drbd_receiver.c 155.6 KB
Newer Older
P
Philipp Reisner 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
/*
   drbd_receiver.c

   This file is part of DRBD by Philipp Reisner and Lars Ellenberg.

   Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
   Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
   Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.

   drbd is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.

   drbd is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with drbd; see the file COPYING.  If not, write to
   the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
 */


#include <linux/module.h>

#include <asm/uaccess.h>
#include <net/sock.h>

#include <linux/drbd.h>
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/in.h>
#include <linux/mm.h>
#include <linux/memcontrol.h>
#include <linux/mm_inline.h>
#include <linux/slab.h>
#include <linux/pkt_sched.h>
#define __KERNEL_SYSCALLS__
#include <linux/unistd.h>
#include <linux/vmalloc.h>
#include <linux/random.h>
#include <linux/string.h>
#include <linux/scatterlist.h>
#include "drbd_int.h"
47
#include "drbd_protocol.h"
P
Philipp Reisner 已提交
48 49 50 51
#include "drbd_req.h"

#include "drbd_vli.h"

52 53
struct packet_info {
	enum drbd_packet cmd;
54 55
	unsigned int size;
	unsigned int vnr;
56
	void *data;
57 58
};

P
Philipp Reisner 已提交
59 60 61 62 63 64
enum finish_epoch {
	FE_STILL_LIVE,
	FE_DESTROYED,
	FE_RECYCLED,
};

65 66
static int drbd_do_features(struct drbd_connection *connection);
static int drbd_do_auth(struct drbd_connection *connection);
67
static int drbd_disconnected(struct drbd_peer_device *);
P
Philipp Reisner 已提交
68

69
static enum finish_epoch drbd_may_finish_epoch(struct drbd_connection *, struct drbd_epoch *, enum epoch_event);
70
static int e_end_block(struct drbd_work *, int);
P
Philipp Reisner 已提交
71 72 73 74


#define GFP_TRY	(__GFP_HIGHMEM | __GFP_NOWARN)

75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
/*
 * some helper functions to deal with single linked page lists,
 * page->private being our "next" pointer.
 */

/* If at least n pages are linked at head, get n pages off.
 * Otherwise, don't modify head, and return NULL.
 * Locking is the responsibility of the caller.
 */
static struct page *page_chain_del(struct page **head, int n)
{
	struct page *page;
	struct page *tmp;

	BUG_ON(!n);
	BUG_ON(!head);

	page = *head;
93 94 95 96

	if (!page)
		return NULL;

97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
	while (page) {
		tmp = page_chain_next(page);
		if (--n == 0)
			break; /* found sufficient pages */
		if (tmp == NULL)
			/* insufficient pages, don't use any of them. */
			return NULL;
		page = tmp;
	}

	/* add end of list marker for the returned list */
	set_page_private(page, 0);
	/* actual return value, and adjustment of head */
	page = *head;
	*head = tmp;
	return page;
}

/* may be used outside of locks to find the tail of a (usually short)
 * "private" page chain, before adding it back to a global chain head
 * with page_chain_add() under a spinlock. */
static struct page *page_chain_tail(struct page *page, int *len)
{
	struct page *tmp;
	int i = 1;
	while ((tmp = page_chain_next(page)))
		++i, page = tmp;
	if (len)
		*len = i;
	return page;
}

static int page_chain_free(struct page *page)
{
	struct page *tmp;
	int i = 0;
	page_chain_for_each_safe(page, tmp) {
		put_page(page);
		++i;
	}
	return i;
}

static void page_chain_add(struct page **head,
		struct page *chain_first, struct page *chain_last)
{
#if 1
	struct page *tmp;
	tmp = page_chain_tail(chain_first, NULL);
	BUG_ON(tmp != chain_last);
#endif

	/* add chain to head */
	set_page_private(chain_last, (unsigned long)*head);
	*head = chain_first;
}

154
static struct page *__drbd_alloc_pages(struct drbd_device *device,
155
				       unsigned int number)
P
Philipp Reisner 已提交
156 157
{
	struct page *page = NULL;
158
	struct page *tmp = NULL;
159
	unsigned int i = 0;
P
Philipp Reisner 已提交
160 161 162

	/* Yes, testing drbd_pp_vacant outside the lock is racy.
	 * So what. It saves a spin_lock. */
163
	if (drbd_pp_vacant >= number) {
P
Philipp Reisner 已提交
164
		spin_lock(&drbd_pp_lock);
165 166 167
		page = page_chain_del(&drbd_pp_pool, number);
		if (page)
			drbd_pp_vacant -= number;
P
Philipp Reisner 已提交
168
		spin_unlock(&drbd_pp_lock);
169 170
		if (page)
			return page;
P
Philipp Reisner 已提交
171
	}
172

P
Philipp Reisner 已提交
173 174 175
	/* GFP_TRY, because we must not cause arbitrary write-out: in a DRBD
	 * "criss-cross" setup, that might cause write-out on some other DRBD,
	 * which in turn might block on the other node at this very place.  */
176 177 178 179 180 181 182 183 184 185 186 187
	for (i = 0; i < number; i++) {
		tmp = alloc_page(GFP_TRY);
		if (!tmp)
			break;
		set_page_private(tmp, (unsigned long)page);
		page = tmp;
	}

	if (i == number)
		return page;

	/* Not enough pages immediately available this time.
188
	 * No need to jump around here, drbd_alloc_pages will retry this
189 190 191 192 193 194 195 196 197
	 * function "soon". */
	if (page) {
		tmp = page_chain_tail(page, NULL);
		spin_lock(&drbd_pp_lock);
		page_chain_add(&drbd_pp_pool, page, tmp);
		drbd_pp_vacant += i;
		spin_unlock(&drbd_pp_lock);
	}
	return NULL;
P
Philipp Reisner 已提交
198 199
}

200
static void reclaim_finished_net_peer_reqs(struct drbd_device *device,
201
					   struct list_head *to_be_freed)
P
Philipp Reisner 已提交
202
{
203
	struct drbd_peer_request *peer_req, *tmp;
P
Philipp Reisner 已提交
204 205 206 207 208 209

	/* The EEs are always appended to the end of the list. Since
	   they are sent in order over the wire, they have to finish
	   in order. As soon as we see the first not finished we can
	   stop to examine the list... */

210
	list_for_each_entry_safe(peer_req, tmp, &device->net_ee, w.list) {
211
		if (drbd_peer_req_has_active_page(peer_req))
P
Philipp Reisner 已提交
212
			break;
213
		list_move(&peer_req->w.list, to_be_freed);
P
Philipp Reisner 已提交
214 215 216
	}
}

217
static void drbd_kick_lo_and_reclaim_net(struct drbd_device *device)
P
Philipp Reisner 已提交
218 219
{
	LIST_HEAD(reclaimed);
220
	struct drbd_peer_request *peer_req, *t;
P
Philipp Reisner 已提交
221

222
	spin_lock_irq(&device->resource->req_lock);
223
	reclaim_finished_net_peer_reqs(device, &reclaimed);
224
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
225

226
	list_for_each_entry_safe(peer_req, t, &reclaimed, w.list)
227
		drbd_free_net_peer_req(device, peer_req);
P
Philipp Reisner 已提交
228 229 230
}

/**
231
 * drbd_alloc_pages() - Returns @number pages, retries forever (or until signalled)
232
 * @device:	DRBD device.
233 234 235 236
 * @number:	number of pages requested
 * @retry:	whether to retry, if not enough pages are available right now
 *
 * Tries to allocate number pages, first from our own page pool, then from
237
 * the kernel.
238
 * Possibly retry until DRBD frees sufficient pages somewhere else.
P
Philipp Reisner 已提交
239
 *
240 241 242 243 244 245 246 247
 * If this allocation would exceed the max_buffers setting, we throttle
 * allocation (schedule_timeout) to give the system some room to breathe.
 *
 * We do not use max-buffers as hard limit, because it could lead to
 * congestion and further to a distributed deadlock during online-verify or
 * (checksum based) resync, if the max-buffers, socket buffer sizes and
 * resync-rate settings are mis-configured.
 *
248
 * Returns a page chain linked via page->private.
P
Philipp Reisner 已提交
249
 */
250
struct page *drbd_alloc_pages(struct drbd_peer_device *peer_device, unsigned int number,
251
			      bool retry)
P
Philipp Reisner 已提交
252
{
253
	struct drbd_device *device = peer_device->device;
P
Philipp Reisner 已提交
254
	struct page *page = NULL;
255
	struct net_conf *nc;
P
Philipp Reisner 已提交
256
	DEFINE_WAIT(wait);
257
	unsigned int mxb;
P
Philipp Reisner 已提交
258

259
	rcu_read_lock();
260
	nc = rcu_dereference(peer_device->connection->net_conf);
261 262 263
	mxb = nc ? nc->max_buffers : 1000000;
	rcu_read_unlock();

264 265
	if (atomic_read(&device->pp_in_use) < mxb)
		page = __drbd_alloc_pages(device, number);
P
Philipp Reisner 已提交
266

267
	while (page == NULL) {
P
Philipp Reisner 已提交
268 269
		prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE);

270
		drbd_kick_lo_and_reclaim_net(device);
P
Philipp Reisner 已提交
271

272 273
		if (atomic_read(&device->pp_in_use) < mxb) {
			page = __drbd_alloc_pages(device, number);
P
Philipp Reisner 已提交
274 275 276 277 278 279 280 281
			if (page)
				break;
		}

		if (!retry)
			break;

		if (signal_pending(current)) {
282
			drbd_warn(device, "drbd_alloc_pages interrupted!\n");
P
Philipp Reisner 已提交
283 284 285
			break;
		}

286 287
		if (schedule_timeout(HZ/10) == 0)
			mxb = UINT_MAX;
P
Philipp Reisner 已提交
288 289 290
	}
	finish_wait(&drbd_pp_wait, &wait);

291
	if (page)
292
		atomic_add(number, &device->pp_in_use);
P
Philipp Reisner 已提交
293 294 295
	return page;
}

296
/* Must not be used from irq, as that may deadlock: see drbd_alloc_pages.
297
 * Is also used from inside an other spin_lock_irq(&resource->req_lock);
298 299
 * Either links the page chain back to the global pool,
 * or returns all pages to the system. */
300
static void drbd_free_pages(struct drbd_device *device, struct page *page, int is_net)
P
Philipp Reisner 已提交
301
{
302
	atomic_t *a = is_net ? &device->pp_in_use_by_net : &device->pp_in_use;
P
Philipp Reisner 已提交
303
	int i;
304

305 306 307
	if (page == NULL)
		return;

308
	if (drbd_pp_vacant > (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count)
309 310 311 312 313 314 315 316
		i = page_chain_free(page);
	else {
		struct page *tmp;
		tmp = page_chain_tail(page, &i);
		spin_lock(&drbd_pp_lock);
		page_chain_add(&drbd_pp_pool, page, tmp);
		drbd_pp_vacant += i;
		spin_unlock(&drbd_pp_lock);
P
Philipp Reisner 已提交
317
	}
318
	i = atomic_sub_return(i, a);
319
	if (i < 0)
320
		drbd_warn(device, "ASSERTION FAILED: %s: %d < 0\n",
321
			is_net ? "pp_in_use_by_net" : "pp_in_use", i);
P
Philipp Reisner 已提交
322 323 324 325 326 327 328 329
	wake_up(&drbd_pp_wait);
}

/*
You need to hold the req_lock:
 _drbd_wait_ee_list_empty()

You must not have the req_lock:
330
 drbd_free_peer_req()
331
 drbd_alloc_peer_req()
332
 drbd_free_peer_reqs()
P
Philipp Reisner 已提交
333
 drbd_ee_fix_bhs()
334
 drbd_finish_peer_reqs()
P
Philipp Reisner 已提交
335 336 337 338
 drbd_clear_done_ee()
 drbd_wait_ee_list_empty()
*/

339
struct drbd_peer_request *
340
drbd_alloc_peer_req(struct drbd_peer_device *peer_device, u64 id, sector_t sector,
341
		    unsigned int data_size, gfp_t gfp_mask) __must_hold(local)
P
Philipp Reisner 已提交
342
{
343
	struct drbd_device *device = peer_device->device;
344
	struct drbd_peer_request *peer_req;
345
	struct page *page = NULL;
346
	unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT;
P
Philipp Reisner 已提交
347

348
	if (drbd_insert_fault(device, DRBD_FAULT_AL_EE))
P
Philipp Reisner 已提交
349 350
		return NULL;

351 352
	peer_req = mempool_alloc(drbd_ee_mempool, gfp_mask & ~__GFP_HIGHMEM);
	if (!peer_req) {
P
Philipp Reisner 已提交
353
		if (!(gfp_mask & __GFP_NOWARN))
354
			drbd_err(device, "%s: allocation failed\n", __func__);
P
Philipp Reisner 已提交
355 356 357
		return NULL;
	}

358
	if (data_size) {
359
		page = drbd_alloc_pages(peer_device, nr_pages, (gfp_mask & __GFP_WAIT));
360 361 362
		if (!page)
			goto fail;
	}
P
Philipp Reisner 已提交
363

364 365 366 367 368 369 370
	drbd_clear_interval(&peer_req->i);
	peer_req->i.size = data_size;
	peer_req->i.sector = sector;
	peer_req->i.local = false;
	peer_req->i.waiting = false;

	peer_req->epoch = NULL;
371
	peer_req->peer_device = peer_device;
372 373 374
	peer_req->pages = page;
	atomic_set(&peer_req->pending_bios, 0);
	peer_req->flags = 0;
375 376 377 378
	/*
	 * The block_id is opaque to the receiver.  It is not endianness
	 * converted, and sent back to the sender unchanged.
	 */
379
	peer_req->block_id = id;
P
Philipp Reisner 已提交
380

381
	return peer_req;
P
Philipp Reisner 已提交
382

383
 fail:
384
	mempool_free(peer_req, drbd_ee_mempool);
P
Philipp Reisner 已提交
385 386 387
	return NULL;
}

388
void __drbd_free_peer_req(struct drbd_device *device, struct drbd_peer_request *peer_req,
389
		       int is_net)
P
Philipp Reisner 已提交
390
{
391 392
	if (peer_req->flags & EE_HAS_DIGEST)
		kfree(peer_req->digest);
393
	drbd_free_pages(device, peer_req->pages, is_net);
394 395
	D_ASSERT(device, atomic_read(&peer_req->pending_bios) == 0);
	D_ASSERT(device, drbd_interval_empty(&peer_req->i));
396
	mempool_free(peer_req, drbd_ee_mempool);
P
Philipp Reisner 已提交
397 398
}

399
int drbd_free_peer_reqs(struct drbd_device *device, struct list_head *list)
P
Philipp Reisner 已提交
400 401
{
	LIST_HEAD(work_list);
402
	struct drbd_peer_request *peer_req, *t;
P
Philipp Reisner 已提交
403
	int count = 0;
404
	int is_net = list == &device->net_ee;
P
Philipp Reisner 已提交
405

406
	spin_lock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
407
	list_splice_init(list, &work_list);
408
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
409

410
	list_for_each_entry_safe(peer_req, t, &work_list, w.list) {
411
		__drbd_free_peer_req(device, peer_req, is_net);
P
Philipp Reisner 已提交
412 413 414 415 416 417
		count++;
	}
	return count;
}

/*
418
 * See also comments in _req_mod(,BARRIER_ACKED) and receive_Barrier.
P
Philipp Reisner 已提交
419
 */
420
static int drbd_finish_peer_reqs(struct drbd_device *device)
P
Philipp Reisner 已提交
421 422 423
{
	LIST_HEAD(work_list);
	LIST_HEAD(reclaimed);
424
	struct drbd_peer_request *peer_req, *t;
425
	int err = 0;
P
Philipp Reisner 已提交
426

427
	spin_lock_irq(&device->resource->req_lock);
428 429
	reclaim_finished_net_peer_reqs(device, &reclaimed);
	list_splice_init(&device->done_ee, &work_list);
430
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
431

432
	list_for_each_entry_safe(peer_req, t, &reclaimed, w.list)
433
		drbd_free_net_peer_req(device, peer_req);
P
Philipp Reisner 已提交
434 435

	/* possible callbacks here:
436
	 * e_end_block, and e_end_resync_block, e_send_superseded.
P
Philipp Reisner 已提交
437 438
	 * all ignore the last argument.
	 */
439
	list_for_each_entry_safe(peer_req, t, &work_list, w.list) {
440 441
		int err2;

P
Philipp Reisner 已提交
442
		/* list_del not necessary, next/prev members not touched */
443
		err2 = peer_req->w.cb(&peer_req->w, !!err);
444 445
		if (!err)
			err = err2;
446
		drbd_free_peer_req(device, peer_req);
P
Philipp Reisner 已提交
447
	}
448
	wake_up(&device->ee_wait);
P
Philipp Reisner 已提交
449

450
	return err;
P
Philipp Reisner 已提交
451 452
}

453
static void _drbd_wait_ee_list_empty(struct drbd_device *device,
454
				     struct list_head *head)
P
Philipp Reisner 已提交
455 456 457 458 459 460
{
	DEFINE_WAIT(wait);

	/* avoids spin_lock/unlock
	 * and calling prepare_to_wait in the fast path */
	while (!list_empty(head)) {
461
		prepare_to_wait(&device->ee_wait, &wait, TASK_UNINTERRUPTIBLE);
462
		spin_unlock_irq(&device->resource->req_lock);
J
Jens Axboe 已提交
463
		io_schedule();
464
		finish_wait(&device->ee_wait, &wait);
465
		spin_lock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
466 467 468
	}
}

469
static void drbd_wait_ee_list_empty(struct drbd_device *device,
470
				    struct list_head *head)
P
Philipp Reisner 已提交
471
{
472
	spin_lock_irq(&device->resource->req_lock);
473
	_drbd_wait_ee_list_empty(device, head);
474
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
475 476
}

477
static int drbd_recv_short(struct socket *sock, void *buf, size_t size, int flags)
P
Philipp Reisner 已提交
478 479 480 481 482 483 484 485
{
	struct kvec iov = {
		.iov_base = buf,
		.iov_len = size,
	};
	struct msghdr msg = {
		.msg_flags = (flags ? flags : MSG_WAITALL | MSG_NOSIGNAL)
	};
A
Al Viro 已提交
486
	return kernel_recvmsg(sock, &msg, &iov, 1, size, msg.msg_flags);
P
Philipp Reisner 已提交
487 488
}

489
static int drbd_recv(struct drbd_connection *connection, void *buf, size_t size)
P
Philipp Reisner 已提交
490 491 492
{
	int rv;

493
	rv = drbd_recv_short(connection->data.socket, buf, size, 0);
P
Philipp Reisner 已提交
494

P
Philipp Reisner 已提交
495 496
	if (rv < 0) {
		if (rv == -ECONNRESET)
497
			drbd_info(connection, "sock was reset by peer\n");
P
Philipp Reisner 已提交
498
		else if (rv != -ERESTARTSYS)
499
			drbd_err(connection, "sock_recvmsg returned %d\n", rv);
P
Philipp Reisner 已提交
500
	} else if (rv == 0) {
501
		if (test_bit(DISCONNECT_SENT, &connection->flags)) {
502 503
			long t;
			rcu_read_lock();
504
			t = rcu_dereference(connection->net_conf)->ping_timeo * HZ/10;
505 506
			rcu_read_unlock();

507
			t = wait_event_timeout(connection->ping_wait, connection->cstate < C_WF_REPORT_PARAMS, t);
508

509 510 511
			if (t)
				goto out;
		}
512
		drbd_info(connection, "sock was shut down by peer\n");
513 514
	}

P
Philipp Reisner 已提交
515
	if (rv != size)
516
		conn_request_state(connection, NS(conn, C_BROKEN_PIPE), CS_HARD);
P
Philipp Reisner 已提交
517

518
out:
P
Philipp Reisner 已提交
519 520 521
	return rv;
}

522
static int drbd_recv_all(struct drbd_connection *connection, void *buf, size_t size)
523 524 525
{
	int err;

526
	err = drbd_recv(connection, buf, size);
527 528 529 530 531 532 533 534
	if (err != size) {
		if (err >= 0)
			err = -EIO;
	} else
		err = 0;
	return err;
}

535
static int drbd_recv_all_warn(struct drbd_connection *connection, void *buf, size_t size)
536 537 538
{
	int err;

539
	err = drbd_recv_all(connection, buf, size);
540
	if (err && !signal_pending(current))
541
		drbd_warn(connection, "short read (expected size %d)\n", (int)size);
542 543 544
	return err;
}

545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
/* quoting tcp(7):
 *   On individual connections, the socket buffer size must be set prior to the
 *   listen(2) or connect(2) calls in order to have it take effect.
 * This is our wrapper to do so.
 */
static void drbd_setbufsize(struct socket *sock, unsigned int snd,
		unsigned int rcv)
{
	/* open coded SO_SNDBUF, SO_RCVBUF */
	if (snd) {
		sock->sk->sk_sndbuf = snd;
		sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
	}
	if (rcv) {
		sock->sk->sk_rcvbuf = rcv;
		sock->sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
	}
}

564
static struct socket *drbd_try_connect(struct drbd_connection *connection)
P
Philipp Reisner 已提交
565 566 567 568
{
	const char *what;
	struct socket *sock;
	struct sockaddr_in6 src_in6;
569 570 571
	struct sockaddr_in6 peer_in6;
	struct net_conf *nc;
	int err, peer_addr_len, my_addr_len;
572
	int sndbuf_size, rcvbuf_size, connect_int;
P
Philipp Reisner 已提交
573 574
	int disconnect_on_error = 1;

575
	rcu_read_lock();
576
	nc = rcu_dereference(connection->net_conf);
577 578
	if (!nc) {
		rcu_read_unlock();
P
Philipp Reisner 已提交
579
		return NULL;
580 581 582
	}
	sndbuf_size = nc->sndbuf_size;
	rcvbuf_size = nc->rcvbuf_size;
583
	connect_int = nc->connect_int;
584
	rcu_read_unlock();
585

586 587
	my_addr_len = min_t(int, connection->my_addr_len, sizeof(src_in6));
	memcpy(&src_in6, &connection->my_addr, my_addr_len);
588

589
	if (((struct sockaddr *)&connection->my_addr)->sa_family == AF_INET6)
590 591 592 593
		src_in6.sin6_port = 0;
	else
		((struct sockaddr_in *)&src_in6)->sin_port = 0; /* AF_INET & AF_SCI */

594 595
	peer_addr_len = min_t(int, connection->peer_addr_len, sizeof(src_in6));
	memcpy(&peer_in6, &connection->peer_addr, peer_addr_len);
P
Philipp Reisner 已提交
596 597

	what = "sock_create_kern";
598 599
	err = sock_create_kern(((struct sockaddr *)&src_in6)->sa_family,
			       SOCK_STREAM, IPPROTO_TCP, &sock);
P
Philipp Reisner 已提交
600 601 602 603 604 605
	if (err < 0) {
		sock = NULL;
		goto out;
	}

	sock->sk->sk_rcvtimeo =
606
	sock->sk->sk_sndtimeo = connect_int * HZ;
607
	drbd_setbufsize(sock, sndbuf_size, rcvbuf_size);
P
Philipp Reisner 已提交
608 609 610 611 612 613 614 615 616

       /* explicitly bind to the configured IP as source IP
	*  for the outgoing connections.
	*  This is needed for multihomed hosts and to be
	*  able to use lo: interfaces for drbd.
	* Make sure to use 0 as port number, so linux selects
	*  a free one dynamically.
	*/
	what = "bind before connect";
617
	err = sock->ops->bind(sock, (struct sockaddr *) &src_in6, my_addr_len);
P
Philipp Reisner 已提交
618 619 620 621 622 623 624
	if (err < 0)
		goto out;

	/* connect may fail, peer not yet available.
	 * stay C_WF_CONNECTION, don't go Disconnecting! */
	disconnect_on_error = 0;
	what = "connect";
625
	err = sock->ops->connect(sock, (struct sockaddr *) &peer_in6, peer_addr_len, 0);
P
Philipp Reisner 已提交
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642

out:
	if (err < 0) {
		if (sock) {
			sock_release(sock);
			sock = NULL;
		}
		switch (-err) {
			/* timeout, busy, signal pending */
		case ETIMEDOUT: case EAGAIN: case EINPROGRESS:
		case EINTR: case ERESTARTSYS:
			/* peer not (yet) available, network problem */
		case ECONNREFUSED: case ENETUNREACH:
		case EHOSTDOWN:    case EHOSTUNREACH:
			disconnect_on_error = 0;
			break;
		default:
643
			drbd_err(connection, "%s failed, err = %d\n", what, err);
P
Philipp Reisner 已提交
644 645
		}
		if (disconnect_on_error)
646
			conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
647
	}
648

P
Philipp Reisner 已提交
649 650 651
	return sock;
}

652
struct accept_wait_data {
653
	struct drbd_connection *connection;
654 655 656 657 658 659
	struct socket *s_listen;
	struct completion door_bell;
	void (*original_sk_state_change)(struct sock *sk);

};

660
static void drbd_incoming_connection(struct sock *sk)
661 662
{
	struct accept_wait_data *ad = sk->sk_user_data;
663
	void (*state_change)(struct sock *sk);
664

665 666 667 668
	state_change = ad->original_sk_state_change;
	if (sk->sk_state == TCP_ESTABLISHED)
		complete(&ad->door_bell);
	state_change(sk);
669 670
}

671
static int prepare_listen_socket(struct drbd_connection *connection, struct accept_wait_data *ad)
P
Philipp Reisner 已提交
672
{
673
	int err, sndbuf_size, rcvbuf_size, my_addr_len;
674
	struct sockaddr_in6 my_addr;
675
	struct socket *s_listen;
676
	struct net_conf *nc;
P
Philipp Reisner 已提交
677 678
	const char *what;

679
	rcu_read_lock();
680
	nc = rcu_dereference(connection->net_conf);
681 682
	if (!nc) {
		rcu_read_unlock();
683
		return -EIO;
684 685 686 687
	}
	sndbuf_size = nc->sndbuf_size;
	rcvbuf_size = nc->rcvbuf_size;
	rcu_read_unlock();
P
Philipp Reisner 已提交
688

689 690
	my_addr_len = min_t(int, connection->my_addr_len, sizeof(struct sockaddr_in6));
	memcpy(&my_addr, &connection->my_addr, my_addr_len);
P
Philipp Reisner 已提交
691 692

	what = "sock_create_kern";
693
	err = sock_create_kern(((struct sockaddr *)&my_addr)->sa_family,
694
			       SOCK_STREAM, IPPROTO_TCP, &s_listen);
P
Philipp Reisner 已提交
695 696 697 698 699
	if (err) {
		s_listen = NULL;
		goto out;
	}

700
	s_listen->sk->sk_reuse = SK_CAN_REUSE; /* SO_REUSEADDR */
701
	drbd_setbufsize(s_listen, sndbuf_size, rcvbuf_size);
P
Philipp Reisner 已提交
702 703

	what = "bind before listen";
704
	err = s_listen->ops->bind(s_listen, (struct sockaddr *)&my_addr, my_addr_len);
P
Philipp Reisner 已提交
705 706 707
	if (err < 0)
		goto out;

708 709 710
	ad->s_listen = s_listen;
	write_lock_bh(&s_listen->sk->sk_callback_lock);
	ad->original_sk_state_change = s_listen->sk->sk_state_change;
711
	s_listen->sk->sk_state_change = drbd_incoming_connection;
712 713
	s_listen->sk->sk_user_data = ad;
	write_unlock_bh(&s_listen->sk->sk_callback_lock);
P
Philipp Reisner 已提交
714

715 716 717 718 719
	what = "listen";
	err = s_listen->ops->listen(s_listen, 5);
	if (err < 0)
		goto out;

720
	return 0;
P
Philipp Reisner 已提交
721 722 723 724 725
out:
	if (s_listen)
		sock_release(s_listen);
	if (err < 0) {
		if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) {
726
			drbd_err(connection, "%s failed, err = %d\n", what, err);
727
			conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
728 729 730
		}
	}

731
	return -EIO;
P
Philipp Reisner 已提交
732 733
}

734
static void unregister_state_change(struct sock *sk, struct accept_wait_data *ad)
P
Philipp Reisner 已提交
735
{
736 737 738 739
	write_lock_bh(&sk->sk_callback_lock);
	sk->sk_state_change = ad->original_sk_state_change;
	sk->sk_user_data = NULL;
	write_unlock_bh(&sk->sk_callback_lock);
P
Philipp Reisner 已提交
740 741
}

742
static struct socket *drbd_wait_for_connect(struct drbd_connection *connection, struct accept_wait_data *ad)
P
Philipp Reisner 已提交
743
{
744 745 746 747 748
	int timeo, connect_int, err = 0;
	struct socket *s_estab = NULL;
	struct net_conf *nc;

	rcu_read_lock();
749
	nc = rcu_dereference(connection->net_conf);
750 751 752 753 754 755 756 757
	if (!nc) {
		rcu_read_unlock();
		return NULL;
	}
	connect_int = nc->connect_int;
	rcu_read_unlock();

	timeo = connect_int * HZ;
758 759
	/* 28.5% random jitter */
	timeo += (prandom_u32() & 1) ? timeo / 7 : -timeo / 7;
760

761 762 763
	err = wait_for_completion_interruptible_timeout(&ad->door_bell, timeo);
	if (err <= 0)
		return NULL;
P
Philipp Reisner 已提交
764

765
	err = kernel_accept(ad->s_listen, &s_estab, 0);
P
Philipp Reisner 已提交
766 767
	if (err < 0) {
		if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) {
768
			drbd_err(connection, "accept failed, err = %d\n", err);
769
			conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
770 771 772
		}
	}

773 774
	if (s_estab)
		unregister_state_change(s_estab->sk, ad);
P
Philipp Reisner 已提交
775 776 777 778

	return s_estab;
}

779
static int decode_header(struct drbd_connection *, void *, struct packet_info *);
P
Philipp Reisner 已提交
780

781
static int send_first_packet(struct drbd_connection *connection, struct drbd_socket *sock,
782 783
			     enum drbd_packet cmd)
{
784
	if (!conn_prepare_command(connection, sock))
785
		return -EIO;
786
	return conn_send_command(connection, sock, cmd, 0, NULL, 0);
P
Philipp Reisner 已提交
787 788
}

789
static int receive_first_packet(struct drbd_connection *connection, struct socket *sock)
P
Philipp Reisner 已提交
790
{
791
	unsigned int header_size = drbd_header_size(connection);
792 793
	struct packet_info pi;
	int err;
P
Philipp Reisner 已提交
794

795
	err = drbd_recv_short(sock, connection->data.rbuf, header_size, 0);
796 797 798 799 800
	if (err != header_size) {
		if (err >= 0)
			err = -EIO;
		return err;
	}
801
	err = decode_header(connection, connection->data.rbuf, &pi);
802 803 804
	if (err)
		return err;
	return pi.cmd;
P
Philipp Reisner 已提交
805 806 807 808 809 810
}

/**
 * drbd_socket_okay() - Free the socket if its connection is not okay
 * @sock:	pointer to the pointer to the socket.
 */
811
static int drbd_socket_okay(struct socket **sock)
P
Philipp Reisner 已提交
812 813 814 815 816
{
	int rr;
	char tb[4];

	if (!*sock)
817
		return false;
P
Philipp Reisner 已提交
818

819
	rr = drbd_recv_short(*sock, tb, 4, MSG_DONTWAIT | MSG_PEEK);
P
Philipp Reisner 已提交
820 821

	if (rr > 0 || rr == -EAGAIN) {
822
		return true;
P
Philipp Reisner 已提交
823 824 825
	} else {
		sock_release(*sock);
		*sock = NULL;
826
		return false;
P
Philipp Reisner 已提交
827 828
	}
}
829 830
/* Gets called if a connection is established, or if a new minor gets created
   in a connection */
831
int drbd_connected(struct drbd_peer_device *peer_device)
832
{
833
	struct drbd_device *device = peer_device->device;
834
	int err;
835

836 837
	atomic_set(&device->packet_seq, 0);
	device->peer_seq = 0;
838

839 840
	device->state_mutex = peer_device->connection->agreed_pro_version < 100 ?
		&peer_device->connection->cstate_mutex :
841
		&device->own_state_mutex;
842

843
	err = drbd_send_sync_param(peer_device);
844
	if (!err)
845
		err = drbd_send_sizes(peer_device, 0, 0);
846
	if (!err)
847
		err = drbd_send_uuids(peer_device);
848
	if (!err)
849
		err = drbd_send_current_state(peer_device);
850 851 852 853
	clear_bit(USE_DEGR_WFC_T, &device->flags);
	clear_bit(RESIZE_PENDING, &device->flags);
	atomic_set(&device->ap_in_flight, 0);
	mod_timer(&device->request_timer, jiffies + HZ); /* just start it here. */
854
	return err;
855
}
P
Philipp Reisner 已提交
856 857 858 859 860 861 862 863 864

/*
 * return values:
 *   1 yes, we have a valid connection
 *   0 oops, did not work out, please try again
 *  -1 peer talks different language,
 *     no point in trying again, please go standalone.
 *  -2 We do not have a network config...
 */
865
static int conn_connect(struct drbd_connection *connection)
P
Philipp Reisner 已提交
866
{
867
	struct drbd_socket sock, msock;
868
	struct drbd_peer_device *peer_device;
869
	struct net_conf *nc;
870
	int vnr, timeout, h, ok;
871
	bool discard_my_data;
872
	enum drbd_state_rv rv;
873
	struct accept_wait_data ad = {
874
		.connection = connection,
875 876
		.door_bell = COMPLETION_INITIALIZER_ONSTACK(ad.door_bell),
	};
P
Philipp Reisner 已提交
877

878 879
	clear_bit(DISCONNECT_SENT, &connection->flags);
	if (conn_request_state(connection, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS)
P
Philipp Reisner 已提交
880 881
		return -2;

882
	mutex_init(&sock.mutex);
883 884
	sock.sbuf = connection->data.sbuf;
	sock.rbuf = connection->data.rbuf;
885 886
	sock.socket = NULL;
	mutex_init(&msock.mutex);
887 888
	msock.sbuf = connection->meta.sbuf;
	msock.rbuf = connection->meta.rbuf;
889 890
	msock.socket = NULL;

891
	/* Assume that the peer only understands protocol 80 until we know better.  */
892
	connection->agreed_pro_version = 80;
P
Philipp Reisner 已提交
893

894
	if (prepare_listen_socket(connection, &ad))
895
		return 0;
P
Philipp Reisner 已提交
896 897

	do {
898
		struct socket *s;
P
Philipp Reisner 已提交
899

900
		s = drbd_try_connect(connection);
P
Philipp Reisner 已提交
901
		if (s) {
902 903
			if (!sock.socket) {
				sock.socket = s;
904
				send_first_packet(connection, &sock, P_INITIAL_DATA);
905
			} else if (!msock.socket) {
906
				clear_bit(RESOLVE_CONFLICTS, &connection->flags);
907
				msock.socket = s;
908
				send_first_packet(connection, &msock, P_INITIAL_META);
P
Philipp Reisner 已提交
909
			} else {
910
				drbd_err(connection, "Logic error in conn_connect()\n");
P
Philipp Reisner 已提交
911 912 913 914
				goto out_release_sockets;
			}
		}

915 916
		if (sock.socket && msock.socket) {
			rcu_read_lock();
917
			nc = rcu_dereference(connection->net_conf);
918 919 920 921 922
			timeout = nc->ping_timeo * HZ / 10;
			rcu_read_unlock();
			schedule_timeout_interruptible(timeout);
			ok = drbd_socket_okay(&sock.socket);
			ok = drbd_socket_okay(&msock.socket) && ok;
P
Philipp Reisner 已提交
923 924 925 926 927
			if (ok)
				break;
		}

retry:
928
		s = drbd_wait_for_connect(connection, &ad);
P
Philipp Reisner 已提交
929
		if (s) {
930
			int fp = receive_first_packet(connection, s);
931 932
			drbd_socket_okay(&sock.socket);
			drbd_socket_okay(&msock.socket);
933
			switch (fp) {
934
			case P_INITIAL_DATA:
935
				if (sock.socket) {
936
					drbd_warn(connection, "initial packet S crossed\n");
937
					sock_release(sock.socket);
938 939
					sock.socket = s;
					goto randomize;
P
Philipp Reisner 已提交
940
				}
941
				sock.socket = s;
P
Philipp Reisner 已提交
942
				break;
943
			case P_INITIAL_META:
944
				set_bit(RESOLVE_CONFLICTS, &connection->flags);
945
				if (msock.socket) {
946
					drbd_warn(connection, "initial packet M crossed\n");
947
					sock_release(msock.socket);
948 949
					msock.socket = s;
					goto randomize;
P
Philipp Reisner 已提交
950
				}
951
				msock.socket = s;
P
Philipp Reisner 已提交
952 953
				break;
			default:
954
				drbd_warn(connection, "Error receiving initial packet\n");
P
Philipp Reisner 已提交
955
				sock_release(s);
956
randomize:
957
				if (prandom_u32() & 1)
P
Philipp Reisner 已提交
958 959 960 961
					goto retry;
			}
		}

962
		if (connection->cstate <= C_DISCONNECTING)
P
Philipp Reisner 已提交
963 964 965 966
			goto out_release_sockets;
		if (signal_pending(current)) {
			flush_signals(current);
			smp_rmb();
967
			if (get_t_state(&connection->receiver) == EXITING)
P
Philipp Reisner 已提交
968 969 970
				goto out_release_sockets;
		}

971 972 973
		ok = drbd_socket_okay(&sock.socket);
		ok = drbd_socket_okay(&msock.socket) && ok;
	} while (!ok);
P
Philipp Reisner 已提交
974

975 976
	if (ad.s_listen)
		sock_release(ad.s_listen);
P
Philipp Reisner 已提交
977

978 979
	sock.socket->sk->sk_reuse = SK_CAN_REUSE; /* SO_REUSEADDR */
	msock.socket->sk->sk_reuse = SK_CAN_REUSE; /* SO_REUSEADDR */
P
Philipp Reisner 已提交
980

981 982
	sock.socket->sk->sk_allocation = GFP_NOIO;
	msock.socket->sk->sk_allocation = GFP_NOIO;
P
Philipp Reisner 已提交
983

984 985
	sock.socket->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK;
	msock.socket->sk->sk_priority = TC_PRIO_INTERACTIVE;
P
Philipp Reisner 已提交
986 987

	/* NOT YET ...
988
	 * sock.socket->sk->sk_sndtimeo = connection->net_conf->timeout*HZ/10;
989
	 * sock.socket->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
990
	 * first set it to the P_CONNECTION_FEATURES timeout,
P
Philipp Reisner 已提交
991
	 * which we set to 4x the configured ping_timeout. */
992
	rcu_read_lock();
993
	nc = rcu_dereference(connection->net_conf);
994

995 996
	sock.socket->sk->sk_sndtimeo =
	sock.socket->sk->sk_rcvtimeo = nc->ping_timeo*4*HZ/10;
P
Philipp Reisner 已提交
997

998
	msock.socket->sk->sk_rcvtimeo = nc->ping_int*HZ;
999
	timeout = nc->timeout * HZ / 10;
1000
	discard_my_data = nc->discard_my_data;
1001
	rcu_read_unlock();
P
Philipp Reisner 已提交
1002

1003
	msock.socket->sk->sk_sndtimeo = timeout;
P
Philipp Reisner 已提交
1004 1005

	/* we don't want delays.
L
Lucas De Marchi 已提交
1006
	 * we use TCP_CORK where appropriate, though */
1007 1008
	drbd_tcp_nodelay(sock.socket);
	drbd_tcp_nodelay(msock.socket);
P
Philipp Reisner 已提交
1009

1010 1011 1012
	connection->data.socket = sock.socket;
	connection->meta.socket = msock.socket;
	connection->last_received = jiffies;
P
Philipp Reisner 已提交
1013

1014
	h = drbd_do_features(connection);
P
Philipp Reisner 已提交
1015 1016 1017
	if (h <= 0)
		return h;

1018
	if (connection->cram_hmac_tfm) {
1019
		/* drbd_request_state(device, NS(conn, WFAuth)); */
1020
		switch (drbd_do_auth(connection)) {
1021
		case -1:
1022
			drbd_err(connection, "Authentication of peer failed\n");
P
Philipp Reisner 已提交
1023
			return -1;
1024
		case 0:
1025
			drbd_err(connection, "Authentication of peer failed, trying again.\n");
1026
			return 0;
P
Philipp Reisner 已提交
1027 1028 1029
		}
	}

1030 1031
	connection->data.socket->sk->sk_sndtimeo = timeout;
	connection->data.socket->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
P
Philipp Reisner 已提交
1032

1033
	if (drbd_send_protocol(connection) == -EOPNOTSUPP)
1034
		return -1;
P
Philipp Reisner 已提交
1035

1036 1037 1038 1039 1040 1041 1042 1043 1044 1045
	/* Prevent a race between resync-handshake and
	 * being promoted to Primary.
	 *
	 * Grab and release the state mutex, so we know that any current
	 * drbd_set_role() is finished, and any incoming drbd_set_role
	 * will see the STATE_SENT flag, and wait for it to be cleared.
	 */
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr)
		mutex_lock(peer_device->device->state_mutex);

1046
	set_bit(STATE_SENT, &connection->flags);
1047

1048 1049 1050
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr)
		mutex_unlock(peer_device->device->state_mutex);

P
Philipp Reisner 已提交
1051
	rcu_read_lock();
1052 1053
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
		struct drbd_device *device = peer_device->device;
1054
		kref_get(&device->kref);
1055 1056
		rcu_read_unlock();

1057
		if (discard_my_data)
1058
			set_bit(DISCARD_MY_DATA, &device->flags);
1059
		else
1060
			clear_bit(DISCARD_MY_DATA, &device->flags);
1061

1062
		drbd_connected(peer_device);
1063
		kref_put(&device->kref, drbd_destroy_device);
P
Philipp Reisner 已提交
1064 1065 1066 1067
		rcu_read_lock();
	}
	rcu_read_unlock();

1068 1069 1070
	rv = conn_request_state(connection, NS(conn, C_WF_REPORT_PARAMS), CS_VERBOSE);
	if (rv < SS_SUCCESS || connection->cstate != C_WF_REPORT_PARAMS) {
		clear_bit(STATE_SENT, &connection->flags);
1071
		return 0;
1072
	}
1073

1074
	drbd_thread_start(&connection->asender);
P
Philipp Reisner 已提交
1075

1076
	mutex_lock(&connection->resource->conf_update);
1077 1078 1079 1080
	/* The discard_my_data flag is a single-shot modifier to the next
	 * connection attempt, the handshake of which is now well underway.
	 * No need for rcu style copying of the whole struct
	 * just to clear a single value. */
1081
	connection->net_conf->discard_my_data = 0;
1082
	mutex_unlock(&connection->resource->conf_update);
1083

1084
	return h;
P
Philipp Reisner 已提交
1085 1086

out_release_sockets:
1087 1088
	if (ad.s_listen)
		sock_release(ad.s_listen);
1089 1090 1091 1092
	if (sock.socket)
		sock_release(sock.socket);
	if (msock.socket)
		sock_release(msock.socket);
P
Philipp Reisner 已提交
1093 1094 1095
	return -1;
}

1096
static int decode_header(struct drbd_connection *connection, void *header, struct packet_info *pi)
P
Philipp Reisner 已提交
1097
{
1098
	unsigned int header_size = drbd_header_size(connection);
1099

1100 1101 1102 1103
	if (header_size == sizeof(struct p_header100) &&
	    *(__be32 *)header == cpu_to_be32(DRBD_MAGIC_100)) {
		struct p_header100 *h = header;
		if (h->pad != 0) {
1104
			drbd_err(connection, "Header padding is not zero\n");
1105 1106 1107 1108 1109 1110 1111
			return -EINVAL;
		}
		pi->vnr = be16_to_cpu(h->volume);
		pi->cmd = be16_to_cpu(h->command);
		pi->size = be32_to_cpu(h->length);
	} else if (header_size == sizeof(struct p_header95) &&
		   *(__be16 *)header == cpu_to_be16(DRBD_MAGIC_BIG)) {
1112 1113
		struct p_header95 *h = header;
		pi->cmd = be16_to_cpu(h->command);
1114 1115
		pi->size = be32_to_cpu(h->length);
		pi->vnr = 0;
1116 1117 1118 1119 1120
	} else if (header_size == sizeof(struct p_header80) &&
		   *(__be32 *)header == cpu_to_be32(DRBD_MAGIC)) {
		struct p_header80 *h = header;
		pi->cmd = be16_to_cpu(h->command);
		pi->size = be16_to_cpu(h->length);
1121
		pi->vnr = 0;
1122
	} else {
1123
		drbd_err(connection, "Wrong magic value 0x%08x in protocol version %d\n",
1124
			 be32_to_cpu(*(__be32 *)header),
1125
			 connection->agreed_pro_version);
1126
		return -EINVAL;
P
Philipp Reisner 已提交
1127
	}
1128
	pi->data = header + header_size;
1129
	return 0;
1130
}
P
Philipp Reisner 已提交
1131

1132
static int drbd_recv_header(struct drbd_connection *connection, struct packet_info *pi)
1133
{
1134
	void *buffer = connection->data.rbuf;
1135
	int err;
1136

1137
	err = drbd_recv_all_warn(connection, buffer, drbd_header_size(connection));
1138
	if (err)
1139
		return err;
1140

1141 1142
	err = decode_header(connection, buffer, pi);
	connection->last_received = jiffies;
P
Philipp Reisner 已提交
1143

1144
	return err;
P
Philipp Reisner 已提交
1145 1146
}

1147
static void drbd_flush(struct drbd_connection *connection)
P
Philipp Reisner 已提交
1148 1149
{
	int rv;
1150
	struct drbd_peer_device *peer_device;
1151 1152
	int vnr;

1153
	if (connection->write_ordering >= WO_bdev_flush) {
1154
		rcu_read_lock();
1155 1156 1157
		idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
			struct drbd_device *device = peer_device->device;

1158
			if (!get_ldev(device))
1159
				continue;
1160
			kref_get(&device->kref);
1161 1162
			rcu_read_unlock();

1163
			rv = blkdev_issue_flush(device->ldev->backing_bdev,
1164 1165
					GFP_NOIO, NULL);
			if (rv) {
1166
				drbd_info(device, "local disk flush failed with status %d\n", rv);
1167 1168 1169
				/* would rather check on EOPNOTSUPP, but that is not reliable.
				 * don't try again for ANY return value != 0
				 * if (rv == -EOPNOTSUPP) */
1170
				drbd_bump_write_ordering(connection, WO_drain_io);
1171
			}
1172
			put_ldev(device);
1173
			kref_put(&device->kref, drbd_destroy_device);
P
Philipp Reisner 已提交
1174

1175 1176 1177
			rcu_read_lock();
			if (rv)
				break;
P
Philipp Reisner 已提交
1178
		}
1179
		rcu_read_unlock();
P
Philipp Reisner 已提交
1180 1181 1182 1183 1184
	}
}

/**
 * drbd_may_finish_epoch() - Applies an epoch_event to the epoch's state, eventually finishes it.
1185
 * @device:	DRBD device.
P
Philipp Reisner 已提交
1186 1187 1188
 * @epoch:	Epoch object.
 * @ev:		Epoch event.
 */
1189
static enum finish_epoch drbd_may_finish_epoch(struct drbd_connection *connection,
P
Philipp Reisner 已提交
1190 1191 1192
					       struct drbd_epoch *epoch,
					       enum epoch_event ev)
{
1193
	int epoch_size;
P
Philipp Reisner 已提交
1194 1195 1196
	struct drbd_epoch *next_epoch;
	enum finish_epoch rv = FE_STILL_LIVE;

1197
	spin_lock(&connection->epoch_lock);
P
Philipp Reisner 已提交
1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
	do {
		next_epoch = NULL;

		epoch_size = atomic_read(&epoch->epoch_size);

		switch (ev & ~EV_CLEANUP) {
		case EV_PUT:
			atomic_dec(&epoch->active);
			break;
		case EV_GOT_BARRIER_NR:
			set_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags);
			break;
		case EV_BECAME_LAST:
			/* nothing to do*/
			break;
		}

		if (epoch_size != 0 &&
		    atomic_read(&epoch->active) == 0 &&
1217
		    (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags) || ev & EV_CLEANUP)) {
P
Philipp Reisner 已提交
1218
			if (!(ev & EV_CLEANUP)) {
1219 1220 1221
				spin_unlock(&connection->epoch_lock);
				drbd_send_b_ack(epoch->connection, epoch->barrier_nr, epoch_size);
				spin_lock(&connection->epoch_lock);
P
Philipp Reisner 已提交
1222
			}
1223 1224 1225
#if 0
			/* FIXME: dec unacked on connection, once we have
			 * something to count pending connection packets in. */
1226
			if (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags))
1227
				dec_unacked(epoch->connection);
1228
#endif
P
Philipp Reisner 已提交
1229

1230
			if (connection->current_epoch != epoch) {
P
Philipp Reisner 已提交
1231 1232 1233
				next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list);
				list_del(&epoch->list);
				ev = EV_BECAME_LAST | (ev & EV_CLEANUP);
1234
				connection->epochs--;
P
Philipp Reisner 已提交
1235 1236 1237 1238 1239 1240 1241
				kfree(epoch);

				if (rv == FE_STILL_LIVE)
					rv = FE_DESTROYED;
			} else {
				epoch->flags = 0;
				atomic_set(&epoch->epoch_size, 0);
1242
				/* atomic_set(&epoch->active, 0); is already zero */
P
Philipp Reisner 已提交
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253
				if (rv == FE_STILL_LIVE)
					rv = FE_RECYCLED;
			}
		}

		if (!next_epoch)
			break;

		epoch = next_epoch;
	} while (1);

1254
	spin_unlock(&connection->epoch_lock);
P
Philipp Reisner 已提交
1255 1256 1257 1258 1259 1260

	return rv;
}

/**
 * drbd_bump_write_ordering() - Fall back to an other write ordering method
1261
 * @connection:	DRBD connection.
P
Philipp Reisner 已提交
1262 1263
 * @wo:		Write ordering method to try.
 */
1264
void drbd_bump_write_ordering(struct drbd_connection *connection, enum write_ordering_e wo)
P
Philipp Reisner 已提交
1265
{
P
Philipp Reisner 已提交
1266
	struct disk_conf *dc;
1267
	struct drbd_peer_device *peer_device;
P
Philipp Reisner 已提交
1268
	enum write_ordering_e pwo;
1269
	int vnr;
P
Philipp Reisner 已提交
1270 1271 1272 1273 1274 1275
	static char *write_ordering_str[] = {
		[WO_none] = "none",
		[WO_drain_io] = "drain",
		[WO_bdev_flush] = "flush",
	};

1276
	pwo = connection->write_ordering;
P
Philipp Reisner 已提交
1277
	wo = min(pwo, wo);
P
Philipp Reisner 已提交
1278
	rcu_read_lock();
1279 1280 1281
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
		struct drbd_device *device = peer_device->device;

1282
		if (!get_ldev_if_state(device, D_ATTACHING))
1283
			continue;
1284
		dc = rcu_dereference(device->ldev->disk_conf);
1285 1286 1287 1288 1289

		if (wo == WO_bdev_flush && !dc->disk_flushes)
			wo = WO_drain_io;
		if (wo == WO_drain_io && !dc->disk_drain)
			wo = WO_none;
1290
		put_ldev(device);
1291
	}
P
Philipp Reisner 已提交
1292
	rcu_read_unlock();
1293 1294
	connection->write_ordering = wo;
	if (pwo != connection->write_ordering || wo == WO_bdev_flush)
1295
		drbd_info(connection, "Method to ensure write ordering: %s\n", write_ordering_str[connection->write_ordering]);
P
Philipp Reisner 已提交
1296 1297
}

1298
/**
1299
 * drbd_submit_peer_request()
1300
 * @device:	DRBD device.
1301
 * @peer_req:	peer request
1302
 * @rw:		flag field, see bio->bi_rw
1303 1304 1305 1306 1307 1308 1309 1310 1311 1312
 *
 * May spread the pages to multiple bios,
 * depending on bio_add_page restrictions.
 *
 * Returns 0 if all bios have been submitted,
 * -ENOMEM if we could not allocate enough bios,
 * -ENOSPC (any better suggestion?) if we have not been able to bio_add_page a
 *  single page to an empty bio (which should never happen and likely indicates
 *  that the lower level IO stack is in some way broken). This has been observed
 *  on certain Xen deployments.
1313 1314
 */
/* TODO allocate from our own bio_set. */
1315
int drbd_submit_peer_request(struct drbd_device *device,
1316 1317
			     struct drbd_peer_request *peer_req,
			     const unsigned rw, const int fault_type)
1318 1319 1320
{
	struct bio *bios = NULL;
	struct bio *bio;
1321 1322 1323
	struct page *page = peer_req->pages;
	sector_t sector = peer_req->i.sector;
	unsigned ds = peer_req->i.size;
1324 1325
	unsigned n_bios = 0;
	unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT;
1326
	int err = -ENOMEM;
1327 1328 1329 1330

	/* In most cases, we will only need one bio.  But in case the lower
	 * level restrictions happen to be different at this offset on this
	 * side than those of the sending peer, we may need to submit the
1331 1332 1333 1334 1335
	 * request in more than one bio.
	 *
	 * Plain bio_alloc is good enough here, this is no DRBD internally
	 * generated bio, but a bio allocated on behalf of the peer.
	 */
1336 1337 1338
next_bio:
	bio = bio_alloc(GFP_NOIO, nr_pages);
	if (!bio) {
1339
		drbd_err(device, "submit_ee: Allocation of a bio failed\n");
1340 1341
		goto fail;
	}
1342
	/* > peer_req->i.sector, unless this is the first bio */
1343
	bio->bi_iter.bi_sector = sector;
1344
	bio->bi_bdev = device->ldev->backing_bdev;
1345
	bio->bi_rw = rw;
1346
	bio->bi_private = peer_req;
1347
	bio->bi_end_io = drbd_peer_request_endio;
1348 1349 1350 1351 1352 1353 1354 1355

	bio->bi_next = bios;
	bios = bio;
	++n_bios;

	page_chain_for_each(page) {
		unsigned len = min_t(unsigned, ds, PAGE_SIZE);
		if (!bio_add_page(bio, page, len, 0)) {
1356 1357 1358 1359
			/* A single page must always be possible!
			 * But in case it fails anyways,
			 * we deal with it, and complain (below). */
			if (bio->bi_vcnt == 0) {
1360
				drbd_err(device,
1361 1362
					"bio_add_page failed for len=%u, "
					"bi_vcnt=0 (bi_sector=%llu)\n",
1363
					len, (uint64_t)bio->bi_iter.bi_sector);
1364 1365 1366
				err = -ENOSPC;
				goto fail;
			}
1367 1368 1369 1370 1371 1372
			goto next_bio;
		}
		ds -= len;
		sector += len >> 9;
		--nr_pages;
	}
1373 1374
	D_ASSERT(device, page == NULL);
	D_ASSERT(device, ds == 0);
1375

1376
	atomic_set(&peer_req->pending_bios, n_bios);
1377 1378 1379 1380 1381
	do {
		bio = bios;
		bios = bios->bi_next;
		bio->bi_next = NULL;

1382
		drbd_generic_make_request(device, fault_type, bio);
1383 1384 1385 1386 1387 1388 1389 1390 1391
	} while (bios);
	return 0;

fail:
	while (bios) {
		bio = bios;
		bios = bios->bi_next;
		bio_put(bio);
	}
1392
	return err;
1393 1394
}

1395
static void drbd_remove_epoch_entry_interval(struct drbd_device *device,
1396
					     struct drbd_peer_request *peer_req)
1397
{
1398
	struct drbd_interval *i = &peer_req->i;
1399

1400
	drbd_remove_interval(&device->write_requests, i);
1401 1402
	drbd_clear_interval(i);

A
Andreas Gruenbacher 已提交
1403
	/* Wake up any processes waiting for this peer request to complete.  */
1404
	if (i->waiting)
1405
		wake_up(&device->misc_wait);
1406 1407
}

1408
static void conn_wait_active_ee_empty(struct drbd_connection *connection)
1409
{
1410
	struct drbd_peer_device *peer_device;
1411 1412 1413
	int vnr;

	rcu_read_lock();
1414 1415 1416
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
		struct drbd_device *device = peer_device->device;

1417
		kref_get(&device->kref);
1418
		rcu_read_unlock();
1419
		drbd_wait_ee_list_empty(device, &device->active_ee);
1420
		kref_put(&device->kref, drbd_destroy_device);
1421 1422 1423 1424 1425
		rcu_read_lock();
	}
	rcu_read_unlock();
}

1426 1427 1428 1429 1430 1431
static struct drbd_peer_device *
conn_peer_device(struct drbd_connection *connection, int volume_number)
{
	return idr_find(&connection->peer_devices, volume_number);
}

1432
static int receive_Barrier(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
1433
{
1434
	int rv;
1435
	struct p_barrier *p = pi->data;
P
Philipp Reisner 已提交
1436 1437
	struct drbd_epoch *epoch;

1438 1439 1440
	/* FIXME these are unacked on connection,
	 * not a specific (peer)device.
	 */
1441 1442 1443
	connection->current_epoch->barrier_nr = p->barrier;
	connection->current_epoch->connection = connection;
	rv = drbd_may_finish_epoch(connection, connection->current_epoch, EV_GOT_BARRIER_NR);
P
Philipp Reisner 已提交
1444 1445 1446 1447 1448 1449

	/* P_BARRIER_ACK may imply that the corresponding extent is dropped from
	 * the activity log, which means it would not be resynced in case the
	 * R_PRIMARY crashes now.
	 * Therefore we must send the barrier_ack after the barrier request was
	 * completed. */
1450
	switch (connection->write_ordering) {
P
Philipp Reisner 已提交
1451 1452
	case WO_none:
		if (rv == FE_RECYCLED)
1453
			return 0;
1454 1455 1456 1457 1458 1459 1460

		/* receiver context, in the writeout path of the other node.
		 * avoid potential distributed deadlock */
		epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO);
		if (epoch)
			break;
		else
1461
			drbd_warn(connection, "Allocation of an epoch failed, slowing down\n");
1462
			/* Fall through */
P
Philipp Reisner 已提交
1463 1464 1465

	case WO_bdev_flush:
	case WO_drain_io:
1466 1467
		conn_wait_active_ee_empty(connection);
		drbd_flush(connection);
1468

1469
		if (atomic_read(&connection->current_epoch->epoch_size)) {
1470 1471 1472
			epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO);
			if (epoch)
				break;
P
Philipp Reisner 已提交
1473 1474
		}

1475
		return 0;
1476
	default:
1477
		drbd_err(connection, "Strangeness in connection->write_ordering %d\n", connection->write_ordering);
1478
		return -EIO;
P
Philipp Reisner 已提交
1479 1480 1481 1482 1483 1484
	}

	epoch->flags = 0;
	atomic_set(&epoch->epoch_size, 0);
	atomic_set(&epoch->active, 0);

1485 1486 1487 1488 1489
	spin_lock(&connection->epoch_lock);
	if (atomic_read(&connection->current_epoch->epoch_size)) {
		list_add(&epoch->list, &connection->current_epoch->list);
		connection->current_epoch = epoch;
		connection->epochs++;
P
Philipp Reisner 已提交
1490 1491 1492 1493
	} else {
		/* The current_epoch got recycled while we allocated this one... */
		kfree(epoch);
	}
1494
	spin_unlock(&connection->epoch_lock);
P
Philipp Reisner 已提交
1495

1496
	return 0;
P
Philipp Reisner 已提交
1497 1498 1499 1500
}

/* used from receive_RSDataReply (recv_resync_read)
 * and from receive_Data */
1501
static struct drbd_peer_request *
1502
read_in_block(struct drbd_peer_device *peer_device, u64 id, sector_t sector,
1503
	      int data_size) __must_hold(local)
P
Philipp Reisner 已提交
1504
{
1505
	struct drbd_device *device = peer_device->device;
1506
	const sector_t capacity = drbd_get_capacity(device->this_bdev);
1507
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
1508
	struct page *page;
1509
	int dgs, ds, err;
1510 1511
	void *dig_in = peer_device->connection->int_dig_in;
	void *dig_vv = peer_device->connection->int_dig_vv;
1512
	unsigned long *data;
P
Philipp Reisner 已提交
1513

1514
	dgs = 0;
1515 1516
	if (peer_device->connection->peer_integrity_tfm) {
		dgs = crypto_hash_digestsize(peer_device->connection->peer_integrity_tfm);
1517 1518 1519 1520
		/*
		 * FIXME: Receive the incoming digest into the receive buffer
		 *	  here, together with its struct p_data?
		 */
1521
		err = drbd_recv_all_warn(peer_device->connection, dig_in, dgs);
1522
		if (err)
P
Philipp Reisner 已提交
1523
			return NULL;
1524
		data_size -= dgs;
P
Philipp Reisner 已提交
1525 1526
	}

1527 1528 1529 1530
	if (!expect(IS_ALIGNED(data_size, 512)))
		return NULL;
	if (!expect(data_size <= DRBD_MAX_BIO_SIZE))
		return NULL;
P
Philipp Reisner 已提交
1531

1532 1533 1534
	/* even though we trust out peer,
	 * we sometimes have to double check. */
	if (sector + (data_size>>9) > capacity) {
1535
		drbd_err(device, "request from peer beyond end of local disk: "
1536
			"capacity: %llus < sector: %llus + size: %u\n",
1537 1538 1539 1540 1541
			(unsigned long long)capacity,
			(unsigned long long)sector, data_size);
		return NULL;
	}

P
Philipp Reisner 已提交
1542 1543 1544
	/* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD
	 * "criss-cross" setup, that might cause write-out on some other DRBD,
	 * which in turn might block on the other node at this very place.  */
1545
	peer_req = drbd_alloc_peer_req(peer_device, id, sector, data_size, GFP_NOIO);
1546
	if (!peer_req)
P
Philipp Reisner 已提交
1547
		return NULL;
1548

1549
	if (!data_size)
1550
		return peer_req;
1551

P
Philipp Reisner 已提交
1552
	ds = data_size;
1553
	page = peer_req->pages;
1554 1555
	page_chain_for_each(page) {
		unsigned len = min_t(int, ds, PAGE_SIZE);
1556
		data = kmap(page);
1557
		err = drbd_recv_all_warn(peer_device->connection, data, len);
1558
		if (drbd_insert_fault(device, DRBD_FAULT_RECEIVE)) {
1559
			drbd_err(device, "Fault injection: Corrupting data on receive\n");
1560 1561
			data[0] = data[0] ^ (unsigned long)-1;
		}
P
Philipp Reisner 已提交
1562
		kunmap(page);
1563
		if (err) {
1564
			drbd_free_peer_req(device, peer_req);
P
Philipp Reisner 已提交
1565 1566
			return NULL;
		}
1567
		ds -= len;
P
Philipp Reisner 已提交
1568 1569 1570
	}

	if (dgs) {
1571
		drbd_csum_ee(peer_device->connection->peer_integrity_tfm, peer_req, dig_vv);
P
Philipp Reisner 已提交
1572
		if (memcmp(dig_in, dig_vv, dgs)) {
1573
			drbd_err(device, "Digest integrity check FAILED: %llus +%u\n",
1574
				(unsigned long long)sector, data_size);
1575
			drbd_free_peer_req(device, peer_req);
P
Philipp Reisner 已提交
1576 1577 1578
			return NULL;
		}
	}
1579
	device->recv_cnt += data_size>>9;
1580
	return peer_req;
P
Philipp Reisner 已提交
1581 1582 1583 1584 1585
}

/* drbd_drain_block() just takes a data block
 * out of the socket input buffer, and discards it.
 */
1586
static int drbd_drain_block(struct drbd_peer_device *peer_device, int data_size)
P
Philipp Reisner 已提交
1587 1588
{
	struct page *page;
1589
	int err = 0;
P
Philipp Reisner 已提交
1590 1591
	void *data;

1592
	if (!data_size)
1593
		return 0;
1594

1595
	page = drbd_alloc_pages(peer_device, 1, 1);
P
Philipp Reisner 已提交
1596 1597 1598

	data = kmap(page);
	while (data_size) {
1599 1600
		unsigned int len = min_t(int, data_size, PAGE_SIZE);

1601
		err = drbd_recv_all_warn(peer_device->connection, data, len);
1602
		if (err)
P
Philipp Reisner 已提交
1603
			break;
1604
		data_size -= len;
P
Philipp Reisner 已提交
1605 1606
	}
	kunmap(page);
1607
	drbd_free_pages(peer_device->device, page, 0);
1608
	return err;
P
Philipp Reisner 已提交
1609 1610
}

1611
static int recv_dless_read(struct drbd_peer_device *peer_device, struct drbd_request *req,
P
Philipp Reisner 已提交
1612 1613
			   sector_t sector, int data_size)
{
1614 1615
	struct bio_vec bvec;
	struct bvec_iter iter;
P
Philipp Reisner 已提交
1616
	struct bio *bio;
1617
	int dgs, err, expect;
1618 1619
	void *dig_in = peer_device->connection->int_dig_in;
	void *dig_vv = peer_device->connection->int_dig_vv;
P
Philipp Reisner 已提交
1620

1621
	dgs = 0;
1622 1623 1624
	if (peer_device->connection->peer_integrity_tfm) {
		dgs = crypto_hash_digestsize(peer_device->connection->peer_integrity_tfm);
		err = drbd_recv_all_warn(peer_device->connection, dig_in, dgs);
1625 1626
		if (err)
			return err;
1627
		data_size -= dgs;
P
Philipp Reisner 已提交
1628 1629 1630 1631
	}

	/* optimistically update recv_cnt.  if receiving fails below,
	 * we disconnect anyways, and counters will be reset. */
1632
	peer_device->device->recv_cnt += data_size>>9;
P
Philipp Reisner 已提交
1633 1634

	bio = req->master_bio;
1635
	D_ASSERT(peer_device->device, sector == bio->bi_iter.bi_sector);
P
Philipp Reisner 已提交
1636

1637 1638 1639
	bio_for_each_segment(bvec, bio, iter) {
		void *mapped = kmap(bvec.bv_page) + bvec.bv_offset;
		expect = min_t(int, data_size, bvec.bv_len);
1640
		err = drbd_recv_all_warn(peer_device->connection, mapped, expect);
1641
		kunmap(bvec.bv_page);
1642 1643 1644
		if (err)
			return err;
		data_size -= expect;
P
Philipp Reisner 已提交
1645 1646 1647
	}

	if (dgs) {
1648
		drbd_csum_bio(peer_device->connection->peer_integrity_tfm, bio, dig_vv);
P
Philipp Reisner 已提交
1649
		if (memcmp(dig_in, dig_vv, dgs)) {
1650
			drbd_err(peer_device, "Digest integrity check FAILED. Broken NICs?\n");
1651
			return -EINVAL;
P
Philipp Reisner 已提交
1652 1653 1654
		}
	}

1655
	D_ASSERT(peer_device->device, data_size == 0);
1656
	return 0;
P
Philipp Reisner 已提交
1657 1658
}

1659 1660 1661 1662
/*
 * e_end_resync_block() is called in asender context via
 * drbd_finish_peer_reqs().
 */
1663
static int e_end_resync_block(struct drbd_work *w, int unused)
P
Philipp Reisner 已提交
1664
{
1665
	struct drbd_peer_request *peer_req =
1666 1667 1668
		container_of(w, struct drbd_peer_request, w);
	struct drbd_peer_device *peer_device = peer_req->peer_device;
	struct drbd_device *device = peer_device->device;
1669
	sector_t sector = peer_req->i.sector;
1670
	int err;
P
Philipp Reisner 已提交
1671

1672
	D_ASSERT(device, drbd_interval_empty(&peer_req->i));
P
Philipp Reisner 已提交
1673

1674
	if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
1675
		drbd_set_in_sync(device, sector, peer_req->i.size);
1676
		err = drbd_send_ack(peer_device, P_RS_WRITE_ACK, peer_req);
P
Philipp Reisner 已提交
1677 1678
	} else {
		/* Record failure to sync */
1679
		drbd_rs_failed_io(device, sector, peer_req->i.size);
P
Philipp Reisner 已提交
1680

1681
		err  = drbd_send_ack(peer_device, P_NEG_ACK, peer_req);
P
Philipp Reisner 已提交
1682
	}
1683
	dec_unacked(device);
P
Philipp Reisner 已提交
1684

1685
	return err;
P
Philipp Reisner 已提交
1686 1687
}

1688 1689
static int recv_resync_read(struct drbd_peer_device *peer_device, sector_t sector,
			    int data_size) __releases(local)
P
Philipp Reisner 已提交
1690
{
1691
	struct drbd_device *device = peer_device->device;
1692
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
1693

1694
	peer_req = read_in_block(peer_device, ID_SYNCER, sector, data_size);
1695
	if (!peer_req)
1696
		goto fail;
P
Philipp Reisner 已提交
1697

1698
	dec_rs_pending(device);
P
Philipp Reisner 已提交
1699

1700
	inc_unacked(device);
P
Philipp Reisner 已提交
1701 1702 1703
	/* corresponding dec_unacked() in e_end_resync_block()
	 * respective _drbd_clear_done_ee */

1704
	peer_req->w.cb = e_end_resync_block;
1705

1706
	spin_lock_irq(&device->resource->req_lock);
1707
	list_add(&peer_req->w.list, &device->sync_ee);
1708
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
1709

1710 1711
	atomic_add(data_size >> 9, &device->rs_sect_ev);
	if (drbd_submit_peer_request(device, peer_req, WRITE, DRBD_FAULT_RS_WR) == 0)
1712
		return 0;
P
Philipp Reisner 已提交
1713

1714
	/* don't care for the reason here */
1715
	drbd_err(device, "submit failed, triggering re-connect\n");
1716
	spin_lock_irq(&device->resource->req_lock);
1717
	list_del(&peer_req->w.list);
1718
	spin_unlock_irq(&device->resource->req_lock);
1719

1720
	drbd_free_peer_req(device, peer_req);
1721
fail:
1722
	put_ldev(device);
1723
	return -EIO;
P
Philipp Reisner 已提交
1724 1725
}

1726
static struct drbd_request *
1727
find_request(struct drbd_device *device, struct rb_root *root, u64 id,
1728
	     sector_t sector, bool missing_ok, const char *func)
1729 1730 1731
{
	struct drbd_request *req;

1732 1733
	/* Request object according to our peer */
	req = (struct drbd_request *)(unsigned long)id;
1734
	if (drbd_contains_interval(root, sector, &req->i) && req->i.local)
1735
		return req;
1736
	if (!missing_ok) {
1737
		drbd_err(device, "%s: failed to find request 0x%lx, sector %llus\n", func,
1738 1739
			(unsigned long)id, (unsigned long long)sector);
	}
1740
	return NULL;
P
Philipp Reisner 已提交
1741 1742
}

1743
static int receive_DataReply(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
1744
{
1745
	struct drbd_peer_device *peer_device;
1746
	struct drbd_device *device;
P
Philipp Reisner 已提交
1747 1748
	struct drbd_request *req;
	sector_t sector;
1749
	int err;
1750
	struct p_data *p = pi->data;
1751

1752 1753
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
1754
		return -EIO;
1755
	device = peer_device->device;
P
Philipp Reisner 已提交
1756 1757 1758

	sector = be64_to_cpu(p->sector);

1759
	spin_lock_irq(&device->resource->req_lock);
1760
	req = find_request(device, &device->read_requests, p->block_id, sector, false, __func__);
1761
	spin_unlock_irq(&device->resource->req_lock);
1762
	if (unlikely(!req))
1763
		return -EIO;
P
Philipp Reisner 已提交
1764

B
Bart Van Assche 已提交
1765
	/* hlist_del(&req->collision) is done in _req_may_be_done, to avoid
P
Philipp Reisner 已提交
1766 1767
	 * special casing it there for the various failure cases.
	 * still no race with drbd_fail_pending_reads */
1768
	err = recv_dless_read(peer_device, req, sector, pi->size);
1769
	if (!err)
1770
		req_mod(req, DATA_RECEIVED);
P
Philipp Reisner 已提交
1771 1772 1773 1774
	/* else: nothing. handled from drbd_disconnect...
	 * I don't think we may complete this just yet
	 * in case we are "on-disconnect: freeze" */

1775
	return err;
P
Philipp Reisner 已提交
1776 1777
}

1778
static int receive_RSDataReply(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
1779
{
1780
	struct drbd_peer_device *peer_device;
1781
	struct drbd_device *device;
P
Philipp Reisner 已提交
1782
	sector_t sector;
1783
	int err;
1784
	struct p_data *p = pi->data;
1785

1786 1787
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
1788
		return -EIO;
1789
	device = peer_device->device;
P
Philipp Reisner 已提交
1790 1791

	sector = be64_to_cpu(p->sector);
1792
	D_ASSERT(device, p->block_id == ID_SYNCER);
P
Philipp Reisner 已提交
1793

1794
	if (get_ldev(device)) {
P
Philipp Reisner 已提交
1795 1796
		/* data is submitted to disk within recv_resync_read.
		 * corresponding put_ldev done below on error,
1797
		 * or in drbd_peer_request_endio. */
1798
		err = recv_resync_read(peer_device, sector, pi->size);
P
Philipp Reisner 已提交
1799 1800
	} else {
		if (__ratelimit(&drbd_ratelimit_state))
1801
			drbd_err(device, "Can not write resync data to local disk.\n");
P
Philipp Reisner 已提交
1802

1803
		err = drbd_drain_block(peer_device, pi->size);
P
Philipp Reisner 已提交
1804

1805
		drbd_send_ack_dp(peer_device, P_NEG_ACK, p, pi->size);
P
Philipp Reisner 已提交
1806 1807
	}

1808
	atomic_add(pi->size >> 9, &device->rs_sect_in);
1809

1810
	return err;
P
Philipp Reisner 已提交
1811 1812
}

1813
static void restart_conflicting_writes(struct drbd_device *device,
1814
				       sector_t sector, int size)
P
Philipp Reisner 已提交
1815
{
1816 1817 1818
	struct drbd_interval *i;
	struct drbd_request *req;

1819
	drbd_for_each_overlap(i, &device->write_requests, sector, size) {
1820 1821 1822 1823 1824 1825
		if (!i->local)
			continue;
		req = container_of(i, struct drbd_request, i);
		if (req->rq_state & RQ_LOCAL_PENDING ||
		    !(req->rq_state & RQ_POSTPONED))
			continue;
1826 1827
		/* as it is RQ_POSTPONED, this will cause it to
		 * be queued on the retry workqueue. */
1828
		__req_mod(req, CONFLICT_RESOLVED, NULL);
1829 1830
	}
}
P
Philipp Reisner 已提交
1831

1832 1833
/*
 * e_end_block() is called in asender context via drbd_finish_peer_reqs().
P
Philipp Reisner 已提交
1834
 */
1835
static int e_end_block(struct drbd_work *w, int cancel)
P
Philipp Reisner 已提交
1836
{
1837
	struct drbd_peer_request *peer_req =
1838 1839 1840
		container_of(w, struct drbd_peer_request, w);
	struct drbd_peer_device *peer_device = peer_req->peer_device;
	struct drbd_device *device = peer_device->device;
1841
	sector_t sector = peer_req->i.sector;
1842
	int err = 0, pcmd;
P
Philipp Reisner 已提交
1843

1844
	if (peer_req->flags & EE_SEND_WRITE_ACK) {
1845
		if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
1846 1847
			pcmd = (device->state.conn >= C_SYNC_SOURCE &&
				device->state.conn <= C_PAUSED_SYNC_T &&
1848
				peer_req->flags & EE_MAY_SET_IN_SYNC) ?
P
Philipp Reisner 已提交
1849
				P_RS_WRITE_ACK : P_WRITE_ACK;
1850
			err = drbd_send_ack(peer_device, pcmd, peer_req);
P
Philipp Reisner 已提交
1851
			if (pcmd == P_RS_WRITE_ACK)
1852
				drbd_set_in_sync(device, sector, peer_req->i.size);
P
Philipp Reisner 已提交
1853
		} else {
1854
			err = drbd_send_ack(peer_device, P_NEG_ACK, peer_req);
P
Philipp Reisner 已提交
1855 1856 1857
			/* we expect it to be marked out of sync anyways...
			 * maybe assert this?  */
		}
1858
		dec_unacked(device);
P
Philipp Reisner 已提交
1859 1860 1861
	}
	/* we delete from the conflict detection hash _after_ we sent out the
	 * P_WRITE_ACK / P_NEG_ACK, to get the sequence number right.  */
1862
	if (peer_req->flags & EE_IN_INTERVAL_TREE) {
1863
		spin_lock_irq(&device->resource->req_lock);
1864
		D_ASSERT(device, !drbd_interval_empty(&peer_req->i));
1865
		drbd_remove_epoch_entry_interval(device, peer_req);
1866
		if (peer_req->flags & EE_RESTART_REQUESTS)
1867
			restart_conflicting_writes(device, sector, peer_req->i.size);
1868
		spin_unlock_irq(&device->resource->req_lock);
1869
	} else
1870
		D_ASSERT(device, drbd_interval_empty(&peer_req->i));
P
Philipp Reisner 已提交
1871

1872
	drbd_may_finish_epoch(first_peer_device(device)->connection, peer_req->epoch, EV_PUT + (cancel ? EV_CLEANUP : 0));
P
Philipp Reisner 已提交
1873

1874
	return err;
P
Philipp Reisner 已提交
1875 1876
}

1877
static int e_send_ack(struct drbd_work *w, enum drbd_packet ack)
P
Philipp Reisner 已提交
1878
{
1879
	struct drbd_peer_request *peer_req =
1880 1881
		container_of(w, struct drbd_peer_request, w);
	struct drbd_peer_device *peer_device = peer_req->peer_device;
1882
	int err;
P
Philipp Reisner 已提交
1883

1884 1885
	err = drbd_send_ack(peer_device, ack, peer_req);
	dec_unacked(peer_device->device);
P
Philipp Reisner 已提交
1886

1887
	return err;
P
Philipp Reisner 已提交
1888 1889
}

1890
static int e_send_superseded(struct drbd_work *w, int unused)
1891
{
1892
	return e_send_ack(w, P_SUPERSEDED);
1893 1894
}

1895
static int e_send_retry_write(struct drbd_work *w, int unused)
1896
{
1897 1898 1899
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
	struct drbd_connection *connection = peer_req->peer_device->connection;
1900

1901
	return e_send_ack(w, connection->agreed_pro_version >= 100 ?
1902
			     P_RETRY_WRITE : P_SUPERSEDED);
1903
}
P
Philipp Reisner 已提交
1904

1905 1906 1907 1908 1909 1910 1911 1912 1913
static bool seq_greater(u32 a, u32 b)
{
	/*
	 * We assume 32-bit wrap-around here.
	 * For 24-bit wrap-around, we would have to shift:
	 *  a <<= 8; b <<= 8;
	 */
	return (s32)a - (s32)b > 0;
}
P
Philipp Reisner 已提交
1914

1915 1916 1917
static u32 seq_max(u32 a, u32 b)
{
	return seq_greater(a, b) ? a : b;
P
Philipp Reisner 已提交
1918 1919
}

1920
static void update_peer_seq(struct drbd_peer_device *peer_device, unsigned int peer_seq)
1921
{
1922
	struct drbd_device *device = peer_device->device;
1923
	unsigned int newest_peer_seq;
1924

1925
	if (test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags)) {
1926 1927 1928 1929 1930
		spin_lock(&device->peer_seq_lock);
		newest_peer_seq = seq_max(device->peer_seq, peer_seq);
		device->peer_seq = newest_peer_seq;
		spin_unlock(&device->peer_seq_lock);
		/* wake up only if we actually changed device->peer_seq */
1931
		if (peer_seq == newest_peer_seq)
1932
			wake_up(&device->seq_wait);
1933
	}
P
Philipp Reisner 已提交
1934 1935
}

1936
static inline int overlaps(sector_t s1, int l1, sector_t s2, int l2)
1937
{
1938 1939
	return !((s1 + (l1>>9) <= s2) || (s1 >= s2 + (l2>>9)));
}
1940

1941
/* maybe change sync_ee into interval trees as well? */
1942
static bool overlapping_resync_write(struct drbd_device *device, struct drbd_peer_request *peer_req)
1943 1944
{
	struct drbd_peer_request *rs_req;
1945 1946
	bool rv = 0;

1947
	spin_lock_irq(&device->resource->req_lock);
1948
	list_for_each_entry(rs_req, &device->sync_ee, w.list) {
1949 1950
		if (overlaps(peer_req->i.sector, peer_req->i.size,
			     rs_req->i.sector, rs_req->i.size)) {
1951 1952 1953 1954
			rv = 1;
			break;
		}
	}
1955
	spin_unlock_irq(&device->resource->req_lock);
1956 1957 1958 1959

	return rv;
}

P
Philipp Reisner 已提交
1960 1961 1962 1963 1964 1965 1966 1967 1968
/* Called from receive_Data.
 * Synchronize packets on sock with packets on msock.
 *
 * This is here so even when a P_DATA packet traveling via sock overtook an Ack
 * packet traveling on msock, they are still processed in the order they have
 * been sent.
 *
 * Note: we don't care for Ack packets overtaking P_DATA packets.
 *
1969
 * In case packet_seq is larger than device->peer_seq number, there are
P
Philipp Reisner 已提交
1970
 * outstanding packets on the msock. We wait for them to arrive.
1971
 * In case we are the logically next packet, we update device->peer_seq
P
Philipp Reisner 已提交
1972 1973 1974 1975 1976 1977 1978 1979 1980
 * ourselves. Correctly handles 32bit wrap around.
 *
 * Assume we have a 10 GBit connection, that is about 1<<30 byte per second,
 * about 1<<21 sectors per second. So "worst" case, we have 1<<3 == 8 seconds
 * for the 24bit wrap (historical atomic_t guarantee on some archs), and we have
 * 1<<9 == 512 seconds aka ages for the 32bit wrap around...
 *
 * returns 0 if we may process the packet,
 * -ERESTARTSYS if we were interrupted (by disconnect signal). */
1981
static int wait_for_and_update_peer_seq(struct drbd_peer_device *peer_device, const u32 peer_seq)
P
Philipp Reisner 已提交
1982
{
1983
	struct drbd_device *device = peer_device->device;
P
Philipp Reisner 已提交
1984 1985
	DEFINE_WAIT(wait);
	long timeout;
1986
	int ret = 0, tp;
1987

1988
	if (!test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags))
1989 1990
		return 0;

1991
	spin_lock(&device->peer_seq_lock);
P
Philipp Reisner 已提交
1992
	for (;;) {
1993 1994
		if (!seq_greater(peer_seq - 1, device->peer_seq)) {
			device->peer_seq = seq_max(device->peer_seq, peer_seq);
P
Philipp Reisner 已提交
1995
			break;
1996
		}
1997

P
Philipp Reisner 已提交
1998 1999 2000 2001
		if (signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}
2002 2003

		rcu_read_lock();
2004
		tp = rcu_dereference(first_peer_device(device)->connection->net_conf)->two_primaries;
2005 2006 2007 2008 2009 2010
		rcu_read_unlock();

		if (!tp)
			break;

		/* Only need to wait if two_primaries is enabled */
2011 2012
		prepare_to_wait(&device->seq_wait, &wait, TASK_INTERRUPTIBLE);
		spin_unlock(&device->peer_seq_lock);
2013
		rcu_read_lock();
2014
		timeout = rcu_dereference(peer_device->connection->net_conf)->ping_timeo*HZ/10;
2015
		rcu_read_unlock();
2016
		timeout = schedule_timeout(timeout);
2017
		spin_lock(&device->peer_seq_lock);
2018
		if (!timeout) {
P
Philipp Reisner 已提交
2019
			ret = -ETIMEDOUT;
2020
			drbd_err(device, "Timed out waiting for missing ack packets; disconnecting\n");
P
Philipp Reisner 已提交
2021 2022 2023
			break;
		}
	}
2024 2025
	spin_unlock(&device->peer_seq_lock);
	finish_wait(&device->seq_wait, &wait);
P
Philipp Reisner 已提交
2026 2027 2028
	return ret;
}

2029 2030 2031
/* see also bio_flags_to_wire()
 * DRBD_REQ_*, because we need to semantically map the flags to data packet
 * flags and back. We may replicate to other kernel versions. */
2032
static unsigned long wire_flags_to_bio(u32 dpf)
2033
{
2034 2035 2036 2037
	return  (dpf & DP_RW_SYNC ? REQ_SYNC : 0) |
		(dpf & DP_FUA ? REQ_FUA : 0) |
		(dpf & DP_FLUSH ? REQ_FLUSH : 0) |
		(dpf & DP_DISCARD ? REQ_DISCARD : 0);
2038 2039
}

2040
static void fail_postponed_requests(struct drbd_device *device, sector_t sector,
2041 2042 2043 2044 2045
				    unsigned int size)
{
	struct drbd_interval *i;

    repeat:
2046
	drbd_for_each_overlap(i, &device->write_requests, sector, size) {
2047 2048 2049 2050 2051 2052 2053 2054 2055 2056
		struct drbd_request *req;
		struct bio_and_error m;

		if (!i->local)
			continue;
		req = container_of(i, struct drbd_request, i);
		if (!(req->rq_state & RQ_POSTPONED))
			continue;
		req->rq_state &= ~RQ_POSTPONED;
		__req_mod(req, NEG_ACKED, &m);
2057
		spin_unlock_irq(&device->resource->req_lock);
2058
		if (m.bio)
2059
			complete_master_bio(device, &m);
2060
		spin_lock_irq(&device->resource->req_lock);
2061 2062 2063 2064
		goto repeat;
	}
}

2065
static int handle_write_conflicts(struct drbd_device *device,
2066 2067
				  struct drbd_peer_request *peer_req)
{
2068
	struct drbd_connection *connection = peer_req->peer_device->connection;
2069
	bool resolve_conflicts = test_bit(RESOLVE_CONFLICTS, &connection->flags);
2070 2071 2072 2073 2074 2075 2076 2077 2078 2079
	sector_t sector = peer_req->i.sector;
	const unsigned int size = peer_req->i.size;
	struct drbd_interval *i;
	bool equal;
	int err;

	/*
	 * Inserting the peer request into the write_requests tree will prevent
	 * new conflicting local requests from being added.
	 */
2080
	drbd_insert_interval(&device->write_requests, &peer_req->i);
2081 2082

    repeat:
2083
	drbd_for_each_overlap(i, &device->write_requests, sector, size) {
2084 2085 2086 2087 2088 2089 2090 2091 2092
		if (i == &peer_req->i)
			continue;

		if (!i->local) {
			/*
			 * Our peer has sent a conflicting remote request; this
			 * should not happen in a two-node setup.  Wait for the
			 * earlier peer request to complete.
			 */
2093
			err = drbd_wait_misc(device, i);
2094 2095 2096 2097 2098 2099 2100 2101 2102
			if (err)
				goto out;
			goto repeat;
		}

		equal = i->sector == sector && i->size == size;
		if (resolve_conflicts) {
			/*
			 * If the peer request is fully contained within the
2103 2104 2105
			 * overlapping request, it can be considered overwritten
			 * and thus superseded; otherwise, it will be retried
			 * once all overlapping requests have completed.
2106
			 */
2107
			bool superseded = i->sector <= sector && i->sector +
2108 2109 2110
				       (i->size >> 9) >= sector + (size >> 9);

			if (!equal)
2111
				drbd_alert(device, "Concurrent writes detected: "
2112 2113 2114 2115
					       "local=%llus +%u, remote=%llus +%u, "
					       "assuming %s came first\n",
					  (unsigned long long)i->sector, i->size,
					  (unsigned long long)sector, size,
2116
					  superseded ? "local" : "remote");
2117

2118
			inc_unacked(device);
2119
			peer_req->w.cb = superseded ? e_send_superseded :
2120
						   e_send_retry_write;
2121
			list_add_tail(&peer_req->w.list, &device->done_ee);
2122
			wake_asender(connection);
2123 2124 2125 2126 2127 2128 2129 2130

			err = -ENOENT;
			goto out;
		} else {
			struct drbd_request *req =
				container_of(i, struct drbd_request, i);

			if (!equal)
2131
				drbd_alert(device, "Concurrent writes detected: "
2132 2133 2134 2135 2136 2137 2138 2139
					       "local=%llus +%u, remote=%llus +%u\n",
					  (unsigned long long)i->sector, i->size,
					  (unsigned long long)sector, size);

			if (req->rq_state & RQ_LOCAL_PENDING ||
			    !(req->rq_state & RQ_POSTPONED)) {
				/*
				 * Wait for the node with the discard flag to
2140 2141 2142
				 * decide if this request has been superseded
				 * or needs to be retried.
				 * Requests that have been superseded will
2143 2144 2145 2146 2147 2148
				 * disappear from the write_requests tree.
				 *
				 * In addition, wait for the conflicting
				 * request to finish locally before submitting
				 * the conflicting peer request.
				 */
2149
				err = drbd_wait_misc(device, &req->i);
2150
				if (err) {
2151
					_conn_request_state(connection, NS(conn, C_TIMEOUT), CS_HARD);
2152
					fail_postponed_requests(device, sector, size);
2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167
					goto out;
				}
				goto repeat;
			}
			/*
			 * Remember to restart the conflicting requests after
			 * the new peer request has completed.
			 */
			peer_req->flags |= EE_RESTART_REQUESTS;
		}
	}
	err = 0;

    out:
	if (err)
2168
		drbd_remove_epoch_entry_interval(device, peer_req);
2169 2170 2171
	return err;
}

P
Philipp Reisner 已提交
2172
/* mirrored write */
2173
static int receive_Data(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
2174
{
2175
	struct drbd_peer_device *peer_device;
2176
	struct drbd_device *device;
P
Philipp Reisner 已提交
2177
	sector_t sector;
2178
	struct drbd_peer_request *peer_req;
2179
	struct p_data *p = pi->data;
2180
	u32 peer_seq = be32_to_cpu(p->seq_num);
P
Philipp Reisner 已提交
2181 2182
	int rw = WRITE;
	u32 dp_flags;
2183
	int err, tp;
P
Philipp Reisner 已提交
2184

2185 2186
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
2187
		return -EIO;
2188
	device = peer_device->device;
P
Philipp Reisner 已提交
2189

2190
	if (!get_ldev(device)) {
2191 2192
		int err2;

2193 2194
		err = wait_for_and_update_peer_seq(peer_device, peer_seq);
		drbd_send_ack_dp(peer_device, P_NEG_ACK, p, pi->size);
2195
		atomic_inc(&connection->current_epoch->epoch_size);
2196
		err2 = drbd_drain_block(peer_device, pi->size);
2197 2198 2199
		if (!err)
			err = err2;
		return err;
P
Philipp Reisner 已提交
2200 2201
	}

2202 2203 2204 2205 2206
	/*
	 * Corresponding put_ldev done either below (on various errors), or in
	 * drbd_peer_request_endio, if we successfully submit the data at the
	 * end of this function.
	 */
P
Philipp Reisner 已提交
2207 2208

	sector = be64_to_cpu(p->sector);
2209
	peer_req = read_in_block(peer_device, p->block_id, sector, pi->size);
2210
	if (!peer_req) {
2211
		put_ldev(device);
2212
		return -EIO;
P
Philipp Reisner 已提交
2213 2214
	}

2215
	peer_req->w.cb = e_end_block;
P
Philipp Reisner 已提交
2216

2217
	dp_flags = be32_to_cpu(p->dp_flags);
2218
	rw |= wire_flags_to_bio(dp_flags);
2219
	if (peer_req->pages == NULL) {
2220 2221
		D_ASSERT(device, peer_req->i.size == 0);
		D_ASSERT(device, dp_flags & DP_FLUSH);
2222
	}
2223 2224

	if (dp_flags & DP_MAY_SET_IN_SYNC)
2225
		peer_req->flags |= EE_MAY_SET_IN_SYNC;
2226

2227 2228
	spin_lock(&connection->epoch_lock);
	peer_req->epoch = connection->current_epoch;
2229 2230
	atomic_inc(&peer_req->epoch->epoch_size);
	atomic_inc(&peer_req->epoch->active);
2231
	spin_unlock(&connection->epoch_lock);
P
Philipp Reisner 已提交
2232

2233
	rcu_read_lock();
2234
	tp = rcu_dereference(peer_device->connection->net_conf)->two_primaries;
2235 2236 2237
	rcu_read_unlock();
	if (tp) {
		peer_req->flags |= EE_IN_INTERVAL_TREE;
2238
		err = wait_for_and_update_peer_seq(peer_device, peer_seq);
2239
		if (err)
P
Philipp Reisner 已提交
2240
			goto out_interrupted;
2241
		spin_lock_irq(&device->resource->req_lock);
2242
		err = handle_write_conflicts(device, peer_req);
2243
		if (err) {
2244
			spin_unlock_irq(&device->resource->req_lock);
2245
			if (err == -ENOENT) {
2246
				put_ldev(device);
2247
				return 0;
P
Philipp Reisner 已提交
2248
			}
2249
			goto out_interrupted;
P
Philipp Reisner 已提交
2250
		}
2251
	} else {
2252
		update_peer_seq(peer_device, peer_seq);
2253
		spin_lock_irq(&device->resource->req_lock);
2254
	}
2255
	list_add(&peer_req->w.list, &device->active_ee);
2256
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
2257

2258 2259
	if (device->state.conn == C_SYNC_TARGET)
		wait_event(device->ee_wait, !overlapping_resync_write(device, peer_req));
P
Philipp Reisner 已提交
2260

2261
	if (peer_device->connection->agreed_pro_version < 100) {
2262
		rcu_read_lock();
2263
		switch (rcu_dereference(peer_device->connection->net_conf)->wire_protocol) {
2264 2265 2266 2267 2268 2269
		case DRBD_PROT_C:
			dp_flags |= DP_SEND_WRITE_ACK;
			break;
		case DRBD_PROT_B:
			dp_flags |= DP_SEND_RECEIVE_ACK;
			break;
P
Philipp Reisner 已提交
2270
		}
2271
		rcu_read_unlock();
P
Philipp Reisner 已提交
2272 2273
	}

2274 2275
	if (dp_flags & DP_SEND_WRITE_ACK) {
		peer_req->flags |= EE_SEND_WRITE_ACK;
2276
		inc_unacked(device);
P
Philipp Reisner 已提交
2277 2278
		/* corresponding dec_unacked() in e_end_block()
		 * respective _drbd_clear_done_ee */
2279 2280 2281
	}

	if (dp_flags & DP_SEND_RECEIVE_ACK) {
P
Philipp Reisner 已提交
2282 2283
		/* I really don't like it that the receiver thread
		 * sends on the msock, but anyways */
2284
		drbd_send_ack(first_peer_device(device), P_RECV_ACK, peer_req);
P
Philipp Reisner 已提交
2285 2286
	}

2287
	if (device->state.pdsk < D_INCONSISTENT) {
P
Philipp Reisner 已提交
2288
		/* In case we have the only disk of the cluster, */
2289
		drbd_set_out_of_sync(device, peer_req->i.sector, peer_req->i.size);
2290 2291
		peer_req->flags |= EE_CALL_AL_COMPLETE_IO;
		peer_req->flags &= ~EE_MAY_SET_IN_SYNC;
2292
		drbd_al_begin_io(device, &peer_req->i, true);
P
Philipp Reisner 已提交
2293 2294
	}

2295
	err = drbd_submit_peer_request(device, peer_req, rw, DRBD_FAULT_DT_WR);
2296 2297
	if (!err)
		return 0;
P
Philipp Reisner 已提交
2298

2299
	/* don't care for the reason here */
2300
	drbd_err(device, "submit failed, triggering re-connect\n");
2301
	spin_lock_irq(&device->resource->req_lock);
2302
	list_del(&peer_req->w.list);
2303
	drbd_remove_epoch_entry_interval(device, peer_req);
2304
	spin_unlock_irq(&device->resource->req_lock);
2305
	if (peer_req->flags & EE_CALL_AL_COMPLETE_IO)
2306
		drbd_al_complete_io(device, &peer_req->i);
2307

P
Philipp Reisner 已提交
2308
out_interrupted:
2309
	drbd_may_finish_epoch(connection, peer_req->epoch, EV_PUT + EV_CLEANUP);
2310 2311
	put_ldev(device);
	drbd_free_peer_req(device, peer_req);
2312
	return err;
P
Philipp Reisner 已提交
2313 2314
}

2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325
/* We may throttle resync, if the lower device seems to be busy,
 * and current sync rate is above c_min_rate.
 *
 * To decide whether or not the lower device is busy, we use a scheme similar
 * to MD RAID is_mddev_idle(): if the partition stats reveal "significant"
 * (more than 64 sectors) of activity we cannot account for with our own resync
 * activity, it obviously is "busy".
 *
 * The current sync rate used here uses only the most recent two step marks,
 * to have a short time average so we can react faster.
 */
2326
int drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector)
2327
{
2328
	struct gendisk *disk = device->ldev->backing_bdev->bd_contains->bd_disk;
2329
	unsigned long db, dt, dbdt;
2330
	struct lc_element *tmp;
2331 2332
	int curr_events;
	int throttle = 0;
P
Philipp Reisner 已提交
2333 2334 2335
	unsigned int c_min_rate;

	rcu_read_lock();
2336
	c_min_rate = rcu_dereference(device->ldev->disk_conf)->c_min_rate;
P
Philipp Reisner 已提交
2337
	rcu_read_unlock();
2338 2339

	/* feature disabled? */
P
Philipp Reisner 已提交
2340
	if (c_min_rate == 0)
2341 2342
		return 0;

2343 2344
	spin_lock_irq(&device->al_lock);
	tmp = lc_find(device->resync, BM_SECT_TO_EXT(sector));
2345 2346 2347
	if (tmp) {
		struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce);
		if (test_bit(BME_PRIORITY, &bm_ext->flags)) {
2348
			spin_unlock_irq(&device->al_lock);
2349 2350 2351 2352
			return 0;
		}
		/* Do not slow down if app IO is already waiting for this extent */
	}
2353
	spin_unlock_irq(&device->al_lock);
2354

2355 2356
	curr_events = (int)part_stat_read(&disk->part0, sectors[0]) +
		      (int)part_stat_read(&disk->part0, sectors[1]) -
2357
			atomic_read(&device->rs_sect_ev);
2358

2359
	if (!device->rs_last_events || curr_events - device->rs_last_events > 64) {
2360 2361 2362
		unsigned long rs_left;
		int i;

2363
		device->rs_last_events = curr_events;
2364 2365 2366

		/* sync speed average over the last 2*DRBD_SYNC_MARK_STEP,
		 * approx. */
2367
		i = (device->rs_last_mark + DRBD_SYNC_MARKS-1) % DRBD_SYNC_MARKS;
2368

2369 2370
		if (device->state.conn == C_VERIFY_S || device->state.conn == C_VERIFY_T)
			rs_left = device->ov_left;
2371
		else
2372
			rs_left = drbd_bm_total_weight(device) - device->rs_failed;
2373

2374
		dt = ((long)jiffies - (long)device->rs_mark_time[i]) / HZ;
2375 2376
		if (!dt)
			dt++;
2377
		db = device->rs_mark_left[i] - rs_left;
2378 2379
		dbdt = Bit2KB(db/dt);

P
Philipp Reisner 已提交
2380
		if (dbdt > c_min_rate)
2381 2382 2383 2384 2385 2386
			throttle = 1;
	}
	return throttle;
}


2387
static int receive_DataRequest(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
2388
{
2389
	struct drbd_peer_device *peer_device;
2390
	struct drbd_device *device;
P
Philipp Reisner 已提交
2391
	sector_t sector;
2392
	sector_t capacity;
2393
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
2394
	struct digest_info *di = NULL;
2395
	int size, verb;
P
Philipp Reisner 已提交
2396
	unsigned int fault_type;
2397
	struct p_block_req *p =	pi->data;
2398

2399 2400
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
2401
		return -EIO;
2402
	device = peer_device->device;
2403
	capacity = drbd_get_capacity(device->this_bdev);
P
Philipp Reisner 已提交
2404 2405 2406 2407

	sector = be64_to_cpu(p->sector);
	size   = be32_to_cpu(p->blksize);

2408
	if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) {
2409
		drbd_err(device, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__,
P
Philipp Reisner 已提交
2410
				(unsigned long long)sector, size);
2411
		return -EINVAL;
P
Philipp Reisner 已提交
2412 2413
	}
	if (sector + (size>>9) > capacity) {
2414
		drbd_err(device, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__,
P
Philipp Reisner 已提交
2415
				(unsigned long long)sector, size);
2416
		return -EINVAL;
P
Philipp Reisner 已提交
2417 2418
	}

2419
	if (!get_ldev_if_state(device, D_UP_TO_DATE)) {
2420
		verb = 1;
2421
		switch (pi->cmd) {
2422
		case P_DATA_REQUEST:
2423
			drbd_send_ack_rp(peer_device, P_NEG_DREPLY, p);
2424 2425 2426 2427
			break;
		case P_RS_DATA_REQUEST:
		case P_CSUM_RS_REQUEST:
		case P_OV_REQUEST:
2428
			drbd_send_ack_rp(peer_device, P_NEG_RS_DREPLY , p);
2429 2430 2431
			break;
		case P_OV_REPLY:
			verb = 0;
2432
			dec_rs_pending(device);
2433
			drbd_send_ack_ex(peer_device, P_OV_RESULT, sector, size, ID_IN_SYNC);
2434 2435
			break;
		default:
2436
			BUG();
2437 2438
		}
		if (verb && __ratelimit(&drbd_ratelimit_state))
2439
			drbd_err(device, "Can not satisfy peer's read request, "
P
Philipp Reisner 已提交
2440
			    "no local data.\n");
2441

L
Lars Ellenberg 已提交
2442
		/* drain possibly payload */
2443
		return drbd_drain_block(peer_device, pi->size);
P
Philipp Reisner 已提交
2444 2445 2446 2447 2448
	}

	/* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD
	 * "criss-cross" setup, that might cause write-out on some other DRBD,
	 * which in turn might block on the other node at this very place.  */
2449
	peer_req = drbd_alloc_peer_req(peer_device, p->block_id, sector, size, GFP_NOIO);
2450
	if (!peer_req) {
2451
		put_ldev(device);
2452
		return -ENOMEM;
P
Philipp Reisner 已提交
2453 2454
	}

2455
	switch (pi->cmd) {
P
Philipp Reisner 已提交
2456
	case P_DATA_REQUEST:
2457
		peer_req->w.cb = w_e_end_data_req;
P
Philipp Reisner 已提交
2458
		fault_type = DRBD_FAULT_DT_RD;
2459 2460 2461
		/* application IO, don't drbd_rs_begin_io */
		goto submit;

P
Philipp Reisner 已提交
2462
	case P_RS_DATA_REQUEST:
2463
		peer_req->w.cb = w_e_end_rsdata_req;
P
Philipp Reisner 已提交
2464
		fault_type = DRBD_FAULT_RS_RD;
2465
		/* used in the sector offset progress display */
2466
		device->bm_resync_fo = BM_SECT_TO_BIT(sector);
P
Philipp Reisner 已提交
2467 2468 2469 2470 2471
		break;

	case P_OV_REPLY:
	case P_CSUM_RS_REQUEST:
		fault_type = DRBD_FAULT_RS_RD;
2472
		di = kmalloc(sizeof(*di) + pi->size, GFP_NOIO);
P
Philipp Reisner 已提交
2473 2474 2475
		if (!di)
			goto out_free_e;

2476
		di->digest_size = pi->size;
P
Philipp Reisner 已提交
2477 2478
		di->digest = (((char *)di)+sizeof(struct digest_info));

2479 2480
		peer_req->digest = di;
		peer_req->flags |= EE_HAS_DIGEST;
2481

2482
		if (drbd_recv_all(peer_device->connection, di->digest, pi->size))
P
Philipp Reisner 已提交
2483 2484
			goto out_free_e;

2485
		if (pi->cmd == P_CSUM_RS_REQUEST) {
2486
			D_ASSERT(device, peer_device->connection->agreed_pro_version >= 89);
2487
			peer_req->w.cb = w_e_end_csum_rs_req;
2488
			/* used in the sector offset progress display */
2489
			device->bm_resync_fo = BM_SECT_TO_BIT(sector);
2490
		} else if (pi->cmd == P_OV_REPLY) {
2491
			/* track progress, we may need to throttle */
2492
			atomic_add(size >> 9, &device->rs_sect_in);
2493
			peer_req->w.cb = w_e_end_ov_reply;
2494
			dec_rs_pending(device);
2495 2496 2497
			/* drbd_rs_begin_io done when we sent this request,
			 * but accounting still needs to be done. */
			goto submit_for_resync;
P
Philipp Reisner 已提交
2498 2499 2500 2501
		}
		break;

	case P_OV_REQUEST:
2502
		if (device->ov_start_sector == ~(sector_t)0 &&
2503
		    peer_device->connection->agreed_pro_version >= 90) {
2504 2505
			unsigned long now = jiffies;
			int i;
2506 2507 2508 2509
			device->ov_start_sector = sector;
			device->ov_position = sector;
			device->ov_left = drbd_bm_bits(device) - BM_SECT_TO_BIT(sector);
			device->rs_total = device->ov_left;
2510
			for (i = 0; i < DRBD_SYNC_MARKS; i++) {
2511 2512
				device->rs_mark_left[i] = device->ov_left;
				device->rs_mark_time[i] = now;
2513
			}
2514
			drbd_info(device, "Online Verify start sector: %llu\n",
P
Philipp Reisner 已提交
2515 2516
					(unsigned long long)sector);
		}
2517
		peer_req->w.cb = w_e_end_ov_req;
P
Philipp Reisner 已提交
2518 2519 2520 2521
		fault_type = DRBD_FAULT_RS_RD;
		break;

	default:
2522
		BUG();
P
Philipp Reisner 已提交
2523 2524
	}

2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546
	/* Throttle, drbd_rs_begin_io and submit should become asynchronous
	 * wrt the receiver, but it is not as straightforward as it may seem.
	 * Various places in the resync start and stop logic assume resync
	 * requests are processed in order, requeuing this on the worker thread
	 * introduces a bunch of new code for synchronization between threads.
	 *
	 * Unlimited throttling before drbd_rs_begin_io may stall the resync
	 * "forever", throttling after drbd_rs_begin_io will lock that extent
	 * for application writes for the same time.  For now, just throttle
	 * here, where the rest of the code expects the receiver to sleep for
	 * a while, anyways.
	 */

	/* Throttle before drbd_rs_begin_io, as that locks out application IO;
	 * this defers syncer requests for some time, before letting at least
	 * on request through.  The resync controller on the receiving side
	 * will adapt to the incoming rate accordingly.
	 *
	 * We cannot throttle here if remote is Primary/SyncTarget:
	 * we would also throttle its application reads.
	 * In that case, throttling is done on the SyncTarget only.
	 */
2547
	if (device->state.peer != R_PRIMARY && drbd_rs_should_slow_down(device, sector))
2548
		schedule_timeout_uninterruptible(HZ/10);
2549
	if (drbd_rs_begin_io(device, sector))
2550
		goto out_free_e;
P
Philipp Reisner 已提交
2551

2552
submit_for_resync:
2553
	atomic_add(size >> 9, &device->rs_sect_ev);
2554

2555
submit:
2556
	inc_unacked(device);
2557
	spin_lock_irq(&device->resource->req_lock);
2558
	list_add_tail(&peer_req->w.list, &device->read_ee);
2559
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
2560

2561
	if (drbd_submit_peer_request(device, peer_req, READ, fault_type) == 0)
2562
		return 0;
P
Philipp Reisner 已提交
2563

2564
	/* don't care for the reason here */
2565
	drbd_err(device, "submit failed, triggering re-connect\n");
2566
	spin_lock_irq(&device->resource->req_lock);
2567
	list_del(&peer_req->w.list);
2568
	spin_unlock_irq(&device->resource->req_lock);
2569 2570
	/* no drbd_rs_complete_io(), we are dropping the connection anyways */

P
Philipp Reisner 已提交
2571
out_free_e:
2572 2573
	put_ldev(device);
	drbd_free_peer_req(device, peer_req);
2574
	return -EIO;
P
Philipp Reisner 已提交
2575 2576
}

2577 2578 2579 2580
/**
 * drbd_asb_recover_0p  -  Recover after split-brain with no remaining primaries
 */
static int drbd_asb_recover_0p(struct drbd_peer_device *peer_device) __must_hold(local)
P
Philipp Reisner 已提交
2581
{
2582
	struct drbd_device *device = peer_device->device;
P
Philipp Reisner 已提交
2583 2584
	int self, peer, rv = -100;
	unsigned long ch_self, ch_peer;
2585
	enum drbd_after_sb_p after_sb_0p;
P
Philipp Reisner 已提交
2586

2587 2588
	self = device->ldev->md.uuid[UI_BITMAP] & 1;
	peer = device->p_uuid[UI_BITMAP] & 1;
P
Philipp Reisner 已提交
2589

2590 2591
	ch_peer = device->p_uuid[UI_SIZE];
	ch_self = device->comm_bm_set;
P
Philipp Reisner 已提交
2592

2593
	rcu_read_lock();
2594
	after_sb_0p = rcu_dereference(peer_device->connection->net_conf)->after_sb_0p;
2595 2596
	rcu_read_unlock();
	switch (after_sb_0p) {
P
Philipp Reisner 已提交
2597 2598 2599
	case ASB_CONSENSUS:
	case ASB_DISCARD_SECONDARY:
	case ASB_CALL_HELPER:
2600
	case ASB_VIOLENTLY:
2601
		drbd_err(device, "Configuration error.\n");
P
Philipp Reisner 已提交
2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624
		break;
	case ASB_DISCONNECT:
		break;
	case ASB_DISCARD_YOUNGER_PRI:
		if (self == 0 && peer == 1) {
			rv = -1;
			break;
		}
		if (self == 1 && peer == 0) {
			rv =  1;
			break;
		}
		/* Else fall through to one of the other strategies... */
	case ASB_DISCARD_OLDER_PRI:
		if (self == 0 && peer == 1) {
			rv = 1;
			break;
		}
		if (self == 1 && peer == 0) {
			rv = -1;
			break;
		}
		/* Else fall through to one of the other strategies... */
2625
		drbd_warn(device, "Discard younger/older primary did not find a decision\n"
P
Philipp Reisner 已提交
2626 2627 2628
		     "Using discard-least-changes instead\n");
	case ASB_DISCARD_ZERO_CHG:
		if (ch_peer == 0 && ch_self == 0) {
2629
			rv = test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags)
P
Philipp Reisner 已提交
2630 2631 2632 2633 2634 2635
				? -1 : 1;
			break;
		} else {
			if (ch_peer == 0) { rv =  1; break; }
			if (ch_self == 0) { rv = -1; break; }
		}
2636
		if (after_sb_0p == ASB_DISCARD_ZERO_CHG)
P
Philipp Reisner 已提交
2637 2638 2639 2640 2641 2642 2643 2644
			break;
	case ASB_DISCARD_LEAST_CHG:
		if	(ch_self < ch_peer)
			rv = -1;
		else if (ch_self > ch_peer)
			rv =  1;
		else /* ( ch_self == ch_peer ) */
		     /* Well, then use something else. */
2645
			rv = test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags)
P
Philipp Reisner 已提交
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657
				? -1 : 1;
		break;
	case ASB_DISCARD_LOCAL:
		rv = -1;
		break;
	case ASB_DISCARD_REMOTE:
		rv =  1;
	}

	return rv;
}

2658 2659 2660 2661
/**
 * drbd_asb_recover_1p  -  Recover after split-brain with one remaining primary
 */
static int drbd_asb_recover_1p(struct drbd_peer_device *peer_device) __must_hold(local)
P
Philipp Reisner 已提交
2662
{
2663
	struct drbd_device *device = peer_device->device;
2664
	int hg, rv = -100;
2665
	enum drbd_after_sb_p after_sb_1p;
P
Philipp Reisner 已提交
2666

2667
	rcu_read_lock();
2668
	after_sb_1p = rcu_dereference(peer_device->connection->net_conf)->after_sb_1p;
2669 2670
	rcu_read_unlock();
	switch (after_sb_1p) {
P
Philipp Reisner 已提交
2671 2672 2673 2674 2675
	case ASB_DISCARD_YOUNGER_PRI:
	case ASB_DISCARD_OLDER_PRI:
	case ASB_DISCARD_LEAST_CHG:
	case ASB_DISCARD_LOCAL:
	case ASB_DISCARD_REMOTE:
2676
	case ASB_DISCARD_ZERO_CHG:
2677
		drbd_err(device, "Configuration error.\n");
P
Philipp Reisner 已提交
2678 2679 2680 2681
		break;
	case ASB_DISCONNECT:
		break;
	case ASB_CONSENSUS:
2682
		hg = drbd_asb_recover_0p(peer_device);
2683
		if (hg == -1 && device->state.role == R_SECONDARY)
P
Philipp Reisner 已提交
2684
			rv = hg;
2685
		if (hg == 1  && device->state.role == R_PRIMARY)
P
Philipp Reisner 已提交
2686 2687 2688
			rv = hg;
		break;
	case ASB_VIOLENTLY:
2689
		rv = drbd_asb_recover_0p(peer_device);
P
Philipp Reisner 已提交
2690 2691
		break;
	case ASB_DISCARD_SECONDARY:
2692
		return device->state.role == R_PRIMARY ? 1 : -1;
P
Philipp Reisner 已提交
2693
	case ASB_CALL_HELPER:
2694
		hg = drbd_asb_recover_0p(peer_device);
2695
		if (hg == -1 && device->state.role == R_PRIMARY) {
2696 2697
			enum drbd_state_rv rv2;

P
Philipp Reisner 已提交
2698 2699 2700
			 /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE,
			  * we might be here in C_WF_REPORT_PARAMS which is transient.
			  * we do not need to wait for the after state change work either. */
2701
			rv2 = drbd_change_state(device, CS_VERBOSE, NS(role, R_SECONDARY));
2702
			if (rv2 != SS_SUCCESS) {
2703
				drbd_khelper(device, "pri-lost-after-sb");
P
Philipp Reisner 已提交
2704
			} else {
2705
				drbd_warn(device, "Successfully gave up primary role.\n");
P
Philipp Reisner 已提交
2706 2707 2708 2709 2710 2711 2712 2713 2714
				rv = hg;
			}
		} else
			rv = hg;
	}

	return rv;
}

2715 2716 2717 2718
/**
 * drbd_asb_recover_2p  -  Recover after split-brain with two remaining primaries
 */
static int drbd_asb_recover_2p(struct drbd_peer_device *peer_device) __must_hold(local)
P
Philipp Reisner 已提交
2719
{
2720
	struct drbd_device *device = peer_device->device;
2721
	int hg, rv = -100;
2722
	enum drbd_after_sb_p after_sb_2p;
P
Philipp Reisner 已提交
2723

2724
	rcu_read_lock();
2725
	after_sb_2p = rcu_dereference(peer_device->connection->net_conf)->after_sb_2p;
2726 2727
	rcu_read_unlock();
	switch (after_sb_2p) {
P
Philipp Reisner 已提交
2728 2729 2730 2731 2732 2733 2734
	case ASB_DISCARD_YOUNGER_PRI:
	case ASB_DISCARD_OLDER_PRI:
	case ASB_DISCARD_LEAST_CHG:
	case ASB_DISCARD_LOCAL:
	case ASB_DISCARD_REMOTE:
	case ASB_CONSENSUS:
	case ASB_DISCARD_SECONDARY:
2735
	case ASB_DISCARD_ZERO_CHG:
2736
		drbd_err(device, "Configuration error.\n");
P
Philipp Reisner 已提交
2737 2738
		break;
	case ASB_VIOLENTLY:
2739
		rv = drbd_asb_recover_0p(peer_device);
P
Philipp Reisner 已提交
2740 2741 2742 2743
		break;
	case ASB_DISCONNECT:
		break;
	case ASB_CALL_HELPER:
2744
		hg = drbd_asb_recover_0p(peer_device);
P
Philipp Reisner 已提交
2745
		if (hg == -1) {
2746 2747
			enum drbd_state_rv rv2;

P
Philipp Reisner 已提交
2748 2749 2750
			 /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE,
			  * we might be here in C_WF_REPORT_PARAMS which is transient.
			  * we do not need to wait for the after state change work either. */
2751
			rv2 = drbd_change_state(device, CS_VERBOSE, NS(role, R_SECONDARY));
2752
			if (rv2 != SS_SUCCESS) {
2753
				drbd_khelper(device, "pri-lost-after-sb");
P
Philipp Reisner 已提交
2754
			} else {
2755
				drbd_warn(device, "Successfully gave up primary role.\n");
P
Philipp Reisner 已提交
2756 2757 2758 2759 2760 2761 2762 2763 2764
				rv = hg;
			}
		} else
			rv = hg;
	}

	return rv;
}

2765
static void drbd_uuid_dump(struct drbd_device *device, char *text, u64 *uuid,
P
Philipp Reisner 已提交
2766 2767 2768
			   u64 bits, u64 flags)
{
	if (!uuid) {
2769
		drbd_info(device, "%s uuid info vanished while I was looking!\n", text);
P
Philipp Reisner 已提交
2770 2771
		return;
	}
2772
	drbd_info(device, "%s %016llX:%016llX:%016llX:%016llX bits:%llu flags:%llX\n",
P
Philipp Reisner 已提交
2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790
	     text,
	     (unsigned long long)uuid[UI_CURRENT],
	     (unsigned long long)uuid[UI_BITMAP],
	     (unsigned long long)uuid[UI_HISTORY_START],
	     (unsigned long long)uuid[UI_HISTORY_END],
	     (unsigned long long)bits,
	     (unsigned long long)flags);
}

/*
  100	after split brain try auto recover
    2	C_SYNC_SOURCE set BitMap
    1	C_SYNC_SOURCE use BitMap
    0	no Sync
   -1	C_SYNC_TARGET use BitMap
   -2	C_SYNC_TARGET set BitMap
 -100	after split brain, disconnect
-1000	unrelated data
2791 2792
-1091   requires proto 91
-1096   requires proto 96
P
Philipp Reisner 已提交
2793
 */
2794
static int drbd_uuid_compare(struct drbd_device *device, int *rule_nr) __must_hold(local)
P
Philipp Reisner 已提交
2795 2796 2797 2798
{
	u64 self, peer;
	int i, j;

2799 2800
	self = device->ldev->md.uuid[UI_CURRENT] & ~((u64)1);
	peer = device->p_uuid[UI_CURRENT] & ~((u64)1);
P
Philipp Reisner 已提交
2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818

	*rule_nr = 10;
	if (self == UUID_JUST_CREATED && peer == UUID_JUST_CREATED)
		return 0;

	*rule_nr = 20;
	if ((self == UUID_JUST_CREATED || self == (u64)0) &&
	     peer != UUID_JUST_CREATED)
		return -2;

	*rule_nr = 30;
	if (self != UUID_JUST_CREATED &&
	    (peer == UUID_JUST_CREATED || peer == (u64)0))
		return 2;

	if (self == peer) {
		int rct, dc; /* roles at crash time */

2819
		if (device->p_uuid[UI_BITMAP] == (u64)0 && device->ldev->md.uuid[UI_BITMAP] != (u64)0) {
P
Philipp Reisner 已提交
2820

2821
			if (first_peer_device(device)->connection->agreed_pro_version < 91)
2822
				return -1091;
P
Philipp Reisner 已提交
2823

2824 2825
			if ((device->ldev->md.uuid[UI_BITMAP] & ~((u64)1)) == (device->p_uuid[UI_HISTORY_START] & ~((u64)1)) &&
			    (device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (device->p_uuid[UI_HISTORY_START + 1] & ~((u64)1))) {
2826
				drbd_info(device, "was SyncSource, missed the resync finished event, corrected myself:\n");
2827 2828 2829
				drbd_uuid_move_history(device);
				device->ldev->md.uuid[UI_HISTORY_START] = device->ldev->md.uuid[UI_BITMAP];
				device->ldev->md.uuid[UI_BITMAP] = 0;
P
Philipp Reisner 已提交
2830

2831 2832
				drbd_uuid_dump(device, "self", device->ldev->md.uuid,
					       device->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(device) : 0, 0);
P
Philipp Reisner 已提交
2833 2834
				*rule_nr = 34;
			} else {
2835
				drbd_info(device, "was SyncSource (peer failed to write sync_uuid)\n");
P
Philipp Reisner 已提交
2836 2837 2838 2839 2840 2841
				*rule_nr = 36;
			}

			return 1;
		}

2842
		if (device->ldev->md.uuid[UI_BITMAP] == (u64)0 && device->p_uuid[UI_BITMAP] != (u64)0) {
P
Philipp Reisner 已提交
2843

2844
			if (first_peer_device(device)->connection->agreed_pro_version < 91)
2845
				return -1091;
P
Philipp Reisner 已提交
2846

2847 2848
			if ((device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (device->p_uuid[UI_BITMAP] & ~((u64)1)) &&
			    (device->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == (device->p_uuid[UI_HISTORY_START] & ~((u64)1))) {
2849
				drbd_info(device, "was SyncTarget, peer missed the resync finished event, corrected peer:\n");
P
Philipp Reisner 已提交
2850

2851 2852 2853
				device->p_uuid[UI_HISTORY_START + 1] = device->p_uuid[UI_HISTORY_START];
				device->p_uuid[UI_HISTORY_START] = device->p_uuid[UI_BITMAP];
				device->p_uuid[UI_BITMAP] = 0UL;
P
Philipp Reisner 已提交
2854

2855
				drbd_uuid_dump(device, "peer", device->p_uuid, device->p_uuid[UI_SIZE], device->p_uuid[UI_FLAGS]);
P
Philipp Reisner 已提交
2856 2857
				*rule_nr = 35;
			} else {
2858
				drbd_info(device, "was SyncTarget (failed to write sync_uuid)\n");
P
Philipp Reisner 已提交
2859 2860 2861 2862 2863 2864 2865
				*rule_nr = 37;
			}

			return -1;
		}

		/* Common power [off|failure] */
2866 2867
		rct = (test_bit(CRASHED_PRIMARY, &device->flags) ? 1 : 0) +
			(device->p_uuid[UI_FLAGS] & 2);
P
Philipp Reisner 已提交
2868 2869 2870 2871 2872 2873 2874 2875 2876
		/* lowest bit is set when we were primary,
		 * next bit (weight 2) is set when peer was primary */
		*rule_nr = 40;

		switch (rct) {
		case 0: /* !self_pri && !peer_pri */ return 0;
		case 1: /*  self_pri && !peer_pri */ return 1;
		case 2: /* !self_pri &&  peer_pri */ return -1;
		case 3: /*  self_pri &&  peer_pri */
2877
			dc = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags);
P
Philipp Reisner 已提交
2878 2879 2880 2881 2882
			return dc ? -1 : 1;
		}
	}

	*rule_nr = 50;
2883
	peer = device->p_uuid[UI_BITMAP] & ~((u64)1);
P
Philipp Reisner 已提交
2884 2885 2886 2887
	if (self == peer)
		return -1;

	*rule_nr = 51;
2888
	peer = device->p_uuid[UI_HISTORY_START] & ~((u64)1);
P
Philipp Reisner 已提交
2889
	if (self == peer) {
2890
		if (first_peer_device(device)->connection->agreed_pro_version < 96 ?
2891 2892 2893
		    (device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) ==
		    (device->p_uuid[UI_HISTORY_START + 1] & ~((u64)1)) :
		    peer + UUID_NEW_BM_OFFSET == (device->p_uuid[UI_BITMAP] & ~((u64)1))) {
P
Philipp Reisner 已提交
2894 2895 2896
			/* The last P_SYNC_UUID did not get though. Undo the last start of
			   resync as sync source modifications of the peer's UUIDs. */

2897
			if (first_peer_device(device)->connection->agreed_pro_version < 91)
2898
				return -1091;
P
Philipp Reisner 已提交
2899

2900 2901
			device->p_uuid[UI_BITMAP] = device->p_uuid[UI_HISTORY_START];
			device->p_uuid[UI_HISTORY_START] = device->p_uuid[UI_HISTORY_START + 1];
2902

2903
			drbd_info(device, "Lost last syncUUID packet, corrected:\n");
2904
			drbd_uuid_dump(device, "peer", device->p_uuid, device->p_uuid[UI_SIZE], device->p_uuid[UI_FLAGS]);
2905

P
Philipp Reisner 已提交
2906 2907 2908 2909 2910
			return -1;
		}
	}

	*rule_nr = 60;
2911
	self = device->ldev->md.uuid[UI_CURRENT] & ~((u64)1);
P
Philipp Reisner 已提交
2912
	for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) {
2913
		peer = device->p_uuid[i] & ~((u64)1);
P
Philipp Reisner 已提交
2914 2915 2916 2917 2918
		if (self == peer)
			return -2;
	}

	*rule_nr = 70;
2919 2920
	self = device->ldev->md.uuid[UI_BITMAP] & ~((u64)1);
	peer = device->p_uuid[UI_CURRENT] & ~((u64)1);
P
Philipp Reisner 已提交
2921 2922 2923 2924
	if (self == peer)
		return 1;

	*rule_nr = 71;
2925
	self = device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1);
P
Philipp Reisner 已提交
2926
	if (self == peer) {
2927
		if (first_peer_device(device)->connection->agreed_pro_version < 96 ?
2928 2929 2930
		    (device->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) ==
		    (device->p_uuid[UI_HISTORY_START] & ~((u64)1)) :
		    self + UUID_NEW_BM_OFFSET == (device->ldev->md.uuid[UI_BITMAP] & ~((u64)1))) {
P
Philipp Reisner 已提交
2931 2932 2933
			/* The last P_SYNC_UUID did not get though. Undo the last start of
			   resync as sync source modifications of our UUIDs. */

2934
			if (first_peer_device(device)->connection->agreed_pro_version < 91)
2935
				return -1091;
P
Philipp Reisner 已提交
2936

2937 2938
			__drbd_uuid_set(device, UI_BITMAP, device->ldev->md.uuid[UI_HISTORY_START]);
			__drbd_uuid_set(device, UI_HISTORY_START, device->ldev->md.uuid[UI_HISTORY_START + 1]);
P
Philipp Reisner 已提交
2939

2940
			drbd_info(device, "Last syncUUID did not get through, corrected:\n");
2941 2942
			drbd_uuid_dump(device, "self", device->ldev->md.uuid,
				       device->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(device) : 0, 0);
P
Philipp Reisner 已提交
2943 2944 2945 2946 2947 2948 2949

			return 1;
		}
	}


	*rule_nr = 80;
2950
	peer = device->p_uuid[UI_CURRENT] & ~((u64)1);
P
Philipp Reisner 已提交
2951
	for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) {
2952
		self = device->ldev->md.uuid[i] & ~((u64)1);
P
Philipp Reisner 已提交
2953 2954 2955 2956 2957
		if (self == peer)
			return 2;
	}

	*rule_nr = 90;
2958 2959
	self = device->ldev->md.uuid[UI_BITMAP] & ~((u64)1);
	peer = device->p_uuid[UI_BITMAP] & ~((u64)1);
P
Philipp Reisner 已提交
2960 2961 2962 2963 2964
	if (self == peer && self != ((u64)0))
		return 100;

	*rule_nr = 100;
	for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) {
2965
		self = device->ldev->md.uuid[i] & ~((u64)1);
P
Philipp Reisner 已提交
2966
		for (j = UI_HISTORY_START; j <= UI_HISTORY_END; j++) {
2967
			peer = device->p_uuid[j] & ~((u64)1);
P
Philipp Reisner 已提交
2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978
			if (self == peer)
				return -100;
		}
	}

	return -1000;
}

/* drbd_sync_handshake() returns the new conn state on success, or
   CONN_MASK (-1) on failure.
 */
2979 2980
static enum drbd_conns drbd_sync_handshake(struct drbd_peer_device *peer_device,
					   enum drbd_role peer_role,
P
Philipp Reisner 已提交
2981 2982
					   enum drbd_disk_state peer_disk) __must_hold(local)
{
2983
	struct drbd_device *device = peer_device->device;
P
Philipp Reisner 已提交
2984 2985
	enum drbd_conns rv = C_MASK;
	enum drbd_disk_state mydisk;
2986
	struct net_conf *nc;
2987
	int hg, rule_nr, rr_conflict, tentative;
P
Philipp Reisner 已提交
2988

2989
	mydisk = device->state.disk;
P
Philipp Reisner 已提交
2990
	if (mydisk == D_NEGOTIATING)
2991
		mydisk = device->new_state_tmp.disk;
P
Philipp Reisner 已提交
2992

2993
	drbd_info(device, "drbd_sync_handshake:\n");
2994

2995 2996 2997 2998
	spin_lock_irq(&device->ldev->md.uuid_lock);
	drbd_uuid_dump(device, "self", device->ldev->md.uuid, device->comm_bm_set, 0);
	drbd_uuid_dump(device, "peer", device->p_uuid,
		       device->p_uuid[UI_SIZE], device->p_uuid[UI_FLAGS]);
P
Philipp Reisner 已提交
2999

3000 3001
	hg = drbd_uuid_compare(device, &rule_nr);
	spin_unlock_irq(&device->ldev->md.uuid_lock);
P
Philipp Reisner 已提交
3002

3003
	drbd_info(device, "uuid_compare()=%d by rule %d\n", hg, rule_nr);
P
Philipp Reisner 已提交
3004 3005

	if (hg == -1000) {
3006
		drbd_alert(device, "Unrelated data, aborting!\n");
P
Philipp Reisner 已提交
3007 3008
		return C_MASK;
	}
3009
	if (hg < -1000) {
3010
		drbd_alert(device, "To resolve this both sides have to support at least protocol %d\n", -hg - 1000);
P
Philipp Reisner 已提交
3011 3012 3013 3014 3015 3016 3017 3018 3019
		return C_MASK;
	}

	if    ((mydisk == D_INCONSISTENT && peer_disk > D_INCONSISTENT) ||
	    (peer_disk == D_INCONSISTENT && mydisk    > D_INCONSISTENT)) {
		int f = (hg == -100) || abs(hg) == 2;
		hg = mydisk > D_INCONSISTENT ? 1 : -1;
		if (f)
			hg = hg*2;
3020
		drbd_info(device, "Becoming sync %s due to disk states.\n",
P
Philipp Reisner 已提交
3021 3022 3023
		     hg > 0 ? "source" : "target");
	}

3024
	if (abs(hg) == 100)
3025
		drbd_khelper(device, "initial-split-brain");
3026

3027
	rcu_read_lock();
3028
	nc = rcu_dereference(peer_device->connection->net_conf);
3029 3030

	if (hg == 100 || (hg == -100 && nc->always_asbp)) {
3031
		int pcount = (device->state.role == R_PRIMARY)
P
Philipp Reisner 已提交
3032 3033 3034 3035 3036
			   + (peer_role == R_PRIMARY);
		int forced = (hg == -100);

		switch (pcount) {
		case 0:
3037
			hg = drbd_asb_recover_0p(peer_device);
P
Philipp Reisner 已提交
3038 3039
			break;
		case 1:
3040
			hg = drbd_asb_recover_1p(peer_device);
P
Philipp Reisner 已提交
3041 3042
			break;
		case 2:
3043
			hg = drbd_asb_recover_2p(peer_device);
P
Philipp Reisner 已提交
3044 3045 3046
			break;
		}
		if (abs(hg) < 100) {
3047
			drbd_warn(device, "Split-Brain detected, %d primaries, "
P
Philipp Reisner 已提交
3048 3049 3050
			     "automatically solved. Sync from %s node\n",
			     pcount, (hg < 0) ? "peer" : "this");
			if (forced) {
3051
				drbd_warn(device, "Doing a full sync, since"
P
Philipp Reisner 已提交
3052 3053 3054 3055 3056 3057 3058
				     " UUIDs where ambiguous.\n");
				hg = hg*2;
			}
		}
	}

	if (hg == -100) {
3059
		if (test_bit(DISCARD_MY_DATA, &device->flags) && !(device->p_uuid[UI_FLAGS]&1))
P
Philipp Reisner 已提交
3060
			hg = -1;
3061
		if (!test_bit(DISCARD_MY_DATA, &device->flags) && (device->p_uuid[UI_FLAGS]&1))
P
Philipp Reisner 已提交
3062 3063 3064
			hg = 1;

		if (abs(hg) < 100)
3065
			drbd_warn(device, "Split-Brain detected, manually solved. "
P
Philipp Reisner 已提交
3066 3067 3068
			     "Sync from %s node\n",
			     (hg < 0) ? "peer" : "this");
	}
3069
	rr_conflict = nc->rr_conflict;
3070
	tentative = nc->tentative;
3071
	rcu_read_unlock();
P
Philipp Reisner 已提交
3072 3073

	if (hg == -100) {
3074 3075 3076 3077
		/* FIXME this log message is not correct if we end up here
		 * after an attempted attach on a diskless node.
		 * We just refuse to attach -- well, we drop the "connection"
		 * to that disk, in a way... */
3078
		drbd_alert(device, "Split-Brain detected but unresolved, dropping connection!\n");
3079
		drbd_khelper(device, "split-brain");
P
Philipp Reisner 已提交
3080 3081 3082 3083
		return C_MASK;
	}

	if (hg > 0 && mydisk <= D_INCONSISTENT) {
3084
		drbd_err(device, "I shall become SyncSource, but I am inconsistent!\n");
P
Philipp Reisner 已提交
3085 3086 3087 3088
		return C_MASK;
	}

	if (hg < 0 && /* by intention we do not use mydisk here. */
3089
	    device->state.role == R_PRIMARY && device->state.disk >= D_CONSISTENT) {
3090
		switch (rr_conflict) {
P
Philipp Reisner 已提交
3091
		case ASB_CALL_HELPER:
3092
			drbd_khelper(device, "pri-lost");
P
Philipp Reisner 已提交
3093 3094
			/* fall through */
		case ASB_DISCONNECT:
3095
			drbd_err(device, "I shall become SyncTarget, but I am primary!\n");
P
Philipp Reisner 已提交
3096 3097
			return C_MASK;
		case ASB_VIOLENTLY:
3098
			drbd_warn(device, "Becoming SyncTarget, violating the stable-data"
P
Philipp Reisner 已提交
3099 3100 3101 3102
			     "assumption\n");
		}
	}

3103
	if (tentative || test_bit(CONN_DRY_RUN, &peer_device->connection->flags)) {
3104
		if (hg == 0)
3105
			drbd_info(device, "dry-run connect: No resync, would become Connected immediately.\n");
3106
		else
3107
			drbd_info(device, "dry-run connect: Would become %s, doing a %s resync.",
3108 3109 3110 3111 3112
				 drbd_conn_str(hg > 0 ? C_SYNC_SOURCE : C_SYNC_TARGET),
				 abs(hg) >= 2 ? "full" : "bit-map based");
		return C_MASK;
	}

P
Philipp Reisner 已提交
3113
	if (abs(hg) >= 2) {
3114
		drbd_info(device, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n");
3115
		if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, "set_n_write from sync_handshake",
3116
					BM_LOCKED_SET_ALLOWED))
P
Philipp Reisner 已提交
3117 3118 3119 3120 3121 3122 3123 3124 3125
			return C_MASK;
	}

	if (hg > 0) { /* become sync source. */
		rv = C_WF_BITMAP_S;
	} else if (hg < 0) { /* become sync target */
		rv = C_WF_BITMAP_T;
	} else {
		rv = C_CONNECTED;
3126
		if (drbd_bm_total_weight(device)) {
3127
			drbd_info(device, "No resync, but %lu bits in bitmap!\n",
3128
			     drbd_bm_total_weight(device));
P
Philipp Reisner 已提交
3129 3130 3131 3132 3133 3134
		}
	}

	return rv;
}

3135
static enum drbd_after_sb_p convert_after_sb(enum drbd_after_sb_p peer)
P
Philipp Reisner 已提交
3136 3137
{
	/* ASB_DISCARD_REMOTE - ASB_DISCARD_LOCAL is valid */
3138 3139
	if (peer == ASB_DISCARD_REMOTE)
		return ASB_DISCARD_LOCAL;
P
Philipp Reisner 已提交
3140 3141

	/* any other things with ASB_DISCARD_REMOTE or ASB_DISCARD_LOCAL are invalid */
3142 3143
	if (peer == ASB_DISCARD_LOCAL)
		return ASB_DISCARD_REMOTE;
P
Philipp Reisner 已提交
3144 3145

	/* everything else is valid if they are equal on both sides. */
3146
	return peer;
P
Philipp Reisner 已提交
3147 3148
}

3149
static int receive_protocol(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3150
{
3151
	struct p_protocol *p = pi->data;
3152 3153 3154 3155
	enum drbd_after_sb_p p_after_sb_0p, p_after_sb_1p, p_after_sb_2p;
	int p_proto, p_discard_my_data, p_two_primaries, cf;
	struct net_conf *nc, *old_net_conf, *new_net_conf = NULL;
	char integrity_alg[SHARED_SECRET_MAX] = "";
3156
	struct crypto_hash *peer_integrity_tfm = NULL;
3157
	void *int_dig_in = NULL, *int_dig_vv = NULL;
P
Philipp Reisner 已提交
3158 3159 3160 3161 3162 3163

	p_proto		= be32_to_cpu(p->protocol);
	p_after_sb_0p	= be32_to_cpu(p->after_sb_0p);
	p_after_sb_1p	= be32_to_cpu(p->after_sb_1p);
	p_after_sb_2p	= be32_to_cpu(p->after_sb_2p);
	p_two_primaries = be32_to_cpu(p->two_primaries);
3164
	cf		= be32_to_cpu(p->conn_flags);
3165
	p_discard_my_data = cf & CF_DISCARD_MY_DATA;
3166

3167
	if (connection->agreed_pro_version >= 87) {
3168
		int err;
3169

3170
		if (pi->size > sizeof(integrity_alg))
3171
			return -EIO;
3172
		err = drbd_recv_all(connection, integrity_alg, pi->size);
3173 3174
		if (err)
			return err;
3175
		integrity_alg[SHARED_SECRET_MAX - 1] = 0;
P
Philipp Reisner 已提交
3176 3177
	}

3178
	if (pi->cmd != P_PROTOCOL_UPDATE) {
3179
		clear_bit(CONN_DRY_RUN, &connection->flags);
P
Philipp Reisner 已提交
3180

3181
		if (cf & CF_DRY_RUN)
3182
			set_bit(CONN_DRY_RUN, &connection->flags);
P
Philipp Reisner 已提交
3183

3184
		rcu_read_lock();
3185
		nc = rcu_dereference(connection->net_conf);
P
Philipp Reisner 已提交
3186

3187
		if (p_proto != nc->wire_protocol) {
3188
			drbd_err(connection, "incompatible %s settings\n", "protocol");
3189 3190
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3191

3192
		if (convert_after_sb(p_after_sb_0p) != nc->after_sb_0p) {
3193
			drbd_err(connection, "incompatible %s settings\n", "after-sb-0pri");
3194 3195
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3196

3197
		if (convert_after_sb(p_after_sb_1p) != nc->after_sb_1p) {
3198
			drbd_err(connection, "incompatible %s settings\n", "after-sb-1pri");
3199 3200
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3201

3202
		if (convert_after_sb(p_after_sb_2p) != nc->after_sb_2p) {
3203
			drbd_err(connection, "incompatible %s settings\n", "after-sb-2pri");
3204 3205
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3206

3207
		if (p_discard_my_data && nc->discard_my_data) {
3208
			drbd_err(connection, "incompatible %s settings\n", "discard-my-data");
3209 3210
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3211

3212
		if (p_two_primaries != nc->two_primaries) {
3213
			drbd_err(connection, "incompatible %s settings\n", "allow-two-primaries");
3214 3215
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3216

3217
		if (strcmp(integrity_alg, nc->integrity_alg)) {
3218
			drbd_err(connection, "incompatible %s settings\n", "data-integrity-alg");
3219 3220
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3221

3222
		rcu_read_unlock();
P
Philipp Reisner 已提交
3223 3224
	}

3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235
	if (integrity_alg[0]) {
		int hash_size;

		/*
		 * We can only change the peer data integrity algorithm
		 * here.  Changing our own data integrity algorithm
		 * requires that we send a P_PROTOCOL_UPDATE packet at
		 * the same time; otherwise, the peer has no way to
		 * tell between which packets the algorithm should
		 * change.
		 */
P
Philipp Reisner 已提交
3236

3237 3238
		peer_integrity_tfm = crypto_alloc_hash(integrity_alg, 0, CRYPTO_ALG_ASYNC);
		if (!peer_integrity_tfm) {
3239
			drbd_err(connection, "peer data-integrity-alg %s not supported\n",
3240 3241 3242
				 integrity_alg);
			goto disconnect;
		}
P
Philipp Reisner 已提交
3243

3244 3245 3246 3247
		hash_size = crypto_hash_digestsize(peer_integrity_tfm);
		int_dig_in = kmalloc(hash_size, GFP_KERNEL);
		int_dig_vv = kmalloc(hash_size, GFP_KERNEL);
		if (!(int_dig_in && int_dig_vv)) {
3248
			drbd_err(connection, "Allocation of buffers for data integrity checking failed\n");
P
Philipp Reisner 已提交
3249 3250 3251 3252
			goto disconnect;
		}
	}

3253 3254
	new_net_conf = kmalloc(sizeof(struct net_conf), GFP_KERNEL);
	if (!new_net_conf) {
3255
		drbd_err(connection, "Allocation of new net_conf failed\n");
3256 3257 3258
		goto disconnect;
	}

3259
	mutex_lock(&connection->data.mutex);
3260
	mutex_lock(&connection->resource->conf_update);
3261
	old_net_conf = connection->net_conf;
3262 3263 3264 3265 3266 3267 3268 3269
	*new_net_conf = *old_net_conf;

	new_net_conf->wire_protocol = p_proto;
	new_net_conf->after_sb_0p = convert_after_sb(p_after_sb_0p);
	new_net_conf->after_sb_1p = convert_after_sb(p_after_sb_1p);
	new_net_conf->after_sb_2p = convert_after_sb(p_after_sb_2p);
	new_net_conf->two_primaries = p_two_primaries;

3270
	rcu_assign_pointer(connection->net_conf, new_net_conf);
3271
	mutex_unlock(&connection->resource->conf_update);
3272
	mutex_unlock(&connection->data.mutex);
3273

3274 3275 3276 3277 3278 3279
	crypto_free_hash(connection->peer_integrity_tfm);
	kfree(connection->int_dig_in);
	kfree(connection->int_dig_vv);
	connection->peer_integrity_tfm = peer_integrity_tfm;
	connection->int_dig_in = int_dig_in;
	connection->int_dig_vv = int_dig_vv;
3280 3281

	if (strcmp(old_net_conf->integrity_alg, integrity_alg))
3282
		drbd_info(connection, "peer data-integrity-alg: %s\n",
3283 3284 3285 3286
			  integrity_alg[0] ? integrity_alg : "(none)");

	synchronize_rcu();
	kfree(old_net_conf);
3287
	return 0;
P
Philipp Reisner 已提交
3288

3289 3290
disconnect_rcu_unlock:
	rcu_read_unlock();
P
Philipp Reisner 已提交
3291
disconnect:
3292
	crypto_free_hash(peer_integrity_tfm);
3293 3294
	kfree(int_dig_in);
	kfree(int_dig_vv);
3295
	conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD);
3296
	return -EIO;
P
Philipp Reisner 已提交
3297 3298 3299 3300 3301 3302 3303
}

/* helper function
 * input: alg name, feature name
 * return: NULL (alg name was "")
 *         ERR_PTR(error) if something goes wrong
 *         or the crypto hash ptr, if it worked out ok. */
3304
static
3305
struct crypto_hash *drbd_crypto_alloc_digest_safe(const struct drbd_device *device,
P
Philipp Reisner 已提交
3306 3307 3308 3309 3310 3311 3312 3313 3314
		const char *alg, const char *name)
{
	struct crypto_hash *tfm;

	if (!alg[0])
		return NULL;

	tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
	if (IS_ERR(tfm)) {
3315
		drbd_err(device, "Can not allocate \"%s\" as %s (reason: %ld)\n",
P
Philipp Reisner 已提交
3316 3317 3318 3319 3320 3321
			alg, name, PTR_ERR(tfm));
		return tfm;
	}
	return tfm;
}

3322
static int ignore_remaining_packet(struct drbd_connection *connection, struct packet_info *pi)
3323
{
3324
	void *buffer = connection->data.rbuf;
3325 3326 3327 3328
	int size = pi->size;

	while (size) {
		int s = min_t(int, size, DRBD_SOCKET_BUFFER_SIZE);
3329
		s = drbd_recv(connection, buffer, s);
3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352
		if (s <= 0) {
			if (s < 0)
				return s;
			break;
		}
		size -= s;
	}
	if (size)
		return -EIO;
	return 0;
}

/*
 * config_unknown_volume  -  device configuration command for unknown volume
 *
 * When a device is added to an existing connection, the node on which the
 * device is added first will send configuration commands to its peer but the
 * peer will not know about the device yet.  It will warn and ignore these
 * commands.  Once the device is added on the second node, the second node will
 * send the same device configuration commands, but in the other direction.
 *
 * (We can also end up here if drbd is misconfigured.)
 */
3353
static int config_unknown_volume(struct drbd_connection *connection, struct packet_info *pi)
3354
{
3355
	drbd_warn(connection, "%s packet received for volume %u, which is not configured locally\n",
3356
		  cmdname(pi->cmd), pi->vnr);
3357
	return ignore_remaining_packet(connection, pi);
3358 3359
}

3360
static int receive_SyncParam(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3361
{
3362
	struct drbd_peer_device *peer_device;
3363
	struct drbd_device *device;
3364
	struct p_rs_param_95 *p;
P
Philipp Reisner 已提交
3365 3366 3367
	unsigned int header_size, data_size, exp_max_sz;
	struct crypto_hash *verify_tfm = NULL;
	struct crypto_hash *csums_tfm = NULL;
3368
	struct net_conf *old_net_conf, *new_net_conf = NULL;
P
Philipp Reisner 已提交
3369
	struct disk_conf *old_disk_conf = NULL, *new_disk_conf = NULL;
3370
	const int apv = connection->agreed_pro_version;
P
Philipp Reisner 已提交
3371
	struct fifo_buffer *old_plan = NULL, *new_plan = NULL;
3372
	int fifo_size = 0;
3373
	int err;
P
Philipp Reisner 已提交
3374

3375 3376
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
3377
		return config_unknown_volume(connection, pi);
3378
	device = peer_device->device;
P
Philipp Reisner 已提交
3379 3380 3381 3382

	exp_max_sz  = apv <= 87 ? sizeof(struct p_rs_param)
		    : apv == 88 ? sizeof(struct p_rs_param)
					+ SHARED_SECRET_MAX
3383 3384
		    : apv <= 94 ? sizeof(struct p_rs_param_89)
		    : /* apv >= 95 */ sizeof(struct p_rs_param_95);
P
Philipp Reisner 已提交
3385

3386
	if (pi->size > exp_max_sz) {
3387
		drbd_err(device, "SyncParam packet too long: received %u, expected <= %u bytes\n",
3388
		    pi->size, exp_max_sz);
3389
		return -EIO;
P
Philipp Reisner 已提交
3390 3391 3392
	}

	if (apv <= 88) {
3393
		header_size = sizeof(struct p_rs_param);
3394
		data_size = pi->size - header_size;
3395
	} else if (apv <= 94) {
3396
		header_size = sizeof(struct p_rs_param_89);
3397
		data_size = pi->size - header_size;
3398
		D_ASSERT(device, data_size == 0);
3399
	} else {
3400
		header_size = sizeof(struct p_rs_param_95);
3401
		data_size = pi->size - header_size;
3402
		D_ASSERT(device, data_size == 0);
P
Philipp Reisner 已提交
3403 3404 3405
	}

	/* initialize verify_alg and csums_alg */
3406
	p = pi->data;
P
Philipp Reisner 已提交
3407 3408
	memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX);

3409
	err = drbd_recv_all(peer_device->connection, p, header_size);
3410 3411
	if (err)
		return err;
P
Philipp Reisner 已提交
3412

3413
	mutex_lock(&connection->resource->conf_update);
3414
	old_net_conf = peer_device->connection->net_conf;
3415
	if (get_ldev(device)) {
P
Philipp Reisner 已提交
3416 3417
		new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL);
		if (!new_disk_conf) {
3418
			put_ldev(device);
3419
			mutex_unlock(&connection->resource->conf_update);
3420
			drbd_err(device, "Allocation of new disk_conf failed\n");
P
Philipp Reisner 已提交
3421 3422
			return -ENOMEM;
		}
P
Philipp Reisner 已提交
3423

3424
		old_disk_conf = device->ldev->disk_conf;
P
Philipp Reisner 已提交
3425
		*new_disk_conf = *old_disk_conf;
P
Philipp Reisner 已提交
3426

3427
		new_disk_conf->resync_rate = be32_to_cpu(p->resync_rate);
P
Philipp Reisner 已提交
3428
	}
P
Philipp Reisner 已提交
3429 3430 3431

	if (apv >= 88) {
		if (apv == 88) {
3432
			if (data_size > SHARED_SECRET_MAX || data_size == 0) {
3433
				drbd_err(device, "verify-alg of wrong size, "
3434 3435
					"peer wants %u, accepting only up to %u byte\n",
					data_size, SHARED_SECRET_MAX);
P
Philipp Reisner 已提交
3436 3437
				err = -EIO;
				goto reconnect;
P
Philipp Reisner 已提交
3438 3439
			}

3440
			err = drbd_recv_all(peer_device->connection, p->verify_alg, data_size);
P
Philipp Reisner 已提交
3441 3442
			if (err)
				goto reconnect;
P
Philipp Reisner 已提交
3443 3444
			/* we expect NUL terminated string */
			/* but just in case someone tries to be evil */
3445
			D_ASSERT(device, p->verify_alg[data_size-1] == 0);
P
Philipp Reisner 已提交
3446 3447 3448 3449 3450
			p->verify_alg[data_size-1] = 0;

		} else /* apv >= 89 */ {
			/* we still expect NUL terminated strings */
			/* but just in case someone tries to be evil */
3451 3452
			D_ASSERT(device, p->verify_alg[SHARED_SECRET_MAX-1] == 0);
			D_ASSERT(device, p->csums_alg[SHARED_SECRET_MAX-1] == 0);
P
Philipp Reisner 已提交
3453 3454 3455 3456
			p->verify_alg[SHARED_SECRET_MAX-1] = 0;
			p->csums_alg[SHARED_SECRET_MAX-1] = 0;
		}

3457
		if (strcmp(old_net_conf->verify_alg, p->verify_alg)) {
3458
			if (device->state.conn == C_WF_REPORT_PARAMS) {
3459
				drbd_err(device, "Different verify-alg settings. me=\"%s\" peer=\"%s\"\n",
3460
				    old_net_conf->verify_alg, p->verify_alg);
P
Philipp Reisner 已提交
3461 3462
				goto disconnect;
			}
3463
			verify_tfm = drbd_crypto_alloc_digest_safe(device,
P
Philipp Reisner 已提交
3464 3465 3466 3467 3468 3469 3470
					p->verify_alg, "verify-alg");
			if (IS_ERR(verify_tfm)) {
				verify_tfm = NULL;
				goto disconnect;
			}
		}

3471
		if (apv >= 89 && strcmp(old_net_conf->csums_alg, p->csums_alg)) {
3472
			if (device->state.conn == C_WF_REPORT_PARAMS) {
3473
				drbd_err(device, "Different csums-alg settings. me=\"%s\" peer=\"%s\"\n",
3474
				    old_net_conf->csums_alg, p->csums_alg);
P
Philipp Reisner 已提交
3475 3476
				goto disconnect;
			}
3477
			csums_tfm = drbd_crypto_alloc_digest_safe(device,
P
Philipp Reisner 已提交
3478 3479 3480 3481 3482 3483 3484
					p->csums_alg, "csums-alg");
			if (IS_ERR(csums_tfm)) {
				csums_tfm = NULL;
				goto disconnect;
			}
		}

P
Philipp Reisner 已提交
3485
		if (apv > 94 && new_disk_conf) {
P
Philipp Reisner 已提交
3486 3487 3488 3489
			new_disk_conf->c_plan_ahead = be32_to_cpu(p->c_plan_ahead);
			new_disk_conf->c_delay_target = be32_to_cpu(p->c_delay_target);
			new_disk_conf->c_fill_target = be32_to_cpu(p->c_fill_target);
			new_disk_conf->c_max_rate = be32_to_cpu(p->c_max_rate);
3490

P
Philipp Reisner 已提交
3491
			fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ;
3492
			if (fifo_size != device->rs_plan_s->size) {
P
Philipp Reisner 已提交
3493 3494
				new_plan = fifo_alloc(fifo_size);
				if (!new_plan) {
3495
					drbd_err(device, "kmalloc of fifo_buffer failed");
3496
					put_ldev(device);
3497 3498 3499
					goto disconnect;
				}
			}
3500
		}
P
Philipp Reisner 已提交
3501

3502
		if (verify_tfm || csums_tfm) {
3503 3504
			new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
			if (!new_net_conf) {
3505
				drbd_err(device, "Allocation of new net_conf failed\n");
3506 3507 3508
				goto disconnect;
			}

3509
			*new_net_conf = *old_net_conf;
3510 3511

			if (verify_tfm) {
3512 3513
				strcpy(new_net_conf->verify_alg, p->verify_alg);
				new_net_conf->verify_alg_len = strlen(p->verify_alg) + 1;
3514 3515
				crypto_free_hash(peer_device->connection->verify_tfm);
				peer_device->connection->verify_tfm = verify_tfm;
3516
				drbd_info(device, "using verify-alg: \"%s\"\n", p->verify_alg);
3517 3518
			}
			if (csums_tfm) {
3519 3520
				strcpy(new_net_conf->csums_alg, p->csums_alg);
				new_net_conf->csums_alg_len = strlen(p->csums_alg) + 1;
3521 3522
				crypto_free_hash(peer_device->connection->csums_tfm);
				peer_device->connection->csums_tfm = csums_tfm;
3523
				drbd_info(device, "using csums-alg: \"%s\"\n", p->csums_alg);
3524
			}
3525
			rcu_assign_pointer(connection->net_conf, new_net_conf);
3526
		}
P
Philipp Reisner 已提交
3527 3528
	}

P
Philipp Reisner 已提交
3529
	if (new_disk_conf) {
3530 3531
		rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf);
		put_ldev(device);
P
Philipp Reisner 已提交
3532 3533 3534
	}

	if (new_plan) {
3535 3536
		old_plan = device->rs_plan_s;
		rcu_assign_pointer(device->rs_plan_s, new_plan);
P
Philipp Reisner 已提交
3537
	}
P
Philipp Reisner 已提交
3538

3539
	mutex_unlock(&connection->resource->conf_update);
P
Philipp Reisner 已提交
3540 3541 3542 3543
	synchronize_rcu();
	if (new_net_conf)
		kfree(old_net_conf);
	kfree(old_disk_conf);
P
Philipp Reisner 已提交
3544
	kfree(old_plan);
P
Philipp Reisner 已提交
3545

3546
	return 0;
P
Philipp Reisner 已提交
3547

P
Philipp Reisner 已提交
3548 3549
reconnect:
	if (new_disk_conf) {
3550
		put_ldev(device);
P
Philipp Reisner 已提交
3551 3552
		kfree(new_disk_conf);
	}
3553
	mutex_unlock(&connection->resource->conf_update);
P
Philipp Reisner 已提交
3554 3555
	return -EIO;

P
Philipp Reisner 已提交
3556
disconnect:
P
Philipp Reisner 已提交
3557 3558
	kfree(new_plan);
	if (new_disk_conf) {
3559
		put_ldev(device);
P
Philipp Reisner 已提交
3560 3561
		kfree(new_disk_conf);
	}
3562
	mutex_unlock(&connection->resource->conf_update);
P
Philipp Reisner 已提交
3563 3564 3565 3566 3567
	/* just for completeness: actually not needed,
	 * as this is not reached if csums_tfm was ok. */
	crypto_free_hash(csums_tfm);
	/* but free the verify_tfm again, if csums_tfm did not work out */
	crypto_free_hash(verify_tfm);
3568
	conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD);
3569
	return -EIO;
P
Philipp Reisner 已提交
3570 3571 3572
}

/* warn if the arguments differ by more than 12.5% */
3573
static void warn_if_differ_considerably(struct drbd_device *device,
P
Philipp Reisner 已提交
3574 3575 3576 3577 3578 3579 3580
	const char *s, sector_t a, sector_t b)
{
	sector_t d;
	if (a == 0 || b == 0)
		return;
	d = (a > b) ? (a - b) : (b - a);
	if (d > (a>>3) || d > (b>>3))
3581
		drbd_warn(device, "Considerable difference in %s: %llus vs. %llus\n", s,
P
Philipp Reisner 已提交
3582 3583 3584
		     (unsigned long long)a, (unsigned long long)b);
}

3585
static int receive_sizes(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3586
{
3587
	struct drbd_peer_device *peer_device;
3588
	struct drbd_device *device;
3589
	struct p_sizes *p = pi->data;
3590
	enum determine_dev_size dd = DS_UNCHANGED;
P
Philipp Reisner 已提交
3591 3592
	sector_t p_size, p_usize, my_usize;
	int ldsc = 0; /* local disk size changed */
3593
	enum dds_flags ddsf;
P
Philipp Reisner 已提交
3594

3595 3596
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
3597
		return config_unknown_volume(connection, pi);
3598
	device = peer_device->device;
3599

P
Philipp Reisner 已提交
3600 3601 3602 3603 3604
	p_size = be64_to_cpu(p->d_size);
	p_usize = be64_to_cpu(p->u_size);

	/* just store the peer's disk size for now.
	 * we still need to figure out whether we accept that. */
3605
	device->p_size = p_size;
P
Philipp Reisner 已提交
3606

3607
	if (get_ldev(device)) {
P
Philipp Reisner 已提交
3608
		rcu_read_lock();
3609
		my_usize = rcu_dereference(device->ldev->disk_conf)->disk_size;
P
Philipp Reisner 已提交
3610 3611
		rcu_read_unlock();

3612 3613 3614
		warn_if_differ_considerably(device, "lower level device sizes",
			   p_size, drbd_get_max_capacity(device->ldev));
		warn_if_differ_considerably(device, "user requested size",
P
Philipp Reisner 已提交
3615
					    p_usize, my_usize);
P
Philipp Reisner 已提交
3616 3617 3618

		/* if this is the first connect, or an otherwise expected
		 * param exchange, choose the minimum */
3619
		if (device->state.conn == C_WF_REPORT_PARAMS)
P
Philipp Reisner 已提交
3620
			p_usize = min_not_zero(my_usize, p_usize);
P
Philipp Reisner 已提交
3621 3622 3623

		/* Never shrink a device with usable data during connect.
		   But allow online shrinking if we are connected. */
3624 3625 3626 3627
		if (drbd_new_dev_size(device, device->ldev, p_usize, 0) <
		    drbd_get_capacity(device->this_bdev) &&
		    device->state.disk >= D_OUTDATED &&
		    device->state.conn < C_CONNECTED) {
3628
			drbd_err(device, "The peer's disk size is too small!\n");
3629
			conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD);
3630
			put_ldev(device);
3631
			return -EIO;
P
Philipp Reisner 已提交
3632
		}
P
Philipp Reisner 已提交
3633 3634 3635 3636 3637 3638

		if (my_usize != p_usize) {
			struct disk_conf *old_disk_conf, *new_disk_conf = NULL;

			new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL);
			if (!new_disk_conf) {
3639
				drbd_err(device, "Allocation of new disk_conf failed\n");
3640
				put_ldev(device);
P
Philipp Reisner 已提交
3641 3642 3643
				return -ENOMEM;
			}

3644
			mutex_lock(&connection->resource->conf_update);
3645
			old_disk_conf = device->ldev->disk_conf;
P
Philipp Reisner 已提交
3646 3647 3648
			*new_disk_conf = *old_disk_conf;
			new_disk_conf->disk_size = p_usize;

3649
			rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf);
3650
			mutex_unlock(&connection->resource->conf_update);
P
Philipp Reisner 已提交
3651 3652 3653
			synchronize_rcu();
			kfree(old_disk_conf);

3654
			drbd_info(device, "Peer sets u_size to %lu sectors\n",
P
Philipp Reisner 已提交
3655
				 (unsigned long)my_usize);
P
Philipp Reisner 已提交
3656
		}
P
Philipp Reisner 已提交
3657

3658
		put_ldev(device);
P
Philipp Reisner 已提交
3659 3660
	}

3661
	ddsf = be16_to_cpu(p->dds_flags);
3662 3663 3664
	if (get_ldev(device)) {
		dd = drbd_determine_dev_size(device, ddsf, NULL);
		put_ldev(device);
3665
		if (dd == DS_ERROR)
3666
			return -EIO;
3667
		drbd_md_sync(device);
P
Philipp Reisner 已提交
3668 3669
	} else {
		/* I am diskless, need to accept the peer's size. */
3670
		drbd_set_my_capacity(device, p_size);
P
Philipp Reisner 已提交
3671 3672
	}

3673 3674
	device->peer_max_bio_size = be32_to_cpu(p->max_bio_size);
	drbd_reconsider_max_bio_size(device);
3675

3676 3677 3678
	if (get_ldev(device)) {
		if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) {
			device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev);
P
Philipp Reisner 已提交
3679 3680 3681
			ldsc = 1;
		}

3682
		put_ldev(device);
P
Philipp Reisner 已提交
3683 3684
	}

3685
	if (device->state.conn > C_WF_REPORT_PARAMS) {
P
Philipp Reisner 已提交
3686
		if (be64_to_cpu(p->c_size) !=
3687
		    drbd_get_capacity(device->this_bdev) || ldsc) {
P
Philipp Reisner 已提交
3688 3689
			/* we have different sizes, probably peer
			 * needs to know my new size... */
3690
			drbd_send_sizes(peer_device, 0, ddsf);
P
Philipp Reisner 已提交
3691
		}
3692 3693 3694 3695
		if (test_and_clear_bit(RESIZE_PENDING, &device->flags) ||
		    (dd == DS_GREW && device->state.conn == C_CONNECTED)) {
			if (device->state.pdsk >= D_INCONSISTENT &&
			    device->state.disk >= D_INCONSISTENT) {
3696
				if (ddsf & DDSF_NO_RESYNC)
3697
					drbd_info(device, "Resync of new storage suppressed with --assume-clean\n");
3698
				else
3699
					resync_after_online_grow(device);
3700
			} else
3701
				set_bit(RESYNC_AFTER_NEG, &device->flags);
P
Philipp Reisner 已提交
3702 3703 3704
		}
	}

3705
	return 0;
P
Philipp Reisner 已提交
3706 3707
}

3708
static int receive_uuids(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3709
{
3710
	struct drbd_peer_device *peer_device;
3711
	struct drbd_device *device;
3712
	struct p_uuids *p = pi->data;
P
Philipp Reisner 已提交
3713
	u64 *p_uuid;
3714
	int i, updated_uuids = 0;
P
Philipp Reisner 已提交
3715

3716 3717
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
3718
		return config_unknown_volume(connection, pi);
3719
	device = peer_device->device;
3720

P
Philipp Reisner 已提交
3721
	p_uuid = kmalloc(sizeof(u64)*UI_EXTENDED_SIZE, GFP_NOIO);
3722
	if (!p_uuid) {
3723
		drbd_err(device, "kmalloc of p_uuid failed\n");
3724 3725
		return false;
	}
P
Philipp Reisner 已提交
3726 3727 3728 3729

	for (i = UI_CURRENT; i < UI_EXTENDED_SIZE; i++)
		p_uuid[i] = be64_to_cpu(p->uuid[i]);

3730 3731
	kfree(device->p_uuid);
	device->p_uuid = p_uuid;
P
Philipp Reisner 已提交
3732

3733 3734 3735 3736
	if (device->state.conn < C_CONNECTED &&
	    device->state.disk < D_INCONSISTENT &&
	    device->state.role == R_PRIMARY &&
	    (device->ed_uuid & ~((u64)1)) != (p_uuid[UI_CURRENT] & ~((u64)1))) {
3737
		drbd_err(device, "Can only connect to data with current UUID=%016llX\n",
3738
		    (unsigned long long)device->ed_uuid);
3739
		conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD);
3740
		return -EIO;
P
Philipp Reisner 已提交
3741 3742
	}

3743
	if (get_ldev(device)) {
P
Philipp Reisner 已提交
3744
		int skip_initial_sync =
3745
			device->state.conn == C_CONNECTED &&
3746
			peer_device->connection->agreed_pro_version >= 90 &&
3747
			device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED &&
P
Philipp Reisner 已提交
3748 3749
			(p_uuid[UI_FLAGS] & 8);
		if (skip_initial_sync) {
3750
			drbd_info(device, "Accepted new current UUID, preparing to skip initial sync\n");
3751
			drbd_bitmap_io(device, &drbd_bmio_clear_n_write,
3752 3753
					"clear_n_write from receive_uuids",
					BM_LOCKED_TEST_ALLOWED);
3754 3755 3756
			_drbd_uuid_set(device, UI_CURRENT, p_uuid[UI_CURRENT]);
			_drbd_uuid_set(device, UI_BITMAP, 0);
			_drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE),
P
Philipp Reisner 已提交
3757
					CS_VERBOSE, NULL);
3758
			drbd_md_sync(device);
3759
			updated_uuids = 1;
P
Philipp Reisner 已提交
3760
		}
3761 3762 3763
		put_ldev(device);
	} else if (device->state.disk < D_INCONSISTENT &&
		   device->state.role == R_PRIMARY) {
3764 3765
		/* I am a diskless primary, the peer just created a new current UUID
		   for me. */
3766
		updated_uuids = drbd_set_ed_uuid(device, p_uuid[UI_CURRENT]);
P
Philipp Reisner 已提交
3767 3768 3769 3770 3771 3772
	}

	/* Before we test for the disk state, we should wait until an eventually
	   ongoing cluster wide state change is finished. That is important if
	   we are primary and are detaching from our disk. We need to see the
	   new disk state... */
3773 3774 3775 3776
	mutex_lock(device->state_mutex);
	mutex_unlock(device->state_mutex);
	if (device->state.conn >= C_CONNECTED && device->state.disk < D_INCONSISTENT)
		updated_uuids |= drbd_set_ed_uuid(device, p_uuid[UI_CURRENT]);
3777 3778

	if (updated_uuids)
3779
		drbd_print_uuids(device, "receiver updated UUIDs to");
P
Philipp Reisner 已提交
3780

3781
	return 0;
P
Philipp Reisner 已提交
3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792
}

/**
 * convert_state() - Converts the peer's view of the cluster state to our point of view
 * @ps:		The state as seen by the peer.
 */
static union drbd_state convert_state(union drbd_state ps)
{
	union drbd_state ms;

	static enum drbd_conns c_tab[] = {
3793
		[C_WF_REPORT_PARAMS] = C_WF_REPORT_PARAMS,
P
Philipp Reisner 已提交
3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
		[C_CONNECTED] = C_CONNECTED,

		[C_STARTING_SYNC_S] = C_STARTING_SYNC_T,
		[C_STARTING_SYNC_T] = C_STARTING_SYNC_S,
		[C_DISCONNECTING] = C_TEAR_DOWN, /* C_NETWORK_FAILURE, */
		[C_VERIFY_S]       = C_VERIFY_T,
		[C_MASK]   = C_MASK,
	};

	ms.i = ps.i;

	ms.conn = c_tab[ps.conn];
	ms.peer = ps.role;
	ms.role = ps.peer;
	ms.pdsk = ps.disk;
	ms.disk = ps.pdsk;
	ms.peer_isp = (ps.aftr_isp | ps.user_isp);

	return ms;
}

3815
static int receive_req_state(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3816
{
3817
	struct drbd_peer_device *peer_device;
3818
	struct drbd_device *device;
3819
	struct p_req_state *p = pi->data;
P
Philipp Reisner 已提交
3820
	union drbd_state mask, val;
3821
	enum drbd_state_rv rv;
P
Philipp Reisner 已提交
3822

3823 3824
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
3825
		return -EIO;
3826
	device = peer_device->device;
3827

P
Philipp Reisner 已提交
3828 3829 3830
	mask.i = be32_to_cpu(p->mask);
	val.i = be32_to_cpu(p->val);

3831
	if (test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags) &&
3832
	    mutex_is_locked(device->state_mutex)) {
3833
		drbd_send_sr_reply(peer_device, SS_CONCURRENT_ST_CHG);
3834
		return 0;
P
Philipp Reisner 已提交
3835 3836 3837 3838 3839
	}

	mask = convert_state(mask);
	val = convert_state(val);

3840
	rv = drbd_change_state(device, CS_VERBOSE, mask, val);
3841
	drbd_send_sr_reply(peer_device, rv);
P
Philipp Reisner 已提交
3842

3843
	drbd_md_sync(device);
P
Philipp Reisner 已提交
3844

3845
	return 0;
P
Philipp Reisner 已提交
3846 3847
}

3848
static int receive_req_conn_state(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3849
{
3850
	struct p_req_state *p = pi->data;
P
Philipp Reisner 已提交
3851
	union drbd_state mask, val;
3852
	enum drbd_state_rv rv;
P
Philipp Reisner 已提交
3853 3854 3855 3856

	mask.i = be32_to_cpu(p->mask);
	val.i = be32_to_cpu(p->val);

3857 3858 3859
	if (test_bit(RESOLVE_CONFLICTS, &connection->flags) &&
	    mutex_is_locked(&connection->cstate_mutex)) {
		conn_send_sr_reply(connection, SS_CONCURRENT_ST_CHG);
3860
		return 0;
P
Philipp Reisner 已提交
3861 3862 3863 3864 3865
	}

	mask = convert_state(mask);
	val = convert_state(val);

3866 3867
	rv = conn_request_state(connection, mask, val, CS_VERBOSE | CS_LOCAL_ONLY | CS_IGN_OUTD_FAIL);
	conn_send_sr_reply(connection, rv);
P
Philipp Reisner 已提交
3868

3869
	return 0;
P
Philipp Reisner 已提交
3870 3871
}

3872
static int receive_state(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
3873
{
3874
	struct drbd_peer_device *peer_device;
3875
	struct drbd_device *device;
3876
	struct p_state *p = pi->data;
3877
	union drbd_state os, ns, peer_state;
P
Philipp Reisner 已提交
3878
	enum drbd_disk_state real_peer_disk;
3879
	enum chg_state_flags cs_flags;
P
Philipp Reisner 已提交
3880 3881
	int rv;

3882 3883
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
3884
		return config_unknown_volume(connection, pi);
3885
	device = peer_device->device;
3886

P
Philipp Reisner 已提交
3887 3888 3889 3890
	peer_state.i = be32_to_cpu(p->state);

	real_peer_disk = peer_state.disk;
	if (peer_state.disk == D_NEGOTIATING) {
3891
		real_peer_disk = device->p_uuid[UI_FLAGS] & 4 ? D_INCONSISTENT : D_CONSISTENT;
3892
		drbd_info(device, "real peer disk state = %s\n", drbd_disk_str(real_peer_disk));
P
Philipp Reisner 已提交
3893 3894
	}

3895
	spin_lock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
3896
 retry:
3897
	os = ns = drbd_read_state(device);
3898
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
3899

3900 3901 3902 3903
	/* If some other part of the code (asender thread, timeout)
	 * already decided to close the connection again,
	 * we must not "re-establish" it here. */
	if (os.conn <= C_TEAR_DOWN)
3904
		return -ECONNRESET;
3905

3906 3907 3908 3909 3910 3911 3912 3913
	/* If this is the "end of sync" confirmation, usually the peer disk
	 * transitions from D_INCONSISTENT to D_UP_TO_DATE. For empty (0 bits
	 * set) resync started in PausedSyncT, or if the timing of pause-/
	 * unpause-sync events has been "just right", the peer disk may
	 * transition from D_CONSISTENT to D_UP_TO_DATE as well.
	 */
	if ((os.pdsk == D_INCONSISTENT || os.pdsk == D_CONSISTENT) &&
	    real_peer_disk == D_UP_TO_DATE &&
3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929
	    os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) {
		/* If we are (becoming) SyncSource, but peer is still in sync
		 * preparation, ignore its uptodate-ness to avoid flapping, it
		 * will change to inconsistent once the peer reaches active
		 * syncing states.
		 * It may have changed syncer-paused flags, however, so we
		 * cannot ignore this completely. */
		if (peer_state.conn > C_CONNECTED &&
		    peer_state.conn < C_SYNC_SOURCE)
			real_peer_disk = D_INCONSISTENT;

		/* if peer_state changes to connected at the same time,
		 * it explicitly notifies us that it finished resync.
		 * Maybe we should finish it up, too? */
		else if (os.conn >= C_SYNC_SOURCE &&
			 peer_state.conn == C_CONNECTED) {
3930 3931
			if (drbd_bm_total_weight(device) <= device->rs_failed)
				drbd_resync_finished(device);
3932
			return 0;
3933 3934 3935
		}
	}

3936 3937 3938
	/* explicit verify finished notification, stop sector reached. */
	if (os.conn == C_VERIFY_T && os.disk == D_UP_TO_DATE &&
	    peer_state.conn == C_CONNECTED && real_peer_disk == D_UP_TO_DATE) {
3939 3940
		ov_out_of_sync_print(device);
		drbd_resync_finished(device);
3941
		return 0;
3942 3943
	}

3944 3945 3946 3947 3948 3949 3950 3951 3952
	/* peer says his disk is inconsistent, while we think it is uptodate,
	 * and this happens while the peer still thinks we have a sync going on,
	 * but we think we are already done with the sync.
	 * We ignore this to avoid flapping pdsk.
	 * This should not happen, if the peer is a recent version of drbd. */
	if (os.pdsk == D_UP_TO_DATE && real_peer_disk == D_INCONSISTENT &&
	    os.conn == C_CONNECTED && peer_state.conn > C_SYNC_SOURCE)
		real_peer_disk = D_UP_TO_DATE;

3953 3954
	if (ns.conn == C_WF_REPORT_PARAMS)
		ns.conn = C_CONNECTED;
P
Philipp Reisner 已提交
3955

3956 3957 3958
	if (peer_state.conn == C_AHEAD)
		ns.conn = C_BEHIND;

3959 3960
	if (device->p_uuid && peer_state.disk >= D_NEGOTIATING &&
	    get_ldev_if_state(device, D_NEGOTIATING)) {
P
Philipp Reisner 已提交
3961 3962 3963
		int cr; /* consider resync */

		/* if we established a new connection */
3964
		cr  = (os.conn < C_CONNECTED);
P
Philipp Reisner 已提交
3965 3966
		/* if we had an established connection
		 * and one of the nodes newly attaches a disk */
3967
		cr |= (os.conn == C_CONNECTED &&
P
Philipp Reisner 已提交
3968
		       (peer_state.disk == D_NEGOTIATING ||
3969
			os.disk == D_NEGOTIATING));
P
Philipp Reisner 已提交
3970 3971
		/* if we have both been inconsistent, and the peer has been
		 * forced to be UpToDate with --overwrite-data */
3972
		cr |= test_bit(CONSIDER_RESYNC, &device->flags);
P
Philipp Reisner 已提交
3973 3974
		/* if we had been plain connected, and the admin requested to
		 * start a sync by "invalidate" or "invalidate-remote" */
3975
		cr |= (os.conn == C_CONNECTED &&
P
Philipp Reisner 已提交
3976 3977 3978 3979
				(peer_state.conn >= C_STARTING_SYNC_S &&
				 peer_state.conn <= C_WF_BITMAP_T));

		if (cr)
3980
			ns.conn = drbd_sync_handshake(peer_device, peer_state.role, real_peer_disk);
P
Philipp Reisner 已提交
3981

3982
		put_ldev(device);
3983 3984
		if (ns.conn == C_MASK) {
			ns.conn = C_CONNECTED;
3985 3986
			if (device->state.disk == D_NEGOTIATING) {
				drbd_force_state(device, NS(disk, D_FAILED));
P
Philipp Reisner 已提交
3987
			} else if (peer_state.disk == D_NEGOTIATING) {
3988
				drbd_err(device, "Disk attach process on the peer node was aborted.\n");
P
Philipp Reisner 已提交
3989
				peer_state.disk = D_DISKLESS;
3990
				real_peer_disk = D_DISKLESS;
P
Philipp Reisner 已提交
3991
			} else {
3992
				if (test_and_clear_bit(CONN_DRY_RUN, &peer_device->connection->flags))
3993
					return -EIO;
3994
				D_ASSERT(device, os.conn == C_WF_REPORT_PARAMS);
3995
				conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD);
3996
				return -EIO;
P
Philipp Reisner 已提交
3997 3998 3999 4000
			}
		}
	}

4001
	spin_lock_irq(&device->resource->req_lock);
4002
	if (os.i != drbd_read_state(device).i)
P
Philipp Reisner 已提交
4003
		goto retry;
4004
	clear_bit(CONSIDER_RESYNC, &device->flags);
P
Philipp Reisner 已提交
4005 4006 4007
	ns.peer = peer_state.role;
	ns.pdsk = real_peer_disk;
	ns.peer_isp = (peer_state.aftr_isp | peer_state.user_isp);
4008
	if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING)
4009
		ns.disk = device->new_state_tmp.disk;
4010
	cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD);
4011 4012
	if (ns.pdsk == D_CONSISTENT && drbd_suspended(device) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED &&
	    test_bit(NEW_CUR_UUID, &device->flags)) {
4013
		/* Do not allow tl_restart(RESEND) for a rebooted peer. We can only allow this
4014
		   for temporal network outages! */
4015
		spin_unlock_irq(&device->resource->req_lock);
4016
		drbd_err(device, "Aborting Connect, can not thaw IO with an only Consistent peer\n");
4017
		tl_clear(peer_device->connection);
4018 4019
		drbd_uuid_new_current(device);
		clear_bit(NEW_CUR_UUID, &device->flags);
4020
		conn_request_state(peer_device->connection, NS2(conn, C_PROTOCOL_ERROR, susp, 0), CS_HARD);
4021
		return -EIO;
4022
	}
4023 4024
	rv = _drbd_set_state(device, ns, cs_flags, NULL);
	ns = drbd_read_state(device);
4025
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
4026 4027

	if (rv < SS_SUCCESS) {
4028
		conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD);
4029
		return -EIO;
P
Philipp Reisner 已提交
4030 4031
	}

4032 4033
	if (os.conn > C_WF_REPORT_PARAMS) {
		if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED &&
P
Philipp Reisner 已提交
4034 4035 4036 4037
		    peer_state.disk != D_NEGOTIATING ) {
			/* we want resync, peer has not yet decided to sync... */
			/* Nowadays only used when forcing a node into primary role and
			   setting its disk to UpToDate with that */
4038 4039
			drbd_send_uuids(peer_device);
			drbd_send_current_state(peer_device);
P
Philipp Reisner 已提交
4040 4041 4042
		}
	}

4043
	clear_bit(DISCARD_MY_DATA, &device->flags);
P
Philipp Reisner 已提交
4044

4045
	drbd_md_sync(device); /* update connected indicator, la_size_sect, ... */
P
Philipp Reisner 已提交
4046

4047
	return 0;
P
Philipp Reisner 已提交
4048 4049
}

4050
static int receive_sync_uuid(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4051
{
4052
	struct drbd_peer_device *peer_device;
4053
	struct drbd_device *device;
4054
	struct p_rs_uuid *p = pi->data;
4055

4056 4057
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
4058
		return -EIO;
4059
	device = peer_device->device;
P
Philipp Reisner 已提交
4060

4061 4062 4063 4064 4065
	wait_event(device->misc_wait,
		   device->state.conn == C_WF_SYNC_UUID ||
		   device->state.conn == C_BEHIND ||
		   device->state.conn < C_CONNECTED ||
		   device->state.disk < D_NEGOTIATING);
P
Philipp Reisner 已提交
4066

4067
	/* D_ASSERT(device,  device->state.conn == C_WF_SYNC_UUID ); */
P
Philipp Reisner 已提交
4068 4069 4070

	/* Here the _drbd_uuid_ functions are right, current should
	   _not_ be rotated into the history */
4071 4072 4073
	if (get_ldev_if_state(device, D_NEGOTIATING)) {
		_drbd_uuid_set(device, UI_CURRENT, be64_to_cpu(p->uuid));
		_drbd_uuid_set(device, UI_BITMAP, 0UL);
P
Philipp Reisner 已提交
4074

4075 4076
		drbd_print_uuids(device, "updated sync uuid");
		drbd_start_resync(device, C_SYNC_TARGET);
P
Philipp Reisner 已提交
4077

4078
		put_ldev(device);
P
Philipp Reisner 已提交
4079
	} else
4080
		drbd_err(device, "Ignoring SyncUUID packet!\n");
P
Philipp Reisner 已提交
4081

4082
	return 0;
P
Philipp Reisner 已提交
4083 4084
}

4085 4086 4087 4088 4089 4090 4091
/**
 * receive_bitmap_plain
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
4092
receive_bitmap_plain(struct drbd_peer_device *peer_device, unsigned int size,
4093
		     unsigned long *p, struct bm_xfer_ctx *c)
P
Philipp Reisner 已提交
4094
{
4095
	unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE -
4096
				 drbd_header_size(peer_device->connection);
4097
	unsigned int num_words = min_t(size_t, data_size / sizeof(*p),
4098
				       c->bm_words - c->word_offset);
4099
	unsigned int want = num_words * sizeof(*p);
4100
	int err;
P
Philipp Reisner 已提交
4101

4102
	if (want != size) {
4103
		drbd_err(peer_device, "%s:want (%u) != size (%u)\n", __func__, want, size);
4104
		return -EIO;
P
Philipp Reisner 已提交
4105 4106
	}
	if (want == 0)
4107
		return 0;
4108
	err = drbd_recv_all(peer_device->connection, p, want);
4109
	if (err)
4110
		return err;
P
Philipp Reisner 已提交
4111

4112
	drbd_bm_merge_lel(peer_device->device, c->word_offset, num_words, p);
P
Philipp Reisner 已提交
4113 4114 4115 4116 4117 4118

	c->word_offset += num_words;
	c->bit_offset = c->word_offset * BITS_PER_LONG;
	if (c->bit_offset > c->bm_bits)
		c->bit_offset = c->bm_bits;

4119
	return 1;
P
Philipp Reisner 已提交
4120 4121
}

4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136
static enum drbd_bitmap_code dcbp_get_code(struct p_compressed_bm *p)
{
	return (enum drbd_bitmap_code)(p->encoding & 0x0f);
}

static int dcbp_get_start(struct p_compressed_bm *p)
{
	return (p->encoding & 0x80) != 0;
}

static int dcbp_get_pad_bits(struct p_compressed_bm *p)
{
	return (p->encoding >> 4) & 0x7;
}

4137 4138 4139 4140 4141 4142 4143
/**
 * recv_bm_rle_bits
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
4144
recv_bm_rle_bits(struct drbd_peer_device *peer_device,
P
Philipp Reisner 已提交
4145
		struct p_compressed_bm *p,
4146 4147
		 struct bm_xfer_ctx *c,
		 unsigned int len)
P
Philipp Reisner 已提交
4148 4149 4150 4151 4152 4153 4154
{
	struct bitstream bs;
	u64 look_ahead;
	u64 rl;
	u64 tmp;
	unsigned long s = c->bit_offset;
	unsigned long e;
4155
	int toggle = dcbp_get_start(p);
P
Philipp Reisner 已提交
4156 4157 4158
	int have;
	int bits;

4159
	bitstream_init(&bs, p->code, len, dcbp_get_pad_bits(p));
P
Philipp Reisner 已提交
4160 4161 4162

	bits = bitstream_get_bits(&bs, &look_ahead, 64);
	if (bits < 0)
4163
		return -EIO;
P
Philipp Reisner 已提交
4164 4165 4166 4167

	for (have = bits; have > 0; s += rl, toggle = !toggle) {
		bits = vli_decode_bits(&rl, look_ahead);
		if (bits <= 0)
4168
			return -EIO;
P
Philipp Reisner 已提交
4169 4170 4171 4172

		if (toggle) {
			e = s + rl -1;
			if (e >= c->bm_bits) {
4173
				drbd_err(peer_device, "bitmap overflow (e:%lu) while decoding bm RLE packet\n", e);
4174
				return -EIO;
P
Philipp Reisner 已提交
4175
			}
4176
			_drbd_bm_set_bits(peer_device->device, s, e);
P
Philipp Reisner 已提交
4177 4178 4179
		}

		if (have < bits) {
4180
			drbd_err(peer_device, "bitmap decoding error: h:%d b:%d la:0x%08llx l:%u/%u\n",
P
Philipp Reisner 已提交
4181 4182 4183
				have, bits, look_ahead,
				(unsigned int)(bs.cur.b - p->code),
				(unsigned int)bs.buf_len);
4184
			return -EIO;
P
Philipp Reisner 已提交
4185
		}
4186 4187 4188 4189 4190
		/* if we consumed all 64 bits, assign 0; >> 64 is "undefined"; */
		if (likely(bits < 64))
			look_ahead >>= bits;
		else
			look_ahead = 0;
P
Philipp Reisner 已提交
4191 4192 4193 4194
		have -= bits;

		bits = bitstream_get_bits(&bs, &tmp, 64 - have);
		if (bits < 0)
4195
			return -EIO;
P
Philipp Reisner 已提交
4196 4197 4198 4199 4200 4201 4202
		look_ahead |= tmp << have;
		have += bits;
	}

	c->bit_offset = s;
	bm_xfer_ctx_bit_to_word_offset(c);

4203
	return (s != c->bm_bits);
P
Philipp Reisner 已提交
4204 4205
}

4206 4207 4208 4209 4210 4211 4212
/**
 * decode_bitmap_c
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
4213
decode_bitmap_c(struct drbd_peer_device *peer_device,
P
Philipp Reisner 已提交
4214
		struct p_compressed_bm *p,
4215 4216
		struct bm_xfer_ctx *c,
		unsigned int len)
P
Philipp Reisner 已提交
4217
{
4218
	if (dcbp_get_code(p) == RLE_VLI_Bits)
4219
		return recv_bm_rle_bits(peer_device, p, c, len - sizeof(*p));
P
Philipp Reisner 已提交
4220 4221 4222 4223 4224

	/* other variants had been implemented for evaluation,
	 * but have been dropped as this one turned out to be "best"
	 * during all our tests. */

4225 4226
	drbd_err(peer_device, "receive_bitmap_c: unknown encoding %u\n", p->encoding);
	conn_request_state(peer_device->connection, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
4227
	return -EIO;
P
Philipp Reisner 已提交
4228 4229
}

4230
void INFO_bm_xfer_stats(struct drbd_device *device,
P
Philipp Reisner 已提交
4231 4232 4233
		const char *direction, struct bm_xfer_ctx *c)
{
	/* what would it take to transfer it "plaintext" */
4234
	unsigned int header_size = drbd_header_size(first_peer_device(device)->connection);
4235 4236 4237 4238 4239 4240
	unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE - header_size;
	unsigned int plain =
		header_size * (DIV_ROUND_UP(c->bm_words, data_size) + 1) +
		c->bm_words * sizeof(unsigned long);
	unsigned int total = c->bytes[0] + c->bytes[1];
	unsigned int r;
P
Philipp Reisner 已提交
4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257

	/* total can not be zero. but just in case: */
	if (total == 0)
		return;

	/* don't report if not compressed */
	if (total >= plain)
		return;

	/* total < plain. check for overflow, still */
	r = (total > UINT_MAX/1000) ? (total / (plain/1000))
		                    : (1000 * total / plain);

	if (r > 1000)
		r = 1000;

	r = 1000 - r;
4258
	drbd_info(device, "%s bitmap stats [Bytes(packets)]: plain %u(%u), RLE %u(%u), "
P
Philipp Reisner 已提交
4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273
	     "total %u; compression: %u.%u%%\n",
			direction,
			c->bytes[1], c->packets[1],
			c->bytes[0], c->packets[0],
			total, r/10, r % 10);
}

/* Since we are processing the bitfield from lower addresses to higher,
   it does not matter if the process it in 32 bit chunks or 64 bit
   chunks as long as it is little endian. (Understand it as byte stream,
   beginning with the lowest byte...) If we would use big endian
   we would need to process it from the highest address to the lowest,
   in order to be agnostic to the 32 vs 64 bits issue.

   returns 0 on failure, 1 if we successfully received it. */
4274
static int receive_bitmap(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4275
{
4276
	struct drbd_peer_device *peer_device;
4277
	struct drbd_device *device;
P
Philipp Reisner 已提交
4278
	struct bm_xfer_ctx c;
4279
	int err;
4280

4281 4282
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
4283
		return -EIO;
4284
	device = peer_device->device;
P
Philipp Reisner 已提交
4285

4286
	drbd_bm_lock(device, "receive bitmap", BM_LOCKED_SET_ALLOWED);
4287 4288
	/* you are supposed to send additional out-of-sync information
	 * if you actually set bits during this phase */
P
Philipp Reisner 已提交
4289 4290

	c = (struct bm_xfer_ctx) {
4291 4292
		.bm_bits = drbd_bm_bits(device),
		.bm_words = drbd_bm_words(device),
P
Philipp Reisner 已提交
4293 4294
	};

4295
	for(;;) {
4296
		if (pi->cmd == P_BITMAP)
4297
			err = receive_bitmap_plain(peer_device, pi->size, pi->data, &c);
4298
		else if (pi->cmd == P_COMPRESSED_BITMAP) {
P
Philipp Reisner 已提交
4299 4300
			/* MAYBE: sanity check that we speak proto >= 90,
			 * and the feature is enabled! */
4301
			struct p_compressed_bm *p = pi->data;
P
Philipp Reisner 已提交
4302

4303
			if (pi->size > DRBD_SOCKET_BUFFER_SIZE - drbd_header_size(connection)) {
4304
				drbd_err(device, "ReportCBitmap packet too large\n");
4305
				err = -EIO;
P
Philipp Reisner 已提交
4306 4307
				goto out;
			}
4308
			if (pi->size <= sizeof(*p)) {
4309
				drbd_err(device, "ReportCBitmap packet too small (l:%u)\n", pi->size);
4310
				err = -EIO;
4311
				goto out;
P
Philipp Reisner 已提交
4312
			}
4313
			err = drbd_recv_all(peer_device->connection, p, pi->size);
4314 4315
			if (err)
			       goto out;
4316
			err = decode_bitmap_c(peer_device, p, &c, pi->size);
P
Philipp Reisner 已提交
4317
		} else {
4318
			drbd_warn(device, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", pi->cmd);
4319
			err = -EIO;
P
Philipp Reisner 已提交
4320 4321 4322
			goto out;
		}

4323
		c.packets[pi->cmd == P_BITMAP]++;
4324
		c.bytes[pi->cmd == P_BITMAP] += drbd_header_size(connection) + pi->size;
P
Philipp Reisner 已提交
4325

4326 4327 4328
		if (err <= 0) {
			if (err < 0)
				goto out;
P
Philipp Reisner 已提交
4329
			break;
4330
		}
4331
		err = drbd_recv_header(peer_device->connection, pi);
4332
		if (err)
P
Philipp Reisner 已提交
4333
			goto out;
4334
	}
P
Philipp Reisner 已提交
4335

4336
	INFO_bm_xfer_stats(device, "receive", &c);
P
Philipp Reisner 已提交
4337

4338
	if (device->state.conn == C_WF_BITMAP_T) {
4339 4340
		enum drbd_state_rv rv;

4341
		err = drbd_send_bitmap(device);
4342
		if (err)
P
Philipp Reisner 已提交
4343 4344
			goto out;
		/* Omit CS_ORDERED with this state transition to avoid deadlocks. */
4345
		rv = _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE);
4346
		D_ASSERT(device, rv == SS_SUCCESS);
4347
	} else if (device->state.conn != C_WF_BITMAP_S) {
P
Philipp Reisner 已提交
4348 4349
		/* admin may have requested C_DISCONNECTING,
		 * other threads may have noticed network errors */
4350
		drbd_info(device, "unexpected cstate (%s) in receive_bitmap\n",
4351
		    drbd_conn_str(device->state.conn));
P
Philipp Reisner 已提交
4352
	}
4353
	err = 0;
P
Philipp Reisner 已提交
4354 4355

 out:
4356 4357 4358
	drbd_bm_unlock(device);
	if (!err && device->state.conn == C_WF_BITMAP_S)
		drbd_start_resync(device, C_SYNC_SOURCE);
4359
	return err;
P
Philipp Reisner 已提交
4360 4361
}

4362
static int receive_skip(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4363
{
4364
	drbd_warn(connection, "skipping unknown optional packet type %d, l: %d!\n",
4365
		 pi->cmd, pi->size);
P
Philipp Reisner 已提交
4366

4367
	return ignore_remaining_packet(connection, pi);
P
Philipp Reisner 已提交
4368 4369
}

4370
static int receive_UnplugRemote(struct drbd_connection *connection, struct packet_info *pi)
4371
{
4372 4373
	/* Make sure we've acked all the TCP data associated
	 * with the data requests being unplugged */
4374
	drbd_tcp_quickack(connection->data.socket);
4375

4376
	return 0;
4377 4378
}

4379
static int receive_out_of_sync(struct drbd_connection *connection, struct packet_info *pi)
4380
{
4381
	struct drbd_peer_device *peer_device;
4382
	struct drbd_device *device;
4383
	struct p_block_desc *p = pi->data;
4384

4385 4386
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
4387
		return -EIO;
4388
	device = peer_device->device;
4389

4390
	switch (device->state.conn) {
4391 4392 4393 4394 4395
	case C_WF_SYNC_UUID:
	case C_WF_BITMAP_T:
	case C_BEHIND:
			break;
	default:
4396
		drbd_err(device, "ASSERT FAILED cstate = %s, expected: WFSyncUUID|WFBitMapT|Behind\n",
4397
				drbd_conn_str(device->state.conn));
4398 4399
	}

4400
	drbd_set_out_of_sync(device, be64_to_cpu(p->sector), be32_to_cpu(p->blksize));
4401

4402
	return 0;
4403 4404
}

4405 4406 4407
struct data_cmd {
	int expect_payload;
	size_t pkt_size;
4408
	int (*fn)(struct drbd_connection *, struct packet_info *);
4409 4410 4411 4412 4413 4414 4415
};

static struct data_cmd drbd_cmd_handler[] = {
	[P_DATA]	    = { 1, sizeof(struct p_data), receive_Data },
	[P_DATA_REPLY]	    = { 1, sizeof(struct p_data), receive_DataReply },
	[P_RS_DATA_REPLY]   = { 1, sizeof(struct p_data), receive_RSDataReply } ,
	[P_BARRIER]	    = { 0, sizeof(struct p_barrier), receive_Barrier } ,
4416 4417 4418
	[P_BITMAP]	    = { 1, 0, receive_bitmap } ,
	[P_COMPRESSED_BITMAP] = { 1, 0, receive_bitmap } ,
	[P_UNPLUG_REMOTE]   = { 0, 0, receive_UnplugRemote },
4419 4420
	[P_DATA_REQUEST]    = { 0, sizeof(struct p_block_req), receive_DataRequest },
	[P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest },
4421 4422
	[P_SYNC_PARAM]	    = { 1, 0, receive_SyncParam },
	[P_SYNC_PARAM89]    = { 1, 0, receive_SyncParam },
4423 4424 4425 4426 4427 4428 4429 4430 4431 4432
	[P_PROTOCOL]        = { 1, sizeof(struct p_protocol), receive_protocol },
	[P_UUIDS]	    = { 0, sizeof(struct p_uuids), receive_uuids },
	[P_SIZES]	    = { 0, sizeof(struct p_sizes), receive_sizes },
	[P_STATE]	    = { 0, sizeof(struct p_state), receive_state },
	[P_STATE_CHG_REQ]   = { 0, sizeof(struct p_req_state), receive_req_state },
	[P_SYNC_UUID]       = { 0, sizeof(struct p_rs_uuid), receive_sync_uuid },
	[P_OV_REQUEST]      = { 0, sizeof(struct p_block_req), receive_DataRequest },
	[P_OV_REPLY]        = { 1, sizeof(struct p_block_req), receive_DataRequest },
	[P_CSUM_RS_REQUEST] = { 1, sizeof(struct p_block_req), receive_DataRequest },
	[P_DELAY_PROBE]     = { 0, sizeof(struct p_delay_probe93), receive_skip },
4433
	[P_OUT_OF_SYNC]     = { 0, sizeof(struct p_block_desc), receive_out_of_sync },
4434
	[P_CONN_ST_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_conn_state },
4435
	[P_PROTOCOL_UPDATE] = { 1, sizeof(struct p_protocol), receive_protocol },
P
Philipp Reisner 已提交
4436 4437
};

4438
static void drbdd(struct drbd_connection *connection)
P
Philipp Reisner 已提交
4439
{
4440
	struct packet_info pi;
4441
	size_t shs; /* sub header size */
4442
	int err;
P
Philipp Reisner 已提交
4443

4444
	while (get_t_state(&connection->receiver) == RUNNING) {
4445
		struct data_cmd *cmd;
P
Philipp Reisner 已提交
4446

4447 4448
		drbd_thread_current_set_cpu(&connection->receiver);
		if (drbd_recv_header(connection, &pi))
4449
			goto err_out;
P
Philipp Reisner 已提交
4450

4451
		cmd = &drbd_cmd_handler[pi.cmd];
4452
		if (unlikely(pi.cmd >= ARRAY_SIZE(drbd_cmd_handler) || !cmd->fn)) {
4453
			drbd_err(connection, "Unexpected data packet %s (0x%04x)",
4454
				 cmdname(pi.cmd), pi.cmd);
4455
			goto err_out;
4456
		}
P
Philipp Reisner 已提交
4457

4458 4459
		shs = cmd->pkt_size;
		if (pi.size > shs && !cmd->expect_payload) {
4460
			drbd_err(connection, "No payload expected %s l:%d\n",
4461
				 cmdname(pi.cmd), pi.size);
4462
			goto err_out;
P
Philipp Reisner 已提交
4463 4464
		}

4465
		if (shs) {
4466
			err = drbd_recv_all_warn(connection, pi.data, shs);
4467
			if (err)
4468
				goto err_out;
4469
			pi.size -= shs;
4470 4471
		}

4472
		err = cmd->fn(connection, &pi);
4473
		if (err) {
4474
			drbd_err(connection, "error receiving %s, e: %d l: %d!\n",
4475
				 cmdname(pi.cmd), err, pi.size);
4476
			goto err_out;
P
Philipp Reisner 已提交
4477 4478
		}
	}
4479
	return;
P
Philipp Reisner 已提交
4480

4481
    err_out:
4482
	conn_request_state(connection, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
P
Philipp Reisner 已提交
4483 4484
}

4485
static void conn_disconnect(struct drbd_connection *connection)
P
Philipp Reisner 已提交
4486
{
4487
	struct drbd_peer_device *peer_device;
4488
	enum drbd_conns oc;
P
Philipp Reisner 已提交
4489
	int vnr;
P
Philipp Reisner 已提交
4490

4491
	if (connection->cstate == C_STANDALONE)
P
Philipp Reisner 已提交
4492 4493
		return;

4494 4495 4496 4497 4498
	/* We are about to start the cleanup after connection loss.
	 * Make sure drbd_make_request knows about that.
	 * Usually we should be in some network failure state already,
	 * but just in case we are not, we fix it up here.
	 */
4499
	conn_request_state(connection, NS(conn, C_NETWORK_FAILURE), CS_HARD);
4500

P
Philipp Reisner 已提交
4501
	/* asender does not clean up anything. it must not interfere, either */
4502 4503
	drbd_thread_stop(&connection->asender);
	drbd_free_sock(connection);
4504

P
Philipp Reisner 已提交
4505
	rcu_read_lock();
4506 4507
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
		struct drbd_device *device = peer_device->device;
4508
		kref_get(&device->kref);
P
Philipp Reisner 已提交
4509
		rcu_read_unlock();
4510
		drbd_disconnected(peer_device);
4511
		kref_put(&device->kref, drbd_destroy_device);
P
Philipp Reisner 已提交
4512 4513 4514 4515
		rcu_read_lock();
	}
	rcu_read_unlock();

4516
	if (!list_empty(&connection->current_epoch->list))
4517
		drbd_err(connection, "ASSERTION FAILED: connection->current_epoch->list not empty\n");
4518
	/* ok, no more ee's on the fly, it is safe to reset the epoch_size */
4519 4520
	atomic_set(&connection->current_epoch->epoch_size, 0);
	connection->send.seen_any_write_yet = false;
4521

4522
	drbd_info(connection, "Connection closed\n");
4523

4524 4525
	if (conn_highest_role(connection) == R_PRIMARY && conn_highest_pdsk(connection) >= D_UNKNOWN)
		conn_try_outdate_peer_async(connection);
4526

4527
	spin_lock_irq(&connection->resource->req_lock);
4528
	oc = connection->cstate;
4529
	if (oc >= C_UNCONNECTED)
4530
		_conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE);
4531

4532
	spin_unlock_irq(&connection->resource->req_lock);
4533

4534
	if (oc == C_DISCONNECTING)
4535
		conn_request_state(connection, NS(conn, C_STANDALONE), CS_VERBOSE | CS_HARD);
4536 4537
}

4538
static int drbd_disconnected(struct drbd_peer_device *peer_device)
4539
{
4540
	struct drbd_device *device = peer_device->device;
4541
	unsigned int i;
P
Philipp Reisner 已提交
4542

4543
	/* wait for current activity to cease. */
4544
	spin_lock_irq(&device->resource->req_lock);
4545 4546 4547
	_drbd_wait_ee_list_empty(device, &device->active_ee);
	_drbd_wait_ee_list_empty(device, &device->sync_ee);
	_drbd_wait_ee_list_empty(device, &device->read_ee);
4548
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559

	/* We do not have data structures that would allow us to
	 * get the rs_pending_cnt down to 0 again.
	 *  * On C_SYNC_TARGET we do not have any data structures describing
	 *    the pending RSDataRequest's we have sent.
	 *  * On C_SYNC_SOURCE there is no data structure that tracks
	 *    the P_RS_DATA_REPLY blocks that we sent to the SyncTarget.
	 *  And no, it is not the sum of the reference counts in the
	 *  resync_LRU. The resync_LRU tracks the whole operation including
	 *  the disk-IO, while the rs_pending_cnt only tracks the blocks
	 *  on the fly. */
4560 4561 4562 4563 4564
	drbd_rs_cancel_all(device);
	device->rs_total = 0;
	device->rs_failed = 0;
	atomic_set(&device->rs_pending_cnt, 0);
	wake_up(&device->misc_wait);
P
Philipp Reisner 已提交
4565

4566 4567
	del_timer_sync(&device->resync_timer);
	resync_timer_fn((unsigned long)device);
P
Philipp Reisner 已提交
4568 4569 4570 4571

	/* wait for all w_e_end_data_req, w_e_end_rsdata_req, w_send_barrier,
	 * w_make_resync_request etc. which may still be on the worker queue
	 * to be "canceled" */
4572
	drbd_flush_workqueue(&peer_device->connection->sender_work);
P
Philipp Reisner 已提交
4573

4574
	drbd_finish_peer_reqs(device);
P
Philipp Reisner 已提交
4575

4576 4577 4578
	/* This second workqueue flush is necessary, since drbd_finish_peer_reqs()
	   might have issued a work again. The one before drbd_finish_peer_reqs() is
	   necessary to reclain net_ee in drbd_finish_peer_reqs(). */
4579
	drbd_flush_workqueue(&peer_device->connection->sender_work);
4580

4581 4582
	/* need to do it again, drbd_finish_peer_reqs() may have populated it
	 * again via drbd_try_clear_on_disk_bm(). */
4583
	drbd_rs_cancel_all(device);
P
Philipp Reisner 已提交
4584

4585 4586
	kfree(device->p_uuid);
	device->p_uuid = NULL;
P
Philipp Reisner 已提交
4587

4588
	if (!drbd_suspended(device))
4589
		tl_clear(peer_device->connection);
P
Philipp Reisner 已提交
4590

4591
	drbd_md_sync(device);
P
Philipp Reisner 已提交
4592

4593 4594
	/* serialize with bitmap writeout triggered by the state change,
	 * if any. */
4595
	wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags));
4596

P
Philipp Reisner 已提交
4597 4598 4599 4600 4601 4602 4603
	/* tcp_close and release of sendpage pages can be deferred.  I don't
	 * want to use SO_LINGER, because apparently it can be deferred for
	 * more than 20 seconds (longest time I checked).
	 *
	 * Actually we don't care for exactly when the network stack does its
	 * put_page(), but release our reference on these pages right here.
	 */
4604
	i = drbd_free_peer_reqs(device, &device->net_ee);
P
Philipp Reisner 已提交
4605
	if (i)
4606
		drbd_info(device, "net_ee not empty, killed %u entries\n", i);
4607
	i = atomic_read(&device->pp_in_use_by_net);
4608
	if (i)
4609
		drbd_info(device, "pp_in_use_by_net = %d, expected 0\n", i);
4610
	i = atomic_read(&device->pp_in_use);
P
Philipp Reisner 已提交
4611
	if (i)
4612
		drbd_info(device, "pp_in_use = %d, expected 0\n", i);
P
Philipp Reisner 已提交
4613

4614 4615 4616 4617
	D_ASSERT(device, list_empty(&device->read_ee));
	D_ASSERT(device, list_empty(&device->active_ee));
	D_ASSERT(device, list_empty(&device->sync_ee));
	D_ASSERT(device, list_empty(&device->done_ee));
P
Philipp Reisner 已提交
4618

4619
	return 0;
P
Philipp Reisner 已提交
4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630
}

/*
 * We support PRO_VERSION_MIN to PRO_VERSION_MAX. The protocol version
 * we can agree on is stored in agreed_pro_version.
 *
 * feature flags and the reserved array should be enough room for future
 * enhancements of the handshake protocol, and possible plugins...
 *
 * for now, they are expected to be zero, but ignored.
 */
4631
static int drbd_send_features(struct drbd_connection *connection)
P
Philipp Reisner 已提交
4632
{
4633 4634
	struct drbd_socket *sock;
	struct p_connection_features *p;
P
Philipp Reisner 已提交
4635

4636 4637
	sock = &connection->data;
	p = conn_prepare_command(connection, sock);
4638
	if (!p)
4639
		return -EIO;
P
Philipp Reisner 已提交
4640 4641 4642
	memset(p, 0, sizeof(*p));
	p->protocol_min = cpu_to_be32(PRO_VERSION_MIN);
	p->protocol_max = cpu_to_be32(PRO_VERSION_MAX);
4643
	return conn_send_command(connection, sock, P_CONNECTION_FEATURES, sizeof(*p), NULL, 0);
P
Philipp Reisner 已提交
4644 4645 4646 4647 4648 4649 4650 4651 4652
}

/*
 * return values:
 *   1 yes, we have a valid connection
 *   0 oops, did not work out, please try again
 *  -1 peer talks different language,
 *     no point in trying again, please go standalone.
 */
4653
static int drbd_do_features(struct drbd_connection *connection)
P
Philipp Reisner 已提交
4654
{
4655
	/* ASSERT current == connection->receiver ... */
4656 4657
	struct p_connection_features *p;
	const int expect = sizeof(struct p_connection_features);
4658
	struct packet_info pi;
4659
	int err;
P
Philipp Reisner 已提交
4660

4661
	err = drbd_send_features(connection);
4662
	if (err)
P
Philipp Reisner 已提交
4663 4664
		return 0;

4665
	err = drbd_recv_header(connection, &pi);
4666
	if (err)
P
Philipp Reisner 已提交
4667 4668
		return 0;

4669
	if (pi.cmd != P_CONNECTION_FEATURES) {
4670
		drbd_err(connection, "expected ConnectionFeatures packet, received: %s (0x%04x)\n",
4671
			 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4672 4673 4674
		return -1;
	}

4675
	if (pi.size != expect) {
4676
		drbd_err(connection, "expected ConnectionFeatures length: %u, received: %u\n",
4677
		     expect, pi.size);
P
Philipp Reisner 已提交
4678 4679 4680
		return -1;
	}

4681
	p = pi.data;
4682
	err = drbd_recv_all_warn(connection, p, expect);
4683
	if (err)
P
Philipp Reisner 已提交
4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694
		return 0;

	p->protocol_min = be32_to_cpu(p->protocol_min);
	p->protocol_max = be32_to_cpu(p->protocol_max);
	if (p->protocol_max == 0)
		p->protocol_max = p->protocol_min;

	if (PRO_VERSION_MAX < p->protocol_min ||
	    PRO_VERSION_MIN > p->protocol_max)
		goto incompat;

4695
	connection->agreed_pro_version = min_t(int, PRO_VERSION_MAX, p->protocol_max);
P
Philipp Reisner 已提交
4696

4697
	drbd_info(connection, "Handshake successful: "
4698
	     "Agreed network protocol version %d\n", connection->agreed_pro_version);
P
Philipp Reisner 已提交
4699 4700 4701 4702

	return 1;

 incompat:
4703
	drbd_err(connection, "incompatible DRBD dialects: "
P
Philipp Reisner 已提交
4704 4705 4706 4707 4708 4709 4710
	    "I support %d-%d, peer supports %d-%d\n",
	    PRO_VERSION_MIN, PRO_VERSION_MAX,
	    p->protocol_min, p->protocol_max);
	return -1;
}

#if !defined(CONFIG_CRYPTO_HMAC) && !defined(CONFIG_CRYPTO_HMAC_MODULE)
4711
static int drbd_do_auth(struct drbd_connection *connection)
P
Philipp Reisner 已提交
4712
{
4713 4714
	drbd_err(connection, "This kernel was build without CONFIG_CRYPTO_HMAC.\n");
	drbd_err(connection, "You need to disable 'cram-hmac-alg' in drbd.conf.\n");
4715
	return -1;
P
Philipp Reisner 已提交
4716 4717 4718
}
#else
#define CHALLENGE_LEN 64
4719 4720 4721 4722 4723 4724 4725

/* Return value:
	1 - auth succeeded,
	0 - failed, try again (network error),
	-1 - auth failed, don't try again.
*/

4726
static int drbd_do_auth(struct drbd_connection *connection)
P
Philipp Reisner 已提交
4727
{
4728
	struct drbd_socket *sock;
P
Philipp Reisner 已提交
4729 4730 4731 4732 4733
	char my_challenge[CHALLENGE_LEN];  /* 64 Bytes... */
	struct scatterlist sg;
	char *response = NULL;
	char *right_response = NULL;
	char *peers_ch = NULL;
4734 4735
	unsigned int key_len;
	char secret[SHARED_SECRET_MAX]; /* 64 byte */
P
Philipp Reisner 已提交
4736 4737
	unsigned int resp_size;
	struct hash_desc desc;
4738
	struct packet_info pi;
4739
	struct net_conf *nc;
4740
	int err, rv;
P
Philipp Reisner 已提交
4741

4742
	/* FIXME: Put the challenge/response into the preallocated socket buffer.  */
P
Philipp Reisner 已提交
4743

4744
	rcu_read_lock();
4745
	nc = rcu_dereference(connection->net_conf);
4746 4747 4748 4749
	key_len = strlen(nc->shared_secret);
	memcpy(secret, nc->shared_secret, key_len);
	rcu_read_unlock();

4750
	desc.tfm = connection->cram_hmac_tfm;
P
Philipp Reisner 已提交
4751 4752
	desc.flags = 0;

4753
	rv = crypto_hash_setkey(connection->cram_hmac_tfm, (u8 *)secret, key_len);
P
Philipp Reisner 已提交
4754
	if (rv) {
4755
		drbd_err(connection, "crypto_hash_setkey() failed with %d\n", rv);
4756
		rv = -1;
P
Philipp Reisner 已提交
4757 4758 4759 4760 4761
		goto fail;
	}

	get_random_bytes(my_challenge, CHALLENGE_LEN);

4762 4763
	sock = &connection->data;
	if (!conn_prepare_command(connection, sock)) {
4764 4765 4766
		rv = 0;
		goto fail;
	}
4767
	rv = !conn_send_command(connection, sock, P_AUTH_CHALLENGE, 0,
4768
				my_challenge, CHALLENGE_LEN);
P
Philipp Reisner 已提交
4769 4770 4771
	if (!rv)
		goto fail;

4772
	err = drbd_recv_header(connection, &pi);
4773 4774
	if (err) {
		rv = 0;
P
Philipp Reisner 已提交
4775
		goto fail;
4776
	}
P
Philipp Reisner 已提交
4777

4778
	if (pi.cmd != P_AUTH_CHALLENGE) {
4779
		drbd_err(connection, "expected AuthChallenge packet, received: %s (0x%04x)\n",
4780
			 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4781 4782 4783 4784
		rv = 0;
		goto fail;
	}

4785
	if (pi.size > CHALLENGE_LEN * 2) {
4786
		drbd_err(connection, "expected AuthChallenge payload too big.\n");
4787
		rv = -1;
P
Philipp Reisner 已提交
4788 4789 4790
		goto fail;
	}

4791
	peers_ch = kmalloc(pi.size, GFP_NOIO);
P
Philipp Reisner 已提交
4792
	if (peers_ch == NULL) {
4793
		drbd_err(connection, "kmalloc of peers_ch failed\n");
4794
		rv = -1;
P
Philipp Reisner 已提交
4795 4796 4797
		goto fail;
	}

4798
	err = drbd_recv_all_warn(connection, peers_ch, pi.size);
4799
	if (err) {
P
Philipp Reisner 已提交
4800 4801 4802 4803
		rv = 0;
		goto fail;
	}

4804
	resp_size = crypto_hash_digestsize(connection->cram_hmac_tfm);
P
Philipp Reisner 已提交
4805 4806
	response = kmalloc(resp_size, GFP_NOIO);
	if (response == NULL) {
4807
		drbd_err(connection, "kmalloc of response failed\n");
4808
		rv = -1;
P
Philipp Reisner 已提交
4809 4810 4811 4812
		goto fail;
	}

	sg_init_table(&sg, 1);
4813
	sg_set_buf(&sg, peers_ch, pi.size);
P
Philipp Reisner 已提交
4814 4815 4816

	rv = crypto_hash_digest(&desc, &sg, sg.length, response);
	if (rv) {
4817
		drbd_err(connection, "crypto_hash_digest() failed with %d\n", rv);
4818
		rv = -1;
P
Philipp Reisner 已提交
4819 4820 4821
		goto fail;
	}

4822
	if (!conn_prepare_command(connection, sock)) {
4823
		rv = 0;
P
Philipp Reisner 已提交
4824
		goto fail;
4825
	}
4826
	rv = !conn_send_command(connection, sock, P_AUTH_RESPONSE, 0,
4827
				response, resp_size);
P
Philipp Reisner 已提交
4828 4829 4830
	if (!rv)
		goto fail;

4831
	err = drbd_recv_header(connection, &pi);
4832
	if (err) {
P
Philipp Reisner 已提交
4833 4834 4835 4836
		rv = 0;
		goto fail;
	}

4837
	if (pi.cmd != P_AUTH_RESPONSE) {
4838
		drbd_err(connection, "expected AuthResponse packet, received: %s (0x%04x)\n",
4839
			 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4840 4841 4842 4843
		rv = 0;
		goto fail;
	}

4844
	if (pi.size != resp_size) {
4845
		drbd_err(connection, "expected AuthResponse payload of wrong size\n");
P
Philipp Reisner 已提交
4846 4847 4848 4849
		rv = 0;
		goto fail;
	}

4850
	err = drbd_recv_all_warn(connection, response , resp_size);
4851
	if (err) {
P
Philipp Reisner 已提交
4852 4853 4854 4855 4856
		rv = 0;
		goto fail;
	}

	right_response = kmalloc(resp_size, GFP_NOIO);
4857
	if (right_response == NULL) {
4858
		drbd_err(connection, "kmalloc of right_response failed\n");
4859
		rv = -1;
P
Philipp Reisner 已提交
4860 4861 4862 4863 4864 4865 4866
		goto fail;
	}

	sg_set_buf(&sg, my_challenge, CHALLENGE_LEN);

	rv = crypto_hash_digest(&desc, &sg, sg.length, right_response);
	if (rv) {
4867
		drbd_err(connection, "crypto_hash_digest() failed with %d\n", rv);
4868
		rv = -1;
P
Philipp Reisner 已提交
4869 4870 4871 4872 4873 4874
		goto fail;
	}

	rv = !memcmp(response, right_response, resp_size);

	if (rv)
4875
		drbd_info(connection, "Peer authenticated using %d bytes HMAC\n",
4876
		     resp_size);
4877 4878
	else
		rv = -1;
P
Philipp Reisner 已提交
4879 4880 4881 4882 4883 4884 4885 4886 4887 4888

 fail:
	kfree(peers_ch);
	kfree(response);
	kfree(right_response);

	return rv;
}
#endif

4889
int drbd_receiver(struct drbd_thread *thi)
P
Philipp Reisner 已提交
4890
{
4891
	struct drbd_connection *connection = thi->connection;
P
Philipp Reisner 已提交
4892 4893
	int h;

4894
	drbd_info(connection, "receiver (re)started\n");
P
Philipp Reisner 已提交
4895 4896

	do {
4897
		h = conn_connect(connection);
P
Philipp Reisner 已提交
4898
		if (h == 0) {
4899
			conn_disconnect(connection);
4900
			schedule_timeout_interruptible(HZ);
P
Philipp Reisner 已提交
4901 4902
		}
		if (h == -1) {
4903
			drbd_warn(connection, "Discarding network configuration.\n");
4904
			conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
4905 4906 4907
		}
	} while (h == 0);

4908
	if (h > 0)
4909
		drbdd(connection);
P
Philipp Reisner 已提交
4910

4911
	conn_disconnect(connection);
P
Philipp Reisner 已提交
4912

4913
	drbd_info(connection, "receiver terminated\n");
P
Philipp Reisner 已提交
4914 4915 4916 4917 4918
	return 0;
}

/* ********* acknowledge sender ******** */

4919
static int got_conn_RqSReply(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4920
{
4921
	struct p_req_state_reply *p = pi->data;
4922 4923 4924
	int retcode = be32_to_cpu(p->retcode);

	if (retcode >= SS_SUCCESS) {
4925
		set_bit(CONN_WD_ST_CHG_OKAY, &connection->flags);
4926
	} else {
4927
		set_bit(CONN_WD_ST_CHG_FAIL, &connection->flags);
4928
		drbd_err(connection, "Requested state change failed by peer: %s (%d)\n",
4929 4930
			 drbd_set_st_err_str(retcode), retcode);
	}
4931
	wake_up(&connection->ping_wait);
4932

4933
	return 0;
4934
}
P
Philipp Reisner 已提交
4935

4936
static int got_RqSReply(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4937
{
4938
	struct drbd_peer_device *peer_device;
4939
	struct drbd_device *device;
4940
	struct p_req_state_reply *p = pi->data;
P
Philipp Reisner 已提交
4941 4942
	int retcode = be32_to_cpu(p->retcode);

4943 4944
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
4945
		return -EIO;
4946
	device = peer_device->device;
4947

4948
	if (test_bit(CONN_WD_ST_CHG_REQ, &connection->flags)) {
4949
		D_ASSERT(device, connection->agreed_pro_version < 100);
4950
		return got_conn_RqSReply(connection, pi);
4951 4952
	}

P
Philipp Reisner 已提交
4953
	if (retcode >= SS_SUCCESS) {
4954
		set_bit(CL_ST_CHG_SUCCESS, &device->flags);
P
Philipp Reisner 已提交
4955
	} else {
4956
		set_bit(CL_ST_CHG_FAIL, &device->flags);
4957
		drbd_err(device, "Requested state change failed by peer: %s (%d)\n",
4958
			drbd_set_st_err_str(retcode), retcode);
P
Philipp Reisner 已提交
4959
	}
4960
	wake_up(&device->state_wait);
P
Philipp Reisner 已提交
4961

4962
	return 0;
P
Philipp Reisner 已提交
4963 4964
}

4965
static int got_Ping(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4966
{
4967
	return drbd_send_ping_ack(connection);
P
Philipp Reisner 已提交
4968 4969 4970

}

4971
static int got_PingAck(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4972 4973
{
	/* restore idle timeout */
4974 4975 4976
	connection->meta.socket->sk->sk_rcvtimeo = connection->net_conf->ping_int*HZ;
	if (!test_and_set_bit(GOT_PING_ACK, &connection->flags))
		wake_up(&connection->ping_wait);
P
Philipp Reisner 已提交
4977

4978
	return 0;
P
Philipp Reisner 已提交
4979 4980
}

4981
static int got_IsInSync(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
4982
{
4983
	struct drbd_peer_device *peer_device;
4984
	struct drbd_device *device;
4985
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4986 4987 4988
	sector_t sector = be64_to_cpu(p->sector);
	int blksize = be32_to_cpu(p->blksize);

4989 4990
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
4991
		return -EIO;
4992
	device = peer_device->device;
4993

4994
	D_ASSERT(device, peer_device->connection->agreed_pro_version >= 89);
P
Philipp Reisner 已提交
4995

4996
	update_peer_seq(peer_device, be32_to_cpu(p->seq_num));
P
Philipp Reisner 已提交
4997

4998 4999 5000
	if (get_ldev(device)) {
		drbd_rs_complete_io(device, sector);
		drbd_set_in_sync(device, sector, blksize);
5001
		/* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */
5002 5003
		device->rs_same_csum += (blksize >> BM_BLOCK_SHIFT);
		put_ldev(device);
5004
	}
5005 5006
	dec_rs_pending(device);
	atomic_add(blksize >> 9, &device->rs_sect_in);
P
Philipp Reisner 已提交
5007

5008
	return 0;
P
Philipp Reisner 已提交
5009 5010
}

5011
static int
5012
validate_req_change_req_state(struct drbd_device *device, u64 id, sector_t sector,
5013 5014
			      struct rb_root *root, const char *func,
			      enum drbd_req_event what, bool missing_ok)
P
Philipp Reisner 已提交
5015 5016 5017 5018
{
	struct drbd_request *req;
	struct bio_and_error m;

5019
	spin_lock_irq(&device->resource->req_lock);
5020
	req = find_request(device, root, id, sector, missing_ok, func);
P
Philipp Reisner 已提交
5021
	if (unlikely(!req)) {
5022
		spin_unlock_irq(&device->resource->req_lock);
5023
		return -EIO;
P
Philipp Reisner 已提交
5024 5025
	}
	__req_mod(req, what, &m);
5026
	spin_unlock_irq(&device->resource->req_lock);
P
Philipp Reisner 已提交
5027 5028

	if (m.bio)
5029
		complete_master_bio(device, &m);
5030
	return 0;
P
Philipp Reisner 已提交
5031 5032
}

5033
static int got_BlockAck(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
5034
{
5035
	struct drbd_peer_device *peer_device;
5036
	struct drbd_device *device;
5037
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
5038 5039 5040 5041
	sector_t sector = be64_to_cpu(p->sector);
	int blksize = be32_to_cpu(p->blksize);
	enum drbd_req_event what;

5042 5043
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
5044
		return -EIO;
5045
	device = peer_device->device;
5046

5047
	update_peer_seq(peer_device, be32_to_cpu(p->seq_num));
P
Philipp Reisner 已提交
5048

5049
	if (p->block_id == ID_SYNCER) {
5050 5051
		drbd_set_in_sync(device, sector, blksize);
		dec_rs_pending(device);
5052
		return 0;
P
Philipp Reisner 已提交
5053
	}
5054
	switch (pi->cmd) {
P
Philipp Reisner 已提交
5055
	case P_RS_WRITE_ACK:
5056
		what = WRITE_ACKED_BY_PEER_AND_SIS;
P
Philipp Reisner 已提交
5057 5058
		break;
	case P_WRITE_ACK:
5059
		what = WRITE_ACKED_BY_PEER;
P
Philipp Reisner 已提交
5060 5061
		break;
	case P_RECV_ACK:
5062
		what = RECV_ACKED_BY_PEER;
P
Philipp Reisner 已提交
5063
		break;
5064 5065
	case P_SUPERSEDED:
		what = CONFLICT_RESOLVED;
P
Philipp Reisner 已提交
5066
		break;
5067 5068
	case P_RETRY_WRITE:
		what = POSTPONE_WRITE;
P
Philipp Reisner 已提交
5069 5070
		break;
	default:
5071
		BUG();
P
Philipp Reisner 已提交
5072 5073
	}

5074 5075
	return validate_req_change_req_state(device, p->block_id, sector,
					     &device->write_requests, __func__,
5076
					     what, false);
P
Philipp Reisner 已提交
5077 5078
}

5079
static int got_NegAck(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
5080
{
5081
	struct drbd_peer_device *peer_device;
5082
	struct drbd_device *device;
5083
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
5084
	sector_t sector = be64_to_cpu(p->sector);
5085
	int size = be32_to_cpu(p->blksize);
5086
	int err;
P
Philipp Reisner 已提交
5087

5088 5089
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
5090
		return -EIO;
5091
	device = peer_device->device;
P
Philipp Reisner 已提交
5092

5093
	update_peer_seq(peer_device, be32_to_cpu(p->seq_num));
P
Philipp Reisner 已提交
5094

5095
	if (p->block_id == ID_SYNCER) {
5096 5097
		dec_rs_pending(device);
		drbd_rs_failed_io(device, sector, size);
5098
		return 0;
P
Philipp Reisner 已提交
5099
	}
5100

5101 5102
	err = validate_req_change_req_state(device, p->block_id, sector,
					    &device->write_requests, __func__,
5103
					    NEG_ACKED, true);
5104
	if (err) {
5105 5106 5107 5108 5109
		/* Protocol A has no P_WRITE_ACKs, but has P_NEG_ACKs.
		   The master bio might already be completed, therefore the
		   request is no longer in the collision hash. */
		/* In Protocol B we might already have got a P_RECV_ACK
		   but then get a P_NEG_ACK afterwards. */
5110
		drbd_set_out_of_sync(device, sector, size);
5111
	}
5112
	return 0;
P
Philipp Reisner 已提交
5113 5114
}

5115
static int got_NegDReply(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
5116
{
5117
	struct drbd_peer_device *peer_device;
5118
	struct drbd_device *device;
5119
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
5120 5121
	sector_t sector = be64_to_cpu(p->sector);

5122 5123
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
5124
		return -EIO;
5125
	device = peer_device->device;
5126

5127
	update_peer_seq(peer_device, be32_to_cpu(p->seq_num));
5128

5129
	drbd_err(device, "Got NegDReply; Sector %llus, len %u.\n",
P
Philipp Reisner 已提交
5130 5131
	    (unsigned long long)sector, be32_to_cpu(p->blksize));

5132 5133
	return validate_req_change_req_state(device, p->block_id, sector,
					     &device->read_requests, __func__,
5134
					     NEG_ACKED, false);
P
Philipp Reisner 已提交
5135 5136
}

5137
static int got_NegRSDReply(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
5138
{
5139
	struct drbd_peer_device *peer_device;
5140
	struct drbd_device *device;
P
Philipp Reisner 已提交
5141 5142
	sector_t sector;
	int size;
5143
	struct p_block_ack *p = pi->data;
5144

5145 5146
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
5147
		return -EIO;
5148
	device = peer_device->device;
P
Philipp Reisner 已提交
5149 5150 5151 5152

	sector = be64_to_cpu(p->sector);
	size = be32_to_cpu(p->blksize);

5153
	update_peer_seq(peer_device, be32_to_cpu(p->seq_num));
P
Philipp Reisner 已提交
5154

5155
	dec_rs_pending(device);
P
Philipp Reisner 已提交
5156

5157 5158
	if (get_ldev_if_state(device, D_FAILED)) {
		drbd_rs_complete_io(device, sector);
5159
		switch (pi->cmd) {
5160
		case P_NEG_RS_DREPLY:
5161
			drbd_rs_failed_io(device, sector, size);
5162 5163 5164
		case P_RS_CANCEL:
			break;
		default:
5165
			BUG();
5166
		}
5167
		put_ldev(device);
P
Philipp Reisner 已提交
5168 5169
	}

5170
	return 0;
P
Philipp Reisner 已提交
5171 5172
}

5173
static int got_BarrierAck(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
5174
{
5175
	struct p_barrier_ack *p = pi->data;
5176
	struct drbd_peer_device *peer_device;
5177
	int vnr;
5178

5179
	tl_release(connection, p->barrier, be32_to_cpu(p->set_size));
P
Philipp Reisner 已提交
5180

5181
	rcu_read_lock();
5182 5183 5184
	idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
		struct drbd_device *device = peer_device->device;

5185 5186 5187 5188 5189
		if (device->state.conn == C_AHEAD &&
		    atomic_read(&device->ap_in_flight) == 0 &&
		    !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &device->flags)) {
			device->start_resync_timer.expires = jiffies + HZ;
			add_timer(&device->start_resync_timer);
5190
		}
5191
	}
5192
	rcu_read_unlock();
5193

5194
	return 0;
P
Philipp Reisner 已提交
5195 5196
}

5197
static int got_OVResult(struct drbd_connection *connection, struct packet_info *pi)
P
Philipp Reisner 已提交
5198
{
5199
	struct drbd_peer_device *peer_device;
5200
	struct drbd_device *device;
5201
	struct p_block_ack *p = pi->data;
5202
	struct drbd_device_work *dw;
P
Philipp Reisner 已提交
5203 5204 5205
	sector_t sector;
	int size;

5206 5207
	peer_device = conn_peer_device(connection, pi->vnr);
	if (!peer_device)
5208
		return -EIO;
5209
	device = peer_device->device;
5210

P
Philipp Reisner 已提交
5211 5212 5213
	sector = be64_to_cpu(p->sector);
	size = be32_to_cpu(p->blksize);

5214
	update_peer_seq(peer_device, be32_to_cpu(p->seq_num));
P
Philipp Reisner 已提交
5215 5216

	if (be64_to_cpu(p->block_id) == ID_OUT_OF_SYNC)
5217
		drbd_ov_out_of_sync_found(device, sector, size);
P
Philipp Reisner 已提交
5218
	else
5219
		ov_out_of_sync_print(device);
P
Philipp Reisner 已提交
5220

5221
	if (!get_ldev(device))
5222
		return 0;
5223

5224 5225
	drbd_rs_complete_io(device, sector);
	dec_rs_pending(device);
P
Philipp Reisner 已提交
5226

5227
	--device->ov_left;
5228 5229

	/* let's advance progress step marks only for every other megabyte */
5230 5231
	if ((device->ov_left & 0x200) == 0x200)
		drbd_advance_rs_marks(device, device->ov_left);
5232

5233
	if (device->ov_left == 0) {
5234 5235 5236 5237 5238
		dw = kmalloc(sizeof(*dw), GFP_NOIO);
		if (dw) {
			dw->w.cb = w_ov_finished;
			dw->device = device;
			drbd_queue_work(&peer_device->connection->sender_work, &dw->w);
P
Philipp Reisner 已提交
5239
		} else {
5240
			drbd_err(device, "kmalloc(dw) failed.");
5241 5242
			ov_out_of_sync_print(device);
			drbd_resync_finished(device);
P
Philipp Reisner 已提交
5243 5244
		}
	}
5245
	put_ldev(device);
5246
	return 0;
P
Philipp Reisner 已提交
5247 5248
}

5249
static int got_skip(struct drbd_connection *connection, struct packet_info *pi)
5250
{
5251
	return 0;
P
Philipp Reisner 已提交
5252 5253
}

5254
static int connection_finish_peer_reqs(struct drbd_connection *connection)
5255
{
5256
	struct drbd_peer_device *peer_device;
P
Philipp Reisner 已提交
5257
	int vnr, not_empty = 0;
5258 5259

	do {
5260
		clear_bit(SIGNAL_ASENDER, &connection->flags);
5261
		flush_signals(current);
P
Philipp Reisner 已提交
5262 5263

		rcu_read_lock();
5264 5265
		idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
			struct drbd_device *device = peer_device->device;
5266
			kref_get(&device->kref);
P
Philipp Reisner 已提交
5267
			rcu_read_unlock();
5268
			if (drbd_finish_peer_reqs(device)) {
5269
				kref_put(&device->kref, drbd_destroy_device);
P
Philipp Reisner 已提交
5270
				return 1;
5271
			}
5272
			kref_put(&device->kref, drbd_destroy_device);
P
Philipp Reisner 已提交
5273
			rcu_read_lock();
5274
		}
5275
		set_bit(SIGNAL_ASENDER, &connection->flags);
5276

5277
		spin_lock_irq(&connection->resource->req_lock);
5278 5279
		idr_for_each_entry(&connection->peer_devices, peer_device, vnr) {
			struct drbd_device *device = peer_device->device;
5280
			not_empty = !list_empty(&device->done_ee);
5281 5282 5283
			if (not_empty)
				break;
		}
5284
		spin_unlock_irq(&connection->resource->req_lock);
P
Philipp Reisner 已提交
5285
		rcu_read_unlock();
5286 5287 5288
	} while (not_empty);

	return 0;
5289 5290
}

P
Philipp Reisner 已提交
5291 5292
struct asender_cmd {
	size_t pkt_size;
5293
	int (*fn)(struct drbd_connection *connection, struct packet_info *);
P
Philipp Reisner 已提交
5294 5295
};

5296
static struct asender_cmd asender_tbl[] = {
5297 5298
	[P_PING]	    = { 0, got_Ping },
	[P_PING_ACK]	    = { 0, got_PingAck },
P
Philipp Reisner 已提交
5299 5300 5301
	[P_RECV_ACK]	    = { sizeof(struct p_block_ack), got_BlockAck },
	[P_WRITE_ACK]	    = { sizeof(struct p_block_ack), got_BlockAck },
	[P_RS_WRITE_ACK]    = { sizeof(struct p_block_ack), got_BlockAck },
5302
	[P_SUPERSEDED]   = { sizeof(struct p_block_ack), got_BlockAck },
P
Philipp Reisner 已提交
5303 5304
	[P_NEG_ACK]	    = { sizeof(struct p_block_ack), got_NegAck },
	[P_NEG_DREPLY]	    = { sizeof(struct p_block_ack), got_NegDReply },
5305
	[P_NEG_RS_DREPLY]   = { sizeof(struct p_block_ack), got_NegRSDReply },
P
Philipp Reisner 已提交
5306 5307 5308 5309
	[P_OV_RESULT]	    = { sizeof(struct p_block_ack), got_OVResult },
	[P_BARRIER_ACK]	    = { sizeof(struct p_barrier_ack), got_BarrierAck },
	[P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply },
	[P_RS_IS_IN_SYNC]   = { sizeof(struct p_block_ack), got_IsInSync },
5310
	[P_DELAY_PROBE]     = { sizeof(struct p_delay_probe93), got_skip },
5311 5312 5313
	[P_RS_CANCEL]       = { sizeof(struct p_block_ack), got_NegRSDReply },
	[P_CONN_ST_CHG_REPLY]={ sizeof(struct p_req_state_reply), got_conn_RqSReply },
	[P_RETRY_WRITE]	    = { sizeof(struct p_block_ack), got_BlockAck },
5314
};
P
Philipp Reisner 已提交
5315 5316 5317

int drbd_asender(struct drbd_thread *thi)
{
5318
	struct drbd_connection *connection = thi->connection;
P
Philipp Reisner 已提交
5319
	struct asender_cmd *cmd = NULL;
5320
	struct packet_info pi;
5321
	int rv;
5322
	void *buf    = connection->meta.rbuf;
P
Philipp Reisner 已提交
5323
	int received = 0;
5324
	unsigned int header_size = drbd_header_size(connection);
5325
	int expect   = header_size;
5326 5327
	bool ping_timeout_active = false;
	struct net_conf *nc;
5328
	int ping_timeo, tcp_cork, ping_int;
P
Philipp Reisner 已提交
5329
	struct sched_param param = { .sched_priority = 2 };
P
Philipp Reisner 已提交
5330

P
Philipp Reisner 已提交
5331 5332
	rv = sched_setscheduler(current, SCHED_RR, &param);
	if (rv < 0)
5333
		drbd_err(connection, "drbd_asender: ERROR set priority, ret=%d\n", rv);
P
Philipp Reisner 已提交
5334

5335
	while (get_t_state(thi) == RUNNING) {
5336
		drbd_thread_current_set_cpu(thi);
P
Philipp Reisner 已提交
5337

5338
		rcu_read_lock();
5339
		nc = rcu_dereference(connection->net_conf);
5340
		ping_timeo = nc->ping_timeo;
5341
		tcp_cork = nc->tcp_cork;
5342 5343 5344
		ping_int = nc->ping_int;
		rcu_read_unlock();

5345 5346
		if (test_and_clear_bit(SEND_PING, &connection->flags)) {
			if (drbd_send_ping(connection)) {
5347
				drbd_err(connection, "drbd_send_ping has failed\n");
P
Philipp Reisner 已提交
5348
				goto reconnect;
5349
			}
5350
			connection->meta.socket->sk->sk_rcvtimeo = ping_timeo * HZ / 10;
5351
			ping_timeout_active = true;
P
Philipp Reisner 已提交
5352 5353
		}

5354 5355
		/* TODO: conditionally cork; it may hurt latency if we cork without
		   much to send */
5356
		if (tcp_cork)
5357 5358
			drbd_tcp_cork(connection->meta.socket);
		if (connection_finish_peer_reqs(connection)) {
5359
			drbd_err(connection, "connection_finish_peer_reqs() failed\n");
5360
			goto reconnect;
P
Philipp Reisner 已提交
5361 5362
		}
		/* but unconditionally uncork unless disabled */
5363
		if (tcp_cork)
5364
			drbd_tcp_uncork(connection->meta.socket);
P
Philipp Reisner 已提交
5365 5366 5367 5368 5369

		/* short circuit, recv_msg would return EINTR anyways. */
		if (signal_pending(current))
			continue;

5370 5371
		rv = drbd_recv_short(connection->meta.socket, buf, expect-received, 0);
		clear_bit(SIGNAL_ASENDER, &connection->flags);
P
Philipp Reisner 已提交
5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388

		flush_signals(current);

		/* Note:
		 * -EINTR	 (on meta) we got a signal
		 * -EAGAIN	 (on meta) rcvtimeo expired
		 * -ECONNRESET	 other side closed the connection
		 * -ERESTARTSYS  (on data) we got a signal
		 * rv <  0	 other than above: unexpected error!
		 * rv == expected: full header or command
		 * rv <  expected: "woken" by signal during receive
		 * rv == 0	 : "connection shut down by peer"
		 */
		if (likely(rv > 0)) {
			received += rv;
			buf	 += rv;
		} else if (rv == 0) {
5389
			if (test_bit(DISCONNECT_SENT, &connection->flags)) {
5390 5391
				long t;
				rcu_read_lock();
5392
				t = rcu_dereference(connection->net_conf)->ping_timeo * HZ/10;
5393 5394
				rcu_read_unlock();

5395 5396
				t = wait_event_timeout(connection->ping_wait,
						       connection->cstate < C_WF_REPORT_PARAMS,
5397
						       t);
5398 5399 5400
				if (t)
					break;
			}
5401
			drbd_err(connection, "meta connection shut down by peer.\n");
P
Philipp Reisner 已提交
5402 5403
			goto reconnect;
		} else if (rv == -EAGAIN) {
5404 5405
			/* If the data socket received something meanwhile,
			 * that is good enough: peer is still alive. */
5406 5407
			if (time_after(connection->last_received,
				jiffies - connection->meta.socket->sk->sk_rcvtimeo))
5408
				continue;
5409
			if (ping_timeout_active) {
5410
				drbd_err(connection, "PingAck did not arrive in time.\n");
P
Philipp Reisner 已提交
5411 5412
				goto reconnect;
			}
5413
			set_bit(SEND_PING, &connection->flags);
P
Philipp Reisner 已提交
5414 5415 5416 5417
			continue;
		} else if (rv == -EINTR) {
			continue;
		} else {
5418
			drbd_err(connection, "sock_recvmsg returned %d\n", rv);
P
Philipp Reisner 已提交
5419 5420 5421 5422
			goto reconnect;
		}

		if (received == expect && cmd == NULL) {
5423
			if (decode_header(connection, connection->meta.rbuf, &pi))
P
Philipp Reisner 已提交
5424
				goto reconnect;
5425
			cmd = &asender_tbl[pi.cmd];
5426
			if (pi.cmd >= ARRAY_SIZE(asender_tbl) || !cmd->fn) {
5427
				drbd_err(connection, "Unexpected meta packet %s (0x%04x)\n",
5428
					 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
5429 5430
				goto disconnect;
			}
5431
			expect = header_size + cmd->pkt_size;
5432
			if (pi.size != expect - header_size) {
5433
				drbd_err(connection, "Wrong packet size on meta (c: %d, l: %d)\n",
5434
					pi.cmd, pi.size);
P
Philipp Reisner 已提交
5435
				goto reconnect;
5436
			}
P
Philipp Reisner 已提交
5437 5438
		}
		if (received == expect) {
5439
			bool err;
5440

5441
			err = cmd->fn(connection, &pi);
5442
			if (err) {
5443
				drbd_err(connection, "%pf failed\n", cmd->fn);
P
Philipp Reisner 已提交
5444
				goto reconnect;
5445
			}
P
Philipp Reisner 已提交
5446

5447
			connection->last_received = jiffies;
5448

5449 5450
			if (cmd == &asender_tbl[P_PING_ACK]) {
				/* restore idle timeout */
5451
				connection->meta.socket->sk->sk_rcvtimeo = ping_int * HZ;
5452 5453
				ping_timeout_active = false;
			}
5454

5455
			buf	 = connection->meta.rbuf;
P
Philipp Reisner 已提交
5456
			received = 0;
5457
			expect	 = header_size;
P
Philipp Reisner 已提交
5458 5459 5460 5461 5462 5463
			cmd	 = NULL;
		}
	}

	if (0) {
reconnect:
5464 5465
		conn_request_state(connection, NS(conn, C_NETWORK_FAILURE), CS_HARD);
		conn_md_sync(connection);
P
Philipp Reisner 已提交
5466 5467 5468
	}
	if (0) {
disconnect:
5469
		conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
5470
	}
5471
	clear_bit(SIGNAL_ASENDER, &connection->flags);
P
Philipp Reisner 已提交
5472

5473
	drbd_info(connection, "asender terminated\n");
P
Philipp Reisner 已提交
5474 5475 5476

	return 0;
}