |
1472 | unsigned long flags; | |
1473 | int retval; | |
1474 | ||
1475 | restart: | |
1476 | retval = thi->function(thi); | |
1477 | ||
1478 | spin_lock_irqsave(&thi->t_lock, flags); | |
1479 | ||
1480 | /* if the receiver has been "Exiting", the last thing it did | |
1481 | * was set the conn state to "StandAlone", | |
1482 | * if now a re-connect request comes in, conn state goes C_UNCONNECTED, | |
1483 | * and receiver thread will be "started". | |
1484 | * drbd_thread_start needs to set "Restarting" in that case. | |
1485 | * t_state check and assignment needs to be within the same spinlock, | |
1486 | * so either thread_start sees Exiting, and can remap to Restarting, | |
1487 | * or thread_start see None, and can proceed as normal. | |
1488 | */ | |
1489 | ||
1490 | if (thi->t_state == Restarting) { | |
1491 | dev_info(DEV, "Restarting %s\n", current->comm); | |
1492 | thi->t_state = Running; | |
1493 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1494 | goto restart; | |
1495 | } | |
1496 | ||
1497 | thi->task = NULL; | |
1498 | thi->t_state = None; | |
1499 | smp_mb(); | |
1500 | complete(&thi->stop); | |
1501 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1502 | ||
1503 | dev_info(DEV, "Terminating %s\n", current->comm); | |
1504 | ||
1505 | /* Release mod reference taken when thread was started */ | |
1506 | module_put(THIS_MODULE); | |
1507 | return retval; | |
1508 | } | |
1509 | ||
1510 | static void drbd_thread_init(struct drbd_conf *mdev, struct drbd_thread *thi, | |
1511 | int (*func) (struct drbd_thread *)) | |
1512 | { | |
1513 | spin_lock_init(&thi->t_lock); | |
1514 | thi->task = NULL; | |
1515 | thi->t_state = None; | |
1516 | thi->function = func; | |
1517 | thi->mdev = mdev; | |
1518 | } | |
1519 | ||
1520 | int drbd_thread_start(struct drbd_thread *thi) | |
1521 | { | |
1522 | struct drbd_conf *mdev = thi->mdev; | |
1523 | struct task_struct *nt; | |
1524 | unsigned long flags; | |
1525 | ||
1526 | const char *me = | |
1527 | thi == &mdev->receiver ? "receiver" : | |
1528 | thi == &mdev->asender ? "asender" : | |
1529 | thi == &mdev->worker ? "worker" : "NONSENSE"; | |
1530 | ||
1531 | /* is used from state engine doing drbd_thread_stop_nowait, | |
1532 | * while holding the req lock irqsave */ | |
1533 | spin_lock_irqsave(&thi->t_lock, flags); | |
1534 | ||
1535 | switch (thi->t_state) { | |
1536 | case None: | |
1537 | dev_info(DEV, "Starting %s thread (from %s [%d])\n", | |
1538 | me, current->comm, current->pid); | |
1539 | ||
1540 | /* Get ref on module for thread - this is released when thread exits */ | |
1541 | if (!try_module_get(THIS_MODULE)) { | |
1542 | dev_err(DEV, "Failed to get module reference in drbd_thread_start\n"); | |
1543 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1544 | return FALSE; | |
1545 | } | |
1546 | ||
1547 | init_completion(&thi->stop); | |
1548 | D_ASSERT(thi->task == NULL); | |
1549 | thi->reset_cpu_mask = 1; | |
1550 | thi->t_state = Running; | |
1551 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1552 | flush_signals(current); /* otherw. may get -ERESTARTNOINTR */ | |
1553 | ||
1554 | nt = kthread_create(drbd_thread_setup, (void *) thi, | |
1555 | "drbd%d_%s", mdev_to_minor(mdev), me); | |
1556 | ||
1557 | if (IS_ERR(nt)) { | |
1558 | dev_err(DEV, "Couldn't start thread\n"); | |
1559 | ||
1560 | module_put(THIS_MODULE); | |
1561 | return FALSE; | |
1562 | } | |
1563 | spin_lock_irqsave(&thi->t_lock, flags); | |
1564 | thi->task = nt; | |
1565 | thi->t_state = Running; | |
1566 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1567 | wake_up_process(nt); | |
1568 | break; | |
1569 | case Exiting: | |
1570 | thi->t_state = Restarting; | |
1571 | dev_info(DEV, "Restarting %s thread (from %s [%d])\n", | |
1572 | me, current->comm, current->pid); | |
1573 | /* fall through */ | |
1574 | case Running: | |
1575 | case Restarting: | |
1576 | default: | |
1577 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1578 | break; | |
1579 | } | |
1580 | ||
1581 | return TRUE; | |
1582 | } | |
1583 | ||
1584 | ||
1585 | void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait) | |
1586 | { | |
1587 | unsigned long flags; | |
1588 | ||
1589 | enum drbd_thread_state ns = restart ? Restarting : Exiting; | |
1590 | ||
1591 | /* may be called from state engine, holding the req lock irqsave */ | |
1592 | spin_lock_irqsave(&thi->t_lock, flags); | |
1593 | ||
1594 | if (thi->t_state == None) { | |
1595 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1596 | if (restart) | |
1597 | drbd_thread_start(thi); | |
1598 | return; | |
1599 | } | |
1600 | ||
1601 | if (thi->t_state != ns) { | |
1602 | if (thi->task == NULL) { | |
1603 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1604 | return; | |
1605 | } | |
1606 | ||
1607 | thi->t_state = ns; | |
1608 | smp_mb(); | |
1609 | init_completion(&thi->stop); | |
1610 | if (thi->task != current) | |
1611 | force_sig(DRBD_SIGKILL, thi->task); | |
1612 | ||
1613 | } | |
1614 | ||
1615 | spin_unlock_irqrestore(&thi->t_lock, flags); | |
1616 | ||
1617 | if (wait) | |
1618 | wait_for_completion(&thi->stop); | |
1619 | } | |
1620 | ||
1621 | #ifdef CONFIG_SMP | |
1622 | /** | |
1623 | * drbd_calc_cpu_mask() - Generate CPU masks, spread over all CPUs | |
1624 | * @mdev: DRBD device. | |
1625 | * | |
1626 | * Forces all threads of a device onto the same CPU. This is beneficial for | |
1627 | * DRBD's performance. May be overwritten by user's configuration. | |
1628 | */ | |
1629 | void drbd_calc_cpu_mask(struct drbd_conf *mdev) | |
1630 | { | |
1631 | int ord, cpu; | |
1632 | ||
1633 | /* user override. */ | |
1634 | if (cpumask_weight(mdev->cpu_mask)) | |
1635 | return; | |
1636 | ||
1637 | ord = mdev_to_minor(mdev) % cpumask_weight(cpu_online_mask); | |
1638 | for_each_online_cpu(cpu) { | |
1639 | if (ord-- == 0) { | |
1640 | cpumask_set_cpu(cpu, mdev->cpu_mask); | |
1641 | return; | |
1642 | } | |
1643 | } | |
1644 | /* should not be reached */ | |
1645 | cpumask_setall(mdev->cpu_mask); | |
1646 | } | |
1647 | ||
1648 | /** | |
1649 | * drbd_thread_current_set_cpu() - modifies the cpu mask of the _current_ thread | |
1650 | * @mdev: DRBD device. | |
1651 | * | |
1652 | * call in the "main loop" of _all_ threads, no need for any mutex, current won't die | |
1653 | * prematurely. | |
1654 | */ | |
1655 | void drbd_thread_current_set_cpu(struct drbd_conf *mdev) | |
1656 | { | |
1657 | struct task_struct *p = current; | |
1658 | struct drbd_thread *thi = | |
1659 | p == mdev->asender.task ? &mdev->asender : | |
1660 | p == mdev->receiver.task ? &mdev->receiver : | |
1661 | p == mdev->worker.task ? &mdev->worker : | |
1662 | NULL; | |
1663 | ERR_IF(thi == NULL) | |
1664 | return; | |
1665 | if (!thi->reset_cpu_mask) | |
1666 | return; | |
1667 | thi->reset_cpu_mask = 0; | |
1668 | set_cpus_allowed_ptr(p, mdev->cpu_mask); | |
1669 | } | |
1670 | #endif | |
1671 | ||
1672 | /* the appropriate socket mutex must be held already */ | |
1673 | int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, | |
0b70a13d | 1674 | enum drbd_packets cmd, struct p_header80 *h, |
b411b363 PR |
1675 | size_t size, unsigned msg_flags) |
1676 | { | |
1677 | int sent, ok; | |
1678 | ||
1679 | ERR_IF(!h) return FALSE; | |
1680 | ERR_IF(!size) return FALSE; | |
1681 | ||
1682 | h->magic = BE_DRBD_MAGIC; | |
1683 | h->command = cpu_to_be16(cmd); | |
0b70a13d | 1684 | h->length = cpu_to_be16(size-sizeof(struct p_header80)); |
b411b363 | 1685 | |
b411b363 PR |
1686 | sent = drbd_send(mdev, sock, h, size, msg_flags); |
1687 | ||
1688 | ok = (sent == size); | |
1689 | if (!ok) | |
1690 | dev_err(DEV, "short sent %s size=%d sent=%d\n", | |
1691 | cmdname(cmd), (int)size, sent); | |
1692 | return ok; | |
1693 | } | |
1694 | ||
1695 | /* don't pass the socket. we may only look at it | |
1696 | * when we hold the appropriate socket mutex. | |
1697 | */ | |
1698 | int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket, | |
0b70a13d | 1699 | enum drbd_packets cmd, struct p_header80 *h, size_t size) |
b411b363 PR |
1700 | { |
1701 | int ok = 0; | |
1702 | struct socket *sock; | |
1703 | ||
1704 | if (use_data_socket) { | |
1705 | mutex_lock(&mdev->data.mutex); | |
1706 | sock = mdev->data.socket; | |
1707 | } else { | |
1708 | mutex_lock(&mdev->meta.mutex); | |
1709 | sock = mdev->meta.socket; | |
1710 | } | |
1711 | ||
1712 | /* drbd_disconnect() could have called drbd_free_sock() | |
1713 | * while we were waiting in down()... */ | |
1714 | if (likely(sock != NULL)) | |
1715 | ok = _drbd_send_cmd(mdev, sock, cmd, h, size, 0); | |
1716 | ||
1717 | if (use_data_socket) | |
1718 | mutex_unlock(&mdev->data.mutex); | |
1719 | else | |
1720 | mutex_unlock(&mdev->meta.mutex); | |
1721 | return ok; | |
1722 | } | |
1723 | ||
1724 | int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, char *data, | |
1725 | size_t size) | |
1726 | { | |
0b70a13d | 1727 | struct p_header80 h; |
b411b363 PR |
1728 | int ok; |
1729 | ||
1730 | h.magic = BE_DRBD_MAGIC; | |
1731 | h.command = cpu_to_be16(cmd); | |
1732 | h.length = cpu_to_be16(size); | |
1733 | ||
1734 | if (!drbd_get_data_sock(mdev)) | |
1735 | return 0; | |
1736 | ||
b411b363 PR |
1737 | ok = (sizeof(h) == |
1738 | drbd_send(mdev, mdev->data.socket, &h, sizeof(h), 0)); | |
1739 | ok = ok && (size == | |
1740 | drbd_send(mdev, mdev->data.socket, data, size, 0)); | |
1741 | ||
1742 | drbd_put_data_sock(mdev); | |
1743 | ||
1744 | return ok; | |
1745 | } | |
1746 | ||
1747 | int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc) | |
1748 | { | |
8e26f9cc | 1749 | struct p_rs_param_95 *p; |
b411b363 PR |
1750 | struct socket *sock; |
1751 | int size, rv; | |
1752 | const int apv = mdev->agreed_pro_version; | |
1753 | ||
1754 | size = apv <= 87 ? sizeof(struct p_rs_param) | |
1755 | : apv == 88 ? sizeof(struct p_rs_param) | |
1756 | + strlen(mdev->sync_conf.verify_alg) + 1 | |
8e26f9cc PR |
1757 | : apv <= 94 ? sizeof(struct p_rs_param_89) |
1758 | : /* apv >= 95 */ sizeof(struct p_rs_param_95); | |
b411b363 PR |
1759 | |
1760 | /* used from admin command context and receiver/worker context. | |
1761 | * to avoid kmalloc, grab the socket right here, | |
1762 | * then use the pre-allocated sbuf there */ | |
1763 | mutex_lock(&mdev->data.mutex); | |
1764 | sock = mdev->data.socket; | |
1765 | ||
1766 | if (likely(sock != NULL)) { | |
1767 | enum drbd_packets cmd = apv >= 89 ? P_SYNC_PARAM89 : P_SYNC_PARAM; | |
1768 | ||
8e26f9cc | 1769 | p = &mdev->data.sbuf.rs_param_95; |
b411b363 PR |
1770 | |
1771 | /* initialize verify_alg and csums_alg */ | |
1772 | memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); | |
1773 | ||
1774 | p->rate = cpu_to_be32(sc->rate); | |
8e26f9cc PR |
1775 | p->c_plan_ahead = cpu_to_be32(sc->c_plan_ahead); |
1776 | p->c_delay_target = cpu_to_be32(sc->c_delay_target); | |
1777 | p->c_fill_target = cpu_to_be32(sc->c_fill_target); | |
1778 | p->c_max_rate = cpu_to_be32(sc->c_max_rate); | |
b411b363 PR |
1779 | |
1780 | if (apv >= 88) | |
1781 | strcpy(p->verify_alg, mdev->sync_conf.verify_alg); | |
1782 | if (apv >= 89) | |
1783 | strcpy(p->csums_alg, mdev->sync_conf.csums_alg); | |
1784 | ||
1785 | rv = _drbd_send_cmd(mdev, sock, cmd, &p->head, size, 0); | |
1786 | } else | |
1787 | rv = 0; /* not ok */ | |
1788 | ||
1789 | mutex_unlock(&mdev->data.mutex); | |
1790 | ||
1791 | return rv; | |
1792 | } | |
1793 | ||
1794 | int drbd_send_protocol(struct drbd_conf *mdev) | |
1795 | { | |
1796 | struct p_protocol *p; | |
cf14c2e9 | 1797 | int size, cf, rv; |
b411b363 PR |
1798 | |
1799 | size = sizeof(struct p_protocol); | |
1800 | ||
1801 | if (mdev->agreed_pro_version >= 87) | |
1802 | size += strlen(mdev->net_conf->integrity_alg) + 1; | |
1803 | ||
1804 | /* we must not recurse into our own queue, | |
1805 | * as that is blocked during handshake */ | |
1806 | p = kmalloc(size, GFP_NOIO); | |
1807 | if (p == NULL) | |
1808 | return 0; | |
1809 | ||
1810 | p->protocol = cpu_to_be32(mdev->net_conf->wire_protocol); | |
1811 | p->after_sb_0p = cpu_to_be32(mdev->net_conf->after_sb_0p); | |
1812 | p->after_sb_1p = cpu_to_be32(mdev->net_conf->after_sb_1p); | |
1813 | p->after_sb_2p = cpu_to_be32(mdev->net_conf->after_sb_2p); | |
b411b363 PR |
1814 | p->two_primaries = cpu_to_be32(mdev->net_conf->two_primaries); |
1815 | ||
cf14c2e9 PR |
1816 | cf = 0; |
1817 | if (mdev->net_conf->want_lose) | |
1818 | cf |= CF_WANT_LOSE; | |
1819 | if (mdev->net_conf->dry_run) { | |
1820 | if (mdev->agreed_pro_version >= 92) | |
1821 | cf |= CF_DRY_RUN; | |
1822 | else { | |
1823 | dev_err(DEV, "--dry-run is not supported by peer"); | |
7ac314c8 | 1824 | kfree(p); |
cf14c2e9 PR |
1825 | return 0; |
1826 | } | |
1827 | } | |
1828 | p->conn_flags = cpu_to_be32(cf); | |
1829 | ||
b411b363 PR |
1830 | if (mdev->agreed_pro_version >= 87) |
1831 | strcpy(p->integrity_alg, mdev->net_conf->integrity_alg); | |
1832 | ||
1833 | rv = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_PROTOCOL, | |
0b70a13d | 1834 | (struct p_header80 *)p, size); |
b411b363 PR |
1835 | kfree(p); |
1836 | return rv; | |
1837 | } | |
1838 | ||
1839 | int _drbd_send_uuids(struct drbd_conf *mdev, u64 uuid_flags) | |
1840 | { | |
1841 | struct p_uuids p; | |
1842 | int i; | |
1843 | ||
1844 | if (!get_ldev_if_state(mdev, D_NEGOTIATING)) | |
1845 | return 1; | |
1846 | ||
1847 | for (i = UI_CURRENT; i < UI_SIZE; i++) | |
1848 | p.uuid[i] = mdev->ldev ? cpu_to_be64(mdev->ldev->md.uuid[i]) : 0; | |
1849 | ||
1850 | mdev->comm_bm_set = drbd_bm_total_weight(mdev); | |
1851 | p.uuid[UI_SIZE] = cpu_to_be64(mdev->comm_bm_set); | |
1852 | uuid_flags |= mdev->net_conf->want_lose ? 1 : 0; | |
1853 | uuid_flags |= test_bit(CRASHED_PRIMARY, &mdev->flags) ? 2 : 0; | |
1854 | uuid_flags |= mdev->new_state_tmp.disk == D_INCONSISTENT ? 4 : 0; | |
1855 | p.uuid[UI_FLAGS] = cpu_to_be64(uuid_flags); | |
1856 | ||
1857 | put_ldev(mdev); | |
1858 | ||
1859 | return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_UUIDS, | |
0b70a13d | 1860 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
1861 | } |
1862 | ||
1863 | int drbd_send_uuids(struct drbd_conf *mdev) | |
1864 | { | |
1865 | return _drbd_send_uuids(mdev, 0); | |
1866 | } | |
1867 | ||
1868 | int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev) | |
1869 | { | |
1870 | return _drbd_send_uuids(mdev, 8); | |
1871 | } | |
1872 | ||
1873 | ||
1874 | int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val) | |
1875 | { | |
1876 | struct p_rs_uuid p; | |
1877 | ||
1878 | p.uuid = cpu_to_be64(val); | |
1879 | ||
1880 | return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SYNC_UUID, | |
0b70a13d | 1881 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
1882 | } |
1883 | ||
e89b591c | 1884 | int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags) |
b411b363 PR |
1885 | { |
1886 | struct p_sizes p; | |
1887 | sector_t d_size, u_size; | |
1888 | int q_order_type; | |
1889 | int ok; | |
1890 | ||
1891 | if (get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
1892 | D_ASSERT(mdev->ldev->backing_bdev); | |
1893 | d_size = drbd_get_max_capacity(mdev->ldev); | |
1894 | u_size = mdev->ldev->dc.disk_size; | |
1895 | q_order_type = drbd_queue_order_type(mdev); | |
b411b363 PR |
1896 | put_ldev(mdev); |
1897 | } else { | |
1898 | d_size = 0; | |
1899 | u_size = 0; | |
1900 | q_order_type = QUEUE_ORDERED_NONE; | |
1901 | } | |
1902 | ||
1903 | p.d_size = cpu_to_be64(d_size); | |
1904 | p.u_size = cpu_to_be64(u_size); | |
1905 | p.c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(mdev->this_bdev)); | |
1906 | p.max_segment_size = cpu_to_be32(queue_max_segment_size(mdev->rq_queue)); | |
e89b591c PR |
1907 | p.queue_order_type = cpu_to_be16(q_order_type); |
1908 | p.dds_flags = cpu_to_be16(flags); | |
b411b363 PR |
1909 | |
1910 | ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SIZES, | |
0b70a13d | 1911 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
1912 | return ok; |
1913 | } | |
1914 | ||
1915 | /** | |
1916 | * drbd_send_state() - Sends the drbd state to the peer | |
1917 | * @mdev: DRBD device. | |
1918 | */ | |
1919 | int drbd_send_state(struct drbd_conf *mdev) | |
1920 | { | |
1921 | struct socket *sock; | |
1922 | struct p_state p; | |
1923 | int ok = 0; | |
1924 | ||
1925 | /* Grab state lock so we wont send state if we're in the middle | |
1926 | * of a cluster wide state change on another thread */ | |
1927 | drbd_state_lock(mdev); | |
1928 | ||
1929 | mutex_lock(&mdev->data.mutex); | |
1930 | ||
1931 | p.state = cpu_to_be32(mdev->state.i); /* Within the send mutex */ | |
1932 | sock = mdev->data.socket; | |
1933 | ||
1934 | if (likely(sock != NULL)) { | |
1935 | ok = _drbd_send_cmd(mdev, sock, P_STATE, | |
0b70a13d | 1936 | (struct p_header80 *)&p, sizeof(p), 0); |
b411b363 PR |
1937 | } |
1938 | ||
1939 | mutex_unlock(&mdev->data.mutex); | |
1940 | ||
1941 | drbd_state_unlock(mdev); | |
1942 | return ok; | |
1943 | } | |
1944 | ||
1945 | int drbd_send_state_req(struct drbd_conf *mdev, | |
1946 | union drbd_state mask, union drbd_state val) | |
1947 | { | |
1948 | struct p_req_state p; | |
1949 | ||
1950 | p.mask = cpu_to_be32(mask.i); | |
1951 | p.val = cpu_to_be32(val.i); | |
1952 | ||
1953 | return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_STATE_CHG_REQ, | |
0b70a13d | 1954 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
1955 | } |
1956 | ||
1957 | int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode) | |
1958 | { | |
1959 | struct p_req_state_reply p; | |
1960 | ||
1961 | p.retcode = cpu_to_be32(retcode); | |
1962 | ||
1963 | return drbd_send_cmd(mdev, USE_META_SOCKET, P_STATE_CHG_REPLY, | |
0b70a13d | 1964 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
1965 | } |
1966 | ||
1967 | int fill_bitmap_rle_bits(struct drbd_conf *mdev, | |
1968 | struct p_compressed_bm *p, | |
1969 | struct bm_xfer_ctx *c) | |
1970 | { | |
1971 | struct bitstream bs; | |
1972 | unsigned long plain_bits; | |
1973 | unsigned long tmp; | |
1974 | unsigned long rl; | |
1975 | unsigned len; | |
1976 | unsigned toggle; | |
1977 | int bits; | |
1978 | ||
1979 | /* may we use this feature? */ | |
1980 | if ((mdev->sync_conf.use_rle == 0) || | |
1981 | (mdev->agreed_pro_version < 90)) | |
1982 | return 0; | |
1983 | ||
1984 | if (c->bit_offset >= c->bm_bits) | |
1985 | return 0; /* nothing to do. */ | |
1986 | ||
1987 | /* use at most thus many bytes */ | |
1988 | bitstream_init(&bs, p->code, BM_PACKET_VLI_BYTES_MAX, 0); | |
1989 | memset(p->code, 0, BM_PACKET_VLI_BYTES_MAX); | |
1990 | /* plain bits covered in this code string */ | |
1991 | plain_bits = 0; | |
1992 | ||
1993 | /* p->encoding & 0x80 stores whether the first run length is set. | |
1994 | * bit offset is implicit. | |
1995 | * start with toggle == 2 to be able to tell the first iteration */ | |
1996 | toggle = 2; | |
1997 | ||
1998 | /* see how much plain bits we can stuff into one packet | |
1999 | * using RLE and VLI. */ | |
2000 | do { | |
2001 | tmp = (toggle == 0) ? _drbd_bm_find_next_zero(mdev, c->bit_offset) | |
2002 | : _drbd_bm_find_next(mdev, c->bit_offset); | |
2003 | if (tmp == -1UL) | |
2004 | tmp = c->bm_bits; | |
2005 | rl = tmp - c->bit_offset; | |
2006 | ||
2007 | if (toggle == 2) { /* first iteration */ | |
2008 | if (rl == 0) { | |
2009 | /* the first checked bit was set, | |
2010 | * store start value, */ | |
2011 | DCBP_set_start(p, 1); | |
2012 | /* but skip encoding of zero run length */ | |
2013 | toggle = !toggle; | |
2014 | continue; | |
2015 | } | |
2016 | DCBP_set_start(p, 0); | |
2017 | } | |
2018 | ||
2019 | /* paranoia: catch zero runlength. | |
2020 | * can only happen if bitmap is modified while we scan it. */ | |
2021 | if (rl == 0) { | |
2022 | dev_err(DEV, "unexpected zero runlength while encoding bitmap " | |
2023 | "t:%u bo:%lu\n", toggle, c->bit_offset); | |
2024 | return -1; | |
2025 | } | |
2026 | ||
2027 | bits = vli_encode_bits(&bs, rl); | |
2028 | if (bits == -ENOBUFS) /* buffer full */ | |
2029 | break; | |
2030 | if (bits <= 0) { | |
2031 | dev_err(DEV, "error while encoding bitmap: %d\n", bits); | |
2032 | return 0; | |
2033 | } | |
2034 | ||
2035 | toggle = !toggle; | |
2036 | plain_bits += rl; | |
2037 | c->bit_offset = tmp; | |
2038 | } while (c->bit_offset < c->bm_bits); | |
2039 | ||
2040 | len = bs.cur.b - p->code + !!bs.cur.bit; | |
2041 | ||
2042 | if (plain_bits < (len << 3)) { | |
2043 | /* incompressible with this method. | |
2044 | * we need to rewind both word and bit position. */ | |
2045 | c->bit_offset -= plain_bits; | |
2046 | bm_xfer_ctx_bit_to_word_offset(c); | |
2047 | c->bit_offset = c->word_offset * BITS_PER_LONG; | |
2048 | return 0; | |
2049 | } | |
2050 | ||
2051 | /* RLE + VLI was able to compress it just fine. | |
2052 | * update c->word_offset. */ | |
2053 | bm_xfer_ctx_bit_to_word_offset(c); | |
2054 | ||
2055 | /* store pad_bits */ | |
2056 | DCBP_set_pad_bits(p, (8 - bs.cur.bit) & 0x7); | |
2057 | ||
2058 | return len; | |
2059 | } | |
2060 | ||
2061 | enum { OK, FAILED, DONE } | |
2062 | send_bitmap_rle_or_plain(struct drbd_conf *mdev, | |
0b70a13d | 2063 | struct p_header80 *h, struct bm_xfer_ctx *c) |
b411b363 PR |
2064 | { |
2065 | struct p_compressed_bm *p = (void*)h; | |
2066 | unsigned long num_words; | |
2067 | int len; | |
2068 | int ok; | |
2069 | ||
2070 | len = fill_bitmap_rle_bits(mdev, p, c); | |
2071 | ||
2072 | if (len < 0) | |
2073 | return FAILED; | |
2074 | ||
2075 | if (len) { | |
2076 | DCBP_set_code(p, RLE_VLI_Bits); | |
2077 | ok = _drbd_send_cmd(mdev, mdev->data.socket, P_COMPRESSED_BITMAP, h, | |
2078 | sizeof(*p) + len, 0); | |
2079 | ||
2080 | c->packets[0]++; | |
2081 | c->bytes[0] += sizeof(*p) + len; | |
2082 | ||
2083 | if (c->bit_offset >= c->bm_bits) | |
2084 | len = 0; /* DONE */ | |
2085 | } else { | |
2086 | /* was not compressible. | |
2087 | * send a buffer full of plain text bits instead. */ | |
2088 | num_words = min_t(size_t, BM_PACKET_WORDS, c->bm_words - c->word_offset); | |
2089 | len = num_words * sizeof(long); | |
2090 | if (len) | |
2091 | drbd_bm_get_lel(mdev, c->word_offset, num_words, (unsigned long*)h->payload); | |
2092 | ok = _drbd_send_cmd(mdev, mdev->data.socket, P_BITMAP, | |
0b70a13d | 2093 | h, sizeof(struct p_header80) + len, 0); |
b411b363 PR |
2094 | c->word_offset += num_words; |
2095 | c->bit_offset = c->word_offset * BITS_PER_LONG; | |
2096 | ||
2097 | c->packets[1]++; | |
0b70a13d | 2098 | c->bytes[1] += sizeof(struct p_header80) + len; |
b411b363 PR |
2099 | |
2100 | if (c->bit_offset > c->bm_bits) | |
2101 | c->bit_offset = c->bm_bits; | |
2102 | } | |
2103 | ok = ok ? ((len == 0) ? DONE : OK) : FAILED; | |
2104 | ||
2105 | if (ok == DONE) | |
2106 | INFO_bm_xfer_stats(mdev, "send", c); | |
2107 | return ok; | |
2108 | } | |
2109 | ||
2110 | /* See the comment at receive_bitmap() */ | |
2111 | int _drbd_send_bitmap(struct drbd_conf *mdev) | |
2112 | { | |
2113 | struct bm_xfer_ctx c; | |
0b70a13d | 2114 | struct p_header80 *p; |
b411b363 PR |
2115 | int ret; |
2116 | ||
2117 | ERR_IF(!mdev->bitmap) return FALSE; | |
2118 | ||
2119 | /* maybe we should use some per thread scratch page, | |
2120 | * and allocate that during initial device creation? */ | |
0b70a13d | 2121 | p = (struct p_header80 *) __get_free_page(GFP_NOIO); |
b411b363 PR |
2122 | if (!p) { |
2123 | dev_err(DEV, "failed to allocate one page buffer in %s\n", __func__); | |
2124 | return FALSE; | |
2125 | } | |
2126 | ||
2127 | if (get_ldev(mdev)) { | |
2128 | if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) { | |
2129 | dev_info(DEV, "Writing the whole bitmap, MDF_FullSync was set.\n"); | |
2130 | drbd_bm_set_all(mdev); | |
2131 | if (drbd_bm_write(mdev)) { | |
2132 | /* write_bm did fail! Leave full sync flag set in Meta P_DATA | |
2133 | * but otherwise process as per normal - need to tell other | |
2134 | * side that a full resync is required! */ | |
2135 | dev_err(DEV, "Failed to write bitmap to disk!\n"); | |
2136 | } else { | |
2137 | drbd_md_clear_flag(mdev, MDF_FULL_SYNC); | |
2138 | drbd_md_sync(mdev); | |
2139 | } | |
2140 | } | |
2141 | put_ldev(mdev); | |
2142 | } | |
2143 | ||
2144 | c = (struct bm_xfer_ctx) { | |
2145 | .bm_bits = drbd_bm_bits(mdev), | |
2146 | .bm_words = drbd_bm_words(mdev), | |
2147 | }; | |
2148 | ||
2149 | do { | |
2150 | ret = send_bitmap_rle_or_plain(mdev, p, &c); | |
2151 | } while (ret == OK); | |
2152 | ||
2153 | free_page((unsigned long) p); | |
2154 | return (ret == DONE); | |
2155 | } | |
2156 | ||
2157 | int drbd_send_bitmap(struct drbd_conf *mdev) | |
2158 | { | |
2159 | int err; | |
2160 | ||
2161 | if (!drbd_get_data_sock(mdev)) | |
2162 | return -1; | |
2163 | err = !_drbd_send_bitmap(mdev); | |
2164 | drbd_put_data_sock(mdev); | |
2165 | return err; | |
2166 | } | |
2167 | ||
2168 | int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, u32 set_size) | |
2169 | { | |
2170 | int ok; | |
2171 | struct p_barrier_ack p; | |
2172 | ||
2173 | p.barrier = barrier_nr; | |
2174 | p.set_size = cpu_to_be32(set_size); | |
2175 | ||
2176 | if (mdev->state.conn < C_CONNECTED) | |
2177 | return FALSE; | |
2178 | ok = drbd_send_cmd(mdev, USE_META_SOCKET, P_BARRIER_ACK, | |
0b70a13d | 2179 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
2180 | return ok; |
2181 | } | |
2182 | ||
2183 | /** | |
2184 | * _drbd_send_ack() - Sends an ack packet | |
2185 | * @mdev: DRBD device. | |
2186 | * @cmd: Packet command code. | |
2187 | * @sector: sector, needs to be in big endian byte order | |
2188 | * @blksize: size in byte, needs to be in big endian byte order | |
2189 | * @block_id: Id, big endian byte order | |
2190 | */ | |
2191 | static int _drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd, | |
2192 | u64 sector, | |
2193 | u32 blksize, | |
2194 | u64 block_id) | |
2195 | { | |
2196 | int ok; | |
2197 | struct p_block_ack p; | |
2198 | ||
2199 | p.sector = sector; | |
2200 | p.block_id = block_id; | |
2201 | p.blksize = blksize; | |
2202 | p.seq_num = cpu_to_be32(atomic_add_return(1, &mdev->packet_seq)); | |
2203 | ||
2204 | if (!mdev->meta.socket || mdev->state.conn < C_CONNECTED) | |
2205 | return FALSE; | |
2206 | ok = drbd_send_cmd(mdev, USE_META_SOCKET, cmd, | |
0b70a13d | 2207 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
2208 | return ok; |
2209 | } | |
2210 | ||
2b2bf214 LE |
2211 | /* dp->sector and dp->block_id already/still in network byte order, |
2212 | * data_size is payload size according to dp->head, | |
2213 | * and may need to be corrected for digest size. */ | |
b411b363 | 2214 | int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd, |
2b2bf214 | 2215 | struct p_data *dp, int data_size) |
b411b363 | 2216 | { |
2b2bf214 LE |
2217 | data_size -= (mdev->agreed_pro_version >= 87 && mdev->integrity_r_tfm) ? |
2218 | crypto_hash_digestsize(mdev->integrity_r_tfm) : 0; | |
b411b363 PR |
2219 | return _drbd_send_ack(mdev, cmd, dp->sector, cpu_to_be32(data_size), |
2220 | dp->block_id); | |
2221 | } | |
2222 | ||
2223 | int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd, | |
2224 | struct p_block_req *rp) | |
2225 | { | |
2226 | return _drbd_send_ack(mdev, cmd, rp->sector, rp->blksize, rp->block_id); | |
2227 | } | |
2228 | ||
2229 | /** | |
2230 | * drbd_send_ack() - Sends an ack packet | |
2231 | * @mdev: DRBD device. | |
2232 | * @cmd: Packet command code. | |
2233 | * @e: Epoch entry. | |
2234 | */ | |
2235 | int drbd_send_ack(struct drbd_conf *mdev, | |
2236 | enum drbd_packets cmd, struct drbd_epoch_entry *e) | |
2237 | { | |
2238 | return _drbd_send_ack(mdev, cmd, | |
2239 | cpu_to_be64(e->sector), | |
2240 | cpu_to_be32(e->size), | |
2241 | e->block_id); | |
2242 | } | |
2243 | ||
2244 | /* This function misuses the block_id field to signal if the blocks | |
2245 | * are is sync or not. */ | |
2246 | int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd, | |
2247 | sector_t sector, int blksize, u64 block_id) | |
2248 | { | |
2249 | return _drbd_send_ack(mdev, cmd, | |
2250 | cpu_to_be64(sector), | |
2251 | cpu_to_be32(blksize), | |
2252 | cpu_to_be64(block_id)); | |
2253 | } | |
2254 | ||
2255 | int drbd_send_drequest(struct drbd_conf *mdev, int cmd, | |
2256 | sector_t sector, int size, u64 block_id) | |
2257 | { | |
2258 | int ok; | |
2259 | struct p_block_req p; | |
2260 | ||
2261 | p.sector = cpu_to_be64(sector); | |
2262 | p.block_id = block_id; | |
2263 | p.blksize = cpu_to_be32(size); | |
2264 | ||
2265 | ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, | |
0b70a13d | 2266 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
2267 | return ok; |
2268 | } | |
2269 | ||
2270 | int drbd_send_drequest_csum(struct drbd_conf *mdev, | |
2271 | sector_t sector, int size, | |
2272 | void *digest, int digest_size, | |
2273 | enum drbd_packets cmd) | |
2274 | { | |
2275 | int ok; | |
2276 | struct p_block_req p; | |
2277 | ||
2278 | p.sector = cpu_to_be64(sector); | |
2279 | p.block_id = BE_DRBD_MAGIC + 0xbeef; | |
2280 | p.blksize = cpu_to_be32(size); | |
2281 | ||
2282 | p.head.magic = BE_DRBD_MAGIC; | |
2283 | p.head.command = cpu_to_be16(cmd); | |
0b70a13d | 2284 | p.head.length = cpu_to_be16(sizeof(p) - sizeof(struct p_header80) + digest_size); |
b411b363 PR |
2285 | |
2286 | mutex_lock(&mdev->data.mutex); | |
2287 | ||
2288 | ok = (sizeof(p) == drbd_send(mdev, mdev->data.socket, &p, sizeof(p), 0)); | |
2289 | ok = ok && (digest_size == drbd_send(mdev, mdev->data.socket, digest, digest_size, 0)); | |
2290 | ||
2291 | mutex_unlock(&mdev->data.mutex); | |
2292 | ||
2293 | return ok; | |
2294 | } | |
2295 | ||
2296 | int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size) | |
2297 | { | |
2298 | int ok; | |
2299 | struct p_block_req p; | |
2300 | ||
2301 | p.sector = cpu_to_be64(sector); | |
2302 | p.block_id = BE_DRBD_MAGIC + 0xbabe; | |
2303 | p.blksize = cpu_to_be32(size); | |
2304 | ||
2305 | ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_OV_REQUEST, | |
0b70a13d | 2306 | (struct p_header80 *)&p, sizeof(p)); |
b411b363 PR |
2307 | return ok; |
2308 | } | |
2309 | ||
2310 | /* called on sndtimeo | |
2311 | * returns FALSE if we should retry, | |
2312 | * TRUE if we think connection is dead | |
2313 | */ | |
2314 | static int we_should_drop_the_connection(struct drbd_conf *mdev, struct socket *sock) | |
2315 | { | |
2316 | int drop_it; | |
2317 | /* long elapsed = (long)(jiffies - mdev->last_received); */ | |
2318 | ||
2319 | drop_it = mdev->meta.socket == sock | |
2320 | || !mdev->asender.task | |
2321 | || get_t_state(&mdev->asender) != Running | |
2322 | || mdev->state.conn < C_CONNECTED; | |
2323 | ||
2324 | if (drop_it) | |
2325 | return TRUE; | |
2326 | ||
2327 | drop_it = !--mdev->ko_count; | |
2328 | if (!drop_it) { | |
2329 | dev_err(DEV, "[%s/%d] sock_sendmsg time expired, ko = %u\n", | |
2330 | current->comm, current->pid, mdev->ko_count); | |
2331 | request_ping(mdev); | |
2332 | } | |
2333 | ||
2334 | return drop_it; /* && (mdev->state == R_PRIMARY) */; | |
2335 | } | |
2336 | ||
2337 | /* The idea of sendpage seems to be to put some kind of reference | |
2338 | * to the page into the skb, and to hand it over to the NIC. In | |
2339 | * this process get_page() gets called. | |
2340 | * | |
2341 | * As soon as the page was really sent over the network put_page() | |
2342 | * gets called by some part of the network layer. [ NIC driver? ] | |
2343 | * | |
2344 | * [ get_page() / put_page() increment/decrement the count. If count | |
2345 | * reaches 0 the page will be freed. ] | |
2346 | * | |
2347 | * This works nicely with pages from FSs. | |
2348 | * But this means that in protocol A we might signal IO completion too early! | |
2349 | * | |
2350 | * In order not to corrupt data during a resync we must make sure | |
2351 | * that we do not reuse our own buffer pages (EEs) to early, therefore | |
2352 | * we have the net_ee list. | |
2353 | * | |
2354 | * XFS seems to have problems, still, it submits pages with page_count == 0! | |
2355 | * As a workaround, we disable sendpage on pages | |
2356 | * with page_count == 0 or PageSlab. | |
2357 | */ | |
2358 | static int _drbd_no_send_page(struct drbd_conf *mdev, struct page *page, | |
ba11ad9a | 2359 | int offset, size_t size, unsigned msg_flags) |
b411b363 | 2360 | { |
ba11ad9a | 2361 | int sent = drbd_send(mdev, mdev->data.socket, kmap(page) + offset, size, msg_flags); |
b411b363 PR |
2362 | kunmap(page); |
2363 | if (sent == size) | |
2364 | mdev->send_cnt += size>>9; | |
2365 | return sent == size; | |
2366 | } | |
2367 | ||
2368 | static int _drbd_send_page(struct drbd_conf *mdev, struct page *page, | |
ba11ad9a | 2369 | int offset, size_t size, unsigned msg_flags) |
b411b363 PR |
2370 | { |
2371 | mm_segment_t oldfs = get_fs(); | |
2372 | int sent, ok; | |
2373 | int len = size; | |
2374 | ||
2375 | /* e.g. XFS meta- & log-data is in slab pages, which have a | |
2376 | * page_count of 0 and/or have PageSlab() set. | |
2377 | * we cannot use send_page for those, as that does get_page(); | |
2378 | * put_page(); and would cause either a VM_BUG directly, or | |
2379 | * __page_cache_release a page that would actually still be referenced | |
2380 | * by someone, leading to some obscure delayed Oops somewhere else. */ | |
2381 | if (disable_sendpage || (page_count(page) < 1) || PageSlab(page)) | |
ba11ad9a | 2382 | return _drbd_no_send_page(mdev, page, offset, size, msg_flags); |
b411b363 | 2383 | |
ba11ad9a | 2384 | msg_flags |= MSG_NOSIGNAL; |
b411b363 PR |
2385 | drbd_update_congested(mdev); |
2386 | set_fs(KERNEL_DS); | |
2387 | do { | |
2388 | sent = mdev->data.socket->ops->sendpage(mdev->data.socket, page, | |
2389 | offset, len, | |
ba11ad9a | 2390 | msg_flags); |
b411b363 PR |
2391 | if (sent == -EAGAIN) { |
2392 | if (we_should_drop_the_connection(mdev, | |
2393 | mdev->data.socket)) | |
2394 | break; | |
2395 | else | |
2396 | continue; | |
2397 | } | |
2398 | if (sent <= 0) { | |
2399 | dev_warn(DEV, "%s: size=%d len=%d sent=%d\n", | |
2400 | __func__, (int)size, len, sent); | |
2401 | break; | |
2402 | } | |
2403 | len -= sent; | |
2404 | offset += sent; | |
2405 | } while (len > 0 /* THINK && mdev->cstate >= C_CONNECTED*/); | |
2406 | set_fs(oldfs); | |
2407 | clear_bit(NET_CONGESTED, &mdev->flags); | |
2408 | ||
2409 | ok = (len == 0); | |
2410 | if (likely(ok)) | |
2411 | mdev->send_cnt += size>>9; | |
2412 | return ok; | |
2413 | } | |
2414 | ||
2415 | static int _drbd_send_bio(struct drbd_conf *mdev, struct bio *bio) | |
2416 | { | |
2417 | struct bio_vec *bvec; | |
2418 | int i; | |
ba11ad9a | 2419 | /* hint all but last page with MSG_MORE */ |
b411b363 PR |
2420 | __bio_for_each_segment(bvec, bio, i, 0) { |
2421 | if (!_drbd_no_send_page(mdev, bvec->bv_page, | |
ba11ad9a LE |
2422 | bvec->bv_offset, bvec->bv_len, |
2423 | i == bio->bi_vcnt -1 ? 0 : MSG_MORE)) | |
b411b363 PR |
2424 | return 0; |
2425 | } | |
2426 | return 1; | |
2427 | } | |
2428 | ||
2429 | static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio) | |
2430 | { | |
2431 | struct bio_vec *bvec; | |
2432 | int i; | |
ba11ad9a | 2433 | /* hint all but last page with MSG_MORE */ |
b411b363 PR |
2434 | __bio_for_each_segment(bvec, bio, i, 0) { |
2435 | if (!_drbd_send_page(mdev, bvec->bv_page, | |
ba11ad9a LE |
2436 | bvec->bv_offset, bvec->bv_len, |
2437 | i == bio->bi_vcnt -1 ? 0 : MSG_MORE)) | |
b411b363 PR |
2438 | return 0; |
2439 | } | |
b411b363 PR |
2440 | return 1; |
2441 | } | |
2442 | ||
45bb912b LE |
2443 | static int _drbd_send_zc_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e) |
2444 | { | |
2445 | struct page *page = e->pages; | |
2446 | unsigned len = e->size; | |
ba11ad9a | 2447 | /* hint all but last page with MSG_MORE */ |
45bb912b LE |
2448 | page_chain_for_each(page) { |
2449 | unsigned l = min_t(unsigned, len, PAGE_SIZE); | |
ba11ad9a LE |
2450 | if (!_drbd_send_page(mdev, page, 0, l, |
2451 | page_chain_next(page) ? MSG_MORE : 0)) | |
45bb912b LE |
2452 | return 0; |
2453 | len -= l; | |
2454 | } | |
2455 | return 1; | |
2456 | } | |
2457 | ||
76d2e7ec PR |
2458 | static u32 bio_flags_to_wire(struct drbd_conf *mdev, unsigned long bi_rw) |
2459 | { | |
2460 | if (mdev->agreed_pro_version >= 95) | |
2461 | return (bi_rw & REQ_SYNC ? DP_RW_SYNC : 0) | | |
2462 | (bi_rw & REQ_UNPLUG ? DP_UNPLUG : 0) | | |
2463 | (bi_rw & REQ_FUA ? DP_FUA : 0) | | |
2464 | (bi_rw & REQ_FLUSH ? DP_FLUSH : 0) | | |
2465 | (bi_rw & REQ_DISCARD ? DP_DISCARD : 0); | |
2466 | else | |
2467 | return bi_rw & (REQ_SYNC | REQ_UNPLUG) ? DP_RW_SYNC : 0; | |
2468 | } | |
2469 | ||
b411b363 PR |
2470 | /* Used to send write requests |
2471 | * R_PRIMARY -> Peer (P_DATA) | |
2472 | */ | |
2473 | int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req) | |
2474 | { | |
2475 | int ok = 1; | |
2476 | struct p_data p; | |
2477 | unsigned int dp_flags = 0; | |
2478 | void *dgb; | |
2479 | int dgs; | |
2480 | ||
2481 | if (!drbd_get_data_sock(mdev)) | |
2482 | return 0; | |
2483 | ||
2484 | dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_w_tfm) ? | |
2485 | crypto_hash_digestsize(mdev->integrity_w_tfm) : 0; | |
2486 | ||
d5373389 | 2487 | if (req->size <= DRBD_MAX_SIZE_H80_PACKET) { |
0b70a13d PR |
2488 | p.head.h80.magic = BE_DRBD_MAGIC; |
2489 | p.head.h80.command = cpu_to_be16(P_DATA); | |
2490 | p.head.h80.length = | |
2491 | cpu_to_be16(sizeof(p) - sizeof(union p_header) + dgs + req->size); | |
2492 | } else { | |
2493 | p.head.h95.magic = BE_DRBD_MAGIC_BIG; | |
2494 | p.head.h95.command = cpu_to_be16(P_DATA); | |
2495 | p.head.h95.length = | |
2496 | cpu_to_be32(sizeof(p) - sizeof(union p_header) + dgs + req->size); | |
2497 | } | |
b411b363 PR |
2498 | |
2499 | p.sector = cpu_to_be64(req->sector); | |
2500 | p.block_id = (unsigned long)req; | |
2501 | p.seq_num = cpu_to_be32(req->seq_num = | |
2502 | atomic_add_return(1, &mdev->packet_seq)); | |
b411b363 | 2503 | |
76d2e7ec PR |
2504 | dp_flags = bio_flags_to_wire(mdev, req->master_bio->bi_rw); |
2505 | ||
b411b363 PR |
2506 | if (mdev->state.conn >= C_SYNC_SOURCE && |
2507 | mdev->state.conn <= C_PAUSED_SYNC_T) | |
2508 | dp_flags |= DP_MAY_SET_IN_SYNC; | |
2509 | ||
2510 | p.dp_flags = cpu_to_be32(dp_flags); | |
b411b363 PR |
2511 | set_bit(UNPLUG_REMOTE, &mdev->flags); |
2512 | ok = (sizeof(p) == | |
ba11ad9a | 2513 | drbd_send(mdev, mdev->data.socket, &p, sizeof(p), dgs ? MSG_MORE : 0)); |
b411b363 PR |
2514 | if (ok && dgs) { |
2515 | dgb = mdev->int_dig_out; | |
45bb912b | 2516 | drbd_csum_bio(mdev, mdev->integrity_w_tfm, req->master_bio, dgb); |
ba11ad9a | 2517 | ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, 0); |
b411b363 PR |
2518 | } |
2519 | if (ok) { | |
2520 | if (mdev->net_conf->wire_protocol == DRBD_PROT_A) | |
2521 | ok = _drbd_send_bio(mdev, req->master_bio); | |
2522 | else | |
2523 | ok = _drbd_send_zc_bio(mdev, req->master_bio); | |
2524 | } | |
2525 | ||
2526 | drbd_put_data_sock(mdev); | |
bd26bfc5 | 2527 | |
b411b363 PR |
2528 | return ok; |
2529 | } | |
2530 | ||
2531 | /* answer packet, used to send data back for read requests: | |
2532 | * Peer -> (diskless) R_PRIMARY (P_DATA_REPLY) | |
2533 | * C_SYNC_SOURCE -> C_SYNC_TARGET (P_RS_DATA_REPLY) | |
2534 | */ | |
2535 | int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, | |
2536 | struct drbd_epoch_entry *e) | |
2537 | { | |
2538 | int ok; | |
2539 | struct p_data p; | |
2540 | void *dgb; | |
2541 | int dgs; | |
2542 | ||
2543 | dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_w_tfm) ? | |
2544 | crypto_hash_digestsize(mdev->integrity_w_tfm) : 0; | |
2545 | ||
d5373389 | 2546 | if (e->size <= DRBD_MAX_SIZE_H80_PACKET) { |
0b70a13d PR |
2547 | p.head.h80.magic = BE_DRBD_MAGIC; |
2548 | p.head.h80.command = cpu_to_be16(cmd); | |
2549 | p.head.h80.length = | |
2550 | cpu_to_be16(sizeof(p) - sizeof(struct p_header80) + dgs + e->size); | |
2551 | } else { | |
2552 | p.head.h95.magic = BE_DRBD_MAGIC_BIG; | |
2553 | p.head.h95.command = cpu_to_be16(cmd); | |
2554 | p.head.h95.length = | |
2555 | cpu_to_be32(sizeof(p) - sizeof(struct p_header80) + dgs + e->size); | |
2556 | } | |
b411b363 PR |
2557 | |
2558 | p.sector = cpu_to_be64(e->sector); | |
2559 | p.block_id = e->block_id; | |
2560 | /* p.seq_num = 0; No sequence numbers here.. */ | |
2561 | ||
2562 | /* Only called by our kernel thread. | |
2563 | * This one may be interrupted by DRBD_SIG and/or DRBD_SIGKILL | |
2564 | * in response to admin command or module unload. | |
2565 | */ | |
2566 | if (!drbd_get_data_sock(mdev)) | |
2567 | return 0; | |
2568 | ||
0b70a13d | 2569 | ok = sizeof(p) == drbd_send(mdev, mdev->data.socket, &p, sizeof(p), dgs ? MSG_MORE : 0); |
b411b363 PR |
2570 | if (ok && dgs) { |
2571 | dgb = mdev->int_dig_out; | |
45bb912b | 2572 | drbd_csum_ee(mdev, mdev->integrity_w_tfm, e, dgb); |
ba11ad9a | 2573 | ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, 0); |
b411b363 PR |
2574 | } |
2575 | if (ok) | |
45bb912b | 2576 | ok = _drbd_send_zc_ee(mdev, e); |
b411b363 PR |
2577 | |
2578 | drbd_put_data_sock(mdev); | |
bd26bfc5 | 2579 | |
b411b363 PR |
2580 | return ok; |
2581 | } | |
2582 | ||
2583 | /* | |
2584 | drbd_send distinguishes two cases: | |
2585 | ||
2586 | Packets sent via the data socket "sock" | |
2587 | and packets sent via the meta data socket "msock" | |
2588 | ||
2589 | sock msock | |
2590 | -----------------+-------------------------+------------------------------ | |
2591 | timeout conf.timeout / 2 conf.timeout / 2 | |
2592 | timeout action send a ping via msock Abort communication | |
2593 | and close all sockets | |
2594 | */ | |
2595 | ||
2596 | /* | |
2597 | * you must have down()ed the appropriate [m]sock_mutex elsewhere! | |
2598 | */ | |
2599 | int drbd_send(struct drbd_conf *mdev, struct socket *sock, | |
2600 | void *buf, size_t size, unsigned msg_flags) | |
2601 | { | |
2602 | struct kvec iov; | |
2603 | struct msghdr msg; | |
2604 | int rv, sent = 0; | |
2605 | ||
2606 | if (!sock) | |
2607 | return -1000; | |
2608 | ||
2609 | /* THINK if (signal_pending) return ... ? */ | |
2610 | ||
2611 | iov.iov_base = buf; | |
2612 | iov.iov_len = size; | |
2613 | ||
2614 | msg.msg_name = NULL; | |
2615 | msg.msg_namelen = 0; | |
2616 | msg.msg_control = NULL; | |
2617 | msg.msg_controllen = 0; | |
2618 | msg.msg_flags = msg_flags | MSG_NOSIGNAL; | |
2619 | ||
2620 | if (sock == mdev->data.socket) { | |
2621 | mdev->ko_count = mdev->net_conf->ko_count; | |
2622 | drbd_update_congested(mdev); | |
2623 | } | |
2624 | do { | |
2625 | /* STRANGE | |
2626 | * tcp_sendmsg does _not_ use its size parameter at all ? | |
2627 | * | |
2628 | * -EAGAIN on timeout, -EINTR on signal. | |
2629 | */ | |
2630 | /* THINK | |
2631 | * do we need to block DRBD_SIG if sock == &meta.socket ?? | |
2632 | * otherwise wake_asender() might interrupt some send_*Ack ! | |
2633 | */ | |
2634 | rv = kernel_sendmsg(sock, &msg, &iov, 1, size); | |
2635 | if (rv == -EAGAIN) { | |
2636 | if (we_should_drop_the_connection(mdev, sock)) | |
2637 | break; | |
2638 | else | |
2639 | continue; | |
2640 | } | |
2641 | D_ASSERT(rv != 0); | |
2642 | if (rv == -EINTR) { | |
2643 | flush_signals(current); | |
2644 | rv = 0; | |
2645 | } | |
2646 | if (rv < 0) | |
2647 | break; | |
2648 | sent += rv; | |
2649 | iov.iov_base += rv; | |
2650 | iov.iov_len -= rv; | |
2651 | } while (sent < size); | |
2652 | ||
2653 | if (sock == mdev->data.socket) | |
2654 | clear_bit(NET_CONGESTED, &mdev->flags); | |
2655 | ||
2656 | if (rv <= 0) { | |
2657 | if (rv != -EAGAIN) { | |
2658 | dev_err(DEV, "%s_sendmsg returned %d\n", | |
2659 | sock == mdev->meta.socket ? "msock" : "sock", | |
2660 | rv); | |
2661 | drbd_force_state(mdev, NS(conn, C_BROKEN_PIPE)); | |
2662 | } else | |
2663 | drbd_force_state(mdev, NS(conn, C_TIMEOUT)); | |
2664 | } | |
2665 | ||
2666 | return sent; | |
2667 | } | |
2668 | ||
2669 | static int drbd_open(struct block_device *bdev, fmode_t mode) | |
2670 | { | |
2671 | struct drbd_conf *mdev = bdev->bd_disk->private_data; | |
2672 | unsigned long flags; | |
2673 | int rv = 0; | |
2674 | ||
6e9624b8 | 2675 | lock_kernel(); |
b411b363 PR |
2676 | spin_lock_irqsave(&mdev->req_lock, flags); |
2677 | /* to have a stable mdev->state.role | |
2678 | * and no race with updating open_cnt */ | |
2679 | ||
2680 | if (mdev->state.role != R_PRIMARY) { | |
2681 | if (mode & FMODE_WRITE) | |
2682 | rv = -EROFS; | |
2683 | else if (!allow_oos) | |
2684 | rv = -EMEDIUMTYPE; | |
2685 | } | |
2686 | ||
2687 | if (!rv) | |
2688 | mdev->open_cnt++; | |
2689 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
6e9624b8 | 2690 | unlock_kernel(); |
b411b363 PR |
2691 | |
2692 | return rv; | |
2693 | } | |
2694 | ||
2695 | static int drbd_release(struct gendisk *gd, fmode_t mode) | |
2696 | { | |
2697 | struct drbd_conf *mdev = gd->private_data; | |
6e9624b8 | 2698 | lock_kernel(); |
b411b363 | 2699 | mdev->open_cnt--; |
6e9624b8 | 2700 | unlock_kernel(); |
b411b363 PR |
2701 | return 0; |
2702 | } | |
2703 | ||
2704 | static void drbd_unplug_fn(struct request_queue *q) | |
2705 | { | |
2706 | struct drbd_conf *mdev = q->queuedata; | |
2707 | ||
b411b363 PR |
2708 | /* unplug FIRST */ |
2709 | spin_lock_irq(q->queue_lock); | |
2710 | blk_remove_plug(q); | |
2711 | spin_unlock_irq(q->queue_lock); | |
2712 | ||
2713 | /* only if connected */ | |
2714 | spin_lock_irq(&mdev->req_lock); | |
2715 | if (mdev->state.pdsk >= D_INCONSISTENT && mdev->state.conn >= C_CONNECTED) { | |
2716 | D_ASSERT(mdev->state.role == R_PRIMARY); | |
2717 | if (test_and_clear_bit(UNPLUG_REMOTE, &mdev->flags)) { | |
2718 | /* add to the data.work queue, | |
2719 | * unless already queued. | |
2720 | * XXX this might be a good addition to drbd_queue_work | |
2721 | * anyways, to detect "double queuing" ... */ | |
2722 | if (list_empty(&mdev->unplug_work.list)) | |
2723 | drbd_queue_work(&mdev->data.work, | |
2724 | &mdev->unplug_work); | |
2725 | } | |
2726 | } | |
2727 | spin_unlock_irq(&mdev->req_lock); | |
2728 | ||
2729 | if (mdev->state.disk >= D_INCONSISTENT) | |
2730 | drbd_kick_lo(mdev); | |
2731 | } | |
2732 | ||
2733 | static void drbd_set_defaults(struct drbd_conf *mdev) | |
2734 | { | |
85f4cc17 PR |
2735 | /* This way we get a compile error when sync_conf grows, |
2736 | and we forgot to initialize it here */ | |
2737 | mdev->sync_conf = (struct syncer_conf) { | |
2738 | /* .rate = */ DRBD_RATE_DEF, | |
2739 | /* .after = */ DRBD_AFTER_DEF, | |
2740 | /* .al_extents = */ DRBD_AL_EXTENTS_DEF, | |
85f4cc17 PR |
2741 | /* .verify_alg = */ {}, 0, |
2742 | /* .cpu_mask = */ {}, 0, | |
2743 | /* .csums_alg = */ {}, 0, | |
e756414f | 2744 | /* .use_rle = */ 0, |
9a31d716 PR |
2745 | /* .on_no_data = */ DRBD_ON_NO_DATA_DEF, |
2746 | /* .c_plan_ahead = */ DRBD_C_PLAN_AHEAD_DEF, | |
2747 | /* .c_delay_target = */ DRBD_C_DELAY_TARGET_DEF, | |
2748 | /* .c_fill_target = */ DRBD_C_FILL_TARGET_DEF, | |
0f0601f4 LE |
2749 | /* .c_max_rate = */ DRBD_C_MAX_RATE_DEF, |
2750 | /* .c_min_rate = */ DRBD_C_MIN_RATE_DEF | |
85f4cc17 PR |
2751 | }; |
2752 | ||
2753 | /* Have to use that way, because the layout differs between | |
2754 | big endian and little endian */ | |
b411b363 PR |
2755 | mdev->state = (union drbd_state) { |
2756 | { .role = R_SECONDARY, | |
2757 | .peer = R_UNKNOWN, | |
2758 | .conn = C_STANDALONE, | |
2759 | .disk = D_DISKLESS, | |
2760 | .pdsk = D_UNKNOWN, | |
fb22c402 PR |
2761 | .susp = 0, |
2762 | .susp_nod = 0, | |
2763 | .susp_fen = 0 | |
b411b363 PR |
2764 | } }; |
2765 | } | |
2766 | ||
2767 | void drbd_init_set_defaults(struct drbd_conf *mdev) | |
2768 | { | |
2769 | /* the memset(,0,) did most of this. | |
2770 | * note: only assignments, no allocation in here */ | |
2771 | ||
2772 | drbd_set_defaults(mdev); | |
2773 | ||
2774 | /* for now, we do NOT yet support it, | |
2775 | * even though we start some framework | |
2776 | * to eventually support barriers */ | |
2777 | set_bit(NO_BARRIER_SUPP, &mdev->flags); | |
2778 | ||
2779 | atomic_set(&mdev->ap_bio_cnt, 0); | |
2780 | atomic_set(&mdev->ap_pending_cnt, 0); | |
2781 | atomic_set(&mdev->rs_pending_cnt, 0); | |
2782 | atomic_set(&mdev->unacked_cnt, 0); | |
2783 | atomic_set(&mdev->local_cnt, 0); | |
2784 | atomic_set(&mdev->net_cnt, 0); | |
2785 | atomic_set(&mdev->packet_seq, 0); | |
2786 | atomic_set(&mdev->pp_in_use, 0); | |
435f0740 | 2787 | atomic_set(&mdev->pp_in_use_by_net, 0); |
778f271d | 2788 | atomic_set(&mdev->rs_sect_in, 0); |
0f0601f4 | 2789 | atomic_set(&mdev->rs_sect_ev, 0); |
b411b363 PR |
2790 | |
2791 | mutex_init(&mdev->md_io_mutex); | |
2792 | mutex_init(&mdev->data.mutex); | |
2793 | mutex_init(&mdev->meta.mutex); | |
2794 | sema_init(&mdev->data.work.s, 0); | |
2795 | sema_init(&mdev->meta.work.s, 0); | |
2796 | mutex_init(&mdev->state_mutex); | |
2797 | ||
2798 | spin_lock_init(&mdev->data.work.q_lock); | |
2799 | spin_lock_init(&mdev->meta.work.q_lock); | |
2800 | ||
2801 | spin_lock_init(&mdev->al_lock); | |
2802 | spin_lock_init(&mdev->req_lock); | |
2803 | spin_lock_init(&mdev->peer_seq_lock); | |
2804 | spin_lock_init(&mdev->epoch_lock); | |
2805 | ||
2806 | INIT_LIST_HEAD(&mdev->active_ee); | |
2807 | INIT_LIST_HEAD(&mdev->sync_ee); | |
2808 | INIT_LIST_HEAD(&mdev->done_ee); | |
2809 | INIT_LIST_HEAD(&mdev->read_ee); | |
2810 | INIT_LIST_HEAD(&mdev->net_ee); | |
2811 | INIT_LIST_HEAD(&mdev->resync_reads); | |
2812 | INIT_LIST_HEAD(&mdev->data.work.q); | |
2813 | INIT_LIST_HEAD(&mdev->meta.work.q); | |
2814 | INIT_LIST_HEAD(&mdev->resync_work.list); | |
2815 | INIT_LIST_HEAD(&mdev->unplug_work.list); | |
e9e6f3ec | 2816 | INIT_LIST_HEAD(&mdev->go_diskless.list); |
b411b363 PR |
2817 | INIT_LIST_HEAD(&mdev->md_sync_work.list); |
2818 | INIT_LIST_HEAD(&mdev->bm_io_work.w.list); | |
0ced55a3 | 2819 | |
b411b363 PR |
2820 | mdev->resync_work.cb = w_resync_inactive; |
2821 | mdev->unplug_work.cb = w_send_write_hint; | |
e9e6f3ec | 2822 | mdev->go_diskless.cb = w_go_diskless; |
b411b363 PR |
2823 | mdev->md_sync_work.cb = w_md_sync; |
2824 | mdev->bm_io_work.w.cb = w_bitmap_io; | |
2825 | init_timer(&mdev->resync_timer); | |
2826 | init_timer(&mdev->md_sync_timer); | |
2827 | mdev->resync_timer.function = resync_timer_fn; | |
2828 | mdev->resync_timer.data = (unsigned long) mdev; | |
2829 | mdev->md_sync_timer.function = md_sync_timer_fn; | |
2830 | mdev->md_sync_timer.data = (unsigned long) mdev; | |
2831 | ||
2832 | init_waitqueue_head(&mdev->misc_wait); | |
2833 | init_waitqueue_head(&mdev->state_wait); | |
84dfb9f5 | 2834 | init_waitqueue_head(&mdev->net_cnt_wait); |
b411b363 PR |
2835 | init_waitqueue_head(&mdev->ee_wait); |
2836 | init_waitqueue_head(&mdev->al_wait); | |
2837 | init_waitqueue_head(&mdev->seq_wait); | |
2838 | ||
2839 | drbd_thread_init(mdev, &mdev->receiver, drbdd_init); | |
2840 | drbd_thread_init(mdev, &mdev->worker, drbd_worker); | |
2841 | drbd_thread_init(mdev, &mdev->asender, drbd_asender); | |
2842 | ||
2843 | mdev->agreed_pro_version = PRO_VERSION_MAX; | |
2844 | mdev->write_ordering = WO_bio_barrier; | |
2845 | mdev->resync_wenr = LC_FREE; | |
2846 | } | |
2847 | ||
2848 | void drbd_mdev_cleanup(struct drbd_conf *mdev) | |
2849 | { | |
1d7734a0 | 2850 | int i; |
b411b363 PR |
2851 | if (mdev->receiver.t_state != None) |
2852 | dev_err(DEV, "ASSERT FAILED: receiver t_state == %d expected 0.\n", | |
2853 | mdev->receiver.t_state); | |
2854 | ||
2855 | /* no need to lock it, I'm the only thread alive */ | |
2856 | if (atomic_read(&mdev->current_epoch->epoch_size) != 0) | |
2857 | dev_err(DEV, "epoch_size:%d\n", atomic_read(&mdev->current_epoch->epoch_size)); | |
2858 | mdev->al_writ_cnt = | |
2859 | mdev->bm_writ_cnt = | |
2860 | mdev->read_cnt = | |
2861 | mdev->recv_cnt = | |
2862 | mdev->send_cnt = | |
2863 | mdev->writ_cnt = | |
2864 | mdev->p_size = | |
2865 | mdev->rs_start = | |
2866 | mdev->rs_total = | |
1d7734a0 LE |
2867 | mdev->rs_failed = 0; |
2868 | mdev->rs_last_events = 0; | |
0f0601f4 | 2869 | mdev->rs_last_sect_ev = 0; |
1d7734a0 LE |
2870 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { |
2871 | mdev->rs_mark_left[i] = 0; | |
2872 | mdev->rs_mark_time[i] = 0; | |
2873 | } | |
b411b363 PR |
2874 | D_ASSERT(mdev->net_conf == NULL); |
2875 | ||
2876 | drbd_set_my_capacity(mdev, 0); | |
2877 | if (mdev->bitmap) { | |
2878 | /* maybe never allocated. */ | |
02d9a94b | 2879 | drbd_bm_resize(mdev, 0, 1); |
b411b363 PR |
2880 | drbd_bm_cleanup(mdev); |
2881 | } | |
2882 | ||
2883 | drbd_free_resources(mdev); | |
0778286a | 2884 | clear_bit(AL_SUSPENDED, &mdev->flags); |
b411b363 PR |
2885 | |
2886 | /* | |
2887 | * currently we drbd_init_ee only on module load, so | |
2888 | * we may do drbd_release_ee only on module unload! | |
2889 | */ | |
2890 | D_ASSERT(list_empty(&mdev->active_ee)); | |
2891 | D_ASSERT(list_empty(&mdev->sync_ee)); | |
2892 | D_ASSERT(list_empty(&mdev->done_ee)); | |
2893 | D_ASSERT(list_empty(&mdev->read_ee)); | |
2894 | D_ASSERT(list_empty(&mdev->net_ee)); | |
2895 | D_ASSERT(list_empty(&mdev->resync_reads)); | |
2896 | D_ASSERT(list_empty(&mdev->data.work.q)); | |
2897 | D_ASSERT(list_empty(&mdev->meta.work.q)); | |
2898 | D_ASSERT(list_empty(&mdev->resync_work.list)); | |
2899 | D_ASSERT(list_empty(&mdev->unplug_work.list)); | |
e9e6f3ec | 2900 | D_ASSERT(list_empty(&mdev->go_diskless.list)); |
b411b363 PR |
2901 | |
2902 | } | |
2903 | ||
2904 | ||
2905 | static void drbd_destroy_mempools(void) | |
2906 | { | |
2907 | struct page *page; | |
2908 | ||
2909 | while (drbd_pp_pool) { | |
2910 | page = drbd_pp_pool; | |
2911 | drbd_pp_pool = (struct page *)page_private(page); | |
2912 | __free_page(page); | |
2913 | drbd_pp_vacant--; | |
2914 | } | |
2915 | ||
2916 | /* D_ASSERT(atomic_read(&drbd_pp_vacant)==0); */ | |
2917 | ||
2918 | if (drbd_ee_mempool) | |
2919 | mempool_destroy(drbd_ee_mempool); | |
2920 | if (drbd_request_mempool) | |
2921 | mempool_destroy(drbd_request_mempool); | |
2922 | if (drbd_ee_cache) | |
2923 | kmem_cache_destroy(drbd_ee_cache); | |
2924 | if (drbd_request_cache) | |
2925 | kmem_cache_destroy(drbd_request_cache); | |
2926 | if (drbd_bm_ext_cache) | |
2927 | kmem_cache_destroy(drbd_bm_ext_cache); | |
2928 | if (drbd_al_ext_cache) | |
2929 | kmem_cache_destroy(drbd_al_ext_cache); | |
2930 | ||
2931 | drbd_ee_mempool = NULL; | |
2932 | drbd_request_mempool = NULL; | |
2933 | drbd_ee_cache = NULL; | |
2934 | drbd_request_cache = NULL; | |
2935 | drbd_bm_ext_cache = NULL; | |
2936 | drbd_al_ext_cache = NULL; | |
2937 | ||
2938 | return; | |
2939 | } | |
2940 | ||
2941 | static int drbd_create_mempools(void) | |
2942 | { | |
2943 | struct page *page; | |
2944 | const int number = (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE) * minor_count; | |
2945 | int i; | |
2946 | ||
2947 | /* prepare our caches and mempools */ | |
2948 | drbd_request_mempool = NULL; | |
2949 | drbd_ee_cache = NULL; | |
2950 | drbd_request_cache = NULL; | |
2951 | drbd_bm_ext_cache = NULL; | |
2952 | drbd_al_ext_cache = NULL; | |
2953 | drbd_pp_pool = NULL; | |
2954 | ||
2955 | /* caches */ | |
2956 | drbd_request_cache = kmem_cache_create( | |
2957 | "drbd_req", sizeof(struct drbd_request), 0, 0, NULL); | |
2958 | if (drbd_request_cache == NULL) | |
2959 | goto Enomem; | |
2960 | ||
2961 | drbd_ee_cache = kmem_cache_create( | |
2962 | "drbd_ee", sizeof(struct drbd_epoch_entry), 0, 0, NULL); | |
2963 | if (drbd_ee_cache == NULL) | |
2964 | goto Enomem; | |
2965 | ||
2966 | drbd_bm_ext_cache = kmem_cache_create( | |
2967 | "drbd_bm", sizeof(struct bm_extent), 0, 0, NULL); | |
2968 | if (drbd_bm_ext_cache == NULL) | |
2969 | goto Enomem; | |
2970 | ||
2971 | drbd_al_ext_cache = kmem_cache_create( | |
2972 | "drbd_al", sizeof(struct lc_element), 0, 0, NULL); | |
2973 | if (drbd_al_ext_cache == NULL) | |
2974 | goto Enomem; | |
2975 | ||
2976 | /* mempools */ | |
2977 | drbd_request_mempool = mempool_create(number, | |
2978 | mempool_alloc_slab, mempool_free_slab, drbd_request_cache); | |
2979 | if (drbd_request_mempool == NULL) | |
2980 | goto Enomem; | |
2981 | ||
2982 | drbd_ee_mempool = mempool_create(number, | |
2983 | mempool_alloc_slab, mempool_free_slab, drbd_ee_cache); | |
2984 | if (drbd_request_mempool == NULL) | |
2985 | goto Enomem; | |
2986 | ||
2987 | /* drbd's page pool */ | |
2988 | spin_lock_init(&drbd_pp_lock); | |
2989 | ||
2990 | for (i = 0; i < number; i++) { | |
2991 | page = alloc_page(GFP_HIGHUSER); | |
2992 | if (!page) | |
2993 | goto Enomem; | |
2994 | set_page_private(page, (unsigned long)drbd_pp_pool); | |
2995 | drbd_pp_pool = page; | |
2996 | } | |
2997 | drbd_pp_vacant = number; | |
2998 | ||
2999 | return 0; | |
3000 | ||
3001 | Enomem: | |
3002 | drbd_destroy_mempools(); /* in case we allocated some */ | |
3003 | return -ENOMEM; | |
3004 | } | |
3005 | ||
3006 | static int drbd_notify_sys(struct notifier_block *this, unsigned long code, | |
3007 | void *unused) | |
3008 | { | |
3009 | /* just so we have it. you never know what interesting things we | |
3010 | * might want to do here some day... | |
3011 | */ | |
3012 | ||
3013 | return NOTIFY_DONE; | |
3014 | } | |
3015 | ||
3016 | static struct notifier_block drbd_notifier = { | |
3017 | .notifier_call = drbd_notify_sys, | |
3018 | }; | |
3019 | ||
3020 | static void drbd_release_ee_lists(struct drbd_conf *mdev) | |
3021 | { | |
3022 | int rr; | |
3023 | ||
3024 | rr = drbd_release_ee(mdev, &mdev->active_ee); | |
3025 | if (rr) | |
3026 | dev_err(DEV, "%d EEs in active list found!\n", rr); | |
3027 | ||
3028 | rr = drbd_release_ee(mdev, &mdev->sync_ee); | |
3029 | if (rr) | |
3030 | dev_err(DEV, "%d EEs in sync list found!\n", rr); | |
3031 | ||
3032 | rr = drbd_release_ee(mdev, &mdev->read_ee); | |
3033 | if (rr) | |
3034 | dev_err(DEV, "%d EEs in read list found!\n", rr); | |
3035 | ||
3036 | rr = drbd_release_ee(mdev, &mdev->done_ee); | |
3037 | if (rr) | |
3038 | dev_err(DEV, "%d EEs in done list found!\n", rr); | |
3039 | ||
3040 | rr = drbd_release_ee(mdev, &mdev->net_ee); | |
3041 | if (rr) | |
3042 | dev_err(DEV, "%d EEs in net list found!\n", rr); | |
3043 | } | |
3044 | ||
3045 | /* caution. no locking. | |
3046 | * currently only used from module cleanup code. */ | |
3047 | static void drbd_delete_device(unsigned int minor) | |
3048 | { | |
3049 | struct drbd_conf *mdev = minor_to_mdev(minor); | |
3050 | ||
3051 | if (!mdev) | |
3052 | return; | |
3053 | ||
3054 | /* paranoia asserts */ | |
3055 | if (mdev->open_cnt != 0) | |
3056 | dev_err(DEV, "open_cnt = %d in %s:%u", mdev->open_cnt, | |
3057 | __FILE__ , __LINE__); | |
3058 | ||
3059 | ERR_IF (!list_empty(&mdev->data.work.q)) { | |
3060 | struct list_head *lp; | |
3061 | list_for_each(lp, &mdev->data.work.q) { | |
3062 | dev_err(DEV, "lp = %p\n", lp); | |
3063 | } | |
3064 | }; | |
3065 | /* end paranoia asserts */ | |
3066 | ||
3067 | del_gendisk(mdev->vdisk); | |
3068 | ||
3069 | /* cleanup stuff that may have been allocated during | |
3070 | * device (re-)configuration or state changes */ | |
3071 | ||
3072 | if (mdev->this_bdev) | |
3073 | bdput(mdev->this_bdev); | |
3074 | ||
3075 | drbd_free_resources(mdev); | |
3076 | ||
3077 | drbd_release_ee_lists(mdev); | |
3078 | ||
3079 | /* should be free'd on disconnect? */ | |
3080 | kfree(mdev->ee_hash); | |
3081 | /* | |
3082 | mdev->ee_hash_s = 0; | |
3083 | mdev->ee_hash = NULL; | |
3084 | */ | |
3085 | ||
3086 | lc_destroy(mdev->act_log); | |
3087 | lc_destroy(mdev->resync); | |
3088 | ||
3089 | kfree(mdev->p_uuid); | |
3090 | /* mdev->p_uuid = NULL; */ | |
3091 | ||
3092 | kfree(mdev->int_dig_out); | |
3093 | kfree(mdev->int_dig_in); | |
3094 | kfree(mdev->int_dig_vv); | |
3095 | ||
3096 | /* cleanup the rest that has been | |
3097 | * allocated from drbd_new_device | |
3098 | * and actually free the mdev itself */ | |
3099 | drbd_free_mdev(mdev); | |
3100 | } | |
3101 | ||
3102 | static void drbd_cleanup(void) | |
3103 | { | |
3104 | unsigned int i; | |
3105 | ||
3106 | unregister_reboot_notifier(&drbd_notifier); | |
3107 | ||
3108 | drbd_nl_cleanup(); | |
3109 | ||
3110 | if (minor_table) { | |
3111 | if (drbd_proc) | |
3112 | remove_proc_entry("drbd", NULL); | |
3113 | i = minor_count; | |
3114 | while (i--) | |
3115 | drbd_delete_device(i); | |
3116 | drbd_destroy_mempools(); | |
3117 | } | |
3118 | ||
3119 | kfree(minor_table); | |
3120 | ||
3121 | unregister_blkdev(DRBD_MAJOR, "drbd"); | |
3122 | ||
3123 | printk(KERN_INFO "drbd: module cleanup done.\n"); | |
3124 | } | |
3125 | ||
3126 | /** | |
3127 | * drbd_congested() - Callback for pdflush | |
3128 | * @congested_data: User data | |
3129 | * @bdi_bits: Bits pdflush is currently interested in | |
3130 | * | |
3131 | * Returns 1<<BDI_async_congested and/or 1<<BDI_sync_congested if we are congested. | |
3132 | */ | |
3133 | static int drbd_congested(void *congested_data, int bdi_bits) | |
3134 | { | |
3135 | struct drbd_conf *mdev = congested_data; | |
3136 | struct request_queue *q; | |
3137 | char reason = '-'; | |
3138 | int r = 0; | |
3139 | ||
3140 | if (!__inc_ap_bio_cond(mdev)) { | |
3141 | /* DRBD has frozen IO */ | |
3142 | r = bdi_bits; | |
3143 | reason = 'd'; | |
3144 | goto out; | |
3145 | } | |
3146 | ||
3147 | if (get_ldev(mdev)) { | |
3148 | q = bdev_get_queue(mdev->ldev->backing_bdev); | |
3149 | r = bdi_congested(&q->backing_dev_info, bdi_bits); | |
3150 | put_ldev(mdev); | |
3151 | if (r) | |
3152 | reason = 'b'; | |
3153 | } | |
3154 | ||
3155 | if (bdi_bits & (1 << BDI_async_congested) && test_bit(NET_CONGESTED, &mdev->flags)) { | |
3156 | r |= (1 << BDI_async_congested); | |
3157 | reason = reason == 'b' ? 'a' : 'n'; | |
3158 | } | |
3159 | ||
3160 | out: | |
3161 | mdev->congestion_reason = reason; | |
3162 | return r; | |
3163 | } | |
3164 | ||
3165 | struct drbd_conf *drbd_new_device(unsigned int minor) | |
3166 | { | |
3167 | struct drbd_conf *mdev; | |
3168 | struct gendisk *disk; | |
3169 | struct request_queue *q; | |
3170 | ||
3171 | /* GFP_KERNEL, we are outside of all write-out paths */ | |
3172 | mdev = kzalloc(sizeof(struct drbd_conf), GFP_KERNEL); | |
3173 | if (!mdev) | |
3174 | return NULL; | |
3175 | if (!zalloc_cpumask_var(&mdev->cpu_mask, GFP_KERNEL)) | |
3176 | goto out_no_cpumask; | |
3177 | ||
3178 | mdev->minor = minor; | |
3179 | ||
3180 | drbd_init_set_defaults(mdev); | |
3181 | ||
3182 | q = blk_alloc_queue(GFP_KERNEL); | |
3183 | if (!q) | |
3184 | goto out_no_q; | |
3185 | mdev->rq_queue = q; | |
3186 | q->queuedata = mdev; | |
b411b363 PR |
3187 | |
3188 | disk = alloc_disk(1); | |
3189 | if (!disk) | |
3190 | goto out_no_disk; | |
3191 | mdev->vdisk = disk; | |
3192 | ||
3193 | set_disk_ro(disk, TRUE); | |
3194 | ||
3195 | disk->queue = q; | |
3196 | disk->major = DRBD_MAJOR; | |
3197 | disk->first_minor = minor; | |
3198 | disk->fops = &drbd_ops; | |
3199 | sprintf(disk->disk_name, "drbd%d", minor); | |
3200 | disk->private_data = mdev; | |
3201 | ||
3202 | mdev->this_bdev = bdget(MKDEV(DRBD_MAJOR, minor)); | |
3203 | /* we have no partitions. we contain only ourselves. */ | |
3204 | mdev->this_bdev->bd_contains = mdev->this_bdev; | |
3205 | ||
3206 | q->backing_dev_info.congested_fn = drbd_congested; | |
3207 | q->backing_dev_info.congested_data = mdev; | |
3208 | ||
3209 | blk_queue_make_request(q, drbd_make_request_26); | |
98ec286e | 3210 | blk_queue_max_segment_size(q, DRBD_MAX_SEGMENT_SIZE); |
b411b363 PR |
3211 | blk_queue_bounce_limit(q, BLK_BOUNCE_ANY); |
3212 | blk_queue_merge_bvec(q, drbd_merge_bvec); | |
3213 | q->queue_lock = &mdev->req_lock; /* needed since we use */ | |
3214 | /* plugging on a queue, that actually has no requests! */ | |
3215 | q->unplug_fn = drbd_unplug_fn; | |
3216 | ||
3217 | mdev->md_io_page = alloc_page(GFP_KERNEL); | |
3218 | if (!mdev->md_io_page) | |
3219 | goto out_no_io_page; | |
3220 | ||
3221 | if (drbd_bm_init(mdev)) | |
3222 | goto out_no_bitmap; | |
3223 | /* no need to lock access, we are still initializing this minor device. */ | |
3224 | if (!tl_init(mdev)) | |
3225 | goto out_no_tl; | |
3226 | ||
3227 | mdev->app_reads_hash = kzalloc(APP_R_HSIZE*sizeof(void *), GFP_KERNEL); | |
3228 | if (!mdev->app_reads_hash) | |
3229 | goto out_no_app_reads; | |
3230 | ||
3231 | mdev->current_epoch = kzalloc(sizeof(struct drbd_epoch), GFP_KERNEL); | |
3232 | if (!mdev->current_epoch) | |
3233 | goto out_no_epoch; | |
3234 | ||
3235 | INIT_LIST_HEAD(&mdev->current_epoch->list); | |
3236 | mdev->epochs = 1; | |
3237 | ||
3238 | return mdev; | |
3239 | ||
3240 | /* out_whatever_else: | |
3241 | kfree(mdev->current_epoch); */ | |
3242 | out_no_epoch: | |
3243 | kfree(mdev->app_reads_hash); | |
3244 | out_no_app_reads: | |
3245 | tl_cleanup(mdev); | |
3246 | out_no_tl: | |
3247 | drbd_bm_cleanup(mdev); | |
3248 | out_no_bitmap: | |
3249 | __free_page(mdev->md_io_page); | |
3250 | out_no_io_page: | |
3251 | put_disk(disk); | |
3252 | out_no_disk: | |
3253 | blk_cleanup_queue(q); | |
3254 | out_no_q: | |
3255 | free_cpumask_var(mdev->cpu_mask); | |
3256 | out_no_cpumask: | |
3257 | kfree(mdev); | |
3258 | return NULL; | |
3259 | } | |
3260 | ||
3261 | /* counterpart of drbd_new_device. | |
3262 | * last part of drbd_delete_device. */ | |
3263 | void drbd_free_mdev(struct drbd_conf *mdev) | |
3264 | { | |
3265 | kfree(mdev->current_epoch); | |
3266 | kfree(mdev->app_reads_hash); | |
3267 | tl_cleanup(mdev); | |
3268 | if (mdev->bitmap) /* should no longer be there. */ | |
3269 | drbd_bm_cleanup(mdev); | |
3270 | __free_page(mdev->md_io_page); | |
3271 | put_disk(mdev->vdisk); | |
3272 | blk_cleanup_queue(mdev->rq_queue); | |
3273 | free_cpumask_var(mdev->cpu_mask); | |
3274 | kfree(mdev); | |
3275 | } | |
3276 | ||
3277 | ||
3278 | int __init drbd_init(void) | |
3279 | { | |
3280 | int err; | |
3281 | ||
3282 | if (sizeof(struct p_handshake) != 80) { | |
3283 | printk(KERN_ERR | |
3284 | "drbd: never change the size or layout " | |
3285 | "of the HandShake packet.\n"); | |
3286 | return -EINVAL; | |
3287 | } | |
3288 | ||
3289 | if (1 > minor_count || minor_count > 255) { | |
3290 | printk(KERN_ERR | |
3291 | "drbd: invalid minor_count (%d)\n", minor_count); | |
3292 | #ifdef MODULE | |
3293 | return -EINVAL; | |
3294 | #else | |
3295 | minor_count = 8; | |
3296 | #endif | |
3297 | } | |
3298 | ||
3299 | err = drbd_nl_init(); | |
3300 | if (err) | |
3301 | return err; | |
3302 | ||
3303 | err = register_blkdev(DRBD_MAJOR, "drbd"); | |
3304 | if (err) { | |
3305 | printk(KERN_ERR | |
3306 | "drbd: unable to register block device major %d\n", | |
3307 | DRBD_MAJOR); | |
3308 | return err; | |
3309 | } | |
3310 | ||
3311 | register_reboot_notifier(&drbd_notifier); | |
3312 | ||
3313 | /* | |
3314 | * allocate all necessary structs | |
3315 | */ | |
3316 | err = -ENOMEM; | |
3317 | ||
3318 | init_waitqueue_head(&drbd_pp_wait); | |
3319 | ||
3320 | drbd_proc = NULL; /* play safe for drbd_cleanup */ | |
3321 | minor_table = kzalloc(sizeof(struct drbd_conf *)*minor_count, | |
3322 | GFP_KERNEL); | |
3323 | if (!minor_table) | |
3324 | goto Enomem; | |
3325 | ||
3326 | err = drbd_create_mempools(); | |
3327 | if (err) | |
3328 | goto Enomem; | |
3329 | ||
8c484ee4 | 3330 | drbd_proc = proc_create_data("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops, NULL); |
b411b363 PR |
3331 | if (!drbd_proc) { |
3332 | printk(KERN_ERR "drbd: unable to register proc file\n"); | |
3333 | goto Enomem; | |
3334 | } | |
3335 | ||
3336 | rwlock_init(&global_state_lock); | |
3337 | ||
3338 | printk(KERN_INFO "drbd: initialized. " | |
3339 | "Version: " REL_VERSION " (api:%d/proto:%d-%d)\n", | |
3340 | API_VERSION, PRO_VERSION_MIN, PRO_VERSION_MAX); | |
3341 | printk(KERN_INFO "drbd: %s\n", drbd_buildtag()); | |
3342 | printk(KERN_INFO "drbd: registered as block device major %d\n", | |
3343 | DRBD_MAJOR); | |
3344 | printk(KERN_INFO "drbd: minor_table @ 0x%p\n", minor_table); | |
3345 | ||
3346 | return 0; /* Success! */ | |
3347 | ||
3348 | Enomem: | |
3349 | drbd_cleanup(); | |
3350 | if (err == -ENOMEM) | |
3351 | /* currently always the case */ | |
3352 | printk(KERN_ERR "drbd: ran out of memory\n"); | |
3353 | else | |
3354 | printk(KERN_ERR "drbd: initialization failure\n"); | |
3355 | return err; | |
3356 | } | |
3357 | ||
3358 | void drbd_free_bc(struct drbd_backing_dev *ldev) | |
3359 | { | |
3360 | if (ldev == NULL) | |
3361 | return; | |
3362 | ||
3363 | bd_release(ldev->backing_bdev); | |
3364 | bd_release(ldev->md_bdev); | |
3365 | ||
3366 | fput(ldev->lo_file); | |
3367 | fput(ldev->md_file); | |
3368 | ||
3369 | kfree(ldev); | |
3370 | } | |
3371 | ||
3372 | void drbd_free_sock(struct drbd_conf *mdev) | |
3373 | { | |
3374 | if (mdev->data.socket) { | |
4589d7f8 | 3375 | mutex_lock(&mdev->data.mutex); |
b411b363 PR |
3376 | kernel_sock_shutdown(mdev->data.socket, SHUT_RDWR); |
3377 | sock_release(mdev->data.socket); | |
3378 | mdev->data.socket = NULL; | |
4589d7f8 | 3379 | mutex_unlock(&mdev->data.mutex); |
b411b363 PR |
3380 | } |
3381 | if (mdev->meta.socket) { | |
4589d7f8 | 3382 | mutex_lock(&mdev->meta.mutex); |
b411b363 PR |
3383 | kernel_sock_shutdown(mdev->meta.socket, SHUT_RDWR); |
3384 | sock_release(mdev->meta.socket); | |
3385 | mdev->meta.socket = NULL; | |
4589d7f8 | 3386 | mutex_unlock(&mdev->meta.mutex); |
b411b363 PR |
3387 | } |
3388 | } | |
3389 | ||
3390 | ||
3391 | void drbd_free_resources(struct drbd_conf *mdev) | |
3392 | { | |
3393 | crypto_free_hash(mdev->csums_tfm); | |
3394 | mdev->csums_tfm = NULL; | |
3395 | crypto_free_hash(mdev->verify_tfm); | |
3396 | mdev->verify_tfm = NULL; | |
3397 | crypto_free_hash(mdev->cram_hmac_tfm); | |
3398 | mdev->cram_hmac_tfm = NULL; | |
3399 | crypto_free_hash(mdev->integrity_w_tfm); | |
3400 | mdev->integrity_w_tfm = NULL; | |
3401 | crypto_free_hash(mdev->integrity_r_tfm); | |
3402 | mdev->integrity_r_tfm = NULL; | |
3403 | ||
3404 | drbd_free_sock(mdev); | |
3405 | ||
3406 | __no_warn(local, | |
3407 | drbd_free_bc(mdev->ldev); | |
3408 | mdev->ldev = NULL;); | |
3409 | } | |
3410 | ||
3411 | /* meta data management */ | |
3412 | ||
3413 | struct meta_data_on_disk { | |
3414 | u64 la_size; /* last agreed size. */ | |
3415 | u64 uuid[UI_SIZE]; /* UUIDs. */ | |
3416 | u64 device_uuid; | |
3417 | u64 reserved_u64_1; | |
3418 | u32 flags; /* MDF */ | |
3419 | u32 magic; | |
3420 | u32 md_size_sect; | |
3421 | u32 al_offset; /* offset to this block */ | |
3422 | u32 al_nr_extents; /* important for restoring the AL */ | |
3423 | /* `-- act_log->nr_elements <-- sync_conf.al_extents */ | |
3424 | u32 bm_offset; /* offset to the bitmap, from here */ | |
3425 | u32 bm_bytes_per_bit; /* BM_BLOCK_SIZE */ | |
3426 | u32 reserved_u32[4]; | |
3427 | ||
3428 | } __packed; | |
3429 | ||
3430 | /** | |
3431 | * drbd_md_sync() - Writes the meta data super block if the MD_DIRTY flag bit is set | |
3432 | * @mdev: DRBD device. | |
3433 | */ | |
3434 | void drbd_md_sync(struct drbd_conf *mdev) | |
3435 | { | |
3436 | struct meta_data_on_disk *buffer; | |
3437 | sector_t sector; | |
3438 | int i; | |
3439 | ||
ee15b038 LE |
3440 | del_timer(&mdev->md_sync_timer); |
3441 | /* timer may be rearmed by drbd_md_mark_dirty() now. */ | |
b411b363 PR |
3442 | if (!test_and_clear_bit(MD_DIRTY, &mdev->flags)) |
3443 | return; | |
b411b363 PR |
3444 | |
3445 | /* We use here D_FAILED and not D_ATTACHING because we try to write | |
3446 | * metadata even if we detach due to a disk failure! */ | |
3447 | if (!get_ldev_if_state(mdev, D_FAILED)) | |
3448 | return; | |
3449 | ||
b411b363 PR |
3450 | mutex_lock(&mdev->md_io_mutex); |
3451 | buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page); | |
3452 | memset(buffer, 0, 512); | |
3453 | ||
3454 | buffer->la_size = cpu_to_be64(drbd_get_capacity(mdev->this_bdev)); | |
3455 | for (i = UI_CURRENT; i < UI_SIZE; i++) | |
3456 | buffer->uuid[i] = cpu_to_be64(mdev->ldev->md.uuid[i]); | |
3457 | buffer->flags = cpu_to_be32(mdev->ldev->md.flags); | |
3458 | buffer->magic = cpu_to_be32(DRBD_MD_MAGIC); | |
3459 | ||
3460 | buffer->md_size_sect = cpu_to_be32(mdev->ldev->md.md_size_sect); | |
3461 | buffer->al_offset = cpu_to_be32(mdev->ldev->md.al_offset); | |
3462 | buffer->al_nr_extents = cpu_to_be32(mdev->act_log->nr_elements); | |
3463 | buffer->bm_bytes_per_bit = cpu_to_be32(BM_BLOCK_SIZE); | |
3464 | buffer->device_uuid = cpu_to_be64(mdev->ldev->md.device_uuid); | |
3465 | ||
3466 | buffer->bm_offset = cpu_to_be32(mdev->ldev->md.bm_offset); | |
3467 | ||
3468 | D_ASSERT(drbd_md_ss__(mdev, mdev->ldev) == mdev->ldev->md.md_offset); | |
3469 | sector = mdev->ldev->md.md_offset; | |
3470 | ||
3f3a9b84 | 3471 | if (!drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) { |
b411b363 PR |
3472 | /* this was a try anyways ... */ |
3473 | dev_err(DEV, "meta data update failed!\n"); | |
b411b363 PR |
3474 | drbd_chk_io_error(mdev, 1, TRUE); |
3475 | } | |
3476 | ||
3477 | /* Update mdev->ldev->md.la_size_sect, | |
3478 | * since we updated it on metadata. */ | |
3479 | mdev->ldev->md.la_size_sect = drbd_get_capacity(mdev->this_bdev); | |
3480 | ||
3481 | mutex_unlock(&mdev->md_io_mutex); | |
3482 | put_ldev(mdev); | |
3483 | } | |
3484 | ||
3485 | /** | |
3486 | * drbd_md_read() - Reads in the meta data super block | |
3487 | * @mdev: DRBD device. | |
3488 | * @bdev: Device from which the meta data should be read in. | |
3489 | * | |
3490 | * Return 0 (NO_ERROR) on success, and an enum drbd_ret_codes in case | |
3491 | * something goes wrong. Currently only: ERR_IO_MD_DISK, ERR_MD_INVALID. | |
3492 | */ | |
3493 | int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) | |
3494 | { | |
3495 | struct meta_data_on_disk *buffer; | |
3496 | int i, rv = NO_ERROR; | |
3497 | ||
3498 | if (!get_ldev_if_state(mdev, D_ATTACHING)) | |
3499 | return ERR_IO_MD_DISK; | |
3500 | ||
b411b363 PR |
3501 | mutex_lock(&mdev->md_io_mutex); |
3502 | buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page); | |
3503 | ||
3504 | if (!drbd_md_sync_page_io(mdev, bdev, bdev->md.md_offset, READ)) { | |
3505 | /* NOTE: cant do normal error processing here as this is | |
3506 | called BEFORE disk is attached */ | |
3507 | dev_err(DEV, "Error while reading metadata.\n"); | |
3508 | rv = ERR_IO_MD_DISK; | |
3509 | goto err; | |
3510 | } | |
3511 | ||
3512 | if (be32_to_cpu(buffer->magic) != DRBD_MD_MAGIC) { | |
3513 | dev_err(DEV, "Error while reading metadata, magic not found.\n"); | |
3514 | rv = ERR_MD_INVALID; | |
3515 | goto err; | |
3516 | } | |
3517 | if (be32_to_cpu(buffer->al_offset) != bdev->md.al_offset) { | |
3518 | dev_err(DEV, "unexpected al_offset: %d (expected %d)\n", | |
3519 | be32_to_cpu(buffer->al_offset), bdev->md.al_offset); | |
3520 | rv = ERR_MD_INVALID; | |
3521 | goto err; | |
3522 | } | |
3523 | if (be32_to_cpu(buffer->bm_offset) != bdev->md.bm_offset) { | |
3524 | dev_err(DEV, "unexpected bm_offset: %d (expected %d)\n", | |
3525 | be32_to_cpu(buffer->bm_offset), bdev->md.bm_offset); | |
3526 | rv = ERR_MD_INVALID; | |
3527 | goto err; | |
3528 | } | |
3529 | if (be32_to_cpu(buffer->md_size_sect) != bdev->md.md_size_sect) { | |
3530 | dev_err(DEV, "unexpected md_size: %u (expected %u)\n", | |
3531 | be32_to_cpu(buffer->md_size_sect), bdev->md.md_size_sect); | |
3532 | rv = ERR_MD_INVALID; | |
3533 | goto err; | |
3534 | } | |
3535 | ||
3536 | if (be32_to_cpu(buffer->bm_bytes_per_bit) != BM_BLOCK_SIZE) { | |
3537 | dev_err(DEV, "unexpected bm_bytes_per_bit: %u (expected %u)\n", | |
3538 | be32_to_cpu(buffer->bm_bytes_per_bit), BM_BLOCK_SIZE); | |
3539 | rv = ERR_MD_INVALID; | |
3540 | goto err; | |
3541 | } | |
3542 | ||
3543 | bdev->md.la_size_sect = be64_to_cpu(buffer->la_size); | |
3544 | for (i = UI_CURRENT; i < UI_SIZE; i++) | |
3545 | bdev->md.uuid[i] = be64_to_cpu(buffer->uuid[i]); | |
3546 | bdev->md.flags = be32_to_cpu(buffer->flags); | |
3547 | mdev->sync_conf.al_extents = be32_to_cpu(buffer->al_nr_extents); | |
3548 | bdev->md.device_uuid = be64_to_cpu(buffer->device_uuid); | |
3549 | ||
3550 | if (mdev->sync_conf.al_extents < 7) | |
3551 | mdev->sync_conf.al_extents = 127; | |
3552 | ||
3553 | err: | |
3554 | mutex_unlock(&mdev->md_io_mutex); | |
3555 | put_ldev(mdev); | |
3556 | ||
3557 | return rv; | |
3558 | } | |
3559 | ||
3560 | /** | |
3561 | * drbd_md_mark_dirty() - Mark meta data super block as dirty | |
3562 | * @mdev: DRBD device. | |
3563 | * | |
3564 | * Call this function if you change anything that should be written to | |
3565 | * the meta-data super block. This function sets MD_DIRTY, and starts a | |
3566 | * timer that ensures that within five seconds you have to call drbd_md_sync(). | |
3567 | */ | |
ca0e6098 | 3568 | #ifdef DEBUG |
ee15b038 LE |
3569 | void drbd_md_mark_dirty_(struct drbd_conf *mdev, unsigned int line, const char *func) |
3570 | { | |
3571 | if (!test_and_set_bit(MD_DIRTY, &mdev->flags)) { | |
3572 | mod_timer(&mdev->md_sync_timer, jiffies + HZ); | |
3573 | mdev->last_md_mark_dirty.line = line; | |
3574 | mdev->last_md_mark_dirty.func = func; | |
3575 | } | |
3576 | } | |
3577 | #else | |
b411b363 PR |
3578 | void drbd_md_mark_dirty(struct drbd_conf *mdev) |
3579 | { | |
ee15b038 | 3580 | if (!test_and_set_bit(MD_DIRTY, &mdev->flags)) |
ca0e6098 | 3581 | mod_timer(&mdev->md_sync_timer, jiffies + 5*HZ); |
b411b363 | 3582 | } |
ee15b038 | 3583 | #endif |
b411b363 PR |
3584 | |
3585 | static void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local) | |
3586 | { | |
3587 | int i; | |
3588 | ||
6a0afdf5 | 3589 | for (i = UI_HISTORY_START; i < UI_HISTORY_END; i++) |
b411b363 | 3590 | mdev->ldev->md.uuid[i+1] = mdev->ldev->md.uuid[i]; |
b411b363 PR |
3591 | } |
3592 | ||
3593 | void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local) | |
3594 | { | |
3595 | if (idx == UI_CURRENT) { | |
3596 | if (mdev->state.role == R_PRIMARY) | |
3597 | val |= 1; | |
3598 | else | |
3599 | val &= ~((u64)1); | |
3600 | ||
3601 | drbd_set_ed_uuid(mdev, val); | |
3602 | } | |
3603 | ||
3604 | mdev->ldev->md.uuid[idx] = val; | |
b411b363 PR |
3605 | drbd_md_mark_dirty(mdev); |
3606 | } | |
3607 | ||
3608 | ||
3609 | void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local) | |
3610 | { | |
3611 | if (mdev->ldev->md.uuid[idx]) { | |
3612 | drbd_uuid_move_history(mdev); | |
3613 | mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[idx]; | |
b411b363 PR |
3614 | } |
3615 | _drbd_uuid_set(mdev, idx, val); | |
3616 | } | |
3617 | ||
3618 | /** | |
3619 | * drbd_uuid_new_current() - Creates a new current UUID | |
3620 | * @mdev: DRBD device. | |
3621 | * | |
3622 | * Creates a new current UUID, and rotates the old current UUID into | |
3623 | * the bitmap slot. Causes an incremental resync upon next connect. | |
3624 | */ | |
3625 | void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local) | |
3626 | { | |
3627 | u64 val; | |
3628 | ||
3629 | dev_info(DEV, "Creating new current UUID\n"); | |
3630 | D_ASSERT(mdev->ldev->md.uuid[UI_BITMAP] == 0); | |
3631 | mdev->ldev->md.uuid[UI_BITMAP] = mdev->ldev->md.uuid[UI_CURRENT]; | |
b411b363 PR |
3632 | |
3633 | get_random_bytes(&val, sizeof(u64)); | |
3634 | _drbd_uuid_set(mdev, UI_CURRENT, val); | |
3635 | } | |
3636 | ||
3637 | void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local) | |
3638 | { | |
3639 | if (mdev->ldev->md.uuid[UI_BITMAP] == 0 && val == 0) | |
3640 | return; | |
3641 | ||
3642 | if (val == 0) { | |
3643 | drbd_uuid_move_history(mdev); | |
3644 | mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP]; | |
3645 | mdev->ldev->md.uuid[UI_BITMAP] = 0; | |
b411b363 PR |
3646 | } else { |
3647 | if (mdev->ldev->md.uuid[UI_BITMAP]) | |
3648 | dev_warn(DEV, "bm UUID already set"); | |
3649 | ||
3650 | mdev->ldev->md.uuid[UI_BITMAP] = val; | |
3651 | mdev->ldev->md.uuid[UI_BITMAP] &= ~((u64)1); | |
3652 | ||
b411b363 PR |
3653 | } |
3654 | drbd_md_mark_dirty(mdev); | |
3655 | } | |
3656 | ||
3657 | /** | |
3658 | * drbd_bmio_set_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io() | |
3659 | * @mdev: DRBD device. | |
3660 | * | |
3661 | * Sets all bits in the bitmap and writes the whole bitmap to stable storage. | |
3662 | */ | |
3663 | int drbd_bmio_set_n_write(struct drbd_conf *mdev) | |
3664 | { | |
3665 | int rv = -EIO; | |
3666 | ||
3667 | if (get_ldev_if_state(mdev, D_ATTACHING)) { | |
3668 | drbd_md_set_flag(mdev, MDF_FULL_SYNC); | |
3669 | drbd_md_sync(mdev); | |
3670 | drbd_bm_set_all(mdev); | |
3671 | ||
3672 | rv = drbd_bm_write(mdev); | |
3673 | ||
3674 | if (!rv) { | |
3675 | drbd_md_clear_flag(mdev, MDF_FULL_SYNC); | |
3676 | drbd_md_sync(mdev); | |
3677 | } | |
3678 | ||
3679 | put_ldev(mdev); | |
3680 | } | |
3681 | ||
3682 | return rv; | |
3683 | } | |
3684 | ||
3685 | /** | |
3686 | * drbd_bmio_clear_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io() | |
3687 | * @mdev: DRBD device. | |
3688 | * | |
3689 | * Clears all bits in the bitmap and writes the whole bitmap to stable storage. | |
3690 | */ | |
3691 | int drbd_bmio_clear_n_write(struct drbd_conf *mdev) | |
3692 | { | |
3693 | int rv = -EIO; | |
3694 | ||
0778286a | 3695 | drbd_resume_al(mdev); |
b411b363 PR |
3696 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
3697 | drbd_bm_clear_all(mdev); | |
3698 | rv = drbd_bm_write(mdev); | |
3699 | put_ldev(mdev); | |
3700 | } | |
3701 | ||
3702 | return rv; | |
3703 | } | |
3704 | ||
3705 | static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused) | |
3706 | { | |
3707 | struct bm_io_work *work = container_of(w, struct bm_io_work, w); | |
3708 | int rv; | |
3709 | ||
3710 | D_ASSERT(atomic_read(&mdev->ap_bio_cnt) == 0); | |
3711 | ||
3712 | drbd_bm_lock(mdev, work->why); | |
3713 | rv = work->io_fn(mdev); | |
3714 | drbd_bm_unlock(mdev); | |
3715 | ||
3716 | clear_bit(BITMAP_IO, &mdev->flags); | |
3717 | wake_up(&mdev->misc_wait); | |
3718 | ||
3719 | if (work->done) | |
3720 | work->done(mdev, rv); | |
3721 | ||
3722 | clear_bit(BITMAP_IO_QUEUED, &mdev->flags); | |
3723 | work->why = NULL; | |
3724 | ||
3725 | return 1; | |
3726 | } | |
3727 | ||
e9e6f3ec LE |
3728 | static int w_go_diskless(struct drbd_conf *mdev, struct drbd_work *w, int unused) |
3729 | { | |
3730 | D_ASSERT(mdev->state.disk == D_FAILED); | |
9d282875 LE |
3731 | /* we cannot assert local_cnt == 0 here, as get_ldev_if_state will |
3732 | * inc/dec it frequently. Once we are D_DISKLESS, no one will touch | |
3733 | * the protected members anymore, though, so in the after_state_ch work | |
3734 | * it will be safe to free them. */ | |
e9e6f3ec LE |
3735 | drbd_force_state(mdev, NS(disk, D_DISKLESS)); |
3736 | ||
3737 | clear_bit(GO_DISKLESS, &mdev->flags); | |
3738 | return 1; | |
3739 | } | |
3740 | ||
3741 | void drbd_go_diskless(struct drbd_conf *mdev) | |
3742 | { | |
3743 | D_ASSERT(mdev->state.disk == D_FAILED); | |
3744 | if (!test_and_set_bit(GO_DISKLESS, &mdev->flags)) | |
9d282875 LE |
3745 | drbd_queue_work(&mdev->data.work, &mdev->go_diskless); |
3746 | /* don't drbd_queue_work_front, | |
3747 | * we need to serialize with the after_state_ch work | |
3748 | * of the -> D_FAILED transition. */ | |
e9e6f3ec LE |
3749 | } |
3750 | ||
b411b363 PR |
3751 | /** |
3752 | * drbd_queue_bitmap_io() - Queues an IO operation on the whole bitmap | |
3753 | * @mdev: DRBD device. | |
3754 | * @io_fn: IO callback to be called when bitmap IO is possible | |
3755 | * @done: callback to be called after the bitmap IO was performed | |
3756 | * @why: Descriptive text of the reason for doing the IO | |
3757 | * | |
3758 | * While IO on the bitmap happens we freeze application IO thus we ensure | |
3759 | * that drbd_set_out_of_sync() can not be called. This function MAY ONLY be | |
3760 | * called from worker context. It MUST NOT be used while a previous such | |
3761 | * work is still pending! | |
3762 | */ | |
3763 | void drbd_queue_bitmap_io(struct drbd_conf *mdev, | |
3764 | int (*io_fn)(struct drbd_conf *), | |
3765 | void (*done)(struct drbd_conf *, int), | |
3766 | char *why) | |
3767 | { | |
3768 | D_ASSERT(current == mdev->worker.task); | |
3769 | ||
3770 | D_ASSERT(!test_bit(BITMAP_IO_QUEUED, &mdev->flags)); | |
3771 | D_ASSERT(!test_bit(BITMAP_IO, &mdev->flags)); | |
3772 | D_ASSERT(list_empty(&mdev->bm_io_work.w.list)); | |
3773 | if (mdev->bm_io_work.why) | |
3774 | dev_err(DEV, "FIXME going to queue '%s' but '%s' still pending?\n", | |
3775 | why, mdev->bm_io_work.why); | |
3776 | ||
3777 | mdev->bm_io_work.io_fn = io_fn; | |
3778 | mdev->bm_io_work.done = done; | |
3779 | mdev->bm_io_work.why = why; | |
3780 | ||
3781 | set_bit(BITMAP_IO, &mdev->flags); | |
3782 | if (atomic_read(&mdev->ap_bio_cnt) == 0) { | |
3783 | if (list_empty(&mdev->bm_io_work.w.list)) { | |
3784 | set_bit(BITMAP_IO_QUEUED, &mdev->flags); | |
3785 | drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w); | |
3786 | } else | |
3787 | dev_err(DEV, "FIXME avoided double queuing bm_io_work\n"); | |
3788 | } | |
3789 | } | |
3790 | ||
3791 | /** | |
3792 | * drbd_bitmap_io() - Does an IO operation on the whole bitmap | |
3793 | * @mdev: DRBD device. | |
3794 | * @io_fn: IO callback to be called when bitmap IO is possible | |
3795 | * @why: Descriptive text of the reason for doing the IO | |
3796 | * | |
3797 | * freezes application IO while that the actual IO operations runs. This | |
3798 | * functions MAY NOT be called from worker context. | |
3799 | */ | |
3800 | int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why) | |
3801 | { | |
3802 | int rv; | |
3803 | ||
3804 | D_ASSERT(current != mdev->worker.task); | |
3805 | ||
3806 | drbd_suspend_io(mdev); | |
3807 | ||
3808 | drbd_bm_lock(mdev, why); | |
3809 | rv = io_fn(mdev); | |
3810 | drbd_bm_unlock(mdev); | |
3811 | ||
3812 | drbd_resume_io(mdev); | |
3813 | ||
3814 | return rv; | |
3815 | } | |
3816 | ||
3817 | void drbd_md_set_flag(struct drbd_conf *mdev, int flag) __must_hold(local) | |
3818 | { | |
3819 | if ((mdev->ldev->md.flags & flag) != flag) { | |
3820 | drbd_md_mark_dirty(mdev); | |
3821 | mdev->ldev->md.flags |= flag; | |
3822 | } | |
3823 | } | |
3824 | ||
3825 | void drbd_md_clear_flag(struct drbd_conf *mdev, int flag) __must_hold(local) | |
3826 | { | |
3827 | if ((mdev->ldev->md.flags & flag) != 0) { | |
3828 | drbd_md_mark_dirty(mdev); | |
3829 | mdev->ldev->md.flags &= ~flag; | |
3830 | } | |
3831 | } | |
3832 | int drbd_md_test_flag(struct drbd_backing_dev *bdev, int flag) | |
3833 | { | |
3834 | return (bdev->md.flags & flag) != 0; | |
3835 | } | |
3836 | ||
3837 | static void md_sync_timer_fn(unsigned long data) | |
3838 | { | |
3839 | struct drbd_conf *mdev = (struct drbd_conf *) data; | |
3840 | ||
3841 | drbd_queue_work_front(&mdev->data.work, &mdev->md_sync_work); | |
3842 | } | |
3843 | ||
3844 | static int w_md_sync(struct drbd_conf *mdev, struct drbd_work *w, int unused) | |
3845 | { | |
3846 | dev_warn(DEV, "md_sync_timer expired! Worker calls drbd_md_sync().\n"); | |
ee15b038 LE |
3847 | #ifdef DEBUG |
3848 | dev_warn(DEV, "last md_mark_dirty: %s:%u\n", | |
3849 | mdev->last_md_mark_dirty.func, mdev->last_md_mark_dirty.line); | |
3850 | #endif | |
b411b363 | 3851 | drbd_md_sync(mdev); |
b411b363 PR |
3852 | return 1; |
3853 | } | |
3854 | ||
3855 | #ifdef CONFIG_DRBD_FAULT_INJECTION | |
3856 | /* Fault insertion support including random number generator shamelessly | |
3857 | * stolen from kernel/rcutorture.c */ | |
3858 | struct fault_random_state { | |
3859 | unsigned long state; | |
3860 | unsigned long count; | |
3861 | }; | |
3862 | ||
3863 | #define FAULT_RANDOM_MULT 39916801 /* prime */ | |
3864 | #define FAULT_RANDOM_ADD 479001701 /* prime */ | |
3865 | #define FAULT_RANDOM_REFRESH 10000 | |
3866 | ||
3867 | /* | |
3868 | * Crude but fast random-number generator. Uses a linear congruential | |
3869 | * generator, with occasional help from get_random_bytes(). | |
3870 | */ | |
3871 | static unsigned long | |
3872 | _drbd_fault_random(struct fault_random_state *rsp) | |
3873 | { | |
3874 | long refresh; | |
3875 | ||
49829ea7 | 3876 | if (!rsp->count--) { |
b411b363 PR |
3877 | get_random_bytes(&refresh, sizeof(refresh)); |
3878 | rsp->state += refresh; | |
3879 | rsp->count = FAULT_RANDOM_REFRESH; | |
3880 | } | |
3881 | rsp->state = rsp->state * FAULT_RANDOM_MULT + FAULT_RANDOM_ADD; | |
3882 | return swahw32(rsp->state); | |
3883 | } | |
3884 | ||
3885 | static char * | |
3886 | _drbd_fault_str(unsigned int type) { | |
3887 | static char *_faults[] = { | |
3888 | [DRBD_FAULT_MD_WR] = "Meta-data write", | |
3889 | [DRBD_FAULT_MD_RD] = "Meta-data read", | |
3890 | [DRBD_FAULT_RS_WR] = "Resync write", | |
3891 | [DRBD_FAULT_RS_RD] = "Resync read", | |
3892 | [DRBD_FAULT_DT_WR] = "Data write", | |
3893 | [DRBD_FAULT_DT_RD] = "Data read", | |
3894 | [DRBD_FAULT_DT_RA] = "Data read ahead", | |
3895 | [DRBD_FAULT_BM_ALLOC] = "BM allocation", | |
6b4388ac PR |
3896 | [DRBD_FAULT_AL_EE] = "EE allocation", |
3897 | [DRBD_FAULT_RECEIVE] = "receive data corruption", | |
b411b363 PR |
3898 | }; |
3899 | ||
3900 | return (type < DRBD_FAULT_MAX) ? _faults[type] : "**Unknown**"; | |
3901 | } | |
3902 | ||
3903 | unsigned int | |
3904 | _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) | |
3905 | { | |
3906 | static struct fault_random_state rrs = {0, 0}; | |
3907 | ||
3908 | unsigned int ret = ( | |
3909 | (fault_devs == 0 || | |
3910 | ((1 << mdev_to_minor(mdev)) & fault_devs) != 0) && | |
3911 | (((_drbd_fault_random(&rrs) % 100) + 1) <= fault_rate)); | |
3912 | ||
3913 | if (ret) { | |
3914 | fault_count++; | |
3915 | ||
7383506c | 3916 | if (__ratelimit(&drbd_ratelimit_state)) |
b411b363 PR |
3917 | dev_warn(DEV, "***Simulating %s failure\n", |
3918 | _drbd_fault_str(type)); | |
3919 | } | |
3920 | ||
3921 | return ret; | |
3922 | } | |
3923 | #endif | |
3924 | ||
3925 | const char *drbd_buildtag(void) | |
3926 | { | |
3927 | /* DRBD built from external sources has here a reference to the | |
3928 | git hash of the source code. */ | |
3929 | ||
3930 | static char buildtag[38] = "\0uilt-in"; | |
3931 | ||
3932 | if (buildtag[0] == 0) { | |
3933 | #ifdef CONFIG_MODULES | |
3934 | if (THIS_MODULE != NULL) | |
3935 | sprintf(buildtag, "srcversion: %-24s", THIS_MODULE->srcversion); | |
3936 | else | |
3937 | #endif | |
3938 | buildtag[0] = 'b'; | |
3939 | } | |
3940 | ||
3941 | return buildtag; | |
3942 | } | |
3943 | ||
3944 | module_init(drbd_init) | |
3945 | module_exit(drbd_cleanup) | |
3946 | ||
b411b363 PR |
3947 | EXPORT_SYMBOL(drbd_conn_str); |
3948 | EXPORT_SYMBOL(drbd_role_str); | |
3949 | EXPORT_SYMBOL(drbd_disk_str); | |
3950 | EXPORT_SYMBOL(drbd_set_st_err_str); |