]>
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; | |
e2857216 AG |
53 | unsigned int size; |
54 | unsigned int vnr; | |
e658983a | 55 | void *data; |
77351055 PR |
56 | }; |
57 | ||
b411b363 PR |
58 | enum finish_epoch { |
59 | FE_STILL_LIVE, | |
60 | FE_DESTROYED, | |
61 | FE_RECYCLED, | |
62 | }; | |
63 | ||
6038178e | 64 | static int drbd_do_features(struct drbd_tconn *tconn); |
13e6037d | 65 | static int drbd_do_auth(struct drbd_tconn *tconn); |
360cc740 | 66 | static int drbd_disconnected(int vnr, void *p, void *data); |
b411b363 PR |
67 | |
68 | static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *, struct drbd_epoch *, enum epoch_event); | |
99920dc5 | 69 | static int e_end_block(struct drbd_work *, int); |
b411b363 | 70 | |
b411b363 PR |
71 | |
72 | #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN) | |
73 | ||
45bb912b LE |
74 | /* |
75 | * some helper functions to deal with single linked page lists, | |
76 | * page->private being our "next" pointer. | |
77 | */ | |
78 | ||
79 | /* If at least n pages are linked at head, get n pages off. | |
80 | * Otherwise, don't modify head, and return NULL. | |
81 | * Locking is the responsibility of the caller. | |
82 | */ | |
83 | static struct page *page_chain_del(struct page **head, int n) | |
84 | { | |
85 | struct page *page; | |
86 | struct page *tmp; | |
87 | ||
88 | BUG_ON(!n); | |
89 | BUG_ON(!head); | |
90 | ||
91 | page = *head; | |
23ce4227 PR |
92 | |
93 | if (!page) | |
94 | return NULL; | |
95 | ||
45bb912b LE |
96 | while (page) { |
97 | tmp = page_chain_next(page); | |
98 | if (--n == 0) | |
99 | break; /* found sufficient pages */ | |
100 | if (tmp == NULL) | |
101 | /* insufficient pages, don't use any of them. */ | |
102 | return NULL; | |
103 | page = tmp; | |
104 | } | |
105 | ||
106 | /* add end of list marker for the returned list */ | |
107 | set_page_private(page, 0); | |
108 | /* actual return value, and adjustment of head */ | |
109 | page = *head; | |
110 | *head = tmp; | |
111 | return page; | |
112 | } | |
113 | ||
114 | /* may be used outside of locks to find the tail of a (usually short) | |
115 | * "private" page chain, before adding it back to a global chain head | |
116 | * with page_chain_add() under a spinlock. */ | |
117 | static struct page *page_chain_tail(struct page *page, int *len) | |
118 | { | |
119 | struct page *tmp; | |
120 | int i = 1; | |
121 | while ((tmp = page_chain_next(page))) | |
122 | ++i, page = tmp; | |
123 | if (len) | |
124 | *len = i; | |
125 | return page; | |
126 | } | |
127 | ||
128 | static int page_chain_free(struct page *page) | |
129 | { | |
130 | struct page *tmp; | |
131 | int i = 0; | |
132 | page_chain_for_each_safe(page, tmp) { | |
133 | put_page(page); | |
134 | ++i; | |
135 | } | |
136 | return i; | |
137 | } | |
138 | ||
139 | static void page_chain_add(struct page **head, | |
140 | struct page *chain_first, struct page *chain_last) | |
141 | { | |
142 | #if 1 | |
143 | struct page *tmp; | |
144 | tmp = page_chain_tail(chain_first, NULL); | |
145 | BUG_ON(tmp != chain_last); | |
146 | #endif | |
147 | ||
148 | /* add chain to head */ | |
149 | set_page_private(chain_last, (unsigned long)*head); | |
150 | *head = chain_first; | |
151 | } | |
152 | ||
153 | static struct page *drbd_pp_first_pages_or_try_alloc(struct drbd_conf *mdev, int number) | |
b411b363 PR |
154 | { |
155 | struct page *page = NULL; | |
45bb912b LE |
156 | struct page *tmp = NULL; |
157 | int i = 0; | |
b411b363 PR |
158 | |
159 | /* Yes, testing drbd_pp_vacant outside the lock is racy. | |
160 | * So what. It saves a spin_lock. */ | |
45bb912b | 161 | if (drbd_pp_vacant >= number) { |
b411b363 | 162 | spin_lock(&drbd_pp_lock); |
45bb912b LE |
163 | page = page_chain_del(&drbd_pp_pool, number); |
164 | if (page) | |
165 | drbd_pp_vacant -= number; | |
b411b363 | 166 | spin_unlock(&drbd_pp_lock); |
45bb912b LE |
167 | if (page) |
168 | return page; | |
b411b363 | 169 | } |
45bb912b | 170 | |
b411b363 PR |
171 | /* GFP_TRY, because we must not cause arbitrary write-out: in a DRBD |
172 | * "criss-cross" setup, that might cause write-out on some other DRBD, | |
173 | * which in turn might block on the other node at this very place. */ | |
45bb912b LE |
174 | for (i = 0; i < number; i++) { |
175 | tmp = alloc_page(GFP_TRY); | |
176 | if (!tmp) | |
177 | break; | |
178 | set_page_private(tmp, (unsigned long)page); | |
179 | page = tmp; | |
180 | } | |
181 | ||
182 | if (i == number) | |
183 | return page; | |
184 | ||
185 | /* Not enough pages immediately available this time. | |
186 | * No need to jump around here, drbd_pp_alloc will retry this | |
187 | * function "soon". */ | |
188 | if (page) { | |
189 | tmp = page_chain_tail(page, NULL); | |
190 | spin_lock(&drbd_pp_lock); | |
191 | page_chain_add(&drbd_pp_pool, page, tmp); | |
192 | drbd_pp_vacant += i; | |
193 | spin_unlock(&drbd_pp_lock); | |
194 | } | |
195 | return NULL; | |
b411b363 PR |
196 | } |
197 | ||
a990be46 AG |
198 | static void reclaim_finished_net_peer_reqs(struct drbd_conf *mdev, |
199 | struct list_head *to_be_freed) | |
b411b363 | 200 | { |
db830c46 | 201 | struct drbd_peer_request *peer_req; |
b411b363 PR |
202 | struct list_head *le, *tle; |
203 | ||
204 | /* The EEs are always appended to the end of the list. Since | |
205 | they are sent in order over the wire, they have to finish | |
206 | in order. As soon as we see the first not finished we can | |
207 | stop to examine the list... */ | |
208 | ||
209 | list_for_each_safe(le, tle, &mdev->net_ee) { | |
db830c46 AG |
210 | peer_req = list_entry(le, struct drbd_peer_request, w.list); |
211 | if (drbd_ee_has_active_page(peer_req)) | |
b411b363 PR |
212 | break; |
213 | list_move(le, to_be_freed); | |
214 | } | |
215 | } | |
216 | ||
217 | static void drbd_kick_lo_and_reclaim_net(struct drbd_conf *mdev) | |
218 | { | |
219 | LIST_HEAD(reclaimed); | |
db830c46 | 220 | struct drbd_peer_request *peer_req, *t; |
b411b363 | 221 | |
87eeee41 | 222 | spin_lock_irq(&mdev->tconn->req_lock); |
a990be46 | 223 | reclaim_finished_net_peer_reqs(mdev, &reclaimed); |
87eeee41 | 224 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 225 | |
db830c46 | 226 | list_for_each_entry_safe(peer_req, t, &reclaimed, w.list) |
3967deb1 | 227 | drbd_free_net_peer_req(mdev, peer_req); |
b411b363 PR |
228 | } |
229 | ||
230 | /** | |
45bb912b | 231 | * drbd_pp_alloc() - Returns @number pages, retries forever (or until signalled) |
b411b363 | 232 | * @mdev: DRBD device. |
45bb912b LE |
233 | * @number: number of pages requested |
234 | * @retry: whether to retry, if not enough pages are available right now | |
235 | * | |
236 | * Tries to allocate number pages, first from our own page pool, then from | |
237 | * the kernel, unless this allocation would exceed the max_buffers setting. | |
238 | * Possibly retry until DRBD frees sufficient pages somewhere else. | |
b411b363 | 239 | * |
45bb912b | 240 | * Returns a page chain linked via page->private. |
b411b363 | 241 | */ |
45bb912b | 242 | static struct page *drbd_pp_alloc(struct drbd_conf *mdev, unsigned number, bool retry) |
b411b363 PR |
243 | { |
244 | struct page *page = NULL; | |
245 | DEFINE_WAIT(wait); | |
246 | ||
45bb912b LE |
247 | /* Yes, we may run up to @number over max_buffers. If we |
248 | * follow it strictly, the admin will get it wrong anyways. */ | |
89e58e75 | 249 | if (atomic_read(&mdev->pp_in_use) < mdev->tconn->net_conf->max_buffers) |
45bb912b | 250 | page = drbd_pp_first_pages_or_try_alloc(mdev, number); |
b411b363 | 251 | |
45bb912b | 252 | while (page == NULL) { |
b411b363 PR |
253 | prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE); |
254 | ||
255 | drbd_kick_lo_and_reclaim_net(mdev); | |
256 | ||
89e58e75 | 257 | if (atomic_read(&mdev->pp_in_use) < mdev->tconn->net_conf->max_buffers) { |
45bb912b | 258 | page = drbd_pp_first_pages_or_try_alloc(mdev, number); |
b411b363 PR |
259 | if (page) |
260 | break; | |
261 | } | |
262 | ||
263 | if (!retry) | |
264 | break; | |
265 | ||
266 | if (signal_pending(current)) { | |
267 | dev_warn(DEV, "drbd_pp_alloc interrupted!\n"); | |
268 | break; | |
269 | } | |
270 | ||
271 | schedule(); | |
272 | } | |
273 | finish_wait(&drbd_pp_wait, &wait); | |
274 | ||
45bb912b LE |
275 | if (page) |
276 | atomic_add(number, &mdev->pp_in_use); | |
b411b363 PR |
277 | return page; |
278 | } | |
279 | ||
280 | /* Must not be used from irq, as that may deadlock: see drbd_pp_alloc. | |
87eeee41 | 281 | * Is also used from inside an other spin_lock_irq(&mdev->tconn->req_lock); |
45bb912b LE |
282 | * Either links the page chain back to the global pool, |
283 | * or returns all pages to the system. */ | |
435f0740 | 284 | static void drbd_pp_free(struct drbd_conf *mdev, struct page *page, int is_net) |
b411b363 | 285 | { |
435f0740 | 286 | atomic_t *a = is_net ? &mdev->pp_in_use_by_net : &mdev->pp_in_use; |
b411b363 | 287 | int i; |
435f0740 | 288 | |
81a5d60e | 289 | if (drbd_pp_vacant > (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count) |
45bb912b LE |
290 | i = page_chain_free(page); |
291 | else { | |
292 | struct page *tmp; | |
293 | tmp = page_chain_tail(page, &i); | |
294 | spin_lock(&drbd_pp_lock); | |
295 | page_chain_add(&drbd_pp_pool, page, tmp); | |
296 | drbd_pp_vacant += i; | |
297 | spin_unlock(&drbd_pp_lock); | |
b411b363 | 298 | } |
435f0740 | 299 | i = atomic_sub_return(i, a); |
45bb912b | 300 | if (i < 0) |
435f0740 LE |
301 | dev_warn(DEV, "ASSERTION FAILED: %s: %d < 0\n", |
302 | is_net ? "pp_in_use_by_net" : "pp_in_use", i); | |
b411b363 PR |
303 | wake_up(&drbd_pp_wait); |
304 | } | |
305 | ||
306 | /* | |
307 | You need to hold the req_lock: | |
308 | _drbd_wait_ee_list_empty() | |
309 | ||
310 | You must not have the req_lock: | |
3967deb1 | 311 | drbd_free_peer_req() |
0db55363 | 312 | drbd_alloc_peer_req() |
7721f567 | 313 | drbd_free_peer_reqs() |
b411b363 | 314 | drbd_ee_fix_bhs() |
a990be46 | 315 | drbd_finish_peer_reqs() |
b411b363 PR |
316 | drbd_clear_done_ee() |
317 | drbd_wait_ee_list_empty() | |
318 | */ | |
319 | ||
f6ffca9f | 320 | struct drbd_peer_request * |
0db55363 AG |
321 | drbd_alloc_peer_req(struct drbd_conf *mdev, u64 id, sector_t sector, |
322 | unsigned int data_size, gfp_t gfp_mask) __must_hold(local) | |
b411b363 | 323 | { |
db830c46 | 324 | struct drbd_peer_request *peer_req; |
b411b363 | 325 | struct page *page; |
45bb912b | 326 | unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT; |
b411b363 | 327 | |
0cf9d27e | 328 | if (drbd_insert_fault(mdev, DRBD_FAULT_AL_EE)) |
b411b363 PR |
329 | return NULL; |
330 | ||
db830c46 AG |
331 | peer_req = mempool_alloc(drbd_ee_mempool, gfp_mask & ~__GFP_HIGHMEM); |
332 | if (!peer_req) { | |
b411b363 | 333 | if (!(gfp_mask & __GFP_NOWARN)) |
0db55363 | 334 | dev_err(DEV, "%s: allocation failed\n", __func__); |
b411b363 PR |
335 | return NULL; |
336 | } | |
337 | ||
45bb912b LE |
338 | page = drbd_pp_alloc(mdev, nr_pages, (gfp_mask & __GFP_WAIT)); |
339 | if (!page) | |
340 | goto fail; | |
b411b363 | 341 | |
db830c46 AG |
342 | drbd_clear_interval(&peer_req->i); |
343 | peer_req->i.size = data_size; | |
344 | peer_req->i.sector = sector; | |
345 | peer_req->i.local = false; | |
346 | peer_req->i.waiting = false; | |
347 | ||
348 | peer_req->epoch = NULL; | |
a21e9298 | 349 | peer_req->w.mdev = mdev; |
db830c46 AG |
350 | peer_req->pages = page; |
351 | atomic_set(&peer_req->pending_bios, 0); | |
352 | peer_req->flags = 0; | |
9a8e7753 AG |
353 | /* |
354 | * The block_id is opaque to the receiver. It is not endianness | |
355 | * converted, and sent back to the sender unchanged. | |
356 | */ | |
db830c46 | 357 | peer_req->block_id = id; |
b411b363 | 358 | |
db830c46 | 359 | return peer_req; |
b411b363 | 360 | |
45bb912b | 361 | fail: |
db830c46 | 362 | mempool_free(peer_req, drbd_ee_mempool); |
b411b363 PR |
363 | return NULL; |
364 | } | |
365 | ||
3967deb1 | 366 | void __drbd_free_peer_req(struct drbd_conf *mdev, struct drbd_peer_request *peer_req, |
f6ffca9f | 367 | int is_net) |
b411b363 | 368 | { |
db830c46 AG |
369 | if (peer_req->flags & EE_HAS_DIGEST) |
370 | kfree(peer_req->digest); | |
371 | drbd_pp_free(mdev, peer_req->pages, is_net); | |
372 | D_ASSERT(atomic_read(&peer_req->pending_bios) == 0); | |
373 | D_ASSERT(drbd_interval_empty(&peer_req->i)); | |
374 | mempool_free(peer_req, drbd_ee_mempool); | |
b411b363 PR |
375 | } |
376 | ||
7721f567 | 377 | int drbd_free_peer_reqs(struct drbd_conf *mdev, struct list_head *list) |
b411b363 PR |
378 | { |
379 | LIST_HEAD(work_list); | |
db830c46 | 380 | struct drbd_peer_request *peer_req, *t; |
b411b363 | 381 | int count = 0; |
435f0740 | 382 | int is_net = list == &mdev->net_ee; |
b411b363 | 383 | |
87eeee41 | 384 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 385 | list_splice_init(list, &work_list); |
87eeee41 | 386 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 387 | |
db830c46 | 388 | list_for_each_entry_safe(peer_req, t, &work_list, w.list) { |
3967deb1 | 389 | __drbd_free_peer_req(mdev, peer_req, is_net); |
b411b363 PR |
390 | count++; |
391 | } | |
392 | return count; | |
393 | } | |
394 | ||
a990be46 AG |
395 | /* |
396 | * See also comments in _req_mod(,BARRIER_ACKED) and receive_Barrier. | |
b411b363 | 397 | */ |
a990be46 | 398 | static int drbd_finish_peer_reqs(struct drbd_conf *mdev) |
b411b363 PR |
399 | { |
400 | LIST_HEAD(work_list); | |
401 | LIST_HEAD(reclaimed); | |
db830c46 | 402 | struct drbd_peer_request *peer_req, *t; |
e2b3032b | 403 | int err = 0; |
b411b363 | 404 | |
87eeee41 | 405 | spin_lock_irq(&mdev->tconn->req_lock); |
a990be46 | 406 | reclaim_finished_net_peer_reqs(mdev, &reclaimed); |
b411b363 | 407 | list_splice_init(&mdev->done_ee, &work_list); |
87eeee41 | 408 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 409 | |
db830c46 | 410 | list_for_each_entry_safe(peer_req, t, &reclaimed, w.list) |
3967deb1 | 411 | drbd_free_net_peer_req(mdev, peer_req); |
b411b363 PR |
412 | |
413 | /* possible callbacks here: | |
7be8da07 | 414 | * e_end_block, and e_end_resync_block, e_send_discard_write. |
b411b363 PR |
415 | * all ignore the last argument. |
416 | */ | |
db830c46 | 417 | list_for_each_entry_safe(peer_req, t, &work_list, w.list) { |
e2b3032b AG |
418 | int err2; |
419 | ||
b411b363 | 420 | /* list_del not necessary, next/prev members not touched */ |
e2b3032b AG |
421 | err2 = peer_req->w.cb(&peer_req->w, !!err); |
422 | if (!err) | |
423 | err = err2; | |
3967deb1 | 424 | drbd_free_peer_req(mdev, peer_req); |
b411b363 PR |
425 | } |
426 | wake_up(&mdev->ee_wait); | |
427 | ||
e2b3032b | 428 | return err; |
b411b363 PR |
429 | } |
430 | ||
431 | void _drbd_wait_ee_list_empty(struct drbd_conf *mdev, struct list_head *head) | |
432 | { | |
433 | DEFINE_WAIT(wait); | |
434 | ||
435 | /* avoids spin_lock/unlock | |
436 | * and calling prepare_to_wait in the fast path */ | |
437 | while (!list_empty(head)) { | |
438 | prepare_to_wait(&mdev->ee_wait, &wait, TASK_UNINTERRUPTIBLE); | |
87eeee41 | 439 | spin_unlock_irq(&mdev->tconn->req_lock); |
7eaceacc | 440 | io_schedule(); |
b411b363 | 441 | finish_wait(&mdev->ee_wait, &wait); |
87eeee41 | 442 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
443 | } |
444 | } | |
445 | ||
446 | void drbd_wait_ee_list_empty(struct drbd_conf *mdev, struct list_head *head) | |
447 | { | |
87eeee41 | 448 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 449 | _drbd_wait_ee_list_empty(mdev, head); |
87eeee41 | 450 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
451 | } |
452 | ||
453 | /* see also kernel_accept; which is only present since 2.6.18. | |
454 | * also we want to log which part of it failed, exactly */ | |
7653620d | 455 | static int drbd_accept(const char **what, struct socket *sock, struct socket **newsock) |
b411b363 PR |
456 | { |
457 | struct sock *sk = sock->sk; | |
458 | int err = 0; | |
459 | ||
460 | *what = "listen"; | |
461 | err = sock->ops->listen(sock, 5); | |
462 | if (err < 0) | |
463 | goto out; | |
464 | ||
465 | *what = "sock_create_lite"; | |
466 | err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol, | |
467 | newsock); | |
468 | if (err < 0) | |
469 | goto out; | |
470 | ||
471 | *what = "accept"; | |
472 | err = sock->ops->accept(sock, *newsock, 0); | |
473 | if (err < 0) { | |
474 | sock_release(*newsock); | |
475 | *newsock = NULL; | |
476 | goto out; | |
477 | } | |
478 | (*newsock)->ops = sock->ops; | |
479 | ||
480 | out: | |
481 | return err; | |
482 | } | |
483 | ||
dbd9eea0 | 484 | static int drbd_recv_short(struct socket *sock, void *buf, size_t size, int flags) |
b411b363 PR |
485 | { |
486 | mm_segment_t oldfs; | |
487 | struct kvec iov = { | |
488 | .iov_base = buf, | |
489 | .iov_len = size, | |
490 | }; | |
491 | struct msghdr msg = { | |
492 | .msg_iovlen = 1, | |
493 | .msg_iov = (struct iovec *)&iov, | |
494 | .msg_flags = (flags ? flags : MSG_WAITALL | MSG_NOSIGNAL) | |
495 | }; | |
496 | int rv; | |
497 | ||
498 | oldfs = get_fs(); | |
499 | set_fs(KERNEL_DS); | |
500 | rv = sock_recvmsg(sock, &msg, size, msg.msg_flags); | |
501 | set_fs(oldfs); | |
502 | ||
503 | return rv; | |
504 | } | |
505 | ||
de0ff338 | 506 | static int drbd_recv(struct drbd_tconn *tconn, void *buf, size_t size) |
b411b363 PR |
507 | { |
508 | mm_segment_t oldfs; | |
509 | struct kvec iov = { | |
510 | .iov_base = buf, | |
511 | .iov_len = size, | |
512 | }; | |
513 | struct msghdr msg = { | |
514 | .msg_iovlen = 1, | |
515 | .msg_iov = (struct iovec *)&iov, | |
516 | .msg_flags = MSG_WAITALL | MSG_NOSIGNAL | |
517 | }; | |
518 | int rv; | |
519 | ||
520 | oldfs = get_fs(); | |
521 | set_fs(KERNEL_DS); | |
522 | ||
523 | for (;;) { | |
de0ff338 | 524 | rv = sock_recvmsg(tconn->data.socket, &msg, size, msg.msg_flags); |
b411b363 PR |
525 | if (rv == size) |
526 | break; | |
527 | ||
528 | /* Note: | |
529 | * ECONNRESET other side closed the connection | |
530 | * ERESTARTSYS (on sock) we got a signal | |
531 | */ | |
532 | ||
533 | if (rv < 0) { | |
534 | if (rv == -ECONNRESET) | |
de0ff338 | 535 | conn_info(tconn, "sock was reset by peer\n"); |
b411b363 | 536 | else if (rv != -ERESTARTSYS) |
de0ff338 | 537 | conn_err(tconn, "sock_recvmsg returned %d\n", rv); |
b411b363 PR |
538 | break; |
539 | } else if (rv == 0) { | |
de0ff338 | 540 | conn_info(tconn, "sock was shut down by peer\n"); |
b411b363 PR |
541 | break; |
542 | } else { | |
543 | /* signal came in, or peer/link went down, | |
544 | * after we read a partial message | |
545 | */ | |
546 | /* D_ASSERT(signal_pending(current)); */ | |
547 | break; | |
548 | } | |
549 | }; | |
550 | ||
551 | set_fs(oldfs); | |
552 | ||
553 | if (rv != size) | |
bbeb641c | 554 | conn_request_state(tconn, NS(conn, C_BROKEN_PIPE), CS_HARD); |
b411b363 PR |
555 | |
556 | return rv; | |
557 | } | |
558 | ||
c6967746 AG |
559 | static int drbd_recv_all(struct drbd_tconn *tconn, void *buf, size_t size) |
560 | { | |
561 | int err; | |
562 | ||
563 | err = drbd_recv(tconn, buf, size); | |
564 | if (err != size) { | |
565 | if (err >= 0) | |
566 | err = -EIO; | |
567 | } else | |
568 | err = 0; | |
569 | return err; | |
570 | } | |
571 | ||
a5c31904 AG |
572 | static int drbd_recv_all_warn(struct drbd_tconn *tconn, void *buf, size_t size) |
573 | { | |
574 | int err; | |
575 | ||
576 | err = drbd_recv_all(tconn, buf, size); | |
577 | if (err && !signal_pending(current)) | |
578 | conn_warn(tconn, "short read (expected size %d)\n", (int)size); | |
579 | return err; | |
580 | } | |
581 | ||
5dbf1673 LE |
582 | /* quoting tcp(7): |
583 | * On individual connections, the socket buffer size must be set prior to the | |
584 | * listen(2) or connect(2) calls in order to have it take effect. | |
585 | * This is our wrapper to do so. | |
586 | */ | |
587 | static void drbd_setbufsize(struct socket *sock, unsigned int snd, | |
588 | unsigned int rcv) | |
589 | { | |
590 | /* open coded SO_SNDBUF, SO_RCVBUF */ | |
591 | if (snd) { | |
592 | sock->sk->sk_sndbuf = snd; | |
593 | sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK; | |
594 | } | |
595 | if (rcv) { | |
596 | sock->sk->sk_rcvbuf = rcv; | |
597 | sock->sk->sk_userlocks |= SOCK_RCVBUF_LOCK; | |
598 | } | |
599 | } | |
600 | ||
eac3e990 | 601 | static struct socket *drbd_try_connect(struct drbd_tconn *tconn) |
b411b363 PR |
602 | { |
603 | const char *what; | |
604 | struct socket *sock; | |
605 | struct sockaddr_in6 src_in6; | |
606 | int err; | |
607 | int disconnect_on_error = 1; | |
608 | ||
eac3e990 | 609 | if (!get_net_conf(tconn)) |
b411b363 PR |
610 | return NULL; |
611 | ||
612 | what = "sock_create_kern"; | |
eac3e990 | 613 | err = sock_create_kern(((struct sockaddr *)tconn->net_conf->my_addr)->sa_family, |
b411b363 PR |
614 | SOCK_STREAM, IPPROTO_TCP, &sock); |
615 | if (err < 0) { | |
616 | sock = NULL; | |
617 | goto out; | |
618 | } | |
619 | ||
620 | sock->sk->sk_rcvtimeo = | |
eac3e990 PR |
621 | sock->sk->sk_sndtimeo = tconn->net_conf->try_connect_int*HZ; |
622 | drbd_setbufsize(sock, tconn->net_conf->sndbuf_size, | |
623 | tconn->net_conf->rcvbuf_size); | |
b411b363 PR |
624 | |
625 | /* explicitly bind to the configured IP as source IP | |
626 | * for the outgoing connections. | |
627 | * This is needed for multihomed hosts and to be | |
628 | * able to use lo: interfaces for drbd. | |
629 | * Make sure to use 0 as port number, so linux selects | |
630 | * a free one dynamically. | |
631 | */ | |
eac3e990 PR |
632 | memcpy(&src_in6, tconn->net_conf->my_addr, |
633 | min_t(int, tconn->net_conf->my_addr_len, sizeof(src_in6))); | |
634 | if (((struct sockaddr *)tconn->net_conf->my_addr)->sa_family == AF_INET6) | |
b411b363 PR |
635 | src_in6.sin6_port = 0; |
636 | else | |
637 | ((struct sockaddr_in *)&src_in6)->sin_port = 0; /* AF_INET & AF_SCI */ | |
638 | ||
639 | what = "bind before connect"; | |
640 | err = sock->ops->bind(sock, | |
641 | (struct sockaddr *) &src_in6, | |
eac3e990 | 642 | tconn->net_conf->my_addr_len); |
b411b363 PR |
643 | if (err < 0) |
644 | goto out; | |
645 | ||
646 | /* connect may fail, peer not yet available. | |
647 | * stay C_WF_CONNECTION, don't go Disconnecting! */ | |
648 | disconnect_on_error = 0; | |
649 | what = "connect"; | |
650 | err = sock->ops->connect(sock, | |
eac3e990 PR |
651 | (struct sockaddr *)tconn->net_conf->peer_addr, |
652 | tconn->net_conf->peer_addr_len, 0); | |
b411b363 PR |
653 | |
654 | out: | |
655 | if (err < 0) { | |
656 | if (sock) { | |
657 | sock_release(sock); | |
658 | sock = NULL; | |
659 | } | |
660 | switch (-err) { | |
661 | /* timeout, busy, signal pending */ | |
662 | case ETIMEDOUT: case EAGAIN: case EINPROGRESS: | |
663 | case EINTR: case ERESTARTSYS: | |
664 | /* peer not (yet) available, network problem */ | |
665 | case ECONNREFUSED: case ENETUNREACH: | |
666 | case EHOSTDOWN: case EHOSTUNREACH: | |
667 | disconnect_on_error = 0; | |
668 | break; | |
669 | default: | |
eac3e990 | 670 | conn_err(tconn, "%s failed, err = %d\n", what, err); |
b411b363 PR |
671 | } |
672 | if (disconnect_on_error) | |
bbeb641c | 673 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 | 674 | } |
eac3e990 | 675 | put_net_conf(tconn); |
b411b363 PR |
676 | return sock; |
677 | } | |
678 | ||
7653620d | 679 | static struct socket *drbd_wait_for_connect(struct drbd_tconn *tconn) |
b411b363 PR |
680 | { |
681 | int timeo, err; | |
682 | struct socket *s_estab = NULL, *s_listen; | |
683 | const char *what; | |
684 | ||
7653620d | 685 | if (!get_net_conf(tconn)) |
b411b363 PR |
686 | return NULL; |
687 | ||
688 | what = "sock_create_kern"; | |
7653620d | 689 | err = sock_create_kern(((struct sockaddr *)tconn->net_conf->my_addr)->sa_family, |
b411b363 PR |
690 | SOCK_STREAM, IPPROTO_TCP, &s_listen); |
691 | if (err) { | |
692 | s_listen = NULL; | |
693 | goto out; | |
694 | } | |
695 | ||
7653620d | 696 | timeo = tconn->net_conf->try_connect_int * HZ; |
b411b363 PR |
697 | timeo += (random32() & 1) ? timeo / 7 : -timeo / 7; /* 28.5% random jitter */ |
698 | ||
699 | s_listen->sk->sk_reuse = 1; /* SO_REUSEADDR */ | |
700 | s_listen->sk->sk_rcvtimeo = timeo; | |
701 | s_listen->sk->sk_sndtimeo = timeo; | |
7653620d PR |
702 | drbd_setbufsize(s_listen, tconn->net_conf->sndbuf_size, |
703 | tconn->net_conf->rcvbuf_size); | |
b411b363 PR |
704 | |
705 | what = "bind before listen"; | |
706 | err = s_listen->ops->bind(s_listen, | |
7653620d PR |
707 | (struct sockaddr *) tconn->net_conf->my_addr, |
708 | tconn->net_conf->my_addr_len); | |
b411b363 PR |
709 | if (err < 0) |
710 | goto out; | |
711 | ||
7653620d | 712 | err = drbd_accept(&what, s_listen, &s_estab); |
b411b363 PR |
713 | |
714 | out: | |
715 | if (s_listen) | |
716 | sock_release(s_listen); | |
717 | if (err < 0) { | |
718 | if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) { | |
7653620d | 719 | conn_err(tconn, "%s failed, err = %d\n", what, err); |
bbeb641c | 720 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 PR |
721 | } |
722 | } | |
7653620d | 723 | put_net_conf(tconn); |
b411b363 PR |
724 | |
725 | return s_estab; | |
726 | } | |
727 | ||
e658983a | 728 | static int decode_header(struct drbd_tconn *, void *, struct packet_info *); |
b411b363 | 729 | |
9f5bdc33 AG |
730 | static int send_first_packet(struct drbd_tconn *tconn, struct drbd_socket *sock, |
731 | enum drbd_packet cmd) | |
732 | { | |
733 | if (!conn_prepare_command(tconn, sock)) | |
734 | return -EIO; | |
e658983a | 735 | return conn_send_command(tconn, sock, cmd, 0, NULL, 0); |
b411b363 PR |
736 | } |
737 | ||
9f5bdc33 | 738 | static int receive_first_packet(struct drbd_tconn *tconn, struct socket *sock) |
b411b363 | 739 | { |
9f5bdc33 AG |
740 | unsigned int header_size = drbd_header_size(tconn); |
741 | struct packet_info pi; | |
742 | int err; | |
b411b363 | 743 | |
9f5bdc33 AG |
744 | err = drbd_recv_short(sock, tconn->data.rbuf, header_size, 0); |
745 | if (err != header_size) { | |
746 | if (err >= 0) | |
747 | err = -EIO; | |
748 | return err; | |
749 | } | |
750 | err = decode_header(tconn, tconn->data.rbuf, &pi); | |
751 | if (err) | |
752 | return err; | |
753 | return pi.cmd; | |
b411b363 PR |
754 | } |
755 | ||
756 | /** | |
757 | * drbd_socket_okay() - Free the socket if its connection is not okay | |
b411b363 PR |
758 | * @sock: pointer to the pointer to the socket. |
759 | */ | |
dbd9eea0 | 760 | static int drbd_socket_okay(struct socket **sock) |
b411b363 PR |
761 | { |
762 | int rr; | |
763 | char tb[4]; | |
764 | ||
765 | if (!*sock) | |
81e84650 | 766 | return false; |
b411b363 | 767 | |
dbd9eea0 | 768 | rr = drbd_recv_short(*sock, tb, 4, MSG_DONTWAIT | MSG_PEEK); |
b411b363 PR |
769 | |
770 | if (rr > 0 || rr == -EAGAIN) { | |
81e84650 | 771 | return true; |
b411b363 PR |
772 | } else { |
773 | sock_release(*sock); | |
774 | *sock = NULL; | |
81e84650 | 775 | return false; |
b411b363 PR |
776 | } |
777 | } | |
2325eb66 PR |
778 | /* Gets called if a connection is established, or if a new minor gets created |
779 | in a connection */ | |
780 | int drbd_connected(int vnr, void *p, void *data) | |
907599e0 PR |
781 | { |
782 | struct drbd_conf *mdev = (struct drbd_conf *)p; | |
0829f5ed | 783 | int err; |
907599e0 PR |
784 | |
785 | atomic_set(&mdev->packet_seq, 0); | |
786 | mdev->peer_seq = 0; | |
787 | ||
8410da8f PR |
788 | mdev->state_mutex = mdev->tconn->agreed_pro_version < 100 ? |
789 | &mdev->tconn->cstate_mutex : | |
790 | &mdev->own_state_mutex; | |
791 | ||
0829f5ed AG |
792 | err = drbd_send_sync_param(mdev); |
793 | if (!err) | |
794 | err = drbd_send_sizes(mdev, 0, 0); | |
795 | if (!err) | |
796 | err = drbd_send_uuids(mdev); | |
797 | if (!err) | |
798 | err = drbd_send_state(mdev); | |
907599e0 PR |
799 | clear_bit(USE_DEGR_WFC_T, &mdev->flags); |
800 | clear_bit(RESIZE_PENDING, &mdev->flags); | |
8b924f1d | 801 | mod_timer(&mdev->request_timer, jiffies + HZ); /* just start it here. */ |
0829f5ed | 802 | return err; |
907599e0 PR |
803 | } |
804 | ||
b411b363 PR |
805 | /* |
806 | * return values: | |
807 | * 1 yes, we have a valid connection | |
808 | * 0 oops, did not work out, please try again | |
809 | * -1 peer talks different language, | |
810 | * no point in trying again, please go standalone. | |
811 | * -2 We do not have a network config... | |
812 | */ | |
907599e0 | 813 | static int drbd_connect(struct drbd_tconn *tconn) |
b411b363 | 814 | { |
2bf89621 | 815 | struct socket *sock, *msock; |
b411b363 PR |
816 | int try, h, ok; |
817 | ||
bbeb641c | 818 | if (conn_request_state(tconn, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS) |
b411b363 PR |
819 | return -2; |
820 | ||
907599e0 | 821 | clear_bit(DISCARD_CONCURRENT, &tconn->flags); |
0916e0e3 AG |
822 | |
823 | /* Assume that the peer only understands protocol 80 until we know better. */ | |
824 | tconn->agreed_pro_version = 80; | |
b411b363 | 825 | |
b411b363 | 826 | do { |
2bf89621 AG |
827 | struct socket *s; |
828 | ||
b411b363 PR |
829 | for (try = 0;;) { |
830 | /* 3 tries, this should take less than a second! */ | |
907599e0 | 831 | s = drbd_try_connect(tconn); |
b411b363 PR |
832 | if (s || ++try >= 3) |
833 | break; | |
834 | /* give the other side time to call bind() & listen() */ | |
20ee6390 | 835 | schedule_timeout_interruptible(HZ / 10); |
b411b363 PR |
836 | } |
837 | ||
838 | if (s) { | |
2bf89621 AG |
839 | if (!tconn->data.socket) { |
840 | tconn->data.socket = s; | |
9f5bdc33 | 841 | send_first_packet(tconn, &tconn->data, P_INITIAL_DATA); |
2bf89621 AG |
842 | } else if (!tconn->meta.socket) { |
843 | tconn->meta.socket = s; | |
9f5bdc33 | 844 | send_first_packet(tconn, &tconn->meta, P_INITIAL_META); |
b411b363 | 845 | } else { |
907599e0 | 846 | conn_err(tconn, "Logic error in drbd_connect()\n"); |
b411b363 PR |
847 | goto out_release_sockets; |
848 | } | |
849 | } | |
850 | ||
2bf89621 | 851 | if (tconn->data.socket && tconn->meta.socket) { |
907599e0 | 852 | schedule_timeout_interruptible(tconn->net_conf->ping_timeo*HZ/10); |
2bf89621 AG |
853 | ok = drbd_socket_okay(&tconn->data.socket); |
854 | ok = drbd_socket_okay(&tconn->meta.socket) && ok; | |
b411b363 PR |
855 | if (ok) |
856 | break; | |
857 | } | |
858 | ||
859 | retry: | |
907599e0 | 860 | s = drbd_wait_for_connect(tconn); |
b411b363 | 861 | if (s) { |
9f5bdc33 | 862 | try = receive_first_packet(tconn, s); |
2bf89621 AG |
863 | drbd_socket_okay(&tconn->data.socket); |
864 | drbd_socket_okay(&tconn->meta.socket); | |
b411b363 | 865 | switch (try) { |
e5d6f33a | 866 | case P_INITIAL_DATA: |
2bf89621 | 867 | if (tconn->data.socket) { |
907599e0 | 868 | conn_warn(tconn, "initial packet S crossed\n"); |
2bf89621 | 869 | sock_release(tconn->data.socket); |
b411b363 | 870 | } |
2bf89621 | 871 | tconn->data.socket = s; |
b411b363 | 872 | break; |
e5d6f33a | 873 | case P_INITIAL_META: |
2bf89621 | 874 | if (tconn->meta.socket) { |
907599e0 | 875 | conn_warn(tconn, "initial packet M crossed\n"); |
2bf89621 | 876 | sock_release(tconn->meta.socket); |
b411b363 | 877 | } |
2bf89621 | 878 | tconn->meta.socket = s; |
907599e0 | 879 | set_bit(DISCARD_CONCURRENT, &tconn->flags); |
b411b363 PR |
880 | break; |
881 | default: | |
907599e0 | 882 | conn_warn(tconn, "Error receiving initial packet\n"); |
b411b363 PR |
883 | sock_release(s); |
884 | if (random32() & 1) | |
885 | goto retry; | |
886 | } | |
887 | } | |
888 | ||
bbeb641c | 889 | if (tconn->cstate <= C_DISCONNECTING) |
b411b363 PR |
890 | goto out_release_sockets; |
891 | if (signal_pending(current)) { | |
892 | flush_signals(current); | |
893 | smp_rmb(); | |
907599e0 | 894 | if (get_t_state(&tconn->receiver) == EXITING) |
b411b363 PR |
895 | goto out_release_sockets; |
896 | } | |
897 | ||
2bf89621 AG |
898 | if (tconn->data.socket && &tconn->meta.socket) { |
899 | ok = drbd_socket_okay(&tconn->data.socket); | |
900 | ok = drbd_socket_okay(&tconn->meta.socket) && ok; | |
b411b363 PR |
901 | if (ok) |
902 | break; | |
903 | } | |
904 | } while (1); | |
905 | ||
2bf89621 AG |
906 | sock = tconn->data.socket; |
907 | msock = tconn->meta.socket; | |
908 | ||
b411b363 PR |
909 | msock->sk->sk_reuse = 1; /* SO_REUSEADDR */ |
910 | sock->sk->sk_reuse = 1; /* SO_REUSEADDR */ | |
911 | ||
912 | sock->sk->sk_allocation = GFP_NOIO; | |
913 | msock->sk->sk_allocation = GFP_NOIO; | |
914 | ||
915 | sock->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK; | |
916 | msock->sk->sk_priority = TC_PRIO_INTERACTIVE; | |
917 | ||
b411b363 | 918 | /* NOT YET ... |
907599e0 | 919 | * sock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10; |
b411b363 | 920 | * sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; |
6038178e | 921 | * first set it to the P_CONNECTION_FEATURES timeout, |
b411b363 PR |
922 | * which we set to 4x the configured ping_timeout. */ |
923 | sock->sk->sk_sndtimeo = | |
907599e0 | 924 | sock->sk->sk_rcvtimeo = tconn->net_conf->ping_timeo*4*HZ/10; |
b411b363 | 925 | |
907599e0 PR |
926 | msock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10; |
927 | msock->sk->sk_rcvtimeo = tconn->net_conf->ping_int*HZ; | |
b411b363 PR |
928 | |
929 | /* we don't want delays. | |
25985edc | 930 | * we use TCP_CORK where appropriate, though */ |
b411b363 PR |
931 | drbd_tcp_nodelay(sock); |
932 | drbd_tcp_nodelay(msock); | |
933 | ||
907599e0 | 934 | tconn->last_received = jiffies; |
b411b363 | 935 | |
6038178e | 936 | h = drbd_do_features(tconn); |
b411b363 PR |
937 | if (h <= 0) |
938 | return h; | |
939 | ||
907599e0 | 940 | if (tconn->cram_hmac_tfm) { |
b411b363 | 941 | /* drbd_request_state(mdev, NS(conn, WFAuth)); */ |
907599e0 | 942 | switch (drbd_do_auth(tconn)) { |
b10d96cb | 943 | case -1: |
907599e0 | 944 | conn_err(tconn, "Authentication of peer failed\n"); |
b411b363 | 945 | return -1; |
b10d96cb | 946 | case 0: |
907599e0 | 947 | conn_err(tconn, "Authentication of peer failed, trying again.\n"); |
b10d96cb | 948 | return 0; |
b411b363 PR |
949 | } |
950 | } | |
951 | ||
bbeb641c | 952 | if (conn_request_state(tconn, NS(conn, C_WF_REPORT_PARAMS), CS_VERBOSE) < SS_SUCCESS) |
b411b363 PR |
953 | return 0; |
954 | ||
907599e0 | 955 | sock->sk->sk_sndtimeo = tconn->net_conf->timeout*HZ/10; |
b411b363 PR |
956 | sock->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; |
957 | ||
907599e0 | 958 | drbd_thread_start(&tconn->asender); |
b411b363 | 959 | |
387eb308 | 960 | if (drbd_send_protocol(tconn) == -EOPNOTSUPP) |
7e2455c1 | 961 | return -1; |
b411b363 | 962 | |
907599e0 | 963 | return !idr_for_each(&tconn->volumes, drbd_connected, tconn); |
b411b363 PR |
964 | |
965 | out_release_sockets: | |
2bf89621 AG |
966 | if (tconn->data.socket) { |
967 | sock_release(tconn->data.socket); | |
968 | tconn->data.socket = NULL; | |
969 | } | |
970 | if (tconn->meta.socket) { | |
971 | sock_release(tconn->meta.socket); | |
972 | tconn->meta.socket = NULL; | |
973 | } | |
b411b363 PR |
974 | return -1; |
975 | } | |
976 | ||
e658983a | 977 | static int decode_header(struct drbd_tconn *tconn, void *header, struct packet_info *pi) |
b411b363 | 978 | { |
e658983a AG |
979 | unsigned int header_size = drbd_header_size(tconn); |
980 | ||
0c8e36d9 AG |
981 | if (header_size == sizeof(struct p_header100) && |
982 | *(__be32 *)header == cpu_to_be32(DRBD_MAGIC_100)) { | |
983 | struct p_header100 *h = header; | |
984 | if (h->pad != 0) { | |
985 | conn_err(tconn, "Header padding is not zero\n"); | |
986 | return -EINVAL; | |
987 | } | |
988 | pi->vnr = be16_to_cpu(h->volume); | |
989 | pi->cmd = be16_to_cpu(h->command); | |
990 | pi->size = be32_to_cpu(h->length); | |
991 | } else if (header_size == sizeof(struct p_header95) && | |
992 | *(__be16 *)header == cpu_to_be16(DRBD_MAGIC_BIG)) { | |
e658983a | 993 | struct p_header95 *h = header; |
e658983a | 994 | pi->cmd = be16_to_cpu(h->command); |
b55d84ba AG |
995 | pi->size = be32_to_cpu(h->length); |
996 | pi->vnr = 0; | |
e658983a AG |
997 | } else if (header_size == sizeof(struct p_header80) && |
998 | *(__be32 *)header == cpu_to_be32(DRBD_MAGIC)) { | |
999 | struct p_header80 *h = header; | |
1000 | pi->cmd = be16_to_cpu(h->command); | |
1001 | pi->size = be16_to_cpu(h->length); | |
77351055 | 1002 | pi->vnr = 0; |
02918be2 | 1003 | } else { |
e658983a AG |
1004 | conn_err(tconn, "Wrong magic value 0x%08x in protocol version %d\n", |
1005 | be32_to_cpu(*(__be32 *)header), | |
1006 | tconn->agreed_pro_version); | |
8172f3e9 | 1007 | return -EINVAL; |
b411b363 | 1008 | } |
e658983a | 1009 | pi->data = header + header_size; |
8172f3e9 | 1010 | return 0; |
257d0af6 PR |
1011 | } |
1012 | ||
9ba7aa00 | 1013 | static int drbd_recv_header(struct drbd_tconn *tconn, struct packet_info *pi) |
257d0af6 | 1014 | { |
e658983a | 1015 | void *buffer = tconn->data.rbuf; |
69bc7bc3 | 1016 | int err; |
257d0af6 | 1017 | |
e658983a | 1018 | err = drbd_recv_all_warn(tconn, buffer, drbd_header_size(tconn)); |
a5c31904 | 1019 | if (err) |
69bc7bc3 | 1020 | return err; |
257d0af6 | 1021 | |
e658983a | 1022 | err = decode_header(tconn, buffer, pi); |
9ba7aa00 | 1023 | tconn->last_received = jiffies; |
b411b363 | 1024 | |
69bc7bc3 | 1025 | return err; |
b411b363 PR |
1026 | } |
1027 | ||
2451fc3b | 1028 | static void drbd_flush(struct drbd_conf *mdev) |
b411b363 PR |
1029 | { |
1030 | int rv; | |
1031 | ||
1032 | if (mdev->write_ordering >= WO_bdev_flush && get_ldev(mdev)) { | |
fbd9b09a | 1033 | rv = blkdev_issue_flush(mdev->ldev->backing_bdev, GFP_KERNEL, |
dd3932ed | 1034 | NULL); |
b411b363 PR |
1035 | if (rv) { |
1036 | dev_err(DEV, "local disk flush failed with status %d\n", rv); | |
1037 | /* would rather check on EOPNOTSUPP, but that is not reliable. | |
1038 | * don't try again for ANY return value != 0 | |
1039 | * if (rv == -EOPNOTSUPP) */ | |
1040 | drbd_bump_write_ordering(mdev, WO_drain_io); | |
1041 | } | |
1042 | put_ldev(mdev); | |
1043 | } | |
b411b363 PR |
1044 | } |
1045 | ||
1046 | /** | |
1047 | * drbd_may_finish_epoch() - Applies an epoch_event to the epoch's state, eventually finishes it. | |
1048 | * @mdev: DRBD device. | |
1049 | * @epoch: Epoch object. | |
1050 | * @ev: Epoch event. | |
1051 | */ | |
1052 | static enum finish_epoch drbd_may_finish_epoch(struct drbd_conf *mdev, | |
1053 | struct drbd_epoch *epoch, | |
1054 | enum epoch_event ev) | |
1055 | { | |
2451fc3b | 1056 | int epoch_size; |
b411b363 | 1057 | struct drbd_epoch *next_epoch; |
b411b363 PR |
1058 | enum finish_epoch rv = FE_STILL_LIVE; |
1059 | ||
1060 | spin_lock(&mdev->epoch_lock); | |
1061 | do { | |
1062 | next_epoch = NULL; | |
b411b363 PR |
1063 | |
1064 | epoch_size = atomic_read(&epoch->epoch_size); | |
1065 | ||
1066 | switch (ev & ~EV_CLEANUP) { | |
1067 | case EV_PUT: | |
1068 | atomic_dec(&epoch->active); | |
1069 | break; | |
1070 | case EV_GOT_BARRIER_NR: | |
1071 | set_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags); | |
b411b363 PR |
1072 | break; |
1073 | case EV_BECAME_LAST: | |
1074 | /* nothing to do*/ | |
1075 | break; | |
1076 | } | |
1077 | ||
b411b363 PR |
1078 | if (epoch_size != 0 && |
1079 | atomic_read(&epoch->active) == 0 && | |
2451fc3b | 1080 | test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags)) { |
b411b363 PR |
1081 | if (!(ev & EV_CLEANUP)) { |
1082 | spin_unlock(&mdev->epoch_lock); | |
1083 | drbd_send_b_ack(mdev, epoch->barrier_nr, epoch_size); | |
1084 | spin_lock(&mdev->epoch_lock); | |
1085 | } | |
1086 | dec_unacked(mdev); | |
1087 | ||
1088 | if (mdev->current_epoch != epoch) { | |
1089 | next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list); | |
1090 | list_del(&epoch->list); | |
1091 | ev = EV_BECAME_LAST | (ev & EV_CLEANUP); | |
1092 | mdev->epochs--; | |
b411b363 PR |
1093 | kfree(epoch); |
1094 | ||
1095 | if (rv == FE_STILL_LIVE) | |
1096 | rv = FE_DESTROYED; | |
1097 | } else { | |
1098 | epoch->flags = 0; | |
1099 | atomic_set(&epoch->epoch_size, 0); | |
698f9315 | 1100 | /* atomic_set(&epoch->active, 0); is already zero */ |
b411b363 PR |
1101 | if (rv == FE_STILL_LIVE) |
1102 | rv = FE_RECYCLED; | |
2451fc3b | 1103 | wake_up(&mdev->ee_wait); |
b411b363 PR |
1104 | } |
1105 | } | |
1106 | ||
1107 | if (!next_epoch) | |
1108 | break; | |
1109 | ||
1110 | epoch = next_epoch; | |
1111 | } while (1); | |
1112 | ||
1113 | spin_unlock(&mdev->epoch_lock); | |
1114 | ||
b411b363 PR |
1115 | return rv; |
1116 | } | |
1117 | ||
1118 | /** | |
1119 | * drbd_bump_write_ordering() - Fall back to an other write ordering method | |
1120 | * @mdev: DRBD device. | |
1121 | * @wo: Write ordering method to try. | |
1122 | */ | |
1123 | void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo) __must_hold(local) | |
1124 | { | |
1125 | enum write_ordering_e pwo; | |
1126 | static char *write_ordering_str[] = { | |
1127 | [WO_none] = "none", | |
1128 | [WO_drain_io] = "drain", | |
1129 | [WO_bdev_flush] = "flush", | |
b411b363 PR |
1130 | }; |
1131 | ||
1132 | pwo = mdev->write_ordering; | |
1133 | wo = min(pwo, wo); | |
b411b363 PR |
1134 | if (wo == WO_bdev_flush && mdev->ldev->dc.no_disk_flush) |
1135 | wo = WO_drain_io; | |
1136 | if (wo == WO_drain_io && mdev->ldev->dc.no_disk_drain) | |
1137 | wo = WO_none; | |
1138 | mdev->write_ordering = wo; | |
2451fc3b | 1139 | if (pwo != mdev->write_ordering || wo == WO_bdev_flush) |
b411b363 PR |
1140 | dev_info(DEV, "Method to ensure write ordering: %s\n", write_ordering_str[mdev->write_ordering]); |
1141 | } | |
1142 | ||
45bb912b | 1143 | /** |
fbe29dec | 1144 | * drbd_submit_peer_request() |
45bb912b | 1145 | * @mdev: DRBD device. |
db830c46 | 1146 | * @peer_req: peer request |
45bb912b | 1147 | * @rw: flag field, see bio->bi_rw |
10f6d992 LE |
1148 | * |
1149 | * May spread the pages to multiple bios, | |
1150 | * depending on bio_add_page restrictions. | |
1151 | * | |
1152 | * Returns 0 if all bios have been submitted, | |
1153 | * -ENOMEM if we could not allocate enough bios, | |
1154 | * -ENOSPC (any better suggestion?) if we have not been able to bio_add_page a | |
1155 | * single page to an empty bio (which should never happen and likely indicates | |
1156 | * that the lower level IO stack is in some way broken). This has been observed | |
1157 | * on certain Xen deployments. | |
45bb912b LE |
1158 | */ |
1159 | /* TODO allocate from our own bio_set. */ | |
fbe29dec AG |
1160 | int drbd_submit_peer_request(struct drbd_conf *mdev, |
1161 | struct drbd_peer_request *peer_req, | |
1162 | const unsigned rw, const int fault_type) | |
45bb912b LE |
1163 | { |
1164 | struct bio *bios = NULL; | |
1165 | struct bio *bio; | |
db830c46 AG |
1166 | struct page *page = peer_req->pages; |
1167 | sector_t sector = peer_req->i.sector; | |
1168 | unsigned ds = peer_req->i.size; | |
45bb912b LE |
1169 | unsigned n_bios = 0; |
1170 | unsigned nr_pages = (ds + PAGE_SIZE -1) >> PAGE_SHIFT; | |
10f6d992 | 1171 | int err = -ENOMEM; |
45bb912b LE |
1172 | |
1173 | /* In most cases, we will only need one bio. But in case the lower | |
1174 | * level restrictions happen to be different at this offset on this | |
1175 | * side than those of the sending peer, we may need to submit the | |
da4a75d2 LE |
1176 | * request in more than one bio. |
1177 | * | |
1178 | * Plain bio_alloc is good enough here, this is no DRBD internally | |
1179 | * generated bio, but a bio allocated on behalf of the peer. | |
1180 | */ | |
45bb912b LE |
1181 | next_bio: |
1182 | bio = bio_alloc(GFP_NOIO, nr_pages); | |
1183 | if (!bio) { | |
1184 | dev_err(DEV, "submit_ee: Allocation of a bio failed\n"); | |
1185 | goto fail; | |
1186 | } | |
db830c46 | 1187 | /* > peer_req->i.sector, unless this is the first bio */ |
45bb912b LE |
1188 | bio->bi_sector = sector; |
1189 | bio->bi_bdev = mdev->ldev->backing_bdev; | |
45bb912b | 1190 | bio->bi_rw = rw; |
db830c46 | 1191 | bio->bi_private = peer_req; |
fcefa62e | 1192 | bio->bi_end_io = drbd_peer_request_endio; |
45bb912b LE |
1193 | |
1194 | bio->bi_next = bios; | |
1195 | bios = bio; | |
1196 | ++n_bios; | |
1197 | ||
1198 | page_chain_for_each(page) { | |
1199 | unsigned len = min_t(unsigned, ds, PAGE_SIZE); | |
1200 | if (!bio_add_page(bio, page, len, 0)) { | |
10f6d992 LE |
1201 | /* A single page must always be possible! |
1202 | * But in case it fails anyways, | |
1203 | * we deal with it, and complain (below). */ | |
1204 | if (bio->bi_vcnt == 0) { | |
1205 | dev_err(DEV, | |
1206 | "bio_add_page failed for len=%u, " | |
1207 | "bi_vcnt=0 (bi_sector=%llu)\n", | |
1208 | len, (unsigned long long)bio->bi_sector); | |
1209 | err = -ENOSPC; | |
1210 | goto fail; | |
1211 | } | |
45bb912b LE |
1212 | goto next_bio; |
1213 | } | |
1214 | ds -= len; | |
1215 | sector += len >> 9; | |
1216 | --nr_pages; | |
1217 | } | |
1218 | D_ASSERT(page == NULL); | |
1219 | D_ASSERT(ds == 0); | |
1220 | ||
db830c46 | 1221 | atomic_set(&peer_req->pending_bios, n_bios); |
45bb912b LE |
1222 | do { |
1223 | bio = bios; | |
1224 | bios = bios->bi_next; | |
1225 | bio->bi_next = NULL; | |
1226 | ||
45bb912b | 1227 | drbd_generic_make_request(mdev, fault_type, bio); |
45bb912b | 1228 | } while (bios); |
45bb912b LE |
1229 | return 0; |
1230 | ||
1231 | fail: | |
1232 | while (bios) { | |
1233 | bio = bios; | |
1234 | bios = bios->bi_next; | |
1235 | bio_put(bio); | |
1236 | } | |
10f6d992 | 1237 | return err; |
45bb912b LE |
1238 | } |
1239 | ||
53840641 | 1240 | static void drbd_remove_epoch_entry_interval(struct drbd_conf *mdev, |
db830c46 | 1241 | struct drbd_peer_request *peer_req) |
53840641 | 1242 | { |
db830c46 | 1243 | struct drbd_interval *i = &peer_req->i; |
53840641 AG |
1244 | |
1245 | drbd_remove_interval(&mdev->write_requests, i); | |
1246 | drbd_clear_interval(i); | |
1247 | ||
6c852bec | 1248 | /* Wake up any processes waiting for this peer request to complete. */ |
53840641 AG |
1249 | if (i->waiting) |
1250 | wake_up(&mdev->misc_wait); | |
1251 | } | |
1252 | ||
4a76b161 | 1253 | static int receive_Barrier(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 1254 | { |
4a76b161 | 1255 | struct drbd_conf *mdev; |
2451fc3b | 1256 | int rv; |
e658983a | 1257 | struct p_barrier *p = pi->data; |
b411b363 PR |
1258 | struct drbd_epoch *epoch; |
1259 | ||
4a76b161 AG |
1260 | mdev = vnr_to_mdev(tconn, pi->vnr); |
1261 | if (!mdev) | |
1262 | return -EIO; | |
1263 | ||
b411b363 PR |
1264 | inc_unacked(mdev); |
1265 | ||
b411b363 PR |
1266 | mdev->current_epoch->barrier_nr = p->barrier; |
1267 | rv = drbd_may_finish_epoch(mdev, mdev->current_epoch, EV_GOT_BARRIER_NR); | |
1268 | ||
1269 | /* P_BARRIER_ACK may imply that the corresponding extent is dropped from | |
1270 | * the activity log, which means it would not be resynced in case the | |
1271 | * R_PRIMARY crashes now. | |
1272 | * Therefore we must send the barrier_ack after the barrier request was | |
1273 | * completed. */ | |
1274 | switch (mdev->write_ordering) { | |
b411b363 PR |
1275 | case WO_none: |
1276 | if (rv == FE_RECYCLED) | |
82bc0194 | 1277 | return 0; |
2451fc3b PR |
1278 | |
1279 | /* receiver context, in the writeout path of the other node. | |
1280 | * avoid potential distributed deadlock */ | |
1281 | epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); | |
1282 | if (epoch) | |
1283 | break; | |
1284 | else | |
1285 | dev_warn(DEV, "Allocation of an epoch failed, slowing down\n"); | |
1286 | /* Fall through */ | |
b411b363 PR |
1287 | |
1288 | case WO_bdev_flush: | |
1289 | case WO_drain_io: | |
b411b363 | 1290 | drbd_wait_ee_list_empty(mdev, &mdev->active_ee); |
2451fc3b PR |
1291 | drbd_flush(mdev); |
1292 | ||
1293 | if (atomic_read(&mdev->current_epoch->epoch_size)) { | |
1294 | epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); | |
1295 | if (epoch) | |
1296 | break; | |
b411b363 PR |
1297 | } |
1298 | ||
2451fc3b PR |
1299 | epoch = mdev->current_epoch; |
1300 | wait_event(mdev->ee_wait, atomic_read(&epoch->epoch_size) == 0); | |
1301 | ||
1302 | D_ASSERT(atomic_read(&epoch->active) == 0); | |
1303 | D_ASSERT(epoch->flags == 0); | |
b411b363 | 1304 | |
82bc0194 | 1305 | return 0; |
2451fc3b PR |
1306 | default: |
1307 | dev_err(DEV, "Strangeness in mdev->write_ordering %d\n", mdev->write_ordering); | |
82bc0194 | 1308 | return -EIO; |
b411b363 PR |
1309 | } |
1310 | ||
1311 | epoch->flags = 0; | |
1312 | atomic_set(&epoch->epoch_size, 0); | |
1313 | atomic_set(&epoch->active, 0); | |
1314 | ||
1315 | spin_lock(&mdev->epoch_lock); | |
1316 | if (atomic_read(&mdev->current_epoch->epoch_size)) { | |
1317 | list_add(&epoch->list, &mdev->current_epoch->list); | |
1318 | mdev->current_epoch = epoch; | |
1319 | mdev->epochs++; | |
b411b363 PR |
1320 | } else { |
1321 | /* The current_epoch got recycled while we allocated this one... */ | |
1322 | kfree(epoch); | |
1323 | } | |
1324 | spin_unlock(&mdev->epoch_lock); | |
1325 | ||
82bc0194 | 1326 | return 0; |
b411b363 PR |
1327 | } |
1328 | ||
1329 | /* used from receive_RSDataReply (recv_resync_read) | |
1330 | * and from receive_Data */ | |
f6ffca9f AG |
1331 | static struct drbd_peer_request * |
1332 | read_in_block(struct drbd_conf *mdev, u64 id, sector_t sector, | |
1333 | int data_size) __must_hold(local) | |
b411b363 | 1334 | { |
6666032a | 1335 | const sector_t capacity = drbd_get_capacity(mdev->this_bdev); |
db830c46 | 1336 | struct drbd_peer_request *peer_req; |
b411b363 | 1337 | struct page *page; |
a5c31904 | 1338 | int dgs, ds, err; |
a0638456 PR |
1339 | void *dig_in = mdev->tconn->int_dig_in; |
1340 | void *dig_vv = mdev->tconn->int_dig_vv; | |
6b4388ac | 1341 | unsigned long *data; |
b411b363 | 1342 | |
a0638456 PR |
1343 | dgs = (mdev->tconn->agreed_pro_version >= 87 && mdev->tconn->integrity_r_tfm) ? |
1344 | crypto_hash_digestsize(mdev->tconn->integrity_r_tfm) : 0; | |
b411b363 PR |
1345 | |
1346 | if (dgs) { | |
9f5bdc33 AG |
1347 | /* |
1348 | * FIXME: Receive the incoming digest into the receive buffer | |
1349 | * here, together with its struct p_data? | |
1350 | */ | |
a5c31904 AG |
1351 | err = drbd_recv_all_warn(mdev->tconn, dig_in, dgs); |
1352 | if (err) | |
b411b363 | 1353 | return NULL; |
b411b363 PR |
1354 | } |
1355 | ||
1356 | data_size -= dgs; | |
1357 | ||
841ce241 AG |
1358 | if (!expect(data_size != 0)) |
1359 | return NULL; | |
1360 | if (!expect(IS_ALIGNED(data_size, 512))) | |
1361 | return NULL; | |
1362 | if (!expect(data_size <= DRBD_MAX_BIO_SIZE)) | |
1363 | return NULL; | |
b411b363 | 1364 | |
6666032a LE |
1365 | /* even though we trust out peer, |
1366 | * we sometimes have to double check. */ | |
1367 | if (sector + (data_size>>9) > capacity) { | |
fdda6544 LE |
1368 | dev_err(DEV, "request from peer beyond end of local disk: " |
1369 | "capacity: %llus < sector: %llus + size: %u\n", | |
6666032a LE |
1370 | (unsigned long long)capacity, |
1371 | (unsigned long long)sector, data_size); | |
1372 | return NULL; | |
1373 | } | |
1374 | ||
b411b363 PR |
1375 | /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD |
1376 | * "criss-cross" setup, that might cause write-out on some other DRBD, | |
1377 | * which in turn might block on the other node at this very place. */ | |
0db55363 | 1378 | peer_req = drbd_alloc_peer_req(mdev, id, sector, data_size, GFP_NOIO); |
db830c46 | 1379 | if (!peer_req) |
b411b363 | 1380 | return NULL; |
45bb912b | 1381 | |
b411b363 | 1382 | ds = data_size; |
db830c46 | 1383 | page = peer_req->pages; |
45bb912b LE |
1384 | page_chain_for_each(page) { |
1385 | unsigned len = min_t(int, ds, PAGE_SIZE); | |
6b4388ac | 1386 | data = kmap(page); |
a5c31904 | 1387 | err = drbd_recv_all_warn(mdev->tconn, data, len); |
0cf9d27e | 1388 | if (drbd_insert_fault(mdev, DRBD_FAULT_RECEIVE)) { |
6b4388ac PR |
1389 | dev_err(DEV, "Fault injection: Corrupting data on receive\n"); |
1390 | data[0] = data[0] ^ (unsigned long)-1; | |
1391 | } | |
b411b363 | 1392 | kunmap(page); |
a5c31904 | 1393 | if (err) { |
3967deb1 | 1394 | drbd_free_peer_req(mdev, peer_req); |
b411b363 PR |
1395 | return NULL; |
1396 | } | |
a5c31904 | 1397 | ds -= len; |
b411b363 PR |
1398 | } |
1399 | ||
1400 | if (dgs) { | |
db830c46 | 1401 | drbd_csum_ee(mdev, mdev->tconn->integrity_r_tfm, peer_req, dig_vv); |
b411b363 | 1402 | if (memcmp(dig_in, dig_vv, dgs)) { |
470be44a LE |
1403 | dev_err(DEV, "Digest integrity check FAILED: %llus +%u\n", |
1404 | (unsigned long long)sector, data_size); | |
3967deb1 | 1405 | drbd_free_peer_req(mdev, peer_req); |
b411b363 PR |
1406 | return NULL; |
1407 | } | |
1408 | } | |
1409 | mdev->recv_cnt += data_size>>9; | |
db830c46 | 1410 | return peer_req; |
b411b363 PR |
1411 | } |
1412 | ||
1413 | /* drbd_drain_block() just takes a data block | |
1414 | * out of the socket input buffer, and discards it. | |
1415 | */ | |
1416 | static int drbd_drain_block(struct drbd_conf *mdev, int data_size) | |
1417 | { | |
1418 | struct page *page; | |
a5c31904 | 1419 | int err = 0; |
b411b363 PR |
1420 | void *data; |
1421 | ||
c3470cde | 1422 | if (!data_size) |
fc5be839 | 1423 | return 0; |
c3470cde | 1424 | |
45bb912b | 1425 | page = drbd_pp_alloc(mdev, 1, 1); |
b411b363 PR |
1426 | |
1427 | data = kmap(page); | |
1428 | while (data_size) { | |
fc5be839 AG |
1429 | unsigned int len = min_t(int, data_size, PAGE_SIZE); |
1430 | ||
a5c31904 AG |
1431 | err = drbd_recv_all_warn(mdev->tconn, data, len); |
1432 | if (err) | |
b411b363 | 1433 | break; |
a5c31904 | 1434 | data_size -= len; |
b411b363 PR |
1435 | } |
1436 | kunmap(page); | |
435f0740 | 1437 | drbd_pp_free(mdev, page, 0); |
fc5be839 | 1438 | return err; |
b411b363 PR |
1439 | } |
1440 | ||
1441 | static int recv_dless_read(struct drbd_conf *mdev, struct drbd_request *req, | |
1442 | sector_t sector, int data_size) | |
1443 | { | |
1444 | struct bio_vec *bvec; | |
1445 | struct bio *bio; | |
a5c31904 | 1446 | int dgs, err, i, expect; |
a0638456 PR |
1447 | void *dig_in = mdev->tconn->int_dig_in; |
1448 | void *dig_vv = mdev->tconn->int_dig_vv; | |
b411b363 | 1449 | |
a0638456 PR |
1450 | dgs = (mdev->tconn->agreed_pro_version >= 87 && mdev->tconn->integrity_r_tfm) ? |
1451 | crypto_hash_digestsize(mdev->tconn->integrity_r_tfm) : 0; | |
b411b363 PR |
1452 | |
1453 | if (dgs) { | |
a5c31904 AG |
1454 | err = drbd_recv_all_warn(mdev->tconn, dig_in, dgs); |
1455 | if (err) | |
1456 | return err; | |
b411b363 PR |
1457 | } |
1458 | ||
1459 | data_size -= dgs; | |
1460 | ||
1461 | /* optimistically update recv_cnt. if receiving fails below, | |
1462 | * we disconnect anyways, and counters will be reset. */ | |
1463 | mdev->recv_cnt += data_size>>9; | |
1464 | ||
1465 | bio = req->master_bio; | |
1466 | D_ASSERT(sector == bio->bi_sector); | |
1467 | ||
1468 | bio_for_each_segment(bvec, bio, i) { | |
a5c31904 | 1469 | void *mapped = kmap(bvec->bv_page) + bvec->bv_offset; |
b411b363 | 1470 | expect = min_t(int, data_size, bvec->bv_len); |
a5c31904 | 1471 | err = drbd_recv_all_warn(mdev->tconn, mapped, expect); |
b411b363 | 1472 | kunmap(bvec->bv_page); |
a5c31904 AG |
1473 | if (err) |
1474 | return err; | |
1475 | data_size -= expect; | |
b411b363 PR |
1476 | } |
1477 | ||
1478 | if (dgs) { | |
a0638456 | 1479 | drbd_csum_bio(mdev, mdev->tconn->integrity_r_tfm, bio, dig_vv); |
b411b363 PR |
1480 | if (memcmp(dig_in, dig_vv, dgs)) { |
1481 | dev_err(DEV, "Digest integrity check FAILED. Broken NICs?\n"); | |
28284cef | 1482 | return -EINVAL; |
b411b363 PR |
1483 | } |
1484 | } | |
1485 | ||
1486 | D_ASSERT(data_size == 0); | |
28284cef | 1487 | return 0; |
b411b363 PR |
1488 | } |
1489 | ||
a990be46 AG |
1490 | /* |
1491 | * e_end_resync_block() is called in asender context via | |
1492 | * drbd_finish_peer_reqs(). | |
1493 | */ | |
99920dc5 | 1494 | static int e_end_resync_block(struct drbd_work *w, int unused) |
b411b363 | 1495 | { |
8050e6d0 AG |
1496 | struct drbd_peer_request *peer_req = |
1497 | container_of(w, struct drbd_peer_request, w); | |
00d56944 | 1498 | struct drbd_conf *mdev = w->mdev; |
db830c46 | 1499 | sector_t sector = peer_req->i.sector; |
99920dc5 | 1500 | int err; |
b411b363 | 1501 | |
db830c46 | 1502 | D_ASSERT(drbd_interval_empty(&peer_req->i)); |
b411b363 | 1503 | |
db830c46 AG |
1504 | if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) { |
1505 | drbd_set_in_sync(mdev, sector, peer_req->i.size); | |
99920dc5 | 1506 | err = drbd_send_ack(mdev, P_RS_WRITE_ACK, peer_req); |
b411b363 PR |
1507 | } else { |
1508 | /* Record failure to sync */ | |
db830c46 | 1509 | drbd_rs_failed_io(mdev, sector, peer_req->i.size); |
b411b363 | 1510 | |
99920dc5 | 1511 | err = drbd_send_ack(mdev, P_NEG_ACK, peer_req); |
b411b363 PR |
1512 | } |
1513 | dec_unacked(mdev); | |
1514 | ||
99920dc5 | 1515 | return err; |
b411b363 PR |
1516 | } |
1517 | ||
1518 | static int recv_resync_read(struct drbd_conf *mdev, sector_t sector, int data_size) __releases(local) | |
1519 | { | |
db830c46 | 1520 | struct drbd_peer_request *peer_req; |
b411b363 | 1521 | |
db830c46 AG |
1522 | peer_req = read_in_block(mdev, ID_SYNCER, sector, data_size); |
1523 | if (!peer_req) | |
45bb912b | 1524 | goto fail; |
b411b363 PR |
1525 | |
1526 | dec_rs_pending(mdev); | |
1527 | ||
b411b363 PR |
1528 | inc_unacked(mdev); |
1529 | /* corresponding dec_unacked() in e_end_resync_block() | |
1530 | * respective _drbd_clear_done_ee */ | |
1531 | ||
db830c46 | 1532 | peer_req->w.cb = e_end_resync_block; |
45bb912b | 1533 | |
87eeee41 | 1534 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 1535 | list_add(&peer_req->w.list, &mdev->sync_ee); |
87eeee41 | 1536 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 1537 | |
0f0601f4 | 1538 | atomic_add(data_size >> 9, &mdev->rs_sect_ev); |
fbe29dec | 1539 | if (drbd_submit_peer_request(mdev, peer_req, WRITE, DRBD_FAULT_RS_WR) == 0) |
e1c1b0fc | 1540 | return 0; |
b411b363 | 1541 | |
10f6d992 LE |
1542 | /* don't care for the reason here */ |
1543 | dev_err(DEV, "submit failed, triggering re-connect\n"); | |
87eeee41 | 1544 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 1545 | list_del(&peer_req->w.list); |
87eeee41 | 1546 | spin_unlock_irq(&mdev->tconn->req_lock); |
22cc37a9 | 1547 | |
3967deb1 | 1548 | drbd_free_peer_req(mdev, peer_req); |
45bb912b LE |
1549 | fail: |
1550 | put_ldev(mdev); | |
e1c1b0fc | 1551 | return -EIO; |
b411b363 PR |
1552 | } |
1553 | ||
668eebc6 | 1554 | static struct drbd_request * |
bc9c5c41 AG |
1555 | find_request(struct drbd_conf *mdev, struct rb_root *root, u64 id, |
1556 | sector_t sector, bool missing_ok, const char *func) | |
51624585 | 1557 | { |
51624585 AG |
1558 | struct drbd_request *req; |
1559 | ||
bc9c5c41 AG |
1560 | /* Request object according to our peer */ |
1561 | req = (struct drbd_request *)(unsigned long)id; | |
5e472264 | 1562 | if (drbd_contains_interval(root, sector, &req->i) && req->i.local) |
668eebc6 | 1563 | return req; |
c3afd8f5 AG |
1564 | if (!missing_ok) { |
1565 | dev_err(DEV, "%s: failed to find request %lu, sector %llus\n", func, | |
1566 | (unsigned long)id, (unsigned long long)sector); | |
1567 | } | |
51624585 AG |
1568 | return NULL; |
1569 | } | |
1570 | ||
4a76b161 | 1571 | static int receive_DataReply(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 1572 | { |
4a76b161 | 1573 | struct drbd_conf *mdev; |
b411b363 PR |
1574 | struct drbd_request *req; |
1575 | sector_t sector; | |
82bc0194 | 1576 | int err; |
e658983a | 1577 | struct p_data *p = pi->data; |
4a76b161 AG |
1578 | |
1579 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
1580 | if (!mdev) | |
1581 | return -EIO; | |
b411b363 PR |
1582 | |
1583 | sector = be64_to_cpu(p->sector); | |
1584 | ||
87eeee41 | 1585 | spin_lock_irq(&mdev->tconn->req_lock); |
bc9c5c41 | 1586 | req = find_request(mdev, &mdev->read_requests, p->block_id, sector, false, __func__); |
87eeee41 | 1587 | spin_unlock_irq(&mdev->tconn->req_lock); |
c3afd8f5 | 1588 | if (unlikely(!req)) |
82bc0194 | 1589 | return -EIO; |
b411b363 | 1590 | |
24c4830c | 1591 | /* hlist_del(&req->collision) is done in _req_may_be_done, to avoid |
b411b363 PR |
1592 | * special casing it there for the various failure cases. |
1593 | * still no race with drbd_fail_pending_reads */ | |
e2857216 | 1594 | err = recv_dless_read(mdev, req, sector, pi->size); |
82bc0194 | 1595 | if (!err) |
8554df1c | 1596 | req_mod(req, DATA_RECEIVED); |
b411b363 PR |
1597 | /* else: nothing. handled from drbd_disconnect... |
1598 | * I don't think we may complete this just yet | |
1599 | * in case we are "on-disconnect: freeze" */ | |
1600 | ||
82bc0194 | 1601 | return err; |
b411b363 PR |
1602 | } |
1603 | ||
4a76b161 | 1604 | static int receive_RSDataReply(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 1605 | { |
4a76b161 | 1606 | struct drbd_conf *mdev; |
b411b363 | 1607 | sector_t sector; |
82bc0194 | 1608 | int err; |
e658983a | 1609 | struct p_data *p = pi->data; |
4a76b161 AG |
1610 | |
1611 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
1612 | if (!mdev) | |
1613 | return -EIO; | |
b411b363 PR |
1614 | |
1615 | sector = be64_to_cpu(p->sector); | |
1616 | D_ASSERT(p->block_id == ID_SYNCER); | |
1617 | ||
1618 | if (get_ldev(mdev)) { | |
1619 | /* data is submitted to disk within recv_resync_read. | |
1620 | * corresponding put_ldev done below on error, | |
fcefa62e | 1621 | * or in drbd_peer_request_endio. */ |
e2857216 | 1622 | err = recv_resync_read(mdev, sector, pi->size); |
b411b363 PR |
1623 | } else { |
1624 | if (__ratelimit(&drbd_ratelimit_state)) | |
1625 | dev_err(DEV, "Can not write resync data to local disk.\n"); | |
1626 | ||
e2857216 | 1627 | err = drbd_drain_block(mdev, pi->size); |
b411b363 | 1628 | |
e2857216 | 1629 | drbd_send_ack_dp(mdev, P_NEG_ACK, p, pi->size); |
b411b363 PR |
1630 | } |
1631 | ||
e2857216 | 1632 | atomic_add(pi->size >> 9, &mdev->rs_sect_in); |
778f271d | 1633 | |
82bc0194 | 1634 | return err; |
b411b363 PR |
1635 | } |
1636 | ||
99920dc5 | 1637 | static int w_restart_write(struct drbd_work *w, int cancel) |
7be8da07 AG |
1638 | { |
1639 | struct drbd_request *req = container_of(w, struct drbd_request, w); | |
1640 | struct drbd_conf *mdev = w->mdev; | |
1641 | struct bio *bio; | |
1642 | unsigned long start_time; | |
1643 | unsigned long flags; | |
1644 | ||
1645 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); | |
1646 | if (!expect(req->rq_state & RQ_POSTPONED)) { | |
1647 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
99920dc5 | 1648 | return -EIO; |
7be8da07 AG |
1649 | } |
1650 | bio = req->master_bio; | |
1651 | start_time = req->start_time; | |
1652 | /* Postponed requests will not have their master_bio completed! */ | |
1653 | __req_mod(req, DISCARD_WRITE, NULL); | |
1654 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); | |
1655 | ||
1656 | while (__drbd_make_request(mdev, bio, start_time)) | |
1657 | /* retry */ ; | |
99920dc5 | 1658 | return 0; |
7be8da07 AG |
1659 | } |
1660 | ||
1661 | static void restart_conflicting_writes(struct drbd_conf *mdev, | |
1662 | sector_t sector, int size) | |
1663 | { | |
1664 | struct drbd_interval *i; | |
1665 | struct drbd_request *req; | |
1666 | ||
1667 | drbd_for_each_overlap(i, &mdev->write_requests, sector, size) { | |
1668 | if (!i->local) | |
1669 | continue; | |
1670 | req = container_of(i, struct drbd_request, i); | |
1671 | if (req->rq_state & RQ_LOCAL_PENDING || | |
1672 | !(req->rq_state & RQ_POSTPONED)) | |
1673 | continue; | |
1674 | if (expect(list_empty(&req->w.list))) { | |
1675 | req->w.mdev = mdev; | |
1676 | req->w.cb = w_restart_write; | |
1677 | drbd_queue_work(&mdev->tconn->data.work, &req->w); | |
1678 | } | |
1679 | } | |
1680 | } | |
1681 | ||
a990be46 AG |
1682 | /* |
1683 | * e_end_block() is called in asender context via drbd_finish_peer_reqs(). | |
b411b363 | 1684 | */ |
99920dc5 | 1685 | static int e_end_block(struct drbd_work *w, int cancel) |
b411b363 | 1686 | { |
8050e6d0 AG |
1687 | struct drbd_peer_request *peer_req = |
1688 | container_of(w, struct drbd_peer_request, w); | |
00d56944 | 1689 | struct drbd_conf *mdev = w->mdev; |
db830c46 | 1690 | sector_t sector = peer_req->i.sector; |
99920dc5 | 1691 | int err = 0, pcmd; |
b411b363 | 1692 | |
89e58e75 | 1693 | if (mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C) { |
db830c46 | 1694 | if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
1695 | pcmd = (mdev->state.conn >= C_SYNC_SOURCE && |
1696 | mdev->state.conn <= C_PAUSED_SYNC_T && | |
db830c46 | 1697 | peer_req->flags & EE_MAY_SET_IN_SYNC) ? |
b411b363 | 1698 | P_RS_WRITE_ACK : P_WRITE_ACK; |
99920dc5 | 1699 | err = drbd_send_ack(mdev, pcmd, peer_req); |
b411b363 | 1700 | if (pcmd == P_RS_WRITE_ACK) |
db830c46 | 1701 | drbd_set_in_sync(mdev, sector, peer_req->i.size); |
b411b363 | 1702 | } else { |
99920dc5 | 1703 | err = drbd_send_ack(mdev, P_NEG_ACK, peer_req); |
b411b363 PR |
1704 | /* we expect it to be marked out of sync anyways... |
1705 | * maybe assert this? */ | |
1706 | } | |
1707 | dec_unacked(mdev); | |
1708 | } | |
1709 | /* we delete from the conflict detection hash _after_ we sent out the | |
1710 | * P_WRITE_ACK / P_NEG_ACK, to get the sequence number right. */ | |
89e58e75 | 1711 | if (mdev->tconn->net_conf->two_primaries) { |
87eeee41 | 1712 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 AG |
1713 | D_ASSERT(!drbd_interval_empty(&peer_req->i)); |
1714 | drbd_remove_epoch_entry_interval(mdev, peer_req); | |
7be8da07 AG |
1715 | if (peer_req->flags & EE_RESTART_REQUESTS) |
1716 | restart_conflicting_writes(mdev, sector, peer_req->i.size); | |
87eeee41 | 1717 | spin_unlock_irq(&mdev->tconn->req_lock); |
bb3bfe96 | 1718 | } else |
db830c46 | 1719 | D_ASSERT(drbd_interval_empty(&peer_req->i)); |
b411b363 | 1720 | |
db830c46 | 1721 | drbd_may_finish_epoch(mdev, peer_req->epoch, EV_PUT + (cancel ? EV_CLEANUP : 0)); |
b411b363 | 1722 | |
99920dc5 | 1723 | return err; |
b411b363 PR |
1724 | } |
1725 | ||
7be8da07 | 1726 | static int e_send_ack(struct drbd_work *w, enum drbd_packet ack) |
b411b363 | 1727 | { |
7be8da07 | 1728 | struct drbd_conf *mdev = w->mdev; |
8050e6d0 AG |
1729 | struct drbd_peer_request *peer_req = |
1730 | container_of(w, struct drbd_peer_request, w); | |
99920dc5 | 1731 | int err; |
b411b363 | 1732 | |
99920dc5 | 1733 | err = drbd_send_ack(mdev, ack, peer_req); |
b411b363 PR |
1734 | dec_unacked(mdev); |
1735 | ||
99920dc5 | 1736 | return err; |
b411b363 PR |
1737 | } |
1738 | ||
99920dc5 | 1739 | static int e_send_discard_write(struct drbd_work *w, int unused) |
7be8da07 AG |
1740 | { |
1741 | return e_send_ack(w, P_DISCARD_WRITE); | |
1742 | } | |
1743 | ||
99920dc5 | 1744 | static int e_send_retry_write(struct drbd_work *w, int unused) |
7be8da07 AG |
1745 | { |
1746 | struct drbd_tconn *tconn = w->mdev->tconn; | |
1747 | ||
1748 | return e_send_ack(w, tconn->agreed_pro_version >= 100 ? | |
1749 | P_RETRY_WRITE : P_DISCARD_WRITE); | |
1750 | } | |
1751 | ||
3e394da1 AG |
1752 | static bool seq_greater(u32 a, u32 b) |
1753 | { | |
1754 | /* | |
1755 | * We assume 32-bit wrap-around here. | |
1756 | * For 24-bit wrap-around, we would have to shift: | |
1757 | * a <<= 8; b <<= 8; | |
1758 | */ | |
1759 | return (s32)a - (s32)b > 0; | |
1760 | } | |
1761 | ||
1762 | static u32 seq_max(u32 a, u32 b) | |
1763 | { | |
1764 | return seq_greater(a, b) ? a : b; | |
1765 | } | |
1766 | ||
7be8da07 AG |
1767 | static bool need_peer_seq(struct drbd_conf *mdev) |
1768 | { | |
1769 | struct drbd_tconn *tconn = mdev->tconn; | |
1770 | ||
1771 | /* | |
1772 | * We only need to keep track of the last packet_seq number of our peer | |
1773 | * if we are in dual-primary mode and we have the discard flag set; see | |
1774 | * handle_write_conflicts(). | |
1775 | */ | |
1776 | return tconn->net_conf->two_primaries && | |
1777 | test_bit(DISCARD_CONCURRENT, &tconn->flags); | |
1778 | } | |
1779 | ||
43ae077d | 1780 | static void update_peer_seq(struct drbd_conf *mdev, unsigned int peer_seq) |
3e394da1 | 1781 | { |
3c13b680 | 1782 | unsigned int newest_peer_seq; |
3e394da1 | 1783 | |
7be8da07 AG |
1784 | if (need_peer_seq(mdev)) { |
1785 | spin_lock(&mdev->peer_seq_lock); | |
3c13b680 LE |
1786 | newest_peer_seq = seq_max(mdev->peer_seq, peer_seq); |
1787 | mdev->peer_seq = newest_peer_seq; | |
7be8da07 | 1788 | spin_unlock(&mdev->peer_seq_lock); |
3c13b680 LE |
1789 | /* wake up only if we actually changed mdev->peer_seq */ |
1790 | if (peer_seq == newest_peer_seq) | |
7be8da07 AG |
1791 | wake_up(&mdev->seq_wait); |
1792 | } | |
3e394da1 AG |
1793 | } |
1794 | ||
b411b363 PR |
1795 | /* Called from receive_Data. |
1796 | * Synchronize packets on sock with packets on msock. | |
1797 | * | |
1798 | * This is here so even when a P_DATA packet traveling via sock overtook an Ack | |
1799 | * packet traveling on msock, they are still processed in the order they have | |
1800 | * been sent. | |
1801 | * | |
1802 | * Note: we don't care for Ack packets overtaking P_DATA packets. | |
1803 | * | |
1804 | * In case packet_seq is larger than mdev->peer_seq number, there are | |
1805 | * outstanding packets on the msock. We wait for them to arrive. | |
1806 | * In case we are the logically next packet, we update mdev->peer_seq | |
1807 | * ourselves. Correctly handles 32bit wrap around. | |
1808 | * | |
1809 | * Assume we have a 10 GBit connection, that is about 1<<30 byte per second, | |
1810 | * about 1<<21 sectors per second. So "worst" case, we have 1<<3 == 8 seconds | |
1811 | * for the 24bit wrap (historical atomic_t guarantee on some archs), and we have | |
1812 | * 1<<9 == 512 seconds aka ages for the 32bit wrap around... | |
1813 | * | |
1814 | * returns 0 if we may process the packet, | |
1815 | * -ERESTARTSYS if we were interrupted (by disconnect signal). */ | |
7be8da07 | 1816 | static int wait_for_and_update_peer_seq(struct drbd_conf *mdev, const u32 peer_seq) |
b411b363 PR |
1817 | { |
1818 | DEFINE_WAIT(wait); | |
b411b363 | 1819 | long timeout; |
7be8da07 AG |
1820 | int ret; |
1821 | ||
1822 | if (!need_peer_seq(mdev)) | |
1823 | return 0; | |
1824 | ||
b411b363 PR |
1825 | spin_lock(&mdev->peer_seq_lock); |
1826 | for (;;) { | |
7be8da07 AG |
1827 | if (!seq_greater(peer_seq - 1, mdev->peer_seq)) { |
1828 | mdev->peer_seq = seq_max(mdev->peer_seq, peer_seq); | |
1829 | ret = 0; | |
b411b363 | 1830 | break; |
7be8da07 | 1831 | } |
b411b363 PR |
1832 | if (signal_pending(current)) { |
1833 | ret = -ERESTARTSYS; | |
1834 | break; | |
1835 | } | |
7be8da07 | 1836 | prepare_to_wait(&mdev->seq_wait, &wait, TASK_INTERRUPTIBLE); |
b411b363 | 1837 | spin_unlock(&mdev->peer_seq_lock); |
71b1c1eb AG |
1838 | timeout = mdev->tconn->net_conf->ping_timeo*HZ/10; |
1839 | timeout = schedule_timeout(timeout); | |
b411b363 | 1840 | spin_lock(&mdev->peer_seq_lock); |
7be8da07 | 1841 | if (!timeout) { |
b411b363 | 1842 | ret = -ETIMEDOUT; |
71b1c1eb | 1843 | dev_err(DEV, "Timed out waiting for missing ack packets; disconnecting\n"); |
b411b363 PR |
1844 | break; |
1845 | } | |
1846 | } | |
b411b363 | 1847 | spin_unlock(&mdev->peer_seq_lock); |
7be8da07 | 1848 | finish_wait(&mdev->seq_wait, &wait); |
b411b363 PR |
1849 | return ret; |
1850 | } | |
1851 | ||
688593c5 LE |
1852 | /* see also bio_flags_to_wire() |
1853 | * DRBD_REQ_*, because we need to semantically map the flags to data packet | |
1854 | * flags and back. We may replicate to other kernel versions. */ | |
1855 | static unsigned long wire_flags_to_bio(struct drbd_conf *mdev, u32 dpf) | |
76d2e7ec | 1856 | { |
688593c5 LE |
1857 | return (dpf & DP_RW_SYNC ? REQ_SYNC : 0) | |
1858 | (dpf & DP_FUA ? REQ_FUA : 0) | | |
1859 | (dpf & DP_FLUSH ? REQ_FLUSH : 0) | | |
1860 | (dpf & DP_DISCARD ? REQ_DISCARD : 0); | |
76d2e7ec PR |
1861 | } |
1862 | ||
7be8da07 AG |
1863 | static void fail_postponed_requests(struct drbd_conf *mdev, sector_t sector, |
1864 | unsigned int size) | |
1865 | { | |
1866 | struct drbd_interval *i; | |
1867 | ||
1868 | repeat: | |
1869 | drbd_for_each_overlap(i, &mdev->write_requests, sector, size) { | |
1870 | struct drbd_request *req; | |
1871 | struct bio_and_error m; | |
1872 | ||
1873 | if (!i->local) | |
1874 | continue; | |
1875 | req = container_of(i, struct drbd_request, i); | |
1876 | if (!(req->rq_state & RQ_POSTPONED)) | |
1877 | continue; | |
1878 | req->rq_state &= ~RQ_POSTPONED; | |
1879 | __req_mod(req, NEG_ACKED, &m); | |
1880 | spin_unlock_irq(&mdev->tconn->req_lock); | |
1881 | if (m.bio) | |
1882 | complete_master_bio(mdev, &m); | |
1883 | spin_lock_irq(&mdev->tconn->req_lock); | |
1884 | goto repeat; | |
1885 | } | |
1886 | } | |
1887 | ||
1888 | static int handle_write_conflicts(struct drbd_conf *mdev, | |
1889 | struct drbd_peer_request *peer_req) | |
1890 | { | |
1891 | struct drbd_tconn *tconn = mdev->tconn; | |
1892 | bool resolve_conflicts = test_bit(DISCARD_CONCURRENT, &tconn->flags); | |
1893 | sector_t sector = peer_req->i.sector; | |
1894 | const unsigned int size = peer_req->i.size; | |
1895 | struct drbd_interval *i; | |
1896 | bool equal; | |
1897 | int err; | |
1898 | ||
1899 | /* | |
1900 | * Inserting the peer request into the write_requests tree will prevent | |
1901 | * new conflicting local requests from being added. | |
1902 | */ | |
1903 | drbd_insert_interval(&mdev->write_requests, &peer_req->i); | |
1904 | ||
1905 | repeat: | |
1906 | drbd_for_each_overlap(i, &mdev->write_requests, sector, size) { | |
1907 | if (i == &peer_req->i) | |
1908 | continue; | |
1909 | ||
1910 | if (!i->local) { | |
1911 | /* | |
1912 | * Our peer has sent a conflicting remote request; this | |
1913 | * should not happen in a two-node setup. Wait for the | |
1914 | * earlier peer request to complete. | |
1915 | */ | |
1916 | err = drbd_wait_misc(mdev, i); | |
1917 | if (err) | |
1918 | goto out; | |
1919 | goto repeat; | |
1920 | } | |
1921 | ||
1922 | equal = i->sector == sector && i->size == size; | |
1923 | if (resolve_conflicts) { | |
1924 | /* | |
1925 | * If the peer request is fully contained within the | |
1926 | * overlapping request, it can be discarded; otherwise, | |
1927 | * it will be retried once all overlapping requests | |
1928 | * have completed. | |
1929 | */ | |
1930 | bool discard = i->sector <= sector && i->sector + | |
1931 | (i->size >> 9) >= sector + (size >> 9); | |
1932 | ||
1933 | if (!equal) | |
1934 | dev_alert(DEV, "Concurrent writes detected: " | |
1935 | "local=%llus +%u, remote=%llus +%u, " | |
1936 | "assuming %s came first\n", | |
1937 | (unsigned long long)i->sector, i->size, | |
1938 | (unsigned long long)sector, size, | |
1939 | discard ? "local" : "remote"); | |
1940 | ||
1941 | inc_unacked(mdev); | |
1942 | peer_req->w.cb = discard ? e_send_discard_write : | |
1943 | e_send_retry_write; | |
1944 | list_add_tail(&peer_req->w.list, &mdev->done_ee); | |
1945 | wake_asender(mdev->tconn); | |
1946 | ||
1947 | err = -ENOENT; | |
1948 | goto out; | |
1949 | } else { | |
1950 | struct drbd_request *req = | |
1951 | container_of(i, struct drbd_request, i); | |
1952 | ||
1953 | if (!equal) | |
1954 | dev_alert(DEV, "Concurrent writes detected: " | |
1955 | "local=%llus +%u, remote=%llus +%u\n", | |
1956 | (unsigned long long)i->sector, i->size, | |
1957 | (unsigned long long)sector, size); | |
1958 | ||
1959 | if (req->rq_state & RQ_LOCAL_PENDING || | |
1960 | !(req->rq_state & RQ_POSTPONED)) { | |
1961 | /* | |
1962 | * Wait for the node with the discard flag to | |
1963 | * decide if this request will be discarded or | |
1964 | * retried. Requests that are discarded will | |
1965 | * disappear from the write_requests tree. | |
1966 | * | |
1967 | * In addition, wait for the conflicting | |
1968 | * request to finish locally before submitting | |
1969 | * the conflicting peer request. | |
1970 | */ | |
1971 | err = drbd_wait_misc(mdev, &req->i); | |
1972 | if (err) { | |
1973 | _conn_request_state(mdev->tconn, | |
1974 | NS(conn, C_TIMEOUT), | |
1975 | CS_HARD); | |
1976 | fail_postponed_requests(mdev, sector, size); | |
1977 | goto out; | |
1978 | } | |
1979 | goto repeat; | |
1980 | } | |
1981 | /* | |
1982 | * Remember to restart the conflicting requests after | |
1983 | * the new peer request has completed. | |
1984 | */ | |
1985 | peer_req->flags |= EE_RESTART_REQUESTS; | |
1986 | } | |
1987 | } | |
1988 | err = 0; | |
1989 | ||
1990 | out: | |
1991 | if (err) | |
1992 | drbd_remove_epoch_entry_interval(mdev, peer_req); | |
1993 | return err; | |
1994 | } | |
1995 | ||
b411b363 | 1996 | /* mirrored write */ |
4a76b161 | 1997 | static int receive_Data(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 1998 | { |
4a76b161 | 1999 | struct drbd_conf *mdev; |
b411b363 | 2000 | sector_t sector; |
db830c46 | 2001 | struct drbd_peer_request *peer_req; |
e658983a | 2002 | struct p_data *p = pi->data; |
7be8da07 | 2003 | u32 peer_seq = be32_to_cpu(p->seq_num); |
b411b363 PR |
2004 | int rw = WRITE; |
2005 | u32 dp_flags; | |
7be8da07 | 2006 | int err; |
b411b363 | 2007 | |
4a76b161 AG |
2008 | mdev = vnr_to_mdev(tconn, pi->vnr); |
2009 | if (!mdev) | |
2010 | return -EIO; | |
2011 | ||
7be8da07 | 2012 | if (!get_ldev(mdev)) { |
82bc0194 AG |
2013 | int err2; |
2014 | ||
7be8da07 | 2015 | err = wait_for_and_update_peer_seq(mdev, peer_seq); |
e2857216 | 2016 | drbd_send_ack_dp(mdev, P_NEG_ACK, p, pi->size); |
b411b363 | 2017 | atomic_inc(&mdev->current_epoch->epoch_size); |
e2857216 | 2018 | err2 = drbd_drain_block(mdev, pi->size); |
82bc0194 AG |
2019 | if (!err) |
2020 | err = err2; | |
2021 | return err; | |
b411b363 PR |
2022 | } |
2023 | ||
fcefa62e AG |
2024 | /* |
2025 | * Corresponding put_ldev done either below (on various errors), or in | |
2026 | * drbd_peer_request_endio, if we successfully submit the data at the | |
2027 | * end of this function. | |
2028 | */ | |
b411b363 PR |
2029 | |
2030 | sector = be64_to_cpu(p->sector); | |
e2857216 | 2031 | peer_req = read_in_block(mdev, p->block_id, sector, pi->size); |
db830c46 | 2032 | if (!peer_req) { |
b411b363 | 2033 | put_ldev(mdev); |
82bc0194 | 2034 | return -EIO; |
b411b363 PR |
2035 | } |
2036 | ||
db830c46 | 2037 | peer_req->w.cb = e_end_block; |
b411b363 | 2038 | |
688593c5 LE |
2039 | dp_flags = be32_to_cpu(p->dp_flags); |
2040 | rw |= wire_flags_to_bio(mdev, dp_flags); | |
2041 | ||
2042 | if (dp_flags & DP_MAY_SET_IN_SYNC) | |
db830c46 | 2043 | peer_req->flags |= EE_MAY_SET_IN_SYNC; |
688593c5 | 2044 | |
b411b363 | 2045 | spin_lock(&mdev->epoch_lock); |
db830c46 AG |
2046 | peer_req->epoch = mdev->current_epoch; |
2047 | atomic_inc(&peer_req->epoch->epoch_size); | |
2048 | atomic_inc(&peer_req->epoch->active); | |
b411b363 PR |
2049 | spin_unlock(&mdev->epoch_lock); |
2050 | ||
7be8da07 AG |
2051 | if (mdev->tconn->net_conf->two_primaries) { |
2052 | err = wait_for_and_update_peer_seq(mdev, peer_seq); | |
2053 | if (err) | |
b411b363 | 2054 | goto out_interrupted; |
87eeee41 | 2055 | spin_lock_irq(&mdev->tconn->req_lock); |
7be8da07 AG |
2056 | err = handle_write_conflicts(mdev, peer_req); |
2057 | if (err) { | |
2058 | spin_unlock_irq(&mdev->tconn->req_lock); | |
2059 | if (err == -ENOENT) { | |
b411b363 | 2060 | put_ldev(mdev); |
82bc0194 | 2061 | return 0; |
b411b363 | 2062 | } |
7be8da07 | 2063 | goto out_interrupted; |
b411b363 | 2064 | } |
7be8da07 AG |
2065 | } else |
2066 | spin_lock_irq(&mdev->tconn->req_lock); | |
db830c46 | 2067 | list_add(&peer_req->w.list, &mdev->active_ee); |
87eeee41 | 2068 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 2069 | |
89e58e75 | 2070 | switch (mdev->tconn->net_conf->wire_protocol) { |
b411b363 PR |
2071 | case DRBD_PROT_C: |
2072 | inc_unacked(mdev); | |
2073 | /* corresponding dec_unacked() in e_end_block() | |
2074 | * respective _drbd_clear_done_ee */ | |
2075 | break; | |
2076 | case DRBD_PROT_B: | |
2077 | /* I really don't like it that the receiver thread | |
2078 | * sends on the msock, but anyways */ | |
db830c46 | 2079 | drbd_send_ack(mdev, P_RECV_ACK, peer_req); |
b411b363 PR |
2080 | break; |
2081 | case DRBD_PROT_A: | |
2082 | /* nothing to do */ | |
2083 | break; | |
2084 | } | |
2085 | ||
6719fb03 | 2086 | if (mdev->state.pdsk < D_INCONSISTENT) { |
b411b363 | 2087 | /* In case we have the only disk of the cluster, */ |
db830c46 AG |
2088 | drbd_set_out_of_sync(mdev, peer_req->i.sector, peer_req->i.size); |
2089 | peer_req->flags |= EE_CALL_AL_COMPLETE_IO; | |
2090 | peer_req->flags &= ~EE_MAY_SET_IN_SYNC; | |
181286ad | 2091 | drbd_al_begin_io(mdev, &peer_req->i); |
b411b363 PR |
2092 | } |
2093 | ||
82bc0194 AG |
2094 | err = drbd_submit_peer_request(mdev, peer_req, rw, DRBD_FAULT_DT_WR); |
2095 | if (!err) | |
2096 | return 0; | |
b411b363 | 2097 | |
10f6d992 LE |
2098 | /* don't care for the reason here */ |
2099 | dev_err(DEV, "submit failed, triggering re-connect\n"); | |
87eeee41 | 2100 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 AG |
2101 | list_del(&peer_req->w.list); |
2102 | drbd_remove_epoch_entry_interval(mdev, peer_req); | |
87eeee41 | 2103 | spin_unlock_irq(&mdev->tconn->req_lock); |
db830c46 | 2104 | if (peer_req->flags & EE_CALL_AL_COMPLETE_IO) |
181286ad | 2105 | drbd_al_complete_io(mdev, &peer_req->i); |
22cc37a9 | 2106 | |
b411b363 | 2107 | out_interrupted: |
db830c46 | 2108 | drbd_may_finish_epoch(mdev, peer_req->epoch, EV_PUT + EV_CLEANUP); |
b411b363 | 2109 | put_ldev(mdev); |
3967deb1 | 2110 | drbd_free_peer_req(mdev, peer_req); |
82bc0194 | 2111 | return err; |
b411b363 PR |
2112 | } |
2113 | ||
0f0601f4 LE |
2114 | /* We may throttle resync, if the lower device seems to be busy, |
2115 | * and current sync rate is above c_min_rate. | |
2116 | * | |
2117 | * To decide whether or not the lower device is busy, we use a scheme similar | |
2118 | * to MD RAID is_mddev_idle(): if the partition stats reveal "significant" | |
2119 | * (more than 64 sectors) of activity we cannot account for with our own resync | |
2120 | * activity, it obviously is "busy". | |
2121 | * | |
2122 | * The current sync rate used here uses only the most recent two step marks, | |
2123 | * to have a short time average so we can react faster. | |
2124 | */ | |
e3555d85 | 2125 | int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector) |
0f0601f4 LE |
2126 | { |
2127 | struct gendisk *disk = mdev->ldev->backing_bdev->bd_contains->bd_disk; | |
2128 | unsigned long db, dt, dbdt; | |
e3555d85 | 2129 | struct lc_element *tmp; |
0f0601f4 LE |
2130 | int curr_events; |
2131 | int throttle = 0; | |
2132 | ||
2133 | /* feature disabled? */ | |
f399002e | 2134 | if (mdev->ldev->dc.c_min_rate == 0) |
0f0601f4 LE |
2135 | return 0; |
2136 | ||
e3555d85 PR |
2137 | spin_lock_irq(&mdev->al_lock); |
2138 | tmp = lc_find(mdev->resync, BM_SECT_TO_EXT(sector)); | |
2139 | if (tmp) { | |
2140 | struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce); | |
2141 | if (test_bit(BME_PRIORITY, &bm_ext->flags)) { | |
2142 | spin_unlock_irq(&mdev->al_lock); | |
2143 | return 0; | |
2144 | } | |
2145 | /* Do not slow down if app IO is already waiting for this extent */ | |
2146 | } | |
2147 | spin_unlock_irq(&mdev->al_lock); | |
2148 | ||
0f0601f4 LE |
2149 | curr_events = (int)part_stat_read(&disk->part0, sectors[0]) + |
2150 | (int)part_stat_read(&disk->part0, sectors[1]) - | |
2151 | atomic_read(&mdev->rs_sect_ev); | |
e3555d85 | 2152 | |
0f0601f4 LE |
2153 | if (!mdev->rs_last_events || curr_events - mdev->rs_last_events > 64) { |
2154 | unsigned long rs_left; | |
2155 | int i; | |
2156 | ||
2157 | mdev->rs_last_events = curr_events; | |
2158 | ||
2159 | /* sync speed average over the last 2*DRBD_SYNC_MARK_STEP, | |
2160 | * approx. */ | |
2649f080 LE |
2161 | i = (mdev->rs_last_mark + DRBD_SYNC_MARKS-1) % DRBD_SYNC_MARKS; |
2162 | ||
2163 | if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T) | |
2164 | rs_left = mdev->ov_left; | |
2165 | else | |
2166 | rs_left = drbd_bm_total_weight(mdev) - mdev->rs_failed; | |
0f0601f4 LE |
2167 | |
2168 | dt = ((long)jiffies - (long)mdev->rs_mark_time[i]) / HZ; | |
2169 | if (!dt) | |
2170 | dt++; | |
2171 | db = mdev->rs_mark_left[i] - rs_left; | |
2172 | dbdt = Bit2KB(db/dt); | |
2173 | ||
f399002e | 2174 | if (dbdt > mdev->ldev->dc.c_min_rate) |
0f0601f4 LE |
2175 | throttle = 1; |
2176 | } | |
2177 | return throttle; | |
2178 | } | |
2179 | ||
2180 | ||
4a76b161 | 2181 | static int receive_DataRequest(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 2182 | { |
4a76b161 | 2183 | struct drbd_conf *mdev; |
b411b363 | 2184 | sector_t sector; |
4a76b161 | 2185 | sector_t capacity; |
db830c46 | 2186 | struct drbd_peer_request *peer_req; |
b411b363 | 2187 | struct digest_info *di = NULL; |
b18b37be | 2188 | int size, verb; |
b411b363 | 2189 | unsigned int fault_type; |
e658983a | 2190 | struct p_block_req *p = pi->data; |
4a76b161 AG |
2191 | |
2192 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
2193 | if (!mdev) | |
2194 | return -EIO; | |
2195 | capacity = drbd_get_capacity(mdev->this_bdev); | |
b411b363 PR |
2196 | |
2197 | sector = be64_to_cpu(p->sector); | |
2198 | size = be32_to_cpu(p->blksize); | |
2199 | ||
c670a398 | 2200 | if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) { |
b411b363 PR |
2201 | dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, |
2202 | (unsigned long long)sector, size); | |
82bc0194 | 2203 | return -EINVAL; |
b411b363 PR |
2204 | } |
2205 | if (sector + (size>>9) > capacity) { | |
2206 | dev_err(DEV, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, | |
2207 | (unsigned long long)sector, size); | |
82bc0194 | 2208 | return -EINVAL; |
b411b363 PR |
2209 | } |
2210 | ||
2211 | if (!get_ldev_if_state(mdev, D_UP_TO_DATE)) { | |
b18b37be | 2212 | verb = 1; |
e2857216 | 2213 | switch (pi->cmd) { |
b18b37be PR |
2214 | case P_DATA_REQUEST: |
2215 | drbd_send_ack_rp(mdev, P_NEG_DREPLY, p); | |
2216 | break; | |
2217 | case P_RS_DATA_REQUEST: | |
2218 | case P_CSUM_RS_REQUEST: | |
2219 | case P_OV_REQUEST: | |
2220 | drbd_send_ack_rp(mdev, P_NEG_RS_DREPLY , p); | |
2221 | break; | |
2222 | case P_OV_REPLY: | |
2223 | verb = 0; | |
2224 | dec_rs_pending(mdev); | |
2225 | drbd_send_ack_ex(mdev, P_OV_RESULT, sector, size, ID_IN_SYNC); | |
2226 | break; | |
2227 | default: | |
49ba9b1b | 2228 | BUG(); |
b18b37be PR |
2229 | } |
2230 | if (verb && __ratelimit(&drbd_ratelimit_state)) | |
b411b363 PR |
2231 | dev_err(DEV, "Can not satisfy peer's read request, " |
2232 | "no local data.\n"); | |
b18b37be | 2233 | |
a821cc4a | 2234 | /* drain possibly payload */ |
e2857216 | 2235 | return drbd_drain_block(mdev, pi->size); |
b411b363 PR |
2236 | } |
2237 | ||
2238 | /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD | |
2239 | * "criss-cross" setup, that might cause write-out on some other DRBD, | |
2240 | * which in turn might block on the other node at this very place. */ | |
0db55363 | 2241 | peer_req = drbd_alloc_peer_req(mdev, p->block_id, sector, size, GFP_NOIO); |
db830c46 | 2242 | if (!peer_req) { |
b411b363 | 2243 | put_ldev(mdev); |
82bc0194 | 2244 | return -ENOMEM; |
b411b363 PR |
2245 | } |
2246 | ||
e2857216 | 2247 | switch (pi->cmd) { |
b411b363 | 2248 | case P_DATA_REQUEST: |
db830c46 | 2249 | peer_req->w.cb = w_e_end_data_req; |
b411b363 | 2250 | fault_type = DRBD_FAULT_DT_RD; |
80a40e43 LE |
2251 | /* application IO, don't drbd_rs_begin_io */ |
2252 | goto submit; | |
2253 | ||
b411b363 | 2254 | case P_RS_DATA_REQUEST: |
db830c46 | 2255 | peer_req->w.cb = w_e_end_rsdata_req; |
b411b363 | 2256 | fault_type = DRBD_FAULT_RS_RD; |
5f9915bb LE |
2257 | /* used in the sector offset progress display */ |
2258 | mdev->bm_resync_fo = BM_SECT_TO_BIT(sector); | |
b411b363 PR |
2259 | break; |
2260 | ||
2261 | case P_OV_REPLY: | |
2262 | case P_CSUM_RS_REQUEST: | |
2263 | fault_type = DRBD_FAULT_RS_RD; | |
e2857216 | 2264 | di = kmalloc(sizeof(*di) + pi->size, GFP_NOIO); |
b411b363 PR |
2265 | if (!di) |
2266 | goto out_free_e; | |
2267 | ||
e2857216 | 2268 | di->digest_size = pi->size; |
b411b363 PR |
2269 | di->digest = (((char *)di)+sizeof(struct digest_info)); |
2270 | ||
db830c46 AG |
2271 | peer_req->digest = di; |
2272 | peer_req->flags |= EE_HAS_DIGEST; | |
c36c3ced | 2273 | |
e2857216 | 2274 | if (drbd_recv_all(mdev->tconn, di->digest, pi->size)) |
b411b363 PR |
2275 | goto out_free_e; |
2276 | ||
e2857216 | 2277 | if (pi->cmd == P_CSUM_RS_REQUEST) { |
31890f4a | 2278 | D_ASSERT(mdev->tconn->agreed_pro_version >= 89); |
db830c46 | 2279 | peer_req->w.cb = w_e_end_csum_rs_req; |
5f9915bb LE |
2280 | /* used in the sector offset progress display */ |
2281 | mdev->bm_resync_fo = BM_SECT_TO_BIT(sector); | |
e2857216 | 2282 | } else if (pi->cmd == P_OV_REPLY) { |
2649f080 LE |
2283 | /* track progress, we may need to throttle */ |
2284 | atomic_add(size >> 9, &mdev->rs_sect_in); | |
db830c46 | 2285 | peer_req->w.cb = w_e_end_ov_reply; |
b411b363 | 2286 | dec_rs_pending(mdev); |
0f0601f4 LE |
2287 | /* drbd_rs_begin_io done when we sent this request, |
2288 | * but accounting still needs to be done. */ | |
2289 | goto submit_for_resync; | |
b411b363 PR |
2290 | } |
2291 | break; | |
2292 | ||
2293 | case P_OV_REQUEST: | |
b411b363 | 2294 | if (mdev->ov_start_sector == ~(sector_t)0 && |
31890f4a | 2295 | mdev->tconn->agreed_pro_version >= 90) { |
de228bba LE |
2296 | unsigned long now = jiffies; |
2297 | int i; | |
b411b363 PR |
2298 | mdev->ov_start_sector = sector; |
2299 | mdev->ov_position = sector; | |
30b743a2 LE |
2300 | mdev->ov_left = drbd_bm_bits(mdev) - BM_SECT_TO_BIT(sector); |
2301 | mdev->rs_total = mdev->ov_left; | |
de228bba LE |
2302 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { |
2303 | mdev->rs_mark_left[i] = mdev->ov_left; | |
2304 | mdev->rs_mark_time[i] = now; | |
2305 | } | |
b411b363 PR |
2306 | dev_info(DEV, "Online Verify start sector: %llu\n", |
2307 | (unsigned long long)sector); | |
2308 | } | |
db830c46 | 2309 | peer_req->w.cb = w_e_end_ov_req; |
b411b363 | 2310 | fault_type = DRBD_FAULT_RS_RD; |
b411b363 PR |
2311 | break; |
2312 | ||
b411b363 | 2313 | default: |
49ba9b1b | 2314 | BUG(); |
b411b363 PR |
2315 | } |
2316 | ||
0f0601f4 LE |
2317 | /* Throttle, drbd_rs_begin_io and submit should become asynchronous |
2318 | * wrt the receiver, but it is not as straightforward as it may seem. | |
2319 | * Various places in the resync start and stop logic assume resync | |
2320 | * requests are processed in order, requeuing this on the worker thread | |
2321 | * introduces a bunch of new code for synchronization between threads. | |
2322 | * | |
2323 | * Unlimited throttling before drbd_rs_begin_io may stall the resync | |
2324 | * "forever", throttling after drbd_rs_begin_io will lock that extent | |
2325 | * for application writes for the same time. For now, just throttle | |
2326 | * here, where the rest of the code expects the receiver to sleep for | |
2327 | * a while, anyways. | |
2328 | */ | |
2329 | ||
2330 | /* Throttle before drbd_rs_begin_io, as that locks out application IO; | |
2331 | * this defers syncer requests for some time, before letting at least | |
2332 | * on request through. The resync controller on the receiving side | |
2333 | * will adapt to the incoming rate accordingly. | |
2334 | * | |
2335 | * We cannot throttle here if remote is Primary/SyncTarget: | |
2336 | * we would also throttle its application reads. | |
2337 | * In that case, throttling is done on the SyncTarget only. | |
2338 | */ | |
e3555d85 PR |
2339 | if (mdev->state.peer != R_PRIMARY && drbd_rs_should_slow_down(mdev, sector)) |
2340 | schedule_timeout_uninterruptible(HZ/10); | |
2341 | if (drbd_rs_begin_io(mdev, sector)) | |
80a40e43 | 2342 | goto out_free_e; |
b411b363 | 2343 | |
0f0601f4 LE |
2344 | submit_for_resync: |
2345 | atomic_add(size >> 9, &mdev->rs_sect_ev); | |
2346 | ||
80a40e43 | 2347 | submit: |
b411b363 | 2348 | inc_unacked(mdev); |
87eeee41 | 2349 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 2350 | list_add_tail(&peer_req->w.list, &mdev->read_ee); |
87eeee41 | 2351 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 2352 | |
fbe29dec | 2353 | if (drbd_submit_peer_request(mdev, peer_req, READ, fault_type) == 0) |
82bc0194 | 2354 | return 0; |
b411b363 | 2355 | |
10f6d992 LE |
2356 | /* don't care for the reason here */ |
2357 | dev_err(DEV, "submit failed, triggering re-connect\n"); | |
87eeee41 | 2358 | spin_lock_irq(&mdev->tconn->req_lock); |
db830c46 | 2359 | list_del(&peer_req->w.list); |
87eeee41 | 2360 | spin_unlock_irq(&mdev->tconn->req_lock); |
22cc37a9 LE |
2361 | /* no drbd_rs_complete_io(), we are dropping the connection anyways */ |
2362 | ||
b411b363 | 2363 | out_free_e: |
b411b363 | 2364 | put_ldev(mdev); |
3967deb1 | 2365 | drbd_free_peer_req(mdev, peer_req); |
82bc0194 | 2366 | return -EIO; |
b411b363 PR |
2367 | } |
2368 | ||
2369 | static int drbd_asb_recover_0p(struct drbd_conf *mdev) __must_hold(local) | |
2370 | { | |
2371 | int self, peer, rv = -100; | |
2372 | unsigned long ch_self, ch_peer; | |
2373 | ||
2374 | self = mdev->ldev->md.uuid[UI_BITMAP] & 1; | |
2375 | peer = mdev->p_uuid[UI_BITMAP] & 1; | |
2376 | ||
2377 | ch_peer = mdev->p_uuid[UI_SIZE]; | |
2378 | ch_self = mdev->comm_bm_set; | |
2379 | ||
89e58e75 | 2380 | switch (mdev->tconn->net_conf->after_sb_0p) { |
b411b363 PR |
2381 | case ASB_CONSENSUS: |
2382 | case ASB_DISCARD_SECONDARY: | |
2383 | case ASB_CALL_HELPER: | |
2384 | dev_err(DEV, "Configuration error.\n"); | |
2385 | break; | |
2386 | case ASB_DISCONNECT: | |
2387 | break; | |
2388 | case ASB_DISCARD_YOUNGER_PRI: | |
2389 | if (self == 0 && peer == 1) { | |
2390 | rv = -1; | |
2391 | break; | |
2392 | } | |
2393 | if (self == 1 && peer == 0) { | |
2394 | rv = 1; | |
2395 | break; | |
2396 | } | |
2397 | /* Else fall through to one of the other strategies... */ | |
2398 | case ASB_DISCARD_OLDER_PRI: | |
2399 | if (self == 0 && peer == 1) { | |
2400 | rv = 1; | |
2401 | break; | |
2402 | } | |
2403 | if (self == 1 && peer == 0) { | |
2404 | rv = -1; | |
2405 | break; | |
2406 | } | |
2407 | /* Else fall through to one of the other strategies... */ | |
ad19bf6e | 2408 | dev_warn(DEV, "Discard younger/older primary did not find a decision\n" |
b411b363 PR |
2409 | "Using discard-least-changes instead\n"); |
2410 | case ASB_DISCARD_ZERO_CHG: | |
2411 | if (ch_peer == 0 && ch_self == 0) { | |
25703f83 | 2412 | rv = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) |
b411b363 PR |
2413 | ? -1 : 1; |
2414 | break; | |
2415 | } else { | |
2416 | if (ch_peer == 0) { rv = 1; break; } | |
2417 | if (ch_self == 0) { rv = -1; break; } | |
2418 | } | |
89e58e75 | 2419 | if (mdev->tconn->net_conf->after_sb_0p == ASB_DISCARD_ZERO_CHG) |
b411b363 PR |
2420 | break; |
2421 | case ASB_DISCARD_LEAST_CHG: | |
2422 | if (ch_self < ch_peer) | |
2423 | rv = -1; | |
2424 | else if (ch_self > ch_peer) | |
2425 | rv = 1; | |
2426 | else /* ( ch_self == ch_peer ) */ | |
2427 | /* Well, then use something else. */ | |
25703f83 | 2428 | rv = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) |
b411b363 PR |
2429 | ? -1 : 1; |
2430 | break; | |
2431 | case ASB_DISCARD_LOCAL: | |
2432 | rv = -1; | |
2433 | break; | |
2434 | case ASB_DISCARD_REMOTE: | |
2435 | rv = 1; | |
2436 | } | |
2437 | ||
2438 | return rv; | |
2439 | } | |
2440 | ||
2441 | static int drbd_asb_recover_1p(struct drbd_conf *mdev) __must_hold(local) | |
2442 | { | |
6184ea21 | 2443 | int hg, rv = -100; |
b411b363 | 2444 | |
89e58e75 | 2445 | switch (mdev->tconn->net_conf->after_sb_1p) { |
b411b363 PR |
2446 | case ASB_DISCARD_YOUNGER_PRI: |
2447 | case ASB_DISCARD_OLDER_PRI: | |
2448 | case ASB_DISCARD_LEAST_CHG: | |
2449 | case ASB_DISCARD_LOCAL: | |
2450 | case ASB_DISCARD_REMOTE: | |
2451 | dev_err(DEV, "Configuration error.\n"); | |
2452 | break; | |
2453 | case ASB_DISCONNECT: | |
2454 | break; | |
2455 | case ASB_CONSENSUS: | |
2456 | hg = drbd_asb_recover_0p(mdev); | |
2457 | if (hg == -1 && mdev->state.role == R_SECONDARY) | |
2458 | rv = hg; | |
2459 | if (hg == 1 && mdev->state.role == R_PRIMARY) | |
2460 | rv = hg; | |
2461 | break; | |
2462 | case ASB_VIOLENTLY: | |
2463 | rv = drbd_asb_recover_0p(mdev); | |
2464 | break; | |
2465 | case ASB_DISCARD_SECONDARY: | |
2466 | return mdev->state.role == R_PRIMARY ? 1 : -1; | |
2467 | case ASB_CALL_HELPER: | |
2468 | hg = drbd_asb_recover_0p(mdev); | |
2469 | if (hg == -1 && mdev->state.role == R_PRIMARY) { | |
bb437946 AG |
2470 | enum drbd_state_rv rv2; |
2471 | ||
2472 | drbd_set_role(mdev, R_SECONDARY, 0); | |
b411b363 PR |
2473 | /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE, |
2474 | * we might be here in C_WF_REPORT_PARAMS which is transient. | |
2475 | * we do not need to wait for the after state change work either. */ | |
bb437946 AG |
2476 | rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY)); |
2477 | if (rv2 != SS_SUCCESS) { | |
b411b363 PR |
2478 | drbd_khelper(mdev, "pri-lost-after-sb"); |
2479 | } else { | |
2480 | dev_warn(DEV, "Successfully gave up primary role.\n"); | |
2481 | rv = hg; | |
2482 | } | |
2483 | } else | |
2484 | rv = hg; | |
2485 | } | |
2486 | ||
2487 | return rv; | |
2488 | } | |
2489 | ||
2490 | static int drbd_asb_recover_2p(struct drbd_conf *mdev) __must_hold(local) | |
2491 | { | |
6184ea21 | 2492 | int hg, rv = -100; |
b411b363 | 2493 | |
89e58e75 | 2494 | switch (mdev->tconn->net_conf->after_sb_2p) { |
b411b363 PR |
2495 | case ASB_DISCARD_YOUNGER_PRI: |
2496 | case ASB_DISCARD_OLDER_PRI: | |
2497 | case ASB_DISCARD_LEAST_CHG: | |
2498 | case ASB_DISCARD_LOCAL: | |
2499 | case ASB_DISCARD_REMOTE: | |
2500 | case ASB_CONSENSUS: | |
2501 | case ASB_DISCARD_SECONDARY: | |
2502 | dev_err(DEV, "Configuration error.\n"); | |
2503 | break; | |
2504 | case ASB_VIOLENTLY: | |
2505 | rv = drbd_asb_recover_0p(mdev); | |
2506 | break; | |
2507 | case ASB_DISCONNECT: | |
2508 | break; | |
2509 | case ASB_CALL_HELPER: | |
2510 | hg = drbd_asb_recover_0p(mdev); | |
2511 | if (hg == -1) { | |
bb437946 AG |
2512 | enum drbd_state_rv rv2; |
2513 | ||
b411b363 PR |
2514 | /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE, |
2515 | * we might be here in C_WF_REPORT_PARAMS which is transient. | |
2516 | * we do not need to wait for the after state change work either. */ | |
bb437946 AG |
2517 | rv2 = drbd_change_state(mdev, CS_VERBOSE, NS(role, R_SECONDARY)); |
2518 | if (rv2 != SS_SUCCESS) { | |
b411b363 PR |
2519 | drbd_khelper(mdev, "pri-lost-after-sb"); |
2520 | } else { | |
2521 | dev_warn(DEV, "Successfully gave up primary role.\n"); | |
2522 | rv = hg; | |
2523 | } | |
2524 | } else | |
2525 | rv = hg; | |
2526 | } | |
2527 | ||
2528 | return rv; | |
2529 | } | |
2530 | ||
2531 | static void drbd_uuid_dump(struct drbd_conf *mdev, char *text, u64 *uuid, | |
2532 | u64 bits, u64 flags) | |
2533 | { | |
2534 | if (!uuid) { | |
2535 | dev_info(DEV, "%s uuid info vanished while I was looking!\n", text); | |
2536 | return; | |
2537 | } | |
2538 | dev_info(DEV, "%s %016llX:%016llX:%016llX:%016llX bits:%llu flags:%llX\n", | |
2539 | text, | |
2540 | (unsigned long long)uuid[UI_CURRENT], | |
2541 | (unsigned long long)uuid[UI_BITMAP], | |
2542 | (unsigned long long)uuid[UI_HISTORY_START], | |
2543 | (unsigned long long)uuid[UI_HISTORY_END], | |
2544 | (unsigned long long)bits, | |
2545 | (unsigned long long)flags); | |
2546 | } | |
2547 | ||
2548 | /* | |
2549 | 100 after split brain try auto recover | |
2550 | 2 C_SYNC_SOURCE set BitMap | |
2551 | 1 C_SYNC_SOURCE use BitMap | |
2552 | 0 no Sync | |
2553 | -1 C_SYNC_TARGET use BitMap | |
2554 | -2 C_SYNC_TARGET set BitMap | |
2555 | -100 after split brain, disconnect | |
2556 | -1000 unrelated data | |
4a23f264 PR |
2557 | -1091 requires proto 91 |
2558 | -1096 requires proto 96 | |
b411b363 PR |
2559 | */ |
2560 | static int drbd_uuid_compare(struct drbd_conf *mdev, int *rule_nr) __must_hold(local) | |
2561 | { | |
2562 | u64 self, peer; | |
2563 | int i, j; | |
2564 | ||
2565 | self = mdev->ldev->md.uuid[UI_CURRENT] & ~((u64)1); | |
2566 | peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1); | |
2567 | ||
2568 | *rule_nr = 10; | |
2569 | if (self == UUID_JUST_CREATED && peer == UUID_JUST_CREATED) | |
2570 | return 0; | |
2571 | ||
2572 | *rule_nr = 20; | |
2573 | if ((self == UUID_JUST_CREATED || self == (u64)0) && | |
2574 | peer != UUID_JUST_CREATED) | |
2575 | return -2; | |
2576 | ||
2577 | *rule_nr = 30; | |
2578 | if (self != UUID_JUST_CREATED && | |
2579 | (peer == UUID_JUST_CREATED || peer == (u64)0)) | |
2580 | return 2; | |
2581 | ||
2582 | if (self == peer) { | |
2583 | int rct, dc; /* roles at crash time */ | |
2584 | ||
2585 | if (mdev->p_uuid[UI_BITMAP] == (u64)0 && mdev->ldev->md.uuid[UI_BITMAP] != (u64)0) { | |
2586 | ||
31890f4a | 2587 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2588 | return -1091; |
b411b363 PR |
2589 | |
2590 | if ((mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) && | |
2591 | (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1))) { | |
2592 | dev_info(DEV, "was SyncSource, missed the resync finished event, corrected myself:\n"); | |
2593 | drbd_uuid_set_bm(mdev, 0UL); | |
2594 | ||
2595 | drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, | |
2596 | mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0); | |
2597 | *rule_nr = 34; | |
2598 | } else { | |
2599 | dev_info(DEV, "was SyncSource (peer failed to write sync_uuid)\n"); | |
2600 | *rule_nr = 36; | |
2601 | } | |
2602 | ||
2603 | return 1; | |
2604 | } | |
2605 | ||
2606 | if (mdev->ldev->md.uuid[UI_BITMAP] == (u64)0 && mdev->p_uuid[UI_BITMAP] != (u64)0) { | |
2607 | ||
31890f4a | 2608 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2609 | return -1091; |
b411b363 PR |
2610 | |
2611 | if ((mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (mdev->p_uuid[UI_BITMAP] & ~((u64)1)) && | |
2612 | (mdev->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1))) { | |
2613 | dev_info(DEV, "was SyncTarget, peer missed the resync finished event, corrected peer:\n"); | |
2614 | ||
2615 | mdev->p_uuid[UI_HISTORY_START + 1] = mdev->p_uuid[UI_HISTORY_START]; | |
2616 | mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_BITMAP]; | |
2617 | mdev->p_uuid[UI_BITMAP] = 0UL; | |
2618 | ||
2619 | drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]); | |
2620 | *rule_nr = 35; | |
2621 | } else { | |
2622 | dev_info(DEV, "was SyncTarget (failed to write sync_uuid)\n"); | |
2623 | *rule_nr = 37; | |
2624 | } | |
2625 | ||
2626 | return -1; | |
2627 | } | |
2628 | ||
2629 | /* Common power [off|failure] */ | |
2630 | rct = (test_bit(CRASHED_PRIMARY, &mdev->flags) ? 1 : 0) + | |
2631 | (mdev->p_uuid[UI_FLAGS] & 2); | |
2632 | /* lowest bit is set when we were primary, | |
2633 | * next bit (weight 2) is set when peer was primary */ | |
2634 | *rule_nr = 40; | |
2635 | ||
2636 | switch (rct) { | |
2637 | case 0: /* !self_pri && !peer_pri */ return 0; | |
2638 | case 1: /* self_pri && !peer_pri */ return 1; | |
2639 | case 2: /* !self_pri && peer_pri */ return -1; | |
2640 | case 3: /* self_pri && peer_pri */ | |
25703f83 | 2641 | dc = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags); |
b411b363 PR |
2642 | return dc ? -1 : 1; |
2643 | } | |
2644 | } | |
2645 | ||
2646 | *rule_nr = 50; | |
2647 | peer = mdev->p_uuid[UI_BITMAP] & ~((u64)1); | |
2648 | if (self == peer) | |
2649 | return -1; | |
2650 | ||
2651 | *rule_nr = 51; | |
2652 | peer = mdev->p_uuid[UI_HISTORY_START] & ~((u64)1); | |
2653 | if (self == peer) { | |
31890f4a | 2654 | if (mdev->tconn->agreed_pro_version < 96 ? |
4a23f264 PR |
2655 | (mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == |
2656 | (mdev->p_uuid[UI_HISTORY_START + 1] & ~((u64)1)) : | |
2657 | peer + UUID_NEW_BM_OFFSET == (mdev->p_uuid[UI_BITMAP] & ~((u64)1))) { | |
b411b363 PR |
2658 | /* The last P_SYNC_UUID did not get though. Undo the last start of |
2659 | resync as sync source modifications of the peer's UUIDs. */ | |
2660 | ||
31890f4a | 2661 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2662 | return -1091; |
b411b363 PR |
2663 | |
2664 | mdev->p_uuid[UI_BITMAP] = mdev->p_uuid[UI_HISTORY_START]; | |
2665 | mdev->p_uuid[UI_HISTORY_START] = mdev->p_uuid[UI_HISTORY_START + 1]; | |
4a23f264 PR |
2666 | |
2667 | dev_info(DEV, "Did not got last syncUUID packet, corrected:\n"); | |
2668 | drbd_uuid_dump(mdev, "peer", mdev->p_uuid, mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]); | |
2669 | ||
b411b363 PR |
2670 | return -1; |
2671 | } | |
2672 | } | |
2673 | ||
2674 | *rule_nr = 60; | |
2675 | self = mdev->ldev->md.uuid[UI_CURRENT] & ~((u64)1); | |
2676 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { | |
2677 | peer = mdev->p_uuid[i] & ~((u64)1); | |
2678 | if (self == peer) | |
2679 | return -2; | |
2680 | } | |
2681 | ||
2682 | *rule_nr = 70; | |
2683 | self = mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1); | |
2684 | peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1); | |
2685 | if (self == peer) | |
2686 | return 1; | |
2687 | ||
2688 | *rule_nr = 71; | |
2689 | self = mdev->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1); | |
2690 | if (self == peer) { | |
31890f4a | 2691 | if (mdev->tconn->agreed_pro_version < 96 ? |
4a23f264 PR |
2692 | (mdev->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == |
2693 | (mdev->p_uuid[UI_HISTORY_START] & ~((u64)1)) : | |
2694 | self + UUID_NEW_BM_OFFSET == (mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1))) { | |
b411b363 PR |
2695 | /* The last P_SYNC_UUID did not get though. Undo the last start of |
2696 | resync as sync source modifications of our UUIDs. */ | |
2697 | ||
31890f4a | 2698 | if (mdev->tconn->agreed_pro_version < 91) |
4a23f264 | 2699 | return -1091; |
b411b363 PR |
2700 | |
2701 | _drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_HISTORY_START]); | |
2702 | _drbd_uuid_set(mdev, UI_HISTORY_START, mdev->ldev->md.uuid[UI_HISTORY_START + 1]); | |
2703 | ||
4a23f264 | 2704 | dev_info(DEV, "Last syncUUID did not get through, corrected:\n"); |
b411b363 PR |
2705 | drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, |
2706 | mdev->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(mdev) : 0, 0); | |
2707 | ||
2708 | return 1; | |
2709 | } | |
2710 | } | |
2711 | ||
2712 | ||
2713 | *rule_nr = 80; | |
d8c2a36b | 2714 | peer = mdev->p_uuid[UI_CURRENT] & ~((u64)1); |
b411b363 PR |
2715 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { |
2716 | self = mdev->ldev->md.uuid[i] & ~((u64)1); | |
2717 | if (self == peer) | |
2718 | return 2; | |
2719 | } | |
2720 | ||
2721 | *rule_nr = 90; | |
2722 | self = mdev->ldev->md.uuid[UI_BITMAP] & ~((u64)1); | |
2723 | peer = mdev->p_uuid[UI_BITMAP] & ~((u64)1); | |
2724 | if (self == peer && self != ((u64)0)) | |
2725 | return 100; | |
2726 | ||
2727 | *rule_nr = 100; | |
2728 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { | |
2729 | self = mdev->ldev->md.uuid[i] & ~((u64)1); | |
2730 | for (j = UI_HISTORY_START; j <= UI_HISTORY_END; j++) { | |
2731 | peer = mdev->p_uuid[j] & ~((u64)1); | |
2732 | if (self == peer) | |
2733 | return -100; | |
2734 | } | |
2735 | } | |
2736 | ||
2737 | return -1000; | |
2738 | } | |
2739 | ||
2740 | /* drbd_sync_handshake() returns the new conn state on success, or | |
2741 | CONN_MASK (-1) on failure. | |
2742 | */ | |
2743 | static enum drbd_conns drbd_sync_handshake(struct drbd_conf *mdev, enum drbd_role peer_role, | |
2744 | enum drbd_disk_state peer_disk) __must_hold(local) | |
2745 | { | |
2746 | int hg, rule_nr; | |
2747 | enum drbd_conns rv = C_MASK; | |
2748 | enum drbd_disk_state mydisk; | |
2749 | ||
2750 | mydisk = mdev->state.disk; | |
2751 | if (mydisk == D_NEGOTIATING) | |
2752 | mydisk = mdev->new_state_tmp.disk; | |
2753 | ||
2754 | dev_info(DEV, "drbd_sync_handshake:\n"); | |
2755 | drbd_uuid_dump(mdev, "self", mdev->ldev->md.uuid, mdev->comm_bm_set, 0); | |
2756 | drbd_uuid_dump(mdev, "peer", mdev->p_uuid, | |
2757 | mdev->p_uuid[UI_SIZE], mdev->p_uuid[UI_FLAGS]); | |
2758 | ||
2759 | hg = drbd_uuid_compare(mdev, &rule_nr); | |
2760 | ||
2761 | dev_info(DEV, "uuid_compare()=%d by rule %d\n", hg, rule_nr); | |
2762 | ||
2763 | if (hg == -1000) { | |
2764 | dev_alert(DEV, "Unrelated data, aborting!\n"); | |
2765 | return C_MASK; | |
2766 | } | |
4a23f264 PR |
2767 | if (hg < -1000) { |
2768 | dev_alert(DEV, "To resolve this both sides have to support at least protocol %d\n", -hg - 1000); | |
b411b363 PR |
2769 | return C_MASK; |
2770 | } | |
2771 | ||
2772 | if ((mydisk == D_INCONSISTENT && peer_disk > D_INCONSISTENT) || | |
2773 | (peer_disk == D_INCONSISTENT && mydisk > D_INCONSISTENT)) { | |
2774 | int f = (hg == -100) || abs(hg) == 2; | |
2775 | hg = mydisk > D_INCONSISTENT ? 1 : -1; | |
2776 | if (f) | |
2777 | hg = hg*2; | |
2778 | dev_info(DEV, "Becoming sync %s due to disk states.\n", | |
2779 | hg > 0 ? "source" : "target"); | |
2780 | } | |
2781 | ||
3a11a487 AG |
2782 | if (abs(hg) == 100) |
2783 | drbd_khelper(mdev, "initial-split-brain"); | |
2784 | ||
89e58e75 | 2785 | if (hg == 100 || (hg == -100 && mdev->tconn->net_conf->always_asbp)) { |
b411b363 PR |
2786 | int pcount = (mdev->state.role == R_PRIMARY) |
2787 | + (peer_role == R_PRIMARY); | |
2788 | int forced = (hg == -100); | |
2789 | ||
2790 | switch (pcount) { | |
2791 | case 0: | |
2792 | hg = drbd_asb_recover_0p(mdev); | |
2793 | break; | |
2794 | case 1: | |
2795 | hg = drbd_asb_recover_1p(mdev); | |
2796 | break; | |
2797 | case 2: | |
2798 | hg = drbd_asb_recover_2p(mdev); | |
2799 | break; | |
2800 | } | |
2801 | if (abs(hg) < 100) { | |
2802 | dev_warn(DEV, "Split-Brain detected, %d primaries, " | |
2803 | "automatically solved. Sync from %s node\n", | |
2804 | pcount, (hg < 0) ? "peer" : "this"); | |
2805 | if (forced) { | |
2806 | dev_warn(DEV, "Doing a full sync, since" | |
2807 | " UUIDs where ambiguous.\n"); | |
2808 | hg = hg*2; | |
2809 | } | |
2810 | } | |
2811 | } | |
2812 | ||
2813 | if (hg == -100) { | |
89e58e75 | 2814 | if (mdev->tconn->net_conf->want_lose && !(mdev->p_uuid[UI_FLAGS]&1)) |
b411b363 | 2815 | hg = -1; |
89e58e75 | 2816 | if (!mdev->tconn->net_conf->want_lose && (mdev->p_uuid[UI_FLAGS]&1)) |
b411b363 PR |
2817 | hg = 1; |
2818 | ||
2819 | if (abs(hg) < 100) | |
2820 | dev_warn(DEV, "Split-Brain detected, manually solved. " | |
2821 | "Sync from %s node\n", | |
2822 | (hg < 0) ? "peer" : "this"); | |
2823 | } | |
2824 | ||
2825 | if (hg == -100) { | |
580b9767 LE |
2826 | /* FIXME this log message is not correct if we end up here |
2827 | * after an attempted attach on a diskless node. | |
2828 | * We just refuse to attach -- well, we drop the "connection" | |
2829 | * to that disk, in a way... */ | |
3a11a487 | 2830 | dev_alert(DEV, "Split-Brain detected but unresolved, dropping connection!\n"); |
b411b363 PR |
2831 | drbd_khelper(mdev, "split-brain"); |
2832 | return C_MASK; | |
2833 | } | |
2834 | ||
2835 | if (hg > 0 && mydisk <= D_INCONSISTENT) { | |
2836 | dev_err(DEV, "I shall become SyncSource, but I am inconsistent!\n"); | |
2837 | return C_MASK; | |
2838 | } | |
2839 | ||
2840 | if (hg < 0 && /* by intention we do not use mydisk here. */ | |
2841 | mdev->state.role == R_PRIMARY && mdev->state.disk >= D_CONSISTENT) { | |
89e58e75 | 2842 | switch (mdev->tconn->net_conf->rr_conflict) { |
b411b363 PR |
2843 | case ASB_CALL_HELPER: |
2844 | drbd_khelper(mdev, "pri-lost"); | |
2845 | /* fall through */ | |
2846 | case ASB_DISCONNECT: | |
2847 | dev_err(DEV, "I shall become SyncTarget, but I am primary!\n"); | |
2848 | return C_MASK; | |
2849 | case ASB_VIOLENTLY: | |
2850 | dev_warn(DEV, "Becoming SyncTarget, violating the stable-data" | |
2851 | "assumption\n"); | |
2852 | } | |
2853 | } | |
2854 | ||
8169e41b | 2855 | if (mdev->tconn->net_conf->dry_run || test_bit(CONN_DRY_RUN, &mdev->tconn->flags)) { |
cf14c2e9 PR |
2856 | if (hg == 0) |
2857 | dev_info(DEV, "dry-run connect: No resync, would become Connected immediately.\n"); | |
2858 | else | |
2859 | dev_info(DEV, "dry-run connect: Would become %s, doing a %s resync.", | |
2860 | drbd_conn_str(hg > 0 ? C_SYNC_SOURCE : C_SYNC_TARGET), | |
2861 | abs(hg) >= 2 ? "full" : "bit-map based"); | |
2862 | return C_MASK; | |
2863 | } | |
2864 | ||
b411b363 PR |
2865 | if (abs(hg) >= 2) { |
2866 | dev_info(DEV, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n"); | |
20ceb2b2 LE |
2867 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from sync_handshake", |
2868 | BM_LOCKED_SET_ALLOWED)) | |
b411b363 PR |
2869 | return C_MASK; |
2870 | } | |
2871 | ||
2872 | if (hg > 0) { /* become sync source. */ | |
2873 | rv = C_WF_BITMAP_S; | |
2874 | } else if (hg < 0) { /* become sync target */ | |
2875 | rv = C_WF_BITMAP_T; | |
2876 | } else { | |
2877 | rv = C_CONNECTED; | |
2878 | if (drbd_bm_total_weight(mdev)) { | |
2879 | dev_info(DEV, "No resync, but %lu bits in bitmap!\n", | |
2880 | drbd_bm_total_weight(mdev)); | |
2881 | } | |
2882 | } | |
2883 | ||
2884 | return rv; | |
2885 | } | |
2886 | ||
2887 | /* returns 1 if invalid */ | |
2888 | static int cmp_after_sb(enum drbd_after_sb_p peer, enum drbd_after_sb_p self) | |
2889 | { | |
2890 | /* ASB_DISCARD_REMOTE - ASB_DISCARD_LOCAL is valid */ | |
2891 | if ((peer == ASB_DISCARD_REMOTE && self == ASB_DISCARD_LOCAL) || | |
2892 | (self == ASB_DISCARD_REMOTE && peer == ASB_DISCARD_LOCAL)) | |
2893 | return 0; | |
2894 | ||
2895 | /* any other things with ASB_DISCARD_REMOTE or ASB_DISCARD_LOCAL are invalid */ | |
2896 | if (peer == ASB_DISCARD_REMOTE || peer == ASB_DISCARD_LOCAL || | |
2897 | self == ASB_DISCARD_REMOTE || self == ASB_DISCARD_LOCAL) | |
2898 | return 1; | |
2899 | ||
2900 | /* everything else is valid if they are equal on both sides. */ | |
2901 | if (peer == self) | |
2902 | return 0; | |
2903 | ||
2904 | /* everything es is invalid. */ | |
2905 | return 1; | |
2906 | } | |
2907 | ||
e2857216 | 2908 | static int receive_protocol(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 2909 | { |
e658983a | 2910 | struct p_protocol *p = pi->data; |
b411b363 | 2911 | int p_proto, p_after_sb_0p, p_after_sb_1p, p_after_sb_2p; |
cf14c2e9 | 2912 | int p_want_lose, p_two_primaries, cf; |
b411b363 PR |
2913 | char p_integrity_alg[SHARED_SECRET_MAX] = ""; |
2914 | ||
b411b363 PR |
2915 | p_proto = be32_to_cpu(p->protocol); |
2916 | p_after_sb_0p = be32_to_cpu(p->after_sb_0p); | |
2917 | p_after_sb_1p = be32_to_cpu(p->after_sb_1p); | |
2918 | p_after_sb_2p = be32_to_cpu(p->after_sb_2p); | |
b411b363 | 2919 | p_two_primaries = be32_to_cpu(p->two_primaries); |
cf14c2e9 PR |
2920 | cf = be32_to_cpu(p->conn_flags); |
2921 | p_want_lose = cf & CF_WANT_LOSE; | |
2922 | ||
7204624c | 2923 | clear_bit(CONN_DRY_RUN, &tconn->flags); |
cf14c2e9 PR |
2924 | |
2925 | if (cf & CF_DRY_RUN) | |
7204624c | 2926 | set_bit(CONN_DRY_RUN, &tconn->flags); |
b411b363 | 2927 | |
7204624c PR |
2928 | if (p_proto != tconn->net_conf->wire_protocol) { |
2929 | conn_err(tconn, "incompatible communication protocols\n"); | |
b411b363 PR |
2930 | goto disconnect; |
2931 | } | |
2932 | ||
7204624c PR |
2933 | if (cmp_after_sb(p_after_sb_0p, tconn->net_conf->after_sb_0p)) { |
2934 | conn_err(tconn, "incompatible after-sb-0pri settings\n"); | |
b411b363 PR |
2935 | goto disconnect; |
2936 | } | |
2937 | ||
7204624c PR |
2938 | if (cmp_after_sb(p_after_sb_1p, tconn->net_conf->after_sb_1p)) { |
2939 | conn_err(tconn, "incompatible after-sb-1pri settings\n"); | |
b411b363 PR |
2940 | goto disconnect; |
2941 | } | |
2942 | ||
7204624c PR |
2943 | if (cmp_after_sb(p_after_sb_2p, tconn->net_conf->after_sb_2p)) { |
2944 | conn_err(tconn, "incompatible after-sb-2pri settings\n"); | |
b411b363 PR |
2945 | goto disconnect; |
2946 | } | |
2947 | ||
7204624c PR |
2948 | if (p_want_lose && tconn->net_conf->want_lose) { |
2949 | conn_err(tconn, "both sides have the 'want_lose' flag set\n"); | |
b411b363 PR |
2950 | goto disconnect; |
2951 | } | |
2952 | ||
7204624c PR |
2953 | if (p_two_primaries != tconn->net_conf->two_primaries) { |
2954 | conn_err(tconn, "incompatible setting of the two-primaries options\n"); | |
b411b363 PR |
2955 | goto disconnect; |
2956 | } | |
2957 | ||
7204624c PR |
2958 | if (tconn->agreed_pro_version >= 87) { |
2959 | unsigned char *my_alg = tconn->net_conf->integrity_alg; | |
82bc0194 | 2960 | int err; |
b411b363 | 2961 | |
e2857216 | 2962 | err = drbd_recv_all(tconn, p_integrity_alg, pi->size); |
82bc0194 AG |
2963 | if (err) |
2964 | return err; | |
b411b363 PR |
2965 | |
2966 | p_integrity_alg[SHARED_SECRET_MAX-1] = 0; | |
2967 | if (strcmp(p_integrity_alg, my_alg)) { | |
7204624c | 2968 | conn_err(tconn, "incompatible setting of the data-integrity-alg\n"); |
b411b363 PR |
2969 | goto disconnect; |
2970 | } | |
7204624c | 2971 | conn_info(tconn, "data-integrity-alg: %s\n", |
b411b363 PR |
2972 | my_alg[0] ? my_alg : (unsigned char *)"<not-used>"); |
2973 | } | |
2974 | ||
82bc0194 | 2975 | return 0; |
b411b363 PR |
2976 | |
2977 | disconnect: | |
7204624c | 2978 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
82bc0194 | 2979 | return -EIO; |
b411b363 PR |
2980 | } |
2981 | ||
2982 | /* helper function | |
2983 | * input: alg name, feature name | |
2984 | * return: NULL (alg name was "") | |
2985 | * ERR_PTR(error) if something goes wrong | |
2986 | * or the crypto hash ptr, if it worked out ok. */ | |
2987 | struct crypto_hash *drbd_crypto_alloc_digest_safe(const struct drbd_conf *mdev, | |
2988 | const char *alg, const char *name) | |
2989 | { | |
2990 | struct crypto_hash *tfm; | |
2991 | ||
2992 | if (!alg[0]) | |
2993 | return NULL; | |
2994 | ||
2995 | tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC); | |
2996 | if (IS_ERR(tfm)) { | |
2997 | dev_err(DEV, "Can not allocate \"%s\" as %s (reason: %ld)\n", | |
2998 | alg, name, PTR_ERR(tfm)); | |
2999 | return tfm; | |
3000 | } | |
3001 | if (!drbd_crypto_is_hash(crypto_hash_tfm(tfm))) { | |
3002 | crypto_free_hash(tfm); | |
3003 | dev_err(DEV, "\"%s\" is not a digest (%s)\n", alg, name); | |
3004 | return ERR_PTR(-EINVAL); | |
3005 | } | |
3006 | return tfm; | |
3007 | } | |
3008 | ||
4a76b161 AG |
3009 | static int ignore_remaining_packet(struct drbd_tconn *tconn, struct packet_info *pi) |
3010 | { | |
3011 | void *buffer = tconn->data.rbuf; | |
3012 | int size = pi->size; | |
3013 | ||
3014 | while (size) { | |
3015 | int s = min_t(int, size, DRBD_SOCKET_BUFFER_SIZE); | |
3016 | s = drbd_recv(tconn, buffer, s); | |
3017 | if (s <= 0) { | |
3018 | if (s < 0) | |
3019 | return s; | |
3020 | break; | |
3021 | } | |
3022 | size -= s; | |
3023 | } | |
3024 | if (size) | |
3025 | return -EIO; | |
3026 | return 0; | |
3027 | } | |
3028 | ||
3029 | /* | |
3030 | * config_unknown_volume - device configuration command for unknown volume | |
3031 | * | |
3032 | * When a device is added to an existing connection, the node on which the | |
3033 | * device is added first will send configuration commands to its peer but the | |
3034 | * peer will not know about the device yet. It will warn and ignore these | |
3035 | * commands. Once the device is added on the second node, the second node will | |
3036 | * send the same device configuration commands, but in the other direction. | |
3037 | * | |
3038 | * (We can also end up here if drbd is misconfigured.) | |
3039 | */ | |
3040 | static int config_unknown_volume(struct drbd_tconn *tconn, struct packet_info *pi) | |
3041 | { | |
3042 | conn_warn(tconn, "Volume %u unknown; ignoring %s packet\n", | |
3043 | pi->vnr, cmdname(pi->cmd)); | |
3044 | return ignore_remaining_packet(tconn, pi); | |
3045 | } | |
3046 | ||
3047 | static int receive_SyncParam(struct drbd_tconn *tconn, struct packet_info *pi) | |
b411b363 | 3048 | { |
4a76b161 | 3049 | struct drbd_conf *mdev; |
e658983a | 3050 | struct p_rs_param_95 *p; |
b411b363 PR |
3051 | unsigned int header_size, data_size, exp_max_sz; |
3052 | struct crypto_hash *verify_tfm = NULL; | |
3053 | struct crypto_hash *csums_tfm = NULL; | |
4a76b161 | 3054 | const int apv = tconn->agreed_pro_version; |
778f271d PR |
3055 | int *rs_plan_s = NULL; |
3056 | int fifo_size = 0; | |
82bc0194 | 3057 | int err; |
b411b363 | 3058 | |
4a76b161 AG |
3059 | mdev = vnr_to_mdev(tconn, pi->vnr); |
3060 | if (!mdev) | |
3061 | return config_unknown_volume(tconn, pi); | |
3062 | ||
b411b363 PR |
3063 | exp_max_sz = apv <= 87 ? sizeof(struct p_rs_param) |
3064 | : apv == 88 ? sizeof(struct p_rs_param) | |
3065 | + SHARED_SECRET_MAX | |
8e26f9cc PR |
3066 | : apv <= 94 ? sizeof(struct p_rs_param_89) |
3067 | : /* apv >= 95 */ sizeof(struct p_rs_param_95); | |
b411b363 | 3068 | |
e2857216 | 3069 | if (pi->size > exp_max_sz) { |
b411b363 | 3070 | dev_err(DEV, "SyncParam packet too long: received %u, expected <= %u bytes\n", |
e2857216 | 3071 | pi->size, exp_max_sz); |
82bc0194 | 3072 | return -EIO; |
b411b363 PR |
3073 | } |
3074 | ||
3075 | if (apv <= 88) { | |
e658983a | 3076 | header_size = sizeof(struct p_rs_param); |
e2857216 | 3077 | data_size = pi->size - header_size; |
8e26f9cc | 3078 | } else if (apv <= 94) { |
e658983a | 3079 | header_size = sizeof(struct p_rs_param_89); |
e2857216 | 3080 | data_size = pi->size - header_size; |
b411b363 | 3081 | D_ASSERT(data_size == 0); |
8e26f9cc | 3082 | } else { |
e658983a | 3083 | header_size = sizeof(struct p_rs_param_95); |
e2857216 | 3084 | data_size = pi->size - header_size; |
b411b363 PR |
3085 | D_ASSERT(data_size == 0); |
3086 | } | |
3087 | ||
3088 | /* initialize verify_alg and csums_alg */ | |
e658983a | 3089 | p = pi->data; |
b411b363 PR |
3090 | memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); |
3091 | ||
e658983a | 3092 | err = drbd_recv_all(mdev->tconn, p, header_size); |
82bc0194 AG |
3093 | if (err) |
3094 | return err; | |
b411b363 | 3095 | |
f399002e LE |
3096 | if (get_ldev(mdev)) { |
3097 | mdev->ldev->dc.resync_rate = be32_to_cpu(p->rate); | |
3098 | put_ldev(mdev); | |
3099 | } | |
b411b363 PR |
3100 | |
3101 | if (apv >= 88) { | |
3102 | if (apv == 88) { | |
3103 | if (data_size > SHARED_SECRET_MAX) { | |
3104 | dev_err(DEV, "verify-alg too long, " | |
3105 | "peer wants %u, accepting only %u byte\n", | |
3106 | data_size, SHARED_SECRET_MAX); | |
82bc0194 | 3107 | return -EIO; |
b411b363 PR |
3108 | } |
3109 | ||
82bc0194 AG |
3110 | err = drbd_recv_all(mdev->tconn, p->verify_alg, data_size); |
3111 | if (err) | |
3112 | return err; | |
b411b363 PR |
3113 | |
3114 | /* we expect NUL terminated string */ | |
3115 | /* but just in case someone tries to be evil */ | |
3116 | D_ASSERT(p->verify_alg[data_size-1] == 0); | |
3117 | p->verify_alg[data_size-1] = 0; | |
3118 | ||
3119 | } else /* apv >= 89 */ { | |
3120 | /* we still expect NUL terminated strings */ | |
3121 | /* but just in case someone tries to be evil */ | |
3122 | D_ASSERT(p->verify_alg[SHARED_SECRET_MAX-1] == 0); | |
3123 | D_ASSERT(p->csums_alg[SHARED_SECRET_MAX-1] == 0); | |
3124 | p->verify_alg[SHARED_SECRET_MAX-1] = 0; | |
3125 | p->csums_alg[SHARED_SECRET_MAX-1] = 0; | |
3126 | } | |
3127 | ||
f399002e | 3128 | if (strcmp(mdev->tconn->net_conf->verify_alg, p->verify_alg)) { |
b411b363 PR |
3129 | if (mdev->state.conn == C_WF_REPORT_PARAMS) { |
3130 | dev_err(DEV, "Different verify-alg settings. me=\"%s\" peer=\"%s\"\n", | |
f399002e | 3131 | mdev->tconn->net_conf->verify_alg, p->verify_alg); |
b411b363 PR |
3132 | goto disconnect; |
3133 | } | |
3134 | verify_tfm = drbd_crypto_alloc_digest_safe(mdev, | |
3135 | p->verify_alg, "verify-alg"); | |
3136 | if (IS_ERR(verify_tfm)) { | |
3137 | verify_tfm = NULL; | |
3138 | goto disconnect; | |
3139 | } | |
3140 | } | |
3141 | ||
f399002e | 3142 | if (apv >= 89 && strcmp(mdev->tconn->net_conf->csums_alg, p->csums_alg)) { |
b411b363 PR |
3143 | if (mdev->state.conn == C_WF_REPORT_PARAMS) { |
3144 | dev_err(DEV, "Different csums-alg settings. me=\"%s\" peer=\"%s\"\n", | |
f399002e | 3145 | mdev->tconn->net_conf->csums_alg, p->csums_alg); |
b411b363 PR |
3146 | goto disconnect; |
3147 | } | |
3148 | csums_tfm = drbd_crypto_alloc_digest_safe(mdev, | |
3149 | p->csums_alg, "csums-alg"); | |
3150 | if (IS_ERR(csums_tfm)) { | |
3151 | csums_tfm = NULL; | |
3152 | goto disconnect; | |
3153 | } | |
3154 | } | |
3155 | ||
f399002e LE |
3156 | if (apv > 94 && get_ldev(mdev)) { |
3157 | mdev->ldev->dc.resync_rate = be32_to_cpu(p->rate); | |
3158 | mdev->ldev->dc.c_plan_ahead = be32_to_cpu(p->c_plan_ahead); | |
3159 | mdev->ldev->dc.c_delay_target = be32_to_cpu(p->c_delay_target); | |
3160 | mdev->ldev->dc.c_fill_target = be32_to_cpu(p->c_fill_target); | |
3161 | mdev->ldev->dc.c_max_rate = be32_to_cpu(p->c_max_rate); | |
778f271d | 3162 | |
f399002e | 3163 | fifo_size = (mdev->ldev->dc.c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
778f271d PR |
3164 | if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) { |
3165 | rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL); | |
3166 | if (!rs_plan_s) { | |
3167 | dev_err(DEV, "kmalloc of fifo_buffer failed"); | |
f399002e | 3168 | put_ldev(mdev); |
778f271d PR |
3169 | goto disconnect; |
3170 | } | |
3171 | } | |
f399002e | 3172 | put_ldev(mdev); |
8e26f9cc | 3173 | } |
b411b363 PR |
3174 | |
3175 | spin_lock(&mdev->peer_seq_lock); | |
3176 | /* lock against drbd_nl_syncer_conf() */ | |
3177 | if (verify_tfm) { | |
f399002e LE |
3178 | strcpy(mdev->tconn->net_conf->verify_alg, p->verify_alg); |
3179 | mdev->tconn->net_conf->verify_alg_len = strlen(p->verify_alg) + 1; | |
3180 | crypto_free_hash(mdev->tconn->verify_tfm); | |
3181 | mdev->tconn->verify_tfm = verify_tfm; | |
b411b363 PR |
3182 | dev_info(DEV, "using verify-alg: \"%s\"\n", p->verify_alg); |
3183 | } | |
3184 | if (csums_tfm) { | |
f399002e LE |
3185 | strcpy(mdev->tconn->net_conf->csums_alg, p->csums_alg); |
3186 | mdev->tconn->net_conf->csums_alg_len = strlen(p->csums_alg) + 1; | |
3187 | crypto_free_hash(mdev->tconn->csums_tfm); | |
3188 | mdev->tconn->csums_tfm = csums_tfm; | |
b411b363 PR |
3189 | dev_info(DEV, "using csums-alg: \"%s\"\n", p->csums_alg); |
3190 | } | |
778f271d PR |
3191 | if (fifo_size != mdev->rs_plan_s.size) { |
3192 | kfree(mdev->rs_plan_s.values); | |
3193 | mdev->rs_plan_s.values = rs_plan_s; | |
3194 | mdev->rs_plan_s.size = fifo_size; | |
3195 | mdev->rs_planed = 0; | |
3196 | } | |
b411b363 PR |
3197 | spin_unlock(&mdev->peer_seq_lock); |
3198 | } | |
82bc0194 | 3199 | return 0; |
b411b363 | 3200 | |
b411b363 PR |
3201 | disconnect: |
3202 | /* just for completeness: actually not needed, | |
3203 | * as this is not reached if csums_tfm was ok. */ | |
3204 | crypto_free_hash(csums_tfm); | |
3205 | /* but free the verify_tfm again, if csums_tfm did not work out */ | |
3206 | crypto_free_hash(verify_tfm); | |
38fa9988 | 3207 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
82bc0194 | 3208 | return -EIO; |
b411b363 PR |
3209 | } |
3210 | ||
b411b363 PR |
3211 | /* warn if the arguments differ by more than 12.5% */ |
3212 | static void warn_if_differ_considerably(struct drbd_conf *mdev, | |
3213 | const char *s, sector_t a, sector_t b) | |
3214 | { | |
3215 | sector_t d; | |
3216 | if (a == 0 || b == 0) | |
3217 | return; | |
3218 | d = (a > b) ? (a - b) : (b - a); | |
3219 | if (d > (a>>3) || d > (b>>3)) | |
3220 | dev_warn(DEV, "Considerable difference in %s: %llus vs. %llus\n", s, | |
3221 | (unsigned long long)a, (unsigned long long)b); | |
3222 | } | |
3223 | ||
4a76b161 | 3224 | static int receive_sizes(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3225 | { |
4a76b161 | 3226 | struct drbd_conf *mdev; |
e658983a | 3227 | struct p_sizes *p = pi->data; |
b411b363 | 3228 | enum determine_dev_size dd = unchanged; |
b411b363 PR |
3229 | sector_t p_size, p_usize, my_usize; |
3230 | int ldsc = 0; /* local disk size changed */ | |
e89b591c | 3231 | enum dds_flags ddsf; |
b411b363 | 3232 | |
4a76b161 AG |
3233 | mdev = vnr_to_mdev(tconn, pi->vnr); |
3234 | if (!mdev) | |
3235 | return config_unknown_volume(tconn, pi); | |
3236 | ||
b411b363 PR |
3237 | p_size = be64_to_cpu(p->d_size); |
3238 | p_usize = be64_to_cpu(p->u_size); | |
3239 | ||
b411b363 PR |
3240 | /* just store the peer's disk size for now. |
3241 | * we still need to figure out whether we accept that. */ | |
3242 | mdev->p_size = p_size; | |
3243 | ||
b411b363 PR |
3244 | if (get_ldev(mdev)) { |
3245 | warn_if_differ_considerably(mdev, "lower level device sizes", | |
3246 | p_size, drbd_get_max_capacity(mdev->ldev)); | |
3247 | warn_if_differ_considerably(mdev, "user requested size", | |
3248 | p_usize, mdev->ldev->dc.disk_size); | |
3249 | ||
3250 | /* if this is the first connect, or an otherwise expected | |
3251 | * param exchange, choose the minimum */ | |
3252 | if (mdev->state.conn == C_WF_REPORT_PARAMS) | |
3253 | p_usize = min_not_zero((sector_t)mdev->ldev->dc.disk_size, | |
3254 | p_usize); | |
3255 | ||
3256 | my_usize = mdev->ldev->dc.disk_size; | |
3257 | ||
3258 | if (mdev->ldev->dc.disk_size != p_usize) { | |
3259 | mdev->ldev->dc.disk_size = p_usize; | |
3260 | dev_info(DEV, "Peer sets u_size to %lu sectors\n", | |
3261 | (unsigned long)mdev->ldev->dc.disk_size); | |
3262 | } | |
3263 | ||
3264 | /* Never shrink a device with usable data during connect. | |
3265 | But allow online shrinking if we are connected. */ | |
a393db6f | 3266 | if (drbd_new_dev_size(mdev, mdev->ldev, 0) < |
b411b363 PR |
3267 | drbd_get_capacity(mdev->this_bdev) && |
3268 | mdev->state.disk >= D_OUTDATED && | |
3269 | mdev->state.conn < C_CONNECTED) { | |
3270 | dev_err(DEV, "The peer's disk size is too small!\n"); | |
38fa9988 | 3271 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 PR |
3272 | mdev->ldev->dc.disk_size = my_usize; |
3273 | put_ldev(mdev); | |
82bc0194 | 3274 | return -EIO; |
b411b363 PR |
3275 | } |
3276 | put_ldev(mdev); | |
3277 | } | |
b411b363 | 3278 | |
e89b591c | 3279 | ddsf = be16_to_cpu(p->dds_flags); |
b411b363 | 3280 | if (get_ldev(mdev)) { |
24c4830c | 3281 | dd = drbd_determine_dev_size(mdev, ddsf); |
b411b363 PR |
3282 | put_ldev(mdev); |
3283 | if (dd == dev_size_error) | |
82bc0194 | 3284 | return -EIO; |
b411b363 PR |
3285 | drbd_md_sync(mdev); |
3286 | } else { | |
3287 | /* I am diskless, need to accept the peer's size. */ | |
3288 | drbd_set_my_capacity(mdev, p_size); | |
3289 | } | |
3290 | ||
99432fcc PR |
3291 | mdev->peer_max_bio_size = be32_to_cpu(p->max_bio_size); |
3292 | drbd_reconsider_max_bio_size(mdev); | |
3293 | ||
b411b363 PR |
3294 | if (get_ldev(mdev)) { |
3295 | if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) { | |
3296 | mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); | |
3297 | ldsc = 1; | |
3298 | } | |
3299 | ||
b411b363 PR |
3300 | put_ldev(mdev); |
3301 | } | |
3302 | ||
3303 | if (mdev->state.conn > C_WF_REPORT_PARAMS) { | |
3304 | if (be64_to_cpu(p->c_size) != | |
3305 | drbd_get_capacity(mdev->this_bdev) || ldsc) { | |
3306 | /* we have different sizes, probably peer | |
3307 | * needs to know my new size... */ | |
e89b591c | 3308 | drbd_send_sizes(mdev, 0, ddsf); |
b411b363 PR |
3309 | } |
3310 | if (test_and_clear_bit(RESIZE_PENDING, &mdev->flags) || | |
3311 | (dd == grew && mdev->state.conn == C_CONNECTED)) { | |
3312 | if (mdev->state.pdsk >= D_INCONSISTENT && | |
e89b591c PR |
3313 | mdev->state.disk >= D_INCONSISTENT) { |
3314 | if (ddsf & DDSF_NO_RESYNC) | |
3315 | dev_info(DEV, "Resync of new storage suppressed with --assume-clean\n"); | |
3316 | else | |
3317 | resync_after_online_grow(mdev); | |
3318 | } else | |
b411b363 PR |
3319 | set_bit(RESYNC_AFTER_NEG, &mdev->flags); |
3320 | } | |
3321 | } | |
3322 | ||
82bc0194 | 3323 | return 0; |
b411b363 PR |
3324 | } |
3325 | ||
4a76b161 | 3326 | static int receive_uuids(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3327 | { |
4a76b161 | 3328 | struct drbd_conf *mdev; |
e658983a | 3329 | struct p_uuids *p = pi->data; |
b411b363 | 3330 | u64 *p_uuid; |
62b0da3a | 3331 | int i, updated_uuids = 0; |
b411b363 | 3332 | |
4a76b161 AG |
3333 | mdev = vnr_to_mdev(tconn, pi->vnr); |
3334 | if (!mdev) | |
3335 | return config_unknown_volume(tconn, pi); | |
3336 | ||
b411b363 PR |
3337 | p_uuid = kmalloc(sizeof(u64)*UI_EXTENDED_SIZE, GFP_NOIO); |
3338 | ||
3339 | for (i = UI_CURRENT; i < UI_EXTENDED_SIZE; i++) | |
3340 | p_uuid[i] = be64_to_cpu(p->uuid[i]); | |
3341 | ||
3342 | kfree(mdev->p_uuid); | |
3343 | mdev->p_uuid = p_uuid; | |
3344 | ||
3345 | if (mdev->state.conn < C_CONNECTED && | |
3346 | mdev->state.disk < D_INCONSISTENT && | |
3347 | mdev->state.role == R_PRIMARY && | |
3348 | (mdev->ed_uuid & ~((u64)1)) != (p_uuid[UI_CURRENT] & ~((u64)1))) { | |
3349 | dev_err(DEV, "Can only connect to data with current UUID=%016llX\n", | |
3350 | (unsigned long long)mdev->ed_uuid); | |
38fa9988 | 3351 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
82bc0194 | 3352 | return -EIO; |
b411b363 PR |
3353 | } |
3354 | ||
3355 | if (get_ldev(mdev)) { | |
3356 | int skip_initial_sync = | |
3357 | mdev->state.conn == C_CONNECTED && | |
31890f4a | 3358 | mdev->tconn->agreed_pro_version >= 90 && |
b411b363 PR |
3359 | mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && |
3360 | (p_uuid[UI_FLAGS] & 8); | |
3361 | if (skip_initial_sync) { | |
3362 | dev_info(DEV, "Accepted new current UUID, preparing to skip initial sync\n"); | |
3363 | drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, | |
20ceb2b2 LE |
3364 | "clear_n_write from receive_uuids", |
3365 | BM_LOCKED_TEST_ALLOWED); | |
b411b363 PR |
3366 | _drbd_uuid_set(mdev, UI_CURRENT, p_uuid[UI_CURRENT]); |
3367 | _drbd_uuid_set(mdev, UI_BITMAP, 0); | |
3368 | _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), | |
3369 | CS_VERBOSE, NULL); | |
3370 | drbd_md_sync(mdev); | |
62b0da3a | 3371 | updated_uuids = 1; |
b411b363 PR |
3372 | } |
3373 | put_ldev(mdev); | |
18a50fa2 PR |
3374 | } else if (mdev->state.disk < D_INCONSISTENT && |
3375 | mdev->state.role == R_PRIMARY) { | |
3376 | /* I am a diskless primary, the peer just created a new current UUID | |
3377 | for me. */ | |
62b0da3a | 3378 | updated_uuids = drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]); |
b411b363 PR |
3379 | } |
3380 | ||
3381 | /* Before we test for the disk state, we should wait until an eventually | |
3382 | ongoing cluster wide state change is finished. That is important if | |
3383 | we are primary and are detaching from our disk. We need to see the | |
3384 | new disk state... */ | |
8410da8f PR |
3385 | mutex_lock(mdev->state_mutex); |
3386 | mutex_unlock(mdev->state_mutex); | |
b411b363 | 3387 | if (mdev->state.conn >= C_CONNECTED && mdev->state.disk < D_INCONSISTENT) |
62b0da3a LE |
3388 | updated_uuids |= drbd_set_ed_uuid(mdev, p_uuid[UI_CURRENT]); |
3389 | ||
3390 | if (updated_uuids) | |
3391 | drbd_print_uuids(mdev, "receiver updated UUIDs to"); | |
b411b363 | 3392 | |
82bc0194 | 3393 | return 0; |
b411b363 PR |
3394 | } |
3395 | ||
3396 | /** | |
3397 | * convert_state() - Converts the peer's view of the cluster state to our point of view | |
3398 | * @ps: The state as seen by the peer. | |
3399 | */ | |
3400 | static union drbd_state convert_state(union drbd_state ps) | |
3401 | { | |
3402 | union drbd_state ms; | |
3403 | ||
3404 | static enum drbd_conns c_tab[] = { | |
3405 | [C_CONNECTED] = C_CONNECTED, | |
3406 | ||
3407 | [C_STARTING_SYNC_S] = C_STARTING_SYNC_T, | |
3408 | [C_STARTING_SYNC_T] = C_STARTING_SYNC_S, | |
3409 | [C_DISCONNECTING] = C_TEAR_DOWN, /* C_NETWORK_FAILURE, */ | |
3410 | [C_VERIFY_S] = C_VERIFY_T, | |
3411 | [C_MASK] = C_MASK, | |
3412 | }; | |
3413 | ||
3414 | ms.i = ps.i; | |
3415 | ||
3416 | ms.conn = c_tab[ps.conn]; | |
3417 | ms.peer = ps.role; | |
3418 | ms.role = ps.peer; | |
3419 | ms.pdsk = ps.disk; | |
3420 | ms.disk = ps.pdsk; | |
3421 | ms.peer_isp = (ps.aftr_isp | ps.user_isp); | |
3422 | ||
3423 | return ms; | |
3424 | } | |
3425 | ||
4a76b161 | 3426 | static int receive_req_state(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3427 | { |
4a76b161 | 3428 | struct drbd_conf *mdev; |
e658983a | 3429 | struct p_req_state *p = pi->data; |
b411b363 | 3430 | union drbd_state mask, val; |
bf885f8a | 3431 | enum drbd_state_rv rv; |
b411b363 | 3432 | |
4a76b161 AG |
3433 | mdev = vnr_to_mdev(tconn, pi->vnr); |
3434 | if (!mdev) | |
3435 | return -EIO; | |
3436 | ||
b411b363 PR |
3437 | mask.i = be32_to_cpu(p->mask); |
3438 | val.i = be32_to_cpu(p->val); | |
3439 | ||
25703f83 | 3440 | if (test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags) && |
8410da8f | 3441 | mutex_is_locked(mdev->state_mutex)) { |
b411b363 | 3442 | drbd_send_sr_reply(mdev, SS_CONCURRENT_ST_CHG); |
82bc0194 | 3443 | return 0; |
b411b363 PR |
3444 | } |
3445 | ||
3446 | mask = convert_state(mask); | |
3447 | val = convert_state(val); | |
3448 | ||
dfafcc8a PR |
3449 | rv = drbd_change_state(mdev, CS_VERBOSE, mask, val); |
3450 | drbd_send_sr_reply(mdev, rv); | |
b411b363 | 3451 | |
b411b363 PR |
3452 | drbd_md_sync(mdev); |
3453 | ||
82bc0194 | 3454 | return 0; |
b411b363 PR |
3455 | } |
3456 | ||
e2857216 | 3457 | static int receive_req_conn_state(struct drbd_tconn *tconn, struct packet_info *pi) |
dfafcc8a | 3458 | { |
e658983a | 3459 | struct p_req_state *p = pi->data; |
dfafcc8a PR |
3460 | union drbd_state mask, val; |
3461 | enum drbd_state_rv rv; | |
3462 | ||
3463 | mask.i = be32_to_cpu(p->mask); | |
3464 | val.i = be32_to_cpu(p->val); | |
3465 | ||
3466 | if (test_bit(DISCARD_CONCURRENT, &tconn->flags) && | |
3467 | mutex_is_locked(&tconn->cstate_mutex)) { | |
3468 | conn_send_sr_reply(tconn, SS_CONCURRENT_ST_CHG); | |
82bc0194 | 3469 | return 0; |
dfafcc8a PR |
3470 | } |
3471 | ||
3472 | mask = convert_state(mask); | |
3473 | val = convert_state(val); | |
3474 | ||
778bcf2e | 3475 | rv = conn_request_state(tconn, mask, val, CS_VERBOSE | CS_LOCAL_ONLY | CS_IGN_OUTD_FAIL); |
dfafcc8a PR |
3476 | conn_send_sr_reply(tconn, rv); |
3477 | ||
82bc0194 | 3478 | return 0; |
dfafcc8a PR |
3479 | } |
3480 | ||
4a76b161 | 3481 | static int receive_state(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3482 | { |
4a76b161 | 3483 | struct drbd_conf *mdev; |
e658983a | 3484 | struct p_state *p = pi->data; |
4ac4aada | 3485 | union drbd_state os, ns, peer_state; |
b411b363 | 3486 | enum drbd_disk_state real_peer_disk; |
65d922c3 | 3487 | enum chg_state_flags cs_flags; |
b411b363 PR |
3488 | int rv; |
3489 | ||
4a76b161 AG |
3490 | mdev = vnr_to_mdev(tconn, pi->vnr); |
3491 | if (!mdev) | |
3492 | return config_unknown_volume(tconn, pi); | |
3493 | ||
b411b363 PR |
3494 | peer_state.i = be32_to_cpu(p->state); |
3495 | ||
3496 | real_peer_disk = peer_state.disk; | |
3497 | if (peer_state.disk == D_NEGOTIATING) { | |
3498 | real_peer_disk = mdev->p_uuid[UI_FLAGS] & 4 ? D_INCONSISTENT : D_CONSISTENT; | |
3499 | dev_info(DEV, "real peer disk state = %s\n", drbd_disk_str(real_peer_disk)); | |
3500 | } | |
3501 | ||
87eeee41 | 3502 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 | 3503 | retry: |
78bae59b | 3504 | os = ns = drbd_read_state(mdev); |
87eeee41 | 3505 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 | 3506 | |
e9ef7bb6 LE |
3507 | /* peer says his disk is uptodate, while we think it is inconsistent, |
3508 | * and this happens while we think we have a sync going on. */ | |
3509 | if (os.pdsk == D_INCONSISTENT && real_peer_disk == D_UP_TO_DATE && | |
3510 | os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) { | |
3511 | /* If we are (becoming) SyncSource, but peer is still in sync | |
3512 | * preparation, ignore its uptodate-ness to avoid flapping, it | |
3513 | * will change to inconsistent once the peer reaches active | |
3514 | * syncing states. | |
3515 | * It may have changed syncer-paused flags, however, so we | |
3516 | * cannot ignore this completely. */ | |
3517 | if (peer_state.conn > C_CONNECTED && | |
3518 | peer_state.conn < C_SYNC_SOURCE) | |
3519 | real_peer_disk = D_INCONSISTENT; | |
3520 | ||
3521 | /* if peer_state changes to connected at the same time, | |
3522 | * it explicitly notifies us that it finished resync. | |
3523 | * Maybe we should finish it up, too? */ | |
3524 | else if (os.conn >= C_SYNC_SOURCE && | |
3525 | peer_state.conn == C_CONNECTED) { | |
3526 | if (drbd_bm_total_weight(mdev) <= mdev->rs_failed) | |
3527 | drbd_resync_finished(mdev); | |
82bc0194 | 3528 | return 0; |
e9ef7bb6 LE |
3529 | } |
3530 | } | |
3531 | ||
3532 | /* peer says his disk is inconsistent, while we think it is uptodate, | |
3533 | * and this happens while the peer still thinks we have a sync going on, | |
3534 | * but we think we are already done with the sync. | |
3535 | * We ignore this to avoid flapping pdsk. | |
3536 | * This should not happen, if the peer is a recent version of drbd. */ | |
3537 | if (os.pdsk == D_UP_TO_DATE && real_peer_disk == D_INCONSISTENT && | |
3538 | os.conn == C_CONNECTED && peer_state.conn > C_SYNC_SOURCE) | |
3539 | real_peer_disk = D_UP_TO_DATE; | |
3540 | ||
4ac4aada LE |
3541 | if (ns.conn == C_WF_REPORT_PARAMS) |
3542 | ns.conn = C_CONNECTED; | |
b411b363 | 3543 | |
67531718 PR |
3544 | if (peer_state.conn == C_AHEAD) |
3545 | ns.conn = C_BEHIND; | |
3546 | ||
b411b363 PR |
3547 | if (mdev->p_uuid && peer_state.disk >= D_NEGOTIATING && |
3548 | get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
3549 | int cr; /* consider resync */ | |
3550 | ||
3551 | /* if we established a new connection */ | |
4ac4aada | 3552 | cr = (os.conn < C_CONNECTED); |
b411b363 PR |
3553 | /* if we had an established connection |
3554 | * and one of the nodes newly attaches a disk */ | |
4ac4aada | 3555 | cr |= (os.conn == C_CONNECTED && |
b411b363 | 3556 | (peer_state.disk == D_NEGOTIATING || |
4ac4aada | 3557 | os.disk == D_NEGOTIATING)); |
b411b363 PR |
3558 | /* if we have both been inconsistent, and the peer has been |
3559 | * forced to be UpToDate with --overwrite-data */ | |
3560 | cr |= test_bit(CONSIDER_RESYNC, &mdev->flags); | |
3561 | /* if we had been plain connected, and the admin requested to | |
3562 | * start a sync by "invalidate" or "invalidate-remote" */ | |
4ac4aada | 3563 | cr |= (os.conn == C_CONNECTED && |
b411b363 PR |
3564 | (peer_state.conn >= C_STARTING_SYNC_S && |
3565 | peer_state.conn <= C_WF_BITMAP_T)); | |
3566 | ||
3567 | if (cr) | |
4ac4aada | 3568 | ns.conn = drbd_sync_handshake(mdev, peer_state.role, real_peer_disk); |
b411b363 PR |
3569 | |
3570 | put_ldev(mdev); | |
4ac4aada LE |
3571 | if (ns.conn == C_MASK) { |
3572 | ns.conn = C_CONNECTED; | |
b411b363 | 3573 | if (mdev->state.disk == D_NEGOTIATING) { |
82f59cc6 | 3574 | drbd_force_state(mdev, NS(disk, D_FAILED)); |
b411b363 PR |
3575 | } else if (peer_state.disk == D_NEGOTIATING) { |
3576 | dev_err(DEV, "Disk attach process on the peer node was aborted.\n"); | |
3577 | peer_state.disk = D_DISKLESS; | |
580b9767 | 3578 | real_peer_disk = D_DISKLESS; |
b411b363 | 3579 | } else { |
8169e41b | 3580 | if (test_and_clear_bit(CONN_DRY_RUN, &mdev->tconn->flags)) |
82bc0194 | 3581 | return -EIO; |
4ac4aada | 3582 | D_ASSERT(os.conn == C_WF_REPORT_PARAMS); |
38fa9988 | 3583 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
82bc0194 | 3584 | return -EIO; |
b411b363 PR |
3585 | } |
3586 | } | |
3587 | } | |
3588 | ||
87eeee41 | 3589 | spin_lock_irq(&mdev->tconn->req_lock); |
78bae59b | 3590 | if (os.i != drbd_read_state(mdev).i) |
b411b363 PR |
3591 | goto retry; |
3592 | clear_bit(CONSIDER_RESYNC, &mdev->flags); | |
b411b363 PR |
3593 | ns.peer = peer_state.role; |
3594 | ns.pdsk = real_peer_disk; | |
3595 | ns.peer_isp = (peer_state.aftr_isp | peer_state.user_isp); | |
4ac4aada | 3596 | if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING) |
b411b363 | 3597 | ns.disk = mdev->new_state_tmp.disk; |
4ac4aada | 3598 | cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD); |
2aebfabb | 3599 | if (ns.pdsk == D_CONSISTENT && drbd_suspended(mdev) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED && |
481c6f50 | 3600 | test_bit(NEW_CUR_UUID, &mdev->flags)) { |
8554df1c | 3601 | /* Do not allow tl_restart(RESEND) for a rebooted peer. We can only allow this |
481c6f50 | 3602 | for temporal network outages! */ |
87eeee41 | 3603 | spin_unlock_irq(&mdev->tconn->req_lock); |
481c6f50 | 3604 | dev_err(DEV, "Aborting Connect, can not thaw IO with an only Consistent peer\n"); |
2f5cdd0b | 3605 | tl_clear(mdev->tconn); |
481c6f50 PR |
3606 | drbd_uuid_new_current(mdev); |
3607 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
38fa9988 | 3608 | conn_request_state(mdev->tconn, NS2(conn, C_PROTOCOL_ERROR, susp, 0), CS_HARD); |
82bc0194 | 3609 | return -EIO; |
481c6f50 | 3610 | } |
65d922c3 | 3611 | rv = _drbd_set_state(mdev, ns, cs_flags, NULL); |
78bae59b | 3612 | ns = drbd_read_state(mdev); |
87eeee41 | 3613 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
3614 | |
3615 | if (rv < SS_SUCCESS) { | |
38fa9988 | 3616 | conn_request_state(mdev->tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
82bc0194 | 3617 | return -EIO; |
b411b363 PR |
3618 | } |
3619 | ||
4ac4aada LE |
3620 | if (os.conn > C_WF_REPORT_PARAMS) { |
3621 | if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED && | |
b411b363 PR |
3622 | peer_state.disk != D_NEGOTIATING ) { |
3623 | /* we want resync, peer has not yet decided to sync... */ | |
3624 | /* Nowadays only used when forcing a node into primary role and | |
3625 | setting its disk to UpToDate with that */ | |
3626 | drbd_send_uuids(mdev); | |
3627 | drbd_send_state(mdev); | |
3628 | } | |
3629 | } | |
3630 | ||
89e58e75 | 3631 | mdev->tconn->net_conf->want_lose = 0; |
b411b363 PR |
3632 | |
3633 | drbd_md_sync(mdev); /* update connected indicator, la_size, ... */ | |
3634 | ||
82bc0194 | 3635 | return 0; |
b411b363 PR |
3636 | } |
3637 | ||
4a76b161 | 3638 | static int receive_sync_uuid(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3639 | { |
4a76b161 | 3640 | struct drbd_conf *mdev; |
e658983a | 3641 | struct p_rs_uuid *p = pi->data; |
4a76b161 AG |
3642 | |
3643 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
3644 | if (!mdev) | |
3645 | return -EIO; | |
b411b363 PR |
3646 | |
3647 | wait_event(mdev->misc_wait, | |
3648 | mdev->state.conn == C_WF_SYNC_UUID || | |
c4752ef1 | 3649 | mdev->state.conn == C_BEHIND || |
b411b363 PR |
3650 | mdev->state.conn < C_CONNECTED || |
3651 | mdev->state.disk < D_NEGOTIATING); | |
3652 | ||
3653 | /* D_ASSERT( mdev->state.conn == C_WF_SYNC_UUID ); */ | |
3654 | ||
b411b363 PR |
3655 | /* Here the _drbd_uuid_ functions are right, current should |
3656 | _not_ be rotated into the history */ | |
3657 | if (get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
3658 | _drbd_uuid_set(mdev, UI_CURRENT, be64_to_cpu(p->uuid)); | |
3659 | _drbd_uuid_set(mdev, UI_BITMAP, 0UL); | |
3660 | ||
62b0da3a | 3661 | drbd_print_uuids(mdev, "updated sync uuid"); |
b411b363 PR |
3662 | drbd_start_resync(mdev, C_SYNC_TARGET); |
3663 | ||
3664 | put_ldev(mdev); | |
3665 | } else | |
3666 | dev_err(DEV, "Ignoring SyncUUID packet!\n"); | |
3667 | ||
82bc0194 | 3668 | return 0; |
b411b363 PR |
3669 | } |
3670 | ||
2c46407d AG |
3671 | /** |
3672 | * receive_bitmap_plain | |
3673 | * | |
3674 | * Return 0 when done, 1 when another iteration is needed, and a negative error | |
3675 | * code upon failure. | |
3676 | */ | |
3677 | static int | |
50d0b1ad | 3678 | receive_bitmap_plain(struct drbd_conf *mdev, unsigned int size, |
e658983a | 3679 | unsigned long *p, struct bm_xfer_ctx *c) |
b411b363 | 3680 | { |
50d0b1ad AG |
3681 | unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE - |
3682 | drbd_header_size(mdev->tconn); | |
e658983a | 3683 | unsigned int num_words = min_t(size_t, data_size / sizeof(*p), |
50d0b1ad | 3684 | c->bm_words - c->word_offset); |
e658983a | 3685 | unsigned int want = num_words * sizeof(*p); |
2c46407d | 3686 | int err; |
b411b363 | 3687 | |
50d0b1ad AG |
3688 | if (want != size) { |
3689 | dev_err(DEV, "%s:want (%u) != size (%u)\n", __func__, want, size); | |
2c46407d | 3690 | return -EIO; |
b411b363 PR |
3691 | } |
3692 | if (want == 0) | |
2c46407d | 3693 | return 0; |
e658983a | 3694 | err = drbd_recv_all(mdev->tconn, p, want); |
82bc0194 | 3695 | if (err) |
2c46407d | 3696 | return err; |
b411b363 | 3697 | |
e658983a | 3698 | drbd_bm_merge_lel(mdev, c->word_offset, num_words, p); |
b411b363 PR |
3699 | |
3700 | c->word_offset += num_words; | |
3701 | c->bit_offset = c->word_offset * BITS_PER_LONG; | |
3702 | if (c->bit_offset > c->bm_bits) | |
3703 | c->bit_offset = c->bm_bits; | |
3704 | ||
2c46407d | 3705 | return 1; |
b411b363 PR |
3706 | } |
3707 | ||
a02d1240 AG |
3708 | static enum drbd_bitmap_code dcbp_get_code(struct p_compressed_bm *p) |
3709 | { | |
3710 | return (enum drbd_bitmap_code)(p->encoding & 0x0f); | |
3711 | } | |
3712 | ||
3713 | static int dcbp_get_start(struct p_compressed_bm *p) | |
3714 | { | |
3715 | return (p->encoding & 0x80) != 0; | |
3716 | } | |
3717 | ||
3718 | static int dcbp_get_pad_bits(struct p_compressed_bm *p) | |
3719 | { | |
3720 | return (p->encoding >> 4) & 0x7; | |
3721 | } | |
3722 | ||
2c46407d AG |
3723 | /** |
3724 | * recv_bm_rle_bits | |
3725 | * | |
3726 | * Return 0 when done, 1 when another iteration is needed, and a negative error | |
3727 | * code upon failure. | |
3728 | */ | |
3729 | static int | |
b411b363 PR |
3730 | recv_bm_rle_bits(struct drbd_conf *mdev, |
3731 | struct p_compressed_bm *p, | |
c6d25cfe PR |
3732 | struct bm_xfer_ctx *c, |
3733 | unsigned int len) | |
b411b363 PR |
3734 | { |
3735 | struct bitstream bs; | |
3736 | u64 look_ahead; | |
3737 | u64 rl; | |
3738 | u64 tmp; | |
3739 | unsigned long s = c->bit_offset; | |
3740 | unsigned long e; | |
a02d1240 | 3741 | int toggle = dcbp_get_start(p); |
b411b363 PR |
3742 | int have; |
3743 | int bits; | |
3744 | ||
a02d1240 | 3745 | bitstream_init(&bs, p->code, len, dcbp_get_pad_bits(p)); |
b411b363 PR |
3746 | |
3747 | bits = bitstream_get_bits(&bs, &look_ahead, 64); | |
3748 | if (bits < 0) | |
2c46407d | 3749 | return -EIO; |
b411b363 PR |
3750 | |
3751 | for (have = bits; have > 0; s += rl, toggle = !toggle) { | |
3752 | bits = vli_decode_bits(&rl, look_ahead); | |
3753 | if (bits <= 0) | |
2c46407d | 3754 | return -EIO; |
b411b363 PR |
3755 | |
3756 | if (toggle) { | |
3757 | e = s + rl -1; | |
3758 | if (e >= c->bm_bits) { | |
3759 | dev_err(DEV, "bitmap overflow (e:%lu) while decoding bm RLE packet\n", e); | |
2c46407d | 3760 | return -EIO; |
b411b363 PR |
3761 | } |
3762 | _drbd_bm_set_bits(mdev, s, e); | |
3763 | } | |
3764 | ||
3765 | if (have < bits) { | |
3766 | dev_err(DEV, "bitmap decoding error: h:%d b:%d la:0x%08llx l:%u/%u\n", | |
3767 | have, bits, look_ahead, | |
3768 | (unsigned int)(bs.cur.b - p->code), | |
3769 | (unsigned int)bs.buf_len); | |
2c46407d | 3770 | return -EIO; |
b411b363 PR |
3771 | } |
3772 | look_ahead >>= bits; | |
3773 | have -= bits; | |
3774 | ||
3775 | bits = bitstream_get_bits(&bs, &tmp, 64 - have); | |
3776 | if (bits < 0) | |
2c46407d | 3777 | return -EIO; |
b411b363 PR |
3778 | look_ahead |= tmp << have; |
3779 | have += bits; | |
3780 | } | |
3781 | ||
3782 | c->bit_offset = s; | |
3783 | bm_xfer_ctx_bit_to_word_offset(c); | |
3784 | ||
2c46407d | 3785 | return (s != c->bm_bits); |
b411b363 PR |
3786 | } |
3787 | ||
2c46407d AG |
3788 | /** |
3789 | * decode_bitmap_c | |
3790 | * | |
3791 | * Return 0 when done, 1 when another iteration is needed, and a negative error | |
3792 | * code upon failure. | |
3793 | */ | |
3794 | static int | |
b411b363 PR |
3795 | decode_bitmap_c(struct drbd_conf *mdev, |
3796 | struct p_compressed_bm *p, | |
c6d25cfe PR |
3797 | struct bm_xfer_ctx *c, |
3798 | unsigned int len) | |
b411b363 | 3799 | { |
a02d1240 | 3800 | if (dcbp_get_code(p) == RLE_VLI_Bits) |
e658983a | 3801 | return recv_bm_rle_bits(mdev, p, c, len - sizeof(*p)); |
b411b363 PR |
3802 | |
3803 | /* other variants had been implemented for evaluation, | |
3804 | * but have been dropped as this one turned out to be "best" | |
3805 | * during all our tests. */ | |
3806 | ||
3807 | dev_err(DEV, "receive_bitmap_c: unknown encoding %u\n", p->encoding); | |
38fa9988 | 3808 | conn_request_state(mdev->tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD); |
2c46407d | 3809 | return -EIO; |
b411b363 PR |
3810 | } |
3811 | ||
3812 | void INFO_bm_xfer_stats(struct drbd_conf *mdev, | |
3813 | const char *direction, struct bm_xfer_ctx *c) | |
3814 | { | |
3815 | /* what would it take to transfer it "plaintext" */ | |
50d0b1ad AG |
3816 | unsigned int header_size = drbd_header_size(mdev->tconn); |
3817 | unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE - header_size; | |
3818 | unsigned int plain = | |
3819 | header_size * (DIV_ROUND_UP(c->bm_words, data_size) + 1) + | |
3820 | c->bm_words * sizeof(unsigned long); | |
3821 | unsigned int total = c->bytes[0] + c->bytes[1]; | |
3822 | unsigned int r; | |
b411b363 PR |
3823 | |
3824 | /* total can not be zero. but just in case: */ | |
3825 | if (total == 0) | |
3826 | return; | |
3827 | ||
3828 | /* don't report if not compressed */ | |
3829 | if (total >= plain) | |
3830 | return; | |
3831 | ||
3832 | /* total < plain. check for overflow, still */ | |
3833 | r = (total > UINT_MAX/1000) ? (total / (plain/1000)) | |
3834 | : (1000 * total / plain); | |
3835 | ||
3836 | if (r > 1000) | |
3837 | r = 1000; | |
3838 | ||
3839 | r = 1000 - r; | |
3840 | dev_info(DEV, "%s bitmap stats [Bytes(packets)]: plain %u(%u), RLE %u(%u), " | |
3841 | "total %u; compression: %u.%u%%\n", | |
3842 | direction, | |
3843 | c->bytes[1], c->packets[1], | |
3844 | c->bytes[0], c->packets[0], | |
3845 | total, r/10, r % 10); | |
3846 | } | |
3847 | ||
3848 | /* Since we are processing the bitfield from lower addresses to higher, | |
3849 | it does not matter if the process it in 32 bit chunks or 64 bit | |
3850 | chunks as long as it is little endian. (Understand it as byte stream, | |
3851 | beginning with the lowest byte...) If we would use big endian | |
3852 | we would need to process it from the highest address to the lowest, | |
3853 | in order to be agnostic to the 32 vs 64 bits issue. | |
3854 | ||
3855 | returns 0 on failure, 1 if we successfully received it. */ | |
4a76b161 | 3856 | static int receive_bitmap(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3857 | { |
4a76b161 | 3858 | struct drbd_conf *mdev; |
b411b363 | 3859 | struct bm_xfer_ctx c; |
2c46407d | 3860 | int err; |
4a76b161 AG |
3861 | |
3862 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
3863 | if (!mdev) | |
3864 | return -EIO; | |
b411b363 | 3865 | |
20ceb2b2 LE |
3866 | drbd_bm_lock(mdev, "receive bitmap", BM_LOCKED_SET_ALLOWED); |
3867 | /* you are supposed to send additional out-of-sync information | |
3868 | * if you actually set bits during this phase */ | |
b411b363 | 3869 | |
b411b363 PR |
3870 | c = (struct bm_xfer_ctx) { |
3871 | .bm_bits = drbd_bm_bits(mdev), | |
3872 | .bm_words = drbd_bm_words(mdev), | |
3873 | }; | |
3874 | ||
2c46407d | 3875 | for(;;) { |
e658983a AG |
3876 | if (pi->cmd == P_BITMAP) |
3877 | err = receive_bitmap_plain(mdev, pi->size, pi->data, &c); | |
3878 | else if (pi->cmd == P_COMPRESSED_BITMAP) { | |
b411b363 PR |
3879 | /* MAYBE: sanity check that we speak proto >= 90, |
3880 | * and the feature is enabled! */ | |
e658983a | 3881 | struct p_compressed_bm *p = pi->data; |
b411b363 | 3882 | |
50d0b1ad | 3883 | if (pi->size > DRBD_SOCKET_BUFFER_SIZE - drbd_header_size(tconn)) { |
b411b363 | 3884 | dev_err(DEV, "ReportCBitmap packet too large\n"); |
82bc0194 | 3885 | err = -EIO; |
b411b363 PR |
3886 | goto out; |
3887 | } | |
e658983a | 3888 | if (pi->size <= sizeof(*p)) { |
e2857216 | 3889 | dev_err(DEV, "ReportCBitmap packet too small (l:%u)\n", pi->size); |
82bc0194 | 3890 | err = -EIO; |
78fcbdae | 3891 | goto out; |
b411b363 | 3892 | } |
e658983a AG |
3893 | err = drbd_recv_all(mdev->tconn, p, pi->size); |
3894 | if (err) | |
3895 | goto out; | |
e2857216 | 3896 | err = decode_bitmap_c(mdev, p, &c, pi->size); |
b411b363 | 3897 | } else { |
e2857216 | 3898 | dev_warn(DEV, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", pi->cmd); |
82bc0194 | 3899 | err = -EIO; |
b411b363 PR |
3900 | goto out; |
3901 | } | |
3902 | ||
e2857216 | 3903 | c.packets[pi->cmd == P_BITMAP]++; |
50d0b1ad | 3904 | c.bytes[pi->cmd == P_BITMAP] += drbd_header_size(tconn) + pi->size; |
b411b363 | 3905 | |
2c46407d AG |
3906 | if (err <= 0) { |
3907 | if (err < 0) | |
3908 | goto out; | |
b411b363 | 3909 | break; |
2c46407d | 3910 | } |
e2857216 | 3911 | err = drbd_recv_header(mdev->tconn, pi); |
82bc0194 | 3912 | if (err) |
b411b363 | 3913 | goto out; |
2c46407d | 3914 | } |
b411b363 PR |
3915 | |
3916 | INFO_bm_xfer_stats(mdev, "receive", &c); | |
3917 | ||
3918 | if (mdev->state.conn == C_WF_BITMAP_T) { | |
de1f8e4a AG |
3919 | enum drbd_state_rv rv; |
3920 | ||
82bc0194 AG |
3921 | err = drbd_send_bitmap(mdev); |
3922 | if (err) | |
b411b363 PR |
3923 | goto out; |
3924 | /* Omit CS_ORDERED with this state transition to avoid deadlocks. */ | |
de1f8e4a AG |
3925 | rv = _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); |
3926 | D_ASSERT(rv == SS_SUCCESS); | |
b411b363 PR |
3927 | } else if (mdev->state.conn != C_WF_BITMAP_S) { |
3928 | /* admin may have requested C_DISCONNECTING, | |
3929 | * other threads may have noticed network errors */ | |
3930 | dev_info(DEV, "unexpected cstate (%s) in receive_bitmap\n", | |
3931 | drbd_conn_str(mdev->state.conn)); | |
3932 | } | |
82bc0194 | 3933 | err = 0; |
b411b363 | 3934 | |
b411b363 | 3935 | out: |
20ceb2b2 | 3936 | drbd_bm_unlock(mdev); |
82bc0194 | 3937 | if (!err && mdev->state.conn == C_WF_BITMAP_S) |
b411b363 | 3938 | drbd_start_resync(mdev, C_SYNC_SOURCE); |
82bc0194 | 3939 | return err; |
b411b363 PR |
3940 | } |
3941 | ||
4a76b161 | 3942 | static int receive_skip(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 3943 | { |
4a76b161 | 3944 | conn_warn(tconn, "skipping unknown optional packet type %d, l: %d!\n", |
e2857216 | 3945 | pi->cmd, pi->size); |
2de876ef | 3946 | |
4a76b161 | 3947 | return ignore_remaining_packet(tconn, pi); |
2de876ef PR |
3948 | } |
3949 | ||
4a76b161 | 3950 | static int receive_UnplugRemote(struct drbd_tconn *tconn, struct packet_info *pi) |
0ced55a3 | 3951 | { |
e7f52dfb LE |
3952 | /* Make sure we've acked all the TCP data associated |
3953 | * with the data requests being unplugged */ | |
4a76b161 | 3954 | drbd_tcp_quickack(tconn->data.socket); |
0ced55a3 | 3955 | |
82bc0194 | 3956 | return 0; |
0ced55a3 PR |
3957 | } |
3958 | ||
4a76b161 | 3959 | static int receive_out_of_sync(struct drbd_tconn *tconn, struct packet_info *pi) |
73a01a18 | 3960 | { |
4a76b161 | 3961 | struct drbd_conf *mdev; |
e658983a | 3962 | struct p_block_desc *p = pi->data; |
4a76b161 AG |
3963 | |
3964 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
3965 | if (!mdev) | |
3966 | return -EIO; | |
73a01a18 | 3967 | |
f735e363 LE |
3968 | switch (mdev->state.conn) { |
3969 | case C_WF_SYNC_UUID: | |
3970 | case C_WF_BITMAP_T: | |
3971 | case C_BEHIND: | |
3972 | break; | |
3973 | default: | |
3974 | dev_err(DEV, "ASSERT FAILED cstate = %s, expected: WFSyncUUID|WFBitMapT|Behind\n", | |
3975 | drbd_conn_str(mdev->state.conn)); | |
3976 | } | |
3977 | ||
73a01a18 PR |
3978 | drbd_set_out_of_sync(mdev, be64_to_cpu(p->sector), be32_to_cpu(p->blksize)); |
3979 | ||
82bc0194 | 3980 | return 0; |
73a01a18 PR |
3981 | } |
3982 | ||
02918be2 PR |
3983 | struct data_cmd { |
3984 | int expect_payload; | |
3985 | size_t pkt_size; | |
4a76b161 | 3986 | int (*fn)(struct drbd_tconn *, struct packet_info *); |
02918be2 PR |
3987 | }; |
3988 | ||
3989 | static struct data_cmd drbd_cmd_handler[] = { | |
4a76b161 AG |
3990 | [P_DATA] = { 1, sizeof(struct p_data), receive_Data }, |
3991 | [P_DATA_REPLY] = { 1, sizeof(struct p_data), receive_DataReply }, | |
3992 | [P_RS_DATA_REPLY] = { 1, sizeof(struct p_data), receive_RSDataReply } , | |
3993 | [P_BARRIER] = { 0, sizeof(struct p_barrier), receive_Barrier } , | |
e658983a AG |
3994 | [P_BITMAP] = { 1, 0, receive_bitmap } , |
3995 | [P_COMPRESSED_BITMAP] = { 1, 0, receive_bitmap } , | |
3996 | [P_UNPLUG_REMOTE] = { 0, 0, receive_UnplugRemote }, | |
4a76b161 AG |
3997 | [P_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, |
3998 | [P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, | |
e658983a AG |
3999 | [P_SYNC_PARAM] = { 1, 0, receive_SyncParam }, |
4000 | [P_SYNC_PARAM89] = { 1, 0, receive_SyncParam }, | |
4a76b161 AG |
4001 | [P_PROTOCOL] = { 1, sizeof(struct p_protocol), receive_protocol }, |
4002 | [P_UUIDS] = { 0, sizeof(struct p_uuids), receive_uuids }, | |
4003 | [P_SIZES] = { 0, sizeof(struct p_sizes), receive_sizes }, | |
4004 | [P_STATE] = { 0, sizeof(struct p_state), receive_state }, | |
4005 | [P_STATE_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_state }, | |
4006 | [P_SYNC_UUID] = { 0, sizeof(struct p_rs_uuid), receive_sync_uuid }, | |
4007 | [P_OV_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, | |
4008 | [P_OV_REPLY] = { 1, sizeof(struct p_block_req), receive_DataRequest }, | |
4009 | [P_CSUM_RS_REQUEST] = { 1, sizeof(struct p_block_req), receive_DataRequest }, | |
4010 | [P_DELAY_PROBE] = { 0, sizeof(struct p_delay_probe93), receive_skip }, | |
4011 | [P_OUT_OF_SYNC] = { 0, sizeof(struct p_block_desc), receive_out_of_sync }, | |
4012 | [P_CONN_ST_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_conn_state }, | |
b411b363 PR |
4013 | }; |
4014 | ||
eefc2f7d | 4015 | static void drbdd(struct drbd_tconn *tconn) |
b411b363 | 4016 | { |
77351055 | 4017 | struct packet_info pi; |
02918be2 | 4018 | size_t shs; /* sub header size */ |
82bc0194 | 4019 | int err; |
b411b363 | 4020 | |
eefc2f7d | 4021 | while (get_t_state(&tconn->receiver) == RUNNING) { |
deebe195 AG |
4022 | struct data_cmd *cmd; |
4023 | ||
eefc2f7d | 4024 | drbd_thread_current_set_cpu(&tconn->receiver); |
69bc7bc3 | 4025 | if (drbd_recv_header(tconn, &pi)) |
02918be2 | 4026 | goto err_out; |
b411b363 | 4027 | |
deebe195 | 4028 | cmd = &drbd_cmd_handler[pi.cmd]; |
4a76b161 | 4029 | if (unlikely(pi.cmd >= ARRAY_SIZE(drbd_cmd_handler) || !cmd->fn)) { |
eefc2f7d | 4030 | conn_err(tconn, "unknown packet type %d, l: %d!\n", pi.cmd, pi.size); |
02918be2 | 4031 | goto err_out; |
0b33a916 | 4032 | } |
b411b363 | 4033 | |
e658983a AG |
4034 | shs = cmd->pkt_size; |
4035 | if (pi.size > shs && !cmd->expect_payload) { | |
eefc2f7d | 4036 | conn_err(tconn, "No payload expected %s l:%d\n", cmdname(pi.cmd), pi.size); |
02918be2 | 4037 | goto err_out; |
b411b363 | 4038 | } |
b411b363 | 4039 | |
c13f7e1a | 4040 | if (shs) { |
e658983a | 4041 | err = drbd_recv_all_warn(tconn, pi.data, shs); |
a5c31904 | 4042 | if (err) |
c13f7e1a | 4043 | goto err_out; |
e2857216 | 4044 | pi.size -= shs; |
c13f7e1a LE |
4045 | } |
4046 | ||
4a76b161 AG |
4047 | err = cmd->fn(tconn, &pi); |
4048 | if (err) { | |
9f5bdc33 AG |
4049 | conn_err(tconn, "error receiving %s, e: %d l: %d!\n", |
4050 | cmdname(pi.cmd), err, pi.size); | |
02918be2 | 4051 | goto err_out; |
b411b363 PR |
4052 | } |
4053 | } | |
82bc0194 | 4054 | return; |
b411b363 | 4055 | |
82bc0194 AG |
4056 | err_out: |
4057 | conn_request_state(tconn, NS(conn, C_PROTOCOL_ERROR), CS_HARD); | |
b411b363 PR |
4058 | } |
4059 | ||
0e29d163 | 4060 | void conn_flush_workqueue(struct drbd_tconn *tconn) |
b411b363 PR |
4061 | { |
4062 | struct drbd_wq_barrier barr; | |
4063 | ||
4064 | barr.w.cb = w_prev_work_done; | |
0e29d163 | 4065 | barr.w.tconn = tconn; |
b411b363 | 4066 | init_completion(&barr.done); |
0e29d163 | 4067 | drbd_queue_work(&tconn->data.work, &barr.w); |
b411b363 PR |
4068 | wait_for_completion(&barr.done); |
4069 | } | |
4070 | ||
360cc740 | 4071 | static void drbd_disconnect(struct drbd_tconn *tconn) |
b411b363 | 4072 | { |
bbeb641c | 4073 | enum drbd_conns oc; |
b411b363 | 4074 | int rv = SS_UNKNOWN_ERROR; |
b411b363 | 4075 | |
bbeb641c | 4076 | if (tconn->cstate == C_STANDALONE) |
b411b363 | 4077 | return; |
b411b363 PR |
4078 | |
4079 | /* asender does not clean up anything. it must not interfere, either */ | |
360cc740 PR |
4080 | drbd_thread_stop(&tconn->asender); |
4081 | drbd_free_sock(tconn); | |
4082 | ||
4083 | idr_for_each(&tconn->volumes, drbd_disconnected, tconn); | |
360cc740 PR |
4084 | conn_info(tconn, "Connection closed\n"); |
4085 | ||
cb703454 PR |
4086 | if (conn_highest_role(tconn) == R_PRIMARY && conn_highest_pdsk(tconn) >= D_UNKNOWN) |
4087 | conn_try_outdate_peer_async(tconn); | |
4088 | ||
360cc740 | 4089 | spin_lock_irq(&tconn->req_lock); |
bbeb641c PR |
4090 | oc = tconn->cstate; |
4091 | if (oc >= C_UNCONNECTED) | |
4092 | rv = _conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE); | |
4093 | ||
360cc740 PR |
4094 | spin_unlock_irq(&tconn->req_lock); |
4095 | ||
bbeb641c | 4096 | if (oc == C_DISCONNECTING) { |
360cc740 PR |
4097 | wait_event(tconn->net_cnt_wait, atomic_read(&tconn->net_cnt) == 0); |
4098 | ||
4099 | crypto_free_hash(tconn->cram_hmac_tfm); | |
4100 | tconn->cram_hmac_tfm = NULL; | |
4101 | ||
4102 | kfree(tconn->net_conf); | |
4103 | tconn->net_conf = NULL; | |
bbeb641c | 4104 | conn_request_state(tconn, NS(conn, C_STANDALONE), CS_VERBOSE); |
360cc740 PR |
4105 | } |
4106 | } | |
4107 | ||
4108 | static int drbd_disconnected(int vnr, void *p, void *data) | |
4109 | { | |
4110 | struct drbd_conf *mdev = (struct drbd_conf *)p; | |
4111 | enum drbd_fencing_p fp; | |
4112 | unsigned int i; | |
b411b363 | 4113 | |
85719573 | 4114 | /* wait for current activity to cease. */ |
87eeee41 | 4115 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
4116 | _drbd_wait_ee_list_empty(mdev, &mdev->active_ee); |
4117 | _drbd_wait_ee_list_empty(mdev, &mdev->sync_ee); | |
4118 | _drbd_wait_ee_list_empty(mdev, &mdev->read_ee); | |
87eeee41 | 4119 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
4120 | |
4121 | /* We do not have data structures that would allow us to | |
4122 | * get the rs_pending_cnt down to 0 again. | |
4123 | * * On C_SYNC_TARGET we do not have any data structures describing | |
4124 | * the pending RSDataRequest's we have sent. | |
4125 | * * On C_SYNC_SOURCE there is no data structure that tracks | |
4126 | * the P_RS_DATA_REPLY blocks that we sent to the SyncTarget. | |
4127 | * And no, it is not the sum of the reference counts in the | |
4128 | * resync_LRU. The resync_LRU tracks the whole operation including | |
4129 | * the disk-IO, while the rs_pending_cnt only tracks the blocks | |
4130 | * on the fly. */ | |
4131 | drbd_rs_cancel_all(mdev); | |
4132 | mdev->rs_total = 0; | |
4133 | mdev->rs_failed = 0; | |
4134 | atomic_set(&mdev->rs_pending_cnt, 0); | |
4135 | wake_up(&mdev->misc_wait); | |
4136 | ||
7fde2be9 PR |
4137 | del_timer(&mdev->request_timer); |
4138 | ||
b411b363 | 4139 | del_timer_sync(&mdev->resync_timer); |
b411b363 PR |
4140 | resync_timer_fn((unsigned long)mdev); |
4141 | ||
b411b363 PR |
4142 | /* wait for all w_e_end_data_req, w_e_end_rsdata_req, w_send_barrier, |
4143 | * w_make_resync_request etc. which may still be on the worker queue | |
4144 | * to be "canceled" */ | |
a21e9298 | 4145 | drbd_flush_workqueue(mdev); |
b411b363 | 4146 | |
a990be46 | 4147 | drbd_finish_peer_reqs(mdev); |
b411b363 PR |
4148 | |
4149 | kfree(mdev->p_uuid); | |
4150 | mdev->p_uuid = NULL; | |
4151 | ||
2aebfabb | 4152 | if (!drbd_suspended(mdev)) |
2f5cdd0b | 4153 | tl_clear(mdev->tconn); |
b411b363 | 4154 | |
b411b363 PR |
4155 | drbd_md_sync(mdev); |
4156 | ||
4157 | fp = FP_DONT_CARE; | |
4158 | if (get_ldev(mdev)) { | |
4159 | fp = mdev->ldev->dc.fencing; | |
4160 | put_ldev(mdev); | |
4161 | } | |
4162 | ||
20ceb2b2 LE |
4163 | /* serialize with bitmap writeout triggered by the state change, |
4164 | * if any. */ | |
4165 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
4166 | ||
b411b363 PR |
4167 | /* tcp_close and release of sendpage pages can be deferred. I don't |
4168 | * want to use SO_LINGER, because apparently it can be deferred for | |
4169 | * more than 20 seconds (longest time I checked). | |
4170 | * | |
4171 | * Actually we don't care for exactly when the network stack does its | |
4172 | * put_page(), but release our reference on these pages right here. | |
4173 | */ | |
7721f567 | 4174 | i = drbd_free_peer_reqs(mdev, &mdev->net_ee); |
b411b363 PR |
4175 | if (i) |
4176 | dev_info(DEV, "net_ee not empty, killed %u entries\n", i); | |
435f0740 LE |
4177 | i = atomic_read(&mdev->pp_in_use_by_net); |
4178 | if (i) | |
4179 | dev_info(DEV, "pp_in_use_by_net = %d, expected 0\n", i); | |
b411b363 PR |
4180 | i = atomic_read(&mdev->pp_in_use); |
4181 | if (i) | |
45bb912b | 4182 | dev_info(DEV, "pp_in_use = %d, expected 0\n", i); |
b411b363 PR |
4183 | |
4184 | D_ASSERT(list_empty(&mdev->read_ee)); | |
4185 | D_ASSERT(list_empty(&mdev->active_ee)); | |
4186 | D_ASSERT(list_empty(&mdev->sync_ee)); | |
4187 | D_ASSERT(list_empty(&mdev->done_ee)); | |
4188 | ||
4189 | /* ok, no more ee's on the fly, it is safe to reset the epoch_size */ | |
4190 | atomic_set(&mdev->current_epoch->epoch_size, 0); | |
4191 | D_ASSERT(list_empty(&mdev->current_epoch->list)); | |
360cc740 PR |
4192 | |
4193 | return 0; | |
b411b363 PR |
4194 | } |
4195 | ||
4196 | /* | |
4197 | * We support PRO_VERSION_MIN to PRO_VERSION_MAX. The protocol version | |
4198 | * we can agree on is stored in agreed_pro_version. | |
4199 | * | |
4200 | * feature flags and the reserved array should be enough room for future | |
4201 | * enhancements of the handshake protocol, and possible plugins... | |
4202 | * | |
4203 | * for now, they are expected to be zero, but ignored. | |
4204 | */ | |
6038178e | 4205 | static int drbd_send_features(struct drbd_tconn *tconn) |
b411b363 | 4206 | { |
9f5bdc33 AG |
4207 | struct drbd_socket *sock; |
4208 | struct p_connection_features *p; | |
b411b363 | 4209 | |
9f5bdc33 AG |
4210 | sock = &tconn->data; |
4211 | p = conn_prepare_command(tconn, sock); | |
4212 | if (!p) | |
e8d17b01 | 4213 | return -EIO; |
b411b363 PR |
4214 | memset(p, 0, sizeof(*p)); |
4215 | p->protocol_min = cpu_to_be32(PRO_VERSION_MIN); | |
4216 | p->protocol_max = cpu_to_be32(PRO_VERSION_MAX); | |
9f5bdc33 | 4217 | return conn_send_command(tconn, sock, P_CONNECTION_FEATURES, sizeof(*p), NULL, 0); |
b411b363 PR |
4218 | } |
4219 | ||
4220 | /* | |
4221 | * return values: | |
4222 | * 1 yes, we have a valid connection | |
4223 | * 0 oops, did not work out, please try again | |
4224 | * -1 peer talks different language, | |
4225 | * no point in trying again, please go standalone. | |
4226 | */ | |
6038178e | 4227 | static int drbd_do_features(struct drbd_tconn *tconn) |
b411b363 | 4228 | { |
65d11ed6 | 4229 | /* ASSERT current == tconn->receiver ... */ |
e658983a AG |
4230 | struct p_connection_features *p; |
4231 | const int expect = sizeof(struct p_connection_features); | |
77351055 | 4232 | struct packet_info pi; |
a5c31904 | 4233 | int err; |
b411b363 | 4234 | |
6038178e | 4235 | err = drbd_send_features(tconn); |
e8d17b01 | 4236 | if (err) |
b411b363 PR |
4237 | return 0; |
4238 | ||
69bc7bc3 AG |
4239 | err = drbd_recv_header(tconn, &pi); |
4240 | if (err) | |
b411b363 PR |
4241 | return 0; |
4242 | ||
6038178e AG |
4243 | if (pi.cmd != P_CONNECTION_FEATURES) { |
4244 | conn_err(tconn, "expected ConnectionFeatures packet, received: %s (0x%04x)\n", | |
77351055 | 4245 | cmdname(pi.cmd), pi.cmd); |
b411b363 PR |
4246 | return -1; |
4247 | } | |
4248 | ||
77351055 | 4249 | if (pi.size != expect) { |
6038178e | 4250 | conn_err(tconn, "expected ConnectionFeatures length: %u, received: %u\n", |
77351055 | 4251 | expect, pi.size); |
b411b363 PR |
4252 | return -1; |
4253 | } | |
4254 | ||
e658983a AG |
4255 | p = pi.data; |
4256 | err = drbd_recv_all_warn(tconn, p, expect); | |
a5c31904 | 4257 | if (err) |
b411b363 | 4258 | return 0; |
b411b363 | 4259 | |
b411b363 PR |
4260 | p->protocol_min = be32_to_cpu(p->protocol_min); |
4261 | p->protocol_max = be32_to_cpu(p->protocol_max); | |
4262 | if (p->protocol_max == 0) | |
4263 | p->protocol_max = p->protocol_min; | |
4264 | ||
4265 | if (PRO_VERSION_MAX < p->protocol_min || | |
4266 | PRO_VERSION_MIN > p->protocol_max) | |
4267 | goto incompat; | |
4268 | ||
65d11ed6 | 4269 | tconn->agreed_pro_version = min_t(int, PRO_VERSION_MAX, p->protocol_max); |
b411b363 | 4270 | |
65d11ed6 PR |
4271 | conn_info(tconn, "Handshake successful: " |
4272 | "Agreed network protocol version %d\n", tconn->agreed_pro_version); | |
b411b363 PR |
4273 | |
4274 | return 1; | |
4275 | ||
4276 | incompat: | |
65d11ed6 | 4277 | conn_err(tconn, "incompatible DRBD dialects: " |
b411b363 PR |
4278 | "I support %d-%d, peer supports %d-%d\n", |
4279 | PRO_VERSION_MIN, PRO_VERSION_MAX, | |
4280 | p->protocol_min, p->protocol_max); | |
4281 | return -1; | |
4282 | } | |
4283 | ||
4284 | #if !defined(CONFIG_CRYPTO_HMAC) && !defined(CONFIG_CRYPTO_HMAC_MODULE) | |
13e6037d | 4285 | static int drbd_do_auth(struct drbd_tconn *tconn) |
b411b363 PR |
4286 | { |
4287 | dev_err(DEV, "This kernel was build without CONFIG_CRYPTO_HMAC.\n"); | |
4288 | dev_err(DEV, "You need to disable 'cram-hmac-alg' in drbd.conf.\n"); | |
b10d96cb | 4289 | return -1; |
b411b363 PR |
4290 | } |
4291 | #else | |
4292 | #define CHALLENGE_LEN 64 | |
b10d96cb JT |
4293 | |
4294 | /* Return value: | |
4295 | 1 - auth succeeded, | |
4296 | 0 - failed, try again (network error), | |
4297 | -1 - auth failed, don't try again. | |
4298 | */ | |
4299 | ||
13e6037d | 4300 | static int drbd_do_auth(struct drbd_tconn *tconn) |
b411b363 | 4301 | { |
9f5bdc33 | 4302 | struct drbd_socket *sock; |
b411b363 PR |
4303 | char my_challenge[CHALLENGE_LEN]; /* 64 Bytes... */ |
4304 | struct scatterlist sg; | |
4305 | char *response = NULL; | |
4306 | char *right_response = NULL; | |
4307 | char *peers_ch = NULL; | |
13e6037d | 4308 | unsigned int key_len = strlen(tconn->net_conf->shared_secret); |
b411b363 PR |
4309 | unsigned int resp_size; |
4310 | struct hash_desc desc; | |
77351055 | 4311 | struct packet_info pi; |
69bc7bc3 | 4312 | int err, rv; |
b411b363 | 4313 | |
9f5bdc33 AG |
4314 | /* FIXME: Put the challenge/response into the preallocated socket buffer. */ |
4315 | ||
13e6037d | 4316 | desc.tfm = tconn->cram_hmac_tfm; |
b411b363 PR |
4317 | desc.flags = 0; |
4318 | ||
13e6037d PR |
4319 | rv = crypto_hash_setkey(tconn->cram_hmac_tfm, |
4320 | (u8 *)tconn->net_conf->shared_secret, key_len); | |
b411b363 | 4321 | if (rv) { |
13e6037d | 4322 | conn_err(tconn, "crypto_hash_setkey() failed with %d\n", rv); |
b10d96cb | 4323 | rv = -1; |
b411b363 PR |
4324 | goto fail; |
4325 | } | |
4326 | ||
4327 | get_random_bytes(my_challenge, CHALLENGE_LEN); | |
4328 | ||
9f5bdc33 AG |
4329 | sock = &tconn->data; |
4330 | if (!conn_prepare_command(tconn, sock)) { | |
4331 | rv = 0; | |
4332 | goto fail; | |
4333 | } | |
e658983a | 4334 | rv = !conn_send_command(tconn, sock, P_AUTH_CHALLENGE, 0, |
9f5bdc33 | 4335 | my_challenge, CHALLENGE_LEN); |
b411b363 PR |
4336 | if (!rv) |
4337 | goto fail; | |
4338 | ||
69bc7bc3 AG |
4339 | err = drbd_recv_header(tconn, &pi); |
4340 | if (err) { | |
4341 | rv = 0; | |
b411b363 | 4342 | goto fail; |
69bc7bc3 | 4343 | } |
b411b363 | 4344 | |
77351055 | 4345 | if (pi.cmd != P_AUTH_CHALLENGE) { |
13e6037d | 4346 | conn_err(tconn, "expected AuthChallenge packet, received: %s (0x%04x)\n", |
77351055 | 4347 | cmdname(pi.cmd), pi.cmd); |
b411b363 PR |
4348 | rv = 0; |
4349 | goto fail; | |
4350 | } | |
4351 | ||
77351055 | 4352 | if (pi.size > CHALLENGE_LEN * 2) { |
13e6037d | 4353 | conn_err(tconn, "expected AuthChallenge payload too big.\n"); |
b10d96cb | 4354 | rv = -1; |
b411b363 PR |
4355 | goto fail; |
4356 | } | |
4357 | ||
77351055 | 4358 | peers_ch = kmalloc(pi.size, GFP_NOIO); |
b411b363 | 4359 | if (peers_ch == NULL) { |
13e6037d | 4360 | conn_err(tconn, "kmalloc of peers_ch failed\n"); |
b10d96cb | 4361 | rv = -1; |
b411b363 PR |
4362 | goto fail; |
4363 | } | |
4364 | ||
a5c31904 AG |
4365 | err = drbd_recv_all_warn(tconn, peers_ch, pi.size); |
4366 | if (err) { | |
b411b363 PR |
4367 | rv = 0; |
4368 | goto fail; | |
4369 | } | |
4370 | ||
13e6037d | 4371 | resp_size = crypto_hash_digestsize(tconn->cram_hmac_tfm); |
b411b363 PR |
4372 | response = kmalloc(resp_size, GFP_NOIO); |
4373 | if (response == NULL) { | |
13e6037d | 4374 | conn_err(tconn, "kmalloc of response failed\n"); |
b10d96cb | 4375 | rv = -1; |
b411b363 PR |
4376 | goto fail; |
4377 | } | |
4378 | ||
4379 | sg_init_table(&sg, 1); | |
77351055 | 4380 | sg_set_buf(&sg, peers_ch, pi.size); |
b411b363 PR |
4381 | |
4382 | rv = crypto_hash_digest(&desc, &sg, sg.length, response); | |
4383 | if (rv) { | |
13e6037d | 4384 | conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv); |
b10d96cb | 4385 | rv = -1; |
b411b363 PR |
4386 | goto fail; |
4387 | } | |
4388 | ||
9f5bdc33 AG |
4389 | if (!conn_prepare_command(tconn, sock)) { |
4390 | rv = 0; | |
4391 | goto fail; | |
4392 | } | |
e658983a | 4393 | rv = !conn_send_command(tconn, sock, P_AUTH_RESPONSE, 0, |
9f5bdc33 | 4394 | response, resp_size); |
b411b363 PR |
4395 | if (!rv) |
4396 | goto fail; | |
4397 | ||
69bc7bc3 AG |
4398 | err = drbd_recv_header(tconn, &pi); |
4399 | if (err) { | |
4400 | rv = 0; | |
b411b363 | 4401 | goto fail; |
69bc7bc3 | 4402 | } |
b411b363 | 4403 | |
77351055 | 4404 | if (pi.cmd != P_AUTH_RESPONSE) { |
13e6037d | 4405 | conn_err(tconn, "expected AuthResponse packet, received: %s (0x%04x)\n", |
77351055 | 4406 | cmdname(pi.cmd), pi.cmd); |
b411b363 PR |
4407 | rv = 0; |
4408 | goto fail; | |
4409 | } | |
4410 | ||
77351055 | 4411 | if (pi.size != resp_size) { |
13e6037d | 4412 | conn_err(tconn, "expected AuthResponse payload of wrong size\n"); |
b411b363 PR |
4413 | rv = 0; |
4414 | goto fail; | |
4415 | } | |
4416 | ||
a5c31904 AG |
4417 | err = drbd_recv_all_warn(tconn, response , resp_size); |
4418 | if (err) { | |
b411b363 PR |
4419 | rv = 0; |
4420 | goto fail; | |
4421 | } | |
4422 | ||
4423 | right_response = kmalloc(resp_size, GFP_NOIO); | |
2d1ee87d | 4424 | if (right_response == NULL) { |
13e6037d | 4425 | conn_err(tconn, "kmalloc of right_response failed\n"); |
b10d96cb | 4426 | rv = -1; |
b411b363 PR |
4427 | goto fail; |
4428 | } | |
4429 | ||
4430 | sg_set_buf(&sg, my_challenge, CHALLENGE_LEN); | |
4431 | ||
4432 | rv = crypto_hash_digest(&desc, &sg, sg.length, right_response); | |
4433 | if (rv) { | |
13e6037d | 4434 | conn_err(tconn, "crypto_hash_digest() failed with %d\n", rv); |
b10d96cb | 4435 | rv = -1; |
b411b363 PR |
4436 | goto fail; |
4437 | } | |
4438 | ||
4439 | rv = !memcmp(response, right_response, resp_size); | |
4440 | ||
4441 | if (rv) | |
13e6037d PR |
4442 | conn_info(tconn, "Peer authenticated using %d bytes of '%s' HMAC\n", |
4443 | resp_size, tconn->net_conf->cram_hmac_alg); | |
b10d96cb JT |
4444 | else |
4445 | rv = -1; | |
b411b363 PR |
4446 | |
4447 | fail: | |
4448 | kfree(peers_ch); | |
4449 | kfree(response); | |
4450 | kfree(right_response); | |
4451 | ||
4452 | return rv; | |
4453 | } | |
4454 | #endif | |
4455 | ||
4456 | int drbdd_init(struct drbd_thread *thi) | |
4457 | { | |
392c8801 | 4458 | struct drbd_tconn *tconn = thi->tconn; |
b411b363 PR |
4459 | int h; |
4460 | ||
4d641dd7 | 4461 | conn_info(tconn, "receiver (re)started\n"); |
b411b363 PR |
4462 | |
4463 | do { | |
4d641dd7 | 4464 | h = drbd_connect(tconn); |
b411b363 | 4465 | if (h == 0) { |
4d641dd7 | 4466 | drbd_disconnect(tconn); |
20ee6390 | 4467 | schedule_timeout_interruptible(HZ); |
b411b363 PR |
4468 | } |
4469 | if (h == -1) { | |
4d641dd7 | 4470 | conn_warn(tconn, "Discarding network configuration.\n"); |
bbeb641c | 4471 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 PR |
4472 | } |
4473 | } while (h == 0); | |
4474 | ||
4475 | if (h > 0) { | |
4d641dd7 PR |
4476 | if (get_net_conf(tconn)) { |
4477 | drbdd(tconn); | |
4478 | put_net_conf(tconn); | |
b411b363 PR |
4479 | } |
4480 | } | |
4481 | ||
4d641dd7 | 4482 | drbd_disconnect(tconn); |
b411b363 | 4483 | |
4d641dd7 | 4484 | conn_info(tconn, "receiver terminated\n"); |
b411b363 PR |
4485 | return 0; |
4486 | } | |
4487 | ||
4488 | /* ********* acknowledge sender ******** */ | |
4489 | ||
e05e1e59 | 4490 | static int got_conn_RqSReply(struct drbd_tconn *tconn, struct packet_info *pi) |
e4f78ede | 4491 | { |
e658983a | 4492 | struct p_req_state_reply *p = pi->data; |
e4f78ede PR |
4493 | int retcode = be32_to_cpu(p->retcode); |
4494 | ||
4495 | if (retcode >= SS_SUCCESS) { | |
4496 | set_bit(CONN_WD_ST_CHG_OKAY, &tconn->flags); | |
4497 | } else { | |
4498 | set_bit(CONN_WD_ST_CHG_FAIL, &tconn->flags); | |
4499 | conn_err(tconn, "Requested state change failed by peer: %s (%d)\n", | |
4500 | drbd_set_st_err_str(retcode), retcode); | |
4501 | } | |
4502 | wake_up(&tconn->ping_wait); | |
4503 | ||
2735a594 | 4504 | return 0; |
e4f78ede PR |
4505 | } |
4506 | ||
1952e916 | 4507 | static int got_RqSReply(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4508 | { |
1952e916 | 4509 | struct drbd_conf *mdev; |
e658983a | 4510 | struct p_req_state_reply *p = pi->data; |
b411b363 PR |
4511 | int retcode = be32_to_cpu(p->retcode); |
4512 | ||
1952e916 AG |
4513 | mdev = vnr_to_mdev(tconn, pi->vnr); |
4514 | if (!mdev) | |
2735a594 | 4515 | return -EIO; |
1952e916 | 4516 | |
e4f78ede PR |
4517 | if (retcode >= SS_SUCCESS) { |
4518 | set_bit(CL_ST_CHG_SUCCESS, &mdev->flags); | |
4519 | } else { | |
4520 | set_bit(CL_ST_CHG_FAIL, &mdev->flags); | |
4521 | dev_err(DEV, "Requested state change failed by peer: %s (%d)\n", | |
4522 | drbd_set_st_err_str(retcode), retcode); | |
b411b363 | 4523 | } |
e4f78ede PR |
4524 | wake_up(&mdev->state_wait); |
4525 | ||
2735a594 | 4526 | return 0; |
b411b363 PR |
4527 | } |
4528 | ||
e05e1e59 | 4529 | static int got_Ping(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4530 | { |
2735a594 | 4531 | return drbd_send_ping_ack(tconn); |
b411b363 PR |
4532 | |
4533 | } | |
4534 | ||
e05e1e59 | 4535 | static int got_PingAck(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 PR |
4536 | { |
4537 | /* restore idle timeout */ | |
2a67d8b9 PR |
4538 | tconn->meta.socket->sk->sk_rcvtimeo = tconn->net_conf->ping_int*HZ; |
4539 | if (!test_and_set_bit(GOT_PING_ACK, &tconn->flags)) | |
4540 | wake_up(&tconn->ping_wait); | |
b411b363 | 4541 | |
2735a594 | 4542 | return 0; |
b411b363 PR |
4543 | } |
4544 | ||
1952e916 | 4545 | static int got_IsInSync(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4546 | { |
1952e916 | 4547 | struct drbd_conf *mdev; |
e658983a | 4548 | struct p_block_ack *p = pi->data; |
b411b363 PR |
4549 | sector_t sector = be64_to_cpu(p->sector); |
4550 | int blksize = be32_to_cpu(p->blksize); | |
4551 | ||
1952e916 AG |
4552 | mdev = vnr_to_mdev(tconn, pi->vnr); |
4553 | if (!mdev) | |
2735a594 | 4554 | return -EIO; |
1952e916 | 4555 | |
31890f4a | 4556 | D_ASSERT(mdev->tconn->agreed_pro_version >= 89); |
b411b363 PR |
4557 | |
4558 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4559 | ||
1d53f09e LE |
4560 | if (get_ldev(mdev)) { |
4561 | drbd_rs_complete_io(mdev, sector); | |
4562 | drbd_set_in_sync(mdev, sector, blksize); | |
4563 | /* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */ | |
4564 | mdev->rs_same_csum += (blksize >> BM_BLOCK_SHIFT); | |
4565 | put_ldev(mdev); | |
4566 | } | |
b411b363 | 4567 | dec_rs_pending(mdev); |
778f271d | 4568 | atomic_add(blksize >> 9, &mdev->rs_sect_in); |
b411b363 | 4569 | |
2735a594 | 4570 | return 0; |
b411b363 PR |
4571 | } |
4572 | ||
bc9c5c41 AG |
4573 | static int |
4574 | validate_req_change_req_state(struct drbd_conf *mdev, u64 id, sector_t sector, | |
4575 | struct rb_root *root, const char *func, | |
4576 | enum drbd_req_event what, bool missing_ok) | |
b411b363 PR |
4577 | { |
4578 | struct drbd_request *req; | |
4579 | struct bio_and_error m; | |
4580 | ||
87eeee41 | 4581 | spin_lock_irq(&mdev->tconn->req_lock); |
bc9c5c41 | 4582 | req = find_request(mdev, root, id, sector, missing_ok, func); |
b411b363 | 4583 | if (unlikely(!req)) { |
87eeee41 | 4584 | spin_unlock_irq(&mdev->tconn->req_lock); |
85997675 | 4585 | return -EIO; |
b411b363 PR |
4586 | } |
4587 | __req_mod(req, what, &m); | |
87eeee41 | 4588 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
4589 | |
4590 | if (m.bio) | |
4591 | complete_master_bio(mdev, &m); | |
85997675 | 4592 | return 0; |
b411b363 PR |
4593 | } |
4594 | ||
1952e916 | 4595 | static int got_BlockAck(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4596 | { |
1952e916 | 4597 | struct drbd_conf *mdev; |
e658983a | 4598 | struct p_block_ack *p = pi->data; |
b411b363 PR |
4599 | sector_t sector = be64_to_cpu(p->sector); |
4600 | int blksize = be32_to_cpu(p->blksize); | |
4601 | enum drbd_req_event what; | |
4602 | ||
1952e916 AG |
4603 | mdev = vnr_to_mdev(tconn, pi->vnr); |
4604 | if (!mdev) | |
2735a594 | 4605 | return -EIO; |
1952e916 | 4606 | |
b411b363 PR |
4607 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); |
4608 | ||
579b57ed | 4609 | if (p->block_id == ID_SYNCER) { |
b411b363 PR |
4610 | drbd_set_in_sync(mdev, sector, blksize); |
4611 | dec_rs_pending(mdev); | |
2735a594 | 4612 | return 0; |
b411b363 | 4613 | } |
e05e1e59 | 4614 | switch (pi->cmd) { |
b411b363 | 4615 | case P_RS_WRITE_ACK: |
89e58e75 | 4616 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); |
8554df1c | 4617 | what = WRITE_ACKED_BY_PEER_AND_SIS; |
b411b363 PR |
4618 | break; |
4619 | case P_WRITE_ACK: | |
89e58e75 | 4620 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); |
8554df1c | 4621 | what = WRITE_ACKED_BY_PEER; |
b411b363 PR |
4622 | break; |
4623 | case P_RECV_ACK: | |
89e58e75 | 4624 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_B); |
8554df1c | 4625 | what = RECV_ACKED_BY_PEER; |
b411b363 | 4626 | break; |
7be8da07 | 4627 | case P_DISCARD_WRITE: |
89e58e75 | 4628 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); |
7be8da07 AG |
4629 | what = DISCARD_WRITE; |
4630 | break; | |
4631 | case P_RETRY_WRITE: | |
4632 | D_ASSERT(mdev->tconn->net_conf->wire_protocol == DRBD_PROT_C); | |
4633 | what = POSTPONE_WRITE; | |
b411b363 PR |
4634 | break; |
4635 | default: | |
2735a594 | 4636 | BUG(); |
b411b363 PR |
4637 | } |
4638 | ||
2735a594 AG |
4639 | return validate_req_change_req_state(mdev, p->block_id, sector, |
4640 | &mdev->write_requests, __func__, | |
4641 | what, false); | |
b411b363 PR |
4642 | } |
4643 | ||
1952e916 | 4644 | static int got_NegAck(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4645 | { |
1952e916 | 4646 | struct drbd_conf *mdev; |
e658983a | 4647 | struct p_block_ack *p = pi->data; |
b411b363 | 4648 | sector_t sector = be64_to_cpu(p->sector); |
2deb8336 | 4649 | int size = be32_to_cpu(p->blksize); |
1952e916 AG |
4650 | bool missing_ok = tconn->net_conf->wire_protocol == DRBD_PROT_A || |
4651 | tconn->net_conf->wire_protocol == DRBD_PROT_B; | |
85997675 | 4652 | int err; |
b411b363 | 4653 | |
1952e916 AG |
4654 | mdev = vnr_to_mdev(tconn, pi->vnr); |
4655 | if (!mdev) | |
2735a594 | 4656 | return -EIO; |
1952e916 | 4657 | |
b411b363 PR |
4658 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); |
4659 | ||
579b57ed | 4660 | if (p->block_id == ID_SYNCER) { |
b411b363 PR |
4661 | dec_rs_pending(mdev); |
4662 | drbd_rs_failed_io(mdev, sector, size); | |
2735a594 | 4663 | return 0; |
b411b363 | 4664 | } |
2deb8336 | 4665 | |
85997675 AG |
4666 | err = validate_req_change_req_state(mdev, p->block_id, sector, |
4667 | &mdev->write_requests, __func__, | |
4668 | NEG_ACKED, missing_ok); | |
4669 | if (err) { | |
c3afd8f5 AG |
4670 | /* Protocol A has no P_WRITE_ACKs, but has P_NEG_ACKs. |
4671 | The master bio might already be completed, therefore the | |
4672 | request is no longer in the collision hash. */ | |
4673 | /* In Protocol B we might already have got a P_RECV_ACK | |
4674 | but then get a P_NEG_ACK afterwards. */ | |
4675 | if (!missing_ok) | |
2735a594 | 4676 | return err; |
c3afd8f5 | 4677 | drbd_set_out_of_sync(mdev, sector, size); |
2deb8336 | 4678 | } |
2735a594 | 4679 | return 0; |
b411b363 PR |
4680 | } |
4681 | ||
1952e916 | 4682 | static int got_NegDReply(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4683 | { |
1952e916 | 4684 | struct drbd_conf *mdev; |
e658983a | 4685 | struct p_block_ack *p = pi->data; |
b411b363 PR |
4686 | sector_t sector = be64_to_cpu(p->sector); |
4687 | ||
1952e916 AG |
4688 | mdev = vnr_to_mdev(tconn, pi->vnr); |
4689 | if (!mdev) | |
2735a594 | 4690 | return -EIO; |
1952e916 | 4691 | |
b411b363 | 4692 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); |
7be8da07 | 4693 | |
b411b363 PR |
4694 | dev_err(DEV, "Got NegDReply; Sector %llus, len %u; Fail original request.\n", |
4695 | (unsigned long long)sector, be32_to_cpu(p->blksize)); | |
4696 | ||
2735a594 AG |
4697 | return validate_req_change_req_state(mdev, p->block_id, sector, |
4698 | &mdev->read_requests, __func__, | |
4699 | NEG_ACKED, false); | |
b411b363 PR |
4700 | } |
4701 | ||
1952e916 | 4702 | static int got_NegRSDReply(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4703 | { |
1952e916 | 4704 | struct drbd_conf *mdev; |
b411b363 PR |
4705 | sector_t sector; |
4706 | int size; | |
e658983a | 4707 | struct p_block_ack *p = pi->data; |
1952e916 AG |
4708 | |
4709 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
4710 | if (!mdev) | |
2735a594 | 4711 | return -EIO; |
b411b363 PR |
4712 | |
4713 | sector = be64_to_cpu(p->sector); | |
4714 | size = be32_to_cpu(p->blksize); | |
b411b363 PR |
4715 | |
4716 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4717 | ||
4718 | dec_rs_pending(mdev); | |
4719 | ||
4720 | if (get_ldev_if_state(mdev, D_FAILED)) { | |
4721 | drbd_rs_complete_io(mdev, sector); | |
e05e1e59 | 4722 | switch (pi->cmd) { |
d612d309 PR |
4723 | case P_NEG_RS_DREPLY: |
4724 | drbd_rs_failed_io(mdev, sector, size); | |
4725 | case P_RS_CANCEL: | |
4726 | break; | |
4727 | default: | |
2735a594 | 4728 | BUG(); |
d612d309 | 4729 | } |
b411b363 PR |
4730 | put_ldev(mdev); |
4731 | } | |
4732 | ||
2735a594 | 4733 | return 0; |
b411b363 PR |
4734 | } |
4735 | ||
1952e916 | 4736 | static int got_BarrierAck(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4737 | { |
1952e916 | 4738 | struct drbd_conf *mdev; |
e658983a | 4739 | struct p_barrier_ack *p = pi->data; |
1952e916 AG |
4740 | |
4741 | mdev = vnr_to_mdev(tconn, pi->vnr); | |
4742 | if (!mdev) | |
2735a594 | 4743 | return -EIO; |
b411b363 | 4744 | |
2f5cdd0b | 4745 | tl_release(mdev->tconn, p->barrier, be32_to_cpu(p->set_size)); |
b411b363 | 4746 | |
c4752ef1 PR |
4747 | if (mdev->state.conn == C_AHEAD && |
4748 | atomic_read(&mdev->ap_in_flight) == 0 && | |
370a43e7 PR |
4749 | !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &mdev->current_epoch->flags)) { |
4750 | mdev->start_resync_timer.expires = jiffies + HZ; | |
4751 | add_timer(&mdev->start_resync_timer); | |
c4752ef1 PR |
4752 | } |
4753 | ||
2735a594 | 4754 | return 0; |
b411b363 PR |
4755 | } |
4756 | ||
1952e916 | 4757 | static int got_OVResult(struct drbd_tconn *tconn, struct packet_info *pi) |
b411b363 | 4758 | { |
1952e916 | 4759 | struct drbd_conf *mdev; |
e658983a | 4760 | struct p_block_ack *p = pi->data; |
b411b363 PR |
4761 | struct drbd_work *w; |
4762 | sector_t sector; | |
4763 | int size; | |
4764 | ||
1952e916 AG |
4765 | mdev = vnr_to_mdev(tconn, pi->vnr); |
4766 | if (!mdev) | |
2735a594 | 4767 | return -EIO; |
1952e916 | 4768 | |
b411b363 PR |
4769 | sector = be64_to_cpu(p->sector); |
4770 | size = be32_to_cpu(p->blksize); | |
4771 | ||
4772 | update_peer_seq(mdev, be32_to_cpu(p->seq_num)); | |
4773 | ||
4774 | if (be64_to_cpu(p->block_id) == ID_OUT_OF_SYNC) | |
8f7bed77 | 4775 | drbd_ov_out_of_sync_found(mdev, sector, size); |
b411b363 | 4776 | else |
8f7bed77 | 4777 | ov_out_of_sync_print(mdev); |
b411b363 | 4778 | |
1d53f09e | 4779 | if (!get_ldev(mdev)) |
2735a594 | 4780 | return 0; |
1d53f09e | 4781 | |
b411b363 PR |
4782 | drbd_rs_complete_io(mdev, sector); |
4783 | dec_rs_pending(mdev); | |
4784 | ||
ea5442af LE |
4785 | --mdev->ov_left; |
4786 | ||
4787 | /* let's advance progress step marks only for every other megabyte */ | |
4788 | if ((mdev->ov_left & 0x200) == 0x200) | |
4789 | drbd_advance_rs_marks(mdev, mdev->ov_left); | |
4790 | ||
4791 | if (mdev->ov_left == 0) { | |
b411b363 PR |
4792 | w = kmalloc(sizeof(*w), GFP_NOIO); |
4793 | if (w) { | |
4794 | w->cb = w_ov_finished; | |
a21e9298 | 4795 | w->mdev = mdev; |
e42325a5 | 4796 | drbd_queue_work_front(&mdev->tconn->data.work, w); |
b411b363 PR |
4797 | } else { |
4798 | dev_err(DEV, "kmalloc(w) failed."); | |
8f7bed77 | 4799 | ov_out_of_sync_print(mdev); |
b411b363 PR |
4800 | drbd_resync_finished(mdev); |
4801 | } | |
4802 | } | |
1d53f09e | 4803 | put_ldev(mdev); |
2735a594 | 4804 | return 0; |
b411b363 PR |
4805 | } |
4806 | ||
1952e916 | 4807 | static int got_skip(struct drbd_tconn *tconn, struct packet_info *pi) |
0ced55a3 | 4808 | { |
2735a594 | 4809 | return 0; |
0ced55a3 PR |
4810 | } |
4811 | ||
a990be46 | 4812 | static int tconn_finish_peer_reqs(struct drbd_tconn *tconn) |
32862ec7 | 4813 | { |
082a3439 PR |
4814 | struct drbd_conf *mdev; |
4815 | int i, not_empty = 0; | |
32862ec7 PR |
4816 | |
4817 | do { | |
4818 | clear_bit(SIGNAL_ASENDER, &tconn->flags); | |
4819 | flush_signals(current); | |
082a3439 | 4820 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
a990be46 | 4821 | if (drbd_finish_peer_reqs(mdev)) |
082a3439 PR |
4822 | return 1; /* error */ |
4823 | } | |
32862ec7 | 4824 | set_bit(SIGNAL_ASENDER, &tconn->flags); |
082a3439 PR |
4825 | |
4826 | spin_lock_irq(&tconn->req_lock); | |
4827 | idr_for_each_entry(&tconn->volumes, mdev, i) { | |
4828 | not_empty = !list_empty(&mdev->done_ee); | |
4829 | if (not_empty) | |
4830 | break; | |
4831 | } | |
4832 | spin_unlock_irq(&tconn->req_lock); | |
32862ec7 PR |
4833 | } while (not_empty); |
4834 | ||
4835 | return 0; | |
4836 | } | |
4837 | ||
7201b972 AG |
4838 | struct asender_cmd { |
4839 | size_t pkt_size; | |
1952e916 | 4840 | int (*fn)(struct drbd_tconn *tconn, struct packet_info *); |
7201b972 AG |
4841 | }; |
4842 | ||
4843 | static struct asender_cmd asender_tbl[] = { | |
e658983a AG |
4844 | [P_PING] = { 0, got_Ping }, |
4845 | [P_PING_ACK] = { 0, got_PingAck }, | |
1952e916 AG |
4846 | [P_RECV_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, |
4847 | [P_WRITE_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, | |
4848 | [P_RS_WRITE_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, | |
4849 | [P_DISCARD_WRITE] = { sizeof(struct p_block_ack), got_BlockAck }, | |
4850 | [P_NEG_ACK] = { sizeof(struct p_block_ack), got_NegAck }, | |
4851 | [P_NEG_DREPLY] = { sizeof(struct p_block_ack), got_NegDReply }, | |
4852 | [P_NEG_RS_DREPLY] = { sizeof(struct p_block_ack), got_NegRSDReply }, | |
4853 | [P_OV_RESULT] = { sizeof(struct p_block_ack), got_OVResult }, | |
4854 | [P_BARRIER_ACK] = { sizeof(struct p_barrier_ack), got_BarrierAck }, | |
4855 | [P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply }, | |
4856 | [P_RS_IS_IN_SYNC] = { sizeof(struct p_block_ack), got_IsInSync }, | |
4857 | [P_DELAY_PROBE] = { sizeof(struct p_delay_probe93), got_skip }, | |
4858 | [P_RS_CANCEL] = { sizeof(struct p_block_ack), got_NegRSDReply }, | |
4859 | [P_CONN_ST_CHG_REPLY]={ sizeof(struct p_req_state_reply), got_conn_RqSReply }, | |
4860 | [P_RETRY_WRITE] = { sizeof(struct p_block_ack), got_BlockAck }, | |
7201b972 AG |
4861 | }; |
4862 | ||
b411b363 PR |
4863 | int drbd_asender(struct drbd_thread *thi) |
4864 | { | |
392c8801 | 4865 | struct drbd_tconn *tconn = thi->tconn; |
b411b363 | 4866 | struct asender_cmd *cmd = NULL; |
77351055 | 4867 | struct packet_info pi; |
257d0af6 | 4868 | int rv; |
e658983a | 4869 | void *buf = tconn->meta.rbuf; |
b411b363 | 4870 | int received = 0; |
52b061a4 AG |
4871 | unsigned int header_size = drbd_header_size(tconn); |
4872 | int expect = header_size; | |
f36af18c | 4873 | int ping_timeout_active = 0; |
b411b363 | 4874 | |
b411b363 PR |
4875 | current->policy = SCHED_RR; /* Make this a realtime task! */ |
4876 | current->rt_priority = 2; /* more important than all other tasks */ | |
4877 | ||
e77a0a5c | 4878 | while (get_t_state(thi) == RUNNING) { |
80822284 | 4879 | drbd_thread_current_set_cpu(thi); |
32862ec7 | 4880 | if (test_and_clear_bit(SEND_PING, &tconn->flags)) { |
a17647aa | 4881 | if (drbd_send_ping(tconn)) { |
32862ec7 | 4882 | conn_err(tconn, "drbd_send_ping has failed\n"); |
841ce241 AG |
4883 | goto reconnect; |
4884 | } | |
32862ec7 PR |
4885 | tconn->meta.socket->sk->sk_rcvtimeo = |
4886 | tconn->net_conf->ping_timeo*HZ/10; | |
f36af18c | 4887 | ping_timeout_active = 1; |
b411b363 PR |
4888 | } |
4889 | ||
32862ec7 PR |
4890 | /* TODO: conditionally cork; it may hurt latency if we cork without |
4891 | much to send */ | |
4892 | if (!tconn->net_conf->no_cork) | |
4893 | drbd_tcp_cork(tconn->meta.socket); | |
a990be46 AG |
4894 | if (tconn_finish_peer_reqs(tconn)) { |
4895 | conn_err(tconn, "tconn_finish_peer_reqs() failed\n"); | |
32862ec7 | 4896 | goto reconnect; |
082a3439 | 4897 | } |
b411b363 | 4898 | /* but unconditionally uncork unless disabled */ |
32862ec7 PR |
4899 | if (!tconn->net_conf->no_cork) |
4900 | drbd_tcp_uncork(tconn->meta.socket); | |
b411b363 PR |
4901 | |
4902 | /* short circuit, recv_msg would return EINTR anyways. */ | |
4903 | if (signal_pending(current)) | |
4904 | continue; | |
4905 | ||
32862ec7 PR |
4906 | rv = drbd_recv_short(tconn->meta.socket, buf, expect-received, 0); |
4907 | clear_bit(SIGNAL_ASENDER, &tconn->flags); | |
b411b363 PR |
4908 | |
4909 | flush_signals(current); | |
4910 | ||
4911 | /* Note: | |
4912 | * -EINTR (on meta) we got a signal | |
4913 | * -EAGAIN (on meta) rcvtimeo expired | |
4914 | * -ECONNRESET other side closed the connection | |
4915 | * -ERESTARTSYS (on data) we got a signal | |
4916 | * rv < 0 other than above: unexpected error! | |
4917 | * rv == expected: full header or command | |
4918 | * rv < expected: "woken" by signal during receive | |
4919 | * rv == 0 : "connection shut down by peer" | |
4920 | */ | |
4921 | if (likely(rv > 0)) { | |
4922 | received += rv; | |
4923 | buf += rv; | |
4924 | } else if (rv == 0) { | |
32862ec7 | 4925 | conn_err(tconn, "meta connection shut down by peer.\n"); |
b411b363 PR |
4926 | goto reconnect; |
4927 | } else if (rv == -EAGAIN) { | |
cb6518cb LE |
4928 | /* If the data socket received something meanwhile, |
4929 | * that is good enough: peer is still alive. */ | |
32862ec7 PR |
4930 | if (time_after(tconn->last_received, |
4931 | jiffies - tconn->meta.socket->sk->sk_rcvtimeo)) | |
cb6518cb | 4932 | continue; |
f36af18c | 4933 | if (ping_timeout_active) { |
32862ec7 | 4934 | conn_err(tconn, "PingAck did not arrive in time.\n"); |
b411b363 PR |
4935 | goto reconnect; |
4936 | } | |
32862ec7 | 4937 | set_bit(SEND_PING, &tconn->flags); |
b411b363 PR |
4938 | continue; |
4939 | } else if (rv == -EINTR) { | |
4940 | continue; | |
4941 | } else { | |
32862ec7 | 4942 | conn_err(tconn, "sock_recvmsg returned %d\n", rv); |
b411b363 PR |
4943 | goto reconnect; |
4944 | } | |
4945 | ||
4946 | if (received == expect && cmd == NULL) { | |
e658983a | 4947 | if (decode_header(tconn, tconn->meta.rbuf, &pi)) |
b411b363 | 4948 | goto reconnect; |
7201b972 | 4949 | cmd = &asender_tbl[pi.cmd]; |
1952e916 | 4950 | if (pi.cmd >= ARRAY_SIZE(asender_tbl) || !cmd->fn) { |
32862ec7 | 4951 | conn_err(tconn, "unknown command %d on meta (l: %d)\n", |
77351055 | 4952 | pi.cmd, pi.size); |
b411b363 PR |
4953 | goto disconnect; |
4954 | } | |
e658983a | 4955 | expect = header_size + cmd->pkt_size; |
52b061a4 | 4956 | if (pi.size != expect - header_size) { |
32862ec7 | 4957 | conn_err(tconn, "Wrong packet size on meta (c: %d, l: %d)\n", |
77351055 | 4958 | pi.cmd, pi.size); |
b411b363 | 4959 | goto reconnect; |
257d0af6 | 4960 | } |
b411b363 PR |
4961 | } |
4962 | if (received == expect) { | |
2735a594 | 4963 | bool err; |
a4fbda8e | 4964 | |
2735a594 AG |
4965 | err = cmd->fn(tconn, &pi); |
4966 | if (err) { | |
1952e916 | 4967 | conn_err(tconn, "%pf failed\n", cmd->fn); |
b411b363 | 4968 | goto reconnect; |
1952e916 | 4969 | } |
b411b363 | 4970 | |
a4fbda8e PR |
4971 | tconn->last_received = jiffies; |
4972 | ||
f36af18c LE |
4973 | /* the idle_timeout (ping-int) |
4974 | * has been restored in got_PingAck() */ | |
7201b972 | 4975 | if (cmd == &asender_tbl[P_PING_ACK]) |
f36af18c LE |
4976 | ping_timeout_active = 0; |
4977 | ||
e658983a | 4978 | buf = tconn->meta.rbuf; |
b411b363 | 4979 | received = 0; |
52b061a4 | 4980 | expect = header_size; |
b411b363 PR |
4981 | cmd = NULL; |
4982 | } | |
4983 | } | |
4984 | ||
4985 | if (0) { | |
4986 | reconnect: | |
bbeb641c | 4987 | conn_request_state(tconn, NS(conn, C_NETWORK_FAILURE), CS_HARD); |
b411b363 PR |
4988 | } |
4989 | if (0) { | |
4990 | disconnect: | |
bbeb641c | 4991 | conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD); |
b411b363 | 4992 | } |
32862ec7 | 4993 | clear_bit(SIGNAL_ASENDER, &tconn->flags); |
b411b363 | 4994 | |
32862ec7 | 4995 | conn_info(tconn, "asender terminated\n"); |
b411b363 PR |
4996 | |
4997 | return 0; | |
4998 | } |