drbd_receiver.c 137.5 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 47 48 49 50
/*
   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"
#include "drbd_req.h"

#include "drbd_vli.h"

51 52
struct packet_info {
	enum drbd_packet cmd;
53 54
	unsigned int size;
	unsigned int vnr;
55
	void *data;
56 57
};

P
Philipp Reisner 已提交
58 59 60 61 62 63
enum finish_epoch {
	FE_STILL_LIVE,
	FE_DESTROYED,
	FE_RECYCLED,
};

64
static int drbd_do_features(struct drbd_tconn *tconn);
65
static int drbd_do_auth(struct drbd_tconn *tconn);
66
static int drbd_disconnected(int vnr, void *p, void *data);
P
Philipp Reisner 已提交
67 68

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


#define GFP_TRY	(__GFP_HIGHMEM | __GFP_NOWARN)

74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
/*
 * 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;
92 93 94 95

	if (!page)
		return NULL;

96 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
	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;
}

153 154
static struct page *__drbd_alloc_pages(struct drbd_conf *mdev,
				       unsigned int number)
P
Philipp Reisner 已提交
155 156
{
	struct page *page = NULL;
157
	struct page *tmp = NULL;
158
	unsigned int i = 0;
P
Philipp Reisner 已提交
159 160 161

	/* Yes, testing drbd_pp_vacant outside the lock is racy.
	 * So what. It saves a spin_lock. */
162
	if (drbd_pp_vacant >= number) {
P
Philipp Reisner 已提交
163
		spin_lock(&drbd_pp_lock);
164 165 166
		page = page_chain_del(&drbd_pp_pool, number);
		if (page)
			drbd_pp_vacant -= number;
P
Philipp Reisner 已提交
167
		spin_unlock(&drbd_pp_lock);
168 169
		if (page)
			return page;
P
Philipp Reisner 已提交
170
	}
171

P
Philipp Reisner 已提交
172 173 174
	/* 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.  */
175 176 177 178 179 180 181 182 183 184 185 186
	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.
187
	 * No need to jump around here, drbd_alloc_pages will retry this
188 189 190 191 192 193 194 195 196
	 * 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
197 198
}

199 200
static void reclaim_finished_net_peer_reqs(struct drbd_conf *mdev,
					   struct list_head *to_be_freed)
P
Philipp Reisner 已提交
201
{
202
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
203 204 205 206 207 208 209 210
	struct list_head *le, *tle;

	/* 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... */

	list_for_each_safe(le, tle, &mdev->net_ee) {
211
		peer_req = list_entry(le, struct drbd_peer_request, w.list);
212
		if (drbd_peer_req_has_active_page(peer_req))
P
Philipp Reisner 已提交
213 214 215 216 217 218 219 220
			break;
		list_move(le, to_be_freed);
	}
}

static void drbd_kick_lo_and_reclaim_net(struct drbd_conf *mdev)
{
	LIST_HEAD(reclaimed);
221
	struct drbd_peer_request *peer_req, *t;
P
Philipp Reisner 已提交
222

223
	spin_lock_irq(&mdev->tconn->req_lock);
224
	reclaim_finished_net_peer_reqs(mdev, &reclaimed);
225
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
226

227
	list_for_each_entry_safe(peer_req, t, &reclaimed, w.list)
228
		drbd_free_net_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
229 230 231
}

/**
232
 * drbd_alloc_pages() - Returns @number pages, retries forever (or until signalled)
P
Philipp Reisner 已提交
233
 * @mdev:	DRBD device.
234 235 236 237 238 239
 * @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
 * the kernel, unless this allocation would exceed the max_buffers setting.
 * Possibly retry until DRBD frees sufficient pages somewhere else.
P
Philipp Reisner 已提交
240
 *
241
 * Returns a page chain linked via page->private.
P
Philipp Reisner 已提交
242
 */
243 244
struct page *drbd_alloc_pages(struct drbd_conf *mdev, unsigned int number,
			      bool retry)
P
Philipp Reisner 已提交
245 246
{
	struct page *page = NULL;
247
	struct net_conf *nc;
P
Philipp Reisner 已提交
248
	DEFINE_WAIT(wait);
249
	int mxb;
P
Philipp Reisner 已提交
250

251 252
	/* Yes, we may run up to @number over max_buffers. If we
	 * follow it strictly, the admin will get it wrong anyways. */
253 254 255 256 257 258
	rcu_read_lock();
	nc = rcu_dereference(mdev->tconn->net_conf);
	mxb = nc ? nc->max_buffers : 1000000;
	rcu_read_unlock();

	if (atomic_read(&mdev->pp_in_use) < mxb)
259
		page = __drbd_alloc_pages(mdev, number);
P
Philipp Reisner 已提交
260

261
	while (page == NULL) {
P
Philipp Reisner 已提交
262 263 264 265
		prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE);

		drbd_kick_lo_and_reclaim_net(mdev);

266
		if (atomic_read(&mdev->pp_in_use) < mxb) {
267
			page = __drbd_alloc_pages(mdev, number);
P
Philipp Reisner 已提交
268 269 270 271 272 273 274 275
			if (page)
				break;
		}

		if (!retry)
			break;

		if (signal_pending(current)) {
276
			dev_warn(DEV, "drbd_alloc_pages interrupted!\n");
P
Philipp Reisner 已提交
277 278 279 280 281 282 283
			break;
		}

		schedule();
	}
	finish_wait(&drbd_pp_wait, &wait);

284 285
	if (page)
		atomic_add(number, &mdev->pp_in_use);
P
Philipp Reisner 已提交
286 287 288
	return page;
}

289
/* Must not be used from irq, as that may deadlock: see drbd_alloc_pages.
290
 * Is also used from inside an other spin_lock_irq(&mdev->tconn->req_lock);
291 292
 * Either links the page chain back to the global pool,
 * or returns all pages to the system. */
293
static void drbd_free_pages(struct drbd_conf *mdev, struct page *page, int is_net)
P
Philipp Reisner 已提交
294
{
295
	atomic_t *a = is_net ? &mdev->pp_in_use_by_net : &mdev->pp_in_use;
P
Philipp Reisner 已提交
296
	int i;
297

298
	if (drbd_pp_vacant > (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count)
299 300 301 302 303 304 305 306
		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
307
	}
308
	i = atomic_sub_return(i, a);
309
	if (i < 0)
310 311
		dev_warn(DEV, "ASSERTION FAILED: %s: %d < 0\n",
			is_net ? "pp_in_use_by_net" : "pp_in_use", i);
P
Philipp Reisner 已提交
312 313 314 315 316 317 318 319
	wake_up(&drbd_pp_wait);
}

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

You must not have the req_lock:
320
 drbd_free_peer_req()
321
 drbd_alloc_peer_req()
322
 drbd_free_peer_reqs()
P
Philipp Reisner 已提交
323
 drbd_ee_fix_bhs()
324
 drbd_finish_peer_reqs()
P
Philipp Reisner 已提交
325 326 327 328
 drbd_clear_done_ee()
 drbd_wait_ee_list_empty()
*/

329
struct drbd_peer_request *
330 331
drbd_alloc_peer_req(struct drbd_conf *mdev, u64 id, sector_t sector,
		    unsigned int data_size, gfp_t gfp_mask) __must_hold(local)
P
Philipp Reisner 已提交
332
{
333
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
334
	struct page *page;
335
	unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT;
P
Philipp Reisner 已提交
336

337
	if (drbd_insert_fault(mdev, DRBD_FAULT_AL_EE))
P
Philipp Reisner 已提交
338 339
		return NULL;

340 341
	peer_req = mempool_alloc(drbd_ee_mempool, gfp_mask & ~__GFP_HIGHMEM);
	if (!peer_req) {
P
Philipp Reisner 已提交
342
		if (!(gfp_mask & __GFP_NOWARN))
343
			dev_err(DEV, "%s: allocation failed\n", __func__);
P
Philipp Reisner 已提交
344 345 346
		return NULL;
	}

347
	page = drbd_alloc_pages(mdev, nr_pages, (gfp_mask & __GFP_WAIT));
348 349
	if (!page)
		goto fail;
P
Philipp Reisner 已提交
350

351 352 353 354 355 356 357
	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;
358
	peer_req->w.mdev = mdev;
359 360 361
	peer_req->pages = page;
	atomic_set(&peer_req->pending_bios, 0);
	peer_req->flags = 0;
362 363 364 365
	/*
	 * The block_id is opaque to the receiver.  It is not endianness
	 * converted, and sent back to the sender unchanged.
	 */
366
	peer_req->block_id = id;
P
Philipp Reisner 已提交
367

368
	return peer_req;
P
Philipp Reisner 已提交
369

370
 fail:
371
	mempool_free(peer_req, drbd_ee_mempool);
P
Philipp Reisner 已提交
372 373 374
	return NULL;
}

375
void __drbd_free_peer_req(struct drbd_conf *mdev, struct drbd_peer_request *peer_req,
376
		       int is_net)
P
Philipp Reisner 已提交
377
{
378 379
	if (peer_req->flags & EE_HAS_DIGEST)
		kfree(peer_req->digest);
380
	drbd_free_pages(mdev, peer_req->pages, is_net);
381 382 383
	D_ASSERT(atomic_read(&peer_req->pending_bios) == 0);
	D_ASSERT(drbd_interval_empty(&peer_req->i));
	mempool_free(peer_req, drbd_ee_mempool);
P
Philipp Reisner 已提交
384 385
}

386
int drbd_free_peer_reqs(struct drbd_conf *mdev, struct list_head *list)
P
Philipp Reisner 已提交
387 388
{
	LIST_HEAD(work_list);
389
	struct drbd_peer_request *peer_req, *t;
P
Philipp Reisner 已提交
390
	int count = 0;
391
	int is_net = list == &mdev->net_ee;
P
Philipp Reisner 已提交
392

393
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
394
	list_splice_init(list, &work_list);
395
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
396

397
	list_for_each_entry_safe(peer_req, t, &work_list, w.list) {
398
		__drbd_free_peer_req(mdev, peer_req, is_net);
P
Philipp Reisner 已提交
399 400 401 402 403
		count++;
	}
	return count;
}

404 405
/*
 * See also comments in _req_mod(,BARRIER_ACKED) and receive_Barrier.
P
Philipp Reisner 已提交
406
 */
407
static int drbd_finish_peer_reqs(struct drbd_conf *mdev)
P
Philipp Reisner 已提交
408 409 410
{
	LIST_HEAD(work_list);
	LIST_HEAD(reclaimed);
411
	struct drbd_peer_request *peer_req, *t;
412
	int err = 0;
P
Philipp Reisner 已提交
413

414
	spin_lock_irq(&mdev->tconn->req_lock);
415
	reclaim_finished_net_peer_reqs(mdev, &reclaimed);
P
Philipp Reisner 已提交
416
	list_splice_init(&mdev->done_ee, &work_list);
417
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
418

419
	list_for_each_entry_safe(peer_req, t, &reclaimed, w.list)
420
		drbd_free_net_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
421 422

	/* possible callbacks here:
423
	 * e_end_block, and e_end_resync_block, e_send_discard_write.
P
Philipp Reisner 已提交
424 425
	 * all ignore the last argument.
	 */
426
	list_for_each_entry_safe(peer_req, t, &work_list, w.list) {
427 428
		int err2;

P
Philipp Reisner 已提交
429
		/* list_del not necessary, next/prev members not touched */
430 431 432
		err2 = peer_req->w.cb(&peer_req->w, !!err);
		if (!err)
			err = err2;
433
		drbd_free_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
434 435 436
	}
	wake_up(&mdev->ee_wait);

437
	return err;
P
Philipp Reisner 已提交
438 439
}

440 441
static void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
				     struct list_head *head)
P
Philipp Reisner 已提交
442 443 444 445 446 447 448
{
	DEFINE_WAIT(wait);

	/* avoids spin_lock/unlock
	 * and calling prepare_to_wait in the fast path */
	while (!list_empty(head)) {
		prepare_to_wait(&mdev->ee_wait, &wait, TASK_UNINTERRUPTIBLE);
449
		spin_unlock_irq(&mdev->tconn->req_lock);
J
Jens Axboe 已提交
450
		io_schedule();
P
Philipp Reisner 已提交
451
		finish_wait(&mdev->ee_wait, &wait);
452
		spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
453 454 455
	}
}

456 457
static void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
				    struct list_head *head)
P
Philipp Reisner 已提交
458
{
459
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
460
	_drbd_wait_ee_list_empty(mdev, head);
461
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
462 463 464 465
}

/* see also kernel_accept; which is only present since 2.6.18.
 * also we want to log which part of it failed, exactly */
466
static int drbd_accept(const char **what, struct socket *sock, struct socket **newsock)
P
Philipp Reisner 已提交
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
{
	struct sock *sk = sock->sk;
	int err = 0;

	*what = "listen";
	err = sock->ops->listen(sock, 5);
	if (err < 0)
		goto out;

	*what = "sock_create_lite";
	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
			       newsock);
	if (err < 0)
		goto out;

	*what = "accept";
	err = sock->ops->accept(sock, *newsock, 0);
	if (err < 0) {
		sock_release(*newsock);
		*newsock = NULL;
		goto out;
	}
	(*newsock)->ops  = sock->ops;

out:
	return err;
}

495
static int drbd_recv_short(struct socket *sock, void *buf, size_t size, int flags)
P
Philipp Reisner 已提交
496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516
{
	mm_segment_t oldfs;
	struct kvec iov = {
		.iov_base = buf,
		.iov_len = size,
	};
	struct msghdr msg = {
		.msg_iovlen = 1,
		.msg_iov = (struct iovec *)&iov,
		.msg_flags = (flags ? flags : MSG_WAITALL | MSG_NOSIGNAL)
	};
	int rv;

	oldfs = get_fs();
	set_fs(KERNEL_DS);
	rv = sock_recvmsg(sock, &msg, size, msg.msg_flags);
	set_fs(oldfs);

	return rv;
}

517
static int drbd_recv(struct drbd_tconn *tconn, void *buf, size_t size)
P
Philipp Reisner 已提交
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
{
	mm_segment_t oldfs;
	struct kvec iov = {
		.iov_base = buf,
		.iov_len = size,
	};
	struct msghdr msg = {
		.msg_iovlen = 1,
		.msg_iov = (struct iovec *)&iov,
		.msg_flags = MSG_WAITALL | MSG_NOSIGNAL
	};
	int rv;

	oldfs = get_fs();
	set_fs(KERNEL_DS);

	for (;;) {
535
		rv = sock_recvmsg(tconn->data.socket, &msg, size, msg.msg_flags);
P
Philipp Reisner 已提交
536 537 538 539 540 541 542 543 544 545
		if (rv == size)
			break;

		/* Note:
		 * ECONNRESET	other side closed the connection
		 * ERESTARTSYS	(on  sock) we got a signal
		 */

		if (rv < 0) {
			if (rv == -ECONNRESET)
546
				conn_info(tconn, "sock was reset by peer\n");
P
Philipp Reisner 已提交
547
			else if (rv != -ERESTARTSYS)
548
				conn_err(tconn, "sock_recvmsg returned %d\n", rv);
P
Philipp Reisner 已提交
549 550
			break;
		} else if (rv == 0) {
551
			conn_info(tconn, "sock was shut down by peer\n");
P
Philipp Reisner 已提交
552 553 554 555 556 557 558 559 560 561 562 563 564
			break;
		} else	{
			/* signal came in, or peer/link went down,
			 * after we read a partial message
			 */
			/* D_ASSERT(signal_pending(current)); */
			break;
		}
	};

	set_fs(oldfs);

	if (rv != size)
565
		conn_request_state(tconn, NS(conn, C_BROKEN_PIPE), CS_HARD);
P
Philipp Reisner 已提交
566 567 568 569

	return rv;
}

570 571 572 573 574 575 576 577 578 579 580 581 582
static int drbd_recv_all(struct drbd_tconn *tconn, void *buf, size_t size)
{
	int err;

	err = drbd_recv(tconn, buf, size);
	if (err != size) {
		if (err >= 0)
			err = -EIO;
	} else
		err = 0;
	return err;
}

583 584 585 586 587 588 589 590 591 592
static int drbd_recv_all_warn(struct drbd_tconn *tconn, void *buf, size_t size)
{
	int err;

	err = drbd_recv_all(tconn, buf, size);
	if (err && !signal_pending(current))
		conn_warn(tconn, "short read (expected size %d)\n", (int)size);
	return err;
}

593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
/* 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;
	}
}

612
static struct socket *drbd_try_connect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
613 614 615 616
{
	const char *what;
	struct socket *sock;
	struct sockaddr_in6 src_in6;
617 618 619 620
	struct sockaddr_in6 peer_in6;
	struct net_conf *nc;
	int err, peer_addr_len, my_addr_len;
	int sndbuf_size, rcvbuf_size, try_connect_int;
P
Philipp Reisner 已提交
621 622
	int disconnect_on_error = 1;

623 624 625 626
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);
	if (!nc) {
		rcu_read_unlock();
P
Philipp Reisner 已提交
627
		return NULL;
628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
	}

	sndbuf_size = nc->sndbuf_size;
	rcvbuf_size = nc->rcvbuf_size;
	try_connect_int = nc->try_connect_int;

	my_addr_len = min_t(int, nc->my_addr_len, sizeof(src_in6));
	memcpy(&src_in6, nc->my_addr, my_addr_len);

	if (((struct sockaddr *)nc->my_addr)->sa_family == AF_INET6)
		src_in6.sin6_port = 0;
	else
		((struct sockaddr_in *)&src_in6)->sin_port = 0; /* AF_INET & AF_SCI */

	peer_addr_len = min_t(int, nc->peer_addr_len, sizeof(src_in6));
	memcpy(&peer_in6, nc->peer_addr, peer_addr_len);

	rcu_read_unlock();
P
Philipp Reisner 已提交
646 647

	what = "sock_create_kern";
648 649
	err = sock_create_kern(((struct sockaddr *)&src_in6)->sa_family,
			       SOCK_STREAM, IPPROTO_TCP, &sock);
P
Philipp Reisner 已提交
650 651 652 653 654 655
	if (err < 0) {
		sock = NULL;
		goto out;
	}

	sock->sk->sk_rcvtimeo =
656 657
	sock->sk->sk_sndtimeo = try_connect_int * HZ;
	drbd_setbufsize(sock, sndbuf_size, rcvbuf_size);
P
Philipp Reisner 已提交
658 659 660 661 662 663 664 665 666

       /* 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";
667
	err = sock->ops->bind(sock, (struct sockaddr *) &src_in6, my_addr_len);
P
Philipp Reisner 已提交
668 669 670 671 672 673 674
	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";
675
	err = sock->ops->connect(sock, (struct sockaddr *) &peer_in6, peer_addr_len, 0);
P
Philipp Reisner 已提交
676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692

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:
693
			conn_err(tconn, "%s failed, err = %d\n", what, err);
P
Philipp Reisner 已提交
694 695
		}
		if (disconnect_on_error)
696
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
697
	}
698

P
Philipp Reisner 已提交
699 700 701
	return sock;
}

702
static struct socket *drbd_wait_for_connect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
703
{
704 705
	int timeo, err, my_addr_len;
	int sndbuf_size, rcvbuf_size, try_connect_int;
P
Philipp Reisner 已提交
706
	struct socket *s_estab = NULL, *s_listen;
707 708
	struct sockaddr_in6 my_addr;
	struct net_conf *nc;
P
Philipp Reisner 已提交
709 710
	const char *what;

711 712 713 714
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);
	if (!nc) {
		rcu_read_unlock();
P
Philipp Reisner 已提交
715
		return NULL;
716 717 718 719 720 721 722 723 724
	}

	sndbuf_size = nc->sndbuf_size;
	rcvbuf_size = nc->rcvbuf_size;
	try_connect_int = nc->try_connect_int;

	my_addr_len = min_t(int, nc->my_addr_len, sizeof(struct sockaddr_in6));
	memcpy(&my_addr, nc->my_addr, my_addr_len);
	rcu_read_unlock();
P
Philipp Reisner 已提交
725 726

	what = "sock_create_kern";
727
	err = sock_create_kern(((struct sockaddr *)&my_addr)->sa_family,
P
Philipp Reisner 已提交
728 729 730 731 732 733
		SOCK_STREAM, IPPROTO_TCP, &s_listen);
	if (err) {
		s_listen = NULL;
		goto out;
	}

734
	timeo = try_connect_int * HZ;
P
Philipp Reisner 已提交
735 736 737 738 739
	timeo += (random32() & 1) ? timeo / 7 : -timeo / 7; /* 28.5% random jitter */

	s_listen->sk->sk_reuse    = 1; /* SO_REUSEADDR */
	s_listen->sk->sk_rcvtimeo = timeo;
	s_listen->sk->sk_sndtimeo = timeo;
740
	drbd_setbufsize(s_listen, sndbuf_size, rcvbuf_size);
P
Philipp Reisner 已提交
741 742

	what = "bind before listen";
743
	err = s_listen->ops->bind(s_listen, (struct sockaddr *)&my_addr, my_addr_len);
P
Philipp Reisner 已提交
744 745 746
	if (err < 0)
		goto out;

747
	err = drbd_accept(&what, s_listen, &s_estab);
P
Philipp Reisner 已提交
748 749 750 751 752 753

out:
	if (s_listen)
		sock_release(s_listen);
	if (err < 0) {
		if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) {
754
			conn_err(tconn, "%s failed, err = %d\n", what, err);
755
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
756 757 758 759 760 761
		}
	}

	return s_estab;
}

762
static int decode_header(struct drbd_tconn *, void *, struct packet_info *);
P
Philipp Reisner 已提交
763

764 765 766 767 768
static int send_first_packet(struct drbd_tconn *tconn, struct drbd_socket *sock,
			     enum drbd_packet cmd)
{
	if (!conn_prepare_command(tconn, sock))
		return -EIO;
769
	return conn_send_command(tconn, sock, cmd, 0, NULL, 0);
P
Philipp Reisner 已提交
770 771
}

772
static int receive_first_packet(struct drbd_tconn *tconn, struct socket *sock)
P
Philipp Reisner 已提交
773
{
774 775 776
	unsigned int header_size = drbd_header_size(tconn);
	struct packet_info pi;
	int err;
P
Philipp Reisner 已提交
777

778 779 780 781 782 783 784 785 786 787
	err = drbd_recv_short(sock, tconn->data.rbuf, header_size, 0);
	if (err != header_size) {
		if (err >= 0)
			err = -EIO;
		return err;
	}
	err = decode_header(tconn, tconn->data.rbuf, &pi);
	if (err)
		return err;
	return pi.cmd;
P
Philipp Reisner 已提交
788 789 790 791 792 793
}

/**
 * drbd_socket_okay() - Free the socket if its connection is not okay
 * @sock:	pointer to the pointer to the socket.
 */
794
static int drbd_socket_okay(struct socket **sock)
P
Philipp Reisner 已提交
795 796 797 798 799
{
	int rr;
	char tb[4];

	if (!*sock)
800
		return false;
P
Philipp Reisner 已提交
801

802
	rr = drbd_recv_short(*sock, tb, 4, MSG_DONTWAIT | MSG_PEEK);
P
Philipp Reisner 已提交
803 804

	if (rr > 0 || rr == -EAGAIN) {
805
		return true;
P
Philipp Reisner 已提交
806 807 808
	} else {
		sock_release(*sock);
		*sock = NULL;
809
		return false;
P
Philipp Reisner 已提交
810 811
	}
}
812 813 814
/* Gets called if a connection is established, or if a new minor gets created
   in a connection */
int drbd_connected(int vnr, void *p, void *data)
815 816
{
	struct drbd_conf *mdev = (struct drbd_conf *)p;
817
	int err;
818 819 820 821

	atomic_set(&mdev->packet_seq, 0);
	mdev->peer_seq = 0;

822 823 824 825
	mdev->state_mutex = mdev->tconn->agreed_pro_version < 100 ?
		&mdev->tconn->cstate_mutex :
		&mdev->own_state_mutex;

826 827 828 829 830 831 832
	err = drbd_send_sync_param(mdev);
	if (!err)
		err = drbd_send_sizes(mdev, 0, 0);
	if (!err)
		err = drbd_send_uuids(mdev);
	if (!err)
		err = drbd_send_state(mdev);
833 834
	clear_bit(USE_DEGR_WFC_T, &mdev->flags);
	clear_bit(RESIZE_PENDING, &mdev->flags);
835
	mod_timer(&mdev->request_timer, jiffies + HZ); /* just start it here. */
836
	return err;
837 838
}

P
Philipp Reisner 已提交
839 840 841 842 843 844 845 846
/*
 * 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...
 */
847
static int drbd_connect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
848
{
849
	struct socket *sock, *msock;
850 851
	struct net_conf *nc;
	int timeout, try, h, ok;
P
Philipp Reisner 已提交
852

853
	if (conn_request_state(tconn, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS)
P
Philipp Reisner 已提交
854 855
		return -2;

856
	clear_bit(DISCARD_CONCURRENT, &tconn->flags);
857 858 859

	/* Assume that the peer only understands protocol 80 until we know better.  */
	tconn->agreed_pro_version = 80;
P
Philipp Reisner 已提交
860 861

	do {
862 863
		struct socket *s;

P
Philipp Reisner 已提交
864 865
		for (try = 0;;) {
			/* 3 tries, this should take less than a second! */
866
			s = drbd_try_connect(tconn);
P
Philipp Reisner 已提交
867 868 869
			if (s || ++try >= 3)
				break;
			/* give the other side time to call bind() & listen() */
870
			schedule_timeout_interruptible(HZ / 10);
P
Philipp Reisner 已提交
871 872 873
		}

		if (s) {
874 875
			if (!tconn->data.socket) {
				tconn->data.socket = s;
876
				send_first_packet(tconn, &tconn->data, P_INITIAL_DATA);
877 878
			} else if (!tconn->meta.socket) {
				tconn->meta.socket = s;
879
				send_first_packet(tconn, &tconn->meta, P_INITIAL_META);
P
Philipp Reisner 已提交
880
			} else {
881
				conn_err(tconn, "Logic error in drbd_connect()\n");
P
Philipp Reisner 已提交
882 883 884 885
				goto out_release_sockets;
			}
		}

886
		if (tconn->data.socket && tconn->meta.socket) {
887
			schedule_timeout_interruptible(tconn->net_conf->ping_timeo*HZ/10);
888 889
			ok = drbd_socket_okay(&tconn->data.socket);
			ok = drbd_socket_okay(&tconn->meta.socket) && ok;
P
Philipp Reisner 已提交
890 891 892 893 894
			if (ok)
				break;
		}

retry:
895
		s = drbd_wait_for_connect(tconn);
P
Philipp Reisner 已提交
896
		if (s) {
897
			try = receive_first_packet(tconn, s);
898 899
			drbd_socket_okay(&tconn->data.socket);
			drbd_socket_okay(&tconn->meta.socket);
P
Philipp Reisner 已提交
900
			switch (try) {
901
			case P_INITIAL_DATA:
902
				if (tconn->data.socket) {
903
					conn_warn(tconn, "initial packet S crossed\n");
904
					sock_release(tconn->data.socket);
P
Philipp Reisner 已提交
905
				}
906
				tconn->data.socket = s;
P
Philipp Reisner 已提交
907
				break;
908
			case P_INITIAL_META:
909
				if (tconn->meta.socket) {
910
					conn_warn(tconn, "initial packet M crossed\n");
911
					sock_release(tconn->meta.socket);
P
Philipp Reisner 已提交
912
				}
913
				tconn->meta.socket = s;
914
				set_bit(DISCARD_CONCURRENT, &tconn->flags);
P
Philipp Reisner 已提交
915 916
				break;
			default:
917
				conn_warn(tconn, "Error receiving initial packet\n");
P
Philipp Reisner 已提交
918 919 920 921 922 923
				sock_release(s);
				if (random32() & 1)
					goto retry;
			}
		}

924
		if (tconn->cstate <= C_DISCONNECTING)
P
Philipp Reisner 已提交
925 926 927 928
			goto out_release_sockets;
		if (signal_pending(current)) {
			flush_signals(current);
			smp_rmb();
929
			if (get_t_state(&tconn->receiver) == EXITING)
P
Philipp Reisner 已提交
930 931 932
				goto out_release_sockets;
		}

933 934 935
		if (tconn->data.socket && &tconn->meta.socket) {
			ok = drbd_socket_okay(&tconn->data.socket);
			ok = drbd_socket_okay(&tconn->meta.socket) && ok;
P
Philipp Reisner 已提交
936 937 938 939 940
			if (ok)
				break;
		}
	} while (1);

941 942 943
	sock  = tconn->data.socket;
	msock = tconn->meta.socket;

P
Philipp Reisner 已提交
944 945 946 947 948 949 950 951 952 953
	msock->sk->sk_reuse = 1; /* SO_REUSEADDR */
	sock->sk->sk_reuse = 1; /* SO_REUSEADDR */

	sock->sk->sk_allocation = GFP_NOIO;
	msock->sk->sk_allocation = GFP_NOIO;

	sock->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK;
	msock->sk->sk_priority = TC_PRIO_INTERACTIVE;

	/* NOT YET ...
954
	 * sock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10;
P
Philipp Reisner 已提交
955
	 * sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
956
	 * first set it to the P_CONNECTION_FEATURES timeout,
P
Philipp Reisner 已提交
957
	 * which we set to 4x the configured ping_timeout. */
958 959 960
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);

P
Philipp Reisner 已提交
961
	sock->sk->sk_sndtimeo =
962 963 964 965 966
	sock->sk->sk_rcvtimeo = nc->ping_timeo*4*HZ/10;

	msock->sk->sk_rcvtimeo = nc->ping_int*HZ;
	timeout = nc->timeout * HZ / 10;
	rcu_read_unlock();
P
Philipp Reisner 已提交
967

968
	msock->sk->sk_sndtimeo = timeout;
P
Philipp Reisner 已提交
969 970

	/* we don't want delays.
L
Lucas De Marchi 已提交
971
	 * we use TCP_CORK where appropriate, though */
P
Philipp Reisner 已提交
972 973 974
	drbd_tcp_nodelay(sock);
	drbd_tcp_nodelay(msock);

975
	tconn->last_received = jiffies;
P
Philipp Reisner 已提交
976

977
	h = drbd_do_features(tconn);
P
Philipp Reisner 已提交
978 979 980
	if (h <= 0)
		return h;

981
	if (tconn->cram_hmac_tfm) {
P
Philipp Reisner 已提交
982
		/* drbd_request_state(mdev, NS(conn, WFAuth)); */
983
		switch (drbd_do_auth(tconn)) {
984
		case -1:
985
			conn_err(tconn, "Authentication of peer failed\n");
P
Philipp Reisner 已提交
986
			return -1;
987
		case 0:
988
			conn_err(tconn, "Authentication of peer failed, trying again.\n");
989
			return 0;
P
Philipp Reisner 已提交
990 991 992
		}
	}

993
	if (conn_request_state(tconn, NS(conn, C_WF_REPORT_PARAMS), CS_VERBOSE) < SS_SUCCESS)
P
Philipp Reisner 已提交
994 995
		return 0;

996
	sock->sk->sk_sndtimeo = timeout;
P
Philipp Reisner 已提交
997 998
	sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;

999
	drbd_thread_start(&tconn->asender);
P
Philipp Reisner 已提交
1000

1001
	if (drbd_send_protocol(tconn) == -EOPNOTSUPP)
1002
		return -1;
P
Philipp Reisner 已提交
1003

1004 1005 1006 1007
	down_read(&drbd_cfg_rwsem);
	h = !idr_for_each(&tconn->volumes, drbd_connected, tconn);
	up_read(&drbd_cfg_rwsem);
	return h;
P
Philipp Reisner 已提交
1008 1009

out_release_sockets:
1010 1011 1012 1013 1014 1015 1016 1017
	if (tconn->data.socket) {
		sock_release(tconn->data.socket);
		tconn->data.socket = NULL;
	}
	if (tconn->meta.socket) {
		sock_release(tconn->meta.socket);
		tconn->meta.socket = NULL;
	}
P
Philipp Reisner 已提交
1018 1019 1020
	return -1;
}

1021
static int decode_header(struct drbd_tconn *tconn, void *header, struct packet_info *pi)
P
Philipp Reisner 已提交
1022
{
1023 1024
	unsigned int header_size = drbd_header_size(tconn);

1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
	if (header_size == sizeof(struct p_header100) &&
	    *(__be32 *)header == cpu_to_be32(DRBD_MAGIC_100)) {
		struct p_header100 *h = header;
		if (h->pad != 0) {
			conn_err(tconn, "Header padding is not zero\n");
			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)) {
1037 1038
		struct p_header95 *h = header;
		pi->cmd = be16_to_cpu(h->command);
1039 1040
		pi->size = be32_to_cpu(h->length);
		pi->vnr = 0;
1041 1042 1043 1044 1045
	} 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);
1046
		pi->vnr = 0;
1047
	} else {
1048 1049 1050
		conn_err(tconn, "Wrong magic value 0x%08x in protocol version %d\n",
			 be32_to_cpu(*(__be32 *)header),
			 tconn->agreed_pro_version);
1051
		return -EINVAL;
P
Philipp Reisner 已提交
1052
	}
1053
	pi->data = header + header_size;
1054
	return 0;
1055 1056
}

1057
static int drbd_recv_header(struct drbd_tconn *tconn, struct packet_info *pi)
1058
{
1059
	void *buffer = tconn->data.rbuf;
1060
	int err;
1061

1062
	err = drbd_recv_all_warn(tconn, buffer, drbd_header_size(tconn));
1063
	if (err)
1064
		return err;
1065

1066
	err = decode_header(tconn, buffer, pi);
1067
	tconn->last_received = jiffies;
P
Philipp Reisner 已提交
1068

1069
	return err;
P
Philipp Reisner 已提交
1070 1071
}

1072
static void drbd_flush(struct drbd_conf *mdev)
P
Philipp Reisner 已提交
1073 1074 1075 1076
{
	int rv;

	if (mdev->write_ordering >= WO_bdev_flush && get_ldev(mdev)) {
1077
		rv = blkdev_issue_flush(mdev->ldev->backing_bdev, GFP_KERNEL,
1078
					NULL);
P
Philipp Reisner 已提交
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
		if (rv) {
			dev_err(DEV, "local disk flush failed with status %d\n", rv);
			/* would rather check on EOPNOTSUPP, but that is not reliable.
			 * don't try again for ANY return value != 0
			 * if (rv == -EOPNOTSUPP) */
			drbd_bump_write_ordering(mdev, WO_drain_io);
		}
		put_ldev(mdev);
	}
}

/**
 * drbd_may_finish_epoch() - Applies an epoch_event to the epoch's state, eventually finishes it.
 * @mdev:	DRBD device.
 * @epoch:	Epoch object.
 * @ev:		Epoch event.
 */
static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev,
					       struct drbd_epoch *epoch,
					       enum epoch_event ev)
{
1100
	int epoch_size;
P
Philipp Reisner 已提交
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123
	struct drbd_epoch *next_epoch;
	enum finish_epoch rv = FE_STILL_LIVE;

	spin_lock(&mdev->epoch_lock);
	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 &&
1124
		    test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags)) {
P
Philipp Reisner 已提交
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
			if (!(ev & EV_CLEANUP)) {
				spin_unlock(&mdev->epoch_lock);
				drbd_send_b_ack(mdev, epoch->barrier_nr, epoch_size);
				spin_lock(&mdev->epoch_lock);
			}
			dec_unacked(mdev);

			if (mdev->current_epoch != epoch) {
				next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list);
				list_del(&epoch->list);
				ev = EV_BECAME_LAST | (ev & EV_CLEANUP);
				mdev->epochs--;
				kfree(epoch);

				if (rv == FE_STILL_LIVE)
					rv = FE_DESTROYED;
			} else {
				epoch->flags = 0;
				atomic_set(&epoch->epoch_size, 0);
1144
				/* atomic_set(&epoch->active, 0); is already zero */
P
Philipp Reisner 已提交
1145 1146
				if (rv == FE_STILL_LIVE)
					rv = FE_RECYCLED;
1147
				wake_up(&mdev->ee_wait);
P
Philipp Reisner 已提交
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
			}
		}

		if (!next_epoch)
			break;

		epoch = next_epoch;
	} while (1);

	spin_unlock(&mdev->epoch_lock);

	return rv;
}

/**
 * drbd_bump_write_ordering() - Fall back to an other write ordering method
 * @mdev:	DRBD device.
 * @wo:		Write ordering method to try.
 */
void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo) __must_hold(local)
{
	enum write_ordering_e pwo;
	static char *write_ordering_str[] = {
		[WO_none] = "none",
		[WO_drain_io] = "drain",
		[WO_bdev_flush] = "flush",
	};

	pwo = mdev->write_ordering;
	wo = min(pwo, wo);
	if (wo == WO_bdev_flush && mdev->ldev->dc.no_disk_flush)
		wo = WO_drain_io;
	if (wo == WO_drain_io && mdev->ldev->dc.no_disk_drain)
		wo = WO_none;
	mdev->write_ordering = wo;
1183
	if (pwo != mdev->write_ordering || wo == WO_bdev_flush)
P
Philipp Reisner 已提交
1184 1185 1186
		dev_info(DEV, "Method to ensure write ordering: %s\n", write_ordering_str[mdev->write_ordering]);
}

1187
/**
1188
 * drbd_submit_peer_request()
1189
 * @mdev:	DRBD device.
1190
 * @peer_req:	peer request
1191
 * @rw:		flag field, see bio->bi_rw
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201
 *
 * 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.
1202 1203
 */
/* TODO allocate from our own bio_set. */
1204 1205 1206
int drbd_submit_peer_request(struct drbd_conf *mdev,
			     struct drbd_peer_request *peer_req,
			     const unsigned rw, const int fault_type)
1207 1208 1209
{
	struct bio *bios = NULL;
	struct bio *bio;
1210 1211 1212
	struct page *page = peer_req->pages;
	sector_t sector = peer_req->i.sector;
	unsigned ds = peer_req->i.size;
1213 1214
	unsigned n_bios = 0;
	unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT;
1215
	int err = -ENOMEM;
1216 1217 1218 1219

	/* 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
1220 1221 1222 1223 1224
	 * 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.
	 */
1225 1226 1227 1228 1229 1230
next_bio:
	bio = bio_alloc(GFP_NOIO, nr_pages);
	if (!bio) {
		dev_err(DEV, "submit_ee: Allocation of a bio failed\n");
		goto fail;
	}
1231
	/* > peer_req->i.sector, unless this is the first bio */
1232 1233 1234
	bio->bi_sector = sector;
	bio->bi_bdev = mdev->ldev->backing_bdev;
	bio->bi_rw = rw;
1235
	bio->bi_private = peer_req;
1236
	bio->bi_end_io = drbd_peer_request_endio;
1237 1238 1239 1240 1241 1242 1243 1244

	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)) {
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
			/* 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) {
				dev_err(DEV,
					"bio_add_page failed for len=%u, "
					"bi_vcnt=0 (bi_sector=%llu)\n",
					len, (unsigned long long)bio->bi_sector);
				err = -ENOSPC;
				goto fail;
			}
1256 1257 1258 1259 1260 1261 1262 1263 1264
			goto next_bio;
		}
		ds -= len;
		sector += len >> 9;
		--nr_pages;
	}
	D_ASSERT(page == NULL);
	D_ASSERT(ds == 0);

1265
	atomic_set(&peer_req->pending_bios, n_bios);
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
	do {
		bio = bios;
		bios = bios->bi_next;
		bio->bi_next = NULL;

		drbd_generic_make_request(mdev, fault_type, bio);
	} while (bios);
	return 0;

fail:
	while (bios) {
		bio = bios;
		bios = bios->bi_next;
		bio_put(bio);
	}
1281
	return err;
1282 1283
}

1284
static void drbd_remove_epoch_entry_interval(struct drbd_conf *mdev,
1285
					     struct drbd_peer_request *peer_req)
1286
{
1287
	struct drbd_interval *i = &peer_req->i;
1288 1289 1290 1291

	drbd_remove_interval(&mdev->write_requests, i);
	drbd_clear_interval(i);

A
Andreas Gruenbacher 已提交
1292
	/* Wake up any processes waiting for this peer request to complete.  */
1293 1294 1295 1296
	if (i->waiting)
		wake_up(&mdev->misc_wait);
}

1297
static int receive_Barrier(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
1298
{
1299
	struct drbd_conf *mdev;
1300
	int rv;
1301
	struct p_barrier *p = pi->data;
P
Philipp Reisner 已提交
1302 1303
	struct drbd_epoch *epoch;

1304 1305 1306 1307
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;

P
Philipp Reisner 已提交
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
	inc_unacked(mdev);

	mdev->current_epoch->barrier_nr = p->barrier;
	rv = drbd_may_finish_epoch(mdev, mdev->current_epoch, EV_GOT_BARRIER_NR);

	/* 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. */
	switch (mdev->write_ordering) {
	case WO_none:
		if (rv == FE_RECYCLED)
1321
			return 0;
1322 1323 1324 1325 1326 1327 1328 1329 1330

		/* 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
			dev_warn(DEV, "Allocation of an epoch failed, slowing down\n");
			/* Fall through */
P
Philipp Reisner 已提交
1331 1332 1333 1334

	case WO_bdev_flush:
	case WO_drain_io:
		drbd_wait_ee_list_empty(mdev, &mdev->active_ee);
1335 1336 1337 1338 1339 1340
		drbd_flush(mdev);

		if (atomic_read(&mdev->current_epoch->epoch_size)) {
			epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO);
			if (epoch)
				break;
P
Philipp Reisner 已提交
1341 1342
		}

1343 1344 1345 1346 1347
		epoch = mdev->current_epoch;
		wait_event(mdev->ee_wait, atomic_read(&epoch->epoch_size) == 0);

		D_ASSERT(atomic_read(&epoch->active) == 0);
		D_ASSERT(epoch->flags == 0);
P
Philipp Reisner 已提交
1348

1349
		return 0;
1350 1351
	default:
		dev_err(DEV, "Strangeness in mdev->write_ordering %d\n", mdev->write_ordering);
1352
		return -EIO;
P
Philipp Reisner 已提交
1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
	}

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

	spin_lock(&mdev->epoch_lock);
	if (atomic_read(&mdev->current_epoch->epoch_size)) {
		list_add(&epoch->list, &mdev->current_epoch->list);
		mdev->current_epoch = epoch;
		mdev->epochs++;
	} else {
		/* The current_epoch got recycled while we allocated this one... */
		kfree(epoch);
	}
	spin_unlock(&mdev->epoch_lock);

1370
	return 0;
P
Philipp Reisner 已提交
1371 1372 1373 1374
}

/* used from receive_RSDataReply (recv_resync_read)
 * and from receive_Data */
1375 1376 1377
static struct drbd_peer_request *
read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector,
	      int data_size) __must_hold(local)
P
Philipp Reisner 已提交
1378
{
1379
	const sector_t capacity = drbd_get_capacity(mdev->this_bdev);
1380
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
1381
	struct page *page;
1382
	int dgs, ds, err;
1383 1384
	void *dig_in = mdev->tconn->int_dig_in;
	void *dig_vv = mdev->tconn->int_dig_vv;
1385
	unsigned long *data;
P
Philipp Reisner 已提交
1386

1387 1388 1389
	dgs = 0;
	if (mdev->tconn->peer_integrity_tfm) {
		dgs = crypto_hash_digestsize(mdev->tconn->peer_integrity_tfm);
1390 1391 1392 1393
		/*
		 * FIXME: Receive the incoming digest into the receive buffer
		 *	  here, together with its struct p_data?
		 */
1394 1395
		err = drbd_recv_all_warn(mdev->tconn, dig_in, dgs);
		if (err)
P
Philipp Reisner 已提交
1396
			return NULL;
1397
		data_size -= dgs;
P
Philipp Reisner 已提交
1398 1399
	}

1400 1401 1402 1403 1404 1405
	if (!expect(data_size != 0))
		return NULL;
	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
1406

1407 1408 1409
	/* even though we trust out peer,
	 * we sometimes have to double check. */
	if (sector + (data_size>>9) > capacity) {
1410 1411
		dev_err(DEV, "request from peer beyond end of local disk: "
			"capacity: %llus < sector: %llus + size: %u\n",
1412 1413 1414 1415 1416
			(unsigned long long)capacity,
			(unsigned long long)sector, data_size);
		return NULL;
	}

P
Philipp Reisner 已提交
1417 1418 1419
	/* 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.  */
1420
	peer_req = drbd_alloc_peer_req(mdev, id, sector, data_size, GFP_NOIO);
1421
	if (!peer_req)
P
Philipp Reisner 已提交
1422
		return NULL;
1423

P
Philipp Reisner 已提交
1424
	ds = data_size;
1425
	page = peer_req->pages;
1426 1427
	page_chain_for_each(page) {
		unsigned len = min_t(int, ds, PAGE_SIZE);
1428
		data = kmap(page);
1429
		err = drbd_recv_all_warn(mdev->tconn, data, len);
1430
		if (drbd_insert_fault(mdev, DRBD_FAULT_RECEIVE)) {
1431 1432 1433
			dev_err(DEV, "Fault injection: Corrupting data on receive\n");
			data[0] = data[0] ^ (unsigned long)-1;
		}
P
Philipp Reisner 已提交
1434
		kunmap(page);
1435
		if (err) {
1436
			drbd_free_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
1437 1438
			return NULL;
		}
1439
		ds -= len;
P
Philipp Reisner 已提交
1440 1441 1442
	}

	if (dgs) {
1443
		drbd_csum_ee(mdev, mdev->tconn->peer_integrity_tfm, peer_req, dig_vv);
P
Philipp Reisner 已提交
1444
		if (memcmp(dig_in, dig_vv, dgs)) {
1445 1446
			dev_err(DEV, "Digest integrity check FAILED: %llus +%u\n",
				(unsigned long long)sector, data_size);
1447
			drbd_free_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
1448 1449 1450 1451
			return NULL;
		}
	}
	mdev->recv_cnt += data_size>>9;
1452
	return peer_req;
P
Philipp Reisner 已提交
1453 1454 1455 1456 1457 1458 1459 1460
}

/* drbd_drain_block() just takes a data block
 * out of the socket input buffer, and discards it.
 */
static int drbd_drain_block(struct drbd_conf *mdev, int data_size)
{
	struct page *page;
1461
	int err = 0;
P
Philipp Reisner 已提交
1462 1463
	void *data;

1464
	if (!data_size)
1465
		return 0;
1466

1467
	page = drbd_alloc_pages(mdev, 1, 1);
P
Philipp Reisner 已提交
1468 1469 1470

	data = kmap(page);
	while (data_size) {
1471 1472
		unsigned int len = min_t(int, data_size, PAGE_SIZE);

1473 1474
		err = drbd_recv_all_warn(mdev->tconn, data, len);
		if (err)
P
Philipp Reisner 已提交
1475
			break;
1476
		data_size -= len;
P
Philipp Reisner 已提交
1477 1478
	}
	kunmap(page);
1479
	drbd_free_pages(mdev, page, 0);
1480
	return err;
P
Philipp Reisner 已提交
1481 1482 1483 1484 1485 1486 1487
}

static int recv_dless_read(struct drbd_conf *mdev, struct drbd_request *req,
			   sector_t sector, int data_size)
{
	struct bio_vec *bvec;
	struct bio *bio;
1488
	int dgs, err, i, expect;
1489 1490
	void *dig_in = mdev->tconn->int_dig_in;
	void *dig_vv = mdev->tconn->int_dig_vv;
P
Philipp Reisner 已提交
1491

1492 1493 1494
	dgs = 0;
	if (mdev->tconn->peer_integrity_tfm) {
		dgs = crypto_hash_digestsize(mdev->tconn->peer_integrity_tfm);
1495 1496 1497
		err = drbd_recv_all_warn(mdev->tconn, dig_in, dgs);
		if (err)
			return err;
1498
		data_size -= dgs;
P
Philipp Reisner 已提交
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508
	}

	/* optimistically update recv_cnt.  if receiving fails below,
	 * we disconnect anyways, and counters will be reset. */
	mdev->recv_cnt += data_size>>9;

	bio = req->master_bio;
	D_ASSERT(sector == bio->bi_sector);

	bio_for_each_segment(bvec, bio, i) {
1509
		void *mapped = kmap(bvec->bv_page) + bvec->bv_offset;
P
Philipp Reisner 已提交
1510
		expect = min_t(int, data_size, bvec->bv_len);
1511
		err = drbd_recv_all_warn(mdev->tconn, mapped, expect);
P
Philipp Reisner 已提交
1512
		kunmap(bvec->bv_page);
1513 1514 1515
		if (err)
			return err;
		data_size -= expect;
P
Philipp Reisner 已提交
1516 1517 1518
	}

	if (dgs) {
1519
		drbd_csum_bio(mdev, mdev->tconn->peer_integrity_tfm, bio, dig_vv);
P
Philipp Reisner 已提交
1520 1521
		if (memcmp(dig_in, dig_vv, dgs)) {
			dev_err(DEV, "Digest integrity check FAILED. Broken NICs?\n");
1522
			return -EINVAL;
P
Philipp Reisner 已提交
1523 1524 1525 1526
		}
	}

	D_ASSERT(data_size == 0);
1527
	return 0;
P
Philipp Reisner 已提交
1528 1529
}

1530 1531 1532 1533
/*
 * e_end_resync_block() is called in asender context via
 * drbd_finish_peer_reqs().
 */
1534
static int e_end_resync_block(struct drbd_work *w, int unused)
P
Philipp Reisner 已提交
1535
{
1536 1537
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
1538
	struct drbd_conf *mdev = w->mdev;
1539
	sector_t sector = peer_req->i.sector;
1540
	int err;
P
Philipp Reisner 已提交
1541

1542
	D_ASSERT(drbd_interval_empty(&peer_req->i));
P
Philipp Reisner 已提交
1543

1544 1545
	if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
		drbd_set_in_sync(mdev, sector, peer_req->i.size);
1546
		err = drbd_send_ack(mdev, P_RS_WRITE_ACK, peer_req);
P
Philipp Reisner 已提交
1547 1548
	} else {
		/* Record failure to sync */
1549
		drbd_rs_failed_io(mdev, sector, peer_req->i.size);
P
Philipp Reisner 已提交
1550

1551
		err  = drbd_send_ack(mdev, P_NEG_ACK, peer_req);
P
Philipp Reisner 已提交
1552 1553 1554
	}
	dec_unacked(mdev);

1555
	return err;
P
Philipp Reisner 已提交
1556 1557 1558 1559
}

static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_size) __releases(local)
{
1560
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
1561

1562 1563
	peer_req = read_in_block(mdev, ID_SYNCER, sector, data_size);
	if (!peer_req)
1564
		goto fail;
P
Philipp Reisner 已提交
1565 1566 1567 1568 1569 1570 1571

	dec_rs_pending(mdev);

	inc_unacked(mdev);
	/* corresponding dec_unacked() in e_end_resync_block()
	 * respective _drbd_clear_done_ee */

1572
	peer_req->w.cb = e_end_resync_block;
1573

1574
	spin_lock_irq(&mdev->tconn->req_lock);
1575
	list_add(&peer_req->w.list, &mdev->sync_ee);
1576
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
1577

1578
	atomic_add(data_size >> 9, &mdev->rs_sect_ev);
1579
	if (drbd_submit_peer_request(mdev, peer_req, WRITE, DRBD_FAULT_RS_WR) == 0)
1580
		return 0;
P
Philipp Reisner 已提交
1581

1582 1583
	/* don't care for the reason here */
	dev_err(DEV, "submit failed, triggering re-connect\n");
1584
	spin_lock_irq(&mdev->tconn->req_lock);
1585
	list_del(&peer_req->w.list);
1586
	spin_unlock_irq(&mdev->tconn->req_lock);
1587

1588
	drbd_free_peer_req(mdev, peer_req);
1589 1590
fail:
	put_ldev(mdev);
1591
	return -EIO;
P
Philipp Reisner 已提交
1592 1593
}

1594
static struct drbd_request *
1595 1596
find_request(struct drbd_conf *mdev, struct rb_root *root, u64 id,
	     sector_t sector, bool missing_ok, const char *func)
1597 1598 1599
{
	struct drbd_request *req;

1600 1601
	/* Request object according to our peer */
	req = (struct drbd_request *)(unsigned long)id;
1602
	if (drbd_contains_interval(root, sector, &req->i) && req->i.local)
1603
		return req;
1604 1605 1606 1607
	if (!missing_ok) {
		dev_err(DEV, "%s: failed to find request %lu, sector %llus\n", func,
			(unsigned long)id, (unsigned long long)sector);
	}
1608 1609 1610
	return NULL;
}

1611
static int receive_DataReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
1612
{
1613
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
1614 1615
	struct drbd_request *req;
	sector_t sector;
1616
	int err;
1617
	struct p_data *p = pi->data;
1618 1619 1620 1621

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
1622 1623 1624

	sector = be64_to_cpu(p->sector);

1625
	spin_lock_irq(&mdev->tconn->req_lock);
1626
	req = find_request(mdev, &mdev->read_requests, p->block_id, sector, false, __func__);
1627
	spin_unlock_irq(&mdev->tconn->req_lock);
1628
	if (unlikely(!req))
1629
		return -EIO;
P
Philipp Reisner 已提交
1630

B
Bart Van Assche 已提交
1631
	/* hlist_del(&req->collision) is done in _req_may_be_done, to avoid
P
Philipp Reisner 已提交
1632 1633
	 * special casing it there for the various failure cases.
	 * still no race with drbd_fail_pending_reads */
1634
	err = recv_dless_read(mdev, req, sector, pi->size);
1635
	if (!err)
1636
		req_mod(req, DATA_RECEIVED);
P
Philipp Reisner 已提交
1637 1638 1639 1640
	/* else: nothing. handled from drbd_disconnect...
	 * I don't think we may complete this just yet
	 * in case we are "on-disconnect: freeze" */

1641
	return err;
P
Philipp Reisner 已提交
1642 1643
}

1644
static int receive_RSDataReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
1645
{
1646
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
1647
	sector_t sector;
1648
	int err;
1649
	struct p_data *p = pi->data;
1650 1651 1652 1653

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
1654 1655 1656 1657 1658 1659 1660

	sector = be64_to_cpu(p->sector);
	D_ASSERT(p->block_id == ID_SYNCER);

	if (get_ldev(mdev)) {
		/* data is submitted to disk within recv_resync_read.
		 * corresponding put_ldev done below on error,
1661
		 * or in drbd_peer_request_endio. */
1662
		err = recv_resync_read(mdev, sector, pi->size);
P
Philipp Reisner 已提交
1663 1664 1665 1666
	} else {
		if (__ratelimit(&drbd_ratelimit_state))
			dev_err(DEV, "Can not write resync data to local disk.\n");

1667
		err = drbd_drain_block(mdev, pi->size);
P
Philipp Reisner 已提交
1668

1669
		drbd_send_ack_dp(mdev, P_NEG_ACK, p, pi->size);
P
Philipp Reisner 已提交
1670 1671
	}

1672
	atomic_add(pi->size >> 9, &mdev->rs_sect_in);
1673

1674
	return err;
P
Philipp Reisner 已提交
1675 1676
}

1677
static int w_restart_write(struct drbd_work *w, int cancel)
1678 1679 1680 1681 1682 1683 1684 1685 1686 1687
{
	struct drbd_request *req = container_of(w, struct drbd_request, w);
	struct drbd_conf *mdev = w->mdev;
	struct bio *bio;
	unsigned long start_time;
	unsigned long flags;

	spin_lock_irqsave(&mdev->tconn->req_lock, flags);
	if (!expect(req->rq_state & RQ_POSTPONED)) {
		spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
1688
		return -EIO;
1689 1690 1691 1692 1693 1694 1695 1696 1697
	}
	bio = req->master_bio;
	start_time = req->start_time;
	/* Postponed requests will not have their master_bio completed!  */
	__req_mod(req, DISCARD_WRITE, NULL);
	spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);

	while (__drbd_make_request(mdev, bio, start_time))
		/* retry */ ;
1698
	return 0;
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721
}

static void restart_conflicting_writes(struct drbd_conf *mdev,
				       sector_t sector, int size)
{
	struct drbd_interval *i;
	struct drbd_request *req;

	drbd_for_each_overlap(i, &mdev->write_requests, sector, size) {
		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;
		if (expect(list_empty(&req->w.list))) {
			req->w.mdev = mdev;
			req->w.cb = w_restart_write;
			drbd_queue_work(&mdev->tconn->data.work, &req->w);
		}
	}
}

1722 1723
/*
 * e_end_block() is called in asender context via drbd_finish_peer_reqs().
P
Philipp Reisner 已提交
1724
 */
1725
static int e_end_block(struct drbd_work *w, int cancel)
P
Philipp Reisner 已提交
1726
{
1727 1728
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
1729
	struct drbd_conf *mdev = w->mdev;
1730
	sector_t sector = peer_req->i.sector;
1731
	int err = 0, pcmd;
P
Philipp Reisner 已提交
1732

1733
	if (peer_req->flags & EE_SEND_WRITE_ACK) {
1734
		if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
P
Philipp Reisner 已提交
1735 1736
			pcmd = (mdev->state.conn >= C_SYNC_SOURCE &&
				mdev->state.conn <= C_PAUSED_SYNC_T &&
1737
				peer_req->flags & EE_MAY_SET_IN_SYNC) ?
P
Philipp Reisner 已提交
1738
				P_RS_WRITE_ACK : P_WRITE_ACK;
1739
			err = drbd_send_ack(mdev, pcmd, peer_req);
P
Philipp Reisner 已提交
1740
			if (pcmd == P_RS_WRITE_ACK)
1741
				drbd_set_in_sync(mdev, sector, peer_req->i.size);
P
Philipp Reisner 已提交
1742
		} else {
1743
			err = drbd_send_ack(mdev, P_NEG_ACK, peer_req);
P
Philipp Reisner 已提交
1744 1745 1746 1747 1748 1749 1750
			/* we expect it to be marked out of sync anyways...
			 * maybe assert this?  */
		}
		dec_unacked(mdev);
	}
	/* we delete from the conflict detection hash _after_ we sent out the
	 * P_WRITE_ACK / P_NEG_ACK, to get the sequence number right.  */
1751
	if (peer_req->flags & EE_IN_INTERVAL_TREE) {
1752
		spin_lock_irq(&mdev->tconn->req_lock);
1753 1754
		D_ASSERT(!drbd_interval_empty(&peer_req->i));
		drbd_remove_epoch_entry_interval(mdev, peer_req);
1755 1756
		if (peer_req->flags & EE_RESTART_REQUESTS)
			restart_conflicting_writes(mdev, sector, peer_req->i.size);
1757
		spin_unlock_irq(&mdev->tconn->req_lock);
1758
	} else
1759
		D_ASSERT(drbd_interval_empty(&peer_req->i));
P
Philipp Reisner 已提交
1760

1761
	drbd_may_finish_epoch(mdev, peer_req->epoch, EV_PUT + (cancel ? EV_CLEANUP : 0));
P
Philipp Reisner 已提交
1762

1763
	return err;
P
Philipp Reisner 已提交
1764 1765
}

1766
static int e_send_ack(struct drbd_work *w, enum drbd_packet ack)
P
Philipp Reisner 已提交
1767
{
1768
	struct drbd_conf *mdev = w->mdev;
1769 1770
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
1771
	int err;
P
Philipp Reisner 已提交
1772

1773
	err = drbd_send_ack(mdev, ack, peer_req);
P
Philipp Reisner 已提交
1774 1775
	dec_unacked(mdev);

1776
	return err;
P
Philipp Reisner 已提交
1777 1778
}

1779
static int e_send_discard_write(struct drbd_work *w, int unused)
1780 1781 1782 1783
{
	return e_send_ack(w, P_DISCARD_WRITE);
}

1784
static int e_send_retry_write(struct drbd_work *w, int unused)
1785 1786 1787 1788 1789 1790 1791
{
	struct drbd_tconn *tconn = w->mdev->tconn;

	return e_send_ack(w, tconn->agreed_pro_version >= 100 ?
			     P_RETRY_WRITE : P_DISCARD_WRITE);
}

1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806
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;
}

static u32 seq_max(u32 a, u32 b)
{
	return seq_greater(a, b) ? a : b;
}

1807 1808 1809
static bool need_peer_seq(struct drbd_conf *mdev)
{
	struct drbd_tconn *tconn = mdev->tconn;
1810
	int tp;
1811 1812 1813 1814 1815 1816

	/*
	 * We only need to keep track of the last packet_seq number of our peer
	 * if we are in dual-primary mode and we have the discard flag set; see
	 * handle_write_conflicts().
	 */
1817 1818 1819 1820 1821 1822

	rcu_read_lock();
	tp = rcu_dereference(mdev->tconn->net_conf)->two_primaries;
	rcu_read_unlock();

	return tp && test_bit(DISCARD_CONCURRENT, &tconn->flags);
1823 1824
}

1825
static void update_peer_seq(struct drbd_conf *mdev, unsigned int peer_seq)
1826
{
1827
	unsigned int newest_peer_seq;
1828

1829 1830
	if (need_peer_seq(mdev)) {
		spin_lock(&mdev->peer_seq_lock);
1831 1832
		newest_peer_seq = seq_max(mdev->peer_seq, peer_seq);
		mdev->peer_seq = newest_peer_seq;
1833
		spin_unlock(&mdev->peer_seq_lock);
1834 1835
		/* wake up only if we actually changed mdev->peer_seq */
		if (peer_seq == newest_peer_seq)
1836 1837
			wake_up(&mdev->seq_wait);
	}
1838 1839
}

P
Philipp Reisner 已提交
1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
/* 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.
 *
 * In case packet_seq is larger than mdev->peer_seq number, there are
 * outstanding packets on the msock. We wait for them to arrive.
 * In case we are the logically next packet, we update mdev->peer_seq
 * 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). */
1861
static int wait_for_and_update_peer_seq(struct drbd_conf *mdev, const u32 peer_seq)
P
Philipp Reisner 已提交
1862 1863 1864
{
	DEFINE_WAIT(wait);
	long timeout;
1865 1866 1867 1868 1869
	int ret;

	if (!need_peer_seq(mdev))
		return 0;

P
Philipp Reisner 已提交
1870 1871
	spin_lock(&mdev->peer_seq_lock);
	for (;;) {
1872 1873 1874
		if (!seq_greater(peer_seq - 1, mdev->peer_seq)) {
			mdev->peer_seq = seq_max(mdev->peer_seq, peer_seq);
			ret = 0;
P
Philipp Reisner 已提交
1875
			break;
1876
		}
P
Philipp Reisner 已提交
1877 1878 1879 1880
		if (signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}
1881
		prepare_to_wait(&mdev->seq_wait, &wait, TASK_INTERRUPTIBLE);
P
Philipp Reisner 已提交
1882
		spin_unlock(&mdev->peer_seq_lock);
1883 1884 1885
		rcu_read_lock();
		timeout = rcu_dereference(mdev->tconn->net_conf)->ping_timeo*HZ/10;
		rcu_read_unlock();
1886
		timeout = schedule_timeout(timeout);
P
Philipp Reisner 已提交
1887
		spin_lock(&mdev->peer_seq_lock);
1888
		if (!timeout) {
P
Philipp Reisner 已提交
1889
			ret = -ETIMEDOUT;
1890
			dev_err(DEV, "Timed out waiting for missing ack packets; disconnecting\n");
P
Philipp Reisner 已提交
1891 1892 1893 1894
			break;
		}
	}
	spin_unlock(&mdev->peer_seq_lock);
1895
	finish_wait(&mdev->seq_wait, &wait);
P
Philipp Reisner 已提交
1896 1897 1898
	return ret;
}

1899 1900 1901 1902
/* 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. */
static unsigned long wire_flags_to_bio(struct drbd_conf *mdev, u32 dpf)
1903
{
1904 1905 1906 1907
	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);
1908 1909
}

1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042
static void fail_postponed_requests(struct drbd_conf *mdev, sector_t sector,
				    unsigned int size)
{
	struct drbd_interval *i;

    repeat:
	drbd_for_each_overlap(i, &mdev->write_requests, sector, size) {
		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);
		spin_unlock_irq(&mdev->tconn->req_lock);
		if (m.bio)
			complete_master_bio(mdev, &m);
		spin_lock_irq(&mdev->tconn->req_lock);
		goto repeat;
	}
}

static int handle_write_conflicts(struct drbd_conf *mdev,
				  struct drbd_peer_request *peer_req)
{
	struct drbd_tconn *tconn = mdev->tconn;
	bool resolve_conflicts = test_bit(DISCARD_CONCURRENT, &tconn->flags);
	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.
	 */
	drbd_insert_interval(&mdev->write_requests, &peer_req->i);

    repeat:
	drbd_for_each_overlap(i, &mdev->write_requests, sector, size) {
		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.
			 */
			err = drbd_wait_misc(mdev, i);
			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
			 * overlapping request, it can be discarded; otherwise,
			 * it will be retried once all overlapping requests
			 * have completed.
			 */
			bool discard = i->sector <= sector && i->sector +
				       (i->size >> 9) >= sector + (size >> 9);

			if (!equal)
				dev_alert(DEV, "Concurrent writes detected: "
					       "local=%llus +%u, remote=%llus +%u, "
					       "assuming %s came first\n",
					  (unsigned long long)i->sector, i->size,
					  (unsigned long long)sector, size,
					  discard ? "local" : "remote");

			inc_unacked(mdev);
			peer_req->w.cb = discard ? e_send_discard_write :
						   e_send_retry_write;
			list_add_tail(&peer_req->w.list, &mdev->done_ee);
			wake_asender(mdev->tconn);

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

			if (!equal)
				dev_alert(DEV, "Concurrent writes detected: "
					       "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
				 * decide if this request will be discarded or
				 * retried.  Requests that are discarded will
				 * disappear from the write_requests tree.
				 *
				 * In addition, wait for the conflicting
				 * request to finish locally before submitting
				 * the conflicting peer request.
				 */
				err = drbd_wait_misc(mdev, &req->i);
				if (err) {
					_conn_request_state(mdev->tconn,
							    NS(conn, C_TIMEOUT),
							    CS_HARD);
					fail_postponed_requests(mdev, sector, size);
					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)
		drbd_remove_epoch_entry_interval(mdev, peer_req);
	return err;
}

P
Philipp Reisner 已提交
2043
/* mirrored write */
2044
static int receive_Data(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
2045
{
2046
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
2047
	sector_t sector;
2048
	struct drbd_peer_request *peer_req;
2049
	struct p_data *p = pi->data;
2050
	u32 peer_seq = be32_to_cpu(p->seq_num);
P
Philipp Reisner 已提交
2051 2052
	int rw = WRITE;
	u32 dp_flags;
2053
	int err, tp;
P
Philipp Reisner 已提交
2054

2055 2056 2057 2058
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;

2059
	if (!get_ldev(mdev)) {
2060 2061
		int err2;

2062
		err = wait_for_and_update_peer_seq(mdev, peer_seq);
2063
		drbd_send_ack_dp(mdev, P_NEG_ACK, p, pi->size);
P
Philipp Reisner 已提交
2064
		atomic_inc(&mdev->current_epoch->epoch_size);
2065
		err2 = drbd_drain_block(mdev, pi->size);
2066 2067 2068
		if (!err)
			err = err2;
		return err;
P
Philipp Reisner 已提交
2069 2070
	}

2071 2072 2073 2074 2075
	/*
	 * 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
2076 2077

	sector = be64_to_cpu(p->sector);
2078
	peer_req = read_in_block(mdev, p->block_id, sector, pi->size);
2079
	if (!peer_req) {
P
Philipp Reisner 已提交
2080
		put_ldev(mdev);
2081
		return -EIO;
P
Philipp Reisner 已提交
2082 2083
	}

2084
	peer_req->w.cb = e_end_block;
P
Philipp Reisner 已提交
2085

2086 2087 2088 2089
	dp_flags = be32_to_cpu(p->dp_flags);
	rw |= wire_flags_to_bio(mdev, dp_flags);

	if (dp_flags & DP_MAY_SET_IN_SYNC)
2090
		peer_req->flags |= EE_MAY_SET_IN_SYNC;
2091

P
Philipp Reisner 已提交
2092
	spin_lock(&mdev->epoch_lock);
2093 2094 2095
	peer_req->epoch = mdev->current_epoch;
	atomic_inc(&peer_req->epoch->epoch_size);
	atomic_inc(&peer_req->epoch->active);
P
Philipp Reisner 已提交
2096 2097
	spin_unlock(&mdev->epoch_lock);

2098 2099 2100 2101 2102
	rcu_read_lock();
	tp = rcu_dereference(mdev->tconn->net_conf)->two_primaries;
	rcu_read_unlock();
	if (tp) {
		peer_req->flags |= EE_IN_INTERVAL_TREE;
2103 2104
		err = wait_for_and_update_peer_seq(mdev, peer_seq);
		if (err)
P
Philipp Reisner 已提交
2105
			goto out_interrupted;
2106
		spin_lock_irq(&mdev->tconn->req_lock);
2107 2108 2109 2110
		err = handle_write_conflicts(mdev, peer_req);
		if (err) {
			spin_unlock_irq(&mdev->tconn->req_lock);
			if (err == -ENOENT) {
P
Philipp Reisner 已提交
2111
				put_ldev(mdev);
2112
				return 0;
P
Philipp Reisner 已提交
2113
			}
2114
			goto out_interrupted;
P
Philipp Reisner 已提交
2115
		}
2116 2117
	} else
		spin_lock_irq(&mdev->tconn->req_lock);
2118
	list_add(&peer_req->w.list, &mdev->active_ee);
2119
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
2120

2121
	if (mdev->tconn->agreed_pro_version < 100) {
2122 2123
		rcu_read_lock();
		switch (rcu_dereference(mdev->tconn->net_conf)->wire_protocol) {
2124 2125 2126 2127 2128 2129 2130
		case DRBD_PROT_C:
			dp_flags |= DP_SEND_WRITE_ACK;
			break;
		case DRBD_PROT_B:
			dp_flags |= DP_SEND_RECEIVE_ACK;
			break;
		}
2131
		rcu_read_unlock();
2132 2133 2134 2135
	}

	if (dp_flags & DP_SEND_WRITE_ACK) {
		peer_req->flags |= EE_SEND_WRITE_ACK;
P
Philipp Reisner 已提交
2136 2137 2138
		inc_unacked(mdev);
		/* corresponding dec_unacked() in e_end_block()
		 * respective _drbd_clear_done_ee */
2139 2140 2141
	}

	if (dp_flags & DP_SEND_RECEIVE_ACK) {
P
Philipp Reisner 已提交
2142 2143
		/* I really don't like it that the receiver thread
		 * sends on the msock, but anyways */
2144
		drbd_send_ack(mdev, P_RECV_ACK, peer_req);
P
Philipp Reisner 已提交
2145 2146
	}

2147
	if (mdev->state.pdsk < D_INCONSISTENT) {
P
Philipp Reisner 已提交
2148
		/* In case we have the only disk of the cluster, */
2149 2150 2151
		drbd_set_out_of_sync(mdev, peer_req->i.sector, peer_req->i.size);
		peer_req->flags |= EE_CALL_AL_COMPLETE_IO;
		peer_req->flags &= ~EE_MAY_SET_IN_SYNC;
2152
		drbd_al_begin_io(mdev, &peer_req->i);
P
Philipp Reisner 已提交
2153 2154
	}

2155 2156 2157
	err = drbd_submit_peer_request(mdev, peer_req, rw, DRBD_FAULT_DT_WR);
	if (!err)
		return 0;
P
Philipp Reisner 已提交
2158

2159 2160
	/* don't care for the reason here */
	dev_err(DEV, "submit failed, triggering re-connect\n");
2161
	spin_lock_irq(&mdev->tconn->req_lock);
2162 2163
	list_del(&peer_req->w.list);
	drbd_remove_epoch_entry_interval(mdev, peer_req);
2164
	spin_unlock_irq(&mdev->tconn->req_lock);
2165
	if (peer_req->flags & EE_CALL_AL_COMPLETE_IO)
2166
		drbd_al_complete_io(mdev, &peer_req->i);
2167

P
Philipp Reisner 已提交
2168
out_interrupted:
2169
	drbd_may_finish_epoch(mdev, peer_req->epoch, EV_PUT + EV_CLEANUP);
P
Philipp Reisner 已提交
2170
	put_ldev(mdev);
2171
	drbd_free_peer_req(mdev, peer_req);
2172
	return err;
P
Philipp Reisner 已提交
2173 2174
}

2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
/* 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.
 */
2186
int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector)
2187 2188 2189
{
	struct gendisk *disk = mdev->ldev->backing_bdev->bd_contains->bd_disk;
	unsigned long db, dt, dbdt;
2190
	struct lc_element *tmp;
2191 2192 2193 2194
	int curr_events;
	int throttle = 0;

	/* feature disabled? */
2195
	if (mdev->ldev->dc.c_min_rate == 0)
2196 2197
		return 0;

2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
	spin_lock_irq(&mdev->al_lock);
	tmp = lc_find(mdev->resync, BM_SECT_TO_EXT(sector));
	if (tmp) {
		struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce);
		if (test_bit(BME_PRIORITY, &bm_ext->flags)) {
			spin_unlock_irq(&mdev->al_lock);
			return 0;
		}
		/* Do not slow down if app IO is already waiting for this extent */
	}
	spin_unlock_irq(&mdev->al_lock);

2210 2211 2212
	curr_events = (int)part_stat_read(&disk->part0, sectors[0]) +
		      (int)part_stat_read(&disk->part0, sectors[1]) -
			atomic_read(&mdev->rs_sect_ev);
2213

2214 2215 2216 2217 2218 2219 2220 2221
	if (!mdev->rs_last_events || curr_events - mdev->rs_last_events > 64) {
		unsigned long rs_left;
		int i;

		mdev->rs_last_events = curr_events;

		/* sync speed average over the last 2*DRBD_SYNC_MARK_STEP,
		 * approx. */
2222 2223 2224 2225 2226 2227
		i = (mdev->rs_last_mark + DRBD_SYNC_MARKS-1) % DRBD_SYNC_MARKS;

		if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
			rs_left = mdev->ov_left;
		else
			rs_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2228 2229 2230 2231 2232 2233 2234

		dt = ((long)jiffies - (long)mdev->rs_mark_time[i]) / HZ;
		if (!dt)
			dt++;
		db = mdev->rs_mark_left[i] - rs_left;
		dbdt = Bit2KB(db/dt);

2235
		if (dbdt > mdev->ldev->dc.c_min_rate)
2236 2237 2238 2239 2240 2241
			throttle = 1;
	}
	return throttle;
}


2242
static int receive_DataRequest(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
2243
{
2244
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
2245
	sector_t sector;
2246
	sector_t capacity;
2247
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
2248
	struct digest_info *di = NULL;
2249
	int size, verb;
P
Philipp Reisner 已提交
2250
	unsigned int fault_type;
2251
	struct p_block_req *p =	pi->data;
2252 2253 2254 2255 2256

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
	capacity = drbd_get_capacity(mdev->this_bdev);
P
Philipp Reisner 已提交
2257 2258 2259 2260

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

2261
	if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) {
P
Philipp Reisner 已提交
2262 2263
		dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__,
				(unsigned long long)sector, size);
2264
		return -EINVAL;
P
Philipp Reisner 已提交
2265 2266 2267 2268
	}
	if (sector + (size>>9) > capacity) {
		dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__,
				(unsigned long long)sector, size);
2269
		return -EINVAL;
P
Philipp Reisner 已提交
2270 2271 2272
	}

	if (!get_ldev_if_state(mdev, D_UP_TO_DATE)) {
2273
		verb = 1;
2274
		switch (pi->cmd) {
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288
		case P_DATA_REQUEST:
			drbd_send_ack_rp(mdev, P_NEG_DREPLY, p);
			break;
		case P_RS_DATA_REQUEST:
		case P_CSUM_RS_REQUEST:
		case P_OV_REQUEST:
			drbd_send_ack_rp(mdev, P_NEG_RS_DREPLY , p);
			break;
		case P_OV_REPLY:
			verb = 0;
			dec_rs_pending(mdev);
			drbd_send_ack_ex(mdev, P_OV_RESULT, sector, size, ID_IN_SYNC);
			break;
		default:
2289
			BUG();
2290 2291
		}
		if (verb && __ratelimit(&drbd_ratelimit_state))
P
Philipp Reisner 已提交
2292 2293
			dev_err(DEV, "Can not satisfy peer's read request, "
			    "no local data.\n");
2294

L
Lars Ellenberg 已提交
2295
		/* drain possibly payload */
2296
		return drbd_drain_block(mdev, pi->size);
P
Philipp Reisner 已提交
2297 2298 2299 2300 2301
	}

	/* 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.  */
2302
	peer_req = drbd_alloc_peer_req(mdev, p->block_id, sector, size, GFP_NOIO);
2303
	if (!peer_req) {
P
Philipp Reisner 已提交
2304
		put_ldev(mdev);
2305
		return -ENOMEM;
P
Philipp Reisner 已提交
2306 2307
	}

2308
	switch (pi->cmd) {
P
Philipp Reisner 已提交
2309
	case P_DATA_REQUEST:
2310
		peer_req->w.cb = w_e_end_data_req;
P
Philipp Reisner 已提交
2311
		fault_type = DRBD_FAULT_DT_RD;
2312 2313 2314
		/* application IO, don't drbd_rs_begin_io */
		goto submit;

P
Philipp Reisner 已提交
2315
	case P_RS_DATA_REQUEST:
2316
		peer_req->w.cb = w_e_end_rsdata_req;
P
Philipp Reisner 已提交
2317
		fault_type = DRBD_FAULT_RS_RD;
2318 2319
		/* used in the sector offset progress display */
		mdev->bm_resync_fo = BM_SECT_TO_BIT(sector);
P
Philipp Reisner 已提交
2320 2321 2322 2323 2324
		break;

	case P_OV_REPLY:
	case P_CSUM_RS_REQUEST:
		fault_type = DRBD_FAULT_RS_RD;
2325
		di = kmalloc(sizeof(*di) + pi->size, GFP_NOIO);
P
Philipp Reisner 已提交
2326 2327 2328
		if (!di)
			goto out_free_e;

2329
		di->digest_size = pi->size;
P
Philipp Reisner 已提交
2330 2331
		di->digest = (((char *)di)+sizeof(struct digest_info));

2332 2333
		peer_req->digest = di;
		peer_req->flags |= EE_HAS_DIGEST;
2334

2335
		if (drbd_recv_all(mdev->tconn, di->digest, pi->size))
P
Philipp Reisner 已提交
2336 2337
			goto out_free_e;

2338
		if (pi->cmd == P_CSUM_RS_REQUEST) {
2339
			D_ASSERT(mdev->tconn->agreed_pro_version >= 89);
2340
			peer_req->w.cb = w_e_end_csum_rs_req;
2341 2342
			/* used in the sector offset progress display */
			mdev->bm_resync_fo = BM_SECT_TO_BIT(sector);
2343
		} else if (pi->cmd == P_OV_REPLY) {
2344 2345
			/* track progress, we may need to throttle */
			atomic_add(size >> 9, &mdev->rs_sect_in);
2346
			peer_req->w.cb = w_e_end_ov_reply;
P
Philipp Reisner 已提交
2347
			dec_rs_pending(mdev);
2348 2349 2350
			/* 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
2351 2352 2353 2354 2355
		}
		break;

	case P_OV_REQUEST:
		if (mdev->ov_start_sector == ~(sector_t)0 &&
2356
		    mdev->tconn->agreed_pro_version >= 90) {
2357 2358
			unsigned long now = jiffies;
			int i;
P
Philipp Reisner 已提交
2359 2360
			mdev->ov_start_sector = sector;
			mdev->ov_position = sector;
2361 2362
			mdev->ov_left = drbd_bm_bits(mdev) - BM_SECT_TO_BIT(sector);
			mdev->rs_total = mdev->ov_left;
2363 2364 2365 2366
			for (i = 0; i < DRBD_SYNC_MARKS; i++) {
				mdev->rs_mark_left[i] = mdev->ov_left;
				mdev->rs_mark_time[i] = now;
			}
P
Philipp Reisner 已提交
2367 2368 2369
			dev_info(DEV, "Online Verify start sector: %llu\n",
					(unsigned long long)sector);
		}
2370
		peer_req->w.cb = w_e_end_ov_req;
P
Philipp Reisner 已提交
2371 2372 2373 2374
		fault_type = DRBD_FAULT_RS_RD;
		break;

	default:
2375
		BUG();
P
Philipp Reisner 已提交
2376 2377
	}

2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399
	/* 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.
	 */
2400 2401 2402
	if (mdev->state.peer != R_PRIMARY && drbd_rs_should_slow_down(mdev, sector))
		schedule_timeout_uninterruptible(HZ/10);
	if (drbd_rs_begin_io(mdev, sector))
2403
		goto out_free_e;
P
Philipp Reisner 已提交
2404

2405 2406 2407
submit_for_resync:
	atomic_add(size >> 9, &mdev->rs_sect_ev);

2408
submit:
P
Philipp Reisner 已提交
2409
	inc_unacked(mdev);
2410
	spin_lock_irq(&mdev->tconn->req_lock);
2411
	list_add_tail(&peer_req->w.list, &mdev->read_ee);
2412
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
2413

2414
	if (drbd_submit_peer_request(mdev, peer_req, READ, fault_type) == 0)
2415
		return 0;
P
Philipp Reisner 已提交
2416

2417 2418
	/* don't care for the reason here */
	dev_err(DEV, "submit failed, triggering re-connect\n");
2419
	spin_lock_irq(&mdev->tconn->req_lock);
2420
	list_del(&peer_req->w.list);
2421
	spin_unlock_irq(&mdev->tconn->req_lock);
2422 2423
	/* no drbd_rs_complete_io(), we are dropping the connection anyways */

P
Philipp Reisner 已提交
2424 2425
out_free_e:
	put_ldev(mdev);
2426
	drbd_free_peer_req(mdev, peer_req);
2427
	return -EIO;
P
Philipp Reisner 已提交
2428 2429 2430 2431 2432 2433
}

static int drbd_asb_recover_0p(struct drbd_conf *mdev) __must_hold(local)
{
	int self, peer, rv = -100;
	unsigned long ch_self, ch_peer;
2434
	enum drbd_after_sb_p after_sb_0p;
P
Philipp Reisner 已提交
2435 2436 2437 2438 2439 2440 2441

	self = mdev->ldev->md.uuid[UI_BITMAP] & 1;
	peer = mdev->p_uuid[UI_BITMAP] & 1;

	ch_peer = mdev->p_uuid[UI_SIZE];
	ch_self = mdev->comm_bm_set;

2442 2443 2444 2445
	rcu_read_lock();
	after_sb_0p = rcu_dereference(mdev->tconn->net_conf)->after_sb_0p;
	rcu_read_unlock();
	switch (after_sb_0p) {
P
Philipp Reisner 已提交
2446 2447 2448
	case ASB_CONSENSUS:
	case ASB_DISCARD_SECONDARY:
	case ASB_CALL_HELPER:
2449
	case ASB_VIOLENTLY:
P
Philipp Reisner 已提交
2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473
		dev_err(DEV, "Configuration error.\n");
		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... */
L
Lars Ellenberg 已提交
2474
		dev_warn(DEV, "Discard younger/older primary did not find a decision\n"
P
Philipp Reisner 已提交
2475 2476 2477
		     "Using discard-least-changes instead\n");
	case ASB_DISCARD_ZERO_CHG:
		if (ch_peer == 0 && ch_self == 0) {
2478
			rv = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags)
P
Philipp Reisner 已提交
2479 2480 2481 2482 2483 2484
				? -1 : 1;
			break;
		} else {
			if (ch_peer == 0) { rv =  1; break; }
			if (ch_self == 0) { rv = -1; break; }
		}
2485
		if (after_sb_0p == ASB_DISCARD_ZERO_CHG)
P
Philipp Reisner 已提交
2486 2487 2488 2489 2490 2491 2492 2493
			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. */
2494
			rv = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags)
P
Philipp Reisner 已提交
2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508
				? -1 : 1;
		break;
	case ASB_DISCARD_LOCAL:
		rv = -1;
		break;
	case ASB_DISCARD_REMOTE:
		rv =  1;
	}

	return rv;
}

static int drbd_asb_recover_1p(struct drbd_conf *mdev) __must_hold(local)
{
2509
	int hg, rv = -100;
2510
	enum drbd_after_sb_p after_sb_1p;
P
Philipp Reisner 已提交
2511

2512 2513 2514 2515
	rcu_read_lock();
	after_sb_1p = rcu_dereference(mdev->tconn->net_conf)->after_sb_1p;
	rcu_read_unlock();
	switch (after_sb_1p) {
P
Philipp Reisner 已提交
2516 2517 2518 2519 2520
	case ASB_DISCARD_YOUNGER_PRI:
	case ASB_DISCARD_OLDER_PRI:
	case ASB_DISCARD_LEAST_CHG:
	case ASB_DISCARD_LOCAL:
	case ASB_DISCARD_REMOTE:
2521
	case ASB_DISCARD_ZERO_CHG:
P
Philipp Reisner 已提交
2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540
		dev_err(DEV, "Configuration error.\n");
		break;
	case ASB_DISCONNECT:
		break;
	case ASB_CONSENSUS:
		hg = drbd_asb_recover_0p(mdev);
		if (hg == -1 && mdev->state.role == R_SECONDARY)
			rv = hg;
		if (hg == 1  && mdev->state.role == R_PRIMARY)
			rv = hg;
		break;
	case ASB_VIOLENTLY:
		rv = drbd_asb_recover_0p(mdev);
		break;
	case ASB_DISCARD_SECONDARY:
		return mdev->state.role == R_PRIMARY ? 1 : -1;
	case ASB_CALL_HELPER:
		hg = drbd_asb_recover_0p(mdev);
		if (hg == -1 && mdev->state.role == R_PRIMARY) {
2541 2542 2543
			enum drbd_state_rv rv2;

			drbd_set_role(mdev, R_SECONDARY, 0);
P
Philipp Reisner 已提交
2544 2545 2546
			 /* 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. */
2547 2548
			rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY));
			if (rv2 != SS_SUCCESS) {
P
Philipp Reisner 已提交
2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562
				drbd_khelper(mdev, "pri-lost-after-sb");
			} else {
				dev_warn(DEV, "Successfully gave up primary role.\n");
				rv = hg;
			}
		} else
			rv = hg;
	}

	return rv;
}

static int drbd_asb_recover_2p(struct drbd_conf *mdev) __must_hold(local)
{
2563
	int hg, rv = -100;
2564
	enum drbd_after_sb_p after_sb_2p;
P
Philipp Reisner 已提交
2565

2566 2567 2568 2569
	rcu_read_lock();
	after_sb_2p = rcu_dereference(mdev->tconn->net_conf)->after_sb_2p;
	rcu_read_unlock();
	switch (after_sb_2p) {
P
Philipp Reisner 已提交
2570 2571 2572 2573 2574 2575 2576
	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:
2577
	case ASB_DISCARD_ZERO_CHG:
P
Philipp Reisner 已提交
2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
		dev_err(DEV, "Configuration error.\n");
		break;
	case ASB_VIOLENTLY:
		rv = drbd_asb_recover_0p(mdev);
		break;
	case ASB_DISCONNECT:
		break;
	case ASB_CALL_HELPER:
		hg = drbd_asb_recover_0p(mdev);
		if (hg == -1) {
2588 2589
			enum drbd_state_rv rv2;

P
Philipp Reisner 已提交
2590 2591 2592
			 /* 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. */
2593 2594
			rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY));
			if (rv2 != SS_SUCCESS) {
P
Philipp Reisner 已提交
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632
				drbd_khelper(mdev, "pri-lost-after-sb");
			} else {
				dev_warn(DEV, "Successfully gave up primary role.\n");
				rv = hg;
			}
		} else
			rv = hg;
	}

	return rv;
}

static void drbd_uuid_dump(struct drbd_conf *mdev, char *text, u64 *uuid,
			   u64 bits, u64 flags)
{
	if (!uuid) {
		dev_info(DEV, "%s uuid info vanished while I was looking!\n", text);
		return;
	}
	dev_info(DEV, "%s %016llX:%016llX:%016llX:%016llX bits:%llu flags:%llX\n",
	     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
2633 2634
-1091   requires proto 91
-1096   requires proto 96
P
Philipp Reisner 已提交
2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662
 */
static int drbd_uuid_compare(struct drbd_conf *mdev, int *rule_nr) __must_hold(local)
{
	u64 self, peer;
	int i, j;

	self = mdev->ldev->md.uuid[UI_CURRENT] & ~((u64)1);
	peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1);

	*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 */

		if (mdev->p_uuid[UI_BITMAP] == (u64)0 && mdev->ldev->md.uuid[UI_BITMAP] != (u64)0) {

2663
			if (mdev->tconn->agreed_pro_version < 91)
2664
				return -1091;
P
Philipp Reisner 已提交
2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683

			if ((mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) &&
			    (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1))) {
				dev_info(DEV, "was SyncSource, missed the resync finished event, corrected myself:\n");
				drbd_uuid_set_bm(mdev, 0UL);

				drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid,
					       mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0);
				*rule_nr = 34;
			} else {
				dev_info(DEV, "was SyncSource (peer failed to write sync_uuid)\n");
				*rule_nr = 36;
			}

			return 1;
		}

		if (mdev->ldev->md.uuid[UI_BITMAP] == (u64)0 && mdev->p_uuid[UI_BITMAP] != (u64)0) {

2684
			if (mdev->tconn->agreed_pro_version < 91)
2685
				return -1091;
P
Philipp Reisner 已提交
2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716

			if ((mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_BITMAP] & ~((u64)1)) &&
			    (mdev->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1))) {
				dev_info(DEV, "was SyncTarget, peer missed the resync finished event, corrected peer:\n");

				mdev->p_uuid[UI_HISTORY_START + 1] = mdev->p_uuid[UI_HISTORY_START];
				mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_BITMAP];
				mdev->p_uuid[UI_BITMAP] = 0UL;

				drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]);
				*rule_nr = 35;
			} else {
				dev_info(DEV, "was SyncTarget (failed to write sync_uuid)\n");
				*rule_nr = 37;
			}

			return -1;
		}

		/* Common power [off|failure] */
		rct = (test_bit(CRASHED_PRIMARY, &mdev->flags) ? 1 : 0) +
			(mdev->p_uuid[UI_FLAGS] & 2);
		/* 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 */
2717
			dc = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags);
P
Philipp Reisner 已提交
2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729
			return dc ? -1 : 1;
		}
	}

	*rule_nr = 50;
	peer = mdev->p_uuid[UI_BITMAP] & ~((u64)1);
	if (self == peer)
		return -1;

	*rule_nr = 51;
	peer = mdev->p_uuid[UI_HISTORY_START] & ~((u64)1);
	if (self == peer) {
2730
		if (mdev->tconn->agreed_pro_version < 96 ?
2731 2732 2733
		    (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) ==
		    (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1)) :
		    peer + UUID_NEW_BM_OFFSET == (mdev->p_uuid[UI_BITMAP] & ~((u64)1))) {
P
Philipp Reisner 已提交
2734 2735 2736
			/* The last P_SYNC_UUID did not get though. Undo the last start of
			   resync as sync source modifications of the peer's UUIDs. */

2737
			if (mdev->tconn->agreed_pro_version < 91)
2738
				return -1091;
P
Philipp Reisner 已提交
2739 2740 2741

			mdev->p_uuid[UI_BITMAP] = mdev->p_uuid[UI_HISTORY_START];
			mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_HISTORY_START + 1];
2742 2743 2744 2745

			dev_info(DEV, "Did not got last syncUUID packet, corrected:\n");
			drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]);

P
Philipp Reisner 已提交
2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
			return -1;
		}
	}

	*rule_nr = 60;
	self = mdev->ldev->md.uuid[UI_CURRENT] & ~((u64)1);
	for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) {
		peer = mdev->p_uuid[i] & ~((u64)1);
		if (self == peer)
			return -2;
	}

	*rule_nr = 70;
	self = mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1);
	peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1);
	if (self == peer)
		return 1;

	*rule_nr = 71;
	self = mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1);
	if (self == peer) {
2767
		if (mdev->tconn->agreed_pro_version < 96 ?
2768 2769 2770
		    (mdev->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) ==
		    (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) :
		    self + UUID_NEW_BM_OFFSET == (mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1))) {
P
Philipp Reisner 已提交
2771 2772 2773
			/* The last P_SYNC_UUID did not get though. Undo the last start of
			   resync as sync source modifications of our UUIDs. */

2774
			if (mdev->tconn->agreed_pro_version < 91)
2775
				return -1091;
P
Philipp Reisner 已提交
2776 2777 2778 2779

			_drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_HISTORY_START]);
			_drbd_uuid_set(mdev, UI_HISTORY_START, mdev->ldev->md.uuid[UI_HISTORY_START + 1]);

2780
			dev_info(DEV, "Last syncUUID did not get through, corrected:\n");
P
Philipp Reisner 已提交
2781 2782 2783 2784 2785 2786 2787 2788 2789
			drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid,
				       mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0);

			return 1;
		}
	}


	*rule_nr = 80;
2790
	peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1);
P
Philipp Reisner 已提交
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823
	for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) {
		self = mdev->ldev->md.uuid[i] & ~((u64)1);
		if (self == peer)
			return 2;
	}

	*rule_nr = 90;
	self = mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1);
	peer = mdev->p_uuid[UI_BITMAP] & ~((u64)1);
	if (self == peer && self != ((u64)0))
		return 100;

	*rule_nr = 100;
	for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) {
		self = mdev->ldev->md.uuid[i] & ~((u64)1);
		for (j = UI_HISTORY_START; j <= UI_HISTORY_END; j++) {
			peer = mdev->p_uuid[j] & ~((u64)1);
			if (self == peer)
				return -100;
		}
	}

	return -1000;
}

/* drbd_sync_handshake() returns the new conn state on success, or
   CONN_MASK (-1) on failure.
 */
static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_role peer_role,
					   enum drbd_disk_state peer_disk) __must_hold(local)
{
	enum drbd_conns rv = C_MASK;
	enum drbd_disk_state mydisk;
2824 2825
	struct net_conf *nc;
	int hg, rule_nr, rr_conflict, dry_run;
P
Philipp Reisner 已提交
2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843

	mydisk = mdev->state.disk;
	if (mydisk == D_NEGOTIATING)
		mydisk = mdev->new_state_tmp.disk;

	dev_info(DEV, "drbd_sync_handshake:\n");
	drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, mdev->comm_bm_set, 0);
	drbd_uuid_dump(mdev, "peer", mdev->p_uuid,
		       mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]);

	hg = drbd_uuid_compare(mdev, &rule_nr);

	dev_info(DEV, "uuid_compare()=%d by rule %d\n", hg, rule_nr);

	if (hg == -1000) {
		dev_alert(DEV, "Unrelated data, aborting!\n");
		return C_MASK;
	}
2844 2845
	if (hg < -1000) {
		dev_alert(DEV, "To resolve this both sides have to support at least protocol %d\n", -hg - 1000);
P
Philipp Reisner 已提交
2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858
		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;
		dev_info(DEV, "Becoming sync %s due to disk states.\n",
		     hg > 0 ? "source" : "target");
	}

2859 2860 2861
	if (abs(hg) == 100)
		drbd_khelper(mdev, "initial-split-brain");

2862 2863 2864 2865
	rcu_read_lock();
	nc = rcu_dereference(mdev->tconn->net_conf);

	if (hg == 100 || (hg == -100 && nc->always_asbp)) {
P
Philipp Reisner 已提交
2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893
		int pcount = (mdev->state.role == R_PRIMARY)
			   + (peer_role == R_PRIMARY);
		int forced = (hg == -100);

		switch (pcount) {
		case 0:
			hg = drbd_asb_recover_0p(mdev);
			break;
		case 1:
			hg = drbd_asb_recover_1p(mdev);
			break;
		case 2:
			hg = drbd_asb_recover_2p(mdev);
			break;
		}
		if (abs(hg) < 100) {
			dev_warn(DEV, "Split-Brain detected, %d primaries, "
			     "automatically solved. Sync from %s node\n",
			     pcount, (hg < 0) ? "peer" : "this");
			if (forced) {
				dev_warn(DEV, "Doing a full sync, since"
				     " UUIDs where ambiguous.\n");
				hg = hg*2;
			}
		}
	}

	if (hg == -100) {
2894
		if (nc->want_lose && !(mdev->p_uuid[UI_FLAGS]&1))
P
Philipp Reisner 已提交
2895
			hg = -1;
2896
		if (!nc->want_lose && (mdev->p_uuid[UI_FLAGS]&1))
P
Philipp Reisner 已提交
2897 2898 2899 2900 2901 2902 2903
			hg = 1;

		if (abs(hg) < 100)
			dev_warn(DEV, "Split-Brain detected, manually solved. "
			     "Sync from %s node\n",
			     (hg < 0) ? "peer" : "this");
	}
2904 2905 2906
	rr_conflict = nc->rr_conflict;
	dry_run = nc->dry_run;
	rcu_read_unlock();
P
Philipp Reisner 已提交
2907 2908

	if (hg == -100) {
2909 2910 2911 2912
		/* 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... */
2913
		dev_alert(DEV, "Split-Brain detected but unresolved, dropping connection!\n");
P
Philipp Reisner 已提交
2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924
		drbd_khelper(mdev, "split-brain");
		return C_MASK;
	}

	if (hg > 0 && mydisk <= D_INCONSISTENT) {
		dev_err(DEV, "I shall become SyncSource, but I am inconsistent!\n");
		return C_MASK;
	}

	if (hg < 0 && /* by intention we do not use mydisk here. */
	    mdev->state.role == R_PRIMARY && mdev->state.disk >= D_CONSISTENT) {
2925
		switch (rr_conflict) {
P
Philipp Reisner 已提交
2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937
		case ASB_CALL_HELPER:
			drbd_khelper(mdev, "pri-lost");
			/* fall through */
		case ASB_DISCONNECT:
			dev_err(DEV, "I shall become SyncTarget, but I am primary!\n");
			return C_MASK;
		case ASB_VIOLENTLY:
			dev_warn(DEV, "Becoming SyncTarget, violating the stable-data"
			     "assumption\n");
		}
	}

2938
	if (dry_run || test_bit(CONN_DRY_RUN, &mdev->tconn->flags)) {
2939 2940 2941 2942 2943 2944 2945 2946 2947
		if (hg == 0)
			dev_info(DEV, "dry-run connect: No resync, would become Connected immediately.\n");
		else
			dev_info(DEV, "dry-run connect: Would become %s, doing a %s resync.",
				 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
2948 2949
	if (abs(hg) >= 2) {
		dev_info(DEV, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n");
2950 2951
		if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from sync_handshake",
					BM_LOCKED_SET_ALLOWED))
P
Philipp Reisner 已提交
2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990
			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;
		if (drbd_bm_total_weight(mdev)) {
			dev_info(DEV, "No resync, but %lu bits in bitmap!\n",
			     drbd_bm_total_weight(mdev));
		}
	}

	return rv;
}

/* returns 1 if invalid */
static int cmp_after_sb(enum drbd_after_sb_p peer, enum drbd_after_sb_p self)
{
	/* ASB_DISCARD_REMOTE - ASB_DISCARD_LOCAL is valid */
	if ((peer == ASB_DISCARD_REMOTE && self == ASB_DISCARD_LOCAL) ||
	    (self == ASB_DISCARD_REMOTE && peer == ASB_DISCARD_LOCAL))
		return 0;

	/* any other things with ASB_DISCARD_REMOTE or ASB_DISCARD_LOCAL are invalid */
	if (peer == ASB_DISCARD_REMOTE || peer == ASB_DISCARD_LOCAL ||
	    self == ASB_DISCARD_REMOTE || self == ASB_DISCARD_LOCAL)
		return 1;

	/* everything else is valid if they are equal on both sides. */
	if (peer == self)
		return 0;

	/* everything es is invalid. */
	return 1;
}

2991
static int receive_protocol(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
2992
{
2993
	struct p_protocol *p = pi->data;
P
Philipp Reisner 已提交
2994
	int p_proto, p_after_sb_0p, p_after_sb_1p, p_after_sb_2p;
2995
	int p_want_lose, p_two_primaries, cf;
2996
	struct net_conf *nc;
P
Philipp Reisner 已提交
2997 2998 2999 3000 3001 3002

	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);
3003 3004 3005
	cf		= be32_to_cpu(p->conn_flags);
	p_want_lose = cf & CF_WANT_LOSE;

3006
	if (tconn->agreed_pro_version >= 87) {
3007 3008
		char integrity_alg[SHARED_SECRET_MAX];
		struct crypto_hash *tfm = NULL;
3009 3010
		int err;

3011
		if (pi->size > sizeof(integrity_alg))
3012
			return -EIO;
3013
		err = drbd_recv_all(tconn, integrity_alg, pi->size);
3014 3015
		if (err)
			return err;
3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026
		integrity_alg[SHARED_SECRET_MAX-1] = 0;

		if (integrity_alg[0]) {
			tfm = crypto_alloc_hash(integrity_alg, 0, CRYPTO_ALG_ASYNC);
			if (!tfm) {
				conn_err(tconn, "peer data-integrity-alg %s not supported\n",
					 integrity_alg);
				goto disconnect;
			}
			conn_info(tconn, "peer data-integrity-alg: %s\n", integrity_alg);
		}
3027

3028 3029 3030
		if (tconn->peer_integrity_tfm)
			crypto_free_hash(tconn->peer_integrity_tfm);
		tconn->peer_integrity_tfm = tfm;
3031 3032
	}

3033
	clear_bit(CONN_DRY_RUN, &tconn->flags);
3034 3035

	if (cf & CF_DRY_RUN)
3036
		set_bit(CONN_DRY_RUN, &tconn->flags);
P
Philipp Reisner 已提交
3037

3038 3039 3040 3041
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);

	if (p_proto != nc->wire_protocol && tconn->agreed_pro_version < 100) {
3042
		conn_err(tconn, "incompatible communication protocols\n");
3043
		goto disconnect_rcu_unlock;
P
Philipp Reisner 已提交
3044 3045
	}

3046
	if (cmp_after_sb(p_after_sb_0p, nc->after_sb_0p)) {
3047
		conn_err(tconn, "incompatible after-sb-0pri settings\n");
3048
		goto disconnect_rcu_unlock;
P
Philipp Reisner 已提交
3049 3050
	}

3051
	if (cmp_after_sb(p_after_sb_1p, nc->after_sb_1p)) {
3052
		conn_err(tconn, "incompatible after-sb-1pri settings\n");
3053
		goto disconnect_rcu_unlock;
P
Philipp Reisner 已提交
3054 3055
	}

3056
	if (cmp_after_sb(p_after_sb_2p, nc->after_sb_2p)) {
3057
		conn_err(tconn, "incompatible after-sb-2pri settings\n");
3058
		goto disconnect_rcu_unlock;
P
Philipp Reisner 已提交
3059 3060
	}

3061
	if (p_want_lose && nc->want_lose) {
3062
		conn_err(tconn, "both sides have the 'want_lose' flag set\n");
3063
		goto disconnect_rcu_unlock;
P
Philipp Reisner 已提交
3064 3065
	}

3066
	if (p_two_primaries != nc->two_primaries) {
3067
		conn_err(tconn, "incompatible setting of the two-primaries options\n");
3068
		goto disconnect_rcu_unlock;
P
Philipp Reisner 已提交
3069 3070
	}

3071 3072
	rcu_read_unlock();

3073
	return 0;
P
Philipp Reisner 已提交
3074

3075 3076
disconnect_rcu_unlock:
	rcu_read_unlock();
P
Philipp Reisner 已提交
3077
disconnect:
3078
	conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3079
	return -EIO;
P
Philipp Reisner 已提交
3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103
}

/* 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. */
struct crypto_hash *drbd_crypto_alloc_digest_safe(const struct drbd_conf *mdev,
		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)) {
		dev_err(DEV, "Can not allocate \"%s\" as %s (reason: %ld)\n",
			alg, name, PTR_ERR(tfm));
		return tfm;
	}
	return tfm;
}

3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142
static int ignore_remaining_packet(struct drbd_tconn *tconn, struct packet_info *pi)
{
	void *buffer = tconn->data.rbuf;
	int size = pi->size;

	while (size) {
		int s = min_t(int, size, DRBD_SOCKET_BUFFER_SIZE);
		s = drbd_recv(tconn, buffer, s);
		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.)
 */
static int config_unknown_volume(struct drbd_tconn *tconn, struct packet_info *pi)
{
	conn_warn(tconn, "Volume %u unknown; ignoring %s packet\n",
		  pi->vnr, cmdname(pi->cmd));
	return ignore_remaining_packet(tconn, pi);
}

static int receive_SyncParam(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3143
{
3144
	struct drbd_conf *mdev;
3145
	struct p_rs_param_95 *p;
P
Philipp Reisner 已提交
3146 3147 3148
	unsigned int header_size, data_size, exp_max_sz;
	struct crypto_hash *verify_tfm = NULL;
	struct crypto_hash *csums_tfm = NULL;
3149
	struct net_conf *old_conf, *new_conf = NULL;
3150
	const int apv = tconn->agreed_pro_version;
3151 3152
	int *rs_plan_s = NULL;
	int fifo_size = 0;
3153
	int err;
P
Philipp Reisner 已提交
3154

3155 3156 3157 3158
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3159 3160 3161
	exp_max_sz  = apv <= 87 ? sizeof(struct p_rs_param)
		    : apv == 88 ? sizeof(struct p_rs_param)
					+ SHARED_SECRET_MAX
3162 3163
		    : apv <= 94 ? sizeof(struct p_rs_param_89)
		    : /* apv >= 95 */ sizeof(struct p_rs_param_95);
P
Philipp Reisner 已提交
3164

3165
	if (pi->size > exp_max_sz) {
P
Philipp Reisner 已提交
3166
		dev_err(DEV, "SyncParam packet too long: received %u, expected <= %u bytes\n",
3167
		    pi->size, exp_max_sz);
3168
		return -EIO;
P
Philipp Reisner 已提交
3169 3170 3171
	}

	if (apv <= 88) {
3172
		header_size = sizeof(struct p_rs_param);
3173
		data_size = pi->size - header_size;
3174
	} else if (apv <= 94) {
3175
		header_size = sizeof(struct p_rs_param_89);
3176
		data_size = pi->size - header_size;
P
Philipp Reisner 已提交
3177
		D_ASSERT(data_size == 0);
3178
	} else {
3179
		header_size = sizeof(struct p_rs_param_95);
3180
		data_size = pi->size - header_size;
P
Philipp Reisner 已提交
3181 3182 3183 3184
		D_ASSERT(data_size == 0);
	}

	/* initialize verify_alg and csums_alg */
3185
	p = pi->data;
P
Philipp Reisner 已提交
3186 3187
	memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX);

3188
	err = drbd_recv_all(mdev->tconn, p, header_size);
3189 3190
	if (err)
		return err;
P
Philipp Reisner 已提交
3191

3192 3193 3194 3195
	if (get_ldev(mdev)) {
		mdev->ldev->dc.resync_rate = be32_to_cpu(p->rate);
		put_ldev(mdev);
	}
P
Philipp Reisner 已提交
3196 3197 3198 3199 3200 3201 3202

	if (apv >= 88) {
		if (apv == 88) {
			if (data_size > SHARED_SECRET_MAX) {
				dev_err(DEV, "verify-alg too long, "
				    "peer wants %u, accepting only %u byte\n",
						data_size, SHARED_SECRET_MAX);
3203
				return -EIO;
P
Philipp Reisner 已提交
3204 3205
			}

3206 3207 3208
			err = drbd_recv_all(mdev->tconn, p->verify_alg, data_size);
			if (err)
				return err;
P
Philipp Reisner 已提交
3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223

			/* we expect NUL terminated string */
			/* but just in case someone tries to be evil */
			D_ASSERT(p->verify_alg[data_size-1] == 0);
			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 */
			D_ASSERT(p->verify_alg[SHARED_SECRET_MAX-1] == 0);
			D_ASSERT(p->csums_alg[SHARED_SECRET_MAX-1] == 0);
			p->verify_alg[SHARED_SECRET_MAX-1] = 0;
			p->csums_alg[SHARED_SECRET_MAX-1] = 0;
		}

3224
		mutex_lock(&mdev->tconn->conf_update);
3225 3226 3227
		old_conf = mdev->tconn->net_conf;

		if (strcmp(old_conf->verify_alg, p->verify_alg)) {
P
Philipp Reisner 已提交
3228 3229
			if (mdev->state.conn == C_WF_REPORT_PARAMS) {
				dev_err(DEV, "Different verify-alg settings. me=\"%s\" peer=\"%s\"\n",
3230
				    old_conf->verify_alg, p->verify_alg);
P
Philipp Reisner 已提交
3231 3232 3233 3234 3235 3236 3237 3238 3239 3240
				goto disconnect;
			}
			verify_tfm = drbd_crypto_alloc_digest_safe(mdev,
					p->verify_alg, "verify-alg");
			if (IS_ERR(verify_tfm)) {
				verify_tfm = NULL;
				goto disconnect;
			}
		}

3241
		if (apv >= 89 && strcmp(old_conf->csums_alg, p->csums_alg)) {
P
Philipp Reisner 已提交
3242 3243
			if (mdev->state.conn == C_WF_REPORT_PARAMS) {
				dev_err(DEV, "Different csums-alg settings. me=\"%s\" peer=\"%s\"\n",
3244
				    old_conf->csums_alg, p->csums_alg);
P
Philipp Reisner 已提交
3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
				goto disconnect;
			}
			csums_tfm = drbd_crypto_alloc_digest_safe(mdev,
					p->csums_alg, "csums-alg");
			if (IS_ERR(csums_tfm)) {
				csums_tfm = NULL;
				goto disconnect;
			}
		}

3255 3256 3257 3258 3259 3260
		if (apv > 94 && get_ldev(mdev)) {
			mdev->ldev->dc.resync_rate = be32_to_cpu(p->rate);
			mdev->ldev->dc.c_plan_ahead = be32_to_cpu(p->c_plan_ahead);
			mdev->ldev->dc.c_delay_target = be32_to_cpu(p->c_delay_target);
			mdev->ldev->dc.c_fill_target = be32_to_cpu(p->c_fill_target);
			mdev->ldev->dc.c_max_rate = be32_to_cpu(p->c_max_rate);
3261

3262
			fifo_size = (mdev->ldev->dc.c_plan_ahead * 10 * SLEEP_TIME) / HZ;
3263 3264 3265 3266
			if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) {
				rs_plan_s   = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL);
				if (!rs_plan_s) {
					dev_err(DEV, "kmalloc of fifo_buffer failed");
3267
					put_ldev(mdev);
3268 3269 3270
					goto disconnect;
				}
			}
3271
			put_ldev(mdev);
3272
		}
P
Philipp Reisner 已提交
3273

3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297
		if (verify_tfm || csums_tfm) {
			new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
			if (!new_conf) {
				dev_err(DEV, "Allocation of new net_conf failed\n");
				goto disconnect;
			}

			*new_conf = *old_conf;

			if (verify_tfm) {
				strcpy(new_conf->verify_alg, p->verify_alg);
				new_conf->verify_alg_len = strlen(p->verify_alg) + 1;
				crypto_free_hash(mdev->tconn->verify_tfm);
				mdev->tconn->verify_tfm = verify_tfm;
				dev_info(DEV, "using verify-alg: \"%s\"\n", p->verify_alg);
			}
			if (csums_tfm) {
				strcpy(new_conf->csums_alg, p->csums_alg);
				new_conf->csums_alg_len = strlen(p->csums_alg) + 1;
				crypto_free_hash(mdev->tconn->csums_tfm);
				mdev->tconn->csums_tfm = csums_tfm;
				dev_info(DEV, "using csums-alg: \"%s\"\n", p->csums_alg);
			}
			rcu_assign_pointer(tconn->net_conf, new_conf);
P
Philipp Reisner 已提交
3298
		}
3299
		mutex_unlock(&mdev->tconn->conf_update);
3300 3301 3302
		if (new_conf) {
			synchronize_rcu();
			kfree(old_conf);
P
Philipp Reisner 已提交
3303
		}
3304 3305

		spin_lock(&mdev->peer_seq_lock);
3306 3307 3308 3309 3310 3311
		if (fifo_size != mdev->rs_plan_s.size) {
			kfree(mdev->rs_plan_s.values);
			mdev->rs_plan_s.values = rs_plan_s;
			mdev->rs_plan_s.size   = fifo_size;
			mdev->rs_planed = 0;
		}
P
Philipp Reisner 已提交
3312 3313
		spin_unlock(&mdev->peer_seq_lock);
	}
3314
	return 0;
P
Philipp Reisner 已提交
3315 3316

disconnect:
3317
	mutex_unlock(&mdev->tconn->conf_update);
P
Philipp Reisner 已提交
3318 3319 3320 3321 3322
	/* 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);
3323
	conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3324
	return -EIO;
P
Philipp Reisner 已提交
3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339
}

/* warn if the arguments differ by more than 12.5% */
static void warn_if_differ_considerably(struct drbd_conf *mdev,
	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))
		dev_warn(DEV, "Considerable difference in %s: %llus vs. %llus\n", s,
		     (unsigned long long)a, (unsigned long long)b);
}

3340
static int receive_sizes(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3341
{
3342
	struct drbd_conf *mdev;
3343
	struct p_sizes *p = pi->data;
P
Philipp Reisner 已提交
3344 3345 3346
	enum determine_dev_size dd = unchanged;
	sector_t p_size, p_usize, my_usize;
	int ldsc = 0; /* local disk size changed */
3347
	enum dds_flags ddsf;
P
Philipp Reisner 已提交
3348

3349 3350 3351 3352
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381
	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. */
	mdev->p_size = p_size;

	if (get_ldev(mdev)) {
		warn_if_differ_considerably(mdev, "lower level device sizes",
			   p_size, drbd_get_max_capacity(mdev->ldev));
		warn_if_differ_considerably(mdev, "user requested size",
					    p_usize, mdev->ldev->dc.disk_size);

		/* if this is the first connect, or an otherwise expected
		 * param exchange, choose the minimum */
		if (mdev->state.conn == C_WF_REPORT_PARAMS)
			p_usize = min_not_zero((sector_t)mdev->ldev->dc.disk_size,
					     p_usize);

		my_usize = mdev->ldev->dc.disk_size;

		if (mdev->ldev->dc.disk_size != p_usize) {
			mdev->ldev->dc.disk_size = p_usize;
			dev_info(DEV, "Peer sets u_size to %lu sectors\n",
			     (unsigned long)mdev->ldev->dc.disk_size);
		}

		/* Never shrink a device with usable data during connect.
		   But allow online shrinking if we are connected. */
3382
		if (drbd_new_dev_size(mdev, mdev->ldev, p_usize, 0) <
P
Philipp Reisner 已提交
3383 3384 3385 3386
		   drbd_get_capacity(mdev->this_bdev) &&
		   mdev->state.disk >= D_OUTDATED &&
		   mdev->state.conn < C_CONNECTED) {
			dev_err(DEV, "The peer's disk size is too small!\n");
3387
			conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
3388 3389
			mdev->ldev->dc.disk_size = my_usize;
			put_ldev(mdev);
3390
			return -EIO;
P
Philipp Reisner 已提交
3391 3392 3393 3394
		}
		put_ldev(mdev);
	}

3395
	ddsf = be16_to_cpu(p->dds_flags);
P
Philipp Reisner 已提交
3396
	if (get_ldev(mdev)) {
B
Bart Van Assche 已提交
3397
		dd = drbd_determine_dev_size(mdev, ddsf);
P
Philipp Reisner 已提交
3398 3399
		put_ldev(mdev);
		if (dd == dev_size_error)
3400
			return -EIO;
P
Philipp Reisner 已提交
3401 3402 3403 3404 3405 3406
		drbd_md_sync(mdev);
	} else {
		/* I am diskless, need to accept the peer's size. */
		drbd_set_my_capacity(mdev, p_size);
	}

3407 3408 3409
	mdev->peer_max_bio_size = be32_to_cpu(p->max_bio_size);
	drbd_reconsider_max_bio_size(mdev);

P
Philipp Reisner 已提交
3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423
	if (get_ldev(mdev)) {
		if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) {
			mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
			ldsc = 1;
		}

		put_ldev(mdev);
	}

	if (mdev->state.conn > C_WF_REPORT_PARAMS) {
		if (be64_to_cpu(p->c_size) !=
		    drbd_get_capacity(mdev->this_bdev) || ldsc) {
			/* we have different sizes, probably peer
			 * needs to know my new size... */
3424
			drbd_send_sizes(mdev, 0, ddsf);
P
Philipp Reisner 已提交
3425 3426 3427 3428
		}
		if (test_and_clear_bit(RESIZE_PENDING, &mdev->flags) ||
		    (dd == grew && mdev->state.conn == C_CONNECTED)) {
			if (mdev->state.pdsk >= D_INCONSISTENT &&
3429 3430 3431 3432 3433 3434
			    mdev->state.disk >= D_INCONSISTENT) {
				if (ddsf & DDSF_NO_RESYNC)
					dev_info(DEV, "Resync of new storage suppressed with --assume-clean\n");
				else
					resync_after_online_grow(mdev);
			} else
P
Philipp Reisner 已提交
3435 3436 3437 3438
				set_bit(RESYNC_AFTER_NEG, &mdev->flags);
		}
	}

3439
	return 0;
P
Philipp Reisner 已提交
3440 3441
}

3442
static int receive_uuids(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3443
{
3444
	struct drbd_conf *mdev;
3445
	struct p_uuids *p = pi->data;
P
Philipp Reisner 已提交
3446
	u64 *p_uuid;
3447
	int i, updated_uuids = 0;
P
Philipp Reisner 已提交
3448

3449 3450 3451 3452
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466
	p_uuid = kmalloc(sizeof(u64)*UI_EXTENDED_SIZE, GFP_NOIO);

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

	kfree(mdev->p_uuid);
	mdev->p_uuid = p_uuid;

	if (mdev->state.conn < C_CONNECTED &&
	    mdev->state.disk < D_INCONSISTENT &&
	    mdev->state.role == R_PRIMARY &&
	    (mdev->ed_uuid & ~((u64)1)) != (p_uuid[UI_CURRENT] & ~((u64)1))) {
		dev_err(DEV, "Can only connect to data with current UUID=%016llX\n",
		    (unsigned long long)mdev->ed_uuid);
3467
		conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3468
		return -EIO;
P
Philipp Reisner 已提交
3469 3470 3471 3472 3473
	}

	if (get_ldev(mdev)) {
		int skip_initial_sync =
			mdev->state.conn == C_CONNECTED &&
3474
			mdev->tconn->agreed_pro_version >= 90 &&
P
Philipp Reisner 已提交
3475 3476 3477 3478 3479
			mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED &&
			(p_uuid[UI_FLAGS] & 8);
		if (skip_initial_sync) {
			dev_info(DEV, "Accepted new current UUID, preparing to skip initial sync\n");
			drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write,
3480 3481
					"clear_n_write from receive_uuids",
					BM_LOCKED_TEST_ALLOWED);
P
Philipp Reisner 已提交
3482 3483 3484 3485 3486
			_drbd_uuid_set(mdev, UI_CURRENT, p_uuid[UI_CURRENT]);
			_drbd_uuid_set(mdev, UI_BITMAP, 0);
			_drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE),
					CS_VERBOSE, NULL);
			drbd_md_sync(mdev);
3487
			updated_uuids = 1;
P
Philipp Reisner 已提交
3488 3489
		}
		put_ldev(mdev);
3490 3491 3492 3493
	} else if (mdev->state.disk < D_INCONSISTENT &&
		   mdev->state.role == R_PRIMARY) {
		/* I am a diskless primary, the peer just created a new current UUID
		   for me. */
3494
		updated_uuids = drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]);
P
Philipp Reisner 已提交
3495 3496 3497 3498 3499 3500
	}

	/* 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... */
3501 3502
	mutex_lock(mdev->state_mutex);
	mutex_unlock(mdev->state_mutex);
P
Philipp Reisner 已提交
3503
	if (mdev->state.conn >= C_CONNECTED && mdev->state.disk < D_INCONSISTENT)
3504 3505 3506 3507
		updated_uuids |= drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]);

	if (updated_uuids)
		drbd_print_uuids(mdev, "receiver updated UUIDs to");
P
Philipp Reisner 已提交
3508

3509
	return 0;
P
Philipp Reisner 已提交
3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541
}

/**
 * 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[] = {
		[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;
}

3542
static int receive_req_state(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3543
{
3544
	struct drbd_conf *mdev;
3545
	struct p_req_state *p = pi->data;
P
Philipp Reisner 已提交
3546
	union drbd_state mask, val;
3547
	enum drbd_state_rv rv;
P
Philipp Reisner 已提交
3548

3549 3550 3551 3552
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;

P
Philipp Reisner 已提交
3553 3554 3555
	mask.i = be32_to_cpu(p->mask);
	val.i = be32_to_cpu(p->val);

3556
	if (test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) &&
3557
	    mutex_is_locked(mdev->state_mutex)) {
P
Philipp Reisner 已提交
3558
		drbd_send_sr_reply(mdev, SS_CONCURRENT_ST_CHG);
3559
		return 0;
P
Philipp Reisner 已提交
3560 3561 3562 3563 3564
	}

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

3565 3566
	rv = drbd_change_state(mdev, CS_VERBOSE, mask, val);
	drbd_send_sr_reply(mdev, rv);
P
Philipp Reisner 已提交
3567 3568 3569

	drbd_md_sync(mdev);

3570
	return 0;
P
Philipp Reisner 已提交
3571 3572
}

3573
static int receive_req_conn_state(struct drbd_tconn *tconn, struct packet_info *pi)
3574
{
3575
	struct p_req_state *p = pi->data;
3576 3577 3578 3579 3580 3581 3582 3583 3584
	union drbd_state mask, val;
	enum drbd_state_rv rv;

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

	if (test_bit(DISCARD_CONCURRENT, &tconn->flags) &&
	    mutex_is_locked(&tconn->cstate_mutex)) {
		conn_send_sr_reply(tconn, SS_CONCURRENT_ST_CHG);
3585
		return 0;
3586 3587 3588 3589 3590
	}

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

3591
	rv = conn_request_state(tconn, mask, val, CS_VERBOSE | CS_LOCAL_ONLY | CS_IGN_OUTD_FAIL);
3592 3593
	conn_send_sr_reply(tconn, rv);

3594
	return 0;
3595 3596
}

3597
static int receive_state(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3598
{
3599
	struct drbd_conf *mdev;
3600
	struct p_state *p = pi->data;
3601
	union drbd_state os, ns, peer_state;
P
Philipp Reisner 已提交
3602
	enum drbd_disk_state real_peer_disk;
3603
	enum chg_state_flags cs_flags;
P
Philipp Reisner 已提交
3604 3605
	int rv;

3606 3607 3608 3609
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3610 3611 3612 3613 3614 3615 3616 3617
	peer_state.i = be32_to_cpu(p->state);

	real_peer_disk = peer_state.disk;
	if (peer_state.disk == D_NEGOTIATING) {
		real_peer_disk = mdev->p_uuid[UI_FLAGS] & 4 ? D_INCONSISTENT : D_CONSISTENT;
		dev_info(DEV, "real peer disk state = %s\n", drbd_disk_str(real_peer_disk));
	}

3618
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
3619
 retry:
3620
	os = ns = drbd_read_state(mdev);
3621
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
3622

3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643
	/* peer says his disk is uptodate, while we think it is inconsistent,
	 * and this happens while we think we have a sync going on. */
	if (os.pdsk == D_INCONSISTENT && real_peer_disk == D_UP_TO_DATE &&
	    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) {
			if (drbd_bm_total_weight(mdev) <= mdev->rs_failed)
				drbd_resync_finished(mdev);
3644
			return 0;
3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656
		}
	}

	/* 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;

3657 3658
	if (ns.conn == C_WF_REPORT_PARAMS)
		ns.conn = C_CONNECTED;
P
Philipp Reisner 已提交
3659

3660 3661 3662
	if (peer_state.conn == C_AHEAD)
		ns.conn = C_BEHIND;

P
Philipp Reisner 已提交
3663 3664 3665 3666 3667
	if (mdev->p_uuid && peer_state.disk >= D_NEGOTIATING &&
	    get_ldev_if_state(mdev, D_NEGOTIATING)) {
		int cr; /* consider resync */

		/* if we established a new connection */
3668
		cr  = (os.conn < C_CONNECTED);
P
Philipp Reisner 已提交
3669 3670
		/* if we had an established connection
		 * and one of the nodes newly attaches a disk */
3671
		cr |= (os.conn == C_CONNECTED &&
P
Philipp Reisner 已提交
3672
		       (peer_state.disk == D_NEGOTIATING ||
3673
			os.disk == D_NEGOTIATING));
P
Philipp Reisner 已提交
3674 3675 3676 3677 3678
		/* if we have both been inconsistent, and the peer has been
		 * forced to be UpToDate with --overwrite-data */
		cr |= test_bit(CONSIDER_RESYNC, &mdev->flags);
		/* if we had been plain connected, and the admin requested to
		 * start a sync by "invalidate" or "invalidate-remote" */
3679
		cr |= (os.conn == C_CONNECTED &&
P
Philipp Reisner 已提交
3680 3681 3682 3683
				(peer_state.conn >= C_STARTING_SYNC_S &&
				 peer_state.conn <= C_WF_BITMAP_T));

		if (cr)
3684
			ns.conn = drbd_sync_handshake(mdev, peer_state.role, real_peer_disk);
P
Philipp Reisner 已提交
3685 3686

		put_ldev(mdev);
3687 3688
		if (ns.conn == C_MASK) {
			ns.conn = C_CONNECTED;
P
Philipp Reisner 已提交
3689
			if (mdev->state.disk == D_NEGOTIATING) {
3690
				drbd_force_state(mdev, NS(disk, D_FAILED));
P
Philipp Reisner 已提交
3691 3692 3693
			} else if (peer_state.disk == D_NEGOTIATING) {
				dev_err(DEV, "Disk attach process on the peer node was aborted.\n");
				peer_state.disk = D_DISKLESS;
3694
				real_peer_disk = D_DISKLESS;
P
Philipp Reisner 已提交
3695
			} else {
3696
				if (test_and_clear_bit(CONN_DRY_RUN, &mdev->tconn->flags))
3697
					return -EIO;
3698
				D_ASSERT(os.conn == C_WF_REPORT_PARAMS);
3699
				conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3700
				return -EIO;
P
Philipp Reisner 已提交
3701 3702 3703 3704
			}
		}
	}

3705
	spin_lock_irq(&mdev->tconn->req_lock);
3706
	if (os.i != drbd_read_state(mdev).i)
P
Philipp Reisner 已提交
3707 3708 3709 3710 3711
		goto retry;
	clear_bit(CONSIDER_RESYNC, &mdev->flags);
	ns.peer = peer_state.role;
	ns.pdsk = real_peer_disk;
	ns.peer_isp = (peer_state.aftr_isp | peer_state.user_isp);
3712
	if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING)
P
Philipp Reisner 已提交
3713
		ns.disk = mdev->new_state_tmp.disk;
3714
	cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD);
3715
	if (ns.pdsk == D_CONSISTENT && drbd_suspended(mdev) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED &&
3716
	    test_bit(NEW_CUR_UUID, &mdev->flags)) {
3717
		/* Do not allow tl_restart(RESEND) for a rebooted peer. We can only allow this
3718
		   for temporal network outages! */
3719
		spin_unlock_irq(&mdev->tconn->req_lock);
3720
		dev_err(DEV, "Aborting Connect, can not thaw IO with an only Consistent peer\n");
3721
		tl_clear(mdev->tconn);
3722 3723
		drbd_uuid_new_current(mdev);
		clear_bit(NEW_CUR_UUID, &mdev->flags);
3724
		conn_request_state(mdev->tconn, NS2(conn, C_PROTOCOL_ERROR, susp, 0), CS_HARD);
3725
		return -EIO;
3726
	}
3727
	rv = _drbd_set_state(mdev, ns, cs_flags, NULL);
3728
	ns = drbd_read_state(mdev);
3729
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
3730 3731

	if (rv < SS_SUCCESS) {
3732
		conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3733
		return -EIO;
P
Philipp Reisner 已提交
3734 3735
	}

3736 3737
	if (os.conn > C_WF_REPORT_PARAMS) {
		if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED &&
P
Philipp Reisner 已提交
3738 3739 3740 3741 3742 3743 3744 3745 3746
		    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 */
			drbd_send_uuids(mdev);
			drbd_send_state(mdev);
		}
	}

3747
	mutex_lock(&mdev->tconn->conf_update);
3748
	mdev->tconn->net_conf->want_lose = 0; /* without copy; single bit op is atomic */
3749
	mutex_unlock(&mdev->tconn->conf_update);
P
Philipp Reisner 已提交
3750 3751 3752

	drbd_md_sync(mdev); /* update connected indicator, la_size, ... */

3753
	return 0;
P
Philipp Reisner 已提交
3754 3755
}

3756
static int receive_sync_uuid(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3757
{
3758
	struct drbd_conf *mdev;
3759
	struct p_rs_uuid *p = pi->data;
3760 3761 3762 3763

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
3764 3765 3766

	wait_event(mdev->misc_wait,
		   mdev->state.conn == C_WF_SYNC_UUID ||
3767
		   mdev->state.conn == C_BEHIND ||
P
Philipp Reisner 已提交
3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778
		   mdev->state.conn < C_CONNECTED ||
		   mdev->state.disk < D_NEGOTIATING);

	/* D_ASSERT( mdev->state.conn == C_WF_SYNC_UUID ); */

	/* Here the _drbd_uuid_ functions are right, current should
	   _not_ be rotated into the history */
	if (get_ldev_if_state(mdev, D_NEGOTIATING)) {
		_drbd_uuid_set(mdev, UI_CURRENT, be64_to_cpu(p->uuid));
		_drbd_uuid_set(mdev, UI_BITMAP, 0UL);

3779
		drbd_print_uuids(mdev, "updated sync uuid");
P
Philipp Reisner 已提交
3780 3781 3782 3783 3784 3785
		drbd_start_resync(mdev, C_SYNC_TARGET);

		put_ldev(mdev);
	} else
		dev_err(DEV, "Ignoring SyncUUID packet!\n");

3786
	return 0;
P
Philipp Reisner 已提交
3787 3788
}

3789 3790 3791 3792 3793 3794 3795
/**
 * receive_bitmap_plain
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
3796
receive_bitmap_plain(struct drbd_conf *mdev, unsigned int size,
3797
		     unsigned long *p, struct bm_xfer_ctx *c)
P
Philipp Reisner 已提交
3798
{
3799 3800
	unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE -
				 drbd_header_size(mdev->tconn);
3801
	unsigned int num_words = min_t(size_t, data_size / sizeof(*p),
3802
				       c->bm_words - c->word_offset);
3803
	unsigned int want = num_words * sizeof(*p);
3804
	int err;
P
Philipp Reisner 已提交
3805

3806 3807
	if (want != size) {
		dev_err(DEV, "%s:want (%u) != size (%u)\n", __func__, want, size);
3808
		return -EIO;
P
Philipp Reisner 已提交
3809 3810
	}
	if (want == 0)
3811
		return 0;
3812
	err = drbd_recv_all(mdev->tconn, p, want);
3813
	if (err)
3814
		return err;
P
Philipp Reisner 已提交
3815

3816
	drbd_bm_merge_lel(mdev, c->word_offset, num_words, p);
P
Philipp Reisner 已提交
3817 3818 3819 3820 3821 3822

	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;

3823
	return 1;
P
Philipp Reisner 已提交
3824 3825
}

3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840
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;
}

3841 3842 3843 3844 3845 3846 3847
/**
 * recv_bm_rle_bits
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
P
Philipp Reisner 已提交
3848 3849
recv_bm_rle_bits(struct drbd_conf *mdev,
		struct p_compressed_bm *p,
3850 3851
		 struct bm_xfer_ctx *c,
		 unsigned int len)
P
Philipp Reisner 已提交
3852 3853 3854 3855 3856 3857 3858
{
	struct bitstream bs;
	u64 look_ahead;
	u64 rl;
	u64 tmp;
	unsigned long s = c->bit_offset;
	unsigned long e;
3859
	int toggle = dcbp_get_start(p);
P
Philipp Reisner 已提交
3860 3861 3862
	int have;
	int bits;

3863
	bitstream_init(&bs, p->code, len, dcbp_get_pad_bits(p));
P
Philipp Reisner 已提交
3864 3865 3866

	bits = bitstream_get_bits(&bs, &look_ahead, 64);
	if (bits < 0)
3867
		return -EIO;
P
Philipp Reisner 已提交
3868 3869 3870 3871

	for (have = bits; have > 0; s += rl, toggle = !toggle) {
		bits = vli_decode_bits(&rl, look_ahead);
		if (bits <= 0)
3872
			return -EIO;
P
Philipp Reisner 已提交
3873 3874 3875 3876 3877

		if (toggle) {
			e = s + rl -1;
			if (e >= c->bm_bits) {
				dev_err(DEV, "bitmap overflow (e:%lu) while decoding bm RLE packet\n", e);
3878
				return -EIO;
P
Philipp Reisner 已提交
3879 3880 3881 3882 3883 3884 3885 3886 3887
			}
			_drbd_bm_set_bits(mdev, s, e);
		}

		if (have < bits) {
			dev_err(DEV, "bitmap decoding error: h:%d b:%d la:0x%08llx l:%u/%u\n",
				have, bits, look_ahead,
				(unsigned int)(bs.cur.b - p->code),
				(unsigned int)bs.buf_len);
3888
			return -EIO;
P
Philipp Reisner 已提交
3889 3890 3891 3892 3893 3894
		}
		look_ahead >>= bits;
		have -= bits;

		bits = bitstream_get_bits(&bs, &tmp, 64 - have);
		if (bits < 0)
3895
			return -EIO;
P
Philipp Reisner 已提交
3896 3897 3898 3899 3900 3901 3902
		look_ahead |= tmp << have;
		have += bits;
	}

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

3903
	return (s != c->bm_bits);
P
Philipp Reisner 已提交
3904 3905
}

3906 3907 3908 3909 3910 3911 3912
/**
 * decode_bitmap_c
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
P
Philipp Reisner 已提交
3913 3914
decode_bitmap_c(struct drbd_conf *mdev,
		struct p_compressed_bm *p,
3915 3916
		struct bm_xfer_ctx *c,
		unsigned int len)
P
Philipp Reisner 已提交
3917
{
3918
	if (dcbp_get_code(p) == RLE_VLI_Bits)
3919
		return recv_bm_rle_bits(mdev, p, c, len - sizeof(*p));
P
Philipp Reisner 已提交
3920 3921 3922 3923 3924 3925

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

	dev_err(DEV, "receive_bitmap_c: unknown encoding %u\n", p->encoding);
3926
	conn_request_state(mdev->tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
3927
	return -EIO;
P
Philipp Reisner 已提交
3928 3929 3930 3931 3932 3933
}

void INFO_bm_xfer_stats(struct drbd_conf *mdev,
		const char *direction, struct bm_xfer_ctx *c)
{
	/* what would it take to transfer it "plaintext" */
3934 3935 3936 3937 3938 3939 3940
	unsigned int header_size = drbd_header_size(mdev->tconn);
	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
3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973

	/* 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;
	dev_info(DEV, "%s bitmap stats [Bytes(packets)]: plain %u(%u), RLE %u(%u), "
	     "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. */
3974
static int receive_bitmap(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3975
{
3976
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
3977
	struct bm_xfer_ctx c;
3978
	int err;
3979 3980 3981 3982

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
3983

3984 3985 3986
	drbd_bm_lock(mdev, "receive bitmap", BM_LOCKED_SET_ALLOWED);
	/* you are supposed to send additional out-of-sync information
	 * if you actually set bits during this phase */
P
Philipp Reisner 已提交
3987 3988 3989 3990 3991 3992

	c = (struct bm_xfer_ctx) {
		.bm_bits = drbd_bm_bits(mdev),
		.bm_words = drbd_bm_words(mdev),
	};

3993
	for(;;) {
3994 3995 3996
		if (pi->cmd == P_BITMAP)
			err = receive_bitmap_plain(mdev, pi->size, pi->data, &c);
		else if (pi->cmd == P_COMPRESSED_BITMAP) {
P
Philipp Reisner 已提交
3997 3998
			/* MAYBE: sanity check that we speak proto >= 90,
			 * and the feature is enabled! */
3999
			struct p_compressed_bm *p = pi->data;
P
Philipp Reisner 已提交
4000

4001
			if (pi->size > DRBD_SOCKET_BUFFER_SIZE - drbd_header_size(tconn)) {
P
Philipp Reisner 已提交
4002
				dev_err(DEV, "ReportCBitmap packet too large\n");
4003
				err = -EIO;
P
Philipp Reisner 已提交
4004 4005
				goto out;
			}
4006
			if (pi->size <= sizeof(*p)) {
4007
				dev_err(DEV, "ReportCBitmap packet too small (l:%u)\n", pi->size);
4008
				err = -EIO;
4009
				goto out;
P
Philipp Reisner 已提交
4010
			}
4011 4012 4013
			err = drbd_recv_all(mdev->tconn, p, pi->size);
			if (err)
			       goto out;
4014
			err = decode_bitmap_c(mdev, p, &c, pi->size);
P
Philipp Reisner 已提交
4015
		} else {
4016
			dev_warn(DEV, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", pi->cmd);
4017
			err = -EIO;
P
Philipp Reisner 已提交
4018 4019 4020
			goto out;
		}

4021
		c.packets[pi->cmd == P_BITMAP]++;
4022
		c.bytes[pi->cmd == P_BITMAP] += drbd_header_size(tconn) + pi->size;
P
Philipp Reisner 已提交
4023

4024 4025 4026
		if (err <= 0) {
			if (err < 0)
				goto out;
P
Philipp Reisner 已提交
4027
			break;
4028
		}
4029
		err = drbd_recv_header(mdev->tconn, pi);
4030
		if (err)
P
Philipp Reisner 已提交
4031
			goto out;
4032
	}
P
Philipp Reisner 已提交
4033 4034 4035 4036

	INFO_bm_xfer_stats(mdev, "receive", &c);

	if (mdev->state.conn == C_WF_BITMAP_T) {
4037 4038
		enum drbd_state_rv rv;

4039 4040
		err = drbd_send_bitmap(mdev);
		if (err)
P
Philipp Reisner 已提交
4041 4042
			goto out;
		/* Omit CS_ORDERED with this state transition to avoid deadlocks. */
4043 4044
		rv = _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE);
		D_ASSERT(rv == SS_SUCCESS);
P
Philipp Reisner 已提交
4045 4046 4047 4048 4049 4050
	} else if (mdev->state.conn != C_WF_BITMAP_S) {
		/* admin may have requested C_DISCONNECTING,
		 * other threads may have noticed network errors */
		dev_info(DEV, "unexpected cstate (%s) in receive_bitmap\n",
		    drbd_conn_str(mdev->state.conn));
	}
4051
	err = 0;
P
Philipp Reisner 已提交
4052 4053

 out:
4054
	drbd_bm_unlock(mdev);
4055
	if (!err && mdev->state.conn == C_WF_BITMAP_S)
P
Philipp Reisner 已提交
4056
		drbd_start_resync(mdev, C_SYNC_SOURCE);
4057
	return err;
P
Philipp Reisner 已提交
4058 4059
}

4060
static int receive_skip(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4061
{
4062
	conn_warn(tconn, "skipping unknown optional packet type %d, l: %d!\n",
4063
		 pi->cmd, pi->size);
4064

4065
	return ignore_remaining_packet(tconn, pi);
4066 4067
}

4068
static int receive_UnplugRemote(struct drbd_tconn *tconn, struct packet_info *pi)
4069
{
4070 4071
	/* Make sure we've acked all the TCP data associated
	 * with the data requests being unplugged */
4072
	drbd_tcp_quickack(tconn->data.socket);
4073

4074
	return 0;
4075 4076
}

4077
static int receive_out_of_sync(struct drbd_tconn *tconn, struct packet_info *pi)
4078
{
4079
	struct drbd_conf *mdev;
4080
	struct p_block_desc *p = pi->data;
4081 4082 4083 4084

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
4085

4086 4087 4088 4089 4090 4091 4092 4093 4094 4095
	switch (mdev->state.conn) {
	case C_WF_SYNC_UUID:
	case C_WF_BITMAP_T:
	case C_BEHIND:
			break;
	default:
		dev_err(DEV, "ASSERT FAILED cstate = %s, expected: WFSyncUUID|WFBitMapT|Behind\n",
				drbd_conn_str(mdev->state.conn));
	}

4096 4097
	drbd_set_out_of_sync(mdev, be64_to_cpu(p->sector), be32_to_cpu(p->blksize));

4098
	return 0;
4099 4100
}

4101 4102 4103
struct data_cmd {
	int expect_payload;
	size_t pkt_size;
4104
	int (*fn)(struct drbd_tconn *, struct packet_info *);
4105 4106 4107
};

static struct data_cmd drbd_cmd_handler[] = {
4108 4109 4110 4111
	[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 } ,
4112 4113 4114
	[P_BITMAP]	    = { 1, 0, receive_bitmap } ,
	[P_COMPRESSED_BITMAP] = { 1, 0, receive_bitmap } ,
	[P_UNPLUG_REMOTE]   = { 0, 0, receive_UnplugRemote },
4115 4116
	[P_DATA_REQUEST]    = { 0, sizeof(struct p_block_req), receive_DataRequest },
	[P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest },
4117 4118
	[P_SYNC_PARAM]	    = { 1, 0, receive_SyncParam },
	[P_SYNC_PARAM89]    = { 1, 0, receive_SyncParam },
4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130
	[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 },
	[P_OUT_OF_SYNC]     = { 0, sizeof(struct p_block_desc), receive_out_of_sync },
	[P_CONN_ST_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_conn_state },
P
Philipp Reisner 已提交
4131 4132
};

4133
static void drbdd(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4134
{
4135
	struct packet_info pi;
4136
	size_t shs; /* sub header size */
4137
	int err;
P
Philipp Reisner 已提交
4138

4139
	while (get_t_state(&tconn->receiver) == RUNNING) {
4140 4141
		struct data_cmd *cmd;

4142
		drbd_thread_current_set_cpu(&tconn->receiver);
4143
		if (drbd_recv_header(tconn, &pi))
4144
			goto err_out;
P
Philipp Reisner 已提交
4145

4146
		cmd = &drbd_cmd_handler[pi.cmd];
4147
		if (unlikely(pi.cmd >= ARRAY_SIZE(drbd_cmd_handler) || !cmd->fn)) {
4148
			conn_err(tconn, "unknown packet type %d, l: %d!\n", pi.cmd, pi.size);
4149
			goto err_out;
4150
		}
P
Philipp Reisner 已提交
4151

4152 4153
		shs = cmd->pkt_size;
		if (pi.size > shs && !cmd->expect_payload) {
4154
			conn_err(tconn, "No payload expected %s l:%d\n", cmdname(pi.cmd), pi.size);
4155
			goto err_out;
P
Philipp Reisner 已提交
4156 4157
		}

4158
		if (shs) {
4159
			err = drbd_recv_all_warn(tconn, pi.data, shs);
4160
			if (err)
4161
				goto err_out;
4162
			pi.size -= shs;
4163 4164
		}

4165 4166
		err = cmd->fn(tconn, &pi);
		if (err) {
4167 4168
			conn_err(tconn, "error receiving %s, e: %d l: %d!\n",
				 cmdname(pi.cmd), err, pi.size);
4169
			goto err_out;
P
Philipp Reisner 已提交
4170 4171
		}
	}
4172
	return;
P
Philipp Reisner 已提交
4173

4174 4175
    err_out:
	conn_request_state(tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
P
Philipp Reisner 已提交
4176 4177
}

4178
void conn_flush_workqueue(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4179 4180 4181 4182
{
	struct drbd_wq_barrier barr;

	barr.w.cb = w_prev_work_done;
4183
	barr.w.tconn = tconn;
P
Philipp Reisner 已提交
4184
	init_completion(&barr.done);
4185
	drbd_queue_work(&tconn->data.work, &barr.w);
P
Philipp Reisner 已提交
4186 4187 4188
	wait_for_completion(&barr.done);
}

4189
static void drbd_disconnect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4190
{
4191
	enum drbd_conns oc;
P
Philipp Reisner 已提交
4192 4193
	int rv = SS_UNKNOWN_ERROR;

4194
	if (tconn->cstate == C_STANDALONE)
P
Philipp Reisner 已提交
4195 4196 4197
		return;

	/* asender does not clean up anything. it must not interfere, either */
4198 4199 4200
	drbd_thread_stop(&tconn->asender);
	drbd_free_sock(tconn);

4201
	down_read(&drbd_cfg_rwsem);
4202
	idr_for_each(&tconn->volumes, drbd_disconnected, tconn);
4203
	up_read(&drbd_cfg_rwsem);
4204 4205
	conn_info(tconn, "Connection closed\n");

4206 4207 4208
	if (conn_highest_role(tconn) == R_PRIMARY && conn_highest_pdsk(tconn) >= D_UNKNOWN)
		conn_try_outdate_peer_async(tconn);

4209
	spin_lock_irq(&tconn->req_lock);
4210 4211 4212 4213
	oc = tconn->cstate;
	if (oc >= C_UNCONNECTED)
		rv = _conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE);

4214 4215
	spin_unlock_irq(&tconn->req_lock);

4216
	if (oc == C_DISCONNECTING)
4217
		conn_request_state(tconn, NS(conn, C_STANDALONE), CS_VERBOSE | CS_HARD);
4218 4219 4220 4221 4222 4223 4224
}

static int drbd_disconnected(int vnr, void *p, void *data)
{
	struct drbd_conf *mdev = (struct drbd_conf *)p;
	enum drbd_fencing_p fp;
	unsigned int i;
P
Philipp Reisner 已提交
4225

4226
	/* wait for current activity to cease. */
4227
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
4228 4229 4230
	_drbd_wait_ee_list_empty(mdev, &mdev->active_ee);
	_drbd_wait_ee_list_empty(mdev, &mdev->sync_ee);
	_drbd_wait_ee_list_empty(mdev, &mdev->read_ee);
4231
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248

	/* 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. */
	drbd_rs_cancel_all(mdev);
	mdev->rs_total = 0;
	mdev->rs_failed = 0;
	atomic_set(&mdev->rs_pending_cnt, 0);
	wake_up(&mdev->misc_wait);

4249 4250
	del_timer(&mdev->request_timer);

P
Philipp Reisner 已提交
4251 4252 4253 4254 4255 4256
	del_timer_sync(&mdev->resync_timer);
	resync_timer_fn((unsigned long)mdev);

	/* 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" */
4257
	drbd_flush_workqueue(mdev);
P
Philipp Reisner 已提交
4258

4259
	drbd_finish_peer_reqs(mdev);
P
Philipp Reisner 已提交
4260 4261 4262 4263

	kfree(mdev->p_uuid);
	mdev->p_uuid = NULL;

4264
	if (!drbd_suspended(mdev))
4265
		tl_clear(mdev->tconn);
P
Philipp Reisner 已提交
4266 4267 4268 4269 4270 4271 4272 4273 4274

	drbd_md_sync(mdev);

	fp = FP_DONT_CARE;
	if (get_ldev(mdev)) {
		fp = mdev->ldev->dc.fencing;
		put_ldev(mdev);
	}

4275 4276 4277 4278
	/* serialize with bitmap writeout triggered by the state change,
	 * if any. */
	wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));

P
Philipp Reisner 已提交
4279 4280 4281 4282 4283 4284 4285
	/* 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.
	 */
4286
	i = drbd_free_peer_reqs(mdev, &mdev->net_ee);
P
Philipp Reisner 已提交
4287 4288
	if (i)
		dev_info(DEV, "net_ee not empty, killed %u entries\n", i);
4289 4290 4291
	i = atomic_read(&mdev->pp_in_use_by_net);
	if (i)
		dev_info(DEV, "pp_in_use_by_net = %d, expected 0\n", i);
P
Philipp Reisner 已提交
4292 4293
	i = atomic_read(&mdev->pp_in_use);
	if (i)
4294
		dev_info(DEV, "pp_in_use = %d, expected 0\n", i);
P
Philipp Reisner 已提交
4295 4296 4297 4298 4299 4300 4301 4302 4303

	D_ASSERT(list_empty(&mdev->read_ee));
	D_ASSERT(list_empty(&mdev->active_ee));
	D_ASSERT(list_empty(&mdev->sync_ee));
	D_ASSERT(list_empty(&mdev->done_ee));

	/* ok, no more ee's on the fly, it is safe to reset the epoch_size */
	atomic_set(&mdev->current_epoch->epoch_size, 0);
	D_ASSERT(list_empty(&mdev->current_epoch->list));
4304 4305

	return 0;
P
Philipp Reisner 已提交
4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316
}

/*
 * 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.
 */
4317
static int drbd_send_features(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4318
{
4319 4320
	struct drbd_socket *sock;
	struct p_connection_features *p;
P
Philipp Reisner 已提交
4321

4322 4323 4324
	sock = &tconn->data;
	p = conn_prepare_command(tconn, sock);
	if (!p)
4325
		return -EIO;
P
Philipp Reisner 已提交
4326 4327 4328
	memset(p, 0, sizeof(*p));
	p->protocol_min = cpu_to_be32(PRO_VERSION_MIN);
	p->protocol_max = cpu_to_be32(PRO_VERSION_MAX);
4329
	return conn_send_command(tconn, sock, P_CONNECTION_FEATURES, sizeof(*p), NULL, 0);
P
Philipp Reisner 已提交
4330 4331 4332 4333 4334 4335 4336 4337 4338
}

/*
 * 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.
 */
4339
static int drbd_do_features(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4340
{
4341
	/* ASSERT current == tconn->receiver ... */
4342 4343
	struct p_connection_features *p;
	const int expect = sizeof(struct p_connection_features);
4344
	struct packet_info pi;
4345
	int err;
P
Philipp Reisner 已提交
4346

4347
	err = drbd_send_features(tconn);
4348
	if (err)
P
Philipp Reisner 已提交
4349 4350
		return 0;

4351 4352
	err = drbd_recv_header(tconn, &pi);
	if (err)
P
Philipp Reisner 已提交
4353 4354
		return 0;

4355 4356
	if (pi.cmd != P_CONNECTION_FEATURES) {
		conn_err(tconn, "expected ConnectionFeatures packet, received: %s (0x%04x)\n",
4357
		     cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4358 4359 4360
		return -1;
	}

4361
	if (pi.size != expect) {
4362
		conn_err(tconn, "expected ConnectionFeatures length: %u, received: %u\n",
4363
		     expect, pi.size);
P
Philipp Reisner 已提交
4364 4365 4366
		return -1;
	}

4367 4368
	p = pi.data;
	err = drbd_recv_all_warn(tconn, p, expect);
4369
	if (err)
P
Philipp Reisner 已提交
4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380
		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;

4381
	tconn->agreed_pro_version = min_t(int, PRO_VERSION_MAX, p->protocol_max);
P
Philipp Reisner 已提交
4382

4383 4384
	conn_info(tconn, "Handshake successful: "
	     "Agreed network protocol version %d\n", tconn->agreed_pro_version);
P
Philipp Reisner 已提交
4385 4386 4387 4388

	return 1;

 incompat:
4389
	conn_err(tconn, "incompatible DRBD dialects: "
P
Philipp Reisner 已提交
4390 4391 4392 4393 4394 4395 4396
	    "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)
4397
static int drbd_do_auth(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4398 4399 4400
{
	dev_err(DEV, "This kernel was build without CONFIG_CRYPTO_HMAC.\n");
	dev_err(DEV, "You need to disable 'cram-hmac-alg' in drbd.conf.\n");
4401
	return -1;
P
Philipp Reisner 已提交
4402 4403 4404
}
#else
#define CHALLENGE_LEN 64
4405 4406 4407 4408 4409 4410 4411

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

4412
static int drbd_do_auth(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4413
{
4414
	struct drbd_socket *sock;
P
Philipp Reisner 已提交
4415 4416 4417 4418 4419
	char my_challenge[CHALLENGE_LEN];  /* 64 Bytes... */
	struct scatterlist sg;
	char *response = NULL;
	char *right_response = NULL;
	char *peers_ch = NULL;
4420 4421
	unsigned int key_len;
	char secret[SHARED_SECRET_MAX]; /* 64 byte */
P
Philipp Reisner 已提交
4422 4423
	unsigned int resp_size;
	struct hash_desc desc;
4424
	struct packet_info pi;
4425
	struct net_conf *nc;
4426
	int err, rv;
P
Philipp Reisner 已提交
4427

4428 4429
	/* FIXME: Put the challenge/response into the preallocated socket buffer.  */

4430 4431 4432 4433 4434 4435
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);
	key_len = strlen(nc->shared_secret);
	memcpy(secret, nc->shared_secret, key_len);
	rcu_read_unlock();

4436
	desc.tfm = tconn->cram_hmac_tfm;
P
Philipp Reisner 已提交
4437 4438
	desc.flags = 0;

4439
	rv = crypto_hash_setkey(tconn->cram_hmac_tfm, (u8 *)secret, key_len);
P
Philipp Reisner 已提交
4440
	if (rv) {
4441
		conn_err(tconn, "crypto_hash_setkey() failed with %d\n", rv);
4442
		rv = -1;
P
Philipp Reisner 已提交
4443 4444 4445 4446 4447
		goto fail;
	}

	get_random_bytes(my_challenge, CHALLENGE_LEN);

4448 4449 4450 4451 4452
	sock = &tconn->data;
	if (!conn_prepare_command(tconn, sock)) {
		rv = 0;
		goto fail;
	}
4453
	rv = !conn_send_command(tconn, sock, P_AUTH_CHALLENGE, 0,
4454
				my_challenge, CHALLENGE_LEN);
P
Philipp Reisner 已提交
4455 4456 4457
	if (!rv)
		goto fail;

4458 4459 4460
	err = drbd_recv_header(tconn, &pi);
	if (err) {
		rv = 0;
P
Philipp Reisner 已提交
4461
		goto fail;
4462
	}
P
Philipp Reisner 已提交
4463

4464
	if (pi.cmd != P_AUTH_CHALLENGE) {
4465
		conn_err(tconn, "expected AuthChallenge packet, received: %s (0x%04x)\n",
4466
		    cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4467 4468 4469 4470
		rv = 0;
		goto fail;
	}

4471
	if (pi.size > CHALLENGE_LEN * 2) {
4472
		conn_err(tconn, "expected AuthChallenge payload too big.\n");
4473
		rv = -1;
P
Philipp Reisner 已提交
4474 4475 4476
		goto fail;
	}

4477
	peers_ch = kmalloc(pi.size, GFP_NOIO);
P
Philipp Reisner 已提交
4478
	if (peers_ch == NULL) {
4479
		conn_err(tconn, "kmalloc of peers_ch failed\n");
4480
		rv = -1;
P
Philipp Reisner 已提交
4481 4482 4483
		goto fail;
	}

4484 4485
	err = drbd_recv_all_warn(tconn, peers_ch, pi.size);
	if (err) {
P
Philipp Reisner 已提交
4486 4487 4488 4489
		rv = 0;
		goto fail;
	}

4490
	resp_size = crypto_hash_digestsize(tconn->cram_hmac_tfm);
P
Philipp Reisner 已提交
4491 4492
	response = kmalloc(resp_size, GFP_NOIO);
	if (response == NULL) {
4493
		conn_err(tconn, "kmalloc of response failed\n");
4494
		rv = -1;
P
Philipp Reisner 已提交
4495 4496 4497 4498
		goto fail;
	}

	sg_init_table(&sg, 1);
4499
	sg_set_buf(&sg, peers_ch, pi.size);
P
Philipp Reisner 已提交
4500 4501 4502

	rv = crypto_hash_digest(&desc, &sg, sg.length, response);
	if (rv) {
4503
		conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv);
4504
		rv = -1;
P
Philipp Reisner 已提交
4505 4506 4507
		goto fail;
	}

4508 4509 4510 4511
	if (!conn_prepare_command(tconn, sock)) {
		rv = 0;
		goto fail;
	}
4512
	rv = !conn_send_command(tconn, sock, P_AUTH_RESPONSE, 0,
4513
				response, resp_size);
P
Philipp Reisner 已提交
4514 4515 4516
	if (!rv)
		goto fail;

4517 4518 4519
	err = drbd_recv_header(tconn, &pi);
	if (err) {
		rv = 0;
P
Philipp Reisner 已提交
4520
		goto fail;
4521
	}
P
Philipp Reisner 已提交
4522

4523
	if (pi.cmd != P_AUTH_RESPONSE) {
4524
		conn_err(tconn, "expected AuthResponse packet, received: %s (0x%04x)\n",
4525
			cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4526 4527 4528 4529
		rv = 0;
		goto fail;
	}

4530
	if (pi.size != resp_size) {
4531
		conn_err(tconn, "expected AuthResponse payload of wrong size\n");
P
Philipp Reisner 已提交
4532 4533 4534 4535
		rv = 0;
		goto fail;
	}

4536 4537
	err = drbd_recv_all_warn(tconn, response , resp_size);
	if (err) {
P
Philipp Reisner 已提交
4538 4539 4540 4541 4542
		rv = 0;
		goto fail;
	}

	right_response = kmalloc(resp_size, GFP_NOIO);
4543
	if (right_response == NULL) {
4544
		conn_err(tconn, "kmalloc of right_response failed\n");
4545
		rv = -1;
P
Philipp Reisner 已提交
4546 4547 4548 4549 4550 4551 4552
		goto fail;
	}

	sg_set_buf(&sg, my_challenge, CHALLENGE_LEN);

	rv = crypto_hash_digest(&desc, &sg, sg.length, right_response);
	if (rv) {
4553
		conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv);
4554
		rv = -1;
P
Philipp Reisner 已提交
4555 4556 4557 4558 4559 4560
		goto fail;
	}

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

	if (rv)
4561 4562
		conn_info(tconn, "Peer authenticated using %d bytes HMAC\n",
		     resp_size);
4563 4564
	else
		rv = -1;
P
Philipp Reisner 已提交
4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576

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

	return rv;
}
#endif

int drbdd_init(struct drbd_thread *thi)
{
4577
	struct drbd_tconn *tconn = thi->tconn;
P
Philipp Reisner 已提交
4578 4579
	int h;

4580
	conn_info(tconn, "receiver (re)started\n");
P
Philipp Reisner 已提交
4581 4582

	do {
4583
		h = drbd_connect(tconn);
P
Philipp Reisner 已提交
4584
		if (h == 0) {
4585
			drbd_disconnect(tconn);
4586
			schedule_timeout_interruptible(HZ);
P
Philipp Reisner 已提交
4587 4588
		}
		if (h == -1) {
4589
			conn_warn(tconn, "Discarding network configuration.\n");
4590
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
4591 4592 4593
		}
	} while (h == 0);

4594 4595
	if (h > 0)
		drbdd(tconn);
P
Philipp Reisner 已提交
4596

4597
	drbd_disconnect(tconn);
P
Philipp Reisner 已提交
4598

4599
	conn_info(tconn, "receiver terminated\n");
P
Philipp Reisner 已提交
4600 4601 4602 4603 4604
	return 0;
}

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

4605
static int got_conn_RqSReply(struct drbd_tconn *tconn, struct packet_info *pi)
4606
{
4607
	struct p_req_state_reply *p = pi->data;
4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618
	int retcode = be32_to_cpu(p->retcode);

	if (retcode >= SS_SUCCESS) {
		set_bit(CONN_WD_ST_CHG_OKAY, &tconn->flags);
	} else {
		set_bit(CONN_WD_ST_CHG_FAIL, &tconn->flags);
		conn_err(tconn, "Requested state change failed by peer: %s (%d)\n",
			 drbd_set_st_err_str(retcode), retcode);
	}
	wake_up(&tconn->ping_wait);

4619
	return 0;
4620 4621
}

4622
static int got_RqSReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4623
{
4624
	struct drbd_conf *mdev;
4625
	struct p_req_state_reply *p = pi->data;
P
Philipp Reisner 已提交
4626 4627
	int retcode = be32_to_cpu(p->retcode);

4628 4629
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4630
		return -EIO;
4631

4632 4633 4634 4635 4636 4637
	if (retcode >= SS_SUCCESS) {
		set_bit(CL_ST_CHG_SUCCESS, &mdev->flags);
	} else {
		set_bit(CL_ST_CHG_FAIL, &mdev->flags);
		dev_err(DEV, "Requested state change failed by peer: %s (%d)\n",
			drbd_set_st_err_str(retcode), retcode);
P
Philipp Reisner 已提交
4638
	}
4639 4640
	wake_up(&mdev->state_wait);

4641
	return 0;
P
Philipp Reisner 已提交
4642 4643
}

4644
static int got_Ping(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4645
{
4646
	return drbd_send_ping_ack(tconn);
P
Philipp Reisner 已提交
4647 4648 4649

}

4650
static int got_PingAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4651 4652
{
	/* restore idle timeout */
4653 4654 4655
	tconn->meta.socket->sk->sk_rcvtimeo = tconn->net_conf->ping_int*HZ;
	if (!test_and_set_bit(GOT_PING_ACK, &tconn->flags))
		wake_up(&tconn->ping_wait);
P
Philipp Reisner 已提交
4656

4657
	return 0;
P
Philipp Reisner 已提交
4658 4659
}

4660
static int got_IsInSync(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4661
{
4662
	struct drbd_conf *mdev;
4663
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4664 4665 4666
	sector_t sector = be64_to_cpu(p->sector);
	int blksize = be32_to_cpu(p->blksize);

4667 4668
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4669
		return -EIO;
4670

4671
	D_ASSERT(mdev->tconn->agreed_pro_version >= 89);
P
Philipp Reisner 已提交
4672 4673 4674

	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

4675 4676 4677 4678 4679 4680 4681
	if (get_ldev(mdev)) {
		drbd_rs_complete_io(mdev, sector);
		drbd_set_in_sync(mdev, sector, blksize);
		/* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */
		mdev->rs_same_csum += (blksize >> BM_BLOCK_SHIFT);
		put_ldev(mdev);
	}
P
Philipp Reisner 已提交
4682
	dec_rs_pending(mdev);
4683
	atomic_add(blksize >> 9, &mdev->rs_sect_in);
P
Philipp Reisner 已提交
4684

4685
	return 0;
P
Philipp Reisner 已提交
4686 4687
}

4688 4689 4690 4691
static int
validate_req_change_req_state(struct drbd_conf *mdev, u64 id, sector_t sector,
			      struct rb_root *root, const char *func,
			      enum drbd_req_event what, bool missing_ok)
P
Philipp Reisner 已提交
4692 4693 4694 4695
{
	struct drbd_request *req;
	struct bio_and_error m;

4696
	spin_lock_irq(&mdev->tconn->req_lock);
4697
	req = find_request(mdev, root, id, sector, missing_ok, func);
P
Philipp Reisner 已提交
4698
	if (unlikely(!req)) {
4699
		spin_unlock_irq(&mdev->tconn->req_lock);
4700
		return -EIO;
P
Philipp Reisner 已提交
4701 4702
	}
	__req_mod(req, what, &m);
4703
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
4704 4705 4706

	if (m.bio)
		complete_master_bio(mdev, &m);
4707
	return 0;
P
Philipp Reisner 已提交
4708 4709
}

4710
static int got_BlockAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4711
{
4712
	struct drbd_conf *mdev;
4713
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4714 4715 4716 4717
	sector_t sector = be64_to_cpu(p->sector);
	int blksize = be32_to_cpu(p->blksize);
	enum drbd_req_event what;

4718 4719
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4720
		return -EIO;
4721

P
Philipp Reisner 已提交
4722 4723
	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

4724
	if (p->block_id == ID_SYNCER) {
P
Philipp Reisner 已提交
4725 4726
		drbd_set_in_sync(mdev, sector, blksize);
		dec_rs_pending(mdev);
4727
		return 0;
P
Philipp Reisner 已提交
4728
	}
4729
	switch (pi->cmd) {
P
Philipp Reisner 已提交
4730
	case P_RS_WRITE_ACK:
4731
		what = WRITE_ACKED_BY_PEER_AND_SIS;
P
Philipp Reisner 已提交
4732 4733
		break;
	case P_WRITE_ACK:
4734
		what = WRITE_ACKED_BY_PEER;
P
Philipp Reisner 已提交
4735 4736
		break;
	case P_RECV_ACK:
4737
		what = RECV_ACKED_BY_PEER;
P
Philipp Reisner 已提交
4738
		break;
4739 4740 4741 4742 4743
	case P_DISCARD_WRITE:
		what = DISCARD_WRITE;
		break;
	case P_RETRY_WRITE:
		what = POSTPONE_WRITE;
P
Philipp Reisner 已提交
4744 4745
		break;
	default:
4746
		BUG();
P
Philipp Reisner 已提交
4747 4748
	}

4749 4750 4751
	return validate_req_change_req_state(mdev, p->block_id, sector,
					     &mdev->write_requests, __func__,
					     what, false);
P
Philipp Reisner 已提交
4752 4753
}

4754
static int got_NegAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4755
{
4756
	struct drbd_conf *mdev;
4757
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4758
	sector_t sector = be64_to_cpu(p->sector);
4759
	int size = be32_to_cpu(p->blksize);
4760
	int err;
P
Philipp Reisner 已提交
4761

4762 4763
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4764
		return -EIO;
4765

P
Philipp Reisner 已提交
4766 4767
	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

4768
	if (p->block_id == ID_SYNCER) {
P
Philipp Reisner 已提交
4769 4770
		dec_rs_pending(mdev);
		drbd_rs_failed_io(mdev, sector, size);
4771
		return 0;
P
Philipp Reisner 已提交
4772
	}
4773

4774 4775
	err = validate_req_change_req_state(mdev, p->block_id, sector,
					    &mdev->write_requests, __func__,
4776
					    NEG_ACKED, true);
4777
	if (err) {
4778 4779 4780 4781 4782 4783
		/* 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. */
		drbd_set_out_of_sync(mdev, sector, size);
4784
	}
4785
	return 0;
P
Philipp Reisner 已提交
4786 4787
}

4788
static int got_NegDReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4789
{
4790
	struct drbd_conf *mdev;
4791
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4792 4793
	sector_t sector = be64_to_cpu(p->sector);

4794 4795
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4796
		return -EIO;
4797

P
Philipp Reisner 已提交
4798
	update_peer_seq(mdev, be32_to_cpu(p->seq_num));
4799

P
Philipp Reisner 已提交
4800 4801 4802
	dev_err(DEV, "Got NegDReply; Sector %llus, len %u; Fail original request.\n",
	    (unsigned long long)sector, be32_to_cpu(p->blksize));

4803 4804 4805
	return validate_req_change_req_state(mdev, p->block_id, sector,
					     &mdev->read_requests, __func__,
					     NEG_ACKED, false);
P
Philipp Reisner 已提交
4806 4807
}

4808
static int got_NegRSDReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4809
{
4810
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
4811 4812
	sector_t sector;
	int size;
4813
	struct p_block_ack *p = pi->data;
4814 4815 4816

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4817
		return -EIO;
P
Philipp Reisner 已提交
4818 4819 4820 4821 4822 4823 4824 4825 4826 4827

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

	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

	dec_rs_pending(mdev);

	if (get_ldev_if_state(mdev, D_FAILED)) {
		drbd_rs_complete_io(mdev, sector);
4828
		switch (pi->cmd) {
4829 4830 4831 4832 4833
		case P_NEG_RS_DREPLY:
			drbd_rs_failed_io(mdev, sector, size);
		case P_RS_CANCEL:
			break;
		default:
4834
			BUG();
4835
		}
P
Philipp Reisner 已提交
4836 4837 4838
		put_ldev(mdev);
	}

4839
	return 0;
P
Philipp Reisner 已提交
4840 4841
}

4842
static int got_BarrierAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4843
{
4844
	struct drbd_conf *mdev;
4845
	struct p_barrier_ack *p = pi->data;
4846 4847 4848

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4849
		return -EIO;
P
Philipp Reisner 已提交
4850

4851
	tl_release(mdev->tconn, p->barrier, be32_to_cpu(p->set_size));
P
Philipp Reisner 已提交
4852

4853 4854
	if (mdev->state.conn == C_AHEAD &&
	    atomic_read(&mdev->ap_in_flight) == 0 &&
4855 4856 4857
	    !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &mdev->current_epoch->flags)) {
		mdev->start_resync_timer.expires = jiffies + HZ;
		add_timer(&mdev->start_resync_timer);
4858 4859
	}

4860
	return 0;
P
Philipp Reisner 已提交
4861 4862
}

4863
static int got_OVResult(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4864
{
4865
	struct drbd_conf *mdev;
4866
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4867 4868 4869 4870
	struct drbd_work *w;
	sector_t sector;
	int size;

4871 4872
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4873
		return -EIO;
4874

P
Philipp Reisner 已提交
4875 4876 4877 4878 4879 4880
	sector = be64_to_cpu(p->sector);
	size = be32_to_cpu(p->blksize);

	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

	if (be64_to_cpu(p->block_id) == ID_OUT_OF_SYNC)
4881
		drbd_ov_out_of_sync_found(mdev, sector, size);
P
Philipp Reisner 已提交
4882
	else
4883
		ov_out_of_sync_print(mdev);
P
Philipp Reisner 已提交
4884

4885
	if (!get_ldev(mdev))
4886
		return 0;
4887

P
Philipp Reisner 已提交
4888 4889 4890
	drbd_rs_complete_io(mdev, sector);
	dec_rs_pending(mdev);

4891 4892 4893 4894 4895 4896 4897
	--mdev->ov_left;

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

	if (mdev->ov_left == 0) {
P
Philipp Reisner 已提交
4898 4899 4900
		w = kmalloc(sizeof(*w), GFP_NOIO);
		if (w) {
			w->cb = w_ov_finished;
4901
			w->mdev = mdev;
4902
			drbd_queue_work_front(&mdev->tconn->data.work, w);
P
Philipp Reisner 已提交
4903 4904
		} else {
			dev_err(DEV, "kmalloc(w) failed.");
4905
			ov_out_of_sync_print(mdev);
P
Philipp Reisner 已提交
4906 4907 4908
			drbd_resync_finished(mdev);
		}
	}
4909
	put_ldev(mdev);
4910
	return 0;
P
Philipp Reisner 已提交
4911 4912
}

4913
static int got_skip(struct drbd_tconn *tconn, struct packet_info *pi)
4914
{
4915
	return 0;
4916 4917
}

4918
static int tconn_finish_peer_reqs(struct drbd_tconn *tconn)
4919
{
4920 4921
	struct drbd_conf *mdev;
	int i, not_empty = 0;
4922 4923 4924 4925

	do {
		clear_bit(SIGNAL_ASENDER, &tconn->flags);
		flush_signals(current);
4926
		down_read(&drbd_cfg_rwsem);
4927
		idr_for_each_entry(&tconn->volumes, mdev, i) {
4928 4929
			if (drbd_finish_peer_reqs(mdev)) {
				up_read(&drbd_cfg_rwsem);
4930
				return 1; /* error */
4931
			}
4932
		}
4933
		up_read(&drbd_cfg_rwsem);
4934
		set_bit(SIGNAL_ASENDER, &tconn->flags);
4935 4936

		spin_lock_irq(&tconn->req_lock);
4937
		rcu_read_lock();
4938 4939 4940 4941 4942
		idr_for_each_entry(&tconn->volumes, mdev, i) {
			not_empty = !list_empty(&mdev->done_ee);
			if (not_empty)
				break;
		}
4943
		rcu_read_unlock();
4944
		spin_unlock_irq(&tconn->req_lock);
4945 4946 4947 4948 4949
	} while (not_empty);

	return 0;
}

4950 4951
struct asender_cmd {
	size_t pkt_size;
4952
	int (*fn)(struct drbd_tconn *tconn, struct packet_info *);
4953 4954 4955
};

static struct asender_cmd asender_tbl[] = {
4956 4957
	[P_PING]	    = { 0, got_Ping },
	[P_PING_ACK]	    = { 0, got_PingAck },
4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972
	[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 },
	[P_DISCARD_WRITE]   = { sizeof(struct p_block_ack), got_BlockAck },
	[P_NEG_ACK]	    = { sizeof(struct p_block_ack), got_NegAck },
	[P_NEG_DREPLY]	    = { sizeof(struct p_block_ack), got_NegDReply },
	[P_NEG_RS_DREPLY]   = { sizeof(struct p_block_ack), got_NegRSDReply },
	[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 },
	[P_DELAY_PROBE]     = { sizeof(struct p_delay_probe93), got_skip },
	[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 },
4973 4974
};

P
Philipp Reisner 已提交
4975 4976
int drbd_asender(struct drbd_thread *thi)
{
4977
	struct drbd_tconn *tconn = thi->tconn;
P
Philipp Reisner 已提交
4978
	struct asender_cmd *cmd = NULL;
4979
	struct packet_info pi;
4980
	int rv;
4981
	void *buf    = tconn->meta.rbuf;
P
Philipp Reisner 已提交
4982
	int received = 0;
4983 4984
	unsigned int header_size = drbd_header_size(tconn);
	int expect   = header_size;
4985 4986 4987
	bool ping_timeout_active = false;
	struct net_conf *nc;
	int ping_timeo, no_cork, ping_int;
P
Philipp Reisner 已提交
4988 4989 4990 4991

	current->policy = SCHED_RR;  /* Make this a realtime task! */
	current->rt_priority = 2;    /* more important than all other tasks */

4992
	while (get_t_state(thi) == RUNNING) {
4993
		drbd_thread_current_set_cpu(thi);
4994 4995 4996 4997 4998 4999 5000 5001

		rcu_read_lock();
		nc = rcu_dereference(tconn->net_conf);
		ping_timeo = nc->ping_timeo;
		no_cork = nc->no_cork;
		ping_int = nc->ping_int;
		rcu_read_unlock();

5002
		if (test_and_clear_bit(SEND_PING, &tconn->flags)) {
5003
			if (drbd_send_ping(tconn)) {
5004
				conn_err(tconn, "drbd_send_ping has failed\n");
5005 5006
				goto reconnect;
			}
5007 5008
			tconn->meta.socket->sk->sk_rcvtimeo = ping_timeo * HZ / 10;
			ping_timeout_active = true;
P
Philipp Reisner 已提交
5009 5010
		}

5011 5012
		/* TODO: conditionally cork; it may hurt latency if we cork without
		   much to send */
5013
		if (!no_cork)
5014
			drbd_tcp_cork(tconn->meta.socket);
5015 5016
		if (tconn_finish_peer_reqs(tconn)) {
			conn_err(tconn, "tconn_finish_peer_reqs() failed\n");
5017
			goto reconnect;
5018
		}
P
Philipp Reisner 已提交
5019
		/* but unconditionally uncork unless disabled */
5020
		if (!no_cork)
5021
			drbd_tcp_uncork(tconn->meta.socket);
P
Philipp Reisner 已提交
5022 5023 5024 5025 5026

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

5027 5028
		rv = drbd_recv_short(tconn->meta.socket, buf, expect-received, 0);
		clear_bit(SIGNAL_ASENDER, &tconn->flags);
P
Philipp Reisner 已提交
5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045

		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) {
5046
			conn_err(tconn, "meta connection shut down by peer.\n");
P
Philipp Reisner 已提交
5047 5048
			goto reconnect;
		} else if (rv == -EAGAIN) {
5049 5050
			/* If the data socket received something meanwhile,
			 * that is good enough: peer is still alive. */
5051 5052
			if (time_after(tconn->last_received,
				jiffies - tconn->meta.socket->sk->sk_rcvtimeo))
5053
				continue;
5054
			if (ping_timeout_active) {
5055
				conn_err(tconn, "PingAck did not arrive in time.\n");
P
Philipp Reisner 已提交
5056 5057
				goto reconnect;
			}
5058
			set_bit(SEND_PING, &tconn->flags);
P
Philipp Reisner 已提交
5059 5060 5061 5062
			continue;
		} else if (rv == -EINTR) {
			continue;
		} else {
5063
			conn_err(tconn, "sock_recvmsg returned %d\n", rv);
P
Philipp Reisner 已提交
5064 5065 5066 5067
			goto reconnect;
		}

		if (received == expect && cmd == NULL) {
5068
			if (decode_header(tconn, tconn->meta.rbuf, &pi))
P
Philipp Reisner 已提交
5069
				goto reconnect;
5070
			cmd = &asender_tbl[pi.cmd];
5071
			if (pi.cmd >= ARRAY_SIZE(asender_tbl) || !cmd->fn) {
5072
				conn_err(tconn, "unknown command %d on meta (l: %d)\n",
5073
					pi.cmd, pi.size);
P
Philipp Reisner 已提交
5074 5075
				goto disconnect;
			}
5076
			expect = header_size + cmd->pkt_size;
5077
			if (pi.size != expect - header_size) {
5078
				conn_err(tconn, "Wrong packet size on meta (c: %d, l: %d)\n",
5079
					pi.cmd, pi.size);
P
Philipp Reisner 已提交
5080
				goto reconnect;
5081
			}
P
Philipp Reisner 已提交
5082 5083
		}
		if (received == expect) {
5084
			bool err;
5085

5086 5087
			err = cmd->fn(tconn, &pi);
			if (err) {
5088
				conn_err(tconn, "%pf failed\n", cmd->fn);
P
Philipp Reisner 已提交
5089
				goto reconnect;
5090
			}
P
Philipp Reisner 已提交
5091

5092 5093
			tconn->last_received = jiffies;

5094 5095 5096 5097 5098
			if (cmd == &asender_tbl[P_PING_ACK]) {
				/* restore idle timeout */
				tconn->meta.socket->sk->sk_rcvtimeo = ping_int * HZ;
				ping_timeout_active = false;
			}
5099

5100
			buf	 = tconn->meta.rbuf;
P
Philipp Reisner 已提交
5101
			received = 0;
5102
			expect	 = header_size;
P
Philipp Reisner 已提交
5103 5104 5105 5106 5107 5108
			cmd	 = NULL;
		}
	}

	if (0) {
reconnect:
5109
		conn_request_state(tconn, NS(conn, C_NETWORK_FAILURE), CS_HARD);
P
Philipp Reisner 已提交
5110 5111 5112
	}
	if (0) {
disconnect:
5113
		conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
5114
	}
5115
	clear_bit(SIGNAL_ASENDER, &tconn->flags);
P
Philipp Reisner 已提交
5116

5117
	conn_info(tconn, "asender terminated\n");
P
Philipp Reisner 已提交
5118 5119 5120

	return 0;
}