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/mutex.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 static DEFINE_MUTEX(drbd_main_mutex);
60 int drbdd_init(struct drbd_thread *);
61 int drbd_worker(struct drbd_thread *);
62 int drbd_asender(struct drbd_thread *);
65 static int drbd_open(struct block_device *bdev, fmode_t mode);
66 static int drbd_release(struct gendisk *gd, fmode_t mode);
67 static int w_md_sync(struct drbd_work *w, int unused);
68 static void md_sync_timer_fn(unsigned long data);
69 static int w_bitmap_io(struct drbd_work *w, int unused);
70 static int w_go_diskless(struct drbd_work *w, int unused);
72 MODULE_AUTHOR("Philipp Reisner <phil@linbit.com>, "
73 "Lars Ellenberg <lars@linbit.com>");
74 MODULE_DESCRIPTION("drbd - Distributed Replicated Block Device v" REL_VERSION);
75 MODULE_VERSION(REL_VERSION);
76 MODULE_LICENSE("GPL");
77 MODULE_PARM_DESC(minor_count, "Approximate number of drbd devices ("
78 __stringify(DRBD_MINOR_COUNT_MIN) "-" __stringify(DRBD_MINOR_COUNT_MAX) ")");
79 MODULE_ALIAS_BLOCKDEV_MAJOR(DRBD_MAJOR);
81 #include <linux/moduleparam.h>
82 /* allow_open_on_secondary */
83 MODULE_PARM_DESC(allow_oos, "DONT USE!");
84 /* thanks to these macros, if compiled into the kernel (not-module),
85 * this becomes the boot parameter drbd.minor_count */
86 module_param(minor_count, uint, 0444);
87 module_param(disable_sendpage, bool, 0644);
88 module_param(allow_oos, bool, 0);
89 module_param(proc_details, int, 0644);
91 #ifdef CONFIG_DRBD_FAULT_INJECTION
94 static int fault_count;
96 /* bitmap of enabled faults */
97 module_param(enable_faults, int, 0664);
98 /* fault rate % value - applies to all enabled faults */
99 module_param(fault_rate, int, 0664);
100 /* count of faults inserted */
101 module_param(fault_count, int, 0664);
102 /* bitmap of devices to insert faults on */
103 module_param(fault_devs, int, 0644);
106 /* module parameter, defined */
107 unsigned int minor_count = DRBD_MINOR_COUNT_DEF;
108 int disable_sendpage;
110 int proc_details; /* Detail level in proc drbd*/
112 /* Module parameter for setting the user mode helper program
113 * to run. Default is /sbin/drbdadm */
114 char usermode_helper[80] = "/sbin/drbdadm";
116 module_param_string(usermode_helper, usermode_helper, sizeof(usermode_helper), 0644);
118 /* in 2.6.x, our device mapping and config info contains our virtual gendisks
119 * as member "struct gendisk *vdisk;"
122 struct list_head drbd_tconns; /* list of struct drbd_tconn */
124 struct kmem_cache *drbd_request_cache;
125 struct kmem_cache *drbd_ee_cache; /* peer requests */
126 struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
127 struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
128 mempool_t *drbd_request_mempool;
129 mempool_t *drbd_ee_mempool;
130 mempool_t *drbd_md_io_page_pool;
131 struct bio_set *drbd_md_io_bio_set;
133 /* I do not use a standard mempool, because:
134 1) I want to hand out the pre-allocated objects first.
135 2) I want to be able to interrupt sleeping allocation with a signal.
136 Note: This is a single linked list, the next pointer is the private
137 member of struct page.
139 struct page *drbd_pp_pool;
140 spinlock_t drbd_pp_lock;
142 wait_queue_head_t drbd_pp_wait;
144 DEFINE_RATELIMIT_STATE(drbd_ratelimit_state, 5 * HZ, 5);
146 static const struct block_device_operations drbd_ops = {
147 .owner = THIS_MODULE,
149 .release = drbd_release,
152 static void bio_destructor_drbd(struct bio *bio)
154 bio_free(bio, drbd_md_io_bio_set);
157 struct bio *bio_alloc_drbd(gfp_t gfp_mask)
161 if (!drbd_md_io_bio_set)
162 return bio_alloc(gfp_mask, 1);
164 bio = bio_alloc_bioset(gfp_mask, 1, drbd_md_io_bio_set);
167 bio->bi_destructor = bio_destructor_drbd;
172 /* When checking with sparse, and this is an inline function, sparse will
173 give tons of false positives. When this is a real functions sparse works.
175 int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
179 atomic_inc(&mdev->local_cnt);
180 io_allowed = (mdev->state.disk >= mins);
182 if (atomic_dec_and_test(&mdev->local_cnt))
183 wake_up(&mdev->misc_wait);
191 * DOC: The transfer log
193 * The transfer log is a single linked list of &struct drbd_tl_epoch objects.
194 * mdev->tconn->newest_tle points to the head, mdev->tconn->oldest_tle points to the tail
195 * of the list. There is always at least one &struct drbd_tl_epoch object.
197 * Each &struct drbd_tl_epoch has a circular double linked list of requests
200 static int tl_init(struct drbd_tconn *tconn)
202 struct drbd_tl_epoch *b;
204 /* during device minor initialization, we may well use GFP_KERNEL */
205 b = kmalloc(sizeof(struct drbd_tl_epoch), GFP_KERNEL);
208 INIT_LIST_HEAD(&b->requests);
209 INIT_LIST_HEAD(&b->w.list);
213 b->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */
215 tconn->oldest_tle = b;
216 tconn->newest_tle = b;
217 INIT_LIST_HEAD(&tconn->out_of_sequence_requests);
218 INIT_LIST_HEAD(&tconn->barrier_acked_requests);
223 static void tl_cleanup(struct drbd_tconn *tconn)
225 if (tconn->oldest_tle != tconn->newest_tle)
226 conn_err(tconn, "ASSERT FAILED: oldest_tle == newest_tle\n");
227 if (!list_empty(&tconn->out_of_sequence_requests))
228 conn_err(tconn, "ASSERT FAILED: list_empty(out_of_sequence_requests)\n");
229 kfree(tconn->oldest_tle);
230 tconn->oldest_tle = NULL;
231 kfree(tconn->unused_spare_tle);
232 tconn->unused_spare_tle = NULL;
236 * _tl_add_barrier() - Adds a barrier to the transfer log
237 * @mdev: DRBD device.
238 * @new: Barrier to be added before the current head of the TL.
240 * The caller must hold the req_lock.
242 void _tl_add_barrier(struct drbd_tconn *tconn, struct drbd_tl_epoch *new)
244 struct drbd_tl_epoch *newest_before;
246 INIT_LIST_HEAD(&new->requests);
247 INIT_LIST_HEAD(&new->w.list);
248 new->w.cb = NULL; /* if this is != NULL, we need to dec_ap_pending in tl_clear */
252 newest_before = tconn->newest_tle;
253 /* never send a barrier number == 0, because that is special-cased
254 * when using TCQ for our write ordering code */
255 new->br_number = (newest_before->br_number+1) ?: 1;
256 if (tconn->newest_tle != new) {
257 tconn->newest_tle->next = new;
258 tconn->newest_tle = new;
263 * tl_release() - Free or recycle the oldest &struct drbd_tl_epoch object of the TL
264 * @mdev: DRBD device.
265 * @barrier_nr: Expected identifier of the DRBD write barrier packet.
266 * @set_size: Expected number of requests before that barrier.
268 * In case the passed barrier_nr or set_size does not match the oldest
269 * &struct drbd_tl_epoch objects this function will cause a termination
272 void tl_release(struct drbd_tconn *tconn, unsigned int barrier_nr,
273 unsigned int set_size)
275 struct drbd_conf *mdev;
276 struct drbd_tl_epoch *b, *nob; /* next old barrier */
277 struct list_head *le, *tle;
278 struct drbd_request *r;
280 spin_lock_irq(&tconn->req_lock);
282 b = tconn->oldest_tle;
284 /* first some paranoia code */
286 conn_err(tconn, "BAD! BarrierAck #%u received, but no epoch in tl!?\n",
290 if (b->br_number != barrier_nr) {
291 conn_err(tconn, "BAD! BarrierAck #%u received, expected #%u!\n",
292 barrier_nr, b->br_number);
295 if (b->n_writes != set_size) {
296 conn_err(tconn, "BAD! BarrierAck #%u received with n_writes=%u, expected n_writes=%u!\n",
297 barrier_nr, set_size, b->n_writes);
301 /* Clean up list of requests processed during current epoch */
302 list_for_each_safe(le, tle, &b->requests) {
303 r = list_entry(le, struct drbd_request, tl_requests);
304 _req_mod(r, BARRIER_ACKED);
306 /* There could be requests on the list waiting for completion
307 of the write to the local disk. To avoid corruptions of
308 slab's data structures we have to remove the lists head.
310 Also there could have been a barrier ack out of sequence, overtaking
311 the write acks - which would be a bug and violating write ordering.
312 To not deadlock in case we lose connection while such requests are
313 still pending, we need some way to find them for the
314 _req_mode(CONNECTION_LOST_WHILE_PENDING).
316 These have been list_move'd to the out_of_sequence_requests list in
317 _req_mod(, BARRIER_ACKED) above.
319 list_splice_init(&b->requests, &tconn->barrier_acked_requests);
323 if (test_and_clear_bit(CREATE_BARRIER, &tconn->flags)) {
324 _tl_add_barrier(tconn, b);
326 tconn->oldest_tle = nob;
327 /* if nob == NULL b was the only barrier, and becomes the new
328 barrier. Therefore tconn->oldest_tle points already to b */
330 D_ASSERT(nob != NULL);
331 tconn->oldest_tle = nob;
335 spin_unlock_irq(&tconn->req_lock);
336 dec_ap_pending(mdev);
341 spin_unlock_irq(&tconn->req_lock);
342 conn_request_state(tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
347 * _tl_restart() - Walks the transfer log, and applies an action to all requests
348 * @mdev: DRBD device.
349 * @what: The action/event to perform with all request objects
351 * @what might be one of CONNECTION_LOST_WHILE_PENDING, RESEND, FAIL_FROZEN_DISK_IO,
352 * RESTART_FROZEN_DISK_IO.
354 void _tl_restart(struct drbd_tconn *tconn, enum drbd_req_event what)
356 struct drbd_tl_epoch *b, *tmp, **pn;
357 struct list_head *le, *tle, carry_reads;
358 struct drbd_request *req;
359 int rv, n_writes, n_reads;
361 b = tconn->oldest_tle;
362 pn = &tconn->oldest_tle;
366 INIT_LIST_HEAD(&carry_reads);
367 list_for_each_safe(le, tle, &b->requests) {
368 req = list_entry(le, struct drbd_request, tl_requests);
369 rv = _req_mod(req, what);
379 if (what == RESEND) {
380 b->n_writes = n_writes;
381 if (b->w.cb == NULL) {
382 b->w.cb = w_send_barrier;
383 inc_ap_pending(b->w.mdev);
384 set_bit(CREATE_BARRIER, &tconn->flags);
387 drbd_queue_work(&tconn->data.work, &b->w);
392 list_add(&carry_reads, &b->requests);
393 /* there could still be requests on that ring list,
394 * in case local io is still pending */
395 list_del(&b->requests);
397 /* dec_ap_pending corresponding to queue_barrier.
398 * the newest barrier may not have been queued yet,
399 * in which case w.cb is still NULL. */
401 dec_ap_pending(b->w.mdev);
403 if (b == tconn->newest_tle) {
404 /* recycle, but reinit! */
406 conn_err(tconn, "ASSERT FAILED tmp == NULL");
407 INIT_LIST_HEAD(&b->requests);
408 list_splice(&carry_reads, &b->requests);
409 INIT_LIST_HEAD(&b->w.list);
411 b->br_number = net_random();
421 list_splice(&carry_reads, &b->requests);
424 /* Actions operating on the disk state, also want to work on
425 requests that got barrier acked. */
427 case FAIL_FROZEN_DISK_IO:
428 case RESTART_FROZEN_DISK_IO:
429 list_for_each_safe(le, tle, &tconn->barrier_acked_requests) {
430 req = list_entry(le, struct drbd_request, tl_requests);
433 case CONNECTION_LOST_WHILE_PENDING:
437 conn_err(tconn, "what = %d in _tl_restart()\n", what);
442 * tl_clear() - Clears all requests and &struct drbd_tl_epoch objects out of the TL
443 * @mdev: DRBD device.
445 * This is called after the connection to the peer was lost. The storage covered
446 * by the requests on the transfer gets marked as our of sync. Called from the
447 * receiver thread and the worker thread.
449 void tl_clear(struct drbd_tconn *tconn)
451 struct list_head *le, *tle;
452 struct drbd_request *r;
454 spin_lock_irq(&tconn->req_lock);
456 _tl_restart(tconn, CONNECTION_LOST_WHILE_PENDING);
458 /* we expect this list to be empty. */
459 if (!list_empty(&tconn->out_of_sequence_requests))
460 conn_err(tconn, "ASSERT FAILED list_empty(&out_of_sequence_requests)\n");
462 /* but just in case, clean it up anyways! */
463 list_for_each_safe(le, tle, &tconn->out_of_sequence_requests) {
464 r = list_entry(le, struct drbd_request, tl_requests);
465 /* It would be nice to complete outside of spinlock.
466 * But this is easier for now. */
467 _req_mod(r, CONNECTION_LOST_WHILE_PENDING);
470 /* ensure bit indicating barrier is required is clear */
471 clear_bit(CREATE_BARRIER, &tconn->flags);
473 spin_unlock_irq(&tconn->req_lock);
476 void tl_restart(struct drbd_tconn *tconn, enum drbd_req_event what)
478 spin_lock_irq(&tconn->req_lock);
479 _tl_restart(tconn, what);
480 spin_unlock_irq(&tconn->req_lock);
484 * tl_abort_disk_io() - Abort disk I/O for all requests for a certain mdev in the TL
485 * @mdev: DRBD device.
487 void tl_abort_disk_io(struct drbd_conf *mdev)
489 struct drbd_tconn *tconn = mdev->tconn;
490 struct drbd_tl_epoch *b;
491 struct list_head *le, *tle;
492 struct drbd_request *req;
494 spin_lock_irq(&tconn->req_lock);
495 b = tconn->oldest_tle;
497 list_for_each_safe(le, tle, &b->requests) {
498 req = list_entry(le, struct drbd_request, tl_requests);
499 if (!(req->rq_state & RQ_LOCAL_PENDING))
501 if (req->w.mdev == mdev)
502 _req_mod(req, ABORT_DISK_IO);
507 list_for_each_safe(le, tle, &tconn->barrier_acked_requests) {
508 req = list_entry(le, struct drbd_request, tl_requests);
509 if (!(req->rq_state & RQ_LOCAL_PENDING))
511 if (req->w.mdev == mdev)
512 _req_mod(req, ABORT_DISK_IO);
515 spin_unlock_irq(&tconn->req_lock);
518 static int drbd_thread_setup(void *arg)
520 struct drbd_thread *thi = (struct drbd_thread *) arg;
521 struct drbd_tconn *tconn = thi->tconn;
525 snprintf(current->comm, sizeof(current->comm), "drbd_%c_%s",
526 thi->name[0], thi->tconn->name);
529 retval = thi->function(thi);
531 spin_lock_irqsave(&thi->t_lock, flags);
533 /* if the receiver has been "EXITING", the last thing it did
534 * was set the conn state to "StandAlone",
535 * if now a re-connect request comes in, conn state goes C_UNCONNECTED,
536 * and receiver thread will be "started".
537 * drbd_thread_start needs to set "RESTARTING" in that case.
538 * t_state check and assignment needs to be within the same spinlock,
539 * so either thread_start sees EXITING, and can remap to RESTARTING,
540 * or thread_start see NONE, and can proceed as normal.
543 if (thi->t_state == RESTARTING) {
544 conn_info(tconn, "Restarting %s thread\n", thi->name);
545 thi->t_state = RUNNING;
546 spin_unlock_irqrestore(&thi->t_lock, flags);
553 complete_all(&thi->stop);
554 spin_unlock_irqrestore(&thi->t_lock, flags);
556 conn_info(tconn, "Terminating %s\n", current->comm);
558 /* Release mod reference taken when thread was started */
560 kref_put(&tconn->kref, &conn_destroy);
561 module_put(THIS_MODULE);
565 static void drbd_thread_init(struct drbd_tconn *tconn, struct drbd_thread *thi,
566 int (*func) (struct drbd_thread *), char *name)
568 spin_lock_init(&thi->t_lock);
571 thi->function = func;
573 strncpy(thi->name, name, ARRAY_SIZE(thi->name));
576 int drbd_thread_start(struct drbd_thread *thi)
578 struct drbd_tconn *tconn = thi->tconn;
579 struct task_struct *nt;
582 /* is used from state engine doing drbd_thread_stop_nowait,
583 * while holding the req lock irqsave */
584 spin_lock_irqsave(&thi->t_lock, flags);
586 switch (thi->t_state) {
588 conn_info(tconn, "Starting %s thread (from %s [%d])\n",
589 thi->name, current->comm, current->pid);
591 /* Get ref on module for thread - this is released when thread exits */
592 if (!try_module_get(THIS_MODULE)) {
593 conn_err(tconn, "Failed to get module reference in drbd_thread_start\n");
594 spin_unlock_irqrestore(&thi->t_lock, flags);
598 kref_get(&thi->tconn->kref);
600 init_completion(&thi->stop);
601 thi->reset_cpu_mask = 1;
602 thi->t_state = RUNNING;
603 spin_unlock_irqrestore(&thi->t_lock, flags);
604 flush_signals(current); /* otherw. may get -ERESTARTNOINTR */
606 nt = kthread_create(drbd_thread_setup, (void *) thi,
607 "drbd_%c_%s", thi->name[0], thi->tconn->name);
610 conn_err(tconn, "Couldn't start thread\n");
612 kref_put(&tconn->kref, &conn_destroy);
613 module_put(THIS_MODULE);
616 spin_lock_irqsave(&thi->t_lock, flags);
618 thi->t_state = RUNNING;
619 spin_unlock_irqrestore(&thi->t_lock, flags);
623 thi->t_state = RESTARTING;
624 conn_info(tconn, "Restarting %s thread (from %s [%d])\n",
625 thi->name, current->comm, current->pid);
630 spin_unlock_irqrestore(&thi->t_lock, flags);
638 void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait)
642 enum drbd_thread_state ns = restart ? RESTARTING : EXITING;
644 /* may be called from state engine, holding the req lock irqsave */
645 spin_lock_irqsave(&thi->t_lock, flags);
647 if (thi->t_state == NONE) {
648 spin_unlock_irqrestore(&thi->t_lock, flags);
650 drbd_thread_start(thi);
654 if (thi->t_state != ns) {
655 if (thi->task == NULL) {
656 spin_unlock_irqrestore(&thi->t_lock, flags);
662 init_completion(&thi->stop);
663 if (thi->task != current)
664 force_sig(DRBD_SIGKILL, thi->task);
667 spin_unlock_irqrestore(&thi->t_lock, flags);
670 wait_for_completion(&thi->stop);
673 static struct drbd_thread *drbd_task_to_thread(struct drbd_tconn *tconn, struct task_struct *task)
675 struct drbd_thread *thi =
676 task == tconn->receiver.task ? &tconn->receiver :
677 task == tconn->asender.task ? &tconn->asender :
678 task == tconn->worker.task ? &tconn->worker : NULL;
683 char *drbd_task_to_thread_name(struct drbd_tconn *tconn, struct task_struct *task)
685 struct drbd_thread *thi = drbd_task_to_thread(tconn, task);
686 return thi ? thi->name : task->comm;
689 int conn_lowest_minor(struct drbd_tconn *tconn)
691 struct drbd_conf *mdev;
695 mdev = idr_get_next(&tconn->volumes, &vnr);
696 m = mdev ? mdev_to_minor(mdev) : -1;
704 * drbd_calc_cpu_mask() - Generate CPU masks, spread over all CPUs
705 * @mdev: DRBD device.
707 * Forces all threads of a device onto the same CPU. This is beneficial for
708 * DRBD's performance. May be overwritten by user's configuration.
710 void drbd_calc_cpu_mask(struct drbd_tconn *tconn)
715 if (cpumask_weight(tconn->cpu_mask))
718 ord = conn_lowest_minor(tconn) % cpumask_weight(cpu_online_mask);
719 for_each_online_cpu(cpu) {
721 cpumask_set_cpu(cpu, tconn->cpu_mask);
725 /* should not be reached */
726 cpumask_setall(tconn->cpu_mask);
730 * drbd_thread_current_set_cpu() - modifies the cpu mask of the _current_ thread
731 * @mdev: DRBD device.
732 * @thi: drbd_thread object
734 * call in the "main loop" of _all_ threads, no need for any mutex, current won't die
737 void drbd_thread_current_set_cpu(struct drbd_thread *thi)
739 struct task_struct *p = current;
741 if (!thi->reset_cpu_mask)
743 thi->reset_cpu_mask = 0;
744 set_cpus_allowed_ptr(p, thi->tconn->cpu_mask);
749 * drbd_header_size - size of a packet header
751 * The header size is a multiple of 8, so any payload following the header is
752 * word aligned on 64-bit architectures. (The bitmap send and receive code
755 unsigned int drbd_header_size(struct drbd_tconn *tconn)
757 if (tconn->agreed_pro_version >= 100) {
758 BUILD_BUG_ON(!IS_ALIGNED(sizeof(struct p_header100), 8));
759 return sizeof(struct p_header100);
761 BUILD_BUG_ON(sizeof(struct p_header80) !=
762 sizeof(struct p_header95));
763 BUILD_BUG_ON(!IS_ALIGNED(sizeof(struct p_header80), 8));
764 return sizeof(struct p_header80);
768 static unsigned int prepare_header80(struct p_header80 *h, enum drbd_packet cmd, int size)
770 h->magic = cpu_to_be32(DRBD_MAGIC);
771 h->command = cpu_to_be16(cmd);
772 h->length = cpu_to_be16(size);
773 return sizeof(struct p_header80);
776 static unsigned int prepare_header95(struct p_header95 *h, enum drbd_packet cmd, int size)
778 h->magic = cpu_to_be16(DRBD_MAGIC_BIG);
779 h->command = cpu_to_be16(cmd);
780 h->length = cpu_to_be32(size);
781 return sizeof(struct p_header95);
784 static unsigned int prepare_header100(struct p_header100 *h, enum drbd_packet cmd,
787 h->magic = cpu_to_be32(DRBD_MAGIC_100);
788 h->volume = cpu_to_be16(vnr);
789 h->command = cpu_to_be16(cmd);
790 h->length = cpu_to_be32(size);
792 return sizeof(struct p_header100);
795 static unsigned int prepare_header(struct drbd_tconn *tconn, int vnr,
796 void *buffer, enum drbd_packet cmd, int size)
798 if (tconn->agreed_pro_version >= 100)
799 return prepare_header100(buffer, cmd, size, vnr);
800 else if (tconn->agreed_pro_version >= 95 &&
801 size > DRBD_MAX_SIZE_H80_PACKET)
802 return prepare_header95(buffer, cmd, size);
804 return prepare_header80(buffer, cmd, size);
807 static void *__conn_prepare_command(struct drbd_tconn *tconn,
808 struct drbd_socket *sock)
812 return sock->sbuf + drbd_header_size(tconn);
815 void *conn_prepare_command(struct drbd_tconn *tconn, struct drbd_socket *sock)
819 mutex_lock(&sock->mutex);
820 p = __conn_prepare_command(tconn, sock);
822 mutex_unlock(&sock->mutex);
827 void *drbd_prepare_command(struct drbd_conf *mdev, struct drbd_socket *sock)
829 return conn_prepare_command(mdev->tconn, sock);
832 static int __send_command(struct drbd_tconn *tconn, int vnr,
833 struct drbd_socket *sock, enum drbd_packet cmd,
834 unsigned int header_size, void *data,
841 * Called with @data == NULL and the size of the data blocks in @size
842 * for commands that send data blocks. For those commands, omit the
843 * MSG_MORE flag: this will increase the likelihood that data blocks
844 * which are page aligned on the sender will end up page aligned on the
847 msg_flags = data ? MSG_MORE : 0;
849 header_size += prepare_header(tconn, vnr, sock->sbuf, cmd,
851 err = drbd_send_all(tconn, sock->socket, sock->sbuf, header_size,
854 err = drbd_send_all(tconn, sock->socket, data, size, 0);
858 static int __conn_send_command(struct drbd_tconn *tconn, struct drbd_socket *sock,
859 enum drbd_packet cmd, unsigned int header_size,
860 void *data, unsigned int size)
862 return __send_command(tconn, 0, sock, cmd, header_size, data, size);
865 int conn_send_command(struct drbd_tconn *tconn, struct drbd_socket *sock,
866 enum drbd_packet cmd, unsigned int header_size,
867 void *data, unsigned int size)
871 err = __conn_send_command(tconn, sock, cmd, header_size, data, size);
872 mutex_unlock(&sock->mutex);
876 int drbd_send_command(struct drbd_conf *mdev, struct drbd_socket *sock,
877 enum drbd_packet cmd, unsigned int header_size,
878 void *data, unsigned int size)
882 err = __send_command(mdev->tconn, mdev->vnr, sock, cmd, header_size,
884 mutex_unlock(&sock->mutex);
888 int drbd_send_ping(struct drbd_tconn *tconn)
890 struct drbd_socket *sock;
893 if (!conn_prepare_command(tconn, sock))
895 return conn_send_command(tconn, sock, P_PING, 0, NULL, 0);
898 int drbd_send_ping_ack(struct drbd_tconn *tconn)
900 struct drbd_socket *sock;
903 if (!conn_prepare_command(tconn, sock))
905 return conn_send_command(tconn, sock, P_PING_ACK, 0, NULL, 0);
908 int drbd_send_sync_param(struct drbd_conf *mdev)
910 struct drbd_socket *sock;
911 struct p_rs_param_95 *p;
913 const int apv = mdev->tconn->agreed_pro_version;
914 enum drbd_packet cmd;
916 struct disk_conf *dc;
918 sock = &mdev->tconn->data;
919 p = drbd_prepare_command(mdev, sock);
924 nc = rcu_dereference(mdev->tconn->net_conf);
926 size = apv <= 87 ? sizeof(struct p_rs_param)
927 : apv == 88 ? sizeof(struct p_rs_param)
928 + strlen(nc->verify_alg) + 1
929 : apv <= 94 ? sizeof(struct p_rs_param_89)
930 : /* apv >= 95 */ sizeof(struct p_rs_param_95);
932 cmd = apv >= 89 ? P_SYNC_PARAM89 : P_SYNC_PARAM;
934 /* initialize verify_alg and csums_alg */
935 memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX);
937 if (get_ldev(mdev)) {
938 dc = rcu_dereference(mdev->ldev->disk_conf);
939 p->resync_rate = cpu_to_be32(dc->resync_rate);
940 p->c_plan_ahead = cpu_to_be32(dc->c_plan_ahead);
941 p->c_delay_target = cpu_to_be32(dc->c_delay_target);
942 p->c_fill_target = cpu_to_be32(dc->c_fill_target);
943 p->c_max_rate = cpu_to_be32(dc->c_max_rate);
946 p->resync_rate = cpu_to_be32(DRBD_RESYNC_RATE_DEF);
947 p->c_plan_ahead = cpu_to_be32(DRBD_C_PLAN_AHEAD_DEF);
948 p->c_delay_target = cpu_to_be32(DRBD_C_DELAY_TARGET_DEF);
949 p->c_fill_target = cpu_to_be32(DRBD_C_FILL_TARGET_DEF);
950 p->c_max_rate = cpu_to_be32(DRBD_C_MAX_RATE_DEF);
954 strcpy(p->verify_alg, nc->verify_alg);
956 strcpy(p->csums_alg, nc->csums_alg);
959 return drbd_send_command(mdev, sock, cmd, size, NULL, 0);
962 int __drbd_send_protocol(struct drbd_tconn *tconn, enum drbd_packet cmd)
964 struct drbd_socket *sock;
965 struct p_protocol *p;
970 p = __conn_prepare_command(tconn, sock);
975 nc = rcu_dereference(tconn->net_conf);
977 if (nc->tentative && tconn->agreed_pro_version < 92) {
979 mutex_unlock(&sock->mutex);
980 conn_err(tconn, "--dry-run is not supported by peer");
985 if (tconn->agreed_pro_version >= 87)
986 size += strlen(nc->integrity_alg) + 1;
988 p->protocol = cpu_to_be32(nc->wire_protocol);
989 p->after_sb_0p = cpu_to_be32(nc->after_sb_0p);
990 p->after_sb_1p = cpu_to_be32(nc->after_sb_1p);
991 p->after_sb_2p = cpu_to_be32(nc->after_sb_2p);
992 p->two_primaries = cpu_to_be32(nc->two_primaries);
994 if (nc->discard_my_data)
995 cf |= CF_DISCARD_MY_DATA;
998 p->conn_flags = cpu_to_be32(cf);
1000 if (tconn->agreed_pro_version >= 87)
1001 strcpy(p->integrity_alg, nc->integrity_alg);
1004 return __conn_send_command(tconn, sock, cmd, size, NULL, 0);
1007 int drbd_send_protocol(struct drbd_tconn *tconn)
1011 mutex_lock(&tconn->data.mutex);
1012 err = __drbd_send_protocol(tconn, P_PROTOCOL);
1013 mutex_unlock(&tconn->data.mutex);
1018 int _drbd_send_uuids(struct drbd_conf *mdev, u64 uuid_flags)
1020 struct drbd_socket *sock;
1024 if (!get_ldev_if_state(mdev, D_NEGOTIATING))
1027 sock = &mdev->tconn->data;
1028 p = drbd_prepare_command(mdev, sock);
1033 for (i = UI_CURRENT; i < UI_SIZE; i++)
1034 p->uuid[i] = mdev->ldev ? cpu_to_be64(mdev->ldev->md.uuid[i]) : 0;
1036 mdev->comm_bm_set = drbd_bm_total_weight(mdev);
1037 p->uuid[UI_SIZE] = cpu_to_be64(mdev->comm_bm_set);
1039 uuid_flags |= rcu_dereference(mdev->tconn->net_conf)->discard_my_data ? 1 : 0;
1041 uuid_flags |= test_bit(CRASHED_PRIMARY, &mdev->flags) ? 2 : 0;
1042 uuid_flags |= mdev->new_state_tmp.disk == D_INCONSISTENT ? 4 : 0;
1043 p->uuid[UI_FLAGS] = cpu_to_be64(uuid_flags);
1046 return drbd_send_command(mdev, sock, P_UUIDS, sizeof(*p), NULL, 0);
1049 int drbd_send_uuids(struct drbd_conf *mdev)
1051 return _drbd_send_uuids(mdev, 0);
1054 int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev)
1056 return _drbd_send_uuids(mdev, 8);
1059 void drbd_print_uuids(struct drbd_conf *mdev, const char *text)
1061 if (get_ldev_if_state(mdev, D_NEGOTIATING)) {
1062 u64 *uuid = mdev->ldev->md.uuid;
1063 dev_info(DEV, "%s %016llX:%016llX:%016llX:%016llX\n",
1065 (unsigned long long)uuid[UI_CURRENT],
1066 (unsigned long long)uuid[UI_BITMAP],
1067 (unsigned long long)uuid[UI_HISTORY_START],
1068 (unsigned long long)uuid[UI_HISTORY_END]);
1071 dev_info(DEV, "%s effective data uuid: %016llX\n",
1073 (unsigned long long)mdev->ed_uuid);
1077 void drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev)
1079 struct drbd_socket *sock;
1080 struct p_rs_uuid *p;
1083 D_ASSERT(mdev->state.disk == D_UP_TO_DATE);
1085 uuid = mdev->ldev->md.uuid[UI_BITMAP];
1086 if (uuid && uuid != UUID_JUST_CREATED)
1087 uuid = uuid + UUID_NEW_BM_OFFSET;
1089 get_random_bytes(&uuid, sizeof(u64));
1090 drbd_uuid_set(mdev, UI_BITMAP, uuid);
1091 drbd_print_uuids(mdev, "updated sync UUID");
1094 sock = &mdev->tconn->data;
1095 p = drbd_prepare_command(mdev, sock);
1097 p->uuid = cpu_to_be64(uuid);
1098 drbd_send_command(mdev, sock, P_SYNC_UUID, sizeof(*p), NULL, 0);
1102 int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags)
1104 struct drbd_socket *sock;
1106 sector_t d_size, u_size;
1107 int q_order_type, max_bio_size;
1109 if (get_ldev_if_state(mdev, D_NEGOTIATING)) {
1110 D_ASSERT(mdev->ldev->backing_bdev);
1111 d_size = drbd_get_max_capacity(mdev->ldev);
1113 u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size;
1115 q_order_type = drbd_queue_order_type(mdev);
1116 max_bio_size = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9;
1117 max_bio_size = min_t(int, max_bio_size, DRBD_MAX_BIO_SIZE);
1122 q_order_type = QUEUE_ORDERED_NONE;
1123 max_bio_size = DRBD_MAX_BIO_SIZE; /* ... multiple BIOs per peer_request */
1126 sock = &mdev->tconn->data;
1127 p = drbd_prepare_command(mdev, sock);
1131 if (mdev->tconn->agreed_pro_version <= 94)
1132 max_bio_size = min_t(int, max_bio_size, DRBD_MAX_SIZE_H80_PACKET);
1133 else if (mdev->tconn->agreed_pro_version < 100)
1134 max_bio_size = min_t(int, max_bio_size, DRBD_MAX_BIO_SIZE_P95);
1136 p->d_size = cpu_to_be64(d_size);
1137 p->u_size = cpu_to_be64(u_size);
1138 p->c_size = cpu_to_be64(trigger_reply ? 0 : drbd_get_capacity(mdev->this_bdev));
1139 p->max_bio_size = cpu_to_be32(max_bio_size);
1140 p->queue_order_type = cpu_to_be16(q_order_type);
1141 p->dds_flags = cpu_to_be16(flags);
1142 return drbd_send_command(mdev, sock, P_SIZES, sizeof(*p), NULL, 0);
1146 * drbd_send_current_state() - Sends the drbd state to the peer
1147 * @mdev: DRBD device.
1149 int drbd_send_current_state(struct drbd_conf *mdev)
1151 struct drbd_socket *sock;
1154 sock = &mdev->tconn->data;
1155 p = drbd_prepare_command(mdev, sock);
1158 p->state = cpu_to_be32(mdev->state.i); /* Within the send mutex */
1159 return drbd_send_command(mdev, sock, P_STATE, sizeof(*p), NULL, 0);
1163 * drbd_send_state() - After a state change, sends the new state to the peer
1164 * @mdev: DRBD device.
1165 * @state: the state to send, not necessarily the current state.
1167 * Each state change queues an "after_state_ch" work, which will eventually
1168 * send the resulting new state to the peer. If more state changes happen
1169 * between queuing and processing of the after_state_ch work, we still
1170 * want to send each intermediary state in the order it occurred.
1172 int drbd_send_state(struct drbd_conf *mdev, union drbd_state state)
1174 struct drbd_socket *sock;
1177 sock = &mdev->tconn->data;
1178 p = drbd_prepare_command(mdev, sock);
1181 p->state = cpu_to_be32(state.i); /* Within the send mutex */
1182 return drbd_send_command(mdev, sock, P_STATE, sizeof(*p), NULL, 0);
1185 int drbd_send_state_req(struct drbd_conf *mdev, union drbd_state mask, union drbd_state val)
1187 struct drbd_socket *sock;
1188 struct p_req_state *p;
1190 sock = &mdev->tconn->data;
1191 p = drbd_prepare_command(mdev, sock);
1194 p->mask = cpu_to_be32(mask.i);
1195 p->val = cpu_to_be32(val.i);
1196 return drbd_send_command(mdev, sock, P_STATE_CHG_REQ, sizeof(*p), NULL, 0);
1199 int conn_send_state_req(struct drbd_tconn *tconn, union drbd_state mask, union drbd_state val)
1201 enum drbd_packet cmd;
1202 struct drbd_socket *sock;
1203 struct p_req_state *p;
1205 cmd = tconn->agreed_pro_version < 100 ? P_STATE_CHG_REQ : P_CONN_ST_CHG_REQ;
1206 sock = &tconn->data;
1207 p = conn_prepare_command(tconn, sock);
1210 p->mask = cpu_to_be32(mask.i);
1211 p->val = cpu_to_be32(val.i);
1212 return conn_send_command(tconn, sock, cmd, sizeof(*p), NULL, 0);
1215 void drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode)
1217 struct drbd_socket *sock;
1218 struct p_req_state_reply *p;
1220 sock = &mdev->tconn->meta;
1221 p = drbd_prepare_command(mdev, sock);
1223 p->retcode = cpu_to_be32(retcode);
1224 drbd_send_command(mdev, sock, P_STATE_CHG_REPLY, sizeof(*p), NULL, 0);
1228 void conn_send_sr_reply(struct drbd_tconn *tconn, enum drbd_state_rv retcode)
1230 struct drbd_socket *sock;
1231 struct p_req_state_reply *p;
1232 enum drbd_packet cmd = tconn->agreed_pro_version < 100 ? P_STATE_CHG_REPLY : P_CONN_ST_CHG_REPLY;
1234 sock = &tconn->meta;
1235 p = conn_prepare_command(tconn, sock);
1237 p->retcode = cpu_to_be32(retcode);
1238 conn_send_command(tconn, sock, cmd, sizeof(*p), NULL, 0);
1242 static void dcbp_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
1244 BUG_ON(code & ~0xf);
1245 p->encoding = (p->encoding & ~0xf) | code;
1248 static void dcbp_set_start(struct p_compressed_bm *p, int set)
1250 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
1253 static void dcbp_set_pad_bits(struct p_compressed_bm *p, int n)
1256 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
1259 int fill_bitmap_rle_bits(struct drbd_conf *mdev,
1260 struct p_compressed_bm *p,
1262 struct bm_xfer_ctx *c)
1264 struct bitstream bs;
1265 unsigned long plain_bits;
1272 /* may we use this feature? */
1274 use_rle = rcu_dereference(mdev->tconn->net_conf)->use_rle;
1276 if (!use_rle || mdev->tconn->agreed_pro_version < 90)
1279 if (c->bit_offset >= c->bm_bits)
1280 return 0; /* nothing to do. */
1282 /* use at most thus many bytes */
1283 bitstream_init(&bs, p->code, size, 0);
1284 memset(p->code, 0, size);
1285 /* plain bits covered in this code string */
1288 /* p->encoding & 0x80 stores whether the first run length is set.
1289 * bit offset is implicit.
1290 * start with toggle == 2 to be able to tell the first iteration */
1293 /* see how much plain bits we can stuff into one packet
1294 * using RLE and VLI. */
1296 tmp = (toggle == 0) ? _drbd_bm_find_next_zero(mdev, c->bit_offset)
1297 : _drbd_bm_find_next(mdev, c->bit_offset);
1300 rl = tmp - c->bit_offset;
1302 if (toggle == 2) { /* first iteration */
1304 /* the first checked bit was set,
1305 * store start value, */
1306 dcbp_set_start(p, 1);
1307 /* but skip encoding of zero run length */
1311 dcbp_set_start(p, 0);
1314 /* paranoia: catch zero runlength.
1315 * can only happen if bitmap is modified while we scan it. */
1317 dev_err(DEV, "unexpected zero runlength while encoding bitmap "
1318 "t:%u bo:%lu\n", toggle, c->bit_offset);
1322 bits = vli_encode_bits(&bs, rl);
1323 if (bits == -ENOBUFS) /* buffer full */
1326 dev_err(DEV, "error while encoding bitmap: %d\n", bits);
1332 c->bit_offset = tmp;
1333 } while (c->bit_offset < c->bm_bits);
1335 len = bs.cur.b - p->code + !!bs.cur.bit;
1337 if (plain_bits < (len << 3)) {
1338 /* incompressible with this method.
1339 * we need to rewind both word and bit position. */
1340 c->bit_offset -= plain_bits;
1341 bm_xfer_ctx_bit_to_word_offset(c);
1342 c->bit_offset = c->word_offset * BITS_PER_LONG;
1346 /* RLE + VLI was able to compress it just fine.
1347 * update c->word_offset. */
1348 bm_xfer_ctx_bit_to_word_offset(c);
1350 /* store pad_bits */
1351 dcbp_set_pad_bits(p, (8 - bs.cur.bit) & 0x7);
1357 * send_bitmap_rle_or_plain
1359 * Return 0 when done, 1 when another iteration is needed, and a negative error
1360 * code upon failure.
1363 send_bitmap_rle_or_plain(struct drbd_conf *mdev, struct bm_xfer_ctx *c)
1365 struct drbd_socket *sock = &mdev->tconn->data;
1366 unsigned int header_size = drbd_header_size(mdev->tconn);
1367 struct p_compressed_bm *p = sock->sbuf + header_size;
1370 len = fill_bitmap_rle_bits(mdev, p,
1371 DRBD_SOCKET_BUFFER_SIZE - header_size - sizeof(*p), c);
1376 dcbp_set_code(p, RLE_VLI_Bits);
1377 err = __send_command(mdev->tconn, mdev->vnr, sock,
1378 P_COMPRESSED_BITMAP, sizeof(*p) + len,
1381 c->bytes[0] += header_size + sizeof(*p) + len;
1383 if (c->bit_offset >= c->bm_bits)
1386 /* was not compressible.
1387 * send a buffer full of plain text bits instead. */
1388 unsigned int data_size;
1389 unsigned long num_words;
1390 unsigned long *p = sock->sbuf + header_size;
1392 data_size = DRBD_SOCKET_BUFFER_SIZE - header_size;
1393 num_words = min_t(size_t, data_size / sizeof(*p),
1394 c->bm_words - c->word_offset);
1395 len = num_words * sizeof(*p);
1397 drbd_bm_get_lel(mdev, c->word_offset, num_words, p);
1398 err = __send_command(mdev->tconn, mdev->vnr, sock, P_BITMAP, len, NULL, 0);
1399 c->word_offset += num_words;
1400 c->bit_offset = c->word_offset * BITS_PER_LONG;
1403 c->bytes[1] += header_size + len;
1405 if (c->bit_offset > c->bm_bits)
1406 c->bit_offset = c->bm_bits;
1410 INFO_bm_xfer_stats(mdev, "send", c);
1418 /* See the comment at receive_bitmap() */
1419 static int _drbd_send_bitmap(struct drbd_conf *mdev)
1421 struct bm_xfer_ctx c;
1424 if (!expect(mdev->bitmap))
1427 if (get_ldev(mdev)) {
1428 if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) {
1429 dev_info(DEV, "Writing the whole bitmap, MDF_FullSync was set.\n");
1430 drbd_bm_set_all(mdev);
1431 if (drbd_bm_write(mdev)) {
1432 /* write_bm did fail! Leave full sync flag set in Meta P_DATA
1433 * but otherwise process as per normal - need to tell other
1434 * side that a full resync is required! */
1435 dev_err(DEV, "Failed to write bitmap to disk!\n");
1437 drbd_md_clear_flag(mdev, MDF_FULL_SYNC);
1444 c = (struct bm_xfer_ctx) {
1445 .bm_bits = drbd_bm_bits(mdev),
1446 .bm_words = drbd_bm_words(mdev),
1450 err = send_bitmap_rle_or_plain(mdev, &c);
1456 int drbd_send_bitmap(struct drbd_conf *mdev)
1458 struct drbd_socket *sock = &mdev->tconn->data;
1461 mutex_lock(&sock->mutex);
1463 err = !_drbd_send_bitmap(mdev);
1464 mutex_unlock(&sock->mutex);
1468 void drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, u32 set_size)
1470 struct drbd_socket *sock;
1471 struct p_barrier_ack *p;
1473 if (mdev->state.conn < C_CONNECTED)
1476 sock = &mdev->tconn->meta;
1477 p = drbd_prepare_command(mdev, sock);
1480 p->barrier = barrier_nr;
1481 p->set_size = cpu_to_be32(set_size);
1482 drbd_send_command(mdev, sock, P_BARRIER_ACK, sizeof(*p), NULL, 0);
1486 * _drbd_send_ack() - Sends an ack packet
1487 * @mdev: DRBD device.
1488 * @cmd: Packet command code.
1489 * @sector: sector, needs to be in big endian byte order
1490 * @blksize: size in byte, needs to be in big endian byte order
1491 * @block_id: Id, big endian byte order
1493 static int _drbd_send_ack(struct drbd_conf *mdev, enum drbd_packet cmd,
1494 u64 sector, u32 blksize, u64 block_id)
1496 struct drbd_socket *sock;
1497 struct p_block_ack *p;
1499 if (mdev->state.conn < C_CONNECTED)
1502 sock = &mdev->tconn->meta;
1503 p = drbd_prepare_command(mdev, sock);
1507 p->block_id = block_id;
1508 p->blksize = blksize;
1509 p->seq_num = cpu_to_be32(atomic_inc_return(&mdev->packet_seq));
1510 return drbd_send_command(mdev, sock, cmd, sizeof(*p), NULL, 0);
1513 /* dp->sector and dp->block_id already/still in network byte order,
1514 * data_size is payload size according to dp->head,
1515 * and may need to be corrected for digest size. */
1516 void drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packet cmd,
1517 struct p_data *dp, int data_size)
1519 if (mdev->tconn->peer_integrity_tfm)
1520 data_size -= crypto_hash_digestsize(mdev->tconn->peer_integrity_tfm);
1521 _drbd_send_ack(mdev, cmd, dp->sector, cpu_to_be32(data_size),
1525 void drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packet cmd,
1526 struct p_block_req *rp)
1528 _drbd_send_ack(mdev, cmd, rp->sector, rp->blksize, rp->block_id);
1532 * drbd_send_ack() - Sends an ack packet
1533 * @mdev: DRBD device
1534 * @cmd: packet command code
1535 * @peer_req: peer request
1537 int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packet cmd,
1538 struct drbd_peer_request *peer_req)
1540 return _drbd_send_ack(mdev, cmd,
1541 cpu_to_be64(peer_req->i.sector),
1542 cpu_to_be32(peer_req->i.size),
1543 peer_req->block_id);
1546 /* This function misuses the block_id field to signal if the blocks
1547 * are is sync or not. */
1548 int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packet cmd,
1549 sector_t sector, int blksize, u64 block_id)
1551 return _drbd_send_ack(mdev, cmd,
1552 cpu_to_be64(sector),
1553 cpu_to_be32(blksize),
1554 cpu_to_be64(block_id));
1557 int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1558 sector_t sector, int size, u64 block_id)
1560 struct drbd_socket *sock;
1561 struct p_block_req *p;
1563 sock = &mdev->tconn->data;
1564 p = drbd_prepare_command(mdev, sock);
1567 p->sector = cpu_to_be64(sector);
1568 p->block_id = block_id;
1569 p->blksize = cpu_to_be32(size);
1570 return drbd_send_command(mdev, sock, cmd, sizeof(*p), NULL, 0);
1573 int drbd_send_drequest_csum(struct drbd_conf *mdev, sector_t sector, int size,
1574 void *digest, int digest_size, enum drbd_packet cmd)
1576 struct drbd_socket *sock;
1577 struct p_block_req *p;
1579 /* FIXME: Put the digest into the preallocated socket buffer. */
1581 sock = &mdev->tconn->data;
1582 p = drbd_prepare_command(mdev, sock);
1585 p->sector = cpu_to_be64(sector);
1586 p->block_id = ID_SYNCER /* unused */;
1587 p->blksize = cpu_to_be32(size);
1588 return drbd_send_command(mdev, sock, cmd, sizeof(*p),
1589 digest, digest_size);
1592 int drbd_send_ov_request(struct drbd_conf *mdev, sector_t sector, int size)
1594 struct drbd_socket *sock;
1595 struct p_block_req *p;
1597 sock = &mdev->tconn->data;
1598 p = drbd_prepare_command(mdev, sock);
1601 p->sector = cpu_to_be64(sector);
1602 p->block_id = ID_SYNCER /* unused */;
1603 p->blksize = cpu_to_be32(size);
1604 return drbd_send_command(mdev, sock, P_OV_REQUEST, sizeof(*p), NULL, 0);
1607 /* called on sndtimeo
1608 * returns false if we should retry,
1609 * true if we think connection is dead
1611 static int we_should_drop_the_connection(struct drbd_tconn *tconn, struct socket *sock)
1614 /* long elapsed = (long)(jiffies - mdev->last_received); */
1616 drop_it = tconn->meta.socket == sock
1617 || !tconn->asender.task
1618 || get_t_state(&tconn->asender) != RUNNING
1619 || tconn->cstate < C_WF_REPORT_PARAMS;
1624 drop_it = !--tconn->ko_count;
1626 conn_err(tconn, "[%s/%d] sock_sendmsg time expired, ko = %u\n",
1627 current->comm, current->pid, tconn->ko_count);
1628 request_ping(tconn);
1631 return drop_it; /* && (mdev->state == R_PRIMARY) */;
1634 static void drbd_update_congested(struct drbd_tconn *tconn)
1636 struct sock *sk = tconn->data.socket->sk;
1637 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
1638 set_bit(NET_CONGESTED, &tconn->flags);
1641 /* The idea of sendpage seems to be to put some kind of reference
1642 * to the page into the skb, and to hand it over to the NIC. In
1643 * this process get_page() gets called.
1645 * As soon as the page was really sent over the network put_page()
1646 * gets called by some part of the network layer. [ NIC driver? ]
1648 * [ get_page() / put_page() increment/decrement the count. If count
1649 * reaches 0 the page will be freed. ]
1651 * This works nicely with pages from FSs.
1652 * But this means that in protocol A we might signal IO completion too early!
1654 * In order not to corrupt data during a resync we must make sure
1655 * that we do not reuse our own buffer pages (EEs) to early, therefore
1656 * we have the net_ee list.
1658 * XFS seems to have problems, still, it submits pages with page_count == 0!
1659 * As a workaround, we disable sendpage on pages
1660 * with page_count == 0 or PageSlab.
1662 static int _drbd_no_send_page(struct drbd_conf *mdev, struct page *page,
1663 int offset, size_t size, unsigned msg_flags)
1665 struct socket *socket;
1669 socket = mdev->tconn->data.socket;
1670 addr = kmap(page) + offset;
1671 err = drbd_send_all(mdev->tconn, socket, addr, size, msg_flags);
1674 mdev->send_cnt += size >> 9;
1678 static int _drbd_send_page(struct drbd_conf *mdev, struct page *page,
1679 int offset, size_t size, unsigned msg_flags)
1681 struct socket *socket = mdev->tconn->data.socket;
1682 mm_segment_t oldfs = get_fs();
1686 /* e.g. XFS meta- & log-data is in slab pages, which have a
1687 * page_count of 0 and/or have PageSlab() set.
1688 * we cannot use send_page for those, as that does get_page();
1689 * put_page(); and would cause either a VM_BUG directly, or
1690 * __page_cache_release a page that would actually still be referenced
1691 * by someone, leading to some obscure delayed Oops somewhere else. */
1692 if (disable_sendpage || (page_count(page) < 1) || PageSlab(page))
1693 return _drbd_no_send_page(mdev, page, offset, size, msg_flags);
1695 msg_flags |= MSG_NOSIGNAL;
1696 drbd_update_congested(mdev->tconn);
1701 sent = socket->ops->sendpage(socket, page, offset, len, msg_flags);
1703 if (sent == -EAGAIN) {
1704 if (we_should_drop_the_connection(mdev->tconn, socket))
1708 dev_warn(DEV, "%s: size=%d len=%d sent=%d\n",
1709 __func__, (int)size, len, sent);
1716 } while (len > 0 /* THINK && mdev->cstate >= C_CONNECTED*/);
1718 clear_bit(NET_CONGESTED, &mdev->tconn->flags);
1722 mdev->send_cnt += size >> 9;
1727 static int _drbd_send_bio(struct drbd_conf *mdev, struct bio *bio)
1729 struct bio_vec *bvec;
1731 /* hint all but last page with MSG_MORE */
1732 __bio_for_each_segment(bvec, bio, i, 0) {
1735 err = _drbd_no_send_page(mdev, bvec->bv_page,
1736 bvec->bv_offset, bvec->bv_len,
1737 i == bio->bi_vcnt - 1 ? 0 : MSG_MORE);
1744 static int _drbd_send_zc_bio(struct drbd_conf *mdev, struct bio *bio)
1746 struct bio_vec *bvec;
1748 /* hint all but last page with MSG_MORE */
1749 __bio_for_each_segment(bvec, bio, i, 0) {
1752 err = _drbd_send_page(mdev, bvec->bv_page,
1753 bvec->bv_offset, bvec->bv_len,
1754 i == bio->bi_vcnt - 1 ? 0 : MSG_MORE);
1761 static int _drbd_send_zc_ee(struct drbd_conf *mdev,
1762 struct drbd_peer_request *peer_req)
1764 struct page *page = peer_req->pages;
1765 unsigned len = peer_req->i.size;
1768 /* hint all but last page with MSG_MORE */
1769 page_chain_for_each(page) {
1770 unsigned l = min_t(unsigned, len, PAGE_SIZE);
1772 err = _drbd_send_page(mdev, page, 0, l,
1773 page_chain_next(page) ? MSG_MORE : 0);
1781 static u32 bio_flags_to_wire(struct drbd_conf *mdev, unsigned long bi_rw)
1783 if (mdev->tconn->agreed_pro_version >= 95)
1784 return (bi_rw & REQ_SYNC ? DP_RW_SYNC : 0) |
1785 (bi_rw & REQ_FUA ? DP_FUA : 0) |
1786 (bi_rw & REQ_FLUSH ? DP_FLUSH : 0) |
1787 (bi_rw & REQ_DISCARD ? DP_DISCARD : 0);
1789 return bi_rw & REQ_SYNC ? DP_RW_SYNC : 0;
1792 /* Used to send write requests
1793 * R_PRIMARY -> Peer (P_DATA)
1795 int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req)
1797 struct drbd_socket *sock;
1799 unsigned int dp_flags = 0;
1803 sock = &mdev->tconn->data;
1804 p = drbd_prepare_command(mdev, sock);
1805 dgs = mdev->tconn->integrity_tfm ? crypto_hash_digestsize(mdev->tconn->integrity_tfm) : 0;
1809 p->sector = cpu_to_be64(req->i.sector);
1810 p->block_id = (unsigned long)req;
1811 p->seq_num = cpu_to_be32(req->seq_num = atomic_inc_return(&mdev->packet_seq));
1812 dp_flags = bio_flags_to_wire(mdev, req->master_bio->bi_rw);
1813 if (mdev->state.conn >= C_SYNC_SOURCE &&
1814 mdev->state.conn <= C_PAUSED_SYNC_T)
1815 dp_flags |= DP_MAY_SET_IN_SYNC;
1816 if (mdev->tconn->agreed_pro_version >= 100) {
1817 if (req->rq_state & RQ_EXP_RECEIVE_ACK)
1818 dp_flags |= DP_SEND_RECEIVE_ACK;
1819 if (req->rq_state & RQ_EXP_WRITE_ACK)
1820 dp_flags |= DP_SEND_WRITE_ACK;
1822 p->dp_flags = cpu_to_be32(dp_flags);
1824 drbd_csum_bio(mdev, mdev->tconn->integrity_tfm, req->master_bio, p + 1);
1825 err = __send_command(mdev->tconn, mdev->vnr, sock, P_DATA, sizeof(*p) + dgs, NULL, req->i.size);
1827 /* For protocol A, we have to memcpy the payload into
1828 * socket buffers, as we may complete right away
1829 * as soon as we handed it over to tcp, at which point the data
1830 * pages may become invalid.
1832 * For data-integrity enabled, we copy it as well, so we can be
1833 * sure that even if the bio pages may still be modified, it
1834 * won't change the data on the wire, thus if the digest checks
1835 * out ok after sending on this side, but does not fit on the
1836 * receiving side, we sure have detected corruption elsewhere.
1838 if (!(req->rq_state & (RQ_EXP_RECEIVE_ACK | RQ_EXP_WRITE_ACK)) || dgs)
1839 err = _drbd_send_bio(mdev, req->master_bio);
1841 err = _drbd_send_zc_bio(mdev, req->master_bio);
1843 /* double check digest, sometimes buffers have been modified in flight. */
1844 if (dgs > 0 && dgs <= 64) {
1845 /* 64 byte, 512 bit, is the largest digest size
1846 * currently supported in kernel crypto. */
1847 unsigned char digest[64];
1848 drbd_csum_bio(mdev, mdev->tconn->integrity_tfm, req->master_bio, digest);
1849 if (memcmp(p + 1, digest, dgs)) {
1851 "Digest mismatch, buffer modified by upper layers during write: %llus +%u\n",
1852 (unsigned long long)req->i.sector, req->i.size);
1854 } /* else if (dgs > 64) {
1855 ... Be noisy about digest too large ...
1858 mutex_unlock(&sock->mutex); /* locked by drbd_prepare_command() */
1863 /* answer packet, used to send data back for read requests:
1864 * Peer -> (diskless) R_PRIMARY (P_DATA_REPLY)
1865 * C_SYNC_SOURCE -> C_SYNC_TARGET (P_RS_DATA_REPLY)
1867 int drbd_send_block(struct drbd_conf *mdev, enum drbd_packet cmd,
1868 struct drbd_peer_request *peer_req)
1870 struct drbd_socket *sock;
1875 sock = &mdev->tconn->data;
1876 p = drbd_prepare_command(mdev, sock);
1878 dgs = mdev->tconn->integrity_tfm ? crypto_hash_digestsize(mdev->tconn->integrity_tfm) : 0;
1882 p->sector = cpu_to_be64(peer_req->i.sector);
1883 p->block_id = peer_req->block_id;
1884 p->seq_num = 0; /* unused */
1887 drbd_csum_ee(mdev, mdev->tconn->integrity_tfm, peer_req, p + 1);
1888 err = __send_command(mdev->tconn, mdev->vnr, sock, cmd, sizeof(*p) + dgs, NULL, peer_req->i.size);
1890 err = _drbd_send_zc_ee(mdev, peer_req);
1891 mutex_unlock(&sock->mutex); /* locked by drbd_prepare_command() */
1896 int drbd_send_out_of_sync(struct drbd_conf *mdev, struct drbd_request *req)
1898 struct drbd_socket *sock;
1899 struct p_block_desc *p;
1901 sock = &mdev->tconn->data;
1902 p = drbd_prepare_command(mdev, sock);
1905 p->sector = cpu_to_be64(req->i.sector);
1906 p->blksize = cpu_to_be32(req->i.size);
1907 return drbd_send_command(mdev, sock, P_OUT_OF_SYNC, sizeof(*p), NULL, 0);
1911 drbd_send distinguishes two cases:
1913 Packets sent via the data socket "sock"
1914 and packets sent via the meta data socket "msock"
1917 -----------------+-------------------------+------------------------------
1918 timeout conf.timeout / 2 conf.timeout / 2
1919 timeout action send a ping via msock Abort communication
1920 and close all sockets
1924 * you must have down()ed the appropriate [m]sock_mutex elsewhere!
1926 int drbd_send(struct drbd_tconn *tconn, struct socket *sock,
1927 void *buf, size_t size, unsigned msg_flags)
1936 /* THINK if (signal_pending) return ... ? */
1941 msg.msg_name = NULL;
1942 msg.msg_namelen = 0;
1943 msg.msg_control = NULL;
1944 msg.msg_controllen = 0;
1945 msg.msg_flags = msg_flags | MSG_NOSIGNAL;
1947 if (sock == tconn->data.socket) {
1949 tconn->ko_count = rcu_dereference(tconn->net_conf)->ko_count;
1951 drbd_update_congested(tconn);
1955 * tcp_sendmsg does _not_ use its size parameter at all ?
1957 * -EAGAIN on timeout, -EINTR on signal.
1960 * do we need to block DRBD_SIG if sock == &meta.socket ??
1961 * otherwise wake_asender() might interrupt some send_*Ack !
1963 rv = kernel_sendmsg(sock, &msg, &iov, 1, size);
1964 if (rv == -EAGAIN) {
1965 if (we_should_drop_the_connection(tconn, sock))
1971 flush_signals(current);
1979 } while (sent < size);
1981 if (sock == tconn->data.socket)
1982 clear_bit(NET_CONGESTED, &tconn->flags);
1985 if (rv != -EAGAIN) {
1986 conn_err(tconn, "%s_sendmsg returned %d\n",
1987 sock == tconn->meta.socket ? "msock" : "sock",
1989 conn_request_state(tconn, NS(conn, C_BROKEN_PIPE), CS_HARD);
1991 conn_request_state(tconn, NS(conn, C_TIMEOUT), CS_HARD);
1998 * drbd_send_all - Send an entire buffer
2000 * Returns 0 upon success and a negative error value otherwise.
2002 int drbd_send_all(struct drbd_tconn *tconn, struct socket *sock, void *buffer,
2003 size_t size, unsigned msg_flags)
2007 err = drbd_send(tconn, sock, buffer, size, msg_flags);
2015 static int drbd_open(struct block_device *bdev, fmode_t mode)
2017 struct drbd_conf *mdev = bdev->bd_disk->private_data;
2018 unsigned long flags;
2021 mutex_lock(&drbd_main_mutex);
2022 spin_lock_irqsave(&mdev->tconn->req_lock, flags);
2023 /* to have a stable mdev->state.role
2024 * and no race with updating open_cnt */
2026 if (mdev->state.role != R_PRIMARY) {
2027 if (mode & FMODE_WRITE)
2029 else if (!allow_oos)
2035 spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
2036 mutex_unlock(&drbd_main_mutex);
2041 static int drbd_release(struct gendisk *gd, fmode_t mode)
2043 struct drbd_conf *mdev = gd->private_data;
2044 mutex_lock(&drbd_main_mutex);
2046 mutex_unlock(&drbd_main_mutex);
2050 static void drbd_set_defaults(struct drbd_conf *mdev)
2052 /* Beware! The actual layout differs
2053 * between big endian and little endian */
2054 mdev->state = (union drbd_dev_state) {
2055 { .role = R_SECONDARY,
2057 .conn = C_STANDALONE,
2063 void drbd_init_set_defaults(struct drbd_conf *mdev)
2065 /* the memset(,0,) did most of this.
2066 * note: only assignments, no allocation in here */
2068 drbd_set_defaults(mdev);
2070 atomic_set(&mdev->ap_bio_cnt, 0);
2071 atomic_set(&mdev->ap_pending_cnt, 0);
2072 atomic_set(&mdev->rs_pending_cnt, 0);
2073 atomic_set(&mdev->unacked_cnt, 0);
2074 atomic_set(&mdev->local_cnt, 0);
2075 atomic_set(&mdev->pp_in_use_by_net, 0);
2076 atomic_set(&mdev->rs_sect_in, 0);
2077 atomic_set(&mdev->rs_sect_ev, 0);
2078 atomic_set(&mdev->ap_in_flight, 0);
2079 atomic_set(&mdev->md_io_in_use, 0);
2081 mutex_init(&mdev->own_state_mutex);
2082 mdev->state_mutex = &mdev->own_state_mutex;
2084 spin_lock_init(&mdev->al_lock);
2085 spin_lock_init(&mdev->peer_seq_lock);
2087 INIT_LIST_HEAD(&mdev->active_ee);
2088 INIT_LIST_HEAD(&mdev->sync_ee);
2089 INIT_LIST_HEAD(&mdev->done_ee);
2090 INIT_LIST_HEAD(&mdev->read_ee);
2091 INIT_LIST_HEAD(&mdev->net_ee);
2092 INIT_LIST_HEAD(&mdev->resync_reads);
2093 INIT_LIST_HEAD(&mdev->resync_work.list);
2094 INIT_LIST_HEAD(&mdev->unplug_work.list);
2095 INIT_LIST_HEAD(&mdev->go_diskless.list);
2096 INIT_LIST_HEAD(&mdev->md_sync_work.list);
2097 INIT_LIST_HEAD(&mdev->start_resync_work.list);
2098 INIT_LIST_HEAD(&mdev->bm_io_work.w.list);
2100 mdev->resync_work.cb = w_resync_timer;
2101 mdev->unplug_work.cb = w_send_write_hint;
2102 mdev->go_diskless.cb = w_go_diskless;
2103 mdev->md_sync_work.cb = w_md_sync;
2104 mdev->bm_io_work.w.cb = w_bitmap_io;
2105 mdev->start_resync_work.cb = w_start_resync;
2107 mdev->resync_work.mdev = mdev;
2108 mdev->unplug_work.mdev = mdev;
2109 mdev->go_diskless.mdev = mdev;
2110 mdev->md_sync_work.mdev = mdev;
2111 mdev->bm_io_work.w.mdev = mdev;
2112 mdev->start_resync_work.mdev = mdev;
2114 init_timer(&mdev->resync_timer);
2115 init_timer(&mdev->md_sync_timer);
2116 init_timer(&mdev->start_resync_timer);
2117 init_timer(&mdev->request_timer);
2118 mdev->resync_timer.function = resync_timer_fn;
2119 mdev->resync_timer.data = (unsigned long) mdev;
2120 mdev->md_sync_timer.function = md_sync_timer_fn;
2121 mdev->md_sync_timer.data = (unsigned long) mdev;
2122 mdev->start_resync_timer.function = start_resync_timer_fn;
2123 mdev->start_resync_timer.data = (unsigned long) mdev;
2124 mdev->request_timer.function = request_timer_fn;
2125 mdev->request_timer.data = (unsigned long) mdev;
2127 init_waitqueue_head(&mdev->misc_wait);
2128 init_waitqueue_head(&mdev->state_wait);
2129 init_waitqueue_head(&mdev->ee_wait);
2130 init_waitqueue_head(&mdev->al_wait);
2131 init_waitqueue_head(&mdev->seq_wait);
2133 mdev->resync_wenr = LC_FREE;
2134 mdev->peer_max_bio_size = DRBD_MAX_BIO_SIZE_SAFE;
2135 mdev->local_max_bio_size = DRBD_MAX_BIO_SIZE_SAFE;
2138 void drbd_mdev_cleanup(struct drbd_conf *mdev)
2141 if (mdev->tconn->receiver.t_state != NONE)
2142 dev_err(DEV, "ASSERT FAILED: receiver t_state == %d expected 0.\n",
2143 mdev->tconn->receiver.t_state);
2154 mdev->rs_failed = 0;
2155 mdev->rs_last_events = 0;
2156 mdev->rs_last_sect_ev = 0;
2157 for (i = 0; i < DRBD_SYNC_MARKS; i++) {
2158 mdev->rs_mark_left[i] = 0;
2159 mdev->rs_mark_time[i] = 0;
2161 D_ASSERT(mdev->tconn->net_conf == NULL);
2163 drbd_set_my_capacity(mdev, 0);
2165 /* maybe never allocated. */
2166 drbd_bm_resize(mdev, 0, 1);
2167 drbd_bm_cleanup(mdev);
2170 drbd_free_bc(mdev->ldev);
2173 clear_bit(AL_SUSPENDED, &mdev->flags);
2175 D_ASSERT(list_empty(&mdev->active_ee));
2176 D_ASSERT(list_empty(&mdev->sync_ee));
2177 D_ASSERT(list_empty(&mdev->done_ee));
2178 D_ASSERT(list_empty(&mdev->read_ee));
2179 D_ASSERT(list_empty(&mdev->net_ee));
2180 D_ASSERT(list_empty(&mdev->resync_reads));
2181 D_ASSERT(list_empty(&mdev->tconn->data.work.q));
2182 D_ASSERT(list_empty(&mdev->tconn->meta.work.q));
2183 D_ASSERT(list_empty(&mdev->resync_work.list));
2184 D_ASSERT(list_empty(&mdev->unplug_work.list));
2185 D_ASSERT(list_empty(&mdev->go_diskless.list));
2187 drbd_set_defaults(mdev);
2191 static void drbd_destroy_mempools(void)
2195 while (drbd_pp_pool) {
2196 page = drbd_pp_pool;
2197 drbd_pp_pool = (struct page *)page_private(page);
2202 /* D_ASSERT(atomic_read(&drbd_pp_vacant)==0); */
2204 if (drbd_md_io_bio_set)
2205 bioset_free(drbd_md_io_bio_set);
2206 if (drbd_md_io_page_pool)
2207 mempool_destroy(drbd_md_io_page_pool);
2208 if (drbd_ee_mempool)
2209 mempool_destroy(drbd_ee_mempool);
2210 if (drbd_request_mempool)
2211 mempool_destroy(drbd_request_mempool);
2213 kmem_cache_destroy(drbd_ee_cache);
2214 if (drbd_request_cache)
2215 kmem_cache_destroy(drbd_request_cache);
2216 if (drbd_bm_ext_cache)
2217 kmem_cache_destroy(drbd_bm_ext_cache);
2218 if (drbd_al_ext_cache)
2219 kmem_cache_destroy(drbd_al_ext_cache);
2221 drbd_md_io_bio_set = NULL;
2222 drbd_md_io_page_pool = NULL;
2223 drbd_ee_mempool = NULL;
2224 drbd_request_mempool = NULL;
2225 drbd_ee_cache = NULL;
2226 drbd_request_cache = NULL;
2227 drbd_bm_ext_cache = NULL;
2228 drbd_al_ext_cache = NULL;
2233 static int drbd_create_mempools(void)
2236 const int number = (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count;
2239 /* prepare our caches and mempools */
2240 drbd_request_mempool = NULL;
2241 drbd_ee_cache = NULL;
2242 drbd_request_cache = NULL;
2243 drbd_bm_ext_cache = NULL;
2244 drbd_al_ext_cache = NULL;
2245 drbd_pp_pool = NULL;
2246 drbd_md_io_page_pool = NULL;
2247 drbd_md_io_bio_set = NULL;
2250 drbd_request_cache = kmem_cache_create(
2251 "drbd_req", sizeof(struct drbd_request), 0, 0, NULL);
2252 if (drbd_request_cache == NULL)
2255 drbd_ee_cache = kmem_cache_create(
2256 "drbd_ee", sizeof(struct drbd_peer_request), 0, 0, NULL);
2257 if (drbd_ee_cache == NULL)
2260 drbd_bm_ext_cache = kmem_cache_create(
2261 "drbd_bm", sizeof(struct bm_extent), 0, 0, NULL);
2262 if (drbd_bm_ext_cache == NULL)
2265 drbd_al_ext_cache = kmem_cache_create(
2266 "drbd_al", sizeof(struct lc_element), 0, 0, NULL);
2267 if (drbd_al_ext_cache == NULL)
2271 drbd_md_io_bio_set = bioset_create(DRBD_MIN_POOL_PAGES, 0);
2272 if (drbd_md_io_bio_set == NULL)
2275 drbd_md_io_page_pool = mempool_create_page_pool(DRBD_MIN_POOL_PAGES, 0);
2276 if (drbd_md_io_page_pool == NULL)
2279 drbd_request_mempool = mempool_create(number,
2280 mempool_alloc_slab, mempool_free_slab, drbd_request_cache);
2281 if (drbd_request_mempool == NULL)
2284 drbd_ee_mempool = mempool_create(number,
2285 mempool_alloc_slab, mempool_free_slab, drbd_ee_cache);
2286 if (drbd_ee_mempool == NULL)
2289 /* drbd's page pool */
2290 spin_lock_init(&drbd_pp_lock);
2292 for (i = 0; i < number; i++) {
2293 page = alloc_page(GFP_HIGHUSER);
2296 set_page_private(page, (unsigned long)drbd_pp_pool);
2297 drbd_pp_pool = page;
2299 drbd_pp_vacant = number;
2304 drbd_destroy_mempools(); /* in case we allocated some */
2308 static int drbd_notify_sys(struct notifier_block *this, unsigned long code,
2311 /* just so we have it. you never know what interesting things we
2312 * might want to do here some day...
2318 static struct notifier_block drbd_notifier = {
2319 .notifier_call = drbd_notify_sys,
2322 static void drbd_release_all_peer_reqs(struct drbd_conf *mdev)
2326 rr = drbd_free_peer_reqs(mdev, &mdev->active_ee);
2328 dev_err(DEV, "%d EEs in active list found!\n", rr);
2330 rr = drbd_free_peer_reqs(mdev, &mdev->sync_ee);
2332 dev_err(DEV, "%d EEs in sync list found!\n", rr);
2334 rr = drbd_free_peer_reqs(mdev, &mdev->read_ee);
2336 dev_err(DEV, "%d EEs in read list found!\n", rr);
2338 rr = drbd_free_peer_reqs(mdev, &mdev->done_ee);
2340 dev_err(DEV, "%d EEs in done list found!\n", rr);
2342 rr = drbd_free_peer_reqs(mdev, &mdev->net_ee);
2344 dev_err(DEV, "%d EEs in net list found!\n", rr);
2347 /* caution. no locking. */
2348 void drbd_minor_destroy(struct kref *kref)
2350 struct drbd_conf *mdev = container_of(kref, struct drbd_conf, kref);
2351 struct drbd_tconn *tconn = mdev->tconn;
2353 del_timer_sync(&mdev->request_timer);
2355 /* paranoia asserts */
2356 D_ASSERT(mdev->open_cnt == 0);
2357 D_ASSERT(list_empty(&mdev->tconn->data.work.q));
2358 /* end paranoia asserts */
2360 /* cleanup stuff that may have been allocated during
2361 * device (re-)configuration or state changes */
2363 if (mdev->this_bdev)
2364 bdput(mdev->this_bdev);
2366 drbd_free_bc(mdev->ldev);
2369 drbd_release_all_peer_reqs(mdev);
2371 lc_destroy(mdev->act_log);
2372 lc_destroy(mdev->resync);
2374 kfree(mdev->p_uuid);
2375 /* mdev->p_uuid = NULL; */
2377 if (mdev->bitmap) /* should no longer be there. */
2378 drbd_bm_cleanup(mdev);
2379 __free_page(mdev->md_io_page);
2380 put_disk(mdev->vdisk);
2381 blk_cleanup_queue(mdev->rq_queue);
2382 kfree(mdev->rs_plan_s);
2385 kref_put(&tconn->kref, &conn_destroy);
2388 static void drbd_cleanup(void)
2391 struct drbd_conf *mdev;
2392 struct drbd_tconn *tconn, *tmp;
2394 unregister_reboot_notifier(&drbd_notifier);
2396 /* first remove proc,
2397 * drbdsetup uses it's presence to detect
2398 * whether DRBD is loaded.
2399 * If we would get stuck in proc removal,
2400 * but have netlink already deregistered,
2401 * some drbdsetup commands may wait forever
2405 remove_proc_entry("drbd", NULL);
2407 drbd_genl_unregister();
2409 idr_for_each_entry(&minors, mdev, i) {
2410 idr_remove(&minors, mdev_to_minor(mdev));
2411 idr_remove(&mdev->tconn->volumes, mdev->vnr);
2412 del_gendisk(mdev->vdisk);
2413 /* synchronize_rcu(); No other threads running at this point */
2414 kref_put(&mdev->kref, &drbd_minor_destroy);
2417 /* not _rcu since, no other updater anymore. Genl already unregistered */
2418 list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) {
2419 list_del(&tconn->all_tconn); /* not _rcu no proc, not other threads */
2420 /* synchronize_rcu(); */
2421 kref_put(&tconn->kref, &conn_destroy);
2424 drbd_destroy_mempools();
2425 unregister_blkdev(DRBD_MAJOR, "drbd");
2427 idr_destroy(&minors);
2429 printk(KERN_INFO "drbd: module cleanup done.\n");
2433 * drbd_congested() - Callback for pdflush
2434 * @congested_data: User data
2435 * @bdi_bits: Bits pdflush is currently interested in
2437 * Returns 1<<BDI_async_congested and/or 1<<BDI_sync_congested if we are congested.
2439 static int drbd_congested(void *congested_data, int bdi_bits)
2441 struct drbd_conf *mdev = congested_data;
2442 struct request_queue *q;
2446 if (!may_inc_ap_bio(mdev)) {
2447 /* DRBD has frozen IO */
2453 if (get_ldev(mdev)) {
2454 q = bdev_get_queue(mdev->ldev->backing_bdev);
2455 r = bdi_congested(&q->backing_dev_info, bdi_bits);
2461 if (bdi_bits & (1 << BDI_async_congested) && test_bit(NET_CONGESTED, &mdev->tconn->flags)) {
2462 r |= (1 << BDI_async_congested);
2463 reason = reason == 'b' ? 'a' : 'n';
2467 mdev->congestion_reason = reason;
2471 static void drbd_init_workqueue(struct drbd_work_queue* wq)
2473 sema_init(&wq->s, 0);
2474 spin_lock_init(&wq->q_lock);
2475 INIT_LIST_HEAD(&wq->q);
2478 struct drbd_tconn *conn_get_by_name(const char *name)
2480 struct drbd_tconn *tconn;
2482 if (!name || !name[0])
2486 list_for_each_entry_rcu(tconn, &drbd_tconns, all_tconn) {
2487 if (!strcmp(tconn->name, name)) {
2488 kref_get(&tconn->kref);
2498 struct drbd_tconn *conn_get_by_addrs(void *my_addr, int my_addr_len,
2499 void *peer_addr, int peer_addr_len)
2501 struct drbd_tconn *tconn;
2504 list_for_each_entry_rcu(tconn, &drbd_tconns, all_tconn) {
2505 if (tconn->my_addr_len == my_addr_len &&
2506 tconn->peer_addr_len == peer_addr_len &&
2507 !memcmp(&tconn->my_addr, my_addr, my_addr_len) &&
2508 !memcmp(&tconn->peer_addr, peer_addr, peer_addr_len)) {
2509 kref_get(&tconn->kref);
2519 static int drbd_alloc_socket(struct drbd_socket *socket)
2521 socket->rbuf = (void *) __get_free_page(GFP_KERNEL);
2524 socket->sbuf = (void *) __get_free_page(GFP_KERNEL);
2530 static void drbd_free_socket(struct drbd_socket *socket)
2532 free_page((unsigned long) socket->sbuf);
2533 free_page((unsigned long) socket->rbuf);
2536 void conn_free_crypto(struct drbd_tconn *tconn)
2538 drbd_free_sock(tconn);
2540 crypto_free_hash(tconn->csums_tfm);
2541 crypto_free_hash(tconn->verify_tfm);
2542 crypto_free_hash(tconn->cram_hmac_tfm);
2543 crypto_free_hash(tconn->integrity_tfm);
2544 crypto_free_hash(tconn->peer_integrity_tfm);
2545 kfree(tconn->int_dig_in);
2546 kfree(tconn->int_dig_vv);
2548 tconn->csums_tfm = NULL;
2549 tconn->verify_tfm = NULL;
2550 tconn->cram_hmac_tfm = NULL;
2551 tconn->integrity_tfm = NULL;
2552 tconn->peer_integrity_tfm = NULL;
2553 tconn->int_dig_in = NULL;
2554 tconn->int_dig_vv = NULL;
2557 int set_resource_options(struct drbd_tconn *tconn, struct res_opts *res_opts)
2559 cpumask_var_t new_cpu_mask;
2562 if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL))
2565 retcode = ERR_NOMEM;
2566 drbd_msg_put_info("unable to allocate cpumask");
2569 /* silently ignore cpu mask on UP kernel */
2570 if (nr_cpu_ids > 1 && res_opts->cpu_mask[0] != 0) {
2571 /* FIXME: Get rid of constant 32 here */
2572 err = __bitmap_parse(res_opts->cpu_mask, 32, 0,
2573 cpumask_bits(new_cpu_mask), nr_cpu_ids);
2575 conn_warn(tconn, "__bitmap_parse() failed with %d\n", err);
2576 /* retcode = ERR_CPU_MASK_PARSE; */
2580 tconn->res_opts = *res_opts;
2581 if (!cpumask_equal(tconn->cpu_mask, new_cpu_mask)) {
2582 cpumask_copy(tconn->cpu_mask, new_cpu_mask);
2583 drbd_calc_cpu_mask(tconn);
2584 tconn->receiver.reset_cpu_mask = 1;
2585 tconn->asender.reset_cpu_mask = 1;
2586 tconn->worker.reset_cpu_mask = 1;
2591 free_cpumask_var(new_cpu_mask);
2596 /* caller must be under genl_lock() */
2597 struct drbd_tconn *conn_create(const char *name, struct res_opts *res_opts)
2599 struct drbd_tconn *tconn;
2601 tconn = kzalloc(sizeof(struct drbd_tconn), GFP_KERNEL);
2605 tconn->name = kstrdup(name, GFP_KERNEL);
2609 if (drbd_alloc_socket(&tconn->data))
2611 if (drbd_alloc_socket(&tconn->meta))
2614 if (!zalloc_cpumask_var(&tconn->cpu_mask, GFP_KERNEL))
2617 if (set_resource_options(tconn, res_opts))
2620 if (!tl_init(tconn))
2623 tconn->current_epoch = kzalloc(sizeof(struct drbd_epoch), GFP_KERNEL);
2624 if (!tconn->current_epoch)
2626 INIT_LIST_HEAD(&tconn->current_epoch->list);
2628 spin_lock_init(&tconn->epoch_lock);
2629 tconn->write_ordering = WO_bdev_flush;
2631 tconn->cstate = C_STANDALONE;
2632 mutex_init(&tconn->cstate_mutex);
2633 spin_lock_init(&tconn->req_lock);
2634 mutex_init(&tconn->conf_update);
2635 init_waitqueue_head(&tconn->ping_wait);
2636 idr_init(&tconn->volumes);
2638 drbd_init_workqueue(&tconn->data.work);
2639 mutex_init(&tconn->data.mutex);
2641 drbd_init_workqueue(&tconn->meta.work);
2642 mutex_init(&tconn->meta.mutex);
2644 drbd_thread_init(tconn, &tconn->receiver, drbdd_init, "receiver");
2645 drbd_thread_init(tconn, &tconn->worker, drbd_worker, "worker");
2646 drbd_thread_init(tconn, &tconn->asender, drbd_asender, "asender");
2648 kref_init(&tconn->kref);
2649 list_add_tail_rcu(&tconn->all_tconn, &drbd_tconns);
2654 kfree(tconn->current_epoch);
2656 free_cpumask_var(tconn->cpu_mask);
2657 drbd_free_socket(&tconn->meta);
2658 drbd_free_socket(&tconn->data);
2665 void conn_destroy(struct kref *kref)
2667 struct drbd_tconn *tconn = container_of(kref, struct drbd_tconn, kref);
2669 if (atomic_read(&tconn->current_epoch->epoch_size) != 0)
2670 conn_err(tconn, "epoch_size:%d\n", atomic_read(&tconn->current_epoch->epoch_size));
2671 kfree(tconn->current_epoch);
2673 idr_destroy(&tconn->volumes);
2675 free_cpumask_var(tconn->cpu_mask);
2676 drbd_free_socket(&tconn->meta);
2677 drbd_free_socket(&tconn->data);
2679 kfree(tconn->int_dig_in);
2680 kfree(tconn->int_dig_vv);
2684 enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor, int vnr)
2686 struct drbd_conf *mdev;
2687 struct gendisk *disk;
2688 struct request_queue *q;
2690 int minor_got = minor;
2691 enum drbd_ret_code err = ERR_NOMEM;
2693 mdev = minor_to_mdev(minor);
2695 return ERR_MINOR_EXISTS;
2697 /* GFP_KERNEL, we are outside of all write-out paths */
2698 mdev = kzalloc(sizeof(struct drbd_conf), GFP_KERNEL);
2702 kref_get(&tconn->kref);
2703 mdev->tconn = tconn;
2705 mdev->minor = minor;
2708 drbd_init_set_defaults(mdev);
2710 q = blk_alloc_queue(GFP_KERNEL);
2714 q->queuedata = mdev;
2716 disk = alloc_disk(1);
2721 set_disk_ro(disk, true);
2724 disk->major = DRBD_MAJOR;
2725 disk->first_minor = minor;
2726 disk->fops = &drbd_ops;
2727 sprintf(disk->disk_name, "drbd%d", minor);
2728 disk->private_data = mdev;
2730 mdev->this_bdev = bdget(MKDEV(DRBD_MAJOR, minor));
2731 /* we have no partitions. we contain only ourselves. */
2732 mdev->this_bdev->bd_contains = mdev->this_bdev;
2734 q->backing_dev_info.congested_fn = drbd_congested;
2735 q->backing_dev_info.congested_data = mdev;
2737 blk_queue_make_request(q, drbd_make_request);
2738 /* Setting the max_hw_sectors to an odd value of 8kibyte here
2739 This triggers a max_bio_size message upon first attach or connect */
2740 blk_queue_max_hw_sectors(q, DRBD_MAX_BIO_SIZE_SAFE >> 8);
2741 blk_queue_bounce_limit(q, BLK_BOUNCE_ANY);
2742 blk_queue_merge_bvec(q, drbd_merge_bvec);
2743 q->queue_lock = &mdev->tconn->req_lock; /* needed since we use */
2745 mdev->md_io_page = alloc_page(GFP_KERNEL);
2746 if (!mdev->md_io_page)
2747 goto out_no_io_page;
2749 if (drbd_bm_init(mdev))
2751 mdev->read_requests = RB_ROOT;
2752 mdev->write_requests = RB_ROOT;
2754 if (!idr_pre_get(&minors, GFP_KERNEL))
2755 goto out_no_minor_idr;
2756 if (idr_get_new_above(&minors, mdev, minor, &minor_got))
2757 goto out_no_minor_idr;
2758 if (minor_got != minor) {
2759 err = ERR_MINOR_EXISTS;
2760 drbd_msg_put_info("requested minor exists already");
2761 goto out_idr_remove_minor;
2764 if (!idr_pre_get(&tconn->volumes, GFP_KERNEL))
2765 goto out_idr_remove_minor;
2766 if (idr_get_new_above(&tconn->volumes, mdev, vnr, &vnr_got))
2767 goto out_idr_remove_minor;
2768 if (vnr_got != vnr) {
2769 err = ERR_INVALID_REQUEST;
2770 drbd_msg_put_info("requested volume exists already");
2771 goto out_idr_remove_vol;
2774 kref_init(&mdev->kref); /* one ref for both idrs and the the add_disk */
2776 /* inherit the connection state */
2777 mdev->state.conn = tconn->cstate;
2778 if (mdev->state.conn == C_WF_REPORT_PARAMS)
2779 drbd_connected(mdev);
2784 idr_remove(&tconn->volumes, vnr_got);
2785 out_idr_remove_minor:
2786 idr_remove(&minors, minor_got);
2789 drbd_bm_cleanup(mdev);
2791 __free_page(mdev->md_io_page);
2795 blk_cleanup_queue(q);
2798 kref_put(&tconn->kref, &conn_destroy);
2802 int __init drbd_init(void)
2806 if (minor_count < DRBD_MINOR_COUNT_MIN || minor_count > DRBD_MINOR_COUNT_MAX) {
2808 "drbd: invalid minor_count (%d)\n", minor_count);
2812 minor_count = DRBD_MINOR_COUNT_DEF;
2816 err = register_blkdev(DRBD_MAJOR, "drbd");
2819 "drbd: unable to register block device major %d\n",
2824 err = drbd_genl_register();
2826 printk(KERN_ERR "drbd: unable to register generic netlink family\n");
2831 register_reboot_notifier(&drbd_notifier);
2834 * allocate all necessary structs
2838 init_waitqueue_head(&drbd_pp_wait);
2840 drbd_proc = NULL; /* play safe for drbd_cleanup */
2843 err = drbd_create_mempools();
2847 drbd_proc = proc_create_data("drbd", S_IFREG | S_IRUGO , NULL, &drbd_proc_fops, NULL);
2849 printk(KERN_ERR "drbd: unable to register proc file\n");
2853 rwlock_init(&global_state_lock);
2854 INIT_LIST_HEAD(&drbd_tconns);
2856 printk(KERN_INFO "drbd: initialized. "
2857 "Version: " REL_VERSION " (api:%d/proto:%d-%d)\n",
2858 API_VERSION, PRO_VERSION_MIN, PRO_VERSION_MAX);
2859 printk(KERN_INFO "drbd: %s\n", drbd_buildtag());
2860 printk(KERN_INFO "drbd: registered as block device major %d\n",
2863 return 0; /* Success! */
2868 /* currently always the case */
2869 printk(KERN_ERR "drbd: ran out of memory\n");
2871 printk(KERN_ERR "drbd: initialization failure\n");
2875 void drbd_free_bc(struct drbd_backing_dev *ldev)
2880 blkdev_put(ldev->backing_bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
2881 blkdev_put(ldev->md_bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL);
2886 void drbd_free_sock(struct drbd_tconn *tconn)
2888 if (tconn->data.socket) {
2889 mutex_lock(&tconn->data.mutex);
2890 kernel_sock_shutdown(tconn->data.socket, SHUT_RDWR);
2891 sock_release(tconn->data.socket);
2892 tconn->data.socket = NULL;
2893 mutex_unlock(&tconn->data.mutex);
2895 if (tconn->meta.socket) {
2896 mutex_lock(&tconn->meta.mutex);
2897 kernel_sock_shutdown(tconn->meta.socket, SHUT_RDWR);
2898 sock_release(tconn->meta.socket);
2899 tconn->meta.socket = NULL;
2900 mutex_unlock(&tconn->meta.mutex);
2904 /* meta data management */
2906 struct meta_data_on_disk {
2907 u64 la_size; /* last agreed size. */
2908 u64 uuid[UI_SIZE]; /* UUIDs. */
2911 u32 flags; /* MDF */
2914 u32 al_offset; /* offset to this block */
2915 u32 al_nr_extents; /* important for restoring the AL */
2916 /* `-- act_log->nr_elements <-- ldev->dc.al_extents */
2917 u32 bm_offset; /* offset to the bitmap, from here */
2918 u32 bm_bytes_per_bit; /* BM_BLOCK_SIZE */
2919 u32 la_peer_max_bio_size; /* last peer max_bio_size */
2920 u32 reserved_u32[3];
2925 * drbd_md_sync() - Writes the meta data super block if the MD_DIRTY flag bit is set
2926 * @mdev: DRBD device.
2928 void drbd_md_sync(struct drbd_conf *mdev)
2930 struct meta_data_on_disk *buffer;
2934 del_timer(&mdev->md_sync_timer);
2935 /* timer may be rearmed by drbd_md_mark_dirty() now. */
2936 if (!test_and_clear_bit(MD_DIRTY, &mdev->flags))
2939 /* We use here D_FAILED and not D_ATTACHING because we try to write
2940 * metadata even if we detach due to a disk failure! */
2941 if (!get_ldev_if_state(mdev, D_FAILED))
2944 buffer = drbd_md_get_buffer(mdev);
2948 memset(buffer, 0, 512);
2950 buffer->la_size = cpu_to_be64(drbd_get_capacity(mdev->this_bdev));
2951 for (i = UI_CURRENT; i < UI_SIZE; i++)
2952 buffer->uuid[i] = cpu_to_be64(mdev->ldev->md.uuid[i]);
2953 buffer->flags = cpu_to_be32(mdev->ldev->md.flags);
2954 buffer->magic = cpu_to_be32(DRBD_MD_MAGIC_84_UNCLEAN);
2956 buffer->md_size_sect = cpu_to_be32(mdev->ldev->md.md_size_sect);
2957 buffer->al_offset = cpu_to_be32(mdev->ldev->md.al_offset);
2958 buffer->al_nr_extents = cpu_to_be32(mdev->act_log->nr_elements);
2959 buffer->bm_bytes_per_bit = cpu_to_be32(BM_BLOCK_SIZE);
2960 buffer->device_uuid = cpu_to_be64(mdev->ldev->md.device_uuid);
2962 buffer->bm_offset = cpu_to_be32(mdev->ldev->md.bm_offset);
2963 buffer->la_peer_max_bio_size = cpu_to_be32(mdev->peer_max_bio_size);
2965 D_ASSERT(drbd_md_ss__(mdev, mdev->ldev) == mdev->ldev->md.md_offset);
2966 sector = mdev->ldev->md.md_offset;
2968 if (drbd_md_sync_page_io(mdev, mdev->ldev, sector, WRITE)) {
2969 /* this was a try anyways ... */
2970 dev_err(DEV, "meta data update failed!\n");
2971 drbd_chk_io_error(mdev, 1, true);
2974 /* Update mdev->ldev->md.la_size_sect,
2975 * since we updated it on metadata. */
2976 mdev->ldev->md.la_size_sect = drbd_get_capacity(mdev->this_bdev);
2978 drbd_md_put_buffer(mdev);
2984 * drbd_md_read() - Reads in the meta data super block
2985 * @mdev: DRBD device.
2986 * @bdev: Device from which the meta data should be read in.
2988 * Return 0 (NO_ERROR) on success, and an enum drbd_ret_code in case
2989 * something goes wrong.
2991 int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev)
2993 struct meta_data_on_disk *buffer;
2995 int i, rv = NO_ERROR;
2997 if (!get_ldev_if_state(mdev, D_ATTACHING))
2998 return ERR_IO_MD_DISK;
3000 buffer = drbd_md_get_buffer(mdev);
3004 if (drbd_md_sync_page_io(mdev, bdev, bdev->md.md_offset, READ)) {
3005 /* NOTE: can't do normal error processing here as this is
3006 called BEFORE disk is attached */
3007 dev_err(DEV, "Error while reading metadata.\n");
3008 rv = ERR_IO_MD_DISK;
3012 magic = be32_to_cpu(buffer->magic);
3013 flags = be32_to_cpu(buffer->flags);
3014 if (magic == DRBD_MD_MAGIC_84_UNCLEAN ||
3015 (magic == DRBD_MD_MAGIC_08 && !(flags & MDF_AL_CLEAN))) {
3016 /* btw: that's Activity Log clean, not "all" clean. */
3017 dev_err(DEV, "Found unclean meta data. Did you \"drbdadm apply-al\"?\n");
3018 rv = ERR_MD_UNCLEAN;
3021 if (magic != DRBD_MD_MAGIC_08) {
3022 if (magic == DRBD_MD_MAGIC_07)
3023 dev_err(DEV, "Found old (0.7) meta data magic. Did you \"drbdadm create-md\"?\n");
3025 dev_err(DEV, "Meta data magic not found. Did you \"drbdadm create-md\"?\n");
3026 rv = ERR_MD_INVALID;
3029 if (be32_to_cpu(buffer->al_offset) != bdev->md.al_offset) {
3030 dev_err(DEV, "unexpected al_offset: %d (expected %d)\n",
3031 be32_to_cpu(buffer->al_offset), bdev->md.al_offset);
3032 rv = ERR_MD_INVALID;
3035 if (be32_to_cpu(buffer->bm_offset) != bdev->md.bm_offset) {
3036 dev_err(DEV, "unexpected bm_offset: %d (expected %d)\n",
3037 be32_to_cpu(buffer->bm_offset), bdev->md.bm_offset);
3038 rv = ERR_MD_INVALID;
3041 if (be32_to_cpu(buffer->md_size_sect) != bdev->md.md_size_sect) {
3042 dev_err(DEV, "unexpected md_size: %u (expected %u)\n",
3043 be32_to_cpu(buffer->md_size_sect), bdev->md.md_size_sect);
3044 rv = ERR_MD_INVALID;
3048 if (be32_to_cpu(buffer->bm_bytes_per_bit) != BM_BLOCK_SIZE) {
3049 dev_err(DEV, "unexpected bm_bytes_per_bit: %u (expected %u)\n",
3050 be32_to_cpu(buffer->bm_bytes_per_bit), BM_BLOCK_SIZE);
3051 rv = ERR_MD_INVALID;
3055 bdev->md.la_size_sect = be64_to_cpu(buffer->la_size);
3056 for (i = UI_CURRENT; i < UI_SIZE; i++)
3057 bdev->md.uuid[i] = be64_to_cpu(buffer->uuid[i]);
3058 bdev->md.flags = be32_to_cpu(buffer->flags);
3059 bdev->md.device_uuid = be64_to_cpu(buffer->device_uuid);
3061 spin_lock_irq(&mdev->tconn->req_lock);
3062 if (mdev->state.conn < C_CONNECTED) {
3064 peer = be32_to_cpu(buffer->la_peer_max_bio_size);
3065 peer = max_t(int, peer, DRBD_MAX_BIO_SIZE_SAFE);
3066 mdev->peer_max_bio_size = peer;
3068 spin_unlock_irq(&mdev->tconn->req_lock);
3071 drbd_md_put_buffer(mdev);
3079 * drbd_md_mark_dirty() - Mark meta data super block as dirty
3080 * @mdev: DRBD device.
3082 * Call this function if you change anything that should be written to
3083 * the meta-data super block. This function sets MD_DIRTY, and starts a
3084 * timer that ensures that within five seconds you have to call drbd_md_sync().
3087 void drbd_md_mark_dirty_(struct drbd_conf *mdev, unsigned int line, const char *func)
3089 if (!test_and_set_bit(MD_DIRTY, &mdev->flags)) {
3090 mod_timer(&mdev->md_sync_timer, jiffies + HZ);
3091 mdev->last_md_mark_dirty.line = line;
3092 mdev->last_md_mark_dirty.func = func;
3096 void drbd_md_mark_dirty(struct drbd_conf *mdev)
3098 if (!test_and_set_bit(MD_DIRTY, &mdev->flags))
3099 mod_timer(&mdev->md_sync_timer, jiffies + 5*HZ);
3103 static void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local)
3107 for (i = UI_HISTORY_START; i < UI_HISTORY_END; i++)
3108 mdev->ldev->md.uuid[i+1] = mdev->ldev->md.uuid[i];
3111 void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
3113 if (idx == UI_CURRENT) {
3114 if (mdev->state.role == R_PRIMARY)
3119 drbd_set_ed_uuid(mdev, val);
3122 mdev->ldev->md.uuid[idx] = val;
3123 drbd_md_mark_dirty(mdev);
3127 void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local)
3129 if (mdev->ldev->md.uuid[idx]) {
3130 drbd_uuid_move_history(mdev);
3131 mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[idx];
3133 _drbd_uuid_set(mdev, idx, val);
3137 * drbd_uuid_new_current() - Creates a new current UUID
3138 * @mdev: DRBD device.
3140 * Creates a new current UUID, and rotates the old current UUID into
3141 * the bitmap slot. Causes an incremental resync upon next connect.
3143 void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local)
3146 unsigned long long bm_uuid = mdev->ldev->md.uuid[UI_BITMAP];
3149 dev_warn(DEV, "bm UUID was already set: %llX\n", bm_uuid);
3151 mdev->ldev->md.uuid[UI_BITMAP] = mdev->ldev->md.uuid[UI_CURRENT];
3153 get_random_bytes(&val, sizeof(u64));
3154 _drbd_uuid_set(mdev, UI_CURRENT, val);
3155 drbd_print_uuids(mdev, "new current UUID");
3156 /* get it to stable storage _now_ */
3160 void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local)
3162 if (mdev->ldev->md.uuid[UI_BITMAP] == 0 && val == 0)
3166 drbd_uuid_move_history(mdev);
3167 mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP];
3168 mdev->ldev->md.uuid[UI_BITMAP] = 0;
3170 unsigned long long bm_uuid = mdev->ldev->md.uuid[UI_BITMAP];
3172 dev_warn(DEV, "bm UUID was already set: %llX\n", bm_uuid);
3174 mdev->ldev->md.uuid[UI_BITMAP] = val & ~((u64)1);
3176 drbd_md_mark_dirty(mdev);
3180 * drbd_bmio_set_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io()
3181 * @mdev: DRBD device.
3183 * Sets all bits in the bitmap and writes the whole bitmap to stable storage.
3185 int drbd_bmio_set_n_write(struct drbd_conf *mdev)
3189 if (get_ldev_if_state(mdev, D_ATTACHING)) {
3190 drbd_md_set_flag(mdev, MDF_FULL_SYNC);
3192 drbd_bm_set_all(mdev);
3194 rv = drbd_bm_write(mdev);
3197 drbd_md_clear_flag(mdev, MDF_FULL_SYNC);
3208 * drbd_bmio_clear_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io()
3209 * @mdev: DRBD device.
3211 * Clears all bits in the bitmap and writes the whole bitmap to stable storage.
3213 int drbd_bmio_clear_n_write(struct drbd_conf *mdev)
3217 drbd_resume_al(mdev);
3218 if (get_ldev_if_state(mdev, D_ATTACHING)) {
3219 drbd_bm_clear_all(mdev);
3220 rv = drbd_bm_write(mdev);
3227 static int w_bitmap_io(struct drbd_work *w, int unused)
3229 struct bm_io_work *work = container_of(w, struct bm_io_work, w);
3230 struct drbd_conf *mdev = w->mdev;
3233 D_ASSERT(atomic_read(&mdev->ap_bio_cnt) == 0);
3235 if (get_ldev(mdev)) {
3236 drbd_bm_lock(mdev, work->why, work->flags);
3237 rv = work->io_fn(mdev);
3238 drbd_bm_unlock(mdev);
3242 clear_bit_unlock(BITMAP_IO, &mdev->flags);
3243 wake_up(&mdev->misc_wait);
3246 work->done(mdev, rv);
3248 clear_bit(BITMAP_IO_QUEUED, &mdev->flags);
3255 void drbd_ldev_destroy(struct drbd_conf *mdev)
3257 lc_destroy(mdev->resync);
3258 mdev->resync = NULL;
3259 lc_destroy(mdev->act_log);
3260 mdev->act_log = NULL;
3262 drbd_free_bc(mdev->ldev);
3263 mdev->ldev = NULL;);
3265 clear_bit(GO_DISKLESS, &mdev->flags);
3268 static int w_go_diskless(struct drbd_work *w, int unused)
3270 struct drbd_conf *mdev = w->mdev;
3272 D_ASSERT(mdev->state.disk == D_FAILED);
3273 /* we cannot assert local_cnt == 0 here, as get_ldev_if_state will
3274 * inc/dec it frequently. Once we are D_DISKLESS, no one will touch
3275 * the protected members anymore, though, so once put_ldev reaches zero
3276 * again, it will be safe to free them. */
3277 drbd_force_state(mdev, NS(disk, D_DISKLESS));
3281 void drbd_go_diskless(struct drbd_conf *mdev)
3283 D_ASSERT(mdev->state.disk == D_FAILED);
3284 if (!test_and_set_bit(GO_DISKLESS, &mdev->flags))
3285 drbd_queue_work(&mdev->tconn->data.work, &mdev->go_diskless);
3289 * drbd_queue_bitmap_io() - Queues an IO operation on the whole bitmap
3290 * @mdev: DRBD device.
3291 * @io_fn: IO callback to be called when bitmap IO is possible
3292 * @done: callback to be called after the bitmap IO was performed
3293 * @why: Descriptive text of the reason for doing the IO
3295 * While IO on the bitmap happens we freeze application IO thus we ensure
3296 * that drbd_set_out_of_sync() can not be called. This function MAY ONLY be
3297 * called from worker context. It MUST NOT be used while a previous such
3298 * work is still pending!
3300 void drbd_queue_bitmap_io(struct drbd_conf *mdev,
3301 int (*io_fn)(struct drbd_conf *),
3302 void (*done)(struct drbd_conf *, int),
3303 char *why, enum bm_flag flags)
3305 D_ASSERT(current == mdev->tconn->worker.task);
3307 D_ASSERT(!test_bit(BITMAP_IO_QUEUED, &mdev->flags));
3308 D_ASSERT(!test_bit(BITMAP_IO, &mdev->flags));
3309 D_ASSERT(list_empty(&mdev->bm_io_work.w.list));
3310 if (mdev->bm_io_work.why)
3311 dev_err(DEV, "FIXME going to queue '%s' but '%s' still pending?\n",
3312 why, mdev->bm_io_work.why);
3314 mdev->bm_io_work.io_fn = io_fn;
3315 mdev->bm_io_work.done = done;
3316 mdev->bm_io_work.why = why;
3317 mdev->bm_io_work.flags = flags;
3319 spin_lock_irq(&mdev->tconn->req_lock);
3320 set_bit(BITMAP_IO, &mdev->flags);
3321 if (atomic_read(&mdev->ap_bio_cnt) == 0) {
3322 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
3323 drbd_queue_work(&mdev->tconn->data.work, &mdev->bm_io_work.w);
3325 spin_unlock_irq(&mdev->tconn->req_lock);
3329 * drbd_bitmap_io() - Does an IO operation on the whole bitmap
3330 * @mdev: DRBD device.
3331 * @io_fn: IO callback to be called when bitmap IO is possible
3332 * @why: Descriptive text of the reason for doing the IO
3334 * freezes application IO while that the actual IO operations runs. This
3335 * functions MAY NOT be called from worker context.
3337 int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *),
3338 char *why, enum bm_flag flags)
3342 D_ASSERT(current != mdev->tconn->worker.task);
3344 if ((flags & BM_LOCKED_SET_ALLOWED) == 0)
3345 drbd_suspend_io(mdev);
3347 drbd_bm_lock(mdev, why, flags);
3349 drbd_bm_unlock(mdev);
3351 if ((flags & BM_LOCKED_SET_ALLOWED) == 0)
3352 drbd_resume_io(mdev);
3357 void drbd_md_set_flag(struct drbd_conf *mdev, int flag) __must_hold(local)
3359 if ((mdev->ldev->md.flags & flag) != flag) {
3360 drbd_md_mark_dirty(mdev);
3361 mdev->ldev->md.flags |= flag;
3365 void drbd_md_clear_flag(struct drbd_conf *mdev, int flag) __must_hold(local)
3367 if ((mdev->ldev->md.flags & flag) != 0) {
3368 drbd_md_mark_dirty(mdev);
3369 mdev->ldev->md.flags &= ~flag;
3372 int drbd_md_test_flag(struct drbd_backing_dev *bdev, int flag)
3374 return (bdev->md.flags & flag) != 0;
3377 static void md_sync_timer_fn(unsigned long data)
3379 struct drbd_conf *mdev = (struct drbd_conf *) data;
3381 drbd_queue_work_front(&mdev->tconn->data.work, &mdev->md_sync_work);
3384 static int w_md_sync(struct drbd_work *w, int unused)
3386 struct drbd_conf *mdev = w->mdev;
3388 dev_warn(DEV, "md_sync_timer expired! Worker calls drbd_md_sync().\n");
3390 dev_warn(DEV, "last md_mark_dirty: %s:%u\n",
3391 mdev->last_md_mark_dirty.func, mdev->last_md_mark_dirty.line);
3397 const char *cmdname(enum drbd_packet cmd)
3399 /* THINK may need to become several global tables
3400 * when we want to support more than
3401 * one PRO_VERSION */
3402 static const char *cmdnames[] = {
3404 [P_DATA_REPLY] = "DataReply",
3405 [P_RS_DATA_REPLY] = "RSDataReply",
3406 [P_BARRIER] = "Barrier",
3407 [P_BITMAP] = "ReportBitMap",
3408 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
3409 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
3410 [P_UNPLUG_REMOTE] = "UnplugRemote",
3411 [P_DATA_REQUEST] = "DataRequest",
3412 [P_RS_DATA_REQUEST] = "RSDataRequest",
3413 [P_SYNC_PARAM] = "SyncParam",
3414 [P_SYNC_PARAM89] = "SyncParam89",
3415 [P_PROTOCOL] = "ReportProtocol",
3416 [P_UUIDS] = "ReportUUIDs",
3417 [P_SIZES] = "ReportSizes",
3418 [P_STATE] = "ReportState",
3419 [P_SYNC_UUID] = "ReportSyncUUID",
3420 [P_AUTH_CHALLENGE] = "AuthChallenge",
3421 [P_AUTH_RESPONSE] = "AuthResponse",
3423 [P_PING_ACK] = "PingAck",
3424 [P_RECV_ACK] = "RecvAck",
3425 [P_WRITE_ACK] = "WriteAck",
3426 [P_RS_WRITE_ACK] = "RSWriteAck",
3427 [P_DISCARD_WRITE] = "DiscardWrite",
3428 [P_NEG_ACK] = "NegAck",
3429 [P_NEG_DREPLY] = "NegDReply",
3430 [P_NEG_RS_DREPLY] = "NegRSDReply",
3431 [P_BARRIER_ACK] = "BarrierAck",
3432 [P_STATE_CHG_REQ] = "StateChgRequest",
3433 [P_STATE_CHG_REPLY] = "StateChgReply",
3434 [P_OV_REQUEST] = "OVRequest",
3435 [P_OV_REPLY] = "OVReply",
3436 [P_OV_RESULT] = "OVResult",
3437 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
3438 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
3439 [P_COMPRESSED_BITMAP] = "CBitmap",
3440 [P_DELAY_PROBE] = "DelayProbe",
3441 [P_OUT_OF_SYNC] = "OutOfSync",
3442 [P_RETRY_WRITE] = "RetryWrite",
3443 [P_RS_CANCEL] = "RSCancel",
3444 [P_CONN_ST_CHG_REQ] = "conn_st_chg_req",
3445 [P_CONN_ST_CHG_REPLY] = "conn_st_chg_reply",
3446 [P_RETRY_WRITE] = "retry_write",
3447 [P_PROTOCOL_UPDATE] = "protocol_update",
3449 /* enum drbd_packet, but not commands - obsoleted flags:
3455 /* too big for the array: 0xfffX */
3456 if (cmd == P_INITIAL_META)
3457 return "InitialMeta";
3458 if (cmd == P_INITIAL_DATA)
3459 return "InitialData";
3460 if (cmd == P_CONNECTION_FEATURES)
3461 return "ConnectionFeatures";
3462 if (cmd >= ARRAY_SIZE(cmdnames))
3464 return cmdnames[cmd];
3468 * drbd_wait_misc - wait for a request to make progress
3469 * @mdev: device associated with the request
3470 * @i: the struct drbd_interval embedded in struct drbd_request or
3471 * struct drbd_peer_request
3473 int drbd_wait_misc(struct drbd_conf *mdev, struct drbd_interval *i)
3475 struct net_conf *nc;
3480 nc = rcu_dereference(mdev->tconn->net_conf);
3485 timeout = nc->ko_count ? nc->timeout * HZ / 10 * nc->ko_count : MAX_SCHEDULE_TIMEOUT;
3488 /* Indicate to wake up mdev->misc_wait on progress. */
3490 prepare_to_wait(&mdev->misc_wait, &wait, TASK_INTERRUPTIBLE);
3491 spin_unlock_irq(&mdev->tconn->req_lock);
3492 timeout = schedule_timeout(timeout);
3493 finish_wait(&mdev->misc_wait, &wait);
3494 spin_lock_irq(&mdev->tconn->req_lock);
3495 if (!timeout || mdev->state.conn < C_CONNECTED)
3497 if (signal_pending(current))
3498 return -ERESTARTSYS;
3502 #ifdef CONFIG_DRBD_FAULT_INJECTION
3503 /* Fault insertion support including random number generator shamelessly
3504 * stolen from kernel/rcutorture.c */
3505 struct fault_random_state {
3506 unsigned long state;
3507 unsigned long count;
3510 #define FAULT_RANDOM_MULT 39916801 /* prime */
3511 #define FAULT_RANDOM_ADD 479001701 /* prime */
3512 #define FAULT_RANDOM_REFRESH 10000
3515 * Crude but fast random-number generator. Uses a linear congruential
3516 * generator, with occasional help from get_random_bytes().
3518 static unsigned long
3519 _drbd_fault_random(struct fault_random_state *rsp)
3523 if (!rsp->count--) {
3524 get_random_bytes(&refresh, sizeof(refresh));
3525 rsp->state += refresh;
3526 rsp->count = FAULT_RANDOM_REFRESH;
3528 rsp->state = rsp->state * FAULT_RANDOM_MULT + FAULT_RANDOM_ADD;
3529 return swahw32(rsp->state);
3533 _drbd_fault_str(unsigned int type) {
3534 static char *_faults[] = {
3535 [DRBD_FAULT_MD_WR] = "Meta-data write",
3536 [DRBD_FAULT_MD_RD] = "Meta-data read",
3537 [DRBD_FAULT_RS_WR] = "Resync write",
3538 [DRBD_FAULT_RS_RD] = "Resync read",
3539 [DRBD_FAULT_DT_WR] = "Data write",
3540 [DRBD_FAULT_DT_RD] = "Data read",
3541 [DRBD_FAULT_DT_RA] = "Data read ahead",
3542 [DRBD_FAULT_BM_ALLOC] = "BM allocation",
3543 [DRBD_FAULT_AL_EE] = "EE allocation",
3544 [DRBD_FAULT_RECEIVE] = "receive data corruption",
3547 return (type < DRBD_FAULT_MAX) ? _faults[type] : "**Unknown**";
3551 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type)
3553 static struct fault_random_state rrs = {0, 0};
3555 unsigned int ret = (
3557 ((1 << mdev_to_minor(mdev)) & fault_devs) != 0) &&
3558 (((_drbd_fault_random(&rrs) % 100) + 1) <= fault_rate));
3563 if (__ratelimit(&drbd_ratelimit_state))
3564 dev_warn(DEV, "***Simulating %s failure\n",
3565 _drbd_fault_str(type));
3572 const char *drbd_buildtag(void)
3574 /* DRBD built from external sources has here a reference to the
3575 git hash of the source code. */
3577 static char buildtag[38] = "\0uilt-in";
3579 if (buildtag[0] == 0) {
3580 #ifdef CONFIG_MODULES
3581 if (THIS_MODULE != NULL)
3582 sprintf(buildtag, "srcversion: %-24s", THIS_MODULE->srcversion);
3591 module_init(drbd_init)
3592 module_exit(drbd_cleanup)
3594 EXPORT_SYMBOL(drbd_conn_str);
3595 EXPORT_SYMBOL(drbd_role_str);
3596 EXPORT_SYMBOL(drbd_disk_str);
3597 EXPORT_SYMBOL(drbd_set_st_err_str);