1 // SPDX-License-Identifier: GPL-2.0-or-later
5 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
7 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
8 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
9 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
11 Thanks to Carter Burden, Bart Grantham and Gennadiy Nerubayev
12 from Logicworks, Inc. for making SDP replication support possible.
17 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
19 #include <linux/module.h>
20 #include <linux/jiffies.h>
21 #include <linux/drbd.h>
22 #include <linux/uaccess.h>
23 #include <asm/types.h>
25 #include <linux/ctype.h>
26 #include <linux/mutex.h>
28 #include <linux/file.h>
29 #include <linux/proc_fs.h>
30 #include <linux/init.h>
32 #include <linux/memcontrol.h>
33 #include <linux/mm_inline.h>
34 #include <linux/slab.h>
35 #include <linux/random.h>
36 #include <linux/reboot.h>
37 #include <linux/notifier.h>
38 #include <linux/kthread.h>
39 #include <linux/workqueue.h>
40 #define __KERNEL_SYSCALLS__
41 #include <linux/unistd.h>
42 #include <linux/vmalloc.h>
43 #include <linux/sched/signal.h>
45 #include <linux/drbd_limits.h>
47 #include "drbd_protocol.h"
48 #include "drbd_req.h" /* only for _req_mod in tl_release and tl_clear */
50 #include "drbd_debugfs.h"
52 static DEFINE_MUTEX(drbd_main_mutex);
53 static int drbd_open(struct block_device *bdev, fmode_t mode);
54 static void drbd_release(struct gendisk *gd, fmode_t mode);
55 static void md_sync_timer_fn(struct timer_list *t);
56 static int w_bitmap_io(struct drbd_work *w, int unused);
58 MODULE_AUTHOR("Philipp Reisner <phil@linbit.com>, "
59 "Lars Ellenberg <lars@linbit.com>");
60 MODULE_DESCRIPTION("drbd - Distributed Replicated Block Device v" REL_VERSION);
61 MODULE_VERSION(REL_VERSION);
62 MODULE_LICENSE("GPL");
63 MODULE_PARM_DESC(minor_count, "Approximate number of drbd devices ("
64 __stringify(DRBD_MINOR_COUNT_MIN) "-" __stringify(DRBD_MINOR_COUNT_MAX) ")");
65 MODULE_ALIAS_BLOCKDEV_MAJOR(DRBD_MAJOR);
67 #include <linux/moduleparam.h>
68 /* thanks to these macros, if compiled into the kernel (not-module),
69 * these become boot parameters (e.g., drbd.minor_count) */
71 #ifdef CONFIG_DRBD_FAULT_INJECTION
72 int drbd_enable_faults;
74 static int drbd_fault_count;
75 static int drbd_fault_devs;
76 /* bitmap of enabled faults */
77 module_param_named(enable_faults, drbd_enable_faults, int, 0664);
78 /* fault rate % value - applies to all enabled faults */
79 module_param_named(fault_rate, drbd_fault_rate, int, 0664);
80 /* count of faults inserted */
81 module_param_named(fault_count, drbd_fault_count, int, 0664);
82 /* bitmap of devices to insert faults on */
83 module_param_named(fault_devs, drbd_fault_devs, int, 0644);
86 /* module parameters we can keep static */
87 static bool drbd_allow_oos; /* allow_open_on_secondary */
88 static bool drbd_disable_sendpage;
89 MODULE_PARM_DESC(allow_oos, "DONT USE!");
90 module_param_named(allow_oos, drbd_allow_oos, bool, 0);
91 module_param_named(disable_sendpage, drbd_disable_sendpage, bool, 0644);
93 /* module parameters we share */
94 int drbd_proc_details; /* Detail level in proc drbd*/
95 module_param_named(proc_details, drbd_proc_details, int, 0644);
96 /* module parameters shared with defaults */
97 unsigned int drbd_minor_count = DRBD_MINOR_COUNT_DEF;
98 /* Module parameter for setting the user mode helper program
99 * to run. Default is /sbin/drbdadm */
100 char drbd_usermode_helper[80] = "/sbin/drbdadm";
101 module_param_named(minor_count, drbd_minor_count, uint, 0444);
102 module_param_string(usermode_helper, drbd_usermode_helper, sizeof(drbd_usermode_helper), 0644);
104 /* in 2.6.x, our device mapping and config info contains our virtual gendisks
105 * as member "struct gendisk *vdisk;"
107 struct idr drbd_devices;
108 struct list_head drbd_resources;
109 struct mutex resources_mutex;
111 struct kmem_cache *drbd_request_cache;
112 struct kmem_cache *drbd_ee_cache; /* peer requests */
113 struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
114 struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
115 mempool_t drbd_request_mempool;
116 mempool_t drbd_ee_mempool;
117 mempool_t drbd_md_io_page_pool;
118 struct bio_set drbd_md_io_bio_set;
119 struct bio_set drbd_io_bio_set;
121 /* I do not use a standard mempool, because:
122 1) I want to hand out the pre-allocated objects first.
123 2) I want to be able to interrupt sleeping allocation with a signal.
124 Note: This is a single linked list, the next pointer is the private
125 member of struct page.
127 struct page *drbd_pp_pool;
128 DEFINE_SPINLOCK(drbd_pp_lock);
130 wait_queue_head_t drbd_pp_wait;
132 DEFINE_RATELIMIT_STATE(drbd_ratelimit_state, 5 * HZ, 5);
134 static const struct block_device_operations drbd_ops = {
135 .owner = THIS_MODULE,
136 .submit_bio = drbd_submit_bio,
138 .release = drbd_release,
142 /* When checking with sparse, and this is an inline function, sparse will
143 give tons of false positives. When this is a real functions sparse works.
145 int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
149 atomic_inc(&device->local_cnt);
150 io_allowed = (device->state.disk >= mins);
152 if (atomic_dec_and_test(&device->local_cnt))
153 wake_up(&device->misc_wait);
161 * tl_release() - mark as BARRIER_ACKED all requests in the corresponding transfer log epoch
162 * @connection: DRBD connection.
163 * @barrier_nr: Expected identifier of the DRBD write barrier packet.
164 * @set_size: Expected number of requests before that barrier.
166 * In case the passed barrier_nr or set_size does not match the oldest
167 * epoch of not yet barrier-acked requests, this function will cause a
168 * termination of the connection.
170 void tl_release(struct drbd_connection *connection, unsigned int barrier_nr,
171 unsigned int set_size)
173 struct drbd_request *r;
174 struct drbd_request *req = NULL, *tmp = NULL;
175 int expect_epoch = 0;
178 spin_lock_irq(&connection->resource->req_lock);
180 /* find oldest not yet barrier-acked write request,
181 * count writes in its epoch. */
182 list_for_each_entry(r, &connection->transfer_log, tl_requests) {
183 const unsigned s = r->rq_state;
187 if (!(s & RQ_NET_MASK))
192 expect_epoch = req->epoch;
195 if (r->epoch != expect_epoch)
199 /* if (s & RQ_DONE): not expected */
200 /* if (!(s & RQ_NET_MASK)): not expected */
205 /* first some paranoia code */
207 drbd_err(connection, "BAD! BarrierAck #%u received, but no epoch in tl!?\n",
211 if (expect_epoch != barrier_nr) {
212 drbd_err(connection, "BAD! BarrierAck #%u received, expected #%u!\n",
213 barrier_nr, expect_epoch);
217 if (expect_size != set_size) {
218 drbd_err(connection, "BAD! BarrierAck #%u received with n_writes=%u, expected n_writes=%u!\n",
219 barrier_nr, set_size, expect_size);
223 /* Clean up list of requests processed during current epoch. */
224 /* this extra list walk restart is paranoia,
225 * to catch requests being barrier-acked "unexpectedly".
226 * It usually should find the same req again, or some READ preceding it. */
227 list_for_each_entry(req, &connection->transfer_log, tl_requests)
228 if (req->epoch == expect_epoch) {
232 req = list_prepare_entry(tmp, &connection->transfer_log, tl_requests);
233 list_for_each_entry_safe_from(req, r, &connection->transfer_log, tl_requests) {
234 if (req->epoch != expect_epoch)
236 _req_mod(req, BARRIER_ACKED);
238 spin_unlock_irq(&connection->resource->req_lock);
243 spin_unlock_irq(&connection->resource->req_lock);
244 conn_request_state(connection, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
249 * _tl_restart() - Walks the transfer log, and applies an action to all requests
250 * @connection: DRBD connection to operate on.
251 * @what: The action/event to perform with all request objects
253 * @what might be one of CONNECTION_LOST_WHILE_PENDING, RESEND, FAIL_FROZEN_DISK_IO,
254 * RESTART_FROZEN_DISK_IO.
256 /* must hold resource->req_lock */
257 void _tl_restart(struct drbd_connection *connection, enum drbd_req_event what)
259 struct drbd_request *req, *r;
261 list_for_each_entry_safe(req, r, &connection->transfer_log, tl_requests)
265 void tl_restart(struct drbd_connection *connection, enum drbd_req_event what)
267 spin_lock_irq(&connection->resource->req_lock);
268 _tl_restart(connection, what);
269 spin_unlock_irq(&connection->resource->req_lock);
273 * tl_clear() - Clears all requests and &struct drbd_tl_epoch objects out of the TL
274 * @connection: DRBD connection.
276 * This is called after the connection to the peer was lost. The storage covered
277 * by the requests on the transfer gets marked as our of sync. Called from the
278 * receiver thread and the worker thread.
280 void tl_clear(struct drbd_connection *connection)
282 tl_restart(connection, CONNECTION_LOST_WHILE_PENDING);
286 * tl_abort_disk_io() - Abort disk I/O for all requests for a certain device in the TL
287 * @device: DRBD device.
289 void tl_abort_disk_io(struct drbd_device *device)
291 struct drbd_connection *connection = first_peer_device(device)->connection;
292 struct drbd_request *req, *r;
294 spin_lock_irq(&connection->resource->req_lock);
295 list_for_each_entry_safe(req, r, &connection->transfer_log, tl_requests) {
296 if (!(req->rq_state & RQ_LOCAL_PENDING))
298 if (req->device != device)
300 _req_mod(req, ABORT_DISK_IO);
302 spin_unlock_irq(&connection->resource->req_lock);
305 static int drbd_thread_setup(void *arg)
307 struct drbd_thread *thi = (struct drbd_thread *) arg;
308 struct drbd_resource *resource = thi->resource;
312 snprintf(current->comm, sizeof(current->comm), "drbd_%c_%s",
316 allow_kernel_signal(DRBD_SIGKILL);
317 allow_kernel_signal(SIGXCPU);
319 retval = thi->function(thi);
321 spin_lock_irqsave(&thi->t_lock, flags);
323 /* if the receiver has been "EXITING", the last thing it did
324 * was set the conn state to "StandAlone",
325 * if now a re-connect request comes in, conn state goes C_UNCONNECTED,
326 * and receiver thread will be "started".
327 * drbd_thread_start needs to set "RESTARTING" in that case.
328 * t_state check and assignment needs to be within the same spinlock,
329 * so either thread_start sees EXITING, and can remap to RESTARTING,
330 * or thread_start see NONE, and can proceed as normal.
333 if (thi->t_state == RESTARTING) {
334 drbd_info(resource, "Restarting %s thread\n", thi->name);
335 thi->t_state = RUNNING;
336 spin_unlock_irqrestore(&thi->t_lock, flags);
343 complete_all(&thi->stop);
344 spin_unlock_irqrestore(&thi->t_lock, flags);
346 drbd_info(resource, "Terminating %s\n", current->comm);
348 /* Release mod reference taken when thread was started */
351 kref_put(&thi->connection->kref, drbd_destroy_connection);
352 kref_put(&resource->kref, drbd_destroy_resource);
353 module_put(THIS_MODULE);
357 static void drbd_thread_init(struct drbd_resource *resource, struct drbd_thread *thi,
358 int (*func) (struct drbd_thread *), const char *name)
360 spin_lock_init(&thi->t_lock);
363 thi->function = func;
364 thi->resource = resource;
365 thi->connection = NULL;
369 int drbd_thread_start(struct drbd_thread *thi)
371 struct drbd_resource *resource = thi->resource;
372 struct task_struct *nt;
375 /* is used from state engine doing drbd_thread_stop_nowait,
376 * while holding the req lock irqsave */
377 spin_lock_irqsave(&thi->t_lock, flags);
379 switch (thi->t_state) {
381 drbd_info(resource, "Starting %s thread (from %s [%d])\n",
382 thi->name, current->comm, current->pid);
384 /* Get ref on module for thread - this is released when thread exits */
385 if (!try_module_get(THIS_MODULE)) {
386 drbd_err(resource, "Failed to get module reference in drbd_thread_start\n");
387 spin_unlock_irqrestore(&thi->t_lock, flags);
391 kref_get(&resource->kref);
393 kref_get(&thi->connection->kref);
395 init_completion(&thi->stop);
396 thi->reset_cpu_mask = 1;
397 thi->t_state = RUNNING;
398 spin_unlock_irqrestore(&thi->t_lock, flags);
399 flush_signals(current); /* otherw. may get -ERESTARTNOINTR */
401 nt = kthread_create(drbd_thread_setup, (void *) thi,
402 "drbd_%c_%s", thi->name[0], thi->resource->name);
405 drbd_err(resource, "Couldn't start thread\n");
408 kref_put(&thi->connection->kref, drbd_destroy_connection);
409 kref_put(&resource->kref, drbd_destroy_resource);
410 module_put(THIS_MODULE);
413 spin_lock_irqsave(&thi->t_lock, flags);
415 thi->t_state = RUNNING;
416 spin_unlock_irqrestore(&thi->t_lock, flags);
420 thi->t_state = RESTARTING;
421 drbd_info(resource, "Restarting %s thread (from %s [%d])\n",
422 thi->name, current->comm, current->pid);
427 spin_unlock_irqrestore(&thi->t_lock, flags);
435 void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait)
439 enum drbd_thread_state ns = restart ? RESTARTING : EXITING;
441 /* may be called from state engine, holding the req lock irqsave */
442 spin_lock_irqsave(&thi->t_lock, flags);
444 if (thi->t_state == NONE) {
445 spin_unlock_irqrestore(&thi->t_lock, flags);
447 drbd_thread_start(thi);
451 if (thi->t_state != ns) {
452 if (thi->task == NULL) {
453 spin_unlock_irqrestore(&thi->t_lock, flags);
459 init_completion(&thi->stop);
460 if (thi->task != current)
461 send_sig(DRBD_SIGKILL, thi->task, 1);
464 spin_unlock_irqrestore(&thi->t_lock, flags);
467 wait_for_completion(&thi->stop);
470 int conn_lowest_minor(struct drbd_connection *connection)
472 struct drbd_peer_device *peer_device;
473 int vnr = 0, minor = -1;
476 peer_device = idr_get_next(&connection->peer_devices, &vnr);
478 minor = device_to_minor(peer_device->device);
486 * drbd_calc_cpu_mask() - Generate CPU masks, spread over all CPUs
488 * Forces all threads of a resource onto the same CPU. This is beneficial for
489 * DRBD's performance. May be overwritten by user's configuration.
491 static void drbd_calc_cpu_mask(cpumask_var_t *cpu_mask)
493 unsigned int *resources_per_cpu, min_index = ~0;
495 resources_per_cpu = kcalloc(nr_cpu_ids, sizeof(*resources_per_cpu),
497 if (resources_per_cpu) {
498 struct drbd_resource *resource;
499 unsigned int cpu, min = ~0;
502 for_each_resource_rcu(resource, &drbd_resources) {
503 for_each_cpu(cpu, resource->cpu_mask)
504 resources_per_cpu[cpu]++;
507 for_each_online_cpu(cpu) {
508 if (resources_per_cpu[cpu] < min) {
509 min = resources_per_cpu[cpu];
513 kfree(resources_per_cpu);
515 if (min_index == ~0) {
516 cpumask_setall(*cpu_mask);
519 cpumask_set_cpu(min_index, *cpu_mask);
523 * drbd_thread_current_set_cpu() - modifies the cpu mask of the _current_ thread
524 * @thi: drbd_thread object
526 * call in the "main loop" of _all_ threads, no need for any mutex, current won't die
529 void drbd_thread_current_set_cpu(struct drbd_thread *thi)
531 struct drbd_resource *resource = thi->resource;
532 struct task_struct *p = current;
534 if (!thi->reset_cpu_mask)
536 thi->reset_cpu_mask = 0;
537 set_cpus_allowed_ptr(p, resource->cpu_mask);
540 #define drbd_calc_cpu_mask(A) ({})
544 * drbd_header_size - size of a packet header
546 * The header size is a multiple of 8, so any payload following the header is
547 * word aligned on 64-bit architectures. (The bitmap send and receive code
550 unsigned int drbd_header_size(struct drbd_connection *connection)
552 if (connection->agreed_pro_version >= 100) {
553 BUILD_BUG_ON(!IS_ALIGNED(sizeof(struct p_header100), 8));
554 return sizeof(struct p_header100);
556 BUILD_BUG_ON(sizeof(struct p_header80) !=
557 sizeof(struct p_header95));
558 BUILD_BUG_ON(!IS_ALIGNED(sizeof(struct p_header80), 8));
559 return sizeof(struct p_header80);
563 static unsigned int prepare_header80(struct p_header80 *h, enum drbd_packet cmd, int size)
565 h->magic = cpu_to_be32(DRBD_MAGIC);
566 h->command = cpu_to_be16(cmd);
567 h->length = cpu_to_be16(size);
568 return sizeof(struct p_header80);
571 static unsigned int prepare_header95(struct p_header95 *h, enum drbd_packet cmd, int size)
573 h->magic = cpu_to_be16(DRBD_MAGIC_BIG);
574 h->command = cpu_to_be16(cmd);
575 h->length = cpu_to_be32(size);
576 return sizeof(struct p_header95);
579 static unsigned int prepare_header100(struct p_header100 *h, enum drbd_packet cmd,
582 h->magic = cpu_to_be32(DRBD_MAGIC_100);
583 h->volume = cpu_to_be16(vnr);
584 h->command = cpu_to_be16(cmd);
585 h->length = cpu_to_be32(size);
587 return sizeof(struct p_header100);
590 static unsigned int prepare_header(struct drbd_connection *connection, int vnr,
591 void *buffer, enum drbd_packet cmd, int size)
593 if (connection->agreed_pro_version >= 100)
594 return prepare_header100(buffer, cmd, size, vnr);
595 else if (connection->agreed_pro_version >= 95 &&
596 size > DRBD_MAX_SIZE_H80_PACKET)
597 return prepare_header95(buffer, cmd, size);
599 return prepare_header80(buffer, cmd, size);
602 static void *__conn_prepare_command(struct drbd_connection *connection,
603 struct drbd_socket *sock)
607 return sock->sbuf + drbd_header_size(connection);
610 void *conn_prepare_command(struct drbd_connection *connection, struct drbd_socket *sock)
614 mutex_lock(&sock->mutex);
615 p = __conn_prepare_command(connection, sock);
617 mutex_unlock(&sock->mutex);
622 void *drbd_prepare_command(struct drbd_peer_device *peer_device, struct drbd_socket *sock)
624 return conn_prepare_command(peer_device->connection, sock);
627 static int __send_command(struct drbd_connection *connection, int vnr,
628 struct drbd_socket *sock, enum drbd_packet cmd,
629 unsigned int header_size, void *data,
636 * Called with @data == NULL and the size of the data blocks in @size
637 * for commands that send data blocks. For those commands, omit the
638 * MSG_MORE flag: this will increase the likelihood that data blocks
639 * which are page aligned on the sender will end up page aligned on the
642 msg_flags = data ? MSG_MORE : 0;
644 header_size += prepare_header(connection, vnr, sock->sbuf, cmd,
646 err = drbd_send_all(connection, sock->socket, sock->sbuf, header_size,
649 err = drbd_send_all(connection, sock->socket, data, size, 0);
650 /* DRBD protocol "pings" are latency critical.
651 * This is supposed to trigger tcp_push_pending_frames() */
652 if (!err && (cmd == P_PING || cmd == P_PING_ACK))
653 tcp_sock_set_nodelay(sock->socket->sk);
658 static int __conn_send_command(struct drbd_connection *connection, struct drbd_socket *sock,
659 enum drbd_packet cmd, unsigned int header_size,
660 void *data, unsigned int size)
662 return __send_command(connection, 0, sock, cmd, header_size, data, size);
665 int conn_send_command(struct drbd_connection *connection, struct drbd_socket *sock,
666 enum drbd_packet cmd, unsigned int header_size,
667 void *data, unsigned int size)
671 err = __conn_send_command(connection, sock, cmd, header_size, data, size);
672 mutex_unlock(&sock->mutex);
676 int drbd_send_command(struct drbd_peer_device *peer_device, struct drbd_socket *sock,
677 enum drbd_packet cmd, unsigned int header_size,
678 void *data, unsigned int size)
682 err = __send_command(peer_device->connection, peer_device->device->vnr,
683 sock, cmd, header_size, data, size);
684 mutex_unlock(&sock->mutex);
688 int drbd_send_ping(struct drbd_connection *connection)
690 struct drbd_socket *sock;
692 sock = &connection->meta;
693 if (!conn_prepare_command(connection, sock))
695 return conn_send_command(connection, sock, P_PING, 0, NULL, 0);
698 int drbd_send_ping_ack(struct drbd_connection *connection)
700 struct drbd_socket *sock;
702 sock = &connection->meta;
703 if (!conn_prepare_command(connection, sock))
705 return conn_send_command(connection, sock, P_PING_ACK, 0, NULL, 0);
708 int drbd_send_sync_param(struct drbd_peer_device *peer_device)
710 struct drbd_socket *sock;
711 struct p_rs_param_95 *p;
713 const int apv = peer_device->connection->agreed_pro_version;
714 enum drbd_packet cmd;
716 struct disk_conf *dc;
718 sock = &peer_device->connection->data;
719 p = drbd_prepare_command(peer_device, sock);
724 nc = rcu_dereference(peer_device->connection->net_conf);
726 size = apv <= 87 ? sizeof(struct p_rs_param)
727 : apv == 88 ? sizeof(struct p_rs_param)
728 + strlen(nc->verify_alg) + 1
729 : apv <= 94 ? sizeof(struct p_rs_param_89)
730 : /* apv >= 95 */ sizeof(struct p_rs_param_95);
732 cmd = apv >= 89 ? P_SYNC_PARAM89 : P_SYNC_PARAM;
734 /* initialize verify_alg and csums_alg */
735 BUILD_BUG_ON(sizeof(p->algs) != 2 * SHARED_SECRET_MAX);
736 memset(&p->algs, 0, sizeof(p->algs));
738 if (get_ldev(peer_device->device)) {
739 dc = rcu_dereference(peer_device->device->ldev->disk_conf);
740 p->resync_rate = cpu_to_be32(dc->resync_rate);
741 p->c_plan_ahead = cpu_to_be32(dc->c_plan_ahead);
742 p->c_delay_target = cpu_to_be32(dc->c_delay_target);
743 p->c_fill_target = cpu_to_be32(dc->c_fill_target);
744 p->c_max_rate = cpu_to_be32(dc->c_max_rate);
745 put_ldev(peer_device->device);
747 p->resync_rate = cpu_to_be32(DRBD_RESYNC_RATE_DEF);
748 p->c_plan_ahead = cpu_to_be32(DRBD_C_PLAN_AHEAD_DEF);
749 p->c_delay_target = cpu_to_be32(DRBD_C_DELAY_TARGET_DEF);
750 p->c_fill_target = cpu_to_be32(DRBD_C_FILL_TARGET_DEF);
751 p->c_max_rate = cpu_to_be32(DRBD_C_MAX_RATE_DEF);
755 strcpy(p->verify_alg, nc->verify_alg);
757 strcpy(p->csums_alg, nc->csums_alg);
760 return drbd_send_command(peer_device, sock, cmd, size, NULL, 0);
763 int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd)
765 struct drbd_socket *sock;
766 struct p_protocol *p;
770 sock = &connection->data;
771 p = __conn_prepare_command(connection, sock);
776 nc = rcu_dereference(connection->net_conf);
778 if (nc->tentative && connection->agreed_pro_version < 92) {
780 drbd_err(connection, "--dry-run is not supported by peer");
785 if (connection->agreed_pro_version >= 87)
786 size += strlen(nc->integrity_alg) + 1;
788 p->protocol = cpu_to_be32(nc->wire_protocol);
789 p->after_sb_0p = cpu_to_be32(nc->after_sb_0p);
790 p->after_sb_1p = cpu_to_be32(nc->after_sb_1p);
791 p->after_sb_2p = cpu_to_be32(nc->after_sb_2p);
792 p->two_primaries = cpu_to_be32(nc->two_primaries);
794 if (nc->discard_my_data)
795 cf |= CF_DISCARD_MY_DATA;
798 p->conn_flags = cpu_to_be32(cf);
800 if (connection->agreed_pro_version >= 87)
801 strcpy(p->integrity_alg, nc->integrity_alg);
804 return __conn_send_command(connection, sock, cmd, size, NULL, 0);
807 int drbd_send_protocol(struct drbd_connection *connection)
811 mutex_lock(&connection->data.mutex);
812 err = __drbd_send_protocol(connection, P_PROTOCOL);
813 mutex_unlock(&connection->data.mutex);
818 static int _drbd_send_uuids(struct drbd_peer_device *peer_device, u64 uuid_flags)
820 struct drbd_device *device = peer_device->device;
821 struct drbd_socket *sock;
825 if (!get_ldev_if_state(device, D_NEGOTIATING))
828 sock = &peer_device->connection->data;
829 p = drbd_prepare_command(peer_device, sock);
834 spin_lock_irq(&device->ldev->md.uuid_lock);
835 for (i = UI_CURRENT; i < UI_SIZE; i++)
836 p->uuid[i] = cpu_to_be64(device->ldev->md.uuid[i]);
837 spin_unlock_irq(&device->ldev->md.uuid_lock);
839 device->comm_bm_set = drbd_bm_total_weight(device);
840 p->uuid[UI_SIZE] = cpu_to_be64(device->comm_bm_set);
842 uuid_flags |= rcu_dereference(peer_device->connection->net_conf)->discard_my_data ? 1 : 0;
844 uuid_flags |= test_bit(CRASHED_PRIMARY, &device->flags) ? 2 : 0;
845 uuid_flags |= device->new_state_tmp.disk == D_INCONSISTENT ? 4 : 0;
846 p->uuid[UI_FLAGS] = cpu_to_be64(uuid_flags);
849 return drbd_send_command(peer_device, sock, P_UUIDS, sizeof(*p), NULL, 0);
852 int drbd_send_uuids(struct drbd_peer_device *peer_device)
854 return _drbd_send_uuids(peer_device, 0);
857 int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *peer_device)
859 return _drbd_send_uuids(peer_device, 8);
862 void drbd_print_uuids(struct drbd_device *device, const char *text)
864 if (get_ldev_if_state(device, D_NEGOTIATING)) {
865 u64 *uuid = device->ldev->md.uuid;
866 drbd_info(device, "%s %016llX:%016llX:%016llX:%016llX\n",
868 (unsigned long long)uuid[UI_CURRENT],
869 (unsigned long long)uuid[UI_BITMAP],
870 (unsigned long long)uuid[UI_HISTORY_START],
871 (unsigned long long)uuid[UI_HISTORY_END]);
874 drbd_info(device, "%s effective data uuid: %016llX\n",
876 (unsigned long long)device->ed_uuid);
880 void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *peer_device)
882 struct drbd_device *device = peer_device->device;
883 struct drbd_socket *sock;
887 D_ASSERT(device, device->state.disk == D_UP_TO_DATE);
889 uuid = device->ldev->md.uuid[UI_BITMAP];
890 if (uuid && uuid != UUID_JUST_CREATED)
891 uuid = uuid + UUID_NEW_BM_OFFSET;
893 get_random_bytes(&uuid, sizeof(u64));
894 drbd_uuid_set(device, UI_BITMAP, uuid);
895 drbd_print_uuids(device, "updated sync UUID");
896 drbd_md_sync(device);
898 sock = &peer_device->connection->data;
899 p = drbd_prepare_command(peer_device, sock);
901 p->uuid = cpu_to_be64(uuid);
902 drbd_send_command(peer_device, sock, P_SYNC_UUID, sizeof(*p), NULL, 0);
906 int drbd_send_sizes(struct drbd_peer_device *peer_device, int trigger_reply, enum dds_flags flags)
908 struct drbd_device *device = peer_device->device;
909 struct drbd_socket *sock;
911 sector_t d_size, u_size;
913 unsigned int max_bio_size;
914 unsigned int packet_size;
916 sock = &peer_device->connection->data;
917 p = drbd_prepare_command(peer_device, sock);
921 packet_size = sizeof(*p);
922 if (peer_device->connection->agreed_features & DRBD_FF_WSAME)
923 packet_size += sizeof(p->qlim[0]);
925 memset(p, 0, packet_size);
926 if (get_ldev_if_state(device, D_NEGOTIATING)) {
927 struct block_device *bdev = device->ldev->backing_bdev;
928 struct request_queue *q = bdev_get_queue(bdev);
930 d_size = drbd_get_max_capacity(device->ldev);
932 u_size = rcu_dereference(device->ldev->disk_conf)->disk_size;
934 q_order_type = drbd_queue_order_type(device);
935 max_bio_size = queue_max_hw_sectors(q) << 9;
936 max_bio_size = min(max_bio_size, DRBD_MAX_BIO_SIZE);
937 p->qlim->physical_block_size =
938 cpu_to_be32(bdev_physical_block_size(bdev));
939 p->qlim->logical_block_size =
940 cpu_to_be32(bdev_logical_block_size(bdev));
941 p->qlim->alignment_offset =
942 cpu_to_be32(bdev_alignment_offset(bdev));
943 p->qlim->io_min = cpu_to_be32(bdev_io_min(bdev));
944 p->qlim->io_opt = cpu_to_be32(bdev_io_opt(bdev));
945 p->qlim->discard_enabled = !!bdev_max_discard_sectors(bdev);
948 struct request_queue *q = device->rq_queue;
950 p->qlim->physical_block_size =
951 cpu_to_be32(queue_physical_block_size(q));
952 p->qlim->logical_block_size =
953 cpu_to_be32(queue_logical_block_size(q));
954 p->qlim->alignment_offset = 0;
955 p->qlim->io_min = cpu_to_be32(queue_io_min(q));
956 p->qlim->io_opt = cpu_to_be32(queue_io_opt(q));
957 p->qlim->discard_enabled = 0;
961 q_order_type = QUEUE_ORDERED_NONE;
962 max_bio_size = DRBD_MAX_BIO_SIZE; /* ... multiple BIOs per peer_request */
965 if (peer_device->connection->agreed_pro_version <= 94)
966 max_bio_size = min(max_bio_size, DRBD_MAX_SIZE_H80_PACKET);
967 else if (peer_device->connection->agreed_pro_version < 100)
968 max_bio_size = min(max_bio_size, DRBD_MAX_BIO_SIZE_P95);
970 p->d_size = cpu_to_be64(d_size);
971 p->u_size = cpu_to_be64(u_size);
975 p->c_size = cpu_to_be64(get_capacity(device->vdisk));
976 p->max_bio_size = cpu_to_be32(max_bio_size);
977 p->queue_order_type = cpu_to_be16(q_order_type);
978 p->dds_flags = cpu_to_be16(flags);
980 return drbd_send_command(peer_device, sock, P_SIZES, packet_size, NULL, 0);
984 * drbd_send_current_state() - Sends the drbd state to the peer
985 * @peer_device: DRBD peer device.
987 int drbd_send_current_state(struct drbd_peer_device *peer_device)
989 struct drbd_socket *sock;
992 sock = &peer_device->connection->data;
993 p = drbd_prepare_command(peer_device, sock);
996 p->state = cpu_to_be32(peer_device->device->state.i); /* Within the send mutex */
997 return drbd_send_command(peer_device, sock, P_STATE, sizeof(*p), NULL, 0);
1001 * drbd_send_state() - After a state change, sends the new state to the peer
1002 * @peer_device: DRBD peer device.
1003 * @state: the state to send, not necessarily the current state.
1005 * Each state change queues an "after_state_ch" work, which will eventually
1006 * send the resulting new state to the peer. If more state changes happen
1007 * between queuing and processing of the after_state_ch work, we still
1008 * want to send each intermediary state in the order it occurred.
1010 int drbd_send_state(struct drbd_peer_device *peer_device, union drbd_state state)
1012 struct drbd_socket *sock;
1015 sock = &peer_device->connection->data;
1016 p = drbd_prepare_command(peer_device, sock);
1019 p->state = cpu_to_be32(state.i); /* Within the send mutex */
1020 return drbd_send_command(peer_device, sock, P_STATE, sizeof(*p), NULL, 0);
1023 int drbd_send_state_req(struct drbd_peer_device *peer_device, union drbd_state mask, union drbd_state val)
1025 struct drbd_socket *sock;
1026 struct p_req_state *p;
1028 sock = &peer_device->connection->data;
1029 p = drbd_prepare_command(peer_device, sock);
1032 p->mask = cpu_to_be32(mask.i);
1033 p->val = cpu_to_be32(val.i);
1034 return drbd_send_command(peer_device, sock, P_STATE_CHG_REQ, sizeof(*p), NULL, 0);
1037 int conn_send_state_req(struct drbd_connection *connection, union drbd_state mask, union drbd_state val)
1039 enum drbd_packet cmd;
1040 struct drbd_socket *sock;
1041 struct p_req_state *p;
1043 cmd = connection->agreed_pro_version < 100 ? P_STATE_CHG_REQ : P_CONN_ST_CHG_REQ;
1044 sock = &connection->data;
1045 p = conn_prepare_command(connection, sock);
1048 p->mask = cpu_to_be32(mask.i);
1049 p->val = cpu_to_be32(val.i);
1050 return conn_send_command(connection, sock, cmd, sizeof(*p), NULL, 0);
1053 void drbd_send_sr_reply(struct drbd_peer_device *peer_device, enum drbd_state_rv retcode)
1055 struct drbd_socket *sock;
1056 struct p_req_state_reply *p;
1058 sock = &peer_device->connection->meta;
1059 p = drbd_prepare_command(peer_device, sock);
1061 p->retcode = cpu_to_be32(retcode);
1062 drbd_send_command(peer_device, sock, P_STATE_CHG_REPLY, sizeof(*p), NULL, 0);
1066 void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode)
1068 struct drbd_socket *sock;
1069 struct p_req_state_reply *p;
1070 enum drbd_packet cmd = connection->agreed_pro_version < 100 ? P_STATE_CHG_REPLY : P_CONN_ST_CHG_REPLY;
1072 sock = &connection->meta;
1073 p = conn_prepare_command(connection, sock);
1075 p->retcode = cpu_to_be32(retcode);
1076 conn_send_command(connection, sock, cmd, sizeof(*p), NULL, 0);
1080 static void dcbp_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
1082 BUG_ON(code & ~0xf);
1083 p->encoding = (p->encoding & ~0xf) | code;
1086 static void dcbp_set_start(struct p_compressed_bm *p, int set)
1088 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
1091 static void dcbp_set_pad_bits(struct p_compressed_bm *p, int n)
1094 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
1097 static int fill_bitmap_rle_bits(struct drbd_device *device,
1098 struct p_compressed_bm *p,
1100 struct bm_xfer_ctx *c)
1102 struct bitstream bs;
1103 unsigned long plain_bits;
1110 /* may we use this feature? */
1112 use_rle = rcu_dereference(first_peer_device(device)->connection->net_conf)->use_rle;
1114 if (!use_rle || first_peer_device(device)->connection->agreed_pro_version < 90)
1117 if (c->bit_offset >= c->bm_bits)
1118 return 0; /* nothing to do. */
1120 /* use at most thus many bytes */
1121 bitstream_init(&bs, p->code, size, 0);
1122 memset(p->code, 0, size);
1123 /* plain bits covered in this code string */
1126 /* p->encoding & 0x80 stores whether the first run length is set.
1127 * bit offset is implicit.
1128 * start with toggle == 2 to be able to tell the first iteration */
1131 /* see how much plain bits we can stuff into one packet
1132 * using RLE and VLI. */
1134 tmp = (toggle == 0) ? _drbd_bm_find_next_zero(device, c->bit_offset)
1135 : _drbd_bm_find_next(device, c->bit_offset);
1138 rl = tmp - c->bit_offset;
1140 if (toggle == 2) { /* first iteration */
1142 /* the first checked bit was set,
1143 * store start value, */
1144 dcbp_set_start(p, 1);
1145 /* but skip encoding of zero run length */
1149 dcbp_set_start(p, 0);
1152 /* paranoia: catch zero runlength.
1153 * can only happen if bitmap is modified while we scan it. */
1155 drbd_err(device, "unexpected zero runlength while encoding bitmap "
1156 "t:%u bo:%lu\n", toggle, c->bit_offset);
1160 bits = vli_encode_bits(&bs, rl);
1161 if (bits == -ENOBUFS) /* buffer full */
1164 drbd_err(device, "error while encoding bitmap: %d\n", bits);
1170 c->bit_offset = tmp;
1171 } while (c->bit_offset < c->bm_bits);
1173 len = bs.cur.b - p->code + !!bs.cur.bit;
1175 if (plain_bits < (len << 3)) {
1176 /* incompressible with this method.
1177 * we need to rewind both word and bit position. */
1178 c->bit_offset -= plain_bits;
1179 bm_xfer_ctx_bit_to_word_offset(c);
1180 c->bit_offset = c->word_offset * BITS_PER_LONG;
1184 /* RLE + VLI was able to compress it just fine.
1185 * update c->word_offset. */
1186 bm_xfer_ctx_bit_to_word_offset(c);
1188 /* store pad_bits */
1189 dcbp_set_pad_bits(p, (8 - bs.cur.bit) & 0x7);
1195 * send_bitmap_rle_or_plain
1197 * Return 0 when done, 1 when another iteration is needed, and a negative error
1198 * code upon failure.
1201 send_bitmap_rle_or_plain(struct drbd_device *device, struct bm_xfer_ctx *c)
1203 struct drbd_socket *sock = &first_peer_device(device)->connection->data;
1204 unsigned int header_size = drbd_header_size(first_peer_device(device)->connection);
1205 struct p_compressed_bm *p = sock->sbuf + header_size;
1208 len = fill_bitmap_rle_bits(device, p,
1209 DRBD_SOCKET_BUFFER_SIZE - header_size - sizeof(*p), c);
1214 dcbp_set_code(p, RLE_VLI_Bits);
1215 err = __send_command(first_peer_device(device)->connection, device->vnr, sock,
1216 P_COMPRESSED_BITMAP, sizeof(*p) + len,
1219 c->bytes[0] += header_size + sizeof(*p) + len;
1221 if (c->bit_offset >= c->bm_bits)
1224 /* was not compressible.
1225 * send a buffer full of plain text bits instead. */
1226 unsigned int data_size;
1227 unsigned long num_words;
1228 unsigned long *p = sock->sbuf + header_size;
1230 data_size = DRBD_SOCKET_BUFFER_SIZE - header_size;
1231 num_words = min_t(size_t, data_size / sizeof(*p),
1232 c->bm_words - c->word_offset);
1233 len = num_words * sizeof(*p);
1235 drbd_bm_get_lel(device, c->word_offset, num_words, p);
1236 err = __send_command(first_peer_device(device)->connection, device->vnr, sock, P_BITMAP, len, NULL, 0);
1237 c->word_offset += num_words;
1238 c->bit_offset = c->word_offset * BITS_PER_LONG;
1241 c->bytes[1] += header_size + len;
1243 if (c->bit_offset > c->bm_bits)
1244 c->bit_offset = c->bm_bits;
1248 INFO_bm_xfer_stats(device, "send", c);
1256 /* See the comment at receive_bitmap() */
1257 static int _drbd_send_bitmap(struct drbd_device *device)
1259 struct bm_xfer_ctx c;
1262 if (!expect(device->bitmap))
1265 if (get_ldev(device)) {
1266 if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC)) {
1267 drbd_info(device, "Writing the whole bitmap, MDF_FullSync was set.\n");
1268 drbd_bm_set_all(device);
1269 if (drbd_bm_write(device)) {
1270 /* write_bm did fail! Leave full sync flag set in Meta P_DATA
1271 * but otherwise process as per normal - need to tell other
1272 * side that a full resync is required! */
1273 drbd_err(device, "Failed to write bitmap to disk!\n");
1275 drbd_md_clear_flag(device, MDF_FULL_SYNC);
1276 drbd_md_sync(device);
1282 c = (struct bm_xfer_ctx) {
1283 .bm_bits = drbd_bm_bits(device),
1284 .bm_words = drbd_bm_words(device),
1288 err = send_bitmap_rle_or_plain(device, &c);
1294 int drbd_send_bitmap(struct drbd_device *device)
1296 struct drbd_socket *sock = &first_peer_device(device)->connection->data;
1299 mutex_lock(&sock->mutex);
1301 err = !_drbd_send_bitmap(device);
1302 mutex_unlock(&sock->mutex);
1306 void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, u32 set_size)
1308 struct drbd_socket *sock;
1309 struct p_barrier_ack *p;
1311 if (connection->cstate < C_WF_REPORT_PARAMS)
1314 sock = &connection->meta;
1315 p = conn_prepare_command(connection, sock);
1318 p->barrier = barrier_nr;
1319 p->set_size = cpu_to_be32(set_size);
1320 conn_send_command(connection, sock, P_BARRIER_ACK, sizeof(*p), NULL, 0);
1324 * _drbd_send_ack() - Sends an ack packet
1325 * @peer_device: DRBD peer device.
1326 * @cmd: Packet command code.
1327 * @sector: sector, needs to be in big endian byte order
1328 * @blksize: size in byte, needs to be in big endian byte order
1329 * @block_id: Id, big endian byte order
1331 static int _drbd_send_ack(struct drbd_peer_device *peer_device, enum drbd_packet cmd,
1332 u64 sector, u32 blksize, u64 block_id)
1334 struct drbd_socket *sock;
1335 struct p_block_ack *p;
1337 if (peer_device->device->state.conn < C_CONNECTED)
1340 sock = &peer_device->connection->meta;
1341 p = drbd_prepare_command(peer_device, sock);
1345 p->block_id = block_id;
1346 p->blksize = blksize;
1347 p->seq_num = cpu_to_be32(atomic_inc_return(&peer_device->device->packet_seq));
1348 return drbd_send_command(peer_device, sock, cmd, sizeof(*p), NULL, 0);
1351 /* dp->sector and dp->block_id already/still in network byte order,
1352 * data_size is payload size according to dp->head,
1353 * and may need to be corrected for digest size. */
1354 void drbd_send_ack_dp(struct drbd_peer_device *peer_device, enum drbd_packet cmd,
1355 struct p_data *dp, int data_size)
1357 if (peer_device->connection->peer_integrity_tfm)
1358 data_size -= crypto_shash_digestsize(peer_device->connection->peer_integrity_tfm);
1359 _drbd_send_ack(peer_device, cmd, dp->sector, cpu_to_be32(data_size),
1363 void drbd_send_ack_rp(struct drbd_peer_device *peer_device, enum drbd_packet cmd,
1364 struct p_block_req *rp)
1366 _drbd_send_ack(peer_device, cmd, rp->sector, rp->blksize, rp->block_id);
1370 * drbd_send_ack() - Sends an ack packet
1371 * @peer_device: DRBD peer device
1372 * @cmd: packet command code
1373 * @peer_req: peer request
1375 int drbd_send_ack(struct drbd_peer_device *peer_device, enum drbd_packet cmd,
1376 struct drbd_peer_request *peer_req)
1378 return _drbd_send_ack(peer_device, cmd,
1379 cpu_to_be64(peer_req->i.sector),
1380 cpu_to_be32(peer_req->i.size),
1381 peer_req->block_id);
1384 /* This function misuses the block_id field to signal if the blocks
1385 * are is sync or not. */
1386 int drbd_send_ack_ex(struct drbd_peer_device *peer_device, enum drbd_packet cmd,
1387 sector_t sector, int blksize, u64 block_id)
1389 return _drbd_send_ack(peer_device, cmd,
1390 cpu_to_be64(sector),
1391 cpu_to_be32(blksize),
1392 cpu_to_be64(block_id));
1395 int drbd_send_rs_deallocated(struct drbd_peer_device *peer_device,
1396 struct drbd_peer_request *peer_req)
1398 struct drbd_socket *sock;
1399 struct p_block_desc *p;
1401 sock = &peer_device->connection->data;
1402 p = drbd_prepare_command(peer_device, sock);
1405 p->sector = cpu_to_be64(peer_req->i.sector);
1406 p->blksize = cpu_to_be32(peer_req->i.size);
1408 return drbd_send_command(peer_device, sock, P_RS_DEALLOCATED, sizeof(*p), NULL, 0);
1411 int drbd_send_drequest(struct drbd_peer_device *peer_device, int cmd,
1412 sector_t sector, int size, u64 block_id)
1414 struct drbd_socket *sock;
1415 struct p_block_req *p;
1417 sock = &peer_device->connection->data;
1418 p = drbd_prepare_command(peer_device, sock);
1421 p->sector = cpu_to_be64(sector);
1422 p->block_id = block_id;
1423 p->blksize = cpu_to_be32(size);
1424 return drbd_send_command(peer_device, sock, cmd, sizeof(*p), NULL, 0);
1427 int drbd_send_drequest_csum(struct drbd_peer_device *peer_device, sector_t sector, int size,
1428 void *digest, int digest_size, enum drbd_packet cmd)
1430 struct drbd_socket *sock;
1431 struct p_block_req *p;
1433 /* FIXME: Put the digest into the preallocated socket buffer. */
1435 sock = &peer_device->connection->data;
1436 p = drbd_prepare_command(peer_device, sock);
1439 p->sector = cpu_to_be64(sector);
1440 p->block_id = ID_SYNCER /* unused */;
1441 p->blksize = cpu_to_be32(size);
1442 return drbd_send_command(peer_device, sock, cmd, sizeof(*p), digest, digest_size);
1445 int drbd_send_ov_request(struct drbd_peer_device *peer_device, sector_t sector, int size)
1447 struct drbd_socket *sock;
1448 struct p_block_req *p;
1450 sock = &peer_device->connection->data;
1451 p = drbd_prepare_command(peer_device, sock);
1454 p->sector = cpu_to_be64(sector);
1455 p->block_id = ID_SYNCER /* unused */;
1456 p->blksize = cpu_to_be32(size);
1457 return drbd_send_command(peer_device, sock, P_OV_REQUEST, sizeof(*p), NULL, 0);
1460 /* called on sndtimeo
1461 * returns false if we should retry,
1462 * true if we think connection is dead
1464 static int we_should_drop_the_connection(struct drbd_connection *connection, struct socket *sock)
1467 /* long elapsed = (long)(jiffies - device->last_received); */
1469 drop_it = connection->meta.socket == sock
1470 || !connection->ack_receiver.task
1471 || get_t_state(&connection->ack_receiver) != RUNNING
1472 || connection->cstate < C_WF_REPORT_PARAMS;
1477 drop_it = !--connection->ko_count;
1479 drbd_err(connection, "[%s/%d] sock_sendmsg time expired, ko = %u\n",
1480 current->comm, current->pid, connection->ko_count);
1481 request_ping(connection);
1484 return drop_it; /* && (device->state == R_PRIMARY) */;
1487 static void drbd_update_congested(struct drbd_connection *connection)
1489 struct sock *sk = connection->data.socket->sk;
1490 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
1491 set_bit(NET_CONGESTED, &connection->flags);
1494 /* The idea of sendpage seems to be to put some kind of reference
1495 * to the page into the skb, and to hand it over to the NIC. In
1496 * this process get_page() gets called.
1498 * As soon as the page was really sent over the network put_page()
1499 * gets called by some part of the network layer. [ NIC driver? ]
1501 * [ get_page() / put_page() increment/decrement the count. If count
1502 * reaches 0 the page will be freed. ]
1504 * This works nicely with pages from FSs.
1505 * But this means that in protocol A we might signal IO completion too early!
1507 * In order not to corrupt data during a resync we must make sure
1508 * that we do not reuse our own buffer pages (EEs) to early, therefore
1509 * we have the net_ee list.
1511 * XFS seems to have problems, still, it submits pages with page_count == 0!
1512 * As a workaround, we disable sendpage on pages
1513 * with page_count == 0 or PageSlab.
1515 static int _drbd_no_send_page(struct drbd_peer_device *peer_device, struct page *page,
1516 int offset, size_t size, unsigned msg_flags)
1518 struct socket *socket;
1522 socket = peer_device->connection->data.socket;
1523 addr = kmap(page) + offset;
1524 err = drbd_send_all(peer_device->connection, socket, addr, size, msg_flags);
1527 peer_device->device->send_cnt += size >> 9;
1531 static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *page,
1532 int offset, size_t size, unsigned msg_flags)
1534 struct socket *socket = peer_device->connection->data.socket;
1538 /* e.g. XFS meta- & log-data is in slab pages, which have a
1539 * page_count of 0 and/or have PageSlab() set.
1540 * we cannot use send_page for those, as that does get_page();
1541 * put_page(); and would cause either a VM_BUG directly, or
1542 * __page_cache_release a page that would actually still be referenced
1543 * by someone, leading to some obscure delayed Oops somewhere else. */
1544 if (drbd_disable_sendpage || !sendpage_ok(page))
1545 return _drbd_no_send_page(peer_device, page, offset, size, msg_flags);
1547 msg_flags |= MSG_NOSIGNAL;
1548 drbd_update_congested(peer_device->connection);
1552 sent = socket->ops->sendpage(socket, page, offset, len, msg_flags);
1554 if (sent == -EAGAIN) {
1555 if (we_should_drop_the_connection(peer_device->connection, socket))
1559 drbd_warn(peer_device->device, "%s: size=%d len=%d sent=%d\n",
1560 __func__, (int)size, len, sent);
1567 } while (len > 0 /* THINK && device->cstate >= C_CONNECTED*/);
1568 clear_bit(NET_CONGESTED, &peer_device->connection->flags);
1572 peer_device->device->send_cnt += size >> 9;
1577 static int _drbd_send_bio(struct drbd_peer_device *peer_device, struct bio *bio)
1579 struct bio_vec bvec;
1580 struct bvec_iter iter;
1582 /* hint all but last page with MSG_MORE */
1583 bio_for_each_segment(bvec, bio, iter) {
1586 err = _drbd_no_send_page(peer_device, bvec.bv_page,
1587 bvec.bv_offset, bvec.bv_len,
1588 bio_iter_last(bvec, iter)
1596 static int _drbd_send_zc_bio(struct drbd_peer_device *peer_device, struct bio *bio)
1598 struct bio_vec bvec;
1599 struct bvec_iter iter;
1601 /* hint all but last page with MSG_MORE */
1602 bio_for_each_segment(bvec, bio, iter) {
1605 err = _drbd_send_page(peer_device, bvec.bv_page,
1606 bvec.bv_offset, bvec.bv_len,
1607 bio_iter_last(bvec, iter) ? 0 : MSG_MORE);
1614 static int _drbd_send_zc_ee(struct drbd_peer_device *peer_device,
1615 struct drbd_peer_request *peer_req)
1617 struct page *page = peer_req->pages;
1618 unsigned len = peer_req->i.size;
1621 /* hint all but last page with MSG_MORE */
1622 page_chain_for_each(page) {
1623 unsigned l = min_t(unsigned, len, PAGE_SIZE);
1625 err = _drbd_send_page(peer_device, page, 0, l,
1626 page_chain_next(page) ? MSG_MORE : 0);
1634 static u32 bio_flags_to_wire(struct drbd_connection *connection,
1637 if (connection->agreed_pro_version >= 95)
1638 return (bio->bi_opf & REQ_SYNC ? DP_RW_SYNC : 0) |
1639 (bio->bi_opf & REQ_FUA ? DP_FUA : 0) |
1640 (bio->bi_opf & REQ_PREFLUSH ? DP_FLUSH : 0) |
1641 (bio_op(bio) == REQ_OP_DISCARD ? DP_DISCARD : 0) |
1642 (bio_op(bio) == REQ_OP_WRITE_ZEROES ?
1643 ((connection->agreed_features & DRBD_FF_WZEROES) ?
1644 (DP_ZEROES |(!(bio->bi_opf & REQ_NOUNMAP) ? DP_DISCARD : 0))
1648 return bio->bi_opf & REQ_SYNC ? DP_RW_SYNC : 0;
1651 /* Used to send write or TRIM aka REQ_OP_DISCARD requests
1652 * R_PRIMARY -> Peer (P_DATA, P_TRIM)
1654 int drbd_send_dblock(struct drbd_peer_device *peer_device, struct drbd_request *req)
1656 struct drbd_device *device = peer_device->device;
1657 struct drbd_socket *sock;
1660 unsigned int dp_flags = 0;
1664 sock = &peer_device->connection->data;
1665 p = drbd_prepare_command(peer_device, sock);
1666 digest_size = peer_device->connection->integrity_tfm ?
1667 crypto_shash_digestsize(peer_device->connection->integrity_tfm) : 0;
1671 p->sector = cpu_to_be64(req->i.sector);
1672 p->block_id = (unsigned long)req;
1673 p->seq_num = cpu_to_be32(atomic_inc_return(&device->packet_seq));
1674 dp_flags = bio_flags_to_wire(peer_device->connection, req->master_bio);
1675 if (device->state.conn >= C_SYNC_SOURCE &&
1676 device->state.conn <= C_PAUSED_SYNC_T)
1677 dp_flags |= DP_MAY_SET_IN_SYNC;
1678 if (peer_device->connection->agreed_pro_version >= 100) {
1679 if (req->rq_state & RQ_EXP_RECEIVE_ACK)
1680 dp_flags |= DP_SEND_RECEIVE_ACK;
1681 /* During resync, request an explicit write ack,
1682 * even in protocol != C */
1683 if (req->rq_state & RQ_EXP_WRITE_ACK
1684 || (dp_flags & DP_MAY_SET_IN_SYNC))
1685 dp_flags |= DP_SEND_WRITE_ACK;
1687 p->dp_flags = cpu_to_be32(dp_flags);
1689 if (dp_flags & (DP_DISCARD|DP_ZEROES)) {
1690 enum drbd_packet cmd = (dp_flags & DP_ZEROES) ? P_ZEROES : P_TRIM;
1691 struct p_trim *t = (struct p_trim*)p;
1692 t->size = cpu_to_be32(req->i.size);
1693 err = __send_command(peer_device->connection, device->vnr, sock, cmd, sizeof(*t), NULL, 0);
1698 /* our digest is still only over the payload.
1699 * TRIM does not carry any payload. */
1701 drbd_csum_bio(peer_device->connection->integrity_tfm, req->master_bio, digest_out);
1702 err = __send_command(peer_device->connection, device->vnr, sock, P_DATA,
1703 sizeof(*p) + digest_size, NULL, req->i.size);
1705 /* For protocol A, we have to memcpy the payload into
1706 * socket buffers, as we may complete right away
1707 * as soon as we handed it over to tcp, at which point the data
1708 * pages may become invalid.
1710 * For data-integrity enabled, we copy it as well, so we can be
1711 * sure that even if the bio pages may still be modified, it
1712 * won't change the data on the wire, thus if the digest checks
1713 * out ok after sending on this side, but does not fit on the
1714 * receiving side, we sure have detected corruption elsewhere.
1716 if (!(req->rq_state & (RQ_EXP_RECEIVE_ACK | RQ_EXP_WRITE_ACK)) || digest_size)
1717 err = _drbd_send_bio(peer_device, req->master_bio);
1719 err = _drbd_send_zc_bio(peer_device, req->master_bio);
1721 /* double check digest, sometimes buffers have been modified in flight. */
1722 if (digest_size > 0 && digest_size <= 64) {
1723 /* 64 byte, 512 bit, is the largest digest size
1724 * currently supported in kernel crypto. */
1725 unsigned char digest[64];
1726 drbd_csum_bio(peer_device->connection->integrity_tfm, req->master_bio, digest);
1727 if (memcmp(p + 1, digest, digest_size)) {
1729 "Digest mismatch, buffer modified by upper layers during write: %llus +%u\n",
1730 (unsigned long long)req->i.sector, req->i.size);
1732 } /* else if (digest_size > 64) {
1733 ... Be noisy about digest too large ...
1737 mutex_unlock(&sock->mutex); /* locked by drbd_prepare_command() */
1742 /* answer packet, used to send data back for read requests:
1743 * Peer -> (diskless) R_PRIMARY (P_DATA_REPLY)
1744 * C_SYNC_SOURCE -> C_SYNC_TARGET (P_RS_DATA_REPLY)
1746 int drbd_send_block(struct drbd_peer_device *peer_device, enum drbd_packet cmd,
1747 struct drbd_peer_request *peer_req)
1749 struct drbd_device *device = peer_device->device;
1750 struct drbd_socket *sock;
1755 sock = &peer_device->connection->data;
1756 p = drbd_prepare_command(peer_device, sock);
1758 digest_size = peer_device->connection->integrity_tfm ?
1759 crypto_shash_digestsize(peer_device->connection->integrity_tfm) : 0;
1763 p->sector = cpu_to_be64(peer_req->i.sector);
1764 p->block_id = peer_req->block_id;
1765 p->seq_num = 0; /* unused */
1768 drbd_csum_ee(peer_device->connection->integrity_tfm, peer_req, p + 1);
1769 err = __send_command(peer_device->connection, device->vnr, sock, cmd, sizeof(*p) + digest_size, NULL, peer_req->i.size);
1771 err = _drbd_send_zc_ee(peer_device, peer_req);
1772 mutex_unlock(&sock->mutex); /* locked by drbd_prepare_command() */
1777 int drbd_send_out_of_sync(struct drbd_peer_device *peer_device, struct drbd_request *req)
1779 struct drbd_socket *sock;
1780 struct p_block_desc *p;
1782 sock = &peer_device->connection->data;
1783 p = drbd_prepare_command(peer_device, sock);
1786 p->sector = cpu_to_be64(req->i.sector);
1787 p->blksize = cpu_to_be32(req->i.size);
1788 return drbd_send_command(peer_device, sock, P_OUT_OF_SYNC, sizeof(*p), NULL, 0);
1792 drbd_send distinguishes two cases:
1794 Packets sent via the data socket "sock"
1795 and packets sent via the meta data socket "msock"
1798 -----------------+-------------------------+------------------------------
1799 timeout conf.timeout / 2 conf.timeout / 2
1800 timeout action send a ping via msock Abort communication
1801 and close all sockets
1805 * you must have down()ed the appropriate [m]sock_mutex elsewhere!
1807 int drbd_send(struct drbd_connection *connection, struct socket *sock,
1808 void *buf, size_t size, unsigned msg_flags)
1810 struct kvec iov = {.iov_base = buf, .iov_len = size};
1811 struct msghdr msg = {.msg_flags = msg_flags | MSG_NOSIGNAL};
1817 /* THINK if (signal_pending) return ... ? */
1819 iov_iter_kvec(&msg.msg_iter, WRITE, &iov, 1, size);
1821 if (sock == connection->data.socket) {
1823 connection->ko_count = rcu_dereference(connection->net_conf)->ko_count;
1825 drbd_update_congested(connection);
1828 rv = sock_sendmsg(sock, &msg);
1829 if (rv == -EAGAIN) {
1830 if (we_should_drop_the_connection(connection, sock))
1836 flush_signals(current);
1842 } while (sent < size);
1844 if (sock == connection->data.socket)
1845 clear_bit(NET_CONGESTED, &connection->flags);
1848 if (rv != -EAGAIN) {
1849 drbd_err(connection, "%s_sendmsg returned %d\n",
1850 sock == connection->meta.socket ? "msock" : "sock",
1852 conn_request_state(connection, NS(conn, C_BROKEN_PIPE), CS_HARD);
1854 conn_request_state(connection, NS(conn, C_TIMEOUT), CS_HARD);
1861 * drbd_send_all - Send an entire buffer
1863 * Returns 0 upon success and a negative error value otherwise.
1865 int drbd_send_all(struct drbd_connection *connection, struct socket *sock, void *buffer,
1866 size_t size, unsigned msg_flags)
1870 err = drbd_send(connection, sock, buffer, size, msg_flags);
1878 static int drbd_open(struct block_device *bdev, fmode_t mode)
1880 struct drbd_device *device = bdev->bd_disk->private_data;
1881 unsigned long flags;
1884 mutex_lock(&drbd_main_mutex);
1885 spin_lock_irqsave(&device->resource->req_lock, flags);
1886 /* to have a stable device->state.role
1887 * and no race with updating open_cnt */
1889 if (device->state.role != R_PRIMARY) {
1890 if (mode & FMODE_WRITE)
1892 else if (!drbd_allow_oos)
1898 spin_unlock_irqrestore(&device->resource->req_lock, flags);
1899 mutex_unlock(&drbd_main_mutex);
1904 static void drbd_release(struct gendisk *gd, fmode_t mode)
1906 struct drbd_device *device = gd->private_data;
1907 mutex_lock(&drbd_main_mutex);
1909 mutex_unlock(&drbd_main_mutex);
1912 /* need to hold resource->req_lock */
1913 void drbd_queue_unplug(struct drbd_device *device)
1915 if (device->state.pdsk >= D_INCONSISTENT && device->state.conn >= C_CONNECTED) {
1916 D_ASSERT(device, device->state.role == R_PRIMARY);
1917 if (test_and_clear_bit(UNPLUG_REMOTE, &device->flags)) {
1918 drbd_queue_work_if_unqueued(
1919 &first_peer_device(device)->connection->sender_work,
1920 &device->unplug_work);
1925 static void drbd_set_defaults(struct drbd_device *device)
1927 /* Beware! The actual layout differs
1928 * between big endian and little endian */
1929 device->state = (union drbd_dev_state) {
1930 { .role = R_SECONDARY,
1932 .conn = C_STANDALONE,
1938 void drbd_init_set_defaults(struct drbd_device *device)
1940 /* the memset(,0,) did most of this.
1941 * note: only assignments, no allocation in here */
1943 drbd_set_defaults(device);
1945 atomic_set(&device->ap_bio_cnt, 0);
1946 atomic_set(&device->ap_actlog_cnt, 0);
1947 atomic_set(&device->ap_pending_cnt, 0);
1948 atomic_set(&device->rs_pending_cnt, 0);
1949 atomic_set(&device->unacked_cnt, 0);
1950 atomic_set(&device->local_cnt, 0);
1951 atomic_set(&device->pp_in_use_by_net, 0);
1952 atomic_set(&device->rs_sect_in, 0);
1953 atomic_set(&device->rs_sect_ev, 0);
1954 atomic_set(&device->ap_in_flight, 0);
1955 atomic_set(&device->md_io.in_use, 0);
1957 mutex_init(&device->own_state_mutex);
1958 device->state_mutex = &device->own_state_mutex;
1960 spin_lock_init(&device->al_lock);
1961 spin_lock_init(&device->peer_seq_lock);
1963 INIT_LIST_HEAD(&device->active_ee);
1964 INIT_LIST_HEAD(&device->sync_ee);
1965 INIT_LIST_HEAD(&device->done_ee);
1966 INIT_LIST_HEAD(&device->read_ee);
1967 INIT_LIST_HEAD(&device->net_ee);
1968 INIT_LIST_HEAD(&device->resync_reads);
1969 INIT_LIST_HEAD(&device->resync_work.list);
1970 INIT_LIST_HEAD(&device->unplug_work.list);
1971 INIT_LIST_HEAD(&device->bm_io_work.w.list);
1972 INIT_LIST_HEAD(&device->pending_master_completion[0]);
1973 INIT_LIST_HEAD(&device->pending_master_completion[1]);
1974 INIT_LIST_HEAD(&device->pending_completion[0]);
1975 INIT_LIST_HEAD(&device->pending_completion[1]);
1977 device->resync_work.cb = w_resync_timer;
1978 device->unplug_work.cb = w_send_write_hint;
1979 device->bm_io_work.w.cb = w_bitmap_io;
1981 timer_setup(&device->resync_timer, resync_timer_fn, 0);
1982 timer_setup(&device->md_sync_timer, md_sync_timer_fn, 0);
1983 timer_setup(&device->start_resync_timer, start_resync_timer_fn, 0);
1984 timer_setup(&device->request_timer, request_timer_fn, 0);
1986 init_waitqueue_head(&device->misc_wait);
1987 init_waitqueue_head(&device->state_wait);
1988 init_waitqueue_head(&device->ee_wait);
1989 init_waitqueue_head(&device->al_wait);
1990 init_waitqueue_head(&device->seq_wait);
1992 device->resync_wenr = LC_FREE;
1993 device->peer_max_bio_size = DRBD_MAX_BIO_SIZE_SAFE;
1994 device->local_max_bio_size = DRBD_MAX_BIO_SIZE_SAFE;
1997 void drbd_set_my_capacity(struct drbd_device *device, sector_t size)
2001 set_capacity_and_notify(device->vdisk, size);
2003 drbd_info(device, "size = %s (%llu KB)\n",
2004 ppsize(ppb, size>>1), (unsigned long long)size>>1);
2007 void drbd_device_cleanup(struct drbd_device *device)
2010 if (first_peer_device(device)->connection->receiver.t_state != NONE)
2011 drbd_err(device, "ASSERT FAILED: receiver t_state == %d expected 0.\n",
2012 first_peer_device(device)->connection->receiver.t_state);
2014 device->al_writ_cnt =
2015 device->bm_writ_cnt =
2023 device->rs_failed = 0;
2024 device->rs_last_events = 0;
2025 device->rs_last_sect_ev = 0;
2026 for (i = 0; i < DRBD_SYNC_MARKS; i++) {
2027 device->rs_mark_left[i] = 0;
2028 device->rs_mark_time[i] = 0;
2030 D_ASSERT(device, first_peer_device(device)->connection->net_conf == NULL);
2032 set_capacity_and_notify(device->vdisk, 0);
2033 if (device->bitmap) {
2034 /* maybe never allocated. */
2035 drbd_bm_resize(device, 0, 1);
2036 drbd_bm_cleanup(device);
2039 drbd_backing_dev_free(device, device->ldev);
2040 device->ldev = NULL;
2042 clear_bit(AL_SUSPENDED, &device->flags);
2044 D_ASSERT(device, list_empty(&device->active_ee));
2045 D_ASSERT(device, list_empty(&device->sync_ee));
2046 D_ASSERT(device, list_empty(&device->done_ee));
2047 D_ASSERT(device, list_empty(&device->read_ee));
2048 D_ASSERT(device, list_empty(&device->net_ee));
2049 D_ASSERT(device, list_empty(&device->resync_reads));
2050 D_ASSERT(device, list_empty(&first_peer_device(device)->connection->sender_work.q));
2051 D_ASSERT(device, list_empty(&device->resync_work.list));
2052 D_ASSERT(device, list_empty(&device->unplug_work.list));
2054 drbd_set_defaults(device);
2058 static void drbd_destroy_mempools(void)
2062 while (drbd_pp_pool) {
2063 page = drbd_pp_pool;
2064 drbd_pp_pool = (struct page *)page_private(page);
2069 /* D_ASSERT(device, atomic_read(&drbd_pp_vacant)==0); */
2071 bioset_exit(&drbd_io_bio_set);
2072 bioset_exit(&drbd_md_io_bio_set);
2073 mempool_exit(&drbd_md_io_page_pool);
2074 mempool_exit(&drbd_ee_mempool);
2075 mempool_exit(&drbd_request_mempool);
2076 kmem_cache_destroy(drbd_ee_cache);
2077 kmem_cache_destroy(drbd_request_cache);
2078 kmem_cache_destroy(drbd_bm_ext_cache);
2079 kmem_cache_destroy(drbd_al_ext_cache);
2081 drbd_ee_cache = NULL;
2082 drbd_request_cache = NULL;
2083 drbd_bm_ext_cache = NULL;
2084 drbd_al_ext_cache = NULL;
2089 static int drbd_create_mempools(void)
2092 const int number = (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * drbd_minor_count;
2096 drbd_request_cache = kmem_cache_create(
2097 "drbd_req", sizeof(struct drbd_request), 0, 0, NULL);
2098 if (drbd_request_cache == NULL)
2101 drbd_ee_cache = kmem_cache_create(
2102 "drbd_ee", sizeof(struct drbd_peer_request), 0, 0, NULL);
2103 if (drbd_ee_cache == NULL)
2106 drbd_bm_ext_cache = kmem_cache_create(
2107 "drbd_bm", sizeof(struct bm_extent), 0, 0, NULL);
2108 if (drbd_bm_ext_cache == NULL)
2111 drbd_al_ext_cache = kmem_cache_create(
2112 "drbd_al", sizeof(struct lc_element), 0, 0, NULL);
2113 if (drbd_al_ext_cache == NULL)
2117 ret = bioset_init(&drbd_io_bio_set, BIO_POOL_SIZE, 0, 0);
2121 ret = bioset_init(&drbd_md_io_bio_set, DRBD_MIN_POOL_PAGES, 0,
2126 ret = mempool_init_page_pool(&drbd_md_io_page_pool, DRBD_MIN_POOL_PAGES, 0);
2130 ret = mempool_init_slab_pool(&drbd_request_mempool, number,
2131 drbd_request_cache);
2135 ret = mempool_init_slab_pool(&drbd_ee_mempool, number, drbd_ee_cache);
2139 for (i = 0; i < number; i++) {
2140 page = alloc_page(GFP_HIGHUSER);
2143 set_page_private(page, (unsigned long)drbd_pp_pool);
2144 drbd_pp_pool = page;
2146 drbd_pp_vacant = number;
2151 drbd_destroy_mempools(); /* in case we allocated some */
2155 static void drbd_release_all_peer_reqs(struct drbd_device *device)
2159 rr = drbd_free_peer_reqs(device, &device->active_ee);
2161 drbd_err(device, "%d EEs in active list found!\n", rr);
2163 rr = drbd_free_peer_reqs(device, &device->sync_ee);
2165 drbd_err(device, "%d EEs in sync list found!\n", rr);
2167 rr = drbd_free_peer_reqs(device, &device->read_ee);
2169 drbd_err(device, "%d EEs in read list found!\n", rr);
2171 rr = drbd_free_peer_reqs(device, &device->done_ee);
2173 drbd_err(device, "%d EEs in done list found!\n", rr);
2175 rr = drbd_free_peer_reqs(device, &device->net_ee);
2177 drbd_err(device, "%d EEs in net list found!\n", rr);
2180 /* caution. no locking. */
2181 void drbd_destroy_device(struct kref *kref)
2183 struct drbd_device *device = container_of(kref, struct drbd_device, kref);
2184 struct drbd_resource *resource = device->resource;
2185 struct drbd_peer_device *peer_device, *tmp_peer_device;
2187 del_timer_sync(&device->request_timer);
2189 /* paranoia asserts */
2190 D_ASSERT(device, device->open_cnt == 0);
2191 /* end paranoia asserts */
2193 /* cleanup stuff that may have been allocated during
2194 * device (re-)configuration or state changes */
2196 drbd_backing_dev_free(device, device->ldev);
2197 device->ldev = NULL;
2199 drbd_release_all_peer_reqs(device);
2201 lc_destroy(device->act_log);
2202 lc_destroy(device->resync);
2204 kfree(device->p_uuid);
2205 /* device->p_uuid = NULL; */
2207 if (device->bitmap) /* should no longer be there. */
2208 drbd_bm_cleanup(device);
2209 __free_page(device->md_io.page);
2210 put_disk(device->vdisk);
2211 kfree(device->rs_plan_s);
2213 /* not for_each_connection(connection, resource):
2214 * those may have been cleaned up and disassociated already.
2216 for_each_peer_device_safe(peer_device, tmp_peer_device, device) {
2217 kref_put(&peer_device->connection->kref, drbd_destroy_connection);
2220 memset(device, 0xfd, sizeof(*device));
2222 kref_put(&resource->kref, drbd_destroy_resource);
2225 /* One global retry thread, if we need to push back some bio and have it
2226 * reinserted through our make request function.
2228 static struct retry_worker {
2229 struct workqueue_struct *wq;
2230 struct work_struct worker;
2233 struct list_head writes;
2236 static void do_retry(struct work_struct *ws)
2238 struct retry_worker *retry = container_of(ws, struct retry_worker, worker);
2240 struct drbd_request *req, *tmp;
2242 spin_lock_irq(&retry->lock);
2243 list_splice_init(&retry->writes, &writes);
2244 spin_unlock_irq(&retry->lock);
2246 list_for_each_entry_safe(req, tmp, &writes, tl_requests) {
2247 struct drbd_device *device = req->device;
2248 struct bio *bio = req->master_bio;
2252 expect(atomic_read(&req->completion_ref) == 0) &&
2253 expect(req->rq_state & RQ_POSTPONED) &&
2254 expect((req->rq_state & RQ_LOCAL_PENDING) == 0 ||
2255 (req->rq_state & RQ_LOCAL_ABORTED) != 0);
2258 drbd_err(device, "req=%p completion_ref=%d rq_state=%x\n",
2259 req, atomic_read(&req->completion_ref),
2262 /* We still need to put one kref associated with the
2263 * "completion_ref" going zero in the code path that queued it
2264 * here. The request object may still be referenced by a
2265 * frozen local req->private_bio, in case we force-detached.
2267 kref_put(&req->kref, drbd_req_destroy);
2269 /* A single suspended or otherwise blocking device may stall
2270 * all others as well. Fortunately, this code path is to
2271 * recover from a situation that "should not happen":
2272 * concurrent writes in multi-primary setup.
2273 * In a "normal" lifecycle, this workqueue is supposed to be
2274 * destroyed without ever doing anything.
2275 * If it turns out to be an issue anyways, we can do per
2276 * resource (replication group) or per device (minor) retry
2277 * workqueues instead.
2280 /* We are not just doing submit_bio_noacct(),
2281 * as we want to keep the start_time information. */
2283 __drbd_make_request(device, bio);
2287 /* called via drbd_req_put_completion_ref(),
2288 * holds resource->req_lock */
2289 void drbd_restart_request(struct drbd_request *req)
2291 unsigned long flags;
2292 spin_lock_irqsave(&retry.lock, flags);
2293 list_move_tail(&req->tl_requests, &retry.writes);
2294 spin_unlock_irqrestore(&retry.lock, flags);
2296 /* Drop the extra reference that would otherwise
2297 * have been dropped by complete_master_bio.
2298 * do_retry() needs to grab a new one. */
2299 dec_ap_bio(req->device);
2301 queue_work(retry.wq, &retry.worker);
2304 void drbd_destroy_resource(struct kref *kref)
2306 struct drbd_resource *resource =
2307 container_of(kref, struct drbd_resource, kref);
2309 idr_destroy(&resource->devices);
2310 free_cpumask_var(resource->cpu_mask);
2311 kfree(resource->name);
2312 memset(resource, 0xf2, sizeof(*resource));
2316 void drbd_free_resource(struct drbd_resource *resource)
2318 struct drbd_connection *connection, *tmp;
2320 for_each_connection_safe(connection, tmp, resource) {
2321 list_del(&connection->connections);
2322 drbd_debugfs_connection_cleanup(connection);
2323 kref_put(&connection->kref, drbd_destroy_connection);
2325 drbd_debugfs_resource_cleanup(resource);
2326 kref_put(&resource->kref, drbd_destroy_resource);
2329 static void drbd_cleanup(void)
2332 struct drbd_device *device;
2333 struct drbd_resource *resource, *tmp;
2335 /* first remove proc,
2336 * drbdsetup uses it's presence to detect
2337 * whether DRBD is loaded.
2338 * If we would get stuck in proc removal,
2339 * but have netlink already deregistered,
2340 * some drbdsetup commands may wait forever
2344 remove_proc_entry("drbd", NULL);
2347 destroy_workqueue(retry.wq);
2349 drbd_genl_unregister();
2351 idr_for_each_entry(&drbd_devices, device, i)
2352 drbd_delete_device(device);
2354 /* not _rcu since, no other updater anymore. Genl already unregistered */
2355 for_each_resource_safe(resource, tmp, &drbd_resources) {
2356 list_del(&resource->resources);
2357 drbd_free_resource(resource);
2360 drbd_debugfs_cleanup();
2362 drbd_destroy_mempools();
2363 unregister_blkdev(DRBD_MAJOR, "drbd");
2365 idr_destroy(&drbd_devices);
2367 pr_info("module cleanup done.\n");
2370 static void drbd_init_workqueue(struct drbd_work_queue* wq)
2372 spin_lock_init(&wq->q_lock);
2373 INIT_LIST_HEAD(&wq->q);
2374 init_waitqueue_head(&wq->q_wait);
2377 struct completion_work {
2379 struct completion done;
2382 static int w_complete(struct drbd_work *w, int cancel)
2384 struct completion_work *completion_work =
2385 container_of(w, struct completion_work, w);
2387 complete(&completion_work->done);
2391 void drbd_flush_workqueue(struct drbd_work_queue *work_queue)
2393 struct completion_work completion_work;
2395 completion_work.w.cb = w_complete;
2396 init_completion(&completion_work.done);
2397 drbd_queue_work(work_queue, &completion_work.w);
2398 wait_for_completion(&completion_work.done);
2401 struct drbd_resource *drbd_find_resource(const char *name)
2403 struct drbd_resource *resource;
2405 if (!name || !name[0])
2409 for_each_resource_rcu(resource, &drbd_resources) {
2410 if (!strcmp(resource->name, name)) {
2411 kref_get(&resource->kref);
2421 struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
2422 void *peer_addr, int peer_addr_len)
2424 struct drbd_resource *resource;
2425 struct drbd_connection *connection;
2428 for_each_resource_rcu(resource, &drbd_resources) {
2429 for_each_connection_rcu(connection, resource) {
2430 if (connection->my_addr_len == my_addr_len &&
2431 connection->peer_addr_len == peer_addr_len &&
2432 !memcmp(&connection->my_addr, my_addr, my_addr_len) &&
2433 !memcmp(&connection->peer_addr, peer_addr, peer_addr_len)) {
2434 kref_get(&connection->kref);
2445 static int drbd_alloc_socket(struct drbd_socket *socket)
2447 socket->rbuf = (void *) __get_free_page(GFP_KERNEL);
2450 socket->sbuf = (void *) __get_free_page(GFP_KERNEL);
2456 static void drbd_free_socket(struct drbd_socket *socket)
2458 free_page((unsigned long) socket->sbuf);
2459 free_page((unsigned long) socket->rbuf);
2462 void conn_free_crypto(struct drbd_connection *connection)
2464 drbd_free_sock(connection);
2466 crypto_free_shash(connection->csums_tfm);
2467 crypto_free_shash(connection->verify_tfm);
2468 crypto_free_shash(connection->cram_hmac_tfm);
2469 crypto_free_shash(connection->integrity_tfm);
2470 crypto_free_shash(connection->peer_integrity_tfm);
2471 kfree(connection->int_dig_in);
2472 kfree(connection->int_dig_vv);
2474 connection->csums_tfm = NULL;
2475 connection->verify_tfm = NULL;
2476 connection->cram_hmac_tfm = NULL;
2477 connection->integrity_tfm = NULL;
2478 connection->peer_integrity_tfm = NULL;
2479 connection->int_dig_in = NULL;
2480 connection->int_dig_vv = NULL;
2483 int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts)
2485 struct drbd_connection *connection;
2486 cpumask_var_t new_cpu_mask;
2489 if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL))
2492 /* silently ignore cpu mask on UP kernel */
2493 if (nr_cpu_ids > 1 && res_opts->cpu_mask[0] != 0) {
2494 err = bitmap_parse(res_opts->cpu_mask, DRBD_CPU_MASK_SIZE,
2495 cpumask_bits(new_cpu_mask), nr_cpu_ids);
2496 if (err == -EOVERFLOW) {
2497 /* So what. mask it out. */
2498 cpumask_var_t tmp_cpu_mask;
2499 if (zalloc_cpumask_var(&tmp_cpu_mask, GFP_KERNEL)) {
2500 cpumask_setall(tmp_cpu_mask);
2501 cpumask_and(new_cpu_mask, new_cpu_mask, tmp_cpu_mask);
2502 drbd_warn(resource, "Overflow in bitmap_parse(%.12s%s), truncating to %u bits\n",
2504 strlen(res_opts->cpu_mask) > 12 ? "..." : "",
2506 free_cpumask_var(tmp_cpu_mask);
2511 drbd_warn(resource, "bitmap_parse() failed with %d\n", err);
2512 /* retcode = ERR_CPU_MASK_PARSE; */
2516 resource->res_opts = *res_opts;
2517 if (cpumask_empty(new_cpu_mask))
2518 drbd_calc_cpu_mask(&new_cpu_mask);
2519 if (!cpumask_equal(resource->cpu_mask, new_cpu_mask)) {
2520 cpumask_copy(resource->cpu_mask, new_cpu_mask);
2521 for_each_connection_rcu(connection, resource) {
2522 connection->receiver.reset_cpu_mask = 1;
2523 connection->ack_receiver.reset_cpu_mask = 1;
2524 connection->worker.reset_cpu_mask = 1;
2530 free_cpumask_var(new_cpu_mask);
2535 struct drbd_resource *drbd_create_resource(const char *name)
2537 struct drbd_resource *resource;
2539 resource = kzalloc(sizeof(struct drbd_resource), GFP_KERNEL);
2542 resource->name = kstrdup(name, GFP_KERNEL);
2543 if (!resource->name)
2544 goto fail_free_resource;
2545 if (!zalloc_cpumask_var(&resource->cpu_mask, GFP_KERNEL))
2546 goto fail_free_name;
2547 kref_init(&resource->kref);
2548 idr_init(&resource->devices);
2549 INIT_LIST_HEAD(&resource->connections);
2550 resource->write_ordering = WO_BDEV_FLUSH;
2551 list_add_tail_rcu(&resource->resources, &drbd_resources);
2552 mutex_init(&resource->conf_update);
2553 mutex_init(&resource->adm_mutex);
2554 spin_lock_init(&resource->req_lock);
2555 drbd_debugfs_resource_add(resource);
2559 kfree(resource->name);
2566 /* caller must be under adm_mutex */
2567 struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts)
2569 struct drbd_resource *resource;
2570 struct drbd_connection *connection;
2572 connection = kzalloc(sizeof(struct drbd_connection), GFP_KERNEL);
2576 if (drbd_alloc_socket(&connection->data))
2578 if (drbd_alloc_socket(&connection->meta))
2581 connection->current_epoch = kzalloc(sizeof(struct drbd_epoch), GFP_KERNEL);
2582 if (!connection->current_epoch)
2585 INIT_LIST_HEAD(&connection->transfer_log);
2587 INIT_LIST_HEAD(&connection->current_epoch->list);
2588 connection->epochs = 1;
2589 spin_lock_init(&connection->epoch_lock);
2591 connection->send.seen_any_write_yet = false;
2592 connection->send.current_epoch_nr = 0;
2593 connection->send.current_epoch_writes = 0;
2595 resource = drbd_create_resource(name);
2599 connection->cstate = C_STANDALONE;
2600 mutex_init(&connection->cstate_mutex);
2601 init_waitqueue_head(&connection->ping_wait);
2602 idr_init(&connection->peer_devices);
2604 drbd_init_workqueue(&connection->sender_work);
2605 mutex_init(&connection->data.mutex);
2606 mutex_init(&connection->meta.mutex);
2608 drbd_thread_init(resource, &connection->receiver, drbd_receiver, "receiver");
2609 connection->receiver.connection = connection;
2610 drbd_thread_init(resource, &connection->worker, drbd_worker, "worker");
2611 connection->worker.connection = connection;
2612 drbd_thread_init(resource, &connection->ack_receiver, drbd_ack_receiver, "ack_recv");
2613 connection->ack_receiver.connection = connection;
2615 kref_init(&connection->kref);
2617 connection->resource = resource;
2619 if (set_resource_options(resource, res_opts))
2622 kref_get(&resource->kref);
2623 list_add_tail_rcu(&connection->connections, &resource->connections);
2624 drbd_debugfs_connection_add(connection);
2628 list_del(&resource->resources);
2629 drbd_free_resource(resource);
2631 kfree(connection->current_epoch);
2632 drbd_free_socket(&connection->meta);
2633 drbd_free_socket(&connection->data);
2638 void drbd_destroy_connection(struct kref *kref)
2640 struct drbd_connection *connection = container_of(kref, struct drbd_connection, kref);
2641 struct drbd_resource *resource = connection->resource;
2643 if (atomic_read(&connection->current_epoch->epoch_size) != 0)
2644 drbd_err(connection, "epoch_size:%d\n", atomic_read(&connection->current_epoch->epoch_size));
2645 kfree(connection->current_epoch);
2647 idr_destroy(&connection->peer_devices);
2649 drbd_free_socket(&connection->meta);
2650 drbd_free_socket(&connection->data);
2651 kfree(connection->int_dig_in);
2652 kfree(connection->int_dig_vv);
2653 memset(connection, 0xfc, sizeof(*connection));
2655 kref_put(&resource->kref, drbd_destroy_resource);
2658 static int init_submitter(struct drbd_device *device)
2660 /* opencoded create_singlethread_workqueue(),
2661 * to be able to say "drbd%d", ..., minor */
2663 alloc_ordered_workqueue("drbd%u_submit", WQ_MEM_RECLAIM, device->minor);
2664 if (!device->submit.wq)
2667 INIT_WORK(&device->submit.worker, do_submit);
2668 INIT_LIST_HEAD(&device->submit.writes);
2672 enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor)
2674 struct drbd_resource *resource = adm_ctx->resource;
2675 struct drbd_connection *connection, *n;
2676 struct drbd_device *device;
2677 struct drbd_peer_device *peer_device, *tmp_peer_device;
2678 struct gendisk *disk;
2680 int vnr = adm_ctx->volume;
2681 enum drbd_ret_code err = ERR_NOMEM;
2683 device = minor_to_device(minor);
2685 return ERR_MINOR_OR_VOLUME_EXISTS;
2687 /* GFP_KERNEL, we are outside of all write-out paths */
2688 device = kzalloc(sizeof(struct drbd_device), GFP_KERNEL);
2691 kref_init(&device->kref);
2693 kref_get(&resource->kref);
2694 device->resource = resource;
2695 device->minor = minor;
2698 drbd_init_set_defaults(device);
2700 disk = blk_alloc_disk(NUMA_NO_NODE);
2704 device->vdisk = disk;
2705 device->rq_queue = disk->queue;
2707 set_disk_ro(disk, true);
2709 disk->major = DRBD_MAJOR;
2710 disk->first_minor = minor;
2712 disk->fops = &drbd_ops;
2713 disk->flags |= GENHD_FL_NO_PART;
2714 sprintf(disk->disk_name, "drbd%d", minor);
2715 disk->private_data = device;
2717 blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue);
2718 blk_queue_write_cache(disk->queue, true, true);
2719 /* Setting the max_hw_sectors to an odd value of 8kibyte here
2720 This triggers a max_bio_size message upon first attach or connect */
2721 blk_queue_max_hw_sectors(disk->queue, DRBD_MAX_BIO_SIZE_SAFE >> 8);
2723 device->md_io.page = alloc_page(GFP_KERNEL);
2724 if (!device->md_io.page)
2725 goto out_no_io_page;
2727 if (drbd_bm_init(device))
2729 device->read_requests = RB_ROOT;
2730 device->write_requests = RB_ROOT;
2732 id = idr_alloc(&drbd_devices, device, minor, minor + 1, GFP_KERNEL);
2735 err = ERR_MINOR_OR_VOLUME_EXISTS;
2736 goto out_no_minor_idr;
2738 kref_get(&device->kref);
2740 id = idr_alloc(&resource->devices, device, vnr, vnr + 1, GFP_KERNEL);
2743 err = ERR_MINOR_OR_VOLUME_EXISTS;
2744 goto out_idr_remove_minor;
2746 kref_get(&device->kref);
2748 INIT_LIST_HEAD(&device->peer_devices);
2749 INIT_LIST_HEAD(&device->pending_bitmap_io);
2750 for_each_connection(connection, resource) {
2751 peer_device = kzalloc(sizeof(struct drbd_peer_device), GFP_KERNEL);
2753 goto out_idr_remove_from_resource;
2754 peer_device->connection = connection;
2755 peer_device->device = device;
2757 list_add(&peer_device->peer_devices, &device->peer_devices);
2758 kref_get(&device->kref);
2760 id = idr_alloc(&connection->peer_devices, peer_device, vnr, vnr + 1, GFP_KERNEL);
2763 err = ERR_INVALID_REQUEST;
2764 goto out_idr_remove_from_resource;
2766 kref_get(&connection->kref);
2767 INIT_WORK(&peer_device->send_acks_work, drbd_send_acks_wf);
2770 if (init_submitter(device)) {
2772 goto out_idr_remove_from_resource;
2775 err = add_disk(disk);
2777 goto out_idr_remove_from_resource;
2779 /* inherit the connection state */
2780 device->state.conn = first_connection(resource)->cstate;
2781 if (device->state.conn == C_WF_REPORT_PARAMS) {
2782 for_each_peer_device(peer_device, device)
2783 drbd_connected(peer_device);
2785 /* move to create_peer_device() */
2786 for_each_peer_device(peer_device, device)
2787 drbd_debugfs_peer_device_add(peer_device);
2788 drbd_debugfs_device_add(device);
2791 out_idr_remove_from_resource:
2792 for_each_connection_safe(connection, n, resource) {
2793 peer_device = idr_remove(&connection->peer_devices, vnr);
2795 kref_put(&connection->kref, drbd_destroy_connection);
2797 for_each_peer_device_safe(peer_device, tmp_peer_device, device) {
2798 list_del(&peer_device->peer_devices);
2801 idr_remove(&resource->devices, vnr);
2802 out_idr_remove_minor:
2803 idr_remove(&drbd_devices, minor);
2806 drbd_bm_cleanup(device);
2808 __free_page(device->md_io.page);
2812 kref_put(&resource->kref, drbd_destroy_resource);
2817 void drbd_delete_device(struct drbd_device *device)
2819 struct drbd_resource *resource = device->resource;
2820 struct drbd_connection *connection;
2821 struct drbd_peer_device *peer_device;
2823 /* move to free_peer_device() */
2824 for_each_peer_device(peer_device, device)
2825 drbd_debugfs_peer_device_cleanup(peer_device);
2826 drbd_debugfs_device_cleanup(device);
2827 for_each_connection(connection, resource) {
2828 idr_remove(&connection->peer_devices, device->vnr);
2829 kref_put(&device->kref, drbd_destroy_device);
2831 idr_remove(&resource->devices, device->vnr);
2832 kref_put(&device->kref, drbd_destroy_device);
2833 idr_remove(&drbd_devices, device_to_minor(device));
2834 kref_put(&device->kref, drbd_destroy_device);
2835 del_gendisk(device->vdisk);
2837 kref_put(&device->kref, drbd_destroy_device);
2840 static int __init drbd_init(void)
2844 if (drbd_minor_count < DRBD_MINOR_COUNT_MIN || drbd_minor_count > DRBD_MINOR_COUNT_MAX) {
2845 pr_err("invalid minor_count (%d)\n", drbd_minor_count);
2849 drbd_minor_count = DRBD_MINOR_COUNT_DEF;
2853 err = register_blkdev(DRBD_MAJOR, "drbd");
2855 pr_err("unable to register block device major %d\n",
2861 * allocate all necessary structs
2863 init_waitqueue_head(&drbd_pp_wait);
2865 drbd_proc = NULL; /* play safe for drbd_cleanup */
2866 idr_init(&drbd_devices);
2868 mutex_init(&resources_mutex);
2869 INIT_LIST_HEAD(&drbd_resources);
2871 err = drbd_genl_register();
2873 pr_err("unable to register generic netlink family\n");
2877 err = drbd_create_mempools();
2882 drbd_proc = proc_create_single("drbd", S_IFREG | 0444 , NULL, drbd_seq_show);
2884 pr_err("unable to register proc file\n");
2888 retry.wq = create_singlethread_workqueue("drbd-reissue");
2890 pr_err("unable to create retry workqueue\n");
2893 INIT_WORK(&retry.worker, do_retry);
2894 spin_lock_init(&retry.lock);
2895 INIT_LIST_HEAD(&retry.writes);
2897 drbd_debugfs_init();
2899 pr_info("initialized. "
2900 "Version: " REL_VERSION " (api:%d/proto:%d-%d)\n",
2901 API_VERSION, PRO_VERSION_MIN, PRO_VERSION_MAX);
2902 pr_info("%s\n", drbd_buildtag());
2903 pr_info("registered as block device major %d\n", DRBD_MAJOR);
2904 return 0; /* Success! */
2909 pr_err("ran out of memory\n");
2911 pr_err("initialization failure\n");
2915 static void drbd_free_one_sock(struct drbd_socket *ds)
2918 mutex_lock(&ds->mutex);
2921 mutex_unlock(&ds->mutex);
2923 /* so debugfs does not need to mutex_lock() */
2925 kernel_sock_shutdown(s, SHUT_RDWR);
2930 void drbd_free_sock(struct drbd_connection *connection)
2932 if (connection->data.socket)
2933 drbd_free_one_sock(&connection->data);
2934 if (connection->meta.socket)
2935 drbd_free_one_sock(&connection->meta);
2938 /* meta data management */
2940 void conn_md_sync(struct drbd_connection *connection)
2942 struct drbd_peer_device *peer_device;
2946 idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
2947 struct drbd_device *device = peer_device->device;
2949 kref_get(&device->kref);
2951 drbd_md_sync(device);
2952 kref_put(&device->kref, drbd_destroy_device);
2958 /* aligned 4kByte */
2959 struct meta_data_on_disk {
2960 u64 la_size_sect; /* last agreed size. */
2961 u64 uuid[UI_SIZE]; /* UUIDs. */
2964 u32 flags; /* MDF */
2967 u32 al_offset; /* offset to this block */
2968 u32 al_nr_extents; /* important for restoring the AL (userspace) */
2969 /* `-- act_log->nr_elements <-- ldev->dc.al_extents */
2970 u32 bm_offset; /* offset to the bitmap, from here */
2971 u32 bm_bytes_per_bit; /* BM_BLOCK_SIZE */
2972 u32 la_peer_max_bio_size; /* last peer max_bio_size */
2974 /* see al_tr_number_to_on_disk_sector() */
2976 u32 al_stripe_size_4k;
2978 u8 reserved_u8[4096 - (7*8 + 10*4)];
2983 void drbd_md_write(struct drbd_device *device, void *b)
2985 struct meta_data_on_disk *buffer = b;
2989 memset(buffer, 0, sizeof(*buffer));
2991 buffer->la_size_sect = cpu_to_be64(get_capacity(device->vdisk));
2992 for (i = UI_CURRENT; i < UI_SIZE; i++)
2993 buffer->uuid[i] = cpu_to_be64(device->ldev->md.uuid[i]);
2994 buffer->flags = cpu_to_be32(device->ldev->md.flags);
2995 buffer->magic = cpu_to_be32(DRBD_MD_MAGIC_84_UNCLEAN);
2997 buffer->md_size_sect = cpu_to_be32(device->ldev->md.md_size_sect);
2998 buffer->al_offset = cpu_to_be32(device->ldev->md.al_offset);
2999 buffer->al_nr_extents = cpu_to_be32(device->act_log->nr_elements);
3000 buffer->bm_bytes_per_bit = cpu_to_be32(BM_BLOCK_SIZE);
3001 buffer->device_uuid = cpu_to_be64(device->ldev->md.device_uuid);
3003 buffer->bm_offset = cpu_to_be32(device->ldev->md.bm_offset);
3004 buffer->la_peer_max_bio_size = cpu_to_be32(device->peer_max_bio_size);
3006 buffer->al_stripes = cpu_to_be32(device->ldev->md.al_stripes);
3007 buffer->al_stripe_size_4k = cpu_to_be32(device->ldev->md.al_stripe_size_4k);
3009 D_ASSERT(device, drbd_md_ss(device->ldev) == device->ldev->md.md_offset);
3010 sector = device->ldev->md.md_offset;
3012 if (drbd_md_sync_page_io(device, device->ldev, sector, REQ_OP_WRITE)) {
3013 /* this was a try anyways ... */
3014 drbd_err(device, "meta data update failed!\n");
3015 drbd_chk_io_error(device, 1, DRBD_META_IO_ERROR);
3020 * drbd_md_sync() - Writes the meta data super block if the MD_DIRTY flag bit is set
3021 * @device: DRBD device.
3023 void drbd_md_sync(struct drbd_device *device)
3025 struct meta_data_on_disk *buffer;
3027 /* Don't accidentally change the DRBD meta data layout. */
3028 BUILD_BUG_ON(UI_SIZE != 4);
3029 BUILD_BUG_ON(sizeof(struct meta_data_on_disk) != 4096);
3031 del_timer(&device->md_sync_timer);
3032 /* timer may be rearmed by drbd_md_mark_dirty() now. */
3033 if (!test_and_clear_bit(MD_DIRTY, &device->flags))
3036 /* We use here D_FAILED and not D_ATTACHING because we try to write
3037 * metadata even if we detach due to a disk failure! */
3038 if (!get_ldev_if_state(device, D_FAILED))
3041 buffer = drbd_md_get_buffer(device, __func__);
3045 drbd_md_write(device, buffer);
3047 /* Update device->ldev->md.la_size_sect,
3048 * since we updated it on metadata. */
3049 device->ldev->md.la_size_sect = get_capacity(device->vdisk);
3051 drbd_md_put_buffer(device);
3056 static int check_activity_log_stripe_size(struct drbd_device *device,
3057 struct meta_data_on_disk *on_disk,
3058 struct drbd_md *in_core)
3060 u32 al_stripes = be32_to_cpu(on_disk->al_stripes);
3061 u32 al_stripe_size_4k = be32_to_cpu(on_disk->al_stripe_size_4k);
3064 /* both not set: default to old fixed size activity log */
3065 if (al_stripes == 0 && al_stripe_size_4k == 0) {
3067 al_stripe_size_4k = MD_32kB_SECT/8;
3070 /* some paranoia plausibility checks */
3072 /* we need both values to be set */
3073 if (al_stripes == 0 || al_stripe_size_4k == 0)
3076 al_size_4k = (u64)al_stripes * al_stripe_size_4k;
3078 /* Upper limit of activity log area, to avoid potential overflow
3079 * problems in al_tr_number_to_on_disk_sector(). As right now, more
3080 * than 72 * 4k blocks total only increases the amount of history,
3081 * limiting this arbitrarily to 16 GB is not a real limitation ;-) */
3082 if (al_size_4k > (16 * 1024 * 1024/4))
3085 /* Lower limit: we need at least 8 transaction slots (32kB)
3086 * to not break existing setups */
3087 if (al_size_4k < MD_32kB_SECT/8)
3090 in_core->al_stripe_size_4k = al_stripe_size_4k;
3091 in_core->al_stripes = al_stripes;
3092 in_core->al_size_4k = al_size_4k;
3096 drbd_err(device, "invalid activity log striping: al_stripes=%u, al_stripe_size_4k=%u\n",
3097 al_stripes, al_stripe_size_4k);
3101 static int check_offsets_and_sizes(struct drbd_device *device, struct drbd_backing_dev *bdev)
3103 sector_t capacity = drbd_get_capacity(bdev->md_bdev);
3104 struct drbd_md *in_core = &bdev->md;
3105 s32 on_disk_al_sect;
3106 s32 on_disk_bm_sect;
3108 /* The on-disk size of the activity log, calculated from offsets, and
3109 * the size of the activity log calculated from the stripe settings,
3111 * Though we could relax this a bit: it is ok, if the striped activity log
3112 * fits in the available on-disk activity log size.
3113 * Right now, that would break how resize is implemented.
3114 * TODO: make drbd_determine_dev_size() (and the drbdmeta tool) aware
3115 * of possible unused padding space in the on disk layout. */
3116 if (in_core->al_offset < 0) {
3117 if (in_core->bm_offset > in_core->al_offset)
3119 on_disk_al_sect = -in_core->al_offset;
3120 on_disk_bm_sect = in_core->al_offset - in_core->bm_offset;
3122 if (in_core->al_offset != MD_4kB_SECT)
3124 if (in_core->bm_offset < in_core->al_offset + in_core->al_size_4k * MD_4kB_SECT)
3127 on_disk_al_sect = in_core->bm_offset - MD_4kB_SECT;
3128 on_disk_bm_sect = in_core->md_size_sect - in_core->bm_offset;
3131 /* old fixed size meta data is exactly that: fixed. */
3132 if (in_core->meta_dev_idx >= 0) {
3133 if (in_core->md_size_sect != MD_128MB_SECT
3134 || in_core->al_offset != MD_4kB_SECT
3135 || in_core->bm_offset != MD_4kB_SECT + MD_32kB_SECT
3136 || in_core->al_stripes != 1
3137 || in_core->al_stripe_size_4k != MD_32kB_SECT/8)
3141 if (capacity < in_core->md_size_sect)
3143 if (capacity - in_core->md_size_sect < drbd_md_first_sector(bdev))
3146 /* should be aligned, and at least 32k */
3147 if ((on_disk_al_sect & 7) || (on_disk_al_sect < MD_32kB_SECT))
3150 /* should fit (for now: exactly) into the available on-disk space;
3151 * overflow prevention is in check_activity_log_stripe_size() above. */
3152 if (on_disk_al_sect != in_core->al_size_4k * MD_4kB_SECT)
3155 /* again, should be aligned */
3156 if (in_core->bm_offset & 7)
3159 /* FIXME check for device grow with flex external meta data? */
3161 /* can the available bitmap space cover the last agreed device size? */
3162 if (on_disk_bm_sect < (in_core->la_size_sect+7)/MD_4kB_SECT/8/512)
3168 drbd_err(device, "meta data offsets don't make sense: idx=%d "
3169 "al_s=%u, al_sz4k=%u, al_offset=%d, bm_offset=%d, "
3170 "md_size_sect=%u, la_size=%llu, md_capacity=%llu\n",
3171 in_core->meta_dev_idx,
3172 in_core->al_stripes, in_core->al_stripe_size_4k,
3173 in_core->al_offset, in_core->bm_offset, in_core->md_size_sect,
3174 (unsigned long long)in_core->la_size_sect,
3175 (unsigned long long)capacity);
3182 * drbd_md_read() - Reads in the meta data super block
3183 * @device: DRBD device.
3184 * @bdev: Device from which the meta data should be read in.
3186 * Return NO_ERROR on success, and an enum drbd_ret_code in case
3187 * something goes wrong.
3189 * Called exactly once during drbd_adm_attach(), while still being D_DISKLESS,
3190 * even before @bdev is assigned to @device->ldev.
3192 int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev)
3194 struct meta_data_on_disk *buffer;
3196 int i, rv = NO_ERROR;
3198 if (device->state.disk != D_DISKLESS)
3199 return ERR_DISK_CONFIGURED;
3201 buffer = drbd_md_get_buffer(device, __func__);
3205 /* First, figure out where our meta data superblock is located,
3207 bdev->md.meta_dev_idx = bdev->disk_conf->meta_dev_idx;
3208 bdev->md.md_offset = drbd_md_ss(bdev);
3209 /* Even for (flexible or indexed) external meta data,
3210 * initially restrict us to the 4k superblock for now.
3211 * Affects the paranoia out-of-range access check in drbd_md_sync_page_io(). */
3212 bdev->md.md_size_sect = 8;
3214 if (drbd_md_sync_page_io(device, bdev, bdev->md.md_offset,
3216 /* NOTE: can't do normal error processing here as this is
3217 called BEFORE disk is attached */
3218 drbd_err(device, "Error while reading metadata.\n");
3219 rv = ERR_IO_MD_DISK;
3223 magic = be32_to_cpu(buffer->magic);
3224 flags = be32_to_cpu(buffer->flags);
3225 if (magic == DRBD_MD_MAGIC_84_UNCLEAN ||
3226 (magic == DRBD_MD_MAGIC_08 && !(flags & MDF_AL_CLEAN))) {
3227 /* btw: that's Activity Log clean, not "all" clean. */
3228 drbd_err(device, "Found unclean meta data. Did you \"drbdadm apply-al\"?\n");
3229 rv = ERR_MD_UNCLEAN;
3233 rv = ERR_MD_INVALID;
3234 if (magic != DRBD_MD_MAGIC_08) {
3235 if (magic == DRBD_MD_MAGIC_07)
3236 drbd_err(device, "Found old (0.7) meta data magic. Did you \"drbdadm create-md\"?\n");
3238 drbd_err(device, "Meta data magic not found. Did you \"drbdadm create-md\"?\n");
3242 if (be32_to_cpu(buffer->bm_bytes_per_bit) != BM_BLOCK_SIZE) {
3243 drbd_err(device, "unexpected bm_bytes_per_bit: %u (expected %u)\n",
3244 be32_to_cpu(buffer->bm_bytes_per_bit), BM_BLOCK_SIZE);
3249 /* convert to in_core endian */
3250 bdev->md.la_size_sect = be64_to_cpu(buffer->la_size_sect);
3251 for (i = UI_CURRENT; i < UI_SIZE; i++)
3252 bdev->md.uuid[i] = be64_to_cpu(buffer->uuid[i]);
3253 bdev->md.flags = be32_to_cpu(buffer->flags);
3254 bdev->md.device_uuid = be64_to_cpu(buffer->device_uuid);
3256 bdev->md.md_size_sect = be32_to_cpu(buffer->md_size_sect);
3257 bdev->md.al_offset = be32_to_cpu(buffer->al_offset);
3258 bdev->md.bm_offset = be32_to_cpu(buffer->bm_offset);
3260 if (check_activity_log_stripe_size(device, buffer, &bdev->md))
3262 if (check_offsets_and_sizes(device, bdev))
3265 if (be32_to_cpu(buffer->bm_offset) != bdev->md.bm_offset) {
3266 drbd_err(device, "unexpected bm_offset: %d (expected %d)\n",
3267 be32_to_cpu(buffer->bm_offset), bdev->md.bm_offset);
3270 if (be32_to_cpu(buffer->md_size_sect) != bdev->md.md_size_sect) {
3271 drbd_err(device, "unexpected md_size: %u (expected %u)\n",
3272 be32_to_cpu(buffer->md_size_sect), bdev->md.md_size_sect);
3278 spin_lock_irq(&device->resource->req_lock);
3279 if (device->state.conn < C_CONNECTED) {
3281 peer = be32_to_cpu(buffer->la_peer_max_bio_size);
3282 peer = max(peer, DRBD_MAX_BIO_SIZE_SAFE);
3283 device->peer_max_bio_size = peer;
3285 spin_unlock_irq(&device->resource->req_lock);
3288 drbd_md_put_buffer(device);
3294 * drbd_md_mark_dirty() - Mark meta data super block as dirty
3295 * @device: DRBD device.
3297 * Call this function if you change anything that should be written to
3298 * the meta-data super block. This function sets MD_DIRTY, and starts a
3299 * timer that ensures that within five seconds you have to call drbd_md_sync().
3301 void drbd_md_mark_dirty(struct drbd_device *device)
3303 if (!test_and_set_bit(MD_DIRTY, &device->flags))
3304 mod_timer(&device->md_sync_timer, jiffies + 5*HZ);
3307 void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local)
3311 for (i = UI_HISTORY_START; i < UI_HISTORY_END; i++)
3312 device->ldev->md.uuid[i+1] = device->ldev->md.uuid[i];
3315 void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local)
3317 if (idx == UI_CURRENT) {
3318 if (device->state.role == R_PRIMARY)
3323 drbd_set_ed_uuid(device, val);
3326 device->ldev->md.uuid[idx] = val;
3327 drbd_md_mark_dirty(device);
3330 void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local)
3332 unsigned long flags;
3333 spin_lock_irqsave(&device->ldev->md.uuid_lock, flags);
3334 __drbd_uuid_set(device, idx, val);
3335 spin_unlock_irqrestore(&device->ldev->md.uuid_lock, flags);
3338 void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local)
3340 unsigned long flags;
3341 spin_lock_irqsave(&device->ldev->md.uuid_lock, flags);
3342 if (device->ldev->md.uuid[idx]) {
3343 drbd_uuid_move_history(device);
3344 device->ldev->md.uuid[UI_HISTORY_START] = device->ldev->md.uuid[idx];
3346 __drbd_uuid_set(device, idx, val);
3347 spin_unlock_irqrestore(&device->ldev->md.uuid_lock, flags);
3351 * drbd_uuid_new_current() - Creates a new current UUID
3352 * @device: DRBD device.
3354 * Creates a new current UUID, and rotates the old current UUID into
3355 * the bitmap slot. Causes an incremental resync upon next connect.
3357 void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local)
3360 unsigned long long bm_uuid;
3362 get_random_bytes(&val, sizeof(u64));
3364 spin_lock_irq(&device->ldev->md.uuid_lock);
3365 bm_uuid = device->ldev->md.uuid[UI_BITMAP];
3368 drbd_warn(device, "bm UUID was already set: %llX\n", bm_uuid);
3370 device->ldev->md.uuid[UI_BITMAP] = device->ldev->md.uuid[UI_CURRENT];
3371 __drbd_uuid_set(device, UI_CURRENT, val);
3372 spin_unlock_irq(&device->ldev->md.uuid_lock);
3374 drbd_print_uuids(device, "new current UUID");
3375 /* get it to stable storage _now_ */
3376 drbd_md_sync(device);
3379 void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local)
3381 unsigned long flags;
3382 if (device->ldev->md.uuid[UI_BITMAP] == 0 && val == 0)
3385 spin_lock_irqsave(&device->ldev->md.uuid_lock, flags);
3387 drbd_uuid_move_history(device);
3388 device->ldev->md.uuid[UI_HISTORY_START] = device->ldev->md.uuid[UI_BITMAP];
3389 device->ldev->md.uuid[UI_BITMAP] = 0;
3391 unsigned long long bm_uuid = device->ldev->md.uuid[UI_BITMAP];
3393 drbd_warn(device, "bm UUID was already set: %llX\n", bm_uuid);
3395 device->ldev->md.uuid[UI_BITMAP] = val & ~((u64)1);
3397 spin_unlock_irqrestore(&device->ldev->md.uuid_lock, flags);
3399 drbd_md_mark_dirty(device);
3403 * drbd_bmio_set_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io()
3404 * @device: DRBD device.
3406 * Sets all bits in the bitmap and writes the whole bitmap to stable storage.
3408 int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local)
3412 drbd_md_set_flag(device, MDF_FULL_SYNC);
3413 drbd_md_sync(device);
3414 drbd_bm_set_all(device);
3416 rv = drbd_bm_write(device);
3419 drbd_md_clear_flag(device, MDF_FULL_SYNC);
3420 drbd_md_sync(device);
3427 * drbd_bmio_clear_n_write() - io_fn for drbd_queue_bitmap_io() or drbd_bitmap_io()
3428 * @device: DRBD device.
3430 * Clears all bits in the bitmap and writes the whole bitmap to stable storage.
3432 int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local)
3434 drbd_resume_al(device);
3435 drbd_bm_clear_all(device);
3436 return drbd_bm_write(device);
3439 static int w_bitmap_io(struct drbd_work *w, int unused)
3441 struct drbd_device *device =
3442 container_of(w, struct drbd_device, bm_io_work.w);
3443 struct bm_io_work *work = &device->bm_io_work;
3446 if (work->flags != BM_LOCKED_CHANGE_ALLOWED) {
3447 int cnt = atomic_read(&device->ap_bio_cnt);
3449 drbd_err(device, "FIXME: ap_bio_cnt %d, expected 0; queued for '%s'\n",
3453 if (get_ldev(device)) {
3454 drbd_bm_lock(device, work->why, work->flags);
3455 rv = work->io_fn(device);
3456 drbd_bm_unlock(device);
3460 clear_bit_unlock(BITMAP_IO, &device->flags);
3461 wake_up(&device->misc_wait);
3464 work->done(device, rv);
3466 clear_bit(BITMAP_IO_QUEUED, &device->flags);
3474 * drbd_queue_bitmap_io() - Queues an IO operation on the whole bitmap
3475 * @device: DRBD device.
3476 * @io_fn: IO callback to be called when bitmap IO is possible
3477 * @done: callback to be called after the bitmap IO was performed
3478 * @why: Descriptive text of the reason for doing the IO
3479 * @flags: Bitmap flags
3481 * While IO on the bitmap happens we freeze application IO thus we ensure
3482 * that drbd_set_out_of_sync() can not be called. This function MAY ONLY be
3483 * called from worker context. It MUST NOT be used while a previous such
3484 * work is still pending!
3486 * Its worker function encloses the call of io_fn() by get_ldev() and
3489 void drbd_queue_bitmap_io(struct drbd_device *device,
3490 int (*io_fn)(struct drbd_device *),
3491 void (*done)(struct drbd_device *, int),
3492 char *why, enum bm_flag flags)
3494 D_ASSERT(device, current == first_peer_device(device)->connection->worker.task);
3496 D_ASSERT(device, !test_bit(BITMAP_IO_QUEUED, &device->flags));
3497 D_ASSERT(device, !test_bit(BITMAP_IO, &device->flags));
3498 D_ASSERT(device, list_empty(&device->bm_io_work.w.list));
3499 if (device->bm_io_work.why)
3500 drbd_err(device, "FIXME going to queue '%s' but '%s' still pending?\n",
3501 why, device->bm_io_work.why);
3503 device->bm_io_work.io_fn = io_fn;
3504 device->bm_io_work.done = done;
3505 device->bm_io_work.why = why;
3506 device->bm_io_work.flags = flags;
3508 spin_lock_irq(&device->resource->req_lock);
3509 set_bit(BITMAP_IO, &device->flags);
3510 /* don't wait for pending application IO if the caller indicates that
3511 * application IO does not conflict anyways. */
3512 if (flags == BM_LOCKED_CHANGE_ALLOWED || atomic_read(&device->ap_bio_cnt) == 0) {
3513 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
3514 drbd_queue_work(&first_peer_device(device)->connection->sender_work,
3515 &device->bm_io_work.w);
3517 spin_unlock_irq(&device->resource->req_lock);
3521 * drbd_bitmap_io() - Does an IO operation on the whole bitmap
3522 * @device: DRBD device.
3523 * @io_fn: IO callback to be called when bitmap IO is possible
3524 * @why: Descriptive text of the reason for doing the IO
3525 * @flags: Bitmap flags
3527 * freezes application IO while that the actual IO operations runs. This
3528 * functions MAY NOT be called from worker context.
3530 int drbd_bitmap_io(struct drbd_device *device, int (*io_fn)(struct drbd_device *),
3531 char *why, enum bm_flag flags)
3533 /* Only suspend io, if some operation is supposed to be locked out */
3534 const bool do_suspend_io = flags & (BM_DONT_CLEAR|BM_DONT_SET|BM_DONT_TEST);
3537 D_ASSERT(device, current != first_peer_device(device)->connection->worker.task);
3540 drbd_suspend_io(device);
3542 drbd_bm_lock(device, why, flags);
3544 drbd_bm_unlock(device);
3547 drbd_resume_io(device);
3552 void drbd_md_set_flag(struct drbd_device *device, int flag) __must_hold(local)
3554 if ((device->ldev->md.flags & flag) != flag) {
3555 drbd_md_mark_dirty(device);
3556 device->ldev->md.flags |= flag;
3560 void drbd_md_clear_flag(struct drbd_device *device, int flag) __must_hold(local)
3562 if ((device->ldev->md.flags & flag) != 0) {
3563 drbd_md_mark_dirty(device);
3564 device->ldev->md.flags &= ~flag;
3567 int drbd_md_test_flag(struct drbd_backing_dev *bdev, int flag)
3569 return (bdev->md.flags & flag) != 0;
3572 static void md_sync_timer_fn(struct timer_list *t)
3574 struct drbd_device *device = from_timer(device, t, md_sync_timer);
3575 drbd_device_post_work(device, MD_SYNC);
3578 const char *cmdname(enum drbd_packet cmd)
3580 /* THINK may need to become several global tables
3581 * when we want to support more than
3582 * one PRO_VERSION */
3583 static const char *cmdnames[] = {
3586 [P_DATA_REPLY] = "DataReply",
3587 [P_RS_DATA_REPLY] = "RSDataReply",
3588 [P_BARRIER] = "Barrier",
3589 [P_BITMAP] = "ReportBitMap",
3590 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
3591 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
3592 [P_UNPLUG_REMOTE] = "UnplugRemote",
3593 [P_DATA_REQUEST] = "DataRequest",
3594 [P_RS_DATA_REQUEST] = "RSDataRequest",
3595 [P_SYNC_PARAM] = "SyncParam",
3596 [P_PROTOCOL] = "ReportProtocol",
3597 [P_UUIDS] = "ReportUUIDs",
3598 [P_SIZES] = "ReportSizes",
3599 [P_STATE] = "ReportState",
3600 [P_SYNC_UUID] = "ReportSyncUUID",
3601 [P_AUTH_CHALLENGE] = "AuthChallenge",
3602 [P_AUTH_RESPONSE] = "AuthResponse",
3603 [P_STATE_CHG_REQ] = "StateChgRequest",
3605 [P_PING_ACK] = "PingAck",
3606 [P_RECV_ACK] = "RecvAck",
3607 [P_WRITE_ACK] = "WriteAck",
3608 [P_RS_WRITE_ACK] = "RSWriteAck",
3609 [P_SUPERSEDED] = "Superseded",
3610 [P_NEG_ACK] = "NegAck",
3611 [P_NEG_DREPLY] = "NegDReply",
3612 [P_NEG_RS_DREPLY] = "NegRSDReply",
3613 [P_BARRIER_ACK] = "BarrierAck",
3614 [P_STATE_CHG_REPLY] = "StateChgReply",
3615 [P_OV_REQUEST] = "OVRequest",
3616 [P_OV_REPLY] = "OVReply",
3617 [P_OV_RESULT] = "OVResult",
3618 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
3619 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
3620 [P_SYNC_PARAM89] = "SyncParam89",
3621 [P_COMPRESSED_BITMAP] = "CBitmap",
3622 [P_DELAY_PROBE] = "DelayProbe",
3623 [P_OUT_OF_SYNC] = "OutOfSync",
3624 [P_RS_CANCEL] = "RSCancel",
3625 [P_CONN_ST_CHG_REQ] = "conn_st_chg_req",
3626 [P_CONN_ST_CHG_REPLY] = "conn_st_chg_reply",
3627 [P_PROTOCOL_UPDATE] = "protocol_update",
3629 [P_RS_THIN_REQ] = "rs_thin_req",
3630 [P_RS_DEALLOCATED] = "rs_deallocated",
3631 [P_WSAME] = "WriteSame",
3632 [P_ZEROES] = "Zeroes",
3634 /* enum drbd_packet, but not commands - obsoleted flags:
3640 /* too big for the array: 0xfffX */
3641 if (cmd == P_INITIAL_META)
3642 return "InitialMeta";
3643 if (cmd == P_INITIAL_DATA)
3644 return "InitialData";
3645 if (cmd == P_CONNECTION_FEATURES)
3646 return "ConnectionFeatures";
3647 if (cmd >= ARRAY_SIZE(cmdnames))
3649 return cmdnames[cmd];
3653 * drbd_wait_misc - wait for a request to make progress
3654 * @device: device associated with the request
3655 * @i: the struct drbd_interval embedded in struct drbd_request or
3656 * struct drbd_peer_request
3658 int drbd_wait_misc(struct drbd_device *device, struct drbd_interval *i)
3660 struct net_conf *nc;
3665 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
3670 timeout = nc->ko_count ? nc->timeout * HZ / 10 * nc->ko_count : MAX_SCHEDULE_TIMEOUT;
3673 /* Indicate to wake up device->misc_wait on progress. */
3675 prepare_to_wait(&device->misc_wait, &wait, TASK_INTERRUPTIBLE);
3676 spin_unlock_irq(&device->resource->req_lock);
3677 timeout = schedule_timeout(timeout);
3678 finish_wait(&device->misc_wait, &wait);
3679 spin_lock_irq(&device->resource->req_lock);
3680 if (!timeout || device->state.conn < C_CONNECTED)
3682 if (signal_pending(current))
3683 return -ERESTARTSYS;
3687 void lock_all_resources(void)
3689 struct drbd_resource *resource;
3690 int __maybe_unused i = 0;
3692 mutex_lock(&resources_mutex);
3693 local_irq_disable();
3694 for_each_resource(resource, &drbd_resources)
3695 spin_lock_nested(&resource->req_lock, i++);
3698 void unlock_all_resources(void)
3700 struct drbd_resource *resource;
3702 for_each_resource(resource, &drbd_resources)
3703 spin_unlock(&resource->req_lock);
3705 mutex_unlock(&resources_mutex);
3708 #ifdef CONFIG_DRBD_FAULT_INJECTION
3709 /* Fault insertion support including random number generator shamelessly
3710 * stolen from kernel/rcutorture.c */
3711 struct fault_random_state {
3712 unsigned long state;
3713 unsigned long count;
3716 #define FAULT_RANDOM_MULT 39916801 /* prime */
3717 #define FAULT_RANDOM_ADD 479001701 /* prime */
3718 #define FAULT_RANDOM_REFRESH 10000
3721 * Crude but fast random-number generator. Uses a linear congruential
3722 * generator, with occasional help from get_random_bytes().
3724 static unsigned long
3725 _drbd_fault_random(struct fault_random_state *rsp)
3729 if (!rsp->count--) {
3730 get_random_bytes(&refresh, sizeof(refresh));
3731 rsp->state += refresh;
3732 rsp->count = FAULT_RANDOM_REFRESH;
3734 rsp->state = rsp->state * FAULT_RANDOM_MULT + FAULT_RANDOM_ADD;
3735 return swahw32(rsp->state);
3739 _drbd_fault_str(unsigned int type) {
3740 static char *_faults[] = {
3741 [DRBD_FAULT_MD_WR] = "Meta-data write",
3742 [DRBD_FAULT_MD_RD] = "Meta-data read",
3743 [DRBD_FAULT_RS_WR] = "Resync write",
3744 [DRBD_FAULT_RS_RD] = "Resync read",
3745 [DRBD_FAULT_DT_WR] = "Data write",
3746 [DRBD_FAULT_DT_RD] = "Data read",
3747 [DRBD_FAULT_DT_RA] = "Data read ahead",
3748 [DRBD_FAULT_BM_ALLOC] = "BM allocation",
3749 [DRBD_FAULT_AL_EE] = "EE allocation",
3750 [DRBD_FAULT_RECEIVE] = "receive data corruption",
3753 return (type < DRBD_FAULT_MAX) ? _faults[type] : "**Unknown**";
3757 _drbd_insert_fault(struct drbd_device *device, unsigned int type)
3759 static struct fault_random_state rrs = {0, 0};
3761 unsigned int ret = (
3762 (drbd_fault_devs == 0 ||
3763 ((1 << device_to_minor(device)) & drbd_fault_devs) != 0) &&
3764 (((_drbd_fault_random(&rrs) % 100) + 1) <= drbd_fault_rate));
3769 if (__ratelimit(&drbd_ratelimit_state))
3770 drbd_warn(device, "***Simulating %s failure\n",
3771 _drbd_fault_str(type));
3778 const char *drbd_buildtag(void)
3780 /* DRBD built from external sources has here a reference to the
3781 git hash of the source code. */
3783 static char buildtag[38] = "\0uilt-in";
3785 if (buildtag[0] == 0) {
3787 sprintf(buildtag, "srcversion: %-24s", THIS_MODULE->srcversion);
3796 module_init(drbd_init)
3797 module_exit(drbd_cleanup)
3799 EXPORT_SYMBOL(drbd_conn_str);
3800 EXPORT_SYMBOL(drbd_role_str);
3801 EXPORT_SYMBOL(drbd_disk_str);
3802 EXPORT_SYMBOL(drbd_set_st_err_str);