Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_receiver.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | */ | |
24 | ||
25 | ||
b411b363 PR |
26 | #include <linux/module.h> |
27 | ||
28 | #include <asm/uaccess.h> | |
29 | #include <net/sock.h> | |
30 | ||
b411b363 PR |
31 | #include <linux/drbd.h> |
32 | #include <linux/fs.h> | |
33 | #include <linux/file.h> | |
34 | #include <linux/in.h> | |
35 | #include <linux/mm.h> | |
36 | #include <linux/memcontrol.h> | |
37 | #include <linux/mm_inline.h> | |
38 | #include <linux/slab.h> | |
b411b363 PR |
39 | #include <linux/pkt_sched.h> |
40 | #define __KERNEL_SYSCALLS__ | |
41 | #include <linux/unistd.h> | |
42 | #include <linux/vmalloc.h> | |
43 | #include <linux/random.h> | |
b411b363 PR |
44 | #include <linux/string.h> |
45 | #include <linux/scatterlist.h> | |
46 | #include "drbd_int.h" | |
b411b363 PR |
47 | #include "drbd_req.h" |
48 | ||
49 | #include "drbd_vli.h" | |
50 | ||
77351055 PR |
51 | struct packet_info { |
52 | enum drbd_packet cmd; | |
53 | int size; | |
54 | int vnr; | |
55 | }; | |
56 | ||
b411b363 PR |
57 | enum finish_epoch { |
58 | FE_STILL_LIVE, | |
59 | FE_DESTROYED, | |
60 | FE_RECYCLED, | |
61 | }; | |
62 | ||
a4fbda8e PR |
63 | enum mdev_or_conn { |
64 | MDEV, | |
65 | CONN, | |
66 | }; | |
67 | ||
65d11ed6 | 68 | static int drbd_do_handshake(struct drbd_tconn *tconn); |
13e6037d | 69 | static int drbd_do_auth(struct drbd_tconn *tconn); |
360cc740 | 70 | static int drbd_disconnected(int vnr, void *p, void *data); |
b411b363 PR |
71 | |
72 | static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *, struct drbd_epoch *, enum epoch_event); | |
00d56944 | 73 | static int e_end_block(struct drbd_work *, int); |
b411b363 | 74 | |
b411b363 PR |
75 | |
76 | #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN) | |
77 | ||
45bb912b LE |
78 | /* |
79 | * some helper functions to deal with single linked page lists, | |
80 | * page->private being our "next" pointer. | |
81 | */ | |
82 | ||
83 | /* If at least n pages are linked at head, get n pages off. | |
84 | * Otherwise, don't modify head, and return NULL. | |
85 | * Locking is the responsibility of the caller. | |
86 | */ | |
87 | static struct page *page_chain_del(struct page **head, int n) | |
88 | { | |
89 | struct page *page; | |
90 | struct page *tmp; | |
91 | ||
92 | BUG_ON(!n); | |
93 | BUG_ON(!head); | |
94 | ||
95 | page = *head; | |
23ce4227 PR |
96 | |
97 | if (!page) | |
98 | return NULL; | |
99 | ||
45bb912b LE |
100 | while (page) { |
101 | tmp = page_chain_next(page); | |
102 | if (--n == 0) | |
103 | break; /* found sufficient pages */ | |
104 | if (tmp == NULL) | |
105 | /* insufficient pages, don't use any of them. */ | |
106 | return NULL; | |
107 | page = tmp; | |
108 | } | |
109 | ||
110 | /* add end of list marker for the returned list */ | |
111 | set_page_private(page, 0); | |
112 | /* actual return value, and adjustment of head */ | |
113 | page = *head; | |
114 | *head = tmp; | |
115 | return page; | |
116 | } | |
117 | ||
118 | /* may be used outside of locks to find the tail of a (usually short) | |
119 | * "private" page chain, before adding it back to a global chain head | |
120 | * with page_chain_add() under a spinlock. */ | |
121 | static struct page *page_chain_tail(struct page *page, int *len) | |
122 | { | |
123 | struct page *tmp; | |
124 | int i = 1; | |
125 | while ((tmp = page_chain_next(page))) | |
126 | ++i, page = tmp; | |
127 | if (len) | |
128 | *len = i; | |
129 | return page; | |
130 | } | |
131 | ||
132 | static int page_chain_free(struct page *page) | |
133 | { | |
134 | struct page *tmp; | |
135 | int i = 0; | |
136 | page_chain_for_each_safe(page, tmp) { | |
137 | put_page(page); | |
138 | ++i; | |
139 | } | |
140 | return i; | |
141 | } | |
142 | ||
143 | static void page_chain_add(struct page **head, | |
144 | struct page *chain_first, struct page *chain_last) | |
145 | { | |
146 | #if 1 | |
147 | struct page *tmp; | |
148 | tmp = page_chain_tail(chain_first, NULL); | |
149 | BUG_ON(tmp != chain_last); | |
150 | #endif | |
151 | ||
152 | /* add chain to head */ | |
153 | set_page_private(chain_last, (unsigned long)*head); | |
154 | *head = chain_first; | |
155 | } | |
156 | ||
157 | static struct page *drbd_pp_first_pages_or_try_alloc(struct drbd_conf *mdev, int number) | |
b411b363 PR |
158 | { |
159 | struct page *page = NULL; | |
45bb912b LE |
160 | struct page *tmp = NULL; |
161 | int i = 0; | |
b411b363 PR |
162 | |
163 | /* Yes, testing drbd_pp_vacant outside the lock is racy. | |
164 | * So what. It saves a spin_lock. */ | |
45bb912b | 165 | if (drbd_pp_vacant >= number) { |
b411b363 | 166 | spin_lock(&drbd_pp_lock); |
45bb912b LE |
167 | page = page_chain_del(&drbd_pp_pool, number); |
168 | if (page) | |
169 | drbd_pp_vacant -= number; | |
b411b363 | 170 | spin_unlock(&drbd_pp_lock); |
45bb912b LE |
171 | if (page) |
172 | return page; | |
b411b363 | 173 | } |
45bb912b | 174 | |
b411b363 PR |
175 | /* GFP_TRY, because we must not cause arbitrary write-out: in a DRBD |
176 | * "criss-cross" setup, that might cause write-out on some other DRBD, | |
177 | * which in turn might block on the other node at this very place. */ | |
45bb912b LE |
178 | for (i = 0; i < number; i++) { |
179 | tmp = alloc_page(GFP_TRY); | |
180 | if (!tmp) | |
181 | break; | |
182 | set_page_private(tmp, (unsigned long)page); | |
183 | page = tmp; | |
184 | } | |
185 | ||
186 | if (i == number) | |
187 | return page; | |
188 | ||
189 | /* Not enough pages immediately available this time. | |
190 | * No need to jump around here, drbd_pp_alloc will retry this | |
191 | * function "soon". */ | |
192 | if (page) { | |
193 | tmp = page_chain_tail(page, NULL); | |
194 | spin_lock(&drbd_pp_lock); | |
195 | page_chain_add(&drbd_pp_pool, page, tmp); | |
196 | drbd_pp_vacant += i; | |
197 | spin_unlock(&drbd_pp_lock); | |
198 | } | |
199 | return NULL; | |
b411b363 PR |
200 | } |
201 | ||
b411b363 PR |
202 | static void reclaim_net_ee(struct drbd_conf *mdev, struct list_head *to_be_freed) |
203 | { | |
db830c46 | 204 | struct drbd_peer_request *peer_req; |
b411b363 PR |
205 | struct list_head *le, *tle; |
206 | ||
207 | /* The EEs are always appended to the end of the list. Since | |
208 | they are sent in order over the wire, they have to finish | |
209 | in order. As soon as we see the first not finished we can | |
210 | stop to examine the list... */ | |
211 | ||
212 | list_for_each_safe(le, tle, &mdev->net_ee) { | |
db830c46 AG |
213 | peer_req = list_entry(le, struct drbd_peer_request, w.list); |
214 | if (drbd_ee_has_active_page(peer_req)) | |
b411b363 PR |
215 | break; |
216 | list_move(le, to_be_freed); | |
217 | } | |
218 | } | |
219 | ||
220 | static void drbd_kick_lo_and_reclaim_net(struct drbd_conf *mdev) | |
221 | { | |
222 | LIST_HEAD(reclaimed); | |
db830c46 | 223 | struct drbd_peer_request *peer_req, *t; |
b411b363 | 224 | |
87eeee41 | 225 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 226 | reclaim_net_ee(mdev, &reclaimed); |
87eeee41 | 227 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 228 | |
db830c46 AG |
229 | list_for_each_entry_safe(peer_req, t, &reclaimed, w.list) |
230 | drbd_free_net_ee(mdev, peer_req); | |
b411b363 PR |
231 | } |
232 | ||
233 | /** | |
45bb912b | 234 | * drbd_pp_alloc() - Returns @number pages, retries forever (or until signalled) |
b411b363 | 235 | * @mdev: DRBD device. |
45bb912b LE |
236 | * @number: number of pages requested |
237 | * @retry: whether to retry, if not enough pages are available right now | |
238 | * | |
239 | * Tries to allocate number pages, first from our own page pool, then from | |
240 | * the kernel, unless this allocation would exceed the max_buffers setting. | |
241 | * Possibly retry until DRBD frees sufficient pages somewhere else. | |
b411b363 | 242 | * |
45bb912b | 243 | * Returns a page chain linked via page->private. |
b411b363 | 244 | */ |
45bb912b | 245 | static struct page *drbd_pp_alloc(struct drbd_conf *mdev, unsigned number, bool retry) |
b411b363 PR |
246 | { |
247 | struct page *page = NULL; | |
248 | DEFINE_WAIT(wait); | |
249 | ||
45bb912b LE |
250 | /* Yes, we may run up to @number over max_buffers. If we |
251 | * follow it strictly, the admin will get it wrong anyways. */ | |
89e58e75 | 252 | if (atomic_read(&mdev->pp_in_use) < mdev->tconn->net_conf->max_buffers) |
45bb912b | 253 | page = drbd_pp_first_pages_or_try_alloc(mdev, number); |
b411b363 | 254 | |
45bb912b | 255 | while (page == NULL) { |
b411b363 PR |
256 | prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE); |
257 | ||
258 | drbd_kick_lo_and_reclaim_net(mdev); | |
259 | ||
89e58e75 | 260 | if (atomic_read(&mdev->pp_in_use) < mdev->tconn->net_conf->max_buffers) { |
45bb912b | 261 | page = drbd_pp_first_pages_or_try_alloc(mdev, number); |
b411b363 PR |
262 | if (page) |
263 | break; | |
264 | } | |
265 | ||
266 | if (!retry) | |
267 | break; | |
268 | ||
269 | if (signal_pending(current)) { | |
270 | dev_warn(DEV, "drbd_pp_alloc interrupted!\n"); | |
271 | break; | |
272 | } | |
273 | ||
274 | schedule(); | |
275 | } | |
276 | finish_wait(&drbd_pp_wait, &wait); | |
277 | ||
45bb912b LE |
278 | if (page) |
279 | atomic_add(number, &mdev->pp_in_use); | |
b411b363 PR |
280 | return page; |
281 | } | |
282 | ||
283 | /* Must not be used from irq, as that may deadlock: see drbd_pp_alloc. | |
87eeee41 | 284 | * Is also used from inside an other spin_lock_irq(&mdev->tconn->req_lock); |
45bb912b LE |
285 | * Either links the page chain back to the global pool, |
286 | * or returns all pages to the system. */ | |
435f0740 | 287 | static void drbd_pp_free(struct drbd_conf *mdev, struct page *page, int is_net) |
b411b363 | 288 | { |
435f0740 | 289 | atomic_t *a = is_net ? &mdev->pp_in_use_by_net : &mdev->pp_in_use; |
b411b363 | 290 | int i; |
435f0740 | 291 | |
81a5d60e | 292 | if (drbd_pp_vacant > (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count) |
45bb912b LE |
293 | i = page_chain_free(page); |
294 | else { | |
295 | struct page *tmp; | |
296 | tmp = page_chain_tail(page, &i); | |
297 | spin_lock(&drbd_pp_lock); | |
298 | page_chain_add(&drbd_pp_pool, page, tmp); | |
299 | drbd_pp_vacant += i; | |
300 | spin_unlock(&drbd_pp_lock); | |
b411b363 | 301 | } |
435f0740 | 302 | i = atomic_sub_return(i, a); |
45bb912b | 303 | if (i < 0) |
435f0740 LE |
304 | dev_warn(DEV, "ASSERTION FAILED: %s: %d < 0\n", |
305 | is_net ? "pp_in_use_by_net" : "pp_in_use", i); | |
b411b363 PR |
306 | wake_up(&drbd_pp_wait); |
307 | } | |
308 | ||
309 | /* | |
310 | You need to hold the req_lock: | |
311 | _drbd_wait_ee_list_empty() | |
312 | ||
313 | You must not have the req_lock: | |
314 | drbd_free_ee() | |
315 | drbd_alloc_ee() | |
316 | drbd_init_ee() | |
317 | drbd_release_ee() | |
318 | drbd_ee_fix_bhs() | |
319 | drbd_process_done_ee() | |
320 | drbd_clear_done_ee() | |
321 | drbd_wait_ee_list_empty() | |
322 | */ | |
323 | ||
f6ffca9f AG |
324 | struct drbd_peer_request * |
325 | drbd_alloc_ee(struct drbd_conf *mdev, u64 id, sector_t sector, | |
326 | unsigned int data_size, gfp_t gfp_mask) __must_hold(local) | |
b411b363 | 327 | { |
db830c46 | 328 | struct drbd_peer_request *peer_req; |
b411b363 | 329 | struct page *page; |
45bb912b | 330 | unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT; |
b411b363 | 331 | |
0cf9d27e | 332 | if (drbd_insert_fault(mdev, DRBD_FAULT_AL_EE)) |
b411b363 PR |
333 | return NULL; |
334 | ||
db830c46 AG |
335 | peer_req = mempool_alloc(drbd_ee_mempool, gfp_mask & ~__GFP_HIGHMEM); |
336 | if (!peer_req) { | |
b411b363 PR |
337 | if (!(gfp_mask & __GFP_NOWARN)) |
338 | dev_err(DEV, "alloc_ee: Allocation of an EE failed\n"); | |
339 | return NULL; | |
340 | } | |
341 | ||
45bb912b LE |
342 | page = drbd_pp_alloc(mdev, nr_pages, (gfp_mask & __GFP_WAIT)); |
343 | if (!page) | |
344 | goto fail; | |
b411b363 | 345 | |
db830c46 AG |
346 | drbd_clear_interval(&peer_req->i); |
347 | peer_req->i.size = data_size; | |
348 | peer_req->i.sector = sector; | |
349 | peer_req->i.local = false; | |
350 | peer_req->i.waiting = false; | |
351 | ||
352 | peer_req->epoch = NULL; | |
a21e9298 | 353 | peer_req->w.mdev = mdev; |
db830c46 AG |
354 | peer_req->pages = page; |
355 | atomic_set(&peer_req->pending_bios, 0); | |
356 | peer_req->flags = 0; | |
9a8e7753 AG |
357 | /* |
358 | * The block_id is opaque to the receiver. It is not endianness | |
359 | * converted, and sent back to the sender unchanged. | |
360 | */ | |
db830c46 | 361 | peer_req->block_id = id; |
b411b363 | 362 | |
db830c46 | 363 | return peer_req; |
b411b363 | 364 | |
45bb912b | 365 | fail: |
db830c46 | 366 | mempool_free(peer_req, drbd_ee_mempool); |
b411b363 PR |
367 | return NULL; |
368 | } | |
369 | ||
db830c46 | 370 | void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_peer_request *peer_req, |
f6ffca9f | 371 | int is_net) |
b411b363 | 372 | { |
db830c46 AG |
373 | if (peer_req->flags & EE_HAS_DIGEST) |
374 | kfree(peer_req->digest); | |
375 | drbd_pp_free(mdev, peer_req->pages, is_net); | |
376 | D_ASSERT(atomic_read(&peer_req->pending_bios) == 0); | |
377 | D_ASSERT(drbd_interval_empty(&peer_req->i)); | |
378 | mempool_free(peer_req, drbd_ee_mempool); | |
b411b363 PR |
379 | } |
380 | ||
381 | int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list) | |
382 | { | |
383 | LIST_HEAD(work_list); | |
db830c46 | 384 | struct drbd_peer_request *peer_req, *t; |
b411b363 | 385 | int count = 0; |
435f0740 | 386 | int is_net = list == &mdev->net_ee; |
b411b363 | 387 | |
87eeee41 | 388 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 389 | list_splice_init(list, &work_list); |
87eeee41 | 390 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 391 | |
db830c46 AG |
392 | list_for_each_entry_safe(peer_req, t, &work_list, w.list) { |
393 | drbd_free_some_ee(mdev, peer_req, is_net); | |
b411b363 PR |
394 | count++; |
395 | } | |
396 | return count; | |
397 | } | |
398 | ||
399 | ||
32862ec7 | 400 | /* See also comments in _req_mod(,BARRIER_ACKED) |
b411b363 PR |
401 | * and receive_Barrier. |
402 | * | |
403 | * Move entries from net_ee to done_ee, if ready. | |
404 | * Grab done_ee, call all callbacks, free the entries. | |
405 | * The callbacks typically send out ACKs. | |
406 | */ | |
407 | static int drbd_process_done_ee(struct drbd_conf *mdev) | |
408 | { | |
409 | LIST_HEAD(work_list); | |
410 | LIST_HEAD(reclaimed); | |
db830c46 | 411 | struct drbd_peer_request *peer_req, *t; |
082a3439 | 412 | int ok = 1; |
b411b363 | 413 | |
87eeee41 | 414 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
415 | reclaim_net_ee(mdev, &reclaimed); |
416 | list_splice_init(&mdev->done_ee, &work_list); | |
87eeee41 | 417 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 418 | |
db830c46 AG |
419 | list_for_each_entry_safe(peer_req, t, &reclaimed, w.list) |
420 | drbd_free_net_ee(mdev, peer_req); | |
b411b363 PR |
421 | |
422 | /* possible callbacks here: | |
7be8da07 | 423 | * e_end_block, and e_end_resync_block, e_send_discard_write. |
b411b363 PR |
424 | * all ignore the last argument. |
425 | */ | |
db830c46 | 426 | list_for_each_entry_safe(peer_req, t, &work_list, w.list) { |
b411b363 | 427 | /* list_del not necessary, next/prev members not touched */ |
00d56944 | 428 | ok = peer_req->w.cb(&peer_req->w, !ok) && ok; |
db830c46 | 429 | drbd_free_ee(mdev, peer_req); |
b411b363 PR |
430 | } |
431 | wake_up(&mdev->ee_wait); | |
432 | ||
433 | return ok; | |
434 | } | |
435 | ||
436 | void _drbd_wait_ee_list_empty(struct drbd_conf *mdev, struct list_head *head) | |
437 | { | |
438 | DEFINE_WAIT(wait); | |
439 | ||
440 | /* avoids spin_lock/unlock | |
441 | * and calling prepare_to_wait in the fast path */ | |
442 | while (!list_empty(head)) { | |
443 | prepare_to_wait(&mdev->ee_wait, &wait, TASK_UNINTERRUPTIBLE); | |
87eeee41 | 444 | spin_unlock_irq(&mdev->tconn->req_lock); |
7eaceacc | 445 | io_schedule(); |
b411b363 | 446 | finish_wait(&mdev->ee_wait, &wait); |
87eeee41 | 447 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
448 | } |
449 | } | |
450 | ||
451 | void drbd_wait_ee_list_empty(struct drbd_conf *mdev, struct list_head *head) | |
452 | { | |
87eeee41 | 453 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 454 | _drbd_wait_ee_list_empty(mdev, head); |
87eeee41 | 455 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
456 | } |
457 | ||
458 | /* see also kernel_accept; which is only present since 2.6.18. | |
459 | * also we want to log which part of it failed, exactly */ | |
7653620d | 460 | static int drbd_accept(const char **what, struct socket *sock, struct socket **newsock) |
b411b363 PR |
461 | { |
462 | struct sock *sk = sock->sk; | |
463 | int err = 0; | |
464 | ||
465 | *what = "listen"; | |
466 | err = sock->ops->listen(sock, 5); | |
467 | if (err < 0) | |
468 | goto out; | |
469 | ||
470 | *what = "sock_create_lite"; | |
471 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
472 | newsock); | |
473 | if (err < 0) | |
474 | goto out; | |
475 | ||
476 | *what = "accept"; | |
477 | err = sock->ops->accept(sock, *newsock, 0); | |
478 | if (err < 0) { | |
479 | sock_release(*newsock); | |
480 | *newsock = NULL; | |
481 | goto out; | |
482 | } | |
483 | (*newsock)->ops = sock->ops; | |
484 | ||
485 | out: | |
486 | return err; | |
487 | } | |
488 | ||
dbd9eea0 | 489 | static int drbd_recv_short(struct socket *sock, void *buf, size_t size, int flags) |
b411b363 PR |
490 | { |
491 | mm_segment_t oldfs; | |
492 | struct kvec iov = { | |
493 | .iov_base = buf, | |
494 | .iov_len = size, | |
495 | }; | |
496 | struct msghdr msg = { | |
497 | .msg_iovlen = 1, | |
498 | .msg_iov = (struct iovec *)&iov, | |
499 | .msg_flags = (flags ? flags : MSG_WAITALL | MSG_NOSIGNAL) | |
500 | }; | |
501 | int rv; | |
502 | ||
503 | oldfs = get_fs(); | |
504 | set_fs(KERNEL_DS); | |
505 | rv = sock_recvmsg(sock, &msg, size, msg.msg_flags); | |
506 | set_fs(oldfs); | |
507 | ||
508 | return rv; | |
509 | } | |
510 | ||
de0ff338 | 511 | static int drbd_recv(struct drbd_tconn *tconn, void *buf, size_t size) |
b411b363 PR |
512 | { |
513 | mm_segment_t oldfs; | |
514 | struct kvec iov = { | |
515 | .iov_base = buf, | |
516 | .iov_len = size, | |
517 | }; | |
518 | struct msghdr msg = { | |
519 | .msg_iovlen = 1, | |
520 | .msg_iov = (struct iovec *)&iov, | |
521 | .msg_flags = MSG_WAITALL | MSG_NOSIGNAL | |
522 | }; | |
523 | int rv; | |
524 | ||
525 | oldfs = get_fs(); | |
526 | set_fs(KERNEL_DS); | |
527 | ||
528 | for (;;) { | |
de0ff338 | 529 | rv = sock_recvmsg(tconn->data.socket, &msg, size, msg.msg_flags); |
b411b363 PR |
530 | if (rv == size) |
531 | break; | |
532 | ||
533 | /* Note: | |
534 | * ECONNRESET other side closed the connection | |
535 | * ERESTARTSYS (on sock) we got a signal | |
536 | */ | |
537 | ||
538 | if (rv < 0) { | |
539 | if (rv == -ECONNRESET) | |
de0ff338 | 540 | conn_info(tconn, "sock was reset by peer\n"); |
b411b363 | 541 | else if (rv != -ERESTARTSYS) |
de0ff338 | 542 | conn_err(tconn, "sock_recvmsg returned %d\n", rv); |
b411b363 PR |
543 | break; |
544 | } else if (rv == 0) { | |
de0ff338 | 545 | conn_info(tconn, "sock was shut down by peer\n"); |
b411b363 PR |
546 | break; |
547 | } else { | |
548 | /* signal came in, or peer/link went down, | |
549 | * after we read a partial message | |
550 | */ | |
551 | /* D_ASSERT(signal_pending(current)); */ | |
552 | break; | |
553 | } | |
554 | }; | |
555 | ||
556 | set_fs(oldfs); | |
557 | ||
558 | if (rv != size) | |
bbeb641c | 559 | conn_request_state(tconn, NS(conn, C_BROKEN_PIPE), CS_HARD); |
b411b363 PR |
560 | |
561 | return rv; | |
562 | } | |
563 | ||
5dbf1673 LE |
564 | /* quoting tcp(7): |
565 | * On individual connections, the socket buffer size must be set prior to the | |
566 | * listen(2) or connect(2) calls in order to have it take effect. | |
567 | * This is our wrapper to do so. | |
568 | */ | |
569 | static void drbd_setbufsize(struct socket *sock, unsigned int snd, | |
570 | unsigned int rcv) | |
571 | { | |
572 | /* open coded SO_SNDBUF, SO_RCVBUF */ | |
573 | if (snd) { | |
574 | sock->sk->sk_sndbuf = snd; | |
575 | sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK; | |
576 | } | |
577 | if (rcv) { | |
578 | sock->sk->sk_rcvbuf = rcv; | |
579 | sock->sk->sk_userlocks |= SOCK_RCVBUF_LOCK; | |
580 | } | |
581 | } | |
582 | ||
eac3e990 | 583 | static struct socket *drbd_try_connect(struct drbd_tconn *tconn) |
b411b363 PR |
584 | { |
585 | const char *what; | |
586 | struct socket *sock; | |
587 | struct sockaddr_in6 src_in6; | |
588 | int err; | |
589 | int disconnect_on_error = 1; | |
590 | ||
eac3e990 | 591 | if (!get_net_conf(tconn)) |
b411b363 PR |
592 | return NULL; |
593 | ||
594 | what = "sock_create_kern"; | |
eac3e990 | 595 | err = sock_create_kern(((struct sockaddr *)tconn->net_conf->my_addr)->sa_family, |
b411b363 PR |
596 | SOCK_STREAM, IPPROTO_TCP, &sock); |
597 | if (err < 0) { | |
598 | sock = NULL; | |
599 | goto out; | |
600 | } | |
601 | ||
602 | sock->sk->sk_rcvtimeo = | |
eac3e990 PR |
603 | sock->sk->sk_sndtimeo = tconn->net_conf->try_connect_int*HZ; |
604 | drbd_setbufsize(sock, tconn->net_conf->sndbuf_size, | |
605 | tconn->net_conf->rcvbuf_size); | |
b411b363 PR |
606 | |
607 | /* explicitly bind to the configured IP as source IP | |
608 | * for the outgoing connections. | |
609 | * This is needed for multihomed hosts and to be | |
610 | * able to use lo: interfaces for drbd. | |
611 | * Make sure to use 0 as port number, so linux selects | |
612 | * a free one dynamically. | |
613 | */ | |
eac3e990 PR |
614 | memcpy(&src_in6, tconn->net_conf->my_addr, |
615 | min_t(int, tconn->net_conf->my_addr_len, sizeof(src_in6))); | |
616 | if (((struct sockaddr *)tconn->net_conf->my_addr)->sa_family == AF_INET6) | |
b411b363 PR |
617 | src_in6.sin6_port = 0; |
618 | else | |
619 | ((struct sockaddr_in *)&src_in6)->sin_port = 0; /* AF_INET & AF_SCI */ | |
620 | ||
621 | what = "bind before connect"; | |
622 | err = sock->ops->bind(sock, | |
623 | (struct sockaddr *) &src_in6, | |
eac3e990 | 624 | tconn->net_conf->my_addr_len); |
b411b363 PR |
625 | if (err < 0) |
626 | goto out; | |
627 | ||
628 | /* connect may fail, peer not yet available. | |
629 | * stay C_WF_CONNECTION, don't go Disconnecting! */ | |
630 | disconnect_on_error = 0; | |
631 | what = "connect"; | |
632 | err = sock->ops->connect(sock, | |
eac3e990 PR |
633 | (struct sockaddr *)tconn->net_conf->peer_addr, |
634 | tconn->net_conf->peer_addr_len, 0); | |
b411b363 PR |
635 | |
636 | out: | |
637 | if (err < 0) { | |
638 | if (sock) { | |
639 | sock_release(sock); | |
640 | sock = NULL; | |
641 | } | |
642 | switch (-err) { | |
643 | /* timeout, busy, signal pending */ | |
644 | case ETIMEDOUT: case EAGAIN: case EINPROGRESS: | |
645 | case EINTR: case ERESTARTSYS: | |
646 | /* peer not (yet) available, network problem */ | |
647 | case ECONNREFUSED: case ENETUNREACH: | |
648 | case EHOSTDOWN: case EHOSTUNREACH: | |
649 | disconnect_on_error = 0; | |
650 | break; | |
651 | default: | |
eac3e990 | 652 | conn_err(tconn, "%s failed, err = %d\n", what, err); |
b411b363 PR |
653 | } |
654 | if (disconnect_on_error) | |
bbeb641c | 655 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 | 656 | } |
eac3e990 | 657 | put_net_conf(tconn); |
b411b363 PR |
658 | return sock; |
659 | } | |
660 | ||
7653620d | 661 | static struct socket *drbd_wait_for_connect(struct drbd_tconn *tconn) |
b411b363 PR |
662 | { |
663 | int timeo, err; | |
664 | struct socket *s_estab = NULL, *s_listen; | |
665 | const char *what; | |
666 | ||
7653620d | 667 | if (!get_net_conf(tconn)) |
b411b363 PR |
668 | return NULL; |
669 | ||
670 | what = "sock_create_kern"; | |
7653620d | 671 | err = sock_create_kern(((struct sockaddr *)tconn->net_conf->my_addr)->sa_family, |
b411b363 PR |
672 | SOCK_STREAM, IPPROTO_TCP, &s_listen); |
673 | if (err) { | |
674 | s_listen = NULL; | |
675 | goto out; | |
676 | } | |
677 | ||
7653620d | 678 | timeo = tconn->net_conf->try_connect_int * HZ; |
b411b363 PR |
679 | timeo += (random32() & 1) ? timeo / 7 : -timeo / 7; /* 28.5% random jitter */ |
680 | ||
681 | s_listen->sk->sk_reuse = 1; /* SO_REUSEADDR */ | |
682 | s_listen->sk->sk_rcvtimeo = timeo; | |
683 | s_listen->sk->sk_sndtimeo = timeo; | |
7653620d PR |
684 | drbd_setbufsize(s_listen, tconn->net_conf->sndbuf_size, |
685 | tconn->net_conf->rcvbuf_size); | |
b411b363 PR |
686 | |
687 | what = "bind before listen"; | |
688 | err = s_listen->ops->bind(s_listen, | |
7653620d PR |
689 | (struct sockaddr *) tconn->net_conf->my_addr, |
690 | tconn->net_conf->my_addr_len); | |
b411b363 PR |
691 | if (err < 0) |
692 | goto out; | |
693 | ||
7653620d | 694 | err = drbd_accept(&what, s_listen, &s_estab); |
b411b363 PR |
695 | |
696 | out: | |
697 | if (s_listen) | |
698 | sock_release(s_listen); | |
699 | if (err < 0) { | |
700 | if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) { | |
7653620d | 701 | conn_err(tconn, "%s failed, err = %d\n", what, err); |
bbeb641c | 702 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 PR |
703 | } |
704 | } | |
7653620d | 705 | put_net_conf(tconn); |
b411b363 PR |
706 | |
707 | return s_estab; | |
708 | } | |
709 | ||
d38e787e | 710 | static int drbd_send_fp(struct drbd_tconn *tconn, struct socket *sock, enum drbd_packet cmd) |
b411b363 | 711 | { |
d38e787e | 712 | struct p_header *h = &tconn->data.sbuf.header; |
b411b363 | 713 | |
d38e787e | 714 | return _conn_send_cmd(tconn, 0, sock, cmd, h, sizeof(*h), 0); |
b411b363 PR |
715 | } |
716 | ||
a25b63f1 | 717 | static enum drbd_packet drbd_recv_fp(struct drbd_tconn *tconn, struct socket *sock) |
b411b363 | 718 | { |
a25b63f1 | 719 | struct p_header80 *h = &tconn->data.rbuf.header.h80; |
b411b363 PR |
720 | int rr; |
721 | ||
dbd9eea0 | 722 | rr = drbd_recv_short(sock, h, sizeof(*h), 0); |
b411b363 | 723 | |
ca9bc12b | 724 | if (rr == sizeof(*h) && h->magic == cpu_to_be32(DRBD_MAGIC)) |
b411b363 PR |
725 | return be16_to_cpu(h->command); |
726 | ||
727 | return 0xffff; | |
728 | } | |
729 | ||
730 | /** | |
731 | * drbd_socket_okay() - Free the socket if its connection is not okay | |
b411b363 PR |
732 | * @sock: pointer to the pointer to the socket. |
733 | */ | |
dbd9eea0 | 734 | static int drbd_socket_okay(struct socket **sock) |
b411b363 PR |
735 | { |
736 | int rr; | |
737 | char tb[4]; | |
738 | ||
739 | if (!*sock) | |
81e84650 | 740 | return false; |
b411b363 | 741 | |
dbd9eea0 | 742 | rr = drbd_recv_short(*sock, tb, 4, MSG_DONTWAIT | MSG_PEEK); |
b411b363 PR |
743 | |
744 | if (rr > 0 || rr == -EAGAIN) { | |
81e84650 | 745 | return true; |
b411b363 PR |
746 | } else { |
747 | sock_release(*sock); | |
748 | *sock = NULL; | |
81e84650 | 749 | return false; |
b411b363 PR |
750 | } |
751 | } | |
2325eb66 PR |
752 | /* Gets called if a connection is established, or if a new minor gets created |
753 | in a connection */ | |
754 | int drbd_connected(int vnr, void *p, void *data) | |
907599e0 PR |
755 | { |
756 | struct drbd_conf *mdev = (struct drbd_conf *)p; | |
757 | int ok = 1; | |
758 | ||
759 | atomic_set(&mdev->packet_seq, 0); | |
760 | mdev->peer_seq = 0; | |
761 | ||
8410da8f PR |
762 | mdev->state_mutex = mdev->tconn->agreed_pro_version < 100 ? |
763 | &mdev->tconn->cstate_mutex : | |
764 | &mdev->own_state_mutex; | |
765 | ||
f399002e | 766 | ok &= drbd_send_sync_param(mdev); |
907599e0 PR |
767 | ok &= drbd_send_sizes(mdev, 0, 0); |
768 | ok &= drbd_send_uuids(mdev); | |
769 | ok &= drbd_send_state(mdev); | |
770 | clear_bit(USE_DEGR_WFC_T, &mdev->flags); | |
771 | clear_bit(RESIZE_PENDING, &mdev->flags); | |
772 | ||
8410da8f | 773 | |
907599e0 PR |
774 | return !ok; |
775 | } | |
776 | ||
b411b363 PR |
777 | /* |
778 | * return values: | |
779 | * 1 yes, we have a valid connection | |
780 | * 0 oops, did not work out, please try again | |
781 | * -1 peer talks different language, | |
782 | * no point in trying again, please go standalone. | |
783 | * -2 We do not have a network config... | |
784 | */ | |
907599e0 | 785 | static int drbd_connect(struct drbd_tconn *tconn) |
b411b363 PR |
786 | { |
787 | struct socket *s, *sock, *msock; | |
788 | int try, h, ok; | |
789 | ||
bbeb641c | 790 | if (conn_request_state(tconn, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS) |
b411b363 PR |
791 | return -2; |
792 | ||
907599e0 PR |
793 | clear_bit(DISCARD_CONCURRENT, &tconn->flags); |
794 | tconn->agreed_pro_version = 99; | |
fd340c12 PR |
795 | /* agreed_pro_version must be smaller than 100 so we send the old |
796 | header (h80) in the first packet and in the handshake packet. */ | |
b411b363 PR |
797 | |
798 | sock = NULL; | |
799 | msock = NULL; | |
800 | ||
801 | do { | |
802 | for (try = 0;;) { | |
803 | /* 3 tries, this should take less than a second! */ | |
907599e0 | 804 | s = drbd_try_connect(tconn); |
b411b363 PR |
805 | if (s || ++try >= 3) |
806 | break; | |
807 | /* give the other side time to call bind() & listen() */ | |
20ee6390 | 808 | schedule_timeout_interruptible(HZ / 10); |
b411b363 PR |
809 | } |
810 | ||
811 | if (s) { | |
812 | if (!sock) { | |
907599e0 | 813 | drbd_send_fp(tconn, s, P_HAND_SHAKE_S); |
b411b363 PR |
814 | sock = s; |
815 | s = NULL; | |
816 | } else if (!msock) { | |
907599e0 | 817 | drbd_send_fp(tconn, s, P_HAND_SHAKE_M); |
b411b363 PR |
818 | msock = s; |
819 | s = NULL; | |
820 | } else { | |
907599e0 | 821 | conn_err(tconn, "Logic error in drbd_connect()\n"); |
b411b363 PR |
822 | goto out_release_sockets; |
823 | } | |
824 | } | |
825 | ||
826 | if (sock && msock) { | |
907599e0 | 827 | schedule_timeout_interruptible(tconn->net_conf->ping_timeo*HZ/10); |
dbd9eea0 PR |
828 | ok = drbd_socket_okay(&sock); |
829 | ok = drbd_socket_okay(&msock) && ok; | |
b411b363 PR |
830 | if (ok) |
831 | break; | |
832 | } | |
833 | ||
834 | retry: | |
907599e0 | 835 | s = drbd_wait_for_connect(tconn); |
b411b363 | 836 | if (s) { |
907599e0 | 837 | try = drbd_recv_fp(tconn, s); |
dbd9eea0 PR |
838 | drbd_socket_okay(&sock); |
839 | drbd_socket_okay(&msock); | |
b411b363 PR |
840 | switch (try) { |
841 | case P_HAND_SHAKE_S: | |
842 | if (sock) { | |
907599e0 | 843 | conn_warn(tconn, "initial packet S crossed\n"); |
b411b363 PR |
844 | sock_release(sock); |
845 | } | |
846 | sock = s; | |
847 | break; | |
848 | case P_HAND_SHAKE_M: | |
849 | if (msock) { | |
907599e0 | 850 | conn_warn(tconn, "initial packet M crossed\n"); |
b411b363 PR |
851 | sock_release(msock); |
852 | } | |
853 | msock = s; | |
907599e0 | 854 | set_bit(DISCARD_CONCURRENT, &tconn->flags); |
b411b363 PR |
855 | break; |
856 | default: | |
907599e0 | 857 | conn_warn(tconn, "Error receiving initial packet\n"); |
b411b363 PR |
858 | sock_release(s); |
859 | if (random32() & 1) | |
860 | goto retry; | |
861 | } | |
862 | } | |
863 | ||
bbeb641c | 864 | if (tconn->cstate <= C_DISCONNECTING) |
b411b363 PR |
865 | goto out_release_sockets; |
866 | if (signal_pending(current)) { | |
867 | flush_signals(current); | |
868 | smp_rmb(); | |
907599e0 | 869 | if (get_t_state(&tconn->receiver) == EXITING) |
b411b363 PR |
870 | goto out_release_sockets; |
871 | } | |
872 | ||
873 | if (sock && msock) { | |
dbd9eea0 PR |
874 | ok = drbd_socket_okay(&sock); |
875 | ok = drbd_socket_okay(&msock) && ok; | |
b411b363 PR |
876 | if (ok) |
877 | break; | |
878 | } | |
879 | } while (1); | |
880 | ||
881 | msock->sk->sk_reuse = 1; /* SO_REUSEADDR */ | |
882 | sock->sk->sk_reuse = 1; /* SO_REUSEADDR */ | |
883 | ||
884 | sock->sk->sk_allocation = GFP_NOIO; | |
885 | msock->sk->sk_allocation = GFP_NOIO; | |
886 | ||
887 | sock->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK; | |
888 | msock->sk->sk_priority = TC_PRIO_INTERACTIVE; | |
889 | ||
b411b363 | 890 | /* NOT YET ... |
907599e0 | 891 | * sock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10; |
b411b363 PR |
892 | * sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; |
893 | * first set it to the P_HAND_SHAKE timeout, | |
894 | * which we set to 4x the configured ping_timeout. */ | |
895 | sock->sk->sk_sndtimeo = | |
907599e0 | 896 | sock->sk->sk_rcvtimeo = tconn->net_conf->ping_timeo*4*HZ/10; |
b411b363 | 897 | |
907599e0 PR |
898 | msock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10; |
899 | msock->sk->sk_rcvtimeo = tconn->net_conf->ping_int*HZ; | |
b411b363 PR |
900 | |
901 | /* we don't want delays. | |
25985edc | 902 | * we use TCP_CORK where appropriate, though */ |
b411b363 PR |
903 | drbd_tcp_nodelay(sock); |
904 | drbd_tcp_nodelay(msock); | |
905 | ||
907599e0 PR |
906 | tconn->data.socket = sock; |
907 | tconn->meta.socket = msock; | |
908 | tconn->last_received = jiffies; | |
b411b363 | 909 | |
907599e0 | 910 | h = drbd_do_handshake(tconn); |
b411b363 PR |
911 | if (h <= 0) |
912 | return h; | |
913 | ||
907599e0 | 914 | if (tconn->cram_hmac_tfm) { |
b411b363 | 915 | /* drbd_request_state(mdev, NS(conn, WFAuth)); */ |
907599e0 | 916 | switch (drbd_do_auth(tconn)) { |
b10d96cb | 917 | case -1: |
907599e0 | 918 | conn_err(tconn, "Authentication of peer failed\n"); |
b411b363 | 919 | return -1; |
b10d96cb | 920 | case 0: |
907599e0 | 921 | conn_err(tconn, "Authentication of peer failed, trying again.\n"); |
b10d96cb | 922 | return 0; |
b411b363 PR |
923 | } |
924 | } | |
925 | ||
bbeb641c | 926 | if (conn_request_state(tconn, NS(conn, C_WF_REPORT_PARAMS), CS_VERBOSE) < SS_SUCCESS) |
b411b363 PR |
927 | return 0; |
928 | ||
907599e0 | 929 | sock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10; |
b411b363 PR |
930 | sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; |
931 | ||
907599e0 | 932 | drbd_thread_start(&tconn->asender); |
b411b363 | 933 | |
907599e0 | 934 | if (drbd_send_protocol(tconn) == -1) |
7e2455c1 | 935 | return -1; |
b411b363 | 936 | |
907599e0 | 937 | return !idr_for_each(&tconn->volumes, drbd_connected, tconn); |
b411b363 PR |
938 | |
939 | out_release_sockets: | |
940 | if (sock) | |
941 | sock_release(sock); | |
942 | if (msock) | |
943 | sock_release(msock); | |
944 | return -1; | |
945 | } | |
946 | ||
ce243853 | 947 | static bool decode_header(struct drbd_tconn *tconn, struct p_header *h, struct packet_info *pi) |
b411b363 | 948 | { |
fd340c12 | 949 | if (h->h80.magic == cpu_to_be32(DRBD_MAGIC)) { |
77351055 PR |
950 | pi->cmd = be16_to_cpu(h->h80.command); |
951 | pi->size = be16_to_cpu(h->h80.length); | |
eefc2f7d | 952 | pi->vnr = 0; |
ca9bc12b | 953 | } else if (h->h95.magic == cpu_to_be16(DRBD_MAGIC_BIG)) { |
77351055 PR |
954 | pi->cmd = be16_to_cpu(h->h95.command); |
955 | pi->size = be32_to_cpu(h->h95.length) & 0x00ffffff; | |
956 | pi->vnr = 0; | |
02918be2 | 957 | } else { |
ce243853 | 958 | conn_err(tconn, "magic?? on data m: 0x%08x c: %d l: %d\n", |
004352fa LE |
959 | be32_to_cpu(h->h80.magic), |
960 | be16_to_cpu(h->h80.command), | |
961 | be16_to_cpu(h->h80.length)); | |
81e84650 | 962 | return false; |
b411b363 | 963 | } |
257d0af6 PR |
964 | return true; |
965 | } | |
966 | ||
9ba7aa00 | 967 | static int drbd_recv_header(struct drbd_tconn *tconn, struct packet_info *pi) |
257d0af6 | 968 | { |
9ba7aa00 | 969 | struct p_header *h = &tconn->data.rbuf.header; |
257d0af6 PR |
970 | int r; |
971 | ||
9ba7aa00 | 972 | r = drbd_recv(tconn, h, sizeof(*h)); |
257d0af6 PR |
973 | if (unlikely(r != sizeof(*h))) { |
974 | if (!signal_pending(current)) | |
9ba7aa00 | 975 | conn_warn(tconn, "short read expecting header on sock: r=%d\n", r); |
257d0af6 PR |
976 | return false; |
977 | } | |
978 | ||
9ba7aa00 PR |
979 | r = decode_header(tconn, h, pi); |
980 | tconn->last_received = jiffies; | |
b411b363 | 981 | |
257d0af6 | 982 | return r; |
b411b363 PR |
983 | } |
984 | ||
2451fc3b | 985 | static void drbd_flush(struct drbd_conf *mdev) |
b411b363 PR |
986 | { |
987 | int rv; | |
988 | ||
989 | if (mdev->write_ordering >= WO_bdev_flush && get_ldev(mdev)) { | |
fbd9b09a | 990 | rv = blkdev_issue_flush(mdev->ldev->backing_bdev, GFP_KERNEL, |
dd3932ed | 991 | NULL); |
b411b363 PR |
992 | if (rv) { |
993 | dev_err(DEV, "local disk flush failed with status %d\n", rv); | |
994 | /* would rather check on EOPNOTSUPP, but that is not reliable. | |
995 | * don't try again for ANY return value != 0 | |
996 | * if (rv == -EOPNOTSUPP) */ | |
997 | drbd_bump_write_ordering(mdev, WO_drain_io); | |
998 | } | |
999 | put_ldev(mdev); | |
1000 | } | |
b411b363 PR |
1001 | } |
1002 | ||
1003 | /** | |
1004 | * drbd_may_finish_epoch() - Applies an epoch_event to the epoch's state, eventually finishes it. | |
1005 | * @mdev: DRBD device. | |
1006 | * @epoch: Epoch object. | |
1007 | * @ev: Epoch event. | |
1008 | */ | |
1009 | static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, | |
1010 | struct drbd_epoch *epoch, | |
1011 | enum epoch_event ev) | |
1012 | { | |
2451fc3b | 1013 | int epoch_size; |
b411b363 | 1014 | struct drbd_epoch *next_epoch; |
b411b363 PR |
1015 | enum finish_epoch rv = FE_STILL_LIVE; |
1016 | ||
1017 | spin_lock(&mdev->epoch_lock); | |
1018 | do { | |
1019 | next_epoch = NULL; | |
b411b363 PR |
1020 | |
1021 | epoch_size = atomic_read(&epoch->epoch_size); | |
1022 | ||
1023 | switch (ev & ~EV_CLEANUP) { | |
1024 | case EV_PUT: | |
1025 | atomic_dec(&epoch->active); | |
1026 | break; | |
1027 | case EV_GOT_BARRIER_NR: | |
1028 | set_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags); | |
b411b363 PR |
1029 | break; |
1030 | case EV_BECAME_LAST: | |
1031 | /* nothing to do*/ | |
1032 | break; | |
1033 | } | |
1034 | ||
b411b363 PR |
1035 | if (epoch_size != 0 && |
1036 | atomic_read(&epoch->active) == 0 && | |
2451fc3b | 1037 | test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags)) { |
b411b363 PR |
1038 | if (!(ev & EV_CLEANUP)) { |
1039 | spin_unlock(&mdev->epoch_lock); | |
1040 | drbd_send_b_ack(mdev, epoch->barrier_nr, epoch_size); | |
1041 | spin_lock(&mdev->epoch_lock); | |
1042 | } | |
1043 | dec_unacked(mdev); | |
1044 | ||
1045 | if (mdev->current_epoch != epoch) { | |
1046 | next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list); | |
1047 | list_del(&epoch->list); | |
1048 | ev = EV_BECAME_LAST | (ev & EV_CLEANUP); | |
1049 | mdev->epochs--; | |
b411b363 PR |
1050 | kfree(epoch); |
1051 | ||
1052 | if (rv == FE_STILL_LIVE) | |
1053 | rv = FE_DESTROYED; | |
1054 | } else { | |
1055 | epoch->flags = 0; | |
1056 | atomic_set(&epoch->epoch_size, 0); | |
698f9315 | 1057 | /* atomic_set(&epoch->active, 0); is already zero */ |
b411b363 PR |
1058 | if (rv == FE_STILL_LIVE) |
1059 | rv = FE_RECYCLED; | |
2451fc3b | 1060 | wake_up(&mdev->ee_wait); |
b411b363 PR |
1061 | } |
1062 | } | |
1063 | ||
1064 | if (!next_epoch) | |
1065 | break; | |
1066 | ||
1067 | epoch = next_epoch; | |
1068 | } while (1); | |
1069 | ||
1070 | spin_unlock(&mdev->epoch_lock); | |
1071 | ||
b411b363 PR |
1072 | return rv; |
1073 | } | |
1074 | ||
1075 | /** | |
1076 | * drbd_bump_write_ordering() - Fall back to an other write ordering method | |
1077 | * @mdev: DRBD device. | |
1078 | * @wo: Write ordering method to try. | |
1079 | */ | |
1080 | void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo) __must_hold(local) | |
1081 | { | |
1082 | enum write_ordering_e pwo; | |
1083 | static char *write_ordering_str[] = { | |
1084 | [WO_none] = "none", | |
1085 | [WO_drain_io] = "drain", | |
1086 | [WO_bdev_flush] = "flush", | |
b411b363 PR |
1087 | }; |
1088 | ||
1089 | pwo = mdev->write_ordering; | |
1090 | wo = min(pwo, wo); | |
b411b363 PR |
1091 | if (wo == WO_bdev_flush && mdev->ldev->dc.no_disk_flush) |
1092 | wo = WO_drain_io; | |
1093 | if (wo == WO_drain_io && mdev->ldev->dc.no_disk_drain) | |
1094 | wo = WO_none; | |
1095 | mdev->write_ordering = wo; | |
2451fc3b | 1096 | if (pwo != mdev->write_ordering || wo == WO_bdev_flush) |
b411b363 PR |
1097 | dev_info(DEV, "Method to ensure write ordering: %s\n", write_ordering_str[mdev->write_ordering]); |
1098 | } | |
1099 | ||
45bb912b | 1100 | /** |
fbe29dec | 1101 | * drbd_submit_peer_request() |
45bb912b | 1102 | * @mdev: DRBD device. |
db830c46 | 1103 | * @peer_req: peer request |
45bb912b | 1104 | * @rw: flag field, see bio->bi_rw |
10f6d992 LE |
1105 | * |
1106 | * May spread the pages to multiple bios, | |
1107 | * depending on bio_add_page restrictions. | |
1108 | * | |
1109 | * Returns 0 if all bios have been submitted, | |
1110 | * -ENOMEM if we could not allocate enough bios, | |
1111 | * -ENOSPC (any better suggestion?) if we have not been able to bio_add_page a | |
1112 | * single page to an empty bio (which should never happen and likely indicates | |
1113 | * that the lower level IO stack is in some way broken). This has been observed | |
1114 | * on certain Xen deployments. | |
45bb912b LE |
1115 | */ |
1116 | /* TODO allocate from our own bio_set. */ | |
fbe29dec AG |
1117 | int drbd_submit_peer_request(struct drbd_conf *mdev, |
1118 | struct drbd_peer_request *peer_req, | |
1119 | const unsigned rw, const int fault_type) | |
45bb912b LE |
1120 | { |
1121 | struct bio *bios = NULL; | |
1122 | struct bio *bio; | |
db830c46 AG |
1123 | struct page *page = peer_req->pages; |
1124 | sector_t sector = peer_req->i.sector; | |
1125 | unsigned ds = peer_req->i.size; | |
45bb912b LE |
1126 | unsigned n_bios = 0; |
1127 | unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT; | |
10f6d992 | 1128 | int err = -ENOMEM; |
45bb912b LE |
1129 | |
1130 | /* In most cases, we will only need one bio. But in case the lower | |
1131 | * level restrictions happen to be different at this offset on this | |
1132 | * side than those of the sending peer, we may need to submit the | |
da4a75d2 LE |
1133 | * request in more than one bio. |
1134 | * | |
1135 | * Plain bio_alloc is good enough here, this is no DRBD internally | |
1136 | * generated bio, but a bio allocated on behalf of the peer. | |
1137 | */ | |
45bb912b LE |
1138 | next_bio: |
1139 | bio = bio_alloc(GFP_NOIO, nr_pages); | |
1140 | if (!bio) { | |
1141 | dev_err(DEV, "submit_ee: Allocation of a bio failed\n"); | |
1142 | goto fail; | |
1143 | } | |
db830c46 | 1144 | /* > peer_req->i.sector, unless this is the first bio */ |
45bb912b LE |
1145 | bio->bi_sector = sector; |
1146 | bio->bi_bdev = mdev->ldev->backing_bdev; | |
45bb912b | 1147 | bio->bi_rw = rw; |
db830c46 | 1148 | bio->bi_private = peer_req; |
fcefa62e | 1149 | bio->bi_end_io = drbd_peer_request_endio; |
45bb912b LE |
1150 | |
1151 | bio->bi_next = bios; | |
1152 | bios = bio; | |
1153 | ++n_bios; | |
1154 | ||
1155 | page_chain_for_each(page) { | |
1156 | unsigned len = min_t(unsigned, ds, PAGE_SIZE); | |
1157 | if (!bio_add_page(bio, page, len, 0)) { | |
10f6d992 LE |
1158 | /* A single page must always be possible! |
1159 | * But in case it fails anyways, | |
1160 | * we deal with it, and complain (below). */ | |
1161 | if (bio->bi_vcnt == 0) { | |
1162 | dev_err(DEV, | |
1163 | "bio_add_page failed for len=%u, " | |
1164 | "bi_vcnt=0 (bi_sector=%llu)\n", | |
1165 | len, (unsigned long long)bio->bi_sector); | |
1166 | err = -ENOSPC; | |
1167 | goto fail; | |
1168 | } | |
45bb912b LE |
1169 | goto next_bio; |
1170 | } | |
1171 | ds -= len; | |
1172 | sector += len >> 9; | |
1173 | --nr_pages; | |
1174 | } | |
1175 | D_ASSERT(page == NULL); | |
1176 | D_ASSERT(ds == 0); | |
1177 | ||
db830c46 | 1178 | atomic_set(&peer_req->pending_bios, n_bios); |
45bb912b LE |
1179 | do { |
1180 | bio = bios; | |
1181 | bios = bios->bi_next; | |
1182 | bio->bi_next = NULL; | |
1183 | ||
45bb912b | 1184 | drbd_generic_make_request(mdev, fault_type, bio); |
45bb912b | 1185 | } while (bios); |
45bb912b LE |
1186 | return 0; |
1187 | ||
1188 | fail: | |
1189 | while (bios) { | |
1190 | bio = bios; | |
1191 | bios = bios->bi_next; | |
1192 | bio_put(bio); | |
1193 | } | |
10f6d992 | 1194 | return err; |
45bb912b LE |
1195 | } |
1196 | ||
53840641 | 1197 | static void drbd_remove_epoch_entry_interval(struct drbd_conf *mdev, |
db830c46 | 1198 | struct drbd_peer_request *peer_req) |
53840641 | 1199 | { |
db830c46 | 1200 | struct drbd_interval *i = &peer_req->i; |
53840641 AG |
1201 | |
1202 | drbd_remove_interval(&mdev->write_requests, i); | |
1203 | drbd_clear_interval(i); | |
1204 | ||
6c852bec | 1205 | /* Wake up any processes waiting for this peer request to complete. */ |
53840641 AG |
1206 | if (i->waiting) |
1207 | wake_up(&mdev->misc_wait); | |
1208 | } | |
1209 | ||
d8763023 AG |
1210 | static int receive_Barrier(struct drbd_conf *mdev, enum drbd_packet cmd, |
1211 | unsigned int data_size) | |
b411b363 | 1212 | { |
2451fc3b | 1213 | int rv; |
e42325a5 | 1214 | struct p_barrier *p = &mdev->tconn->data.rbuf.barrier; |
b411b363 PR |
1215 | struct drbd_epoch *epoch; |
1216 | ||
b411b363 PR |
1217 | inc_unacked(mdev); |
1218 | ||
b411b363 PR |
1219 | mdev->current_epoch->barrier_nr = p->barrier; |
1220 | rv = drbd_may_finish_epoch(mdev, mdev->current_epoch, EV_GOT_BARRIER_NR); | |
1221 | ||
1222 | /* P_BARRIER_ACK may imply that the corresponding extent is dropped from | |
1223 | * the activity log, which means it would not be resynced in case the | |
1224 | * R_PRIMARY crashes now. | |
1225 | * Therefore we must send the barrier_ack after the barrier request was | |
1226 | * completed. */ | |
1227 | switch (mdev->write_ordering) { | |
b411b363 PR |
1228 | case WO_none: |
1229 | if (rv == FE_RECYCLED) | |
81e84650 | 1230 | return true; |
2451fc3b PR |
1231 | |
1232 | /* receiver context, in the writeout path of the other node. | |
1233 | * avoid potential distributed deadlock */ | |
1234 | epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); | |
1235 | if (epoch) | |
1236 | break; | |
1237 | else | |
1238 | dev_warn(DEV, "Allocation of an epoch failed, slowing down\n"); | |
1239 | /* Fall through */ | |
b411b363 PR |
1240 | |
1241 | case WO_bdev_flush: | |
1242 | case WO_drain_io: | |
b411b363 | 1243 | drbd_wait_ee_list_empty(mdev, &mdev->active_ee); |
2451fc3b PR |
1244 | drbd_flush(mdev); |
1245 | ||
1246 | if (atomic_read(&mdev->current_epoch->epoch_size)) { | |
1247 | epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); | |
1248 | if (epoch) | |
1249 | break; | |
b411b363 PR |
1250 | } |
1251 | ||
2451fc3b PR |
1252 | epoch = mdev->current_epoch; |
1253 | wait_event(mdev->ee_wait, atomic_read(&epoch->epoch_size) == 0); | |
1254 | ||
1255 | D_ASSERT(atomic_read(&epoch->active) == 0); | |
1256 | D_ASSERT(epoch->flags == 0); | |
b411b363 | 1257 | |
81e84650 | 1258 | return true; |
2451fc3b PR |
1259 | default: |
1260 | dev_err(DEV, "Strangeness in mdev->write_ordering %d\n", mdev->write_ordering); | |
81e84650 | 1261 | return false; |
b411b363 PR |
1262 | } |
1263 | ||
1264 | epoch->flags = 0; | |
1265 | atomic_set(&epoch->epoch_size, 0); | |
1266 | atomic_set(&epoch->active, 0); | |
1267 | ||
1268 | spin_lock(&mdev->epoch_lock); | |
1269 | if (atomic_read(&mdev->current_epoch->epoch_size)) { | |
1270 | list_add(&epoch->list, &mdev->current_epoch->list); | |
1271 | mdev->current_epoch = epoch; | |
1272 | mdev->epochs++; | |
b411b363 PR |
1273 | } else { |
1274 | /* The current_epoch got recycled while we allocated this one... */ | |
1275 | kfree(epoch); | |
1276 | } | |
1277 | spin_unlock(&mdev->epoch_lock); | |
1278 | ||
81e84650 | 1279 | return true; |
b411b363 PR |
1280 | } |
1281 | ||
1282 | /* used from receive_RSDataReply (recv_resync_read) | |
1283 | * and from receive_Data */ | |
f6ffca9f AG |
1284 | static struct drbd_peer_request * |
1285 | read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, | |
1286 | int data_size) __must_hold(local) | |
b411b363 | 1287 | { |
6666032a | 1288 | const sector_t capacity = drbd_get_capacity(mdev->this_bdev); |
db830c46 | 1289 | struct drbd_peer_request *peer_req; |
b411b363 | 1290 | struct page *page; |
45bb912b | 1291 | int dgs, ds, rr; |
a0638456 PR |
1292 | void *dig_in = mdev->tconn->int_dig_in; |
1293 | void *dig_vv = mdev->tconn->int_dig_vv; | |
6b4388ac | 1294 | unsigned long *data; |
b411b363 | 1295 | |
a0638456 PR |
1296 | dgs = (mdev->tconn->agreed_pro_version >= 87 && mdev->tconn->integrity_r_tfm) ? |
1297 | crypto_hash_digestsize(mdev->tconn->integrity_r_tfm) : 0; | |
b411b363 PR |
1298 | |
1299 | if (dgs) { | |
de0ff338 | 1300 | rr = drbd_recv(mdev->tconn, dig_in, dgs); |
b411b363 | 1301 | if (rr != dgs) { |
0ddc5549 LE |
1302 | if (!signal_pending(current)) |
1303 | dev_warn(DEV, | |
1304 | "short read receiving data digest: read %d expected %d\n", | |
1305 | rr, dgs); | |
b411b363 PR |
1306 | return NULL; |
1307 | } | |
1308 | } | |
1309 | ||
1310 | data_size -= dgs; | |
1311 | ||
841ce241 AG |
1312 | if (!expect(data_size != 0)) |
1313 | return NULL; | |
1314 | if (!expect(IS_ALIGNED(data_size, 512))) | |
1315 | return NULL; | |
1316 | if (!expect(data_size <= DRBD_MAX_BIO_SIZE)) | |
1317 | return NULL; | |
b411b363 | 1318 | |
6666032a LE |
1319 | /* even though we trust out peer, |
1320 | * we sometimes have to double check. */ | |
1321 | if (sector + (data_size>>9) > capacity) { | |
fdda6544 LE |
1322 | dev_err(DEV, "request from peer beyond end of local disk: " |
1323 | "capacity: %llus < sector: %llus + size: %u\n", | |
6666032a LE |
1324 | (unsigned long long)capacity, |
1325 | (unsigned long long)sector, data_size); | |
1326 | return NULL; | |
1327 | } | |
1328 | ||
b411b363 PR |
1329 | /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD |
1330 | * "criss-cross" setup, that might cause write-out on some other DRBD, | |
1331 | * which in turn might block on the other node at this very place. */ | |
db830c46 AG |
1332 | peer_req = drbd_alloc_ee(mdev, id, sector, data_size, GFP_NOIO); |
1333 | if (!peer_req) | |
b411b363 | 1334 | return NULL; |
45bb912b | 1335 | |
b411b363 | 1336 | ds = data_size; |
db830c46 | 1337 | page = peer_req->pages; |
45bb912b LE |
1338 | page_chain_for_each(page) { |
1339 | unsigned len = min_t(int, ds, PAGE_SIZE); | |
6b4388ac | 1340 | data = kmap(page); |
de0ff338 | 1341 | rr = drbd_recv(mdev->tconn, data, len); |
0cf9d27e | 1342 | if (drbd_insert_fault(mdev, DRBD_FAULT_RECEIVE)) { |
6b4388ac PR |
1343 | dev_err(DEV, "Fault injection: Corrupting data on receive\n"); |
1344 | data[0] = data[0] ^ (unsigned long)-1; | |
1345 | } | |
b411b363 | 1346 | kunmap(page); |
45bb912b | 1347 | if (rr != len) { |
db830c46 | 1348 | drbd_free_ee(mdev, peer_req); |
0ddc5549 LE |
1349 | if (!signal_pending(current)) |
1350 | dev_warn(DEV, "short read receiving data: read %d expected %d\n", | |
1351 | rr, len); | |
b411b363 PR |
1352 | return NULL; |
1353 | } | |
1354 | ds -= rr; | |
1355 | } | |
1356 | ||
1357 | if (dgs) { | |
db830c46 | 1358 | drbd_csum_ee(mdev, mdev->tconn->integrity_r_tfm, peer_req, dig_vv); |
b411b363 | 1359 | if (memcmp(dig_in, dig_vv, dgs)) { |
470be44a LE |
1360 | dev_err(DEV, "Digest integrity check FAILED: %llus +%u\n", |
1361 | (unsigned long long)sector, data_size); | |
db830c46 | 1362 | drbd_free_ee(mdev, peer_req); |
b411b363 PR |
1363 | return NULL; |
1364 | } | |
1365 | } | |
1366 | mdev->recv_cnt += data_size>>9; | |
db830c46 | 1367 | return peer_req; |
b411b363 PR |
1368 | } |
1369 | ||
1370 | /* drbd_drain_block() just takes a data block | |
1371 | * out of the socket input buffer, and discards it. | |
1372 | */ | |
1373 | static int drbd_drain_block(struct drbd_conf *mdev, int data_size) | |
1374 | { | |
1375 | struct page *page; | |
1376 | int rr, rv = 1; | |
1377 | void *data; | |
1378 | ||
c3470cde | 1379 | if (!data_size) |
81e84650 | 1380 | return true; |
c3470cde | 1381 | |
45bb912b | 1382 | page = drbd_pp_alloc(mdev, 1, 1); |
b411b363 PR |
1383 | |
1384 | data = kmap(page); | |
1385 | while (data_size) { | |
de0ff338 | 1386 | rr = drbd_recv(mdev->tconn, data, min_t(int, data_size, PAGE_SIZE)); |
b411b363 PR |
1387 | if (rr != min_t(int, data_size, PAGE_SIZE)) { |
1388 | rv = 0; | |
0ddc5549 LE |
1389 | if (!signal_pending(current)) |
1390 | dev_warn(DEV, | |
1391 | "short read receiving data: read %d expected %d\n", | |
1392 | rr, min_t(int, data_size, PAGE_SIZE)); | |
b411b363 PR |
1393 | break; |
1394 | } | |
1395 | data_size -= rr; | |
1396 | } | |
1397 | kunmap(page); | |
435f0740 | 1398 | drbd_pp_free(mdev, page, 0); |
b411b363 PR |
1399 | return rv; |
1400 | } | |
1401 | ||
1402 | static int recv_dless_read(struct drbd_conf *mdev, struct drbd_request *req, | |
1403 | sector_t sector, int data_size) | |
1404 | { | |
1405 | struct bio_vec *bvec; | |
1406 | struct bio *bio; | |
1407 | int dgs, rr, i, expect; | |
a0638456 PR |
1408 | void *dig_in = mdev->tconn->int_dig_in; |
1409 | void *dig_vv = mdev->tconn->int_dig_vv; | |
b411b363 | 1410 | |
a0638456 PR |
1411 | dgs = (mdev->tconn->agreed_pro_version >= 87 && mdev->tconn->integrity_r_tfm) ? |
1412 | crypto_hash_digestsize(mdev->tconn->integrity_r_tfm) : 0; | |
b411b363 PR |
1413 | |
1414 | if (dgs) { | |
de0ff338 | 1415 | rr = drbd_recv(mdev->tconn, dig_in, dgs); |
b411b363 | 1416 | if (rr != dgs) { |
0ddc5549 LE |
1417 | if (!signal_pending(current)) |
1418 | dev_warn(DEV, | |
1419 | "short read receiving data reply digest: read %d expected %d\n", | |
1420 | rr, dgs); | |
b411b363 PR |
1421 | return 0; |
1422 | } | |
1423 | } | |
1424 | ||
1425 | data_size -= dgs; | |
1426 | ||
1427 | /* optimistically update recv_cnt. if receiving fails below, | |
1428 | * we disconnect anyways, and counters will be reset. */ | |
1429 | mdev->recv_cnt += data_size>>9; | |
1430 | ||
1431 | bio = req->master_bio; | |
1432 | D_ASSERT(sector == bio->bi_sector); | |
1433 | ||
1434 | bio_for_each_segment(bvec, bio, i) { | |
1435 | expect = min_t(int, data_size, bvec->bv_len); | |
de0ff338 | 1436 | rr = drbd_recv(mdev->tconn, |
b411b363 PR |
1437 | kmap(bvec->bv_page)+bvec->bv_offset, |
1438 | expect); | |
1439 | kunmap(bvec->bv_page); | |
1440 | if (rr != expect) { | |
0ddc5549 LE |
1441 | if (!signal_pending(current)) |
1442 | dev_warn(DEV, "short read receiving data reply: " | |
1443 | "read %d expected %d\n", | |
1444 | rr, expect); | |
b411b363 PR |
1445 | return 0; |
1446 | } | |
1447 | data_size -= rr; | |
1448 | } | |
1449 | ||
1450 | if (dgs) { | |
a0638456 | 1451 | drbd_csum_bio(mdev, mdev->tconn->integrity_r_tfm, bio, dig_vv); |
b411b363 PR |
1452 | if (memcmp(dig_in, dig_vv, dgs)) { |
1453 | dev_err(DEV, "Digest integrity check FAILED. Broken NICs?\n"); | |
1454 | return 0; | |
1455 | } | |
1456 | } | |
1457 | ||
1458 | D_ASSERT(data_size == 0); | |
1459 | return 1; | |
1460 | } | |
1461 | ||
1462 | /* e_end_resync_block() is called via | |
1463 | * drbd_process_done_ee() by asender only */ | |
00d56944 | 1464 | static int e_end_resync_block(struct drbd_work *w, int unused) |
b411b363 | 1465 | { |
8050e6d0 AG |
1466 | struct drbd_peer_request *peer_req = |
1467 | container_of(w, struct drbd_peer_request, w); | |
00d56944 | 1468 | struct drbd_conf *mdev = w->mdev; |
db830c46 | 1469 | sector_t sector = peer_req->i.sector; |
b411b363 PR |
1470 | int ok; |
1471 | ||
db830c46 | 1472 | D_ASSERT(drbd_interval_empty(&peer_req->i)); |
b411b363 | 1473 | |
db830c46 AG |
1474 | if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) { |
1475 | drbd_set_in_sync(mdev, sector, peer_req->i.size); | |
1476 | ok = drbd_send_ack(mdev, P_RS_WRITE_ACK, peer_req); | |
b411b363 PR |
1477 | } else { |
1478 | /* Record failure to sync */ | |
db830c46 | 1479 | drbd_rs_failed_io(mdev, sector, peer_req->i.size); |
b411b363 | 1480 | |
db830c46 | 1481 | ok = drbd_send_ack(mdev, P_NEG_ACK, peer_req); |
b411b363 PR |
1482 | } |
1483 | dec_unacked(mdev); | |
1484 | ||
1485 | return ok; | |
1486 | } | |
1487 | ||
1488 | static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_size) __releases(local) | |
1489 | { | |
db830c46 | 1490 | struct drbd_peer_request *peer_req; |
b411b363 | 1491 | |
db830c46 AG |
1492 | peer_req = read_in_block(mdev, ID_SYNCER, sector, data_size); |
1493 | if (!peer_req) | |
45bb912b | 1494 | goto fail; |
b411b363 PR |
1495 | |
1496 | dec_rs_pending(mdev); | |
1497 | ||
b411b363 PR |
1498 | inc_unacked(mdev); |
1499 | /* corresponding dec_unacked() in e_end_resync_block() | |
1500 | * respective _drbd_clear_done_ee */ | |
1501 | ||
db830c46 | 1502 | peer_req->w.cb = e_end_resync_block; |
45bb912b | 1503 | |
87eeee41 | 1504 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 1505 | list_add(&peer_req->w.list, &mdev->sync_ee); |
87eeee41 | 1506 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 1507 | |
0f0601f4 | 1508 | atomic_add(data_size >> 9, &mdev->rs_sect_ev); |
fbe29dec | 1509 | if (drbd_submit_peer_request(mdev, peer_req, WRITE, DRBD_FAULT_RS_WR) == 0) |
81e84650 | 1510 | return true; |
b411b363 | 1511 | |
10f6d992 LE |
1512 | /* don't care for the reason here */ |
1513 | dev_err(DEV, "submit failed, triggering re-connect\n"); | |
87eeee41 | 1514 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 1515 | list_del(&peer_req->w.list); |
87eeee41 | 1516 | spin_unlock_irq(&mdev->tconn->req_lock); |
22cc37a9 | 1517 | |
db830c46 | 1518 | drbd_free_ee(mdev, peer_req); |
45bb912b LE |
1519 | fail: |
1520 | put_ldev(mdev); | |
81e84650 | 1521 | return false; |
b411b363 PR |
1522 | } |
1523 | ||
668eebc6 | 1524 | static struct drbd_request * |
bc9c5c41 AG |
1525 | find_request(struct drbd_conf *mdev, struct rb_root *root, u64 id, |
1526 | sector_t sector, bool missing_ok, const char *func) | |
51624585 | 1527 | { |
51624585 AG |
1528 | struct drbd_request *req; |
1529 | ||
bc9c5c41 AG |
1530 | /* Request object according to our peer */ |
1531 | req = (struct drbd_request *)(unsigned long)id; | |
5e472264 | 1532 | if (drbd_contains_interval(root, sector, &req->i) && req->i.local) |
668eebc6 | 1533 | return req; |
c3afd8f5 AG |
1534 | if (!missing_ok) { |
1535 | dev_err(DEV, "%s: failed to find request %lu, sector %llus\n", func, | |
1536 | (unsigned long)id, (unsigned long long)sector); | |
1537 | } | |
51624585 AG |
1538 | return NULL; |
1539 | } | |
1540 | ||
d8763023 AG |
1541 | static int receive_DataReply(struct drbd_conf *mdev, enum drbd_packet cmd, |
1542 | unsigned int data_size) | |
b411b363 PR |
1543 | { |
1544 | struct drbd_request *req; | |
1545 | sector_t sector; | |
b411b363 | 1546 | int ok; |
e42325a5 | 1547 | struct p_data *p = &mdev->tconn->data.rbuf.data; |
b411b363 PR |
1548 | |
1549 | sector = be64_to_cpu(p->sector); | |
1550 | ||
87eeee41 | 1551 | spin_lock_irq(&mdev->tconn->req_lock); |
bc9c5c41 | 1552 | req = find_request(mdev, &mdev->read_requests, p->block_id, sector, false, __func__); |
87eeee41 | 1553 | spin_unlock_irq(&mdev->tconn->req_lock); |
c3afd8f5 | 1554 | if (unlikely(!req)) |
81e84650 | 1555 | return false; |
b411b363 | 1556 | |
24c4830c | 1557 | /* hlist_del(&req->collision) is done in _req_may_be_done, to avoid |
b411b363 PR |
1558 | * special casing it there for the various failure cases. |
1559 | * still no race with drbd_fail_pending_reads */ | |
1560 | ok = recv_dless_read(mdev, req, sector, data_size); | |
1561 | ||
1562 | if (ok) | |
8554df1c | 1563 | req_mod(req, DATA_RECEIVED); |
b411b363 PR |
1564 | /* else: nothing. handled from drbd_disconnect... |
1565 | * I don't think we may complete this just yet | |
1566 | * in case we are "on-disconnect: freeze" */ | |
1567 | ||
1568 | return ok; | |
1569 | } | |
1570 | ||
d8763023 AG |
1571 | static int receive_RSDataReply(struct drbd_conf *mdev, enum drbd_packet cmd, |
1572 | unsigned int data_size) | |
b411b363 PR |
1573 | { |
1574 | sector_t sector; | |
b411b363 | 1575 | int ok; |
e42325a5 | 1576 | struct p_data *p = &mdev->tconn->data.rbuf.data; |
b411b363 PR |
1577 | |
1578 | sector = be64_to_cpu(p->sector); | |
1579 | D_ASSERT(p->block_id == ID_SYNCER); | |
1580 | ||
1581 | if (get_ldev(mdev)) { | |
1582 | /* data is submitted to disk within recv_resync_read. | |
1583 | * corresponding put_ldev done below on error, | |
fcefa62e | 1584 | * or in drbd_peer_request_endio. */ |
b411b363 PR |
1585 | ok = recv_resync_read(mdev, sector, data_size); |
1586 | } else { | |
1587 | if (__ratelimit(&drbd_ratelimit_state)) | |
1588 | dev_err(DEV, "Can not write resync data to local disk.\n"); | |
1589 | ||
1590 | ok = drbd_drain_block(mdev, data_size); | |
1591 | ||
2b2bf214 | 1592 | drbd_send_ack_dp(mdev, P_NEG_ACK, p, data_size); |
b411b363 PR |
1593 | } |
1594 | ||
778f271d PR |
1595 | atomic_add(data_size >> 9, &mdev->rs_sect_in); |
1596 | ||
b411b363 PR |
1597 | return ok; |
1598 | } | |
1599 | ||
7be8da07 AG |
1600 | static int w_restart_write(struct drbd_work *w, int cancel) |
1601 | { | |
1602 | struct drbd_request *req = container_of(w, struct drbd_request, w); | |
1603 | struct drbd_conf *mdev = w->mdev; | |
1604 | struct bio *bio; | |
1605 | unsigned long start_time; | |
1606 | unsigned long flags; | |
1607 | ||
1608 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); | |
1609 | if (!expect(req->rq_state & RQ_POSTPONED)) { | |
1610 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
1611 | return 0; | |
1612 | } | |
1613 | bio = req->master_bio; | |
1614 | start_time = req->start_time; | |
1615 | /* Postponed requests will not have their master_bio completed! */ | |
1616 | __req_mod(req, DISCARD_WRITE, NULL); | |
1617 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
1618 | ||
1619 | while (__drbd_make_request(mdev, bio, start_time)) | |
1620 | /* retry */ ; | |
1621 | return 1; | |
1622 | } | |
1623 | ||
1624 | static void restart_conflicting_writes(struct drbd_conf *mdev, | |
1625 | sector_t sector, int size) | |
1626 | { | |
1627 | struct drbd_interval *i; | |
1628 | struct drbd_request *req; | |
1629 | ||
1630 | drbd_for_each_overlap(i, &mdev->write_requests, sector, size) { | |
1631 | if (!i->local) | |
1632 | continue; | |
1633 | req = container_of(i, struct drbd_request, i); | |
1634 | if (req->rq_state & RQ_LOCAL_PENDING || | |
1635 | !(req->rq_state & RQ_POSTPONED)) | |
1636 | continue; | |
1637 | if (expect(list_empty(&req->w.list))) { | |
1638 | req->w.mdev = mdev; | |
1639 | req->w.cb = w_restart_write; | |
1640 | drbd_queue_work(&mdev->tconn->data.work, &req->w); | |
1641 | } | |
1642 | } | |
1643 | } | |
1644 | ||
b411b363 PR |
1645 | /* e_end_block() is called via drbd_process_done_ee(). |
1646 | * this means this function only runs in the asender thread | |
1647 | */ | |
00d56944 | 1648 | static int e_end_block(struct drbd_work *w, int cancel) |
b411b363 | 1649 | { |
8050e6d0 AG |
1650 | struct drbd_peer_request *peer_req = |
1651 | container_of(w, struct drbd_peer_request, w); | |
00d56944 | 1652 | struct drbd_conf *mdev = w->mdev; |
db830c46 | 1653 | sector_t sector = peer_req->i.sector; |
b411b363 PR |
1654 | int ok = 1, pcmd; |
1655 | ||
89e58e75 | 1656 | if (mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C) { |
db830c46 | 1657 | if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
1658 | pcmd = (mdev->state.conn >= C_SYNC_SOURCE && |
1659 | mdev->state.conn <= C_PAUSED_SYNC_T && | |
db830c46 | 1660 | peer_req->flags & EE_MAY_SET_IN_SYNC) ? |
b411b363 | 1661 | P_RS_WRITE_ACK : P_WRITE_ACK; |
db830c46 | 1662 | ok &= drbd_send_ack(mdev, pcmd, peer_req); |
b411b363 | 1663 | if (pcmd == P_RS_WRITE_ACK) |
db830c46 | 1664 | drbd_set_in_sync(mdev, sector, peer_req->i.size); |
b411b363 | 1665 | } else { |
db830c46 | 1666 | ok = drbd_send_ack(mdev, P_NEG_ACK, peer_req); |
b411b363 PR |
1667 | /* we expect it to be marked out of sync anyways... |
1668 | * maybe assert this? */ | |
1669 | } | |
1670 | dec_unacked(mdev); | |
1671 | } | |
1672 | /* we delete from the conflict detection hash _after_ we sent out the | |
1673 | * P_WRITE_ACK / P_NEG_ACK, to get the sequence number right. */ | |
89e58e75 | 1674 | if (mdev->tconn->net_conf->two_primaries) { |
87eeee41 | 1675 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 AG |
1676 | D_ASSERT(!drbd_interval_empty(&peer_req->i)); |
1677 | drbd_remove_epoch_entry_interval(mdev, peer_req); | |
7be8da07 AG |
1678 | if (peer_req->flags & EE_RESTART_REQUESTS) |
1679 | restart_conflicting_writes(mdev, sector, peer_req->i.size); | |
87eeee41 | 1680 | spin_unlock_irq(&mdev->tconn->req_lock); |
bb3bfe96 | 1681 | } else |
db830c46 | 1682 | D_ASSERT(drbd_interval_empty(&peer_req->i)); |
b411b363 | 1683 | |
db830c46 | 1684 | drbd_may_finish_epoch(mdev, peer_req->epoch, EV_PUT + (cancel ? EV_CLEANUP : 0)); |
b411b363 PR |
1685 | |
1686 | return ok; | |
1687 | } | |
1688 | ||
7be8da07 | 1689 | static int e_send_ack(struct drbd_work *w, enum drbd_packet ack) |
b411b363 | 1690 | { |
7be8da07 | 1691 | struct drbd_conf *mdev = w->mdev; |
8050e6d0 AG |
1692 | struct drbd_peer_request *peer_req = |
1693 | container_of(w, struct drbd_peer_request, w); | |
206d3589 | 1694 | int ok; |
b411b363 | 1695 | |
7be8da07 | 1696 | ok = drbd_send_ack(mdev, ack, peer_req); |
b411b363 PR |
1697 | dec_unacked(mdev); |
1698 | ||
1699 | return ok; | |
1700 | } | |
1701 | ||
7be8da07 AG |
1702 | static int e_send_discard_write(struct drbd_work *w, int unused) |
1703 | { | |
1704 | return e_send_ack(w, P_DISCARD_WRITE); | |
1705 | } | |
1706 | ||
1707 | static int e_send_retry_write(struct drbd_work *w, int unused) | |
1708 | { | |
1709 | struct drbd_tconn *tconn = w->mdev->tconn; | |
1710 | ||
1711 | return e_send_ack(w, tconn->agreed_pro_version >= 100 ? | |
1712 | P_RETRY_WRITE : P_DISCARD_WRITE); | |
1713 | } | |
1714 | ||
3e394da1 AG |
1715 | static bool seq_greater(u32 a, u32 b) |
1716 | { | |
1717 | /* | |
1718 | * We assume 32-bit wrap-around here. | |
1719 | * For 24-bit wrap-around, we would have to shift: | |
1720 | * a <<= 8; b <<= 8; | |
1721 | */ | |
1722 | return (s32)a - (s32)b > 0; | |
1723 | } | |
1724 | ||
1725 | static u32 seq_max(u32 a, u32 b) | |
1726 | { | |
1727 | return seq_greater(a, b) ? a : b; | |
1728 | } | |
1729 | ||
7be8da07 AG |
1730 | static bool need_peer_seq(struct drbd_conf *mdev) |
1731 | { | |
1732 | struct drbd_tconn *tconn = mdev->tconn; | |
1733 | ||
1734 | /* | |
1735 | * We only need to keep track of the last packet_seq number of our peer | |
1736 | * if we are in dual-primary mode and we have the discard flag set; see | |
1737 | * handle_write_conflicts(). | |
1738 | */ | |
1739 | return tconn->net_conf->two_primaries && | |
1740 | test_bit(DISCARD_CONCURRENT, &tconn->flags); | |
1741 | } | |
1742 | ||
43ae077d | 1743 | static void update_peer_seq(struct drbd_conf *mdev, unsigned int peer_seq) |
3e394da1 | 1744 | { |
3c13b680 | 1745 | unsigned int newest_peer_seq; |
3e394da1 | 1746 | |
7be8da07 AG |
1747 | if (need_peer_seq(mdev)) { |
1748 | spin_lock(&mdev->peer_seq_lock); | |
3c13b680 LE |
1749 | newest_peer_seq = seq_max(mdev->peer_seq, peer_seq); |
1750 | mdev->peer_seq = newest_peer_seq; | |
7be8da07 | 1751 | spin_unlock(&mdev->peer_seq_lock); |
3c13b680 LE |
1752 | /* wake up only if we actually changed mdev->peer_seq */ |
1753 | if (peer_seq == newest_peer_seq) | |
7be8da07 AG |
1754 | wake_up(&mdev->seq_wait); |
1755 | } | |
3e394da1 AG |
1756 | } |
1757 | ||
b411b363 PR |
1758 | /* Called from receive_Data. |
1759 | * Synchronize packets on sock with packets on msock. | |
1760 | * | |
1761 | * This is here so even when a P_DATA packet traveling via sock overtook an Ack | |
1762 | * packet traveling on msock, they are still processed in the order they have | |
1763 | * been sent. | |
1764 | * | |
1765 | * Note: we don't care for Ack packets overtaking P_DATA packets. | |
1766 | * | |
1767 | * In case packet_seq is larger than mdev->peer_seq number, there are | |
1768 | * outstanding packets on the msock. We wait for them to arrive. | |
1769 | * In case we are the logically next packet, we update mdev->peer_seq | |
1770 | * ourselves. Correctly handles 32bit wrap around. | |
1771 | * | |
1772 | * Assume we have a 10 GBit connection, that is about 1<<30 byte per second, | |
1773 | * about 1<<21 sectors per second. So "worst" case, we have 1<<3 == 8 seconds | |
1774 | * for the 24bit wrap (historical atomic_t guarantee on some archs), and we have | |
1775 | * 1<<9 == 512 seconds aka ages for the 32bit wrap around... | |
1776 | * | |
1777 | * returns 0 if we may process the packet, | |
1778 | * -ERESTARTSYS if we were interrupted (by disconnect signal). */ | |
7be8da07 | 1779 | static int wait_for_and_update_peer_seq(struct drbd_conf *mdev, const u32 peer_seq) |
b411b363 PR |
1780 | { |
1781 | DEFINE_WAIT(wait); | |
b411b363 | 1782 | long timeout; |
7be8da07 AG |
1783 | int ret; |
1784 | ||
1785 | if (!need_peer_seq(mdev)) | |
1786 | return 0; | |
1787 | ||
b411b363 PR |
1788 | spin_lock(&mdev->peer_seq_lock); |
1789 | for (;;) { | |
7be8da07 AG |
1790 | if (!seq_greater(peer_seq - 1, mdev->peer_seq)) { |
1791 | mdev->peer_seq = seq_max(mdev->peer_seq, peer_seq); | |
1792 | ret = 0; | |
b411b363 | 1793 | break; |
7be8da07 | 1794 | } |
b411b363 PR |
1795 | if (signal_pending(current)) { |
1796 | ret = -ERESTARTSYS; | |
1797 | break; | |
1798 | } | |
7be8da07 | 1799 | prepare_to_wait(&mdev->seq_wait, &wait, TASK_INTERRUPTIBLE); |
b411b363 | 1800 | spin_unlock(&mdev->peer_seq_lock); |
71b1c1eb AG |
1801 | timeout = mdev->tconn->net_conf->ping_timeo*HZ/10; |
1802 | timeout = schedule_timeout(timeout); | |
b411b363 | 1803 | spin_lock(&mdev->peer_seq_lock); |
7be8da07 | 1804 | if (!timeout) { |
b411b363 | 1805 | ret = -ETIMEDOUT; |
71b1c1eb | 1806 | dev_err(DEV, "Timed out waiting for missing ack packets; disconnecting\n"); |
b411b363 PR |
1807 | break; |
1808 | } | |
1809 | } | |
b411b363 | 1810 | spin_unlock(&mdev->peer_seq_lock); |
7be8da07 | 1811 | finish_wait(&mdev->seq_wait, &wait); |
b411b363 PR |
1812 | return ret; |
1813 | } | |
1814 | ||
688593c5 LE |
1815 | /* see also bio_flags_to_wire() |
1816 | * DRBD_REQ_*, because we need to semantically map the flags to data packet | |
1817 | * flags and back. We may replicate to other kernel versions. */ | |
1818 | static unsigned long wire_flags_to_bio(struct drbd_conf *mdev, u32 dpf) | |
76d2e7ec | 1819 | { |
688593c5 LE |
1820 | return (dpf & DP_RW_SYNC ? REQ_SYNC : 0) | |
1821 | (dpf & DP_FUA ? REQ_FUA : 0) | | |
1822 | (dpf & DP_FLUSH ? REQ_FLUSH : 0) | | |
1823 | (dpf & DP_DISCARD ? REQ_DISCARD : 0); | |
76d2e7ec PR |
1824 | } |
1825 | ||
7be8da07 AG |
1826 | static void fail_postponed_requests(struct drbd_conf *mdev, sector_t sector, |
1827 | unsigned int size) | |
1828 | { | |
1829 | struct drbd_interval *i; | |
1830 | ||
1831 | repeat: | |
1832 | drbd_for_each_overlap(i, &mdev->write_requests, sector, size) { | |
1833 | struct drbd_request *req; | |
1834 | struct bio_and_error m; | |
1835 | ||
1836 | if (!i->local) | |
1837 | continue; | |
1838 | req = container_of(i, struct drbd_request, i); | |
1839 | if (!(req->rq_state & RQ_POSTPONED)) | |
1840 | continue; | |
1841 | req->rq_state &= ~RQ_POSTPONED; | |
1842 | __req_mod(req, NEG_ACKED, &m); | |
1843 | spin_unlock_irq(&mdev->tconn->req_lock); | |
1844 | if (m.bio) | |
1845 | complete_master_bio(mdev, &m); | |
1846 | spin_lock_irq(&mdev->tconn->req_lock); | |
1847 | goto repeat; | |
1848 | } | |
1849 | } | |
1850 | ||
1851 | static int handle_write_conflicts(struct drbd_conf *mdev, | |
1852 | struct drbd_peer_request *peer_req) | |
1853 | { | |
1854 | struct drbd_tconn *tconn = mdev->tconn; | |
1855 | bool resolve_conflicts = test_bit(DISCARD_CONCURRENT, &tconn->flags); | |
1856 | sector_t sector = peer_req->i.sector; | |
1857 | const unsigned int size = peer_req->i.size; | |
1858 | struct drbd_interval *i; | |
1859 | bool equal; | |
1860 | int err; | |
1861 | ||
1862 | /* | |
1863 | * Inserting the peer request into the write_requests tree will prevent | |
1864 | * new conflicting local requests from being added. | |
1865 | */ | |
1866 | drbd_insert_interval(&mdev->write_requests, &peer_req->i); | |
1867 | ||
1868 | repeat: | |
1869 | drbd_for_each_overlap(i, &mdev->write_requests, sector, size) { | |
1870 | if (i == &peer_req->i) | |
1871 | continue; | |
1872 | ||
1873 | if (!i->local) { | |
1874 | /* | |
1875 | * Our peer has sent a conflicting remote request; this | |
1876 | * should not happen in a two-node setup. Wait for the | |
1877 | * earlier peer request to complete. | |
1878 | */ | |
1879 | err = drbd_wait_misc(mdev, i); | |
1880 | if (err) | |
1881 | goto out; | |
1882 | goto repeat; | |
1883 | } | |
1884 | ||
1885 | equal = i->sector == sector && i->size == size; | |
1886 | if (resolve_conflicts) { | |
1887 | /* | |
1888 | * If the peer request is fully contained within the | |
1889 | * overlapping request, it can be discarded; otherwise, | |
1890 | * it will be retried once all overlapping requests | |
1891 | * have completed. | |
1892 | */ | |
1893 | bool discard = i->sector <= sector && i->sector + | |
1894 | (i->size >> 9) >= sector + (size >> 9); | |
1895 | ||
1896 | if (!equal) | |
1897 | dev_alert(DEV, "Concurrent writes detected: " | |
1898 | "local=%llus +%u, remote=%llus +%u, " | |
1899 | "assuming %s came first\n", | |
1900 | (unsigned long long)i->sector, i->size, | |
1901 | (unsigned long long)sector, size, | |
1902 | discard ? "local" : "remote"); | |
1903 | ||
1904 | inc_unacked(mdev); | |
1905 | peer_req->w.cb = discard ? e_send_discard_write : | |
1906 | e_send_retry_write; | |
1907 | list_add_tail(&peer_req->w.list, &mdev->done_ee); | |
1908 | wake_asender(mdev->tconn); | |
1909 | ||
1910 | err = -ENOENT; | |
1911 | goto out; | |
1912 | } else { | |
1913 | struct drbd_request *req = | |
1914 | container_of(i, struct drbd_request, i); | |
1915 | ||
1916 | if (!equal) | |
1917 | dev_alert(DEV, "Concurrent writes detected: " | |
1918 | "local=%llus +%u, remote=%llus +%u\n", | |
1919 | (unsigned long long)i->sector, i->size, | |
1920 | (unsigned long long)sector, size); | |
1921 | ||
1922 | if (req->rq_state & RQ_LOCAL_PENDING || | |
1923 | !(req->rq_state & RQ_POSTPONED)) { | |
1924 | /* | |
1925 | * Wait for the node with the discard flag to | |
1926 | * decide if this request will be discarded or | |
1927 | * retried. Requests that are discarded will | |
1928 | * disappear from the write_requests tree. | |
1929 | * | |
1930 | * In addition, wait for the conflicting | |
1931 | * request to finish locally before submitting | |
1932 | * the conflicting peer request. | |
1933 | */ | |
1934 | err = drbd_wait_misc(mdev, &req->i); | |
1935 | if (err) { | |
1936 | _conn_request_state(mdev->tconn, | |
1937 | NS(conn, C_TIMEOUT), | |
1938 | CS_HARD); | |
1939 | fail_postponed_requests(mdev, sector, size); | |
1940 | goto out; | |
1941 | } | |
1942 | goto repeat; | |
1943 | } | |
1944 | /* | |
1945 | * Remember to restart the conflicting requests after | |
1946 | * the new peer request has completed. | |
1947 | */ | |
1948 | peer_req->flags |= EE_RESTART_REQUESTS; | |
1949 | } | |
1950 | } | |
1951 | err = 0; | |
1952 | ||
1953 | out: | |
1954 | if (err) | |
1955 | drbd_remove_epoch_entry_interval(mdev, peer_req); | |
1956 | return err; | |
1957 | } | |
1958 | ||
b411b363 | 1959 | /* mirrored write */ |
d8763023 AG |
1960 | static int receive_Data(struct drbd_conf *mdev, enum drbd_packet cmd, |
1961 | unsigned int data_size) | |
b411b363 PR |
1962 | { |
1963 | sector_t sector; | |
db830c46 | 1964 | struct drbd_peer_request *peer_req; |
e42325a5 | 1965 | struct p_data *p = &mdev->tconn->data.rbuf.data; |
7be8da07 | 1966 | u32 peer_seq = be32_to_cpu(p->seq_num); |
b411b363 PR |
1967 | int rw = WRITE; |
1968 | u32 dp_flags; | |
7be8da07 | 1969 | int err; |
b411b363 | 1970 | |
b411b363 | 1971 | |
7be8da07 AG |
1972 | if (!get_ldev(mdev)) { |
1973 | err = wait_for_and_update_peer_seq(mdev, peer_seq); | |
2b2bf214 | 1974 | drbd_send_ack_dp(mdev, P_NEG_ACK, p, data_size); |
b411b363 | 1975 | atomic_inc(&mdev->current_epoch->epoch_size); |
7be8da07 | 1976 | return drbd_drain_block(mdev, data_size) && err == 0; |
b411b363 PR |
1977 | } |
1978 | ||
fcefa62e AG |
1979 | /* |
1980 | * Corresponding put_ldev done either below (on various errors), or in | |
1981 | * drbd_peer_request_endio, if we successfully submit the data at the | |
1982 | * end of this function. | |
1983 | */ | |
b411b363 PR |
1984 | |
1985 | sector = be64_to_cpu(p->sector); | |
db830c46 AG |
1986 | peer_req = read_in_block(mdev, p->block_id, sector, data_size); |
1987 | if (!peer_req) { | |
b411b363 | 1988 | put_ldev(mdev); |
81e84650 | 1989 | return false; |
b411b363 PR |
1990 | } |
1991 | ||
db830c46 | 1992 | peer_req->w.cb = e_end_block; |
b411b363 | 1993 | |
688593c5 LE |
1994 | dp_flags = be32_to_cpu(p->dp_flags); |
1995 | rw |= wire_flags_to_bio(mdev, dp_flags); | |
1996 | ||
1997 | if (dp_flags & DP_MAY_SET_IN_SYNC) | |
db830c46 | 1998 | peer_req->flags |= EE_MAY_SET_IN_SYNC; |
688593c5 | 1999 | |
b411b363 | 2000 | spin_lock(&mdev->epoch_lock); |
db830c46 AG |
2001 | peer_req->epoch = mdev->current_epoch; |
2002 | atomic_inc(&peer_req->epoch->epoch_size); | |
2003 | atomic_inc(&peer_req->epoch->active); | |
b411b363 PR |
2004 | spin_unlock(&mdev->epoch_lock); |
2005 | ||
7be8da07 AG |
2006 | if (mdev->tconn->net_conf->two_primaries) { |
2007 | err = wait_for_and_update_peer_seq(mdev, peer_seq); | |
2008 | if (err) | |
b411b363 | 2009 | goto out_interrupted; |
87eeee41 | 2010 | spin_lock_irq(&mdev->tconn->req_lock); |
7be8da07 AG |
2011 | err = handle_write_conflicts(mdev, peer_req); |
2012 | if (err) { | |
2013 | spin_unlock_irq(&mdev->tconn->req_lock); | |
2014 | if (err == -ENOENT) { | |
b411b363 | 2015 | put_ldev(mdev); |
81e84650 | 2016 | return true; |
b411b363 | 2017 | } |
7be8da07 | 2018 | goto out_interrupted; |
b411b363 | 2019 | } |
7be8da07 AG |
2020 | } else |
2021 | spin_lock_irq(&mdev->tconn->req_lock); | |
db830c46 | 2022 | list_add(&peer_req->w.list, &mdev->active_ee); |
87eeee41 | 2023 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 2024 | |
89e58e75 | 2025 | switch (mdev->tconn->net_conf->wire_protocol) { |
b411b363 PR |
2026 | case DRBD_PROT_C: |
2027 | inc_unacked(mdev); | |
2028 | /* corresponding dec_unacked() in e_end_block() | |
2029 | * respective _drbd_clear_done_ee */ | |
2030 | break; | |
2031 | case DRBD_PROT_B: | |
2032 | /* I really don't like it that the receiver thread | |
2033 | * sends on the msock, but anyways */ | |
db830c46 | 2034 | drbd_send_ack(mdev, P_RECV_ACK, peer_req); |
b411b363 PR |
2035 | break; |
2036 | case DRBD_PROT_A: | |
2037 | /* nothing to do */ | |
2038 | break; | |
2039 | } | |
2040 | ||
6719fb03 | 2041 | if (mdev->state.pdsk < D_INCONSISTENT) { |
b411b363 | 2042 | /* In case we have the only disk of the cluster, */ |
db830c46 AG |
2043 | drbd_set_out_of_sync(mdev, peer_req->i.sector, peer_req->i.size); |
2044 | peer_req->flags |= EE_CALL_AL_COMPLETE_IO; | |
2045 | peer_req->flags &= ~EE_MAY_SET_IN_SYNC; | |
2046 | drbd_al_begin_io(mdev, peer_req->i.sector); | |
b411b363 PR |
2047 | } |
2048 | ||
fbe29dec | 2049 | if (drbd_submit_peer_request(mdev, peer_req, rw, DRBD_FAULT_DT_WR) == 0) |
81e84650 | 2050 | return true; |
b411b363 | 2051 | |
10f6d992 LE |
2052 | /* don't care for the reason here */ |
2053 | dev_err(DEV, "submit failed, triggering re-connect\n"); | |
87eeee41 | 2054 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 AG |
2055 | list_del(&peer_req->w.list); |
2056 | drbd_remove_epoch_entry_interval(mdev, peer_req); | |
87eeee41 | 2057 | spin_unlock_irq(&mdev->tconn->req_lock); |
db830c46 AG |
2058 | if (peer_req->flags & EE_CALL_AL_COMPLETE_IO) |
2059 | drbd_al_complete_io(mdev, peer_req->i.sector); | |
22cc37a9 | 2060 | |
b411b363 | 2061 | out_interrupted: |
db830c46 | 2062 | drbd_may_finish_epoch(mdev, peer_req->epoch, EV_PUT + EV_CLEANUP); |
b411b363 | 2063 | put_ldev(mdev); |
db830c46 | 2064 | drbd_free_ee(mdev, peer_req); |
81e84650 | 2065 | return false; |
b411b363 PR |
2066 | } |
2067 | ||
0f0601f4 LE |
2068 | /* We may throttle resync, if the lower device seems to be busy, |
2069 | * and current sync rate is above c_min_rate. | |
2070 | * | |
2071 | * To decide whether or not the lower device is busy, we use a scheme similar | |
2072 | * to MD RAID is_mddev_idle(): if the partition stats reveal "significant" | |
2073 | * (more than 64 sectors) of activity we cannot account for with our own resync | |
2074 | * activity, it obviously is "busy". | |
2075 | * | |
2076 | * The current sync rate used here uses only the most recent two step marks, | |
2077 | * to have a short time average so we can react faster. | |
2078 | */ | |
e3555d85 | 2079 | int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector) |
0f0601f4 LE |
2080 | { |
2081 | struct gendisk *disk = mdev->ldev->backing_bdev->bd_contains->bd_disk; | |
2082 | unsigned long db, dt, dbdt; | |
e3555d85 | 2083 | struct lc_element *tmp; |
0f0601f4 LE |
2084 | int curr_events; |
2085 | int throttle = 0; | |
2086 | ||
2087 | /* feature disabled? */ | |
f399002e | 2088 | if (mdev->ldev->dc.c_min_rate == 0) |
0f0601f4 LE |
2089 | return 0; |
2090 | ||
e3555d85 PR |
2091 | spin_lock_irq(&mdev->al_lock); |
2092 | tmp = lc_find(mdev->resync, BM_SECT_TO_EXT(sector)); | |
2093 | if (tmp) { | |
2094 | struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce); | |
2095 | if (test_bit(BME_PRIORITY, &bm_ext->flags)) { | |
2096 | spin_unlock_irq(&mdev->al_lock); | |
2097 | return 0; | |
2098 | } | |
2099 | /* Do not slow down if app IO is already waiting for this extent */ | |
2100 | } | |
2101 | spin_unlock_irq(&mdev->al_lock); | |
2102 | ||
0f0601f4 LE |
2103 | curr_events = (int)part_stat_read(&disk->part0, sectors[0]) + |
2104 | (int)part_stat_read(&disk->part0, sectors[1]) - | |
2105 | atomic_read(&mdev->rs_sect_ev); | |
e3555d85 | 2106 | |
0f0601f4 LE |
2107 | if (!mdev->rs_last_events || curr_events - mdev->rs_last_events > 64) { |
2108 | unsigned long rs_left; | |
2109 | int i; | |
2110 | ||
2111 | mdev->rs_last_events = curr_events; | |
2112 | ||
2113 | /* sync speed average over the last 2*DRBD_SYNC_MARK_STEP, | |
2114 | * approx. */ | |
2649f080 LE |
2115 | i = (mdev->rs_last_mark + DRBD_SYNC_MARKS-1) % DRBD_SYNC_MARKS; |
2116 | ||
2117 | if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T) | |
2118 | rs_left = mdev->ov_left; | |
2119 | else | |
2120 | rs_left = drbd_bm_total_weight(mdev) - mdev->rs_failed; | |
0f0601f4 LE |
2121 | |
2122 | dt = ((long)jiffies - (long)mdev->rs_mark_time[i]) / HZ; | |
2123 | if (!dt) | |
2124 | dt++; | |
2125 | db = mdev->rs_mark_left[i] - rs_left; | |
2126 | dbdt = Bit2KB(db/dt); | |
2127 | ||
f399002e | 2128 | if (dbdt > mdev->ldev->dc.c_min_rate) |
0f0601f4 LE |
2129 | throttle = 1; |
2130 | } | |
2131 | return throttle; | |
2132 | } | |
2133 | ||
2134 | ||
d8763023 AG |
2135 | static int receive_DataRequest(struct drbd_conf *mdev, enum drbd_packet cmd, |
2136 | unsigned int digest_size) | |
b411b363 PR |
2137 | { |
2138 | sector_t sector; | |
2139 | const sector_t capacity = drbd_get_capacity(mdev->this_bdev); | |
db830c46 | 2140 | struct drbd_peer_request *peer_req; |
b411b363 | 2141 | struct digest_info *di = NULL; |
b18b37be | 2142 | int size, verb; |
b411b363 | 2143 | unsigned int fault_type; |
e42325a5 | 2144 | struct p_block_req *p = &mdev->tconn->data.rbuf.block_req; |
b411b363 PR |
2145 | |
2146 | sector = be64_to_cpu(p->sector); | |
2147 | size = be32_to_cpu(p->blksize); | |
2148 | ||
c670a398 | 2149 | if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) { |
b411b363 PR |
2150 | dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, |
2151 | (unsigned long long)sector, size); | |
81e84650 | 2152 | return false; |
b411b363 PR |
2153 | } |
2154 | if (sector + (size>>9) > capacity) { | |
2155 | dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, | |
2156 | (unsigned long long)sector, size); | |
81e84650 | 2157 | return false; |
b411b363 PR |
2158 | } |
2159 | ||
2160 | if (!get_ldev_if_state(mdev, D_UP_TO_DATE)) { | |
b18b37be PR |
2161 | verb = 1; |
2162 | switch (cmd) { | |
2163 | case P_DATA_REQUEST: | |
2164 | drbd_send_ack_rp(mdev, P_NEG_DREPLY, p); | |
2165 | break; | |
2166 | case P_RS_DATA_REQUEST: | |
2167 | case P_CSUM_RS_REQUEST: | |
2168 | case P_OV_REQUEST: | |
2169 | drbd_send_ack_rp(mdev, P_NEG_RS_DREPLY , p); | |
2170 | break; | |
2171 | case P_OV_REPLY: | |
2172 | verb = 0; | |
2173 | dec_rs_pending(mdev); | |
2174 | drbd_send_ack_ex(mdev, P_OV_RESULT, sector, size, ID_IN_SYNC); | |
2175 | break; | |
2176 | default: | |
2177 | dev_err(DEV, "unexpected command (%s) in receive_DataRequest\n", | |
2178 | cmdname(cmd)); | |
2179 | } | |
2180 | if (verb && __ratelimit(&drbd_ratelimit_state)) | |
b411b363 PR |
2181 | dev_err(DEV, "Can not satisfy peer's read request, " |
2182 | "no local data.\n"); | |
b18b37be | 2183 | |
a821cc4a LE |
2184 | /* drain possibly payload */ |
2185 | return drbd_drain_block(mdev, digest_size); | |
b411b363 PR |
2186 | } |
2187 | ||
2188 | /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD | |
2189 | * "criss-cross" setup, that might cause write-out on some other DRBD, | |
2190 | * which in turn might block on the other node at this very place. */ | |
db830c46 AG |
2191 | peer_req = drbd_alloc_ee(mdev, p->block_id, sector, size, GFP_NOIO); |
2192 | if (!peer_req) { | |
b411b363 | 2193 | put_ldev(mdev); |
81e84650 | 2194 | return false; |
b411b363 PR |
2195 | } |
2196 | ||
02918be2 | 2197 | switch (cmd) { |
b411b363 | 2198 | case P_DATA_REQUEST: |
db830c46 | 2199 | peer_req->w.cb = w_e_end_data_req; |
b411b363 | 2200 | fault_type = DRBD_FAULT_DT_RD; |
80a40e43 LE |
2201 | /* application IO, don't drbd_rs_begin_io */ |
2202 | goto submit; | |
2203 | ||
b411b363 | 2204 | case P_RS_DATA_REQUEST: |
db830c46 | 2205 | peer_req->w.cb = w_e_end_rsdata_req; |
b411b363 | 2206 | fault_type = DRBD_FAULT_RS_RD; |
5f9915bb LE |
2207 | /* used in the sector offset progress display */ |
2208 | mdev->bm_resync_fo = BM_SECT_TO_BIT(sector); | |
b411b363 PR |
2209 | break; |
2210 | ||
2211 | case P_OV_REPLY: | |
2212 | case P_CSUM_RS_REQUEST: | |
2213 | fault_type = DRBD_FAULT_RS_RD; | |
b411b363 PR |
2214 | di = kmalloc(sizeof(*di) + digest_size, GFP_NOIO); |
2215 | if (!di) | |
2216 | goto out_free_e; | |
2217 | ||
2218 | di->digest_size = digest_size; | |
2219 | di->digest = (((char *)di)+sizeof(struct digest_info)); | |
2220 | ||
db830c46 AG |
2221 | peer_req->digest = di; |
2222 | peer_req->flags |= EE_HAS_DIGEST; | |
c36c3ced | 2223 | |
de0ff338 | 2224 | if (drbd_recv(mdev->tconn, di->digest, digest_size) != digest_size) |
b411b363 PR |
2225 | goto out_free_e; |
2226 | ||
02918be2 | 2227 | if (cmd == P_CSUM_RS_REQUEST) { |
31890f4a | 2228 | D_ASSERT(mdev->tconn->agreed_pro_version >= 89); |
db830c46 | 2229 | peer_req->w.cb = w_e_end_csum_rs_req; |
5f9915bb LE |
2230 | /* used in the sector offset progress display */ |
2231 | mdev->bm_resync_fo = BM_SECT_TO_BIT(sector); | |
02918be2 | 2232 | } else if (cmd == P_OV_REPLY) { |
2649f080 LE |
2233 | /* track progress, we may need to throttle */ |
2234 | atomic_add(size >> 9, &mdev->rs_sect_in); | |
db830c46 | 2235 | peer_req->w.cb = w_e_end_ov_reply; |
b411b363 | 2236 | dec_rs_pending(mdev); |
0f0601f4 LE |
2237 | /* drbd_rs_begin_io done when we sent this request, |
2238 | * but accounting still needs to be done. */ | |
2239 | goto submit_for_resync; | |
b411b363 PR |
2240 | } |
2241 | break; | |
2242 | ||
2243 | case P_OV_REQUEST: | |
b411b363 | 2244 | if (mdev->ov_start_sector == ~(sector_t)0 && |
31890f4a | 2245 | mdev->tconn->agreed_pro_version >= 90) { |
de228bba LE |
2246 | unsigned long now = jiffies; |
2247 | int i; | |
b411b363 PR |
2248 | mdev->ov_start_sector = sector; |
2249 | mdev->ov_position = sector; | |
30b743a2 LE |
2250 | mdev->ov_left = drbd_bm_bits(mdev) - BM_SECT_TO_BIT(sector); |
2251 | mdev->rs_total = mdev->ov_left; | |
de228bba LE |
2252 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { |
2253 | mdev->rs_mark_left[i] = mdev->ov_left; | |
2254 | mdev->rs_mark_time[i] = now; | |
2255 | } | |
b411b363 PR |
2256 | dev_info(DEV, "Online Verify start sector: %llu\n", |
2257 | (unsigned long long)sector); | |
2258 | } | |
db830c46 | 2259 | peer_req->w.cb = w_e_end_ov_req; |
b411b363 | 2260 | fault_type = DRBD_FAULT_RS_RD; |
b411b363 PR |
2261 | break; |
2262 | ||
b411b363 PR |
2263 | default: |
2264 | dev_err(DEV, "unexpected command (%s) in receive_DataRequest\n", | |
02918be2 | 2265 | cmdname(cmd)); |
b411b363 | 2266 | fault_type = DRBD_FAULT_MAX; |
80a40e43 | 2267 | goto out_free_e; |
b411b363 PR |
2268 | } |
2269 | ||
0f0601f4 LE |
2270 | /* Throttle, drbd_rs_begin_io and submit should become asynchronous |
2271 | * wrt the receiver, but it is not as straightforward as it may seem. | |
2272 | * Various places in the resync start and stop logic assume resync | |
2273 | * requests are processed in order, requeuing this on the worker thread | |
2274 | * introduces a bunch of new code for synchronization between threads. | |
2275 | * | |
2276 | * Unlimited throttling before drbd_rs_begin_io may stall the resync | |
2277 | * "forever", throttling after drbd_rs_begin_io will lock that extent | |
2278 | * for application writes for the same time. For now, just throttle | |
2279 | * here, where the rest of the code expects the receiver to sleep for | |
2280 | * a while, anyways. | |
2281 | */ | |
2282 | ||
2283 | /* Throttle before drbd_rs_begin_io, as that locks out application IO; | |
2284 | * this defers syncer requests for some time, before letting at least | |
2285 | * on request through. The resync controller on the receiving side | |
2286 | * will adapt to the incoming rate accordingly. | |
2287 | * | |
2288 | * We cannot throttle here if remote is Primary/SyncTarget: | |
2289 | * we would also throttle its application reads. | |
2290 | * In that case, throttling is done on the SyncTarget only. | |
2291 | */ | |
e3555d85 PR |
2292 | if (mdev->state.peer != R_PRIMARY && drbd_rs_should_slow_down(mdev, sector)) |
2293 | schedule_timeout_uninterruptible(HZ/10); | |
2294 | if (drbd_rs_begin_io(mdev, sector)) | |
80a40e43 | 2295 | goto out_free_e; |
b411b363 | 2296 | |
0f0601f4 LE |
2297 | submit_for_resync: |
2298 | atomic_add(size >> 9, &mdev->rs_sect_ev); | |
2299 | ||
80a40e43 | 2300 | submit: |
b411b363 | 2301 | inc_unacked(mdev); |
87eeee41 | 2302 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 2303 | list_add_tail(&peer_req->w.list, &mdev->read_ee); |
87eeee41 | 2304 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 2305 | |
fbe29dec | 2306 | if (drbd_submit_peer_request(mdev, peer_req, READ, fault_type) == 0) |
81e84650 | 2307 | return true; |
b411b363 | 2308 | |
10f6d992 LE |
2309 | /* don't care for the reason here */ |
2310 | dev_err(DEV, "submit failed, triggering re-connect\n"); | |
87eeee41 | 2311 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 2312 | list_del(&peer_req->w.list); |
87eeee41 | 2313 | spin_unlock_irq(&mdev->tconn->req_lock); |
22cc37a9 LE |
2314 | /* no drbd_rs_complete_io(), we are dropping the connection anyways */ |
2315 | ||
b411b363 | 2316 | out_free_e: |
b411b363 | 2317 | put_ldev(mdev); |
db830c46 | 2318 | drbd_free_ee(mdev, peer_req); |
81e84650 | 2319 | return false; |
b411b363 PR |
2320 | } |
2321 | ||
2322 | static int drbd_asb_recover_0p(struct drbd_conf *mdev) __must_hold(local) | |
2323 | { | |
2324 | int self, peer, rv = -100; | |
2325 | unsigned long ch_self, ch_peer; | |
2326 | ||
2327 | self = mdev->ldev->md.uuid[UI_BITMAP] & 1; | |
2328 | peer = mdev->p_uuid[UI_BITMAP] & 1; | |
2329 | ||
2330 | ch_peer = mdev->p_uuid[UI_SIZE]; | |
2331 | ch_self = mdev->comm_bm_set; | |
2332 | ||
89e58e75 | 2333 | switch (mdev->tconn->net_conf->after_sb_0p) { |
b411b363 PR |
2334 | case ASB_CONSENSUS: |
2335 | case ASB_DISCARD_SECONDARY: | |
2336 | case ASB_CALL_HELPER: | |
2337 | dev_err(DEV, "Configuration error.\n"); | |
2338 | break; | |
2339 | case ASB_DISCONNECT: | |
2340 | break; | |
2341 | case ASB_DISCARD_YOUNGER_PRI: | |
2342 | if (self == 0 && peer == 1) { | |
2343 | rv = -1; | |
2344 | break; | |
2345 | } | |
2346 | if (self == 1 && peer == 0) { | |
2347 | rv = 1; | |
2348 | break; | |
2349 | } | |
2350 | /* Else fall through to one of the other strategies... */ | |
2351 | case ASB_DISCARD_OLDER_PRI: | |
2352 | if (self == 0 && peer == 1) { | |
2353 | rv = 1; | |
2354 | break; | |
2355 | } | |
2356 | if (self == 1 && peer == 0) { | |
2357 | rv = -1; | |
2358 | break; | |
2359 | } | |
2360 | /* Else fall through to one of the other strategies... */ | |
ad19bf6e | 2361 | dev_warn(DEV, "Discard younger/older primary did not find a decision\n" |
b411b363 PR |
2362 | "Using discard-least-changes instead\n"); |
2363 | case ASB_DISCARD_ZERO_CHG: | |
2364 | if (ch_peer == 0 && ch_self == 0) { | |
25703f83 | 2365 | rv = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) |
b411b363 PR |
2366 | ? -1 : 1; |
2367 | break; | |
2368 | } else { | |
2369 | if (ch_peer == 0) { rv = 1; break; } | |
2370 | if (ch_self == 0) { rv = -1; break; } | |
2371 | } | |
89e58e75 | 2372 | if (mdev->tconn->net_conf->after_sb_0p == ASB_DISCARD_ZERO_CHG) |
b411b363 PR |
2373 | break; |
2374 | case ASB_DISCARD_LEAST_CHG: | |
2375 | if (ch_self < ch_peer) | |
2376 | rv = -1; | |
2377 | else if (ch_self > ch_peer) | |
2378 | rv = 1; | |
2379 | else /* ( ch_self == ch_peer ) */ | |
2380 | /* Well, then use something else. */ | |
25703f83 | 2381 | rv = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) |
b411b363 PR |
2382 | ? -1 : 1; |
2383 | break; | |
2384 | case ASB_DISCARD_LOCAL: | |
2385 | rv = -1; | |
2386 | break; | |
2387 | case ASB_DISCARD_REMOTE: | |
2388 | rv = 1; | |
2389 | } | |
2390 | ||
2391 | return rv; | |
2392 | } | |
2393 | ||
2394 | static int drbd_asb_recover_1p(struct drbd_conf *mdev) __must_hold(local) | |
2395 | { | |
6184ea21 | 2396 | int hg, rv = -100; |
b411b363 | 2397 | |
89e58e75 | 2398 | switch (mdev->tconn->net_conf->after_sb_1p) { |
b411b363 PR |
2399 | case ASB_DISCARD_YOUNGER_PRI: |
2400 | case ASB_DISCARD_OLDER_PRI: | |
2401 | case ASB_DISCARD_LEAST_CHG: | |
2402 | case ASB_DISCARD_LOCAL: | |
2403 | case ASB_DISCARD_REMOTE: | |
2404 | dev_err(DEV, "Configuration error.\n"); | |
2405 | break; | |
2406 | case ASB_DISCONNECT: | |
2407 | break; | |
2408 | case ASB_CONSENSUS: | |
2409 | hg = drbd_asb_recover_0p(mdev); | |
2410 | if (hg == -1 && mdev->state.role == R_SECONDARY) | |
2411 | rv = hg; | |
2412 | if (hg == 1 && mdev->state.role == R_PRIMARY) | |
2413 | rv = hg; | |
2414 | break; | |
2415 | case ASB_VIOLENTLY: | |
2416 | rv = drbd_asb_recover_0p(mdev); | |
2417 | break; | |
2418 | case ASB_DISCARD_SECONDARY: | |
2419 | return mdev->state.role == R_PRIMARY ? 1 : -1; | |
2420 | case ASB_CALL_HELPER: | |
2421 | hg = drbd_asb_recover_0p(mdev); | |
2422 | if (hg == -1 && mdev->state.role == R_PRIMARY) { | |
bb437946 AG |
2423 | enum drbd_state_rv rv2; |
2424 | ||
2425 | drbd_set_role(mdev, R_SECONDARY, 0); | |
b411b363 PR |
2426 | /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE, |
2427 | * we might be here in C_WF_REPORT_PARAMS which is transient. | |
2428 | * we do not need to wait for the after state change work either. */ | |
bb437946 AG |
2429 | rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY)); |
2430 | if (rv2 != SS_SUCCESS) { | |
b411b363 PR |
2431 | drbd_khelper(mdev, "pri-lost-after-sb"); |
2432 | } else { | |
2433 | dev_warn(DEV, "Successfully gave up primary role.\n"); | |
2434 | rv = hg; | |
2435 | } | |
2436 | } else | |
2437 | rv = hg; | |
2438 | } | |
2439 | ||
2440 | return rv; | |
2441 | } | |
2442 | ||
2443 | static int drbd_asb_recover_2p(struct drbd_conf *mdev) __must_hold(local) | |
2444 | { | |
6184ea21 | 2445 | int hg, rv = -100; |
b411b363 | 2446 | |
89e58e75 | 2447 | switch (mdev->tconn->net_conf->after_sb_2p) { |
b411b363 PR |
2448 | case ASB_DISCARD_YOUNGER_PRI: |
2449 | case ASB_DISCARD_OLDER_PRI: | |
2450 | case ASB_DISCARD_LEAST_CHG: | |
2451 | case ASB_DISCARD_LOCAL: | |
2452 | case ASB_DISCARD_REMOTE: | |
2453 | case ASB_CONSENSUS: | |
2454 | case ASB_DISCARD_SECONDARY: | |
2455 | dev_err(DEV, "Configuration error.\n"); | |
2456 | break; | |
2457 | case ASB_VIOLENTLY: | |
2458 | rv = drbd_asb_recover_0p(mdev); | |
2459 | break; | |
2460 | case ASB_DISCONNECT: | |
2461 | break; | |
2462 | case ASB_CALL_HELPER: | |
2463 | hg = drbd_asb_recover_0p(mdev); | |
2464 | if (hg == -1) { | |
bb437946 AG |
2465 | enum drbd_state_rv rv2; |
2466 | ||
b411b363 PR |
2467 | /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE, |
2468 | * we might be here in C_WF_REPORT_PARAMS which is transient. | |
2469 | * we do not need to wait for the after state change work either. */ | |
bb437946 AG |
2470 | rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY)); |
2471 | if (rv2 != SS_SUCCESS) { | |
b411b363 PR |
2472 | drbd_khelper(mdev, "pri-lost-after-sb"); |
2473 | } else { | |
2474 | dev_warn(DEV, "Successfully gave up primary role.\n"); | |
2475 | rv = hg; | |
2476 | } | |
2477 | } else | |
2478 | rv = hg; | |
2479 | } | |
2480 | ||
2481 | return rv; | |
2482 | } | |
2483 | ||
2484 | static void drbd_uuid_dump(struct drbd_conf *mdev, char *text, u64 *uuid, | |
2485 | u64 bits, u64 flags) | |
2486 | { | |
2487 | if (!uuid) { | |
2488 | dev_info(DEV, "%s uuid info vanished while I was looking!\n", text); | |
2489 | return; | |
2490 | } | |
2491 | dev_info(DEV, "%s %016llX:%016llX:%016llX:%016llX bits:%llu flags:%llX\n", | |
2492 | text, | |
2493 | (unsigned long long)uuid[UI_CURRENT], | |
2494 | (unsigned long long)uuid[UI_BITMAP], | |
2495 | (unsigned long long)uuid[UI_HISTORY_START], | |
2496 | (unsigned long long)uuid[UI_HISTORY_END], | |
2497 | (unsigned long long)bits, | |
2498 | (unsigned long long)flags); | |
2499 | } | |
2500 | ||
2501 | /* | |
2502 | 100 after split brain try auto recover | |
2503 | 2 C_SYNC_SOURCE set BitMap | |
2504 | 1 C_SYNC_SOURCE use BitMap | |
2505 | 0 no Sync | |
2506 | -1 C_SYNC_TARGET use BitMap | |
2507 | -2 C_SYNC_TARGET set BitMap | |
2508 | -100 after split brain, disconnect | |
2509 | -1000 unrelated data | |
4a23f264 PR |
2510 | -1091 requires proto 91 |
2511 | -1096 requires proto 96 | |
b411b363 PR |
2512 | */ |
2513 | static int drbd_uuid_compare(struct drbd_conf *mdev, int *rule_nr) __must_hold(local) | |
2514 | { | |
2515 | u64 self, peer; | |
2516 | int i, j; | |
2517 | ||
2518 | self = mdev->ldev->md.uuid[UI_CURRENT] & ~((u64)1); | |
2519 | peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1); | |
2520 | ||
2521 | *rule_nr = 10; | |
2522 | if (self == UUID_JUST_CREATED && peer == UUID_JUST_CREATED) | |
2523 | return 0; | |
2524 | ||
2525 | *rule_nr = 20; | |
2526 | if ((self == UUID_JUST_CREATED || self == (u64)0) && | |
2527 | peer != UUID_JUST_CREATED) | |
2528 | return -2; | |
2529 | ||
2530 | *rule_nr = 30; | |
2531 | if (self != UUID_JUST_CREATED && | |
2532 | (peer == UUID_JUST_CREATED || peer == (u64)0)) | |
2533 | return 2; | |
2534 | ||
2535 | if (self == peer) { | |
2536 | int rct, dc; /* roles at crash time */ | |
2537 | ||
2538 | if (mdev->p_uuid[UI_BITMAP] == (u64)0 && mdev->ldev->md.uuid[UI_BITMAP] != (u64)0) { | |
2539 | ||
31890f4a | 2540 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2541 | return -1091; |
b411b363 PR |
2542 | |
2543 | if ((mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) && | |
2544 | (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1))) { | |
2545 | dev_info(DEV, "was SyncSource, missed the resync finished event, corrected myself:\n"); | |
2546 | drbd_uuid_set_bm(mdev, 0UL); | |
2547 | ||
2548 | drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, | |
2549 | mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0); | |
2550 | *rule_nr = 34; | |
2551 | } else { | |
2552 | dev_info(DEV, "was SyncSource (peer failed to write sync_uuid)\n"); | |
2553 | *rule_nr = 36; | |
2554 | } | |
2555 | ||
2556 | return 1; | |
2557 | } | |
2558 | ||
2559 | if (mdev->ldev->md.uuid[UI_BITMAP] == (u64)0 && mdev->p_uuid[UI_BITMAP] != (u64)0) { | |
2560 | ||
31890f4a | 2561 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2562 | return -1091; |
b411b363 PR |
2563 | |
2564 | if ((mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_BITMAP] & ~((u64)1)) && | |
2565 | (mdev->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1))) { | |
2566 | dev_info(DEV, "was SyncTarget, peer missed the resync finished event, corrected peer:\n"); | |
2567 | ||
2568 | mdev->p_uuid[UI_HISTORY_START + 1] = mdev->p_uuid[UI_HISTORY_START]; | |
2569 | mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_BITMAP]; | |
2570 | mdev->p_uuid[UI_BITMAP] = 0UL; | |
2571 | ||
2572 | drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]); | |
2573 | *rule_nr = 35; | |
2574 | } else { | |
2575 | dev_info(DEV, "was SyncTarget (failed to write sync_uuid)\n"); | |
2576 | *rule_nr = 37; | |
2577 | } | |
2578 | ||
2579 | return -1; | |
2580 | } | |
2581 | ||
2582 | /* Common power [off|failure] */ | |
2583 | rct = (test_bit(CRASHED_PRIMARY, &mdev->flags) ? 1 : 0) + | |
2584 | (mdev->p_uuid[UI_FLAGS] & 2); | |
2585 | /* lowest bit is set when we were primary, | |
2586 | * next bit (weight 2) is set when peer was primary */ | |
2587 | *rule_nr = 40; | |
2588 | ||
2589 | switch (rct) { | |
2590 | case 0: /* !self_pri && !peer_pri */ return 0; | |
2591 | case 1: /* self_pri && !peer_pri */ return 1; | |
2592 | case 2: /* !self_pri && peer_pri */ return -1; | |
2593 | case 3: /* self_pri && peer_pri */ | |
25703f83 | 2594 | dc = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags); |
b411b363 PR |
2595 | return dc ? -1 : 1; |
2596 | } | |
2597 | } | |
2598 | ||
2599 | *rule_nr = 50; | |
2600 | peer = mdev->p_uuid[UI_BITMAP] & ~((u64)1); | |
2601 | if (self == peer) | |
2602 | return -1; | |
2603 | ||
2604 | *rule_nr = 51; | |
2605 | peer = mdev->p_uuid[UI_HISTORY_START] & ~((u64)1); | |
2606 | if (self == peer) { | |
31890f4a | 2607 | if (mdev->tconn->agreed_pro_version < 96 ? |
4a23f264 PR |
2608 | (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == |
2609 | (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1)) : | |
2610 | peer + UUID_NEW_BM_OFFSET == (mdev->p_uuid[UI_BITMAP] & ~((u64)1))) { | |
b411b363 PR |
2611 | /* The last P_SYNC_UUID did not get though. Undo the last start of |
2612 | resync as sync source modifications of the peer's UUIDs. */ | |
2613 | ||
31890f4a | 2614 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2615 | return -1091; |
b411b363 PR |
2616 | |
2617 | mdev->p_uuid[UI_BITMAP] = mdev->p_uuid[UI_HISTORY_START]; | |
2618 | mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_HISTORY_START + 1]; | |
4a23f264 PR |
2619 | |
2620 | dev_info(DEV, "Did not got last syncUUID packet, corrected:\n"); | |
2621 | drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]); | |
2622 | ||
b411b363 PR |
2623 | return -1; |
2624 | } | |
2625 | } | |
2626 | ||
2627 | *rule_nr = 60; | |
2628 | self = mdev->ldev->md.uuid[UI_CURRENT] & ~((u64)1); | |
2629 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { | |
2630 | peer = mdev->p_uuid[i] & ~((u64)1); | |
2631 | if (self == peer) | |
2632 | return -2; | |
2633 | } | |
2634 | ||
2635 | *rule_nr = 70; | |
2636 | self = mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1); | |
2637 | peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1); | |
2638 | if (self == peer) | |
2639 | return 1; | |
2640 | ||
2641 | *rule_nr = 71; | |
2642 | self = mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1); | |
2643 | if (self == peer) { | |
31890f4a | 2644 | if (mdev->tconn->agreed_pro_version < 96 ? |
4a23f264 PR |
2645 | (mdev->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == |
2646 | (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) : | |
2647 | self + UUID_NEW_BM_OFFSET == (mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1))) { | |
b411b363 PR |
2648 | /* The last P_SYNC_UUID did not get though. Undo the last start of |
2649 | resync as sync source modifications of our UUIDs. */ | |
2650 | ||
31890f4a | 2651 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2652 | return -1091; |
b411b363 PR |
2653 | |
2654 | _drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_HISTORY_START]); | |
2655 | _drbd_uuid_set(mdev, UI_HISTORY_START, mdev->ldev->md.uuid[UI_HISTORY_START + 1]); | |
2656 | ||
4a23f264 | 2657 | dev_info(DEV, "Last syncUUID did not get through, corrected:\n"); |
b411b363 PR |
2658 | drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, |
2659 | mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0); | |
2660 | ||
2661 | return 1; | |
2662 | } | |
2663 | } | |
2664 | ||
2665 | ||
2666 | *rule_nr = 80; | |
d8c2a36b | 2667 | peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1); |
b411b363 PR |
2668 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { |
2669 | self = mdev->ldev->md.uuid[i] & ~((u64)1); | |
2670 | if (self == peer) | |
2671 | return 2; | |
2672 | } | |
2673 | ||
2674 | *rule_nr = 90; | |
2675 | self = mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1); | |
2676 | peer = mdev->p_uuid[UI_BITMAP] & ~((u64)1); | |
2677 | if (self == peer && self != ((u64)0)) | |
2678 | return 100; | |
2679 | ||
2680 | *rule_nr = 100; | |
2681 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { | |
2682 | self = mdev->ldev->md.uuid[i] & ~((u64)1); | |
2683 | for (j = UI_HISTORY_START; j <= UI_HISTORY_END; j++) { | |
2684 | peer = mdev->p_uuid[j] & ~((u64)1); | |
2685 | if (self == peer) | |
2686 | return -100; | |
2687 | } | |
2688 | } | |
2689 | ||
2690 | return -1000; | |
2691 | } | |
2692 | ||
2693 | /* drbd_sync_handshake() returns the new conn state on success, or | |
2694 | CONN_MASK (-1) on failure. | |
2695 | */ | |
2696 | static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_role peer_role, | |
2697 | enum drbd_disk_state peer_disk) __must_hold(local) | |
2698 | { | |
2699 | int hg, rule_nr; | |
2700 | enum drbd_conns rv = C_MASK; | |
2701 | enum drbd_disk_state mydisk; | |
2702 | ||
2703 | mydisk = mdev->state.disk; | |
2704 | if (mydisk == D_NEGOTIATING) | |
2705 | mydisk = mdev->new_state_tmp.disk; | |
2706 | ||
2707 | dev_info(DEV, "drbd_sync_handshake:\n"); | |
2708 | drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, mdev->comm_bm_set, 0); | |
2709 | drbd_uuid_dump(mdev, "peer", mdev->p_uuid, | |
2710 | mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]); | |
2711 | ||
2712 | hg = drbd_uuid_compare(mdev, &rule_nr); | |
2713 | ||
2714 | dev_info(DEV, "uuid_compare()=%d by rule %d\n", hg, rule_nr); | |
2715 | ||
2716 | if (hg == -1000) { | |
2717 | dev_alert(DEV, "Unrelated data, aborting!\n"); | |
2718 | return C_MASK; | |
2719 | } | |
4a23f264 PR |
2720 | if (hg < -1000) { |
2721 | dev_alert(DEV, "To resolve this both sides have to support at least protocol %d\n", -hg - 1000); | |
b411b363 PR |
2722 | return C_MASK; |
2723 | } | |
2724 | ||
2725 | if ((mydisk == D_INCONSISTENT && peer_disk > D_INCONSISTENT) || | |
2726 | (peer_disk == D_INCONSISTENT && mydisk > D_INCONSISTENT)) { | |
2727 | int f = (hg == -100) || abs(hg) == 2; | |
2728 | hg = mydisk > D_INCONSISTENT ? 1 : -1; | |
2729 | if (f) | |
2730 | hg = hg*2; | |
2731 | dev_info(DEV, "Becoming sync %s due to disk states.\n", | |
2732 | hg > 0 ? "source" : "target"); | |
2733 | } | |
2734 | ||
3a11a487 AG |
2735 | if (abs(hg) == 100) |
2736 | drbd_khelper(mdev, "initial-split-brain"); | |
2737 | ||
89e58e75 | 2738 | if (hg == 100 || (hg == -100 && mdev->tconn->net_conf->always_asbp)) { |
b411b363 PR |
2739 | int pcount = (mdev->state.role == R_PRIMARY) |
2740 | + (peer_role == R_PRIMARY); | |
2741 | int forced = (hg == -100); | |
2742 | ||
2743 | switch (pcount) { | |
2744 | case 0: | |
2745 | hg = drbd_asb_recover_0p(mdev); | |
2746 | break; | |
2747 | case 1: | |
2748 | hg = drbd_asb_recover_1p(mdev); | |
2749 | break; | |
2750 | case 2: | |
2751 | hg = drbd_asb_recover_2p(mdev); | |
2752 | break; | |
2753 | } | |
2754 | if (abs(hg) < 100) { | |
2755 | dev_warn(DEV, "Split-Brain detected, %d primaries, " | |
2756 | "automatically solved. Sync from %s node\n", | |
2757 | pcount, (hg < 0) ? "peer" : "this"); | |
2758 | if (forced) { | |
2759 | dev_warn(DEV, "Doing a full sync, since" | |
2760 | " UUIDs where ambiguous.\n"); | |
2761 | hg = hg*2; | |
2762 | } | |
2763 | } | |
2764 | } | |
2765 | ||
2766 | if (hg == -100) { | |
89e58e75 | 2767 | if (mdev->tconn->net_conf->want_lose && !(mdev->p_uuid[UI_FLAGS]&1)) |
b411b363 | 2768 | hg = -1; |
89e58e75 | 2769 | if (!mdev->tconn->net_conf->want_lose && (mdev->p_uuid[UI_FLAGS]&1)) |
b411b363 PR |
2770 | hg = 1; |
2771 | ||
2772 | if (abs(hg) < 100) | |
2773 | dev_warn(DEV, "Split-Brain detected, manually solved. " | |
2774 | "Sync from %s node\n", | |
2775 | (hg < 0) ? "peer" : "this"); | |
2776 | } | |
2777 | ||
2778 | if (hg == -100) { | |
580b9767 LE |
2779 | /* FIXME this log message is not correct if we end up here |
2780 | * after an attempted attach on a diskless node. | |
2781 | * We just refuse to attach -- well, we drop the "connection" | |
2782 | * to that disk, in a way... */ | |
3a11a487 | 2783 | dev_alert(DEV, "Split-Brain detected but unresolved, dropping connection!\n"); |
b411b363 PR |
2784 | drbd_khelper(mdev, "split-brain"); |
2785 | return C_MASK; | |
2786 | } | |
2787 | ||
2788 | if (hg > 0 && mydisk <= D_INCONSISTENT) { | |
2789 | dev_err(DEV, "I shall become SyncSource, but I am inconsistent!\n"); | |
2790 | return C_MASK; | |
2791 | } | |
2792 | ||
2793 | if (hg < 0 && /* by intention we do not use mydisk here. */ | |
2794 | mdev->state.role == R_PRIMARY && mdev->state.disk >= D_CONSISTENT) { | |
89e58e75 | 2795 | switch (mdev->tconn->net_conf->rr_conflict) { |
b411b363 PR |
2796 | case ASB_CALL_HELPER: |
2797 | drbd_khelper(mdev, "pri-lost"); | |
2798 | /* fall through */ | |
2799 | case ASB_DISCONNECT: | |
2800 | dev_err(DEV, "I shall become SyncTarget, but I am primary!\n"); | |
2801 | return C_MASK; | |
2802 | case ASB_VIOLENTLY: | |
2803 | dev_warn(DEV, "Becoming SyncTarget, violating the stable-data" | |
2804 | "assumption\n"); | |
2805 | } | |
2806 | } | |
2807 | ||
8169e41b | 2808 | if (mdev->tconn->net_conf->dry_run || test_bit(CONN_DRY_RUN, &mdev->tconn->flags)) { |
cf14c2e9 PR |
2809 | if (hg == 0) |
2810 | dev_info(DEV, "dry-run connect: No resync, would become Connected immediately.\n"); | |
2811 | else | |
2812 | dev_info(DEV, "dry-run connect: Would become %s, doing a %s resync.", | |
2813 | drbd_conn_str(hg > 0 ? C_SYNC_SOURCE : C_SYNC_TARGET), | |
2814 | abs(hg) >= 2 ? "full" : "bit-map based"); | |
2815 | return C_MASK; | |
2816 | } | |
2817 | ||
b411b363 PR |
2818 | if (abs(hg) >= 2) { |
2819 | dev_info(DEV, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n"); | |
20ceb2b2 LE |
2820 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from sync_handshake", |
2821 | BM_LOCKED_SET_ALLOWED)) | |
b411b363 PR |
2822 | return C_MASK; |
2823 | } | |
2824 | ||
2825 | if (hg > 0) { /* become sync source. */ | |
2826 | rv = C_WF_BITMAP_S; | |
2827 | } else if (hg < 0) { /* become sync target */ | |
2828 | rv = C_WF_BITMAP_T; | |
2829 | } else { | |
2830 | rv = C_CONNECTED; | |
2831 | if (drbd_bm_total_weight(mdev)) { | |
2832 | dev_info(DEV, "No resync, but %lu bits in bitmap!\n", | |
2833 | drbd_bm_total_weight(mdev)); | |
2834 | } | |
2835 | } | |
2836 | ||
2837 | return rv; | |
2838 | } | |
2839 | ||
2840 | /* returns 1 if invalid */ | |
2841 | static int cmp_after_sb(enum drbd_after_sb_p peer, enum drbd_after_sb_p self) | |
2842 | { | |
2843 | /* ASB_DISCARD_REMOTE - ASB_DISCARD_LOCAL is valid */ | |
2844 | if ((peer == ASB_DISCARD_REMOTE && self == ASB_DISCARD_LOCAL) || | |
2845 | (self == ASB_DISCARD_REMOTE && peer == ASB_DISCARD_LOCAL)) | |
2846 | return 0; | |
2847 | ||
2848 | /* any other things with ASB_DISCARD_REMOTE or ASB_DISCARD_LOCAL are invalid */ | |
2849 | if (peer == ASB_DISCARD_REMOTE || peer == ASB_DISCARD_LOCAL || | |
2850 | self == ASB_DISCARD_REMOTE || self == ASB_DISCARD_LOCAL) | |
2851 | return 1; | |
2852 | ||
2853 | /* everything else is valid if they are equal on both sides. */ | |
2854 | if (peer == self) | |
2855 | return 0; | |
2856 | ||
2857 | /* everything es is invalid. */ | |
2858 | return 1; | |
2859 | } | |
2860 | ||
7204624c | 2861 | static int receive_protocol(struct drbd_tconn *tconn, enum drbd_packet cmd, |
d8763023 | 2862 | unsigned int data_size) |
b411b363 | 2863 | { |
7204624c | 2864 | struct p_protocol *p = &tconn->data.rbuf.protocol; |
b411b363 | 2865 | int p_proto, p_after_sb_0p, p_after_sb_1p, p_after_sb_2p; |
cf14c2e9 | 2866 | int p_want_lose, p_two_primaries, cf; |
b411b363 PR |
2867 | char p_integrity_alg[SHARED_SECRET_MAX] = ""; |
2868 | ||
b411b363 PR |
2869 | p_proto = be32_to_cpu(p->protocol); |
2870 | p_after_sb_0p = be32_to_cpu(p->after_sb_0p); | |
2871 | p_after_sb_1p = be32_to_cpu(p->after_sb_1p); | |
2872 | p_after_sb_2p = be32_to_cpu(p->after_sb_2p); | |
b411b363 | 2873 | p_two_primaries = be32_to_cpu(p->two_primaries); |
cf14c2e9 PR |
2874 | cf = be32_to_cpu(p->conn_flags); |
2875 | p_want_lose = cf & CF_WANT_LOSE; | |
2876 | ||
7204624c | 2877 | clear_bit(CONN_DRY_RUN, &tconn->flags); |
cf14c2e9 PR |
2878 | |
2879 | if (cf & CF_DRY_RUN) | |
7204624c | 2880 | set_bit(CONN_DRY_RUN, &tconn->flags); |
b411b363 | 2881 | |
7204624c PR |
2882 | if (p_proto != tconn->net_conf->wire_protocol) { |
2883 | conn_err(tconn, "incompatible communication protocols\n"); | |
b411b363 PR |
2884 | goto disconnect; |
2885 | } | |
2886 | ||
7204624c PR |
2887 | if (cmp_after_sb(p_after_sb_0p, tconn->net_conf->after_sb_0p)) { |
2888 | conn_err(tconn, "incompatible after-sb-0pri settings\n"); | |
b411b363 PR |
2889 | goto disconnect; |
2890 | } | |
2891 | ||
7204624c PR |
2892 | if (cmp_after_sb(p_after_sb_1p, tconn->net_conf->after_sb_1p)) { |
2893 | conn_err(tconn, "incompatible after-sb-1pri settings\n"); | |
b411b363 PR |
2894 | goto disconnect; |
2895 | } | |
2896 | ||
7204624c PR |
2897 | if (cmp_after_sb(p_after_sb_2p, tconn->net_conf->after_sb_2p)) { |
2898 | conn_err(tconn, "incompatible after-sb-2pri settings\n"); | |
b411b363 PR |
2899 | goto disconnect; |
2900 | } | |
2901 | ||
7204624c PR |
2902 | if (p_want_lose && tconn->net_conf->want_lose) { |
2903 | conn_err(tconn, "both sides have the 'want_lose' flag set\n"); | |
b411b363 PR |
2904 | goto disconnect; |
2905 | } | |
2906 | ||
7204624c PR |
2907 | if (p_two_primaries != tconn->net_conf->two_primaries) { |
2908 | conn_err(tconn, "incompatible setting of the two-primaries options\n"); | |
b411b363 PR |
2909 | goto disconnect; |
2910 | } | |
2911 | ||
7204624c PR |
2912 | if (tconn->agreed_pro_version >= 87) { |
2913 | unsigned char *my_alg = tconn->net_conf->integrity_alg; | |
b411b363 | 2914 | |
7204624c | 2915 | if (drbd_recv(tconn, p_integrity_alg, data_size) != data_size) |
81e84650 | 2916 | return false; |
b411b363 PR |
2917 | |
2918 | p_integrity_alg[SHARED_SECRET_MAX-1] = 0; | |
2919 | if (strcmp(p_integrity_alg, my_alg)) { | |
7204624c | 2920 | conn_err(tconn, "incompatible setting of the data-integrity-alg\n"); |
b411b363 PR |
2921 | goto disconnect; |
2922 | } | |
7204624c | 2923 | conn_info(tconn, "data-integrity-alg: %s\n", |
b411b363 PR |
2924 | my_alg[0] ? my_alg : (unsigned char *)"<not-used>"); |
2925 | } | |
2926 | ||
81e84650 | 2927 | return true; |
b411b363 PR |
2928 | |
2929 | disconnect: | |
7204624c | 2930 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
81e84650 | 2931 | return false; |
b411b363 PR |
2932 | } |
2933 | ||
2934 | /* helper function | |
2935 | * input: alg name, feature name | |
2936 | * return: NULL (alg name was "") | |
2937 | * ERR_PTR(error) if something goes wrong | |
2938 | * or the crypto hash ptr, if it worked out ok. */ | |
2939 | struct crypto_hash *drbd_crypto_alloc_digest_safe(const struct drbd_conf *mdev, | |
2940 | const char *alg, const char *name) | |
2941 | { | |
2942 | struct crypto_hash *tfm; | |
2943 | ||
2944 | if (!alg[0]) | |
2945 | return NULL; | |
2946 | ||
2947 | tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC); | |
2948 | if (IS_ERR(tfm)) { | |
2949 | dev_err(DEV, "Can not allocate \"%s\" as %s (reason: %ld)\n", | |
2950 | alg, name, PTR_ERR(tfm)); | |
2951 | return tfm; | |
2952 | } | |
2953 | if (!drbd_crypto_is_hash(crypto_hash_tfm(tfm))) { | |
2954 | crypto_free_hash(tfm); | |
2955 | dev_err(DEV, "\"%s\" is not a digest (%s)\n", alg, name); | |
2956 | return ERR_PTR(-EINVAL); | |
2957 | } | |
2958 | return tfm; | |
2959 | } | |
2960 | ||
d8763023 AG |
2961 | static int receive_SyncParam(struct drbd_conf *mdev, enum drbd_packet cmd, |
2962 | unsigned int packet_size) | |
b411b363 | 2963 | { |
81e84650 | 2964 | int ok = true; |
e42325a5 | 2965 | struct p_rs_param_95 *p = &mdev->tconn->data.rbuf.rs_param_95; |
b411b363 PR |
2966 | unsigned int header_size, data_size, exp_max_sz; |
2967 | struct crypto_hash *verify_tfm = NULL; | |
2968 | struct crypto_hash *csums_tfm = NULL; | |
31890f4a | 2969 | const int apv = mdev->tconn->agreed_pro_version; |
778f271d PR |
2970 | int *rs_plan_s = NULL; |
2971 | int fifo_size = 0; | |
b411b363 PR |
2972 | |
2973 | exp_max_sz = apv <= 87 ? sizeof(struct p_rs_param) | |
2974 | : apv == 88 ? sizeof(struct p_rs_param) | |
2975 | + SHARED_SECRET_MAX | |
8e26f9cc PR |
2976 | : apv <= 94 ? sizeof(struct p_rs_param_89) |
2977 | : /* apv >= 95 */ sizeof(struct p_rs_param_95); | |
b411b363 | 2978 | |
02918be2 | 2979 | if (packet_size > exp_max_sz) { |
b411b363 | 2980 | dev_err(DEV, "SyncParam packet too long: received %u, expected <= %u bytes\n", |
02918be2 | 2981 | packet_size, exp_max_sz); |
81e84650 | 2982 | return false; |
b411b363 PR |
2983 | } |
2984 | ||
2985 | if (apv <= 88) { | |
257d0af6 | 2986 | header_size = sizeof(struct p_rs_param) - sizeof(struct p_header); |
02918be2 | 2987 | data_size = packet_size - header_size; |
8e26f9cc | 2988 | } else if (apv <= 94) { |
257d0af6 | 2989 | header_size = sizeof(struct p_rs_param_89) - sizeof(struct p_header); |
02918be2 | 2990 | data_size = packet_size - header_size; |
b411b363 | 2991 | D_ASSERT(data_size == 0); |
8e26f9cc | 2992 | } else { |
257d0af6 | 2993 | header_size = sizeof(struct p_rs_param_95) - sizeof(struct p_header); |
02918be2 | 2994 | data_size = packet_size - header_size; |
b411b363 PR |
2995 | D_ASSERT(data_size == 0); |
2996 | } | |
2997 | ||
2998 | /* initialize verify_alg and csums_alg */ | |
2999 | memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); | |
3000 | ||
de0ff338 | 3001 | if (drbd_recv(mdev->tconn, &p->head.payload, header_size) != header_size) |
81e84650 | 3002 | return false; |
b411b363 | 3003 | |
f399002e LE |
3004 | if (get_ldev(mdev)) { |
3005 | mdev->ldev->dc.resync_rate = be32_to_cpu(p->rate); | |
3006 | put_ldev(mdev); | |
3007 | } | |
b411b363 PR |
3008 | |
3009 | if (apv >= 88) { | |
3010 | if (apv == 88) { | |
3011 | if (data_size > SHARED_SECRET_MAX) { | |
3012 | dev_err(DEV, "verify-alg too long, " | |
3013 | "peer wants %u, accepting only %u byte\n", | |
3014 | data_size, SHARED_SECRET_MAX); | |
81e84650 | 3015 | return false; |
b411b363 PR |
3016 | } |
3017 | ||
de0ff338 | 3018 | if (drbd_recv(mdev->tconn, p->verify_alg, data_size) != data_size) |
81e84650 | 3019 | return false; |
b411b363 PR |
3020 | |
3021 | /* we expect NUL terminated string */ | |
3022 | /* but just in case someone tries to be evil */ | |
3023 | D_ASSERT(p->verify_alg[data_size-1] == 0); | |
3024 | p->verify_alg[data_size-1] = 0; | |
3025 | ||
3026 | } else /* apv >= 89 */ { | |
3027 | /* we still expect NUL terminated strings */ | |
3028 | /* but just in case someone tries to be evil */ | |
3029 | D_ASSERT(p->verify_alg[SHARED_SECRET_MAX-1] == 0); | |
3030 | D_ASSERT(p->csums_alg[SHARED_SECRET_MAX-1] == 0); | |
3031 | p->verify_alg[SHARED_SECRET_MAX-1] = 0; | |
3032 | p->csums_alg[SHARED_SECRET_MAX-1] = 0; | |
3033 | } | |
3034 | ||
f399002e | 3035 | if (strcmp(mdev->tconn->net_conf->verify_alg, p->verify_alg)) { |
b411b363 PR |
3036 | if (mdev->state.conn == C_WF_REPORT_PARAMS) { |
3037 | dev_err(DEV, "Different verify-alg settings. me=\"%s\" peer=\"%s\"\n", | |
f399002e | 3038 | mdev->tconn->net_conf->verify_alg, p->verify_alg); |
b411b363 PR |
3039 | goto disconnect; |
3040 | } | |
3041 | verify_tfm = drbd_crypto_alloc_digest_safe(mdev, | |
3042 | p->verify_alg, "verify-alg"); | |
3043 | if (IS_ERR(verify_tfm)) { | |
3044 | verify_tfm = NULL; | |
3045 | goto disconnect; | |
3046 | } | |
3047 | } | |
3048 | ||
f399002e | 3049 | if (apv >= 89 && strcmp(mdev->tconn->net_conf->csums_alg, p->csums_alg)) { |
b411b363 PR |
3050 | if (mdev->state.conn == C_WF_REPORT_PARAMS) { |
3051 | dev_err(DEV, "Different csums-alg settings. me=\"%s\" peer=\"%s\"\n", | |
f399002e | 3052 | mdev->tconn->net_conf->csums_alg, p->csums_alg); |
b411b363 PR |
3053 | goto disconnect; |
3054 | } | |
3055 | csums_tfm = drbd_crypto_alloc_digest_safe(mdev, | |
3056 | p->csums_alg, "csums-alg"); | |
3057 | if (IS_ERR(csums_tfm)) { | |
3058 | csums_tfm = NULL; | |
3059 | goto disconnect; | |
3060 | } | |
3061 | } | |
3062 | ||
f399002e LE |
3063 | if (apv > 94 && get_ldev(mdev)) { |
3064 | mdev->ldev->dc.resync_rate = be32_to_cpu(p->rate); | |
3065 | mdev->ldev->dc.c_plan_ahead = be32_to_cpu(p->c_plan_ahead); | |
3066 | mdev->ldev->dc.c_delay_target = be32_to_cpu(p->c_delay_target); | |
3067 | mdev->ldev->dc.c_fill_target = be32_to_cpu(p->c_fill_target); | |
3068 | mdev->ldev->dc.c_max_rate = be32_to_cpu(p->c_max_rate); | |
778f271d | 3069 | |
f399002e | 3070 | fifo_size = (mdev->ldev->dc.c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
778f271d PR |
3071 | if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) { |
3072 | rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL); | |
3073 | if (!rs_plan_s) { | |
3074 | dev_err(DEV, "kmalloc of fifo_buffer failed"); | |
f399002e | 3075 | put_ldev(mdev); |
778f271d PR |
3076 | goto disconnect; |
3077 | } | |
3078 | } | |
f399002e | 3079 | put_ldev(mdev); |
8e26f9cc | 3080 | } |
b411b363 PR |
3081 | |
3082 | spin_lock(&mdev->peer_seq_lock); | |
3083 | /* lock against drbd_nl_syncer_conf() */ | |
3084 | if (verify_tfm) { | |
f399002e LE |
3085 | strcpy(mdev->tconn->net_conf->verify_alg, p->verify_alg); |
3086 | mdev->tconn->net_conf->verify_alg_len = strlen(p->verify_alg) + 1; | |
3087 | crypto_free_hash(mdev->tconn->verify_tfm); | |
3088 | mdev->tconn->verify_tfm = verify_tfm; | |
b411b363 PR |
3089 | dev_info(DEV, "using verify-alg: \"%s\"\n", p->verify_alg); |
3090 | } | |
3091 | if (csums_tfm) { | |
f399002e LE |
3092 | strcpy(mdev->tconn->net_conf->csums_alg, p->csums_alg); |
3093 | mdev->tconn->net_conf->csums_alg_len = strlen(p->csums_alg) + 1; | |
3094 | crypto_free_hash(mdev->tconn->csums_tfm); | |
3095 | mdev->tconn->csums_tfm = csums_tfm; | |
b411b363 PR |
3096 | dev_info(DEV, "using csums-alg: \"%s\"\n", p->csums_alg); |
3097 | } | |
778f271d PR |
3098 | if (fifo_size != mdev->rs_plan_s.size) { |
3099 | kfree(mdev->rs_plan_s.values); | |
3100 | mdev->rs_plan_s.values = rs_plan_s; | |
3101 | mdev->rs_plan_s.size = fifo_size; | |
3102 | mdev->rs_planed = 0; | |
3103 | } | |
b411b363 PR |
3104 | spin_unlock(&mdev->peer_seq_lock); |
3105 | } | |
3106 | ||
3107 | return ok; | |
3108 | disconnect: | |
3109 | /* just for completeness: actually not needed, | |
3110 | * as this is not reached if csums_tfm was ok. */ | |
3111 | crypto_free_hash(csums_tfm); | |
3112 | /* but free the verify_tfm again, if csums_tfm did not work out */ | |
3113 | crypto_free_hash(verify_tfm); | |
38fa9988 | 3114 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
81e84650 | 3115 | return false; |
b411b363 PR |
3116 | } |
3117 | ||
b411b363 PR |
3118 | /* warn if the arguments differ by more than 12.5% */ |
3119 | static void warn_if_differ_considerably(struct drbd_conf *mdev, | |
3120 | const char *s, sector_t a, sector_t b) | |
3121 | { | |
3122 | sector_t d; | |
3123 | if (a == 0 || b == 0) | |
3124 | return; | |
3125 | d = (a > b) ? (a - b) : (b - a); | |
3126 | if (d > (a>>3) || d > (b>>3)) | |
3127 | dev_warn(DEV, "Considerable difference in %s: %llus vs. %llus\n", s, | |
3128 | (unsigned long long)a, (unsigned long long)b); | |
3129 | } | |
3130 | ||
d8763023 AG |
3131 | static int receive_sizes(struct drbd_conf *mdev, enum drbd_packet cmd, |
3132 | unsigned int data_size) | |
b411b363 | 3133 | { |
e42325a5 | 3134 | struct p_sizes *p = &mdev->tconn->data.rbuf.sizes; |
b411b363 | 3135 | enum determine_dev_size dd = unchanged; |
b411b363 PR |
3136 | sector_t p_size, p_usize, my_usize; |
3137 | int ldsc = 0; /* local disk size changed */ | |
e89b591c | 3138 | enum dds_flags ddsf; |
b411b363 | 3139 | |
b411b363 PR |
3140 | p_size = be64_to_cpu(p->d_size); |
3141 | p_usize = be64_to_cpu(p->u_size); | |
3142 | ||
b411b363 PR |
3143 | /* just store the peer's disk size for now. |
3144 | * we still need to figure out whether we accept that. */ | |
3145 | mdev->p_size = p_size; | |
3146 | ||
b411b363 PR |
3147 | if (get_ldev(mdev)) { |
3148 | warn_if_differ_considerably(mdev, "lower level device sizes", | |
3149 | p_size, drbd_get_max_capacity(mdev->ldev)); | |
3150 | warn_if_differ_considerably(mdev, "user requested size", | |
3151 | p_usize, mdev->ldev->dc.disk_size); | |
3152 | ||
3153 | /* if this is the first connect, or an otherwise expected | |
3154 | * param exchange, choose the minimum */ | |
3155 | if (mdev->state.conn == C_WF_REPORT_PARAMS) | |
3156 | p_usize = min_not_zero((sector_t)mdev->ldev->dc.disk_size, | |
3157 | p_usize); | |
3158 | ||
3159 | my_usize = mdev->ldev->dc.disk_size; | |
3160 | ||
3161 | if (mdev->ldev->dc.disk_size != p_usize) { | |
3162 | mdev->ldev->dc.disk_size = p_usize; | |
3163 | dev_info(DEV, "Peer sets u_size to %lu sectors\n", | |
3164 | (unsigned long)mdev->ldev->dc.disk_size); | |
3165 | } | |
3166 | ||
3167 | /* Never shrink a device with usable data during connect. | |
3168 | But allow online shrinking if we are connected. */ | |
a393db6f | 3169 | if (drbd_new_dev_size(mdev, mdev->ldev, 0) < |
b411b363 PR |
3170 | drbd_get_capacity(mdev->this_bdev) && |
3171 | mdev->state.disk >= D_OUTDATED && | |
3172 | mdev->state.conn < C_CONNECTED) { | |
3173 | dev_err(DEV, "The peer's disk size is too small!\n"); | |
38fa9988 | 3174 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 PR |
3175 | mdev->ldev->dc.disk_size = my_usize; |
3176 | put_ldev(mdev); | |
81e84650 | 3177 | return false; |
b411b363 PR |
3178 | } |
3179 | put_ldev(mdev); | |
3180 | } | |
b411b363 | 3181 | |
e89b591c | 3182 | ddsf = be16_to_cpu(p->dds_flags); |
b411b363 | 3183 | if (get_ldev(mdev)) { |
24c4830c | 3184 | dd = drbd_determine_dev_size(mdev, ddsf); |
b411b363 PR |
3185 | put_ldev(mdev); |
3186 | if (dd == dev_size_error) | |
81e84650 | 3187 | return false; |
b411b363 PR |
3188 | drbd_md_sync(mdev); |
3189 | } else { | |
3190 | /* I am diskless, need to accept the peer's size. */ | |
3191 | drbd_set_my_capacity(mdev, p_size); | |
3192 | } | |
3193 | ||
99432fcc PR |
3194 | mdev->peer_max_bio_size = be32_to_cpu(p->max_bio_size); |
3195 | drbd_reconsider_max_bio_size(mdev); | |
3196 | ||
b411b363 PR |
3197 | if (get_ldev(mdev)) { |
3198 | if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) { | |
3199 | mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); | |
3200 | ldsc = 1; | |
3201 | } | |
3202 | ||
b411b363 PR |
3203 | put_ldev(mdev); |
3204 | } | |
3205 | ||
3206 | if (mdev->state.conn > C_WF_REPORT_PARAMS) { | |
3207 | if (be64_to_cpu(p->c_size) != | |
3208 | drbd_get_capacity(mdev->this_bdev) || ldsc) { | |
3209 | /* we have different sizes, probably peer | |
3210 | * needs to know my new size... */ | |
e89b591c | 3211 | drbd_send_sizes(mdev, 0, ddsf); |
b411b363 PR |
3212 | } |
3213 | if (test_and_clear_bit(RESIZE_PENDING, &mdev->flags) || | |
3214 | (dd == grew && mdev->state.conn == C_CONNECTED)) { | |
3215 | if (mdev->state.pdsk >= D_INCONSISTENT && | |
e89b591c PR |
3216 | mdev->state.disk >= D_INCONSISTENT) { |
3217 | if (ddsf & DDSF_NO_RESYNC) | |
3218 | dev_info(DEV, "Resync of new storage suppressed with --assume-clean\n"); | |
3219 | else | |
3220 | resync_after_online_grow(mdev); | |
3221 | } else | |
b411b363 PR |
3222 | set_bit(RESYNC_AFTER_NEG, &mdev->flags); |
3223 | } | |
3224 | } | |
3225 | ||
81e84650 | 3226 | return true; |
b411b363 PR |
3227 | } |
3228 | ||
d8763023 AG |
3229 | static int receive_uuids(struct drbd_conf *mdev, enum drbd_packet cmd, |
3230 | unsigned int data_size) | |
b411b363 | 3231 | { |
e42325a5 | 3232 | struct p_uuids *p = &mdev->tconn->data.rbuf.uuids; |
b411b363 | 3233 | u64 *p_uuid; |
62b0da3a | 3234 | int i, updated_uuids = 0; |
b411b363 | 3235 | |
b411b363 PR |
3236 | p_uuid = kmalloc(sizeof(u64)*UI_EXTENDED_SIZE, GFP_NOIO); |
3237 | ||
3238 | for (i = UI_CURRENT; i < UI_EXTENDED_SIZE; i++) | |
3239 | p_uuid[i] = be64_to_cpu(p->uuid[i]); | |
3240 | ||
3241 | kfree(mdev->p_uuid); | |
3242 | mdev->p_uuid = p_uuid; | |
3243 | ||
3244 | if (mdev->state.conn < C_CONNECTED && | |
3245 | mdev->state.disk < D_INCONSISTENT && | |
3246 | mdev->state.role == R_PRIMARY && | |
3247 | (mdev->ed_uuid & ~((u64)1)) != (p_uuid[UI_CURRENT] & ~((u64)1))) { | |
3248 | dev_err(DEV, "Can only connect to data with current UUID=%016llX\n", | |
3249 | (unsigned long long)mdev->ed_uuid); | |
38fa9988 | 3250 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
81e84650 | 3251 | return false; |
b411b363 PR |
3252 | } |
3253 | ||
3254 | if (get_ldev(mdev)) { | |
3255 | int skip_initial_sync = | |
3256 | mdev->state.conn == C_CONNECTED && | |
31890f4a | 3257 | mdev->tconn->agreed_pro_version >= 90 && |
b411b363 PR |
3258 | mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && |
3259 | (p_uuid[UI_FLAGS] & 8); | |
3260 | if (skip_initial_sync) { | |
3261 | dev_info(DEV, "Accepted new current UUID, preparing to skip initial sync\n"); | |
3262 | drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, | |
20ceb2b2 LE |
3263 | "clear_n_write from receive_uuids", |
3264 | BM_LOCKED_TEST_ALLOWED); | |
b411b363 PR |
3265 | _drbd_uuid_set(mdev, UI_CURRENT, p_uuid[UI_CURRENT]); |
3266 | _drbd_uuid_set(mdev, UI_BITMAP, 0); | |
3267 | _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), | |
3268 | CS_VERBOSE, NULL); | |
3269 | drbd_md_sync(mdev); | |
62b0da3a | 3270 | updated_uuids = 1; |
b411b363 PR |
3271 | } |
3272 | put_ldev(mdev); | |
18a50fa2 PR |
3273 | } else if (mdev->state.disk < D_INCONSISTENT && |
3274 | mdev->state.role == R_PRIMARY) { | |
3275 | /* I am a diskless primary, the peer just created a new current UUID | |
3276 | for me. */ | |
62b0da3a | 3277 | updated_uuids = drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]); |
b411b363 PR |
3278 | } |
3279 | ||
3280 | /* Before we test for the disk state, we should wait until an eventually | |
3281 | ongoing cluster wide state change is finished. That is important if | |
3282 | we are primary and are detaching from our disk. We need to see the | |
3283 | new disk state... */ | |
8410da8f PR |
3284 | mutex_lock(mdev->state_mutex); |
3285 | mutex_unlock(mdev->state_mutex); | |
b411b363 | 3286 | if (mdev->state.conn >= C_CONNECTED && mdev->state.disk < D_INCONSISTENT) |
62b0da3a LE |
3287 | updated_uuids |= drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]); |
3288 | ||
3289 | if (updated_uuids) | |
3290 | drbd_print_uuids(mdev, "receiver updated UUIDs to"); | |
b411b363 | 3291 | |
81e84650 | 3292 | return true; |
b411b363 PR |
3293 | } |
3294 | ||
3295 | /** | |
3296 | * convert_state() - Converts the peer's view of the cluster state to our point of view | |
3297 | * @ps: The state as seen by the peer. | |
3298 | */ | |
3299 | static union drbd_state convert_state(union drbd_state ps) | |
3300 | { | |
3301 | union drbd_state ms; | |
3302 | ||
3303 | static enum drbd_conns c_tab[] = { | |
3304 | [C_CONNECTED] = C_CONNECTED, | |
3305 | ||
3306 | [C_STARTING_SYNC_S] = C_STARTING_SYNC_T, | |
3307 | [C_STARTING_SYNC_T] = C_STARTING_SYNC_S, | |
3308 | [C_DISCONNECTING] = C_TEAR_DOWN, /* C_NETWORK_FAILURE, */ | |
3309 | [C_VERIFY_S] = C_VERIFY_T, | |
3310 | [C_MASK] = C_MASK, | |
3311 | }; | |
3312 | ||
3313 | ms.i = ps.i; | |
3314 | ||
3315 | ms.conn = c_tab[ps.conn]; | |
3316 | ms.peer = ps.role; | |
3317 | ms.role = ps.peer; | |
3318 | ms.pdsk = ps.disk; | |
3319 | ms.disk = ps.pdsk; | |
3320 | ms.peer_isp = (ps.aftr_isp | ps.user_isp); | |
3321 | ||
3322 | return ms; | |
3323 | } | |
3324 | ||
d8763023 AG |
3325 | static int receive_req_state(struct drbd_conf *mdev, enum drbd_packet cmd, |
3326 | unsigned int data_size) | |
b411b363 | 3327 | { |
e42325a5 | 3328 | struct p_req_state *p = &mdev->tconn->data.rbuf.req_state; |
b411b363 | 3329 | union drbd_state mask, val; |
bf885f8a | 3330 | enum drbd_state_rv rv; |
b411b363 | 3331 | |
b411b363 PR |
3332 | mask.i = be32_to_cpu(p->mask); |
3333 | val.i = be32_to_cpu(p->val); | |
3334 | ||
25703f83 | 3335 | if (test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) && |
8410da8f | 3336 | mutex_is_locked(mdev->state_mutex)) { |
b411b363 | 3337 | drbd_send_sr_reply(mdev, SS_CONCURRENT_ST_CHG); |
81e84650 | 3338 | return true; |
b411b363 PR |
3339 | } |
3340 | ||
3341 | mask = convert_state(mask); | |
3342 | val = convert_state(val); | |
3343 | ||
dfafcc8a PR |
3344 | rv = drbd_change_state(mdev, CS_VERBOSE, mask, val); |
3345 | drbd_send_sr_reply(mdev, rv); | |
b411b363 | 3346 | |
b411b363 PR |
3347 | drbd_md_sync(mdev); |
3348 | ||
81e84650 | 3349 | return true; |
b411b363 PR |
3350 | } |
3351 | ||
dfafcc8a PR |
3352 | static int receive_req_conn_state(struct drbd_tconn *tconn, enum drbd_packet cmd, |
3353 | unsigned int data_size) | |
3354 | { | |
3355 | struct p_req_state *p = &tconn->data.rbuf.req_state; | |
3356 | union drbd_state mask, val; | |
3357 | enum drbd_state_rv rv; | |
3358 | ||
3359 | mask.i = be32_to_cpu(p->mask); | |
3360 | val.i = be32_to_cpu(p->val); | |
3361 | ||
3362 | if (test_bit(DISCARD_CONCURRENT, &tconn->flags) && | |
3363 | mutex_is_locked(&tconn->cstate_mutex)) { | |
3364 | conn_send_sr_reply(tconn, SS_CONCURRENT_ST_CHG); | |
3365 | return true; | |
3366 | } | |
3367 | ||
3368 | mask = convert_state(mask); | |
3369 | val = convert_state(val); | |
3370 | ||
3371 | rv = conn_request_state(tconn, mask, val, CS_VERBOSE | CS_LOCAL_ONLY); | |
3372 | conn_send_sr_reply(tconn, rv); | |
3373 | ||
3374 | return true; | |
3375 | } | |
3376 | ||
d8763023 AG |
3377 | static int receive_state(struct drbd_conf *mdev, enum drbd_packet cmd, |
3378 | unsigned int data_size) | |
b411b363 | 3379 | { |
e42325a5 | 3380 | struct p_state *p = &mdev->tconn->data.rbuf.state; |
4ac4aada | 3381 | union drbd_state os, ns, peer_state; |
b411b363 | 3382 | enum drbd_disk_state real_peer_disk; |
65d922c3 | 3383 | enum chg_state_flags cs_flags; |
b411b363 PR |
3384 | int rv; |
3385 | ||
b411b363 PR |
3386 | peer_state.i = be32_to_cpu(p->state); |
3387 | ||
3388 | real_peer_disk = peer_state.disk; | |
3389 | if (peer_state.disk == D_NEGOTIATING) { | |
3390 | real_peer_disk = mdev->p_uuid[UI_FLAGS] & 4 ? D_INCONSISTENT : D_CONSISTENT; | |
3391 | dev_info(DEV, "real peer disk state = %s\n", drbd_disk_str(real_peer_disk)); | |
3392 | } | |
3393 | ||
87eeee41 | 3394 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 3395 | retry: |
4ac4aada | 3396 | os = ns = mdev->state; |
87eeee41 | 3397 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 3398 | |
e9ef7bb6 LE |
3399 | /* peer says his disk is uptodate, while we think it is inconsistent, |
3400 | * and this happens while we think we have a sync going on. */ | |
3401 | if (os.pdsk == D_INCONSISTENT && real_peer_disk == D_UP_TO_DATE && | |
3402 | os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) { | |
3403 | /* If we are (becoming) SyncSource, but peer is still in sync | |
3404 | * preparation, ignore its uptodate-ness to avoid flapping, it | |
3405 | * will change to inconsistent once the peer reaches active | |
3406 | * syncing states. | |
3407 | * It may have changed syncer-paused flags, however, so we | |
3408 | * cannot ignore this completely. */ | |
3409 | if (peer_state.conn > C_CONNECTED && | |
3410 | peer_state.conn < C_SYNC_SOURCE) | |
3411 | real_peer_disk = D_INCONSISTENT; | |
3412 | ||
3413 | /* if peer_state changes to connected at the same time, | |
3414 | * it explicitly notifies us that it finished resync. | |
3415 | * Maybe we should finish it up, too? */ | |
3416 | else if (os.conn >= C_SYNC_SOURCE && | |
3417 | peer_state.conn == C_CONNECTED) { | |
3418 | if (drbd_bm_total_weight(mdev) <= mdev->rs_failed) | |
3419 | drbd_resync_finished(mdev); | |
81e84650 | 3420 | return true; |
e9ef7bb6 LE |
3421 | } |
3422 | } | |
3423 | ||
3424 | /* peer says his disk is inconsistent, while we think it is uptodate, | |
3425 | * and this happens while the peer still thinks we have a sync going on, | |
3426 | * but we think we are already done with the sync. | |
3427 | * We ignore this to avoid flapping pdsk. | |
3428 | * This should not happen, if the peer is a recent version of drbd. */ | |
3429 | if (os.pdsk == D_UP_TO_DATE && real_peer_disk == D_INCONSISTENT && | |
3430 | os.conn == C_CONNECTED && peer_state.conn > C_SYNC_SOURCE) | |
3431 | real_peer_disk = D_UP_TO_DATE; | |
3432 | ||
4ac4aada LE |
3433 | if (ns.conn == C_WF_REPORT_PARAMS) |
3434 | ns.conn = C_CONNECTED; | |
b411b363 | 3435 | |
67531718 PR |
3436 | if (peer_state.conn == C_AHEAD) |
3437 | ns.conn = C_BEHIND; | |
3438 | ||
b411b363 PR |
3439 | if (mdev->p_uuid && peer_state.disk >= D_NEGOTIATING && |
3440 | get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
3441 | int cr; /* consider resync */ | |
3442 | ||
3443 | /* if we established a new connection */ | |
4ac4aada | 3444 | cr = (os.conn < C_CONNECTED); |
b411b363 PR |
3445 | /* if we had an established connection |
3446 | * and one of the nodes newly attaches a disk */ | |
4ac4aada | 3447 | cr |= (os.conn == C_CONNECTED && |
b411b363 | 3448 | (peer_state.disk == D_NEGOTIATING || |
4ac4aada | 3449 | os.disk == D_NEGOTIATING)); |
b411b363 PR |
3450 | /* if we have both been inconsistent, and the peer has been |
3451 | * forced to be UpToDate with --overwrite-data */ | |
3452 | cr |= test_bit(CONSIDER_RESYNC, &mdev->flags); | |
3453 | /* if we had been plain connected, and the admin requested to | |
3454 | * start a sync by "invalidate" or "invalidate-remote" */ | |
4ac4aada | 3455 | cr |= (os.conn == C_CONNECTED && |
b411b363 PR |
3456 | (peer_state.conn >= C_STARTING_SYNC_S && |
3457 | peer_state.conn <= C_WF_BITMAP_T)); | |
3458 | ||
3459 | if (cr) | |
4ac4aada | 3460 | ns.conn = drbd_sync_handshake(mdev, peer_state.role, real_peer_disk); |
b411b363 PR |
3461 | |
3462 | put_ldev(mdev); | |
4ac4aada LE |
3463 | if (ns.conn == C_MASK) { |
3464 | ns.conn = C_CONNECTED; | |
b411b363 | 3465 | if (mdev->state.disk == D_NEGOTIATING) { |
82f59cc6 | 3466 | drbd_force_state(mdev, NS(disk, D_FAILED)); |
b411b363 PR |
3467 | } else if (peer_state.disk == D_NEGOTIATING) { |
3468 | dev_err(DEV, "Disk attach process on the peer node was aborted.\n"); | |
3469 | peer_state.disk = D_DISKLESS; | |
580b9767 | 3470 | real_peer_disk = D_DISKLESS; |
b411b363 | 3471 | } else { |
8169e41b | 3472 | if (test_and_clear_bit(CONN_DRY_RUN, &mdev->tconn->flags)) |
81e84650 | 3473 | return false; |
4ac4aada | 3474 | D_ASSERT(os.conn == C_WF_REPORT_PARAMS); |
38fa9988 | 3475 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
81e84650 | 3476 | return false; |
b411b363 PR |
3477 | } |
3478 | } | |
3479 | } | |
3480 | ||
87eeee41 | 3481 | spin_lock_irq(&mdev->tconn->req_lock); |
4ac4aada | 3482 | if (mdev->state.i != os.i) |
b411b363 PR |
3483 | goto retry; |
3484 | clear_bit(CONSIDER_RESYNC, &mdev->flags); | |
b411b363 PR |
3485 | ns.peer = peer_state.role; |
3486 | ns.pdsk = real_peer_disk; | |
3487 | ns.peer_isp = (peer_state.aftr_isp | peer_state.user_isp); | |
4ac4aada | 3488 | if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING) |
b411b363 | 3489 | ns.disk = mdev->new_state_tmp.disk; |
4ac4aada LE |
3490 | cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD); |
3491 | if (ns.pdsk == D_CONSISTENT && is_susp(ns) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED && | |
481c6f50 | 3492 | test_bit(NEW_CUR_UUID, &mdev->flags)) { |
8554df1c | 3493 | /* Do not allow tl_restart(RESEND) for a rebooted peer. We can only allow this |
481c6f50 | 3494 | for temporal network outages! */ |
87eeee41 | 3495 | spin_unlock_irq(&mdev->tconn->req_lock); |
481c6f50 | 3496 | dev_err(DEV, "Aborting Connect, can not thaw IO with an only Consistent peer\n"); |
2f5cdd0b | 3497 | tl_clear(mdev->tconn); |
481c6f50 PR |
3498 | drbd_uuid_new_current(mdev); |
3499 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
38fa9988 | 3500 | conn_request_state(mdev->tconn, NS2(conn, C_PROTOCOL_ERROR, susp, 0), CS_HARD); |
81e84650 | 3501 | return false; |
481c6f50 | 3502 | } |
65d922c3 | 3503 | rv = _drbd_set_state(mdev, ns, cs_flags, NULL); |
b411b363 | 3504 | ns = mdev->state; |
87eeee41 | 3505 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
3506 | |
3507 | if (rv < SS_SUCCESS) { | |
38fa9988 | 3508 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
81e84650 | 3509 | return false; |
b411b363 PR |
3510 | } |
3511 | ||
4ac4aada LE |
3512 | if (os.conn > C_WF_REPORT_PARAMS) { |
3513 | if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED && | |
b411b363 PR |
3514 | peer_state.disk != D_NEGOTIATING ) { |
3515 | /* we want resync, peer has not yet decided to sync... */ | |
3516 | /* Nowadays only used when forcing a node into primary role and | |
3517 | setting its disk to UpToDate with that */ | |
3518 | drbd_send_uuids(mdev); | |
3519 | drbd_send_state(mdev); | |
3520 | } | |
3521 | } | |
3522 | ||
89e58e75 | 3523 | mdev->tconn->net_conf->want_lose = 0; |
b411b363 PR |
3524 | |
3525 | drbd_md_sync(mdev); /* update connected indicator, la_size, ... */ | |
3526 | ||
81e84650 | 3527 | return true; |
b411b363 PR |
3528 | } |
3529 | ||
d8763023 AG |
3530 | static int receive_sync_uuid(struct drbd_conf *mdev, enum drbd_packet cmd, |
3531 | unsigned int data_size) | |
b411b363 | 3532 | { |
e42325a5 | 3533 | struct p_rs_uuid *p = &mdev->tconn->data.rbuf.rs_uuid; |
b411b363 PR |
3534 | |
3535 | wait_event(mdev->misc_wait, | |
3536 | mdev->state.conn == C_WF_SYNC_UUID || | |
c4752ef1 | 3537 | mdev->state.conn == C_BEHIND || |
b411b363 PR |
3538 | mdev->state.conn < C_CONNECTED || |
3539 | mdev->state.disk < D_NEGOTIATING); | |
3540 | ||
3541 | /* D_ASSERT( mdev->state.conn == C_WF_SYNC_UUID ); */ | |
3542 | ||
b411b363 PR |
3543 | /* Here the _drbd_uuid_ functions are right, current should |
3544 | _not_ be rotated into the history */ | |
3545 | if (get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
3546 | _drbd_uuid_set(mdev, UI_CURRENT, be64_to_cpu(p->uuid)); | |
3547 | _drbd_uuid_set(mdev, UI_BITMAP, 0UL); | |
3548 | ||
62b0da3a | 3549 | drbd_print_uuids(mdev, "updated sync uuid"); |
b411b363 PR |
3550 | drbd_start_resync(mdev, C_SYNC_TARGET); |
3551 | ||
3552 | put_ldev(mdev); | |
3553 | } else | |
3554 | dev_err(DEV, "Ignoring SyncUUID packet!\n"); | |
3555 | ||
81e84650 | 3556 | return true; |
b411b363 PR |
3557 | } |
3558 | ||
2c46407d AG |
3559 | /** |
3560 | * receive_bitmap_plain | |
3561 | * | |
3562 | * Return 0 when done, 1 when another iteration is needed, and a negative error | |
3563 | * code upon failure. | |
3564 | */ | |
3565 | static int | |
02918be2 PR |
3566 | receive_bitmap_plain(struct drbd_conf *mdev, unsigned int data_size, |
3567 | unsigned long *buffer, struct bm_xfer_ctx *c) | |
b411b363 PR |
3568 | { |
3569 | unsigned num_words = min_t(size_t, BM_PACKET_WORDS, c->bm_words - c->word_offset); | |
3570 | unsigned want = num_words * sizeof(long); | |
2c46407d | 3571 | int err; |
b411b363 | 3572 | |
02918be2 PR |
3573 | if (want != data_size) { |
3574 | dev_err(DEV, "%s:want (%u) != data_size (%u)\n", __func__, want, data_size); | |
2c46407d | 3575 | return -EIO; |
b411b363 PR |
3576 | } |
3577 | if (want == 0) | |
2c46407d | 3578 | return 0; |
de0ff338 | 3579 | err = drbd_recv(mdev->tconn, buffer, want); |
2c46407d AG |
3580 | if (err != want) { |
3581 | if (err >= 0) | |
3582 | err = -EIO; | |
3583 | return err; | |
3584 | } | |
b411b363 PR |
3585 | |
3586 | drbd_bm_merge_lel(mdev, c->word_offset, num_words, buffer); | |
3587 | ||
3588 | c->word_offset += num_words; | |
3589 | c->bit_offset = c->word_offset * BITS_PER_LONG; | |
3590 | if (c->bit_offset > c->bm_bits) | |
3591 | c->bit_offset = c->bm_bits; | |
3592 | ||
2c46407d | 3593 | return 1; |
b411b363 PR |
3594 | } |
3595 | ||
2c46407d AG |
3596 | /** |
3597 | * recv_bm_rle_bits | |
3598 | * | |
3599 | * Return 0 when done, 1 when another iteration is needed, and a negative error | |
3600 | * code upon failure. | |
3601 | */ | |
3602 | static int | |
b411b363 PR |
3603 | recv_bm_rle_bits(struct drbd_conf *mdev, |
3604 | struct p_compressed_bm *p, | |
c6d25cfe PR |
3605 | struct bm_xfer_ctx *c, |
3606 | unsigned int len) | |
b411b363 PR |
3607 | { |
3608 | struct bitstream bs; | |
3609 | u64 look_ahead; | |
3610 | u64 rl; | |
3611 | u64 tmp; | |
3612 | unsigned long s = c->bit_offset; | |
3613 | unsigned long e; | |
b411b363 PR |
3614 | int toggle = DCBP_get_start(p); |
3615 | int have; | |
3616 | int bits; | |
3617 | ||
3618 | bitstream_init(&bs, p->code, len, DCBP_get_pad_bits(p)); | |
3619 | ||
3620 | bits = bitstream_get_bits(&bs, &look_ahead, 64); | |
3621 | if (bits < 0) | |
2c46407d | 3622 | return -EIO; |
b411b363 PR |
3623 | |
3624 | for (have = bits; have > 0; s += rl, toggle = !toggle) { | |
3625 | bits = vli_decode_bits(&rl, look_ahead); | |
3626 | if (bits <= 0) | |
2c46407d | 3627 | return -EIO; |
b411b363 PR |
3628 | |
3629 | if (toggle) { | |
3630 | e = s + rl -1; | |
3631 | if (e >= c->bm_bits) { | |
3632 | dev_err(DEV, "bitmap overflow (e:%lu) while decoding bm RLE packet\n", e); | |
2c46407d | 3633 | return -EIO; |
b411b363 PR |
3634 | } |
3635 | _drbd_bm_set_bits(mdev, s, e); | |
3636 | } | |
3637 | ||
3638 | if (have < bits) { | |
3639 | dev_err(DEV, "bitmap decoding error: h:%d b:%d la:0x%08llx l:%u/%u\n", | |
3640 | have, bits, look_ahead, | |
3641 | (unsigned int)(bs.cur.b - p->code), | |
3642 | (unsigned int)bs.buf_len); | |
2c46407d | 3643 | return -EIO; |
b411b363 PR |
3644 | } |
3645 | look_ahead >>= bits; | |
3646 | have -= bits; | |
3647 | ||
3648 | bits = bitstream_get_bits(&bs, &tmp, 64 - have); | |
3649 | if (bits < 0) | |
2c46407d | 3650 | return -EIO; |
b411b363 PR |
3651 | look_ahead |= tmp << have; |
3652 | have += bits; | |
3653 | } | |
3654 | ||
3655 | c->bit_offset = s; | |
3656 | bm_xfer_ctx_bit_to_word_offset(c); | |
3657 | ||
2c46407d | 3658 | return (s != c->bm_bits); |
b411b363 PR |
3659 | } |
3660 | ||
2c46407d AG |
3661 | /** |
3662 | * decode_bitmap_c | |
3663 | * | |
3664 | * Return 0 when done, 1 when another iteration is needed, and a negative error | |
3665 | * code upon failure. | |
3666 | */ | |
3667 | static int | |
b411b363 PR |
3668 | decode_bitmap_c(struct drbd_conf *mdev, |
3669 | struct p_compressed_bm *p, | |
c6d25cfe PR |
3670 | struct bm_xfer_ctx *c, |
3671 | unsigned int len) | |
b411b363 PR |
3672 | { |
3673 | if (DCBP_get_code(p) == RLE_VLI_Bits) | |
c6d25cfe | 3674 | return recv_bm_rle_bits(mdev, p, c, len); |
b411b363 PR |
3675 | |
3676 | /* other variants had been implemented for evaluation, | |
3677 | * but have been dropped as this one turned out to be "best" | |
3678 | * during all our tests. */ | |
3679 | ||
3680 | dev_err(DEV, "receive_bitmap_c: unknown encoding %u\n", p->encoding); | |
38fa9988 | 3681 | conn_request_state(mdev->tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD); |
2c46407d | 3682 | return -EIO; |
b411b363 PR |
3683 | } |
3684 | ||
3685 | void INFO_bm_xfer_stats(struct drbd_conf *mdev, | |
3686 | const char *direction, struct bm_xfer_ctx *c) | |
3687 | { | |
3688 | /* what would it take to transfer it "plaintext" */ | |
c012949a | 3689 | unsigned plain = sizeof(struct p_header) * |
b411b363 PR |
3690 | ((c->bm_words+BM_PACKET_WORDS-1)/BM_PACKET_WORDS+1) |
3691 | + c->bm_words * sizeof(long); | |
3692 | unsigned total = c->bytes[0] + c->bytes[1]; | |
3693 | unsigned r; | |
3694 | ||
3695 | /* total can not be zero. but just in case: */ | |
3696 | if (total == 0) | |
3697 | return; | |
3698 | ||
3699 | /* don't report if not compressed */ | |
3700 | if (total >= plain) | |
3701 | return; | |
3702 | ||
3703 | /* total < plain. check for overflow, still */ | |
3704 | r = (total > UINT_MAX/1000) ? (total / (plain/1000)) | |
3705 | : (1000 * total / plain); | |
3706 | ||
3707 | if (r > 1000) | |
3708 | r = 1000; | |
3709 | ||
3710 | r = 1000 - r; | |
3711 | dev_info(DEV, "%s bitmap stats [Bytes(packets)]: plain %u(%u), RLE %u(%u), " | |
3712 | "total %u; compression: %u.%u%%\n", | |
3713 | direction, | |
3714 | c->bytes[1], c->packets[1], | |
3715 | c->bytes[0], c->packets[0], | |
3716 | total, r/10, r % 10); | |
3717 | } | |
3718 | ||
3719 | /* Since we are processing the bitfield from lower addresses to higher, | |
3720 | it does not matter if the process it in 32 bit chunks or 64 bit | |
3721 | chunks as long as it is little endian. (Understand it as byte stream, | |
3722 | beginning with the lowest byte...) If we would use big endian | |
3723 | we would need to process it from the highest address to the lowest, | |
3724 | in order to be agnostic to the 32 vs 64 bits issue. | |
3725 | ||
3726 | returns 0 on failure, 1 if we successfully received it. */ | |
d8763023 AG |
3727 | static int receive_bitmap(struct drbd_conf *mdev, enum drbd_packet cmd, |
3728 | unsigned int data_size) | |
b411b363 PR |
3729 | { |
3730 | struct bm_xfer_ctx c; | |
3731 | void *buffer; | |
2c46407d | 3732 | int err; |
81e84650 | 3733 | int ok = false; |
257d0af6 | 3734 | struct p_header *h = &mdev->tconn->data.rbuf.header; |
77351055 | 3735 | struct packet_info pi; |
b411b363 | 3736 | |
20ceb2b2 LE |
3737 | drbd_bm_lock(mdev, "receive bitmap", BM_LOCKED_SET_ALLOWED); |
3738 | /* you are supposed to send additional out-of-sync information | |
3739 | * if you actually set bits during this phase */ | |
b411b363 PR |
3740 | |
3741 | /* maybe we should use some per thread scratch page, | |
3742 | * and allocate that during initial device creation? */ | |
3743 | buffer = (unsigned long *) __get_free_page(GFP_NOIO); | |
3744 | if (!buffer) { | |
3745 | dev_err(DEV, "failed to allocate one page buffer in %s\n", __func__); | |
3746 | goto out; | |
3747 | } | |
3748 | ||
3749 | c = (struct bm_xfer_ctx) { | |
3750 | .bm_bits = drbd_bm_bits(mdev), | |
3751 | .bm_words = drbd_bm_words(mdev), | |
3752 | }; | |
3753 | ||
2c46407d | 3754 | for(;;) { |
02918be2 | 3755 | if (cmd == P_BITMAP) { |
2c46407d | 3756 | err = receive_bitmap_plain(mdev, data_size, buffer, &c); |
02918be2 | 3757 | } else if (cmd == P_COMPRESSED_BITMAP) { |
b411b363 PR |
3758 | /* MAYBE: sanity check that we speak proto >= 90, |
3759 | * and the feature is enabled! */ | |
3760 | struct p_compressed_bm *p; | |
3761 | ||
02918be2 | 3762 | if (data_size > BM_PACKET_PAYLOAD_BYTES) { |
b411b363 PR |
3763 | dev_err(DEV, "ReportCBitmap packet too large\n"); |
3764 | goto out; | |
3765 | } | |
3766 | /* use the page buff */ | |
3767 | p = buffer; | |
3768 | memcpy(p, h, sizeof(*h)); | |
de0ff338 | 3769 | if (drbd_recv(mdev->tconn, p->head.payload, data_size) != data_size) |
b411b363 | 3770 | goto out; |
004352fa LE |
3771 | if (data_size <= (sizeof(*p) - sizeof(p->head))) { |
3772 | dev_err(DEV, "ReportCBitmap packet too small (l:%u)\n", data_size); | |
78fcbdae | 3773 | goto out; |
b411b363 | 3774 | } |
c6d25cfe | 3775 | err = decode_bitmap_c(mdev, p, &c, data_size); |
b411b363 | 3776 | } else { |
02918be2 | 3777 | dev_warn(DEV, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", cmd); |
b411b363 PR |
3778 | goto out; |
3779 | } | |
3780 | ||
02918be2 | 3781 | c.packets[cmd == P_BITMAP]++; |
257d0af6 | 3782 | c.bytes[cmd == P_BITMAP] += sizeof(struct p_header) + data_size; |
b411b363 | 3783 | |
2c46407d AG |
3784 | if (err <= 0) { |
3785 | if (err < 0) | |
3786 | goto out; | |
b411b363 | 3787 | break; |
2c46407d | 3788 | } |
9ba7aa00 | 3789 | if (!drbd_recv_header(mdev->tconn, &pi)) |
b411b363 | 3790 | goto out; |
77351055 PR |
3791 | cmd = pi.cmd; |
3792 | data_size = pi.size; | |
2c46407d | 3793 | } |
b411b363 PR |
3794 | |
3795 | INFO_bm_xfer_stats(mdev, "receive", &c); | |
3796 | ||
3797 | if (mdev->state.conn == C_WF_BITMAP_T) { | |
de1f8e4a AG |
3798 | enum drbd_state_rv rv; |
3799 | ||
b411b363 PR |
3800 | ok = !drbd_send_bitmap(mdev); |
3801 | if (!ok) | |
3802 | goto out; | |
3803 | /* Omit CS_ORDERED with this state transition to avoid deadlocks. */ | |
de1f8e4a AG |
3804 | rv = _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); |
3805 | D_ASSERT(rv == SS_SUCCESS); | |
b411b363 PR |
3806 | } else if (mdev->state.conn != C_WF_BITMAP_S) { |
3807 | /* admin may have requested C_DISCONNECTING, | |
3808 | * other threads may have noticed network errors */ | |
3809 | dev_info(DEV, "unexpected cstate (%s) in receive_bitmap\n", | |
3810 | drbd_conn_str(mdev->state.conn)); | |
3811 | } | |
3812 | ||
81e84650 | 3813 | ok = true; |
b411b363 | 3814 | out: |
20ceb2b2 | 3815 | drbd_bm_unlock(mdev); |
b411b363 PR |
3816 | if (ok && mdev->state.conn == C_WF_BITMAP_S) |
3817 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
3818 | free_page((unsigned long) buffer); | |
3819 | return ok; | |
3820 | } | |
3821 | ||
2de876ef | 3822 | static int _tconn_receive_skip(struct drbd_tconn *tconn, unsigned int data_size) |
b411b363 PR |
3823 | { |
3824 | /* TODO zero copy sink :) */ | |
3825 | static char sink[128]; | |
3826 | int size, want, r; | |
3827 | ||
02918be2 | 3828 | size = data_size; |
b411b363 PR |
3829 | while (size > 0) { |
3830 | want = min_t(int, size, sizeof(sink)); | |
2de876ef PR |
3831 | r = drbd_recv(tconn, sink, want); |
3832 | if (r <= 0) | |
841ce241 | 3833 | break; |
b411b363 PR |
3834 | size -= r; |
3835 | } | |
3836 | return size == 0; | |
3837 | } | |
3838 | ||
2de876ef PR |
3839 | static int receive_skip(struct drbd_conf *mdev, enum drbd_packet cmd, |
3840 | unsigned int data_size) | |
3841 | { | |
3842 | dev_warn(DEV, "skipping unknown optional packet type %d, l: %d!\n", | |
3843 | cmd, data_size); | |
3844 | ||
3845 | return _tconn_receive_skip(mdev->tconn, data_size); | |
3846 | } | |
3847 | ||
3848 | static int tconn_receive_skip(struct drbd_tconn *tconn, enum drbd_packet cmd, unsigned int data_size) | |
3849 | { | |
3850 | conn_warn(tconn, "skipping packet for non existing volume type %d, l: %d!\n", | |
3851 | cmd, data_size); | |
3852 | ||
3853 | return _tconn_receive_skip(tconn, data_size); | |
3854 | } | |
3855 | ||
d8763023 AG |
3856 | static int receive_UnplugRemote(struct drbd_conf *mdev, enum drbd_packet cmd, |
3857 | unsigned int data_size) | |
0ced55a3 | 3858 | { |
e7f52dfb LE |
3859 | /* Make sure we've acked all the TCP data associated |
3860 | * with the data requests being unplugged */ | |
e42325a5 | 3861 | drbd_tcp_quickack(mdev->tconn->data.socket); |
0ced55a3 | 3862 | |
81e84650 | 3863 | return true; |
0ced55a3 PR |
3864 | } |
3865 | ||
d8763023 AG |
3866 | static int receive_out_of_sync(struct drbd_conf *mdev, enum drbd_packet cmd, |
3867 | unsigned int data_size) | |
73a01a18 | 3868 | { |
e42325a5 | 3869 | struct p_block_desc *p = &mdev->tconn->data.rbuf.block_desc; |
73a01a18 | 3870 | |
f735e363 LE |
3871 | switch (mdev->state.conn) { |
3872 | case C_WF_SYNC_UUID: | |
3873 | case C_WF_BITMAP_T: | |
3874 | case C_BEHIND: | |
3875 | break; | |
3876 | default: | |
3877 | dev_err(DEV, "ASSERT FAILED cstate = %s, expected: WFSyncUUID|WFBitMapT|Behind\n", | |
3878 | drbd_conn_str(mdev->state.conn)); | |
3879 | } | |
3880 | ||
73a01a18 PR |
3881 | drbd_set_out_of_sync(mdev, be64_to_cpu(p->sector), be32_to_cpu(p->blksize)); |
3882 | ||
81e84650 | 3883 | return true; |
73a01a18 PR |
3884 | } |
3885 | ||
02918be2 PR |
3886 | struct data_cmd { |
3887 | int expect_payload; | |
3888 | size_t pkt_size; | |
a4fbda8e | 3889 | enum mdev_or_conn fa_type; /* first argument's type */ |
d9ae84e7 PR |
3890 | union { |
3891 | int (*mdev_fn)(struct drbd_conf *, enum drbd_packet cmd, | |
3892 | unsigned int to_receive); | |
3893 | int (*conn_fn)(struct drbd_tconn *, enum drbd_packet cmd, | |
3894 | unsigned int to_receive); | |
3895 | }; | |
02918be2 PR |
3896 | }; |
3897 | ||
3898 | static struct data_cmd drbd_cmd_handler[] = { | |
d9ae84e7 PR |
3899 | [P_DATA] = { 1, sizeof(struct p_data), MDEV, { receive_Data } }, |
3900 | [P_DATA_REPLY] = { 1, sizeof(struct p_data), MDEV, { receive_DataReply } }, | |
3901 | [P_RS_DATA_REPLY] = { 1, sizeof(struct p_data), MDEV, { receive_RSDataReply } } , | |
3902 | [P_BARRIER] = { 0, sizeof(struct p_barrier), MDEV, { receive_Barrier } } , | |
3903 | [P_BITMAP] = { 1, sizeof(struct p_header), MDEV, { receive_bitmap } } , | |
3904 | [P_COMPRESSED_BITMAP] = { 1, sizeof(struct p_header), MDEV, { receive_bitmap } } , | |
3905 | [P_UNPLUG_REMOTE] = { 0, sizeof(struct p_header), MDEV, { receive_UnplugRemote } }, | |
3906 | [P_DATA_REQUEST] = { 0, sizeof(struct p_block_req), MDEV, { receive_DataRequest } }, | |
3907 | [P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), MDEV, { receive_DataRequest } }, | |
3908 | [P_SYNC_PARAM] = { 1, sizeof(struct p_header), MDEV, { receive_SyncParam } }, | |
3909 | [P_SYNC_PARAM89] = { 1, sizeof(struct p_header), MDEV, { receive_SyncParam } }, | |
7204624c | 3910 | [P_PROTOCOL] = { 1, sizeof(struct p_protocol), CONN, { .conn_fn = receive_protocol } }, |
d9ae84e7 PR |
3911 | [P_UUIDS] = { 0, sizeof(struct p_uuids), MDEV, { receive_uuids } }, |
3912 | [P_SIZES] = { 0, sizeof(struct p_sizes), MDEV, { receive_sizes } }, | |
3913 | [P_STATE] = { 0, sizeof(struct p_state), MDEV, { receive_state } }, | |
3914 | [P_STATE_CHG_REQ] = { 0, sizeof(struct p_req_state), MDEV, { receive_req_state } }, | |
3915 | [P_SYNC_UUID] = { 0, sizeof(struct p_rs_uuid), MDEV, { receive_sync_uuid } }, | |
3916 | [P_OV_REQUEST] = { 0, sizeof(struct p_block_req), MDEV, { receive_DataRequest } }, | |
3917 | [P_OV_REPLY] = { 1, sizeof(struct p_block_req), MDEV, { receive_DataRequest } }, | |
3918 | [P_CSUM_RS_REQUEST] = { 1, sizeof(struct p_block_req), MDEV, { receive_DataRequest } }, | |
3919 | [P_DELAY_PROBE] = { 0, sizeof(struct p_delay_probe93), MDEV, { receive_skip } }, | |
3920 | [P_OUT_OF_SYNC] = { 0, sizeof(struct p_block_desc), MDEV, { receive_out_of_sync } }, | |
dfafcc8a | 3921 | [P_CONN_ST_CHG_REQ] = { 0, sizeof(struct p_req_state), CONN, { .conn_fn = receive_req_conn_state } }, |
b411b363 PR |
3922 | }; |
3923 | ||
02918be2 | 3924 | /* All handler functions that expect a sub-header get that sub-heder in |
e42325a5 | 3925 | mdev->tconn->data.rbuf.header.head.payload. |
02918be2 | 3926 | |
e42325a5 | 3927 | Usually in mdev->tconn->data.rbuf.header.head the callback can find the usual |
02918be2 PR |
3928 | p_header, but they may not rely on that. Since there is also p_header95 ! |
3929 | */ | |
b411b363 | 3930 | |
eefc2f7d | 3931 | static void drbdd(struct drbd_tconn *tconn) |
b411b363 | 3932 | { |
eefc2f7d | 3933 | struct p_header *header = &tconn->data.rbuf.header; |
77351055 | 3934 | struct packet_info pi; |
02918be2 PR |
3935 | size_t shs; /* sub header size */ |
3936 | int rv; | |
b411b363 | 3937 | |
eefc2f7d PR |
3938 | while (get_t_state(&tconn->receiver) == RUNNING) { |
3939 | drbd_thread_current_set_cpu(&tconn->receiver); | |
3940 | if (!drbd_recv_header(tconn, &pi)) | |
02918be2 | 3941 | goto err_out; |
b411b363 | 3942 | |
6e849ce8 | 3943 | if (unlikely(pi.cmd >= ARRAY_SIZE(drbd_cmd_handler) || |
d9ae84e7 | 3944 | !drbd_cmd_handler[pi.cmd].mdev_fn)) { |
eefc2f7d | 3945 | conn_err(tconn, "unknown packet type %d, l: %d!\n", pi.cmd, pi.size); |
02918be2 | 3946 | goto err_out; |
0b33a916 | 3947 | } |
b411b363 | 3948 | |
77351055 PR |
3949 | shs = drbd_cmd_handler[pi.cmd].pkt_size - sizeof(struct p_header); |
3950 | if (pi.size - shs > 0 && !drbd_cmd_handler[pi.cmd].expect_payload) { | |
eefc2f7d | 3951 | conn_err(tconn, "No payload expected %s l:%d\n", cmdname(pi.cmd), pi.size); |
02918be2 | 3952 | goto err_out; |
b411b363 | 3953 | } |
b411b363 | 3954 | |
c13f7e1a | 3955 | if (shs) { |
eefc2f7d | 3956 | rv = drbd_recv(tconn, &header->payload, shs); |
c13f7e1a | 3957 | if (unlikely(rv != shs)) { |
0ddc5549 | 3958 | if (!signal_pending(current)) |
eefc2f7d | 3959 | conn_warn(tconn, "short read while reading sub header: rv=%d\n", rv); |
c13f7e1a LE |
3960 | goto err_out; |
3961 | } | |
3962 | } | |
3963 | ||
a4fbda8e | 3964 | if (drbd_cmd_handler[pi.cmd].fa_type == CONN) { |
d9ae84e7 PR |
3965 | rv = drbd_cmd_handler[pi.cmd].conn_fn(tconn, pi.cmd, pi.size - shs); |
3966 | } else { | |
3967 | struct drbd_conf *mdev = vnr_to_mdev(tconn, pi.vnr); | |
3968 | rv = mdev ? | |
3969 | drbd_cmd_handler[pi.cmd].mdev_fn(mdev, pi.cmd, pi.size - shs) : | |
3970 | tconn_receive_skip(tconn, pi.cmd, pi.size - shs); | |
3971 | } | |
b411b363 | 3972 | |
02918be2 | 3973 | if (unlikely(!rv)) { |
eefc2f7d | 3974 | conn_err(tconn, "error receiving %s, l: %d!\n", |
77351055 | 3975 | cmdname(pi.cmd), pi.size); |
02918be2 | 3976 | goto err_out; |
b411b363 PR |
3977 | } |
3978 | } | |
b411b363 | 3979 | |
02918be2 PR |
3980 | if (0) { |
3981 | err_out: | |
bbeb641c | 3982 | conn_request_state(tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD); |
02918be2 | 3983 | } |
b411b363 PR |
3984 | } |
3985 | ||
0e29d163 | 3986 | void conn_flush_workqueue(struct drbd_tconn *tconn) |
b411b363 PR |
3987 | { |
3988 | struct drbd_wq_barrier barr; | |
3989 | ||
3990 | barr.w.cb = w_prev_work_done; | |
0e29d163 | 3991 | barr.w.tconn = tconn; |
b411b363 | 3992 | init_completion(&barr.done); |
0e29d163 | 3993 | drbd_queue_work(&tconn->data.work, &barr.w); |
b411b363 PR |
3994 | wait_for_completion(&barr.done); |
3995 | } | |
3996 | ||
360cc740 | 3997 | static void drbd_disconnect(struct drbd_tconn *tconn) |
b411b363 | 3998 | { |
bbeb641c | 3999 | enum drbd_conns oc; |
b411b363 | 4000 | int rv = SS_UNKNOWN_ERROR; |
b411b363 | 4001 | |
bbeb641c | 4002 | if (tconn->cstate == C_STANDALONE) |
b411b363 | 4003 | return; |
b411b363 PR |
4004 | |
4005 | /* asender does not clean up anything. it must not interfere, either */ | |
360cc740 PR |
4006 | drbd_thread_stop(&tconn->asender); |
4007 | drbd_free_sock(tconn); | |
4008 | ||
4009 | idr_for_each(&tconn->volumes, drbd_disconnected, tconn); | |
4010 | ||
4011 | conn_info(tconn, "Connection closed\n"); | |
4012 | ||
4013 | spin_lock_irq(&tconn->req_lock); | |
bbeb641c PR |
4014 | oc = tconn->cstate; |
4015 | if (oc >= C_UNCONNECTED) | |
4016 | rv = _conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE); | |
4017 | ||
360cc740 PR |
4018 | spin_unlock_irq(&tconn->req_lock); |
4019 | ||
bbeb641c | 4020 | if (oc == C_DISCONNECTING) { |
360cc740 PR |
4021 | wait_event(tconn->net_cnt_wait, atomic_read(&tconn->net_cnt) == 0); |
4022 | ||
4023 | crypto_free_hash(tconn->cram_hmac_tfm); | |
4024 | tconn->cram_hmac_tfm = NULL; | |
4025 | ||
4026 | kfree(tconn->net_conf); | |
4027 | tconn->net_conf = NULL; | |
bbeb641c | 4028 | conn_request_state(tconn, NS(conn, C_STANDALONE), CS_VERBOSE); |
360cc740 PR |
4029 | } |
4030 | } | |
4031 | ||
4032 | static int drbd_disconnected(int vnr, void *p, void *data) | |
4033 | { | |
4034 | struct drbd_conf *mdev = (struct drbd_conf *)p; | |
4035 | enum drbd_fencing_p fp; | |
4036 | unsigned int i; | |
b411b363 | 4037 | |
85719573 | 4038 | /* wait for current activity to cease. */ |
87eeee41 | 4039 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
4040 | _drbd_wait_ee_list_empty(mdev, &mdev->active_ee); |
4041 | _drbd_wait_ee_list_empty(mdev, &mdev->sync_ee); | |
4042 | _drbd_wait_ee_list_empty(mdev, &mdev->read_ee); | |
87eeee41 | 4043 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
4044 | |
4045 | /* We do not have data structures that would allow us to | |
4046 | * get the rs_pending_cnt down to 0 again. | |
4047 | * * On C_SYNC_TARGET we do not have any data structures describing | |
4048 | * the pending RSDataRequest's we have sent. | |
4049 | * * On C_SYNC_SOURCE there is no data structure that tracks | |
4050 | * the P_RS_DATA_REPLY blocks that we sent to the SyncTarget. | |
4051 | * And no, it is not the sum of the reference counts in the | |
4052 | * resync_LRU. The resync_LRU tracks the whole operation including | |
4053 | * the disk-IO, while the rs_pending_cnt only tracks the blocks | |
4054 | * on the fly. */ | |
4055 | drbd_rs_cancel_all(mdev); | |
4056 | mdev->rs_total = 0; | |
4057 | mdev->rs_failed = 0; | |
4058 | atomic_set(&mdev->rs_pending_cnt, 0); | |
4059 | wake_up(&mdev->misc_wait); | |
4060 | ||
7fde2be9 PR |
4061 | del_timer(&mdev->request_timer); |
4062 | ||
b411b363 | 4063 | del_timer_sync(&mdev->resync_timer); |
b411b363 PR |
4064 | resync_timer_fn((unsigned long)mdev); |
4065 | ||
b411b363 PR |
4066 | /* wait for all w_e_end_data_req, w_e_end_rsdata_req, w_send_barrier, |
4067 | * w_make_resync_request etc. which may still be on the worker queue | |
4068 | * to be "canceled" */ | |
a21e9298 | 4069 | drbd_flush_workqueue(mdev); |
b411b363 PR |
4070 | |
4071 | /* This also does reclaim_net_ee(). If we do this too early, we might | |
4072 | * miss some resync ee and pages.*/ | |
4073 | drbd_process_done_ee(mdev); | |
4074 | ||
4075 | kfree(mdev->p_uuid); | |
4076 | mdev->p_uuid = NULL; | |
4077 | ||
fb22c402 | 4078 | if (!is_susp(mdev->state)) |
2f5cdd0b | 4079 | tl_clear(mdev->tconn); |
b411b363 | 4080 | |
b411b363 PR |
4081 | drbd_md_sync(mdev); |
4082 | ||
4083 | fp = FP_DONT_CARE; | |
4084 | if (get_ldev(mdev)) { | |
4085 | fp = mdev->ldev->dc.fencing; | |
4086 | put_ldev(mdev); | |
4087 | } | |
4088 | ||
87f7be4c PR |
4089 | if (mdev->state.role == R_PRIMARY && fp >= FP_RESOURCE && mdev->state.pdsk >= D_UNKNOWN) |
4090 | drbd_try_outdate_peer_async(mdev); | |
b411b363 | 4091 | |
20ceb2b2 LE |
4092 | /* serialize with bitmap writeout triggered by the state change, |
4093 | * if any. */ | |
4094 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
4095 | ||
b411b363 PR |
4096 | /* tcp_close and release of sendpage pages can be deferred. I don't |
4097 | * want to use SO_LINGER, because apparently it can be deferred for | |
4098 | * more than 20 seconds (longest time I checked). | |
4099 | * | |
4100 | * Actually we don't care for exactly when the network stack does its | |
4101 | * put_page(), but release our reference on these pages right here. | |
4102 | */ | |
4103 | i = drbd_release_ee(mdev, &mdev->net_ee); | |
4104 | if (i) | |
4105 | dev_info(DEV, "net_ee not empty, killed %u entries\n", i); | |
435f0740 LE |
4106 | i = atomic_read(&mdev->pp_in_use_by_net); |
4107 | if (i) | |
4108 | dev_info(DEV, "pp_in_use_by_net = %d, expected 0\n", i); | |
b411b363 PR |
4109 | i = atomic_read(&mdev->pp_in_use); |
4110 | if (i) | |
45bb912b | 4111 | dev_info(DEV, "pp_in_use = %d, expected 0\n", i); |
b411b363 PR |
4112 | |
4113 | D_ASSERT(list_empty(&mdev->read_ee)); | |
4114 | D_ASSERT(list_empty(&mdev->active_ee)); | |
4115 | D_ASSERT(list_empty(&mdev->sync_ee)); | |
4116 | D_ASSERT(list_empty(&mdev->done_ee)); | |
4117 | ||
4118 | /* ok, no more ee's on the fly, it is safe to reset the epoch_size */ | |
4119 | atomic_set(&mdev->current_epoch->epoch_size, 0); | |
4120 | D_ASSERT(list_empty(&mdev->current_epoch->list)); | |
360cc740 PR |
4121 | |
4122 | return 0; | |
b411b363 PR |
4123 | } |
4124 | ||
4125 | /* | |
4126 | * We support PRO_VERSION_MIN to PRO_VERSION_MAX. The protocol version | |
4127 | * we can agree on is stored in agreed_pro_version. | |
4128 | * | |
4129 | * feature flags and the reserved array should be enough room for future | |
4130 | * enhancements of the handshake protocol, and possible plugins... | |
4131 | * | |
4132 | * for now, they are expected to be zero, but ignored. | |
4133 | */ | |
8a22cccc | 4134 | static int drbd_send_handshake(struct drbd_tconn *tconn) |
b411b363 | 4135 | { |
e6b3ea83 | 4136 | /* ASSERT current == mdev->tconn->receiver ... */ |
8a22cccc | 4137 | struct p_handshake *p = &tconn->data.sbuf.handshake; |
b411b363 PR |
4138 | int ok; |
4139 | ||
8a22cccc PR |
4140 | if (mutex_lock_interruptible(&tconn->data.mutex)) { |
4141 | conn_err(tconn, "interrupted during initial handshake\n"); | |
b411b363 PR |
4142 | return 0; /* interrupted. not ok. */ |
4143 | } | |
4144 | ||
8a22cccc PR |
4145 | if (tconn->data.socket == NULL) { |
4146 | mutex_unlock(&tconn->data.mutex); | |
b411b363 PR |
4147 | return 0; |
4148 | } | |
4149 | ||
4150 | memset(p, 0, sizeof(*p)); | |
4151 | p->protocol_min = cpu_to_be32(PRO_VERSION_MIN); | |
4152 | p->protocol_max = cpu_to_be32(PRO_VERSION_MAX); | |
8a22cccc PR |
4153 | ok = _conn_send_cmd(tconn, 0, tconn->data.socket, P_HAND_SHAKE, |
4154 | &p->head, sizeof(*p), 0); | |
4155 | mutex_unlock(&tconn->data.mutex); | |
b411b363 PR |
4156 | return ok; |
4157 | } | |
4158 | ||
4159 | /* | |
4160 | * return values: | |
4161 | * 1 yes, we have a valid connection | |
4162 | * 0 oops, did not work out, please try again | |
4163 | * -1 peer talks different language, | |
4164 | * no point in trying again, please go standalone. | |
4165 | */ | |
65d11ed6 | 4166 | static int drbd_do_handshake(struct drbd_tconn *tconn) |
b411b363 | 4167 | { |
65d11ed6 PR |
4168 | /* ASSERT current == tconn->receiver ... */ |
4169 | struct p_handshake *p = &tconn->data.rbuf.handshake; | |
02918be2 | 4170 | const int expect = sizeof(struct p_handshake) - sizeof(struct p_header80); |
77351055 | 4171 | struct packet_info pi; |
b411b363 PR |
4172 | int rv; |
4173 | ||
65d11ed6 | 4174 | rv = drbd_send_handshake(tconn); |
b411b363 PR |
4175 | if (!rv) |
4176 | return 0; | |
4177 | ||
65d11ed6 | 4178 | rv = drbd_recv_header(tconn, &pi); |
b411b363 PR |
4179 | if (!rv) |
4180 | return 0; | |
4181 | ||
77351055 | 4182 | if (pi.cmd != P_HAND_SHAKE) { |
65d11ed6 | 4183 | conn_err(tconn, "expected HandShake packet, received: %s (0x%04x)\n", |
77351055 | 4184 | cmdname(pi.cmd), pi.cmd); |
b411b363 PR |
4185 | return -1; |
4186 | } | |
4187 | ||
77351055 | 4188 | if (pi.size != expect) { |
65d11ed6 | 4189 | conn_err(tconn, "expected HandShake length: %u, received: %u\n", |
77351055 | 4190 | expect, pi.size); |
b411b363 PR |
4191 | return -1; |
4192 | } | |
4193 | ||
65d11ed6 | 4194 | rv = drbd_recv(tconn, &p->head.payload, expect); |
b411b363 PR |
4195 | |
4196 | if (rv != expect) { | |
0ddc5549 | 4197 | if (!signal_pending(current)) |
65d11ed6 | 4198 | conn_warn(tconn, "short read receiving handshake packet: l=%u\n", rv); |
b411b363 PR |
4199 | return 0; |
4200 | } | |
4201 | ||
b411b363 PR |
4202 | p->protocol_min = be32_to_cpu(p->protocol_min); |
4203 | p->protocol_max = be32_to_cpu(p->protocol_max); | |
4204 | if (p->protocol_max == 0) | |
4205 | p->protocol_max = p->protocol_min; | |
4206 | ||
4207 | if (PRO_VERSION_MAX < p->protocol_min || | |
4208 | PRO_VERSION_MIN > p->protocol_max) | |
4209 | goto incompat; | |
4210 | ||
65d11ed6 | 4211 | tconn->agreed_pro_version = min_t(int, PRO_VERSION_MAX, p->protocol_max); |
b411b363 | 4212 | |
65d11ed6 PR |
4213 | conn_info(tconn, "Handshake successful: " |
4214 | "Agreed network protocol version %d\n", tconn->agreed_pro_version); | |
b411b363 PR |
4215 | |
4216 | return 1; | |
4217 | ||
4218 | incompat: | |
65d11ed6 | 4219 | conn_err(tconn, "incompatible DRBD dialects: " |
b411b363 PR |
4220 | "I support %d-%d, peer supports %d-%d\n", |
4221 | PRO_VERSION_MIN, PRO_VERSION_MAX, | |
4222 | p->protocol_min, p->protocol_max); | |
4223 | return -1; | |
4224 | } | |
4225 | ||
4226 | #if !defined(CONFIG_CRYPTO_HMAC) && !defined(CONFIG_CRYPTO_HMAC_MODULE) | |
13e6037d | 4227 | static int drbd_do_auth(struct drbd_tconn *tconn) |
b411b363 PR |
4228 | { |
4229 | dev_err(DEV, "This kernel was build without CONFIG_CRYPTO_HMAC.\n"); | |
4230 | dev_err(DEV, "You need to disable 'cram-hmac-alg' in drbd.conf.\n"); | |
b10d96cb | 4231 | return -1; |
b411b363 PR |
4232 | } |
4233 | #else | |
4234 | #define CHALLENGE_LEN 64 | |
b10d96cb JT |
4235 | |
4236 | /* Return value: | |
4237 | 1 - auth succeeded, | |
4238 | 0 - failed, try again (network error), | |
4239 | -1 - auth failed, don't try again. | |
4240 | */ | |
4241 | ||
13e6037d | 4242 | static int drbd_do_auth(struct drbd_tconn *tconn) |
b411b363 PR |
4243 | { |
4244 | char my_challenge[CHALLENGE_LEN]; /* 64 Bytes... */ | |
4245 | struct scatterlist sg; | |
4246 | char *response = NULL; | |
4247 | char *right_response = NULL; | |
4248 | char *peers_ch = NULL; | |
13e6037d | 4249 | unsigned int key_len = strlen(tconn->net_conf->shared_secret); |
b411b363 PR |
4250 | unsigned int resp_size; |
4251 | struct hash_desc desc; | |
77351055 | 4252 | struct packet_info pi; |
b411b363 PR |
4253 | int rv; |
4254 | ||
13e6037d | 4255 | desc.tfm = tconn->cram_hmac_tfm; |
b411b363 PR |
4256 | desc.flags = 0; |
4257 | ||
13e6037d PR |
4258 | rv = crypto_hash_setkey(tconn->cram_hmac_tfm, |
4259 | (u8 *)tconn->net_conf->shared_secret, key_len); | |
b411b363 | 4260 | if (rv) { |
13e6037d | 4261 | conn_err(tconn, "crypto_hash_setkey() failed with %d\n", rv); |
b10d96cb | 4262 | rv = -1; |
b411b363 PR |
4263 | goto fail; |
4264 | } | |
4265 | ||
4266 | get_random_bytes(my_challenge, CHALLENGE_LEN); | |
4267 | ||
13e6037d | 4268 | rv = conn_send_cmd2(tconn, P_AUTH_CHALLENGE, my_challenge, CHALLENGE_LEN); |
b411b363 PR |
4269 | if (!rv) |
4270 | goto fail; | |
4271 | ||
13e6037d | 4272 | rv = drbd_recv_header(tconn, &pi); |
b411b363 PR |
4273 | if (!rv) |
4274 | goto fail; | |
4275 | ||
77351055 | 4276 | if (pi.cmd != P_AUTH_CHALLENGE) { |
13e6037d | 4277 | conn_err(tconn, "expected AuthChallenge packet, received: %s (0x%04x)\n", |
77351055 | 4278 | cmdname(pi.cmd), pi.cmd); |
b411b363 PR |
4279 | rv = 0; |
4280 | goto fail; | |
4281 | } | |
4282 | ||
77351055 | 4283 | if (pi.size > CHALLENGE_LEN * 2) { |
13e6037d | 4284 | conn_err(tconn, "expected AuthChallenge payload too big.\n"); |
b10d96cb | 4285 | rv = -1; |
b411b363 PR |
4286 | goto fail; |
4287 | } | |
4288 | ||
77351055 | 4289 | peers_ch = kmalloc(pi.size, GFP_NOIO); |
b411b363 | 4290 | if (peers_ch == NULL) { |
13e6037d | 4291 | conn_err(tconn, "kmalloc of peers_ch failed\n"); |
b10d96cb | 4292 | rv = -1; |
b411b363 PR |
4293 | goto fail; |
4294 | } | |
4295 | ||
13e6037d | 4296 | rv = drbd_recv(tconn, peers_ch, pi.size); |
b411b363 | 4297 | |
77351055 | 4298 | if (rv != pi.size) { |
0ddc5549 | 4299 | if (!signal_pending(current)) |
13e6037d | 4300 | conn_warn(tconn, "short read AuthChallenge: l=%u\n", rv); |
b411b363 PR |
4301 | rv = 0; |
4302 | goto fail; | |
4303 | } | |
4304 | ||
13e6037d | 4305 | resp_size = crypto_hash_digestsize(tconn->cram_hmac_tfm); |
b411b363 PR |
4306 | response = kmalloc(resp_size, GFP_NOIO); |
4307 | if (response == NULL) { | |
13e6037d | 4308 | conn_err(tconn, "kmalloc of response failed\n"); |
b10d96cb | 4309 | rv = -1; |
b411b363 PR |
4310 | goto fail; |
4311 | } | |
4312 | ||
4313 | sg_init_table(&sg, 1); | |
77351055 | 4314 | sg_set_buf(&sg, peers_ch, pi.size); |
b411b363 PR |
4315 | |
4316 | rv = crypto_hash_digest(&desc, &sg, sg.length, response); | |
4317 | if (rv) { | |
13e6037d | 4318 | conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv); |
b10d96cb | 4319 | rv = -1; |
b411b363 PR |
4320 | goto fail; |
4321 | } | |
4322 | ||
13e6037d | 4323 | rv = conn_send_cmd2(tconn, P_AUTH_RESPONSE, response, resp_size); |
b411b363 PR |
4324 | if (!rv) |
4325 | goto fail; | |
4326 | ||
13e6037d | 4327 | rv = drbd_recv_header(tconn, &pi); |
b411b363 PR |
4328 | if (!rv) |
4329 | goto fail; | |
4330 | ||
77351055 | 4331 | if (pi.cmd != P_AUTH_RESPONSE) { |
13e6037d | 4332 | conn_err(tconn, "expected AuthResponse packet, received: %s (0x%04x)\n", |
77351055 | 4333 | cmdname(pi.cmd), pi.cmd); |
b411b363 PR |
4334 | rv = 0; |
4335 | goto fail; | |
4336 | } | |
4337 | ||
77351055 | 4338 | if (pi.size != resp_size) { |
13e6037d | 4339 | conn_err(tconn, "expected AuthResponse payload of wrong size\n"); |
b411b363 PR |
4340 | rv = 0; |
4341 | goto fail; | |
4342 | } | |
4343 | ||
13e6037d | 4344 | rv = drbd_recv(tconn, response , resp_size); |
b411b363 PR |
4345 | |
4346 | if (rv != resp_size) { | |
0ddc5549 | 4347 | if (!signal_pending(current)) |
13e6037d | 4348 | conn_warn(tconn, "short read receiving AuthResponse: l=%u\n", rv); |
b411b363 PR |
4349 | rv = 0; |
4350 | goto fail; | |
4351 | } | |
4352 | ||
4353 | right_response = kmalloc(resp_size, GFP_NOIO); | |
2d1ee87d | 4354 | if (right_response == NULL) { |
13e6037d | 4355 | conn_err(tconn, "kmalloc of right_response failed\n"); |
b10d96cb | 4356 | rv = -1; |
b411b363 PR |
4357 | goto fail; |
4358 | } | |
4359 | ||
4360 | sg_set_buf(&sg, my_challenge, CHALLENGE_LEN); | |
4361 | ||
4362 | rv = crypto_hash_digest(&desc, &sg, sg.length, right_response); | |
4363 | if (rv) { | |
13e6037d | 4364 | conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv); |
b10d96cb | 4365 | rv = -1; |
b411b363 PR |
4366 | goto fail; |
4367 | } | |
4368 | ||
4369 | rv = !memcmp(response, right_response, resp_size); | |
4370 | ||
4371 | if (rv) | |
13e6037d PR |
4372 | conn_info(tconn, "Peer authenticated using %d bytes of '%s' HMAC\n", |
4373 | resp_size, tconn->net_conf->cram_hmac_alg); | |
b10d96cb JT |
4374 | else |
4375 | rv = -1; | |
b411b363 PR |
4376 | |
4377 | fail: | |
4378 | kfree(peers_ch); | |
4379 | kfree(response); | |
4380 | kfree(right_response); | |
4381 | ||
4382 | return rv; | |
4383 | } | |
4384 | #endif | |
4385 | ||
4386 | int drbdd_init(struct drbd_thread *thi) | |
4387 | { | |
392c8801 | 4388 | struct drbd_tconn *tconn = thi->tconn; |
b411b363 PR |
4389 | int h; |
4390 | ||
4d641dd7 | 4391 | conn_info(tconn, "receiver (re)started\n"); |
b411b363 PR |
4392 | |
4393 | do { | |
4d641dd7 | 4394 | h = drbd_connect(tconn); |
b411b363 | 4395 | if (h == 0) { |
4d641dd7 | 4396 | drbd_disconnect(tconn); |
20ee6390 | 4397 | schedule_timeout_interruptible(HZ); |
b411b363 PR |
4398 | } |
4399 | if (h == -1) { | |
4d641dd7 | 4400 | conn_warn(tconn, "Discarding network configuration.\n"); |
bbeb641c | 4401 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 PR |
4402 | } |
4403 | } while (h == 0); | |
4404 | ||
4405 | if (h > 0) { | |
4d641dd7 PR |
4406 | if (get_net_conf(tconn)) { |
4407 | drbdd(tconn); | |
4408 | put_net_conf(tconn); | |
b411b363 PR |
4409 | } |
4410 | } | |
4411 | ||
4d641dd7 | 4412 | drbd_disconnect(tconn); |
b411b363 | 4413 | |
4d641dd7 | 4414 | conn_info(tconn, "receiver terminated\n"); |
b411b363 PR |
4415 | return 0; |
4416 | } | |
4417 | ||
4418 | /* ********* acknowledge sender ******** */ | |
4419 | ||
e4f78ede PR |
4420 | static int got_conn_RqSReply(struct drbd_tconn *tconn, enum drbd_packet cmd) |
4421 | { | |
4422 | struct p_req_state_reply *p = &tconn->meta.rbuf.req_state_reply; | |
4423 | int retcode = be32_to_cpu(p->retcode); | |
4424 | ||
4425 | if (retcode >= SS_SUCCESS) { | |
4426 | set_bit(CONN_WD_ST_CHG_OKAY, &tconn->flags); | |
4427 | } else { | |
4428 | set_bit(CONN_WD_ST_CHG_FAIL, &tconn->flags); | |
4429 | conn_err(tconn, "Requested state change failed by peer: %s (%d)\n", | |
4430 | drbd_set_st_err_str(retcode), retcode); | |
4431 | } | |
4432 | wake_up(&tconn->ping_wait); | |
4433 | ||
4434 | return true; | |
4435 | } | |
4436 | ||
d8763023 | 4437 | static int got_RqSReply(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4438 | { |
257d0af6 | 4439 | struct p_req_state_reply *p = &mdev->tconn->meta.rbuf.req_state_reply; |
b411b363 PR |
4440 | int retcode = be32_to_cpu(p->retcode); |
4441 | ||
e4f78ede PR |
4442 | if (retcode >= SS_SUCCESS) { |
4443 | set_bit(CL_ST_CHG_SUCCESS, &mdev->flags); | |
4444 | } else { | |
4445 | set_bit(CL_ST_CHG_FAIL, &mdev->flags); | |
4446 | dev_err(DEV, "Requested state change failed by peer: %s (%d)\n", | |
4447 | drbd_set_st_err_str(retcode), retcode); | |
b411b363 | 4448 | } |
e4f78ede PR |
4449 | wake_up(&mdev->state_wait); |
4450 | ||
81e84650 | 4451 | return true; |
b411b363 PR |
4452 | } |
4453 | ||
f19e4f8b | 4454 | static int got_Ping(struct drbd_tconn *tconn, enum drbd_packet cmd) |
b411b363 | 4455 | { |
f19e4f8b | 4456 | return drbd_send_ping_ack(tconn); |
b411b363 PR |
4457 | |
4458 | } | |
4459 | ||
f19e4f8b | 4460 | static int got_PingAck(struct drbd_tconn *tconn, enum drbd_packet cmd) |
b411b363 PR |
4461 | { |
4462 | /* restore idle timeout */ | |
2a67d8b9 PR |
4463 | tconn->meta.socket->sk->sk_rcvtimeo = tconn->net_conf->ping_int*HZ; |
4464 | if (!test_and_set_bit(GOT_PING_ACK, &tconn->flags)) | |
4465 | wake_up(&tconn->ping_wait); | |
b411b363 | 4466 | |
81e84650 | 4467 | return true; |
b411b363 PR |
4468 | } |
4469 | ||
d8763023 | 4470 | static int got_IsInSync(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4471 | { |
257d0af6 | 4472 | struct p_block_ack *p = &mdev->tconn->meta.rbuf.block_ack; |
b411b363 PR |
4473 | sector_t sector = be64_to_cpu(p->sector); |
4474 | int blksize = be32_to_cpu(p->blksize); | |
4475 | ||
31890f4a | 4476 | D_ASSERT(mdev->tconn->agreed_pro_version >= 89); |
b411b363 PR |
4477 | |
4478 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4479 | ||
1d53f09e LE |
4480 | if (get_ldev(mdev)) { |
4481 | drbd_rs_complete_io(mdev, sector); | |
4482 | drbd_set_in_sync(mdev, sector, blksize); | |
4483 | /* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */ | |
4484 | mdev->rs_same_csum += (blksize >> BM_BLOCK_SHIFT); | |
4485 | put_ldev(mdev); | |
4486 | } | |
b411b363 | 4487 | dec_rs_pending(mdev); |
778f271d | 4488 | atomic_add(blksize >> 9, &mdev->rs_sect_in); |
b411b363 | 4489 | |
81e84650 | 4490 | return true; |
b411b363 PR |
4491 | } |
4492 | ||
bc9c5c41 AG |
4493 | static int |
4494 | validate_req_change_req_state(struct drbd_conf *mdev, u64 id, sector_t sector, | |
4495 | struct rb_root *root, const char *func, | |
4496 | enum drbd_req_event what, bool missing_ok) | |
b411b363 PR |
4497 | { |
4498 | struct drbd_request *req; | |
4499 | struct bio_and_error m; | |
4500 | ||
87eeee41 | 4501 | spin_lock_irq(&mdev->tconn->req_lock); |
bc9c5c41 | 4502 | req = find_request(mdev, root, id, sector, missing_ok, func); |
b411b363 | 4503 | if (unlikely(!req)) { |
87eeee41 | 4504 | spin_unlock_irq(&mdev->tconn->req_lock); |
81e84650 | 4505 | return false; |
b411b363 PR |
4506 | } |
4507 | __req_mod(req, what, &m); | |
87eeee41 | 4508 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
4509 | |
4510 | if (m.bio) | |
4511 | complete_master_bio(mdev, &m); | |
81e84650 | 4512 | return true; |
b411b363 PR |
4513 | } |
4514 | ||
d8763023 | 4515 | static int got_BlockAck(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4516 | { |
257d0af6 | 4517 | struct p_block_ack *p = &mdev->tconn->meta.rbuf.block_ack; |
b411b363 PR |
4518 | sector_t sector = be64_to_cpu(p->sector); |
4519 | int blksize = be32_to_cpu(p->blksize); | |
4520 | enum drbd_req_event what; | |
4521 | ||
4522 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4523 | ||
579b57ed | 4524 | if (p->block_id == ID_SYNCER) { |
b411b363 PR |
4525 | drbd_set_in_sync(mdev, sector, blksize); |
4526 | dec_rs_pending(mdev); | |
81e84650 | 4527 | return true; |
b411b363 | 4528 | } |
257d0af6 | 4529 | switch (cmd) { |
b411b363 | 4530 | case P_RS_WRITE_ACK: |
89e58e75 | 4531 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); |
8554df1c | 4532 | what = WRITE_ACKED_BY_PEER_AND_SIS; |
b411b363 PR |
4533 | break; |
4534 | case P_WRITE_ACK: | |
89e58e75 | 4535 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); |
8554df1c | 4536 | what = WRITE_ACKED_BY_PEER; |
b411b363 PR |
4537 | break; |
4538 | case P_RECV_ACK: | |
89e58e75 | 4539 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_B); |
8554df1c | 4540 | what = RECV_ACKED_BY_PEER; |
b411b363 | 4541 | break; |
7be8da07 | 4542 | case P_DISCARD_WRITE: |
89e58e75 | 4543 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); |
7be8da07 AG |
4544 | what = DISCARD_WRITE; |
4545 | break; | |
4546 | case P_RETRY_WRITE: | |
4547 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); | |
4548 | what = POSTPONE_WRITE; | |
b411b363 PR |
4549 | break; |
4550 | default: | |
4551 | D_ASSERT(0); | |
81e84650 | 4552 | return false; |
b411b363 PR |
4553 | } |
4554 | ||
4555 | return validate_req_change_req_state(mdev, p->block_id, sector, | |
bc9c5c41 AG |
4556 | &mdev->write_requests, __func__, |
4557 | what, false); | |
b411b363 PR |
4558 | } |
4559 | ||
d8763023 | 4560 | static int got_NegAck(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4561 | { |
257d0af6 | 4562 | struct p_block_ack *p = &mdev->tconn->meta.rbuf.block_ack; |
b411b363 | 4563 | sector_t sector = be64_to_cpu(p->sector); |
2deb8336 | 4564 | int size = be32_to_cpu(p->blksize); |
89e58e75 PR |
4565 | bool missing_ok = mdev->tconn->net_conf->wire_protocol == DRBD_PROT_A || |
4566 | mdev->tconn->net_conf->wire_protocol == DRBD_PROT_B; | |
c3afd8f5 | 4567 | bool found; |
b411b363 PR |
4568 | |
4569 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4570 | ||
579b57ed | 4571 | if (p->block_id == ID_SYNCER) { |
b411b363 PR |
4572 | dec_rs_pending(mdev); |
4573 | drbd_rs_failed_io(mdev, sector, size); | |
81e84650 | 4574 | return true; |
b411b363 | 4575 | } |
2deb8336 | 4576 | |
c3afd8f5 | 4577 | found = validate_req_change_req_state(mdev, p->block_id, sector, |
bc9c5c41 | 4578 | &mdev->write_requests, __func__, |
8554df1c | 4579 | NEG_ACKED, missing_ok); |
c3afd8f5 AG |
4580 | if (!found) { |
4581 | /* Protocol A has no P_WRITE_ACKs, but has P_NEG_ACKs. | |
4582 | The master bio might already be completed, therefore the | |
4583 | request is no longer in the collision hash. */ | |
4584 | /* In Protocol B we might already have got a P_RECV_ACK | |
4585 | but then get a P_NEG_ACK afterwards. */ | |
4586 | if (!missing_ok) | |
2deb8336 | 4587 | return false; |
c3afd8f5 | 4588 | drbd_set_out_of_sync(mdev, sector, size); |
2deb8336 | 4589 | } |
2deb8336 | 4590 | return true; |
b411b363 PR |
4591 | } |
4592 | ||
d8763023 | 4593 | static int got_NegDReply(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4594 | { |
257d0af6 | 4595 | struct p_block_ack *p = &mdev->tconn->meta.rbuf.block_ack; |
b411b363 PR |
4596 | sector_t sector = be64_to_cpu(p->sector); |
4597 | ||
4598 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
7be8da07 | 4599 | |
b411b363 PR |
4600 | dev_err(DEV, "Got NegDReply; Sector %llus, len %u; Fail original request.\n", |
4601 | (unsigned long long)sector, be32_to_cpu(p->blksize)); | |
4602 | ||
4603 | return validate_req_change_req_state(mdev, p->block_id, sector, | |
bc9c5c41 | 4604 | &mdev->read_requests, __func__, |
8554df1c | 4605 | NEG_ACKED, false); |
b411b363 PR |
4606 | } |
4607 | ||
d8763023 | 4608 | static int got_NegRSDReply(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 PR |
4609 | { |
4610 | sector_t sector; | |
4611 | int size; | |
257d0af6 | 4612 | struct p_block_ack *p = &mdev->tconn->meta.rbuf.block_ack; |
b411b363 PR |
4613 | |
4614 | sector = be64_to_cpu(p->sector); | |
4615 | size = be32_to_cpu(p->blksize); | |
b411b363 PR |
4616 | |
4617 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4618 | ||
4619 | dec_rs_pending(mdev); | |
4620 | ||
4621 | if (get_ldev_if_state(mdev, D_FAILED)) { | |
4622 | drbd_rs_complete_io(mdev, sector); | |
257d0af6 | 4623 | switch (cmd) { |
d612d309 PR |
4624 | case P_NEG_RS_DREPLY: |
4625 | drbd_rs_failed_io(mdev, sector, size); | |
4626 | case P_RS_CANCEL: | |
4627 | break; | |
4628 | default: | |
4629 | D_ASSERT(0); | |
4630 | put_ldev(mdev); | |
4631 | return false; | |
4632 | } | |
b411b363 PR |
4633 | put_ldev(mdev); |
4634 | } | |
4635 | ||
81e84650 | 4636 | return true; |
b411b363 PR |
4637 | } |
4638 | ||
d8763023 | 4639 | static int got_BarrierAck(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4640 | { |
257d0af6 | 4641 | struct p_barrier_ack *p = &mdev->tconn->meta.rbuf.barrier_ack; |
b411b363 | 4642 | |
2f5cdd0b | 4643 | tl_release(mdev->tconn, p->barrier, be32_to_cpu(p->set_size)); |
b411b363 | 4644 | |
c4752ef1 PR |
4645 | if (mdev->state.conn == C_AHEAD && |
4646 | atomic_read(&mdev->ap_in_flight) == 0 && | |
370a43e7 PR |
4647 | !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &mdev->current_epoch->flags)) { |
4648 | mdev->start_resync_timer.expires = jiffies + HZ; | |
4649 | add_timer(&mdev->start_resync_timer); | |
c4752ef1 PR |
4650 | } |
4651 | ||
81e84650 | 4652 | return true; |
b411b363 PR |
4653 | } |
4654 | ||
d8763023 | 4655 | static int got_OVResult(struct drbd_conf *mdev, enum drbd_packet cmd) |
b411b363 | 4656 | { |
257d0af6 | 4657 | struct p_block_ack *p = &mdev->tconn->meta.rbuf.block_ack; |
b411b363 PR |
4658 | struct drbd_work *w; |
4659 | sector_t sector; | |
4660 | int size; | |
4661 | ||
4662 | sector = be64_to_cpu(p->sector); | |
4663 | size = be32_to_cpu(p->blksize); | |
4664 | ||
4665 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4666 | ||
4667 | if (be64_to_cpu(p->block_id) == ID_OUT_OF_SYNC) | |
4668 | drbd_ov_oos_found(mdev, sector, size); | |
4669 | else | |
4670 | ov_oos_print(mdev); | |
4671 | ||
1d53f09e | 4672 | if (!get_ldev(mdev)) |
81e84650 | 4673 | return true; |
1d53f09e | 4674 | |
b411b363 PR |
4675 | drbd_rs_complete_io(mdev, sector); |
4676 | dec_rs_pending(mdev); | |
4677 | ||
ea5442af LE |
4678 | --mdev->ov_left; |
4679 | ||
4680 | /* let's advance progress step marks only for every other megabyte */ | |
4681 | if ((mdev->ov_left & 0x200) == 0x200) | |
4682 | drbd_advance_rs_marks(mdev, mdev->ov_left); | |
4683 | ||
4684 | if (mdev->ov_left == 0) { | |
b411b363 PR |
4685 | w = kmalloc(sizeof(*w), GFP_NOIO); |
4686 | if (w) { | |
4687 | w->cb = w_ov_finished; | |
a21e9298 | 4688 | w->mdev = mdev; |
e42325a5 | 4689 | drbd_queue_work_front(&mdev->tconn->data.work, w); |
b411b363 PR |
4690 | } else { |
4691 | dev_err(DEV, "kmalloc(w) failed."); | |
4692 | ov_oos_print(mdev); | |
4693 | drbd_resync_finished(mdev); | |
4694 | } | |
4695 | } | |
1d53f09e | 4696 | put_ldev(mdev); |
81e84650 | 4697 | return true; |
b411b363 PR |
4698 | } |
4699 | ||
d8763023 | 4700 | static int got_skip(struct drbd_conf *mdev, enum drbd_packet cmd) |
0ced55a3 | 4701 | { |
81e84650 | 4702 | return true; |
0ced55a3 PR |
4703 | } |
4704 | ||
32862ec7 PR |
4705 | static int tconn_process_done_ee(struct drbd_tconn *tconn) |
4706 | { | |
082a3439 PR |
4707 | struct drbd_conf *mdev; |
4708 | int i, not_empty = 0; | |
32862ec7 PR |
4709 | |
4710 | do { | |
4711 | clear_bit(SIGNAL_ASENDER, &tconn->flags); | |
4712 | flush_signals(current); | |
082a3439 PR |
4713 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
4714 | if (!drbd_process_done_ee(mdev)) | |
4715 | return 1; /* error */ | |
4716 | } | |
32862ec7 | 4717 | set_bit(SIGNAL_ASENDER, &tconn->flags); |
082a3439 PR |
4718 | |
4719 | spin_lock_irq(&tconn->req_lock); | |
4720 | idr_for_each_entry(&tconn->volumes, mdev, i) { | |
4721 | not_empty = !list_empty(&mdev->done_ee); | |
4722 | if (not_empty) | |
4723 | break; | |
4724 | } | |
4725 | spin_unlock_irq(&tconn->req_lock); | |
32862ec7 PR |
4726 | } while (not_empty); |
4727 | ||
4728 | return 0; | |
4729 | } | |
4730 | ||
7201b972 AG |
4731 | struct asender_cmd { |
4732 | size_t pkt_size; | |
a4fbda8e PR |
4733 | enum mdev_or_conn fa_type; /* first argument's type */ |
4734 | union { | |
4735 | int (*mdev_fn)(struct drbd_conf *mdev, enum drbd_packet cmd); | |
4736 | int (*conn_fn)(struct drbd_tconn *tconn, enum drbd_packet cmd); | |
4737 | }; | |
7201b972 AG |
4738 | }; |
4739 | ||
4740 | static struct asender_cmd asender_tbl[] = { | |
f19e4f8b PR |
4741 | [P_PING] = { sizeof(struct p_header), CONN, { .conn_fn = got_Ping } }, |
4742 | [P_PING_ACK] = { sizeof(struct p_header), CONN, { .conn_fn = got_PingAck } }, | |
a4fbda8e PR |
4743 | [P_RECV_ACK] = { sizeof(struct p_block_ack), MDEV, { got_BlockAck } }, |
4744 | [P_WRITE_ACK] = { sizeof(struct p_block_ack), MDEV, { got_BlockAck } }, | |
4745 | [P_RS_WRITE_ACK] = { sizeof(struct p_block_ack), MDEV, { got_BlockAck } }, | |
4746 | [P_DISCARD_WRITE] = { sizeof(struct p_block_ack), MDEV, { got_BlockAck } }, | |
4747 | [P_NEG_ACK] = { sizeof(struct p_block_ack), MDEV, { got_NegAck } }, | |
4748 | [P_NEG_DREPLY] = { sizeof(struct p_block_ack), MDEV, { got_NegDReply } }, | |
4749 | [P_NEG_RS_DREPLY] = { sizeof(struct p_block_ack), MDEV, { got_NegRSDReply } }, | |
4750 | [P_OV_RESULT] = { sizeof(struct p_block_ack), MDEV, { got_OVResult } }, | |
4751 | [P_BARRIER_ACK] = { sizeof(struct p_barrier_ack), MDEV, { got_BarrierAck } }, | |
4752 | [P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), MDEV, { got_RqSReply } }, | |
4753 | [P_RS_IS_IN_SYNC] = { sizeof(struct p_block_ack), MDEV, { got_IsInSync } }, | |
4754 | [P_DELAY_PROBE] = { sizeof(struct p_delay_probe93), MDEV, { got_skip } }, | |
4755 | [P_RS_CANCEL] = { sizeof(struct p_block_ack), MDEV, { got_NegRSDReply } }, | |
e4f78ede | 4756 | [P_CONN_ST_CHG_REPLY]={ sizeof(struct p_req_state_reply), CONN, {.conn_fn = got_conn_RqSReply}}, |
a4fbda8e | 4757 | [P_RETRY_WRITE] = { sizeof(struct p_block_ack), MDEV, { got_BlockAck } }, |
7201b972 AG |
4758 | }; |
4759 | ||
b411b363 PR |
4760 | int drbd_asender(struct drbd_thread *thi) |
4761 | { | |
392c8801 | 4762 | struct drbd_tconn *tconn = thi->tconn; |
32862ec7 | 4763 | struct p_header *h = &tconn->meta.rbuf.header; |
b411b363 | 4764 | struct asender_cmd *cmd = NULL; |
77351055 | 4765 | struct packet_info pi; |
257d0af6 | 4766 | int rv; |
b411b363 PR |
4767 | void *buf = h; |
4768 | int received = 0; | |
257d0af6 | 4769 | int expect = sizeof(struct p_header); |
f36af18c | 4770 | int ping_timeout_active = 0; |
b411b363 | 4771 | |
b411b363 PR |
4772 | current->policy = SCHED_RR; /* Make this a realtime task! */ |
4773 | current->rt_priority = 2; /* more important than all other tasks */ | |
4774 | ||
e77a0a5c | 4775 | while (get_t_state(thi) == RUNNING) { |
80822284 | 4776 | drbd_thread_current_set_cpu(thi); |
32862ec7 | 4777 | if (test_and_clear_bit(SEND_PING, &tconn->flags)) { |
2a67d8b9 | 4778 | if (!drbd_send_ping(tconn)) { |
32862ec7 | 4779 | conn_err(tconn, "drbd_send_ping has failed\n"); |
841ce241 AG |
4780 | goto reconnect; |
4781 | } | |
32862ec7 PR |
4782 | tconn->meta.socket->sk->sk_rcvtimeo = |
4783 | tconn->net_conf->ping_timeo*HZ/10; | |
f36af18c | 4784 | ping_timeout_active = 1; |
b411b363 PR |
4785 | } |
4786 | ||
32862ec7 PR |
4787 | /* TODO: conditionally cork; it may hurt latency if we cork without |
4788 | much to send */ | |
4789 | if (!tconn->net_conf->no_cork) | |
4790 | drbd_tcp_cork(tconn->meta.socket); | |
082a3439 PR |
4791 | if (tconn_process_done_ee(tconn)) { |
4792 | conn_err(tconn, "tconn_process_done_ee() failed\n"); | |
32862ec7 | 4793 | goto reconnect; |
082a3439 | 4794 | } |
b411b363 | 4795 | /* but unconditionally uncork unless disabled */ |
32862ec7 PR |
4796 | if (!tconn->net_conf->no_cork) |
4797 | drbd_tcp_uncork(tconn->meta.socket); | |
b411b363 PR |
4798 | |
4799 | /* short circuit, recv_msg would return EINTR anyways. */ | |
4800 | if (signal_pending(current)) | |
4801 | continue; | |
4802 | ||
32862ec7 PR |
4803 | rv = drbd_recv_short(tconn->meta.socket, buf, expect-received, 0); |
4804 | clear_bit(SIGNAL_ASENDER, &tconn->flags); | |
b411b363 PR |
4805 | |
4806 | flush_signals(current); | |
4807 | ||
4808 | /* Note: | |
4809 | * -EINTR (on meta) we got a signal | |
4810 | * -EAGAIN (on meta) rcvtimeo expired | |
4811 | * -ECONNRESET other side closed the connection | |
4812 | * -ERESTARTSYS (on data) we got a signal | |
4813 | * rv < 0 other than above: unexpected error! | |
4814 | * rv == expected: full header or command | |
4815 | * rv < expected: "woken" by signal during receive | |
4816 | * rv == 0 : "connection shut down by peer" | |
4817 | */ | |
4818 | if (likely(rv > 0)) { | |
4819 | received += rv; | |
4820 | buf += rv; | |
4821 | } else if (rv == 0) { | |
32862ec7 | 4822 | conn_err(tconn, "meta connection shut down by peer.\n"); |
b411b363 PR |
4823 | goto reconnect; |
4824 | } else if (rv == -EAGAIN) { | |
cb6518cb LE |
4825 | /* If the data socket received something meanwhile, |
4826 | * that is good enough: peer is still alive. */ | |
32862ec7 PR |
4827 | if (time_after(tconn->last_received, |
4828 | jiffies - tconn->meta.socket->sk->sk_rcvtimeo)) | |
cb6518cb | 4829 | continue; |
f36af18c | 4830 | if (ping_timeout_active) { |
32862ec7 | 4831 | conn_err(tconn, "PingAck did not arrive in time.\n"); |
b411b363 PR |
4832 | goto reconnect; |
4833 | } | |
32862ec7 | 4834 | set_bit(SEND_PING, &tconn->flags); |
b411b363 PR |
4835 | continue; |
4836 | } else if (rv == -EINTR) { | |
4837 | continue; | |
4838 | } else { | |
32862ec7 | 4839 | conn_err(tconn, "sock_recvmsg returned %d\n", rv); |
b411b363 PR |
4840 | goto reconnect; |
4841 | } | |
4842 | ||
4843 | if (received == expect && cmd == NULL) { | |
32862ec7 | 4844 | if (!decode_header(tconn, h, &pi)) |
b411b363 | 4845 | goto reconnect; |
7201b972 AG |
4846 | cmd = &asender_tbl[pi.cmd]; |
4847 | if (pi.cmd >= ARRAY_SIZE(asender_tbl) || !cmd) { | |
32862ec7 | 4848 | conn_err(tconn, "unknown command %d on meta (l: %d)\n", |
77351055 | 4849 | pi.cmd, pi.size); |
b411b363 PR |
4850 | goto disconnect; |
4851 | } | |
4852 | expect = cmd->pkt_size; | |
77351055 | 4853 | if (pi.size != expect - sizeof(struct p_header)) { |
32862ec7 | 4854 | conn_err(tconn, "Wrong packet size on meta (c: %d, l: %d)\n", |
77351055 | 4855 | pi.cmd, pi.size); |
b411b363 | 4856 | goto reconnect; |
257d0af6 | 4857 | } |
b411b363 PR |
4858 | } |
4859 | if (received == expect) { | |
a4fbda8e PR |
4860 | bool rv; |
4861 | ||
4862 | if (cmd->fa_type == CONN) { | |
4863 | rv = cmd->conn_fn(tconn, pi.cmd); | |
4864 | } else { | |
4865 | struct drbd_conf *mdev = vnr_to_mdev(tconn, pi.vnr); | |
4866 | rv = cmd->mdev_fn(mdev, pi.cmd); | |
4867 | } | |
4868 | ||
4869 | if (!rv) | |
b411b363 PR |
4870 | goto reconnect; |
4871 | ||
a4fbda8e PR |
4872 | tconn->last_received = jiffies; |
4873 | ||
f36af18c LE |
4874 | /* the idle_timeout (ping-int) |
4875 | * has been restored in got_PingAck() */ | |
7201b972 | 4876 | if (cmd == &asender_tbl[P_PING_ACK]) |
f36af18c LE |
4877 | ping_timeout_active = 0; |
4878 | ||
b411b363 PR |
4879 | buf = h; |
4880 | received = 0; | |
257d0af6 | 4881 | expect = sizeof(struct p_header); |
b411b363 PR |
4882 | cmd = NULL; |
4883 | } | |
4884 | } | |
4885 | ||
4886 | if (0) { | |
4887 | reconnect: | |
bbeb641c | 4888 | conn_request_state(tconn, NS(conn, C_NETWORK_FAILURE), CS_HARD); |
b411b363 PR |
4889 | } |
4890 | if (0) { | |
4891 | disconnect: | |
bbeb641c | 4892 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 | 4893 | } |
32862ec7 | 4894 | clear_bit(SIGNAL_ASENDER, &tconn->flags); |
b411b363 | 4895 | |
32862ec7 | 4896 | conn_info(tconn, "asender terminated\n"); |
b411b363 PR |
4897 | |
4898 | return 0; | |
4899 | } |