4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
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>.
10 Thanks to Carter Burden, Bart Grantham and Gennadiy Nerubayev
11 from Logicworks, Inc. for making SDP replication support possible.
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)
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.
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.
29 #include <linux/module.h>
30 #include <linux/drbd.h>
31 #include <asm/uaccess.h>
32 #include <asm/types.h>
34 #include <linux/ctype.h>
35 #include <linux/smp_lock.h>
37 #include <linux/file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/init.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>
49 #define __KERNEL_SYSCALLS__
50 #include <linux/unistd.h>
51 #include <linux/vmalloc.h>
53 #include <linux/drbd_limits.h>
55 #include "drbd_req.h" /* only for _req_mod in tl_release and tl_clear */
59 struct after_state_chg_work {
63 enum chg_state_flags flags;
64 struct completion *done;
67 int drbdd_init(struct drbd_thread *);
68 int drbd_worker(struct drbd_thread *);
69 int drbd_asender(struct drbd_thread *);
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);
81 MODULE_AUTHOR("Philipp Reisner <phil@linbit.com>, "
82 "Lars Ellenberg <lars@linbit.com>");
83 MODULE_DESCRIPTION("drbd - Distributed Replicated Block Device v" REL_VERSION);
84 MODULE_VERSION(REL_VERSION);
85 MODULE_LICENSE("GPL");
86 MODULE_PARM_DESC(minor_count, "Maximum number of drbd devices (1-255)");
87 MODULE_ALIAS_BLOCKDEV_MAJOR(DRBD_MAJOR);
89 #include <linux/moduleparam.h>
90 /* allow_open_on_secondary */
91 MODULE_PARM_DESC(allow_oos, "DONT USE!");
92 /* thanks to these macros, if compiled into the kernel (not-module),
93 * this becomes the boot parameter drbd.minor_count */
94 module_param(minor_count, uint, 0444);
95 module_param(disable_sendpage, bool, 0644);
96 module_param(allow_oos, bool, 0);
97 module_param(cn_idx, uint, 0444);
98 module_param(proc_details, int, 0644);
100 #ifdef CONFIG_DRBD_FAULT_INJECTION
103 static int fault_count;
105 /* bitmap of enabled faults */
106 module_param(enable_faults, int, 0664);
107 /* fault rate % value - applies to all enabled faults */
108 module_param(fault_rate, int, 0664);
109 /* count of faults inserted */
110 module_param(fault_count, int, 0664);
111 /* bitmap of devices to insert faults on */
112 module_param(fault_devs, int, 0644);
115 /* module parameter, defined */
116 unsigned int minor_count = 32;
117 int disable_sendpage;
119 unsigned int cn_idx = CN_IDX_DRBD;
120 int proc_details; /* Detail level in proc drbd*/
122 /* Module parameter for setting the user mode helper program
123 * to run. Default is /sbin/drbdadm */
124 char usermode_helper[80] = "/sbin/drbdadm";
126 module_param_string(usermode_helper, usermode_helper, sizeof(usermode_helper), 0644);
128 /* in 2.6.x, our device mapping and config info contains our virtual gendisks
129 * as member "struct gendisk *vdisk;"
131 struct drbd_conf **minor_table;
133 struct kmem_cache *drbd_request_cache;
134 struct kmem_cache *drbd_ee_cache; /* epoch entries */
135 struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
136 struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
137 mempool_t *drbd_request_mempool;
138 mempool_t *drbd_ee_mempool;
140 /* I do not use a standard mempool, because:
141 1) I want to hand out the pre-allocated objects first.
142 2) I want to be able to interrupt sleeping allocation with a signal.
143 Note: This is a single linked list, the next pointer is the private
144 member of struct page.
146 struct page *drbd_pp_pool;
147 spinlock_t drbd_pp_lock;
149 wait_queue_head_t drbd_pp_wait;
151 DEFINE_RATELIMIT_STATE(drbd_ratelimit_state, 5 * HZ, 5);
153 static const struct block_device_operations drbd_ops = {
154 .owner = THIS_MODULE,
156 .release = drbd_release,
159 #define ARRY_SIZE(A) (sizeof(A)/sizeof(A[0]))
162 /* When checking with sparse, and this is an inline function, sparse will
163 give tons of false positives. When this is a real functions sparse works.
165 int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
169 atomic_inc(&mdev->local_cnt);
170 io_allowed = (mdev->state.disk >= mins);
172 if (atomic_dec_and_test(&mdev->local_cnt))
173 wake_up(&mdev->misc_wait);
181 * DOC: The transfer log
183 * The transfer log is a single linked list of &struct drbd_tl_epoch objects.
184 * mdev->newest_tle points to the head, mdev->oldest_tle points to the tail
185 * of the list. There is always at least one &struct drbd_tl_epoch object.
187 * Each &struct drbd_tl_epoch has a circular double linked list of requests
190 static int tl_init(struct drbd_conf *mdev)
192 struct drbd_tl_epoch *b;
194 /* during device minor initialization, we may well use GFP_KERNEL */
195 b = kmalloc(sizeof(struct drbd_tl_epoch), GFP_KERNEL);
198 INIT_LIST_HEAD(&b->requests);
199 INIT_LIST_HEAD(&b->w.list);
203 b->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */
205 mdev->oldest_tle = b;
206 mdev->newest_tle = b;
207 INIT_LIST_HEAD(&mdev->out_of_sequence_requests);
209 mdev->tl_hash = NULL;
215 static void tl_cleanup(struct drbd_conf *mdev)
217 D_ASSERT(mdev->oldest_tle == mdev->newest_tle);
218 D_ASSERT(list_empty(&mdev->out_of_sequence_requests));
219 kfree(mdev->oldest_tle);
220 mdev->oldest_tle = NULL;
221 kfree(mdev->unused_spare_tle);
222 mdev->unused_spare_tle = NULL;
223 kfree(mdev->tl_hash);
224 mdev->tl_hash = NULL;
229 * _tl_add_barrier() - Adds a barrier to the transfer log
230 * @mdev: DRBD device.
231 * @new: Barrier to be added before the current head of the TL.
233 * The caller must hold the req_lock.
235 void _tl_add_barrier(struct drbd_conf *mdev, struct drbd_tl_epoch *new)
237 struct drbd_tl_epoch *newest_before;
239 INIT_LIST_HEAD(&new->requests);
240 INIT_LIST_HEAD(&new->w.list);
241 new->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */
245 newest_before = mdev->newest_tle;
246 /* never send a barrier number == 0, because that is special-cased
247 * when using TCQ for our write ordering code */
248 new->br_number = (newest_before->br_number+1) ?: 1;
249 if (mdev->newest_tle != new) {
250 mdev->newest_tle->next = new;
251 mdev->newest_tle = new;
256 * tl_release() - Free or recycle the oldest &struct drbd_tl_epoch object of the TL
257 * @mdev: DRBD device.
258 * @barrier_nr: Expected identifier of the DRBD write barrier packet.
259 * @set_size: Expected number of requests before that barrier.
261 * In case the passed barrier_nr or set_size does not match the oldest
262 * &struct drbd_tl_epoch objects this function will cause a termination
265 void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
266 unsigned int set_size)
268 struct drbd_tl_epoch *b, *nob; /* next old barrier */
269 struct list_head *le, *tle;
270 struct drbd_request *r;
272 spin_lock_irq(&mdev->req_lock);
274 b = mdev->oldest_tle;
276 /* first some paranoia code */
278 dev_err(DEV, "BAD! BarrierAck #%u received, but no epoch in tl!?\n",
282 if (b->br_number != barrier_nr) {
283 dev_err(DEV, "BAD! BarrierAck #%u received, expected #%u!\n",
284 barrier_nr, b->br_number);
287 if (b->n_req != set_size) {
288 dev_err(DEV, "BAD! BarrierAck #%u received with n_req=%u, expected n_req=%u!\n",
289 barrier_nr, set_size, b->n_req);
293 /* Clean up list of requests processed during current epoch */
294 list_for_each_safe(le, tle, &b->requests) {
295 r = list_entry(le, struct drbd_request, tl_requests);
296 _req_mod(r, barrier_acked);
298 /* There could be requests on the list waiting for completion
299 of the write to the local disk. To avoid corruptions of
300 slab's data structures we have to remove the lists head.
302 Also there could have been a barrier ack out of sequence, overtaking
303 the write acks - which would be a bug and violating write ordering.
304 To not deadlock in case we lose connection while such requests are
305 still pending, we need some way to find them for the
306 _req_mode(connection_lost_while_pending).
308 These have been list_move'd to the out_of_sequence_requests list in
309 _req_mod(, barrier_acked) above.
311 list_del_init(&b->requests);
314 if (test_and_clear_bit(CREATE_BARRIER, &mdev->flags)) {
315 _tl_add_barrier(mdev, b);
317 mdev->oldest_tle = nob;
318 /* if nob == NULL b was the only barrier, and becomes the new
319 barrier. Therefore mdev->oldest_tle points already to b */
321 D_ASSERT(nob != NULL);
322 mdev->oldest_tle = nob;
326 spin_unlock_irq(&mdev->req_lock);
327 dec_ap_pending(mdev);
332 spin_unlock_irq(&mdev->req_lock);
333 drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR));
338 * tl_clear() - Clears all requests and &struct drbd_tl_epoch objects out of the TL
339 * @mdev: DRBD device.
341 * This is called after the connection to the peer was lost. The storage covered
342 * by the requests on the transfer gets marked as our of sync. Called from the
343 * receiver thread and the worker thread.
345 void tl_clear(struct drbd_conf *mdev)
347 struct drbd_tl_epoch *b, *tmp;
348 struct list_head *le, *tle;
349 struct drbd_request *r;
350 int new_initial_bnr = net_random();
352 spin_lock_irq(&mdev->req_lock);
354 b = mdev->oldest_tle;
356 list_for_each_safe(le, tle, &b->requests) {
357 r = list_entry(le, struct drbd_request, tl_requests);
358 /* It would be nice to complete outside of spinlock.
359 * But this is easier for now. */
360 _req_mod(r, connection_lost_while_pending);
364 /* there could still be requests on that ring list,
365 * in case local io is still pending */
366 list_del(&b->requests);
368 /* dec_ap_pending corresponding to queue_barrier.
369 * the newest barrier may not have been queued yet,
370 * in which case w.cb is still NULL. */
372 dec_ap_pending(mdev);
374 if (b == mdev->newest_tle) {
375 /* recycle, but reinit! */
376 D_ASSERT(tmp == NULL);
377 INIT_LIST_HEAD(&b->requests);
378 INIT_LIST_HEAD(&b->w.list);
380 b->br_number = new_initial_bnr;
383 mdev->oldest_tle = b;
390 /* we expect this list to be empty. */
391 D_ASSERT(list_empty(&mdev->out_of_sequence_requests));
393 /* but just in case, clean it up anyways! */
394 list_for_each_safe(le, tle, &mdev->out_of_sequence_requests) {
395 r = list_entry(le, struct drbd_request, tl_requests);
396 /* It would be nice to complete outside of spinlock.
397 * But this is easier for now. */
398 _req_mod(r, connection_lost_while_pending);
401 /* ensure bit indicating barrier is required is clear */
402 clear_bit(CREATE_BARRIER, &mdev->flags);
404 spin_unlock_irq(&mdev->req_lock);
408 * cl_wide_st_chg() - TRUE if the state change is a cluster wide one
409 * @mdev: DRBD device.
410 * @os: old (current) state.
411 * @ns: new (wanted) state.
413 static int cl_wide_st_chg(struct drbd_conf *mdev,
414 union drbd_state os, union drbd_state ns)
416 return (os.conn >= C_CONNECTED && ns.conn >= C_CONNECTED &&
417 ((os.role != R_PRIMARY && ns.role == R_PRIMARY) ||
418 (os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) ||
419 (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S) ||
420 (os.disk != D_DISKLESS && ns.disk == D_DISKLESS))) ||
421 (os.conn >= C_CONNECTED && ns.conn == C_DISCONNECTING) ||
422 (os.conn == C_CONNECTED && ns.conn == C_VERIFY_S);
425 int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
426 union drbd_state mask, union drbd_state val)
429 union drbd_state os, ns;
432 spin_lock_irqsave(&mdev->req_lock, flags);
434 ns.i = (os.i & ~mask.i) | val.i;
435 rv = _drbd_set_state(mdev, ns, f, NULL);
437 spin_unlock_irqrestore(&mdev->req_lock, flags);
443 * drbd_force_state() - Impose a change which happens outside our control on our state
444 * @mdev: DRBD device.
445 * @mask: mask of state bits to change.
446 * @val: value of new state bits.
448 void drbd_force_state(struct drbd_conf *mdev,
449 union drbd_state mask, union drbd_state val)
451 drbd_change_state(mdev, CS_HARD, mask, val);
454 static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns);
455 static int is_valid_state_transition(struct drbd_conf *,
456 union drbd_state, union drbd_state);
457 static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os,
458 union drbd_state ns, int *warn_sync_abort);
459 int drbd_send_state_req(struct drbd_conf *,
460 union drbd_state, union drbd_state);
462 static enum drbd_state_ret_codes _req_st_cond(struct drbd_conf *mdev,
463 union drbd_state mask, union drbd_state val)
465 union drbd_state os, ns;
469 if (test_and_clear_bit(CL_ST_CHG_SUCCESS, &mdev->flags))
470 return SS_CW_SUCCESS;
472 if (test_and_clear_bit(CL_ST_CHG_FAIL, &mdev->flags))
473 return SS_CW_FAILED_BY_PEER;
476 spin_lock_irqsave(&mdev->req_lock, flags);
478 ns.i = (os.i & ~mask.i) | val.i;
479 ns = sanitize_state(mdev, os, ns, NULL);
481 if (!cl_wide_st_chg(mdev, os, ns))
484 rv = is_valid_state(mdev, ns);
485 if (rv == SS_SUCCESS) {
486 rv = is_valid_state_transition(mdev, ns, os);
487 if (rv == SS_SUCCESS)
488 rv = 0; /* cont waiting, otherwise fail. */
491 spin_unlock_irqrestore(&mdev->req_lock, flags);
497 * drbd_req_state() - Perform an eventually cluster wide state change
498 * @mdev: DRBD device.
499 * @mask: mask of state bits to change.
500 * @val: value of new state bits.
503 * Should not be called directly, use drbd_request_state() or
504 * _drbd_request_state().
506 static int drbd_req_state(struct drbd_conf *mdev,
507 union drbd_state mask, union drbd_state val,
508 enum chg_state_flags f)
510 struct completion done;
512 union drbd_state os, ns;
515 init_completion(&done);
517 if (f & CS_SERIALIZE)
518 mutex_lock(&mdev->state_mutex);
520 spin_lock_irqsave(&mdev->req_lock, flags);
522 ns.i = (os.i & ~mask.i) | val.i;
523 ns = sanitize_state(mdev, os, ns, NULL);
525 if (cl_wide_st_chg(mdev, os, ns)) {
526 rv = is_valid_state(mdev, ns);
527 if (rv == SS_SUCCESS)
528 rv = is_valid_state_transition(mdev, ns, os);
529 spin_unlock_irqrestore(&mdev->req_lock, flags);
531 if (rv < SS_SUCCESS) {
533 print_st_err(mdev, os, ns, rv);
537 drbd_state_lock(mdev);
538 if (!drbd_send_state_req(mdev, mask, val)) {
539 drbd_state_unlock(mdev);
540 rv = SS_CW_FAILED_BY_PEER;
542 print_st_err(mdev, os, ns, rv);
546 wait_event(mdev->state_wait,
547 (rv = _req_st_cond(mdev, mask, val)));
549 if (rv < SS_SUCCESS) {
550 drbd_state_unlock(mdev);
552 print_st_err(mdev, os, ns, rv);
555 spin_lock_irqsave(&mdev->req_lock, flags);
557 ns.i = (os.i & ~mask.i) | val.i;
558 rv = _drbd_set_state(mdev, ns, f, &done);
559 drbd_state_unlock(mdev);
561 rv = _drbd_set_state(mdev, ns, f, &done);
564 spin_unlock_irqrestore(&mdev->req_lock, flags);
566 if (f & CS_WAIT_COMPLETE && rv == SS_SUCCESS) {
567 D_ASSERT(current != mdev->worker.task);
568 wait_for_completion(&done);
572 if (f & CS_SERIALIZE)
573 mutex_unlock(&mdev->state_mutex);
579 * _drbd_request_state() - Request a state change (with flags)
580 * @mdev: DRBD device.
581 * @mask: mask of state bits to change.
582 * @val: value of new state bits.
585 * Cousin of drbd_request_state(), useful with the CS_WAIT_COMPLETE
586 * flag, or when logging of failed state change requests is not desired.
588 int _drbd_request_state(struct drbd_conf *mdev, union drbd_state mask,
589 union drbd_state val, enum chg_state_flags f)
593 wait_event(mdev->state_wait,
594 (rv = drbd_req_state(mdev, mask, val, f)) != SS_IN_TRANSIENT_STATE);
599 static void print_st(struct drbd_conf *mdev, char *name, union drbd_state ns)
601 dev_err(DEV, " %s = { cs:%s ro:%s/%s ds:%s/%s %c%c%c%c }\n",
603 drbd_conn_str(ns.conn),
604 drbd_role_str(ns.role),
605 drbd_role_str(ns.peer),
606 drbd_disk_str(ns.disk),
607 drbd_disk_str(ns.pdsk),
609 ns.aftr_isp ? 'a' : '-',
610 ns.peer_isp ? 'p' : '-',
611 ns.user_isp ? 'u' : '-'
615 void print_st_err(struct drbd_conf *mdev,
616 union drbd_state os, union drbd_state ns, int err)
618 if (err == SS_IN_TRANSIENT_STATE)
620 dev_err(DEV, "State change failed: %s\n", drbd_set_st_err_str(err));
621 print_st(mdev, " state", os);
622 print_st(mdev, "wanted", ns);
626 #define drbd_peer_str drbd_role_str
627 #define drbd_pdsk_str drbd_disk_str
629 #define drbd_susp_str(A) ((A) ? "1" : "0")
630 #define drbd_aftr_isp_str(A) ((A) ? "1" : "0")
631 #define drbd_peer_isp_str(A) ((A) ? "1" : "0")
632 #define drbd_user_isp_str(A) ((A) ? "1" : "0")
635 ({ if (ns.A != os.A) { \
636 pbp += sprintf(pbp, #A "( %s -> %s ) ", \
637 drbd_##A##_str(os.A), \
638 drbd_##A##_str(ns.A)); \
642 * is_valid_state() - Returns an SS_ error code if ns is not valid
643 * @mdev: DRBD device.
644 * @ns: State to consider.
646 static int is_valid_state(struct drbd_conf *mdev, union drbd_state ns)
648 /* See drbd_state_sw_errors in drbd_strings.c */
650 enum drbd_fencing_p fp;
654 if (get_ldev(mdev)) {
655 fp = mdev->ldev->dc.fencing;
659 if (get_net_conf(mdev)) {
660 if (!mdev->net_conf->two_primaries &&
661 ns.role == R_PRIMARY && ns.peer == R_PRIMARY)
662 rv = SS_TWO_PRIMARIES;
667 /* already found a reason to abort */;
668 else if (ns.role == R_SECONDARY && mdev->open_cnt)
669 rv = SS_DEVICE_IN_USE;
671 else if (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.disk < D_UP_TO_DATE)
672 rv = SS_NO_UP_TO_DATE_DISK;
674 else if (fp >= FP_RESOURCE &&
675 ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk >= D_UNKNOWN)
678 else if (ns.role == R_PRIMARY && ns.disk <= D_INCONSISTENT && ns.pdsk <= D_INCONSISTENT)
679 rv = SS_NO_UP_TO_DATE_DISK;
681 else if (ns.conn > C_CONNECTED && ns.disk < D_INCONSISTENT)
682 rv = SS_NO_LOCAL_DISK;
684 else if (ns.conn > C_CONNECTED && ns.pdsk < D_INCONSISTENT)
685 rv = SS_NO_REMOTE_DISK;
687 else if ((ns.conn == C_CONNECTED ||
688 ns.conn == C_WF_BITMAP_S ||
689 ns.conn == C_SYNC_SOURCE ||
690 ns.conn == C_PAUSED_SYNC_S) &&
691 ns.disk == D_OUTDATED)
692 rv = SS_CONNECTED_OUTDATES;
694 else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) &&
695 (mdev->sync_conf.verify_alg[0] == 0))
696 rv = SS_NO_VERIFY_ALG;
698 else if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) &&
699 mdev->agreed_pro_version < 88)
700 rv = SS_NOT_SUPPORTED;
706 * is_valid_state_transition() - Returns an SS_ error code if the state transition is not possible
707 * @mdev: DRBD device.
711 static int is_valid_state_transition(struct drbd_conf *mdev,
712 union drbd_state ns, union drbd_state os)
716 if ((ns.conn == C_STARTING_SYNC_T || ns.conn == C_STARTING_SYNC_S) &&
717 os.conn > C_CONNECTED)
718 rv = SS_RESYNC_RUNNING;
720 if (ns.conn == C_DISCONNECTING && os.conn == C_STANDALONE)
721 rv = SS_ALREADY_STANDALONE;
723 if (ns.disk > D_ATTACHING && os.disk == D_DISKLESS)
726 if (ns.conn == C_WF_CONNECTION && os.conn < C_UNCONNECTED)
727 rv = SS_NO_NET_CONFIG;
729 if (ns.disk == D_OUTDATED && os.disk < D_OUTDATED && os.disk != D_ATTACHING)
730 rv = SS_LOWER_THAN_OUTDATED;
732 if (ns.conn == C_DISCONNECTING && os.conn == C_UNCONNECTED)
733 rv = SS_IN_TRANSIENT_STATE;
735 if (ns.conn == os.conn && ns.conn == C_WF_REPORT_PARAMS)
736 rv = SS_IN_TRANSIENT_STATE;
738 if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) && os.conn < C_CONNECTED)
739 rv = SS_NEED_CONNECTION;
741 if ((ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T) &&
742 ns.conn != os.conn && os.conn > C_CONNECTED)
743 rv = SS_RESYNC_RUNNING;
745 if ((ns.conn == C_STARTING_SYNC_S || ns.conn == C_STARTING_SYNC_T) &&
746 os.conn < C_CONNECTED)
747 rv = SS_NEED_CONNECTION;
753 * sanitize_state() - Resolves implicitly necessary additional changes to a state transition
754 * @mdev: DRBD device.
759 * When we loose connection, we have to set the state of the peers disk (pdsk)
760 * to D_UNKNOWN. This rule and many more along those lines are in this function.
762 static union drbd_state sanitize_state(struct drbd_conf *mdev, union drbd_state os,
763 union drbd_state ns, int *warn_sync_abort)
765 enum drbd_fencing_p fp;
768 if (get_ldev(mdev)) {
769 fp = mdev->ldev->dc.fencing;
773 /* Disallow Network errors to configure a device's network part */
774 if ((ns.conn >= C_TIMEOUT && ns.conn <= C_TEAR_DOWN) &&
775 os.conn <= C_DISCONNECTING)
778 /* After a network error (+C_TEAR_DOWN) only C_UNCONNECTED or C_DISCONNECTING can follow */
779 if (os.conn >= C_TIMEOUT && os.conn <= C_TEAR_DOWN &&
780 ns.conn != C_UNCONNECTED && ns.conn != C_DISCONNECTING)
783 /* After C_DISCONNECTING only C_STANDALONE may follow */
784 if (os.conn == C_DISCONNECTING && ns.conn != C_STANDALONE)
787 if (ns.conn < C_CONNECTED) {
790 if (ns.pdsk > D_UNKNOWN || ns.pdsk < D_INCONSISTENT)
794 /* Clear the aftr_isp when becoming unconfigured */
795 if (ns.conn == C_STANDALONE && ns.disk == D_DISKLESS && ns.role == R_SECONDARY)
798 if (ns.conn <= C_DISCONNECTING && ns.disk == D_DISKLESS)
801 /* Abort resync if a disk fails/detaches */
802 if (os.conn > C_CONNECTED && ns.conn > C_CONNECTED &&
803 (ns.disk <= D_FAILED || ns.pdsk <= D_FAILED)) {
805 *warn_sync_abort = 1;
806 ns.conn = C_CONNECTED;
809 if (ns.conn >= C_CONNECTED &&
810 ((ns.disk == D_CONSISTENT || ns.disk == D_OUTDATED) ||
811 (ns.disk == D_NEGOTIATING && ns.conn == C_WF_BITMAP_T))) {
814 case C_PAUSED_SYNC_T:
815 ns.disk = D_OUTDATED;
820 case C_PAUSED_SYNC_S:
821 ns.disk = D_UP_TO_DATE;
824 ns.disk = D_INCONSISTENT;
825 dev_warn(DEV, "Implicitly set disk state Inconsistent!\n");
828 if (os.disk == D_OUTDATED && ns.disk == D_UP_TO_DATE)
829 dev_warn(DEV, "Implicitly set disk from Outdated to UpToDate\n");
832 if (ns.conn >= C_CONNECTED &&
833 (ns.pdsk == D_CONSISTENT || ns.pdsk == D_OUTDATED)) {
837 case C_PAUSED_SYNC_T:
839 ns.pdsk = D_UP_TO_DATE;
842 case C_PAUSED_SYNC_S:
843 ns.pdsk = D_OUTDATED;
846 ns.pdsk = D_INCONSISTENT;
847 dev_warn(DEV, "Implicitly set pdsk Inconsistent!\n");
850 if (os.pdsk == D_OUTDATED && ns.pdsk == D_UP_TO_DATE)
851 dev_warn(DEV, "Implicitly set pdsk from Outdated to UpToDate\n");
854 /* Connection breaks down before we finished "Negotiating" */
855 if (ns.conn < C_CONNECTED && ns.disk == D_NEGOTIATING &&
856 get_ldev_if_state(mdev, D_NEGOTIATING)) {
857 if (mdev->ed_uuid == mdev->ldev->md.uuid[UI_CURRENT]) {
858 ns.disk = mdev->new_state_tmp.disk;
859 ns.pdsk = mdev->new_state_tmp.pdsk;
861 dev_alert(DEV, "Connection lost while negotiating, no data!\n");
862 ns.disk = D_DISKLESS;
868 if (fp == FP_STONITH &&
869 (ns.role == R_PRIMARY && ns.conn < C_CONNECTED && ns.pdsk > D_OUTDATED) &&
870 !(os.role == R_PRIMARY && os.conn < C_CONNECTED && os.pdsk > D_OUTDATED))
873 if (ns.aftr_isp || ns.peer_isp || ns.user_isp) {
874 if (ns.conn == C_SYNC_SOURCE)
875 ns.conn = C_PAUSED_SYNC_S;
876 if (ns.conn == C_SYNC_TARGET)
877 ns.conn = C_PAUSED_SYNC_T;
879 if (ns.conn == C_PAUSED_SYNC_S)
880 ns.conn = C_SYNC_SOURCE;
881 if (ns.conn == C_PAUSED_SYNC_T)
882 ns.conn = C_SYNC_TARGET;
888 /* helper for __drbd_set_state */
889 static void set_ov_position(struct drbd_conf *mdev, enum drbd_conns cs)
891 if (cs == C_VERIFY_T) {
892 /* starting online verify from an arbitrary position
893 * does not fit well into the existing protocol.
894 * on C_VERIFY_T, we initialize ov_left and friends
895 * implicitly in receive_DataRequest once the
896 * first P_OV_REQUEST is received */
897 mdev->ov_start_sector = ~(sector_t)0;
899 unsigned long bit = BM_SECT_TO_BIT(mdev->ov_start_sector);
900 if (bit >= mdev->rs_total)
901 mdev->ov_start_sector =
902 BM_BIT_TO_SECT(mdev->rs_total - 1);
903 mdev->ov_position = mdev->ov_start_sector;
908 * __drbd_set_state() - Set a new DRBD state
909 * @mdev: DRBD device.
912 * @done: Optional completion, that will get completed after the after_state_ch() finished
914 * Caller needs to hold req_lock, and global_state_lock. Do not call directly.
916 int __drbd_set_state(struct drbd_conf *mdev,
917 union drbd_state ns, enum chg_state_flags flags,
918 struct completion *done)
922 int warn_sync_abort = 0;
923 struct after_state_chg_work *ascw;
927 ns = sanitize_state(mdev, os, ns, &warn_sync_abort);
930 return SS_NOTHING_TO_DO;
932 if (!(flags & CS_HARD)) {
933 /* pre-state-change checks ; only look at ns */
934 /* See drbd_state_sw_errors in drbd_strings.c */
936 rv = is_valid_state(mdev, ns);
937 if (rv < SS_SUCCESS) {
938 /* If the old state was illegal as well, then let
941 if (is_valid_state(mdev, os) == rv) {
942 dev_err(DEV, "Considering state change from bad state. "
943 "Error would be: '%s'\n",
944 drbd_set_st_err_str(rv));
945 print_st(mdev, "old", os);
946 print_st(mdev, "new", ns);
947 rv = is_valid_state_transition(mdev, ns, os);
950 rv = is_valid_state_transition(mdev, ns, os);
953 if (rv < SS_SUCCESS) {
954 if (flags & CS_VERBOSE)
955 print_st_err(mdev, os, ns, rv);
960 dev_warn(DEV, "Resync aborted.\n");
975 dev_info(DEV, "%s\n", pb);
978 /* solve the race between becoming unconfigured,
979 * worker doing the cleanup, and
980 * admin reconfiguring us:
981 * on (re)configure, first set CONFIG_PENDING,
982 * then wait for a potentially exiting worker,
983 * start the worker, and schedule one no_op.
984 * then proceed with configuration.
986 if (ns.disk == D_DISKLESS &&
987 ns.conn == C_STANDALONE &&
988 ns.role == R_SECONDARY &&
989 !test_and_set_bit(CONFIG_PENDING, &mdev->flags))
990 set_bit(DEVICE_DYING, &mdev->flags);
992 mdev->state.i = ns.i;
993 wake_up(&mdev->misc_wait);
994 wake_up(&mdev->state_wait);
996 /* post-state-change actions */
997 if (os.conn >= C_SYNC_SOURCE && ns.conn <= C_CONNECTED) {
998 set_bit(STOP_SYNC_TIMER, &mdev->flags);
999 mod_timer(&mdev->resync_timer, jiffies);
1002 /* aborted verify run. log the last position */
1003 if ((os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) &&
1004 ns.conn < C_CONNECTED) {
1005 mdev->ov_start_sector =
1006 BM_BIT_TO_SECT(mdev->rs_total - mdev->ov_left);
1007 dev_info(DEV, "Online Verify reached sector %llu\n",
1008 (unsigned long long)mdev->ov_start_sector);
1011 if ((os.conn == C_PAUSED_SYNC_T || os.conn == C_PAUSED_SYNC_S) &&
1012 (ns.conn == C_SYNC_TARGET || ns.conn == C_SYNC_SOURCE)) {
1013 dev_info(DEV, "Syncer continues.\n");
1014 mdev->rs_paused += (long)jiffies-(long)mdev->rs_mark_time;
1015 if (ns.conn == C_SYNC_TARGET) {
1016 if (!test_and_clear_bit(STOP_SYNC_TIMER, &mdev->flags))
1017 mod_timer(&mdev->resync_timer, jiffies);
1018 /* This if (!test_bit) is only needed for the case
1019 that a device that has ceased to used its timer,
1020 i.e. it is already in drbd_resync_finished() gets
1021 paused and resumed. */
1025 if ((os.conn == C_SYNC_TARGET || os.conn == C_SYNC_SOURCE) &&
1026 (ns.conn == C_PAUSED_SYNC_T || ns.conn == C_PAUSED_SYNC_S)) {
1027 dev_info(DEV, "Resync suspended\n");
1028 mdev->rs_mark_time = jiffies;
1029 if (ns.conn == C_PAUSED_SYNC_T)
1030 set_bit(STOP_SYNC_TIMER, &mdev->flags);
1033 if (os.conn == C_CONNECTED &&
1034 (ns.conn == C_VERIFY_S || ns.conn == C_VERIFY_T)) {
1035 mdev->ov_position = 0;
1037 mdev->rs_mark_left = drbd_bm_bits(mdev);
1038 if (mdev->agreed_pro_version >= 90)
1039 set_ov_position(mdev, ns.conn);
1041 mdev->ov_start_sector = 0;
1042 mdev->ov_left = mdev->rs_total
1043 - BM_SECT_TO_BIT(mdev->ov_position);
1045 mdev->rs_mark_time = jiffies;
1046 mdev->ov_last_oos_size = 0;
1047 mdev->ov_last_oos_start = 0;
1049 if (ns.conn == C_VERIFY_S) {
1050 dev_info(DEV, "Starting Online Verify from sector %llu\n",
1051 (unsigned long long)mdev->ov_position);
1052 mod_timer(&mdev->resync_timer, jiffies);
1056 if (get_ldev(mdev)) {
1057 u32 mdf = mdev->ldev->md.flags & ~(MDF_CONSISTENT|MDF_PRIMARY_IND|
1058 MDF_CONNECTED_IND|MDF_WAS_UP_TO_DATE|
1059 MDF_PEER_OUT_DATED|MDF_CRASHED_PRIMARY);
1061 if (test_bit(CRASHED_PRIMARY, &mdev->flags))
1062 mdf |= MDF_CRASHED_PRIMARY;
1063 if (mdev->state.role == R_PRIMARY ||
1064 (mdev->state.pdsk < D_INCONSISTENT && mdev->state.peer == R_PRIMARY))
1065 mdf |= MDF_PRIMARY_IND;
1066 if (mdev->state.conn > C_WF_REPORT_PARAMS)
1067 mdf |= MDF_CONNECTED_IND;
1068 if (mdev->state.disk > D_INCONSISTENT)
1069 mdf |= MDF_CONSISTENT;
1070 if (mdev->state.disk > D_OUTDATED)
1071 mdf |= MDF_WAS_UP_TO_DATE;
1072 if (mdev->state.pdsk <= D_OUTDATED && mdev->state.pdsk >= D_INCONSISTENT)
1073 mdf |= MDF_PEER_OUT_DATED;
1074 if (mdf != mdev->ldev->md.flags) {
1075 mdev->ldev->md.flags = mdf;
1076 drbd_md_mark_dirty(mdev);
1078 if (os.disk < D_CONSISTENT && ns.disk >= D_CONSISTENT)
1079 drbd_set_ed_uuid(mdev, mdev->ldev->md.uuid[UI_CURRENT]);
1083 /* Peer was forced D_UP_TO_DATE & R_PRIMARY, consider to resync */
1084 if (os.disk == D_INCONSISTENT && os.pdsk == D_INCONSISTENT &&
1085 os.peer == R_SECONDARY && ns.peer == R_PRIMARY)
1086 set_bit(CONSIDER_RESYNC, &mdev->flags);
1088 /* Receiver should clean up itself */
1089 if (os.conn != C_DISCONNECTING && ns.conn == C_DISCONNECTING)
1090 drbd_thread_stop_nowait(&mdev->receiver);
1092 /* Now the receiver finished cleaning up itself, it should die */
1093 if (os.conn != C_STANDALONE && ns.conn == C_STANDALONE)
1094 drbd_thread_stop_nowait(&mdev->receiver);
1096 /* Upon network failure, we need to restart the receiver. */
1097 if (os.conn > C_TEAR_DOWN &&
1098 ns.conn <= C_TEAR_DOWN && ns.conn >= C_TIMEOUT)
1099 drbd_thread_restart_nowait(&mdev->receiver);
1101 ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC);
1105 ascw->flags = flags;
1106 ascw->w.cb = w_after_state_ch;
1108 drbd_queue_work(&mdev->data.work, &ascw->w);
1110 dev_warn(DEV, "Could not kmalloc an ascw\n");
1116 static int w_after_state_ch(struct drbd_conf *mdev, struct drbd_work *w, int unused)
1118 struct after_state_chg_work *ascw =
1119 container_of(w, struct after_state_chg_work, w);
1120 after_state_ch(mdev, ascw->os, ascw->ns, ascw->flags);
1121 if (ascw->flags & CS_WAIT_COMPLETE) {
1122 D_ASSERT(ascw->done != NULL);
1123 complete(ascw->done);
1130 static void abw_start_sync(struct drbd_conf *mdev, int rv)
1133 dev_err(DEV, "Writing the bitmap failed not starting resync.\n");
1134 _drbd_request_state(mdev, NS(conn, C_CONNECTED), CS_VERBOSE);
1138 switch (mdev->state.conn) {
1139 case C_STARTING_SYNC_T:
1140 _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE);
1142 case C_STARTING_SYNC_S:
1143 drbd_start_resync(mdev, C_SYNC_SOURCE);
1149 * after_state_ch() - Perform after state change actions that may sleep
1150 * @mdev: DRBD device.
1155 static void after_state_ch(struct drbd_conf *mdev, union drbd_state os,
1156 union drbd_state ns, enum chg_state_flags flags)
1158 enum drbd_fencing_p fp;
1160 if (os.conn != C_CONNECTED && ns.conn == C_CONNECTED) {
1161 clear_bit(CRASHED_PRIMARY, &mdev->flags);
1163 mdev->p_uuid[UI_FLAGS] &= ~((u64)2);
1167 if (get_ldev(mdev)) {
1168 fp = mdev->ldev->dc.fencing;
1172 /* Inform userspace about the change... */
1173 drbd_bcast_state(mdev, ns);
1175 if (!(os.role == R_PRIMARY && os.disk < D_UP_TO_DATE && os.pdsk < D_UP_TO_DATE) &&
1176 (ns.role == R_PRIMARY && ns.disk < D_UP_TO_DATE && ns.pdsk < D_UP_TO_DATE))
1177 drbd_khelper(mdev, "pri-on-incon-degr");
1179 /* Here we have the actions that are performed after a
1180 state change. This function might sleep */
1182 if (fp == FP_STONITH && ns.susp) {
1183 /* case1: The outdate peer handler is successful:
1184 * case2: The connection was established again: */
1185 if ((os.pdsk > D_OUTDATED && ns.pdsk <= D_OUTDATED) ||
1186 (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED)) {
1188 spin_lock_irq(&mdev->req_lock);
1189 _drbd_set_state(_NS(mdev, susp, 0), CS_VERBOSE, NULL);
1190 spin_unlock_irq(&mdev->req_lock);
1193 /* Do not change the order of the if above and the two below... */
1194 if (os.pdsk == D_DISKLESS && ns.pdsk > D_DISKLESS) { /* attach on the peer */
1195 drbd_send_uuids(mdev);
1196 drbd_send_state(mdev);
1198 if (os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S)
1199 drbd_queue_bitmap_io(mdev, &drbd_send_bitmap, NULL, "send_bitmap (WFBitMapS)");
1201 /* Lost contact to peer's copy of the data */
1202 if ((os.pdsk >= D_INCONSISTENT &&
1203 os.pdsk != D_UNKNOWN &&
1204 os.pdsk != D_OUTDATED)
1205 && (ns.pdsk < D_INCONSISTENT ||
1206 ns.pdsk == D_UNKNOWN ||
1207 ns.pdsk == D_OUTDATED)) {
1208 kfree(mdev->p_uuid);
1209 mdev->p_uuid = NULL;
1210 if (get_ldev(mdev)) {
1211 if ((ns.role == R_PRIMARY || ns.peer == R_PRIMARY) &&
1212 mdev->ldev->md.uuid[UI_BITMAP] == 0 && ns.disk >= D_UP_TO_DATE) {
1213 drbd_uuid_new_current(mdev);
1214 drbd_send_uuids(mdev);
1220 if (ns.pdsk < D_INCONSISTENT && get_ldev(mdev)) {
1221 if (ns.peer == R_PRIMARY && mdev->ldev->md.uuid[UI_BITMAP] == 0)
1222 drbd_uuid_new_current(mdev);
1224 /* D_DISKLESS Peer becomes secondary */
1225 if (os.peer == R_PRIMARY && ns.peer == R_SECONDARY)
1226 drbd_al_to_on_disk_bm(mdev);
1230 /* Last part of the attaching process ... */
1231 if (ns.conn >= C_CONNECTED &&
1232 os.disk == D_ATTACHING && ns.disk == D_NEGOTIATING) {
1233 kfree(mdev->p_uuid); /* We expect to receive up-to-date UUIDs soon. */
1234 mdev->p_uuid = NULL; /* ...to not use the old ones in the mean time */
1235 drbd_send_sizes(mdev, 0); /* to start sync... */
1236 drbd_send_uuids(mdev);
1237 drbd_send_state(mdev);
1240 /* We want to pause/continue resync, tell peer. */
1241 if (ns.conn >= C_CONNECTED &&
1242 ((os.aftr_isp != ns.aftr_isp) ||
1243 (os.user_isp != ns.user_isp)))
1244 drbd_send_state(mdev);
1246 /* In case one of the isp bits got set, suspend other devices. */
1247 if ((!os.aftr_isp && !os.peer_isp && !os.user_isp) &&
1248 (ns.aftr_isp || ns.peer_isp || ns.user_isp))
1249 suspend_other_sg(mdev);
1251 /* Make sure the peer gets informed about eventual state
1252 changes (ISP bits) while we were in WFReportParams. */
1253 if (os.conn == C_WF_REPORT_PARAMS && ns.conn >= C_CONNECTED)
1254 drbd_send_state(mdev);
1256 /* We are in the progress to start a full sync... */
1257 if ((os.conn != C_STARTING_SYNC_T && ns.conn == C_STARTING_SYNC_T) ||
1258 (os.conn != C_STARTING_SYNC_S && ns.conn == C_STARTING_SYNC_S))
1259 drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, &abw_start_sync, "set_n_write from StartingSync");
1261 /* We are invalidating our self... */
1262 if (os.conn < C_CONNECTED && ns.conn < C_CONNECTED &&
1263 os.disk > D_INCONSISTENT && ns.disk == D_INCONSISTENT)
1264 drbd_queue_bitmap_io(mdev, &drbd_bmio_set_n_write, NULL, "set_n_write from invalidate");
1266 if (os.disk > D_FAILED && ns.disk == D_FAILED) {
1267 enum drbd_io_error_p eh;
1270 if (get_ldev_if_state(mdev, D_FAILED)) {
1271 eh = mdev->ldev->dc.on_io_error;
1275 drbd_rs_cancel_all(mdev);
1276 /* since get_ldev() only works as long as disk>=D_INCONSISTENT,
1277 and it is D_DISKLESS here, local_cnt can only go down, it can
1278 not increase... It will reach zero */
1279 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
1281 mdev->rs_failed = 0;
1282 atomic_set(&mdev->rs_pending_cnt, 0);
1284 spin_lock_irq(&mdev->req_lock);
1285 _drbd_set_state(_NS(mdev, disk, D_DISKLESS), CS_HARD, NULL);
1286 spin_unlock_irq(&mdev->req_lock);
1288 if (eh == EP_CALL_HELPER)
1289 drbd_khelper(mdev, "local-io-error");
1292 if (os.disk > D_DISKLESS && ns.disk == D_DISKLESS) {
1294 if (os.disk == D_FAILED) /* && ns.disk == D_DISKLESS*/ {
1295 if (drbd_send_state(mdev))
1296 dev_warn(DEV, "Notified peer that my disk is broken.\n");
1298 dev_err(DEV, "Sending state in drbd_io_error() failed\n");
1301 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
1302 lc_destroy(mdev->resync);
1303 mdev->resync = NULL;
1304 lc_destroy(mdev->act_log);
1305 mdev->act_log = NULL;
1307 drbd_free_bc(mdev->ldev);
1308 mdev->ldev = NULL;);
1310 if (mdev->md_io_tmpp)
1311 __free_page(mdev->md_io_tmpp);
1314 /* Disks got bigger while they were detached */
1315 if (ns.disk > D_NEGOTIATING && ns.pdsk > D_NEGOTIATING &&
1316 test_and_clear_bit(RESYNC_AFTER_NEG, &mdev->flags)) {
1317 if (ns.conn == C_CONNECTED)
1318 resync_after_online_grow(mdev);
1321 /* A resync finished or aborted, wake paused devices... */
1322 if ((os.conn > C_CONNECTED && ns.conn <= C_CONNECTED) ||
1323 (os.peer_isp && !ns.peer_isp) ||
1324 (os.user_isp && !ns.user_isp))
1325 resume_next_sg(mdev);
1327 /* Upon network connection, we need to start the receiver */
1328 if (os.conn == C_STANDALONE && ns.conn == C_UNCONNECTED)
1329 drbd_thread_start(&mdev->receiver);
1331 /* Terminate worker thread if we are unconfigured - it will be
1332 restarted as needed... */
1333 if (ns.disk == D_DISKLESS &&
1334 ns.conn == C_STANDALONE &&
1335 ns.role == R_SECONDARY) {
1336 if (os.aftr_isp != ns.aftr_isp)
1337 resume_next_sg(mdev);
1338 /* set in __drbd_set_state, unless CONFIG_PENDING was set */
1339 if (test_bit(DEVICE_DYING, &mdev->flags))
1340 drbd_thread_stop_nowait(&mdev->worker);
1347 static int drbd_thread_setup(void *arg)
1349 struct drbd_thread *thi = (struct drbd_thread *) arg;
1350 struct drbd_conf *mdev = thi->mdev;
1351 unsigned long flags;
1355 retval = thi->function(thi);
1357 spin_lock_irqsave(&thi->t_lock, flags);
1359 /* if the receiver has been "Exiting", the last thing it did
1360 * was set the conn state to "StandAlone",
1361 * if now a re-connect request comes in, conn state goes C_UNCONNECTED,
1362 * and receiver thread will be "started".
1363 * drbd_thread_start needs to set "Restarting" in that case.
1364 * t_state check and assignment needs to be within the same spinlock,
1365 * so either thread_start sees Exiting, and can remap to Restarting,
1366 * or thread_start see None, and can proceed as normal.
1369 if (thi->t_state == Restarting) {
1370 dev_info(DEV, "Restarting %s\n", current->comm);
1371 thi->t_state = Running;
1372 spin_unlock_irqrestore(&thi->t_lock, flags);
1377 thi->t_state = None;
1379 complete(&thi->stop);
1380 spin_unlock_irqrestore(&thi->t_lock, flags);
1382 dev_info(DEV, "Terminating %s\n", current->comm);
1384 /* Release mod reference taken when thread was started */
1385 module_put(THIS_MODULE);
1389 static void drbd_thread_init(struct drbd_conf *mdev, struct drbd_thread *thi,
1390 int (*func) (struct drbd_thread *))
1392 spin_lock_init(&thi->t_lock);
1394 thi->t_state = None;
1395 thi->function = func;
1399 int drbd_thread_start(struct drbd_thread *thi)
1401 struct drbd_conf *mdev = thi->mdev;
1402 struct task_struct *nt;
1403 unsigned long flags;
1406 thi == &mdev->receiver ? "receiver" :
1407 thi == &mdev->asender ? "asender" :
1408 thi == &mdev->worker ? "worker" : "NONSENSE";
1410 /* is used from state engine doing drbd_thread_stop_nowait,
1411 * while holding the req lock irqsave */
1412 spin_lock_irqsave(&thi->t_lock, flags);
1414 switch (thi->t_state) {
1416 dev_info(DEV, "Starting %s thread (from %s [%d])\n",
1417 me, current->comm, current->pid);
1419 /* Get ref on module for thread - this is released when thread exits */
1420 if (!try_module_get(THIS_MODULE)) {
1421 dev_err(DEV, "Failed to get module reference in drbd_thread_start\n");
1422 spin_unlock_irqrestore(&thi->t_lock, flags);
1426 init_completion(&thi->stop);
1427 D_ASSERT(thi->task == NULL);
1428 thi->reset_cpu_mask = 1;
1429 thi->t_state = Running;
1430 spin_unlock_irqrestore(&thi->t_lock, flags);
1431 flush_signals(current); /* otherw. may get -ERESTARTNOINTR */
1433 nt = kthread_create(drbd_thread_setup, (void *) thi,
1434 "drbd%d_%s", mdev_to_minor(mdev), me);
1437 dev_err(DEV, "Couldn't start thread\n");
1439 module_put(THIS_MODULE);
1442 spin_lock_irqsave(&thi->t_lock, flags);
1444 thi->t_state = Running;
1445 spin_unlock_irqrestore(&thi->t_lock, flags);
1446 wake_up_process(nt);
1449 thi->t_state = Restarting;
1450 dev_info(DEV, "Restarting %s thread (from %s [%d])\n",
1451 me, current->comm, current->pid);
1456 spin_unlock_irqrestore(&thi->t_lock, flags);
1464 void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait)
1466 unsigned long flags;
1468 enum drbd_thread_state ns = restart ? Restarting : Exiting;
1470 /* may be called from state engine, holding the req lock irqsave */
1471 spin_lock_irqsave(&thi->t_lock, flags);
1473 if (thi->t_state == None) {
1474 spin_unlock_irqrestore(&thi->t_lock, flags);
1476 drbd_thread_start(thi);
1480 if (thi->t_state != ns) {
1481 if (thi->task == NULL) {
1482 spin_unlock_irqrestore(&thi->t_lock, flags);
1488 init_completion(&thi->stop);
1489 if (thi->task != current)
1490 force_sig(DRBD_SIGKILL, thi->task);
1494 spin_unlock_irqrestore(&thi->t_lock, flags);
1497 wait_for_completion(&thi->stop);
1502 * drbd_calc_cpu_mask() - Generate CPU masks, spread over all CPUs
1503 * @mdev: DRBD device.
1505 * Forces all threads of a device onto the same CPU. This is beneficial for
1506 * DRBD's performance. May be overwritten by user's configuration.
1508 void drbd_calc_cpu_mask(struct drbd_conf *mdev)
1512 /* user override. */
1513 if (cpumask_weight(mdev->cpu_mask))
1516 ord = mdev_to_minor(mdev) % cpumask_weight(cpu_online_mask);
1517 for_each_online_cpu(cpu) {
1519 cpumask_set_cpu(cpu, mdev->cpu_mask);
1523 /* should not be reached */
1524 cpumask_setall(mdev->cpu_mask);
1528 * drbd_thread_current_set_cpu() - modifies the cpu mask of the _current_ thread
1529 * @mdev: DRBD device.
1531 * call in the "main loop" of _all_ threads, no need for any mutex, current won't die
1534 void drbd_thread_current_set_cpu(struct drbd_conf *mdev)
1536 struct task_struct *p = current;
1537 struct drbd_thread *thi =
1538 p == mdev->asender.task ? &mdev->asender :
1539 p == mdev->receiver.task ? &mdev->receiver :
1540 p == mdev->worker.task ? &mdev->worker :
1544 if (!thi->reset_cpu_mask)
1546 thi->reset_cpu_mask = 0;
1547 set_cpus_allowed_ptr(p, mdev->cpu_mask);
1551 /* the appropriate socket mutex must be held already */
1552 int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1553 enum drbd_packets cmd, struct p_header *h,
1554 size_t size, unsigned msg_flags)
1558 ERR_IF(!h) return FALSE;
1559 ERR_IF(!size) return FALSE;
1561 h->magic = BE_DRBD_MAGIC;
1562 h->command = cpu_to_be16(cmd);
1563 h->length = cpu_to_be16(size-sizeof(struct p_header));
1565 sent = drbd_send(mdev, sock, h, size, msg_flags);
1567 ok = (sent == size);
1569 dev_err(DEV, "short sent %s size=%d sent=%d\n",
1570 cmdname(cmd), (int)size, sent);
1574 /* don't pass the socket. we may only look at it
1575 * when we hold the appropriate socket mutex.
1577 int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1578 enum drbd_packets cmd, struct p_header *h, size_t size)
1581 struct socket *sock;
1583 if (use_data_socket) {
1584 mutex_lock(&mdev->data.mutex);
1585 sock = mdev->data.socket;
1587 mutex_lock(&mdev->meta.mutex);
1588 sock = mdev->meta.socket;
1591 /* drbd_disconnect() could have called drbd_free_sock()
1592 * while we were waiting in down()... */
1593 if (likely(sock != NULL))
1594 ok = _drbd_send_cmd(mdev, sock, cmd, h, size, 0);
1596 if (use_data_socket)
1597 mutex_unlock(&mdev->data.mutex);
1599 mutex_unlock(&mdev->meta.mutex);
1603 int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, char *data,
1609 h.magic = BE_DRBD_MAGIC;
1610 h.command = cpu_to_be16(cmd);
1611 h.length = cpu_to_be16(size);
1613 if (!drbd_get_data_sock(mdev))
1617 drbd_send(mdev, mdev->data.socket, &h, sizeof(h), 0));
1619 drbd_send(mdev, mdev->data.socket, data, size, 0));
1621 drbd_put_data_sock(mdev);
1626 int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc)
1628 struct p_rs_param_89 *p;
1629 struct socket *sock;
1631 const int apv = mdev->agreed_pro_version;
1633 size = apv <= 87 ? sizeof(struct p_rs_param)
1634 : apv == 88 ? sizeof(struct p_rs_param)
1635 + strlen(mdev->sync_conf.verify_alg) + 1
1636 : /* 89 */ sizeof(struct p_rs_param_89);
1638 /* used from admin command context and receiver/worker context.
1639 * to avoid kmalloc, grab the socket right here,
1640 * then use the pre-allocated sbuf there */
1641 mutex_lock(&mdev->data.mutex);
1642 sock = mdev->data.socket;
1644 if (likely(sock != NULL)) {
1645 enum drbd_packets cmd = apv >= 89 ? P_SYNC_PARAM89 : P_SYNC_PARAM;
1647 p = &mdev->data.sbuf.rs_param_89;
1649 /* initialize verify_alg and csums_alg */
1650 memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX);
1652 p->rate = cpu_to_be32(sc->rate);
1655 strcpy(p->verify_alg, mdev->sync_conf.verify_alg);
1657 strcpy(p->csums_alg, mdev->sync_conf.csums_alg);
1659 rv = _drbd_send_cmd(mdev, sock, cmd, &p->head, size, 0);
1661 rv = 0; /* not ok */
1663 mutex_unlock(&mdev->data.mutex);
1668 int drbd_send_protocol(struct drbd_conf *mdev)
1670 struct p_protocol *p;
1673 size = sizeof(struct p_protocol);
1675 if (mdev->agreed_pro_version >= 87)
1676 size += strlen(mdev->net_conf->integrity_alg) + 1;
1678 /* we must not recurse into our own queue,
1679 * as that is blocked during handshake */
1680 p = kmalloc(size, GFP_NOIO);
1684 p->protocol = cpu_to_be32(mdev->net_conf->wire_protocol);
1685 p->after_sb_0p = cpu_to_be32(mdev->net_conf->after_sb_0p);
1686 p->after_sb_1p = cpu_to_be32(mdev->net_conf->after_sb_1p);
1687 p->after_sb_2p = cpu_to_be32(mdev->net_conf->after_sb_2p);
1688 p->two_primaries = cpu_to_be32(mdev->net_conf->two_primaries);
1691 if (mdev->net_conf->want_lose)
1693 if (mdev->net_conf->dry_run) {
1694 if (mdev->agreed_pro_version >= 92)
1697 dev_err(DEV, "--dry-run is not supported by peer");
1701 p->conn_flags = cpu_to_be32(cf);
1703 if (mdev->agreed_pro_version >= 87)
1704 strcpy(p->integrity_alg, mdev->net_conf->integrity_alg);
1706 rv = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_PROTOCOL,
1707 (struct p_header *)p, size);
1712 int _drbd_send_uuids(struct drbd_conf *mdev, u64 uuid_flags)
1717 if (!get_ldev_if_state(mdev, D_NEGOTIATING))
1720 for (i = UI_CURRENT; i < UI_SIZE; i++)
1721 p.uuid[i] = mdev->ldev ? cpu_to_be64(mdev->ldev->md.uuid[i]) : 0;
1723 mdev->comm_bm_set = drbd_bm_total_weight(mdev);
1724 p.uuid[UI_SIZE] = cpu_to_be64(mdev->comm_bm_set);
1725 uuid_flags |= mdev->net_conf->want_lose ? 1 : 0;
1726 uuid_flags |= test_bit(CRASHED_PRIMARY, &mdev->flags) ? 2 : 0;
1727 uuid_flags |= mdev->new_state_tmp.disk == D_INCONSISTENT ? 4 : 0;
1728 p.uuid[UI_FLAGS] = cpu_to_be64(uuid_flags);
1732 return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_UUIDS,
1733 (struct p_header *)&p, sizeof(p));
1736 int drbd_send_uuids(struct drbd_conf *mdev)
1738 return _drbd_send_uuids(mdev, 0);
1741 int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev)
1743 return _drbd_send_uuids(mdev, 8);
1747 int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val)
1751 p.uuid = cpu_to_be64(val);
1753 return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SYNC_UUID,
1754 (struct p_header *)&p, sizeof(p));
1757 int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply)
1760 sector_t d_size, u_size;
1764 if (get_ldev_if_state(mdev, D_NEGOTIATING)) {
1765 D_ASSERT(mdev->ldev->backing_bdev);
1766 d_size = drbd_get_max_capacity(mdev->ldev);
1767 u_size = mdev->ldev->dc.disk_size;
1768 q_order_type = drbd_queue_order_type(mdev);
1769 p.queue_order_type = cpu_to_be32(drbd_queue_order_type(mdev));
1774 q_order_type = QUEUE_ORDERED_NONE;
1777 p.d_size = cpu_to_be64(d_size);
1778 p.u_size = cpu_to_be64(u_size);
1779 p.c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(mdev->this_bdev));
1780 p.max_segment_size = cpu_to_be32(queue_max_segment_size(mdev->rq_queue));
1781 p.queue_order_type = cpu_to_be32(q_order_type);
1783 ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_SIZES,
1784 (struct p_header *)&p, sizeof(p));
1789 * drbd_send_state() - Sends the drbd state to the peer
1790 * @mdev: DRBD device.
1792 int drbd_send_state(struct drbd_conf *mdev)
1794 struct socket *sock;
1798 /* Grab state lock so we wont send state if we're in the middle
1799 * of a cluster wide state change on another thread */
1800 drbd_state_lock(mdev);
1802 mutex_lock(&mdev->data.mutex);
1804 p.state = cpu_to_be32(mdev->state.i); /* Within the send mutex */
1805 sock = mdev->data.socket;
1807 if (likely(sock != NULL)) {
1808 ok = _drbd_send_cmd(mdev, sock, P_STATE,
1809 (struct p_header *)&p, sizeof(p), 0);
1812 mutex_unlock(&mdev->data.mutex);
1814 drbd_state_unlock(mdev);
1818 int drbd_send_state_req(struct drbd_conf *mdev,
1819 union drbd_state mask, union drbd_state val)
1821 struct p_req_state p;
1823 p.mask = cpu_to_be32(mask.i);
1824 p.val = cpu_to_be32(val.i);
1826 return drbd_send_cmd(mdev, USE_DATA_SOCKET, P_STATE_CHG_REQ,
1827 (struct p_header *)&p, sizeof(p));
1830 int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode)
1832 struct p_req_state_reply p;
1834 p.retcode = cpu_to_be32(retcode);
1836 return drbd_send_cmd(mdev, USE_META_SOCKET, P_STATE_CHG_REPLY,
1837 (struct p_header *)&p, sizeof(p));
1840 int fill_bitmap_rle_bits(struct drbd_conf *mdev,
1841 struct p_compressed_bm *p,
1842 struct bm_xfer_ctx *c)
1844 struct bitstream bs;
1845 unsigned long plain_bits;
1852 /* may we use this feature? */
1853 if ((mdev->sync_conf.use_rle == 0) ||
1854 (mdev->agreed_pro_version < 90))
1857 if (c->bit_offset >= c->bm_bits)
1858 return 0; /* nothing to do. */
1860 /* use at most thus many bytes */
1861 bitstream_init(&bs, p->code, BM_PACKET_VLI_BYTES_MAX, 0);
1862 memset(p->code, 0, BM_PACKET_VLI_BYTES_MAX);
1863 /* plain bits covered in this code string */
1866 /* p->encoding & 0x80 stores whether the first run length is set.
1867 * bit offset is implicit.
1868 * start with toggle == 2 to be able to tell the first iteration */
1871 /* see how much plain bits we can stuff into one packet
1872 * using RLE and VLI. */
1874 tmp = (toggle == 0) ? _drbd_bm_find_next_zero(mdev, c->bit_offset)
1875 : _drbd_bm_find_next(mdev, c->bit_offset);
1878 rl = tmp - c->bit_offset;
1880 if (toggle == 2) { /* first iteration */
1882 /* the first checked bit was set,
1883 * store start value, */
1884 DCBP_set_start(p, 1);
1885 /* but skip encoding of zero run length */
1889 DCBP_set_start(p, 0);
1892 /* paranoia: catch zero runlength.
1893 * can only happen if bitmap is modified while we scan it. */
1895 dev_err(DEV, "unexpected zero runlength while encoding bitmap "
1896 "t:%u bo:%lu\n", toggle, c->bit_offset);
1900 bits = vli_encode_bits(&bs, rl);
1901 if (bits == -ENOBUFS) /* buffer full */
1904 dev_err(DEV, "error while encoding bitmap: %d\n", bits);
1910 c->bit_offset = tmp;
1911 } while (c->bit_offset < c->bm_bits);
1913 len = bs.cur.b - p->code + !!bs.cur.bit;
1915 if (plain_bits < (len << 3)) {
1916 /* incompressible with this method.
1917 * we need to rewind both word and bit position. */
1918 c->bit_offset -= plain_bits;
1919 bm_xfer_ctx_bit_to_word_offset(c);
1920 c->bit_offset = c->word_offset * BITS_PER_LONG;
1924 /* RLE + VLI was able to compress it just fine.
1925 * update c->word_offset. */
1926 bm_xfer_ctx_bit_to_word_offset(c);
1928 /* store pad_bits */
1929 DCBP_set_pad_bits(p, (8 - bs.cur.bit) & 0x7);
1934 enum { OK, FAILED, DONE }
1935 send_bitmap_rle_or_plain(struct drbd_conf *mdev,
1936 struct p_header *h, struct bm_xfer_ctx *c)
1938 struct p_compressed_bm *p = (void*)h;
1939 unsigned long num_words;
1943 len = fill_bitmap_rle_bits(mdev, p, c);
1949 DCBP_set_code(p, RLE_VLI_Bits);
1950 ok = _drbd_send_cmd(mdev, mdev->data.socket, P_COMPRESSED_BITMAP, h,
1951 sizeof(*p) + len, 0);
1954 c->bytes[0] += sizeof(*p) + len;
1956 if (c->bit_offset >= c->bm_bits)
1959 /* was not compressible.
1960 * send a buffer full of plain text bits instead. */
1961 num_words = min_t(size_t, BM_PACKET_WORDS, c->bm_words - c->word_offset);
1962 len = num_words * sizeof(long);
1964 drbd_bm_get_lel(mdev, c->word_offset, num_words, (unsigned long*)h->payload);
1965 ok = _drbd_send_cmd(mdev, mdev->data.socket, P_BITMAP,
1966 h, sizeof(struct p_header) + len, 0);
1967 c->word_offset += num_words;
1968 c->bit_offset = c->word_offset * BITS_PER_LONG;
1971 c->bytes[1] += sizeof(struct p_header) + len;
1973 if (c->bit_offset > c->bm_bits)
1974 c->bit_offset = c->bm_bits;
1976 ok = ok ? ((len == 0) ? DONE : OK) : FAILED;
1979 INFO_bm_xfer_stats(mdev, "send", c);
1983 /* See the comment at receive_bitmap() */
1984 int _drbd_send_bitmap(struct drbd_conf *mdev)
1986 struct bm_xfer_ctx c;
1990 ERR_IF(!mdev->bitmap) return FALSE;
1992 /* maybe we should use some per thread scratch page,
1993 * and allocate that during initial device creation? */
1994 p = (struct p_header *) __get_free_page(GFP_NOIO);
1996 dev_err(DEV, "failed to allocate one page buffer in %s\n", __func__);
2000 if (get_ldev(mdev)) {
2001 if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) {
2002 dev_info(DEV, "Writing the whole bitmap, MDF_FullSync was set.\n");
2003 drbd_bm_set_all(mdev);
2004 if (drbd_bm_write(mdev)) {
2005 /* write_bm did fail! Leave full sync flag set in Meta P_DATA
2006 * but otherwise process as per normal - need to tell other
2007 * side that a full resync is required! */
2008 dev_err(DEV, "Failed to write bitmap to disk!\n");
2010 drbd_md_clear_flag(mdev, MDF_FULL_SYNC);
2017 c = (struct bm_xfer_ctx) {
2018 .bm_bits = drbd_bm_bits(mdev),
2019 .bm_words = drbd_bm_words(mdev),
2023 ret = send_bitmap_rle_or_plain(mdev, p, &c);
2024 } while (ret == OK);
2026 free_page((unsigned long) p);
2027 return (ret == DONE);
2030 int drbd_send_bitmap(struct drbd_conf *mdev)
2034 if (!drbd_get_data_sock(mdev))
2036 err = !_drbd_send_bitmap(mdev);
2037 drbd_put_data_sock(mdev);
2041 int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, u32 set_size)
2044 struct p_barrier_ack p;
2046 p.barrier = barrier_nr;
2047 p.set_size = cpu_to_be32(set_size);
2049 if (mdev->state.conn < C_CONNECTED)
2051 ok = drbd_send_cmd(mdev, USE_META_SOCKET, P_BARRIER_ACK,
2052 (struct p_header *)&p, sizeof(p));
2057 * _drbd_send_ack() - Sends an ack packet
2058 * @mdev: DRBD device.
2059 * @cmd: Packet command code.
2060 * @sector: sector, needs to be in big endian byte order
2061 * @blksize: size in byte, needs to be in big endian byte order
2062 * @block_id: Id, big endian byte order
2064 static int _drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
2070 struct p_block_ack p;
2073 p.block_id = block_id;
2074 p.blksize = blksize;
2075 p.seq_num = cpu_to_be32(atomic_add_return(1, &mdev->packet_seq));
2077 if (!mdev->meta.socket || mdev->state.conn < C_CONNECTED)
2079 ok = drbd_send_cmd(mdev, USE_META_SOCKET, cmd,
2080 (struct p_header *)&p, sizeof(p));
2084 int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
2087 const int header_size = sizeof(struct p_data)
2088 - sizeof(struct p_header);
2089 int data_size = ((struct p_header *)dp)->length - header_size;
2091 return _drbd_send_ack(mdev, cmd, dp->sector, cpu_to_be32(data_size),
2095 int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
2096 struct p_block_req *rp)
2098 return _drbd_send_ack(mdev, cmd, rp->sector, rp->blksize, rp->block_id);
2102 * drbd_send_ack() - Sends an ack packet
2103 * @mdev: DRBD device.
2104 * @cmd: Packet command code.
2107 int drbd_send_ack(struct drbd_conf *mdev,
2108 enum drbd_packets cmd, struct drbd_epoch_entry *e)
2110 return _drbd_send_ack(mdev, cmd,
2111 cpu_to_be64(e->sector),
2112 cpu_to_be32(e->size),
2116 /* This function misuses the block_id field to signal if the blocks
2117 * are is sync or not. */
2118 int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
2119 sector_t sector, int blksize, u64 block_id)
2121 return _drbd_send_ack(mdev, cmd,
2122 cpu_to_be64(sector),
2123 cpu_to_be32(blksize),
2124 cpu_to_be64(block_id));
2127 int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
2128 sector_t sector, int size, u64 block_id)
2131 struct p_block_req p;
2133 p.sector = cpu_to_be64(sector);
2134 p.block_id = block_id;
2135 p.blksize = cpu_to_be32(size);
2137 ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd,
2138 (struct p_header *)&p, sizeof(p));
2142 int drbd_send_drequest_csum(struct drbd_conf *mdev,
2143 sector_t sector, int size,
2144 void *digest, int digest_size,
2145 enum drbd_packets cmd)
2148 struct p_block_req p;
2150 p.sector = cpu_to_be64(sector);
2151 p.block_id = BE_DRBD_MAGIC + 0xbeef;
2152 p.blksize = cpu_to_be32(size);
2154 p.head.magic = BE_DRBD_MAGIC;
2155 p.head.command = cpu_to_be16(cmd);
2156 p.head.length = cpu_to_be16(sizeof(p) - sizeof(struct p_header) + digest_size);
2158 mutex_lock(&mdev->data.mutex);
2160 ok = (sizeof(p) == drbd_send(mdev, mdev->data.socket, &p, sizeof(p), 0));
2161 ok = ok && (digest_size == drbd_send(mdev, mdev->data.socket, digest, digest_size, 0));
2163 mutex_unlock(&mdev->data.mutex);
2168 int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size)
2171 struct p_block_req p;
2173 p.sector = cpu_to_be64(sector);
2174 p.block_id = BE_DRBD_MAGIC + 0xbabe;
2175 p.blksize = cpu_to_be32(size);
2177 ok = drbd_send_cmd(mdev, USE_DATA_SOCKET, P_OV_REQUEST,
2178 (struct p_header *)&p, sizeof(p));
2182 /* called on sndtimeo
2183 * returns FALSE if we should retry,
2184 * TRUE if we think connection is dead
2186 static int we_should_drop_the_connection(struct drbd_conf *mdev, struct socket *sock)
2189 /* long elapsed = (long)(jiffies - mdev->last_received); */
2191 drop_it = mdev->meta.socket == sock
2192 || !mdev->asender.task
2193 || get_t_state(&mdev->asender) != Running
2194 || mdev->state.conn < C_CONNECTED;
2199 drop_it = !--mdev->ko_count;
2201 dev_err(DEV, "[%s/%d] sock_sendmsg time expired, ko = %u\n",
2202 current->comm, current->pid, mdev->ko_count);
2206 return drop_it; /* && (mdev->state == R_PRIMARY) */;
2209 /* The idea of sendpage seems to be to put some kind of reference
2210 * to the page into the skb, and to hand it over to the NIC. In
2211 * this process get_page() gets called.
2213 * As soon as the page was really sent over the network put_page()
2214 * gets called by some part of the network layer. [ NIC driver? ]
2216 * [ get_page() / put_page() increment/decrement the count. If count
2217 * reaches 0 the page will be freed. ]
2219 * This works nicely with pages from FSs.
2220 * But this means that in protocol A we might signal IO completion too early!
2222 * In order not to corrupt data during a resync we must make sure
2223 * that we do not reuse our own buffer pages (EEs) to early, therefore
2224 * we have the net_ee list.
2226 * XFS seems to have problems, still, it submits pages with page_count == 0!
2227 * As a workaround, we disable sendpage on pages
2228 * with page_count == 0 or PageSlab.
2230 static int _drbd_no_send_page(struct drbd_conf *mdev, struct page *page,
2231 int offset, size_t size)
2233 int sent = drbd_send(mdev, mdev->data.socket, kmap(page) + offset, size, 0);
2236 mdev->send_cnt += size>>9;
2237 return sent == size;
2240 static int _drbd_send_page(struct drbd_conf *mdev, struct page *page,
2241 int offset, size_t size)
2243 mm_segment_t oldfs = get_fs();
2247 /* e.g. XFS meta- & log-data is in slab pages, which have a
2248 * page_count of 0 and/or have PageSlab() set.
2249 * we cannot use send_page for those, as that does get_page();
2250 * put_page(); and would cause either a VM_BUG directly, or
2251 * __page_cache_release a page that would actually still be referenced
2252 * by someone, leading to some obscure delayed Oops somewhere else. */
2253 if (disable_sendpage || (page_count(page) < 1) || PageSlab(page))
2254 return _drbd_no_send_page(mdev, page, offset, size);
2256 drbd_update_congested(mdev);
2259 sent = mdev->data.socket->ops->sendpage(mdev->data.socket, page,
2262 if (sent == -EAGAIN) {
2263 if (we_should_drop_the_connection(mdev,
2270 dev_warn(DEV, "%s: size=%d len=%d sent=%d\n",
2271 __func__, (int)size, len, sent);
2276 } while (len > 0 /* THINK && mdev->cstate >= C_CONNECTED*/);
2278 clear_bit(NET_CONGESTED, &mdev->flags);
2282 mdev->send_cnt += size>>9;
2286 static int _drbd_send_bio(struct drbd_conf *mdev, struct bio *bio)
2288 struct bio_vec *bvec;
2290 __bio_for_each_segment(bvec, bio, i, 0) {
2291 if (!_drbd_no_send_page(mdev, bvec->bv_page,
2292 bvec->bv_offset, bvec->bv_len))
2298 static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio)
2300 struct bio_vec *bvec;
2302 __bio_for_each_segment(bvec, bio, i, 0) {
2303 if (!_drbd_send_page(mdev, bvec->bv_page,
2304 bvec->bv_offset, bvec->bv_len))
2311 /* Used to send write requests
2312 * R_PRIMARY -> Peer (P_DATA)
2314 int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req)
2318 unsigned int dp_flags = 0;
2322 if (!drbd_get_data_sock(mdev))
2325 dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_w_tfm) ?
2326 crypto_hash_digestsize(mdev->integrity_w_tfm) : 0;
2328 p.head.magic = BE_DRBD_MAGIC;
2329 p.head.command = cpu_to_be16(P_DATA);
2331 cpu_to_be16(sizeof(p) - sizeof(struct p_header) + dgs + req->size);
2333 p.sector = cpu_to_be64(req->sector);
2334 p.block_id = (unsigned long)req;
2335 p.seq_num = cpu_to_be32(req->seq_num =
2336 atomic_add_return(1, &mdev->packet_seq));
2339 /* NOTE: no need to check if barriers supported here as we would
2340 * not pass the test in make_request_common in that case
2342 if (bio_rw_flagged(req->master_bio, BIO_RW_BARRIER)) {
2343 dev_err(DEV, "ASSERT FAILED would have set DP_HARDBARRIER\n");
2344 /* dp_flags |= DP_HARDBARRIER; */
2346 if (bio_rw_flagged(req->master_bio, BIO_RW_SYNCIO))
2347 dp_flags |= DP_RW_SYNC;
2348 /* for now handle SYNCIO and UNPLUG
2349 * as if they still were one and the same flag */
2350 if (bio_rw_flagged(req->master_bio, BIO_RW_UNPLUG))
2351 dp_flags |= DP_RW_SYNC;
2352 if (mdev->state.conn >= C_SYNC_SOURCE &&
2353 mdev->state.conn <= C_PAUSED_SYNC_T)
2354 dp_flags |= DP_MAY_SET_IN_SYNC;
2356 p.dp_flags = cpu_to_be32(dp_flags);
2357 set_bit(UNPLUG_REMOTE, &mdev->flags);
2359 drbd_send(mdev, mdev->data.socket, &p, sizeof(p), MSG_MORE));
2361 dgb = mdev->int_dig_out;
2362 drbd_csum(mdev, mdev->integrity_w_tfm, req->master_bio, dgb);
2363 ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, MSG_MORE);
2366 if (mdev->net_conf->wire_protocol == DRBD_PROT_A)
2367 ok = _drbd_send_bio(mdev, req->master_bio);
2369 ok = _drbd_send_zc_bio(mdev, req->master_bio);
2372 drbd_put_data_sock(mdev);
2376 /* answer packet, used to send data back for read requests:
2377 * Peer -> (diskless) R_PRIMARY (P_DATA_REPLY)
2378 * C_SYNC_SOURCE -> C_SYNC_TARGET (P_RS_DATA_REPLY)
2380 int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
2381 struct drbd_epoch_entry *e)
2388 dgs = (mdev->agreed_pro_version >= 87 && mdev->integrity_w_tfm) ?
2389 crypto_hash_digestsize(mdev->integrity_w_tfm) : 0;
2391 p.head.magic = BE_DRBD_MAGIC;
2392 p.head.command = cpu_to_be16(cmd);
2394 cpu_to_be16(sizeof(p) - sizeof(struct p_header) + dgs + e->size);
2396 p.sector = cpu_to_be64(e->sector);
2397 p.block_id = e->block_id;
2398 /* p.seq_num = 0; No sequence numbers here.. */
2400 /* Only called by our kernel thread.
2401 * This one may be interrupted by DRBD_SIG and/or DRBD_SIGKILL
2402 * in response to admin command or module unload.
2404 if (!drbd_get_data_sock(mdev))
2407 ok = sizeof(p) == drbd_send(mdev, mdev->data.socket, &p,
2408 sizeof(p), MSG_MORE);
2410 dgb = mdev->int_dig_out;
2411 drbd_csum(mdev, mdev->integrity_w_tfm, e->private_bio, dgb);
2412 ok = drbd_send(mdev, mdev->data.socket, dgb, dgs, MSG_MORE);
2415 ok = _drbd_send_zc_bio(mdev, e->private_bio);
2417 drbd_put_data_sock(mdev);
2422 drbd_send distinguishes two cases:
2424 Packets sent via the data socket "sock"
2425 and packets sent via the meta data socket "msock"
2428 -----------------+-------------------------+------------------------------
2429 timeout conf.timeout / 2 conf.timeout / 2
2430 timeout action send a ping via msock Abort communication
2431 and close all sockets
2435 * you must have down()ed the appropriate [m]sock_mutex elsewhere!
2437 int drbd_send(struct drbd_conf *mdev, struct socket *sock,
2438 void *buf, size_t size, unsigned msg_flags)
2447 /* THINK if (signal_pending) return ... ? */
2452 msg.msg_name = NULL;
2453 msg.msg_namelen = 0;
2454 msg.msg_control = NULL;
2455 msg.msg_controllen = 0;
2456 msg.msg_flags = msg_flags | MSG_NOSIGNAL;
2458 if (sock == mdev->data.socket) {
2459 mdev->ko_count = mdev->net_conf->ko_count;
2460 drbd_update_congested(mdev);
2464 * tcp_sendmsg does _not_ use its size parameter at all ?
2466 * -EAGAIN on timeout, -EINTR on signal.
2469 * do we need to block DRBD_SIG if sock == &meta.socket ??
2470 * otherwise wake_asender() might interrupt some send_*Ack !
2472 rv = kernel_sendmsg(sock, &msg, &iov, 1, size);
2473 if (rv == -EAGAIN) {
2474 if (we_should_drop_the_connection(mdev, sock))
2481 flush_signals(current);
2489 } while (sent < size);
2491 if (sock == mdev->data.socket)
2492 clear_bit(NET_CONGESTED, &mdev->flags);
2495 if (rv != -EAGAIN) {
2496 dev_err(DEV, "%s_sendmsg returned %d\n",
2497 sock == mdev->meta.socket ? "msock" : "sock",
2499 drbd_force_state(mdev, NS(conn, C_BROKEN_PIPE));
2501 drbd_force_state(mdev, NS(conn, C_TIMEOUT));
2507 static int drbd_open(struct block_device *bdev, fmode_t mode)
2509 struct drbd_conf *mdev = bdev->bd_disk->private_data;
2510 unsigned long flags;
2513 spin_lock_irqsave(&mdev->req_lock, flags);
2514 /* to have a stable mdev->state.role
2515 * and no race with updating open_cnt */
2517 if (mdev->state.role != R_PRIMARY) {
2518 if (mode & FMODE_WRITE)
2520 else if (!allow_oos)
2526 spin_unlock_irqrestore(&mdev->req_lock, flags);
2531 static int drbd_release(struct gendisk *gd, fmode_t mode)
2533 struct drbd_conf *mdev = gd->private_data;
2538 static void drbd_unplug_fn(struct request_queue *q)
2540 struct drbd_conf *mdev = q->queuedata;
2543 spin_lock_irq(q->queue_lock);
2545 spin_unlock_irq(q->queue_lock);
2547 /* only if connected */
2548 spin_lock_irq(&mdev->req_lock);
2549 if (mdev->state.pdsk >= D_INCONSISTENT && mdev->state.conn >= C_CONNECTED) {
2550 D_ASSERT(mdev->state.role == R_PRIMARY);
2551 if (test_and_clear_bit(UNPLUG_REMOTE, &mdev->flags)) {
2552 /* add to the data.work queue,
2553 * unless already queued.
2554 * XXX this might be a good addition to drbd_queue_work
2555 * anyways, to detect "double queuing" ... */
2556 if (list_empty(&mdev->unplug_work.list))
2557 drbd_queue_work(&mdev->data.work,
2558 &mdev->unplug_work);
2561 spin_unlock_irq(&mdev->req_lock);
2563 if (mdev->state.disk >= D_INCONSISTENT)
2567 static void drbd_set_defaults(struct drbd_conf *mdev)
2569 mdev->sync_conf.after = DRBD_AFTER_DEF;
2570 mdev->sync_conf.rate = DRBD_RATE_DEF;
2571 mdev->sync_conf.al_extents = DRBD_AL_EXTENTS_DEF;
2572 mdev->state = (union drbd_state) {
2573 { .role = R_SECONDARY,
2575 .conn = C_STANDALONE,
2582 void drbd_init_set_defaults(struct drbd_conf *mdev)
2584 /* the memset(,0,) did most of this.
2585 * note: only assignments, no allocation in here */
2587 drbd_set_defaults(mdev);
2589 /* for now, we do NOT yet support it,
2590 * even though we start some framework
2591 * to eventually support barriers */
2592 set_bit(NO_BARRIER_SUPP, &mdev->flags);
2594 atomic_set(&mdev->ap_bio_cnt, 0);
2595 atomic_set(&mdev->ap_pending_cnt, 0);
2596 atomic_set(&mdev->rs_pending_cnt, 0);
2597 atomic_set(&mdev->unacked_cnt, 0);
2598 atomic_set(&mdev->local_cnt, 0);
2599 atomic_set(&mdev->net_cnt, 0);
2600 atomic_set(&mdev->packet_seq, 0);
2601 atomic_set(&mdev->pp_in_use, 0);
2603 mutex_init(&mdev->md_io_mutex);
2604 mutex_init(&mdev->data.mutex);
2605 mutex_init(&mdev->meta.mutex);
2606 sema_init(&mdev->data.work.s, 0);
2607 sema_init(&mdev->meta.work.s, 0);
2608 mutex_init(&mdev->state_mutex);
2610 spin_lock_init(&mdev->data.work.q_lock);
2611 spin_lock_init(&mdev->meta.work.q_lock);
2613 spin_lock_init(&mdev->al_lock);
2614 spin_lock_init(&mdev->req_lock);
2615 spin_lock_init(&mdev->peer_seq_lock);
2616 spin_lock_init(&mdev->epoch_lock);
2618 INIT_LIST_HEAD(&mdev->active_ee);
2619 INIT_LIST_HEAD(&mdev->sync_ee);
2620 INIT_LIST_HEAD(&mdev->done_ee);
2621 INIT_LIST_HEAD(&mdev->read_ee);
2622 INIT_LIST_HEAD(&mdev->net_ee);
2623 INIT_LIST_HEAD(&mdev->resync_reads);
2624 INIT_LIST_HEAD(&mdev->data.work.q);
2625 INIT_LIST_HEAD(&mdev->meta.work.q);
2626 INIT_LIST_HEAD(&mdev->resync_work.list);
2627 INIT_LIST_HEAD(&mdev->unplug_work.list);
2628 INIT_LIST_HEAD(&mdev->md_sync_work.list);
2629 INIT_LIST_HEAD(&mdev->bm_io_work.w.list);
2630 mdev->resync_work.cb = w_resync_inactive;
2631 mdev->unplug_work.cb = w_send_write_hint;
2632 mdev->md_sync_work.cb = w_md_sync;
2633 mdev->bm_io_work.w.cb = w_bitmap_io;
2634 init_timer(&mdev->resync_timer);
2635 init_timer(&mdev->md_sync_timer);
2636 mdev->resync_timer.function = resync_timer_fn;
2637 mdev->resync_timer.data = (unsigned long) mdev;
2638 mdev->md_sync_timer.function = md_sync_timer_fn;
2639 mdev->md_sync_timer.data = (unsigned long) mdev;
2641 init_waitqueue_head(&mdev->misc_wait);
2642 init_waitqueue_head(&mdev->state_wait);
2643 init_waitqueue_head(&mdev->ee_wait);
2644 init_waitqueue_head(&mdev->al_wait);
2645 init_waitqueue_head(&mdev->seq_wait);
2647 drbd_thread_init(mdev, &mdev->receiver, drbdd_init);
2648 drbd_thread_init(mdev, &mdev->worker, drbd_worker);
2649 drbd_thread_init(mdev, &mdev->asender, drbd_asender);
2651 mdev->agreed_pro_version = PRO_VERSION_MAX;
2652 mdev->write_ordering = WO_bio_barrier;
2653 mdev->resync_wenr = LC_FREE;
2656 void drbd_mdev_cleanup(struct drbd_conf *mdev)
2658 if (mdev->receiver.t_state != None)
2659 dev_err(DEV, "ASSERT FAILED: receiver t_state == %d expected 0.\n",
2660 mdev->receiver.t_state);
2662 /* no need to lock it, I'm the only thread alive */
2663 if (atomic_read(&mdev->current_epoch->epoch_size) != 0)
2664 dev_err(DEV, "epoch_size:%d\n", atomic_read(&mdev->current_epoch->epoch_size));
2675 mdev->rs_mark_left =
2676 mdev->rs_mark_time = 0;
2677 D_ASSERT(mdev->net_conf == NULL);
2679 drbd_set_my_capacity(mdev, 0);
2681 /* maybe never allocated. */
2682 drbd_bm_resize(mdev, 0);
2683 drbd_bm_cleanup(mdev);
2686 drbd_free_resources(mdev);
2689 * currently we drbd_init_ee only on module load, so
2690 * we may do drbd_release_ee only on module unload!
2692 D_ASSERT(list_empty(&mdev->active_ee));
2693 D_ASSERT(list_empty(&mdev->sync_ee));
2694 D_ASSERT(list_empty(&mdev->done_ee));
2695 D_ASSERT(list_empty(&mdev->read_ee));
2696 D_ASSERT(list_empty(&mdev->net_ee));
2697 D_ASSERT(list_empty(&mdev->resync_reads));
2698 D_ASSERT(list_empty(&mdev->data.work.q));
2699 D_ASSERT(list_empty(&mdev->meta.work.q));
2700 D_ASSERT(list_empty(&mdev->resync_work.list));
2701 D_ASSERT(list_empty(&mdev->unplug_work.list));
2706 static void drbd_destroy_mempools(void)
2710 while (drbd_pp_pool) {
2711 page = drbd_pp_pool;
2712 drbd_pp_pool = (struct page *)page_private(page);
2717 /* D_ASSERT(atomic_read(&drbd_pp_vacant)==0); */
2719 if (drbd_ee_mempool)
2720 mempool_destroy(drbd_ee_mempool);
2721 if (drbd_request_mempool)
2722 mempool_destroy(drbd_request_mempool);
2724 kmem_cache_destroy(drbd_ee_cache);
2725 if (drbd_request_cache)
2726 kmem_cache_destroy(drbd_request_cache);
2727 if (drbd_bm_ext_cache)
2728 kmem_cache_destroy(drbd_bm_ext_cache);
2729 if (drbd_al_ext_cache)
2730 kmem_cache_destroy(drbd_al_ext_cache);
2732 drbd_ee_mempool = NULL;
2733 drbd_request_mempool = NULL;
2734 drbd_ee_cache = NULL;
2735 drbd_request_cache = NULL;
2736 drbd_bm_ext_cache = NULL;
2737 drbd_al_ext_cache = NULL;
2742 static int drbd_create_mempools(void)
2745 const int number = (DRBD_MAX_SEGMENT_SIZE/PAGE_SIZE) * minor_count;
2748 /* prepare our caches and mempools */
2749 drbd_request_mempool = NULL;
2750 drbd_ee_cache = NULL;
2751 drbd_request_cache = NULL;
2752 drbd_bm_ext_cache = NULL;
2753 drbd_al_ext_cache = NULL;
2754 drbd_pp_pool = NULL;
2757 drbd_request_cache = kmem_cache_create(
2758 "drbd_req", sizeof(struct drbd_request), 0, 0, NULL);
2759 if (drbd_request_cache == NULL)
2762 drbd_ee_cache = kmem_cache_create(
2763 "drbd_ee", sizeof(struct drbd_epoch_entry), 0, 0, NULL);
2764 if (drbd_ee_cache == NULL)
2767 drbd_bm_ext_cache = kmem_cache_create(
2768 "drbd_bm", sizeof(struct bm_extent), 0, 0, NULL);
2769 if (drbd_bm_ext_cache == NULL)
2772 drbd_al_ext_cache = kmem_cache_create(
2773 "drbd_al", sizeof(struct lc_element), 0, 0, NULL);
2774 if (drbd_al_ext_cache == NULL)
2778 drbd_request_mempool = mempool_create(number,
2779 mempool_alloc_slab, mempool_free_slab, drbd_request_cache);
2780 if (drbd_request_mempool == NULL)
2783 drbd_ee_mempool = mempool_create(number,
2784 mempool_alloc_slab, mempool_free_slab, drbd_ee_cache);
2785 if (drbd_request_mempool == NULL)
2788 /* drbd's page pool */
2789 spin_lock_init(&drbd_pp_lock);
2791 for (i = 0; i < number; i++) {
2792 page = alloc_page(GFP_HIGHUSER);
2795 set_page_private(page, (unsigned long)drbd_pp_pool);
2796 drbd_pp_pool = page;
2798 drbd_pp_vacant = number;
2803 drbd_destroy_mempools(); /* in case we allocated some */
2807 static int drbd_notify_sys(struct notifier_block *this, unsigned long code,
2810 /* just so we have it. you never know what interesting things we
2811 * might want to do here some day...
2817 static struct notifier_block drbd_notifier = {
2818 .notifier_call = drbd_notify_sys,
2821 static void drbd_release_ee_lists(struct drbd_conf *mdev)
2825 rr = drbd_release_ee(mdev, &mdev->active_ee);
2827 dev_err(DEV, "%d EEs in active list found!\n", rr);
2829 rr = drbd_release_ee(mdev, &mdev->sync_ee);
2831 dev_err(DEV, "%d EEs in sync list found!\n", rr);
2833 rr = drbd_release_ee(mdev, &mdev->read_ee);
2835 dev_err(DEV, "%d EEs in read list found!\n", rr);
2837 rr = drbd_release_ee(mdev, &mdev->done_ee);
2839 dev_err(DEV, "%d EEs in done list found!\n", rr);
2841 rr = drbd_release_ee(mdev, &mdev->net_ee);
2843 dev_err(DEV, "%d EEs in net list found!\n", rr);
2846 /* caution. no locking.
2847 * currently only used from module cleanup code. */
2848 static void drbd_delete_device(unsigned int minor)
2850 struct drbd_conf *mdev = minor_to_mdev(minor);
2855 /* paranoia asserts */
2856 if (mdev->open_cnt != 0)
2857 dev_err(DEV, "open_cnt = %d in %s:%u", mdev->open_cnt,
2858 __FILE__ , __LINE__);
2860 ERR_IF (!list_empty(&mdev->data.work.q)) {
2861 struct list_head *lp;
2862 list_for_each(lp, &mdev->data.work.q) {
2863 dev_err(DEV, "lp = %p\n", lp);
2866 /* end paranoia asserts */
2868 del_gendisk(mdev->vdisk);
2870 /* cleanup stuff that may have been allocated during
2871 * device (re-)configuration or state changes */
2873 if (mdev->this_bdev)
2874 bdput(mdev->this_bdev);
2876 drbd_free_resources(mdev);
2878 drbd_release_ee_lists(mdev);
2880 /* should be free'd on disconnect? */
2881 kfree(mdev->ee_hash);
2883 mdev->ee_hash_s = 0;
2884 mdev->ee_hash = NULL;
2887 lc_destroy(mdev->act_log);
2888 lc_destroy(mdev->resync);
2890 kfree(mdev->p_uuid);
2891 /* mdev->p_uuid = NULL; */
2893 kfree(mdev->int_dig_out);
2894 kfree(mdev->int_dig_in);
2895 kfree(mdev->int_dig_vv);
2897 /* cleanup the rest that has been
2898 * allocated from drbd_new_device
2899 * and actually free the mdev itself */
2900 drbd_free_mdev(mdev);
2903 static void drbd_cleanup(void)
2907 unregister_reboot_notifier(&drbd_notifier);
2913 remove_proc_entry("drbd", NULL);
2916 drbd_delete_device(i);
2917 drbd_destroy_mempools();
2922 unregister_blkdev(DRBD_MAJOR, "drbd");
2924 printk(KERN_INFO "drbd: module cleanup done.\n");
2928 * drbd_congested() - Callback for pdflush
2929 * @congested_data: User data
2930 * @bdi_bits: Bits pdflush is currently interested in
2932 * Returns 1<<BDI_async_congested and/or 1<<BDI_sync_congested if we are congested.
2934 static int drbd_congested(void *congested_data, int bdi_bits)
2936 struct drbd_conf *mdev = congested_data;
2937 struct request_queue *q;
2941 if (!__inc_ap_bio_cond(mdev)) {
2942 /* DRBD has frozen IO */
2948 if (get_ldev(mdev)) {
2949 q = bdev_get_queue(mdev->ldev->backing_bdev);
2950 r = bdi_congested(&q->backing_dev_info, bdi_bits);
2956 if (bdi_bits & (1 << BDI_async_congested) && test_bit(NET_CONGESTED, &mdev->flags)) {
2957 r |= (1 << BDI_async_congested);
2958 reason = reason == 'b' ? 'a' : 'n';
2962 mdev->congestion_reason = reason;
2966 struct drbd_conf *drbd_new_device(unsigned int minor)
2968 struct drbd_conf *mdev;
2969 struct gendisk *disk;
2970 struct request_queue *q;
2972 /* GFP_KERNEL, we are outside of all write-out paths */
2973 mdev = kzalloc(sizeof(struct drbd_conf), GFP_KERNEL);
2976 if (!zalloc_cpumask_var(&mdev->cpu_mask, GFP_KERNEL))
2977 goto out_no_cpumask;
2979 mdev->minor = minor;
2981 drbd_init_set_defaults(mdev);
2983 q = blk_alloc_queue(GFP_KERNEL);
2987 q->queuedata = mdev;
2989 disk = alloc_disk(1);
2994 set_disk_ro(disk, TRUE);
2997 disk->major = DRBD_MAJOR;
2998 disk->first_minor = minor;
2999 disk->fops = &drbd_ops;
3000 sprintf(disk->disk_name, "drbd%d", minor);
3001 disk->private_data = mdev;
3003 mdev->this_bdev = bdget(MKDEV(DRBD_MAJOR, minor));
3004 /* we have no partitions. we contain only ourselves. */
3005 mdev->this_bdev->bd_contains = mdev->this_bdev;
3007 q->backing_dev_info.congested_fn = drbd_congested;
3008 q->backing_dev_info.congested_data = mdev;
3010 blk_queue_make_request(q, drbd_make_request_26);
3011 blk_queue_max_segment_size(q, DRBD_MAX_SEGMENT_SIZE);
3012 blk_queue_bounce_limit(q, BLK_BOUNCE_ANY);
3013 blk_queue_merge_bvec(q, drbd_merge_bvec);
3014 q->queue_lock = &mdev->req_lock; /* needed since we use */
3015 /* plugging on a queue, that actually has no requests! */
3016 q->unplug_fn = drbd_unplug_fn;
3018 mdev->md_io_page = alloc_page(GFP_KERNEL);
3019 if (!mdev->md_io_page)
3020 goto out_no_io_page;
3022 if (drbd_bm_init(mdev))
3024 /* no need to lock access, we are still initializing this minor device. */
3028 mdev->app_reads_hash = kzalloc(APP_R_HSIZE*sizeof(void *), GFP_KERNEL);
3029 if (!mdev->app_reads_hash)
3030 goto out_no_app_reads;
3032 mdev->current_epoch = kzalloc(sizeof(struct drbd_epoch), GFP_KERNEL);
3033 if (!mdev->current_epoch)
3036 INIT_LIST_HEAD(&mdev->current_epoch->list);
3041 /* out_whatever_else:
3042 kfree(mdev->current_epoch); */
3044 kfree(mdev->app_reads_hash);
3048 drbd_bm_cleanup(mdev);
3050 __free_page(mdev->md_io_page);
3054 blk_cleanup_queue(q);
3056 free_cpumask_var(mdev->cpu_mask);
3062 /* counterpart of drbd_new_device.
3063 * last part of drbd_delete_device. */
3064 void drbd_free_mdev(struct drbd_conf *mdev)
3066 kfree(mdev->current_epoch);
3067 kfree(mdev->app_reads_hash);
3069 if (mdev->bitmap) /* should no longer be there. */
3070 drbd_bm_cleanup(mdev);
3071 __free_page(mdev->md_io_page);
3072 put_disk(mdev->vdisk);
3073 blk_cleanup_queue(mdev->rq_queue);
3074 free_cpumask_var(mdev->cpu_mask);
3079 int __init drbd_init(void)
3083 if (sizeof(struct p_handshake) != 80) {
3085 "drbd: never change the size or layout "
3086 "of the HandShake packet.\n");
3090 if (1 > minor_count || minor_count > 255) {
3092 "drbd: invalid minor_count (%d)\n", minor_count);
3100 err = drbd_nl_init();
3104 err = register_blkdev(DRBD_MAJOR, "drbd");
3107 "drbd: unable to register block device major %d\n",
3112 register_reboot_notifier(&drbd_notifier);
3115 * allocate all necessary structs
3119 init_waitqueue_head(&drbd_pp_wait);
3121 drbd_proc = NULL; /* play safe for drbd_cleanup */
3122 minor_table = kzalloc(sizeof(struct drbd_conf *)*minor_count,
3127 err = drbd_create_mempools();
3131 drbd_proc = proc_create("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops);
3133 printk(KERN_ERR "drbd: unable to register proc file\n");
3137 rwlock_init(&global_state_lock);
3139 printk(KERN_INFO "drbd: initialized. "
3140 "Version: " REL_VERSION " (api:%d/proto:%d-%d)\n",
3141 API_VERSION, PRO_VERSION_MIN, PRO_VERSION_MAX);
3142 printk(KERN_INFO "drbd: %s\n", drbd_buildtag());
3143 printk(KERN_INFO "drbd: registered as block device major %d\n",
3145 printk(KERN_INFO "drbd: minor_table @ 0x%p\n", minor_table);
3147 return 0; /* Success! */
3152 /* currently always the case */
3153 printk(KERN_ERR "drbd: ran out of memory\n");
3155 printk(KERN_ERR "drbd: initialization failure\n");
3159 void drbd_free_bc(struct drbd_backing_dev *ldev)
3164 bd_release(ldev->backing_bdev);
3165 bd_release(ldev->md_bdev);
3167 fput(ldev->lo_file);
3168 fput(ldev->md_file);
3173 void drbd_free_sock(struct drbd_conf *mdev)
3175 if (mdev->data.socket) {
3176 mutex_lock(&mdev->data.mutex);
3177 kernel_sock_shutdown(mdev->data.socket, SHUT_RDWR);
3178 sock_release(mdev->data.socket);
3179 mdev->data.socket = NULL;
3180 mutex_unlock(&mdev->data.mutex);
3182 if (mdev->meta.socket) {
3183 mutex_lock(&mdev->meta.mutex);
3184 kernel_sock_shutdown(mdev->meta.socket, SHUT_RDWR);
3185 sock_release(mdev->meta.socket);
3186 mdev->meta.socket = NULL;
3187 mutex_unlock(&mdev->meta.mutex);
3192 void drbd_free_resources(struct drbd_conf *mdev)
3194 crypto_free_hash(mdev->csums_tfm);
3195 mdev->csums_tfm = NULL;
3196 crypto_free_hash(mdev->verify_tfm);
3197 mdev->verify_tfm = NULL;
3198 crypto_free_hash(mdev->cram_hmac_tfm);
3199 mdev->cram_hmac_tfm = NULL;
3200 crypto_free_hash(mdev->integrity_w_tfm);
3201 mdev->integrity_w_tfm = NULL;
3202 crypto_free_hash(mdev->integrity_r_tfm);
3203 mdev->integrity_r_tfm = NULL;
3205 drbd_free_sock(mdev);
3208 drbd_free_bc(mdev->ldev);
3209 mdev->ldev = NULL;);
3212 /* meta data management */
3214 struct meta_data_on_disk {
3215 u64 la_size; /* last agreed size. */
3216 u64 uuid[UI_SIZE]; /* UUIDs. */
3219 u32 flags; /* MDF */
3222 u32 al_offset; /* offset to this block */
3223 u32 al_nr_extents; /* important for restoring the AL */
3224 /* `-- act_log->nr_elements <-- sync_conf.al_extents */
3225 u32 bm_offset; /* offset to the bitmap, from here */
3226 u32 bm_bytes_per_bit; /* BM_BLOCK_SIZE */
3227 u32 reserved_u32[4];
3232 * drbd_md_sync() - Writes the meta data super block if the MD_DIRTY flag bit is set
3233 * @mdev: DRBD device.
3235 void drbd_md_sync(struct drbd_conf *mdev)
3237 struct meta_data_on_disk *buffer;
3241 if (!test_and_clear_bit(MD_DIRTY, &mdev->flags))
3243 del_timer(&mdev->md_sync_timer);
3245 /* We use here D_FAILED and not D_ATTACHING because we try to write
3246 * metadata even if we detach due to a disk failure! */
3247 if (!get_ldev_if_state(mdev, D_FAILED))
3250 mutex_lock(&mdev->md_io_mutex);
3251 buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page);
3252 memset(buffer, 0, 512);
3254 buffer->la_size = cpu_to_be64(drbd_get_capacity(mdev->this_bdev));
3255 for (i = UI_CURRENT; i < UI_SIZE; i++)
3256 buffer->uuid[i] = cpu_to_be64(mdev->ldev->md.uuid[i]);
3257 buffer->flags = cpu_to_be32(mdev->ldev->md.flags);
3258 buffer->magic = cpu_to_be32(DRBD_MD_MAGIC);
3260 buffer->md_size_sect = cpu_to_be32(mdev->ldev->md.md_size_sect);
3261 buffer->al_offset = cpu_to_be32(mdev->ldev->md.al_offset);
3262 buffer->al_nr_extents = cpu_to_be32(mdev->act_log->nr_elements);
3263 buffer->bm_bytes_per_bit = cpu_to_be32(BM_BLOCK_SIZE);
3264 buffer->device_uuid = cpu_to_be64(mdev->ldev->md.device_uuid);
3266 buffer->bm_offset = cpu_to_be32(mdev->ldev->md.bm_offset);
3268 D_ASSERT(drbd_md_ss__(mdev, mdev->ldev) == mdev->ldev->md.md_offset);
3269 sector = mdev->ldev->md.md_offset;
3271 if (drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) {
3272 clear_bit(MD_DIRTY, &mdev->flags);
3274 /* this was a try anyways ... */
3275 dev_err(DEV, "meta data update failed!\n");
3277 drbd_chk_io_error(mdev, 1, TRUE);
3280 /* Update mdev->ldev->md.la_size_sect,
3281 * since we updated it on metadata. */
3282 mdev->ldev->md.la_size_sect = drbd_get_capacity(mdev->this_bdev);
3284 mutex_unlock(&mdev->md_io_mutex);
3289 * drbd_md_read() - Reads in the meta data super block
3290 * @mdev: DRBD device.
3291 * @bdev: Device from which the meta data should be read in.
3293 * Return 0 (NO_ERROR) on success, and an enum drbd_ret_codes in case
3294 * something goes wrong. Currently only: ERR_IO_MD_DISK, ERR_MD_INVALID.
3296 int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev)
3298 struct meta_data_on_disk *buffer;
3299 int i, rv = NO_ERROR;
3301 if (!get_ldev_if_state(mdev, D_ATTACHING))
3302 return ERR_IO_MD_DISK;
3304 mutex_lock(&mdev->md_io_mutex);
3305 buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page);
3307 if (!drbd_md_sync_page_io(mdev, bdev, bdev->md.md_offset, READ)) {
3308 /* NOTE: cant do normal error processing here as this is
3309 called BEFORE disk is attached */
3310 dev_err(DEV, "Error while reading metadata.\n");
3311 rv = ERR_IO_MD_DISK;
3315 if (be32_to_cpu(buffer->magic) != DRBD_MD_MAGIC) {
3316 dev_err(DEV, "Error while reading metadata, magic not found.\n");
3317 rv = ERR_MD_INVALID;
3320 if (be32_to_cpu(buffer->al_offset) != bdev->md.al_offset) {
3321 dev_err(DEV, "unexpected al_offset: %d (expected %d)\n",
3322 be32_to_cpu(buffer->al_offset), bdev->md.al_offset);
3323 rv = ERR_MD_INVALID;
3326 if (be32_to_cpu(buffer->bm_offset) != bdev->md.bm_offset) {
3327 dev_err(DEV, "unexpected bm_offset: %d (expected %d)\n",
3328 be32_to_cpu(buffer->bm_offset), bdev->md.bm_offset);
3329 rv = ERR_MD_INVALID;
3332 if (be32_to_cpu(buffer->md_size_sect) != bdev->md.md_size_sect) {
3333 dev_err(DEV, "unexpected md_size: %u (expected %u)\n",
3334 be32_to_cpu(buffer->md_size_sect), bdev->md.md_size_sect);
3335 rv = ERR_MD_INVALID;
3339 if (be32_to_cpu(buffer->bm_bytes_per_bit) != BM_BLOCK_SIZE) {
3340 dev_err(DEV, "unexpected bm_bytes_per_bit: %u (expected %u)\n",
3341 be32_to_cpu(buffer->bm_bytes_per_bit), BM_BLOCK_SIZE);
3342 rv = ERR_MD_INVALID;
3346 bdev->md.la_size_sect = be64_to_cpu(buffer->la_size);
3347 for (i = UI_CURRENT; i < UI_SIZE; i++)
3348 bdev->md.uuid[i] = be64_to_cpu(buffer->uuid[i]);
3349 bdev->md.flags = be32_to_cpu(buffer->flags);
3350 mdev->sync_conf.al_extents = be32_to_cpu(buffer->al_nr_extents);
3351 bdev->md.device_uuid = be64_to_cpu(buffer->device_uuid);
3353 if (mdev->sync_conf.al_extents < 7)
3354 mdev->sync_conf.al_extents = 127;
3357 mutex_unlock(&mdev->md_io_mutex);
3364 * drbd_md_mark_dirty() - Mark meta data super block as dirty
3365 * @mdev: DRBD device.
3367 * Call this function if you change anything that should be written to
3368 * the meta-data super block. This function sets MD_DIRTY, and starts a
3369 * timer that ensures that within five seconds you have to call drbd_md_sync().
3371 void drbd_md_mark_dirty(struct drbd_conf *mdev)
3373 set_bit(MD_DIRTY, &mdev->flags);
3374 mod_timer(&mdev->md_sync_timer, jiffies + 5*HZ);
3378 static void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local)
3382 for (i = UI_HISTORY_START; i < UI_HISTORY_END; i++)
3383 mdev->ldev->md.uuid[i+1] = mdev->ldev->md.uuid[i];
3386 void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
3388 if (idx == UI_CURRENT) {
3389 if (mdev->state.role == R_PRIMARY)
3394 drbd_set_ed_uuid(mdev, val);
3397 mdev->ldev->md.uuid[idx] = val;
3398 drbd_md_mark_dirty(mdev);
3402 void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
3404 if (mdev->ldev->md.uuid[idx]) {
3405 drbd_uuid_move_history(mdev);
3406 mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[idx];
3408 _drbd_uuid_set(mdev, idx, val);
3412 * drbd_uuid_new_current() - Creates a new current UUID
3413 * @mdev: DRBD device.
3415 * Creates a new current UUID, and rotates the old current UUID into
3416 * the bitmap slot. Causes an incremental resync upon next connect.
3418 void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local)
3422 dev_info(DEV, "Creating new current UUID\n");
3423 D_ASSERT(mdev->ldev->md.uuid[UI_BITMAP] == 0);
3424 mdev->ldev->md.uuid[UI_BITMAP] = mdev->ldev->md.uuid[UI_CURRENT];
3426 get_random_bytes(&val, sizeof(u64));
3427 _drbd_uuid_set(mdev, UI_CURRENT, val);
3430 void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local)
3432 if (mdev->ldev->md.uuid[UI_BITMAP] == 0 && val == 0)
3436 drbd_uuid_move_history(mdev);
3437 mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP];
3438 mdev->ldev->md.uuid[UI_BITMAP] = 0;
3440 if (mdev->ldev->md.uuid[UI_BITMAP])
3441 dev_warn(DEV, "bm UUID already set");
3443 mdev->ldev->md.uuid[UI_BITMAP] = val;
3444 mdev->ldev->md.uuid[UI_BITMAP] &= ~((u64)1);
3447 drbd_md_mark_dirty(mdev);
3451 * drbd_bmio_set_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io()
3452 * @mdev: DRBD device.
3454 * Sets all bits in the bitmap and writes the whole bitmap to stable storage.
3456 int drbd_bmio_set_n_write(struct drbd_conf *mdev)
3460 if (get_ldev_if_state(mdev, D_ATTACHING)) {
3461 drbd_md_set_flag(mdev, MDF_FULL_SYNC);
3463 drbd_bm_set_all(mdev);
3465 rv = drbd_bm_write(mdev);
3468 drbd_md_clear_flag(mdev, MDF_FULL_SYNC);
3479 * drbd_bmio_clear_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io()
3480 * @mdev: DRBD device.
3482 * Clears all bits in the bitmap and writes the whole bitmap to stable storage.
3484 int drbd_bmio_clear_n_write(struct drbd_conf *mdev)
3488 if (get_ldev_if_state(mdev, D_ATTACHING)) {
3489 drbd_bm_clear_all(mdev);
3490 rv = drbd_bm_write(mdev);
3497 static int w_bitmap_io(struct drbd_conf *mdev, struct drbd_work *w, int unused)
3499 struct bm_io_work *work = container_of(w, struct bm_io_work, w);
3502 D_ASSERT(atomic_read(&mdev->ap_bio_cnt) == 0);
3504 drbd_bm_lock(mdev, work->why);
3505 rv = work->io_fn(mdev);
3506 drbd_bm_unlock(mdev);
3508 clear_bit(BITMAP_IO, &mdev->flags);
3509 wake_up(&mdev->misc_wait);
3512 work->done(mdev, rv);
3514 clear_bit(BITMAP_IO_QUEUED, &mdev->flags);
3521 * drbd_queue_bitmap_io() - Queues an IO operation on the whole bitmap
3522 * @mdev: DRBD device.
3523 * @io_fn: IO callback to be called when bitmap IO is possible
3524 * @done: callback to be called after the bitmap IO was performed
3525 * @why: Descriptive text of the reason for doing the IO
3527 * While IO on the bitmap happens we freeze application IO thus we ensure
3528 * that drbd_set_out_of_sync() can not be called. This function MAY ONLY be
3529 * called from worker context. It MUST NOT be used while a previous such
3530 * work is still pending!
3532 void drbd_queue_bitmap_io(struct drbd_conf *mdev,
3533 int (*io_fn)(struct drbd_conf *),
3534 void (*done)(struct drbd_conf *, int),
3537 D_ASSERT(current == mdev->worker.task);
3539 D_ASSERT(!test_bit(BITMAP_IO_QUEUED, &mdev->flags));
3540 D_ASSERT(!test_bit(BITMAP_IO, &mdev->flags));
3541 D_ASSERT(list_empty(&mdev->bm_io_work.w.list));
3542 if (mdev->bm_io_work.why)
3543 dev_err(DEV, "FIXME going to queue '%s' but '%s' still pending?\n",
3544 why, mdev->bm_io_work.why);
3546 mdev->bm_io_work.io_fn = io_fn;
3547 mdev->bm_io_work.done = done;
3548 mdev->bm_io_work.why = why;
3550 set_bit(BITMAP_IO, &mdev->flags);
3551 if (atomic_read(&mdev->ap_bio_cnt) == 0) {
3552 if (list_empty(&mdev->bm_io_work.w.list)) {
3553 set_bit(BITMAP_IO_QUEUED, &mdev->flags);
3554 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
3556 dev_err(DEV, "FIXME avoided double queuing bm_io_work\n");
3561 * drbd_bitmap_io() - Does an IO operation on the whole bitmap
3562 * @mdev: DRBD device.
3563 * @io_fn: IO callback to be called when bitmap IO is possible
3564 * @why: Descriptive text of the reason for doing the IO
3566 * freezes application IO while that the actual IO operations runs. This
3567 * functions MAY NOT be called from worker context.
3569 int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why)
3573 D_ASSERT(current != mdev->worker.task);
3575 drbd_suspend_io(mdev);
3577 drbd_bm_lock(mdev, why);
3579 drbd_bm_unlock(mdev);
3581 drbd_resume_io(mdev);
3586 void drbd_md_set_flag(struct drbd_conf *mdev, int flag) __must_hold(local)
3588 if ((mdev->ldev->md.flags & flag) != flag) {
3589 drbd_md_mark_dirty(mdev);
3590 mdev->ldev->md.flags |= flag;
3594 void drbd_md_clear_flag(struct drbd_conf *mdev, int flag) __must_hold(local)
3596 if ((mdev->ldev->md.flags & flag) != 0) {
3597 drbd_md_mark_dirty(mdev);
3598 mdev->ldev->md.flags &= ~flag;
3601 int drbd_md_test_flag(struct drbd_backing_dev *bdev, int flag)
3603 return (bdev->md.flags & flag) != 0;
3606 static void md_sync_timer_fn(unsigned long data)
3608 struct drbd_conf *mdev = (struct drbd_conf *) data;
3610 drbd_queue_work_front(&mdev->data.work, &mdev->md_sync_work);
3613 static int w_md_sync(struct drbd_conf *mdev, struct drbd_work *w, int unused)
3615 dev_warn(DEV, "md_sync_timer expired! Worker calls drbd_md_sync().\n");
3621 #ifdef CONFIG_DRBD_FAULT_INJECTION
3622 /* Fault insertion support including random number generator shamelessly
3623 * stolen from kernel/rcutorture.c */
3624 struct fault_random_state {
3625 unsigned long state;
3626 unsigned long count;
3629 #define FAULT_RANDOM_MULT 39916801 /* prime */
3630 #define FAULT_RANDOM_ADD 479001701 /* prime */
3631 #define FAULT_RANDOM_REFRESH 10000
3634 * Crude but fast random-number generator. Uses a linear congruential
3635 * generator, with occasional help from get_random_bytes().
3637 static unsigned long
3638 _drbd_fault_random(struct fault_random_state *rsp)
3642 if (!rsp->count--) {
3643 get_random_bytes(&refresh, sizeof(refresh));
3644 rsp->state += refresh;
3645 rsp->count = FAULT_RANDOM_REFRESH;
3647 rsp->state = rsp->state * FAULT_RANDOM_MULT + FAULT_RANDOM_ADD;
3648 return swahw32(rsp->state);
3652 _drbd_fault_str(unsigned int type) {
3653 static char *_faults[] = {
3654 [DRBD_FAULT_MD_WR] = "Meta-data write",
3655 [DRBD_FAULT_MD_RD] = "Meta-data read",
3656 [DRBD_FAULT_RS_WR] = "Resync write",
3657 [DRBD_FAULT_RS_RD] = "Resync read",
3658 [DRBD_FAULT_DT_WR] = "Data write",
3659 [DRBD_FAULT_DT_RD] = "Data read",
3660 [DRBD_FAULT_DT_RA] = "Data read ahead",
3661 [DRBD_FAULT_BM_ALLOC] = "BM allocation",
3662 [DRBD_FAULT_AL_EE] = "EE allocation"
3665 return (type < DRBD_FAULT_MAX) ? _faults[type] : "**Unknown**";
3669 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type)
3671 static struct fault_random_state rrs = {0, 0};
3673 unsigned int ret = (
3675 ((1 << mdev_to_minor(mdev)) & fault_devs) != 0) &&
3676 (((_drbd_fault_random(&rrs) % 100) + 1) <= fault_rate));
3681 if (printk_ratelimit())
3682 dev_warn(DEV, "***Simulating %s failure\n",
3683 _drbd_fault_str(type));
3690 const char *drbd_buildtag(void)
3692 /* DRBD built from external sources has here a reference to the
3693 git hash of the source code. */
3695 static char buildtag[38] = "\0uilt-in";
3697 if (buildtag[0] == 0) {
3698 #ifdef CONFIG_MODULES
3699 if (THIS_MODULE != NULL)
3700 sprintf(buildtag, "srcversion: %-24s", THIS_MODULE->srcversion);
3709 module_init(drbd_init)
3710 module_exit(drbd_cleanup)
3712 EXPORT_SYMBOL(drbd_conn_str);
3713 EXPORT_SYMBOL(drbd_role_str);
3714 EXPORT_SYMBOL(drbd_disk_str);
3715 EXPORT_SYMBOL(drbd_set_st_err_str);