drbd_receiver.c 145.7 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);
P
Philipp Reisner 已提交
66
static int drbd_disconnected(struct drbd_conf *mdev);
P
Philipp Reisner 已提交
67

68
static enum finish_epoch drbd_may_finish_epoch(struct drbd_tconn *, 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

298 299 300
	if (page == NULL)
		return;

301
	if (drbd_pp_vacant > (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count)
302 303 304 305 306 307 308 309
		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
310
	}
311
	i = atomic_sub_return(i, a);
312
	if (i < 0)
313 314
		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
315 316 317 318 319 320 321 322
	wake_up(&drbd_pp_wait);
}

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

You must not have the req_lock:
323
 drbd_free_peer_req()
324
 drbd_alloc_peer_req()
325
 drbd_free_peer_reqs()
P
Philipp Reisner 已提交
326
 drbd_ee_fix_bhs()
327
 drbd_finish_peer_reqs()
P
Philipp Reisner 已提交
328 329 330 331
 drbd_clear_done_ee()
 drbd_wait_ee_list_empty()
*/

332
struct drbd_peer_request *
333 334
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
335
{
336
	struct drbd_peer_request *peer_req;
337
	struct page *page = NULL;
338
	unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT;
P
Philipp Reisner 已提交
339

340
	if (drbd_insert_fault(mdev, DRBD_FAULT_AL_EE))
P
Philipp Reisner 已提交
341 342
		return NULL;

343 344
	peer_req = mempool_alloc(drbd_ee_mempool, gfp_mask & ~__GFP_HIGHMEM);
	if (!peer_req) {
P
Philipp Reisner 已提交
345
		if (!(gfp_mask & __GFP_NOWARN))
346
			dev_err(DEV, "%s: allocation failed\n", __func__);
P
Philipp Reisner 已提交
347 348 349
		return NULL;
	}

350 351 352 353 354
	if (data_size) {
		page = drbd_alloc_pages(mdev, nr_pages, (gfp_mask & __GFP_WAIT));
		if (!page)
			goto fail;
	}
P
Philipp Reisner 已提交
355

356 357 358 359 360 361 362
	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;
363
	peer_req->w.mdev = mdev;
364 365 366
	peer_req->pages = page;
	atomic_set(&peer_req->pending_bios, 0);
	peer_req->flags = 0;
367 368 369 370
	/*
	 * The block_id is opaque to the receiver.  It is not endianness
	 * converted, and sent back to the sender unchanged.
	 */
371
	peer_req->block_id = id;
P
Philipp Reisner 已提交
372

373
	return peer_req;
P
Philipp Reisner 已提交
374

375
 fail:
376
	mempool_free(peer_req, drbd_ee_mempool);
P
Philipp Reisner 已提交
377 378 379
	return NULL;
}

380
void __drbd_free_peer_req(struct drbd_conf *mdev, struct drbd_peer_request *peer_req,
381
		       int is_net)
P
Philipp Reisner 已提交
382
{
383 384
	if (peer_req->flags & EE_HAS_DIGEST)
		kfree(peer_req->digest);
385
	drbd_free_pages(mdev, peer_req->pages, is_net);
386 387 388
	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
389 390
}

391
int drbd_free_peer_reqs(struct drbd_conf *mdev, struct list_head *list)
P
Philipp Reisner 已提交
392 393
{
	LIST_HEAD(work_list);
394
	struct drbd_peer_request *peer_req, *t;
P
Philipp Reisner 已提交
395
	int count = 0;
396
	int is_net = list == &mdev->net_ee;
P
Philipp Reisner 已提交
397

398
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
399
	list_splice_init(list, &work_list);
400
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
401

402
	list_for_each_entry_safe(peer_req, t, &work_list, w.list) {
403
		__drbd_free_peer_req(mdev, peer_req, is_net);
P
Philipp Reisner 已提交
404 405 406 407 408
		count++;
	}
	return count;
}

409 410
/*
 * See also comments in _req_mod(,BARRIER_ACKED) and receive_Barrier.
P
Philipp Reisner 已提交
411
 */
412
static int drbd_finish_peer_reqs(struct drbd_conf *mdev)
P
Philipp Reisner 已提交
413 414 415
{
	LIST_HEAD(work_list);
	LIST_HEAD(reclaimed);
416
	struct drbd_peer_request *peer_req, *t;
417
	int err = 0;
P
Philipp Reisner 已提交
418

419
	spin_lock_irq(&mdev->tconn->req_lock);
420
	reclaim_finished_net_peer_reqs(mdev, &reclaimed);
P
Philipp Reisner 已提交
421
	list_splice_init(&mdev->done_ee, &work_list);
422
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
423

424
	list_for_each_entry_safe(peer_req, t, &reclaimed, w.list)
425
		drbd_free_net_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
426 427

	/* possible callbacks here:
428
	 * e_end_block, and e_end_resync_block, e_send_superseded.
P
Philipp Reisner 已提交
429 430
	 * all ignore the last argument.
	 */
431
	list_for_each_entry_safe(peer_req, t, &work_list, w.list) {
432 433
		int err2;

P
Philipp Reisner 已提交
434
		/* list_del not necessary, next/prev members not touched */
435 436 437
		err2 = peer_req->w.cb(&peer_req->w, !!err);
		if (!err)
			err = err2;
438
		drbd_free_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
439 440 441
	}
	wake_up(&mdev->ee_wait);

442
	return err;
P
Philipp Reisner 已提交
443 444
}

445 446
static void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
				     struct list_head *head)
P
Philipp Reisner 已提交
447 448 449 450 451 452 453
{
	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);
454
		spin_unlock_irq(&mdev->tconn->req_lock);
J
Jens Axboe 已提交
455
		io_schedule();
P
Philipp Reisner 已提交
456
		finish_wait(&mdev->ee_wait, &wait);
457
		spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
458 459 460
	}
}

461 462
static void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
				    struct list_head *head)
P
Philipp Reisner 已提交
463
{
464
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
465
	_drbd_wait_ee_list_empty(mdev, head);
466
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
467 468
}

469
static int drbd_recv_short(struct socket *sock, void *buf, size_t size, int flags)
P
Philipp Reisner 已提交
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
{
	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;
}

491
static int drbd_recv(struct drbd_tconn *tconn, void *buf, size_t size)
P
Philipp Reisner 已提交
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
{
	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);
P
Philipp Reisner 已提交
507
	rv = sock_recvmsg(tconn->data.socket, &msg, size, msg.msg_flags);
P
Philipp Reisner 已提交
508 509
	set_fs(oldfs);

P
Philipp Reisner 已提交
510 511 512 513 514 515
	if (rv < 0) {
		if (rv == -ECONNRESET)
			conn_info(tconn, "sock was reset by peer\n");
		else if (rv != -ERESTARTSYS)
			conn_err(tconn, "sock_recvmsg returned %d\n", rv);
	} else if (rv == 0) {
516 517 518 519 520 521 522 523 524 525 526 527 528 529
		if (test_bit(DISCONNECT_SENT, &tconn->flags)) {
			long t;
			rcu_read_lock();
			t = rcu_dereference(tconn->net_conf)->ping_timeo * HZ/10;
			rcu_read_unlock();

			t = wait_event_timeout(tconn->ping_wait, tconn->cstate < C_WF_REPORT_PARAMS, t);

			if (t)
				goto out;
		}
		conn_info(tconn, "sock was shut down by peer\n");
	}

P
Philipp Reisner 已提交
530
	if (rv != size)
531
		conn_request_state(tconn, NS(conn, C_BROKEN_PIPE), CS_HARD);
P
Philipp Reisner 已提交
532

533
out:
P
Philipp Reisner 已提交
534 535 536
	return rv;
}

537 538 539 540 541 542 543 544 545 546 547 548 549
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;
}

550 551 552 553 554 555 556 557 558 559
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;
}

560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578
/* 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;
	}
}

579
static struct socket *drbd_try_connect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
580 581 582 583
{
	const char *what;
	struct socket *sock;
	struct sockaddr_in6 src_in6;
584 585 586
	struct sockaddr_in6 peer_in6;
	struct net_conf *nc;
	int err, peer_addr_len, my_addr_len;
587
	int sndbuf_size, rcvbuf_size, connect_int;
P
Philipp Reisner 已提交
588 589
	int disconnect_on_error = 1;

590 591 592 593
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);
	if (!nc) {
		rcu_read_unlock();
P
Philipp Reisner 已提交
594
		return NULL;
595 596 597
	}
	sndbuf_size = nc->sndbuf_size;
	rcvbuf_size = nc->rcvbuf_size;
598
	connect_int = nc->connect_int;
599
	rcu_read_unlock();
600

601 602
	my_addr_len = min_t(int, tconn->my_addr_len, sizeof(src_in6));
	memcpy(&src_in6, &tconn->my_addr, my_addr_len);
603

604
	if (((struct sockaddr *)&tconn->my_addr)->sa_family == AF_INET6)
605 606 607 608
		src_in6.sin6_port = 0;
	else
		((struct sockaddr_in *)&src_in6)->sin_port = 0; /* AF_INET & AF_SCI */

609 610
	peer_addr_len = min_t(int, tconn->peer_addr_len, sizeof(src_in6));
	memcpy(&peer_in6, &tconn->peer_addr, peer_addr_len);
P
Philipp Reisner 已提交
611 612

	what = "sock_create_kern";
613 614
	err = sock_create_kern(((struct sockaddr *)&src_in6)->sa_family,
			       SOCK_STREAM, IPPROTO_TCP, &sock);
P
Philipp Reisner 已提交
615 616 617 618 619 620
	if (err < 0) {
		sock = NULL;
		goto out;
	}

	sock->sk->sk_rcvtimeo =
621
	sock->sk->sk_sndtimeo = connect_int * HZ;
622
	drbd_setbufsize(sock, sndbuf_size, rcvbuf_size);
P
Philipp Reisner 已提交
623 624 625 626 627 628 629 630 631

       /* 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";
632
	err = sock->ops->bind(sock, (struct sockaddr *) &src_in6, my_addr_len);
P
Philipp Reisner 已提交
633 634 635 636 637 638 639
	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";
640
	err = sock->ops->connect(sock, (struct sockaddr *) &peer_in6, peer_addr_len, 0);
P
Philipp Reisner 已提交
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657

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:
658
			conn_err(tconn, "%s failed, err = %d\n", what, err);
P
Philipp Reisner 已提交
659 660
		}
		if (disconnect_on_error)
661
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
662
	}
663

P
Philipp Reisner 已提交
664 665 666
	return sock;
}

667 668 669 670 671 672 673 674
struct accept_wait_data {
	struct drbd_tconn *tconn;
	struct socket *s_listen;
	struct completion door_bell;
	void (*original_sk_state_change)(struct sock *sk);

};

675
static void drbd_incoming_connection(struct sock *sk)
676 677
{
	struct accept_wait_data *ad = sk->sk_user_data;
678
	void (*state_change)(struct sock *sk);
679

680 681 682 683
	state_change = ad->original_sk_state_change;
	if (sk->sk_state == TCP_ESTABLISHED)
		complete(&ad->door_bell);
	state_change(sk);
684 685 686
}

static int prepare_listen_socket(struct drbd_tconn *tconn, struct accept_wait_data *ad)
P
Philipp Reisner 已提交
687
{
688
	int err, sndbuf_size, rcvbuf_size, my_addr_len;
689
	struct sockaddr_in6 my_addr;
690
	struct socket *s_listen;
691
	struct net_conf *nc;
P
Philipp Reisner 已提交
692 693
	const char *what;

694 695 696 697
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);
	if (!nc) {
		rcu_read_unlock();
698
		return -EIO;
699 700 701 702
	}
	sndbuf_size = nc->sndbuf_size;
	rcvbuf_size = nc->rcvbuf_size;
	rcu_read_unlock();
P
Philipp Reisner 已提交
703

704 705 706
	my_addr_len = min_t(int, tconn->my_addr_len, sizeof(struct sockaddr_in6));
	memcpy(&my_addr, &tconn->my_addr, my_addr_len);

P
Philipp Reisner 已提交
707
	what = "sock_create_kern";
708
	err = sock_create_kern(((struct sockaddr *)&my_addr)->sa_family,
709
			       SOCK_STREAM, IPPROTO_TCP, &s_listen);
P
Philipp Reisner 已提交
710 711 712 713 714
	if (err) {
		s_listen = NULL;
		goto out;
	}

715
	s_listen->sk->sk_reuse = 1; /* SO_REUSEADDR */
716
	drbd_setbufsize(s_listen, sndbuf_size, rcvbuf_size);
P
Philipp Reisner 已提交
717 718

	what = "bind before listen";
719
	err = s_listen->ops->bind(s_listen, (struct sockaddr *)&my_addr, my_addr_len);
P
Philipp Reisner 已提交
720 721 722
	if (err < 0)
		goto out;

723 724 725
	ad->s_listen = s_listen;
	write_lock_bh(&s_listen->sk->sk_callback_lock);
	ad->original_sk_state_change = s_listen->sk->sk_state_change;
726
	s_listen->sk->sk_state_change = drbd_incoming_connection;
727 728 729
	s_listen->sk->sk_user_data = ad;
	write_unlock_bh(&s_listen->sk->sk_callback_lock);

730 731 732 733 734
	what = "listen";
	err = s_listen->ops->listen(s_listen, 5);
	if (err < 0)
		goto out;

735
	return 0;
736 737 738 739 740 741 742 743 744 745
out:
	if (s_listen)
		sock_release(s_listen);
	if (err < 0) {
		if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) {
			conn_err(tconn, "%s failed, err = %d\n", what, err);
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
		}
	}

746
	return -EIO;
747 748
}

749 750 751 752 753 754 755 756
static void unregister_state_change(struct sock *sk, struct accept_wait_data *ad)
{
	write_lock_bh(&sk->sk_callback_lock);
	sk->sk_state_change = ad->original_sk_state_change;
	sk->sk_user_data = NULL;
	write_unlock_bh(&sk->sk_callback_lock);
}

757
static struct socket *drbd_wait_for_connect(struct drbd_tconn *tconn, struct accept_wait_data *ad)
758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
{
	int timeo, connect_int, err = 0;
	struct socket *s_estab = NULL;
	struct net_conf *nc;

	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);
	if (!nc) {
		rcu_read_unlock();
		return NULL;
	}
	connect_int = nc->connect_int;
	rcu_read_unlock();

	timeo = connect_int * HZ;
	timeo += (random32() & 1) ? timeo / 7 : -timeo / 7; /* 28.5% random jitter */

775 776 777
	err = wait_for_completion_interruptible_timeout(&ad->door_bell, timeo);
	if (err <= 0)
		return NULL;
P
Philipp Reisner 已提交
778

779
	err = kernel_accept(ad->s_listen, &s_estab, 0);
P
Philipp Reisner 已提交
780 781
	if (err < 0) {
		if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) {
782
			conn_err(tconn, "accept failed, err = %d\n", err);
783
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
784 785 786
		}
	}

787 788 789
	if (s_estab)
		unregister_state_change(s_estab->sk, ad);

P
Philipp Reisner 已提交
790 791 792
	return s_estab;
}

793
static int decode_header(struct drbd_tconn *, void *, struct packet_info *);
P
Philipp Reisner 已提交
794

795 796 797 798 799
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;
800
	return conn_send_command(tconn, sock, cmd, 0, NULL, 0);
P
Philipp Reisner 已提交
801 802
}

803
static int receive_first_packet(struct drbd_tconn *tconn, struct socket *sock)
P
Philipp Reisner 已提交
804
{
805 806 807
	unsigned int header_size = drbd_header_size(tconn);
	struct packet_info pi;
	int err;
P
Philipp Reisner 已提交
808

809 810 811 812 813 814 815 816 817 818
	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
819 820 821 822 823 824
}

/**
 * drbd_socket_okay() - Free the socket if its connection is not okay
 * @sock:	pointer to the pointer to the socket.
 */
825
static int drbd_socket_okay(struct socket **sock)
P
Philipp Reisner 已提交
826 827 828 829 830
{
	int rr;
	char tb[4];

	if (!*sock)
831
		return false;
P
Philipp Reisner 已提交
832

833
	rr = drbd_recv_short(*sock, tb, 4, MSG_DONTWAIT | MSG_PEEK);
P
Philipp Reisner 已提交
834 835

	if (rr > 0 || rr == -EAGAIN) {
836
		return true;
P
Philipp Reisner 已提交
837 838 839
	} else {
		sock_release(*sock);
		*sock = NULL;
840
		return false;
P
Philipp Reisner 已提交
841 842
	}
}
843 844
/* Gets called if a connection is established, or if a new minor gets created
   in a connection */
P
Philipp Reisner 已提交
845
int drbd_connected(struct drbd_conf *mdev)
846
{
847
	int err;
848 849 850 851

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

852 853 854 855
	mdev->state_mutex = mdev->tconn->agreed_pro_version < 100 ?
		&mdev->tconn->cstate_mutex :
		&mdev->own_state_mutex;

856 857 858 859 860 861
	err = drbd_send_sync_param(mdev);
	if (!err)
		err = drbd_send_sizes(mdev, 0, 0);
	if (!err)
		err = drbd_send_uuids(mdev);
	if (!err)
862
		err = drbd_send_current_state(mdev);
863 864
	clear_bit(USE_DEGR_WFC_T, &mdev->flags);
	clear_bit(RESIZE_PENDING, &mdev->flags);
865
	mod_timer(&mdev->request_timer, jiffies + HZ); /* just start it here. */
866
	return err;
867 868
}

P
Philipp Reisner 已提交
869 870 871 872 873 874 875 876
/*
 * 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...
 */
877
static int conn_connect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
878
{
879
	struct drbd_socket sock, msock;
P
Philipp Reisner 已提交
880
	struct drbd_conf *mdev;
881
	struct net_conf *nc;
882
	int vnr, timeout, h, ok;
883
	bool discard_my_data;
884
	enum drbd_state_rv rv;
885 886 887 888
	struct accept_wait_data ad = {
		.tconn = tconn,
		.door_bell = COMPLETION_INITIALIZER_ONSTACK(ad.door_bell),
	};
P
Philipp Reisner 已提交
889

890
	clear_bit(DISCONNECT_SENT, &tconn->flags);
891
	if (conn_request_state(tconn, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS)
P
Philipp Reisner 已提交
892 893
		return -2;

894 895 896 897 898 899 900 901 902
	mutex_init(&sock.mutex);
	sock.sbuf = tconn->data.sbuf;
	sock.rbuf = tconn->data.rbuf;
	sock.socket = NULL;
	mutex_init(&msock.mutex);
	msock.sbuf = tconn->meta.sbuf;
	msock.rbuf = tconn->meta.rbuf;
	msock.socket = NULL;

903 904
	/* Assume that the peer only understands protocol 80 until we know better.  */
	tconn->agreed_pro_version = 80;
P
Philipp Reisner 已提交
905

906 907 908
	if (prepare_listen_socket(tconn, &ad))
		return 0;

P
Philipp Reisner 已提交
909
	do {
910 911
		struct socket *s;

912
		s = drbd_try_connect(tconn);
P
Philipp Reisner 已提交
913
		if (s) {
914 915 916 917
			if (!sock.socket) {
				sock.socket = s;
				send_first_packet(tconn, &sock, P_INITIAL_DATA);
			} else if (!msock.socket) {
918
				clear_bit(RESOLVE_CONFLICTS, &tconn->flags);
919 920
				msock.socket = s;
				send_first_packet(tconn, &msock, P_INITIAL_META);
P
Philipp Reisner 已提交
921
			} else {
922
				conn_err(tconn, "Logic error in conn_connect()\n");
P
Philipp Reisner 已提交
923 924 925 926
				goto out_release_sockets;
			}
		}

927 928 929 930 931 932 933 934
		if (sock.socket && msock.socket) {
			rcu_read_lock();
			nc = rcu_dereference(tconn->net_conf);
			timeout = nc->ping_timeo * HZ / 10;
			rcu_read_unlock();
			schedule_timeout_interruptible(timeout);
			ok = drbd_socket_okay(&sock.socket);
			ok = drbd_socket_okay(&msock.socket) && ok;
P
Philipp Reisner 已提交
935 936 937 938 939
			if (ok)
				break;
		}

retry:
940
		s = drbd_wait_for_connect(tconn, &ad);
P
Philipp Reisner 已提交
941
		if (s) {
942
			int fp = receive_first_packet(tconn, s);
943 944
			drbd_socket_okay(&sock.socket);
			drbd_socket_okay(&msock.socket);
945
			switch (fp) {
946
			case P_INITIAL_DATA:
947
				if (sock.socket) {
948
					conn_warn(tconn, "initial packet S crossed\n");
949
					sock_release(sock.socket);
950 951
					sock.socket = s;
					goto randomize;
P
Philipp Reisner 已提交
952
				}
953
				sock.socket = s;
P
Philipp Reisner 已提交
954
				break;
955
			case P_INITIAL_META:
956
				set_bit(RESOLVE_CONFLICTS, &tconn->flags);
957
				if (msock.socket) {
958
					conn_warn(tconn, "initial packet M crossed\n");
959
					sock_release(msock.socket);
960 961
					msock.socket = s;
					goto randomize;
P
Philipp Reisner 已提交
962
				}
963
				msock.socket = s;
P
Philipp Reisner 已提交
964 965
				break;
			default:
966
				conn_warn(tconn, "Error receiving initial packet\n");
P
Philipp Reisner 已提交
967
				sock_release(s);
968
randomize:
P
Philipp Reisner 已提交
969 970 971 972 973
				if (random32() & 1)
					goto retry;
			}
		}

974
		if (tconn->cstate <= C_DISCONNECTING)
P
Philipp Reisner 已提交
975 976 977 978
			goto out_release_sockets;
		if (signal_pending(current)) {
			flush_signals(current);
			smp_rmb();
979
			if (get_t_state(&tconn->receiver) == EXITING)
P
Philipp Reisner 已提交
980 981 982
				goto out_release_sockets;
		}

983 984 985
		ok = drbd_socket_okay(&sock.socket);
		ok = drbd_socket_okay(&msock.socket) && ok;
	} while (!ok);
P
Philipp Reisner 已提交
986

987 988 989
	if (ad.s_listen)
		sock_release(ad.s_listen);

990 991
	sock.socket->sk->sk_reuse = 1; /* SO_REUSEADDR */
	msock.socket->sk->sk_reuse = 1; /* SO_REUSEADDR */
P
Philipp Reisner 已提交
992

993 994
	sock.socket->sk->sk_allocation = GFP_NOIO;
	msock.socket->sk->sk_allocation = GFP_NOIO;
P
Philipp Reisner 已提交
995

996 997
	sock.socket->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK;
	msock.socket->sk->sk_priority = TC_PRIO_INTERACTIVE;
P
Philipp Reisner 已提交
998 999

	/* NOT YET ...
1000 1001
	 * sock.socket->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10;
	 * sock.socket->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
1002
	 * first set it to the P_CONNECTION_FEATURES timeout,
P
Philipp Reisner 已提交
1003
	 * which we set to 4x the configured ping_timeout. */
1004 1005 1006
	rcu_read_lock();
	nc = rcu_dereference(tconn->net_conf);

1007 1008
	sock.socket->sk->sk_sndtimeo =
	sock.socket->sk->sk_rcvtimeo = nc->ping_timeo*4*HZ/10;
1009

1010
	msock.socket->sk->sk_rcvtimeo = nc->ping_int*HZ;
1011
	timeout = nc->timeout * HZ / 10;
1012
	discard_my_data = nc->discard_my_data;
1013
	rcu_read_unlock();
P
Philipp Reisner 已提交
1014

1015
	msock.socket->sk->sk_sndtimeo = timeout;
P
Philipp Reisner 已提交
1016 1017

	/* we don't want delays.
L
Lucas De Marchi 已提交
1018
	 * we use TCP_CORK where appropriate, though */
1019 1020
	drbd_tcp_nodelay(sock.socket);
	drbd_tcp_nodelay(msock.socket);
P
Philipp Reisner 已提交
1021

1022 1023
	tconn->data.socket = sock.socket;
	tconn->meta.socket = msock.socket;
1024
	tconn->last_received = jiffies;
P
Philipp Reisner 已提交
1025

1026
	h = drbd_do_features(tconn);
P
Philipp Reisner 已提交
1027 1028 1029
	if (h <= 0)
		return h;

1030
	if (tconn->cram_hmac_tfm) {
P
Philipp Reisner 已提交
1031
		/* drbd_request_state(mdev, NS(conn, WFAuth)); */
1032
		switch (drbd_do_auth(tconn)) {
1033
		case -1:
1034
			conn_err(tconn, "Authentication of peer failed\n");
P
Philipp Reisner 已提交
1035
			return -1;
1036
		case 0:
1037
			conn_err(tconn, "Authentication of peer failed, trying again.\n");
1038
			return 0;
P
Philipp Reisner 已提交
1039 1040 1041
		}
	}

1042 1043
	tconn->data.socket->sk->sk_sndtimeo = timeout;
	tconn->data.socket->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
P
Philipp Reisner 已提交
1044

1045
	if (drbd_send_protocol(tconn) == -EOPNOTSUPP)
1046
		return -1;
P
Philipp Reisner 已提交
1047

1048 1049
	set_bit(STATE_SENT, &tconn->flags);

P
Philipp Reisner 已提交
1050 1051 1052 1053
	rcu_read_lock();
	idr_for_each_entry(&tconn->volumes, mdev, vnr) {
		kref_get(&mdev->kref);
		rcu_read_unlock();
1054 1055 1056 1057 1058 1059

		if (discard_my_data)
			set_bit(DISCARD_MY_DATA, &mdev->flags);
		else
			clear_bit(DISCARD_MY_DATA, &mdev->flags);

P
Philipp Reisner 已提交
1060 1061 1062 1063 1064 1065
		drbd_connected(mdev);
		kref_put(&mdev->kref, &drbd_minor_destroy);
		rcu_read_lock();
	}
	rcu_read_unlock();

1066 1067 1068
	rv = conn_request_state(tconn, NS(conn, C_WF_REPORT_PARAMS), CS_VERBOSE);
	if (rv < SS_SUCCESS) {
		clear_bit(STATE_SENT, &tconn->flags);
1069
		return 0;
1070
	}
1071 1072 1073

	drbd_thread_start(&tconn->asender);

1074 1075 1076 1077 1078 1079 1080 1081
	mutex_lock(&tconn->conf_update);
	/* The discard_my_data flag is a single-shot modifier to the next
	 * connection attempt, the handshake of which is now well underway.
	 * No need for rcu style copying of the whole struct
	 * just to clear a single value. */
	tconn->net_conf->discard_my_data = 0;
	mutex_unlock(&tconn->conf_update);

1082
	return h;
P
Philipp Reisner 已提交
1083 1084

out_release_sockets:
1085 1086
	if (ad.s_listen)
		sock_release(ad.s_listen);
1087 1088 1089 1090
	if (sock.socket)
		sock_release(sock.socket);
	if (msock.socket)
		sock_release(msock.socket);
P
Philipp Reisner 已提交
1091 1092 1093
	return -1;
}

1094
static int decode_header(struct drbd_tconn *tconn, void *header, struct packet_info *pi)
P
Philipp Reisner 已提交
1095
{
1096 1097
	unsigned int header_size = drbd_header_size(tconn);

1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
	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)) {
1110 1111
		struct p_header95 *h = header;
		pi->cmd = be16_to_cpu(h->command);
1112 1113
		pi->size = be32_to_cpu(h->length);
		pi->vnr = 0;
1114 1115 1116 1117 1118
	} 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);
1119
		pi->vnr = 0;
1120
	} else {
1121 1122 1123
		conn_err(tconn, "Wrong magic value 0x%08x in protocol version %d\n",
			 be32_to_cpu(*(__be32 *)header),
			 tconn->agreed_pro_version);
1124
		return -EINVAL;
P
Philipp Reisner 已提交
1125
	}
1126
	pi->data = header + header_size;
1127
	return 0;
1128 1129
}

1130
static int drbd_recv_header(struct drbd_tconn *tconn, struct packet_info *pi)
1131
{
1132
	void *buffer = tconn->data.rbuf;
1133
	int err;
1134

1135
	err = drbd_recv_all_warn(tconn, buffer, drbd_header_size(tconn));
1136
	if (err)
1137
		return err;
1138

1139
	err = decode_header(tconn, buffer, pi);
1140
	tconn->last_received = jiffies;
P
Philipp Reisner 已提交
1141

1142
	return err;
P
Philipp Reisner 已提交
1143 1144
}

1145
static void drbd_flush(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
1146 1147
{
	int rv;
1148 1149 1150 1151
	struct drbd_conf *mdev;
	int vnr;

	if (tconn->write_ordering >= WO_bdev_flush) {
1152
		rcu_read_lock();
1153
		idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
			if (!get_ldev(mdev))
				continue;
			kref_get(&mdev->kref);
			rcu_read_unlock();

			rv = blkdev_issue_flush(mdev->ldev->backing_bdev,
					GFP_NOIO, NULL);
			if (rv) {
				dev_info(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(tconn, WO_drain_io);
1167
			}
1168 1169 1170 1171 1172 1173
			put_ldev(mdev);
			kref_put(&mdev->kref, &drbd_minor_destroy);

			rcu_read_lock();
			if (rv)
				break;
P
Philipp Reisner 已提交
1174
		}
1175
		rcu_read_unlock();
P
Philipp Reisner 已提交
1176 1177 1178 1179 1180 1181 1182 1183 1184
	}
}

/**
 * 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.
 */
1185
static enum finish_epoch drbd_may_finish_epoch(struct drbd_tconn *tconn,
P
Philipp Reisner 已提交
1186 1187 1188
					       struct drbd_epoch *epoch,
					       enum epoch_event ev)
{
1189
	int epoch_size;
P
Philipp Reisner 已提交
1190 1191 1192
	struct drbd_epoch *next_epoch;
	enum finish_epoch rv = FE_STILL_LIVE;

1193
	spin_lock(&tconn->epoch_lock);
P
Philipp Reisner 已提交
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
	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 &&
1213
		    (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags) || ev & EV_CLEANUP)) {
P
Philipp Reisner 已提交
1214
			if (!(ev & EV_CLEANUP)) {
1215
				spin_unlock(&tconn->epoch_lock);
1216
				drbd_send_b_ack(epoch->tconn, epoch->barrier_nr, epoch_size);
1217
				spin_lock(&tconn->epoch_lock);
P
Philipp Reisner 已提交
1218
			}
1219 1220 1221
#if 0
			/* FIXME: dec unacked on connection, once we have
			 * something to count pending connection packets in. */
1222
			if (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags))
1223 1224
				dec_unacked(epoch->tconn);
#endif
P
Philipp Reisner 已提交
1225

1226
			if (tconn->current_epoch != epoch) {
P
Philipp Reisner 已提交
1227 1228 1229
				next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list);
				list_del(&epoch->list);
				ev = EV_BECAME_LAST | (ev & EV_CLEANUP);
1230
				tconn->epochs--;
P
Philipp Reisner 已提交
1231 1232 1233 1234 1235 1236 1237
				kfree(epoch);

				if (rv == FE_STILL_LIVE)
					rv = FE_DESTROYED;
			} else {
				epoch->flags = 0;
				atomic_set(&epoch->epoch_size, 0);
1238
				/* atomic_set(&epoch->active, 0); is already zero */
P
Philipp Reisner 已提交
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
				if (rv == FE_STILL_LIVE)
					rv = FE_RECYCLED;
			}
		}

		if (!next_epoch)
			break;

		epoch = next_epoch;
	} while (1);

1250
	spin_unlock(&tconn->epoch_lock);
P
Philipp Reisner 已提交
1251 1252 1253 1254 1255 1256

	return rv;
}

/**
 * drbd_bump_write_ordering() - Fall back to an other write ordering method
1257
 * @tconn:	DRBD connection.
P
Philipp Reisner 已提交
1258 1259
 * @wo:		Write ordering method to try.
 */
1260
void drbd_bump_write_ordering(struct drbd_tconn *tconn, enum write_ordering_e wo)
P
Philipp Reisner 已提交
1261
{
P
Philipp Reisner 已提交
1262
	struct disk_conf *dc;
1263
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
1264
	enum write_ordering_e pwo;
1265
	int vnr;
P
Philipp Reisner 已提交
1266 1267 1268 1269 1270 1271
	static char *write_ordering_str[] = {
		[WO_none] = "none",
		[WO_drain_io] = "drain",
		[WO_bdev_flush] = "flush",
	};

1272
	pwo = tconn->write_ordering;
P
Philipp Reisner 已提交
1273
	wo = min(pwo, wo);
P
Philipp Reisner 已提交
1274
	rcu_read_lock();
1275
	idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1276
		if (!get_ldev_if_state(mdev, D_ATTACHING))
1277 1278 1279 1280 1281 1282 1283 1284 1285
			continue;
		dc = rcu_dereference(mdev->ldev->disk_conf);

		if (wo == WO_bdev_flush && !dc->disk_flushes)
			wo = WO_drain_io;
		if (wo == WO_drain_io && !dc->disk_drain)
			wo = WO_none;
		put_ldev(mdev);
	}
P
Philipp Reisner 已提交
1286
	rcu_read_unlock();
1287 1288 1289
	tconn->write_ordering = wo;
	if (pwo != tconn->write_ordering || wo == WO_bdev_flush)
		conn_info(tconn, "Method to ensure write ordering: %s\n", write_ordering_str[tconn->write_ordering]);
P
Philipp Reisner 已提交
1290 1291
}

1292
/**
1293
 * drbd_submit_peer_request()
1294
 * @mdev:	DRBD device.
1295
 * @peer_req:	peer request
1296
 * @rw:		flag field, see bio->bi_rw
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306
 *
 * 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.
1307 1308
 */
/* TODO allocate from our own bio_set. */
1309 1310 1311
int drbd_submit_peer_request(struct drbd_conf *mdev,
			     struct drbd_peer_request *peer_req,
			     const unsigned rw, const int fault_type)
1312 1313 1314
{
	struct bio *bios = NULL;
	struct bio *bio;
1315 1316 1317
	struct page *page = peer_req->pages;
	sector_t sector = peer_req->i.sector;
	unsigned ds = peer_req->i.size;
1318 1319
	unsigned n_bios = 0;
	unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT;
1320
	int err = -ENOMEM;
1321 1322 1323 1324

	/* 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
1325 1326 1327 1328 1329
	 * 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.
	 */
1330 1331 1332 1333 1334 1335
next_bio:
	bio = bio_alloc(GFP_NOIO, nr_pages);
	if (!bio) {
		dev_err(DEV, "submit_ee: Allocation of a bio failed\n");
		goto fail;
	}
1336
	/* > peer_req->i.sector, unless this is the first bio */
1337 1338 1339
	bio->bi_sector = sector;
	bio->bi_bdev = mdev->ldev->backing_bdev;
	bio->bi_rw = rw;
1340
	bio->bi_private = peer_req;
1341
	bio->bi_end_io = drbd_peer_request_endio;
1342 1343 1344 1345 1346 1347 1348 1349

	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)) {
1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360
			/* 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;
			}
1361 1362 1363 1364 1365 1366 1367 1368 1369
			goto next_bio;
		}
		ds -= len;
		sector += len >> 9;
		--nr_pages;
	}
	D_ASSERT(page == NULL);
	D_ASSERT(ds == 0);

1370
	atomic_set(&peer_req->pending_bios, n_bios);
1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
	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);
	}
1386
	return err;
1387 1388
}

1389
static void drbd_remove_epoch_entry_interval(struct drbd_conf *mdev,
1390
					     struct drbd_peer_request *peer_req)
1391
{
1392
	struct drbd_interval *i = &peer_req->i;
1393 1394 1395 1396

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

A
Andreas Gruenbacher 已提交
1397
	/* Wake up any processes waiting for this peer request to complete.  */
1398 1399 1400 1401
	if (i->waiting)
		wake_up(&mdev->misc_wait);
}

1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
void conn_wait_active_ee_empty(struct drbd_tconn *tconn)
{
	struct drbd_conf *mdev;
	int vnr;

	rcu_read_lock();
	idr_for_each_entry(&tconn->volumes, mdev, vnr) {
		kref_get(&mdev->kref);
		rcu_read_unlock();
		drbd_wait_ee_list_empty(mdev, &mdev->active_ee);
		kref_put(&mdev->kref, &drbd_minor_destroy);
		rcu_read_lock();
	}
	rcu_read_unlock();
}

1418
static int receive_Barrier(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
1419
{
1420
	int rv;
1421
	struct p_barrier *p = pi->data;
P
Philipp Reisner 已提交
1422 1423
	struct drbd_epoch *epoch;

1424 1425 1426
	/* FIXME these are unacked on connection,
	 * not a specific (peer)device.
	 */
1427
	tconn->current_epoch->barrier_nr = p->barrier;
1428
	tconn->current_epoch->tconn = tconn;
1429
	rv = drbd_may_finish_epoch(tconn, tconn->current_epoch, EV_GOT_BARRIER_NR);
P
Philipp Reisner 已提交
1430 1431 1432 1433 1434 1435

	/* 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. */
1436
	switch (tconn->write_ordering) {
P
Philipp Reisner 已提交
1437 1438
	case WO_none:
		if (rv == FE_RECYCLED)
1439
			return 0;
1440 1441 1442 1443 1444 1445 1446

		/* 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
1447
			conn_warn(tconn, "Allocation of an epoch failed, slowing down\n");
1448
			/* Fall through */
P
Philipp Reisner 已提交
1449 1450 1451

	case WO_bdev_flush:
	case WO_drain_io:
1452
		conn_wait_active_ee_empty(tconn);
1453
		drbd_flush(tconn);
1454

1455
		if (atomic_read(&tconn->current_epoch->epoch_size)) {
1456 1457 1458
			epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO);
			if (epoch)
				break;
P
Philipp Reisner 已提交
1459 1460
		}

1461
		return 0;
1462
	default:
1463
		conn_err(tconn, "Strangeness in tconn->write_ordering %d\n", tconn->write_ordering);
1464
		return -EIO;
P
Philipp Reisner 已提交
1465 1466 1467 1468 1469 1470
	}

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

1471 1472 1473 1474 1475
	spin_lock(&tconn->epoch_lock);
	if (atomic_read(&tconn->current_epoch->epoch_size)) {
		list_add(&epoch->list, &tconn->current_epoch->list);
		tconn->current_epoch = epoch;
		tconn->epochs++;
P
Philipp Reisner 已提交
1476 1477 1478 1479
	} else {
		/* The current_epoch got recycled while we allocated this one... */
		kfree(epoch);
	}
1480
	spin_unlock(&tconn->epoch_lock);
P
Philipp Reisner 已提交
1481

1482
	return 0;
P
Philipp Reisner 已提交
1483 1484 1485 1486
}

/* used from receive_RSDataReply (recv_resync_read)
 * and from receive_Data */
1487 1488 1489
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
1490
{
1491
	const sector_t capacity = drbd_get_capacity(mdev->this_bdev);
1492
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
1493
	struct page *page;
1494
	int dgs, ds, err;
1495 1496
	void *dig_in = mdev->tconn->int_dig_in;
	void *dig_vv = mdev->tconn->int_dig_vv;
1497
	unsigned long *data;
P
Philipp Reisner 已提交
1498

1499 1500 1501
	dgs = 0;
	if (mdev->tconn->peer_integrity_tfm) {
		dgs = crypto_hash_digestsize(mdev->tconn->peer_integrity_tfm);
1502 1503 1504 1505
		/*
		 * FIXME: Receive the incoming digest into the receive buffer
		 *	  here, together with its struct p_data?
		 */
1506 1507
		err = drbd_recv_all_warn(mdev->tconn, dig_in, dgs);
		if (err)
P
Philipp Reisner 已提交
1508
			return NULL;
1509
		data_size -= dgs;
P
Philipp Reisner 已提交
1510 1511
	}

1512 1513 1514 1515
	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
1516

1517 1518 1519
	/* even though we trust out peer,
	 * we sometimes have to double check. */
	if (sector + (data_size>>9) > capacity) {
1520 1521
		dev_err(DEV, "request from peer beyond end of local disk: "
			"capacity: %llus < sector: %llus + size: %u\n",
1522 1523 1524 1525 1526
			(unsigned long long)capacity,
			(unsigned long long)sector, data_size);
		return NULL;
	}

P
Philipp Reisner 已提交
1527 1528 1529
	/* 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.  */
1530
	peer_req = drbd_alloc_peer_req(mdev, id, sector, data_size, GFP_NOIO);
1531
	if (!peer_req)
P
Philipp Reisner 已提交
1532
		return NULL;
1533

1534 1535 1536
	if (!data_size)
		return peer_req;

P
Philipp Reisner 已提交
1537
	ds = data_size;
1538
	page = peer_req->pages;
1539 1540
	page_chain_for_each(page) {
		unsigned len = min_t(int, ds, PAGE_SIZE);
1541
		data = kmap(page);
1542
		err = drbd_recv_all_warn(mdev->tconn, data, len);
1543
		if (drbd_insert_fault(mdev, DRBD_FAULT_RECEIVE)) {
1544 1545 1546
			dev_err(DEV, "Fault injection: Corrupting data on receive\n");
			data[0] = data[0] ^ (unsigned long)-1;
		}
P
Philipp Reisner 已提交
1547
		kunmap(page);
1548
		if (err) {
1549
			drbd_free_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
1550 1551
			return NULL;
		}
1552
		ds -= len;
P
Philipp Reisner 已提交
1553 1554 1555
	}

	if (dgs) {
1556
		drbd_csum_ee(mdev, mdev->tconn->peer_integrity_tfm, peer_req, dig_vv);
P
Philipp Reisner 已提交
1557
		if (memcmp(dig_in, dig_vv, dgs)) {
1558 1559
			dev_err(DEV, "Digest integrity check FAILED: %llus +%u\n",
				(unsigned long long)sector, data_size);
1560
			drbd_free_peer_req(mdev, peer_req);
P
Philipp Reisner 已提交
1561 1562 1563 1564
			return NULL;
		}
	}
	mdev->recv_cnt += data_size>>9;
1565
	return peer_req;
P
Philipp Reisner 已提交
1566 1567 1568 1569 1570 1571 1572 1573
}

/* 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;
1574
	int err = 0;
P
Philipp Reisner 已提交
1575 1576
	void *data;

1577
	if (!data_size)
1578
		return 0;
1579

1580
	page = drbd_alloc_pages(mdev, 1, 1);
P
Philipp Reisner 已提交
1581 1582 1583

	data = kmap(page);
	while (data_size) {
1584 1585
		unsigned int len = min_t(int, data_size, PAGE_SIZE);

1586 1587
		err = drbd_recv_all_warn(mdev->tconn, data, len);
		if (err)
P
Philipp Reisner 已提交
1588
			break;
1589
		data_size -= len;
P
Philipp Reisner 已提交
1590 1591
	}
	kunmap(page);
1592
	drbd_free_pages(mdev, page, 0);
1593
	return err;
P
Philipp Reisner 已提交
1594 1595 1596 1597 1598 1599 1600
}

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;
1601
	int dgs, err, i, expect;
1602 1603
	void *dig_in = mdev->tconn->int_dig_in;
	void *dig_vv = mdev->tconn->int_dig_vv;
P
Philipp Reisner 已提交
1604

1605 1606 1607
	dgs = 0;
	if (mdev->tconn->peer_integrity_tfm) {
		dgs = crypto_hash_digestsize(mdev->tconn->peer_integrity_tfm);
1608 1609 1610
		err = drbd_recv_all_warn(mdev->tconn, dig_in, dgs);
		if (err)
			return err;
1611
		data_size -= dgs;
P
Philipp Reisner 已提交
1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
	}

	/* 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) {
1622
		void *mapped = kmap(bvec->bv_page) + bvec->bv_offset;
P
Philipp Reisner 已提交
1623
		expect = min_t(int, data_size, bvec->bv_len);
1624
		err = drbd_recv_all_warn(mdev->tconn, mapped, expect);
P
Philipp Reisner 已提交
1625
		kunmap(bvec->bv_page);
1626 1627 1628
		if (err)
			return err;
		data_size -= expect;
P
Philipp Reisner 已提交
1629 1630 1631
	}

	if (dgs) {
1632
		drbd_csum_bio(mdev, mdev->tconn->peer_integrity_tfm, bio, dig_vv);
P
Philipp Reisner 已提交
1633 1634
		if (memcmp(dig_in, dig_vv, dgs)) {
			dev_err(DEV, "Digest integrity check FAILED. Broken NICs?\n");
1635
			return -EINVAL;
P
Philipp Reisner 已提交
1636 1637 1638 1639
		}
	}

	D_ASSERT(data_size == 0);
1640
	return 0;
P
Philipp Reisner 已提交
1641 1642
}

1643 1644 1645 1646
/*
 * e_end_resync_block() is called in asender context via
 * drbd_finish_peer_reqs().
 */
1647
static int e_end_resync_block(struct drbd_work *w, int unused)
P
Philipp Reisner 已提交
1648
{
1649 1650
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
1651
	struct drbd_conf *mdev = w->mdev;
1652
	sector_t sector = peer_req->i.sector;
1653
	int err;
P
Philipp Reisner 已提交
1654

1655
	D_ASSERT(drbd_interval_empty(&peer_req->i));
P
Philipp Reisner 已提交
1656

1657 1658
	if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
		drbd_set_in_sync(mdev, sector, peer_req->i.size);
1659
		err = drbd_send_ack(mdev, P_RS_WRITE_ACK, peer_req);
P
Philipp Reisner 已提交
1660 1661
	} else {
		/* Record failure to sync */
1662
		drbd_rs_failed_io(mdev, sector, peer_req->i.size);
P
Philipp Reisner 已提交
1663

1664
		err  = drbd_send_ack(mdev, P_NEG_ACK, peer_req);
P
Philipp Reisner 已提交
1665 1666 1667
	}
	dec_unacked(mdev);

1668
	return err;
P
Philipp Reisner 已提交
1669 1670 1671 1672
}

static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_size) __releases(local)
{
1673
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
1674

1675 1676
	peer_req = read_in_block(mdev, ID_SYNCER, sector, data_size);
	if (!peer_req)
1677
		goto fail;
P
Philipp Reisner 已提交
1678 1679 1680 1681 1682 1683 1684

	dec_rs_pending(mdev);

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

1685
	peer_req->w.cb = e_end_resync_block;
1686

1687
	spin_lock_irq(&mdev->tconn->req_lock);
1688
	list_add(&peer_req->w.list, &mdev->sync_ee);
1689
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
1690

1691
	atomic_add(data_size >> 9, &mdev->rs_sect_ev);
1692
	if (drbd_submit_peer_request(mdev, peer_req, WRITE, DRBD_FAULT_RS_WR) == 0)
1693
		return 0;
P
Philipp Reisner 已提交
1694

1695 1696
	/* don't care for the reason here */
	dev_err(DEV, "submit failed, triggering re-connect\n");
1697
	spin_lock_irq(&mdev->tconn->req_lock);
1698
	list_del(&peer_req->w.list);
1699
	spin_unlock_irq(&mdev->tconn->req_lock);
1700

1701
	drbd_free_peer_req(mdev, peer_req);
1702 1703
fail:
	put_ldev(mdev);
1704
	return -EIO;
P
Philipp Reisner 已提交
1705 1706
}

1707
static struct drbd_request *
1708 1709
find_request(struct drbd_conf *mdev, struct rb_root *root, u64 id,
	     sector_t sector, bool missing_ok, const char *func)
1710 1711 1712
{
	struct drbd_request *req;

1713 1714
	/* Request object according to our peer */
	req = (struct drbd_request *)(unsigned long)id;
1715
	if (drbd_contains_interval(root, sector, &req->i) && req->i.local)
1716
		return req;
1717
	if (!missing_ok) {
1718
		dev_err(DEV, "%s: failed to find request 0x%lx, sector %llus\n", func,
1719 1720
			(unsigned long)id, (unsigned long long)sector);
	}
1721 1722 1723
	return NULL;
}

1724
static int receive_DataReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
1725
{
1726
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
1727 1728
	struct drbd_request *req;
	sector_t sector;
1729
	int err;
1730
	struct p_data *p = pi->data;
1731 1732 1733 1734

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
1735 1736 1737

	sector = be64_to_cpu(p->sector);

1738
	spin_lock_irq(&mdev->tconn->req_lock);
1739
	req = find_request(mdev, &mdev->read_requests, p->block_id, sector, false, __func__);
1740
	spin_unlock_irq(&mdev->tconn->req_lock);
1741
	if (unlikely(!req))
1742
		return -EIO;
P
Philipp Reisner 已提交
1743

B
Bart Van Assche 已提交
1744
	/* hlist_del(&req->collision) is done in _req_may_be_done, to avoid
P
Philipp Reisner 已提交
1745 1746
	 * special casing it there for the various failure cases.
	 * still no race with drbd_fail_pending_reads */
1747
	err = recv_dless_read(mdev, req, sector, pi->size);
1748
	if (!err)
1749
		req_mod(req, DATA_RECEIVED);
P
Philipp Reisner 已提交
1750 1751 1752 1753
	/* else: nothing. handled from drbd_disconnect...
	 * I don't think we may complete this just yet
	 * in case we are "on-disconnect: freeze" */

1754
	return err;
P
Philipp Reisner 已提交
1755 1756
}

1757
static int receive_RSDataReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
1758
{
1759
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
1760
	sector_t sector;
1761
	int err;
1762
	struct p_data *p = pi->data;
1763 1764 1765 1766

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
1767 1768 1769 1770 1771 1772 1773

	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,
1774
		 * or in drbd_peer_request_endio. */
1775
		err = recv_resync_read(mdev, sector, pi->size);
P
Philipp Reisner 已提交
1776 1777 1778 1779
	} else {
		if (__ratelimit(&drbd_ratelimit_state))
			dev_err(DEV, "Can not write resync data to local disk.\n");

1780
		err = drbd_drain_block(mdev, pi->size);
P
Philipp Reisner 已提交
1781

1782
		drbd_send_ack_dp(mdev, P_NEG_ACK, p, pi->size);
P
Philipp Reisner 已提交
1783 1784
	}

1785
	atomic_add(pi->size >> 9, &mdev->rs_sect_in);
1786

1787
	return err;
P
Philipp Reisner 已提交
1788 1789
}

1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802
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;
1803 1804
		/* as it is RQ_POSTPONED, this will cause it to
		 * be queued on the retry workqueue. */
1805
		__req_mod(req, CONFLICT_RESOLVED, NULL);
1806 1807 1808
	}
}

1809 1810
/*
 * e_end_block() is called in asender context via drbd_finish_peer_reqs().
P
Philipp Reisner 已提交
1811
 */
1812
static int e_end_block(struct drbd_work *w, int cancel)
P
Philipp Reisner 已提交
1813
{
1814 1815
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
1816
	struct drbd_conf *mdev = w->mdev;
1817
	sector_t sector = peer_req->i.sector;
1818
	int err = 0, pcmd;
P
Philipp Reisner 已提交
1819

1820
	if (peer_req->flags & EE_SEND_WRITE_ACK) {
1821
		if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) {
P
Philipp Reisner 已提交
1822 1823
			pcmd = (mdev->state.conn >= C_SYNC_SOURCE &&
				mdev->state.conn <= C_PAUSED_SYNC_T &&
1824
				peer_req->flags & EE_MAY_SET_IN_SYNC) ?
P
Philipp Reisner 已提交
1825
				P_RS_WRITE_ACK : P_WRITE_ACK;
1826
			err = drbd_send_ack(mdev, pcmd, peer_req);
P
Philipp Reisner 已提交
1827
			if (pcmd == P_RS_WRITE_ACK)
1828
				drbd_set_in_sync(mdev, sector, peer_req->i.size);
P
Philipp Reisner 已提交
1829
		} else {
1830
			err = drbd_send_ack(mdev, P_NEG_ACK, peer_req);
P
Philipp Reisner 已提交
1831 1832 1833 1834 1835 1836 1837
			/* 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.  */
1838
	if (peer_req->flags & EE_IN_INTERVAL_TREE) {
1839
		spin_lock_irq(&mdev->tconn->req_lock);
1840 1841
		D_ASSERT(!drbd_interval_empty(&peer_req->i));
		drbd_remove_epoch_entry_interval(mdev, peer_req);
1842 1843
		if (peer_req->flags & EE_RESTART_REQUESTS)
			restart_conflicting_writes(mdev, sector, peer_req->i.size);
1844
		spin_unlock_irq(&mdev->tconn->req_lock);
1845
	} else
1846
		D_ASSERT(drbd_interval_empty(&peer_req->i));
P
Philipp Reisner 已提交
1847

1848
	drbd_may_finish_epoch(mdev->tconn, peer_req->epoch, EV_PUT + (cancel ? EV_CLEANUP : 0));
P
Philipp Reisner 已提交
1849

1850
	return err;
P
Philipp Reisner 已提交
1851 1852
}

1853
static int e_send_ack(struct drbd_work *w, enum drbd_packet ack)
P
Philipp Reisner 已提交
1854
{
1855
	struct drbd_conf *mdev = w->mdev;
1856 1857
	struct drbd_peer_request *peer_req =
		container_of(w, struct drbd_peer_request, w);
1858
	int err;
P
Philipp Reisner 已提交
1859

1860
	err = drbd_send_ack(mdev, ack, peer_req);
P
Philipp Reisner 已提交
1861 1862
	dec_unacked(mdev);

1863
	return err;
P
Philipp Reisner 已提交
1864 1865
}

1866
static int e_send_superseded(struct drbd_work *w, int unused)
1867
{
1868
	return e_send_ack(w, P_SUPERSEDED);
1869 1870
}

1871
static int e_send_retry_write(struct drbd_work *w, int unused)
1872 1873 1874 1875
{
	struct drbd_tconn *tconn = w->mdev->tconn;

	return e_send_ack(w, tconn->agreed_pro_version >= 100 ?
1876
			     P_RETRY_WRITE : P_SUPERSEDED);
1877 1878
}

1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893
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;
}

1894 1895 1896
static bool need_peer_seq(struct drbd_conf *mdev)
{
	struct drbd_tconn *tconn = mdev->tconn;
1897
	int tp;
1898 1899 1900

	/*
	 * We only need to keep track of the last packet_seq number of our peer
1901
	 * if we are in dual-primary mode and we have the resolve-conflicts flag set; see
1902 1903
	 * handle_write_conflicts().
	 */
1904 1905 1906 1907 1908

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

1909
	return tp && test_bit(RESOLVE_CONFLICTS, &tconn->flags);
1910 1911
}

1912
static void update_peer_seq(struct drbd_conf *mdev, unsigned int peer_seq)
1913
{
1914
	unsigned int newest_peer_seq;
1915

1916 1917
	if (need_peer_seq(mdev)) {
		spin_lock(&mdev->peer_seq_lock);
1918 1919
		newest_peer_seq = seq_max(mdev->peer_seq, peer_seq);
		mdev->peer_seq = newest_peer_seq;
1920
		spin_unlock(&mdev->peer_seq_lock);
1921 1922
		/* wake up only if we actually changed mdev->peer_seq */
		if (peer_seq == newest_peer_seq)
1923 1924
			wake_up(&mdev->seq_wait);
	}
1925 1926
}

1927 1928 1929 1930 1931 1932
static inline int overlaps(sector_t s1, int l1, sector_t s2, int l2)
{
	return !((s1 + (l1>>9) <= s2) || (s1 >= s2 + (l2>>9)));
}

/* maybe change sync_ee into interval trees as well? */
1933
static bool overlapping_resync_write(struct drbd_conf *mdev, struct drbd_peer_request *peer_req)
1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
{
	struct drbd_peer_request *rs_req;
	bool rv = 0;

	spin_lock_irq(&mdev->tconn->req_lock);
	list_for_each_entry(rs_req, &mdev->sync_ee, w.list) {
		if (overlaps(peer_req->i.sector, peer_req->i.size,
			     rs_req->i.sector, rs_req->i.size)) {
			rv = 1;
			break;
		}
	}
	spin_unlock_irq(&mdev->tconn->req_lock);

	return rv;
}

P
Philipp Reisner 已提交
1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971
/* 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). */
1972
static int wait_for_and_update_peer_seq(struct drbd_conf *mdev, const u32 peer_seq)
P
Philipp Reisner 已提交
1973 1974 1975
{
	DEFINE_WAIT(wait);
	long timeout;
1976 1977 1978 1979 1980
	int ret;

	if (!need_peer_seq(mdev))
		return 0;

P
Philipp Reisner 已提交
1981 1982
	spin_lock(&mdev->peer_seq_lock);
	for (;;) {
1983 1984 1985
		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
1986
			break;
1987
		}
P
Philipp Reisner 已提交
1988 1989 1990 1991
		if (signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}
1992
		prepare_to_wait(&mdev->seq_wait, &wait, TASK_INTERRUPTIBLE);
P
Philipp Reisner 已提交
1993
		spin_unlock(&mdev->peer_seq_lock);
1994 1995 1996
		rcu_read_lock();
		timeout = rcu_dereference(mdev->tconn->net_conf)->ping_timeo*HZ/10;
		rcu_read_unlock();
1997
		timeout = schedule_timeout(timeout);
P
Philipp Reisner 已提交
1998
		spin_lock(&mdev->peer_seq_lock);
1999
		if (!timeout) {
P
Philipp Reisner 已提交
2000
			ret = -ETIMEDOUT;
2001
			dev_err(DEV, "Timed out waiting for missing ack packets; disconnecting\n");
P
Philipp Reisner 已提交
2002 2003 2004 2005
			break;
		}
	}
	spin_unlock(&mdev->peer_seq_lock);
2006
	finish_wait(&mdev->seq_wait, &wait);
P
Philipp Reisner 已提交
2007 2008 2009
	return ret;
}

2010 2011 2012 2013
/* 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)
2014
{
2015 2016 2017 2018
	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);
2019 2020
}

2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049
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;
2050
	bool resolve_conflicts = test_bit(RESOLVE_CONFLICTS, &tconn->flags);
2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
	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
2084 2085 2086
			 * overlapping request, it can be considered overwritten
			 * and thus superseded; otherwise, it will be retried
			 * once all overlapping requests have completed.
2087
			 */
2088
			bool superseded = i->sector <= sector && i->sector +
2089 2090 2091 2092 2093 2094 2095 2096
				       (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,
2097
					  superseded ? "local" : "remote");
2098 2099

			inc_unacked(mdev);
2100
			peer_req->w.cb = superseded ? e_send_superseded :
2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
						   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
2121 2122 2123
				 * decide if this request has been superseded
				 * or needs to be retried.
				 * Requests that have been superseded will
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154
				 * 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
2155
/* mirrored write */
2156
static int receive_Data(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
2157
{
2158
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
2159
	sector_t sector;
2160
	struct drbd_peer_request *peer_req;
2161
	struct p_data *p = pi->data;
2162
	u32 peer_seq = be32_to_cpu(p->seq_num);
P
Philipp Reisner 已提交
2163 2164
	int rw = WRITE;
	u32 dp_flags;
2165
	int err, tp;
P
Philipp Reisner 已提交
2166

2167 2168 2169 2170
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;

2171
	if (!get_ldev(mdev)) {
2172 2173
		int err2;

2174
		err = wait_for_and_update_peer_seq(mdev, peer_seq);
2175
		drbd_send_ack_dp(mdev, P_NEG_ACK, p, pi->size);
2176
		atomic_inc(&tconn->current_epoch->epoch_size);
2177
		err2 = drbd_drain_block(mdev, pi->size);
2178 2179 2180
		if (!err)
			err = err2;
		return err;
P
Philipp Reisner 已提交
2181 2182
	}

2183 2184 2185 2186 2187
	/*
	 * 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
2188 2189

	sector = be64_to_cpu(p->sector);
2190
	peer_req = read_in_block(mdev, p->block_id, sector, pi->size);
2191
	if (!peer_req) {
P
Philipp Reisner 已提交
2192
		put_ldev(mdev);
2193
		return -EIO;
P
Philipp Reisner 已提交
2194 2195
	}

2196
	peer_req->w.cb = e_end_block;
P
Philipp Reisner 已提交
2197

2198 2199
	dp_flags = be32_to_cpu(p->dp_flags);
	rw |= wire_flags_to_bio(mdev, dp_flags);
2200 2201 2202 2203
	if (peer_req->pages == NULL) {
		D_ASSERT(peer_req->i.size == 0);
		D_ASSERT(dp_flags & DP_FLUSH);
	}
2204 2205

	if (dp_flags & DP_MAY_SET_IN_SYNC)
2206
		peer_req->flags |= EE_MAY_SET_IN_SYNC;
2207

2208 2209
	spin_lock(&tconn->epoch_lock);
	peer_req->epoch = tconn->current_epoch;
2210 2211
	atomic_inc(&peer_req->epoch->epoch_size);
	atomic_inc(&peer_req->epoch->active);
2212
	spin_unlock(&tconn->epoch_lock);
P
Philipp Reisner 已提交
2213

2214 2215 2216 2217 2218
	rcu_read_lock();
	tp = rcu_dereference(mdev->tconn->net_conf)->two_primaries;
	rcu_read_unlock();
	if (tp) {
		peer_req->flags |= EE_IN_INTERVAL_TREE;
2219 2220
		err = wait_for_and_update_peer_seq(mdev, peer_seq);
		if (err)
P
Philipp Reisner 已提交
2221
			goto out_interrupted;
2222
		spin_lock_irq(&mdev->tconn->req_lock);
2223 2224 2225 2226
		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
2227
				put_ldev(mdev);
2228
				return 0;
P
Philipp Reisner 已提交
2229
			}
2230
			goto out_interrupted;
P
Philipp Reisner 已提交
2231
		}
2232 2233
	} else
		spin_lock_irq(&mdev->tconn->req_lock);
2234
	list_add(&peer_req->w.list, &mdev->active_ee);
2235
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
2236

2237
	if (mdev->state.conn == C_SYNC_TARGET)
2238
		wait_event(mdev->ee_wait, !overlapping_resync_write(mdev, peer_req));
2239

2240
	if (mdev->tconn->agreed_pro_version < 100) {
2241 2242
		rcu_read_lock();
		switch (rcu_dereference(mdev->tconn->net_conf)->wire_protocol) {
2243 2244 2245 2246 2247 2248 2249
		case DRBD_PROT_C:
			dp_flags |= DP_SEND_WRITE_ACK;
			break;
		case DRBD_PROT_B:
			dp_flags |= DP_SEND_RECEIVE_ACK;
			break;
		}
2250
		rcu_read_unlock();
2251 2252 2253 2254
	}

	if (dp_flags & DP_SEND_WRITE_ACK) {
		peer_req->flags |= EE_SEND_WRITE_ACK;
P
Philipp Reisner 已提交
2255 2256 2257
		inc_unacked(mdev);
		/* corresponding dec_unacked() in e_end_block()
		 * respective _drbd_clear_done_ee */
2258 2259 2260
	}

	if (dp_flags & DP_SEND_RECEIVE_ACK) {
P
Philipp Reisner 已提交
2261 2262
		/* I really don't like it that the receiver thread
		 * sends on the msock, but anyways */
2263
		drbd_send_ack(mdev, P_RECV_ACK, peer_req);
P
Philipp Reisner 已提交
2264 2265
	}

2266
	if (mdev->state.pdsk < D_INCONSISTENT) {
P
Philipp Reisner 已提交
2267
		/* In case we have the only disk of the cluster, */
2268 2269 2270
		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;
2271
		drbd_al_begin_io(mdev, &peer_req->i);
P
Philipp Reisner 已提交
2272 2273
	}

2274 2275 2276
	err = drbd_submit_peer_request(mdev, peer_req, rw, DRBD_FAULT_DT_WR);
	if (!err)
		return 0;
P
Philipp Reisner 已提交
2277

2278 2279
	/* don't care for the reason here */
	dev_err(DEV, "submit failed, triggering re-connect\n");
2280
	spin_lock_irq(&mdev->tconn->req_lock);
2281 2282
	list_del(&peer_req->w.list);
	drbd_remove_epoch_entry_interval(mdev, peer_req);
2283
	spin_unlock_irq(&mdev->tconn->req_lock);
2284
	if (peer_req->flags & EE_CALL_AL_COMPLETE_IO)
2285
		drbd_al_complete_io(mdev, &peer_req->i);
2286

P
Philipp Reisner 已提交
2287
out_interrupted:
2288
	drbd_may_finish_epoch(tconn, peer_req->epoch, EV_PUT + EV_CLEANUP);
P
Philipp Reisner 已提交
2289
	put_ldev(mdev);
2290
	drbd_free_peer_req(mdev, peer_req);
2291
	return err;
P
Philipp Reisner 已提交
2292 2293
}

2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
/* 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.
 */
2305
int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector)
2306 2307 2308
{
	struct gendisk *disk = mdev->ldev->backing_bdev->bd_contains->bd_disk;
	unsigned long db, dt, dbdt;
2309
	struct lc_element *tmp;
2310 2311
	int curr_events;
	int throttle = 0;
P
Philipp Reisner 已提交
2312 2313 2314 2315 2316
	unsigned int c_min_rate;

	rcu_read_lock();
	c_min_rate = rcu_dereference(mdev->ldev->disk_conf)->c_min_rate;
	rcu_read_unlock();
2317 2318

	/* feature disabled? */
P
Philipp Reisner 已提交
2319
	if (c_min_rate == 0)
2320 2321
		return 0;

2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333
	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);

2334 2335 2336
	curr_events = (int)part_stat_read(&disk->part0, sectors[0]) +
		      (int)part_stat_read(&disk->part0, sectors[1]) -
			atomic_read(&mdev->rs_sect_ev);
2337

2338 2339 2340 2341 2342 2343 2344 2345
	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. */
2346 2347 2348 2349 2350 2351
		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;
2352 2353 2354 2355 2356 2357 2358

		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);

P
Philipp Reisner 已提交
2359
		if (dbdt > c_min_rate)
2360 2361 2362 2363 2364 2365
			throttle = 1;
	}
	return throttle;
}


2366
static int receive_DataRequest(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
2367
{
2368
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
2369
	sector_t sector;
2370
	sector_t capacity;
2371
	struct drbd_peer_request *peer_req;
P
Philipp Reisner 已提交
2372
	struct digest_info *di = NULL;
2373
	int size, verb;
P
Philipp Reisner 已提交
2374
	unsigned int fault_type;
2375
	struct p_block_req *p =	pi->data;
2376 2377 2378 2379 2380

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
	capacity = drbd_get_capacity(mdev->this_bdev);
P
Philipp Reisner 已提交
2381 2382 2383 2384

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

2385
	if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) {
P
Philipp Reisner 已提交
2386 2387
		dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__,
				(unsigned long long)sector, size);
2388
		return -EINVAL;
P
Philipp Reisner 已提交
2389 2390 2391 2392
	}
	if (sector + (size>>9) > capacity) {
		dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__,
				(unsigned long long)sector, size);
2393
		return -EINVAL;
P
Philipp Reisner 已提交
2394 2395 2396
	}

	if (!get_ldev_if_state(mdev, D_UP_TO_DATE)) {
2397
		verb = 1;
2398
		switch (pi->cmd) {
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412
		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:
2413
			BUG();
2414 2415
		}
		if (verb && __ratelimit(&drbd_ratelimit_state))
P
Philipp Reisner 已提交
2416 2417
			dev_err(DEV, "Can not satisfy peer's read request, "
			    "no local data.\n");
2418

L
Lars Ellenberg 已提交
2419
		/* drain possibly payload */
2420
		return drbd_drain_block(mdev, pi->size);
P
Philipp Reisner 已提交
2421 2422 2423 2424 2425
	}

	/* 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.  */
2426
	peer_req = drbd_alloc_peer_req(mdev, p->block_id, sector, size, GFP_NOIO);
2427
	if (!peer_req) {
P
Philipp Reisner 已提交
2428
		put_ldev(mdev);
2429
		return -ENOMEM;
P
Philipp Reisner 已提交
2430 2431
	}

2432
	switch (pi->cmd) {
P
Philipp Reisner 已提交
2433
	case P_DATA_REQUEST:
2434
		peer_req->w.cb = w_e_end_data_req;
P
Philipp Reisner 已提交
2435
		fault_type = DRBD_FAULT_DT_RD;
2436 2437 2438
		/* application IO, don't drbd_rs_begin_io */
		goto submit;

P
Philipp Reisner 已提交
2439
	case P_RS_DATA_REQUEST:
2440
		peer_req->w.cb = w_e_end_rsdata_req;
P
Philipp Reisner 已提交
2441
		fault_type = DRBD_FAULT_RS_RD;
2442 2443
		/* used in the sector offset progress display */
		mdev->bm_resync_fo = BM_SECT_TO_BIT(sector);
P
Philipp Reisner 已提交
2444 2445 2446 2447 2448
		break;

	case P_OV_REPLY:
	case P_CSUM_RS_REQUEST:
		fault_type = DRBD_FAULT_RS_RD;
2449
		di = kmalloc(sizeof(*di) + pi->size, GFP_NOIO);
P
Philipp Reisner 已提交
2450 2451 2452
		if (!di)
			goto out_free_e;

2453
		di->digest_size = pi->size;
P
Philipp Reisner 已提交
2454 2455
		di->digest = (((char *)di)+sizeof(struct digest_info));

2456 2457
		peer_req->digest = di;
		peer_req->flags |= EE_HAS_DIGEST;
2458

2459
		if (drbd_recv_all(mdev->tconn, di->digest, pi->size))
P
Philipp Reisner 已提交
2460 2461
			goto out_free_e;

2462
		if (pi->cmd == P_CSUM_RS_REQUEST) {
2463
			D_ASSERT(mdev->tconn->agreed_pro_version >= 89);
2464
			peer_req->w.cb = w_e_end_csum_rs_req;
2465 2466
			/* used in the sector offset progress display */
			mdev->bm_resync_fo = BM_SECT_TO_BIT(sector);
2467
		} else if (pi->cmd == P_OV_REPLY) {
2468 2469
			/* track progress, we may need to throttle */
			atomic_add(size >> 9, &mdev->rs_sect_in);
2470
			peer_req->w.cb = w_e_end_ov_reply;
P
Philipp Reisner 已提交
2471
			dec_rs_pending(mdev);
2472 2473 2474
			/* 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
2475 2476 2477 2478 2479
		}
		break;

	case P_OV_REQUEST:
		if (mdev->ov_start_sector == ~(sector_t)0 &&
2480
		    mdev->tconn->agreed_pro_version >= 90) {
2481 2482
			unsigned long now = jiffies;
			int i;
P
Philipp Reisner 已提交
2483 2484
			mdev->ov_start_sector = sector;
			mdev->ov_position = sector;
2485 2486
			mdev->ov_left = drbd_bm_bits(mdev) - BM_SECT_TO_BIT(sector);
			mdev->rs_total = mdev->ov_left;
2487 2488 2489 2490
			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
2491 2492 2493
			dev_info(DEV, "Online Verify start sector: %llu\n",
					(unsigned long long)sector);
		}
2494
		peer_req->w.cb = w_e_end_ov_req;
P
Philipp Reisner 已提交
2495 2496 2497 2498
		fault_type = DRBD_FAULT_RS_RD;
		break;

	default:
2499
		BUG();
P
Philipp Reisner 已提交
2500 2501
	}

2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523
	/* 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.
	 */
2524 2525 2526
	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))
2527
		goto out_free_e;
P
Philipp Reisner 已提交
2528

2529 2530 2531
submit_for_resync:
	atomic_add(size >> 9, &mdev->rs_sect_ev);

2532
submit:
P
Philipp Reisner 已提交
2533
	inc_unacked(mdev);
2534
	spin_lock_irq(&mdev->tconn->req_lock);
2535
	list_add_tail(&peer_req->w.list, &mdev->read_ee);
2536
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
2537

2538
	if (drbd_submit_peer_request(mdev, peer_req, READ, fault_type) == 0)
2539
		return 0;
P
Philipp Reisner 已提交
2540

2541 2542
	/* don't care for the reason here */
	dev_err(DEV, "submit failed, triggering re-connect\n");
2543
	spin_lock_irq(&mdev->tconn->req_lock);
2544
	list_del(&peer_req->w.list);
2545
	spin_unlock_irq(&mdev->tconn->req_lock);
2546 2547
	/* no drbd_rs_complete_io(), we are dropping the connection anyways */

P
Philipp Reisner 已提交
2548 2549
out_free_e:
	put_ldev(mdev);
2550
	drbd_free_peer_req(mdev, peer_req);
2551
	return -EIO;
P
Philipp Reisner 已提交
2552 2553 2554 2555 2556 2557
}

static int drbd_asb_recover_0p(struct drbd_conf *mdev) __must_hold(local)
{
	int self, peer, rv = -100;
	unsigned long ch_self, ch_peer;
2558
	enum drbd_after_sb_p after_sb_0p;
P
Philipp Reisner 已提交
2559 2560 2561 2562 2563 2564 2565

	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;

2566 2567 2568 2569
	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
2570 2571 2572
	case ASB_CONSENSUS:
	case ASB_DISCARD_SECONDARY:
	case ASB_CALL_HELPER:
2573
	case ASB_VIOLENTLY:
P
Philipp Reisner 已提交
2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597
		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
2598
		dev_warn(DEV, "Discard younger/older primary did not find a decision\n"
P
Philipp Reisner 已提交
2599 2600 2601
		     "Using discard-least-changes instead\n");
	case ASB_DISCARD_ZERO_CHG:
		if (ch_peer == 0 && ch_self == 0) {
2602
			rv = test_bit(RESOLVE_CONFLICTS, &mdev->tconn->flags)
P
Philipp Reisner 已提交
2603 2604 2605 2606 2607 2608
				? -1 : 1;
			break;
		} else {
			if (ch_peer == 0) { rv =  1; break; }
			if (ch_self == 0) { rv = -1; break; }
		}
2609
		if (after_sb_0p == ASB_DISCARD_ZERO_CHG)
P
Philipp Reisner 已提交
2610 2611 2612 2613 2614 2615 2616 2617
			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. */
2618
			rv = test_bit(RESOLVE_CONFLICTS, &mdev->tconn->flags)
P
Philipp Reisner 已提交
2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632
				? -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)
{
2633
	int hg, rv = -100;
2634
	enum drbd_after_sb_p after_sb_1p;
P
Philipp Reisner 已提交
2635

2636 2637 2638 2639
	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
2640 2641 2642 2643 2644
	case ASB_DISCARD_YOUNGER_PRI:
	case ASB_DISCARD_OLDER_PRI:
	case ASB_DISCARD_LEAST_CHG:
	case ASB_DISCARD_LOCAL:
	case ASB_DISCARD_REMOTE:
2645
	case ASB_DISCARD_ZERO_CHG:
P
Philipp Reisner 已提交
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664
		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) {
2665 2666 2667
			enum drbd_state_rv rv2;

			drbd_set_role(mdev, R_SECONDARY, 0);
P
Philipp Reisner 已提交
2668 2669 2670
			 /* 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. */
2671 2672
			rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY));
			if (rv2 != SS_SUCCESS) {
P
Philipp Reisner 已提交
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686
				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)
{
2687
	int hg, rv = -100;
2688
	enum drbd_after_sb_p after_sb_2p;
P
Philipp Reisner 已提交
2689

2690 2691 2692 2693
	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
2694 2695 2696 2697 2698 2699 2700
	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:
2701
	case ASB_DISCARD_ZERO_CHG:
P
Philipp Reisner 已提交
2702 2703 2704 2705 2706 2707 2708 2709 2710 2711
		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) {
2712 2713
			enum drbd_state_rv rv2;

P
Philipp Reisner 已提交
2714 2715 2716
			 /* 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. */
2717 2718
			rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY));
			if (rv2 != SS_SUCCESS) {
P
Philipp Reisner 已提交
2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756
				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
2757 2758
-1091   requires proto 91
-1096   requires proto 96
P
Philipp Reisner 已提交
2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786
 */
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) {

2787
			if (mdev->tconn->agreed_pro_version < 91)
2788
				return -1091;
P
Philipp Reisner 已提交
2789 2790 2791 2792

			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");
2793 2794 2795
				drbd_uuid_move_history(mdev);
				mdev->ldev->md.uuid[UI_HISTORY_START] = mdev->ldev->md.uuid[UI_BITMAP];
				mdev->ldev->md.uuid[UI_BITMAP] = 0;
P
Philipp Reisner 已提交
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809

				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) {

2810
			if (mdev->tconn->agreed_pro_version < 91)
2811
				return -1091;
P
Philipp Reisner 已提交
2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842

			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 */
2843
			dc = test_bit(RESOLVE_CONFLICTS, &mdev->tconn->flags);
P
Philipp Reisner 已提交
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855
			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) {
2856
		if (mdev->tconn->agreed_pro_version < 96 ?
2857 2858 2859
		    (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
2860 2861 2862
			/* The last P_SYNC_UUID did not get though. Undo the last start of
			   resync as sync source modifications of the peer's UUIDs. */

2863
			if (mdev->tconn->agreed_pro_version < 91)
2864
				return -1091;
P
Philipp Reisner 已提交
2865 2866 2867

			mdev->p_uuid[UI_BITMAP] = mdev->p_uuid[UI_HISTORY_START];
			mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_HISTORY_START + 1];
2868

L
Lars Ellenberg 已提交
2869
			dev_info(DEV, "Lost last syncUUID packet, corrected:\n");
2870 2871
			drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]);

P
Philipp Reisner 已提交
2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892
			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) {
2893
		if (mdev->tconn->agreed_pro_version < 96 ?
2894 2895 2896
		    (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
2897 2898 2899
			/* The last P_SYNC_UUID did not get though. Undo the last start of
			   resync as sync source modifications of our UUIDs. */

2900
			if (mdev->tconn->agreed_pro_version < 91)
2901
				return -1091;
P
Philipp Reisner 已提交
2902

2903 2904
			__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]);
P
Philipp Reisner 已提交
2905

2906
			dev_info(DEV, "Last syncUUID did not get through, corrected:\n");
P
Philipp Reisner 已提交
2907 2908 2909 2910 2911 2912 2913 2914 2915
			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;
2916
	peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1);
P
Philipp Reisner 已提交
2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949
	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;
2950
	struct net_conf *nc;
2951
	int hg, rule_nr, rr_conflict, tentative;
P
Philipp Reisner 已提交
2952 2953 2954 2955 2956 2957

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

	dev_info(DEV, "drbd_sync_handshake:\n");
2958 2959

	spin_lock_irq(&mdev->ldev->md.uuid_lock);
P
Philipp Reisner 已提交
2960 2961 2962 2963 2964
	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);
2965
	spin_unlock_irq(&mdev->ldev->md.uuid_lock);
P
Philipp Reisner 已提交
2966 2967 2968 2969 2970 2971 2972

	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;
	}
2973 2974
	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
2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987
		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");
	}

2988 2989 2990
	if (abs(hg) == 100)
		drbd_khelper(mdev, "initial-split-brain");

2991 2992 2993 2994
	rcu_read_lock();
	nc = rcu_dereference(mdev->tconn->net_conf);

	if (hg == 100 || (hg == -100 && nc->always_asbp)) {
P
Philipp Reisner 已提交
2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022
		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) {
3023
		if (test_bit(DISCARD_MY_DATA, &mdev->flags) && !(mdev->p_uuid[UI_FLAGS]&1))
P
Philipp Reisner 已提交
3024
			hg = -1;
3025
		if (!test_bit(DISCARD_MY_DATA, &mdev->flags) && (mdev->p_uuid[UI_FLAGS]&1))
P
Philipp Reisner 已提交
3026 3027 3028 3029 3030 3031 3032
			hg = 1;

		if (abs(hg) < 100)
			dev_warn(DEV, "Split-Brain detected, manually solved. "
			     "Sync from %s node\n",
			     (hg < 0) ? "peer" : "this");
	}
3033
	rr_conflict = nc->rr_conflict;
3034
	tentative = nc->tentative;
3035
	rcu_read_unlock();
P
Philipp Reisner 已提交
3036 3037

	if (hg == -100) {
3038 3039 3040 3041
		/* 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... */
3042
		dev_alert(DEV, "Split-Brain detected but unresolved, dropping connection!\n");
P
Philipp Reisner 已提交
3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
		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) {
3054
		switch (rr_conflict) {
P
Philipp Reisner 已提交
3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066
		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");
		}
	}

3067
	if (tentative || test_bit(CONN_DRY_RUN, &mdev->tconn->flags)) {
3068 3069 3070 3071 3072 3073 3074 3075 3076
		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
3077 3078
	if (abs(hg) >= 2) {
		dev_info(DEV, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n");
3079 3080
		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
3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098
			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;
}

3099
static enum drbd_after_sb_p convert_after_sb(enum drbd_after_sb_p peer)
P
Philipp Reisner 已提交
3100 3101
{
	/* ASB_DISCARD_REMOTE - ASB_DISCARD_LOCAL is valid */
3102 3103
	if (peer == ASB_DISCARD_REMOTE)
		return ASB_DISCARD_LOCAL;
P
Philipp Reisner 已提交
3104 3105

	/* any other things with ASB_DISCARD_REMOTE or ASB_DISCARD_LOCAL are invalid */
3106 3107
	if (peer == ASB_DISCARD_LOCAL)
		return ASB_DISCARD_REMOTE;
P
Philipp Reisner 已提交
3108 3109

	/* everything else is valid if they are equal on both sides. */
3110
	return peer;
P
Philipp Reisner 已提交
3111 3112
}

3113
static int receive_protocol(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3114
{
3115
	struct p_protocol *p = pi->data;
3116 3117 3118 3119
	enum drbd_after_sb_p p_after_sb_0p, p_after_sb_1p, p_after_sb_2p;
	int p_proto, p_discard_my_data, p_two_primaries, cf;
	struct net_conf *nc, *old_net_conf, *new_net_conf = NULL;
	char integrity_alg[SHARED_SECRET_MAX] = "";
3120
	struct crypto_hash *peer_integrity_tfm = NULL;
3121
	void *int_dig_in = NULL, *int_dig_vv = NULL;
P
Philipp Reisner 已提交
3122 3123 3124 3125 3126 3127

	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);
3128
	cf		= be32_to_cpu(p->conn_flags);
3129
	p_discard_my_data = cf & CF_DISCARD_MY_DATA;
3130

3131 3132 3133
	if (tconn->agreed_pro_version >= 87) {
		int err;

3134
		if (pi->size > sizeof(integrity_alg))
3135
			return -EIO;
3136
		err = drbd_recv_all(tconn, integrity_alg, pi->size);
3137 3138
		if (err)
			return err;
3139 3140
		integrity_alg[SHARED_SECRET_MAX - 1] = 0;
	}
3141

3142
	if (pi->cmd != P_PROTOCOL_UPDATE) {
3143
		clear_bit(CONN_DRY_RUN, &tconn->flags);
3144

3145 3146
		if (cf & CF_DRY_RUN)
			set_bit(CONN_DRY_RUN, &tconn->flags);
3147

3148 3149
		rcu_read_lock();
		nc = rcu_dereference(tconn->net_conf);
P
Philipp Reisner 已提交
3150

3151
		if (p_proto != nc->wire_protocol) {
3152
			conn_err(tconn, "incompatible %s settings\n", "protocol");
3153 3154
			goto disconnect_rcu_unlock;
		}
3155

3156
		if (convert_after_sb(p_after_sb_0p) != nc->after_sb_0p) {
3157
			conn_err(tconn, "incompatible %s settings\n", "after-sb-0pri");
3158 3159
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3160

3161
		if (convert_after_sb(p_after_sb_1p) != nc->after_sb_1p) {
3162
			conn_err(tconn, "incompatible %s settings\n", "after-sb-1pri");
3163 3164
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3165

3166
		if (convert_after_sb(p_after_sb_2p) != nc->after_sb_2p) {
3167
			conn_err(tconn, "incompatible %s settings\n", "after-sb-2pri");
3168 3169
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3170

3171
		if (p_discard_my_data && nc->discard_my_data) {
3172
			conn_err(tconn, "incompatible %s settings\n", "discard-my-data");
3173 3174
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3175

3176
		if (p_two_primaries != nc->two_primaries) {
3177
			conn_err(tconn, "incompatible %s settings\n", "allow-two-primaries");
3178 3179
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3180

3181
		if (strcmp(integrity_alg, nc->integrity_alg)) {
3182
			conn_err(tconn, "incompatible %s settings\n", "data-integrity-alg");
3183 3184
			goto disconnect_rcu_unlock;
		}
P
Philipp Reisner 已提交
3185

3186
		rcu_read_unlock();
3187
	}
3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250

	if (integrity_alg[0]) {
		int hash_size;

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

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

		hash_size = crypto_hash_digestsize(peer_integrity_tfm);
		int_dig_in = kmalloc(hash_size, GFP_KERNEL);
		int_dig_vv = kmalloc(hash_size, GFP_KERNEL);
		if (!(int_dig_in && int_dig_vv)) {
			conn_err(tconn, "Allocation of buffers for data integrity checking failed\n");
			goto disconnect;
		}
	}

	new_net_conf = kmalloc(sizeof(struct net_conf), GFP_KERNEL);
	if (!new_net_conf) {
		conn_err(tconn, "Allocation of new net_conf failed\n");
		goto disconnect;
	}

	mutex_lock(&tconn->data.mutex);
	mutex_lock(&tconn->conf_update);
	old_net_conf = tconn->net_conf;
	*new_net_conf = *old_net_conf;

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

	rcu_assign_pointer(tconn->net_conf, new_net_conf);
	mutex_unlock(&tconn->conf_update);
	mutex_unlock(&tconn->data.mutex);

	crypto_free_hash(tconn->peer_integrity_tfm);
	kfree(tconn->int_dig_in);
	kfree(tconn->int_dig_vv);
	tconn->peer_integrity_tfm = peer_integrity_tfm;
	tconn->int_dig_in = int_dig_in;
	tconn->int_dig_vv = int_dig_vv;

	if (strcmp(old_net_conf->integrity_alg, integrity_alg))
		conn_info(tconn, "peer data-integrity-alg: %s\n",
			  integrity_alg[0] ? integrity_alg : "(none)");

	synchronize_rcu();
	kfree(old_net_conf);
3251
	return 0;
P
Philipp Reisner 已提交
3252

3253 3254
disconnect_rcu_unlock:
	rcu_read_unlock();
P
Philipp Reisner 已提交
3255
disconnect:
3256
	crypto_free_hash(peer_integrity_tfm);
3257 3258
	kfree(int_dig_in);
	kfree(int_dig_vv);
3259
	conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3260
	return -EIO;
P
Philipp Reisner 已提交
3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284
}

/* 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;
}

3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317
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)
{
3318 3319
	conn_warn(tconn, "%s packet received for volume %u, which is not configured locally\n",
		  cmdname(pi->cmd), pi->vnr);
3320 3321 3322 3323
	return ignore_remaining_packet(tconn, pi);
}

static int receive_SyncParam(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3324
{
3325
	struct drbd_conf *mdev;
3326
	struct p_rs_param_95 *p;
P
Philipp Reisner 已提交
3327 3328 3329
	unsigned int header_size, data_size, exp_max_sz;
	struct crypto_hash *verify_tfm = NULL;
	struct crypto_hash *csums_tfm = NULL;
3330
	struct net_conf *old_net_conf, *new_net_conf = NULL;
P
Philipp Reisner 已提交
3331
	struct disk_conf *old_disk_conf = NULL, *new_disk_conf = NULL;
3332
	const int apv = tconn->agreed_pro_version;
P
Philipp Reisner 已提交
3333
	struct fifo_buffer *old_plan = NULL, *new_plan = NULL;
3334
	int fifo_size = 0;
3335
	int err;
P
Philipp Reisner 已提交
3336

3337 3338 3339 3340
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3341 3342 3343
	exp_max_sz  = apv <= 87 ? sizeof(struct p_rs_param)
		    : apv == 88 ? sizeof(struct p_rs_param)
					+ SHARED_SECRET_MAX
3344 3345
		    : apv <= 94 ? sizeof(struct p_rs_param_89)
		    : /* apv >= 95 */ sizeof(struct p_rs_param_95);
P
Philipp Reisner 已提交
3346

3347
	if (pi->size > exp_max_sz) {
P
Philipp Reisner 已提交
3348
		dev_err(DEV, "SyncParam packet too long: received %u, expected <= %u bytes\n",
3349
		    pi->size, exp_max_sz);
3350
		return -EIO;
P
Philipp Reisner 已提交
3351 3352 3353
	}

	if (apv <= 88) {
3354
		header_size = sizeof(struct p_rs_param);
3355
		data_size = pi->size - header_size;
3356
	} else if (apv <= 94) {
3357
		header_size = sizeof(struct p_rs_param_89);
3358
		data_size = pi->size - header_size;
P
Philipp Reisner 已提交
3359
		D_ASSERT(data_size == 0);
3360
	} else {
3361
		header_size = sizeof(struct p_rs_param_95);
3362
		data_size = pi->size - header_size;
P
Philipp Reisner 已提交
3363 3364 3365 3366
		D_ASSERT(data_size == 0);
	}

	/* initialize verify_alg and csums_alg */
3367
	p = pi->data;
P
Philipp Reisner 已提交
3368 3369
	memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX);

3370
	err = drbd_recv_all(mdev->tconn, p, header_size);
3371 3372
	if (err)
		return err;
P
Philipp Reisner 已提交
3373

P
Philipp Reisner 已提交
3374 3375
	mutex_lock(&mdev->tconn->conf_update);
	old_net_conf = mdev->tconn->net_conf;
P
Philipp Reisner 已提交
3376 3377 3378 3379 3380 3381 3382 3383
	if (get_ldev(mdev)) {
		new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL);
		if (!new_disk_conf) {
			put_ldev(mdev);
			mutex_unlock(&mdev->tconn->conf_update);
			dev_err(DEV, "Allocation of new disk_conf failed\n");
			return -ENOMEM;
		}
P
Philipp Reisner 已提交
3384

P
Philipp Reisner 已提交
3385 3386 3387
		old_disk_conf = mdev->ldev->disk_conf;
		*new_disk_conf = *old_disk_conf;

3388
		new_disk_conf->resync_rate = be32_to_cpu(p->resync_rate);
P
Philipp Reisner 已提交
3389
	}
P
Philipp Reisner 已提交
3390

P
Philipp Reisner 已提交
3391 3392
	if (apv >= 88) {
		if (apv == 88) {
3393 3394 3395 3396
			if (data_size > SHARED_SECRET_MAX || data_size == 0) {
				dev_err(DEV, "verify-alg of wrong size, "
					"peer wants %u, accepting only up to %u byte\n",
					data_size, SHARED_SECRET_MAX);
P
Philipp Reisner 已提交
3397 3398
				err = -EIO;
				goto reconnect;
P
Philipp Reisner 已提交
3399 3400
			}

3401
			err = drbd_recv_all(mdev->tconn, p->verify_alg, data_size);
P
Philipp Reisner 已提交
3402 3403
			if (err)
				goto reconnect;
P
Philipp Reisner 已提交
3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417
			/* 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;
		}

3418
		if (strcmp(old_net_conf->verify_alg, p->verify_alg)) {
P
Philipp Reisner 已提交
3419 3420
			if (mdev->state.conn == C_WF_REPORT_PARAMS) {
				dev_err(DEV, "Different verify-alg settings. me=\"%s\" peer=\"%s\"\n",
3421
				    old_net_conf->verify_alg, p->verify_alg);
P
Philipp Reisner 已提交
3422 3423 3424 3425 3426 3427 3428 3429 3430 3431
				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;
			}
		}

3432
		if (apv >= 89 && strcmp(old_net_conf->csums_alg, p->csums_alg)) {
P
Philipp Reisner 已提交
3433 3434
			if (mdev->state.conn == C_WF_REPORT_PARAMS) {
				dev_err(DEV, "Different csums-alg settings. me=\"%s\" peer=\"%s\"\n",
3435
				    old_net_conf->csums_alg, p->csums_alg);
P
Philipp Reisner 已提交
3436 3437 3438 3439 3440 3441 3442 3443 3444 3445
				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;
			}
		}

P
Philipp Reisner 已提交
3446
		if (apv > 94 && new_disk_conf) {
P
Philipp Reisner 已提交
3447 3448 3449 3450
			new_disk_conf->c_plan_ahead = be32_to_cpu(p->c_plan_ahead);
			new_disk_conf->c_delay_target = be32_to_cpu(p->c_delay_target);
			new_disk_conf->c_fill_target = be32_to_cpu(p->c_fill_target);
			new_disk_conf->c_max_rate = be32_to_cpu(p->c_max_rate);
3451

P
Philipp Reisner 已提交
3452
			fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ;
3453
			if (fifo_size != mdev->rs_plan_s->size) {
P
Philipp Reisner 已提交
3454 3455
				new_plan = fifo_alloc(fifo_size);
				if (!new_plan) {
3456
					dev_err(DEV, "kmalloc of fifo_buffer failed");
3457
					put_ldev(mdev);
3458 3459 3460
					goto disconnect;
				}
			}
3461
		}
P
Philipp Reisner 已提交
3462

3463
		if (verify_tfm || csums_tfm) {
3464 3465
			new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
			if (!new_net_conf) {
3466 3467 3468 3469
				dev_err(DEV, "Allocation of new net_conf failed\n");
				goto disconnect;
			}

3470
			*new_net_conf = *old_net_conf;
3471 3472

			if (verify_tfm) {
3473 3474
				strcpy(new_net_conf->verify_alg, p->verify_alg);
				new_net_conf->verify_alg_len = strlen(p->verify_alg) + 1;
3475 3476 3477 3478 3479
				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) {
3480 3481
				strcpy(new_net_conf->csums_alg, p->csums_alg);
				new_net_conf->csums_alg_len = strlen(p->csums_alg) + 1;
3482 3483 3484 3485
				crypto_free_hash(mdev->tconn->csums_tfm);
				mdev->tconn->csums_tfm = csums_tfm;
				dev_info(DEV, "using csums-alg: \"%s\"\n", p->csums_alg);
			}
3486
			rcu_assign_pointer(tconn->net_conf, new_net_conf);
P
Philipp Reisner 已提交
3487
		}
P
Philipp Reisner 已提交
3488
	}
3489

P
Philipp Reisner 已提交
3490 3491 3492 3493 3494 3495 3496 3497
	if (new_disk_conf) {
		rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf);
		put_ldev(mdev);
	}

	if (new_plan) {
		old_plan = mdev->rs_plan_s;
		rcu_assign_pointer(mdev->rs_plan_s, new_plan);
P
Philipp Reisner 已提交
3498
	}
P
Philipp Reisner 已提交
3499 3500 3501 3502 3503 3504

	mutex_unlock(&mdev->tconn->conf_update);
	synchronize_rcu();
	if (new_net_conf)
		kfree(old_net_conf);
	kfree(old_disk_conf);
P
Philipp Reisner 已提交
3505
	kfree(old_plan);
P
Philipp Reisner 已提交
3506

3507
	return 0;
P
Philipp Reisner 已提交
3508

P
Philipp Reisner 已提交
3509 3510 3511 3512 3513 3514 3515 3516
reconnect:
	if (new_disk_conf) {
		put_ldev(mdev);
		kfree(new_disk_conf);
	}
	mutex_unlock(&mdev->tconn->conf_update);
	return -EIO;

P
Philipp Reisner 已提交
3517
disconnect:
P
Philipp Reisner 已提交
3518 3519 3520 3521 3522
	kfree(new_plan);
	if (new_disk_conf) {
		put_ldev(mdev);
		kfree(new_disk_conf);
	}
3523
	mutex_unlock(&mdev->tconn->conf_update);
P
Philipp Reisner 已提交
3524 3525 3526 3527 3528
	/* 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);
3529
	conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3530
	return -EIO;
P
Philipp Reisner 已提交
3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545
}

/* 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);
}

3546
static int receive_sizes(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3547
{
3548
	struct drbd_conf *mdev;
3549
	struct p_sizes *p = pi->data;
P
Philipp Reisner 已提交
3550 3551 3552
	enum determine_dev_size dd = unchanged;
	sector_t p_size, p_usize, my_usize;
	int ldsc = 0; /* local disk size changed */
3553
	enum dds_flags ddsf;
P
Philipp Reisner 已提交
3554

3555 3556 3557 3558
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3559 3560 3561 3562 3563 3564 3565 3566
	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)) {
P
Philipp Reisner 已提交
3567 3568 3569 3570
		rcu_read_lock();
		my_usize = rcu_dereference(mdev->ldev->disk_conf)->disk_size;
		rcu_read_unlock();

P
Philipp Reisner 已提交
3571 3572 3573
		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
Philipp Reisner 已提交
3574
					    p_usize, my_usize);
P
Philipp Reisner 已提交
3575 3576 3577 3578

		/* if this is the first connect, or an otherwise expected
		 * param exchange, choose the minimum */
		if (mdev->state.conn == C_WF_REPORT_PARAMS)
P
Philipp Reisner 已提交
3579
			p_usize = min_not_zero(my_usize, p_usize);
P
Philipp Reisner 已提交
3580 3581 3582

		/* Never shrink a device with usable data during connect.
		   But allow online shrinking if we are connected. */
3583
		if (drbd_new_dev_size(mdev, mdev->ldev, p_usize, 0) <
P
Philipp Reisner 已提交
3584 3585 3586
		    drbd_get_capacity(mdev->this_bdev) &&
		    mdev->state.disk >= D_OUTDATED &&
		    mdev->state.conn < C_CONNECTED) {
P
Philipp Reisner 已提交
3587
			dev_err(DEV, "The peer's disk size is too small!\n");
3588
			conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
3589
			put_ldev(mdev);
3590
			return -EIO;
P
Philipp Reisner 已提交
3591
		}
P
Philipp Reisner 已提交
3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616

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

			new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL);
			if (!new_disk_conf) {
				dev_err(DEV, "Allocation of new disk_conf failed\n");
				put_ldev(mdev);
				return -ENOMEM;
			}

			mutex_lock(&mdev->tconn->conf_update);
			old_disk_conf = mdev->ldev->disk_conf;
			*new_disk_conf = *old_disk_conf;
			new_disk_conf->disk_size = p_usize;

			rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf);
			mutex_unlock(&mdev->tconn->conf_update);
			synchronize_rcu();
			kfree(old_disk_conf);

			dev_info(DEV, "Peer sets u_size to %lu sectors\n",
				 (unsigned long)my_usize);
		}

P
Philipp Reisner 已提交
3617 3618 3619
		put_ldev(mdev);
	}

3620
	ddsf = be16_to_cpu(p->dds_flags);
P
Philipp Reisner 已提交
3621
	if (get_ldev(mdev)) {
B
Bart Van Assche 已提交
3622
		dd = drbd_determine_dev_size(mdev, ddsf);
P
Philipp Reisner 已提交
3623 3624
		put_ldev(mdev);
		if (dd == dev_size_error)
3625
			return -EIO;
P
Philipp Reisner 已提交
3626 3627 3628 3629 3630 3631
		drbd_md_sync(mdev);
	} else {
		/* I am diskless, need to accept the peer's size. */
		drbd_set_my_capacity(mdev, p_size);
	}

3632 3633 3634
	mdev->peer_max_bio_size = be32_to_cpu(p->max_bio_size);
	drbd_reconsider_max_bio_size(mdev);

P
Philipp Reisner 已提交
3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648
	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... */
3649
			drbd_send_sizes(mdev, 0, ddsf);
P
Philipp Reisner 已提交
3650 3651 3652 3653
		}
		if (test_and_clear_bit(RESIZE_PENDING, &mdev->flags) ||
		    (dd == grew && mdev->state.conn == C_CONNECTED)) {
			if (mdev->state.pdsk >= D_INCONSISTENT &&
3654 3655 3656 3657 3658 3659
			    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
3660 3661 3662 3663
				set_bit(RESYNC_AFTER_NEG, &mdev->flags);
		}
	}

3664
	return 0;
P
Philipp Reisner 已提交
3665 3666
}

3667
static int receive_uuids(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3668
{
3669
	struct drbd_conf *mdev;
3670
	struct p_uuids *p = pi->data;
P
Philipp Reisner 已提交
3671
	u64 *p_uuid;
3672
	int i, updated_uuids = 0;
P
Philipp Reisner 已提交
3673

3674 3675 3676 3677
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691
	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);
3692
		conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3693
		return -EIO;
P
Philipp Reisner 已提交
3694 3695 3696 3697 3698
	}

	if (get_ldev(mdev)) {
		int skip_initial_sync =
			mdev->state.conn == C_CONNECTED &&
3699
			mdev->tconn->agreed_pro_version >= 90 &&
P
Philipp Reisner 已提交
3700 3701 3702 3703 3704
			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,
3705 3706
					"clear_n_write from receive_uuids",
					BM_LOCKED_TEST_ALLOWED);
P
Philipp Reisner 已提交
3707 3708 3709 3710 3711
			_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);
3712
			updated_uuids = 1;
P
Philipp Reisner 已提交
3713 3714
		}
		put_ldev(mdev);
3715 3716 3717 3718
	} 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. */
3719
		updated_uuids = drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]);
P
Philipp Reisner 已提交
3720 3721 3722 3723 3724 3725
	}

	/* 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... */
3726 3727
	mutex_lock(mdev->state_mutex);
	mutex_unlock(mdev->state_mutex);
P
Philipp Reisner 已提交
3728
	if (mdev->state.conn >= C_CONNECTED && mdev->state.disk < D_INCONSISTENT)
3729 3730 3731 3732
		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
3733

3734
	return 0;
P
Philipp Reisner 已提交
3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745
}

/**
 * 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[] = {
3746
		[C_WF_REPORT_PARAMS] = C_WF_REPORT_PARAMS,
P
Philipp Reisner 已提交
3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767
		[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;
}

3768
static int receive_req_state(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3769
{
3770
	struct drbd_conf *mdev;
3771
	struct p_req_state *p = pi->data;
P
Philipp Reisner 已提交
3772
	union drbd_state mask, val;
3773
	enum drbd_state_rv rv;
P
Philipp Reisner 已提交
3774

3775 3776 3777 3778
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;

P
Philipp Reisner 已提交
3779 3780 3781
	mask.i = be32_to_cpu(p->mask);
	val.i = be32_to_cpu(p->val);

3782
	if (test_bit(RESOLVE_CONFLICTS, &mdev->tconn->flags) &&
3783
	    mutex_is_locked(mdev->state_mutex)) {
P
Philipp Reisner 已提交
3784
		drbd_send_sr_reply(mdev, SS_CONCURRENT_ST_CHG);
3785
		return 0;
P
Philipp Reisner 已提交
3786 3787 3788 3789 3790
	}

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

3791 3792
	rv = drbd_change_state(mdev, CS_VERBOSE, mask, val);
	drbd_send_sr_reply(mdev, rv);
P
Philipp Reisner 已提交
3793 3794 3795

	drbd_md_sync(mdev);

3796
	return 0;
P
Philipp Reisner 已提交
3797 3798
}

3799
static int receive_req_conn_state(struct drbd_tconn *tconn, struct packet_info *pi)
3800
{
3801
	struct p_req_state *p = pi->data;
3802 3803 3804 3805 3806 3807
	union drbd_state mask, val;
	enum drbd_state_rv rv;

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

3808
	if (test_bit(RESOLVE_CONFLICTS, &tconn->flags) &&
3809 3810
	    mutex_is_locked(&tconn->cstate_mutex)) {
		conn_send_sr_reply(tconn, SS_CONCURRENT_ST_CHG);
3811
		return 0;
3812 3813 3814 3815 3816
	}

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

3817
	rv = conn_request_state(tconn, mask, val, CS_VERBOSE | CS_LOCAL_ONLY | CS_IGN_OUTD_FAIL);
3818 3819
	conn_send_sr_reply(tconn, rv);

3820
	return 0;
3821 3822
}

3823
static int receive_state(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
3824
{
3825
	struct drbd_conf *mdev;
3826
	struct p_state *p = pi->data;
3827
	union drbd_state os, ns, peer_state;
P
Philipp Reisner 已提交
3828
	enum drbd_disk_state real_peer_disk;
3829
	enum chg_state_flags cs_flags;
P
Philipp Reisner 已提交
3830 3831
	int rv;

3832 3833 3834 3835
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return config_unknown_volume(tconn, pi);

P
Philipp Reisner 已提交
3836 3837 3838 3839 3840 3841 3842 3843
	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));
	}

3844
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
3845
 retry:
3846
	os = ns = drbd_read_state(mdev);
3847
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
3848

3849 3850 3851 3852
	/* If some other part of the code (asender thread, timeout)
	 * already decided to close the connection again,
	 * we must not "re-establish" it here. */
	if (os.conn <= C_TEAR_DOWN)
3853
		return -ECONNRESET;
3854

P
Philipp Reisner 已提交
3855 3856 3857 3858 3859 3860 3861 3862
	/* If this is the "end of sync" confirmation, usually the peer disk
	 * transitions from D_INCONSISTENT to D_UP_TO_DATE. For empty (0 bits
	 * set) resync started in PausedSyncT, or if the timing of pause-/
	 * unpause-sync events has been "just right", the peer disk may
	 * transition from D_CONSISTENT to D_UP_TO_DATE as well.
	 */
	if ((os.pdsk == D_INCONSISTENT || os.pdsk == D_CONSISTENT) &&
	    real_peer_disk == D_UP_TO_DATE &&
3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880
	    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);
3881
			return 0;
3882 3883 3884
		}
	}

3885 3886 3887 3888 3889 3890 3891 3892
	/* explicit verify finished notification, stop sector reached. */
	if (os.conn == C_VERIFY_T && os.disk == D_UP_TO_DATE &&
	    peer_state.conn == C_CONNECTED && real_peer_disk == D_UP_TO_DATE) {
		ov_out_of_sync_print(mdev);
		drbd_resync_finished(mdev);
		return 0;
	}

3893 3894 3895 3896 3897 3898 3899 3900 3901
	/* 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;

3902 3903
	if (ns.conn == C_WF_REPORT_PARAMS)
		ns.conn = C_CONNECTED;
P
Philipp Reisner 已提交
3904

3905 3906 3907
	if (peer_state.conn == C_AHEAD)
		ns.conn = C_BEHIND;

P
Philipp Reisner 已提交
3908 3909 3910 3911 3912
	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 */
3913
		cr  = (os.conn < C_CONNECTED);
P
Philipp Reisner 已提交
3914 3915
		/* if we had an established connection
		 * and one of the nodes newly attaches a disk */
3916
		cr |= (os.conn == C_CONNECTED &&
P
Philipp Reisner 已提交
3917
		       (peer_state.disk == D_NEGOTIATING ||
3918
			os.disk == D_NEGOTIATING));
P
Philipp Reisner 已提交
3919 3920 3921 3922 3923
		/* 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" */
3924
		cr |= (os.conn == C_CONNECTED &&
P
Philipp Reisner 已提交
3925 3926 3927 3928
				(peer_state.conn >= C_STARTING_SYNC_S &&
				 peer_state.conn <= C_WF_BITMAP_T));

		if (cr)
3929
			ns.conn = drbd_sync_handshake(mdev, peer_state.role, real_peer_disk);
P
Philipp Reisner 已提交
3930 3931

		put_ldev(mdev);
3932 3933
		if (ns.conn == C_MASK) {
			ns.conn = C_CONNECTED;
P
Philipp Reisner 已提交
3934
			if (mdev->state.disk == D_NEGOTIATING) {
3935
				drbd_force_state(mdev, NS(disk, D_FAILED));
P
Philipp Reisner 已提交
3936 3937 3938
			} 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;
3939
				real_peer_disk = D_DISKLESS;
P
Philipp Reisner 已提交
3940
			} else {
3941
				if (test_and_clear_bit(CONN_DRY_RUN, &mdev->tconn->flags))
3942
					return -EIO;
3943
				D_ASSERT(os.conn == C_WF_REPORT_PARAMS);
3944
				conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3945
				return -EIO;
P
Philipp Reisner 已提交
3946 3947 3948 3949
			}
		}
	}

3950
	spin_lock_irq(&mdev->tconn->req_lock);
3951
	if (os.i != drbd_read_state(mdev).i)
P
Philipp Reisner 已提交
3952 3953 3954 3955 3956
		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);
3957
	if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING)
P
Philipp Reisner 已提交
3958
		ns.disk = mdev->new_state_tmp.disk;
3959
	cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD);
3960
	if (ns.pdsk == D_CONSISTENT && drbd_suspended(mdev) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED &&
3961
	    test_bit(NEW_CUR_UUID, &mdev->flags)) {
3962
		/* Do not allow tl_restart(RESEND) for a rebooted peer. We can only allow this
3963
		   for temporal network outages! */
3964
		spin_unlock_irq(&mdev->tconn->req_lock);
3965
		dev_err(DEV, "Aborting Connect, can not thaw IO with an only Consistent peer\n");
3966
		tl_clear(mdev->tconn);
3967 3968
		drbd_uuid_new_current(mdev);
		clear_bit(NEW_CUR_UUID, &mdev->flags);
3969
		conn_request_state(mdev->tconn, NS2(conn, C_PROTOCOL_ERROR, susp, 0), CS_HARD);
3970
		return -EIO;
3971
	}
3972
	rv = _drbd_set_state(mdev, ns, cs_flags, NULL);
3973
	ns = drbd_read_state(mdev);
3974
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
3975 3976

	if (rv < SS_SUCCESS) {
3977
		conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD);
3978
		return -EIO;
P
Philipp Reisner 已提交
3979 3980
	}

3981 3982
	if (os.conn > C_WF_REPORT_PARAMS) {
		if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED &&
P
Philipp Reisner 已提交
3983 3984 3985 3986 3987
		    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);
3988
			drbd_send_current_state(mdev);
P
Philipp Reisner 已提交
3989 3990 3991
		}
	}

3992
	clear_bit(DISCARD_MY_DATA, &mdev->flags);
P
Philipp Reisner 已提交
3993 3994 3995

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

3996
	return 0;
P
Philipp Reisner 已提交
3997 3998
}

3999
static int receive_sync_uuid(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4000
{
4001
	struct drbd_conf *mdev;
4002
	struct p_rs_uuid *p = pi->data;
4003 4004 4005 4006

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
4007 4008 4009

	wait_event(mdev->misc_wait,
		   mdev->state.conn == C_WF_SYNC_UUID ||
4010
		   mdev->state.conn == C_BEHIND ||
P
Philipp Reisner 已提交
4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021
		   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);

4022
		drbd_print_uuids(mdev, "updated sync uuid");
P
Philipp Reisner 已提交
4023 4024 4025 4026 4027 4028
		drbd_start_resync(mdev, C_SYNC_TARGET);

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

4029
	return 0;
P
Philipp Reisner 已提交
4030 4031
}

4032 4033 4034 4035 4036 4037 4038
/**
 * receive_bitmap_plain
 *
 * Return 0 when done, 1 when another iteration is needed, and a negative error
 * code upon failure.
 */
static int
4039
receive_bitmap_plain(struct drbd_conf *mdev, unsigned int size,
4040
		     unsigned long *p, struct bm_xfer_ctx *c)
P
Philipp Reisner 已提交
4041
{
4042 4043
	unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE -
				 drbd_header_size(mdev->tconn);
4044
	unsigned int num_words = min_t(size_t, data_size / sizeof(*p),
4045
				       c->bm_words - c->word_offset);
4046
	unsigned int want = num_words * sizeof(*p);
4047
	int err;
P
Philipp Reisner 已提交
4048

4049 4050
	if (want != size) {
		dev_err(DEV, "%s:want (%u) != size (%u)\n", __func__, want, size);
4051
		return -EIO;
P
Philipp Reisner 已提交
4052 4053
	}
	if (want == 0)
4054
		return 0;
4055
	err = drbd_recv_all(mdev->tconn, p, want);
4056
	if (err)
4057
		return err;
P
Philipp Reisner 已提交
4058

4059
	drbd_bm_merge_lel(mdev, c->word_offset, num_words, p);
P
Philipp Reisner 已提交
4060 4061 4062 4063 4064 4065

	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;

4066
	return 1;
P
Philipp Reisner 已提交
4067 4068
}

4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083
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;
}

4084 4085 4086 4087 4088 4089 4090
/**
 * 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
4091 4092
recv_bm_rle_bits(struct drbd_conf *mdev,
		struct p_compressed_bm *p,
4093 4094
		 struct bm_xfer_ctx *c,
		 unsigned int len)
P
Philipp Reisner 已提交
4095 4096 4097 4098 4099 4100 4101
{
	struct bitstream bs;
	u64 look_ahead;
	u64 rl;
	u64 tmp;
	unsigned long s = c->bit_offset;
	unsigned long e;
4102
	int toggle = dcbp_get_start(p);
P
Philipp Reisner 已提交
4103 4104 4105
	int have;
	int bits;

4106
	bitstream_init(&bs, p->code, len, dcbp_get_pad_bits(p));
P
Philipp Reisner 已提交
4107 4108 4109

	bits = bitstream_get_bits(&bs, &look_ahead, 64);
	if (bits < 0)
4110
		return -EIO;
P
Philipp Reisner 已提交
4111 4112 4113 4114

	for (have = bits; have > 0; s += rl, toggle = !toggle) {
		bits = vli_decode_bits(&rl, look_ahead);
		if (bits <= 0)
4115
			return -EIO;
P
Philipp Reisner 已提交
4116 4117 4118 4119 4120

		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);
4121
				return -EIO;
P
Philipp Reisner 已提交
4122 4123 4124 4125 4126 4127 4128 4129 4130
			}
			_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);
4131
			return -EIO;
P
Philipp Reisner 已提交
4132 4133 4134 4135 4136 4137
		}
		look_ahead >>= bits;
		have -= bits;

		bits = bitstream_get_bits(&bs, &tmp, 64 - have);
		if (bits < 0)
4138
			return -EIO;
P
Philipp Reisner 已提交
4139 4140 4141 4142 4143 4144 4145
		look_ahead |= tmp << have;
		have += bits;
	}

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

4146
	return (s != c->bm_bits);
P
Philipp Reisner 已提交
4147 4148
}

4149 4150 4151 4152 4153 4154 4155
/**
 * 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
4156 4157
decode_bitmap_c(struct drbd_conf *mdev,
		struct p_compressed_bm *p,
4158 4159
		struct bm_xfer_ctx *c,
		unsigned int len)
P
Philipp Reisner 已提交
4160
{
4161
	if (dcbp_get_code(p) == RLE_VLI_Bits)
4162
		return recv_bm_rle_bits(mdev, p, c, len - sizeof(*p));
P
Philipp Reisner 已提交
4163 4164 4165 4166 4167 4168

	/* 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);
4169
	conn_request_state(mdev->tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
4170
	return -EIO;
P
Philipp Reisner 已提交
4171 4172 4173 4174 4175 4176
}

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" */
4177 4178 4179 4180 4181 4182 4183
	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
4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216

	/* 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. */
4217
static int receive_bitmap(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4218
{
4219
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
4220
	struct bm_xfer_ctx c;
4221
	int err;
4222 4223 4224 4225

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
P
Philipp Reisner 已提交
4226

4227 4228 4229
	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
4230 4231 4232 4233 4234 4235

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

4236
	for(;;) {
4237 4238 4239
		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
4240 4241
			/* MAYBE: sanity check that we speak proto >= 90,
			 * and the feature is enabled! */
4242
			struct p_compressed_bm *p = pi->data;
P
Philipp Reisner 已提交
4243

4244
			if (pi->size > DRBD_SOCKET_BUFFER_SIZE - drbd_header_size(tconn)) {
P
Philipp Reisner 已提交
4245
				dev_err(DEV, "ReportCBitmap packet too large\n");
4246
				err = -EIO;
P
Philipp Reisner 已提交
4247 4248
				goto out;
			}
4249
			if (pi->size <= sizeof(*p)) {
4250
				dev_err(DEV, "ReportCBitmap packet too small (l:%u)\n", pi->size);
4251
				err = -EIO;
4252
				goto out;
P
Philipp Reisner 已提交
4253
			}
4254 4255 4256
			err = drbd_recv_all(mdev->tconn, p, pi->size);
			if (err)
			       goto out;
4257
			err = decode_bitmap_c(mdev, p, &c, pi->size);
P
Philipp Reisner 已提交
4258
		} else {
4259
			dev_warn(DEV, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", pi->cmd);
4260
			err = -EIO;
P
Philipp Reisner 已提交
4261 4262 4263
			goto out;
		}

4264
		c.packets[pi->cmd == P_BITMAP]++;
4265
		c.bytes[pi->cmd == P_BITMAP] += drbd_header_size(tconn) + pi->size;
P
Philipp Reisner 已提交
4266

4267 4268 4269
		if (err <= 0) {
			if (err < 0)
				goto out;
P
Philipp Reisner 已提交
4270
			break;
4271
		}
4272
		err = drbd_recv_header(mdev->tconn, pi);
4273
		if (err)
P
Philipp Reisner 已提交
4274
			goto out;
4275
	}
P
Philipp Reisner 已提交
4276 4277 4278 4279

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

	if (mdev->state.conn == C_WF_BITMAP_T) {
4280 4281
		enum drbd_state_rv rv;

4282 4283
		err = drbd_send_bitmap(mdev);
		if (err)
P
Philipp Reisner 已提交
4284 4285
			goto out;
		/* Omit CS_ORDERED with this state transition to avoid deadlocks. */
4286 4287
		rv = _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE);
		D_ASSERT(rv == SS_SUCCESS);
P
Philipp Reisner 已提交
4288 4289 4290 4291 4292 4293
	} 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));
	}
4294
	err = 0;
P
Philipp Reisner 已提交
4295 4296

 out:
4297
	drbd_bm_unlock(mdev);
4298
	if (!err && mdev->state.conn == C_WF_BITMAP_S)
P
Philipp Reisner 已提交
4299
		drbd_start_resync(mdev, C_SYNC_SOURCE);
4300
	return err;
P
Philipp Reisner 已提交
4301 4302
}

4303
static int receive_skip(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4304
{
4305
	conn_warn(tconn, "skipping unknown optional packet type %d, l: %d!\n",
4306
		 pi->cmd, pi->size);
4307

4308
	return ignore_remaining_packet(tconn, pi);
4309 4310
}

4311
static int receive_UnplugRemote(struct drbd_tconn *tconn, struct packet_info *pi)
4312
{
4313 4314
	/* Make sure we've acked all the TCP data associated
	 * with the data requests being unplugged */
4315
	drbd_tcp_quickack(tconn->data.socket);
4316

4317
	return 0;
4318 4319
}

4320
static int receive_out_of_sync(struct drbd_tconn *tconn, struct packet_info *pi)
4321
{
4322
	struct drbd_conf *mdev;
4323
	struct p_block_desc *p = pi->data;
4324 4325 4326 4327

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
		return -EIO;
4328

4329 4330 4331 4332 4333 4334 4335 4336 4337 4338
	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));
	}

4339 4340
	drbd_set_out_of_sync(mdev, be64_to_cpu(p->sector), be32_to_cpu(p->blksize));

4341
	return 0;
4342 4343
}

4344 4345 4346
struct data_cmd {
	int expect_payload;
	size_t pkt_size;
4347
	int (*fn)(struct drbd_tconn *, struct packet_info *);
4348 4349 4350
};

static struct data_cmd drbd_cmd_handler[] = {
4351 4352 4353 4354
	[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 } ,
4355 4356 4357
	[P_BITMAP]	    = { 1, 0, receive_bitmap } ,
	[P_COMPRESSED_BITMAP] = { 1, 0, receive_bitmap } ,
	[P_UNPLUG_REMOTE]   = { 0, 0, receive_UnplugRemote },
4358 4359
	[P_DATA_REQUEST]    = { 0, sizeof(struct p_block_req), receive_DataRequest },
	[P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest },
4360 4361
	[P_SYNC_PARAM]	    = { 1, 0, receive_SyncParam },
	[P_SYNC_PARAM89]    = { 1, 0, receive_SyncParam },
4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373
	[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 },
4374
	[P_PROTOCOL_UPDATE] = { 1, sizeof(struct p_protocol), receive_protocol },
P
Philipp Reisner 已提交
4375 4376
};

4377
static void drbdd(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4378
{
4379
	struct packet_info pi;
4380
	size_t shs; /* sub header size */
4381
	int err;
P
Philipp Reisner 已提交
4382

4383
	while (get_t_state(&tconn->receiver) == RUNNING) {
4384 4385
		struct data_cmd *cmd;

4386
		drbd_thread_current_set_cpu(&tconn->receiver);
4387
		if (drbd_recv_header(tconn, &pi))
4388
			goto err_out;
P
Philipp Reisner 已提交
4389

4390
		cmd = &drbd_cmd_handler[pi.cmd];
4391
		if (unlikely(pi.cmd >= ARRAY_SIZE(drbd_cmd_handler) || !cmd->fn)) {
4392 4393
			conn_err(tconn, "Unexpected data packet %s (0x%04x)",
				 cmdname(pi.cmd), pi.cmd);
4394
			goto err_out;
4395
		}
P
Philipp Reisner 已提交
4396

4397 4398
		shs = cmd->pkt_size;
		if (pi.size > shs && !cmd->expect_payload) {
4399 4400
			conn_err(tconn, "No payload expected %s l:%d\n",
				 cmdname(pi.cmd), pi.size);
4401
			goto err_out;
P
Philipp Reisner 已提交
4402 4403
		}

4404
		if (shs) {
4405
			err = drbd_recv_all_warn(tconn, pi.data, shs);
4406
			if (err)
4407
				goto err_out;
4408
			pi.size -= shs;
4409 4410
		}

4411 4412
		err = cmd->fn(tconn, &pi);
		if (err) {
4413 4414
			conn_err(tconn, "error receiving %s, e: %d l: %d!\n",
				 cmdname(pi.cmd), err, pi.size);
4415
			goto err_out;
P
Philipp Reisner 已提交
4416 4417
		}
	}
4418
	return;
P
Philipp Reisner 已提交
4419

4420 4421
    err_out:
	conn_request_state(tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD);
P
Philipp Reisner 已提交
4422 4423
}

4424
void conn_flush_workqueue(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4425 4426 4427 4428
{
	struct drbd_wq_barrier barr;

	barr.w.cb = w_prev_work_done;
4429
	barr.w.tconn = tconn;
P
Philipp Reisner 已提交
4430
	init_completion(&barr.done);
4431
	drbd_queue_work(&tconn->sender_work, &barr.w);
P
Philipp Reisner 已提交
4432 4433 4434
	wait_for_completion(&barr.done);
}

4435
static void conn_disconnect(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4436
{
P
Philipp Reisner 已提交
4437
	struct drbd_conf *mdev;
4438
	enum drbd_conns oc;
P
Philipp Reisner 已提交
4439
	int vnr;
P
Philipp Reisner 已提交
4440

4441
	if (tconn->cstate == C_STANDALONE)
P
Philipp Reisner 已提交
4442 4443
		return;

4444 4445 4446 4447 4448 4449 4450
	/* We are about to start the cleanup after connection loss.
	 * Make sure drbd_make_request knows about that.
	 * Usually we should be in some network failure state already,
	 * but just in case we are not, we fix it up here.
	 */
	conn_request_state(tconn, NS(conn, C_NETWORK_FAILURE), CS_HARD);

P
Philipp Reisner 已提交
4451
	/* asender does not clean up anything. it must not interfere, either */
4452 4453 4454
	drbd_thread_stop(&tconn->asender);
	drbd_free_sock(tconn);

P
Philipp Reisner 已提交
4455 4456 4457 4458 4459 4460 4461 4462 4463 4464
	rcu_read_lock();
	idr_for_each_entry(&tconn->volumes, mdev, vnr) {
		kref_get(&mdev->kref);
		rcu_read_unlock();
		drbd_disconnected(mdev);
		kref_put(&mdev->kref, &drbd_minor_destroy);
		rcu_read_lock();
	}
	rcu_read_unlock();

4465 4466 4467 4468
	if (!list_empty(&tconn->current_epoch->list))
		conn_err(tconn, "ASSERTION FAILED: tconn->current_epoch->list not empty\n");
	/* ok, no more ee's on the fly, it is safe to reset the epoch_size */
	atomic_set(&tconn->current_epoch->epoch_size, 0);
4469
	tconn->send.seen_any_write_yet = false;
4470

4471 4472
	conn_info(tconn, "Connection closed\n");

4473 4474 4475
	if (conn_highest_role(tconn) == R_PRIMARY && conn_highest_pdsk(tconn) >= D_UNKNOWN)
		conn_try_outdate_peer_async(tconn);

4476
	spin_lock_irq(&tconn->req_lock);
4477 4478
	oc = tconn->cstate;
	if (oc >= C_UNCONNECTED)
P
Philipp Reisner 已提交
4479
		_conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE);
4480

4481 4482
	spin_unlock_irq(&tconn->req_lock);

4483
	if (oc == C_DISCONNECTING)
4484
		conn_request_state(tconn, NS(conn, C_STANDALONE), CS_VERBOSE | CS_HARD);
4485 4486
}

P
Philipp Reisner 已提交
4487
static int drbd_disconnected(struct drbd_conf *mdev)
4488 4489
{
	unsigned int i;
P
Philipp Reisner 已提交
4490

4491
	/* wait for current activity to cease. */
4492
	spin_lock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
4493 4494 4495
	_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);
4496
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519

	/* 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);

	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" */
4520
	drbd_flush_workqueue(mdev);
P
Philipp Reisner 已提交
4521

4522
	drbd_finish_peer_reqs(mdev);
P
Philipp Reisner 已提交
4523

4524 4525 4526 4527 4528
	/* This second workqueue flush is necessary, since drbd_finish_peer_reqs()
	   might have issued a work again. The one before drbd_finish_peer_reqs() is
	   necessary to reclain net_ee in drbd_finish_peer_reqs(). */
	drbd_flush_workqueue(mdev);

4529 4530 4531 4532
	/* need to do it again, drbd_finish_peer_reqs() may have populated it
	 * again via drbd_try_clear_on_disk_bm(). */
	drbd_rs_cancel_all(mdev);

P
Philipp Reisner 已提交
4533 4534 4535
	kfree(mdev->p_uuid);
	mdev->p_uuid = NULL;

4536
	if (!drbd_suspended(mdev))
4537
		tl_clear(mdev->tconn);
P
Philipp Reisner 已提交
4538 4539 4540

	drbd_md_sync(mdev);

4541 4542 4543 4544
	/* 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
4545 4546 4547 4548 4549 4550 4551
	/* 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.
	 */
4552
	i = drbd_free_peer_reqs(mdev, &mdev->net_ee);
P
Philipp Reisner 已提交
4553 4554
	if (i)
		dev_info(DEV, "net_ee not empty, killed %u entries\n", i);
4555 4556 4557
	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
4558 4559
	i = atomic_read(&mdev->pp_in_use);
	if (i)
4560
		dev_info(DEV, "pp_in_use = %d, expected 0\n", i);
P
Philipp Reisner 已提交
4561 4562 4563 4564 4565 4566

	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));

4567
	return 0;
P
Philipp Reisner 已提交
4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578
}

/*
 * 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.
 */
4579
static int drbd_send_features(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4580
{
4581 4582
	struct drbd_socket *sock;
	struct p_connection_features *p;
P
Philipp Reisner 已提交
4583

4584 4585 4586
	sock = &tconn->data;
	p = conn_prepare_command(tconn, sock);
	if (!p)
4587
		return -EIO;
P
Philipp Reisner 已提交
4588 4589 4590
	memset(p, 0, sizeof(*p));
	p->protocol_min = cpu_to_be32(PRO_VERSION_MIN);
	p->protocol_max = cpu_to_be32(PRO_VERSION_MAX);
4591
	return conn_send_command(tconn, sock, P_CONNECTION_FEATURES, sizeof(*p), NULL, 0);
P
Philipp Reisner 已提交
4592 4593 4594 4595 4596 4597 4598 4599 4600
}

/*
 * 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.
 */
4601
static int drbd_do_features(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4602
{
4603
	/* ASSERT current == tconn->receiver ... */
4604 4605
	struct p_connection_features *p;
	const int expect = sizeof(struct p_connection_features);
4606
	struct packet_info pi;
4607
	int err;
P
Philipp Reisner 已提交
4608

4609
	err = drbd_send_features(tconn);
4610
	if (err)
P
Philipp Reisner 已提交
4611 4612
		return 0;

4613 4614
	err = drbd_recv_header(tconn, &pi);
	if (err)
P
Philipp Reisner 已提交
4615 4616
		return 0;

4617 4618
	if (pi.cmd != P_CONNECTION_FEATURES) {
		conn_err(tconn, "expected ConnectionFeatures packet, received: %s (0x%04x)\n",
4619
			 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4620 4621 4622
		return -1;
	}

4623
	if (pi.size != expect) {
4624
		conn_err(tconn, "expected ConnectionFeatures length: %u, received: %u\n",
4625
		     expect, pi.size);
P
Philipp Reisner 已提交
4626 4627 4628
		return -1;
	}

4629 4630
	p = pi.data;
	err = drbd_recv_all_warn(tconn, p, expect);
4631
	if (err)
P
Philipp Reisner 已提交
4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642
		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;

4643
	tconn->agreed_pro_version = min_t(int, PRO_VERSION_MAX, p->protocol_max);
P
Philipp Reisner 已提交
4644

4645 4646
	conn_info(tconn, "Handshake successful: "
	     "Agreed network protocol version %d\n", tconn->agreed_pro_version);
P
Philipp Reisner 已提交
4647 4648 4649 4650

	return 1;

 incompat:
4651
	conn_err(tconn, "incompatible DRBD dialects: "
P
Philipp Reisner 已提交
4652 4653 4654 4655 4656 4657 4658
	    "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)
4659
static int drbd_do_auth(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4660 4661 4662
{
	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");
4663
	return -1;
P
Philipp Reisner 已提交
4664 4665 4666
}
#else
#define CHALLENGE_LEN 64
4667 4668 4669 4670 4671 4672 4673

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

4674
static int drbd_do_auth(struct drbd_tconn *tconn)
P
Philipp Reisner 已提交
4675
{
4676
	struct drbd_socket *sock;
P
Philipp Reisner 已提交
4677 4678 4679 4680 4681
	char my_challenge[CHALLENGE_LEN];  /* 64 Bytes... */
	struct scatterlist sg;
	char *response = NULL;
	char *right_response = NULL;
	char *peers_ch = NULL;
4682 4683
	unsigned int key_len;
	char secret[SHARED_SECRET_MAX]; /* 64 byte */
P
Philipp Reisner 已提交
4684 4685
	unsigned int resp_size;
	struct hash_desc desc;
4686
	struct packet_info pi;
4687
	struct net_conf *nc;
4688
	int err, rv;
P
Philipp Reisner 已提交
4689

4690 4691
	/* FIXME: Put the challenge/response into the preallocated socket buffer.  */

4692 4693 4694 4695 4696 4697
	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();

4698
	desc.tfm = tconn->cram_hmac_tfm;
P
Philipp Reisner 已提交
4699 4700
	desc.flags = 0;

4701
	rv = crypto_hash_setkey(tconn->cram_hmac_tfm, (u8 *)secret, key_len);
P
Philipp Reisner 已提交
4702
	if (rv) {
4703
		conn_err(tconn, "crypto_hash_setkey() failed with %d\n", rv);
4704
		rv = -1;
P
Philipp Reisner 已提交
4705 4706 4707 4708 4709
		goto fail;
	}

	get_random_bytes(my_challenge, CHALLENGE_LEN);

4710 4711 4712 4713 4714
	sock = &tconn->data;
	if (!conn_prepare_command(tconn, sock)) {
		rv = 0;
		goto fail;
	}
4715
	rv = !conn_send_command(tconn, sock, P_AUTH_CHALLENGE, 0,
4716
				my_challenge, CHALLENGE_LEN);
P
Philipp Reisner 已提交
4717 4718 4719
	if (!rv)
		goto fail;

4720 4721 4722
	err = drbd_recv_header(tconn, &pi);
	if (err) {
		rv = 0;
P
Philipp Reisner 已提交
4723
		goto fail;
4724
	}
P
Philipp Reisner 已提交
4725

4726
	if (pi.cmd != P_AUTH_CHALLENGE) {
4727
		conn_err(tconn, "expected AuthChallenge packet, received: %s (0x%04x)\n",
4728
			 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4729 4730 4731 4732
		rv = 0;
		goto fail;
	}

4733
	if (pi.size > CHALLENGE_LEN * 2) {
4734
		conn_err(tconn, "expected AuthChallenge payload too big.\n");
4735
		rv = -1;
P
Philipp Reisner 已提交
4736 4737 4738
		goto fail;
	}

4739
	peers_ch = kmalloc(pi.size, GFP_NOIO);
P
Philipp Reisner 已提交
4740
	if (peers_ch == NULL) {
4741
		conn_err(tconn, "kmalloc of peers_ch failed\n");
4742
		rv = -1;
P
Philipp Reisner 已提交
4743 4744 4745
		goto fail;
	}

4746 4747
	err = drbd_recv_all_warn(tconn, peers_ch, pi.size);
	if (err) {
P
Philipp Reisner 已提交
4748 4749 4750 4751
		rv = 0;
		goto fail;
	}

4752
	resp_size = crypto_hash_digestsize(tconn->cram_hmac_tfm);
P
Philipp Reisner 已提交
4753 4754
	response = kmalloc(resp_size, GFP_NOIO);
	if (response == NULL) {
4755
		conn_err(tconn, "kmalloc of response failed\n");
4756
		rv = -1;
P
Philipp Reisner 已提交
4757 4758 4759 4760
		goto fail;
	}

	sg_init_table(&sg, 1);
4761
	sg_set_buf(&sg, peers_ch, pi.size);
P
Philipp Reisner 已提交
4762 4763 4764

	rv = crypto_hash_digest(&desc, &sg, sg.length, response);
	if (rv) {
4765
		conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv);
4766
		rv = -1;
P
Philipp Reisner 已提交
4767 4768 4769
		goto fail;
	}

4770 4771 4772 4773
	if (!conn_prepare_command(tconn, sock)) {
		rv = 0;
		goto fail;
	}
4774
	rv = !conn_send_command(tconn, sock, P_AUTH_RESPONSE, 0,
4775
				response, resp_size);
P
Philipp Reisner 已提交
4776 4777 4778
	if (!rv)
		goto fail;

4779 4780 4781
	err = drbd_recv_header(tconn, &pi);
	if (err) {
		rv = 0;
P
Philipp Reisner 已提交
4782
		goto fail;
4783
	}
P
Philipp Reisner 已提交
4784

4785
	if (pi.cmd != P_AUTH_RESPONSE) {
4786
		conn_err(tconn, "expected AuthResponse packet, received: %s (0x%04x)\n",
4787
			 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
4788 4789 4790 4791
		rv = 0;
		goto fail;
	}

4792
	if (pi.size != resp_size) {
4793
		conn_err(tconn, "expected AuthResponse payload of wrong size\n");
P
Philipp Reisner 已提交
4794 4795 4796 4797
		rv = 0;
		goto fail;
	}

4798 4799
	err = drbd_recv_all_warn(tconn, response , resp_size);
	if (err) {
P
Philipp Reisner 已提交
4800 4801 4802 4803 4804
		rv = 0;
		goto fail;
	}

	right_response = kmalloc(resp_size, GFP_NOIO);
4805
	if (right_response == NULL) {
4806
		conn_err(tconn, "kmalloc of right_response failed\n");
4807
		rv = -1;
P
Philipp Reisner 已提交
4808 4809 4810 4811 4812 4813 4814
		goto fail;
	}

	sg_set_buf(&sg, my_challenge, CHALLENGE_LEN);

	rv = crypto_hash_digest(&desc, &sg, sg.length, right_response);
	if (rv) {
4815
		conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv);
4816
		rv = -1;
P
Philipp Reisner 已提交
4817 4818 4819 4820 4821 4822
		goto fail;
	}

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

	if (rv)
4823 4824
		conn_info(tconn, "Peer authenticated using %d bytes HMAC\n",
		     resp_size);
4825 4826
	else
		rv = -1;
P
Philipp Reisner 已提交
4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838

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

	return rv;
}
#endif

int drbdd_init(struct drbd_thread *thi)
{
4839
	struct drbd_tconn *tconn = thi->tconn;
P
Philipp Reisner 已提交
4840 4841
	int h;

4842
	conn_info(tconn, "receiver (re)started\n");
P
Philipp Reisner 已提交
4843 4844

	do {
4845
		h = conn_connect(tconn);
P
Philipp Reisner 已提交
4846
		if (h == 0) {
4847
			conn_disconnect(tconn);
4848
			schedule_timeout_interruptible(HZ);
P
Philipp Reisner 已提交
4849 4850
		}
		if (h == -1) {
4851
			conn_warn(tconn, "Discarding network configuration.\n");
4852
			conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
4853 4854 4855
		}
	} while (h == 0);

4856 4857
	if (h > 0)
		drbdd(tconn);
P
Philipp Reisner 已提交
4858

4859
	conn_disconnect(tconn);
P
Philipp Reisner 已提交
4860

4861
	conn_info(tconn, "receiver terminated\n");
P
Philipp Reisner 已提交
4862 4863 4864 4865 4866
	return 0;
}

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

4867
static int got_conn_RqSReply(struct drbd_tconn *tconn, struct packet_info *pi)
4868
{
4869
	struct p_req_state_reply *p = pi->data;
4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880
	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);

4881
	return 0;
4882 4883
}

4884
static int got_RqSReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4885
{
4886
	struct drbd_conf *mdev;
4887
	struct p_req_state_reply *p = pi->data;
P
Philipp Reisner 已提交
4888 4889
	int retcode = be32_to_cpu(p->retcode);

4890 4891
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4892
		return -EIO;
4893

4894 4895 4896 4897 4898
	if (test_bit(CONN_WD_ST_CHG_REQ, &tconn->flags)) {
		D_ASSERT(tconn->agreed_pro_version < 100);
		return got_conn_RqSReply(tconn, pi);
	}

4899 4900 4901 4902 4903 4904
	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
4905
	}
4906 4907
	wake_up(&mdev->state_wait);

4908
	return 0;
P
Philipp Reisner 已提交
4909 4910
}

4911
static int got_Ping(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4912
{
4913
	return drbd_send_ping_ack(tconn);
P
Philipp Reisner 已提交
4914 4915 4916

}

4917
static int got_PingAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4918 4919
{
	/* restore idle timeout */
4920 4921 4922
	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
4923

4924
	return 0;
P
Philipp Reisner 已提交
4925 4926
}

4927
static int got_IsInSync(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4928
{
4929
	struct drbd_conf *mdev;
4930
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4931 4932 4933
	sector_t sector = be64_to_cpu(p->sector);
	int blksize = be32_to_cpu(p->blksize);

4934 4935
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4936
		return -EIO;
4937

4938
	D_ASSERT(mdev->tconn->agreed_pro_version >= 89);
P
Philipp Reisner 已提交
4939 4940 4941

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

4942 4943 4944 4945 4946 4947 4948
	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
4949
	dec_rs_pending(mdev);
4950
	atomic_add(blksize >> 9, &mdev->rs_sect_in);
P
Philipp Reisner 已提交
4951

4952
	return 0;
P
Philipp Reisner 已提交
4953 4954
}

4955 4956 4957 4958
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
4959 4960 4961 4962
{
	struct drbd_request *req;
	struct bio_and_error m;

4963
	spin_lock_irq(&mdev->tconn->req_lock);
4964
	req = find_request(mdev, root, id, sector, missing_ok, func);
P
Philipp Reisner 已提交
4965
	if (unlikely(!req)) {
4966
		spin_unlock_irq(&mdev->tconn->req_lock);
4967
		return -EIO;
P
Philipp Reisner 已提交
4968 4969
	}
	__req_mod(req, what, &m);
4970
	spin_unlock_irq(&mdev->tconn->req_lock);
P
Philipp Reisner 已提交
4971 4972 4973

	if (m.bio)
		complete_master_bio(mdev, &m);
4974
	return 0;
P
Philipp Reisner 已提交
4975 4976
}

4977
static int got_BlockAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
4978
{
4979
	struct drbd_conf *mdev;
4980
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
4981 4982 4983 4984
	sector_t sector = be64_to_cpu(p->sector);
	int blksize = be32_to_cpu(p->blksize);
	enum drbd_req_event what;

4985 4986
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
4987
		return -EIO;
4988

P
Philipp Reisner 已提交
4989 4990
	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

4991
	if (p->block_id == ID_SYNCER) {
P
Philipp Reisner 已提交
4992 4993
		drbd_set_in_sync(mdev, sector, blksize);
		dec_rs_pending(mdev);
4994
		return 0;
P
Philipp Reisner 已提交
4995
	}
4996
	switch (pi->cmd) {
P
Philipp Reisner 已提交
4997
	case P_RS_WRITE_ACK:
4998
		what = WRITE_ACKED_BY_PEER_AND_SIS;
P
Philipp Reisner 已提交
4999 5000
		break;
	case P_WRITE_ACK:
5001
		what = WRITE_ACKED_BY_PEER;
P
Philipp Reisner 已提交
5002 5003
		break;
	case P_RECV_ACK:
5004
		what = RECV_ACKED_BY_PEER;
P
Philipp Reisner 已提交
5005
		break;
5006 5007
	case P_SUPERSEDED:
		what = CONFLICT_RESOLVED;
5008 5009 5010
		break;
	case P_RETRY_WRITE:
		what = POSTPONE_WRITE;
P
Philipp Reisner 已提交
5011 5012
		break;
	default:
5013
		BUG();
P
Philipp Reisner 已提交
5014 5015
	}

5016 5017 5018
	return validate_req_change_req_state(mdev, p->block_id, sector,
					     &mdev->write_requests, __func__,
					     what, false);
P
Philipp Reisner 已提交
5019 5020
}

5021
static int got_NegAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
5022
{
5023
	struct drbd_conf *mdev;
5024
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
5025
	sector_t sector = be64_to_cpu(p->sector);
5026
	int size = be32_to_cpu(p->blksize);
5027
	int err;
P
Philipp Reisner 已提交
5028

5029 5030
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
5031
		return -EIO;
5032

P
Philipp Reisner 已提交
5033 5034
	update_peer_seq(mdev, be32_to_cpu(p->seq_num));

5035
	if (p->block_id == ID_SYNCER) {
P
Philipp Reisner 已提交
5036 5037
		dec_rs_pending(mdev);
		drbd_rs_failed_io(mdev, sector, size);
5038
		return 0;
P
Philipp Reisner 已提交
5039
	}
5040

5041 5042
	err = validate_req_change_req_state(mdev, p->block_id, sector,
					    &mdev->write_requests, __func__,
5043
					    NEG_ACKED, true);
5044
	if (err) {
5045 5046 5047 5048 5049 5050
		/* 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);
5051
	}
5052
	return 0;
P
Philipp Reisner 已提交
5053 5054
}

5055
static int got_NegDReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
5056
{
5057
	struct drbd_conf *mdev;
5058
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
5059 5060
	sector_t sector = be64_to_cpu(p->sector);

5061 5062
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
5063
		return -EIO;
5064

P
Philipp Reisner 已提交
5065
	update_peer_seq(mdev, be32_to_cpu(p->seq_num));
5066

5067
	dev_err(DEV, "Got NegDReply; Sector %llus, len %u.\n",
P
Philipp Reisner 已提交
5068 5069
	    (unsigned long long)sector, be32_to_cpu(p->blksize));

5070 5071 5072
	return validate_req_change_req_state(mdev, p->block_id, sector,
					     &mdev->read_requests, __func__,
					     NEG_ACKED, false);
P
Philipp Reisner 已提交
5073 5074
}

5075
static int got_NegRSDReply(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
5076
{
5077
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
5078 5079
	sector_t sector;
	int size;
5080
	struct p_block_ack *p = pi->data;
5081 5082 5083

	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
5084
		return -EIO;
P
Philipp Reisner 已提交
5085 5086 5087 5088 5089 5090 5091 5092 5093 5094

	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);
5095
		switch (pi->cmd) {
5096 5097 5098 5099 5100
		case P_NEG_RS_DREPLY:
			drbd_rs_failed_io(mdev, sector, size);
		case P_RS_CANCEL:
			break;
		default:
5101
			BUG();
5102
		}
P
Philipp Reisner 已提交
5103 5104 5105
		put_ldev(mdev);
	}

5106
	return 0;
P
Philipp Reisner 已提交
5107 5108
}

5109
static int got_BarrierAck(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
5110
{
5111
	struct p_barrier_ack *p = pi->data;
5112 5113
	struct drbd_conf *mdev;
	int vnr;
5114

5115
	tl_release(tconn, p->barrier, be32_to_cpu(p->set_size));
P
Philipp Reisner 已提交
5116

5117 5118 5119 5120 5121 5122 5123 5124
	rcu_read_lock();
	idr_for_each_entry(&tconn->volumes, mdev, vnr) {
		if (mdev->state.conn == C_AHEAD &&
		    atomic_read(&mdev->ap_in_flight) == 0 &&
		    !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &mdev->flags)) {
			mdev->start_resync_timer.expires = jiffies + HZ;
			add_timer(&mdev->start_resync_timer);
		}
5125
	}
5126
	rcu_read_unlock();
5127

5128
	return 0;
P
Philipp Reisner 已提交
5129 5130
}

5131
static int got_OVResult(struct drbd_tconn *tconn, struct packet_info *pi)
P
Philipp Reisner 已提交
5132
{
5133
	struct drbd_conf *mdev;
5134
	struct p_block_ack *p = pi->data;
P
Philipp Reisner 已提交
5135 5136 5137 5138
	struct drbd_work *w;
	sector_t sector;
	int size;

5139 5140
	mdev = vnr_to_mdev(tconn, pi->vnr);
	if (!mdev)
5141
		return -EIO;
5142

P
Philipp Reisner 已提交
5143 5144 5145 5146 5147 5148
	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)
5149
		drbd_ov_out_of_sync_found(mdev, sector, size);
P
Philipp Reisner 已提交
5150
	else
5151
		ov_out_of_sync_print(mdev);
P
Philipp Reisner 已提交
5152

5153
	if (!get_ldev(mdev))
5154
		return 0;
5155

P
Philipp Reisner 已提交
5156 5157 5158
	drbd_rs_complete_io(mdev, sector);
	dec_rs_pending(mdev);

5159 5160 5161 5162 5163 5164 5165
	--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
5166 5167 5168
		w = kmalloc(sizeof(*w), GFP_NOIO);
		if (w) {
			w->cb = w_ov_finished;
5169
			w->mdev = mdev;
5170
			drbd_queue_work(&mdev->tconn->sender_work, w);
P
Philipp Reisner 已提交
5171 5172
		} else {
			dev_err(DEV, "kmalloc(w) failed.");
5173
			ov_out_of_sync_print(mdev);
P
Philipp Reisner 已提交
5174 5175 5176
			drbd_resync_finished(mdev);
		}
	}
5177
	put_ldev(mdev);
5178
	return 0;
P
Philipp Reisner 已提交
5179 5180
}

5181
static int got_skip(struct drbd_tconn *tconn, struct packet_info *pi)
5182
{
5183
	return 0;
5184 5185
}

5186
static int tconn_finish_peer_reqs(struct drbd_tconn *tconn)
5187
{
5188
	struct drbd_conf *mdev;
P
Philipp Reisner 已提交
5189
	int vnr, not_empty = 0;
5190 5191 5192 5193

	do {
		clear_bit(SIGNAL_ASENDER, &tconn->flags);
		flush_signals(current);
P
Philipp Reisner 已提交
5194 5195 5196 5197 5198

		rcu_read_lock();
		idr_for_each_entry(&tconn->volumes, mdev, vnr) {
			kref_get(&mdev->kref);
			rcu_read_unlock();
5199
			if (drbd_finish_peer_reqs(mdev)) {
P
Philipp Reisner 已提交
5200 5201
				kref_put(&mdev->kref, &drbd_minor_destroy);
				return 1;
5202
			}
P
Philipp Reisner 已提交
5203 5204
			kref_put(&mdev->kref, &drbd_minor_destroy);
			rcu_read_lock();
5205
		}
5206
		set_bit(SIGNAL_ASENDER, &tconn->flags);
5207 5208

		spin_lock_irq(&tconn->req_lock);
P
Philipp Reisner 已提交
5209
		idr_for_each_entry(&tconn->volumes, mdev, vnr) {
5210 5211 5212 5213 5214
			not_empty = !list_empty(&mdev->done_ee);
			if (not_empty)
				break;
		}
		spin_unlock_irq(&tconn->req_lock);
P
Philipp Reisner 已提交
5215
		rcu_read_unlock();
5216 5217 5218 5219 5220
	} while (not_empty);

	return 0;
}

5221 5222
struct asender_cmd {
	size_t pkt_size;
5223
	int (*fn)(struct drbd_tconn *tconn, struct packet_info *);
5224 5225 5226
};

static struct asender_cmd asender_tbl[] = {
5227 5228
	[P_PING]	    = { 0, got_Ping },
	[P_PING_ACK]	    = { 0, got_PingAck },
5229 5230 5231
	[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 },
5232
	[P_SUPERSEDED]   = { sizeof(struct p_block_ack), got_BlockAck },
5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243
	[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 },
5244 5245
};

P
Philipp Reisner 已提交
5246 5247
int drbd_asender(struct drbd_thread *thi)
{
5248
	struct drbd_tconn *tconn = thi->tconn;
P
Philipp Reisner 已提交
5249
	struct asender_cmd *cmd = NULL;
5250
	struct packet_info pi;
5251
	int rv;
5252
	void *buf    = tconn->meta.rbuf;
P
Philipp Reisner 已提交
5253
	int received = 0;
5254 5255
	unsigned int header_size = drbd_header_size(tconn);
	int expect   = header_size;
5256 5257
	bool ping_timeout_active = false;
	struct net_conf *nc;
5258
	int ping_timeo, tcp_cork, ping_int;
P
Philipp Reisner 已提交
5259 5260 5261 5262

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

5263
	while (get_t_state(thi) == RUNNING) {
5264
		drbd_thread_current_set_cpu(thi);
5265 5266 5267 5268

		rcu_read_lock();
		nc = rcu_dereference(tconn->net_conf);
		ping_timeo = nc->ping_timeo;
5269
		tcp_cork = nc->tcp_cork;
5270 5271 5272
		ping_int = nc->ping_int;
		rcu_read_unlock();

5273
		if (test_and_clear_bit(SEND_PING, &tconn->flags)) {
5274
			if (drbd_send_ping(tconn)) {
5275
				conn_err(tconn, "drbd_send_ping has failed\n");
5276 5277
				goto reconnect;
			}
5278 5279
			tconn->meta.socket->sk->sk_rcvtimeo = ping_timeo * HZ / 10;
			ping_timeout_active = true;
P
Philipp Reisner 已提交
5280 5281
		}

5282 5283
		/* TODO: conditionally cork; it may hurt latency if we cork without
		   much to send */
5284
		if (tcp_cork)
5285
			drbd_tcp_cork(tconn->meta.socket);
5286 5287
		if (tconn_finish_peer_reqs(tconn)) {
			conn_err(tconn, "tconn_finish_peer_reqs() failed\n");
5288
			goto reconnect;
5289
		}
P
Philipp Reisner 已提交
5290
		/* but unconditionally uncork unless disabled */
5291
		if (tcp_cork)
5292
			drbd_tcp_uncork(tconn->meta.socket);
P
Philipp Reisner 已提交
5293 5294 5295 5296 5297

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

5298 5299
		rv = drbd_recv_short(tconn->meta.socket, buf, expect-received, 0);
		clear_bit(SIGNAL_ASENDER, &tconn->flags);
P
Philipp Reisner 已提交
5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316

		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) {
5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328
			if (test_bit(DISCONNECT_SENT, &tconn->flags)) {
				long t;
				rcu_read_lock();
				t = rcu_dereference(tconn->net_conf)->ping_timeo * HZ/10;
				rcu_read_unlock();

				t = wait_event_timeout(tconn->ping_wait,
						       tconn->cstate < C_WF_REPORT_PARAMS,
						       t);
				if (t)
					break;
			}
5329
			conn_err(tconn, "meta connection shut down by peer.\n");
P
Philipp Reisner 已提交
5330 5331
			goto reconnect;
		} else if (rv == -EAGAIN) {
5332 5333
			/* If the data socket received something meanwhile,
			 * that is good enough: peer is still alive. */
5334 5335
			if (time_after(tconn->last_received,
				jiffies - tconn->meta.socket->sk->sk_rcvtimeo))
5336
				continue;
5337
			if (ping_timeout_active) {
5338
				conn_err(tconn, "PingAck did not arrive in time.\n");
P
Philipp Reisner 已提交
5339 5340
				goto reconnect;
			}
5341
			set_bit(SEND_PING, &tconn->flags);
P
Philipp Reisner 已提交
5342 5343 5344 5345
			continue;
		} else if (rv == -EINTR) {
			continue;
		} else {
5346
			conn_err(tconn, "sock_recvmsg returned %d\n", rv);
P
Philipp Reisner 已提交
5347 5348 5349 5350
			goto reconnect;
		}

		if (received == expect && cmd == NULL) {
5351
			if (decode_header(tconn, tconn->meta.rbuf, &pi))
P
Philipp Reisner 已提交
5352
				goto reconnect;
5353
			cmd = &asender_tbl[pi.cmd];
5354
			if (pi.cmd >= ARRAY_SIZE(asender_tbl) || !cmd->fn) {
5355 5356
				conn_err(tconn, "Unexpected meta packet %s (0x%04x)\n",
					 cmdname(pi.cmd), pi.cmd);
P
Philipp Reisner 已提交
5357 5358
				goto disconnect;
			}
5359
			expect = header_size + cmd->pkt_size;
5360
			if (pi.size != expect - header_size) {
5361
				conn_err(tconn, "Wrong packet size on meta (c: %d, l: %d)\n",
5362
					pi.cmd, pi.size);
P
Philipp Reisner 已提交
5363
				goto reconnect;
5364
			}
P
Philipp Reisner 已提交
5365 5366
		}
		if (received == expect) {
5367
			bool err;
5368

5369 5370
			err = cmd->fn(tconn, &pi);
			if (err) {
5371
				conn_err(tconn, "%pf failed\n", cmd->fn);
P
Philipp Reisner 已提交
5372
				goto reconnect;
5373
			}
P
Philipp Reisner 已提交
5374

5375 5376
			tconn->last_received = jiffies;

5377 5378 5379 5380 5381
			if (cmd == &asender_tbl[P_PING_ACK]) {
				/* restore idle timeout */
				tconn->meta.socket->sk->sk_rcvtimeo = ping_int * HZ;
				ping_timeout_active = false;
			}
5382

5383
			buf	 = tconn->meta.rbuf;
P
Philipp Reisner 已提交
5384
			received = 0;
5385
			expect	 = header_size;
P
Philipp Reisner 已提交
5386 5387 5388 5389 5390 5391
			cmd	 = NULL;
		}
	}

	if (0) {
reconnect:
5392
		conn_request_state(tconn, NS(conn, C_NETWORK_FAILURE), CS_HARD);
5393
		conn_md_sync(tconn);
P
Philipp Reisner 已提交
5394 5395 5396
	}
	if (0) {
disconnect:
5397
		conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
P
Philipp Reisner 已提交
5398
	}
5399
	clear_bit(SIGNAL_ASENDER, &tconn->flags);
P
Philipp Reisner 已提交
5400

5401
	conn_info(tconn, "asender terminated\n");
P
Philipp Reisner 已提交
5402 5403 5404

	return 0;
}